31
Origins of Xenolithic Eclogites and Pyroxenites from the Central Slave Craton, Canada SONJA AULBACH 1 *, NORMAN J. PEARSON 1 , SUZANNE Y. O’REILLY 1 AND BUDDY J. DOYLE 2 1 GEMOC ARC NATIONAL KEY CENTRE, DEPARTMENT OF EARTH AND PLANETARY SCIENCES, MACQUARIE UNIVERSITY, NSW 2109, AUSTRALIA 2 LITHOSPHERE SERVICES, 4009 EDINBURGH STREET, BURNABY, BC, CANADA V5C 1R4 RECEIVED SEPTEMBER 27, 2006; ACCEPTEDJULY 4, 2007 ADVANCE ACCESS PUBLICATION SEPTEMBER 3, 2007 Major- and trace-element and Sr^Nd^Hf isotopic compositions of garnet and clinopyroxene in kimberlite-borne eclogite and pyroxe- nite xenoliths were used to establish their origins and evolution in the subcontinental lithospheric mantle beneath the central Slave Craton, Canada. The majority of eclogites can be assigned to three groups (high-Mg, high-Ca or low-Mg eclogites) that have distinct trace-element patterns. Although post-formation metasomatism involving high field strength element (HFSE) and light rare earth element (LREE) addition has partially obscured the primary compositional features of the high-Mg and high-Ca eclo- gites, trace-element features, such as unfractionated middle REE (MREE) to heavy REE (HREE) patterns suggestive of garnet-free residues and low Zr/Sm consistent with plagioclase accu- mulation, could indicate a subduction origin from a broadly gabbroic protolith. In this scenario, the low P REE and small positive Eu anomalies of the high-Mg eclogites suggest more primitive, plagi- oclase-rich protoliths, whereas the high-Ca eclogites are proposed to have more evolved protoliths with higher (normative) clinopyrox- ene/plagioclase ratios plus trapped melt, consistent with their lower Mg-numbers, higher P REE and absence of Eu anomalies. In con- trast, the subchondritic Zr/Hf and positive slope in the HREE of the low-Mg eclogites are similar to Archaean second-stage melts and point to a previously depleted source for their precursors. Low ratios of fluid-mobile to less fluid-mobile elements and of LREE to HREE are consistent with dehydration and partial melt loss for some eclogites.The trace-element characteristics of the different eclogite types translate into lower e Nd for high-Mg eclogites than for low-Mg eclogites.Within the low-Mg group, samples that show evidence for metasomatic enrichment in LREE and HFSE have lower e Nd and e Hf than a sample that was apparently not enriched, pointing to long-term evolution at their respective parent^daughter ratios. Garnet and clinopyroxene in pyroxenites show different major-element relationships from those in eclogites, such as an opposite CaO^Na 2 O trend and the presence of a CaO^Cr 2 O 3 trend, independent of whether or not opx is part of the assemblage. Therefore, these two rock types are probably not related by fractiona- tion processes.The presence of opx in about half of the samples precludes direct crystallization from eclogite-derived melts. They probably formed from hybridized melts that reacted with the peridotitic mantle. KEY WORDS: eclogites; pyroxenite xenoliths; mantle xenoliths; eclogite trace elements; eclogite Sr isotopes; eclogite Hf isotopes; eclogite Nd isotopes INTRODUCTION Eclogites are a volumetrically small but compositionally significant component in the mantle because of their fertile composition, rapidly evolving radiogenic isotope compositions (as a result of parent^daughter ratios distinct from depleted or primitive upper mantle reservoirs), and lower melting point relative to peridotite, making them a source of isotopically distinct components in the refractory continental lithosphere. In addition, the presence of eclo- gites in the mantle column indicates the operation of processes that are likely to have significantly affected the lithosphere, regardless of their origin. There are two main hypotheses for the origin of kimberlite-borne eclogite xenoliths [for a recent review, see Jacob (2004)]: they might have an intrusive high- pressure mantle melt origin (e.g. Hills & Haggerty, 1989; Smyth et al ., 1989; Caporuscio & Smyth, 1990; *Corresponding author. Present address: Earth and Atmospheric Sciences, University of Alberta, Edmonton, AB, T6G 2E3, Canada. Telephone: þ61 2 9850 8953. Fax: þ61 2 9850 8943. E-mail: saulbach@ els.mq.edu.au ß The Author 2007. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@ oxfordjournals.org JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 PAGES 1843^1873 2007 doi:10.1093/petrology/egm041 Downloaded from https://academic.oup.com/petrology/article/48/10/1843/1512585 by guest on 01 January 2022

Origins of Xenolithic Eclogites and Pyroxenites from the Central

  • Upload
    others

  • View
    5

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Origins of Xenolithic Eclogites and Pyroxenites from the Central

Origins of Xenolithic Eclogites and Pyroxenitesfrom the Central Slave Craton Canada

SONJA AULBACH1 NORMAN J PEARSON1SUZANNE Y OrsquoREILLY1 AND BUDDY J DOYLE2

1GEMOC ARC NATIONAL KEY CENTRE DEPARTMENT OF EARTH AND PLANETARY SCIENCES MACQUARIE

UNIVERSITY NSW 2109 AUSTRALIA2LITHOSPHERE SERVICES 4009 EDINBURGH STREET BURNABY BC CANADA V5C 1R4

RECEIVED SEPTEMBER 27 2006 ACCEPTEDJULY 4 2007ADVANCE ACCESS PUBLICATION SEPTEMBER 3 2007

Major- and trace-element and Sr^Nd^Hf isotopic compositions

of garnet and clinopyroxene in kimberlite-borne eclogite and pyroxe-

nite xenoliths were used to establish their origins and evolution in

the subcontinental lithospheric mantle beneath the central Slave

Craton Canada The majority of eclogites can be assigned to three

groups (high-Mg high-Ca or low-Mg eclogites) that have distinct

trace-element patterns Although post-formation metasomatism

involving high field strength element (HFSE) and light rare

earth element (LREE) addition has partially obscured the

primary compositional features of the high-Mg and high-Ca eclo-

gites trace-element features such as unfractionated middle

REE (MREE) to heavy REE (HREE) patterns suggestive of

garnet-free residues and low ZrSm consistent with plagioclase accu-

mulation could indicate a subduction origin from a broadly gabbroic

protolith In this scenario the lowP

REE and small positive

Eu anomalies of the high-Mg eclogites suggest more primitive plagi-

oclase-rich protoliths whereas the high-Ca eclogites are proposed to

have more evolved protoliths with higher (normative) clinopyrox-

eneplagioclase ratios plus trapped melt consistent with their lower

Mg-numbers higherP

REE and absence of Eu anomalies In con-

trast the subchondritic ZrHf and positive slope in the HREE of the

low-Mg eclogites are similar to Archaean second-stage melts and

point to a previously depleted source for their precursors Low ratios

of fluid-mobile to less fluid-mobile elements and of LREE to

HREE are consistent with dehydration and partial melt loss

for some eclogites The trace-element characteristics of the different

eclogite types translate into lower eNd for high-Mg eclogites than

for low-Mg eclogitesWithin the low-Mg group samples that show

evidence for metasomatic enrichment in LREE and HFSE have

lower eNd and eHf than a sample that was apparently not enriched

pointing to long-term evolution at their respective parent^daughter

ratios Garnet and clinopyroxene in pyroxenites show different

major-element relationships from those in eclogites such as an

opposite CaO^Na2O trend and the presence of a CaO^Cr2O3

trend independent of whether or not opx is part of the assemblage

Therefore these two rock types are probably not related by fractiona-

tion processes The presence of opx in about half of the samples

precludes direct crystallization from eclogite-derived melts

They probably formed from hybridized melts that reacted with the

peridotitic mantle

KEY WORDS eclogites pyroxenite xenoliths mantle xenoliths

eclogite trace elements eclogite Sr isotopes eclogite Hf isotopes

eclogite Nd isotopes

I NTRODUCTIONEclogites are a volumetrically small but compositionallysignificant component in the mantle because of theirfertile composition rapidly evolving radiogenic isotopecompositions (as a result of parent^daughter ratios distinctfrom depleted or primitive upper mantle reservoirs) andlower melting point relative to peridotite making them asource of isotopically distinct components in the refractorycontinental lithosphere In addition the presence of eclo-gites in the mantle column indicates the operation ofprocesses that are likely to have significantly affected thelithosphere regardless of their originThere are two main hypotheses for the origin of

kimberlite-borne eclogite xenoliths [for a recent reviewsee Jacob (2004)] they might have an intrusive high-pressure mantle melt origin (eg Hills amp Haggerty1989 Smyth et al 1989 Caporuscio amp Smyth 1990

Corresponding author Present address Earth and AtmosphericSciences University of Alberta Edmonton AB T6G 2E3 CanadaTelephone thorn612 9850 8953 Fax thorn612 9850 8943 E-mail saulbachelsmqeduau

The Author 2007 Published by Oxford University Press Allrights reserved For Permissions please e-mail journalspermissionsoxfordjournalsorg

JOURNALOFPETROLOGY VOLUME 48 NUMBER10 PAGES1843^1873 2007 doi101093petrologyegm041D

ownloaded from

httpsacademicoupcom

petrologyarticle481018431512585 by guest on 01 January 2022

Griffin amp OrsquoReilly 2007) or represent subducted oceaniccrust (eg MacGregor amp Manton 1986 Schulze ampHelmstaedt 1988 Jacob et al 1994 Beard et al 1996)Eclogite suites from many localities in the Kaapvaal andSiberian cratons fall into groups with one group havingcharacteristics consistent with a mantle origin whereasanother group (or groups) has been interpreted as crus-tally derived (Taylor amp Neal 1989 Viljoen et al 1996Kopylova et al1999 Barth et al 2002b) In addition to pri-mary differences eclogite compositions can be modifiedduring melting and metasomatism following their forma-tion (Ireland et al 1994 Barth et al 2001)We have carried out a detailed petrographic major- and

trace-element and multi-isotope study to constrain the gen-esis of different eclogite xenolith types found in kimberlitesintruded in the central Slave Craton Canada bearing inmind that they may have been subjected to a range ofpost-formation processes during their residence time inthe Archaean lithosphere including melting and metaso-matic overprinting

GEOLOGICAL SETT INGThe samples are derived from the Diavik A154 kimberlitepipe in the Lac de Gras area of the Slave Craton in theCanadian Northwest Territories (Fig 1) The Slave Cratonis a small (400 km 500 km) Archean block that is com-posed of two major basement domains (1) a Hadean to

Mesoarcheaean domain in the west (Central SlaveBasement Complex CSBC Bleeker et al 1999) with agesof 40^28Ga onto which 273^270Ga tholeiites wereextruded (Padgham amp Fyson 1992 van Breemen et al1992 Isachsen amp Bowring 1994) (2) a Neoarchaean isoto-pically juvenile domain in the east where the tholeiitesequence is notably absent The Mesoarchaean basementthe eastern extent of which is not accurately known dipsunder the eastern domain (Bleeker et al 1999) The originof these domains has been ascribed to arc^continent colli-sion (Kusky 1989 Davis amp Hegner 1992) or to the easterndomain representing attenuated modified Mesoarchaeanlithosphere (Bleeker 2003) A major north^south-trendingprovinciality in the Slave Craton is evident in basementNd and Pb isotope data (Davis amp Hegner 1992) and acoupling of the two domains by 27Ga may be indicatedby pan-Slave calc-alkaline volcanism (van Breemen et al1992) Younger events include 22^18Ga collisions withneighbouring terranes (Hoffman1989) numerous episodesof Proterozoic dike emplacement such as the Malley^McKay dike swarm at 221^223Ga (LeCheminant et al1996) and the 127Ga Mackenzie swarm (LeCheminantamp Heaman 1989) followed by kimberlite volcanism inCretaceous to Eocene time (Creaser et al 2004 Heamanet al 2004)Distinct mantle xenolith suites and spatial distributions

of mantle rock types in the lithosphere beneath different

66deg

64deg

108deg

16deg

Pb

Great Slave Lake Shear Z

one

Bathurst Fault

112deg

Central Slave(40ndash28 Gabasement)

Protero-

zoic

Nd

Eastern Slave(lt28 Ga)

Lac de Graskimberlites

NunavutNWT

Jericho

Ekati

Diavik

Contwoyto Lake

Lac De Gras

Fig 1 Map of the Slave Craton (after Davis et al 2003) showing Pb and Nd isotope lines that separate ancient basement in the west fromjuvenile rocks in the east Localities of the Diavik Ekati and Jericho kimberlites are shown in more detail on the left

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1844

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

parts of the craton have been recognized (Griffin et al1999 2004 Grulaquo tter et al 1999 Kopylova et al 1999MacKenzie amp Canil 1999 Pearson et al 1999 Carbno ampCanil 2002) which have been substantiated by magnetotel-luric and elastic thickness data (Jones et al 2001 PoudjomDjomani et al 2005) Here we focus on our currentknowledge of the mantle evolution in the central SlaveCraton 340 km NE from Yellowknife NT where the Lacde Gras kimberlites were intruded The subcontinentallithospheric mantle (SCLM) sampled by the Lac deGras kimberlites is strongly layered with an ultra-depletedshallow layer and a less depleted deep layer the latter sug-gested to have formed by subcretion of a plume head thatdelivered diamonds containing lower mantle inclusions(Griffin et al 1999 2004 Davies et al 1999 Aulbach et al2007) A Re^Os isochron age of 327034Ga for somesulfides from the deep layer may date the formationof this layer and shows that significantly oldermantle resides beneath some of the 27Ga crust of thejuvenile eastern domain (Contwoyto Terrane) in the SlaveCraton (Aulbach et al 2004a) This isochron ageagrees within uncertainty with one derived from sulfideinclusions in diamonds from the Panda kimberlite(Westerlund et al 2006) just north of Lac de Gras and isconsistent with previous findings that many kimberliteslying in the younger eastern domain appear to havesampled older lower crust and mantle indicative of aneast-dipping trans-lithospheric boundary dating backto 27Ga craton amalgamation (Grulaquo tter et al 1999Irvine et al 2003)

SAMPLES AND ANALYTICALTECHNIQUESAll samples of the present study were retrieved fromdrill core of kimberlite pipe A154S in the Lac de Grasarea from depths between 100 and 470m Samples ofat least 1cm in size were cut out of the drill core andprocessed to minimize sampling bias although thispossibly discriminates against more friable eclogitetypes that are not as well preserved Eighteen eclogitesand pyroxenites were selected for the present study andresults combined with those of Pearson et al (1999and unpublished data) to make up a total of 35 eclogitesand 30 pyroxenitesMajor-element analyses were obtained using a

CAMECA Camebax SX50 electron microprobeIn situ trace-element compositions were determinedwith a custom-built laser-ablation system (designed byS E Jackson) or a Merchantek LUV 266 NdYAG UVlaser system both linked to an Agilent 7500 inductivelycoupled plasma mass spectrometry (ICP-MS) system andreduced using the GLITTER software (van Achterberget al 1999) Accuracy and precision were monitored by

analysing basalt standard BCR-2G with each batch ofsamples as an unknown Results standard deviations anddetection limits are given in Electronic Appendix 1 whichcan be downloaded from httppetrologyoxfordjournalsorg Sr^Nd^Hf isotope data for garnet and clinopyroxene(cpx) separates leached and ultrasonicated in 6N HClfor 30min followed by ultrasonication in three aliquots ofMQ prior to dissolution were obtained using a Nu Plasmamulti-collector (MC) ICP-MS system All analytical workwas carried out in the GEMOC National Key Centreat Macquarie University (wwwesmqeduauGEMOC)following the techniques described by Aulbach et al(2004b) Repeated measurements of standard materialsduring data acquisition (February^July 2003) yielded thefollowing values 87Sr86Sr of 07102570000045 (2 SDnfrac14 42) for the SRM-987 standard 87Sr86Sr of0703520000040 (nfrac1415) for BHVO-1 143Nd144Nd of05111380000038 (nfrac14 42) for the JMC-321 standard143Nd144Nd of 05130030000043 (nfrac1410) for BHVO-1176Hf177Hf of 02821630000002 (nfrac14 20) for the JMC-475 standard and 176Hf177Hf of 02831280000090 forBHVO-1 (nfrac14 8) (Aulbach et al 2004b)Although the uncertainties of Sr and Nd isotope ratios

for pure Sr and Nd standards obtained by MC-ICP-MSare two to three times higher than those reported for ther-mal ionization mass spectrometry (TIMS) measurements(httpgeoremmpch-mainzgwdgde) they are adequatefor the samples analysed in the present study whichmostly have significantly evolved isotopic compositionscompared with depleted mantle or CHUR The signifi-cantly higher uncertainty for 176Hf177Hf of the rock stan-dard than for the pure Hf standard may reflect matrixinterference as a result of imperfectly purified solutions(eg Blichert-Toft et al 1997)

PETROGRAPHY ANDCLASSI F ICAT IONThe petrography of the studied eclogites from Lac deGras is summarized in Table 1 The majority of samplesthat were large enough to obtain microstructural informa-tion have medium- to coarse-grained granoblasticmicrostructures Three eclogites have a weakly tabularmicrostructure which is due to the weak elongation andsubparallel alignment of garnet and cpx in sampleYK3528 and to the preferred orientation of phlogopite insamplesVR19674ecl-g7 and YK1911Most samples are fresh or show minor (510) altera-

tion that is restricted to veins and grain boundaries A fewsamples are more strongly altered with pyroxenesusually more affected than garnet Altered areas arecomposed of small secondary phlogopite sulfide flakesand unidentifiable microcrystalline phases with a feltybrown to green appearance

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1845

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 1 Petrographic data and eclogite classification

Sample no Ref1 Type2 GpMampG3 GpTampN

4 Size5 Microstructure6 Assemblagemode7

gt cpx rut sf ky opx other

Eclogites

SE01 3 high-Mg I A 51 na x x

SE02 3 high-Ca I B 51 na x x x

SE03 3 high-Mg II A 51 na x x

SE04 3 high-Mg II B 51 na x x

SE11 3 low-Mg II B 51 na x x x x

vr09399 2 high-Mg II A 51 na x x

vr19674ecl-g4 2 low-Mg II B 16 na x x

vr19674ecl-g5 2 high-Ca II C 51 na x x

vr19674ecl-g6 2 high-Mg I A 51 na x x

vr19674ecl-g7 1 volatile-rich II A 15 medium-tabular phlog (44) ap (51)

vr19677ecl-g1 2 high-Mg II B 51 na x x

vr19677ecl-g2 2 high-Mg II B 17 na x x x

vr19677ecl-g4 2 high-Mg II B 51 na x x

vr19677ecl-g5 2 high-Mg I B 51 na x x

vr19677ecl-g6 2 high-Mg II A 51 na x x

vr40345 3 high-Ca I B 51 na x x x dia (trace)

vr40364 2 high-Mg II A 51 na x x

vr43452 1 low-Mg I B 45 coarse-grano 48 51 1

vr43465 1 high-Ca I B 2 coarse-grano 41 55 1 graph (3)

vr43469 1 low-Mg II B 15 coarse-grano 60 40 51 51

vr43477 1 low-Mg II B 15 coarse-grano x x x

vr43479 1 low-Mg II A 25 medium-grano 50 50 51

vr43480 1 high-Mg I A 2 medium-grano 89 11 51

vr50858 2 high-Ca II B 51 na x x x

vr50860 2 high-Ca I C 51 na x x

vr50909 3 high-Ca II B 51 na x x x x

vr50916 2 high-Mg II A 51 na x x

vr67112b 1 volatile-rich II C 2 medium-grano 51 ilm (3) cc (2)

vr67360 1 high-Mg II A 3 coarse-grano 62 38 51

yk1911 1 volatile-rich II B 2 medium-tabular 31 18 phlog (49) ap (51) plag (51) cc (51)

yk1926 1 low-Mg II B 17 coarse-grano 70 30 51

yk1943 1 low-Mg I B 2 coarse-grano 57 41 2 51 qz (51)

yk1946 1 high-Mg II A 3 coarse-grano 62 37 1 51

yk1949 1 high-Mg II B 3 coarse-grano 70 30 51 51

yk3528 1 high-Ca I B 25 medium-tabular 57 43 51

Pyroxenites

SE08 3 na II A 51 na x x

SE09 3 na II A 51 na x x x

vr09353 2 na II A 51 na x x x

vr09384 2 na II B 51 na x x x

vr09405 2 na II A 51 na x x

vr09406 2 na II A 51 na x x x

vr09424 3 na II A 51 na x x x

(continued)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1846

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Some xenoliths have subhedral or rounded garnets in alsquomatrixrsquo of interstitial cpx (Group I of MacGregor ampCarter 1970 Fig 2a) whereas in others garnet and cpxhave straight grain boundaries and an interlocking fabric(Group II Fig 2b) The classification of McCandless ampGurney (1989) builds on work of MacGregor amp Carter(1970) and distinguishes group I eclogites by their higherNa2O in garnet (009wt ) and K2O in cpx (008wt) from group II eclogites The classification used inTable 1 (GpMampG) is that based on Na2O contents ingarnet A different classification (GpTampN) places eclogitesinto groups A B and C distinguished by the MgO FeOand CaO contents of garnets and Na2O and MgO contentsof clinopyroxenes (Coleman et al 1965 Taylor amp Neal

1989) These classifications are based on distinct eclogitesuites from Southern African kimberlites and do not neces-sarily reflect the particularities of eclogites from otherlocalities We therefore use a slightly different scheme(lsquotypesrsquo in Table 1) that reflects the specific groupings ofeclogites in this study with regard to Cr2O3^CaO CaO(MgOthornFeO)^MgOFeO and CaO^Na2O relationshipsin garnet (see below)Garnet modes in volatile-free eclogites range from 40

to 90 vol and cpx modes from 10 to 55 with anaverage and median garnet mode of 60 and 65respectively Two samples have high modal amounts ofphlogopite (44 and 48 respectively) in addition togarnet and cpx and are therefore not eclogites sensu stricto

Table 1 Continued

Sample no Ref1 Type2 GpMampG3 GpTampN

4 Size5 Microstructure6 Assemblagemode7

gt cpx rut sf ky opx other

vr19673ecl-g1 2 na II B 51 na x x x

vr19673ecl-g2 2 na II B 51 na x x x x

vr19674ecl-g1 2 na II A 51 na x x x

vr19674ecl-g2 2 na II A 51 na x x x

vr19674ecl-g3 2 na II A 51 na x x

vr19674per-g1 2 na II A 51 na x x

vr19674per-g4 2 na II A 51 na x x x

vr40302 2 na II A 51 na x x x

vr40336 3 na II A 51 na x x x

vr40337 2 na I A 51 na x x

vr40374 2 na II A 51 na x x x

vr40382 2 na II A 51 na x x x

vr40384 ecl-g4 2 na II A 51 na x x x

vr40399 2 na II B 51 na x x x

vr50856 2 na II A 51 na x x x

vr50887 2 na II A 51 na x x

vr50888 2 na II A 51 na x x

vr50899 2 na II A 51 na x x x

vr50900 2 na II A 51 na x x x

vr50925 2 na II B 51 na x x

yk1914 1 na II A 25 medium-grano 22 78

yk1915 1 na II A 15 coarse-grano x x x x

yk1952 1 na II A 2 medium-grano x x x x

1References 1 this study 2 Pearson et al (1999) 3 Pearson et al (unpublished data) trace-element analysesthis study2Classification based on chemical parameters (see text for details)3Classification of McCandless amp Gurney (1989) based on MacGregor amp Carter (1970)4Classification of Taylor amp Neal (1989) based on Coleman et al (1965)5Longest dimension (cm) across xenolith6Only given for samples41 cm in size grano granoblastic7Modes (vol ) are given where available otherwise presence of mineral is indicated as identified in thin section or stubgt garnet cpx clinopyroxene ky kyanite rut rutile sulf sulfide phlog phlogopite ap apatite dia diamond graphgraphite cc calcite ilm ilmenite pl plagioclase qz quartz opx orthopyroxene na not applicable or not available

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1847

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(phlogopite occurring as small secondary grains at grainboundaries and in veins is not considered in this study)Ten samples of all eclogite types show secondary spongyrims of cpx around cpx cores Secondary rims containsmall (mm-scale) patches of glassRutile is the most common primary accessory phase

(15 of 65 eclogites and pyroxenites) with modes between51 and 2 vol Sulfide occurs in 14 samples and ispresent as rounded to subhedral to irregular grains

Kyanite has been identified in three samples studied byPearson et al (1999) One of the kyanite-bearing eclogitesalso contains diamond (VR40345) Kyanite as well asdiamond or graphite associated with kyanite-bearing andcompositionally similar eclogites (high-Ca eclogites) hasbeen frequently observed in drill core (Pearson et al1999) but because these eclogite types are often alteredand friable they are not proportionally represented in thepresent study One high-Ca eclogite (VR43465) contains

Fig 2 Eclogites in thin section (plane-polarized light) illustrating the microstructures mentioned in the text (a) high-Mg eclogiteVR67360(b) high-Mg eclogiteYK1946 (c) phlogopitized eclogiteYK1911 (d) garnet pyroxeniteYK1915 (e) garnet pyroxeniteYK1914 (f) garnet pyrox-enite schlieren in a matrix of garnet spongy opx and glass (YK1952) Grt garnet cpx clinopyroxene rut rutile phlog phlogopite Scalebar in(c) applies to all sections (1cm)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1848

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

3 vol graphite which occurs as disseminated plates sev-eral millimetres long with prismatic tabular or irregularhabit Quartz has been retrieved from a mineral separateof low-Mg eclogite YK1943 The phlogopitized samples(VR19674ecl-g7 and YK1911) also contain apatite andglass as accessory phases Vesicles in sample VR67112bare partially filled with single or multiple large carbonatecrystalsMost pyroxenites from Lac de Gras have fine- to coarse-

grained granoblastic equilibrated microstructures(Pearson et al 1999) Modal information for pyroxenitesin the present study is available for only the larger sampleYK1914 (22 garnet 78 cpx) (Fig 2e)

MAJOR ELEMENTSAverage major-element compositions of garnet cpx andopx in the different eclogite types (except for the threevolatile-rich eclogites which are not averaged) and in pyr-oxenites are given in Table 2 the full datasets includingaccessory minerals are available as ElectronicAppendices 2^5 which can be downloaded from httppetrologyoxfordjournalsorg

GarnetGarnets have XMg [pyrope component 100Mg(MgthornFethornCathornMn)] ranging from 178 to 756 andXCa (grossular component mol) ranging from 72 to359 (Electronic Appendix 2) In a diagram of CaO(MgOthornFeO) vs MgOFeO (Fig 3) eclogitic garnetsfall into three groups (1) low-Mg garnets with low

CaO(MgOthornFeO) and low MgOFeO (2) high-Mggarnets with low CaO(MgOthornFeO) and high MgOFeO (3) high-Ca those with high CaO(MgOthornFeO)and intermediate MgOFeO Volatile-rich eclogites havegarnets with the lowest MgOFeO whereas pyroxenitestrend towards higher values than eclogitesMost eclogite garnets have Cr2O3 contents502wt

with variable CaO contents (Fig 4) Garnets in pyroxeniteshave low CaO contents which correlate positively withCr2O3 More than half of the pyroxenites are opx-free andall pyroxenites are olivine-free on thin-section scalePyroxenitic garnets show an opposing trend of CaOvs Na2O (Fig 5) compared with the eclogitic garnetsBecause the positive correlation of Cr2O3 and CaO forpyroxenitic garnet indicates buffering by both cpx andopx opx-bearing or opx-free pyroxenites will not befurther distinguished subsequently (except in the geother-mobarometry section where opx-bearing assemblagesallow simultaneous calculation of pressure andtemperature)Distinct rim compositions in most garnets are

characterized by lower CaO contents and higherMg-number [frac14 100Mg(MgthornFe)] Garnet compositionsin three samples are inhomogeneous with regard to CaOMgO and FeO (SE01 VR43479 and VR43480) withoutclear core^rim zonations similar to coexisting cpx Likerims CaO-poor compositions are characterized by higherMg-number Visibly secondary spongy cpx rims also havehigher MgO contents (see below) and MgO-rich garnetsare therefore regarded as affected by secondary changesBased on Cr^Ca Mg^Ca^Fe and Ca^Na composi-

tional relationships in garnet the following eclogitetypes are distinguished (1) eclogites with high-Ca

0

05

10

15

20

25

30

0 01 02 03 04 05 06 07 08 09

high-Ca

high-Mg+pyroxenites

low-Mg

CaO(MgO+FeO)

MgO

FeO

GARNETpyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropic Jericho massiveDiavik low-temperatureDiavik high-temperatureDiavik diamond-bearing

eclogites

Fig 3 CaO(MgOthornFeO) vs MgOFeO (wt ) in garnet Fieldsdistinguish low-Mg and high-Ca eclogites as well as high-Mg eclogitesplus pyroxenites in this study Garnets in eclogites from Jericho(Kopylova et al 1999) and Diavik (Schmidberger et al 2007) areshown for comparison Garnet inclusions in diamond from Davieset al (1999 2004)

-135 wt (high-Ca)

2

3

4

5

6

7

8

9

0 05 10 15 20 25Cr2O3

CaO

GARNET

opx-bearing

opx-free

pyroxenites

low-Mghigh-Mgvolatile-rich

Jericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

VR19673ecl-g1g2

Fig 4 Cr2O3 vs CaO (wt ) in garnet Garnets in eclogites fromJericho and Diavik as in Fig 3

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1849

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 2 Summary of major-element contents in garnet and cpx (wt )

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO Na2O Total Mg-no T(Krogh) Type

Garnet

Low-Mg ecl 3964 026 2214 011 1843 037 1050 846 009 1000 503 976

minimum 3893 008 2130 5009 1071 024 849 567 006 990 423 860

maximum 4094 036 2275 011 2093 046 1457 1062 012 1004 708 1088

1s 070 012 052 na 354 007 227 185 003 05 99 98

High-Mg ecl 4081 026 2261 013 1419 030 1537 624 007 1000 659 969

minimum 3970 008 2224 5009 951 023 1127 418 003 991 551 820

maximum 4250 052 2323 035 1795 036 2030 882 013 1012 792 1241

1s 077 013 031 na 225 004 189 141 003 05 57 88

High-Ca ecl 3979 034 2211 na 1516 027 958 1240 011 998 528 1078

minimum 3909 011 2117 5009 1164 023 801 1144 007 989 456 910

maximum 4039 073 2242 011 1713 035 1249 1351 018 1004 657 1246

1s 042 026 042 na 173 004 155 073 004 06 67 151

vr19674ecl-g7 3867 003 2187 5009 2380 153 793 627 5004 1002 373 volatile-rich

vr67112b 3783 013 2097 009 2620 204 466 790 5004 999 241 volatile-rich

yk1911 3874 5006 2200 5009 2120 221 802 747 5004 997 403 volatile-rich

Pyroxenites 4163 021 2257 075 1148 034 1906 407 006 1002 747 909

minimum 3989 005 2081 009 729 025 1535 277 004 996 593 699

maximum 4279 048 2330 260 1881 054 2215 566 009 1011 842 1158

1s 068 010 057 057 323 007 209 065 001 04 73 117

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO Na2O Total Mg-no Type

Cpx

Low-Mg ecl 5453 028 739 na 500 5009 1117 1662 434 994 801

minimum 5160 017 564 5009 337 na 876 1388 197 981 741

maximum 5570 044 1067 014 687 na 1324 1926 632 1001 865

1s 128 011 155 na 124 na 153 177 126 06 40

High-Mg ecl 5483 023 483 na 381 na 1368 1875 318 994 865

minimum 5353 009 263 5009 245 5009 1156 1617 204 986 778

maximum 5645 031 810 016 674 011 1588 2101 495 1011 920

1s 071 007 137 na 095 na 104 138 075 06 30

High-Ca ecl 5563 027 1184 na 318 5009 868 1368 615 995 836

minimum 5448 014 769 5009 168 na 619 1060 391 987 751

maximum 5670 049 1579 010 603 na 1148 1689 807 1003 878

1s 076 015 334 na 151 na 189 261 167 06 45

vr19674 ecl-g7 5322 010 185 5009 536 015 1448 2339 062 992 828 volatile-rich

vr67112b 5040 041 328 5009 1079 024 1108 2229 071 992 647 volatile-rich

yk1911 5231 014 301 5009 544 024 1450 2340 056 996 826 volatile-rich

Pyroxenites 5527 na 347 038 304 na 1543 1959 245 999 899

minimum 5299 5006 037 010 149 5009 1128 1408 030 990 804

maximum 5666 037 786 134 490 014 1863 2312 537 1009 947

1s 068 na 189 028 091 na 193 226 131 05 37

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO NiO Total Mg-no

Opx

Pyroxenites 5781 na 059 na 627 na 3508 037 011 1006 909

minimum 5546 5006 032 5009 462 5009 2999 020 008 996 804

maximum 5888 009 318 054 1304 014 3628 077 014 1014 933

1s 098 na 067 na 212 na 160 019 002 06 32

Averaged for the three eclogite types and for pyroxenites minimum and maximum values as well as standard deviationsalso given Concentrations in volatile-rich eclogites are shown individually ecl eclogite Mg-numberfrac14 100Mg(Mgthorn Fe)Temperatures of last equilibration [T(Krogh)] as in Electronic Appendix 8

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1850

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

garnets (these include kyanite- graphite- and diamond-bearing varieties) (2) eclogites with low-Mg garnets (theseinclude quartz-bearing eclogite) (3) eclogites with high-Mggarnets and Cr2O3 contents502wt (4) volatile-richeclogites (phlogopite- apatite- carbonate-bearing)

Clinopyroxene (cpx)Clinopyroxenes in all eclogite groups have higher averageAl2O3 contents than those in pyroxenites with the formerhaving about 90 of the total Al in the jadeite molecule(AlVI) versus about 35 for the latter corresponding toan average of 41 and 25wt Na2O respectivelyThe highest Al2O3 and Na2O contents in the dataset areobserved in cpx from the high-Ca eclogites (up to158wt Al2O3) (Electronic Appendix 3) Contents ofK2O are generally below the detection limit(5004wt ) but values up to 024wt are observed insome cpx in high-Ca eclogites Average cpx Mg-numberare highest in pyroxenites (899) followed by cpx inhigh-Mg eclogites (865) high-Ca eclogites (836) andlow-Mg eclogites (801) Pyroxenitic cpx tend to havelower TiO2 and higher NiO and distinctly higherCr2O3 contents compared with those in eclogites A plotof MgO vs Al2O3 shows a negative correlation (Fig 6)Clinopyroxenes in low-Mg eclogites have lower Al2O3

contents at a given MgO content than other eclogitetypes Between 15 and 17wt MgO the Al2O3^MgOcorrelation for cpx in pyroxenites has a different slopefrom that in the eclogites Clinopyroxenes in pyroxenites

VR19673ecl-g1 and VR19673ecl-g2 have distinctly lowerMgO contents and higher Al2O3 contents than inthe other pyroxenites They coexist with garnets havingthe lowest Cr2O3 and CaO contents of the dataset whichdefine one end of the positive Cr2O3^CaO correlationobserved for garnetsClinopyroxenes in volatile-rich eclogites have distinctly

high MnO contents (up to 024wt ) Mg-numbers forcpx in the phlogopite-bearing variety are 826 and 828whereas that in the calcite-bearing eclogite (VR67112b)has an Mg-number of 647Some cpx grains have spongy rims that contain small

(mm-scale) patches of glass these rims are richer in Caand Mg and markedly poorer in Na than the cores similarto cpx in eclogites from southern Africa (Taylor amp Neal1989) The rim texture is unequilibrated suggesting lategrowth The formation of the spongy cpx rims might berelated to incongruent melting of cpx or be a result ofmetasomatism (Taylor amp Neal 1989 and referencestherein) Distinct although not spongy rims in two eclo-gites (VR67360 and YK1946) are similarly enriched inCaO and depleted in Na2O Two types of cpx with respectto CaO Na2O and Mg are present in VR43477VR43479 and VR43480 without clear core^rim zonationor spongy rims No significant differences are recognizedwith respect to other oxides

Orthopyroxene (opx)Orthopyroxenes have Mg-number ranging from 804 to933 (17 of 19 samples have values between 895 and 933)and typically contain minor amounts of Al2O3

(032^118wt one outlier has 318wt ) CaO(020^118 wt ) and NiO (008^014wt ) (Electronic

0

2

4

6

8

10

12

14

7 9 11 13 15 17 19

Al 2

O3

CPX

pyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

cpx in pyroxenitesVR19673ecl-g1g2

MgO

Fig 6 MgO vs Al2O3 (wt ) in cpx Data sources as in Figs 3and 4

0

005

010

015

020

2 4 6 8 10 12 14 16

CaO

Na 2

O

dl

group Igroup II

pyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

GARNET

Fig 5 CaO vs Na2O (wt ) in garnet Arrows show trends foreclogites and pyroxenites respectively [note that part of thepyroxenite trend extends into the field falling below the detectionlimit (dl) for Na2O] Group I and group II refer to classificationof McCandless amp Gurney (1989) Data sources as in Figs 3 and 4

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1851

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Appendix 4) The contents of all other elements are mostlybelow their respective detection limits

Accessory phasesRutile and ilmenite occur in some of the eclogites (Table 1)All rutile grains have exsolved 10 ilmenite as lamellaeand rims Ilmenite-free areas of rutile in eclogites containbetween 962 and 988wt TiO2 and minor amounts ofAl2O3 (01^04wt ) Cr2O3 (5009 to 02wt ) andFeO (03^07wt ) Rutile in pyroxenite YK1915 contains004wt Al2O3 01wt FeO and 15wt Cr2O3Discrete ilmenite grains in sample VR67112b haveTiO2 contents of 499 and 500wt MgO of 23 and15wt and FeO of 399 and 403wt respectivelySulfides occur interstitially or enclosed in cpx garnet

and opx they are mostly pyrrhotite and monosulfide solidsolution with variable Ni contents (unpublished data)They will not be considered further hereAbundant platy phlogopite is observed in two samples

(VR19674ecl-g7 and YK1911) It is Cl-rich in sampleVR19674ecl-g7 (063wt vs 003wt in YK1911) andF-rich in sample YK1911 (097wt vs 005wt inVR19674ecl-g7) (Electronic Appendix 5) The same is truefor coexisting apatite (VR19674ecl-g7 254wt Cl and014wt F YK1911 006wt Cl and 306wt F)

Comparison with mineral inclusions indiamonds from Lac de Gras and witheclogite xenoliths from other Slave CratonlocalitiesMineral inclusions in diamonds from the Lac de Graskimberlites have been investigated by Davies et al (19992004) For garnets there is a striking similarity betweenthe inclusions in diamond and those in the high-Ca eclogitexenoliths although the inclusions in diamond trendtowards higher Na2O (Fig 5) and lower Al2O3 contentsThe overlap is not as marked for cpx where inclusionsin diamond span almost the entire range observed foreclogitic cpx (Fig 6)A comparison with eclogites from other localities in

the Slave Craton shows that garnets in most Diavik low-temperature eclogites reported by Schmidberger et al(2007) plot with garnets in high-Mg eclogites and inpyroxenites from this study (Figs 3^5) whereas garnets inhigh-temperature and diamond-bearing eclogites(Schmidberger et al 2007) overlap most with garnets inhigh-Ca eclogites from this study and with garnetsincluded in diamond Eclogites from the Jericho kimberlitehave either massive or anisotropic fabrics (Kopylova et al1999) Garnets in almost all massive eclogites plot withgarnets in low-Mg eclogites from this study with regard toCaO^MgO^FeO relationships (Fig 3) but many havelower Na2O contents at a given CaO content (Fig 5)Garnets in anisotropic eclogites are restricted to lowMgOFeO (507) and show some overlap with garnets

in high-Ca eclogites although trending towards lowerCaO(MgOthornFeO) (Fig 3) while coexisting cpx is alsorestricted to low MgO and trends toward high Al2O3 con-tents (Fig 6)

TRACE ELEMENTSGarnetTrace-element abundances in garnet are summarized inTable 3 and the complete dataset is reported in ElectronicAppendix 1 Garnets in low-Mg eclogites show steep posi-tive slopes between LaN and SmN and shallower positiveslopes in the normalized heavy rare earth elements(HREEN Fig 7a) The normalized light REE (LREEN)show an order of magnitude variability ZrHf is mostlysupra-chondritic and Ti can be enriched or depleted rela-tive to elements of similar compatibility Niobium in threesamples is strongly depleted at 002^003 chondriticGarnets in high-Mg eclogites also show smooth positiveslopes in LREEN but have flat HREEN and variablenormalized middle REE (MREEN) to HREEN (Fig 7b)ZrHf is always supra-chondritic and Ti enriched ordepleted relative to similarly compatible elementsNiobium is most samples is below detection Small positiveEu anomalies though not outside the analytical uncer-tainty for Eu relative to either Sm or Gd are observed forgarnet in some high-Mg eclogites (eg EuEu up to 15for VR67360 where Eu is the average of the chondrite-normalized Sm and Gd concentrations) Garnets in high-Ca eclogites are slightly enriched in the MREEN relative tothe HREEN or have flat MREEN and HREEN with ahump between EuN and HoN (Fig 7c) One sample(VR40345) shows strong La enrichment (greater thanchondrite) Two samples have supra-chondritic ZrHfwhereas three samples have close to chondritic ratios andone sample has a subchondritic ratio As is true for theother eclogite typesTiN is either enriched or depleted rela-tive to similarly compatible elementsPyroxenitic garnets have variable slopes in the LREEN

and flat MREEN to HREEN patterns and two samplesshow high Sc relative to neighbouring elements Garnet inone of the pyroxenites (YK1952) has a highly distinctivepattern relative to other pyroxenites with very low LREEand HFSE (Fig 7d) This sample consists of coarse pyroxe-nite schlieren in which the garnet was analysed in a finer-grained lsquomatrixrsquo of spongy cpx with secondary rims finecpx^opx intergrowths embayed garnet and melt patchesVolatile-rich eclogites have garnets that are slightly

enriched in MREEN relative to HREEN or have flatMREE to HREE patterns One garnet has a pronouncednegative Eu anomaly (EuEufrac14 07 forVR67112b Fig7e)Garnets in volatile-rich samples have the lowest Ni Coand Ti and the highest V ScYand HREE abundances inthe dataset

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1852

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 3 Summary of trace-element contents in garnet and cpx (ppm)

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR19674

ecl-g7

VR67112b YK1911 av min max 1s

n 6 6 7 3 5 4 4

P 169 54 270 88 196 70 310 78 282 80 450 138 160 60 80 214 48 370 144

Sc 62 40 89 16 48 36 67 11 53 42 79 15 69 92 93 186 52 440 179

Ti 1667 490 2500 680 1073 620 2000 455 2141 700 4700 1582 270 690 260 1097 540 1900 573

V 169 78 250 68 150 78 280 72 178 110 390 101 140 340 100 390 71 820 326

Co 59 42 70 11 58 39 84 13 58 47 65 5 30 41 27 68 55 90 16

Ni 17 5 32 10 21 11 34 10 24 11 35 8 6 10 51 54 25 110 38

Ga 12 7 15 3 13 8 16 3 12 9 20 4 8 86 58 12 10 14 2

Sr na 512 4 na na 522 522 na na 57 57 na 52 5025 510 513 513 513 na

Y 37 6 61 18 21 7 38 10 31 20 38 5 60 81 70 19 12 26 6

Zr 15 5 29 9 28 6 83 26 24 9 43 12 60 22 90 18 1 41 17

Nb 009 0005 048 019 na 50015 05 na na 5005 05 na 5003 0012 5001 022 0030 057 025

La 005 001 009 004 na 50007 00 na na 5003 04 na 004 002 0009 003 000 009 005

Ce 03 01 05 02 01 00 01 00 05 01 18 05 03 022 015 01 00 03 01

Pr 01 00 02 01 00 00 01 00 02 01 03 00 017 014 012 00 00 01 00

Nd 12 03 24 07 07 04 14 03 21 12 32 07 24 23 20 03 00 07 03

Sm 11 04 22 06 10 03 18 05 23 08 36 10 29 31 33 05 01 09 04

Eu 06 03 10 03 06 02 10 03 11 04 17 04 16 114 16 03 01 06 02

Gd 27 07 41 12 23 07 42 12 43 16 53 13 8 81 90 13 06 24 08

Dy 54 11 70 23 36 13 65 18 58 35 71 11 12 144 12 29 18 42 10

Ho 14 02 23 07 08 03 15 04 12 09 14 02 24 30 23 07 04 10 02

Er 46 07 80 24 24 09 40 10 34 25 40 05 8 85 6 22 14 28 06

Yb 52 06 98 30 24 10 37 08 32 22 39 05 8 70 5 23 13 30 07

Lu 09 01 16 05 04 02 05 01 05 03 07 01 11 09 08 04 02 05 01

Hf 03 02 05 01 04 01 10 03 05 02 09 02 09 027 007 03 00 06 03

Ta na 50008 5004 na na 50005 001 na na 50004 na na 003 50002 50005 na 50007 011 na

Pb na 5005 02 na na 5002 02 na na 5003 07 na 02 004 5002 na 5003 01 na

Th na 5001 005 na na 50007 003 na na 50008 011 na 5007 0012 50004 002 0005 006 002

U na 5003 5003 na na 5001 01 na na 5007 na na 01 0023 010 na 5001 003 na

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR67112b YK1911 av min max 1s

n 6 4 8 5 4 4

P 65 15 120 41 39 14 56 16 53 30 70 12 20 9 80 27 120 40

Sc 22 10 30 7 15 14 19 2 14 7 24 5 80 43 17 11 21 5

Ti 1557 990 2400 508 1168 1010 1490 191 1988 810 3500 1154 2390 1700 689 102 1685 694

V 415 180 480 107 318 210 400 69 324 259 450 72 600 180 250 139 480 156

Co 35 26 50 8 20 14 31 7 26 14 42 11 49 23 29 20 42 10

Ni 164 87 220 47 228 100 430 128 188 135 320 56 59 34 428 410 670 184

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1853

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

ClinopyroxeneTrace-element abundances in cpx are summarized inTable 3 and the complete dataset is reported in ElectronicAppendix 6 Clinopyroxenes in low-Mg eclogites havemildly positive to negative slopes from LaN to NdN andsmooth negative slopes between SmN and HoN thatmostly continue to LuN (Fig 8a) Niobium is below detec-tion or 5007 chondritic and SrN is variably enrichedrelative to neighbouring elements with the peak size corre-lating negatively with LREEN ZrHf is subchondritic andHfN is enriched relative to SmN whereasTiN is enriched ordepleted relative to similarly compatible elementsWe ana-lysed the spongy rim in cpxYK1943 and its trace-elementpattern is similar to that of cpx cores for more compatibleelements whereas the most incompatible elements areseveral orders of magnitude higher Clinopyroxenes in

high-Mg eclogites have steeper negative slopes betweenNdN and LuN than those in low-Mg eclogites combinedwith lower HREE abundances HfN is depleted relative toSmN (Fig 8b) High-Ca eclogites contain cpx with sinusoi-dal REEN patterns with lower LREE than other eclogitetypes and the heaviest REE below detection They haveuniformly strong positive Sr peaks relative to elements ofsimilar compatibility (Fig 9) The diamond-bearing high-Ca eclogite (VR40345) has cpx with unusually high Nband LREE abundancesClinopyroxenes in four pyroxenites have disparate trace

element patterns They all share depletions in TiN relativeto similarly compatible elements and two show positiveSrN peaks Two samples have cpx with high and flatLREEN to MREEN patterns and steep negative slopes inthe HREEN The trace-element patterns of cpx in YK1914

Table 3 Continued

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR67112b YK1911 av min max 1s

n 6 4 8 5 4 4

Ga 16 10 20 3 12 9 15 2 26 15 39 9 129 123 6 11 11 5

Rb na 5008 04 na na 5007 05 02 na 5003 03 na 02 5004 03 010 05 0

Sr 166 85 290 74 222 153 285 50 243 83 440 142 85 120 200 11 414 209

Y 2 04 4 10 13 07 2 04 11 03 3 10 28 51 2 03 3 2

Zr 19 4 48 15 25 8 45 13 17 8 28 7 111 129 36 008 81 42

Nb 01 0003 03 01 na 5002 00 na na 5002 16 9 007 011 010 003 02 008

Cs na 5001 6 na na 5006 5006 na5013 5004 5013 na bdl 5002 na 5001 5002 na

Ba 09 02 2 07 09 01 2 07 10 03 3 10 06 01 13 06 3 10

La 3 03 8 3 3 23 3 04 4 01 14 6 51 97 4 01 10 4

Ce 7 13 20 6 10 7 13 2 7 05 24 10 25 28 11 01 28 13

Pr 11 02 3 08 2 10 3 06 08 02 2 09 47 46 2 002 4 2

Nd 5 12 12 4 10 5 14 4 4 09 8 26 23 22 8 01 22 10

Sm 11 05 21 06 20 10 31 10 07 02 11 03 47 53 2 005 5 2

Eu 03 01 05 01 06 028 09 03 02 010 03 007 087 123 06 002 15 07

Gd 08 03 13 03 11 05 19 06 05 03 06 011 28 37 15 010 36 2

Dy 05 01 07 02 04 022 06 01 02 013 03 007 11 16 07 008 14 06

Ho 01 00 02 004 006 003 007 002 na 5003 012 na 011 021 009 001 02 007

Er na 507 507 na 011 004 02 00 na 5006 03 na 019 033 02 002 03 012

Yb na 5001 02 na na 5001 011 na na 5006 03 na 007 015 009 001 02 006

Lu na 5001 003 na na50002 002 na na50008 01 na 0006 0013 0013 0002 003 001

Hf 12 03 24 07 11 04 18 05 10 06 16 04 34 53 na50015 35 na

Ta na50003 50015 na na50006 006 na na50004 5004 na 0014 00236 na50003 503 na

Pb 06 02 12 04 06 04 09 02 na 5016 07 02 065 098 08 01 18 1

Th na50005 06 na 012 005 02 005 na 5002 5006 na 021 02 001 06 03

U na 5005 5005 na 002 0009 005 001 na 5003 5005 na 0013 024 na50005 003 na

Averaged for the three eclogite types and for pyroxenites as in Table 2 pyx pyroxenite n number of analyses averagedbdl below detection limit

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1854

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

and YK1952 are similar although at different abundanceswith only modest LREEN over HREEN enrichment andstrong depletions in ZrN relative to neighbouring elements(Fig 8d)Clinopyroxenes in the volatile-rich samples have flat pat-

terns and high abundances from LaN to SmN and smoothnegative slopes from SmN to LuN Niobium Sr and Ti are

noticeably depleted relative to elements of similar compat-ibility (Fig 8e)

Trace-element partitioningTo assess whether coexisting phases are in trace-elementequilibrium we calculated apparent cpx^garnet distribu-tion coefficients and compared them with those

(a)

(d)

(b)

(e)

(c)

GarnetCI-chondrite GarnetCI-chondrite

low-Mghigh-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1952

1

10

Nd Sm Eu Gd Dy

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Fig 7 Extended trace-element patterns of garnet in (a) low-Mg eclogites (b) high-Mg eclogites with inset showing enlargement of MREE tohighlight Eu anomalies and 1s uncertainties (c) high-Ca eclogites (d) pyroxenites and (e) volatile-bearing eclogites Normalized to chondritevalues of McDonough amp Sun (1995)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1855

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

determined experimentally or in demonstrably equili-brated natural samples Low-Mg and high-Mg eclogiteshave Dcpxgarnet for most elements that are within therange determined for natural equilibrated eclogites andexperimental assemblages (Harte amp Kirkley 1997 Greenet al 2000 Barth et al 2002a) (Electronic Appendix 7)High-Ca eclogites plot at the low end of the range

of distribution coefficients and have very low DMREE

and DHREE outside the range indicated by the datasetof Harte amp Kirkley (1997) who noted that DREE decreaseswith increasing Ca content in eclogitic minerals Theseassemblages are therefore considered to be in equilibriumPyroxenite VR40384 has D values similar to eclogites

but at the high end of the range determined by Harte amp

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

(a)

(d)

(b)

(e)

(c)

CpxCI-chondrite CpxCI-chondrite

low-Mg high-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1914

YK1952

spongy rim

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

VR40345

Fig 8 Extended trace-element patterns of cpx as in Fig 7

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1856

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Kirkley (1997) (Electronic Appendix 7) Another group(YK1914 and YK1952) plots at the low end but DHf DZrDTi and DV are far below the values indicated by experi-mental datasets Because pyroxenites compositionallygrade into eclogites their drastically different DHFSE

values suggest disequilibrium In contrast volatile-bearingsamples have DHFSE values that appear to be too high to bein equilibrium however this may be a consequence oftheir unusual bulk chemical composition (Ti- and K-rich)which differs substantially from that of lsquonormalrsquo eclogitesand experimental charges

GEOTHERMOBAROMETRYFollowing Schmidberger et al (2007) temperatures oflast equilibration were calculated for eclogitic andopx-free pyroxenitic assemblages using the formulationof Krogh (1988) The results are summarized in Table 2and Electronic Appendix 8 for a preset pressure of5 GPa and illustrated as a histogram in Fig 9 Foropx-bearing pyroxenites temperatures were calculatedsimultaneously with pressure [Ca-in-opx thermometryand Al-in-opx barometry using the formulations of Breyet al (1990)] to constrain their relative positions in themantle columnMinerals in some eclogites from Lac de Gras show dis-

tinct core and rim compositions or inhomogeneitiesTemperatures were not calculated for samples withspatially irresolvable inhomogeneous garnet compositions(SE01VR43479VR43480) For eclogites with spongy cpxcore compositions were used in the calculationTemperatures were calculated for both core and rim com-positions where applicable For all samples these

temperatures agree within the uncertainty of this thermo-meter (508C Brey amp Kolaquo hler 1990)Most equilibration temperatures for eclogites lie between

800 and10008C (Fig 9)The transition between the shallowand the deep lithospheric mantle layers in the central Lacde Gras area lies in this temperature interval (Griffin et al1999 Pearson et al 1999) Eclogites in this study lack thedistinct bimodal temperature distribution observed bySchmidberger et al (2007) but as in their study high-Caeclogites give higher average temperatures (10808C corre-sponding to a depth of 155 km if equilibrated along a 40mWm2 geotherm) than other eclogites types (9808Cdepth 135 km) the diamond-bearing sample (VR40345)gives thehighest temperature of the datasetHighNa2Ocon-tents in garnet from high-Ca eclogites are qualitativelyconsistent with higher equilibration pressures (Sobolev ampLavrentrsquoyev 1971) as are high K2O contents in coexistingcpx (Erlank ampKushiro1970) suggesting that high-Ca eclo-gites not only equilibrated at high temperature but also highpressure rather than representing a temperature excursionPyroxenites both opx-bearing and opx-free give somewhatlower average temperatures than eclogites (mean 9258C)and phlogopite-rich eclogites give the lowest temperatures(685 and 7698C respectively) The calcite-bearing eclogiteVR67112b gives a relatively low equilibration temperatureof 8558C

MEASURED ANDRECONSTRUCTEDWHOLE -ROCKSRationale and approachKimberlite-borne xenoliths sometimes show evidence forinfiltration of the host kimberlite (eg Taylor amp Neal1989 Barth et al 2001 2002b Heaman et al 2006) Wetherefore follow the standard approach of calculatingwhole-rock compositions from the compositions of the con-stituent phases weighted by their modal abundances Onlycore compositions were used as distinct rim compositionsmay point to late diffusion of elements from kimberlite-related melt veinsMany eclogitic whole-rock compositions reconstructed

with only garnet and cpx have negative Ti anomalies(not shown) For most of these samples rutile or ilmenitehas been identified as part of the mineral assemblage andthese phases should therefore be considered in the whole-rock reconstruction Rutile or ilmenite modes were esti-mated by assuming that the bulk-rock has no Ti anomalywhen normalized to a normal mid-ocean ridge basalt(N-MORB) composition and calculating the amount ofrutile required to erase theTi anomalyIt is of course possible that the eclogites had negative

or positive Ti anomalies relative to elements of similarcompatibility (Eu and Dy) Because we reconstructed

0

2

4

6

8

10

12volatile-rich

pyroxenite opx-bearing

pyroxenite opx-free

high-Ca

high-Mg

low-Mg

700750

800850

900950

10001050

11001150

12001250

Fre

quen

cy

Temperature

Fig 9 Histogram ofTKrogh (Krogh 1988) obtained at a preset pres-sure of 5 GPa except opx-bearing pyroxenites where TBKN wassolved simultaneously with PBKN (Brey amp Kolaquo hler 1990)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1857

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

whole-rock compositions from mineral core compositionskimberlite infiltration is unlikely to have caused fractiona-tion of Ti from Eu and Dy In a study in which rutilemodes were estimated from the differences between Ti inmeasured bulk-rocks and in bulk-rocks reconstructed withonly garnet and cpx only three of 19 eclogites recon-structed with rutile show Ti anomalies (Barth et al2002a) Even eclogite xenoliths with depleted Zr and Hfabundances that resemble those of modern island arcbasalts do not show Ti anomalies (Jacob amp Foley 1999)We therefore consider our approach to be a reasonablyrobust means of obtaining meaningful rutile modes wheresmall sample sizes preclude reliable determination of themodes of accessory phases by point-countingWhere rutilewas not recognized on thin-section scale or not analysedthe average eclogitic rutile composition was used for thereconstruction

UncertaintiesReconstructing whole-rock compositions in this way is arather crude method given the coarse grain size of theeclogites relative to the sample size and the non-uniformdistribution of the principal minerals on thin-sectionscale The uncertainty in mineral modes of medium- tocoarse-grained eclogites with grains45mm has been esti-mated by Jerde et al (1993) to be 10 plus uncertaintiesrelated to partial alteration which is minor in our samplesNevertheless the whole-rock trace-element budget israther insensitive to the exact modes for common eclogiticassemblages and REE patterns which encompass mostother elements in terms of compatibility in garnet andcpx do not change significantly even for modal variationsof 30 (Jerde et al 1993) This is not true for major-element concentrations which vary substantially with var-iations in mineral modes and we therefore do not calculatewhole-rock major-element contentsBecause the discussion that follows is based on trace-

element patterns and trends rather than absolute abun-dances whole-rock reconstruction based on mineralmodes is adequate We are confident that the uncertaintyinherent in this approach is not leading to significant mis-representation of eclogite whole-rock compositions becausethe trace-element patterns of the reconstructed whole-rocks show distinct differences (see below) that correlatewith the assignment to eclogite types based on garnetmajor-element composition and this is independent ofmineral mode Following Jerde et al (1993) in ElectronicAppendix 9 we show variations in trace-element patternfor one sample in each eclogite type for variations ingarnet and cpx abundance from 7030 to 3070 vol This shows that the trace-element patterns are generallyinsensitive to cpxgt ratios with regard to slopes andanomalies such as positive Sr Pb and Eu anomalies withthe exception of the positive Eu anomaly in high-Mgeclogite VR43479 which disappears if a high cpx mode is

assumed (not shown) Variations in rutile modes by 50(Electronic Appendix 9d) lead as would be expected tothe appearance of Ti anomalies but a rutile mode of 05instead of 1 vol does not erase the positive Nb anomalyin low-Mg eclogiteVR50909Because of the robustness of the relative trace-element

abundances in the reconstructed eclogites a garnet propor-tion of 065 and cpx proportion of 035 approximately themedian modal abundances of these minerals in volatile-free eclogites from Lac de Gras are assumed for eclogitesamples that are too small for determination of meaningfulmodes by point-counting

Within- and between-group variationsThe trace-element compositions of reconstructed whole-rocks are given in Table 4 N-MORB-normalized trace-element patterns of reconstructed whole-rocks areshown in Fig 10 The three eclogite groups have distincttrace element patterns that are particularly obvious forthe REE (Fig 11) Low-Mg eclogites have positive Pb andSr anomalies and smooth positive slopes in their MREEN

and HREEN LREEN are enriched relative to MREEN inthree samples (Fig 11a) Three samples have high Nbrelative to similarly compatible elements and comparedwith most other eclogites (Fig 10a) All high-Mg eclogitesdisplay more or less pronounced positive Eu and Sranomalies and most have positive Pb and negative Zrand P anomalies Their N-MORB-normalized REE pat-terns show shallow positive slopes from LaN to PrN orNdNand aweak negative slope from NdN to LuN (Fig11b)VR43479 is a low-Mg eclogite with regard to its garnet

Ca^Fe^Mg composition but has a trace-element patterntypical of high-Mg eclogites and will be discussed withthe latter Conversely sample YK1949 a high-Mg eclogitein terms of its garnet Ca^Fe^Mg compostion is more simi-lar to low-Mg eclogites in terms of its whole-rock trace-element pattern and is plotted in Figs 10a and 11aWhereas the patterns of most samples within the groupresemble each other the N-MORB-normalized incompati-ble-element abundances of the high-Ca eclogites vary bymore than an order of magnitude (Fig 10c) All sampleshave positive Pb anomalies and most have positive Sranomalies Some samples have negative P andor Zranomalies Niobium abundances in three of the samplesare very high similar to those for some low-Mg eclogitesHigh-Ca eclogites tend to have flatter MREEN andHREEN patterns and more uniform HREE abundancesthat the other eclogite types with humps between SmN

and HoNThe phlogopitized sample YK1911 shows a strong posi-

tive Pb anomaly and lower HREE abundances than thecalcite-bearing sample VR67112b which shows a distinctnegative Eu anomaly and a depletion of La relative to Ce(Figs 10d and 11d)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1858

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Trace-element compositions of reconstructed whole-rocks (ppm)

Sample P Sc Ti V Co Ni Ga Sr Y Zr Nb La Ce Pr Nd

Low-Mg eclogites

vr19674-g4 186 50 4649 317 57 59 144 77 24 30 36 13 31 05 24

vr43452 154 39 2253 328 49 84 176 52 32 11 01 02 07 01 08

vr43469 38 65 4377 349 54 91 142 34 28 15 02 05 15 03 17

yk1926 144 45 3406 258 52 87 145 89 18 18 20 09 29 06 33

yk1943 173 46 5732 295 41 39 146 122 25 48 82 34 85 13 64

yk19491 102 34 1693 158 31 49 100 46 10 14 03 09 21 03 17

High-Mg eclogites

vr19677-g2 138 39 5160 291 41 53 119 98 14 22 04 13 53 11 64

vr434792 46 25 1265 129 36 126 84 56 3 7 04 06 25 04 23

vr67360 49 36 1573 138 64 184 111 76 5 9 04 09 38 07 33

yk1946 153 28 4877 192 45 105 124 108 12 48 03 10 38 08 50

High-Ca eclogites

vr19674-g5 267 30 7663 200 41 67 195 32 23 43 77 01 04 02 16

vr40345 0 33 2821 175 47 98 210 179 13 13 66 60 109 11 38

vr43465 142 27 5104 220 31 83 121 197 14 35 29 05 22 05 40

vr50858 295 30 6898 201 44 94 150 48 22 38 06 01 03 02 20

vr50860 68 57 5672 418 57 71 186 109 19 11 02 02 13 03 26

vr50909 238 28 7295 201 42 92 171 48 20 57 13 01 04 02 21

yk3528 91 51 3712 341 53 155 177 49 16 9 01 02 08 02 11

Volatile-rich eclogites

vr67112b 40 84 9242 521 46 30 107 4 40 66 31 26 126 24 125

yk1911 31 39 9428 286 44 82 164 3 23 52 64 18 52 09 46

Pyroxenites

yk1914 111 49 707 207 48 547 44 31 4 3 02 06 19 03 10

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

Low-Mg eclogites

vr19674-g4 11 05 10 22 39 10 29 33 06 12 05 028 24 008

vr43452 04 02 08 13 38 12 41 49 08 05 01 nd 02 na

vr43469 11 05 09 25 44 11 34 36 06 07 02 001 09 129

yk1926 12 05 11 19 30 07 22 24 04 07 03 004 23 090

yk1943 21 08 10 29 42 10 30 32 05 16 05 013 65 017

yk19491 06 03 11 08 14 04 14 18 03 04 02 003 29 115

High-Mg eclogites

vr19677-g2 20 08 12 21 24 05 15 15 03 05 05 007 52 6811

vr434792 05 02 12 05 06 01 04 03 00 03 01 001 98 037

vr67360 06 03 14 06 09 02 06 06 01 02 02 002 86 000

yk1946 18 07 12 21 22 05 13 12 02 11 03 006 56 112

High-Ca eclogites

vr19674-g5 17 09 11 33 43 09 24 23 04 16 05 007 02 011

vr40345 09 04 10 12 22 06 16 14 02 05 nd nd 298 na

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1859

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Continued

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

vr43465 22 09 12 25 26 06 15 14 02 10 04 001 25 040

vr50858 20 08 09 33 41 08 22 21 03 10 01 nd 02 na

vr50860 14 07 11 27 37 08 22 18 03 06 00 nd 06 na

vr50909 20 10 12 33 38 07 20 20 03 16 01 002 02 181

yk3528 08 05 11 20 30 07 18 18 02 05 01 nd 08 na

Volatile-rich eclogites

vr67112b 38 10 07 53 74 15 41 34 04 18 03 012 60 004

yk1911 20 07 08 35 39 07 18 17 02 10 14 nd 73 na

Pyroxenites

yk1914 02 01 11 04 08 02 05 06 01 01 03 003 68 101

aGarnet has CandashMgndashFe relationship typical of high-Mg eclogites but whole-rock trace-element pattern is typical oflow-Mg eclogitesbGarnet has CandashMgndashFe relationship typical of low-Mg eclogites but whole-rock trace-element pattern is typical ofhigh-Mg eclogites

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

(a)

001

01

1

10(c)

001

01

10(b)

VR43452

YK1949

RockNMORB

low-Mg

RockNMORB

high-Mg

(d)

volatile-richhigh-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

phlogopite-bearingYK1911

calcite-bearingVR67112b

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 10 Extended trace-element patterns of reconstructed (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-bearing eclogites Normalized to N-MORB of Sun amp McDonough (1989)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1860

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

RockNMORB

(b)

RockNMORB

(d)

010

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu001

01

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

(a)

001

01

1

10(b)

001

01

1

10(c)

001

01

1

10(e)

high-TSchmidberger et al

low-TSchmidberger et al

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

high-Mglow-Mg

high-Cavolatile-rich

phlogopite-bearingYK1911

calcite-bearingVR67112b

YK1949VR43479

Fig 11 N-MORB normalized REE patterns for (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-richeclogites Shown for comparison in (e) are reconstructed eclogites from Schmidberger et al (2007)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1861

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Both volatile-rich whole rocks display strong negativeSr anomalies Strontium can be sited in calcite or apatiteandor indicate the presence of a Sr-rich mineral that wasnot identified in thin section such as bariteFour of the six reconstructed whole-rocks in the study

of Schmidberger et al (2007) (Fig 11e) show markedLREEHREEN enrichment that is observed only in onehigh-Ca eclogite in our study (Fig 11c) suggesting thatmany of the eclogites they studied were affected by metaso-matic LREE enrichment Two samples in their study thatare not affected by this enrichment show positive Euanomalies similar to the high-Mg eclogites in this studywhereas the distinctive REE pattern of the low-Mg eclo-gites is not recognized among their samples

SR^ND^HF I SOTOPESYSTEMATICSStrontium Nd and Hf isotope data for optically puremineral separates and reconstructed whole-rocks aregiven in Table 5 Spongy rims and alteration are likely tobe mostly removed during leaching in 6N HCl in combina-tion with repeated ultrasonication Despite this rigoroustreatment the reproducibility of Nd in garnet VR43469and in cpx YK1914 is worse than for other samples Thismay be due to inhomogeneities related to late Nd diffusionpossibly from grain boundaries containing a kimberlitecomponent The reproducibility of the 176Hf177Hf data isgenerally poorer than that of Nd which may reflect a com-bination of matrix effects caused by imperfectly purifiedsolutions and sample heterogeneity Parent^daughterratios are available only from in situ analyses which arenot precise enough to allow meaningful model age orinitial isotope ratio calculations (ie at the time of kimber-lite eruption) Because the Lac de Gras kimberlites arerelatively young (55^56 Ma Graham et al 1999) age cor-rection would result in only a minor change in isotopecomposition which is inconsequential in view of the rangeof compositions displayed by most samples and hence hasno effect on the conclusions drawnClinopyroxenes in five volatile-free eclogites and a

pyroxenite have Sr isotope ratios ranging from 0701662 to0703546 this range is similar to or somewhat lower thanthat of the depleted mantle at the time of kimberliteemplacement (0703 Bell amp Tilton 2002) Clinopyroxenein the phlogopitized sample has a distinctly high 87Sr86Srof 0751186Clinopyroxene controls 56^92 of the Nd budget (Nd

concentration in cpx weighted by mode) Neodymiumisotope ratios range from 0511660 to 0514844 Thiscorresponds to eNd variation (per 10 000 deviationof 143Nd144Nd in the sample from 143Nd144Nd in the chon-dritic mantle model reservoir) from 17 to thorn43 cpx inthe low-Mg eclogites have higher eNd than those in

high-Mg eclogites Garnets have generally have higher143Nd144Nd from 0512984 (eNdfrac14thorn 7) to 0520965(eNdfrac14thorn 162) Garnets in the two volatile-rich sampleshave the highest ratios 0522790 for garnet in the calcite-bearing sample VR67112b (eNdfrac14thorn198) and 0525721 inthe phlogopitized eclogiteYK1911 (eNdfrac14thorn 255)Garnet contributes 18^32 and cpx 48^78 of the Hf

budget The proportion controlled by rutile is 4^30(Aulbach et al in preparation) Measured 176Hf177Hf incpx ranges from 0281263 in phlogopite-bearing sampleYK1911 (eHffrac1453) to 0286130 in low-Mg eclogiteVR43469 (eHffrac14thorn119) and in garnet from 0283958 inpyroxeniteYK1915 (eHffrac14thorn42) to 0295718 in low-Mg eclo-gite YK1926 (eHffrac14thorn458) Hafnium in garnet in the phlo-gopite^apatite-bearing sample is extremely radiogenic(176Hf177Hffrac14 0354367) corresponding to an eHf of thorn2532which is close to a value determined for a garnet separatefrom a Roberts Victor eclogite (eHffrac14thorn2561 Jacob et al2005) Garnet in the calcite-bearing sample has extremelyradiogenic Nd and Hf (eNdfrac14thorn193 eHffrac14thorn2516) Garnetin pyroxenite YK1915 also has radiogenic Nd and Hf(eNdfrac14thorn27 eHffrac14thorn42)Despite the large uncertainties for reconstructed whole-

rocks (calculated as the square root of the sum of squareduncertainties of two standard errors for garnet and cpxanalyses combined with 20 propagated uncertainty onthe proportion of Nd or Hf respectively contributed byeach mineral) the isotopic data are consistent in thatthey correlate with the independently determinedparent^daughter ratios (1) Low-Mg eclogites have higherSmNd reflected in the steeper positive slope of theirN-MORB-normalized REE patterns (Fig 11a) comparedwith the high-Mg eclogites (Fig 11b) Accordingly recon-structed low-Mg eclogites have higher present-day eNd

(0 to thorn47) than high-Mg eclogites (4 and 17) (2) Low-Mg eclogite YK1943 has very high Nb and LREE abun-dances that could have been added during metasomatismthat also enhanced its Hf abundance This sample has thelowest eNd and eHf of the low-Mg eclogites (0 and thorn20respectively) In contrast apparently unenriched low-Mgeclogites have more radiogenic Nd and Hf (eNdfrac14thorn47and thorn9 eHffrac14thorn128 and thorn113 respectively)The Sr^Nd isotope diagram reveals that eNd in cpx

varies from 19 to 43 at nearly constant depleted-mantle-like 87Sr86Sr (Fig 12a) In the Nd^Hf isotope diagram(Fig 12b) the reconstructed whole-rocks lie on curvedmixing lines if the lsquomixingrsquo proportions of Nd and Hffor the two lsquoend-memberrsquo minerals are very different(eg YK1911) Garnets have more radiogenic Nd andHf than coexisting cpx All reconstructed volatile-freeeclogites plot in the radiogenic field with regard to Hfor are isotopically similar to the depleted mantle Twohigh-Mg eclogites and one low-Mg eclogite have unradio-genic Nd one high-Mg eclogite has radiogenic Nd and

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1862

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 5 Individual and averaged Sr Nd and Hf isotope ratios obtained from garnet and cpx solutions

Sample 87Sr86Sr 2SE2SD 143Nd144Nd 2SE2SD eNd 2SE2SD 176Hf177Hf 2SE2SD eHf 2SE2SD

Low-Mg eclogites

VR43469 cp 0702661 0000010 0514844 0000006 43 01 0286130 0000012 119 04

VR43469 gt 0515508 0000014 56 03 0286506 0000012 132 04

VR43469 gt r 0515453 0000010 55 02 0286548 0000011 134 04

VR43469 gt r 0515384 0000007 54 01 0287244 0000010 158 04

VR43469 gt r 0515487 0000014 56 03 0286996 0000014 149 05

VR43469avg gt 0515458 0000109 55 21 0286823 0000715 143 253

WR VR43469 0515059 0000273 47 5 0286405 0001079 128 38

YK1926 cp 0703546 0000011 0513043 0000007 8 0 0283154 0000009 13 03

YK1926 gt 0513364 0000008 14 01 0295718 0000005 458 02

WR YK1926 0513120 0000026 9 1 0285980 0000716 113 25

YK1943 cp 0512143 0000007 10 01 0282213 0000005 20 02

YK1943 gt 0514461 0000012 36 02 0287510 0000008 168 03

YK1943 gt r 0514491 0000009 36 02 0287332 0000005 161 02

YK1943 gt r 0514517 0000010 37 02 0287290 0000007 160 02

YK1943avg gt 0514489 0000056 36 11 0287377 0000233 163 82

WR YK1943 0512625 0000228 0 5 0283138 0000879 13 31

High-Mg eclogites

VR43480 gt 0520965 0000017 162 03 0284950 0000010 77 04

VR67360 cp 0283509 0000074 26 26

VR67360 cp r 0283433 0000031 23 11

VR67360avg cp 0701662 0000018 0511660 0000010 19 02 0283471 0000108 25 38

VR67360 gt 0512984 0000101 7 20 0287429 0000180 165 64

WR VR67360 0511761 0000132 17 3 0284734 0000469 69 17

YK1949 cp 0703019 0000013 0512231 0000009 8 02 0282513 0000009 9 03

YK1949 gt 0513511 0000010 17 02 0290297 0000042 266 15

YK1949 gt r 0513508 0000012 17 02 0289947 0000015 254 05

YK1949 gt r 0513552 0000014 18 03

YK1949avg gt 0513523 0000049 17 10 0291569 0000494 311 175

WR YK1949 0512446 0000103 4 2 0284622 0000847 65 30

Volatile-rich eclogites

VR67112b gt 0522790 0000008 198 02 0354367 0000036 2532 13

YK1911 cp 0751186 0000009 0512165 0000007 9 01 0281263 0000009 53 03

YK1911 gt 0525721 0000013 255 03 0288051 0000012 187 04

WR YK1911 0513940 0000342 25 7 0281414 0000018 48 1

Pyroxenites

YK1914 cp 0703015 0000034 0513283 0000041 13 08

YK1914 cp r 0702956 0000005 0513200 0000027 11 05

YK1914 cp r 0513527 0000059 17 11

YK1914 cp r 0513045 0000010 8 02

YK1914avg cp 0702986 0000083 0513264 0000402 12 78

YK1915 gt 0514003 0000011 27 02 0283958 0000021 42

Analyses of full procedural replicates are designated with an lsquorrsquo after sample name SE standard error (internal errorsingle analyses) SD standard deviation (external error averaged multiple analyses) Italics used for reconstructed whole-rocks (mineral isotopic compositions and concentrations weighted by modes) Uncertainties for reconstructed whole-rocksinclude 2SE or 2SD uncertainties on each mineralrsquos isotope composition combined with 20 uncertainty in the amountof Nd or Hf contributed by the constituent minerals eNd and eHf are per 10 000 deviation from model chondritic mantle[model of Wasserburg et al (1981) and Blichert-Toft amp Albarede (1997) respectively]

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1863

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

εHf

εNd

Ekati eclogitesminus100

0

100

200

300

400

500

minus40 0 40 80 120 160 200 240 280

gt in YK1915

gt in VR67112bεNd=198 εHf=2532

minus40

0

40

80

120

160

minus40 minus20 0 20 40 60

minus60

minus40

minus20

0

20

40

070 071 072 073 074 075 076

North Americankimberlite field

εNd

87Sr86Sr

cpx in YK1914

Ekatieclogites

DM

DM

DM

DM

(a)

(b)

pyroxenites

low-Mg

high-Mg

volatile-rich

eclogites

Diavik low-TA154 S kimberlite

cpx in YK1911

Fig 12 (a) Sr^Nd for cpx and (b) Nd-Hf isotope diagram for garnet cpx and reconstructed whole-rocks (symbols with bold outline) Finedashed lines in (b) connect data points from the same sample with garnets having more radiogenic isotopic compositions Uncertainties forreconstructed whole-rocks combine 20 modal uncertainty with uncertainties on each mineralrsquos isotope composition Error bars on individualminerals may be smaller than the symbols Also shown are the isotopic compositions of the depleted mantle (DM DePaolo 1981 Griffin et al2000 Bell amp Tilton 2002) Range of 87Sr86Sr in eclogites from Ekati (Lac de Gras area 100 km north of Diavik Jacob et al 2003) in NorthAmerican kimberlites and compositions of low-temperature eclogites from Diavik and of Lac de Gras A154 South kimberlite fromSchmidberger et al (2007)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1864

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

one low-Mg eclogite has eNd similar to the depleted mantle(Fig 12b)Clinopyroxene in the phlogopitized eclogite YK1911

has unradiogenic Nd and Hf and strongly radiogenicSr whereas the coexisting garnet has radiogenic Nd andHf Although this sample contains 50 vol phlogopiteits whole-rock isotopic composition was calculated

from garnet and cpx alone because the contributionfrom phlogopite to the Hf and Nd budget is 52 and51 respectively (unpublished data) The whole-rockhas radiogenic Nd but has lower eHf than anyother sample analysed in accord with the extreme Hfabundance (53 ppm) and low LuHf of the cpx

DISCUSS IONOrigins of eclogites beneath the SlaveCratonHigh-Mg and high-Ca eclogites

With regard to their major- and trace-element contentsthe high-Mg and high-Ca eclogites from Lac de Gras arebroadly similar to variably evolved oceanic gabbros(Fig 13b and c) which may indicate an origin as subductedoceanic crust Compatible trace-element abundancesare useful to deduce potential crustal protoliths as theseelements are less susceptible to post-formation modification(Jacob 2004) The flat HREE patterns in the high-Mg andhigh-Ca eclogites are compatible with a low-pressureorigin as the oceanic crust is a product of moderate degreesof partial melting mainly within the spinel stability field(eg Presnall et al 2002) whereas melting with residualgarnet would qualitatively be expected to produce stronglyfractionated REE patterns Positive Sr and Pb anomaliesalthough compatible with a plagioclase signature couldalso be due to cryptic metasomatism (Dawson 1984) ofthe eclogites during their residence in the lithosphericmantle These anomalies are probably not related to kim-berlite infiltration as the whole-rock eclogite compositionswere reconstructed from analyses of fresh mineral coresLow ZrSm similar to that in primitive gabbros is

observed for many of the high-Mg and high-Caeclogites (Fig 13b and c) and is independent of the ratioof garnetcpx and whether or not the whole-rock wasreconstructed with rutile (Electronic Appendix 10) Thenegative trend of Mg-number against Yb in garnet(Fig 14) argues against igneous processes involving garnetaccumulation (Taylor amp Neal 1989) as suggested forexample for some high-Mg eclogites from West Africa(Barth et al 2002b) Rather the anti-correlation suggeststhat olivine fractionation played a role in the formation ofthe protoliths Fractionating olivine would excludeYb (Dolivinebasalt

frac14 002 Nikogosian amp Sobolev 1997)and incorporate Mg relative to Fe (Dfrac1415 and 05 respec-tively Taura et al1998) leading to progressive depletion ofMg and enrichment of Yb in the residual melt and henceanti-correlated Mg and Yb in the whole-rock This isinherited after eclogitization All eclogite types (exceptVR43452) plot along the olivine fractionation trend sug-gesting that they formed by a similar processHigh-Mg eclogites have lower

PREE than high-Ca

eclogites and some appear to have positive Eu anomalies

high-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

gabbrosSE Indian ridge

gabbrosOman ophiolite

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

high-Mg

RockNMORB(a)

VR43452

YK1949

low-Mg

boninite-like melts

01

001

1

10

(b)

01

001

1

10

(c)

01

001

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

Fig 13 Trace-element patterns of (a) low-Mg (b) high-Mg and(c) high-Ca eclogites compared with boninite-like volcanic rocksfrom the Isua greenstone belt (Polat et al 2002) gabbros from theOman ophiolite (Benoit et al 1996) and the SE Indian Ridge (Hartet al 1999) normalized to N-MORB of Sun amp McDonough (1989)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1865

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(EuEu up to 14) similar to more primitive cumulates(troctolite gabbros and olivine gabbros) reported in the lit-erature (Fig 13b) Although the sizes of these Eu anomaliesare not outside the analytical uncertainty their correlationwith CeYb in reconstructed eclogites supports their signif-icance (r2frac14066 nfrac14 4) The more primitive nature ofhigh-Mg eclogites is consistent with higher Mg-numbersrelative to high-Ca eclogites and may be due to olivineaccumulation during formation of their protolithsThe two samples with the lowest HREE abundanceshave the highest EuEu which is compatible with varia-tions in plagioclase modesMore evolved protoliths with higher pyroxeneplagio-

clase plus trapped melt could account for small or absentpositive Eu anomalies and higher

PHREE of high-Ca

eclogites (Fig 13c) Although the reason for the frequentassociation of graphite or diamond with high-Ca eclogitesis unclear (Aulbach et al 2003) the frequent occurrence ofkyanite in high-Ca eclogites also observed at Ekati justnorth of Lac de Gras has been argued to support a crustalprotolith (Jacob 2004) Kyanite andor corundum havebeen suggested to exsolve from high-temperature high-Alclinopyroxenes (Caporuscio amp Smyth 1990) Eclogitesfrom Lac de Gras show no exsolution textures suggestiveof such an origin although recrystallization can obscureor destroy microstructural evidence for exsolution(eg Griffin et al 1984)Modelling shows that qualitatively plagioclase accumu-

lation from a low-pressure melt derived from MORBmantle can account for positive Pb Sr and Eu and negativeZr anomalies relative to N-MORB whereas cpx accumula-tion or trapped melt increases Zr Y MREE and HREE

abundances (Fig 15) Some melt addition (for exampleby entrapment) seems to be required to increase all trace-element abundances to those observed for the eclogitesAlthough high-Ca and high-Mg eclogites could representshallow and deep portions of the same oceanic crust thehigher average temperatures obtained for the high-Caeclogites the frequent occurrence of carbon in high-Caeclogites and its absence in high-Mg eclogites the strongLREEHREE fractionation observed for six of seven high-Ca eclogites and the narrow range of compositions of thehigh-Mg eclogites suggest that they may have protolithsthat are not cogenetic Some discrepancies in particularaffecting the incompatible elements may be due to post-crystallization processes acting upon the protoliths suchas dehydration melting and later metasomatism that isevident in peridotites from the same locality (Griffin et al1999)Low K2OTh for high-Ca and high-Mg eclogites

(average 08 Table 4) relative to assumed crustal

0

2

4

6

8

10

30 40 50 60 70 80

Mg

VR43452

Yb

GARNET

pyroxenites

low-Mghigh-Mghigh-Ca

eclogites

VR19673ecl-g2

Fig 14 Garnet Mg-number vs Yb (ppm including 2s uncertainty)Lines qualitatively show separate correlations for eclogites (exceptoutlierVR43452) and pyroxenites (except outlierVR19673ecl-g2)

high-Ca

high-Mg

001

01

1

10

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

melt08 melt02 cryst cpx01 melt09 cryst

plag

(a)

001

01

1

10(b)

RockNMORB

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 15 Reconstructed (a) high-Ca and (b) high-Mg eclogitescompared with compositions of plagioclase and cpx crystallizingfrom a MORB melt modelled using distribution coefficients ofNorman et al (2005)The MORB was calculated as a 10 equilibriummelt of MORB pyrolite [composition from Kelemen et al (1993)except Pb Pr Sr Y and Lu which were calculated as residues from1 fractional melting of pyrolite from McDonough amp Sun (1995)using melt modes of Kinzler (1997) and distribution coefficients ofNorman et al (2005)] Also shown are mixtures of crystals (consistingof 50 plagioclase plus 50 cpx) and melt in the proportions 0901and 0208

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1866

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

precursors (eg gabbros SE Indian Ridge average 22Hart et al 1999) may reflect the preferred loss of fluid-mobile elements during dehydration (Becker et al 2000)although large uncertainties on these low-abundance ele-ments in the constituent minerals and reconstructedwhole-rocks makes this interpretation less than robustThe REE patterns of most high-Mg eclogites are too flatto represent melting residues as partial melting shouldproduce low LREEHREE (Electronic Appendix 10) Incontrast the marked LREE depletion of five of sevenhigh-Ca eclogites (LaLufrac14 02^08) contrasts with that ofassumed evolved gabbroic protoliths (LaLufrac1415^32Hart et al 1999) and can be modelled as a consequence ofpartial melting during subduction after eclogitization (iein the presence of garnet which retains HREE relative toLREE eg Barth et al 2002a) has taken place (Aulbachet al 2003)Our results are compatible with previous studies on

eclogite xenoliths from the Slave Craton includingthose from the same kimberlite that have linked thesesamples or some of these samples to Paleoproterozoicsubduction events (the 195^191Ga formation of theHottah terrane and the 18Ga formation of the GreatBear magmatic arc) in the Slave craton (Griffin et al1999 Kopylova et al 1999 Heaman et al 2002 2006Schmidberger et al 2005 2007)The compositional features discussed above although

compatible with crustal protoliths are not necessarilydefinitive of a subduction origin However if direct crystal-lization from basaltic melts at high pressure (ie mantledepths) is invoked instead specific conditions would haveto be met to satisfy the observations The Mg-number ineclogites from Lac de Gras are lower than those expectedof primary high-pressure mantle melts (eg Walter 1998OrsquoHara amp Herzberg 2002) requiring crystallization athigh pressures after olivine fractionation This originwould further require shielding from peridotitic wallrockwith which the fractionated melts would no longer be inequilibrium (Barth et al 2001) In such a setting the lowrelative abundances of fluid-mobile elements comparedwith those of less fluid-mobile elements and of LREErelative to HREE (Table 4) would reflect an expulsion offluids during crystallization of the protoliths and latermelt mobilization respectively The eclogites clearly donot represent garnet plus cpx cumulates which wouldrequire a positive correlation of Mg-number with Yb thatis at variance with the observed negative correlation ineclogitic garnets from this study (Fig 14)

Low-Mg eclogites

Garnets in low-Mg eclogites compositionally overlap withthose in massive eclogites from Jericho (NW of the Lacde Gras kimberlites) that were suggested to have a high-pressure intrusive origin based on the similarity of theirmajor-element compositions to those of Group A eclogites

fromsouthernAfrica forwhichamantleoriginhasbeenpro-posed (Kopylova et al1999) Garnets in low-Mg eclogites inthe present study have major-element compositions moresimilar to garnets in groupB rather thangroupAeclogitesThe positive slope in N-MORB-normalized HREE

abundances displayed by the low-Mg eclogites is clearlydistinct from other eclogite types and is similar to those ofboninite-like second-stage melts that are closely associatedwith ultramafic volcanics (Smithies et al 2004) (Fig 13a)The range of Mg-numbers in the low-Mg eclogites (55^66)overlaps with that of Archaean second-stage melts fromIsua (58^75) and Abitibi (52^82 Smithies et al 2004)These similarities might indicate that the low-Mg eclogiteshave an origin in a previously depleted mantle source suchas supra-subduction zone mantle Subchondritic ZrHf ofthe low-Mg eclogites contrasts with the supra-chondriticZrHf of the high-Mg and high-Ca eclogites (Aulbach et

al in preparation) and is consistent with previous cpx-controlled depletion in an arc mantle source This sourcecould have been subsequently remelted to generate theboninite-like precursor to the low-Mg eclogites as sug-gested for the protoliths of eclogites from Udachnaya(Jacob amp Foley 1999)Although some second-stage melts to which the low-Mg

eclogites are similar are closely associated with komatiitesand could contain a plume component the radiogenic ordepleted mantle-like Nd and Hf and unradiogenic Sr inthe low-Mg eclogites do not support plume involvement

Trace-element and Sr^Nd^Hf isotopeconstraints on post-formationmetasomatismSeveral low-Mg and high-Ca eclogites show negativeslopes in the LREE which may be indicative of interactionwith small amounts of LREE-enriched melt (Navon ampStolper 1987) amounting to cryptic metasomatism(ie not accompanied by crystallization of new phases) inthe nomenclature of Dawson (1984) High Nb abundancesrelative to elements of similar compatibility displayed bysome low-Mg and high-Ca eclogites (Fig 10) could resultfrom overestimation of rutile modes However high-CaeclogiteVR40345 is rutile-free and has been reconstructedwithout rutile its high Nb abundance is accompanied byLREE enrichment indicating that metasomatic additionmay be responsibleLow-Mg eclogites show a positive correlation of La with

Nb (Electronic Appendix 11a) indicating concomitantaddition of LREE and Nb This suggests that the metaso-matic agent was a solute-rich fluid or silicate silicic melt ashydrous fluids and carbonatites do not carry enough HFSEto affect the budget of eclogites (Green amp Wallace 1988Klemme et al 1995 Adam et al 1997 Stalder et al 1998)Of the samples that are enriched in Nb all but one (high-Ca eclogiteVR50909) show a simultaneous enrichment in

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1867

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Hf (Electronic Appendix 11b) Together these trace-element relationships suggest that both SmNd and LuHfwere lowered in these samples Their low parent^daughterratios contrast with higher ratios in unenriched samplesand resulted in the ingrowth of a range of Nd and Hf iso-tope compositions Compared with the low-Mg and high-Ca eclogites the high-Mg eclogites have a narrow range ofLa and Nb concentrations which correlates with a narrowrange in Nd and Hf isotope compositions (Fig 12)Precise model ages cannot be calculated because parent^

daughter ratios are available only from in situ analyseswhich have relatively large uncertainties Howeverevidence for long-term enrichment and depletion may bediscerned Low-Mg eclogites even those that are enrichedin LREE have relatively high SmNd as a result of havingoverall positive slopes in their REE patterns relative toMORB whereas the high-Mg eclogites have relativelyflat REE patterns and therefore lower SmNd (Fig 11)This translates into lower eNd for high-Mg eclogites thanfor low-Mg eclogites In contrast they show no systematicdifference with respect to 176Hf177Hf Because the low- andhigh-Mg eclogites are proposed to have different origins asdiscussed above and therefore did not necessarily have thesame initial isotopic composition and formation age theseisotopic differences cannot be used to estimate their agesA long-term coupling of trace-element and isotopic

abundances is also indicated by the fact that low-Mg eclo-gite YK1943 which has experienced Nb and LREE addi-tion has lower eNd (0) and eHf (thorn13) than low-Mg eclogiteVR43469 with higher LuHf and SmNd (eNdfrac14thorn47eHffrac14thorn128) If low-Mg eclogites YK1943 and VR43469originally had identical initial Nd and Hf isotope composi-tions and if the parent^daughter ratios are accurate andwere modified only once during their evolution shortlyafter metamorphism (when their isotopic compositionwould have been reset) this overprint can be calculatedfor both isotopic systems to have occurred at 17GaThis would suggest that Hf and Nd were added duringthe same event In contrast high-Mg eclogites VR67360and YK1949 would have had identical Ndisotope compositions at 10Ga whereas their Hf isotopecompositions are disturbed in that the sample with thehigher parent^daughter ratio has a lower 176Hf177Hf thanthe sample with the lower parent^daughter ratio Theseresults suggest that the LREE and HFSE enrichmentobserved in different samples occurred during separateevents that did not affect the eclogites equallyExtreme HFSE enrichment leading to the growth of

zircon and rutile and LREE addition leading to thegrowth of apatite in eclogites from the Jericho kimberlitein the northern Slave craton have been ascribed to twoseparate metasomatic events at around 18Ga (quasi-coincident with the age of formation and metamorphism)and 13^10Ga respectively (Heaman et al 2002 2006

Schmidberger et al 2005) The authors associated theseevents with Paleoproterozoic subduction and aMesoproterozoic thermal disturbance in the Slave litho-spheric mantle possibly linked to the Mackenzie igneousevent Mesoproterozoic Nd model ages mostly between13 and 14Ga were reported for eclogites from thenearby Ekati kimberlites (Jacob et al 2003) Although wedo not observe zircon or apatite in volatile-free eclogitesfrom Lac de Gras and the age constraints obtained fromour data are weak they support a link between theseevents described for other eclogite populations and themetasomatic enrichment observed in our studyThe Sr isotope data in the present study are similar to

those of Schmidberger et al (2007) who suggested that thedepleted-mantle-like values are consistent with derivationof the protoliths from an oceanic mantle source However87Sr86Sr in volatile-free eclogites in the present study isremarkably constant considering their Nd and Hf isotopevariability (Fig 12) and the variability of trace-elementpatterns (Fig 10) Given the evidence for dehydration andmelting in many of the eclogites fractionation of highlyincompatible Rb from Sr and differential ingrowth ofradiogenic Sr would be expectedThis may indicate a rela-tively recent overprint and considering the similarityof 87Sr86Sr to present-day depleted mantle which is signif-icantly lower than 87Sr86Sr of North American kimber-lites a juvenile (asthenospheric) source

PyroxenitesPyroxenites from Lac de Gras often show relationshipsin their major-element contents that are different fromthose observed in the eclogites For example pyroxeniticgarnets show a negative correlation between CaO andNa2O contents whereas these oxides are positively corre-lated for garnets in eclogites (Fig 5) CaO is well correlatedwith C2O3 in pyroxenitic garnets but not in eclogitic gar-nets (Fig 4) Mg-number is positively correlated withYb abundance for pyroxenitic garnets but anti-correlatedfor eclogitic garnets (Fig 14) This suggests that the pyrox-enites in the present study are not related to the eclogitesby fractionation or accumulation processes as proposedfor the origin of some pyroxenites (Foley et al 2003)Pyroxenites have also been interpreted as the products

of crystal segregation in magma channels with alteredoceanic crust as the magma source this would explaintheir Eu anomalies and stable and radiogenic isotopecompositions (Pearson et al 1993 Becker 1996) As thehigh-pressure liquidus phases of most terrestrial basaltsare cpx and garnet (OrsquoHara amp Herzberg 2002) partialmelts of eclogites with basaltic composition wouldcrystallize cpx and garnet but not opx More than half ofthe pyroxenites from Lac de Gras contain opx suggestingthat they probably did not crystallize directly fromeclogite-derived melts

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1868

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Low-Cr websterites and pyroxenites could be crystalli-zation products of mafic melts that interacted with perido-titic mantle thus becoming more refractory The positivecorrelation of Yb and Mg-number in garnet would allowan origin by fractional crystallization from such meltsAlternatively they formed by reaction of silicic (possiblyeclogite-derived) melts with mantle peridotites(eg Kelemen et al 1993 Aulbach et al 2002) This agreeswith the similarity of hypothetical melts in equilibriumwith pyroxenitic garnets in most samples to eclogite-derived melts (Electronic Appendix 12) Niobium concen-trations in all hypothetical melts are much higher than ineclogite-derived melts regardless of whether or not rutileis assumed to be in the residue The same is true for LaCe and Zr concentrations in some samples which mayindicate that NbZr and LREE were added to thepyroxenites some time after their formation as discussedin the previous section for some eclogites Because mini-mum ages for pyroxenitic rutile with an unradiogenicHf isotope composition (calculated by projecting the176Hf177Hf to the model depleted mantle isotopic evolutioncurve assuming zero ingrowth) range to 178066Ga(Aulbach et al in preparation) it appears likely that thepyroxenites were not formed recently (eg during kimber-lite-related or precursory melt migration) and that some ofthe samples are at least MesoproterozoicGarnet pyroxeniteYK1952 consists of schlieren of equili-

brated garnet pyroxenite in a lsquomatrixrsquo of spongy cpxfine cpx^opx intergrowths embayed garnet and meltpatches (Fig 2f) Garnet in the pyroxenite schlieren hasvery low LREE and HFSE abundances Orthopyroxenein the matrix has distinctly higher Mg-number andCr-number and contains more CaO than opx from thepyroxenite schlieren whereas opx rims in the schlierenare compositionally similar to opx in the matrix The con-siderable compositional and textural disequilibrium inthis sample suggests that this is a young feature causedby reaction with fluids shortly before or at the time ofentrainmentAs a consequence of the small sample sizes of most of

the pyroxenites studied here a comprehensive isotopicdataset is not available Clearly more such data areneeded in addition to stable isotope data to obtain betterconstraints on the origins of this rock type Consideringthe microstructural and compositional variability ofthe pyroxenites and major-element relationships oppositeto those observed for eclogites it appears certain that pyr-oxenites form by a variety of processes some of which areunrelated to the formation of eclogites and their protoliths

Volatile-rich eclogitesVolatile-rich eclogites from Lac de Gras display the mostextreme Hf^Nd isotope systematics of all the analysedsamples Garnets in phlogopitized (YK1911) and in cal-cite-bearing eclogite (VR67112b) have the most radiogenic

Nd (eNdfrac14 2486 and 1927 respectively) Garnet inVR67112b has extremely high eHf (2516) Clinopyroxene inYK1911 has the lowest eHf measured for any mineralseparate from Lac de Gras (53) in accord with the veryhigh Hf abundance in its cpx (5 ppm) In contrastgarnet in this sample has radiogenic Hf (eHffrac14 136) Thisdiscrepancy is at least in part due to their relatively lowequilibration temperatures (7698C and 8558C respec-tively) which allowed the isotopic compositions of cpxand garnet to evolve separately at their low andhigh parent^daughter element ratios respectively forextended periods of time and possibly since formationAccordingly the cpx in YK1911 gives a minimum Hf ageof 27GaSlab-derived fluids could have been introduced during

the 27Ga collision of the western Slave Craton domainincluding the thick plume-derived deep lithosphericmantle layer and the eastern domain with its highlydepleted lithosphere This would also be consistent withthe observation that volatile-rich eclogites occur at shallowdepths in the mantle (TKroghfrac14 685^8558C correspondingto depths of 85^115 km along a 40 mWm2 geotherm)where the bulk of dehydration and production of solute-rich melts in a subducting slab might be expected to occur(Becker et al 2000 Hermann et al 2006) Their negativeEu anomalies might indicate that the fluid was derivedfrom the shallow portion of the slab that experiencedplagioclase fractionation The presence of carbonate inVR67112b requires a carbonated sourceHigh 87Sr86Sr in cpx in YK1911 is probably the conse-

quence of isotopic equilibration with high-RbSr phlogo-pite after long-term radiogenic ingrowth while Hfaddition led to retarded ingrowth of radiogenic HfThese systematics combined with radiogenic Nd appear

inconsistent with the involvement of sediments in the pet-rogenesis of this sample (eg Perfit et al 1980 Vervoortet al 1999) or a hydrous fluid such as those responsible forMARID metasomatism in the Kaapvaal craton (Konzettet al 1998)Whatever the origin of the fluids the apparentHf mobility suggests that they were not in equilibriumwith a rutile-bearing source and also that they weresolute-rich in nature (ie hydrous melts able to dissolveHFSE) and the relatively high HREE abundances of thevolatile-rich eclogites indicate a garnet-free source suchas amphibolite This allowed evolution of the sample atrelatively high time-integrated SmNd

SUMMARY AND CONCLUSIONS

(1) The Lac de Gras kimberlites carried abundant eclo-gite and pyroxenite xenoliths to the surface Threemain eclogite types are distinguished in terms ofmajor-element composition and trace element

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1869

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

patterns high-Mg high-Ca and low-Mg eclogitesThe garnets of all eclogite types have anti-correlatedMg-number and Yb pointing to olivine fractionationas a controlling factor in the formation of their pre-cursor rocks Their flat or positively sloped HREEpatterns qualitatively indicate that the melt parentalto the precursor rocks was not in equilibrium with agarnet-bearing residue

(2) Reconstructed high-Mg eclogites and high-Caeclogites have flat HREE patterns positive Pb andSr anomalies and low ZrSm similar to oceanic gab-bros reported in the literature High-Mg eclogiteshave variably low HREE contents and some havesmall positive Eu anomalies possibly suggestive ofmore primitive relatively plagioclase-rich protolithswhereas high-Ca eclogites have higher HREE abun-dances and small or absent Eu anomalies that requiremore cpx- andor melt-rich protoliths if a subductionorigin is assumed

(3) Low-Mg eclogites have distinctive trace-element pat-terns that show a marked positive slope in the HREEand subchondritic ZrHf They resemble someArchaean second-stage melts (lsquoboninitesrsquo) and theirprotoliths appear to have formed from a previouslydepleted source

(4) After formation the eclogites were subject to a varietyof processes Most have low ratios of fluid-mobile vsless fluid-mobile elements possibly as a result of dehy-dration of the protoliths In addition some of the low-Mg and high-Ca eclogites have low LREEHREEconsistent with partial melt lossWhereas the low-Mgeclogites show concomitant enrichment in La Nb andZr some high-Mg and high-Ca eclogites show evi-dence for variable and unsystematic (decoupled)LREE and Nb addition

(5) Isotopically the trace-element characteristics of thedifferent eclogite types translate into lower eNd forhigh-Mg eclogites (negative slope in MREEfrac14 lowerSmNd) than for low-Mg eclogites (positive slope inMREEfrac14higher SmNd) Within the low-Mg groupsamples that show evidence for metasomatism (lowLuHf and SmNd) have lower eNd and eHf than asample that was apparently not metasomatized BothNd and Hf addition can be dated to 17Ga if identi-cal initial isotope compsitions are assumed for thelow-Mg eclogites In contrast the relatively constantand depleted mantle-like 87Sr86Sr of eclogitic cpxdespite variable Nd and Hf isotope compositions andparent^daughter ratios suggests recent input of juve-nile material

(6) Garnet and cpx in pyroxenites show major-elementrelationships different from those in eclogitesTherefore these two rock types are probably notrelated by fractionation processes The presence of

opx precludes direct crystallization from slab-derivedmelts Pyroxenites may be the product of (slab-derived) melts hybridized during assimilation of peri-dotitic mantle

(7) Garnets in phlogopite- and carbonate-bearing eclo-gites have extreme isotopic compositions includingthe most radiogenic Nd and Hf in the datasetNegative Eu anomalies may point to the involvementof a crustal protolith Highly radiogenic Sr radiogenicNd and unradiogenic Hf in the reconstructedphlogopite-bearing sample combined with highHREE abundances and evidence for Hf mobilitymay indicate involvement of solute-rich fluids origi-nating in a garnet- and rutile-free source

ACKNOWLEDGEMENTSHelp in the clean laboratories and with the analyticalfacilities by S Elhlou C Lawson A Sharma P Wielandand M Zhang is greatly appreciated Bill Griffin isthanked for his support and for agreeing to reviewseveral versions of the manuscript despite disagreeingwith many of the conclusions presented here The paperbenefited greatly from the detailed and constructivereviews provided by the editors Gareth Davies andMarjorie Wilson and by Matthias Barth Larry Heamanand an anonymous reviewer This work was funded by aMacquarie University International PostgraduateAward and Postgraduate Research Fund (SA) by anARC SPIRT grant sponsored by Kennecott Canada Incand by ARC Large and Discovery Grants to SYOrsquoRAnalytical data were obtained using instrumentationfunded by ARC LIEF and DEST Systemic InfrastructureGrants industry partners and Macquarie UniversityThis is Publication 481 from the ARC National KeyCentre for Geochemical Evolution and Metallogeny ofContinents (wwwesmqeduauGEMOC)

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online

REFERENCESAdam J GreenT H Sie S H amp Ryan C G (1997) Trace element

partitioning between aqueous fluids silicate melts and mineralsEuropeanJournal of Mineralogy 9 569^584

Aulbach S Stachel T Viljoen K S Brey G P amp Harris J W(2002) Eclogitic and websteritic diamond sources beneath theLimpopo Beltccedilis slab-melting the link Contributions to Mineralogy

and Petrology 143 56^70Aulbach S GriffinW L Pearson N J OrsquoReilly S Y Kivi K amp

Doyle B (2003) Origins of eclogites beneath the central SlaveCraton Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA-0011

Aulbach S Griffin W L Pearson N J OrsquoReilly S Y amp Kivi K(2004a) Mantle formation and evolution Slave Craton constraints

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1870

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

from HSE abundances and Re^Os isotope systematics of sulfideinclusions in mantle xenocrysts Chemical Geology 208 61^88

Aulbach S Griffin W L OrsquoReilly S Y amp McCandless T E(2004b) Genesis and evolution of the lithospheric mantle beneaththe Buffalo HeadTerrane Alberta (Canada) Lithos 77 413^451

Aulbach S GriffinW L Pearson N J OrsquoReilly S Y amp Doyle B(2007) Lithosphere formation in the central Slave Craton(Canada) plume subcretion or lithosphere accretion Contributionsto Mineralogy and Petrology (in press) doi 101007s00410-007-0200-1

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2001)Geochemistry of xenolithic eclogites from West Africa Part IA link between low MgO eclogites and Archean crust formationGeochimica et Cosmochimica Acta 65 1499^1527

Barth M G Foley S F amp Horn I (2002a) Partial melting inArchean subduction zones constraints from experimentally deter-mined trace element partition coefficients between eclogitic miner-als and tonalitic melts under upper mantle conditions PrecambrianResearch 113 323^340

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2002b)Geochemistry of xenolithic eclogites from West Africa part 2Origins of the high MgO eclogites Geochimica et Cosmochimica Acta

66 4325^4345Beard B L Fraracci K NTaylor L A Snyder G A Clayton R

A Mayeda T K amp Sobolev NV (1996) Petrography andgeochemistry of eclogites from the Mir kimberliteYakutia RussiaContributions to Mineralogy and Petrology 125 293^310

Becker H (1996) Geochemistry of garnet peridotite massifs fromlower Austria and the composition of deep lithosphere beneatha Palaeozoic convergent plate margin Chemical Geology 134 49^65

Becker H Jochum K P amp Carlson R W (2000) Trace elementfractionation during dehydration of eclogites from high-pressureterranes and the implications for element fluxes in subductionzones Chemical Geology 163 65^99

Bell K amp Tilton G R (2002) Probing the mantle The story fromcarbonatites EOS Transactions American Geophysical Union 83275^277

Benoit M Polvecurren M amp Ceuleneer G (1996) Trace element andisotopic characterisation of mafic cumulates in a fossil mantlediapir (Oman ophiolite) Chemical Geology 134 199^214

BleekerW (2003) The late Archean record a puzzle in ca 35 piecesLithos 71 99^134

BleekerW Ketchum JW F JacksonV A amp Villeneuve M (1999)The central Slave Basement Complex Part I Its structural topol-ogy and autochthonous core Canadian Journal of Earth Sciences 361083^1109

Blichert-Toft J amp Albare de F (1997) The Lu^Hf isotope geochemis-try of chondrites and the evolution of the mantle^crust systemEarth and Planetary Science Letters 148 243^258

Blichert-Toft J Chauvel C amp Albare de F (1997) Separation of Hfand Lu for high precision isotope analysis of rock samples by mag-netic sector-multiple collector ICP-MS Contributions to Mineralogy

and Petrology 127 248^260Brey G P amp Kolaquo hler T (1990) Geothermobarometry in four-phase

lherzolites II New thermobarometers and practical assessment ofexisting thermobarometers Journal of Petrology 31 1353^1378

Brey G P Kolaquo hlerT amp Nickel K G (1990) Geothermobarometry infour-phase lherzolites I Experimental results from 10 to 60 kbJournal of Petrology 31 1313^1352

Caporuscio F amp Smyth J (1990) Trace element crystal chemistryof mantle eclogites Contributions to Mineralogy and Petrology 105550^561

Carbno G B amp Canil D (2002) Mantle structure beneath the SWSlave Craton Canada constraints from garnet geochemistry in theDrybones Bay Kimberlite Journal of Petrology 43 129^142

Coleman R G Lee E D Beatty L B amp Brannock W W (1965)Eclogites and eclogites their differences and similaritiesGeological Society of America Bulletin 76 483^508

Creaser R A Grulaquo tter H Carlson J amp Crawford B (2004)Macrocrystal phlogopite Rb^Sr dates for the Ekati propertykimberlites Slave Province Canada evidence for multiple intru-sive episodes in the Paleocene and Eocene Lithos 76 399^414

Davies R M GriffinW L Pearson N J Andrew A S Doyle BJ amp OrsquoReilly S Y (1999) Diamonds from the deep Pipe DO-27Slave Craton Canada In Gurney J J Gurney J L Pascoe MD amp Richardson S H (eds) Proceedings of the 7th International

Kimberlite Conference CapeTown Red Roof Design pp 148^155Davies R M Griffin W L OrsquoReilly S Y amp Doyle B J (2004)

Mineral inclusions and geochemical characteristics of microdia-monds from the DO27 A154 A21 A418 DO18 DD17 and RanchLake kimberlites at Lac de Gras Slave Craton Canada Lithos 7739^55

DavisW J amp Hegner E (1992) Neodymium isotopic evidence for thetectonic assembly of Late Archean crust in the Slave Provincenorthwest Canada Contributions to Mineralogy and Petrology 111493^504

Davis W J Jones A G Bleeker W amp Grutter H (2003)Lithosphere development in the Slave craton a linked crustal andmantle perspective Lithos 71 575^589

Dawson J B (1984) Contrasting types of upper-mantle metasoma-tism In Kornprobst J (ed) Kimberlites II The Mantle and

Crust^Mantle Relationships Amsterdam Elsevier pp 289^294DePaolo D J (1981) Neodymium isotopes in the Colorado Front

Range and crust^mantle evolution in the Proterozoic Nature 291193^196

Erlank A J amp Kushiro I (1970) Potassium contents of syntheticpyroxenes at high temperatures and pressures Carnegie Institution ofWashingtonYearbook 68 433^439

Foley S F Barth M G amp Jenner G A (2000) Rutilemelt partitioncoefficients for trace elements and an assessment of the influence ofrutile on the trace element characteristics of subduction zonemagmas Geochimica et Cosmochimica Acta 64 933^938

Foley S F Buhre S amp Jacob D E (2003) Evolution of the Archaeancrust by delamination and shallow subduction Nature 421 249^252

Graham I Burgess J L Bryan D Ravenscroft P J Thomas EDoyle B J Hopkins R amp Armstrong K A (1999) Explorationhistory and geology of the Diavik Kimberlites Lac de GrasNorthwest Territories Canada In Gurney J J Gurney J LPascoe M D amp Richardson S H (eds) Proceedings of the 7th

International Kimberlite Conference Cape Town Red Roof Designpp 262^279

Green D H amp Wallace M E (1988) Mantle metasomatism byephemeral carbonatite melts Nature 336 459^462

Green T H Blundy J D Adam J amp Yaxley G M (2000) SIMSdetermination of trace element partition coefficients betweengarnet clinopyroxene and hydrous basaltic liquids at 2^75 GPaand 1080^12008C Lithos 53 165^187

GriffinW L amp OrsquoReilly SY (2007) Cratonic lithospheric mantle Isanything subducted Episodes 30(1) 43^53

Griffin W L Wass S Y amp Hollis J D (1984) Ultramafic xenolithsfrom Bullenmerri and Gnotuk maars Victoria Australia petrol-ogy of a sub-continental crust^mantle transition Journal of

Petrology 25 53^87GriffinW L Doyle B J Ryan C G Pearson N J OrsquoReilly SY

Davies R Kivi K van Achterbergh E amp Natapov LM (1999)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1871

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Layered mantle lithosphere in the Lac de Gras Area SlaveCraton composition structure and origin Journal of Petrology 40705^727

Griffin W L Pearson N J Belousova E Jackson S E vanAchterbergh E OrsquoReilly SY amp Shee S R (2000)The Hf isotopecomposition of cratonic mantle LAM-MC-ICPMS analysis ofzircon megacrysts in kimberlites Geochimica et Cosmochimica Acta

64 133^147Griffin W L OrsquoReilly S Y Doyle B J Pearson N J

Coopersmith H Kivi K Malkovets V amp Pokhilenko N(2004) Lithosphere mapping beneath the North American plateLithos 77 873^922

Grulaquo tter H S Apter D B amp Kong J (1999) Crust^mantle couplingevidence from mantle-derived xenocrystic garnets In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 307^312

Hart S R Blusztajn J Dick H J B Meyer P S ampMuehlenbachs K (1999) The fingerprint of seawater circulationin a 500-meter section of ocean crust gabbros Geochimica et

Cosmochimica Acta 63 4059^4080Harte B amp Kirkley M B (1997) Partitioning of trace elements

between clinopyroxene and garnet Data from mantle eclogitesChemical Geology 136 1^24

Heaman L M Creaser R A amp Cookenboo H O (2002) Extremeenrichment of high field strength elements in Jericho eclogitexenoliths A cryptic record of Paleoproterozoic subduction partialmelting and metasomatism beneath the Slave craton CanadaGeology 30 507^510

Heaman L M Kjarsgaard B A amp Creaser R A (2004) The tem-poral evolution of North American kimberlites Lithos 76 377^397

Heaman L M Creaser R A Cookenboo H O amp Chacko T(2006) Multi-stage modification of the northern Slave mantlelithosphere Evidence from zircon- and diamond-bearing eclogitexenoliths entrained in Jericho Kimberlite Journal of Petrology 47821^858

Hermann J Cpandler C Hack A amp Korsakov A (2006) Aqueousfluids and hydrous melts in high-pressure and ultra-high pressurerocks implications for element transfer in subduction zones Lithos92 399^417

Hills DV amp Haggerty S E (1989) Petrochemistry of eclogites fromthe Koidu Kimberlite Complex Sierra Leone Contributions to

Mineralogy and Petrology 103 397^422Hoffman P F (1989) Precambrian geology and tectonic history of

North America In Bally A amp Palmer A (eds) The Geology of

North AmericaccedilAn Overview Boulder CO Geological Society ofAmerica pp 447^512

IrelandT R Rudnick R L amp Spetsius Z (1994) Trace-elements indiamond inclusions from eclogites reveal link to Archean granitesEarth and Planetary Science Letters 128 199^213

Irvine G J Pearson D G Kjarsgaard B A Carlson R WKopylova M G amp Dreibus G (2003) Evolution of the litho-spheric mantle beneath Northern Canada a Re^Os isotope andPGE study of kimberlite-derived peridotite xenoliths fromSomerset Island and a comparison to the Slave and Kaapvaalcratons Lithos 71 461^488

Isachsen C E amp Bowring S A (1994) Evolution of the Slave cratonGeology 22 917^920

Jacob D E (2004) Nature and origin of eclogite xenoliths from kim-berlites Lithos 77 295^316

Jacob D E amp Foley S F (1999) Evidence for Archean ocean crustwith low high field strength element signature from diamondifer-ous eclogite xenoliths Lithos 48 317^336

Jacob D Jagoutz E Lowry D Mattey D amp Kudrjavtseva G(1994) Diamondiferous eclogites from Siberiaccedilremnants ofArchean oceanic-crust Geochimica et Cosmochimica Acta 585191^5207

Jacob D E Fung A Jagoutz E amp Pearson D G (2003) Petrologyand geochemistry of eclogite xenoliths from the Ekati kimberlitesarea Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA_239

Jacob D E Bizimis M amp Salters V J M (2005) Lu^Hf and geo-chemical systematics of recycled ancient oceanic crust evidencefrom Roberts Victor eclogites Contributions to Mineralogy and

Petrology 148 707^720Jerde E A Taylor L A Crozaz G Sobolev NV amp SobolevV N

(1993) Diamondiferous eclogites from Yakutia Siberiaccedilevidencefor a diversity of protoliths Contributions to Mineralogy and Petrology

114 189^202Jones A G Ferguson I J Chave A D Evans R L amp

McNeice GW (2001) The electric lithosphere of the Slave cratonGeology 29 423^426

Kelemen P B Shimizu N amp Dunn T (1993) Relative depletionof niobium in some arc magmas and the continental-crust-partitioning of K Nb La and Ce during meltrockreaction in the upper-mantle Earth and Planetary Science Letters 120111^134

Kinzler R J (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition Application to mid-ocean ridge basalt petrogenesis Journal of Geophysical Research 102853^874

Klemme S van der Laan S R Foley S F amp Guenther D (1995)Experimentally determined trace and minor element partitioningbetween clinopyroxene and carbonatite melt under upper mantleconditions Earth and Planetary Science Letters 133 439^448

Konzett J Armstrong R A Sweeney R J amp CompstonW (1998)The timing of MARID metasomatism in the Kaapvaal mantle anion probe study of zircons from MARID xenoliths Earth and

Planetary Science Letters 160 133^145Kopylova M G Russell J K amp Cookenboo H (1999) Mapping the

lithopshere beneath the north central Slave Craton In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 468^479

Krogh E (1988) The garnet^clinopyroxene iron^magnesiumgeothermometerccedila reinterpretation of existing experimentaldata Contributions to Mineralogy and Petrology 99 44^48

Kusky T M (1989) Accretion of Archean Slave Province Geology 1763^67

LeCheminant A N amp Heaman L M (1989) Mackenzie igneousevents Canada Middle Proterozoic hotspot magmatismassociated with ocean opening Earth and Planetary Science Letters

96 38^48LeCheminant A N Richardson D G DiLabio R N W amp

Richardson K A (1996) La recherche de diamants au CanadaGeological Survey of Canada Open File 3228 268pp

MacGregor I D amp Carter J L (1970) The chemistry of clinopyrox-enes and garnets of eclogite and peridotite xenoliths from theRoberts Victor mine South Africa Physics of the Earth and PlanetaryInteriors 3 391^397

MacGregor I D amp Manton W I (1986) Roberts Victor eclogitesancient oceanic crust Journal of Geophysical Research 91 14063^14079

MacKenzie J M amp Canil D (1999) Composition and thermal evo-lution of cratonic mantle beneath the central Archean SlaveProvince NWT Canada Contributions to Mineralogy and Petrology

134 313^324

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1872

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

McCandlessT E amp Gurney J J (1989) Sodium in garnet and potas-sum in clinopyroxene criteria for classifying mantle eclogites InRoss J (ed) Kimberlites and Related Rocks Carlton Blackwellpp 827^832

McDonoughW F amp Sun S-S (1995) The composition of the EarthChemical Geology 120 223^253

Navon O amp Stolper E (1987) Geochemical consequences of meltpercolation the upper mantle as a chromatographic columnJournal of Geology 95 285^307

Nikogosian I K amp Sobolev AV (1997) Ion-microprobe analysis ofmelt inclusions in olivine experience in estimating the olivine^meltpartition coefficients of trace elements Geochemistry International 35119^126

Norman M D Griffin W L Pearson N J Garcia M O ampOrsquoReilly S Y (1998) Quantitative analysis of trace element abun-dances in glasses and minerals a comparison of laser ablationinductively coupled plasma mass spectrometry solution inductivelycoupled plasma mass spectrometry proton microprobe and elec-tron microprobe data Journal of Analytical Atomic Spectrometry 13477^482

Norman M D Garcia M O amp Pietruszka A J (2005) Trace-element distribution coefficients for pyroxenes plagioclase andolivine in evolved tholeiites from the 1955 eruption of KilaueaVolcano Hawairsquoi and petrogenesis of differentiated rift-zonelavas American Mineralogist 90 888^899

OrsquoHara M J amp Herzberg C (2002) Interpretation of trace elementand isotope features of basalts Relevance of field relations petrol-ogy major element data phase equilibria and magma chambermodeling in basalt petrogenesis Geochimica et Cosmochimica Acta 662167^2191

PadghamW A amp FysonW K (1992) The Slave Province a distinctcraton CanadianJournal of Earth Sciences 29 2072^2086

Pearson D G Davies G R amp Nixon P H (1993) Geochemical con-straints on the petrogenesis of diamond facies pyroxenites from theBeni Bousera peridotite massif North Morocco Journal of Petrology34 125^172

Pearson N J Griffin W L Doyle B J OrsquoReilly S Y vanAchterbergh E amp Kivi K (1999) Xenoliths from kimberlitepipes of the Lac de gras area Slave Craton Canada In GurneyJ J Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite ConferenceCapeTown Red Roof Design pp 644^658

Perfit N R Gust D A Bence A E Arculus R J amp Taylor S R(1980) Chemical characteristics of island-arc basalts implicationsfor mantle sources Chemical Geology 30 227^256

Polat A Hofmann A W amp Rosing M T (2002) Boninite-likevolcanic rocks in the 37^38 Ga Isua greenstone belt WestGreenland geochemical evidence for intra-oceanic subductionzone processes in the early Earth Chemical Geology 184 231^254

Poudjom DjomaniY H GriffinW L OrsquoReilly SY amp Doyle B J(2005) Lithospheric domains and controls on kimberlite emplace-ment Slave Province Canada evidence from elastic thickness andupper mantle composition Geochemistry Geophysics Geosystems 6doi1010292005GC0009978

Presnall D C Gudfinnsson G H amp Walter M J (2002)Generation of mid-ocean ridge basalts at pressures from 1 to 7GPa Geochimica et Cosmochimica Acta 66 2073^2090

Schmidberger S S Heaman L M Simonetti A Creaser R A ampCookenboo H O (2005) Formation of Paleoproterozoiceclogitic mantle Slave Province (Canada) Insights from in-situHf and U^Pb isotopic analyses of mantle zircons Earth and

Planetary Science Letters 240 621^633

Schmidberger S S Simonetti A Heaman L M Creaser R A ampWhiteford S (2007) Lu^Hf in-situ Sr and Pb isotope and traceelement systematics for mantle eclogites from the Diavik diamondmine Evidence for Paleoproterozoic subduction beneath the Slavecraton Canada Earth and Planetary Science Letters 254 55^68

Schulze D J amp Helmstaedt H (1988) Coesite-sanidine eclogitesfrom kimberlite products of mantle fractionation or subductionJournal of Geology 96 435^443

Smithies R H Champion D C amp Sun S S (2004) The case forArchaean boninites Contributions to Mineralogy and Petrology 147705^721

Smyth J R Caporuscio F A amp McCormick T C (1989) Mantleeclogites evidence of igneous fractionation in the mantle Earth andPlanetary Science Letters 93 133^141

Sobolev NV amp Lavrentrsquoyev Y G (1971) Isomorphic sodium admix-ture in garnets formed at high pressure Contributions to Mineralogy

and Petrology 31 1^12Stalder R Foley S F Brey G P amp Horn I (1998) Mineral aqueous

fluid partitioning of trace elements at 900^12008C and 30^57GPa New experimental data for garnet clinopyroxene and rutileand implications for mantle metasomatism Geochimica et

Cosmochimica Acta 62 1781^1801Sun S-S amp McDonough W F (1989) Chemical and isotopic sys-

tematics of oceanic basalts implications for mantle compositionsand processes In Saunders A D amp Norris M J (eds)Magmatism in the Ocean Basins Geological Society London Special

Publications 42 313^345Taura H Yurimoto H Kurita K amp Sueno S (1998) Pressure

dependence on partition coefficients for trace elements between oli-vine and the coexisting melts Physics and Chemistry of Minerals 25469^484

Taylor L A amp Neal C R (1989) Eclogites with oceanic crustal andmantle signatures from the Bellsbank kimberlite South AfricaPart I Mineralogy petrography and whole rock chemistryJournal of Geology 97 551^567

van Achterbergh E Ryan C G amp Griffin W L (1999) Glitter online intensity reduction for the laser ablation inductively coupledplasma spectrometry Abstract 9th V M GoldschmidtConference LPI Boston MA 905 pp Abstract no 7215

van Breemen O DavisW J amp King J E (1992) Temporal distribu-tion of granitoid plutonic rocks in the Archean Slave ProvinceNorthwest Canadian Shield Canadian Journal of Earth Sciences 292186^2199

Vervoort J D Patchett P J Blichert-Toft J amp Albare de F (1999)Relationship between Lu^Hf and Sm^Nd isotopic systems in theglobal sedimentary system Earth and Planetary Science Letters 16879^99

Viljoen K S Smith C B amp Sharp Z D (1996) Stable and radio-genic isotope study of eclogite xenoliths from the Orapa kimberliteBotswana Chemical Geology 131 235^255

Walter M J (1998) Melting of garnet peridotite and theorigin of komatiite and depleted lithosphere Journal of Petrology 3929^60

Wasserburg G J Jacobsen S B DePaolo D J McCulloch M Tamp Wen T (1981) Precise determination of SmNd ratios Sm andNd isotopic abundances in standard solutions Geochimica et

Cosmochimica Acta 45 2311^2323Westerlund K J Shirey S B Richardson S H Carlson R W

Gurney J J amp Harris JW (2006) A subduction wedge origin forPaleoarchean peridotitic diamonds and harzburgites from thePanda kimberlite Slave craton evidence from Re^Os isotope sys-tematics Contributions to Mineralogy and Petrology 152 275^294

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1873

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Page 2: Origins of Xenolithic Eclogites and Pyroxenites from the Central

Griffin amp OrsquoReilly 2007) or represent subducted oceaniccrust (eg MacGregor amp Manton 1986 Schulze ampHelmstaedt 1988 Jacob et al 1994 Beard et al 1996)Eclogite suites from many localities in the Kaapvaal andSiberian cratons fall into groups with one group havingcharacteristics consistent with a mantle origin whereasanother group (or groups) has been interpreted as crus-tally derived (Taylor amp Neal 1989 Viljoen et al 1996Kopylova et al1999 Barth et al 2002b) In addition to pri-mary differences eclogite compositions can be modifiedduring melting and metasomatism following their forma-tion (Ireland et al 1994 Barth et al 2001)We have carried out a detailed petrographic major- and

trace-element and multi-isotope study to constrain the gen-esis of different eclogite xenolith types found in kimberlitesintruded in the central Slave Craton Canada bearing inmind that they may have been subjected to a range ofpost-formation processes during their residence time inthe Archaean lithosphere including melting and metaso-matic overprinting

GEOLOGICAL SETT INGThe samples are derived from the Diavik A154 kimberlitepipe in the Lac de Gras area of the Slave Craton in theCanadian Northwest Territories (Fig 1) The Slave Cratonis a small (400 km 500 km) Archean block that is com-posed of two major basement domains (1) a Hadean to

Mesoarcheaean domain in the west (Central SlaveBasement Complex CSBC Bleeker et al 1999) with agesof 40^28Ga onto which 273^270Ga tholeiites wereextruded (Padgham amp Fyson 1992 van Breemen et al1992 Isachsen amp Bowring 1994) (2) a Neoarchaean isoto-pically juvenile domain in the east where the tholeiitesequence is notably absent The Mesoarchaean basementthe eastern extent of which is not accurately known dipsunder the eastern domain (Bleeker et al 1999) The originof these domains has been ascribed to arc^continent colli-sion (Kusky 1989 Davis amp Hegner 1992) or to the easterndomain representing attenuated modified Mesoarchaeanlithosphere (Bleeker 2003) A major north^south-trendingprovinciality in the Slave Craton is evident in basementNd and Pb isotope data (Davis amp Hegner 1992) and acoupling of the two domains by 27Ga may be indicatedby pan-Slave calc-alkaline volcanism (van Breemen et al1992) Younger events include 22^18Ga collisions withneighbouring terranes (Hoffman1989) numerous episodesof Proterozoic dike emplacement such as the Malley^McKay dike swarm at 221^223Ga (LeCheminant et al1996) and the 127Ga Mackenzie swarm (LeCheminantamp Heaman 1989) followed by kimberlite volcanism inCretaceous to Eocene time (Creaser et al 2004 Heamanet al 2004)Distinct mantle xenolith suites and spatial distributions

of mantle rock types in the lithosphere beneath different

66deg

64deg

108deg

16deg

Pb

Great Slave Lake Shear Z

one

Bathurst Fault

112deg

Central Slave(40ndash28 Gabasement)

Protero-

zoic

Nd

Eastern Slave(lt28 Ga)

Lac de Graskimberlites

NunavutNWT

Jericho

Ekati

Diavik

Contwoyto Lake

Lac De Gras

Fig 1 Map of the Slave Craton (after Davis et al 2003) showing Pb and Nd isotope lines that separate ancient basement in the west fromjuvenile rocks in the east Localities of the Diavik Ekati and Jericho kimberlites are shown in more detail on the left

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1844

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

parts of the craton have been recognized (Griffin et al1999 2004 Grulaquo tter et al 1999 Kopylova et al 1999MacKenzie amp Canil 1999 Pearson et al 1999 Carbno ampCanil 2002) which have been substantiated by magnetotel-luric and elastic thickness data (Jones et al 2001 PoudjomDjomani et al 2005) Here we focus on our currentknowledge of the mantle evolution in the central SlaveCraton 340 km NE from Yellowknife NT where the Lacde Gras kimberlites were intruded The subcontinentallithospheric mantle (SCLM) sampled by the Lac deGras kimberlites is strongly layered with an ultra-depletedshallow layer and a less depleted deep layer the latter sug-gested to have formed by subcretion of a plume head thatdelivered diamonds containing lower mantle inclusions(Griffin et al 1999 2004 Davies et al 1999 Aulbach et al2007) A Re^Os isochron age of 327034Ga for somesulfides from the deep layer may date the formationof this layer and shows that significantly oldermantle resides beneath some of the 27Ga crust of thejuvenile eastern domain (Contwoyto Terrane) in the SlaveCraton (Aulbach et al 2004a) This isochron ageagrees within uncertainty with one derived from sulfideinclusions in diamonds from the Panda kimberlite(Westerlund et al 2006) just north of Lac de Gras and isconsistent with previous findings that many kimberliteslying in the younger eastern domain appear to havesampled older lower crust and mantle indicative of aneast-dipping trans-lithospheric boundary dating backto 27Ga craton amalgamation (Grulaquo tter et al 1999Irvine et al 2003)

SAMPLES AND ANALYTICALTECHNIQUESAll samples of the present study were retrieved fromdrill core of kimberlite pipe A154S in the Lac de Grasarea from depths between 100 and 470m Samples ofat least 1cm in size were cut out of the drill core andprocessed to minimize sampling bias although thispossibly discriminates against more friable eclogitetypes that are not as well preserved Eighteen eclogitesand pyroxenites were selected for the present study andresults combined with those of Pearson et al (1999and unpublished data) to make up a total of 35 eclogitesand 30 pyroxenitesMajor-element analyses were obtained using a

CAMECA Camebax SX50 electron microprobeIn situ trace-element compositions were determinedwith a custom-built laser-ablation system (designed byS E Jackson) or a Merchantek LUV 266 NdYAG UVlaser system both linked to an Agilent 7500 inductivelycoupled plasma mass spectrometry (ICP-MS) system andreduced using the GLITTER software (van Achterberget al 1999) Accuracy and precision were monitored by

analysing basalt standard BCR-2G with each batch ofsamples as an unknown Results standard deviations anddetection limits are given in Electronic Appendix 1 whichcan be downloaded from httppetrologyoxfordjournalsorg Sr^Nd^Hf isotope data for garnet and clinopyroxene(cpx) separates leached and ultrasonicated in 6N HClfor 30min followed by ultrasonication in three aliquots ofMQ prior to dissolution were obtained using a Nu Plasmamulti-collector (MC) ICP-MS system All analytical workwas carried out in the GEMOC National Key Centreat Macquarie University (wwwesmqeduauGEMOC)following the techniques described by Aulbach et al(2004b) Repeated measurements of standard materialsduring data acquisition (February^July 2003) yielded thefollowing values 87Sr86Sr of 07102570000045 (2 SDnfrac14 42) for the SRM-987 standard 87Sr86Sr of0703520000040 (nfrac1415) for BHVO-1 143Nd144Nd of05111380000038 (nfrac14 42) for the JMC-321 standard143Nd144Nd of 05130030000043 (nfrac1410) for BHVO-1176Hf177Hf of 02821630000002 (nfrac14 20) for the JMC-475 standard and 176Hf177Hf of 02831280000090 forBHVO-1 (nfrac14 8) (Aulbach et al 2004b)Although the uncertainties of Sr and Nd isotope ratios

for pure Sr and Nd standards obtained by MC-ICP-MSare two to three times higher than those reported for ther-mal ionization mass spectrometry (TIMS) measurements(httpgeoremmpch-mainzgwdgde) they are adequatefor the samples analysed in the present study whichmostly have significantly evolved isotopic compositionscompared with depleted mantle or CHUR The signifi-cantly higher uncertainty for 176Hf177Hf of the rock stan-dard than for the pure Hf standard may reflect matrixinterference as a result of imperfectly purified solutions(eg Blichert-Toft et al 1997)

PETROGRAPHY ANDCLASSI F ICAT IONThe petrography of the studied eclogites from Lac deGras is summarized in Table 1 The majority of samplesthat were large enough to obtain microstructural informa-tion have medium- to coarse-grained granoblasticmicrostructures Three eclogites have a weakly tabularmicrostructure which is due to the weak elongation andsubparallel alignment of garnet and cpx in sampleYK3528 and to the preferred orientation of phlogopite insamplesVR19674ecl-g7 and YK1911Most samples are fresh or show minor (510) altera-

tion that is restricted to veins and grain boundaries A fewsamples are more strongly altered with pyroxenesusually more affected than garnet Altered areas arecomposed of small secondary phlogopite sulfide flakesand unidentifiable microcrystalline phases with a feltybrown to green appearance

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1845

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 1 Petrographic data and eclogite classification

Sample no Ref1 Type2 GpMampG3 GpTampN

4 Size5 Microstructure6 Assemblagemode7

gt cpx rut sf ky opx other

Eclogites

SE01 3 high-Mg I A 51 na x x

SE02 3 high-Ca I B 51 na x x x

SE03 3 high-Mg II A 51 na x x

SE04 3 high-Mg II B 51 na x x

SE11 3 low-Mg II B 51 na x x x x

vr09399 2 high-Mg II A 51 na x x

vr19674ecl-g4 2 low-Mg II B 16 na x x

vr19674ecl-g5 2 high-Ca II C 51 na x x

vr19674ecl-g6 2 high-Mg I A 51 na x x

vr19674ecl-g7 1 volatile-rich II A 15 medium-tabular phlog (44) ap (51)

vr19677ecl-g1 2 high-Mg II B 51 na x x

vr19677ecl-g2 2 high-Mg II B 17 na x x x

vr19677ecl-g4 2 high-Mg II B 51 na x x

vr19677ecl-g5 2 high-Mg I B 51 na x x

vr19677ecl-g6 2 high-Mg II A 51 na x x

vr40345 3 high-Ca I B 51 na x x x dia (trace)

vr40364 2 high-Mg II A 51 na x x

vr43452 1 low-Mg I B 45 coarse-grano 48 51 1

vr43465 1 high-Ca I B 2 coarse-grano 41 55 1 graph (3)

vr43469 1 low-Mg II B 15 coarse-grano 60 40 51 51

vr43477 1 low-Mg II B 15 coarse-grano x x x

vr43479 1 low-Mg II A 25 medium-grano 50 50 51

vr43480 1 high-Mg I A 2 medium-grano 89 11 51

vr50858 2 high-Ca II B 51 na x x x

vr50860 2 high-Ca I C 51 na x x

vr50909 3 high-Ca II B 51 na x x x x

vr50916 2 high-Mg II A 51 na x x

vr67112b 1 volatile-rich II C 2 medium-grano 51 ilm (3) cc (2)

vr67360 1 high-Mg II A 3 coarse-grano 62 38 51

yk1911 1 volatile-rich II B 2 medium-tabular 31 18 phlog (49) ap (51) plag (51) cc (51)

yk1926 1 low-Mg II B 17 coarse-grano 70 30 51

yk1943 1 low-Mg I B 2 coarse-grano 57 41 2 51 qz (51)

yk1946 1 high-Mg II A 3 coarse-grano 62 37 1 51

yk1949 1 high-Mg II B 3 coarse-grano 70 30 51 51

yk3528 1 high-Ca I B 25 medium-tabular 57 43 51

Pyroxenites

SE08 3 na II A 51 na x x

SE09 3 na II A 51 na x x x

vr09353 2 na II A 51 na x x x

vr09384 2 na II B 51 na x x x

vr09405 2 na II A 51 na x x

vr09406 2 na II A 51 na x x x

vr09424 3 na II A 51 na x x x

(continued)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1846

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Some xenoliths have subhedral or rounded garnets in alsquomatrixrsquo of interstitial cpx (Group I of MacGregor ampCarter 1970 Fig 2a) whereas in others garnet and cpxhave straight grain boundaries and an interlocking fabric(Group II Fig 2b) The classification of McCandless ampGurney (1989) builds on work of MacGregor amp Carter(1970) and distinguishes group I eclogites by their higherNa2O in garnet (009wt ) and K2O in cpx (008wt) from group II eclogites The classification used inTable 1 (GpMampG) is that based on Na2O contents ingarnet A different classification (GpTampN) places eclogitesinto groups A B and C distinguished by the MgO FeOand CaO contents of garnets and Na2O and MgO contentsof clinopyroxenes (Coleman et al 1965 Taylor amp Neal

1989) These classifications are based on distinct eclogitesuites from Southern African kimberlites and do not neces-sarily reflect the particularities of eclogites from otherlocalities We therefore use a slightly different scheme(lsquotypesrsquo in Table 1) that reflects the specific groupings ofeclogites in this study with regard to Cr2O3^CaO CaO(MgOthornFeO)^MgOFeO and CaO^Na2O relationshipsin garnet (see below)Garnet modes in volatile-free eclogites range from 40

to 90 vol and cpx modes from 10 to 55 with anaverage and median garnet mode of 60 and 65respectively Two samples have high modal amounts ofphlogopite (44 and 48 respectively) in addition togarnet and cpx and are therefore not eclogites sensu stricto

Table 1 Continued

Sample no Ref1 Type2 GpMampG3 GpTampN

4 Size5 Microstructure6 Assemblagemode7

gt cpx rut sf ky opx other

vr19673ecl-g1 2 na II B 51 na x x x

vr19673ecl-g2 2 na II B 51 na x x x x

vr19674ecl-g1 2 na II A 51 na x x x

vr19674ecl-g2 2 na II A 51 na x x x

vr19674ecl-g3 2 na II A 51 na x x

vr19674per-g1 2 na II A 51 na x x

vr19674per-g4 2 na II A 51 na x x x

vr40302 2 na II A 51 na x x x

vr40336 3 na II A 51 na x x x

vr40337 2 na I A 51 na x x

vr40374 2 na II A 51 na x x x

vr40382 2 na II A 51 na x x x

vr40384 ecl-g4 2 na II A 51 na x x x

vr40399 2 na II B 51 na x x x

vr50856 2 na II A 51 na x x x

vr50887 2 na II A 51 na x x

vr50888 2 na II A 51 na x x

vr50899 2 na II A 51 na x x x

vr50900 2 na II A 51 na x x x

vr50925 2 na II B 51 na x x

yk1914 1 na II A 25 medium-grano 22 78

yk1915 1 na II A 15 coarse-grano x x x x

yk1952 1 na II A 2 medium-grano x x x x

1References 1 this study 2 Pearson et al (1999) 3 Pearson et al (unpublished data) trace-element analysesthis study2Classification based on chemical parameters (see text for details)3Classification of McCandless amp Gurney (1989) based on MacGregor amp Carter (1970)4Classification of Taylor amp Neal (1989) based on Coleman et al (1965)5Longest dimension (cm) across xenolith6Only given for samples41 cm in size grano granoblastic7Modes (vol ) are given where available otherwise presence of mineral is indicated as identified in thin section or stubgt garnet cpx clinopyroxene ky kyanite rut rutile sulf sulfide phlog phlogopite ap apatite dia diamond graphgraphite cc calcite ilm ilmenite pl plagioclase qz quartz opx orthopyroxene na not applicable or not available

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1847

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(phlogopite occurring as small secondary grains at grainboundaries and in veins is not considered in this study)Ten samples of all eclogite types show secondary spongyrims of cpx around cpx cores Secondary rims containsmall (mm-scale) patches of glassRutile is the most common primary accessory phase

(15 of 65 eclogites and pyroxenites) with modes between51 and 2 vol Sulfide occurs in 14 samples and ispresent as rounded to subhedral to irregular grains

Kyanite has been identified in three samples studied byPearson et al (1999) One of the kyanite-bearing eclogitesalso contains diamond (VR40345) Kyanite as well asdiamond or graphite associated with kyanite-bearing andcompositionally similar eclogites (high-Ca eclogites) hasbeen frequently observed in drill core (Pearson et al1999) but because these eclogite types are often alteredand friable they are not proportionally represented in thepresent study One high-Ca eclogite (VR43465) contains

Fig 2 Eclogites in thin section (plane-polarized light) illustrating the microstructures mentioned in the text (a) high-Mg eclogiteVR67360(b) high-Mg eclogiteYK1946 (c) phlogopitized eclogiteYK1911 (d) garnet pyroxeniteYK1915 (e) garnet pyroxeniteYK1914 (f) garnet pyrox-enite schlieren in a matrix of garnet spongy opx and glass (YK1952) Grt garnet cpx clinopyroxene rut rutile phlog phlogopite Scalebar in(c) applies to all sections (1cm)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1848

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

3 vol graphite which occurs as disseminated plates sev-eral millimetres long with prismatic tabular or irregularhabit Quartz has been retrieved from a mineral separateof low-Mg eclogite YK1943 The phlogopitized samples(VR19674ecl-g7 and YK1911) also contain apatite andglass as accessory phases Vesicles in sample VR67112bare partially filled with single or multiple large carbonatecrystalsMost pyroxenites from Lac de Gras have fine- to coarse-

grained granoblastic equilibrated microstructures(Pearson et al 1999) Modal information for pyroxenitesin the present study is available for only the larger sampleYK1914 (22 garnet 78 cpx) (Fig 2e)

MAJOR ELEMENTSAverage major-element compositions of garnet cpx andopx in the different eclogite types (except for the threevolatile-rich eclogites which are not averaged) and in pyr-oxenites are given in Table 2 the full datasets includingaccessory minerals are available as ElectronicAppendices 2^5 which can be downloaded from httppetrologyoxfordjournalsorg

GarnetGarnets have XMg [pyrope component 100Mg(MgthornFethornCathornMn)] ranging from 178 to 756 andXCa (grossular component mol) ranging from 72 to359 (Electronic Appendix 2) In a diagram of CaO(MgOthornFeO) vs MgOFeO (Fig 3) eclogitic garnetsfall into three groups (1) low-Mg garnets with low

CaO(MgOthornFeO) and low MgOFeO (2) high-Mggarnets with low CaO(MgOthornFeO) and high MgOFeO (3) high-Ca those with high CaO(MgOthornFeO)and intermediate MgOFeO Volatile-rich eclogites havegarnets with the lowest MgOFeO whereas pyroxenitestrend towards higher values than eclogitesMost eclogite garnets have Cr2O3 contents502wt

with variable CaO contents (Fig 4) Garnets in pyroxeniteshave low CaO contents which correlate positively withCr2O3 More than half of the pyroxenites are opx-free andall pyroxenites are olivine-free on thin-section scalePyroxenitic garnets show an opposing trend of CaOvs Na2O (Fig 5) compared with the eclogitic garnetsBecause the positive correlation of Cr2O3 and CaO forpyroxenitic garnet indicates buffering by both cpx andopx opx-bearing or opx-free pyroxenites will not befurther distinguished subsequently (except in the geother-mobarometry section where opx-bearing assemblagesallow simultaneous calculation of pressure andtemperature)Distinct rim compositions in most garnets are

characterized by lower CaO contents and higherMg-number [frac14 100Mg(MgthornFe)] Garnet compositionsin three samples are inhomogeneous with regard to CaOMgO and FeO (SE01 VR43479 and VR43480) withoutclear core^rim zonations similar to coexisting cpx Likerims CaO-poor compositions are characterized by higherMg-number Visibly secondary spongy cpx rims also havehigher MgO contents (see below) and MgO-rich garnetsare therefore regarded as affected by secondary changesBased on Cr^Ca Mg^Ca^Fe and Ca^Na composi-

tional relationships in garnet the following eclogitetypes are distinguished (1) eclogites with high-Ca

0

05

10

15

20

25

30

0 01 02 03 04 05 06 07 08 09

high-Ca

high-Mg+pyroxenites

low-Mg

CaO(MgO+FeO)

MgO

FeO

GARNETpyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropic Jericho massiveDiavik low-temperatureDiavik high-temperatureDiavik diamond-bearing

eclogites

Fig 3 CaO(MgOthornFeO) vs MgOFeO (wt ) in garnet Fieldsdistinguish low-Mg and high-Ca eclogites as well as high-Mg eclogitesplus pyroxenites in this study Garnets in eclogites from Jericho(Kopylova et al 1999) and Diavik (Schmidberger et al 2007) areshown for comparison Garnet inclusions in diamond from Davieset al (1999 2004)

-135 wt (high-Ca)

2

3

4

5

6

7

8

9

0 05 10 15 20 25Cr2O3

CaO

GARNET

opx-bearing

opx-free

pyroxenites

low-Mghigh-Mgvolatile-rich

Jericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

VR19673ecl-g1g2

Fig 4 Cr2O3 vs CaO (wt ) in garnet Garnets in eclogites fromJericho and Diavik as in Fig 3

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1849

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 2 Summary of major-element contents in garnet and cpx (wt )

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO Na2O Total Mg-no T(Krogh) Type

Garnet

Low-Mg ecl 3964 026 2214 011 1843 037 1050 846 009 1000 503 976

minimum 3893 008 2130 5009 1071 024 849 567 006 990 423 860

maximum 4094 036 2275 011 2093 046 1457 1062 012 1004 708 1088

1s 070 012 052 na 354 007 227 185 003 05 99 98

High-Mg ecl 4081 026 2261 013 1419 030 1537 624 007 1000 659 969

minimum 3970 008 2224 5009 951 023 1127 418 003 991 551 820

maximum 4250 052 2323 035 1795 036 2030 882 013 1012 792 1241

1s 077 013 031 na 225 004 189 141 003 05 57 88

High-Ca ecl 3979 034 2211 na 1516 027 958 1240 011 998 528 1078

minimum 3909 011 2117 5009 1164 023 801 1144 007 989 456 910

maximum 4039 073 2242 011 1713 035 1249 1351 018 1004 657 1246

1s 042 026 042 na 173 004 155 073 004 06 67 151

vr19674ecl-g7 3867 003 2187 5009 2380 153 793 627 5004 1002 373 volatile-rich

vr67112b 3783 013 2097 009 2620 204 466 790 5004 999 241 volatile-rich

yk1911 3874 5006 2200 5009 2120 221 802 747 5004 997 403 volatile-rich

Pyroxenites 4163 021 2257 075 1148 034 1906 407 006 1002 747 909

minimum 3989 005 2081 009 729 025 1535 277 004 996 593 699

maximum 4279 048 2330 260 1881 054 2215 566 009 1011 842 1158

1s 068 010 057 057 323 007 209 065 001 04 73 117

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO Na2O Total Mg-no Type

Cpx

Low-Mg ecl 5453 028 739 na 500 5009 1117 1662 434 994 801

minimum 5160 017 564 5009 337 na 876 1388 197 981 741

maximum 5570 044 1067 014 687 na 1324 1926 632 1001 865

1s 128 011 155 na 124 na 153 177 126 06 40

High-Mg ecl 5483 023 483 na 381 na 1368 1875 318 994 865

minimum 5353 009 263 5009 245 5009 1156 1617 204 986 778

maximum 5645 031 810 016 674 011 1588 2101 495 1011 920

1s 071 007 137 na 095 na 104 138 075 06 30

High-Ca ecl 5563 027 1184 na 318 5009 868 1368 615 995 836

minimum 5448 014 769 5009 168 na 619 1060 391 987 751

maximum 5670 049 1579 010 603 na 1148 1689 807 1003 878

1s 076 015 334 na 151 na 189 261 167 06 45

vr19674 ecl-g7 5322 010 185 5009 536 015 1448 2339 062 992 828 volatile-rich

vr67112b 5040 041 328 5009 1079 024 1108 2229 071 992 647 volatile-rich

yk1911 5231 014 301 5009 544 024 1450 2340 056 996 826 volatile-rich

Pyroxenites 5527 na 347 038 304 na 1543 1959 245 999 899

minimum 5299 5006 037 010 149 5009 1128 1408 030 990 804

maximum 5666 037 786 134 490 014 1863 2312 537 1009 947

1s 068 na 189 028 091 na 193 226 131 05 37

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO NiO Total Mg-no

Opx

Pyroxenites 5781 na 059 na 627 na 3508 037 011 1006 909

minimum 5546 5006 032 5009 462 5009 2999 020 008 996 804

maximum 5888 009 318 054 1304 014 3628 077 014 1014 933

1s 098 na 067 na 212 na 160 019 002 06 32

Averaged for the three eclogite types and for pyroxenites minimum and maximum values as well as standard deviationsalso given Concentrations in volatile-rich eclogites are shown individually ecl eclogite Mg-numberfrac14 100Mg(Mgthorn Fe)Temperatures of last equilibration [T(Krogh)] as in Electronic Appendix 8

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1850

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

garnets (these include kyanite- graphite- and diamond-bearing varieties) (2) eclogites with low-Mg garnets (theseinclude quartz-bearing eclogite) (3) eclogites with high-Mggarnets and Cr2O3 contents502wt (4) volatile-richeclogites (phlogopite- apatite- carbonate-bearing)

Clinopyroxene (cpx)Clinopyroxenes in all eclogite groups have higher averageAl2O3 contents than those in pyroxenites with the formerhaving about 90 of the total Al in the jadeite molecule(AlVI) versus about 35 for the latter corresponding toan average of 41 and 25wt Na2O respectivelyThe highest Al2O3 and Na2O contents in the dataset areobserved in cpx from the high-Ca eclogites (up to158wt Al2O3) (Electronic Appendix 3) Contents ofK2O are generally below the detection limit(5004wt ) but values up to 024wt are observed insome cpx in high-Ca eclogites Average cpx Mg-numberare highest in pyroxenites (899) followed by cpx inhigh-Mg eclogites (865) high-Ca eclogites (836) andlow-Mg eclogites (801) Pyroxenitic cpx tend to havelower TiO2 and higher NiO and distinctly higherCr2O3 contents compared with those in eclogites A plotof MgO vs Al2O3 shows a negative correlation (Fig 6)Clinopyroxenes in low-Mg eclogites have lower Al2O3

contents at a given MgO content than other eclogitetypes Between 15 and 17wt MgO the Al2O3^MgOcorrelation for cpx in pyroxenites has a different slopefrom that in the eclogites Clinopyroxenes in pyroxenites

VR19673ecl-g1 and VR19673ecl-g2 have distinctly lowerMgO contents and higher Al2O3 contents than inthe other pyroxenites They coexist with garnets havingthe lowest Cr2O3 and CaO contents of the dataset whichdefine one end of the positive Cr2O3^CaO correlationobserved for garnetsClinopyroxenes in volatile-rich eclogites have distinctly

high MnO contents (up to 024wt ) Mg-numbers forcpx in the phlogopite-bearing variety are 826 and 828whereas that in the calcite-bearing eclogite (VR67112b)has an Mg-number of 647Some cpx grains have spongy rims that contain small

(mm-scale) patches of glass these rims are richer in Caand Mg and markedly poorer in Na than the cores similarto cpx in eclogites from southern Africa (Taylor amp Neal1989) The rim texture is unequilibrated suggesting lategrowth The formation of the spongy cpx rims might berelated to incongruent melting of cpx or be a result ofmetasomatism (Taylor amp Neal 1989 and referencestherein) Distinct although not spongy rims in two eclo-gites (VR67360 and YK1946) are similarly enriched inCaO and depleted in Na2O Two types of cpx with respectto CaO Na2O and Mg are present in VR43477VR43479 and VR43480 without clear core^rim zonationor spongy rims No significant differences are recognizedwith respect to other oxides

Orthopyroxene (opx)Orthopyroxenes have Mg-number ranging from 804 to933 (17 of 19 samples have values between 895 and 933)and typically contain minor amounts of Al2O3

(032^118wt one outlier has 318wt ) CaO(020^118 wt ) and NiO (008^014wt ) (Electronic

0

2

4

6

8

10

12

14

7 9 11 13 15 17 19

Al 2

O3

CPX

pyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

cpx in pyroxenitesVR19673ecl-g1g2

MgO

Fig 6 MgO vs Al2O3 (wt ) in cpx Data sources as in Figs 3and 4

0

005

010

015

020

2 4 6 8 10 12 14 16

CaO

Na 2

O

dl

group Igroup II

pyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

GARNET

Fig 5 CaO vs Na2O (wt ) in garnet Arrows show trends foreclogites and pyroxenites respectively [note that part of thepyroxenite trend extends into the field falling below the detectionlimit (dl) for Na2O] Group I and group II refer to classificationof McCandless amp Gurney (1989) Data sources as in Figs 3 and 4

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1851

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Appendix 4) The contents of all other elements are mostlybelow their respective detection limits

Accessory phasesRutile and ilmenite occur in some of the eclogites (Table 1)All rutile grains have exsolved 10 ilmenite as lamellaeand rims Ilmenite-free areas of rutile in eclogites containbetween 962 and 988wt TiO2 and minor amounts ofAl2O3 (01^04wt ) Cr2O3 (5009 to 02wt ) andFeO (03^07wt ) Rutile in pyroxenite YK1915 contains004wt Al2O3 01wt FeO and 15wt Cr2O3Discrete ilmenite grains in sample VR67112b haveTiO2 contents of 499 and 500wt MgO of 23 and15wt and FeO of 399 and 403wt respectivelySulfides occur interstitially or enclosed in cpx garnet

and opx they are mostly pyrrhotite and monosulfide solidsolution with variable Ni contents (unpublished data)They will not be considered further hereAbundant platy phlogopite is observed in two samples

(VR19674ecl-g7 and YK1911) It is Cl-rich in sampleVR19674ecl-g7 (063wt vs 003wt in YK1911) andF-rich in sample YK1911 (097wt vs 005wt inVR19674ecl-g7) (Electronic Appendix 5) The same is truefor coexisting apatite (VR19674ecl-g7 254wt Cl and014wt F YK1911 006wt Cl and 306wt F)

Comparison with mineral inclusions indiamonds from Lac de Gras and witheclogite xenoliths from other Slave CratonlocalitiesMineral inclusions in diamonds from the Lac de Graskimberlites have been investigated by Davies et al (19992004) For garnets there is a striking similarity betweenthe inclusions in diamond and those in the high-Ca eclogitexenoliths although the inclusions in diamond trendtowards higher Na2O (Fig 5) and lower Al2O3 contentsThe overlap is not as marked for cpx where inclusionsin diamond span almost the entire range observed foreclogitic cpx (Fig 6)A comparison with eclogites from other localities in

the Slave Craton shows that garnets in most Diavik low-temperature eclogites reported by Schmidberger et al(2007) plot with garnets in high-Mg eclogites and inpyroxenites from this study (Figs 3^5) whereas garnets inhigh-temperature and diamond-bearing eclogites(Schmidberger et al 2007) overlap most with garnets inhigh-Ca eclogites from this study and with garnetsincluded in diamond Eclogites from the Jericho kimberlitehave either massive or anisotropic fabrics (Kopylova et al1999) Garnets in almost all massive eclogites plot withgarnets in low-Mg eclogites from this study with regard toCaO^MgO^FeO relationships (Fig 3) but many havelower Na2O contents at a given CaO content (Fig 5)Garnets in anisotropic eclogites are restricted to lowMgOFeO (507) and show some overlap with garnets

in high-Ca eclogites although trending towards lowerCaO(MgOthornFeO) (Fig 3) while coexisting cpx is alsorestricted to low MgO and trends toward high Al2O3 con-tents (Fig 6)

TRACE ELEMENTSGarnetTrace-element abundances in garnet are summarized inTable 3 and the complete dataset is reported in ElectronicAppendix 1 Garnets in low-Mg eclogites show steep posi-tive slopes between LaN and SmN and shallower positiveslopes in the normalized heavy rare earth elements(HREEN Fig 7a) The normalized light REE (LREEN)show an order of magnitude variability ZrHf is mostlysupra-chondritic and Ti can be enriched or depleted rela-tive to elements of similar compatibility Niobium in threesamples is strongly depleted at 002^003 chondriticGarnets in high-Mg eclogites also show smooth positiveslopes in LREEN but have flat HREEN and variablenormalized middle REE (MREEN) to HREEN (Fig 7b)ZrHf is always supra-chondritic and Ti enriched ordepleted relative to similarly compatible elementsNiobium is most samples is below detection Small positiveEu anomalies though not outside the analytical uncer-tainty for Eu relative to either Sm or Gd are observed forgarnet in some high-Mg eclogites (eg EuEu up to 15for VR67360 where Eu is the average of the chondrite-normalized Sm and Gd concentrations) Garnets in high-Ca eclogites are slightly enriched in the MREEN relative tothe HREEN or have flat MREEN and HREEN with ahump between EuN and HoN (Fig 7c) One sample(VR40345) shows strong La enrichment (greater thanchondrite) Two samples have supra-chondritic ZrHfwhereas three samples have close to chondritic ratios andone sample has a subchondritic ratio As is true for theother eclogite typesTiN is either enriched or depleted rela-tive to similarly compatible elementsPyroxenitic garnets have variable slopes in the LREEN

and flat MREEN to HREEN patterns and two samplesshow high Sc relative to neighbouring elements Garnet inone of the pyroxenites (YK1952) has a highly distinctivepattern relative to other pyroxenites with very low LREEand HFSE (Fig 7d) This sample consists of coarse pyroxe-nite schlieren in which the garnet was analysed in a finer-grained lsquomatrixrsquo of spongy cpx with secondary rims finecpx^opx intergrowths embayed garnet and melt patchesVolatile-rich eclogites have garnets that are slightly

enriched in MREEN relative to HREEN or have flatMREE to HREE patterns One garnet has a pronouncednegative Eu anomaly (EuEufrac14 07 forVR67112b Fig7e)Garnets in volatile-rich samples have the lowest Ni Coand Ti and the highest V ScYand HREE abundances inthe dataset

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1852

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 3 Summary of trace-element contents in garnet and cpx (ppm)

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR19674

ecl-g7

VR67112b YK1911 av min max 1s

n 6 6 7 3 5 4 4

P 169 54 270 88 196 70 310 78 282 80 450 138 160 60 80 214 48 370 144

Sc 62 40 89 16 48 36 67 11 53 42 79 15 69 92 93 186 52 440 179

Ti 1667 490 2500 680 1073 620 2000 455 2141 700 4700 1582 270 690 260 1097 540 1900 573

V 169 78 250 68 150 78 280 72 178 110 390 101 140 340 100 390 71 820 326

Co 59 42 70 11 58 39 84 13 58 47 65 5 30 41 27 68 55 90 16

Ni 17 5 32 10 21 11 34 10 24 11 35 8 6 10 51 54 25 110 38

Ga 12 7 15 3 13 8 16 3 12 9 20 4 8 86 58 12 10 14 2

Sr na 512 4 na na 522 522 na na 57 57 na 52 5025 510 513 513 513 na

Y 37 6 61 18 21 7 38 10 31 20 38 5 60 81 70 19 12 26 6

Zr 15 5 29 9 28 6 83 26 24 9 43 12 60 22 90 18 1 41 17

Nb 009 0005 048 019 na 50015 05 na na 5005 05 na 5003 0012 5001 022 0030 057 025

La 005 001 009 004 na 50007 00 na na 5003 04 na 004 002 0009 003 000 009 005

Ce 03 01 05 02 01 00 01 00 05 01 18 05 03 022 015 01 00 03 01

Pr 01 00 02 01 00 00 01 00 02 01 03 00 017 014 012 00 00 01 00

Nd 12 03 24 07 07 04 14 03 21 12 32 07 24 23 20 03 00 07 03

Sm 11 04 22 06 10 03 18 05 23 08 36 10 29 31 33 05 01 09 04

Eu 06 03 10 03 06 02 10 03 11 04 17 04 16 114 16 03 01 06 02

Gd 27 07 41 12 23 07 42 12 43 16 53 13 8 81 90 13 06 24 08

Dy 54 11 70 23 36 13 65 18 58 35 71 11 12 144 12 29 18 42 10

Ho 14 02 23 07 08 03 15 04 12 09 14 02 24 30 23 07 04 10 02

Er 46 07 80 24 24 09 40 10 34 25 40 05 8 85 6 22 14 28 06

Yb 52 06 98 30 24 10 37 08 32 22 39 05 8 70 5 23 13 30 07

Lu 09 01 16 05 04 02 05 01 05 03 07 01 11 09 08 04 02 05 01

Hf 03 02 05 01 04 01 10 03 05 02 09 02 09 027 007 03 00 06 03

Ta na 50008 5004 na na 50005 001 na na 50004 na na 003 50002 50005 na 50007 011 na

Pb na 5005 02 na na 5002 02 na na 5003 07 na 02 004 5002 na 5003 01 na

Th na 5001 005 na na 50007 003 na na 50008 011 na 5007 0012 50004 002 0005 006 002

U na 5003 5003 na na 5001 01 na na 5007 na na 01 0023 010 na 5001 003 na

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR67112b YK1911 av min max 1s

n 6 4 8 5 4 4

P 65 15 120 41 39 14 56 16 53 30 70 12 20 9 80 27 120 40

Sc 22 10 30 7 15 14 19 2 14 7 24 5 80 43 17 11 21 5

Ti 1557 990 2400 508 1168 1010 1490 191 1988 810 3500 1154 2390 1700 689 102 1685 694

V 415 180 480 107 318 210 400 69 324 259 450 72 600 180 250 139 480 156

Co 35 26 50 8 20 14 31 7 26 14 42 11 49 23 29 20 42 10

Ni 164 87 220 47 228 100 430 128 188 135 320 56 59 34 428 410 670 184

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1853

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

ClinopyroxeneTrace-element abundances in cpx are summarized inTable 3 and the complete dataset is reported in ElectronicAppendix 6 Clinopyroxenes in low-Mg eclogites havemildly positive to negative slopes from LaN to NdN andsmooth negative slopes between SmN and HoN thatmostly continue to LuN (Fig 8a) Niobium is below detec-tion or 5007 chondritic and SrN is variably enrichedrelative to neighbouring elements with the peak size corre-lating negatively with LREEN ZrHf is subchondritic andHfN is enriched relative to SmN whereasTiN is enriched ordepleted relative to similarly compatible elementsWe ana-lysed the spongy rim in cpxYK1943 and its trace-elementpattern is similar to that of cpx cores for more compatibleelements whereas the most incompatible elements areseveral orders of magnitude higher Clinopyroxenes in

high-Mg eclogites have steeper negative slopes betweenNdN and LuN than those in low-Mg eclogites combinedwith lower HREE abundances HfN is depleted relative toSmN (Fig 8b) High-Ca eclogites contain cpx with sinusoi-dal REEN patterns with lower LREE than other eclogitetypes and the heaviest REE below detection They haveuniformly strong positive Sr peaks relative to elements ofsimilar compatibility (Fig 9) The diamond-bearing high-Ca eclogite (VR40345) has cpx with unusually high Nband LREE abundancesClinopyroxenes in four pyroxenites have disparate trace

element patterns They all share depletions in TiN relativeto similarly compatible elements and two show positiveSrN peaks Two samples have cpx with high and flatLREEN to MREEN patterns and steep negative slopes inthe HREEN The trace-element patterns of cpx in YK1914

Table 3 Continued

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR67112b YK1911 av min max 1s

n 6 4 8 5 4 4

Ga 16 10 20 3 12 9 15 2 26 15 39 9 129 123 6 11 11 5

Rb na 5008 04 na na 5007 05 02 na 5003 03 na 02 5004 03 010 05 0

Sr 166 85 290 74 222 153 285 50 243 83 440 142 85 120 200 11 414 209

Y 2 04 4 10 13 07 2 04 11 03 3 10 28 51 2 03 3 2

Zr 19 4 48 15 25 8 45 13 17 8 28 7 111 129 36 008 81 42

Nb 01 0003 03 01 na 5002 00 na na 5002 16 9 007 011 010 003 02 008

Cs na 5001 6 na na 5006 5006 na5013 5004 5013 na bdl 5002 na 5001 5002 na

Ba 09 02 2 07 09 01 2 07 10 03 3 10 06 01 13 06 3 10

La 3 03 8 3 3 23 3 04 4 01 14 6 51 97 4 01 10 4

Ce 7 13 20 6 10 7 13 2 7 05 24 10 25 28 11 01 28 13

Pr 11 02 3 08 2 10 3 06 08 02 2 09 47 46 2 002 4 2

Nd 5 12 12 4 10 5 14 4 4 09 8 26 23 22 8 01 22 10

Sm 11 05 21 06 20 10 31 10 07 02 11 03 47 53 2 005 5 2

Eu 03 01 05 01 06 028 09 03 02 010 03 007 087 123 06 002 15 07

Gd 08 03 13 03 11 05 19 06 05 03 06 011 28 37 15 010 36 2

Dy 05 01 07 02 04 022 06 01 02 013 03 007 11 16 07 008 14 06

Ho 01 00 02 004 006 003 007 002 na 5003 012 na 011 021 009 001 02 007

Er na 507 507 na 011 004 02 00 na 5006 03 na 019 033 02 002 03 012

Yb na 5001 02 na na 5001 011 na na 5006 03 na 007 015 009 001 02 006

Lu na 5001 003 na na50002 002 na na50008 01 na 0006 0013 0013 0002 003 001

Hf 12 03 24 07 11 04 18 05 10 06 16 04 34 53 na50015 35 na

Ta na50003 50015 na na50006 006 na na50004 5004 na 0014 00236 na50003 503 na

Pb 06 02 12 04 06 04 09 02 na 5016 07 02 065 098 08 01 18 1

Th na50005 06 na 012 005 02 005 na 5002 5006 na 021 02 001 06 03

U na 5005 5005 na 002 0009 005 001 na 5003 5005 na 0013 024 na50005 003 na

Averaged for the three eclogite types and for pyroxenites as in Table 2 pyx pyroxenite n number of analyses averagedbdl below detection limit

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1854

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

and YK1952 are similar although at different abundanceswith only modest LREEN over HREEN enrichment andstrong depletions in ZrN relative to neighbouring elements(Fig 8d)Clinopyroxenes in the volatile-rich samples have flat pat-

terns and high abundances from LaN to SmN and smoothnegative slopes from SmN to LuN Niobium Sr and Ti are

noticeably depleted relative to elements of similar compat-ibility (Fig 8e)

Trace-element partitioningTo assess whether coexisting phases are in trace-elementequilibrium we calculated apparent cpx^garnet distribu-tion coefficients and compared them with those

(a)

(d)

(b)

(e)

(c)

GarnetCI-chondrite GarnetCI-chondrite

low-Mghigh-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1952

1

10

Nd Sm Eu Gd Dy

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Fig 7 Extended trace-element patterns of garnet in (a) low-Mg eclogites (b) high-Mg eclogites with inset showing enlargement of MREE tohighlight Eu anomalies and 1s uncertainties (c) high-Ca eclogites (d) pyroxenites and (e) volatile-bearing eclogites Normalized to chondritevalues of McDonough amp Sun (1995)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1855

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

determined experimentally or in demonstrably equili-brated natural samples Low-Mg and high-Mg eclogiteshave Dcpxgarnet for most elements that are within therange determined for natural equilibrated eclogites andexperimental assemblages (Harte amp Kirkley 1997 Greenet al 2000 Barth et al 2002a) (Electronic Appendix 7)High-Ca eclogites plot at the low end of the range

of distribution coefficients and have very low DMREE

and DHREE outside the range indicated by the datasetof Harte amp Kirkley (1997) who noted that DREE decreaseswith increasing Ca content in eclogitic minerals Theseassemblages are therefore considered to be in equilibriumPyroxenite VR40384 has D values similar to eclogites

but at the high end of the range determined by Harte amp

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

(a)

(d)

(b)

(e)

(c)

CpxCI-chondrite CpxCI-chondrite

low-Mg high-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1914

YK1952

spongy rim

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

VR40345

Fig 8 Extended trace-element patterns of cpx as in Fig 7

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1856

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Kirkley (1997) (Electronic Appendix 7) Another group(YK1914 and YK1952) plots at the low end but DHf DZrDTi and DV are far below the values indicated by experi-mental datasets Because pyroxenites compositionallygrade into eclogites their drastically different DHFSE

values suggest disequilibrium In contrast volatile-bearingsamples have DHFSE values that appear to be too high to bein equilibrium however this may be a consequence oftheir unusual bulk chemical composition (Ti- and K-rich)which differs substantially from that of lsquonormalrsquo eclogitesand experimental charges

GEOTHERMOBAROMETRYFollowing Schmidberger et al (2007) temperatures oflast equilibration were calculated for eclogitic andopx-free pyroxenitic assemblages using the formulationof Krogh (1988) The results are summarized in Table 2and Electronic Appendix 8 for a preset pressure of5 GPa and illustrated as a histogram in Fig 9 Foropx-bearing pyroxenites temperatures were calculatedsimultaneously with pressure [Ca-in-opx thermometryand Al-in-opx barometry using the formulations of Breyet al (1990)] to constrain their relative positions in themantle columnMinerals in some eclogites from Lac de Gras show dis-

tinct core and rim compositions or inhomogeneitiesTemperatures were not calculated for samples withspatially irresolvable inhomogeneous garnet compositions(SE01VR43479VR43480) For eclogites with spongy cpxcore compositions were used in the calculationTemperatures were calculated for both core and rim com-positions where applicable For all samples these

temperatures agree within the uncertainty of this thermo-meter (508C Brey amp Kolaquo hler 1990)Most equilibration temperatures for eclogites lie between

800 and10008C (Fig 9)The transition between the shallowand the deep lithospheric mantle layers in the central Lacde Gras area lies in this temperature interval (Griffin et al1999 Pearson et al 1999) Eclogites in this study lack thedistinct bimodal temperature distribution observed bySchmidberger et al (2007) but as in their study high-Caeclogites give higher average temperatures (10808C corre-sponding to a depth of 155 km if equilibrated along a 40mWm2 geotherm) than other eclogites types (9808Cdepth 135 km) the diamond-bearing sample (VR40345)gives thehighest temperature of the datasetHighNa2Ocon-tents in garnet from high-Ca eclogites are qualitativelyconsistent with higher equilibration pressures (Sobolev ampLavrentrsquoyev 1971) as are high K2O contents in coexistingcpx (Erlank ampKushiro1970) suggesting that high-Ca eclo-gites not only equilibrated at high temperature but also highpressure rather than representing a temperature excursionPyroxenites both opx-bearing and opx-free give somewhatlower average temperatures than eclogites (mean 9258C)and phlogopite-rich eclogites give the lowest temperatures(685 and 7698C respectively) The calcite-bearing eclogiteVR67112b gives a relatively low equilibration temperatureof 8558C

MEASURED ANDRECONSTRUCTEDWHOLE -ROCKSRationale and approachKimberlite-borne xenoliths sometimes show evidence forinfiltration of the host kimberlite (eg Taylor amp Neal1989 Barth et al 2001 2002b Heaman et al 2006) Wetherefore follow the standard approach of calculatingwhole-rock compositions from the compositions of the con-stituent phases weighted by their modal abundances Onlycore compositions were used as distinct rim compositionsmay point to late diffusion of elements from kimberlite-related melt veinsMany eclogitic whole-rock compositions reconstructed

with only garnet and cpx have negative Ti anomalies(not shown) For most of these samples rutile or ilmenitehas been identified as part of the mineral assemblage andthese phases should therefore be considered in the whole-rock reconstruction Rutile or ilmenite modes were esti-mated by assuming that the bulk-rock has no Ti anomalywhen normalized to a normal mid-ocean ridge basalt(N-MORB) composition and calculating the amount ofrutile required to erase theTi anomalyIt is of course possible that the eclogites had negative

or positive Ti anomalies relative to elements of similarcompatibility (Eu and Dy) Because we reconstructed

0

2

4

6

8

10

12volatile-rich

pyroxenite opx-bearing

pyroxenite opx-free

high-Ca

high-Mg

low-Mg

700750

800850

900950

10001050

11001150

12001250

Fre

quen

cy

Temperature

Fig 9 Histogram ofTKrogh (Krogh 1988) obtained at a preset pres-sure of 5 GPa except opx-bearing pyroxenites where TBKN wassolved simultaneously with PBKN (Brey amp Kolaquo hler 1990)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1857

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

whole-rock compositions from mineral core compositionskimberlite infiltration is unlikely to have caused fractiona-tion of Ti from Eu and Dy In a study in which rutilemodes were estimated from the differences between Ti inmeasured bulk-rocks and in bulk-rocks reconstructed withonly garnet and cpx only three of 19 eclogites recon-structed with rutile show Ti anomalies (Barth et al2002a) Even eclogite xenoliths with depleted Zr and Hfabundances that resemble those of modern island arcbasalts do not show Ti anomalies (Jacob amp Foley 1999)We therefore consider our approach to be a reasonablyrobust means of obtaining meaningful rutile modes wheresmall sample sizes preclude reliable determination of themodes of accessory phases by point-countingWhere rutilewas not recognized on thin-section scale or not analysedthe average eclogitic rutile composition was used for thereconstruction

UncertaintiesReconstructing whole-rock compositions in this way is arather crude method given the coarse grain size of theeclogites relative to the sample size and the non-uniformdistribution of the principal minerals on thin-sectionscale The uncertainty in mineral modes of medium- tocoarse-grained eclogites with grains45mm has been esti-mated by Jerde et al (1993) to be 10 plus uncertaintiesrelated to partial alteration which is minor in our samplesNevertheless the whole-rock trace-element budget israther insensitive to the exact modes for common eclogiticassemblages and REE patterns which encompass mostother elements in terms of compatibility in garnet andcpx do not change significantly even for modal variationsof 30 (Jerde et al 1993) This is not true for major-element concentrations which vary substantially with var-iations in mineral modes and we therefore do not calculatewhole-rock major-element contentsBecause the discussion that follows is based on trace-

element patterns and trends rather than absolute abun-dances whole-rock reconstruction based on mineralmodes is adequate We are confident that the uncertaintyinherent in this approach is not leading to significant mis-representation of eclogite whole-rock compositions becausethe trace-element patterns of the reconstructed whole-rocks show distinct differences (see below) that correlatewith the assignment to eclogite types based on garnetmajor-element composition and this is independent ofmineral mode Following Jerde et al (1993) in ElectronicAppendix 9 we show variations in trace-element patternfor one sample in each eclogite type for variations ingarnet and cpx abundance from 7030 to 3070 vol This shows that the trace-element patterns are generallyinsensitive to cpxgt ratios with regard to slopes andanomalies such as positive Sr Pb and Eu anomalies withthe exception of the positive Eu anomaly in high-Mgeclogite VR43479 which disappears if a high cpx mode is

assumed (not shown) Variations in rutile modes by 50(Electronic Appendix 9d) lead as would be expected tothe appearance of Ti anomalies but a rutile mode of 05instead of 1 vol does not erase the positive Nb anomalyin low-Mg eclogiteVR50909Because of the robustness of the relative trace-element

abundances in the reconstructed eclogites a garnet propor-tion of 065 and cpx proportion of 035 approximately themedian modal abundances of these minerals in volatile-free eclogites from Lac de Gras are assumed for eclogitesamples that are too small for determination of meaningfulmodes by point-counting

Within- and between-group variationsThe trace-element compositions of reconstructed whole-rocks are given in Table 4 N-MORB-normalized trace-element patterns of reconstructed whole-rocks areshown in Fig 10 The three eclogite groups have distincttrace element patterns that are particularly obvious forthe REE (Fig 11) Low-Mg eclogites have positive Pb andSr anomalies and smooth positive slopes in their MREEN

and HREEN LREEN are enriched relative to MREEN inthree samples (Fig 11a) Three samples have high Nbrelative to similarly compatible elements and comparedwith most other eclogites (Fig 10a) All high-Mg eclogitesdisplay more or less pronounced positive Eu and Sranomalies and most have positive Pb and negative Zrand P anomalies Their N-MORB-normalized REE pat-terns show shallow positive slopes from LaN to PrN orNdNand aweak negative slope from NdN to LuN (Fig11b)VR43479 is a low-Mg eclogite with regard to its garnet

Ca^Fe^Mg composition but has a trace-element patterntypical of high-Mg eclogites and will be discussed withthe latter Conversely sample YK1949 a high-Mg eclogitein terms of its garnet Ca^Fe^Mg compostion is more simi-lar to low-Mg eclogites in terms of its whole-rock trace-element pattern and is plotted in Figs 10a and 11aWhereas the patterns of most samples within the groupresemble each other the N-MORB-normalized incompati-ble-element abundances of the high-Ca eclogites vary bymore than an order of magnitude (Fig 10c) All sampleshave positive Pb anomalies and most have positive Sranomalies Some samples have negative P andor Zranomalies Niobium abundances in three of the samplesare very high similar to those for some low-Mg eclogitesHigh-Ca eclogites tend to have flatter MREEN andHREEN patterns and more uniform HREE abundancesthat the other eclogite types with humps between SmN

and HoNThe phlogopitized sample YK1911 shows a strong posi-

tive Pb anomaly and lower HREE abundances than thecalcite-bearing sample VR67112b which shows a distinctnegative Eu anomaly and a depletion of La relative to Ce(Figs 10d and 11d)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1858

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Trace-element compositions of reconstructed whole-rocks (ppm)

Sample P Sc Ti V Co Ni Ga Sr Y Zr Nb La Ce Pr Nd

Low-Mg eclogites

vr19674-g4 186 50 4649 317 57 59 144 77 24 30 36 13 31 05 24

vr43452 154 39 2253 328 49 84 176 52 32 11 01 02 07 01 08

vr43469 38 65 4377 349 54 91 142 34 28 15 02 05 15 03 17

yk1926 144 45 3406 258 52 87 145 89 18 18 20 09 29 06 33

yk1943 173 46 5732 295 41 39 146 122 25 48 82 34 85 13 64

yk19491 102 34 1693 158 31 49 100 46 10 14 03 09 21 03 17

High-Mg eclogites

vr19677-g2 138 39 5160 291 41 53 119 98 14 22 04 13 53 11 64

vr434792 46 25 1265 129 36 126 84 56 3 7 04 06 25 04 23

vr67360 49 36 1573 138 64 184 111 76 5 9 04 09 38 07 33

yk1946 153 28 4877 192 45 105 124 108 12 48 03 10 38 08 50

High-Ca eclogites

vr19674-g5 267 30 7663 200 41 67 195 32 23 43 77 01 04 02 16

vr40345 0 33 2821 175 47 98 210 179 13 13 66 60 109 11 38

vr43465 142 27 5104 220 31 83 121 197 14 35 29 05 22 05 40

vr50858 295 30 6898 201 44 94 150 48 22 38 06 01 03 02 20

vr50860 68 57 5672 418 57 71 186 109 19 11 02 02 13 03 26

vr50909 238 28 7295 201 42 92 171 48 20 57 13 01 04 02 21

yk3528 91 51 3712 341 53 155 177 49 16 9 01 02 08 02 11

Volatile-rich eclogites

vr67112b 40 84 9242 521 46 30 107 4 40 66 31 26 126 24 125

yk1911 31 39 9428 286 44 82 164 3 23 52 64 18 52 09 46

Pyroxenites

yk1914 111 49 707 207 48 547 44 31 4 3 02 06 19 03 10

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

Low-Mg eclogites

vr19674-g4 11 05 10 22 39 10 29 33 06 12 05 028 24 008

vr43452 04 02 08 13 38 12 41 49 08 05 01 nd 02 na

vr43469 11 05 09 25 44 11 34 36 06 07 02 001 09 129

yk1926 12 05 11 19 30 07 22 24 04 07 03 004 23 090

yk1943 21 08 10 29 42 10 30 32 05 16 05 013 65 017

yk19491 06 03 11 08 14 04 14 18 03 04 02 003 29 115

High-Mg eclogites

vr19677-g2 20 08 12 21 24 05 15 15 03 05 05 007 52 6811

vr434792 05 02 12 05 06 01 04 03 00 03 01 001 98 037

vr67360 06 03 14 06 09 02 06 06 01 02 02 002 86 000

yk1946 18 07 12 21 22 05 13 12 02 11 03 006 56 112

High-Ca eclogites

vr19674-g5 17 09 11 33 43 09 24 23 04 16 05 007 02 011

vr40345 09 04 10 12 22 06 16 14 02 05 nd nd 298 na

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1859

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Continued

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

vr43465 22 09 12 25 26 06 15 14 02 10 04 001 25 040

vr50858 20 08 09 33 41 08 22 21 03 10 01 nd 02 na

vr50860 14 07 11 27 37 08 22 18 03 06 00 nd 06 na

vr50909 20 10 12 33 38 07 20 20 03 16 01 002 02 181

yk3528 08 05 11 20 30 07 18 18 02 05 01 nd 08 na

Volatile-rich eclogites

vr67112b 38 10 07 53 74 15 41 34 04 18 03 012 60 004

yk1911 20 07 08 35 39 07 18 17 02 10 14 nd 73 na

Pyroxenites

yk1914 02 01 11 04 08 02 05 06 01 01 03 003 68 101

aGarnet has CandashMgndashFe relationship typical of high-Mg eclogites but whole-rock trace-element pattern is typical oflow-Mg eclogitesbGarnet has CandashMgndashFe relationship typical of low-Mg eclogites but whole-rock trace-element pattern is typical ofhigh-Mg eclogites

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

(a)

001

01

1

10(c)

001

01

10(b)

VR43452

YK1949

RockNMORB

low-Mg

RockNMORB

high-Mg

(d)

volatile-richhigh-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

phlogopite-bearingYK1911

calcite-bearingVR67112b

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 10 Extended trace-element patterns of reconstructed (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-bearing eclogites Normalized to N-MORB of Sun amp McDonough (1989)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1860

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

RockNMORB

(b)

RockNMORB

(d)

010

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu001

01

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

(a)

001

01

1

10(b)

001

01

1

10(c)

001

01

1

10(e)

high-TSchmidberger et al

low-TSchmidberger et al

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

high-Mglow-Mg

high-Cavolatile-rich

phlogopite-bearingYK1911

calcite-bearingVR67112b

YK1949VR43479

Fig 11 N-MORB normalized REE patterns for (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-richeclogites Shown for comparison in (e) are reconstructed eclogites from Schmidberger et al (2007)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1861

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Both volatile-rich whole rocks display strong negativeSr anomalies Strontium can be sited in calcite or apatiteandor indicate the presence of a Sr-rich mineral that wasnot identified in thin section such as bariteFour of the six reconstructed whole-rocks in the study

of Schmidberger et al (2007) (Fig 11e) show markedLREEHREEN enrichment that is observed only in onehigh-Ca eclogite in our study (Fig 11c) suggesting thatmany of the eclogites they studied were affected by metaso-matic LREE enrichment Two samples in their study thatare not affected by this enrichment show positive Euanomalies similar to the high-Mg eclogites in this studywhereas the distinctive REE pattern of the low-Mg eclo-gites is not recognized among their samples

SR^ND^HF I SOTOPESYSTEMATICSStrontium Nd and Hf isotope data for optically puremineral separates and reconstructed whole-rocks aregiven in Table 5 Spongy rims and alteration are likely tobe mostly removed during leaching in 6N HCl in combina-tion with repeated ultrasonication Despite this rigoroustreatment the reproducibility of Nd in garnet VR43469and in cpx YK1914 is worse than for other samples Thismay be due to inhomogeneities related to late Nd diffusionpossibly from grain boundaries containing a kimberlitecomponent The reproducibility of the 176Hf177Hf data isgenerally poorer than that of Nd which may reflect a com-bination of matrix effects caused by imperfectly purifiedsolutions and sample heterogeneity Parent^daughterratios are available only from in situ analyses which arenot precise enough to allow meaningful model age orinitial isotope ratio calculations (ie at the time of kimber-lite eruption) Because the Lac de Gras kimberlites arerelatively young (55^56 Ma Graham et al 1999) age cor-rection would result in only a minor change in isotopecomposition which is inconsequential in view of the rangeof compositions displayed by most samples and hence hasno effect on the conclusions drawnClinopyroxenes in five volatile-free eclogites and a

pyroxenite have Sr isotope ratios ranging from 0701662 to0703546 this range is similar to or somewhat lower thanthat of the depleted mantle at the time of kimberliteemplacement (0703 Bell amp Tilton 2002) Clinopyroxenein the phlogopitized sample has a distinctly high 87Sr86Srof 0751186Clinopyroxene controls 56^92 of the Nd budget (Nd

concentration in cpx weighted by mode) Neodymiumisotope ratios range from 0511660 to 0514844 Thiscorresponds to eNd variation (per 10 000 deviationof 143Nd144Nd in the sample from 143Nd144Nd in the chon-dritic mantle model reservoir) from 17 to thorn43 cpx inthe low-Mg eclogites have higher eNd than those in

high-Mg eclogites Garnets have generally have higher143Nd144Nd from 0512984 (eNdfrac14thorn 7) to 0520965(eNdfrac14thorn 162) Garnets in the two volatile-rich sampleshave the highest ratios 0522790 for garnet in the calcite-bearing sample VR67112b (eNdfrac14thorn198) and 0525721 inthe phlogopitized eclogiteYK1911 (eNdfrac14thorn 255)Garnet contributes 18^32 and cpx 48^78 of the Hf

budget The proportion controlled by rutile is 4^30(Aulbach et al in preparation) Measured 176Hf177Hf incpx ranges from 0281263 in phlogopite-bearing sampleYK1911 (eHffrac1453) to 0286130 in low-Mg eclogiteVR43469 (eHffrac14thorn119) and in garnet from 0283958 inpyroxeniteYK1915 (eHffrac14thorn42) to 0295718 in low-Mg eclo-gite YK1926 (eHffrac14thorn458) Hafnium in garnet in the phlo-gopite^apatite-bearing sample is extremely radiogenic(176Hf177Hffrac14 0354367) corresponding to an eHf of thorn2532which is close to a value determined for a garnet separatefrom a Roberts Victor eclogite (eHffrac14thorn2561 Jacob et al2005) Garnet in the calcite-bearing sample has extremelyradiogenic Nd and Hf (eNdfrac14thorn193 eHffrac14thorn2516) Garnetin pyroxenite YK1915 also has radiogenic Nd and Hf(eNdfrac14thorn27 eHffrac14thorn42)Despite the large uncertainties for reconstructed whole-

rocks (calculated as the square root of the sum of squareduncertainties of two standard errors for garnet and cpxanalyses combined with 20 propagated uncertainty onthe proportion of Nd or Hf respectively contributed byeach mineral) the isotopic data are consistent in thatthey correlate with the independently determinedparent^daughter ratios (1) Low-Mg eclogites have higherSmNd reflected in the steeper positive slope of theirN-MORB-normalized REE patterns (Fig 11a) comparedwith the high-Mg eclogites (Fig 11b) Accordingly recon-structed low-Mg eclogites have higher present-day eNd

(0 to thorn47) than high-Mg eclogites (4 and 17) (2) Low-Mg eclogite YK1943 has very high Nb and LREE abun-dances that could have been added during metasomatismthat also enhanced its Hf abundance This sample has thelowest eNd and eHf of the low-Mg eclogites (0 and thorn20respectively) In contrast apparently unenriched low-Mgeclogites have more radiogenic Nd and Hf (eNdfrac14thorn47and thorn9 eHffrac14thorn128 and thorn113 respectively)The Sr^Nd isotope diagram reveals that eNd in cpx

varies from 19 to 43 at nearly constant depleted-mantle-like 87Sr86Sr (Fig 12a) In the Nd^Hf isotope diagram(Fig 12b) the reconstructed whole-rocks lie on curvedmixing lines if the lsquomixingrsquo proportions of Nd and Hffor the two lsquoend-memberrsquo minerals are very different(eg YK1911) Garnets have more radiogenic Nd andHf than coexisting cpx All reconstructed volatile-freeeclogites plot in the radiogenic field with regard to Hfor are isotopically similar to the depleted mantle Twohigh-Mg eclogites and one low-Mg eclogite have unradio-genic Nd one high-Mg eclogite has radiogenic Nd and

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1862

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 5 Individual and averaged Sr Nd and Hf isotope ratios obtained from garnet and cpx solutions

Sample 87Sr86Sr 2SE2SD 143Nd144Nd 2SE2SD eNd 2SE2SD 176Hf177Hf 2SE2SD eHf 2SE2SD

Low-Mg eclogites

VR43469 cp 0702661 0000010 0514844 0000006 43 01 0286130 0000012 119 04

VR43469 gt 0515508 0000014 56 03 0286506 0000012 132 04

VR43469 gt r 0515453 0000010 55 02 0286548 0000011 134 04

VR43469 gt r 0515384 0000007 54 01 0287244 0000010 158 04

VR43469 gt r 0515487 0000014 56 03 0286996 0000014 149 05

VR43469avg gt 0515458 0000109 55 21 0286823 0000715 143 253

WR VR43469 0515059 0000273 47 5 0286405 0001079 128 38

YK1926 cp 0703546 0000011 0513043 0000007 8 0 0283154 0000009 13 03

YK1926 gt 0513364 0000008 14 01 0295718 0000005 458 02

WR YK1926 0513120 0000026 9 1 0285980 0000716 113 25

YK1943 cp 0512143 0000007 10 01 0282213 0000005 20 02

YK1943 gt 0514461 0000012 36 02 0287510 0000008 168 03

YK1943 gt r 0514491 0000009 36 02 0287332 0000005 161 02

YK1943 gt r 0514517 0000010 37 02 0287290 0000007 160 02

YK1943avg gt 0514489 0000056 36 11 0287377 0000233 163 82

WR YK1943 0512625 0000228 0 5 0283138 0000879 13 31

High-Mg eclogites

VR43480 gt 0520965 0000017 162 03 0284950 0000010 77 04

VR67360 cp 0283509 0000074 26 26

VR67360 cp r 0283433 0000031 23 11

VR67360avg cp 0701662 0000018 0511660 0000010 19 02 0283471 0000108 25 38

VR67360 gt 0512984 0000101 7 20 0287429 0000180 165 64

WR VR67360 0511761 0000132 17 3 0284734 0000469 69 17

YK1949 cp 0703019 0000013 0512231 0000009 8 02 0282513 0000009 9 03

YK1949 gt 0513511 0000010 17 02 0290297 0000042 266 15

YK1949 gt r 0513508 0000012 17 02 0289947 0000015 254 05

YK1949 gt r 0513552 0000014 18 03

YK1949avg gt 0513523 0000049 17 10 0291569 0000494 311 175

WR YK1949 0512446 0000103 4 2 0284622 0000847 65 30

Volatile-rich eclogites

VR67112b gt 0522790 0000008 198 02 0354367 0000036 2532 13

YK1911 cp 0751186 0000009 0512165 0000007 9 01 0281263 0000009 53 03

YK1911 gt 0525721 0000013 255 03 0288051 0000012 187 04

WR YK1911 0513940 0000342 25 7 0281414 0000018 48 1

Pyroxenites

YK1914 cp 0703015 0000034 0513283 0000041 13 08

YK1914 cp r 0702956 0000005 0513200 0000027 11 05

YK1914 cp r 0513527 0000059 17 11

YK1914 cp r 0513045 0000010 8 02

YK1914avg cp 0702986 0000083 0513264 0000402 12 78

YK1915 gt 0514003 0000011 27 02 0283958 0000021 42

Analyses of full procedural replicates are designated with an lsquorrsquo after sample name SE standard error (internal errorsingle analyses) SD standard deviation (external error averaged multiple analyses) Italics used for reconstructed whole-rocks (mineral isotopic compositions and concentrations weighted by modes) Uncertainties for reconstructed whole-rocksinclude 2SE or 2SD uncertainties on each mineralrsquos isotope composition combined with 20 uncertainty in the amountof Nd or Hf contributed by the constituent minerals eNd and eHf are per 10 000 deviation from model chondritic mantle[model of Wasserburg et al (1981) and Blichert-Toft amp Albarede (1997) respectively]

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1863

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

εHf

εNd

Ekati eclogitesminus100

0

100

200

300

400

500

minus40 0 40 80 120 160 200 240 280

gt in YK1915

gt in VR67112bεNd=198 εHf=2532

minus40

0

40

80

120

160

minus40 minus20 0 20 40 60

minus60

minus40

minus20

0

20

40

070 071 072 073 074 075 076

North Americankimberlite field

εNd

87Sr86Sr

cpx in YK1914

Ekatieclogites

DM

DM

DM

DM

(a)

(b)

pyroxenites

low-Mg

high-Mg

volatile-rich

eclogites

Diavik low-TA154 S kimberlite

cpx in YK1911

Fig 12 (a) Sr^Nd for cpx and (b) Nd-Hf isotope diagram for garnet cpx and reconstructed whole-rocks (symbols with bold outline) Finedashed lines in (b) connect data points from the same sample with garnets having more radiogenic isotopic compositions Uncertainties forreconstructed whole-rocks combine 20 modal uncertainty with uncertainties on each mineralrsquos isotope composition Error bars on individualminerals may be smaller than the symbols Also shown are the isotopic compositions of the depleted mantle (DM DePaolo 1981 Griffin et al2000 Bell amp Tilton 2002) Range of 87Sr86Sr in eclogites from Ekati (Lac de Gras area 100 km north of Diavik Jacob et al 2003) in NorthAmerican kimberlites and compositions of low-temperature eclogites from Diavik and of Lac de Gras A154 South kimberlite fromSchmidberger et al (2007)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1864

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

one low-Mg eclogite has eNd similar to the depleted mantle(Fig 12b)Clinopyroxene in the phlogopitized eclogite YK1911

has unradiogenic Nd and Hf and strongly radiogenicSr whereas the coexisting garnet has radiogenic Nd andHf Although this sample contains 50 vol phlogopiteits whole-rock isotopic composition was calculated

from garnet and cpx alone because the contributionfrom phlogopite to the Hf and Nd budget is 52 and51 respectively (unpublished data) The whole-rockhas radiogenic Nd but has lower eHf than anyother sample analysed in accord with the extreme Hfabundance (53 ppm) and low LuHf of the cpx

DISCUSS IONOrigins of eclogites beneath the SlaveCratonHigh-Mg and high-Ca eclogites

With regard to their major- and trace-element contentsthe high-Mg and high-Ca eclogites from Lac de Gras arebroadly similar to variably evolved oceanic gabbros(Fig 13b and c) which may indicate an origin as subductedoceanic crust Compatible trace-element abundancesare useful to deduce potential crustal protoliths as theseelements are less susceptible to post-formation modification(Jacob 2004) The flat HREE patterns in the high-Mg andhigh-Ca eclogites are compatible with a low-pressureorigin as the oceanic crust is a product of moderate degreesof partial melting mainly within the spinel stability field(eg Presnall et al 2002) whereas melting with residualgarnet would qualitatively be expected to produce stronglyfractionated REE patterns Positive Sr and Pb anomaliesalthough compatible with a plagioclase signature couldalso be due to cryptic metasomatism (Dawson 1984) ofthe eclogites during their residence in the lithosphericmantle These anomalies are probably not related to kim-berlite infiltration as the whole-rock eclogite compositionswere reconstructed from analyses of fresh mineral coresLow ZrSm similar to that in primitive gabbros is

observed for many of the high-Mg and high-Caeclogites (Fig 13b and c) and is independent of the ratioof garnetcpx and whether or not the whole-rock wasreconstructed with rutile (Electronic Appendix 10) Thenegative trend of Mg-number against Yb in garnet(Fig 14) argues against igneous processes involving garnetaccumulation (Taylor amp Neal 1989) as suggested forexample for some high-Mg eclogites from West Africa(Barth et al 2002b) Rather the anti-correlation suggeststhat olivine fractionation played a role in the formation ofthe protoliths Fractionating olivine would excludeYb (Dolivinebasalt

frac14 002 Nikogosian amp Sobolev 1997)and incorporate Mg relative to Fe (Dfrac1415 and 05 respec-tively Taura et al1998) leading to progressive depletion ofMg and enrichment of Yb in the residual melt and henceanti-correlated Mg and Yb in the whole-rock This isinherited after eclogitization All eclogite types (exceptVR43452) plot along the olivine fractionation trend sug-gesting that they formed by a similar processHigh-Mg eclogites have lower

PREE than high-Ca

eclogites and some appear to have positive Eu anomalies

high-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

gabbrosSE Indian ridge

gabbrosOman ophiolite

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

high-Mg

RockNMORB(a)

VR43452

YK1949

low-Mg

boninite-like melts

01

001

1

10

(b)

01

001

1

10

(c)

01

001

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

Fig 13 Trace-element patterns of (a) low-Mg (b) high-Mg and(c) high-Ca eclogites compared with boninite-like volcanic rocksfrom the Isua greenstone belt (Polat et al 2002) gabbros from theOman ophiolite (Benoit et al 1996) and the SE Indian Ridge (Hartet al 1999) normalized to N-MORB of Sun amp McDonough (1989)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1865

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(EuEu up to 14) similar to more primitive cumulates(troctolite gabbros and olivine gabbros) reported in the lit-erature (Fig 13b) Although the sizes of these Eu anomaliesare not outside the analytical uncertainty their correlationwith CeYb in reconstructed eclogites supports their signif-icance (r2frac14066 nfrac14 4) The more primitive nature ofhigh-Mg eclogites is consistent with higher Mg-numbersrelative to high-Ca eclogites and may be due to olivineaccumulation during formation of their protolithsThe two samples with the lowest HREE abundanceshave the highest EuEu which is compatible with varia-tions in plagioclase modesMore evolved protoliths with higher pyroxeneplagio-

clase plus trapped melt could account for small or absentpositive Eu anomalies and higher

PHREE of high-Ca

eclogites (Fig 13c) Although the reason for the frequentassociation of graphite or diamond with high-Ca eclogitesis unclear (Aulbach et al 2003) the frequent occurrence ofkyanite in high-Ca eclogites also observed at Ekati justnorth of Lac de Gras has been argued to support a crustalprotolith (Jacob 2004) Kyanite andor corundum havebeen suggested to exsolve from high-temperature high-Alclinopyroxenes (Caporuscio amp Smyth 1990) Eclogitesfrom Lac de Gras show no exsolution textures suggestiveof such an origin although recrystallization can obscureor destroy microstructural evidence for exsolution(eg Griffin et al 1984)Modelling shows that qualitatively plagioclase accumu-

lation from a low-pressure melt derived from MORBmantle can account for positive Pb Sr and Eu and negativeZr anomalies relative to N-MORB whereas cpx accumula-tion or trapped melt increases Zr Y MREE and HREE

abundances (Fig 15) Some melt addition (for exampleby entrapment) seems to be required to increase all trace-element abundances to those observed for the eclogitesAlthough high-Ca and high-Mg eclogites could representshallow and deep portions of the same oceanic crust thehigher average temperatures obtained for the high-Caeclogites the frequent occurrence of carbon in high-Caeclogites and its absence in high-Mg eclogites the strongLREEHREE fractionation observed for six of seven high-Ca eclogites and the narrow range of compositions of thehigh-Mg eclogites suggest that they may have protolithsthat are not cogenetic Some discrepancies in particularaffecting the incompatible elements may be due to post-crystallization processes acting upon the protoliths suchas dehydration melting and later metasomatism that isevident in peridotites from the same locality (Griffin et al1999)Low K2OTh for high-Ca and high-Mg eclogites

(average 08 Table 4) relative to assumed crustal

0

2

4

6

8

10

30 40 50 60 70 80

Mg

VR43452

Yb

GARNET

pyroxenites

low-Mghigh-Mghigh-Ca

eclogites

VR19673ecl-g2

Fig 14 Garnet Mg-number vs Yb (ppm including 2s uncertainty)Lines qualitatively show separate correlations for eclogites (exceptoutlierVR43452) and pyroxenites (except outlierVR19673ecl-g2)

high-Ca

high-Mg

001

01

1

10

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

melt08 melt02 cryst cpx01 melt09 cryst

plag

(a)

001

01

1

10(b)

RockNMORB

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 15 Reconstructed (a) high-Ca and (b) high-Mg eclogitescompared with compositions of plagioclase and cpx crystallizingfrom a MORB melt modelled using distribution coefficients ofNorman et al (2005)The MORB was calculated as a 10 equilibriummelt of MORB pyrolite [composition from Kelemen et al (1993)except Pb Pr Sr Y and Lu which were calculated as residues from1 fractional melting of pyrolite from McDonough amp Sun (1995)using melt modes of Kinzler (1997) and distribution coefficients ofNorman et al (2005)] Also shown are mixtures of crystals (consistingof 50 plagioclase plus 50 cpx) and melt in the proportions 0901and 0208

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1866

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

precursors (eg gabbros SE Indian Ridge average 22Hart et al 1999) may reflect the preferred loss of fluid-mobile elements during dehydration (Becker et al 2000)although large uncertainties on these low-abundance ele-ments in the constituent minerals and reconstructedwhole-rocks makes this interpretation less than robustThe REE patterns of most high-Mg eclogites are too flatto represent melting residues as partial melting shouldproduce low LREEHREE (Electronic Appendix 10) Incontrast the marked LREE depletion of five of sevenhigh-Ca eclogites (LaLufrac14 02^08) contrasts with that ofassumed evolved gabbroic protoliths (LaLufrac1415^32Hart et al 1999) and can be modelled as a consequence ofpartial melting during subduction after eclogitization (iein the presence of garnet which retains HREE relative toLREE eg Barth et al 2002a) has taken place (Aulbachet al 2003)Our results are compatible with previous studies on

eclogite xenoliths from the Slave Craton includingthose from the same kimberlite that have linked thesesamples or some of these samples to Paleoproterozoicsubduction events (the 195^191Ga formation of theHottah terrane and the 18Ga formation of the GreatBear magmatic arc) in the Slave craton (Griffin et al1999 Kopylova et al 1999 Heaman et al 2002 2006Schmidberger et al 2005 2007)The compositional features discussed above although

compatible with crustal protoliths are not necessarilydefinitive of a subduction origin However if direct crystal-lization from basaltic melts at high pressure (ie mantledepths) is invoked instead specific conditions would haveto be met to satisfy the observations The Mg-number ineclogites from Lac de Gras are lower than those expectedof primary high-pressure mantle melts (eg Walter 1998OrsquoHara amp Herzberg 2002) requiring crystallization athigh pressures after olivine fractionation This originwould further require shielding from peridotitic wallrockwith which the fractionated melts would no longer be inequilibrium (Barth et al 2001) In such a setting the lowrelative abundances of fluid-mobile elements comparedwith those of less fluid-mobile elements and of LREErelative to HREE (Table 4) would reflect an expulsion offluids during crystallization of the protoliths and latermelt mobilization respectively The eclogites clearly donot represent garnet plus cpx cumulates which wouldrequire a positive correlation of Mg-number with Yb thatis at variance with the observed negative correlation ineclogitic garnets from this study (Fig 14)

Low-Mg eclogites

Garnets in low-Mg eclogites compositionally overlap withthose in massive eclogites from Jericho (NW of the Lacde Gras kimberlites) that were suggested to have a high-pressure intrusive origin based on the similarity of theirmajor-element compositions to those of Group A eclogites

fromsouthernAfrica forwhichamantleoriginhasbeenpro-posed (Kopylova et al1999) Garnets in low-Mg eclogites inthe present study have major-element compositions moresimilar to garnets in groupB rather thangroupAeclogitesThe positive slope in N-MORB-normalized HREE

abundances displayed by the low-Mg eclogites is clearlydistinct from other eclogite types and is similar to those ofboninite-like second-stage melts that are closely associatedwith ultramafic volcanics (Smithies et al 2004) (Fig 13a)The range of Mg-numbers in the low-Mg eclogites (55^66)overlaps with that of Archaean second-stage melts fromIsua (58^75) and Abitibi (52^82 Smithies et al 2004)These similarities might indicate that the low-Mg eclogiteshave an origin in a previously depleted mantle source suchas supra-subduction zone mantle Subchondritic ZrHf ofthe low-Mg eclogites contrasts with the supra-chondriticZrHf of the high-Mg and high-Ca eclogites (Aulbach et

al in preparation) and is consistent with previous cpx-controlled depletion in an arc mantle source This sourcecould have been subsequently remelted to generate theboninite-like precursor to the low-Mg eclogites as sug-gested for the protoliths of eclogites from Udachnaya(Jacob amp Foley 1999)Although some second-stage melts to which the low-Mg

eclogites are similar are closely associated with komatiitesand could contain a plume component the radiogenic ordepleted mantle-like Nd and Hf and unradiogenic Sr inthe low-Mg eclogites do not support plume involvement

Trace-element and Sr^Nd^Hf isotopeconstraints on post-formationmetasomatismSeveral low-Mg and high-Ca eclogites show negativeslopes in the LREE which may be indicative of interactionwith small amounts of LREE-enriched melt (Navon ampStolper 1987) amounting to cryptic metasomatism(ie not accompanied by crystallization of new phases) inthe nomenclature of Dawson (1984) High Nb abundancesrelative to elements of similar compatibility displayed bysome low-Mg and high-Ca eclogites (Fig 10) could resultfrom overestimation of rutile modes However high-CaeclogiteVR40345 is rutile-free and has been reconstructedwithout rutile its high Nb abundance is accompanied byLREE enrichment indicating that metasomatic additionmay be responsibleLow-Mg eclogites show a positive correlation of La with

Nb (Electronic Appendix 11a) indicating concomitantaddition of LREE and Nb This suggests that the metaso-matic agent was a solute-rich fluid or silicate silicic melt ashydrous fluids and carbonatites do not carry enough HFSEto affect the budget of eclogites (Green amp Wallace 1988Klemme et al 1995 Adam et al 1997 Stalder et al 1998)Of the samples that are enriched in Nb all but one (high-Ca eclogiteVR50909) show a simultaneous enrichment in

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1867

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Hf (Electronic Appendix 11b) Together these trace-element relationships suggest that both SmNd and LuHfwere lowered in these samples Their low parent^daughterratios contrast with higher ratios in unenriched samplesand resulted in the ingrowth of a range of Nd and Hf iso-tope compositions Compared with the low-Mg and high-Ca eclogites the high-Mg eclogites have a narrow range ofLa and Nb concentrations which correlates with a narrowrange in Nd and Hf isotope compositions (Fig 12)Precise model ages cannot be calculated because parent^

daughter ratios are available only from in situ analyseswhich have relatively large uncertainties Howeverevidence for long-term enrichment and depletion may bediscerned Low-Mg eclogites even those that are enrichedin LREE have relatively high SmNd as a result of havingoverall positive slopes in their REE patterns relative toMORB whereas the high-Mg eclogites have relativelyflat REE patterns and therefore lower SmNd (Fig 11)This translates into lower eNd for high-Mg eclogites thanfor low-Mg eclogites In contrast they show no systematicdifference with respect to 176Hf177Hf Because the low- andhigh-Mg eclogites are proposed to have different origins asdiscussed above and therefore did not necessarily have thesame initial isotopic composition and formation age theseisotopic differences cannot be used to estimate their agesA long-term coupling of trace-element and isotopic

abundances is also indicated by the fact that low-Mg eclo-gite YK1943 which has experienced Nb and LREE addi-tion has lower eNd (0) and eHf (thorn13) than low-Mg eclogiteVR43469 with higher LuHf and SmNd (eNdfrac14thorn47eHffrac14thorn128) If low-Mg eclogites YK1943 and VR43469originally had identical initial Nd and Hf isotope composi-tions and if the parent^daughter ratios are accurate andwere modified only once during their evolution shortlyafter metamorphism (when their isotopic compositionwould have been reset) this overprint can be calculatedfor both isotopic systems to have occurred at 17GaThis would suggest that Hf and Nd were added duringthe same event In contrast high-Mg eclogites VR67360and YK1949 would have had identical Ndisotope compositions at 10Ga whereas their Hf isotopecompositions are disturbed in that the sample with thehigher parent^daughter ratio has a lower 176Hf177Hf thanthe sample with the lower parent^daughter ratio Theseresults suggest that the LREE and HFSE enrichmentobserved in different samples occurred during separateevents that did not affect the eclogites equallyExtreme HFSE enrichment leading to the growth of

zircon and rutile and LREE addition leading to thegrowth of apatite in eclogites from the Jericho kimberlitein the northern Slave craton have been ascribed to twoseparate metasomatic events at around 18Ga (quasi-coincident with the age of formation and metamorphism)and 13^10Ga respectively (Heaman et al 2002 2006

Schmidberger et al 2005) The authors associated theseevents with Paleoproterozoic subduction and aMesoproterozoic thermal disturbance in the Slave litho-spheric mantle possibly linked to the Mackenzie igneousevent Mesoproterozoic Nd model ages mostly between13 and 14Ga were reported for eclogites from thenearby Ekati kimberlites (Jacob et al 2003) Although wedo not observe zircon or apatite in volatile-free eclogitesfrom Lac de Gras and the age constraints obtained fromour data are weak they support a link between theseevents described for other eclogite populations and themetasomatic enrichment observed in our studyThe Sr isotope data in the present study are similar to

those of Schmidberger et al (2007) who suggested that thedepleted-mantle-like values are consistent with derivationof the protoliths from an oceanic mantle source However87Sr86Sr in volatile-free eclogites in the present study isremarkably constant considering their Nd and Hf isotopevariability (Fig 12) and the variability of trace-elementpatterns (Fig 10) Given the evidence for dehydration andmelting in many of the eclogites fractionation of highlyincompatible Rb from Sr and differential ingrowth ofradiogenic Sr would be expectedThis may indicate a rela-tively recent overprint and considering the similarityof 87Sr86Sr to present-day depleted mantle which is signif-icantly lower than 87Sr86Sr of North American kimber-lites a juvenile (asthenospheric) source

PyroxenitesPyroxenites from Lac de Gras often show relationshipsin their major-element contents that are different fromthose observed in the eclogites For example pyroxeniticgarnets show a negative correlation between CaO andNa2O contents whereas these oxides are positively corre-lated for garnets in eclogites (Fig 5) CaO is well correlatedwith C2O3 in pyroxenitic garnets but not in eclogitic gar-nets (Fig 4) Mg-number is positively correlated withYb abundance for pyroxenitic garnets but anti-correlatedfor eclogitic garnets (Fig 14) This suggests that the pyrox-enites in the present study are not related to the eclogitesby fractionation or accumulation processes as proposedfor the origin of some pyroxenites (Foley et al 2003)Pyroxenites have also been interpreted as the products

of crystal segregation in magma channels with alteredoceanic crust as the magma source this would explaintheir Eu anomalies and stable and radiogenic isotopecompositions (Pearson et al 1993 Becker 1996) As thehigh-pressure liquidus phases of most terrestrial basaltsare cpx and garnet (OrsquoHara amp Herzberg 2002) partialmelts of eclogites with basaltic composition wouldcrystallize cpx and garnet but not opx More than half ofthe pyroxenites from Lac de Gras contain opx suggestingthat they probably did not crystallize directly fromeclogite-derived melts

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1868

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Low-Cr websterites and pyroxenites could be crystalli-zation products of mafic melts that interacted with perido-titic mantle thus becoming more refractory The positivecorrelation of Yb and Mg-number in garnet would allowan origin by fractional crystallization from such meltsAlternatively they formed by reaction of silicic (possiblyeclogite-derived) melts with mantle peridotites(eg Kelemen et al 1993 Aulbach et al 2002) This agreeswith the similarity of hypothetical melts in equilibriumwith pyroxenitic garnets in most samples to eclogite-derived melts (Electronic Appendix 12) Niobium concen-trations in all hypothetical melts are much higher than ineclogite-derived melts regardless of whether or not rutileis assumed to be in the residue The same is true for LaCe and Zr concentrations in some samples which mayindicate that NbZr and LREE were added to thepyroxenites some time after their formation as discussedin the previous section for some eclogites Because mini-mum ages for pyroxenitic rutile with an unradiogenicHf isotope composition (calculated by projecting the176Hf177Hf to the model depleted mantle isotopic evolutioncurve assuming zero ingrowth) range to 178066Ga(Aulbach et al in preparation) it appears likely that thepyroxenites were not formed recently (eg during kimber-lite-related or precursory melt migration) and that some ofthe samples are at least MesoproterozoicGarnet pyroxeniteYK1952 consists of schlieren of equili-

brated garnet pyroxenite in a lsquomatrixrsquo of spongy cpxfine cpx^opx intergrowths embayed garnet and meltpatches (Fig 2f) Garnet in the pyroxenite schlieren hasvery low LREE and HFSE abundances Orthopyroxenein the matrix has distinctly higher Mg-number andCr-number and contains more CaO than opx from thepyroxenite schlieren whereas opx rims in the schlierenare compositionally similar to opx in the matrix The con-siderable compositional and textural disequilibrium inthis sample suggests that this is a young feature causedby reaction with fluids shortly before or at the time ofentrainmentAs a consequence of the small sample sizes of most of

the pyroxenites studied here a comprehensive isotopicdataset is not available Clearly more such data areneeded in addition to stable isotope data to obtain betterconstraints on the origins of this rock type Consideringthe microstructural and compositional variability ofthe pyroxenites and major-element relationships oppositeto those observed for eclogites it appears certain that pyr-oxenites form by a variety of processes some of which areunrelated to the formation of eclogites and their protoliths

Volatile-rich eclogitesVolatile-rich eclogites from Lac de Gras display the mostextreme Hf^Nd isotope systematics of all the analysedsamples Garnets in phlogopitized (YK1911) and in cal-cite-bearing eclogite (VR67112b) have the most radiogenic

Nd (eNdfrac14 2486 and 1927 respectively) Garnet inVR67112b has extremely high eHf (2516) Clinopyroxene inYK1911 has the lowest eHf measured for any mineralseparate from Lac de Gras (53) in accord with the veryhigh Hf abundance in its cpx (5 ppm) In contrastgarnet in this sample has radiogenic Hf (eHffrac14 136) Thisdiscrepancy is at least in part due to their relatively lowequilibration temperatures (7698C and 8558C respec-tively) which allowed the isotopic compositions of cpxand garnet to evolve separately at their low andhigh parent^daughter element ratios respectively forextended periods of time and possibly since formationAccordingly the cpx in YK1911 gives a minimum Hf ageof 27GaSlab-derived fluids could have been introduced during

the 27Ga collision of the western Slave Craton domainincluding the thick plume-derived deep lithosphericmantle layer and the eastern domain with its highlydepleted lithosphere This would also be consistent withthe observation that volatile-rich eclogites occur at shallowdepths in the mantle (TKroghfrac14 685^8558C correspondingto depths of 85^115 km along a 40 mWm2 geotherm)where the bulk of dehydration and production of solute-rich melts in a subducting slab might be expected to occur(Becker et al 2000 Hermann et al 2006) Their negativeEu anomalies might indicate that the fluid was derivedfrom the shallow portion of the slab that experiencedplagioclase fractionation The presence of carbonate inVR67112b requires a carbonated sourceHigh 87Sr86Sr in cpx in YK1911 is probably the conse-

quence of isotopic equilibration with high-RbSr phlogo-pite after long-term radiogenic ingrowth while Hfaddition led to retarded ingrowth of radiogenic HfThese systematics combined with radiogenic Nd appear

inconsistent with the involvement of sediments in the pet-rogenesis of this sample (eg Perfit et al 1980 Vervoortet al 1999) or a hydrous fluid such as those responsible forMARID metasomatism in the Kaapvaal craton (Konzettet al 1998)Whatever the origin of the fluids the apparentHf mobility suggests that they were not in equilibriumwith a rutile-bearing source and also that they weresolute-rich in nature (ie hydrous melts able to dissolveHFSE) and the relatively high HREE abundances of thevolatile-rich eclogites indicate a garnet-free source suchas amphibolite This allowed evolution of the sample atrelatively high time-integrated SmNd

SUMMARY AND CONCLUSIONS

(1) The Lac de Gras kimberlites carried abundant eclo-gite and pyroxenite xenoliths to the surface Threemain eclogite types are distinguished in terms ofmajor-element composition and trace element

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1869

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

patterns high-Mg high-Ca and low-Mg eclogitesThe garnets of all eclogite types have anti-correlatedMg-number and Yb pointing to olivine fractionationas a controlling factor in the formation of their pre-cursor rocks Their flat or positively sloped HREEpatterns qualitatively indicate that the melt parentalto the precursor rocks was not in equilibrium with agarnet-bearing residue

(2) Reconstructed high-Mg eclogites and high-Caeclogites have flat HREE patterns positive Pb andSr anomalies and low ZrSm similar to oceanic gab-bros reported in the literature High-Mg eclogiteshave variably low HREE contents and some havesmall positive Eu anomalies possibly suggestive ofmore primitive relatively plagioclase-rich protolithswhereas high-Ca eclogites have higher HREE abun-dances and small or absent Eu anomalies that requiremore cpx- andor melt-rich protoliths if a subductionorigin is assumed

(3) Low-Mg eclogites have distinctive trace-element pat-terns that show a marked positive slope in the HREEand subchondritic ZrHf They resemble someArchaean second-stage melts (lsquoboninitesrsquo) and theirprotoliths appear to have formed from a previouslydepleted source

(4) After formation the eclogites were subject to a varietyof processes Most have low ratios of fluid-mobile vsless fluid-mobile elements possibly as a result of dehy-dration of the protoliths In addition some of the low-Mg and high-Ca eclogites have low LREEHREEconsistent with partial melt lossWhereas the low-Mgeclogites show concomitant enrichment in La Nb andZr some high-Mg and high-Ca eclogites show evi-dence for variable and unsystematic (decoupled)LREE and Nb addition

(5) Isotopically the trace-element characteristics of thedifferent eclogite types translate into lower eNd forhigh-Mg eclogites (negative slope in MREEfrac14 lowerSmNd) than for low-Mg eclogites (positive slope inMREEfrac14higher SmNd) Within the low-Mg groupsamples that show evidence for metasomatism (lowLuHf and SmNd) have lower eNd and eHf than asample that was apparently not metasomatized BothNd and Hf addition can be dated to 17Ga if identi-cal initial isotope compsitions are assumed for thelow-Mg eclogites In contrast the relatively constantand depleted mantle-like 87Sr86Sr of eclogitic cpxdespite variable Nd and Hf isotope compositions andparent^daughter ratios suggests recent input of juve-nile material

(6) Garnet and cpx in pyroxenites show major-elementrelationships different from those in eclogitesTherefore these two rock types are probably notrelated by fractionation processes The presence of

opx precludes direct crystallization from slab-derivedmelts Pyroxenites may be the product of (slab-derived) melts hybridized during assimilation of peri-dotitic mantle

(7) Garnets in phlogopite- and carbonate-bearing eclo-gites have extreme isotopic compositions includingthe most radiogenic Nd and Hf in the datasetNegative Eu anomalies may point to the involvementof a crustal protolith Highly radiogenic Sr radiogenicNd and unradiogenic Hf in the reconstructedphlogopite-bearing sample combined with highHREE abundances and evidence for Hf mobilitymay indicate involvement of solute-rich fluids origi-nating in a garnet- and rutile-free source

ACKNOWLEDGEMENTSHelp in the clean laboratories and with the analyticalfacilities by S Elhlou C Lawson A Sharma P Wielandand M Zhang is greatly appreciated Bill Griffin isthanked for his support and for agreeing to reviewseveral versions of the manuscript despite disagreeingwith many of the conclusions presented here The paperbenefited greatly from the detailed and constructivereviews provided by the editors Gareth Davies andMarjorie Wilson and by Matthias Barth Larry Heamanand an anonymous reviewer This work was funded by aMacquarie University International PostgraduateAward and Postgraduate Research Fund (SA) by anARC SPIRT grant sponsored by Kennecott Canada Incand by ARC Large and Discovery Grants to SYOrsquoRAnalytical data were obtained using instrumentationfunded by ARC LIEF and DEST Systemic InfrastructureGrants industry partners and Macquarie UniversityThis is Publication 481 from the ARC National KeyCentre for Geochemical Evolution and Metallogeny ofContinents (wwwesmqeduauGEMOC)

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online

REFERENCESAdam J GreenT H Sie S H amp Ryan C G (1997) Trace element

partitioning between aqueous fluids silicate melts and mineralsEuropeanJournal of Mineralogy 9 569^584

Aulbach S Stachel T Viljoen K S Brey G P amp Harris J W(2002) Eclogitic and websteritic diamond sources beneath theLimpopo Beltccedilis slab-melting the link Contributions to Mineralogy

and Petrology 143 56^70Aulbach S GriffinW L Pearson N J OrsquoReilly S Y Kivi K amp

Doyle B (2003) Origins of eclogites beneath the central SlaveCraton Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA-0011

Aulbach S Griffin W L Pearson N J OrsquoReilly S Y amp Kivi K(2004a) Mantle formation and evolution Slave Craton constraints

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1870

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

from HSE abundances and Re^Os isotope systematics of sulfideinclusions in mantle xenocrysts Chemical Geology 208 61^88

Aulbach S Griffin W L OrsquoReilly S Y amp McCandless T E(2004b) Genesis and evolution of the lithospheric mantle beneaththe Buffalo HeadTerrane Alberta (Canada) Lithos 77 413^451

Aulbach S GriffinW L Pearson N J OrsquoReilly S Y amp Doyle B(2007) Lithosphere formation in the central Slave Craton(Canada) plume subcretion or lithosphere accretion Contributionsto Mineralogy and Petrology (in press) doi 101007s00410-007-0200-1

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2001)Geochemistry of xenolithic eclogites from West Africa Part IA link between low MgO eclogites and Archean crust formationGeochimica et Cosmochimica Acta 65 1499^1527

Barth M G Foley S F amp Horn I (2002a) Partial melting inArchean subduction zones constraints from experimentally deter-mined trace element partition coefficients between eclogitic miner-als and tonalitic melts under upper mantle conditions PrecambrianResearch 113 323^340

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2002b)Geochemistry of xenolithic eclogites from West Africa part 2Origins of the high MgO eclogites Geochimica et Cosmochimica Acta

66 4325^4345Beard B L Fraracci K NTaylor L A Snyder G A Clayton R

A Mayeda T K amp Sobolev NV (1996) Petrography andgeochemistry of eclogites from the Mir kimberliteYakutia RussiaContributions to Mineralogy and Petrology 125 293^310

Becker H (1996) Geochemistry of garnet peridotite massifs fromlower Austria and the composition of deep lithosphere beneatha Palaeozoic convergent plate margin Chemical Geology 134 49^65

Becker H Jochum K P amp Carlson R W (2000) Trace elementfractionation during dehydration of eclogites from high-pressureterranes and the implications for element fluxes in subductionzones Chemical Geology 163 65^99

Bell K amp Tilton G R (2002) Probing the mantle The story fromcarbonatites EOS Transactions American Geophysical Union 83275^277

Benoit M Polvecurren M amp Ceuleneer G (1996) Trace element andisotopic characterisation of mafic cumulates in a fossil mantlediapir (Oman ophiolite) Chemical Geology 134 199^214

BleekerW (2003) The late Archean record a puzzle in ca 35 piecesLithos 71 99^134

BleekerW Ketchum JW F JacksonV A amp Villeneuve M (1999)The central Slave Basement Complex Part I Its structural topol-ogy and autochthonous core Canadian Journal of Earth Sciences 361083^1109

Blichert-Toft J amp Albare de F (1997) The Lu^Hf isotope geochemis-try of chondrites and the evolution of the mantle^crust systemEarth and Planetary Science Letters 148 243^258

Blichert-Toft J Chauvel C amp Albare de F (1997) Separation of Hfand Lu for high precision isotope analysis of rock samples by mag-netic sector-multiple collector ICP-MS Contributions to Mineralogy

and Petrology 127 248^260Brey G P amp Kolaquo hler T (1990) Geothermobarometry in four-phase

lherzolites II New thermobarometers and practical assessment ofexisting thermobarometers Journal of Petrology 31 1353^1378

Brey G P Kolaquo hlerT amp Nickel K G (1990) Geothermobarometry infour-phase lherzolites I Experimental results from 10 to 60 kbJournal of Petrology 31 1313^1352

Caporuscio F amp Smyth J (1990) Trace element crystal chemistryof mantle eclogites Contributions to Mineralogy and Petrology 105550^561

Carbno G B amp Canil D (2002) Mantle structure beneath the SWSlave Craton Canada constraints from garnet geochemistry in theDrybones Bay Kimberlite Journal of Petrology 43 129^142

Coleman R G Lee E D Beatty L B amp Brannock W W (1965)Eclogites and eclogites their differences and similaritiesGeological Society of America Bulletin 76 483^508

Creaser R A Grulaquo tter H Carlson J amp Crawford B (2004)Macrocrystal phlogopite Rb^Sr dates for the Ekati propertykimberlites Slave Province Canada evidence for multiple intru-sive episodes in the Paleocene and Eocene Lithos 76 399^414

Davies R M GriffinW L Pearson N J Andrew A S Doyle BJ amp OrsquoReilly S Y (1999) Diamonds from the deep Pipe DO-27Slave Craton Canada In Gurney J J Gurney J L Pascoe MD amp Richardson S H (eds) Proceedings of the 7th International

Kimberlite Conference CapeTown Red Roof Design pp 148^155Davies R M Griffin W L OrsquoReilly S Y amp Doyle B J (2004)

Mineral inclusions and geochemical characteristics of microdia-monds from the DO27 A154 A21 A418 DO18 DD17 and RanchLake kimberlites at Lac de Gras Slave Craton Canada Lithos 7739^55

DavisW J amp Hegner E (1992) Neodymium isotopic evidence for thetectonic assembly of Late Archean crust in the Slave Provincenorthwest Canada Contributions to Mineralogy and Petrology 111493^504

Davis W J Jones A G Bleeker W amp Grutter H (2003)Lithosphere development in the Slave craton a linked crustal andmantle perspective Lithos 71 575^589

Dawson J B (1984) Contrasting types of upper-mantle metasoma-tism In Kornprobst J (ed) Kimberlites II The Mantle and

Crust^Mantle Relationships Amsterdam Elsevier pp 289^294DePaolo D J (1981) Neodymium isotopes in the Colorado Front

Range and crust^mantle evolution in the Proterozoic Nature 291193^196

Erlank A J amp Kushiro I (1970) Potassium contents of syntheticpyroxenes at high temperatures and pressures Carnegie Institution ofWashingtonYearbook 68 433^439

Foley S F Barth M G amp Jenner G A (2000) Rutilemelt partitioncoefficients for trace elements and an assessment of the influence ofrutile on the trace element characteristics of subduction zonemagmas Geochimica et Cosmochimica Acta 64 933^938

Foley S F Buhre S amp Jacob D E (2003) Evolution of the Archaeancrust by delamination and shallow subduction Nature 421 249^252

Graham I Burgess J L Bryan D Ravenscroft P J Thomas EDoyle B J Hopkins R amp Armstrong K A (1999) Explorationhistory and geology of the Diavik Kimberlites Lac de GrasNorthwest Territories Canada In Gurney J J Gurney J LPascoe M D amp Richardson S H (eds) Proceedings of the 7th

International Kimberlite Conference Cape Town Red Roof Designpp 262^279

Green D H amp Wallace M E (1988) Mantle metasomatism byephemeral carbonatite melts Nature 336 459^462

Green T H Blundy J D Adam J amp Yaxley G M (2000) SIMSdetermination of trace element partition coefficients betweengarnet clinopyroxene and hydrous basaltic liquids at 2^75 GPaand 1080^12008C Lithos 53 165^187

GriffinW L amp OrsquoReilly SY (2007) Cratonic lithospheric mantle Isanything subducted Episodes 30(1) 43^53

Griffin W L Wass S Y amp Hollis J D (1984) Ultramafic xenolithsfrom Bullenmerri and Gnotuk maars Victoria Australia petrol-ogy of a sub-continental crust^mantle transition Journal of

Petrology 25 53^87GriffinW L Doyle B J Ryan C G Pearson N J OrsquoReilly SY

Davies R Kivi K van Achterbergh E amp Natapov LM (1999)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1871

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Layered mantle lithosphere in the Lac de Gras Area SlaveCraton composition structure and origin Journal of Petrology 40705^727

Griffin W L Pearson N J Belousova E Jackson S E vanAchterbergh E OrsquoReilly SY amp Shee S R (2000)The Hf isotopecomposition of cratonic mantle LAM-MC-ICPMS analysis ofzircon megacrysts in kimberlites Geochimica et Cosmochimica Acta

64 133^147Griffin W L OrsquoReilly S Y Doyle B J Pearson N J

Coopersmith H Kivi K Malkovets V amp Pokhilenko N(2004) Lithosphere mapping beneath the North American plateLithos 77 873^922

Grulaquo tter H S Apter D B amp Kong J (1999) Crust^mantle couplingevidence from mantle-derived xenocrystic garnets In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 307^312

Hart S R Blusztajn J Dick H J B Meyer P S ampMuehlenbachs K (1999) The fingerprint of seawater circulationin a 500-meter section of ocean crust gabbros Geochimica et

Cosmochimica Acta 63 4059^4080Harte B amp Kirkley M B (1997) Partitioning of trace elements

between clinopyroxene and garnet Data from mantle eclogitesChemical Geology 136 1^24

Heaman L M Creaser R A amp Cookenboo H O (2002) Extremeenrichment of high field strength elements in Jericho eclogitexenoliths A cryptic record of Paleoproterozoic subduction partialmelting and metasomatism beneath the Slave craton CanadaGeology 30 507^510

Heaman L M Kjarsgaard B A amp Creaser R A (2004) The tem-poral evolution of North American kimberlites Lithos 76 377^397

Heaman L M Creaser R A Cookenboo H O amp Chacko T(2006) Multi-stage modification of the northern Slave mantlelithosphere Evidence from zircon- and diamond-bearing eclogitexenoliths entrained in Jericho Kimberlite Journal of Petrology 47821^858

Hermann J Cpandler C Hack A amp Korsakov A (2006) Aqueousfluids and hydrous melts in high-pressure and ultra-high pressurerocks implications for element transfer in subduction zones Lithos92 399^417

Hills DV amp Haggerty S E (1989) Petrochemistry of eclogites fromthe Koidu Kimberlite Complex Sierra Leone Contributions to

Mineralogy and Petrology 103 397^422Hoffman P F (1989) Precambrian geology and tectonic history of

North America In Bally A amp Palmer A (eds) The Geology of

North AmericaccedilAn Overview Boulder CO Geological Society ofAmerica pp 447^512

IrelandT R Rudnick R L amp Spetsius Z (1994) Trace-elements indiamond inclusions from eclogites reveal link to Archean granitesEarth and Planetary Science Letters 128 199^213

Irvine G J Pearson D G Kjarsgaard B A Carlson R WKopylova M G amp Dreibus G (2003) Evolution of the litho-spheric mantle beneath Northern Canada a Re^Os isotope andPGE study of kimberlite-derived peridotite xenoliths fromSomerset Island and a comparison to the Slave and Kaapvaalcratons Lithos 71 461^488

Isachsen C E amp Bowring S A (1994) Evolution of the Slave cratonGeology 22 917^920

Jacob D E (2004) Nature and origin of eclogite xenoliths from kim-berlites Lithos 77 295^316

Jacob D E amp Foley S F (1999) Evidence for Archean ocean crustwith low high field strength element signature from diamondifer-ous eclogite xenoliths Lithos 48 317^336

Jacob D Jagoutz E Lowry D Mattey D amp Kudrjavtseva G(1994) Diamondiferous eclogites from Siberiaccedilremnants ofArchean oceanic-crust Geochimica et Cosmochimica Acta 585191^5207

Jacob D E Fung A Jagoutz E amp Pearson D G (2003) Petrologyand geochemistry of eclogite xenoliths from the Ekati kimberlitesarea Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA_239

Jacob D E Bizimis M amp Salters V J M (2005) Lu^Hf and geo-chemical systematics of recycled ancient oceanic crust evidencefrom Roberts Victor eclogites Contributions to Mineralogy and

Petrology 148 707^720Jerde E A Taylor L A Crozaz G Sobolev NV amp SobolevV N

(1993) Diamondiferous eclogites from Yakutia Siberiaccedilevidencefor a diversity of protoliths Contributions to Mineralogy and Petrology

114 189^202Jones A G Ferguson I J Chave A D Evans R L amp

McNeice GW (2001) The electric lithosphere of the Slave cratonGeology 29 423^426

Kelemen P B Shimizu N amp Dunn T (1993) Relative depletionof niobium in some arc magmas and the continental-crust-partitioning of K Nb La and Ce during meltrockreaction in the upper-mantle Earth and Planetary Science Letters 120111^134

Kinzler R J (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition Application to mid-ocean ridge basalt petrogenesis Journal of Geophysical Research 102853^874

Klemme S van der Laan S R Foley S F amp Guenther D (1995)Experimentally determined trace and minor element partitioningbetween clinopyroxene and carbonatite melt under upper mantleconditions Earth and Planetary Science Letters 133 439^448

Konzett J Armstrong R A Sweeney R J amp CompstonW (1998)The timing of MARID metasomatism in the Kaapvaal mantle anion probe study of zircons from MARID xenoliths Earth and

Planetary Science Letters 160 133^145Kopylova M G Russell J K amp Cookenboo H (1999) Mapping the

lithopshere beneath the north central Slave Craton In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 468^479

Krogh E (1988) The garnet^clinopyroxene iron^magnesiumgeothermometerccedila reinterpretation of existing experimentaldata Contributions to Mineralogy and Petrology 99 44^48

Kusky T M (1989) Accretion of Archean Slave Province Geology 1763^67

LeCheminant A N amp Heaman L M (1989) Mackenzie igneousevents Canada Middle Proterozoic hotspot magmatismassociated with ocean opening Earth and Planetary Science Letters

96 38^48LeCheminant A N Richardson D G DiLabio R N W amp

Richardson K A (1996) La recherche de diamants au CanadaGeological Survey of Canada Open File 3228 268pp

MacGregor I D amp Carter J L (1970) The chemistry of clinopyrox-enes and garnets of eclogite and peridotite xenoliths from theRoberts Victor mine South Africa Physics of the Earth and PlanetaryInteriors 3 391^397

MacGregor I D amp Manton W I (1986) Roberts Victor eclogitesancient oceanic crust Journal of Geophysical Research 91 14063^14079

MacKenzie J M amp Canil D (1999) Composition and thermal evo-lution of cratonic mantle beneath the central Archean SlaveProvince NWT Canada Contributions to Mineralogy and Petrology

134 313^324

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1872

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

McCandlessT E amp Gurney J J (1989) Sodium in garnet and potas-sum in clinopyroxene criteria for classifying mantle eclogites InRoss J (ed) Kimberlites and Related Rocks Carlton Blackwellpp 827^832

McDonoughW F amp Sun S-S (1995) The composition of the EarthChemical Geology 120 223^253

Navon O amp Stolper E (1987) Geochemical consequences of meltpercolation the upper mantle as a chromatographic columnJournal of Geology 95 285^307

Nikogosian I K amp Sobolev AV (1997) Ion-microprobe analysis ofmelt inclusions in olivine experience in estimating the olivine^meltpartition coefficients of trace elements Geochemistry International 35119^126

Norman M D Griffin W L Pearson N J Garcia M O ampOrsquoReilly S Y (1998) Quantitative analysis of trace element abun-dances in glasses and minerals a comparison of laser ablationinductively coupled plasma mass spectrometry solution inductivelycoupled plasma mass spectrometry proton microprobe and elec-tron microprobe data Journal of Analytical Atomic Spectrometry 13477^482

Norman M D Garcia M O amp Pietruszka A J (2005) Trace-element distribution coefficients for pyroxenes plagioclase andolivine in evolved tholeiites from the 1955 eruption of KilaueaVolcano Hawairsquoi and petrogenesis of differentiated rift-zonelavas American Mineralogist 90 888^899

OrsquoHara M J amp Herzberg C (2002) Interpretation of trace elementand isotope features of basalts Relevance of field relations petrol-ogy major element data phase equilibria and magma chambermodeling in basalt petrogenesis Geochimica et Cosmochimica Acta 662167^2191

PadghamW A amp FysonW K (1992) The Slave Province a distinctcraton CanadianJournal of Earth Sciences 29 2072^2086

Pearson D G Davies G R amp Nixon P H (1993) Geochemical con-straints on the petrogenesis of diamond facies pyroxenites from theBeni Bousera peridotite massif North Morocco Journal of Petrology34 125^172

Pearson N J Griffin W L Doyle B J OrsquoReilly S Y vanAchterbergh E amp Kivi K (1999) Xenoliths from kimberlitepipes of the Lac de gras area Slave Craton Canada In GurneyJ J Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite ConferenceCapeTown Red Roof Design pp 644^658

Perfit N R Gust D A Bence A E Arculus R J amp Taylor S R(1980) Chemical characteristics of island-arc basalts implicationsfor mantle sources Chemical Geology 30 227^256

Polat A Hofmann A W amp Rosing M T (2002) Boninite-likevolcanic rocks in the 37^38 Ga Isua greenstone belt WestGreenland geochemical evidence for intra-oceanic subductionzone processes in the early Earth Chemical Geology 184 231^254

Poudjom DjomaniY H GriffinW L OrsquoReilly SY amp Doyle B J(2005) Lithospheric domains and controls on kimberlite emplace-ment Slave Province Canada evidence from elastic thickness andupper mantle composition Geochemistry Geophysics Geosystems 6doi1010292005GC0009978

Presnall D C Gudfinnsson G H amp Walter M J (2002)Generation of mid-ocean ridge basalts at pressures from 1 to 7GPa Geochimica et Cosmochimica Acta 66 2073^2090

Schmidberger S S Heaman L M Simonetti A Creaser R A ampCookenboo H O (2005) Formation of Paleoproterozoiceclogitic mantle Slave Province (Canada) Insights from in-situHf and U^Pb isotopic analyses of mantle zircons Earth and

Planetary Science Letters 240 621^633

Schmidberger S S Simonetti A Heaman L M Creaser R A ampWhiteford S (2007) Lu^Hf in-situ Sr and Pb isotope and traceelement systematics for mantle eclogites from the Diavik diamondmine Evidence for Paleoproterozoic subduction beneath the Slavecraton Canada Earth and Planetary Science Letters 254 55^68

Schulze D J amp Helmstaedt H (1988) Coesite-sanidine eclogitesfrom kimberlite products of mantle fractionation or subductionJournal of Geology 96 435^443

Smithies R H Champion D C amp Sun S S (2004) The case forArchaean boninites Contributions to Mineralogy and Petrology 147705^721

Smyth J R Caporuscio F A amp McCormick T C (1989) Mantleeclogites evidence of igneous fractionation in the mantle Earth andPlanetary Science Letters 93 133^141

Sobolev NV amp Lavrentrsquoyev Y G (1971) Isomorphic sodium admix-ture in garnets formed at high pressure Contributions to Mineralogy

and Petrology 31 1^12Stalder R Foley S F Brey G P amp Horn I (1998) Mineral aqueous

fluid partitioning of trace elements at 900^12008C and 30^57GPa New experimental data for garnet clinopyroxene and rutileand implications for mantle metasomatism Geochimica et

Cosmochimica Acta 62 1781^1801Sun S-S amp McDonough W F (1989) Chemical and isotopic sys-

tematics of oceanic basalts implications for mantle compositionsand processes In Saunders A D amp Norris M J (eds)Magmatism in the Ocean Basins Geological Society London Special

Publications 42 313^345Taura H Yurimoto H Kurita K amp Sueno S (1998) Pressure

dependence on partition coefficients for trace elements between oli-vine and the coexisting melts Physics and Chemistry of Minerals 25469^484

Taylor L A amp Neal C R (1989) Eclogites with oceanic crustal andmantle signatures from the Bellsbank kimberlite South AfricaPart I Mineralogy petrography and whole rock chemistryJournal of Geology 97 551^567

van Achterbergh E Ryan C G amp Griffin W L (1999) Glitter online intensity reduction for the laser ablation inductively coupledplasma spectrometry Abstract 9th V M GoldschmidtConference LPI Boston MA 905 pp Abstract no 7215

van Breemen O DavisW J amp King J E (1992) Temporal distribu-tion of granitoid plutonic rocks in the Archean Slave ProvinceNorthwest Canadian Shield Canadian Journal of Earth Sciences 292186^2199

Vervoort J D Patchett P J Blichert-Toft J amp Albare de F (1999)Relationship between Lu^Hf and Sm^Nd isotopic systems in theglobal sedimentary system Earth and Planetary Science Letters 16879^99

Viljoen K S Smith C B amp Sharp Z D (1996) Stable and radio-genic isotope study of eclogite xenoliths from the Orapa kimberliteBotswana Chemical Geology 131 235^255

Walter M J (1998) Melting of garnet peridotite and theorigin of komatiite and depleted lithosphere Journal of Petrology 3929^60

Wasserburg G J Jacobsen S B DePaolo D J McCulloch M Tamp Wen T (1981) Precise determination of SmNd ratios Sm andNd isotopic abundances in standard solutions Geochimica et

Cosmochimica Acta 45 2311^2323Westerlund K J Shirey S B Richardson S H Carlson R W

Gurney J J amp Harris JW (2006) A subduction wedge origin forPaleoarchean peridotitic diamonds and harzburgites from thePanda kimberlite Slave craton evidence from Re^Os isotope sys-tematics Contributions to Mineralogy and Petrology 152 275^294

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1873

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Page 3: Origins of Xenolithic Eclogites and Pyroxenites from the Central

parts of the craton have been recognized (Griffin et al1999 2004 Grulaquo tter et al 1999 Kopylova et al 1999MacKenzie amp Canil 1999 Pearson et al 1999 Carbno ampCanil 2002) which have been substantiated by magnetotel-luric and elastic thickness data (Jones et al 2001 PoudjomDjomani et al 2005) Here we focus on our currentknowledge of the mantle evolution in the central SlaveCraton 340 km NE from Yellowknife NT where the Lacde Gras kimberlites were intruded The subcontinentallithospheric mantle (SCLM) sampled by the Lac deGras kimberlites is strongly layered with an ultra-depletedshallow layer and a less depleted deep layer the latter sug-gested to have formed by subcretion of a plume head thatdelivered diamonds containing lower mantle inclusions(Griffin et al 1999 2004 Davies et al 1999 Aulbach et al2007) A Re^Os isochron age of 327034Ga for somesulfides from the deep layer may date the formationof this layer and shows that significantly oldermantle resides beneath some of the 27Ga crust of thejuvenile eastern domain (Contwoyto Terrane) in the SlaveCraton (Aulbach et al 2004a) This isochron ageagrees within uncertainty with one derived from sulfideinclusions in diamonds from the Panda kimberlite(Westerlund et al 2006) just north of Lac de Gras and isconsistent with previous findings that many kimberliteslying in the younger eastern domain appear to havesampled older lower crust and mantle indicative of aneast-dipping trans-lithospheric boundary dating backto 27Ga craton amalgamation (Grulaquo tter et al 1999Irvine et al 2003)

SAMPLES AND ANALYTICALTECHNIQUESAll samples of the present study were retrieved fromdrill core of kimberlite pipe A154S in the Lac de Grasarea from depths between 100 and 470m Samples ofat least 1cm in size were cut out of the drill core andprocessed to minimize sampling bias although thispossibly discriminates against more friable eclogitetypes that are not as well preserved Eighteen eclogitesand pyroxenites were selected for the present study andresults combined with those of Pearson et al (1999and unpublished data) to make up a total of 35 eclogitesand 30 pyroxenitesMajor-element analyses were obtained using a

CAMECA Camebax SX50 electron microprobeIn situ trace-element compositions were determinedwith a custom-built laser-ablation system (designed byS E Jackson) or a Merchantek LUV 266 NdYAG UVlaser system both linked to an Agilent 7500 inductivelycoupled plasma mass spectrometry (ICP-MS) system andreduced using the GLITTER software (van Achterberget al 1999) Accuracy and precision were monitored by

analysing basalt standard BCR-2G with each batch ofsamples as an unknown Results standard deviations anddetection limits are given in Electronic Appendix 1 whichcan be downloaded from httppetrologyoxfordjournalsorg Sr^Nd^Hf isotope data for garnet and clinopyroxene(cpx) separates leached and ultrasonicated in 6N HClfor 30min followed by ultrasonication in three aliquots ofMQ prior to dissolution were obtained using a Nu Plasmamulti-collector (MC) ICP-MS system All analytical workwas carried out in the GEMOC National Key Centreat Macquarie University (wwwesmqeduauGEMOC)following the techniques described by Aulbach et al(2004b) Repeated measurements of standard materialsduring data acquisition (February^July 2003) yielded thefollowing values 87Sr86Sr of 07102570000045 (2 SDnfrac14 42) for the SRM-987 standard 87Sr86Sr of0703520000040 (nfrac1415) for BHVO-1 143Nd144Nd of05111380000038 (nfrac14 42) for the JMC-321 standard143Nd144Nd of 05130030000043 (nfrac1410) for BHVO-1176Hf177Hf of 02821630000002 (nfrac14 20) for the JMC-475 standard and 176Hf177Hf of 02831280000090 forBHVO-1 (nfrac14 8) (Aulbach et al 2004b)Although the uncertainties of Sr and Nd isotope ratios

for pure Sr and Nd standards obtained by MC-ICP-MSare two to three times higher than those reported for ther-mal ionization mass spectrometry (TIMS) measurements(httpgeoremmpch-mainzgwdgde) they are adequatefor the samples analysed in the present study whichmostly have significantly evolved isotopic compositionscompared with depleted mantle or CHUR The signifi-cantly higher uncertainty for 176Hf177Hf of the rock stan-dard than for the pure Hf standard may reflect matrixinterference as a result of imperfectly purified solutions(eg Blichert-Toft et al 1997)

PETROGRAPHY ANDCLASSI F ICAT IONThe petrography of the studied eclogites from Lac deGras is summarized in Table 1 The majority of samplesthat were large enough to obtain microstructural informa-tion have medium- to coarse-grained granoblasticmicrostructures Three eclogites have a weakly tabularmicrostructure which is due to the weak elongation andsubparallel alignment of garnet and cpx in sampleYK3528 and to the preferred orientation of phlogopite insamplesVR19674ecl-g7 and YK1911Most samples are fresh or show minor (510) altera-

tion that is restricted to veins and grain boundaries A fewsamples are more strongly altered with pyroxenesusually more affected than garnet Altered areas arecomposed of small secondary phlogopite sulfide flakesand unidentifiable microcrystalline phases with a feltybrown to green appearance

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1845

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 1 Petrographic data and eclogite classification

Sample no Ref1 Type2 GpMampG3 GpTampN

4 Size5 Microstructure6 Assemblagemode7

gt cpx rut sf ky opx other

Eclogites

SE01 3 high-Mg I A 51 na x x

SE02 3 high-Ca I B 51 na x x x

SE03 3 high-Mg II A 51 na x x

SE04 3 high-Mg II B 51 na x x

SE11 3 low-Mg II B 51 na x x x x

vr09399 2 high-Mg II A 51 na x x

vr19674ecl-g4 2 low-Mg II B 16 na x x

vr19674ecl-g5 2 high-Ca II C 51 na x x

vr19674ecl-g6 2 high-Mg I A 51 na x x

vr19674ecl-g7 1 volatile-rich II A 15 medium-tabular phlog (44) ap (51)

vr19677ecl-g1 2 high-Mg II B 51 na x x

vr19677ecl-g2 2 high-Mg II B 17 na x x x

vr19677ecl-g4 2 high-Mg II B 51 na x x

vr19677ecl-g5 2 high-Mg I B 51 na x x

vr19677ecl-g6 2 high-Mg II A 51 na x x

vr40345 3 high-Ca I B 51 na x x x dia (trace)

vr40364 2 high-Mg II A 51 na x x

vr43452 1 low-Mg I B 45 coarse-grano 48 51 1

vr43465 1 high-Ca I B 2 coarse-grano 41 55 1 graph (3)

vr43469 1 low-Mg II B 15 coarse-grano 60 40 51 51

vr43477 1 low-Mg II B 15 coarse-grano x x x

vr43479 1 low-Mg II A 25 medium-grano 50 50 51

vr43480 1 high-Mg I A 2 medium-grano 89 11 51

vr50858 2 high-Ca II B 51 na x x x

vr50860 2 high-Ca I C 51 na x x

vr50909 3 high-Ca II B 51 na x x x x

vr50916 2 high-Mg II A 51 na x x

vr67112b 1 volatile-rich II C 2 medium-grano 51 ilm (3) cc (2)

vr67360 1 high-Mg II A 3 coarse-grano 62 38 51

yk1911 1 volatile-rich II B 2 medium-tabular 31 18 phlog (49) ap (51) plag (51) cc (51)

yk1926 1 low-Mg II B 17 coarse-grano 70 30 51

yk1943 1 low-Mg I B 2 coarse-grano 57 41 2 51 qz (51)

yk1946 1 high-Mg II A 3 coarse-grano 62 37 1 51

yk1949 1 high-Mg II B 3 coarse-grano 70 30 51 51

yk3528 1 high-Ca I B 25 medium-tabular 57 43 51

Pyroxenites

SE08 3 na II A 51 na x x

SE09 3 na II A 51 na x x x

vr09353 2 na II A 51 na x x x

vr09384 2 na II B 51 na x x x

vr09405 2 na II A 51 na x x

vr09406 2 na II A 51 na x x x

vr09424 3 na II A 51 na x x x

(continued)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1846

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Some xenoliths have subhedral or rounded garnets in alsquomatrixrsquo of interstitial cpx (Group I of MacGregor ampCarter 1970 Fig 2a) whereas in others garnet and cpxhave straight grain boundaries and an interlocking fabric(Group II Fig 2b) The classification of McCandless ampGurney (1989) builds on work of MacGregor amp Carter(1970) and distinguishes group I eclogites by their higherNa2O in garnet (009wt ) and K2O in cpx (008wt) from group II eclogites The classification used inTable 1 (GpMampG) is that based on Na2O contents ingarnet A different classification (GpTampN) places eclogitesinto groups A B and C distinguished by the MgO FeOand CaO contents of garnets and Na2O and MgO contentsof clinopyroxenes (Coleman et al 1965 Taylor amp Neal

1989) These classifications are based on distinct eclogitesuites from Southern African kimberlites and do not neces-sarily reflect the particularities of eclogites from otherlocalities We therefore use a slightly different scheme(lsquotypesrsquo in Table 1) that reflects the specific groupings ofeclogites in this study with regard to Cr2O3^CaO CaO(MgOthornFeO)^MgOFeO and CaO^Na2O relationshipsin garnet (see below)Garnet modes in volatile-free eclogites range from 40

to 90 vol and cpx modes from 10 to 55 with anaverage and median garnet mode of 60 and 65respectively Two samples have high modal amounts ofphlogopite (44 and 48 respectively) in addition togarnet and cpx and are therefore not eclogites sensu stricto

Table 1 Continued

Sample no Ref1 Type2 GpMampG3 GpTampN

4 Size5 Microstructure6 Assemblagemode7

gt cpx rut sf ky opx other

vr19673ecl-g1 2 na II B 51 na x x x

vr19673ecl-g2 2 na II B 51 na x x x x

vr19674ecl-g1 2 na II A 51 na x x x

vr19674ecl-g2 2 na II A 51 na x x x

vr19674ecl-g3 2 na II A 51 na x x

vr19674per-g1 2 na II A 51 na x x

vr19674per-g4 2 na II A 51 na x x x

vr40302 2 na II A 51 na x x x

vr40336 3 na II A 51 na x x x

vr40337 2 na I A 51 na x x

vr40374 2 na II A 51 na x x x

vr40382 2 na II A 51 na x x x

vr40384 ecl-g4 2 na II A 51 na x x x

vr40399 2 na II B 51 na x x x

vr50856 2 na II A 51 na x x x

vr50887 2 na II A 51 na x x

vr50888 2 na II A 51 na x x

vr50899 2 na II A 51 na x x x

vr50900 2 na II A 51 na x x x

vr50925 2 na II B 51 na x x

yk1914 1 na II A 25 medium-grano 22 78

yk1915 1 na II A 15 coarse-grano x x x x

yk1952 1 na II A 2 medium-grano x x x x

1References 1 this study 2 Pearson et al (1999) 3 Pearson et al (unpublished data) trace-element analysesthis study2Classification based on chemical parameters (see text for details)3Classification of McCandless amp Gurney (1989) based on MacGregor amp Carter (1970)4Classification of Taylor amp Neal (1989) based on Coleman et al (1965)5Longest dimension (cm) across xenolith6Only given for samples41 cm in size grano granoblastic7Modes (vol ) are given where available otherwise presence of mineral is indicated as identified in thin section or stubgt garnet cpx clinopyroxene ky kyanite rut rutile sulf sulfide phlog phlogopite ap apatite dia diamond graphgraphite cc calcite ilm ilmenite pl plagioclase qz quartz opx orthopyroxene na not applicable or not available

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1847

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(phlogopite occurring as small secondary grains at grainboundaries and in veins is not considered in this study)Ten samples of all eclogite types show secondary spongyrims of cpx around cpx cores Secondary rims containsmall (mm-scale) patches of glassRutile is the most common primary accessory phase

(15 of 65 eclogites and pyroxenites) with modes between51 and 2 vol Sulfide occurs in 14 samples and ispresent as rounded to subhedral to irregular grains

Kyanite has been identified in three samples studied byPearson et al (1999) One of the kyanite-bearing eclogitesalso contains diamond (VR40345) Kyanite as well asdiamond or graphite associated with kyanite-bearing andcompositionally similar eclogites (high-Ca eclogites) hasbeen frequently observed in drill core (Pearson et al1999) but because these eclogite types are often alteredand friable they are not proportionally represented in thepresent study One high-Ca eclogite (VR43465) contains

Fig 2 Eclogites in thin section (plane-polarized light) illustrating the microstructures mentioned in the text (a) high-Mg eclogiteVR67360(b) high-Mg eclogiteYK1946 (c) phlogopitized eclogiteYK1911 (d) garnet pyroxeniteYK1915 (e) garnet pyroxeniteYK1914 (f) garnet pyrox-enite schlieren in a matrix of garnet spongy opx and glass (YK1952) Grt garnet cpx clinopyroxene rut rutile phlog phlogopite Scalebar in(c) applies to all sections (1cm)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1848

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

3 vol graphite which occurs as disseminated plates sev-eral millimetres long with prismatic tabular or irregularhabit Quartz has been retrieved from a mineral separateof low-Mg eclogite YK1943 The phlogopitized samples(VR19674ecl-g7 and YK1911) also contain apatite andglass as accessory phases Vesicles in sample VR67112bare partially filled with single or multiple large carbonatecrystalsMost pyroxenites from Lac de Gras have fine- to coarse-

grained granoblastic equilibrated microstructures(Pearson et al 1999) Modal information for pyroxenitesin the present study is available for only the larger sampleYK1914 (22 garnet 78 cpx) (Fig 2e)

MAJOR ELEMENTSAverage major-element compositions of garnet cpx andopx in the different eclogite types (except for the threevolatile-rich eclogites which are not averaged) and in pyr-oxenites are given in Table 2 the full datasets includingaccessory minerals are available as ElectronicAppendices 2^5 which can be downloaded from httppetrologyoxfordjournalsorg

GarnetGarnets have XMg [pyrope component 100Mg(MgthornFethornCathornMn)] ranging from 178 to 756 andXCa (grossular component mol) ranging from 72 to359 (Electronic Appendix 2) In a diagram of CaO(MgOthornFeO) vs MgOFeO (Fig 3) eclogitic garnetsfall into three groups (1) low-Mg garnets with low

CaO(MgOthornFeO) and low MgOFeO (2) high-Mggarnets with low CaO(MgOthornFeO) and high MgOFeO (3) high-Ca those with high CaO(MgOthornFeO)and intermediate MgOFeO Volatile-rich eclogites havegarnets with the lowest MgOFeO whereas pyroxenitestrend towards higher values than eclogitesMost eclogite garnets have Cr2O3 contents502wt

with variable CaO contents (Fig 4) Garnets in pyroxeniteshave low CaO contents which correlate positively withCr2O3 More than half of the pyroxenites are opx-free andall pyroxenites are olivine-free on thin-section scalePyroxenitic garnets show an opposing trend of CaOvs Na2O (Fig 5) compared with the eclogitic garnetsBecause the positive correlation of Cr2O3 and CaO forpyroxenitic garnet indicates buffering by both cpx andopx opx-bearing or opx-free pyroxenites will not befurther distinguished subsequently (except in the geother-mobarometry section where opx-bearing assemblagesallow simultaneous calculation of pressure andtemperature)Distinct rim compositions in most garnets are

characterized by lower CaO contents and higherMg-number [frac14 100Mg(MgthornFe)] Garnet compositionsin three samples are inhomogeneous with regard to CaOMgO and FeO (SE01 VR43479 and VR43480) withoutclear core^rim zonations similar to coexisting cpx Likerims CaO-poor compositions are characterized by higherMg-number Visibly secondary spongy cpx rims also havehigher MgO contents (see below) and MgO-rich garnetsare therefore regarded as affected by secondary changesBased on Cr^Ca Mg^Ca^Fe and Ca^Na composi-

tional relationships in garnet the following eclogitetypes are distinguished (1) eclogites with high-Ca

0

05

10

15

20

25

30

0 01 02 03 04 05 06 07 08 09

high-Ca

high-Mg+pyroxenites

low-Mg

CaO(MgO+FeO)

MgO

FeO

GARNETpyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropic Jericho massiveDiavik low-temperatureDiavik high-temperatureDiavik diamond-bearing

eclogites

Fig 3 CaO(MgOthornFeO) vs MgOFeO (wt ) in garnet Fieldsdistinguish low-Mg and high-Ca eclogites as well as high-Mg eclogitesplus pyroxenites in this study Garnets in eclogites from Jericho(Kopylova et al 1999) and Diavik (Schmidberger et al 2007) areshown for comparison Garnet inclusions in diamond from Davieset al (1999 2004)

-135 wt (high-Ca)

2

3

4

5

6

7

8

9

0 05 10 15 20 25Cr2O3

CaO

GARNET

opx-bearing

opx-free

pyroxenites

low-Mghigh-Mgvolatile-rich

Jericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

VR19673ecl-g1g2

Fig 4 Cr2O3 vs CaO (wt ) in garnet Garnets in eclogites fromJericho and Diavik as in Fig 3

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1849

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 2 Summary of major-element contents in garnet and cpx (wt )

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO Na2O Total Mg-no T(Krogh) Type

Garnet

Low-Mg ecl 3964 026 2214 011 1843 037 1050 846 009 1000 503 976

minimum 3893 008 2130 5009 1071 024 849 567 006 990 423 860

maximum 4094 036 2275 011 2093 046 1457 1062 012 1004 708 1088

1s 070 012 052 na 354 007 227 185 003 05 99 98

High-Mg ecl 4081 026 2261 013 1419 030 1537 624 007 1000 659 969

minimum 3970 008 2224 5009 951 023 1127 418 003 991 551 820

maximum 4250 052 2323 035 1795 036 2030 882 013 1012 792 1241

1s 077 013 031 na 225 004 189 141 003 05 57 88

High-Ca ecl 3979 034 2211 na 1516 027 958 1240 011 998 528 1078

minimum 3909 011 2117 5009 1164 023 801 1144 007 989 456 910

maximum 4039 073 2242 011 1713 035 1249 1351 018 1004 657 1246

1s 042 026 042 na 173 004 155 073 004 06 67 151

vr19674ecl-g7 3867 003 2187 5009 2380 153 793 627 5004 1002 373 volatile-rich

vr67112b 3783 013 2097 009 2620 204 466 790 5004 999 241 volatile-rich

yk1911 3874 5006 2200 5009 2120 221 802 747 5004 997 403 volatile-rich

Pyroxenites 4163 021 2257 075 1148 034 1906 407 006 1002 747 909

minimum 3989 005 2081 009 729 025 1535 277 004 996 593 699

maximum 4279 048 2330 260 1881 054 2215 566 009 1011 842 1158

1s 068 010 057 057 323 007 209 065 001 04 73 117

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO Na2O Total Mg-no Type

Cpx

Low-Mg ecl 5453 028 739 na 500 5009 1117 1662 434 994 801

minimum 5160 017 564 5009 337 na 876 1388 197 981 741

maximum 5570 044 1067 014 687 na 1324 1926 632 1001 865

1s 128 011 155 na 124 na 153 177 126 06 40

High-Mg ecl 5483 023 483 na 381 na 1368 1875 318 994 865

minimum 5353 009 263 5009 245 5009 1156 1617 204 986 778

maximum 5645 031 810 016 674 011 1588 2101 495 1011 920

1s 071 007 137 na 095 na 104 138 075 06 30

High-Ca ecl 5563 027 1184 na 318 5009 868 1368 615 995 836

minimum 5448 014 769 5009 168 na 619 1060 391 987 751

maximum 5670 049 1579 010 603 na 1148 1689 807 1003 878

1s 076 015 334 na 151 na 189 261 167 06 45

vr19674 ecl-g7 5322 010 185 5009 536 015 1448 2339 062 992 828 volatile-rich

vr67112b 5040 041 328 5009 1079 024 1108 2229 071 992 647 volatile-rich

yk1911 5231 014 301 5009 544 024 1450 2340 056 996 826 volatile-rich

Pyroxenites 5527 na 347 038 304 na 1543 1959 245 999 899

minimum 5299 5006 037 010 149 5009 1128 1408 030 990 804

maximum 5666 037 786 134 490 014 1863 2312 537 1009 947

1s 068 na 189 028 091 na 193 226 131 05 37

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO NiO Total Mg-no

Opx

Pyroxenites 5781 na 059 na 627 na 3508 037 011 1006 909

minimum 5546 5006 032 5009 462 5009 2999 020 008 996 804

maximum 5888 009 318 054 1304 014 3628 077 014 1014 933

1s 098 na 067 na 212 na 160 019 002 06 32

Averaged for the three eclogite types and for pyroxenites minimum and maximum values as well as standard deviationsalso given Concentrations in volatile-rich eclogites are shown individually ecl eclogite Mg-numberfrac14 100Mg(Mgthorn Fe)Temperatures of last equilibration [T(Krogh)] as in Electronic Appendix 8

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1850

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

garnets (these include kyanite- graphite- and diamond-bearing varieties) (2) eclogites with low-Mg garnets (theseinclude quartz-bearing eclogite) (3) eclogites with high-Mggarnets and Cr2O3 contents502wt (4) volatile-richeclogites (phlogopite- apatite- carbonate-bearing)

Clinopyroxene (cpx)Clinopyroxenes in all eclogite groups have higher averageAl2O3 contents than those in pyroxenites with the formerhaving about 90 of the total Al in the jadeite molecule(AlVI) versus about 35 for the latter corresponding toan average of 41 and 25wt Na2O respectivelyThe highest Al2O3 and Na2O contents in the dataset areobserved in cpx from the high-Ca eclogites (up to158wt Al2O3) (Electronic Appendix 3) Contents ofK2O are generally below the detection limit(5004wt ) but values up to 024wt are observed insome cpx in high-Ca eclogites Average cpx Mg-numberare highest in pyroxenites (899) followed by cpx inhigh-Mg eclogites (865) high-Ca eclogites (836) andlow-Mg eclogites (801) Pyroxenitic cpx tend to havelower TiO2 and higher NiO and distinctly higherCr2O3 contents compared with those in eclogites A plotof MgO vs Al2O3 shows a negative correlation (Fig 6)Clinopyroxenes in low-Mg eclogites have lower Al2O3

contents at a given MgO content than other eclogitetypes Between 15 and 17wt MgO the Al2O3^MgOcorrelation for cpx in pyroxenites has a different slopefrom that in the eclogites Clinopyroxenes in pyroxenites

VR19673ecl-g1 and VR19673ecl-g2 have distinctly lowerMgO contents and higher Al2O3 contents than inthe other pyroxenites They coexist with garnets havingthe lowest Cr2O3 and CaO contents of the dataset whichdefine one end of the positive Cr2O3^CaO correlationobserved for garnetsClinopyroxenes in volatile-rich eclogites have distinctly

high MnO contents (up to 024wt ) Mg-numbers forcpx in the phlogopite-bearing variety are 826 and 828whereas that in the calcite-bearing eclogite (VR67112b)has an Mg-number of 647Some cpx grains have spongy rims that contain small

(mm-scale) patches of glass these rims are richer in Caand Mg and markedly poorer in Na than the cores similarto cpx in eclogites from southern Africa (Taylor amp Neal1989) The rim texture is unequilibrated suggesting lategrowth The formation of the spongy cpx rims might berelated to incongruent melting of cpx or be a result ofmetasomatism (Taylor amp Neal 1989 and referencestherein) Distinct although not spongy rims in two eclo-gites (VR67360 and YK1946) are similarly enriched inCaO and depleted in Na2O Two types of cpx with respectto CaO Na2O and Mg are present in VR43477VR43479 and VR43480 without clear core^rim zonationor spongy rims No significant differences are recognizedwith respect to other oxides

Orthopyroxene (opx)Orthopyroxenes have Mg-number ranging from 804 to933 (17 of 19 samples have values between 895 and 933)and typically contain minor amounts of Al2O3

(032^118wt one outlier has 318wt ) CaO(020^118 wt ) and NiO (008^014wt ) (Electronic

0

2

4

6

8

10

12

14

7 9 11 13 15 17 19

Al 2

O3

CPX

pyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

cpx in pyroxenitesVR19673ecl-g1g2

MgO

Fig 6 MgO vs Al2O3 (wt ) in cpx Data sources as in Figs 3and 4

0

005

010

015

020

2 4 6 8 10 12 14 16

CaO

Na 2

O

dl

group Igroup II

pyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

GARNET

Fig 5 CaO vs Na2O (wt ) in garnet Arrows show trends foreclogites and pyroxenites respectively [note that part of thepyroxenite trend extends into the field falling below the detectionlimit (dl) for Na2O] Group I and group II refer to classificationof McCandless amp Gurney (1989) Data sources as in Figs 3 and 4

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1851

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Appendix 4) The contents of all other elements are mostlybelow their respective detection limits

Accessory phasesRutile and ilmenite occur in some of the eclogites (Table 1)All rutile grains have exsolved 10 ilmenite as lamellaeand rims Ilmenite-free areas of rutile in eclogites containbetween 962 and 988wt TiO2 and minor amounts ofAl2O3 (01^04wt ) Cr2O3 (5009 to 02wt ) andFeO (03^07wt ) Rutile in pyroxenite YK1915 contains004wt Al2O3 01wt FeO and 15wt Cr2O3Discrete ilmenite grains in sample VR67112b haveTiO2 contents of 499 and 500wt MgO of 23 and15wt and FeO of 399 and 403wt respectivelySulfides occur interstitially or enclosed in cpx garnet

and opx they are mostly pyrrhotite and monosulfide solidsolution with variable Ni contents (unpublished data)They will not be considered further hereAbundant platy phlogopite is observed in two samples

(VR19674ecl-g7 and YK1911) It is Cl-rich in sampleVR19674ecl-g7 (063wt vs 003wt in YK1911) andF-rich in sample YK1911 (097wt vs 005wt inVR19674ecl-g7) (Electronic Appendix 5) The same is truefor coexisting apatite (VR19674ecl-g7 254wt Cl and014wt F YK1911 006wt Cl and 306wt F)

Comparison with mineral inclusions indiamonds from Lac de Gras and witheclogite xenoliths from other Slave CratonlocalitiesMineral inclusions in diamonds from the Lac de Graskimberlites have been investigated by Davies et al (19992004) For garnets there is a striking similarity betweenthe inclusions in diamond and those in the high-Ca eclogitexenoliths although the inclusions in diamond trendtowards higher Na2O (Fig 5) and lower Al2O3 contentsThe overlap is not as marked for cpx where inclusionsin diamond span almost the entire range observed foreclogitic cpx (Fig 6)A comparison with eclogites from other localities in

the Slave Craton shows that garnets in most Diavik low-temperature eclogites reported by Schmidberger et al(2007) plot with garnets in high-Mg eclogites and inpyroxenites from this study (Figs 3^5) whereas garnets inhigh-temperature and diamond-bearing eclogites(Schmidberger et al 2007) overlap most with garnets inhigh-Ca eclogites from this study and with garnetsincluded in diamond Eclogites from the Jericho kimberlitehave either massive or anisotropic fabrics (Kopylova et al1999) Garnets in almost all massive eclogites plot withgarnets in low-Mg eclogites from this study with regard toCaO^MgO^FeO relationships (Fig 3) but many havelower Na2O contents at a given CaO content (Fig 5)Garnets in anisotropic eclogites are restricted to lowMgOFeO (507) and show some overlap with garnets

in high-Ca eclogites although trending towards lowerCaO(MgOthornFeO) (Fig 3) while coexisting cpx is alsorestricted to low MgO and trends toward high Al2O3 con-tents (Fig 6)

TRACE ELEMENTSGarnetTrace-element abundances in garnet are summarized inTable 3 and the complete dataset is reported in ElectronicAppendix 1 Garnets in low-Mg eclogites show steep posi-tive slopes between LaN and SmN and shallower positiveslopes in the normalized heavy rare earth elements(HREEN Fig 7a) The normalized light REE (LREEN)show an order of magnitude variability ZrHf is mostlysupra-chondritic and Ti can be enriched or depleted rela-tive to elements of similar compatibility Niobium in threesamples is strongly depleted at 002^003 chondriticGarnets in high-Mg eclogites also show smooth positiveslopes in LREEN but have flat HREEN and variablenormalized middle REE (MREEN) to HREEN (Fig 7b)ZrHf is always supra-chondritic and Ti enriched ordepleted relative to similarly compatible elementsNiobium is most samples is below detection Small positiveEu anomalies though not outside the analytical uncer-tainty for Eu relative to either Sm or Gd are observed forgarnet in some high-Mg eclogites (eg EuEu up to 15for VR67360 where Eu is the average of the chondrite-normalized Sm and Gd concentrations) Garnets in high-Ca eclogites are slightly enriched in the MREEN relative tothe HREEN or have flat MREEN and HREEN with ahump between EuN and HoN (Fig 7c) One sample(VR40345) shows strong La enrichment (greater thanchondrite) Two samples have supra-chondritic ZrHfwhereas three samples have close to chondritic ratios andone sample has a subchondritic ratio As is true for theother eclogite typesTiN is either enriched or depleted rela-tive to similarly compatible elementsPyroxenitic garnets have variable slopes in the LREEN

and flat MREEN to HREEN patterns and two samplesshow high Sc relative to neighbouring elements Garnet inone of the pyroxenites (YK1952) has a highly distinctivepattern relative to other pyroxenites with very low LREEand HFSE (Fig 7d) This sample consists of coarse pyroxe-nite schlieren in which the garnet was analysed in a finer-grained lsquomatrixrsquo of spongy cpx with secondary rims finecpx^opx intergrowths embayed garnet and melt patchesVolatile-rich eclogites have garnets that are slightly

enriched in MREEN relative to HREEN or have flatMREE to HREE patterns One garnet has a pronouncednegative Eu anomaly (EuEufrac14 07 forVR67112b Fig7e)Garnets in volatile-rich samples have the lowest Ni Coand Ti and the highest V ScYand HREE abundances inthe dataset

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1852

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 3 Summary of trace-element contents in garnet and cpx (ppm)

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR19674

ecl-g7

VR67112b YK1911 av min max 1s

n 6 6 7 3 5 4 4

P 169 54 270 88 196 70 310 78 282 80 450 138 160 60 80 214 48 370 144

Sc 62 40 89 16 48 36 67 11 53 42 79 15 69 92 93 186 52 440 179

Ti 1667 490 2500 680 1073 620 2000 455 2141 700 4700 1582 270 690 260 1097 540 1900 573

V 169 78 250 68 150 78 280 72 178 110 390 101 140 340 100 390 71 820 326

Co 59 42 70 11 58 39 84 13 58 47 65 5 30 41 27 68 55 90 16

Ni 17 5 32 10 21 11 34 10 24 11 35 8 6 10 51 54 25 110 38

Ga 12 7 15 3 13 8 16 3 12 9 20 4 8 86 58 12 10 14 2

Sr na 512 4 na na 522 522 na na 57 57 na 52 5025 510 513 513 513 na

Y 37 6 61 18 21 7 38 10 31 20 38 5 60 81 70 19 12 26 6

Zr 15 5 29 9 28 6 83 26 24 9 43 12 60 22 90 18 1 41 17

Nb 009 0005 048 019 na 50015 05 na na 5005 05 na 5003 0012 5001 022 0030 057 025

La 005 001 009 004 na 50007 00 na na 5003 04 na 004 002 0009 003 000 009 005

Ce 03 01 05 02 01 00 01 00 05 01 18 05 03 022 015 01 00 03 01

Pr 01 00 02 01 00 00 01 00 02 01 03 00 017 014 012 00 00 01 00

Nd 12 03 24 07 07 04 14 03 21 12 32 07 24 23 20 03 00 07 03

Sm 11 04 22 06 10 03 18 05 23 08 36 10 29 31 33 05 01 09 04

Eu 06 03 10 03 06 02 10 03 11 04 17 04 16 114 16 03 01 06 02

Gd 27 07 41 12 23 07 42 12 43 16 53 13 8 81 90 13 06 24 08

Dy 54 11 70 23 36 13 65 18 58 35 71 11 12 144 12 29 18 42 10

Ho 14 02 23 07 08 03 15 04 12 09 14 02 24 30 23 07 04 10 02

Er 46 07 80 24 24 09 40 10 34 25 40 05 8 85 6 22 14 28 06

Yb 52 06 98 30 24 10 37 08 32 22 39 05 8 70 5 23 13 30 07

Lu 09 01 16 05 04 02 05 01 05 03 07 01 11 09 08 04 02 05 01

Hf 03 02 05 01 04 01 10 03 05 02 09 02 09 027 007 03 00 06 03

Ta na 50008 5004 na na 50005 001 na na 50004 na na 003 50002 50005 na 50007 011 na

Pb na 5005 02 na na 5002 02 na na 5003 07 na 02 004 5002 na 5003 01 na

Th na 5001 005 na na 50007 003 na na 50008 011 na 5007 0012 50004 002 0005 006 002

U na 5003 5003 na na 5001 01 na na 5007 na na 01 0023 010 na 5001 003 na

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR67112b YK1911 av min max 1s

n 6 4 8 5 4 4

P 65 15 120 41 39 14 56 16 53 30 70 12 20 9 80 27 120 40

Sc 22 10 30 7 15 14 19 2 14 7 24 5 80 43 17 11 21 5

Ti 1557 990 2400 508 1168 1010 1490 191 1988 810 3500 1154 2390 1700 689 102 1685 694

V 415 180 480 107 318 210 400 69 324 259 450 72 600 180 250 139 480 156

Co 35 26 50 8 20 14 31 7 26 14 42 11 49 23 29 20 42 10

Ni 164 87 220 47 228 100 430 128 188 135 320 56 59 34 428 410 670 184

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1853

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

ClinopyroxeneTrace-element abundances in cpx are summarized inTable 3 and the complete dataset is reported in ElectronicAppendix 6 Clinopyroxenes in low-Mg eclogites havemildly positive to negative slopes from LaN to NdN andsmooth negative slopes between SmN and HoN thatmostly continue to LuN (Fig 8a) Niobium is below detec-tion or 5007 chondritic and SrN is variably enrichedrelative to neighbouring elements with the peak size corre-lating negatively with LREEN ZrHf is subchondritic andHfN is enriched relative to SmN whereasTiN is enriched ordepleted relative to similarly compatible elementsWe ana-lysed the spongy rim in cpxYK1943 and its trace-elementpattern is similar to that of cpx cores for more compatibleelements whereas the most incompatible elements areseveral orders of magnitude higher Clinopyroxenes in

high-Mg eclogites have steeper negative slopes betweenNdN and LuN than those in low-Mg eclogites combinedwith lower HREE abundances HfN is depleted relative toSmN (Fig 8b) High-Ca eclogites contain cpx with sinusoi-dal REEN patterns with lower LREE than other eclogitetypes and the heaviest REE below detection They haveuniformly strong positive Sr peaks relative to elements ofsimilar compatibility (Fig 9) The diamond-bearing high-Ca eclogite (VR40345) has cpx with unusually high Nband LREE abundancesClinopyroxenes in four pyroxenites have disparate trace

element patterns They all share depletions in TiN relativeto similarly compatible elements and two show positiveSrN peaks Two samples have cpx with high and flatLREEN to MREEN patterns and steep negative slopes inthe HREEN The trace-element patterns of cpx in YK1914

Table 3 Continued

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR67112b YK1911 av min max 1s

n 6 4 8 5 4 4

Ga 16 10 20 3 12 9 15 2 26 15 39 9 129 123 6 11 11 5

Rb na 5008 04 na na 5007 05 02 na 5003 03 na 02 5004 03 010 05 0

Sr 166 85 290 74 222 153 285 50 243 83 440 142 85 120 200 11 414 209

Y 2 04 4 10 13 07 2 04 11 03 3 10 28 51 2 03 3 2

Zr 19 4 48 15 25 8 45 13 17 8 28 7 111 129 36 008 81 42

Nb 01 0003 03 01 na 5002 00 na na 5002 16 9 007 011 010 003 02 008

Cs na 5001 6 na na 5006 5006 na5013 5004 5013 na bdl 5002 na 5001 5002 na

Ba 09 02 2 07 09 01 2 07 10 03 3 10 06 01 13 06 3 10

La 3 03 8 3 3 23 3 04 4 01 14 6 51 97 4 01 10 4

Ce 7 13 20 6 10 7 13 2 7 05 24 10 25 28 11 01 28 13

Pr 11 02 3 08 2 10 3 06 08 02 2 09 47 46 2 002 4 2

Nd 5 12 12 4 10 5 14 4 4 09 8 26 23 22 8 01 22 10

Sm 11 05 21 06 20 10 31 10 07 02 11 03 47 53 2 005 5 2

Eu 03 01 05 01 06 028 09 03 02 010 03 007 087 123 06 002 15 07

Gd 08 03 13 03 11 05 19 06 05 03 06 011 28 37 15 010 36 2

Dy 05 01 07 02 04 022 06 01 02 013 03 007 11 16 07 008 14 06

Ho 01 00 02 004 006 003 007 002 na 5003 012 na 011 021 009 001 02 007

Er na 507 507 na 011 004 02 00 na 5006 03 na 019 033 02 002 03 012

Yb na 5001 02 na na 5001 011 na na 5006 03 na 007 015 009 001 02 006

Lu na 5001 003 na na50002 002 na na50008 01 na 0006 0013 0013 0002 003 001

Hf 12 03 24 07 11 04 18 05 10 06 16 04 34 53 na50015 35 na

Ta na50003 50015 na na50006 006 na na50004 5004 na 0014 00236 na50003 503 na

Pb 06 02 12 04 06 04 09 02 na 5016 07 02 065 098 08 01 18 1

Th na50005 06 na 012 005 02 005 na 5002 5006 na 021 02 001 06 03

U na 5005 5005 na 002 0009 005 001 na 5003 5005 na 0013 024 na50005 003 na

Averaged for the three eclogite types and for pyroxenites as in Table 2 pyx pyroxenite n number of analyses averagedbdl below detection limit

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1854

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

and YK1952 are similar although at different abundanceswith only modest LREEN over HREEN enrichment andstrong depletions in ZrN relative to neighbouring elements(Fig 8d)Clinopyroxenes in the volatile-rich samples have flat pat-

terns and high abundances from LaN to SmN and smoothnegative slopes from SmN to LuN Niobium Sr and Ti are

noticeably depleted relative to elements of similar compat-ibility (Fig 8e)

Trace-element partitioningTo assess whether coexisting phases are in trace-elementequilibrium we calculated apparent cpx^garnet distribu-tion coefficients and compared them with those

(a)

(d)

(b)

(e)

(c)

GarnetCI-chondrite GarnetCI-chondrite

low-Mghigh-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1952

1

10

Nd Sm Eu Gd Dy

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Fig 7 Extended trace-element patterns of garnet in (a) low-Mg eclogites (b) high-Mg eclogites with inset showing enlargement of MREE tohighlight Eu anomalies and 1s uncertainties (c) high-Ca eclogites (d) pyroxenites and (e) volatile-bearing eclogites Normalized to chondritevalues of McDonough amp Sun (1995)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1855

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

determined experimentally or in demonstrably equili-brated natural samples Low-Mg and high-Mg eclogiteshave Dcpxgarnet for most elements that are within therange determined for natural equilibrated eclogites andexperimental assemblages (Harte amp Kirkley 1997 Greenet al 2000 Barth et al 2002a) (Electronic Appendix 7)High-Ca eclogites plot at the low end of the range

of distribution coefficients and have very low DMREE

and DHREE outside the range indicated by the datasetof Harte amp Kirkley (1997) who noted that DREE decreaseswith increasing Ca content in eclogitic minerals Theseassemblages are therefore considered to be in equilibriumPyroxenite VR40384 has D values similar to eclogites

but at the high end of the range determined by Harte amp

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

(a)

(d)

(b)

(e)

(c)

CpxCI-chondrite CpxCI-chondrite

low-Mg high-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1914

YK1952

spongy rim

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

VR40345

Fig 8 Extended trace-element patterns of cpx as in Fig 7

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1856

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Kirkley (1997) (Electronic Appendix 7) Another group(YK1914 and YK1952) plots at the low end but DHf DZrDTi and DV are far below the values indicated by experi-mental datasets Because pyroxenites compositionallygrade into eclogites their drastically different DHFSE

values suggest disequilibrium In contrast volatile-bearingsamples have DHFSE values that appear to be too high to bein equilibrium however this may be a consequence oftheir unusual bulk chemical composition (Ti- and K-rich)which differs substantially from that of lsquonormalrsquo eclogitesand experimental charges

GEOTHERMOBAROMETRYFollowing Schmidberger et al (2007) temperatures oflast equilibration were calculated for eclogitic andopx-free pyroxenitic assemblages using the formulationof Krogh (1988) The results are summarized in Table 2and Electronic Appendix 8 for a preset pressure of5 GPa and illustrated as a histogram in Fig 9 Foropx-bearing pyroxenites temperatures were calculatedsimultaneously with pressure [Ca-in-opx thermometryand Al-in-opx barometry using the formulations of Breyet al (1990)] to constrain their relative positions in themantle columnMinerals in some eclogites from Lac de Gras show dis-

tinct core and rim compositions or inhomogeneitiesTemperatures were not calculated for samples withspatially irresolvable inhomogeneous garnet compositions(SE01VR43479VR43480) For eclogites with spongy cpxcore compositions were used in the calculationTemperatures were calculated for both core and rim com-positions where applicable For all samples these

temperatures agree within the uncertainty of this thermo-meter (508C Brey amp Kolaquo hler 1990)Most equilibration temperatures for eclogites lie between

800 and10008C (Fig 9)The transition between the shallowand the deep lithospheric mantle layers in the central Lacde Gras area lies in this temperature interval (Griffin et al1999 Pearson et al 1999) Eclogites in this study lack thedistinct bimodal temperature distribution observed bySchmidberger et al (2007) but as in their study high-Caeclogites give higher average temperatures (10808C corre-sponding to a depth of 155 km if equilibrated along a 40mWm2 geotherm) than other eclogites types (9808Cdepth 135 km) the diamond-bearing sample (VR40345)gives thehighest temperature of the datasetHighNa2Ocon-tents in garnet from high-Ca eclogites are qualitativelyconsistent with higher equilibration pressures (Sobolev ampLavrentrsquoyev 1971) as are high K2O contents in coexistingcpx (Erlank ampKushiro1970) suggesting that high-Ca eclo-gites not only equilibrated at high temperature but also highpressure rather than representing a temperature excursionPyroxenites both opx-bearing and opx-free give somewhatlower average temperatures than eclogites (mean 9258C)and phlogopite-rich eclogites give the lowest temperatures(685 and 7698C respectively) The calcite-bearing eclogiteVR67112b gives a relatively low equilibration temperatureof 8558C

MEASURED ANDRECONSTRUCTEDWHOLE -ROCKSRationale and approachKimberlite-borne xenoliths sometimes show evidence forinfiltration of the host kimberlite (eg Taylor amp Neal1989 Barth et al 2001 2002b Heaman et al 2006) Wetherefore follow the standard approach of calculatingwhole-rock compositions from the compositions of the con-stituent phases weighted by their modal abundances Onlycore compositions were used as distinct rim compositionsmay point to late diffusion of elements from kimberlite-related melt veinsMany eclogitic whole-rock compositions reconstructed

with only garnet and cpx have negative Ti anomalies(not shown) For most of these samples rutile or ilmenitehas been identified as part of the mineral assemblage andthese phases should therefore be considered in the whole-rock reconstruction Rutile or ilmenite modes were esti-mated by assuming that the bulk-rock has no Ti anomalywhen normalized to a normal mid-ocean ridge basalt(N-MORB) composition and calculating the amount ofrutile required to erase theTi anomalyIt is of course possible that the eclogites had negative

or positive Ti anomalies relative to elements of similarcompatibility (Eu and Dy) Because we reconstructed

0

2

4

6

8

10

12volatile-rich

pyroxenite opx-bearing

pyroxenite opx-free

high-Ca

high-Mg

low-Mg

700750

800850

900950

10001050

11001150

12001250

Fre

quen

cy

Temperature

Fig 9 Histogram ofTKrogh (Krogh 1988) obtained at a preset pres-sure of 5 GPa except opx-bearing pyroxenites where TBKN wassolved simultaneously with PBKN (Brey amp Kolaquo hler 1990)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1857

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

whole-rock compositions from mineral core compositionskimberlite infiltration is unlikely to have caused fractiona-tion of Ti from Eu and Dy In a study in which rutilemodes were estimated from the differences between Ti inmeasured bulk-rocks and in bulk-rocks reconstructed withonly garnet and cpx only three of 19 eclogites recon-structed with rutile show Ti anomalies (Barth et al2002a) Even eclogite xenoliths with depleted Zr and Hfabundances that resemble those of modern island arcbasalts do not show Ti anomalies (Jacob amp Foley 1999)We therefore consider our approach to be a reasonablyrobust means of obtaining meaningful rutile modes wheresmall sample sizes preclude reliable determination of themodes of accessory phases by point-countingWhere rutilewas not recognized on thin-section scale or not analysedthe average eclogitic rutile composition was used for thereconstruction

UncertaintiesReconstructing whole-rock compositions in this way is arather crude method given the coarse grain size of theeclogites relative to the sample size and the non-uniformdistribution of the principal minerals on thin-sectionscale The uncertainty in mineral modes of medium- tocoarse-grained eclogites with grains45mm has been esti-mated by Jerde et al (1993) to be 10 plus uncertaintiesrelated to partial alteration which is minor in our samplesNevertheless the whole-rock trace-element budget israther insensitive to the exact modes for common eclogiticassemblages and REE patterns which encompass mostother elements in terms of compatibility in garnet andcpx do not change significantly even for modal variationsof 30 (Jerde et al 1993) This is not true for major-element concentrations which vary substantially with var-iations in mineral modes and we therefore do not calculatewhole-rock major-element contentsBecause the discussion that follows is based on trace-

element patterns and trends rather than absolute abun-dances whole-rock reconstruction based on mineralmodes is adequate We are confident that the uncertaintyinherent in this approach is not leading to significant mis-representation of eclogite whole-rock compositions becausethe trace-element patterns of the reconstructed whole-rocks show distinct differences (see below) that correlatewith the assignment to eclogite types based on garnetmajor-element composition and this is independent ofmineral mode Following Jerde et al (1993) in ElectronicAppendix 9 we show variations in trace-element patternfor one sample in each eclogite type for variations ingarnet and cpx abundance from 7030 to 3070 vol This shows that the trace-element patterns are generallyinsensitive to cpxgt ratios with regard to slopes andanomalies such as positive Sr Pb and Eu anomalies withthe exception of the positive Eu anomaly in high-Mgeclogite VR43479 which disappears if a high cpx mode is

assumed (not shown) Variations in rutile modes by 50(Electronic Appendix 9d) lead as would be expected tothe appearance of Ti anomalies but a rutile mode of 05instead of 1 vol does not erase the positive Nb anomalyin low-Mg eclogiteVR50909Because of the robustness of the relative trace-element

abundances in the reconstructed eclogites a garnet propor-tion of 065 and cpx proportion of 035 approximately themedian modal abundances of these minerals in volatile-free eclogites from Lac de Gras are assumed for eclogitesamples that are too small for determination of meaningfulmodes by point-counting

Within- and between-group variationsThe trace-element compositions of reconstructed whole-rocks are given in Table 4 N-MORB-normalized trace-element patterns of reconstructed whole-rocks areshown in Fig 10 The three eclogite groups have distincttrace element patterns that are particularly obvious forthe REE (Fig 11) Low-Mg eclogites have positive Pb andSr anomalies and smooth positive slopes in their MREEN

and HREEN LREEN are enriched relative to MREEN inthree samples (Fig 11a) Three samples have high Nbrelative to similarly compatible elements and comparedwith most other eclogites (Fig 10a) All high-Mg eclogitesdisplay more or less pronounced positive Eu and Sranomalies and most have positive Pb and negative Zrand P anomalies Their N-MORB-normalized REE pat-terns show shallow positive slopes from LaN to PrN orNdNand aweak negative slope from NdN to LuN (Fig11b)VR43479 is a low-Mg eclogite with regard to its garnet

Ca^Fe^Mg composition but has a trace-element patterntypical of high-Mg eclogites and will be discussed withthe latter Conversely sample YK1949 a high-Mg eclogitein terms of its garnet Ca^Fe^Mg compostion is more simi-lar to low-Mg eclogites in terms of its whole-rock trace-element pattern and is plotted in Figs 10a and 11aWhereas the patterns of most samples within the groupresemble each other the N-MORB-normalized incompati-ble-element abundances of the high-Ca eclogites vary bymore than an order of magnitude (Fig 10c) All sampleshave positive Pb anomalies and most have positive Sranomalies Some samples have negative P andor Zranomalies Niobium abundances in three of the samplesare very high similar to those for some low-Mg eclogitesHigh-Ca eclogites tend to have flatter MREEN andHREEN patterns and more uniform HREE abundancesthat the other eclogite types with humps between SmN

and HoNThe phlogopitized sample YK1911 shows a strong posi-

tive Pb anomaly and lower HREE abundances than thecalcite-bearing sample VR67112b which shows a distinctnegative Eu anomaly and a depletion of La relative to Ce(Figs 10d and 11d)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1858

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Trace-element compositions of reconstructed whole-rocks (ppm)

Sample P Sc Ti V Co Ni Ga Sr Y Zr Nb La Ce Pr Nd

Low-Mg eclogites

vr19674-g4 186 50 4649 317 57 59 144 77 24 30 36 13 31 05 24

vr43452 154 39 2253 328 49 84 176 52 32 11 01 02 07 01 08

vr43469 38 65 4377 349 54 91 142 34 28 15 02 05 15 03 17

yk1926 144 45 3406 258 52 87 145 89 18 18 20 09 29 06 33

yk1943 173 46 5732 295 41 39 146 122 25 48 82 34 85 13 64

yk19491 102 34 1693 158 31 49 100 46 10 14 03 09 21 03 17

High-Mg eclogites

vr19677-g2 138 39 5160 291 41 53 119 98 14 22 04 13 53 11 64

vr434792 46 25 1265 129 36 126 84 56 3 7 04 06 25 04 23

vr67360 49 36 1573 138 64 184 111 76 5 9 04 09 38 07 33

yk1946 153 28 4877 192 45 105 124 108 12 48 03 10 38 08 50

High-Ca eclogites

vr19674-g5 267 30 7663 200 41 67 195 32 23 43 77 01 04 02 16

vr40345 0 33 2821 175 47 98 210 179 13 13 66 60 109 11 38

vr43465 142 27 5104 220 31 83 121 197 14 35 29 05 22 05 40

vr50858 295 30 6898 201 44 94 150 48 22 38 06 01 03 02 20

vr50860 68 57 5672 418 57 71 186 109 19 11 02 02 13 03 26

vr50909 238 28 7295 201 42 92 171 48 20 57 13 01 04 02 21

yk3528 91 51 3712 341 53 155 177 49 16 9 01 02 08 02 11

Volatile-rich eclogites

vr67112b 40 84 9242 521 46 30 107 4 40 66 31 26 126 24 125

yk1911 31 39 9428 286 44 82 164 3 23 52 64 18 52 09 46

Pyroxenites

yk1914 111 49 707 207 48 547 44 31 4 3 02 06 19 03 10

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

Low-Mg eclogites

vr19674-g4 11 05 10 22 39 10 29 33 06 12 05 028 24 008

vr43452 04 02 08 13 38 12 41 49 08 05 01 nd 02 na

vr43469 11 05 09 25 44 11 34 36 06 07 02 001 09 129

yk1926 12 05 11 19 30 07 22 24 04 07 03 004 23 090

yk1943 21 08 10 29 42 10 30 32 05 16 05 013 65 017

yk19491 06 03 11 08 14 04 14 18 03 04 02 003 29 115

High-Mg eclogites

vr19677-g2 20 08 12 21 24 05 15 15 03 05 05 007 52 6811

vr434792 05 02 12 05 06 01 04 03 00 03 01 001 98 037

vr67360 06 03 14 06 09 02 06 06 01 02 02 002 86 000

yk1946 18 07 12 21 22 05 13 12 02 11 03 006 56 112

High-Ca eclogites

vr19674-g5 17 09 11 33 43 09 24 23 04 16 05 007 02 011

vr40345 09 04 10 12 22 06 16 14 02 05 nd nd 298 na

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1859

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Continued

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

vr43465 22 09 12 25 26 06 15 14 02 10 04 001 25 040

vr50858 20 08 09 33 41 08 22 21 03 10 01 nd 02 na

vr50860 14 07 11 27 37 08 22 18 03 06 00 nd 06 na

vr50909 20 10 12 33 38 07 20 20 03 16 01 002 02 181

yk3528 08 05 11 20 30 07 18 18 02 05 01 nd 08 na

Volatile-rich eclogites

vr67112b 38 10 07 53 74 15 41 34 04 18 03 012 60 004

yk1911 20 07 08 35 39 07 18 17 02 10 14 nd 73 na

Pyroxenites

yk1914 02 01 11 04 08 02 05 06 01 01 03 003 68 101

aGarnet has CandashMgndashFe relationship typical of high-Mg eclogites but whole-rock trace-element pattern is typical oflow-Mg eclogitesbGarnet has CandashMgndashFe relationship typical of low-Mg eclogites but whole-rock trace-element pattern is typical ofhigh-Mg eclogites

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

(a)

001

01

1

10(c)

001

01

10(b)

VR43452

YK1949

RockNMORB

low-Mg

RockNMORB

high-Mg

(d)

volatile-richhigh-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

phlogopite-bearingYK1911

calcite-bearingVR67112b

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 10 Extended trace-element patterns of reconstructed (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-bearing eclogites Normalized to N-MORB of Sun amp McDonough (1989)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1860

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

RockNMORB

(b)

RockNMORB

(d)

010

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu001

01

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

(a)

001

01

1

10(b)

001

01

1

10(c)

001

01

1

10(e)

high-TSchmidberger et al

low-TSchmidberger et al

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

high-Mglow-Mg

high-Cavolatile-rich

phlogopite-bearingYK1911

calcite-bearingVR67112b

YK1949VR43479

Fig 11 N-MORB normalized REE patterns for (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-richeclogites Shown for comparison in (e) are reconstructed eclogites from Schmidberger et al (2007)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1861

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Both volatile-rich whole rocks display strong negativeSr anomalies Strontium can be sited in calcite or apatiteandor indicate the presence of a Sr-rich mineral that wasnot identified in thin section such as bariteFour of the six reconstructed whole-rocks in the study

of Schmidberger et al (2007) (Fig 11e) show markedLREEHREEN enrichment that is observed only in onehigh-Ca eclogite in our study (Fig 11c) suggesting thatmany of the eclogites they studied were affected by metaso-matic LREE enrichment Two samples in their study thatare not affected by this enrichment show positive Euanomalies similar to the high-Mg eclogites in this studywhereas the distinctive REE pattern of the low-Mg eclo-gites is not recognized among their samples

SR^ND^HF I SOTOPESYSTEMATICSStrontium Nd and Hf isotope data for optically puremineral separates and reconstructed whole-rocks aregiven in Table 5 Spongy rims and alteration are likely tobe mostly removed during leaching in 6N HCl in combina-tion with repeated ultrasonication Despite this rigoroustreatment the reproducibility of Nd in garnet VR43469and in cpx YK1914 is worse than for other samples Thismay be due to inhomogeneities related to late Nd diffusionpossibly from grain boundaries containing a kimberlitecomponent The reproducibility of the 176Hf177Hf data isgenerally poorer than that of Nd which may reflect a com-bination of matrix effects caused by imperfectly purifiedsolutions and sample heterogeneity Parent^daughterratios are available only from in situ analyses which arenot precise enough to allow meaningful model age orinitial isotope ratio calculations (ie at the time of kimber-lite eruption) Because the Lac de Gras kimberlites arerelatively young (55^56 Ma Graham et al 1999) age cor-rection would result in only a minor change in isotopecomposition which is inconsequential in view of the rangeof compositions displayed by most samples and hence hasno effect on the conclusions drawnClinopyroxenes in five volatile-free eclogites and a

pyroxenite have Sr isotope ratios ranging from 0701662 to0703546 this range is similar to or somewhat lower thanthat of the depleted mantle at the time of kimberliteemplacement (0703 Bell amp Tilton 2002) Clinopyroxenein the phlogopitized sample has a distinctly high 87Sr86Srof 0751186Clinopyroxene controls 56^92 of the Nd budget (Nd

concentration in cpx weighted by mode) Neodymiumisotope ratios range from 0511660 to 0514844 Thiscorresponds to eNd variation (per 10 000 deviationof 143Nd144Nd in the sample from 143Nd144Nd in the chon-dritic mantle model reservoir) from 17 to thorn43 cpx inthe low-Mg eclogites have higher eNd than those in

high-Mg eclogites Garnets have generally have higher143Nd144Nd from 0512984 (eNdfrac14thorn 7) to 0520965(eNdfrac14thorn 162) Garnets in the two volatile-rich sampleshave the highest ratios 0522790 for garnet in the calcite-bearing sample VR67112b (eNdfrac14thorn198) and 0525721 inthe phlogopitized eclogiteYK1911 (eNdfrac14thorn 255)Garnet contributes 18^32 and cpx 48^78 of the Hf

budget The proportion controlled by rutile is 4^30(Aulbach et al in preparation) Measured 176Hf177Hf incpx ranges from 0281263 in phlogopite-bearing sampleYK1911 (eHffrac1453) to 0286130 in low-Mg eclogiteVR43469 (eHffrac14thorn119) and in garnet from 0283958 inpyroxeniteYK1915 (eHffrac14thorn42) to 0295718 in low-Mg eclo-gite YK1926 (eHffrac14thorn458) Hafnium in garnet in the phlo-gopite^apatite-bearing sample is extremely radiogenic(176Hf177Hffrac14 0354367) corresponding to an eHf of thorn2532which is close to a value determined for a garnet separatefrom a Roberts Victor eclogite (eHffrac14thorn2561 Jacob et al2005) Garnet in the calcite-bearing sample has extremelyradiogenic Nd and Hf (eNdfrac14thorn193 eHffrac14thorn2516) Garnetin pyroxenite YK1915 also has radiogenic Nd and Hf(eNdfrac14thorn27 eHffrac14thorn42)Despite the large uncertainties for reconstructed whole-

rocks (calculated as the square root of the sum of squareduncertainties of two standard errors for garnet and cpxanalyses combined with 20 propagated uncertainty onthe proportion of Nd or Hf respectively contributed byeach mineral) the isotopic data are consistent in thatthey correlate with the independently determinedparent^daughter ratios (1) Low-Mg eclogites have higherSmNd reflected in the steeper positive slope of theirN-MORB-normalized REE patterns (Fig 11a) comparedwith the high-Mg eclogites (Fig 11b) Accordingly recon-structed low-Mg eclogites have higher present-day eNd

(0 to thorn47) than high-Mg eclogites (4 and 17) (2) Low-Mg eclogite YK1943 has very high Nb and LREE abun-dances that could have been added during metasomatismthat also enhanced its Hf abundance This sample has thelowest eNd and eHf of the low-Mg eclogites (0 and thorn20respectively) In contrast apparently unenriched low-Mgeclogites have more radiogenic Nd and Hf (eNdfrac14thorn47and thorn9 eHffrac14thorn128 and thorn113 respectively)The Sr^Nd isotope diagram reveals that eNd in cpx

varies from 19 to 43 at nearly constant depleted-mantle-like 87Sr86Sr (Fig 12a) In the Nd^Hf isotope diagram(Fig 12b) the reconstructed whole-rocks lie on curvedmixing lines if the lsquomixingrsquo proportions of Nd and Hffor the two lsquoend-memberrsquo minerals are very different(eg YK1911) Garnets have more radiogenic Nd andHf than coexisting cpx All reconstructed volatile-freeeclogites plot in the radiogenic field with regard to Hfor are isotopically similar to the depleted mantle Twohigh-Mg eclogites and one low-Mg eclogite have unradio-genic Nd one high-Mg eclogite has radiogenic Nd and

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1862

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 5 Individual and averaged Sr Nd and Hf isotope ratios obtained from garnet and cpx solutions

Sample 87Sr86Sr 2SE2SD 143Nd144Nd 2SE2SD eNd 2SE2SD 176Hf177Hf 2SE2SD eHf 2SE2SD

Low-Mg eclogites

VR43469 cp 0702661 0000010 0514844 0000006 43 01 0286130 0000012 119 04

VR43469 gt 0515508 0000014 56 03 0286506 0000012 132 04

VR43469 gt r 0515453 0000010 55 02 0286548 0000011 134 04

VR43469 gt r 0515384 0000007 54 01 0287244 0000010 158 04

VR43469 gt r 0515487 0000014 56 03 0286996 0000014 149 05

VR43469avg gt 0515458 0000109 55 21 0286823 0000715 143 253

WR VR43469 0515059 0000273 47 5 0286405 0001079 128 38

YK1926 cp 0703546 0000011 0513043 0000007 8 0 0283154 0000009 13 03

YK1926 gt 0513364 0000008 14 01 0295718 0000005 458 02

WR YK1926 0513120 0000026 9 1 0285980 0000716 113 25

YK1943 cp 0512143 0000007 10 01 0282213 0000005 20 02

YK1943 gt 0514461 0000012 36 02 0287510 0000008 168 03

YK1943 gt r 0514491 0000009 36 02 0287332 0000005 161 02

YK1943 gt r 0514517 0000010 37 02 0287290 0000007 160 02

YK1943avg gt 0514489 0000056 36 11 0287377 0000233 163 82

WR YK1943 0512625 0000228 0 5 0283138 0000879 13 31

High-Mg eclogites

VR43480 gt 0520965 0000017 162 03 0284950 0000010 77 04

VR67360 cp 0283509 0000074 26 26

VR67360 cp r 0283433 0000031 23 11

VR67360avg cp 0701662 0000018 0511660 0000010 19 02 0283471 0000108 25 38

VR67360 gt 0512984 0000101 7 20 0287429 0000180 165 64

WR VR67360 0511761 0000132 17 3 0284734 0000469 69 17

YK1949 cp 0703019 0000013 0512231 0000009 8 02 0282513 0000009 9 03

YK1949 gt 0513511 0000010 17 02 0290297 0000042 266 15

YK1949 gt r 0513508 0000012 17 02 0289947 0000015 254 05

YK1949 gt r 0513552 0000014 18 03

YK1949avg gt 0513523 0000049 17 10 0291569 0000494 311 175

WR YK1949 0512446 0000103 4 2 0284622 0000847 65 30

Volatile-rich eclogites

VR67112b gt 0522790 0000008 198 02 0354367 0000036 2532 13

YK1911 cp 0751186 0000009 0512165 0000007 9 01 0281263 0000009 53 03

YK1911 gt 0525721 0000013 255 03 0288051 0000012 187 04

WR YK1911 0513940 0000342 25 7 0281414 0000018 48 1

Pyroxenites

YK1914 cp 0703015 0000034 0513283 0000041 13 08

YK1914 cp r 0702956 0000005 0513200 0000027 11 05

YK1914 cp r 0513527 0000059 17 11

YK1914 cp r 0513045 0000010 8 02

YK1914avg cp 0702986 0000083 0513264 0000402 12 78

YK1915 gt 0514003 0000011 27 02 0283958 0000021 42

Analyses of full procedural replicates are designated with an lsquorrsquo after sample name SE standard error (internal errorsingle analyses) SD standard deviation (external error averaged multiple analyses) Italics used for reconstructed whole-rocks (mineral isotopic compositions and concentrations weighted by modes) Uncertainties for reconstructed whole-rocksinclude 2SE or 2SD uncertainties on each mineralrsquos isotope composition combined with 20 uncertainty in the amountof Nd or Hf contributed by the constituent minerals eNd and eHf are per 10 000 deviation from model chondritic mantle[model of Wasserburg et al (1981) and Blichert-Toft amp Albarede (1997) respectively]

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1863

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

εHf

εNd

Ekati eclogitesminus100

0

100

200

300

400

500

minus40 0 40 80 120 160 200 240 280

gt in YK1915

gt in VR67112bεNd=198 εHf=2532

minus40

0

40

80

120

160

minus40 minus20 0 20 40 60

minus60

minus40

minus20

0

20

40

070 071 072 073 074 075 076

North Americankimberlite field

εNd

87Sr86Sr

cpx in YK1914

Ekatieclogites

DM

DM

DM

DM

(a)

(b)

pyroxenites

low-Mg

high-Mg

volatile-rich

eclogites

Diavik low-TA154 S kimberlite

cpx in YK1911

Fig 12 (a) Sr^Nd for cpx and (b) Nd-Hf isotope diagram for garnet cpx and reconstructed whole-rocks (symbols with bold outline) Finedashed lines in (b) connect data points from the same sample with garnets having more radiogenic isotopic compositions Uncertainties forreconstructed whole-rocks combine 20 modal uncertainty with uncertainties on each mineralrsquos isotope composition Error bars on individualminerals may be smaller than the symbols Also shown are the isotopic compositions of the depleted mantle (DM DePaolo 1981 Griffin et al2000 Bell amp Tilton 2002) Range of 87Sr86Sr in eclogites from Ekati (Lac de Gras area 100 km north of Diavik Jacob et al 2003) in NorthAmerican kimberlites and compositions of low-temperature eclogites from Diavik and of Lac de Gras A154 South kimberlite fromSchmidberger et al (2007)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1864

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

one low-Mg eclogite has eNd similar to the depleted mantle(Fig 12b)Clinopyroxene in the phlogopitized eclogite YK1911

has unradiogenic Nd and Hf and strongly radiogenicSr whereas the coexisting garnet has radiogenic Nd andHf Although this sample contains 50 vol phlogopiteits whole-rock isotopic composition was calculated

from garnet and cpx alone because the contributionfrom phlogopite to the Hf and Nd budget is 52 and51 respectively (unpublished data) The whole-rockhas radiogenic Nd but has lower eHf than anyother sample analysed in accord with the extreme Hfabundance (53 ppm) and low LuHf of the cpx

DISCUSS IONOrigins of eclogites beneath the SlaveCratonHigh-Mg and high-Ca eclogites

With regard to their major- and trace-element contentsthe high-Mg and high-Ca eclogites from Lac de Gras arebroadly similar to variably evolved oceanic gabbros(Fig 13b and c) which may indicate an origin as subductedoceanic crust Compatible trace-element abundancesare useful to deduce potential crustal protoliths as theseelements are less susceptible to post-formation modification(Jacob 2004) The flat HREE patterns in the high-Mg andhigh-Ca eclogites are compatible with a low-pressureorigin as the oceanic crust is a product of moderate degreesof partial melting mainly within the spinel stability field(eg Presnall et al 2002) whereas melting with residualgarnet would qualitatively be expected to produce stronglyfractionated REE patterns Positive Sr and Pb anomaliesalthough compatible with a plagioclase signature couldalso be due to cryptic metasomatism (Dawson 1984) ofthe eclogites during their residence in the lithosphericmantle These anomalies are probably not related to kim-berlite infiltration as the whole-rock eclogite compositionswere reconstructed from analyses of fresh mineral coresLow ZrSm similar to that in primitive gabbros is

observed for many of the high-Mg and high-Caeclogites (Fig 13b and c) and is independent of the ratioof garnetcpx and whether or not the whole-rock wasreconstructed with rutile (Electronic Appendix 10) Thenegative trend of Mg-number against Yb in garnet(Fig 14) argues against igneous processes involving garnetaccumulation (Taylor amp Neal 1989) as suggested forexample for some high-Mg eclogites from West Africa(Barth et al 2002b) Rather the anti-correlation suggeststhat olivine fractionation played a role in the formation ofthe protoliths Fractionating olivine would excludeYb (Dolivinebasalt

frac14 002 Nikogosian amp Sobolev 1997)and incorporate Mg relative to Fe (Dfrac1415 and 05 respec-tively Taura et al1998) leading to progressive depletion ofMg and enrichment of Yb in the residual melt and henceanti-correlated Mg and Yb in the whole-rock This isinherited after eclogitization All eclogite types (exceptVR43452) plot along the olivine fractionation trend sug-gesting that they formed by a similar processHigh-Mg eclogites have lower

PREE than high-Ca

eclogites and some appear to have positive Eu anomalies

high-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

gabbrosSE Indian ridge

gabbrosOman ophiolite

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

high-Mg

RockNMORB(a)

VR43452

YK1949

low-Mg

boninite-like melts

01

001

1

10

(b)

01

001

1

10

(c)

01

001

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

Fig 13 Trace-element patterns of (a) low-Mg (b) high-Mg and(c) high-Ca eclogites compared with boninite-like volcanic rocksfrom the Isua greenstone belt (Polat et al 2002) gabbros from theOman ophiolite (Benoit et al 1996) and the SE Indian Ridge (Hartet al 1999) normalized to N-MORB of Sun amp McDonough (1989)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1865

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(EuEu up to 14) similar to more primitive cumulates(troctolite gabbros and olivine gabbros) reported in the lit-erature (Fig 13b) Although the sizes of these Eu anomaliesare not outside the analytical uncertainty their correlationwith CeYb in reconstructed eclogites supports their signif-icance (r2frac14066 nfrac14 4) The more primitive nature ofhigh-Mg eclogites is consistent with higher Mg-numbersrelative to high-Ca eclogites and may be due to olivineaccumulation during formation of their protolithsThe two samples with the lowest HREE abundanceshave the highest EuEu which is compatible with varia-tions in plagioclase modesMore evolved protoliths with higher pyroxeneplagio-

clase plus trapped melt could account for small or absentpositive Eu anomalies and higher

PHREE of high-Ca

eclogites (Fig 13c) Although the reason for the frequentassociation of graphite or diamond with high-Ca eclogitesis unclear (Aulbach et al 2003) the frequent occurrence ofkyanite in high-Ca eclogites also observed at Ekati justnorth of Lac de Gras has been argued to support a crustalprotolith (Jacob 2004) Kyanite andor corundum havebeen suggested to exsolve from high-temperature high-Alclinopyroxenes (Caporuscio amp Smyth 1990) Eclogitesfrom Lac de Gras show no exsolution textures suggestiveof such an origin although recrystallization can obscureor destroy microstructural evidence for exsolution(eg Griffin et al 1984)Modelling shows that qualitatively plagioclase accumu-

lation from a low-pressure melt derived from MORBmantle can account for positive Pb Sr and Eu and negativeZr anomalies relative to N-MORB whereas cpx accumula-tion or trapped melt increases Zr Y MREE and HREE

abundances (Fig 15) Some melt addition (for exampleby entrapment) seems to be required to increase all trace-element abundances to those observed for the eclogitesAlthough high-Ca and high-Mg eclogites could representshallow and deep portions of the same oceanic crust thehigher average temperatures obtained for the high-Caeclogites the frequent occurrence of carbon in high-Caeclogites and its absence in high-Mg eclogites the strongLREEHREE fractionation observed for six of seven high-Ca eclogites and the narrow range of compositions of thehigh-Mg eclogites suggest that they may have protolithsthat are not cogenetic Some discrepancies in particularaffecting the incompatible elements may be due to post-crystallization processes acting upon the protoliths suchas dehydration melting and later metasomatism that isevident in peridotites from the same locality (Griffin et al1999)Low K2OTh for high-Ca and high-Mg eclogites

(average 08 Table 4) relative to assumed crustal

0

2

4

6

8

10

30 40 50 60 70 80

Mg

VR43452

Yb

GARNET

pyroxenites

low-Mghigh-Mghigh-Ca

eclogites

VR19673ecl-g2

Fig 14 Garnet Mg-number vs Yb (ppm including 2s uncertainty)Lines qualitatively show separate correlations for eclogites (exceptoutlierVR43452) and pyroxenites (except outlierVR19673ecl-g2)

high-Ca

high-Mg

001

01

1

10

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

melt08 melt02 cryst cpx01 melt09 cryst

plag

(a)

001

01

1

10(b)

RockNMORB

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 15 Reconstructed (a) high-Ca and (b) high-Mg eclogitescompared with compositions of plagioclase and cpx crystallizingfrom a MORB melt modelled using distribution coefficients ofNorman et al (2005)The MORB was calculated as a 10 equilibriummelt of MORB pyrolite [composition from Kelemen et al (1993)except Pb Pr Sr Y and Lu which were calculated as residues from1 fractional melting of pyrolite from McDonough amp Sun (1995)using melt modes of Kinzler (1997) and distribution coefficients ofNorman et al (2005)] Also shown are mixtures of crystals (consistingof 50 plagioclase plus 50 cpx) and melt in the proportions 0901and 0208

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1866

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

precursors (eg gabbros SE Indian Ridge average 22Hart et al 1999) may reflect the preferred loss of fluid-mobile elements during dehydration (Becker et al 2000)although large uncertainties on these low-abundance ele-ments in the constituent minerals and reconstructedwhole-rocks makes this interpretation less than robustThe REE patterns of most high-Mg eclogites are too flatto represent melting residues as partial melting shouldproduce low LREEHREE (Electronic Appendix 10) Incontrast the marked LREE depletion of five of sevenhigh-Ca eclogites (LaLufrac14 02^08) contrasts with that ofassumed evolved gabbroic protoliths (LaLufrac1415^32Hart et al 1999) and can be modelled as a consequence ofpartial melting during subduction after eclogitization (iein the presence of garnet which retains HREE relative toLREE eg Barth et al 2002a) has taken place (Aulbachet al 2003)Our results are compatible with previous studies on

eclogite xenoliths from the Slave Craton includingthose from the same kimberlite that have linked thesesamples or some of these samples to Paleoproterozoicsubduction events (the 195^191Ga formation of theHottah terrane and the 18Ga formation of the GreatBear magmatic arc) in the Slave craton (Griffin et al1999 Kopylova et al 1999 Heaman et al 2002 2006Schmidberger et al 2005 2007)The compositional features discussed above although

compatible with crustal protoliths are not necessarilydefinitive of a subduction origin However if direct crystal-lization from basaltic melts at high pressure (ie mantledepths) is invoked instead specific conditions would haveto be met to satisfy the observations The Mg-number ineclogites from Lac de Gras are lower than those expectedof primary high-pressure mantle melts (eg Walter 1998OrsquoHara amp Herzberg 2002) requiring crystallization athigh pressures after olivine fractionation This originwould further require shielding from peridotitic wallrockwith which the fractionated melts would no longer be inequilibrium (Barth et al 2001) In such a setting the lowrelative abundances of fluid-mobile elements comparedwith those of less fluid-mobile elements and of LREErelative to HREE (Table 4) would reflect an expulsion offluids during crystallization of the protoliths and latermelt mobilization respectively The eclogites clearly donot represent garnet plus cpx cumulates which wouldrequire a positive correlation of Mg-number with Yb thatis at variance with the observed negative correlation ineclogitic garnets from this study (Fig 14)

Low-Mg eclogites

Garnets in low-Mg eclogites compositionally overlap withthose in massive eclogites from Jericho (NW of the Lacde Gras kimberlites) that were suggested to have a high-pressure intrusive origin based on the similarity of theirmajor-element compositions to those of Group A eclogites

fromsouthernAfrica forwhichamantleoriginhasbeenpro-posed (Kopylova et al1999) Garnets in low-Mg eclogites inthe present study have major-element compositions moresimilar to garnets in groupB rather thangroupAeclogitesThe positive slope in N-MORB-normalized HREE

abundances displayed by the low-Mg eclogites is clearlydistinct from other eclogite types and is similar to those ofboninite-like second-stage melts that are closely associatedwith ultramafic volcanics (Smithies et al 2004) (Fig 13a)The range of Mg-numbers in the low-Mg eclogites (55^66)overlaps with that of Archaean second-stage melts fromIsua (58^75) and Abitibi (52^82 Smithies et al 2004)These similarities might indicate that the low-Mg eclogiteshave an origin in a previously depleted mantle source suchas supra-subduction zone mantle Subchondritic ZrHf ofthe low-Mg eclogites contrasts with the supra-chondriticZrHf of the high-Mg and high-Ca eclogites (Aulbach et

al in preparation) and is consistent with previous cpx-controlled depletion in an arc mantle source This sourcecould have been subsequently remelted to generate theboninite-like precursor to the low-Mg eclogites as sug-gested for the protoliths of eclogites from Udachnaya(Jacob amp Foley 1999)Although some second-stage melts to which the low-Mg

eclogites are similar are closely associated with komatiitesand could contain a plume component the radiogenic ordepleted mantle-like Nd and Hf and unradiogenic Sr inthe low-Mg eclogites do not support plume involvement

Trace-element and Sr^Nd^Hf isotopeconstraints on post-formationmetasomatismSeveral low-Mg and high-Ca eclogites show negativeslopes in the LREE which may be indicative of interactionwith small amounts of LREE-enriched melt (Navon ampStolper 1987) amounting to cryptic metasomatism(ie not accompanied by crystallization of new phases) inthe nomenclature of Dawson (1984) High Nb abundancesrelative to elements of similar compatibility displayed bysome low-Mg and high-Ca eclogites (Fig 10) could resultfrom overestimation of rutile modes However high-CaeclogiteVR40345 is rutile-free and has been reconstructedwithout rutile its high Nb abundance is accompanied byLREE enrichment indicating that metasomatic additionmay be responsibleLow-Mg eclogites show a positive correlation of La with

Nb (Electronic Appendix 11a) indicating concomitantaddition of LREE and Nb This suggests that the metaso-matic agent was a solute-rich fluid or silicate silicic melt ashydrous fluids and carbonatites do not carry enough HFSEto affect the budget of eclogites (Green amp Wallace 1988Klemme et al 1995 Adam et al 1997 Stalder et al 1998)Of the samples that are enriched in Nb all but one (high-Ca eclogiteVR50909) show a simultaneous enrichment in

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1867

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Hf (Electronic Appendix 11b) Together these trace-element relationships suggest that both SmNd and LuHfwere lowered in these samples Their low parent^daughterratios contrast with higher ratios in unenriched samplesand resulted in the ingrowth of a range of Nd and Hf iso-tope compositions Compared with the low-Mg and high-Ca eclogites the high-Mg eclogites have a narrow range ofLa and Nb concentrations which correlates with a narrowrange in Nd and Hf isotope compositions (Fig 12)Precise model ages cannot be calculated because parent^

daughter ratios are available only from in situ analyseswhich have relatively large uncertainties Howeverevidence for long-term enrichment and depletion may bediscerned Low-Mg eclogites even those that are enrichedin LREE have relatively high SmNd as a result of havingoverall positive slopes in their REE patterns relative toMORB whereas the high-Mg eclogites have relativelyflat REE patterns and therefore lower SmNd (Fig 11)This translates into lower eNd for high-Mg eclogites thanfor low-Mg eclogites In contrast they show no systematicdifference with respect to 176Hf177Hf Because the low- andhigh-Mg eclogites are proposed to have different origins asdiscussed above and therefore did not necessarily have thesame initial isotopic composition and formation age theseisotopic differences cannot be used to estimate their agesA long-term coupling of trace-element and isotopic

abundances is also indicated by the fact that low-Mg eclo-gite YK1943 which has experienced Nb and LREE addi-tion has lower eNd (0) and eHf (thorn13) than low-Mg eclogiteVR43469 with higher LuHf and SmNd (eNdfrac14thorn47eHffrac14thorn128) If low-Mg eclogites YK1943 and VR43469originally had identical initial Nd and Hf isotope composi-tions and if the parent^daughter ratios are accurate andwere modified only once during their evolution shortlyafter metamorphism (when their isotopic compositionwould have been reset) this overprint can be calculatedfor both isotopic systems to have occurred at 17GaThis would suggest that Hf and Nd were added duringthe same event In contrast high-Mg eclogites VR67360and YK1949 would have had identical Ndisotope compositions at 10Ga whereas their Hf isotopecompositions are disturbed in that the sample with thehigher parent^daughter ratio has a lower 176Hf177Hf thanthe sample with the lower parent^daughter ratio Theseresults suggest that the LREE and HFSE enrichmentobserved in different samples occurred during separateevents that did not affect the eclogites equallyExtreme HFSE enrichment leading to the growth of

zircon and rutile and LREE addition leading to thegrowth of apatite in eclogites from the Jericho kimberlitein the northern Slave craton have been ascribed to twoseparate metasomatic events at around 18Ga (quasi-coincident with the age of formation and metamorphism)and 13^10Ga respectively (Heaman et al 2002 2006

Schmidberger et al 2005) The authors associated theseevents with Paleoproterozoic subduction and aMesoproterozoic thermal disturbance in the Slave litho-spheric mantle possibly linked to the Mackenzie igneousevent Mesoproterozoic Nd model ages mostly between13 and 14Ga were reported for eclogites from thenearby Ekati kimberlites (Jacob et al 2003) Although wedo not observe zircon or apatite in volatile-free eclogitesfrom Lac de Gras and the age constraints obtained fromour data are weak they support a link between theseevents described for other eclogite populations and themetasomatic enrichment observed in our studyThe Sr isotope data in the present study are similar to

those of Schmidberger et al (2007) who suggested that thedepleted-mantle-like values are consistent with derivationof the protoliths from an oceanic mantle source However87Sr86Sr in volatile-free eclogites in the present study isremarkably constant considering their Nd and Hf isotopevariability (Fig 12) and the variability of trace-elementpatterns (Fig 10) Given the evidence for dehydration andmelting in many of the eclogites fractionation of highlyincompatible Rb from Sr and differential ingrowth ofradiogenic Sr would be expectedThis may indicate a rela-tively recent overprint and considering the similarityof 87Sr86Sr to present-day depleted mantle which is signif-icantly lower than 87Sr86Sr of North American kimber-lites a juvenile (asthenospheric) source

PyroxenitesPyroxenites from Lac de Gras often show relationshipsin their major-element contents that are different fromthose observed in the eclogites For example pyroxeniticgarnets show a negative correlation between CaO andNa2O contents whereas these oxides are positively corre-lated for garnets in eclogites (Fig 5) CaO is well correlatedwith C2O3 in pyroxenitic garnets but not in eclogitic gar-nets (Fig 4) Mg-number is positively correlated withYb abundance for pyroxenitic garnets but anti-correlatedfor eclogitic garnets (Fig 14) This suggests that the pyrox-enites in the present study are not related to the eclogitesby fractionation or accumulation processes as proposedfor the origin of some pyroxenites (Foley et al 2003)Pyroxenites have also been interpreted as the products

of crystal segregation in magma channels with alteredoceanic crust as the magma source this would explaintheir Eu anomalies and stable and radiogenic isotopecompositions (Pearson et al 1993 Becker 1996) As thehigh-pressure liquidus phases of most terrestrial basaltsare cpx and garnet (OrsquoHara amp Herzberg 2002) partialmelts of eclogites with basaltic composition wouldcrystallize cpx and garnet but not opx More than half ofthe pyroxenites from Lac de Gras contain opx suggestingthat they probably did not crystallize directly fromeclogite-derived melts

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1868

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Low-Cr websterites and pyroxenites could be crystalli-zation products of mafic melts that interacted with perido-titic mantle thus becoming more refractory The positivecorrelation of Yb and Mg-number in garnet would allowan origin by fractional crystallization from such meltsAlternatively they formed by reaction of silicic (possiblyeclogite-derived) melts with mantle peridotites(eg Kelemen et al 1993 Aulbach et al 2002) This agreeswith the similarity of hypothetical melts in equilibriumwith pyroxenitic garnets in most samples to eclogite-derived melts (Electronic Appendix 12) Niobium concen-trations in all hypothetical melts are much higher than ineclogite-derived melts regardless of whether or not rutileis assumed to be in the residue The same is true for LaCe and Zr concentrations in some samples which mayindicate that NbZr and LREE were added to thepyroxenites some time after their formation as discussedin the previous section for some eclogites Because mini-mum ages for pyroxenitic rutile with an unradiogenicHf isotope composition (calculated by projecting the176Hf177Hf to the model depleted mantle isotopic evolutioncurve assuming zero ingrowth) range to 178066Ga(Aulbach et al in preparation) it appears likely that thepyroxenites were not formed recently (eg during kimber-lite-related or precursory melt migration) and that some ofthe samples are at least MesoproterozoicGarnet pyroxeniteYK1952 consists of schlieren of equili-

brated garnet pyroxenite in a lsquomatrixrsquo of spongy cpxfine cpx^opx intergrowths embayed garnet and meltpatches (Fig 2f) Garnet in the pyroxenite schlieren hasvery low LREE and HFSE abundances Orthopyroxenein the matrix has distinctly higher Mg-number andCr-number and contains more CaO than opx from thepyroxenite schlieren whereas opx rims in the schlierenare compositionally similar to opx in the matrix The con-siderable compositional and textural disequilibrium inthis sample suggests that this is a young feature causedby reaction with fluids shortly before or at the time ofentrainmentAs a consequence of the small sample sizes of most of

the pyroxenites studied here a comprehensive isotopicdataset is not available Clearly more such data areneeded in addition to stable isotope data to obtain betterconstraints on the origins of this rock type Consideringthe microstructural and compositional variability ofthe pyroxenites and major-element relationships oppositeto those observed for eclogites it appears certain that pyr-oxenites form by a variety of processes some of which areunrelated to the formation of eclogites and their protoliths

Volatile-rich eclogitesVolatile-rich eclogites from Lac de Gras display the mostextreme Hf^Nd isotope systematics of all the analysedsamples Garnets in phlogopitized (YK1911) and in cal-cite-bearing eclogite (VR67112b) have the most radiogenic

Nd (eNdfrac14 2486 and 1927 respectively) Garnet inVR67112b has extremely high eHf (2516) Clinopyroxene inYK1911 has the lowest eHf measured for any mineralseparate from Lac de Gras (53) in accord with the veryhigh Hf abundance in its cpx (5 ppm) In contrastgarnet in this sample has radiogenic Hf (eHffrac14 136) Thisdiscrepancy is at least in part due to their relatively lowequilibration temperatures (7698C and 8558C respec-tively) which allowed the isotopic compositions of cpxand garnet to evolve separately at their low andhigh parent^daughter element ratios respectively forextended periods of time and possibly since formationAccordingly the cpx in YK1911 gives a minimum Hf ageof 27GaSlab-derived fluids could have been introduced during

the 27Ga collision of the western Slave Craton domainincluding the thick plume-derived deep lithosphericmantle layer and the eastern domain with its highlydepleted lithosphere This would also be consistent withthe observation that volatile-rich eclogites occur at shallowdepths in the mantle (TKroghfrac14 685^8558C correspondingto depths of 85^115 km along a 40 mWm2 geotherm)where the bulk of dehydration and production of solute-rich melts in a subducting slab might be expected to occur(Becker et al 2000 Hermann et al 2006) Their negativeEu anomalies might indicate that the fluid was derivedfrom the shallow portion of the slab that experiencedplagioclase fractionation The presence of carbonate inVR67112b requires a carbonated sourceHigh 87Sr86Sr in cpx in YK1911 is probably the conse-

quence of isotopic equilibration with high-RbSr phlogo-pite after long-term radiogenic ingrowth while Hfaddition led to retarded ingrowth of radiogenic HfThese systematics combined with radiogenic Nd appear

inconsistent with the involvement of sediments in the pet-rogenesis of this sample (eg Perfit et al 1980 Vervoortet al 1999) or a hydrous fluid such as those responsible forMARID metasomatism in the Kaapvaal craton (Konzettet al 1998)Whatever the origin of the fluids the apparentHf mobility suggests that they were not in equilibriumwith a rutile-bearing source and also that they weresolute-rich in nature (ie hydrous melts able to dissolveHFSE) and the relatively high HREE abundances of thevolatile-rich eclogites indicate a garnet-free source suchas amphibolite This allowed evolution of the sample atrelatively high time-integrated SmNd

SUMMARY AND CONCLUSIONS

(1) The Lac de Gras kimberlites carried abundant eclo-gite and pyroxenite xenoliths to the surface Threemain eclogite types are distinguished in terms ofmajor-element composition and trace element

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1869

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

patterns high-Mg high-Ca and low-Mg eclogitesThe garnets of all eclogite types have anti-correlatedMg-number and Yb pointing to olivine fractionationas a controlling factor in the formation of their pre-cursor rocks Their flat or positively sloped HREEpatterns qualitatively indicate that the melt parentalto the precursor rocks was not in equilibrium with agarnet-bearing residue

(2) Reconstructed high-Mg eclogites and high-Caeclogites have flat HREE patterns positive Pb andSr anomalies and low ZrSm similar to oceanic gab-bros reported in the literature High-Mg eclogiteshave variably low HREE contents and some havesmall positive Eu anomalies possibly suggestive ofmore primitive relatively plagioclase-rich protolithswhereas high-Ca eclogites have higher HREE abun-dances and small or absent Eu anomalies that requiremore cpx- andor melt-rich protoliths if a subductionorigin is assumed

(3) Low-Mg eclogites have distinctive trace-element pat-terns that show a marked positive slope in the HREEand subchondritic ZrHf They resemble someArchaean second-stage melts (lsquoboninitesrsquo) and theirprotoliths appear to have formed from a previouslydepleted source

(4) After formation the eclogites were subject to a varietyof processes Most have low ratios of fluid-mobile vsless fluid-mobile elements possibly as a result of dehy-dration of the protoliths In addition some of the low-Mg and high-Ca eclogites have low LREEHREEconsistent with partial melt lossWhereas the low-Mgeclogites show concomitant enrichment in La Nb andZr some high-Mg and high-Ca eclogites show evi-dence for variable and unsystematic (decoupled)LREE and Nb addition

(5) Isotopically the trace-element characteristics of thedifferent eclogite types translate into lower eNd forhigh-Mg eclogites (negative slope in MREEfrac14 lowerSmNd) than for low-Mg eclogites (positive slope inMREEfrac14higher SmNd) Within the low-Mg groupsamples that show evidence for metasomatism (lowLuHf and SmNd) have lower eNd and eHf than asample that was apparently not metasomatized BothNd and Hf addition can be dated to 17Ga if identi-cal initial isotope compsitions are assumed for thelow-Mg eclogites In contrast the relatively constantand depleted mantle-like 87Sr86Sr of eclogitic cpxdespite variable Nd and Hf isotope compositions andparent^daughter ratios suggests recent input of juve-nile material

(6) Garnet and cpx in pyroxenites show major-elementrelationships different from those in eclogitesTherefore these two rock types are probably notrelated by fractionation processes The presence of

opx precludes direct crystallization from slab-derivedmelts Pyroxenites may be the product of (slab-derived) melts hybridized during assimilation of peri-dotitic mantle

(7) Garnets in phlogopite- and carbonate-bearing eclo-gites have extreme isotopic compositions includingthe most radiogenic Nd and Hf in the datasetNegative Eu anomalies may point to the involvementof a crustal protolith Highly radiogenic Sr radiogenicNd and unradiogenic Hf in the reconstructedphlogopite-bearing sample combined with highHREE abundances and evidence for Hf mobilitymay indicate involvement of solute-rich fluids origi-nating in a garnet- and rutile-free source

ACKNOWLEDGEMENTSHelp in the clean laboratories and with the analyticalfacilities by S Elhlou C Lawson A Sharma P Wielandand M Zhang is greatly appreciated Bill Griffin isthanked for his support and for agreeing to reviewseveral versions of the manuscript despite disagreeingwith many of the conclusions presented here The paperbenefited greatly from the detailed and constructivereviews provided by the editors Gareth Davies andMarjorie Wilson and by Matthias Barth Larry Heamanand an anonymous reviewer This work was funded by aMacquarie University International PostgraduateAward and Postgraduate Research Fund (SA) by anARC SPIRT grant sponsored by Kennecott Canada Incand by ARC Large and Discovery Grants to SYOrsquoRAnalytical data were obtained using instrumentationfunded by ARC LIEF and DEST Systemic InfrastructureGrants industry partners and Macquarie UniversityThis is Publication 481 from the ARC National KeyCentre for Geochemical Evolution and Metallogeny ofContinents (wwwesmqeduauGEMOC)

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online

REFERENCESAdam J GreenT H Sie S H amp Ryan C G (1997) Trace element

partitioning between aqueous fluids silicate melts and mineralsEuropeanJournal of Mineralogy 9 569^584

Aulbach S Stachel T Viljoen K S Brey G P amp Harris J W(2002) Eclogitic and websteritic diamond sources beneath theLimpopo Beltccedilis slab-melting the link Contributions to Mineralogy

and Petrology 143 56^70Aulbach S GriffinW L Pearson N J OrsquoReilly S Y Kivi K amp

Doyle B (2003) Origins of eclogites beneath the central SlaveCraton Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA-0011

Aulbach S Griffin W L Pearson N J OrsquoReilly S Y amp Kivi K(2004a) Mantle formation and evolution Slave Craton constraints

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1870

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

from HSE abundances and Re^Os isotope systematics of sulfideinclusions in mantle xenocrysts Chemical Geology 208 61^88

Aulbach S Griffin W L OrsquoReilly S Y amp McCandless T E(2004b) Genesis and evolution of the lithospheric mantle beneaththe Buffalo HeadTerrane Alberta (Canada) Lithos 77 413^451

Aulbach S GriffinW L Pearson N J OrsquoReilly S Y amp Doyle B(2007) Lithosphere formation in the central Slave Craton(Canada) plume subcretion or lithosphere accretion Contributionsto Mineralogy and Petrology (in press) doi 101007s00410-007-0200-1

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2001)Geochemistry of xenolithic eclogites from West Africa Part IA link between low MgO eclogites and Archean crust formationGeochimica et Cosmochimica Acta 65 1499^1527

Barth M G Foley S F amp Horn I (2002a) Partial melting inArchean subduction zones constraints from experimentally deter-mined trace element partition coefficients between eclogitic miner-als and tonalitic melts under upper mantle conditions PrecambrianResearch 113 323^340

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2002b)Geochemistry of xenolithic eclogites from West Africa part 2Origins of the high MgO eclogites Geochimica et Cosmochimica Acta

66 4325^4345Beard B L Fraracci K NTaylor L A Snyder G A Clayton R

A Mayeda T K amp Sobolev NV (1996) Petrography andgeochemistry of eclogites from the Mir kimberliteYakutia RussiaContributions to Mineralogy and Petrology 125 293^310

Becker H (1996) Geochemistry of garnet peridotite massifs fromlower Austria and the composition of deep lithosphere beneatha Palaeozoic convergent plate margin Chemical Geology 134 49^65

Becker H Jochum K P amp Carlson R W (2000) Trace elementfractionation during dehydration of eclogites from high-pressureterranes and the implications for element fluxes in subductionzones Chemical Geology 163 65^99

Bell K amp Tilton G R (2002) Probing the mantle The story fromcarbonatites EOS Transactions American Geophysical Union 83275^277

Benoit M Polvecurren M amp Ceuleneer G (1996) Trace element andisotopic characterisation of mafic cumulates in a fossil mantlediapir (Oman ophiolite) Chemical Geology 134 199^214

BleekerW (2003) The late Archean record a puzzle in ca 35 piecesLithos 71 99^134

BleekerW Ketchum JW F JacksonV A amp Villeneuve M (1999)The central Slave Basement Complex Part I Its structural topol-ogy and autochthonous core Canadian Journal of Earth Sciences 361083^1109

Blichert-Toft J amp Albare de F (1997) The Lu^Hf isotope geochemis-try of chondrites and the evolution of the mantle^crust systemEarth and Planetary Science Letters 148 243^258

Blichert-Toft J Chauvel C amp Albare de F (1997) Separation of Hfand Lu for high precision isotope analysis of rock samples by mag-netic sector-multiple collector ICP-MS Contributions to Mineralogy

and Petrology 127 248^260Brey G P amp Kolaquo hler T (1990) Geothermobarometry in four-phase

lherzolites II New thermobarometers and practical assessment ofexisting thermobarometers Journal of Petrology 31 1353^1378

Brey G P Kolaquo hlerT amp Nickel K G (1990) Geothermobarometry infour-phase lherzolites I Experimental results from 10 to 60 kbJournal of Petrology 31 1313^1352

Caporuscio F amp Smyth J (1990) Trace element crystal chemistryof mantle eclogites Contributions to Mineralogy and Petrology 105550^561

Carbno G B amp Canil D (2002) Mantle structure beneath the SWSlave Craton Canada constraints from garnet geochemistry in theDrybones Bay Kimberlite Journal of Petrology 43 129^142

Coleman R G Lee E D Beatty L B amp Brannock W W (1965)Eclogites and eclogites their differences and similaritiesGeological Society of America Bulletin 76 483^508

Creaser R A Grulaquo tter H Carlson J amp Crawford B (2004)Macrocrystal phlogopite Rb^Sr dates for the Ekati propertykimberlites Slave Province Canada evidence for multiple intru-sive episodes in the Paleocene and Eocene Lithos 76 399^414

Davies R M GriffinW L Pearson N J Andrew A S Doyle BJ amp OrsquoReilly S Y (1999) Diamonds from the deep Pipe DO-27Slave Craton Canada In Gurney J J Gurney J L Pascoe MD amp Richardson S H (eds) Proceedings of the 7th International

Kimberlite Conference CapeTown Red Roof Design pp 148^155Davies R M Griffin W L OrsquoReilly S Y amp Doyle B J (2004)

Mineral inclusions and geochemical characteristics of microdia-monds from the DO27 A154 A21 A418 DO18 DD17 and RanchLake kimberlites at Lac de Gras Slave Craton Canada Lithos 7739^55

DavisW J amp Hegner E (1992) Neodymium isotopic evidence for thetectonic assembly of Late Archean crust in the Slave Provincenorthwest Canada Contributions to Mineralogy and Petrology 111493^504

Davis W J Jones A G Bleeker W amp Grutter H (2003)Lithosphere development in the Slave craton a linked crustal andmantle perspective Lithos 71 575^589

Dawson J B (1984) Contrasting types of upper-mantle metasoma-tism In Kornprobst J (ed) Kimberlites II The Mantle and

Crust^Mantle Relationships Amsterdam Elsevier pp 289^294DePaolo D J (1981) Neodymium isotopes in the Colorado Front

Range and crust^mantle evolution in the Proterozoic Nature 291193^196

Erlank A J amp Kushiro I (1970) Potassium contents of syntheticpyroxenes at high temperatures and pressures Carnegie Institution ofWashingtonYearbook 68 433^439

Foley S F Barth M G amp Jenner G A (2000) Rutilemelt partitioncoefficients for trace elements and an assessment of the influence ofrutile on the trace element characteristics of subduction zonemagmas Geochimica et Cosmochimica Acta 64 933^938

Foley S F Buhre S amp Jacob D E (2003) Evolution of the Archaeancrust by delamination and shallow subduction Nature 421 249^252

Graham I Burgess J L Bryan D Ravenscroft P J Thomas EDoyle B J Hopkins R amp Armstrong K A (1999) Explorationhistory and geology of the Diavik Kimberlites Lac de GrasNorthwest Territories Canada In Gurney J J Gurney J LPascoe M D amp Richardson S H (eds) Proceedings of the 7th

International Kimberlite Conference Cape Town Red Roof Designpp 262^279

Green D H amp Wallace M E (1988) Mantle metasomatism byephemeral carbonatite melts Nature 336 459^462

Green T H Blundy J D Adam J amp Yaxley G M (2000) SIMSdetermination of trace element partition coefficients betweengarnet clinopyroxene and hydrous basaltic liquids at 2^75 GPaand 1080^12008C Lithos 53 165^187

GriffinW L amp OrsquoReilly SY (2007) Cratonic lithospheric mantle Isanything subducted Episodes 30(1) 43^53

Griffin W L Wass S Y amp Hollis J D (1984) Ultramafic xenolithsfrom Bullenmerri and Gnotuk maars Victoria Australia petrol-ogy of a sub-continental crust^mantle transition Journal of

Petrology 25 53^87GriffinW L Doyle B J Ryan C G Pearson N J OrsquoReilly SY

Davies R Kivi K van Achterbergh E amp Natapov LM (1999)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1871

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Layered mantle lithosphere in the Lac de Gras Area SlaveCraton composition structure and origin Journal of Petrology 40705^727

Griffin W L Pearson N J Belousova E Jackson S E vanAchterbergh E OrsquoReilly SY amp Shee S R (2000)The Hf isotopecomposition of cratonic mantle LAM-MC-ICPMS analysis ofzircon megacrysts in kimberlites Geochimica et Cosmochimica Acta

64 133^147Griffin W L OrsquoReilly S Y Doyle B J Pearson N J

Coopersmith H Kivi K Malkovets V amp Pokhilenko N(2004) Lithosphere mapping beneath the North American plateLithos 77 873^922

Grulaquo tter H S Apter D B amp Kong J (1999) Crust^mantle couplingevidence from mantle-derived xenocrystic garnets In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 307^312

Hart S R Blusztajn J Dick H J B Meyer P S ampMuehlenbachs K (1999) The fingerprint of seawater circulationin a 500-meter section of ocean crust gabbros Geochimica et

Cosmochimica Acta 63 4059^4080Harte B amp Kirkley M B (1997) Partitioning of trace elements

between clinopyroxene and garnet Data from mantle eclogitesChemical Geology 136 1^24

Heaman L M Creaser R A amp Cookenboo H O (2002) Extremeenrichment of high field strength elements in Jericho eclogitexenoliths A cryptic record of Paleoproterozoic subduction partialmelting and metasomatism beneath the Slave craton CanadaGeology 30 507^510

Heaman L M Kjarsgaard B A amp Creaser R A (2004) The tem-poral evolution of North American kimberlites Lithos 76 377^397

Heaman L M Creaser R A Cookenboo H O amp Chacko T(2006) Multi-stage modification of the northern Slave mantlelithosphere Evidence from zircon- and diamond-bearing eclogitexenoliths entrained in Jericho Kimberlite Journal of Petrology 47821^858

Hermann J Cpandler C Hack A amp Korsakov A (2006) Aqueousfluids and hydrous melts in high-pressure and ultra-high pressurerocks implications for element transfer in subduction zones Lithos92 399^417

Hills DV amp Haggerty S E (1989) Petrochemistry of eclogites fromthe Koidu Kimberlite Complex Sierra Leone Contributions to

Mineralogy and Petrology 103 397^422Hoffman P F (1989) Precambrian geology and tectonic history of

North America In Bally A amp Palmer A (eds) The Geology of

North AmericaccedilAn Overview Boulder CO Geological Society ofAmerica pp 447^512

IrelandT R Rudnick R L amp Spetsius Z (1994) Trace-elements indiamond inclusions from eclogites reveal link to Archean granitesEarth and Planetary Science Letters 128 199^213

Irvine G J Pearson D G Kjarsgaard B A Carlson R WKopylova M G amp Dreibus G (2003) Evolution of the litho-spheric mantle beneath Northern Canada a Re^Os isotope andPGE study of kimberlite-derived peridotite xenoliths fromSomerset Island and a comparison to the Slave and Kaapvaalcratons Lithos 71 461^488

Isachsen C E amp Bowring S A (1994) Evolution of the Slave cratonGeology 22 917^920

Jacob D E (2004) Nature and origin of eclogite xenoliths from kim-berlites Lithos 77 295^316

Jacob D E amp Foley S F (1999) Evidence for Archean ocean crustwith low high field strength element signature from diamondifer-ous eclogite xenoliths Lithos 48 317^336

Jacob D Jagoutz E Lowry D Mattey D amp Kudrjavtseva G(1994) Diamondiferous eclogites from Siberiaccedilremnants ofArchean oceanic-crust Geochimica et Cosmochimica Acta 585191^5207

Jacob D E Fung A Jagoutz E amp Pearson D G (2003) Petrologyand geochemistry of eclogite xenoliths from the Ekati kimberlitesarea Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA_239

Jacob D E Bizimis M amp Salters V J M (2005) Lu^Hf and geo-chemical systematics of recycled ancient oceanic crust evidencefrom Roberts Victor eclogites Contributions to Mineralogy and

Petrology 148 707^720Jerde E A Taylor L A Crozaz G Sobolev NV amp SobolevV N

(1993) Diamondiferous eclogites from Yakutia Siberiaccedilevidencefor a diversity of protoliths Contributions to Mineralogy and Petrology

114 189^202Jones A G Ferguson I J Chave A D Evans R L amp

McNeice GW (2001) The electric lithosphere of the Slave cratonGeology 29 423^426

Kelemen P B Shimizu N amp Dunn T (1993) Relative depletionof niobium in some arc magmas and the continental-crust-partitioning of K Nb La and Ce during meltrockreaction in the upper-mantle Earth and Planetary Science Letters 120111^134

Kinzler R J (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition Application to mid-ocean ridge basalt petrogenesis Journal of Geophysical Research 102853^874

Klemme S van der Laan S R Foley S F amp Guenther D (1995)Experimentally determined trace and minor element partitioningbetween clinopyroxene and carbonatite melt under upper mantleconditions Earth and Planetary Science Letters 133 439^448

Konzett J Armstrong R A Sweeney R J amp CompstonW (1998)The timing of MARID metasomatism in the Kaapvaal mantle anion probe study of zircons from MARID xenoliths Earth and

Planetary Science Letters 160 133^145Kopylova M G Russell J K amp Cookenboo H (1999) Mapping the

lithopshere beneath the north central Slave Craton In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 468^479

Krogh E (1988) The garnet^clinopyroxene iron^magnesiumgeothermometerccedila reinterpretation of existing experimentaldata Contributions to Mineralogy and Petrology 99 44^48

Kusky T M (1989) Accretion of Archean Slave Province Geology 1763^67

LeCheminant A N amp Heaman L M (1989) Mackenzie igneousevents Canada Middle Proterozoic hotspot magmatismassociated with ocean opening Earth and Planetary Science Letters

96 38^48LeCheminant A N Richardson D G DiLabio R N W amp

Richardson K A (1996) La recherche de diamants au CanadaGeological Survey of Canada Open File 3228 268pp

MacGregor I D amp Carter J L (1970) The chemistry of clinopyrox-enes and garnets of eclogite and peridotite xenoliths from theRoberts Victor mine South Africa Physics of the Earth and PlanetaryInteriors 3 391^397

MacGregor I D amp Manton W I (1986) Roberts Victor eclogitesancient oceanic crust Journal of Geophysical Research 91 14063^14079

MacKenzie J M amp Canil D (1999) Composition and thermal evo-lution of cratonic mantle beneath the central Archean SlaveProvince NWT Canada Contributions to Mineralogy and Petrology

134 313^324

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1872

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

McCandlessT E amp Gurney J J (1989) Sodium in garnet and potas-sum in clinopyroxene criteria for classifying mantle eclogites InRoss J (ed) Kimberlites and Related Rocks Carlton Blackwellpp 827^832

McDonoughW F amp Sun S-S (1995) The composition of the EarthChemical Geology 120 223^253

Navon O amp Stolper E (1987) Geochemical consequences of meltpercolation the upper mantle as a chromatographic columnJournal of Geology 95 285^307

Nikogosian I K amp Sobolev AV (1997) Ion-microprobe analysis ofmelt inclusions in olivine experience in estimating the olivine^meltpartition coefficients of trace elements Geochemistry International 35119^126

Norman M D Griffin W L Pearson N J Garcia M O ampOrsquoReilly S Y (1998) Quantitative analysis of trace element abun-dances in glasses and minerals a comparison of laser ablationinductively coupled plasma mass spectrometry solution inductivelycoupled plasma mass spectrometry proton microprobe and elec-tron microprobe data Journal of Analytical Atomic Spectrometry 13477^482

Norman M D Garcia M O amp Pietruszka A J (2005) Trace-element distribution coefficients for pyroxenes plagioclase andolivine in evolved tholeiites from the 1955 eruption of KilaueaVolcano Hawairsquoi and petrogenesis of differentiated rift-zonelavas American Mineralogist 90 888^899

OrsquoHara M J amp Herzberg C (2002) Interpretation of trace elementand isotope features of basalts Relevance of field relations petrol-ogy major element data phase equilibria and magma chambermodeling in basalt petrogenesis Geochimica et Cosmochimica Acta 662167^2191

PadghamW A amp FysonW K (1992) The Slave Province a distinctcraton CanadianJournal of Earth Sciences 29 2072^2086

Pearson D G Davies G R amp Nixon P H (1993) Geochemical con-straints on the petrogenesis of diamond facies pyroxenites from theBeni Bousera peridotite massif North Morocco Journal of Petrology34 125^172

Pearson N J Griffin W L Doyle B J OrsquoReilly S Y vanAchterbergh E amp Kivi K (1999) Xenoliths from kimberlitepipes of the Lac de gras area Slave Craton Canada In GurneyJ J Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite ConferenceCapeTown Red Roof Design pp 644^658

Perfit N R Gust D A Bence A E Arculus R J amp Taylor S R(1980) Chemical characteristics of island-arc basalts implicationsfor mantle sources Chemical Geology 30 227^256

Polat A Hofmann A W amp Rosing M T (2002) Boninite-likevolcanic rocks in the 37^38 Ga Isua greenstone belt WestGreenland geochemical evidence for intra-oceanic subductionzone processes in the early Earth Chemical Geology 184 231^254

Poudjom DjomaniY H GriffinW L OrsquoReilly SY amp Doyle B J(2005) Lithospheric domains and controls on kimberlite emplace-ment Slave Province Canada evidence from elastic thickness andupper mantle composition Geochemistry Geophysics Geosystems 6doi1010292005GC0009978

Presnall D C Gudfinnsson G H amp Walter M J (2002)Generation of mid-ocean ridge basalts at pressures from 1 to 7GPa Geochimica et Cosmochimica Acta 66 2073^2090

Schmidberger S S Heaman L M Simonetti A Creaser R A ampCookenboo H O (2005) Formation of Paleoproterozoiceclogitic mantle Slave Province (Canada) Insights from in-situHf and U^Pb isotopic analyses of mantle zircons Earth and

Planetary Science Letters 240 621^633

Schmidberger S S Simonetti A Heaman L M Creaser R A ampWhiteford S (2007) Lu^Hf in-situ Sr and Pb isotope and traceelement systematics for mantle eclogites from the Diavik diamondmine Evidence for Paleoproterozoic subduction beneath the Slavecraton Canada Earth and Planetary Science Letters 254 55^68

Schulze D J amp Helmstaedt H (1988) Coesite-sanidine eclogitesfrom kimberlite products of mantle fractionation or subductionJournal of Geology 96 435^443

Smithies R H Champion D C amp Sun S S (2004) The case forArchaean boninites Contributions to Mineralogy and Petrology 147705^721

Smyth J R Caporuscio F A amp McCormick T C (1989) Mantleeclogites evidence of igneous fractionation in the mantle Earth andPlanetary Science Letters 93 133^141

Sobolev NV amp Lavrentrsquoyev Y G (1971) Isomorphic sodium admix-ture in garnets formed at high pressure Contributions to Mineralogy

and Petrology 31 1^12Stalder R Foley S F Brey G P amp Horn I (1998) Mineral aqueous

fluid partitioning of trace elements at 900^12008C and 30^57GPa New experimental data for garnet clinopyroxene and rutileand implications for mantle metasomatism Geochimica et

Cosmochimica Acta 62 1781^1801Sun S-S amp McDonough W F (1989) Chemical and isotopic sys-

tematics of oceanic basalts implications for mantle compositionsand processes In Saunders A D amp Norris M J (eds)Magmatism in the Ocean Basins Geological Society London Special

Publications 42 313^345Taura H Yurimoto H Kurita K amp Sueno S (1998) Pressure

dependence on partition coefficients for trace elements between oli-vine and the coexisting melts Physics and Chemistry of Minerals 25469^484

Taylor L A amp Neal C R (1989) Eclogites with oceanic crustal andmantle signatures from the Bellsbank kimberlite South AfricaPart I Mineralogy petrography and whole rock chemistryJournal of Geology 97 551^567

van Achterbergh E Ryan C G amp Griffin W L (1999) Glitter online intensity reduction for the laser ablation inductively coupledplasma spectrometry Abstract 9th V M GoldschmidtConference LPI Boston MA 905 pp Abstract no 7215

van Breemen O DavisW J amp King J E (1992) Temporal distribu-tion of granitoid plutonic rocks in the Archean Slave ProvinceNorthwest Canadian Shield Canadian Journal of Earth Sciences 292186^2199

Vervoort J D Patchett P J Blichert-Toft J amp Albare de F (1999)Relationship between Lu^Hf and Sm^Nd isotopic systems in theglobal sedimentary system Earth and Planetary Science Letters 16879^99

Viljoen K S Smith C B amp Sharp Z D (1996) Stable and radio-genic isotope study of eclogite xenoliths from the Orapa kimberliteBotswana Chemical Geology 131 235^255

Walter M J (1998) Melting of garnet peridotite and theorigin of komatiite and depleted lithosphere Journal of Petrology 3929^60

Wasserburg G J Jacobsen S B DePaolo D J McCulloch M Tamp Wen T (1981) Precise determination of SmNd ratios Sm andNd isotopic abundances in standard solutions Geochimica et

Cosmochimica Acta 45 2311^2323Westerlund K J Shirey S B Richardson S H Carlson R W

Gurney J J amp Harris JW (2006) A subduction wedge origin forPaleoarchean peridotitic diamonds and harzburgites from thePanda kimberlite Slave craton evidence from Re^Os isotope sys-tematics Contributions to Mineralogy and Petrology 152 275^294

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1873

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Page 4: Origins of Xenolithic Eclogites and Pyroxenites from the Central

Table 1 Petrographic data and eclogite classification

Sample no Ref1 Type2 GpMampG3 GpTampN

4 Size5 Microstructure6 Assemblagemode7

gt cpx rut sf ky opx other

Eclogites

SE01 3 high-Mg I A 51 na x x

SE02 3 high-Ca I B 51 na x x x

SE03 3 high-Mg II A 51 na x x

SE04 3 high-Mg II B 51 na x x

SE11 3 low-Mg II B 51 na x x x x

vr09399 2 high-Mg II A 51 na x x

vr19674ecl-g4 2 low-Mg II B 16 na x x

vr19674ecl-g5 2 high-Ca II C 51 na x x

vr19674ecl-g6 2 high-Mg I A 51 na x x

vr19674ecl-g7 1 volatile-rich II A 15 medium-tabular phlog (44) ap (51)

vr19677ecl-g1 2 high-Mg II B 51 na x x

vr19677ecl-g2 2 high-Mg II B 17 na x x x

vr19677ecl-g4 2 high-Mg II B 51 na x x

vr19677ecl-g5 2 high-Mg I B 51 na x x

vr19677ecl-g6 2 high-Mg II A 51 na x x

vr40345 3 high-Ca I B 51 na x x x dia (trace)

vr40364 2 high-Mg II A 51 na x x

vr43452 1 low-Mg I B 45 coarse-grano 48 51 1

vr43465 1 high-Ca I B 2 coarse-grano 41 55 1 graph (3)

vr43469 1 low-Mg II B 15 coarse-grano 60 40 51 51

vr43477 1 low-Mg II B 15 coarse-grano x x x

vr43479 1 low-Mg II A 25 medium-grano 50 50 51

vr43480 1 high-Mg I A 2 medium-grano 89 11 51

vr50858 2 high-Ca II B 51 na x x x

vr50860 2 high-Ca I C 51 na x x

vr50909 3 high-Ca II B 51 na x x x x

vr50916 2 high-Mg II A 51 na x x

vr67112b 1 volatile-rich II C 2 medium-grano 51 ilm (3) cc (2)

vr67360 1 high-Mg II A 3 coarse-grano 62 38 51

yk1911 1 volatile-rich II B 2 medium-tabular 31 18 phlog (49) ap (51) plag (51) cc (51)

yk1926 1 low-Mg II B 17 coarse-grano 70 30 51

yk1943 1 low-Mg I B 2 coarse-grano 57 41 2 51 qz (51)

yk1946 1 high-Mg II A 3 coarse-grano 62 37 1 51

yk1949 1 high-Mg II B 3 coarse-grano 70 30 51 51

yk3528 1 high-Ca I B 25 medium-tabular 57 43 51

Pyroxenites

SE08 3 na II A 51 na x x

SE09 3 na II A 51 na x x x

vr09353 2 na II A 51 na x x x

vr09384 2 na II B 51 na x x x

vr09405 2 na II A 51 na x x

vr09406 2 na II A 51 na x x x

vr09424 3 na II A 51 na x x x

(continued)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1846

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Some xenoliths have subhedral or rounded garnets in alsquomatrixrsquo of interstitial cpx (Group I of MacGregor ampCarter 1970 Fig 2a) whereas in others garnet and cpxhave straight grain boundaries and an interlocking fabric(Group II Fig 2b) The classification of McCandless ampGurney (1989) builds on work of MacGregor amp Carter(1970) and distinguishes group I eclogites by their higherNa2O in garnet (009wt ) and K2O in cpx (008wt) from group II eclogites The classification used inTable 1 (GpMampG) is that based on Na2O contents ingarnet A different classification (GpTampN) places eclogitesinto groups A B and C distinguished by the MgO FeOand CaO contents of garnets and Na2O and MgO contentsof clinopyroxenes (Coleman et al 1965 Taylor amp Neal

1989) These classifications are based on distinct eclogitesuites from Southern African kimberlites and do not neces-sarily reflect the particularities of eclogites from otherlocalities We therefore use a slightly different scheme(lsquotypesrsquo in Table 1) that reflects the specific groupings ofeclogites in this study with regard to Cr2O3^CaO CaO(MgOthornFeO)^MgOFeO and CaO^Na2O relationshipsin garnet (see below)Garnet modes in volatile-free eclogites range from 40

to 90 vol and cpx modes from 10 to 55 with anaverage and median garnet mode of 60 and 65respectively Two samples have high modal amounts ofphlogopite (44 and 48 respectively) in addition togarnet and cpx and are therefore not eclogites sensu stricto

Table 1 Continued

Sample no Ref1 Type2 GpMampG3 GpTampN

4 Size5 Microstructure6 Assemblagemode7

gt cpx rut sf ky opx other

vr19673ecl-g1 2 na II B 51 na x x x

vr19673ecl-g2 2 na II B 51 na x x x x

vr19674ecl-g1 2 na II A 51 na x x x

vr19674ecl-g2 2 na II A 51 na x x x

vr19674ecl-g3 2 na II A 51 na x x

vr19674per-g1 2 na II A 51 na x x

vr19674per-g4 2 na II A 51 na x x x

vr40302 2 na II A 51 na x x x

vr40336 3 na II A 51 na x x x

vr40337 2 na I A 51 na x x

vr40374 2 na II A 51 na x x x

vr40382 2 na II A 51 na x x x

vr40384 ecl-g4 2 na II A 51 na x x x

vr40399 2 na II B 51 na x x x

vr50856 2 na II A 51 na x x x

vr50887 2 na II A 51 na x x

vr50888 2 na II A 51 na x x

vr50899 2 na II A 51 na x x x

vr50900 2 na II A 51 na x x x

vr50925 2 na II B 51 na x x

yk1914 1 na II A 25 medium-grano 22 78

yk1915 1 na II A 15 coarse-grano x x x x

yk1952 1 na II A 2 medium-grano x x x x

1References 1 this study 2 Pearson et al (1999) 3 Pearson et al (unpublished data) trace-element analysesthis study2Classification based on chemical parameters (see text for details)3Classification of McCandless amp Gurney (1989) based on MacGregor amp Carter (1970)4Classification of Taylor amp Neal (1989) based on Coleman et al (1965)5Longest dimension (cm) across xenolith6Only given for samples41 cm in size grano granoblastic7Modes (vol ) are given where available otherwise presence of mineral is indicated as identified in thin section or stubgt garnet cpx clinopyroxene ky kyanite rut rutile sulf sulfide phlog phlogopite ap apatite dia diamond graphgraphite cc calcite ilm ilmenite pl plagioclase qz quartz opx orthopyroxene na not applicable or not available

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1847

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(phlogopite occurring as small secondary grains at grainboundaries and in veins is not considered in this study)Ten samples of all eclogite types show secondary spongyrims of cpx around cpx cores Secondary rims containsmall (mm-scale) patches of glassRutile is the most common primary accessory phase

(15 of 65 eclogites and pyroxenites) with modes between51 and 2 vol Sulfide occurs in 14 samples and ispresent as rounded to subhedral to irregular grains

Kyanite has been identified in three samples studied byPearson et al (1999) One of the kyanite-bearing eclogitesalso contains diamond (VR40345) Kyanite as well asdiamond or graphite associated with kyanite-bearing andcompositionally similar eclogites (high-Ca eclogites) hasbeen frequently observed in drill core (Pearson et al1999) but because these eclogite types are often alteredand friable they are not proportionally represented in thepresent study One high-Ca eclogite (VR43465) contains

Fig 2 Eclogites in thin section (plane-polarized light) illustrating the microstructures mentioned in the text (a) high-Mg eclogiteVR67360(b) high-Mg eclogiteYK1946 (c) phlogopitized eclogiteYK1911 (d) garnet pyroxeniteYK1915 (e) garnet pyroxeniteYK1914 (f) garnet pyrox-enite schlieren in a matrix of garnet spongy opx and glass (YK1952) Grt garnet cpx clinopyroxene rut rutile phlog phlogopite Scalebar in(c) applies to all sections (1cm)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1848

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

3 vol graphite which occurs as disseminated plates sev-eral millimetres long with prismatic tabular or irregularhabit Quartz has been retrieved from a mineral separateof low-Mg eclogite YK1943 The phlogopitized samples(VR19674ecl-g7 and YK1911) also contain apatite andglass as accessory phases Vesicles in sample VR67112bare partially filled with single or multiple large carbonatecrystalsMost pyroxenites from Lac de Gras have fine- to coarse-

grained granoblastic equilibrated microstructures(Pearson et al 1999) Modal information for pyroxenitesin the present study is available for only the larger sampleYK1914 (22 garnet 78 cpx) (Fig 2e)

MAJOR ELEMENTSAverage major-element compositions of garnet cpx andopx in the different eclogite types (except for the threevolatile-rich eclogites which are not averaged) and in pyr-oxenites are given in Table 2 the full datasets includingaccessory minerals are available as ElectronicAppendices 2^5 which can be downloaded from httppetrologyoxfordjournalsorg

GarnetGarnets have XMg [pyrope component 100Mg(MgthornFethornCathornMn)] ranging from 178 to 756 andXCa (grossular component mol) ranging from 72 to359 (Electronic Appendix 2) In a diagram of CaO(MgOthornFeO) vs MgOFeO (Fig 3) eclogitic garnetsfall into three groups (1) low-Mg garnets with low

CaO(MgOthornFeO) and low MgOFeO (2) high-Mggarnets with low CaO(MgOthornFeO) and high MgOFeO (3) high-Ca those with high CaO(MgOthornFeO)and intermediate MgOFeO Volatile-rich eclogites havegarnets with the lowest MgOFeO whereas pyroxenitestrend towards higher values than eclogitesMost eclogite garnets have Cr2O3 contents502wt

with variable CaO contents (Fig 4) Garnets in pyroxeniteshave low CaO contents which correlate positively withCr2O3 More than half of the pyroxenites are opx-free andall pyroxenites are olivine-free on thin-section scalePyroxenitic garnets show an opposing trend of CaOvs Na2O (Fig 5) compared with the eclogitic garnetsBecause the positive correlation of Cr2O3 and CaO forpyroxenitic garnet indicates buffering by both cpx andopx opx-bearing or opx-free pyroxenites will not befurther distinguished subsequently (except in the geother-mobarometry section where opx-bearing assemblagesallow simultaneous calculation of pressure andtemperature)Distinct rim compositions in most garnets are

characterized by lower CaO contents and higherMg-number [frac14 100Mg(MgthornFe)] Garnet compositionsin three samples are inhomogeneous with regard to CaOMgO and FeO (SE01 VR43479 and VR43480) withoutclear core^rim zonations similar to coexisting cpx Likerims CaO-poor compositions are characterized by higherMg-number Visibly secondary spongy cpx rims also havehigher MgO contents (see below) and MgO-rich garnetsare therefore regarded as affected by secondary changesBased on Cr^Ca Mg^Ca^Fe and Ca^Na composi-

tional relationships in garnet the following eclogitetypes are distinguished (1) eclogites with high-Ca

0

05

10

15

20

25

30

0 01 02 03 04 05 06 07 08 09

high-Ca

high-Mg+pyroxenites

low-Mg

CaO(MgO+FeO)

MgO

FeO

GARNETpyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropic Jericho massiveDiavik low-temperatureDiavik high-temperatureDiavik diamond-bearing

eclogites

Fig 3 CaO(MgOthornFeO) vs MgOFeO (wt ) in garnet Fieldsdistinguish low-Mg and high-Ca eclogites as well as high-Mg eclogitesplus pyroxenites in this study Garnets in eclogites from Jericho(Kopylova et al 1999) and Diavik (Schmidberger et al 2007) areshown for comparison Garnet inclusions in diamond from Davieset al (1999 2004)

-135 wt (high-Ca)

2

3

4

5

6

7

8

9

0 05 10 15 20 25Cr2O3

CaO

GARNET

opx-bearing

opx-free

pyroxenites

low-Mghigh-Mgvolatile-rich

Jericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

VR19673ecl-g1g2

Fig 4 Cr2O3 vs CaO (wt ) in garnet Garnets in eclogites fromJericho and Diavik as in Fig 3

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1849

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 2 Summary of major-element contents in garnet and cpx (wt )

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO Na2O Total Mg-no T(Krogh) Type

Garnet

Low-Mg ecl 3964 026 2214 011 1843 037 1050 846 009 1000 503 976

minimum 3893 008 2130 5009 1071 024 849 567 006 990 423 860

maximum 4094 036 2275 011 2093 046 1457 1062 012 1004 708 1088

1s 070 012 052 na 354 007 227 185 003 05 99 98

High-Mg ecl 4081 026 2261 013 1419 030 1537 624 007 1000 659 969

minimum 3970 008 2224 5009 951 023 1127 418 003 991 551 820

maximum 4250 052 2323 035 1795 036 2030 882 013 1012 792 1241

1s 077 013 031 na 225 004 189 141 003 05 57 88

High-Ca ecl 3979 034 2211 na 1516 027 958 1240 011 998 528 1078

minimum 3909 011 2117 5009 1164 023 801 1144 007 989 456 910

maximum 4039 073 2242 011 1713 035 1249 1351 018 1004 657 1246

1s 042 026 042 na 173 004 155 073 004 06 67 151

vr19674ecl-g7 3867 003 2187 5009 2380 153 793 627 5004 1002 373 volatile-rich

vr67112b 3783 013 2097 009 2620 204 466 790 5004 999 241 volatile-rich

yk1911 3874 5006 2200 5009 2120 221 802 747 5004 997 403 volatile-rich

Pyroxenites 4163 021 2257 075 1148 034 1906 407 006 1002 747 909

minimum 3989 005 2081 009 729 025 1535 277 004 996 593 699

maximum 4279 048 2330 260 1881 054 2215 566 009 1011 842 1158

1s 068 010 057 057 323 007 209 065 001 04 73 117

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO Na2O Total Mg-no Type

Cpx

Low-Mg ecl 5453 028 739 na 500 5009 1117 1662 434 994 801

minimum 5160 017 564 5009 337 na 876 1388 197 981 741

maximum 5570 044 1067 014 687 na 1324 1926 632 1001 865

1s 128 011 155 na 124 na 153 177 126 06 40

High-Mg ecl 5483 023 483 na 381 na 1368 1875 318 994 865

minimum 5353 009 263 5009 245 5009 1156 1617 204 986 778

maximum 5645 031 810 016 674 011 1588 2101 495 1011 920

1s 071 007 137 na 095 na 104 138 075 06 30

High-Ca ecl 5563 027 1184 na 318 5009 868 1368 615 995 836

minimum 5448 014 769 5009 168 na 619 1060 391 987 751

maximum 5670 049 1579 010 603 na 1148 1689 807 1003 878

1s 076 015 334 na 151 na 189 261 167 06 45

vr19674 ecl-g7 5322 010 185 5009 536 015 1448 2339 062 992 828 volatile-rich

vr67112b 5040 041 328 5009 1079 024 1108 2229 071 992 647 volatile-rich

yk1911 5231 014 301 5009 544 024 1450 2340 056 996 826 volatile-rich

Pyroxenites 5527 na 347 038 304 na 1543 1959 245 999 899

minimum 5299 5006 037 010 149 5009 1128 1408 030 990 804

maximum 5666 037 786 134 490 014 1863 2312 537 1009 947

1s 068 na 189 028 091 na 193 226 131 05 37

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO NiO Total Mg-no

Opx

Pyroxenites 5781 na 059 na 627 na 3508 037 011 1006 909

minimum 5546 5006 032 5009 462 5009 2999 020 008 996 804

maximum 5888 009 318 054 1304 014 3628 077 014 1014 933

1s 098 na 067 na 212 na 160 019 002 06 32

Averaged for the three eclogite types and for pyroxenites minimum and maximum values as well as standard deviationsalso given Concentrations in volatile-rich eclogites are shown individually ecl eclogite Mg-numberfrac14 100Mg(Mgthorn Fe)Temperatures of last equilibration [T(Krogh)] as in Electronic Appendix 8

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1850

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

garnets (these include kyanite- graphite- and diamond-bearing varieties) (2) eclogites with low-Mg garnets (theseinclude quartz-bearing eclogite) (3) eclogites with high-Mggarnets and Cr2O3 contents502wt (4) volatile-richeclogites (phlogopite- apatite- carbonate-bearing)

Clinopyroxene (cpx)Clinopyroxenes in all eclogite groups have higher averageAl2O3 contents than those in pyroxenites with the formerhaving about 90 of the total Al in the jadeite molecule(AlVI) versus about 35 for the latter corresponding toan average of 41 and 25wt Na2O respectivelyThe highest Al2O3 and Na2O contents in the dataset areobserved in cpx from the high-Ca eclogites (up to158wt Al2O3) (Electronic Appendix 3) Contents ofK2O are generally below the detection limit(5004wt ) but values up to 024wt are observed insome cpx in high-Ca eclogites Average cpx Mg-numberare highest in pyroxenites (899) followed by cpx inhigh-Mg eclogites (865) high-Ca eclogites (836) andlow-Mg eclogites (801) Pyroxenitic cpx tend to havelower TiO2 and higher NiO and distinctly higherCr2O3 contents compared with those in eclogites A plotof MgO vs Al2O3 shows a negative correlation (Fig 6)Clinopyroxenes in low-Mg eclogites have lower Al2O3

contents at a given MgO content than other eclogitetypes Between 15 and 17wt MgO the Al2O3^MgOcorrelation for cpx in pyroxenites has a different slopefrom that in the eclogites Clinopyroxenes in pyroxenites

VR19673ecl-g1 and VR19673ecl-g2 have distinctly lowerMgO contents and higher Al2O3 contents than inthe other pyroxenites They coexist with garnets havingthe lowest Cr2O3 and CaO contents of the dataset whichdefine one end of the positive Cr2O3^CaO correlationobserved for garnetsClinopyroxenes in volatile-rich eclogites have distinctly

high MnO contents (up to 024wt ) Mg-numbers forcpx in the phlogopite-bearing variety are 826 and 828whereas that in the calcite-bearing eclogite (VR67112b)has an Mg-number of 647Some cpx grains have spongy rims that contain small

(mm-scale) patches of glass these rims are richer in Caand Mg and markedly poorer in Na than the cores similarto cpx in eclogites from southern Africa (Taylor amp Neal1989) The rim texture is unequilibrated suggesting lategrowth The formation of the spongy cpx rims might berelated to incongruent melting of cpx or be a result ofmetasomatism (Taylor amp Neal 1989 and referencestherein) Distinct although not spongy rims in two eclo-gites (VR67360 and YK1946) are similarly enriched inCaO and depleted in Na2O Two types of cpx with respectto CaO Na2O and Mg are present in VR43477VR43479 and VR43480 without clear core^rim zonationor spongy rims No significant differences are recognizedwith respect to other oxides

Orthopyroxene (opx)Orthopyroxenes have Mg-number ranging from 804 to933 (17 of 19 samples have values between 895 and 933)and typically contain minor amounts of Al2O3

(032^118wt one outlier has 318wt ) CaO(020^118 wt ) and NiO (008^014wt ) (Electronic

0

2

4

6

8

10

12

14

7 9 11 13 15 17 19

Al 2

O3

CPX

pyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

cpx in pyroxenitesVR19673ecl-g1g2

MgO

Fig 6 MgO vs Al2O3 (wt ) in cpx Data sources as in Figs 3and 4

0

005

010

015

020

2 4 6 8 10 12 14 16

CaO

Na 2

O

dl

group Igroup II

pyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

GARNET

Fig 5 CaO vs Na2O (wt ) in garnet Arrows show trends foreclogites and pyroxenites respectively [note that part of thepyroxenite trend extends into the field falling below the detectionlimit (dl) for Na2O] Group I and group II refer to classificationof McCandless amp Gurney (1989) Data sources as in Figs 3 and 4

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1851

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Appendix 4) The contents of all other elements are mostlybelow their respective detection limits

Accessory phasesRutile and ilmenite occur in some of the eclogites (Table 1)All rutile grains have exsolved 10 ilmenite as lamellaeand rims Ilmenite-free areas of rutile in eclogites containbetween 962 and 988wt TiO2 and minor amounts ofAl2O3 (01^04wt ) Cr2O3 (5009 to 02wt ) andFeO (03^07wt ) Rutile in pyroxenite YK1915 contains004wt Al2O3 01wt FeO and 15wt Cr2O3Discrete ilmenite grains in sample VR67112b haveTiO2 contents of 499 and 500wt MgO of 23 and15wt and FeO of 399 and 403wt respectivelySulfides occur interstitially or enclosed in cpx garnet

and opx they are mostly pyrrhotite and monosulfide solidsolution with variable Ni contents (unpublished data)They will not be considered further hereAbundant platy phlogopite is observed in two samples

(VR19674ecl-g7 and YK1911) It is Cl-rich in sampleVR19674ecl-g7 (063wt vs 003wt in YK1911) andF-rich in sample YK1911 (097wt vs 005wt inVR19674ecl-g7) (Electronic Appendix 5) The same is truefor coexisting apatite (VR19674ecl-g7 254wt Cl and014wt F YK1911 006wt Cl and 306wt F)

Comparison with mineral inclusions indiamonds from Lac de Gras and witheclogite xenoliths from other Slave CratonlocalitiesMineral inclusions in diamonds from the Lac de Graskimberlites have been investigated by Davies et al (19992004) For garnets there is a striking similarity betweenthe inclusions in diamond and those in the high-Ca eclogitexenoliths although the inclusions in diamond trendtowards higher Na2O (Fig 5) and lower Al2O3 contentsThe overlap is not as marked for cpx where inclusionsin diamond span almost the entire range observed foreclogitic cpx (Fig 6)A comparison with eclogites from other localities in

the Slave Craton shows that garnets in most Diavik low-temperature eclogites reported by Schmidberger et al(2007) plot with garnets in high-Mg eclogites and inpyroxenites from this study (Figs 3^5) whereas garnets inhigh-temperature and diamond-bearing eclogites(Schmidberger et al 2007) overlap most with garnets inhigh-Ca eclogites from this study and with garnetsincluded in diamond Eclogites from the Jericho kimberlitehave either massive or anisotropic fabrics (Kopylova et al1999) Garnets in almost all massive eclogites plot withgarnets in low-Mg eclogites from this study with regard toCaO^MgO^FeO relationships (Fig 3) but many havelower Na2O contents at a given CaO content (Fig 5)Garnets in anisotropic eclogites are restricted to lowMgOFeO (507) and show some overlap with garnets

in high-Ca eclogites although trending towards lowerCaO(MgOthornFeO) (Fig 3) while coexisting cpx is alsorestricted to low MgO and trends toward high Al2O3 con-tents (Fig 6)

TRACE ELEMENTSGarnetTrace-element abundances in garnet are summarized inTable 3 and the complete dataset is reported in ElectronicAppendix 1 Garnets in low-Mg eclogites show steep posi-tive slopes between LaN and SmN and shallower positiveslopes in the normalized heavy rare earth elements(HREEN Fig 7a) The normalized light REE (LREEN)show an order of magnitude variability ZrHf is mostlysupra-chondritic and Ti can be enriched or depleted rela-tive to elements of similar compatibility Niobium in threesamples is strongly depleted at 002^003 chondriticGarnets in high-Mg eclogites also show smooth positiveslopes in LREEN but have flat HREEN and variablenormalized middle REE (MREEN) to HREEN (Fig 7b)ZrHf is always supra-chondritic and Ti enriched ordepleted relative to similarly compatible elementsNiobium is most samples is below detection Small positiveEu anomalies though not outside the analytical uncer-tainty for Eu relative to either Sm or Gd are observed forgarnet in some high-Mg eclogites (eg EuEu up to 15for VR67360 where Eu is the average of the chondrite-normalized Sm and Gd concentrations) Garnets in high-Ca eclogites are slightly enriched in the MREEN relative tothe HREEN or have flat MREEN and HREEN with ahump between EuN and HoN (Fig 7c) One sample(VR40345) shows strong La enrichment (greater thanchondrite) Two samples have supra-chondritic ZrHfwhereas three samples have close to chondritic ratios andone sample has a subchondritic ratio As is true for theother eclogite typesTiN is either enriched or depleted rela-tive to similarly compatible elementsPyroxenitic garnets have variable slopes in the LREEN

and flat MREEN to HREEN patterns and two samplesshow high Sc relative to neighbouring elements Garnet inone of the pyroxenites (YK1952) has a highly distinctivepattern relative to other pyroxenites with very low LREEand HFSE (Fig 7d) This sample consists of coarse pyroxe-nite schlieren in which the garnet was analysed in a finer-grained lsquomatrixrsquo of spongy cpx with secondary rims finecpx^opx intergrowths embayed garnet and melt patchesVolatile-rich eclogites have garnets that are slightly

enriched in MREEN relative to HREEN or have flatMREE to HREE patterns One garnet has a pronouncednegative Eu anomaly (EuEufrac14 07 forVR67112b Fig7e)Garnets in volatile-rich samples have the lowest Ni Coand Ti and the highest V ScYand HREE abundances inthe dataset

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1852

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 3 Summary of trace-element contents in garnet and cpx (ppm)

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR19674

ecl-g7

VR67112b YK1911 av min max 1s

n 6 6 7 3 5 4 4

P 169 54 270 88 196 70 310 78 282 80 450 138 160 60 80 214 48 370 144

Sc 62 40 89 16 48 36 67 11 53 42 79 15 69 92 93 186 52 440 179

Ti 1667 490 2500 680 1073 620 2000 455 2141 700 4700 1582 270 690 260 1097 540 1900 573

V 169 78 250 68 150 78 280 72 178 110 390 101 140 340 100 390 71 820 326

Co 59 42 70 11 58 39 84 13 58 47 65 5 30 41 27 68 55 90 16

Ni 17 5 32 10 21 11 34 10 24 11 35 8 6 10 51 54 25 110 38

Ga 12 7 15 3 13 8 16 3 12 9 20 4 8 86 58 12 10 14 2

Sr na 512 4 na na 522 522 na na 57 57 na 52 5025 510 513 513 513 na

Y 37 6 61 18 21 7 38 10 31 20 38 5 60 81 70 19 12 26 6

Zr 15 5 29 9 28 6 83 26 24 9 43 12 60 22 90 18 1 41 17

Nb 009 0005 048 019 na 50015 05 na na 5005 05 na 5003 0012 5001 022 0030 057 025

La 005 001 009 004 na 50007 00 na na 5003 04 na 004 002 0009 003 000 009 005

Ce 03 01 05 02 01 00 01 00 05 01 18 05 03 022 015 01 00 03 01

Pr 01 00 02 01 00 00 01 00 02 01 03 00 017 014 012 00 00 01 00

Nd 12 03 24 07 07 04 14 03 21 12 32 07 24 23 20 03 00 07 03

Sm 11 04 22 06 10 03 18 05 23 08 36 10 29 31 33 05 01 09 04

Eu 06 03 10 03 06 02 10 03 11 04 17 04 16 114 16 03 01 06 02

Gd 27 07 41 12 23 07 42 12 43 16 53 13 8 81 90 13 06 24 08

Dy 54 11 70 23 36 13 65 18 58 35 71 11 12 144 12 29 18 42 10

Ho 14 02 23 07 08 03 15 04 12 09 14 02 24 30 23 07 04 10 02

Er 46 07 80 24 24 09 40 10 34 25 40 05 8 85 6 22 14 28 06

Yb 52 06 98 30 24 10 37 08 32 22 39 05 8 70 5 23 13 30 07

Lu 09 01 16 05 04 02 05 01 05 03 07 01 11 09 08 04 02 05 01

Hf 03 02 05 01 04 01 10 03 05 02 09 02 09 027 007 03 00 06 03

Ta na 50008 5004 na na 50005 001 na na 50004 na na 003 50002 50005 na 50007 011 na

Pb na 5005 02 na na 5002 02 na na 5003 07 na 02 004 5002 na 5003 01 na

Th na 5001 005 na na 50007 003 na na 50008 011 na 5007 0012 50004 002 0005 006 002

U na 5003 5003 na na 5001 01 na na 5007 na na 01 0023 010 na 5001 003 na

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR67112b YK1911 av min max 1s

n 6 4 8 5 4 4

P 65 15 120 41 39 14 56 16 53 30 70 12 20 9 80 27 120 40

Sc 22 10 30 7 15 14 19 2 14 7 24 5 80 43 17 11 21 5

Ti 1557 990 2400 508 1168 1010 1490 191 1988 810 3500 1154 2390 1700 689 102 1685 694

V 415 180 480 107 318 210 400 69 324 259 450 72 600 180 250 139 480 156

Co 35 26 50 8 20 14 31 7 26 14 42 11 49 23 29 20 42 10

Ni 164 87 220 47 228 100 430 128 188 135 320 56 59 34 428 410 670 184

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1853

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

ClinopyroxeneTrace-element abundances in cpx are summarized inTable 3 and the complete dataset is reported in ElectronicAppendix 6 Clinopyroxenes in low-Mg eclogites havemildly positive to negative slopes from LaN to NdN andsmooth negative slopes between SmN and HoN thatmostly continue to LuN (Fig 8a) Niobium is below detec-tion or 5007 chondritic and SrN is variably enrichedrelative to neighbouring elements with the peak size corre-lating negatively with LREEN ZrHf is subchondritic andHfN is enriched relative to SmN whereasTiN is enriched ordepleted relative to similarly compatible elementsWe ana-lysed the spongy rim in cpxYK1943 and its trace-elementpattern is similar to that of cpx cores for more compatibleelements whereas the most incompatible elements areseveral orders of magnitude higher Clinopyroxenes in

high-Mg eclogites have steeper negative slopes betweenNdN and LuN than those in low-Mg eclogites combinedwith lower HREE abundances HfN is depleted relative toSmN (Fig 8b) High-Ca eclogites contain cpx with sinusoi-dal REEN patterns with lower LREE than other eclogitetypes and the heaviest REE below detection They haveuniformly strong positive Sr peaks relative to elements ofsimilar compatibility (Fig 9) The diamond-bearing high-Ca eclogite (VR40345) has cpx with unusually high Nband LREE abundancesClinopyroxenes in four pyroxenites have disparate trace

element patterns They all share depletions in TiN relativeto similarly compatible elements and two show positiveSrN peaks Two samples have cpx with high and flatLREEN to MREEN patterns and steep negative slopes inthe HREEN The trace-element patterns of cpx in YK1914

Table 3 Continued

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR67112b YK1911 av min max 1s

n 6 4 8 5 4 4

Ga 16 10 20 3 12 9 15 2 26 15 39 9 129 123 6 11 11 5

Rb na 5008 04 na na 5007 05 02 na 5003 03 na 02 5004 03 010 05 0

Sr 166 85 290 74 222 153 285 50 243 83 440 142 85 120 200 11 414 209

Y 2 04 4 10 13 07 2 04 11 03 3 10 28 51 2 03 3 2

Zr 19 4 48 15 25 8 45 13 17 8 28 7 111 129 36 008 81 42

Nb 01 0003 03 01 na 5002 00 na na 5002 16 9 007 011 010 003 02 008

Cs na 5001 6 na na 5006 5006 na5013 5004 5013 na bdl 5002 na 5001 5002 na

Ba 09 02 2 07 09 01 2 07 10 03 3 10 06 01 13 06 3 10

La 3 03 8 3 3 23 3 04 4 01 14 6 51 97 4 01 10 4

Ce 7 13 20 6 10 7 13 2 7 05 24 10 25 28 11 01 28 13

Pr 11 02 3 08 2 10 3 06 08 02 2 09 47 46 2 002 4 2

Nd 5 12 12 4 10 5 14 4 4 09 8 26 23 22 8 01 22 10

Sm 11 05 21 06 20 10 31 10 07 02 11 03 47 53 2 005 5 2

Eu 03 01 05 01 06 028 09 03 02 010 03 007 087 123 06 002 15 07

Gd 08 03 13 03 11 05 19 06 05 03 06 011 28 37 15 010 36 2

Dy 05 01 07 02 04 022 06 01 02 013 03 007 11 16 07 008 14 06

Ho 01 00 02 004 006 003 007 002 na 5003 012 na 011 021 009 001 02 007

Er na 507 507 na 011 004 02 00 na 5006 03 na 019 033 02 002 03 012

Yb na 5001 02 na na 5001 011 na na 5006 03 na 007 015 009 001 02 006

Lu na 5001 003 na na50002 002 na na50008 01 na 0006 0013 0013 0002 003 001

Hf 12 03 24 07 11 04 18 05 10 06 16 04 34 53 na50015 35 na

Ta na50003 50015 na na50006 006 na na50004 5004 na 0014 00236 na50003 503 na

Pb 06 02 12 04 06 04 09 02 na 5016 07 02 065 098 08 01 18 1

Th na50005 06 na 012 005 02 005 na 5002 5006 na 021 02 001 06 03

U na 5005 5005 na 002 0009 005 001 na 5003 5005 na 0013 024 na50005 003 na

Averaged for the three eclogite types and for pyroxenites as in Table 2 pyx pyroxenite n number of analyses averagedbdl below detection limit

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1854

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

and YK1952 are similar although at different abundanceswith only modest LREEN over HREEN enrichment andstrong depletions in ZrN relative to neighbouring elements(Fig 8d)Clinopyroxenes in the volatile-rich samples have flat pat-

terns and high abundances from LaN to SmN and smoothnegative slopes from SmN to LuN Niobium Sr and Ti are

noticeably depleted relative to elements of similar compat-ibility (Fig 8e)

Trace-element partitioningTo assess whether coexisting phases are in trace-elementequilibrium we calculated apparent cpx^garnet distribu-tion coefficients and compared them with those

(a)

(d)

(b)

(e)

(c)

GarnetCI-chondrite GarnetCI-chondrite

low-Mghigh-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1952

1

10

Nd Sm Eu Gd Dy

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Fig 7 Extended trace-element patterns of garnet in (a) low-Mg eclogites (b) high-Mg eclogites with inset showing enlargement of MREE tohighlight Eu anomalies and 1s uncertainties (c) high-Ca eclogites (d) pyroxenites and (e) volatile-bearing eclogites Normalized to chondritevalues of McDonough amp Sun (1995)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1855

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

determined experimentally or in demonstrably equili-brated natural samples Low-Mg and high-Mg eclogiteshave Dcpxgarnet for most elements that are within therange determined for natural equilibrated eclogites andexperimental assemblages (Harte amp Kirkley 1997 Greenet al 2000 Barth et al 2002a) (Electronic Appendix 7)High-Ca eclogites plot at the low end of the range

of distribution coefficients and have very low DMREE

and DHREE outside the range indicated by the datasetof Harte amp Kirkley (1997) who noted that DREE decreaseswith increasing Ca content in eclogitic minerals Theseassemblages are therefore considered to be in equilibriumPyroxenite VR40384 has D values similar to eclogites

but at the high end of the range determined by Harte amp

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

(a)

(d)

(b)

(e)

(c)

CpxCI-chondrite CpxCI-chondrite

low-Mg high-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1914

YK1952

spongy rim

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

VR40345

Fig 8 Extended trace-element patterns of cpx as in Fig 7

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1856

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Kirkley (1997) (Electronic Appendix 7) Another group(YK1914 and YK1952) plots at the low end but DHf DZrDTi and DV are far below the values indicated by experi-mental datasets Because pyroxenites compositionallygrade into eclogites their drastically different DHFSE

values suggest disequilibrium In contrast volatile-bearingsamples have DHFSE values that appear to be too high to bein equilibrium however this may be a consequence oftheir unusual bulk chemical composition (Ti- and K-rich)which differs substantially from that of lsquonormalrsquo eclogitesand experimental charges

GEOTHERMOBAROMETRYFollowing Schmidberger et al (2007) temperatures oflast equilibration were calculated for eclogitic andopx-free pyroxenitic assemblages using the formulationof Krogh (1988) The results are summarized in Table 2and Electronic Appendix 8 for a preset pressure of5 GPa and illustrated as a histogram in Fig 9 Foropx-bearing pyroxenites temperatures were calculatedsimultaneously with pressure [Ca-in-opx thermometryand Al-in-opx barometry using the formulations of Breyet al (1990)] to constrain their relative positions in themantle columnMinerals in some eclogites from Lac de Gras show dis-

tinct core and rim compositions or inhomogeneitiesTemperatures were not calculated for samples withspatially irresolvable inhomogeneous garnet compositions(SE01VR43479VR43480) For eclogites with spongy cpxcore compositions were used in the calculationTemperatures were calculated for both core and rim com-positions where applicable For all samples these

temperatures agree within the uncertainty of this thermo-meter (508C Brey amp Kolaquo hler 1990)Most equilibration temperatures for eclogites lie between

800 and10008C (Fig 9)The transition between the shallowand the deep lithospheric mantle layers in the central Lacde Gras area lies in this temperature interval (Griffin et al1999 Pearson et al 1999) Eclogites in this study lack thedistinct bimodal temperature distribution observed bySchmidberger et al (2007) but as in their study high-Caeclogites give higher average temperatures (10808C corre-sponding to a depth of 155 km if equilibrated along a 40mWm2 geotherm) than other eclogites types (9808Cdepth 135 km) the diamond-bearing sample (VR40345)gives thehighest temperature of the datasetHighNa2Ocon-tents in garnet from high-Ca eclogites are qualitativelyconsistent with higher equilibration pressures (Sobolev ampLavrentrsquoyev 1971) as are high K2O contents in coexistingcpx (Erlank ampKushiro1970) suggesting that high-Ca eclo-gites not only equilibrated at high temperature but also highpressure rather than representing a temperature excursionPyroxenites both opx-bearing and opx-free give somewhatlower average temperatures than eclogites (mean 9258C)and phlogopite-rich eclogites give the lowest temperatures(685 and 7698C respectively) The calcite-bearing eclogiteVR67112b gives a relatively low equilibration temperatureof 8558C

MEASURED ANDRECONSTRUCTEDWHOLE -ROCKSRationale and approachKimberlite-borne xenoliths sometimes show evidence forinfiltration of the host kimberlite (eg Taylor amp Neal1989 Barth et al 2001 2002b Heaman et al 2006) Wetherefore follow the standard approach of calculatingwhole-rock compositions from the compositions of the con-stituent phases weighted by their modal abundances Onlycore compositions were used as distinct rim compositionsmay point to late diffusion of elements from kimberlite-related melt veinsMany eclogitic whole-rock compositions reconstructed

with only garnet and cpx have negative Ti anomalies(not shown) For most of these samples rutile or ilmenitehas been identified as part of the mineral assemblage andthese phases should therefore be considered in the whole-rock reconstruction Rutile or ilmenite modes were esti-mated by assuming that the bulk-rock has no Ti anomalywhen normalized to a normal mid-ocean ridge basalt(N-MORB) composition and calculating the amount ofrutile required to erase theTi anomalyIt is of course possible that the eclogites had negative

or positive Ti anomalies relative to elements of similarcompatibility (Eu and Dy) Because we reconstructed

0

2

4

6

8

10

12volatile-rich

pyroxenite opx-bearing

pyroxenite opx-free

high-Ca

high-Mg

low-Mg

700750

800850

900950

10001050

11001150

12001250

Fre

quen

cy

Temperature

Fig 9 Histogram ofTKrogh (Krogh 1988) obtained at a preset pres-sure of 5 GPa except opx-bearing pyroxenites where TBKN wassolved simultaneously with PBKN (Brey amp Kolaquo hler 1990)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1857

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

whole-rock compositions from mineral core compositionskimberlite infiltration is unlikely to have caused fractiona-tion of Ti from Eu and Dy In a study in which rutilemodes were estimated from the differences between Ti inmeasured bulk-rocks and in bulk-rocks reconstructed withonly garnet and cpx only three of 19 eclogites recon-structed with rutile show Ti anomalies (Barth et al2002a) Even eclogite xenoliths with depleted Zr and Hfabundances that resemble those of modern island arcbasalts do not show Ti anomalies (Jacob amp Foley 1999)We therefore consider our approach to be a reasonablyrobust means of obtaining meaningful rutile modes wheresmall sample sizes preclude reliable determination of themodes of accessory phases by point-countingWhere rutilewas not recognized on thin-section scale or not analysedthe average eclogitic rutile composition was used for thereconstruction

UncertaintiesReconstructing whole-rock compositions in this way is arather crude method given the coarse grain size of theeclogites relative to the sample size and the non-uniformdistribution of the principal minerals on thin-sectionscale The uncertainty in mineral modes of medium- tocoarse-grained eclogites with grains45mm has been esti-mated by Jerde et al (1993) to be 10 plus uncertaintiesrelated to partial alteration which is minor in our samplesNevertheless the whole-rock trace-element budget israther insensitive to the exact modes for common eclogiticassemblages and REE patterns which encompass mostother elements in terms of compatibility in garnet andcpx do not change significantly even for modal variationsof 30 (Jerde et al 1993) This is not true for major-element concentrations which vary substantially with var-iations in mineral modes and we therefore do not calculatewhole-rock major-element contentsBecause the discussion that follows is based on trace-

element patterns and trends rather than absolute abun-dances whole-rock reconstruction based on mineralmodes is adequate We are confident that the uncertaintyinherent in this approach is not leading to significant mis-representation of eclogite whole-rock compositions becausethe trace-element patterns of the reconstructed whole-rocks show distinct differences (see below) that correlatewith the assignment to eclogite types based on garnetmajor-element composition and this is independent ofmineral mode Following Jerde et al (1993) in ElectronicAppendix 9 we show variations in trace-element patternfor one sample in each eclogite type for variations ingarnet and cpx abundance from 7030 to 3070 vol This shows that the trace-element patterns are generallyinsensitive to cpxgt ratios with regard to slopes andanomalies such as positive Sr Pb and Eu anomalies withthe exception of the positive Eu anomaly in high-Mgeclogite VR43479 which disappears if a high cpx mode is

assumed (not shown) Variations in rutile modes by 50(Electronic Appendix 9d) lead as would be expected tothe appearance of Ti anomalies but a rutile mode of 05instead of 1 vol does not erase the positive Nb anomalyin low-Mg eclogiteVR50909Because of the robustness of the relative trace-element

abundances in the reconstructed eclogites a garnet propor-tion of 065 and cpx proportion of 035 approximately themedian modal abundances of these minerals in volatile-free eclogites from Lac de Gras are assumed for eclogitesamples that are too small for determination of meaningfulmodes by point-counting

Within- and between-group variationsThe trace-element compositions of reconstructed whole-rocks are given in Table 4 N-MORB-normalized trace-element patterns of reconstructed whole-rocks areshown in Fig 10 The three eclogite groups have distincttrace element patterns that are particularly obvious forthe REE (Fig 11) Low-Mg eclogites have positive Pb andSr anomalies and smooth positive slopes in their MREEN

and HREEN LREEN are enriched relative to MREEN inthree samples (Fig 11a) Three samples have high Nbrelative to similarly compatible elements and comparedwith most other eclogites (Fig 10a) All high-Mg eclogitesdisplay more or less pronounced positive Eu and Sranomalies and most have positive Pb and negative Zrand P anomalies Their N-MORB-normalized REE pat-terns show shallow positive slopes from LaN to PrN orNdNand aweak negative slope from NdN to LuN (Fig11b)VR43479 is a low-Mg eclogite with regard to its garnet

Ca^Fe^Mg composition but has a trace-element patterntypical of high-Mg eclogites and will be discussed withthe latter Conversely sample YK1949 a high-Mg eclogitein terms of its garnet Ca^Fe^Mg compostion is more simi-lar to low-Mg eclogites in terms of its whole-rock trace-element pattern and is plotted in Figs 10a and 11aWhereas the patterns of most samples within the groupresemble each other the N-MORB-normalized incompati-ble-element abundances of the high-Ca eclogites vary bymore than an order of magnitude (Fig 10c) All sampleshave positive Pb anomalies and most have positive Sranomalies Some samples have negative P andor Zranomalies Niobium abundances in three of the samplesare very high similar to those for some low-Mg eclogitesHigh-Ca eclogites tend to have flatter MREEN andHREEN patterns and more uniform HREE abundancesthat the other eclogite types with humps between SmN

and HoNThe phlogopitized sample YK1911 shows a strong posi-

tive Pb anomaly and lower HREE abundances than thecalcite-bearing sample VR67112b which shows a distinctnegative Eu anomaly and a depletion of La relative to Ce(Figs 10d and 11d)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1858

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Trace-element compositions of reconstructed whole-rocks (ppm)

Sample P Sc Ti V Co Ni Ga Sr Y Zr Nb La Ce Pr Nd

Low-Mg eclogites

vr19674-g4 186 50 4649 317 57 59 144 77 24 30 36 13 31 05 24

vr43452 154 39 2253 328 49 84 176 52 32 11 01 02 07 01 08

vr43469 38 65 4377 349 54 91 142 34 28 15 02 05 15 03 17

yk1926 144 45 3406 258 52 87 145 89 18 18 20 09 29 06 33

yk1943 173 46 5732 295 41 39 146 122 25 48 82 34 85 13 64

yk19491 102 34 1693 158 31 49 100 46 10 14 03 09 21 03 17

High-Mg eclogites

vr19677-g2 138 39 5160 291 41 53 119 98 14 22 04 13 53 11 64

vr434792 46 25 1265 129 36 126 84 56 3 7 04 06 25 04 23

vr67360 49 36 1573 138 64 184 111 76 5 9 04 09 38 07 33

yk1946 153 28 4877 192 45 105 124 108 12 48 03 10 38 08 50

High-Ca eclogites

vr19674-g5 267 30 7663 200 41 67 195 32 23 43 77 01 04 02 16

vr40345 0 33 2821 175 47 98 210 179 13 13 66 60 109 11 38

vr43465 142 27 5104 220 31 83 121 197 14 35 29 05 22 05 40

vr50858 295 30 6898 201 44 94 150 48 22 38 06 01 03 02 20

vr50860 68 57 5672 418 57 71 186 109 19 11 02 02 13 03 26

vr50909 238 28 7295 201 42 92 171 48 20 57 13 01 04 02 21

yk3528 91 51 3712 341 53 155 177 49 16 9 01 02 08 02 11

Volatile-rich eclogites

vr67112b 40 84 9242 521 46 30 107 4 40 66 31 26 126 24 125

yk1911 31 39 9428 286 44 82 164 3 23 52 64 18 52 09 46

Pyroxenites

yk1914 111 49 707 207 48 547 44 31 4 3 02 06 19 03 10

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

Low-Mg eclogites

vr19674-g4 11 05 10 22 39 10 29 33 06 12 05 028 24 008

vr43452 04 02 08 13 38 12 41 49 08 05 01 nd 02 na

vr43469 11 05 09 25 44 11 34 36 06 07 02 001 09 129

yk1926 12 05 11 19 30 07 22 24 04 07 03 004 23 090

yk1943 21 08 10 29 42 10 30 32 05 16 05 013 65 017

yk19491 06 03 11 08 14 04 14 18 03 04 02 003 29 115

High-Mg eclogites

vr19677-g2 20 08 12 21 24 05 15 15 03 05 05 007 52 6811

vr434792 05 02 12 05 06 01 04 03 00 03 01 001 98 037

vr67360 06 03 14 06 09 02 06 06 01 02 02 002 86 000

yk1946 18 07 12 21 22 05 13 12 02 11 03 006 56 112

High-Ca eclogites

vr19674-g5 17 09 11 33 43 09 24 23 04 16 05 007 02 011

vr40345 09 04 10 12 22 06 16 14 02 05 nd nd 298 na

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1859

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Continued

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

vr43465 22 09 12 25 26 06 15 14 02 10 04 001 25 040

vr50858 20 08 09 33 41 08 22 21 03 10 01 nd 02 na

vr50860 14 07 11 27 37 08 22 18 03 06 00 nd 06 na

vr50909 20 10 12 33 38 07 20 20 03 16 01 002 02 181

yk3528 08 05 11 20 30 07 18 18 02 05 01 nd 08 na

Volatile-rich eclogites

vr67112b 38 10 07 53 74 15 41 34 04 18 03 012 60 004

yk1911 20 07 08 35 39 07 18 17 02 10 14 nd 73 na

Pyroxenites

yk1914 02 01 11 04 08 02 05 06 01 01 03 003 68 101

aGarnet has CandashMgndashFe relationship typical of high-Mg eclogites but whole-rock trace-element pattern is typical oflow-Mg eclogitesbGarnet has CandashMgndashFe relationship typical of low-Mg eclogites but whole-rock trace-element pattern is typical ofhigh-Mg eclogites

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

(a)

001

01

1

10(c)

001

01

10(b)

VR43452

YK1949

RockNMORB

low-Mg

RockNMORB

high-Mg

(d)

volatile-richhigh-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

phlogopite-bearingYK1911

calcite-bearingVR67112b

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 10 Extended trace-element patterns of reconstructed (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-bearing eclogites Normalized to N-MORB of Sun amp McDonough (1989)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1860

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

RockNMORB

(b)

RockNMORB

(d)

010

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu001

01

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

(a)

001

01

1

10(b)

001

01

1

10(c)

001

01

1

10(e)

high-TSchmidberger et al

low-TSchmidberger et al

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

high-Mglow-Mg

high-Cavolatile-rich

phlogopite-bearingYK1911

calcite-bearingVR67112b

YK1949VR43479

Fig 11 N-MORB normalized REE patterns for (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-richeclogites Shown for comparison in (e) are reconstructed eclogites from Schmidberger et al (2007)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1861

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Both volatile-rich whole rocks display strong negativeSr anomalies Strontium can be sited in calcite or apatiteandor indicate the presence of a Sr-rich mineral that wasnot identified in thin section such as bariteFour of the six reconstructed whole-rocks in the study

of Schmidberger et al (2007) (Fig 11e) show markedLREEHREEN enrichment that is observed only in onehigh-Ca eclogite in our study (Fig 11c) suggesting thatmany of the eclogites they studied were affected by metaso-matic LREE enrichment Two samples in their study thatare not affected by this enrichment show positive Euanomalies similar to the high-Mg eclogites in this studywhereas the distinctive REE pattern of the low-Mg eclo-gites is not recognized among their samples

SR^ND^HF I SOTOPESYSTEMATICSStrontium Nd and Hf isotope data for optically puremineral separates and reconstructed whole-rocks aregiven in Table 5 Spongy rims and alteration are likely tobe mostly removed during leaching in 6N HCl in combina-tion with repeated ultrasonication Despite this rigoroustreatment the reproducibility of Nd in garnet VR43469and in cpx YK1914 is worse than for other samples Thismay be due to inhomogeneities related to late Nd diffusionpossibly from grain boundaries containing a kimberlitecomponent The reproducibility of the 176Hf177Hf data isgenerally poorer than that of Nd which may reflect a com-bination of matrix effects caused by imperfectly purifiedsolutions and sample heterogeneity Parent^daughterratios are available only from in situ analyses which arenot precise enough to allow meaningful model age orinitial isotope ratio calculations (ie at the time of kimber-lite eruption) Because the Lac de Gras kimberlites arerelatively young (55^56 Ma Graham et al 1999) age cor-rection would result in only a minor change in isotopecomposition which is inconsequential in view of the rangeof compositions displayed by most samples and hence hasno effect on the conclusions drawnClinopyroxenes in five volatile-free eclogites and a

pyroxenite have Sr isotope ratios ranging from 0701662 to0703546 this range is similar to or somewhat lower thanthat of the depleted mantle at the time of kimberliteemplacement (0703 Bell amp Tilton 2002) Clinopyroxenein the phlogopitized sample has a distinctly high 87Sr86Srof 0751186Clinopyroxene controls 56^92 of the Nd budget (Nd

concentration in cpx weighted by mode) Neodymiumisotope ratios range from 0511660 to 0514844 Thiscorresponds to eNd variation (per 10 000 deviationof 143Nd144Nd in the sample from 143Nd144Nd in the chon-dritic mantle model reservoir) from 17 to thorn43 cpx inthe low-Mg eclogites have higher eNd than those in

high-Mg eclogites Garnets have generally have higher143Nd144Nd from 0512984 (eNdfrac14thorn 7) to 0520965(eNdfrac14thorn 162) Garnets in the two volatile-rich sampleshave the highest ratios 0522790 for garnet in the calcite-bearing sample VR67112b (eNdfrac14thorn198) and 0525721 inthe phlogopitized eclogiteYK1911 (eNdfrac14thorn 255)Garnet contributes 18^32 and cpx 48^78 of the Hf

budget The proportion controlled by rutile is 4^30(Aulbach et al in preparation) Measured 176Hf177Hf incpx ranges from 0281263 in phlogopite-bearing sampleYK1911 (eHffrac1453) to 0286130 in low-Mg eclogiteVR43469 (eHffrac14thorn119) and in garnet from 0283958 inpyroxeniteYK1915 (eHffrac14thorn42) to 0295718 in low-Mg eclo-gite YK1926 (eHffrac14thorn458) Hafnium in garnet in the phlo-gopite^apatite-bearing sample is extremely radiogenic(176Hf177Hffrac14 0354367) corresponding to an eHf of thorn2532which is close to a value determined for a garnet separatefrom a Roberts Victor eclogite (eHffrac14thorn2561 Jacob et al2005) Garnet in the calcite-bearing sample has extremelyradiogenic Nd and Hf (eNdfrac14thorn193 eHffrac14thorn2516) Garnetin pyroxenite YK1915 also has radiogenic Nd and Hf(eNdfrac14thorn27 eHffrac14thorn42)Despite the large uncertainties for reconstructed whole-

rocks (calculated as the square root of the sum of squareduncertainties of two standard errors for garnet and cpxanalyses combined with 20 propagated uncertainty onthe proportion of Nd or Hf respectively contributed byeach mineral) the isotopic data are consistent in thatthey correlate with the independently determinedparent^daughter ratios (1) Low-Mg eclogites have higherSmNd reflected in the steeper positive slope of theirN-MORB-normalized REE patterns (Fig 11a) comparedwith the high-Mg eclogites (Fig 11b) Accordingly recon-structed low-Mg eclogites have higher present-day eNd

(0 to thorn47) than high-Mg eclogites (4 and 17) (2) Low-Mg eclogite YK1943 has very high Nb and LREE abun-dances that could have been added during metasomatismthat also enhanced its Hf abundance This sample has thelowest eNd and eHf of the low-Mg eclogites (0 and thorn20respectively) In contrast apparently unenriched low-Mgeclogites have more radiogenic Nd and Hf (eNdfrac14thorn47and thorn9 eHffrac14thorn128 and thorn113 respectively)The Sr^Nd isotope diagram reveals that eNd in cpx

varies from 19 to 43 at nearly constant depleted-mantle-like 87Sr86Sr (Fig 12a) In the Nd^Hf isotope diagram(Fig 12b) the reconstructed whole-rocks lie on curvedmixing lines if the lsquomixingrsquo proportions of Nd and Hffor the two lsquoend-memberrsquo minerals are very different(eg YK1911) Garnets have more radiogenic Nd andHf than coexisting cpx All reconstructed volatile-freeeclogites plot in the radiogenic field with regard to Hfor are isotopically similar to the depleted mantle Twohigh-Mg eclogites and one low-Mg eclogite have unradio-genic Nd one high-Mg eclogite has radiogenic Nd and

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1862

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 5 Individual and averaged Sr Nd and Hf isotope ratios obtained from garnet and cpx solutions

Sample 87Sr86Sr 2SE2SD 143Nd144Nd 2SE2SD eNd 2SE2SD 176Hf177Hf 2SE2SD eHf 2SE2SD

Low-Mg eclogites

VR43469 cp 0702661 0000010 0514844 0000006 43 01 0286130 0000012 119 04

VR43469 gt 0515508 0000014 56 03 0286506 0000012 132 04

VR43469 gt r 0515453 0000010 55 02 0286548 0000011 134 04

VR43469 gt r 0515384 0000007 54 01 0287244 0000010 158 04

VR43469 gt r 0515487 0000014 56 03 0286996 0000014 149 05

VR43469avg gt 0515458 0000109 55 21 0286823 0000715 143 253

WR VR43469 0515059 0000273 47 5 0286405 0001079 128 38

YK1926 cp 0703546 0000011 0513043 0000007 8 0 0283154 0000009 13 03

YK1926 gt 0513364 0000008 14 01 0295718 0000005 458 02

WR YK1926 0513120 0000026 9 1 0285980 0000716 113 25

YK1943 cp 0512143 0000007 10 01 0282213 0000005 20 02

YK1943 gt 0514461 0000012 36 02 0287510 0000008 168 03

YK1943 gt r 0514491 0000009 36 02 0287332 0000005 161 02

YK1943 gt r 0514517 0000010 37 02 0287290 0000007 160 02

YK1943avg gt 0514489 0000056 36 11 0287377 0000233 163 82

WR YK1943 0512625 0000228 0 5 0283138 0000879 13 31

High-Mg eclogites

VR43480 gt 0520965 0000017 162 03 0284950 0000010 77 04

VR67360 cp 0283509 0000074 26 26

VR67360 cp r 0283433 0000031 23 11

VR67360avg cp 0701662 0000018 0511660 0000010 19 02 0283471 0000108 25 38

VR67360 gt 0512984 0000101 7 20 0287429 0000180 165 64

WR VR67360 0511761 0000132 17 3 0284734 0000469 69 17

YK1949 cp 0703019 0000013 0512231 0000009 8 02 0282513 0000009 9 03

YK1949 gt 0513511 0000010 17 02 0290297 0000042 266 15

YK1949 gt r 0513508 0000012 17 02 0289947 0000015 254 05

YK1949 gt r 0513552 0000014 18 03

YK1949avg gt 0513523 0000049 17 10 0291569 0000494 311 175

WR YK1949 0512446 0000103 4 2 0284622 0000847 65 30

Volatile-rich eclogites

VR67112b gt 0522790 0000008 198 02 0354367 0000036 2532 13

YK1911 cp 0751186 0000009 0512165 0000007 9 01 0281263 0000009 53 03

YK1911 gt 0525721 0000013 255 03 0288051 0000012 187 04

WR YK1911 0513940 0000342 25 7 0281414 0000018 48 1

Pyroxenites

YK1914 cp 0703015 0000034 0513283 0000041 13 08

YK1914 cp r 0702956 0000005 0513200 0000027 11 05

YK1914 cp r 0513527 0000059 17 11

YK1914 cp r 0513045 0000010 8 02

YK1914avg cp 0702986 0000083 0513264 0000402 12 78

YK1915 gt 0514003 0000011 27 02 0283958 0000021 42

Analyses of full procedural replicates are designated with an lsquorrsquo after sample name SE standard error (internal errorsingle analyses) SD standard deviation (external error averaged multiple analyses) Italics used for reconstructed whole-rocks (mineral isotopic compositions and concentrations weighted by modes) Uncertainties for reconstructed whole-rocksinclude 2SE or 2SD uncertainties on each mineralrsquos isotope composition combined with 20 uncertainty in the amountof Nd or Hf contributed by the constituent minerals eNd and eHf are per 10 000 deviation from model chondritic mantle[model of Wasserburg et al (1981) and Blichert-Toft amp Albarede (1997) respectively]

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1863

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

εHf

εNd

Ekati eclogitesminus100

0

100

200

300

400

500

minus40 0 40 80 120 160 200 240 280

gt in YK1915

gt in VR67112bεNd=198 εHf=2532

minus40

0

40

80

120

160

minus40 minus20 0 20 40 60

minus60

minus40

minus20

0

20

40

070 071 072 073 074 075 076

North Americankimberlite field

εNd

87Sr86Sr

cpx in YK1914

Ekatieclogites

DM

DM

DM

DM

(a)

(b)

pyroxenites

low-Mg

high-Mg

volatile-rich

eclogites

Diavik low-TA154 S kimberlite

cpx in YK1911

Fig 12 (a) Sr^Nd for cpx and (b) Nd-Hf isotope diagram for garnet cpx and reconstructed whole-rocks (symbols with bold outline) Finedashed lines in (b) connect data points from the same sample with garnets having more radiogenic isotopic compositions Uncertainties forreconstructed whole-rocks combine 20 modal uncertainty with uncertainties on each mineralrsquos isotope composition Error bars on individualminerals may be smaller than the symbols Also shown are the isotopic compositions of the depleted mantle (DM DePaolo 1981 Griffin et al2000 Bell amp Tilton 2002) Range of 87Sr86Sr in eclogites from Ekati (Lac de Gras area 100 km north of Diavik Jacob et al 2003) in NorthAmerican kimberlites and compositions of low-temperature eclogites from Diavik and of Lac de Gras A154 South kimberlite fromSchmidberger et al (2007)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1864

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

one low-Mg eclogite has eNd similar to the depleted mantle(Fig 12b)Clinopyroxene in the phlogopitized eclogite YK1911

has unradiogenic Nd and Hf and strongly radiogenicSr whereas the coexisting garnet has radiogenic Nd andHf Although this sample contains 50 vol phlogopiteits whole-rock isotopic composition was calculated

from garnet and cpx alone because the contributionfrom phlogopite to the Hf and Nd budget is 52 and51 respectively (unpublished data) The whole-rockhas radiogenic Nd but has lower eHf than anyother sample analysed in accord with the extreme Hfabundance (53 ppm) and low LuHf of the cpx

DISCUSS IONOrigins of eclogites beneath the SlaveCratonHigh-Mg and high-Ca eclogites

With regard to their major- and trace-element contentsthe high-Mg and high-Ca eclogites from Lac de Gras arebroadly similar to variably evolved oceanic gabbros(Fig 13b and c) which may indicate an origin as subductedoceanic crust Compatible trace-element abundancesare useful to deduce potential crustal protoliths as theseelements are less susceptible to post-formation modification(Jacob 2004) The flat HREE patterns in the high-Mg andhigh-Ca eclogites are compatible with a low-pressureorigin as the oceanic crust is a product of moderate degreesof partial melting mainly within the spinel stability field(eg Presnall et al 2002) whereas melting with residualgarnet would qualitatively be expected to produce stronglyfractionated REE patterns Positive Sr and Pb anomaliesalthough compatible with a plagioclase signature couldalso be due to cryptic metasomatism (Dawson 1984) ofthe eclogites during their residence in the lithosphericmantle These anomalies are probably not related to kim-berlite infiltration as the whole-rock eclogite compositionswere reconstructed from analyses of fresh mineral coresLow ZrSm similar to that in primitive gabbros is

observed for many of the high-Mg and high-Caeclogites (Fig 13b and c) and is independent of the ratioof garnetcpx and whether or not the whole-rock wasreconstructed with rutile (Electronic Appendix 10) Thenegative trend of Mg-number against Yb in garnet(Fig 14) argues against igneous processes involving garnetaccumulation (Taylor amp Neal 1989) as suggested forexample for some high-Mg eclogites from West Africa(Barth et al 2002b) Rather the anti-correlation suggeststhat olivine fractionation played a role in the formation ofthe protoliths Fractionating olivine would excludeYb (Dolivinebasalt

frac14 002 Nikogosian amp Sobolev 1997)and incorporate Mg relative to Fe (Dfrac1415 and 05 respec-tively Taura et al1998) leading to progressive depletion ofMg and enrichment of Yb in the residual melt and henceanti-correlated Mg and Yb in the whole-rock This isinherited after eclogitization All eclogite types (exceptVR43452) plot along the olivine fractionation trend sug-gesting that they formed by a similar processHigh-Mg eclogites have lower

PREE than high-Ca

eclogites and some appear to have positive Eu anomalies

high-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

gabbrosSE Indian ridge

gabbrosOman ophiolite

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

high-Mg

RockNMORB(a)

VR43452

YK1949

low-Mg

boninite-like melts

01

001

1

10

(b)

01

001

1

10

(c)

01

001

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

Fig 13 Trace-element patterns of (a) low-Mg (b) high-Mg and(c) high-Ca eclogites compared with boninite-like volcanic rocksfrom the Isua greenstone belt (Polat et al 2002) gabbros from theOman ophiolite (Benoit et al 1996) and the SE Indian Ridge (Hartet al 1999) normalized to N-MORB of Sun amp McDonough (1989)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1865

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(EuEu up to 14) similar to more primitive cumulates(troctolite gabbros and olivine gabbros) reported in the lit-erature (Fig 13b) Although the sizes of these Eu anomaliesare not outside the analytical uncertainty their correlationwith CeYb in reconstructed eclogites supports their signif-icance (r2frac14066 nfrac14 4) The more primitive nature ofhigh-Mg eclogites is consistent with higher Mg-numbersrelative to high-Ca eclogites and may be due to olivineaccumulation during formation of their protolithsThe two samples with the lowest HREE abundanceshave the highest EuEu which is compatible with varia-tions in plagioclase modesMore evolved protoliths with higher pyroxeneplagio-

clase plus trapped melt could account for small or absentpositive Eu anomalies and higher

PHREE of high-Ca

eclogites (Fig 13c) Although the reason for the frequentassociation of graphite or diamond with high-Ca eclogitesis unclear (Aulbach et al 2003) the frequent occurrence ofkyanite in high-Ca eclogites also observed at Ekati justnorth of Lac de Gras has been argued to support a crustalprotolith (Jacob 2004) Kyanite andor corundum havebeen suggested to exsolve from high-temperature high-Alclinopyroxenes (Caporuscio amp Smyth 1990) Eclogitesfrom Lac de Gras show no exsolution textures suggestiveof such an origin although recrystallization can obscureor destroy microstructural evidence for exsolution(eg Griffin et al 1984)Modelling shows that qualitatively plagioclase accumu-

lation from a low-pressure melt derived from MORBmantle can account for positive Pb Sr and Eu and negativeZr anomalies relative to N-MORB whereas cpx accumula-tion or trapped melt increases Zr Y MREE and HREE

abundances (Fig 15) Some melt addition (for exampleby entrapment) seems to be required to increase all trace-element abundances to those observed for the eclogitesAlthough high-Ca and high-Mg eclogites could representshallow and deep portions of the same oceanic crust thehigher average temperatures obtained for the high-Caeclogites the frequent occurrence of carbon in high-Caeclogites and its absence in high-Mg eclogites the strongLREEHREE fractionation observed for six of seven high-Ca eclogites and the narrow range of compositions of thehigh-Mg eclogites suggest that they may have protolithsthat are not cogenetic Some discrepancies in particularaffecting the incompatible elements may be due to post-crystallization processes acting upon the protoliths suchas dehydration melting and later metasomatism that isevident in peridotites from the same locality (Griffin et al1999)Low K2OTh for high-Ca and high-Mg eclogites

(average 08 Table 4) relative to assumed crustal

0

2

4

6

8

10

30 40 50 60 70 80

Mg

VR43452

Yb

GARNET

pyroxenites

low-Mghigh-Mghigh-Ca

eclogites

VR19673ecl-g2

Fig 14 Garnet Mg-number vs Yb (ppm including 2s uncertainty)Lines qualitatively show separate correlations for eclogites (exceptoutlierVR43452) and pyroxenites (except outlierVR19673ecl-g2)

high-Ca

high-Mg

001

01

1

10

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

melt08 melt02 cryst cpx01 melt09 cryst

plag

(a)

001

01

1

10(b)

RockNMORB

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 15 Reconstructed (a) high-Ca and (b) high-Mg eclogitescompared with compositions of plagioclase and cpx crystallizingfrom a MORB melt modelled using distribution coefficients ofNorman et al (2005)The MORB was calculated as a 10 equilibriummelt of MORB pyrolite [composition from Kelemen et al (1993)except Pb Pr Sr Y and Lu which were calculated as residues from1 fractional melting of pyrolite from McDonough amp Sun (1995)using melt modes of Kinzler (1997) and distribution coefficients ofNorman et al (2005)] Also shown are mixtures of crystals (consistingof 50 plagioclase plus 50 cpx) and melt in the proportions 0901and 0208

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1866

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

precursors (eg gabbros SE Indian Ridge average 22Hart et al 1999) may reflect the preferred loss of fluid-mobile elements during dehydration (Becker et al 2000)although large uncertainties on these low-abundance ele-ments in the constituent minerals and reconstructedwhole-rocks makes this interpretation less than robustThe REE patterns of most high-Mg eclogites are too flatto represent melting residues as partial melting shouldproduce low LREEHREE (Electronic Appendix 10) Incontrast the marked LREE depletion of five of sevenhigh-Ca eclogites (LaLufrac14 02^08) contrasts with that ofassumed evolved gabbroic protoliths (LaLufrac1415^32Hart et al 1999) and can be modelled as a consequence ofpartial melting during subduction after eclogitization (iein the presence of garnet which retains HREE relative toLREE eg Barth et al 2002a) has taken place (Aulbachet al 2003)Our results are compatible with previous studies on

eclogite xenoliths from the Slave Craton includingthose from the same kimberlite that have linked thesesamples or some of these samples to Paleoproterozoicsubduction events (the 195^191Ga formation of theHottah terrane and the 18Ga formation of the GreatBear magmatic arc) in the Slave craton (Griffin et al1999 Kopylova et al 1999 Heaman et al 2002 2006Schmidberger et al 2005 2007)The compositional features discussed above although

compatible with crustal protoliths are not necessarilydefinitive of a subduction origin However if direct crystal-lization from basaltic melts at high pressure (ie mantledepths) is invoked instead specific conditions would haveto be met to satisfy the observations The Mg-number ineclogites from Lac de Gras are lower than those expectedof primary high-pressure mantle melts (eg Walter 1998OrsquoHara amp Herzberg 2002) requiring crystallization athigh pressures after olivine fractionation This originwould further require shielding from peridotitic wallrockwith which the fractionated melts would no longer be inequilibrium (Barth et al 2001) In such a setting the lowrelative abundances of fluid-mobile elements comparedwith those of less fluid-mobile elements and of LREErelative to HREE (Table 4) would reflect an expulsion offluids during crystallization of the protoliths and latermelt mobilization respectively The eclogites clearly donot represent garnet plus cpx cumulates which wouldrequire a positive correlation of Mg-number with Yb thatis at variance with the observed negative correlation ineclogitic garnets from this study (Fig 14)

Low-Mg eclogites

Garnets in low-Mg eclogites compositionally overlap withthose in massive eclogites from Jericho (NW of the Lacde Gras kimberlites) that were suggested to have a high-pressure intrusive origin based on the similarity of theirmajor-element compositions to those of Group A eclogites

fromsouthernAfrica forwhichamantleoriginhasbeenpro-posed (Kopylova et al1999) Garnets in low-Mg eclogites inthe present study have major-element compositions moresimilar to garnets in groupB rather thangroupAeclogitesThe positive slope in N-MORB-normalized HREE

abundances displayed by the low-Mg eclogites is clearlydistinct from other eclogite types and is similar to those ofboninite-like second-stage melts that are closely associatedwith ultramafic volcanics (Smithies et al 2004) (Fig 13a)The range of Mg-numbers in the low-Mg eclogites (55^66)overlaps with that of Archaean second-stage melts fromIsua (58^75) and Abitibi (52^82 Smithies et al 2004)These similarities might indicate that the low-Mg eclogiteshave an origin in a previously depleted mantle source suchas supra-subduction zone mantle Subchondritic ZrHf ofthe low-Mg eclogites contrasts with the supra-chondriticZrHf of the high-Mg and high-Ca eclogites (Aulbach et

al in preparation) and is consistent with previous cpx-controlled depletion in an arc mantle source This sourcecould have been subsequently remelted to generate theboninite-like precursor to the low-Mg eclogites as sug-gested for the protoliths of eclogites from Udachnaya(Jacob amp Foley 1999)Although some second-stage melts to which the low-Mg

eclogites are similar are closely associated with komatiitesand could contain a plume component the radiogenic ordepleted mantle-like Nd and Hf and unradiogenic Sr inthe low-Mg eclogites do not support plume involvement

Trace-element and Sr^Nd^Hf isotopeconstraints on post-formationmetasomatismSeveral low-Mg and high-Ca eclogites show negativeslopes in the LREE which may be indicative of interactionwith small amounts of LREE-enriched melt (Navon ampStolper 1987) amounting to cryptic metasomatism(ie not accompanied by crystallization of new phases) inthe nomenclature of Dawson (1984) High Nb abundancesrelative to elements of similar compatibility displayed bysome low-Mg and high-Ca eclogites (Fig 10) could resultfrom overestimation of rutile modes However high-CaeclogiteVR40345 is rutile-free and has been reconstructedwithout rutile its high Nb abundance is accompanied byLREE enrichment indicating that metasomatic additionmay be responsibleLow-Mg eclogites show a positive correlation of La with

Nb (Electronic Appendix 11a) indicating concomitantaddition of LREE and Nb This suggests that the metaso-matic agent was a solute-rich fluid or silicate silicic melt ashydrous fluids and carbonatites do not carry enough HFSEto affect the budget of eclogites (Green amp Wallace 1988Klemme et al 1995 Adam et al 1997 Stalder et al 1998)Of the samples that are enriched in Nb all but one (high-Ca eclogiteVR50909) show a simultaneous enrichment in

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1867

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Hf (Electronic Appendix 11b) Together these trace-element relationships suggest that both SmNd and LuHfwere lowered in these samples Their low parent^daughterratios contrast with higher ratios in unenriched samplesand resulted in the ingrowth of a range of Nd and Hf iso-tope compositions Compared with the low-Mg and high-Ca eclogites the high-Mg eclogites have a narrow range ofLa and Nb concentrations which correlates with a narrowrange in Nd and Hf isotope compositions (Fig 12)Precise model ages cannot be calculated because parent^

daughter ratios are available only from in situ analyseswhich have relatively large uncertainties Howeverevidence for long-term enrichment and depletion may bediscerned Low-Mg eclogites even those that are enrichedin LREE have relatively high SmNd as a result of havingoverall positive slopes in their REE patterns relative toMORB whereas the high-Mg eclogites have relativelyflat REE patterns and therefore lower SmNd (Fig 11)This translates into lower eNd for high-Mg eclogites thanfor low-Mg eclogites In contrast they show no systematicdifference with respect to 176Hf177Hf Because the low- andhigh-Mg eclogites are proposed to have different origins asdiscussed above and therefore did not necessarily have thesame initial isotopic composition and formation age theseisotopic differences cannot be used to estimate their agesA long-term coupling of trace-element and isotopic

abundances is also indicated by the fact that low-Mg eclo-gite YK1943 which has experienced Nb and LREE addi-tion has lower eNd (0) and eHf (thorn13) than low-Mg eclogiteVR43469 with higher LuHf and SmNd (eNdfrac14thorn47eHffrac14thorn128) If low-Mg eclogites YK1943 and VR43469originally had identical initial Nd and Hf isotope composi-tions and if the parent^daughter ratios are accurate andwere modified only once during their evolution shortlyafter metamorphism (when their isotopic compositionwould have been reset) this overprint can be calculatedfor both isotopic systems to have occurred at 17GaThis would suggest that Hf and Nd were added duringthe same event In contrast high-Mg eclogites VR67360and YK1949 would have had identical Ndisotope compositions at 10Ga whereas their Hf isotopecompositions are disturbed in that the sample with thehigher parent^daughter ratio has a lower 176Hf177Hf thanthe sample with the lower parent^daughter ratio Theseresults suggest that the LREE and HFSE enrichmentobserved in different samples occurred during separateevents that did not affect the eclogites equallyExtreme HFSE enrichment leading to the growth of

zircon and rutile and LREE addition leading to thegrowth of apatite in eclogites from the Jericho kimberlitein the northern Slave craton have been ascribed to twoseparate metasomatic events at around 18Ga (quasi-coincident with the age of formation and metamorphism)and 13^10Ga respectively (Heaman et al 2002 2006

Schmidberger et al 2005) The authors associated theseevents with Paleoproterozoic subduction and aMesoproterozoic thermal disturbance in the Slave litho-spheric mantle possibly linked to the Mackenzie igneousevent Mesoproterozoic Nd model ages mostly between13 and 14Ga were reported for eclogites from thenearby Ekati kimberlites (Jacob et al 2003) Although wedo not observe zircon or apatite in volatile-free eclogitesfrom Lac de Gras and the age constraints obtained fromour data are weak they support a link between theseevents described for other eclogite populations and themetasomatic enrichment observed in our studyThe Sr isotope data in the present study are similar to

those of Schmidberger et al (2007) who suggested that thedepleted-mantle-like values are consistent with derivationof the protoliths from an oceanic mantle source However87Sr86Sr in volatile-free eclogites in the present study isremarkably constant considering their Nd and Hf isotopevariability (Fig 12) and the variability of trace-elementpatterns (Fig 10) Given the evidence for dehydration andmelting in many of the eclogites fractionation of highlyincompatible Rb from Sr and differential ingrowth ofradiogenic Sr would be expectedThis may indicate a rela-tively recent overprint and considering the similarityof 87Sr86Sr to present-day depleted mantle which is signif-icantly lower than 87Sr86Sr of North American kimber-lites a juvenile (asthenospheric) source

PyroxenitesPyroxenites from Lac de Gras often show relationshipsin their major-element contents that are different fromthose observed in the eclogites For example pyroxeniticgarnets show a negative correlation between CaO andNa2O contents whereas these oxides are positively corre-lated for garnets in eclogites (Fig 5) CaO is well correlatedwith C2O3 in pyroxenitic garnets but not in eclogitic gar-nets (Fig 4) Mg-number is positively correlated withYb abundance for pyroxenitic garnets but anti-correlatedfor eclogitic garnets (Fig 14) This suggests that the pyrox-enites in the present study are not related to the eclogitesby fractionation or accumulation processes as proposedfor the origin of some pyroxenites (Foley et al 2003)Pyroxenites have also been interpreted as the products

of crystal segregation in magma channels with alteredoceanic crust as the magma source this would explaintheir Eu anomalies and stable and radiogenic isotopecompositions (Pearson et al 1993 Becker 1996) As thehigh-pressure liquidus phases of most terrestrial basaltsare cpx and garnet (OrsquoHara amp Herzberg 2002) partialmelts of eclogites with basaltic composition wouldcrystallize cpx and garnet but not opx More than half ofthe pyroxenites from Lac de Gras contain opx suggestingthat they probably did not crystallize directly fromeclogite-derived melts

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1868

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Low-Cr websterites and pyroxenites could be crystalli-zation products of mafic melts that interacted with perido-titic mantle thus becoming more refractory The positivecorrelation of Yb and Mg-number in garnet would allowan origin by fractional crystallization from such meltsAlternatively they formed by reaction of silicic (possiblyeclogite-derived) melts with mantle peridotites(eg Kelemen et al 1993 Aulbach et al 2002) This agreeswith the similarity of hypothetical melts in equilibriumwith pyroxenitic garnets in most samples to eclogite-derived melts (Electronic Appendix 12) Niobium concen-trations in all hypothetical melts are much higher than ineclogite-derived melts regardless of whether or not rutileis assumed to be in the residue The same is true for LaCe and Zr concentrations in some samples which mayindicate that NbZr and LREE were added to thepyroxenites some time after their formation as discussedin the previous section for some eclogites Because mini-mum ages for pyroxenitic rutile with an unradiogenicHf isotope composition (calculated by projecting the176Hf177Hf to the model depleted mantle isotopic evolutioncurve assuming zero ingrowth) range to 178066Ga(Aulbach et al in preparation) it appears likely that thepyroxenites were not formed recently (eg during kimber-lite-related or precursory melt migration) and that some ofthe samples are at least MesoproterozoicGarnet pyroxeniteYK1952 consists of schlieren of equili-

brated garnet pyroxenite in a lsquomatrixrsquo of spongy cpxfine cpx^opx intergrowths embayed garnet and meltpatches (Fig 2f) Garnet in the pyroxenite schlieren hasvery low LREE and HFSE abundances Orthopyroxenein the matrix has distinctly higher Mg-number andCr-number and contains more CaO than opx from thepyroxenite schlieren whereas opx rims in the schlierenare compositionally similar to opx in the matrix The con-siderable compositional and textural disequilibrium inthis sample suggests that this is a young feature causedby reaction with fluids shortly before or at the time ofentrainmentAs a consequence of the small sample sizes of most of

the pyroxenites studied here a comprehensive isotopicdataset is not available Clearly more such data areneeded in addition to stable isotope data to obtain betterconstraints on the origins of this rock type Consideringthe microstructural and compositional variability ofthe pyroxenites and major-element relationships oppositeto those observed for eclogites it appears certain that pyr-oxenites form by a variety of processes some of which areunrelated to the formation of eclogites and their protoliths

Volatile-rich eclogitesVolatile-rich eclogites from Lac de Gras display the mostextreme Hf^Nd isotope systematics of all the analysedsamples Garnets in phlogopitized (YK1911) and in cal-cite-bearing eclogite (VR67112b) have the most radiogenic

Nd (eNdfrac14 2486 and 1927 respectively) Garnet inVR67112b has extremely high eHf (2516) Clinopyroxene inYK1911 has the lowest eHf measured for any mineralseparate from Lac de Gras (53) in accord with the veryhigh Hf abundance in its cpx (5 ppm) In contrastgarnet in this sample has radiogenic Hf (eHffrac14 136) Thisdiscrepancy is at least in part due to their relatively lowequilibration temperatures (7698C and 8558C respec-tively) which allowed the isotopic compositions of cpxand garnet to evolve separately at their low andhigh parent^daughter element ratios respectively forextended periods of time and possibly since formationAccordingly the cpx in YK1911 gives a minimum Hf ageof 27GaSlab-derived fluids could have been introduced during

the 27Ga collision of the western Slave Craton domainincluding the thick plume-derived deep lithosphericmantle layer and the eastern domain with its highlydepleted lithosphere This would also be consistent withthe observation that volatile-rich eclogites occur at shallowdepths in the mantle (TKroghfrac14 685^8558C correspondingto depths of 85^115 km along a 40 mWm2 geotherm)where the bulk of dehydration and production of solute-rich melts in a subducting slab might be expected to occur(Becker et al 2000 Hermann et al 2006) Their negativeEu anomalies might indicate that the fluid was derivedfrom the shallow portion of the slab that experiencedplagioclase fractionation The presence of carbonate inVR67112b requires a carbonated sourceHigh 87Sr86Sr in cpx in YK1911 is probably the conse-

quence of isotopic equilibration with high-RbSr phlogo-pite after long-term radiogenic ingrowth while Hfaddition led to retarded ingrowth of radiogenic HfThese systematics combined with radiogenic Nd appear

inconsistent with the involvement of sediments in the pet-rogenesis of this sample (eg Perfit et al 1980 Vervoortet al 1999) or a hydrous fluid such as those responsible forMARID metasomatism in the Kaapvaal craton (Konzettet al 1998)Whatever the origin of the fluids the apparentHf mobility suggests that they were not in equilibriumwith a rutile-bearing source and also that they weresolute-rich in nature (ie hydrous melts able to dissolveHFSE) and the relatively high HREE abundances of thevolatile-rich eclogites indicate a garnet-free source suchas amphibolite This allowed evolution of the sample atrelatively high time-integrated SmNd

SUMMARY AND CONCLUSIONS

(1) The Lac de Gras kimberlites carried abundant eclo-gite and pyroxenite xenoliths to the surface Threemain eclogite types are distinguished in terms ofmajor-element composition and trace element

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1869

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

patterns high-Mg high-Ca and low-Mg eclogitesThe garnets of all eclogite types have anti-correlatedMg-number and Yb pointing to olivine fractionationas a controlling factor in the formation of their pre-cursor rocks Their flat or positively sloped HREEpatterns qualitatively indicate that the melt parentalto the precursor rocks was not in equilibrium with agarnet-bearing residue

(2) Reconstructed high-Mg eclogites and high-Caeclogites have flat HREE patterns positive Pb andSr anomalies and low ZrSm similar to oceanic gab-bros reported in the literature High-Mg eclogiteshave variably low HREE contents and some havesmall positive Eu anomalies possibly suggestive ofmore primitive relatively plagioclase-rich protolithswhereas high-Ca eclogites have higher HREE abun-dances and small or absent Eu anomalies that requiremore cpx- andor melt-rich protoliths if a subductionorigin is assumed

(3) Low-Mg eclogites have distinctive trace-element pat-terns that show a marked positive slope in the HREEand subchondritic ZrHf They resemble someArchaean second-stage melts (lsquoboninitesrsquo) and theirprotoliths appear to have formed from a previouslydepleted source

(4) After formation the eclogites were subject to a varietyof processes Most have low ratios of fluid-mobile vsless fluid-mobile elements possibly as a result of dehy-dration of the protoliths In addition some of the low-Mg and high-Ca eclogites have low LREEHREEconsistent with partial melt lossWhereas the low-Mgeclogites show concomitant enrichment in La Nb andZr some high-Mg and high-Ca eclogites show evi-dence for variable and unsystematic (decoupled)LREE and Nb addition

(5) Isotopically the trace-element characteristics of thedifferent eclogite types translate into lower eNd forhigh-Mg eclogites (negative slope in MREEfrac14 lowerSmNd) than for low-Mg eclogites (positive slope inMREEfrac14higher SmNd) Within the low-Mg groupsamples that show evidence for metasomatism (lowLuHf and SmNd) have lower eNd and eHf than asample that was apparently not metasomatized BothNd and Hf addition can be dated to 17Ga if identi-cal initial isotope compsitions are assumed for thelow-Mg eclogites In contrast the relatively constantand depleted mantle-like 87Sr86Sr of eclogitic cpxdespite variable Nd and Hf isotope compositions andparent^daughter ratios suggests recent input of juve-nile material

(6) Garnet and cpx in pyroxenites show major-elementrelationships different from those in eclogitesTherefore these two rock types are probably notrelated by fractionation processes The presence of

opx precludes direct crystallization from slab-derivedmelts Pyroxenites may be the product of (slab-derived) melts hybridized during assimilation of peri-dotitic mantle

(7) Garnets in phlogopite- and carbonate-bearing eclo-gites have extreme isotopic compositions includingthe most radiogenic Nd and Hf in the datasetNegative Eu anomalies may point to the involvementof a crustal protolith Highly radiogenic Sr radiogenicNd and unradiogenic Hf in the reconstructedphlogopite-bearing sample combined with highHREE abundances and evidence for Hf mobilitymay indicate involvement of solute-rich fluids origi-nating in a garnet- and rutile-free source

ACKNOWLEDGEMENTSHelp in the clean laboratories and with the analyticalfacilities by S Elhlou C Lawson A Sharma P Wielandand M Zhang is greatly appreciated Bill Griffin isthanked for his support and for agreeing to reviewseveral versions of the manuscript despite disagreeingwith many of the conclusions presented here The paperbenefited greatly from the detailed and constructivereviews provided by the editors Gareth Davies andMarjorie Wilson and by Matthias Barth Larry Heamanand an anonymous reviewer This work was funded by aMacquarie University International PostgraduateAward and Postgraduate Research Fund (SA) by anARC SPIRT grant sponsored by Kennecott Canada Incand by ARC Large and Discovery Grants to SYOrsquoRAnalytical data were obtained using instrumentationfunded by ARC LIEF and DEST Systemic InfrastructureGrants industry partners and Macquarie UniversityThis is Publication 481 from the ARC National KeyCentre for Geochemical Evolution and Metallogeny ofContinents (wwwesmqeduauGEMOC)

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online

REFERENCESAdam J GreenT H Sie S H amp Ryan C G (1997) Trace element

partitioning between aqueous fluids silicate melts and mineralsEuropeanJournal of Mineralogy 9 569^584

Aulbach S Stachel T Viljoen K S Brey G P amp Harris J W(2002) Eclogitic and websteritic diamond sources beneath theLimpopo Beltccedilis slab-melting the link Contributions to Mineralogy

and Petrology 143 56^70Aulbach S GriffinW L Pearson N J OrsquoReilly S Y Kivi K amp

Doyle B (2003) Origins of eclogites beneath the central SlaveCraton Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA-0011

Aulbach S Griffin W L Pearson N J OrsquoReilly S Y amp Kivi K(2004a) Mantle formation and evolution Slave Craton constraints

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1870

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

from HSE abundances and Re^Os isotope systematics of sulfideinclusions in mantle xenocrysts Chemical Geology 208 61^88

Aulbach S Griffin W L OrsquoReilly S Y amp McCandless T E(2004b) Genesis and evolution of the lithospheric mantle beneaththe Buffalo HeadTerrane Alberta (Canada) Lithos 77 413^451

Aulbach S GriffinW L Pearson N J OrsquoReilly S Y amp Doyle B(2007) Lithosphere formation in the central Slave Craton(Canada) plume subcretion or lithosphere accretion Contributionsto Mineralogy and Petrology (in press) doi 101007s00410-007-0200-1

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2001)Geochemistry of xenolithic eclogites from West Africa Part IA link between low MgO eclogites and Archean crust formationGeochimica et Cosmochimica Acta 65 1499^1527

Barth M G Foley S F amp Horn I (2002a) Partial melting inArchean subduction zones constraints from experimentally deter-mined trace element partition coefficients between eclogitic miner-als and tonalitic melts under upper mantle conditions PrecambrianResearch 113 323^340

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2002b)Geochemistry of xenolithic eclogites from West Africa part 2Origins of the high MgO eclogites Geochimica et Cosmochimica Acta

66 4325^4345Beard B L Fraracci K NTaylor L A Snyder G A Clayton R

A Mayeda T K amp Sobolev NV (1996) Petrography andgeochemistry of eclogites from the Mir kimberliteYakutia RussiaContributions to Mineralogy and Petrology 125 293^310

Becker H (1996) Geochemistry of garnet peridotite massifs fromlower Austria and the composition of deep lithosphere beneatha Palaeozoic convergent plate margin Chemical Geology 134 49^65

Becker H Jochum K P amp Carlson R W (2000) Trace elementfractionation during dehydration of eclogites from high-pressureterranes and the implications for element fluxes in subductionzones Chemical Geology 163 65^99

Bell K amp Tilton G R (2002) Probing the mantle The story fromcarbonatites EOS Transactions American Geophysical Union 83275^277

Benoit M Polvecurren M amp Ceuleneer G (1996) Trace element andisotopic characterisation of mafic cumulates in a fossil mantlediapir (Oman ophiolite) Chemical Geology 134 199^214

BleekerW (2003) The late Archean record a puzzle in ca 35 piecesLithos 71 99^134

BleekerW Ketchum JW F JacksonV A amp Villeneuve M (1999)The central Slave Basement Complex Part I Its structural topol-ogy and autochthonous core Canadian Journal of Earth Sciences 361083^1109

Blichert-Toft J amp Albare de F (1997) The Lu^Hf isotope geochemis-try of chondrites and the evolution of the mantle^crust systemEarth and Planetary Science Letters 148 243^258

Blichert-Toft J Chauvel C amp Albare de F (1997) Separation of Hfand Lu for high precision isotope analysis of rock samples by mag-netic sector-multiple collector ICP-MS Contributions to Mineralogy

and Petrology 127 248^260Brey G P amp Kolaquo hler T (1990) Geothermobarometry in four-phase

lherzolites II New thermobarometers and practical assessment ofexisting thermobarometers Journal of Petrology 31 1353^1378

Brey G P Kolaquo hlerT amp Nickel K G (1990) Geothermobarometry infour-phase lherzolites I Experimental results from 10 to 60 kbJournal of Petrology 31 1313^1352

Caporuscio F amp Smyth J (1990) Trace element crystal chemistryof mantle eclogites Contributions to Mineralogy and Petrology 105550^561

Carbno G B amp Canil D (2002) Mantle structure beneath the SWSlave Craton Canada constraints from garnet geochemistry in theDrybones Bay Kimberlite Journal of Petrology 43 129^142

Coleman R G Lee E D Beatty L B amp Brannock W W (1965)Eclogites and eclogites their differences and similaritiesGeological Society of America Bulletin 76 483^508

Creaser R A Grulaquo tter H Carlson J amp Crawford B (2004)Macrocrystal phlogopite Rb^Sr dates for the Ekati propertykimberlites Slave Province Canada evidence for multiple intru-sive episodes in the Paleocene and Eocene Lithos 76 399^414

Davies R M GriffinW L Pearson N J Andrew A S Doyle BJ amp OrsquoReilly S Y (1999) Diamonds from the deep Pipe DO-27Slave Craton Canada In Gurney J J Gurney J L Pascoe MD amp Richardson S H (eds) Proceedings of the 7th International

Kimberlite Conference CapeTown Red Roof Design pp 148^155Davies R M Griffin W L OrsquoReilly S Y amp Doyle B J (2004)

Mineral inclusions and geochemical characteristics of microdia-monds from the DO27 A154 A21 A418 DO18 DD17 and RanchLake kimberlites at Lac de Gras Slave Craton Canada Lithos 7739^55

DavisW J amp Hegner E (1992) Neodymium isotopic evidence for thetectonic assembly of Late Archean crust in the Slave Provincenorthwest Canada Contributions to Mineralogy and Petrology 111493^504

Davis W J Jones A G Bleeker W amp Grutter H (2003)Lithosphere development in the Slave craton a linked crustal andmantle perspective Lithos 71 575^589

Dawson J B (1984) Contrasting types of upper-mantle metasoma-tism In Kornprobst J (ed) Kimberlites II The Mantle and

Crust^Mantle Relationships Amsterdam Elsevier pp 289^294DePaolo D J (1981) Neodymium isotopes in the Colorado Front

Range and crust^mantle evolution in the Proterozoic Nature 291193^196

Erlank A J amp Kushiro I (1970) Potassium contents of syntheticpyroxenes at high temperatures and pressures Carnegie Institution ofWashingtonYearbook 68 433^439

Foley S F Barth M G amp Jenner G A (2000) Rutilemelt partitioncoefficients for trace elements and an assessment of the influence ofrutile on the trace element characteristics of subduction zonemagmas Geochimica et Cosmochimica Acta 64 933^938

Foley S F Buhre S amp Jacob D E (2003) Evolution of the Archaeancrust by delamination and shallow subduction Nature 421 249^252

Graham I Burgess J L Bryan D Ravenscroft P J Thomas EDoyle B J Hopkins R amp Armstrong K A (1999) Explorationhistory and geology of the Diavik Kimberlites Lac de GrasNorthwest Territories Canada In Gurney J J Gurney J LPascoe M D amp Richardson S H (eds) Proceedings of the 7th

International Kimberlite Conference Cape Town Red Roof Designpp 262^279

Green D H amp Wallace M E (1988) Mantle metasomatism byephemeral carbonatite melts Nature 336 459^462

Green T H Blundy J D Adam J amp Yaxley G M (2000) SIMSdetermination of trace element partition coefficients betweengarnet clinopyroxene and hydrous basaltic liquids at 2^75 GPaand 1080^12008C Lithos 53 165^187

GriffinW L amp OrsquoReilly SY (2007) Cratonic lithospheric mantle Isanything subducted Episodes 30(1) 43^53

Griffin W L Wass S Y amp Hollis J D (1984) Ultramafic xenolithsfrom Bullenmerri and Gnotuk maars Victoria Australia petrol-ogy of a sub-continental crust^mantle transition Journal of

Petrology 25 53^87GriffinW L Doyle B J Ryan C G Pearson N J OrsquoReilly SY

Davies R Kivi K van Achterbergh E amp Natapov LM (1999)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1871

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Layered mantle lithosphere in the Lac de Gras Area SlaveCraton composition structure and origin Journal of Petrology 40705^727

Griffin W L Pearson N J Belousova E Jackson S E vanAchterbergh E OrsquoReilly SY amp Shee S R (2000)The Hf isotopecomposition of cratonic mantle LAM-MC-ICPMS analysis ofzircon megacrysts in kimberlites Geochimica et Cosmochimica Acta

64 133^147Griffin W L OrsquoReilly S Y Doyle B J Pearson N J

Coopersmith H Kivi K Malkovets V amp Pokhilenko N(2004) Lithosphere mapping beneath the North American plateLithos 77 873^922

Grulaquo tter H S Apter D B amp Kong J (1999) Crust^mantle couplingevidence from mantle-derived xenocrystic garnets In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 307^312

Hart S R Blusztajn J Dick H J B Meyer P S ampMuehlenbachs K (1999) The fingerprint of seawater circulationin a 500-meter section of ocean crust gabbros Geochimica et

Cosmochimica Acta 63 4059^4080Harte B amp Kirkley M B (1997) Partitioning of trace elements

between clinopyroxene and garnet Data from mantle eclogitesChemical Geology 136 1^24

Heaman L M Creaser R A amp Cookenboo H O (2002) Extremeenrichment of high field strength elements in Jericho eclogitexenoliths A cryptic record of Paleoproterozoic subduction partialmelting and metasomatism beneath the Slave craton CanadaGeology 30 507^510

Heaman L M Kjarsgaard B A amp Creaser R A (2004) The tem-poral evolution of North American kimberlites Lithos 76 377^397

Heaman L M Creaser R A Cookenboo H O amp Chacko T(2006) Multi-stage modification of the northern Slave mantlelithosphere Evidence from zircon- and diamond-bearing eclogitexenoliths entrained in Jericho Kimberlite Journal of Petrology 47821^858

Hermann J Cpandler C Hack A amp Korsakov A (2006) Aqueousfluids and hydrous melts in high-pressure and ultra-high pressurerocks implications for element transfer in subduction zones Lithos92 399^417

Hills DV amp Haggerty S E (1989) Petrochemistry of eclogites fromthe Koidu Kimberlite Complex Sierra Leone Contributions to

Mineralogy and Petrology 103 397^422Hoffman P F (1989) Precambrian geology and tectonic history of

North America In Bally A amp Palmer A (eds) The Geology of

North AmericaccedilAn Overview Boulder CO Geological Society ofAmerica pp 447^512

IrelandT R Rudnick R L amp Spetsius Z (1994) Trace-elements indiamond inclusions from eclogites reveal link to Archean granitesEarth and Planetary Science Letters 128 199^213

Irvine G J Pearson D G Kjarsgaard B A Carlson R WKopylova M G amp Dreibus G (2003) Evolution of the litho-spheric mantle beneath Northern Canada a Re^Os isotope andPGE study of kimberlite-derived peridotite xenoliths fromSomerset Island and a comparison to the Slave and Kaapvaalcratons Lithos 71 461^488

Isachsen C E amp Bowring S A (1994) Evolution of the Slave cratonGeology 22 917^920

Jacob D E (2004) Nature and origin of eclogite xenoliths from kim-berlites Lithos 77 295^316

Jacob D E amp Foley S F (1999) Evidence for Archean ocean crustwith low high field strength element signature from diamondifer-ous eclogite xenoliths Lithos 48 317^336

Jacob D Jagoutz E Lowry D Mattey D amp Kudrjavtseva G(1994) Diamondiferous eclogites from Siberiaccedilremnants ofArchean oceanic-crust Geochimica et Cosmochimica Acta 585191^5207

Jacob D E Fung A Jagoutz E amp Pearson D G (2003) Petrologyand geochemistry of eclogite xenoliths from the Ekati kimberlitesarea Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA_239

Jacob D E Bizimis M amp Salters V J M (2005) Lu^Hf and geo-chemical systematics of recycled ancient oceanic crust evidencefrom Roberts Victor eclogites Contributions to Mineralogy and

Petrology 148 707^720Jerde E A Taylor L A Crozaz G Sobolev NV amp SobolevV N

(1993) Diamondiferous eclogites from Yakutia Siberiaccedilevidencefor a diversity of protoliths Contributions to Mineralogy and Petrology

114 189^202Jones A G Ferguson I J Chave A D Evans R L amp

McNeice GW (2001) The electric lithosphere of the Slave cratonGeology 29 423^426

Kelemen P B Shimizu N amp Dunn T (1993) Relative depletionof niobium in some arc magmas and the continental-crust-partitioning of K Nb La and Ce during meltrockreaction in the upper-mantle Earth and Planetary Science Letters 120111^134

Kinzler R J (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition Application to mid-ocean ridge basalt petrogenesis Journal of Geophysical Research 102853^874

Klemme S van der Laan S R Foley S F amp Guenther D (1995)Experimentally determined trace and minor element partitioningbetween clinopyroxene and carbonatite melt under upper mantleconditions Earth and Planetary Science Letters 133 439^448

Konzett J Armstrong R A Sweeney R J amp CompstonW (1998)The timing of MARID metasomatism in the Kaapvaal mantle anion probe study of zircons from MARID xenoliths Earth and

Planetary Science Letters 160 133^145Kopylova M G Russell J K amp Cookenboo H (1999) Mapping the

lithopshere beneath the north central Slave Craton In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 468^479

Krogh E (1988) The garnet^clinopyroxene iron^magnesiumgeothermometerccedila reinterpretation of existing experimentaldata Contributions to Mineralogy and Petrology 99 44^48

Kusky T M (1989) Accretion of Archean Slave Province Geology 1763^67

LeCheminant A N amp Heaman L M (1989) Mackenzie igneousevents Canada Middle Proterozoic hotspot magmatismassociated with ocean opening Earth and Planetary Science Letters

96 38^48LeCheminant A N Richardson D G DiLabio R N W amp

Richardson K A (1996) La recherche de diamants au CanadaGeological Survey of Canada Open File 3228 268pp

MacGregor I D amp Carter J L (1970) The chemistry of clinopyrox-enes and garnets of eclogite and peridotite xenoliths from theRoberts Victor mine South Africa Physics of the Earth and PlanetaryInteriors 3 391^397

MacGregor I D amp Manton W I (1986) Roberts Victor eclogitesancient oceanic crust Journal of Geophysical Research 91 14063^14079

MacKenzie J M amp Canil D (1999) Composition and thermal evo-lution of cratonic mantle beneath the central Archean SlaveProvince NWT Canada Contributions to Mineralogy and Petrology

134 313^324

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1872

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

McCandlessT E amp Gurney J J (1989) Sodium in garnet and potas-sum in clinopyroxene criteria for classifying mantle eclogites InRoss J (ed) Kimberlites and Related Rocks Carlton Blackwellpp 827^832

McDonoughW F amp Sun S-S (1995) The composition of the EarthChemical Geology 120 223^253

Navon O amp Stolper E (1987) Geochemical consequences of meltpercolation the upper mantle as a chromatographic columnJournal of Geology 95 285^307

Nikogosian I K amp Sobolev AV (1997) Ion-microprobe analysis ofmelt inclusions in olivine experience in estimating the olivine^meltpartition coefficients of trace elements Geochemistry International 35119^126

Norman M D Griffin W L Pearson N J Garcia M O ampOrsquoReilly S Y (1998) Quantitative analysis of trace element abun-dances in glasses and minerals a comparison of laser ablationinductively coupled plasma mass spectrometry solution inductivelycoupled plasma mass spectrometry proton microprobe and elec-tron microprobe data Journal of Analytical Atomic Spectrometry 13477^482

Norman M D Garcia M O amp Pietruszka A J (2005) Trace-element distribution coefficients for pyroxenes plagioclase andolivine in evolved tholeiites from the 1955 eruption of KilaueaVolcano Hawairsquoi and petrogenesis of differentiated rift-zonelavas American Mineralogist 90 888^899

OrsquoHara M J amp Herzberg C (2002) Interpretation of trace elementand isotope features of basalts Relevance of field relations petrol-ogy major element data phase equilibria and magma chambermodeling in basalt petrogenesis Geochimica et Cosmochimica Acta 662167^2191

PadghamW A amp FysonW K (1992) The Slave Province a distinctcraton CanadianJournal of Earth Sciences 29 2072^2086

Pearson D G Davies G R amp Nixon P H (1993) Geochemical con-straints on the petrogenesis of diamond facies pyroxenites from theBeni Bousera peridotite massif North Morocco Journal of Petrology34 125^172

Pearson N J Griffin W L Doyle B J OrsquoReilly S Y vanAchterbergh E amp Kivi K (1999) Xenoliths from kimberlitepipes of the Lac de gras area Slave Craton Canada In GurneyJ J Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite ConferenceCapeTown Red Roof Design pp 644^658

Perfit N R Gust D A Bence A E Arculus R J amp Taylor S R(1980) Chemical characteristics of island-arc basalts implicationsfor mantle sources Chemical Geology 30 227^256

Polat A Hofmann A W amp Rosing M T (2002) Boninite-likevolcanic rocks in the 37^38 Ga Isua greenstone belt WestGreenland geochemical evidence for intra-oceanic subductionzone processes in the early Earth Chemical Geology 184 231^254

Poudjom DjomaniY H GriffinW L OrsquoReilly SY amp Doyle B J(2005) Lithospheric domains and controls on kimberlite emplace-ment Slave Province Canada evidence from elastic thickness andupper mantle composition Geochemistry Geophysics Geosystems 6doi1010292005GC0009978

Presnall D C Gudfinnsson G H amp Walter M J (2002)Generation of mid-ocean ridge basalts at pressures from 1 to 7GPa Geochimica et Cosmochimica Acta 66 2073^2090

Schmidberger S S Heaman L M Simonetti A Creaser R A ampCookenboo H O (2005) Formation of Paleoproterozoiceclogitic mantle Slave Province (Canada) Insights from in-situHf and U^Pb isotopic analyses of mantle zircons Earth and

Planetary Science Letters 240 621^633

Schmidberger S S Simonetti A Heaman L M Creaser R A ampWhiteford S (2007) Lu^Hf in-situ Sr and Pb isotope and traceelement systematics for mantle eclogites from the Diavik diamondmine Evidence for Paleoproterozoic subduction beneath the Slavecraton Canada Earth and Planetary Science Letters 254 55^68

Schulze D J amp Helmstaedt H (1988) Coesite-sanidine eclogitesfrom kimberlite products of mantle fractionation or subductionJournal of Geology 96 435^443

Smithies R H Champion D C amp Sun S S (2004) The case forArchaean boninites Contributions to Mineralogy and Petrology 147705^721

Smyth J R Caporuscio F A amp McCormick T C (1989) Mantleeclogites evidence of igneous fractionation in the mantle Earth andPlanetary Science Letters 93 133^141

Sobolev NV amp Lavrentrsquoyev Y G (1971) Isomorphic sodium admix-ture in garnets formed at high pressure Contributions to Mineralogy

and Petrology 31 1^12Stalder R Foley S F Brey G P amp Horn I (1998) Mineral aqueous

fluid partitioning of trace elements at 900^12008C and 30^57GPa New experimental data for garnet clinopyroxene and rutileand implications for mantle metasomatism Geochimica et

Cosmochimica Acta 62 1781^1801Sun S-S amp McDonough W F (1989) Chemical and isotopic sys-

tematics of oceanic basalts implications for mantle compositionsand processes In Saunders A D amp Norris M J (eds)Magmatism in the Ocean Basins Geological Society London Special

Publications 42 313^345Taura H Yurimoto H Kurita K amp Sueno S (1998) Pressure

dependence on partition coefficients for trace elements between oli-vine and the coexisting melts Physics and Chemistry of Minerals 25469^484

Taylor L A amp Neal C R (1989) Eclogites with oceanic crustal andmantle signatures from the Bellsbank kimberlite South AfricaPart I Mineralogy petrography and whole rock chemistryJournal of Geology 97 551^567

van Achterbergh E Ryan C G amp Griffin W L (1999) Glitter online intensity reduction for the laser ablation inductively coupledplasma spectrometry Abstract 9th V M GoldschmidtConference LPI Boston MA 905 pp Abstract no 7215

van Breemen O DavisW J amp King J E (1992) Temporal distribu-tion of granitoid plutonic rocks in the Archean Slave ProvinceNorthwest Canadian Shield Canadian Journal of Earth Sciences 292186^2199

Vervoort J D Patchett P J Blichert-Toft J amp Albare de F (1999)Relationship between Lu^Hf and Sm^Nd isotopic systems in theglobal sedimentary system Earth and Planetary Science Letters 16879^99

Viljoen K S Smith C B amp Sharp Z D (1996) Stable and radio-genic isotope study of eclogite xenoliths from the Orapa kimberliteBotswana Chemical Geology 131 235^255

Walter M J (1998) Melting of garnet peridotite and theorigin of komatiite and depleted lithosphere Journal of Petrology 3929^60

Wasserburg G J Jacobsen S B DePaolo D J McCulloch M Tamp Wen T (1981) Precise determination of SmNd ratios Sm andNd isotopic abundances in standard solutions Geochimica et

Cosmochimica Acta 45 2311^2323Westerlund K J Shirey S B Richardson S H Carlson R W

Gurney J J amp Harris JW (2006) A subduction wedge origin forPaleoarchean peridotitic diamonds and harzburgites from thePanda kimberlite Slave craton evidence from Re^Os isotope sys-tematics Contributions to Mineralogy and Petrology 152 275^294

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1873

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Page 5: Origins of Xenolithic Eclogites and Pyroxenites from the Central

Some xenoliths have subhedral or rounded garnets in alsquomatrixrsquo of interstitial cpx (Group I of MacGregor ampCarter 1970 Fig 2a) whereas in others garnet and cpxhave straight grain boundaries and an interlocking fabric(Group II Fig 2b) The classification of McCandless ampGurney (1989) builds on work of MacGregor amp Carter(1970) and distinguishes group I eclogites by their higherNa2O in garnet (009wt ) and K2O in cpx (008wt) from group II eclogites The classification used inTable 1 (GpMampG) is that based on Na2O contents ingarnet A different classification (GpTampN) places eclogitesinto groups A B and C distinguished by the MgO FeOand CaO contents of garnets and Na2O and MgO contentsof clinopyroxenes (Coleman et al 1965 Taylor amp Neal

1989) These classifications are based on distinct eclogitesuites from Southern African kimberlites and do not neces-sarily reflect the particularities of eclogites from otherlocalities We therefore use a slightly different scheme(lsquotypesrsquo in Table 1) that reflects the specific groupings ofeclogites in this study with regard to Cr2O3^CaO CaO(MgOthornFeO)^MgOFeO and CaO^Na2O relationshipsin garnet (see below)Garnet modes in volatile-free eclogites range from 40

to 90 vol and cpx modes from 10 to 55 with anaverage and median garnet mode of 60 and 65respectively Two samples have high modal amounts ofphlogopite (44 and 48 respectively) in addition togarnet and cpx and are therefore not eclogites sensu stricto

Table 1 Continued

Sample no Ref1 Type2 GpMampG3 GpTampN

4 Size5 Microstructure6 Assemblagemode7

gt cpx rut sf ky opx other

vr19673ecl-g1 2 na II B 51 na x x x

vr19673ecl-g2 2 na II B 51 na x x x x

vr19674ecl-g1 2 na II A 51 na x x x

vr19674ecl-g2 2 na II A 51 na x x x

vr19674ecl-g3 2 na II A 51 na x x

vr19674per-g1 2 na II A 51 na x x

vr19674per-g4 2 na II A 51 na x x x

vr40302 2 na II A 51 na x x x

vr40336 3 na II A 51 na x x x

vr40337 2 na I A 51 na x x

vr40374 2 na II A 51 na x x x

vr40382 2 na II A 51 na x x x

vr40384 ecl-g4 2 na II A 51 na x x x

vr40399 2 na II B 51 na x x x

vr50856 2 na II A 51 na x x x

vr50887 2 na II A 51 na x x

vr50888 2 na II A 51 na x x

vr50899 2 na II A 51 na x x x

vr50900 2 na II A 51 na x x x

vr50925 2 na II B 51 na x x

yk1914 1 na II A 25 medium-grano 22 78

yk1915 1 na II A 15 coarse-grano x x x x

yk1952 1 na II A 2 medium-grano x x x x

1References 1 this study 2 Pearson et al (1999) 3 Pearson et al (unpublished data) trace-element analysesthis study2Classification based on chemical parameters (see text for details)3Classification of McCandless amp Gurney (1989) based on MacGregor amp Carter (1970)4Classification of Taylor amp Neal (1989) based on Coleman et al (1965)5Longest dimension (cm) across xenolith6Only given for samples41 cm in size grano granoblastic7Modes (vol ) are given where available otherwise presence of mineral is indicated as identified in thin section or stubgt garnet cpx clinopyroxene ky kyanite rut rutile sulf sulfide phlog phlogopite ap apatite dia diamond graphgraphite cc calcite ilm ilmenite pl plagioclase qz quartz opx orthopyroxene na not applicable or not available

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1847

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(phlogopite occurring as small secondary grains at grainboundaries and in veins is not considered in this study)Ten samples of all eclogite types show secondary spongyrims of cpx around cpx cores Secondary rims containsmall (mm-scale) patches of glassRutile is the most common primary accessory phase

(15 of 65 eclogites and pyroxenites) with modes between51 and 2 vol Sulfide occurs in 14 samples and ispresent as rounded to subhedral to irregular grains

Kyanite has been identified in three samples studied byPearson et al (1999) One of the kyanite-bearing eclogitesalso contains diamond (VR40345) Kyanite as well asdiamond or graphite associated with kyanite-bearing andcompositionally similar eclogites (high-Ca eclogites) hasbeen frequently observed in drill core (Pearson et al1999) but because these eclogite types are often alteredand friable they are not proportionally represented in thepresent study One high-Ca eclogite (VR43465) contains

Fig 2 Eclogites in thin section (plane-polarized light) illustrating the microstructures mentioned in the text (a) high-Mg eclogiteVR67360(b) high-Mg eclogiteYK1946 (c) phlogopitized eclogiteYK1911 (d) garnet pyroxeniteYK1915 (e) garnet pyroxeniteYK1914 (f) garnet pyrox-enite schlieren in a matrix of garnet spongy opx and glass (YK1952) Grt garnet cpx clinopyroxene rut rutile phlog phlogopite Scalebar in(c) applies to all sections (1cm)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1848

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

3 vol graphite which occurs as disseminated plates sev-eral millimetres long with prismatic tabular or irregularhabit Quartz has been retrieved from a mineral separateof low-Mg eclogite YK1943 The phlogopitized samples(VR19674ecl-g7 and YK1911) also contain apatite andglass as accessory phases Vesicles in sample VR67112bare partially filled with single or multiple large carbonatecrystalsMost pyroxenites from Lac de Gras have fine- to coarse-

grained granoblastic equilibrated microstructures(Pearson et al 1999) Modal information for pyroxenitesin the present study is available for only the larger sampleYK1914 (22 garnet 78 cpx) (Fig 2e)

MAJOR ELEMENTSAverage major-element compositions of garnet cpx andopx in the different eclogite types (except for the threevolatile-rich eclogites which are not averaged) and in pyr-oxenites are given in Table 2 the full datasets includingaccessory minerals are available as ElectronicAppendices 2^5 which can be downloaded from httppetrologyoxfordjournalsorg

GarnetGarnets have XMg [pyrope component 100Mg(MgthornFethornCathornMn)] ranging from 178 to 756 andXCa (grossular component mol) ranging from 72 to359 (Electronic Appendix 2) In a diagram of CaO(MgOthornFeO) vs MgOFeO (Fig 3) eclogitic garnetsfall into three groups (1) low-Mg garnets with low

CaO(MgOthornFeO) and low MgOFeO (2) high-Mggarnets with low CaO(MgOthornFeO) and high MgOFeO (3) high-Ca those with high CaO(MgOthornFeO)and intermediate MgOFeO Volatile-rich eclogites havegarnets with the lowest MgOFeO whereas pyroxenitestrend towards higher values than eclogitesMost eclogite garnets have Cr2O3 contents502wt

with variable CaO contents (Fig 4) Garnets in pyroxeniteshave low CaO contents which correlate positively withCr2O3 More than half of the pyroxenites are opx-free andall pyroxenites are olivine-free on thin-section scalePyroxenitic garnets show an opposing trend of CaOvs Na2O (Fig 5) compared with the eclogitic garnetsBecause the positive correlation of Cr2O3 and CaO forpyroxenitic garnet indicates buffering by both cpx andopx opx-bearing or opx-free pyroxenites will not befurther distinguished subsequently (except in the geother-mobarometry section where opx-bearing assemblagesallow simultaneous calculation of pressure andtemperature)Distinct rim compositions in most garnets are

characterized by lower CaO contents and higherMg-number [frac14 100Mg(MgthornFe)] Garnet compositionsin three samples are inhomogeneous with regard to CaOMgO and FeO (SE01 VR43479 and VR43480) withoutclear core^rim zonations similar to coexisting cpx Likerims CaO-poor compositions are characterized by higherMg-number Visibly secondary spongy cpx rims also havehigher MgO contents (see below) and MgO-rich garnetsare therefore regarded as affected by secondary changesBased on Cr^Ca Mg^Ca^Fe and Ca^Na composi-

tional relationships in garnet the following eclogitetypes are distinguished (1) eclogites with high-Ca

0

05

10

15

20

25

30

0 01 02 03 04 05 06 07 08 09

high-Ca

high-Mg+pyroxenites

low-Mg

CaO(MgO+FeO)

MgO

FeO

GARNETpyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropic Jericho massiveDiavik low-temperatureDiavik high-temperatureDiavik diamond-bearing

eclogites

Fig 3 CaO(MgOthornFeO) vs MgOFeO (wt ) in garnet Fieldsdistinguish low-Mg and high-Ca eclogites as well as high-Mg eclogitesplus pyroxenites in this study Garnets in eclogites from Jericho(Kopylova et al 1999) and Diavik (Schmidberger et al 2007) areshown for comparison Garnet inclusions in diamond from Davieset al (1999 2004)

-135 wt (high-Ca)

2

3

4

5

6

7

8

9

0 05 10 15 20 25Cr2O3

CaO

GARNET

opx-bearing

opx-free

pyroxenites

low-Mghigh-Mgvolatile-rich

Jericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

VR19673ecl-g1g2

Fig 4 Cr2O3 vs CaO (wt ) in garnet Garnets in eclogites fromJericho and Diavik as in Fig 3

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1849

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 2 Summary of major-element contents in garnet and cpx (wt )

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO Na2O Total Mg-no T(Krogh) Type

Garnet

Low-Mg ecl 3964 026 2214 011 1843 037 1050 846 009 1000 503 976

minimum 3893 008 2130 5009 1071 024 849 567 006 990 423 860

maximum 4094 036 2275 011 2093 046 1457 1062 012 1004 708 1088

1s 070 012 052 na 354 007 227 185 003 05 99 98

High-Mg ecl 4081 026 2261 013 1419 030 1537 624 007 1000 659 969

minimum 3970 008 2224 5009 951 023 1127 418 003 991 551 820

maximum 4250 052 2323 035 1795 036 2030 882 013 1012 792 1241

1s 077 013 031 na 225 004 189 141 003 05 57 88

High-Ca ecl 3979 034 2211 na 1516 027 958 1240 011 998 528 1078

minimum 3909 011 2117 5009 1164 023 801 1144 007 989 456 910

maximum 4039 073 2242 011 1713 035 1249 1351 018 1004 657 1246

1s 042 026 042 na 173 004 155 073 004 06 67 151

vr19674ecl-g7 3867 003 2187 5009 2380 153 793 627 5004 1002 373 volatile-rich

vr67112b 3783 013 2097 009 2620 204 466 790 5004 999 241 volatile-rich

yk1911 3874 5006 2200 5009 2120 221 802 747 5004 997 403 volatile-rich

Pyroxenites 4163 021 2257 075 1148 034 1906 407 006 1002 747 909

minimum 3989 005 2081 009 729 025 1535 277 004 996 593 699

maximum 4279 048 2330 260 1881 054 2215 566 009 1011 842 1158

1s 068 010 057 057 323 007 209 065 001 04 73 117

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO Na2O Total Mg-no Type

Cpx

Low-Mg ecl 5453 028 739 na 500 5009 1117 1662 434 994 801

minimum 5160 017 564 5009 337 na 876 1388 197 981 741

maximum 5570 044 1067 014 687 na 1324 1926 632 1001 865

1s 128 011 155 na 124 na 153 177 126 06 40

High-Mg ecl 5483 023 483 na 381 na 1368 1875 318 994 865

minimum 5353 009 263 5009 245 5009 1156 1617 204 986 778

maximum 5645 031 810 016 674 011 1588 2101 495 1011 920

1s 071 007 137 na 095 na 104 138 075 06 30

High-Ca ecl 5563 027 1184 na 318 5009 868 1368 615 995 836

minimum 5448 014 769 5009 168 na 619 1060 391 987 751

maximum 5670 049 1579 010 603 na 1148 1689 807 1003 878

1s 076 015 334 na 151 na 189 261 167 06 45

vr19674 ecl-g7 5322 010 185 5009 536 015 1448 2339 062 992 828 volatile-rich

vr67112b 5040 041 328 5009 1079 024 1108 2229 071 992 647 volatile-rich

yk1911 5231 014 301 5009 544 024 1450 2340 056 996 826 volatile-rich

Pyroxenites 5527 na 347 038 304 na 1543 1959 245 999 899

minimum 5299 5006 037 010 149 5009 1128 1408 030 990 804

maximum 5666 037 786 134 490 014 1863 2312 537 1009 947

1s 068 na 189 028 091 na 193 226 131 05 37

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO NiO Total Mg-no

Opx

Pyroxenites 5781 na 059 na 627 na 3508 037 011 1006 909

minimum 5546 5006 032 5009 462 5009 2999 020 008 996 804

maximum 5888 009 318 054 1304 014 3628 077 014 1014 933

1s 098 na 067 na 212 na 160 019 002 06 32

Averaged for the three eclogite types and for pyroxenites minimum and maximum values as well as standard deviationsalso given Concentrations in volatile-rich eclogites are shown individually ecl eclogite Mg-numberfrac14 100Mg(Mgthorn Fe)Temperatures of last equilibration [T(Krogh)] as in Electronic Appendix 8

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1850

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

garnets (these include kyanite- graphite- and diamond-bearing varieties) (2) eclogites with low-Mg garnets (theseinclude quartz-bearing eclogite) (3) eclogites with high-Mggarnets and Cr2O3 contents502wt (4) volatile-richeclogites (phlogopite- apatite- carbonate-bearing)

Clinopyroxene (cpx)Clinopyroxenes in all eclogite groups have higher averageAl2O3 contents than those in pyroxenites with the formerhaving about 90 of the total Al in the jadeite molecule(AlVI) versus about 35 for the latter corresponding toan average of 41 and 25wt Na2O respectivelyThe highest Al2O3 and Na2O contents in the dataset areobserved in cpx from the high-Ca eclogites (up to158wt Al2O3) (Electronic Appendix 3) Contents ofK2O are generally below the detection limit(5004wt ) but values up to 024wt are observed insome cpx in high-Ca eclogites Average cpx Mg-numberare highest in pyroxenites (899) followed by cpx inhigh-Mg eclogites (865) high-Ca eclogites (836) andlow-Mg eclogites (801) Pyroxenitic cpx tend to havelower TiO2 and higher NiO and distinctly higherCr2O3 contents compared with those in eclogites A plotof MgO vs Al2O3 shows a negative correlation (Fig 6)Clinopyroxenes in low-Mg eclogites have lower Al2O3

contents at a given MgO content than other eclogitetypes Between 15 and 17wt MgO the Al2O3^MgOcorrelation for cpx in pyroxenites has a different slopefrom that in the eclogites Clinopyroxenes in pyroxenites

VR19673ecl-g1 and VR19673ecl-g2 have distinctly lowerMgO contents and higher Al2O3 contents than inthe other pyroxenites They coexist with garnets havingthe lowest Cr2O3 and CaO contents of the dataset whichdefine one end of the positive Cr2O3^CaO correlationobserved for garnetsClinopyroxenes in volatile-rich eclogites have distinctly

high MnO contents (up to 024wt ) Mg-numbers forcpx in the phlogopite-bearing variety are 826 and 828whereas that in the calcite-bearing eclogite (VR67112b)has an Mg-number of 647Some cpx grains have spongy rims that contain small

(mm-scale) patches of glass these rims are richer in Caand Mg and markedly poorer in Na than the cores similarto cpx in eclogites from southern Africa (Taylor amp Neal1989) The rim texture is unequilibrated suggesting lategrowth The formation of the spongy cpx rims might berelated to incongruent melting of cpx or be a result ofmetasomatism (Taylor amp Neal 1989 and referencestherein) Distinct although not spongy rims in two eclo-gites (VR67360 and YK1946) are similarly enriched inCaO and depleted in Na2O Two types of cpx with respectto CaO Na2O and Mg are present in VR43477VR43479 and VR43480 without clear core^rim zonationor spongy rims No significant differences are recognizedwith respect to other oxides

Orthopyroxene (opx)Orthopyroxenes have Mg-number ranging from 804 to933 (17 of 19 samples have values between 895 and 933)and typically contain minor amounts of Al2O3

(032^118wt one outlier has 318wt ) CaO(020^118 wt ) and NiO (008^014wt ) (Electronic

0

2

4

6

8

10

12

14

7 9 11 13 15 17 19

Al 2

O3

CPX

pyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

cpx in pyroxenitesVR19673ecl-g1g2

MgO

Fig 6 MgO vs Al2O3 (wt ) in cpx Data sources as in Figs 3and 4

0

005

010

015

020

2 4 6 8 10 12 14 16

CaO

Na 2

O

dl

group Igroup II

pyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

GARNET

Fig 5 CaO vs Na2O (wt ) in garnet Arrows show trends foreclogites and pyroxenites respectively [note that part of thepyroxenite trend extends into the field falling below the detectionlimit (dl) for Na2O] Group I and group II refer to classificationof McCandless amp Gurney (1989) Data sources as in Figs 3 and 4

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1851

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Appendix 4) The contents of all other elements are mostlybelow their respective detection limits

Accessory phasesRutile and ilmenite occur in some of the eclogites (Table 1)All rutile grains have exsolved 10 ilmenite as lamellaeand rims Ilmenite-free areas of rutile in eclogites containbetween 962 and 988wt TiO2 and minor amounts ofAl2O3 (01^04wt ) Cr2O3 (5009 to 02wt ) andFeO (03^07wt ) Rutile in pyroxenite YK1915 contains004wt Al2O3 01wt FeO and 15wt Cr2O3Discrete ilmenite grains in sample VR67112b haveTiO2 contents of 499 and 500wt MgO of 23 and15wt and FeO of 399 and 403wt respectivelySulfides occur interstitially or enclosed in cpx garnet

and opx they are mostly pyrrhotite and monosulfide solidsolution with variable Ni contents (unpublished data)They will not be considered further hereAbundant platy phlogopite is observed in two samples

(VR19674ecl-g7 and YK1911) It is Cl-rich in sampleVR19674ecl-g7 (063wt vs 003wt in YK1911) andF-rich in sample YK1911 (097wt vs 005wt inVR19674ecl-g7) (Electronic Appendix 5) The same is truefor coexisting apatite (VR19674ecl-g7 254wt Cl and014wt F YK1911 006wt Cl and 306wt F)

Comparison with mineral inclusions indiamonds from Lac de Gras and witheclogite xenoliths from other Slave CratonlocalitiesMineral inclusions in diamonds from the Lac de Graskimberlites have been investigated by Davies et al (19992004) For garnets there is a striking similarity betweenthe inclusions in diamond and those in the high-Ca eclogitexenoliths although the inclusions in diamond trendtowards higher Na2O (Fig 5) and lower Al2O3 contentsThe overlap is not as marked for cpx where inclusionsin diamond span almost the entire range observed foreclogitic cpx (Fig 6)A comparison with eclogites from other localities in

the Slave Craton shows that garnets in most Diavik low-temperature eclogites reported by Schmidberger et al(2007) plot with garnets in high-Mg eclogites and inpyroxenites from this study (Figs 3^5) whereas garnets inhigh-temperature and diamond-bearing eclogites(Schmidberger et al 2007) overlap most with garnets inhigh-Ca eclogites from this study and with garnetsincluded in diamond Eclogites from the Jericho kimberlitehave either massive or anisotropic fabrics (Kopylova et al1999) Garnets in almost all massive eclogites plot withgarnets in low-Mg eclogites from this study with regard toCaO^MgO^FeO relationships (Fig 3) but many havelower Na2O contents at a given CaO content (Fig 5)Garnets in anisotropic eclogites are restricted to lowMgOFeO (507) and show some overlap with garnets

in high-Ca eclogites although trending towards lowerCaO(MgOthornFeO) (Fig 3) while coexisting cpx is alsorestricted to low MgO and trends toward high Al2O3 con-tents (Fig 6)

TRACE ELEMENTSGarnetTrace-element abundances in garnet are summarized inTable 3 and the complete dataset is reported in ElectronicAppendix 1 Garnets in low-Mg eclogites show steep posi-tive slopes between LaN and SmN and shallower positiveslopes in the normalized heavy rare earth elements(HREEN Fig 7a) The normalized light REE (LREEN)show an order of magnitude variability ZrHf is mostlysupra-chondritic and Ti can be enriched or depleted rela-tive to elements of similar compatibility Niobium in threesamples is strongly depleted at 002^003 chondriticGarnets in high-Mg eclogites also show smooth positiveslopes in LREEN but have flat HREEN and variablenormalized middle REE (MREEN) to HREEN (Fig 7b)ZrHf is always supra-chondritic and Ti enriched ordepleted relative to similarly compatible elementsNiobium is most samples is below detection Small positiveEu anomalies though not outside the analytical uncer-tainty for Eu relative to either Sm or Gd are observed forgarnet in some high-Mg eclogites (eg EuEu up to 15for VR67360 where Eu is the average of the chondrite-normalized Sm and Gd concentrations) Garnets in high-Ca eclogites are slightly enriched in the MREEN relative tothe HREEN or have flat MREEN and HREEN with ahump between EuN and HoN (Fig 7c) One sample(VR40345) shows strong La enrichment (greater thanchondrite) Two samples have supra-chondritic ZrHfwhereas three samples have close to chondritic ratios andone sample has a subchondritic ratio As is true for theother eclogite typesTiN is either enriched or depleted rela-tive to similarly compatible elementsPyroxenitic garnets have variable slopes in the LREEN

and flat MREEN to HREEN patterns and two samplesshow high Sc relative to neighbouring elements Garnet inone of the pyroxenites (YK1952) has a highly distinctivepattern relative to other pyroxenites with very low LREEand HFSE (Fig 7d) This sample consists of coarse pyroxe-nite schlieren in which the garnet was analysed in a finer-grained lsquomatrixrsquo of spongy cpx with secondary rims finecpx^opx intergrowths embayed garnet and melt patchesVolatile-rich eclogites have garnets that are slightly

enriched in MREEN relative to HREEN or have flatMREE to HREE patterns One garnet has a pronouncednegative Eu anomaly (EuEufrac14 07 forVR67112b Fig7e)Garnets in volatile-rich samples have the lowest Ni Coand Ti and the highest V ScYand HREE abundances inthe dataset

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1852

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 3 Summary of trace-element contents in garnet and cpx (ppm)

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR19674

ecl-g7

VR67112b YK1911 av min max 1s

n 6 6 7 3 5 4 4

P 169 54 270 88 196 70 310 78 282 80 450 138 160 60 80 214 48 370 144

Sc 62 40 89 16 48 36 67 11 53 42 79 15 69 92 93 186 52 440 179

Ti 1667 490 2500 680 1073 620 2000 455 2141 700 4700 1582 270 690 260 1097 540 1900 573

V 169 78 250 68 150 78 280 72 178 110 390 101 140 340 100 390 71 820 326

Co 59 42 70 11 58 39 84 13 58 47 65 5 30 41 27 68 55 90 16

Ni 17 5 32 10 21 11 34 10 24 11 35 8 6 10 51 54 25 110 38

Ga 12 7 15 3 13 8 16 3 12 9 20 4 8 86 58 12 10 14 2

Sr na 512 4 na na 522 522 na na 57 57 na 52 5025 510 513 513 513 na

Y 37 6 61 18 21 7 38 10 31 20 38 5 60 81 70 19 12 26 6

Zr 15 5 29 9 28 6 83 26 24 9 43 12 60 22 90 18 1 41 17

Nb 009 0005 048 019 na 50015 05 na na 5005 05 na 5003 0012 5001 022 0030 057 025

La 005 001 009 004 na 50007 00 na na 5003 04 na 004 002 0009 003 000 009 005

Ce 03 01 05 02 01 00 01 00 05 01 18 05 03 022 015 01 00 03 01

Pr 01 00 02 01 00 00 01 00 02 01 03 00 017 014 012 00 00 01 00

Nd 12 03 24 07 07 04 14 03 21 12 32 07 24 23 20 03 00 07 03

Sm 11 04 22 06 10 03 18 05 23 08 36 10 29 31 33 05 01 09 04

Eu 06 03 10 03 06 02 10 03 11 04 17 04 16 114 16 03 01 06 02

Gd 27 07 41 12 23 07 42 12 43 16 53 13 8 81 90 13 06 24 08

Dy 54 11 70 23 36 13 65 18 58 35 71 11 12 144 12 29 18 42 10

Ho 14 02 23 07 08 03 15 04 12 09 14 02 24 30 23 07 04 10 02

Er 46 07 80 24 24 09 40 10 34 25 40 05 8 85 6 22 14 28 06

Yb 52 06 98 30 24 10 37 08 32 22 39 05 8 70 5 23 13 30 07

Lu 09 01 16 05 04 02 05 01 05 03 07 01 11 09 08 04 02 05 01

Hf 03 02 05 01 04 01 10 03 05 02 09 02 09 027 007 03 00 06 03

Ta na 50008 5004 na na 50005 001 na na 50004 na na 003 50002 50005 na 50007 011 na

Pb na 5005 02 na na 5002 02 na na 5003 07 na 02 004 5002 na 5003 01 na

Th na 5001 005 na na 50007 003 na na 50008 011 na 5007 0012 50004 002 0005 006 002

U na 5003 5003 na na 5001 01 na na 5007 na na 01 0023 010 na 5001 003 na

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR67112b YK1911 av min max 1s

n 6 4 8 5 4 4

P 65 15 120 41 39 14 56 16 53 30 70 12 20 9 80 27 120 40

Sc 22 10 30 7 15 14 19 2 14 7 24 5 80 43 17 11 21 5

Ti 1557 990 2400 508 1168 1010 1490 191 1988 810 3500 1154 2390 1700 689 102 1685 694

V 415 180 480 107 318 210 400 69 324 259 450 72 600 180 250 139 480 156

Co 35 26 50 8 20 14 31 7 26 14 42 11 49 23 29 20 42 10

Ni 164 87 220 47 228 100 430 128 188 135 320 56 59 34 428 410 670 184

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1853

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

ClinopyroxeneTrace-element abundances in cpx are summarized inTable 3 and the complete dataset is reported in ElectronicAppendix 6 Clinopyroxenes in low-Mg eclogites havemildly positive to negative slopes from LaN to NdN andsmooth negative slopes between SmN and HoN thatmostly continue to LuN (Fig 8a) Niobium is below detec-tion or 5007 chondritic and SrN is variably enrichedrelative to neighbouring elements with the peak size corre-lating negatively with LREEN ZrHf is subchondritic andHfN is enriched relative to SmN whereasTiN is enriched ordepleted relative to similarly compatible elementsWe ana-lysed the spongy rim in cpxYK1943 and its trace-elementpattern is similar to that of cpx cores for more compatibleelements whereas the most incompatible elements areseveral orders of magnitude higher Clinopyroxenes in

high-Mg eclogites have steeper negative slopes betweenNdN and LuN than those in low-Mg eclogites combinedwith lower HREE abundances HfN is depleted relative toSmN (Fig 8b) High-Ca eclogites contain cpx with sinusoi-dal REEN patterns with lower LREE than other eclogitetypes and the heaviest REE below detection They haveuniformly strong positive Sr peaks relative to elements ofsimilar compatibility (Fig 9) The diamond-bearing high-Ca eclogite (VR40345) has cpx with unusually high Nband LREE abundancesClinopyroxenes in four pyroxenites have disparate trace

element patterns They all share depletions in TiN relativeto similarly compatible elements and two show positiveSrN peaks Two samples have cpx with high and flatLREEN to MREEN patterns and steep negative slopes inthe HREEN The trace-element patterns of cpx in YK1914

Table 3 Continued

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR67112b YK1911 av min max 1s

n 6 4 8 5 4 4

Ga 16 10 20 3 12 9 15 2 26 15 39 9 129 123 6 11 11 5

Rb na 5008 04 na na 5007 05 02 na 5003 03 na 02 5004 03 010 05 0

Sr 166 85 290 74 222 153 285 50 243 83 440 142 85 120 200 11 414 209

Y 2 04 4 10 13 07 2 04 11 03 3 10 28 51 2 03 3 2

Zr 19 4 48 15 25 8 45 13 17 8 28 7 111 129 36 008 81 42

Nb 01 0003 03 01 na 5002 00 na na 5002 16 9 007 011 010 003 02 008

Cs na 5001 6 na na 5006 5006 na5013 5004 5013 na bdl 5002 na 5001 5002 na

Ba 09 02 2 07 09 01 2 07 10 03 3 10 06 01 13 06 3 10

La 3 03 8 3 3 23 3 04 4 01 14 6 51 97 4 01 10 4

Ce 7 13 20 6 10 7 13 2 7 05 24 10 25 28 11 01 28 13

Pr 11 02 3 08 2 10 3 06 08 02 2 09 47 46 2 002 4 2

Nd 5 12 12 4 10 5 14 4 4 09 8 26 23 22 8 01 22 10

Sm 11 05 21 06 20 10 31 10 07 02 11 03 47 53 2 005 5 2

Eu 03 01 05 01 06 028 09 03 02 010 03 007 087 123 06 002 15 07

Gd 08 03 13 03 11 05 19 06 05 03 06 011 28 37 15 010 36 2

Dy 05 01 07 02 04 022 06 01 02 013 03 007 11 16 07 008 14 06

Ho 01 00 02 004 006 003 007 002 na 5003 012 na 011 021 009 001 02 007

Er na 507 507 na 011 004 02 00 na 5006 03 na 019 033 02 002 03 012

Yb na 5001 02 na na 5001 011 na na 5006 03 na 007 015 009 001 02 006

Lu na 5001 003 na na50002 002 na na50008 01 na 0006 0013 0013 0002 003 001

Hf 12 03 24 07 11 04 18 05 10 06 16 04 34 53 na50015 35 na

Ta na50003 50015 na na50006 006 na na50004 5004 na 0014 00236 na50003 503 na

Pb 06 02 12 04 06 04 09 02 na 5016 07 02 065 098 08 01 18 1

Th na50005 06 na 012 005 02 005 na 5002 5006 na 021 02 001 06 03

U na 5005 5005 na 002 0009 005 001 na 5003 5005 na 0013 024 na50005 003 na

Averaged for the three eclogite types and for pyroxenites as in Table 2 pyx pyroxenite n number of analyses averagedbdl below detection limit

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1854

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

and YK1952 are similar although at different abundanceswith only modest LREEN over HREEN enrichment andstrong depletions in ZrN relative to neighbouring elements(Fig 8d)Clinopyroxenes in the volatile-rich samples have flat pat-

terns and high abundances from LaN to SmN and smoothnegative slopes from SmN to LuN Niobium Sr and Ti are

noticeably depleted relative to elements of similar compat-ibility (Fig 8e)

Trace-element partitioningTo assess whether coexisting phases are in trace-elementequilibrium we calculated apparent cpx^garnet distribu-tion coefficients and compared them with those

(a)

(d)

(b)

(e)

(c)

GarnetCI-chondrite GarnetCI-chondrite

low-Mghigh-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1952

1

10

Nd Sm Eu Gd Dy

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Fig 7 Extended trace-element patterns of garnet in (a) low-Mg eclogites (b) high-Mg eclogites with inset showing enlargement of MREE tohighlight Eu anomalies and 1s uncertainties (c) high-Ca eclogites (d) pyroxenites and (e) volatile-bearing eclogites Normalized to chondritevalues of McDonough amp Sun (1995)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1855

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

determined experimentally or in demonstrably equili-brated natural samples Low-Mg and high-Mg eclogiteshave Dcpxgarnet for most elements that are within therange determined for natural equilibrated eclogites andexperimental assemblages (Harte amp Kirkley 1997 Greenet al 2000 Barth et al 2002a) (Electronic Appendix 7)High-Ca eclogites plot at the low end of the range

of distribution coefficients and have very low DMREE

and DHREE outside the range indicated by the datasetof Harte amp Kirkley (1997) who noted that DREE decreaseswith increasing Ca content in eclogitic minerals Theseassemblages are therefore considered to be in equilibriumPyroxenite VR40384 has D values similar to eclogites

but at the high end of the range determined by Harte amp

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

(a)

(d)

(b)

(e)

(c)

CpxCI-chondrite CpxCI-chondrite

low-Mg high-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1914

YK1952

spongy rim

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

VR40345

Fig 8 Extended trace-element patterns of cpx as in Fig 7

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1856

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Kirkley (1997) (Electronic Appendix 7) Another group(YK1914 and YK1952) plots at the low end but DHf DZrDTi and DV are far below the values indicated by experi-mental datasets Because pyroxenites compositionallygrade into eclogites their drastically different DHFSE

values suggest disequilibrium In contrast volatile-bearingsamples have DHFSE values that appear to be too high to bein equilibrium however this may be a consequence oftheir unusual bulk chemical composition (Ti- and K-rich)which differs substantially from that of lsquonormalrsquo eclogitesand experimental charges

GEOTHERMOBAROMETRYFollowing Schmidberger et al (2007) temperatures oflast equilibration were calculated for eclogitic andopx-free pyroxenitic assemblages using the formulationof Krogh (1988) The results are summarized in Table 2and Electronic Appendix 8 for a preset pressure of5 GPa and illustrated as a histogram in Fig 9 Foropx-bearing pyroxenites temperatures were calculatedsimultaneously with pressure [Ca-in-opx thermometryand Al-in-opx barometry using the formulations of Breyet al (1990)] to constrain their relative positions in themantle columnMinerals in some eclogites from Lac de Gras show dis-

tinct core and rim compositions or inhomogeneitiesTemperatures were not calculated for samples withspatially irresolvable inhomogeneous garnet compositions(SE01VR43479VR43480) For eclogites with spongy cpxcore compositions were used in the calculationTemperatures were calculated for both core and rim com-positions where applicable For all samples these

temperatures agree within the uncertainty of this thermo-meter (508C Brey amp Kolaquo hler 1990)Most equilibration temperatures for eclogites lie between

800 and10008C (Fig 9)The transition between the shallowand the deep lithospheric mantle layers in the central Lacde Gras area lies in this temperature interval (Griffin et al1999 Pearson et al 1999) Eclogites in this study lack thedistinct bimodal temperature distribution observed bySchmidberger et al (2007) but as in their study high-Caeclogites give higher average temperatures (10808C corre-sponding to a depth of 155 km if equilibrated along a 40mWm2 geotherm) than other eclogites types (9808Cdepth 135 km) the diamond-bearing sample (VR40345)gives thehighest temperature of the datasetHighNa2Ocon-tents in garnet from high-Ca eclogites are qualitativelyconsistent with higher equilibration pressures (Sobolev ampLavrentrsquoyev 1971) as are high K2O contents in coexistingcpx (Erlank ampKushiro1970) suggesting that high-Ca eclo-gites not only equilibrated at high temperature but also highpressure rather than representing a temperature excursionPyroxenites both opx-bearing and opx-free give somewhatlower average temperatures than eclogites (mean 9258C)and phlogopite-rich eclogites give the lowest temperatures(685 and 7698C respectively) The calcite-bearing eclogiteVR67112b gives a relatively low equilibration temperatureof 8558C

MEASURED ANDRECONSTRUCTEDWHOLE -ROCKSRationale and approachKimberlite-borne xenoliths sometimes show evidence forinfiltration of the host kimberlite (eg Taylor amp Neal1989 Barth et al 2001 2002b Heaman et al 2006) Wetherefore follow the standard approach of calculatingwhole-rock compositions from the compositions of the con-stituent phases weighted by their modal abundances Onlycore compositions were used as distinct rim compositionsmay point to late diffusion of elements from kimberlite-related melt veinsMany eclogitic whole-rock compositions reconstructed

with only garnet and cpx have negative Ti anomalies(not shown) For most of these samples rutile or ilmenitehas been identified as part of the mineral assemblage andthese phases should therefore be considered in the whole-rock reconstruction Rutile or ilmenite modes were esti-mated by assuming that the bulk-rock has no Ti anomalywhen normalized to a normal mid-ocean ridge basalt(N-MORB) composition and calculating the amount ofrutile required to erase theTi anomalyIt is of course possible that the eclogites had negative

or positive Ti anomalies relative to elements of similarcompatibility (Eu and Dy) Because we reconstructed

0

2

4

6

8

10

12volatile-rich

pyroxenite opx-bearing

pyroxenite opx-free

high-Ca

high-Mg

low-Mg

700750

800850

900950

10001050

11001150

12001250

Fre

quen

cy

Temperature

Fig 9 Histogram ofTKrogh (Krogh 1988) obtained at a preset pres-sure of 5 GPa except opx-bearing pyroxenites where TBKN wassolved simultaneously with PBKN (Brey amp Kolaquo hler 1990)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1857

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

whole-rock compositions from mineral core compositionskimberlite infiltration is unlikely to have caused fractiona-tion of Ti from Eu and Dy In a study in which rutilemodes were estimated from the differences between Ti inmeasured bulk-rocks and in bulk-rocks reconstructed withonly garnet and cpx only three of 19 eclogites recon-structed with rutile show Ti anomalies (Barth et al2002a) Even eclogite xenoliths with depleted Zr and Hfabundances that resemble those of modern island arcbasalts do not show Ti anomalies (Jacob amp Foley 1999)We therefore consider our approach to be a reasonablyrobust means of obtaining meaningful rutile modes wheresmall sample sizes preclude reliable determination of themodes of accessory phases by point-countingWhere rutilewas not recognized on thin-section scale or not analysedthe average eclogitic rutile composition was used for thereconstruction

UncertaintiesReconstructing whole-rock compositions in this way is arather crude method given the coarse grain size of theeclogites relative to the sample size and the non-uniformdistribution of the principal minerals on thin-sectionscale The uncertainty in mineral modes of medium- tocoarse-grained eclogites with grains45mm has been esti-mated by Jerde et al (1993) to be 10 plus uncertaintiesrelated to partial alteration which is minor in our samplesNevertheless the whole-rock trace-element budget israther insensitive to the exact modes for common eclogiticassemblages and REE patterns which encompass mostother elements in terms of compatibility in garnet andcpx do not change significantly even for modal variationsof 30 (Jerde et al 1993) This is not true for major-element concentrations which vary substantially with var-iations in mineral modes and we therefore do not calculatewhole-rock major-element contentsBecause the discussion that follows is based on trace-

element patterns and trends rather than absolute abun-dances whole-rock reconstruction based on mineralmodes is adequate We are confident that the uncertaintyinherent in this approach is not leading to significant mis-representation of eclogite whole-rock compositions becausethe trace-element patterns of the reconstructed whole-rocks show distinct differences (see below) that correlatewith the assignment to eclogite types based on garnetmajor-element composition and this is independent ofmineral mode Following Jerde et al (1993) in ElectronicAppendix 9 we show variations in trace-element patternfor one sample in each eclogite type for variations ingarnet and cpx abundance from 7030 to 3070 vol This shows that the trace-element patterns are generallyinsensitive to cpxgt ratios with regard to slopes andanomalies such as positive Sr Pb and Eu anomalies withthe exception of the positive Eu anomaly in high-Mgeclogite VR43479 which disappears if a high cpx mode is

assumed (not shown) Variations in rutile modes by 50(Electronic Appendix 9d) lead as would be expected tothe appearance of Ti anomalies but a rutile mode of 05instead of 1 vol does not erase the positive Nb anomalyin low-Mg eclogiteVR50909Because of the robustness of the relative trace-element

abundances in the reconstructed eclogites a garnet propor-tion of 065 and cpx proportion of 035 approximately themedian modal abundances of these minerals in volatile-free eclogites from Lac de Gras are assumed for eclogitesamples that are too small for determination of meaningfulmodes by point-counting

Within- and between-group variationsThe trace-element compositions of reconstructed whole-rocks are given in Table 4 N-MORB-normalized trace-element patterns of reconstructed whole-rocks areshown in Fig 10 The three eclogite groups have distincttrace element patterns that are particularly obvious forthe REE (Fig 11) Low-Mg eclogites have positive Pb andSr anomalies and smooth positive slopes in their MREEN

and HREEN LREEN are enriched relative to MREEN inthree samples (Fig 11a) Three samples have high Nbrelative to similarly compatible elements and comparedwith most other eclogites (Fig 10a) All high-Mg eclogitesdisplay more or less pronounced positive Eu and Sranomalies and most have positive Pb and negative Zrand P anomalies Their N-MORB-normalized REE pat-terns show shallow positive slopes from LaN to PrN orNdNand aweak negative slope from NdN to LuN (Fig11b)VR43479 is a low-Mg eclogite with regard to its garnet

Ca^Fe^Mg composition but has a trace-element patterntypical of high-Mg eclogites and will be discussed withthe latter Conversely sample YK1949 a high-Mg eclogitein terms of its garnet Ca^Fe^Mg compostion is more simi-lar to low-Mg eclogites in terms of its whole-rock trace-element pattern and is plotted in Figs 10a and 11aWhereas the patterns of most samples within the groupresemble each other the N-MORB-normalized incompati-ble-element abundances of the high-Ca eclogites vary bymore than an order of magnitude (Fig 10c) All sampleshave positive Pb anomalies and most have positive Sranomalies Some samples have negative P andor Zranomalies Niobium abundances in three of the samplesare very high similar to those for some low-Mg eclogitesHigh-Ca eclogites tend to have flatter MREEN andHREEN patterns and more uniform HREE abundancesthat the other eclogite types with humps between SmN

and HoNThe phlogopitized sample YK1911 shows a strong posi-

tive Pb anomaly and lower HREE abundances than thecalcite-bearing sample VR67112b which shows a distinctnegative Eu anomaly and a depletion of La relative to Ce(Figs 10d and 11d)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1858

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Trace-element compositions of reconstructed whole-rocks (ppm)

Sample P Sc Ti V Co Ni Ga Sr Y Zr Nb La Ce Pr Nd

Low-Mg eclogites

vr19674-g4 186 50 4649 317 57 59 144 77 24 30 36 13 31 05 24

vr43452 154 39 2253 328 49 84 176 52 32 11 01 02 07 01 08

vr43469 38 65 4377 349 54 91 142 34 28 15 02 05 15 03 17

yk1926 144 45 3406 258 52 87 145 89 18 18 20 09 29 06 33

yk1943 173 46 5732 295 41 39 146 122 25 48 82 34 85 13 64

yk19491 102 34 1693 158 31 49 100 46 10 14 03 09 21 03 17

High-Mg eclogites

vr19677-g2 138 39 5160 291 41 53 119 98 14 22 04 13 53 11 64

vr434792 46 25 1265 129 36 126 84 56 3 7 04 06 25 04 23

vr67360 49 36 1573 138 64 184 111 76 5 9 04 09 38 07 33

yk1946 153 28 4877 192 45 105 124 108 12 48 03 10 38 08 50

High-Ca eclogites

vr19674-g5 267 30 7663 200 41 67 195 32 23 43 77 01 04 02 16

vr40345 0 33 2821 175 47 98 210 179 13 13 66 60 109 11 38

vr43465 142 27 5104 220 31 83 121 197 14 35 29 05 22 05 40

vr50858 295 30 6898 201 44 94 150 48 22 38 06 01 03 02 20

vr50860 68 57 5672 418 57 71 186 109 19 11 02 02 13 03 26

vr50909 238 28 7295 201 42 92 171 48 20 57 13 01 04 02 21

yk3528 91 51 3712 341 53 155 177 49 16 9 01 02 08 02 11

Volatile-rich eclogites

vr67112b 40 84 9242 521 46 30 107 4 40 66 31 26 126 24 125

yk1911 31 39 9428 286 44 82 164 3 23 52 64 18 52 09 46

Pyroxenites

yk1914 111 49 707 207 48 547 44 31 4 3 02 06 19 03 10

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

Low-Mg eclogites

vr19674-g4 11 05 10 22 39 10 29 33 06 12 05 028 24 008

vr43452 04 02 08 13 38 12 41 49 08 05 01 nd 02 na

vr43469 11 05 09 25 44 11 34 36 06 07 02 001 09 129

yk1926 12 05 11 19 30 07 22 24 04 07 03 004 23 090

yk1943 21 08 10 29 42 10 30 32 05 16 05 013 65 017

yk19491 06 03 11 08 14 04 14 18 03 04 02 003 29 115

High-Mg eclogites

vr19677-g2 20 08 12 21 24 05 15 15 03 05 05 007 52 6811

vr434792 05 02 12 05 06 01 04 03 00 03 01 001 98 037

vr67360 06 03 14 06 09 02 06 06 01 02 02 002 86 000

yk1946 18 07 12 21 22 05 13 12 02 11 03 006 56 112

High-Ca eclogites

vr19674-g5 17 09 11 33 43 09 24 23 04 16 05 007 02 011

vr40345 09 04 10 12 22 06 16 14 02 05 nd nd 298 na

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1859

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Continued

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

vr43465 22 09 12 25 26 06 15 14 02 10 04 001 25 040

vr50858 20 08 09 33 41 08 22 21 03 10 01 nd 02 na

vr50860 14 07 11 27 37 08 22 18 03 06 00 nd 06 na

vr50909 20 10 12 33 38 07 20 20 03 16 01 002 02 181

yk3528 08 05 11 20 30 07 18 18 02 05 01 nd 08 na

Volatile-rich eclogites

vr67112b 38 10 07 53 74 15 41 34 04 18 03 012 60 004

yk1911 20 07 08 35 39 07 18 17 02 10 14 nd 73 na

Pyroxenites

yk1914 02 01 11 04 08 02 05 06 01 01 03 003 68 101

aGarnet has CandashMgndashFe relationship typical of high-Mg eclogites but whole-rock trace-element pattern is typical oflow-Mg eclogitesbGarnet has CandashMgndashFe relationship typical of low-Mg eclogites but whole-rock trace-element pattern is typical ofhigh-Mg eclogites

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

(a)

001

01

1

10(c)

001

01

10(b)

VR43452

YK1949

RockNMORB

low-Mg

RockNMORB

high-Mg

(d)

volatile-richhigh-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

phlogopite-bearingYK1911

calcite-bearingVR67112b

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 10 Extended trace-element patterns of reconstructed (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-bearing eclogites Normalized to N-MORB of Sun amp McDonough (1989)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1860

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

RockNMORB

(b)

RockNMORB

(d)

010

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu001

01

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

(a)

001

01

1

10(b)

001

01

1

10(c)

001

01

1

10(e)

high-TSchmidberger et al

low-TSchmidberger et al

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

high-Mglow-Mg

high-Cavolatile-rich

phlogopite-bearingYK1911

calcite-bearingVR67112b

YK1949VR43479

Fig 11 N-MORB normalized REE patterns for (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-richeclogites Shown for comparison in (e) are reconstructed eclogites from Schmidberger et al (2007)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1861

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Both volatile-rich whole rocks display strong negativeSr anomalies Strontium can be sited in calcite or apatiteandor indicate the presence of a Sr-rich mineral that wasnot identified in thin section such as bariteFour of the six reconstructed whole-rocks in the study

of Schmidberger et al (2007) (Fig 11e) show markedLREEHREEN enrichment that is observed only in onehigh-Ca eclogite in our study (Fig 11c) suggesting thatmany of the eclogites they studied were affected by metaso-matic LREE enrichment Two samples in their study thatare not affected by this enrichment show positive Euanomalies similar to the high-Mg eclogites in this studywhereas the distinctive REE pattern of the low-Mg eclo-gites is not recognized among their samples

SR^ND^HF I SOTOPESYSTEMATICSStrontium Nd and Hf isotope data for optically puremineral separates and reconstructed whole-rocks aregiven in Table 5 Spongy rims and alteration are likely tobe mostly removed during leaching in 6N HCl in combina-tion with repeated ultrasonication Despite this rigoroustreatment the reproducibility of Nd in garnet VR43469and in cpx YK1914 is worse than for other samples Thismay be due to inhomogeneities related to late Nd diffusionpossibly from grain boundaries containing a kimberlitecomponent The reproducibility of the 176Hf177Hf data isgenerally poorer than that of Nd which may reflect a com-bination of matrix effects caused by imperfectly purifiedsolutions and sample heterogeneity Parent^daughterratios are available only from in situ analyses which arenot precise enough to allow meaningful model age orinitial isotope ratio calculations (ie at the time of kimber-lite eruption) Because the Lac de Gras kimberlites arerelatively young (55^56 Ma Graham et al 1999) age cor-rection would result in only a minor change in isotopecomposition which is inconsequential in view of the rangeof compositions displayed by most samples and hence hasno effect on the conclusions drawnClinopyroxenes in five volatile-free eclogites and a

pyroxenite have Sr isotope ratios ranging from 0701662 to0703546 this range is similar to or somewhat lower thanthat of the depleted mantle at the time of kimberliteemplacement (0703 Bell amp Tilton 2002) Clinopyroxenein the phlogopitized sample has a distinctly high 87Sr86Srof 0751186Clinopyroxene controls 56^92 of the Nd budget (Nd

concentration in cpx weighted by mode) Neodymiumisotope ratios range from 0511660 to 0514844 Thiscorresponds to eNd variation (per 10 000 deviationof 143Nd144Nd in the sample from 143Nd144Nd in the chon-dritic mantle model reservoir) from 17 to thorn43 cpx inthe low-Mg eclogites have higher eNd than those in

high-Mg eclogites Garnets have generally have higher143Nd144Nd from 0512984 (eNdfrac14thorn 7) to 0520965(eNdfrac14thorn 162) Garnets in the two volatile-rich sampleshave the highest ratios 0522790 for garnet in the calcite-bearing sample VR67112b (eNdfrac14thorn198) and 0525721 inthe phlogopitized eclogiteYK1911 (eNdfrac14thorn 255)Garnet contributes 18^32 and cpx 48^78 of the Hf

budget The proportion controlled by rutile is 4^30(Aulbach et al in preparation) Measured 176Hf177Hf incpx ranges from 0281263 in phlogopite-bearing sampleYK1911 (eHffrac1453) to 0286130 in low-Mg eclogiteVR43469 (eHffrac14thorn119) and in garnet from 0283958 inpyroxeniteYK1915 (eHffrac14thorn42) to 0295718 in low-Mg eclo-gite YK1926 (eHffrac14thorn458) Hafnium in garnet in the phlo-gopite^apatite-bearing sample is extremely radiogenic(176Hf177Hffrac14 0354367) corresponding to an eHf of thorn2532which is close to a value determined for a garnet separatefrom a Roberts Victor eclogite (eHffrac14thorn2561 Jacob et al2005) Garnet in the calcite-bearing sample has extremelyradiogenic Nd and Hf (eNdfrac14thorn193 eHffrac14thorn2516) Garnetin pyroxenite YK1915 also has radiogenic Nd and Hf(eNdfrac14thorn27 eHffrac14thorn42)Despite the large uncertainties for reconstructed whole-

rocks (calculated as the square root of the sum of squareduncertainties of two standard errors for garnet and cpxanalyses combined with 20 propagated uncertainty onthe proportion of Nd or Hf respectively contributed byeach mineral) the isotopic data are consistent in thatthey correlate with the independently determinedparent^daughter ratios (1) Low-Mg eclogites have higherSmNd reflected in the steeper positive slope of theirN-MORB-normalized REE patterns (Fig 11a) comparedwith the high-Mg eclogites (Fig 11b) Accordingly recon-structed low-Mg eclogites have higher present-day eNd

(0 to thorn47) than high-Mg eclogites (4 and 17) (2) Low-Mg eclogite YK1943 has very high Nb and LREE abun-dances that could have been added during metasomatismthat also enhanced its Hf abundance This sample has thelowest eNd and eHf of the low-Mg eclogites (0 and thorn20respectively) In contrast apparently unenriched low-Mgeclogites have more radiogenic Nd and Hf (eNdfrac14thorn47and thorn9 eHffrac14thorn128 and thorn113 respectively)The Sr^Nd isotope diagram reveals that eNd in cpx

varies from 19 to 43 at nearly constant depleted-mantle-like 87Sr86Sr (Fig 12a) In the Nd^Hf isotope diagram(Fig 12b) the reconstructed whole-rocks lie on curvedmixing lines if the lsquomixingrsquo proportions of Nd and Hffor the two lsquoend-memberrsquo minerals are very different(eg YK1911) Garnets have more radiogenic Nd andHf than coexisting cpx All reconstructed volatile-freeeclogites plot in the radiogenic field with regard to Hfor are isotopically similar to the depleted mantle Twohigh-Mg eclogites and one low-Mg eclogite have unradio-genic Nd one high-Mg eclogite has radiogenic Nd and

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1862

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 5 Individual and averaged Sr Nd and Hf isotope ratios obtained from garnet and cpx solutions

Sample 87Sr86Sr 2SE2SD 143Nd144Nd 2SE2SD eNd 2SE2SD 176Hf177Hf 2SE2SD eHf 2SE2SD

Low-Mg eclogites

VR43469 cp 0702661 0000010 0514844 0000006 43 01 0286130 0000012 119 04

VR43469 gt 0515508 0000014 56 03 0286506 0000012 132 04

VR43469 gt r 0515453 0000010 55 02 0286548 0000011 134 04

VR43469 gt r 0515384 0000007 54 01 0287244 0000010 158 04

VR43469 gt r 0515487 0000014 56 03 0286996 0000014 149 05

VR43469avg gt 0515458 0000109 55 21 0286823 0000715 143 253

WR VR43469 0515059 0000273 47 5 0286405 0001079 128 38

YK1926 cp 0703546 0000011 0513043 0000007 8 0 0283154 0000009 13 03

YK1926 gt 0513364 0000008 14 01 0295718 0000005 458 02

WR YK1926 0513120 0000026 9 1 0285980 0000716 113 25

YK1943 cp 0512143 0000007 10 01 0282213 0000005 20 02

YK1943 gt 0514461 0000012 36 02 0287510 0000008 168 03

YK1943 gt r 0514491 0000009 36 02 0287332 0000005 161 02

YK1943 gt r 0514517 0000010 37 02 0287290 0000007 160 02

YK1943avg gt 0514489 0000056 36 11 0287377 0000233 163 82

WR YK1943 0512625 0000228 0 5 0283138 0000879 13 31

High-Mg eclogites

VR43480 gt 0520965 0000017 162 03 0284950 0000010 77 04

VR67360 cp 0283509 0000074 26 26

VR67360 cp r 0283433 0000031 23 11

VR67360avg cp 0701662 0000018 0511660 0000010 19 02 0283471 0000108 25 38

VR67360 gt 0512984 0000101 7 20 0287429 0000180 165 64

WR VR67360 0511761 0000132 17 3 0284734 0000469 69 17

YK1949 cp 0703019 0000013 0512231 0000009 8 02 0282513 0000009 9 03

YK1949 gt 0513511 0000010 17 02 0290297 0000042 266 15

YK1949 gt r 0513508 0000012 17 02 0289947 0000015 254 05

YK1949 gt r 0513552 0000014 18 03

YK1949avg gt 0513523 0000049 17 10 0291569 0000494 311 175

WR YK1949 0512446 0000103 4 2 0284622 0000847 65 30

Volatile-rich eclogites

VR67112b gt 0522790 0000008 198 02 0354367 0000036 2532 13

YK1911 cp 0751186 0000009 0512165 0000007 9 01 0281263 0000009 53 03

YK1911 gt 0525721 0000013 255 03 0288051 0000012 187 04

WR YK1911 0513940 0000342 25 7 0281414 0000018 48 1

Pyroxenites

YK1914 cp 0703015 0000034 0513283 0000041 13 08

YK1914 cp r 0702956 0000005 0513200 0000027 11 05

YK1914 cp r 0513527 0000059 17 11

YK1914 cp r 0513045 0000010 8 02

YK1914avg cp 0702986 0000083 0513264 0000402 12 78

YK1915 gt 0514003 0000011 27 02 0283958 0000021 42

Analyses of full procedural replicates are designated with an lsquorrsquo after sample name SE standard error (internal errorsingle analyses) SD standard deviation (external error averaged multiple analyses) Italics used for reconstructed whole-rocks (mineral isotopic compositions and concentrations weighted by modes) Uncertainties for reconstructed whole-rocksinclude 2SE or 2SD uncertainties on each mineralrsquos isotope composition combined with 20 uncertainty in the amountof Nd or Hf contributed by the constituent minerals eNd and eHf are per 10 000 deviation from model chondritic mantle[model of Wasserburg et al (1981) and Blichert-Toft amp Albarede (1997) respectively]

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1863

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

εHf

εNd

Ekati eclogitesminus100

0

100

200

300

400

500

minus40 0 40 80 120 160 200 240 280

gt in YK1915

gt in VR67112bεNd=198 εHf=2532

minus40

0

40

80

120

160

minus40 minus20 0 20 40 60

minus60

minus40

minus20

0

20

40

070 071 072 073 074 075 076

North Americankimberlite field

εNd

87Sr86Sr

cpx in YK1914

Ekatieclogites

DM

DM

DM

DM

(a)

(b)

pyroxenites

low-Mg

high-Mg

volatile-rich

eclogites

Diavik low-TA154 S kimberlite

cpx in YK1911

Fig 12 (a) Sr^Nd for cpx and (b) Nd-Hf isotope diagram for garnet cpx and reconstructed whole-rocks (symbols with bold outline) Finedashed lines in (b) connect data points from the same sample with garnets having more radiogenic isotopic compositions Uncertainties forreconstructed whole-rocks combine 20 modal uncertainty with uncertainties on each mineralrsquos isotope composition Error bars on individualminerals may be smaller than the symbols Also shown are the isotopic compositions of the depleted mantle (DM DePaolo 1981 Griffin et al2000 Bell amp Tilton 2002) Range of 87Sr86Sr in eclogites from Ekati (Lac de Gras area 100 km north of Diavik Jacob et al 2003) in NorthAmerican kimberlites and compositions of low-temperature eclogites from Diavik and of Lac de Gras A154 South kimberlite fromSchmidberger et al (2007)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1864

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

one low-Mg eclogite has eNd similar to the depleted mantle(Fig 12b)Clinopyroxene in the phlogopitized eclogite YK1911

has unradiogenic Nd and Hf and strongly radiogenicSr whereas the coexisting garnet has radiogenic Nd andHf Although this sample contains 50 vol phlogopiteits whole-rock isotopic composition was calculated

from garnet and cpx alone because the contributionfrom phlogopite to the Hf and Nd budget is 52 and51 respectively (unpublished data) The whole-rockhas radiogenic Nd but has lower eHf than anyother sample analysed in accord with the extreme Hfabundance (53 ppm) and low LuHf of the cpx

DISCUSS IONOrigins of eclogites beneath the SlaveCratonHigh-Mg and high-Ca eclogites

With regard to their major- and trace-element contentsthe high-Mg and high-Ca eclogites from Lac de Gras arebroadly similar to variably evolved oceanic gabbros(Fig 13b and c) which may indicate an origin as subductedoceanic crust Compatible trace-element abundancesare useful to deduce potential crustal protoliths as theseelements are less susceptible to post-formation modification(Jacob 2004) The flat HREE patterns in the high-Mg andhigh-Ca eclogites are compatible with a low-pressureorigin as the oceanic crust is a product of moderate degreesof partial melting mainly within the spinel stability field(eg Presnall et al 2002) whereas melting with residualgarnet would qualitatively be expected to produce stronglyfractionated REE patterns Positive Sr and Pb anomaliesalthough compatible with a plagioclase signature couldalso be due to cryptic metasomatism (Dawson 1984) ofthe eclogites during their residence in the lithosphericmantle These anomalies are probably not related to kim-berlite infiltration as the whole-rock eclogite compositionswere reconstructed from analyses of fresh mineral coresLow ZrSm similar to that in primitive gabbros is

observed for many of the high-Mg and high-Caeclogites (Fig 13b and c) and is independent of the ratioof garnetcpx and whether or not the whole-rock wasreconstructed with rutile (Electronic Appendix 10) Thenegative trend of Mg-number against Yb in garnet(Fig 14) argues against igneous processes involving garnetaccumulation (Taylor amp Neal 1989) as suggested forexample for some high-Mg eclogites from West Africa(Barth et al 2002b) Rather the anti-correlation suggeststhat olivine fractionation played a role in the formation ofthe protoliths Fractionating olivine would excludeYb (Dolivinebasalt

frac14 002 Nikogosian amp Sobolev 1997)and incorporate Mg relative to Fe (Dfrac1415 and 05 respec-tively Taura et al1998) leading to progressive depletion ofMg and enrichment of Yb in the residual melt and henceanti-correlated Mg and Yb in the whole-rock This isinherited after eclogitization All eclogite types (exceptVR43452) plot along the olivine fractionation trend sug-gesting that they formed by a similar processHigh-Mg eclogites have lower

PREE than high-Ca

eclogites and some appear to have positive Eu anomalies

high-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

gabbrosSE Indian ridge

gabbrosOman ophiolite

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

high-Mg

RockNMORB(a)

VR43452

YK1949

low-Mg

boninite-like melts

01

001

1

10

(b)

01

001

1

10

(c)

01

001

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

Fig 13 Trace-element patterns of (a) low-Mg (b) high-Mg and(c) high-Ca eclogites compared with boninite-like volcanic rocksfrom the Isua greenstone belt (Polat et al 2002) gabbros from theOman ophiolite (Benoit et al 1996) and the SE Indian Ridge (Hartet al 1999) normalized to N-MORB of Sun amp McDonough (1989)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1865

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(EuEu up to 14) similar to more primitive cumulates(troctolite gabbros and olivine gabbros) reported in the lit-erature (Fig 13b) Although the sizes of these Eu anomaliesare not outside the analytical uncertainty their correlationwith CeYb in reconstructed eclogites supports their signif-icance (r2frac14066 nfrac14 4) The more primitive nature ofhigh-Mg eclogites is consistent with higher Mg-numbersrelative to high-Ca eclogites and may be due to olivineaccumulation during formation of their protolithsThe two samples with the lowest HREE abundanceshave the highest EuEu which is compatible with varia-tions in plagioclase modesMore evolved protoliths with higher pyroxeneplagio-

clase plus trapped melt could account for small or absentpositive Eu anomalies and higher

PHREE of high-Ca

eclogites (Fig 13c) Although the reason for the frequentassociation of graphite or diamond with high-Ca eclogitesis unclear (Aulbach et al 2003) the frequent occurrence ofkyanite in high-Ca eclogites also observed at Ekati justnorth of Lac de Gras has been argued to support a crustalprotolith (Jacob 2004) Kyanite andor corundum havebeen suggested to exsolve from high-temperature high-Alclinopyroxenes (Caporuscio amp Smyth 1990) Eclogitesfrom Lac de Gras show no exsolution textures suggestiveof such an origin although recrystallization can obscureor destroy microstructural evidence for exsolution(eg Griffin et al 1984)Modelling shows that qualitatively plagioclase accumu-

lation from a low-pressure melt derived from MORBmantle can account for positive Pb Sr and Eu and negativeZr anomalies relative to N-MORB whereas cpx accumula-tion or trapped melt increases Zr Y MREE and HREE

abundances (Fig 15) Some melt addition (for exampleby entrapment) seems to be required to increase all trace-element abundances to those observed for the eclogitesAlthough high-Ca and high-Mg eclogites could representshallow and deep portions of the same oceanic crust thehigher average temperatures obtained for the high-Caeclogites the frequent occurrence of carbon in high-Caeclogites and its absence in high-Mg eclogites the strongLREEHREE fractionation observed for six of seven high-Ca eclogites and the narrow range of compositions of thehigh-Mg eclogites suggest that they may have protolithsthat are not cogenetic Some discrepancies in particularaffecting the incompatible elements may be due to post-crystallization processes acting upon the protoliths suchas dehydration melting and later metasomatism that isevident in peridotites from the same locality (Griffin et al1999)Low K2OTh for high-Ca and high-Mg eclogites

(average 08 Table 4) relative to assumed crustal

0

2

4

6

8

10

30 40 50 60 70 80

Mg

VR43452

Yb

GARNET

pyroxenites

low-Mghigh-Mghigh-Ca

eclogites

VR19673ecl-g2

Fig 14 Garnet Mg-number vs Yb (ppm including 2s uncertainty)Lines qualitatively show separate correlations for eclogites (exceptoutlierVR43452) and pyroxenites (except outlierVR19673ecl-g2)

high-Ca

high-Mg

001

01

1

10

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

melt08 melt02 cryst cpx01 melt09 cryst

plag

(a)

001

01

1

10(b)

RockNMORB

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 15 Reconstructed (a) high-Ca and (b) high-Mg eclogitescompared with compositions of plagioclase and cpx crystallizingfrom a MORB melt modelled using distribution coefficients ofNorman et al (2005)The MORB was calculated as a 10 equilibriummelt of MORB pyrolite [composition from Kelemen et al (1993)except Pb Pr Sr Y and Lu which were calculated as residues from1 fractional melting of pyrolite from McDonough amp Sun (1995)using melt modes of Kinzler (1997) and distribution coefficients ofNorman et al (2005)] Also shown are mixtures of crystals (consistingof 50 plagioclase plus 50 cpx) and melt in the proportions 0901and 0208

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1866

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

precursors (eg gabbros SE Indian Ridge average 22Hart et al 1999) may reflect the preferred loss of fluid-mobile elements during dehydration (Becker et al 2000)although large uncertainties on these low-abundance ele-ments in the constituent minerals and reconstructedwhole-rocks makes this interpretation less than robustThe REE patterns of most high-Mg eclogites are too flatto represent melting residues as partial melting shouldproduce low LREEHREE (Electronic Appendix 10) Incontrast the marked LREE depletion of five of sevenhigh-Ca eclogites (LaLufrac14 02^08) contrasts with that ofassumed evolved gabbroic protoliths (LaLufrac1415^32Hart et al 1999) and can be modelled as a consequence ofpartial melting during subduction after eclogitization (iein the presence of garnet which retains HREE relative toLREE eg Barth et al 2002a) has taken place (Aulbachet al 2003)Our results are compatible with previous studies on

eclogite xenoliths from the Slave Craton includingthose from the same kimberlite that have linked thesesamples or some of these samples to Paleoproterozoicsubduction events (the 195^191Ga formation of theHottah terrane and the 18Ga formation of the GreatBear magmatic arc) in the Slave craton (Griffin et al1999 Kopylova et al 1999 Heaman et al 2002 2006Schmidberger et al 2005 2007)The compositional features discussed above although

compatible with crustal protoliths are not necessarilydefinitive of a subduction origin However if direct crystal-lization from basaltic melts at high pressure (ie mantledepths) is invoked instead specific conditions would haveto be met to satisfy the observations The Mg-number ineclogites from Lac de Gras are lower than those expectedof primary high-pressure mantle melts (eg Walter 1998OrsquoHara amp Herzberg 2002) requiring crystallization athigh pressures after olivine fractionation This originwould further require shielding from peridotitic wallrockwith which the fractionated melts would no longer be inequilibrium (Barth et al 2001) In such a setting the lowrelative abundances of fluid-mobile elements comparedwith those of less fluid-mobile elements and of LREErelative to HREE (Table 4) would reflect an expulsion offluids during crystallization of the protoliths and latermelt mobilization respectively The eclogites clearly donot represent garnet plus cpx cumulates which wouldrequire a positive correlation of Mg-number with Yb thatis at variance with the observed negative correlation ineclogitic garnets from this study (Fig 14)

Low-Mg eclogites

Garnets in low-Mg eclogites compositionally overlap withthose in massive eclogites from Jericho (NW of the Lacde Gras kimberlites) that were suggested to have a high-pressure intrusive origin based on the similarity of theirmajor-element compositions to those of Group A eclogites

fromsouthernAfrica forwhichamantleoriginhasbeenpro-posed (Kopylova et al1999) Garnets in low-Mg eclogites inthe present study have major-element compositions moresimilar to garnets in groupB rather thangroupAeclogitesThe positive slope in N-MORB-normalized HREE

abundances displayed by the low-Mg eclogites is clearlydistinct from other eclogite types and is similar to those ofboninite-like second-stage melts that are closely associatedwith ultramafic volcanics (Smithies et al 2004) (Fig 13a)The range of Mg-numbers in the low-Mg eclogites (55^66)overlaps with that of Archaean second-stage melts fromIsua (58^75) and Abitibi (52^82 Smithies et al 2004)These similarities might indicate that the low-Mg eclogiteshave an origin in a previously depleted mantle source suchas supra-subduction zone mantle Subchondritic ZrHf ofthe low-Mg eclogites contrasts with the supra-chondriticZrHf of the high-Mg and high-Ca eclogites (Aulbach et

al in preparation) and is consistent with previous cpx-controlled depletion in an arc mantle source This sourcecould have been subsequently remelted to generate theboninite-like precursor to the low-Mg eclogites as sug-gested for the protoliths of eclogites from Udachnaya(Jacob amp Foley 1999)Although some second-stage melts to which the low-Mg

eclogites are similar are closely associated with komatiitesand could contain a plume component the radiogenic ordepleted mantle-like Nd and Hf and unradiogenic Sr inthe low-Mg eclogites do not support plume involvement

Trace-element and Sr^Nd^Hf isotopeconstraints on post-formationmetasomatismSeveral low-Mg and high-Ca eclogites show negativeslopes in the LREE which may be indicative of interactionwith small amounts of LREE-enriched melt (Navon ampStolper 1987) amounting to cryptic metasomatism(ie not accompanied by crystallization of new phases) inthe nomenclature of Dawson (1984) High Nb abundancesrelative to elements of similar compatibility displayed bysome low-Mg and high-Ca eclogites (Fig 10) could resultfrom overestimation of rutile modes However high-CaeclogiteVR40345 is rutile-free and has been reconstructedwithout rutile its high Nb abundance is accompanied byLREE enrichment indicating that metasomatic additionmay be responsibleLow-Mg eclogites show a positive correlation of La with

Nb (Electronic Appendix 11a) indicating concomitantaddition of LREE and Nb This suggests that the metaso-matic agent was a solute-rich fluid or silicate silicic melt ashydrous fluids and carbonatites do not carry enough HFSEto affect the budget of eclogites (Green amp Wallace 1988Klemme et al 1995 Adam et al 1997 Stalder et al 1998)Of the samples that are enriched in Nb all but one (high-Ca eclogiteVR50909) show a simultaneous enrichment in

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1867

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Hf (Electronic Appendix 11b) Together these trace-element relationships suggest that both SmNd and LuHfwere lowered in these samples Their low parent^daughterratios contrast with higher ratios in unenriched samplesand resulted in the ingrowth of a range of Nd and Hf iso-tope compositions Compared with the low-Mg and high-Ca eclogites the high-Mg eclogites have a narrow range ofLa and Nb concentrations which correlates with a narrowrange in Nd and Hf isotope compositions (Fig 12)Precise model ages cannot be calculated because parent^

daughter ratios are available only from in situ analyseswhich have relatively large uncertainties Howeverevidence for long-term enrichment and depletion may bediscerned Low-Mg eclogites even those that are enrichedin LREE have relatively high SmNd as a result of havingoverall positive slopes in their REE patterns relative toMORB whereas the high-Mg eclogites have relativelyflat REE patterns and therefore lower SmNd (Fig 11)This translates into lower eNd for high-Mg eclogites thanfor low-Mg eclogites In contrast they show no systematicdifference with respect to 176Hf177Hf Because the low- andhigh-Mg eclogites are proposed to have different origins asdiscussed above and therefore did not necessarily have thesame initial isotopic composition and formation age theseisotopic differences cannot be used to estimate their agesA long-term coupling of trace-element and isotopic

abundances is also indicated by the fact that low-Mg eclo-gite YK1943 which has experienced Nb and LREE addi-tion has lower eNd (0) and eHf (thorn13) than low-Mg eclogiteVR43469 with higher LuHf and SmNd (eNdfrac14thorn47eHffrac14thorn128) If low-Mg eclogites YK1943 and VR43469originally had identical initial Nd and Hf isotope composi-tions and if the parent^daughter ratios are accurate andwere modified only once during their evolution shortlyafter metamorphism (when their isotopic compositionwould have been reset) this overprint can be calculatedfor both isotopic systems to have occurred at 17GaThis would suggest that Hf and Nd were added duringthe same event In contrast high-Mg eclogites VR67360and YK1949 would have had identical Ndisotope compositions at 10Ga whereas their Hf isotopecompositions are disturbed in that the sample with thehigher parent^daughter ratio has a lower 176Hf177Hf thanthe sample with the lower parent^daughter ratio Theseresults suggest that the LREE and HFSE enrichmentobserved in different samples occurred during separateevents that did not affect the eclogites equallyExtreme HFSE enrichment leading to the growth of

zircon and rutile and LREE addition leading to thegrowth of apatite in eclogites from the Jericho kimberlitein the northern Slave craton have been ascribed to twoseparate metasomatic events at around 18Ga (quasi-coincident with the age of formation and metamorphism)and 13^10Ga respectively (Heaman et al 2002 2006

Schmidberger et al 2005) The authors associated theseevents with Paleoproterozoic subduction and aMesoproterozoic thermal disturbance in the Slave litho-spheric mantle possibly linked to the Mackenzie igneousevent Mesoproterozoic Nd model ages mostly between13 and 14Ga were reported for eclogites from thenearby Ekati kimberlites (Jacob et al 2003) Although wedo not observe zircon or apatite in volatile-free eclogitesfrom Lac de Gras and the age constraints obtained fromour data are weak they support a link between theseevents described for other eclogite populations and themetasomatic enrichment observed in our studyThe Sr isotope data in the present study are similar to

those of Schmidberger et al (2007) who suggested that thedepleted-mantle-like values are consistent with derivationof the protoliths from an oceanic mantle source However87Sr86Sr in volatile-free eclogites in the present study isremarkably constant considering their Nd and Hf isotopevariability (Fig 12) and the variability of trace-elementpatterns (Fig 10) Given the evidence for dehydration andmelting in many of the eclogites fractionation of highlyincompatible Rb from Sr and differential ingrowth ofradiogenic Sr would be expectedThis may indicate a rela-tively recent overprint and considering the similarityof 87Sr86Sr to present-day depleted mantle which is signif-icantly lower than 87Sr86Sr of North American kimber-lites a juvenile (asthenospheric) source

PyroxenitesPyroxenites from Lac de Gras often show relationshipsin their major-element contents that are different fromthose observed in the eclogites For example pyroxeniticgarnets show a negative correlation between CaO andNa2O contents whereas these oxides are positively corre-lated for garnets in eclogites (Fig 5) CaO is well correlatedwith C2O3 in pyroxenitic garnets but not in eclogitic gar-nets (Fig 4) Mg-number is positively correlated withYb abundance for pyroxenitic garnets but anti-correlatedfor eclogitic garnets (Fig 14) This suggests that the pyrox-enites in the present study are not related to the eclogitesby fractionation or accumulation processes as proposedfor the origin of some pyroxenites (Foley et al 2003)Pyroxenites have also been interpreted as the products

of crystal segregation in magma channels with alteredoceanic crust as the magma source this would explaintheir Eu anomalies and stable and radiogenic isotopecompositions (Pearson et al 1993 Becker 1996) As thehigh-pressure liquidus phases of most terrestrial basaltsare cpx and garnet (OrsquoHara amp Herzberg 2002) partialmelts of eclogites with basaltic composition wouldcrystallize cpx and garnet but not opx More than half ofthe pyroxenites from Lac de Gras contain opx suggestingthat they probably did not crystallize directly fromeclogite-derived melts

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1868

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Low-Cr websterites and pyroxenites could be crystalli-zation products of mafic melts that interacted with perido-titic mantle thus becoming more refractory The positivecorrelation of Yb and Mg-number in garnet would allowan origin by fractional crystallization from such meltsAlternatively they formed by reaction of silicic (possiblyeclogite-derived) melts with mantle peridotites(eg Kelemen et al 1993 Aulbach et al 2002) This agreeswith the similarity of hypothetical melts in equilibriumwith pyroxenitic garnets in most samples to eclogite-derived melts (Electronic Appendix 12) Niobium concen-trations in all hypothetical melts are much higher than ineclogite-derived melts regardless of whether or not rutileis assumed to be in the residue The same is true for LaCe and Zr concentrations in some samples which mayindicate that NbZr and LREE were added to thepyroxenites some time after their formation as discussedin the previous section for some eclogites Because mini-mum ages for pyroxenitic rutile with an unradiogenicHf isotope composition (calculated by projecting the176Hf177Hf to the model depleted mantle isotopic evolutioncurve assuming zero ingrowth) range to 178066Ga(Aulbach et al in preparation) it appears likely that thepyroxenites were not formed recently (eg during kimber-lite-related or precursory melt migration) and that some ofthe samples are at least MesoproterozoicGarnet pyroxeniteYK1952 consists of schlieren of equili-

brated garnet pyroxenite in a lsquomatrixrsquo of spongy cpxfine cpx^opx intergrowths embayed garnet and meltpatches (Fig 2f) Garnet in the pyroxenite schlieren hasvery low LREE and HFSE abundances Orthopyroxenein the matrix has distinctly higher Mg-number andCr-number and contains more CaO than opx from thepyroxenite schlieren whereas opx rims in the schlierenare compositionally similar to opx in the matrix The con-siderable compositional and textural disequilibrium inthis sample suggests that this is a young feature causedby reaction with fluids shortly before or at the time ofentrainmentAs a consequence of the small sample sizes of most of

the pyroxenites studied here a comprehensive isotopicdataset is not available Clearly more such data areneeded in addition to stable isotope data to obtain betterconstraints on the origins of this rock type Consideringthe microstructural and compositional variability ofthe pyroxenites and major-element relationships oppositeto those observed for eclogites it appears certain that pyr-oxenites form by a variety of processes some of which areunrelated to the formation of eclogites and their protoliths

Volatile-rich eclogitesVolatile-rich eclogites from Lac de Gras display the mostextreme Hf^Nd isotope systematics of all the analysedsamples Garnets in phlogopitized (YK1911) and in cal-cite-bearing eclogite (VR67112b) have the most radiogenic

Nd (eNdfrac14 2486 and 1927 respectively) Garnet inVR67112b has extremely high eHf (2516) Clinopyroxene inYK1911 has the lowest eHf measured for any mineralseparate from Lac de Gras (53) in accord with the veryhigh Hf abundance in its cpx (5 ppm) In contrastgarnet in this sample has radiogenic Hf (eHffrac14 136) Thisdiscrepancy is at least in part due to their relatively lowequilibration temperatures (7698C and 8558C respec-tively) which allowed the isotopic compositions of cpxand garnet to evolve separately at their low andhigh parent^daughter element ratios respectively forextended periods of time and possibly since formationAccordingly the cpx in YK1911 gives a minimum Hf ageof 27GaSlab-derived fluids could have been introduced during

the 27Ga collision of the western Slave Craton domainincluding the thick plume-derived deep lithosphericmantle layer and the eastern domain with its highlydepleted lithosphere This would also be consistent withthe observation that volatile-rich eclogites occur at shallowdepths in the mantle (TKroghfrac14 685^8558C correspondingto depths of 85^115 km along a 40 mWm2 geotherm)where the bulk of dehydration and production of solute-rich melts in a subducting slab might be expected to occur(Becker et al 2000 Hermann et al 2006) Their negativeEu anomalies might indicate that the fluid was derivedfrom the shallow portion of the slab that experiencedplagioclase fractionation The presence of carbonate inVR67112b requires a carbonated sourceHigh 87Sr86Sr in cpx in YK1911 is probably the conse-

quence of isotopic equilibration with high-RbSr phlogo-pite after long-term radiogenic ingrowth while Hfaddition led to retarded ingrowth of radiogenic HfThese systematics combined with radiogenic Nd appear

inconsistent with the involvement of sediments in the pet-rogenesis of this sample (eg Perfit et al 1980 Vervoortet al 1999) or a hydrous fluid such as those responsible forMARID metasomatism in the Kaapvaal craton (Konzettet al 1998)Whatever the origin of the fluids the apparentHf mobility suggests that they were not in equilibriumwith a rutile-bearing source and also that they weresolute-rich in nature (ie hydrous melts able to dissolveHFSE) and the relatively high HREE abundances of thevolatile-rich eclogites indicate a garnet-free source suchas amphibolite This allowed evolution of the sample atrelatively high time-integrated SmNd

SUMMARY AND CONCLUSIONS

(1) The Lac de Gras kimberlites carried abundant eclo-gite and pyroxenite xenoliths to the surface Threemain eclogite types are distinguished in terms ofmajor-element composition and trace element

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1869

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

patterns high-Mg high-Ca and low-Mg eclogitesThe garnets of all eclogite types have anti-correlatedMg-number and Yb pointing to olivine fractionationas a controlling factor in the formation of their pre-cursor rocks Their flat or positively sloped HREEpatterns qualitatively indicate that the melt parentalto the precursor rocks was not in equilibrium with agarnet-bearing residue

(2) Reconstructed high-Mg eclogites and high-Caeclogites have flat HREE patterns positive Pb andSr anomalies and low ZrSm similar to oceanic gab-bros reported in the literature High-Mg eclogiteshave variably low HREE contents and some havesmall positive Eu anomalies possibly suggestive ofmore primitive relatively plagioclase-rich protolithswhereas high-Ca eclogites have higher HREE abun-dances and small or absent Eu anomalies that requiremore cpx- andor melt-rich protoliths if a subductionorigin is assumed

(3) Low-Mg eclogites have distinctive trace-element pat-terns that show a marked positive slope in the HREEand subchondritic ZrHf They resemble someArchaean second-stage melts (lsquoboninitesrsquo) and theirprotoliths appear to have formed from a previouslydepleted source

(4) After formation the eclogites were subject to a varietyof processes Most have low ratios of fluid-mobile vsless fluid-mobile elements possibly as a result of dehy-dration of the protoliths In addition some of the low-Mg and high-Ca eclogites have low LREEHREEconsistent with partial melt lossWhereas the low-Mgeclogites show concomitant enrichment in La Nb andZr some high-Mg and high-Ca eclogites show evi-dence for variable and unsystematic (decoupled)LREE and Nb addition

(5) Isotopically the trace-element characteristics of thedifferent eclogite types translate into lower eNd forhigh-Mg eclogites (negative slope in MREEfrac14 lowerSmNd) than for low-Mg eclogites (positive slope inMREEfrac14higher SmNd) Within the low-Mg groupsamples that show evidence for metasomatism (lowLuHf and SmNd) have lower eNd and eHf than asample that was apparently not metasomatized BothNd and Hf addition can be dated to 17Ga if identi-cal initial isotope compsitions are assumed for thelow-Mg eclogites In contrast the relatively constantand depleted mantle-like 87Sr86Sr of eclogitic cpxdespite variable Nd and Hf isotope compositions andparent^daughter ratios suggests recent input of juve-nile material

(6) Garnet and cpx in pyroxenites show major-elementrelationships different from those in eclogitesTherefore these two rock types are probably notrelated by fractionation processes The presence of

opx precludes direct crystallization from slab-derivedmelts Pyroxenites may be the product of (slab-derived) melts hybridized during assimilation of peri-dotitic mantle

(7) Garnets in phlogopite- and carbonate-bearing eclo-gites have extreme isotopic compositions includingthe most radiogenic Nd and Hf in the datasetNegative Eu anomalies may point to the involvementof a crustal protolith Highly radiogenic Sr radiogenicNd and unradiogenic Hf in the reconstructedphlogopite-bearing sample combined with highHREE abundances and evidence for Hf mobilitymay indicate involvement of solute-rich fluids origi-nating in a garnet- and rutile-free source

ACKNOWLEDGEMENTSHelp in the clean laboratories and with the analyticalfacilities by S Elhlou C Lawson A Sharma P Wielandand M Zhang is greatly appreciated Bill Griffin isthanked for his support and for agreeing to reviewseveral versions of the manuscript despite disagreeingwith many of the conclusions presented here The paperbenefited greatly from the detailed and constructivereviews provided by the editors Gareth Davies andMarjorie Wilson and by Matthias Barth Larry Heamanand an anonymous reviewer This work was funded by aMacquarie University International PostgraduateAward and Postgraduate Research Fund (SA) by anARC SPIRT grant sponsored by Kennecott Canada Incand by ARC Large and Discovery Grants to SYOrsquoRAnalytical data were obtained using instrumentationfunded by ARC LIEF and DEST Systemic InfrastructureGrants industry partners and Macquarie UniversityThis is Publication 481 from the ARC National KeyCentre for Geochemical Evolution and Metallogeny ofContinents (wwwesmqeduauGEMOC)

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online

REFERENCESAdam J GreenT H Sie S H amp Ryan C G (1997) Trace element

partitioning between aqueous fluids silicate melts and mineralsEuropeanJournal of Mineralogy 9 569^584

Aulbach S Stachel T Viljoen K S Brey G P amp Harris J W(2002) Eclogitic and websteritic diamond sources beneath theLimpopo Beltccedilis slab-melting the link Contributions to Mineralogy

and Petrology 143 56^70Aulbach S GriffinW L Pearson N J OrsquoReilly S Y Kivi K amp

Doyle B (2003) Origins of eclogites beneath the central SlaveCraton Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA-0011

Aulbach S Griffin W L Pearson N J OrsquoReilly S Y amp Kivi K(2004a) Mantle formation and evolution Slave Craton constraints

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1870

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

from HSE abundances and Re^Os isotope systematics of sulfideinclusions in mantle xenocrysts Chemical Geology 208 61^88

Aulbach S Griffin W L OrsquoReilly S Y amp McCandless T E(2004b) Genesis and evolution of the lithospheric mantle beneaththe Buffalo HeadTerrane Alberta (Canada) Lithos 77 413^451

Aulbach S GriffinW L Pearson N J OrsquoReilly S Y amp Doyle B(2007) Lithosphere formation in the central Slave Craton(Canada) plume subcretion or lithosphere accretion Contributionsto Mineralogy and Petrology (in press) doi 101007s00410-007-0200-1

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2001)Geochemistry of xenolithic eclogites from West Africa Part IA link between low MgO eclogites and Archean crust formationGeochimica et Cosmochimica Acta 65 1499^1527

Barth M G Foley S F amp Horn I (2002a) Partial melting inArchean subduction zones constraints from experimentally deter-mined trace element partition coefficients between eclogitic miner-als and tonalitic melts under upper mantle conditions PrecambrianResearch 113 323^340

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2002b)Geochemistry of xenolithic eclogites from West Africa part 2Origins of the high MgO eclogites Geochimica et Cosmochimica Acta

66 4325^4345Beard B L Fraracci K NTaylor L A Snyder G A Clayton R

A Mayeda T K amp Sobolev NV (1996) Petrography andgeochemistry of eclogites from the Mir kimberliteYakutia RussiaContributions to Mineralogy and Petrology 125 293^310

Becker H (1996) Geochemistry of garnet peridotite massifs fromlower Austria and the composition of deep lithosphere beneatha Palaeozoic convergent plate margin Chemical Geology 134 49^65

Becker H Jochum K P amp Carlson R W (2000) Trace elementfractionation during dehydration of eclogites from high-pressureterranes and the implications for element fluxes in subductionzones Chemical Geology 163 65^99

Bell K amp Tilton G R (2002) Probing the mantle The story fromcarbonatites EOS Transactions American Geophysical Union 83275^277

Benoit M Polvecurren M amp Ceuleneer G (1996) Trace element andisotopic characterisation of mafic cumulates in a fossil mantlediapir (Oman ophiolite) Chemical Geology 134 199^214

BleekerW (2003) The late Archean record a puzzle in ca 35 piecesLithos 71 99^134

BleekerW Ketchum JW F JacksonV A amp Villeneuve M (1999)The central Slave Basement Complex Part I Its structural topol-ogy and autochthonous core Canadian Journal of Earth Sciences 361083^1109

Blichert-Toft J amp Albare de F (1997) The Lu^Hf isotope geochemis-try of chondrites and the evolution of the mantle^crust systemEarth and Planetary Science Letters 148 243^258

Blichert-Toft J Chauvel C amp Albare de F (1997) Separation of Hfand Lu for high precision isotope analysis of rock samples by mag-netic sector-multiple collector ICP-MS Contributions to Mineralogy

and Petrology 127 248^260Brey G P amp Kolaquo hler T (1990) Geothermobarometry in four-phase

lherzolites II New thermobarometers and practical assessment ofexisting thermobarometers Journal of Petrology 31 1353^1378

Brey G P Kolaquo hlerT amp Nickel K G (1990) Geothermobarometry infour-phase lherzolites I Experimental results from 10 to 60 kbJournal of Petrology 31 1313^1352

Caporuscio F amp Smyth J (1990) Trace element crystal chemistryof mantle eclogites Contributions to Mineralogy and Petrology 105550^561

Carbno G B amp Canil D (2002) Mantle structure beneath the SWSlave Craton Canada constraints from garnet geochemistry in theDrybones Bay Kimberlite Journal of Petrology 43 129^142

Coleman R G Lee E D Beatty L B amp Brannock W W (1965)Eclogites and eclogites their differences and similaritiesGeological Society of America Bulletin 76 483^508

Creaser R A Grulaquo tter H Carlson J amp Crawford B (2004)Macrocrystal phlogopite Rb^Sr dates for the Ekati propertykimberlites Slave Province Canada evidence for multiple intru-sive episodes in the Paleocene and Eocene Lithos 76 399^414

Davies R M GriffinW L Pearson N J Andrew A S Doyle BJ amp OrsquoReilly S Y (1999) Diamonds from the deep Pipe DO-27Slave Craton Canada In Gurney J J Gurney J L Pascoe MD amp Richardson S H (eds) Proceedings of the 7th International

Kimberlite Conference CapeTown Red Roof Design pp 148^155Davies R M Griffin W L OrsquoReilly S Y amp Doyle B J (2004)

Mineral inclusions and geochemical characteristics of microdia-monds from the DO27 A154 A21 A418 DO18 DD17 and RanchLake kimberlites at Lac de Gras Slave Craton Canada Lithos 7739^55

DavisW J amp Hegner E (1992) Neodymium isotopic evidence for thetectonic assembly of Late Archean crust in the Slave Provincenorthwest Canada Contributions to Mineralogy and Petrology 111493^504

Davis W J Jones A G Bleeker W amp Grutter H (2003)Lithosphere development in the Slave craton a linked crustal andmantle perspective Lithos 71 575^589

Dawson J B (1984) Contrasting types of upper-mantle metasoma-tism In Kornprobst J (ed) Kimberlites II The Mantle and

Crust^Mantle Relationships Amsterdam Elsevier pp 289^294DePaolo D J (1981) Neodymium isotopes in the Colorado Front

Range and crust^mantle evolution in the Proterozoic Nature 291193^196

Erlank A J amp Kushiro I (1970) Potassium contents of syntheticpyroxenes at high temperatures and pressures Carnegie Institution ofWashingtonYearbook 68 433^439

Foley S F Barth M G amp Jenner G A (2000) Rutilemelt partitioncoefficients for trace elements and an assessment of the influence ofrutile on the trace element characteristics of subduction zonemagmas Geochimica et Cosmochimica Acta 64 933^938

Foley S F Buhre S amp Jacob D E (2003) Evolution of the Archaeancrust by delamination and shallow subduction Nature 421 249^252

Graham I Burgess J L Bryan D Ravenscroft P J Thomas EDoyle B J Hopkins R amp Armstrong K A (1999) Explorationhistory and geology of the Diavik Kimberlites Lac de GrasNorthwest Territories Canada In Gurney J J Gurney J LPascoe M D amp Richardson S H (eds) Proceedings of the 7th

International Kimberlite Conference Cape Town Red Roof Designpp 262^279

Green D H amp Wallace M E (1988) Mantle metasomatism byephemeral carbonatite melts Nature 336 459^462

Green T H Blundy J D Adam J amp Yaxley G M (2000) SIMSdetermination of trace element partition coefficients betweengarnet clinopyroxene and hydrous basaltic liquids at 2^75 GPaand 1080^12008C Lithos 53 165^187

GriffinW L amp OrsquoReilly SY (2007) Cratonic lithospheric mantle Isanything subducted Episodes 30(1) 43^53

Griffin W L Wass S Y amp Hollis J D (1984) Ultramafic xenolithsfrom Bullenmerri and Gnotuk maars Victoria Australia petrol-ogy of a sub-continental crust^mantle transition Journal of

Petrology 25 53^87GriffinW L Doyle B J Ryan C G Pearson N J OrsquoReilly SY

Davies R Kivi K van Achterbergh E amp Natapov LM (1999)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1871

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Layered mantle lithosphere in the Lac de Gras Area SlaveCraton composition structure and origin Journal of Petrology 40705^727

Griffin W L Pearson N J Belousova E Jackson S E vanAchterbergh E OrsquoReilly SY amp Shee S R (2000)The Hf isotopecomposition of cratonic mantle LAM-MC-ICPMS analysis ofzircon megacrysts in kimberlites Geochimica et Cosmochimica Acta

64 133^147Griffin W L OrsquoReilly S Y Doyle B J Pearson N J

Coopersmith H Kivi K Malkovets V amp Pokhilenko N(2004) Lithosphere mapping beneath the North American plateLithos 77 873^922

Grulaquo tter H S Apter D B amp Kong J (1999) Crust^mantle couplingevidence from mantle-derived xenocrystic garnets In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 307^312

Hart S R Blusztajn J Dick H J B Meyer P S ampMuehlenbachs K (1999) The fingerprint of seawater circulationin a 500-meter section of ocean crust gabbros Geochimica et

Cosmochimica Acta 63 4059^4080Harte B amp Kirkley M B (1997) Partitioning of trace elements

between clinopyroxene and garnet Data from mantle eclogitesChemical Geology 136 1^24

Heaman L M Creaser R A amp Cookenboo H O (2002) Extremeenrichment of high field strength elements in Jericho eclogitexenoliths A cryptic record of Paleoproterozoic subduction partialmelting and metasomatism beneath the Slave craton CanadaGeology 30 507^510

Heaman L M Kjarsgaard B A amp Creaser R A (2004) The tem-poral evolution of North American kimberlites Lithos 76 377^397

Heaman L M Creaser R A Cookenboo H O amp Chacko T(2006) Multi-stage modification of the northern Slave mantlelithosphere Evidence from zircon- and diamond-bearing eclogitexenoliths entrained in Jericho Kimberlite Journal of Petrology 47821^858

Hermann J Cpandler C Hack A amp Korsakov A (2006) Aqueousfluids and hydrous melts in high-pressure and ultra-high pressurerocks implications for element transfer in subduction zones Lithos92 399^417

Hills DV amp Haggerty S E (1989) Petrochemistry of eclogites fromthe Koidu Kimberlite Complex Sierra Leone Contributions to

Mineralogy and Petrology 103 397^422Hoffman P F (1989) Precambrian geology and tectonic history of

North America In Bally A amp Palmer A (eds) The Geology of

North AmericaccedilAn Overview Boulder CO Geological Society ofAmerica pp 447^512

IrelandT R Rudnick R L amp Spetsius Z (1994) Trace-elements indiamond inclusions from eclogites reveal link to Archean granitesEarth and Planetary Science Letters 128 199^213

Irvine G J Pearson D G Kjarsgaard B A Carlson R WKopylova M G amp Dreibus G (2003) Evolution of the litho-spheric mantle beneath Northern Canada a Re^Os isotope andPGE study of kimberlite-derived peridotite xenoliths fromSomerset Island and a comparison to the Slave and Kaapvaalcratons Lithos 71 461^488

Isachsen C E amp Bowring S A (1994) Evolution of the Slave cratonGeology 22 917^920

Jacob D E (2004) Nature and origin of eclogite xenoliths from kim-berlites Lithos 77 295^316

Jacob D E amp Foley S F (1999) Evidence for Archean ocean crustwith low high field strength element signature from diamondifer-ous eclogite xenoliths Lithos 48 317^336

Jacob D Jagoutz E Lowry D Mattey D amp Kudrjavtseva G(1994) Diamondiferous eclogites from Siberiaccedilremnants ofArchean oceanic-crust Geochimica et Cosmochimica Acta 585191^5207

Jacob D E Fung A Jagoutz E amp Pearson D G (2003) Petrologyand geochemistry of eclogite xenoliths from the Ekati kimberlitesarea Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA_239

Jacob D E Bizimis M amp Salters V J M (2005) Lu^Hf and geo-chemical systematics of recycled ancient oceanic crust evidencefrom Roberts Victor eclogites Contributions to Mineralogy and

Petrology 148 707^720Jerde E A Taylor L A Crozaz G Sobolev NV amp SobolevV N

(1993) Diamondiferous eclogites from Yakutia Siberiaccedilevidencefor a diversity of protoliths Contributions to Mineralogy and Petrology

114 189^202Jones A G Ferguson I J Chave A D Evans R L amp

McNeice GW (2001) The electric lithosphere of the Slave cratonGeology 29 423^426

Kelemen P B Shimizu N amp Dunn T (1993) Relative depletionof niobium in some arc magmas and the continental-crust-partitioning of K Nb La and Ce during meltrockreaction in the upper-mantle Earth and Planetary Science Letters 120111^134

Kinzler R J (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition Application to mid-ocean ridge basalt petrogenesis Journal of Geophysical Research 102853^874

Klemme S van der Laan S R Foley S F amp Guenther D (1995)Experimentally determined trace and minor element partitioningbetween clinopyroxene and carbonatite melt under upper mantleconditions Earth and Planetary Science Letters 133 439^448

Konzett J Armstrong R A Sweeney R J amp CompstonW (1998)The timing of MARID metasomatism in the Kaapvaal mantle anion probe study of zircons from MARID xenoliths Earth and

Planetary Science Letters 160 133^145Kopylova M G Russell J K amp Cookenboo H (1999) Mapping the

lithopshere beneath the north central Slave Craton In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 468^479

Krogh E (1988) The garnet^clinopyroxene iron^magnesiumgeothermometerccedila reinterpretation of existing experimentaldata Contributions to Mineralogy and Petrology 99 44^48

Kusky T M (1989) Accretion of Archean Slave Province Geology 1763^67

LeCheminant A N amp Heaman L M (1989) Mackenzie igneousevents Canada Middle Proterozoic hotspot magmatismassociated with ocean opening Earth and Planetary Science Letters

96 38^48LeCheminant A N Richardson D G DiLabio R N W amp

Richardson K A (1996) La recherche de diamants au CanadaGeological Survey of Canada Open File 3228 268pp

MacGregor I D amp Carter J L (1970) The chemistry of clinopyrox-enes and garnets of eclogite and peridotite xenoliths from theRoberts Victor mine South Africa Physics of the Earth and PlanetaryInteriors 3 391^397

MacGregor I D amp Manton W I (1986) Roberts Victor eclogitesancient oceanic crust Journal of Geophysical Research 91 14063^14079

MacKenzie J M amp Canil D (1999) Composition and thermal evo-lution of cratonic mantle beneath the central Archean SlaveProvince NWT Canada Contributions to Mineralogy and Petrology

134 313^324

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1872

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

McCandlessT E amp Gurney J J (1989) Sodium in garnet and potas-sum in clinopyroxene criteria for classifying mantle eclogites InRoss J (ed) Kimberlites and Related Rocks Carlton Blackwellpp 827^832

McDonoughW F amp Sun S-S (1995) The composition of the EarthChemical Geology 120 223^253

Navon O amp Stolper E (1987) Geochemical consequences of meltpercolation the upper mantle as a chromatographic columnJournal of Geology 95 285^307

Nikogosian I K amp Sobolev AV (1997) Ion-microprobe analysis ofmelt inclusions in olivine experience in estimating the olivine^meltpartition coefficients of trace elements Geochemistry International 35119^126

Norman M D Griffin W L Pearson N J Garcia M O ampOrsquoReilly S Y (1998) Quantitative analysis of trace element abun-dances in glasses and minerals a comparison of laser ablationinductively coupled plasma mass spectrometry solution inductivelycoupled plasma mass spectrometry proton microprobe and elec-tron microprobe data Journal of Analytical Atomic Spectrometry 13477^482

Norman M D Garcia M O amp Pietruszka A J (2005) Trace-element distribution coefficients for pyroxenes plagioclase andolivine in evolved tholeiites from the 1955 eruption of KilaueaVolcano Hawairsquoi and petrogenesis of differentiated rift-zonelavas American Mineralogist 90 888^899

OrsquoHara M J amp Herzberg C (2002) Interpretation of trace elementand isotope features of basalts Relevance of field relations petrol-ogy major element data phase equilibria and magma chambermodeling in basalt petrogenesis Geochimica et Cosmochimica Acta 662167^2191

PadghamW A amp FysonW K (1992) The Slave Province a distinctcraton CanadianJournal of Earth Sciences 29 2072^2086

Pearson D G Davies G R amp Nixon P H (1993) Geochemical con-straints on the petrogenesis of diamond facies pyroxenites from theBeni Bousera peridotite massif North Morocco Journal of Petrology34 125^172

Pearson N J Griffin W L Doyle B J OrsquoReilly S Y vanAchterbergh E amp Kivi K (1999) Xenoliths from kimberlitepipes of the Lac de gras area Slave Craton Canada In GurneyJ J Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite ConferenceCapeTown Red Roof Design pp 644^658

Perfit N R Gust D A Bence A E Arculus R J amp Taylor S R(1980) Chemical characteristics of island-arc basalts implicationsfor mantle sources Chemical Geology 30 227^256

Polat A Hofmann A W amp Rosing M T (2002) Boninite-likevolcanic rocks in the 37^38 Ga Isua greenstone belt WestGreenland geochemical evidence for intra-oceanic subductionzone processes in the early Earth Chemical Geology 184 231^254

Poudjom DjomaniY H GriffinW L OrsquoReilly SY amp Doyle B J(2005) Lithospheric domains and controls on kimberlite emplace-ment Slave Province Canada evidence from elastic thickness andupper mantle composition Geochemistry Geophysics Geosystems 6doi1010292005GC0009978

Presnall D C Gudfinnsson G H amp Walter M J (2002)Generation of mid-ocean ridge basalts at pressures from 1 to 7GPa Geochimica et Cosmochimica Acta 66 2073^2090

Schmidberger S S Heaman L M Simonetti A Creaser R A ampCookenboo H O (2005) Formation of Paleoproterozoiceclogitic mantle Slave Province (Canada) Insights from in-situHf and U^Pb isotopic analyses of mantle zircons Earth and

Planetary Science Letters 240 621^633

Schmidberger S S Simonetti A Heaman L M Creaser R A ampWhiteford S (2007) Lu^Hf in-situ Sr and Pb isotope and traceelement systematics for mantle eclogites from the Diavik diamondmine Evidence for Paleoproterozoic subduction beneath the Slavecraton Canada Earth and Planetary Science Letters 254 55^68

Schulze D J amp Helmstaedt H (1988) Coesite-sanidine eclogitesfrom kimberlite products of mantle fractionation or subductionJournal of Geology 96 435^443

Smithies R H Champion D C amp Sun S S (2004) The case forArchaean boninites Contributions to Mineralogy and Petrology 147705^721

Smyth J R Caporuscio F A amp McCormick T C (1989) Mantleeclogites evidence of igneous fractionation in the mantle Earth andPlanetary Science Letters 93 133^141

Sobolev NV amp Lavrentrsquoyev Y G (1971) Isomorphic sodium admix-ture in garnets formed at high pressure Contributions to Mineralogy

and Petrology 31 1^12Stalder R Foley S F Brey G P amp Horn I (1998) Mineral aqueous

fluid partitioning of trace elements at 900^12008C and 30^57GPa New experimental data for garnet clinopyroxene and rutileand implications for mantle metasomatism Geochimica et

Cosmochimica Acta 62 1781^1801Sun S-S amp McDonough W F (1989) Chemical and isotopic sys-

tematics of oceanic basalts implications for mantle compositionsand processes In Saunders A D amp Norris M J (eds)Magmatism in the Ocean Basins Geological Society London Special

Publications 42 313^345Taura H Yurimoto H Kurita K amp Sueno S (1998) Pressure

dependence on partition coefficients for trace elements between oli-vine and the coexisting melts Physics and Chemistry of Minerals 25469^484

Taylor L A amp Neal C R (1989) Eclogites with oceanic crustal andmantle signatures from the Bellsbank kimberlite South AfricaPart I Mineralogy petrography and whole rock chemistryJournal of Geology 97 551^567

van Achterbergh E Ryan C G amp Griffin W L (1999) Glitter online intensity reduction for the laser ablation inductively coupledplasma spectrometry Abstract 9th V M GoldschmidtConference LPI Boston MA 905 pp Abstract no 7215

van Breemen O DavisW J amp King J E (1992) Temporal distribu-tion of granitoid plutonic rocks in the Archean Slave ProvinceNorthwest Canadian Shield Canadian Journal of Earth Sciences 292186^2199

Vervoort J D Patchett P J Blichert-Toft J amp Albare de F (1999)Relationship between Lu^Hf and Sm^Nd isotopic systems in theglobal sedimentary system Earth and Planetary Science Letters 16879^99

Viljoen K S Smith C B amp Sharp Z D (1996) Stable and radio-genic isotope study of eclogite xenoliths from the Orapa kimberliteBotswana Chemical Geology 131 235^255

Walter M J (1998) Melting of garnet peridotite and theorigin of komatiite and depleted lithosphere Journal of Petrology 3929^60

Wasserburg G J Jacobsen S B DePaolo D J McCulloch M Tamp Wen T (1981) Precise determination of SmNd ratios Sm andNd isotopic abundances in standard solutions Geochimica et

Cosmochimica Acta 45 2311^2323Westerlund K J Shirey S B Richardson S H Carlson R W

Gurney J J amp Harris JW (2006) A subduction wedge origin forPaleoarchean peridotitic diamonds and harzburgites from thePanda kimberlite Slave craton evidence from Re^Os isotope sys-tematics Contributions to Mineralogy and Petrology 152 275^294

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1873

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Page 6: Origins of Xenolithic Eclogites and Pyroxenites from the Central

(phlogopite occurring as small secondary grains at grainboundaries and in veins is not considered in this study)Ten samples of all eclogite types show secondary spongyrims of cpx around cpx cores Secondary rims containsmall (mm-scale) patches of glassRutile is the most common primary accessory phase

(15 of 65 eclogites and pyroxenites) with modes between51 and 2 vol Sulfide occurs in 14 samples and ispresent as rounded to subhedral to irregular grains

Kyanite has been identified in three samples studied byPearson et al (1999) One of the kyanite-bearing eclogitesalso contains diamond (VR40345) Kyanite as well asdiamond or graphite associated with kyanite-bearing andcompositionally similar eclogites (high-Ca eclogites) hasbeen frequently observed in drill core (Pearson et al1999) but because these eclogite types are often alteredand friable they are not proportionally represented in thepresent study One high-Ca eclogite (VR43465) contains

Fig 2 Eclogites in thin section (plane-polarized light) illustrating the microstructures mentioned in the text (a) high-Mg eclogiteVR67360(b) high-Mg eclogiteYK1946 (c) phlogopitized eclogiteYK1911 (d) garnet pyroxeniteYK1915 (e) garnet pyroxeniteYK1914 (f) garnet pyrox-enite schlieren in a matrix of garnet spongy opx and glass (YK1952) Grt garnet cpx clinopyroxene rut rutile phlog phlogopite Scalebar in(c) applies to all sections (1cm)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1848

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

3 vol graphite which occurs as disseminated plates sev-eral millimetres long with prismatic tabular or irregularhabit Quartz has been retrieved from a mineral separateof low-Mg eclogite YK1943 The phlogopitized samples(VR19674ecl-g7 and YK1911) also contain apatite andglass as accessory phases Vesicles in sample VR67112bare partially filled with single or multiple large carbonatecrystalsMost pyroxenites from Lac de Gras have fine- to coarse-

grained granoblastic equilibrated microstructures(Pearson et al 1999) Modal information for pyroxenitesin the present study is available for only the larger sampleYK1914 (22 garnet 78 cpx) (Fig 2e)

MAJOR ELEMENTSAverage major-element compositions of garnet cpx andopx in the different eclogite types (except for the threevolatile-rich eclogites which are not averaged) and in pyr-oxenites are given in Table 2 the full datasets includingaccessory minerals are available as ElectronicAppendices 2^5 which can be downloaded from httppetrologyoxfordjournalsorg

GarnetGarnets have XMg [pyrope component 100Mg(MgthornFethornCathornMn)] ranging from 178 to 756 andXCa (grossular component mol) ranging from 72 to359 (Electronic Appendix 2) In a diagram of CaO(MgOthornFeO) vs MgOFeO (Fig 3) eclogitic garnetsfall into three groups (1) low-Mg garnets with low

CaO(MgOthornFeO) and low MgOFeO (2) high-Mggarnets with low CaO(MgOthornFeO) and high MgOFeO (3) high-Ca those with high CaO(MgOthornFeO)and intermediate MgOFeO Volatile-rich eclogites havegarnets with the lowest MgOFeO whereas pyroxenitestrend towards higher values than eclogitesMost eclogite garnets have Cr2O3 contents502wt

with variable CaO contents (Fig 4) Garnets in pyroxeniteshave low CaO contents which correlate positively withCr2O3 More than half of the pyroxenites are opx-free andall pyroxenites are olivine-free on thin-section scalePyroxenitic garnets show an opposing trend of CaOvs Na2O (Fig 5) compared with the eclogitic garnetsBecause the positive correlation of Cr2O3 and CaO forpyroxenitic garnet indicates buffering by both cpx andopx opx-bearing or opx-free pyroxenites will not befurther distinguished subsequently (except in the geother-mobarometry section where opx-bearing assemblagesallow simultaneous calculation of pressure andtemperature)Distinct rim compositions in most garnets are

characterized by lower CaO contents and higherMg-number [frac14 100Mg(MgthornFe)] Garnet compositionsin three samples are inhomogeneous with regard to CaOMgO and FeO (SE01 VR43479 and VR43480) withoutclear core^rim zonations similar to coexisting cpx Likerims CaO-poor compositions are characterized by higherMg-number Visibly secondary spongy cpx rims also havehigher MgO contents (see below) and MgO-rich garnetsare therefore regarded as affected by secondary changesBased on Cr^Ca Mg^Ca^Fe and Ca^Na composi-

tional relationships in garnet the following eclogitetypes are distinguished (1) eclogites with high-Ca

0

05

10

15

20

25

30

0 01 02 03 04 05 06 07 08 09

high-Ca

high-Mg+pyroxenites

low-Mg

CaO(MgO+FeO)

MgO

FeO

GARNETpyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropic Jericho massiveDiavik low-temperatureDiavik high-temperatureDiavik diamond-bearing

eclogites

Fig 3 CaO(MgOthornFeO) vs MgOFeO (wt ) in garnet Fieldsdistinguish low-Mg and high-Ca eclogites as well as high-Mg eclogitesplus pyroxenites in this study Garnets in eclogites from Jericho(Kopylova et al 1999) and Diavik (Schmidberger et al 2007) areshown for comparison Garnet inclusions in diamond from Davieset al (1999 2004)

-135 wt (high-Ca)

2

3

4

5

6

7

8

9

0 05 10 15 20 25Cr2O3

CaO

GARNET

opx-bearing

opx-free

pyroxenites

low-Mghigh-Mgvolatile-rich

Jericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

VR19673ecl-g1g2

Fig 4 Cr2O3 vs CaO (wt ) in garnet Garnets in eclogites fromJericho and Diavik as in Fig 3

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1849

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 2 Summary of major-element contents in garnet and cpx (wt )

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO Na2O Total Mg-no T(Krogh) Type

Garnet

Low-Mg ecl 3964 026 2214 011 1843 037 1050 846 009 1000 503 976

minimum 3893 008 2130 5009 1071 024 849 567 006 990 423 860

maximum 4094 036 2275 011 2093 046 1457 1062 012 1004 708 1088

1s 070 012 052 na 354 007 227 185 003 05 99 98

High-Mg ecl 4081 026 2261 013 1419 030 1537 624 007 1000 659 969

minimum 3970 008 2224 5009 951 023 1127 418 003 991 551 820

maximum 4250 052 2323 035 1795 036 2030 882 013 1012 792 1241

1s 077 013 031 na 225 004 189 141 003 05 57 88

High-Ca ecl 3979 034 2211 na 1516 027 958 1240 011 998 528 1078

minimum 3909 011 2117 5009 1164 023 801 1144 007 989 456 910

maximum 4039 073 2242 011 1713 035 1249 1351 018 1004 657 1246

1s 042 026 042 na 173 004 155 073 004 06 67 151

vr19674ecl-g7 3867 003 2187 5009 2380 153 793 627 5004 1002 373 volatile-rich

vr67112b 3783 013 2097 009 2620 204 466 790 5004 999 241 volatile-rich

yk1911 3874 5006 2200 5009 2120 221 802 747 5004 997 403 volatile-rich

Pyroxenites 4163 021 2257 075 1148 034 1906 407 006 1002 747 909

minimum 3989 005 2081 009 729 025 1535 277 004 996 593 699

maximum 4279 048 2330 260 1881 054 2215 566 009 1011 842 1158

1s 068 010 057 057 323 007 209 065 001 04 73 117

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO Na2O Total Mg-no Type

Cpx

Low-Mg ecl 5453 028 739 na 500 5009 1117 1662 434 994 801

minimum 5160 017 564 5009 337 na 876 1388 197 981 741

maximum 5570 044 1067 014 687 na 1324 1926 632 1001 865

1s 128 011 155 na 124 na 153 177 126 06 40

High-Mg ecl 5483 023 483 na 381 na 1368 1875 318 994 865

minimum 5353 009 263 5009 245 5009 1156 1617 204 986 778

maximum 5645 031 810 016 674 011 1588 2101 495 1011 920

1s 071 007 137 na 095 na 104 138 075 06 30

High-Ca ecl 5563 027 1184 na 318 5009 868 1368 615 995 836

minimum 5448 014 769 5009 168 na 619 1060 391 987 751

maximum 5670 049 1579 010 603 na 1148 1689 807 1003 878

1s 076 015 334 na 151 na 189 261 167 06 45

vr19674 ecl-g7 5322 010 185 5009 536 015 1448 2339 062 992 828 volatile-rich

vr67112b 5040 041 328 5009 1079 024 1108 2229 071 992 647 volatile-rich

yk1911 5231 014 301 5009 544 024 1450 2340 056 996 826 volatile-rich

Pyroxenites 5527 na 347 038 304 na 1543 1959 245 999 899

minimum 5299 5006 037 010 149 5009 1128 1408 030 990 804

maximum 5666 037 786 134 490 014 1863 2312 537 1009 947

1s 068 na 189 028 091 na 193 226 131 05 37

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO NiO Total Mg-no

Opx

Pyroxenites 5781 na 059 na 627 na 3508 037 011 1006 909

minimum 5546 5006 032 5009 462 5009 2999 020 008 996 804

maximum 5888 009 318 054 1304 014 3628 077 014 1014 933

1s 098 na 067 na 212 na 160 019 002 06 32

Averaged for the three eclogite types and for pyroxenites minimum and maximum values as well as standard deviationsalso given Concentrations in volatile-rich eclogites are shown individually ecl eclogite Mg-numberfrac14 100Mg(Mgthorn Fe)Temperatures of last equilibration [T(Krogh)] as in Electronic Appendix 8

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1850

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

garnets (these include kyanite- graphite- and diamond-bearing varieties) (2) eclogites with low-Mg garnets (theseinclude quartz-bearing eclogite) (3) eclogites with high-Mggarnets and Cr2O3 contents502wt (4) volatile-richeclogites (phlogopite- apatite- carbonate-bearing)

Clinopyroxene (cpx)Clinopyroxenes in all eclogite groups have higher averageAl2O3 contents than those in pyroxenites with the formerhaving about 90 of the total Al in the jadeite molecule(AlVI) versus about 35 for the latter corresponding toan average of 41 and 25wt Na2O respectivelyThe highest Al2O3 and Na2O contents in the dataset areobserved in cpx from the high-Ca eclogites (up to158wt Al2O3) (Electronic Appendix 3) Contents ofK2O are generally below the detection limit(5004wt ) but values up to 024wt are observed insome cpx in high-Ca eclogites Average cpx Mg-numberare highest in pyroxenites (899) followed by cpx inhigh-Mg eclogites (865) high-Ca eclogites (836) andlow-Mg eclogites (801) Pyroxenitic cpx tend to havelower TiO2 and higher NiO and distinctly higherCr2O3 contents compared with those in eclogites A plotof MgO vs Al2O3 shows a negative correlation (Fig 6)Clinopyroxenes in low-Mg eclogites have lower Al2O3

contents at a given MgO content than other eclogitetypes Between 15 and 17wt MgO the Al2O3^MgOcorrelation for cpx in pyroxenites has a different slopefrom that in the eclogites Clinopyroxenes in pyroxenites

VR19673ecl-g1 and VR19673ecl-g2 have distinctly lowerMgO contents and higher Al2O3 contents than inthe other pyroxenites They coexist with garnets havingthe lowest Cr2O3 and CaO contents of the dataset whichdefine one end of the positive Cr2O3^CaO correlationobserved for garnetsClinopyroxenes in volatile-rich eclogites have distinctly

high MnO contents (up to 024wt ) Mg-numbers forcpx in the phlogopite-bearing variety are 826 and 828whereas that in the calcite-bearing eclogite (VR67112b)has an Mg-number of 647Some cpx grains have spongy rims that contain small

(mm-scale) patches of glass these rims are richer in Caand Mg and markedly poorer in Na than the cores similarto cpx in eclogites from southern Africa (Taylor amp Neal1989) The rim texture is unequilibrated suggesting lategrowth The formation of the spongy cpx rims might berelated to incongruent melting of cpx or be a result ofmetasomatism (Taylor amp Neal 1989 and referencestherein) Distinct although not spongy rims in two eclo-gites (VR67360 and YK1946) are similarly enriched inCaO and depleted in Na2O Two types of cpx with respectto CaO Na2O and Mg are present in VR43477VR43479 and VR43480 without clear core^rim zonationor spongy rims No significant differences are recognizedwith respect to other oxides

Orthopyroxene (opx)Orthopyroxenes have Mg-number ranging from 804 to933 (17 of 19 samples have values between 895 and 933)and typically contain minor amounts of Al2O3

(032^118wt one outlier has 318wt ) CaO(020^118 wt ) and NiO (008^014wt ) (Electronic

0

2

4

6

8

10

12

14

7 9 11 13 15 17 19

Al 2

O3

CPX

pyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

cpx in pyroxenitesVR19673ecl-g1g2

MgO

Fig 6 MgO vs Al2O3 (wt ) in cpx Data sources as in Figs 3and 4

0

005

010

015

020

2 4 6 8 10 12 14 16

CaO

Na 2

O

dl

group Igroup II

pyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

GARNET

Fig 5 CaO vs Na2O (wt ) in garnet Arrows show trends foreclogites and pyroxenites respectively [note that part of thepyroxenite trend extends into the field falling below the detectionlimit (dl) for Na2O] Group I and group II refer to classificationof McCandless amp Gurney (1989) Data sources as in Figs 3 and 4

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1851

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Appendix 4) The contents of all other elements are mostlybelow their respective detection limits

Accessory phasesRutile and ilmenite occur in some of the eclogites (Table 1)All rutile grains have exsolved 10 ilmenite as lamellaeand rims Ilmenite-free areas of rutile in eclogites containbetween 962 and 988wt TiO2 and minor amounts ofAl2O3 (01^04wt ) Cr2O3 (5009 to 02wt ) andFeO (03^07wt ) Rutile in pyroxenite YK1915 contains004wt Al2O3 01wt FeO and 15wt Cr2O3Discrete ilmenite grains in sample VR67112b haveTiO2 contents of 499 and 500wt MgO of 23 and15wt and FeO of 399 and 403wt respectivelySulfides occur interstitially or enclosed in cpx garnet

and opx they are mostly pyrrhotite and monosulfide solidsolution with variable Ni contents (unpublished data)They will not be considered further hereAbundant platy phlogopite is observed in two samples

(VR19674ecl-g7 and YK1911) It is Cl-rich in sampleVR19674ecl-g7 (063wt vs 003wt in YK1911) andF-rich in sample YK1911 (097wt vs 005wt inVR19674ecl-g7) (Electronic Appendix 5) The same is truefor coexisting apatite (VR19674ecl-g7 254wt Cl and014wt F YK1911 006wt Cl and 306wt F)

Comparison with mineral inclusions indiamonds from Lac de Gras and witheclogite xenoliths from other Slave CratonlocalitiesMineral inclusions in diamonds from the Lac de Graskimberlites have been investigated by Davies et al (19992004) For garnets there is a striking similarity betweenthe inclusions in diamond and those in the high-Ca eclogitexenoliths although the inclusions in diamond trendtowards higher Na2O (Fig 5) and lower Al2O3 contentsThe overlap is not as marked for cpx where inclusionsin diamond span almost the entire range observed foreclogitic cpx (Fig 6)A comparison with eclogites from other localities in

the Slave Craton shows that garnets in most Diavik low-temperature eclogites reported by Schmidberger et al(2007) plot with garnets in high-Mg eclogites and inpyroxenites from this study (Figs 3^5) whereas garnets inhigh-temperature and diamond-bearing eclogites(Schmidberger et al 2007) overlap most with garnets inhigh-Ca eclogites from this study and with garnetsincluded in diamond Eclogites from the Jericho kimberlitehave either massive or anisotropic fabrics (Kopylova et al1999) Garnets in almost all massive eclogites plot withgarnets in low-Mg eclogites from this study with regard toCaO^MgO^FeO relationships (Fig 3) but many havelower Na2O contents at a given CaO content (Fig 5)Garnets in anisotropic eclogites are restricted to lowMgOFeO (507) and show some overlap with garnets

in high-Ca eclogites although trending towards lowerCaO(MgOthornFeO) (Fig 3) while coexisting cpx is alsorestricted to low MgO and trends toward high Al2O3 con-tents (Fig 6)

TRACE ELEMENTSGarnetTrace-element abundances in garnet are summarized inTable 3 and the complete dataset is reported in ElectronicAppendix 1 Garnets in low-Mg eclogites show steep posi-tive slopes between LaN and SmN and shallower positiveslopes in the normalized heavy rare earth elements(HREEN Fig 7a) The normalized light REE (LREEN)show an order of magnitude variability ZrHf is mostlysupra-chondritic and Ti can be enriched or depleted rela-tive to elements of similar compatibility Niobium in threesamples is strongly depleted at 002^003 chondriticGarnets in high-Mg eclogites also show smooth positiveslopes in LREEN but have flat HREEN and variablenormalized middle REE (MREEN) to HREEN (Fig 7b)ZrHf is always supra-chondritic and Ti enriched ordepleted relative to similarly compatible elementsNiobium is most samples is below detection Small positiveEu anomalies though not outside the analytical uncer-tainty for Eu relative to either Sm or Gd are observed forgarnet in some high-Mg eclogites (eg EuEu up to 15for VR67360 where Eu is the average of the chondrite-normalized Sm and Gd concentrations) Garnets in high-Ca eclogites are slightly enriched in the MREEN relative tothe HREEN or have flat MREEN and HREEN with ahump between EuN and HoN (Fig 7c) One sample(VR40345) shows strong La enrichment (greater thanchondrite) Two samples have supra-chondritic ZrHfwhereas three samples have close to chondritic ratios andone sample has a subchondritic ratio As is true for theother eclogite typesTiN is either enriched or depleted rela-tive to similarly compatible elementsPyroxenitic garnets have variable slopes in the LREEN

and flat MREEN to HREEN patterns and two samplesshow high Sc relative to neighbouring elements Garnet inone of the pyroxenites (YK1952) has a highly distinctivepattern relative to other pyroxenites with very low LREEand HFSE (Fig 7d) This sample consists of coarse pyroxe-nite schlieren in which the garnet was analysed in a finer-grained lsquomatrixrsquo of spongy cpx with secondary rims finecpx^opx intergrowths embayed garnet and melt patchesVolatile-rich eclogites have garnets that are slightly

enriched in MREEN relative to HREEN or have flatMREE to HREE patterns One garnet has a pronouncednegative Eu anomaly (EuEufrac14 07 forVR67112b Fig7e)Garnets in volatile-rich samples have the lowest Ni Coand Ti and the highest V ScYand HREE abundances inthe dataset

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1852

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 3 Summary of trace-element contents in garnet and cpx (ppm)

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR19674

ecl-g7

VR67112b YK1911 av min max 1s

n 6 6 7 3 5 4 4

P 169 54 270 88 196 70 310 78 282 80 450 138 160 60 80 214 48 370 144

Sc 62 40 89 16 48 36 67 11 53 42 79 15 69 92 93 186 52 440 179

Ti 1667 490 2500 680 1073 620 2000 455 2141 700 4700 1582 270 690 260 1097 540 1900 573

V 169 78 250 68 150 78 280 72 178 110 390 101 140 340 100 390 71 820 326

Co 59 42 70 11 58 39 84 13 58 47 65 5 30 41 27 68 55 90 16

Ni 17 5 32 10 21 11 34 10 24 11 35 8 6 10 51 54 25 110 38

Ga 12 7 15 3 13 8 16 3 12 9 20 4 8 86 58 12 10 14 2

Sr na 512 4 na na 522 522 na na 57 57 na 52 5025 510 513 513 513 na

Y 37 6 61 18 21 7 38 10 31 20 38 5 60 81 70 19 12 26 6

Zr 15 5 29 9 28 6 83 26 24 9 43 12 60 22 90 18 1 41 17

Nb 009 0005 048 019 na 50015 05 na na 5005 05 na 5003 0012 5001 022 0030 057 025

La 005 001 009 004 na 50007 00 na na 5003 04 na 004 002 0009 003 000 009 005

Ce 03 01 05 02 01 00 01 00 05 01 18 05 03 022 015 01 00 03 01

Pr 01 00 02 01 00 00 01 00 02 01 03 00 017 014 012 00 00 01 00

Nd 12 03 24 07 07 04 14 03 21 12 32 07 24 23 20 03 00 07 03

Sm 11 04 22 06 10 03 18 05 23 08 36 10 29 31 33 05 01 09 04

Eu 06 03 10 03 06 02 10 03 11 04 17 04 16 114 16 03 01 06 02

Gd 27 07 41 12 23 07 42 12 43 16 53 13 8 81 90 13 06 24 08

Dy 54 11 70 23 36 13 65 18 58 35 71 11 12 144 12 29 18 42 10

Ho 14 02 23 07 08 03 15 04 12 09 14 02 24 30 23 07 04 10 02

Er 46 07 80 24 24 09 40 10 34 25 40 05 8 85 6 22 14 28 06

Yb 52 06 98 30 24 10 37 08 32 22 39 05 8 70 5 23 13 30 07

Lu 09 01 16 05 04 02 05 01 05 03 07 01 11 09 08 04 02 05 01

Hf 03 02 05 01 04 01 10 03 05 02 09 02 09 027 007 03 00 06 03

Ta na 50008 5004 na na 50005 001 na na 50004 na na 003 50002 50005 na 50007 011 na

Pb na 5005 02 na na 5002 02 na na 5003 07 na 02 004 5002 na 5003 01 na

Th na 5001 005 na na 50007 003 na na 50008 011 na 5007 0012 50004 002 0005 006 002

U na 5003 5003 na na 5001 01 na na 5007 na na 01 0023 010 na 5001 003 na

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR67112b YK1911 av min max 1s

n 6 4 8 5 4 4

P 65 15 120 41 39 14 56 16 53 30 70 12 20 9 80 27 120 40

Sc 22 10 30 7 15 14 19 2 14 7 24 5 80 43 17 11 21 5

Ti 1557 990 2400 508 1168 1010 1490 191 1988 810 3500 1154 2390 1700 689 102 1685 694

V 415 180 480 107 318 210 400 69 324 259 450 72 600 180 250 139 480 156

Co 35 26 50 8 20 14 31 7 26 14 42 11 49 23 29 20 42 10

Ni 164 87 220 47 228 100 430 128 188 135 320 56 59 34 428 410 670 184

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1853

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

ClinopyroxeneTrace-element abundances in cpx are summarized inTable 3 and the complete dataset is reported in ElectronicAppendix 6 Clinopyroxenes in low-Mg eclogites havemildly positive to negative slopes from LaN to NdN andsmooth negative slopes between SmN and HoN thatmostly continue to LuN (Fig 8a) Niobium is below detec-tion or 5007 chondritic and SrN is variably enrichedrelative to neighbouring elements with the peak size corre-lating negatively with LREEN ZrHf is subchondritic andHfN is enriched relative to SmN whereasTiN is enriched ordepleted relative to similarly compatible elementsWe ana-lysed the spongy rim in cpxYK1943 and its trace-elementpattern is similar to that of cpx cores for more compatibleelements whereas the most incompatible elements areseveral orders of magnitude higher Clinopyroxenes in

high-Mg eclogites have steeper negative slopes betweenNdN and LuN than those in low-Mg eclogites combinedwith lower HREE abundances HfN is depleted relative toSmN (Fig 8b) High-Ca eclogites contain cpx with sinusoi-dal REEN patterns with lower LREE than other eclogitetypes and the heaviest REE below detection They haveuniformly strong positive Sr peaks relative to elements ofsimilar compatibility (Fig 9) The diamond-bearing high-Ca eclogite (VR40345) has cpx with unusually high Nband LREE abundancesClinopyroxenes in four pyroxenites have disparate trace

element patterns They all share depletions in TiN relativeto similarly compatible elements and two show positiveSrN peaks Two samples have cpx with high and flatLREEN to MREEN patterns and steep negative slopes inthe HREEN The trace-element patterns of cpx in YK1914

Table 3 Continued

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR67112b YK1911 av min max 1s

n 6 4 8 5 4 4

Ga 16 10 20 3 12 9 15 2 26 15 39 9 129 123 6 11 11 5

Rb na 5008 04 na na 5007 05 02 na 5003 03 na 02 5004 03 010 05 0

Sr 166 85 290 74 222 153 285 50 243 83 440 142 85 120 200 11 414 209

Y 2 04 4 10 13 07 2 04 11 03 3 10 28 51 2 03 3 2

Zr 19 4 48 15 25 8 45 13 17 8 28 7 111 129 36 008 81 42

Nb 01 0003 03 01 na 5002 00 na na 5002 16 9 007 011 010 003 02 008

Cs na 5001 6 na na 5006 5006 na5013 5004 5013 na bdl 5002 na 5001 5002 na

Ba 09 02 2 07 09 01 2 07 10 03 3 10 06 01 13 06 3 10

La 3 03 8 3 3 23 3 04 4 01 14 6 51 97 4 01 10 4

Ce 7 13 20 6 10 7 13 2 7 05 24 10 25 28 11 01 28 13

Pr 11 02 3 08 2 10 3 06 08 02 2 09 47 46 2 002 4 2

Nd 5 12 12 4 10 5 14 4 4 09 8 26 23 22 8 01 22 10

Sm 11 05 21 06 20 10 31 10 07 02 11 03 47 53 2 005 5 2

Eu 03 01 05 01 06 028 09 03 02 010 03 007 087 123 06 002 15 07

Gd 08 03 13 03 11 05 19 06 05 03 06 011 28 37 15 010 36 2

Dy 05 01 07 02 04 022 06 01 02 013 03 007 11 16 07 008 14 06

Ho 01 00 02 004 006 003 007 002 na 5003 012 na 011 021 009 001 02 007

Er na 507 507 na 011 004 02 00 na 5006 03 na 019 033 02 002 03 012

Yb na 5001 02 na na 5001 011 na na 5006 03 na 007 015 009 001 02 006

Lu na 5001 003 na na50002 002 na na50008 01 na 0006 0013 0013 0002 003 001

Hf 12 03 24 07 11 04 18 05 10 06 16 04 34 53 na50015 35 na

Ta na50003 50015 na na50006 006 na na50004 5004 na 0014 00236 na50003 503 na

Pb 06 02 12 04 06 04 09 02 na 5016 07 02 065 098 08 01 18 1

Th na50005 06 na 012 005 02 005 na 5002 5006 na 021 02 001 06 03

U na 5005 5005 na 002 0009 005 001 na 5003 5005 na 0013 024 na50005 003 na

Averaged for the three eclogite types and for pyroxenites as in Table 2 pyx pyroxenite n number of analyses averagedbdl below detection limit

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1854

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

and YK1952 are similar although at different abundanceswith only modest LREEN over HREEN enrichment andstrong depletions in ZrN relative to neighbouring elements(Fig 8d)Clinopyroxenes in the volatile-rich samples have flat pat-

terns and high abundances from LaN to SmN and smoothnegative slopes from SmN to LuN Niobium Sr and Ti are

noticeably depleted relative to elements of similar compat-ibility (Fig 8e)

Trace-element partitioningTo assess whether coexisting phases are in trace-elementequilibrium we calculated apparent cpx^garnet distribu-tion coefficients and compared them with those

(a)

(d)

(b)

(e)

(c)

GarnetCI-chondrite GarnetCI-chondrite

low-Mghigh-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1952

1

10

Nd Sm Eu Gd Dy

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Fig 7 Extended trace-element patterns of garnet in (a) low-Mg eclogites (b) high-Mg eclogites with inset showing enlargement of MREE tohighlight Eu anomalies and 1s uncertainties (c) high-Ca eclogites (d) pyroxenites and (e) volatile-bearing eclogites Normalized to chondritevalues of McDonough amp Sun (1995)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1855

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

determined experimentally or in demonstrably equili-brated natural samples Low-Mg and high-Mg eclogiteshave Dcpxgarnet for most elements that are within therange determined for natural equilibrated eclogites andexperimental assemblages (Harte amp Kirkley 1997 Greenet al 2000 Barth et al 2002a) (Electronic Appendix 7)High-Ca eclogites plot at the low end of the range

of distribution coefficients and have very low DMREE

and DHREE outside the range indicated by the datasetof Harte amp Kirkley (1997) who noted that DREE decreaseswith increasing Ca content in eclogitic minerals Theseassemblages are therefore considered to be in equilibriumPyroxenite VR40384 has D values similar to eclogites

but at the high end of the range determined by Harte amp

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

(a)

(d)

(b)

(e)

(c)

CpxCI-chondrite CpxCI-chondrite

low-Mg high-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1914

YK1952

spongy rim

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

VR40345

Fig 8 Extended trace-element patterns of cpx as in Fig 7

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1856

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Kirkley (1997) (Electronic Appendix 7) Another group(YK1914 and YK1952) plots at the low end but DHf DZrDTi and DV are far below the values indicated by experi-mental datasets Because pyroxenites compositionallygrade into eclogites their drastically different DHFSE

values suggest disequilibrium In contrast volatile-bearingsamples have DHFSE values that appear to be too high to bein equilibrium however this may be a consequence oftheir unusual bulk chemical composition (Ti- and K-rich)which differs substantially from that of lsquonormalrsquo eclogitesand experimental charges

GEOTHERMOBAROMETRYFollowing Schmidberger et al (2007) temperatures oflast equilibration were calculated for eclogitic andopx-free pyroxenitic assemblages using the formulationof Krogh (1988) The results are summarized in Table 2and Electronic Appendix 8 for a preset pressure of5 GPa and illustrated as a histogram in Fig 9 Foropx-bearing pyroxenites temperatures were calculatedsimultaneously with pressure [Ca-in-opx thermometryand Al-in-opx barometry using the formulations of Breyet al (1990)] to constrain their relative positions in themantle columnMinerals in some eclogites from Lac de Gras show dis-

tinct core and rim compositions or inhomogeneitiesTemperatures were not calculated for samples withspatially irresolvable inhomogeneous garnet compositions(SE01VR43479VR43480) For eclogites with spongy cpxcore compositions were used in the calculationTemperatures were calculated for both core and rim com-positions where applicable For all samples these

temperatures agree within the uncertainty of this thermo-meter (508C Brey amp Kolaquo hler 1990)Most equilibration temperatures for eclogites lie between

800 and10008C (Fig 9)The transition between the shallowand the deep lithospheric mantle layers in the central Lacde Gras area lies in this temperature interval (Griffin et al1999 Pearson et al 1999) Eclogites in this study lack thedistinct bimodal temperature distribution observed bySchmidberger et al (2007) but as in their study high-Caeclogites give higher average temperatures (10808C corre-sponding to a depth of 155 km if equilibrated along a 40mWm2 geotherm) than other eclogites types (9808Cdepth 135 km) the diamond-bearing sample (VR40345)gives thehighest temperature of the datasetHighNa2Ocon-tents in garnet from high-Ca eclogites are qualitativelyconsistent with higher equilibration pressures (Sobolev ampLavrentrsquoyev 1971) as are high K2O contents in coexistingcpx (Erlank ampKushiro1970) suggesting that high-Ca eclo-gites not only equilibrated at high temperature but also highpressure rather than representing a temperature excursionPyroxenites both opx-bearing and opx-free give somewhatlower average temperatures than eclogites (mean 9258C)and phlogopite-rich eclogites give the lowest temperatures(685 and 7698C respectively) The calcite-bearing eclogiteVR67112b gives a relatively low equilibration temperatureof 8558C

MEASURED ANDRECONSTRUCTEDWHOLE -ROCKSRationale and approachKimberlite-borne xenoliths sometimes show evidence forinfiltration of the host kimberlite (eg Taylor amp Neal1989 Barth et al 2001 2002b Heaman et al 2006) Wetherefore follow the standard approach of calculatingwhole-rock compositions from the compositions of the con-stituent phases weighted by their modal abundances Onlycore compositions were used as distinct rim compositionsmay point to late diffusion of elements from kimberlite-related melt veinsMany eclogitic whole-rock compositions reconstructed

with only garnet and cpx have negative Ti anomalies(not shown) For most of these samples rutile or ilmenitehas been identified as part of the mineral assemblage andthese phases should therefore be considered in the whole-rock reconstruction Rutile or ilmenite modes were esti-mated by assuming that the bulk-rock has no Ti anomalywhen normalized to a normal mid-ocean ridge basalt(N-MORB) composition and calculating the amount ofrutile required to erase theTi anomalyIt is of course possible that the eclogites had negative

or positive Ti anomalies relative to elements of similarcompatibility (Eu and Dy) Because we reconstructed

0

2

4

6

8

10

12volatile-rich

pyroxenite opx-bearing

pyroxenite opx-free

high-Ca

high-Mg

low-Mg

700750

800850

900950

10001050

11001150

12001250

Fre

quen

cy

Temperature

Fig 9 Histogram ofTKrogh (Krogh 1988) obtained at a preset pres-sure of 5 GPa except opx-bearing pyroxenites where TBKN wassolved simultaneously with PBKN (Brey amp Kolaquo hler 1990)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1857

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

whole-rock compositions from mineral core compositionskimberlite infiltration is unlikely to have caused fractiona-tion of Ti from Eu and Dy In a study in which rutilemodes were estimated from the differences between Ti inmeasured bulk-rocks and in bulk-rocks reconstructed withonly garnet and cpx only three of 19 eclogites recon-structed with rutile show Ti anomalies (Barth et al2002a) Even eclogite xenoliths with depleted Zr and Hfabundances that resemble those of modern island arcbasalts do not show Ti anomalies (Jacob amp Foley 1999)We therefore consider our approach to be a reasonablyrobust means of obtaining meaningful rutile modes wheresmall sample sizes preclude reliable determination of themodes of accessory phases by point-countingWhere rutilewas not recognized on thin-section scale or not analysedthe average eclogitic rutile composition was used for thereconstruction

UncertaintiesReconstructing whole-rock compositions in this way is arather crude method given the coarse grain size of theeclogites relative to the sample size and the non-uniformdistribution of the principal minerals on thin-sectionscale The uncertainty in mineral modes of medium- tocoarse-grained eclogites with grains45mm has been esti-mated by Jerde et al (1993) to be 10 plus uncertaintiesrelated to partial alteration which is minor in our samplesNevertheless the whole-rock trace-element budget israther insensitive to the exact modes for common eclogiticassemblages and REE patterns which encompass mostother elements in terms of compatibility in garnet andcpx do not change significantly even for modal variationsof 30 (Jerde et al 1993) This is not true for major-element concentrations which vary substantially with var-iations in mineral modes and we therefore do not calculatewhole-rock major-element contentsBecause the discussion that follows is based on trace-

element patterns and trends rather than absolute abun-dances whole-rock reconstruction based on mineralmodes is adequate We are confident that the uncertaintyinherent in this approach is not leading to significant mis-representation of eclogite whole-rock compositions becausethe trace-element patterns of the reconstructed whole-rocks show distinct differences (see below) that correlatewith the assignment to eclogite types based on garnetmajor-element composition and this is independent ofmineral mode Following Jerde et al (1993) in ElectronicAppendix 9 we show variations in trace-element patternfor one sample in each eclogite type for variations ingarnet and cpx abundance from 7030 to 3070 vol This shows that the trace-element patterns are generallyinsensitive to cpxgt ratios with regard to slopes andanomalies such as positive Sr Pb and Eu anomalies withthe exception of the positive Eu anomaly in high-Mgeclogite VR43479 which disappears if a high cpx mode is

assumed (not shown) Variations in rutile modes by 50(Electronic Appendix 9d) lead as would be expected tothe appearance of Ti anomalies but a rutile mode of 05instead of 1 vol does not erase the positive Nb anomalyin low-Mg eclogiteVR50909Because of the robustness of the relative trace-element

abundances in the reconstructed eclogites a garnet propor-tion of 065 and cpx proportion of 035 approximately themedian modal abundances of these minerals in volatile-free eclogites from Lac de Gras are assumed for eclogitesamples that are too small for determination of meaningfulmodes by point-counting

Within- and between-group variationsThe trace-element compositions of reconstructed whole-rocks are given in Table 4 N-MORB-normalized trace-element patterns of reconstructed whole-rocks areshown in Fig 10 The three eclogite groups have distincttrace element patterns that are particularly obvious forthe REE (Fig 11) Low-Mg eclogites have positive Pb andSr anomalies and smooth positive slopes in their MREEN

and HREEN LREEN are enriched relative to MREEN inthree samples (Fig 11a) Three samples have high Nbrelative to similarly compatible elements and comparedwith most other eclogites (Fig 10a) All high-Mg eclogitesdisplay more or less pronounced positive Eu and Sranomalies and most have positive Pb and negative Zrand P anomalies Their N-MORB-normalized REE pat-terns show shallow positive slopes from LaN to PrN orNdNand aweak negative slope from NdN to LuN (Fig11b)VR43479 is a low-Mg eclogite with regard to its garnet

Ca^Fe^Mg composition but has a trace-element patterntypical of high-Mg eclogites and will be discussed withthe latter Conversely sample YK1949 a high-Mg eclogitein terms of its garnet Ca^Fe^Mg compostion is more simi-lar to low-Mg eclogites in terms of its whole-rock trace-element pattern and is plotted in Figs 10a and 11aWhereas the patterns of most samples within the groupresemble each other the N-MORB-normalized incompati-ble-element abundances of the high-Ca eclogites vary bymore than an order of magnitude (Fig 10c) All sampleshave positive Pb anomalies and most have positive Sranomalies Some samples have negative P andor Zranomalies Niobium abundances in three of the samplesare very high similar to those for some low-Mg eclogitesHigh-Ca eclogites tend to have flatter MREEN andHREEN patterns and more uniform HREE abundancesthat the other eclogite types with humps between SmN

and HoNThe phlogopitized sample YK1911 shows a strong posi-

tive Pb anomaly and lower HREE abundances than thecalcite-bearing sample VR67112b which shows a distinctnegative Eu anomaly and a depletion of La relative to Ce(Figs 10d and 11d)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1858

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Trace-element compositions of reconstructed whole-rocks (ppm)

Sample P Sc Ti V Co Ni Ga Sr Y Zr Nb La Ce Pr Nd

Low-Mg eclogites

vr19674-g4 186 50 4649 317 57 59 144 77 24 30 36 13 31 05 24

vr43452 154 39 2253 328 49 84 176 52 32 11 01 02 07 01 08

vr43469 38 65 4377 349 54 91 142 34 28 15 02 05 15 03 17

yk1926 144 45 3406 258 52 87 145 89 18 18 20 09 29 06 33

yk1943 173 46 5732 295 41 39 146 122 25 48 82 34 85 13 64

yk19491 102 34 1693 158 31 49 100 46 10 14 03 09 21 03 17

High-Mg eclogites

vr19677-g2 138 39 5160 291 41 53 119 98 14 22 04 13 53 11 64

vr434792 46 25 1265 129 36 126 84 56 3 7 04 06 25 04 23

vr67360 49 36 1573 138 64 184 111 76 5 9 04 09 38 07 33

yk1946 153 28 4877 192 45 105 124 108 12 48 03 10 38 08 50

High-Ca eclogites

vr19674-g5 267 30 7663 200 41 67 195 32 23 43 77 01 04 02 16

vr40345 0 33 2821 175 47 98 210 179 13 13 66 60 109 11 38

vr43465 142 27 5104 220 31 83 121 197 14 35 29 05 22 05 40

vr50858 295 30 6898 201 44 94 150 48 22 38 06 01 03 02 20

vr50860 68 57 5672 418 57 71 186 109 19 11 02 02 13 03 26

vr50909 238 28 7295 201 42 92 171 48 20 57 13 01 04 02 21

yk3528 91 51 3712 341 53 155 177 49 16 9 01 02 08 02 11

Volatile-rich eclogites

vr67112b 40 84 9242 521 46 30 107 4 40 66 31 26 126 24 125

yk1911 31 39 9428 286 44 82 164 3 23 52 64 18 52 09 46

Pyroxenites

yk1914 111 49 707 207 48 547 44 31 4 3 02 06 19 03 10

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

Low-Mg eclogites

vr19674-g4 11 05 10 22 39 10 29 33 06 12 05 028 24 008

vr43452 04 02 08 13 38 12 41 49 08 05 01 nd 02 na

vr43469 11 05 09 25 44 11 34 36 06 07 02 001 09 129

yk1926 12 05 11 19 30 07 22 24 04 07 03 004 23 090

yk1943 21 08 10 29 42 10 30 32 05 16 05 013 65 017

yk19491 06 03 11 08 14 04 14 18 03 04 02 003 29 115

High-Mg eclogites

vr19677-g2 20 08 12 21 24 05 15 15 03 05 05 007 52 6811

vr434792 05 02 12 05 06 01 04 03 00 03 01 001 98 037

vr67360 06 03 14 06 09 02 06 06 01 02 02 002 86 000

yk1946 18 07 12 21 22 05 13 12 02 11 03 006 56 112

High-Ca eclogites

vr19674-g5 17 09 11 33 43 09 24 23 04 16 05 007 02 011

vr40345 09 04 10 12 22 06 16 14 02 05 nd nd 298 na

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1859

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Continued

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

vr43465 22 09 12 25 26 06 15 14 02 10 04 001 25 040

vr50858 20 08 09 33 41 08 22 21 03 10 01 nd 02 na

vr50860 14 07 11 27 37 08 22 18 03 06 00 nd 06 na

vr50909 20 10 12 33 38 07 20 20 03 16 01 002 02 181

yk3528 08 05 11 20 30 07 18 18 02 05 01 nd 08 na

Volatile-rich eclogites

vr67112b 38 10 07 53 74 15 41 34 04 18 03 012 60 004

yk1911 20 07 08 35 39 07 18 17 02 10 14 nd 73 na

Pyroxenites

yk1914 02 01 11 04 08 02 05 06 01 01 03 003 68 101

aGarnet has CandashMgndashFe relationship typical of high-Mg eclogites but whole-rock trace-element pattern is typical oflow-Mg eclogitesbGarnet has CandashMgndashFe relationship typical of low-Mg eclogites but whole-rock trace-element pattern is typical ofhigh-Mg eclogites

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

(a)

001

01

1

10(c)

001

01

10(b)

VR43452

YK1949

RockNMORB

low-Mg

RockNMORB

high-Mg

(d)

volatile-richhigh-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

phlogopite-bearingYK1911

calcite-bearingVR67112b

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 10 Extended trace-element patterns of reconstructed (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-bearing eclogites Normalized to N-MORB of Sun amp McDonough (1989)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1860

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

RockNMORB

(b)

RockNMORB

(d)

010

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu001

01

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

(a)

001

01

1

10(b)

001

01

1

10(c)

001

01

1

10(e)

high-TSchmidberger et al

low-TSchmidberger et al

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

high-Mglow-Mg

high-Cavolatile-rich

phlogopite-bearingYK1911

calcite-bearingVR67112b

YK1949VR43479

Fig 11 N-MORB normalized REE patterns for (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-richeclogites Shown for comparison in (e) are reconstructed eclogites from Schmidberger et al (2007)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1861

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Both volatile-rich whole rocks display strong negativeSr anomalies Strontium can be sited in calcite or apatiteandor indicate the presence of a Sr-rich mineral that wasnot identified in thin section such as bariteFour of the six reconstructed whole-rocks in the study

of Schmidberger et al (2007) (Fig 11e) show markedLREEHREEN enrichment that is observed only in onehigh-Ca eclogite in our study (Fig 11c) suggesting thatmany of the eclogites they studied were affected by metaso-matic LREE enrichment Two samples in their study thatare not affected by this enrichment show positive Euanomalies similar to the high-Mg eclogites in this studywhereas the distinctive REE pattern of the low-Mg eclo-gites is not recognized among their samples

SR^ND^HF I SOTOPESYSTEMATICSStrontium Nd and Hf isotope data for optically puremineral separates and reconstructed whole-rocks aregiven in Table 5 Spongy rims and alteration are likely tobe mostly removed during leaching in 6N HCl in combina-tion with repeated ultrasonication Despite this rigoroustreatment the reproducibility of Nd in garnet VR43469and in cpx YK1914 is worse than for other samples Thismay be due to inhomogeneities related to late Nd diffusionpossibly from grain boundaries containing a kimberlitecomponent The reproducibility of the 176Hf177Hf data isgenerally poorer than that of Nd which may reflect a com-bination of matrix effects caused by imperfectly purifiedsolutions and sample heterogeneity Parent^daughterratios are available only from in situ analyses which arenot precise enough to allow meaningful model age orinitial isotope ratio calculations (ie at the time of kimber-lite eruption) Because the Lac de Gras kimberlites arerelatively young (55^56 Ma Graham et al 1999) age cor-rection would result in only a minor change in isotopecomposition which is inconsequential in view of the rangeof compositions displayed by most samples and hence hasno effect on the conclusions drawnClinopyroxenes in five volatile-free eclogites and a

pyroxenite have Sr isotope ratios ranging from 0701662 to0703546 this range is similar to or somewhat lower thanthat of the depleted mantle at the time of kimberliteemplacement (0703 Bell amp Tilton 2002) Clinopyroxenein the phlogopitized sample has a distinctly high 87Sr86Srof 0751186Clinopyroxene controls 56^92 of the Nd budget (Nd

concentration in cpx weighted by mode) Neodymiumisotope ratios range from 0511660 to 0514844 Thiscorresponds to eNd variation (per 10 000 deviationof 143Nd144Nd in the sample from 143Nd144Nd in the chon-dritic mantle model reservoir) from 17 to thorn43 cpx inthe low-Mg eclogites have higher eNd than those in

high-Mg eclogites Garnets have generally have higher143Nd144Nd from 0512984 (eNdfrac14thorn 7) to 0520965(eNdfrac14thorn 162) Garnets in the two volatile-rich sampleshave the highest ratios 0522790 for garnet in the calcite-bearing sample VR67112b (eNdfrac14thorn198) and 0525721 inthe phlogopitized eclogiteYK1911 (eNdfrac14thorn 255)Garnet contributes 18^32 and cpx 48^78 of the Hf

budget The proportion controlled by rutile is 4^30(Aulbach et al in preparation) Measured 176Hf177Hf incpx ranges from 0281263 in phlogopite-bearing sampleYK1911 (eHffrac1453) to 0286130 in low-Mg eclogiteVR43469 (eHffrac14thorn119) and in garnet from 0283958 inpyroxeniteYK1915 (eHffrac14thorn42) to 0295718 in low-Mg eclo-gite YK1926 (eHffrac14thorn458) Hafnium in garnet in the phlo-gopite^apatite-bearing sample is extremely radiogenic(176Hf177Hffrac14 0354367) corresponding to an eHf of thorn2532which is close to a value determined for a garnet separatefrom a Roberts Victor eclogite (eHffrac14thorn2561 Jacob et al2005) Garnet in the calcite-bearing sample has extremelyradiogenic Nd and Hf (eNdfrac14thorn193 eHffrac14thorn2516) Garnetin pyroxenite YK1915 also has radiogenic Nd and Hf(eNdfrac14thorn27 eHffrac14thorn42)Despite the large uncertainties for reconstructed whole-

rocks (calculated as the square root of the sum of squareduncertainties of two standard errors for garnet and cpxanalyses combined with 20 propagated uncertainty onthe proportion of Nd or Hf respectively contributed byeach mineral) the isotopic data are consistent in thatthey correlate with the independently determinedparent^daughter ratios (1) Low-Mg eclogites have higherSmNd reflected in the steeper positive slope of theirN-MORB-normalized REE patterns (Fig 11a) comparedwith the high-Mg eclogites (Fig 11b) Accordingly recon-structed low-Mg eclogites have higher present-day eNd

(0 to thorn47) than high-Mg eclogites (4 and 17) (2) Low-Mg eclogite YK1943 has very high Nb and LREE abun-dances that could have been added during metasomatismthat also enhanced its Hf abundance This sample has thelowest eNd and eHf of the low-Mg eclogites (0 and thorn20respectively) In contrast apparently unenriched low-Mgeclogites have more radiogenic Nd and Hf (eNdfrac14thorn47and thorn9 eHffrac14thorn128 and thorn113 respectively)The Sr^Nd isotope diagram reveals that eNd in cpx

varies from 19 to 43 at nearly constant depleted-mantle-like 87Sr86Sr (Fig 12a) In the Nd^Hf isotope diagram(Fig 12b) the reconstructed whole-rocks lie on curvedmixing lines if the lsquomixingrsquo proportions of Nd and Hffor the two lsquoend-memberrsquo minerals are very different(eg YK1911) Garnets have more radiogenic Nd andHf than coexisting cpx All reconstructed volatile-freeeclogites plot in the radiogenic field with regard to Hfor are isotopically similar to the depleted mantle Twohigh-Mg eclogites and one low-Mg eclogite have unradio-genic Nd one high-Mg eclogite has radiogenic Nd and

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1862

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 5 Individual and averaged Sr Nd and Hf isotope ratios obtained from garnet and cpx solutions

Sample 87Sr86Sr 2SE2SD 143Nd144Nd 2SE2SD eNd 2SE2SD 176Hf177Hf 2SE2SD eHf 2SE2SD

Low-Mg eclogites

VR43469 cp 0702661 0000010 0514844 0000006 43 01 0286130 0000012 119 04

VR43469 gt 0515508 0000014 56 03 0286506 0000012 132 04

VR43469 gt r 0515453 0000010 55 02 0286548 0000011 134 04

VR43469 gt r 0515384 0000007 54 01 0287244 0000010 158 04

VR43469 gt r 0515487 0000014 56 03 0286996 0000014 149 05

VR43469avg gt 0515458 0000109 55 21 0286823 0000715 143 253

WR VR43469 0515059 0000273 47 5 0286405 0001079 128 38

YK1926 cp 0703546 0000011 0513043 0000007 8 0 0283154 0000009 13 03

YK1926 gt 0513364 0000008 14 01 0295718 0000005 458 02

WR YK1926 0513120 0000026 9 1 0285980 0000716 113 25

YK1943 cp 0512143 0000007 10 01 0282213 0000005 20 02

YK1943 gt 0514461 0000012 36 02 0287510 0000008 168 03

YK1943 gt r 0514491 0000009 36 02 0287332 0000005 161 02

YK1943 gt r 0514517 0000010 37 02 0287290 0000007 160 02

YK1943avg gt 0514489 0000056 36 11 0287377 0000233 163 82

WR YK1943 0512625 0000228 0 5 0283138 0000879 13 31

High-Mg eclogites

VR43480 gt 0520965 0000017 162 03 0284950 0000010 77 04

VR67360 cp 0283509 0000074 26 26

VR67360 cp r 0283433 0000031 23 11

VR67360avg cp 0701662 0000018 0511660 0000010 19 02 0283471 0000108 25 38

VR67360 gt 0512984 0000101 7 20 0287429 0000180 165 64

WR VR67360 0511761 0000132 17 3 0284734 0000469 69 17

YK1949 cp 0703019 0000013 0512231 0000009 8 02 0282513 0000009 9 03

YK1949 gt 0513511 0000010 17 02 0290297 0000042 266 15

YK1949 gt r 0513508 0000012 17 02 0289947 0000015 254 05

YK1949 gt r 0513552 0000014 18 03

YK1949avg gt 0513523 0000049 17 10 0291569 0000494 311 175

WR YK1949 0512446 0000103 4 2 0284622 0000847 65 30

Volatile-rich eclogites

VR67112b gt 0522790 0000008 198 02 0354367 0000036 2532 13

YK1911 cp 0751186 0000009 0512165 0000007 9 01 0281263 0000009 53 03

YK1911 gt 0525721 0000013 255 03 0288051 0000012 187 04

WR YK1911 0513940 0000342 25 7 0281414 0000018 48 1

Pyroxenites

YK1914 cp 0703015 0000034 0513283 0000041 13 08

YK1914 cp r 0702956 0000005 0513200 0000027 11 05

YK1914 cp r 0513527 0000059 17 11

YK1914 cp r 0513045 0000010 8 02

YK1914avg cp 0702986 0000083 0513264 0000402 12 78

YK1915 gt 0514003 0000011 27 02 0283958 0000021 42

Analyses of full procedural replicates are designated with an lsquorrsquo after sample name SE standard error (internal errorsingle analyses) SD standard deviation (external error averaged multiple analyses) Italics used for reconstructed whole-rocks (mineral isotopic compositions and concentrations weighted by modes) Uncertainties for reconstructed whole-rocksinclude 2SE or 2SD uncertainties on each mineralrsquos isotope composition combined with 20 uncertainty in the amountof Nd or Hf contributed by the constituent minerals eNd and eHf are per 10 000 deviation from model chondritic mantle[model of Wasserburg et al (1981) and Blichert-Toft amp Albarede (1997) respectively]

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1863

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

εHf

εNd

Ekati eclogitesminus100

0

100

200

300

400

500

minus40 0 40 80 120 160 200 240 280

gt in YK1915

gt in VR67112bεNd=198 εHf=2532

minus40

0

40

80

120

160

minus40 minus20 0 20 40 60

minus60

minus40

minus20

0

20

40

070 071 072 073 074 075 076

North Americankimberlite field

εNd

87Sr86Sr

cpx in YK1914

Ekatieclogites

DM

DM

DM

DM

(a)

(b)

pyroxenites

low-Mg

high-Mg

volatile-rich

eclogites

Diavik low-TA154 S kimberlite

cpx in YK1911

Fig 12 (a) Sr^Nd for cpx and (b) Nd-Hf isotope diagram for garnet cpx and reconstructed whole-rocks (symbols with bold outline) Finedashed lines in (b) connect data points from the same sample with garnets having more radiogenic isotopic compositions Uncertainties forreconstructed whole-rocks combine 20 modal uncertainty with uncertainties on each mineralrsquos isotope composition Error bars on individualminerals may be smaller than the symbols Also shown are the isotopic compositions of the depleted mantle (DM DePaolo 1981 Griffin et al2000 Bell amp Tilton 2002) Range of 87Sr86Sr in eclogites from Ekati (Lac de Gras area 100 km north of Diavik Jacob et al 2003) in NorthAmerican kimberlites and compositions of low-temperature eclogites from Diavik and of Lac de Gras A154 South kimberlite fromSchmidberger et al (2007)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1864

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

one low-Mg eclogite has eNd similar to the depleted mantle(Fig 12b)Clinopyroxene in the phlogopitized eclogite YK1911

has unradiogenic Nd and Hf and strongly radiogenicSr whereas the coexisting garnet has radiogenic Nd andHf Although this sample contains 50 vol phlogopiteits whole-rock isotopic composition was calculated

from garnet and cpx alone because the contributionfrom phlogopite to the Hf and Nd budget is 52 and51 respectively (unpublished data) The whole-rockhas radiogenic Nd but has lower eHf than anyother sample analysed in accord with the extreme Hfabundance (53 ppm) and low LuHf of the cpx

DISCUSS IONOrigins of eclogites beneath the SlaveCratonHigh-Mg and high-Ca eclogites

With regard to their major- and trace-element contentsthe high-Mg and high-Ca eclogites from Lac de Gras arebroadly similar to variably evolved oceanic gabbros(Fig 13b and c) which may indicate an origin as subductedoceanic crust Compatible trace-element abundancesare useful to deduce potential crustal protoliths as theseelements are less susceptible to post-formation modification(Jacob 2004) The flat HREE patterns in the high-Mg andhigh-Ca eclogites are compatible with a low-pressureorigin as the oceanic crust is a product of moderate degreesof partial melting mainly within the spinel stability field(eg Presnall et al 2002) whereas melting with residualgarnet would qualitatively be expected to produce stronglyfractionated REE patterns Positive Sr and Pb anomaliesalthough compatible with a plagioclase signature couldalso be due to cryptic metasomatism (Dawson 1984) ofthe eclogites during their residence in the lithosphericmantle These anomalies are probably not related to kim-berlite infiltration as the whole-rock eclogite compositionswere reconstructed from analyses of fresh mineral coresLow ZrSm similar to that in primitive gabbros is

observed for many of the high-Mg and high-Caeclogites (Fig 13b and c) and is independent of the ratioof garnetcpx and whether or not the whole-rock wasreconstructed with rutile (Electronic Appendix 10) Thenegative trend of Mg-number against Yb in garnet(Fig 14) argues against igneous processes involving garnetaccumulation (Taylor amp Neal 1989) as suggested forexample for some high-Mg eclogites from West Africa(Barth et al 2002b) Rather the anti-correlation suggeststhat olivine fractionation played a role in the formation ofthe protoliths Fractionating olivine would excludeYb (Dolivinebasalt

frac14 002 Nikogosian amp Sobolev 1997)and incorporate Mg relative to Fe (Dfrac1415 and 05 respec-tively Taura et al1998) leading to progressive depletion ofMg and enrichment of Yb in the residual melt and henceanti-correlated Mg and Yb in the whole-rock This isinherited after eclogitization All eclogite types (exceptVR43452) plot along the olivine fractionation trend sug-gesting that they formed by a similar processHigh-Mg eclogites have lower

PREE than high-Ca

eclogites and some appear to have positive Eu anomalies

high-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

gabbrosSE Indian ridge

gabbrosOman ophiolite

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

high-Mg

RockNMORB(a)

VR43452

YK1949

low-Mg

boninite-like melts

01

001

1

10

(b)

01

001

1

10

(c)

01

001

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

Fig 13 Trace-element patterns of (a) low-Mg (b) high-Mg and(c) high-Ca eclogites compared with boninite-like volcanic rocksfrom the Isua greenstone belt (Polat et al 2002) gabbros from theOman ophiolite (Benoit et al 1996) and the SE Indian Ridge (Hartet al 1999) normalized to N-MORB of Sun amp McDonough (1989)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1865

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(EuEu up to 14) similar to more primitive cumulates(troctolite gabbros and olivine gabbros) reported in the lit-erature (Fig 13b) Although the sizes of these Eu anomaliesare not outside the analytical uncertainty their correlationwith CeYb in reconstructed eclogites supports their signif-icance (r2frac14066 nfrac14 4) The more primitive nature ofhigh-Mg eclogites is consistent with higher Mg-numbersrelative to high-Ca eclogites and may be due to olivineaccumulation during formation of their protolithsThe two samples with the lowest HREE abundanceshave the highest EuEu which is compatible with varia-tions in plagioclase modesMore evolved protoliths with higher pyroxeneplagio-

clase plus trapped melt could account for small or absentpositive Eu anomalies and higher

PHREE of high-Ca

eclogites (Fig 13c) Although the reason for the frequentassociation of graphite or diamond with high-Ca eclogitesis unclear (Aulbach et al 2003) the frequent occurrence ofkyanite in high-Ca eclogites also observed at Ekati justnorth of Lac de Gras has been argued to support a crustalprotolith (Jacob 2004) Kyanite andor corundum havebeen suggested to exsolve from high-temperature high-Alclinopyroxenes (Caporuscio amp Smyth 1990) Eclogitesfrom Lac de Gras show no exsolution textures suggestiveof such an origin although recrystallization can obscureor destroy microstructural evidence for exsolution(eg Griffin et al 1984)Modelling shows that qualitatively plagioclase accumu-

lation from a low-pressure melt derived from MORBmantle can account for positive Pb Sr and Eu and negativeZr anomalies relative to N-MORB whereas cpx accumula-tion or trapped melt increases Zr Y MREE and HREE

abundances (Fig 15) Some melt addition (for exampleby entrapment) seems to be required to increase all trace-element abundances to those observed for the eclogitesAlthough high-Ca and high-Mg eclogites could representshallow and deep portions of the same oceanic crust thehigher average temperatures obtained for the high-Caeclogites the frequent occurrence of carbon in high-Caeclogites and its absence in high-Mg eclogites the strongLREEHREE fractionation observed for six of seven high-Ca eclogites and the narrow range of compositions of thehigh-Mg eclogites suggest that they may have protolithsthat are not cogenetic Some discrepancies in particularaffecting the incompatible elements may be due to post-crystallization processes acting upon the protoliths suchas dehydration melting and later metasomatism that isevident in peridotites from the same locality (Griffin et al1999)Low K2OTh for high-Ca and high-Mg eclogites

(average 08 Table 4) relative to assumed crustal

0

2

4

6

8

10

30 40 50 60 70 80

Mg

VR43452

Yb

GARNET

pyroxenites

low-Mghigh-Mghigh-Ca

eclogites

VR19673ecl-g2

Fig 14 Garnet Mg-number vs Yb (ppm including 2s uncertainty)Lines qualitatively show separate correlations for eclogites (exceptoutlierVR43452) and pyroxenites (except outlierVR19673ecl-g2)

high-Ca

high-Mg

001

01

1

10

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

melt08 melt02 cryst cpx01 melt09 cryst

plag

(a)

001

01

1

10(b)

RockNMORB

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 15 Reconstructed (a) high-Ca and (b) high-Mg eclogitescompared with compositions of plagioclase and cpx crystallizingfrom a MORB melt modelled using distribution coefficients ofNorman et al (2005)The MORB was calculated as a 10 equilibriummelt of MORB pyrolite [composition from Kelemen et al (1993)except Pb Pr Sr Y and Lu which were calculated as residues from1 fractional melting of pyrolite from McDonough amp Sun (1995)using melt modes of Kinzler (1997) and distribution coefficients ofNorman et al (2005)] Also shown are mixtures of crystals (consistingof 50 plagioclase plus 50 cpx) and melt in the proportions 0901and 0208

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1866

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

precursors (eg gabbros SE Indian Ridge average 22Hart et al 1999) may reflect the preferred loss of fluid-mobile elements during dehydration (Becker et al 2000)although large uncertainties on these low-abundance ele-ments in the constituent minerals and reconstructedwhole-rocks makes this interpretation less than robustThe REE patterns of most high-Mg eclogites are too flatto represent melting residues as partial melting shouldproduce low LREEHREE (Electronic Appendix 10) Incontrast the marked LREE depletion of five of sevenhigh-Ca eclogites (LaLufrac14 02^08) contrasts with that ofassumed evolved gabbroic protoliths (LaLufrac1415^32Hart et al 1999) and can be modelled as a consequence ofpartial melting during subduction after eclogitization (iein the presence of garnet which retains HREE relative toLREE eg Barth et al 2002a) has taken place (Aulbachet al 2003)Our results are compatible with previous studies on

eclogite xenoliths from the Slave Craton includingthose from the same kimberlite that have linked thesesamples or some of these samples to Paleoproterozoicsubduction events (the 195^191Ga formation of theHottah terrane and the 18Ga formation of the GreatBear magmatic arc) in the Slave craton (Griffin et al1999 Kopylova et al 1999 Heaman et al 2002 2006Schmidberger et al 2005 2007)The compositional features discussed above although

compatible with crustal protoliths are not necessarilydefinitive of a subduction origin However if direct crystal-lization from basaltic melts at high pressure (ie mantledepths) is invoked instead specific conditions would haveto be met to satisfy the observations The Mg-number ineclogites from Lac de Gras are lower than those expectedof primary high-pressure mantle melts (eg Walter 1998OrsquoHara amp Herzberg 2002) requiring crystallization athigh pressures after olivine fractionation This originwould further require shielding from peridotitic wallrockwith which the fractionated melts would no longer be inequilibrium (Barth et al 2001) In such a setting the lowrelative abundances of fluid-mobile elements comparedwith those of less fluid-mobile elements and of LREErelative to HREE (Table 4) would reflect an expulsion offluids during crystallization of the protoliths and latermelt mobilization respectively The eclogites clearly donot represent garnet plus cpx cumulates which wouldrequire a positive correlation of Mg-number with Yb thatis at variance with the observed negative correlation ineclogitic garnets from this study (Fig 14)

Low-Mg eclogites

Garnets in low-Mg eclogites compositionally overlap withthose in massive eclogites from Jericho (NW of the Lacde Gras kimberlites) that were suggested to have a high-pressure intrusive origin based on the similarity of theirmajor-element compositions to those of Group A eclogites

fromsouthernAfrica forwhichamantleoriginhasbeenpro-posed (Kopylova et al1999) Garnets in low-Mg eclogites inthe present study have major-element compositions moresimilar to garnets in groupB rather thangroupAeclogitesThe positive slope in N-MORB-normalized HREE

abundances displayed by the low-Mg eclogites is clearlydistinct from other eclogite types and is similar to those ofboninite-like second-stage melts that are closely associatedwith ultramafic volcanics (Smithies et al 2004) (Fig 13a)The range of Mg-numbers in the low-Mg eclogites (55^66)overlaps with that of Archaean second-stage melts fromIsua (58^75) and Abitibi (52^82 Smithies et al 2004)These similarities might indicate that the low-Mg eclogiteshave an origin in a previously depleted mantle source suchas supra-subduction zone mantle Subchondritic ZrHf ofthe low-Mg eclogites contrasts with the supra-chondriticZrHf of the high-Mg and high-Ca eclogites (Aulbach et

al in preparation) and is consistent with previous cpx-controlled depletion in an arc mantle source This sourcecould have been subsequently remelted to generate theboninite-like precursor to the low-Mg eclogites as sug-gested for the protoliths of eclogites from Udachnaya(Jacob amp Foley 1999)Although some second-stage melts to which the low-Mg

eclogites are similar are closely associated with komatiitesand could contain a plume component the radiogenic ordepleted mantle-like Nd and Hf and unradiogenic Sr inthe low-Mg eclogites do not support plume involvement

Trace-element and Sr^Nd^Hf isotopeconstraints on post-formationmetasomatismSeveral low-Mg and high-Ca eclogites show negativeslopes in the LREE which may be indicative of interactionwith small amounts of LREE-enriched melt (Navon ampStolper 1987) amounting to cryptic metasomatism(ie not accompanied by crystallization of new phases) inthe nomenclature of Dawson (1984) High Nb abundancesrelative to elements of similar compatibility displayed bysome low-Mg and high-Ca eclogites (Fig 10) could resultfrom overestimation of rutile modes However high-CaeclogiteVR40345 is rutile-free and has been reconstructedwithout rutile its high Nb abundance is accompanied byLREE enrichment indicating that metasomatic additionmay be responsibleLow-Mg eclogites show a positive correlation of La with

Nb (Electronic Appendix 11a) indicating concomitantaddition of LREE and Nb This suggests that the metaso-matic agent was a solute-rich fluid or silicate silicic melt ashydrous fluids and carbonatites do not carry enough HFSEto affect the budget of eclogites (Green amp Wallace 1988Klemme et al 1995 Adam et al 1997 Stalder et al 1998)Of the samples that are enriched in Nb all but one (high-Ca eclogiteVR50909) show a simultaneous enrichment in

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1867

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Hf (Electronic Appendix 11b) Together these trace-element relationships suggest that both SmNd and LuHfwere lowered in these samples Their low parent^daughterratios contrast with higher ratios in unenriched samplesand resulted in the ingrowth of a range of Nd and Hf iso-tope compositions Compared with the low-Mg and high-Ca eclogites the high-Mg eclogites have a narrow range ofLa and Nb concentrations which correlates with a narrowrange in Nd and Hf isotope compositions (Fig 12)Precise model ages cannot be calculated because parent^

daughter ratios are available only from in situ analyseswhich have relatively large uncertainties Howeverevidence for long-term enrichment and depletion may bediscerned Low-Mg eclogites even those that are enrichedin LREE have relatively high SmNd as a result of havingoverall positive slopes in their REE patterns relative toMORB whereas the high-Mg eclogites have relativelyflat REE patterns and therefore lower SmNd (Fig 11)This translates into lower eNd for high-Mg eclogites thanfor low-Mg eclogites In contrast they show no systematicdifference with respect to 176Hf177Hf Because the low- andhigh-Mg eclogites are proposed to have different origins asdiscussed above and therefore did not necessarily have thesame initial isotopic composition and formation age theseisotopic differences cannot be used to estimate their agesA long-term coupling of trace-element and isotopic

abundances is also indicated by the fact that low-Mg eclo-gite YK1943 which has experienced Nb and LREE addi-tion has lower eNd (0) and eHf (thorn13) than low-Mg eclogiteVR43469 with higher LuHf and SmNd (eNdfrac14thorn47eHffrac14thorn128) If low-Mg eclogites YK1943 and VR43469originally had identical initial Nd and Hf isotope composi-tions and if the parent^daughter ratios are accurate andwere modified only once during their evolution shortlyafter metamorphism (when their isotopic compositionwould have been reset) this overprint can be calculatedfor both isotopic systems to have occurred at 17GaThis would suggest that Hf and Nd were added duringthe same event In contrast high-Mg eclogites VR67360and YK1949 would have had identical Ndisotope compositions at 10Ga whereas their Hf isotopecompositions are disturbed in that the sample with thehigher parent^daughter ratio has a lower 176Hf177Hf thanthe sample with the lower parent^daughter ratio Theseresults suggest that the LREE and HFSE enrichmentobserved in different samples occurred during separateevents that did not affect the eclogites equallyExtreme HFSE enrichment leading to the growth of

zircon and rutile and LREE addition leading to thegrowth of apatite in eclogites from the Jericho kimberlitein the northern Slave craton have been ascribed to twoseparate metasomatic events at around 18Ga (quasi-coincident with the age of formation and metamorphism)and 13^10Ga respectively (Heaman et al 2002 2006

Schmidberger et al 2005) The authors associated theseevents with Paleoproterozoic subduction and aMesoproterozoic thermal disturbance in the Slave litho-spheric mantle possibly linked to the Mackenzie igneousevent Mesoproterozoic Nd model ages mostly between13 and 14Ga were reported for eclogites from thenearby Ekati kimberlites (Jacob et al 2003) Although wedo not observe zircon or apatite in volatile-free eclogitesfrom Lac de Gras and the age constraints obtained fromour data are weak they support a link between theseevents described for other eclogite populations and themetasomatic enrichment observed in our studyThe Sr isotope data in the present study are similar to

those of Schmidberger et al (2007) who suggested that thedepleted-mantle-like values are consistent with derivationof the protoliths from an oceanic mantle source However87Sr86Sr in volatile-free eclogites in the present study isremarkably constant considering their Nd and Hf isotopevariability (Fig 12) and the variability of trace-elementpatterns (Fig 10) Given the evidence for dehydration andmelting in many of the eclogites fractionation of highlyincompatible Rb from Sr and differential ingrowth ofradiogenic Sr would be expectedThis may indicate a rela-tively recent overprint and considering the similarityof 87Sr86Sr to present-day depleted mantle which is signif-icantly lower than 87Sr86Sr of North American kimber-lites a juvenile (asthenospheric) source

PyroxenitesPyroxenites from Lac de Gras often show relationshipsin their major-element contents that are different fromthose observed in the eclogites For example pyroxeniticgarnets show a negative correlation between CaO andNa2O contents whereas these oxides are positively corre-lated for garnets in eclogites (Fig 5) CaO is well correlatedwith C2O3 in pyroxenitic garnets but not in eclogitic gar-nets (Fig 4) Mg-number is positively correlated withYb abundance for pyroxenitic garnets but anti-correlatedfor eclogitic garnets (Fig 14) This suggests that the pyrox-enites in the present study are not related to the eclogitesby fractionation or accumulation processes as proposedfor the origin of some pyroxenites (Foley et al 2003)Pyroxenites have also been interpreted as the products

of crystal segregation in magma channels with alteredoceanic crust as the magma source this would explaintheir Eu anomalies and stable and radiogenic isotopecompositions (Pearson et al 1993 Becker 1996) As thehigh-pressure liquidus phases of most terrestrial basaltsare cpx and garnet (OrsquoHara amp Herzberg 2002) partialmelts of eclogites with basaltic composition wouldcrystallize cpx and garnet but not opx More than half ofthe pyroxenites from Lac de Gras contain opx suggestingthat they probably did not crystallize directly fromeclogite-derived melts

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1868

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Low-Cr websterites and pyroxenites could be crystalli-zation products of mafic melts that interacted with perido-titic mantle thus becoming more refractory The positivecorrelation of Yb and Mg-number in garnet would allowan origin by fractional crystallization from such meltsAlternatively they formed by reaction of silicic (possiblyeclogite-derived) melts with mantle peridotites(eg Kelemen et al 1993 Aulbach et al 2002) This agreeswith the similarity of hypothetical melts in equilibriumwith pyroxenitic garnets in most samples to eclogite-derived melts (Electronic Appendix 12) Niobium concen-trations in all hypothetical melts are much higher than ineclogite-derived melts regardless of whether or not rutileis assumed to be in the residue The same is true for LaCe and Zr concentrations in some samples which mayindicate that NbZr and LREE were added to thepyroxenites some time after their formation as discussedin the previous section for some eclogites Because mini-mum ages for pyroxenitic rutile with an unradiogenicHf isotope composition (calculated by projecting the176Hf177Hf to the model depleted mantle isotopic evolutioncurve assuming zero ingrowth) range to 178066Ga(Aulbach et al in preparation) it appears likely that thepyroxenites were not formed recently (eg during kimber-lite-related or precursory melt migration) and that some ofthe samples are at least MesoproterozoicGarnet pyroxeniteYK1952 consists of schlieren of equili-

brated garnet pyroxenite in a lsquomatrixrsquo of spongy cpxfine cpx^opx intergrowths embayed garnet and meltpatches (Fig 2f) Garnet in the pyroxenite schlieren hasvery low LREE and HFSE abundances Orthopyroxenein the matrix has distinctly higher Mg-number andCr-number and contains more CaO than opx from thepyroxenite schlieren whereas opx rims in the schlierenare compositionally similar to opx in the matrix The con-siderable compositional and textural disequilibrium inthis sample suggests that this is a young feature causedby reaction with fluids shortly before or at the time ofentrainmentAs a consequence of the small sample sizes of most of

the pyroxenites studied here a comprehensive isotopicdataset is not available Clearly more such data areneeded in addition to stable isotope data to obtain betterconstraints on the origins of this rock type Consideringthe microstructural and compositional variability ofthe pyroxenites and major-element relationships oppositeto those observed for eclogites it appears certain that pyr-oxenites form by a variety of processes some of which areunrelated to the formation of eclogites and their protoliths

Volatile-rich eclogitesVolatile-rich eclogites from Lac de Gras display the mostextreme Hf^Nd isotope systematics of all the analysedsamples Garnets in phlogopitized (YK1911) and in cal-cite-bearing eclogite (VR67112b) have the most radiogenic

Nd (eNdfrac14 2486 and 1927 respectively) Garnet inVR67112b has extremely high eHf (2516) Clinopyroxene inYK1911 has the lowest eHf measured for any mineralseparate from Lac de Gras (53) in accord with the veryhigh Hf abundance in its cpx (5 ppm) In contrastgarnet in this sample has radiogenic Hf (eHffrac14 136) Thisdiscrepancy is at least in part due to their relatively lowequilibration temperatures (7698C and 8558C respec-tively) which allowed the isotopic compositions of cpxand garnet to evolve separately at their low andhigh parent^daughter element ratios respectively forextended periods of time and possibly since formationAccordingly the cpx in YK1911 gives a minimum Hf ageof 27GaSlab-derived fluids could have been introduced during

the 27Ga collision of the western Slave Craton domainincluding the thick plume-derived deep lithosphericmantle layer and the eastern domain with its highlydepleted lithosphere This would also be consistent withthe observation that volatile-rich eclogites occur at shallowdepths in the mantle (TKroghfrac14 685^8558C correspondingto depths of 85^115 km along a 40 mWm2 geotherm)where the bulk of dehydration and production of solute-rich melts in a subducting slab might be expected to occur(Becker et al 2000 Hermann et al 2006) Their negativeEu anomalies might indicate that the fluid was derivedfrom the shallow portion of the slab that experiencedplagioclase fractionation The presence of carbonate inVR67112b requires a carbonated sourceHigh 87Sr86Sr in cpx in YK1911 is probably the conse-

quence of isotopic equilibration with high-RbSr phlogo-pite after long-term radiogenic ingrowth while Hfaddition led to retarded ingrowth of radiogenic HfThese systematics combined with radiogenic Nd appear

inconsistent with the involvement of sediments in the pet-rogenesis of this sample (eg Perfit et al 1980 Vervoortet al 1999) or a hydrous fluid such as those responsible forMARID metasomatism in the Kaapvaal craton (Konzettet al 1998)Whatever the origin of the fluids the apparentHf mobility suggests that they were not in equilibriumwith a rutile-bearing source and also that they weresolute-rich in nature (ie hydrous melts able to dissolveHFSE) and the relatively high HREE abundances of thevolatile-rich eclogites indicate a garnet-free source suchas amphibolite This allowed evolution of the sample atrelatively high time-integrated SmNd

SUMMARY AND CONCLUSIONS

(1) The Lac de Gras kimberlites carried abundant eclo-gite and pyroxenite xenoliths to the surface Threemain eclogite types are distinguished in terms ofmajor-element composition and trace element

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1869

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

patterns high-Mg high-Ca and low-Mg eclogitesThe garnets of all eclogite types have anti-correlatedMg-number and Yb pointing to olivine fractionationas a controlling factor in the formation of their pre-cursor rocks Their flat or positively sloped HREEpatterns qualitatively indicate that the melt parentalto the precursor rocks was not in equilibrium with agarnet-bearing residue

(2) Reconstructed high-Mg eclogites and high-Caeclogites have flat HREE patterns positive Pb andSr anomalies and low ZrSm similar to oceanic gab-bros reported in the literature High-Mg eclogiteshave variably low HREE contents and some havesmall positive Eu anomalies possibly suggestive ofmore primitive relatively plagioclase-rich protolithswhereas high-Ca eclogites have higher HREE abun-dances and small or absent Eu anomalies that requiremore cpx- andor melt-rich protoliths if a subductionorigin is assumed

(3) Low-Mg eclogites have distinctive trace-element pat-terns that show a marked positive slope in the HREEand subchondritic ZrHf They resemble someArchaean second-stage melts (lsquoboninitesrsquo) and theirprotoliths appear to have formed from a previouslydepleted source

(4) After formation the eclogites were subject to a varietyof processes Most have low ratios of fluid-mobile vsless fluid-mobile elements possibly as a result of dehy-dration of the protoliths In addition some of the low-Mg and high-Ca eclogites have low LREEHREEconsistent with partial melt lossWhereas the low-Mgeclogites show concomitant enrichment in La Nb andZr some high-Mg and high-Ca eclogites show evi-dence for variable and unsystematic (decoupled)LREE and Nb addition

(5) Isotopically the trace-element characteristics of thedifferent eclogite types translate into lower eNd forhigh-Mg eclogites (negative slope in MREEfrac14 lowerSmNd) than for low-Mg eclogites (positive slope inMREEfrac14higher SmNd) Within the low-Mg groupsamples that show evidence for metasomatism (lowLuHf and SmNd) have lower eNd and eHf than asample that was apparently not metasomatized BothNd and Hf addition can be dated to 17Ga if identi-cal initial isotope compsitions are assumed for thelow-Mg eclogites In contrast the relatively constantand depleted mantle-like 87Sr86Sr of eclogitic cpxdespite variable Nd and Hf isotope compositions andparent^daughter ratios suggests recent input of juve-nile material

(6) Garnet and cpx in pyroxenites show major-elementrelationships different from those in eclogitesTherefore these two rock types are probably notrelated by fractionation processes The presence of

opx precludes direct crystallization from slab-derivedmelts Pyroxenites may be the product of (slab-derived) melts hybridized during assimilation of peri-dotitic mantle

(7) Garnets in phlogopite- and carbonate-bearing eclo-gites have extreme isotopic compositions includingthe most radiogenic Nd and Hf in the datasetNegative Eu anomalies may point to the involvementof a crustal protolith Highly radiogenic Sr radiogenicNd and unradiogenic Hf in the reconstructedphlogopite-bearing sample combined with highHREE abundances and evidence for Hf mobilitymay indicate involvement of solute-rich fluids origi-nating in a garnet- and rutile-free source

ACKNOWLEDGEMENTSHelp in the clean laboratories and with the analyticalfacilities by S Elhlou C Lawson A Sharma P Wielandand M Zhang is greatly appreciated Bill Griffin isthanked for his support and for agreeing to reviewseveral versions of the manuscript despite disagreeingwith many of the conclusions presented here The paperbenefited greatly from the detailed and constructivereviews provided by the editors Gareth Davies andMarjorie Wilson and by Matthias Barth Larry Heamanand an anonymous reviewer This work was funded by aMacquarie University International PostgraduateAward and Postgraduate Research Fund (SA) by anARC SPIRT grant sponsored by Kennecott Canada Incand by ARC Large and Discovery Grants to SYOrsquoRAnalytical data were obtained using instrumentationfunded by ARC LIEF and DEST Systemic InfrastructureGrants industry partners and Macquarie UniversityThis is Publication 481 from the ARC National KeyCentre for Geochemical Evolution and Metallogeny ofContinents (wwwesmqeduauGEMOC)

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online

REFERENCESAdam J GreenT H Sie S H amp Ryan C G (1997) Trace element

partitioning between aqueous fluids silicate melts and mineralsEuropeanJournal of Mineralogy 9 569^584

Aulbach S Stachel T Viljoen K S Brey G P amp Harris J W(2002) Eclogitic and websteritic diamond sources beneath theLimpopo Beltccedilis slab-melting the link Contributions to Mineralogy

and Petrology 143 56^70Aulbach S GriffinW L Pearson N J OrsquoReilly S Y Kivi K amp

Doyle B (2003) Origins of eclogites beneath the central SlaveCraton Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA-0011

Aulbach S Griffin W L Pearson N J OrsquoReilly S Y amp Kivi K(2004a) Mantle formation and evolution Slave Craton constraints

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1870

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

from HSE abundances and Re^Os isotope systematics of sulfideinclusions in mantle xenocrysts Chemical Geology 208 61^88

Aulbach S Griffin W L OrsquoReilly S Y amp McCandless T E(2004b) Genesis and evolution of the lithospheric mantle beneaththe Buffalo HeadTerrane Alberta (Canada) Lithos 77 413^451

Aulbach S GriffinW L Pearson N J OrsquoReilly S Y amp Doyle B(2007) Lithosphere formation in the central Slave Craton(Canada) plume subcretion or lithosphere accretion Contributionsto Mineralogy and Petrology (in press) doi 101007s00410-007-0200-1

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2001)Geochemistry of xenolithic eclogites from West Africa Part IA link between low MgO eclogites and Archean crust formationGeochimica et Cosmochimica Acta 65 1499^1527

Barth M G Foley S F amp Horn I (2002a) Partial melting inArchean subduction zones constraints from experimentally deter-mined trace element partition coefficients between eclogitic miner-als and tonalitic melts under upper mantle conditions PrecambrianResearch 113 323^340

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2002b)Geochemistry of xenolithic eclogites from West Africa part 2Origins of the high MgO eclogites Geochimica et Cosmochimica Acta

66 4325^4345Beard B L Fraracci K NTaylor L A Snyder G A Clayton R

A Mayeda T K amp Sobolev NV (1996) Petrography andgeochemistry of eclogites from the Mir kimberliteYakutia RussiaContributions to Mineralogy and Petrology 125 293^310

Becker H (1996) Geochemistry of garnet peridotite massifs fromlower Austria and the composition of deep lithosphere beneatha Palaeozoic convergent plate margin Chemical Geology 134 49^65

Becker H Jochum K P amp Carlson R W (2000) Trace elementfractionation during dehydration of eclogites from high-pressureterranes and the implications for element fluxes in subductionzones Chemical Geology 163 65^99

Bell K amp Tilton G R (2002) Probing the mantle The story fromcarbonatites EOS Transactions American Geophysical Union 83275^277

Benoit M Polvecurren M amp Ceuleneer G (1996) Trace element andisotopic characterisation of mafic cumulates in a fossil mantlediapir (Oman ophiolite) Chemical Geology 134 199^214

BleekerW (2003) The late Archean record a puzzle in ca 35 piecesLithos 71 99^134

BleekerW Ketchum JW F JacksonV A amp Villeneuve M (1999)The central Slave Basement Complex Part I Its structural topol-ogy and autochthonous core Canadian Journal of Earth Sciences 361083^1109

Blichert-Toft J amp Albare de F (1997) The Lu^Hf isotope geochemis-try of chondrites and the evolution of the mantle^crust systemEarth and Planetary Science Letters 148 243^258

Blichert-Toft J Chauvel C amp Albare de F (1997) Separation of Hfand Lu for high precision isotope analysis of rock samples by mag-netic sector-multiple collector ICP-MS Contributions to Mineralogy

and Petrology 127 248^260Brey G P amp Kolaquo hler T (1990) Geothermobarometry in four-phase

lherzolites II New thermobarometers and practical assessment ofexisting thermobarometers Journal of Petrology 31 1353^1378

Brey G P Kolaquo hlerT amp Nickel K G (1990) Geothermobarometry infour-phase lherzolites I Experimental results from 10 to 60 kbJournal of Petrology 31 1313^1352

Caporuscio F amp Smyth J (1990) Trace element crystal chemistryof mantle eclogites Contributions to Mineralogy and Petrology 105550^561

Carbno G B amp Canil D (2002) Mantle structure beneath the SWSlave Craton Canada constraints from garnet geochemistry in theDrybones Bay Kimberlite Journal of Petrology 43 129^142

Coleman R G Lee E D Beatty L B amp Brannock W W (1965)Eclogites and eclogites their differences and similaritiesGeological Society of America Bulletin 76 483^508

Creaser R A Grulaquo tter H Carlson J amp Crawford B (2004)Macrocrystal phlogopite Rb^Sr dates for the Ekati propertykimberlites Slave Province Canada evidence for multiple intru-sive episodes in the Paleocene and Eocene Lithos 76 399^414

Davies R M GriffinW L Pearson N J Andrew A S Doyle BJ amp OrsquoReilly S Y (1999) Diamonds from the deep Pipe DO-27Slave Craton Canada In Gurney J J Gurney J L Pascoe MD amp Richardson S H (eds) Proceedings of the 7th International

Kimberlite Conference CapeTown Red Roof Design pp 148^155Davies R M Griffin W L OrsquoReilly S Y amp Doyle B J (2004)

Mineral inclusions and geochemical characteristics of microdia-monds from the DO27 A154 A21 A418 DO18 DD17 and RanchLake kimberlites at Lac de Gras Slave Craton Canada Lithos 7739^55

DavisW J amp Hegner E (1992) Neodymium isotopic evidence for thetectonic assembly of Late Archean crust in the Slave Provincenorthwest Canada Contributions to Mineralogy and Petrology 111493^504

Davis W J Jones A G Bleeker W amp Grutter H (2003)Lithosphere development in the Slave craton a linked crustal andmantle perspective Lithos 71 575^589

Dawson J B (1984) Contrasting types of upper-mantle metasoma-tism In Kornprobst J (ed) Kimberlites II The Mantle and

Crust^Mantle Relationships Amsterdam Elsevier pp 289^294DePaolo D J (1981) Neodymium isotopes in the Colorado Front

Range and crust^mantle evolution in the Proterozoic Nature 291193^196

Erlank A J amp Kushiro I (1970) Potassium contents of syntheticpyroxenes at high temperatures and pressures Carnegie Institution ofWashingtonYearbook 68 433^439

Foley S F Barth M G amp Jenner G A (2000) Rutilemelt partitioncoefficients for trace elements and an assessment of the influence ofrutile on the trace element characteristics of subduction zonemagmas Geochimica et Cosmochimica Acta 64 933^938

Foley S F Buhre S amp Jacob D E (2003) Evolution of the Archaeancrust by delamination and shallow subduction Nature 421 249^252

Graham I Burgess J L Bryan D Ravenscroft P J Thomas EDoyle B J Hopkins R amp Armstrong K A (1999) Explorationhistory and geology of the Diavik Kimberlites Lac de GrasNorthwest Territories Canada In Gurney J J Gurney J LPascoe M D amp Richardson S H (eds) Proceedings of the 7th

International Kimberlite Conference Cape Town Red Roof Designpp 262^279

Green D H amp Wallace M E (1988) Mantle metasomatism byephemeral carbonatite melts Nature 336 459^462

Green T H Blundy J D Adam J amp Yaxley G M (2000) SIMSdetermination of trace element partition coefficients betweengarnet clinopyroxene and hydrous basaltic liquids at 2^75 GPaand 1080^12008C Lithos 53 165^187

GriffinW L amp OrsquoReilly SY (2007) Cratonic lithospheric mantle Isanything subducted Episodes 30(1) 43^53

Griffin W L Wass S Y amp Hollis J D (1984) Ultramafic xenolithsfrom Bullenmerri and Gnotuk maars Victoria Australia petrol-ogy of a sub-continental crust^mantle transition Journal of

Petrology 25 53^87GriffinW L Doyle B J Ryan C G Pearson N J OrsquoReilly SY

Davies R Kivi K van Achterbergh E amp Natapov LM (1999)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1871

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Layered mantle lithosphere in the Lac de Gras Area SlaveCraton composition structure and origin Journal of Petrology 40705^727

Griffin W L Pearson N J Belousova E Jackson S E vanAchterbergh E OrsquoReilly SY amp Shee S R (2000)The Hf isotopecomposition of cratonic mantle LAM-MC-ICPMS analysis ofzircon megacrysts in kimberlites Geochimica et Cosmochimica Acta

64 133^147Griffin W L OrsquoReilly S Y Doyle B J Pearson N J

Coopersmith H Kivi K Malkovets V amp Pokhilenko N(2004) Lithosphere mapping beneath the North American plateLithos 77 873^922

Grulaquo tter H S Apter D B amp Kong J (1999) Crust^mantle couplingevidence from mantle-derived xenocrystic garnets In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 307^312

Hart S R Blusztajn J Dick H J B Meyer P S ampMuehlenbachs K (1999) The fingerprint of seawater circulationin a 500-meter section of ocean crust gabbros Geochimica et

Cosmochimica Acta 63 4059^4080Harte B amp Kirkley M B (1997) Partitioning of trace elements

between clinopyroxene and garnet Data from mantle eclogitesChemical Geology 136 1^24

Heaman L M Creaser R A amp Cookenboo H O (2002) Extremeenrichment of high field strength elements in Jericho eclogitexenoliths A cryptic record of Paleoproterozoic subduction partialmelting and metasomatism beneath the Slave craton CanadaGeology 30 507^510

Heaman L M Kjarsgaard B A amp Creaser R A (2004) The tem-poral evolution of North American kimberlites Lithos 76 377^397

Heaman L M Creaser R A Cookenboo H O amp Chacko T(2006) Multi-stage modification of the northern Slave mantlelithosphere Evidence from zircon- and diamond-bearing eclogitexenoliths entrained in Jericho Kimberlite Journal of Petrology 47821^858

Hermann J Cpandler C Hack A amp Korsakov A (2006) Aqueousfluids and hydrous melts in high-pressure and ultra-high pressurerocks implications for element transfer in subduction zones Lithos92 399^417

Hills DV amp Haggerty S E (1989) Petrochemistry of eclogites fromthe Koidu Kimberlite Complex Sierra Leone Contributions to

Mineralogy and Petrology 103 397^422Hoffman P F (1989) Precambrian geology and tectonic history of

North America In Bally A amp Palmer A (eds) The Geology of

North AmericaccedilAn Overview Boulder CO Geological Society ofAmerica pp 447^512

IrelandT R Rudnick R L amp Spetsius Z (1994) Trace-elements indiamond inclusions from eclogites reveal link to Archean granitesEarth and Planetary Science Letters 128 199^213

Irvine G J Pearson D G Kjarsgaard B A Carlson R WKopylova M G amp Dreibus G (2003) Evolution of the litho-spheric mantle beneath Northern Canada a Re^Os isotope andPGE study of kimberlite-derived peridotite xenoliths fromSomerset Island and a comparison to the Slave and Kaapvaalcratons Lithos 71 461^488

Isachsen C E amp Bowring S A (1994) Evolution of the Slave cratonGeology 22 917^920

Jacob D E (2004) Nature and origin of eclogite xenoliths from kim-berlites Lithos 77 295^316

Jacob D E amp Foley S F (1999) Evidence for Archean ocean crustwith low high field strength element signature from diamondifer-ous eclogite xenoliths Lithos 48 317^336

Jacob D Jagoutz E Lowry D Mattey D amp Kudrjavtseva G(1994) Diamondiferous eclogites from Siberiaccedilremnants ofArchean oceanic-crust Geochimica et Cosmochimica Acta 585191^5207

Jacob D E Fung A Jagoutz E amp Pearson D G (2003) Petrologyand geochemistry of eclogite xenoliths from the Ekati kimberlitesarea Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA_239

Jacob D E Bizimis M amp Salters V J M (2005) Lu^Hf and geo-chemical systematics of recycled ancient oceanic crust evidencefrom Roberts Victor eclogites Contributions to Mineralogy and

Petrology 148 707^720Jerde E A Taylor L A Crozaz G Sobolev NV amp SobolevV N

(1993) Diamondiferous eclogites from Yakutia Siberiaccedilevidencefor a diversity of protoliths Contributions to Mineralogy and Petrology

114 189^202Jones A G Ferguson I J Chave A D Evans R L amp

McNeice GW (2001) The electric lithosphere of the Slave cratonGeology 29 423^426

Kelemen P B Shimizu N amp Dunn T (1993) Relative depletionof niobium in some arc magmas and the continental-crust-partitioning of K Nb La and Ce during meltrockreaction in the upper-mantle Earth and Planetary Science Letters 120111^134

Kinzler R J (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition Application to mid-ocean ridge basalt petrogenesis Journal of Geophysical Research 102853^874

Klemme S van der Laan S R Foley S F amp Guenther D (1995)Experimentally determined trace and minor element partitioningbetween clinopyroxene and carbonatite melt under upper mantleconditions Earth and Planetary Science Letters 133 439^448

Konzett J Armstrong R A Sweeney R J amp CompstonW (1998)The timing of MARID metasomatism in the Kaapvaal mantle anion probe study of zircons from MARID xenoliths Earth and

Planetary Science Letters 160 133^145Kopylova M G Russell J K amp Cookenboo H (1999) Mapping the

lithopshere beneath the north central Slave Craton In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 468^479

Krogh E (1988) The garnet^clinopyroxene iron^magnesiumgeothermometerccedila reinterpretation of existing experimentaldata Contributions to Mineralogy and Petrology 99 44^48

Kusky T M (1989) Accretion of Archean Slave Province Geology 1763^67

LeCheminant A N amp Heaman L M (1989) Mackenzie igneousevents Canada Middle Proterozoic hotspot magmatismassociated with ocean opening Earth and Planetary Science Letters

96 38^48LeCheminant A N Richardson D G DiLabio R N W amp

Richardson K A (1996) La recherche de diamants au CanadaGeological Survey of Canada Open File 3228 268pp

MacGregor I D amp Carter J L (1970) The chemistry of clinopyrox-enes and garnets of eclogite and peridotite xenoliths from theRoberts Victor mine South Africa Physics of the Earth and PlanetaryInteriors 3 391^397

MacGregor I D amp Manton W I (1986) Roberts Victor eclogitesancient oceanic crust Journal of Geophysical Research 91 14063^14079

MacKenzie J M amp Canil D (1999) Composition and thermal evo-lution of cratonic mantle beneath the central Archean SlaveProvince NWT Canada Contributions to Mineralogy and Petrology

134 313^324

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1872

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

McCandlessT E amp Gurney J J (1989) Sodium in garnet and potas-sum in clinopyroxene criteria for classifying mantle eclogites InRoss J (ed) Kimberlites and Related Rocks Carlton Blackwellpp 827^832

McDonoughW F amp Sun S-S (1995) The composition of the EarthChemical Geology 120 223^253

Navon O amp Stolper E (1987) Geochemical consequences of meltpercolation the upper mantle as a chromatographic columnJournal of Geology 95 285^307

Nikogosian I K amp Sobolev AV (1997) Ion-microprobe analysis ofmelt inclusions in olivine experience in estimating the olivine^meltpartition coefficients of trace elements Geochemistry International 35119^126

Norman M D Griffin W L Pearson N J Garcia M O ampOrsquoReilly S Y (1998) Quantitative analysis of trace element abun-dances in glasses and minerals a comparison of laser ablationinductively coupled plasma mass spectrometry solution inductivelycoupled plasma mass spectrometry proton microprobe and elec-tron microprobe data Journal of Analytical Atomic Spectrometry 13477^482

Norman M D Garcia M O amp Pietruszka A J (2005) Trace-element distribution coefficients for pyroxenes plagioclase andolivine in evolved tholeiites from the 1955 eruption of KilaueaVolcano Hawairsquoi and petrogenesis of differentiated rift-zonelavas American Mineralogist 90 888^899

OrsquoHara M J amp Herzberg C (2002) Interpretation of trace elementand isotope features of basalts Relevance of field relations petrol-ogy major element data phase equilibria and magma chambermodeling in basalt petrogenesis Geochimica et Cosmochimica Acta 662167^2191

PadghamW A amp FysonW K (1992) The Slave Province a distinctcraton CanadianJournal of Earth Sciences 29 2072^2086

Pearson D G Davies G R amp Nixon P H (1993) Geochemical con-straints on the petrogenesis of diamond facies pyroxenites from theBeni Bousera peridotite massif North Morocco Journal of Petrology34 125^172

Pearson N J Griffin W L Doyle B J OrsquoReilly S Y vanAchterbergh E amp Kivi K (1999) Xenoliths from kimberlitepipes of the Lac de gras area Slave Craton Canada In GurneyJ J Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite ConferenceCapeTown Red Roof Design pp 644^658

Perfit N R Gust D A Bence A E Arculus R J amp Taylor S R(1980) Chemical characteristics of island-arc basalts implicationsfor mantle sources Chemical Geology 30 227^256

Polat A Hofmann A W amp Rosing M T (2002) Boninite-likevolcanic rocks in the 37^38 Ga Isua greenstone belt WestGreenland geochemical evidence for intra-oceanic subductionzone processes in the early Earth Chemical Geology 184 231^254

Poudjom DjomaniY H GriffinW L OrsquoReilly SY amp Doyle B J(2005) Lithospheric domains and controls on kimberlite emplace-ment Slave Province Canada evidence from elastic thickness andupper mantle composition Geochemistry Geophysics Geosystems 6doi1010292005GC0009978

Presnall D C Gudfinnsson G H amp Walter M J (2002)Generation of mid-ocean ridge basalts at pressures from 1 to 7GPa Geochimica et Cosmochimica Acta 66 2073^2090

Schmidberger S S Heaman L M Simonetti A Creaser R A ampCookenboo H O (2005) Formation of Paleoproterozoiceclogitic mantle Slave Province (Canada) Insights from in-situHf and U^Pb isotopic analyses of mantle zircons Earth and

Planetary Science Letters 240 621^633

Schmidberger S S Simonetti A Heaman L M Creaser R A ampWhiteford S (2007) Lu^Hf in-situ Sr and Pb isotope and traceelement systematics for mantle eclogites from the Diavik diamondmine Evidence for Paleoproterozoic subduction beneath the Slavecraton Canada Earth and Planetary Science Letters 254 55^68

Schulze D J amp Helmstaedt H (1988) Coesite-sanidine eclogitesfrom kimberlite products of mantle fractionation or subductionJournal of Geology 96 435^443

Smithies R H Champion D C amp Sun S S (2004) The case forArchaean boninites Contributions to Mineralogy and Petrology 147705^721

Smyth J R Caporuscio F A amp McCormick T C (1989) Mantleeclogites evidence of igneous fractionation in the mantle Earth andPlanetary Science Letters 93 133^141

Sobolev NV amp Lavrentrsquoyev Y G (1971) Isomorphic sodium admix-ture in garnets formed at high pressure Contributions to Mineralogy

and Petrology 31 1^12Stalder R Foley S F Brey G P amp Horn I (1998) Mineral aqueous

fluid partitioning of trace elements at 900^12008C and 30^57GPa New experimental data for garnet clinopyroxene and rutileand implications for mantle metasomatism Geochimica et

Cosmochimica Acta 62 1781^1801Sun S-S amp McDonough W F (1989) Chemical and isotopic sys-

tematics of oceanic basalts implications for mantle compositionsand processes In Saunders A D amp Norris M J (eds)Magmatism in the Ocean Basins Geological Society London Special

Publications 42 313^345Taura H Yurimoto H Kurita K amp Sueno S (1998) Pressure

dependence on partition coefficients for trace elements between oli-vine and the coexisting melts Physics and Chemistry of Minerals 25469^484

Taylor L A amp Neal C R (1989) Eclogites with oceanic crustal andmantle signatures from the Bellsbank kimberlite South AfricaPart I Mineralogy petrography and whole rock chemistryJournal of Geology 97 551^567

van Achterbergh E Ryan C G amp Griffin W L (1999) Glitter online intensity reduction for the laser ablation inductively coupledplasma spectrometry Abstract 9th V M GoldschmidtConference LPI Boston MA 905 pp Abstract no 7215

van Breemen O DavisW J amp King J E (1992) Temporal distribu-tion of granitoid plutonic rocks in the Archean Slave ProvinceNorthwest Canadian Shield Canadian Journal of Earth Sciences 292186^2199

Vervoort J D Patchett P J Blichert-Toft J amp Albare de F (1999)Relationship between Lu^Hf and Sm^Nd isotopic systems in theglobal sedimentary system Earth and Planetary Science Letters 16879^99

Viljoen K S Smith C B amp Sharp Z D (1996) Stable and radio-genic isotope study of eclogite xenoliths from the Orapa kimberliteBotswana Chemical Geology 131 235^255

Walter M J (1998) Melting of garnet peridotite and theorigin of komatiite and depleted lithosphere Journal of Petrology 3929^60

Wasserburg G J Jacobsen S B DePaolo D J McCulloch M Tamp Wen T (1981) Precise determination of SmNd ratios Sm andNd isotopic abundances in standard solutions Geochimica et

Cosmochimica Acta 45 2311^2323Westerlund K J Shirey S B Richardson S H Carlson R W

Gurney J J amp Harris JW (2006) A subduction wedge origin forPaleoarchean peridotitic diamonds and harzburgites from thePanda kimberlite Slave craton evidence from Re^Os isotope sys-tematics Contributions to Mineralogy and Petrology 152 275^294

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1873

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Page 7: Origins of Xenolithic Eclogites and Pyroxenites from the Central

3 vol graphite which occurs as disseminated plates sev-eral millimetres long with prismatic tabular or irregularhabit Quartz has been retrieved from a mineral separateof low-Mg eclogite YK1943 The phlogopitized samples(VR19674ecl-g7 and YK1911) also contain apatite andglass as accessory phases Vesicles in sample VR67112bare partially filled with single or multiple large carbonatecrystalsMost pyroxenites from Lac de Gras have fine- to coarse-

grained granoblastic equilibrated microstructures(Pearson et al 1999) Modal information for pyroxenitesin the present study is available for only the larger sampleYK1914 (22 garnet 78 cpx) (Fig 2e)

MAJOR ELEMENTSAverage major-element compositions of garnet cpx andopx in the different eclogite types (except for the threevolatile-rich eclogites which are not averaged) and in pyr-oxenites are given in Table 2 the full datasets includingaccessory minerals are available as ElectronicAppendices 2^5 which can be downloaded from httppetrologyoxfordjournalsorg

GarnetGarnets have XMg [pyrope component 100Mg(MgthornFethornCathornMn)] ranging from 178 to 756 andXCa (grossular component mol) ranging from 72 to359 (Electronic Appendix 2) In a diagram of CaO(MgOthornFeO) vs MgOFeO (Fig 3) eclogitic garnetsfall into three groups (1) low-Mg garnets with low

CaO(MgOthornFeO) and low MgOFeO (2) high-Mggarnets with low CaO(MgOthornFeO) and high MgOFeO (3) high-Ca those with high CaO(MgOthornFeO)and intermediate MgOFeO Volatile-rich eclogites havegarnets with the lowest MgOFeO whereas pyroxenitestrend towards higher values than eclogitesMost eclogite garnets have Cr2O3 contents502wt

with variable CaO contents (Fig 4) Garnets in pyroxeniteshave low CaO contents which correlate positively withCr2O3 More than half of the pyroxenites are opx-free andall pyroxenites are olivine-free on thin-section scalePyroxenitic garnets show an opposing trend of CaOvs Na2O (Fig 5) compared with the eclogitic garnetsBecause the positive correlation of Cr2O3 and CaO forpyroxenitic garnet indicates buffering by both cpx andopx opx-bearing or opx-free pyroxenites will not befurther distinguished subsequently (except in the geother-mobarometry section where opx-bearing assemblagesallow simultaneous calculation of pressure andtemperature)Distinct rim compositions in most garnets are

characterized by lower CaO contents and higherMg-number [frac14 100Mg(MgthornFe)] Garnet compositionsin three samples are inhomogeneous with regard to CaOMgO and FeO (SE01 VR43479 and VR43480) withoutclear core^rim zonations similar to coexisting cpx Likerims CaO-poor compositions are characterized by higherMg-number Visibly secondary spongy cpx rims also havehigher MgO contents (see below) and MgO-rich garnetsare therefore regarded as affected by secondary changesBased on Cr^Ca Mg^Ca^Fe and Ca^Na composi-

tional relationships in garnet the following eclogitetypes are distinguished (1) eclogites with high-Ca

0

05

10

15

20

25

30

0 01 02 03 04 05 06 07 08 09

high-Ca

high-Mg+pyroxenites

low-Mg

CaO(MgO+FeO)

MgO

FeO

GARNETpyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropic Jericho massiveDiavik low-temperatureDiavik high-temperatureDiavik diamond-bearing

eclogites

Fig 3 CaO(MgOthornFeO) vs MgOFeO (wt ) in garnet Fieldsdistinguish low-Mg and high-Ca eclogites as well as high-Mg eclogitesplus pyroxenites in this study Garnets in eclogites from Jericho(Kopylova et al 1999) and Diavik (Schmidberger et al 2007) areshown for comparison Garnet inclusions in diamond from Davieset al (1999 2004)

-135 wt (high-Ca)

2

3

4

5

6

7

8

9

0 05 10 15 20 25Cr2O3

CaO

GARNET

opx-bearing

opx-free

pyroxenites

low-Mghigh-Mgvolatile-rich

Jericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

VR19673ecl-g1g2

Fig 4 Cr2O3 vs CaO (wt ) in garnet Garnets in eclogites fromJericho and Diavik as in Fig 3

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1849

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 2 Summary of major-element contents in garnet and cpx (wt )

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO Na2O Total Mg-no T(Krogh) Type

Garnet

Low-Mg ecl 3964 026 2214 011 1843 037 1050 846 009 1000 503 976

minimum 3893 008 2130 5009 1071 024 849 567 006 990 423 860

maximum 4094 036 2275 011 2093 046 1457 1062 012 1004 708 1088

1s 070 012 052 na 354 007 227 185 003 05 99 98

High-Mg ecl 4081 026 2261 013 1419 030 1537 624 007 1000 659 969

minimum 3970 008 2224 5009 951 023 1127 418 003 991 551 820

maximum 4250 052 2323 035 1795 036 2030 882 013 1012 792 1241

1s 077 013 031 na 225 004 189 141 003 05 57 88

High-Ca ecl 3979 034 2211 na 1516 027 958 1240 011 998 528 1078

minimum 3909 011 2117 5009 1164 023 801 1144 007 989 456 910

maximum 4039 073 2242 011 1713 035 1249 1351 018 1004 657 1246

1s 042 026 042 na 173 004 155 073 004 06 67 151

vr19674ecl-g7 3867 003 2187 5009 2380 153 793 627 5004 1002 373 volatile-rich

vr67112b 3783 013 2097 009 2620 204 466 790 5004 999 241 volatile-rich

yk1911 3874 5006 2200 5009 2120 221 802 747 5004 997 403 volatile-rich

Pyroxenites 4163 021 2257 075 1148 034 1906 407 006 1002 747 909

minimum 3989 005 2081 009 729 025 1535 277 004 996 593 699

maximum 4279 048 2330 260 1881 054 2215 566 009 1011 842 1158

1s 068 010 057 057 323 007 209 065 001 04 73 117

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO Na2O Total Mg-no Type

Cpx

Low-Mg ecl 5453 028 739 na 500 5009 1117 1662 434 994 801

minimum 5160 017 564 5009 337 na 876 1388 197 981 741

maximum 5570 044 1067 014 687 na 1324 1926 632 1001 865

1s 128 011 155 na 124 na 153 177 126 06 40

High-Mg ecl 5483 023 483 na 381 na 1368 1875 318 994 865

minimum 5353 009 263 5009 245 5009 1156 1617 204 986 778

maximum 5645 031 810 016 674 011 1588 2101 495 1011 920

1s 071 007 137 na 095 na 104 138 075 06 30

High-Ca ecl 5563 027 1184 na 318 5009 868 1368 615 995 836

minimum 5448 014 769 5009 168 na 619 1060 391 987 751

maximum 5670 049 1579 010 603 na 1148 1689 807 1003 878

1s 076 015 334 na 151 na 189 261 167 06 45

vr19674 ecl-g7 5322 010 185 5009 536 015 1448 2339 062 992 828 volatile-rich

vr67112b 5040 041 328 5009 1079 024 1108 2229 071 992 647 volatile-rich

yk1911 5231 014 301 5009 544 024 1450 2340 056 996 826 volatile-rich

Pyroxenites 5527 na 347 038 304 na 1543 1959 245 999 899

minimum 5299 5006 037 010 149 5009 1128 1408 030 990 804

maximum 5666 037 786 134 490 014 1863 2312 537 1009 947

1s 068 na 189 028 091 na 193 226 131 05 37

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO NiO Total Mg-no

Opx

Pyroxenites 5781 na 059 na 627 na 3508 037 011 1006 909

minimum 5546 5006 032 5009 462 5009 2999 020 008 996 804

maximum 5888 009 318 054 1304 014 3628 077 014 1014 933

1s 098 na 067 na 212 na 160 019 002 06 32

Averaged for the three eclogite types and for pyroxenites minimum and maximum values as well as standard deviationsalso given Concentrations in volatile-rich eclogites are shown individually ecl eclogite Mg-numberfrac14 100Mg(Mgthorn Fe)Temperatures of last equilibration [T(Krogh)] as in Electronic Appendix 8

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1850

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

garnets (these include kyanite- graphite- and diamond-bearing varieties) (2) eclogites with low-Mg garnets (theseinclude quartz-bearing eclogite) (3) eclogites with high-Mggarnets and Cr2O3 contents502wt (4) volatile-richeclogites (phlogopite- apatite- carbonate-bearing)

Clinopyroxene (cpx)Clinopyroxenes in all eclogite groups have higher averageAl2O3 contents than those in pyroxenites with the formerhaving about 90 of the total Al in the jadeite molecule(AlVI) versus about 35 for the latter corresponding toan average of 41 and 25wt Na2O respectivelyThe highest Al2O3 and Na2O contents in the dataset areobserved in cpx from the high-Ca eclogites (up to158wt Al2O3) (Electronic Appendix 3) Contents ofK2O are generally below the detection limit(5004wt ) but values up to 024wt are observed insome cpx in high-Ca eclogites Average cpx Mg-numberare highest in pyroxenites (899) followed by cpx inhigh-Mg eclogites (865) high-Ca eclogites (836) andlow-Mg eclogites (801) Pyroxenitic cpx tend to havelower TiO2 and higher NiO and distinctly higherCr2O3 contents compared with those in eclogites A plotof MgO vs Al2O3 shows a negative correlation (Fig 6)Clinopyroxenes in low-Mg eclogites have lower Al2O3

contents at a given MgO content than other eclogitetypes Between 15 and 17wt MgO the Al2O3^MgOcorrelation for cpx in pyroxenites has a different slopefrom that in the eclogites Clinopyroxenes in pyroxenites

VR19673ecl-g1 and VR19673ecl-g2 have distinctly lowerMgO contents and higher Al2O3 contents than inthe other pyroxenites They coexist with garnets havingthe lowest Cr2O3 and CaO contents of the dataset whichdefine one end of the positive Cr2O3^CaO correlationobserved for garnetsClinopyroxenes in volatile-rich eclogites have distinctly

high MnO contents (up to 024wt ) Mg-numbers forcpx in the phlogopite-bearing variety are 826 and 828whereas that in the calcite-bearing eclogite (VR67112b)has an Mg-number of 647Some cpx grains have spongy rims that contain small

(mm-scale) patches of glass these rims are richer in Caand Mg and markedly poorer in Na than the cores similarto cpx in eclogites from southern Africa (Taylor amp Neal1989) The rim texture is unequilibrated suggesting lategrowth The formation of the spongy cpx rims might berelated to incongruent melting of cpx or be a result ofmetasomatism (Taylor amp Neal 1989 and referencestherein) Distinct although not spongy rims in two eclo-gites (VR67360 and YK1946) are similarly enriched inCaO and depleted in Na2O Two types of cpx with respectto CaO Na2O and Mg are present in VR43477VR43479 and VR43480 without clear core^rim zonationor spongy rims No significant differences are recognizedwith respect to other oxides

Orthopyroxene (opx)Orthopyroxenes have Mg-number ranging from 804 to933 (17 of 19 samples have values between 895 and 933)and typically contain minor amounts of Al2O3

(032^118wt one outlier has 318wt ) CaO(020^118 wt ) and NiO (008^014wt ) (Electronic

0

2

4

6

8

10

12

14

7 9 11 13 15 17 19

Al 2

O3

CPX

pyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

cpx in pyroxenitesVR19673ecl-g1g2

MgO

Fig 6 MgO vs Al2O3 (wt ) in cpx Data sources as in Figs 3and 4

0

005

010

015

020

2 4 6 8 10 12 14 16

CaO

Na 2

O

dl

group Igroup II

pyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

GARNET

Fig 5 CaO vs Na2O (wt ) in garnet Arrows show trends foreclogites and pyroxenites respectively [note that part of thepyroxenite trend extends into the field falling below the detectionlimit (dl) for Na2O] Group I and group II refer to classificationof McCandless amp Gurney (1989) Data sources as in Figs 3 and 4

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1851

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Appendix 4) The contents of all other elements are mostlybelow their respective detection limits

Accessory phasesRutile and ilmenite occur in some of the eclogites (Table 1)All rutile grains have exsolved 10 ilmenite as lamellaeand rims Ilmenite-free areas of rutile in eclogites containbetween 962 and 988wt TiO2 and minor amounts ofAl2O3 (01^04wt ) Cr2O3 (5009 to 02wt ) andFeO (03^07wt ) Rutile in pyroxenite YK1915 contains004wt Al2O3 01wt FeO and 15wt Cr2O3Discrete ilmenite grains in sample VR67112b haveTiO2 contents of 499 and 500wt MgO of 23 and15wt and FeO of 399 and 403wt respectivelySulfides occur interstitially or enclosed in cpx garnet

and opx they are mostly pyrrhotite and monosulfide solidsolution with variable Ni contents (unpublished data)They will not be considered further hereAbundant platy phlogopite is observed in two samples

(VR19674ecl-g7 and YK1911) It is Cl-rich in sampleVR19674ecl-g7 (063wt vs 003wt in YK1911) andF-rich in sample YK1911 (097wt vs 005wt inVR19674ecl-g7) (Electronic Appendix 5) The same is truefor coexisting apatite (VR19674ecl-g7 254wt Cl and014wt F YK1911 006wt Cl and 306wt F)

Comparison with mineral inclusions indiamonds from Lac de Gras and witheclogite xenoliths from other Slave CratonlocalitiesMineral inclusions in diamonds from the Lac de Graskimberlites have been investigated by Davies et al (19992004) For garnets there is a striking similarity betweenthe inclusions in diamond and those in the high-Ca eclogitexenoliths although the inclusions in diamond trendtowards higher Na2O (Fig 5) and lower Al2O3 contentsThe overlap is not as marked for cpx where inclusionsin diamond span almost the entire range observed foreclogitic cpx (Fig 6)A comparison with eclogites from other localities in

the Slave Craton shows that garnets in most Diavik low-temperature eclogites reported by Schmidberger et al(2007) plot with garnets in high-Mg eclogites and inpyroxenites from this study (Figs 3^5) whereas garnets inhigh-temperature and diamond-bearing eclogites(Schmidberger et al 2007) overlap most with garnets inhigh-Ca eclogites from this study and with garnetsincluded in diamond Eclogites from the Jericho kimberlitehave either massive or anisotropic fabrics (Kopylova et al1999) Garnets in almost all massive eclogites plot withgarnets in low-Mg eclogites from this study with regard toCaO^MgO^FeO relationships (Fig 3) but many havelower Na2O contents at a given CaO content (Fig 5)Garnets in anisotropic eclogites are restricted to lowMgOFeO (507) and show some overlap with garnets

in high-Ca eclogites although trending towards lowerCaO(MgOthornFeO) (Fig 3) while coexisting cpx is alsorestricted to low MgO and trends toward high Al2O3 con-tents (Fig 6)

TRACE ELEMENTSGarnetTrace-element abundances in garnet are summarized inTable 3 and the complete dataset is reported in ElectronicAppendix 1 Garnets in low-Mg eclogites show steep posi-tive slopes between LaN and SmN and shallower positiveslopes in the normalized heavy rare earth elements(HREEN Fig 7a) The normalized light REE (LREEN)show an order of magnitude variability ZrHf is mostlysupra-chondritic and Ti can be enriched or depleted rela-tive to elements of similar compatibility Niobium in threesamples is strongly depleted at 002^003 chondriticGarnets in high-Mg eclogites also show smooth positiveslopes in LREEN but have flat HREEN and variablenormalized middle REE (MREEN) to HREEN (Fig 7b)ZrHf is always supra-chondritic and Ti enriched ordepleted relative to similarly compatible elementsNiobium is most samples is below detection Small positiveEu anomalies though not outside the analytical uncer-tainty for Eu relative to either Sm or Gd are observed forgarnet in some high-Mg eclogites (eg EuEu up to 15for VR67360 where Eu is the average of the chondrite-normalized Sm and Gd concentrations) Garnets in high-Ca eclogites are slightly enriched in the MREEN relative tothe HREEN or have flat MREEN and HREEN with ahump between EuN and HoN (Fig 7c) One sample(VR40345) shows strong La enrichment (greater thanchondrite) Two samples have supra-chondritic ZrHfwhereas three samples have close to chondritic ratios andone sample has a subchondritic ratio As is true for theother eclogite typesTiN is either enriched or depleted rela-tive to similarly compatible elementsPyroxenitic garnets have variable slopes in the LREEN

and flat MREEN to HREEN patterns and two samplesshow high Sc relative to neighbouring elements Garnet inone of the pyroxenites (YK1952) has a highly distinctivepattern relative to other pyroxenites with very low LREEand HFSE (Fig 7d) This sample consists of coarse pyroxe-nite schlieren in which the garnet was analysed in a finer-grained lsquomatrixrsquo of spongy cpx with secondary rims finecpx^opx intergrowths embayed garnet and melt patchesVolatile-rich eclogites have garnets that are slightly

enriched in MREEN relative to HREEN or have flatMREE to HREE patterns One garnet has a pronouncednegative Eu anomaly (EuEufrac14 07 forVR67112b Fig7e)Garnets in volatile-rich samples have the lowest Ni Coand Ti and the highest V ScYand HREE abundances inthe dataset

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1852

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 3 Summary of trace-element contents in garnet and cpx (ppm)

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR19674

ecl-g7

VR67112b YK1911 av min max 1s

n 6 6 7 3 5 4 4

P 169 54 270 88 196 70 310 78 282 80 450 138 160 60 80 214 48 370 144

Sc 62 40 89 16 48 36 67 11 53 42 79 15 69 92 93 186 52 440 179

Ti 1667 490 2500 680 1073 620 2000 455 2141 700 4700 1582 270 690 260 1097 540 1900 573

V 169 78 250 68 150 78 280 72 178 110 390 101 140 340 100 390 71 820 326

Co 59 42 70 11 58 39 84 13 58 47 65 5 30 41 27 68 55 90 16

Ni 17 5 32 10 21 11 34 10 24 11 35 8 6 10 51 54 25 110 38

Ga 12 7 15 3 13 8 16 3 12 9 20 4 8 86 58 12 10 14 2

Sr na 512 4 na na 522 522 na na 57 57 na 52 5025 510 513 513 513 na

Y 37 6 61 18 21 7 38 10 31 20 38 5 60 81 70 19 12 26 6

Zr 15 5 29 9 28 6 83 26 24 9 43 12 60 22 90 18 1 41 17

Nb 009 0005 048 019 na 50015 05 na na 5005 05 na 5003 0012 5001 022 0030 057 025

La 005 001 009 004 na 50007 00 na na 5003 04 na 004 002 0009 003 000 009 005

Ce 03 01 05 02 01 00 01 00 05 01 18 05 03 022 015 01 00 03 01

Pr 01 00 02 01 00 00 01 00 02 01 03 00 017 014 012 00 00 01 00

Nd 12 03 24 07 07 04 14 03 21 12 32 07 24 23 20 03 00 07 03

Sm 11 04 22 06 10 03 18 05 23 08 36 10 29 31 33 05 01 09 04

Eu 06 03 10 03 06 02 10 03 11 04 17 04 16 114 16 03 01 06 02

Gd 27 07 41 12 23 07 42 12 43 16 53 13 8 81 90 13 06 24 08

Dy 54 11 70 23 36 13 65 18 58 35 71 11 12 144 12 29 18 42 10

Ho 14 02 23 07 08 03 15 04 12 09 14 02 24 30 23 07 04 10 02

Er 46 07 80 24 24 09 40 10 34 25 40 05 8 85 6 22 14 28 06

Yb 52 06 98 30 24 10 37 08 32 22 39 05 8 70 5 23 13 30 07

Lu 09 01 16 05 04 02 05 01 05 03 07 01 11 09 08 04 02 05 01

Hf 03 02 05 01 04 01 10 03 05 02 09 02 09 027 007 03 00 06 03

Ta na 50008 5004 na na 50005 001 na na 50004 na na 003 50002 50005 na 50007 011 na

Pb na 5005 02 na na 5002 02 na na 5003 07 na 02 004 5002 na 5003 01 na

Th na 5001 005 na na 50007 003 na na 50008 011 na 5007 0012 50004 002 0005 006 002

U na 5003 5003 na na 5001 01 na na 5007 na na 01 0023 010 na 5001 003 na

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR67112b YK1911 av min max 1s

n 6 4 8 5 4 4

P 65 15 120 41 39 14 56 16 53 30 70 12 20 9 80 27 120 40

Sc 22 10 30 7 15 14 19 2 14 7 24 5 80 43 17 11 21 5

Ti 1557 990 2400 508 1168 1010 1490 191 1988 810 3500 1154 2390 1700 689 102 1685 694

V 415 180 480 107 318 210 400 69 324 259 450 72 600 180 250 139 480 156

Co 35 26 50 8 20 14 31 7 26 14 42 11 49 23 29 20 42 10

Ni 164 87 220 47 228 100 430 128 188 135 320 56 59 34 428 410 670 184

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1853

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

ClinopyroxeneTrace-element abundances in cpx are summarized inTable 3 and the complete dataset is reported in ElectronicAppendix 6 Clinopyroxenes in low-Mg eclogites havemildly positive to negative slopes from LaN to NdN andsmooth negative slopes between SmN and HoN thatmostly continue to LuN (Fig 8a) Niobium is below detec-tion or 5007 chondritic and SrN is variably enrichedrelative to neighbouring elements with the peak size corre-lating negatively with LREEN ZrHf is subchondritic andHfN is enriched relative to SmN whereasTiN is enriched ordepleted relative to similarly compatible elementsWe ana-lysed the spongy rim in cpxYK1943 and its trace-elementpattern is similar to that of cpx cores for more compatibleelements whereas the most incompatible elements areseveral orders of magnitude higher Clinopyroxenes in

high-Mg eclogites have steeper negative slopes betweenNdN and LuN than those in low-Mg eclogites combinedwith lower HREE abundances HfN is depleted relative toSmN (Fig 8b) High-Ca eclogites contain cpx with sinusoi-dal REEN patterns with lower LREE than other eclogitetypes and the heaviest REE below detection They haveuniformly strong positive Sr peaks relative to elements ofsimilar compatibility (Fig 9) The diamond-bearing high-Ca eclogite (VR40345) has cpx with unusually high Nband LREE abundancesClinopyroxenes in four pyroxenites have disparate trace

element patterns They all share depletions in TiN relativeto similarly compatible elements and two show positiveSrN peaks Two samples have cpx with high and flatLREEN to MREEN patterns and steep negative slopes inthe HREEN The trace-element patterns of cpx in YK1914

Table 3 Continued

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR67112b YK1911 av min max 1s

n 6 4 8 5 4 4

Ga 16 10 20 3 12 9 15 2 26 15 39 9 129 123 6 11 11 5

Rb na 5008 04 na na 5007 05 02 na 5003 03 na 02 5004 03 010 05 0

Sr 166 85 290 74 222 153 285 50 243 83 440 142 85 120 200 11 414 209

Y 2 04 4 10 13 07 2 04 11 03 3 10 28 51 2 03 3 2

Zr 19 4 48 15 25 8 45 13 17 8 28 7 111 129 36 008 81 42

Nb 01 0003 03 01 na 5002 00 na na 5002 16 9 007 011 010 003 02 008

Cs na 5001 6 na na 5006 5006 na5013 5004 5013 na bdl 5002 na 5001 5002 na

Ba 09 02 2 07 09 01 2 07 10 03 3 10 06 01 13 06 3 10

La 3 03 8 3 3 23 3 04 4 01 14 6 51 97 4 01 10 4

Ce 7 13 20 6 10 7 13 2 7 05 24 10 25 28 11 01 28 13

Pr 11 02 3 08 2 10 3 06 08 02 2 09 47 46 2 002 4 2

Nd 5 12 12 4 10 5 14 4 4 09 8 26 23 22 8 01 22 10

Sm 11 05 21 06 20 10 31 10 07 02 11 03 47 53 2 005 5 2

Eu 03 01 05 01 06 028 09 03 02 010 03 007 087 123 06 002 15 07

Gd 08 03 13 03 11 05 19 06 05 03 06 011 28 37 15 010 36 2

Dy 05 01 07 02 04 022 06 01 02 013 03 007 11 16 07 008 14 06

Ho 01 00 02 004 006 003 007 002 na 5003 012 na 011 021 009 001 02 007

Er na 507 507 na 011 004 02 00 na 5006 03 na 019 033 02 002 03 012

Yb na 5001 02 na na 5001 011 na na 5006 03 na 007 015 009 001 02 006

Lu na 5001 003 na na50002 002 na na50008 01 na 0006 0013 0013 0002 003 001

Hf 12 03 24 07 11 04 18 05 10 06 16 04 34 53 na50015 35 na

Ta na50003 50015 na na50006 006 na na50004 5004 na 0014 00236 na50003 503 na

Pb 06 02 12 04 06 04 09 02 na 5016 07 02 065 098 08 01 18 1

Th na50005 06 na 012 005 02 005 na 5002 5006 na 021 02 001 06 03

U na 5005 5005 na 002 0009 005 001 na 5003 5005 na 0013 024 na50005 003 na

Averaged for the three eclogite types and for pyroxenites as in Table 2 pyx pyroxenite n number of analyses averagedbdl below detection limit

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1854

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

and YK1952 are similar although at different abundanceswith only modest LREEN over HREEN enrichment andstrong depletions in ZrN relative to neighbouring elements(Fig 8d)Clinopyroxenes in the volatile-rich samples have flat pat-

terns and high abundances from LaN to SmN and smoothnegative slopes from SmN to LuN Niobium Sr and Ti are

noticeably depleted relative to elements of similar compat-ibility (Fig 8e)

Trace-element partitioningTo assess whether coexisting phases are in trace-elementequilibrium we calculated apparent cpx^garnet distribu-tion coefficients and compared them with those

(a)

(d)

(b)

(e)

(c)

GarnetCI-chondrite GarnetCI-chondrite

low-Mghigh-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1952

1

10

Nd Sm Eu Gd Dy

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Fig 7 Extended trace-element patterns of garnet in (a) low-Mg eclogites (b) high-Mg eclogites with inset showing enlargement of MREE tohighlight Eu anomalies and 1s uncertainties (c) high-Ca eclogites (d) pyroxenites and (e) volatile-bearing eclogites Normalized to chondritevalues of McDonough amp Sun (1995)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1855

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

determined experimentally or in demonstrably equili-brated natural samples Low-Mg and high-Mg eclogiteshave Dcpxgarnet for most elements that are within therange determined for natural equilibrated eclogites andexperimental assemblages (Harte amp Kirkley 1997 Greenet al 2000 Barth et al 2002a) (Electronic Appendix 7)High-Ca eclogites plot at the low end of the range

of distribution coefficients and have very low DMREE

and DHREE outside the range indicated by the datasetof Harte amp Kirkley (1997) who noted that DREE decreaseswith increasing Ca content in eclogitic minerals Theseassemblages are therefore considered to be in equilibriumPyroxenite VR40384 has D values similar to eclogites

but at the high end of the range determined by Harte amp

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

(a)

(d)

(b)

(e)

(c)

CpxCI-chondrite CpxCI-chondrite

low-Mg high-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1914

YK1952

spongy rim

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

VR40345

Fig 8 Extended trace-element patterns of cpx as in Fig 7

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1856

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Kirkley (1997) (Electronic Appendix 7) Another group(YK1914 and YK1952) plots at the low end but DHf DZrDTi and DV are far below the values indicated by experi-mental datasets Because pyroxenites compositionallygrade into eclogites their drastically different DHFSE

values suggest disequilibrium In contrast volatile-bearingsamples have DHFSE values that appear to be too high to bein equilibrium however this may be a consequence oftheir unusual bulk chemical composition (Ti- and K-rich)which differs substantially from that of lsquonormalrsquo eclogitesand experimental charges

GEOTHERMOBAROMETRYFollowing Schmidberger et al (2007) temperatures oflast equilibration were calculated for eclogitic andopx-free pyroxenitic assemblages using the formulationof Krogh (1988) The results are summarized in Table 2and Electronic Appendix 8 for a preset pressure of5 GPa and illustrated as a histogram in Fig 9 Foropx-bearing pyroxenites temperatures were calculatedsimultaneously with pressure [Ca-in-opx thermometryand Al-in-opx barometry using the formulations of Breyet al (1990)] to constrain their relative positions in themantle columnMinerals in some eclogites from Lac de Gras show dis-

tinct core and rim compositions or inhomogeneitiesTemperatures were not calculated for samples withspatially irresolvable inhomogeneous garnet compositions(SE01VR43479VR43480) For eclogites with spongy cpxcore compositions were used in the calculationTemperatures were calculated for both core and rim com-positions where applicable For all samples these

temperatures agree within the uncertainty of this thermo-meter (508C Brey amp Kolaquo hler 1990)Most equilibration temperatures for eclogites lie between

800 and10008C (Fig 9)The transition between the shallowand the deep lithospheric mantle layers in the central Lacde Gras area lies in this temperature interval (Griffin et al1999 Pearson et al 1999) Eclogites in this study lack thedistinct bimodal temperature distribution observed bySchmidberger et al (2007) but as in their study high-Caeclogites give higher average temperatures (10808C corre-sponding to a depth of 155 km if equilibrated along a 40mWm2 geotherm) than other eclogites types (9808Cdepth 135 km) the diamond-bearing sample (VR40345)gives thehighest temperature of the datasetHighNa2Ocon-tents in garnet from high-Ca eclogites are qualitativelyconsistent with higher equilibration pressures (Sobolev ampLavrentrsquoyev 1971) as are high K2O contents in coexistingcpx (Erlank ampKushiro1970) suggesting that high-Ca eclo-gites not only equilibrated at high temperature but also highpressure rather than representing a temperature excursionPyroxenites both opx-bearing and opx-free give somewhatlower average temperatures than eclogites (mean 9258C)and phlogopite-rich eclogites give the lowest temperatures(685 and 7698C respectively) The calcite-bearing eclogiteVR67112b gives a relatively low equilibration temperatureof 8558C

MEASURED ANDRECONSTRUCTEDWHOLE -ROCKSRationale and approachKimberlite-borne xenoliths sometimes show evidence forinfiltration of the host kimberlite (eg Taylor amp Neal1989 Barth et al 2001 2002b Heaman et al 2006) Wetherefore follow the standard approach of calculatingwhole-rock compositions from the compositions of the con-stituent phases weighted by their modal abundances Onlycore compositions were used as distinct rim compositionsmay point to late diffusion of elements from kimberlite-related melt veinsMany eclogitic whole-rock compositions reconstructed

with only garnet and cpx have negative Ti anomalies(not shown) For most of these samples rutile or ilmenitehas been identified as part of the mineral assemblage andthese phases should therefore be considered in the whole-rock reconstruction Rutile or ilmenite modes were esti-mated by assuming that the bulk-rock has no Ti anomalywhen normalized to a normal mid-ocean ridge basalt(N-MORB) composition and calculating the amount ofrutile required to erase theTi anomalyIt is of course possible that the eclogites had negative

or positive Ti anomalies relative to elements of similarcompatibility (Eu and Dy) Because we reconstructed

0

2

4

6

8

10

12volatile-rich

pyroxenite opx-bearing

pyroxenite opx-free

high-Ca

high-Mg

low-Mg

700750

800850

900950

10001050

11001150

12001250

Fre

quen

cy

Temperature

Fig 9 Histogram ofTKrogh (Krogh 1988) obtained at a preset pres-sure of 5 GPa except opx-bearing pyroxenites where TBKN wassolved simultaneously with PBKN (Brey amp Kolaquo hler 1990)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1857

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

whole-rock compositions from mineral core compositionskimberlite infiltration is unlikely to have caused fractiona-tion of Ti from Eu and Dy In a study in which rutilemodes were estimated from the differences between Ti inmeasured bulk-rocks and in bulk-rocks reconstructed withonly garnet and cpx only three of 19 eclogites recon-structed with rutile show Ti anomalies (Barth et al2002a) Even eclogite xenoliths with depleted Zr and Hfabundances that resemble those of modern island arcbasalts do not show Ti anomalies (Jacob amp Foley 1999)We therefore consider our approach to be a reasonablyrobust means of obtaining meaningful rutile modes wheresmall sample sizes preclude reliable determination of themodes of accessory phases by point-countingWhere rutilewas not recognized on thin-section scale or not analysedthe average eclogitic rutile composition was used for thereconstruction

UncertaintiesReconstructing whole-rock compositions in this way is arather crude method given the coarse grain size of theeclogites relative to the sample size and the non-uniformdistribution of the principal minerals on thin-sectionscale The uncertainty in mineral modes of medium- tocoarse-grained eclogites with grains45mm has been esti-mated by Jerde et al (1993) to be 10 plus uncertaintiesrelated to partial alteration which is minor in our samplesNevertheless the whole-rock trace-element budget israther insensitive to the exact modes for common eclogiticassemblages and REE patterns which encompass mostother elements in terms of compatibility in garnet andcpx do not change significantly even for modal variationsof 30 (Jerde et al 1993) This is not true for major-element concentrations which vary substantially with var-iations in mineral modes and we therefore do not calculatewhole-rock major-element contentsBecause the discussion that follows is based on trace-

element patterns and trends rather than absolute abun-dances whole-rock reconstruction based on mineralmodes is adequate We are confident that the uncertaintyinherent in this approach is not leading to significant mis-representation of eclogite whole-rock compositions becausethe trace-element patterns of the reconstructed whole-rocks show distinct differences (see below) that correlatewith the assignment to eclogite types based on garnetmajor-element composition and this is independent ofmineral mode Following Jerde et al (1993) in ElectronicAppendix 9 we show variations in trace-element patternfor one sample in each eclogite type for variations ingarnet and cpx abundance from 7030 to 3070 vol This shows that the trace-element patterns are generallyinsensitive to cpxgt ratios with regard to slopes andanomalies such as positive Sr Pb and Eu anomalies withthe exception of the positive Eu anomaly in high-Mgeclogite VR43479 which disappears if a high cpx mode is

assumed (not shown) Variations in rutile modes by 50(Electronic Appendix 9d) lead as would be expected tothe appearance of Ti anomalies but a rutile mode of 05instead of 1 vol does not erase the positive Nb anomalyin low-Mg eclogiteVR50909Because of the robustness of the relative trace-element

abundances in the reconstructed eclogites a garnet propor-tion of 065 and cpx proportion of 035 approximately themedian modal abundances of these minerals in volatile-free eclogites from Lac de Gras are assumed for eclogitesamples that are too small for determination of meaningfulmodes by point-counting

Within- and between-group variationsThe trace-element compositions of reconstructed whole-rocks are given in Table 4 N-MORB-normalized trace-element patterns of reconstructed whole-rocks areshown in Fig 10 The three eclogite groups have distincttrace element patterns that are particularly obvious forthe REE (Fig 11) Low-Mg eclogites have positive Pb andSr anomalies and smooth positive slopes in their MREEN

and HREEN LREEN are enriched relative to MREEN inthree samples (Fig 11a) Three samples have high Nbrelative to similarly compatible elements and comparedwith most other eclogites (Fig 10a) All high-Mg eclogitesdisplay more or less pronounced positive Eu and Sranomalies and most have positive Pb and negative Zrand P anomalies Their N-MORB-normalized REE pat-terns show shallow positive slopes from LaN to PrN orNdNand aweak negative slope from NdN to LuN (Fig11b)VR43479 is a low-Mg eclogite with regard to its garnet

Ca^Fe^Mg composition but has a trace-element patterntypical of high-Mg eclogites and will be discussed withthe latter Conversely sample YK1949 a high-Mg eclogitein terms of its garnet Ca^Fe^Mg compostion is more simi-lar to low-Mg eclogites in terms of its whole-rock trace-element pattern and is plotted in Figs 10a and 11aWhereas the patterns of most samples within the groupresemble each other the N-MORB-normalized incompati-ble-element abundances of the high-Ca eclogites vary bymore than an order of magnitude (Fig 10c) All sampleshave positive Pb anomalies and most have positive Sranomalies Some samples have negative P andor Zranomalies Niobium abundances in three of the samplesare very high similar to those for some low-Mg eclogitesHigh-Ca eclogites tend to have flatter MREEN andHREEN patterns and more uniform HREE abundancesthat the other eclogite types with humps between SmN

and HoNThe phlogopitized sample YK1911 shows a strong posi-

tive Pb anomaly and lower HREE abundances than thecalcite-bearing sample VR67112b which shows a distinctnegative Eu anomaly and a depletion of La relative to Ce(Figs 10d and 11d)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1858

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Trace-element compositions of reconstructed whole-rocks (ppm)

Sample P Sc Ti V Co Ni Ga Sr Y Zr Nb La Ce Pr Nd

Low-Mg eclogites

vr19674-g4 186 50 4649 317 57 59 144 77 24 30 36 13 31 05 24

vr43452 154 39 2253 328 49 84 176 52 32 11 01 02 07 01 08

vr43469 38 65 4377 349 54 91 142 34 28 15 02 05 15 03 17

yk1926 144 45 3406 258 52 87 145 89 18 18 20 09 29 06 33

yk1943 173 46 5732 295 41 39 146 122 25 48 82 34 85 13 64

yk19491 102 34 1693 158 31 49 100 46 10 14 03 09 21 03 17

High-Mg eclogites

vr19677-g2 138 39 5160 291 41 53 119 98 14 22 04 13 53 11 64

vr434792 46 25 1265 129 36 126 84 56 3 7 04 06 25 04 23

vr67360 49 36 1573 138 64 184 111 76 5 9 04 09 38 07 33

yk1946 153 28 4877 192 45 105 124 108 12 48 03 10 38 08 50

High-Ca eclogites

vr19674-g5 267 30 7663 200 41 67 195 32 23 43 77 01 04 02 16

vr40345 0 33 2821 175 47 98 210 179 13 13 66 60 109 11 38

vr43465 142 27 5104 220 31 83 121 197 14 35 29 05 22 05 40

vr50858 295 30 6898 201 44 94 150 48 22 38 06 01 03 02 20

vr50860 68 57 5672 418 57 71 186 109 19 11 02 02 13 03 26

vr50909 238 28 7295 201 42 92 171 48 20 57 13 01 04 02 21

yk3528 91 51 3712 341 53 155 177 49 16 9 01 02 08 02 11

Volatile-rich eclogites

vr67112b 40 84 9242 521 46 30 107 4 40 66 31 26 126 24 125

yk1911 31 39 9428 286 44 82 164 3 23 52 64 18 52 09 46

Pyroxenites

yk1914 111 49 707 207 48 547 44 31 4 3 02 06 19 03 10

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

Low-Mg eclogites

vr19674-g4 11 05 10 22 39 10 29 33 06 12 05 028 24 008

vr43452 04 02 08 13 38 12 41 49 08 05 01 nd 02 na

vr43469 11 05 09 25 44 11 34 36 06 07 02 001 09 129

yk1926 12 05 11 19 30 07 22 24 04 07 03 004 23 090

yk1943 21 08 10 29 42 10 30 32 05 16 05 013 65 017

yk19491 06 03 11 08 14 04 14 18 03 04 02 003 29 115

High-Mg eclogites

vr19677-g2 20 08 12 21 24 05 15 15 03 05 05 007 52 6811

vr434792 05 02 12 05 06 01 04 03 00 03 01 001 98 037

vr67360 06 03 14 06 09 02 06 06 01 02 02 002 86 000

yk1946 18 07 12 21 22 05 13 12 02 11 03 006 56 112

High-Ca eclogites

vr19674-g5 17 09 11 33 43 09 24 23 04 16 05 007 02 011

vr40345 09 04 10 12 22 06 16 14 02 05 nd nd 298 na

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1859

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Continued

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

vr43465 22 09 12 25 26 06 15 14 02 10 04 001 25 040

vr50858 20 08 09 33 41 08 22 21 03 10 01 nd 02 na

vr50860 14 07 11 27 37 08 22 18 03 06 00 nd 06 na

vr50909 20 10 12 33 38 07 20 20 03 16 01 002 02 181

yk3528 08 05 11 20 30 07 18 18 02 05 01 nd 08 na

Volatile-rich eclogites

vr67112b 38 10 07 53 74 15 41 34 04 18 03 012 60 004

yk1911 20 07 08 35 39 07 18 17 02 10 14 nd 73 na

Pyroxenites

yk1914 02 01 11 04 08 02 05 06 01 01 03 003 68 101

aGarnet has CandashMgndashFe relationship typical of high-Mg eclogites but whole-rock trace-element pattern is typical oflow-Mg eclogitesbGarnet has CandashMgndashFe relationship typical of low-Mg eclogites but whole-rock trace-element pattern is typical ofhigh-Mg eclogites

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

(a)

001

01

1

10(c)

001

01

10(b)

VR43452

YK1949

RockNMORB

low-Mg

RockNMORB

high-Mg

(d)

volatile-richhigh-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

phlogopite-bearingYK1911

calcite-bearingVR67112b

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 10 Extended trace-element patterns of reconstructed (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-bearing eclogites Normalized to N-MORB of Sun amp McDonough (1989)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1860

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

RockNMORB

(b)

RockNMORB

(d)

010

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu001

01

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

(a)

001

01

1

10(b)

001

01

1

10(c)

001

01

1

10(e)

high-TSchmidberger et al

low-TSchmidberger et al

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

high-Mglow-Mg

high-Cavolatile-rich

phlogopite-bearingYK1911

calcite-bearingVR67112b

YK1949VR43479

Fig 11 N-MORB normalized REE patterns for (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-richeclogites Shown for comparison in (e) are reconstructed eclogites from Schmidberger et al (2007)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1861

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Both volatile-rich whole rocks display strong negativeSr anomalies Strontium can be sited in calcite or apatiteandor indicate the presence of a Sr-rich mineral that wasnot identified in thin section such as bariteFour of the six reconstructed whole-rocks in the study

of Schmidberger et al (2007) (Fig 11e) show markedLREEHREEN enrichment that is observed only in onehigh-Ca eclogite in our study (Fig 11c) suggesting thatmany of the eclogites they studied were affected by metaso-matic LREE enrichment Two samples in their study thatare not affected by this enrichment show positive Euanomalies similar to the high-Mg eclogites in this studywhereas the distinctive REE pattern of the low-Mg eclo-gites is not recognized among their samples

SR^ND^HF I SOTOPESYSTEMATICSStrontium Nd and Hf isotope data for optically puremineral separates and reconstructed whole-rocks aregiven in Table 5 Spongy rims and alteration are likely tobe mostly removed during leaching in 6N HCl in combina-tion with repeated ultrasonication Despite this rigoroustreatment the reproducibility of Nd in garnet VR43469and in cpx YK1914 is worse than for other samples Thismay be due to inhomogeneities related to late Nd diffusionpossibly from grain boundaries containing a kimberlitecomponent The reproducibility of the 176Hf177Hf data isgenerally poorer than that of Nd which may reflect a com-bination of matrix effects caused by imperfectly purifiedsolutions and sample heterogeneity Parent^daughterratios are available only from in situ analyses which arenot precise enough to allow meaningful model age orinitial isotope ratio calculations (ie at the time of kimber-lite eruption) Because the Lac de Gras kimberlites arerelatively young (55^56 Ma Graham et al 1999) age cor-rection would result in only a minor change in isotopecomposition which is inconsequential in view of the rangeof compositions displayed by most samples and hence hasno effect on the conclusions drawnClinopyroxenes in five volatile-free eclogites and a

pyroxenite have Sr isotope ratios ranging from 0701662 to0703546 this range is similar to or somewhat lower thanthat of the depleted mantle at the time of kimberliteemplacement (0703 Bell amp Tilton 2002) Clinopyroxenein the phlogopitized sample has a distinctly high 87Sr86Srof 0751186Clinopyroxene controls 56^92 of the Nd budget (Nd

concentration in cpx weighted by mode) Neodymiumisotope ratios range from 0511660 to 0514844 Thiscorresponds to eNd variation (per 10 000 deviationof 143Nd144Nd in the sample from 143Nd144Nd in the chon-dritic mantle model reservoir) from 17 to thorn43 cpx inthe low-Mg eclogites have higher eNd than those in

high-Mg eclogites Garnets have generally have higher143Nd144Nd from 0512984 (eNdfrac14thorn 7) to 0520965(eNdfrac14thorn 162) Garnets in the two volatile-rich sampleshave the highest ratios 0522790 for garnet in the calcite-bearing sample VR67112b (eNdfrac14thorn198) and 0525721 inthe phlogopitized eclogiteYK1911 (eNdfrac14thorn 255)Garnet contributes 18^32 and cpx 48^78 of the Hf

budget The proportion controlled by rutile is 4^30(Aulbach et al in preparation) Measured 176Hf177Hf incpx ranges from 0281263 in phlogopite-bearing sampleYK1911 (eHffrac1453) to 0286130 in low-Mg eclogiteVR43469 (eHffrac14thorn119) and in garnet from 0283958 inpyroxeniteYK1915 (eHffrac14thorn42) to 0295718 in low-Mg eclo-gite YK1926 (eHffrac14thorn458) Hafnium in garnet in the phlo-gopite^apatite-bearing sample is extremely radiogenic(176Hf177Hffrac14 0354367) corresponding to an eHf of thorn2532which is close to a value determined for a garnet separatefrom a Roberts Victor eclogite (eHffrac14thorn2561 Jacob et al2005) Garnet in the calcite-bearing sample has extremelyradiogenic Nd and Hf (eNdfrac14thorn193 eHffrac14thorn2516) Garnetin pyroxenite YK1915 also has radiogenic Nd and Hf(eNdfrac14thorn27 eHffrac14thorn42)Despite the large uncertainties for reconstructed whole-

rocks (calculated as the square root of the sum of squareduncertainties of two standard errors for garnet and cpxanalyses combined with 20 propagated uncertainty onthe proportion of Nd or Hf respectively contributed byeach mineral) the isotopic data are consistent in thatthey correlate with the independently determinedparent^daughter ratios (1) Low-Mg eclogites have higherSmNd reflected in the steeper positive slope of theirN-MORB-normalized REE patterns (Fig 11a) comparedwith the high-Mg eclogites (Fig 11b) Accordingly recon-structed low-Mg eclogites have higher present-day eNd

(0 to thorn47) than high-Mg eclogites (4 and 17) (2) Low-Mg eclogite YK1943 has very high Nb and LREE abun-dances that could have been added during metasomatismthat also enhanced its Hf abundance This sample has thelowest eNd and eHf of the low-Mg eclogites (0 and thorn20respectively) In contrast apparently unenriched low-Mgeclogites have more radiogenic Nd and Hf (eNdfrac14thorn47and thorn9 eHffrac14thorn128 and thorn113 respectively)The Sr^Nd isotope diagram reveals that eNd in cpx

varies from 19 to 43 at nearly constant depleted-mantle-like 87Sr86Sr (Fig 12a) In the Nd^Hf isotope diagram(Fig 12b) the reconstructed whole-rocks lie on curvedmixing lines if the lsquomixingrsquo proportions of Nd and Hffor the two lsquoend-memberrsquo minerals are very different(eg YK1911) Garnets have more radiogenic Nd andHf than coexisting cpx All reconstructed volatile-freeeclogites plot in the radiogenic field with regard to Hfor are isotopically similar to the depleted mantle Twohigh-Mg eclogites and one low-Mg eclogite have unradio-genic Nd one high-Mg eclogite has radiogenic Nd and

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1862

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 5 Individual and averaged Sr Nd and Hf isotope ratios obtained from garnet and cpx solutions

Sample 87Sr86Sr 2SE2SD 143Nd144Nd 2SE2SD eNd 2SE2SD 176Hf177Hf 2SE2SD eHf 2SE2SD

Low-Mg eclogites

VR43469 cp 0702661 0000010 0514844 0000006 43 01 0286130 0000012 119 04

VR43469 gt 0515508 0000014 56 03 0286506 0000012 132 04

VR43469 gt r 0515453 0000010 55 02 0286548 0000011 134 04

VR43469 gt r 0515384 0000007 54 01 0287244 0000010 158 04

VR43469 gt r 0515487 0000014 56 03 0286996 0000014 149 05

VR43469avg gt 0515458 0000109 55 21 0286823 0000715 143 253

WR VR43469 0515059 0000273 47 5 0286405 0001079 128 38

YK1926 cp 0703546 0000011 0513043 0000007 8 0 0283154 0000009 13 03

YK1926 gt 0513364 0000008 14 01 0295718 0000005 458 02

WR YK1926 0513120 0000026 9 1 0285980 0000716 113 25

YK1943 cp 0512143 0000007 10 01 0282213 0000005 20 02

YK1943 gt 0514461 0000012 36 02 0287510 0000008 168 03

YK1943 gt r 0514491 0000009 36 02 0287332 0000005 161 02

YK1943 gt r 0514517 0000010 37 02 0287290 0000007 160 02

YK1943avg gt 0514489 0000056 36 11 0287377 0000233 163 82

WR YK1943 0512625 0000228 0 5 0283138 0000879 13 31

High-Mg eclogites

VR43480 gt 0520965 0000017 162 03 0284950 0000010 77 04

VR67360 cp 0283509 0000074 26 26

VR67360 cp r 0283433 0000031 23 11

VR67360avg cp 0701662 0000018 0511660 0000010 19 02 0283471 0000108 25 38

VR67360 gt 0512984 0000101 7 20 0287429 0000180 165 64

WR VR67360 0511761 0000132 17 3 0284734 0000469 69 17

YK1949 cp 0703019 0000013 0512231 0000009 8 02 0282513 0000009 9 03

YK1949 gt 0513511 0000010 17 02 0290297 0000042 266 15

YK1949 gt r 0513508 0000012 17 02 0289947 0000015 254 05

YK1949 gt r 0513552 0000014 18 03

YK1949avg gt 0513523 0000049 17 10 0291569 0000494 311 175

WR YK1949 0512446 0000103 4 2 0284622 0000847 65 30

Volatile-rich eclogites

VR67112b gt 0522790 0000008 198 02 0354367 0000036 2532 13

YK1911 cp 0751186 0000009 0512165 0000007 9 01 0281263 0000009 53 03

YK1911 gt 0525721 0000013 255 03 0288051 0000012 187 04

WR YK1911 0513940 0000342 25 7 0281414 0000018 48 1

Pyroxenites

YK1914 cp 0703015 0000034 0513283 0000041 13 08

YK1914 cp r 0702956 0000005 0513200 0000027 11 05

YK1914 cp r 0513527 0000059 17 11

YK1914 cp r 0513045 0000010 8 02

YK1914avg cp 0702986 0000083 0513264 0000402 12 78

YK1915 gt 0514003 0000011 27 02 0283958 0000021 42

Analyses of full procedural replicates are designated with an lsquorrsquo after sample name SE standard error (internal errorsingle analyses) SD standard deviation (external error averaged multiple analyses) Italics used for reconstructed whole-rocks (mineral isotopic compositions and concentrations weighted by modes) Uncertainties for reconstructed whole-rocksinclude 2SE or 2SD uncertainties on each mineralrsquos isotope composition combined with 20 uncertainty in the amountof Nd or Hf contributed by the constituent minerals eNd and eHf are per 10 000 deviation from model chondritic mantle[model of Wasserburg et al (1981) and Blichert-Toft amp Albarede (1997) respectively]

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1863

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

εHf

εNd

Ekati eclogitesminus100

0

100

200

300

400

500

minus40 0 40 80 120 160 200 240 280

gt in YK1915

gt in VR67112bεNd=198 εHf=2532

minus40

0

40

80

120

160

minus40 minus20 0 20 40 60

minus60

minus40

minus20

0

20

40

070 071 072 073 074 075 076

North Americankimberlite field

εNd

87Sr86Sr

cpx in YK1914

Ekatieclogites

DM

DM

DM

DM

(a)

(b)

pyroxenites

low-Mg

high-Mg

volatile-rich

eclogites

Diavik low-TA154 S kimberlite

cpx in YK1911

Fig 12 (a) Sr^Nd for cpx and (b) Nd-Hf isotope diagram for garnet cpx and reconstructed whole-rocks (symbols with bold outline) Finedashed lines in (b) connect data points from the same sample with garnets having more radiogenic isotopic compositions Uncertainties forreconstructed whole-rocks combine 20 modal uncertainty with uncertainties on each mineralrsquos isotope composition Error bars on individualminerals may be smaller than the symbols Also shown are the isotopic compositions of the depleted mantle (DM DePaolo 1981 Griffin et al2000 Bell amp Tilton 2002) Range of 87Sr86Sr in eclogites from Ekati (Lac de Gras area 100 km north of Diavik Jacob et al 2003) in NorthAmerican kimberlites and compositions of low-temperature eclogites from Diavik and of Lac de Gras A154 South kimberlite fromSchmidberger et al (2007)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1864

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

one low-Mg eclogite has eNd similar to the depleted mantle(Fig 12b)Clinopyroxene in the phlogopitized eclogite YK1911

has unradiogenic Nd and Hf and strongly radiogenicSr whereas the coexisting garnet has radiogenic Nd andHf Although this sample contains 50 vol phlogopiteits whole-rock isotopic composition was calculated

from garnet and cpx alone because the contributionfrom phlogopite to the Hf and Nd budget is 52 and51 respectively (unpublished data) The whole-rockhas radiogenic Nd but has lower eHf than anyother sample analysed in accord with the extreme Hfabundance (53 ppm) and low LuHf of the cpx

DISCUSS IONOrigins of eclogites beneath the SlaveCratonHigh-Mg and high-Ca eclogites

With regard to their major- and trace-element contentsthe high-Mg and high-Ca eclogites from Lac de Gras arebroadly similar to variably evolved oceanic gabbros(Fig 13b and c) which may indicate an origin as subductedoceanic crust Compatible trace-element abundancesare useful to deduce potential crustal protoliths as theseelements are less susceptible to post-formation modification(Jacob 2004) The flat HREE patterns in the high-Mg andhigh-Ca eclogites are compatible with a low-pressureorigin as the oceanic crust is a product of moderate degreesof partial melting mainly within the spinel stability field(eg Presnall et al 2002) whereas melting with residualgarnet would qualitatively be expected to produce stronglyfractionated REE patterns Positive Sr and Pb anomaliesalthough compatible with a plagioclase signature couldalso be due to cryptic metasomatism (Dawson 1984) ofthe eclogites during their residence in the lithosphericmantle These anomalies are probably not related to kim-berlite infiltration as the whole-rock eclogite compositionswere reconstructed from analyses of fresh mineral coresLow ZrSm similar to that in primitive gabbros is

observed for many of the high-Mg and high-Caeclogites (Fig 13b and c) and is independent of the ratioof garnetcpx and whether or not the whole-rock wasreconstructed with rutile (Electronic Appendix 10) Thenegative trend of Mg-number against Yb in garnet(Fig 14) argues against igneous processes involving garnetaccumulation (Taylor amp Neal 1989) as suggested forexample for some high-Mg eclogites from West Africa(Barth et al 2002b) Rather the anti-correlation suggeststhat olivine fractionation played a role in the formation ofthe protoliths Fractionating olivine would excludeYb (Dolivinebasalt

frac14 002 Nikogosian amp Sobolev 1997)and incorporate Mg relative to Fe (Dfrac1415 and 05 respec-tively Taura et al1998) leading to progressive depletion ofMg and enrichment of Yb in the residual melt and henceanti-correlated Mg and Yb in the whole-rock This isinherited after eclogitization All eclogite types (exceptVR43452) plot along the olivine fractionation trend sug-gesting that they formed by a similar processHigh-Mg eclogites have lower

PREE than high-Ca

eclogites and some appear to have positive Eu anomalies

high-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

gabbrosSE Indian ridge

gabbrosOman ophiolite

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

high-Mg

RockNMORB(a)

VR43452

YK1949

low-Mg

boninite-like melts

01

001

1

10

(b)

01

001

1

10

(c)

01

001

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

Fig 13 Trace-element patterns of (a) low-Mg (b) high-Mg and(c) high-Ca eclogites compared with boninite-like volcanic rocksfrom the Isua greenstone belt (Polat et al 2002) gabbros from theOman ophiolite (Benoit et al 1996) and the SE Indian Ridge (Hartet al 1999) normalized to N-MORB of Sun amp McDonough (1989)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1865

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(EuEu up to 14) similar to more primitive cumulates(troctolite gabbros and olivine gabbros) reported in the lit-erature (Fig 13b) Although the sizes of these Eu anomaliesare not outside the analytical uncertainty their correlationwith CeYb in reconstructed eclogites supports their signif-icance (r2frac14066 nfrac14 4) The more primitive nature ofhigh-Mg eclogites is consistent with higher Mg-numbersrelative to high-Ca eclogites and may be due to olivineaccumulation during formation of their protolithsThe two samples with the lowest HREE abundanceshave the highest EuEu which is compatible with varia-tions in plagioclase modesMore evolved protoliths with higher pyroxeneplagio-

clase plus trapped melt could account for small or absentpositive Eu anomalies and higher

PHREE of high-Ca

eclogites (Fig 13c) Although the reason for the frequentassociation of graphite or diamond with high-Ca eclogitesis unclear (Aulbach et al 2003) the frequent occurrence ofkyanite in high-Ca eclogites also observed at Ekati justnorth of Lac de Gras has been argued to support a crustalprotolith (Jacob 2004) Kyanite andor corundum havebeen suggested to exsolve from high-temperature high-Alclinopyroxenes (Caporuscio amp Smyth 1990) Eclogitesfrom Lac de Gras show no exsolution textures suggestiveof such an origin although recrystallization can obscureor destroy microstructural evidence for exsolution(eg Griffin et al 1984)Modelling shows that qualitatively plagioclase accumu-

lation from a low-pressure melt derived from MORBmantle can account for positive Pb Sr and Eu and negativeZr anomalies relative to N-MORB whereas cpx accumula-tion or trapped melt increases Zr Y MREE and HREE

abundances (Fig 15) Some melt addition (for exampleby entrapment) seems to be required to increase all trace-element abundances to those observed for the eclogitesAlthough high-Ca and high-Mg eclogites could representshallow and deep portions of the same oceanic crust thehigher average temperatures obtained for the high-Caeclogites the frequent occurrence of carbon in high-Caeclogites and its absence in high-Mg eclogites the strongLREEHREE fractionation observed for six of seven high-Ca eclogites and the narrow range of compositions of thehigh-Mg eclogites suggest that they may have protolithsthat are not cogenetic Some discrepancies in particularaffecting the incompatible elements may be due to post-crystallization processes acting upon the protoliths suchas dehydration melting and later metasomatism that isevident in peridotites from the same locality (Griffin et al1999)Low K2OTh for high-Ca and high-Mg eclogites

(average 08 Table 4) relative to assumed crustal

0

2

4

6

8

10

30 40 50 60 70 80

Mg

VR43452

Yb

GARNET

pyroxenites

low-Mghigh-Mghigh-Ca

eclogites

VR19673ecl-g2

Fig 14 Garnet Mg-number vs Yb (ppm including 2s uncertainty)Lines qualitatively show separate correlations for eclogites (exceptoutlierVR43452) and pyroxenites (except outlierVR19673ecl-g2)

high-Ca

high-Mg

001

01

1

10

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

melt08 melt02 cryst cpx01 melt09 cryst

plag

(a)

001

01

1

10(b)

RockNMORB

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 15 Reconstructed (a) high-Ca and (b) high-Mg eclogitescompared with compositions of plagioclase and cpx crystallizingfrom a MORB melt modelled using distribution coefficients ofNorman et al (2005)The MORB was calculated as a 10 equilibriummelt of MORB pyrolite [composition from Kelemen et al (1993)except Pb Pr Sr Y and Lu which were calculated as residues from1 fractional melting of pyrolite from McDonough amp Sun (1995)using melt modes of Kinzler (1997) and distribution coefficients ofNorman et al (2005)] Also shown are mixtures of crystals (consistingof 50 plagioclase plus 50 cpx) and melt in the proportions 0901and 0208

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1866

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

precursors (eg gabbros SE Indian Ridge average 22Hart et al 1999) may reflect the preferred loss of fluid-mobile elements during dehydration (Becker et al 2000)although large uncertainties on these low-abundance ele-ments in the constituent minerals and reconstructedwhole-rocks makes this interpretation less than robustThe REE patterns of most high-Mg eclogites are too flatto represent melting residues as partial melting shouldproduce low LREEHREE (Electronic Appendix 10) Incontrast the marked LREE depletion of five of sevenhigh-Ca eclogites (LaLufrac14 02^08) contrasts with that ofassumed evolved gabbroic protoliths (LaLufrac1415^32Hart et al 1999) and can be modelled as a consequence ofpartial melting during subduction after eclogitization (iein the presence of garnet which retains HREE relative toLREE eg Barth et al 2002a) has taken place (Aulbachet al 2003)Our results are compatible with previous studies on

eclogite xenoliths from the Slave Craton includingthose from the same kimberlite that have linked thesesamples or some of these samples to Paleoproterozoicsubduction events (the 195^191Ga formation of theHottah terrane and the 18Ga formation of the GreatBear magmatic arc) in the Slave craton (Griffin et al1999 Kopylova et al 1999 Heaman et al 2002 2006Schmidberger et al 2005 2007)The compositional features discussed above although

compatible with crustal protoliths are not necessarilydefinitive of a subduction origin However if direct crystal-lization from basaltic melts at high pressure (ie mantledepths) is invoked instead specific conditions would haveto be met to satisfy the observations The Mg-number ineclogites from Lac de Gras are lower than those expectedof primary high-pressure mantle melts (eg Walter 1998OrsquoHara amp Herzberg 2002) requiring crystallization athigh pressures after olivine fractionation This originwould further require shielding from peridotitic wallrockwith which the fractionated melts would no longer be inequilibrium (Barth et al 2001) In such a setting the lowrelative abundances of fluid-mobile elements comparedwith those of less fluid-mobile elements and of LREErelative to HREE (Table 4) would reflect an expulsion offluids during crystallization of the protoliths and latermelt mobilization respectively The eclogites clearly donot represent garnet plus cpx cumulates which wouldrequire a positive correlation of Mg-number with Yb thatis at variance with the observed negative correlation ineclogitic garnets from this study (Fig 14)

Low-Mg eclogites

Garnets in low-Mg eclogites compositionally overlap withthose in massive eclogites from Jericho (NW of the Lacde Gras kimberlites) that were suggested to have a high-pressure intrusive origin based on the similarity of theirmajor-element compositions to those of Group A eclogites

fromsouthernAfrica forwhichamantleoriginhasbeenpro-posed (Kopylova et al1999) Garnets in low-Mg eclogites inthe present study have major-element compositions moresimilar to garnets in groupB rather thangroupAeclogitesThe positive slope in N-MORB-normalized HREE

abundances displayed by the low-Mg eclogites is clearlydistinct from other eclogite types and is similar to those ofboninite-like second-stage melts that are closely associatedwith ultramafic volcanics (Smithies et al 2004) (Fig 13a)The range of Mg-numbers in the low-Mg eclogites (55^66)overlaps with that of Archaean second-stage melts fromIsua (58^75) and Abitibi (52^82 Smithies et al 2004)These similarities might indicate that the low-Mg eclogiteshave an origin in a previously depleted mantle source suchas supra-subduction zone mantle Subchondritic ZrHf ofthe low-Mg eclogites contrasts with the supra-chondriticZrHf of the high-Mg and high-Ca eclogites (Aulbach et

al in preparation) and is consistent with previous cpx-controlled depletion in an arc mantle source This sourcecould have been subsequently remelted to generate theboninite-like precursor to the low-Mg eclogites as sug-gested for the protoliths of eclogites from Udachnaya(Jacob amp Foley 1999)Although some second-stage melts to which the low-Mg

eclogites are similar are closely associated with komatiitesand could contain a plume component the radiogenic ordepleted mantle-like Nd and Hf and unradiogenic Sr inthe low-Mg eclogites do not support plume involvement

Trace-element and Sr^Nd^Hf isotopeconstraints on post-formationmetasomatismSeveral low-Mg and high-Ca eclogites show negativeslopes in the LREE which may be indicative of interactionwith small amounts of LREE-enriched melt (Navon ampStolper 1987) amounting to cryptic metasomatism(ie not accompanied by crystallization of new phases) inthe nomenclature of Dawson (1984) High Nb abundancesrelative to elements of similar compatibility displayed bysome low-Mg and high-Ca eclogites (Fig 10) could resultfrom overestimation of rutile modes However high-CaeclogiteVR40345 is rutile-free and has been reconstructedwithout rutile its high Nb abundance is accompanied byLREE enrichment indicating that metasomatic additionmay be responsibleLow-Mg eclogites show a positive correlation of La with

Nb (Electronic Appendix 11a) indicating concomitantaddition of LREE and Nb This suggests that the metaso-matic agent was a solute-rich fluid or silicate silicic melt ashydrous fluids and carbonatites do not carry enough HFSEto affect the budget of eclogites (Green amp Wallace 1988Klemme et al 1995 Adam et al 1997 Stalder et al 1998)Of the samples that are enriched in Nb all but one (high-Ca eclogiteVR50909) show a simultaneous enrichment in

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1867

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Hf (Electronic Appendix 11b) Together these trace-element relationships suggest that both SmNd and LuHfwere lowered in these samples Their low parent^daughterratios contrast with higher ratios in unenriched samplesand resulted in the ingrowth of a range of Nd and Hf iso-tope compositions Compared with the low-Mg and high-Ca eclogites the high-Mg eclogites have a narrow range ofLa and Nb concentrations which correlates with a narrowrange in Nd and Hf isotope compositions (Fig 12)Precise model ages cannot be calculated because parent^

daughter ratios are available only from in situ analyseswhich have relatively large uncertainties Howeverevidence for long-term enrichment and depletion may bediscerned Low-Mg eclogites even those that are enrichedin LREE have relatively high SmNd as a result of havingoverall positive slopes in their REE patterns relative toMORB whereas the high-Mg eclogites have relativelyflat REE patterns and therefore lower SmNd (Fig 11)This translates into lower eNd for high-Mg eclogites thanfor low-Mg eclogites In contrast they show no systematicdifference with respect to 176Hf177Hf Because the low- andhigh-Mg eclogites are proposed to have different origins asdiscussed above and therefore did not necessarily have thesame initial isotopic composition and formation age theseisotopic differences cannot be used to estimate their agesA long-term coupling of trace-element and isotopic

abundances is also indicated by the fact that low-Mg eclo-gite YK1943 which has experienced Nb and LREE addi-tion has lower eNd (0) and eHf (thorn13) than low-Mg eclogiteVR43469 with higher LuHf and SmNd (eNdfrac14thorn47eHffrac14thorn128) If low-Mg eclogites YK1943 and VR43469originally had identical initial Nd and Hf isotope composi-tions and if the parent^daughter ratios are accurate andwere modified only once during their evolution shortlyafter metamorphism (when their isotopic compositionwould have been reset) this overprint can be calculatedfor both isotopic systems to have occurred at 17GaThis would suggest that Hf and Nd were added duringthe same event In contrast high-Mg eclogites VR67360and YK1949 would have had identical Ndisotope compositions at 10Ga whereas their Hf isotopecompositions are disturbed in that the sample with thehigher parent^daughter ratio has a lower 176Hf177Hf thanthe sample with the lower parent^daughter ratio Theseresults suggest that the LREE and HFSE enrichmentobserved in different samples occurred during separateevents that did not affect the eclogites equallyExtreme HFSE enrichment leading to the growth of

zircon and rutile and LREE addition leading to thegrowth of apatite in eclogites from the Jericho kimberlitein the northern Slave craton have been ascribed to twoseparate metasomatic events at around 18Ga (quasi-coincident with the age of formation and metamorphism)and 13^10Ga respectively (Heaman et al 2002 2006

Schmidberger et al 2005) The authors associated theseevents with Paleoproterozoic subduction and aMesoproterozoic thermal disturbance in the Slave litho-spheric mantle possibly linked to the Mackenzie igneousevent Mesoproterozoic Nd model ages mostly between13 and 14Ga were reported for eclogites from thenearby Ekati kimberlites (Jacob et al 2003) Although wedo not observe zircon or apatite in volatile-free eclogitesfrom Lac de Gras and the age constraints obtained fromour data are weak they support a link between theseevents described for other eclogite populations and themetasomatic enrichment observed in our studyThe Sr isotope data in the present study are similar to

those of Schmidberger et al (2007) who suggested that thedepleted-mantle-like values are consistent with derivationof the protoliths from an oceanic mantle source However87Sr86Sr in volatile-free eclogites in the present study isremarkably constant considering their Nd and Hf isotopevariability (Fig 12) and the variability of trace-elementpatterns (Fig 10) Given the evidence for dehydration andmelting in many of the eclogites fractionation of highlyincompatible Rb from Sr and differential ingrowth ofradiogenic Sr would be expectedThis may indicate a rela-tively recent overprint and considering the similarityof 87Sr86Sr to present-day depleted mantle which is signif-icantly lower than 87Sr86Sr of North American kimber-lites a juvenile (asthenospheric) source

PyroxenitesPyroxenites from Lac de Gras often show relationshipsin their major-element contents that are different fromthose observed in the eclogites For example pyroxeniticgarnets show a negative correlation between CaO andNa2O contents whereas these oxides are positively corre-lated for garnets in eclogites (Fig 5) CaO is well correlatedwith C2O3 in pyroxenitic garnets but not in eclogitic gar-nets (Fig 4) Mg-number is positively correlated withYb abundance for pyroxenitic garnets but anti-correlatedfor eclogitic garnets (Fig 14) This suggests that the pyrox-enites in the present study are not related to the eclogitesby fractionation or accumulation processes as proposedfor the origin of some pyroxenites (Foley et al 2003)Pyroxenites have also been interpreted as the products

of crystal segregation in magma channels with alteredoceanic crust as the magma source this would explaintheir Eu anomalies and stable and radiogenic isotopecompositions (Pearson et al 1993 Becker 1996) As thehigh-pressure liquidus phases of most terrestrial basaltsare cpx and garnet (OrsquoHara amp Herzberg 2002) partialmelts of eclogites with basaltic composition wouldcrystallize cpx and garnet but not opx More than half ofthe pyroxenites from Lac de Gras contain opx suggestingthat they probably did not crystallize directly fromeclogite-derived melts

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1868

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Low-Cr websterites and pyroxenites could be crystalli-zation products of mafic melts that interacted with perido-titic mantle thus becoming more refractory The positivecorrelation of Yb and Mg-number in garnet would allowan origin by fractional crystallization from such meltsAlternatively they formed by reaction of silicic (possiblyeclogite-derived) melts with mantle peridotites(eg Kelemen et al 1993 Aulbach et al 2002) This agreeswith the similarity of hypothetical melts in equilibriumwith pyroxenitic garnets in most samples to eclogite-derived melts (Electronic Appendix 12) Niobium concen-trations in all hypothetical melts are much higher than ineclogite-derived melts regardless of whether or not rutileis assumed to be in the residue The same is true for LaCe and Zr concentrations in some samples which mayindicate that NbZr and LREE were added to thepyroxenites some time after their formation as discussedin the previous section for some eclogites Because mini-mum ages for pyroxenitic rutile with an unradiogenicHf isotope composition (calculated by projecting the176Hf177Hf to the model depleted mantle isotopic evolutioncurve assuming zero ingrowth) range to 178066Ga(Aulbach et al in preparation) it appears likely that thepyroxenites were not formed recently (eg during kimber-lite-related or precursory melt migration) and that some ofthe samples are at least MesoproterozoicGarnet pyroxeniteYK1952 consists of schlieren of equili-

brated garnet pyroxenite in a lsquomatrixrsquo of spongy cpxfine cpx^opx intergrowths embayed garnet and meltpatches (Fig 2f) Garnet in the pyroxenite schlieren hasvery low LREE and HFSE abundances Orthopyroxenein the matrix has distinctly higher Mg-number andCr-number and contains more CaO than opx from thepyroxenite schlieren whereas opx rims in the schlierenare compositionally similar to opx in the matrix The con-siderable compositional and textural disequilibrium inthis sample suggests that this is a young feature causedby reaction with fluids shortly before or at the time ofentrainmentAs a consequence of the small sample sizes of most of

the pyroxenites studied here a comprehensive isotopicdataset is not available Clearly more such data areneeded in addition to stable isotope data to obtain betterconstraints on the origins of this rock type Consideringthe microstructural and compositional variability ofthe pyroxenites and major-element relationships oppositeto those observed for eclogites it appears certain that pyr-oxenites form by a variety of processes some of which areunrelated to the formation of eclogites and their protoliths

Volatile-rich eclogitesVolatile-rich eclogites from Lac de Gras display the mostextreme Hf^Nd isotope systematics of all the analysedsamples Garnets in phlogopitized (YK1911) and in cal-cite-bearing eclogite (VR67112b) have the most radiogenic

Nd (eNdfrac14 2486 and 1927 respectively) Garnet inVR67112b has extremely high eHf (2516) Clinopyroxene inYK1911 has the lowest eHf measured for any mineralseparate from Lac de Gras (53) in accord with the veryhigh Hf abundance in its cpx (5 ppm) In contrastgarnet in this sample has radiogenic Hf (eHffrac14 136) Thisdiscrepancy is at least in part due to their relatively lowequilibration temperatures (7698C and 8558C respec-tively) which allowed the isotopic compositions of cpxand garnet to evolve separately at their low andhigh parent^daughter element ratios respectively forextended periods of time and possibly since formationAccordingly the cpx in YK1911 gives a minimum Hf ageof 27GaSlab-derived fluids could have been introduced during

the 27Ga collision of the western Slave Craton domainincluding the thick plume-derived deep lithosphericmantle layer and the eastern domain with its highlydepleted lithosphere This would also be consistent withthe observation that volatile-rich eclogites occur at shallowdepths in the mantle (TKroghfrac14 685^8558C correspondingto depths of 85^115 km along a 40 mWm2 geotherm)where the bulk of dehydration and production of solute-rich melts in a subducting slab might be expected to occur(Becker et al 2000 Hermann et al 2006) Their negativeEu anomalies might indicate that the fluid was derivedfrom the shallow portion of the slab that experiencedplagioclase fractionation The presence of carbonate inVR67112b requires a carbonated sourceHigh 87Sr86Sr in cpx in YK1911 is probably the conse-

quence of isotopic equilibration with high-RbSr phlogo-pite after long-term radiogenic ingrowth while Hfaddition led to retarded ingrowth of radiogenic HfThese systematics combined with radiogenic Nd appear

inconsistent with the involvement of sediments in the pet-rogenesis of this sample (eg Perfit et al 1980 Vervoortet al 1999) or a hydrous fluid such as those responsible forMARID metasomatism in the Kaapvaal craton (Konzettet al 1998)Whatever the origin of the fluids the apparentHf mobility suggests that they were not in equilibriumwith a rutile-bearing source and also that they weresolute-rich in nature (ie hydrous melts able to dissolveHFSE) and the relatively high HREE abundances of thevolatile-rich eclogites indicate a garnet-free source suchas amphibolite This allowed evolution of the sample atrelatively high time-integrated SmNd

SUMMARY AND CONCLUSIONS

(1) The Lac de Gras kimberlites carried abundant eclo-gite and pyroxenite xenoliths to the surface Threemain eclogite types are distinguished in terms ofmajor-element composition and trace element

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1869

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

patterns high-Mg high-Ca and low-Mg eclogitesThe garnets of all eclogite types have anti-correlatedMg-number and Yb pointing to olivine fractionationas a controlling factor in the formation of their pre-cursor rocks Their flat or positively sloped HREEpatterns qualitatively indicate that the melt parentalto the precursor rocks was not in equilibrium with agarnet-bearing residue

(2) Reconstructed high-Mg eclogites and high-Caeclogites have flat HREE patterns positive Pb andSr anomalies and low ZrSm similar to oceanic gab-bros reported in the literature High-Mg eclogiteshave variably low HREE contents and some havesmall positive Eu anomalies possibly suggestive ofmore primitive relatively plagioclase-rich protolithswhereas high-Ca eclogites have higher HREE abun-dances and small or absent Eu anomalies that requiremore cpx- andor melt-rich protoliths if a subductionorigin is assumed

(3) Low-Mg eclogites have distinctive trace-element pat-terns that show a marked positive slope in the HREEand subchondritic ZrHf They resemble someArchaean second-stage melts (lsquoboninitesrsquo) and theirprotoliths appear to have formed from a previouslydepleted source

(4) After formation the eclogites were subject to a varietyof processes Most have low ratios of fluid-mobile vsless fluid-mobile elements possibly as a result of dehy-dration of the protoliths In addition some of the low-Mg and high-Ca eclogites have low LREEHREEconsistent with partial melt lossWhereas the low-Mgeclogites show concomitant enrichment in La Nb andZr some high-Mg and high-Ca eclogites show evi-dence for variable and unsystematic (decoupled)LREE and Nb addition

(5) Isotopically the trace-element characteristics of thedifferent eclogite types translate into lower eNd forhigh-Mg eclogites (negative slope in MREEfrac14 lowerSmNd) than for low-Mg eclogites (positive slope inMREEfrac14higher SmNd) Within the low-Mg groupsamples that show evidence for metasomatism (lowLuHf and SmNd) have lower eNd and eHf than asample that was apparently not metasomatized BothNd and Hf addition can be dated to 17Ga if identi-cal initial isotope compsitions are assumed for thelow-Mg eclogites In contrast the relatively constantand depleted mantle-like 87Sr86Sr of eclogitic cpxdespite variable Nd and Hf isotope compositions andparent^daughter ratios suggests recent input of juve-nile material

(6) Garnet and cpx in pyroxenites show major-elementrelationships different from those in eclogitesTherefore these two rock types are probably notrelated by fractionation processes The presence of

opx precludes direct crystallization from slab-derivedmelts Pyroxenites may be the product of (slab-derived) melts hybridized during assimilation of peri-dotitic mantle

(7) Garnets in phlogopite- and carbonate-bearing eclo-gites have extreme isotopic compositions includingthe most radiogenic Nd and Hf in the datasetNegative Eu anomalies may point to the involvementof a crustal protolith Highly radiogenic Sr radiogenicNd and unradiogenic Hf in the reconstructedphlogopite-bearing sample combined with highHREE abundances and evidence for Hf mobilitymay indicate involvement of solute-rich fluids origi-nating in a garnet- and rutile-free source

ACKNOWLEDGEMENTSHelp in the clean laboratories and with the analyticalfacilities by S Elhlou C Lawson A Sharma P Wielandand M Zhang is greatly appreciated Bill Griffin isthanked for his support and for agreeing to reviewseveral versions of the manuscript despite disagreeingwith many of the conclusions presented here The paperbenefited greatly from the detailed and constructivereviews provided by the editors Gareth Davies andMarjorie Wilson and by Matthias Barth Larry Heamanand an anonymous reviewer This work was funded by aMacquarie University International PostgraduateAward and Postgraduate Research Fund (SA) by anARC SPIRT grant sponsored by Kennecott Canada Incand by ARC Large and Discovery Grants to SYOrsquoRAnalytical data were obtained using instrumentationfunded by ARC LIEF and DEST Systemic InfrastructureGrants industry partners and Macquarie UniversityThis is Publication 481 from the ARC National KeyCentre for Geochemical Evolution and Metallogeny ofContinents (wwwesmqeduauGEMOC)

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online

REFERENCESAdam J GreenT H Sie S H amp Ryan C G (1997) Trace element

partitioning between aqueous fluids silicate melts and mineralsEuropeanJournal of Mineralogy 9 569^584

Aulbach S Stachel T Viljoen K S Brey G P amp Harris J W(2002) Eclogitic and websteritic diamond sources beneath theLimpopo Beltccedilis slab-melting the link Contributions to Mineralogy

and Petrology 143 56^70Aulbach S GriffinW L Pearson N J OrsquoReilly S Y Kivi K amp

Doyle B (2003) Origins of eclogites beneath the central SlaveCraton Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA-0011

Aulbach S Griffin W L Pearson N J OrsquoReilly S Y amp Kivi K(2004a) Mantle formation and evolution Slave Craton constraints

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1870

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

from HSE abundances and Re^Os isotope systematics of sulfideinclusions in mantle xenocrysts Chemical Geology 208 61^88

Aulbach S Griffin W L OrsquoReilly S Y amp McCandless T E(2004b) Genesis and evolution of the lithospheric mantle beneaththe Buffalo HeadTerrane Alberta (Canada) Lithos 77 413^451

Aulbach S GriffinW L Pearson N J OrsquoReilly S Y amp Doyle B(2007) Lithosphere formation in the central Slave Craton(Canada) plume subcretion or lithosphere accretion Contributionsto Mineralogy and Petrology (in press) doi 101007s00410-007-0200-1

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2001)Geochemistry of xenolithic eclogites from West Africa Part IA link between low MgO eclogites and Archean crust formationGeochimica et Cosmochimica Acta 65 1499^1527

Barth M G Foley S F amp Horn I (2002a) Partial melting inArchean subduction zones constraints from experimentally deter-mined trace element partition coefficients between eclogitic miner-als and tonalitic melts under upper mantle conditions PrecambrianResearch 113 323^340

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2002b)Geochemistry of xenolithic eclogites from West Africa part 2Origins of the high MgO eclogites Geochimica et Cosmochimica Acta

66 4325^4345Beard B L Fraracci K NTaylor L A Snyder G A Clayton R

A Mayeda T K amp Sobolev NV (1996) Petrography andgeochemistry of eclogites from the Mir kimberliteYakutia RussiaContributions to Mineralogy and Petrology 125 293^310

Becker H (1996) Geochemistry of garnet peridotite massifs fromlower Austria and the composition of deep lithosphere beneatha Palaeozoic convergent plate margin Chemical Geology 134 49^65

Becker H Jochum K P amp Carlson R W (2000) Trace elementfractionation during dehydration of eclogites from high-pressureterranes and the implications for element fluxes in subductionzones Chemical Geology 163 65^99

Bell K amp Tilton G R (2002) Probing the mantle The story fromcarbonatites EOS Transactions American Geophysical Union 83275^277

Benoit M Polvecurren M amp Ceuleneer G (1996) Trace element andisotopic characterisation of mafic cumulates in a fossil mantlediapir (Oman ophiolite) Chemical Geology 134 199^214

BleekerW (2003) The late Archean record a puzzle in ca 35 piecesLithos 71 99^134

BleekerW Ketchum JW F JacksonV A amp Villeneuve M (1999)The central Slave Basement Complex Part I Its structural topol-ogy and autochthonous core Canadian Journal of Earth Sciences 361083^1109

Blichert-Toft J amp Albare de F (1997) The Lu^Hf isotope geochemis-try of chondrites and the evolution of the mantle^crust systemEarth and Planetary Science Letters 148 243^258

Blichert-Toft J Chauvel C amp Albare de F (1997) Separation of Hfand Lu for high precision isotope analysis of rock samples by mag-netic sector-multiple collector ICP-MS Contributions to Mineralogy

and Petrology 127 248^260Brey G P amp Kolaquo hler T (1990) Geothermobarometry in four-phase

lherzolites II New thermobarometers and practical assessment ofexisting thermobarometers Journal of Petrology 31 1353^1378

Brey G P Kolaquo hlerT amp Nickel K G (1990) Geothermobarometry infour-phase lherzolites I Experimental results from 10 to 60 kbJournal of Petrology 31 1313^1352

Caporuscio F amp Smyth J (1990) Trace element crystal chemistryof mantle eclogites Contributions to Mineralogy and Petrology 105550^561

Carbno G B amp Canil D (2002) Mantle structure beneath the SWSlave Craton Canada constraints from garnet geochemistry in theDrybones Bay Kimberlite Journal of Petrology 43 129^142

Coleman R G Lee E D Beatty L B amp Brannock W W (1965)Eclogites and eclogites their differences and similaritiesGeological Society of America Bulletin 76 483^508

Creaser R A Grulaquo tter H Carlson J amp Crawford B (2004)Macrocrystal phlogopite Rb^Sr dates for the Ekati propertykimberlites Slave Province Canada evidence for multiple intru-sive episodes in the Paleocene and Eocene Lithos 76 399^414

Davies R M GriffinW L Pearson N J Andrew A S Doyle BJ amp OrsquoReilly S Y (1999) Diamonds from the deep Pipe DO-27Slave Craton Canada In Gurney J J Gurney J L Pascoe MD amp Richardson S H (eds) Proceedings of the 7th International

Kimberlite Conference CapeTown Red Roof Design pp 148^155Davies R M Griffin W L OrsquoReilly S Y amp Doyle B J (2004)

Mineral inclusions and geochemical characteristics of microdia-monds from the DO27 A154 A21 A418 DO18 DD17 and RanchLake kimberlites at Lac de Gras Slave Craton Canada Lithos 7739^55

DavisW J amp Hegner E (1992) Neodymium isotopic evidence for thetectonic assembly of Late Archean crust in the Slave Provincenorthwest Canada Contributions to Mineralogy and Petrology 111493^504

Davis W J Jones A G Bleeker W amp Grutter H (2003)Lithosphere development in the Slave craton a linked crustal andmantle perspective Lithos 71 575^589

Dawson J B (1984) Contrasting types of upper-mantle metasoma-tism In Kornprobst J (ed) Kimberlites II The Mantle and

Crust^Mantle Relationships Amsterdam Elsevier pp 289^294DePaolo D J (1981) Neodymium isotopes in the Colorado Front

Range and crust^mantle evolution in the Proterozoic Nature 291193^196

Erlank A J amp Kushiro I (1970) Potassium contents of syntheticpyroxenes at high temperatures and pressures Carnegie Institution ofWashingtonYearbook 68 433^439

Foley S F Barth M G amp Jenner G A (2000) Rutilemelt partitioncoefficients for trace elements and an assessment of the influence ofrutile on the trace element characteristics of subduction zonemagmas Geochimica et Cosmochimica Acta 64 933^938

Foley S F Buhre S amp Jacob D E (2003) Evolution of the Archaeancrust by delamination and shallow subduction Nature 421 249^252

Graham I Burgess J L Bryan D Ravenscroft P J Thomas EDoyle B J Hopkins R amp Armstrong K A (1999) Explorationhistory and geology of the Diavik Kimberlites Lac de GrasNorthwest Territories Canada In Gurney J J Gurney J LPascoe M D amp Richardson S H (eds) Proceedings of the 7th

International Kimberlite Conference Cape Town Red Roof Designpp 262^279

Green D H amp Wallace M E (1988) Mantle metasomatism byephemeral carbonatite melts Nature 336 459^462

Green T H Blundy J D Adam J amp Yaxley G M (2000) SIMSdetermination of trace element partition coefficients betweengarnet clinopyroxene and hydrous basaltic liquids at 2^75 GPaand 1080^12008C Lithos 53 165^187

GriffinW L amp OrsquoReilly SY (2007) Cratonic lithospheric mantle Isanything subducted Episodes 30(1) 43^53

Griffin W L Wass S Y amp Hollis J D (1984) Ultramafic xenolithsfrom Bullenmerri and Gnotuk maars Victoria Australia petrol-ogy of a sub-continental crust^mantle transition Journal of

Petrology 25 53^87GriffinW L Doyle B J Ryan C G Pearson N J OrsquoReilly SY

Davies R Kivi K van Achterbergh E amp Natapov LM (1999)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1871

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Layered mantle lithosphere in the Lac de Gras Area SlaveCraton composition structure and origin Journal of Petrology 40705^727

Griffin W L Pearson N J Belousova E Jackson S E vanAchterbergh E OrsquoReilly SY amp Shee S R (2000)The Hf isotopecomposition of cratonic mantle LAM-MC-ICPMS analysis ofzircon megacrysts in kimberlites Geochimica et Cosmochimica Acta

64 133^147Griffin W L OrsquoReilly S Y Doyle B J Pearson N J

Coopersmith H Kivi K Malkovets V amp Pokhilenko N(2004) Lithosphere mapping beneath the North American plateLithos 77 873^922

Grulaquo tter H S Apter D B amp Kong J (1999) Crust^mantle couplingevidence from mantle-derived xenocrystic garnets In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 307^312

Hart S R Blusztajn J Dick H J B Meyer P S ampMuehlenbachs K (1999) The fingerprint of seawater circulationin a 500-meter section of ocean crust gabbros Geochimica et

Cosmochimica Acta 63 4059^4080Harte B amp Kirkley M B (1997) Partitioning of trace elements

between clinopyroxene and garnet Data from mantle eclogitesChemical Geology 136 1^24

Heaman L M Creaser R A amp Cookenboo H O (2002) Extremeenrichment of high field strength elements in Jericho eclogitexenoliths A cryptic record of Paleoproterozoic subduction partialmelting and metasomatism beneath the Slave craton CanadaGeology 30 507^510

Heaman L M Kjarsgaard B A amp Creaser R A (2004) The tem-poral evolution of North American kimberlites Lithos 76 377^397

Heaman L M Creaser R A Cookenboo H O amp Chacko T(2006) Multi-stage modification of the northern Slave mantlelithosphere Evidence from zircon- and diamond-bearing eclogitexenoliths entrained in Jericho Kimberlite Journal of Petrology 47821^858

Hermann J Cpandler C Hack A amp Korsakov A (2006) Aqueousfluids and hydrous melts in high-pressure and ultra-high pressurerocks implications for element transfer in subduction zones Lithos92 399^417

Hills DV amp Haggerty S E (1989) Petrochemistry of eclogites fromthe Koidu Kimberlite Complex Sierra Leone Contributions to

Mineralogy and Petrology 103 397^422Hoffman P F (1989) Precambrian geology and tectonic history of

North America In Bally A amp Palmer A (eds) The Geology of

North AmericaccedilAn Overview Boulder CO Geological Society ofAmerica pp 447^512

IrelandT R Rudnick R L amp Spetsius Z (1994) Trace-elements indiamond inclusions from eclogites reveal link to Archean granitesEarth and Planetary Science Letters 128 199^213

Irvine G J Pearson D G Kjarsgaard B A Carlson R WKopylova M G amp Dreibus G (2003) Evolution of the litho-spheric mantle beneath Northern Canada a Re^Os isotope andPGE study of kimberlite-derived peridotite xenoliths fromSomerset Island and a comparison to the Slave and Kaapvaalcratons Lithos 71 461^488

Isachsen C E amp Bowring S A (1994) Evolution of the Slave cratonGeology 22 917^920

Jacob D E (2004) Nature and origin of eclogite xenoliths from kim-berlites Lithos 77 295^316

Jacob D E amp Foley S F (1999) Evidence for Archean ocean crustwith low high field strength element signature from diamondifer-ous eclogite xenoliths Lithos 48 317^336

Jacob D Jagoutz E Lowry D Mattey D amp Kudrjavtseva G(1994) Diamondiferous eclogites from Siberiaccedilremnants ofArchean oceanic-crust Geochimica et Cosmochimica Acta 585191^5207

Jacob D E Fung A Jagoutz E amp Pearson D G (2003) Petrologyand geochemistry of eclogite xenoliths from the Ekati kimberlitesarea Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA_239

Jacob D E Bizimis M amp Salters V J M (2005) Lu^Hf and geo-chemical systematics of recycled ancient oceanic crust evidencefrom Roberts Victor eclogites Contributions to Mineralogy and

Petrology 148 707^720Jerde E A Taylor L A Crozaz G Sobolev NV amp SobolevV N

(1993) Diamondiferous eclogites from Yakutia Siberiaccedilevidencefor a diversity of protoliths Contributions to Mineralogy and Petrology

114 189^202Jones A G Ferguson I J Chave A D Evans R L amp

McNeice GW (2001) The electric lithosphere of the Slave cratonGeology 29 423^426

Kelemen P B Shimizu N amp Dunn T (1993) Relative depletionof niobium in some arc magmas and the continental-crust-partitioning of K Nb La and Ce during meltrockreaction in the upper-mantle Earth and Planetary Science Letters 120111^134

Kinzler R J (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition Application to mid-ocean ridge basalt petrogenesis Journal of Geophysical Research 102853^874

Klemme S van der Laan S R Foley S F amp Guenther D (1995)Experimentally determined trace and minor element partitioningbetween clinopyroxene and carbonatite melt under upper mantleconditions Earth and Planetary Science Letters 133 439^448

Konzett J Armstrong R A Sweeney R J amp CompstonW (1998)The timing of MARID metasomatism in the Kaapvaal mantle anion probe study of zircons from MARID xenoliths Earth and

Planetary Science Letters 160 133^145Kopylova M G Russell J K amp Cookenboo H (1999) Mapping the

lithopshere beneath the north central Slave Craton In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 468^479

Krogh E (1988) The garnet^clinopyroxene iron^magnesiumgeothermometerccedila reinterpretation of existing experimentaldata Contributions to Mineralogy and Petrology 99 44^48

Kusky T M (1989) Accretion of Archean Slave Province Geology 1763^67

LeCheminant A N amp Heaman L M (1989) Mackenzie igneousevents Canada Middle Proterozoic hotspot magmatismassociated with ocean opening Earth and Planetary Science Letters

96 38^48LeCheminant A N Richardson D G DiLabio R N W amp

Richardson K A (1996) La recherche de diamants au CanadaGeological Survey of Canada Open File 3228 268pp

MacGregor I D amp Carter J L (1970) The chemistry of clinopyrox-enes and garnets of eclogite and peridotite xenoliths from theRoberts Victor mine South Africa Physics of the Earth and PlanetaryInteriors 3 391^397

MacGregor I D amp Manton W I (1986) Roberts Victor eclogitesancient oceanic crust Journal of Geophysical Research 91 14063^14079

MacKenzie J M amp Canil D (1999) Composition and thermal evo-lution of cratonic mantle beneath the central Archean SlaveProvince NWT Canada Contributions to Mineralogy and Petrology

134 313^324

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1872

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

McCandlessT E amp Gurney J J (1989) Sodium in garnet and potas-sum in clinopyroxene criteria for classifying mantle eclogites InRoss J (ed) Kimberlites and Related Rocks Carlton Blackwellpp 827^832

McDonoughW F amp Sun S-S (1995) The composition of the EarthChemical Geology 120 223^253

Navon O amp Stolper E (1987) Geochemical consequences of meltpercolation the upper mantle as a chromatographic columnJournal of Geology 95 285^307

Nikogosian I K amp Sobolev AV (1997) Ion-microprobe analysis ofmelt inclusions in olivine experience in estimating the olivine^meltpartition coefficients of trace elements Geochemistry International 35119^126

Norman M D Griffin W L Pearson N J Garcia M O ampOrsquoReilly S Y (1998) Quantitative analysis of trace element abun-dances in glasses and minerals a comparison of laser ablationinductively coupled plasma mass spectrometry solution inductivelycoupled plasma mass spectrometry proton microprobe and elec-tron microprobe data Journal of Analytical Atomic Spectrometry 13477^482

Norman M D Garcia M O amp Pietruszka A J (2005) Trace-element distribution coefficients for pyroxenes plagioclase andolivine in evolved tholeiites from the 1955 eruption of KilaueaVolcano Hawairsquoi and petrogenesis of differentiated rift-zonelavas American Mineralogist 90 888^899

OrsquoHara M J amp Herzberg C (2002) Interpretation of trace elementand isotope features of basalts Relevance of field relations petrol-ogy major element data phase equilibria and magma chambermodeling in basalt petrogenesis Geochimica et Cosmochimica Acta 662167^2191

PadghamW A amp FysonW K (1992) The Slave Province a distinctcraton CanadianJournal of Earth Sciences 29 2072^2086

Pearson D G Davies G R amp Nixon P H (1993) Geochemical con-straints on the petrogenesis of diamond facies pyroxenites from theBeni Bousera peridotite massif North Morocco Journal of Petrology34 125^172

Pearson N J Griffin W L Doyle B J OrsquoReilly S Y vanAchterbergh E amp Kivi K (1999) Xenoliths from kimberlitepipes of the Lac de gras area Slave Craton Canada In GurneyJ J Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite ConferenceCapeTown Red Roof Design pp 644^658

Perfit N R Gust D A Bence A E Arculus R J amp Taylor S R(1980) Chemical characteristics of island-arc basalts implicationsfor mantle sources Chemical Geology 30 227^256

Polat A Hofmann A W amp Rosing M T (2002) Boninite-likevolcanic rocks in the 37^38 Ga Isua greenstone belt WestGreenland geochemical evidence for intra-oceanic subductionzone processes in the early Earth Chemical Geology 184 231^254

Poudjom DjomaniY H GriffinW L OrsquoReilly SY amp Doyle B J(2005) Lithospheric domains and controls on kimberlite emplace-ment Slave Province Canada evidence from elastic thickness andupper mantle composition Geochemistry Geophysics Geosystems 6doi1010292005GC0009978

Presnall D C Gudfinnsson G H amp Walter M J (2002)Generation of mid-ocean ridge basalts at pressures from 1 to 7GPa Geochimica et Cosmochimica Acta 66 2073^2090

Schmidberger S S Heaman L M Simonetti A Creaser R A ampCookenboo H O (2005) Formation of Paleoproterozoiceclogitic mantle Slave Province (Canada) Insights from in-situHf and U^Pb isotopic analyses of mantle zircons Earth and

Planetary Science Letters 240 621^633

Schmidberger S S Simonetti A Heaman L M Creaser R A ampWhiteford S (2007) Lu^Hf in-situ Sr and Pb isotope and traceelement systematics for mantle eclogites from the Diavik diamondmine Evidence for Paleoproterozoic subduction beneath the Slavecraton Canada Earth and Planetary Science Letters 254 55^68

Schulze D J amp Helmstaedt H (1988) Coesite-sanidine eclogitesfrom kimberlite products of mantle fractionation or subductionJournal of Geology 96 435^443

Smithies R H Champion D C amp Sun S S (2004) The case forArchaean boninites Contributions to Mineralogy and Petrology 147705^721

Smyth J R Caporuscio F A amp McCormick T C (1989) Mantleeclogites evidence of igneous fractionation in the mantle Earth andPlanetary Science Letters 93 133^141

Sobolev NV amp Lavrentrsquoyev Y G (1971) Isomorphic sodium admix-ture in garnets formed at high pressure Contributions to Mineralogy

and Petrology 31 1^12Stalder R Foley S F Brey G P amp Horn I (1998) Mineral aqueous

fluid partitioning of trace elements at 900^12008C and 30^57GPa New experimental data for garnet clinopyroxene and rutileand implications for mantle metasomatism Geochimica et

Cosmochimica Acta 62 1781^1801Sun S-S amp McDonough W F (1989) Chemical and isotopic sys-

tematics of oceanic basalts implications for mantle compositionsand processes In Saunders A D amp Norris M J (eds)Magmatism in the Ocean Basins Geological Society London Special

Publications 42 313^345Taura H Yurimoto H Kurita K amp Sueno S (1998) Pressure

dependence on partition coefficients for trace elements between oli-vine and the coexisting melts Physics and Chemistry of Minerals 25469^484

Taylor L A amp Neal C R (1989) Eclogites with oceanic crustal andmantle signatures from the Bellsbank kimberlite South AfricaPart I Mineralogy petrography and whole rock chemistryJournal of Geology 97 551^567

van Achterbergh E Ryan C G amp Griffin W L (1999) Glitter online intensity reduction for the laser ablation inductively coupledplasma spectrometry Abstract 9th V M GoldschmidtConference LPI Boston MA 905 pp Abstract no 7215

van Breemen O DavisW J amp King J E (1992) Temporal distribu-tion of granitoid plutonic rocks in the Archean Slave ProvinceNorthwest Canadian Shield Canadian Journal of Earth Sciences 292186^2199

Vervoort J D Patchett P J Blichert-Toft J amp Albare de F (1999)Relationship between Lu^Hf and Sm^Nd isotopic systems in theglobal sedimentary system Earth and Planetary Science Letters 16879^99

Viljoen K S Smith C B amp Sharp Z D (1996) Stable and radio-genic isotope study of eclogite xenoliths from the Orapa kimberliteBotswana Chemical Geology 131 235^255

Walter M J (1998) Melting of garnet peridotite and theorigin of komatiite and depleted lithosphere Journal of Petrology 3929^60

Wasserburg G J Jacobsen S B DePaolo D J McCulloch M Tamp Wen T (1981) Precise determination of SmNd ratios Sm andNd isotopic abundances in standard solutions Geochimica et

Cosmochimica Acta 45 2311^2323Westerlund K J Shirey S B Richardson S H Carlson R W

Gurney J J amp Harris JW (2006) A subduction wedge origin forPaleoarchean peridotitic diamonds and harzburgites from thePanda kimberlite Slave craton evidence from Re^Os isotope sys-tematics Contributions to Mineralogy and Petrology 152 275^294

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1873

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Page 8: Origins of Xenolithic Eclogites and Pyroxenites from the Central

Table 2 Summary of major-element contents in garnet and cpx (wt )

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO Na2O Total Mg-no T(Krogh) Type

Garnet

Low-Mg ecl 3964 026 2214 011 1843 037 1050 846 009 1000 503 976

minimum 3893 008 2130 5009 1071 024 849 567 006 990 423 860

maximum 4094 036 2275 011 2093 046 1457 1062 012 1004 708 1088

1s 070 012 052 na 354 007 227 185 003 05 99 98

High-Mg ecl 4081 026 2261 013 1419 030 1537 624 007 1000 659 969

minimum 3970 008 2224 5009 951 023 1127 418 003 991 551 820

maximum 4250 052 2323 035 1795 036 2030 882 013 1012 792 1241

1s 077 013 031 na 225 004 189 141 003 05 57 88

High-Ca ecl 3979 034 2211 na 1516 027 958 1240 011 998 528 1078

minimum 3909 011 2117 5009 1164 023 801 1144 007 989 456 910

maximum 4039 073 2242 011 1713 035 1249 1351 018 1004 657 1246

1s 042 026 042 na 173 004 155 073 004 06 67 151

vr19674ecl-g7 3867 003 2187 5009 2380 153 793 627 5004 1002 373 volatile-rich

vr67112b 3783 013 2097 009 2620 204 466 790 5004 999 241 volatile-rich

yk1911 3874 5006 2200 5009 2120 221 802 747 5004 997 403 volatile-rich

Pyroxenites 4163 021 2257 075 1148 034 1906 407 006 1002 747 909

minimum 3989 005 2081 009 729 025 1535 277 004 996 593 699

maximum 4279 048 2330 260 1881 054 2215 566 009 1011 842 1158

1s 068 010 057 057 323 007 209 065 001 04 73 117

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO Na2O Total Mg-no Type

Cpx

Low-Mg ecl 5453 028 739 na 500 5009 1117 1662 434 994 801

minimum 5160 017 564 5009 337 na 876 1388 197 981 741

maximum 5570 044 1067 014 687 na 1324 1926 632 1001 865

1s 128 011 155 na 124 na 153 177 126 06 40

High-Mg ecl 5483 023 483 na 381 na 1368 1875 318 994 865

minimum 5353 009 263 5009 245 5009 1156 1617 204 986 778

maximum 5645 031 810 016 674 011 1588 2101 495 1011 920

1s 071 007 137 na 095 na 104 138 075 06 30

High-Ca ecl 5563 027 1184 na 318 5009 868 1368 615 995 836

minimum 5448 014 769 5009 168 na 619 1060 391 987 751

maximum 5670 049 1579 010 603 na 1148 1689 807 1003 878

1s 076 015 334 na 151 na 189 261 167 06 45

vr19674 ecl-g7 5322 010 185 5009 536 015 1448 2339 062 992 828 volatile-rich

vr67112b 5040 041 328 5009 1079 024 1108 2229 071 992 647 volatile-rich

yk1911 5231 014 301 5009 544 024 1450 2340 056 996 826 volatile-rich

Pyroxenites 5527 na 347 038 304 na 1543 1959 245 999 899

minimum 5299 5006 037 010 149 5009 1128 1408 030 990 804

maximum 5666 037 786 134 490 014 1863 2312 537 1009 947

1s 068 na 189 028 091 na 193 226 131 05 37

SiO2 TiO2 Al2O3 Cr2O3 FeOtotal MnO MgO CaO NiO Total Mg-no

Opx

Pyroxenites 5781 na 059 na 627 na 3508 037 011 1006 909

minimum 5546 5006 032 5009 462 5009 2999 020 008 996 804

maximum 5888 009 318 054 1304 014 3628 077 014 1014 933

1s 098 na 067 na 212 na 160 019 002 06 32

Averaged for the three eclogite types and for pyroxenites minimum and maximum values as well as standard deviationsalso given Concentrations in volatile-rich eclogites are shown individually ecl eclogite Mg-numberfrac14 100Mg(Mgthorn Fe)Temperatures of last equilibration [T(Krogh)] as in Electronic Appendix 8

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1850

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

garnets (these include kyanite- graphite- and diamond-bearing varieties) (2) eclogites with low-Mg garnets (theseinclude quartz-bearing eclogite) (3) eclogites with high-Mggarnets and Cr2O3 contents502wt (4) volatile-richeclogites (phlogopite- apatite- carbonate-bearing)

Clinopyroxene (cpx)Clinopyroxenes in all eclogite groups have higher averageAl2O3 contents than those in pyroxenites with the formerhaving about 90 of the total Al in the jadeite molecule(AlVI) versus about 35 for the latter corresponding toan average of 41 and 25wt Na2O respectivelyThe highest Al2O3 and Na2O contents in the dataset areobserved in cpx from the high-Ca eclogites (up to158wt Al2O3) (Electronic Appendix 3) Contents ofK2O are generally below the detection limit(5004wt ) but values up to 024wt are observed insome cpx in high-Ca eclogites Average cpx Mg-numberare highest in pyroxenites (899) followed by cpx inhigh-Mg eclogites (865) high-Ca eclogites (836) andlow-Mg eclogites (801) Pyroxenitic cpx tend to havelower TiO2 and higher NiO and distinctly higherCr2O3 contents compared with those in eclogites A plotof MgO vs Al2O3 shows a negative correlation (Fig 6)Clinopyroxenes in low-Mg eclogites have lower Al2O3

contents at a given MgO content than other eclogitetypes Between 15 and 17wt MgO the Al2O3^MgOcorrelation for cpx in pyroxenites has a different slopefrom that in the eclogites Clinopyroxenes in pyroxenites

VR19673ecl-g1 and VR19673ecl-g2 have distinctly lowerMgO contents and higher Al2O3 contents than inthe other pyroxenites They coexist with garnets havingthe lowest Cr2O3 and CaO contents of the dataset whichdefine one end of the positive Cr2O3^CaO correlationobserved for garnetsClinopyroxenes in volatile-rich eclogites have distinctly

high MnO contents (up to 024wt ) Mg-numbers forcpx in the phlogopite-bearing variety are 826 and 828whereas that in the calcite-bearing eclogite (VR67112b)has an Mg-number of 647Some cpx grains have spongy rims that contain small

(mm-scale) patches of glass these rims are richer in Caand Mg and markedly poorer in Na than the cores similarto cpx in eclogites from southern Africa (Taylor amp Neal1989) The rim texture is unequilibrated suggesting lategrowth The formation of the spongy cpx rims might berelated to incongruent melting of cpx or be a result ofmetasomatism (Taylor amp Neal 1989 and referencestherein) Distinct although not spongy rims in two eclo-gites (VR67360 and YK1946) are similarly enriched inCaO and depleted in Na2O Two types of cpx with respectto CaO Na2O and Mg are present in VR43477VR43479 and VR43480 without clear core^rim zonationor spongy rims No significant differences are recognizedwith respect to other oxides

Orthopyroxene (opx)Orthopyroxenes have Mg-number ranging from 804 to933 (17 of 19 samples have values between 895 and 933)and typically contain minor amounts of Al2O3

(032^118wt one outlier has 318wt ) CaO(020^118 wt ) and NiO (008^014wt ) (Electronic

0

2

4

6

8

10

12

14

7 9 11 13 15 17 19

Al 2

O3

CPX

pyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

cpx in pyroxenitesVR19673ecl-g1g2

MgO

Fig 6 MgO vs Al2O3 (wt ) in cpx Data sources as in Figs 3and 4

0

005

010

015

020

2 4 6 8 10 12 14 16

CaO

Na 2

O

dl

group Igroup II

pyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

GARNET

Fig 5 CaO vs Na2O (wt ) in garnet Arrows show trends foreclogites and pyroxenites respectively [note that part of thepyroxenite trend extends into the field falling below the detectionlimit (dl) for Na2O] Group I and group II refer to classificationof McCandless amp Gurney (1989) Data sources as in Figs 3 and 4

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1851

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Appendix 4) The contents of all other elements are mostlybelow their respective detection limits

Accessory phasesRutile and ilmenite occur in some of the eclogites (Table 1)All rutile grains have exsolved 10 ilmenite as lamellaeand rims Ilmenite-free areas of rutile in eclogites containbetween 962 and 988wt TiO2 and minor amounts ofAl2O3 (01^04wt ) Cr2O3 (5009 to 02wt ) andFeO (03^07wt ) Rutile in pyroxenite YK1915 contains004wt Al2O3 01wt FeO and 15wt Cr2O3Discrete ilmenite grains in sample VR67112b haveTiO2 contents of 499 and 500wt MgO of 23 and15wt and FeO of 399 and 403wt respectivelySulfides occur interstitially or enclosed in cpx garnet

and opx they are mostly pyrrhotite and monosulfide solidsolution with variable Ni contents (unpublished data)They will not be considered further hereAbundant platy phlogopite is observed in two samples

(VR19674ecl-g7 and YK1911) It is Cl-rich in sampleVR19674ecl-g7 (063wt vs 003wt in YK1911) andF-rich in sample YK1911 (097wt vs 005wt inVR19674ecl-g7) (Electronic Appendix 5) The same is truefor coexisting apatite (VR19674ecl-g7 254wt Cl and014wt F YK1911 006wt Cl and 306wt F)

Comparison with mineral inclusions indiamonds from Lac de Gras and witheclogite xenoliths from other Slave CratonlocalitiesMineral inclusions in diamonds from the Lac de Graskimberlites have been investigated by Davies et al (19992004) For garnets there is a striking similarity betweenthe inclusions in diamond and those in the high-Ca eclogitexenoliths although the inclusions in diamond trendtowards higher Na2O (Fig 5) and lower Al2O3 contentsThe overlap is not as marked for cpx where inclusionsin diamond span almost the entire range observed foreclogitic cpx (Fig 6)A comparison with eclogites from other localities in

the Slave Craton shows that garnets in most Diavik low-temperature eclogites reported by Schmidberger et al(2007) plot with garnets in high-Mg eclogites and inpyroxenites from this study (Figs 3^5) whereas garnets inhigh-temperature and diamond-bearing eclogites(Schmidberger et al 2007) overlap most with garnets inhigh-Ca eclogites from this study and with garnetsincluded in diamond Eclogites from the Jericho kimberlitehave either massive or anisotropic fabrics (Kopylova et al1999) Garnets in almost all massive eclogites plot withgarnets in low-Mg eclogites from this study with regard toCaO^MgO^FeO relationships (Fig 3) but many havelower Na2O contents at a given CaO content (Fig 5)Garnets in anisotropic eclogites are restricted to lowMgOFeO (507) and show some overlap with garnets

in high-Ca eclogites although trending towards lowerCaO(MgOthornFeO) (Fig 3) while coexisting cpx is alsorestricted to low MgO and trends toward high Al2O3 con-tents (Fig 6)

TRACE ELEMENTSGarnetTrace-element abundances in garnet are summarized inTable 3 and the complete dataset is reported in ElectronicAppendix 1 Garnets in low-Mg eclogites show steep posi-tive slopes between LaN and SmN and shallower positiveslopes in the normalized heavy rare earth elements(HREEN Fig 7a) The normalized light REE (LREEN)show an order of magnitude variability ZrHf is mostlysupra-chondritic and Ti can be enriched or depleted rela-tive to elements of similar compatibility Niobium in threesamples is strongly depleted at 002^003 chondriticGarnets in high-Mg eclogites also show smooth positiveslopes in LREEN but have flat HREEN and variablenormalized middle REE (MREEN) to HREEN (Fig 7b)ZrHf is always supra-chondritic and Ti enriched ordepleted relative to similarly compatible elementsNiobium is most samples is below detection Small positiveEu anomalies though not outside the analytical uncer-tainty for Eu relative to either Sm or Gd are observed forgarnet in some high-Mg eclogites (eg EuEu up to 15for VR67360 where Eu is the average of the chondrite-normalized Sm and Gd concentrations) Garnets in high-Ca eclogites are slightly enriched in the MREEN relative tothe HREEN or have flat MREEN and HREEN with ahump between EuN and HoN (Fig 7c) One sample(VR40345) shows strong La enrichment (greater thanchondrite) Two samples have supra-chondritic ZrHfwhereas three samples have close to chondritic ratios andone sample has a subchondritic ratio As is true for theother eclogite typesTiN is either enriched or depleted rela-tive to similarly compatible elementsPyroxenitic garnets have variable slopes in the LREEN

and flat MREEN to HREEN patterns and two samplesshow high Sc relative to neighbouring elements Garnet inone of the pyroxenites (YK1952) has a highly distinctivepattern relative to other pyroxenites with very low LREEand HFSE (Fig 7d) This sample consists of coarse pyroxe-nite schlieren in which the garnet was analysed in a finer-grained lsquomatrixrsquo of spongy cpx with secondary rims finecpx^opx intergrowths embayed garnet and melt patchesVolatile-rich eclogites have garnets that are slightly

enriched in MREEN relative to HREEN or have flatMREE to HREE patterns One garnet has a pronouncednegative Eu anomaly (EuEufrac14 07 forVR67112b Fig7e)Garnets in volatile-rich samples have the lowest Ni Coand Ti and the highest V ScYand HREE abundances inthe dataset

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1852

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 3 Summary of trace-element contents in garnet and cpx (ppm)

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR19674

ecl-g7

VR67112b YK1911 av min max 1s

n 6 6 7 3 5 4 4

P 169 54 270 88 196 70 310 78 282 80 450 138 160 60 80 214 48 370 144

Sc 62 40 89 16 48 36 67 11 53 42 79 15 69 92 93 186 52 440 179

Ti 1667 490 2500 680 1073 620 2000 455 2141 700 4700 1582 270 690 260 1097 540 1900 573

V 169 78 250 68 150 78 280 72 178 110 390 101 140 340 100 390 71 820 326

Co 59 42 70 11 58 39 84 13 58 47 65 5 30 41 27 68 55 90 16

Ni 17 5 32 10 21 11 34 10 24 11 35 8 6 10 51 54 25 110 38

Ga 12 7 15 3 13 8 16 3 12 9 20 4 8 86 58 12 10 14 2

Sr na 512 4 na na 522 522 na na 57 57 na 52 5025 510 513 513 513 na

Y 37 6 61 18 21 7 38 10 31 20 38 5 60 81 70 19 12 26 6

Zr 15 5 29 9 28 6 83 26 24 9 43 12 60 22 90 18 1 41 17

Nb 009 0005 048 019 na 50015 05 na na 5005 05 na 5003 0012 5001 022 0030 057 025

La 005 001 009 004 na 50007 00 na na 5003 04 na 004 002 0009 003 000 009 005

Ce 03 01 05 02 01 00 01 00 05 01 18 05 03 022 015 01 00 03 01

Pr 01 00 02 01 00 00 01 00 02 01 03 00 017 014 012 00 00 01 00

Nd 12 03 24 07 07 04 14 03 21 12 32 07 24 23 20 03 00 07 03

Sm 11 04 22 06 10 03 18 05 23 08 36 10 29 31 33 05 01 09 04

Eu 06 03 10 03 06 02 10 03 11 04 17 04 16 114 16 03 01 06 02

Gd 27 07 41 12 23 07 42 12 43 16 53 13 8 81 90 13 06 24 08

Dy 54 11 70 23 36 13 65 18 58 35 71 11 12 144 12 29 18 42 10

Ho 14 02 23 07 08 03 15 04 12 09 14 02 24 30 23 07 04 10 02

Er 46 07 80 24 24 09 40 10 34 25 40 05 8 85 6 22 14 28 06

Yb 52 06 98 30 24 10 37 08 32 22 39 05 8 70 5 23 13 30 07

Lu 09 01 16 05 04 02 05 01 05 03 07 01 11 09 08 04 02 05 01

Hf 03 02 05 01 04 01 10 03 05 02 09 02 09 027 007 03 00 06 03

Ta na 50008 5004 na na 50005 001 na na 50004 na na 003 50002 50005 na 50007 011 na

Pb na 5005 02 na na 5002 02 na na 5003 07 na 02 004 5002 na 5003 01 na

Th na 5001 005 na na 50007 003 na na 50008 011 na 5007 0012 50004 002 0005 006 002

U na 5003 5003 na na 5001 01 na na 5007 na na 01 0023 010 na 5001 003 na

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR67112b YK1911 av min max 1s

n 6 4 8 5 4 4

P 65 15 120 41 39 14 56 16 53 30 70 12 20 9 80 27 120 40

Sc 22 10 30 7 15 14 19 2 14 7 24 5 80 43 17 11 21 5

Ti 1557 990 2400 508 1168 1010 1490 191 1988 810 3500 1154 2390 1700 689 102 1685 694

V 415 180 480 107 318 210 400 69 324 259 450 72 600 180 250 139 480 156

Co 35 26 50 8 20 14 31 7 26 14 42 11 49 23 29 20 42 10

Ni 164 87 220 47 228 100 430 128 188 135 320 56 59 34 428 410 670 184

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1853

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

ClinopyroxeneTrace-element abundances in cpx are summarized inTable 3 and the complete dataset is reported in ElectronicAppendix 6 Clinopyroxenes in low-Mg eclogites havemildly positive to negative slopes from LaN to NdN andsmooth negative slopes between SmN and HoN thatmostly continue to LuN (Fig 8a) Niobium is below detec-tion or 5007 chondritic and SrN is variably enrichedrelative to neighbouring elements with the peak size corre-lating negatively with LREEN ZrHf is subchondritic andHfN is enriched relative to SmN whereasTiN is enriched ordepleted relative to similarly compatible elementsWe ana-lysed the spongy rim in cpxYK1943 and its trace-elementpattern is similar to that of cpx cores for more compatibleelements whereas the most incompatible elements areseveral orders of magnitude higher Clinopyroxenes in

high-Mg eclogites have steeper negative slopes betweenNdN and LuN than those in low-Mg eclogites combinedwith lower HREE abundances HfN is depleted relative toSmN (Fig 8b) High-Ca eclogites contain cpx with sinusoi-dal REEN patterns with lower LREE than other eclogitetypes and the heaviest REE below detection They haveuniformly strong positive Sr peaks relative to elements ofsimilar compatibility (Fig 9) The diamond-bearing high-Ca eclogite (VR40345) has cpx with unusually high Nband LREE abundancesClinopyroxenes in four pyroxenites have disparate trace

element patterns They all share depletions in TiN relativeto similarly compatible elements and two show positiveSrN peaks Two samples have cpx with high and flatLREEN to MREEN patterns and steep negative slopes inthe HREEN The trace-element patterns of cpx in YK1914

Table 3 Continued

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR67112b YK1911 av min max 1s

n 6 4 8 5 4 4

Ga 16 10 20 3 12 9 15 2 26 15 39 9 129 123 6 11 11 5

Rb na 5008 04 na na 5007 05 02 na 5003 03 na 02 5004 03 010 05 0

Sr 166 85 290 74 222 153 285 50 243 83 440 142 85 120 200 11 414 209

Y 2 04 4 10 13 07 2 04 11 03 3 10 28 51 2 03 3 2

Zr 19 4 48 15 25 8 45 13 17 8 28 7 111 129 36 008 81 42

Nb 01 0003 03 01 na 5002 00 na na 5002 16 9 007 011 010 003 02 008

Cs na 5001 6 na na 5006 5006 na5013 5004 5013 na bdl 5002 na 5001 5002 na

Ba 09 02 2 07 09 01 2 07 10 03 3 10 06 01 13 06 3 10

La 3 03 8 3 3 23 3 04 4 01 14 6 51 97 4 01 10 4

Ce 7 13 20 6 10 7 13 2 7 05 24 10 25 28 11 01 28 13

Pr 11 02 3 08 2 10 3 06 08 02 2 09 47 46 2 002 4 2

Nd 5 12 12 4 10 5 14 4 4 09 8 26 23 22 8 01 22 10

Sm 11 05 21 06 20 10 31 10 07 02 11 03 47 53 2 005 5 2

Eu 03 01 05 01 06 028 09 03 02 010 03 007 087 123 06 002 15 07

Gd 08 03 13 03 11 05 19 06 05 03 06 011 28 37 15 010 36 2

Dy 05 01 07 02 04 022 06 01 02 013 03 007 11 16 07 008 14 06

Ho 01 00 02 004 006 003 007 002 na 5003 012 na 011 021 009 001 02 007

Er na 507 507 na 011 004 02 00 na 5006 03 na 019 033 02 002 03 012

Yb na 5001 02 na na 5001 011 na na 5006 03 na 007 015 009 001 02 006

Lu na 5001 003 na na50002 002 na na50008 01 na 0006 0013 0013 0002 003 001

Hf 12 03 24 07 11 04 18 05 10 06 16 04 34 53 na50015 35 na

Ta na50003 50015 na na50006 006 na na50004 5004 na 0014 00236 na50003 503 na

Pb 06 02 12 04 06 04 09 02 na 5016 07 02 065 098 08 01 18 1

Th na50005 06 na 012 005 02 005 na 5002 5006 na 021 02 001 06 03

U na 5005 5005 na 002 0009 005 001 na 5003 5005 na 0013 024 na50005 003 na

Averaged for the three eclogite types and for pyroxenites as in Table 2 pyx pyroxenite n number of analyses averagedbdl below detection limit

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1854

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

and YK1952 are similar although at different abundanceswith only modest LREEN over HREEN enrichment andstrong depletions in ZrN relative to neighbouring elements(Fig 8d)Clinopyroxenes in the volatile-rich samples have flat pat-

terns and high abundances from LaN to SmN and smoothnegative slopes from SmN to LuN Niobium Sr and Ti are

noticeably depleted relative to elements of similar compat-ibility (Fig 8e)

Trace-element partitioningTo assess whether coexisting phases are in trace-elementequilibrium we calculated apparent cpx^garnet distribu-tion coefficients and compared them with those

(a)

(d)

(b)

(e)

(c)

GarnetCI-chondrite GarnetCI-chondrite

low-Mghigh-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1952

1

10

Nd Sm Eu Gd Dy

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Fig 7 Extended trace-element patterns of garnet in (a) low-Mg eclogites (b) high-Mg eclogites with inset showing enlargement of MREE tohighlight Eu anomalies and 1s uncertainties (c) high-Ca eclogites (d) pyroxenites and (e) volatile-bearing eclogites Normalized to chondritevalues of McDonough amp Sun (1995)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1855

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

determined experimentally or in demonstrably equili-brated natural samples Low-Mg and high-Mg eclogiteshave Dcpxgarnet for most elements that are within therange determined for natural equilibrated eclogites andexperimental assemblages (Harte amp Kirkley 1997 Greenet al 2000 Barth et al 2002a) (Electronic Appendix 7)High-Ca eclogites plot at the low end of the range

of distribution coefficients and have very low DMREE

and DHREE outside the range indicated by the datasetof Harte amp Kirkley (1997) who noted that DREE decreaseswith increasing Ca content in eclogitic minerals Theseassemblages are therefore considered to be in equilibriumPyroxenite VR40384 has D values similar to eclogites

but at the high end of the range determined by Harte amp

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

(a)

(d)

(b)

(e)

(c)

CpxCI-chondrite CpxCI-chondrite

low-Mg high-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1914

YK1952

spongy rim

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

VR40345

Fig 8 Extended trace-element patterns of cpx as in Fig 7

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1856

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Kirkley (1997) (Electronic Appendix 7) Another group(YK1914 and YK1952) plots at the low end but DHf DZrDTi and DV are far below the values indicated by experi-mental datasets Because pyroxenites compositionallygrade into eclogites their drastically different DHFSE

values suggest disequilibrium In contrast volatile-bearingsamples have DHFSE values that appear to be too high to bein equilibrium however this may be a consequence oftheir unusual bulk chemical composition (Ti- and K-rich)which differs substantially from that of lsquonormalrsquo eclogitesand experimental charges

GEOTHERMOBAROMETRYFollowing Schmidberger et al (2007) temperatures oflast equilibration were calculated for eclogitic andopx-free pyroxenitic assemblages using the formulationof Krogh (1988) The results are summarized in Table 2and Electronic Appendix 8 for a preset pressure of5 GPa and illustrated as a histogram in Fig 9 Foropx-bearing pyroxenites temperatures were calculatedsimultaneously with pressure [Ca-in-opx thermometryand Al-in-opx barometry using the formulations of Breyet al (1990)] to constrain their relative positions in themantle columnMinerals in some eclogites from Lac de Gras show dis-

tinct core and rim compositions or inhomogeneitiesTemperatures were not calculated for samples withspatially irresolvable inhomogeneous garnet compositions(SE01VR43479VR43480) For eclogites with spongy cpxcore compositions were used in the calculationTemperatures were calculated for both core and rim com-positions where applicable For all samples these

temperatures agree within the uncertainty of this thermo-meter (508C Brey amp Kolaquo hler 1990)Most equilibration temperatures for eclogites lie between

800 and10008C (Fig 9)The transition between the shallowand the deep lithospheric mantle layers in the central Lacde Gras area lies in this temperature interval (Griffin et al1999 Pearson et al 1999) Eclogites in this study lack thedistinct bimodal temperature distribution observed bySchmidberger et al (2007) but as in their study high-Caeclogites give higher average temperatures (10808C corre-sponding to a depth of 155 km if equilibrated along a 40mWm2 geotherm) than other eclogites types (9808Cdepth 135 km) the diamond-bearing sample (VR40345)gives thehighest temperature of the datasetHighNa2Ocon-tents in garnet from high-Ca eclogites are qualitativelyconsistent with higher equilibration pressures (Sobolev ampLavrentrsquoyev 1971) as are high K2O contents in coexistingcpx (Erlank ampKushiro1970) suggesting that high-Ca eclo-gites not only equilibrated at high temperature but also highpressure rather than representing a temperature excursionPyroxenites both opx-bearing and opx-free give somewhatlower average temperatures than eclogites (mean 9258C)and phlogopite-rich eclogites give the lowest temperatures(685 and 7698C respectively) The calcite-bearing eclogiteVR67112b gives a relatively low equilibration temperatureof 8558C

MEASURED ANDRECONSTRUCTEDWHOLE -ROCKSRationale and approachKimberlite-borne xenoliths sometimes show evidence forinfiltration of the host kimberlite (eg Taylor amp Neal1989 Barth et al 2001 2002b Heaman et al 2006) Wetherefore follow the standard approach of calculatingwhole-rock compositions from the compositions of the con-stituent phases weighted by their modal abundances Onlycore compositions were used as distinct rim compositionsmay point to late diffusion of elements from kimberlite-related melt veinsMany eclogitic whole-rock compositions reconstructed

with only garnet and cpx have negative Ti anomalies(not shown) For most of these samples rutile or ilmenitehas been identified as part of the mineral assemblage andthese phases should therefore be considered in the whole-rock reconstruction Rutile or ilmenite modes were esti-mated by assuming that the bulk-rock has no Ti anomalywhen normalized to a normal mid-ocean ridge basalt(N-MORB) composition and calculating the amount ofrutile required to erase theTi anomalyIt is of course possible that the eclogites had negative

or positive Ti anomalies relative to elements of similarcompatibility (Eu and Dy) Because we reconstructed

0

2

4

6

8

10

12volatile-rich

pyroxenite opx-bearing

pyroxenite opx-free

high-Ca

high-Mg

low-Mg

700750

800850

900950

10001050

11001150

12001250

Fre

quen

cy

Temperature

Fig 9 Histogram ofTKrogh (Krogh 1988) obtained at a preset pres-sure of 5 GPa except opx-bearing pyroxenites where TBKN wassolved simultaneously with PBKN (Brey amp Kolaquo hler 1990)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1857

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

whole-rock compositions from mineral core compositionskimberlite infiltration is unlikely to have caused fractiona-tion of Ti from Eu and Dy In a study in which rutilemodes were estimated from the differences between Ti inmeasured bulk-rocks and in bulk-rocks reconstructed withonly garnet and cpx only three of 19 eclogites recon-structed with rutile show Ti anomalies (Barth et al2002a) Even eclogite xenoliths with depleted Zr and Hfabundances that resemble those of modern island arcbasalts do not show Ti anomalies (Jacob amp Foley 1999)We therefore consider our approach to be a reasonablyrobust means of obtaining meaningful rutile modes wheresmall sample sizes preclude reliable determination of themodes of accessory phases by point-countingWhere rutilewas not recognized on thin-section scale or not analysedthe average eclogitic rutile composition was used for thereconstruction

UncertaintiesReconstructing whole-rock compositions in this way is arather crude method given the coarse grain size of theeclogites relative to the sample size and the non-uniformdistribution of the principal minerals on thin-sectionscale The uncertainty in mineral modes of medium- tocoarse-grained eclogites with grains45mm has been esti-mated by Jerde et al (1993) to be 10 plus uncertaintiesrelated to partial alteration which is minor in our samplesNevertheless the whole-rock trace-element budget israther insensitive to the exact modes for common eclogiticassemblages and REE patterns which encompass mostother elements in terms of compatibility in garnet andcpx do not change significantly even for modal variationsof 30 (Jerde et al 1993) This is not true for major-element concentrations which vary substantially with var-iations in mineral modes and we therefore do not calculatewhole-rock major-element contentsBecause the discussion that follows is based on trace-

element patterns and trends rather than absolute abun-dances whole-rock reconstruction based on mineralmodes is adequate We are confident that the uncertaintyinherent in this approach is not leading to significant mis-representation of eclogite whole-rock compositions becausethe trace-element patterns of the reconstructed whole-rocks show distinct differences (see below) that correlatewith the assignment to eclogite types based on garnetmajor-element composition and this is independent ofmineral mode Following Jerde et al (1993) in ElectronicAppendix 9 we show variations in trace-element patternfor one sample in each eclogite type for variations ingarnet and cpx abundance from 7030 to 3070 vol This shows that the trace-element patterns are generallyinsensitive to cpxgt ratios with regard to slopes andanomalies such as positive Sr Pb and Eu anomalies withthe exception of the positive Eu anomaly in high-Mgeclogite VR43479 which disappears if a high cpx mode is

assumed (not shown) Variations in rutile modes by 50(Electronic Appendix 9d) lead as would be expected tothe appearance of Ti anomalies but a rutile mode of 05instead of 1 vol does not erase the positive Nb anomalyin low-Mg eclogiteVR50909Because of the robustness of the relative trace-element

abundances in the reconstructed eclogites a garnet propor-tion of 065 and cpx proportion of 035 approximately themedian modal abundances of these minerals in volatile-free eclogites from Lac de Gras are assumed for eclogitesamples that are too small for determination of meaningfulmodes by point-counting

Within- and between-group variationsThe trace-element compositions of reconstructed whole-rocks are given in Table 4 N-MORB-normalized trace-element patterns of reconstructed whole-rocks areshown in Fig 10 The three eclogite groups have distincttrace element patterns that are particularly obvious forthe REE (Fig 11) Low-Mg eclogites have positive Pb andSr anomalies and smooth positive slopes in their MREEN

and HREEN LREEN are enriched relative to MREEN inthree samples (Fig 11a) Three samples have high Nbrelative to similarly compatible elements and comparedwith most other eclogites (Fig 10a) All high-Mg eclogitesdisplay more or less pronounced positive Eu and Sranomalies and most have positive Pb and negative Zrand P anomalies Their N-MORB-normalized REE pat-terns show shallow positive slopes from LaN to PrN orNdNand aweak negative slope from NdN to LuN (Fig11b)VR43479 is a low-Mg eclogite with regard to its garnet

Ca^Fe^Mg composition but has a trace-element patterntypical of high-Mg eclogites and will be discussed withthe latter Conversely sample YK1949 a high-Mg eclogitein terms of its garnet Ca^Fe^Mg compostion is more simi-lar to low-Mg eclogites in terms of its whole-rock trace-element pattern and is plotted in Figs 10a and 11aWhereas the patterns of most samples within the groupresemble each other the N-MORB-normalized incompati-ble-element abundances of the high-Ca eclogites vary bymore than an order of magnitude (Fig 10c) All sampleshave positive Pb anomalies and most have positive Sranomalies Some samples have negative P andor Zranomalies Niobium abundances in three of the samplesare very high similar to those for some low-Mg eclogitesHigh-Ca eclogites tend to have flatter MREEN andHREEN patterns and more uniform HREE abundancesthat the other eclogite types with humps between SmN

and HoNThe phlogopitized sample YK1911 shows a strong posi-

tive Pb anomaly and lower HREE abundances than thecalcite-bearing sample VR67112b which shows a distinctnegative Eu anomaly and a depletion of La relative to Ce(Figs 10d and 11d)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1858

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Trace-element compositions of reconstructed whole-rocks (ppm)

Sample P Sc Ti V Co Ni Ga Sr Y Zr Nb La Ce Pr Nd

Low-Mg eclogites

vr19674-g4 186 50 4649 317 57 59 144 77 24 30 36 13 31 05 24

vr43452 154 39 2253 328 49 84 176 52 32 11 01 02 07 01 08

vr43469 38 65 4377 349 54 91 142 34 28 15 02 05 15 03 17

yk1926 144 45 3406 258 52 87 145 89 18 18 20 09 29 06 33

yk1943 173 46 5732 295 41 39 146 122 25 48 82 34 85 13 64

yk19491 102 34 1693 158 31 49 100 46 10 14 03 09 21 03 17

High-Mg eclogites

vr19677-g2 138 39 5160 291 41 53 119 98 14 22 04 13 53 11 64

vr434792 46 25 1265 129 36 126 84 56 3 7 04 06 25 04 23

vr67360 49 36 1573 138 64 184 111 76 5 9 04 09 38 07 33

yk1946 153 28 4877 192 45 105 124 108 12 48 03 10 38 08 50

High-Ca eclogites

vr19674-g5 267 30 7663 200 41 67 195 32 23 43 77 01 04 02 16

vr40345 0 33 2821 175 47 98 210 179 13 13 66 60 109 11 38

vr43465 142 27 5104 220 31 83 121 197 14 35 29 05 22 05 40

vr50858 295 30 6898 201 44 94 150 48 22 38 06 01 03 02 20

vr50860 68 57 5672 418 57 71 186 109 19 11 02 02 13 03 26

vr50909 238 28 7295 201 42 92 171 48 20 57 13 01 04 02 21

yk3528 91 51 3712 341 53 155 177 49 16 9 01 02 08 02 11

Volatile-rich eclogites

vr67112b 40 84 9242 521 46 30 107 4 40 66 31 26 126 24 125

yk1911 31 39 9428 286 44 82 164 3 23 52 64 18 52 09 46

Pyroxenites

yk1914 111 49 707 207 48 547 44 31 4 3 02 06 19 03 10

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

Low-Mg eclogites

vr19674-g4 11 05 10 22 39 10 29 33 06 12 05 028 24 008

vr43452 04 02 08 13 38 12 41 49 08 05 01 nd 02 na

vr43469 11 05 09 25 44 11 34 36 06 07 02 001 09 129

yk1926 12 05 11 19 30 07 22 24 04 07 03 004 23 090

yk1943 21 08 10 29 42 10 30 32 05 16 05 013 65 017

yk19491 06 03 11 08 14 04 14 18 03 04 02 003 29 115

High-Mg eclogites

vr19677-g2 20 08 12 21 24 05 15 15 03 05 05 007 52 6811

vr434792 05 02 12 05 06 01 04 03 00 03 01 001 98 037

vr67360 06 03 14 06 09 02 06 06 01 02 02 002 86 000

yk1946 18 07 12 21 22 05 13 12 02 11 03 006 56 112

High-Ca eclogites

vr19674-g5 17 09 11 33 43 09 24 23 04 16 05 007 02 011

vr40345 09 04 10 12 22 06 16 14 02 05 nd nd 298 na

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1859

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Continued

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

vr43465 22 09 12 25 26 06 15 14 02 10 04 001 25 040

vr50858 20 08 09 33 41 08 22 21 03 10 01 nd 02 na

vr50860 14 07 11 27 37 08 22 18 03 06 00 nd 06 na

vr50909 20 10 12 33 38 07 20 20 03 16 01 002 02 181

yk3528 08 05 11 20 30 07 18 18 02 05 01 nd 08 na

Volatile-rich eclogites

vr67112b 38 10 07 53 74 15 41 34 04 18 03 012 60 004

yk1911 20 07 08 35 39 07 18 17 02 10 14 nd 73 na

Pyroxenites

yk1914 02 01 11 04 08 02 05 06 01 01 03 003 68 101

aGarnet has CandashMgndashFe relationship typical of high-Mg eclogites but whole-rock trace-element pattern is typical oflow-Mg eclogitesbGarnet has CandashMgndashFe relationship typical of low-Mg eclogites but whole-rock trace-element pattern is typical ofhigh-Mg eclogites

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

(a)

001

01

1

10(c)

001

01

10(b)

VR43452

YK1949

RockNMORB

low-Mg

RockNMORB

high-Mg

(d)

volatile-richhigh-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

phlogopite-bearingYK1911

calcite-bearingVR67112b

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 10 Extended trace-element patterns of reconstructed (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-bearing eclogites Normalized to N-MORB of Sun amp McDonough (1989)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1860

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

RockNMORB

(b)

RockNMORB

(d)

010

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu001

01

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

(a)

001

01

1

10(b)

001

01

1

10(c)

001

01

1

10(e)

high-TSchmidberger et al

low-TSchmidberger et al

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

high-Mglow-Mg

high-Cavolatile-rich

phlogopite-bearingYK1911

calcite-bearingVR67112b

YK1949VR43479

Fig 11 N-MORB normalized REE patterns for (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-richeclogites Shown for comparison in (e) are reconstructed eclogites from Schmidberger et al (2007)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1861

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Both volatile-rich whole rocks display strong negativeSr anomalies Strontium can be sited in calcite or apatiteandor indicate the presence of a Sr-rich mineral that wasnot identified in thin section such as bariteFour of the six reconstructed whole-rocks in the study

of Schmidberger et al (2007) (Fig 11e) show markedLREEHREEN enrichment that is observed only in onehigh-Ca eclogite in our study (Fig 11c) suggesting thatmany of the eclogites they studied were affected by metaso-matic LREE enrichment Two samples in their study thatare not affected by this enrichment show positive Euanomalies similar to the high-Mg eclogites in this studywhereas the distinctive REE pattern of the low-Mg eclo-gites is not recognized among their samples

SR^ND^HF I SOTOPESYSTEMATICSStrontium Nd and Hf isotope data for optically puremineral separates and reconstructed whole-rocks aregiven in Table 5 Spongy rims and alteration are likely tobe mostly removed during leaching in 6N HCl in combina-tion with repeated ultrasonication Despite this rigoroustreatment the reproducibility of Nd in garnet VR43469and in cpx YK1914 is worse than for other samples Thismay be due to inhomogeneities related to late Nd diffusionpossibly from grain boundaries containing a kimberlitecomponent The reproducibility of the 176Hf177Hf data isgenerally poorer than that of Nd which may reflect a com-bination of matrix effects caused by imperfectly purifiedsolutions and sample heterogeneity Parent^daughterratios are available only from in situ analyses which arenot precise enough to allow meaningful model age orinitial isotope ratio calculations (ie at the time of kimber-lite eruption) Because the Lac de Gras kimberlites arerelatively young (55^56 Ma Graham et al 1999) age cor-rection would result in only a minor change in isotopecomposition which is inconsequential in view of the rangeof compositions displayed by most samples and hence hasno effect on the conclusions drawnClinopyroxenes in five volatile-free eclogites and a

pyroxenite have Sr isotope ratios ranging from 0701662 to0703546 this range is similar to or somewhat lower thanthat of the depleted mantle at the time of kimberliteemplacement (0703 Bell amp Tilton 2002) Clinopyroxenein the phlogopitized sample has a distinctly high 87Sr86Srof 0751186Clinopyroxene controls 56^92 of the Nd budget (Nd

concentration in cpx weighted by mode) Neodymiumisotope ratios range from 0511660 to 0514844 Thiscorresponds to eNd variation (per 10 000 deviationof 143Nd144Nd in the sample from 143Nd144Nd in the chon-dritic mantle model reservoir) from 17 to thorn43 cpx inthe low-Mg eclogites have higher eNd than those in

high-Mg eclogites Garnets have generally have higher143Nd144Nd from 0512984 (eNdfrac14thorn 7) to 0520965(eNdfrac14thorn 162) Garnets in the two volatile-rich sampleshave the highest ratios 0522790 for garnet in the calcite-bearing sample VR67112b (eNdfrac14thorn198) and 0525721 inthe phlogopitized eclogiteYK1911 (eNdfrac14thorn 255)Garnet contributes 18^32 and cpx 48^78 of the Hf

budget The proportion controlled by rutile is 4^30(Aulbach et al in preparation) Measured 176Hf177Hf incpx ranges from 0281263 in phlogopite-bearing sampleYK1911 (eHffrac1453) to 0286130 in low-Mg eclogiteVR43469 (eHffrac14thorn119) and in garnet from 0283958 inpyroxeniteYK1915 (eHffrac14thorn42) to 0295718 in low-Mg eclo-gite YK1926 (eHffrac14thorn458) Hafnium in garnet in the phlo-gopite^apatite-bearing sample is extremely radiogenic(176Hf177Hffrac14 0354367) corresponding to an eHf of thorn2532which is close to a value determined for a garnet separatefrom a Roberts Victor eclogite (eHffrac14thorn2561 Jacob et al2005) Garnet in the calcite-bearing sample has extremelyradiogenic Nd and Hf (eNdfrac14thorn193 eHffrac14thorn2516) Garnetin pyroxenite YK1915 also has radiogenic Nd and Hf(eNdfrac14thorn27 eHffrac14thorn42)Despite the large uncertainties for reconstructed whole-

rocks (calculated as the square root of the sum of squareduncertainties of two standard errors for garnet and cpxanalyses combined with 20 propagated uncertainty onthe proportion of Nd or Hf respectively contributed byeach mineral) the isotopic data are consistent in thatthey correlate with the independently determinedparent^daughter ratios (1) Low-Mg eclogites have higherSmNd reflected in the steeper positive slope of theirN-MORB-normalized REE patterns (Fig 11a) comparedwith the high-Mg eclogites (Fig 11b) Accordingly recon-structed low-Mg eclogites have higher present-day eNd

(0 to thorn47) than high-Mg eclogites (4 and 17) (2) Low-Mg eclogite YK1943 has very high Nb and LREE abun-dances that could have been added during metasomatismthat also enhanced its Hf abundance This sample has thelowest eNd and eHf of the low-Mg eclogites (0 and thorn20respectively) In contrast apparently unenriched low-Mgeclogites have more radiogenic Nd and Hf (eNdfrac14thorn47and thorn9 eHffrac14thorn128 and thorn113 respectively)The Sr^Nd isotope diagram reveals that eNd in cpx

varies from 19 to 43 at nearly constant depleted-mantle-like 87Sr86Sr (Fig 12a) In the Nd^Hf isotope diagram(Fig 12b) the reconstructed whole-rocks lie on curvedmixing lines if the lsquomixingrsquo proportions of Nd and Hffor the two lsquoend-memberrsquo minerals are very different(eg YK1911) Garnets have more radiogenic Nd andHf than coexisting cpx All reconstructed volatile-freeeclogites plot in the radiogenic field with regard to Hfor are isotopically similar to the depleted mantle Twohigh-Mg eclogites and one low-Mg eclogite have unradio-genic Nd one high-Mg eclogite has radiogenic Nd and

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1862

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 5 Individual and averaged Sr Nd and Hf isotope ratios obtained from garnet and cpx solutions

Sample 87Sr86Sr 2SE2SD 143Nd144Nd 2SE2SD eNd 2SE2SD 176Hf177Hf 2SE2SD eHf 2SE2SD

Low-Mg eclogites

VR43469 cp 0702661 0000010 0514844 0000006 43 01 0286130 0000012 119 04

VR43469 gt 0515508 0000014 56 03 0286506 0000012 132 04

VR43469 gt r 0515453 0000010 55 02 0286548 0000011 134 04

VR43469 gt r 0515384 0000007 54 01 0287244 0000010 158 04

VR43469 gt r 0515487 0000014 56 03 0286996 0000014 149 05

VR43469avg gt 0515458 0000109 55 21 0286823 0000715 143 253

WR VR43469 0515059 0000273 47 5 0286405 0001079 128 38

YK1926 cp 0703546 0000011 0513043 0000007 8 0 0283154 0000009 13 03

YK1926 gt 0513364 0000008 14 01 0295718 0000005 458 02

WR YK1926 0513120 0000026 9 1 0285980 0000716 113 25

YK1943 cp 0512143 0000007 10 01 0282213 0000005 20 02

YK1943 gt 0514461 0000012 36 02 0287510 0000008 168 03

YK1943 gt r 0514491 0000009 36 02 0287332 0000005 161 02

YK1943 gt r 0514517 0000010 37 02 0287290 0000007 160 02

YK1943avg gt 0514489 0000056 36 11 0287377 0000233 163 82

WR YK1943 0512625 0000228 0 5 0283138 0000879 13 31

High-Mg eclogites

VR43480 gt 0520965 0000017 162 03 0284950 0000010 77 04

VR67360 cp 0283509 0000074 26 26

VR67360 cp r 0283433 0000031 23 11

VR67360avg cp 0701662 0000018 0511660 0000010 19 02 0283471 0000108 25 38

VR67360 gt 0512984 0000101 7 20 0287429 0000180 165 64

WR VR67360 0511761 0000132 17 3 0284734 0000469 69 17

YK1949 cp 0703019 0000013 0512231 0000009 8 02 0282513 0000009 9 03

YK1949 gt 0513511 0000010 17 02 0290297 0000042 266 15

YK1949 gt r 0513508 0000012 17 02 0289947 0000015 254 05

YK1949 gt r 0513552 0000014 18 03

YK1949avg gt 0513523 0000049 17 10 0291569 0000494 311 175

WR YK1949 0512446 0000103 4 2 0284622 0000847 65 30

Volatile-rich eclogites

VR67112b gt 0522790 0000008 198 02 0354367 0000036 2532 13

YK1911 cp 0751186 0000009 0512165 0000007 9 01 0281263 0000009 53 03

YK1911 gt 0525721 0000013 255 03 0288051 0000012 187 04

WR YK1911 0513940 0000342 25 7 0281414 0000018 48 1

Pyroxenites

YK1914 cp 0703015 0000034 0513283 0000041 13 08

YK1914 cp r 0702956 0000005 0513200 0000027 11 05

YK1914 cp r 0513527 0000059 17 11

YK1914 cp r 0513045 0000010 8 02

YK1914avg cp 0702986 0000083 0513264 0000402 12 78

YK1915 gt 0514003 0000011 27 02 0283958 0000021 42

Analyses of full procedural replicates are designated with an lsquorrsquo after sample name SE standard error (internal errorsingle analyses) SD standard deviation (external error averaged multiple analyses) Italics used for reconstructed whole-rocks (mineral isotopic compositions and concentrations weighted by modes) Uncertainties for reconstructed whole-rocksinclude 2SE or 2SD uncertainties on each mineralrsquos isotope composition combined with 20 uncertainty in the amountof Nd or Hf contributed by the constituent minerals eNd and eHf are per 10 000 deviation from model chondritic mantle[model of Wasserburg et al (1981) and Blichert-Toft amp Albarede (1997) respectively]

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1863

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

εHf

εNd

Ekati eclogitesminus100

0

100

200

300

400

500

minus40 0 40 80 120 160 200 240 280

gt in YK1915

gt in VR67112bεNd=198 εHf=2532

minus40

0

40

80

120

160

minus40 minus20 0 20 40 60

minus60

minus40

minus20

0

20

40

070 071 072 073 074 075 076

North Americankimberlite field

εNd

87Sr86Sr

cpx in YK1914

Ekatieclogites

DM

DM

DM

DM

(a)

(b)

pyroxenites

low-Mg

high-Mg

volatile-rich

eclogites

Diavik low-TA154 S kimberlite

cpx in YK1911

Fig 12 (a) Sr^Nd for cpx and (b) Nd-Hf isotope diagram for garnet cpx and reconstructed whole-rocks (symbols with bold outline) Finedashed lines in (b) connect data points from the same sample with garnets having more radiogenic isotopic compositions Uncertainties forreconstructed whole-rocks combine 20 modal uncertainty with uncertainties on each mineralrsquos isotope composition Error bars on individualminerals may be smaller than the symbols Also shown are the isotopic compositions of the depleted mantle (DM DePaolo 1981 Griffin et al2000 Bell amp Tilton 2002) Range of 87Sr86Sr in eclogites from Ekati (Lac de Gras area 100 km north of Diavik Jacob et al 2003) in NorthAmerican kimberlites and compositions of low-temperature eclogites from Diavik and of Lac de Gras A154 South kimberlite fromSchmidberger et al (2007)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1864

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

one low-Mg eclogite has eNd similar to the depleted mantle(Fig 12b)Clinopyroxene in the phlogopitized eclogite YK1911

has unradiogenic Nd and Hf and strongly radiogenicSr whereas the coexisting garnet has radiogenic Nd andHf Although this sample contains 50 vol phlogopiteits whole-rock isotopic composition was calculated

from garnet and cpx alone because the contributionfrom phlogopite to the Hf and Nd budget is 52 and51 respectively (unpublished data) The whole-rockhas radiogenic Nd but has lower eHf than anyother sample analysed in accord with the extreme Hfabundance (53 ppm) and low LuHf of the cpx

DISCUSS IONOrigins of eclogites beneath the SlaveCratonHigh-Mg and high-Ca eclogites

With regard to their major- and trace-element contentsthe high-Mg and high-Ca eclogites from Lac de Gras arebroadly similar to variably evolved oceanic gabbros(Fig 13b and c) which may indicate an origin as subductedoceanic crust Compatible trace-element abundancesare useful to deduce potential crustal protoliths as theseelements are less susceptible to post-formation modification(Jacob 2004) The flat HREE patterns in the high-Mg andhigh-Ca eclogites are compatible with a low-pressureorigin as the oceanic crust is a product of moderate degreesof partial melting mainly within the spinel stability field(eg Presnall et al 2002) whereas melting with residualgarnet would qualitatively be expected to produce stronglyfractionated REE patterns Positive Sr and Pb anomaliesalthough compatible with a plagioclase signature couldalso be due to cryptic metasomatism (Dawson 1984) ofthe eclogites during their residence in the lithosphericmantle These anomalies are probably not related to kim-berlite infiltration as the whole-rock eclogite compositionswere reconstructed from analyses of fresh mineral coresLow ZrSm similar to that in primitive gabbros is

observed for many of the high-Mg and high-Caeclogites (Fig 13b and c) and is independent of the ratioof garnetcpx and whether or not the whole-rock wasreconstructed with rutile (Electronic Appendix 10) Thenegative trend of Mg-number against Yb in garnet(Fig 14) argues against igneous processes involving garnetaccumulation (Taylor amp Neal 1989) as suggested forexample for some high-Mg eclogites from West Africa(Barth et al 2002b) Rather the anti-correlation suggeststhat olivine fractionation played a role in the formation ofthe protoliths Fractionating olivine would excludeYb (Dolivinebasalt

frac14 002 Nikogosian amp Sobolev 1997)and incorporate Mg relative to Fe (Dfrac1415 and 05 respec-tively Taura et al1998) leading to progressive depletion ofMg and enrichment of Yb in the residual melt and henceanti-correlated Mg and Yb in the whole-rock This isinherited after eclogitization All eclogite types (exceptVR43452) plot along the olivine fractionation trend sug-gesting that they formed by a similar processHigh-Mg eclogites have lower

PREE than high-Ca

eclogites and some appear to have positive Eu anomalies

high-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

gabbrosSE Indian ridge

gabbrosOman ophiolite

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

high-Mg

RockNMORB(a)

VR43452

YK1949

low-Mg

boninite-like melts

01

001

1

10

(b)

01

001

1

10

(c)

01

001

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

Fig 13 Trace-element patterns of (a) low-Mg (b) high-Mg and(c) high-Ca eclogites compared with boninite-like volcanic rocksfrom the Isua greenstone belt (Polat et al 2002) gabbros from theOman ophiolite (Benoit et al 1996) and the SE Indian Ridge (Hartet al 1999) normalized to N-MORB of Sun amp McDonough (1989)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1865

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(EuEu up to 14) similar to more primitive cumulates(troctolite gabbros and olivine gabbros) reported in the lit-erature (Fig 13b) Although the sizes of these Eu anomaliesare not outside the analytical uncertainty their correlationwith CeYb in reconstructed eclogites supports their signif-icance (r2frac14066 nfrac14 4) The more primitive nature ofhigh-Mg eclogites is consistent with higher Mg-numbersrelative to high-Ca eclogites and may be due to olivineaccumulation during formation of their protolithsThe two samples with the lowest HREE abundanceshave the highest EuEu which is compatible with varia-tions in plagioclase modesMore evolved protoliths with higher pyroxeneplagio-

clase plus trapped melt could account for small or absentpositive Eu anomalies and higher

PHREE of high-Ca

eclogites (Fig 13c) Although the reason for the frequentassociation of graphite or diamond with high-Ca eclogitesis unclear (Aulbach et al 2003) the frequent occurrence ofkyanite in high-Ca eclogites also observed at Ekati justnorth of Lac de Gras has been argued to support a crustalprotolith (Jacob 2004) Kyanite andor corundum havebeen suggested to exsolve from high-temperature high-Alclinopyroxenes (Caporuscio amp Smyth 1990) Eclogitesfrom Lac de Gras show no exsolution textures suggestiveof such an origin although recrystallization can obscureor destroy microstructural evidence for exsolution(eg Griffin et al 1984)Modelling shows that qualitatively plagioclase accumu-

lation from a low-pressure melt derived from MORBmantle can account for positive Pb Sr and Eu and negativeZr anomalies relative to N-MORB whereas cpx accumula-tion or trapped melt increases Zr Y MREE and HREE

abundances (Fig 15) Some melt addition (for exampleby entrapment) seems to be required to increase all trace-element abundances to those observed for the eclogitesAlthough high-Ca and high-Mg eclogites could representshallow and deep portions of the same oceanic crust thehigher average temperatures obtained for the high-Caeclogites the frequent occurrence of carbon in high-Caeclogites and its absence in high-Mg eclogites the strongLREEHREE fractionation observed for six of seven high-Ca eclogites and the narrow range of compositions of thehigh-Mg eclogites suggest that they may have protolithsthat are not cogenetic Some discrepancies in particularaffecting the incompatible elements may be due to post-crystallization processes acting upon the protoliths suchas dehydration melting and later metasomatism that isevident in peridotites from the same locality (Griffin et al1999)Low K2OTh for high-Ca and high-Mg eclogites

(average 08 Table 4) relative to assumed crustal

0

2

4

6

8

10

30 40 50 60 70 80

Mg

VR43452

Yb

GARNET

pyroxenites

low-Mghigh-Mghigh-Ca

eclogites

VR19673ecl-g2

Fig 14 Garnet Mg-number vs Yb (ppm including 2s uncertainty)Lines qualitatively show separate correlations for eclogites (exceptoutlierVR43452) and pyroxenites (except outlierVR19673ecl-g2)

high-Ca

high-Mg

001

01

1

10

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

melt08 melt02 cryst cpx01 melt09 cryst

plag

(a)

001

01

1

10(b)

RockNMORB

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 15 Reconstructed (a) high-Ca and (b) high-Mg eclogitescompared with compositions of plagioclase and cpx crystallizingfrom a MORB melt modelled using distribution coefficients ofNorman et al (2005)The MORB was calculated as a 10 equilibriummelt of MORB pyrolite [composition from Kelemen et al (1993)except Pb Pr Sr Y and Lu which were calculated as residues from1 fractional melting of pyrolite from McDonough amp Sun (1995)using melt modes of Kinzler (1997) and distribution coefficients ofNorman et al (2005)] Also shown are mixtures of crystals (consistingof 50 plagioclase plus 50 cpx) and melt in the proportions 0901and 0208

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1866

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

precursors (eg gabbros SE Indian Ridge average 22Hart et al 1999) may reflect the preferred loss of fluid-mobile elements during dehydration (Becker et al 2000)although large uncertainties on these low-abundance ele-ments in the constituent minerals and reconstructedwhole-rocks makes this interpretation less than robustThe REE patterns of most high-Mg eclogites are too flatto represent melting residues as partial melting shouldproduce low LREEHREE (Electronic Appendix 10) Incontrast the marked LREE depletion of five of sevenhigh-Ca eclogites (LaLufrac14 02^08) contrasts with that ofassumed evolved gabbroic protoliths (LaLufrac1415^32Hart et al 1999) and can be modelled as a consequence ofpartial melting during subduction after eclogitization (iein the presence of garnet which retains HREE relative toLREE eg Barth et al 2002a) has taken place (Aulbachet al 2003)Our results are compatible with previous studies on

eclogite xenoliths from the Slave Craton includingthose from the same kimberlite that have linked thesesamples or some of these samples to Paleoproterozoicsubduction events (the 195^191Ga formation of theHottah terrane and the 18Ga formation of the GreatBear magmatic arc) in the Slave craton (Griffin et al1999 Kopylova et al 1999 Heaman et al 2002 2006Schmidberger et al 2005 2007)The compositional features discussed above although

compatible with crustal protoliths are not necessarilydefinitive of a subduction origin However if direct crystal-lization from basaltic melts at high pressure (ie mantledepths) is invoked instead specific conditions would haveto be met to satisfy the observations The Mg-number ineclogites from Lac de Gras are lower than those expectedof primary high-pressure mantle melts (eg Walter 1998OrsquoHara amp Herzberg 2002) requiring crystallization athigh pressures after olivine fractionation This originwould further require shielding from peridotitic wallrockwith which the fractionated melts would no longer be inequilibrium (Barth et al 2001) In such a setting the lowrelative abundances of fluid-mobile elements comparedwith those of less fluid-mobile elements and of LREErelative to HREE (Table 4) would reflect an expulsion offluids during crystallization of the protoliths and latermelt mobilization respectively The eclogites clearly donot represent garnet plus cpx cumulates which wouldrequire a positive correlation of Mg-number with Yb thatis at variance with the observed negative correlation ineclogitic garnets from this study (Fig 14)

Low-Mg eclogites

Garnets in low-Mg eclogites compositionally overlap withthose in massive eclogites from Jericho (NW of the Lacde Gras kimberlites) that were suggested to have a high-pressure intrusive origin based on the similarity of theirmajor-element compositions to those of Group A eclogites

fromsouthernAfrica forwhichamantleoriginhasbeenpro-posed (Kopylova et al1999) Garnets in low-Mg eclogites inthe present study have major-element compositions moresimilar to garnets in groupB rather thangroupAeclogitesThe positive slope in N-MORB-normalized HREE

abundances displayed by the low-Mg eclogites is clearlydistinct from other eclogite types and is similar to those ofboninite-like second-stage melts that are closely associatedwith ultramafic volcanics (Smithies et al 2004) (Fig 13a)The range of Mg-numbers in the low-Mg eclogites (55^66)overlaps with that of Archaean second-stage melts fromIsua (58^75) and Abitibi (52^82 Smithies et al 2004)These similarities might indicate that the low-Mg eclogiteshave an origin in a previously depleted mantle source suchas supra-subduction zone mantle Subchondritic ZrHf ofthe low-Mg eclogites contrasts with the supra-chondriticZrHf of the high-Mg and high-Ca eclogites (Aulbach et

al in preparation) and is consistent with previous cpx-controlled depletion in an arc mantle source This sourcecould have been subsequently remelted to generate theboninite-like precursor to the low-Mg eclogites as sug-gested for the protoliths of eclogites from Udachnaya(Jacob amp Foley 1999)Although some second-stage melts to which the low-Mg

eclogites are similar are closely associated with komatiitesand could contain a plume component the radiogenic ordepleted mantle-like Nd and Hf and unradiogenic Sr inthe low-Mg eclogites do not support plume involvement

Trace-element and Sr^Nd^Hf isotopeconstraints on post-formationmetasomatismSeveral low-Mg and high-Ca eclogites show negativeslopes in the LREE which may be indicative of interactionwith small amounts of LREE-enriched melt (Navon ampStolper 1987) amounting to cryptic metasomatism(ie not accompanied by crystallization of new phases) inthe nomenclature of Dawson (1984) High Nb abundancesrelative to elements of similar compatibility displayed bysome low-Mg and high-Ca eclogites (Fig 10) could resultfrom overestimation of rutile modes However high-CaeclogiteVR40345 is rutile-free and has been reconstructedwithout rutile its high Nb abundance is accompanied byLREE enrichment indicating that metasomatic additionmay be responsibleLow-Mg eclogites show a positive correlation of La with

Nb (Electronic Appendix 11a) indicating concomitantaddition of LREE and Nb This suggests that the metaso-matic agent was a solute-rich fluid or silicate silicic melt ashydrous fluids and carbonatites do not carry enough HFSEto affect the budget of eclogites (Green amp Wallace 1988Klemme et al 1995 Adam et al 1997 Stalder et al 1998)Of the samples that are enriched in Nb all but one (high-Ca eclogiteVR50909) show a simultaneous enrichment in

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1867

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Hf (Electronic Appendix 11b) Together these trace-element relationships suggest that both SmNd and LuHfwere lowered in these samples Their low parent^daughterratios contrast with higher ratios in unenriched samplesand resulted in the ingrowth of a range of Nd and Hf iso-tope compositions Compared with the low-Mg and high-Ca eclogites the high-Mg eclogites have a narrow range ofLa and Nb concentrations which correlates with a narrowrange in Nd and Hf isotope compositions (Fig 12)Precise model ages cannot be calculated because parent^

daughter ratios are available only from in situ analyseswhich have relatively large uncertainties Howeverevidence for long-term enrichment and depletion may bediscerned Low-Mg eclogites even those that are enrichedin LREE have relatively high SmNd as a result of havingoverall positive slopes in their REE patterns relative toMORB whereas the high-Mg eclogites have relativelyflat REE patterns and therefore lower SmNd (Fig 11)This translates into lower eNd for high-Mg eclogites thanfor low-Mg eclogites In contrast they show no systematicdifference with respect to 176Hf177Hf Because the low- andhigh-Mg eclogites are proposed to have different origins asdiscussed above and therefore did not necessarily have thesame initial isotopic composition and formation age theseisotopic differences cannot be used to estimate their agesA long-term coupling of trace-element and isotopic

abundances is also indicated by the fact that low-Mg eclo-gite YK1943 which has experienced Nb and LREE addi-tion has lower eNd (0) and eHf (thorn13) than low-Mg eclogiteVR43469 with higher LuHf and SmNd (eNdfrac14thorn47eHffrac14thorn128) If low-Mg eclogites YK1943 and VR43469originally had identical initial Nd and Hf isotope composi-tions and if the parent^daughter ratios are accurate andwere modified only once during their evolution shortlyafter metamorphism (when their isotopic compositionwould have been reset) this overprint can be calculatedfor both isotopic systems to have occurred at 17GaThis would suggest that Hf and Nd were added duringthe same event In contrast high-Mg eclogites VR67360and YK1949 would have had identical Ndisotope compositions at 10Ga whereas their Hf isotopecompositions are disturbed in that the sample with thehigher parent^daughter ratio has a lower 176Hf177Hf thanthe sample with the lower parent^daughter ratio Theseresults suggest that the LREE and HFSE enrichmentobserved in different samples occurred during separateevents that did not affect the eclogites equallyExtreme HFSE enrichment leading to the growth of

zircon and rutile and LREE addition leading to thegrowth of apatite in eclogites from the Jericho kimberlitein the northern Slave craton have been ascribed to twoseparate metasomatic events at around 18Ga (quasi-coincident with the age of formation and metamorphism)and 13^10Ga respectively (Heaman et al 2002 2006

Schmidberger et al 2005) The authors associated theseevents with Paleoproterozoic subduction and aMesoproterozoic thermal disturbance in the Slave litho-spheric mantle possibly linked to the Mackenzie igneousevent Mesoproterozoic Nd model ages mostly between13 and 14Ga were reported for eclogites from thenearby Ekati kimberlites (Jacob et al 2003) Although wedo not observe zircon or apatite in volatile-free eclogitesfrom Lac de Gras and the age constraints obtained fromour data are weak they support a link between theseevents described for other eclogite populations and themetasomatic enrichment observed in our studyThe Sr isotope data in the present study are similar to

those of Schmidberger et al (2007) who suggested that thedepleted-mantle-like values are consistent with derivationof the protoliths from an oceanic mantle source However87Sr86Sr in volatile-free eclogites in the present study isremarkably constant considering their Nd and Hf isotopevariability (Fig 12) and the variability of trace-elementpatterns (Fig 10) Given the evidence for dehydration andmelting in many of the eclogites fractionation of highlyincompatible Rb from Sr and differential ingrowth ofradiogenic Sr would be expectedThis may indicate a rela-tively recent overprint and considering the similarityof 87Sr86Sr to present-day depleted mantle which is signif-icantly lower than 87Sr86Sr of North American kimber-lites a juvenile (asthenospheric) source

PyroxenitesPyroxenites from Lac de Gras often show relationshipsin their major-element contents that are different fromthose observed in the eclogites For example pyroxeniticgarnets show a negative correlation between CaO andNa2O contents whereas these oxides are positively corre-lated for garnets in eclogites (Fig 5) CaO is well correlatedwith C2O3 in pyroxenitic garnets but not in eclogitic gar-nets (Fig 4) Mg-number is positively correlated withYb abundance for pyroxenitic garnets but anti-correlatedfor eclogitic garnets (Fig 14) This suggests that the pyrox-enites in the present study are not related to the eclogitesby fractionation or accumulation processes as proposedfor the origin of some pyroxenites (Foley et al 2003)Pyroxenites have also been interpreted as the products

of crystal segregation in magma channels with alteredoceanic crust as the magma source this would explaintheir Eu anomalies and stable and radiogenic isotopecompositions (Pearson et al 1993 Becker 1996) As thehigh-pressure liquidus phases of most terrestrial basaltsare cpx and garnet (OrsquoHara amp Herzberg 2002) partialmelts of eclogites with basaltic composition wouldcrystallize cpx and garnet but not opx More than half ofthe pyroxenites from Lac de Gras contain opx suggestingthat they probably did not crystallize directly fromeclogite-derived melts

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1868

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Low-Cr websterites and pyroxenites could be crystalli-zation products of mafic melts that interacted with perido-titic mantle thus becoming more refractory The positivecorrelation of Yb and Mg-number in garnet would allowan origin by fractional crystallization from such meltsAlternatively they formed by reaction of silicic (possiblyeclogite-derived) melts with mantle peridotites(eg Kelemen et al 1993 Aulbach et al 2002) This agreeswith the similarity of hypothetical melts in equilibriumwith pyroxenitic garnets in most samples to eclogite-derived melts (Electronic Appendix 12) Niobium concen-trations in all hypothetical melts are much higher than ineclogite-derived melts regardless of whether or not rutileis assumed to be in the residue The same is true for LaCe and Zr concentrations in some samples which mayindicate that NbZr and LREE were added to thepyroxenites some time after their formation as discussedin the previous section for some eclogites Because mini-mum ages for pyroxenitic rutile with an unradiogenicHf isotope composition (calculated by projecting the176Hf177Hf to the model depleted mantle isotopic evolutioncurve assuming zero ingrowth) range to 178066Ga(Aulbach et al in preparation) it appears likely that thepyroxenites were not formed recently (eg during kimber-lite-related or precursory melt migration) and that some ofthe samples are at least MesoproterozoicGarnet pyroxeniteYK1952 consists of schlieren of equili-

brated garnet pyroxenite in a lsquomatrixrsquo of spongy cpxfine cpx^opx intergrowths embayed garnet and meltpatches (Fig 2f) Garnet in the pyroxenite schlieren hasvery low LREE and HFSE abundances Orthopyroxenein the matrix has distinctly higher Mg-number andCr-number and contains more CaO than opx from thepyroxenite schlieren whereas opx rims in the schlierenare compositionally similar to opx in the matrix The con-siderable compositional and textural disequilibrium inthis sample suggests that this is a young feature causedby reaction with fluids shortly before or at the time ofentrainmentAs a consequence of the small sample sizes of most of

the pyroxenites studied here a comprehensive isotopicdataset is not available Clearly more such data areneeded in addition to stable isotope data to obtain betterconstraints on the origins of this rock type Consideringthe microstructural and compositional variability ofthe pyroxenites and major-element relationships oppositeto those observed for eclogites it appears certain that pyr-oxenites form by a variety of processes some of which areunrelated to the formation of eclogites and their protoliths

Volatile-rich eclogitesVolatile-rich eclogites from Lac de Gras display the mostextreme Hf^Nd isotope systematics of all the analysedsamples Garnets in phlogopitized (YK1911) and in cal-cite-bearing eclogite (VR67112b) have the most radiogenic

Nd (eNdfrac14 2486 and 1927 respectively) Garnet inVR67112b has extremely high eHf (2516) Clinopyroxene inYK1911 has the lowest eHf measured for any mineralseparate from Lac de Gras (53) in accord with the veryhigh Hf abundance in its cpx (5 ppm) In contrastgarnet in this sample has radiogenic Hf (eHffrac14 136) Thisdiscrepancy is at least in part due to their relatively lowequilibration temperatures (7698C and 8558C respec-tively) which allowed the isotopic compositions of cpxand garnet to evolve separately at their low andhigh parent^daughter element ratios respectively forextended periods of time and possibly since formationAccordingly the cpx in YK1911 gives a minimum Hf ageof 27GaSlab-derived fluids could have been introduced during

the 27Ga collision of the western Slave Craton domainincluding the thick plume-derived deep lithosphericmantle layer and the eastern domain with its highlydepleted lithosphere This would also be consistent withthe observation that volatile-rich eclogites occur at shallowdepths in the mantle (TKroghfrac14 685^8558C correspondingto depths of 85^115 km along a 40 mWm2 geotherm)where the bulk of dehydration and production of solute-rich melts in a subducting slab might be expected to occur(Becker et al 2000 Hermann et al 2006) Their negativeEu anomalies might indicate that the fluid was derivedfrom the shallow portion of the slab that experiencedplagioclase fractionation The presence of carbonate inVR67112b requires a carbonated sourceHigh 87Sr86Sr in cpx in YK1911 is probably the conse-

quence of isotopic equilibration with high-RbSr phlogo-pite after long-term radiogenic ingrowth while Hfaddition led to retarded ingrowth of radiogenic HfThese systematics combined with radiogenic Nd appear

inconsistent with the involvement of sediments in the pet-rogenesis of this sample (eg Perfit et al 1980 Vervoortet al 1999) or a hydrous fluid such as those responsible forMARID metasomatism in the Kaapvaal craton (Konzettet al 1998)Whatever the origin of the fluids the apparentHf mobility suggests that they were not in equilibriumwith a rutile-bearing source and also that they weresolute-rich in nature (ie hydrous melts able to dissolveHFSE) and the relatively high HREE abundances of thevolatile-rich eclogites indicate a garnet-free source suchas amphibolite This allowed evolution of the sample atrelatively high time-integrated SmNd

SUMMARY AND CONCLUSIONS

(1) The Lac de Gras kimberlites carried abundant eclo-gite and pyroxenite xenoliths to the surface Threemain eclogite types are distinguished in terms ofmajor-element composition and trace element

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1869

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

patterns high-Mg high-Ca and low-Mg eclogitesThe garnets of all eclogite types have anti-correlatedMg-number and Yb pointing to olivine fractionationas a controlling factor in the formation of their pre-cursor rocks Their flat or positively sloped HREEpatterns qualitatively indicate that the melt parentalto the precursor rocks was not in equilibrium with agarnet-bearing residue

(2) Reconstructed high-Mg eclogites and high-Caeclogites have flat HREE patterns positive Pb andSr anomalies and low ZrSm similar to oceanic gab-bros reported in the literature High-Mg eclogiteshave variably low HREE contents and some havesmall positive Eu anomalies possibly suggestive ofmore primitive relatively plagioclase-rich protolithswhereas high-Ca eclogites have higher HREE abun-dances and small or absent Eu anomalies that requiremore cpx- andor melt-rich protoliths if a subductionorigin is assumed

(3) Low-Mg eclogites have distinctive trace-element pat-terns that show a marked positive slope in the HREEand subchondritic ZrHf They resemble someArchaean second-stage melts (lsquoboninitesrsquo) and theirprotoliths appear to have formed from a previouslydepleted source

(4) After formation the eclogites were subject to a varietyof processes Most have low ratios of fluid-mobile vsless fluid-mobile elements possibly as a result of dehy-dration of the protoliths In addition some of the low-Mg and high-Ca eclogites have low LREEHREEconsistent with partial melt lossWhereas the low-Mgeclogites show concomitant enrichment in La Nb andZr some high-Mg and high-Ca eclogites show evi-dence for variable and unsystematic (decoupled)LREE and Nb addition

(5) Isotopically the trace-element characteristics of thedifferent eclogite types translate into lower eNd forhigh-Mg eclogites (negative slope in MREEfrac14 lowerSmNd) than for low-Mg eclogites (positive slope inMREEfrac14higher SmNd) Within the low-Mg groupsamples that show evidence for metasomatism (lowLuHf and SmNd) have lower eNd and eHf than asample that was apparently not metasomatized BothNd and Hf addition can be dated to 17Ga if identi-cal initial isotope compsitions are assumed for thelow-Mg eclogites In contrast the relatively constantand depleted mantle-like 87Sr86Sr of eclogitic cpxdespite variable Nd and Hf isotope compositions andparent^daughter ratios suggests recent input of juve-nile material

(6) Garnet and cpx in pyroxenites show major-elementrelationships different from those in eclogitesTherefore these two rock types are probably notrelated by fractionation processes The presence of

opx precludes direct crystallization from slab-derivedmelts Pyroxenites may be the product of (slab-derived) melts hybridized during assimilation of peri-dotitic mantle

(7) Garnets in phlogopite- and carbonate-bearing eclo-gites have extreme isotopic compositions includingthe most radiogenic Nd and Hf in the datasetNegative Eu anomalies may point to the involvementof a crustal protolith Highly radiogenic Sr radiogenicNd and unradiogenic Hf in the reconstructedphlogopite-bearing sample combined with highHREE abundances and evidence for Hf mobilitymay indicate involvement of solute-rich fluids origi-nating in a garnet- and rutile-free source

ACKNOWLEDGEMENTSHelp in the clean laboratories and with the analyticalfacilities by S Elhlou C Lawson A Sharma P Wielandand M Zhang is greatly appreciated Bill Griffin isthanked for his support and for agreeing to reviewseveral versions of the manuscript despite disagreeingwith many of the conclusions presented here The paperbenefited greatly from the detailed and constructivereviews provided by the editors Gareth Davies andMarjorie Wilson and by Matthias Barth Larry Heamanand an anonymous reviewer This work was funded by aMacquarie University International PostgraduateAward and Postgraduate Research Fund (SA) by anARC SPIRT grant sponsored by Kennecott Canada Incand by ARC Large and Discovery Grants to SYOrsquoRAnalytical data were obtained using instrumentationfunded by ARC LIEF and DEST Systemic InfrastructureGrants industry partners and Macquarie UniversityThis is Publication 481 from the ARC National KeyCentre for Geochemical Evolution and Metallogeny ofContinents (wwwesmqeduauGEMOC)

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online

REFERENCESAdam J GreenT H Sie S H amp Ryan C G (1997) Trace element

partitioning between aqueous fluids silicate melts and mineralsEuropeanJournal of Mineralogy 9 569^584

Aulbach S Stachel T Viljoen K S Brey G P amp Harris J W(2002) Eclogitic and websteritic diamond sources beneath theLimpopo Beltccedilis slab-melting the link Contributions to Mineralogy

and Petrology 143 56^70Aulbach S GriffinW L Pearson N J OrsquoReilly S Y Kivi K amp

Doyle B (2003) Origins of eclogites beneath the central SlaveCraton Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA-0011

Aulbach S Griffin W L Pearson N J OrsquoReilly S Y amp Kivi K(2004a) Mantle formation and evolution Slave Craton constraints

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1870

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

from HSE abundances and Re^Os isotope systematics of sulfideinclusions in mantle xenocrysts Chemical Geology 208 61^88

Aulbach S Griffin W L OrsquoReilly S Y amp McCandless T E(2004b) Genesis and evolution of the lithospheric mantle beneaththe Buffalo HeadTerrane Alberta (Canada) Lithos 77 413^451

Aulbach S GriffinW L Pearson N J OrsquoReilly S Y amp Doyle B(2007) Lithosphere formation in the central Slave Craton(Canada) plume subcretion or lithosphere accretion Contributionsto Mineralogy and Petrology (in press) doi 101007s00410-007-0200-1

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2001)Geochemistry of xenolithic eclogites from West Africa Part IA link between low MgO eclogites and Archean crust formationGeochimica et Cosmochimica Acta 65 1499^1527

Barth M G Foley S F amp Horn I (2002a) Partial melting inArchean subduction zones constraints from experimentally deter-mined trace element partition coefficients between eclogitic miner-als and tonalitic melts under upper mantle conditions PrecambrianResearch 113 323^340

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2002b)Geochemistry of xenolithic eclogites from West Africa part 2Origins of the high MgO eclogites Geochimica et Cosmochimica Acta

66 4325^4345Beard B L Fraracci K NTaylor L A Snyder G A Clayton R

A Mayeda T K amp Sobolev NV (1996) Petrography andgeochemistry of eclogites from the Mir kimberliteYakutia RussiaContributions to Mineralogy and Petrology 125 293^310

Becker H (1996) Geochemistry of garnet peridotite massifs fromlower Austria and the composition of deep lithosphere beneatha Palaeozoic convergent plate margin Chemical Geology 134 49^65

Becker H Jochum K P amp Carlson R W (2000) Trace elementfractionation during dehydration of eclogites from high-pressureterranes and the implications for element fluxes in subductionzones Chemical Geology 163 65^99

Bell K amp Tilton G R (2002) Probing the mantle The story fromcarbonatites EOS Transactions American Geophysical Union 83275^277

Benoit M Polvecurren M amp Ceuleneer G (1996) Trace element andisotopic characterisation of mafic cumulates in a fossil mantlediapir (Oman ophiolite) Chemical Geology 134 199^214

BleekerW (2003) The late Archean record a puzzle in ca 35 piecesLithos 71 99^134

BleekerW Ketchum JW F JacksonV A amp Villeneuve M (1999)The central Slave Basement Complex Part I Its structural topol-ogy and autochthonous core Canadian Journal of Earth Sciences 361083^1109

Blichert-Toft J amp Albare de F (1997) The Lu^Hf isotope geochemis-try of chondrites and the evolution of the mantle^crust systemEarth and Planetary Science Letters 148 243^258

Blichert-Toft J Chauvel C amp Albare de F (1997) Separation of Hfand Lu for high precision isotope analysis of rock samples by mag-netic sector-multiple collector ICP-MS Contributions to Mineralogy

and Petrology 127 248^260Brey G P amp Kolaquo hler T (1990) Geothermobarometry in four-phase

lherzolites II New thermobarometers and practical assessment ofexisting thermobarometers Journal of Petrology 31 1353^1378

Brey G P Kolaquo hlerT amp Nickel K G (1990) Geothermobarometry infour-phase lherzolites I Experimental results from 10 to 60 kbJournal of Petrology 31 1313^1352

Caporuscio F amp Smyth J (1990) Trace element crystal chemistryof mantle eclogites Contributions to Mineralogy and Petrology 105550^561

Carbno G B amp Canil D (2002) Mantle structure beneath the SWSlave Craton Canada constraints from garnet geochemistry in theDrybones Bay Kimberlite Journal of Petrology 43 129^142

Coleman R G Lee E D Beatty L B amp Brannock W W (1965)Eclogites and eclogites their differences and similaritiesGeological Society of America Bulletin 76 483^508

Creaser R A Grulaquo tter H Carlson J amp Crawford B (2004)Macrocrystal phlogopite Rb^Sr dates for the Ekati propertykimberlites Slave Province Canada evidence for multiple intru-sive episodes in the Paleocene and Eocene Lithos 76 399^414

Davies R M GriffinW L Pearson N J Andrew A S Doyle BJ amp OrsquoReilly S Y (1999) Diamonds from the deep Pipe DO-27Slave Craton Canada In Gurney J J Gurney J L Pascoe MD amp Richardson S H (eds) Proceedings of the 7th International

Kimberlite Conference CapeTown Red Roof Design pp 148^155Davies R M Griffin W L OrsquoReilly S Y amp Doyle B J (2004)

Mineral inclusions and geochemical characteristics of microdia-monds from the DO27 A154 A21 A418 DO18 DD17 and RanchLake kimberlites at Lac de Gras Slave Craton Canada Lithos 7739^55

DavisW J amp Hegner E (1992) Neodymium isotopic evidence for thetectonic assembly of Late Archean crust in the Slave Provincenorthwest Canada Contributions to Mineralogy and Petrology 111493^504

Davis W J Jones A G Bleeker W amp Grutter H (2003)Lithosphere development in the Slave craton a linked crustal andmantle perspective Lithos 71 575^589

Dawson J B (1984) Contrasting types of upper-mantle metasoma-tism In Kornprobst J (ed) Kimberlites II The Mantle and

Crust^Mantle Relationships Amsterdam Elsevier pp 289^294DePaolo D J (1981) Neodymium isotopes in the Colorado Front

Range and crust^mantle evolution in the Proterozoic Nature 291193^196

Erlank A J amp Kushiro I (1970) Potassium contents of syntheticpyroxenes at high temperatures and pressures Carnegie Institution ofWashingtonYearbook 68 433^439

Foley S F Barth M G amp Jenner G A (2000) Rutilemelt partitioncoefficients for trace elements and an assessment of the influence ofrutile on the trace element characteristics of subduction zonemagmas Geochimica et Cosmochimica Acta 64 933^938

Foley S F Buhre S amp Jacob D E (2003) Evolution of the Archaeancrust by delamination and shallow subduction Nature 421 249^252

Graham I Burgess J L Bryan D Ravenscroft P J Thomas EDoyle B J Hopkins R amp Armstrong K A (1999) Explorationhistory and geology of the Diavik Kimberlites Lac de GrasNorthwest Territories Canada In Gurney J J Gurney J LPascoe M D amp Richardson S H (eds) Proceedings of the 7th

International Kimberlite Conference Cape Town Red Roof Designpp 262^279

Green D H amp Wallace M E (1988) Mantle metasomatism byephemeral carbonatite melts Nature 336 459^462

Green T H Blundy J D Adam J amp Yaxley G M (2000) SIMSdetermination of trace element partition coefficients betweengarnet clinopyroxene and hydrous basaltic liquids at 2^75 GPaand 1080^12008C Lithos 53 165^187

GriffinW L amp OrsquoReilly SY (2007) Cratonic lithospheric mantle Isanything subducted Episodes 30(1) 43^53

Griffin W L Wass S Y amp Hollis J D (1984) Ultramafic xenolithsfrom Bullenmerri and Gnotuk maars Victoria Australia petrol-ogy of a sub-continental crust^mantle transition Journal of

Petrology 25 53^87GriffinW L Doyle B J Ryan C G Pearson N J OrsquoReilly SY

Davies R Kivi K van Achterbergh E amp Natapov LM (1999)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1871

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Layered mantle lithosphere in the Lac de Gras Area SlaveCraton composition structure and origin Journal of Petrology 40705^727

Griffin W L Pearson N J Belousova E Jackson S E vanAchterbergh E OrsquoReilly SY amp Shee S R (2000)The Hf isotopecomposition of cratonic mantle LAM-MC-ICPMS analysis ofzircon megacrysts in kimberlites Geochimica et Cosmochimica Acta

64 133^147Griffin W L OrsquoReilly S Y Doyle B J Pearson N J

Coopersmith H Kivi K Malkovets V amp Pokhilenko N(2004) Lithosphere mapping beneath the North American plateLithos 77 873^922

Grulaquo tter H S Apter D B amp Kong J (1999) Crust^mantle couplingevidence from mantle-derived xenocrystic garnets In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 307^312

Hart S R Blusztajn J Dick H J B Meyer P S ampMuehlenbachs K (1999) The fingerprint of seawater circulationin a 500-meter section of ocean crust gabbros Geochimica et

Cosmochimica Acta 63 4059^4080Harte B amp Kirkley M B (1997) Partitioning of trace elements

between clinopyroxene and garnet Data from mantle eclogitesChemical Geology 136 1^24

Heaman L M Creaser R A amp Cookenboo H O (2002) Extremeenrichment of high field strength elements in Jericho eclogitexenoliths A cryptic record of Paleoproterozoic subduction partialmelting and metasomatism beneath the Slave craton CanadaGeology 30 507^510

Heaman L M Kjarsgaard B A amp Creaser R A (2004) The tem-poral evolution of North American kimberlites Lithos 76 377^397

Heaman L M Creaser R A Cookenboo H O amp Chacko T(2006) Multi-stage modification of the northern Slave mantlelithosphere Evidence from zircon- and diamond-bearing eclogitexenoliths entrained in Jericho Kimberlite Journal of Petrology 47821^858

Hermann J Cpandler C Hack A amp Korsakov A (2006) Aqueousfluids and hydrous melts in high-pressure and ultra-high pressurerocks implications for element transfer in subduction zones Lithos92 399^417

Hills DV amp Haggerty S E (1989) Petrochemistry of eclogites fromthe Koidu Kimberlite Complex Sierra Leone Contributions to

Mineralogy and Petrology 103 397^422Hoffman P F (1989) Precambrian geology and tectonic history of

North America In Bally A amp Palmer A (eds) The Geology of

North AmericaccedilAn Overview Boulder CO Geological Society ofAmerica pp 447^512

IrelandT R Rudnick R L amp Spetsius Z (1994) Trace-elements indiamond inclusions from eclogites reveal link to Archean granitesEarth and Planetary Science Letters 128 199^213

Irvine G J Pearson D G Kjarsgaard B A Carlson R WKopylova M G amp Dreibus G (2003) Evolution of the litho-spheric mantle beneath Northern Canada a Re^Os isotope andPGE study of kimberlite-derived peridotite xenoliths fromSomerset Island and a comparison to the Slave and Kaapvaalcratons Lithos 71 461^488

Isachsen C E amp Bowring S A (1994) Evolution of the Slave cratonGeology 22 917^920

Jacob D E (2004) Nature and origin of eclogite xenoliths from kim-berlites Lithos 77 295^316

Jacob D E amp Foley S F (1999) Evidence for Archean ocean crustwith low high field strength element signature from diamondifer-ous eclogite xenoliths Lithos 48 317^336

Jacob D Jagoutz E Lowry D Mattey D amp Kudrjavtseva G(1994) Diamondiferous eclogites from Siberiaccedilremnants ofArchean oceanic-crust Geochimica et Cosmochimica Acta 585191^5207

Jacob D E Fung A Jagoutz E amp Pearson D G (2003) Petrologyand geochemistry of eclogite xenoliths from the Ekati kimberlitesarea Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA_239

Jacob D E Bizimis M amp Salters V J M (2005) Lu^Hf and geo-chemical systematics of recycled ancient oceanic crust evidencefrom Roberts Victor eclogites Contributions to Mineralogy and

Petrology 148 707^720Jerde E A Taylor L A Crozaz G Sobolev NV amp SobolevV N

(1993) Diamondiferous eclogites from Yakutia Siberiaccedilevidencefor a diversity of protoliths Contributions to Mineralogy and Petrology

114 189^202Jones A G Ferguson I J Chave A D Evans R L amp

McNeice GW (2001) The electric lithosphere of the Slave cratonGeology 29 423^426

Kelemen P B Shimizu N amp Dunn T (1993) Relative depletionof niobium in some arc magmas and the continental-crust-partitioning of K Nb La and Ce during meltrockreaction in the upper-mantle Earth and Planetary Science Letters 120111^134

Kinzler R J (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition Application to mid-ocean ridge basalt petrogenesis Journal of Geophysical Research 102853^874

Klemme S van der Laan S R Foley S F amp Guenther D (1995)Experimentally determined trace and minor element partitioningbetween clinopyroxene and carbonatite melt under upper mantleconditions Earth and Planetary Science Letters 133 439^448

Konzett J Armstrong R A Sweeney R J amp CompstonW (1998)The timing of MARID metasomatism in the Kaapvaal mantle anion probe study of zircons from MARID xenoliths Earth and

Planetary Science Letters 160 133^145Kopylova M G Russell J K amp Cookenboo H (1999) Mapping the

lithopshere beneath the north central Slave Craton In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 468^479

Krogh E (1988) The garnet^clinopyroxene iron^magnesiumgeothermometerccedila reinterpretation of existing experimentaldata Contributions to Mineralogy and Petrology 99 44^48

Kusky T M (1989) Accretion of Archean Slave Province Geology 1763^67

LeCheminant A N amp Heaman L M (1989) Mackenzie igneousevents Canada Middle Proterozoic hotspot magmatismassociated with ocean opening Earth and Planetary Science Letters

96 38^48LeCheminant A N Richardson D G DiLabio R N W amp

Richardson K A (1996) La recherche de diamants au CanadaGeological Survey of Canada Open File 3228 268pp

MacGregor I D amp Carter J L (1970) The chemistry of clinopyrox-enes and garnets of eclogite and peridotite xenoliths from theRoberts Victor mine South Africa Physics of the Earth and PlanetaryInteriors 3 391^397

MacGregor I D amp Manton W I (1986) Roberts Victor eclogitesancient oceanic crust Journal of Geophysical Research 91 14063^14079

MacKenzie J M amp Canil D (1999) Composition and thermal evo-lution of cratonic mantle beneath the central Archean SlaveProvince NWT Canada Contributions to Mineralogy and Petrology

134 313^324

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1872

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

McCandlessT E amp Gurney J J (1989) Sodium in garnet and potas-sum in clinopyroxene criteria for classifying mantle eclogites InRoss J (ed) Kimberlites and Related Rocks Carlton Blackwellpp 827^832

McDonoughW F amp Sun S-S (1995) The composition of the EarthChemical Geology 120 223^253

Navon O amp Stolper E (1987) Geochemical consequences of meltpercolation the upper mantle as a chromatographic columnJournal of Geology 95 285^307

Nikogosian I K amp Sobolev AV (1997) Ion-microprobe analysis ofmelt inclusions in olivine experience in estimating the olivine^meltpartition coefficients of trace elements Geochemistry International 35119^126

Norman M D Griffin W L Pearson N J Garcia M O ampOrsquoReilly S Y (1998) Quantitative analysis of trace element abun-dances in glasses and minerals a comparison of laser ablationinductively coupled plasma mass spectrometry solution inductivelycoupled plasma mass spectrometry proton microprobe and elec-tron microprobe data Journal of Analytical Atomic Spectrometry 13477^482

Norman M D Garcia M O amp Pietruszka A J (2005) Trace-element distribution coefficients for pyroxenes plagioclase andolivine in evolved tholeiites from the 1955 eruption of KilaueaVolcano Hawairsquoi and petrogenesis of differentiated rift-zonelavas American Mineralogist 90 888^899

OrsquoHara M J amp Herzberg C (2002) Interpretation of trace elementand isotope features of basalts Relevance of field relations petrol-ogy major element data phase equilibria and magma chambermodeling in basalt petrogenesis Geochimica et Cosmochimica Acta 662167^2191

PadghamW A amp FysonW K (1992) The Slave Province a distinctcraton CanadianJournal of Earth Sciences 29 2072^2086

Pearson D G Davies G R amp Nixon P H (1993) Geochemical con-straints on the petrogenesis of diamond facies pyroxenites from theBeni Bousera peridotite massif North Morocco Journal of Petrology34 125^172

Pearson N J Griffin W L Doyle B J OrsquoReilly S Y vanAchterbergh E amp Kivi K (1999) Xenoliths from kimberlitepipes of the Lac de gras area Slave Craton Canada In GurneyJ J Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite ConferenceCapeTown Red Roof Design pp 644^658

Perfit N R Gust D A Bence A E Arculus R J amp Taylor S R(1980) Chemical characteristics of island-arc basalts implicationsfor mantle sources Chemical Geology 30 227^256

Polat A Hofmann A W amp Rosing M T (2002) Boninite-likevolcanic rocks in the 37^38 Ga Isua greenstone belt WestGreenland geochemical evidence for intra-oceanic subductionzone processes in the early Earth Chemical Geology 184 231^254

Poudjom DjomaniY H GriffinW L OrsquoReilly SY amp Doyle B J(2005) Lithospheric domains and controls on kimberlite emplace-ment Slave Province Canada evidence from elastic thickness andupper mantle composition Geochemistry Geophysics Geosystems 6doi1010292005GC0009978

Presnall D C Gudfinnsson G H amp Walter M J (2002)Generation of mid-ocean ridge basalts at pressures from 1 to 7GPa Geochimica et Cosmochimica Acta 66 2073^2090

Schmidberger S S Heaman L M Simonetti A Creaser R A ampCookenboo H O (2005) Formation of Paleoproterozoiceclogitic mantle Slave Province (Canada) Insights from in-situHf and U^Pb isotopic analyses of mantle zircons Earth and

Planetary Science Letters 240 621^633

Schmidberger S S Simonetti A Heaman L M Creaser R A ampWhiteford S (2007) Lu^Hf in-situ Sr and Pb isotope and traceelement systematics for mantle eclogites from the Diavik diamondmine Evidence for Paleoproterozoic subduction beneath the Slavecraton Canada Earth and Planetary Science Letters 254 55^68

Schulze D J amp Helmstaedt H (1988) Coesite-sanidine eclogitesfrom kimberlite products of mantle fractionation or subductionJournal of Geology 96 435^443

Smithies R H Champion D C amp Sun S S (2004) The case forArchaean boninites Contributions to Mineralogy and Petrology 147705^721

Smyth J R Caporuscio F A amp McCormick T C (1989) Mantleeclogites evidence of igneous fractionation in the mantle Earth andPlanetary Science Letters 93 133^141

Sobolev NV amp Lavrentrsquoyev Y G (1971) Isomorphic sodium admix-ture in garnets formed at high pressure Contributions to Mineralogy

and Petrology 31 1^12Stalder R Foley S F Brey G P amp Horn I (1998) Mineral aqueous

fluid partitioning of trace elements at 900^12008C and 30^57GPa New experimental data for garnet clinopyroxene and rutileand implications for mantle metasomatism Geochimica et

Cosmochimica Acta 62 1781^1801Sun S-S amp McDonough W F (1989) Chemical and isotopic sys-

tematics of oceanic basalts implications for mantle compositionsand processes In Saunders A D amp Norris M J (eds)Magmatism in the Ocean Basins Geological Society London Special

Publications 42 313^345Taura H Yurimoto H Kurita K amp Sueno S (1998) Pressure

dependence on partition coefficients for trace elements between oli-vine and the coexisting melts Physics and Chemistry of Minerals 25469^484

Taylor L A amp Neal C R (1989) Eclogites with oceanic crustal andmantle signatures from the Bellsbank kimberlite South AfricaPart I Mineralogy petrography and whole rock chemistryJournal of Geology 97 551^567

van Achterbergh E Ryan C G amp Griffin W L (1999) Glitter online intensity reduction for the laser ablation inductively coupledplasma spectrometry Abstract 9th V M GoldschmidtConference LPI Boston MA 905 pp Abstract no 7215

van Breemen O DavisW J amp King J E (1992) Temporal distribu-tion of granitoid plutonic rocks in the Archean Slave ProvinceNorthwest Canadian Shield Canadian Journal of Earth Sciences 292186^2199

Vervoort J D Patchett P J Blichert-Toft J amp Albare de F (1999)Relationship between Lu^Hf and Sm^Nd isotopic systems in theglobal sedimentary system Earth and Planetary Science Letters 16879^99

Viljoen K S Smith C B amp Sharp Z D (1996) Stable and radio-genic isotope study of eclogite xenoliths from the Orapa kimberliteBotswana Chemical Geology 131 235^255

Walter M J (1998) Melting of garnet peridotite and theorigin of komatiite and depleted lithosphere Journal of Petrology 3929^60

Wasserburg G J Jacobsen S B DePaolo D J McCulloch M Tamp Wen T (1981) Precise determination of SmNd ratios Sm andNd isotopic abundances in standard solutions Geochimica et

Cosmochimica Acta 45 2311^2323Westerlund K J Shirey S B Richardson S H Carlson R W

Gurney J J amp Harris JW (2006) A subduction wedge origin forPaleoarchean peridotitic diamonds and harzburgites from thePanda kimberlite Slave craton evidence from Re^Os isotope sys-tematics Contributions to Mineralogy and Petrology 152 275^294

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1873

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Page 9: Origins of Xenolithic Eclogites and Pyroxenites from the Central

garnets (these include kyanite- graphite- and diamond-bearing varieties) (2) eclogites with low-Mg garnets (theseinclude quartz-bearing eclogite) (3) eclogites with high-Mggarnets and Cr2O3 contents502wt (4) volatile-richeclogites (phlogopite- apatite- carbonate-bearing)

Clinopyroxene (cpx)Clinopyroxenes in all eclogite groups have higher averageAl2O3 contents than those in pyroxenites with the formerhaving about 90 of the total Al in the jadeite molecule(AlVI) versus about 35 for the latter corresponding toan average of 41 and 25wt Na2O respectivelyThe highest Al2O3 and Na2O contents in the dataset areobserved in cpx from the high-Ca eclogites (up to158wt Al2O3) (Electronic Appendix 3) Contents ofK2O are generally below the detection limit(5004wt ) but values up to 024wt are observed insome cpx in high-Ca eclogites Average cpx Mg-numberare highest in pyroxenites (899) followed by cpx inhigh-Mg eclogites (865) high-Ca eclogites (836) andlow-Mg eclogites (801) Pyroxenitic cpx tend to havelower TiO2 and higher NiO and distinctly higherCr2O3 contents compared with those in eclogites A plotof MgO vs Al2O3 shows a negative correlation (Fig 6)Clinopyroxenes in low-Mg eclogites have lower Al2O3

contents at a given MgO content than other eclogitetypes Between 15 and 17wt MgO the Al2O3^MgOcorrelation for cpx in pyroxenites has a different slopefrom that in the eclogites Clinopyroxenes in pyroxenites

VR19673ecl-g1 and VR19673ecl-g2 have distinctly lowerMgO contents and higher Al2O3 contents than inthe other pyroxenites They coexist with garnets havingthe lowest Cr2O3 and CaO contents of the dataset whichdefine one end of the positive Cr2O3^CaO correlationobserved for garnetsClinopyroxenes in volatile-rich eclogites have distinctly

high MnO contents (up to 024wt ) Mg-numbers forcpx in the phlogopite-bearing variety are 826 and 828whereas that in the calcite-bearing eclogite (VR67112b)has an Mg-number of 647Some cpx grains have spongy rims that contain small

(mm-scale) patches of glass these rims are richer in Caand Mg and markedly poorer in Na than the cores similarto cpx in eclogites from southern Africa (Taylor amp Neal1989) The rim texture is unequilibrated suggesting lategrowth The formation of the spongy cpx rims might berelated to incongruent melting of cpx or be a result ofmetasomatism (Taylor amp Neal 1989 and referencestherein) Distinct although not spongy rims in two eclo-gites (VR67360 and YK1946) are similarly enriched inCaO and depleted in Na2O Two types of cpx with respectto CaO Na2O and Mg are present in VR43477VR43479 and VR43480 without clear core^rim zonationor spongy rims No significant differences are recognizedwith respect to other oxides

Orthopyroxene (opx)Orthopyroxenes have Mg-number ranging from 804 to933 (17 of 19 samples have values between 895 and 933)and typically contain minor amounts of Al2O3

(032^118wt one outlier has 318wt ) CaO(020^118 wt ) and NiO (008^014wt ) (Electronic

0

2

4

6

8

10

12

14

7 9 11 13 15 17 19

Al 2

O3

CPX

pyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

cpx in pyroxenitesVR19673ecl-g1g2

MgO

Fig 6 MgO vs Al2O3 (wt ) in cpx Data sources as in Figs 3and 4

0

005

010

015

020

2 4 6 8 10 12 14 16

CaO

Na 2

O

dl

group Igroup II

pyroxenites

low-Mghigh-Mghigh-Cavolatile-rich

inclusions in diamondJericho anisotropicJericho massiveDiavik low-TDiavik high-TDiavik diamond-bearing

eclogites

GARNET

Fig 5 CaO vs Na2O (wt ) in garnet Arrows show trends foreclogites and pyroxenites respectively [note that part of thepyroxenite trend extends into the field falling below the detectionlimit (dl) for Na2O] Group I and group II refer to classificationof McCandless amp Gurney (1989) Data sources as in Figs 3 and 4

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1851

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Appendix 4) The contents of all other elements are mostlybelow their respective detection limits

Accessory phasesRutile and ilmenite occur in some of the eclogites (Table 1)All rutile grains have exsolved 10 ilmenite as lamellaeand rims Ilmenite-free areas of rutile in eclogites containbetween 962 and 988wt TiO2 and minor amounts ofAl2O3 (01^04wt ) Cr2O3 (5009 to 02wt ) andFeO (03^07wt ) Rutile in pyroxenite YK1915 contains004wt Al2O3 01wt FeO and 15wt Cr2O3Discrete ilmenite grains in sample VR67112b haveTiO2 contents of 499 and 500wt MgO of 23 and15wt and FeO of 399 and 403wt respectivelySulfides occur interstitially or enclosed in cpx garnet

and opx they are mostly pyrrhotite and monosulfide solidsolution with variable Ni contents (unpublished data)They will not be considered further hereAbundant platy phlogopite is observed in two samples

(VR19674ecl-g7 and YK1911) It is Cl-rich in sampleVR19674ecl-g7 (063wt vs 003wt in YK1911) andF-rich in sample YK1911 (097wt vs 005wt inVR19674ecl-g7) (Electronic Appendix 5) The same is truefor coexisting apatite (VR19674ecl-g7 254wt Cl and014wt F YK1911 006wt Cl and 306wt F)

Comparison with mineral inclusions indiamonds from Lac de Gras and witheclogite xenoliths from other Slave CratonlocalitiesMineral inclusions in diamonds from the Lac de Graskimberlites have been investigated by Davies et al (19992004) For garnets there is a striking similarity betweenthe inclusions in diamond and those in the high-Ca eclogitexenoliths although the inclusions in diamond trendtowards higher Na2O (Fig 5) and lower Al2O3 contentsThe overlap is not as marked for cpx where inclusionsin diamond span almost the entire range observed foreclogitic cpx (Fig 6)A comparison with eclogites from other localities in

the Slave Craton shows that garnets in most Diavik low-temperature eclogites reported by Schmidberger et al(2007) plot with garnets in high-Mg eclogites and inpyroxenites from this study (Figs 3^5) whereas garnets inhigh-temperature and diamond-bearing eclogites(Schmidberger et al 2007) overlap most with garnets inhigh-Ca eclogites from this study and with garnetsincluded in diamond Eclogites from the Jericho kimberlitehave either massive or anisotropic fabrics (Kopylova et al1999) Garnets in almost all massive eclogites plot withgarnets in low-Mg eclogites from this study with regard toCaO^MgO^FeO relationships (Fig 3) but many havelower Na2O contents at a given CaO content (Fig 5)Garnets in anisotropic eclogites are restricted to lowMgOFeO (507) and show some overlap with garnets

in high-Ca eclogites although trending towards lowerCaO(MgOthornFeO) (Fig 3) while coexisting cpx is alsorestricted to low MgO and trends toward high Al2O3 con-tents (Fig 6)

TRACE ELEMENTSGarnetTrace-element abundances in garnet are summarized inTable 3 and the complete dataset is reported in ElectronicAppendix 1 Garnets in low-Mg eclogites show steep posi-tive slopes between LaN and SmN and shallower positiveslopes in the normalized heavy rare earth elements(HREEN Fig 7a) The normalized light REE (LREEN)show an order of magnitude variability ZrHf is mostlysupra-chondritic and Ti can be enriched or depleted rela-tive to elements of similar compatibility Niobium in threesamples is strongly depleted at 002^003 chondriticGarnets in high-Mg eclogites also show smooth positiveslopes in LREEN but have flat HREEN and variablenormalized middle REE (MREEN) to HREEN (Fig 7b)ZrHf is always supra-chondritic and Ti enriched ordepleted relative to similarly compatible elementsNiobium is most samples is below detection Small positiveEu anomalies though not outside the analytical uncer-tainty for Eu relative to either Sm or Gd are observed forgarnet in some high-Mg eclogites (eg EuEu up to 15for VR67360 where Eu is the average of the chondrite-normalized Sm and Gd concentrations) Garnets in high-Ca eclogites are slightly enriched in the MREEN relative tothe HREEN or have flat MREEN and HREEN with ahump between EuN and HoN (Fig 7c) One sample(VR40345) shows strong La enrichment (greater thanchondrite) Two samples have supra-chondritic ZrHfwhereas three samples have close to chondritic ratios andone sample has a subchondritic ratio As is true for theother eclogite typesTiN is either enriched or depleted rela-tive to similarly compatible elementsPyroxenitic garnets have variable slopes in the LREEN

and flat MREEN to HREEN patterns and two samplesshow high Sc relative to neighbouring elements Garnet inone of the pyroxenites (YK1952) has a highly distinctivepattern relative to other pyroxenites with very low LREEand HFSE (Fig 7d) This sample consists of coarse pyroxe-nite schlieren in which the garnet was analysed in a finer-grained lsquomatrixrsquo of spongy cpx with secondary rims finecpx^opx intergrowths embayed garnet and melt patchesVolatile-rich eclogites have garnets that are slightly

enriched in MREEN relative to HREEN or have flatMREE to HREE patterns One garnet has a pronouncednegative Eu anomaly (EuEufrac14 07 forVR67112b Fig7e)Garnets in volatile-rich samples have the lowest Ni Coand Ti and the highest V ScYand HREE abundances inthe dataset

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1852

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 3 Summary of trace-element contents in garnet and cpx (ppm)

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR19674

ecl-g7

VR67112b YK1911 av min max 1s

n 6 6 7 3 5 4 4

P 169 54 270 88 196 70 310 78 282 80 450 138 160 60 80 214 48 370 144

Sc 62 40 89 16 48 36 67 11 53 42 79 15 69 92 93 186 52 440 179

Ti 1667 490 2500 680 1073 620 2000 455 2141 700 4700 1582 270 690 260 1097 540 1900 573

V 169 78 250 68 150 78 280 72 178 110 390 101 140 340 100 390 71 820 326

Co 59 42 70 11 58 39 84 13 58 47 65 5 30 41 27 68 55 90 16

Ni 17 5 32 10 21 11 34 10 24 11 35 8 6 10 51 54 25 110 38

Ga 12 7 15 3 13 8 16 3 12 9 20 4 8 86 58 12 10 14 2

Sr na 512 4 na na 522 522 na na 57 57 na 52 5025 510 513 513 513 na

Y 37 6 61 18 21 7 38 10 31 20 38 5 60 81 70 19 12 26 6

Zr 15 5 29 9 28 6 83 26 24 9 43 12 60 22 90 18 1 41 17

Nb 009 0005 048 019 na 50015 05 na na 5005 05 na 5003 0012 5001 022 0030 057 025

La 005 001 009 004 na 50007 00 na na 5003 04 na 004 002 0009 003 000 009 005

Ce 03 01 05 02 01 00 01 00 05 01 18 05 03 022 015 01 00 03 01

Pr 01 00 02 01 00 00 01 00 02 01 03 00 017 014 012 00 00 01 00

Nd 12 03 24 07 07 04 14 03 21 12 32 07 24 23 20 03 00 07 03

Sm 11 04 22 06 10 03 18 05 23 08 36 10 29 31 33 05 01 09 04

Eu 06 03 10 03 06 02 10 03 11 04 17 04 16 114 16 03 01 06 02

Gd 27 07 41 12 23 07 42 12 43 16 53 13 8 81 90 13 06 24 08

Dy 54 11 70 23 36 13 65 18 58 35 71 11 12 144 12 29 18 42 10

Ho 14 02 23 07 08 03 15 04 12 09 14 02 24 30 23 07 04 10 02

Er 46 07 80 24 24 09 40 10 34 25 40 05 8 85 6 22 14 28 06

Yb 52 06 98 30 24 10 37 08 32 22 39 05 8 70 5 23 13 30 07

Lu 09 01 16 05 04 02 05 01 05 03 07 01 11 09 08 04 02 05 01

Hf 03 02 05 01 04 01 10 03 05 02 09 02 09 027 007 03 00 06 03

Ta na 50008 5004 na na 50005 001 na na 50004 na na 003 50002 50005 na 50007 011 na

Pb na 5005 02 na na 5002 02 na na 5003 07 na 02 004 5002 na 5003 01 na

Th na 5001 005 na na 50007 003 na na 50008 011 na 5007 0012 50004 002 0005 006 002

U na 5003 5003 na na 5001 01 na na 5007 na na 01 0023 010 na 5001 003 na

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR67112b YK1911 av min max 1s

n 6 4 8 5 4 4

P 65 15 120 41 39 14 56 16 53 30 70 12 20 9 80 27 120 40

Sc 22 10 30 7 15 14 19 2 14 7 24 5 80 43 17 11 21 5

Ti 1557 990 2400 508 1168 1010 1490 191 1988 810 3500 1154 2390 1700 689 102 1685 694

V 415 180 480 107 318 210 400 69 324 259 450 72 600 180 250 139 480 156

Co 35 26 50 8 20 14 31 7 26 14 42 11 49 23 29 20 42 10

Ni 164 87 220 47 228 100 430 128 188 135 320 56 59 34 428 410 670 184

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1853

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

ClinopyroxeneTrace-element abundances in cpx are summarized inTable 3 and the complete dataset is reported in ElectronicAppendix 6 Clinopyroxenes in low-Mg eclogites havemildly positive to negative slopes from LaN to NdN andsmooth negative slopes between SmN and HoN thatmostly continue to LuN (Fig 8a) Niobium is below detec-tion or 5007 chondritic and SrN is variably enrichedrelative to neighbouring elements with the peak size corre-lating negatively with LREEN ZrHf is subchondritic andHfN is enriched relative to SmN whereasTiN is enriched ordepleted relative to similarly compatible elementsWe ana-lysed the spongy rim in cpxYK1943 and its trace-elementpattern is similar to that of cpx cores for more compatibleelements whereas the most incompatible elements areseveral orders of magnitude higher Clinopyroxenes in

high-Mg eclogites have steeper negative slopes betweenNdN and LuN than those in low-Mg eclogites combinedwith lower HREE abundances HfN is depleted relative toSmN (Fig 8b) High-Ca eclogites contain cpx with sinusoi-dal REEN patterns with lower LREE than other eclogitetypes and the heaviest REE below detection They haveuniformly strong positive Sr peaks relative to elements ofsimilar compatibility (Fig 9) The diamond-bearing high-Ca eclogite (VR40345) has cpx with unusually high Nband LREE abundancesClinopyroxenes in four pyroxenites have disparate trace

element patterns They all share depletions in TiN relativeto similarly compatible elements and two show positiveSrN peaks Two samples have cpx with high and flatLREEN to MREEN patterns and steep negative slopes inthe HREEN The trace-element patterns of cpx in YK1914

Table 3 Continued

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR67112b YK1911 av min max 1s

n 6 4 8 5 4 4

Ga 16 10 20 3 12 9 15 2 26 15 39 9 129 123 6 11 11 5

Rb na 5008 04 na na 5007 05 02 na 5003 03 na 02 5004 03 010 05 0

Sr 166 85 290 74 222 153 285 50 243 83 440 142 85 120 200 11 414 209

Y 2 04 4 10 13 07 2 04 11 03 3 10 28 51 2 03 3 2

Zr 19 4 48 15 25 8 45 13 17 8 28 7 111 129 36 008 81 42

Nb 01 0003 03 01 na 5002 00 na na 5002 16 9 007 011 010 003 02 008

Cs na 5001 6 na na 5006 5006 na5013 5004 5013 na bdl 5002 na 5001 5002 na

Ba 09 02 2 07 09 01 2 07 10 03 3 10 06 01 13 06 3 10

La 3 03 8 3 3 23 3 04 4 01 14 6 51 97 4 01 10 4

Ce 7 13 20 6 10 7 13 2 7 05 24 10 25 28 11 01 28 13

Pr 11 02 3 08 2 10 3 06 08 02 2 09 47 46 2 002 4 2

Nd 5 12 12 4 10 5 14 4 4 09 8 26 23 22 8 01 22 10

Sm 11 05 21 06 20 10 31 10 07 02 11 03 47 53 2 005 5 2

Eu 03 01 05 01 06 028 09 03 02 010 03 007 087 123 06 002 15 07

Gd 08 03 13 03 11 05 19 06 05 03 06 011 28 37 15 010 36 2

Dy 05 01 07 02 04 022 06 01 02 013 03 007 11 16 07 008 14 06

Ho 01 00 02 004 006 003 007 002 na 5003 012 na 011 021 009 001 02 007

Er na 507 507 na 011 004 02 00 na 5006 03 na 019 033 02 002 03 012

Yb na 5001 02 na na 5001 011 na na 5006 03 na 007 015 009 001 02 006

Lu na 5001 003 na na50002 002 na na50008 01 na 0006 0013 0013 0002 003 001

Hf 12 03 24 07 11 04 18 05 10 06 16 04 34 53 na50015 35 na

Ta na50003 50015 na na50006 006 na na50004 5004 na 0014 00236 na50003 503 na

Pb 06 02 12 04 06 04 09 02 na 5016 07 02 065 098 08 01 18 1

Th na50005 06 na 012 005 02 005 na 5002 5006 na 021 02 001 06 03

U na 5005 5005 na 002 0009 005 001 na 5003 5005 na 0013 024 na50005 003 na

Averaged for the three eclogite types and for pyroxenites as in Table 2 pyx pyroxenite n number of analyses averagedbdl below detection limit

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1854

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

and YK1952 are similar although at different abundanceswith only modest LREEN over HREEN enrichment andstrong depletions in ZrN relative to neighbouring elements(Fig 8d)Clinopyroxenes in the volatile-rich samples have flat pat-

terns and high abundances from LaN to SmN and smoothnegative slopes from SmN to LuN Niobium Sr and Ti are

noticeably depleted relative to elements of similar compat-ibility (Fig 8e)

Trace-element partitioningTo assess whether coexisting phases are in trace-elementequilibrium we calculated apparent cpx^garnet distribu-tion coefficients and compared them with those

(a)

(d)

(b)

(e)

(c)

GarnetCI-chondrite GarnetCI-chondrite

low-Mghigh-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1952

1

10

Nd Sm Eu Gd Dy

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Fig 7 Extended trace-element patterns of garnet in (a) low-Mg eclogites (b) high-Mg eclogites with inset showing enlargement of MREE tohighlight Eu anomalies and 1s uncertainties (c) high-Ca eclogites (d) pyroxenites and (e) volatile-bearing eclogites Normalized to chondritevalues of McDonough amp Sun (1995)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1855

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

determined experimentally or in demonstrably equili-brated natural samples Low-Mg and high-Mg eclogiteshave Dcpxgarnet for most elements that are within therange determined for natural equilibrated eclogites andexperimental assemblages (Harte amp Kirkley 1997 Greenet al 2000 Barth et al 2002a) (Electronic Appendix 7)High-Ca eclogites plot at the low end of the range

of distribution coefficients and have very low DMREE

and DHREE outside the range indicated by the datasetof Harte amp Kirkley (1997) who noted that DREE decreaseswith increasing Ca content in eclogitic minerals Theseassemblages are therefore considered to be in equilibriumPyroxenite VR40384 has D values similar to eclogites

but at the high end of the range determined by Harte amp

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

(a)

(d)

(b)

(e)

(c)

CpxCI-chondrite CpxCI-chondrite

low-Mg high-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1914

YK1952

spongy rim

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

VR40345

Fig 8 Extended trace-element patterns of cpx as in Fig 7

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1856

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Kirkley (1997) (Electronic Appendix 7) Another group(YK1914 and YK1952) plots at the low end but DHf DZrDTi and DV are far below the values indicated by experi-mental datasets Because pyroxenites compositionallygrade into eclogites their drastically different DHFSE

values suggest disequilibrium In contrast volatile-bearingsamples have DHFSE values that appear to be too high to bein equilibrium however this may be a consequence oftheir unusual bulk chemical composition (Ti- and K-rich)which differs substantially from that of lsquonormalrsquo eclogitesand experimental charges

GEOTHERMOBAROMETRYFollowing Schmidberger et al (2007) temperatures oflast equilibration were calculated for eclogitic andopx-free pyroxenitic assemblages using the formulationof Krogh (1988) The results are summarized in Table 2and Electronic Appendix 8 for a preset pressure of5 GPa and illustrated as a histogram in Fig 9 Foropx-bearing pyroxenites temperatures were calculatedsimultaneously with pressure [Ca-in-opx thermometryand Al-in-opx barometry using the formulations of Breyet al (1990)] to constrain their relative positions in themantle columnMinerals in some eclogites from Lac de Gras show dis-

tinct core and rim compositions or inhomogeneitiesTemperatures were not calculated for samples withspatially irresolvable inhomogeneous garnet compositions(SE01VR43479VR43480) For eclogites with spongy cpxcore compositions were used in the calculationTemperatures were calculated for both core and rim com-positions where applicable For all samples these

temperatures agree within the uncertainty of this thermo-meter (508C Brey amp Kolaquo hler 1990)Most equilibration temperatures for eclogites lie between

800 and10008C (Fig 9)The transition between the shallowand the deep lithospheric mantle layers in the central Lacde Gras area lies in this temperature interval (Griffin et al1999 Pearson et al 1999) Eclogites in this study lack thedistinct bimodal temperature distribution observed bySchmidberger et al (2007) but as in their study high-Caeclogites give higher average temperatures (10808C corre-sponding to a depth of 155 km if equilibrated along a 40mWm2 geotherm) than other eclogites types (9808Cdepth 135 km) the diamond-bearing sample (VR40345)gives thehighest temperature of the datasetHighNa2Ocon-tents in garnet from high-Ca eclogites are qualitativelyconsistent with higher equilibration pressures (Sobolev ampLavrentrsquoyev 1971) as are high K2O contents in coexistingcpx (Erlank ampKushiro1970) suggesting that high-Ca eclo-gites not only equilibrated at high temperature but also highpressure rather than representing a temperature excursionPyroxenites both opx-bearing and opx-free give somewhatlower average temperatures than eclogites (mean 9258C)and phlogopite-rich eclogites give the lowest temperatures(685 and 7698C respectively) The calcite-bearing eclogiteVR67112b gives a relatively low equilibration temperatureof 8558C

MEASURED ANDRECONSTRUCTEDWHOLE -ROCKSRationale and approachKimberlite-borne xenoliths sometimes show evidence forinfiltration of the host kimberlite (eg Taylor amp Neal1989 Barth et al 2001 2002b Heaman et al 2006) Wetherefore follow the standard approach of calculatingwhole-rock compositions from the compositions of the con-stituent phases weighted by their modal abundances Onlycore compositions were used as distinct rim compositionsmay point to late diffusion of elements from kimberlite-related melt veinsMany eclogitic whole-rock compositions reconstructed

with only garnet and cpx have negative Ti anomalies(not shown) For most of these samples rutile or ilmenitehas been identified as part of the mineral assemblage andthese phases should therefore be considered in the whole-rock reconstruction Rutile or ilmenite modes were esti-mated by assuming that the bulk-rock has no Ti anomalywhen normalized to a normal mid-ocean ridge basalt(N-MORB) composition and calculating the amount ofrutile required to erase theTi anomalyIt is of course possible that the eclogites had negative

or positive Ti anomalies relative to elements of similarcompatibility (Eu and Dy) Because we reconstructed

0

2

4

6

8

10

12volatile-rich

pyroxenite opx-bearing

pyroxenite opx-free

high-Ca

high-Mg

low-Mg

700750

800850

900950

10001050

11001150

12001250

Fre

quen

cy

Temperature

Fig 9 Histogram ofTKrogh (Krogh 1988) obtained at a preset pres-sure of 5 GPa except opx-bearing pyroxenites where TBKN wassolved simultaneously with PBKN (Brey amp Kolaquo hler 1990)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1857

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

whole-rock compositions from mineral core compositionskimberlite infiltration is unlikely to have caused fractiona-tion of Ti from Eu and Dy In a study in which rutilemodes were estimated from the differences between Ti inmeasured bulk-rocks and in bulk-rocks reconstructed withonly garnet and cpx only three of 19 eclogites recon-structed with rutile show Ti anomalies (Barth et al2002a) Even eclogite xenoliths with depleted Zr and Hfabundances that resemble those of modern island arcbasalts do not show Ti anomalies (Jacob amp Foley 1999)We therefore consider our approach to be a reasonablyrobust means of obtaining meaningful rutile modes wheresmall sample sizes preclude reliable determination of themodes of accessory phases by point-countingWhere rutilewas not recognized on thin-section scale or not analysedthe average eclogitic rutile composition was used for thereconstruction

UncertaintiesReconstructing whole-rock compositions in this way is arather crude method given the coarse grain size of theeclogites relative to the sample size and the non-uniformdistribution of the principal minerals on thin-sectionscale The uncertainty in mineral modes of medium- tocoarse-grained eclogites with grains45mm has been esti-mated by Jerde et al (1993) to be 10 plus uncertaintiesrelated to partial alteration which is minor in our samplesNevertheless the whole-rock trace-element budget israther insensitive to the exact modes for common eclogiticassemblages and REE patterns which encompass mostother elements in terms of compatibility in garnet andcpx do not change significantly even for modal variationsof 30 (Jerde et al 1993) This is not true for major-element concentrations which vary substantially with var-iations in mineral modes and we therefore do not calculatewhole-rock major-element contentsBecause the discussion that follows is based on trace-

element patterns and trends rather than absolute abun-dances whole-rock reconstruction based on mineralmodes is adequate We are confident that the uncertaintyinherent in this approach is not leading to significant mis-representation of eclogite whole-rock compositions becausethe trace-element patterns of the reconstructed whole-rocks show distinct differences (see below) that correlatewith the assignment to eclogite types based on garnetmajor-element composition and this is independent ofmineral mode Following Jerde et al (1993) in ElectronicAppendix 9 we show variations in trace-element patternfor one sample in each eclogite type for variations ingarnet and cpx abundance from 7030 to 3070 vol This shows that the trace-element patterns are generallyinsensitive to cpxgt ratios with regard to slopes andanomalies such as positive Sr Pb and Eu anomalies withthe exception of the positive Eu anomaly in high-Mgeclogite VR43479 which disappears if a high cpx mode is

assumed (not shown) Variations in rutile modes by 50(Electronic Appendix 9d) lead as would be expected tothe appearance of Ti anomalies but a rutile mode of 05instead of 1 vol does not erase the positive Nb anomalyin low-Mg eclogiteVR50909Because of the robustness of the relative trace-element

abundances in the reconstructed eclogites a garnet propor-tion of 065 and cpx proportion of 035 approximately themedian modal abundances of these minerals in volatile-free eclogites from Lac de Gras are assumed for eclogitesamples that are too small for determination of meaningfulmodes by point-counting

Within- and between-group variationsThe trace-element compositions of reconstructed whole-rocks are given in Table 4 N-MORB-normalized trace-element patterns of reconstructed whole-rocks areshown in Fig 10 The three eclogite groups have distincttrace element patterns that are particularly obvious forthe REE (Fig 11) Low-Mg eclogites have positive Pb andSr anomalies and smooth positive slopes in their MREEN

and HREEN LREEN are enriched relative to MREEN inthree samples (Fig 11a) Three samples have high Nbrelative to similarly compatible elements and comparedwith most other eclogites (Fig 10a) All high-Mg eclogitesdisplay more or less pronounced positive Eu and Sranomalies and most have positive Pb and negative Zrand P anomalies Their N-MORB-normalized REE pat-terns show shallow positive slopes from LaN to PrN orNdNand aweak negative slope from NdN to LuN (Fig11b)VR43479 is a low-Mg eclogite with regard to its garnet

Ca^Fe^Mg composition but has a trace-element patterntypical of high-Mg eclogites and will be discussed withthe latter Conversely sample YK1949 a high-Mg eclogitein terms of its garnet Ca^Fe^Mg compostion is more simi-lar to low-Mg eclogites in terms of its whole-rock trace-element pattern and is plotted in Figs 10a and 11aWhereas the patterns of most samples within the groupresemble each other the N-MORB-normalized incompati-ble-element abundances of the high-Ca eclogites vary bymore than an order of magnitude (Fig 10c) All sampleshave positive Pb anomalies and most have positive Sranomalies Some samples have negative P andor Zranomalies Niobium abundances in three of the samplesare very high similar to those for some low-Mg eclogitesHigh-Ca eclogites tend to have flatter MREEN andHREEN patterns and more uniform HREE abundancesthat the other eclogite types with humps between SmN

and HoNThe phlogopitized sample YK1911 shows a strong posi-

tive Pb anomaly and lower HREE abundances than thecalcite-bearing sample VR67112b which shows a distinctnegative Eu anomaly and a depletion of La relative to Ce(Figs 10d and 11d)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1858

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Trace-element compositions of reconstructed whole-rocks (ppm)

Sample P Sc Ti V Co Ni Ga Sr Y Zr Nb La Ce Pr Nd

Low-Mg eclogites

vr19674-g4 186 50 4649 317 57 59 144 77 24 30 36 13 31 05 24

vr43452 154 39 2253 328 49 84 176 52 32 11 01 02 07 01 08

vr43469 38 65 4377 349 54 91 142 34 28 15 02 05 15 03 17

yk1926 144 45 3406 258 52 87 145 89 18 18 20 09 29 06 33

yk1943 173 46 5732 295 41 39 146 122 25 48 82 34 85 13 64

yk19491 102 34 1693 158 31 49 100 46 10 14 03 09 21 03 17

High-Mg eclogites

vr19677-g2 138 39 5160 291 41 53 119 98 14 22 04 13 53 11 64

vr434792 46 25 1265 129 36 126 84 56 3 7 04 06 25 04 23

vr67360 49 36 1573 138 64 184 111 76 5 9 04 09 38 07 33

yk1946 153 28 4877 192 45 105 124 108 12 48 03 10 38 08 50

High-Ca eclogites

vr19674-g5 267 30 7663 200 41 67 195 32 23 43 77 01 04 02 16

vr40345 0 33 2821 175 47 98 210 179 13 13 66 60 109 11 38

vr43465 142 27 5104 220 31 83 121 197 14 35 29 05 22 05 40

vr50858 295 30 6898 201 44 94 150 48 22 38 06 01 03 02 20

vr50860 68 57 5672 418 57 71 186 109 19 11 02 02 13 03 26

vr50909 238 28 7295 201 42 92 171 48 20 57 13 01 04 02 21

yk3528 91 51 3712 341 53 155 177 49 16 9 01 02 08 02 11

Volatile-rich eclogites

vr67112b 40 84 9242 521 46 30 107 4 40 66 31 26 126 24 125

yk1911 31 39 9428 286 44 82 164 3 23 52 64 18 52 09 46

Pyroxenites

yk1914 111 49 707 207 48 547 44 31 4 3 02 06 19 03 10

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

Low-Mg eclogites

vr19674-g4 11 05 10 22 39 10 29 33 06 12 05 028 24 008

vr43452 04 02 08 13 38 12 41 49 08 05 01 nd 02 na

vr43469 11 05 09 25 44 11 34 36 06 07 02 001 09 129

yk1926 12 05 11 19 30 07 22 24 04 07 03 004 23 090

yk1943 21 08 10 29 42 10 30 32 05 16 05 013 65 017

yk19491 06 03 11 08 14 04 14 18 03 04 02 003 29 115

High-Mg eclogites

vr19677-g2 20 08 12 21 24 05 15 15 03 05 05 007 52 6811

vr434792 05 02 12 05 06 01 04 03 00 03 01 001 98 037

vr67360 06 03 14 06 09 02 06 06 01 02 02 002 86 000

yk1946 18 07 12 21 22 05 13 12 02 11 03 006 56 112

High-Ca eclogites

vr19674-g5 17 09 11 33 43 09 24 23 04 16 05 007 02 011

vr40345 09 04 10 12 22 06 16 14 02 05 nd nd 298 na

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1859

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Continued

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

vr43465 22 09 12 25 26 06 15 14 02 10 04 001 25 040

vr50858 20 08 09 33 41 08 22 21 03 10 01 nd 02 na

vr50860 14 07 11 27 37 08 22 18 03 06 00 nd 06 na

vr50909 20 10 12 33 38 07 20 20 03 16 01 002 02 181

yk3528 08 05 11 20 30 07 18 18 02 05 01 nd 08 na

Volatile-rich eclogites

vr67112b 38 10 07 53 74 15 41 34 04 18 03 012 60 004

yk1911 20 07 08 35 39 07 18 17 02 10 14 nd 73 na

Pyroxenites

yk1914 02 01 11 04 08 02 05 06 01 01 03 003 68 101

aGarnet has CandashMgndashFe relationship typical of high-Mg eclogites but whole-rock trace-element pattern is typical oflow-Mg eclogitesbGarnet has CandashMgndashFe relationship typical of low-Mg eclogites but whole-rock trace-element pattern is typical ofhigh-Mg eclogites

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

(a)

001

01

1

10(c)

001

01

10(b)

VR43452

YK1949

RockNMORB

low-Mg

RockNMORB

high-Mg

(d)

volatile-richhigh-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

phlogopite-bearingYK1911

calcite-bearingVR67112b

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 10 Extended trace-element patterns of reconstructed (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-bearing eclogites Normalized to N-MORB of Sun amp McDonough (1989)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1860

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

RockNMORB

(b)

RockNMORB

(d)

010

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu001

01

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

(a)

001

01

1

10(b)

001

01

1

10(c)

001

01

1

10(e)

high-TSchmidberger et al

low-TSchmidberger et al

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

high-Mglow-Mg

high-Cavolatile-rich

phlogopite-bearingYK1911

calcite-bearingVR67112b

YK1949VR43479

Fig 11 N-MORB normalized REE patterns for (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-richeclogites Shown for comparison in (e) are reconstructed eclogites from Schmidberger et al (2007)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1861

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Both volatile-rich whole rocks display strong negativeSr anomalies Strontium can be sited in calcite or apatiteandor indicate the presence of a Sr-rich mineral that wasnot identified in thin section such as bariteFour of the six reconstructed whole-rocks in the study

of Schmidberger et al (2007) (Fig 11e) show markedLREEHREEN enrichment that is observed only in onehigh-Ca eclogite in our study (Fig 11c) suggesting thatmany of the eclogites they studied were affected by metaso-matic LREE enrichment Two samples in their study thatare not affected by this enrichment show positive Euanomalies similar to the high-Mg eclogites in this studywhereas the distinctive REE pattern of the low-Mg eclo-gites is not recognized among their samples

SR^ND^HF I SOTOPESYSTEMATICSStrontium Nd and Hf isotope data for optically puremineral separates and reconstructed whole-rocks aregiven in Table 5 Spongy rims and alteration are likely tobe mostly removed during leaching in 6N HCl in combina-tion with repeated ultrasonication Despite this rigoroustreatment the reproducibility of Nd in garnet VR43469and in cpx YK1914 is worse than for other samples Thismay be due to inhomogeneities related to late Nd diffusionpossibly from grain boundaries containing a kimberlitecomponent The reproducibility of the 176Hf177Hf data isgenerally poorer than that of Nd which may reflect a com-bination of matrix effects caused by imperfectly purifiedsolutions and sample heterogeneity Parent^daughterratios are available only from in situ analyses which arenot precise enough to allow meaningful model age orinitial isotope ratio calculations (ie at the time of kimber-lite eruption) Because the Lac de Gras kimberlites arerelatively young (55^56 Ma Graham et al 1999) age cor-rection would result in only a minor change in isotopecomposition which is inconsequential in view of the rangeof compositions displayed by most samples and hence hasno effect on the conclusions drawnClinopyroxenes in five volatile-free eclogites and a

pyroxenite have Sr isotope ratios ranging from 0701662 to0703546 this range is similar to or somewhat lower thanthat of the depleted mantle at the time of kimberliteemplacement (0703 Bell amp Tilton 2002) Clinopyroxenein the phlogopitized sample has a distinctly high 87Sr86Srof 0751186Clinopyroxene controls 56^92 of the Nd budget (Nd

concentration in cpx weighted by mode) Neodymiumisotope ratios range from 0511660 to 0514844 Thiscorresponds to eNd variation (per 10 000 deviationof 143Nd144Nd in the sample from 143Nd144Nd in the chon-dritic mantle model reservoir) from 17 to thorn43 cpx inthe low-Mg eclogites have higher eNd than those in

high-Mg eclogites Garnets have generally have higher143Nd144Nd from 0512984 (eNdfrac14thorn 7) to 0520965(eNdfrac14thorn 162) Garnets in the two volatile-rich sampleshave the highest ratios 0522790 for garnet in the calcite-bearing sample VR67112b (eNdfrac14thorn198) and 0525721 inthe phlogopitized eclogiteYK1911 (eNdfrac14thorn 255)Garnet contributes 18^32 and cpx 48^78 of the Hf

budget The proportion controlled by rutile is 4^30(Aulbach et al in preparation) Measured 176Hf177Hf incpx ranges from 0281263 in phlogopite-bearing sampleYK1911 (eHffrac1453) to 0286130 in low-Mg eclogiteVR43469 (eHffrac14thorn119) and in garnet from 0283958 inpyroxeniteYK1915 (eHffrac14thorn42) to 0295718 in low-Mg eclo-gite YK1926 (eHffrac14thorn458) Hafnium in garnet in the phlo-gopite^apatite-bearing sample is extremely radiogenic(176Hf177Hffrac14 0354367) corresponding to an eHf of thorn2532which is close to a value determined for a garnet separatefrom a Roberts Victor eclogite (eHffrac14thorn2561 Jacob et al2005) Garnet in the calcite-bearing sample has extremelyradiogenic Nd and Hf (eNdfrac14thorn193 eHffrac14thorn2516) Garnetin pyroxenite YK1915 also has radiogenic Nd and Hf(eNdfrac14thorn27 eHffrac14thorn42)Despite the large uncertainties for reconstructed whole-

rocks (calculated as the square root of the sum of squareduncertainties of two standard errors for garnet and cpxanalyses combined with 20 propagated uncertainty onthe proportion of Nd or Hf respectively contributed byeach mineral) the isotopic data are consistent in thatthey correlate with the independently determinedparent^daughter ratios (1) Low-Mg eclogites have higherSmNd reflected in the steeper positive slope of theirN-MORB-normalized REE patterns (Fig 11a) comparedwith the high-Mg eclogites (Fig 11b) Accordingly recon-structed low-Mg eclogites have higher present-day eNd

(0 to thorn47) than high-Mg eclogites (4 and 17) (2) Low-Mg eclogite YK1943 has very high Nb and LREE abun-dances that could have been added during metasomatismthat also enhanced its Hf abundance This sample has thelowest eNd and eHf of the low-Mg eclogites (0 and thorn20respectively) In contrast apparently unenriched low-Mgeclogites have more radiogenic Nd and Hf (eNdfrac14thorn47and thorn9 eHffrac14thorn128 and thorn113 respectively)The Sr^Nd isotope diagram reveals that eNd in cpx

varies from 19 to 43 at nearly constant depleted-mantle-like 87Sr86Sr (Fig 12a) In the Nd^Hf isotope diagram(Fig 12b) the reconstructed whole-rocks lie on curvedmixing lines if the lsquomixingrsquo proportions of Nd and Hffor the two lsquoend-memberrsquo minerals are very different(eg YK1911) Garnets have more radiogenic Nd andHf than coexisting cpx All reconstructed volatile-freeeclogites plot in the radiogenic field with regard to Hfor are isotopically similar to the depleted mantle Twohigh-Mg eclogites and one low-Mg eclogite have unradio-genic Nd one high-Mg eclogite has radiogenic Nd and

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1862

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 5 Individual and averaged Sr Nd and Hf isotope ratios obtained from garnet and cpx solutions

Sample 87Sr86Sr 2SE2SD 143Nd144Nd 2SE2SD eNd 2SE2SD 176Hf177Hf 2SE2SD eHf 2SE2SD

Low-Mg eclogites

VR43469 cp 0702661 0000010 0514844 0000006 43 01 0286130 0000012 119 04

VR43469 gt 0515508 0000014 56 03 0286506 0000012 132 04

VR43469 gt r 0515453 0000010 55 02 0286548 0000011 134 04

VR43469 gt r 0515384 0000007 54 01 0287244 0000010 158 04

VR43469 gt r 0515487 0000014 56 03 0286996 0000014 149 05

VR43469avg gt 0515458 0000109 55 21 0286823 0000715 143 253

WR VR43469 0515059 0000273 47 5 0286405 0001079 128 38

YK1926 cp 0703546 0000011 0513043 0000007 8 0 0283154 0000009 13 03

YK1926 gt 0513364 0000008 14 01 0295718 0000005 458 02

WR YK1926 0513120 0000026 9 1 0285980 0000716 113 25

YK1943 cp 0512143 0000007 10 01 0282213 0000005 20 02

YK1943 gt 0514461 0000012 36 02 0287510 0000008 168 03

YK1943 gt r 0514491 0000009 36 02 0287332 0000005 161 02

YK1943 gt r 0514517 0000010 37 02 0287290 0000007 160 02

YK1943avg gt 0514489 0000056 36 11 0287377 0000233 163 82

WR YK1943 0512625 0000228 0 5 0283138 0000879 13 31

High-Mg eclogites

VR43480 gt 0520965 0000017 162 03 0284950 0000010 77 04

VR67360 cp 0283509 0000074 26 26

VR67360 cp r 0283433 0000031 23 11

VR67360avg cp 0701662 0000018 0511660 0000010 19 02 0283471 0000108 25 38

VR67360 gt 0512984 0000101 7 20 0287429 0000180 165 64

WR VR67360 0511761 0000132 17 3 0284734 0000469 69 17

YK1949 cp 0703019 0000013 0512231 0000009 8 02 0282513 0000009 9 03

YK1949 gt 0513511 0000010 17 02 0290297 0000042 266 15

YK1949 gt r 0513508 0000012 17 02 0289947 0000015 254 05

YK1949 gt r 0513552 0000014 18 03

YK1949avg gt 0513523 0000049 17 10 0291569 0000494 311 175

WR YK1949 0512446 0000103 4 2 0284622 0000847 65 30

Volatile-rich eclogites

VR67112b gt 0522790 0000008 198 02 0354367 0000036 2532 13

YK1911 cp 0751186 0000009 0512165 0000007 9 01 0281263 0000009 53 03

YK1911 gt 0525721 0000013 255 03 0288051 0000012 187 04

WR YK1911 0513940 0000342 25 7 0281414 0000018 48 1

Pyroxenites

YK1914 cp 0703015 0000034 0513283 0000041 13 08

YK1914 cp r 0702956 0000005 0513200 0000027 11 05

YK1914 cp r 0513527 0000059 17 11

YK1914 cp r 0513045 0000010 8 02

YK1914avg cp 0702986 0000083 0513264 0000402 12 78

YK1915 gt 0514003 0000011 27 02 0283958 0000021 42

Analyses of full procedural replicates are designated with an lsquorrsquo after sample name SE standard error (internal errorsingle analyses) SD standard deviation (external error averaged multiple analyses) Italics used for reconstructed whole-rocks (mineral isotopic compositions and concentrations weighted by modes) Uncertainties for reconstructed whole-rocksinclude 2SE or 2SD uncertainties on each mineralrsquos isotope composition combined with 20 uncertainty in the amountof Nd or Hf contributed by the constituent minerals eNd and eHf are per 10 000 deviation from model chondritic mantle[model of Wasserburg et al (1981) and Blichert-Toft amp Albarede (1997) respectively]

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1863

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

εHf

εNd

Ekati eclogitesminus100

0

100

200

300

400

500

minus40 0 40 80 120 160 200 240 280

gt in YK1915

gt in VR67112bεNd=198 εHf=2532

minus40

0

40

80

120

160

minus40 minus20 0 20 40 60

minus60

minus40

minus20

0

20

40

070 071 072 073 074 075 076

North Americankimberlite field

εNd

87Sr86Sr

cpx in YK1914

Ekatieclogites

DM

DM

DM

DM

(a)

(b)

pyroxenites

low-Mg

high-Mg

volatile-rich

eclogites

Diavik low-TA154 S kimberlite

cpx in YK1911

Fig 12 (a) Sr^Nd for cpx and (b) Nd-Hf isotope diagram for garnet cpx and reconstructed whole-rocks (symbols with bold outline) Finedashed lines in (b) connect data points from the same sample with garnets having more radiogenic isotopic compositions Uncertainties forreconstructed whole-rocks combine 20 modal uncertainty with uncertainties on each mineralrsquos isotope composition Error bars on individualminerals may be smaller than the symbols Also shown are the isotopic compositions of the depleted mantle (DM DePaolo 1981 Griffin et al2000 Bell amp Tilton 2002) Range of 87Sr86Sr in eclogites from Ekati (Lac de Gras area 100 km north of Diavik Jacob et al 2003) in NorthAmerican kimberlites and compositions of low-temperature eclogites from Diavik and of Lac de Gras A154 South kimberlite fromSchmidberger et al (2007)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1864

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

one low-Mg eclogite has eNd similar to the depleted mantle(Fig 12b)Clinopyroxene in the phlogopitized eclogite YK1911

has unradiogenic Nd and Hf and strongly radiogenicSr whereas the coexisting garnet has radiogenic Nd andHf Although this sample contains 50 vol phlogopiteits whole-rock isotopic composition was calculated

from garnet and cpx alone because the contributionfrom phlogopite to the Hf and Nd budget is 52 and51 respectively (unpublished data) The whole-rockhas radiogenic Nd but has lower eHf than anyother sample analysed in accord with the extreme Hfabundance (53 ppm) and low LuHf of the cpx

DISCUSS IONOrigins of eclogites beneath the SlaveCratonHigh-Mg and high-Ca eclogites

With regard to their major- and trace-element contentsthe high-Mg and high-Ca eclogites from Lac de Gras arebroadly similar to variably evolved oceanic gabbros(Fig 13b and c) which may indicate an origin as subductedoceanic crust Compatible trace-element abundancesare useful to deduce potential crustal protoliths as theseelements are less susceptible to post-formation modification(Jacob 2004) The flat HREE patterns in the high-Mg andhigh-Ca eclogites are compatible with a low-pressureorigin as the oceanic crust is a product of moderate degreesof partial melting mainly within the spinel stability field(eg Presnall et al 2002) whereas melting with residualgarnet would qualitatively be expected to produce stronglyfractionated REE patterns Positive Sr and Pb anomaliesalthough compatible with a plagioclase signature couldalso be due to cryptic metasomatism (Dawson 1984) ofthe eclogites during their residence in the lithosphericmantle These anomalies are probably not related to kim-berlite infiltration as the whole-rock eclogite compositionswere reconstructed from analyses of fresh mineral coresLow ZrSm similar to that in primitive gabbros is

observed for many of the high-Mg and high-Caeclogites (Fig 13b and c) and is independent of the ratioof garnetcpx and whether or not the whole-rock wasreconstructed with rutile (Electronic Appendix 10) Thenegative trend of Mg-number against Yb in garnet(Fig 14) argues against igneous processes involving garnetaccumulation (Taylor amp Neal 1989) as suggested forexample for some high-Mg eclogites from West Africa(Barth et al 2002b) Rather the anti-correlation suggeststhat olivine fractionation played a role in the formation ofthe protoliths Fractionating olivine would excludeYb (Dolivinebasalt

frac14 002 Nikogosian amp Sobolev 1997)and incorporate Mg relative to Fe (Dfrac1415 and 05 respec-tively Taura et al1998) leading to progressive depletion ofMg and enrichment of Yb in the residual melt and henceanti-correlated Mg and Yb in the whole-rock This isinherited after eclogitization All eclogite types (exceptVR43452) plot along the olivine fractionation trend sug-gesting that they formed by a similar processHigh-Mg eclogites have lower

PREE than high-Ca

eclogites and some appear to have positive Eu anomalies

high-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

gabbrosSE Indian ridge

gabbrosOman ophiolite

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

high-Mg

RockNMORB(a)

VR43452

YK1949

low-Mg

boninite-like melts

01

001

1

10

(b)

01

001

1

10

(c)

01

001

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

Fig 13 Trace-element patterns of (a) low-Mg (b) high-Mg and(c) high-Ca eclogites compared with boninite-like volcanic rocksfrom the Isua greenstone belt (Polat et al 2002) gabbros from theOman ophiolite (Benoit et al 1996) and the SE Indian Ridge (Hartet al 1999) normalized to N-MORB of Sun amp McDonough (1989)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1865

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(EuEu up to 14) similar to more primitive cumulates(troctolite gabbros and olivine gabbros) reported in the lit-erature (Fig 13b) Although the sizes of these Eu anomaliesare not outside the analytical uncertainty their correlationwith CeYb in reconstructed eclogites supports their signif-icance (r2frac14066 nfrac14 4) The more primitive nature ofhigh-Mg eclogites is consistent with higher Mg-numbersrelative to high-Ca eclogites and may be due to olivineaccumulation during formation of their protolithsThe two samples with the lowest HREE abundanceshave the highest EuEu which is compatible with varia-tions in plagioclase modesMore evolved protoliths with higher pyroxeneplagio-

clase plus trapped melt could account for small or absentpositive Eu anomalies and higher

PHREE of high-Ca

eclogites (Fig 13c) Although the reason for the frequentassociation of graphite or diamond with high-Ca eclogitesis unclear (Aulbach et al 2003) the frequent occurrence ofkyanite in high-Ca eclogites also observed at Ekati justnorth of Lac de Gras has been argued to support a crustalprotolith (Jacob 2004) Kyanite andor corundum havebeen suggested to exsolve from high-temperature high-Alclinopyroxenes (Caporuscio amp Smyth 1990) Eclogitesfrom Lac de Gras show no exsolution textures suggestiveof such an origin although recrystallization can obscureor destroy microstructural evidence for exsolution(eg Griffin et al 1984)Modelling shows that qualitatively plagioclase accumu-

lation from a low-pressure melt derived from MORBmantle can account for positive Pb Sr and Eu and negativeZr anomalies relative to N-MORB whereas cpx accumula-tion or trapped melt increases Zr Y MREE and HREE

abundances (Fig 15) Some melt addition (for exampleby entrapment) seems to be required to increase all trace-element abundances to those observed for the eclogitesAlthough high-Ca and high-Mg eclogites could representshallow and deep portions of the same oceanic crust thehigher average temperatures obtained for the high-Caeclogites the frequent occurrence of carbon in high-Caeclogites and its absence in high-Mg eclogites the strongLREEHREE fractionation observed for six of seven high-Ca eclogites and the narrow range of compositions of thehigh-Mg eclogites suggest that they may have protolithsthat are not cogenetic Some discrepancies in particularaffecting the incompatible elements may be due to post-crystallization processes acting upon the protoliths suchas dehydration melting and later metasomatism that isevident in peridotites from the same locality (Griffin et al1999)Low K2OTh for high-Ca and high-Mg eclogites

(average 08 Table 4) relative to assumed crustal

0

2

4

6

8

10

30 40 50 60 70 80

Mg

VR43452

Yb

GARNET

pyroxenites

low-Mghigh-Mghigh-Ca

eclogites

VR19673ecl-g2

Fig 14 Garnet Mg-number vs Yb (ppm including 2s uncertainty)Lines qualitatively show separate correlations for eclogites (exceptoutlierVR43452) and pyroxenites (except outlierVR19673ecl-g2)

high-Ca

high-Mg

001

01

1

10

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

melt08 melt02 cryst cpx01 melt09 cryst

plag

(a)

001

01

1

10(b)

RockNMORB

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 15 Reconstructed (a) high-Ca and (b) high-Mg eclogitescompared with compositions of plagioclase and cpx crystallizingfrom a MORB melt modelled using distribution coefficients ofNorman et al (2005)The MORB was calculated as a 10 equilibriummelt of MORB pyrolite [composition from Kelemen et al (1993)except Pb Pr Sr Y and Lu which were calculated as residues from1 fractional melting of pyrolite from McDonough amp Sun (1995)using melt modes of Kinzler (1997) and distribution coefficients ofNorman et al (2005)] Also shown are mixtures of crystals (consistingof 50 plagioclase plus 50 cpx) and melt in the proportions 0901and 0208

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1866

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

precursors (eg gabbros SE Indian Ridge average 22Hart et al 1999) may reflect the preferred loss of fluid-mobile elements during dehydration (Becker et al 2000)although large uncertainties on these low-abundance ele-ments in the constituent minerals and reconstructedwhole-rocks makes this interpretation less than robustThe REE patterns of most high-Mg eclogites are too flatto represent melting residues as partial melting shouldproduce low LREEHREE (Electronic Appendix 10) Incontrast the marked LREE depletion of five of sevenhigh-Ca eclogites (LaLufrac14 02^08) contrasts with that ofassumed evolved gabbroic protoliths (LaLufrac1415^32Hart et al 1999) and can be modelled as a consequence ofpartial melting during subduction after eclogitization (iein the presence of garnet which retains HREE relative toLREE eg Barth et al 2002a) has taken place (Aulbachet al 2003)Our results are compatible with previous studies on

eclogite xenoliths from the Slave Craton includingthose from the same kimberlite that have linked thesesamples or some of these samples to Paleoproterozoicsubduction events (the 195^191Ga formation of theHottah terrane and the 18Ga formation of the GreatBear magmatic arc) in the Slave craton (Griffin et al1999 Kopylova et al 1999 Heaman et al 2002 2006Schmidberger et al 2005 2007)The compositional features discussed above although

compatible with crustal protoliths are not necessarilydefinitive of a subduction origin However if direct crystal-lization from basaltic melts at high pressure (ie mantledepths) is invoked instead specific conditions would haveto be met to satisfy the observations The Mg-number ineclogites from Lac de Gras are lower than those expectedof primary high-pressure mantle melts (eg Walter 1998OrsquoHara amp Herzberg 2002) requiring crystallization athigh pressures after olivine fractionation This originwould further require shielding from peridotitic wallrockwith which the fractionated melts would no longer be inequilibrium (Barth et al 2001) In such a setting the lowrelative abundances of fluid-mobile elements comparedwith those of less fluid-mobile elements and of LREErelative to HREE (Table 4) would reflect an expulsion offluids during crystallization of the protoliths and latermelt mobilization respectively The eclogites clearly donot represent garnet plus cpx cumulates which wouldrequire a positive correlation of Mg-number with Yb thatis at variance with the observed negative correlation ineclogitic garnets from this study (Fig 14)

Low-Mg eclogites

Garnets in low-Mg eclogites compositionally overlap withthose in massive eclogites from Jericho (NW of the Lacde Gras kimberlites) that were suggested to have a high-pressure intrusive origin based on the similarity of theirmajor-element compositions to those of Group A eclogites

fromsouthernAfrica forwhichamantleoriginhasbeenpro-posed (Kopylova et al1999) Garnets in low-Mg eclogites inthe present study have major-element compositions moresimilar to garnets in groupB rather thangroupAeclogitesThe positive slope in N-MORB-normalized HREE

abundances displayed by the low-Mg eclogites is clearlydistinct from other eclogite types and is similar to those ofboninite-like second-stage melts that are closely associatedwith ultramafic volcanics (Smithies et al 2004) (Fig 13a)The range of Mg-numbers in the low-Mg eclogites (55^66)overlaps with that of Archaean second-stage melts fromIsua (58^75) and Abitibi (52^82 Smithies et al 2004)These similarities might indicate that the low-Mg eclogiteshave an origin in a previously depleted mantle source suchas supra-subduction zone mantle Subchondritic ZrHf ofthe low-Mg eclogites contrasts with the supra-chondriticZrHf of the high-Mg and high-Ca eclogites (Aulbach et

al in preparation) and is consistent with previous cpx-controlled depletion in an arc mantle source This sourcecould have been subsequently remelted to generate theboninite-like precursor to the low-Mg eclogites as sug-gested for the protoliths of eclogites from Udachnaya(Jacob amp Foley 1999)Although some second-stage melts to which the low-Mg

eclogites are similar are closely associated with komatiitesand could contain a plume component the radiogenic ordepleted mantle-like Nd and Hf and unradiogenic Sr inthe low-Mg eclogites do not support plume involvement

Trace-element and Sr^Nd^Hf isotopeconstraints on post-formationmetasomatismSeveral low-Mg and high-Ca eclogites show negativeslopes in the LREE which may be indicative of interactionwith small amounts of LREE-enriched melt (Navon ampStolper 1987) amounting to cryptic metasomatism(ie not accompanied by crystallization of new phases) inthe nomenclature of Dawson (1984) High Nb abundancesrelative to elements of similar compatibility displayed bysome low-Mg and high-Ca eclogites (Fig 10) could resultfrom overestimation of rutile modes However high-CaeclogiteVR40345 is rutile-free and has been reconstructedwithout rutile its high Nb abundance is accompanied byLREE enrichment indicating that metasomatic additionmay be responsibleLow-Mg eclogites show a positive correlation of La with

Nb (Electronic Appendix 11a) indicating concomitantaddition of LREE and Nb This suggests that the metaso-matic agent was a solute-rich fluid or silicate silicic melt ashydrous fluids and carbonatites do not carry enough HFSEto affect the budget of eclogites (Green amp Wallace 1988Klemme et al 1995 Adam et al 1997 Stalder et al 1998)Of the samples that are enriched in Nb all but one (high-Ca eclogiteVR50909) show a simultaneous enrichment in

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1867

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Hf (Electronic Appendix 11b) Together these trace-element relationships suggest that both SmNd and LuHfwere lowered in these samples Their low parent^daughterratios contrast with higher ratios in unenriched samplesand resulted in the ingrowth of a range of Nd and Hf iso-tope compositions Compared with the low-Mg and high-Ca eclogites the high-Mg eclogites have a narrow range ofLa and Nb concentrations which correlates with a narrowrange in Nd and Hf isotope compositions (Fig 12)Precise model ages cannot be calculated because parent^

daughter ratios are available only from in situ analyseswhich have relatively large uncertainties Howeverevidence for long-term enrichment and depletion may bediscerned Low-Mg eclogites even those that are enrichedin LREE have relatively high SmNd as a result of havingoverall positive slopes in their REE patterns relative toMORB whereas the high-Mg eclogites have relativelyflat REE patterns and therefore lower SmNd (Fig 11)This translates into lower eNd for high-Mg eclogites thanfor low-Mg eclogites In contrast they show no systematicdifference with respect to 176Hf177Hf Because the low- andhigh-Mg eclogites are proposed to have different origins asdiscussed above and therefore did not necessarily have thesame initial isotopic composition and formation age theseisotopic differences cannot be used to estimate their agesA long-term coupling of trace-element and isotopic

abundances is also indicated by the fact that low-Mg eclo-gite YK1943 which has experienced Nb and LREE addi-tion has lower eNd (0) and eHf (thorn13) than low-Mg eclogiteVR43469 with higher LuHf and SmNd (eNdfrac14thorn47eHffrac14thorn128) If low-Mg eclogites YK1943 and VR43469originally had identical initial Nd and Hf isotope composi-tions and if the parent^daughter ratios are accurate andwere modified only once during their evolution shortlyafter metamorphism (when their isotopic compositionwould have been reset) this overprint can be calculatedfor both isotopic systems to have occurred at 17GaThis would suggest that Hf and Nd were added duringthe same event In contrast high-Mg eclogites VR67360and YK1949 would have had identical Ndisotope compositions at 10Ga whereas their Hf isotopecompositions are disturbed in that the sample with thehigher parent^daughter ratio has a lower 176Hf177Hf thanthe sample with the lower parent^daughter ratio Theseresults suggest that the LREE and HFSE enrichmentobserved in different samples occurred during separateevents that did not affect the eclogites equallyExtreme HFSE enrichment leading to the growth of

zircon and rutile and LREE addition leading to thegrowth of apatite in eclogites from the Jericho kimberlitein the northern Slave craton have been ascribed to twoseparate metasomatic events at around 18Ga (quasi-coincident with the age of formation and metamorphism)and 13^10Ga respectively (Heaman et al 2002 2006

Schmidberger et al 2005) The authors associated theseevents with Paleoproterozoic subduction and aMesoproterozoic thermal disturbance in the Slave litho-spheric mantle possibly linked to the Mackenzie igneousevent Mesoproterozoic Nd model ages mostly between13 and 14Ga were reported for eclogites from thenearby Ekati kimberlites (Jacob et al 2003) Although wedo not observe zircon or apatite in volatile-free eclogitesfrom Lac de Gras and the age constraints obtained fromour data are weak they support a link between theseevents described for other eclogite populations and themetasomatic enrichment observed in our studyThe Sr isotope data in the present study are similar to

those of Schmidberger et al (2007) who suggested that thedepleted-mantle-like values are consistent with derivationof the protoliths from an oceanic mantle source However87Sr86Sr in volatile-free eclogites in the present study isremarkably constant considering their Nd and Hf isotopevariability (Fig 12) and the variability of trace-elementpatterns (Fig 10) Given the evidence for dehydration andmelting in many of the eclogites fractionation of highlyincompatible Rb from Sr and differential ingrowth ofradiogenic Sr would be expectedThis may indicate a rela-tively recent overprint and considering the similarityof 87Sr86Sr to present-day depleted mantle which is signif-icantly lower than 87Sr86Sr of North American kimber-lites a juvenile (asthenospheric) source

PyroxenitesPyroxenites from Lac de Gras often show relationshipsin their major-element contents that are different fromthose observed in the eclogites For example pyroxeniticgarnets show a negative correlation between CaO andNa2O contents whereas these oxides are positively corre-lated for garnets in eclogites (Fig 5) CaO is well correlatedwith C2O3 in pyroxenitic garnets but not in eclogitic gar-nets (Fig 4) Mg-number is positively correlated withYb abundance for pyroxenitic garnets but anti-correlatedfor eclogitic garnets (Fig 14) This suggests that the pyrox-enites in the present study are not related to the eclogitesby fractionation or accumulation processes as proposedfor the origin of some pyroxenites (Foley et al 2003)Pyroxenites have also been interpreted as the products

of crystal segregation in magma channels with alteredoceanic crust as the magma source this would explaintheir Eu anomalies and stable and radiogenic isotopecompositions (Pearson et al 1993 Becker 1996) As thehigh-pressure liquidus phases of most terrestrial basaltsare cpx and garnet (OrsquoHara amp Herzberg 2002) partialmelts of eclogites with basaltic composition wouldcrystallize cpx and garnet but not opx More than half ofthe pyroxenites from Lac de Gras contain opx suggestingthat they probably did not crystallize directly fromeclogite-derived melts

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1868

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Low-Cr websterites and pyroxenites could be crystalli-zation products of mafic melts that interacted with perido-titic mantle thus becoming more refractory The positivecorrelation of Yb and Mg-number in garnet would allowan origin by fractional crystallization from such meltsAlternatively they formed by reaction of silicic (possiblyeclogite-derived) melts with mantle peridotites(eg Kelemen et al 1993 Aulbach et al 2002) This agreeswith the similarity of hypothetical melts in equilibriumwith pyroxenitic garnets in most samples to eclogite-derived melts (Electronic Appendix 12) Niobium concen-trations in all hypothetical melts are much higher than ineclogite-derived melts regardless of whether or not rutileis assumed to be in the residue The same is true for LaCe and Zr concentrations in some samples which mayindicate that NbZr and LREE were added to thepyroxenites some time after their formation as discussedin the previous section for some eclogites Because mini-mum ages for pyroxenitic rutile with an unradiogenicHf isotope composition (calculated by projecting the176Hf177Hf to the model depleted mantle isotopic evolutioncurve assuming zero ingrowth) range to 178066Ga(Aulbach et al in preparation) it appears likely that thepyroxenites were not formed recently (eg during kimber-lite-related or precursory melt migration) and that some ofthe samples are at least MesoproterozoicGarnet pyroxeniteYK1952 consists of schlieren of equili-

brated garnet pyroxenite in a lsquomatrixrsquo of spongy cpxfine cpx^opx intergrowths embayed garnet and meltpatches (Fig 2f) Garnet in the pyroxenite schlieren hasvery low LREE and HFSE abundances Orthopyroxenein the matrix has distinctly higher Mg-number andCr-number and contains more CaO than opx from thepyroxenite schlieren whereas opx rims in the schlierenare compositionally similar to opx in the matrix The con-siderable compositional and textural disequilibrium inthis sample suggests that this is a young feature causedby reaction with fluids shortly before or at the time ofentrainmentAs a consequence of the small sample sizes of most of

the pyroxenites studied here a comprehensive isotopicdataset is not available Clearly more such data areneeded in addition to stable isotope data to obtain betterconstraints on the origins of this rock type Consideringthe microstructural and compositional variability ofthe pyroxenites and major-element relationships oppositeto those observed for eclogites it appears certain that pyr-oxenites form by a variety of processes some of which areunrelated to the formation of eclogites and their protoliths

Volatile-rich eclogitesVolatile-rich eclogites from Lac de Gras display the mostextreme Hf^Nd isotope systematics of all the analysedsamples Garnets in phlogopitized (YK1911) and in cal-cite-bearing eclogite (VR67112b) have the most radiogenic

Nd (eNdfrac14 2486 and 1927 respectively) Garnet inVR67112b has extremely high eHf (2516) Clinopyroxene inYK1911 has the lowest eHf measured for any mineralseparate from Lac de Gras (53) in accord with the veryhigh Hf abundance in its cpx (5 ppm) In contrastgarnet in this sample has radiogenic Hf (eHffrac14 136) Thisdiscrepancy is at least in part due to their relatively lowequilibration temperatures (7698C and 8558C respec-tively) which allowed the isotopic compositions of cpxand garnet to evolve separately at their low andhigh parent^daughter element ratios respectively forextended periods of time and possibly since formationAccordingly the cpx in YK1911 gives a minimum Hf ageof 27GaSlab-derived fluids could have been introduced during

the 27Ga collision of the western Slave Craton domainincluding the thick plume-derived deep lithosphericmantle layer and the eastern domain with its highlydepleted lithosphere This would also be consistent withthe observation that volatile-rich eclogites occur at shallowdepths in the mantle (TKroghfrac14 685^8558C correspondingto depths of 85^115 km along a 40 mWm2 geotherm)where the bulk of dehydration and production of solute-rich melts in a subducting slab might be expected to occur(Becker et al 2000 Hermann et al 2006) Their negativeEu anomalies might indicate that the fluid was derivedfrom the shallow portion of the slab that experiencedplagioclase fractionation The presence of carbonate inVR67112b requires a carbonated sourceHigh 87Sr86Sr in cpx in YK1911 is probably the conse-

quence of isotopic equilibration with high-RbSr phlogo-pite after long-term radiogenic ingrowth while Hfaddition led to retarded ingrowth of radiogenic HfThese systematics combined with radiogenic Nd appear

inconsistent with the involvement of sediments in the pet-rogenesis of this sample (eg Perfit et al 1980 Vervoortet al 1999) or a hydrous fluid such as those responsible forMARID metasomatism in the Kaapvaal craton (Konzettet al 1998)Whatever the origin of the fluids the apparentHf mobility suggests that they were not in equilibriumwith a rutile-bearing source and also that they weresolute-rich in nature (ie hydrous melts able to dissolveHFSE) and the relatively high HREE abundances of thevolatile-rich eclogites indicate a garnet-free source suchas amphibolite This allowed evolution of the sample atrelatively high time-integrated SmNd

SUMMARY AND CONCLUSIONS

(1) The Lac de Gras kimberlites carried abundant eclo-gite and pyroxenite xenoliths to the surface Threemain eclogite types are distinguished in terms ofmajor-element composition and trace element

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1869

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

patterns high-Mg high-Ca and low-Mg eclogitesThe garnets of all eclogite types have anti-correlatedMg-number and Yb pointing to olivine fractionationas a controlling factor in the formation of their pre-cursor rocks Their flat or positively sloped HREEpatterns qualitatively indicate that the melt parentalto the precursor rocks was not in equilibrium with agarnet-bearing residue

(2) Reconstructed high-Mg eclogites and high-Caeclogites have flat HREE patterns positive Pb andSr anomalies and low ZrSm similar to oceanic gab-bros reported in the literature High-Mg eclogiteshave variably low HREE contents and some havesmall positive Eu anomalies possibly suggestive ofmore primitive relatively plagioclase-rich protolithswhereas high-Ca eclogites have higher HREE abun-dances and small or absent Eu anomalies that requiremore cpx- andor melt-rich protoliths if a subductionorigin is assumed

(3) Low-Mg eclogites have distinctive trace-element pat-terns that show a marked positive slope in the HREEand subchondritic ZrHf They resemble someArchaean second-stage melts (lsquoboninitesrsquo) and theirprotoliths appear to have formed from a previouslydepleted source

(4) After formation the eclogites were subject to a varietyof processes Most have low ratios of fluid-mobile vsless fluid-mobile elements possibly as a result of dehy-dration of the protoliths In addition some of the low-Mg and high-Ca eclogites have low LREEHREEconsistent with partial melt lossWhereas the low-Mgeclogites show concomitant enrichment in La Nb andZr some high-Mg and high-Ca eclogites show evi-dence for variable and unsystematic (decoupled)LREE and Nb addition

(5) Isotopically the trace-element characteristics of thedifferent eclogite types translate into lower eNd forhigh-Mg eclogites (negative slope in MREEfrac14 lowerSmNd) than for low-Mg eclogites (positive slope inMREEfrac14higher SmNd) Within the low-Mg groupsamples that show evidence for metasomatism (lowLuHf and SmNd) have lower eNd and eHf than asample that was apparently not metasomatized BothNd and Hf addition can be dated to 17Ga if identi-cal initial isotope compsitions are assumed for thelow-Mg eclogites In contrast the relatively constantand depleted mantle-like 87Sr86Sr of eclogitic cpxdespite variable Nd and Hf isotope compositions andparent^daughter ratios suggests recent input of juve-nile material

(6) Garnet and cpx in pyroxenites show major-elementrelationships different from those in eclogitesTherefore these two rock types are probably notrelated by fractionation processes The presence of

opx precludes direct crystallization from slab-derivedmelts Pyroxenites may be the product of (slab-derived) melts hybridized during assimilation of peri-dotitic mantle

(7) Garnets in phlogopite- and carbonate-bearing eclo-gites have extreme isotopic compositions includingthe most radiogenic Nd and Hf in the datasetNegative Eu anomalies may point to the involvementof a crustal protolith Highly radiogenic Sr radiogenicNd and unradiogenic Hf in the reconstructedphlogopite-bearing sample combined with highHREE abundances and evidence for Hf mobilitymay indicate involvement of solute-rich fluids origi-nating in a garnet- and rutile-free source

ACKNOWLEDGEMENTSHelp in the clean laboratories and with the analyticalfacilities by S Elhlou C Lawson A Sharma P Wielandand M Zhang is greatly appreciated Bill Griffin isthanked for his support and for agreeing to reviewseveral versions of the manuscript despite disagreeingwith many of the conclusions presented here The paperbenefited greatly from the detailed and constructivereviews provided by the editors Gareth Davies andMarjorie Wilson and by Matthias Barth Larry Heamanand an anonymous reviewer This work was funded by aMacquarie University International PostgraduateAward and Postgraduate Research Fund (SA) by anARC SPIRT grant sponsored by Kennecott Canada Incand by ARC Large and Discovery Grants to SYOrsquoRAnalytical data were obtained using instrumentationfunded by ARC LIEF and DEST Systemic InfrastructureGrants industry partners and Macquarie UniversityThis is Publication 481 from the ARC National KeyCentre for Geochemical Evolution and Metallogeny ofContinents (wwwesmqeduauGEMOC)

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online

REFERENCESAdam J GreenT H Sie S H amp Ryan C G (1997) Trace element

partitioning between aqueous fluids silicate melts and mineralsEuropeanJournal of Mineralogy 9 569^584

Aulbach S Stachel T Viljoen K S Brey G P amp Harris J W(2002) Eclogitic and websteritic diamond sources beneath theLimpopo Beltccedilis slab-melting the link Contributions to Mineralogy

and Petrology 143 56^70Aulbach S GriffinW L Pearson N J OrsquoReilly S Y Kivi K amp

Doyle B (2003) Origins of eclogites beneath the central SlaveCraton Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA-0011

Aulbach S Griffin W L Pearson N J OrsquoReilly S Y amp Kivi K(2004a) Mantle formation and evolution Slave Craton constraints

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1870

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

from HSE abundances and Re^Os isotope systematics of sulfideinclusions in mantle xenocrysts Chemical Geology 208 61^88

Aulbach S Griffin W L OrsquoReilly S Y amp McCandless T E(2004b) Genesis and evolution of the lithospheric mantle beneaththe Buffalo HeadTerrane Alberta (Canada) Lithos 77 413^451

Aulbach S GriffinW L Pearson N J OrsquoReilly S Y amp Doyle B(2007) Lithosphere formation in the central Slave Craton(Canada) plume subcretion or lithosphere accretion Contributionsto Mineralogy and Petrology (in press) doi 101007s00410-007-0200-1

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2001)Geochemistry of xenolithic eclogites from West Africa Part IA link between low MgO eclogites and Archean crust formationGeochimica et Cosmochimica Acta 65 1499^1527

Barth M G Foley S F amp Horn I (2002a) Partial melting inArchean subduction zones constraints from experimentally deter-mined trace element partition coefficients between eclogitic miner-als and tonalitic melts under upper mantle conditions PrecambrianResearch 113 323^340

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2002b)Geochemistry of xenolithic eclogites from West Africa part 2Origins of the high MgO eclogites Geochimica et Cosmochimica Acta

66 4325^4345Beard B L Fraracci K NTaylor L A Snyder G A Clayton R

A Mayeda T K amp Sobolev NV (1996) Petrography andgeochemistry of eclogites from the Mir kimberliteYakutia RussiaContributions to Mineralogy and Petrology 125 293^310

Becker H (1996) Geochemistry of garnet peridotite massifs fromlower Austria and the composition of deep lithosphere beneatha Palaeozoic convergent plate margin Chemical Geology 134 49^65

Becker H Jochum K P amp Carlson R W (2000) Trace elementfractionation during dehydration of eclogites from high-pressureterranes and the implications for element fluxes in subductionzones Chemical Geology 163 65^99

Bell K amp Tilton G R (2002) Probing the mantle The story fromcarbonatites EOS Transactions American Geophysical Union 83275^277

Benoit M Polvecurren M amp Ceuleneer G (1996) Trace element andisotopic characterisation of mafic cumulates in a fossil mantlediapir (Oman ophiolite) Chemical Geology 134 199^214

BleekerW (2003) The late Archean record a puzzle in ca 35 piecesLithos 71 99^134

BleekerW Ketchum JW F JacksonV A amp Villeneuve M (1999)The central Slave Basement Complex Part I Its structural topol-ogy and autochthonous core Canadian Journal of Earth Sciences 361083^1109

Blichert-Toft J amp Albare de F (1997) The Lu^Hf isotope geochemis-try of chondrites and the evolution of the mantle^crust systemEarth and Planetary Science Letters 148 243^258

Blichert-Toft J Chauvel C amp Albare de F (1997) Separation of Hfand Lu for high precision isotope analysis of rock samples by mag-netic sector-multiple collector ICP-MS Contributions to Mineralogy

and Petrology 127 248^260Brey G P amp Kolaquo hler T (1990) Geothermobarometry in four-phase

lherzolites II New thermobarometers and practical assessment ofexisting thermobarometers Journal of Petrology 31 1353^1378

Brey G P Kolaquo hlerT amp Nickel K G (1990) Geothermobarometry infour-phase lherzolites I Experimental results from 10 to 60 kbJournal of Petrology 31 1313^1352

Caporuscio F amp Smyth J (1990) Trace element crystal chemistryof mantle eclogites Contributions to Mineralogy and Petrology 105550^561

Carbno G B amp Canil D (2002) Mantle structure beneath the SWSlave Craton Canada constraints from garnet geochemistry in theDrybones Bay Kimberlite Journal of Petrology 43 129^142

Coleman R G Lee E D Beatty L B amp Brannock W W (1965)Eclogites and eclogites their differences and similaritiesGeological Society of America Bulletin 76 483^508

Creaser R A Grulaquo tter H Carlson J amp Crawford B (2004)Macrocrystal phlogopite Rb^Sr dates for the Ekati propertykimberlites Slave Province Canada evidence for multiple intru-sive episodes in the Paleocene and Eocene Lithos 76 399^414

Davies R M GriffinW L Pearson N J Andrew A S Doyle BJ amp OrsquoReilly S Y (1999) Diamonds from the deep Pipe DO-27Slave Craton Canada In Gurney J J Gurney J L Pascoe MD amp Richardson S H (eds) Proceedings of the 7th International

Kimberlite Conference CapeTown Red Roof Design pp 148^155Davies R M Griffin W L OrsquoReilly S Y amp Doyle B J (2004)

Mineral inclusions and geochemical characteristics of microdia-monds from the DO27 A154 A21 A418 DO18 DD17 and RanchLake kimberlites at Lac de Gras Slave Craton Canada Lithos 7739^55

DavisW J amp Hegner E (1992) Neodymium isotopic evidence for thetectonic assembly of Late Archean crust in the Slave Provincenorthwest Canada Contributions to Mineralogy and Petrology 111493^504

Davis W J Jones A G Bleeker W amp Grutter H (2003)Lithosphere development in the Slave craton a linked crustal andmantle perspective Lithos 71 575^589

Dawson J B (1984) Contrasting types of upper-mantle metasoma-tism In Kornprobst J (ed) Kimberlites II The Mantle and

Crust^Mantle Relationships Amsterdam Elsevier pp 289^294DePaolo D J (1981) Neodymium isotopes in the Colorado Front

Range and crust^mantle evolution in the Proterozoic Nature 291193^196

Erlank A J amp Kushiro I (1970) Potassium contents of syntheticpyroxenes at high temperatures and pressures Carnegie Institution ofWashingtonYearbook 68 433^439

Foley S F Barth M G amp Jenner G A (2000) Rutilemelt partitioncoefficients for trace elements and an assessment of the influence ofrutile on the trace element characteristics of subduction zonemagmas Geochimica et Cosmochimica Acta 64 933^938

Foley S F Buhre S amp Jacob D E (2003) Evolution of the Archaeancrust by delamination and shallow subduction Nature 421 249^252

Graham I Burgess J L Bryan D Ravenscroft P J Thomas EDoyle B J Hopkins R amp Armstrong K A (1999) Explorationhistory and geology of the Diavik Kimberlites Lac de GrasNorthwest Territories Canada In Gurney J J Gurney J LPascoe M D amp Richardson S H (eds) Proceedings of the 7th

International Kimberlite Conference Cape Town Red Roof Designpp 262^279

Green D H amp Wallace M E (1988) Mantle metasomatism byephemeral carbonatite melts Nature 336 459^462

Green T H Blundy J D Adam J amp Yaxley G M (2000) SIMSdetermination of trace element partition coefficients betweengarnet clinopyroxene and hydrous basaltic liquids at 2^75 GPaand 1080^12008C Lithos 53 165^187

GriffinW L amp OrsquoReilly SY (2007) Cratonic lithospheric mantle Isanything subducted Episodes 30(1) 43^53

Griffin W L Wass S Y amp Hollis J D (1984) Ultramafic xenolithsfrom Bullenmerri and Gnotuk maars Victoria Australia petrol-ogy of a sub-continental crust^mantle transition Journal of

Petrology 25 53^87GriffinW L Doyle B J Ryan C G Pearson N J OrsquoReilly SY

Davies R Kivi K van Achterbergh E amp Natapov LM (1999)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1871

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Layered mantle lithosphere in the Lac de Gras Area SlaveCraton composition structure and origin Journal of Petrology 40705^727

Griffin W L Pearson N J Belousova E Jackson S E vanAchterbergh E OrsquoReilly SY amp Shee S R (2000)The Hf isotopecomposition of cratonic mantle LAM-MC-ICPMS analysis ofzircon megacrysts in kimberlites Geochimica et Cosmochimica Acta

64 133^147Griffin W L OrsquoReilly S Y Doyle B J Pearson N J

Coopersmith H Kivi K Malkovets V amp Pokhilenko N(2004) Lithosphere mapping beneath the North American plateLithos 77 873^922

Grulaquo tter H S Apter D B amp Kong J (1999) Crust^mantle couplingevidence from mantle-derived xenocrystic garnets In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 307^312

Hart S R Blusztajn J Dick H J B Meyer P S ampMuehlenbachs K (1999) The fingerprint of seawater circulationin a 500-meter section of ocean crust gabbros Geochimica et

Cosmochimica Acta 63 4059^4080Harte B amp Kirkley M B (1997) Partitioning of trace elements

between clinopyroxene and garnet Data from mantle eclogitesChemical Geology 136 1^24

Heaman L M Creaser R A amp Cookenboo H O (2002) Extremeenrichment of high field strength elements in Jericho eclogitexenoliths A cryptic record of Paleoproterozoic subduction partialmelting and metasomatism beneath the Slave craton CanadaGeology 30 507^510

Heaman L M Kjarsgaard B A amp Creaser R A (2004) The tem-poral evolution of North American kimberlites Lithos 76 377^397

Heaman L M Creaser R A Cookenboo H O amp Chacko T(2006) Multi-stage modification of the northern Slave mantlelithosphere Evidence from zircon- and diamond-bearing eclogitexenoliths entrained in Jericho Kimberlite Journal of Petrology 47821^858

Hermann J Cpandler C Hack A amp Korsakov A (2006) Aqueousfluids and hydrous melts in high-pressure and ultra-high pressurerocks implications for element transfer in subduction zones Lithos92 399^417

Hills DV amp Haggerty S E (1989) Petrochemistry of eclogites fromthe Koidu Kimberlite Complex Sierra Leone Contributions to

Mineralogy and Petrology 103 397^422Hoffman P F (1989) Precambrian geology and tectonic history of

North America In Bally A amp Palmer A (eds) The Geology of

North AmericaccedilAn Overview Boulder CO Geological Society ofAmerica pp 447^512

IrelandT R Rudnick R L amp Spetsius Z (1994) Trace-elements indiamond inclusions from eclogites reveal link to Archean granitesEarth and Planetary Science Letters 128 199^213

Irvine G J Pearson D G Kjarsgaard B A Carlson R WKopylova M G amp Dreibus G (2003) Evolution of the litho-spheric mantle beneath Northern Canada a Re^Os isotope andPGE study of kimberlite-derived peridotite xenoliths fromSomerset Island and a comparison to the Slave and Kaapvaalcratons Lithos 71 461^488

Isachsen C E amp Bowring S A (1994) Evolution of the Slave cratonGeology 22 917^920

Jacob D E (2004) Nature and origin of eclogite xenoliths from kim-berlites Lithos 77 295^316

Jacob D E amp Foley S F (1999) Evidence for Archean ocean crustwith low high field strength element signature from diamondifer-ous eclogite xenoliths Lithos 48 317^336

Jacob D Jagoutz E Lowry D Mattey D amp Kudrjavtseva G(1994) Diamondiferous eclogites from Siberiaccedilremnants ofArchean oceanic-crust Geochimica et Cosmochimica Acta 585191^5207

Jacob D E Fung A Jagoutz E amp Pearson D G (2003) Petrologyand geochemistry of eclogite xenoliths from the Ekati kimberlitesarea Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA_239

Jacob D E Bizimis M amp Salters V J M (2005) Lu^Hf and geo-chemical systematics of recycled ancient oceanic crust evidencefrom Roberts Victor eclogites Contributions to Mineralogy and

Petrology 148 707^720Jerde E A Taylor L A Crozaz G Sobolev NV amp SobolevV N

(1993) Diamondiferous eclogites from Yakutia Siberiaccedilevidencefor a diversity of protoliths Contributions to Mineralogy and Petrology

114 189^202Jones A G Ferguson I J Chave A D Evans R L amp

McNeice GW (2001) The electric lithosphere of the Slave cratonGeology 29 423^426

Kelemen P B Shimizu N amp Dunn T (1993) Relative depletionof niobium in some arc magmas and the continental-crust-partitioning of K Nb La and Ce during meltrockreaction in the upper-mantle Earth and Planetary Science Letters 120111^134

Kinzler R J (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition Application to mid-ocean ridge basalt petrogenesis Journal of Geophysical Research 102853^874

Klemme S van der Laan S R Foley S F amp Guenther D (1995)Experimentally determined trace and minor element partitioningbetween clinopyroxene and carbonatite melt under upper mantleconditions Earth and Planetary Science Letters 133 439^448

Konzett J Armstrong R A Sweeney R J amp CompstonW (1998)The timing of MARID metasomatism in the Kaapvaal mantle anion probe study of zircons from MARID xenoliths Earth and

Planetary Science Letters 160 133^145Kopylova M G Russell J K amp Cookenboo H (1999) Mapping the

lithopshere beneath the north central Slave Craton In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 468^479

Krogh E (1988) The garnet^clinopyroxene iron^magnesiumgeothermometerccedila reinterpretation of existing experimentaldata Contributions to Mineralogy and Petrology 99 44^48

Kusky T M (1989) Accretion of Archean Slave Province Geology 1763^67

LeCheminant A N amp Heaman L M (1989) Mackenzie igneousevents Canada Middle Proterozoic hotspot magmatismassociated with ocean opening Earth and Planetary Science Letters

96 38^48LeCheminant A N Richardson D G DiLabio R N W amp

Richardson K A (1996) La recherche de diamants au CanadaGeological Survey of Canada Open File 3228 268pp

MacGregor I D amp Carter J L (1970) The chemistry of clinopyrox-enes and garnets of eclogite and peridotite xenoliths from theRoberts Victor mine South Africa Physics of the Earth and PlanetaryInteriors 3 391^397

MacGregor I D amp Manton W I (1986) Roberts Victor eclogitesancient oceanic crust Journal of Geophysical Research 91 14063^14079

MacKenzie J M amp Canil D (1999) Composition and thermal evo-lution of cratonic mantle beneath the central Archean SlaveProvince NWT Canada Contributions to Mineralogy and Petrology

134 313^324

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1872

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

McCandlessT E amp Gurney J J (1989) Sodium in garnet and potas-sum in clinopyroxene criteria for classifying mantle eclogites InRoss J (ed) Kimberlites and Related Rocks Carlton Blackwellpp 827^832

McDonoughW F amp Sun S-S (1995) The composition of the EarthChemical Geology 120 223^253

Navon O amp Stolper E (1987) Geochemical consequences of meltpercolation the upper mantle as a chromatographic columnJournal of Geology 95 285^307

Nikogosian I K amp Sobolev AV (1997) Ion-microprobe analysis ofmelt inclusions in olivine experience in estimating the olivine^meltpartition coefficients of trace elements Geochemistry International 35119^126

Norman M D Griffin W L Pearson N J Garcia M O ampOrsquoReilly S Y (1998) Quantitative analysis of trace element abun-dances in glasses and minerals a comparison of laser ablationinductively coupled plasma mass spectrometry solution inductivelycoupled plasma mass spectrometry proton microprobe and elec-tron microprobe data Journal of Analytical Atomic Spectrometry 13477^482

Norman M D Garcia M O amp Pietruszka A J (2005) Trace-element distribution coefficients for pyroxenes plagioclase andolivine in evolved tholeiites from the 1955 eruption of KilaueaVolcano Hawairsquoi and petrogenesis of differentiated rift-zonelavas American Mineralogist 90 888^899

OrsquoHara M J amp Herzberg C (2002) Interpretation of trace elementand isotope features of basalts Relevance of field relations petrol-ogy major element data phase equilibria and magma chambermodeling in basalt petrogenesis Geochimica et Cosmochimica Acta 662167^2191

PadghamW A amp FysonW K (1992) The Slave Province a distinctcraton CanadianJournal of Earth Sciences 29 2072^2086

Pearson D G Davies G R amp Nixon P H (1993) Geochemical con-straints on the petrogenesis of diamond facies pyroxenites from theBeni Bousera peridotite massif North Morocco Journal of Petrology34 125^172

Pearson N J Griffin W L Doyle B J OrsquoReilly S Y vanAchterbergh E amp Kivi K (1999) Xenoliths from kimberlitepipes of the Lac de gras area Slave Craton Canada In GurneyJ J Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite ConferenceCapeTown Red Roof Design pp 644^658

Perfit N R Gust D A Bence A E Arculus R J amp Taylor S R(1980) Chemical characteristics of island-arc basalts implicationsfor mantle sources Chemical Geology 30 227^256

Polat A Hofmann A W amp Rosing M T (2002) Boninite-likevolcanic rocks in the 37^38 Ga Isua greenstone belt WestGreenland geochemical evidence for intra-oceanic subductionzone processes in the early Earth Chemical Geology 184 231^254

Poudjom DjomaniY H GriffinW L OrsquoReilly SY amp Doyle B J(2005) Lithospheric domains and controls on kimberlite emplace-ment Slave Province Canada evidence from elastic thickness andupper mantle composition Geochemistry Geophysics Geosystems 6doi1010292005GC0009978

Presnall D C Gudfinnsson G H amp Walter M J (2002)Generation of mid-ocean ridge basalts at pressures from 1 to 7GPa Geochimica et Cosmochimica Acta 66 2073^2090

Schmidberger S S Heaman L M Simonetti A Creaser R A ampCookenboo H O (2005) Formation of Paleoproterozoiceclogitic mantle Slave Province (Canada) Insights from in-situHf and U^Pb isotopic analyses of mantle zircons Earth and

Planetary Science Letters 240 621^633

Schmidberger S S Simonetti A Heaman L M Creaser R A ampWhiteford S (2007) Lu^Hf in-situ Sr and Pb isotope and traceelement systematics for mantle eclogites from the Diavik diamondmine Evidence for Paleoproterozoic subduction beneath the Slavecraton Canada Earth and Planetary Science Letters 254 55^68

Schulze D J amp Helmstaedt H (1988) Coesite-sanidine eclogitesfrom kimberlite products of mantle fractionation or subductionJournal of Geology 96 435^443

Smithies R H Champion D C amp Sun S S (2004) The case forArchaean boninites Contributions to Mineralogy and Petrology 147705^721

Smyth J R Caporuscio F A amp McCormick T C (1989) Mantleeclogites evidence of igneous fractionation in the mantle Earth andPlanetary Science Letters 93 133^141

Sobolev NV amp Lavrentrsquoyev Y G (1971) Isomorphic sodium admix-ture in garnets formed at high pressure Contributions to Mineralogy

and Petrology 31 1^12Stalder R Foley S F Brey G P amp Horn I (1998) Mineral aqueous

fluid partitioning of trace elements at 900^12008C and 30^57GPa New experimental data for garnet clinopyroxene and rutileand implications for mantle metasomatism Geochimica et

Cosmochimica Acta 62 1781^1801Sun S-S amp McDonough W F (1989) Chemical and isotopic sys-

tematics of oceanic basalts implications for mantle compositionsand processes In Saunders A D amp Norris M J (eds)Magmatism in the Ocean Basins Geological Society London Special

Publications 42 313^345Taura H Yurimoto H Kurita K amp Sueno S (1998) Pressure

dependence on partition coefficients for trace elements between oli-vine and the coexisting melts Physics and Chemistry of Minerals 25469^484

Taylor L A amp Neal C R (1989) Eclogites with oceanic crustal andmantle signatures from the Bellsbank kimberlite South AfricaPart I Mineralogy petrography and whole rock chemistryJournal of Geology 97 551^567

van Achterbergh E Ryan C G amp Griffin W L (1999) Glitter online intensity reduction for the laser ablation inductively coupledplasma spectrometry Abstract 9th V M GoldschmidtConference LPI Boston MA 905 pp Abstract no 7215

van Breemen O DavisW J amp King J E (1992) Temporal distribu-tion of granitoid plutonic rocks in the Archean Slave ProvinceNorthwest Canadian Shield Canadian Journal of Earth Sciences 292186^2199

Vervoort J D Patchett P J Blichert-Toft J amp Albare de F (1999)Relationship between Lu^Hf and Sm^Nd isotopic systems in theglobal sedimentary system Earth and Planetary Science Letters 16879^99

Viljoen K S Smith C B amp Sharp Z D (1996) Stable and radio-genic isotope study of eclogite xenoliths from the Orapa kimberliteBotswana Chemical Geology 131 235^255

Walter M J (1998) Melting of garnet peridotite and theorigin of komatiite and depleted lithosphere Journal of Petrology 3929^60

Wasserburg G J Jacobsen S B DePaolo D J McCulloch M Tamp Wen T (1981) Precise determination of SmNd ratios Sm andNd isotopic abundances in standard solutions Geochimica et

Cosmochimica Acta 45 2311^2323Westerlund K J Shirey S B Richardson S H Carlson R W

Gurney J J amp Harris JW (2006) A subduction wedge origin forPaleoarchean peridotitic diamonds and harzburgites from thePanda kimberlite Slave craton evidence from Re^Os isotope sys-tematics Contributions to Mineralogy and Petrology 152 275^294

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1873

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Page 10: Origins of Xenolithic Eclogites and Pyroxenites from the Central

Appendix 4) The contents of all other elements are mostlybelow their respective detection limits

Accessory phasesRutile and ilmenite occur in some of the eclogites (Table 1)All rutile grains have exsolved 10 ilmenite as lamellaeand rims Ilmenite-free areas of rutile in eclogites containbetween 962 and 988wt TiO2 and minor amounts ofAl2O3 (01^04wt ) Cr2O3 (5009 to 02wt ) andFeO (03^07wt ) Rutile in pyroxenite YK1915 contains004wt Al2O3 01wt FeO and 15wt Cr2O3Discrete ilmenite grains in sample VR67112b haveTiO2 contents of 499 and 500wt MgO of 23 and15wt and FeO of 399 and 403wt respectivelySulfides occur interstitially or enclosed in cpx garnet

and opx they are mostly pyrrhotite and monosulfide solidsolution with variable Ni contents (unpublished data)They will not be considered further hereAbundant platy phlogopite is observed in two samples

(VR19674ecl-g7 and YK1911) It is Cl-rich in sampleVR19674ecl-g7 (063wt vs 003wt in YK1911) andF-rich in sample YK1911 (097wt vs 005wt inVR19674ecl-g7) (Electronic Appendix 5) The same is truefor coexisting apatite (VR19674ecl-g7 254wt Cl and014wt F YK1911 006wt Cl and 306wt F)

Comparison with mineral inclusions indiamonds from Lac de Gras and witheclogite xenoliths from other Slave CratonlocalitiesMineral inclusions in diamonds from the Lac de Graskimberlites have been investigated by Davies et al (19992004) For garnets there is a striking similarity betweenthe inclusions in diamond and those in the high-Ca eclogitexenoliths although the inclusions in diamond trendtowards higher Na2O (Fig 5) and lower Al2O3 contentsThe overlap is not as marked for cpx where inclusionsin diamond span almost the entire range observed foreclogitic cpx (Fig 6)A comparison with eclogites from other localities in

the Slave Craton shows that garnets in most Diavik low-temperature eclogites reported by Schmidberger et al(2007) plot with garnets in high-Mg eclogites and inpyroxenites from this study (Figs 3^5) whereas garnets inhigh-temperature and diamond-bearing eclogites(Schmidberger et al 2007) overlap most with garnets inhigh-Ca eclogites from this study and with garnetsincluded in diamond Eclogites from the Jericho kimberlitehave either massive or anisotropic fabrics (Kopylova et al1999) Garnets in almost all massive eclogites plot withgarnets in low-Mg eclogites from this study with regard toCaO^MgO^FeO relationships (Fig 3) but many havelower Na2O contents at a given CaO content (Fig 5)Garnets in anisotropic eclogites are restricted to lowMgOFeO (507) and show some overlap with garnets

in high-Ca eclogites although trending towards lowerCaO(MgOthornFeO) (Fig 3) while coexisting cpx is alsorestricted to low MgO and trends toward high Al2O3 con-tents (Fig 6)

TRACE ELEMENTSGarnetTrace-element abundances in garnet are summarized inTable 3 and the complete dataset is reported in ElectronicAppendix 1 Garnets in low-Mg eclogites show steep posi-tive slopes between LaN and SmN and shallower positiveslopes in the normalized heavy rare earth elements(HREEN Fig 7a) The normalized light REE (LREEN)show an order of magnitude variability ZrHf is mostlysupra-chondritic and Ti can be enriched or depleted rela-tive to elements of similar compatibility Niobium in threesamples is strongly depleted at 002^003 chondriticGarnets in high-Mg eclogites also show smooth positiveslopes in LREEN but have flat HREEN and variablenormalized middle REE (MREEN) to HREEN (Fig 7b)ZrHf is always supra-chondritic and Ti enriched ordepleted relative to similarly compatible elementsNiobium is most samples is below detection Small positiveEu anomalies though not outside the analytical uncer-tainty for Eu relative to either Sm or Gd are observed forgarnet in some high-Mg eclogites (eg EuEu up to 15for VR67360 where Eu is the average of the chondrite-normalized Sm and Gd concentrations) Garnets in high-Ca eclogites are slightly enriched in the MREEN relative tothe HREEN or have flat MREEN and HREEN with ahump between EuN and HoN (Fig 7c) One sample(VR40345) shows strong La enrichment (greater thanchondrite) Two samples have supra-chondritic ZrHfwhereas three samples have close to chondritic ratios andone sample has a subchondritic ratio As is true for theother eclogite typesTiN is either enriched or depleted rela-tive to similarly compatible elementsPyroxenitic garnets have variable slopes in the LREEN

and flat MREEN to HREEN patterns and two samplesshow high Sc relative to neighbouring elements Garnet inone of the pyroxenites (YK1952) has a highly distinctivepattern relative to other pyroxenites with very low LREEand HFSE (Fig 7d) This sample consists of coarse pyroxe-nite schlieren in which the garnet was analysed in a finer-grained lsquomatrixrsquo of spongy cpx with secondary rims finecpx^opx intergrowths embayed garnet and melt patchesVolatile-rich eclogites have garnets that are slightly

enriched in MREEN relative to HREEN or have flatMREE to HREE patterns One garnet has a pronouncednegative Eu anomaly (EuEufrac14 07 forVR67112b Fig7e)Garnets in volatile-rich samples have the lowest Ni Coand Ti and the highest V ScYand HREE abundances inthe dataset

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1852

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 3 Summary of trace-element contents in garnet and cpx (ppm)

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR19674

ecl-g7

VR67112b YK1911 av min max 1s

n 6 6 7 3 5 4 4

P 169 54 270 88 196 70 310 78 282 80 450 138 160 60 80 214 48 370 144

Sc 62 40 89 16 48 36 67 11 53 42 79 15 69 92 93 186 52 440 179

Ti 1667 490 2500 680 1073 620 2000 455 2141 700 4700 1582 270 690 260 1097 540 1900 573

V 169 78 250 68 150 78 280 72 178 110 390 101 140 340 100 390 71 820 326

Co 59 42 70 11 58 39 84 13 58 47 65 5 30 41 27 68 55 90 16

Ni 17 5 32 10 21 11 34 10 24 11 35 8 6 10 51 54 25 110 38

Ga 12 7 15 3 13 8 16 3 12 9 20 4 8 86 58 12 10 14 2

Sr na 512 4 na na 522 522 na na 57 57 na 52 5025 510 513 513 513 na

Y 37 6 61 18 21 7 38 10 31 20 38 5 60 81 70 19 12 26 6

Zr 15 5 29 9 28 6 83 26 24 9 43 12 60 22 90 18 1 41 17

Nb 009 0005 048 019 na 50015 05 na na 5005 05 na 5003 0012 5001 022 0030 057 025

La 005 001 009 004 na 50007 00 na na 5003 04 na 004 002 0009 003 000 009 005

Ce 03 01 05 02 01 00 01 00 05 01 18 05 03 022 015 01 00 03 01

Pr 01 00 02 01 00 00 01 00 02 01 03 00 017 014 012 00 00 01 00

Nd 12 03 24 07 07 04 14 03 21 12 32 07 24 23 20 03 00 07 03

Sm 11 04 22 06 10 03 18 05 23 08 36 10 29 31 33 05 01 09 04

Eu 06 03 10 03 06 02 10 03 11 04 17 04 16 114 16 03 01 06 02

Gd 27 07 41 12 23 07 42 12 43 16 53 13 8 81 90 13 06 24 08

Dy 54 11 70 23 36 13 65 18 58 35 71 11 12 144 12 29 18 42 10

Ho 14 02 23 07 08 03 15 04 12 09 14 02 24 30 23 07 04 10 02

Er 46 07 80 24 24 09 40 10 34 25 40 05 8 85 6 22 14 28 06

Yb 52 06 98 30 24 10 37 08 32 22 39 05 8 70 5 23 13 30 07

Lu 09 01 16 05 04 02 05 01 05 03 07 01 11 09 08 04 02 05 01

Hf 03 02 05 01 04 01 10 03 05 02 09 02 09 027 007 03 00 06 03

Ta na 50008 5004 na na 50005 001 na na 50004 na na 003 50002 50005 na 50007 011 na

Pb na 5005 02 na na 5002 02 na na 5003 07 na 02 004 5002 na 5003 01 na

Th na 5001 005 na na 50007 003 na na 50008 011 na 5007 0012 50004 002 0005 006 002

U na 5003 5003 na na 5001 01 na na 5007 na na 01 0023 010 na 5001 003 na

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR67112b YK1911 av min max 1s

n 6 4 8 5 4 4

P 65 15 120 41 39 14 56 16 53 30 70 12 20 9 80 27 120 40

Sc 22 10 30 7 15 14 19 2 14 7 24 5 80 43 17 11 21 5

Ti 1557 990 2400 508 1168 1010 1490 191 1988 810 3500 1154 2390 1700 689 102 1685 694

V 415 180 480 107 318 210 400 69 324 259 450 72 600 180 250 139 480 156

Co 35 26 50 8 20 14 31 7 26 14 42 11 49 23 29 20 42 10

Ni 164 87 220 47 228 100 430 128 188 135 320 56 59 34 428 410 670 184

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1853

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

ClinopyroxeneTrace-element abundances in cpx are summarized inTable 3 and the complete dataset is reported in ElectronicAppendix 6 Clinopyroxenes in low-Mg eclogites havemildly positive to negative slopes from LaN to NdN andsmooth negative slopes between SmN and HoN thatmostly continue to LuN (Fig 8a) Niobium is below detec-tion or 5007 chondritic and SrN is variably enrichedrelative to neighbouring elements with the peak size corre-lating negatively with LREEN ZrHf is subchondritic andHfN is enriched relative to SmN whereasTiN is enriched ordepleted relative to similarly compatible elementsWe ana-lysed the spongy rim in cpxYK1943 and its trace-elementpattern is similar to that of cpx cores for more compatibleelements whereas the most incompatible elements areseveral orders of magnitude higher Clinopyroxenes in

high-Mg eclogites have steeper negative slopes betweenNdN and LuN than those in low-Mg eclogites combinedwith lower HREE abundances HfN is depleted relative toSmN (Fig 8b) High-Ca eclogites contain cpx with sinusoi-dal REEN patterns with lower LREE than other eclogitetypes and the heaviest REE below detection They haveuniformly strong positive Sr peaks relative to elements ofsimilar compatibility (Fig 9) The diamond-bearing high-Ca eclogite (VR40345) has cpx with unusually high Nband LREE abundancesClinopyroxenes in four pyroxenites have disparate trace

element patterns They all share depletions in TiN relativeto similarly compatible elements and two show positiveSrN peaks Two samples have cpx with high and flatLREEN to MREEN patterns and steep negative slopes inthe HREEN The trace-element patterns of cpx in YK1914

Table 3 Continued

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR67112b YK1911 av min max 1s

n 6 4 8 5 4 4

Ga 16 10 20 3 12 9 15 2 26 15 39 9 129 123 6 11 11 5

Rb na 5008 04 na na 5007 05 02 na 5003 03 na 02 5004 03 010 05 0

Sr 166 85 290 74 222 153 285 50 243 83 440 142 85 120 200 11 414 209

Y 2 04 4 10 13 07 2 04 11 03 3 10 28 51 2 03 3 2

Zr 19 4 48 15 25 8 45 13 17 8 28 7 111 129 36 008 81 42

Nb 01 0003 03 01 na 5002 00 na na 5002 16 9 007 011 010 003 02 008

Cs na 5001 6 na na 5006 5006 na5013 5004 5013 na bdl 5002 na 5001 5002 na

Ba 09 02 2 07 09 01 2 07 10 03 3 10 06 01 13 06 3 10

La 3 03 8 3 3 23 3 04 4 01 14 6 51 97 4 01 10 4

Ce 7 13 20 6 10 7 13 2 7 05 24 10 25 28 11 01 28 13

Pr 11 02 3 08 2 10 3 06 08 02 2 09 47 46 2 002 4 2

Nd 5 12 12 4 10 5 14 4 4 09 8 26 23 22 8 01 22 10

Sm 11 05 21 06 20 10 31 10 07 02 11 03 47 53 2 005 5 2

Eu 03 01 05 01 06 028 09 03 02 010 03 007 087 123 06 002 15 07

Gd 08 03 13 03 11 05 19 06 05 03 06 011 28 37 15 010 36 2

Dy 05 01 07 02 04 022 06 01 02 013 03 007 11 16 07 008 14 06

Ho 01 00 02 004 006 003 007 002 na 5003 012 na 011 021 009 001 02 007

Er na 507 507 na 011 004 02 00 na 5006 03 na 019 033 02 002 03 012

Yb na 5001 02 na na 5001 011 na na 5006 03 na 007 015 009 001 02 006

Lu na 5001 003 na na50002 002 na na50008 01 na 0006 0013 0013 0002 003 001

Hf 12 03 24 07 11 04 18 05 10 06 16 04 34 53 na50015 35 na

Ta na50003 50015 na na50006 006 na na50004 5004 na 0014 00236 na50003 503 na

Pb 06 02 12 04 06 04 09 02 na 5016 07 02 065 098 08 01 18 1

Th na50005 06 na 012 005 02 005 na 5002 5006 na 021 02 001 06 03

U na 5005 5005 na 002 0009 005 001 na 5003 5005 na 0013 024 na50005 003 na

Averaged for the three eclogite types and for pyroxenites as in Table 2 pyx pyroxenite n number of analyses averagedbdl below detection limit

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1854

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

and YK1952 are similar although at different abundanceswith only modest LREEN over HREEN enrichment andstrong depletions in ZrN relative to neighbouring elements(Fig 8d)Clinopyroxenes in the volatile-rich samples have flat pat-

terns and high abundances from LaN to SmN and smoothnegative slopes from SmN to LuN Niobium Sr and Ti are

noticeably depleted relative to elements of similar compat-ibility (Fig 8e)

Trace-element partitioningTo assess whether coexisting phases are in trace-elementequilibrium we calculated apparent cpx^garnet distribu-tion coefficients and compared them with those

(a)

(d)

(b)

(e)

(c)

GarnetCI-chondrite GarnetCI-chondrite

low-Mghigh-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1952

1

10

Nd Sm Eu Gd Dy

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Fig 7 Extended trace-element patterns of garnet in (a) low-Mg eclogites (b) high-Mg eclogites with inset showing enlargement of MREE tohighlight Eu anomalies and 1s uncertainties (c) high-Ca eclogites (d) pyroxenites and (e) volatile-bearing eclogites Normalized to chondritevalues of McDonough amp Sun (1995)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1855

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

determined experimentally or in demonstrably equili-brated natural samples Low-Mg and high-Mg eclogiteshave Dcpxgarnet for most elements that are within therange determined for natural equilibrated eclogites andexperimental assemblages (Harte amp Kirkley 1997 Greenet al 2000 Barth et al 2002a) (Electronic Appendix 7)High-Ca eclogites plot at the low end of the range

of distribution coefficients and have very low DMREE

and DHREE outside the range indicated by the datasetof Harte amp Kirkley (1997) who noted that DREE decreaseswith increasing Ca content in eclogitic minerals Theseassemblages are therefore considered to be in equilibriumPyroxenite VR40384 has D values similar to eclogites

but at the high end of the range determined by Harte amp

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

(a)

(d)

(b)

(e)

(c)

CpxCI-chondrite CpxCI-chondrite

low-Mg high-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1914

YK1952

spongy rim

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

VR40345

Fig 8 Extended trace-element patterns of cpx as in Fig 7

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1856

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Kirkley (1997) (Electronic Appendix 7) Another group(YK1914 and YK1952) plots at the low end but DHf DZrDTi and DV are far below the values indicated by experi-mental datasets Because pyroxenites compositionallygrade into eclogites their drastically different DHFSE

values suggest disequilibrium In contrast volatile-bearingsamples have DHFSE values that appear to be too high to bein equilibrium however this may be a consequence oftheir unusual bulk chemical composition (Ti- and K-rich)which differs substantially from that of lsquonormalrsquo eclogitesand experimental charges

GEOTHERMOBAROMETRYFollowing Schmidberger et al (2007) temperatures oflast equilibration were calculated for eclogitic andopx-free pyroxenitic assemblages using the formulationof Krogh (1988) The results are summarized in Table 2and Electronic Appendix 8 for a preset pressure of5 GPa and illustrated as a histogram in Fig 9 Foropx-bearing pyroxenites temperatures were calculatedsimultaneously with pressure [Ca-in-opx thermometryand Al-in-opx barometry using the formulations of Breyet al (1990)] to constrain their relative positions in themantle columnMinerals in some eclogites from Lac de Gras show dis-

tinct core and rim compositions or inhomogeneitiesTemperatures were not calculated for samples withspatially irresolvable inhomogeneous garnet compositions(SE01VR43479VR43480) For eclogites with spongy cpxcore compositions were used in the calculationTemperatures were calculated for both core and rim com-positions where applicable For all samples these

temperatures agree within the uncertainty of this thermo-meter (508C Brey amp Kolaquo hler 1990)Most equilibration temperatures for eclogites lie between

800 and10008C (Fig 9)The transition between the shallowand the deep lithospheric mantle layers in the central Lacde Gras area lies in this temperature interval (Griffin et al1999 Pearson et al 1999) Eclogites in this study lack thedistinct bimodal temperature distribution observed bySchmidberger et al (2007) but as in their study high-Caeclogites give higher average temperatures (10808C corre-sponding to a depth of 155 km if equilibrated along a 40mWm2 geotherm) than other eclogites types (9808Cdepth 135 km) the diamond-bearing sample (VR40345)gives thehighest temperature of the datasetHighNa2Ocon-tents in garnet from high-Ca eclogites are qualitativelyconsistent with higher equilibration pressures (Sobolev ampLavrentrsquoyev 1971) as are high K2O contents in coexistingcpx (Erlank ampKushiro1970) suggesting that high-Ca eclo-gites not only equilibrated at high temperature but also highpressure rather than representing a temperature excursionPyroxenites both opx-bearing and opx-free give somewhatlower average temperatures than eclogites (mean 9258C)and phlogopite-rich eclogites give the lowest temperatures(685 and 7698C respectively) The calcite-bearing eclogiteVR67112b gives a relatively low equilibration temperatureof 8558C

MEASURED ANDRECONSTRUCTEDWHOLE -ROCKSRationale and approachKimberlite-borne xenoliths sometimes show evidence forinfiltration of the host kimberlite (eg Taylor amp Neal1989 Barth et al 2001 2002b Heaman et al 2006) Wetherefore follow the standard approach of calculatingwhole-rock compositions from the compositions of the con-stituent phases weighted by their modal abundances Onlycore compositions were used as distinct rim compositionsmay point to late diffusion of elements from kimberlite-related melt veinsMany eclogitic whole-rock compositions reconstructed

with only garnet and cpx have negative Ti anomalies(not shown) For most of these samples rutile or ilmenitehas been identified as part of the mineral assemblage andthese phases should therefore be considered in the whole-rock reconstruction Rutile or ilmenite modes were esti-mated by assuming that the bulk-rock has no Ti anomalywhen normalized to a normal mid-ocean ridge basalt(N-MORB) composition and calculating the amount ofrutile required to erase theTi anomalyIt is of course possible that the eclogites had negative

or positive Ti anomalies relative to elements of similarcompatibility (Eu and Dy) Because we reconstructed

0

2

4

6

8

10

12volatile-rich

pyroxenite opx-bearing

pyroxenite opx-free

high-Ca

high-Mg

low-Mg

700750

800850

900950

10001050

11001150

12001250

Fre

quen

cy

Temperature

Fig 9 Histogram ofTKrogh (Krogh 1988) obtained at a preset pres-sure of 5 GPa except opx-bearing pyroxenites where TBKN wassolved simultaneously with PBKN (Brey amp Kolaquo hler 1990)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1857

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

whole-rock compositions from mineral core compositionskimberlite infiltration is unlikely to have caused fractiona-tion of Ti from Eu and Dy In a study in which rutilemodes were estimated from the differences between Ti inmeasured bulk-rocks and in bulk-rocks reconstructed withonly garnet and cpx only three of 19 eclogites recon-structed with rutile show Ti anomalies (Barth et al2002a) Even eclogite xenoliths with depleted Zr and Hfabundances that resemble those of modern island arcbasalts do not show Ti anomalies (Jacob amp Foley 1999)We therefore consider our approach to be a reasonablyrobust means of obtaining meaningful rutile modes wheresmall sample sizes preclude reliable determination of themodes of accessory phases by point-countingWhere rutilewas not recognized on thin-section scale or not analysedthe average eclogitic rutile composition was used for thereconstruction

UncertaintiesReconstructing whole-rock compositions in this way is arather crude method given the coarse grain size of theeclogites relative to the sample size and the non-uniformdistribution of the principal minerals on thin-sectionscale The uncertainty in mineral modes of medium- tocoarse-grained eclogites with grains45mm has been esti-mated by Jerde et al (1993) to be 10 plus uncertaintiesrelated to partial alteration which is minor in our samplesNevertheless the whole-rock trace-element budget israther insensitive to the exact modes for common eclogiticassemblages and REE patterns which encompass mostother elements in terms of compatibility in garnet andcpx do not change significantly even for modal variationsof 30 (Jerde et al 1993) This is not true for major-element concentrations which vary substantially with var-iations in mineral modes and we therefore do not calculatewhole-rock major-element contentsBecause the discussion that follows is based on trace-

element patterns and trends rather than absolute abun-dances whole-rock reconstruction based on mineralmodes is adequate We are confident that the uncertaintyinherent in this approach is not leading to significant mis-representation of eclogite whole-rock compositions becausethe trace-element patterns of the reconstructed whole-rocks show distinct differences (see below) that correlatewith the assignment to eclogite types based on garnetmajor-element composition and this is independent ofmineral mode Following Jerde et al (1993) in ElectronicAppendix 9 we show variations in trace-element patternfor one sample in each eclogite type for variations ingarnet and cpx abundance from 7030 to 3070 vol This shows that the trace-element patterns are generallyinsensitive to cpxgt ratios with regard to slopes andanomalies such as positive Sr Pb and Eu anomalies withthe exception of the positive Eu anomaly in high-Mgeclogite VR43479 which disappears if a high cpx mode is

assumed (not shown) Variations in rutile modes by 50(Electronic Appendix 9d) lead as would be expected tothe appearance of Ti anomalies but a rutile mode of 05instead of 1 vol does not erase the positive Nb anomalyin low-Mg eclogiteVR50909Because of the robustness of the relative trace-element

abundances in the reconstructed eclogites a garnet propor-tion of 065 and cpx proportion of 035 approximately themedian modal abundances of these minerals in volatile-free eclogites from Lac de Gras are assumed for eclogitesamples that are too small for determination of meaningfulmodes by point-counting

Within- and between-group variationsThe trace-element compositions of reconstructed whole-rocks are given in Table 4 N-MORB-normalized trace-element patterns of reconstructed whole-rocks areshown in Fig 10 The three eclogite groups have distincttrace element patterns that are particularly obvious forthe REE (Fig 11) Low-Mg eclogites have positive Pb andSr anomalies and smooth positive slopes in their MREEN

and HREEN LREEN are enriched relative to MREEN inthree samples (Fig 11a) Three samples have high Nbrelative to similarly compatible elements and comparedwith most other eclogites (Fig 10a) All high-Mg eclogitesdisplay more or less pronounced positive Eu and Sranomalies and most have positive Pb and negative Zrand P anomalies Their N-MORB-normalized REE pat-terns show shallow positive slopes from LaN to PrN orNdNand aweak negative slope from NdN to LuN (Fig11b)VR43479 is a low-Mg eclogite with regard to its garnet

Ca^Fe^Mg composition but has a trace-element patterntypical of high-Mg eclogites and will be discussed withthe latter Conversely sample YK1949 a high-Mg eclogitein terms of its garnet Ca^Fe^Mg compostion is more simi-lar to low-Mg eclogites in terms of its whole-rock trace-element pattern and is plotted in Figs 10a and 11aWhereas the patterns of most samples within the groupresemble each other the N-MORB-normalized incompati-ble-element abundances of the high-Ca eclogites vary bymore than an order of magnitude (Fig 10c) All sampleshave positive Pb anomalies and most have positive Sranomalies Some samples have negative P andor Zranomalies Niobium abundances in three of the samplesare very high similar to those for some low-Mg eclogitesHigh-Ca eclogites tend to have flatter MREEN andHREEN patterns and more uniform HREE abundancesthat the other eclogite types with humps between SmN

and HoNThe phlogopitized sample YK1911 shows a strong posi-

tive Pb anomaly and lower HREE abundances than thecalcite-bearing sample VR67112b which shows a distinctnegative Eu anomaly and a depletion of La relative to Ce(Figs 10d and 11d)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1858

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Trace-element compositions of reconstructed whole-rocks (ppm)

Sample P Sc Ti V Co Ni Ga Sr Y Zr Nb La Ce Pr Nd

Low-Mg eclogites

vr19674-g4 186 50 4649 317 57 59 144 77 24 30 36 13 31 05 24

vr43452 154 39 2253 328 49 84 176 52 32 11 01 02 07 01 08

vr43469 38 65 4377 349 54 91 142 34 28 15 02 05 15 03 17

yk1926 144 45 3406 258 52 87 145 89 18 18 20 09 29 06 33

yk1943 173 46 5732 295 41 39 146 122 25 48 82 34 85 13 64

yk19491 102 34 1693 158 31 49 100 46 10 14 03 09 21 03 17

High-Mg eclogites

vr19677-g2 138 39 5160 291 41 53 119 98 14 22 04 13 53 11 64

vr434792 46 25 1265 129 36 126 84 56 3 7 04 06 25 04 23

vr67360 49 36 1573 138 64 184 111 76 5 9 04 09 38 07 33

yk1946 153 28 4877 192 45 105 124 108 12 48 03 10 38 08 50

High-Ca eclogites

vr19674-g5 267 30 7663 200 41 67 195 32 23 43 77 01 04 02 16

vr40345 0 33 2821 175 47 98 210 179 13 13 66 60 109 11 38

vr43465 142 27 5104 220 31 83 121 197 14 35 29 05 22 05 40

vr50858 295 30 6898 201 44 94 150 48 22 38 06 01 03 02 20

vr50860 68 57 5672 418 57 71 186 109 19 11 02 02 13 03 26

vr50909 238 28 7295 201 42 92 171 48 20 57 13 01 04 02 21

yk3528 91 51 3712 341 53 155 177 49 16 9 01 02 08 02 11

Volatile-rich eclogites

vr67112b 40 84 9242 521 46 30 107 4 40 66 31 26 126 24 125

yk1911 31 39 9428 286 44 82 164 3 23 52 64 18 52 09 46

Pyroxenites

yk1914 111 49 707 207 48 547 44 31 4 3 02 06 19 03 10

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

Low-Mg eclogites

vr19674-g4 11 05 10 22 39 10 29 33 06 12 05 028 24 008

vr43452 04 02 08 13 38 12 41 49 08 05 01 nd 02 na

vr43469 11 05 09 25 44 11 34 36 06 07 02 001 09 129

yk1926 12 05 11 19 30 07 22 24 04 07 03 004 23 090

yk1943 21 08 10 29 42 10 30 32 05 16 05 013 65 017

yk19491 06 03 11 08 14 04 14 18 03 04 02 003 29 115

High-Mg eclogites

vr19677-g2 20 08 12 21 24 05 15 15 03 05 05 007 52 6811

vr434792 05 02 12 05 06 01 04 03 00 03 01 001 98 037

vr67360 06 03 14 06 09 02 06 06 01 02 02 002 86 000

yk1946 18 07 12 21 22 05 13 12 02 11 03 006 56 112

High-Ca eclogites

vr19674-g5 17 09 11 33 43 09 24 23 04 16 05 007 02 011

vr40345 09 04 10 12 22 06 16 14 02 05 nd nd 298 na

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1859

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Continued

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

vr43465 22 09 12 25 26 06 15 14 02 10 04 001 25 040

vr50858 20 08 09 33 41 08 22 21 03 10 01 nd 02 na

vr50860 14 07 11 27 37 08 22 18 03 06 00 nd 06 na

vr50909 20 10 12 33 38 07 20 20 03 16 01 002 02 181

yk3528 08 05 11 20 30 07 18 18 02 05 01 nd 08 na

Volatile-rich eclogites

vr67112b 38 10 07 53 74 15 41 34 04 18 03 012 60 004

yk1911 20 07 08 35 39 07 18 17 02 10 14 nd 73 na

Pyroxenites

yk1914 02 01 11 04 08 02 05 06 01 01 03 003 68 101

aGarnet has CandashMgndashFe relationship typical of high-Mg eclogites but whole-rock trace-element pattern is typical oflow-Mg eclogitesbGarnet has CandashMgndashFe relationship typical of low-Mg eclogites but whole-rock trace-element pattern is typical ofhigh-Mg eclogites

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

(a)

001

01

1

10(c)

001

01

10(b)

VR43452

YK1949

RockNMORB

low-Mg

RockNMORB

high-Mg

(d)

volatile-richhigh-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

phlogopite-bearingYK1911

calcite-bearingVR67112b

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 10 Extended trace-element patterns of reconstructed (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-bearing eclogites Normalized to N-MORB of Sun amp McDonough (1989)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1860

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

RockNMORB

(b)

RockNMORB

(d)

010

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu001

01

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

(a)

001

01

1

10(b)

001

01

1

10(c)

001

01

1

10(e)

high-TSchmidberger et al

low-TSchmidberger et al

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

high-Mglow-Mg

high-Cavolatile-rich

phlogopite-bearingYK1911

calcite-bearingVR67112b

YK1949VR43479

Fig 11 N-MORB normalized REE patterns for (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-richeclogites Shown for comparison in (e) are reconstructed eclogites from Schmidberger et al (2007)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1861

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Both volatile-rich whole rocks display strong negativeSr anomalies Strontium can be sited in calcite or apatiteandor indicate the presence of a Sr-rich mineral that wasnot identified in thin section such as bariteFour of the six reconstructed whole-rocks in the study

of Schmidberger et al (2007) (Fig 11e) show markedLREEHREEN enrichment that is observed only in onehigh-Ca eclogite in our study (Fig 11c) suggesting thatmany of the eclogites they studied were affected by metaso-matic LREE enrichment Two samples in their study thatare not affected by this enrichment show positive Euanomalies similar to the high-Mg eclogites in this studywhereas the distinctive REE pattern of the low-Mg eclo-gites is not recognized among their samples

SR^ND^HF I SOTOPESYSTEMATICSStrontium Nd and Hf isotope data for optically puremineral separates and reconstructed whole-rocks aregiven in Table 5 Spongy rims and alteration are likely tobe mostly removed during leaching in 6N HCl in combina-tion with repeated ultrasonication Despite this rigoroustreatment the reproducibility of Nd in garnet VR43469and in cpx YK1914 is worse than for other samples Thismay be due to inhomogeneities related to late Nd diffusionpossibly from grain boundaries containing a kimberlitecomponent The reproducibility of the 176Hf177Hf data isgenerally poorer than that of Nd which may reflect a com-bination of matrix effects caused by imperfectly purifiedsolutions and sample heterogeneity Parent^daughterratios are available only from in situ analyses which arenot precise enough to allow meaningful model age orinitial isotope ratio calculations (ie at the time of kimber-lite eruption) Because the Lac de Gras kimberlites arerelatively young (55^56 Ma Graham et al 1999) age cor-rection would result in only a minor change in isotopecomposition which is inconsequential in view of the rangeof compositions displayed by most samples and hence hasno effect on the conclusions drawnClinopyroxenes in five volatile-free eclogites and a

pyroxenite have Sr isotope ratios ranging from 0701662 to0703546 this range is similar to or somewhat lower thanthat of the depleted mantle at the time of kimberliteemplacement (0703 Bell amp Tilton 2002) Clinopyroxenein the phlogopitized sample has a distinctly high 87Sr86Srof 0751186Clinopyroxene controls 56^92 of the Nd budget (Nd

concentration in cpx weighted by mode) Neodymiumisotope ratios range from 0511660 to 0514844 Thiscorresponds to eNd variation (per 10 000 deviationof 143Nd144Nd in the sample from 143Nd144Nd in the chon-dritic mantle model reservoir) from 17 to thorn43 cpx inthe low-Mg eclogites have higher eNd than those in

high-Mg eclogites Garnets have generally have higher143Nd144Nd from 0512984 (eNdfrac14thorn 7) to 0520965(eNdfrac14thorn 162) Garnets in the two volatile-rich sampleshave the highest ratios 0522790 for garnet in the calcite-bearing sample VR67112b (eNdfrac14thorn198) and 0525721 inthe phlogopitized eclogiteYK1911 (eNdfrac14thorn 255)Garnet contributes 18^32 and cpx 48^78 of the Hf

budget The proportion controlled by rutile is 4^30(Aulbach et al in preparation) Measured 176Hf177Hf incpx ranges from 0281263 in phlogopite-bearing sampleYK1911 (eHffrac1453) to 0286130 in low-Mg eclogiteVR43469 (eHffrac14thorn119) and in garnet from 0283958 inpyroxeniteYK1915 (eHffrac14thorn42) to 0295718 in low-Mg eclo-gite YK1926 (eHffrac14thorn458) Hafnium in garnet in the phlo-gopite^apatite-bearing sample is extremely radiogenic(176Hf177Hffrac14 0354367) corresponding to an eHf of thorn2532which is close to a value determined for a garnet separatefrom a Roberts Victor eclogite (eHffrac14thorn2561 Jacob et al2005) Garnet in the calcite-bearing sample has extremelyradiogenic Nd and Hf (eNdfrac14thorn193 eHffrac14thorn2516) Garnetin pyroxenite YK1915 also has radiogenic Nd and Hf(eNdfrac14thorn27 eHffrac14thorn42)Despite the large uncertainties for reconstructed whole-

rocks (calculated as the square root of the sum of squareduncertainties of two standard errors for garnet and cpxanalyses combined with 20 propagated uncertainty onthe proportion of Nd or Hf respectively contributed byeach mineral) the isotopic data are consistent in thatthey correlate with the independently determinedparent^daughter ratios (1) Low-Mg eclogites have higherSmNd reflected in the steeper positive slope of theirN-MORB-normalized REE patterns (Fig 11a) comparedwith the high-Mg eclogites (Fig 11b) Accordingly recon-structed low-Mg eclogites have higher present-day eNd

(0 to thorn47) than high-Mg eclogites (4 and 17) (2) Low-Mg eclogite YK1943 has very high Nb and LREE abun-dances that could have been added during metasomatismthat also enhanced its Hf abundance This sample has thelowest eNd and eHf of the low-Mg eclogites (0 and thorn20respectively) In contrast apparently unenriched low-Mgeclogites have more radiogenic Nd and Hf (eNdfrac14thorn47and thorn9 eHffrac14thorn128 and thorn113 respectively)The Sr^Nd isotope diagram reveals that eNd in cpx

varies from 19 to 43 at nearly constant depleted-mantle-like 87Sr86Sr (Fig 12a) In the Nd^Hf isotope diagram(Fig 12b) the reconstructed whole-rocks lie on curvedmixing lines if the lsquomixingrsquo proportions of Nd and Hffor the two lsquoend-memberrsquo minerals are very different(eg YK1911) Garnets have more radiogenic Nd andHf than coexisting cpx All reconstructed volatile-freeeclogites plot in the radiogenic field with regard to Hfor are isotopically similar to the depleted mantle Twohigh-Mg eclogites and one low-Mg eclogite have unradio-genic Nd one high-Mg eclogite has radiogenic Nd and

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1862

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 5 Individual and averaged Sr Nd and Hf isotope ratios obtained from garnet and cpx solutions

Sample 87Sr86Sr 2SE2SD 143Nd144Nd 2SE2SD eNd 2SE2SD 176Hf177Hf 2SE2SD eHf 2SE2SD

Low-Mg eclogites

VR43469 cp 0702661 0000010 0514844 0000006 43 01 0286130 0000012 119 04

VR43469 gt 0515508 0000014 56 03 0286506 0000012 132 04

VR43469 gt r 0515453 0000010 55 02 0286548 0000011 134 04

VR43469 gt r 0515384 0000007 54 01 0287244 0000010 158 04

VR43469 gt r 0515487 0000014 56 03 0286996 0000014 149 05

VR43469avg gt 0515458 0000109 55 21 0286823 0000715 143 253

WR VR43469 0515059 0000273 47 5 0286405 0001079 128 38

YK1926 cp 0703546 0000011 0513043 0000007 8 0 0283154 0000009 13 03

YK1926 gt 0513364 0000008 14 01 0295718 0000005 458 02

WR YK1926 0513120 0000026 9 1 0285980 0000716 113 25

YK1943 cp 0512143 0000007 10 01 0282213 0000005 20 02

YK1943 gt 0514461 0000012 36 02 0287510 0000008 168 03

YK1943 gt r 0514491 0000009 36 02 0287332 0000005 161 02

YK1943 gt r 0514517 0000010 37 02 0287290 0000007 160 02

YK1943avg gt 0514489 0000056 36 11 0287377 0000233 163 82

WR YK1943 0512625 0000228 0 5 0283138 0000879 13 31

High-Mg eclogites

VR43480 gt 0520965 0000017 162 03 0284950 0000010 77 04

VR67360 cp 0283509 0000074 26 26

VR67360 cp r 0283433 0000031 23 11

VR67360avg cp 0701662 0000018 0511660 0000010 19 02 0283471 0000108 25 38

VR67360 gt 0512984 0000101 7 20 0287429 0000180 165 64

WR VR67360 0511761 0000132 17 3 0284734 0000469 69 17

YK1949 cp 0703019 0000013 0512231 0000009 8 02 0282513 0000009 9 03

YK1949 gt 0513511 0000010 17 02 0290297 0000042 266 15

YK1949 gt r 0513508 0000012 17 02 0289947 0000015 254 05

YK1949 gt r 0513552 0000014 18 03

YK1949avg gt 0513523 0000049 17 10 0291569 0000494 311 175

WR YK1949 0512446 0000103 4 2 0284622 0000847 65 30

Volatile-rich eclogites

VR67112b gt 0522790 0000008 198 02 0354367 0000036 2532 13

YK1911 cp 0751186 0000009 0512165 0000007 9 01 0281263 0000009 53 03

YK1911 gt 0525721 0000013 255 03 0288051 0000012 187 04

WR YK1911 0513940 0000342 25 7 0281414 0000018 48 1

Pyroxenites

YK1914 cp 0703015 0000034 0513283 0000041 13 08

YK1914 cp r 0702956 0000005 0513200 0000027 11 05

YK1914 cp r 0513527 0000059 17 11

YK1914 cp r 0513045 0000010 8 02

YK1914avg cp 0702986 0000083 0513264 0000402 12 78

YK1915 gt 0514003 0000011 27 02 0283958 0000021 42

Analyses of full procedural replicates are designated with an lsquorrsquo after sample name SE standard error (internal errorsingle analyses) SD standard deviation (external error averaged multiple analyses) Italics used for reconstructed whole-rocks (mineral isotopic compositions and concentrations weighted by modes) Uncertainties for reconstructed whole-rocksinclude 2SE or 2SD uncertainties on each mineralrsquos isotope composition combined with 20 uncertainty in the amountof Nd or Hf contributed by the constituent minerals eNd and eHf are per 10 000 deviation from model chondritic mantle[model of Wasserburg et al (1981) and Blichert-Toft amp Albarede (1997) respectively]

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1863

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

εHf

εNd

Ekati eclogitesminus100

0

100

200

300

400

500

minus40 0 40 80 120 160 200 240 280

gt in YK1915

gt in VR67112bεNd=198 εHf=2532

minus40

0

40

80

120

160

minus40 minus20 0 20 40 60

minus60

minus40

minus20

0

20

40

070 071 072 073 074 075 076

North Americankimberlite field

εNd

87Sr86Sr

cpx in YK1914

Ekatieclogites

DM

DM

DM

DM

(a)

(b)

pyroxenites

low-Mg

high-Mg

volatile-rich

eclogites

Diavik low-TA154 S kimberlite

cpx in YK1911

Fig 12 (a) Sr^Nd for cpx and (b) Nd-Hf isotope diagram for garnet cpx and reconstructed whole-rocks (symbols with bold outline) Finedashed lines in (b) connect data points from the same sample with garnets having more radiogenic isotopic compositions Uncertainties forreconstructed whole-rocks combine 20 modal uncertainty with uncertainties on each mineralrsquos isotope composition Error bars on individualminerals may be smaller than the symbols Also shown are the isotopic compositions of the depleted mantle (DM DePaolo 1981 Griffin et al2000 Bell amp Tilton 2002) Range of 87Sr86Sr in eclogites from Ekati (Lac de Gras area 100 km north of Diavik Jacob et al 2003) in NorthAmerican kimberlites and compositions of low-temperature eclogites from Diavik and of Lac de Gras A154 South kimberlite fromSchmidberger et al (2007)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1864

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

one low-Mg eclogite has eNd similar to the depleted mantle(Fig 12b)Clinopyroxene in the phlogopitized eclogite YK1911

has unradiogenic Nd and Hf and strongly radiogenicSr whereas the coexisting garnet has radiogenic Nd andHf Although this sample contains 50 vol phlogopiteits whole-rock isotopic composition was calculated

from garnet and cpx alone because the contributionfrom phlogopite to the Hf and Nd budget is 52 and51 respectively (unpublished data) The whole-rockhas radiogenic Nd but has lower eHf than anyother sample analysed in accord with the extreme Hfabundance (53 ppm) and low LuHf of the cpx

DISCUSS IONOrigins of eclogites beneath the SlaveCratonHigh-Mg and high-Ca eclogites

With regard to their major- and trace-element contentsthe high-Mg and high-Ca eclogites from Lac de Gras arebroadly similar to variably evolved oceanic gabbros(Fig 13b and c) which may indicate an origin as subductedoceanic crust Compatible trace-element abundancesare useful to deduce potential crustal protoliths as theseelements are less susceptible to post-formation modification(Jacob 2004) The flat HREE patterns in the high-Mg andhigh-Ca eclogites are compatible with a low-pressureorigin as the oceanic crust is a product of moderate degreesof partial melting mainly within the spinel stability field(eg Presnall et al 2002) whereas melting with residualgarnet would qualitatively be expected to produce stronglyfractionated REE patterns Positive Sr and Pb anomaliesalthough compatible with a plagioclase signature couldalso be due to cryptic metasomatism (Dawson 1984) ofthe eclogites during their residence in the lithosphericmantle These anomalies are probably not related to kim-berlite infiltration as the whole-rock eclogite compositionswere reconstructed from analyses of fresh mineral coresLow ZrSm similar to that in primitive gabbros is

observed for many of the high-Mg and high-Caeclogites (Fig 13b and c) and is independent of the ratioof garnetcpx and whether or not the whole-rock wasreconstructed with rutile (Electronic Appendix 10) Thenegative trend of Mg-number against Yb in garnet(Fig 14) argues against igneous processes involving garnetaccumulation (Taylor amp Neal 1989) as suggested forexample for some high-Mg eclogites from West Africa(Barth et al 2002b) Rather the anti-correlation suggeststhat olivine fractionation played a role in the formation ofthe protoliths Fractionating olivine would excludeYb (Dolivinebasalt

frac14 002 Nikogosian amp Sobolev 1997)and incorporate Mg relative to Fe (Dfrac1415 and 05 respec-tively Taura et al1998) leading to progressive depletion ofMg and enrichment of Yb in the residual melt and henceanti-correlated Mg and Yb in the whole-rock This isinherited after eclogitization All eclogite types (exceptVR43452) plot along the olivine fractionation trend sug-gesting that they formed by a similar processHigh-Mg eclogites have lower

PREE than high-Ca

eclogites and some appear to have positive Eu anomalies

high-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

gabbrosSE Indian ridge

gabbrosOman ophiolite

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

high-Mg

RockNMORB(a)

VR43452

YK1949

low-Mg

boninite-like melts

01

001

1

10

(b)

01

001

1

10

(c)

01

001

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

Fig 13 Trace-element patterns of (a) low-Mg (b) high-Mg and(c) high-Ca eclogites compared with boninite-like volcanic rocksfrom the Isua greenstone belt (Polat et al 2002) gabbros from theOman ophiolite (Benoit et al 1996) and the SE Indian Ridge (Hartet al 1999) normalized to N-MORB of Sun amp McDonough (1989)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1865

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(EuEu up to 14) similar to more primitive cumulates(troctolite gabbros and olivine gabbros) reported in the lit-erature (Fig 13b) Although the sizes of these Eu anomaliesare not outside the analytical uncertainty their correlationwith CeYb in reconstructed eclogites supports their signif-icance (r2frac14066 nfrac14 4) The more primitive nature ofhigh-Mg eclogites is consistent with higher Mg-numbersrelative to high-Ca eclogites and may be due to olivineaccumulation during formation of their protolithsThe two samples with the lowest HREE abundanceshave the highest EuEu which is compatible with varia-tions in plagioclase modesMore evolved protoliths with higher pyroxeneplagio-

clase plus trapped melt could account for small or absentpositive Eu anomalies and higher

PHREE of high-Ca

eclogites (Fig 13c) Although the reason for the frequentassociation of graphite or diamond with high-Ca eclogitesis unclear (Aulbach et al 2003) the frequent occurrence ofkyanite in high-Ca eclogites also observed at Ekati justnorth of Lac de Gras has been argued to support a crustalprotolith (Jacob 2004) Kyanite andor corundum havebeen suggested to exsolve from high-temperature high-Alclinopyroxenes (Caporuscio amp Smyth 1990) Eclogitesfrom Lac de Gras show no exsolution textures suggestiveof such an origin although recrystallization can obscureor destroy microstructural evidence for exsolution(eg Griffin et al 1984)Modelling shows that qualitatively plagioclase accumu-

lation from a low-pressure melt derived from MORBmantle can account for positive Pb Sr and Eu and negativeZr anomalies relative to N-MORB whereas cpx accumula-tion or trapped melt increases Zr Y MREE and HREE

abundances (Fig 15) Some melt addition (for exampleby entrapment) seems to be required to increase all trace-element abundances to those observed for the eclogitesAlthough high-Ca and high-Mg eclogites could representshallow and deep portions of the same oceanic crust thehigher average temperatures obtained for the high-Caeclogites the frequent occurrence of carbon in high-Caeclogites and its absence in high-Mg eclogites the strongLREEHREE fractionation observed for six of seven high-Ca eclogites and the narrow range of compositions of thehigh-Mg eclogites suggest that they may have protolithsthat are not cogenetic Some discrepancies in particularaffecting the incompatible elements may be due to post-crystallization processes acting upon the protoliths suchas dehydration melting and later metasomatism that isevident in peridotites from the same locality (Griffin et al1999)Low K2OTh for high-Ca and high-Mg eclogites

(average 08 Table 4) relative to assumed crustal

0

2

4

6

8

10

30 40 50 60 70 80

Mg

VR43452

Yb

GARNET

pyroxenites

low-Mghigh-Mghigh-Ca

eclogites

VR19673ecl-g2

Fig 14 Garnet Mg-number vs Yb (ppm including 2s uncertainty)Lines qualitatively show separate correlations for eclogites (exceptoutlierVR43452) and pyroxenites (except outlierVR19673ecl-g2)

high-Ca

high-Mg

001

01

1

10

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

melt08 melt02 cryst cpx01 melt09 cryst

plag

(a)

001

01

1

10(b)

RockNMORB

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 15 Reconstructed (a) high-Ca and (b) high-Mg eclogitescompared with compositions of plagioclase and cpx crystallizingfrom a MORB melt modelled using distribution coefficients ofNorman et al (2005)The MORB was calculated as a 10 equilibriummelt of MORB pyrolite [composition from Kelemen et al (1993)except Pb Pr Sr Y and Lu which were calculated as residues from1 fractional melting of pyrolite from McDonough amp Sun (1995)using melt modes of Kinzler (1997) and distribution coefficients ofNorman et al (2005)] Also shown are mixtures of crystals (consistingof 50 plagioclase plus 50 cpx) and melt in the proportions 0901and 0208

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1866

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

precursors (eg gabbros SE Indian Ridge average 22Hart et al 1999) may reflect the preferred loss of fluid-mobile elements during dehydration (Becker et al 2000)although large uncertainties on these low-abundance ele-ments in the constituent minerals and reconstructedwhole-rocks makes this interpretation less than robustThe REE patterns of most high-Mg eclogites are too flatto represent melting residues as partial melting shouldproduce low LREEHREE (Electronic Appendix 10) Incontrast the marked LREE depletion of five of sevenhigh-Ca eclogites (LaLufrac14 02^08) contrasts with that ofassumed evolved gabbroic protoliths (LaLufrac1415^32Hart et al 1999) and can be modelled as a consequence ofpartial melting during subduction after eclogitization (iein the presence of garnet which retains HREE relative toLREE eg Barth et al 2002a) has taken place (Aulbachet al 2003)Our results are compatible with previous studies on

eclogite xenoliths from the Slave Craton includingthose from the same kimberlite that have linked thesesamples or some of these samples to Paleoproterozoicsubduction events (the 195^191Ga formation of theHottah terrane and the 18Ga formation of the GreatBear magmatic arc) in the Slave craton (Griffin et al1999 Kopylova et al 1999 Heaman et al 2002 2006Schmidberger et al 2005 2007)The compositional features discussed above although

compatible with crustal protoliths are not necessarilydefinitive of a subduction origin However if direct crystal-lization from basaltic melts at high pressure (ie mantledepths) is invoked instead specific conditions would haveto be met to satisfy the observations The Mg-number ineclogites from Lac de Gras are lower than those expectedof primary high-pressure mantle melts (eg Walter 1998OrsquoHara amp Herzberg 2002) requiring crystallization athigh pressures after olivine fractionation This originwould further require shielding from peridotitic wallrockwith which the fractionated melts would no longer be inequilibrium (Barth et al 2001) In such a setting the lowrelative abundances of fluid-mobile elements comparedwith those of less fluid-mobile elements and of LREErelative to HREE (Table 4) would reflect an expulsion offluids during crystallization of the protoliths and latermelt mobilization respectively The eclogites clearly donot represent garnet plus cpx cumulates which wouldrequire a positive correlation of Mg-number with Yb thatis at variance with the observed negative correlation ineclogitic garnets from this study (Fig 14)

Low-Mg eclogites

Garnets in low-Mg eclogites compositionally overlap withthose in massive eclogites from Jericho (NW of the Lacde Gras kimberlites) that were suggested to have a high-pressure intrusive origin based on the similarity of theirmajor-element compositions to those of Group A eclogites

fromsouthernAfrica forwhichamantleoriginhasbeenpro-posed (Kopylova et al1999) Garnets in low-Mg eclogites inthe present study have major-element compositions moresimilar to garnets in groupB rather thangroupAeclogitesThe positive slope in N-MORB-normalized HREE

abundances displayed by the low-Mg eclogites is clearlydistinct from other eclogite types and is similar to those ofboninite-like second-stage melts that are closely associatedwith ultramafic volcanics (Smithies et al 2004) (Fig 13a)The range of Mg-numbers in the low-Mg eclogites (55^66)overlaps with that of Archaean second-stage melts fromIsua (58^75) and Abitibi (52^82 Smithies et al 2004)These similarities might indicate that the low-Mg eclogiteshave an origin in a previously depleted mantle source suchas supra-subduction zone mantle Subchondritic ZrHf ofthe low-Mg eclogites contrasts with the supra-chondriticZrHf of the high-Mg and high-Ca eclogites (Aulbach et

al in preparation) and is consistent with previous cpx-controlled depletion in an arc mantle source This sourcecould have been subsequently remelted to generate theboninite-like precursor to the low-Mg eclogites as sug-gested for the protoliths of eclogites from Udachnaya(Jacob amp Foley 1999)Although some second-stage melts to which the low-Mg

eclogites are similar are closely associated with komatiitesand could contain a plume component the radiogenic ordepleted mantle-like Nd and Hf and unradiogenic Sr inthe low-Mg eclogites do not support plume involvement

Trace-element and Sr^Nd^Hf isotopeconstraints on post-formationmetasomatismSeveral low-Mg and high-Ca eclogites show negativeslopes in the LREE which may be indicative of interactionwith small amounts of LREE-enriched melt (Navon ampStolper 1987) amounting to cryptic metasomatism(ie not accompanied by crystallization of new phases) inthe nomenclature of Dawson (1984) High Nb abundancesrelative to elements of similar compatibility displayed bysome low-Mg and high-Ca eclogites (Fig 10) could resultfrom overestimation of rutile modes However high-CaeclogiteVR40345 is rutile-free and has been reconstructedwithout rutile its high Nb abundance is accompanied byLREE enrichment indicating that metasomatic additionmay be responsibleLow-Mg eclogites show a positive correlation of La with

Nb (Electronic Appendix 11a) indicating concomitantaddition of LREE and Nb This suggests that the metaso-matic agent was a solute-rich fluid or silicate silicic melt ashydrous fluids and carbonatites do not carry enough HFSEto affect the budget of eclogites (Green amp Wallace 1988Klemme et al 1995 Adam et al 1997 Stalder et al 1998)Of the samples that are enriched in Nb all but one (high-Ca eclogiteVR50909) show a simultaneous enrichment in

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1867

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Hf (Electronic Appendix 11b) Together these trace-element relationships suggest that both SmNd and LuHfwere lowered in these samples Their low parent^daughterratios contrast with higher ratios in unenriched samplesand resulted in the ingrowth of a range of Nd and Hf iso-tope compositions Compared with the low-Mg and high-Ca eclogites the high-Mg eclogites have a narrow range ofLa and Nb concentrations which correlates with a narrowrange in Nd and Hf isotope compositions (Fig 12)Precise model ages cannot be calculated because parent^

daughter ratios are available only from in situ analyseswhich have relatively large uncertainties Howeverevidence for long-term enrichment and depletion may bediscerned Low-Mg eclogites even those that are enrichedin LREE have relatively high SmNd as a result of havingoverall positive slopes in their REE patterns relative toMORB whereas the high-Mg eclogites have relativelyflat REE patterns and therefore lower SmNd (Fig 11)This translates into lower eNd for high-Mg eclogites thanfor low-Mg eclogites In contrast they show no systematicdifference with respect to 176Hf177Hf Because the low- andhigh-Mg eclogites are proposed to have different origins asdiscussed above and therefore did not necessarily have thesame initial isotopic composition and formation age theseisotopic differences cannot be used to estimate their agesA long-term coupling of trace-element and isotopic

abundances is also indicated by the fact that low-Mg eclo-gite YK1943 which has experienced Nb and LREE addi-tion has lower eNd (0) and eHf (thorn13) than low-Mg eclogiteVR43469 with higher LuHf and SmNd (eNdfrac14thorn47eHffrac14thorn128) If low-Mg eclogites YK1943 and VR43469originally had identical initial Nd and Hf isotope composi-tions and if the parent^daughter ratios are accurate andwere modified only once during their evolution shortlyafter metamorphism (when their isotopic compositionwould have been reset) this overprint can be calculatedfor both isotopic systems to have occurred at 17GaThis would suggest that Hf and Nd were added duringthe same event In contrast high-Mg eclogites VR67360and YK1949 would have had identical Ndisotope compositions at 10Ga whereas their Hf isotopecompositions are disturbed in that the sample with thehigher parent^daughter ratio has a lower 176Hf177Hf thanthe sample with the lower parent^daughter ratio Theseresults suggest that the LREE and HFSE enrichmentobserved in different samples occurred during separateevents that did not affect the eclogites equallyExtreme HFSE enrichment leading to the growth of

zircon and rutile and LREE addition leading to thegrowth of apatite in eclogites from the Jericho kimberlitein the northern Slave craton have been ascribed to twoseparate metasomatic events at around 18Ga (quasi-coincident with the age of formation and metamorphism)and 13^10Ga respectively (Heaman et al 2002 2006

Schmidberger et al 2005) The authors associated theseevents with Paleoproterozoic subduction and aMesoproterozoic thermal disturbance in the Slave litho-spheric mantle possibly linked to the Mackenzie igneousevent Mesoproterozoic Nd model ages mostly between13 and 14Ga were reported for eclogites from thenearby Ekati kimberlites (Jacob et al 2003) Although wedo not observe zircon or apatite in volatile-free eclogitesfrom Lac de Gras and the age constraints obtained fromour data are weak they support a link between theseevents described for other eclogite populations and themetasomatic enrichment observed in our studyThe Sr isotope data in the present study are similar to

those of Schmidberger et al (2007) who suggested that thedepleted-mantle-like values are consistent with derivationof the protoliths from an oceanic mantle source However87Sr86Sr in volatile-free eclogites in the present study isremarkably constant considering their Nd and Hf isotopevariability (Fig 12) and the variability of trace-elementpatterns (Fig 10) Given the evidence for dehydration andmelting in many of the eclogites fractionation of highlyincompatible Rb from Sr and differential ingrowth ofradiogenic Sr would be expectedThis may indicate a rela-tively recent overprint and considering the similarityof 87Sr86Sr to present-day depleted mantle which is signif-icantly lower than 87Sr86Sr of North American kimber-lites a juvenile (asthenospheric) source

PyroxenitesPyroxenites from Lac de Gras often show relationshipsin their major-element contents that are different fromthose observed in the eclogites For example pyroxeniticgarnets show a negative correlation between CaO andNa2O contents whereas these oxides are positively corre-lated for garnets in eclogites (Fig 5) CaO is well correlatedwith C2O3 in pyroxenitic garnets but not in eclogitic gar-nets (Fig 4) Mg-number is positively correlated withYb abundance for pyroxenitic garnets but anti-correlatedfor eclogitic garnets (Fig 14) This suggests that the pyrox-enites in the present study are not related to the eclogitesby fractionation or accumulation processes as proposedfor the origin of some pyroxenites (Foley et al 2003)Pyroxenites have also been interpreted as the products

of crystal segregation in magma channels with alteredoceanic crust as the magma source this would explaintheir Eu anomalies and stable and radiogenic isotopecompositions (Pearson et al 1993 Becker 1996) As thehigh-pressure liquidus phases of most terrestrial basaltsare cpx and garnet (OrsquoHara amp Herzberg 2002) partialmelts of eclogites with basaltic composition wouldcrystallize cpx and garnet but not opx More than half ofthe pyroxenites from Lac de Gras contain opx suggestingthat they probably did not crystallize directly fromeclogite-derived melts

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1868

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Low-Cr websterites and pyroxenites could be crystalli-zation products of mafic melts that interacted with perido-titic mantle thus becoming more refractory The positivecorrelation of Yb and Mg-number in garnet would allowan origin by fractional crystallization from such meltsAlternatively they formed by reaction of silicic (possiblyeclogite-derived) melts with mantle peridotites(eg Kelemen et al 1993 Aulbach et al 2002) This agreeswith the similarity of hypothetical melts in equilibriumwith pyroxenitic garnets in most samples to eclogite-derived melts (Electronic Appendix 12) Niobium concen-trations in all hypothetical melts are much higher than ineclogite-derived melts regardless of whether or not rutileis assumed to be in the residue The same is true for LaCe and Zr concentrations in some samples which mayindicate that NbZr and LREE were added to thepyroxenites some time after their formation as discussedin the previous section for some eclogites Because mini-mum ages for pyroxenitic rutile with an unradiogenicHf isotope composition (calculated by projecting the176Hf177Hf to the model depleted mantle isotopic evolutioncurve assuming zero ingrowth) range to 178066Ga(Aulbach et al in preparation) it appears likely that thepyroxenites were not formed recently (eg during kimber-lite-related or precursory melt migration) and that some ofthe samples are at least MesoproterozoicGarnet pyroxeniteYK1952 consists of schlieren of equili-

brated garnet pyroxenite in a lsquomatrixrsquo of spongy cpxfine cpx^opx intergrowths embayed garnet and meltpatches (Fig 2f) Garnet in the pyroxenite schlieren hasvery low LREE and HFSE abundances Orthopyroxenein the matrix has distinctly higher Mg-number andCr-number and contains more CaO than opx from thepyroxenite schlieren whereas opx rims in the schlierenare compositionally similar to opx in the matrix The con-siderable compositional and textural disequilibrium inthis sample suggests that this is a young feature causedby reaction with fluids shortly before or at the time ofentrainmentAs a consequence of the small sample sizes of most of

the pyroxenites studied here a comprehensive isotopicdataset is not available Clearly more such data areneeded in addition to stable isotope data to obtain betterconstraints on the origins of this rock type Consideringthe microstructural and compositional variability ofthe pyroxenites and major-element relationships oppositeto those observed for eclogites it appears certain that pyr-oxenites form by a variety of processes some of which areunrelated to the formation of eclogites and their protoliths

Volatile-rich eclogitesVolatile-rich eclogites from Lac de Gras display the mostextreme Hf^Nd isotope systematics of all the analysedsamples Garnets in phlogopitized (YK1911) and in cal-cite-bearing eclogite (VR67112b) have the most radiogenic

Nd (eNdfrac14 2486 and 1927 respectively) Garnet inVR67112b has extremely high eHf (2516) Clinopyroxene inYK1911 has the lowest eHf measured for any mineralseparate from Lac de Gras (53) in accord with the veryhigh Hf abundance in its cpx (5 ppm) In contrastgarnet in this sample has radiogenic Hf (eHffrac14 136) Thisdiscrepancy is at least in part due to their relatively lowequilibration temperatures (7698C and 8558C respec-tively) which allowed the isotopic compositions of cpxand garnet to evolve separately at their low andhigh parent^daughter element ratios respectively forextended periods of time and possibly since formationAccordingly the cpx in YK1911 gives a minimum Hf ageof 27GaSlab-derived fluids could have been introduced during

the 27Ga collision of the western Slave Craton domainincluding the thick plume-derived deep lithosphericmantle layer and the eastern domain with its highlydepleted lithosphere This would also be consistent withthe observation that volatile-rich eclogites occur at shallowdepths in the mantle (TKroghfrac14 685^8558C correspondingto depths of 85^115 km along a 40 mWm2 geotherm)where the bulk of dehydration and production of solute-rich melts in a subducting slab might be expected to occur(Becker et al 2000 Hermann et al 2006) Their negativeEu anomalies might indicate that the fluid was derivedfrom the shallow portion of the slab that experiencedplagioclase fractionation The presence of carbonate inVR67112b requires a carbonated sourceHigh 87Sr86Sr in cpx in YK1911 is probably the conse-

quence of isotopic equilibration with high-RbSr phlogo-pite after long-term radiogenic ingrowth while Hfaddition led to retarded ingrowth of radiogenic HfThese systematics combined with radiogenic Nd appear

inconsistent with the involvement of sediments in the pet-rogenesis of this sample (eg Perfit et al 1980 Vervoortet al 1999) or a hydrous fluid such as those responsible forMARID metasomatism in the Kaapvaal craton (Konzettet al 1998)Whatever the origin of the fluids the apparentHf mobility suggests that they were not in equilibriumwith a rutile-bearing source and also that they weresolute-rich in nature (ie hydrous melts able to dissolveHFSE) and the relatively high HREE abundances of thevolatile-rich eclogites indicate a garnet-free source suchas amphibolite This allowed evolution of the sample atrelatively high time-integrated SmNd

SUMMARY AND CONCLUSIONS

(1) The Lac de Gras kimberlites carried abundant eclo-gite and pyroxenite xenoliths to the surface Threemain eclogite types are distinguished in terms ofmajor-element composition and trace element

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1869

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

patterns high-Mg high-Ca and low-Mg eclogitesThe garnets of all eclogite types have anti-correlatedMg-number and Yb pointing to olivine fractionationas a controlling factor in the formation of their pre-cursor rocks Their flat or positively sloped HREEpatterns qualitatively indicate that the melt parentalto the precursor rocks was not in equilibrium with agarnet-bearing residue

(2) Reconstructed high-Mg eclogites and high-Caeclogites have flat HREE patterns positive Pb andSr anomalies and low ZrSm similar to oceanic gab-bros reported in the literature High-Mg eclogiteshave variably low HREE contents and some havesmall positive Eu anomalies possibly suggestive ofmore primitive relatively plagioclase-rich protolithswhereas high-Ca eclogites have higher HREE abun-dances and small or absent Eu anomalies that requiremore cpx- andor melt-rich protoliths if a subductionorigin is assumed

(3) Low-Mg eclogites have distinctive trace-element pat-terns that show a marked positive slope in the HREEand subchondritic ZrHf They resemble someArchaean second-stage melts (lsquoboninitesrsquo) and theirprotoliths appear to have formed from a previouslydepleted source

(4) After formation the eclogites were subject to a varietyof processes Most have low ratios of fluid-mobile vsless fluid-mobile elements possibly as a result of dehy-dration of the protoliths In addition some of the low-Mg and high-Ca eclogites have low LREEHREEconsistent with partial melt lossWhereas the low-Mgeclogites show concomitant enrichment in La Nb andZr some high-Mg and high-Ca eclogites show evi-dence for variable and unsystematic (decoupled)LREE and Nb addition

(5) Isotopically the trace-element characteristics of thedifferent eclogite types translate into lower eNd forhigh-Mg eclogites (negative slope in MREEfrac14 lowerSmNd) than for low-Mg eclogites (positive slope inMREEfrac14higher SmNd) Within the low-Mg groupsamples that show evidence for metasomatism (lowLuHf and SmNd) have lower eNd and eHf than asample that was apparently not metasomatized BothNd and Hf addition can be dated to 17Ga if identi-cal initial isotope compsitions are assumed for thelow-Mg eclogites In contrast the relatively constantand depleted mantle-like 87Sr86Sr of eclogitic cpxdespite variable Nd and Hf isotope compositions andparent^daughter ratios suggests recent input of juve-nile material

(6) Garnet and cpx in pyroxenites show major-elementrelationships different from those in eclogitesTherefore these two rock types are probably notrelated by fractionation processes The presence of

opx precludes direct crystallization from slab-derivedmelts Pyroxenites may be the product of (slab-derived) melts hybridized during assimilation of peri-dotitic mantle

(7) Garnets in phlogopite- and carbonate-bearing eclo-gites have extreme isotopic compositions includingthe most radiogenic Nd and Hf in the datasetNegative Eu anomalies may point to the involvementof a crustal protolith Highly radiogenic Sr radiogenicNd and unradiogenic Hf in the reconstructedphlogopite-bearing sample combined with highHREE abundances and evidence for Hf mobilitymay indicate involvement of solute-rich fluids origi-nating in a garnet- and rutile-free source

ACKNOWLEDGEMENTSHelp in the clean laboratories and with the analyticalfacilities by S Elhlou C Lawson A Sharma P Wielandand M Zhang is greatly appreciated Bill Griffin isthanked for his support and for agreeing to reviewseveral versions of the manuscript despite disagreeingwith many of the conclusions presented here The paperbenefited greatly from the detailed and constructivereviews provided by the editors Gareth Davies andMarjorie Wilson and by Matthias Barth Larry Heamanand an anonymous reviewer This work was funded by aMacquarie University International PostgraduateAward and Postgraduate Research Fund (SA) by anARC SPIRT grant sponsored by Kennecott Canada Incand by ARC Large and Discovery Grants to SYOrsquoRAnalytical data were obtained using instrumentationfunded by ARC LIEF and DEST Systemic InfrastructureGrants industry partners and Macquarie UniversityThis is Publication 481 from the ARC National KeyCentre for Geochemical Evolution and Metallogeny ofContinents (wwwesmqeduauGEMOC)

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online

REFERENCESAdam J GreenT H Sie S H amp Ryan C G (1997) Trace element

partitioning between aqueous fluids silicate melts and mineralsEuropeanJournal of Mineralogy 9 569^584

Aulbach S Stachel T Viljoen K S Brey G P amp Harris J W(2002) Eclogitic and websteritic diamond sources beneath theLimpopo Beltccedilis slab-melting the link Contributions to Mineralogy

and Petrology 143 56^70Aulbach S GriffinW L Pearson N J OrsquoReilly S Y Kivi K amp

Doyle B (2003) Origins of eclogites beneath the central SlaveCraton Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA-0011

Aulbach S Griffin W L Pearson N J OrsquoReilly S Y amp Kivi K(2004a) Mantle formation and evolution Slave Craton constraints

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1870

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

from HSE abundances and Re^Os isotope systematics of sulfideinclusions in mantle xenocrysts Chemical Geology 208 61^88

Aulbach S Griffin W L OrsquoReilly S Y amp McCandless T E(2004b) Genesis and evolution of the lithospheric mantle beneaththe Buffalo HeadTerrane Alberta (Canada) Lithos 77 413^451

Aulbach S GriffinW L Pearson N J OrsquoReilly S Y amp Doyle B(2007) Lithosphere formation in the central Slave Craton(Canada) plume subcretion or lithosphere accretion Contributionsto Mineralogy and Petrology (in press) doi 101007s00410-007-0200-1

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2001)Geochemistry of xenolithic eclogites from West Africa Part IA link between low MgO eclogites and Archean crust formationGeochimica et Cosmochimica Acta 65 1499^1527

Barth M G Foley S F amp Horn I (2002a) Partial melting inArchean subduction zones constraints from experimentally deter-mined trace element partition coefficients between eclogitic miner-als and tonalitic melts under upper mantle conditions PrecambrianResearch 113 323^340

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2002b)Geochemistry of xenolithic eclogites from West Africa part 2Origins of the high MgO eclogites Geochimica et Cosmochimica Acta

66 4325^4345Beard B L Fraracci K NTaylor L A Snyder G A Clayton R

A Mayeda T K amp Sobolev NV (1996) Petrography andgeochemistry of eclogites from the Mir kimberliteYakutia RussiaContributions to Mineralogy and Petrology 125 293^310

Becker H (1996) Geochemistry of garnet peridotite massifs fromlower Austria and the composition of deep lithosphere beneatha Palaeozoic convergent plate margin Chemical Geology 134 49^65

Becker H Jochum K P amp Carlson R W (2000) Trace elementfractionation during dehydration of eclogites from high-pressureterranes and the implications for element fluxes in subductionzones Chemical Geology 163 65^99

Bell K amp Tilton G R (2002) Probing the mantle The story fromcarbonatites EOS Transactions American Geophysical Union 83275^277

Benoit M Polvecurren M amp Ceuleneer G (1996) Trace element andisotopic characterisation of mafic cumulates in a fossil mantlediapir (Oman ophiolite) Chemical Geology 134 199^214

BleekerW (2003) The late Archean record a puzzle in ca 35 piecesLithos 71 99^134

BleekerW Ketchum JW F JacksonV A amp Villeneuve M (1999)The central Slave Basement Complex Part I Its structural topol-ogy and autochthonous core Canadian Journal of Earth Sciences 361083^1109

Blichert-Toft J amp Albare de F (1997) The Lu^Hf isotope geochemis-try of chondrites and the evolution of the mantle^crust systemEarth and Planetary Science Letters 148 243^258

Blichert-Toft J Chauvel C amp Albare de F (1997) Separation of Hfand Lu for high precision isotope analysis of rock samples by mag-netic sector-multiple collector ICP-MS Contributions to Mineralogy

and Petrology 127 248^260Brey G P amp Kolaquo hler T (1990) Geothermobarometry in four-phase

lherzolites II New thermobarometers and practical assessment ofexisting thermobarometers Journal of Petrology 31 1353^1378

Brey G P Kolaquo hlerT amp Nickel K G (1990) Geothermobarometry infour-phase lherzolites I Experimental results from 10 to 60 kbJournal of Petrology 31 1313^1352

Caporuscio F amp Smyth J (1990) Trace element crystal chemistryof mantle eclogites Contributions to Mineralogy and Petrology 105550^561

Carbno G B amp Canil D (2002) Mantle structure beneath the SWSlave Craton Canada constraints from garnet geochemistry in theDrybones Bay Kimberlite Journal of Petrology 43 129^142

Coleman R G Lee E D Beatty L B amp Brannock W W (1965)Eclogites and eclogites their differences and similaritiesGeological Society of America Bulletin 76 483^508

Creaser R A Grulaquo tter H Carlson J amp Crawford B (2004)Macrocrystal phlogopite Rb^Sr dates for the Ekati propertykimberlites Slave Province Canada evidence for multiple intru-sive episodes in the Paleocene and Eocene Lithos 76 399^414

Davies R M GriffinW L Pearson N J Andrew A S Doyle BJ amp OrsquoReilly S Y (1999) Diamonds from the deep Pipe DO-27Slave Craton Canada In Gurney J J Gurney J L Pascoe MD amp Richardson S H (eds) Proceedings of the 7th International

Kimberlite Conference CapeTown Red Roof Design pp 148^155Davies R M Griffin W L OrsquoReilly S Y amp Doyle B J (2004)

Mineral inclusions and geochemical characteristics of microdia-monds from the DO27 A154 A21 A418 DO18 DD17 and RanchLake kimberlites at Lac de Gras Slave Craton Canada Lithos 7739^55

DavisW J amp Hegner E (1992) Neodymium isotopic evidence for thetectonic assembly of Late Archean crust in the Slave Provincenorthwest Canada Contributions to Mineralogy and Petrology 111493^504

Davis W J Jones A G Bleeker W amp Grutter H (2003)Lithosphere development in the Slave craton a linked crustal andmantle perspective Lithos 71 575^589

Dawson J B (1984) Contrasting types of upper-mantle metasoma-tism In Kornprobst J (ed) Kimberlites II The Mantle and

Crust^Mantle Relationships Amsterdam Elsevier pp 289^294DePaolo D J (1981) Neodymium isotopes in the Colorado Front

Range and crust^mantle evolution in the Proterozoic Nature 291193^196

Erlank A J amp Kushiro I (1970) Potassium contents of syntheticpyroxenes at high temperatures and pressures Carnegie Institution ofWashingtonYearbook 68 433^439

Foley S F Barth M G amp Jenner G A (2000) Rutilemelt partitioncoefficients for trace elements and an assessment of the influence ofrutile on the trace element characteristics of subduction zonemagmas Geochimica et Cosmochimica Acta 64 933^938

Foley S F Buhre S amp Jacob D E (2003) Evolution of the Archaeancrust by delamination and shallow subduction Nature 421 249^252

Graham I Burgess J L Bryan D Ravenscroft P J Thomas EDoyle B J Hopkins R amp Armstrong K A (1999) Explorationhistory and geology of the Diavik Kimberlites Lac de GrasNorthwest Territories Canada In Gurney J J Gurney J LPascoe M D amp Richardson S H (eds) Proceedings of the 7th

International Kimberlite Conference Cape Town Red Roof Designpp 262^279

Green D H amp Wallace M E (1988) Mantle metasomatism byephemeral carbonatite melts Nature 336 459^462

Green T H Blundy J D Adam J amp Yaxley G M (2000) SIMSdetermination of trace element partition coefficients betweengarnet clinopyroxene and hydrous basaltic liquids at 2^75 GPaand 1080^12008C Lithos 53 165^187

GriffinW L amp OrsquoReilly SY (2007) Cratonic lithospheric mantle Isanything subducted Episodes 30(1) 43^53

Griffin W L Wass S Y amp Hollis J D (1984) Ultramafic xenolithsfrom Bullenmerri and Gnotuk maars Victoria Australia petrol-ogy of a sub-continental crust^mantle transition Journal of

Petrology 25 53^87GriffinW L Doyle B J Ryan C G Pearson N J OrsquoReilly SY

Davies R Kivi K van Achterbergh E amp Natapov LM (1999)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1871

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Layered mantle lithosphere in the Lac de Gras Area SlaveCraton composition structure and origin Journal of Petrology 40705^727

Griffin W L Pearson N J Belousova E Jackson S E vanAchterbergh E OrsquoReilly SY amp Shee S R (2000)The Hf isotopecomposition of cratonic mantle LAM-MC-ICPMS analysis ofzircon megacrysts in kimberlites Geochimica et Cosmochimica Acta

64 133^147Griffin W L OrsquoReilly S Y Doyle B J Pearson N J

Coopersmith H Kivi K Malkovets V amp Pokhilenko N(2004) Lithosphere mapping beneath the North American plateLithos 77 873^922

Grulaquo tter H S Apter D B amp Kong J (1999) Crust^mantle couplingevidence from mantle-derived xenocrystic garnets In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 307^312

Hart S R Blusztajn J Dick H J B Meyer P S ampMuehlenbachs K (1999) The fingerprint of seawater circulationin a 500-meter section of ocean crust gabbros Geochimica et

Cosmochimica Acta 63 4059^4080Harte B amp Kirkley M B (1997) Partitioning of trace elements

between clinopyroxene and garnet Data from mantle eclogitesChemical Geology 136 1^24

Heaman L M Creaser R A amp Cookenboo H O (2002) Extremeenrichment of high field strength elements in Jericho eclogitexenoliths A cryptic record of Paleoproterozoic subduction partialmelting and metasomatism beneath the Slave craton CanadaGeology 30 507^510

Heaman L M Kjarsgaard B A amp Creaser R A (2004) The tem-poral evolution of North American kimberlites Lithos 76 377^397

Heaman L M Creaser R A Cookenboo H O amp Chacko T(2006) Multi-stage modification of the northern Slave mantlelithosphere Evidence from zircon- and diamond-bearing eclogitexenoliths entrained in Jericho Kimberlite Journal of Petrology 47821^858

Hermann J Cpandler C Hack A amp Korsakov A (2006) Aqueousfluids and hydrous melts in high-pressure and ultra-high pressurerocks implications for element transfer in subduction zones Lithos92 399^417

Hills DV amp Haggerty S E (1989) Petrochemistry of eclogites fromthe Koidu Kimberlite Complex Sierra Leone Contributions to

Mineralogy and Petrology 103 397^422Hoffman P F (1989) Precambrian geology and tectonic history of

North America In Bally A amp Palmer A (eds) The Geology of

North AmericaccedilAn Overview Boulder CO Geological Society ofAmerica pp 447^512

IrelandT R Rudnick R L amp Spetsius Z (1994) Trace-elements indiamond inclusions from eclogites reveal link to Archean granitesEarth and Planetary Science Letters 128 199^213

Irvine G J Pearson D G Kjarsgaard B A Carlson R WKopylova M G amp Dreibus G (2003) Evolution of the litho-spheric mantle beneath Northern Canada a Re^Os isotope andPGE study of kimberlite-derived peridotite xenoliths fromSomerset Island and a comparison to the Slave and Kaapvaalcratons Lithos 71 461^488

Isachsen C E amp Bowring S A (1994) Evolution of the Slave cratonGeology 22 917^920

Jacob D E (2004) Nature and origin of eclogite xenoliths from kim-berlites Lithos 77 295^316

Jacob D E amp Foley S F (1999) Evidence for Archean ocean crustwith low high field strength element signature from diamondifer-ous eclogite xenoliths Lithos 48 317^336

Jacob D Jagoutz E Lowry D Mattey D amp Kudrjavtseva G(1994) Diamondiferous eclogites from Siberiaccedilremnants ofArchean oceanic-crust Geochimica et Cosmochimica Acta 585191^5207

Jacob D E Fung A Jagoutz E amp Pearson D G (2003) Petrologyand geochemistry of eclogite xenoliths from the Ekati kimberlitesarea Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA_239

Jacob D E Bizimis M amp Salters V J M (2005) Lu^Hf and geo-chemical systematics of recycled ancient oceanic crust evidencefrom Roberts Victor eclogites Contributions to Mineralogy and

Petrology 148 707^720Jerde E A Taylor L A Crozaz G Sobolev NV amp SobolevV N

(1993) Diamondiferous eclogites from Yakutia Siberiaccedilevidencefor a diversity of protoliths Contributions to Mineralogy and Petrology

114 189^202Jones A G Ferguson I J Chave A D Evans R L amp

McNeice GW (2001) The electric lithosphere of the Slave cratonGeology 29 423^426

Kelemen P B Shimizu N amp Dunn T (1993) Relative depletionof niobium in some arc magmas and the continental-crust-partitioning of K Nb La and Ce during meltrockreaction in the upper-mantle Earth and Planetary Science Letters 120111^134

Kinzler R J (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition Application to mid-ocean ridge basalt petrogenesis Journal of Geophysical Research 102853^874

Klemme S van der Laan S R Foley S F amp Guenther D (1995)Experimentally determined trace and minor element partitioningbetween clinopyroxene and carbonatite melt under upper mantleconditions Earth and Planetary Science Letters 133 439^448

Konzett J Armstrong R A Sweeney R J amp CompstonW (1998)The timing of MARID metasomatism in the Kaapvaal mantle anion probe study of zircons from MARID xenoliths Earth and

Planetary Science Letters 160 133^145Kopylova M G Russell J K amp Cookenboo H (1999) Mapping the

lithopshere beneath the north central Slave Craton In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 468^479

Krogh E (1988) The garnet^clinopyroxene iron^magnesiumgeothermometerccedila reinterpretation of existing experimentaldata Contributions to Mineralogy and Petrology 99 44^48

Kusky T M (1989) Accretion of Archean Slave Province Geology 1763^67

LeCheminant A N amp Heaman L M (1989) Mackenzie igneousevents Canada Middle Proterozoic hotspot magmatismassociated with ocean opening Earth and Planetary Science Letters

96 38^48LeCheminant A N Richardson D G DiLabio R N W amp

Richardson K A (1996) La recherche de diamants au CanadaGeological Survey of Canada Open File 3228 268pp

MacGregor I D amp Carter J L (1970) The chemistry of clinopyrox-enes and garnets of eclogite and peridotite xenoliths from theRoberts Victor mine South Africa Physics of the Earth and PlanetaryInteriors 3 391^397

MacGregor I D amp Manton W I (1986) Roberts Victor eclogitesancient oceanic crust Journal of Geophysical Research 91 14063^14079

MacKenzie J M amp Canil D (1999) Composition and thermal evo-lution of cratonic mantle beneath the central Archean SlaveProvince NWT Canada Contributions to Mineralogy and Petrology

134 313^324

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1872

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

McCandlessT E amp Gurney J J (1989) Sodium in garnet and potas-sum in clinopyroxene criteria for classifying mantle eclogites InRoss J (ed) Kimberlites and Related Rocks Carlton Blackwellpp 827^832

McDonoughW F amp Sun S-S (1995) The composition of the EarthChemical Geology 120 223^253

Navon O amp Stolper E (1987) Geochemical consequences of meltpercolation the upper mantle as a chromatographic columnJournal of Geology 95 285^307

Nikogosian I K amp Sobolev AV (1997) Ion-microprobe analysis ofmelt inclusions in olivine experience in estimating the olivine^meltpartition coefficients of trace elements Geochemistry International 35119^126

Norman M D Griffin W L Pearson N J Garcia M O ampOrsquoReilly S Y (1998) Quantitative analysis of trace element abun-dances in glasses and minerals a comparison of laser ablationinductively coupled plasma mass spectrometry solution inductivelycoupled plasma mass spectrometry proton microprobe and elec-tron microprobe data Journal of Analytical Atomic Spectrometry 13477^482

Norman M D Garcia M O amp Pietruszka A J (2005) Trace-element distribution coefficients for pyroxenes plagioclase andolivine in evolved tholeiites from the 1955 eruption of KilaueaVolcano Hawairsquoi and petrogenesis of differentiated rift-zonelavas American Mineralogist 90 888^899

OrsquoHara M J amp Herzberg C (2002) Interpretation of trace elementand isotope features of basalts Relevance of field relations petrol-ogy major element data phase equilibria and magma chambermodeling in basalt petrogenesis Geochimica et Cosmochimica Acta 662167^2191

PadghamW A amp FysonW K (1992) The Slave Province a distinctcraton CanadianJournal of Earth Sciences 29 2072^2086

Pearson D G Davies G R amp Nixon P H (1993) Geochemical con-straints on the petrogenesis of diamond facies pyroxenites from theBeni Bousera peridotite massif North Morocco Journal of Petrology34 125^172

Pearson N J Griffin W L Doyle B J OrsquoReilly S Y vanAchterbergh E amp Kivi K (1999) Xenoliths from kimberlitepipes of the Lac de gras area Slave Craton Canada In GurneyJ J Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite ConferenceCapeTown Red Roof Design pp 644^658

Perfit N R Gust D A Bence A E Arculus R J amp Taylor S R(1980) Chemical characteristics of island-arc basalts implicationsfor mantle sources Chemical Geology 30 227^256

Polat A Hofmann A W amp Rosing M T (2002) Boninite-likevolcanic rocks in the 37^38 Ga Isua greenstone belt WestGreenland geochemical evidence for intra-oceanic subductionzone processes in the early Earth Chemical Geology 184 231^254

Poudjom DjomaniY H GriffinW L OrsquoReilly SY amp Doyle B J(2005) Lithospheric domains and controls on kimberlite emplace-ment Slave Province Canada evidence from elastic thickness andupper mantle composition Geochemistry Geophysics Geosystems 6doi1010292005GC0009978

Presnall D C Gudfinnsson G H amp Walter M J (2002)Generation of mid-ocean ridge basalts at pressures from 1 to 7GPa Geochimica et Cosmochimica Acta 66 2073^2090

Schmidberger S S Heaman L M Simonetti A Creaser R A ampCookenboo H O (2005) Formation of Paleoproterozoiceclogitic mantle Slave Province (Canada) Insights from in-situHf and U^Pb isotopic analyses of mantle zircons Earth and

Planetary Science Letters 240 621^633

Schmidberger S S Simonetti A Heaman L M Creaser R A ampWhiteford S (2007) Lu^Hf in-situ Sr and Pb isotope and traceelement systematics for mantle eclogites from the Diavik diamondmine Evidence for Paleoproterozoic subduction beneath the Slavecraton Canada Earth and Planetary Science Letters 254 55^68

Schulze D J amp Helmstaedt H (1988) Coesite-sanidine eclogitesfrom kimberlite products of mantle fractionation or subductionJournal of Geology 96 435^443

Smithies R H Champion D C amp Sun S S (2004) The case forArchaean boninites Contributions to Mineralogy and Petrology 147705^721

Smyth J R Caporuscio F A amp McCormick T C (1989) Mantleeclogites evidence of igneous fractionation in the mantle Earth andPlanetary Science Letters 93 133^141

Sobolev NV amp Lavrentrsquoyev Y G (1971) Isomorphic sodium admix-ture in garnets formed at high pressure Contributions to Mineralogy

and Petrology 31 1^12Stalder R Foley S F Brey G P amp Horn I (1998) Mineral aqueous

fluid partitioning of trace elements at 900^12008C and 30^57GPa New experimental data for garnet clinopyroxene and rutileand implications for mantle metasomatism Geochimica et

Cosmochimica Acta 62 1781^1801Sun S-S amp McDonough W F (1989) Chemical and isotopic sys-

tematics of oceanic basalts implications for mantle compositionsand processes In Saunders A D amp Norris M J (eds)Magmatism in the Ocean Basins Geological Society London Special

Publications 42 313^345Taura H Yurimoto H Kurita K amp Sueno S (1998) Pressure

dependence on partition coefficients for trace elements between oli-vine and the coexisting melts Physics and Chemistry of Minerals 25469^484

Taylor L A amp Neal C R (1989) Eclogites with oceanic crustal andmantle signatures from the Bellsbank kimberlite South AfricaPart I Mineralogy petrography and whole rock chemistryJournal of Geology 97 551^567

van Achterbergh E Ryan C G amp Griffin W L (1999) Glitter online intensity reduction for the laser ablation inductively coupledplasma spectrometry Abstract 9th V M GoldschmidtConference LPI Boston MA 905 pp Abstract no 7215

van Breemen O DavisW J amp King J E (1992) Temporal distribu-tion of granitoid plutonic rocks in the Archean Slave ProvinceNorthwest Canadian Shield Canadian Journal of Earth Sciences 292186^2199

Vervoort J D Patchett P J Blichert-Toft J amp Albare de F (1999)Relationship between Lu^Hf and Sm^Nd isotopic systems in theglobal sedimentary system Earth and Planetary Science Letters 16879^99

Viljoen K S Smith C B amp Sharp Z D (1996) Stable and radio-genic isotope study of eclogite xenoliths from the Orapa kimberliteBotswana Chemical Geology 131 235^255

Walter M J (1998) Melting of garnet peridotite and theorigin of komatiite and depleted lithosphere Journal of Petrology 3929^60

Wasserburg G J Jacobsen S B DePaolo D J McCulloch M Tamp Wen T (1981) Precise determination of SmNd ratios Sm andNd isotopic abundances in standard solutions Geochimica et

Cosmochimica Acta 45 2311^2323Westerlund K J Shirey S B Richardson S H Carlson R W

Gurney J J amp Harris JW (2006) A subduction wedge origin forPaleoarchean peridotitic diamonds and harzburgites from thePanda kimberlite Slave craton evidence from Re^Os isotope sys-tematics Contributions to Mineralogy and Petrology 152 275^294

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1873

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Page 11: Origins of Xenolithic Eclogites and Pyroxenites from the Central

Table 3 Summary of trace-element contents in garnet and cpx (ppm)

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR19674

ecl-g7

VR67112b YK1911 av min max 1s

n 6 6 7 3 5 4 4

P 169 54 270 88 196 70 310 78 282 80 450 138 160 60 80 214 48 370 144

Sc 62 40 89 16 48 36 67 11 53 42 79 15 69 92 93 186 52 440 179

Ti 1667 490 2500 680 1073 620 2000 455 2141 700 4700 1582 270 690 260 1097 540 1900 573

V 169 78 250 68 150 78 280 72 178 110 390 101 140 340 100 390 71 820 326

Co 59 42 70 11 58 39 84 13 58 47 65 5 30 41 27 68 55 90 16

Ni 17 5 32 10 21 11 34 10 24 11 35 8 6 10 51 54 25 110 38

Ga 12 7 15 3 13 8 16 3 12 9 20 4 8 86 58 12 10 14 2

Sr na 512 4 na na 522 522 na na 57 57 na 52 5025 510 513 513 513 na

Y 37 6 61 18 21 7 38 10 31 20 38 5 60 81 70 19 12 26 6

Zr 15 5 29 9 28 6 83 26 24 9 43 12 60 22 90 18 1 41 17

Nb 009 0005 048 019 na 50015 05 na na 5005 05 na 5003 0012 5001 022 0030 057 025

La 005 001 009 004 na 50007 00 na na 5003 04 na 004 002 0009 003 000 009 005

Ce 03 01 05 02 01 00 01 00 05 01 18 05 03 022 015 01 00 03 01

Pr 01 00 02 01 00 00 01 00 02 01 03 00 017 014 012 00 00 01 00

Nd 12 03 24 07 07 04 14 03 21 12 32 07 24 23 20 03 00 07 03

Sm 11 04 22 06 10 03 18 05 23 08 36 10 29 31 33 05 01 09 04

Eu 06 03 10 03 06 02 10 03 11 04 17 04 16 114 16 03 01 06 02

Gd 27 07 41 12 23 07 42 12 43 16 53 13 8 81 90 13 06 24 08

Dy 54 11 70 23 36 13 65 18 58 35 71 11 12 144 12 29 18 42 10

Ho 14 02 23 07 08 03 15 04 12 09 14 02 24 30 23 07 04 10 02

Er 46 07 80 24 24 09 40 10 34 25 40 05 8 85 6 22 14 28 06

Yb 52 06 98 30 24 10 37 08 32 22 39 05 8 70 5 23 13 30 07

Lu 09 01 16 05 04 02 05 01 05 03 07 01 11 09 08 04 02 05 01

Hf 03 02 05 01 04 01 10 03 05 02 09 02 09 027 007 03 00 06 03

Ta na 50008 5004 na na 50005 001 na na 50004 na na 003 50002 50005 na 50007 011 na

Pb na 5005 02 na na 5002 02 na na 5003 07 na 02 004 5002 na 5003 01 na

Th na 5001 005 na na 50007 003 na na 50008 011 na 5007 0012 50004 002 0005 006 002

U na 5003 5003 na na 5001 01 na na 5007 na na 01 0023 010 na 5001 003 na

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR67112b YK1911 av min max 1s

n 6 4 8 5 4 4

P 65 15 120 41 39 14 56 16 53 30 70 12 20 9 80 27 120 40

Sc 22 10 30 7 15 14 19 2 14 7 24 5 80 43 17 11 21 5

Ti 1557 990 2400 508 1168 1010 1490 191 1988 810 3500 1154 2390 1700 689 102 1685 694

V 415 180 480 107 318 210 400 69 324 259 450 72 600 180 250 139 480 156

Co 35 26 50 8 20 14 31 7 26 14 42 11 49 23 29 20 42 10

Ni 164 87 220 47 228 100 430 128 188 135 320 56 59 34 428 410 670 184

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1853

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

ClinopyroxeneTrace-element abundances in cpx are summarized inTable 3 and the complete dataset is reported in ElectronicAppendix 6 Clinopyroxenes in low-Mg eclogites havemildly positive to negative slopes from LaN to NdN andsmooth negative slopes between SmN and HoN thatmostly continue to LuN (Fig 8a) Niobium is below detec-tion or 5007 chondritic and SrN is variably enrichedrelative to neighbouring elements with the peak size corre-lating negatively with LREEN ZrHf is subchondritic andHfN is enriched relative to SmN whereasTiN is enriched ordepleted relative to similarly compatible elementsWe ana-lysed the spongy rim in cpxYK1943 and its trace-elementpattern is similar to that of cpx cores for more compatibleelements whereas the most incompatible elements areseveral orders of magnitude higher Clinopyroxenes in

high-Mg eclogites have steeper negative slopes betweenNdN and LuN than those in low-Mg eclogites combinedwith lower HREE abundances HfN is depleted relative toSmN (Fig 8b) High-Ca eclogites contain cpx with sinusoi-dal REEN patterns with lower LREE than other eclogitetypes and the heaviest REE below detection They haveuniformly strong positive Sr peaks relative to elements ofsimilar compatibility (Fig 9) The diamond-bearing high-Ca eclogite (VR40345) has cpx with unusually high Nband LREE abundancesClinopyroxenes in four pyroxenites have disparate trace

element patterns They all share depletions in TiN relativeto similarly compatible elements and two show positiveSrN peaks Two samples have cpx with high and flatLREEN to MREEN patterns and steep negative slopes inthe HREEN The trace-element patterns of cpx in YK1914

Table 3 Continued

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR67112b YK1911 av min max 1s

n 6 4 8 5 4 4

Ga 16 10 20 3 12 9 15 2 26 15 39 9 129 123 6 11 11 5

Rb na 5008 04 na na 5007 05 02 na 5003 03 na 02 5004 03 010 05 0

Sr 166 85 290 74 222 153 285 50 243 83 440 142 85 120 200 11 414 209

Y 2 04 4 10 13 07 2 04 11 03 3 10 28 51 2 03 3 2

Zr 19 4 48 15 25 8 45 13 17 8 28 7 111 129 36 008 81 42

Nb 01 0003 03 01 na 5002 00 na na 5002 16 9 007 011 010 003 02 008

Cs na 5001 6 na na 5006 5006 na5013 5004 5013 na bdl 5002 na 5001 5002 na

Ba 09 02 2 07 09 01 2 07 10 03 3 10 06 01 13 06 3 10

La 3 03 8 3 3 23 3 04 4 01 14 6 51 97 4 01 10 4

Ce 7 13 20 6 10 7 13 2 7 05 24 10 25 28 11 01 28 13

Pr 11 02 3 08 2 10 3 06 08 02 2 09 47 46 2 002 4 2

Nd 5 12 12 4 10 5 14 4 4 09 8 26 23 22 8 01 22 10

Sm 11 05 21 06 20 10 31 10 07 02 11 03 47 53 2 005 5 2

Eu 03 01 05 01 06 028 09 03 02 010 03 007 087 123 06 002 15 07

Gd 08 03 13 03 11 05 19 06 05 03 06 011 28 37 15 010 36 2

Dy 05 01 07 02 04 022 06 01 02 013 03 007 11 16 07 008 14 06

Ho 01 00 02 004 006 003 007 002 na 5003 012 na 011 021 009 001 02 007

Er na 507 507 na 011 004 02 00 na 5006 03 na 019 033 02 002 03 012

Yb na 5001 02 na na 5001 011 na na 5006 03 na 007 015 009 001 02 006

Lu na 5001 003 na na50002 002 na na50008 01 na 0006 0013 0013 0002 003 001

Hf 12 03 24 07 11 04 18 05 10 06 16 04 34 53 na50015 35 na

Ta na50003 50015 na na50006 006 na na50004 5004 na 0014 00236 na50003 503 na

Pb 06 02 12 04 06 04 09 02 na 5016 07 02 065 098 08 01 18 1

Th na50005 06 na 012 005 02 005 na 5002 5006 na 021 02 001 06 03

U na 5005 5005 na 002 0009 005 001 na 5003 5005 na 0013 024 na50005 003 na

Averaged for the three eclogite types and for pyroxenites as in Table 2 pyx pyroxenite n number of analyses averagedbdl below detection limit

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1854

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

and YK1952 are similar although at different abundanceswith only modest LREEN over HREEN enrichment andstrong depletions in ZrN relative to neighbouring elements(Fig 8d)Clinopyroxenes in the volatile-rich samples have flat pat-

terns and high abundances from LaN to SmN and smoothnegative slopes from SmN to LuN Niobium Sr and Ti are

noticeably depleted relative to elements of similar compat-ibility (Fig 8e)

Trace-element partitioningTo assess whether coexisting phases are in trace-elementequilibrium we calculated apparent cpx^garnet distribu-tion coefficients and compared them with those

(a)

(d)

(b)

(e)

(c)

GarnetCI-chondrite GarnetCI-chondrite

low-Mghigh-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1952

1

10

Nd Sm Eu Gd Dy

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Fig 7 Extended trace-element patterns of garnet in (a) low-Mg eclogites (b) high-Mg eclogites with inset showing enlargement of MREE tohighlight Eu anomalies and 1s uncertainties (c) high-Ca eclogites (d) pyroxenites and (e) volatile-bearing eclogites Normalized to chondritevalues of McDonough amp Sun (1995)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1855

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

determined experimentally or in demonstrably equili-brated natural samples Low-Mg and high-Mg eclogiteshave Dcpxgarnet for most elements that are within therange determined for natural equilibrated eclogites andexperimental assemblages (Harte amp Kirkley 1997 Greenet al 2000 Barth et al 2002a) (Electronic Appendix 7)High-Ca eclogites plot at the low end of the range

of distribution coefficients and have very low DMREE

and DHREE outside the range indicated by the datasetof Harte amp Kirkley (1997) who noted that DREE decreaseswith increasing Ca content in eclogitic minerals Theseassemblages are therefore considered to be in equilibriumPyroxenite VR40384 has D values similar to eclogites

but at the high end of the range determined by Harte amp

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

(a)

(d)

(b)

(e)

(c)

CpxCI-chondrite CpxCI-chondrite

low-Mg high-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1914

YK1952

spongy rim

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

VR40345

Fig 8 Extended trace-element patterns of cpx as in Fig 7

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1856

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Kirkley (1997) (Electronic Appendix 7) Another group(YK1914 and YK1952) plots at the low end but DHf DZrDTi and DV are far below the values indicated by experi-mental datasets Because pyroxenites compositionallygrade into eclogites their drastically different DHFSE

values suggest disequilibrium In contrast volatile-bearingsamples have DHFSE values that appear to be too high to bein equilibrium however this may be a consequence oftheir unusual bulk chemical composition (Ti- and K-rich)which differs substantially from that of lsquonormalrsquo eclogitesand experimental charges

GEOTHERMOBAROMETRYFollowing Schmidberger et al (2007) temperatures oflast equilibration were calculated for eclogitic andopx-free pyroxenitic assemblages using the formulationof Krogh (1988) The results are summarized in Table 2and Electronic Appendix 8 for a preset pressure of5 GPa and illustrated as a histogram in Fig 9 Foropx-bearing pyroxenites temperatures were calculatedsimultaneously with pressure [Ca-in-opx thermometryand Al-in-opx barometry using the formulations of Breyet al (1990)] to constrain their relative positions in themantle columnMinerals in some eclogites from Lac de Gras show dis-

tinct core and rim compositions or inhomogeneitiesTemperatures were not calculated for samples withspatially irresolvable inhomogeneous garnet compositions(SE01VR43479VR43480) For eclogites with spongy cpxcore compositions were used in the calculationTemperatures were calculated for both core and rim com-positions where applicable For all samples these

temperatures agree within the uncertainty of this thermo-meter (508C Brey amp Kolaquo hler 1990)Most equilibration temperatures for eclogites lie between

800 and10008C (Fig 9)The transition between the shallowand the deep lithospheric mantle layers in the central Lacde Gras area lies in this temperature interval (Griffin et al1999 Pearson et al 1999) Eclogites in this study lack thedistinct bimodal temperature distribution observed bySchmidberger et al (2007) but as in their study high-Caeclogites give higher average temperatures (10808C corre-sponding to a depth of 155 km if equilibrated along a 40mWm2 geotherm) than other eclogites types (9808Cdepth 135 km) the diamond-bearing sample (VR40345)gives thehighest temperature of the datasetHighNa2Ocon-tents in garnet from high-Ca eclogites are qualitativelyconsistent with higher equilibration pressures (Sobolev ampLavrentrsquoyev 1971) as are high K2O contents in coexistingcpx (Erlank ampKushiro1970) suggesting that high-Ca eclo-gites not only equilibrated at high temperature but also highpressure rather than representing a temperature excursionPyroxenites both opx-bearing and opx-free give somewhatlower average temperatures than eclogites (mean 9258C)and phlogopite-rich eclogites give the lowest temperatures(685 and 7698C respectively) The calcite-bearing eclogiteVR67112b gives a relatively low equilibration temperatureof 8558C

MEASURED ANDRECONSTRUCTEDWHOLE -ROCKSRationale and approachKimberlite-borne xenoliths sometimes show evidence forinfiltration of the host kimberlite (eg Taylor amp Neal1989 Barth et al 2001 2002b Heaman et al 2006) Wetherefore follow the standard approach of calculatingwhole-rock compositions from the compositions of the con-stituent phases weighted by their modal abundances Onlycore compositions were used as distinct rim compositionsmay point to late diffusion of elements from kimberlite-related melt veinsMany eclogitic whole-rock compositions reconstructed

with only garnet and cpx have negative Ti anomalies(not shown) For most of these samples rutile or ilmenitehas been identified as part of the mineral assemblage andthese phases should therefore be considered in the whole-rock reconstruction Rutile or ilmenite modes were esti-mated by assuming that the bulk-rock has no Ti anomalywhen normalized to a normal mid-ocean ridge basalt(N-MORB) composition and calculating the amount ofrutile required to erase theTi anomalyIt is of course possible that the eclogites had negative

or positive Ti anomalies relative to elements of similarcompatibility (Eu and Dy) Because we reconstructed

0

2

4

6

8

10

12volatile-rich

pyroxenite opx-bearing

pyroxenite opx-free

high-Ca

high-Mg

low-Mg

700750

800850

900950

10001050

11001150

12001250

Fre

quen

cy

Temperature

Fig 9 Histogram ofTKrogh (Krogh 1988) obtained at a preset pres-sure of 5 GPa except opx-bearing pyroxenites where TBKN wassolved simultaneously with PBKN (Brey amp Kolaquo hler 1990)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1857

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

whole-rock compositions from mineral core compositionskimberlite infiltration is unlikely to have caused fractiona-tion of Ti from Eu and Dy In a study in which rutilemodes were estimated from the differences between Ti inmeasured bulk-rocks and in bulk-rocks reconstructed withonly garnet and cpx only three of 19 eclogites recon-structed with rutile show Ti anomalies (Barth et al2002a) Even eclogite xenoliths with depleted Zr and Hfabundances that resemble those of modern island arcbasalts do not show Ti anomalies (Jacob amp Foley 1999)We therefore consider our approach to be a reasonablyrobust means of obtaining meaningful rutile modes wheresmall sample sizes preclude reliable determination of themodes of accessory phases by point-countingWhere rutilewas not recognized on thin-section scale or not analysedthe average eclogitic rutile composition was used for thereconstruction

UncertaintiesReconstructing whole-rock compositions in this way is arather crude method given the coarse grain size of theeclogites relative to the sample size and the non-uniformdistribution of the principal minerals on thin-sectionscale The uncertainty in mineral modes of medium- tocoarse-grained eclogites with grains45mm has been esti-mated by Jerde et al (1993) to be 10 plus uncertaintiesrelated to partial alteration which is minor in our samplesNevertheless the whole-rock trace-element budget israther insensitive to the exact modes for common eclogiticassemblages and REE patterns which encompass mostother elements in terms of compatibility in garnet andcpx do not change significantly even for modal variationsof 30 (Jerde et al 1993) This is not true for major-element concentrations which vary substantially with var-iations in mineral modes and we therefore do not calculatewhole-rock major-element contentsBecause the discussion that follows is based on trace-

element patterns and trends rather than absolute abun-dances whole-rock reconstruction based on mineralmodes is adequate We are confident that the uncertaintyinherent in this approach is not leading to significant mis-representation of eclogite whole-rock compositions becausethe trace-element patterns of the reconstructed whole-rocks show distinct differences (see below) that correlatewith the assignment to eclogite types based on garnetmajor-element composition and this is independent ofmineral mode Following Jerde et al (1993) in ElectronicAppendix 9 we show variations in trace-element patternfor one sample in each eclogite type for variations ingarnet and cpx abundance from 7030 to 3070 vol This shows that the trace-element patterns are generallyinsensitive to cpxgt ratios with regard to slopes andanomalies such as positive Sr Pb and Eu anomalies withthe exception of the positive Eu anomaly in high-Mgeclogite VR43479 which disappears if a high cpx mode is

assumed (not shown) Variations in rutile modes by 50(Electronic Appendix 9d) lead as would be expected tothe appearance of Ti anomalies but a rutile mode of 05instead of 1 vol does not erase the positive Nb anomalyin low-Mg eclogiteVR50909Because of the robustness of the relative trace-element

abundances in the reconstructed eclogites a garnet propor-tion of 065 and cpx proportion of 035 approximately themedian modal abundances of these minerals in volatile-free eclogites from Lac de Gras are assumed for eclogitesamples that are too small for determination of meaningfulmodes by point-counting

Within- and between-group variationsThe trace-element compositions of reconstructed whole-rocks are given in Table 4 N-MORB-normalized trace-element patterns of reconstructed whole-rocks areshown in Fig 10 The three eclogite groups have distincttrace element patterns that are particularly obvious forthe REE (Fig 11) Low-Mg eclogites have positive Pb andSr anomalies and smooth positive slopes in their MREEN

and HREEN LREEN are enriched relative to MREEN inthree samples (Fig 11a) Three samples have high Nbrelative to similarly compatible elements and comparedwith most other eclogites (Fig 10a) All high-Mg eclogitesdisplay more or less pronounced positive Eu and Sranomalies and most have positive Pb and negative Zrand P anomalies Their N-MORB-normalized REE pat-terns show shallow positive slopes from LaN to PrN orNdNand aweak negative slope from NdN to LuN (Fig11b)VR43479 is a low-Mg eclogite with regard to its garnet

Ca^Fe^Mg composition but has a trace-element patterntypical of high-Mg eclogites and will be discussed withthe latter Conversely sample YK1949 a high-Mg eclogitein terms of its garnet Ca^Fe^Mg compostion is more simi-lar to low-Mg eclogites in terms of its whole-rock trace-element pattern and is plotted in Figs 10a and 11aWhereas the patterns of most samples within the groupresemble each other the N-MORB-normalized incompati-ble-element abundances of the high-Ca eclogites vary bymore than an order of magnitude (Fig 10c) All sampleshave positive Pb anomalies and most have positive Sranomalies Some samples have negative P andor Zranomalies Niobium abundances in three of the samplesare very high similar to those for some low-Mg eclogitesHigh-Ca eclogites tend to have flatter MREEN andHREEN patterns and more uniform HREE abundancesthat the other eclogite types with humps between SmN

and HoNThe phlogopitized sample YK1911 shows a strong posi-

tive Pb anomaly and lower HREE abundances than thecalcite-bearing sample VR67112b which shows a distinctnegative Eu anomaly and a depletion of La relative to Ce(Figs 10d and 11d)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1858

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Trace-element compositions of reconstructed whole-rocks (ppm)

Sample P Sc Ti V Co Ni Ga Sr Y Zr Nb La Ce Pr Nd

Low-Mg eclogites

vr19674-g4 186 50 4649 317 57 59 144 77 24 30 36 13 31 05 24

vr43452 154 39 2253 328 49 84 176 52 32 11 01 02 07 01 08

vr43469 38 65 4377 349 54 91 142 34 28 15 02 05 15 03 17

yk1926 144 45 3406 258 52 87 145 89 18 18 20 09 29 06 33

yk1943 173 46 5732 295 41 39 146 122 25 48 82 34 85 13 64

yk19491 102 34 1693 158 31 49 100 46 10 14 03 09 21 03 17

High-Mg eclogites

vr19677-g2 138 39 5160 291 41 53 119 98 14 22 04 13 53 11 64

vr434792 46 25 1265 129 36 126 84 56 3 7 04 06 25 04 23

vr67360 49 36 1573 138 64 184 111 76 5 9 04 09 38 07 33

yk1946 153 28 4877 192 45 105 124 108 12 48 03 10 38 08 50

High-Ca eclogites

vr19674-g5 267 30 7663 200 41 67 195 32 23 43 77 01 04 02 16

vr40345 0 33 2821 175 47 98 210 179 13 13 66 60 109 11 38

vr43465 142 27 5104 220 31 83 121 197 14 35 29 05 22 05 40

vr50858 295 30 6898 201 44 94 150 48 22 38 06 01 03 02 20

vr50860 68 57 5672 418 57 71 186 109 19 11 02 02 13 03 26

vr50909 238 28 7295 201 42 92 171 48 20 57 13 01 04 02 21

yk3528 91 51 3712 341 53 155 177 49 16 9 01 02 08 02 11

Volatile-rich eclogites

vr67112b 40 84 9242 521 46 30 107 4 40 66 31 26 126 24 125

yk1911 31 39 9428 286 44 82 164 3 23 52 64 18 52 09 46

Pyroxenites

yk1914 111 49 707 207 48 547 44 31 4 3 02 06 19 03 10

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

Low-Mg eclogites

vr19674-g4 11 05 10 22 39 10 29 33 06 12 05 028 24 008

vr43452 04 02 08 13 38 12 41 49 08 05 01 nd 02 na

vr43469 11 05 09 25 44 11 34 36 06 07 02 001 09 129

yk1926 12 05 11 19 30 07 22 24 04 07 03 004 23 090

yk1943 21 08 10 29 42 10 30 32 05 16 05 013 65 017

yk19491 06 03 11 08 14 04 14 18 03 04 02 003 29 115

High-Mg eclogites

vr19677-g2 20 08 12 21 24 05 15 15 03 05 05 007 52 6811

vr434792 05 02 12 05 06 01 04 03 00 03 01 001 98 037

vr67360 06 03 14 06 09 02 06 06 01 02 02 002 86 000

yk1946 18 07 12 21 22 05 13 12 02 11 03 006 56 112

High-Ca eclogites

vr19674-g5 17 09 11 33 43 09 24 23 04 16 05 007 02 011

vr40345 09 04 10 12 22 06 16 14 02 05 nd nd 298 na

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1859

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Continued

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

vr43465 22 09 12 25 26 06 15 14 02 10 04 001 25 040

vr50858 20 08 09 33 41 08 22 21 03 10 01 nd 02 na

vr50860 14 07 11 27 37 08 22 18 03 06 00 nd 06 na

vr50909 20 10 12 33 38 07 20 20 03 16 01 002 02 181

yk3528 08 05 11 20 30 07 18 18 02 05 01 nd 08 na

Volatile-rich eclogites

vr67112b 38 10 07 53 74 15 41 34 04 18 03 012 60 004

yk1911 20 07 08 35 39 07 18 17 02 10 14 nd 73 na

Pyroxenites

yk1914 02 01 11 04 08 02 05 06 01 01 03 003 68 101

aGarnet has CandashMgndashFe relationship typical of high-Mg eclogites but whole-rock trace-element pattern is typical oflow-Mg eclogitesbGarnet has CandashMgndashFe relationship typical of low-Mg eclogites but whole-rock trace-element pattern is typical ofhigh-Mg eclogites

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

(a)

001

01

1

10(c)

001

01

10(b)

VR43452

YK1949

RockNMORB

low-Mg

RockNMORB

high-Mg

(d)

volatile-richhigh-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

phlogopite-bearingYK1911

calcite-bearingVR67112b

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 10 Extended trace-element patterns of reconstructed (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-bearing eclogites Normalized to N-MORB of Sun amp McDonough (1989)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1860

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

RockNMORB

(b)

RockNMORB

(d)

010

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu001

01

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

(a)

001

01

1

10(b)

001

01

1

10(c)

001

01

1

10(e)

high-TSchmidberger et al

low-TSchmidberger et al

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

high-Mglow-Mg

high-Cavolatile-rich

phlogopite-bearingYK1911

calcite-bearingVR67112b

YK1949VR43479

Fig 11 N-MORB normalized REE patterns for (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-richeclogites Shown for comparison in (e) are reconstructed eclogites from Schmidberger et al (2007)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1861

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Both volatile-rich whole rocks display strong negativeSr anomalies Strontium can be sited in calcite or apatiteandor indicate the presence of a Sr-rich mineral that wasnot identified in thin section such as bariteFour of the six reconstructed whole-rocks in the study

of Schmidberger et al (2007) (Fig 11e) show markedLREEHREEN enrichment that is observed only in onehigh-Ca eclogite in our study (Fig 11c) suggesting thatmany of the eclogites they studied were affected by metaso-matic LREE enrichment Two samples in their study thatare not affected by this enrichment show positive Euanomalies similar to the high-Mg eclogites in this studywhereas the distinctive REE pattern of the low-Mg eclo-gites is not recognized among their samples

SR^ND^HF I SOTOPESYSTEMATICSStrontium Nd and Hf isotope data for optically puremineral separates and reconstructed whole-rocks aregiven in Table 5 Spongy rims and alteration are likely tobe mostly removed during leaching in 6N HCl in combina-tion with repeated ultrasonication Despite this rigoroustreatment the reproducibility of Nd in garnet VR43469and in cpx YK1914 is worse than for other samples Thismay be due to inhomogeneities related to late Nd diffusionpossibly from grain boundaries containing a kimberlitecomponent The reproducibility of the 176Hf177Hf data isgenerally poorer than that of Nd which may reflect a com-bination of matrix effects caused by imperfectly purifiedsolutions and sample heterogeneity Parent^daughterratios are available only from in situ analyses which arenot precise enough to allow meaningful model age orinitial isotope ratio calculations (ie at the time of kimber-lite eruption) Because the Lac de Gras kimberlites arerelatively young (55^56 Ma Graham et al 1999) age cor-rection would result in only a minor change in isotopecomposition which is inconsequential in view of the rangeof compositions displayed by most samples and hence hasno effect on the conclusions drawnClinopyroxenes in five volatile-free eclogites and a

pyroxenite have Sr isotope ratios ranging from 0701662 to0703546 this range is similar to or somewhat lower thanthat of the depleted mantle at the time of kimberliteemplacement (0703 Bell amp Tilton 2002) Clinopyroxenein the phlogopitized sample has a distinctly high 87Sr86Srof 0751186Clinopyroxene controls 56^92 of the Nd budget (Nd

concentration in cpx weighted by mode) Neodymiumisotope ratios range from 0511660 to 0514844 Thiscorresponds to eNd variation (per 10 000 deviationof 143Nd144Nd in the sample from 143Nd144Nd in the chon-dritic mantle model reservoir) from 17 to thorn43 cpx inthe low-Mg eclogites have higher eNd than those in

high-Mg eclogites Garnets have generally have higher143Nd144Nd from 0512984 (eNdfrac14thorn 7) to 0520965(eNdfrac14thorn 162) Garnets in the two volatile-rich sampleshave the highest ratios 0522790 for garnet in the calcite-bearing sample VR67112b (eNdfrac14thorn198) and 0525721 inthe phlogopitized eclogiteYK1911 (eNdfrac14thorn 255)Garnet contributes 18^32 and cpx 48^78 of the Hf

budget The proportion controlled by rutile is 4^30(Aulbach et al in preparation) Measured 176Hf177Hf incpx ranges from 0281263 in phlogopite-bearing sampleYK1911 (eHffrac1453) to 0286130 in low-Mg eclogiteVR43469 (eHffrac14thorn119) and in garnet from 0283958 inpyroxeniteYK1915 (eHffrac14thorn42) to 0295718 in low-Mg eclo-gite YK1926 (eHffrac14thorn458) Hafnium in garnet in the phlo-gopite^apatite-bearing sample is extremely radiogenic(176Hf177Hffrac14 0354367) corresponding to an eHf of thorn2532which is close to a value determined for a garnet separatefrom a Roberts Victor eclogite (eHffrac14thorn2561 Jacob et al2005) Garnet in the calcite-bearing sample has extremelyradiogenic Nd and Hf (eNdfrac14thorn193 eHffrac14thorn2516) Garnetin pyroxenite YK1915 also has radiogenic Nd and Hf(eNdfrac14thorn27 eHffrac14thorn42)Despite the large uncertainties for reconstructed whole-

rocks (calculated as the square root of the sum of squareduncertainties of two standard errors for garnet and cpxanalyses combined with 20 propagated uncertainty onthe proportion of Nd or Hf respectively contributed byeach mineral) the isotopic data are consistent in thatthey correlate with the independently determinedparent^daughter ratios (1) Low-Mg eclogites have higherSmNd reflected in the steeper positive slope of theirN-MORB-normalized REE patterns (Fig 11a) comparedwith the high-Mg eclogites (Fig 11b) Accordingly recon-structed low-Mg eclogites have higher present-day eNd

(0 to thorn47) than high-Mg eclogites (4 and 17) (2) Low-Mg eclogite YK1943 has very high Nb and LREE abun-dances that could have been added during metasomatismthat also enhanced its Hf abundance This sample has thelowest eNd and eHf of the low-Mg eclogites (0 and thorn20respectively) In contrast apparently unenriched low-Mgeclogites have more radiogenic Nd and Hf (eNdfrac14thorn47and thorn9 eHffrac14thorn128 and thorn113 respectively)The Sr^Nd isotope diagram reveals that eNd in cpx

varies from 19 to 43 at nearly constant depleted-mantle-like 87Sr86Sr (Fig 12a) In the Nd^Hf isotope diagram(Fig 12b) the reconstructed whole-rocks lie on curvedmixing lines if the lsquomixingrsquo proportions of Nd and Hffor the two lsquoend-memberrsquo minerals are very different(eg YK1911) Garnets have more radiogenic Nd andHf than coexisting cpx All reconstructed volatile-freeeclogites plot in the radiogenic field with regard to Hfor are isotopically similar to the depleted mantle Twohigh-Mg eclogites and one low-Mg eclogite have unradio-genic Nd one high-Mg eclogite has radiogenic Nd and

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1862

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 5 Individual and averaged Sr Nd and Hf isotope ratios obtained from garnet and cpx solutions

Sample 87Sr86Sr 2SE2SD 143Nd144Nd 2SE2SD eNd 2SE2SD 176Hf177Hf 2SE2SD eHf 2SE2SD

Low-Mg eclogites

VR43469 cp 0702661 0000010 0514844 0000006 43 01 0286130 0000012 119 04

VR43469 gt 0515508 0000014 56 03 0286506 0000012 132 04

VR43469 gt r 0515453 0000010 55 02 0286548 0000011 134 04

VR43469 gt r 0515384 0000007 54 01 0287244 0000010 158 04

VR43469 gt r 0515487 0000014 56 03 0286996 0000014 149 05

VR43469avg gt 0515458 0000109 55 21 0286823 0000715 143 253

WR VR43469 0515059 0000273 47 5 0286405 0001079 128 38

YK1926 cp 0703546 0000011 0513043 0000007 8 0 0283154 0000009 13 03

YK1926 gt 0513364 0000008 14 01 0295718 0000005 458 02

WR YK1926 0513120 0000026 9 1 0285980 0000716 113 25

YK1943 cp 0512143 0000007 10 01 0282213 0000005 20 02

YK1943 gt 0514461 0000012 36 02 0287510 0000008 168 03

YK1943 gt r 0514491 0000009 36 02 0287332 0000005 161 02

YK1943 gt r 0514517 0000010 37 02 0287290 0000007 160 02

YK1943avg gt 0514489 0000056 36 11 0287377 0000233 163 82

WR YK1943 0512625 0000228 0 5 0283138 0000879 13 31

High-Mg eclogites

VR43480 gt 0520965 0000017 162 03 0284950 0000010 77 04

VR67360 cp 0283509 0000074 26 26

VR67360 cp r 0283433 0000031 23 11

VR67360avg cp 0701662 0000018 0511660 0000010 19 02 0283471 0000108 25 38

VR67360 gt 0512984 0000101 7 20 0287429 0000180 165 64

WR VR67360 0511761 0000132 17 3 0284734 0000469 69 17

YK1949 cp 0703019 0000013 0512231 0000009 8 02 0282513 0000009 9 03

YK1949 gt 0513511 0000010 17 02 0290297 0000042 266 15

YK1949 gt r 0513508 0000012 17 02 0289947 0000015 254 05

YK1949 gt r 0513552 0000014 18 03

YK1949avg gt 0513523 0000049 17 10 0291569 0000494 311 175

WR YK1949 0512446 0000103 4 2 0284622 0000847 65 30

Volatile-rich eclogites

VR67112b gt 0522790 0000008 198 02 0354367 0000036 2532 13

YK1911 cp 0751186 0000009 0512165 0000007 9 01 0281263 0000009 53 03

YK1911 gt 0525721 0000013 255 03 0288051 0000012 187 04

WR YK1911 0513940 0000342 25 7 0281414 0000018 48 1

Pyroxenites

YK1914 cp 0703015 0000034 0513283 0000041 13 08

YK1914 cp r 0702956 0000005 0513200 0000027 11 05

YK1914 cp r 0513527 0000059 17 11

YK1914 cp r 0513045 0000010 8 02

YK1914avg cp 0702986 0000083 0513264 0000402 12 78

YK1915 gt 0514003 0000011 27 02 0283958 0000021 42

Analyses of full procedural replicates are designated with an lsquorrsquo after sample name SE standard error (internal errorsingle analyses) SD standard deviation (external error averaged multiple analyses) Italics used for reconstructed whole-rocks (mineral isotopic compositions and concentrations weighted by modes) Uncertainties for reconstructed whole-rocksinclude 2SE or 2SD uncertainties on each mineralrsquos isotope composition combined with 20 uncertainty in the amountof Nd or Hf contributed by the constituent minerals eNd and eHf are per 10 000 deviation from model chondritic mantle[model of Wasserburg et al (1981) and Blichert-Toft amp Albarede (1997) respectively]

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1863

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

εHf

εNd

Ekati eclogitesminus100

0

100

200

300

400

500

minus40 0 40 80 120 160 200 240 280

gt in YK1915

gt in VR67112bεNd=198 εHf=2532

minus40

0

40

80

120

160

minus40 minus20 0 20 40 60

minus60

minus40

minus20

0

20

40

070 071 072 073 074 075 076

North Americankimberlite field

εNd

87Sr86Sr

cpx in YK1914

Ekatieclogites

DM

DM

DM

DM

(a)

(b)

pyroxenites

low-Mg

high-Mg

volatile-rich

eclogites

Diavik low-TA154 S kimberlite

cpx in YK1911

Fig 12 (a) Sr^Nd for cpx and (b) Nd-Hf isotope diagram for garnet cpx and reconstructed whole-rocks (symbols with bold outline) Finedashed lines in (b) connect data points from the same sample with garnets having more radiogenic isotopic compositions Uncertainties forreconstructed whole-rocks combine 20 modal uncertainty with uncertainties on each mineralrsquos isotope composition Error bars on individualminerals may be smaller than the symbols Also shown are the isotopic compositions of the depleted mantle (DM DePaolo 1981 Griffin et al2000 Bell amp Tilton 2002) Range of 87Sr86Sr in eclogites from Ekati (Lac de Gras area 100 km north of Diavik Jacob et al 2003) in NorthAmerican kimberlites and compositions of low-temperature eclogites from Diavik and of Lac de Gras A154 South kimberlite fromSchmidberger et al (2007)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1864

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

one low-Mg eclogite has eNd similar to the depleted mantle(Fig 12b)Clinopyroxene in the phlogopitized eclogite YK1911

has unradiogenic Nd and Hf and strongly radiogenicSr whereas the coexisting garnet has radiogenic Nd andHf Although this sample contains 50 vol phlogopiteits whole-rock isotopic composition was calculated

from garnet and cpx alone because the contributionfrom phlogopite to the Hf and Nd budget is 52 and51 respectively (unpublished data) The whole-rockhas radiogenic Nd but has lower eHf than anyother sample analysed in accord with the extreme Hfabundance (53 ppm) and low LuHf of the cpx

DISCUSS IONOrigins of eclogites beneath the SlaveCratonHigh-Mg and high-Ca eclogites

With regard to their major- and trace-element contentsthe high-Mg and high-Ca eclogites from Lac de Gras arebroadly similar to variably evolved oceanic gabbros(Fig 13b and c) which may indicate an origin as subductedoceanic crust Compatible trace-element abundancesare useful to deduce potential crustal protoliths as theseelements are less susceptible to post-formation modification(Jacob 2004) The flat HREE patterns in the high-Mg andhigh-Ca eclogites are compatible with a low-pressureorigin as the oceanic crust is a product of moderate degreesof partial melting mainly within the spinel stability field(eg Presnall et al 2002) whereas melting with residualgarnet would qualitatively be expected to produce stronglyfractionated REE patterns Positive Sr and Pb anomaliesalthough compatible with a plagioclase signature couldalso be due to cryptic metasomatism (Dawson 1984) ofthe eclogites during their residence in the lithosphericmantle These anomalies are probably not related to kim-berlite infiltration as the whole-rock eclogite compositionswere reconstructed from analyses of fresh mineral coresLow ZrSm similar to that in primitive gabbros is

observed for many of the high-Mg and high-Caeclogites (Fig 13b and c) and is independent of the ratioof garnetcpx and whether or not the whole-rock wasreconstructed with rutile (Electronic Appendix 10) Thenegative trend of Mg-number against Yb in garnet(Fig 14) argues against igneous processes involving garnetaccumulation (Taylor amp Neal 1989) as suggested forexample for some high-Mg eclogites from West Africa(Barth et al 2002b) Rather the anti-correlation suggeststhat olivine fractionation played a role in the formation ofthe protoliths Fractionating olivine would excludeYb (Dolivinebasalt

frac14 002 Nikogosian amp Sobolev 1997)and incorporate Mg relative to Fe (Dfrac1415 and 05 respec-tively Taura et al1998) leading to progressive depletion ofMg and enrichment of Yb in the residual melt and henceanti-correlated Mg and Yb in the whole-rock This isinherited after eclogitization All eclogite types (exceptVR43452) plot along the olivine fractionation trend sug-gesting that they formed by a similar processHigh-Mg eclogites have lower

PREE than high-Ca

eclogites and some appear to have positive Eu anomalies

high-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

gabbrosSE Indian ridge

gabbrosOman ophiolite

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

high-Mg

RockNMORB(a)

VR43452

YK1949

low-Mg

boninite-like melts

01

001

1

10

(b)

01

001

1

10

(c)

01

001

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

Fig 13 Trace-element patterns of (a) low-Mg (b) high-Mg and(c) high-Ca eclogites compared with boninite-like volcanic rocksfrom the Isua greenstone belt (Polat et al 2002) gabbros from theOman ophiolite (Benoit et al 1996) and the SE Indian Ridge (Hartet al 1999) normalized to N-MORB of Sun amp McDonough (1989)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1865

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(EuEu up to 14) similar to more primitive cumulates(troctolite gabbros and olivine gabbros) reported in the lit-erature (Fig 13b) Although the sizes of these Eu anomaliesare not outside the analytical uncertainty their correlationwith CeYb in reconstructed eclogites supports their signif-icance (r2frac14066 nfrac14 4) The more primitive nature ofhigh-Mg eclogites is consistent with higher Mg-numbersrelative to high-Ca eclogites and may be due to olivineaccumulation during formation of their protolithsThe two samples with the lowest HREE abundanceshave the highest EuEu which is compatible with varia-tions in plagioclase modesMore evolved protoliths with higher pyroxeneplagio-

clase plus trapped melt could account for small or absentpositive Eu anomalies and higher

PHREE of high-Ca

eclogites (Fig 13c) Although the reason for the frequentassociation of graphite or diamond with high-Ca eclogitesis unclear (Aulbach et al 2003) the frequent occurrence ofkyanite in high-Ca eclogites also observed at Ekati justnorth of Lac de Gras has been argued to support a crustalprotolith (Jacob 2004) Kyanite andor corundum havebeen suggested to exsolve from high-temperature high-Alclinopyroxenes (Caporuscio amp Smyth 1990) Eclogitesfrom Lac de Gras show no exsolution textures suggestiveof such an origin although recrystallization can obscureor destroy microstructural evidence for exsolution(eg Griffin et al 1984)Modelling shows that qualitatively plagioclase accumu-

lation from a low-pressure melt derived from MORBmantle can account for positive Pb Sr and Eu and negativeZr anomalies relative to N-MORB whereas cpx accumula-tion or trapped melt increases Zr Y MREE and HREE

abundances (Fig 15) Some melt addition (for exampleby entrapment) seems to be required to increase all trace-element abundances to those observed for the eclogitesAlthough high-Ca and high-Mg eclogites could representshallow and deep portions of the same oceanic crust thehigher average temperatures obtained for the high-Caeclogites the frequent occurrence of carbon in high-Caeclogites and its absence in high-Mg eclogites the strongLREEHREE fractionation observed for six of seven high-Ca eclogites and the narrow range of compositions of thehigh-Mg eclogites suggest that they may have protolithsthat are not cogenetic Some discrepancies in particularaffecting the incompatible elements may be due to post-crystallization processes acting upon the protoliths suchas dehydration melting and later metasomatism that isevident in peridotites from the same locality (Griffin et al1999)Low K2OTh for high-Ca and high-Mg eclogites

(average 08 Table 4) relative to assumed crustal

0

2

4

6

8

10

30 40 50 60 70 80

Mg

VR43452

Yb

GARNET

pyroxenites

low-Mghigh-Mghigh-Ca

eclogites

VR19673ecl-g2

Fig 14 Garnet Mg-number vs Yb (ppm including 2s uncertainty)Lines qualitatively show separate correlations for eclogites (exceptoutlierVR43452) and pyroxenites (except outlierVR19673ecl-g2)

high-Ca

high-Mg

001

01

1

10

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

melt08 melt02 cryst cpx01 melt09 cryst

plag

(a)

001

01

1

10(b)

RockNMORB

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 15 Reconstructed (a) high-Ca and (b) high-Mg eclogitescompared with compositions of plagioclase and cpx crystallizingfrom a MORB melt modelled using distribution coefficients ofNorman et al (2005)The MORB was calculated as a 10 equilibriummelt of MORB pyrolite [composition from Kelemen et al (1993)except Pb Pr Sr Y and Lu which were calculated as residues from1 fractional melting of pyrolite from McDonough amp Sun (1995)using melt modes of Kinzler (1997) and distribution coefficients ofNorman et al (2005)] Also shown are mixtures of crystals (consistingof 50 plagioclase plus 50 cpx) and melt in the proportions 0901and 0208

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1866

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

precursors (eg gabbros SE Indian Ridge average 22Hart et al 1999) may reflect the preferred loss of fluid-mobile elements during dehydration (Becker et al 2000)although large uncertainties on these low-abundance ele-ments in the constituent minerals and reconstructedwhole-rocks makes this interpretation less than robustThe REE patterns of most high-Mg eclogites are too flatto represent melting residues as partial melting shouldproduce low LREEHREE (Electronic Appendix 10) Incontrast the marked LREE depletion of five of sevenhigh-Ca eclogites (LaLufrac14 02^08) contrasts with that ofassumed evolved gabbroic protoliths (LaLufrac1415^32Hart et al 1999) and can be modelled as a consequence ofpartial melting during subduction after eclogitization (iein the presence of garnet which retains HREE relative toLREE eg Barth et al 2002a) has taken place (Aulbachet al 2003)Our results are compatible with previous studies on

eclogite xenoliths from the Slave Craton includingthose from the same kimberlite that have linked thesesamples or some of these samples to Paleoproterozoicsubduction events (the 195^191Ga formation of theHottah terrane and the 18Ga formation of the GreatBear magmatic arc) in the Slave craton (Griffin et al1999 Kopylova et al 1999 Heaman et al 2002 2006Schmidberger et al 2005 2007)The compositional features discussed above although

compatible with crustal protoliths are not necessarilydefinitive of a subduction origin However if direct crystal-lization from basaltic melts at high pressure (ie mantledepths) is invoked instead specific conditions would haveto be met to satisfy the observations The Mg-number ineclogites from Lac de Gras are lower than those expectedof primary high-pressure mantle melts (eg Walter 1998OrsquoHara amp Herzberg 2002) requiring crystallization athigh pressures after olivine fractionation This originwould further require shielding from peridotitic wallrockwith which the fractionated melts would no longer be inequilibrium (Barth et al 2001) In such a setting the lowrelative abundances of fluid-mobile elements comparedwith those of less fluid-mobile elements and of LREErelative to HREE (Table 4) would reflect an expulsion offluids during crystallization of the protoliths and latermelt mobilization respectively The eclogites clearly donot represent garnet plus cpx cumulates which wouldrequire a positive correlation of Mg-number with Yb thatis at variance with the observed negative correlation ineclogitic garnets from this study (Fig 14)

Low-Mg eclogites

Garnets in low-Mg eclogites compositionally overlap withthose in massive eclogites from Jericho (NW of the Lacde Gras kimberlites) that were suggested to have a high-pressure intrusive origin based on the similarity of theirmajor-element compositions to those of Group A eclogites

fromsouthernAfrica forwhichamantleoriginhasbeenpro-posed (Kopylova et al1999) Garnets in low-Mg eclogites inthe present study have major-element compositions moresimilar to garnets in groupB rather thangroupAeclogitesThe positive slope in N-MORB-normalized HREE

abundances displayed by the low-Mg eclogites is clearlydistinct from other eclogite types and is similar to those ofboninite-like second-stage melts that are closely associatedwith ultramafic volcanics (Smithies et al 2004) (Fig 13a)The range of Mg-numbers in the low-Mg eclogites (55^66)overlaps with that of Archaean second-stage melts fromIsua (58^75) and Abitibi (52^82 Smithies et al 2004)These similarities might indicate that the low-Mg eclogiteshave an origin in a previously depleted mantle source suchas supra-subduction zone mantle Subchondritic ZrHf ofthe low-Mg eclogites contrasts with the supra-chondriticZrHf of the high-Mg and high-Ca eclogites (Aulbach et

al in preparation) and is consistent with previous cpx-controlled depletion in an arc mantle source This sourcecould have been subsequently remelted to generate theboninite-like precursor to the low-Mg eclogites as sug-gested for the protoliths of eclogites from Udachnaya(Jacob amp Foley 1999)Although some second-stage melts to which the low-Mg

eclogites are similar are closely associated with komatiitesand could contain a plume component the radiogenic ordepleted mantle-like Nd and Hf and unradiogenic Sr inthe low-Mg eclogites do not support plume involvement

Trace-element and Sr^Nd^Hf isotopeconstraints on post-formationmetasomatismSeveral low-Mg and high-Ca eclogites show negativeslopes in the LREE which may be indicative of interactionwith small amounts of LREE-enriched melt (Navon ampStolper 1987) amounting to cryptic metasomatism(ie not accompanied by crystallization of new phases) inthe nomenclature of Dawson (1984) High Nb abundancesrelative to elements of similar compatibility displayed bysome low-Mg and high-Ca eclogites (Fig 10) could resultfrom overestimation of rutile modes However high-CaeclogiteVR40345 is rutile-free and has been reconstructedwithout rutile its high Nb abundance is accompanied byLREE enrichment indicating that metasomatic additionmay be responsibleLow-Mg eclogites show a positive correlation of La with

Nb (Electronic Appendix 11a) indicating concomitantaddition of LREE and Nb This suggests that the metaso-matic agent was a solute-rich fluid or silicate silicic melt ashydrous fluids and carbonatites do not carry enough HFSEto affect the budget of eclogites (Green amp Wallace 1988Klemme et al 1995 Adam et al 1997 Stalder et al 1998)Of the samples that are enriched in Nb all but one (high-Ca eclogiteVR50909) show a simultaneous enrichment in

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1867

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Hf (Electronic Appendix 11b) Together these trace-element relationships suggest that both SmNd and LuHfwere lowered in these samples Their low parent^daughterratios contrast with higher ratios in unenriched samplesand resulted in the ingrowth of a range of Nd and Hf iso-tope compositions Compared with the low-Mg and high-Ca eclogites the high-Mg eclogites have a narrow range ofLa and Nb concentrations which correlates with a narrowrange in Nd and Hf isotope compositions (Fig 12)Precise model ages cannot be calculated because parent^

daughter ratios are available only from in situ analyseswhich have relatively large uncertainties Howeverevidence for long-term enrichment and depletion may bediscerned Low-Mg eclogites even those that are enrichedin LREE have relatively high SmNd as a result of havingoverall positive slopes in their REE patterns relative toMORB whereas the high-Mg eclogites have relativelyflat REE patterns and therefore lower SmNd (Fig 11)This translates into lower eNd for high-Mg eclogites thanfor low-Mg eclogites In contrast they show no systematicdifference with respect to 176Hf177Hf Because the low- andhigh-Mg eclogites are proposed to have different origins asdiscussed above and therefore did not necessarily have thesame initial isotopic composition and formation age theseisotopic differences cannot be used to estimate their agesA long-term coupling of trace-element and isotopic

abundances is also indicated by the fact that low-Mg eclo-gite YK1943 which has experienced Nb and LREE addi-tion has lower eNd (0) and eHf (thorn13) than low-Mg eclogiteVR43469 with higher LuHf and SmNd (eNdfrac14thorn47eHffrac14thorn128) If low-Mg eclogites YK1943 and VR43469originally had identical initial Nd and Hf isotope composi-tions and if the parent^daughter ratios are accurate andwere modified only once during their evolution shortlyafter metamorphism (when their isotopic compositionwould have been reset) this overprint can be calculatedfor both isotopic systems to have occurred at 17GaThis would suggest that Hf and Nd were added duringthe same event In contrast high-Mg eclogites VR67360and YK1949 would have had identical Ndisotope compositions at 10Ga whereas their Hf isotopecompositions are disturbed in that the sample with thehigher parent^daughter ratio has a lower 176Hf177Hf thanthe sample with the lower parent^daughter ratio Theseresults suggest that the LREE and HFSE enrichmentobserved in different samples occurred during separateevents that did not affect the eclogites equallyExtreme HFSE enrichment leading to the growth of

zircon and rutile and LREE addition leading to thegrowth of apatite in eclogites from the Jericho kimberlitein the northern Slave craton have been ascribed to twoseparate metasomatic events at around 18Ga (quasi-coincident with the age of formation and metamorphism)and 13^10Ga respectively (Heaman et al 2002 2006

Schmidberger et al 2005) The authors associated theseevents with Paleoproterozoic subduction and aMesoproterozoic thermal disturbance in the Slave litho-spheric mantle possibly linked to the Mackenzie igneousevent Mesoproterozoic Nd model ages mostly between13 and 14Ga were reported for eclogites from thenearby Ekati kimberlites (Jacob et al 2003) Although wedo not observe zircon or apatite in volatile-free eclogitesfrom Lac de Gras and the age constraints obtained fromour data are weak they support a link between theseevents described for other eclogite populations and themetasomatic enrichment observed in our studyThe Sr isotope data in the present study are similar to

those of Schmidberger et al (2007) who suggested that thedepleted-mantle-like values are consistent with derivationof the protoliths from an oceanic mantle source However87Sr86Sr in volatile-free eclogites in the present study isremarkably constant considering their Nd and Hf isotopevariability (Fig 12) and the variability of trace-elementpatterns (Fig 10) Given the evidence for dehydration andmelting in many of the eclogites fractionation of highlyincompatible Rb from Sr and differential ingrowth ofradiogenic Sr would be expectedThis may indicate a rela-tively recent overprint and considering the similarityof 87Sr86Sr to present-day depleted mantle which is signif-icantly lower than 87Sr86Sr of North American kimber-lites a juvenile (asthenospheric) source

PyroxenitesPyroxenites from Lac de Gras often show relationshipsin their major-element contents that are different fromthose observed in the eclogites For example pyroxeniticgarnets show a negative correlation between CaO andNa2O contents whereas these oxides are positively corre-lated for garnets in eclogites (Fig 5) CaO is well correlatedwith C2O3 in pyroxenitic garnets but not in eclogitic gar-nets (Fig 4) Mg-number is positively correlated withYb abundance for pyroxenitic garnets but anti-correlatedfor eclogitic garnets (Fig 14) This suggests that the pyrox-enites in the present study are not related to the eclogitesby fractionation or accumulation processes as proposedfor the origin of some pyroxenites (Foley et al 2003)Pyroxenites have also been interpreted as the products

of crystal segregation in magma channels with alteredoceanic crust as the magma source this would explaintheir Eu anomalies and stable and radiogenic isotopecompositions (Pearson et al 1993 Becker 1996) As thehigh-pressure liquidus phases of most terrestrial basaltsare cpx and garnet (OrsquoHara amp Herzberg 2002) partialmelts of eclogites with basaltic composition wouldcrystallize cpx and garnet but not opx More than half ofthe pyroxenites from Lac de Gras contain opx suggestingthat they probably did not crystallize directly fromeclogite-derived melts

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1868

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Low-Cr websterites and pyroxenites could be crystalli-zation products of mafic melts that interacted with perido-titic mantle thus becoming more refractory The positivecorrelation of Yb and Mg-number in garnet would allowan origin by fractional crystallization from such meltsAlternatively they formed by reaction of silicic (possiblyeclogite-derived) melts with mantle peridotites(eg Kelemen et al 1993 Aulbach et al 2002) This agreeswith the similarity of hypothetical melts in equilibriumwith pyroxenitic garnets in most samples to eclogite-derived melts (Electronic Appendix 12) Niobium concen-trations in all hypothetical melts are much higher than ineclogite-derived melts regardless of whether or not rutileis assumed to be in the residue The same is true for LaCe and Zr concentrations in some samples which mayindicate that NbZr and LREE were added to thepyroxenites some time after their formation as discussedin the previous section for some eclogites Because mini-mum ages for pyroxenitic rutile with an unradiogenicHf isotope composition (calculated by projecting the176Hf177Hf to the model depleted mantle isotopic evolutioncurve assuming zero ingrowth) range to 178066Ga(Aulbach et al in preparation) it appears likely that thepyroxenites were not formed recently (eg during kimber-lite-related or precursory melt migration) and that some ofthe samples are at least MesoproterozoicGarnet pyroxeniteYK1952 consists of schlieren of equili-

brated garnet pyroxenite in a lsquomatrixrsquo of spongy cpxfine cpx^opx intergrowths embayed garnet and meltpatches (Fig 2f) Garnet in the pyroxenite schlieren hasvery low LREE and HFSE abundances Orthopyroxenein the matrix has distinctly higher Mg-number andCr-number and contains more CaO than opx from thepyroxenite schlieren whereas opx rims in the schlierenare compositionally similar to opx in the matrix The con-siderable compositional and textural disequilibrium inthis sample suggests that this is a young feature causedby reaction with fluids shortly before or at the time ofentrainmentAs a consequence of the small sample sizes of most of

the pyroxenites studied here a comprehensive isotopicdataset is not available Clearly more such data areneeded in addition to stable isotope data to obtain betterconstraints on the origins of this rock type Consideringthe microstructural and compositional variability ofthe pyroxenites and major-element relationships oppositeto those observed for eclogites it appears certain that pyr-oxenites form by a variety of processes some of which areunrelated to the formation of eclogites and their protoliths

Volatile-rich eclogitesVolatile-rich eclogites from Lac de Gras display the mostextreme Hf^Nd isotope systematics of all the analysedsamples Garnets in phlogopitized (YK1911) and in cal-cite-bearing eclogite (VR67112b) have the most radiogenic

Nd (eNdfrac14 2486 and 1927 respectively) Garnet inVR67112b has extremely high eHf (2516) Clinopyroxene inYK1911 has the lowest eHf measured for any mineralseparate from Lac de Gras (53) in accord with the veryhigh Hf abundance in its cpx (5 ppm) In contrastgarnet in this sample has radiogenic Hf (eHffrac14 136) Thisdiscrepancy is at least in part due to their relatively lowequilibration temperatures (7698C and 8558C respec-tively) which allowed the isotopic compositions of cpxand garnet to evolve separately at their low andhigh parent^daughter element ratios respectively forextended periods of time and possibly since formationAccordingly the cpx in YK1911 gives a minimum Hf ageof 27GaSlab-derived fluids could have been introduced during

the 27Ga collision of the western Slave Craton domainincluding the thick plume-derived deep lithosphericmantle layer and the eastern domain with its highlydepleted lithosphere This would also be consistent withthe observation that volatile-rich eclogites occur at shallowdepths in the mantle (TKroghfrac14 685^8558C correspondingto depths of 85^115 km along a 40 mWm2 geotherm)where the bulk of dehydration and production of solute-rich melts in a subducting slab might be expected to occur(Becker et al 2000 Hermann et al 2006) Their negativeEu anomalies might indicate that the fluid was derivedfrom the shallow portion of the slab that experiencedplagioclase fractionation The presence of carbonate inVR67112b requires a carbonated sourceHigh 87Sr86Sr in cpx in YK1911 is probably the conse-

quence of isotopic equilibration with high-RbSr phlogo-pite after long-term radiogenic ingrowth while Hfaddition led to retarded ingrowth of radiogenic HfThese systematics combined with radiogenic Nd appear

inconsistent with the involvement of sediments in the pet-rogenesis of this sample (eg Perfit et al 1980 Vervoortet al 1999) or a hydrous fluid such as those responsible forMARID metasomatism in the Kaapvaal craton (Konzettet al 1998)Whatever the origin of the fluids the apparentHf mobility suggests that they were not in equilibriumwith a rutile-bearing source and also that they weresolute-rich in nature (ie hydrous melts able to dissolveHFSE) and the relatively high HREE abundances of thevolatile-rich eclogites indicate a garnet-free source suchas amphibolite This allowed evolution of the sample atrelatively high time-integrated SmNd

SUMMARY AND CONCLUSIONS

(1) The Lac de Gras kimberlites carried abundant eclo-gite and pyroxenite xenoliths to the surface Threemain eclogite types are distinguished in terms ofmajor-element composition and trace element

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1869

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

patterns high-Mg high-Ca and low-Mg eclogitesThe garnets of all eclogite types have anti-correlatedMg-number and Yb pointing to olivine fractionationas a controlling factor in the formation of their pre-cursor rocks Their flat or positively sloped HREEpatterns qualitatively indicate that the melt parentalto the precursor rocks was not in equilibrium with agarnet-bearing residue

(2) Reconstructed high-Mg eclogites and high-Caeclogites have flat HREE patterns positive Pb andSr anomalies and low ZrSm similar to oceanic gab-bros reported in the literature High-Mg eclogiteshave variably low HREE contents and some havesmall positive Eu anomalies possibly suggestive ofmore primitive relatively plagioclase-rich protolithswhereas high-Ca eclogites have higher HREE abun-dances and small or absent Eu anomalies that requiremore cpx- andor melt-rich protoliths if a subductionorigin is assumed

(3) Low-Mg eclogites have distinctive trace-element pat-terns that show a marked positive slope in the HREEand subchondritic ZrHf They resemble someArchaean second-stage melts (lsquoboninitesrsquo) and theirprotoliths appear to have formed from a previouslydepleted source

(4) After formation the eclogites were subject to a varietyof processes Most have low ratios of fluid-mobile vsless fluid-mobile elements possibly as a result of dehy-dration of the protoliths In addition some of the low-Mg and high-Ca eclogites have low LREEHREEconsistent with partial melt lossWhereas the low-Mgeclogites show concomitant enrichment in La Nb andZr some high-Mg and high-Ca eclogites show evi-dence for variable and unsystematic (decoupled)LREE and Nb addition

(5) Isotopically the trace-element characteristics of thedifferent eclogite types translate into lower eNd forhigh-Mg eclogites (negative slope in MREEfrac14 lowerSmNd) than for low-Mg eclogites (positive slope inMREEfrac14higher SmNd) Within the low-Mg groupsamples that show evidence for metasomatism (lowLuHf and SmNd) have lower eNd and eHf than asample that was apparently not metasomatized BothNd and Hf addition can be dated to 17Ga if identi-cal initial isotope compsitions are assumed for thelow-Mg eclogites In contrast the relatively constantand depleted mantle-like 87Sr86Sr of eclogitic cpxdespite variable Nd and Hf isotope compositions andparent^daughter ratios suggests recent input of juve-nile material

(6) Garnet and cpx in pyroxenites show major-elementrelationships different from those in eclogitesTherefore these two rock types are probably notrelated by fractionation processes The presence of

opx precludes direct crystallization from slab-derivedmelts Pyroxenites may be the product of (slab-derived) melts hybridized during assimilation of peri-dotitic mantle

(7) Garnets in phlogopite- and carbonate-bearing eclo-gites have extreme isotopic compositions includingthe most radiogenic Nd and Hf in the datasetNegative Eu anomalies may point to the involvementof a crustal protolith Highly radiogenic Sr radiogenicNd and unradiogenic Hf in the reconstructedphlogopite-bearing sample combined with highHREE abundances and evidence for Hf mobilitymay indicate involvement of solute-rich fluids origi-nating in a garnet- and rutile-free source

ACKNOWLEDGEMENTSHelp in the clean laboratories and with the analyticalfacilities by S Elhlou C Lawson A Sharma P Wielandand M Zhang is greatly appreciated Bill Griffin isthanked for his support and for agreeing to reviewseveral versions of the manuscript despite disagreeingwith many of the conclusions presented here The paperbenefited greatly from the detailed and constructivereviews provided by the editors Gareth Davies andMarjorie Wilson and by Matthias Barth Larry Heamanand an anonymous reviewer This work was funded by aMacquarie University International PostgraduateAward and Postgraduate Research Fund (SA) by anARC SPIRT grant sponsored by Kennecott Canada Incand by ARC Large and Discovery Grants to SYOrsquoRAnalytical data were obtained using instrumentationfunded by ARC LIEF and DEST Systemic InfrastructureGrants industry partners and Macquarie UniversityThis is Publication 481 from the ARC National KeyCentre for Geochemical Evolution and Metallogeny ofContinents (wwwesmqeduauGEMOC)

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online

REFERENCESAdam J GreenT H Sie S H amp Ryan C G (1997) Trace element

partitioning between aqueous fluids silicate melts and mineralsEuropeanJournal of Mineralogy 9 569^584

Aulbach S Stachel T Viljoen K S Brey G P amp Harris J W(2002) Eclogitic and websteritic diamond sources beneath theLimpopo Beltccedilis slab-melting the link Contributions to Mineralogy

and Petrology 143 56^70Aulbach S GriffinW L Pearson N J OrsquoReilly S Y Kivi K amp

Doyle B (2003) Origins of eclogites beneath the central SlaveCraton Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA-0011

Aulbach S Griffin W L Pearson N J OrsquoReilly S Y amp Kivi K(2004a) Mantle formation and evolution Slave Craton constraints

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1870

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

from HSE abundances and Re^Os isotope systematics of sulfideinclusions in mantle xenocrysts Chemical Geology 208 61^88

Aulbach S Griffin W L OrsquoReilly S Y amp McCandless T E(2004b) Genesis and evolution of the lithospheric mantle beneaththe Buffalo HeadTerrane Alberta (Canada) Lithos 77 413^451

Aulbach S GriffinW L Pearson N J OrsquoReilly S Y amp Doyle B(2007) Lithosphere formation in the central Slave Craton(Canada) plume subcretion or lithosphere accretion Contributionsto Mineralogy and Petrology (in press) doi 101007s00410-007-0200-1

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2001)Geochemistry of xenolithic eclogites from West Africa Part IA link between low MgO eclogites and Archean crust formationGeochimica et Cosmochimica Acta 65 1499^1527

Barth M G Foley S F amp Horn I (2002a) Partial melting inArchean subduction zones constraints from experimentally deter-mined trace element partition coefficients between eclogitic miner-als and tonalitic melts under upper mantle conditions PrecambrianResearch 113 323^340

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2002b)Geochemistry of xenolithic eclogites from West Africa part 2Origins of the high MgO eclogites Geochimica et Cosmochimica Acta

66 4325^4345Beard B L Fraracci K NTaylor L A Snyder G A Clayton R

A Mayeda T K amp Sobolev NV (1996) Petrography andgeochemistry of eclogites from the Mir kimberliteYakutia RussiaContributions to Mineralogy and Petrology 125 293^310

Becker H (1996) Geochemistry of garnet peridotite massifs fromlower Austria and the composition of deep lithosphere beneatha Palaeozoic convergent plate margin Chemical Geology 134 49^65

Becker H Jochum K P amp Carlson R W (2000) Trace elementfractionation during dehydration of eclogites from high-pressureterranes and the implications for element fluxes in subductionzones Chemical Geology 163 65^99

Bell K amp Tilton G R (2002) Probing the mantle The story fromcarbonatites EOS Transactions American Geophysical Union 83275^277

Benoit M Polvecurren M amp Ceuleneer G (1996) Trace element andisotopic characterisation of mafic cumulates in a fossil mantlediapir (Oman ophiolite) Chemical Geology 134 199^214

BleekerW (2003) The late Archean record a puzzle in ca 35 piecesLithos 71 99^134

BleekerW Ketchum JW F JacksonV A amp Villeneuve M (1999)The central Slave Basement Complex Part I Its structural topol-ogy and autochthonous core Canadian Journal of Earth Sciences 361083^1109

Blichert-Toft J amp Albare de F (1997) The Lu^Hf isotope geochemis-try of chondrites and the evolution of the mantle^crust systemEarth and Planetary Science Letters 148 243^258

Blichert-Toft J Chauvel C amp Albare de F (1997) Separation of Hfand Lu for high precision isotope analysis of rock samples by mag-netic sector-multiple collector ICP-MS Contributions to Mineralogy

and Petrology 127 248^260Brey G P amp Kolaquo hler T (1990) Geothermobarometry in four-phase

lherzolites II New thermobarometers and practical assessment ofexisting thermobarometers Journal of Petrology 31 1353^1378

Brey G P Kolaquo hlerT amp Nickel K G (1990) Geothermobarometry infour-phase lherzolites I Experimental results from 10 to 60 kbJournal of Petrology 31 1313^1352

Caporuscio F amp Smyth J (1990) Trace element crystal chemistryof mantle eclogites Contributions to Mineralogy and Petrology 105550^561

Carbno G B amp Canil D (2002) Mantle structure beneath the SWSlave Craton Canada constraints from garnet geochemistry in theDrybones Bay Kimberlite Journal of Petrology 43 129^142

Coleman R G Lee E D Beatty L B amp Brannock W W (1965)Eclogites and eclogites their differences and similaritiesGeological Society of America Bulletin 76 483^508

Creaser R A Grulaquo tter H Carlson J amp Crawford B (2004)Macrocrystal phlogopite Rb^Sr dates for the Ekati propertykimberlites Slave Province Canada evidence for multiple intru-sive episodes in the Paleocene and Eocene Lithos 76 399^414

Davies R M GriffinW L Pearson N J Andrew A S Doyle BJ amp OrsquoReilly S Y (1999) Diamonds from the deep Pipe DO-27Slave Craton Canada In Gurney J J Gurney J L Pascoe MD amp Richardson S H (eds) Proceedings of the 7th International

Kimberlite Conference CapeTown Red Roof Design pp 148^155Davies R M Griffin W L OrsquoReilly S Y amp Doyle B J (2004)

Mineral inclusions and geochemical characteristics of microdia-monds from the DO27 A154 A21 A418 DO18 DD17 and RanchLake kimberlites at Lac de Gras Slave Craton Canada Lithos 7739^55

DavisW J amp Hegner E (1992) Neodymium isotopic evidence for thetectonic assembly of Late Archean crust in the Slave Provincenorthwest Canada Contributions to Mineralogy and Petrology 111493^504

Davis W J Jones A G Bleeker W amp Grutter H (2003)Lithosphere development in the Slave craton a linked crustal andmantle perspective Lithos 71 575^589

Dawson J B (1984) Contrasting types of upper-mantle metasoma-tism In Kornprobst J (ed) Kimberlites II The Mantle and

Crust^Mantle Relationships Amsterdam Elsevier pp 289^294DePaolo D J (1981) Neodymium isotopes in the Colorado Front

Range and crust^mantle evolution in the Proterozoic Nature 291193^196

Erlank A J amp Kushiro I (1970) Potassium contents of syntheticpyroxenes at high temperatures and pressures Carnegie Institution ofWashingtonYearbook 68 433^439

Foley S F Barth M G amp Jenner G A (2000) Rutilemelt partitioncoefficients for trace elements and an assessment of the influence ofrutile on the trace element characteristics of subduction zonemagmas Geochimica et Cosmochimica Acta 64 933^938

Foley S F Buhre S amp Jacob D E (2003) Evolution of the Archaeancrust by delamination and shallow subduction Nature 421 249^252

Graham I Burgess J L Bryan D Ravenscroft P J Thomas EDoyle B J Hopkins R amp Armstrong K A (1999) Explorationhistory and geology of the Diavik Kimberlites Lac de GrasNorthwest Territories Canada In Gurney J J Gurney J LPascoe M D amp Richardson S H (eds) Proceedings of the 7th

International Kimberlite Conference Cape Town Red Roof Designpp 262^279

Green D H amp Wallace M E (1988) Mantle metasomatism byephemeral carbonatite melts Nature 336 459^462

Green T H Blundy J D Adam J amp Yaxley G M (2000) SIMSdetermination of trace element partition coefficients betweengarnet clinopyroxene and hydrous basaltic liquids at 2^75 GPaand 1080^12008C Lithos 53 165^187

GriffinW L amp OrsquoReilly SY (2007) Cratonic lithospheric mantle Isanything subducted Episodes 30(1) 43^53

Griffin W L Wass S Y amp Hollis J D (1984) Ultramafic xenolithsfrom Bullenmerri and Gnotuk maars Victoria Australia petrol-ogy of a sub-continental crust^mantle transition Journal of

Petrology 25 53^87GriffinW L Doyle B J Ryan C G Pearson N J OrsquoReilly SY

Davies R Kivi K van Achterbergh E amp Natapov LM (1999)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1871

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Layered mantle lithosphere in the Lac de Gras Area SlaveCraton composition structure and origin Journal of Petrology 40705^727

Griffin W L Pearson N J Belousova E Jackson S E vanAchterbergh E OrsquoReilly SY amp Shee S R (2000)The Hf isotopecomposition of cratonic mantle LAM-MC-ICPMS analysis ofzircon megacrysts in kimberlites Geochimica et Cosmochimica Acta

64 133^147Griffin W L OrsquoReilly S Y Doyle B J Pearson N J

Coopersmith H Kivi K Malkovets V amp Pokhilenko N(2004) Lithosphere mapping beneath the North American plateLithos 77 873^922

Grulaquo tter H S Apter D B amp Kong J (1999) Crust^mantle couplingevidence from mantle-derived xenocrystic garnets In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 307^312

Hart S R Blusztajn J Dick H J B Meyer P S ampMuehlenbachs K (1999) The fingerprint of seawater circulationin a 500-meter section of ocean crust gabbros Geochimica et

Cosmochimica Acta 63 4059^4080Harte B amp Kirkley M B (1997) Partitioning of trace elements

between clinopyroxene and garnet Data from mantle eclogitesChemical Geology 136 1^24

Heaman L M Creaser R A amp Cookenboo H O (2002) Extremeenrichment of high field strength elements in Jericho eclogitexenoliths A cryptic record of Paleoproterozoic subduction partialmelting and metasomatism beneath the Slave craton CanadaGeology 30 507^510

Heaman L M Kjarsgaard B A amp Creaser R A (2004) The tem-poral evolution of North American kimberlites Lithos 76 377^397

Heaman L M Creaser R A Cookenboo H O amp Chacko T(2006) Multi-stage modification of the northern Slave mantlelithosphere Evidence from zircon- and diamond-bearing eclogitexenoliths entrained in Jericho Kimberlite Journal of Petrology 47821^858

Hermann J Cpandler C Hack A amp Korsakov A (2006) Aqueousfluids and hydrous melts in high-pressure and ultra-high pressurerocks implications for element transfer in subduction zones Lithos92 399^417

Hills DV amp Haggerty S E (1989) Petrochemistry of eclogites fromthe Koidu Kimberlite Complex Sierra Leone Contributions to

Mineralogy and Petrology 103 397^422Hoffman P F (1989) Precambrian geology and tectonic history of

North America In Bally A amp Palmer A (eds) The Geology of

North AmericaccedilAn Overview Boulder CO Geological Society ofAmerica pp 447^512

IrelandT R Rudnick R L amp Spetsius Z (1994) Trace-elements indiamond inclusions from eclogites reveal link to Archean granitesEarth and Planetary Science Letters 128 199^213

Irvine G J Pearson D G Kjarsgaard B A Carlson R WKopylova M G amp Dreibus G (2003) Evolution of the litho-spheric mantle beneath Northern Canada a Re^Os isotope andPGE study of kimberlite-derived peridotite xenoliths fromSomerset Island and a comparison to the Slave and Kaapvaalcratons Lithos 71 461^488

Isachsen C E amp Bowring S A (1994) Evolution of the Slave cratonGeology 22 917^920

Jacob D E (2004) Nature and origin of eclogite xenoliths from kim-berlites Lithos 77 295^316

Jacob D E amp Foley S F (1999) Evidence for Archean ocean crustwith low high field strength element signature from diamondifer-ous eclogite xenoliths Lithos 48 317^336

Jacob D Jagoutz E Lowry D Mattey D amp Kudrjavtseva G(1994) Diamondiferous eclogites from Siberiaccedilremnants ofArchean oceanic-crust Geochimica et Cosmochimica Acta 585191^5207

Jacob D E Fung A Jagoutz E amp Pearson D G (2003) Petrologyand geochemistry of eclogite xenoliths from the Ekati kimberlitesarea Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA_239

Jacob D E Bizimis M amp Salters V J M (2005) Lu^Hf and geo-chemical systematics of recycled ancient oceanic crust evidencefrom Roberts Victor eclogites Contributions to Mineralogy and

Petrology 148 707^720Jerde E A Taylor L A Crozaz G Sobolev NV amp SobolevV N

(1993) Diamondiferous eclogites from Yakutia Siberiaccedilevidencefor a diversity of protoliths Contributions to Mineralogy and Petrology

114 189^202Jones A G Ferguson I J Chave A D Evans R L amp

McNeice GW (2001) The electric lithosphere of the Slave cratonGeology 29 423^426

Kelemen P B Shimizu N amp Dunn T (1993) Relative depletionof niobium in some arc magmas and the continental-crust-partitioning of K Nb La and Ce during meltrockreaction in the upper-mantle Earth and Planetary Science Letters 120111^134

Kinzler R J (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition Application to mid-ocean ridge basalt petrogenesis Journal of Geophysical Research 102853^874

Klemme S van der Laan S R Foley S F amp Guenther D (1995)Experimentally determined trace and minor element partitioningbetween clinopyroxene and carbonatite melt under upper mantleconditions Earth and Planetary Science Letters 133 439^448

Konzett J Armstrong R A Sweeney R J amp CompstonW (1998)The timing of MARID metasomatism in the Kaapvaal mantle anion probe study of zircons from MARID xenoliths Earth and

Planetary Science Letters 160 133^145Kopylova M G Russell J K amp Cookenboo H (1999) Mapping the

lithopshere beneath the north central Slave Craton In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 468^479

Krogh E (1988) The garnet^clinopyroxene iron^magnesiumgeothermometerccedila reinterpretation of existing experimentaldata Contributions to Mineralogy and Petrology 99 44^48

Kusky T M (1989) Accretion of Archean Slave Province Geology 1763^67

LeCheminant A N amp Heaman L M (1989) Mackenzie igneousevents Canada Middle Proterozoic hotspot magmatismassociated with ocean opening Earth and Planetary Science Letters

96 38^48LeCheminant A N Richardson D G DiLabio R N W amp

Richardson K A (1996) La recherche de diamants au CanadaGeological Survey of Canada Open File 3228 268pp

MacGregor I D amp Carter J L (1970) The chemistry of clinopyrox-enes and garnets of eclogite and peridotite xenoliths from theRoberts Victor mine South Africa Physics of the Earth and PlanetaryInteriors 3 391^397

MacGregor I D amp Manton W I (1986) Roberts Victor eclogitesancient oceanic crust Journal of Geophysical Research 91 14063^14079

MacKenzie J M amp Canil D (1999) Composition and thermal evo-lution of cratonic mantle beneath the central Archean SlaveProvince NWT Canada Contributions to Mineralogy and Petrology

134 313^324

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1872

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

McCandlessT E amp Gurney J J (1989) Sodium in garnet and potas-sum in clinopyroxene criteria for classifying mantle eclogites InRoss J (ed) Kimberlites and Related Rocks Carlton Blackwellpp 827^832

McDonoughW F amp Sun S-S (1995) The composition of the EarthChemical Geology 120 223^253

Navon O amp Stolper E (1987) Geochemical consequences of meltpercolation the upper mantle as a chromatographic columnJournal of Geology 95 285^307

Nikogosian I K amp Sobolev AV (1997) Ion-microprobe analysis ofmelt inclusions in olivine experience in estimating the olivine^meltpartition coefficients of trace elements Geochemistry International 35119^126

Norman M D Griffin W L Pearson N J Garcia M O ampOrsquoReilly S Y (1998) Quantitative analysis of trace element abun-dances in glasses and minerals a comparison of laser ablationinductively coupled plasma mass spectrometry solution inductivelycoupled plasma mass spectrometry proton microprobe and elec-tron microprobe data Journal of Analytical Atomic Spectrometry 13477^482

Norman M D Garcia M O amp Pietruszka A J (2005) Trace-element distribution coefficients for pyroxenes plagioclase andolivine in evolved tholeiites from the 1955 eruption of KilaueaVolcano Hawairsquoi and petrogenesis of differentiated rift-zonelavas American Mineralogist 90 888^899

OrsquoHara M J amp Herzberg C (2002) Interpretation of trace elementand isotope features of basalts Relevance of field relations petrol-ogy major element data phase equilibria and magma chambermodeling in basalt petrogenesis Geochimica et Cosmochimica Acta 662167^2191

PadghamW A amp FysonW K (1992) The Slave Province a distinctcraton CanadianJournal of Earth Sciences 29 2072^2086

Pearson D G Davies G R amp Nixon P H (1993) Geochemical con-straints on the petrogenesis of diamond facies pyroxenites from theBeni Bousera peridotite massif North Morocco Journal of Petrology34 125^172

Pearson N J Griffin W L Doyle B J OrsquoReilly S Y vanAchterbergh E amp Kivi K (1999) Xenoliths from kimberlitepipes of the Lac de gras area Slave Craton Canada In GurneyJ J Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite ConferenceCapeTown Red Roof Design pp 644^658

Perfit N R Gust D A Bence A E Arculus R J amp Taylor S R(1980) Chemical characteristics of island-arc basalts implicationsfor mantle sources Chemical Geology 30 227^256

Polat A Hofmann A W amp Rosing M T (2002) Boninite-likevolcanic rocks in the 37^38 Ga Isua greenstone belt WestGreenland geochemical evidence for intra-oceanic subductionzone processes in the early Earth Chemical Geology 184 231^254

Poudjom DjomaniY H GriffinW L OrsquoReilly SY amp Doyle B J(2005) Lithospheric domains and controls on kimberlite emplace-ment Slave Province Canada evidence from elastic thickness andupper mantle composition Geochemistry Geophysics Geosystems 6doi1010292005GC0009978

Presnall D C Gudfinnsson G H amp Walter M J (2002)Generation of mid-ocean ridge basalts at pressures from 1 to 7GPa Geochimica et Cosmochimica Acta 66 2073^2090

Schmidberger S S Heaman L M Simonetti A Creaser R A ampCookenboo H O (2005) Formation of Paleoproterozoiceclogitic mantle Slave Province (Canada) Insights from in-situHf and U^Pb isotopic analyses of mantle zircons Earth and

Planetary Science Letters 240 621^633

Schmidberger S S Simonetti A Heaman L M Creaser R A ampWhiteford S (2007) Lu^Hf in-situ Sr and Pb isotope and traceelement systematics for mantle eclogites from the Diavik diamondmine Evidence for Paleoproterozoic subduction beneath the Slavecraton Canada Earth and Planetary Science Letters 254 55^68

Schulze D J amp Helmstaedt H (1988) Coesite-sanidine eclogitesfrom kimberlite products of mantle fractionation or subductionJournal of Geology 96 435^443

Smithies R H Champion D C amp Sun S S (2004) The case forArchaean boninites Contributions to Mineralogy and Petrology 147705^721

Smyth J R Caporuscio F A amp McCormick T C (1989) Mantleeclogites evidence of igneous fractionation in the mantle Earth andPlanetary Science Letters 93 133^141

Sobolev NV amp Lavrentrsquoyev Y G (1971) Isomorphic sodium admix-ture in garnets formed at high pressure Contributions to Mineralogy

and Petrology 31 1^12Stalder R Foley S F Brey G P amp Horn I (1998) Mineral aqueous

fluid partitioning of trace elements at 900^12008C and 30^57GPa New experimental data for garnet clinopyroxene and rutileand implications for mantle metasomatism Geochimica et

Cosmochimica Acta 62 1781^1801Sun S-S amp McDonough W F (1989) Chemical and isotopic sys-

tematics of oceanic basalts implications for mantle compositionsand processes In Saunders A D amp Norris M J (eds)Magmatism in the Ocean Basins Geological Society London Special

Publications 42 313^345Taura H Yurimoto H Kurita K amp Sueno S (1998) Pressure

dependence on partition coefficients for trace elements between oli-vine and the coexisting melts Physics and Chemistry of Minerals 25469^484

Taylor L A amp Neal C R (1989) Eclogites with oceanic crustal andmantle signatures from the Bellsbank kimberlite South AfricaPart I Mineralogy petrography and whole rock chemistryJournal of Geology 97 551^567

van Achterbergh E Ryan C G amp Griffin W L (1999) Glitter online intensity reduction for the laser ablation inductively coupledplasma spectrometry Abstract 9th V M GoldschmidtConference LPI Boston MA 905 pp Abstract no 7215

van Breemen O DavisW J amp King J E (1992) Temporal distribu-tion of granitoid plutonic rocks in the Archean Slave ProvinceNorthwest Canadian Shield Canadian Journal of Earth Sciences 292186^2199

Vervoort J D Patchett P J Blichert-Toft J amp Albare de F (1999)Relationship between Lu^Hf and Sm^Nd isotopic systems in theglobal sedimentary system Earth and Planetary Science Letters 16879^99

Viljoen K S Smith C B amp Sharp Z D (1996) Stable and radio-genic isotope study of eclogite xenoliths from the Orapa kimberliteBotswana Chemical Geology 131 235^255

Walter M J (1998) Melting of garnet peridotite and theorigin of komatiite and depleted lithosphere Journal of Petrology 3929^60

Wasserburg G J Jacobsen S B DePaolo D J McCulloch M Tamp Wen T (1981) Precise determination of SmNd ratios Sm andNd isotopic abundances in standard solutions Geochimica et

Cosmochimica Acta 45 2311^2323Westerlund K J Shirey S B Richardson S H Carlson R W

Gurney J J amp Harris JW (2006) A subduction wedge origin forPaleoarchean peridotitic diamonds and harzburgites from thePanda kimberlite Slave craton evidence from Re^Os isotope sys-tematics Contributions to Mineralogy and Petrology 152 275^294

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1873

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Page 12: Origins of Xenolithic Eclogites and Pyroxenites from the Central

ClinopyroxeneTrace-element abundances in cpx are summarized inTable 3 and the complete dataset is reported in ElectronicAppendix 6 Clinopyroxenes in low-Mg eclogites havemildly positive to negative slopes from LaN to NdN andsmooth negative slopes between SmN and HoN thatmostly continue to LuN (Fig 8a) Niobium is below detec-tion or 5007 chondritic and SrN is variably enrichedrelative to neighbouring elements with the peak size corre-lating negatively with LREEN ZrHf is subchondritic andHfN is enriched relative to SmN whereasTiN is enriched ordepleted relative to similarly compatible elementsWe ana-lysed the spongy rim in cpxYK1943 and its trace-elementpattern is similar to that of cpx cores for more compatibleelements whereas the most incompatible elements areseveral orders of magnitude higher Clinopyroxenes in

high-Mg eclogites have steeper negative slopes betweenNdN and LuN than those in low-Mg eclogites combinedwith lower HREE abundances HfN is depleted relative toSmN (Fig 8b) High-Ca eclogites contain cpx with sinusoi-dal REEN patterns with lower LREE than other eclogitetypes and the heaviest REE below detection They haveuniformly strong positive Sr peaks relative to elements ofsimilar compatibility (Fig 9) The diamond-bearing high-Ca eclogite (VR40345) has cpx with unusually high Nband LREE abundancesClinopyroxenes in four pyroxenites have disparate trace

element patterns They all share depletions in TiN relativeto similarly compatible elements and two show positiveSrN peaks Two samples have cpx with high and flatLREEN to MREEN patterns and steep negative slopes inthe HREEN The trace-element patterns of cpx in YK1914

Table 3 Continued

Cpx

Type Low-Mg High-Mg High-Ca volatile-rich Pyx

av min max 1s av min max 1s av min max 1s VR67112b YK1911 av min max 1s

n 6 4 8 5 4 4

Ga 16 10 20 3 12 9 15 2 26 15 39 9 129 123 6 11 11 5

Rb na 5008 04 na na 5007 05 02 na 5003 03 na 02 5004 03 010 05 0

Sr 166 85 290 74 222 153 285 50 243 83 440 142 85 120 200 11 414 209

Y 2 04 4 10 13 07 2 04 11 03 3 10 28 51 2 03 3 2

Zr 19 4 48 15 25 8 45 13 17 8 28 7 111 129 36 008 81 42

Nb 01 0003 03 01 na 5002 00 na na 5002 16 9 007 011 010 003 02 008

Cs na 5001 6 na na 5006 5006 na5013 5004 5013 na bdl 5002 na 5001 5002 na

Ba 09 02 2 07 09 01 2 07 10 03 3 10 06 01 13 06 3 10

La 3 03 8 3 3 23 3 04 4 01 14 6 51 97 4 01 10 4

Ce 7 13 20 6 10 7 13 2 7 05 24 10 25 28 11 01 28 13

Pr 11 02 3 08 2 10 3 06 08 02 2 09 47 46 2 002 4 2

Nd 5 12 12 4 10 5 14 4 4 09 8 26 23 22 8 01 22 10

Sm 11 05 21 06 20 10 31 10 07 02 11 03 47 53 2 005 5 2

Eu 03 01 05 01 06 028 09 03 02 010 03 007 087 123 06 002 15 07

Gd 08 03 13 03 11 05 19 06 05 03 06 011 28 37 15 010 36 2

Dy 05 01 07 02 04 022 06 01 02 013 03 007 11 16 07 008 14 06

Ho 01 00 02 004 006 003 007 002 na 5003 012 na 011 021 009 001 02 007

Er na 507 507 na 011 004 02 00 na 5006 03 na 019 033 02 002 03 012

Yb na 5001 02 na na 5001 011 na na 5006 03 na 007 015 009 001 02 006

Lu na 5001 003 na na50002 002 na na50008 01 na 0006 0013 0013 0002 003 001

Hf 12 03 24 07 11 04 18 05 10 06 16 04 34 53 na50015 35 na

Ta na50003 50015 na na50006 006 na na50004 5004 na 0014 00236 na50003 503 na

Pb 06 02 12 04 06 04 09 02 na 5016 07 02 065 098 08 01 18 1

Th na50005 06 na 012 005 02 005 na 5002 5006 na 021 02 001 06 03

U na 5005 5005 na 002 0009 005 001 na 5003 5005 na 0013 024 na50005 003 na

Averaged for the three eclogite types and for pyroxenites as in Table 2 pyx pyroxenite n number of analyses averagedbdl below detection limit

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1854

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

and YK1952 are similar although at different abundanceswith only modest LREEN over HREEN enrichment andstrong depletions in ZrN relative to neighbouring elements(Fig 8d)Clinopyroxenes in the volatile-rich samples have flat pat-

terns and high abundances from LaN to SmN and smoothnegative slopes from SmN to LuN Niobium Sr and Ti are

noticeably depleted relative to elements of similar compat-ibility (Fig 8e)

Trace-element partitioningTo assess whether coexisting phases are in trace-elementequilibrium we calculated apparent cpx^garnet distribu-tion coefficients and compared them with those

(a)

(d)

(b)

(e)

(c)

GarnetCI-chondrite GarnetCI-chondrite

low-Mghigh-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1952

1

10

Nd Sm Eu Gd Dy

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Fig 7 Extended trace-element patterns of garnet in (a) low-Mg eclogites (b) high-Mg eclogites with inset showing enlargement of MREE tohighlight Eu anomalies and 1s uncertainties (c) high-Ca eclogites (d) pyroxenites and (e) volatile-bearing eclogites Normalized to chondritevalues of McDonough amp Sun (1995)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1855

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

determined experimentally or in demonstrably equili-brated natural samples Low-Mg and high-Mg eclogiteshave Dcpxgarnet for most elements that are within therange determined for natural equilibrated eclogites andexperimental assemblages (Harte amp Kirkley 1997 Greenet al 2000 Barth et al 2002a) (Electronic Appendix 7)High-Ca eclogites plot at the low end of the range

of distribution coefficients and have very low DMREE

and DHREE outside the range indicated by the datasetof Harte amp Kirkley (1997) who noted that DREE decreaseswith increasing Ca content in eclogitic minerals Theseassemblages are therefore considered to be in equilibriumPyroxenite VR40384 has D values similar to eclogites

but at the high end of the range determined by Harte amp

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

(a)

(d)

(b)

(e)

(c)

CpxCI-chondrite CpxCI-chondrite

low-Mg high-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1914

YK1952

spongy rim

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

VR40345

Fig 8 Extended trace-element patterns of cpx as in Fig 7

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1856

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Kirkley (1997) (Electronic Appendix 7) Another group(YK1914 and YK1952) plots at the low end but DHf DZrDTi and DV are far below the values indicated by experi-mental datasets Because pyroxenites compositionallygrade into eclogites their drastically different DHFSE

values suggest disequilibrium In contrast volatile-bearingsamples have DHFSE values that appear to be too high to bein equilibrium however this may be a consequence oftheir unusual bulk chemical composition (Ti- and K-rich)which differs substantially from that of lsquonormalrsquo eclogitesand experimental charges

GEOTHERMOBAROMETRYFollowing Schmidberger et al (2007) temperatures oflast equilibration were calculated for eclogitic andopx-free pyroxenitic assemblages using the formulationof Krogh (1988) The results are summarized in Table 2and Electronic Appendix 8 for a preset pressure of5 GPa and illustrated as a histogram in Fig 9 Foropx-bearing pyroxenites temperatures were calculatedsimultaneously with pressure [Ca-in-opx thermometryand Al-in-opx barometry using the formulations of Breyet al (1990)] to constrain their relative positions in themantle columnMinerals in some eclogites from Lac de Gras show dis-

tinct core and rim compositions or inhomogeneitiesTemperatures were not calculated for samples withspatially irresolvable inhomogeneous garnet compositions(SE01VR43479VR43480) For eclogites with spongy cpxcore compositions were used in the calculationTemperatures were calculated for both core and rim com-positions where applicable For all samples these

temperatures agree within the uncertainty of this thermo-meter (508C Brey amp Kolaquo hler 1990)Most equilibration temperatures for eclogites lie between

800 and10008C (Fig 9)The transition between the shallowand the deep lithospheric mantle layers in the central Lacde Gras area lies in this temperature interval (Griffin et al1999 Pearson et al 1999) Eclogites in this study lack thedistinct bimodal temperature distribution observed bySchmidberger et al (2007) but as in their study high-Caeclogites give higher average temperatures (10808C corre-sponding to a depth of 155 km if equilibrated along a 40mWm2 geotherm) than other eclogites types (9808Cdepth 135 km) the diamond-bearing sample (VR40345)gives thehighest temperature of the datasetHighNa2Ocon-tents in garnet from high-Ca eclogites are qualitativelyconsistent with higher equilibration pressures (Sobolev ampLavrentrsquoyev 1971) as are high K2O contents in coexistingcpx (Erlank ampKushiro1970) suggesting that high-Ca eclo-gites not only equilibrated at high temperature but also highpressure rather than representing a temperature excursionPyroxenites both opx-bearing and opx-free give somewhatlower average temperatures than eclogites (mean 9258C)and phlogopite-rich eclogites give the lowest temperatures(685 and 7698C respectively) The calcite-bearing eclogiteVR67112b gives a relatively low equilibration temperatureof 8558C

MEASURED ANDRECONSTRUCTEDWHOLE -ROCKSRationale and approachKimberlite-borne xenoliths sometimes show evidence forinfiltration of the host kimberlite (eg Taylor amp Neal1989 Barth et al 2001 2002b Heaman et al 2006) Wetherefore follow the standard approach of calculatingwhole-rock compositions from the compositions of the con-stituent phases weighted by their modal abundances Onlycore compositions were used as distinct rim compositionsmay point to late diffusion of elements from kimberlite-related melt veinsMany eclogitic whole-rock compositions reconstructed

with only garnet and cpx have negative Ti anomalies(not shown) For most of these samples rutile or ilmenitehas been identified as part of the mineral assemblage andthese phases should therefore be considered in the whole-rock reconstruction Rutile or ilmenite modes were esti-mated by assuming that the bulk-rock has no Ti anomalywhen normalized to a normal mid-ocean ridge basalt(N-MORB) composition and calculating the amount ofrutile required to erase theTi anomalyIt is of course possible that the eclogites had negative

or positive Ti anomalies relative to elements of similarcompatibility (Eu and Dy) Because we reconstructed

0

2

4

6

8

10

12volatile-rich

pyroxenite opx-bearing

pyroxenite opx-free

high-Ca

high-Mg

low-Mg

700750

800850

900950

10001050

11001150

12001250

Fre

quen

cy

Temperature

Fig 9 Histogram ofTKrogh (Krogh 1988) obtained at a preset pres-sure of 5 GPa except opx-bearing pyroxenites where TBKN wassolved simultaneously with PBKN (Brey amp Kolaquo hler 1990)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1857

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

whole-rock compositions from mineral core compositionskimberlite infiltration is unlikely to have caused fractiona-tion of Ti from Eu and Dy In a study in which rutilemodes were estimated from the differences between Ti inmeasured bulk-rocks and in bulk-rocks reconstructed withonly garnet and cpx only three of 19 eclogites recon-structed with rutile show Ti anomalies (Barth et al2002a) Even eclogite xenoliths with depleted Zr and Hfabundances that resemble those of modern island arcbasalts do not show Ti anomalies (Jacob amp Foley 1999)We therefore consider our approach to be a reasonablyrobust means of obtaining meaningful rutile modes wheresmall sample sizes preclude reliable determination of themodes of accessory phases by point-countingWhere rutilewas not recognized on thin-section scale or not analysedthe average eclogitic rutile composition was used for thereconstruction

UncertaintiesReconstructing whole-rock compositions in this way is arather crude method given the coarse grain size of theeclogites relative to the sample size and the non-uniformdistribution of the principal minerals on thin-sectionscale The uncertainty in mineral modes of medium- tocoarse-grained eclogites with grains45mm has been esti-mated by Jerde et al (1993) to be 10 plus uncertaintiesrelated to partial alteration which is minor in our samplesNevertheless the whole-rock trace-element budget israther insensitive to the exact modes for common eclogiticassemblages and REE patterns which encompass mostother elements in terms of compatibility in garnet andcpx do not change significantly even for modal variationsof 30 (Jerde et al 1993) This is not true for major-element concentrations which vary substantially with var-iations in mineral modes and we therefore do not calculatewhole-rock major-element contentsBecause the discussion that follows is based on trace-

element patterns and trends rather than absolute abun-dances whole-rock reconstruction based on mineralmodes is adequate We are confident that the uncertaintyinherent in this approach is not leading to significant mis-representation of eclogite whole-rock compositions becausethe trace-element patterns of the reconstructed whole-rocks show distinct differences (see below) that correlatewith the assignment to eclogite types based on garnetmajor-element composition and this is independent ofmineral mode Following Jerde et al (1993) in ElectronicAppendix 9 we show variations in trace-element patternfor one sample in each eclogite type for variations ingarnet and cpx abundance from 7030 to 3070 vol This shows that the trace-element patterns are generallyinsensitive to cpxgt ratios with regard to slopes andanomalies such as positive Sr Pb and Eu anomalies withthe exception of the positive Eu anomaly in high-Mgeclogite VR43479 which disappears if a high cpx mode is

assumed (not shown) Variations in rutile modes by 50(Electronic Appendix 9d) lead as would be expected tothe appearance of Ti anomalies but a rutile mode of 05instead of 1 vol does not erase the positive Nb anomalyin low-Mg eclogiteVR50909Because of the robustness of the relative trace-element

abundances in the reconstructed eclogites a garnet propor-tion of 065 and cpx proportion of 035 approximately themedian modal abundances of these minerals in volatile-free eclogites from Lac de Gras are assumed for eclogitesamples that are too small for determination of meaningfulmodes by point-counting

Within- and between-group variationsThe trace-element compositions of reconstructed whole-rocks are given in Table 4 N-MORB-normalized trace-element patterns of reconstructed whole-rocks areshown in Fig 10 The three eclogite groups have distincttrace element patterns that are particularly obvious forthe REE (Fig 11) Low-Mg eclogites have positive Pb andSr anomalies and smooth positive slopes in their MREEN

and HREEN LREEN are enriched relative to MREEN inthree samples (Fig 11a) Three samples have high Nbrelative to similarly compatible elements and comparedwith most other eclogites (Fig 10a) All high-Mg eclogitesdisplay more or less pronounced positive Eu and Sranomalies and most have positive Pb and negative Zrand P anomalies Their N-MORB-normalized REE pat-terns show shallow positive slopes from LaN to PrN orNdNand aweak negative slope from NdN to LuN (Fig11b)VR43479 is a low-Mg eclogite with regard to its garnet

Ca^Fe^Mg composition but has a trace-element patterntypical of high-Mg eclogites and will be discussed withthe latter Conversely sample YK1949 a high-Mg eclogitein terms of its garnet Ca^Fe^Mg compostion is more simi-lar to low-Mg eclogites in terms of its whole-rock trace-element pattern and is plotted in Figs 10a and 11aWhereas the patterns of most samples within the groupresemble each other the N-MORB-normalized incompati-ble-element abundances of the high-Ca eclogites vary bymore than an order of magnitude (Fig 10c) All sampleshave positive Pb anomalies and most have positive Sranomalies Some samples have negative P andor Zranomalies Niobium abundances in three of the samplesare very high similar to those for some low-Mg eclogitesHigh-Ca eclogites tend to have flatter MREEN andHREEN patterns and more uniform HREE abundancesthat the other eclogite types with humps between SmN

and HoNThe phlogopitized sample YK1911 shows a strong posi-

tive Pb anomaly and lower HREE abundances than thecalcite-bearing sample VR67112b which shows a distinctnegative Eu anomaly and a depletion of La relative to Ce(Figs 10d and 11d)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1858

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Trace-element compositions of reconstructed whole-rocks (ppm)

Sample P Sc Ti V Co Ni Ga Sr Y Zr Nb La Ce Pr Nd

Low-Mg eclogites

vr19674-g4 186 50 4649 317 57 59 144 77 24 30 36 13 31 05 24

vr43452 154 39 2253 328 49 84 176 52 32 11 01 02 07 01 08

vr43469 38 65 4377 349 54 91 142 34 28 15 02 05 15 03 17

yk1926 144 45 3406 258 52 87 145 89 18 18 20 09 29 06 33

yk1943 173 46 5732 295 41 39 146 122 25 48 82 34 85 13 64

yk19491 102 34 1693 158 31 49 100 46 10 14 03 09 21 03 17

High-Mg eclogites

vr19677-g2 138 39 5160 291 41 53 119 98 14 22 04 13 53 11 64

vr434792 46 25 1265 129 36 126 84 56 3 7 04 06 25 04 23

vr67360 49 36 1573 138 64 184 111 76 5 9 04 09 38 07 33

yk1946 153 28 4877 192 45 105 124 108 12 48 03 10 38 08 50

High-Ca eclogites

vr19674-g5 267 30 7663 200 41 67 195 32 23 43 77 01 04 02 16

vr40345 0 33 2821 175 47 98 210 179 13 13 66 60 109 11 38

vr43465 142 27 5104 220 31 83 121 197 14 35 29 05 22 05 40

vr50858 295 30 6898 201 44 94 150 48 22 38 06 01 03 02 20

vr50860 68 57 5672 418 57 71 186 109 19 11 02 02 13 03 26

vr50909 238 28 7295 201 42 92 171 48 20 57 13 01 04 02 21

yk3528 91 51 3712 341 53 155 177 49 16 9 01 02 08 02 11

Volatile-rich eclogites

vr67112b 40 84 9242 521 46 30 107 4 40 66 31 26 126 24 125

yk1911 31 39 9428 286 44 82 164 3 23 52 64 18 52 09 46

Pyroxenites

yk1914 111 49 707 207 48 547 44 31 4 3 02 06 19 03 10

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

Low-Mg eclogites

vr19674-g4 11 05 10 22 39 10 29 33 06 12 05 028 24 008

vr43452 04 02 08 13 38 12 41 49 08 05 01 nd 02 na

vr43469 11 05 09 25 44 11 34 36 06 07 02 001 09 129

yk1926 12 05 11 19 30 07 22 24 04 07 03 004 23 090

yk1943 21 08 10 29 42 10 30 32 05 16 05 013 65 017

yk19491 06 03 11 08 14 04 14 18 03 04 02 003 29 115

High-Mg eclogites

vr19677-g2 20 08 12 21 24 05 15 15 03 05 05 007 52 6811

vr434792 05 02 12 05 06 01 04 03 00 03 01 001 98 037

vr67360 06 03 14 06 09 02 06 06 01 02 02 002 86 000

yk1946 18 07 12 21 22 05 13 12 02 11 03 006 56 112

High-Ca eclogites

vr19674-g5 17 09 11 33 43 09 24 23 04 16 05 007 02 011

vr40345 09 04 10 12 22 06 16 14 02 05 nd nd 298 na

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1859

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Continued

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

vr43465 22 09 12 25 26 06 15 14 02 10 04 001 25 040

vr50858 20 08 09 33 41 08 22 21 03 10 01 nd 02 na

vr50860 14 07 11 27 37 08 22 18 03 06 00 nd 06 na

vr50909 20 10 12 33 38 07 20 20 03 16 01 002 02 181

yk3528 08 05 11 20 30 07 18 18 02 05 01 nd 08 na

Volatile-rich eclogites

vr67112b 38 10 07 53 74 15 41 34 04 18 03 012 60 004

yk1911 20 07 08 35 39 07 18 17 02 10 14 nd 73 na

Pyroxenites

yk1914 02 01 11 04 08 02 05 06 01 01 03 003 68 101

aGarnet has CandashMgndashFe relationship typical of high-Mg eclogites but whole-rock trace-element pattern is typical oflow-Mg eclogitesbGarnet has CandashMgndashFe relationship typical of low-Mg eclogites but whole-rock trace-element pattern is typical ofhigh-Mg eclogites

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

(a)

001

01

1

10(c)

001

01

10(b)

VR43452

YK1949

RockNMORB

low-Mg

RockNMORB

high-Mg

(d)

volatile-richhigh-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

phlogopite-bearingYK1911

calcite-bearingVR67112b

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 10 Extended trace-element patterns of reconstructed (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-bearing eclogites Normalized to N-MORB of Sun amp McDonough (1989)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1860

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

RockNMORB

(b)

RockNMORB

(d)

010

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu001

01

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

(a)

001

01

1

10(b)

001

01

1

10(c)

001

01

1

10(e)

high-TSchmidberger et al

low-TSchmidberger et al

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

high-Mglow-Mg

high-Cavolatile-rich

phlogopite-bearingYK1911

calcite-bearingVR67112b

YK1949VR43479

Fig 11 N-MORB normalized REE patterns for (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-richeclogites Shown for comparison in (e) are reconstructed eclogites from Schmidberger et al (2007)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1861

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Both volatile-rich whole rocks display strong negativeSr anomalies Strontium can be sited in calcite or apatiteandor indicate the presence of a Sr-rich mineral that wasnot identified in thin section such as bariteFour of the six reconstructed whole-rocks in the study

of Schmidberger et al (2007) (Fig 11e) show markedLREEHREEN enrichment that is observed only in onehigh-Ca eclogite in our study (Fig 11c) suggesting thatmany of the eclogites they studied were affected by metaso-matic LREE enrichment Two samples in their study thatare not affected by this enrichment show positive Euanomalies similar to the high-Mg eclogites in this studywhereas the distinctive REE pattern of the low-Mg eclo-gites is not recognized among their samples

SR^ND^HF I SOTOPESYSTEMATICSStrontium Nd and Hf isotope data for optically puremineral separates and reconstructed whole-rocks aregiven in Table 5 Spongy rims and alteration are likely tobe mostly removed during leaching in 6N HCl in combina-tion with repeated ultrasonication Despite this rigoroustreatment the reproducibility of Nd in garnet VR43469and in cpx YK1914 is worse than for other samples Thismay be due to inhomogeneities related to late Nd diffusionpossibly from grain boundaries containing a kimberlitecomponent The reproducibility of the 176Hf177Hf data isgenerally poorer than that of Nd which may reflect a com-bination of matrix effects caused by imperfectly purifiedsolutions and sample heterogeneity Parent^daughterratios are available only from in situ analyses which arenot precise enough to allow meaningful model age orinitial isotope ratio calculations (ie at the time of kimber-lite eruption) Because the Lac de Gras kimberlites arerelatively young (55^56 Ma Graham et al 1999) age cor-rection would result in only a minor change in isotopecomposition which is inconsequential in view of the rangeof compositions displayed by most samples and hence hasno effect on the conclusions drawnClinopyroxenes in five volatile-free eclogites and a

pyroxenite have Sr isotope ratios ranging from 0701662 to0703546 this range is similar to or somewhat lower thanthat of the depleted mantle at the time of kimberliteemplacement (0703 Bell amp Tilton 2002) Clinopyroxenein the phlogopitized sample has a distinctly high 87Sr86Srof 0751186Clinopyroxene controls 56^92 of the Nd budget (Nd

concentration in cpx weighted by mode) Neodymiumisotope ratios range from 0511660 to 0514844 Thiscorresponds to eNd variation (per 10 000 deviationof 143Nd144Nd in the sample from 143Nd144Nd in the chon-dritic mantle model reservoir) from 17 to thorn43 cpx inthe low-Mg eclogites have higher eNd than those in

high-Mg eclogites Garnets have generally have higher143Nd144Nd from 0512984 (eNdfrac14thorn 7) to 0520965(eNdfrac14thorn 162) Garnets in the two volatile-rich sampleshave the highest ratios 0522790 for garnet in the calcite-bearing sample VR67112b (eNdfrac14thorn198) and 0525721 inthe phlogopitized eclogiteYK1911 (eNdfrac14thorn 255)Garnet contributes 18^32 and cpx 48^78 of the Hf

budget The proportion controlled by rutile is 4^30(Aulbach et al in preparation) Measured 176Hf177Hf incpx ranges from 0281263 in phlogopite-bearing sampleYK1911 (eHffrac1453) to 0286130 in low-Mg eclogiteVR43469 (eHffrac14thorn119) and in garnet from 0283958 inpyroxeniteYK1915 (eHffrac14thorn42) to 0295718 in low-Mg eclo-gite YK1926 (eHffrac14thorn458) Hafnium in garnet in the phlo-gopite^apatite-bearing sample is extremely radiogenic(176Hf177Hffrac14 0354367) corresponding to an eHf of thorn2532which is close to a value determined for a garnet separatefrom a Roberts Victor eclogite (eHffrac14thorn2561 Jacob et al2005) Garnet in the calcite-bearing sample has extremelyradiogenic Nd and Hf (eNdfrac14thorn193 eHffrac14thorn2516) Garnetin pyroxenite YK1915 also has radiogenic Nd and Hf(eNdfrac14thorn27 eHffrac14thorn42)Despite the large uncertainties for reconstructed whole-

rocks (calculated as the square root of the sum of squareduncertainties of two standard errors for garnet and cpxanalyses combined with 20 propagated uncertainty onthe proportion of Nd or Hf respectively contributed byeach mineral) the isotopic data are consistent in thatthey correlate with the independently determinedparent^daughter ratios (1) Low-Mg eclogites have higherSmNd reflected in the steeper positive slope of theirN-MORB-normalized REE patterns (Fig 11a) comparedwith the high-Mg eclogites (Fig 11b) Accordingly recon-structed low-Mg eclogites have higher present-day eNd

(0 to thorn47) than high-Mg eclogites (4 and 17) (2) Low-Mg eclogite YK1943 has very high Nb and LREE abun-dances that could have been added during metasomatismthat also enhanced its Hf abundance This sample has thelowest eNd and eHf of the low-Mg eclogites (0 and thorn20respectively) In contrast apparently unenriched low-Mgeclogites have more radiogenic Nd and Hf (eNdfrac14thorn47and thorn9 eHffrac14thorn128 and thorn113 respectively)The Sr^Nd isotope diagram reveals that eNd in cpx

varies from 19 to 43 at nearly constant depleted-mantle-like 87Sr86Sr (Fig 12a) In the Nd^Hf isotope diagram(Fig 12b) the reconstructed whole-rocks lie on curvedmixing lines if the lsquomixingrsquo proportions of Nd and Hffor the two lsquoend-memberrsquo minerals are very different(eg YK1911) Garnets have more radiogenic Nd andHf than coexisting cpx All reconstructed volatile-freeeclogites plot in the radiogenic field with regard to Hfor are isotopically similar to the depleted mantle Twohigh-Mg eclogites and one low-Mg eclogite have unradio-genic Nd one high-Mg eclogite has radiogenic Nd and

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1862

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 5 Individual and averaged Sr Nd and Hf isotope ratios obtained from garnet and cpx solutions

Sample 87Sr86Sr 2SE2SD 143Nd144Nd 2SE2SD eNd 2SE2SD 176Hf177Hf 2SE2SD eHf 2SE2SD

Low-Mg eclogites

VR43469 cp 0702661 0000010 0514844 0000006 43 01 0286130 0000012 119 04

VR43469 gt 0515508 0000014 56 03 0286506 0000012 132 04

VR43469 gt r 0515453 0000010 55 02 0286548 0000011 134 04

VR43469 gt r 0515384 0000007 54 01 0287244 0000010 158 04

VR43469 gt r 0515487 0000014 56 03 0286996 0000014 149 05

VR43469avg gt 0515458 0000109 55 21 0286823 0000715 143 253

WR VR43469 0515059 0000273 47 5 0286405 0001079 128 38

YK1926 cp 0703546 0000011 0513043 0000007 8 0 0283154 0000009 13 03

YK1926 gt 0513364 0000008 14 01 0295718 0000005 458 02

WR YK1926 0513120 0000026 9 1 0285980 0000716 113 25

YK1943 cp 0512143 0000007 10 01 0282213 0000005 20 02

YK1943 gt 0514461 0000012 36 02 0287510 0000008 168 03

YK1943 gt r 0514491 0000009 36 02 0287332 0000005 161 02

YK1943 gt r 0514517 0000010 37 02 0287290 0000007 160 02

YK1943avg gt 0514489 0000056 36 11 0287377 0000233 163 82

WR YK1943 0512625 0000228 0 5 0283138 0000879 13 31

High-Mg eclogites

VR43480 gt 0520965 0000017 162 03 0284950 0000010 77 04

VR67360 cp 0283509 0000074 26 26

VR67360 cp r 0283433 0000031 23 11

VR67360avg cp 0701662 0000018 0511660 0000010 19 02 0283471 0000108 25 38

VR67360 gt 0512984 0000101 7 20 0287429 0000180 165 64

WR VR67360 0511761 0000132 17 3 0284734 0000469 69 17

YK1949 cp 0703019 0000013 0512231 0000009 8 02 0282513 0000009 9 03

YK1949 gt 0513511 0000010 17 02 0290297 0000042 266 15

YK1949 gt r 0513508 0000012 17 02 0289947 0000015 254 05

YK1949 gt r 0513552 0000014 18 03

YK1949avg gt 0513523 0000049 17 10 0291569 0000494 311 175

WR YK1949 0512446 0000103 4 2 0284622 0000847 65 30

Volatile-rich eclogites

VR67112b gt 0522790 0000008 198 02 0354367 0000036 2532 13

YK1911 cp 0751186 0000009 0512165 0000007 9 01 0281263 0000009 53 03

YK1911 gt 0525721 0000013 255 03 0288051 0000012 187 04

WR YK1911 0513940 0000342 25 7 0281414 0000018 48 1

Pyroxenites

YK1914 cp 0703015 0000034 0513283 0000041 13 08

YK1914 cp r 0702956 0000005 0513200 0000027 11 05

YK1914 cp r 0513527 0000059 17 11

YK1914 cp r 0513045 0000010 8 02

YK1914avg cp 0702986 0000083 0513264 0000402 12 78

YK1915 gt 0514003 0000011 27 02 0283958 0000021 42

Analyses of full procedural replicates are designated with an lsquorrsquo after sample name SE standard error (internal errorsingle analyses) SD standard deviation (external error averaged multiple analyses) Italics used for reconstructed whole-rocks (mineral isotopic compositions and concentrations weighted by modes) Uncertainties for reconstructed whole-rocksinclude 2SE or 2SD uncertainties on each mineralrsquos isotope composition combined with 20 uncertainty in the amountof Nd or Hf contributed by the constituent minerals eNd and eHf are per 10 000 deviation from model chondritic mantle[model of Wasserburg et al (1981) and Blichert-Toft amp Albarede (1997) respectively]

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1863

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

εHf

εNd

Ekati eclogitesminus100

0

100

200

300

400

500

minus40 0 40 80 120 160 200 240 280

gt in YK1915

gt in VR67112bεNd=198 εHf=2532

minus40

0

40

80

120

160

minus40 minus20 0 20 40 60

minus60

minus40

minus20

0

20

40

070 071 072 073 074 075 076

North Americankimberlite field

εNd

87Sr86Sr

cpx in YK1914

Ekatieclogites

DM

DM

DM

DM

(a)

(b)

pyroxenites

low-Mg

high-Mg

volatile-rich

eclogites

Diavik low-TA154 S kimberlite

cpx in YK1911

Fig 12 (a) Sr^Nd for cpx and (b) Nd-Hf isotope diagram for garnet cpx and reconstructed whole-rocks (symbols with bold outline) Finedashed lines in (b) connect data points from the same sample with garnets having more radiogenic isotopic compositions Uncertainties forreconstructed whole-rocks combine 20 modal uncertainty with uncertainties on each mineralrsquos isotope composition Error bars on individualminerals may be smaller than the symbols Also shown are the isotopic compositions of the depleted mantle (DM DePaolo 1981 Griffin et al2000 Bell amp Tilton 2002) Range of 87Sr86Sr in eclogites from Ekati (Lac de Gras area 100 km north of Diavik Jacob et al 2003) in NorthAmerican kimberlites and compositions of low-temperature eclogites from Diavik and of Lac de Gras A154 South kimberlite fromSchmidberger et al (2007)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1864

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

one low-Mg eclogite has eNd similar to the depleted mantle(Fig 12b)Clinopyroxene in the phlogopitized eclogite YK1911

has unradiogenic Nd and Hf and strongly radiogenicSr whereas the coexisting garnet has radiogenic Nd andHf Although this sample contains 50 vol phlogopiteits whole-rock isotopic composition was calculated

from garnet and cpx alone because the contributionfrom phlogopite to the Hf and Nd budget is 52 and51 respectively (unpublished data) The whole-rockhas radiogenic Nd but has lower eHf than anyother sample analysed in accord with the extreme Hfabundance (53 ppm) and low LuHf of the cpx

DISCUSS IONOrigins of eclogites beneath the SlaveCratonHigh-Mg and high-Ca eclogites

With regard to their major- and trace-element contentsthe high-Mg and high-Ca eclogites from Lac de Gras arebroadly similar to variably evolved oceanic gabbros(Fig 13b and c) which may indicate an origin as subductedoceanic crust Compatible trace-element abundancesare useful to deduce potential crustal protoliths as theseelements are less susceptible to post-formation modification(Jacob 2004) The flat HREE patterns in the high-Mg andhigh-Ca eclogites are compatible with a low-pressureorigin as the oceanic crust is a product of moderate degreesof partial melting mainly within the spinel stability field(eg Presnall et al 2002) whereas melting with residualgarnet would qualitatively be expected to produce stronglyfractionated REE patterns Positive Sr and Pb anomaliesalthough compatible with a plagioclase signature couldalso be due to cryptic metasomatism (Dawson 1984) ofthe eclogites during their residence in the lithosphericmantle These anomalies are probably not related to kim-berlite infiltration as the whole-rock eclogite compositionswere reconstructed from analyses of fresh mineral coresLow ZrSm similar to that in primitive gabbros is

observed for many of the high-Mg and high-Caeclogites (Fig 13b and c) and is independent of the ratioof garnetcpx and whether or not the whole-rock wasreconstructed with rutile (Electronic Appendix 10) Thenegative trend of Mg-number against Yb in garnet(Fig 14) argues against igneous processes involving garnetaccumulation (Taylor amp Neal 1989) as suggested forexample for some high-Mg eclogites from West Africa(Barth et al 2002b) Rather the anti-correlation suggeststhat olivine fractionation played a role in the formation ofthe protoliths Fractionating olivine would excludeYb (Dolivinebasalt

frac14 002 Nikogosian amp Sobolev 1997)and incorporate Mg relative to Fe (Dfrac1415 and 05 respec-tively Taura et al1998) leading to progressive depletion ofMg and enrichment of Yb in the residual melt and henceanti-correlated Mg and Yb in the whole-rock This isinherited after eclogitization All eclogite types (exceptVR43452) plot along the olivine fractionation trend sug-gesting that they formed by a similar processHigh-Mg eclogites have lower

PREE than high-Ca

eclogites and some appear to have positive Eu anomalies

high-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

gabbrosSE Indian ridge

gabbrosOman ophiolite

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

high-Mg

RockNMORB(a)

VR43452

YK1949

low-Mg

boninite-like melts

01

001

1

10

(b)

01

001

1

10

(c)

01

001

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

Fig 13 Trace-element patterns of (a) low-Mg (b) high-Mg and(c) high-Ca eclogites compared with boninite-like volcanic rocksfrom the Isua greenstone belt (Polat et al 2002) gabbros from theOman ophiolite (Benoit et al 1996) and the SE Indian Ridge (Hartet al 1999) normalized to N-MORB of Sun amp McDonough (1989)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1865

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(EuEu up to 14) similar to more primitive cumulates(troctolite gabbros and olivine gabbros) reported in the lit-erature (Fig 13b) Although the sizes of these Eu anomaliesare not outside the analytical uncertainty their correlationwith CeYb in reconstructed eclogites supports their signif-icance (r2frac14066 nfrac14 4) The more primitive nature ofhigh-Mg eclogites is consistent with higher Mg-numbersrelative to high-Ca eclogites and may be due to olivineaccumulation during formation of their protolithsThe two samples with the lowest HREE abundanceshave the highest EuEu which is compatible with varia-tions in plagioclase modesMore evolved protoliths with higher pyroxeneplagio-

clase plus trapped melt could account for small or absentpositive Eu anomalies and higher

PHREE of high-Ca

eclogites (Fig 13c) Although the reason for the frequentassociation of graphite or diamond with high-Ca eclogitesis unclear (Aulbach et al 2003) the frequent occurrence ofkyanite in high-Ca eclogites also observed at Ekati justnorth of Lac de Gras has been argued to support a crustalprotolith (Jacob 2004) Kyanite andor corundum havebeen suggested to exsolve from high-temperature high-Alclinopyroxenes (Caporuscio amp Smyth 1990) Eclogitesfrom Lac de Gras show no exsolution textures suggestiveof such an origin although recrystallization can obscureor destroy microstructural evidence for exsolution(eg Griffin et al 1984)Modelling shows that qualitatively plagioclase accumu-

lation from a low-pressure melt derived from MORBmantle can account for positive Pb Sr and Eu and negativeZr anomalies relative to N-MORB whereas cpx accumula-tion or trapped melt increases Zr Y MREE and HREE

abundances (Fig 15) Some melt addition (for exampleby entrapment) seems to be required to increase all trace-element abundances to those observed for the eclogitesAlthough high-Ca and high-Mg eclogites could representshallow and deep portions of the same oceanic crust thehigher average temperatures obtained for the high-Caeclogites the frequent occurrence of carbon in high-Caeclogites and its absence in high-Mg eclogites the strongLREEHREE fractionation observed for six of seven high-Ca eclogites and the narrow range of compositions of thehigh-Mg eclogites suggest that they may have protolithsthat are not cogenetic Some discrepancies in particularaffecting the incompatible elements may be due to post-crystallization processes acting upon the protoliths suchas dehydration melting and later metasomatism that isevident in peridotites from the same locality (Griffin et al1999)Low K2OTh for high-Ca and high-Mg eclogites

(average 08 Table 4) relative to assumed crustal

0

2

4

6

8

10

30 40 50 60 70 80

Mg

VR43452

Yb

GARNET

pyroxenites

low-Mghigh-Mghigh-Ca

eclogites

VR19673ecl-g2

Fig 14 Garnet Mg-number vs Yb (ppm including 2s uncertainty)Lines qualitatively show separate correlations for eclogites (exceptoutlierVR43452) and pyroxenites (except outlierVR19673ecl-g2)

high-Ca

high-Mg

001

01

1

10

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

melt08 melt02 cryst cpx01 melt09 cryst

plag

(a)

001

01

1

10(b)

RockNMORB

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 15 Reconstructed (a) high-Ca and (b) high-Mg eclogitescompared with compositions of plagioclase and cpx crystallizingfrom a MORB melt modelled using distribution coefficients ofNorman et al (2005)The MORB was calculated as a 10 equilibriummelt of MORB pyrolite [composition from Kelemen et al (1993)except Pb Pr Sr Y and Lu which were calculated as residues from1 fractional melting of pyrolite from McDonough amp Sun (1995)using melt modes of Kinzler (1997) and distribution coefficients ofNorman et al (2005)] Also shown are mixtures of crystals (consistingof 50 plagioclase plus 50 cpx) and melt in the proportions 0901and 0208

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1866

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

precursors (eg gabbros SE Indian Ridge average 22Hart et al 1999) may reflect the preferred loss of fluid-mobile elements during dehydration (Becker et al 2000)although large uncertainties on these low-abundance ele-ments in the constituent minerals and reconstructedwhole-rocks makes this interpretation less than robustThe REE patterns of most high-Mg eclogites are too flatto represent melting residues as partial melting shouldproduce low LREEHREE (Electronic Appendix 10) Incontrast the marked LREE depletion of five of sevenhigh-Ca eclogites (LaLufrac14 02^08) contrasts with that ofassumed evolved gabbroic protoliths (LaLufrac1415^32Hart et al 1999) and can be modelled as a consequence ofpartial melting during subduction after eclogitization (iein the presence of garnet which retains HREE relative toLREE eg Barth et al 2002a) has taken place (Aulbachet al 2003)Our results are compatible with previous studies on

eclogite xenoliths from the Slave Craton includingthose from the same kimberlite that have linked thesesamples or some of these samples to Paleoproterozoicsubduction events (the 195^191Ga formation of theHottah terrane and the 18Ga formation of the GreatBear magmatic arc) in the Slave craton (Griffin et al1999 Kopylova et al 1999 Heaman et al 2002 2006Schmidberger et al 2005 2007)The compositional features discussed above although

compatible with crustal protoliths are not necessarilydefinitive of a subduction origin However if direct crystal-lization from basaltic melts at high pressure (ie mantledepths) is invoked instead specific conditions would haveto be met to satisfy the observations The Mg-number ineclogites from Lac de Gras are lower than those expectedof primary high-pressure mantle melts (eg Walter 1998OrsquoHara amp Herzberg 2002) requiring crystallization athigh pressures after olivine fractionation This originwould further require shielding from peridotitic wallrockwith which the fractionated melts would no longer be inequilibrium (Barth et al 2001) In such a setting the lowrelative abundances of fluid-mobile elements comparedwith those of less fluid-mobile elements and of LREErelative to HREE (Table 4) would reflect an expulsion offluids during crystallization of the protoliths and latermelt mobilization respectively The eclogites clearly donot represent garnet plus cpx cumulates which wouldrequire a positive correlation of Mg-number with Yb thatis at variance with the observed negative correlation ineclogitic garnets from this study (Fig 14)

Low-Mg eclogites

Garnets in low-Mg eclogites compositionally overlap withthose in massive eclogites from Jericho (NW of the Lacde Gras kimberlites) that were suggested to have a high-pressure intrusive origin based on the similarity of theirmajor-element compositions to those of Group A eclogites

fromsouthernAfrica forwhichamantleoriginhasbeenpro-posed (Kopylova et al1999) Garnets in low-Mg eclogites inthe present study have major-element compositions moresimilar to garnets in groupB rather thangroupAeclogitesThe positive slope in N-MORB-normalized HREE

abundances displayed by the low-Mg eclogites is clearlydistinct from other eclogite types and is similar to those ofboninite-like second-stage melts that are closely associatedwith ultramafic volcanics (Smithies et al 2004) (Fig 13a)The range of Mg-numbers in the low-Mg eclogites (55^66)overlaps with that of Archaean second-stage melts fromIsua (58^75) and Abitibi (52^82 Smithies et al 2004)These similarities might indicate that the low-Mg eclogiteshave an origin in a previously depleted mantle source suchas supra-subduction zone mantle Subchondritic ZrHf ofthe low-Mg eclogites contrasts with the supra-chondriticZrHf of the high-Mg and high-Ca eclogites (Aulbach et

al in preparation) and is consistent with previous cpx-controlled depletion in an arc mantle source This sourcecould have been subsequently remelted to generate theboninite-like precursor to the low-Mg eclogites as sug-gested for the protoliths of eclogites from Udachnaya(Jacob amp Foley 1999)Although some second-stage melts to which the low-Mg

eclogites are similar are closely associated with komatiitesand could contain a plume component the radiogenic ordepleted mantle-like Nd and Hf and unradiogenic Sr inthe low-Mg eclogites do not support plume involvement

Trace-element and Sr^Nd^Hf isotopeconstraints on post-formationmetasomatismSeveral low-Mg and high-Ca eclogites show negativeslopes in the LREE which may be indicative of interactionwith small amounts of LREE-enriched melt (Navon ampStolper 1987) amounting to cryptic metasomatism(ie not accompanied by crystallization of new phases) inthe nomenclature of Dawson (1984) High Nb abundancesrelative to elements of similar compatibility displayed bysome low-Mg and high-Ca eclogites (Fig 10) could resultfrom overestimation of rutile modes However high-CaeclogiteVR40345 is rutile-free and has been reconstructedwithout rutile its high Nb abundance is accompanied byLREE enrichment indicating that metasomatic additionmay be responsibleLow-Mg eclogites show a positive correlation of La with

Nb (Electronic Appendix 11a) indicating concomitantaddition of LREE and Nb This suggests that the metaso-matic agent was a solute-rich fluid or silicate silicic melt ashydrous fluids and carbonatites do not carry enough HFSEto affect the budget of eclogites (Green amp Wallace 1988Klemme et al 1995 Adam et al 1997 Stalder et al 1998)Of the samples that are enriched in Nb all but one (high-Ca eclogiteVR50909) show a simultaneous enrichment in

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1867

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Hf (Electronic Appendix 11b) Together these trace-element relationships suggest that both SmNd and LuHfwere lowered in these samples Their low parent^daughterratios contrast with higher ratios in unenriched samplesand resulted in the ingrowth of a range of Nd and Hf iso-tope compositions Compared with the low-Mg and high-Ca eclogites the high-Mg eclogites have a narrow range ofLa and Nb concentrations which correlates with a narrowrange in Nd and Hf isotope compositions (Fig 12)Precise model ages cannot be calculated because parent^

daughter ratios are available only from in situ analyseswhich have relatively large uncertainties Howeverevidence for long-term enrichment and depletion may bediscerned Low-Mg eclogites even those that are enrichedin LREE have relatively high SmNd as a result of havingoverall positive slopes in their REE patterns relative toMORB whereas the high-Mg eclogites have relativelyflat REE patterns and therefore lower SmNd (Fig 11)This translates into lower eNd for high-Mg eclogites thanfor low-Mg eclogites In contrast they show no systematicdifference with respect to 176Hf177Hf Because the low- andhigh-Mg eclogites are proposed to have different origins asdiscussed above and therefore did not necessarily have thesame initial isotopic composition and formation age theseisotopic differences cannot be used to estimate their agesA long-term coupling of trace-element and isotopic

abundances is also indicated by the fact that low-Mg eclo-gite YK1943 which has experienced Nb and LREE addi-tion has lower eNd (0) and eHf (thorn13) than low-Mg eclogiteVR43469 with higher LuHf and SmNd (eNdfrac14thorn47eHffrac14thorn128) If low-Mg eclogites YK1943 and VR43469originally had identical initial Nd and Hf isotope composi-tions and if the parent^daughter ratios are accurate andwere modified only once during their evolution shortlyafter metamorphism (when their isotopic compositionwould have been reset) this overprint can be calculatedfor both isotopic systems to have occurred at 17GaThis would suggest that Hf and Nd were added duringthe same event In contrast high-Mg eclogites VR67360and YK1949 would have had identical Ndisotope compositions at 10Ga whereas their Hf isotopecompositions are disturbed in that the sample with thehigher parent^daughter ratio has a lower 176Hf177Hf thanthe sample with the lower parent^daughter ratio Theseresults suggest that the LREE and HFSE enrichmentobserved in different samples occurred during separateevents that did not affect the eclogites equallyExtreme HFSE enrichment leading to the growth of

zircon and rutile and LREE addition leading to thegrowth of apatite in eclogites from the Jericho kimberlitein the northern Slave craton have been ascribed to twoseparate metasomatic events at around 18Ga (quasi-coincident with the age of formation and metamorphism)and 13^10Ga respectively (Heaman et al 2002 2006

Schmidberger et al 2005) The authors associated theseevents with Paleoproterozoic subduction and aMesoproterozoic thermal disturbance in the Slave litho-spheric mantle possibly linked to the Mackenzie igneousevent Mesoproterozoic Nd model ages mostly between13 and 14Ga were reported for eclogites from thenearby Ekati kimberlites (Jacob et al 2003) Although wedo not observe zircon or apatite in volatile-free eclogitesfrom Lac de Gras and the age constraints obtained fromour data are weak they support a link between theseevents described for other eclogite populations and themetasomatic enrichment observed in our studyThe Sr isotope data in the present study are similar to

those of Schmidberger et al (2007) who suggested that thedepleted-mantle-like values are consistent with derivationof the protoliths from an oceanic mantle source However87Sr86Sr in volatile-free eclogites in the present study isremarkably constant considering their Nd and Hf isotopevariability (Fig 12) and the variability of trace-elementpatterns (Fig 10) Given the evidence for dehydration andmelting in many of the eclogites fractionation of highlyincompatible Rb from Sr and differential ingrowth ofradiogenic Sr would be expectedThis may indicate a rela-tively recent overprint and considering the similarityof 87Sr86Sr to present-day depleted mantle which is signif-icantly lower than 87Sr86Sr of North American kimber-lites a juvenile (asthenospheric) source

PyroxenitesPyroxenites from Lac de Gras often show relationshipsin their major-element contents that are different fromthose observed in the eclogites For example pyroxeniticgarnets show a negative correlation between CaO andNa2O contents whereas these oxides are positively corre-lated for garnets in eclogites (Fig 5) CaO is well correlatedwith C2O3 in pyroxenitic garnets but not in eclogitic gar-nets (Fig 4) Mg-number is positively correlated withYb abundance for pyroxenitic garnets but anti-correlatedfor eclogitic garnets (Fig 14) This suggests that the pyrox-enites in the present study are not related to the eclogitesby fractionation or accumulation processes as proposedfor the origin of some pyroxenites (Foley et al 2003)Pyroxenites have also been interpreted as the products

of crystal segregation in magma channels with alteredoceanic crust as the magma source this would explaintheir Eu anomalies and stable and radiogenic isotopecompositions (Pearson et al 1993 Becker 1996) As thehigh-pressure liquidus phases of most terrestrial basaltsare cpx and garnet (OrsquoHara amp Herzberg 2002) partialmelts of eclogites with basaltic composition wouldcrystallize cpx and garnet but not opx More than half ofthe pyroxenites from Lac de Gras contain opx suggestingthat they probably did not crystallize directly fromeclogite-derived melts

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1868

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Low-Cr websterites and pyroxenites could be crystalli-zation products of mafic melts that interacted with perido-titic mantle thus becoming more refractory The positivecorrelation of Yb and Mg-number in garnet would allowan origin by fractional crystallization from such meltsAlternatively they formed by reaction of silicic (possiblyeclogite-derived) melts with mantle peridotites(eg Kelemen et al 1993 Aulbach et al 2002) This agreeswith the similarity of hypothetical melts in equilibriumwith pyroxenitic garnets in most samples to eclogite-derived melts (Electronic Appendix 12) Niobium concen-trations in all hypothetical melts are much higher than ineclogite-derived melts regardless of whether or not rutileis assumed to be in the residue The same is true for LaCe and Zr concentrations in some samples which mayindicate that NbZr and LREE were added to thepyroxenites some time after their formation as discussedin the previous section for some eclogites Because mini-mum ages for pyroxenitic rutile with an unradiogenicHf isotope composition (calculated by projecting the176Hf177Hf to the model depleted mantle isotopic evolutioncurve assuming zero ingrowth) range to 178066Ga(Aulbach et al in preparation) it appears likely that thepyroxenites were not formed recently (eg during kimber-lite-related or precursory melt migration) and that some ofthe samples are at least MesoproterozoicGarnet pyroxeniteYK1952 consists of schlieren of equili-

brated garnet pyroxenite in a lsquomatrixrsquo of spongy cpxfine cpx^opx intergrowths embayed garnet and meltpatches (Fig 2f) Garnet in the pyroxenite schlieren hasvery low LREE and HFSE abundances Orthopyroxenein the matrix has distinctly higher Mg-number andCr-number and contains more CaO than opx from thepyroxenite schlieren whereas opx rims in the schlierenare compositionally similar to opx in the matrix The con-siderable compositional and textural disequilibrium inthis sample suggests that this is a young feature causedby reaction with fluids shortly before or at the time ofentrainmentAs a consequence of the small sample sizes of most of

the pyroxenites studied here a comprehensive isotopicdataset is not available Clearly more such data areneeded in addition to stable isotope data to obtain betterconstraints on the origins of this rock type Consideringthe microstructural and compositional variability ofthe pyroxenites and major-element relationships oppositeto those observed for eclogites it appears certain that pyr-oxenites form by a variety of processes some of which areunrelated to the formation of eclogites and their protoliths

Volatile-rich eclogitesVolatile-rich eclogites from Lac de Gras display the mostextreme Hf^Nd isotope systematics of all the analysedsamples Garnets in phlogopitized (YK1911) and in cal-cite-bearing eclogite (VR67112b) have the most radiogenic

Nd (eNdfrac14 2486 and 1927 respectively) Garnet inVR67112b has extremely high eHf (2516) Clinopyroxene inYK1911 has the lowest eHf measured for any mineralseparate from Lac de Gras (53) in accord with the veryhigh Hf abundance in its cpx (5 ppm) In contrastgarnet in this sample has radiogenic Hf (eHffrac14 136) Thisdiscrepancy is at least in part due to their relatively lowequilibration temperatures (7698C and 8558C respec-tively) which allowed the isotopic compositions of cpxand garnet to evolve separately at their low andhigh parent^daughter element ratios respectively forextended periods of time and possibly since formationAccordingly the cpx in YK1911 gives a minimum Hf ageof 27GaSlab-derived fluids could have been introduced during

the 27Ga collision of the western Slave Craton domainincluding the thick plume-derived deep lithosphericmantle layer and the eastern domain with its highlydepleted lithosphere This would also be consistent withthe observation that volatile-rich eclogites occur at shallowdepths in the mantle (TKroghfrac14 685^8558C correspondingto depths of 85^115 km along a 40 mWm2 geotherm)where the bulk of dehydration and production of solute-rich melts in a subducting slab might be expected to occur(Becker et al 2000 Hermann et al 2006) Their negativeEu anomalies might indicate that the fluid was derivedfrom the shallow portion of the slab that experiencedplagioclase fractionation The presence of carbonate inVR67112b requires a carbonated sourceHigh 87Sr86Sr in cpx in YK1911 is probably the conse-

quence of isotopic equilibration with high-RbSr phlogo-pite after long-term radiogenic ingrowth while Hfaddition led to retarded ingrowth of radiogenic HfThese systematics combined with radiogenic Nd appear

inconsistent with the involvement of sediments in the pet-rogenesis of this sample (eg Perfit et al 1980 Vervoortet al 1999) or a hydrous fluid such as those responsible forMARID metasomatism in the Kaapvaal craton (Konzettet al 1998)Whatever the origin of the fluids the apparentHf mobility suggests that they were not in equilibriumwith a rutile-bearing source and also that they weresolute-rich in nature (ie hydrous melts able to dissolveHFSE) and the relatively high HREE abundances of thevolatile-rich eclogites indicate a garnet-free source suchas amphibolite This allowed evolution of the sample atrelatively high time-integrated SmNd

SUMMARY AND CONCLUSIONS

(1) The Lac de Gras kimberlites carried abundant eclo-gite and pyroxenite xenoliths to the surface Threemain eclogite types are distinguished in terms ofmajor-element composition and trace element

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1869

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

patterns high-Mg high-Ca and low-Mg eclogitesThe garnets of all eclogite types have anti-correlatedMg-number and Yb pointing to olivine fractionationas a controlling factor in the formation of their pre-cursor rocks Their flat or positively sloped HREEpatterns qualitatively indicate that the melt parentalto the precursor rocks was not in equilibrium with agarnet-bearing residue

(2) Reconstructed high-Mg eclogites and high-Caeclogites have flat HREE patterns positive Pb andSr anomalies and low ZrSm similar to oceanic gab-bros reported in the literature High-Mg eclogiteshave variably low HREE contents and some havesmall positive Eu anomalies possibly suggestive ofmore primitive relatively plagioclase-rich protolithswhereas high-Ca eclogites have higher HREE abun-dances and small or absent Eu anomalies that requiremore cpx- andor melt-rich protoliths if a subductionorigin is assumed

(3) Low-Mg eclogites have distinctive trace-element pat-terns that show a marked positive slope in the HREEand subchondritic ZrHf They resemble someArchaean second-stage melts (lsquoboninitesrsquo) and theirprotoliths appear to have formed from a previouslydepleted source

(4) After formation the eclogites were subject to a varietyof processes Most have low ratios of fluid-mobile vsless fluid-mobile elements possibly as a result of dehy-dration of the protoliths In addition some of the low-Mg and high-Ca eclogites have low LREEHREEconsistent with partial melt lossWhereas the low-Mgeclogites show concomitant enrichment in La Nb andZr some high-Mg and high-Ca eclogites show evi-dence for variable and unsystematic (decoupled)LREE and Nb addition

(5) Isotopically the trace-element characteristics of thedifferent eclogite types translate into lower eNd forhigh-Mg eclogites (negative slope in MREEfrac14 lowerSmNd) than for low-Mg eclogites (positive slope inMREEfrac14higher SmNd) Within the low-Mg groupsamples that show evidence for metasomatism (lowLuHf and SmNd) have lower eNd and eHf than asample that was apparently not metasomatized BothNd and Hf addition can be dated to 17Ga if identi-cal initial isotope compsitions are assumed for thelow-Mg eclogites In contrast the relatively constantand depleted mantle-like 87Sr86Sr of eclogitic cpxdespite variable Nd and Hf isotope compositions andparent^daughter ratios suggests recent input of juve-nile material

(6) Garnet and cpx in pyroxenites show major-elementrelationships different from those in eclogitesTherefore these two rock types are probably notrelated by fractionation processes The presence of

opx precludes direct crystallization from slab-derivedmelts Pyroxenites may be the product of (slab-derived) melts hybridized during assimilation of peri-dotitic mantle

(7) Garnets in phlogopite- and carbonate-bearing eclo-gites have extreme isotopic compositions includingthe most radiogenic Nd and Hf in the datasetNegative Eu anomalies may point to the involvementof a crustal protolith Highly radiogenic Sr radiogenicNd and unradiogenic Hf in the reconstructedphlogopite-bearing sample combined with highHREE abundances and evidence for Hf mobilitymay indicate involvement of solute-rich fluids origi-nating in a garnet- and rutile-free source

ACKNOWLEDGEMENTSHelp in the clean laboratories and with the analyticalfacilities by S Elhlou C Lawson A Sharma P Wielandand M Zhang is greatly appreciated Bill Griffin isthanked for his support and for agreeing to reviewseveral versions of the manuscript despite disagreeingwith many of the conclusions presented here The paperbenefited greatly from the detailed and constructivereviews provided by the editors Gareth Davies andMarjorie Wilson and by Matthias Barth Larry Heamanand an anonymous reviewer This work was funded by aMacquarie University International PostgraduateAward and Postgraduate Research Fund (SA) by anARC SPIRT grant sponsored by Kennecott Canada Incand by ARC Large and Discovery Grants to SYOrsquoRAnalytical data were obtained using instrumentationfunded by ARC LIEF and DEST Systemic InfrastructureGrants industry partners and Macquarie UniversityThis is Publication 481 from the ARC National KeyCentre for Geochemical Evolution and Metallogeny ofContinents (wwwesmqeduauGEMOC)

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online

REFERENCESAdam J GreenT H Sie S H amp Ryan C G (1997) Trace element

partitioning between aqueous fluids silicate melts and mineralsEuropeanJournal of Mineralogy 9 569^584

Aulbach S Stachel T Viljoen K S Brey G P amp Harris J W(2002) Eclogitic and websteritic diamond sources beneath theLimpopo Beltccedilis slab-melting the link Contributions to Mineralogy

and Petrology 143 56^70Aulbach S GriffinW L Pearson N J OrsquoReilly S Y Kivi K amp

Doyle B (2003) Origins of eclogites beneath the central SlaveCraton Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA-0011

Aulbach S Griffin W L Pearson N J OrsquoReilly S Y amp Kivi K(2004a) Mantle formation and evolution Slave Craton constraints

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1870

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

from HSE abundances and Re^Os isotope systematics of sulfideinclusions in mantle xenocrysts Chemical Geology 208 61^88

Aulbach S Griffin W L OrsquoReilly S Y amp McCandless T E(2004b) Genesis and evolution of the lithospheric mantle beneaththe Buffalo HeadTerrane Alberta (Canada) Lithos 77 413^451

Aulbach S GriffinW L Pearson N J OrsquoReilly S Y amp Doyle B(2007) Lithosphere formation in the central Slave Craton(Canada) plume subcretion or lithosphere accretion Contributionsto Mineralogy and Petrology (in press) doi 101007s00410-007-0200-1

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2001)Geochemistry of xenolithic eclogites from West Africa Part IA link between low MgO eclogites and Archean crust formationGeochimica et Cosmochimica Acta 65 1499^1527

Barth M G Foley S F amp Horn I (2002a) Partial melting inArchean subduction zones constraints from experimentally deter-mined trace element partition coefficients between eclogitic miner-als and tonalitic melts under upper mantle conditions PrecambrianResearch 113 323^340

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2002b)Geochemistry of xenolithic eclogites from West Africa part 2Origins of the high MgO eclogites Geochimica et Cosmochimica Acta

66 4325^4345Beard B L Fraracci K NTaylor L A Snyder G A Clayton R

A Mayeda T K amp Sobolev NV (1996) Petrography andgeochemistry of eclogites from the Mir kimberliteYakutia RussiaContributions to Mineralogy and Petrology 125 293^310

Becker H (1996) Geochemistry of garnet peridotite massifs fromlower Austria and the composition of deep lithosphere beneatha Palaeozoic convergent plate margin Chemical Geology 134 49^65

Becker H Jochum K P amp Carlson R W (2000) Trace elementfractionation during dehydration of eclogites from high-pressureterranes and the implications for element fluxes in subductionzones Chemical Geology 163 65^99

Bell K amp Tilton G R (2002) Probing the mantle The story fromcarbonatites EOS Transactions American Geophysical Union 83275^277

Benoit M Polvecurren M amp Ceuleneer G (1996) Trace element andisotopic characterisation of mafic cumulates in a fossil mantlediapir (Oman ophiolite) Chemical Geology 134 199^214

BleekerW (2003) The late Archean record a puzzle in ca 35 piecesLithos 71 99^134

BleekerW Ketchum JW F JacksonV A amp Villeneuve M (1999)The central Slave Basement Complex Part I Its structural topol-ogy and autochthonous core Canadian Journal of Earth Sciences 361083^1109

Blichert-Toft J amp Albare de F (1997) The Lu^Hf isotope geochemis-try of chondrites and the evolution of the mantle^crust systemEarth and Planetary Science Letters 148 243^258

Blichert-Toft J Chauvel C amp Albare de F (1997) Separation of Hfand Lu for high precision isotope analysis of rock samples by mag-netic sector-multiple collector ICP-MS Contributions to Mineralogy

and Petrology 127 248^260Brey G P amp Kolaquo hler T (1990) Geothermobarometry in four-phase

lherzolites II New thermobarometers and practical assessment ofexisting thermobarometers Journal of Petrology 31 1353^1378

Brey G P Kolaquo hlerT amp Nickel K G (1990) Geothermobarometry infour-phase lherzolites I Experimental results from 10 to 60 kbJournal of Petrology 31 1313^1352

Caporuscio F amp Smyth J (1990) Trace element crystal chemistryof mantle eclogites Contributions to Mineralogy and Petrology 105550^561

Carbno G B amp Canil D (2002) Mantle structure beneath the SWSlave Craton Canada constraints from garnet geochemistry in theDrybones Bay Kimberlite Journal of Petrology 43 129^142

Coleman R G Lee E D Beatty L B amp Brannock W W (1965)Eclogites and eclogites their differences and similaritiesGeological Society of America Bulletin 76 483^508

Creaser R A Grulaquo tter H Carlson J amp Crawford B (2004)Macrocrystal phlogopite Rb^Sr dates for the Ekati propertykimberlites Slave Province Canada evidence for multiple intru-sive episodes in the Paleocene and Eocene Lithos 76 399^414

Davies R M GriffinW L Pearson N J Andrew A S Doyle BJ amp OrsquoReilly S Y (1999) Diamonds from the deep Pipe DO-27Slave Craton Canada In Gurney J J Gurney J L Pascoe MD amp Richardson S H (eds) Proceedings of the 7th International

Kimberlite Conference CapeTown Red Roof Design pp 148^155Davies R M Griffin W L OrsquoReilly S Y amp Doyle B J (2004)

Mineral inclusions and geochemical characteristics of microdia-monds from the DO27 A154 A21 A418 DO18 DD17 and RanchLake kimberlites at Lac de Gras Slave Craton Canada Lithos 7739^55

DavisW J amp Hegner E (1992) Neodymium isotopic evidence for thetectonic assembly of Late Archean crust in the Slave Provincenorthwest Canada Contributions to Mineralogy and Petrology 111493^504

Davis W J Jones A G Bleeker W amp Grutter H (2003)Lithosphere development in the Slave craton a linked crustal andmantle perspective Lithos 71 575^589

Dawson J B (1984) Contrasting types of upper-mantle metasoma-tism In Kornprobst J (ed) Kimberlites II The Mantle and

Crust^Mantle Relationships Amsterdam Elsevier pp 289^294DePaolo D J (1981) Neodymium isotopes in the Colorado Front

Range and crust^mantle evolution in the Proterozoic Nature 291193^196

Erlank A J amp Kushiro I (1970) Potassium contents of syntheticpyroxenes at high temperatures and pressures Carnegie Institution ofWashingtonYearbook 68 433^439

Foley S F Barth M G amp Jenner G A (2000) Rutilemelt partitioncoefficients for trace elements and an assessment of the influence ofrutile on the trace element characteristics of subduction zonemagmas Geochimica et Cosmochimica Acta 64 933^938

Foley S F Buhre S amp Jacob D E (2003) Evolution of the Archaeancrust by delamination and shallow subduction Nature 421 249^252

Graham I Burgess J L Bryan D Ravenscroft P J Thomas EDoyle B J Hopkins R amp Armstrong K A (1999) Explorationhistory and geology of the Diavik Kimberlites Lac de GrasNorthwest Territories Canada In Gurney J J Gurney J LPascoe M D amp Richardson S H (eds) Proceedings of the 7th

International Kimberlite Conference Cape Town Red Roof Designpp 262^279

Green D H amp Wallace M E (1988) Mantle metasomatism byephemeral carbonatite melts Nature 336 459^462

Green T H Blundy J D Adam J amp Yaxley G M (2000) SIMSdetermination of trace element partition coefficients betweengarnet clinopyroxene and hydrous basaltic liquids at 2^75 GPaand 1080^12008C Lithos 53 165^187

GriffinW L amp OrsquoReilly SY (2007) Cratonic lithospheric mantle Isanything subducted Episodes 30(1) 43^53

Griffin W L Wass S Y amp Hollis J D (1984) Ultramafic xenolithsfrom Bullenmerri and Gnotuk maars Victoria Australia petrol-ogy of a sub-continental crust^mantle transition Journal of

Petrology 25 53^87GriffinW L Doyle B J Ryan C G Pearson N J OrsquoReilly SY

Davies R Kivi K van Achterbergh E amp Natapov LM (1999)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1871

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Layered mantle lithosphere in the Lac de Gras Area SlaveCraton composition structure and origin Journal of Petrology 40705^727

Griffin W L Pearson N J Belousova E Jackson S E vanAchterbergh E OrsquoReilly SY amp Shee S R (2000)The Hf isotopecomposition of cratonic mantle LAM-MC-ICPMS analysis ofzircon megacrysts in kimberlites Geochimica et Cosmochimica Acta

64 133^147Griffin W L OrsquoReilly S Y Doyle B J Pearson N J

Coopersmith H Kivi K Malkovets V amp Pokhilenko N(2004) Lithosphere mapping beneath the North American plateLithos 77 873^922

Grulaquo tter H S Apter D B amp Kong J (1999) Crust^mantle couplingevidence from mantle-derived xenocrystic garnets In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 307^312

Hart S R Blusztajn J Dick H J B Meyer P S ampMuehlenbachs K (1999) The fingerprint of seawater circulationin a 500-meter section of ocean crust gabbros Geochimica et

Cosmochimica Acta 63 4059^4080Harte B amp Kirkley M B (1997) Partitioning of trace elements

between clinopyroxene and garnet Data from mantle eclogitesChemical Geology 136 1^24

Heaman L M Creaser R A amp Cookenboo H O (2002) Extremeenrichment of high field strength elements in Jericho eclogitexenoliths A cryptic record of Paleoproterozoic subduction partialmelting and metasomatism beneath the Slave craton CanadaGeology 30 507^510

Heaman L M Kjarsgaard B A amp Creaser R A (2004) The tem-poral evolution of North American kimberlites Lithos 76 377^397

Heaman L M Creaser R A Cookenboo H O amp Chacko T(2006) Multi-stage modification of the northern Slave mantlelithosphere Evidence from zircon- and diamond-bearing eclogitexenoliths entrained in Jericho Kimberlite Journal of Petrology 47821^858

Hermann J Cpandler C Hack A amp Korsakov A (2006) Aqueousfluids and hydrous melts in high-pressure and ultra-high pressurerocks implications for element transfer in subduction zones Lithos92 399^417

Hills DV amp Haggerty S E (1989) Petrochemistry of eclogites fromthe Koidu Kimberlite Complex Sierra Leone Contributions to

Mineralogy and Petrology 103 397^422Hoffman P F (1989) Precambrian geology and tectonic history of

North America In Bally A amp Palmer A (eds) The Geology of

North AmericaccedilAn Overview Boulder CO Geological Society ofAmerica pp 447^512

IrelandT R Rudnick R L amp Spetsius Z (1994) Trace-elements indiamond inclusions from eclogites reveal link to Archean granitesEarth and Planetary Science Letters 128 199^213

Irvine G J Pearson D G Kjarsgaard B A Carlson R WKopylova M G amp Dreibus G (2003) Evolution of the litho-spheric mantle beneath Northern Canada a Re^Os isotope andPGE study of kimberlite-derived peridotite xenoliths fromSomerset Island and a comparison to the Slave and Kaapvaalcratons Lithos 71 461^488

Isachsen C E amp Bowring S A (1994) Evolution of the Slave cratonGeology 22 917^920

Jacob D E (2004) Nature and origin of eclogite xenoliths from kim-berlites Lithos 77 295^316

Jacob D E amp Foley S F (1999) Evidence for Archean ocean crustwith low high field strength element signature from diamondifer-ous eclogite xenoliths Lithos 48 317^336

Jacob D Jagoutz E Lowry D Mattey D amp Kudrjavtseva G(1994) Diamondiferous eclogites from Siberiaccedilremnants ofArchean oceanic-crust Geochimica et Cosmochimica Acta 585191^5207

Jacob D E Fung A Jagoutz E amp Pearson D G (2003) Petrologyand geochemistry of eclogite xenoliths from the Ekati kimberlitesarea Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA_239

Jacob D E Bizimis M amp Salters V J M (2005) Lu^Hf and geo-chemical systematics of recycled ancient oceanic crust evidencefrom Roberts Victor eclogites Contributions to Mineralogy and

Petrology 148 707^720Jerde E A Taylor L A Crozaz G Sobolev NV amp SobolevV N

(1993) Diamondiferous eclogites from Yakutia Siberiaccedilevidencefor a diversity of protoliths Contributions to Mineralogy and Petrology

114 189^202Jones A G Ferguson I J Chave A D Evans R L amp

McNeice GW (2001) The electric lithosphere of the Slave cratonGeology 29 423^426

Kelemen P B Shimizu N amp Dunn T (1993) Relative depletionof niobium in some arc magmas and the continental-crust-partitioning of K Nb La and Ce during meltrockreaction in the upper-mantle Earth and Planetary Science Letters 120111^134

Kinzler R J (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition Application to mid-ocean ridge basalt petrogenesis Journal of Geophysical Research 102853^874

Klemme S van der Laan S R Foley S F amp Guenther D (1995)Experimentally determined trace and minor element partitioningbetween clinopyroxene and carbonatite melt under upper mantleconditions Earth and Planetary Science Letters 133 439^448

Konzett J Armstrong R A Sweeney R J amp CompstonW (1998)The timing of MARID metasomatism in the Kaapvaal mantle anion probe study of zircons from MARID xenoliths Earth and

Planetary Science Letters 160 133^145Kopylova M G Russell J K amp Cookenboo H (1999) Mapping the

lithopshere beneath the north central Slave Craton In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 468^479

Krogh E (1988) The garnet^clinopyroxene iron^magnesiumgeothermometerccedila reinterpretation of existing experimentaldata Contributions to Mineralogy and Petrology 99 44^48

Kusky T M (1989) Accretion of Archean Slave Province Geology 1763^67

LeCheminant A N amp Heaman L M (1989) Mackenzie igneousevents Canada Middle Proterozoic hotspot magmatismassociated with ocean opening Earth and Planetary Science Letters

96 38^48LeCheminant A N Richardson D G DiLabio R N W amp

Richardson K A (1996) La recherche de diamants au CanadaGeological Survey of Canada Open File 3228 268pp

MacGregor I D amp Carter J L (1970) The chemistry of clinopyrox-enes and garnets of eclogite and peridotite xenoliths from theRoberts Victor mine South Africa Physics of the Earth and PlanetaryInteriors 3 391^397

MacGregor I D amp Manton W I (1986) Roberts Victor eclogitesancient oceanic crust Journal of Geophysical Research 91 14063^14079

MacKenzie J M amp Canil D (1999) Composition and thermal evo-lution of cratonic mantle beneath the central Archean SlaveProvince NWT Canada Contributions to Mineralogy and Petrology

134 313^324

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1872

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

McCandlessT E amp Gurney J J (1989) Sodium in garnet and potas-sum in clinopyroxene criteria for classifying mantle eclogites InRoss J (ed) Kimberlites and Related Rocks Carlton Blackwellpp 827^832

McDonoughW F amp Sun S-S (1995) The composition of the EarthChemical Geology 120 223^253

Navon O amp Stolper E (1987) Geochemical consequences of meltpercolation the upper mantle as a chromatographic columnJournal of Geology 95 285^307

Nikogosian I K amp Sobolev AV (1997) Ion-microprobe analysis ofmelt inclusions in olivine experience in estimating the olivine^meltpartition coefficients of trace elements Geochemistry International 35119^126

Norman M D Griffin W L Pearson N J Garcia M O ampOrsquoReilly S Y (1998) Quantitative analysis of trace element abun-dances in glasses and minerals a comparison of laser ablationinductively coupled plasma mass spectrometry solution inductivelycoupled plasma mass spectrometry proton microprobe and elec-tron microprobe data Journal of Analytical Atomic Spectrometry 13477^482

Norman M D Garcia M O amp Pietruszka A J (2005) Trace-element distribution coefficients for pyroxenes plagioclase andolivine in evolved tholeiites from the 1955 eruption of KilaueaVolcano Hawairsquoi and petrogenesis of differentiated rift-zonelavas American Mineralogist 90 888^899

OrsquoHara M J amp Herzberg C (2002) Interpretation of trace elementand isotope features of basalts Relevance of field relations petrol-ogy major element data phase equilibria and magma chambermodeling in basalt petrogenesis Geochimica et Cosmochimica Acta 662167^2191

PadghamW A amp FysonW K (1992) The Slave Province a distinctcraton CanadianJournal of Earth Sciences 29 2072^2086

Pearson D G Davies G R amp Nixon P H (1993) Geochemical con-straints on the petrogenesis of diamond facies pyroxenites from theBeni Bousera peridotite massif North Morocco Journal of Petrology34 125^172

Pearson N J Griffin W L Doyle B J OrsquoReilly S Y vanAchterbergh E amp Kivi K (1999) Xenoliths from kimberlitepipes of the Lac de gras area Slave Craton Canada In GurneyJ J Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite ConferenceCapeTown Red Roof Design pp 644^658

Perfit N R Gust D A Bence A E Arculus R J amp Taylor S R(1980) Chemical characteristics of island-arc basalts implicationsfor mantle sources Chemical Geology 30 227^256

Polat A Hofmann A W amp Rosing M T (2002) Boninite-likevolcanic rocks in the 37^38 Ga Isua greenstone belt WestGreenland geochemical evidence for intra-oceanic subductionzone processes in the early Earth Chemical Geology 184 231^254

Poudjom DjomaniY H GriffinW L OrsquoReilly SY amp Doyle B J(2005) Lithospheric domains and controls on kimberlite emplace-ment Slave Province Canada evidence from elastic thickness andupper mantle composition Geochemistry Geophysics Geosystems 6doi1010292005GC0009978

Presnall D C Gudfinnsson G H amp Walter M J (2002)Generation of mid-ocean ridge basalts at pressures from 1 to 7GPa Geochimica et Cosmochimica Acta 66 2073^2090

Schmidberger S S Heaman L M Simonetti A Creaser R A ampCookenboo H O (2005) Formation of Paleoproterozoiceclogitic mantle Slave Province (Canada) Insights from in-situHf and U^Pb isotopic analyses of mantle zircons Earth and

Planetary Science Letters 240 621^633

Schmidberger S S Simonetti A Heaman L M Creaser R A ampWhiteford S (2007) Lu^Hf in-situ Sr and Pb isotope and traceelement systematics for mantle eclogites from the Diavik diamondmine Evidence for Paleoproterozoic subduction beneath the Slavecraton Canada Earth and Planetary Science Letters 254 55^68

Schulze D J amp Helmstaedt H (1988) Coesite-sanidine eclogitesfrom kimberlite products of mantle fractionation or subductionJournal of Geology 96 435^443

Smithies R H Champion D C amp Sun S S (2004) The case forArchaean boninites Contributions to Mineralogy and Petrology 147705^721

Smyth J R Caporuscio F A amp McCormick T C (1989) Mantleeclogites evidence of igneous fractionation in the mantle Earth andPlanetary Science Letters 93 133^141

Sobolev NV amp Lavrentrsquoyev Y G (1971) Isomorphic sodium admix-ture in garnets formed at high pressure Contributions to Mineralogy

and Petrology 31 1^12Stalder R Foley S F Brey G P amp Horn I (1998) Mineral aqueous

fluid partitioning of trace elements at 900^12008C and 30^57GPa New experimental data for garnet clinopyroxene and rutileand implications for mantle metasomatism Geochimica et

Cosmochimica Acta 62 1781^1801Sun S-S amp McDonough W F (1989) Chemical and isotopic sys-

tematics of oceanic basalts implications for mantle compositionsand processes In Saunders A D amp Norris M J (eds)Magmatism in the Ocean Basins Geological Society London Special

Publications 42 313^345Taura H Yurimoto H Kurita K amp Sueno S (1998) Pressure

dependence on partition coefficients for trace elements between oli-vine and the coexisting melts Physics and Chemistry of Minerals 25469^484

Taylor L A amp Neal C R (1989) Eclogites with oceanic crustal andmantle signatures from the Bellsbank kimberlite South AfricaPart I Mineralogy petrography and whole rock chemistryJournal of Geology 97 551^567

van Achterbergh E Ryan C G amp Griffin W L (1999) Glitter online intensity reduction for the laser ablation inductively coupledplasma spectrometry Abstract 9th V M GoldschmidtConference LPI Boston MA 905 pp Abstract no 7215

van Breemen O DavisW J amp King J E (1992) Temporal distribu-tion of granitoid plutonic rocks in the Archean Slave ProvinceNorthwest Canadian Shield Canadian Journal of Earth Sciences 292186^2199

Vervoort J D Patchett P J Blichert-Toft J amp Albare de F (1999)Relationship between Lu^Hf and Sm^Nd isotopic systems in theglobal sedimentary system Earth and Planetary Science Letters 16879^99

Viljoen K S Smith C B amp Sharp Z D (1996) Stable and radio-genic isotope study of eclogite xenoliths from the Orapa kimberliteBotswana Chemical Geology 131 235^255

Walter M J (1998) Melting of garnet peridotite and theorigin of komatiite and depleted lithosphere Journal of Petrology 3929^60

Wasserburg G J Jacobsen S B DePaolo D J McCulloch M Tamp Wen T (1981) Precise determination of SmNd ratios Sm andNd isotopic abundances in standard solutions Geochimica et

Cosmochimica Acta 45 2311^2323Westerlund K J Shirey S B Richardson S H Carlson R W

Gurney J J amp Harris JW (2006) A subduction wedge origin forPaleoarchean peridotitic diamonds and harzburgites from thePanda kimberlite Slave craton evidence from Re^Os isotope sys-tematics Contributions to Mineralogy and Petrology 152 275^294

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1873

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Page 13: Origins of Xenolithic Eclogites and Pyroxenites from the Central

and YK1952 are similar although at different abundanceswith only modest LREEN over HREEN enrichment andstrong depletions in ZrN relative to neighbouring elements(Fig 8d)Clinopyroxenes in the volatile-rich samples have flat pat-

terns and high abundances from LaN to SmN and smoothnegative slopes from SmN to LuN Niobium Sr and Ti are

noticeably depleted relative to elements of similar compat-ibility (Fig 8e)

Trace-element partitioningTo assess whether coexisting phases are in trace-elementequilibrium we calculated apparent cpx^garnet distribu-tion coefficients and compared them with those

(a)

(d)

(b)

(e)

(c)

GarnetCI-chondrite GarnetCI-chondrite

low-Mghigh-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1952

1

10

Nd Sm Eu Gd Dy

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Sm GdDy

001

01

1

10

100

LaCe

PrNd Eu

HoY

ErYb

LuSc

ZrHf

TiV

NbCo

Fig 7 Extended trace-element patterns of garnet in (a) low-Mg eclogites (b) high-Mg eclogites with inset showing enlargement of MREE tohighlight Eu anomalies and 1s uncertainties (c) high-Ca eclogites (d) pyroxenites and (e) volatile-bearing eclogites Normalized to chondritevalues of McDonough amp Sun (1995)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1855

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

determined experimentally or in demonstrably equili-brated natural samples Low-Mg and high-Mg eclogiteshave Dcpxgarnet for most elements that are within therange determined for natural equilibrated eclogites andexperimental assemblages (Harte amp Kirkley 1997 Greenet al 2000 Barth et al 2002a) (Electronic Appendix 7)High-Ca eclogites plot at the low end of the range

of distribution coefficients and have very low DMREE

and DHREE outside the range indicated by the datasetof Harte amp Kirkley (1997) who noted that DREE decreaseswith increasing Ca content in eclogitic minerals Theseassemblages are therefore considered to be in equilibriumPyroxenite VR40384 has D values similar to eclogites

but at the high end of the range determined by Harte amp

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

(a)

(d)

(b)

(e)

(c)

CpxCI-chondrite CpxCI-chondrite

low-Mg high-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1914

YK1952

spongy rim

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

VR40345

Fig 8 Extended trace-element patterns of cpx as in Fig 7

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1856

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Kirkley (1997) (Electronic Appendix 7) Another group(YK1914 and YK1952) plots at the low end but DHf DZrDTi and DV are far below the values indicated by experi-mental datasets Because pyroxenites compositionallygrade into eclogites their drastically different DHFSE

values suggest disequilibrium In contrast volatile-bearingsamples have DHFSE values that appear to be too high to bein equilibrium however this may be a consequence oftheir unusual bulk chemical composition (Ti- and K-rich)which differs substantially from that of lsquonormalrsquo eclogitesand experimental charges

GEOTHERMOBAROMETRYFollowing Schmidberger et al (2007) temperatures oflast equilibration were calculated for eclogitic andopx-free pyroxenitic assemblages using the formulationof Krogh (1988) The results are summarized in Table 2and Electronic Appendix 8 for a preset pressure of5 GPa and illustrated as a histogram in Fig 9 Foropx-bearing pyroxenites temperatures were calculatedsimultaneously with pressure [Ca-in-opx thermometryand Al-in-opx barometry using the formulations of Breyet al (1990)] to constrain their relative positions in themantle columnMinerals in some eclogites from Lac de Gras show dis-

tinct core and rim compositions or inhomogeneitiesTemperatures were not calculated for samples withspatially irresolvable inhomogeneous garnet compositions(SE01VR43479VR43480) For eclogites with spongy cpxcore compositions were used in the calculationTemperatures were calculated for both core and rim com-positions where applicable For all samples these

temperatures agree within the uncertainty of this thermo-meter (508C Brey amp Kolaquo hler 1990)Most equilibration temperatures for eclogites lie between

800 and10008C (Fig 9)The transition between the shallowand the deep lithospheric mantle layers in the central Lacde Gras area lies in this temperature interval (Griffin et al1999 Pearson et al 1999) Eclogites in this study lack thedistinct bimodal temperature distribution observed bySchmidberger et al (2007) but as in their study high-Caeclogites give higher average temperatures (10808C corre-sponding to a depth of 155 km if equilibrated along a 40mWm2 geotherm) than other eclogites types (9808Cdepth 135 km) the diamond-bearing sample (VR40345)gives thehighest temperature of the datasetHighNa2Ocon-tents in garnet from high-Ca eclogites are qualitativelyconsistent with higher equilibration pressures (Sobolev ampLavrentrsquoyev 1971) as are high K2O contents in coexistingcpx (Erlank ampKushiro1970) suggesting that high-Ca eclo-gites not only equilibrated at high temperature but also highpressure rather than representing a temperature excursionPyroxenites both opx-bearing and opx-free give somewhatlower average temperatures than eclogites (mean 9258C)and phlogopite-rich eclogites give the lowest temperatures(685 and 7698C respectively) The calcite-bearing eclogiteVR67112b gives a relatively low equilibration temperatureof 8558C

MEASURED ANDRECONSTRUCTEDWHOLE -ROCKSRationale and approachKimberlite-borne xenoliths sometimes show evidence forinfiltration of the host kimberlite (eg Taylor amp Neal1989 Barth et al 2001 2002b Heaman et al 2006) Wetherefore follow the standard approach of calculatingwhole-rock compositions from the compositions of the con-stituent phases weighted by their modal abundances Onlycore compositions were used as distinct rim compositionsmay point to late diffusion of elements from kimberlite-related melt veinsMany eclogitic whole-rock compositions reconstructed

with only garnet and cpx have negative Ti anomalies(not shown) For most of these samples rutile or ilmenitehas been identified as part of the mineral assemblage andthese phases should therefore be considered in the whole-rock reconstruction Rutile or ilmenite modes were esti-mated by assuming that the bulk-rock has no Ti anomalywhen normalized to a normal mid-ocean ridge basalt(N-MORB) composition and calculating the amount ofrutile required to erase theTi anomalyIt is of course possible that the eclogites had negative

or positive Ti anomalies relative to elements of similarcompatibility (Eu and Dy) Because we reconstructed

0

2

4

6

8

10

12volatile-rich

pyroxenite opx-bearing

pyroxenite opx-free

high-Ca

high-Mg

low-Mg

700750

800850

900950

10001050

11001150

12001250

Fre

quen

cy

Temperature

Fig 9 Histogram ofTKrogh (Krogh 1988) obtained at a preset pres-sure of 5 GPa except opx-bearing pyroxenites where TBKN wassolved simultaneously with PBKN (Brey amp Kolaquo hler 1990)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1857

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

whole-rock compositions from mineral core compositionskimberlite infiltration is unlikely to have caused fractiona-tion of Ti from Eu and Dy In a study in which rutilemodes were estimated from the differences between Ti inmeasured bulk-rocks and in bulk-rocks reconstructed withonly garnet and cpx only three of 19 eclogites recon-structed with rutile show Ti anomalies (Barth et al2002a) Even eclogite xenoliths with depleted Zr and Hfabundances that resemble those of modern island arcbasalts do not show Ti anomalies (Jacob amp Foley 1999)We therefore consider our approach to be a reasonablyrobust means of obtaining meaningful rutile modes wheresmall sample sizes preclude reliable determination of themodes of accessory phases by point-countingWhere rutilewas not recognized on thin-section scale or not analysedthe average eclogitic rutile composition was used for thereconstruction

UncertaintiesReconstructing whole-rock compositions in this way is arather crude method given the coarse grain size of theeclogites relative to the sample size and the non-uniformdistribution of the principal minerals on thin-sectionscale The uncertainty in mineral modes of medium- tocoarse-grained eclogites with grains45mm has been esti-mated by Jerde et al (1993) to be 10 plus uncertaintiesrelated to partial alteration which is minor in our samplesNevertheless the whole-rock trace-element budget israther insensitive to the exact modes for common eclogiticassemblages and REE patterns which encompass mostother elements in terms of compatibility in garnet andcpx do not change significantly even for modal variationsof 30 (Jerde et al 1993) This is not true for major-element concentrations which vary substantially with var-iations in mineral modes and we therefore do not calculatewhole-rock major-element contentsBecause the discussion that follows is based on trace-

element patterns and trends rather than absolute abun-dances whole-rock reconstruction based on mineralmodes is adequate We are confident that the uncertaintyinherent in this approach is not leading to significant mis-representation of eclogite whole-rock compositions becausethe trace-element patterns of the reconstructed whole-rocks show distinct differences (see below) that correlatewith the assignment to eclogite types based on garnetmajor-element composition and this is independent ofmineral mode Following Jerde et al (1993) in ElectronicAppendix 9 we show variations in trace-element patternfor one sample in each eclogite type for variations ingarnet and cpx abundance from 7030 to 3070 vol This shows that the trace-element patterns are generallyinsensitive to cpxgt ratios with regard to slopes andanomalies such as positive Sr Pb and Eu anomalies withthe exception of the positive Eu anomaly in high-Mgeclogite VR43479 which disappears if a high cpx mode is

assumed (not shown) Variations in rutile modes by 50(Electronic Appendix 9d) lead as would be expected tothe appearance of Ti anomalies but a rutile mode of 05instead of 1 vol does not erase the positive Nb anomalyin low-Mg eclogiteVR50909Because of the robustness of the relative trace-element

abundances in the reconstructed eclogites a garnet propor-tion of 065 and cpx proportion of 035 approximately themedian modal abundances of these minerals in volatile-free eclogites from Lac de Gras are assumed for eclogitesamples that are too small for determination of meaningfulmodes by point-counting

Within- and between-group variationsThe trace-element compositions of reconstructed whole-rocks are given in Table 4 N-MORB-normalized trace-element patterns of reconstructed whole-rocks areshown in Fig 10 The three eclogite groups have distincttrace element patterns that are particularly obvious forthe REE (Fig 11) Low-Mg eclogites have positive Pb andSr anomalies and smooth positive slopes in their MREEN

and HREEN LREEN are enriched relative to MREEN inthree samples (Fig 11a) Three samples have high Nbrelative to similarly compatible elements and comparedwith most other eclogites (Fig 10a) All high-Mg eclogitesdisplay more or less pronounced positive Eu and Sranomalies and most have positive Pb and negative Zrand P anomalies Their N-MORB-normalized REE pat-terns show shallow positive slopes from LaN to PrN orNdNand aweak negative slope from NdN to LuN (Fig11b)VR43479 is a low-Mg eclogite with regard to its garnet

Ca^Fe^Mg composition but has a trace-element patterntypical of high-Mg eclogites and will be discussed withthe latter Conversely sample YK1949 a high-Mg eclogitein terms of its garnet Ca^Fe^Mg compostion is more simi-lar to low-Mg eclogites in terms of its whole-rock trace-element pattern and is plotted in Figs 10a and 11aWhereas the patterns of most samples within the groupresemble each other the N-MORB-normalized incompati-ble-element abundances of the high-Ca eclogites vary bymore than an order of magnitude (Fig 10c) All sampleshave positive Pb anomalies and most have positive Sranomalies Some samples have negative P andor Zranomalies Niobium abundances in three of the samplesare very high similar to those for some low-Mg eclogitesHigh-Ca eclogites tend to have flatter MREEN andHREEN patterns and more uniform HREE abundancesthat the other eclogite types with humps between SmN

and HoNThe phlogopitized sample YK1911 shows a strong posi-

tive Pb anomaly and lower HREE abundances than thecalcite-bearing sample VR67112b which shows a distinctnegative Eu anomaly and a depletion of La relative to Ce(Figs 10d and 11d)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1858

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Trace-element compositions of reconstructed whole-rocks (ppm)

Sample P Sc Ti V Co Ni Ga Sr Y Zr Nb La Ce Pr Nd

Low-Mg eclogites

vr19674-g4 186 50 4649 317 57 59 144 77 24 30 36 13 31 05 24

vr43452 154 39 2253 328 49 84 176 52 32 11 01 02 07 01 08

vr43469 38 65 4377 349 54 91 142 34 28 15 02 05 15 03 17

yk1926 144 45 3406 258 52 87 145 89 18 18 20 09 29 06 33

yk1943 173 46 5732 295 41 39 146 122 25 48 82 34 85 13 64

yk19491 102 34 1693 158 31 49 100 46 10 14 03 09 21 03 17

High-Mg eclogites

vr19677-g2 138 39 5160 291 41 53 119 98 14 22 04 13 53 11 64

vr434792 46 25 1265 129 36 126 84 56 3 7 04 06 25 04 23

vr67360 49 36 1573 138 64 184 111 76 5 9 04 09 38 07 33

yk1946 153 28 4877 192 45 105 124 108 12 48 03 10 38 08 50

High-Ca eclogites

vr19674-g5 267 30 7663 200 41 67 195 32 23 43 77 01 04 02 16

vr40345 0 33 2821 175 47 98 210 179 13 13 66 60 109 11 38

vr43465 142 27 5104 220 31 83 121 197 14 35 29 05 22 05 40

vr50858 295 30 6898 201 44 94 150 48 22 38 06 01 03 02 20

vr50860 68 57 5672 418 57 71 186 109 19 11 02 02 13 03 26

vr50909 238 28 7295 201 42 92 171 48 20 57 13 01 04 02 21

yk3528 91 51 3712 341 53 155 177 49 16 9 01 02 08 02 11

Volatile-rich eclogites

vr67112b 40 84 9242 521 46 30 107 4 40 66 31 26 126 24 125

yk1911 31 39 9428 286 44 82 164 3 23 52 64 18 52 09 46

Pyroxenites

yk1914 111 49 707 207 48 547 44 31 4 3 02 06 19 03 10

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

Low-Mg eclogites

vr19674-g4 11 05 10 22 39 10 29 33 06 12 05 028 24 008

vr43452 04 02 08 13 38 12 41 49 08 05 01 nd 02 na

vr43469 11 05 09 25 44 11 34 36 06 07 02 001 09 129

yk1926 12 05 11 19 30 07 22 24 04 07 03 004 23 090

yk1943 21 08 10 29 42 10 30 32 05 16 05 013 65 017

yk19491 06 03 11 08 14 04 14 18 03 04 02 003 29 115

High-Mg eclogites

vr19677-g2 20 08 12 21 24 05 15 15 03 05 05 007 52 6811

vr434792 05 02 12 05 06 01 04 03 00 03 01 001 98 037

vr67360 06 03 14 06 09 02 06 06 01 02 02 002 86 000

yk1946 18 07 12 21 22 05 13 12 02 11 03 006 56 112

High-Ca eclogites

vr19674-g5 17 09 11 33 43 09 24 23 04 16 05 007 02 011

vr40345 09 04 10 12 22 06 16 14 02 05 nd nd 298 na

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1859

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Continued

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

vr43465 22 09 12 25 26 06 15 14 02 10 04 001 25 040

vr50858 20 08 09 33 41 08 22 21 03 10 01 nd 02 na

vr50860 14 07 11 27 37 08 22 18 03 06 00 nd 06 na

vr50909 20 10 12 33 38 07 20 20 03 16 01 002 02 181

yk3528 08 05 11 20 30 07 18 18 02 05 01 nd 08 na

Volatile-rich eclogites

vr67112b 38 10 07 53 74 15 41 34 04 18 03 012 60 004

yk1911 20 07 08 35 39 07 18 17 02 10 14 nd 73 na

Pyroxenites

yk1914 02 01 11 04 08 02 05 06 01 01 03 003 68 101

aGarnet has CandashMgndashFe relationship typical of high-Mg eclogites but whole-rock trace-element pattern is typical oflow-Mg eclogitesbGarnet has CandashMgndashFe relationship typical of low-Mg eclogites but whole-rock trace-element pattern is typical ofhigh-Mg eclogites

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

(a)

001

01

1

10(c)

001

01

10(b)

VR43452

YK1949

RockNMORB

low-Mg

RockNMORB

high-Mg

(d)

volatile-richhigh-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

phlogopite-bearingYK1911

calcite-bearingVR67112b

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 10 Extended trace-element patterns of reconstructed (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-bearing eclogites Normalized to N-MORB of Sun amp McDonough (1989)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1860

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

RockNMORB

(b)

RockNMORB

(d)

010

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu001

01

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

(a)

001

01

1

10(b)

001

01

1

10(c)

001

01

1

10(e)

high-TSchmidberger et al

low-TSchmidberger et al

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

high-Mglow-Mg

high-Cavolatile-rich

phlogopite-bearingYK1911

calcite-bearingVR67112b

YK1949VR43479

Fig 11 N-MORB normalized REE patterns for (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-richeclogites Shown for comparison in (e) are reconstructed eclogites from Schmidberger et al (2007)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1861

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Both volatile-rich whole rocks display strong negativeSr anomalies Strontium can be sited in calcite or apatiteandor indicate the presence of a Sr-rich mineral that wasnot identified in thin section such as bariteFour of the six reconstructed whole-rocks in the study

of Schmidberger et al (2007) (Fig 11e) show markedLREEHREEN enrichment that is observed only in onehigh-Ca eclogite in our study (Fig 11c) suggesting thatmany of the eclogites they studied were affected by metaso-matic LREE enrichment Two samples in their study thatare not affected by this enrichment show positive Euanomalies similar to the high-Mg eclogites in this studywhereas the distinctive REE pattern of the low-Mg eclo-gites is not recognized among their samples

SR^ND^HF I SOTOPESYSTEMATICSStrontium Nd and Hf isotope data for optically puremineral separates and reconstructed whole-rocks aregiven in Table 5 Spongy rims and alteration are likely tobe mostly removed during leaching in 6N HCl in combina-tion with repeated ultrasonication Despite this rigoroustreatment the reproducibility of Nd in garnet VR43469and in cpx YK1914 is worse than for other samples Thismay be due to inhomogeneities related to late Nd diffusionpossibly from grain boundaries containing a kimberlitecomponent The reproducibility of the 176Hf177Hf data isgenerally poorer than that of Nd which may reflect a com-bination of matrix effects caused by imperfectly purifiedsolutions and sample heterogeneity Parent^daughterratios are available only from in situ analyses which arenot precise enough to allow meaningful model age orinitial isotope ratio calculations (ie at the time of kimber-lite eruption) Because the Lac de Gras kimberlites arerelatively young (55^56 Ma Graham et al 1999) age cor-rection would result in only a minor change in isotopecomposition which is inconsequential in view of the rangeof compositions displayed by most samples and hence hasno effect on the conclusions drawnClinopyroxenes in five volatile-free eclogites and a

pyroxenite have Sr isotope ratios ranging from 0701662 to0703546 this range is similar to or somewhat lower thanthat of the depleted mantle at the time of kimberliteemplacement (0703 Bell amp Tilton 2002) Clinopyroxenein the phlogopitized sample has a distinctly high 87Sr86Srof 0751186Clinopyroxene controls 56^92 of the Nd budget (Nd

concentration in cpx weighted by mode) Neodymiumisotope ratios range from 0511660 to 0514844 Thiscorresponds to eNd variation (per 10 000 deviationof 143Nd144Nd in the sample from 143Nd144Nd in the chon-dritic mantle model reservoir) from 17 to thorn43 cpx inthe low-Mg eclogites have higher eNd than those in

high-Mg eclogites Garnets have generally have higher143Nd144Nd from 0512984 (eNdfrac14thorn 7) to 0520965(eNdfrac14thorn 162) Garnets in the two volatile-rich sampleshave the highest ratios 0522790 for garnet in the calcite-bearing sample VR67112b (eNdfrac14thorn198) and 0525721 inthe phlogopitized eclogiteYK1911 (eNdfrac14thorn 255)Garnet contributes 18^32 and cpx 48^78 of the Hf

budget The proportion controlled by rutile is 4^30(Aulbach et al in preparation) Measured 176Hf177Hf incpx ranges from 0281263 in phlogopite-bearing sampleYK1911 (eHffrac1453) to 0286130 in low-Mg eclogiteVR43469 (eHffrac14thorn119) and in garnet from 0283958 inpyroxeniteYK1915 (eHffrac14thorn42) to 0295718 in low-Mg eclo-gite YK1926 (eHffrac14thorn458) Hafnium in garnet in the phlo-gopite^apatite-bearing sample is extremely radiogenic(176Hf177Hffrac14 0354367) corresponding to an eHf of thorn2532which is close to a value determined for a garnet separatefrom a Roberts Victor eclogite (eHffrac14thorn2561 Jacob et al2005) Garnet in the calcite-bearing sample has extremelyradiogenic Nd and Hf (eNdfrac14thorn193 eHffrac14thorn2516) Garnetin pyroxenite YK1915 also has radiogenic Nd and Hf(eNdfrac14thorn27 eHffrac14thorn42)Despite the large uncertainties for reconstructed whole-

rocks (calculated as the square root of the sum of squareduncertainties of two standard errors for garnet and cpxanalyses combined with 20 propagated uncertainty onthe proportion of Nd or Hf respectively contributed byeach mineral) the isotopic data are consistent in thatthey correlate with the independently determinedparent^daughter ratios (1) Low-Mg eclogites have higherSmNd reflected in the steeper positive slope of theirN-MORB-normalized REE patterns (Fig 11a) comparedwith the high-Mg eclogites (Fig 11b) Accordingly recon-structed low-Mg eclogites have higher present-day eNd

(0 to thorn47) than high-Mg eclogites (4 and 17) (2) Low-Mg eclogite YK1943 has very high Nb and LREE abun-dances that could have been added during metasomatismthat also enhanced its Hf abundance This sample has thelowest eNd and eHf of the low-Mg eclogites (0 and thorn20respectively) In contrast apparently unenriched low-Mgeclogites have more radiogenic Nd and Hf (eNdfrac14thorn47and thorn9 eHffrac14thorn128 and thorn113 respectively)The Sr^Nd isotope diagram reveals that eNd in cpx

varies from 19 to 43 at nearly constant depleted-mantle-like 87Sr86Sr (Fig 12a) In the Nd^Hf isotope diagram(Fig 12b) the reconstructed whole-rocks lie on curvedmixing lines if the lsquomixingrsquo proportions of Nd and Hffor the two lsquoend-memberrsquo minerals are very different(eg YK1911) Garnets have more radiogenic Nd andHf than coexisting cpx All reconstructed volatile-freeeclogites plot in the radiogenic field with regard to Hfor are isotopically similar to the depleted mantle Twohigh-Mg eclogites and one low-Mg eclogite have unradio-genic Nd one high-Mg eclogite has radiogenic Nd and

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1862

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 5 Individual and averaged Sr Nd and Hf isotope ratios obtained from garnet and cpx solutions

Sample 87Sr86Sr 2SE2SD 143Nd144Nd 2SE2SD eNd 2SE2SD 176Hf177Hf 2SE2SD eHf 2SE2SD

Low-Mg eclogites

VR43469 cp 0702661 0000010 0514844 0000006 43 01 0286130 0000012 119 04

VR43469 gt 0515508 0000014 56 03 0286506 0000012 132 04

VR43469 gt r 0515453 0000010 55 02 0286548 0000011 134 04

VR43469 gt r 0515384 0000007 54 01 0287244 0000010 158 04

VR43469 gt r 0515487 0000014 56 03 0286996 0000014 149 05

VR43469avg gt 0515458 0000109 55 21 0286823 0000715 143 253

WR VR43469 0515059 0000273 47 5 0286405 0001079 128 38

YK1926 cp 0703546 0000011 0513043 0000007 8 0 0283154 0000009 13 03

YK1926 gt 0513364 0000008 14 01 0295718 0000005 458 02

WR YK1926 0513120 0000026 9 1 0285980 0000716 113 25

YK1943 cp 0512143 0000007 10 01 0282213 0000005 20 02

YK1943 gt 0514461 0000012 36 02 0287510 0000008 168 03

YK1943 gt r 0514491 0000009 36 02 0287332 0000005 161 02

YK1943 gt r 0514517 0000010 37 02 0287290 0000007 160 02

YK1943avg gt 0514489 0000056 36 11 0287377 0000233 163 82

WR YK1943 0512625 0000228 0 5 0283138 0000879 13 31

High-Mg eclogites

VR43480 gt 0520965 0000017 162 03 0284950 0000010 77 04

VR67360 cp 0283509 0000074 26 26

VR67360 cp r 0283433 0000031 23 11

VR67360avg cp 0701662 0000018 0511660 0000010 19 02 0283471 0000108 25 38

VR67360 gt 0512984 0000101 7 20 0287429 0000180 165 64

WR VR67360 0511761 0000132 17 3 0284734 0000469 69 17

YK1949 cp 0703019 0000013 0512231 0000009 8 02 0282513 0000009 9 03

YK1949 gt 0513511 0000010 17 02 0290297 0000042 266 15

YK1949 gt r 0513508 0000012 17 02 0289947 0000015 254 05

YK1949 gt r 0513552 0000014 18 03

YK1949avg gt 0513523 0000049 17 10 0291569 0000494 311 175

WR YK1949 0512446 0000103 4 2 0284622 0000847 65 30

Volatile-rich eclogites

VR67112b gt 0522790 0000008 198 02 0354367 0000036 2532 13

YK1911 cp 0751186 0000009 0512165 0000007 9 01 0281263 0000009 53 03

YK1911 gt 0525721 0000013 255 03 0288051 0000012 187 04

WR YK1911 0513940 0000342 25 7 0281414 0000018 48 1

Pyroxenites

YK1914 cp 0703015 0000034 0513283 0000041 13 08

YK1914 cp r 0702956 0000005 0513200 0000027 11 05

YK1914 cp r 0513527 0000059 17 11

YK1914 cp r 0513045 0000010 8 02

YK1914avg cp 0702986 0000083 0513264 0000402 12 78

YK1915 gt 0514003 0000011 27 02 0283958 0000021 42

Analyses of full procedural replicates are designated with an lsquorrsquo after sample name SE standard error (internal errorsingle analyses) SD standard deviation (external error averaged multiple analyses) Italics used for reconstructed whole-rocks (mineral isotopic compositions and concentrations weighted by modes) Uncertainties for reconstructed whole-rocksinclude 2SE or 2SD uncertainties on each mineralrsquos isotope composition combined with 20 uncertainty in the amountof Nd or Hf contributed by the constituent minerals eNd and eHf are per 10 000 deviation from model chondritic mantle[model of Wasserburg et al (1981) and Blichert-Toft amp Albarede (1997) respectively]

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1863

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

εHf

εNd

Ekati eclogitesminus100

0

100

200

300

400

500

minus40 0 40 80 120 160 200 240 280

gt in YK1915

gt in VR67112bεNd=198 εHf=2532

minus40

0

40

80

120

160

minus40 minus20 0 20 40 60

minus60

minus40

minus20

0

20

40

070 071 072 073 074 075 076

North Americankimberlite field

εNd

87Sr86Sr

cpx in YK1914

Ekatieclogites

DM

DM

DM

DM

(a)

(b)

pyroxenites

low-Mg

high-Mg

volatile-rich

eclogites

Diavik low-TA154 S kimberlite

cpx in YK1911

Fig 12 (a) Sr^Nd for cpx and (b) Nd-Hf isotope diagram for garnet cpx and reconstructed whole-rocks (symbols with bold outline) Finedashed lines in (b) connect data points from the same sample with garnets having more radiogenic isotopic compositions Uncertainties forreconstructed whole-rocks combine 20 modal uncertainty with uncertainties on each mineralrsquos isotope composition Error bars on individualminerals may be smaller than the symbols Also shown are the isotopic compositions of the depleted mantle (DM DePaolo 1981 Griffin et al2000 Bell amp Tilton 2002) Range of 87Sr86Sr in eclogites from Ekati (Lac de Gras area 100 km north of Diavik Jacob et al 2003) in NorthAmerican kimberlites and compositions of low-temperature eclogites from Diavik and of Lac de Gras A154 South kimberlite fromSchmidberger et al (2007)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1864

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

one low-Mg eclogite has eNd similar to the depleted mantle(Fig 12b)Clinopyroxene in the phlogopitized eclogite YK1911

has unradiogenic Nd and Hf and strongly radiogenicSr whereas the coexisting garnet has radiogenic Nd andHf Although this sample contains 50 vol phlogopiteits whole-rock isotopic composition was calculated

from garnet and cpx alone because the contributionfrom phlogopite to the Hf and Nd budget is 52 and51 respectively (unpublished data) The whole-rockhas radiogenic Nd but has lower eHf than anyother sample analysed in accord with the extreme Hfabundance (53 ppm) and low LuHf of the cpx

DISCUSS IONOrigins of eclogites beneath the SlaveCratonHigh-Mg and high-Ca eclogites

With regard to their major- and trace-element contentsthe high-Mg and high-Ca eclogites from Lac de Gras arebroadly similar to variably evolved oceanic gabbros(Fig 13b and c) which may indicate an origin as subductedoceanic crust Compatible trace-element abundancesare useful to deduce potential crustal protoliths as theseelements are less susceptible to post-formation modification(Jacob 2004) The flat HREE patterns in the high-Mg andhigh-Ca eclogites are compatible with a low-pressureorigin as the oceanic crust is a product of moderate degreesof partial melting mainly within the spinel stability field(eg Presnall et al 2002) whereas melting with residualgarnet would qualitatively be expected to produce stronglyfractionated REE patterns Positive Sr and Pb anomaliesalthough compatible with a plagioclase signature couldalso be due to cryptic metasomatism (Dawson 1984) ofthe eclogites during their residence in the lithosphericmantle These anomalies are probably not related to kim-berlite infiltration as the whole-rock eclogite compositionswere reconstructed from analyses of fresh mineral coresLow ZrSm similar to that in primitive gabbros is

observed for many of the high-Mg and high-Caeclogites (Fig 13b and c) and is independent of the ratioof garnetcpx and whether or not the whole-rock wasreconstructed with rutile (Electronic Appendix 10) Thenegative trend of Mg-number against Yb in garnet(Fig 14) argues against igneous processes involving garnetaccumulation (Taylor amp Neal 1989) as suggested forexample for some high-Mg eclogites from West Africa(Barth et al 2002b) Rather the anti-correlation suggeststhat olivine fractionation played a role in the formation ofthe protoliths Fractionating olivine would excludeYb (Dolivinebasalt

frac14 002 Nikogosian amp Sobolev 1997)and incorporate Mg relative to Fe (Dfrac1415 and 05 respec-tively Taura et al1998) leading to progressive depletion ofMg and enrichment of Yb in the residual melt and henceanti-correlated Mg and Yb in the whole-rock This isinherited after eclogitization All eclogite types (exceptVR43452) plot along the olivine fractionation trend sug-gesting that they formed by a similar processHigh-Mg eclogites have lower

PREE than high-Ca

eclogites and some appear to have positive Eu anomalies

high-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

gabbrosSE Indian ridge

gabbrosOman ophiolite

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

high-Mg

RockNMORB(a)

VR43452

YK1949

low-Mg

boninite-like melts

01

001

1

10

(b)

01

001

1

10

(c)

01

001

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

Fig 13 Trace-element patterns of (a) low-Mg (b) high-Mg and(c) high-Ca eclogites compared with boninite-like volcanic rocksfrom the Isua greenstone belt (Polat et al 2002) gabbros from theOman ophiolite (Benoit et al 1996) and the SE Indian Ridge (Hartet al 1999) normalized to N-MORB of Sun amp McDonough (1989)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1865

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(EuEu up to 14) similar to more primitive cumulates(troctolite gabbros and olivine gabbros) reported in the lit-erature (Fig 13b) Although the sizes of these Eu anomaliesare not outside the analytical uncertainty their correlationwith CeYb in reconstructed eclogites supports their signif-icance (r2frac14066 nfrac14 4) The more primitive nature ofhigh-Mg eclogites is consistent with higher Mg-numbersrelative to high-Ca eclogites and may be due to olivineaccumulation during formation of their protolithsThe two samples with the lowest HREE abundanceshave the highest EuEu which is compatible with varia-tions in plagioclase modesMore evolved protoliths with higher pyroxeneplagio-

clase plus trapped melt could account for small or absentpositive Eu anomalies and higher

PHREE of high-Ca

eclogites (Fig 13c) Although the reason for the frequentassociation of graphite or diamond with high-Ca eclogitesis unclear (Aulbach et al 2003) the frequent occurrence ofkyanite in high-Ca eclogites also observed at Ekati justnorth of Lac de Gras has been argued to support a crustalprotolith (Jacob 2004) Kyanite andor corundum havebeen suggested to exsolve from high-temperature high-Alclinopyroxenes (Caporuscio amp Smyth 1990) Eclogitesfrom Lac de Gras show no exsolution textures suggestiveof such an origin although recrystallization can obscureor destroy microstructural evidence for exsolution(eg Griffin et al 1984)Modelling shows that qualitatively plagioclase accumu-

lation from a low-pressure melt derived from MORBmantle can account for positive Pb Sr and Eu and negativeZr anomalies relative to N-MORB whereas cpx accumula-tion or trapped melt increases Zr Y MREE and HREE

abundances (Fig 15) Some melt addition (for exampleby entrapment) seems to be required to increase all trace-element abundances to those observed for the eclogitesAlthough high-Ca and high-Mg eclogites could representshallow and deep portions of the same oceanic crust thehigher average temperatures obtained for the high-Caeclogites the frequent occurrence of carbon in high-Caeclogites and its absence in high-Mg eclogites the strongLREEHREE fractionation observed for six of seven high-Ca eclogites and the narrow range of compositions of thehigh-Mg eclogites suggest that they may have protolithsthat are not cogenetic Some discrepancies in particularaffecting the incompatible elements may be due to post-crystallization processes acting upon the protoliths suchas dehydration melting and later metasomatism that isevident in peridotites from the same locality (Griffin et al1999)Low K2OTh for high-Ca and high-Mg eclogites

(average 08 Table 4) relative to assumed crustal

0

2

4

6

8

10

30 40 50 60 70 80

Mg

VR43452

Yb

GARNET

pyroxenites

low-Mghigh-Mghigh-Ca

eclogites

VR19673ecl-g2

Fig 14 Garnet Mg-number vs Yb (ppm including 2s uncertainty)Lines qualitatively show separate correlations for eclogites (exceptoutlierVR43452) and pyroxenites (except outlierVR19673ecl-g2)

high-Ca

high-Mg

001

01

1

10

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

melt08 melt02 cryst cpx01 melt09 cryst

plag

(a)

001

01

1

10(b)

RockNMORB

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 15 Reconstructed (a) high-Ca and (b) high-Mg eclogitescompared with compositions of plagioclase and cpx crystallizingfrom a MORB melt modelled using distribution coefficients ofNorman et al (2005)The MORB was calculated as a 10 equilibriummelt of MORB pyrolite [composition from Kelemen et al (1993)except Pb Pr Sr Y and Lu which were calculated as residues from1 fractional melting of pyrolite from McDonough amp Sun (1995)using melt modes of Kinzler (1997) and distribution coefficients ofNorman et al (2005)] Also shown are mixtures of crystals (consistingof 50 plagioclase plus 50 cpx) and melt in the proportions 0901and 0208

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1866

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

precursors (eg gabbros SE Indian Ridge average 22Hart et al 1999) may reflect the preferred loss of fluid-mobile elements during dehydration (Becker et al 2000)although large uncertainties on these low-abundance ele-ments in the constituent minerals and reconstructedwhole-rocks makes this interpretation less than robustThe REE patterns of most high-Mg eclogites are too flatto represent melting residues as partial melting shouldproduce low LREEHREE (Electronic Appendix 10) Incontrast the marked LREE depletion of five of sevenhigh-Ca eclogites (LaLufrac14 02^08) contrasts with that ofassumed evolved gabbroic protoliths (LaLufrac1415^32Hart et al 1999) and can be modelled as a consequence ofpartial melting during subduction after eclogitization (iein the presence of garnet which retains HREE relative toLREE eg Barth et al 2002a) has taken place (Aulbachet al 2003)Our results are compatible with previous studies on

eclogite xenoliths from the Slave Craton includingthose from the same kimberlite that have linked thesesamples or some of these samples to Paleoproterozoicsubduction events (the 195^191Ga formation of theHottah terrane and the 18Ga formation of the GreatBear magmatic arc) in the Slave craton (Griffin et al1999 Kopylova et al 1999 Heaman et al 2002 2006Schmidberger et al 2005 2007)The compositional features discussed above although

compatible with crustal protoliths are not necessarilydefinitive of a subduction origin However if direct crystal-lization from basaltic melts at high pressure (ie mantledepths) is invoked instead specific conditions would haveto be met to satisfy the observations The Mg-number ineclogites from Lac de Gras are lower than those expectedof primary high-pressure mantle melts (eg Walter 1998OrsquoHara amp Herzberg 2002) requiring crystallization athigh pressures after olivine fractionation This originwould further require shielding from peridotitic wallrockwith which the fractionated melts would no longer be inequilibrium (Barth et al 2001) In such a setting the lowrelative abundances of fluid-mobile elements comparedwith those of less fluid-mobile elements and of LREErelative to HREE (Table 4) would reflect an expulsion offluids during crystallization of the protoliths and latermelt mobilization respectively The eclogites clearly donot represent garnet plus cpx cumulates which wouldrequire a positive correlation of Mg-number with Yb thatis at variance with the observed negative correlation ineclogitic garnets from this study (Fig 14)

Low-Mg eclogites

Garnets in low-Mg eclogites compositionally overlap withthose in massive eclogites from Jericho (NW of the Lacde Gras kimberlites) that were suggested to have a high-pressure intrusive origin based on the similarity of theirmajor-element compositions to those of Group A eclogites

fromsouthernAfrica forwhichamantleoriginhasbeenpro-posed (Kopylova et al1999) Garnets in low-Mg eclogites inthe present study have major-element compositions moresimilar to garnets in groupB rather thangroupAeclogitesThe positive slope in N-MORB-normalized HREE

abundances displayed by the low-Mg eclogites is clearlydistinct from other eclogite types and is similar to those ofboninite-like second-stage melts that are closely associatedwith ultramafic volcanics (Smithies et al 2004) (Fig 13a)The range of Mg-numbers in the low-Mg eclogites (55^66)overlaps with that of Archaean second-stage melts fromIsua (58^75) and Abitibi (52^82 Smithies et al 2004)These similarities might indicate that the low-Mg eclogiteshave an origin in a previously depleted mantle source suchas supra-subduction zone mantle Subchondritic ZrHf ofthe low-Mg eclogites contrasts with the supra-chondriticZrHf of the high-Mg and high-Ca eclogites (Aulbach et

al in preparation) and is consistent with previous cpx-controlled depletion in an arc mantle source This sourcecould have been subsequently remelted to generate theboninite-like precursor to the low-Mg eclogites as sug-gested for the protoliths of eclogites from Udachnaya(Jacob amp Foley 1999)Although some second-stage melts to which the low-Mg

eclogites are similar are closely associated with komatiitesand could contain a plume component the radiogenic ordepleted mantle-like Nd and Hf and unradiogenic Sr inthe low-Mg eclogites do not support plume involvement

Trace-element and Sr^Nd^Hf isotopeconstraints on post-formationmetasomatismSeveral low-Mg and high-Ca eclogites show negativeslopes in the LREE which may be indicative of interactionwith small amounts of LREE-enriched melt (Navon ampStolper 1987) amounting to cryptic metasomatism(ie not accompanied by crystallization of new phases) inthe nomenclature of Dawson (1984) High Nb abundancesrelative to elements of similar compatibility displayed bysome low-Mg and high-Ca eclogites (Fig 10) could resultfrom overestimation of rutile modes However high-CaeclogiteVR40345 is rutile-free and has been reconstructedwithout rutile its high Nb abundance is accompanied byLREE enrichment indicating that metasomatic additionmay be responsibleLow-Mg eclogites show a positive correlation of La with

Nb (Electronic Appendix 11a) indicating concomitantaddition of LREE and Nb This suggests that the metaso-matic agent was a solute-rich fluid or silicate silicic melt ashydrous fluids and carbonatites do not carry enough HFSEto affect the budget of eclogites (Green amp Wallace 1988Klemme et al 1995 Adam et al 1997 Stalder et al 1998)Of the samples that are enriched in Nb all but one (high-Ca eclogiteVR50909) show a simultaneous enrichment in

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1867

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Hf (Electronic Appendix 11b) Together these trace-element relationships suggest that both SmNd and LuHfwere lowered in these samples Their low parent^daughterratios contrast with higher ratios in unenriched samplesand resulted in the ingrowth of a range of Nd and Hf iso-tope compositions Compared with the low-Mg and high-Ca eclogites the high-Mg eclogites have a narrow range ofLa and Nb concentrations which correlates with a narrowrange in Nd and Hf isotope compositions (Fig 12)Precise model ages cannot be calculated because parent^

daughter ratios are available only from in situ analyseswhich have relatively large uncertainties Howeverevidence for long-term enrichment and depletion may bediscerned Low-Mg eclogites even those that are enrichedin LREE have relatively high SmNd as a result of havingoverall positive slopes in their REE patterns relative toMORB whereas the high-Mg eclogites have relativelyflat REE patterns and therefore lower SmNd (Fig 11)This translates into lower eNd for high-Mg eclogites thanfor low-Mg eclogites In contrast they show no systematicdifference with respect to 176Hf177Hf Because the low- andhigh-Mg eclogites are proposed to have different origins asdiscussed above and therefore did not necessarily have thesame initial isotopic composition and formation age theseisotopic differences cannot be used to estimate their agesA long-term coupling of trace-element and isotopic

abundances is also indicated by the fact that low-Mg eclo-gite YK1943 which has experienced Nb and LREE addi-tion has lower eNd (0) and eHf (thorn13) than low-Mg eclogiteVR43469 with higher LuHf and SmNd (eNdfrac14thorn47eHffrac14thorn128) If low-Mg eclogites YK1943 and VR43469originally had identical initial Nd and Hf isotope composi-tions and if the parent^daughter ratios are accurate andwere modified only once during their evolution shortlyafter metamorphism (when their isotopic compositionwould have been reset) this overprint can be calculatedfor both isotopic systems to have occurred at 17GaThis would suggest that Hf and Nd were added duringthe same event In contrast high-Mg eclogites VR67360and YK1949 would have had identical Ndisotope compositions at 10Ga whereas their Hf isotopecompositions are disturbed in that the sample with thehigher parent^daughter ratio has a lower 176Hf177Hf thanthe sample with the lower parent^daughter ratio Theseresults suggest that the LREE and HFSE enrichmentobserved in different samples occurred during separateevents that did not affect the eclogites equallyExtreme HFSE enrichment leading to the growth of

zircon and rutile and LREE addition leading to thegrowth of apatite in eclogites from the Jericho kimberlitein the northern Slave craton have been ascribed to twoseparate metasomatic events at around 18Ga (quasi-coincident with the age of formation and metamorphism)and 13^10Ga respectively (Heaman et al 2002 2006

Schmidberger et al 2005) The authors associated theseevents with Paleoproterozoic subduction and aMesoproterozoic thermal disturbance in the Slave litho-spheric mantle possibly linked to the Mackenzie igneousevent Mesoproterozoic Nd model ages mostly between13 and 14Ga were reported for eclogites from thenearby Ekati kimberlites (Jacob et al 2003) Although wedo not observe zircon or apatite in volatile-free eclogitesfrom Lac de Gras and the age constraints obtained fromour data are weak they support a link between theseevents described for other eclogite populations and themetasomatic enrichment observed in our studyThe Sr isotope data in the present study are similar to

those of Schmidberger et al (2007) who suggested that thedepleted-mantle-like values are consistent with derivationof the protoliths from an oceanic mantle source However87Sr86Sr in volatile-free eclogites in the present study isremarkably constant considering their Nd and Hf isotopevariability (Fig 12) and the variability of trace-elementpatterns (Fig 10) Given the evidence for dehydration andmelting in many of the eclogites fractionation of highlyincompatible Rb from Sr and differential ingrowth ofradiogenic Sr would be expectedThis may indicate a rela-tively recent overprint and considering the similarityof 87Sr86Sr to present-day depleted mantle which is signif-icantly lower than 87Sr86Sr of North American kimber-lites a juvenile (asthenospheric) source

PyroxenitesPyroxenites from Lac de Gras often show relationshipsin their major-element contents that are different fromthose observed in the eclogites For example pyroxeniticgarnets show a negative correlation between CaO andNa2O contents whereas these oxides are positively corre-lated for garnets in eclogites (Fig 5) CaO is well correlatedwith C2O3 in pyroxenitic garnets but not in eclogitic gar-nets (Fig 4) Mg-number is positively correlated withYb abundance for pyroxenitic garnets but anti-correlatedfor eclogitic garnets (Fig 14) This suggests that the pyrox-enites in the present study are not related to the eclogitesby fractionation or accumulation processes as proposedfor the origin of some pyroxenites (Foley et al 2003)Pyroxenites have also been interpreted as the products

of crystal segregation in magma channels with alteredoceanic crust as the magma source this would explaintheir Eu anomalies and stable and radiogenic isotopecompositions (Pearson et al 1993 Becker 1996) As thehigh-pressure liquidus phases of most terrestrial basaltsare cpx and garnet (OrsquoHara amp Herzberg 2002) partialmelts of eclogites with basaltic composition wouldcrystallize cpx and garnet but not opx More than half ofthe pyroxenites from Lac de Gras contain opx suggestingthat they probably did not crystallize directly fromeclogite-derived melts

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1868

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Low-Cr websterites and pyroxenites could be crystalli-zation products of mafic melts that interacted with perido-titic mantle thus becoming more refractory The positivecorrelation of Yb and Mg-number in garnet would allowan origin by fractional crystallization from such meltsAlternatively they formed by reaction of silicic (possiblyeclogite-derived) melts with mantle peridotites(eg Kelemen et al 1993 Aulbach et al 2002) This agreeswith the similarity of hypothetical melts in equilibriumwith pyroxenitic garnets in most samples to eclogite-derived melts (Electronic Appendix 12) Niobium concen-trations in all hypothetical melts are much higher than ineclogite-derived melts regardless of whether or not rutileis assumed to be in the residue The same is true for LaCe and Zr concentrations in some samples which mayindicate that NbZr and LREE were added to thepyroxenites some time after their formation as discussedin the previous section for some eclogites Because mini-mum ages for pyroxenitic rutile with an unradiogenicHf isotope composition (calculated by projecting the176Hf177Hf to the model depleted mantle isotopic evolutioncurve assuming zero ingrowth) range to 178066Ga(Aulbach et al in preparation) it appears likely that thepyroxenites were not formed recently (eg during kimber-lite-related or precursory melt migration) and that some ofthe samples are at least MesoproterozoicGarnet pyroxeniteYK1952 consists of schlieren of equili-

brated garnet pyroxenite in a lsquomatrixrsquo of spongy cpxfine cpx^opx intergrowths embayed garnet and meltpatches (Fig 2f) Garnet in the pyroxenite schlieren hasvery low LREE and HFSE abundances Orthopyroxenein the matrix has distinctly higher Mg-number andCr-number and contains more CaO than opx from thepyroxenite schlieren whereas opx rims in the schlierenare compositionally similar to opx in the matrix The con-siderable compositional and textural disequilibrium inthis sample suggests that this is a young feature causedby reaction with fluids shortly before or at the time ofentrainmentAs a consequence of the small sample sizes of most of

the pyroxenites studied here a comprehensive isotopicdataset is not available Clearly more such data areneeded in addition to stable isotope data to obtain betterconstraints on the origins of this rock type Consideringthe microstructural and compositional variability ofthe pyroxenites and major-element relationships oppositeto those observed for eclogites it appears certain that pyr-oxenites form by a variety of processes some of which areunrelated to the formation of eclogites and their protoliths

Volatile-rich eclogitesVolatile-rich eclogites from Lac de Gras display the mostextreme Hf^Nd isotope systematics of all the analysedsamples Garnets in phlogopitized (YK1911) and in cal-cite-bearing eclogite (VR67112b) have the most radiogenic

Nd (eNdfrac14 2486 and 1927 respectively) Garnet inVR67112b has extremely high eHf (2516) Clinopyroxene inYK1911 has the lowest eHf measured for any mineralseparate from Lac de Gras (53) in accord with the veryhigh Hf abundance in its cpx (5 ppm) In contrastgarnet in this sample has radiogenic Hf (eHffrac14 136) Thisdiscrepancy is at least in part due to their relatively lowequilibration temperatures (7698C and 8558C respec-tively) which allowed the isotopic compositions of cpxand garnet to evolve separately at their low andhigh parent^daughter element ratios respectively forextended periods of time and possibly since formationAccordingly the cpx in YK1911 gives a minimum Hf ageof 27GaSlab-derived fluids could have been introduced during

the 27Ga collision of the western Slave Craton domainincluding the thick plume-derived deep lithosphericmantle layer and the eastern domain with its highlydepleted lithosphere This would also be consistent withthe observation that volatile-rich eclogites occur at shallowdepths in the mantle (TKroghfrac14 685^8558C correspondingto depths of 85^115 km along a 40 mWm2 geotherm)where the bulk of dehydration and production of solute-rich melts in a subducting slab might be expected to occur(Becker et al 2000 Hermann et al 2006) Their negativeEu anomalies might indicate that the fluid was derivedfrom the shallow portion of the slab that experiencedplagioclase fractionation The presence of carbonate inVR67112b requires a carbonated sourceHigh 87Sr86Sr in cpx in YK1911 is probably the conse-

quence of isotopic equilibration with high-RbSr phlogo-pite after long-term radiogenic ingrowth while Hfaddition led to retarded ingrowth of radiogenic HfThese systematics combined with radiogenic Nd appear

inconsistent with the involvement of sediments in the pet-rogenesis of this sample (eg Perfit et al 1980 Vervoortet al 1999) or a hydrous fluid such as those responsible forMARID metasomatism in the Kaapvaal craton (Konzettet al 1998)Whatever the origin of the fluids the apparentHf mobility suggests that they were not in equilibriumwith a rutile-bearing source and also that they weresolute-rich in nature (ie hydrous melts able to dissolveHFSE) and the relatively high HREE abundances of thevolatile-rich eclogites indicate a garnet-free source suchas amphibolite This allowed evolution of the sample atrelatively high time-integrated SmNd

SUMMARY AND CONCLUSIONS

(1) The Lac de Gras kimberlites carried abundant eclo-gite and pyroxenite xenoliths to the surface Threemain eclogite types are distinguished in terms ofmajor-element composition and trace element

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1869

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

patterns high-Mg high-Ca and low-Mg eclogitesThe garnets of all eclogite types have anti-correlatedMg-number and Yb pointing to olivine fractionationas a controlling factor in the formation of their pre-cursor rocks Their flat or positively sloped HREEpatterns qualitatively indicate that the melt parentalto the precursor rocks was not in equilibrium with agarnet-bearing residue

(2) Reconstructed high-Mg eclogites and high-Caeclogites have flat HREE patterns positive Pb andSr anomalies and low ZrSm similar to oceanic gab-bros reported in the literature High-Mg eclogiteshave variably low HREE contents and some havesmall positive Eu anomalies possibly suggestive ofmore primitive relatively plagioclase-rich protolithswhereas high-Ca eclogites have higher HREE abun-dances and small or absent Eu anomalies that requiremore cpx- andor melt-rich protoliths if a subductionorigin is assumed

(3) Low-Mg eclogites have distinctive trace-element pat-terns that show a marked positive slope in the HREEand subchondritic ZrHf They resemble someArchaean second-stage melts (lsquoboninitesrsquo) and theirprotoliths appear to have formed from a previouslydepleted source

(4) After formation the eclogites were subject to a varietyof processes Most have low ratios of fluid-mobile vsless fluid-mobile elements possibly as a result of dehy-dration of the protoliths In addition some of the low-Mg and high-Ca eclogites have low LREEHREEconsistent with partial melt lossWhereas the low-Mgeclogites show concomitant enrichment in La Nb andZr some high-Mg and high-Ca eclogites show evi-dence for variable and unsystematic (decoupled)LREE and Nb addition

(5) Isotopically the trace-element characteristics of thedifferent eclogite types translate into lower eNd forhigh-Mg eclogites (negative slope in MREEfrac14 lowerSmNd) than for low-Mg eclogites (positive slope inMREEfrac14higher SmNd) Within the low-Mg groupsamples that show evidence for metasomatism (lowLuHf and SmNd) have lower eNd and eHf than asample that was apparently not metasomatized BothNd and Hf addition can be dated to 17Ga if identi-cal initial isotope compsitions are assumed for thelow-Mg eclogites In contrast the relatively constantand depleted mantle-like 87Sr86Sr of eclogitic cpxdespite variable Nd and Hf isotope compositions andparent^daughter ratios suggests recent input of juve-nile material

(6) Garnet and cpx in pyroxenites show major-elementrelationships different from those in eclogitesTherefore these two rock types are probably notrelated by fractionation processes The presence of

opx precludes direct crystallization from slab-derivedmelts Pyroxenites may be the product of (slab-derived) melts hybridized during assimilation of peri-dotitic mantle

(7) Garnets in phlogopite- and carbonate-bearing eclo-gites have extreme isotopic compositions includingthe most radiogenic Nd and Hf in the datasetNegative Eu anomalies may point to the involvementof a crustal protolith Highly radiogenic Sr radiogenicNd and unradiogenic Hf in the reconstructedphlogopite-bearing sample combined with highHREE abundances and evidence for Hf mobilitymay indicate involvement of solute-rich fluids origi-nating in a garnet- and rutile-free source

ACKNOWLEDGEMENTSHelp in the clean laboratories and with the analyticalfacilities by S Elhlou C Lawson A Sharma P Wielandand M Zhang is greatly appreciated Bill Griffin isthanked for his support and for agreeing to reviewseveral versions of the manuscript despite disagreeingwith many of the conclusions presented here The paperbenefited greatly from the detailed and constructivereviews provided by the editors Gareth Davies andMarjorie Wilson and by Matthias Barth Larry Heamanand an anonymous reviewer This work was funded by aMacquarie University International PostgraduateAward and Postgraduate Research Fund (SA) by anARC SPIRT grant sponsored by Kennecott Canada Incand by ARC Large and Discovery Grants to SYOrsquoRAnalytical data were obtained using instrumentationfunded by ARC LIEF and DEST Systemic InfrastructureGrants industry partners and Macquarie UniversityThis is Publication 481 from the ARC National KeyCentre for Geochemical Evolution and Metallogeny ofContinents (wwwesmqeduauGEMOC)

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online

REFERENCESAdam J GreenT H Sie S H amp Ryan C G (1997) Trace element

partitioning between aqueous fluids silicate melts and mineralsEuropeanJournal of Mineralogy 9 569^584

Aulbach S Stachel T Viljoen K S Brey G P amp Harris J W(2002) Eclogitic and websteritic diamond sources beneath theLimpopo Beltccedilis slab-melting the link Contributions to Mineralogy

and Petrology 143 56^70Aulbach S GriffinW L Pearson N J OrsquoReilly S Y Kivi K amp

Doyle B (2003) Origins of eclogites beneath the central SlaveCraton Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA-0011

Aulbach S Griffin W L Pearson N J OrsquoReilly S Y amp Kivi K(2004a) Mantle formation and evolution Slave Craton constraints

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1870

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

from HSE abundances and Re^Os isotope systematics of sulfideinclusions in mantle xenocrysts Chemical Geology 208 61^88

Aulbach S Griffin W L OrsquoReilly S Y amp McCandless T E(2004b) Genesis and evolution of the lithospheric mantle beneaththe Buffalo HeadTerrane Alberta (Canada) Lithos 77 413^451

Aulbach S GriffinW L Pearson N J OrsquoReilly S Y amp Doyle B(2007) Lithosphere formation in the central Slave Craton(Canada) plume subcretion or lithosphere accretion Contributionsto Mineralogy and Petrology (in press) doi 101007s00410-007-0200-1

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2001)Geochemistry of xenolithic eclogites from West Africa Part IA link between low MgO eclogites and Archean crust formationGeochimica et Cosmochimica Acta 65 1499^1527

Barth M G Foley S F amp Horn I (2002a) Partial melting inArchean subduction zones constraints from experimentally deter-mined trace element partition coefficients between eclogitic miner-als and tonalitic melts under upper mantle conditions PrecambrianResearch 113 323^340

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2002b)Geochemistry of xenolithic eclogites from West Africa part 2Origins of the high MgO eclogites Geochimica et Cosmochimica Acta

66 4325^4345Beard B L Fraracci K NTaylor L A Snyder G A Clayton R

A Mayeda T K amp Sobolev NV (1996) Petrography andgeochemistry of eclogites from the Mir kimberliteYakutia RussiaContributions to Mineralogy and Petrology 125 293^310

Becker H (1996) Geochemistry of garnet peridotite massifs fromlower Austria and the composition of deep lithosphere beneatha Palaeozoic convergent plate margin Chemical Geology 134 49^65

Becker H Jochum K P amp Carlson R W (2000) Trace elementfractionation during dehydration of eclogites from high-pressureterranes and the implications for element fluxes in subductionzones Chemical Geology 163 65^99

Bell K amp Tilton G R (2002) Probing the mantle The story fromcarbonatites EOS Transactions American Geophysical Union 83275^277

Benoit M Polvecurren M amp Ceuleneer G (1996) Trace element andisotopic characterisation of mafic cumulates in a fossil mantlediapir (Oman ophiolite) Chemical Geology 134 199^214

BleekerW (2003) The late Archean record a puzzle in ca 35 piecesLithos 71 99^134

BleekerW Ketchum JW F JacksonV A amp Villeneuve M (1999)The central Slave Basement Complex Part I Its structural topol-ogy and autochthonous core Canadian Journal of Earth Sciences 361083^1109

Blichert-Toft J amp Albare de F (1997) The Lu^Hf isotope geochemis-try of chondrites and the evolution of the mantle^crust systemEarth and Planetary Science Letters 148 243^258

Blichert-Toft J Chauvel C amp Albare de F (1997) Separation of Hfand Lu for high precision isotope analysis of rock samples by mag-netic sector-multiple collector ICP-MS Contributions to Mineralogy

and Petrology 127 248^260Brey G P amp Kolaquo hler T (1990) Geothermobarometry in four-phase

lherzolites II New thermobarometers and practical assessment ofexisting thermobarometers Journal of Petrology 31 1353^1378

Brey G P Kolaquo hlerT amp Nickel K G (1990) Geothermobarometry infour-phase lherzolites I Experimental results from 10 to 60 kbJournal of Petrology 31 1313^1352

Caporuscio F amp Smyth J (1990) Trace element crystal chemistryof mantle eclogites Contributions to Mineralogy and Petrology 105550^561

Carbno G B amp Canil D (2002) Mantle structure beneath the SWSlave Craton Canada constraints from garnet geochemistry in theDrybones Bay Kimberlite Journal of Petrology 43 129^142

Coleman R G Lee E D Beatty L B amp Brannock W W (1965)Eclogites and eclogites their differences and similaritiesGeological Society of America Bulletin 76 483^508

Creaser R A Grulaquo tter H Carlson J amp Crawford B (2004)Macrocrystal phlogopite Rb^Sr dates for the Ekati propertykimberlites Slave Province Canada evidence for multiple intru-sive episodes in the Paleocene and Eocene Lithos 76 399^414

Davies R M GriffinW L Pearson N J Andrew A S Doyle BJ amp OrsquoReilly S Y (1999) Diamonds from the deep Pipe DO-27Slave Craton Canada In Gurney J J Gurney J L Pascoe MD amp Richardson S H (eds) Proceedings of the 7th International

Kimberlite Conference CapeTown Red Roof Design pp 148^155Davies R M Griffin W L OrsquoReilly S Y amp Doyle B J (2004)

Mineral inclusions and geochemical characteristics of microdia-monds from the DO27 A154 A21 A418 DO18 DD17 and RanchLake kimberlites at Lac de Gras Slave Craton Canada Lithos 7739^55

DavisW J amp Hegner E (1992) Neodymium isotopic evidence for thetectonic assembly of Late Archean crust in the Slave Provincenorthwest Canada Contributions to Mineralogy and Petrology 111493^504

Davis W J Jones A G Bleeker W amp Grutter H (2003)Lithosphere development in the Slave craton a linked crustal andmantle perspective Lithos 71 575^589

Dawson J B (1984) Contrasting types of upper-mantle metasoma-tism In Kornprobst J (ed) Kimberlites II The Mantle and

Crust^Mantle Relationships Amsterdam Elsevier pp 289^294DePaolo D J (1981) Neodymium isotopes in the Colorado Front

Range and crust^mantle evolution in the Proterozoic Nature 291193^196

Erlank A J amp Kushiro I (1970) Potassium contents of syntheticpyroxenes at high temperatures and pressures Carnegie Institution ofWashingtonYearbook 68 433^439

Foley S F Barth M G amp Jenner G A (2000) Rutilemelt partitioncoefficients for trace elements and an assessment of the influence ofrutile on the trace element characteristics of subduction zonemagmas Geochimica et Cosmochimica Acta 64 933^938

Foley S F Buhre S amp Jacob D E (2003) Evolution of the Archaeancrust by delamination and shallow subduction Nature 421 249^252

Graham I Burgess J L Bryan D Ravenscroft P J Thomas EDoyle B J Hopkins R amp Armstrong K A (1999) Explorationhistory and geology of the Diavik Kimberlites Lac de GrasNorthwest Territories Canada In Gurney J J Gurney J LPascoe M D amp Richardson S H (eds) Proceedings of the 7th

International Kimberlite Conference Cape Town Red Roof Designpp 262^279

Green D H amp Wallace M E (1988) Mantle metasomatism byephemeral carbonatite melts Nature 336 459^462

Green T H Blundy J D Adam J amp Yaxley G M (2000) SIMSdetermination of trace element partition coefficients betweengarnet clinopyroxene and hydrous basaltic liquids at 2^75 GPaand 1080^12008C Lithos 53 165^187

GriffinW L amp OrsquoReilly SY (2007) Cratonic lithospheric mantle Isanything subducted Episodes 30(1) 43^53

Griffin W L Wass S Y amp Hollis J D (1984) Ultramafic xenolithsfrom Bullenmerri and Gnotuk maars Victoria Australia petrol-ogy of a sub-continental crust^mantle transition Journal of

Petrology 25 53^87GriffinW L Doyle B J Ryan C G Pearson N J OrsquoReilly SY

Davies R Kivi K van Achterbergh E amp Natapov LM (1999)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1871

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Layered mantle lithosphere in the Lac de Gras Area SlaveCraton composition structure and origin Journal of Petrology 40705^727

Griffin W L Pearson N J Belousova E Jackson S E vanAchterbergh E OrsquoReilly SY amp Shee S R (2000)The Hf isotopecomposition of cratonic mantle LAM-MC-ICPMS analysis ofzircon megacrysts in kimberlites Geochimica et Cosmochimica Acta

64 133^147Griffin W L OrsquoReilly S Y Doyle B J Pearson N J

Coopersmith H Kivi K Malkovets V amp Pokhilenko N(2004) Lithosphere mapping beneath the North American plateLithos 77 873^922

Grulaquo tter H S Apter D B amp Kong J (1999) Crust^mantle couplingevidence from mantle-derived xenocrystic garnets In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 307^312

Hart S R Blusztajn J Dick H J B Meyer P S ampMuehlenbachs K (1999) The fingerprint of seawater circulationin a 500-meter section of ocean crust gabbros Geochimica et

Cosmochimica Acta 63 4059^4080Harte B amp Kirkley M B (1997) Partitioning of trace elements

between clinopyroxene and garnet Data from mantle eclogitesChemical Geology 136 1^24

Heaman L M Creaser R A amp Cookenboo H O (2002) Extremeenrichment of high field strength elements in Jericho eclogitexenoliths A cryptic record of Paleoproterozoic subduction partialmelting and metasomatism beneath the Slave craton CanadaGeology 30 507^510

Heaman L M Kjarsgaard B A amp Creaser R A (2004) The tem-poral evolution of North American kimberlites Lithos 76 377^397

Heaman L M Creaser R A Cookenboo H O amp Chacko T(2006) Multi-stage modification of the northern Slave mantlelithosphere Evidence from zircon- and diamond-bearing eclogitexenoliths entrained in Jericho Kimberlite Journal of Petrology 47821^858

Hermann J Cpandler C Hack A amp Korsakov A (2006) Aqueousfluids and hydrous melts in high-pressure and ultra-high pressurerocks implications for element transfer in subduction zones Lithos92 399^417

Hills DV amp Haggerty S E (1989) Petrochemistry of eclogites fromthe Koidu Kimberlite Complex Sierra Leone Contributions to

Mineralogy and Petrology 103 397^422Hoffman P F (1989) Precambrian geology and tectonic history of

North America In Bally A amp Palmer A (eds) The Geology of

North AmericaccedilAn Overview Boulder CO Geological Society ofAmerica pp 447^512

IrelandT R Rudnick R L amp Spetsius Z (1994) Trace-elements indiamond inclusions from eclogites reveal link to Archean granitesEarth and Planetary Science Letters 128 199^213

Irvine G J Pearson D G Kjarsgaard B A Carlson R WKopylova M G amp Dreibus G (2003) Evolution of the litho-spheric mantle beneath Northern Canada a Re^Os isotope andPGE study of kimberlite-derived peridotite xenoliths fromSomerset Island and a comparison to the Slave and Kaapvaalcratons Lithos 71 461^488

Isachsen C E amp Bowring S A (1994) Evolution of the Slave cratonGeology 22 917^920

Jacob D E (2004) Nature and origin of eclogite xenoliths from kim-berlites Lithos 77 295^316

Jacob D E amp Foley S F (1999) Evidence for Archean ocean crustwith low high field strength element signature from diamondifer-ous eclogite xenoliths Lithos 48 317^336

Jacob D Jagoutz E Lowry D Mattey D amp Kudrjavtseva G(1994) Diamondiferous eclogites from Siberiaccedilremnants ofArchean oceanic-crust Geochimica et Cosmochimica Acta 585191^5207

Jacob D E Fung A Jagoutz E amp Pearson D G (2003) Petrologyand geochemistry of eclogite xenoliths from the Ekati kimberlitesarea Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA_239

Jacob D E Bizimis M amp Salters V J M (2005) Lu^Hf and geo-chemical systematics of recycled ancient oceanic crust evidencefrom Roberts Victor eclogites Contributions to Mineralogy and

Petrology 148 707^720Jerde E A Taylor L A Crozaz G Sobolev NV amp SobolevV N

(1993) Diamondiferous eclogites from Yakutia Siberiaccedilevidencefor a diversity of protoliths Contributions to Mineralogy and Petrology

114 189^202Jones A G Ferguson I J Chave A D Evans R L amp

McNeice GW (2001) The electric lithosphere of the Slave cratonGeology 29 423^426

Kelemen P B Shimizu N amp Dunn T (1993) Relative depletionof niobium in some arc magmas and the continental-crust-partitioning of K Nb La and Ce during meltrockreaction in the upper-mantle Earth and Planetary Science Letters 120111^134

Kinzler R J (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition Application to mid-ocean ridge basalt petrogenesis Journal of Geophysical Research 102853^874

Klemme S van der Laan S R Foley S F amp Guenther D (1995)Experimentally determined trace and minor element partitioningbetween clinopyroxene and carbonatite melt under upper mantleconditions Earth and Planetary Science Letters 133 439^448

Konzett J Armstrong R A Sweeney R J amp CompstonW (1998)The timing of MARID metasomatism in the Kaapvaal mantle anion probe study of zircons from MARID xenoliths Earth and

Planetary Science Letters 160 133^145Kopylova M G Russell J K amp Cookenboo H (1999) Mapping the

lithopshere beneath the north central Slave Craton In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 468^479

Krogh E (1988) The garnet^clinopyroxene iron^magnesiumgeothermometerccedila reinterpretation of existing experimentaldata Contributions to Mineralogy and Petrology 99 44^48

Kusky T M (1989) Accretion of Archean Slave Province Geology 1763^67

LeCheminant A N amp Heaman L M (1989) Mackenzie igneousevents Canada Middle Proterozoic hotspot magmatismassociated with ocean opening Earth and Planetary Science Letters

96 38^48LeCheminant A N Richardson D G DiLabio R N W amp

Richardson K A (1996) La recherche de diamants au CanadaGeological Survey of Canada Open File 3228 268pp

MacGregor I D amp Carter J L (1970) The chemistry of clinopyrox-enes and garnets of eclogite and peridotite xenoliths from theRoberts Victor mine South Africa Physics of the Earth and PlanetaryInteriors 3 391^397

MacGregor I D amp Manton W I (1986) Roberts Victor eclogitesancient oceanic crust Journal of Geophysical Research 91 14063^14079

MacKenzie J M amp Canil D (1999) Composition and thermal evo-lution of cratonic mantle beneath the central Archean SlaveProvince NWT Canada Contributions to Mineralogy and Petrology

134 313^324

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1872

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

McCandlessT E amp Gurney J J (1989) Sodium in garnet and potas-sum in clinopyroxene criteria for classifying mantle eclogites InRoss J (ed) Kimberlites and Related Rocks Carlton Blackwellpp 827^832

McDonoughW F amp Sun S-S (1995) The composition of the EarthChemical Geology 120 223^253

Navon O amp Stolper E (1987) Geochemical consequences of meltpercolation the upper mantle as a chromatographic columnJournal of Geology 95 285^307

Nikogosian I K amp Sobolev AV (1997) Ion-microprobe analysis ofmelt inclusions in olivine experience in estimating the olivine^meltpartition coefficients of trace elements Geochemistry International 35119^126

Norman M D Griffin W L Pearson N J Garcia M O ampOrsquoReilly S Y (1998) Quantitative analysis of trace element abun-dances in glasses and minerals a comparison of laser ablationinductively coupled plasma mass spectrometry solution inductivelycoupled plasma mass spectrometry proton microprobe and elec-tron microprobe data Journal of Analytical Atomic Spectrometry 13477^482

Norman M D Garcia M O amp Pietruszka A J (2005) Trace-element distribution coefficients for pyroxenes plagioclase andolivine in evolved tholeiites from the 1955 eruption of KilaueaVolcano Hawairsquoi and petrogenesis of differentiated rift-zonelavas American Mineralogist 90 888^899

OrsquoHara M J amp Herzberg C (2002) Interpretation of trace elementand isotope features of basalts Relevance of field relations petrol-ogy major element data phase equilibria and magma chambermodeling in basalt petrogenesis Geochimica et Cosmochimica Acta 662167^2191

PadghamW A amp FysonW K (1992) The Slave Province a distinctcraton CanadianJournal of Earth Sciences 29 2072^2086

Pearson D G Davies G R amp Nixon P H (1993) Geochemical con-straints on the petrogenesis of diamond facies pyroxenites from theBeni Bousera peridotite massif North Morocco Journal of Petrology34 125^172

Pearson N J Griffin W L Doyle B J OrsquoReilly S Y vanAchterbergh E amp Kivi K (1999) Xenoliths from kimberlitepipes of the Lac de gras area Slave Craton Canada In GurneyJ J Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite ConferenceCapeTown Red Roof Design pp 644^658

Perfit N R Gust D A Bence A E Arculus R J amp Taylor S R(1980) Chemical characteristics of island-arc basalts implicationsfor mantle sources Chemical Geology 30 227^256

Polat A Hofmann A W amp Rosing M T (2002) Boninite-likevolcanic rocks in the 37^38 Ga Isua greenstone belt WestGreenland geochemical evidence for intra-oceanic subductionzone processes in the early Earth Chemical Geology 184 231^254

Poudjom DjomaniY H GriffinW L OrsquoReilly SY amp Doyle B J(2005) Lithospheric domains and controls on kimberlite emplace-ment Slave Province Canada evidence from elastic thickness andupper mantle composition Geochemistry Geophysics Geosystems 6doi1010292005GC0009978

Presnall D C Gudfinnsson G H amp Walter M J (2002)Generation of mid-ocean ridge basalts at pressures from 1 to 7GPa Geochimica et Cosmochimica Acta 66 2073^2090

Schmidberger S S Heaman L M Simonetti A Creaser R A ampCookenboo H O (2005) Formation of Paleoproterozoiceclogitic mantle Slave Province (Canada) Insights from in-situHf and U^Pb isotopic analyses of mantle zircons Earth and

Planetary Science Letters 240 621^633

Schmidberger S S Simonetti A Heaman L M Creaser R A ampWhiteford S (2007) Lu^Hf in-situ Sr and Pb isotope and traceelement systematics for mantle eclogites from the Diavik diamondmine Evidence for Paleoproterozoic subduction beneath the Slavecraton Canada Earth and Planetary Science Letters 254 55^68

Schulze D J amp Helmstaedt H (1988) Coesite-sanidine eclogitesfrom kimberlite products of mantle fractionation or subductionJournal of Geology 96 435^443

Smithies R H Champion D C amp Sun S S (2004) The case forArchaean boninites Contributions to Mineralogy and Petrology 147705^721

Smyth J R Caporuscio F A amp McCormick T C (1989) Mantleeclogites evidence of igneous fractionation in the mantle Earth andPlanetary Science Letters 93 133^141

Sobolev NV amp Lavrentrsquoyev Y G (1971) Isomorphic sodium admix-ture in garnets formed at high pressure Contributions to Mineralogy

and Petrology 31 1^12Stalder R Foley S F Brey G P amp Horn I (1998) Mineral aqueous

fluid partitioning of trace elements at 900^12008C and 30^57GPa New experimental data for garnet clinopyroxene and rutileand implications for mantle metasomatism Geochimica et

Cosmochimica Acta 62 1781^1801Sun S-S amp McDonough W F (1989) Chemical and isotopic sys-

tematics of oceanic basalts implications for mantle compositionsand processes In Saunders A D amp Norris M J (eds)Magmatism in the Ocean Basins Geological Society London Special

Publications 42 313^345Taura H Yurimoto H Kurita K amp Sueno S (1998) Pressure

dependence on partition coefficients for trace elements between oli-vine and the coexisting melts Physics and Chemistry of Minerals 25469^484

Taylor L A amp Neal C R (1989) Eclogites with oceanic crustal andmantle signatures from the Bellsbank kimberlite South AfricaPart I Mineralogy petrography and whole rock chemistryJournal of Geology 97 551^567

van Achterbergh E Ryan C G amp Griffin W L (1999) Glitter online intensity reduction for the laser ablation inductively coupledplasma spectrometry Abstract 9th V M GoldschmidtConference LPI Boston MA 905 pp Abstract no 7215

van Breemen O DavisW J amp King J E (1992) Temporal distribu-tion of granitoid plutonic rocks in the Archean Slave ProvinceNorthwest Canadian Shield Canadian Journal of Earth Sciences 292186^2199

Vervoort J D Patchett P J Blichert-Toft J amp Albare de F (1999)Relationship between Lu^Hf and Sm^Nd isotopic systems in theglobal sedimentary system Earth and Planetary Science Letters 16879^99

Viljoen K S Smith C B amp Sharp Z D (1996) Stable and radio-genic isotope study of eclogite xenoliths from the Orapa kimberliteBotswana Chemical Geology 131 235^255

Walter M J (1998) Melting of garnet peridotite and theorigin of komatiite and depleted lithosphere Journal of Petrology 3929^60

Wasserburg G J Jacobsen S B DePaolo D J McCulloch M Tamp Wen T (1981) Precise determination of SmNd ratios Sm andNd isotopic abundances in standard solutions Geochimica et

Cosmochimica Acta 45 2311^2323Westerlund K J Shirey S B Richardson S H Carlson R W

Gurney J J amp Harris JW (2006) A subduction wedge origin forPaleoarchean peridotitic diamonds and harzburgites from thePanda kimberlite Slave craton evidence from Re^Os isotope sys-tematics Contributions to Mineralogy and Petrology 152 275^294

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1873

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Page 14: Origins of Xenolithic Eclogites and Pyroxenites from the Central

determined experimentally or in demonstrably equili-brated natural samples Low-Mg and high-Mg eclogiteshave Dcpxgarnet for most elements that are within therange determined for natural equilibrated eclogites andexperimental assemblages (Harte amp Kirkley 1997 Greenet al 2000 Barth et al 2002a) (Electronic Appendix 7)High-Ca eclogites plot at the low end of the range

of distribution coefficients and have very low DMREE

and DHREE outside the range indicated by the datasetof Harte amp Kirkley (1997) who noted that DREE decreaseswith increasing Ca content in eclogitic minerals Theseassemblages are therefore considered to be in equilibriumPyroxenite VR40384 has D values similar to eclogites

but at the high end of the range determined by Harte amp

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

(a)

(d)

(b)

(e)

(c)

CpxCI-chondrite CpxCI-chondrite

low-Mg high-Mg

high-Ca

phlogopite-bearingcalcite-bearing

pyroxenite

YK1914

YK1952

spongy rim

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

001

01

1

10

100

Rb U La Pr Nd Hf Eu Gd Ho Er Lu VBa Nb Ce Sr Zr Sm Ti Y Yb ScDy

VR40345

Fig 8 Extended trace-element patterns of cpx as in Fig 7

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1856

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Kirkley (1997) (Electronic Appendix 7) Another group(YK1914 and YK1952) plots at the low end but DHf DZrDTi and DV are far below the values indicated by experi-mental datasets Because pyroxenites compositionallygrade into eclogites their drastically different DHFSE

values suggest disequilibrium In contrast volatile-bearingsamples have DHFSE values that appear to be too high to bein equilibrium however this may be a consequence oftheir unusual bulk chemical composition (Ti- and K-rich)which differs substantially from that of lsquonormalrsquo eclogitesand experimental charges

GEOTHERMOBAROMETRYFollowing Schmidberger et al (2007) temperatures oflast equilibration were calculated for eclogitic andopx-free pyroxenitic assemblages using the formulationof Krogh (1988) The results are summarized in Table 2and Electronic Appendix 8 for a preset pressure of5 GPa and illustrated as a histogram in Fig 9 Foropx-bearing pyroxenites temperatures were calculatedsimultaneously with pressure [Ca-in-opx thermometryand Al-in-opx barometry using the formulations of Breyet al (1990)] to constrain their relative positions in themantle columnMinerals in some eclogites from Lac de Gras show dis-

tinct core and rim compositions or inhomogeneitiesTemperatures were not calculated for samples withspatially irresolvable inhomogeneous garnet compositions(SE01VR43479VR43480) For eclogites with spongy cpxcore compositions were used in the calculationTemperatures were calculated for both core and rim com-positions where applicable For all samples these

temperatures agree within the uncertainty of this thermo-meter (508C Brey amp Kolaquo hler 1990)Most equilibration temperatures for eclogites lie between

800 and10008C (Fig 9)The transition between the shallowand the deep lithospheric mantle layers in the central Lacde Gras area lies in this temperature interval (Griffin et al1999 Pearson et al 1999) Eclogites in this study lack thedistinct bimodal temperature distribution observed bySchmidberger et al (2007) but as in their study high-Caeclogites give higher average temperatures (10808C corre-sponding to a depth of 155 km if equilibrated along a 40mWm2 geotherm) than other eclogites types (9808Cdepth 135 km) the diamond-bearing sample (VR40345)gives thehighest temperature of the datasetHighNa2Ocon-tents in garnet from high-Ca eclogites are qualitativelyconsistent with higher equilibration pressures (Sobolev ampLavrentrsquoyev 1971) as are high K2O contents in coexistingcpx (Erlank ampKushiro1970) suggesting that high-Ca eclo-gites not only equilibrated at high temperature but also highpressure rather than representing a temperature excursionPyroxenites both opx-bearing and opx-free give somewhatlower average temperatures than eclogites (mean 9258C)and phlogopite-rich eclogites give the lowest temperatures(685 and 7698C respectively) The calcite-bearing eclogiteVR67112b gives a relatively low equilibration temperatureof 8558C

MEASURED ANDRECONSTRUCTEDWHOLE -ROCKSRationale and approachKimberlite-borne xenoliths sometimes show evidence forinfiltration of the host kimberlite (eg Taylor amp Neal1989 Barth et al 2001 2002b Heaman et al 2006) Wetherefore follow the standard approach of calculatingwhole-rock compositions from the compositions of the con-stituent phases weighted by their modal abundances Onlycore compositions were used as distinct rim compositionsmay point to late diffusion of elements from kimberlite-related melt veinsMany eclogitic whole-rock compositions reconstructed

with only garnet and cpx have negative Ti anomalies(not shown) For most of these samples rutile or ilmenitehas been identified as part of the mineral assemblage andthese phases should therefore be considered in the whole-rock reconstruction Rutile or ilmenite modes were esti-mated by assuming that the bulk-rock has no Ti anomalywhen normalized to a normal mid-ocean ridge basalt(N-MORB) composition and calculating the amount ofrutile required to erase theTi anomalyIt is of course possible that the eclogites had negative

or positive Ti anomalies relative to elements of similarcompatibility (Eu and Dy) Because we reconstructed

0

2

4

6

8

10

12volatile-rich

pyroxenite opx-bearing

pyroxenite opx-free

high-Ca

high-Mg

low-Mg

700750

800850

900950

10001050

11001150

12001250

Fre

quen

cy

Temperature

Fig 9 Histogram ofTKrogh (Krogh 1988) obtained at a preset pres-sure of 5 GPa except opx-bearing pyroxenites where TBKN wassolved simultaneously with PBKN (Brey amp Kolaquo hler 1990)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1857

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

whole-rock compositions from mineral core compositionskimberlite infiltration is unlikely to have caused fractiona-tion of Ti from Eu and Dy In a study in which rutilemodes were estimated from the differences between Ti inmeasured bulk-rocks and in bulk-rocks reconstructed withonly garnet and cpx only three of 19 eclogites recon-structed with rutile show Ti anomalies (Barth et al2002a) Even eclogite xenoliths with depleted Zr and Hfabundances that resemble those of modern island arcbasalts do not show Ti anomalies (Jacob amp Foley 1999)We therefore consider our approach to be a reasonablyrobust means of obtaining meaningful rutile modes wheresmall sample sizes preclude reliable determination of themodes of accessory phases by point-countingWhere rutilewas not recognized on thin-section scale or not analysedthe average eclogitic rutile composition was used for thereconstruction

UncertaintiesReconstructing whole-rock compositions in this way is arather crude method given the coarse grain size of theeclogites relative to the sample size and the non-uniformdistribution of the principal minerals on thin-sectionscale The uncertainty in mineral modes of medium- tocoarse-grained eclogites with grains45mm has been esti-mated by Jerde et al (1993) to be 10 plus uncertaintiesrelated to partial alteration which is minor in our samplesNevertheless the whole-rock trace-element budget israther insensitive to the exact modes for common eclogiticassemblages and REE patterns which encompass mostother elements in terms of compatibility in garnet andcpx do not change significantly even for modal variationsof 30 (Jerde et al 1993) This is not true for major-element concentrations which vary substantially with var-iations in mineral modes and we therefore do not calculatewhole-rock major-element contentsBecause the discussion that follows is based on trace-

element patterns and trends rather than absolute abun-dances whole-rock reconstruction based on mineralmodes is adequate We are confident that the uncertaintyinherent in this approach is not leading to significant mis-representation of eclogite whole-rock compositions becausethe trace-element patterns of the reconstructed whole-rocks show distinct differences (see below) that correlatewith the assignment to eclogite types based on garnetmajor-element composition and this is independent ofmineral mode Following Jerde et al (1993) in ElectronicAppendix 9 we show variations in trace-element patternfor one sample in each eclogite type for variations ingarnet and cpx abundance from 7030 to 3070 vol This shows that the trace-element patterns are generallyinsensitive to cpxgt ratios with regard to slopes andanomalies such as positive Sr Pb and Eu anomalies withthe exception of the positive Eu anomaly in high-Mgeclogite VR43479 which disappears if a high cpx mode is

assumed (not shown) Variations in rutile modes by 50(Electronic Appendix 9d) lead as would be expected tothe appearance of Ti anomalies but a rutile mode of 05instead of 1 vol does not erase the positive Nb anomalyin low-Mg eclogiteVR50909Because of the robustness of the relative trace-element

abundances in the reconstructed eclogites a garnet propor-tion of 065 and cpx proportion of 035 approximately themedian modal abundances of these minerals in volatile-free eclogites from Lac de Gras are assumed for eclogitesamples that are too small for determination of meaningfulmodes by point-counting

Within- and between-group variationsThe trace-element compositions of reconstructed whole-rocks are given in Table 4 N-MORB-normalized trace-element patterns of reconstructed whole-rocks areshown in Fig 10 The three eclogite groups have distincttrace element patterns that are particularly obvious forthe REE (Fig 11) Low-Mg eclogites have positive Pb andSr anomalies and smooth positive slopes in their MREEN

and HREEN LREEN are enriched relative to MREEN inthree samples (Fig 11a) Three samples have high Nbrelative to similarly compatible elements and comparedwith most other eclogites (Fig 10a) All high-Mg eclogitesdisplay more or less pronounced positive Eu and Sranomalies and most have positive Pb and negative Zrand P anomalies Their N-MORB-normalized REE pat-terns show shallow positive slopes from LaN to PrN orNdNand aweak negative slope from NdN to LuN (Fig11b)VR43479 is a low-Mg eclogite with regard to its garnet

Ca^Fe^Mg composition but has a trace-element patterntypical of high-Mg eclogites and will be discussed withthe latter Conversely sample YK1949 a high-Mg eclogitein terms of its garnet Ca^Fe^Mg compostion is more simi-lar to low-Mg eclogites in terms of its whole-rock trace-element pattern and is plotted in Figs 10a and 11aWhereas the patterns of most samples within the groupresemble each other the N-MORB-normalized incompati-ble-element abundances of the high-Ca eclogites vary bymore than an order of magnitude (Fig 10c) All sampleshave positive Pb anomalies and most have positive Sranomalies Some samples have negative P andor Zranomalies Niobium abundances in three of the samplesare very high similar to those for some low-Mg eclogitesHigh-Ca eclogites tend to have flatter MREEN andHREEN patterns and more uniform HREE abundancesthat the other eclogite types with humps between SmN

and HoNThe phlogopitized sample YK1911 shows a strong posi-

tive Pb anomaly and lower HREE abundances than thecalcite-bearing sample VR67112b which shows a distinctnegative Eu anomaly and a depletion of La relative to Ce(Figs 10d and 11d)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1858

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Trace-element compositions of reconstructed whole-rocks (ppm)

Sample P Sc Ti V Co Ni Ga Sr Y Zr Nb La Ce Pr Nd

Low-Mg eclogites

vr19674-g4 186 50 4649 317 57 59 144 77 24 30 36 13 31 05 24

vr43452 154 39 2253 328 49 84 176 52 32 11 01 02 07 01 08

vr43469 38 65 4377 349 54 91 142 34 28 15 02 05 15 03 17

yk1926 144 45 3406 258 52 87 145 89 18 18 20 09 29 06 33

yk1943 173 46 5732 295 41 39 146 122 25 48 82 34 85 13 64

yk19491 102 34 1693 158 31 49 100 46 10 14 03 09 21 03 17

High-Mg eclogites

vr19677-g2 138 39 5160 291 41 53 119 98 14 22 04 13 53 11 64

vr434792 46 25 1265 129 36 126 84 56 3 7 04 06 25 04 23

vr67360 49 36 1573 138 64 184 111 76 5 9 04 09 38 07 33

yk1946 153 28 4877 192 45 105 124 108 12 48 03 10 38 08 50

High-Ca eclogites

vr19674-g5 267 30 7663 200 41 67 195 32 23 43 77 01 04 02 16

vr40345 0 33 2821 175 47 98 210 179 13 13 66 60 109 11 38

vr43465 142 27 5104 220 31 83 121 197 14 35 29 05 22 05 40

vr50858 295 30 6898 201 44 94 150 48 22 38 06 01 03 02 20

vr50860 68 57 5672 418 57 71 186 109 19 11 02 02 13 03 26

vr50909 238 28 7295 201 42 92 171 48 20 57 13 01 04 02 21

yk3528 91 51 3712 341 53 155 177 49 16 9 01 02 08 02 11

Volatile-rich eclogites

vr67112b 40 84 9242 521 46 30 107 4 40 66 31 26 126 24 125

yk1911 31 39 9428 286 44 82 164 3 23 52 64 18 52 09 46

Pyroxenites

yk1914 111 49 707 207 48 547 44 31 4 3 02 06 19 03 10

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

Low-Mg eclogites

vr19674-g4 11 05 10 22 39 10 29 33 06 12 05 028 24 008

vr43452 04 02 08 13 38 12 41 49 08 05 01 nd 02 na

vr43469 11 05 09 25 44 11 34 36 06 07 02 001 09 129

yk1926 12 05 11 19 30 07 22 24 04 07 03 004 23 090

yk1943 21 08 10 29 42 10 30 32 05 16 05 013 65 017

yk19491 06 03 11 08 14 04 14 18 03 04 02 003 29 115

High-Mg eclogites

vr19677-g2 20 08 12 21 24 05 15 15 03 05 05 007 52 6811

vr434792 05 02 12 05 06 01 04 03 00 03 01 001 98 037

vr67360 06 03 14 06 09 02 06 06 01 02 02 002 86 000

yk1946 18 07 12 21 22 05 13 12 02 11 03 006 56 112

High-Ca eclogites

vr19674-g5 17 09 11 33 43 09 24 23 04 16 05 007 02 011

vr40345 09 04 10 12 22 06 16 14 02 05 nd nd 298 na

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1859

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Continued

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

vr43465 22 09 12 25 26 06 15 14 02 10 04 001 25 040

vr50858 20 08 09 33 41 08 22 21 03 10 01 nd 02 na

vr50860 14 07 11 27 37 08 22 18 03 06 00 nd 06 na

vr50909 20 10 12 33 38 07 20 20 03 16 01 002 02 181

yk3528 08 05 11 20 30 07 18 18 02 05 01 nd 08 na

Volatile-rich eclogites

vr67112b 38 10 07 53 74 15 41 34 04 18 03 012 60 004

yk1911 20 07 08 35 39 07 18 17 02 10 14 nd 73 na

Pyroxenites

yk1914 02 01 11 04 08 02 05 06 01 01 03 003 68 101

aGarnet has CandashMgndashFe relationship typical of high-Mg eclogites but whole-rock trace-element pattern is typical oflow-Mg eclogitesbGarnet has CandashMgndashFe relationship typical of low-Mg eclogites but whole-rock trace-element pattern is typical ofhigh-Mg eclogites

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

(a)

001

01

1

10(c)

001

01

10(b)

VR43452

YK1949

RockNMORB

low-Mg

RockNMORB

high-Mg

(d)

volatile-richhigh-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

phlogopite-bearingYK1911

calcite-bearingVR67112b

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 10 Extended trace-element patterns of reconstructed (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-bearing eclogites Normalized to N-MORB of Sun amp McDonough (1989)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1860

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

RockNMORB

(b)

RockNMORB

(d)

010

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu001

01

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

(a)

001

01

1

10(b)

001

01

1

10(c)

001

01

1

10(e)

high-TSchmidberger et al

low-TSchmidberger et al

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

high-Mglow-Mg

high-Cavolatile-rich

phlogopite-bearingYK1911

calcite-bearingVR67112b

YK1949VR43479

Fig 11 N-MORB normalized REE patterns for (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-richeclogites Shown for comparison in (e) are reconstructed eclogites from Schmidberger et al (2007)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1861

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Both volatile-rich whole rocks display strong negativeSr anomalies Strontium can be sited in calcite or apatiteandor indicate the presence of a Sr-rich mineral that wasnot identified in thin section such as bariteFour of the six reconstructed whole-rocks in the study

of Schmidberger et al (2007) (Fig 11e) show markedLREEHREEN enrichment that is observed only in onehigh-Ca eclogite in our study (Fig 11c) suggesting thatmany of the eclogites they studied were affected by metaso-matic LREE enrichment Two samples in their study thatare not affected by this enrichment show positive Euanomalies similar to the high-Mg eclogites in this studywhereas the distinctive REE pattern of the low-Mg eclo-gites is not recognized among their samples

SR^ND^HF I SOTOPESYSTEMATICSStrontium Nd and Hf isotope data for optically puremineral separates and reconstructed whole-rocks aregiven in Table 5 Spongy rims and alteration are likely tobe mostly removed during leaching in 6N HCl in combina-tion with repeated ultrasonication Despite this rigoroustreatment the reproducibility of Nd in garnet VR43469and in cpx YK1914 is worse than for other samples Thismay be due to inhomogeneities related to late Nd diffusionpossibly from grain boundaries containing a kimberlitecomponent The reproducibility of the 176Hf177Hf data isgenerally poorer than that of Nd which may reflect a com-bination of matrix effects caused by imperfectly purifiedsolutions and sample heterogeneity Parent^daughterratios are available only from in situ analyses which arenot precise enough to allow meaningful model age orinitial isotope ratio calculations (ie at the time of kimber-lite eruption) Because the Lac de Gras kimberlites arerelatively young (55^56 Ma Graham et al 1999) age cor-rection would result in only a minor change in isotopecomposition which is inconsequential in view of the rangeof compositions displayed by most samples and hence hasno effect on the conclusions drawnClinopyroxenes in five volatile-free eclogites and a

pyroxenite have Sr isotope ratios ranging from 0701662 to0703546 this range is similar to or somewhat lower thanthat of the depleted mantle at the time of kimberliteemplacement (0703 Bell amp Tilton 2002) Clinopyroxenein the phlogopitized sample has a distinctly high 87Sr86Srof 0751186Clinopyroxene controls 56^92 of the Nd budget (Nd

concentration in cpx weighted by mode) Neodymiumisotope ratios range from 0511660 to 0514844 Thiscorresponds to eNd variation (per 10 000 deviationof 143Nd144Nd in the sample from 143Nd144Nd in the chon-dritic mantle model reservoir) from 17 to thorn43 cpx inthe low-Mg eclogites have higher eNd than those in

high-Mg eclogites Garnets have generally have higher143Nd144Nd from 0512984 (eNdfrac14thorn 7) to 0520965(eNdfrac14thorn 162) Garnets in the two volatile-rich sampleshave the highest ratios 0522790 for garnet in the calcite-bearing sample VR67112b (eNdfrac14thorn198) and 0525721 inthe phlogopitized eclogiteYK1911 (eNdfrac14thorn 255)Garnet contributes 18^32 and cpx 48^78 of the Hf

budget The proportion controlled by rutile is 4^30(Aulbach et al in preparation) Measured 176Hf177Hf incpx ranges from 0281263 in phlogopite-bearing sampleYK1911 (eHffrac1453) to 0286130 in low-Mg eclogiteVR43469 (eHffrac14thorn119) and in garnet from 0283958 inpyroxeniteYK1915 (eHffrac14thorn42) to 0295718 in low-Mg eclo-gite YK1926 (eHffrac14thorn458) Hafnium in garnet in the phlo-gopite^apatite-bearing sample is extremely radiogenic(176Hf177Hffrac14 0354367) corresponding to an eHf of thorn2532which is close to a value determined for a garnet separatefrom a Roberts Victor eclogite (eHffrac14thorn2561 Jacob et al2005) Garnet in the calcite-bearing sample has extremelyradiogenic Nd and Hf (eNdfrac14thorn193 eHffrac14thorn2516) Garnetin pyroxenite YK1915 also has radiogenic Nd and Hf(eNdfrac14thorn27 eHffrac14thorn42)Despite the large uncertainties for reconstructed whole-

rocks (calculated as the square root of the sum of squareduncertainties of two standard errors for garnet and cpxanalyses combined with 20 propagated uncertainty onthe proportion of Nd or Hf respectively contributed byeach mineral) the isotopic data are consistent in thatthey correlate with the independently determinedparent^daughter ratios (1) Low-Mg eclogites have higherSmNd reflected in the steeper positive slope of theirN-MORB-normalized REE patterns (Fig 11a) comparedwith the high-Mg eclogites (Fig 11b) Accordingly recon-structed low-Mg eclogites have higher present-day eNd

(0 to thorn47) than high-Mg eclogites (4 and 17) (2) Low-Mg eclogite YK1943 has very high Nb and LREE abun-dances that could have been added during metasomatismthat also enhanced its Hf abundance This sample has thelowest eNd and eHf of the low-Mg eclogites (0 and thorn20respectively) In contrast apparently unenriched low-Mgeclogites have more radiogenic Nd and Hf (eNdfrac14thorn47and thorn9 eHffrac14thorn128 and thorn113 respectively)The Sr^Nd isotope diagram reveals that eNd in cpx

varies from 19 to 43 at nearly constant depleted-mantle-like 87Sr86Sr (Fig 12a) In the Nd^Hf isotope diagram(Fig 12b) the reconstructed whole-rocks lie on curvedmixing lines if the lsquomixingrsquo proportions of Nd and Hffor the two lsquoend-memberrsquo minerals are very different(eg YK1911) Garnets have more radiogenic Nd andHf than coexisting cpx All reconstructed volatile-freeeclogites plot in the radiogenic field with regard to Hfor are isotopically similar to the depleted mantle Twohigh-Mg eclogites and one low-Mg eclogite have unradio-genic Nd one high-Mg eclogite has radiogenic Nd and

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1862

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 5 Individual and averaged Sr Nd and Hf isotope ratios obtained from garnet and cpx solutions

Sample 87Sr86Sr 2SE2SD 143Nd144Nd 2SE2SD eNd 2SE2SD 176Hf177Hf 2SE2SD eHf 2SE2SD

Low-Mg eclogites

VR43469 cp 0702661 0000010 0514844 0000006 43 01 0286130 0000012 119 04

VR43469 gt 0515508 0000014 56 03 0286506 0000012 132 04

VR43469 gt r 0515453 0000010 55 02 0286548 0000011 134 04

VR43469 gt r 0515384 0000007 54 01 0287244 0000010 158 04

VR43469 gt r 0515487 0000014 56 03 0286996 0000014 149 05

VR43469avg gt 0515458 0000109 55 21 0286823 0000715 143 253

WR VR43469 0515059 0000273 47 5 0286405 0001079 128 38

YK1926 cp 0703546 0000011 0513043 0000007 8 0 0283154 0000009 13 03

YK1926 gt 0513364 0000008 14 01 0295718 0000005 458 02

WR YK1926 0513120 0000026 9 1 0285980 0000716 113 25

YK1943 cp 0512143 0000007 10 01 0282213 0000005 20 02

YK1943 gt 0514461 0000012 36 02 0287510 0000008 168 03

YK1943 gt r 0514491 0000009 36 02 0287332 0000005 161 02

YK1943 gt r 0514517 0000010 37 02 0287290 0000007 160 02

YK1943avg gt 0514489 0000056 36 11 0287377 0000233 163 82

WR YK1943 0512625 0000228 0 5 0283138 0000879 13 31

High-Mg eclogites

VR43480 gt 0520965 0000017 162 03 0284950 0000010 77 04

VR67360 cp 0283509 0000074 26 26

VR67360 cp r 0283433 0000031 23 11

VR67360avg cp 0701662 0000018 0511660 0000010 19 02 0283471 0000108 25 38

VR67360 gt 0512984 0000101 7 20 0287429 0000180 165 64

WR VR67360 0511761 0000132 17 3 0284734 0000469 69 17

YK1949 cp 0703019 0000013 0512231 0000009 8 02 0282513 0000009 9 03

YK1949 gt 0513511 0000010 17 02 0290297 0000042 266 15

YK1949 gt r 0513508 0000012 17 02 0289947 0000015 254 05

YK1949 gt r 0513552 0000014 18 03

YK1949avg gt 0513523 0000049 17 10 0291569 0000494 311 175

WR YK1949 0512446 0000103 4 2 0284622 0000847 65 30

Volatile-rich eclogites

VR67112b gt 0522790 0000008 198 02 0354367 0000036 2532 13

YK1911 cp 0751186 0000009 0512165 0000007 9 01 0281263 0000009 53 03

YK1911 gt 0525721 0000013 255 03 0288051 0000012 187 04

WR YK1911 0513940 0000342 25 7 0281414 0000018 48 1

Pyroxenites

YK1914 cp 0703015 0000034 0513283 0000041 13 08

YK1914 cp r 0702956 0000005 0513200 0000027 11 05

YK1914 cp r 0513527 0000059 17 11

YK1914 cp r 0513045 0000010 8 02

YK1914avg cp 0702986 0000083 0513264 0000402 12 78

YK1915 gt 0514003 0000011 27 02 0283958 0000021 42

Analyses of full procedural replicates are designated with an lsquorrsquo after sample name SE standard error (internal errorsingle analyses) SD standard deviation (external error averaged multiple analyses) Italics used for reconstructed whole-rocks (mineral isotopic compositions and concentrations weighted by modes) Uncertainties for reconstructed whole-rocksinclude 2SE or 2SD uncertainties on each mineralrsquos isotope composition combined with 20 uncertainty in the amountof Nd or Hf contributed by the constituent minerals eNd and eHf are per 10 000 deviation from model chondritic mantle[model of Wasserburg et al (1981) and Blichert-Toft amp Albarede (1997) respectively]

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1863

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

εHf

εNd

Ekati eclogitesminus100

0

100

200

300

400

500

minus40 0 40 80 120 160 200 240 280

gt in YK1915

gt in VR67112bεNd=198 εHf=2532

minus40

0

40

80

120

160

minus40 minus20 0 20 40 60

minus60

minus40

minus20

0

20

40

070 071 072 073 074 075 076

North Americankimberlite field

εNd

87Sr86Sr

cpx in YK1914

Ekatieclogites

DM

DM

DM

DM

(a)

(b)

pyroxenites

low-Mg

high-Mg

volatile-rich

eclogites

Diavik low-TA154 S kimberlite

cpx in YK1911

Fig 12 (a) Sr^Nd for cpx and (b) Nd-Hf isotope diagram for garnet cpx and reconstructed whole-rocks (symbols with bold outline) Finedashed lines in (b) connect data points from the same sample with garnets having more radiogenic isotopic compositions Uncertainties forreconstructed whole-rocks combine 20 modal uncertainty with uncertainties on each mineralrsquos isotope composition Error bars on individualminerals may be smaller than the symbols Also shown are the isotopic compositions of the depleted mantle (DM DePaolo 1981 Griffin et al2000 Bell amp Tilton 2002) Range of 87Sr86Sr in eclogites from Ekati (Lac de Gras area 100 km north of Diavik Jacob et al 2003) in NorthAmerican kimberlites and compositions of low-temperature eclogites from Diavik and of Lac de Gras A154 South kimberlite fromSchmidberger et al (2007)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1864

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

one low-Mg eclogite has eNd similar to the depleted mantle(Fig 12b)Clinopyroxene in the phlogopitized eclogite YK1911

has unradiogenic Nd and Hf and strongly radiogenicSr whereas the coexisting garnet has radiogenic Nd andHf Although this sample contains 50 vol phlogopiteits whole-rock isotopic composition was calculated

from garnet and cpx alone because the contributionfrom phlogopite to the Hf and Nd budget is 52 and51 respectively (unpublished data) The whole-rockhas radiogenic Nd but has lower eHf than anyother sample analysed in accord with the extreme Hfabundance (53 ppm) and low LuHf of the cpx

DISCUSS IONOrigins of eclogites beneath the SlaveCratonHigh-Mg and high-Ca eclogites

With regard to their major- and trace-element contentsthe high-Mg and high-Ca eclogites from Lac de Gras arebroadly similar to variably evolved oceanic gabbros(Fig 13b and c) which may indicate an origin as subductedoceanic crust Compatible trace-element abundancesare useful to deduce potential crustal protoliths as theseelements are less susceptible to post-formation modification(Jacob 2004) The flat HREE patterns in the high-Mg andhigh-Ca eclogites are compatible with a low-pressureorigin as the oceanic crust is a product of moderate degreesof partial melting mainly within the spinel stability field(eg Presnall et al 2002) whereas melting with residualgarnet would qualitatively be expected to produce stronglyfractionated REE patterns Positive Sr and Pb anomaliesalthough compatible with a plagioclase signature couldalso be due to cryptic metasomatism (Dawson 1984) ofthe eclogites during their residence in the lithosphericmantle These anomalies are probably not related to kim-berlite infiltration as the whole-rock eclogite compositionswere reconstructed from analyses of fresh mineral coresLow ZrSm similar to that in primitive gabbros is

observed for many of the high-Mg and high-Caeclogites (Fig 13b and c) and is independent of the ratioof garnetcpx and whether or not the whole-rock wasreconstructed with rutile (Electronic Appendix 10) Thenegative trend of Mg-number against Yb in garnet(Fig 14) argues against igneous processes involving garnetaccumulation (Taylor amp Neal 1989) as suggested forexample for some high-Mg eclogites from West Africa(Barth et al 2002b) Rather the anti-correlation suggeststhat olivine fractionation played a role in the formation ofthe protoliths Fractionating olivine would excludeYb (Dolivinebasalt

frac14 002 Nikogosian amp Sobolev 1997)and incorporate Mg relative to Fe (Dfrac1415 and 05 respec-tively Taura et al1998) leading to progressive depletion ofMg and enrichment of Yb in the residual melt and henceanti-correlated Mg and Yb in the whole-rock This isinherited after eclogitization All eclogite types (exceptVR43452) plot along the olivine fractionation trend sug-gesting that they formed by a similar processHigh-Mg eclogites have lower

PREE than high-Ca

eclogites and some appear to have positive Eu anomalies

high-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

gabbrosSE Indian ridge

gabbrosOman ophiolite

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

high-Mg

RockNMORB(a)

VR43452

YK1949

low-Mg

boninite-like melts

01

001

1

10

(b)

01

001

1

10

(c)

01

001

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

Fig 13 Trace-element patterns of (a) low-Mg (b) high-Mg and(c) high-Ca eclogites compared with boninite-like volcanic rocksfrom the Isua greenstone belt (Polat et al 2002) gabbros from theOman ophiolite (Benoit et al 1996) and the SE Indian Ridge (Hartet al 1999) normalized to N-MORB of Sun amp McDonough (1989)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1865

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(EuEu up to 14) similar to more primitive cumulates(troctolite gabbros and olivine gabbros) reported in the lit-erature (Fig 13b) Although the sizes of these Eu anomaliesare not outside the analytical uncertainty their correlationwith CeYb in reconstructed eclogites supports their signif-icance (r2frac14066 nfrac14 4) The more primitive nature ofhigh-Mg eclogites is consistent with higher Mg-numbersrelative to high-Ca eclogites and may be due to olivineaccumulation during formation of their protolithsThe two samples with the lowest HREE abundanceshave the highest EuEu which is compatible with varia-tions in plagioclase modesMore evolved protoliths with higher pyroxeneplagio-

clase plus trapped melt could account for small or absentpositive Eu anomalies and higher

PHREE of high-Ca

eclogites (Fig 13c) Although the reason for the frequentassociation of graphite or diamond with high-Ca eclogitesis unclear (Aulbach et al 2003) the frequent occurrence ofkyanite in high-Ca eclogites also observed at Ekati justnorth of Lac de Gras has been argued to support a crustalprotolith (Jacob 2004) Kyanite andor corundum havebeen suggested to exsolve from high-temperature high-Alclinopyroxenes (Caporuscio amp Smyth 1990) Eclogitesfrom Lac de Gras show no exsolution textures suggestiveof such an origin although recrystallization can obscureor destroy microstructural evidence for exsolution(eg Griffin et al 1984)Modelling shows that qualitatively plagioclase accumu-

lation from a low-pressure melt derived from MORBmantle can account for positive Pb Sr and Eu and negativeZr anomalies relative to N-MORB whereas cpx accumula-tion or trapped melt increases Zr Y MREE and HREE

abundances (Fig 15) Some melt addition (for exampleby entrapment) seems to be required to increase all trace-element abundances to those observed for the eclogitesAlthough high-Ca and high-Mg eclogites could representshallow and deep portions of the same oceanic crust thehigher average temperatures obtained for the high-Caeclogites the frequent occurrence of carbon in high-Caeclogites and its absence in high-Mg eclogites the strongLREEHREE fractionation observed for six of seven high-Ca eclogites and the narrow range of compositions of thehigh-Mg eclogites suggest that they may have protolithsthat are not cogenetic Some discrepancies in particularaffecting the incompatible elements may be due to post-crystallization processes acting upon the protoliths suchas dehydration melting and later metasomatism that isevident in peridotites from the same locality (Griffin et al1999)Low K2OTh for high-Ca and high-Mg eclogites

(average 08 Table 4) relative to assumed crustal

0

2

4

6

8

10

30 40 50 60 70 80

Mg

VR43452

Yb

GARNET

pyroxenites

low-Mghigh-Mghigh-Ca

eclogites

VR19673ecl-g2

Fig 14 Garnet Mg-number vs Yb (ppm including 2s uncertainty)Lines qualitatively show separate correlations for eclogites (exceptoutlierVR43452) and pyroxenites (except outlierVR19673ecl-g2)

high-Ca

high-Mg

001

01

1

10

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

melt08 melt02 cryst cpx01 melt09 cryst

plag

(a)

001

01

1

10(b)

RockNMORB

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 15 Reconstructed (a) high-Ca and (b) high-Mg eclogitescompared with compositions of plagioclase and cpx crystallizingfrom a MORB melt modelled using distribution coefficients ofNorman et al (2005)The MORB was calculated as a 10 equilibriummelt of MORB pyrolite [composition from Kelemen et al (1993)except Pb Pr Sr Y and Lu which were calculated as residues from1 fractional melting of pyrolite from McDonough amp Sun (1995)using melt modes of Kinzler (1997) and distribution coefficients ofNorman et al (2005)] Also shown are mixtures of crystals (consistingof 50 plagioclase plus 50 cpx) and melt in the proportions 0901and 0208

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1866

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

precursors (eg gabbros SE Indian Ridge average 22Hart et al 1999) may reflect the preferred loss of fluid-mobile elements during dehydration (Becker et al 2000)although large uncertainties on these low-abundance ele-ments in the constituent minerals and reconstructedwhole-rocks makes this interpretation less than robustThe REE patterns of most high-Mg eclogites are too flatto represent melting residues as partial melting shouldproduce low LREEHREE (Electronic Appendix 10) Incontrast the marked LREE depletion of five of sevenhigh-Ca eclogites (LaLufrac14 02^08) contrasts with that ofassumed evolved gabbroic protoliths (LaLufrac1415^32Hart et al 1999) and can be modelled as a consequence ofpartial melting during subduction after eclogitization (iein the presence of garnet which retains HREE relative toLREE eg Barth et al 2002a) has taken place (Aulbachet al 2003)Our results are compatible with previous studies on

eclogite xenoliths from the Slave Craton includingthose from the same kimberlite that have linked thesesamples or some of these samples to Paleoproterozoicsubduction events (the 195^191Ga formation of theHottah terrane and the 18Ga formation of the GreatBear magmatic arc) in the Slave craton (Griffin et al1999 Kopylova et al 1999 Heaman et al 2002 2006Schmidberger et al 2005 2007)The compositional features discussed above although

compatible with crustal protoliths are not necessarilydefinitive of a subduction origin However if direct crystal-lization from basaltic melts at high pressure (ie mantledepths) is invoked instead specific conditions would haveto be met to satisfy the observations The Mg-number ineclogites from Lac de Gras are lower than those expectedof primary high-pressure mantle melts (eg Walter 1998OrsquoHara amp Herzberg 2002) requiring crystallization athigh pressures after olivine fractionation This originwould further require shielding from peridotitic wallrockwith which the fractionated melts would no longer be inequilibrium (Barth et al 2001) In such a setting the lowrelative abundances of fluid-mobile elements comparedwith those of less fluid-mobile elements and of LREErelative to HREE (Table 4) would reflect an expulsion offluids during crystallization of the protoliths and latermelt mobilization respectively The eclogites clearly donot represent garnet plus cpx cumulates which wouldrequire a positive correlation of Mg-number with Yb thatis at variance with the observed negative correlation ineclogitic garnets from this study (Fig 14)

Low-Mg eclogites

Garnets in low-Mg eclogites compositionally overlap withthose in massive eclogites from Jericho (NW of the Lacde Gras kimberlites) that were suggested to have a high-pressure intrusive origin based on the similarity of theirmajor-element compositions to those of Group A eclogites

fromsouthernAfrica forwhichamantleoriginhasbeenpro-posed (Kopylova et al1999) Garnets in low-Mg eclogites inthe present study have major-element compositions moresimilar to garnets in groupB rather thangroupAeclogitesThe positive slope in N-MORB-normalized HREE

abundances displayed by the low-Mg eclogites is clearlydistinct from other eclogite types and is similar to those ofboninite-like second-stage melts that are closely associatedwith ultramafic volcanics (Smithies et al 2004) (Fig 13a)The range of Mg-numbers in the low-Mg eclogites (55^66)overlaps with that of Archaean second-stage melts fromIsua (58^75) and Abitibi (52^82 Smithies et al 2004)These similarities might indicate that the low-Mg eclogiteshave an origin in a previously depleted mantle source suchas supra-subduction zone mantle Subchondritic ZrHf ofthe low-Mg eclogites contrasts with the supra-chondriticZrHf of the high-Mg and high-Ca eclogites (Aulbach et

al in preparation) and is consistent with previous cpx-controlled depletion in an arc mantle source This sourcecould have been subsequently remelted to generate theboninite-like precursor to the low-Mg eclogites as sug-gested for the protoliths of eclogites from Udachnaya(Jacob amp Foley 1999)Although some second-stage melts to which the low-Mg

eclogites are similar are closely associated with komatiitesand could contain a plume component the radiogenic ordepleted mantle-like Nd and Hf and unradiogenic Sr inthe low-Mg eclogites do not support plume involvement

Trace-element and Sr^Nd^Hf isotopeconstraints on post-formationmetasomatismSeveral low-Mg and high-Ca eclogites show negativeslopes in the LREE which may be indicative of interactionwith small amounts of LREE-enriched melt (Navon ampStolper 1987) amounting to cryptic metasomatism(ie not accompanied by crystallization of new phases) inthe nomenclature of Dawson (1984) High Nb abundancesrelative to elements of similar compatibility displayed bysome low-Mg and high-Ca eclogites (Fig 10) could resultfrom overestimation of rutile modes However high-CaeclogiteVR40345 is rutile-free and has been reconstructedwithout rutile its high Nb abundance is accompanied byLREE enrichment indicating that metasomatic additionmay be responsibleLow-Mg eclogites show a positive correlation of La with

Nb (Electronic Appendix 11a) indicating concomitantaddition of LREE and Nb This suggests that the metaso-matic agent was a solute-rich fluid or silicate silicic melt ashydrous fluids and carbonatites do not carry enough HFSEto affect the budget of eclogites (Green amp Wallace 1988Klemme et al 1995 Adam et al 1997 Stalder et al 1998)Of the samples that are enriched in Nb all but one (high-Ca eclogiteVR50909) show a simultaneous enrichment in

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1867

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Hf (Electronic Appendix 11b) Together these trace-element relationships suggest that both SmNd and LuHfwere lowered in these samples Their low parent^daughterratios contrast with higher ratios in unenriched samplesand resulted in the ingrowth of a range of Nd and Hf iso-tope compositions Compared with the low-Mg and high-Ca eclogites the high-Mg eclogites have a narrow range ofLa and Nb concentrations which correlates with a narrowrange in Nd and Hf isotope compositions (Fig 12)Precise model ages cannot be calculated because parent^

daughter ratios are available only from in situ analyseswhich have relatively large uncertainties Howeverevidence for long-term enrichment and depletion may bediscerned Low-Mg eclogites even those that are enrichedin LREE have relatively high SmNd as a result of havingoverall positive slopes in their REE patterns relative toMORB whereas the high-Mg eclogites have relativelyflat REE patterns and therefore lower SmNd (Fig 11)This translates into lower eNd for high-Mg eclogites thanfor low-Mg eclogites In contrast they show no systematicdifference with respect to 176Hf177Hf Because the low- andhigh-Mg eclogites are proposed to have different origins asdiscussed above and therefore did not necessarily have thesame initial isotopic composition and formation age theseisotopic differences cannot be used to estimate their agesA long-term coupling of trace-element and isotopic

abundances is also indicated by the fact that low-Mg eclo-gite YK1943 which has experienced Nb and LREE addi-tion has lower eNd (0) and eHf (thorn13) than low-Mg eclogiteVR43469 with higher LuHf and SmNd (eNdfrac14thorn47eHffrac14thorn128) If low-Mg eclogites YK1943 and VR43469originally had identical initial Nd and Hf isotope composi-tions and if the parent^daughter ratios are accurate andwere modified only once during their evolution shortlyafter metamorphism (when their isotopic compositionwould have been reset) this overprint can be calculatedfor both isotopic systems to have occurred at 17GaThis would suggest that Hf and Nd were added duringthe same event In contrast high-Mg eclogites VR67360and YK1949 would have had identical Ndisotope compositions at 10Ga whereas their Hf isotopecompositions are disturbed in that the sample with thehigher parent^daughter ratio has a lower 176Hf177Hf thanthe sample with the lower parent^daughter ratio Theseresults suggest that the LREE and HFSE enrichmentobserved in different samples occurred during separateevents that did not affect the eclogites equallyExtreme HFSE enrichment leading to the growth of

zircon and rutile and LREE addition leading to thegrowth of apatite in eclogites from the Jericho kimberlitein the northern Slave craton have been ascribed to twoseparate metasomatic events at around 18Ga (quasi-coincident with the age of formation and metamorphism)and 13^10Ga respectively (Heaman et al 2002 2006

Schmidberger et al 2005) The authors associated theseevents with Paleoproterozoic subduction and aMesoproterozoic thermal disturbance in the Slave litho-spheric mantle possibly linked to the Mackenzie igneousevent Mesoproterozoic Nd model ages mostly between13 and 14Ga were reported for eclogites from thenearby Ekati kimberlites (Jacob et al 2003) Although wedo not observe zircon or apatite in volatile-free eclogitesfrom Lac de Gras and the age constraints obtained fromour data are weak they support a link between theseevents described for other eclogite populations and themetasomatic enrichment observed in our studyThe Sr isotope data in the present study are similar to

those of Schmidberger et al (2007) who suggested that thedepleted-mantle-like values are consistent with derivationof the protoliths from an oceanic mantle source However87Sr86Sr in volatile-free eclogites in the present study isremarkably constant considering their Nd and Hf isotopevariability (Fig 12) and the variability of trace-elementpatterns (Fig 10) Given the evidence for dehydration andmelting in many of the eclogites fractionation of highlyincompatible Rb from Sr and differential ingrowth ofradiogenic Sr would be expectedThis may indicate a rela-tively recent overprint and considering the similarityof 87Sr86Sr to present-day depleted mantle which is signif-icantly lower than 87Sr86Sr of North American kimber-lites a juvenile (asthenospheric) source

PyroxenitesPyroxenites from Lac de Gras often show relationshipsin their major-element contents that are different fromthose observed in the eclogites For example pyroxeniticgarnets show a negative correlation between CaO andNa2O contents whereas these oxides are positively corre-lated for garnets in eclogites (Fig 5) CaO is well correlatedwith C2O3 in pyroxenitic garnets but not in eclogitic gar-nets (Fig 4) Mg-number is positively correlated withYb abundance for pyroxenitic garnets but anti-correlatedfor eclogitic garnets (Fig 14) This suggests that the pyrox-enites in the present study are not related to the eclogitesby fractionation or accumulation processes as proposedfor the origin of some pyroxenites (Foley et al 2003)Pyroxenites have also been interpreted as the products

of crystal segregation in magma channels with alteredoceanic crust as the magma source this would explaintheir Eu anomalies and stable and radiogenic isotopecompositions (Pearson et al 1993 Becker 1996) As thehigh-pressure liquidus phases of most terrestrial basaltsare cpx and garnet (OrsquoHara amp Herzberg 2002) partialmelts of eclogites with basaltic composition wouldcrystallize cpx and garnet but not opx More than half ofthe pyroxenites from Lac de Gras contain opx suggestingthat they probably did not crystallize directly fromeclogite-derived melts

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1868

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Low-Cr websterites and pyroxenites could be crystalli-zation products of mafic melts that interacted with perido-titic mantle thus becoming more refractory The positivecorrelation of Yb and Mg-number in garnet would allowan origin by fractional crystallization from such meltsAlternatively they formed by reaction of silicic (possiblyeclogite-derived) melts with mantle peridotites(eg Kelemen et al 1993 Aulbach et al 2002) This agreeswith the similarity of hypothetical melts in equilibriumwith pyroxenitic garnets in most samples to eclogite-derived melts (Electronic Appendix 12) Niobium concen-trations in all hypothetical melts are much higher than ineclogite-derived melts regardless of whether or not rutileis assumed to be in the residue The same is true for LaCe and Zr concentrations in some samples which mayindicate that NbZr and LREE were added to thepyroxenites some time after their formation as discussedin the previous section for some eclogites Because mini-mum ages for pyroxenitic rutile with an unradiogenicHf isotope composition (calculated by projecting the176Hf177Hf to the model depleted mantle isotopic evolutioncurve assuming zero ingrowth) range to 178066Ga(Aulbach et al in preparation) it appears likely that thepyroxenites were not formed recently (eg during kimber-lite-related or precursory melt migration) and that some ofthe samples are at least MesoproterozoicGarnet pyroxeniteYK1952 consists of schlieren of equili-

brated garnet pyroxenite in a lsquomatrixrsquo of spongy cpxfine cpx^opx intergrowths embayed garnet and meltpatches (Fig 2f) Garnet in the pyroxenite schlieren hasvery low LREE and HFSE abundances Orthopyroxenein the matrix has distinctly higher Mg-number andCr-number and contains more CaO than opx from thepyroxenite schlieren whereas opx rims in the schlierenare compositionally similar to opx in the matrix The con-siderable compositional and textural disequilibrium inthis sample suggests that this is a young feature causedby reaction with fluids shortly before or at the time ofentrainmentAs a consequence of the small sample sizes of most of

the pyroxenites studied here a comprehensive isotopicdataset is not available Clearly more such data areneeded in addition to stable isotope data to obtain betterconstraints on the origins of this rock type Consideringthe microstructural and compositional variability ofthe pyroxenites and major-element relationships oppositeto those observed for eclogites it appears certain that pyr-oxenites form by a variety of processes some of which areunrelated to the formation of eclogites and their protoliths

Volatile-rich eclogitesVolatile-rich eclogites from Lac de Gras display the mostextreme Hf^Nd isotope systematics of all the analysedsamples Garnets in phlogopitized (YK1911) and in cal-cite-bearing eclogite (VR67112b) have the most radiogenic

Nd (eNdfrac14 2486 and 1927 respectively) Garnet inVR67112b has extremely high eHf (2516) Clinopyroxene inYK1911 has the lowest eHf measured for any mineralseparate from Lac de Gras (53) in accord with the veryhigh Hf abundance in its cpx (5 ppm) In contrastgarnet in this sample has radiogenic Hf (eHffrac14 136) Thisdiscrepancy is at least in part due to their relatively lowequilibration temperatures (7698C and 8558C respec-tively) which allowed the isotopic compositions of cpxand garnet to evolve separately at their low andhigh parent^daughter element ratios respectively forextended periods of time and possibly since formationAccordingly the cpx in YK1911 gives a minimum Hf ageof 27GaSlab-derived fluids could have been introduced during

the 27Ga collision of the western Slave Craton domainincluding the thick plume-derived deep lithosphericmantle layer and the eastern domain with its highlydepleted lithosphere This would also be consistent withthe observation that volatile-rich eclogites occur at shallowdepths in the mantle (TKroghfrac14 685^8558C correspondingto depths of 85^115 km along a 40 mWm2 geotherm)where the bulk of dehydration and production of solute-rich melts in a subducting slab might be expected to occur(Becker et al 2000 Hermann et al 2006) Their negativeEu anomalies might indicate that the fluid was derivedfrom the shallow portion of the slab that experiencedplagioclase fractionation The presence of carbonate inVR67112b requires a carbonated sourceHigh 87Sr86Sr in cpx in YK1911 is probably the conse-

quence of isotopic equilibration with high-RbSr phlogo-pite after long-term radiogenic ingrowth while Hfaddition led to retarded ingrowth of radiogenic HfThese systematics combined with radiogenic Nd appear

inconsistent with the involvement of sediments in the pet-rogenesis of this sample (eg Perfit et al 1980 Vervoortet al 1999) or a hydrous fluid such as those responsible forMARID metasomatism in the Kaapvaal craton (Konzettet al 1998)Whatever the origin of the fluids the apparentHf mobility suggests that they were not in equilibriumwith a rutile-bearing source and also that they weresolute-rich in nature (ie hydrous melts able to dissolveHFSE) and the relatively high HREE abundances of thevolatile-rich eclogites indicate a garnet-free source suchas amphibolite This allowed evolution of the sample atrelatively high time-integrated SmNd

SUMMARY AND CONCLUSIONS

(1) The Lac de Gras kimberlites carried abundant eclo-gite and pyroxenite xenoliths to the surface Threemain eclogite types are distinguished in terms ofmajor-element composition and trace element

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1869

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

patterns high-Mg high-Ca and low-Mg eclogitesThe garnets of all eclogite types have anti-correlatedMg-number and Yb pointing to olivine fractionationas a controlling factor in the formation of their pre-cursor rocks Their flat or positively sloped HREEpatterns qualitatively indicate that the melt parentalto the precursor rocks was not in equilibrium with agarnet-bearing residue

(2) Reconstructed high-Mg eclogites and high-Caeclogites have flat HREE patterns positive Pb andSr anomalies and low ZrSm similar to oceanic gab-bros reported in the literature High-Mg eclogiteshave variably low HREE contents and some havesmall positive Eu anomalies possibly suggestive ofmore primitive relatively plagioclase-rich protolithswhereas high-Ca eclogites have higher HREE abun-dances and small or absent Eu anomalies that requiremore cpx- andor melt-rich protoliths if a subductionorigin is assumed

(3) Low-Mg eclogites have distinctive trace-element pat-terns that show a marked positive slope in the HREEand subchondritic ZrHf They resemble someArchaean second-stage melts (lsquoboninitesrsquo) and theirprotoliths appear to have formed from a previouslydepleted source

(4) After formation the eclogites were subject to a varietyof processes Most have low ratios of fluid-mobile vsless fluid-mobile elements possibly as a result of dehy-dration of the protoliths In addition some of the low-Mg and high-Ca eclogites have low LREEHREEconsistent with partial melt lossWhereas the low-Mgeclogites show concomitant enrichment in La Nb andZr some high-Mg and high-Ca eclogites show evi-dence for variable and unsystematic (decoupled)LREE and Nb addition

(5) Isotopically the trace-element characteristics of thedifferent eclogite types translate into lower eNd forhigh-Mg eclogites (negative slope in MREEfrac14 lowerSmNd) than for low-Mg eclogites (positive slope inMREEfrac14higher SmNd) Within the low-Mg groupsamples that show evidence for metasomatism (lowLuHf and SmNd) have lower eNd and eHf than asample that was apparently not metasomatized BothNd and Hf addition can be dated to 17Ga if identi-cal initial isotope compsitions are assumed for thelow-Mg eclogites In contrast the relatively constantand depleted mantle-like 87Sr86Sr of eclogitic cpxdespite variable Nd and Hf isotope compositions andparent^daughter ratios suggests recent input of juve-nile material

(6) Garnet and cpx in pyroxenites show major-elementrelationships different from those in eclogitesTherefore these two rock types are probably notrelated by fractionation processes The presence of

opx precludes direct crystallization from slab-derivedmelts Pyroxenites may be the product of (slab-derived) melts hybridized during assimilation of peri-dotitic mantle

(7) Garnets in phlogopite- and carbonate-bearing eclo-gites have extreme isotopic compositions includingthe most radiogenic Nd and Hf in the datasetNegative Eu anomalies may point to the involvementof a crustal protolith Highly radiogenic Sr radiogenicNd and unradiogenic Hf in the reconstructedphlogopite-bearing sample combined with highHREE abundances and evidence for Hf mobilitymay indicate involvement of solute-rich fluids origi-nating in a garnet- and rutile-free source

ACKNOWLEDGEMENTSHelp in the clean laboratories and with the analyticalfacilities by S Elhlou C Lawson A Sharma P Wielandand M Zhang is greatly appreciated Bill Griffin isthanked for his support and for agreeing to reviewseveral versions of the manuscript despite disagreeingwith many of the conclusions presented here The paperbenefited greatly from the detailed and constructivereviews provided by the editors Gareth Davies andMarjorie Wilson and by Matthias Barth Larry Heamanand an anonymous reviewer This work was funded by aMacquarie University International PostgraduateAward and Postgraduate Research Fund (SA) by anARC SPIRT grant sponsored by Kennecott Canada Incand by ARC Large and Discovery Grants to SYOrsquoRAnalytical data were obtained using instrumentationfunded by ARC LIEF and DEST Systemic InfrastructureGrants industry partners and Macquarie UniversityThis is Publication 481 from the ARC National KeyCentre for Geochemical Evolution and Metallogeny ofContinents (wwwesmqeduauGEMOC)

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online

REFERENCESAdam J GreenT H Sie S H amp Ryan C G (1997) Trace element

partitioning between aqueous fluids silicate melts and mineralsEuropeanJournal of Mineralogy 9 569^584

Aulbach S Stachel T Viljoen K S Brey G P amp Harris J W(2002) Eclogitic and websteritic diamond sources beneath theLimpopo Beltccedilis slab-melting the link Contributions to Mineralogy

and Petrology 143 56^70Aulbach S GriffinW L Pearson N J OrsquoReilly S Y Kivi K amp

Doyle B (2003) Origins of eclogites beneath the central SlaveCraton Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA-0011

Aulbach S Griffin W L Pearson N J OrsquoReilly S Y amp Kivi K(2004a) Mantle formation and evolution Slave Craton constraints

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1870

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

from HSE abundances and Re^Os isotope systematics of sulfideinclusions in mantle xenocrysts Chemical Geology 208 61^88

Aulbach S Griffin W L OrsquoReilly S Y amp McCandless T E(2004b) Genesis and evolution of the lithospheric mantle beneaththe Buffalo HeadTerrane Alberta (Canada) Lithos 77 413^451

Aulbach S GriffinW L Pearson N J OrsquoReilly S Y amp Doyle B(2007) Lithosphere formation in the central Slave Craton(Canada) plume subcretion or lithosphere accretion Contributionsto Mineralogy and Petrology (in press) doi 101007s00410-007-0200-1

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2001)Geochemistry of xenolithic eclogites from West Africa Part IA link between low MgO eclogites and Archean crust formationGeochimica et Cosmochimica Acta 65 1499^1527

Barth M G Foley S F amp Horn I (2002a) Partial melting inArchean subduction zones constraints from experimentally deter-mined trace element partition coefficients between eclogitic miner-als and tonalitic melts under upper mantle conditions PrecambrianResearch 113 323^340

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2002b)Geochemistry of xenolithic eclogites from West Africa part 2Origins of the high MgO eclogites Geochimica et Cosmochimica Acta

66 4325^4345Beard B L Fraracci K NTaylor L A Snyder G A Clayton R

A Mayeda T K amp Sobolev NV (1996) Petrography andgeochemistry of eclogites from the Mir kimberliteYakutia RussiaContributions to Mineralogy and Petrology 125 293^310

Becker H (1996) Geochemistry of garnet peridotite massifs fromlower Austria and the composition of deep lithosphere beneatha Palaeozoic convergent plate margin Chemical Geology 134 49^65

Becker H Jochum K P amp Carlson R W (2000) Trace elementfractionation during dehydration of eclogites from high-pressureterranes and the implications for element fluxes in subductionzones Chemical Geology 163 65^99

Bell K amp Tilton G R (2002) Probing the mantle The story fromcarbonatites EOS Transactions American Geophysical Union 83275^277

Benoit M Polvecurren M amp Ceuleneer G (1996) Trace element andisotopic characterisation of mafic cumulates in a fossil mantlediapir (Oman ophiolite) Chemical Geology 134 199^214

BleekerW (2003) The late Archean record a puzzle in ca 35 piecesLithos 71 99^134

BleekerW Ketchum JW F JacksonV A amp Villeneuve M (1999)The central Slave Basement Complex Part I Its structural topol-ogy and autochthonous core Canadian Journal of Earth Sciences 361083^1109

Blichert-Toft J amp Albare de F (1997) The Lu^Hf isotope geochemis-try of chondrites and the evolution of the mantle^crust systemEarth and Planetary Science Letters 148 243^258

Blichert-Toft J Chauvel C amp Albare de F (1997) Separation of Hfand Lu for high precision isotope analysis of rock samples by mag-netic sector-multiple collector ICP-MS Contributions to Mineralogy

and Petrology 127 248^260Brey G P amp Kolaquo hler T (1990) Geothermobarometry in four-phase

lherzolites II New thermobarometers and practical assessment ofexisting thermobarometers Journal of Petrology 31 1353^1378

Brey G P Kolaquo hlerT amp Nickel K G (1990) Geothermobarometry infour-phase lherzolites I Experimental results from 10 to 60 kbJournal of Petrology 31 1313^1352

Caporuscio F amp Smyth J (1990) Trace element crystal chemistryof mantle eclogites Contributions to Mineralogy and Petrology 105550^561

Carbno G B amp Canil D (2002) Mantle structure beneath the SWSlave Craton Canada constraints from garnet geochemistry in theDrybones Bay Kimberlite Journal of Petrology 43 129^142

Coleman R G Lee E D Beatty L B amp Brannock W W (1965)Eclogites and eclogites their differences and similaritiesGeological Society of America Bulletin 76 483^508

Creaser R A Grulaquo tter H Carlson J amp Crawford B (2004)Macrocrystal phlogopite Rb^Sr dates for the Ekati propertykimberlites Slave Province Canada evidence for multiple intru-sive episodes in the Paleocene and Eocene Lithos 76 399^414

Davies R M GriffinW L Pearson N J Andrew A S Doyle BJ amp OrsquoReilly S Y (1999) Diamonds from the deep Pipe DO-27Slave Craton Canada In Gurney J J Gurney J L Pascoe MD amp Richardson S H (eds) Proceedings of the 7th International

Kimberlite Conference CapeTown Red Roof Design pp 148^155Davies R M Griffin W L OrsquoReilly S Y amp Doyle B J (2004)

Mineral inclusions and geochemical characteristics of microdia-monds from the DO27 A154 A21 A418 DO18 DD17 and RanchLake kimberlites at Lac de Gras Slave Craton Canada Lithos 7739^55

DavisW J amp Hegner E (1992) Neodymium isotopic evidence for thetectonic assembly of Late Archean crust in the Slave Provincenorthwest Canada Contributions to Mineralogy and Petrology 111493^504

Davis W J Jones A G Bleeker W amp Grutter H (2003)Lithosphere development in the Slave craton a linked crustal andmantle perspective Lithos 71 575^589

Dawson J B (1984) Contrasting types of upper-mantle metasoma-tism In Kornprobst J (ed) Kimberlites II The Mantle and

Crust^Mantle Relationships Amsterdam Elsevier pp 289^294DePaolo D J (1981) Neodymium isotopes in the Colorado Front

Range and crust^mantle evolution in the Proterozoic Nature 291193^196

Erlank A J amp Kushiro I (1970) Potassium contents of syntheticpyroxenes at high temperatures and pressures Carnegie Institution ofWashingtonYearbook 68 433^439

Foley S F Barth M G amp Jenner G A (2000) Rutilemelt partitioncoefficients for trace elements and an assessment of the influence ofrutile on the trace element characteristics of subduction zonemagmas Geochimica et Cosmochimica Acta 64 933^938

Foley S F Buhre S amp Jacob D E (2003) Evolution of the Archaeancrust by delamination and shallow subduction Nature 421 249^252

Graham I Burgess J L Bryan D Ravenscroft P J Thomas EDoyle B J Hopkins R amp Armstrong K A (1999) Explorationhistory and geology of the Diavik Kimberlites Lac de GrasNorthwest Territories Canada In Gurney J J Gurney J LPascoe M D amp Richardson S H (eds) Proceedings of the 7th

International Kimberlite Conference Cape Town Red Roof Designpp 262^279

Green D H amp Wallace M E (1988) Mantle metasomatism byephemeral carbonatite melts Nature 336 459^462

Green T H Blundy J D Adam J amp Yaxley G M (2000) SIMSdetermination of trace element partition coefficients betweengarnet clinopyroxene and hydrous basaltic liquids at 2^75 GPaand 1080^12008C Lithos 53 165^187

GriffinW L amp OrsquoReilly SY (2007) Cratonic lithospheric mantle Isanything subducted Episodes 30(1) 43^53

Griffin W L Wass S Y amp Hollis J D (1984) Ultramafic xenolithsfrom Bullenmerri and Gnotuk maars Victoria Australia petrol-ogy of a sub-continental crust^mantle transition Journal of

Petrology 25 53^87GriffinW L Doyle B J Ryan C G Pearson N J OrsquoReilly SY

Davies R Kivi K van Achterbergh E amp Natapov LM (1999)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1871

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Layered mantle lithosphere in the Lac de Gras Area SlaveCraton composition structure and origin Journal of Petrology 40705^727

Griffin W L Pearson N J Belousova E Jackson S E vanAchterbergh E OrsquoReilly SY amp Shee S R (2000)The Hf isotopecomposition of cratonic mantle LAM-MC-ICPMS analysis ofzircon megacrysts in kimberlites Geochimica et Cosmochimica Acta

64 133^147Griffin W L OrsquoReilly S Y Doyle B J Pearson N J

Coopersmith H Kivi K Malkovets V amp Pokhilenko N(2004) Lithosphere mapping beneath the North American plateLithos 77 873^922

Grulaquo tter H S Apter D B amp Kong J (1999) Crust^mantle couplingevidence from mantle-derived xenocrystic garnets In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 307^312

Hart S R Blusztajn J Dick H J B Meyer P S ampMuehlenbachs K (1999) The fingerprint of seawater circulationin a 500-meter section of ocean crust gabbros Geochimica et

Cosmochimica Acta 63 4059^4080Harte B amp Kirkley M B (1997) Partitioning of trace elements

between clinopyroxene and garnet Data from mantle eclogitesChemical Geology 136 1^24

Heaman L M Creaser R A amp Cookenboo H O (2002) Extremeenrichment of high field strength elements in Jericho eclogitexenoliths A cryptic record of Paleoproterozoic subduction partialmelting and metasomatism beneath the Slave craton CanadaGeology 30 507^510

Heaman L M Kjarsgaard B A amp Creaser R A (2004) The tem-poral evolution of North American kimberlites Lithos 76 377^397

Heaman L M Creaser R A Cookenboo H O amp Chacko T(2006) Multi-stage modification of the northern Slave mantlelithosphere Evidence from zircon- and diamond-bearing eclogitexenoliths entrained in Jericho Kimberlite Journal of Petrology 47821^858

Hermann J Cpandler C Hack A amp Korsakov A (2006) Aqueousfluids and hydrous melts in high-pressure and ultra-high pressurerocks implications for element transfer in subduction zones Lithos92 399^417

Hills DV amp Haggerty S E (1989) Petrochemistry of eclogites fromthe Koidu Kimberlite Complex Sierra Leone Contributions to

Mineralogy and Petrology 103 397^422Hoffman P F (1989) Precambrian geology and tectonic history of

North America In Bally A amp Palmer A (eds) The Geology of

North AmericaccedilAn Overview Boulder CO Geological Society ofAmerica pp 447^512

IrelandT R Rudnick R L amp Spetsius Z (1994) Trace-elements indiamond inclusions from eclogites reveal link to Archean granitesEarth and Planetary Science Letters 128 199^213

Irvine G J Pearson D G Kjarsgaard B A Carlson R WKopylova M G amp Dreibus G (2003) Evolution of the litho-spheric mantle beneath Northern Canada a Re^Os isotope andPGE study of kimberlite-derived peridotite xenoliths fromSomerset Island and a comparison to the Slave and Kaapvaalcratons Lithos 71 461^488

Isachsen C E amp Bowring S A (1994) Evolution of the Slave cratonGeology 22 917^920

Jacob D E (2004) Nature and origin of eclogite xenoliths from kim-berlites Lithos 77 295^316

Jacob D E amp Foley S F (1999) Evidence for Archean ocean crustwith low high field strength element signature from diamondifer-ous eclogite xenoliths Lithos 48 317^336

Jacob D Jagoutz E Lowry D Mattey D amp Kudrjavtseva G(1994) Diamondiferous eclogites from Siberiaccedilremnants ofArchean oceanic-crust Geochimica et Cosmochimica Acta 585191^5207

Jacob D E Fung A Jagoutz E amp Pearson D G (2003) Petrologyand geochemistry of eclogite xenoliths from the Ekati kimberlitesarea Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA_239

Jacob D E Bizimis M amp Salters V J M (2005) Lu^Hf and geo-chemical systematics of recycled ancient oceanic crust evidencefrom Roberts Victor eclogites Contributions to Mineralogy and

Petrology 148 707^720Jerde E A Taylor L A Crozaz G Sobolev NV amp SobolevV N

(1993) Diamondiferous eclogites from Yakutia Siberiaccedilevidencefor a diversity of protoliths Contributions to Mineralogy and Petrology

114 189^202Jones A G Ferguson I J Chave A D Evans R L amp

McNeice GW (2001) The electric lithosphere of the Slave cratonGeology 29 423^426

Kelemen P B Shimizu N amp Dunn T (1993) Relative depletionof niobium in some arc magmas and the continental-crust-partitioning of K Nb La and Ce during meltrockreaction in the upper-mantle Earth and Planetary Science Letters 120111^134

Kinzler R J (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition Application to mid-ocean ridge basalt petrogenesis Journal of Geophysical Research 102853^874

Klemme S van der Laan S R Foley S F amp Guenther D (1995)Experimentally determined trace and minor element partitioningbetween clinopyroxene and carbonatite melt under upper mantleconditions Earth and Planetary Science Letters 133 439^448

Konzett J Armstrong R A Sweeney R J amp CompstonW (1998)The timing of MARID metasomatism in the Kaapvaal mantle anion probe study of zircons from MARID xenoliths Earth and

Planetary Science Letters 160 133^145Kopylova M G Russell J K amp Cookenboo H (1999) Mapping the

lithopshere beneath the north central Slave Craton In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 468^479

Krogh E (1988) The garnet^clinopyroxene iron^magnesiumgeothermometerccedila reinterpretation of existing experimentaldata Contributions to Mineralogy and Petrology 99 44^48

Kusky T M (1989) Accretion of Archean Slave Province Geology 1763^67

LeCheminant A N amp Heaman L M (1989) Mackenzie igneousevents Canada Middle Proterozoic hotspot magmatismassociated with ocean opening Earth and Planetary Science Letters

96 38^48LeCheminant A N Richardson D G DiLabio R N W amp

Richardson K A (1996) La recherche de diamants au CanadaGeological Survey of Canada Open File 3228 268pp

MacGregor I D amp Carter J L (1970) The chemistry of clinopyrox-enes and garnets of eclogite and peridotite xenoliths from theRoberts Victor mine South Africa Physics of the Earth and PlanetaryInteriors 3 391^397

MacGregor I D amp Manton W I (1986) Roberts Victor eclogitesancient oceanic crust Journal of Geophysical Research 91 14063^14079

MacKenzie J M amp Canil D (1999) Composition and thermal evo-lution of cratonic mantle beneath the central Archean SlaveProvince NWT Canada Contributions to Mineralogy and Petrology

134 313^324

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1872

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

McCandlessT E amp Gurney J J (1989) Sodium in garnet and potas-sum in clinopyroxene criteria for classifying mantle eclogites InRoss J (ed) Kimberlites and Related Rocks Carlton Blackwellpp 827^832

McDonoughW F amp Sun S-S (1995) The composition of the EarthChemical Geology 120 223^253

Navon O amp Stolper E (1987) Geochemical consequences of meltpercolation the upper mantle as a chromatographic columnJournal of Geology 95 285^307

Nikogosian I K amp Sobolev AV (1997) Ion-microprobe analysis ofmelt inclusions in olivine experience in estimating the olivine^meltpartition coefficients of trace elements Geochemistry International 35119^126

Norman M D Griffin W L Pearson N J Garcia M O ampOrsquoReilly S Y (1998) Quantitative analysis of trace element abun-dances in glasses and minerals a comparison of laser ablationinductively coupled plasma mass spectrometry solution inductivelycoupled plasma mass spectrometry proton microprobe and elec-tron microprobe data Journal of Analytical Atomic Spectrometry 13477^482

Norman M D Garcia M O amp Pietruszka A J (2005) Trace-element distribution coefficients for pyroxenes plagioclase andolivine in evolved tholeiites from the 1955 eruption of KilaueaVolcano Hawairsquoi and petrogenesis of differentiated rift-zonelavas American Mineralogist 90 888^899

OrsquoHara M J amp Herzberg C (2002) Interpretation of trace elementand isotope features of basalts Relevance of field relations petrol-ogy major element data phase equilibria and magma chambermodeling in basalt petrogenesis Geochimica et Cosmochimica Acta 662167^2191

PadghamW A amp FysonW K (1992) The Slave Province a distinctcraton CanadianJournal of Earth Sciences 29 2072^2086

Pearson D G Davies G R amp Nixon P H (1993) Geochemical con-straints on the petrogenesis of diamond facies pyroxenites from theBeni Bousera peridotite massif North Morocco Journal of Petrology34 125^172

Pearson N J Griffin W L Doyle B J OrsquoReilly S Y vanAchterbergh E amp Kivi K (1999) Xenoliths from kimberlitepipes of the Lac de gras area Slave Craton Canada In GurneyJ J Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite ConferenceCapeTown Red Roof Design pp 644^658

Perfit N R Gust D A Bence A E Arculus R J amp Taylor S R(1980) Chemical characteristics of island-arc basalts implicationsfor mantle sources Chemical Geology 30 227^256

Polat A Hofmann A W amp Rosing M T (2002) Boninite-likevolcanic rocks in the 37^38 Ga Isua greenstone belt WestGreenland geochemical evidence for intra-oceanic subductionzone processes in the early Earth Chemical Geology 184 231^254

Poudjom DjomaniY H GriffinW L OrsquoReilly SY amp Doyle B J(2005) Lithospheric domains and controls on kimberlite emplace-ment Slave Province Canada evidence from elastic thickness andupper mantle composition Geochemistry Geophysics Geosystems 6doi1010292005GC0009978

Presnall D C Gudfinnsson G H amp Walter M J (2002)Generation of mid-ocean ridge basalts at pressures from 1 to 7GPa Geochimica et Cosmochimica Acta 66 2073^2090

Schmidberger S S Heaman L M Simonetti A Creaser R A ampCookenboo H O (2005) Formation of Paleoproterozoiceclogitic mantle Slave Province (Canada) Insights from in-situHf and U^Pb isotopic analyses of mantle zircons Earth and

Planetary Science Letters 240 621^633

Schmidberger S S Simonetti A Heaman L M Creaser R A ampWhiteford S (2007) Lu^Hf in-situ Sr and Pb isotope and traceelement systematics for mantle eclogites from the Diavik diamondmine Evidence for Paleoproterozoic subduction beneath the Slavecraton Canada Earth and Planetary Science Letters 254 55^68

Schulze D J amp Helmstaedt H (1988) Coesite-sanidine eclogitesfrom kimberlite products of mantle fractionation or subductionJournal of Geology 96 435^443

Smithies R H Champion D C amp Sun S S (2004) The case forArchaean boninites Contributions to Mineralogy and Petrology 147705^721

Smyth J R Caporuscio F A amp McCormick T C (1989) Mantleeclogites evidence of igneous fractionation in the mantle Earth andPlanetary Science Letters 93 133^141

Sobolev NV amp Lavrentrsquoyev Y G (1971) Isomorphic sodium admix-ture in garnets formed at high pressure Contributions to Mineralogy

and Petrology 31 1^12Stalder R Foley S F Brey G P amp Horn I (1998) Mineral aqueous

fluid partitioning of trace elements at 900^12008C and 30^57GPa New experimental data for garnet clinopyroxene and rutileand implications for mantle metasomatism Geochimica et

Cosmochimica Acta 62 1781^1801Sun S-S amp McDonough W F (1989) Chemical and isotopic sys-

tematics of oceanic basalts implications for mantle compositionsand processes In Saunders A D amp Norris M J (eds)Magmatism in the Ocean Basins Geological Society London Special

Publications 42 313^345Taura H Yurimoto H Kurita K amp Sueno S (1998) Pressure

dependence on partition coefficients for trace elements between oli-vine and the coexisting melts Physics and Chemistry of Minerals 25469^484

Taylor L A amp Neal C R (1989) Eclogites with oceanic crustal andmantle signatures from the Bellsbank kimberlite South AfricaPart I Mineralogy petrography and whole rock chemistryJournal of Geology 97 551^567

van Achterbergh E Ryan C G amp Griffin W L (1999) Glitter online intensity reduction for the laser ablation inductively coupledplasma spectrometry Abstract 9th V M GoldschmidtConference LPI Boston MA 905 pp Abstract no 7215

van Breemen O DavisW J amp King J E (1992) Temporal distribu-tion of granitoid plutonic rocks in the Archean Slave ProvinceNorthwest Canadian Shield Canadian Journal of Earth Sciences 292186^2199

Vervoort J D Patchett P J Blichert-Toft J amp Albare de F (1999)Relationship between Lu^Hf and Sm^Nd isotopic systems in theglobal sedimentary system Earth and Planetary Science Letters 16879^99

Viljoen K S Smith C B amp Sharp Z D (1996) Stable and radio-genic isotope study of eclogite xenoliths from the Orapa kimberliteBotswana Chemical Geology 131 235^255

Walter M J (1998) Melting of garnet peridotite and theorigin of komatiite and depleted lithosphere Journal of Petrology 3929^60

Wasserburg G J Jacobsen S B DePaolo D J McCulloch M Tamp Wen T (1981) Precise determination of SmNd ratios Sm andNd isotopic abundances in standard solutions Geochimica et

Cosmochimica Acta 45 2311^2323Westerlund K J Shirey S B Richardson S H Carlson R W

Gurney J J amp Harris JW (2006) A subduction wedge origin forPaleoarchean peridotitic diamonds and harzburgites from thePanda kimberlite Slave craton evidence from Re^Os isotope sys-tematics Contributions to Mineralogy and Petrology 152 275^294

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1873

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Page 15: Origins of Xenolithic Eclogites and Pyroxenites from the Central

Kirkley (1997) (Electronic Appendix 7) Another group(YK1914 and YK1952) plots at the low end but DHf DZrDTi and DV are far below the values indicated by experi-mental datasets Because pyroxenites compositionallygrade into eclogites their drastically different DHFSE

values suggest disequilibrium In contrast volatile-bearingsamples have DHFSE values that appear to be too high to bein equilibrium however this may be a consequence oftheir unusual bulk chemical composition (Ti- and K-rich)which differs substantially from that of lsquonormalrsquo eclogitesand experimental charges

GEOTHERMOBAROMETRYFollowing Schmidberger et al (2007) temperatures oflast equilibration were calculated for eclogitic andopx-free pyroxenitic assemblages using the formulationof Krogh (1988) The results are summarized in Table 2and Electronic Appendix 8 for a preset pressure of5 GPa and illustrated as a histogram in Fig 9 Foropx-bearing pyroxenites temperatures were calculatedsimultaneously with pressure [Ca-in-opx thermometryand Al-in-opx barometry using the formulations of Breyet al (1990)] to constrain their relative positions in themantle columnMinerals in some eclogites from Lac de Gras show dis-

tinct core and rim compositions or inhomogeneitiesTemperatures were not calculated for samples withspatially irresolvable inhomogeneous garnet compositions(SE01VR43479VR43480) For eclogites with spongy cpxcore compositions were used in the calculationTemperatures were calculated for both core and rim com-positions where applicable For all samples these

temperatures agree within the uncertainty of this thermo-meter (508C Brey amp Kolaquo hler 1990)Most equilibration temperatures for eclogites lie between

800 and10008C (Fig 9)The transition between the shallowand the deep lithospheric mantle layers in the central Lacde Gras area lies in this temperature interval (Griffin et al1999 Pearson et al 1999) Eclogites in this study lack thedistinct bimodal temperature distribution observed bySchmidberger et al (2007) but as in their study high-Caeclogites give higher average temperatures (10808C corre-sponding to a depth of 155 km if equilibrated along a 40mWm2 geotherm) than other eclogites types (9808Cdepth 135 km) the diamond-bearing sample (VR40345)gives thehighest temperature of the datasetHighNa2Ocon-tents in garnet from high-Ca eclogites are qualitativelyconsistent with higher equilibration pressures (Sobolev ampLavrentrsquoyev 1971) as are high K2O contents in coexistingcpx (Erlank ampKushiro1970) suggesting that high-Ca eclo-gites not only equilibrated at high temperature but also highpressure rather than representing a temperature excursionPyroxenites both opx-bearing and opx-free give somewhatlower average temperatures than eclogites (mean 9258C)and phlogopite-rich eclogites give the lowest temperatures(685 and 7698C respectively) The calcite-bearing eclogiteVR67112b gives a relatively low equilibration temperatureof 8558C

MEASURED ANDRECONSTRUCTEDWHOLE -ROCKSRationale and approachKimberlite-borne xenoliths sometimes show evidence forinfiltration of the host kimberlite (eg Taylor amp Neal1989 Barth et al 2001 2002b Heaman et al 2006) Wetherefore follow the standard approach of calculatingwhole-rock compositions from the compositions of the con-stituent phases weighted by their modal abundances Onlycore compositions were used as distinct rim compositionsmay point to late diffusion of elements from kimberlite-related melt veinsMany eclogitic whole-rock compositions reconstructed

with only garnet and cpx have negative Ti anomalies(not shown) For most of these samples rutile or ilmenitehas been identified as part of the mineral assemblage andthese phases should therefore be considered in the whole-rock reconstruction Rutile or ilmenite modes were esti-mated by assuming that the bulk-rock has no Ti anomalywhen normalized to a normal mid-ocean ridge basalt(N-MORB) composition and calculating the amount ofrutile required to erase theTi anomalyIt is of course possible that the eclogites had negative

or positive Ti anomalies relative to elements of similarcompatibility (Eu and Dy) Because we reconstructed

0

2

4

6

8

10

12volatile-rich

pyroxenite opx-bearing

pyroxenite opx-free

high-Ca

high-Mg

low-Mg

700750

800850

900950

10001050

11001150

12001250

Fre

quen

cy

Temperature

Fig 9 Histogram ofTKrogh (Krogh 1988) obtained at a preset pres-sure of 5 GPa except opx-bearing pyroxenites where TBKN wassolved simultaneously with PBKN (Brey amp Kolaquo hler 1990)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1857

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

whole-rock compositions from mineral core compositionskimberlite infiltration is unlikely to have caused fractiona-tion of Ti from Eu and Dy In a study in which rutilemodes were estimated from the differences between Ti inmeasured bulk-rocks and in bulk-rocks reconstructed withonly garnet and cpx only three of 19 eclogites recon-structed with rutile show Ti anomalies (Barth et al2002a) Even eclogite xenoliths with depleted Zr and Hfabundances that resemble those of modern island arcbasalts do not show Ti anomalies (Jacob amp Foley 1999)We therefore consider our approach to be a reasonablyrobust means of obtaining meaningful rutile modes wheresmall sample sizes preclude reliable determination of themodes of accessory phases by point-countingWhere rutilewas not recognized on thin-section scale or not analysedthe average eclogitic rutile composition was used for thereconstruction

UncertaintiesReconstructing whole-rock compositions in this way is arather crude method given the coarse grain size of theeclogites relative to the sample size and the non-uniformdistribution of the principal minerals on thin-sectionscale The uncertainty in mineral modes of medium- tocoarse-grained eclogites with grains45mm has been esti-mated by Jerde et al (1993) to be 10 plus uncertaintiesrelated to partial alteration which is minor in our samplesNevertheless the whole-rock trace-element budget israther insensitive to the exact modes for common eclogiticassemblages and REE patterns which encompass mostother elements in terms of compatibility in garnet andcpx do not change significantly even for modal variationsof 30 (Jerde et al 1993) This is not true for major-element concentrations which vary substantially with var-iations in mineral modes and we therefore do not calculatewhole-rock major-element contentsBecause the discussion that follows is based on trace-

element patterns and trends rather than absolute abun-dances whole-rock reconstruction based on mineralmodes is adequate We are confident that the uncertaintyinherent in this approach is not leading to significant mis-representation of eclogite whole-rock compositions becausethe trace-element patterns of the reconstructed whole-rocks show distinct differences (see below) that correlatewith the assignment to eclogite types based on garnetmajor-element composition and this is independent ofmineral mode Following Jerde et al (1993) in ElectronicAppendix 9 we show variations in trace-element patternfor one sample in each eclogite type for variations ingarnet and cpx abundance from 7030 to 3070 vol This shows that the trace-element patterns are generallyinsensitive to cpxgt ratios with regard to slopes andanomalies such as positive Sr Pb and Eu anomalies withthe exception of the positive Eu anomaly in high-Mgeclogite VR43479 which disappears if a high cpx mode is

assumed (not shown) Variations in rutile modes by 50(Electronic Appendix 9d) lead as would be expected tothe appearance of Ti anomalies but a rutile mode of 05instead of 1 vol does not erase the positive Nb anomalyin low-Mg eclogiteVR50909Because of the robustness of the relative trace-element

abundances in the reconstructed eclogites a garnet propor-tion of 065 and cpx proportion of 035 approximately themedian modal abundances of these minerals in volatile-free eclogites from Lac de Gras are assumed for eclogitesamples that are too small for determination of meaningfulmodes by point-counting

Within- and between-group variationsThe trace-element compositions of reconstructed whole-rocks are given in Table 4 N-MORB-normalized trace-element patterns of reconstructed whole-rocks areshown in Fig 10 The three eclogite groups have distincttrace element patterns that are particularly obvious forthe REE (Fig 11) Low-Mg eclogites have positive Pb andSr anomalies and smooth positive slopes in their MREEN

and HREEN LREEN are enriched relative to MREEN inthree samples (Fig 11a) Three samples have high Nbrelative to similarly compatible elements and comparedwith most other eclogites (Fig 10a) All high-Mg eclogitesdisplay more or less pronounced positive Eu and Sranomalies and most have positive Pb and negative Zrand P anomalies Their N-MORB-normalized REE pat-terns show shallow positive slopes from LaN to PrN orNdNand aweak negative slope from NdN to LuN (Fig11b)VR43479 is a low-Mg eclogite with regard to its garnet

Ca^Fe^Mg composition but has a trace-element patterntypical of high-Mg eclogites and will be discussed withthe latter Conversely sample YK1949 a high-Mg eclogitein terms of its garnet Ca^Fe^Mg compostion is more simi-lar to low-Mg eclogites in terms of its whole-rock trace-element pattern and is plotted in Figs 10a and 11aWhereas the patterns of most samples within the groupresemble each other the N-MORB-normalized incompati-ble-element abundances of the high-Ca eclogites vary bymore than an order of magnitude (Fig 10c) All sampleshave positive Pb anomalies and most have positive Sranomalies Some samples have negative P andor Zranomalies Niobium abundances in three of the samplesare very high similar to those for some low-Mg eclogitesHigh-Ca eclogites tend to have flatter MREEN andHREEN patterns and more uniform HREE abundancesthat the other eclogite types with humps between SmN

and HoNThe phlogopitized sample YK1911 shows a strong posi-

tive Pb anomaly and lower HREE abundances than thecalcite-bearing sample VR67112b which shows a distinctnegative Eu anomaly and a depletion of La relative to Ce(Figs 10d and 11d)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1858

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Trace-element compositions of reconstructed whole-rocks (ppm)

Sample P Sc Ti V Co Ni Ga Sr Y Zr Nb La Ce Pr Nd

Low-Mg eclogites

vr19674-g4 186 50 4649 317 57 59 144 77 24 30 36 13 31 05 24

vr43452 154 39 2253 328 49 84 176 52 32 11 01 02 07 01 08

vr43469 38 65 4377 349 54 91 142 34 28 15 02 05 15 03 17

yk1926 144 45 3406 258 52 87 145 89 18 18 20 09 29 06 33

yk1943 173 46 5732 295 41 39 146 122 25 48 82 34 85 13 64

yk19491 102 34 1693 158 31 49 100 46 10 14 03 09 21 03 17

High-Mg eclogites

vr19677-g2 138 39 5160 291 41 53 119 98 14 22 04 13 53 11 64

vr434792 46 25 1265 129 36 126 84 56 3 7 04 06 25 04 23

vr67360 49 36 1573 138 64 184 111 76 5 9 04 09 38 07 33

yk1946 153 28 4877 192 45 105 124 108 12 48 03 10 38 08 50

High-Ca eclogites

vr19674-g5 267 30 7663 200 41 67 195 32 23 43 77 01 04 02 16

vr40345 0 33 2821 175 47 98 210 179 13 13 66 60 109 11 38

vr43465 142 27 5104 220 31 83 121 197 14 35 29 05 22 05 40

vr50858 295 30 6898 201 44 94 150 48 22 38 06 01 03 02 20

vr50860 68 57 5672 418 57 71 186 109 19 11 02 02 13 03 26

vr50909 238 28 7295 201 42 92 171 48 20 57 13 01 04 02 21

yk3528 91 51 3712 341 53 155 177 49 16 9 01 02 08 02 11

Volatile-rich eclogites

vr67112b 40 84 9242 521 46 30 107 4 40 66 31 26 126 24 125

yk1911 31 39 9428 286 44 82 164 3 23 52 64 18 52 09 46

Pyroxenites

yk1914 111 49 707 207 48 547 44 31 4 3 02 06 19 03 10

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

Low-Mg eclogites

vr19674-g4 11 05 10 22 39 10 29 33 06 12 05 028 24 008

vr43452 04 02 08 13 38 12 41 49 08 05 01 nd 02 na

vr43469 11 05 09 25 44 11 34 36 06 07 02 001 09 129

yk1926 12 05 11 19 30 07 22 24 04 07 03 004 23 090

yk1943 21 08 10 29 42 10 30 32 05 16 05 013 65 017

yk19491 06 03 11 08 14 04 14 18 03 04 02 003 29 115

High-Mg eclogites

vr19677-g2 20 08 12 21 24 05 15 15 03 05 05 007 52 6811

vr434792 05 02 12 05 06 01 04 03 00 03 01 001 98 037

vr67360 06 03 14 06 09 02 06 06 01 02 02 002 86 000

yk1946 18 07 12 21 22 05 13 12 02 11 03 006 56 112

High-Ca eclogites

vr19674-g5 17 09 11 33 43 09 24 23 04 16 05 007 02 011

vr40345 09 04 10 12 22 06 16 14 02 05 nd nd 298 na

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1859

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Continued

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

vr43465 22 09 12 25 26 06 15 14 02 10 04 001 25 040

vr50858 20 08 09 33 41 08 22 21 03 10 01 nd 02 na

vr50860 14 07 11 27 37 08 22 18 03 06 00 nd 06 na

vr50909 20 10 12 33 38 07 20 20 03 16 01 002 02 181

yk3528 08 05 11 20 30 07 18 18 02 05 01 nd 08 na

Volatile-rich eclogites

vr67112b 38 10 07 53 74 15 41 34 04 18 03 012 60 004

yk1911 20 07 08 35 39 07 18 17 02 10 14 nd 73 na

Pyroxenites

yk1914 02 01 11 04 08 02 05 06 01 01 03 003 68 101

aGarnet has CandashMgndashFe relationship typical of high-Mg eclogites but whole-rock trace-element pattern is typical oflow-Mg eclogitesbGarnet has CandashMgndashFe relationship typical of low-Mg eclogites but whole-rock trace-element pattern is typical ofhigh-Mg eclogites

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

(a)

001

01

1

10(c)

001

01

10(b)

VR43452

YK1949

RockNMORB

low-Mg

RockNMORB

high-Mg

(d)

volatile-richhigh-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

phlogopite-bearingYK1911

calcite-bearingVR67112b

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 10 Extended trace-element patterns of reconstructed (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-bearing eclogites Normalized to N-MORB of Sun amp McDonough (1989)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1860

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

RockNMORB

(b)

RockNMORB

(d)

010

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu001

01

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

(a)

001

01

1

10(b)

001

01

1

10(c)

001

01

1

10(e)

high-TSchmidberger et al

low-TSchmidberger et al

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

high-Mglow-Mg

high-Cavolatile-rich

phlogopite-bearingYK1911

calcite-bearingVR67112b

YK1949VR43479

Fig 11 N-MORB normalized REE patterns for (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-richeclogites Shown for comparison in (e) are reconstructed eclogites from Schmidberger et al (2007)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1861

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Both volatile-rich whole rocks display strong negativeSr anomalies Strontium can be sited in calcite or apatiteandor indicate the presence of a Sr-rich mineral that wasnot identified in thin section such as bariteFour of the six reconstructed whole-rocks in the study

of Schmidberger et al (2007) (Fig 11e) show markedLREEHREEN enrichment that is observed only in onehigh-Ca eclogite in our study (Fig 11c) suggesting thatmany of the eclogites they studied were affected by metaso-matic LREE enrichment Two samples in their study thatare not affected by this enrichment show positive Euanomalies similar to the high-Mg eclogites in this studywhereas the distinctive REE pattern of the low-Mg eclo-gites is not recognized among their samples

SR^ND^HF I SOTOPESYSTEMATICSStrontium Nd and Hf isotope data for optically puremineral separates and reconstructed whole-rocks aregiven in Table 5 Spongy rims and alteration are likely tobe mostly removed during leaching in 6N HCl in combina-tion with repeated ultrasonication Despite this rigoroustreatment the reproducibility of Nd in garnet VR43469and in cpx YK1914 is worse than for other samples Thismay be due to inhomogeneities related to late Nd diffusionpossibly from grain boundaries containing a kimberlitecomponent The reproducibility of the 176Hf177Hf data isgenerally poorer than that of Nd which may reflect a com-bination of matrix effects caused by imperfectly purifiedsolutions and sample heterogeneity Parent^daughterratios are available only from in situ analyses which arenot precise enough to allow meaningful model age orinitial isotope ratio calculations (ie at the time of kimber-lite eruption) Because the Lac de Gras kimberlites arerelatively young (55^56 Ma Graham et al 1999) age cor-rection would result in only a minor change in isotopecomposition which is inconsequential in view of the rangeof compositions displayed by most samples and hence hasno effect on the conclusions drawnClinopyroxenes in five volatile-free eclogites and a

pyroxenite have Sr isotope ratios ranging from 0701662 to0703546 this range is similar to or somewhat lower thanthat of the depleted mantle at the time of kimberliteemplacement (0703 Bell amp Tilton 2002) Clinopyroxenein the phlogopitized sample has a distinctly high 87Sr86Srof 0751186Clinopyroxene controls 56^92 of the Nd budget (Nd

concentration in cpx weighted by mode) Neodymiumisotope ratios range from 0511660 to 0514844 Thiscorresponds to eNd variation (per 10 000 deviationof 143Nd144Nd in the sample from 143Nd144Nd in the chon-dritic mantle model reservoir) from 17 to thorn43 cpx inthe low-Mg eclogites have higher eNd than those in

high-Mg eclogites Garnets have generally have higher143Nd144Nd from 0512984 (eNdfrac14thorn 7) to 0520965(eNdfrac14thorn 162) Garnets in the two volatile-rich sampleshave the highest ratios 0522790 for garnet in the calcite-bearing sample VR67112b (eNdfrac14thorn198) and 0525721 inthe phlogopitized eclogiteYK1911 (eNdfrac14thorn 255)Garnet contributes 18^32 and cpx 48^78 of the Hf

budget The proportion controlled by rutile is 4^30(Aulbach et al in preparation) Measured 176Hf177Hf incpx ranges from 0281263 in phlogopite-bearing sampleYK1911 (eHffrac1453) to 0286130 in low-Mg eclogiteVR43469 (eHffrac14thorn119) and in garnet from 0283958 inpyroxeniteYK1915 (eHffrac14thorn42) to 0295718 in low-Mg eclo-gite YK1926 (eHffrac14thorn458) Hafnium in garnet in the phlo-gopite^apatite-bearing sample is extremely radiogenic(176Hf177Hffrac14 0354367) corresponding to an eHf of thorn2532which is close to a value determined for a garnet separatefrom a Roberts Victor eclogite (eHffrac14thorn2561 Jacob et al2005) Garnet in the calcite-bearing sample has extremelyradiogenic Nd and Hf (eNdfrac14thorn193 eHffrac14thorn2516) Garnetin pyroxenite YK1915 also has radiogenic Nd and Hf(eNdfrac14thorn27 eHffrac14thorn42)Despite the large uncertainties for reconstructed whole-

rocks (calculated as the square root of the sum of squareduncertainties of two standard errors for garnet and cpxanalyses combined with 20 propagated uncertainty onthe proportion of Nd or Hf respectively contributed byeach mineral) the isotopic data are consistent in thatthey correlate with the independently determinedparent^daughter ratios (1) Low-Mg eclogites have higherSmNd reflected in the steeper positive slope of theirN-MORB-normalized REE patterns (Fig 11a) comparedwith the high-Mg eclogites (Fig 11b) Accordingly recon-structed low-Mg eclogites have higher present-day eNd

(0 to thorn47) than high-Mg eclogites (4 and 17) (2) Low-Mg eclogite YK1943 has very high Nb and LREE abun-dances that could have been added during metasomatismthat also enhanced its Hf abundance This sample has thelowest eNd and eHf of the low-Mg eclogites (0 and thorn20respectively) In contrast apparently unenriched low-Mgeclogites have more radiogenic Nd and Hf (eNdfrac14thorn47and thorn9 eHffrac14thorn128 and thorn113 respectively)The Sr^Nd isotope diagram reveals that eNd in cpx

varies from 19 to 43 at nearly constant depleted-mantle-like 87Sr86Sr (Fig 12a) In the Nd^Hf isotope diagram(Fig 12b) the reconstructed whole-rocks lie on curvedmixing lines if the lsquomixingrsquo proportions of Nd and Hffor the two lsquoend-memberrsquo minerals are very different(eg YK1911) Garnets have more radiogenic Nd andHf than coexisting cpx All reconstructed volatile-freeeclogites plot in the radiogenic field with regard to Hfor are isotopically similar to the depleted mantle Twohigh-Mg eclogites and one low-Mg eclogite have unradio-genic Nd one high-Mg eclogite has radiogenic Nd and

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1862

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 5 Individual and averaged Sr Nd and Hf isotope ratios obtained from garnet and cpx solutions

Sample 87Sr86Sr 2SE2SD 143Nd144Nd 2SE2SD eNd 2SE2SD 176Hf177Hf 2SE2SD eHf 2SE2SD

Low-Mg eclogites

VR43469 cp 0702661 0000010 0514844 0000006 43 01 0286130 0000012 119 04

VR43469 gt 0515508 0000014 56 03 0286506 0000012 132 04

VR43469 gt r 0515453 0000010 55 02 0286548 0000011 134 04

VR43469 gt r 0515384 0000007 54 01 0287244 0000010 158 04

VR43469 gt r 0515487 0000014 56 03 0286996 0000014 149 05

VR43469avg gt 0515458 0000109 55 21 0286823 0000715 143 253

WR VR43469 0515059 0000273 47 5 0286405 0001079 128 38

YK1926 cp 0703546 0000011 0513043 0000007 8 0 0283154 0000009 13 03

YK1926 gt 0513364 0000008 14 01 0295718 0000005 458 02

WR YK1926 0513120 0000026 9 1 0285980 0000716 113 25

YK1943 cp 0512143 0000007 10 01 0282213 0000005 20 02

YK1943 gt 0514461 0000012 36 02 0287510 0000008 168 03

YK1943 gt r 0514491 0000009 36 02 0287332 0000005 161 02

YK1943 gt r 0514517 0000010 37 02 0287290 0000007 160 02

YK1943avg gt 0514489 0000056 36 11 0287377 0000233 163 82

WR YK1943 0512625 0000228 0 5 0283138 0000879 13 31

High-Mg eclogites

VR43480 gt 0520965 0000017 162 03 0284950 0000010 77 04

VR67360 cp 0283509 0000074 26 26

VR67360 cp r 0283433 0000031 23 11

VR67360avg cp 0701662 0000018 0511660 0000010 19 02 0283471 0000108 25 38

VR67360 gt 0512984 0000101 7 20 0287429 0000180 165 64

WR VR67360 0511761 0000132 17 3 0284734 0000469 69 17

YK1949 cp 0703019 0000013 0512231 0000009 8 02 0282513 0000009 9 03

YK1949 gt 0513511 0000010 17 02 0290297 0000042 266 15

YK1949 gt r 0513508 0000012 17 02 0289947 0000015 254 05

YK1949 gt r 0513552 0000014 18 03

YK1949avg gt 0513523 0000049 17 10 0291569 0000494 311 175

WR YK1949 0512446 0000103 4 2 0284622 0000847 65 30

Volatile-rich eclogites

VR67112b gt 0522790 0000008 198 02 0354367 0000036 2532 13

YK1911 cp 0751186 0000009 0512165 0000007 9 01 0281263 0000009 53 03

YK1911 gt 0525721 0000013 255 03 0288051 0000012 187 04

WR YK1911 0513940 0000342 25 7 0281414 0000018 48 1

Pyroxenites

YK1914 cp 0703015 0000034 0513283 0000041 13 08

YK1914 cp r 0702956 0000005 0513200 0000027 11 05

YK1914 cp r 0513527 0000059 17 11

YK1914 cp r 0513045 0000010 8 02

YK1914avg cp 0702986 0000083 0513264 0000402 12 78

YK1915 gt 0514003 0000011 27 02 0283958 0000021 42

Analyses of full procedural replicates are designated with an lsquorrsquo after sample name SE standard error (internal errorsingle analyses) SD standard deviation (external error averaged multiple analyses) Italics used for reconstructed whole-rocks (mineral isotopic compositions and concentrations weighted by modes) Uncertainties for reconstructed whole-rocksinclude 2SE or 2SD uncertainties on each mineralrsquos isotope composition combined with 20 uncertainty in the amountof Nd or Hf contributed by the constituent minerals eNd and eHf are per 10 000 deviation from model chondritic mantle[model of Wasserburg et al (1981) and Blichert-Toft amp Albarede (1997) respectively]

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1863

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

εHf

εNd

Ekati eclogitesminus100

0

100

200

300

400

500

minus40 0 40 80 120 160 200 240 280

gt in YK1915

gt in VR67112bεNd=198 εHf=2532

minus40

0

40

80

120

160

minus40 minus20 0 20 40 60

minus60

minus40

minus20

0

20

40

070 071 072 073 074 075 076

North Americankimberlite field

εNd

87Sr86Sr

cpx in YK1914

Ekatieclogites

DM

DM

DM

DM

(a)

(b)

pyroxenites

low-Mg

high-Mg

volatile-rich

eclogites

Diavik low-TA154 S kimberlite

cpx in YK1911

Fig 12 (a) Sr^Nd for cpx and (b) Nd-Hf isotope diagram for garnet cpx and reconstructed whole-rocks (symbols with bold outline) Finedashed lines in (b) connect data points from the same sample with garnets having more radiogenic isotopic compositions Uncertainties forreconstructed whole-rocks combine 20 modal uncertainty with uncertainties on each mineralrsquos isotope composition Error bars on individualminerals may be smaller than the symbols Also shown are the isotopic compositions of the depleted mantle (DM DePaolo 1981 Griffin et al2000 Bell amp Tilton 2002) Range of 87Sr86Sr in eclogites from Ekati (Lac de Gras area 100 km north of Diavik Jacob et al 2003) in NorthAmerican kimberlites and compositions of low-temperature eclogites from Diavik and of Lac de Gras A154 South kimberlite fromSchmidberger et al (2007)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1864

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

one low-Mg eclogite has eNd similar to the depleted mantle(Fig 12b)Clinopyroxene in the phlogopitized eclogite YK1911

has unradiogenic Nd and Hf and strongly radiogenicSr whereas the coexisting garnet has radiogenic Nd andHf Although this sample contains 50 vol phlogopiteits whole-rock isotopic composition was calculated

from garnet and cpx alone because the contributionfrom phlogopite to the Hf and Nd budget is 52 and51 respectively (unpublished data) The whole-rockhas radiogenic Nd but has lower eHf than anyother sample analysed in accord with the extreme Hfabundance (53 ppm) and low LuHf of the cpx

DISCUSS IONOrigins of eclogites beneath the SlaveCratonHigh-Mg and high-Ca eclogites

With regard to their major- and trace-element contentsthe high-Mg and high-Ca eclogites from Lac de Gras arebroadly similar to variably evolved oceanic gabbros(Fig 13b and c) which may indicate an origin as subductedoceanic crust Compatible trace-element abundancesare useful to deduce potential crustal protoliths as theseelements are less susceptible to post-formation modification(Jacob 2004) The flat HREE patterns in the high-Mg andhigh-Ca eclogites are compatible with a low-pressureorigin as the oceanic crust is a product of moderate degreesof partial melting mainly within the spinel stability field(eg Presnall et al 2002) whereas melting with residualgarnet would qualitatively be expected to produce stronglyfractionated REE patterns Positive Sr and Pb anomaliesalthough compatible with a plagioclase signature couldalso be due to cryptic metasomatism (Dawson 1984) ofthe eclogites during their residence in the lithosphericmantle These anomalies are probably not related to kim-berlite infiltration as the whole-rock eclogite compositionswere reconstructed from analyses of fresh mineral coresLow ZrSm similar to that in primitive gabbros is

observed for many of the high-Mg and high-Caeclogites (Fig 13b and c) and is independent of the ratioof garnetcpx and whether or not the whole-rock wasreconstructed with rutile (Electronic Appendix 10) Thenegative trend of Mg-number against Yb in garnet(Fig 14) argues against igneous processes involving garnetaccumulation (Taylor amp Neal 1989) as suggested forexample for some high-Mg eclogites from West Africa(Barth et al 2002b) Rather the anti-correlation suggeststhat olivine fractionation played a role in the formation ofthe protoliths Fractionating olivine would excludeYb (Dolivinebasalt

frac14 002 Nikogosian amp Sobolev 1997)and incorporate Mg relative to Fe (Dfrac1415 and 05 respec-tively Taura et al1998) leading to progressive depletion ofMg and enrichment of Yb in the residual melt and henceanti-correlated Mg and Yb in the whole-rock This isinherited after eclogitization All eclogite types (exceptVR43452) plot along the olivine fractionation trend sug-gesting that they formed by a similar processHigh-Mg eclogites have lower

PREE than high-Ca

eclogites and some appear to have positive Eu anomalies

high-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

gabbrosSE Indian ridge

gabbrosOman ophiolite

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

high-Mg

RockNMORB(a)

VR43452

YK1949

low-Mg

boninite-like melts

01

001

1

10

(b)

01

001

1

10

(c)

01

001

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

Fig 13 Trace-element patterns of (a) low-Mg (b) high-Mg and(c) high-Ca eclogites compared with boninite-like volcanic rocksfrom the Isua greenstone belt (Polat et al 2002) gabbros from theOman ophiolite (Benoit et al 1996) and the SE Indian Ridge (Hartet al 1999) normalized to N-MORB of Sun amp McDonough (1989)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1865

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(EuEu up to 14) similar to more primitive cumulates(troctolite gabbros and olivine gabbros) reported in the lit-erature (Fig 13b) Although the sizes of these Eu anomaliesare not outside the analytical uncertainty their correlationwith CeYb in reconstructed eclogites supports their signif-icance (r2frac14066 nfrac14 4) The more primitive nature ofhigh-Mg eclogites is consistent with higher Mg-numbersrelative to high-Ca eclogites and may be due to olivineaccumulation during formation of their protolithsThe two samples with the lowest HREE abundanceshave the highest EuEu which is compatible with varia-tions in plagioclase modesMore evolved protoliths with higher pyroxeneplagio-

clase plus trapped melt could account for small or absentpositive Eu anomalies and higher

PHREE of high-Ca

eclogites (Fig 13c) Although the reason for the frequentassociation of graphite or diamond with high-Ca eclogitesis unclear (Aulbach et al 2003) the frequent occurrence ofkyanite in high-Ca eclogites also observed at Ekati justnorth of Lac de Gras has been argued to support a crustalprotolith (Jacob 2004) Kyanite andor corundum havebeen suggested to exsolve from high-temperature high-Alclinopyroxenes (Caporuscio amp Smyth 1990) Eclogitesfrom Lac de Gras show no exsolution textures suggestiveof such an origin although recrystallization can obscureor destroy microstructural evidence for exsolution(eg Griffin et al 1984)Modelling shows that qualitatively plagioclase accumu-

lation from a low-pressure melt derived from MORBmantle can account for positive Pb Sr and Eu and negativeZr anomalies relative to N-MORB whereas cpx accumula-tion or trapped melt increases Zr Y MREE and HREE

abundances (Fig 15) Some melt addition (for exampleby entrapment) seems to be required to increase all trace-element abundances to those observed for the eclogitesAlthough high-Ca and high-Mg eclogites could representshallow and deep portions of the same oceanic crust thehigher average temperatures obtained for the high-Caeclogites the frequent occurrence of carbon in high-Caeclogites and its absence in high-Mg eclogites the strongLREEHREE fractionation observed for six of seven high-Ca eclogites and the narrow range of compositions of thehigh-Mg eclogites suggest that they may have protolithsthat are not cogenetic Some discrepancies in particularaffecting the incompatible elements may be due to post-crystallization processes acting upon the protoliths suchas dehydration melting and later metasomatism that isevident in peridotites from the same locality (Griffin et al1999)Low K2OTh for high-Ca and high-Mg eclogites

(average 08 Table 4) relative to assumed crustal

0

2

4

6

8

10

30 40 50 60 70 80

Mg

VR43452

Yb

GARNET

pyroxenites

low-Mghigh-Mghigh-Ca

eclogites

VR19673ecl-g2

Fig 14 Garnet Mg-number vs Yb (ppm including 2s uncertainty)Lines qualitatively show separate correlations for eclogites (exceptoutlierVR43452) and pyroxenites (except outlierVR19673ecl-g2)

high-Ca

high-Mg

001

01

1

10

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

melt08 melt02 cryst cpx01 melt09 cryst

plag

(a)

001

01

1

10(b)

RockNMORB

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 15 Reconstructed (a) high-Ca and (b) high-Mg eclogitescompared with compositions of plagioclase and cpx crystallizingfrom a MORB melt modelled using distribution coefficients ofNorman et al (2005)The MORB was calculated as a 10 equilibriummelt of MORB pyrolite [composition from Kelemen et al (1993)except Pb Pr Sr Y and Lu which were calculated as residues from1 fractional melting of pyrolite from McDonough amp Sun (1995)using melt modes of Kinzler (1997) and distribution coefficients ofNorman et al (2005)] Also shown are mixtures of crystals (consistingof 50 plagioclase plus 50 cpx) and melt in the proportions 0901and 0208

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1866

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

precursors (eg gabbros SE Indian Ridge average 22Hart et al 1999) may reflect the preferred loss of fluid-mobile elements during dehydration (Becker et al 2000)although large uncertainties on these low-abundance ele-ments in the constituent minerals and reconstructedwhole-rocks makes this interpretation less than robustThe REE patterns of most high-Mg eclogites are too flatto represent melting residues as partial melting shouldproduce low LREEHREE (Electronic Appendix 10) Incontrast the marked LREE depletion of five of sevenhigh-Ca eclogites (LaLufrac14 02^08) contrasts with that ofassumed evolved gabbroic protoliths (LaLufrac1415^32Hart et al 1999) and can be modelled as a consequence ofpartial melting during subduction after eclogitization (iein the presence of garnet which retains HREE relative toLREE eg Barth et al 2002a) has taken place (Aulbachet al 2003)Our results are compatible with previous studies on

eclogite xenoliths from the Slave Craton includingthose from the same kimberlite that have linked thesesamples or some of these samples to Paleoproterozoicsubduction events (the 195^191Ga formation of theHottah terrane and the 18Ga formation of the GreatBear magmatic arc) in the Slave craton (Griffin et al1999 Kopylova et al 1999 Heaman et al 2002 2006Schmidberger et al 2005 2007)The compositional features discussed above although

compatible with crustal protoliths are not necessarilydefinitive of a subduction origin However if direct crystal-lization from basaltic melts at high pressure (ie mantledepths) is invoked instead specific conditions would haveto be met to satisfy the observations The Mg-number ineclogites from Lac de Gras are lower than those expectedof primary high-pressure mantle melts (eg Walter 1998OrsquoHara amp Herzberg 2002) requiring crystallization athigh pressures after olivine fractionation This originwould further require shielding from peridotitic wallrockwith which the fractionated melts would no longer be inequilibrium (Barth et al 2001) In such a setting the lowrelative abundances of fluid-mobile elements comparedwith those of less fluid-mobile elements and of LREErelative to HREE (Table 4) would reflect an expulsion offluids during crystallization of the protoliths and latermelt mobilization respectively The eclogites clearly donot represent garnet plus cpx cumulates which wouldrequire a positive correlation of Mg-number with Yb thatis at variance with the observed negative correlation ineclogitic garnets from this study (Fig 14)

Low-Mg eclogites

Garnets in low-Mg eclogites compositionally overlap withthose in massive eclogites from Jericho (NW of the Lacde Gras kimberlites) that were suggested to have a high-pressure intrusive origin based on the similarity of theirmajor-element compositions to those of Group A eclogites

fromsouthernAfrica forwhichamantleoriginhasbeenpro-posed (Kopylova et al1999) Garnets in low-Mg eclogites inthe present study have major-element compositions moresimilar to garnets in groupB rather thangroupAeclogitesThe positive slope in N-MORB-normalized HREE

abundances displayed by the low-Mg eclogites is clearlydistinct from other eclogite types and is similar to those ofboninite-like second-stage melts that are closely associatedwith ultramafic volcanics (Smithies et al 2004) (Fig 13a)The range of Mg-numbers in the low-Mg eclogites (55^66)overlaps with that of Archaean second-stage melts fromIsua (58^75) and Abitibi (52^82 Smithies et al 2004)These similarities might indicate that the low-Mg eclogiteshave an origin in a previously depleted mantle source suchas supra-subduction zone mantle Subchondritic ZrHf ofthe low-Mg eclogites contrasts with the supra-chondriticZrHf of the high-Mg and high-Ca eclogites (Aulbach et

al in preparation) and is consistent with previous cpx-controlled depletion in an arc mantle source This sourcecould have been subsequently remelted to generate theboninite-like precursor to the low-Mg eclogites as sug-gested for the protoliths of eclogites from Udachnaya(Jacob amp Foley 1999)Although some second-stage melts to which the low-Mg

eclogites are similar are closely associated with komatiitesand could contain a plume component the radiogenic ordepleted mantle-like Nd and Hf and unradiogenic Sr inthe low-Mg eclogites do not support plume involvement

Trace-element and Sr^Nd^Hf isotopeconstraints on post-formationmetasomatismSeveral low-Mg and high-Ca eclogites show negativeslopes in the LREE which may be indicative of interactionwith small amounts of LREE-enriched melt (Navon ampStolper 1987) amounting to cryptic metasomatism(ie not accompanied by crystallization of new phases) inthe nomenclature of Dawson (1984) High Nb abundancesrelative to elements of similar compatibility displayed bysome low-Mg and high-Ca eclogites (Fig 10) could resultfrom overestimation of rutile modes However high-CaeclogiteVR40345 is rutile-free and has been reconstructedwithout rutile its high Nb abundance is accompanied byLREE enrichment indicating that metasomatic additionmay be responsibleLow-Mg eclogites show a positive correlation of La with

Nb (Electronic Appendix 11a) indicating concomitantaddition of LREE and Nb This suggests that the metaso-matic agent was a solute-rich fluid or silicate silicic melt ashydrous fluids and carbonatites do not carry enough HFSEto affect the budget of eclogites (Green amp Wallace 1988Klemme et al 1995 Adam et al 1997 Stalder et al 1998)Of the samples that are enriched in Nb all but one (high-Ca eclogiteVR50909) show a simultaneous enrichment in

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1867

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Hf (Electronic Appendix 11b) Together these trace-element relationships suggest that both SmNd and LuHfwere lowered in these samples Their low parent^daughterratios contrast with higher ratios in unenriched samplesand resulted in the ingrowth of a range of Nd and Hf iso-tope compositions Compared with the low-Mg and high-Ca eclogites the high-Mg eclogites have a narrow range ofLa and Nb concentrations which correlates with a narrowrange in Nd and Hf isotope compositions (Fig 12)Precise model ages cannot be calculated because parent^

daughter ratios are available only from in situ analyseswhich have relatively large uncertainties Howeverevidence for long-term enrichment and depletion may bediscerned Low-Mg eclogites even those that are enrichedin LREE have relatively high SmNd as a result of havingoverall positive slopes in their REE patterns relative toMORB whereas the high-Mg eclogites have relativelyflat REE patterns and therefore lower SmNd (Fig 11)This translates into lower eNd for high-Mg eclogites thanfor low-Mg eclogites In contrast they show no systematicdifference with respect to 176Hf177Hf Because the low- andhigh-Mg eclogites are proposed to have different origins asdiscussed above and therefore did not necessarily have thesame initial isotopic composition and formation age theseisotopic differences cannot be used to estimate their agesA long-term coupling of trace-element and isotopic

abundances is also indicated by the fact that low-Mg eclo-gite YK1943 which has experienced Nb and LREE addi-tion has lower eNd (0) and eHf (thorn13) than low-Mg eclogiteVR43469 with higher LuHf and SmNd (eNdfrac14thorn47eHffrac14thorn128) If low-Mg eclogites YK1943 and VR43469originally had identical initial Nd and Hf isotope composi-tions and if the parent^daughter ratios are accurate andwere modified only once during their evolution shortlyafter metamorphism (when their isotopic compositionwould have been reset) this overprint can be calculatedfor both isotopic systems to have occurred at 17GaThis would suggest that Hf and Nd were added duringthe same event In contrast high-Mg eclogites VR67360and YK1949 would have had identical Ndisotope compositions at 10Ga whereas their Hf isotopecompositions are disturbed in that the sample with thehigher parent^daughter ratio has a lower 176Hf177Hf thanthe sample with the lower parent^daughter ratio Theseresults suggest that the LREE and HFSE enrichmentobserved in different samples occurred during separateevents that did not affect the eclogites equallyExtreme HFSE enrichment leading to the growth of

zircon and rutile and LREE addition leading to thegrowth of apatite in eclogites from the Jericho kimberlitein the northern Slave craton have been ascribed to twoseparate metasomatic events at around 18Ga (quasi-coincident with the age of formation and metamorphism)and 13^10Ga respectively (Heaman et al 2002 2006

Schmidberger et al 2005) The authors associated theseevents with Paleoproterozoic subduction and aMesoproterozoic thermal disturbance in the Slave litho-spheric mantle possibly linked to the Mackenzie igneousevent Mesoproterozoic Nd model ages mostly between13 and 14Ga were reported for eclogites from thenearby Ekati kimberlites (Jacob et al 2003) Although wedo not observe zircon or apatite in volatile-free eclogitesfrom Lac de Gras and the age constraints obtained fromour data are weak they support a link between theseevents described for other eclogite populations and themetasomatic enrichment observed in our studyThe Sr isotope data in the present study are similar to

those of Schmidberger et al (2007) who suggested that thedepleted-mantle-like values are consistent with derivationof the protoliths from an oceanic mantle source However87Sr86Sr in volatile-free eclogites in the present study isremarkably constant considering their Nd and Hf isotopevariability (Fig 12) and the variability of trace-elementpatterns (Fig 10) Given the evidence for dehydration andmelting in many of the eclogites fractionation of highlyincompatible Rb from Sr and differential ingrowth ofradiogenic Sr would be expectedThis may indicate a rela-tively recent overprint and considering the similarityof 87Sr86Sr to present-day depleted mantle which is signif-icantly lower than 87Sr86Sr of North American kimber-lites a juvenile (asthenospheric) source

PyroxenitesPyroxenites from Lac de Gras often show relationshipsin their major-element contents that are different fromthose observed in the eclogites For example pyroxeniticgarnets show a negative correlation between CaO andNa2O contents whereas these oxides are positively corre-lated for garnets in eclogites (Fig 5) CaO is well correlatedwith C2O3 in pyroxenitic garnets but not in eclogitic gar-nets (Fig 4) Mg-number is positively correlated withYb abundance for pyroxenitic garnets but anti-correlatedfor eclogitic garnets (Fig 14) This suggests that the pyrox-enites in the present study are not related to the eclogitesby fractionation or accumulation processes as proposedfor the origin of some pyroxenites (Foley et al 2003)Pyroxenites have also been interpreted as the products

of crystal segregation in magma channels with alteredoceanic crust as the magma source this would explaintheir Eu anomalies and stable and radiogenic isotopecompositions (Pearson et al 1993 Becker 1996) As thehigh-pressure liquidus phases of most terrestrial basaltsare cpx and garnet (OrsquoHara amp Herzberg 2002) partialmelts of eclogites with basaltic composition wouldcrystallize cpx and garnet but not opx More than half ofthe pyroxenites from Lac de Gras contain opx suggestingthat they probably did not crystallize directly fromeclogite-derived melts

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1868

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Low-Cr websterites and pyroxenites could be crystalli-zation products of mafic melts that interacted with perido-titic mantle thus becoming more refractory The positivecorrelation of Yb and Mg-number in garnet would allowan origin by fractional crystallization from such meltsAlternatively they formed by reaction of silicic (possiblyeclogite-derived) melts with mantle peridotites(eg Kelemen et al 1993 Aulbach et al 2002) This agreeswith the similarity of hypothetical melts in equilibriumwith pyroxenitic garnets in most samples to eclogite-derived melts (Electronic Appendix 12) Niobium concen-trations in all hypothetical melts are much higher than ineclogite-derived melts regardless of whether or not rutileis assumed to be in the residue The same is true for LaCe and Zr concentrations in some samples which mayindicate that NbZr and LREE were added to thepyroxenites some time after their formation as discussedin the previous section for some eclogites Because mini-mum ages for pyroxenitic rutile with an unradiogenicHf isotope composition (calculated by projecting the176Hf177Hf to the model depleted mantle isotopic evolutioncurve assuming zero ingrowth) range to 178066Ga(Aulbach et al in preparation) it appears likely that thepyroxenites were not formed recently (eg during kimber-lite-related or precursory melt migration) and that some ofthe samples are at least MesoproterozoicGarnet pyroxeniteYK1952 consists of schlieren of equili-

brated garnet pyroxenite in a lsquomatrixrsquo of spongy cpxfine cpx^opx intergrowths embayed garnet and meltpatches (Fig 2f) Garnet in the pyroxenite schlieren hasvery low LREE and HFSE abundances Orthopyroxenein the matrix has distinctly higher Mg-number andCr-number and contains more CaO than opx from thepyroxenite schlieren whereas opx rims in the schlierenare compositionally similar to opx in the matrix The con-siderable compositional and textural disequilibrium inthis sample suggests that this is a young feature causedby reaction with fluids shortly before or at the time ofentrainmentAs a consequence of the small sample sizes of most of

the pyroxenites studied here a comprehensive isotopicdataset is not available Clearly more such data areneeded in addition to stable isotope data to obtain betterconstraints on the origins of this rock type Consideringthe microstructural and compositional variability ofthe pyroxenites and major-element relationships oppositeto those observed for eclogites it appears certain that pyr-oxenites form by a variety of processes some of which areunrelated to the formation of eclogites and their protoliths

Volatile-rich eclogitesVolatile-rich eclogites from Lac de Gras display the mostextreme Hf^Nd isotope systematics of all the analysedsamples Garnets in phlogopitized (YK1911) and in cal-cite-bearing eclogite (VR67112b) have the most radiogenic

Nd (eNdfrac14 2486 and 1927 respectively) Garnet inVR67112b has extremely high eHf (2516) Clinopyroxene inYK1911 has the lowest eHf measured for any mineralseparate from Lac de Gras (53) in accord with the veryhigh Hf abundance in its cpx (5 ppm) In contrastgarnet in this sample has radiogenic Hf (eHffrac14 136) Thisdiscrepancy is at least in part due to their relatively lowequilibration temperatures (7698C and 8558C respec-tively) which allowed the isotopic compositions of cpxand garnet to evolve separately at their low andhigh parent^daughter element ratios respectively forextended periods of time and possibly since formationAccordingly the cpx in YK1911 gives a minimum Hf ageof 27GaSlab-derived fluids could have been introduced during

the 27Ga collision of the western Slave Craton domainincluding the thick plume-derived deep lithosphericmantle layer and the eastern domain with its highlydepleted lithosphere This would also be consistent withthe observation that volatile-rich eclogites occur at shallowdepths in the mantle (TKroghfrac14 685^8558C correspondingto depths of 85^115 km along a 40 mWm2 geotherm)where the bulk of dehydration and production of solute-rich melts in a subducting slab might be expected to occur(Becker et al 2000 Hermann et al 2006) Their negativeEu anomalies might indicate that the fluid was derivedfrom the shallow portion of the slab that experiencedplagioclase fractionation The presence of carbonate inVR67112b requires a carbonated sourceHigh 87Sr86Sr in cpx in YK1911 is probably the conse-

quence of isotopic equilibration with high-RbSr phlogo-pite after long-term radiogenic ingrowth while Hfaddition led to retarded ingrowth of radiogenic HfThese systematics combined with radiogenic Nd appear

inconsistent with the involvement of sediments in the pet-rogenesis of this sample (eg Perfit et al 1980 Vervoortet al 1999) or a hydrous fluid such as those responsible forMARID metasomatism in the Kaapvaal craton (Konzettet al 1998)Whatever the origin of the fluids the apparentHf mobility suggests that they were not in equilibriumwith a rutile-bearing source and also that they weresolute-rich in nature (ie hydrous melts able to dissolveHFSE) and the relatively high HREE abundances of thevolatile-rich eclogites indicate a garnet-free source suchas amphibolite This allowed evolution of the sample atrelatively high time-integrated SmNd

SUMMARY AND CONCLUSIONS

(1) The Lac de Gras kimberlites carried abundant eclo-gite and pyroxenite xenoliths to the surface Threemain eclogite types are distinguished in terms ofmajor-element composition and trace element

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1869

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

patterns high-Mg high-Ca and low-Mg eclogitesThe garnets of all eclogite types have anti-correlatedMg-number and Yb pointing to olivine fractionationas a controlling factor in the formation of their pre-cursor rocks Their flat or positively sloped HREEpatterns qualitatively indicate that the melt parentalto the precursor rocks was not in equilibrium with agarnet-bearing residue

(2) Reconstructed high-Mg eclogites and high-Caeclogites have flat HREE patterns positive Pb andSr anomalies and low ZrSm similar to oceanic gab-bros reported in the literature High-Mg eclogiteshave variably low HREE contents and some havesmall positive Eu anomalies possibly suggestive ofmore primitive relatively plagioclase-rich protolithswhereas high-Ca eclogites have higher HREE abun-dances and small or absent Eu anomalies that requiremore cpx- andor melt-rich protoliths if a subductionorigin is assumed

(3) Low-Mg eclogites have distinctive trace-element pat-terns that show a marked positive slope in the HREEand subchondritic ZrHf They resemble someArchaean second-stage melts (lsquoboninitesrsquo) and theirprotoliths appear to have formed from a previouslydepleted source

(4) After formation the eclogites were subject to a varietyof processes Most have low ratios of fluid-mobile vsless fluid-mobile elements possibly as a result of dehy-dration of the protoliths In addition some of the low-Mg and high-Ca eclogites have low LREEHREEconsistent with partial melt lossWhereas the low-Mgeclogites show concomitant enrichment in La Nb andZr some high-Mg and high-Ca eclogites show evi-dence for variable and unsystematic (decoupled)LREE and Nb addition

(5) Isotopically the trace-element characteristics of thedifferent eclogite types translate into lower eNd forhigh-Mg eclogites (negative slope in MREEfrac14 lowerSmNd) than for low-Mg eclogites (positive slope inMREEfrac14higher SmNd) Within the low-Mg groupsamples that show evidence for metasomatism (lowLuHf and SmNd) have lower eNd and eHf than asample that was apparently not metasomatized BothNd and Hf addition can be dated to 17Ga if identi-cal initial isotope compsitions are assumed for thelow-Mg eclogites In contrast the relatively constantand depleted mantle-like 87Sr86Sr of eclogitic cpxdespite variable Nd and Hf isotope compositions andparent^daughter ratios suggests recent input of juve-nile material

(6) Garnet and cpx in pyroxenites show major-elementrelationships different from those in eclogitesTherefore these two rock types are probably notrelated by fractionation processes The presence of

opx precludes direct crystallization from slab-derivedmelts Pyroxenites may be the product of (slab-derived) melts hybridized during assimilation of peri-dotitic mantle

(7) Garnets in phlogopite- and carbonate-bearing eclo-gites have extreme isotopic compositions includingthe most radiogenic Nd and Hf in the datasetNegative Eu anomalies may point to the involvementof a crustal protolith Highly radiogenic Sr radiogenicNd and unradiogenic Hf in the reconstructedphlogopite-bearing sample combined with highHREE abundances and evidence for Hf mobilitymay indicate involvement of solute-rich fluids origi-nating in a garnet- and rutile-free source

ACKNOWLEDGEMENTSHelp in the clean laboratories and with the analyticalfacilities by S Elhlou C Lawson A Sharma P Wielandand M Zhang is greatly appreciated Bill Griffin isthanked for his support and for agreeing to reviewseveral versions of the manuscript despite disagreeingwith many of the conclusions presented here The paperbenefited greatly from the detailed and constructivereviews provided by the editors Gareth Davies andMarjorie Wilson and by Matthias Barth Larry Heamanand an anonymous reviewer This work was funded by aMacquarie University International PostgraduateAward and Postgraduate Research Fund (SA) by anARC SPIRT grant sponsored by Kennecott Canada Incand by ARC Large and Discovery Grants to SYOrsquoRAnalytical data were obtained using instrumentationfunded by ARC LIEF and DEST Systemic InfrastructureGrants industry partners and Macquarie UniversityThis is Publication 481 from the ARC National KeyCentre for Geochemical Evolution and Metallogeny ofContinents (wwwesmqeduauGEMOC)

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online

REFERENCESAdam J GreenT H Sie S H amp Ryan C G (1997) Trace element

partitioning between aqueous fluids silicate melts and mineralsEuropeanJournal of Mineralogy 9 569^584

Aulbach S Stachel T Viljoen K S Brey G P amp Harris J W(2002) Eclogitic and websteritic diamond sources beneath theLimpopo Beltccedilis slab-melting the link Contributions to Mineralogy

and Petrology 143 56^70Aulbach S GriffinW L Pearson N J OrsquoReilly S Y Kivi K amp

Doyle B (2003) Origins of eclogites beneath the central SlaveCraton Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA-0011

Aulbach S Griffin W L Pearson N J OrsquoReilly S Y amp Kivi K(2004a) Mantle formation and evolution Slave Craton constraints

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1870

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

from HSE abundances and Re^Os isotope systematics of sulfideinclusions in mantle xenocrysts Chemical Geology 208 61^88

Aulbach S Griffin W L OrsquoReilly S Y amp McCandless T E(2004b) Genesis and evolution of the lithospheric mantle beneaththe Buffalo HeadTerrane Alberta (Canada) Lithos 77 413^451

Aulbach S GriffinW L Pearson N J OrsquoReilly S Y amp Doyle B(2007) Lithosphere formation in the central Slave Craton(Canada) plume subcretion or lithosphere accretion Contributionsto Mineralogy and Petrology (in press) doi 101007s00410-007-0200-1

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2001)Geochemistry of xenolithic eclogites from West Africa Part IA link between low MgO eclogites and Archean crust formationGeochimica et Cosmochimica Acta 65 1499^1527

Barth M G Foley S F amp Horn I (2002a) Partial melting inArchean subduction zones constraints from experimentally deter-mined trace element partition coefficients between eclogitic miner-als and tonalitic melts under upper mantle conditions PrecambrianResearch 113 323^340

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2002b)Geochemistry of xenolithic eclogites from West Africa part 2Origins of the high MgO eclogites Geochimica et Cosmochimica Acta

66 4325^4345Beard B L Fraracci K NTaylor L A Snyder G A Clayton R

A Mayeda T K amp Sobolev NV (1996) Petrography andgeochemistry of eclogites from the Mir kimberliteYakutia RussiaContributions to Mineralogy and Petrology 125 293^310

Becker H (1996) Geochemistry of garnet peridotite massifs fromlower Austria and the composition of deep lithosphere beneatha Palaeozoic convergent plate margin Chemical Geology 134 49^65

Becker H Jochum K P amp Carlson R W (2000) Trace elementfractionation during dehydration of eclogites from high-pressureterranes and the implications for element fluxes in subductionzones Chemical Geology 163 65^99

Bell K amp Tilton G R (2002) Probing the mantle The story fromcarbonatites EOS Transactions American Geophysical Union 83275^277

Benoit M Polvecurren M amp Ceuleneer G (1996) Trace element andisotopic characterisation of mafic cumulates in a fossil mantlediapir (Oman ophiolite) Chemical Geology 134 199^214

BleekerW (2003) The late Archean record a puzzle in ca 35 piecesLithos 71 99^134

BleekerW Ketchum JW F JacksonV A amp Villeneuve M (1999)The central Slave Basement Complex Part I Its structural topol-ogy and autochthonous core Canadian Journal of Earth Sciences 361083^1109

Blichert-Toft J amp Albare de F (1997) The Lu^Hf isotope geochemis-try of chondrites and the evolution of the mantle^crust systemEarth and Planetary Science Letters 148 243^258

Blichert-Toft J Chauvel C amp Albare de F (1997) Separation of Hfand Lu for high precision isotope analysis of rock samples by mag-netic sector-multiple collector ICP-MS Contributions to Mineralogy

and Petrology 127 248^260Brey G P amp Kolaquo hler T (1990) Geothermobarometry in four-phase

lherzolites II New thermobarometers and practical assessment ofexisting thermobarometers Journal of Petrology 31 1353^1378

Brey G P Kolaquo hlerT amp Nickel K G (1990) Geothermobarometry infour-phase lherzolites I Experimental results from 10 to 60 kbJournal of Petrology 31 1313^1352

Caporuscio F amp Smyth J (1990) Trace element crystal chemistryof mantle eclogites Contributions to Mineralogy and Petrology 105550^561

Carbno G B amp Canil D (2002) Mantle structure beneath the SWSlave Craton Canada constraints from garnet geochemistry in theDrybones Bay Kimberlite Journal of Petrology 43 129^142

Coleman R G Lee E D Beatty L B amp Brannock W W (1965)Eclogites and eclogites their differences and similaritiesGeological Society of America Bulletin 76 483^508

Creaser R A Grulaquo tter H Carlson J amp Crawford B (2004)Macrocrystal phlogopite Rb^Sr dates for the Ekati propertykimberlites Slave Province Canada evidence for multiple intru-sive episodes in the Paleocene and Eocene Lithos 76 399^414

Davies R M GriffinW L Pearson N J Andrew A S Doyle BJ amp OrsquoReilly S Y (1999) Diamonds from the deep Pipe DO-27Slave Craton Canada In Gurney J J Gurney J L Pascoe MD amp Richardson S H (eds) Proceedings of the 7th International

Kimberlite Conference CapeTown Red Roof Design pp 148^155Davies R M Griffin W L OrsquoReilly S Y amp Doyle B J (2004)

Mineral inclusions and geochemical characteristics of microdia-monds from the DO27 A154 A21 A418 DO18 DD17 and RanchLake kimberlites at Lac de Gras Slave Craton Canada Lithos 7739^55

DavisW J amp Hegner E (1992) Neodymium isotopic evidence for thetectonic assembly of Late Archean crust in the Slave Provincenorthwest Canada Contributions to Mineralogy and Petrology 111493^504

Davis W J Jones A G Bleeker W amp Grutter H (2003)Lithosphere development in the Slave craton a linked crustal andmantle perspective Lithos 71 575^589

Dawson J B (1984) Contrasting types of upper-mantle metasoma-tism In Kornprobst J (ed) Kimberlites II The Mantle and

Crust^Mantle Relationships Amsterdam Elsevier pp 289^294DePaolo D J (1981) Neodymium isotopes in the Colorado Front

Range and crust^mantle evolution in the Proterozoic Nature 291193^196

Erlank A J amp Kushiro I (1970) Potassium contents of syntheticpyroxenes at high temperatures and pressures Carnegie Institution ofWashingtonYearbook 68 433^439

Foley S F Barth M G amp Jenner G A (2000) Rutilemelt partitioncoefficients for trace elements and an assessment of the influence ofrutile on the trace element characteristics of subduction zonemagmas Geochimica et Cosmochimica Acta 64 933^938

Foley S F Buhre S amp Jacob D E (2003) Evolution of the Archaeancrust by delamination and shallow subduction Nature 421 249^252

Graham I Burgess J L Bryan D Ravenscroft P J Thomas EDoyle B J Hopkins R amp Armstrong K A (1999) Explorationhistory and geology of the Diavik Kimberlites Lac de GrasNorthwest Territories Canada In Gurney J J Gurney J LPascoe M D amp Richardson S H (eds) Proceedings of the 7th

International Kimberlite Conference Cape Town Red Roof Designpp 262^279

Green D H amp Wallace M E (1988) Mantle metasomatism byephemeral carbonatite melts Nature 336 459^462

Green T H Blundy J D Adam J amp Yaxley G M (2000) SIMSdetermination of trace element partition coefficients betweengarnet clinopyroxene and hydrous basaltic liquids at 2^75 GPaand 1080^12008C Lithos 53 165^187

GriffinW L amp OrsquoReilly SY (2007) Cratonic lithospheric mantle Isanything subducted Episodes 30(1) 43^53

Griffin W L Wass S Y amp Hollis J D (1984) Ultramafic xenolithsfrom Bullenmerri and Gnotuk maars Victoria Australia petrol-ogy of a sub-continental crust^mantle transition Journal of

Petrology 25 53^87GriffinW L Doyle B J Ryan C G Pearson N J OrsquoReilly SY

Davies R Kivi K van Achterbergh E amp Natapov LM (1999)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1871

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Layered mantle lithosphere in the Lac de Gras Area SlaveCraton composition structure and origin Journal of Petrology 40705^727

Griffin W L Pearson N J Belousova E Jackson S E vanAchterbergh E OrsquoReilly SY amp Shee S R (2000)The Hf isotopecomposition of cratonic mantle LAM-MC-ICPMS analysis ofzircon megacrysts in kimberlites Geochimica et Cosmochimica Acta

64 133^147Griffin W L OrsquoReilly S Y Doyle B J Pearson N J

Coopersmith H Kivi K Malkovets V amp Pokhilenko N(2004) Lithosphere mapping beneath the North American plateLithos 77 873^922

Grulaquo tter H S Apter D B amp Kong J (1999) Crust^mantle couplingevidence from mantle-derived xenocrystic garnets In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 307^312

Hart S R Blusztajn J Dick H J B Meyer P S ampMuehlenbachs K (1999) The fingerprint of seawater circulationin a 500-meter section of ocean crust gabbros Geochimica et

Cosmochimica Acta 63 4059^4080Harte B amp Kirkley M B (1997) Partitioning of trace elements

between clinopyroxene and garnet Data from mantle eclogitesChemical Geology 136 1^24

Heaman L M Creaser R A amp Cookenboo H O (2002) Extremeenrichment of high field strength elements in Jericho eclogitexenoliths A cryptic record of Paleoproterozoic subduction partialmelting and metasomatism beneath the Slave craton CanadaGeology 30 507^510

Heaman L M Kjarsgaard B A amp Creaser R A (2004) The tem-poral evolution of North American kimberlites Lithos 76 377^397

Heaman L M Creaser R A Cookenboo H O amp Chacko T(2006) Multi-stage modification of the northern Slave mantlelithosphere Evidence from zircon- and diamond-bearing eclogitexenoliths entrained in Jericho Kimberlite Journal of Petrology 47821^858

Hermann J Cpandler C Hack A amp Korsakov A (2006) Aqueousfluids and hydrous melts in high-pressure and ultra-high pressurerocks implications for element transfer in subduction zones Lithos92 399^417

Hills DV amp Haggerty S E (1989) Petrochemistry of eclogites fromthe Koidu Kimberlite Complex Sierra Leone Contributions to

Mineralogy and Petrology 103 397^422Hoffman P F (1989) Precambrian geology and tectonic history of

North America In Bally A amp Palmer A (eds) The Geology of

North AmericaccedilAn Overview Boulder CO Geological Society ofAmerica pp 447^512

IrelandT R Rudnick R L amp Spetsius Z (1994) Trace-elements indiamond inclusions from eclogites reveal link to Archean granitesEarth and Planetary Science Letters 128 199^213

Irvine G J Pearson D G Kjarsgaard B A Carlson R WKopylova M G amp Dreibus G (2003) Evolution of the litho-spheric mantle beneath Northern Canada a Re^Os isotope andPGE study of kimberlite-derived peridotite xenoliths fromSomerset Island and a comparison to the Slave and Kaapvaalcratons Lithos 71 461^488

Isachsen C E amp Bowring S A (1994) Evolution of the Slave cratonGeology 22 917^920

Jacob D E (2004) Nature and origin of eclogite xenoliths from kim-berlites Lithos 77 295^316

Jacob D E amp Foley S F (1999) Evidence for Archean ocean crustwith low high field strength element signature from diamondifer-ous eclogite xenoliths Lithos 48 317^336

Jacob D Jagoutz E Lowry D Mattey D amp Kudrjavtseva G(1994) Diamondiferous eclogites from Siberiaccedilremnants ofArchean oceanic-crust Geochimica et Cosmochimica Acta 585191^5207

Jacob D E Fung A Jagoutz E amp Pearson D G (2003) Petrologyand geochemistry of eclogite xenoliths from the Ekati kimberlitesarea Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA_239

Jacob D E Bizimis M amp Salters V J M (2005) Lu^Hf and geo-chemical systematics of recycled ancient oceanic crust evidencefrom Roberts Victor eclogites Contributions to Mineralogy and

Petrology 148 707^720Jerde E A Taylor L A Crozaz G Sobolev NV amp SobolevV N

(1993) Diamondiferous eclogites from Yakutia Siberiaccedilevidencefor a diversity of protoliths Contributions to Mineralogy and Petrology

114 189^202Jones A G Ferguson I J Chave A D Evans R L amp

McNeice GW (2001) The electric lithosphere of the Slave cratonGeology 29 423^426

Kelemen P B Shimizu N amp Dunn T (1993) Relative depletionof niobium in some arc magmas and the continental-crust-partitioning of K Nb La and Ce during meltrockreaction in the upper-mantle Earth and Planetary Science Letters 120111^134

Kinzler R J (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition Application to mid-ocean ridge basalt petrogenesis Journal of Geophysical Research 102853^874

Klemme S van der Laan S R Foley S F amp Guenther D (1995)Experimentally determined trace and minor element partitioningbetween clinopyroxene and carbonatite melt under upper mantleconditions Earth and Planetary Science Letters 133 439^448

Konzett J Armstrong R A Sweeney R J amp CompstonW (1998)The timing of MARID metasomatism in the Kaapvaal mantle anion probe study of zircons from MARID xenoliths Earth and

Planetary Science Letters 160 133^145Kopylova M G Russell J K amp Cookenboo H (1999) Mapping the

lithopshere beneath the north central Slave Craton In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 468^479

Krogh E (1988) The garnet^clinopyroxene iron^magnesiumgeothermometerccedila reinterpretation of existing experimentaldata Contributions to Mineralogy and Petrology 99 44^48

Kusky T M (1989) Accretion of Archean Slave Province Geology 1763^67

LeCheminant A N amp Heaman L M (1989) Mackenzie igneousevents Canada Middle Proterozoic hotspot magmatismassociated with ocean opening Earth and Planetary Science Letters

96 38^48LeCheminant A N Richardson D G DiLabio R N W amp

Richardson K A (1996) La recherche de diamants au CanadaGeological Survey of Canada Open File 3228 268pp

MacGregor I D amp Carter J L (1970) The chemistry of clinopyrox-enes and garnets of eclogite and peridotite xenoliths from theRoberts Victor mine South Africa Physics of the Earth and PlanetaryInteriors 3 391^397

MacGregor I D amp Manton W I (1986) Roberts Victor eclogitesancient oceanic crust Journal of Geophysical Research 91 14063^14079

MacKenzie J M amp Canil D (1999) Composition and thermal evo-lution of cratonic mantle beneath the central Archean SlaveProvince NWT Canada Contributions to Mineralogy and Petrology

134 313^324

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1872

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

McCandlessT E amp Gurney J J (1989) Sodium in garnet and potas-sum in clinopyroxene criteria for classifying mantle eclogites InRoss J (ed) Kimberlites and Related Rocks Carlton Blackwellpp 827^832

McDonoughW F amp Sun S-S (1995) The composition of the EarthChemical Geology 120 223^253

Navon O amp Stolper E (1987) Geochemical consequences of meltpercolation the upper mantle as a chromatographic columnJournal of Geology 95 285^307

Nikogosian I K amp Sobolev AV (1997) Ion-microprobe analysis ofmelt inclusions in olivine experience in estimating the olivine^meltpartition coefficients of trace elements Geochemistry International 35119^126

Norman M D Griffin W L Pearson N J Garcia M O ampOrsquoReilly S Y (1998) Quantitative analysis of trace element abun-dances in glasses and minerals a comparison of laser ablationinductively coupled plasma mass spectrometry solution inductivelycoupled plasma mass spectrometry proton microprobe and elec-tron microprobe data Journal of Analytical Atomic Spectrometry 13477^482

Norman M D Garcia M O amp Pietruszka A J (2005) Trace-element distribution coefficients for pyroxenes plagioclase andolivine in evolved tholeiites from the 1955 eruption of KilaueaVolcano Hawairsquoi and petrogenesis of differentiated rift-zonelavas American Mineralogist 90 888^899

OrsquoHara M J amp Herzberg C (2002) Interpretation of trace elementand isotope features of basalts Relevance of field relations petrol-ogy major element data phase equilibria and magma chambermodeling in basalt petrogenesis Geochimica et Cosmochimica Acta 662167^2191

PadghamW A amp FysonW K (1992) The Slave Province a distinctcraton CanadianJournal of Earth Sciences 29 2072^2086

Pearson D G Davies G R amp Nixon P H (1993) Geochemical con-straints on the petrogenesis of diamond facies pyroxenites from theBeni Bousera peridotite massif North Morocco Journal of Petrology34 125^172

Pearson N J Griffin W L Doyle B J OrsquoReilly S Y vanAchterbergh E amp Kivi K (1999) Xenoliths from kimberlitepipes of the Lac de gras area Slave Craton Canada In GurneyJ J Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite ConferenceCapeTown Red Roof Design pp 644^658

Perfit N R Gust D A Bence A E Arculus R J amp Taylor S R(1980) Chemical characteristics of island-arc basalts implicationsfor mantle sources Chemical Geology 30 227^256

Polat A Hofmann A W amp Rosing M T (2002) Boninite-likevolcanic rocks in the 37^38 Ga Isua greenstone belt WestGreenland geochemical evidence for intra-oceanic subductionzone processes in the early Earth Chemical Geology 184 231^254

Poudjom DjomaniY H GriffinW L OrsquoReilly SY amp Doyle B J(2005) Lithospheric domains and controls on kimberlite emplace-ment Slave Province Canada evidence from elastic thickness andupper mantle composition Geochemistry Geophysics Geosystems 6doi1010292005GC0009978

Presnall D C Gudfinnsson G H amp Walter M J (2002)Generation of mid-ocean ridge basalts at pressures from 1 to 7GPa Geochimica et Cosmochimica Acta 66 2073^2090

Schmidberger S S Heaman L M Simonetti A Creaser R A ampCookenboo H O (2005) Formation of Paleoproterozoiceclogitic mantle Slave Province (Canada) Insights from in-situHf and U^Pb isotopic analyses of mantle zircons Earth and

Planetary Science Letters 240 621^633

Schmidberger S S Simonetti A Heaman L M Creaser R A ampWhiteford S (2007) Lu^Hf in-situ Sr and Pb isotope and traceelement systematics for mantle eclogites from the Diavik diamondmine Evidence for Paleoproterozoic subduction beneath the Slavecraton Canada Earth and Planetary Science Letters 254 55^68

Schulze D J amp Helmstaedt H (1988) Coesite-sanidine eclogitesfrom kimberlite products of mantle fractionation or subductionJournal of Geology 96 435^443

Smithies R H Champion D C amp Sun S S (2004) The case forArchaean boninites Contributions to Mineralogy and Petrology 147705^721

Smyth J R Caporuscio F A amp McCormick T C (1989) Mantleeclogites evidence of igneous fractionation in the mantle Earth andPlanetary Science Letters 93 133^141

Sobolev NV amp Lavrentrsquoyev Y G (1971) Isomorphic sodium admix-ture in garnets formed at high pressure Contributions to Mineralogy

and Petrology 31 1^12Stalder R Foley S F Brey G P amp Horn I (1998) Mineral aqueous

fluid partitioning of trace elements at 900^12008C and 30^57GPa New experimental data for garnet clinopyroxene and rutileand implications for mantle metasomatism Geochimica et

Cosmochimica Acta 62 1781^1801Sun S-S amp McDonough W F (1989) Chemical and isotopic sys-

tematics of oceanic basalts implications for mantle compositionsand processes In Saunders A D amp Norris M J (eds)Magmatism in the Ocean Basins Geological Society London Special

Publications 42 313^345Taura H Yurimoto H Kurita K amp Sueno S (1998) Pressure

dependence on partition coefficients for trace elements between oli-vine and the coexisting melts Physics and Chemistry of Minerals 25469^484

Taylor L A amp Neal C R (1989) Eclogites with oceanic crustal andmantle signatures from the Bellsbank kimberlite South AfricaPart I Mineralogy petrography and whole rock chemistryJournal of Geology 97 551^567

van Achterbergh E Ryan C G amp Griffin W L (1999) Glitter online intensity reduction for the laser ablation inductively coupledplasma spectrometry Abstract 9th V M GoldschmidtConference LPI Boston MA 905 pp Abstract no 7215

van Breemen O DavisW J amp King J E (1992) Temporal distribu-tion of granitoid plutonic rocks in the Archean Slave ProvinceNorthwest Canadian Shield Canadian Journal of Earth Sciences 292186^2199

Vervoort J D Patchett P J Blichert-Toft J amp Albare de F (1999)Relationship between Lu^Hf and Sm^Nd isotopic systems in theglobal sedimentary system Earth and Planetary Science Letters 16879^99

Viljoen K S Smith C B amp Sharp Z D (1996) Stable and radio-genic isotope study of eclogite xenoliths from the Orapa kimberliteBotswana Chemical Geology 131 235^255

Walter M J (1998) Melting of garnet peridotite and theorigin of komatiite and depleted lithosphere Journal of Petrology 3929^60

Wasserburg G J Jacobsen S B DePaolo D J McCulloch M Tamp Wen T (1981) Precise determination of SmNd ratios Sm andNd isotopic abundances in standard solutions Geochimica et

Cosmochimica Acta 45 2311^2323Westerlund K J Shirey S B Richardson S H Carlson R W

Gurney J J amp Harris JW (2006) A subduction wedge origin forPaleoarchean peridotitic diamonds and harzburgites from thePanda kimberlite Slave craton evidence from Re^Os isotope sys-tematics Contributions to Mineralogy and Petrology 152 275^294

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1873

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Page 16: Origins of Xenolithic Eclogites and Pyroxenites from the Central

whole-rock compositions from mineral core compositionskimberlite infiltration is unlikely to have caused fractiona-tion of Ti from Eu and Dy In a study in which rutilemodes were estimated from the differences between Ti inmeasured bulk-rocks and in bulk-rocks reconstructed withonly garnet and cpx only three of 19 eclogites recon-structed with rutile show Ti anomalies (Barth et al2002a) Even eclogite xenoliths with depleted Zr and Hfabundances that resemble those of modern island arcbasalts do not show Ti anomalies (Jacob amp Foley 1999)We therefore consider our approach to be a reasonablyrobust means of obtaining meaningful rutile modes wheresmall sample sizes preclude reliable determination of themodes of accessory phases by point-countingWhere rutilewas not recognized on thin-section scale or not analysedthe average eclogitic rutile composition was used for thereconstruction

UncertaintiesReconstructing whole-rock compositions in this way is arather crude method given the coarse grain size of theeclogites relative to the sample size and the non-uniformdistribution of the principal minerals on thin-sectionscale The uncertainty in mineral modes of medium- tocoarse-grained eclogites with grains45mm has been esti-mated by Jerde et al (1993) to be 10 plus uncertaintiesrelated to partial alteration which is minor in our samplesNevertheless the whole-rock trace-element budget israther insensitive to the exact modes for common eclogiticassemblages and REE patterns which encompass mostother elements in terms of compatibility in garnet andcpx do not change significantly even for modal variationsof 30 (Jerde et al 1993) This is not true for major-element concentrations which vary substantially with var-iations in mineral modes and we therefore do not calculatewhole-rock major-element contentsBecause the discussion that follows is based on trace-

element patterns and trends rather than absolute abun-dances whole-rock reconstruction based on mineralmodes is adequate We are confident that the uncertaintyinherent in this approach is not leading to significant mis-representation of eclogite whole-rock compositions becausethe trace-element patterns of the reconstructed whole-rocks show distinct differences (see below) that correlatewith the assignment to eclogite types based on garnetmajor-element composition and this is independent ofmineral mode Following Jerde et al (1993) in ElectronicAppendix 9 we show variations in trace-element patternfor one sample in each eclogite type for variations ingarnet and cpx abundance from 7030 to 3070 vol This shows that the trace-element patterns are generallyinsensitive to cpxgt ratios with regard to slopes andanomalies such as positive Sr Pb and Eu anomalies withthe exception of the positive Eu anomaly in high-Mgeclogite VR43479 which disappears if a high cpx mode is

assumed (not shown) Variations in rutile modes by 50(Electronic Appendix 9d) lead as would be expected tothe appearance of Ti anomalies but a rutile mode of 05instead of 1 vol does not erase the positive Nb anomalyin low-Mg eclogiteVR50909Because of the robustness of the relative trace-element

abundances in the reconstructed eclogites a garnet propor-tion of 065 and cpx proportion of 035 approximately themedian modal abundances of these minerals in volatile-free eclogites from Lac de Gras are assumed for eclogitesamples that are too small for determination of meaningfulmodes by point-counting

Within- and between-group variationsThe trace-element compositions of reconstructed whole-rocks are given in Table 4 N-MORB-normalized trace-element patterns of reconstructed whole-rocks areshown in Fig 10 The three eclogite groups have distincttrace element patterns that are particularly obvious forthe REE (Fig 11) Low-Mg eclogites have positive Pb andSr anomalies and smooth positive slopes in their MREEN

and HREEN LREEN are enriched relative to MREEN inthree samples (Fig 11a) Three samples have high Nbrelative to similarly compatible elements and comparedwith most other eclogites (Fig 10a) All high-Mg eclogitesdisplay more or less pronounced positive Eu and Sranomalies and most have positive Pb and negative Zrand P anomalies Their N-MORB-normalized REE pat-terns show shallow positive slopes from LaN to PrN orNdNand aweak negative slope from NdN to LuN (Fig11b)VR43479 is a low-Mg eclogite with regard to its garnet

Ca^Fe^Mg composition but has a trace-element patterntypical of high-Mg eclogites and will be discussed withthe latter Conversely sample YK1949 a high-Mg eclogitein terms of its garnet Ca^Fe^Mg compostion is more simi-lar to low-Mg eclogites in terms of its whole-rock trace-element pattern and is plotted in Figs 10a and 11aWhereas the patterns of most samples within the groupresemble each other the N-MORB-normalized incompati-ble-element abundances of the high-Ca eclogites vary bymore than an order of magnitude (Fig 10c) All sampleshave positive Pb anomalies and most have positive Sranomalies Some samples have negative P andor Zranomalies Niobium abundances in three of the samplesare very high similar to those for some low-Mg eclogitesHigh-Ca eclogites tend to have flatter MREEN andHREEN patterns and more uniform HREE abundancesthat the other eclogite types with humps between SmN

and HoNThe phlogopitized sample YK1911 shows a strong posi-

tive Pb anomaly and lower HREE abundances than thecalcite-bearing sample VR67112b which shows a distinctnegative Eu anomaly and a depletion of La relative to Ce(Figs 10d and 11d)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1858

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Trace-element compositions of reconstructed whole-rocks (ppm)

Sample P Sc Ti V Co Ni Ga Sr Y Zr Nb La Ce Pr Nd

Low-Mg eclogites

vr19674-g4 186 50 4649 317 57 59 144 77 24 30 36 13 31 05 24

vr43452 154 39 2253 328 49 84 176 52 32 11 01 02 07 01 08

vr43469 38 65 4377 349 54 91 142 34 28 15 02 05 15 03 17

yk1926 144 45 3406 258 52 87 145 89 18 18 20 09 29 06 33

yk1943 173 46 5732 295 41 39 146 122 25 48 82 34 85 13 64

yk19491 102 34 1693 158 31 49 100 46 10 14 03 09 21 03 17

High-Mg eclogites

vr19677-g2 138 39 5160 291 41 53 119 98 14 22 04 13 53 11 64

vr434792 46 25 1265 129 36 126 84 56 3 7 04 06 25 04 23

vr67360 49 36 1573 138 64 184 111 76 5 9 04 09 38 07 33

yk1946 153 28 4877 192 45 105 124 108 12 48 03 10 38 08 50

High-Ca eclogites

vr19674-g5 267 30 7663 200 41 67 195 32 23 43 77 01 04 02 16

vr40345 0 33 2821 175 47 98 210 179 13 13 66 60 109 11 38

vr43465 142 27 5104 220 31 83 121 197 14 35 29 05 22 05 40

vr50858 295 30 6898 201 44 94 150 48 22 38 06 01 03 02 20

vr50860 68 57 5672 418 57 71 186 109 19 11 02 02 13 03 26

vr50909 238 28 7295 201 42 92 171 48 20 57 13 01 04 02 21

yk3528 91 51 3712 341 53 155 177 49 16 9 01 02 08 02 11

Volatile-rich eclogites

vr67112b 40 84 9242 521 46 30 107 4 40 66 31 26 126 24 125

yk1911 31 39 9428 286 44 82 164 3 23 52 64 18 52 09 46

Pyroxenites

yk1914 111 49 707 207 48 547 44 31 4 3 02 06 19 03 10

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

Low-Mg eclogites

vr19674-g4 11 05 10 22 39 10 29 33 06 12 05 028 24 008

vr43452 04 02 08 13 38 12 41 49 08 05 01 nd 02 na

vr43469 11 05 09 25 44 11 34 36 06 07 02 001 09 129

yk1926 12 05 11 19 30 07 22 24 04 07 03 004 23 090

yk1943 21 08 10 29 42 10 30 32 05 16 05 013 65 017

yk19491 06 03 11 08 14 04 14 18 03 04 02 003 29 115

High-Mg eclogites

vr19677-g2 20 08 12 21 24 05 15 15 03 05 05 007 52 6811

vr434792 05 02 12 05 06 01 04 03 00 03 01 001 98 037

vr67360 06 03 14 06 09 02 06 06 01 02 02 002 86 000

yk1946 18 07 12 21 22 05 13 12 02 11 03 006 56 112

High-Ca eclogites

vr19674-g5 17 09 11 33 43 09 24 23 04 16 05 007 02 011

vr40345 09 04 10 12 22 06 16 14 02 05 nd nd 298 na

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1859

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Continued

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

vr43465 22 09 12 25 26 06 15 14 02 10 04 001 25 040

vr50858 20 08 09 33 41 08 22 21 03 10 01 nd 02 na

vr50860 14 07 11 27 37 08 22 18 03 06 00 nd 06 na

vr50909 20 10 12 33 38 07 20 20 03 16 01 002 02 181

yk3528 08 05 11 20 30 07 18 18 02 05 01 nd 08 na

Volatile-rich eclogites

vr67112b 38 10 07 53 74 15 41 34 04 18 03 012 60 004

yk1911 20 07 08 35 39 07 18 17 02 10 14 nd 73 na

Pyroxenites

yk1914 02 01 11 04 08 02 05 06 01 01 03 003 68 101

aGarnet has CandashMgndashFe relationship typical of high-Mg eclogites but whole-rock trace-element pattern is typical oflow-Mg eclogitesbGarnet has CandashMgndashFe relationship typical of low-Mg eclogites but whole-rock trace-element pattern is typical ofhigh-Mg eclogites

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

(a)

001

01

1

10(c)

001

01

10(b)

VR43452

YK1949

RockNMORB

low-Mg

RockNMORB

high-Mg

(d)

volatile-richhigh-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

phlogopite-bearingYK1911

calcite-bearingVR67112b

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 10 Extended trace-element patterns of reconstructed (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-bearing eclogites Normalized to N-MORB of Sun amp McDonough (1989)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1860

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

RockNMORB

(b)

RockNMORB

(d)

010

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu001

01

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

(a)

001

01

1

10(b)

001

01

1

10(c)

001

01

1

10(e)

high-TSchmidberger et al

low-TSchmidberger et al

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

high-Mglow-Mg

high-Cavolatile-rich

phlogopite-bearingYK1911

calcite-bearingVR67112b

YK1949VR43479

Fig 11 N-MORB normalized REE patterns for (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-richeclogites Shown for comparison in (e) are reconstructed eclogites from Schmidberger et al (2007)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1861

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Both volatile-rich whole rocks display strong negativeSr anomalies Strontium can be sited in calcite or apatiteandor indicate the presence of a Sr-rich mineral that wasnot identified in thin section such as bariteFour of the six reconstructed whole-rocks in the study

of Schmidberger et al (2007) (Fig 11e) show markedLREEHREEN enrichment that is observed only in onehigh-Ca eclogite in our study (Fig 11c) suggesting thatmany of the eclogites they studied were affected by metaso-matic LREE enrichment Two samples in their study thatare not affected by this enrichment show positive Euanomalies similar to the high-Mg eclogites in this studywhereas the distinctive REE pattern of the low-Mg eclo-gites is not recognized among their samples

SR^ND^HF I SOTOPESYSTEMATICSStrontium Nd and Hf isotope data for optically puremineral separates and reconstructed whole-rocks aregiven in Table 5 Spongy rims and alteration are likely tobe mostly removed during leaching in 6N HCl in combina-tion with repeated ultrasonication Despite this rigoroustreatment the reproducibility of Nd in garnet VR43469and in cpx YK1914 is worse than for other samples Thismay be due to inhomogeneities related to late Nd diffusionpossibly from grain boundaries containing a kimberlitecomponent The reproducibility of the 176Hf177Hf data isgenerally poorer than that of Nd which may reflect a com-bination of matrix effects caused by imperfectly purifiedsolutions and sample heterogeneity Parent^daughterratios are available only from in situ analyses which arenot precise enough to allow meaningful model age orinitial isotope ratio calculations (ie at the time of kimber-lite eruption) Because the Lac de Gras kimberlites arerelatively young (55^56 Ma Graham et al 1999) age cor-rection would result in only a minor change in isotopecomposition which is inconsequential in view of the rangeof compositions displayed by most samples and hence hasno effect on the conclusions drawnClinopyroxenes in five volatile-free eclogites and a

pyroxenite have Sr isotope ratios ranging from 0701662 to0703546 this range is similar to or somewhat lower thanthat of the depleted mantle at the time of kimberliteemplacement (0703 Bell amp Tilton 2002) Clinopyroxenein the phlogopitized sample has a distinctly high 87Sr86Srof 0751186Clinopyroxene controls 56^92 of the Nd budget (Nd

concentration in cpx weighted by mode) Neodymiumisotope ratios range from 0511660 to 0514844 Thiscorresponds to eNd variation (per 10 000 deviationof 143Nd144Nd in the sample from 143Nd144Nd in the chon-dritic mantle model reservoir) from 17 to thorn43 cpx inthe low-Mg eclogites have higher eNd than those in

high-Mg eclogites Garnets have generally have higher143Nd144Nd from 0512984 (eNdfrac14thorn 7) to 0520965(eNdfrac14thorn 162) Garnets in the two volatile-rich sampleshave the highest ratios 0522790 for garnet in the calcite-bearing sample VR67112b (eNdfrac14thorn198) and 0525721 inthe phlogopitized eclogiteYK1911 (eNdfrac14thorn 255)Garnet contributes 18^32 and cpx 48^78 of the Hf

budget The proportion controlled by rutile is 4^30(Aulbach et al in preparation) Measured 176Hf177Hf incpx ranges from 0281263 in phlogopite-bearing sampleYK1911 (eHffrac1453) to 0286130 in low-Mg eclogiteVR43469 (eHffrac14thorn119) and in garnet from 0283958 inpyroxeniteYK1915 (eHffrac14thorn42) to 0295718 in low-Mg eclo-gite YK1926 (eHffrac14thorn458) Hafnium in garnet in the phlo-gopite^apatite-bearing sample is extremely radiogenic(176Hf177Hffrac14 0354367) corresponding to an eHf of thorn2532which is close to a value determined for a garnet separatefrom a Roberts Victor eclogite (eHffrac14thorn2561 Jacob et al2005) Garnet in the calcite-bearing sample has extremelyradiogenic Nd and Hf (eNdfrac14thorn193 eHffrac14thorn2516) Garnetin pyroxenite YK1915 also has radiogenic Nd and Hf(eNdfrac14thorn27 eHffrac14thorn42)Despite the large uncertainties for reconstructed whole-

rocks (calculated as the square root of the sum of squareduncertainties of two standard errors for garnet and cpxanalyses combined with 20 propagated uncertainty onthe proportion of Nd or Hf respectively contributed byeach mineral) the isotopic data are consistent in thatthey correlate with the independently determinedparent^daughter ratios (1) Low-Mg eclogites have higherSmNd reflected in the steeper positive slope of theirN-MORB-normalized REE patterns (Fig 11a) comparedwith the high-Mg eclogites (Fig 11b) Accordingly recon-structed low-Mg eclogites have higher present-day eNd

(0 to thorn47) than high-Mg eclogites (4 and 17) (2) Low-Mg eclogite YK1943 has very high Nb and LREE abun-dances that could have been added during metasomatismthat also enhanced its Hf abundance This sample has thelowest eNd and eHf of the low-Mg eclogites (0 and thorn20respectively) In contrast apparently unenriched low-Mgeclogites have more radiogenic Nd and Hf (eNdfrac14thorn47and thorn9 eHffrac14thorn128 and thorn113 respectively)The Sr^Nd isotope diagram reveals that eNd in cpx

varies from 19 to 43 at nearly constant depleted-mantle-like 87Sr86Sr (Fig 12a) In the Nd^Hf isotope diagram(Fig 12b) the reconstructed whole-rocks lie on curvedmixing lines if the lsquomixingrsquo proportions of Nd and Hffor the two lsquoend-memberrsquo minerals are very different(eg YK1911) Garnets have more radiogenic Nd andHf than coexisting cpx All reconstructed volatile-freeeclogites plot in the radiogenic field with regard to Hfor are isotopically similar to the depleted mantle Twohigh-Mg eclogites and one low-Mg eclogite have unradio-genic Nd one high-Mg eclogite has radiogenic Nd and

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1862

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 5 Individual and averaged Sr Nd and Hf isotope ratios obtained from garnet and cpx solutions

Sample 87Sr86Sr 2SE2SD 143Nd144Nd 2SE2SD eNd 2SE2SD 176Hf177Hf 2SE2SD eHf 2SE2SD

Low-Mg eclogites

VR43469 cp 0702661 0000010 0514844 0000006 43 01 0286130 0000012 119 04

VR43469 gt 0515508 0000014 56 03 0286506 0000012 132 04

VR43469 gt r 0515453 0000010 55 02 0286548 0000011 134 04

VR43469 gt r 0515384 0000007 54 01 0287244 0000010 158 04

VR43469 gt r 0515487 0000014 56 03 0286996 0000014 149 05

VR43469avg gt 0515458 0000109 55 21 0286823 0000715 143 253

WR VR43469 0515059 0000273 47 5 0286405 0001079 128 38

YK1926 cp 0703546 0000011 0513043 0000007 8 0 0283154 0000009 13 03

YK1926 gt 0513364 0000008 14 01 0295718 0000005 458 02

WR YK1926 0513120 0000026 9 1 0285980 0000716 113 25

YK1943 cp 0512143 0000007 10 01 0282213 0000005 20 02

YK1943 gt 0514461 0000012 36 02 0287510 0000008 168 03

YK1943 gt r 0514491 0000009 36 02 0287332 0000005 161 02

YK1943 gt r 0514517 0000010 37 02 0287290 0000007 160 02

YK1943avg gt 0514489 0000056 36 11 0287377 0000233 163 82

WR YK1943 0512625 0000228 0 5 0283138 0000879 13 31

High-Mg eclogites

VR43480 gt 0520965 0000017 162 03 0284950 0000010 77 04

VR67360 cp 0283509 0000074 26 26

VR67360 cp r 0283433 0000031 23 11

VR67360avg cp 0701662 0000018 0511660 0000010 19 02 0283471 0000108 25 38

VR67360 gt 0512984 0000101 7 20 0287429 0000180 165 64

WR VR67360 0511761 0000132 17 3 0284734 0000469 69 17

YK1949 cp 0703019 0000013 0512231 0000009 8 02 0282513 0000009 9 03

YK1949 gt 0513511 0000010 17 02 0290297 0000042 266 15

YK1949 gt r 0513508 0000012 17 02 0289947 0000015 254 05

YK1949 gt r 0513552 0000014 18 03

YK1949avg gt 0513523 0000049 17 10 0291569 0000494 311 175

WR YK1949 0512446 0000103 4 2 0284622 0000847 65 30

Volatile-rich eclogites

VR67112b gt 0522790 0000008 198 02 0354367 0000036 2532 13

YK1911 cp 0751186 0000009 0512165 0000007 9 01 0281263 0000009 53 03

YK1911 gt 0525721 0000013 255 03 0288051 0000012 187 04

WR YK1911 0513940 0000342 25 7 0281414 0000018 48 1

Pyroxenites

YK1914 cp 0703015 0000034 0513283 0000041 13 08

YK1914 cp r 0702956 0000005 0513200 0000027 11 05

YK1914 cp r 0513527 0000059 17 11

YK1914 cp r 0513045 0000010 8 02

YK1914avg cp 0702986 0000083 0513264 0000402 12 78

YK1915 gt 0514003 0000011 27 02 0283958 0000021 42

Analyses of full procedural replicates are designated with an lsquorrsquo after sample name SE standard error (internal errorsingle analyses) SD standard deviation (external error averaged multiple analyses) Italics used for reconstructed whole-rocks (mineral isotopic compositions and concentrations weighted by modes) Uncertainties for reconstructed whole-rocksinclude 2SE or 2SD uncertainties on each mineralrsquos isotope composition combined with 20 uncertainty in the amountof Nd or Hf contributed by the constituent minerals eNd and eHf are per 10 000 deviation from model chondritic mantle[model of Wasserburg et al (1981) and Blichert-Toft amp Albarede (1997) respectively]

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1863

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

εHf

εNd

Ekati eclogitesminus100

0

100

200

300

400

500

minus40 0 40 80 120 160 200 240 280

gt in YK1915

gt in VR67112bεNd=198 εHf=2532

minus40

0

40

80

120

160

minus40 minus20 0 20 40 60

minus60

minus40

minus20

0

20

40

070 071 072 073 074 075 076

North Americankimberlite field

εNd

87Sr86Sr

cpx in YK1914

Ekatieclogites

DM

DM

DM

DM

(a)

(b)

pyroxenites

low-Mg

high-Mg

volatile-rich

eclogites

Diavik low-TA154 S kimberlite

cpx in YK1911

Fig 12 (a) Sr^Nd for cpx and (b) Nd-Hf isotope diagram for garnet cpx and reconstructed whole-rocks (symbols with bold outline) Finedashed lines in (b) connect data points from the same sample with garnets having more radiogenic isotopic compositions Uncertainties forreconstructed whole-rocks combine 20 modal uncertainty with uncertainties on each mineralrsquos isotope composition Error bars on individualminerals may be smaller than the symbols Also shown are the isotopic compositions of the depleted mantle (DM DePaolo 1981 Griffin et al2000 Bell amp Tilton 2002) Range of 87Sr86Sr in eclogites from Ekati (Lac de Gras area 100 km north of Diavik Jacob et al 2003) in NorthAmerican kimberlites and compositions of low-temperature eclogites from Diavik and of Lac de Gras A154 South kimberlite fromSchmidberger et al (2007)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1864

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

one low-Mg eclogite has eNd similar to the depleted mantle(Fig 12b)Clinopyroxene in the phlogopitized eclogite YK1911

has unradiogenic Nd and Hf and strongly radiogenicSr whereas the coexisting garnet has radiogenic Nd andHf Although this sample contains 50 vol phlogopiteits whole-rock isotopic composition was calculated

from garnet and cpx alone because the contributionfrom phlogopite to the Hf and Nd budget is 52 and51 respectively (unpublished data) The whole-rockhas radiogenic Nd but has lower eHf than anyother sample analysed in accord with the extreme Hfabundance (53 ppm) and low LuHf of the cpx

DISCUSS IONOrigins of eclogites beneath the SlaveCratonHigh-Mg and high-Ca eclogites

With regard to their major- and trace-element contentsthe high-Mg and high-Ca eclogites from Lac de Gras arebroadly similar to variably evolved oceanic gabbros(Fig 13b and c) which may indicate an origin as subductedoceanic crust Compatible trace-element abundancesare useful to deduce potential crustal protoliths as theseelements are less susceptible to post-formation modification(Jacob 2004) The flat HREE patterns in the high-Mg andhigh-Ca eclogites are compatible with a low-pressureorigin as the oceanic crust is a product of moderate degreesof partial melting mainly within the spinel stability field(eg Presnall et al 2002) whereas melting with residualgarnet would qualitatively be expected to produce stronglyfractionated REE patterns Positive Sr and Pb anomaliesalthough compatible with a plagioclase signature couldalso be due to cryptic metasomatism (Dawson 1984) ofthe eclogites during their residence in the lithosphericmantle These anomalies are probably not related to kim-berlite infiltration as the whole-rock eclogite compositionswere reconstructed from analyses of fresh mineral coresLow ZrSm similar to that in primitive gabbros is

observed for many of the high-Mg and high-Caeclogites (Fig 13b and c) and is independent of the ratioof garnetcpx and whether or not the whole-rock wasreconstructed with rutile (Electronic Appendix 10) Thenegative trend of Mg-number against Yb in garnet(Fig 14) argues against igneous processes involving garnetaccumulation (Taylor amp Neal 1989) as suggested forexample for some high-Mg eclogites from West Africa(Barth et al 2002b) Rather the anti-correlation suggeststhat olivine fractionation played a role in the formation ofthe protoliths Fractionating olivine would excludeYb (Dolivinebasalt

frac14 002 Nikogosian amp Sobolev 1997)and incorporate Mg relative to Fe (Dfrac1415 and 05 respec-tively Taura et al1998) leading to progressive depletion ofMg and enrichment of Yb in the residual melt and henceanti-correlated Mg and Yb in the whole-rock This isinherited after eclogitization All eclogite types (exceptVR43452) plot along the olivine fractionation trend sug-gesting that they formed by a similar processHigh-Mg eclogites have lower

PREE than high-Ca

eclogites and some appear to have positive Eu anomalies

high-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

gabbrosSE Indian ridge

gabbrosOman ophiolite

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

high-Mg

RockNMORB(a)

VR43452

YK1949

low-Mg

boninite-like melts

01

001

1

10

(b)

01

001

1

10

(c)

01

001

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

Fig 13 Trace-element patterns of (a) low-Mg (b) high-Mg and(c) high-Ca eclogites compared with boninite-like volcanic rocksfrom the Isua greenstone belt (Polat et al 2002) gabbros from theOman ophiolite (Benoit et al 1996) and the SE Indian Ridge (Hartet al 1999) normalized to N-MORB of Sun amp McDonough (1989)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1865

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(EuEu up to 14) similar to more primitive cumulates(troctolite gabbros and olivine gabbros) reported in the lit-erature (Fig 13b) Although the sizes of these Eu anomaliesare not outside the analytical uncertainty their correlationwith CeYb in reconstructed eclogites supports their signif-icance (r2frac14066 nfrac14 4) The more primitive nature ofhigh-Mg eclogites is consistent with higher Mg-numbersrelative to high-Ca eclogites and may be due to olivineaccumulation during formation of their protolithsThe two samples with the lowest HREE abundanceshave the highest EuEu which is compatible with varia-tions in plagioclase modesMore evolved protoliths with higher pyroxeneplagio-

clase plus trapped melt could account for small or absentpositive Eu anomalies and higher

PHREE of high-Ca

eclogites (Fig 13c) Although the reason for the frequentassociation of graphite or diamond with high-Ca eclogitesis unclear (Aulbach et al 2003) the frequent occurrence ofkyanite in high-Ca eclogites also observed at Ekati justnorth of Lac de Gras has been argued to support a crustalprotolith (Jacob 2004) Kyanite andor corundum havebeen suggested to exsolve from high-temperature high-Alclinopyroxenes (Caporuscio amp Smyth 1990) Eclogitesfrom Lac de Gras show no exsolution textures suggestiveof such an origin although recrystallization can obscureor destroy microstructural evidence for exsolution(eg Griffin et al 1984)Modelling shows that qualitatively plagioclase accumu-

lation from a low-pressure melt derived from MORBmantle can account for positive Pb Sr and Eu and negativeZr anomalies relative to N-MORB whereas cpx accumula-tion or trapped melt increases Zr Y MREE and HREE

abundances (Fig 15) Some melt addition (for exampleby entrapment) seems to be required to increase all trace-element abundances to those observed for the eclogitesAlthough high-Ca and high-Mg eclogites could representshallow and deep portions of the same oceanic crust thehigher average temperatures obtained for the high-Caeclogites the frequent occurrence of carbon in high-Caeclogites and its absence in high-Mg eclogites the strongLREEHREE fractionation observed for six of seven high-Ca eclogites and the narrow range of compositions of thehigh-Mg eclogites suggest that they may have protolithsthat are not cogenetic Some discrepancies in particularaffecting the incompatible elements may be due to post-crystallization processes acting upon the protoliths suchas dehydration melting and later metasomatism that isevident in peridotites from the same locality (Griffin et al1999)Low K2OTh for high-Ca and high-Mg eclogites

(average 08 Table 4) relative to assumed crustal

0

2

4

6

8

10

30 40 50 60 70 80

Mg

VR43452

Yb

GARNET

pyroxenites

low-Mghigh-Mghigh-Ca

eclogites

VR19673ecl-g2

Fig 14 Garnet Mg-number vs Yb (ppm including 2s uncertainty)Lines qualitatively show separate correlations for eclogites (exceptoutlierVR43452) and pyroxenites (except outlierVR19673ecl-g2)

high-Ca

high-Mg

001

01

1

10

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

melt08 melt02 cryst cpx01 melt09 cryst

plag

(a)

001

01

1

10(b)

RockNMORB

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 15 Reconstructed (a) high-Ca and (b) high-Mg eclogitescompared with compositions of plagioclase and cpx crystallizingfrom a MORB melt modelled using distribution coefficients ofNorman et al (2005)The MORB was calculated as a 10 equilibriummelt of MORB pyrolite [composition from Kelemen et al (1993)except Pb Pr Sr Y and Lu which were calculated as residues from1 fractional melting of pyrolite from McDonough amp Sun (1995)using melt modes of Kinzler (1997) and distribution coefficients ofNorman et al (2005)] Also shown are mixtures of crystals (consistingof 50 plagioclase plus 50 cpx) and melt in the proportions 0901and 0208

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1866

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

precursors (eg gabbros SE Indian Ridge average 22Hart et al 1999) may reflect the preferred loss of fluid-mobile elements during dehydration (Becker et al 2000)although large uncertainties on these low-abundance ele-ments in the constituent minerals and reconstructedwhole-rocks makes this interpretation less than robustThe REE patterns of most high-Mg eclogites are too flatto represent melting residues as partial melting shouldproduce low LREEHREE (Electronic Appendix 10) Incontrast the marked LREE depletion of five of sevenhigh-Ca eclogites (LaLufrac14 02^08) contrasts with that ofassumed evolved gabbroic protoliths (LaLufrac1415^32Hart et al 1999) and can be modelled as a consequence ofpartial melting during subduction after eclogitization (iein the presence of garnet which retains HREE relative toLREE eg Barth et al 2002a) has taken place (Aulbachet al 2003)Our results are compatible with previous studies on

eclogite xenoliths from the Slave Craton includingthose from the same kimberlite that have linked thesesamples or some of these samples to Paleoproterozoicsubduction events (the 195^191Ga formation of theHottah terrane and the 18Ga formation of the GreatBear magmatic arc) in the Slave craton (Griffin et al1999 Kopylova et al 1999 Heaman et al 2002 2006Schmidberger et al 2005 2007)The compositional features discussed above although

compatible with crustal protoliths are not necessarilydefinitive of a subduction origin However if direct crystal-lization from basaltic melts at high pressure (ie mantledepths) is invoked instead specific conditions would haveto be met to satisfy the observations The Mg-number ineclogites from Lac de Gras are lower than those expectedof primary high-pressure mantle melts (eg Walter 1998OrsquoHara amp Herzberg 2002) requiring crystallization athigh pressures after olivine fractionation This originwould further require shielding from peridotitic wallrockwith which the fractionated melts would no longer be inequilibrium (Barth et al 2001) In such a setting the lowrelative abundances of fluid-mobile elements comparedwith those of less fluid-mobile elements and of LREErelative to HREE (Table 4) would reflect an expulsion offluids during crystallization of the protoliths and latermelt mobilization respectively The eclogites clearly donot represent garnet plus cpx cumulates which wouldrequire a positive correlation of Mg-number with Yb thatis at variance with the observed negative correlation ineclogitic garnets from this study (Fig 14)

Low-Mg eclogites

Garnets in low-Mg eclogites compositionally overlap withthose in massive eclogites from Jericho (NW of the Lacde Gras kimberlites) that were suggested to have a high-pressure intrusive origin based on the similarity of theirmajor-element compositions to those of Group A eclogites

fromsouthernAfrica forwhichamantleoriginhasbeenpro-posed (Kopylova et al1999) Garnets in low-Mg eclogites inthe present study have major-element compositions moresimilar to garnets in groupB rather thangroupAeclogitesThe positive slope in N-MORB-normalized HREE

abundances displayed by the low-Mg eclogites is clearlydistinct from other eclogite types and is similar to those ofboninite-like second-stage melts that are closely associatedwith ultramafic volcanics (Smithies et al 2004) (Fig 13a)The range of Mg-numbers in the low-Mg eclogites (55^66)overlaps with that of Archaean second-stage melts fromIsua (58^75) and Abitibi (52^82 Smithies et al 2004)These similarities might indicate that the low-Mg eclogiteshave an origin in a previously depleted mantle source suchas supra-subduction zone mantle Subchondritic ZrHf ofthe low-Mg eclogites contrasts with the supra-chondriticZrHf of the high-Mg and high-Ca eclogites (Aulbach et

al in preparation) and is consistent with previous cpx-controlled depletion in an arc mantle source This sourcecould have been subsequently remelted to generate theboninite-like precursor to the low-Mg eclogites as sug-gested for the protoliths of eclogites from Udachnaya(Jacob amp Foley 1999)Although some second-stage melts to which the low-Mg

eclogites are similar are closely associated with komatiitesand could contain a plume component the radiogenic ordepleted mantle-like Nd and Hf and unradiogenic Sr inthe low-Mg eclogites do not support plume involvement

Trace-element and Sr^Nd^Hf isotopeconstraints on post-formationmetasomatismSeveral low-Mg and high-Ca eclogites show negativeslopes in the LREE which may be indicative of interactionwith small amounts of LREE-enriched melt (Navon ampStolper 1987) amounting to cryptic metasomatism(ie not accompanied by crystallization of new phases) inthe nomenclature of Dawson (1984) High Nb abundancesrelative to elements of similar compatibility displayed bysome low-Mg and high-Ca eclogites (Fig 10) could resultfrom overestimation of rutile modes However high-CaeclogiteVR40345 is rutile-free and has been reconstructedwithout rutile its high Nb abundance is accompanied byLREE enrichment indicating that metasomatic additionmay be responsibleLow-Mg eclogites show a positive correlation of La with

Nb (Electronic Appendix 11a) indicating concomitantaddition of LREE and Nb This suggests that the metaso-matic agent was a solute-rich fluid or silicate silicic melt ashydrous fluids and carbonatites do not carry enough HFSEto affect the budget of eclogites (Green amp Wallace 1988Klemme et al 1995 Adam et al 1997 Stalder et al 1998)Of the samples that are enriched in Nb all but one (high-Ca eclogiteVR50909) show a simultaneous enrichment in

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1867

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Hf (Electronic Appendix 11b) Together these trace-element relationships suggest that both SmNd and LuHfwere lowered in these samples Their low parent^daughterratios contrast with higher ratios in unenriched samplesand resulted in the ingrowth of a range of Nd and Hf iso-tope compositions Compared with the low-Mg and high-Ca eclogites the high-Mg eclogites have a narrow range ofLa and Nb concentrations which correlates with a narrowrange in Nd and Hf isotope compositions (Fig 12)Precise model ages cannot be calculated because parent^

daughter ratios are available only from in situ analyseswhich have relatively large uncertainties Howeverevidence for long-term enrichment and depletion may bediscerned Low-Mg eclogites even those that are enrichedin LREE have relatively high SmNd as a result of havingoverall positive slopes in their REE patterns relative toMORB whereas the high-Mg eclogites have relativelyflat REE patterns and therefore lower SmNd (Fig 11)This translates into lower eNd for high-Mg eclogites thanfor low-Mg eclogites In contrast they show no systematicdifference with respect to 176Hf177Hf Because the low- andhigh-Mg eclogites are proposed to have different origins asdiscussed above and therefore did not necessarily have thesame initial isotopic composition and formation age theseisotopic differences cannot be used to estimate their agesA long-term coupling of trace-element and isotopic

abundances is also indicated by the fact that low-Mg eclo-gite YK1943 which has experienced Nb and LREE addi-tion has lower eNd (0) and eHf (thorn13) than low-Mg eclogiteVR43469 with higher LuHf and SmNd (eNdfrac14thorn47eHffrac14thorn128) If low-Mg eclogites YK1943 and VR43469originally had identical initial Nd and Hf isotope composi-tions and if the parent^daughter ratios are accurate andwere modified only once during their evolution shortlyafter metamorphism (when their isotopic compositionwould have been reset) this overprint can be calculatedfor both isotopic systems to have occurred at 17GaThis would suggest that Hf and Nd were added duringthe same event In contrast high-Mg eclogites VR67360and YK1949 would have had identical Ndisotope compositions at 10Ga whereas their Hf isotopecompositions are disturbed in that the sample with thehigher parent^daughter ratio has a lower 176Hf177Hf thanthe sample with the lower parent^daughter ratio Theseresults suggest that the LREE and HFSE enrichmentobserved in different samples occurred during separateevents that did not affect the eclogites equallyExtreme HFSE enrichment leading to the growth of

zircon and rutile and LREE addition leading to thegrowth of apatite in eclogites from the Jericho kimberlitein the northern Slave craton have been ascribed to twoseparate metasomatic events at around 18Ga (quasi-coincident with the age of formation and metamorphism)and 13^10Ga respectively (Heaman et al 2002 2006

Schmidberger et al 2005) The authors associated theseevents with Paleoproterozoic subduction and aMesoproterozoic thermal disturbance in the Slave litho-spheric mantle possibly linked to the Mackenzie igneousevent Mesoproterozoic Nd model ages mostly between13 and 14Ga were reported for eclogites from thenearby Ekati kimberlites (Jacob et al 2003) Although wedo not observe zircon or apatite in volatile-free eclogitesfrom Lac de Gras and the age constraints obtained fromour data are weak they support a link between theseevents described for other eclogite populations and themetasomatic enrichment observed in our studyThe Sr isotope data in the present study are similar to

those of Schmidberger et al (2007) who suggested that thedepleted-mantle-like values are consistent with derivationof the protoliths from an oceanic mantle source However87Sr86Sr in volatile-free eclogites in the present study isremarkably constant considering their Nd and Hf isotopevariability (Fig 12) and the variability of trace-elementpatterns (Fig 10) Given the evidence for dehydration andmelting in many of the eclogites fractionation of highlyincompatible Rb from Sr and differential ingrowth ofradiogenic Sr would be expectedThis may indicate a rela-tively recent overprint and considering the similarityof 87Sr86Sr to present-day depleted mantle which is signif-icantly lower than 87Sr86Sr of North American kimber-lites a juvenile (asthenospheric) source

PyroxenitesPyroxenites from Lac de Gras often show relationshipsin their major-element contents that are different fromthose observed in the eclogites For example pyroxeniticgarnets show a negative correlation between CaO andNa2O contents whereas these oxides are positively corre-lated for garnets in eclogites (Fig 5) CaO is well correlatedwith C2O3 in pyroxenitic garnets but not in eclogitic gar-nets (Fig 4) Mg-number is positively correlated withYb abundance for pyroxenitic garnets but anti-correlatedfor eclogitic garnets (Fig 14) This suggests that the pyrox-enites in the present study are not related to the eclogitesby fractionation or accumulation processes as proposedfor the origin of some pyroxenites (Foley et al 2003)Pyroxenites have also been interpreted as the products

of crystal segregation in magma channels with alteredoceanic crust as the magma source this would explaintheir Eu anomalies and stable and radiogenic isotopecompositions (Pearson et al 1993 Becker 1996) As thehigh-pressure liquidus phases of most terrestrial basaltsare cpx and garnet (OrsquoHara amp Herzberg 2002) partialmelts of eclogites with basaltic composition wouldcrystallize cpx and garnet but not opx More than half ofthe pyroxenites from Lac de Gras contain opx suggestingthat they probably did not crystallize directly fromeclogite-derived melts

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1868

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Low-Cr websterites and pyroxenites could be crystalli-zation products of mafic melts that interacted with perido-titic mantle thus becoming more refractory The positivecorrelation of Yb and Mg-number in garnet would allowan origin by fractional crystallization from such meltsAlternatively they formed by reaction of silicic (possiblyeclogite-derived) melts with mantle peridotites(eg Kelemen et al 1993 Aulbach et al 2002) This agreeswith the similarity of hypothetical melts in equilibriumwith pyroxenitic garnets in most samples to eclogite-derived melts (Electronic Appendix 12) Niobium concen-trations in all hypothetical melts are much higher than ineclogite-derived melts regardless of whether or not rutileis assumed to be in the residue The same is true for LaCe and Zr concentrations in some samples which mayindicate that NbZr and LREE were added to thepyroxenites some time after their formation as discussedin the previous section for some eclogites Because mini-mum ages for pyroxenitic rutile with an unradiogenicHf isotope composition (calculated by projecting the176Hf177Hf to the model depleted mantle isotopic evolutioncurve assuming zero ingrowth) range to 178066Ga(Aulbach et al in preparation) it appears likely that thepyroxenites were not formed recently (eg during kimber-lite-related or precursory melt migration) and that some ofthe samples are at least MesoproterozoicGarnet pyroxeniteYK1952 consists of schlieren of equili-

brated garnet pyroxenite in a lsquomatrixrsquo of spongy cpxfine cpx^opx intergrowths embayed garnet and meltpatches (Fig 2f) Garnet in the pyroxenite schlieren hasvery low LREE and HFSE abundances Orthopyroxenein the matrix has distinctly higher Mg-number andCr-number and contains more CaO than opx from thepyroxenite schlieren whereas opx rims in the schlierenare compositionally similar to opx in the matrix The con-siderable compositional and textural disequilibrium inthis sample suggests that this is a young feature causedby reaction with fluids shortly before or at the time ofentrainmentAs a consequence of the small sample sizes of most of

the pyroxenites studied here a comprehensive isotopicdataset is not available Clearly more such data areneeded in addition to stable isotope data to obtain betterconstraints on the origins of this rock type Consideringthe microstructural and compositional variability ofthe pyroxenites and major-element relationships oppositeto those observed for eclogites it appears certain that pyr-oxenites form by a variety of processes some of which areunrelated to the formation of eclogites and their protoliths

Volatile-rich eclogitesVolatile-rich eclogites from Lac de Gras display the mostextreme Hf^Nd isotope systematics of all the analysedsamples Garnets in phlogopitized (YK1911) and in cal-cite-bearing eclogite (VR67112b) have the most radiogenic

Nd (eNdfrac14 2486 and 1927 respectively) Garnet inVR67112b has extremely high eHf (2516) Clinopyroxene inYK1911 has the lowest eHf measured for any mineralseparate from Lac de Gras (53) in accord with the veryhigh Hf abundance in its cpx (5 ppm) In contrastgarnet in this sample has radiogenic Hf (eHffrac14 136) Thisdiscrepancy is at least in part due to their relatively lowequilibration temperatures (7698C and 8558C respec-tively) which allowed the isotopic compositions of cpxand garnet to evolve separately at their low andhigh parent^daughter element ratios respectively forextended periods of time and possibly since formationAccordingly the cpx in YK1911 gives a minimum Hf ageof 27GaSlab-derived fluids could have been introduced during

the 27Ga collision of the western Slave Craton domainincluding the thick plume-derived deep lithosphericmantle layer and the eastern domain with its highlydepleted lithosphere This would also be consistent withthe observation that volatile-rich eclogites occur at shallowdepths in the mantle (TKroghfrac14 685^8558C correspondingto depths of 85^115 km along a 40 mWm2 geotherm)where the bulk of dehydration and production of solute-rich melts in a subducting slab might be expected to occur(Becker et al 2000 Hermann et al 2006) Their negativeEu anomalies might indicate that the fluid was derivedfrom the shallow portion of the slab that experiencedplagioclase fractionation The presence of carbonate inVR67112b requires a carbonated sourceHigh 87Sr86Sr in cpx in YK1911 is probably the conse-

quence of isotopic equilibration with high-RbSr phlogo-pite after long-term radiogenic ingrowth while Hfaddition led to retarded ingrowth of radiogenic HfThese systematics combined with radiogenic Nd appear

inconsistent with the involvement of sediments in the pet-rogenesis of this sample (eg Perfit et al 1980 Vervoortet al 1999) or a hydrous fluid such as those responsible forMARID metasomatism in the Kaapvaal craton (Konzettet al 1998)Whatever the origin of the fluids the apparentHf mobility suggests that they were not in equilibriumwith a rutile-bearing source and also that they weresolute-rich in nature (ie hydrous melts able to dissolveHFSE) and the relatively high HREE abundances of thevolatile-rich eclogites indicate a garnet-free source suchas amphibolite This allowed evolution of the sample atrelatively high time-integrated SmNd

SUMMARY AND CONCLUSIONS

(1) The Lac de Gras kimberlites carried abundant eclo-gite and pyroxenite xenoliths to the surface Threemain eclogite types are distinguished in terms ofmajor-element composition and trace element

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1869

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

patterns high-Mg high-Ca and low-Mg eclogitesThe garnets of all eclogite types have anti-correlatedMg-number and Yb pointing to olivine fractionationas a controlling factor in the formation of their pre-cursor rocks Their flat or positively sloped HREEpatterns qualitatively indicate that the melt parentalto the precursor rocks was not in equilibrium with agarnet-bearing residue

(2) Reconstructed high-Mg eclogites and high-Caeclogites have flat HREE patterns positive Pb andSr anomalies and low ZrSm similar to oceanic gab-bros reported in the literature High-Mg eclogiteshave variably low HREE contents and some havesmall positive Eu anomalies possibly suggestive ofmore primitive relatively plagioclase-rich protolithswhereas high-Ca eclogites have higher HREE abun-dances and small or absent Eu anomalies that requiremore cpx- andor melt-rich protoliths if a subductionorigin is assumed

(3) Low-Mg eclogites have distinctive trace-element pat-terns that show a marked positive slope in the HREEand subchondritic ZrHf They resemble someArchaean second-stage melts (lsquoboninitesrsquo) and theirprotoliths appear to have formed from a previouslydepleted source

(4) After formation the eclogites were subject to a varietyof processes Most have low ratios of fluid-mobile vsless fluid-mobile elements possibly as a result of dehy-dration of the protoliths In addition some of the low-Mg and high-Ca eclogites have low LREEHREEconsistent with partial melt lossWhereas the low-Mgeclogites show concomitant enrichment in La Nb andZr some high-Mg and high-Ca eclogites show evi-dence for variable and unsystematic (decoupled)LREE and Nb addition

(5) Isotopically the trace-element characteristics of thedifferent eclogite types translate into lower eNd forhigh-Mg eclogites (negative slope in MREEfrac14 lowerSmNd) than for low-Mg eclogites (positive slope inMREEfrac14higher SmNd) Within the low-Mg groupsamples that show evidence for metasomatism (lowLuHf and SmNd) have lower eNd and eHf than asample that was apparently not metasomatized BothNd and Hf addition can be dated to 17Ga if identi-cal initial isotope compsitions are assumed for thelow-Mg eclogites In contrast the relatively constantand depleted mantle-like 87Sr86Sr of eclogitic cpxdespite variable Nd and Hf isotope compositions andparent^daughter ratios suggests recent input of juve-nile material

(6) Garnet and cpx in pyroxenites show major-elementrelationships different from those in eclogitesTherefore these two rock types are probably notrelated by fractionation processes The presence of

opx precludes direct crystallization from slab-derivedmelts Pyroxenites may be the product of (slab-derived) melts hybridized during assimilation of peri-dotitic mantle

(7) Garnets in phlogopite- and carbonate-bearing eclo-gites have extreme isotopic compositions includingthe most radiogenic Nd and Hf in the datasetNegative Eu anomalies may point to the involvementof a crustal protolith Highly radiogenic Sr radiogenicNd and unradiogenic Hf in the reconstructedphlogopite-bearing sample combined with highHREE abundances and evidence for Hf mobilitymay indicate involvement of solute-rich fluids origi-nating in a garnet- and rutile-free source

ACKNOWLEDGEMENTSHelp in the clean laboratories and with the analyticalfacilities by S Elhlou C Lawson A Sharma P Wielandand M Zhang is greatly appreciated Bill Griffin isthanked for his support and for agreeing to reviewseveral versions of the manuscript despite disagreeingwith many of the conclusions presented here The paperbenefited greatly from the detailed and constructivereviews provided by the editors Gareth Davies andMarjorie Wilson and by Matthias Barth Larry Heamanand an anonymous reviewer This work was funded by aMacquarie University International PostgraduateAward and Postgraduate Research Fund (SA) by anARC SPIRT grant sponsored by Kennecott Canada Incand by ARC Large and Discovery Grants to SYOrsquoRAnalytical data were obtained using instrumentationfunded by ARC LIEF and DEST Systemic InfrastructureGrants industry partners and Macquarie UniversityThis is Publication 481 from the ARC National KeyCentre for Geochemical Evolution and Metallogeny ofContinents (wwwesmqeduauGEMOC)

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online

REFERENCESAdam J GreenT H Sie S H amp Ryan C G (1997) Trace element

partitioning between aqueous fluids silicate melts and mineralsEuropeanJournal of Mineralogy 9 569^584

Aulbach S Stachel T Viljoen K S Brey G P amp Harris J W(2002) Eclogitic and websteritic diamond sources beneath theLimpopo Beltccedilis slab-melting the link Contributions to Mineralogy

and Petrology 143 56^70Aulbach S GriffinW L Pearson N J OrsquoReilly S Y Kivi K amp

Doyle B (2003) Origins of eclogites beneath the central SlaveCraton Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA-0011

Aulbach S Griffin W L Pearson N J OrsquoReilly S Y amp Kivi K(2004a) Mantle formation and evolution Slave Craton constraints

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1870

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

from HSE abundances and Re^Os isotope systematics of sulfideinclusions in mantle xenocrysts Chemical Geology 208 61^88

Aulbach S Griffin W L OrsquoReilly S Y amp McCandless T E(2004b) Genesis and evolution of the lithospheric mantle beneaththe Buffalo HeadTerrane Alberta (Canada) Lithos 77 413^451

Aulbach S GriffinW L Pearson N J OrsquoReilly S Y amp Doyle B(2007) Lithosphere formation in the central Slave Craton(Canada) plume subcretion or lithosphere accretion Contributionsto Mineralogy and Petrology (in press) doi 101007s00410-007-0200-1

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2001)Geochemistry of xenolithic eclogites from West Africa Part IA link between low MgO eclogites and Archean crust formationGeochimica et Cosmochimica Acta 65 1499^1527

Barth M G Foley S F amp Horn I (2002a) Partial melting inArchean subduction zones constraints from experimentally deter-mined trace element partition coefficients between eclogitic miner-als and tonalitic melts under upper mantle conditions PrecambrianResearch 113 323^340

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2002b)Geochemistry of xenolithic eclogites from West Africa part 2Origins of the high MgO eclogites Geochimica et Cosmochimica Acta

66 4325^4345Beard B L Fraracci K NTaylor L A Snyder G A Clayton R

A Mayeda T K amp Sobolev NV (1996) Petrography andgeochemistry of eclogites from the Mir kimberliteYakutia RussiaContributions to Mineralogy and Petrology 125 293^310

Becker H (1996) Geochemistry of garnet peridotite massifs fromlower Austria and the composition of deep lithosphere beneatha Palaeozoic convergent plate margin Chemical Geology 134 49^65

Becker H Jochum K P amp Carlson R W (2000) Trace elementfractionation during dehydration of eclogites from high-pressureterranes and the implications for element fluxes in subductionzones Chemical Geology 163 65^99

Bell K amp Tilton G R (2002) Probing the mantle The story fromcarbonatites EOS Transactions American Geophysical Union 83275^277

Benoit M Polvecurren M amp Ceuleneer G (1996) Trace element andisotopic characterisation of mafic cumulates in a fossil mantlediapir (Oman ophiolite) Chemical Geology 134 199^214

BleekerW (2003) The late Archean record a puzzle in ca 35 piecesLithos 71 99^134

BleekerW Ketchum JW F JacksonV A amp Villeneuve M (1999)The central Slave Basement Complex Part I Its structural topol-ogy and autochthonous core Canadian Journal of Earth Sciences 361083^1109

Blichert-Toft J amp Albare de F (1997) The Lu^Hf isotope geochemis-try of chondrites and the evolution of the mantle^crust systemEarth and Planetary Science Letters 148 243^258

Blichert-Toft J Chauvel C amp Albare de F (1997) Separation of Hfand Lu for high precision isotope analysis of rock samples by mag-netic sector-multiple collector ICP-MS Contributions to Mineralogy

and Petrology 127 248^260Brey G P amp Kolaquo hler T (1990) Geothermobarometry in four-phase

lherzolites II New thermobarometers and practical assessment ofexisting thermobarometers Journal of Petrology 31 1353^1378

Brey G P Kolaquo hlerT amp Nickel K G (1990) Geothermobarometry infour-phase lherzolites I Experimental results from 10 to 60 kbJournal of Petrology 31 1313^1352

Caporuscio F amp Smyth J (1990) Trace element crystal chemistryof mantle eclogites Contributions to Mineralogy and Petrology 105550^561

Carbno G B amp Canil D (2002) Mantle structure beneath the SWSlave Craton Canada constraints from garnet geochemistry in theDrybones Bay Kimberlite Journal of Petrology 43 129^142

Coleman R G Lee E D Beatty L B amp Brannock W W (1965)Eclogites and eclogites their differences and similaritiesGeological Society of America Bulletin 76 483^508

Creaser R A Grulaquo tter H Carlson J amp Crawford B (2004)Macrocrystal phlogopite Rb^Sr dates for the Ekati propertykimberlites Slave Province Canada evidence for multiple intru-sive episodes in the Paleocene and Eocene Lithos 76 399^414

Davies R M GriffinW L Pearson N J Andrew A S Doyle BJ amp OrsquoReilly S Y (1999) Diamonds from the deep Pipe DO-27Slave Craton Canada In Gurney J J Gurney J L Pascoe MD amp Richardson S H (eds) Proceedings of the 7th International

Kimberlite Conference CapeTown Red Roof Design pp 148^155Davies R M Griffin W L OrsquoReilly S Y amp Doyle B J (2004)

Mineral inclusions and geochemical characteristics of microdia-monds from the DO27 A154 A21 A418 DO18 DD17 and RanchLake kimberlites at Lac de Gras Slave Craton Canada Lithos 7739^55

DavisW J amp Hegner E (1992) Neodymium isotopic evidence for thetectonic assembly of Late Archean crust in the Slave Provincenorthwest Canada Contributions to Mineralogy and Petrology 111493^504

Davis W J Jones A G Bleeker W amp Grutter H (2003)Lithosphere development in the Slave craton a linked crustal andmantle perspective Lithos 71 575^589

Dawson J B (1984) Contrasting types of upper-mantle metasoma-tism In Kornprobst J (ed) Kimberlites II The Mantle and

Crust^Mantle Relationships Amsterdam Elsevier pp 289^294DePaolo D J (1981) Neodymium isotopes in the Colorado Front

Range and crust^mantle evolution in the Proterozoic Nature 291193^196

Erlank A J amp Kushiro I (1970) Potassium contents of syntheticpyroxenes at high temperatures and pressures Carnegie Institution ofWashingtonYearbook 68 433^439

Foley S F Barth M G amp Jenner G A (2000) Rutilemelt partitioncoefficients for trace elements and an assessment of the influence ofrutile on the trace element characteristics of subduction zonemagmas Geochimica et Cosmochimica Acta 64 933^938

Foley S F Buhre S amp Jacob D E (2003) Evolution of the Archaeancrust by delamination and shallow subduction Nature 421 249^252

Graham I Burgess J L Bryan D Ravenscroft P J Thomas EDoyle B J Hopkins R amp Armstrong K A (1999) Explorationhistory and geology of the Diavik Kimberlites Lac de GrasNorthwest Territories Canada In Gurney J J Gurney J LPascoe M D amp Richardson S H (eds) Proceedings of the 7th

International Kimberlite Conference Cape Town Red Roof Designpp 262^279

Green D H amp Wallace M E (1988) Mantle metasomatism byephemeral carbonatite melts Nature 336 459^462

Green T H Blundy J D Adam J amp Yaxley G M (2000) SIMSdetermination of trace element partition coefficients betweengarnet clinopyroxene and hydrous basaltic liquids at 2^75 GPaand 1080^12008C Lithos 53 165^187

GriffinW L amp OrsquoReilly SY (2007) Cratonic lithospheric mantle Isanything subducted Episodes 30(1) 43^53

Griffin W L Wass S Y amp Hollis J D (1984) Ultramafic xenolithsfrom Bullenmerri and Gnotuk maars Victoria Australia petrol-ogy of a sub-continental crust^mantle transition Journal of

Petrology 25 53^87GriffinW L Doyle B J Ryan C G Pearson N J OrsquoReilly SY

Davies R Kivi K van Achterbergh E amp Natapov LM (1999)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1871

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Layered mantle lithosphere in the Lac de Gras Area SlaveCraton composition structure and origin Journal of Petrology 40705^727

Griffin W L Pearson N J Belousova E Jackson S E vanAchterbergh E OrsquoReilly SY amp Shee S R (2000)The Hf isotopecomposition of cratonic mantle LAM-MC-ICPMS analysis ofzircon megacrysts in kimberlites Geochimica et Cosmochimica Acta

64 133^147Griffin W L OrsquoReilly S Y Doyle B J Pearson N J

Coopersmith H Kivi K Malkovets V amp Pokhilenko N(2004) Lithosphere mapping beneath the North American plateLithos 77 873^922

Grulaquo tter H S Apter D B amp Kong J (1999) Crust^mantle couplingevidence from mantle-derived xenocrystic garnets In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 307^312

Hart S R Blusztajn J Dick H J B Meyer P S ampMuehlenbachs K (1999) The fingerprint of seawater circulationin a 500-meter section of ocean crust gabbros Geochimica et

Cosmochimica Acta 63 4059^4080Harte B amp Kirkley M B (1997) Partitioning of trace elements

between clinopyroxene and garnet Data from mantle eclogitesChemical Geology 136 1^24

Heaman L M Creaser R A amp Cookenboo H O (2002) Extremeenrichment of high field strength elements in Jericho eclogitexenoliths A cryptic record of Paleoproterozoic subduction partialmelting and metasomatism beneath the Slave craton CanadaGeology 30 507^510

Heaman L M Kjarsgaard B A amp Creaser R A (2004) The tem-poral evolution of North American kimberlites Lithos 76 377^397

Heaman L M Creaser R A Cookenboo H O amp Chacko T(2006) Multi-stage modification of the northern Slave mantlelithosphere Evidence from zircon- and diamond-bearing eclogitexenoliths entrained in Jericho Kimberlite Journal of Petrology 47821^858

Hermann J Cpandler C Hack A amp Korsakov A (2006) Aqueousfluids and hydrous melts in high-pressure and ultra-high pressurerocks implications for element transfer in subduction zones Lithos92 399^417

Hills DV amp Haggerty S E (1989) Petrochemistry of eclogites fromthe Koidu Kimberlite Complex Sierra Leone Contributions to

Mineralogy and Petrology 103 397^422Hoffman P F (1989) Precambrian geology and tectonic history of

North America In Bally A amp Palmer A (eds) The Geology of

North AmericaccedilAn Overview Boulder CO Geological Society ofAmerica pp 447^512

IrelandT R Rudnick R L amp Spetsius Z (1994) Trace-elements indiamond inclusions from eclogites reveal link to Archean granitesEarth and Planetary Science Letters 128 199^213

Irvine G J Pearson D G Kjarsgaard B A Carlson R WKopylova M G amp Dreibus G (2003) Evolution of the litho-spheric mantle beneath Northern Canada a Re^Os isotope andPGE study of kimberlite-derived peridotite xenoliths fromSomerset Island and a comparison to the Slave and Kaapvaalcratons Lithos 71 461^488

Isachsen C E amp Bowring S A (1994) Evolution of the Slave cratonGeology 22 917^920

Jacob D E (2004) Nature and origin of eclogite xenoliths from kim-berlites Lithos 77 295^316

Jacob D E amp Foley S F (1999) Evidence for Archean ocean crustwith low high field strength element signature from diamondifer-ous eclogite xenoliths Lithos 48 317^336

Jacob D Jagoutz E Lowry D Mattey D amp Kudrjavtseva G(1994) Diamondiferous eclogites from Siberiaccedilremnants ofArchean oceanic-crust Geochimica et Cosmochimica Acta 585191^5207

Jacob D E Fung A Jagoutz E amp Pearson D G (2003) Petrologyand geochemistry of eclogite xenoliths from the Ekati kimberlitesarea Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA_239

Jacob D E Bizimis M amp Salters V J M (2005) Lu^Hf and geo-chemical systematics of recycled ancient oceanic crust evidencefrom Roberts Victor eclogites Contributions to Mineralogy and

Petrology 148 707^720Jerde E A Taylor L A Crozaz G Sobolev NV amp SobolevV N

(1993) Diamondiferous eclogites from Yakutia Siberiaccedilevidencefor a diversity of protoliths Contributions to Mineralogy and Petrology

114 189^202Jones A G Ferguson I J Chave A D Evans R L amp

McNeice GW (2001) The electric lithosphere of the Slave cratonGeology 29 423^426

Kelemen P B Shimizu N amp Dunn T (1993) Relative depletionof niobium in some arc magmas and the continental-crust-partitioning of K Nb La and Ce during meltrockreaction in the upper-mantle Earth and Planetary Science Letters 120111^134

Kinzler R J (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition Application to mid-ocean ridge basalt petrogenesis Journal of Geophysical Research 102853^874

Klemme S van der Laan S R Foley S F amp Guenther D (1995)Experimentally determined trace and minor element partitioningbetween clinopyroxene and carbonatite melt under upper mantleconditions Earth and Planetary Science Letters 133 439^448

Konzett J Armstrong R A Sweeney R J amp CompstonW (1998)The timing of MARID metasomatism in the Kaapvaal mantle anion probe study of zircons from MARID xenoliths Earth and

Planetary Science Letters 160 133^145Kopylova M G Russell J K amp Cookenboo H (1999) Mapping the

lithopshere beneath the north central Slave Craton In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 468^479

Krogh E (1988) The garnet^clinopyroxene iron^magnesiumgeothermometerccedila reinterpretation of existing experimentaldata Contributions to Mineralogy and Petrology 99 44^48

Kusky T M (1989) Accretion of Archean Slave Province Geology 1763^67

LeCheminant A N amp Heaman L M (1989) Mackenzie igneousevents Canada Middle Proterozoic hotspot magmatismassociated with ocean opening Earth and Planetary Science Letters

96 38^48LeCheminant A N Richardson D G DiLabio R N W amp

Richardson K A (1996) La recherche de diamants au CanadaGeological Survey of Canada Open File 3228 268pp

MacGregor I D amp Carter J L (1970) The chemistry of clinopyrox-enes and garnets of eclogite and peridotite xenoliths from theRoberts Victor mine South Africa Physics of the Earth and PlanetaryInteriors 3 391^397

MacGregor I D amp Manton W I (1986) Roberts Victor eclogitesancient oceanic crust Journal of Geophysical Research 91 14063^14079

MacKenzie J M amp Canil D (1999) Composition and thermal evo-lution of cratonic mantle beneath the central Archean SlaveProvince NWT Canada Contributions to Mineralogy and Petrology

134 313^324

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1872

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

McCandlessT E amp Gurney J J (1989) Sodium in garnet and potas-sum in clinopyroxene criteria for classifying mantle eclogites InRoss J (ed) Kimberlites and Related Rocks Carlton Blackwellpp 827^832

McDonoughW F amp Sun S-S (1995) The composition of the EarthChemical Geology 120 223^253

Navon O amp Stolper E (1987) Geochemical consequences of meltpercolation the upper mantle as a chromatographic columnJournal of Geology 95 285^307

Nikogosian I K amp Sobolev AV (1997) Ion-microprobe analysis ofmelt inclusions in olivine experience in estimating the olivine^meltpartition coefficients of trace elements Geochemistry International 35119^126

Norman M D Griffin W L Pearson N J Garcia M O ampOrsquoReilly S Y (1998) Quantitative analysis of trace element abun-dances in glasses and minerals a comparison of laser ablationinductively coupled plasma mass spectrometry solution inductivelycoupled plasma mass spectrometry proton microprobe and elec-tron microprobe data Journal of Analytical Atomic Spectrometry 13477^482

Norman M D Garcia M O amp Pietruszka A J (2005) Trace-element distribution coefficients for pyroxenes plagioclase andolivine in evolved tholeiites from the 1955 eruption of KilaueaVolcano Hawairsquoi and petrogenesis of differentiated rift-zonelavas American Mineralogist 90 888^899

OrsquoHara M J amp Herzberg C (2002) Interpretation of trace elementand isotope features of basalts Relevance of field relations petrol-ogy major element data phase equilibria and magma chambermodeling in basalt petrogenesis Geochimica et Cosmochimica Acta 662167^2191

PadghamW A amp FysonW K (1992) The Slave Province a distinctcraton CanadianJournal of Earth Sciences 29 2072^2086

Pearson D G Davies G R amp Nixon P H (1993) Geochemical con-straints on the petrogenesis of diamond facies pyroxenites from theBeni Bousera peridotite massif North Morocco Journal of Petrology34 125^172

Pearson N J Griffin W L Doyle B J OrsquoReilly S Y vanAchterbergh E amp Kivi K (1999) Xenoliths from kimberlitepipes of the Lac de gras area Slave Craton Canada In GurneyJ J Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite ConferenceCapeTown Red Roof Design pp 644^658

Perfit N R Gust D A Bence A E Arculus R J amp Taylor S R(1980) Chemical characteristics of island-arc basalts implicationsfor mantle sources Chemical Geology 30 227^256

Polat A Hofmann A W amp Rosing M T (2002) Boninite-likevolcanic rocks in the 37^38 Ga Isua greenstone belt WestGreenland geochemical evidence for intra-oceanic subductionzone processes in the early Earth Chemical Geology 184 231^254

Poudjom DjomaniY H GriffinW L OrsquoReilly SY amp Doyle B J(2005) Lithospheric domains and controls on kimberlite emplace-ment Slave Province Canada evidence from elastic thickness andupper mantle composition Geochemistry Geophysics Geosystems 6doi1010292005GC0009978

Presnall D C Gudfinnsson G H amp Walter M J (2002)Generation of mid-ocean ridge basalts at pressures from 1 to 7GPa Geochimica et Cosmochimica Acta 66 2073^2090

Schmidberger S S Heaman L M Simonetti A Creaser R A ampCookenboo H O (2005) Formation of Paleoproterozoiceclogitic mantle Slave Province (Canada) Insights from in-situHf and U^Pb isotopic analyses of mantle zircons Earth and

Planetary Science Letters 240 621^633

Schmidberger S S Simonetti A Heaman L M Creaser R A ampWhiteford S (2007) Lu^Hf in-situ Sr and Pb isotope and traceelement systematics for mantle eclogites from the Diavik diamondmine Evidence for Paleoproterozoic subduction beneath the Slavecraton Canada Earth and Planetary Science Letters 254 55^68

Schulze D J amp Helmstaedt H (1988) Coesite-sanidine eclogitesfrom kimberlite products of mantle fractionation or subductionJournal of Geology 96 435^443

Smithies R H Champion D C amp Sun S S (2004) The case forArchaean boninites Contributions to Mineralogy and Petrology 147705^721

Smyth J R Caporuscio F A amp McCormick T C (1989) Mantleeclogites evidence of igneous fractionation in the mantle Earth andPlanetary Science Letters 93 133^141

Sobolev NV amp Lavrentrsquoyev Y G (1971) Isomorphic sodium admix-ture in garnets formed at high pressure Contributions to Mineralogy

and Petrology 31 1^12Stalder R Foley S F Brey G P amp Horn I (1998) Mineral aqueous

fluid partitioning of trace elements at 900^12008C and 30^57GPa New experimental data for garnet clinopyroxene and rutileand implications for mantle metasomatism Geochimica et

Cosmochimica Acta 62 1781^1801Sun S-S amp McDonough W F (1989) Chemical and isotopic sys-

tematics of oceanic basalts implications for mantle compositionsand processes In Saunders A D amp Norris M J (eds)Magmatism in the Ocean Basins Geological Society London Special

Publications 42 313^345Taura H Yurimoto H Kurita K amp Sueno S (1998) Pressure

dependence on partition coefficients for trace elements between oli-vine and the coexisting melts Physics and Chemistry of Minerals 25469^484

Taylor L A amp Neal C R (1989) Eclogites with oceanic crustal andmantle signatures from the Bellsbank kimberlite South AfricaPart I Mineralogy petrography and whole rock chemistryJournal of Geology 97 551^567

van Achterbergh E Ryan C G amp Griffin W L (1999) Glitter online intensity reduction for the laser ablation inductively coupledplasma spectrometry Abstract 9th V M GoldschmidtConference LPI Boston MA 905 pp Abstract no 7215

van Breemen O DavisW J amp King J E (1992) Temporal distribu-tion of granitoid plutonic rocks in the Archean Slave ProvinceNorthwest Canadian Shield Canadian Journal of Earth Sciences 292186^2199

Vervoort J D Patchett P J Blichert-Toft J amp Albare de F (1999)Relationship between Lu^Hf and Sm^Nd isotopic systems in theglobal sedimentary system Earth and Planetary Science Letters 16879^99

Viljoen K S Smith C B amp Sharp Z D (1996) Stable and radio-genic isotope study of eclogite xenoliths from the Orapa kimberliteBotswana Chemical Geology 131 235^255

Walter M J (1998) Melting of garnet peridotite and theorigin of komatiite and depleted lithosphere Journal of Petrology 3929^60

Wasserburg G J Jacobsen S B DePaolo D J McCulloch M Tamp Wen T (1981) Precise determination of SmNd ratios Sm andNd isotopic abundances in standard solutions Geochimica et

Cosmochimica Acta 45 2311^2323Westerlund K J Shirey S B Richardson S H Carlson R W

Gurney J J amp Harris JW (2006) A subduction wedge origin forPaleoarchean peridotitic diamonds and harzburgites from thePanda kimberlite Slave craton evidence from Re^Os isotope sys-tematics Contributions to Mineralogy and Petrology 152 275^294

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1873

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Page 17: Origins of Xenolithic Eclogites and Pyroxenites from the Central

Table 4 Trace-element compositions of reconstructed whole-rocks (ppm)

Sample P Sc Ti V Co Ni Ga Sr Y Zr Nb La Ce Pr Nd

Low-Mg eclogites

vr19674-g4 186 50 4649 317 57 59 144 77 24 30 36 13 31 05 24

vr43452 154 39 2253 328 49 84 176 52 32 11 01 02 07 01 08

vr43469 38 65 4377 349 54 91 142 34 28 15 02 05 15 03 17

yk1926 144 45 3406 258 52 87 145 89 18 18 20 09 29 06 33

yk1943 173 46 5732 295 41 39 146 122 25 48 82 34 85 13 64

yk19491 102 34 1693 158 31 49 100 46 10 14 03 09 21 03 17

High-Mg eclogites

vr19677-g2 138 39 5160 291 41 53 119 98 14 22 04 13 53 11 64

vr434792 46 25 1265 129 36 126 84 56 3 7 04 06 25 04 23

vr67360 49 36 1573 138 64 184 111 76 5 9 04 09 38 07 33

yk1946 153 28 4877 192 45 105 124 108 12 48 03 10 38 08 50

High-Ca eclogites

vr19674-g5 267 30 7663 200 41 67 195 32 23 43 77 01 04 02 16

vr40345 0 33 2821 175 47 98 210 179 13 13 66 60 109 11 38

vr43465 142 27 5104 220 31 83 121 197 14 35 29 05 22 05 40

vr50858 295 30 6898 201 44 94 150 48 22 38 06 01 03 02 20

vr50860 68 57 5672 418 57 71 186 109 19 11 02 02 13 03 26

vr50909 238 28 7295 201 42 92 171 48 20 57 13 01 04 02 21

yk3528 91 51 3712 341 53 155 177 49 16 9 01 02 08 02 11

Volatile-rich eclogites

vr67112b 40 84 9242 521 46 30 107 4 40 66 31 26 126 24 125

yk1911 31 39 9428 286 44 82 164 3 23 52 64 18 52 09 46

Pyroxenites

yk1914 111 49 707 207 48 547 44 31 4 3 02 06 19 03 10

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

Low-Mg eclogites

vr19674-g4 11 05 10 22 39 10 29 33 06 12 05 028 24 008

vr43452 04 02 08 13 38 12 41 49 08 05 01 nd 02 na

vr43469 11 05 09 25 44 11 34 36 06 07 02 001 09 129

yk1926 12 05 11 19 30 07 22 24 04 07 03 004 23 090

yk1943 21 08 10 29 42 10 30 32 05 16 05 013 65 017

yk19491 06 03 11 08 14 04 14 18 03 04 02 003 29 115

High-Mg eclogites

vr19677-g2 20 08 12 21 24 05 15 15 03 05 05 007 52 6811

vr434792 05 02 12 05 06 01 04 03 00 03 01 001 98 037

vr67360 06 03 14 06 09 02 06 06 01 02 02 002 86 000

yk1946 18 07 12 21 22 05 13 12 02 11 03 006 56 112

High-Ca eclogites

vr19674-g5 17 09 11 33 43 09 24 23 04 16 05 007 02 011

vr40345 09 04 10 12 22 06 16 14 02 05 nd nd 298 na

(continued)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1859

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 4 Continued

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

vr43465 22 09 12 25 26 06 15 14 02 10 04 001 25 040

vr50858 20 08 09 33 41 08 22 21 03 10 01 nd 02 na

vr50860 14 07 11 27 37 08 22 18 03 06 00 nd 06 na

vr50909 20 10 12 33 38 07 20 20 03 16 01 002 02 181

yk3528 08 05 11 20 30 07 18 18 02 05 01 nd 08 na

Volatile-rich eclogites

vr67112b 38 10 07 53 74 15 41 34 04 18 03 012 60 004

yk1911 20 07 08 35 39 07 18 17 02 10 14 nd 73 na

Pyroxenites

yk1914 02 01 11 04 08 02 05 06 01 01 03 003 68 101

aGarnet has CandashMgndashFe relationship typical of high-Mg eclogites but whole-rock trace-element pattern is typical oflow-Mg eclogitesbGarnet has CandashMgndashFe relationship typical of low-Mg eclogites but whole-rock trace-element pattern is typical ofhigh-Mg eclogites

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

(a)

001

01

1

10(c)

001

01

10(b)

VR43452

YK1949

RockNMORB

low-Mg

RockNMORB

high-Mg

(d)

volatile-richhigh-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

phlogopite-bearingYK1911

calcite-bearingVR67112b

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 10 Extended trace-element patterns of reconstructed (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-bearing eclogites Normalized to N-MORB of Sun amp McDonough (1989)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1860

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

RockNMORB

(b)

RockNMORB

(d)

010

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu001

01

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

(a)

001

01

1

10(b)

001

01

1

10(c)

001

01

1

10(e)

high-TSchmidberger et al

low-TSchmidberger et al

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

high-Mglow-Mg

high-Cavolatile-rich

phlogopite-bearingYK1911

calcite-bearingVR67112b

YK1949VR43479

Fig 11 N-MORB normalized REE patterns for (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-richeclogites Shown for comparison in (e) are reconstructed eclogites from Schmidberger et al (2007)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1861

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Both volatile-rich whole rocks display strong negativeSr anomalies Strontium can be sited in calcite or apatiteandor indicate the presence of a Sr-rich mineral that wasnot identified in thin section such as bariteFour of the six reconstructed whole-rocks in the study

of Schmidberger et al (2007) (Fig 11e) show markedLREEHREEN enrichment that is observed only in onehigh-Ca eclogite in our study (Fig 11c) suggesting thatmany of the eclogites they studied were affected by metaso-matic LREE enrichment Two samples in their study thatare not affected by this enrichment show positive Euanomalies similar to the high-Mg eclogites in this studywhereas the distinctive REE pattern of the low-Mg eclo-gites is not recognized among their samples

SR^ND^HF I SOTOPESYSTEMATICSStrontium Nd and Hf isotope data for optically puremineral separates and reconstructed whole-rocks aregiven in Table 5 Spongy rims and alteration are likely tobe mostly removed during leaching in 6N HCl in combina-tion with repeated ultrasonication Despite this rigoroustreatment the reproducibility of Nd in garnet VR43469and in cpx YK1914 is worse than for other samples Thismay be due to inhomogeneities related to late Nd diffusionpossibly from grain boundaries containing a kimberlitecomponent The reproducibility of the 176Hf177Hf data isgenerally poorer than that of Nd which may reflect a com-bination of matrix effects caused by imperfectly purifiedsolutions and sample heterogeneity Parent^daughterratios are available only from in situ analyses which arenot precise enough to allow meaningful model age orinitial isotope ratio calculations (ie at the time of kimber-lite eruption) Because the Lac de Gras kimberlites arerelatively young (55^56 Ma Graham et al 1999) age cor-rection would result in only a minor change in isotopecomposition which is inconsequential in view of the rangeof compositions displayed by most samples and hence hasno effect on the conclusions drawnClinopyroxenes in five volatile-free eclogites and a

pyroxenite have Sr isotope ratios ranging from 0701662 to0703546 this range is similar to or somewhat lower thanthat of the depleted mantle at the time of kimberliteemplacement (0703 Bell amp Tilton 2002) Clinopyroxenein the phlogopitized sample has a distinctly high 87Sr86Srof 0751186Clinopyroxene controls 56^92 of the Nd budget (Nd

concentration in cpx weighted by mode) Neodymiumisotope ratios range from 0511660 to 0514844 Thiscorresponds to eNd variation (per 10 000 deviationof 143Nd144Nd in the sample from 143Nd144Nd in the chon-dritic mantle model reservoir) from 17 to thorn43 cpx inthe low-Mg eclogites have higher eNd than those in

high-Mg eclogites Garnets have generally have higher143Nd144Nd from 0512984 (eNdfrac14thorn 7) to 0520965(eNdfrac14thorn 162) Garnets in the two volatile-rich sampleshave the highest ratios 0522790 for garnet in the calcite-bearing sample VR67112b (eNdfrac14thorn198) and 0525721 inthe phlogopitized eclogiteYK1911 (eNdfrac14thorn 255)Garnet contributes 18^32 and cpx 48^78 of the Hf

budget The proportion controlled by rutile is 4^30(Aulbach et al in preparation) Measured 176Hf177Hf incpx ranges from 0281263 in phlogopite-bearing sampleYK1911 (eHffrac1453) to 0286130 in low-Mg eclogiteVR43469 (eHffrac14thorn119) and in garnet from 0283958 inpyroxeniteYK1915 (eHffrac14thorn42) to 0295718 in low-Mg eclo-gite YK1926 (eHffrac14thorn458) Hafnium in garnet in the phlo-gopite^apatite-bearing sample is extremely radiogenic(176Hf177Hffrac14 0354367) corresponding to an eHf of thorn2532which is close to a value determined for a garnet separatefrom a Roberts Victor eclogite (eHffrac14thorn2561 Jacob et al2005) Garnet in the calcite-bearing sample has extremelyradiogenic Nd and Hf (eNdfrac14thorn193 eHffrac14thorn2516) Garnetin pyroxenite YK1915 also has radiogenic Nd and Hf(eNdfrac14thorn27 eHffrac14thorn42)Despite the large uncertainties for reconstructed whole-

rocks (calculated as the square root of the sum of squareduncertainties of two standard errors for garnet and cpxanalyses combined with 20 propagated uncertainty onthe proportion of Nd or Hf respectively contributed byeach mineral) the isotopic data are consistent in thatthey correlate with the independently determinedparent^daughter ratios (1) Low-Mg eclogites have higherSmNd reflected in the steeper positive slope of theirN-MORB-normalized REE patterns (Fig 11a) comparedwith the high-Mg eclogites (Fig 11b) Accordingly recon-structed low-Mg eclogites have higher present-day eNd

(0 to thorn47) than high-Mg eclogites (4 and 17) (2) Low-Mg eclogite YK1943 has very high Nb and LREE abun-dances that could have been added during metasomatismthat also enhanced its Hf abundance This sample has thelowest eNd and eHf of the low-Mg eclogites (0 and thorn20respectively) In contrast apparently unenriched low-Mgeclogites have more radiogenic Nd and Hf (eNdfrac14thorn47and thorn9 eHffrac14thorn128 and thorn113 respectively)The Sr^Nd isotope diagram reveals that eNd in cpx

varies from 19 to 43 at nearly constant depleted-mantle-like 87Sr86Sr (Fig 12a) In the Nd^Hf isotope diagram(Fig 12b) the reconstructed whole-rocks lie on curvedmixing lines if the lsquomixingrsquo proportions of Nd and Hffor the two lsquoend-memberrsquo minerals are very different(eg YK1911) Garnets have more radiogenic Nd andHf than coexisting cpx All reconstructed volatile-freeeclogites plot in the radiogenic field with regard to Hfor are isotopically similar to the depleted mantle Twohigh-Mg eclogites and one low-Mg eclogite have unradio-genic Nd one high-Mg eclogite has radiogenic Nd and

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1862

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 5 Individual and averaged Sr Nd and Hf isotope ratios obtained from garnet and cpx solutions

Sample 87Sr86Sr 2SE2SD 143Nd144Nd 2SE2SD eNd 2SE2SD 176Hf177Hf 2SE2SD eHf 2SE2SD

Low-Mg eclogites

VR43469 cp 0702661 0000010 0514844 0000006 43 01 0286130 0000012 119 04

VR43469 gt 0515508 0000014 56 03 0286506 0000012 132 04

VR43469 gt r 0515453 0000010 55 02 0286548 0000011 134 04

VR43469 gt r 0515384 0000007 54 01 0287244 0000010 158 04

VR43469 gt r 0515487 0000014 56 03 0286996 0000014 149 05

VR43469avg gt 0515458 0000109 55 21 0286823 0000715 143 253

WR VR43469 0515059 0000273 47 5 0286405 0001079 128 38

YK1926 cp 0703546 0000011 0513043 0000007 8 0 0283154 0000009 13 03

YK1926 gt 0513364 0000008 14 01 0295718 0000005 458 02

WR YK1926 0513120 0000026 9 1 0285980 0000716 113 25

YK1943 cp 0512143 0000007 10 01 0282213 0000005 20 02

YK1943 gt 0514461 0000012 36 02 0287510 0000008 168 03

YK1943 gt r 0514491 0000009 36 02 0287332 0000005 161 02

YK1943 gt r 0514517 0000010 37 02 0287290 0000007 160 02

YK1943avg gt 0514489 0000056 36 11 0287377 0000233 163 82

WR YK1943 0512625 0000228 0 5 0283138 0000879 13 31

High-Mg eclogites

VR43480 gt 0520965 0000017 162 03 0284950 0000010 77 04

VR67360 cp 0283509 0000074 26 26

VR67360 cp r 0283433 0000031 23 11

VR67360avg cp 0701662 0000018 0511660 0000010 19 02 0283471 0000108 25 38

VR67360 gt 0512984 0000101 7 20 0287429 0000180 165 64

WR VR67360 0511761 0000132 17 3 0284734 0000469 69 17

YK1949 cp 0703019 0000013 0512231 0000009 8 02 0282513 0000009 9 03

YK1949 gt 0513511 0000010 17 02 0290297 0000042 266 15

YK1949 gt r 0513508 0000012 17 02 0289947 0000015 254 05

YK1949 gt r 0513552 0000014 18 03

YK1949avg gt 0513523 0000049 17 10 0291569 0000494 311 175

WR YK1949 0512446 0000103 4 2 0284622 0000847 65 30

Volatile-rich eclogites

VR67112b gt 0522790 0000008 198 02 0354367 0000036 2532 13

YK1911 cp 0751186 0000009 0512165 0000007 9 01 0281263 0000009 53 03

YK1911 gt 0525721 0000013 255 03 0288051 0000012 187 04

WR YK1911 0513940 0000342 25 7 0281414 0000018 48 1

Pyroxenites

YK1914 cp 0703015 0000034 0513283 0000041 13 08

YK1914 cp r 0702956 0000005 0513200 0000027 11 05

YK1914 cp r 0513527 0000059 17 11

YK1914 cp r 0513045 0000010 8 02

YK1914avg cp 0702986 0000083 0513264 0000402 12 78

YK1915 gt 0514003 0000011 27 02 0283958 0000021 42

Analyses of full procedural replicates are designated with an lsquorrsquo after sample name SE standard error (internal errorsingle analyses) SD standard deviation (external error averaged multiple analyses) Italics used for reconstructed whole-rocks (mineral isotopic compositions and concentrations weighted by modes) Uncertainties for reconstructed whole-rocksinclude 2SE or 2SD uncertainties on each mineralrsquos isotope composition combined with 20 uncertainty in the amountof Nd or Hf contributed by the constituent minerals eNd and eHf are per 10 000 deviation from model chondritic mantle[model of Wasserburg et al (1981) and Blichert-Toft amp Albarede (1997) respectively]

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1863

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

εHf

εNd

Ekati eclogitesminus100

0

100

200

300

400

500

minus40 0 40 80 120 160 200 240 280

gt in YK1915

gt in VR67112bεNd=198 εHf=2532

minus40

0

40

80

120

160

minus40 minus20 0 20 40 60

minus60

minus40

minus20

0

20

40

070 071 072 073 074 075 076

North Americankimberlite field

εNd

87Sr86Sr

cpx in YK1914

Ekatieclogites

DM

DM

DM

DM

(a)

(b)

pyroxenites

low-Mg

high-Mg

volatile-rich

eclogites

Diavik low-TA154 S kimberlite

cpx in YK1911

Fig 12 (a) Sr^Nd for cpx and (b) Nd-Hf isotope diagram for garnet cpx and reconstructed whole-rocks (symbols with bold outline) Finedashed lines in (b) connect data points from the same sample with garnets having more radiogenic isotopic compositions Uncertainties forreconstructed whole-rocks combine 20 modal uncertainty with uncertainties on each mineralrsquos isotope composition Error bars on individualminerals may be smaller than the symbols Also shown are the isotopic compositions of the depleted mantle (DM DePaolo 1981 Griffin et al2000 Bell amp Tilton 2002) Range of 87Sr86Sr in eclogites from Ekati (Lac de Gras area 100 km north of Diavik Jacob et al 2003) in NorthAmerican kimberlites and compositions of low-temperature eclogites from Diavik and of Lac de Gras A154 South kimberlite fromSchmidberger et al (2007)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1864

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

one low-Mg eclogite has eNd similar to the depleted mantle(Fig 12b)Clinopyroxene in the phlogopitized eclogite YK1911

has unradiogenic Nd and Hf and strongly radiogenicSr whereas the coexisting garnet has radiogenic Nd andHf Although this sample contains 50 vol phlogopiteits whole-rock isotopic composition was calculated

from garnet and cpx alone because the contributionfrom phlogopite to the Hf and Nd budget is 52 and51 respectively (unpublished data) The whole-rockhas radiogenic Nd but has lower eHf than anyother sample analysed in accord with the extreme Hfabundance (53 ppm) and low LuHf of the cpx

DISCUSS IONOrigins of eclogites beneath the SlaveCratonHigh-Mg and high-Ca eclogites

With regard to their major- and trace-element contentsthe high-Mg and high-Ca eclogites from Lac de Gras arebroadly similar to variably evolved oceanic gabbros(Fig 13b and c) which may indicate an origin as subductedoceanic crust Compatible trace-element abundancesare useful to deduce potential crustal protoliths as theseelements are less susceptible to post-formation modification(Jacob 2004) The flat HREE patterns in the high-Mg andhigh-Ca eclogites are compatible with a low-pressureorigin as the oceanic crust is a product of moderate degreesof partial melting mainly within the spinel stability field(eg Presnall et al 2002) whereas melting with residualgarnet would qualitatively be expected to produce stronglyfractionated REE patterns Positive Sr and Pb anomaliesalthough compatible with a plagioclase signature couldalso be due to cryptic metasomatism (Dawson 1984) ofthe eclogites during their residence in the lithosphericmantle These anomalies are probably not related to kim-berlite infiltration as the whole-rock eclogite compositionswere reconstructed from analyses of fresh mineral coresLow ZrSm similar to that in primitive gabbros is

observed for many of the high-Mg and high-Caeclogites (Fig 13b and c) and is independent of the ratioof garnetcpx and whether or not the whole-rock wasreconstructed with rutile (Electronic Appendix 10) Thenegative trend of Mg-number against Yb in garnet(Fig 14) argues against igneous processes involving garnetaccumulation (Taylor amp Neal 1989) as suggested forexample for some high-Mg eclogites from West Africa(Barth et al 2002b) Rather the anti-correlation suggeststhat olivine fractionation played a role in the formation ofthe protoliths Fractionating olivine would excludeYb (Dolivinebasalt

frac14 002 Nikogosian amp Sobolev 1997)and incorporate Mg relative to Fe (Dfrac1415 and 05 respec-tively Taura et al1998) leading to progressive depletion ofMg and enrichment of Yb in the residual melt and henceanti-correlated Mg and Yb in the whole-rock This isinherited after eclogitization All eclogite types (exceptVR43452) plot along the olivine fractionation trend sug-gesting that they formed by a similar processHigh-Mg eclogites have lower

PREE than high-Ca

eclogites and some appear to have positive Eu anomalies

high-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

gabbrosSE Indian ridge

gabbrosOman ophiolite

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

high-Mg

RockNMORB(a)

VR43452

YK1949

low-Mg

boninite-like melts

01

001

1

10

(b)

01

001

1

10

(c)

01

001

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

Fig 13 Trace-element patterns of (a) low-Mg (b) high-Mg and(c) high-Ca eclogites compared with boninite-like volcanic rocksfrom the Isua greenstone belt (Polat et al 2002) gabbros from theOman ophiolite (Benoit et al 1996) and the SE Indian Ridge (Hartet al 1999) normalized to N-MORB of Sun amp McDonough (1989)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1865

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(EuEu up to 14) similar to more primitive cumulates(troctolite gabbros and olivine gabbros) reported in the lit-erature (Fig 13b) Although the sizes of these Eu anomaliesare not outside the analytical uncertainty their correlationwith CeYb in reconstructed eclogites supports their signif-icance (r2frac14066 nfrac14 4) The more primitive nature ofhigh-Mg eclogites is consistent with higher Mg-numbersrelative to high-Ca eclogites and may be due to olivineaccumulation during formation of their protolithsThe two samples with the lowest HREE abundanceshave the highest EuEu which is compatible with varia-tions in plagioclase modesMore evolved protoliths with higher pyroxeneplagio-

clase plus trapped melt could account for small or absentpositive Eu anomalies and higher

PHREE of high-Ca

eclogites (Fig 13c) Although the reason for the frequentassociation of graphite or diamond with high-Ca eclogitesis unclear (Aulbach et al 2003) the frequent occurrence ofkyanite in high-Ca eclogites also observed at Ekati justnorth of Lac de Gras has been argued to support a crustalprotolith (Jacob 2004) Kyanite andor corundum havebeen suggested to exsolve from high-temperature high-Alclinopyroxenes (Caporuscio amp Smyth 1990) Eclogitesfrom Lac de Gras show no exsolution textures suggestiveof such an origin although recrystallization can obscureor destroy microstructural evidence for exsolution(eg Griffin et al 1984)Modelling shows that qualitatively plagioclase accumu-

lation from a low-pressure melt derived from MORBmantle can account for positive Pb Sr and Eu and negativeZr anomalies relative to N-MORB whereas cpx accumula-tion or trapped melt increases Zr Y MREE and HREE

abundances (Fig 15) Some melt addition (for exampleby entrapment) seems to be required to increase all trace-element abundances to those observed for the eclogitesAlthough high-Ca and high-Mg eclogites could representshallow and deep portions of the same oceanic crust thehigher average temperatures obtained for the high-Caeclogites the frequent occurrence of carbon in high-Caeclogites and its absence in high-Mg eclogites the strongLREEHREE fractionation observed for six of seven high-Ca eclogites and the narrow range of compositions of thehigh-Mg eclogites suggest that they may have protolithsthat are not cogenetic Some discrepancies in particularaffecting the incompatible elements may be due to post-crystallization processes acting upon the protoliths suchas dehydration melting and later metasomatism that isevident in peridotites from the same locality (Griffin et al1999)Low K2OTh for high-Ca and high-Mg eclogites

(average 08 Table 4) relative to assumed crustal

0

2

4

6

8

10

30 40 50 60 70 80

Mg

VR43452

Yb

GARNET

pyroxenites

low-Mghigh-Mghigh-Ca

eclogites

VR19673ecl-g2

Fig 14 Garnet Mg-number vs Yb (ppm including 2s uncertainty)Lines qualitatively show separate correlations for eclogites (exceptoutlierVR43452) and pyroxenites (except outlierVR19673ecl-g2)

high-Ca

high-Mg

001

01

1

10

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

melt08 melt02 cryst cpx01 melt09 cryst

plag

(a)

001

01

1

10(b)

RockNMORB

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 15 Reconstructed (a) high-Ca and (b) high-Mg eclogitescompared with compositions of plagioclase and cpx crystallizingfrom a MORB melt modelled using distribution coefficients ofNorman et al (2005)The MORB was calculated as a 10 equilibriummelt of MORB pyrolite [composition from Kelemen et al (1993)except Pb Pr Sr Y and Lu which were calculated as residues from1 fractional melting of pyrolite from McDonough amp Sun (1995)using melt modes of Kinzler (1997) and distribution coefficients ofNorman et al (2005)] Also shown are mixtures of crystals (consistingof 50 plagioclase plus 50 cpx) and melt in the proportions 0901and 0208

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1866

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

precursors (eg gabbros SE Indian Ridge average 22Hart et al 1999) may reflect the preferred loss of fluid-mobile elements during dehydration (Becker et al 2000)although large uncertainties on these low-abundance ele-ments in the constituent minerals and reconstructedwhole-rocks makes this interpretation less than robustThe REE patterns of most high-Mg eclogites are too flatto represent melting residues as partial melting shouldproduce low LREEHREE (Electronic Appendix 10) Incontrast the marked LREE depletion of five of sevenhigh-Ca eclogites (LaLufrac14 02^08) contrasts with that ofassumed evolved gabbroic protoliths (LaLufrac1415^32Hart et al 1999) and can be modelled as a consequence ofpartial melting during subduction after eclogitization (iein the presence of garnet which retains HREE relative toLREE eg Barth et al 2002a) has taken place (Aulbachet al 2003)Our results are compatible with previous studies on

eclogite xenoliths from the Slave Craton includingthose from the same kimberlite that have linked thesesamples or some of these samples to Paleoproterozoicsubduction events (the 195^191Ga formation of theHottah terrane and the 18Ga formation of the GreatBear magmatic arc) in the Slave craton (Griffin et al1999 Kopylova et al 1999 Heaman et al 2002 2006Schmidberger et al 2005 2007)The compositional features discussed above although

compatible with crustal protoliths are not necessarilydefinitive of a subduction origin However if direct crystal-lization from basaltic melts at high pressure (ie mantledepths) is invoked instead specific conditions would haveto be met to satisfy the observations The Mg-number ineclogites from Lac de Gras are lower than those expectedof primary high-pressure mantle melts (eg Walter 1998OrsquoHara amp Herzberg 2002) requiring crystallization athigh pressures after olivine fractionation This originwould further require shielding from peridotitic wallrockwith which the fractionated melts would no longer be inequilibrium (Barth et al 2001) In such a setting the lowrelative abundances of fluid-mobile elements comparedwith those of less fluid-mobile elements and of LREErelative to HREE (Table 4) would reflect an expulsion offluids during crystallization of the protoliths and latermelt mobilization respectively The eclogites clearly donot represent garnet plus cpx cumulates which wouldrequire a positive correlation of Mg-number with Yb thatis at variance with the observed negative correlation ineclogitic garnets from this study (Fig 14)

Low-Mg eclogites

Garnets in low-Mg eclogites compositionally overlap withthose in massive eclogites from Jericho (NW of the Lacde Gras kimberlites) that were suggested to have a high-pressure intrusive origin based on the similarity of theirmajor-element compositions to those of Group A eclogites

fromsouthernAfrica forwhichamantleoriginhasbeenpro-posed (Kopylova et al1999) Garnets in low-Mg eclogites inthe present study have major-element compositions moresimilar to garnets in groupB rather thangroupAeclogitesThe positive slope in N-MORB-normalized HREE

abundances displayed by the low-Mg eclogites is clearlydistinct from other eclogite types and is similar to those ofboninite-like second-stage melts that are closely associatedwith ultramafic volcanics (Smithies et al 2004) (Fig 13a)The range of Mg-numbers in the low-Mg eclogites (55^66)overlaps with that of Archaean second-stage melts fromIsua (58^75) and Abitibi (52^82 Smithies et al 2004)These similarities might indicate that the low-Mg eclogiteshave an origin in a previously depleted mantle source suchas supra-subduction zone mantle Subchondritic ZrHf ofthe low-Mg eclogites contrasts with the supra-chondriticZrHf of the high-Mg and high-Ca eclogites (Aulbach et

al in preparation) and is consistent with previous cpx-controlled depletion in an arc mantle source This sourcecould have been subsequently remelted to generate theboninite-like precursor to the low-Mg eclogites as sug-gested for the protoliths of eclogites from Udachnaya(Jacob amp Foley 1999)Although some second-stage melts to which the low-Mg

eclogites are similar are closely associated with komatiitesand could contain a plume component the radiogenic ordepleted mantle-like Nd and Hf and unradiogenic Sr inthe low-Mg eclogites do not support plume involvement

Trace-element and Sr^Nd^Hf isotopeconstraints on post-formationmetasomatismSeveral low-Mg and high-Ca eclogites show negativeslopes in the LREE which may be indicative of interactionwith small amounts of LREE-enriched melt (Navon ampStolper 1987) amounting to cryptic metasomatism(ie not accompanied by crystallization of new phases) inthe nomenclature of Dawson (1984) High Nb abundancesrelative to elements of similar compatibility displayed bysome low-Mg and high-Ca eclogites (Fig 10) could resultfrom overestimation of rutile modes However high-CaeclogiteVR40345 is rutile-free and has been reconstructedwithout rutile its high Nb abundance is accompanied byLREE enrichment indicating that metasomatic additionmay be responsibleLow-Mg eclogites show a positive correlation of La with

Nb (Electronic Appendix 11a) indicating concomitantaddition of LREE and Nb This suggests that the metaso-matic agent was a solute-rich fluid or silicate silicic melt ashydrous fluids and carbonatites do not carry enough HFSEto affect the budget of eclogites (Green amp Wallace 1988Klemme et al 1995 Adam et al 1997 Stalder et al 1998)Of the samples that are enriched in Nb all but one (high-Ca eclogiteVR50909) show a simultaneous enrichment in

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1867

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Hf (Electronic Appendix 11b) Together these trace-element relationships suggest that both SmNd and LuHfwere lowered in these samples Their low parent^daughterratios contrast with higher ratios in unenriched samplesand resulted in the ingrowth of a range of Nd and Hf iso-tope compositions Compared with the low-Mg and high-Ca eclogites the high-Mg eclogites have a narrow range ofLa and Nb concentrations which correlates with a narrowrange in Nd and Hf isotope compositions (Fig 12)Precise model ages cannot be calculated because parent^

daughter ratios are available only from in situ analyseswhich have relatively large uncertainties Howeverevidence for long-term enrichment and depletion may bediscerned Low-Mg eclogites even those that are enrichedin LREE have relatively high SmNd as a result of havingoverall positive slopes in their REE patterns relative toMORB whereas the high-Mg eclogites have relativelyflat REE patterns and therefore lower SmNd (Fig 11)This translates into lower eNd for high-Mg eclogites thanfor low-Mg eclogites In contrast they show no systematicdifference with respect to 176Hf177Hf Because the low- andhigh-Mg eclogites are proposed to have different origins asdiscussed above and therefore did not necessarily have thesame initial isotopic composition and formation age theseisotopic differences cannot be used to estimate their agesA long-term coupling of trace-element and isotopic

abundances is also indicated by the fact that low-Mg eclo-gite YK1943 which has experienced Nb and LREE addi-tion has lower eNd (0) and eHf (thorn13) than low-Mg eclogiteVR43469 with higher LuHf and SmNd (eNdfrac14thorn47eHffrac14thorn128) If low-Mg eclogites YK1943 and VR43469originally had identical initial Nd and Hf isotope composi-tions and if the parent^daughter ratios are accurate andwere modified only once during their evolution shortlyafter metamorphism (when their isotopic compositionwould have been reset) this overprint can be calculatedfor both isotopic systems to have occurred at 17GaThis would suggest that Hf and Nd were added duringthe same event In contrast high-Mg eclogites VR67360and YK1949 would have had identical Ndisotope compositions at 10Ga whereas their Hf isotopecompositions are disturbed in that the sample with thehigher parent^daughter ratio has a lower 176Hf177Hf thanthe sample with the lower parent^daughter ratio Theseresults suggest that the LREE and HFSE enrichmentobserved in different samples occurred during separateevents that did not affect the eclogites equallyExtreme HFSE enrichment leading to the growth of

zircon and rutile and LREE addition leading to thegrowth of apatite in eclogites from the Jericho kimberlitein the northern Slave craton have been ascribed to twoseparate metasomatic events at around 18Ga (quasi-coincident with the age of formation and metamorphism)and 13^10Ga respectively (Heaman et al 2002 2006

Schmidberger et al 2005) The authors associated theseevents with Paleoproterozoic subduction and aMesoproterozoic thermal disturbance in the Slave litho-spheric mantle possibly linked to the Mackenzie igneousevent Mesoproterozoic Nd model ages mostly between13 and 14Ga were reported for eclogites from thenearby Ekati kimberlites (Jacob et al 2003) Although wedo not observe zircon or apatite in volatile-free eclogitesfrom Lac de Gras and the age constraints obtained fromour data are weak they support a link between theseevents described for other eclogite populations and themetasomatic enrichment observed in our studyThe Sr isotope data in the present study are similar to

those of Schmidberger et al (2007) who suggested that thedepleted-mantle-like values are consistent with derivationof the protoliths from an oceanic mantle source However87Sr86Sr in volatile-free eclogites in the present study isremarkably constant considering their Nd and Hf isotopevariability (Fig 12) and the variability of trace-elementpatterns (Fig 10) Given the evidence for dehydration andmelting in many of the eclogites fractionation of highlyincompatible Rb from Sr and differential ingrowth ofradiogenic Sr would be expectedThis may indicate a rela-tively recent overprint and considering the similarityof 87Sr86Sr to present-day depleted mantle which is signif-icantly lower than 87Sr86Sr of North American kimber-lites a juvenile (asthenospheric) source

PyroxenitesPyroxenites from Lac de Gras often show relationshipsin their major-element contents that are different fromthose observed in the eclogites For example pyroxeniticgarnets show a negative correlation between CaO andNa2O contents whereas these oxides are positively corre-lated for garnets in eclogites (Fig 5) CaO is well correlatedwith C2O3 in pyroxenitic garnets but not in eclogitic gar-nets (Fig 4) Mg-number is positively correlated withYb abundance for pyroxenitic garnets but anti-correlatedfor eclogitic garnets (Fig 14) This suggests that the pyrox-enites in the present study are not related to the eclogitesby fractionation or accumulation processes as proposedfor the origin of some pyroxenites (Foley et al 2003)Pyroxenites have also been interpreted as the products

of crystal segregation in magma channels with alteredoceanic crust as the magma source this would explaintheir Eu anomalies and stable and radiogenic isotopecompositions (Pearson et al 1993 Becker 1996) As thehigh-pressure liquidus phases of most terrestrial basaltsare cpx and garnet (OrsquoHara amp Herzberg 2002) partialmelts of eclogites with basaltic composition wouldcrystallize cpx and garnet but not opx More than half ofthe pyroxenites from Lac de Gras contain opx suggestingthat they probably did not crystallize directly fromeclogite-derived melts

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1868

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Low-Cr websterites and pyroxenites could be crystalli-zation products of mafic melts that interacted with perido-titic mantle thus becoming more refractory The positivecorrelation of Yb and Mg-number in garnet would allowan origin by fractional crystallization from such meltsAlternatively they formed by reaction of silicic (possiblyeclogite-derived) melts with mantle peridotites(eg Kelemen et al 1993 Aulbach et al 2002) This agreeswith the similarity of hypothetical melts in equilibriumwith pyroxenitic garnets in most samples to eclogite-derived melts (Electronic Appendix 12) Niobium concen-trations in all hypothetical melts are much higher than ineclogite-derived melts regardless of whether or not rutileis assumed to be in the residue The same is true for LaCe and Zr concentrations in some samples which mayindicate that NbZr and LREE were added to thepyroxenites some time after their formation as discussedin the previous section for some eclogites Because mini-mum ages for pyroxenitic rutile with an unradiogenicHf isotope composition (calculated by projecting the176Hf177Hf to the model depleted mantle isotopic evolutioncurve assuming zero ingrowth) range to 178066Ga(Aulbach et al in preparation) it appears likely that thepyroxenites were not formed recently (eg during kimber-lite-related or precursory melt migration) and that some ofthe samples are at least MesoproterozoicGarnet pyroxeniteYK1952 consists of schlieren of equili-

brated garnet pyroxenite in a lsquomatrixrsquo of spongy cpxfine cpx^opx intergrowths embayed garnet and meltpatches (Fig 2f) Garnet in the pyroxenite schlieren hasvery low LREE and HFSE abundances Orthopyroxenein the matrix has distinctly higher Mg-number andCr-number and contains more CaO than opx from thepyroxenite schlieren whereas opx rims in the schlierenare compositionally similar to opx in the matrix The con-siderable compositional and textural disequilibrium inthis sample suggests that this is a young feature causedby reaction with fluids shortly before or at the time ofentrainmentAs a consequence of the small sample sizes of most of

the pyroxenites studied here a comprehensive isotopicdataset is not available Clearly more such data areneeded in addition to stable isotope data to obtain betterconstraints on the origins of this rock type Consideringthe microstructural and compositional variability ofthe pyroxenites and major-element relationships oppositeto those observed for eclogites it appears certain that pyr-oxenites form by a variety of processes some of which areunrelated to the formation of eclogites and their protoliths

Volatile-rich eclogitesVolatile-rich eclogites from Lac de Gras display the mostextreme Hf^Nd isotope systematics of all the analysedsamples Garnets in phlogopitized (YK1911) and in cal-cite-bearing eclogite (VR67112b) have the most radiogenic

Nd (eNdfrac14 2486 and 1927 respectively) Garnet inVR67112b has extremely high eHf (2516) Clinopyroxene inYK1911 has the lowest eHf measured for any mineralseparate from Lac de Gras (53) in accord with the veryhigh Hf abundance in its cpx (5 ppm) In contrastgarnet in this sample has radiogenic Hf (eHffrac14 136) Thisdiscrepancy is at least in part due to their relatively lowequilibration temperatures (7698C and 8558C respec-tively) which allowed the isotopic compositions of cpxand garnet to evolve separately at their low andhigh parent^daughter element ratios respectively forextended periods of time and possibly since formationAccordingly the cpx in YK1911 gives a minimum Hf ageof 27GaSlab-derived fluids could have been introduced during

the 27Ga collision of the western Slave Craton domainincluding the thick plume-derived deep lithosphericmantle layer and the eastern domain with its highlydepleted lithosphere This would also be consistent withthe observation that volatile-rich eclogites occur at shallowdepths in the mantle (TKroghfrac14 685^8558C correspondingto depths of 85^115 km along a 40 mWm2 geotherm)where the bulk of dehydration and production of solute-rich melts in a subducting slab might be expected to occur(Becker et al 2000 Hermann et al 2006) Their negativeEu anomalies might indicate that the fluid was derivedfrom the shallow portion of the slab that experiencedplagioclase fractionation The presence of carbonate inVR67112b requires a carbonated sourceHigh 87Sr86Sr in cpx in YK1911 is probably the conse-

quence of isotopic equilibration with high-RbSr phlogo-pite after long-term radiogenic ingrowth while Hfaddition led to retarded ingrowth of radiogenic HfThese systematics combined with radiogenic Nd appear

inconsistent with the involvement of sediments in the pet-rogenesis of this sample (eg Perfit et al 1980 Vervoortet al 1999) or a hydrous fluid such as those responsible forMARID metasomatism in the Kaapvaal craton (Konzettet al 1998)Whatever the origin of the fluids the apparentHf mobility suggests that they were not in equilibriumwith a rutile-bearing source and also that they weresolute-rich in nature (ie hydrous melts able to dissolveHFSE) and the relatively high HREE abundances of thevolatile-rich eclogites indicate a garnet-free source suchas amphibolite This allowed evolution of the sample atrelatively high time-integrated SmNd

SUMMARY AND CONCLUSIONS

(1) The Lac de Gras kimberlites carried abundant eclo-gite and pyroxenite xenoliths to the surface Threemain eclogite types are distinguished in terms ofmajor-element composition and trace element

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1869

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

patterns high-Mg high-Ca and low-Mg eclogitesThe garnets of all eclogite types have anti-correlatedMg-number and Yb pointing to olivine fractionationas a controlling factor in the formation of their pre-cursor rocks Their flat or positively sloped HREEpatterns qualitatively indicate that the melt parentalto the precursor rocks was not in equilibrium with agarnet-bearing residue

(2) Reconstructed high-Mg eclogites and high-Caeclogites have flat HREE patterns positive Pb andSr anomalies and low ZrSm similar to oceanic gab-bros reported in the literature High-Mg eclogiteshave variably low HREE contents and some havesmall positive Eu anomalies possibly suggestive ofmore primitive relatively plagioclase-rich protolithswhereas high-Ca eclogites have higher HREE abun-dances and small or absent Eu anomalies that requiremore cpx- andor melt-rich protoliths if a subductionorigin is assumed

(3) Low-Mg eclogites have distinctive trace-element pat-terns that show a marked positive slope in the HREEand subchondritic ZrHf They resemble someArchaean second-stage melts (lsquoboninitesrsquo) and theirprotoliths appear to have formed from a previouslydepleted source

(4) After formation the eclogites were subject to a varietyof processes Most have low ratios of fluid-mobile vsless fluid-mobile elements possibly as a result of dehy-dration of the protoliths In addition some of the low-Mg and high-Ca eclogites have low LREEHREEconsistent with partial melt lossWhereas the low-Mgeclogites show concomitant enrichment in La Nb andZr some high-Mg and high-Ca eclogites show evi-dence for variable and unsystematic (decoupled)LREE and Nb addition

(5) Isotopically the trace-element characteristics of thedifferent eclogite types translate into lower eNd forhigh-Mg eclogites (negative slope in MREEfrac14 lowerSmNd) than for low-Mg eclogites (positive slope inMREEfrac14higher SmNd) Within the low-Mg groupsamples that show evidence for metasomatism (lowLuHf and SmNd) have lower eNd and eHf than asample that was apparently not metasomatized BothNd and Hf addition can be dated to 17Ga if identi-cal initial isotope compsitions are assumed for thelow-Mg eclogites In contrast the relatively constantand depleted mantle-like 87Sr86Sr of eclogitic cpxdespite variable Nd and Hf isotope compositions andparent^daughter ratios suggests recent input of juve-nile material

(6) Garnet and cpx in pyroxenites show major-elementrelationships different from those in eclogitesTherefore these two rock types are probably notrelated by fractionation processes The presence of

opx precludes direct crystallization from slab-derivedmelts Pyroxenites may be the product of (slab-derived) melts hybridized during assimilation of peri-dotitic mantle

(7) Garnets in phlogopite- and carbonate-bearing eclo-gites have extreme isotopic compositions includingthe most radiogenic Nd and Hf in the datasetNegative Eu anomalies may point to the involvementof a crustal protolith Highly radiogenic Sr radiogenicNd and unradiogenic Hf in the reconstructedphlogopite-bearing sample combined with highHREE abundances and evidence for Hf mobilitymay indicate involvement of solute-rich fluids origi-nating in a garnet- and rutile-free source

ACKNOWLEDGEMENTSHelp in the clean laboratories and with the analyticalfacilities by S Elhlou C Lawson A Sharma P Wielandand M Zhang is greatly appreciated Bill Griffin isthanked for his support and for agreeing to reviewseveral versions of the manuscript despite disagreeingwith many of the conclusions presented here The paperbenefited greatly from the detailed and constructivereviews provided by the editors Gareth Davies andMarjorie Wilson and by Matthias Barth Larry Heamanand an anonymous reviewer This work was funded by aMacquarie University International PostgraduateAward and Postgraduate Research Fund (SA) by anARC SPIRT grant sponsored by Kennecott Canada Incand by ARC Large and Discovery Grants to SYOrsquoRAnalytical data were obtained using instrumentationfunded by ARC LIEF and DEST Systemic InfrastructureGrants industry partners and Macquarie UniversityThis is Publication 481 from the ARC National KeyCentre for Geochemical Evolution and Metallogeny ofContinents (wwwesmqeduauGEMOC)

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online

REFERENCESAdam J GreenT H Sie S H amp Ryan C G (1997) Trace element

partitioning between aqueous fluids silicate melts and mineralsEuropeanJournal of Mineralogy 9 569^584

Aulbach S Stachel T Viljoen K S Brey G P amp Harris J W(2002) Eclogitic and websteritic diamond sources beneath theLimpopo Beltccedilis slab-melting the link Contributions to Mineralogy

and Petrology 143 56^70Aulbach S GriffinW L Pearson N J OrsquoReilly S Y Kivi K amp

Doyle B (2003) Origins of eclogites beneath the central SlaveCraton Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA-0011

Aulbach S Griffin W L Pearson N J OrsquoReilly S Y amp Kivi K(2004a) Mantle formation and evolution Slave Craton constraints

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1870

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

from HSE abundances and Re^Os isotope systematics of sulfideinclusions in mantle xenocrysts Chemical Geology 208 61^88

Aulbach S Griffin W L OrsquoReilly S Y amp McCandless T E(2004b) Genesis and evolution of the lithospheric mantle beneaththe Buffalo HeadTerrane Alberta (Canada) Lithos 77 413^451

Aulbach S GriffinW L Pearson N J OrsquoReilly S Y amp Doyle B(2007) Lithosphere formation in the central Slave Craton(Canada) plume subcretion or lithosphere accretion Contributionsto Mineralogy and Petrology (in press) doi 101007s00410-007-0200-1

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2001)Geochemistry of xenolithic eclogites from West Africa Part IA link between low MgO eclogites and Archean crust formationGeochimica et Cosmochimica Acta 65 1499^1527

Barth M G Foley S F amp Horn I (2002a) Partial melting inArchean subduction zones constraints from experimentally deter-mined trace element partition coefficients between eclogitic miner-als and tonalitic melts under upper mantle conditions PrecambrianResearch 113 323^340

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2002b)Geochemistry of xenolithic eclogites from West Africa part 2Origins of the high MgO eclogites Geochimica et Cosmochimica Acta

66 4325^4345Beard B L Fraracci K NTaylor L A Snyder G A Clayton R

A Mayeda T K amp Sobolev NV (1996) Petrography andgeochemistry of eclogites from the Mir kimberliteYakutia RussiaContributions to Mineralogy and Petrology 125 293^310

Becker H (1996) Geochemistry of garnet peridotite massifs fromlower Austria and the composition of deep lithosphere beneatha Palaeozoic convergent plate margin Chemical Geology 134 49^65

Becker H Jochum K P amp Carlson R W (2000) Trace elementfractionation during dehydration of eclogites from high-pressureterranes and the implications for element fluxes in subductionzones Chemical Geology 163 65^99

Bell K amp Tilton G R (2002) Probing the mantle The story fromcarbonatites EOS Transactions American Geophysical Union 83275^277

Benoit M Polvecurren M amp Ceuleneer G (1996) Trace element andisotopic characterisation of mafic cumulates in a fossil mantlediapir (Oman ophiolite) Chemical Geology 134 199^214

BleekerW (2003) The late Archean record a puzzle in ca 35 piecesLithos 71 99^134

BleekerW Ketchum JW F JacksonV A amp Villeneuve M (1999)The central Slave Basement Complex Part I Its structural topol-ogy and autochthonous core Canadian Journal of Earth Sciences 361083^1109

Blichert-Toft J amp Albare de F (1997) The Lu^Hf isotope geochemis-try of chondrites and the evolution of the mantle^crust systemEarth and Planetary Science Letters 148 243^258

Blichert-Toft J Chauvel C amp Albare de F (1997) Separation of Hfand Lu for high precision isotope analysis of rock samples by mag-netic sector-multiple collector ICP-MS Contributions to Mineralogy

and Petrology 127 248^260Brey G P amp Kolaquo hler T (1990) Geothermobarometry in four-phase

lherzolites II New thermobarometers and practical assessment ofexisting thermobarometers Journal of Petrology 31 1353^1378

Brey G P Kolaquo hlerT amp Nickel K G (1990) Geothermobarometry infour-phase lherzolites I Experimental results from 10 to 60 kbJournal of Petrology 31 1313^1352

Caporuscio F amp Smyth J (1990) Trace element crystal chemistryof mantle eclogites Contributions to Mineralogy and Petrology 105550^561

Carbno G B amp Canil D (2002) Mantle structure beneath the SWSlave Craton Canada constraints from garnet geochemistry in theDrybones Bay Kimberlite Journal of Petrology 43 129^142

Coleman R G Lee E D Beatty L B amp Brannock W W (1965)Eclogites and eclogites their differences and similaritiesGeological Society of America Bulletin 76 483^508

Creaser R A Grulaquo tter H Carlson J amp Crawford B (2004)Macrocrystal phlogopite Rb^Sr dates for the Ekati propertykimberlites Slave Province Canada evidence for multiple intru-sive episodes in the Paleocene and Eocene Lithos 76 399^414

Davies R M GriffinW L Pearson N J Andrew A S Doyle BJ amp OrsquoReilly S Y (1999) Diamonds from the deep Pipe DO-27Slave Craton Canada In Gurney J J Gurney J L Pascoe MD amp Richardson S H (eds) Proceedings of the 7th International

Kimberlite Conference CapeTown Red Roof Design pp 148^155Davies R M Griffin W L OrsquoReilly S Y amp Doyle B J (2004)

Mineral inclusions and geochemical characteristics of microdia-monds from the DO27 A154 A21 A418 DO18 DD17 and RanchLake kimberlites at Lac de Gras Slave Craton Canada Lithos 7739^55

DavisW J amp Hegner E (1992) Neodymium isotopic evidence for thetectonic assembly of Late Archean crust in the Slave Provincenorthwest Canada Contributions to Mineralogy and Petrology 111493^504

Davis W J Jones A G Bleeker W amp Grutter H (2003)Lithosphere development in the Slave craton a linked crustal andmantle perspective Lithos 71 575^589

Dawson J B (1984) Contrasting types of upper-mantle metasoma-tism In Kornprobst J (ed) Kimberlites II The Mantle and

Crust^Mantle Relationships Amsterdam Elsevier pp 289^294DePaolo D J (1981) Neodymium isotopes in the Colorado Front

Range and crust^mantle evolution in the Proterozoic Nature 291193^196

Erlank A J amp Kushiro I (1970) Potassium contents of syntheticpyroxenes at high temperatures and pressures Carnegie Institution ofWashingtonYearbook 68 433^439

Foley S F Barth M G amp Jenner G A (2000) Rutilemelt partitioncoefficients for trace elements and an assessment of the influence ofrutile on the trace element characteristics of subduction zonemagmas Geochimica et Cosmochimica Acta 64 933^938

Foley S F Buhre S amp Jacob D E (2003) Evolution of the Archaeancrust by delamination and shallow subduction Nature 421 249^252

Graham I Burgess J L Bryan D Ravenscroft P J Thomas EDoyle B J Hopkins R amp Armstrong K A (1999) Explorationhistory and geology of the Diavik Kimberlites Lac de GrasNorthwest Territories Canada In Gurney J J Gurney J LPascoe M D amp Richardson S H (eds) Proceedings of the 7th

International Kimberlite Conference Cape Town Red Roof Designpp 262^279

Green D H amp Wallace M E (1988) Mantle metasomatism byephemeral carbonatite melts Nature 336 459^462

Green T H Blundy J D Adam J amp Yaxley G M (2000) SIMSdetermination of trace element partition coefficients betweengarnet clinopyroxene and hydrous basaltic liquids at 2^75 GPaand 1080^12008C Lithos 53 165^187

GriffinW L amp OrsquoReilly SY (2007) Cratonic lithospheric mantle Isanything subducted Episodes 30(1) 43^53

Griffin W L Wass S Y amp Hollis J D (1984) Ultramafic xenolithsfrom Bullenmerri and Gnotuk maars Victoria Australia petrol-ogy of a sub-continental crust^mantle transition Journal of

Petrology 25 53^87GriffinW L Doyle B J Ryan C G Pearson N J OrsquoReilly SY

Davies R Kivi K van Achterbergh E amp Natapov LM (1999)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1871

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Layered mantle lithosphere in the Lac de Gras Area SlaveCraton composition structure and origin Journal of Petrology 40705^727

Griffin W L Pearson N J Belousova E Jackson S E vanAchterbergh E OrsquoReilly SY amp Shee S R (2000)The Hf isotopecomposition of cratonic mantle LAM-MC-ICPMS analysis ofzircon megacrysts in kimberlites Geochimica et Cosmochimica Acta

64 133^147Griffin W L OrsquoReilly S Y Doyle B J Pearson N J

Coopersmith H Kivi K Malkovets V amp Pokhilenko N(2004) Lithosphere mapping beneath the North American plateLithos 77 873^922

Grulaquo tter H S Apter D B amp Kong J (1999) Crust^mantle couplingevidence from mantle-derived xenocrystic garnets In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 307^312

Hart S R Blusztajn J Dick H J B Meyer P S ampMuehlenbachs K (1999) The fingerprint of seawater circulationin a 500-meter section of ocean crust gabbros Geochimica et

Cosmochimica Acta 63 4059^4080Harte B amp Kirkley M B (1997) Partitioning of trace elements

between clinopyroxene and garnet Data from mantle eclogitesChemical Geology 136 1^24

Heaman L M Creaser R A amp Cookenboo H O (2002) Extremeenrichment of high field strength elements in Jericho eclogitexenoliths A cryptic record of Paleoproterozoic subduction partialmelting and metasomatism beneath the Slave craton CanadaGeology 30 507^510

Heaman L M Kjarsgaard B A amp Creaser R A (2004) The tem-poral evolution of North American kimberlites Lithos 76 377^397

Heaman L M Creaser R A Cookenboo H O amp Chacko T(2006) Multi-stage modification of the northern Slave mantlelithosphere Evidence from zircon- and diamond-bearing eclogitexenoliths entrained in Jericho Kimberlite Journal of Petrology 47821^858

Hermann J Cpandler C Hack A amp Korsakov A (2006) Aqueousfluids and hydrous melts in high-pressure and ultra-high pressurerocks implications for element transfer in subduction zones Lithos92 399^417

Hills DV amp Haggerty S E (1989) Petrochemistry of eclogites fromthe Koidu Kimberlite Complex Sierra Leone Contributions to

Mineralogy and Petrology 103 397^422Hoffman P F (1989) Precambrian geology and tectonic history of

North America In Bally A amp Palmer A (eds) The Geology of

North AmericaccedilAn Overview Boulder CO Geological Society ofAmerica pp 447^512

IrelandT R Rudnick R L amp Spetsius Z (1994) Trace-elements indiamond inclusions from eclogites reveal link to Archean granitesEarth and Planetary Science Letters 128 199^213

Irvine G J Pearson D G Kjarsgaard B A Carlson R WKopylova M G amp Dreibus G (2003) Evolution of the litho-spheric mantle beneath Northern Canada a Re^Os isotope andPGE study of kimberlite-derived peridotite xenoliths fromSomerset Island and a comparison to the Slave and Kaapvaalcratons Lithos 71 461^488

Isachsen C E amp Bowring S A (1994) Evolution of the Slave cratonGeology 22 917^920

Jacob D E (2004) Nature and origin of eclogite xenoliths from kim-berlites Lithos 77 295^316

Jacob D E amp Foley S F (1999) Evidence for Archean ocean crustwith low high field strength element signature from diamondifer-ous eclogite xenoliths Lithos 48 317^336

Jacob D Jagoutz E Lowry D Mattey D amp Kudrjavtseva G(1994) Diamondiferous eclogites from Siberiaccedilremnants ofArchean oceanic-crust Geochimica et Cosmochimica Acta 585191^5207

Jacob D E Fung A Jagoutz E amp Pearson D G (2003) Petrologyand geochemistry of eclogite xenoliths from the Ekati kimberlitesarea Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA_239

Jacob D E Bizimis M amp Salters V J M (2005) Lu^Hf and geo-chemical systematics of recycled ancient oceanic crust evidencefrom Roberts Victor eclogites Contributions to Mineralogy and

Petrology 148 707^720Jerde E A Taylor L A Crozaz G Sobolev NV amp SobolevV N

(1993) Diamondiferous eclogites from Yakutia Siberiaccedilevidencefor a diversity of protoliths Contributions to Mineralogy and Petrology

114 189^202Jones A G Ferguson I J Chave A D Evans R L amp

McNeice GW (2001) The electric lithosphere of the Slave cratonGeology 29 423^426

Kelemen P B Shimizu N amp Dunn T (1993) Relative depletionof niobium in some arc magmas and the continental-crust-partitioning of K Nb La and Ce during meltrockreaction in the upper-mantle Earth and Planetary Science Letters 120111^134

Kinzler R J (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition Application to mid-ocean ridge basalt petrogenesis Journal of Geophysical Research 102853^874

Klemme S van der Laan S R Foley S F amp Guenther D (1995)Experimentally determined trace and minor element partitioningbetween clinopyroxene and carbonatite melt under upper mantleconditions Earth and Planetary Science Letters 133 439^448

Konzett J Armstrong R A Sweeney R J amp CompstonW (1998)The timing of MARID metasomatism in the Kaapvaal mantle anion probe study of zircons from MARID xenoliths Earth and

Planetary Science Letters 160 133^145Kopylova M G Russell J K amp Cookenboo H (1999) Mapping the

lithopshere beneath the north central Slave Craton In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 468^479

Krogh E (1988) The garnet^clinopyroxene iron^magnesiumgeothermometerccedila reinterpretation of existing experimentaldata Contributions to Mineralogy and Petrology 99 44^48

Kusky T M (1989) Accretion of Archean Slave Province Geology 1763^67

LeCheminant A N amp Heaman L M (1989) Mackenzie igneousevents Canada Middle Proterozoic hotspot magmatismassociated with ocean opening Earth and Planetary Science Letters

96 38^48LeCheminant A N Richardson D G DiLabio R N W amp

Richardson K A (1996) La recherche de diamants au CanadaGeological Survey of Canada Open File 3228 268pp

MacGregor I D amp Carter J L (1970) The chemistry of clinopyrox-enes and garnets of eclogite and peridotite xenoliths from theRoberts Victor mine South Africa Physics of the Earth and PlanetaryInteriors 3 391^397

MacGregor I D amp Manton W I (1986) Roberts Victor eclogitesancient oceanic crust Journal of Geophysical Research 91 14063^14079

MacKenzie J M amp Canil D (1999) Composition and thermal evo-lution of cratonic mantle beneath the central Archean SlaveProvince NWT Canada Contributions to Mineralogy and Petrology

134 313^324

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1872

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

McCandlessT E amp Gurney J J (1989) Sodium in garnet and potas-sum in clinopyroxene criteria for classifying mantle eclogites InRoss J (ed) Kimberlites and Related Rocks Carlton Blackwellpp 827^832

McDonoughW F amp Sun S-S (1995) The composition of the EarthChemical Geology 120 223^253

Navon O amp Stolper E (1987) Geochemical consequences of meltpercolation the upper mantle as a chromatographic columnJournal of Geology 95 285^307

Nikogosian I K amp Sobolev AV (1997) Ion-microprobe analysis ofmelt inclusions in olivine experience in estimating the olivine^meltpartition coefficients of trace elements Geochemistry International 35119^126

Norman M D Griffin W L Pearson N J Garcia M O ampOrsquoReilly S Y (1998) Quantitative analysis of trace element abun-dances in glasses and minerals a comparison of laser ablationinductively coupled plasma mass spectrometry solution inductivelycoupled plasma mass spectrometry proton microprobe and elec-tron microprobe data Journal of Analytical Atomic Spectrometry 13477^482

Norman M D Garcia M O amp Pietruszka A J (2005) Trace-element distribution coefficients for pyroxenes plagioclase andolivine in evolved tholeiites from the 1955 eruption of KilaueaVolcano Hawairsquoi and petrogenesis of differentiated rift-zonelavas American Mineralogist 90 888^899

OrsquoHara M J amp Herzberg C (2002) Interpretation of trace elementand isotope features of basalts Relevance of field relations petrol-ogy major element data phase equilibria and magma chambermodeling in basalt petrogenesis Geochimica et Cosmochimica Acta 662167^2191

PadghamW A amp FysonW K (1992) The Slave Province a distinctcraton CanadianJournal of Earth Sciences 29 2072^2086

Pearson D G Davies G R amp Nixon P H (1993) Geochemical con-straints on the petrogenesis of diamond facies pyroxenites from theBeni Bousera peridotite massif North Morocco Journal of Petrology34 125^172

Pearson N J Griffin W L Doyle B J OrsquoReilly S Y vanAchterbergh E amp Kivi K (1999) Xenoliths from kimberlitepipes of the Lac de gras area Slave Craton Canada In GurneyJ J Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite ConferenceCapeTown Red Roof Design pp 644^658

Perfit N R Gust D A Bence A E Arculus R J amp Taylor S R(1980) Chemical characteristics of island-arc basalts implicationsfor mantle sources Chemical Geology 30 227^256

Polat A Hofmann A W amp Rosing M T (2002) Boninite-likevolcanic rocks in the 37^38 Ga Isua greenstone belt WestGreenland geochemical evidence for intra-oceanic subductionzone processes in the early Earth Chemical Geology 184 231^254

Poudjom DjomaniY H GriffinW L OrsquoReilly SY amp Doyle B J(2005) Lithospheric domains and controls on kimberlite emplace-ment Slave Province Canada evidence from elastic thickness andupper mantle composition Geochemistry Geophysics Geosystems 6doi1010292005GC0009978

Presnall D C Gudfinnsson G H amp Walter M J (2002)Generation of mid-ocean ridge basalts at pressures from 1 to 7GPa Geochimica et Cosmochimica Acta 66 2073^2090

Schmidberger S S Heaman L M Simonetti A Creaser R A ampCookenboo H O (2005) Formation of Paleoproterozoiceclogitic mantle Slave Province (Canada) Insights from in-situHf and U^Pb isotopic analyses of mantle zircons Earth and

Planetary Science Letters 240 621^633

Schmidberger S S Simonetti A Heaman L M Creaser R A ampWhiteford S (2007) Lu^Hf in-situ Sr and Pb isotope and traceelement systematics for mantle eclogites from the Diavik diamondmine Evidence for Paleoproterozoic subduction beneath the Slavecraton Canada Earth and Planetary Science Letters 254 55^68

Schulze D J amp Helmstaedt H (1988) Coesite-sanidine eclogitesfrom kimberlite products of mantle fractionation or subductionJournal of Geology 96 435^443

Smithies R H Champion D C amp Sun S S (2004) The case forArchaean boninites Contributions to Mineralogy and Petrology 147705^721

Smyth J R Caporuscio F A amp McCormick T C (1989) Mantleeclogites evidence of igneous fractionation in the mantle Earth andPlanetary Science Letters 93 133^141

Sobolev NV amp Lavrentrsquoyev Y G (1971) Isomorphic sodium admix-ture in garnets formed at high pressure Contributions to Mineralogy

and Petrology 31 1^12Stalder R Foley S F Brey G P amp Horn I (1998) Mineral aqueous

fluid partitioning of trace elements at 900^12008C and 30^57GPa New experimental data for garnet clinopyroxene and rutileand implications for mantle metasomatism Geochimica et

Cosmochimica Acta 62 1781^1801Sun S-S amp McDonough W F (1989) Chemical and isotopic sys-

tematics of oceanic basalts implications for mantle compositionsand processes In Saunders A D amp Norris M J (eds)Magmatism in the Ocean Basins Geological Society London Special

Publications 42 313^345Taura H Yurimoto H Kurita K amp Sueno S (1998) Pressure

dependence on partition coefficients for trace elements between oli-vine and the coexisting melts Physics and Chemistry of Minerals 25469^484

Taylor L A amp Neal C R (1989) Eclogites with oceanic crustal andmantle signatures from the Bellsbank kimberlite South AfricaPart I Mineralogy petrography and whole rock chemistryJournal of Geology 97 551^567

van Achterbergh E Ryan C G amp Griffin W L (1999) Glitter online intensity reduction for the laser ablation inductively coupledplasma spectrometry Abstract 9th V M GoldschmidtConference LPI Boston MA 905 pp Abstract no 7215

van Breemen O DavisW J amp King J E (1992) Temporal distribu-tion of granitoid plutonic rocks in the Archean Slave ProvinceNorthwest Canadian Shield Canadian Journal of Earth Sciences 292186^2199

Vervoort J D Patchett P J Blichert-Toft J amp Albare de F (1999)Relationship between Lu^Hf and Sm^Nd isotopic systems in theglobal sedimentary system Earth and Planetary Science Letters 16879^99

Viljoen K S Smith C B amp Sharp Z D (1996) Stable and radio-genic isotope study of eclogite xenoliths from the Orapa kimberliteBotswana Chemical Geology 131 235^255

Walter M J (1998) Melting of garnet peridotite and theorigin of komatiite and depleted lithosphere Journal of Petrology 3929^60

Wasserburg G J Jacobsen S B DePaolo D J McCulloch M Tamp Wen T (1981) Precise determination of SmNd ratios Sm andNd isotopic abundances in standard solutions Geochimica et

Cosmochimica Acta 45 2311^2323Westerlund K J Shirey S B Richardson S H Carlson R W

Gurney J J amp Harris JW (2006) A subduction wedge origin forPaleoarchean peridotitic diamonds and harzburgites from thePanda kimberlite Slave craton evidence from Re^Os isotope sys-tematics Contributions to Mineralogy and Petrology 152 275^294

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1873

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Page 18: Origins of Xenolithic Eclogites and Pyroxenites from the Central

Table 4 Continued

Sample Sm Eu EuEu Gd Dy Ho Er Yb Lu Hf Pb Th LaLu K2OTh

vr43465 22 09 12 25 26 06 15 14 02 10 04 001 25 040

vr50858 20 08 09 33 41 08 22 21 03 10 01 nd 02 na

vr50860 14 07 11 27 37 08 22 18 03 06 00 nd 06 na

vr50909 20 10 12 33 38 07 20 20 03 16 01 002 02 181

yk3528 08 05 11 20 30 07 18 18 02 05 01 nd 08 na

Volatile-rich eclogites

vr67112b 38 10 07 53 74 15 41 34 04 18 03 012 60 004

yk1911 20 07 08 35 39 07 18 17 02 10 14 nd 73 na

Pyroxenites

yk1914 02 01 11 04 08 02 05 06 01 01 03 003 68 101

aGarnet has CandashMgndashFe relationship typical of high-Mg eclogites but whole-rock trace-element pattern is typical oflow-Mg eclogitesbGarnet has CandashMgndashFe relationship typical of low-Mg eclogites but whole-rock trace-element pattern is typical ofhigh-Mg eclogites

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

(a)

001

01

1

10(c)

001

01

10(b)

VR43452

YK1949

RockNMORB

low-Mg

RockNMORB

high-Mg

(d)

volatile-richhigh-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

phlogopite-bearingYK1911

calcite-bearingVR67112b

001

01

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 10 Extended trace-element patterns of reconstructed (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-bearing eclogites Normalized to N-MORB of Sun amp McDonough (1989)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1860

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

RockNMORB

(b)

RockNMORB

(d)

010

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu001

01

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

(a)

001

01

1

10(b)

001

01

1

10(c)

001

01

1

10(e)

high-TSchmidberger et al

low-TSchmidberger et al

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

high-Mglow-Mg

high-Cavolatile-rich

phlogopite-bearingYK1911

calcite-bearingVR67112b

YK1949VR43479

Fig 11 N-MORB normalized REE patterns for (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-richeclogites Shown for comparison in (e) are reconstructed eclogites from Schmidberger et al (2007)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1861

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Both volatile-rich whole rocks display strong negativeSr anomalies Strontium can be sited in calcite or apatiteandor indicate the presence of a Sr-rich mineral that wasnot identified in thin section such as bariteFour of the six reconstructed whole-rocks in the study

of Schmidberger et al (2007) (Fig 11e) show markedLREEHREEN enrichment that is observed only in onehigh-Ca eclogite in our study (Fig 11c) suggesting thatmany of the eclogites they studied were affected by metaso-matic LREE enrichment Two samples in their study thatare not affected by this enrichment show positive Euanomalies similar to the high-Mg eclogites in this studywhereas the distinctive REE pattern of the low-Mg eclo-gites is not recognized among their samples

SR^ND^HF I SOTOPESYSTEMATICSStrontium Nd and Hf isotope data for optically puremineral separates and reconstructed whole-rocks aregiven in Table 5 Spongy rims and alteration are likely tobe mostly removed during leaching in 6N HCl in combina-tion with repeated ultrasonication Despite this rigoroustreatment the reproducibility of Nd in garnet VR43469and in cpx YK1914 is worse than for other samples Thismay be due to inhomogeneities related to late Nd diffusionpossibly from grain boundaries containing a kimberlitecomponent The reproducibility of the 176Hf177Hf data isgenerally poorer than that of Nd which may reflect a com-bination of matrix effects caused by imperfectly purifiedsolutions and sample heterogeneity Parent^daughterratios are available only from in situ analyses which arenot precise enough to allow meaningful model age orinitial isotope ratio calculations (ie at the time of kimber-lite eruption) Because the Lac de Gras kimberlites arerelatively young (55^56 Ma Graham et al 1999) age cor-rection would result in only a minor change in isotopecomposition which is inconsequential in view of the rangeof compositions displayed by most samples and hence hasno effect on the conclusions drawnClinopyroxenes in five volatile-free eclogites and a

pyroxenite have Sr isotope ratios ranging from 0701662 to0703546 this range is similar to or somewhat lower thanthat of the depleted mantle at the time of kimberliteemplacement (0703 Bell amp Tilton 2002) Clinopyroxenein the phlogopitized sample has a distinctly high 87Sr86Srof 0751186Clinopyroxene controls 56^92 of the Nd budget (Nd

concentration in cpx weighted by mode) Neodymiumisotope ratios range from 0511660 to 0514844 Thiscorresponds to eNd variation (per 10 000 deviationof 143Nd144Nd in the sample from 143Nd144Nd in the chon-dritic mantle model reservoir) from 17 to thorn43 cpx inthe low-Mg eclogites have higher eNd than those in

high-Mg eclogites Garnets have generally have higher143Nd144Nd from 0512984 (eNdfrac14thorn 7) to 0520965(eNdfrac14thorn 162) Garnets in the two volatile-rich sampleshave the highest ratios 0522790 for garnet in the calcite-bearing sample VR67112b (eNdfrac14thorn198) and 0525721 inthe phlogopitized eclogiteYK1911 (eNdfrac14thorn 255)Garnet contributes 18^32 and cpx 48^78 of the Hf

budget The proportion controlled by rutile is 4^30(Aulbach et al in preparation) Measured 176Hf177Hf incpx ranges from 0281263 in phlogopite-bearing sampleYK1911 (eHffrac1453) to 0286130 in low-Mg eclogiteVR43469 (eHffrac14thorn119) and in garnet from 0283958 inpyroxeniteYK1915 (eHffrac14thorn42) to 0295718 in low-Mg eclo-gite YK1926 (eHffrac14thorn458) Hafnium in garnet in the phlo-gopite^apatite-bearing sample is extremely radiogenic(176Hf177Hffrac14 0354367) corresponding to an eHf of thorn2532which is close to a value determined for a garnet separatefrom a Roberts Victor eclogite (eHffrac14thorn2561 Jacob et al2005) Garnet in the calcite-bearing sample has extremelyradiogenic Nd and Hf (eNdfrac14thorn193 eHffrac14thorn2516) Garnetin pyroxenite YK1915 also has radiogenic Nd and Hf(eNdfrac14thorn27 eHffrac14thorn42)Despite the large uncertainties for reconstructed whole-

rocks (calculated as the square root of the sum of squareduncertainties of two standard errors for garnet and cpxanalyses combined with 20 propagated uncertainty onthe proportion of Nd or Hf respectively contributed byeach mineral) the isotopic data are consistent in thatthey correlate with the independently determinedparent^daughter ratios (1) Low-Mg eclogites have higherSmNd reflected in the steeper positive slope of theirN-MORB-normalized REE patterns (Fig 11a) comparedwith the high-Mg eclogites (Fig 11b) Accordingly recon-structed low-Mg eclogites have higher present-day eNd

(0 to thorn47) than high-Mg eclogites (4 and 17) (2) Low-Mg eclogite YK1943 has very high Nb and LREE abun-dances that could have been added during metasomatismthat also enhanced its Hf abundance This sample has thelowest eNd and eHf of the low-Mg eclogites (0 and thorn20respectively) In contrast apparently unenriched low-Mgeclogites have more radiogenic Nd and Hf (eNdfrac14thorn47and thorn9 eHffrac14thorn128 and thorn113 respectively)The Sr^Nd isotope diagram reveals that eNd in cpx

varies from 19 to 43 at nearly constant depleted-mantle-like 87Sr86Sr (Fig 12a) In the Nd^Hf isotope diagram(Fig 12b) the reconstructed whole-rocks lie on curvedmixing lines if the lsquomixingrsquo proportions of Nd and Hffor the two lsquoend-memberrsquo minerals are very different(eg YK1911) Garnets have more radiogenic Nd andHf than coexisting cpx All reconstructed volatile-freeeclogites plot in the radiogenic field with regard to Hfor are isotopically similar to the depleted mantle Twohigh-Mg eclogites and one low-Mg eclogite have unradio-genic Nd one high-Mg eclogite has radiogenic Nd and

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1862

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 5 Individual and averaged Sr Nd and Hf isotope ratios obtained from garnet and cpx solutions

Sample 87Sr86Sr 2SE2SD 143Nd144Nd 2SE2SD eNd 2SE2SD 176Hf177Hf 2SE2SD eHf 2SE2SD

Low-Mg eclogites

VR43469 cp 0702661 0000010 0514844 0000006 43 01 0286130 0000012 119 04

VR43469 gt 0515508 0000014 56 03 0286506 0000012 132 04

VR43469 gt r 0515453 0000010 55 02 0286548 0000011 134 04

VR43469 gt r 0515384 0000007 54 01 0287244 0000010 158 04

VR43469 gt r 0515487 0000014 56 03 0286996 0000014 149 05

VR43469avg gt 0515458 0000109 55 21 0286823 0000715 143 253

WR VR43469 0515059 0000273 47 5 0286405 0001079 128 38

YK1926 cp 0703546 0000011 0513043 0000007 8 0 0283154 0000009 13 03

YK1926 gt 0513364 0000008 14 01 0295718 0000005 458 02

WR YK1926 0513120 0000026 9 1 0285980 0000716 113 25

YK1943 cp 0512143 0000007 10 01 0282213 0000005 20 02

YK1943 gt 0514461 0000012 36 02 0287510 0000008 168 03

YK1943 gt r 0514491 0000009 36 02 0287332 0000005 161 02

YK1943 gt r 0514517 0000010 37 02 0287290 0000007 160 02

YK1943avg gt 0514489 0000056 36 11 0287377 0000233 163 82

WR YK1943 0512625 0000228 0 5 0283138 0000879 13 31

High-Mg eclogites

VR43480 gt 0520965 0000017 162 03 0284950 0000010 77 04

VR67360 cp 0283509 0000074 26 26

VR67360 cp r 0283433 0000031 23 11

VR67360avg cp 0701662 0000018 0511660 0000010 19 02 0283471 0000108 25 38

VR67360 gt 0512984 0000101 7 20 0287429 0000180 165 64

WR VR67360 0511761 0000132 17 3 0284734 0000469 69 17

YK1949 cp 0703019 0000013 0512231 0000009 8 02 0282513 0000009 9 03

YK1949 gt 0513511 0000010 17 02 0290297 0000042 266 15

YK1949 gt r 0513508 0000012 17 02 0289947 0000015 254 05

YK1949 gt r 0513552 0000014 18 03

YK1949avg gt 0513523 0000049 17 10 0291569 0000494 311 175

WR YK1949 0512446 0000103 4 2 0284622 0000847 65 30

Volatile-rich eclogites

VR67112b gt 0522790 0000008 198 02 0354367 0000036 2532 13

YK1911 cp 0751186 0000009 0512165 0000007 9 01 0281263 0000009 53 03

YK1911 gt 0525721 0000013 255 03 0288051 0000012 187 04

WR YK1911 0513940 0000342 25 7 0281414 0000018 48 1

Pyroxenites

YK1914 cp 0703015 0000034 0513283 0000041 13 08

YK1914 cp r 0702956 0000005 0513200 0000027 11 05

YK1914 cp r 0513527 0000059 17 11

YK1914 cp r 0513045 0000010 8 02

YK1914avg cp 0702986 0000083 0513264 0000402 12 78

YK1915 gt 0514003 0000011 27 02 0283958 0000021 42

Analyses of full procedural replicates are designated with an lsquorrsquo after sample name SE standard error (internal errorsingle analyses) SD standard deviation (external error averaged multiple analyses) Italics used for reconstructed whole-rocks (mineral isotopic compositions and concentrations weighted by modes) Uncertainties for reconstructed whole-rocksinclude 2SE or 2SD uncertainties on each mineralrsquos isotope composition combined with 20 uncertainty in the amountof Nd or Hf contributed by the constituent minerals eNd and eHf are per 10 000 deviation from model chondritic mantle[model of Wasserburg et al (1981) and Blichert-Toft amp Albarede (1997) respectively]

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1863

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

εHf

εNd

Ekati eclogitesminus100

0

100

200

300

400

500

minus40 0 40 80 120 160 200 240 280

gt in YK1915

gt in VR67112bεNd=198 εHf=2532

minus40

0

40

80

120

160

minus40 minus20 0 20 40 60

minus60

minus40

minus20

0

20

40

070 071 072 073 074 075 076

North Americankimberlite field

εNd

87Sr86Sr

cpx in YK1914

Ekatieclogites

DM

DM

DM

DM

(a)

(b)

pyroxenites

low-Mg

high-Mg

volatile-rich

eclogites

Diavik low-TA154 S kimberlite

cpx in YK1911

Fig 12 (a) Sr^Nd for cpx and (b) Nd-Hf isotope diagram for garnet cpx and reconstructed whole-rocks (symbols with bold outline) Finedashed lines in (b) connect data points from the same sample with garnets having more radiogenic isotopic compositions Uncertainties forreconstructed whole-rocks combine 20 modal uncertainty with uncertainties on each mineralrsquos isotope composition Error bars on individualminerals may be smaller than the symbols Also shown are the isotopic compositions of the depleted mantle (DM DePaolo 1981 Griffin et al2000 Bell amp Tilton 2002) Range of 87Sr86Sr in eclogites from Ekati (Lac de Gras area 100 km north of Diavik Jacob et al 2003) in NorthAmerican kimberlites and compositions of low-temperature eclogites from Diavik and of Lac de Gras A154 South kimberlite fromSchmidberger et al (2007)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1864

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

one low-Mg eclogite has eNd similar to the depleted mantle(Fig 12b)Clinopyroxene in the phlogopitized eclogite YK1911

has unradiogenic Nd and Hf and strongly radiogenicSr whereas the coexisting garnet has radiogenic Nd andHf Although this sample contains 50 vol phlogopiteits whole-rock isotopic composition was calculated

from garnet and cpx alone because the contributionfrom phlogopite to the Hf and Nd budget is 52 and51 respectively (unpublished data) The whole-rockhas radiogenic Nd but has lower eHf than anyother sample analysed in accord with the extreme Hfabundance (53 ppm) and low LuHf of the cpx

DISCUSS IONOrigins of eclogites beneath the SlaveCratonHigh-Mg and high-Ca eclogites

With regard to their major- and trace-element contentsthe high-Mg and high-Ca eclogites from Lac de Gras arebroadly similar to variably evolved oceanic gabbros(Fig 13b and c) which may indicate an origin as subductedoceanic crust Compatible trace-element abundancesare useful to deduce potential crustal protoliths as theseelements are less susceptible to post-formation modification(Jacob 2004) The flat HREE patterns in the high-Mg andhigh-Ca eclogites are compatible with a low-pressureorigin as the oceanic crust is a product of moderate degreesof partial melting mainly within the spinel stability field(eg Presnall et al 2002) whereas melting with residualgarnet would qualitatively be expected to produce stronglyfractionated REE patterns Positive Sr and Pb anomaliesalthough compatible with a plagioclase signature couldalso be due to cryptic metasomatism (Dawson 1984) ofthe eclogites during their residence in the lithosphericmantle These anomalies are probably not related to kim-berlite infiltration as the whole-rock eclogite compositionswere reconstructed from analyses of fresh mineral coresLow ZrSm similar to that in primitive gabbros is

observed for many of the high-Mg and high-Caeclogites (Fig 13b and c) and is independent of the ratioof garnetcpx and whether or not the whole-rock wasreconstructed with rutile (Electronic Appendix 10) Thenegative trend of Mg-number against Yb in garnet(Fig 14) argues against igneous processes involving garnetaccumulation (Taylor amp Neal 1989) as suggested forexample for some high-Mg eclogites from West Africa(Barth et al 2002b) Rather the anti-correlation suggeststhat olivine fractionation played a role in the formation ofthe protoliths Fractionating olivine would excludeYb (Dolivinebasalt

frac14 002 Nikogosian amp Sobolev 1997)and incorporate Mg relative to Fe (Dfrac1415 and 05 respec-tively Taura et al1998) leading to progressive depletion ofMg and enrichment of Yb in the residual melt and henceanti-correlated Mg and Yb in the whole-rock This isinherited after eclogitization All eclogite types (exceptVR43452) plot along the olivine fractionation trend sug-gesting that they formed by a similar processHigh-Mg eclogites have lower

PREE than high-Ca

eclogites and some appear to have positive Eu anomalies

high-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

gabbrosSE Indian ridge

gabbrosOman ophiolite

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

high-Mg

RockNMORB(a)

VR43452

YK1949

low-Mg

boninite-like melts

01

001

1

10

(b)

01

001

1

10

(c)

01

001

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

Fig 13 Trace-element patterns of (a) low-Mg (b) high-Mg and(c) high-Ca eclogites compared with boninite-like volcanic rocksfrom the Isua greenstone belt (Polat et al 2002) gabbros from theOman ophiolite (Benoit et al 1996) and the SE Indian Ridge (Hartet al 1999) normalized to N-MORB of Sun amp McDonough (1989)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1865

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(EuEu up to 14) similar to more primitive cumulates(troctolite gabbros and olivine gabbros) reported in the lit-erature (Fig 13b) Although the sizes of these Eu anomaliesare not outside the analytical uncertainty their correlationwith CeYb in reconstructed eclogites supports their signif-icance (r2frac14066 nfrac14 4) The more primitive nature ofhigh-Mg eclogites is consistent with higher Mg-numbersrelative to high-Ca eclogites and may be due to olivineaccumulation during formation of their protolithsThe two samples with the lowest HREE abundanceshave the highest EuEu which is compatible with varia-tions in plagioclase modesMore evolved protoliths with higher pyroxeneplagio-

clase plus trapped melt could account for small or absentpositive Eu anomalies and higher

PHREE of high-Ca

eclogites (Fig 13c) Although the reason for the frequentassociation of graphite or diamond with high-Ca eclogitesis unclear (Aulbach et al 2003) the frequent occurrence ofkyanite in high-Ca eclogites also observed at Ekati justnorth of Lac de Gras has been argued to support a crustalprotolith (Jacob 2004) Kyanite andor corundum havebeen suggested to exsolve from high-temperature high-Alclinopyroxenes (Caporuscio amp Smyth 1990) Eclogitesfrom Lac de Gras show no exsolution textures suggestiveof such an origin although recrystallization can obscureor destroy microstructural evidence for exsolution(eg Griffin et al 1984)Modelling shows that qualitatively plagioclase accumu-

lation from a low-pressure melt derived from MORBmantle can account for positive Pb Sr and Eu and negativeZr anomalies relative to N-MORB whereas cpx accumula-tion or trapped melt increases Zr Y MREE and HREE

abundances (Fig 15) Some melt addition (for exampleby entrapment) seems to be required to increase all trace-element abundances to those observed for the eclogitesAlthough high-Ca and high-Mg eclogites could representshallow and deep portions of the same oceanic crust thehigher average temperatures obtained for the high-Caeclogites the frequent occurrence of carbon in high-Caeclogites and its absence in high-Mg eclogites the strongLREEHREE fractionation observed for six of seven high-Ca eclogites and the narrow range of compositions of thehigh-Mg eclogites suggest that they may have protolithsthat are not cogenetic Some discrepancies in particularaffecting the incompatible elements may be due to post-crystallization processes acting upon the protoliths suchas dehydration melting and later metasomatism that isevident in peridotites from the same locality (Griffin et al1999)Low K2OTh for high-Ca and high-Mg eclogites

(average 08 Table 4) relative to assumed crustal

0

2

4

6

8

10

30 40 50 60 70 80

Mg

VR43452

Yb

GARNET

pyroxenites

low-Mghigh-Mghigh-Ca

eclogites

VR19673ecl-g2

Fig 14 Garnet Mg-number vs Yb (ppm including 2s uncertainty)Lines qualitatively show separate correlations for eclogites (exceptoutlierVR43452) and pyroxenites (except outlierVR19673ecl-g2)

high-Ca

high-Mg

001

01

1

10

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

melt08 melt02 cryst cpx01 melt09 cryst

plag

(a)

001

01

1

10(b)

RockNMORB

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 15 Reconstructed (a) high-Ca and (b) high-Mg eclogitescompared with compositions of plagioclase and cpx crystallizingfrom a MORB melt modelled using distribution coefficients ofNorman et al (2005)The MORB was calculated as a 10 equilibriummelt of MORB pyrolite [composition from Kelemen et al (1993)except Pb Pr Sr Y and Lu which were calculated as residues from1 fractional melting of pyrolite from McDonough amp Sun (1995)using melt modes of Kinzler (1997) and distribution coefficients ofNorman et al (2005)] Also shown are mixtures of crystals (consistingof 50 plagioclase plus 50 cpx) and melt in the proportions 0901and 0208

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1866

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

precursors (eg gabbros SE Indian Ridge average 22Hart et al 1999) may reflect the preferred loss of fluid-mobile elements during dehydration (Becker et al 2000)although large uncertainties on these low-abundance ele-ments in the constituent minerals and reconstructedwhole-rocks makes this interpretation less than robustThe REE patterns of most high-Mg eclogites are too flatto represent melting residues as partial melting shouldproduce low LREEHREE (Electronic Appendix 10) Incontrast the marked LREE depletion of five of sevenhigh-Ca eclogites (LaLufrac14 02^08) contrasts with that ofassumed evolved gabbroic protoliths (LaLufrac1415^32Hart et al 1999) and can be modelled as a consequence ofpartial melting during subduction after eclogitization (iein the presence of garnet which retains HREE relative toLREE eg Barth et al 2002a) has taken place (Aulbachet al 2003)Our results are compatible with previous studies on

eclogite xenoliths from the Slave Craton includingthose from the same kimberlite that have linked thesesamples or some of these samples to Paleoproterozoicsubduction events (the 195^191Ga formation of theHottah terrane and the 18Ga formation of the GreatBear magmatic arc) in the Slave craton (Griffin et al1999 Kopylova et al 1999 Heaman et al 2002 2006Schmidberger et al 2005 2007)The compositional features discussed above although

compatible with crustal protoliths are not necessarilydefinitive of a subduction origin However if direct crystal-lization from basaltic melts at high pressure (ie mantledepths) is invoked instead specific conditions would haveto be met to satisfy the observations The Mg-number ineclogites from Lac de Gras are lower than those expectedof primary high-pressure mantle melts (eg Walter 1998OrsquoHara amp Herzberg 2002) requiring crystallization athigh pressures after olivine fractionation This originwould further require shielding from peridotitic wallrockwith which the fractionated melts would no longer be inequilibrium (Barth et al 2001) In such a setting the lowrelative abundances of fluid-mobile elements comparedwith those of less fluid-mobile elements and of LREErelative to HREE (Table 4) would reflect an expulsion offluids during crystallization of the protoliths and latermelt mobilization respectively The eclogites clearly donot represent garnet plus cpx cumulates which wouldrequire a positive correlation of Mg-number with Yb thatis at variance with the observed negative correlation ineclogitic garnets from this study (Fig 14)

Low-Mg eclogites

Garnets in low-Mg eclogites compositionally overlap withthose in massive eclogites from Jericho (NW of the Lacde Gras kimberlites) that were suggested to have a high-pressure intrusive origin based on the similarity of theirmajor-element compositions to those of Group A eclogites

fromsouthernAfrica forwhichamantleoriginhasbeenpro-posed (Kopylova et al1999) Garnets in low-Mg eclogites inthe present study have major-element compositions moresimilar to garnets in groupB rather thangroupAeclogitesThe positive slope in N-MORB-normalized HREE

abundances displayed by the low-Mg eclogites is clearlydistinct from other eclogite types and is similar to those ofboninite-like second-stage melts that are closely associatedwith ultramafic volcanics (Smithies et al 2004) (Fig 13a)The range of Mg-numbers in the low-Mg eclogites (55^66)overlaps with that of Archaean second-stage melts fromIsua (58^75) and Abitibi (52^82 Smithies et al 2004)These similarities might indicate that the low-Mg eclogiteshave an origin in a previously depleted mantle source suchas supra-subduction zone mantle Subchondritic ZrHf ofthe low-Mg eclogites contrasts with the supra-chondriticZrHf of the high-Mg and high-Ca eclogites (Aulbach et

al in preparation) and is consistent with previous cpx-controlled depletion in an arc mantle source This sourcecould have been subsequently remelted to generate theboninite-like precursor to the low-Mg eclogites as sug-gested for the protoliths of eclogites from Udachnaya(Jacob amp Foley 1999)Although some second-stage melts to which the low-Mg

eclogites are similar are closely associated with komatiitesand could contain a plume component the radiogenic ordepleted mantle-like Nd and Hf and unradiogenic Sr inthe low-Mg eclogites do not support plume involvement

Trace-element and Sr^Nd^Hf isotopeconstraints on post-formationmetasomatismSeveral low-Mg and high-Ca eclogites show negativeslopes in the LREE which may be indicative of interactionwith small amounts of LREE-enriched melt (Navon ampStolper 1987) amounting to cryptic metasomatism(ie not accompanied by crystallization of new phases) inthe nomenclature of Dawson (1984) High Nb abundancesrelative to elements of similar compatibility displayed bysome low-Mg and high-Ca eclogites (Fig 10) could resultfrom overestimation of rutile modes However high-CaeclogiteVR40345 is rutile-free and has been reconstructedwithout rutile its high Nb abundance is accompanied byLREE enrichment indicating that metasomatic additionmay be responsibleLow-Mg eclogites show a positive correlation of La with

Nb (Electronic Appendix 11a) indicating concomitantaddition of LREE and Nb This suggests that the metaso-matic agent was a solute-rich fluid or silicate silicic melt ashydrous fluids and carbonatites do not carry enough HFSEto affect the budget of eclogites (Green amp Wallace 1988Klemme et al 1995 Adam et al 1997 Stalder et al 1998)Of the samples that are enriched in Nb all but one (high-Ca eclogiteVR50909) show a simultaneous enrichment in

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1867

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Hf (Electronic Appendix 11b) Together these trace-element relationships suggest that both SmNd and LuHfwere lowered in these samples Their low parent^daughterratios contrast with higher ratios in unenriched samplesand resulted in the ingrowth of a range of Nd and Hf iso-tope compositions Compared with the low-Mg and high-Ca eclogites the high-Mg eclogites have a narrow range ofLa and Nb concentrations which correlates with a narrowrange in Nd and Hf isotope compositions (Fig 12)Precise model ages cannot be calculated because parent^

daughter ratios are available only from in situ analyseswhich have relatively large uncertainties Howeverevidence for long-term enrichment and depletion may bediscerned Low-Mg eclogites even those that are enrichedin LREE have relatively high SmNd as a result of havingoverall positive slopes in their REE patterns relative toMORB whereas the high-Mg eclogites have relativelyflat REE patterns and therefore lower SmNd (Fig 11)This translates into lower eNd for high-Mg eclogites thanfor low-Mg eclogites In contrast they show no systematicdifference with respect to 176Hf177Hf Because the low- andhigh-Mg eclogites are proposed to have different origins asdiscussed above and therefore did not necessarily have thesame initial isotopic composition and formation age theseisotopic differences cannot be used to estimate their agesA long-term coupling of trace-element and isotopic

abundances is also indicated by the fact that low-Mg eclo-gite YK1943 which has experienced Nb and LREE addi-tion has lower eNd (0) and eHf (thorn13) than low-Mg eclogiteVR43469 with higher LuHf and SmNd (eNdfrac14thorn47eHffrac14thorn128) If low-Mg eclogites YK1943 and VR43469originally had identical initial Nd and Hf isotope composi-tions and if the parent^daughter ratios are accurate andwere modified only once during their evolution shortlyafter metamorphism (when their isotopic compositionwould have been reset) this overprint can be calculatedfor both isotopic systems to have occurred at 17GaThis would suggest that Hf and Nd were added duringthe same event In contrast high-Mg eclogites VR67360and YK1949 would have had identical Ndisotope compositions at 10Ga whereas their Hf isotopecompositions are disturbed in that the sample with thehigher parent^daughter ratio has a lower 176Hf177Hf thanthe sample with the lower parent^daughter ratio Theseresults suggest that the LREE and HFSE enrichmentobserved in different samples occurred during separateevents that did not affect the eclogites equallyExtreme HFSE enrichment leading to the growth of

zircon and rutile and LREE addition leading to thegrowth of apatite in eclogites from the Jericho kimberlitein the northern Slave craton have been ascribed to twoseparate metasomatic events at around 18Ga (quasi-coincident with the age of formation and metamorphism)and 13^10Ga respectively (Heaman et al 2002 2006

Schmidberger et al 2005) The authors associated theseevents with Paleoproterozoic subduction and aMesoproterozoic thermal disturbance in the Slave litho-spheric mantle possibly linked to the Mackenzie igneousevent Mesoproterozoic Nd model ages mostly between13 and 14Ga were reported for eclogites from thenearby Ekati kimberlites (Jacob et al 2003) Although wedo not observe zircon or apatite in volatile-free eclogitesfrom Lac de Gras and the age constraints obtained fromour data are weak they support a link between theseevents described for other eclogite populations and themetasomatic enrichment observed in our studyThe Sr isotope data in the present study are similar to

those of Schmidberger et al (2007) who suggested that thedepleted-mantle-like values are consistent with derivationof the protoliths from an oceanic mantle source However87Sr86Sr in volatile-free eclogites in the present study isremarkably constant considering their Nd and Hf isotopevariability (Fig 12) and the variability of trace-elementpatterns (Fig 10) Given the evidence for dehydration andmelting in many of the eclogites fractionation of highlyincompatible Rb from Sr and differential ingrowth ofradiogenic Sr would be expectedThis may indicate a rela-tively recent overprint and considering the similarityof 87Sr86Sr to present-day depleted mantle which is signif-icantly lower than 87Sr86Sr of North American kimber-lites a juvenile (asthenospheric) source

PyroxenitesPyroxenites from Lac de Gras often show relationshipsin their major-element contents that are different fromthose observed in the eclogites For example pyroxeniticgarnets show a negative correlation between CaO andNa2O contents whereas these oxides are positively corre-lated for garnets in eclogites (Fig 5) CaO is well correlatedwith C2O3 in pyroxenitic garnets but not in eclogitic gar-nets (Fig 4) Mg-number is positively correlated withYb abundance for pyroxenitic garnets but anti-correlatedfor eclogitic garnets (Fig 14) This suggests that the pyrox-enites in the present study are not related to the eclogitesby fractionation or accumulation processes as proposedfor the origin of some pyroxenites (Foley et al 2003)Pyroxenites have also been interpreted as the products

of crystal segregation in magma channels with alteredoceanic crust as the magma source this would explaintheir Eu anomalies and stable and radiogenic isotopecompositions (Pearson et al 1993 Becker 1996) As thehigh-pressure liquidus phases of most terrestrial basaltsare cpx and garnet (OrsquoHara amp Herzberg 2002) partialmelts of eclogites with basaltic composition wouldcrystallize cpx and garnet but not opx More than half ofthe pyroxenites from Lac de Gras contain opx suggestingthat they probably did not crystallize directly fromeclogite-derived melts

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1868

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Low-Cr websterites and pyroxenites could be crystalli-zation products of mafic melts that interacted with perido-titic mantle thus becoming more refractory The positivecorrelation of Yb and Mg-number in garnet would allowan origin by fractional crystallization from such meltsAlternatively they formed by reaction of silicic (possiblyeclogite-derived) melts with mantle peridotites(eg Kelemen et al 1993 Aulbach et al 2002) This agreeswith the similarity of hypothetical melts in equilibriumwith pyroxenitic garnets in most samples to eclogite-derived melts (Electronic Appendix 12) Niobium concen-trations in all hypothetical melts are much higher than ineclogite-derived melts regardless of whether or not rutileis assumed to be in the residue The same is true for LaCe and Zr concentrations in some samples which mayindicate that NbZr and LREE were added to thepyroxenites some time after their formation as discussedin the previous section for some eclogites Because mini-mum ages for pyroxenitic rutile with an unradiogenicHf isotope composition (calculated by projecting the176Hf177Hf to the model depleted mantle isotopic evolutioncurve assuming zero ingrowth) range to 178066Ga(Aulbach et al in preparation) it appears likely that thepyroxenites were not formed recently (eg during kimber-lite-related or precursory melt migration) and that some ofthe samples are at least MesoproterozoicGarnet pyroxeniteYK1952 consists of schlieren of equili-

brated garnet pyroxenite in a lsquomatrixrsquo of spongy cpxfine cpx^opx intergrowths embayed garnet and meltpatches (Fig 2f) Garnet in the pyroxenite schlieren hasvery low LREE and HFSE abundances Orthopyroxenein the matrix has distinctly higher Mg-number andCr-number and contains more CaO than opx from thepyroxenite schlieren whereas opx rims in the schlierenare compositionally similar to opx in the matrix The con-siderable compositional and textural disequilibrium inthis sample suggests that this is a young feature causedby reaction with fluids shortly before or at the time ofentrainmentAs a consequence of the small sample sizes of most of

the pyroxenites studied here a comprehensive isotopicdataset is not available Clearly more such data areneeded in addition to stable isotope data to obtain betterconstraints on the origins of this rock type Consideringthe microstructural and compositional variability ofthe pyroxenites and major-element relationships oppositeto those observed for eclogites it appears certain that pyr-oxenites form by a variety of processes some of which areunrelated to the formation of eclogites and their protoliths

Volatile-rich eclogitesVolatile-rich eclogites from Lac de Gras display the mostextreme Hf^Nd isotope systematics of all the analysedsamples Garnets in phlogopitized (YK1911) and in cal-cite-bearing eclogite (VR67112b) have the most radiogenic

Nd (eNdfrac14 2486 and 1927 respectively) Garnet inVR67112b has extremely high eHf (2516) Clinopyroxene inYK1911 has the lowest eHf measured for any mineralseparate from Lac de Gras (53) in accord with the veryhigh Hf abundance in its cpx (5 ppm) In contrastgarnet in this sample has radiogenic Hf (eHffrac14 136) Thisdiscrepancy is at least in part due to their relatively lowequilibration temperatures (7698C and 8558C respec-tively) which allowed the isotopic compositions of cpxand garnet to evolve separately at their low andhigh parent^daughter element ratios respectively forextended periods of time and possibly since formationAccordingly the cpx in YK1911 gives a minimum Hf ageof 27GaSlab-derived fluids could have been introduced during

the 27Ga collision of the western Slave Craton domainincluding the thick plume-derived deep lithosphericmantle layer and the eastern domain with its highlydepleted lithosphere This would also be consistent withthe observation that volatile-rich eclogites occur at shallowdepths in the mantle (TKroghfrac14 685^8558C correspondingto depths of 85^115 km along a 40 mWm2 geotherm)where the bulk of dehydration and production of solute-rich melts in a subducting slab might be expected to occur(Becker et al 2000 Hermann et al 2006) Their negativeEu anomalies might indicate that the fluid was derivedfrom the shallow portion of the slab that experiencedplagioclase fractionation The presence of carbonate inVR67112b requires a carbonated sourceHigh 87Sr86Sr in cpx in YK1911 is probably the conse-

quence of isotopic equilibration with high-RbSr phlogo-pite after long-term radiogenic ingrowth while Hfaddition led to retarded ingrowth of radiogenic HfThese systematics combined with radiogenic Nd appear

inconsistent with the involvement of sediments in the pet-rogenesis of this sample (eg Perfit et al 1980 Vervoortet al 1999) or a hydrous fluid such as those responsible forMARID metasomatism in the Kaapvaal craton (Konzettet al 1998)Whatever the origin of the fluids the apparentHf mobility suggests that they were not in equilibriumwith a rutile-bearing source and also that they weresolute-rich in nature (ie hydrous melts able to dissolveHFSE) and the relatively high HREE abundances of thevolatile-rich eclogites indicate a garnet-free source suchas amphibolite This allowed evolution of the sample atrelatively high time-integrated SmNd

SUMMARY AND CONCLUSIONS

(1) The Lac de Gras kimberlites carried abundant eclo-gite and pyroxenite xenoliths to the surface Threemain eclogite types are distinguished in terms ofmajor-element composition and trace element

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1869

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

patterns high-Mg high-Ca and low-Mg eclogitesThe garnets of all eclogite types have anti-correlatedMg-number and Yb pointing to olivine fractionationas a controlling factor in the formation of their pre-cursor rocks Their flat or positively sloped HREEpatterns qualitatively indicate that the melt parentalto the precursor rocks was not in equilibrium with agarnet-bearing residue

(2) Reconstructed high-Mg eclogites and high-Caeclogites have flat HREE patterns positive Pb andSr anomalies and low ZrSm similar to oceanic gab-bros reported in the literature High-Mg eclogiteshave variably low HREE contents and some havesmall positive Eu anomalies possibly suggestive ofmore primitive relatively plagioclase-rich protolithswhereas high-Ca eclogites have higher HREE abun-dances and small or absent Eu anomalies that requiremore cpx- andor melt-rich protoliths if a subductionorigin is assumed

(3) Low-Mg eclogites have distinctive trace-element pat-terns that show a marked positive slope in the HREEand subchondritic ZrHf They resemble someArchaean second-stage melts (lsquoboninitesrsquo) and theirprotoliths appear to have formed from a previouslydepleted source

(4) After formation the eclogites were subject to a varietyof processes Most have low ratios of fluid-mobile vsless fluid-mobile elements possibly as a result of dehy-dration of the protoliths In addition some of the low-Mg and high-Ca eclogites have low LREEHREEconsistent with partial melt lossWhereas the low-Mgeclogites show concomitant enrichment in La Nb andZr some high-Mg and high-Ca eclogites show evi-dence for variable and unsystematic (decoupled)LREE and Nb addition

(5) Isotopically the trace-element characteristics of thedifferent eclogite types translate into lower eNd forhigh-Mg eclogites (negative slope in MREEfrac14 lowerSmNd) than for low-Mg eclogites (positive slope inMREEfrac14higher SmNd) Within the low-Mg groupsamples that show evidence for metasomatism (lowLuHf and SmNd) have lower eNd and eHf than asample that was apparently not metasomatized BothNd and Hf addition can be dated to 17Ga if identi-cal initial isotope compsitions are assumed for thelow-Mg eclogites In contrast the relatively constantand depleted mantle-like 87Sr86Sr of eclogitic cpxdespite variable Nd and Hf isotope compositions andparent^daughter ratios suggests recent input of juve-nile material

(6) Garnet and cpx in pyroxenites show major-elementrelationships different from those in eclogitesTherefore these two rock types are probably notrelated by fractionation processes The presence of

opx precludes direct crystallization from slab-derivedmelts Pyroxenites may be the product of (slab-derived) melts hybridized during assimilation of peri-dotitic mantle

(7) Garnets in phlogopite- and carbonate-bearing eclo-gites have extreme isotopic compositions includingthe most radiogenic Nd and Hf in the datasetNegative Eu anomalies may point to the involvementof a crustal protolith Highly radiogenic Sr radiogenicNd and unradiogenic Hf in the reconstructedphlogopite-bearing sample combined with highHREE abundances and evidence for Hf mobilitymay indicate involvement of solute-rich fluids origi-nating in a garnet- and rutile-free source

ACKNOWLEDGEMENTSHelp in the clean laboratories and with the analyticalfacilities by S Elhlou C Lawson A Sharma P Wielandand M Zhang is greatly appreciated Bill Griffin isthanked for his support and for agreeing to reviewseveral versions of the manuscript despite disagreeingwith many of the conclusions presented here The paperbenefited greatly from the detailed and constructivereviews provided by the editors Gareth Davies andMarjorie Wilson and by Matthias Barth Larry Heamanand an anonymous reviewer This work was funded by aMacquarie University International PostgraduateAward and Postgraduate Research Fund (SA) by anARC SPIRT grant sponsored by Kennecott Canada Incand by ARC Large and Discovery Grants to SYOrsquoRAnalytical data were obtained using instrumentationfunded by ARC LIEF and DEST Systemic InfrastructureGrants industry partners and Macquarie UniversityThis is Publication 481 from the ARC National KeyCentre for Geochemical Evolution and Metallogeny ofContinents (wwwesmqeduauGEMOC)

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online

REFERENCESAdam J GreenT H Sie S H amp Ryan C G (1997) Trace element

partitioning between aqueous fluids silicate melts and mineralsEuropeanJournal of Mineralogy 9 569^584

Aulbach S Stachel T Viljoen K S Brey G P amp Harris J W(2002) Eclogitic and websteritic diamond sources beneath theLimpopo Beltccedilis slab-melting the link Contributions to Mineralogy

and Petrology 143 56^70Aulbach S GriffinW L Pearson N J OrsquoReilly S Y Kivi K amp

Doyle B (2003) Origins of eclogites beneath the central SlaveCraton Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA-0011

Aulbach S Griffin W L Pearson N J OrsquoReilly S Y amp Kivi K(2004a) Mantle formation and evolution Slave Craton constraints

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1870

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

from HSE abundances and Re^Os isotope systematics of sulfideinclusions in mantle xenocrysts Chemical Geology 208 61^88

Aulbach S Griffin W L OrsquoReilly S Y amp McCandless T E(2004b) Genesis and evolution of the lithospheric mantle beneaththe Buffalo HeadTerrane Alberta (Canada) Lithos 77 413^451

Aulbach S GriffinW L Pearson N J OrsquoReilly S Y amp Doyle B(2007) Lithosphere formation in the central Slave Craton(Canada) plume subcretion or lithosphere accretion Contributionsto Mineralogy and Petrology (in press) doi 101007s00410-007-0200-1

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2001)Geochemistry of xenolithic eclogites from West Africa Part IA link between low MgO eclogites and Archean crust formationGeochimica et Cosmochimica Acta 65 1499^1527

Barth M G Foley S F amp Horn I (2002a) Partial melting inArchean subduction zones constraints from experimentally deter-mined trace element partition coefficients between eclogitic miner-als and tonalitic melts under upper mantle conditions PrecambrianResearch 113 323^340

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2002b)Geochemistry of xenolithic eclogites from West Africa part 2Origins of the high MgO eclogites Geochimica et Cosmochimica Acta

66 4325^4345Beard B L Fraracci K NTaylor L A Snyder G A Clayton R

A Mayeda T K amp Sobolev NV (1996) Petrography andgeochemistry of eclogites from the Mir kimberliteYakutia RussiaContributions to Mineralogy and Petrology 125 293^310

Becker H (1996) Geochemistry of garnet peridotite massifs fromlower Austria and the composition of deep lithosphere beneatha Palaeozoic convergent plate margin Chemical Geology 134 49^65

Becker H Jochum K P amp Carlson R W (2000) Trace elementfractionation during dehydration of eclogites from high-pressureterranes and the implications for element fluxes in subductionzones Chemical Geology 163 65^99

Bell K amp Tilton G R (2002) Probing the mantle The story fromcarbonatites EOS Transactions American Geophysical Union 83275^277

Benoit M Polvecurren M amp Ceuleneer G (1996) Trace element andisotopic characterisation of mafic cumulates in a fossil mantlediapir (Oman ophiolite) Chemical Geology 134 199^214

BleekerW (2003) The late Archean record a puzzle in ca 35 piecesLithos 71 99^134

BleekerW Ketchum JW F JacksonV A amp Villeneuve M (1999)The central Slave Basement Complex Part I Its structural topol-ogy and autochthonous core Canadian Journal of Earth Sciences 361083^1109

Blichert-Toft J amp Albare de F (1997) The Lu^Hf isotope geochemis-try of chondrites and the evolution of the mantle^crust systemEarth and Planetary Science Letters 148 243^258

Blichert-Toft J Chauvel C amp Albare de F (1997) Separation of Hfand Lu for high precision isotope analysis of rock samples by mag-netic sector-multiple collector ICP-MS Contributions to Mineralogy

and Petrology 127 248^260Brey G P amp Kolaquo hler T (1990) Geothermobarometry in four-phase

lherzolites II New thermobarometers and practical assessment ofexisting thermobarometers Journal of Petrology 31 1353^1378

Brey G P Kolaquo hlerT amp Nickel K G (1990) Geothermobarometry infour-phase lherzolites I Experimental results from 10 to 60 kbJournal of Petrology 31 1313^1352

Caporuscio F amp Smyth J (1990) Trace element crystal chemistryof mantle eclogites Contributions to Mineralogy and Petrology 105550^561

Carbno G B amp Canil D (2002) Mantle structure beneath the SWSlave Craton Canada constraints from garnet geochemistry in theDrybones Bay Kimberlite Journal of Petrology 43 129^142

Coleman R G Lee E D Beatty L B amp Brannock W W (1965)Eclogites and eclogites their differences and similaritiesGeological Society of America Bulletin 76 483^508

Creaser R A Grulaquo tter H Carlson J amp Crawford B (2004)Macrocrystal phlogopite Rb^Sr dates for the Ekati propertykimberlites Slave Province Canada evidence for multiple intru-sive episodes in the Paleocene and Eocene Lithos 76 399^414

Davies R M GriffinW L Pearson N J Andrew A S Doyle BJ amp OrsquoReilly S Y (1999) Diamonds from the deep Pipe DO-27Slave Craton Canada In Gurney J J Gurney J L Pascoe MD amp Richardson S H (eds) Proceedings of the 7th International

Kimberlite Conference CapeTown Red Roof Design pp 148^155Davies R M Griffin W L OrsquoReilly S Y amp Doyle B J (2004)

Mineral inclusions and geochemical characteristics of microdia-monds from the DO27 A154 A21 A418 DO18 DD17 and RanchLake kimberlites at Lac de Gras Slave Craton Canada Lithos 7739^55

DavisW J amp Hegner E (1992) Neodymium isotopic evidence for thetectonic assembly of Late Archean crust in the Slave Provincenorthwest Canada Contributions to Mineralogy and Petrology 111493^504

Davis W J Jones A G Bleeker W amp Grutter H (2003)Lithosphere development in the Slave craton a linked crustal andmantle perspective Lithos 71 575^589

Dawson J B (1984) Contrasting types of upper-mantle metasoma-tism In Kornprobst J (ed) Kimberlites II The Mantle and

Crust^Mantle Relationships Amsterdam Elsevier pp 289^294DePaolo D J (1981) Neodymium isotopes in the Colorado Front

Range and crust^mantle evolution in the Proterozoic Nature 291193^196

Erlank A J amp Kushiro I (1970) Potassium contents of syntheticpyroxenes at high temperatures and pressures Carnegie Institution ofWashingtonYearbook 68 433^439

Foley S F Barth M G amp Jenner G A (2000) Rutilemelt partitioncoefficients for trace elements and an assessment of the influence ofrutile on the trace element characteristics of subduction zonemagmas Geochimica et Cosmochimica Acta 64 933^938

Foley S F Buhre S amp Jacob D E (2003) Evolution of the Archaeancrust by delamination and shallow subduction Nature 421 249^252

Graham I Burgess J L Bryan D Ravenscroft P J Thomas EDoyle B J Hopkins R amp Armstrong K A (1999) Explorationhistory and geology of the Diavik Kimberlites Lac de GrasNorthwest Territories Canada In Gurney J J Gurney J LPascoe M D amp Richardson S H (eds) Proceedings of the 7th

International Kimberlite Conference Cape Town Red Roof Designpp 262^279

Green D H amp Wallace M E (1988) Mantle metasomatism byephemeral carbonatite melts Nature 336 459^462

Green T H Blundy J D Adam J amp Yaxley G M (2000) SIMSdetermination of trace element partition coefficients betweengarnet clinopyroxene and hydrous basaltic liquids at 2^75 GPaand 1080^12008C Lithos 53 165^187

GriffinW L amp OrsquoReilly SY (2007) Cratonic lithospheric mantle Isanything subducted Episodes 30(1) 43^53

Griffin W L Wass S Y amp Hollis J D (1984) Ultramafic xenolithsfrom Bullenmerri and Gnotuk maars Victoria Australia petrol-ogy of a sub-continental crust^mantle transition Journal of

Petrology 25 53^87GriffinW L Doyle B J Ryan C G Pearson N J OrsquoReilly SY

Davies R Kivi K van Achterbergh E amp Natapov LM (1999)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1871

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Layered mantle lithosphere in the Lac de Gras Area SlaveCraton composition structure and origin Journal of Petrology 40705^727

Griffin W L Pearson N J Belousova E Jackson S E vanAchterbergh E OrsquoReilly SY amp Shee S R (2000)The Hf isotopecomposition of cratonic mantle LAM-MC-ICPMS analysis ofzircon megacrysts in kimberlites Geochimica et Cosmochimica Acta

64 133^147Griffin W L OrsquoReilly S Y Doyle B J Pearson N J

Coopersmith H Kivi K Malkovets V amp Pokhilenko N(2004) Lithosphere mapping beneath the North American plateLithos 77 873^922

Grulaquo tter H S Apter D B amp Kong J (1999) Crust^mantle couplingevidence from mantle-derived xenocrystic garnets In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 307^312

Hart S R Blusztajn J Dick H J B Meyer P S ampMuehlenbachs K (1999) The fingerprint of seawater circulationin a 500-meter section of ocean crust gabbros Geochimica et

Cosmochimica Acta 63 4059^4080Harte B amp Kirkley M B (1997) Partitioning of trace elements

between clinopyroxene and garnet Data from mantle eclogitesChemical Geology 136 1^24

Heaman L M Creaser R A amp Cookenboo H O (2002) Extremeenrichment of high field strength elements in Jericho eclogitexenoliths A cryptic record of Paleoproterozoic subduction partialmelting and metasomatism beneath the Slave craton CanadaGeology 30 507^510

Heaman L M Kjarsgaard B A amp Creaser R A (2004) The tem-poral evolution of North American kimberlites Lithos 76 377^397

Heaman L M Creaser R A Cookenboo H O amp Chacko T(2006) Multi-stage modification of the northern Slave mantlelithosphere Evidence from zircon- and diamond-bearing eclogitexenoliths entrained in Jericho Kimberlite Journal of Petrology 47821^858

Hermann J Cpandler C Hack A amp Korsakov A (2006) Aqueousfluids and hydrous melts in high-pressure and ultra-high pressurerocks implications for element transfer in subduction zones Lithos92 399^417

Hills DV amp Haggerty S E (1989) Petrochemistry of eclogites fromthe Koidu Kimberlite Complex Sierra Leone Contributions to

Mineralogy and Petrology 103 397^422Hoffman P F (1989) Precambrian geology and tectonic history of

North America In Bally A amp Palmer A (eds) The Geology of

North AmericaccedilAn Overview Boulder CO Geological Society ofAmerica pp 447^512

IrelandT R Rudnick R L amp Spetsius Z (1994) Trace-elements indiamond inclusions from eclogites reveal link to Archean granitesEarth and Planetary Science Letters 128 199^213

Irvine G J Pearson D G Kjarsgaard B A Carlson R WKopylova M G amp Dreibus G (2003) Evolution of the litho-spheric mantle beneath Northern Canada a Re^Os isotope andPGE study of kimberlite-derived peridotite xenoliths fromSomerset Island and a comparison to the Slave and Kaapvaalcratons Lithos 71 461^488

Isachsen C E amp Bowring S A (1994) Evolution of the Slave cratonGeology 22 917^920

Jacob D E (2004) Nature and origin of eclogite xenoliths from kim-berlites Lithos 77 295^316

Jacob D E amp Foley S F (1999) Evidence for Archean ocean crustwith low high field strength element signature from diamondifer-ous eclogite xenoliths Lithos 48 317^336

Jacob D Jagoutz E Lowry D Mattey D amp Kudrjavtseva G(1994) Diamondiferous eclogites from Siberiaccedilremnants ofArchean oceanic-crust Geochimica et Cosmochimica Acta 585191^5207

Jacob D E Fung A Jagoutz E amp Pearson D G (2003) Petrologyand geochemistry of eclogite xenoliths from the Ekati kimberlitesarea Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA_239

Jacob D E Bizimis M amp Salters V J M (2005) Lu^Hf and geo-chemical systematics of recycled ancient oceanic crust evidencefrom Roberts Victor eclogites Contributions to Mineralogy and

Petrology 148 707^720Jerde E A Taylor L A Crozaz G Sobolev NV amp SobolevV N

(1993) Diamondiferous eclogites from Yakutia Siberiaccedilevidencefor a diversity of protoliths Contributions to Mineralogy and Petrology

114 189^202Jones A G Ferguson I J Chave A D Evans R L amp

McNeice GW (2001) The electric lithosphere of the Slave cratonGeology 29 423^426

Kelemen P B Shimizu N amp Dunn T (1993) Relative depletionof niobium in some arc magmas and the continental-crust-partitioning of K Nb La and Ce during meltrockreaction in the upper-mantle Earth and Planetary Science Letters 120111^134

Kinzler R J (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition Application to mid-ocean ridge basalt petrogenesis Journal of Geophysical Research 102853^874

Klemme S van der Laan S R Foley S F amp Guenther D (1995)Experimentally determined trace and minor element partitioningbetween clinopyroxene and carbonatite melt under upper mantleconditions Earth and Planetary Science Letters 133 439^448

Konzett J Armstrong R A Sweeney R J amp CompstonW (1998)The timing of MARID metasomatism in the Kaapvaal mantle anion probe study of zircons from MARID xenoliths Earth and

Planetary Science Letters 160 133^145Kopylova M G Russell J K amp Cookenboo H (1999) Mapping the

lithopshere beneath the north central Slave Craton In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 468^479

Krogh E (1988) The garnet^clinopyroxene iron^magnesiumgeothermometerccedila reinterpretation of existing experimentaldata Contributions to Mineralogy and Petrology 99 44^48

Kusky T M (1989) Accretion of Archean Slave Province Geology 1763^67

LeCheminant A N amp Heaman L M (1989) Mackenzie igneousevents Canada Middle Proterozoic hotspot magmatismassociated with ocean opening Earth and Planetary Science Letters

96 38^48LeCheminant A N Richardson D G DiLabio R N W amp

Richardson K A (1996) La recherche de diamants au CanadaGeological Survey of Canada Open File 3228 268pp

MacGregor I D amp Carter J L (1970) The chemistry of clinopyrox-enes and garnets of eclogite and peridotite xenoliths from theRoberts Victor mine South Africa Physics of the Earth and PlanetaryInteriors 3 391^397

MacGregor I D amp Manton W I (1986) Roberts Victor eclogitesancient oceanic crust Journal of Geophysical Research 91 14063^14079

MacKenzie J M amp Canil D (1999) Composition and thermal evo-lution of cratonic mantle beneath the central Archean SlaveProvince NWT Canada Contributions to Mineralogy and Petrology

134 313^324

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1872

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

McCandlessT E amp Gurney J J (1989) Sodium in garnet and potas-sum in clinopyroxene criteria for classifying mantle eclogites InRoss J (ed) Kimberlites and Related Rocks Carlton Blackwellpp 827^832

McDonoughW F amp Sun S-S (1995) The composition of the EarthChemical Geology 120 223^253

Navon O amp Stolper E (1987) Geochemical consequences of meltpercolation the upper mantle as a chromatographic columnJournal of Geology 95 285^307

Nikogosian I K amp Sobolev AV (1997) Ion-microprobe analysis ofmelt inclusions in olivine experience in estimating the olivine^meltpartition coefficients of trace elements Geochemistry International 35119^126

Norman M D Griffin W L Pearson N J Garcia M O ampOrsquoReilly S Y (1998) Quantitative analysis of trace element abun-dances in glasses and minerals a comparison of laser ablationinductively coupled plasma mass spectrometry solution inductivelycoupled plasma mass spectrometry proton microprobe and elec-tron microprobe data Journal of Analytical Atomic Spectrometry 13477^482

Norman M D Garcia M O amp Pietruszka A J (2005) Trace-element distribution coefficients for pyroxenes plagioclase andolivine in evolved tholeiites from the 1955 eruption of KilaueaVolcano Hawairsquoi and petrogenesis of differentiated rift-zonelavas American Mineralogist 90 888^899

OrsquoHara M J amp Herzberg C (2002) Interpretation of trace elementand isotope features of basalts Relevance of field relations petrol-ogy major element data phase equilibria and magma chambermodeling in basalt petrogenesis Geochimica et Cosmochimica Acta 662167^2191

PadghamW A amp FysonW K (1992) The Slave Province a distinctcraton CanadianJournal of Earth Sciences 29 2072^2086

Pearson D G Davies G R amp Nixon P H (1993) Geochemical con-straints on the petrogenesis of diamond facies pyroxenites from theBeni Bousera peridotite massif North Morocco Journal of Petrology34 125^172

Pearson N J Griffin W L Doyle B J OrsquoReilly S Y vanAchterbergh E amp Kivi K (1999) Xenoliths from kimberlitepipes of the Lac de gras area Slave Craton Canada In GurneyJ J Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite ConferenceCapeTown Red Roof Design pp 644^658

Perfit N R Gust D A Bence A E Arculus R J amp Taylor S R(1980) Chemical characteristics of island-arc basalts implicationsfor mantle sources Chemical Geology 30 227^256

Polat A Hofmann A W amp Rosing M T (2002) Boninite-likevolcanic rocks in the 37^38 Ga Isua greenstone belt WestGreenland geochemical evidence for intra-oceanic subductionzone processes in the early Earth Chemical Geology 184 231^254

Poudjom DjomaniY H GriffinW L OrsquoReilly SY amp Doyle B J(2005) Lithospheric domains and controls on kimberlite emplace-ment Slave Province Canada evidence from elastic thickness andupper mantle composition Geochemistry Geophysics Geosystems 6doi1010292005GC0009978

Presnall D C Gudfinnsson G H amp Walter M J (2002)Generation of mid-ocean ridge basalts at pressures from 1 to 7GPa Geochimica et Cosmochimica Acta 66 2073^2090

Schmidberger S S Heaman L M Simonetti A Creaser R A ampCookenboo H O (2005) Formation of Paleoproterozoiceclogitic mantle Slave Province (Canada) Insights from in-situHf and U^Pb isotopic analyses of mantle zircons Earth and

Planetary Science Letters 240 621^633

Schmidberger S S Simonetti A Heaman L M Creaser R A ampWhiteford S (2007) Lu^Hf in-situ Sr and Pb isotope and traceelement systematics for mantle eclogites from the Diavik diamondmine Evidence for Paleoproterozoic subduction beneath the Slavecraton Canada Earth and Planetary Science Letters 254 55^68

Schulze D J amp Helmstaedt H (1988) Coesite-sanidine eclogitesfrom kimberlite products of mantle fractionation or subductionJournal of Geology 96 435^443

Smithies R H Champion D C amp Sun S S (2004) The case forArchaean boninites Contributions to Mineralogy and Petrology 147705^721

Smyth J R Caporuscio F A amp McCormick T C (1989) Mantleeclogites evidence of igneous fractionation in the mantle Earth andPlanetary Science Letters 93 133^141

Sobolev NV amp Lavrentrsquoyev Y G (1971) Isomorphic sodium admix-ture in garnets formed at high pressure Contributions to Mineralogy

and Petrology 31 1^12Stalder R Foley S F Brey G P amp Horn I (1998) Mineral aqueous

fluid partitioning of trace elements at 900^12008C and 30^57GPa New experimental data for garnet clinopyroxene and rutileand implications for mantle metasomatism Geochimica et

Cosmochimica Acta 62 1781^1801Sun S-S amp McDonough W F (1989) Chemical and isotopic sys-

tematics of oceanic basalts implications for mantle compositionsand processes In Saunders A D amp Norris M J (eds)Magmatism in the Ocean Basins Geological Society London Special

Publications 42 313^345Taura H Yurimoto H Kurita K amp Sueno S (1998) Pressure

dependence on partition coefficients for trace elements between oli-vine and the coexisting melts Physics and Chemistry of Minerals 25469^484

Taylor L A amp Neal C R (1989) Eclogites with oceanic crustal andmantle signatures from the Bellsbank kimberlite South AfricaPart I Mineralogy petrography and whole rock chemistryJournal of Geology 97 551^567

van Achterbergh E Ryan C G amp Griffin W L (1999) Glitter online intensity reduction for the laser ablation inductively coupledplasma spectrometry Abstract 9th V M GoldschmidtConference LPI Boston MA 905 pp Abstract no 7215

van Breemen O DavisW J amp King J E (1992) Temporal distribu-tion of granitoid plutonic rocks in the Archean Slave ProvinceNorthwest Canadian Shield Canadian Journal of Earth Sciences 292186^2199

Vervoort J D Patchett P J Blichert-Toft J amp Albare de F (1999)Relationship between Lu^Hf and Sm^Nd isotopic systems in theglobal sedimentary system Earth and Planetary Science Letters 16879^99

Viljoen K S Smith C B amp Sharp Z D (1996) Stable and radio-genic isotope study of eclogite xenoliths from the Orapa kimberliteBotswana Chemical Geology 131 235^255

Walter M J (1998) Melting of garnet peridotite and theorigin of komatiite and depleted lithosphere Journal of Petrology 3929^60

Wasserburg G J Jacobsen S B DePaolo D J McCulloch M Tamp Wen T (1981) Precise determination of SmNd ratios Sm andNd isotopic abundances in standard solutions Geochimica et

Cosmochimica Acta 45 2311^2323Westerlund K J Shirey S B Richardson S H Carlson R W

Gurney J J amp Harris JW (2006) A subduction wedge origin forPaleoarchean peridotitic diamonds and harzburgites from thePanda kimberlite Slave craton evidence from Re^Os isotope sys-tematics Contributions to Mineralogy and Petrology 152 275^294

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1873

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Page 19: Origins of Xenolithic Eclogites and Pyroxenites from the Central

RockNMORB

(b)

RockNMORB

(d)

010

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu001

01

1

10

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

(a)

001

01

1

10(b)

001

01

1

10(c)

001

01

1

10(e)

high-TSchmidberger et al

low-TSchmidberger et al

La Ce Pr Nd Sm Eu Gd Dy Ho Er Yb Lu

high-Mglow-Mg

high-Cavolatile-rich

phlogopite-bearingYK1911

calcite-bearingVR67112b

YK1949VR43479

Fig 11 N-MORB normalized REE patterns for (a) low-Mg eclogites (b) high-Mg eclogites (c) high-Ca eclogites and (d) volatile-richeclogites Shown for comparison in (e) are reconstructed eclogites from Schmidberger et al (2007)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1861

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Both volatile-rich whole rocks display strong negativeSr anomalies Strontium can be sited in calcite or apatiteandor indicate the presence of a Sr-rich mineral that wasnot identified in thin section such as bariteFour of the six reconstructed whole-rocks in the study

of Schmidberger et al (2007) (Fig 11e) show markedLREEHREEN enrichment that is observed only in onehigh-Ca eclogite in our study (Fig 11c) suggesting thatmany of the eclogites they studied were affected by metaso-matic LREE enrichment Two samples in their study thatare not affected by this enrichment show positive Euanomalies similar to the high-Mg eclogites in this studywhereas the distinctive REE pattern of the low-Mg eclo-gites is not recognized among their samples

SR^ND^HF I SOTOPESYSTEMATICSStrontium Nd and Hf isotope data for optically puremineral separates and reconstructed whole-rocks aregiven in Table 5 Spongy rims and alteration are likely tobe mostly removed during leaching in 6N HCl in combina-tion with repeated ultrasonication Despite this rigoroustreatment the reproducibility of Nd in garnet VR43469and in cpx YK1914 is worse than for other samples Thismay be due to inhomogeneities related to late Nd diffusionpossibly from grain boundaries containing a kimberlitecomponent The reproducibility of the 176Hf177Hf data isgenerally poorer than that of Nd which may reflect a com-bination of matrix effects caused by imperfectly purifiedsolutions and sample heterogeneity Parent^daughterratios are available only from in situ analyses which arenot precise enough to allow meaningful model age orinitial isotope ratio calculations (ie at the time of kimber-lite eruption) Because the Lac de Gras kimberlites arerelatively young (55^56 Ma Graham et al 1999) age cor-rection would result in only a minor change in isotopecomposition which is inconsequential in view of the rangeof compositions displayed by most samples and hence hasno effect on the conclusions drawnClinopyroxenes in five volatile-free eclogites and a

pyroxenite have Sr isotope ratios ranging from 0701662 to0703546 this range is similar to or somewhat lower thanthat of the depleted mantle at the time of kimberliteemplacement (0703 Bell amp Tilton 2002) Clinopyroxenein the phlogopitized sample has a distinctly high 87Sr86Srof 0751186Clinopyroxene controls 56^92 of the Nd budget (Nd

concentration in cpx weighted by mode) Neodymiumisotope ratios range from 0511660 to 0514844 Thiscorresponds to eNd variation (per 10 000 deviationof 143Nd144Nd in the sample from 143Nd144Nd in the chon-dritic mantle model reservoir) from 17 to thorn43 cpx inthe low-Mg eclogites have higher eNd than those in

high-Mg eclogites Garnets have generally have higher143Nd144Nd from 0512984 (eNdfrac14thorn 7) to 0520965(eNdfrac14thorn 162) Garnets in the two volatile-rich sampleshave the highest ratios 0522790 for garnet in the calcite-bearing sample VR67112b (eNdfrac14thorn198) and 0525721 inthe phlogopitized eclogiteYK1911 (eNdfrac14thorn 255)Garnet contributes 18^32 and cpx 48^78 of the Hf

budget The proportion controlled by rutile is 4^30(Aulbach et al in preparation) Measured 176Hf177Hf incpx ranges from 0281263 in phlogopite-bearing sampleYK1911 (eHffrac1453) to 0286130 in low-Mg eclogiteVR43469 (eHffrac14thorn119) and in garnet from 0283958 inpyroxeniteYK1915 (eHffrac14thorn42) to 0295718 in low-Mg eclo-gite YK1926 (eHffrac14thorn458) Hafnium in garnet in the phlo-gopite^apatite-bearing sample is extremely radiogenic(176Hf177Hffrac14 0354367) corresponding to an eHf of thorn2532which is close to a value determined for a garnet separatefrom a Roberts Victor eclogite (eHffrac14thorn2561 Jacob et al2005) Garnet in the calcite-bearing sample has extremelyradiogenic Nd and Hf (eNdfrac14thorn193 eHffrac14thorn2516) Garnetin pyroxenite YK1915 also has radiogenic Nd and Hf(eNdfrac14thorn27 eHffrac14thorn42)Despite the large uncertainties for reconstructed whole-

rocks (calculated as the square root of the sum of squareduncertainties of two standard errors for garnet and cpxanalyses combined with 20 propagated uncertainty onthe proportion of Nd or Hf respectively contributed byeach mineral) the isotopic data are consistent in thatthey correlate with the independently determinedparent^daughter ratios (1) Low-Mg eclogites have higherSmNd reflected in the steeper positive slope of theirN-MORB-normalized REE patterns (Fig 11a) comparedwith the high-Mg eclogites (Fig 11b) Accordingly recon-structed low-Mg eclogites have higher present-day eNd

(0 to thorn47) than high-Mg eclogites (4 and 17) (2) Low-Mg eclogite YK1943 has very high Nb and LREE abun-dances that could have been added during metasomatismthat also enhanced its Hf abundance This sample has thelowest eNd and eHf of the low-Mg eclogites (0 and thorn20respectively) In contrast apparently unenriched low-Mgeclogites have more radiogenic Nd and Hf (eNdfrac14thorn47and thorn9 eHffrac14thorn128 and thorn113 respectively)The Sr^Nd isotope diagram reveals that eNd in cpx

varies from 19 to 43 at nearly constant depleted-mantle-like 87Sr86Sr (Fig 12a) In the Nd^Hf isotope diagram(Fig 12b) the reconstructed whole-rocks lie on curvedmixing lines if the lsquomixingrsquo proportions of Nd and Hffor the two lsquoend-memberrsquo minerals are very different(eg YK1911) Garnets have more radiogenic Nd andHf than coexisting cpx All reconstructed volatile-freeeclogites plot in the radiogenic field with regard to Hfor are isotopically similar to the depleted mantle Twohigh-Mg eclogites and one low-Mg eclogite have unradio-genic Nd one high-Mg eclogite has radiogenic Nd and

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1862

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 5 Individual and averaged Sr Nd and Hf isotope ratios obtained from garnet and cpx solutions

Sample 87Sr86Sr 2SE2SD 143Nd144Nd 2SE2SD eNd 2SE2SD 176Hf177Hf 2SE2SD eHf 2SE2SD

Low-Mg eclogites

VR43469 cp 0702661 0000010 0514844 0000006 43 01 0286130 0000012 119 04

VR43469 gt 0515508 0000014 56 03 0286506 0000012 132 04

VR43469 gt r 0515453 0000010 55 02 0286548 0000011 134 04

VR43469 gt r 0515384 0000007 54 01 0287244 0000010 158 04

VR43469 gt r 0515487 0000014 56 03 0286996 0000014 149 05

VR43469avg gt 0515458 0000109 55 21 0286823 0000715 143 253

WR VR43469 0515059 0000273 47 5 0286405 0001079 128 38

YK1926 cp 0703546 0000011 0513043 0000007 8 0 0283154 0000009 13 03

YK1926 gt 0513364 0000008 14 01 0295718 0000005 458 02

WR YK1926 0513120 0000026 9 1 0285980 0000716 113 25

YK1943 cp 0512143 0000007 10 01 0282213 0000005 20 02

YK1943 gt 0514461 0000012 36 02 0287510 0000008 168 03

YK1943 gt r 0514491 0000009 36 02 0287332 0000005 161 02

YK1943 gt r 0514517 0000010 37 02 0287290 0000007 160 02

YK1943avg gt 0514489 0000056 36 11 0287377 0000233 163 82

WR YK1943 0512625 0000228 0 5 0283138 0000879 13 31

High-Mg eclogites

VR43480 gt 0520965 0000017 162 03 0284950 0000010 77 04

VR67360 cp 0283509 0000074 26 26

VR67360 cp r 0283433 0000031 23 11

VR67360avg cp 0701662 0000018 0511660 0000010 19 02 0283471 0000108 25 38

VR67360 gt 0512984 0000101 7 20 0287429 0000180 165 64

WR VR67360 0511761 0000132 17 3 0284734 0000469 69 17

YK1949 cp 0703019 0000013 0512231 0000009 8 02 0282513 0000009 9 03

YK1949 gt 0513511 0000010 17 02 0290297 0000042 266 15

YK1949 gt r 0513508 0000012 17 02 0289947 0000015 254 05

YK1949 gt r 0513552 0000014 18 03

YK1949avg gt 0513523 0000049 17 10 0291569 0000494 311 175

WR YK1949 0512446 0000103 4 2 0284622 0000847 65 30

Volatile-rich eclogites

VR67112b gt 0522790 0000008 198 02 0354367 0000036 2532 13

YK1911 cp 0751186 0000009 0512165 0000007 9 01 0281263 0000009 53 03

YK1911 gt 0525721 0000013 255 03 0288051 0000012 187 04

WR YK1911 0513940 0000342 25 7 0281414 0000018 48 1

Pyroxenites

YK1914 cp 0703015 0000034 0513283 0000041 13 08

YK1914 cp r 0702956 0000005 0513200 0000027 11 05

YK1914 cp r 0513527 0000059 17 11

YK1914 cp r 0513045 0000010 8 02

YK1914avg cp 0702986 0000083 0513264 0000402 12 78

YK1915 gt 0514003 0000011 27 02 0283958 0000021 42

Analyses of full procedural replicates are designated with an lsquorrsquo after sample name SE standard error (internal errorsingle analyses) SD standard deviation (external error averaged multiple analyses) Italics used for reconstructed whole-rocks (mineral isotopic compositions and concentrations weighted by modes) Uncertainties for reconstructed whole-rocksinclude 2SE or 2SD uncertainties on each mineralrsquos isotope composition combined with 20 uncertainty in the amountof Nd or Hf contributed by the constituent minerals eNd and eHf are per 10 000 deviation from model chondritic mantle[model of Wasserburg et al (1981) and Blichert-Toft amp Albarede (1997) respectively]

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1863

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

εHf

εNd

Ekati eclogitesminus100

0

100

200

300

400

500

minus40 0 40 80 120 160 200 240 280

gt in YK1915

gt in VR67112bεNd=198 εHf=2532

minus40

0

40

80

120

160

minus40 minus20 0 20 40 60

minus60

minus40

minus20

0

20

40

070 071 072 073 074 075 076

North Americankimberlite field

εNd

87Sr86Sr

cpx in YK1914

Ekatieclogites

DM

DM

DM

DM

(a)

(b)

pyroxenites

low-Mg

high-Mg

volatile-rich

eclogites

Diavik low-TA154 S kimberlite

cpx in YK1911

Fig 12 (a) Sr^Nd for cpx and (b) Nd-Hf isotope diagram for garnet cpx and reconstructed whole-rocks (symbols with bold outline) Finedashed lines in (b) connect data points from the same sample with garnets having more radiogenic isotopic compositions Uncertainties forreconstructed whole-rocks combine 20 modal uncertainty with uncertainties on each mineralrsquos isotope composition Error bars on individualminerals may be smaller than the symbols Also shown are the isotopic compositions of the depleted mantle (DM DePaolo 1981 Griffin et al2000 Bell amp Tilton 2002) Range of 87Sr86Sr in eclogites from Ekati (Lac de Gras area 100 km north of Diavik Jacob et al 2003) in NorthAmerican kimberlites and compositions of low-temperature eclogites from Diavik and of Lac de Gras A154 South kimberlite fromSchmidberger et al (2007)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1864

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

one low-Mg eclogite has eNd similar to the depleted mantle(Fig 12b)Clinopyroxene in the phlogopitized eclogite YK1911

has unradiogenic Nd and Hf and strongly radiogenicSr whereas the coexisting garnet has radiogenic Nd andHf Although this sample contains 50 vol phlogopiteits whole-rock isotopic composition was calculated

from garnet and cpx alone because the contributionfrom phlogopite to the Hf and Nd budget is 52 and51 respectively (unpublished data) The whole-rockhas radiogenic Nd but has lower eHf than anyother sample analysed in accord with the extreme Hfabundance (53 ppm) and low LuHf of the cpx

DISCUSS IONOrigins of eclogites beneath the SlaveCratonHigh-Mg and high-Ca eclogites

With regard to their major- and trace-element contentsthe high-Mg and high-Ca eclogites from Lac de Gras arebroadly similar to variably evolved oceanic gabbros(Fig 13b and c) which may indicate an origin as subductedoceanic crust Compatible trace-element abundancesare useful to deduce potential crustal protoliths as theseelements are less susceptible to post-formation modification(Jacob 2004) The flat HREE patterns in the high-Mg andhigh-Ca eclogites are compatible with a low-pressureorigin as the oceanic crust is a product of moderate degreesof partial melting mainly within the spinel stability field(eg Presnall et al 2002) whereas melting with residualgarnet would qualitatively be expected to produce stronglyfractionated REE patterns Positive Sr and Pb anomaliesalthough compatible with a plagioclase signature couldalso be due to cryptic metasomatism (Dawson 1984) ofthe eclogites during their residence in the lithosphericmantle These anomalies are probably not related to kim-berlite infiltration as the whole-rock eclogite compositionswere reconstructed from analyses of fresh mineral coresLow ZrSm similar to that in primitive gabbros is

observed for many of the high-Mg and high-Caeclogites (Fig 13b and c) and is independent of the ratioof garnetcpx and whether or not the whole-rock wasreconstructed with rutile (Electronic Appendix 10) Thenegative trend of Mg-number against Yb in garnet(Fig 14) argues against igneous processes involving garnetaccumulation (Taylor amp Neal 1989) as suggested forexample for some high-Mg eclogites from West Africa(Barth et al 2002b) Rather the anti-correlation suggeststhat olivine fractionation played a role in the formation ofthe protoliths Fractionating olivine would excludeYb (Dolivinebasalt

frac14 002 Nikogosian amp Sobolev 1997)and incorporate Mg relative to Fe (Dfrac1415 and 05 respec-tively Taura et al1998) leading to progressive depletion ofMg and enrichment of Yb in the residual melt and henceanti-correlated Mg and Yb in the whole-rock This isinherited after eclogitization All eclogite types (exceptVR43452) plot along the olivine fractionation trend sug-gesting that they formed by a similar processHigh-Mg eclogites have lower

PREE than high-Ca

eclogites and some appear to have positive Eu anomalies

high-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

gabbrosSE Indian ridge

gabbrosOman ophiolite

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

high-Mg

RockNMORB(a)

VR43452

YK1949

low-Mg

boninite-like melts

01

001

1

10

(b)

01

001

1

10

(c)

01

001

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

Fig 13 Trace-element patterns of (a) low-Mg (b) high-Mg and(c) high-Ca eclogites compared with boninite-like volcanic rocksfrom the Isua greenstone belt (Polat et al 2002) gabbros from theOman ophiolite (Benoit et al 1996) and the SE Indian Ridge (Hartet al 1999) normalized to N-MORB of Sun amp McDonough (1989)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1865

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(EuEu up to 14) similar to more primitive cumulates(troctolite gabbros and olivine gabbros) reported in the lit-erature (Fig 13b) Although the sizes of these Eu anomaliesare not outside the analytical uncertainty their correlationwith CeYb in reconstructed eclogites supports their signif-icance (r2frac14066 nfrac14 4) The more primitive nature ofhigh-Mg eclogites is consistent with higher Mg-numbersrelative to high-Ca eclogites and may be due to olivineaccumulation during formation of their protolithsThe two samples with the lowest HREE abundanceshave the highest EuEu which is compatible with varia-tions in plagioclase modesMore evolved protoliths with higher pyroxeneplagio-

clase plus trapped melt could account for small or absentpositive Eu anomalies and higher

PHREE of high-Ca

eclogites (Fig 13c) Although the reason for the frequentassociation of graphite or diamond with high-Ca eclogitesis unclear (Aulbach et al 2003) the frequent occurrence ofkyanite in high-Ca eclogites also observed at Ekati justnorth of Lac de Gras has been argued to support a crustalprotolith (Jacob 2004) Kyanite andor corundum havebeen suggested to exsolve from high-temperature high-Alclinopyroxenes (Caporuscio amp Smyth 1990) Eclogitesfrom Lac de Gras show no exsolution textures suggestiveof such an origin although recrystallization can obscureor destroy microstructural evidence for exsolution(eg Griffin et al 1984)Modelling shows that qualitatively plagioclase accumu-

lation from a low-pressure melt derived from MORBmantle can account for positive Pb Sr and Eu and negativeZr anomalies relative to N-MORB whereas cpx accumula-tion or trapped melt increases Zr Y MREE and HREE

abundances (Fig 15) Some melt addition (for exampleby entrapment) seems to be required to increase all trace-element abundances to those observed for the eclogitesAlthough high-Ca and high-Mg eclogites could representshallow and deep portions of the same oceanic crust thehigher average temperatures obtained for the high-Caeclogites the frequent occurrence of carbon in high-Caeclogites and its absence in high-Mg eclogites the strongLREEHREE fractionation observed for six of seven high-Ca eclogites and the narrow range of compositions of thehigh-Mg eclogites suggest that they may have protolithsthat are not cogenetic Some discrepancies in particularaffecting the incompatible elements may be due to post-crystallization processes acting upon the protoliths suchas dehydration melting and later metasomatism that isevident in peridotites from the same locality (Griffin et al1999)Low K2OTh for high-Ca and high-Mg eclogites

(average 08 Table 4) relative to assumed crustal

0

2

4

6

8

10

30 40 50 60 70 80

Mg

VR43452

Yb

GARNET

pyroxenites

low-Mghigh-Mghigh-Ca

eclogites

VR19673ecl-g2

Fig 14 Garnet Mg-number vs Yb (ppm including 2s uncertainty)Lines qualitatively show separate correlations for eclogites (exceptoutlierVR43452) and pyroxenites (except outlierVR19673ecl-g2)

high-Ca

high-Mg

001

01

1

10

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

melt08 melt02 cryst cpx01 melt09 cryst

plag

(a)

001

01

1

10(b)

RockNMORB

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 15 Reconstructed (a) high-Ca and (b) high-Mg eclogitescompared with compositions of plagioclase and cpx crystallizingfrom a MORB melt modelled using distribution coefficients ofNorman et al (2005)The MORB was calculated as a 10 equilibriummelt of MORB pyrolite [composition from Kelemen et al (1993)except Pb Pr Sr Y and Lu which were calculated as residues from1 fractional melting of pyrolite from McDonough amp Sun (1995)using melt modes of Kinzler (1997) and distribution coefficients ofNorman et al (2005)] Also shown are mixtures of crystals (consistingof 50 plagioclase plus 50 cpx) and melt in the proportions 0901and 0208

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1866

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

precursors (eg gabbros SE Indian Ridge average 22Hart et al 1999) may reflect the preferred loss of fluid-mobile elements during dehydration (Becker et al 2000)although large uncertainties on these low-abundance ele-ments in the constituent minerals and reconstructedwhole-rocks makes this interpretation less than robustThe REE patterns of most high-Mg eclogites are too flatto represent melting residues as partial melting shouldproduce low LREEHREE (Electronic Appendix 10) Incontrast the marked LREE depletion of five of sevenhigh-Ca eclogites (LaLufrac14 02^08) contrasts with that ofassumed evolved gabbroic protoliths (LaLufrac1415^32Hart et al 1999) and can be modelled as a consequence ofpartial melting during subduction after eclogitization (iein the presence of garnet which retains HREE relative toLREE eg Barth et al 2002a) has taken place (Aulbachet al 2003)Our results are compatible with previous studies on

eclogite xenoliths from the Slave Craton includingthose from the same kimberlite that have linked thesesamples or some of these samples to Paleoproterozoicsubduction events (the 195^191Ga formation of theHottah terrane and the 18Ga formation of the GreatBear magmatic arc) in the Slave craton (Griffin et al1999 Kopylova et al 1999 Heaman et al 2002 2006Schmidberger et al 2005 2007)The compositional features discussed above although

compatible with crustal protoliths are not necessarilydefinitive of a subduction origin However if direct crystal-lization from basaltic melts at high pressure (ie mantledepths) is invoked instead specific conditions would haveto be met to satisfy the observations The Mg-number ineclogites from Lac de Gras are lower than those expectedof primary high-pressure mantle melts (eg Walter 1998OrsquoHara amp Herzberg 2002) requiring crystallization athigh pressures after olivine fractionation This originwould further require shielding from peridotitic wallrockwith which the fractionated melts would no longer be inequilibrium (Barth et al 2001) In such a setting the lowrelative abundances of fluid-mobile elements comparedwith those of less fluid-mobile elements and of LREErelative to HREE (Table 4) would reflect an expulsion offluids during crystallization of the protoliths and latermelt mobilization respectively The eclogites clearly donot represent garnet plus cpx cumulates which wouldrequire a positive correlation of Mg-number with Yb thatis at variance with the observed negative correlation ineclogitic garnets from this study (Fig 14)

Low-Mg eclogites

Garnets in low-Mg eclogites compositionally overlap withthose in massive eclogites from Jericho (NW of the Lacde Gras kimberlites) that were suggested to have a high-pressure intrusive origin based on the similarity of theirmajor-element compositions to those of Group A eclogites

fromsouthernAfrica forwhichamantleoriginhasbeenpro-posed (Kopylova et al1999) Garnets in low-Mg eclogites inthe present study have major-element compositions moresimilar to garnets in groupB rather thangroupAeclogitesThe positive slope in N-MORB-normalized HREE

abundances displayed by the low-Mg eclogites is clearlydistinct from other eclogite types and is similar to those ofboninite-like second-stage melts that are closely associatedwith ultramafic volcanics (Smithies et al 2004) (Fig 13a)The range of Mg-numbers in the low-Mg eclogites (55^66)overlaps with that of Archaean second-stage melts fromIsua (58^75) and Abitibi (52^82 Smithies et al 2004)These similarities might indicate that the low-Mg eclogiteshave an origin in a previously depleted mantle source suchas supra-subduction zone mantle Subchondritic ZrHf ofthe low-Mg eclogites contrasts with the supra-chondriticZrHf of the high-Mg and high-Ca eclogites (Aulbach et

al in preparation) and is consistent with previous cpx-controlled depletion in an arc mantle source This sourcecould have been subsequently remelted to generate theboninite-like precursor to the low-Mg eclogites as sug-gested for the protoliths of eclogites from Udachnaya(Jacob amp Foley 1999)Although some second-stage melts to which the low-Mg

eclogites are similar are closely associated with komatiitesand could contain a plume component the radiogenic ordepleted mantle-like Nd and Hf and unradiogenic Sr inthe low-Mg eclogites do not support plume involvement

Trace-element and Sr^Nd^Hf isotopeconstraints on post-formationmetasomatismSeveral low-Mg and high-Ca eclogites show negativeslopes in the LREE which may be indicative of interactionwith small amounts of LREE-enriched melt (Navon ampStolper 1987) amounting to cryptic metasomatism(ie not accompanied by crystallization of new phases) inthe nomenclature of Dawson (1984) High Nb abundancesrelative to elements of similar compatibility displayed bysome low-Mg and high-Ca eclogites (Fig 10) could resultfrom overestimation of rutile modes However high-CaeclogiteVR40345 is rutile-free and has been reconstructedwithout rutile its high Nb abundance is accompanied byLREE enrichment indicating that metasomatic additionmay be responsibleLow-Mg eclogites show a positive correlation of La with

Nb (Electronic Appendix 11a) indicating concomitantaddition of LREE and Nb This suggests that the metaso-matic agent was a solute-rich fluid or silicate silicic melt ashydrous fluids and carbonatites do not carry enough HFSEto affect the budget of eclogites (Green amp Wallace 1988Klemme et al 1995 Adam et al 1997 Stalder et al 1998)Of the samples that are enriched in Nb all but one (high-Ca eclogiteVR50909) show a simultaneous enrichment in

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1867

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Hf (Electronic Appendix 11b) Together these trace-element relationships suggest that both SmNd and LuHfwere lowered in these samples Their low parent^daughterratios contrast with higher ratios in unenriched samplesand resulted in the ingrowth of a range of Nd and Hf iso-tope compositions Compared with the low-Mg and high-Ca eclogites the high-Mg eclogites have a narrow range ofLa and Nb concentrations which correlates with a narrowrange in Nd and Hf isotope compositions (Fig 12)Precise model ages cannot be calculated because parent^

daughter ratios are available only from in situ analyseswhich have relatively large uncertainties Howeverevidence for long-term enrichment and depletion may bediscerned Low-Mg eclogites even those that are enrichedin LREE have relatively high SmNd as a result of havingoverall positive slopes in their REE patterns relative toMORB whereas the high-Mg eclogites have relativelyflat REE patterns and therefore lower SmNd (Fig 11)This translates into lower eNd for high-Mg eclogites thanfor low-Mg eclogites In contrast they show no systematicdifference with respect to 176Hf177Hf Because the low- andhigh-Mg eclogites are proposed to have different origins asdiscussed above and therefore did not necessarily have thesame initial isotopic composition and formation age theseisotopic differences cannot be used to estimate their agesA long-term coupling of trace-element and isotopic

abundances is also indicated by the fact that low-Mg eclo-gite YK1943 which has experienced Nb and LREE addi-tion has lower eNd (0) and eHf (thorn13) than low-Mg eclogiteVR43469 with higher LuHf and SmNd (eNdfrac14thorn47eHffrac14thorn128) If low-Mg eclogites YK1943 and VR43469originally had identical initial Nd and Hf isotope composi-tions and if the parent^daughter ratios are accurate andwere modified only once during their evolution shortlyafter metamorphism (when their isotopic compositionwould have been reset) this overprint can be calculatedfor both isotopic systems to have occurred at 17GaThis would suggest that Hf and Nd were added duringthe same event In contrast high-Mg eclogites VR67360and YK1949 would have had identical Ndisotope compositions at 10Ga whereas their Hf isotopecompositions are disturbed in that the sample with thehigher parent^daughter ratio has a lower 176Hf177Hf thanthe sample with the lower parent^daughter ratio Theseresults suggest that the LREE and HFSE enrichmentobserved in different samples occurred during separateevents that did not affect the eclogites equallyExtreme HFSE enrichment leading to the growth of

zircon and rutile and LREE addition leading to thegrowth of apatite in eclogites from the Jericho kimberlitein the northern Slave craton have been ascribed to twoseparate metasomatic events at around 18Ga (quasi-coincident with the age of formation and metamorphism)and 13^10Ga respectively (Heaman et al 2002 2006

Schmidberger et al 2005) The authors associated theseevents with Paleoproterozoic subduction and aMesoproterozoic thermal disturbance in the Slave litho-spheric mantle possibly linked to the Mackenzie igneousevent Mesoproterozoic Nd model ages mostly between13 and 14Ga were reported for eclogites from thenearby Ekati kimberlites (Jacob et al 2003) Although wedo not observe zircon or apatite in volatile-free eclogitesfrom Lac de Gras and the age constraints obtained fromour data are weak they support a link between theseevents described for other eclogite populations and themetasomatic enrichment observed in our studyThe Sr isotope data in the present study are similar to

those of Schmidberger et al (2007) who suggested that thedepleted-mantle-like values are consistent with derivationof the protoliths from an oceanic mantle source However87Sr86Sr in volatile-free eclogites in the present study isremarkably constant considering their Nd and Hf isotopevariability (Fig 12) and the variability of trace-elementpatterns (Fig 10) Given the evidence for dehydration andmelting in many of the eclogites fractionation of highlyincompatible Rb from Sr and differential ingrowth ofradiogenic Sr would be expectedThis may indicate a rela-tively recent overprint and considering the similarityof 87Sr86Sr to present-day depleted mantle which is signif-icantly lower than 87Sr86Sr of North American kimber-lites a juvenile (asthenospheric) source

PyroxenitesPyroxenites from Lac de Gras often show relationshipsin their major-element contents that are different fromthose observed in the eclogites For example pyroxeniticgarnets show a negative correlation between CaO andNa2O contents whereas these oxides are positively corre-lated for garnets in eclogites (Fig 5) CaO is well correlatedwith C2O3 in pyroxenitic garnets but not in eclogitic gar-nets (Fig 4) Mg-number is positively correlated withYb abundance for pyroxenitic garnets but anti-correlatedfor eclogitic garnets (Fig 14) This suggests that the pyrox-enites in the present study are not related to the eclogitesby fractionation or accumulation processes as proposedfor the origin of some pyroxenites (Foley et al 2003)Pyroxenites have also been interpreted as the products

of crystal segregation in magma channels with alteredoceanic crust as the magma source this would explaintheir Eu anomalies and stable and radiogenic isotopecompositions (Pearson et al 1993 Becker 1996) As thehigh-pressure liquidus phases of most terrestrial basaltsare cpx and garnet (OrsquoHara amp Herzberg 2002) partialmelts of eclogites with basaltic composition wouldcrystallize cpx and garnet but not opx More than half ofthe pyroxenites from Lac de Gras contain opx suggestingthat they probably did not crystallize directly fromeclogite-derived melts

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1868

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Low-Cr websterites and pyroxenites could be crystalli-zation products of mafic melts that interacted with perido-titic mantle thus becoming more refractory The positivecorrelation of Yb and Mg-number in garnet would allowan origin by fractional crystallization from such meltsAlternatively they formed by reaction of silicic (possiblyeclogite-derived) melts with mantle peridotites(eg Kelemen et al 1993 Aulbach et al 2002) This agreeswith the similarity of hypothetical melts in equilibriumwith pyroxenitic garnets in most samples to eclogite-derived melts (Electronic Appendix 12) Niobium concen-trations in all hypothetical melts are much higher than ineclogite-derived melts regardless of whether or not rutileis assumed to be in the residue The same is true for LaCe and Zr concentrations in some samples which mayindicate that NbZr and LREE were added to thepyroxenites some time after their formation as discussedin the previous section for some eclogites Because mini-mum ages for pyroxenitic rutile with an unradiogenicHf isotope composition (calculated by projecting the176Hf177Hf to the model depleted mantle isotopic evolutioncurve assuming zero ingrowth) range to 178066Ga(Aulbach et al in preparation) it appears likely that thepyroxenites were not formed recently (eg during kimber-lite-related or precursory melt migration) and that some ofthe samples are at least MesoproterozoicGarnet pyroxeniteYK1952 consists of schlieren of equili-

brated garnet pyroxenite in a lsquomatrixrsquo of spongy cpxfine cpx^opx intergrowths embayed garnet and meltpatches (Fig 2f) Garnet in the pyroxenite schlieren hasvery low LREE and HFSE abundances Orthopyroxenein the matrix has distinctly higher Mg-number andCr-number and contains more CaO than opx from thepyroxenite schlieren whereas opx rims in the schlierenare compositionally similar to opx in the matrix The con-siderable compositional and textural disequilibrium inthis sample suggests that this is a young feature causedby reaction with fluids shortly before or at the time ofentrainmentAs a consequence of the small sample sizes of most of

the pyroxenites studied here a comprehensive isotopicdataset is not available Clearly more such data areneeded in addition to stable isotope data to obtain betterconstraints on the origins of this rock type Consideringthe microstructural and compositional variability ofthe pyroxenites and major-element relationships oppositeto those observed for eclogites it appears certain that pyr-oxenites form by a variety of processes some of which areunrelated to the formation of eclogites and their protoliths

Volatile-rich eclogitesVolatile-rich eclogites from Lac de Gras display the mostextreme Hf^Nd isotope systematics of all the analysedsamples Garnets in phlogopitized (YK1911) and in cal-cite-bearing eclogite (VR67112b) have the most radiogenic

Nd (eNdfrac14 2486 and 1927 respectively) Garnet inVR67112b has extremely high eHf (2516) Clinopyroxene inYK1911 has the lowest eHf measured for any mineralseparate from Lac de Gras (53) in accord with the veryhigh Hf abundance in its cpx (5 ppm) In contrastgarnet in this sample has radiogenic Hf (eHffrac14 136) Thisdiscrepancy is at least in part due to their relatively lowequilibration temperatures (7698C and 8558C respec-tively) which allowed the isotopic compositions of cpxand garnet to evolve separately at their low andhigh parent^daughter element ratios respectively forextended periods of time and possibly since formationAccordingly the cpx in YK1911 gives a minimum Hf ageof 27GaSlab-derived fluids could have been introduced during

the 27Ga collision of the western Slave Craton domainincluding the thick plume-derived deep lithosphericmantle layer and the eastern domain with its highlydepleted lithosphere This would also be consistent withthe observation that volatile-rich eclogites occur at shallowdepths in the mantle (TKroghfrac14 685^8558C correspondingto depths of 85^115 km along a 40 mWm2 geotherm)where the bulk of dehydration and production of solute-rich melts in a subducting slab might be expected to occur(Becker et al 2000 Hermann et al 2006) Their negativeEu anomalies might indicate that the fluid was derivedfrom the shallow portion of the slab that experiencedplagioclase fractionation The presence of carbonate inVR67112b requires a carbonated sourceHigh 87Sr86Sr in cpx in YK1911 is probably the conse-

quence of isotopic equilibration with high-RbSr phlogo-pite after long-term radiogenic ingrowth while Hfaddition led to retarded ingrowth of radiogenic HfThese systematics combined with radiogenic Nd appear

inconsistent with the involvement of sediments in the pet-rogenesis of this sample (eg Perfit et al 1980 Vervoortet al 1999) or a hydrous fluid such as those responsible forMARID metasomatism in the Kaapvaal craton (Konzettet al 1998)Whatever the origin of the fluids the apparentHf mobility suggests that they were not in equilibriumwith a rutile-bearing source and also that they weresolute-rich in nature (ie hydrous melts able to dissolveHFSE) and the relatively high HREE abundances of thevolatile-rich eclogites indicate a garnet-free source suchas amphibolite This allowed evolution of the sample atrelatively high time-integrated SmNd

SUMMARY AND CONCLUSIONS

(1) The Lac de Gras kimberlites carried abundant eclo-gite and pyroxenite xenoliths to the surface Threemain eclogite types are distinguished in terms ofmajor-element composition and trace element

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1869

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

patterns high-Mg high-Ca and low-Mg eclogitesThe garnets of all eclogite types have anti-correlatedMg-number and Yb pointing to olivine fractionationas a controlling factor in the formation of their pre-cursor rocks Their flat or positively sloped HREEpatterns qualitatively indicate that the melt parentalto the precursor rocks was not in equilibrium with agarnet-bearing residue

(2) Reconstructed high-Mg eclogites and high-Caeclogites have flat HREE patterns positive Pb andSr anomalies and low ZrSm similar to oceanic gab-bros reported in the literature High-Mg eclogiteshave variably low HREE contents and some havesmall positive Eu anomalies possibly suggestive ofmore primitive relatively plagioclase-rich protolithswhereas high-Ca eclogites have higher HREE abun-dances and small or absent Eu anomalies that requiremore cpx- andor melt-rich protoliths if a subductionorigin is assumed

(3) Low-Mg eclogites have distinctive trace-element pat-terns that show a marked positive slope in the HREEand subchondritic ZrHf They resemble someArchaean second-stage melts (lsquoboninitesrsquo) and theirprotoliths appear to have formed from a previouslydepleted source

(4) After formation the eclogites were subject to a varietyof processes Most have low ratios of fluid-mobile vsless fluid-mobile elements possibly as a result of dehy-dration of the protoliths In addition some of the low-Mg and high-Ca eclogites have low LREEHREEconsistent with partial melt lossWhereas the low-Mgeclogites show concomitant enrichment in La Nb andZr some high-Mg and high-Ca eclogites show evi-dence for variable and unsystematic (decoupled)LREE and Nb addition

(5) Isotopically the trace-element characteristics of thedifferent eclogite types translate into lower eNd forhigh-Mg eclogites (negative slope in MREEfrac14 lowerSmNd) than for low-Mg eclogites (positive slope inMREEfrac14higher SmNd) Within the low-Mg groupsamples that show evidence for metasomatism (lowLuHf and SmNd) have lower eNd and eHf than asample that was apparently not metasomatized BothNd and Hf addition can be dated to 17Ga if identi-cal initial isotope compsitions are assumed for thelow-Mg eclogites In contrast the relatively constantand depleted mantle-like 87Sr86Sr of eclogitic cpxdespite variable Nd and Hf isotope compositions andparent^daughter ratios suggests recent input of juve-nile material

(6) Garnet and cpx in pyroxenites show major-elementrelationships different from those in eclogitesTherefore these two rock types are probably notrelated by fractionation processes The presence of

opx precludes direct crystallization from slab-derivedmelts Pyroxenites may be the product of (slab-derived) melts hybridized during assimilation of peri-dotitic mantle

(7) Garnets in phlogopite- and carbonate-bearing eclo-gites have extreme isotopic compositions includingthe most radiogenic Nd and Hf in the datasetNegative Eu anomalies may point to the involvementof a crustal protolith Highly radiogenic Sr radiogenicNd and unradiogenic Hf in the reconstructedphlogopite-bearing sample combined with highHREE abundances and evidence for Hf mobilitymay indicate involvement of solute-rich fluids origi-nating in a garnet- and rutile-free source

ACKNOWLEDGEMENTSHelp in the clean laboratories and with the analyticalfacilities by S Elhlou C Lawson A Sharma P Wielandand M Zhang is greatly appreciated Bill Griffin isthanked for his support and for agreeing to reviewseveral versions of the manuscript despite disagreeingwith many of the conclusions presented here The paperbenefited greatly from the detailed and constructivereviews provided by the editors Gareth Davies andMarjorie Wilson and by Matthias Barth Larry Heamanand an anonymous reviewer This work was funded by aMacquarie University International PostgraduateAward and Postgraduate Research Fund (SA) by anARC SPIRT grant sponsored by Kennecott Canada Incand by ARC Large and Discovery Grants to SYOrsquoRAnalytical data were obtained using instrumentationfunded by ARC LIEF and DEST Systemic InfrastructureGrants industry partners and Macquarie UniversityThis is Publication 481 from the ARC National KeyCentre for Geochemical Evolution and Metallogeny ofContinents (wwwesmqeduauGEMOC)

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online

REFERENCESAdam J GreenT H Sie S H amp Ryan C G (1997) Trace element

partitioning between aqueous fluids silicate melts and mineralsEuropeanJournal of Mineralogy 9 569^584

Aulbach S Stachel T Viljoen K S Brey G P amp Harris J W(2002) Eclogitic and websteritic diamond sources beneath theLimpopo Beltccedilis slab-melting the link Contributions to Mineralogy

and Petrology 143 56^70Aulbach S GriffinW L Pearson N J OrsquoReilly S Y Kivi K amp

Doyle B (2003) Origins of eclogites beneath the central SlaveCraton Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA-0011

Aulbach S Griffin W L Pearson N J OrsquoReilly S Y amp Kivi K(2004a) Mantle formation and evolution Slave Craton constraints

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1870

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

from HSE abundances and Re^Os isotope systematics of sulfideinclusions in mantle xenocrysts Chemical Geology 208 61^88

Aulbach S Griffin W L OrsquoReilly S Y amp McCandless T E(2004b) Genesis and evolution of the lithospheric mantle beneaththe Buffalo HeadTerrane Alberta (Canada) Lithos 77 413^451

Aulbach S GriffinW L Pearson N J OrsquoReilly S Y amp Doyle B(2007) Lithosphere formation in the central Slave Craton(Canada) plume subcretion or lithosphere accretion Contributionsto Mineralogy and Petrology (in press) doi 101007s00410-007-0200-1

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2001)Geochemistry of xenolithic eclogites from West Africa Part IA link between low MgO eclogites and Archean crust formationGeochimica et Cosmochimica Acta 65 1499^1527

Barth M G Foley S F amp Horn I (2002a) Partial melting inArchean subduction zones constraints from experimentally deter-mined trace element partition coefficients between eclogitic miner-als and tonalitic melts under upper mantle conditions PrecambrianResearch 113 323^340

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2002b)Geochemistry of xenolithic eclogites from West Africa part 2Origins of the high MgO eclogites Geochimica et Cosmochimica Acta

66 4325^4345Beard B L Fraracci K NTaylor L A Snyder G A Clayton R

A Mayeda T K amp Sobolev NV (1996) Petrography andgeochemistry of eclogites from the Mir kimberliteYakutia RussiaContributions to Mineralogy and Petrology 125 293^310

Becker H (1996) Geochemistry of garnet peridotite massifs fromlower Austria and the composition of deep lithosphere beneatha Palaeozoic convergent plate margin Chemical Geology 134 49^65

Becker H Jochum K P amp Carlson R W (2000) Trace elementfractionation during dehydration of eclogites from high-pressureterranes and the implications for element fluxes in subductionzones Chemical Geology 163 65^99

Bell K amp Tilton G R (2002) Probing the mantle The story fromcarbonatites EOS Transactions American Geophysical Union 83275^277

Benoit M Polvecurren M amp Ceuleneer G (1996) Trace element andisotopic characterisation of mafic cumulates in a fossil mantlediapir (Oman ophiolite) Chemical Geology 134 199^214

BleekerW (2003) The late Archean record a puzzle in ca 35 piecesLithos 71 99^134

BleekerW Ketchum JW F JacksonV A amp Villeneuve M (1999)The central Slave Basement Complex Part I Its structural topol-ogy and autochthonous core Canadian Journal of Earth Sciences 361083^1109

Blichert-Toft J amp Albare de F (1997) The Lu^Hf isotope geochemis-try of chondrites and the evolution of the mantle^crust systemEarth and Planetary Science Letters 148 243^258

Blichert-Toft J Chauvel C amp Albare de F (1997) Separation of Hfand Lu for high precision isotope analysis of rock samples by mag-netic sector-multiple collector ICP-MS Contributions to Mineralogy

and Petrology 127 248^260Brey G P amp Kolaquo hler T (1990) Geothermobarometry in four-phase

lherzolites II New thermobarometers and practical assessment ofexisting thermobarometers Journal of Petrology 31 1353^1378

Brey G P Kolaquo hlerT amp Nickel K G (1990) Geothermobarometry infour-phase lherzolites I Experimental results from 10 to 60 kbJournal of Petrology 31 1313^1352

Caporuscio F amp Smyth J (1990) Trace element crystal chemistryof mantle eclogites Contributions to Mineralogy and Petrology 105550^561

Carbno G B amp Canil D (2002) Mantle structure beneath the SWSlave Craton Canada constraints from garnet geochemistry in theDrybones Bay Kimberlite Journal of Petrology 43 129^142

Coleman R G Lee E D Beatty L B amp Brannock W W (1965)Eclogites and eclogites their differences and similaritiesGeological Society of America Bulletin 76 483^508

Creaser R A Grulaquo tter H Carlson J amp Crawford B (2004)Macrocrystal phlogopite Rb^Sr dates for the Ekati propertykimberlites Slave Province Canada evidence for multiple intru-sive episodes in the Paleocene and Eocene Lithos 76 399^414

Davies R M GriffinW L Pearson N J Andrew A S Doyle BJ amp OrsquoReilly S Y (1999) Diamonds from the deep Pipe DO-27Slave Craton Canada In Gurney J J Gurney J L Pascoe MD amp Richardson S H (eds) Proceedings of the 7th International

Kimberlite Conference CapeTown Red Roof Design pp 148^155Davies R M Griffin W L OrsquoReilly S Y amp Doyle B J (2004)

Mineral inclusions and geochemical characteristics of microdia-monds from the DO27 A154 A21 A418 DO18 DD17 and RanchLake kimberlites at Lac de Gras Slave Craton Canada Lithos 7739^55

DavisW J amp Hegner E (1992) Neodymium isotopic evidence for thetectonic assembly of Late Archean crust in the Slave Provincenorthwest Canada Contributions to Mineralogy and Petrology 111493^504

Davis W J Jones A G Bleeker W amp Grutter H (2003)Lithosphere development in the Slave craton a linked crustal andmantle perspective Lithos 71 575^589

Dawson J B (1984) Contrasting types of upper-mantle metasoma-tism In Kornprobst J (ed) Kimberlites II The Mantle and

Crust^Mantle Relationships Amsterdam Elsevier pp 289^294DePaolo D J (1981) Neodymium isotopes in the Colorado Front

Range and crust^mantle evolution in the Proterozoic Nature 291193^196

Erlank A J amp Kushiro I (1970) Potassium contents of syntheticpyroxenes at high temperatures and pressures Carnegie Institution ofWashingtonYearbook 68 433^439

Foley S F Barth M G amp Jenner G A (2000) Rutilemelt partitioncoefficients for trace elements and an assessment of the influence ofrutile on the trace element characteristics of subduction zonemagmas Geochimica et Cosmochimica Acta 64 933^938

Foley S F Buhre S amp Jacob D E (2003) Evolution of the Archaeancrust by delamination and shallow subduction Nature 421 249^252

Graham I Burgess J L Bryan D Ravenscroft P J Thomas EDoyle B J Hopkins R amp Armstrong K A (1999) Explorationhistory and geology of the Diavik Kimberlites Lac de GrasNorthwest Territories Canada In Gurney J J Gurney J LPascoe M D amp Richardson S H (eds) Proceedings of the 7th

International Kimberlite Conference Cape Town Red Roof Designpp 262^279

Green D H amp Wallace M E (1988) Mantle metasomatism byephemeral carbonatite melts Nature 336 459^462

Green T H Blundy J D Adam J amp Yaxley G M (2000) SIMSdetermination of trace element partition coefficients betweengarnet clinopyroxene and hydrous basaltic liquids at 2^75 GPaand 1080^12008C Lithos 53 165^187

GriffinW L amp OrsquoReilly SY (2007) Cratonic lithospheric mantle Isanything subducted Episodes 30(1) 43^53

Griffin W L Wass S Y amp Hollis J D (1984) Ultramafic xenolithsfrom Bullenmerri and Gnotuk maars Victoria Australia petrol-ogy of a sub-continental crust^mantle transition Journal of

Petrology 25 53^87GriffinW L Doyle B J Ryan C G Pearson N J OrsquoReilly SY

Davies R Kivi K van Achterbergh E amp Natapov LM (1999)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1871

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Layered mantle lithosphere in the Lac de Gras Area SlaveCraton composition structure and origin Journal of Petrology 40705^727

Griffin W L Pearson N J Belousova E Jackson S E vanAchterbergh E OrsquoReilly SY amp Shee S R (2000)The Hf isotopecomposition of cratonic mantle LAM-MC-ICPMS analysis ofzircon megacrysts in kimberlites Geochimica et Cosmochimica Acta

64 133^147Griffin W L OrsquoReilly S Y Doyle B J Pearson N J

Coopersmith H Kivi K Malkovets V amp Pokhilenko N(2004) Lithosphere mapping beneath the North American plateLithos 77 873^922

Grulaquo tter H S Apter D B amp Kong J (1999) Crust^mantle couplingevidence from mantle-derived xenocrystic garnets In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 307^312

Hart S R Blusztajn J Dick H J B Meyer P S ampMuehlenbachs K (1999) The fingerprint of seawater circulationin a 500-meter section of ocean crust gabbros Geochimica et

Cosmochimica Acta 63 4059^4080Harte B amp Kirkley M B (1997) Partitioning of trace elements

between clinopyroxene and garnet Data from mantle eclogitesChemical Geology 136 1^24

Heaman L M Creaser R A amp Cookenboo H O (2002) Extremeenrichment of high field strength elements in Jericho eclogitexenoliths A cryptic record of Paleoproterozoic subduction partialmelting and metasomatism beneath the Slave craton CanadaGeology 30 507^510

Heaman L M Kjarsgaard B A amp Creaser R A (2004) The tem-poral evolution of North American kimberlites Lithos 76 377^397

Heaman L M Creaser R A Cookenboo H O amp Chacko T(2006) Multi-stage modification of the northern Slave mantlelithosphere Evidence from zircon- and diamond-bearing eclogitexenoliths entrained in Jericho Kimberlite Journal of Petrology 47821^858

Hermann J Cpandler C Hack A amp Korsakov A (2006) Aqueousfluids and hydrous melts in high-pressure and ultra-high pressurerocks implications for element transfer in subduction zones Lithos92 399^417

Hills DV amp Haggerty S E (1989) Petrochemistry of eclogites fromthe Koidu Kimberlite Complex Sierra Leone Contributions to

Mineralogy and Petrology 103 397^422Hoffman P F (1989) Precambrian geology and tectonic history of

North America In Bally A amp Palmer A (eds) The Geology of

North AmericaccedilAn Overview Boulder CO Geological Society ofAmerica pp 447^512

IrelandT R Rudnick R L amp Spetsius Z (1994) Trace-elements indiamond inclusions from eclogites reveal link to Archean granitesEarth and Planetary Science Letters 128 199^213

Irvine G J Pearson D G Kjarsgaard B A Carlson R WKopylova M G amp Dreibus G (2003) Evolution of the litho-spheric mantle beneath Northern Canada a Re^Os isotope andPGE study of kimberlite-derived peridotite xenoliths fromSomerset Island and a comparison to the Slave and Kaapvaalcratons Lithos 71 461^488

Isachsen C E amp Bowring S A (1994) Evolution of the Slave cratonGeology 22 917^920

Jacob D E (2004) Nature and origin of eclogite xenoliths from kim-berlites Lithos 77 295^316

Jacob D E amp Foley S F (1999) Evidence for Archean ocean crustwith low high field strength element signature from diamondifer-ous eclogite xenoliths Lithos 48 317^336

Jacob D Jagoutz E Lowry D Mattey D amp Kudrjavtseva G(1994) Diamondiferous eclogites from Siberiaccedilremnants ofArchean oceanic-crust Geochimica et Cosmochimica Acta 585191^5207

Jacob D E Fung A Jagoutz E amp Pearson D G (2003) Petrologyand geochemistry of eclogite xenoliths from the Ekati kimberlitesarea Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA_239

Jacob D E Bizimis M amp Salters V J M (2005) Lu^Hf and geo-chemical systematics of recycled ancient oceanic crust evidencefrom Roberts Victor eclogites Contributions to Mineralogy and

Petrology 148 707^720Jerde E A Taylor L A Crozaz G Sobolev NV amp SobolevV N

(1993) Diamondiferous eclogites from Yakutia Siberiaccedilevidencefor a diversity of protoliths Contributions to Mineralogy and Petrology

114 189^202Jones A G Ferguson I J Chave A D Evans R L amp

McNeice GW (2001) The electric lithosphere of the Slave cratonGeology 29 423^426

Kelemen P B Shimizu N amp Dunn T (1993) Relative depletionof niobium in some arc magmas and the continental-crust-partitioning of K Nb La and Ce during meltrockreaction in the upper-mantle Earth and Planetary Science Letters 120111^134

Kinzler R J (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition Application to mid-ocean ridge basalt petrogenesis Journal of Geophysical Research 102853^874

Klemme S van der Laan S R Foley S F amp Guenther D (1995)Experimentally determined trace and minor element partitioningbetween clinopyroxene and carbonatite melt under upper mantleconditions Earth and Planetary Science Letters 133 439^448

Konzett J Armstrong R A Sweeney R J amp CompstonW (1998)The timing of MARID metasomatism in the Kaapvaal mantle anion probe study of zircons from MARID xenoliths Earth and

Planetary Science Letters 160 133^145Kopylova M G Russell J K amp Cookenboo H (1999) Mapping the

lithopshere beneath the north central Slave Craton In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 468^479

Krogh E (1988) The garnet^clinopyroxene iron^magnesiumgeothermometerccedila reinterpretation of existing experimentaldata Contributions to Mineralogy and Petrology 99 44^48

Kusky T M (1989) Accretion of Archean Slave Province Geology 1763^67

LeCheminant A N amp Heaman L M (1989) Mackenzie igneousevents Canada Middle Proterozoic hotspot magmatismassociated with ocean opening Earth and Planetary Science Letters

96 38^48LeCheminant A N Richardson D G DiLabio R N W amp

Richardson K A (1996) La recherche de diamants au CanadaGeological Survey of Canada Open File 3228 268pp

MacGregor I D amp Carter J L (1970) The chemistry of clinopyrox-enes and garnets of eclogite and peridotite xenoliths from theRoberts Victor mine South Africa Physics of the Earth and PlanetaryInteriors 3 391^397

MacGregor I D amp Manton W I (1986) Roberts Victor eclogitesancient oceanic crust Journal of Geophysical Research 91 14063^14079

MacKenzie J M amp Canil D (1999) Composition and thermal evo-lution of cratonic mantle beneath the central Archean SlaveProvince NWT Canada Contributions to Mineralogy and Petrology

134 313^324

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1872

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

McCandlessT E amp Gurney J J (1989) Sodium in garnet and potas-sum in clinopyroxene criteria for classifying mantle eclogites InRoss J (ed) Kimberlites and Related Rocks Carlton Blackwellpp 827^832

McDonoughW F amp Sun S-S (1995) The composition of the EarthChemical Geology 120 223^253

Navon O amp Stolper E (1987) Geochemical consequences of meltpercolation the upper mantle as a chromatographic columnJournal of Geology 95 285^307

Nikogosian I K amp Sobolev AV (1997) Ion-microprobe analysis ofmelt inclusions in olivine experience in estimating the olivine^meltpartition coefficients of trace elements Geochemistry International 35119^126

Norman M D Griffin W L Pearson N J Garcia M O ampOrsquoReilly S Y (1998) Quantitative analysis of trace element abun-dances in glasses and minerals a comparison of laser ablationinductively coupled plasma mass spectrometry solution inductivelycoupled plasma mass spectrometry proton microprobe and elec-tron microprobe data Journal of Analytical Atomic Spectrometry 13477^482

Norman M D Garcia M O amp Pietruszka A J (2005) Trace-element distribution coefficients for pyroxenes plagioclase andolivine in evolved tholeiites from the 1955 eruption of KilaueaVolcano Hawairsquoi and petrogenesis of differentiated rift-zonelavas American Mineralogist 90 888^899

OrsquoHara M J amp Herzberg C (2002) Interpretation of trace elementand isotope features of basalts Relevance of field relations petrol-ogy major element data phase equilibria and magma chambermodeling in basalt petrogenesis Geochimica et Cosmochimica Acta 662167^2191

PadghamW A amp FysonW K (1992) The Slave Province a distinctcraton CanadianJournal of Earth Sciences 29 2072^2086

Pearson D G Davies G R amp Nixon P H (1993) Geochemical con-straints on the petrogenesis of diamond facies pyroxenites from theBeni Bousera peridotite massif North Morocco Journal of Petrology34 125^172

Pearson N J Griffin W L Doyle B J OrsquoReilly S Y vanAchterbergh E amp Kivi K (1999) Xenoliths from kimberlitepipes of the Lac de gras area Slave Craton Canada In GurneyJ J Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite ConferenceCapeTown Red Roof Design pp 644^658

Perfit N R Gust D A Bence A E Arculus R J amp Taylor S R(1980) Chemical characteristics of island-arc basalts implicationsfor mantle sources Chemical Geology 30 227^256

Polat A Hofmann A W amp Rosing M T (2002) Boninite-likevolcanic rocks in the 37^38 Ga Isua greenstone belt WestGreenland geochemical evidence for intra-oceanic subductionzone processes in the early Earth Chemical Geology 184 231^254

Poudjom DjomaniY H GriffinW L OrsquoReilly SY amp Doyle B J(2005) Lithospheric domains and controls on kimberlite emplace-ment Slave Province Canada evidence from elastic thickness andupper mantle composition Geochemistry Geophysics Geosystems 6doi1010292005GC0009978

Presnall D C Gudfinnsson G H amp Walter M J (2002)Generation of mid-ocean ridge basalts at pressures from 1 to 7GPa Geochimica et Cosmochimica Acta 66 2073^2090

Schmidberger S S Heaman L M Simonetti A Creaser R A ampCookenboo H O (2005) Formation of Paleoproterozoiceclogitic mantle Slave Province (Canada) Insights from in-situHf and U^Pb isotopic analyses of mantle zircons Earth and

Planetary Science Letters 240 621^633

Schmidberger S S Simonetti A Heaman L M Creaser R A ampWhiteford S (2007) Lu^Hf in-situ Sr and Pb isotope and traceelement systematics for mantle eclogites from the Diavik diamondmine Evidence for Paleoproterozoic subduction beneath the Slavecraton Canada Earth and Planetary Science Letters 254 55^68

Schulze D J amp Helmstaedt H (1988) Coesite-sanidine eclogitesfrom kimberlite products of mantle fractionation or subductionJournal of Geology 96 435^443

Smithies R H Champion D C amp Sun S S (2004) The case forArchaean boninites Contributions to Mineralogy and Petrology 147705^721

Smyth J R Caporuscio F A amp McCormick T C (1989) Mantleeclogites evidence of igneous fractionation in the mantle Earth andPlanetary Science Letters 93 133^141

Sobolev NV amp Lavrentrsquoyev Y G (1971) Isomorphic sodium admix-ture in garnets formed at high pressure Contributions to Mineralogy

and Petrology 31 1^12Stalder R Foley S F Brey G P amp Horn I (1998) Mineral aqueous

fluid partitioning of trace elements at 900^12008C and 30^57GPa New experimental data for garnet clinopyroxene and rutileand implications for mantle metasomatism Geochimica et

Cosmochimica Acta 62 1781^1801Sun S-S amp McDonough W F (1989) Chemical and isotopic sys-

tematics of oceanic basalts implications for mantle compositionsand processes In Saunders A D amp Norris M J (eds)Magmatism in the Ocean Basins Geological Society London Special

Publications 42 313^345Taura H Yurimoto H Kurita K amp Sueno S (1998) Pressure

dependence on partition coefficients for trace elements between oli-vine and the coexisting melts Physics and Chemistry of Minerals 25469^484

Taylor L A amp Neal C R (1989) Eclogites with oceanic crustal andmantle signatures from the Bellsbank kimberlite South AfricaPart I Mineralogy petrography and whole rock chemistryJournal of Geology 97 551^567

van Achterbergh E Ryan C G amp Griffin W L (1999) Glitter online intensity reduction for the laser ablation inductively coupledplasma spectrometry Abstract 9th V M GoldschmidtConference LPI Boston MA 905 pp Abstract no 7215

van Breemen O DavisW J amp King J E (1992) Temporal distribu-tion of granitoid plutonic rocks in the Archean Slave ProvinceNorthwest Canadian Shield Canadian Journal of Earth Sciences 292186^2199

Vervoort J D Patchett P J Blichert-Toft J amp Albare de F (1999)Relationship between Lu^Hf and Sm^Nd isotopic systems in theglobal sedimentary system Earth and Planetary Science Letters 16879^99

Viljoen K S Smith C B amp Sharp Z D (1996) Stable and radio-genic isotope study of eclogite xenoliths from the Orapa kimberliteBotswana Chemical Geology 131 235^255

Walter M J (1998) Melting of garnet peridotite and theorigin of komatiite and depleted lithosphere Journal of Petrology 3929^60

Wasserburg G J Jacobsen S B DePaolo D J McCulloch M Tamp Wen T (1981) Precise determination of SmNd ratios Sm andNd isotopic abundances in standard solutions Geochimica et

Cosmochimica Acta 45 2311^2323Westerlund K J Shirey S B Richardson S H Carlson R W

Gurney J J amp Harris JW (2006) A subduction wedge origin forPaleoarchean peridotitic diamonds and harzburgites from thePanda kimberlite Slave craton evidence from Re^Os isotope sys-tematics Contributions to Mineralogy and Petrology 152 275^294

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1873

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Page 20: Origins of Xenolithic Eclogites and Pyroxenites from the Central

Both volatile-rich whole rocks display strong negativeSr anomalies Strontium can be sited in calcite or apatiteandor indicate the presence of a Sr-rich mineral that wasnot identified in thin section such as bariteFour of the six reconstructed whole-rocks in the study

of Schmidberger et al (2007) (Fig 11e) show markedLREEHREEN enrichment that is observed only in onehigh-Ca eclogite in our study (Fig 11c) suggesting thatmany of the eclogites they studied were affected by metaso-matic LREE enrichment Two samples in their study thatare not affected by this enrichment show positive Euanomalies similar to the high-Mg eclogites in this studywhereas the distinctive REE pattern of the low-Mg eclo-gites is not recognized among their samples

SR^ND^HF I SOTOPESYSTEMATICSStrontium Nd and Hf isotope data for optically puremineral separates and reconstructed whole-rocks aregiven in Table 5 Spongy rims and alteration are likely tobe mostly removed during leaching in 6N HCl in combina-tion with repeated ultrasonication Despite this rigoroustreatment the reproducibility of Nd in garnet VR43469and in cpx YK1914 is worse than for other samples Thismay be due to inhomogeneities related to late Nd diffusionpossibly from grain boundaries containing a kimberlitecomponent The reproducibility of the 176Hf177Hf data isgenerally poorer than that of Nd which may reflect a com-bination of matrix effects caused by imperfectly purifiedsolutions and sample heterogeneity Parent^daughterratios are available only from in situ analyses which arenot precise enough to allow meaningful model age orinitial isotope ratio calculations (ie at the time of kimber-lite eruption) Because the Lac de Gras kimberlites arerelatively young (55^56 Ma Graham et al 1999) age cor-rection would result in only a minor change in isotopecomposition which is inconsequential in view of the rangeof compositions displayed by most samples and hence hasno effect on the conclusions drawnClinopyroxenes in five volatile-free eclogites and a

pyroxenite have Sr isotope ratios ranging from 0701662 to0703546 this range is similar to or somewhat lower thanthat of the depleted mantle at the time of kimberliteemplacement (0703 Bell amp Tilton 2002) Clinopyroxenein the phlogopitized sample has a distinctly high 87Sr86Srof 0751186Clinopyroxene controls 56^92 of the Nd budget (Nd

concentration in cpx weighted by mode) Neodymiumisotope ratios range from 0511660 to 0514844 Thiscorresponds to eNd variation (per 10 000 deviationof 143Nd144Nd in the sample from 143Nd144Nd in the chon-dritic mantle model reservoir) from 17 to thorn43 cpx inthe low-Mg eclogites have higher eNd than those in

high-Mg eclogites Garnets have generally have higher143Nd144Nd from 0512984 (eNdfrac14thorn 7) to 0520965(eNdfrac14thorn 162) Garnets in the two volatile-rich sampleshave the highest ratios 0522790 for garnet in the calcite-bearing sample VR67112b (eNdfrac14thorn198) and 0525721 inthe phlogopitized eclogiteYK1911 (eNdfrac14thorn 255)Garnet contributes 18^32 and cpx 48^78 of the Hf

budget The proportion controlled by rutile is 4^30(Aulbach et al in preparation) Measured 176Hf177Hf incpx ranges from 0281263 in phlogopite-bearing sampleYK1911 (eHffrac1453) to 0286130 in low-Mg eclogiteVR43469 (eHffrac14thorn119) and in garnet from 0283958 inpyroxeniteYK1915 (eHffrac14thorn42) to 0295718 in low-Mg eclo-gite YK1926 (eHffrac14thorn458) Hafnium in garnet in the phlo-gopite^apatite-bearing sample is extremely radiogenic(176Hf177Hffrac14 0354367) corresponding to an eHf of thorn2532which is close to a value determined for a garnet separatefrom a Roberts Victor eclogite (eHffrac14thorn2561 Jacob et al2005) Garnet in the calcite-bearing sample has extremelyradiogenic Nd and Hf (eNdfrac14thorn193 eHffrac14thorn2516) Garnetin pyroxenite YK1915 also has radiogenic Nd and Hf(eNdfrac14thorn27 eHffrac14thorn42)Despite the large uncertainties for reconstructed whole-

rocks (calculated as the square root of the sum of squareduncertainties of two standard errors for garnet and cpxanalyses combined with 20 propagated uncertainty onthe proportion of Nd or Hf respectively contributed byeach mineral) the isotopic data are consistent in thatthey correlate with the independently determinedparent^daughter ratios (1) Low-Mg eclogites have higherSmNd reflected in the steeper positive slope of theirN-MORB-normalized REE patterns (Fig 11a) comparedwith the high-Mg eclogites (Fig 11b) Accordingly recon-structed low-Mg eclogites have higher present-day eNd

(0 to thorn47) than high-Mg eclogites (4 and 17) (2) Low-Mg eclogite YK1943 has very high Nb and LREE abun-dances that could have been added during metasomatismthat also enhanced its Hf abundance This sample has thelowest eNd and eHf of the low-Mg eclogites (0 and thorn20respectively) In contrast apparently unenriched low-Mgeclogites have more radiogenic Nd and Hf (eNdfrac14thorn47and thorn9 eHffrac14thorn128 and thorn113 respectively)The Sr^Nd isotope diagram reveals that eNd in cpx

varies from 19 to 43 at nearly constant depleted-mantle-like 87Sr86Sr (Fig 12a) In the Nd^Hf isotope diagram(Fig 12b) the reconstructed whole-rocks lie on curvedmixing lines if the lsquomixingrsquo proportions of Nd and Hffor the two lsquoend-memberrsquo minerals are very different(eg YK1911) Garnets have more radiogenic Nd andHf than coexisting cpx All reconstructed volatile-freeeclogites plot in the radiogenic field with regard to Hfor are isotopically similar to the depleted mantle Twohigh-Mg eclogites and one low-Mg eclogite have unradio-genic Nd one high-Mg eclogite has radiogenic Nd and

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1862

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Table 5 Individual and averaged Sr Nd and Hf isotope ratios obtained from garnet and cpx solutions

Sample 87Sr86Sr 2SE2SD 143Nd144Nd 2SE2SD eNd 2SE2SD 176Hf177Hf 2SE2SD eHf 2SE2SD

Low-Mg eclogites

VR43469 cp 0702661 0000010 0514844 0000006 43 01 0286130 0000012 119 04

VR43469 gt 0515508 0000014 56 03 0286506 0000012 132 04

VR43469 gt r 0515453 0000010 55 02 0286548 0000011 134 04

VR43469 gt r 0515384 0000007 54 01 0287244 0000010 158 04

VR43469 gt r 0515487 0000014 56 03 0286996 0000014 149 05

VR43469avg gt 0515458 0000109 55 21 0286823 0000715 143 253

WR VR43469 0515059 0000273 47 5 0286405 0001079 128 38

YK1926 cp 0703546 0000011 0513043 0000007 8 0 0283154 0000009 13 03

YK1926 gt 0513364 0000008 14 01 0295718 0000005 458 02

WR YK1926 0513120 0000026 9 1 0285980 0000716 113 25

YK1943 cp 0512143 0000007 10 01 0282213 0000005 20 02

YK1943 gt 0514461 0000012 36 02 0287510 0000008 168 03

YK1943 gt r 0514491 0000009 36 02 0287332 0000005 161 02

YK1943 gt r 0514517 0000010 37 02 0287290 0000007 160 02

YK1943avg gt 0514489 0000056 36 11 0287377 0000233 163 82

WR YK1943 0512625 0000228 0 5 0283138 0000879 13 31

High-Mg eclogites

VR43480 gt 0520965 0000017 162 03 0284950 0000010 77 04

VR67360 cp 0283509 0000074 26 26

VR67360 cp r 0283433 0000031 23 11

VR67360avg cp 0701662 0000018 0511660 0000010 19 02 0283471 0000108 25 38

VR67360 gt 0512984 0000101 7 20 0287429 0000180 165 64

WR VR67360 0511761 0000132 17 3 0284734 0000469 69 17

YK1949 cp 0703019 0000013 0512231 0000009 8 02 0282513 0000009 9 03

YK1949 gt 0513511 0000010 17 02 0290297 0000042 266 15

YK1949 gt r 0513508 0000012 17 02 0289947 0000015 254 05

YK1949 gt r 0513552 0000014 18 03

YK1949avg gt 0513523 0000049 17 10 0291569 0000494 311 175

WR YK1949 0512446 0000103 4 2 0284622 0000847 65 30

Volatile-rich eclogites

VR67112b gt 0522790 0000008 198 02 0354367 0000036 2532 13

YK1911 cp 0751186 0000009 0512165 0000007 9 01 0281263 0000009 53 03

YK1911 gt 0525721 0000013 255 03 0288051 0000012 187 04

WR YK1911 0513940 0000342 25 7 0281414 0000018 48 1

Pyroxenites

YK1914 cp 0703015 0000034 0513283 0000041 13 08

YK1914 cp r 0702956 0000005 0513200 0000027 11 05

YK1914 cp r 0513527 0000059 17 11

YK1914 cp r 0513045 0000010 8 02

YK1914avg cp 0702986 0000083 0513264 0000402 12 78

YK1915 gt 0514003 0000011 27 02 0283958 0000021 42

Analyses of full procedural replicates are designated with an lsquorrsquo after sample name SE standard error (internal errorsingle analyses) SD standard deviation (external error averaged multiple analyses) Italics used for reconstructed whole-rocks (mineral isotopic compositions and concentrations weighted by modes) Uncertainties for reconstructed whole-rocksinclude 2SE or 2SD uncertainties on each mineralrsquos isotope composition combined with 20 uncertainty in the amountof Nd or Hf contributed by the constituent minerals eNd and eHf are per 10 000 deviation from model chondritic mantle[model of Wasserburg et al (1981) and Blichert-Toft amp Albarede (1997) respectively]

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1863

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

εHf

εNd

Ekati eclogitesminus100

0

100

200

300

400

500

minus40 0 40 80 120 160 200 240 280

gt in YK1915

gt in VR67112bεNd=198 εHf=2532

minus40

0

40

80

120

160

minus40 minus20 0 20 40 60

minus60

minus40

minus20

0

20

40

070 071 072 073 074 075 076

North Americankimberlite field

εNd

87Sr86Sr

cpx in YK1914

Ekatieclogites

DM

DM

DM

DM

(a)

(b)

pyroxenites

low-Mg

high-Mg

volatile-rich

eclogites

Diavik low-TA154 S kimberlite

cpx in YK1911

Fig 12 (a) Sr^Nd for cpx and (b) Nd-Hf isotope diagram for garnet cpx and reconstructed whole-rocks (symbols with bold outline) Finedashed lines in (b) connect data points from the same sample with garnets having more radiogenic isotopic compositions Uncertainties forreconstructed whole-rocks combine 20 modal uncertainty with uncertainties on each mineralrsquos isotope composition Error bars on individualminerals may be smaller than the symbols Also shown are the isotopic compositions of the depleted mantle (DM DePaolo 1981 Griffin et al2000 Bell amp Tilton 2002) Range of 87Sr86Sr in eclogites from Ekati (Lac de Gras area 100 km north of Diavik Jacob et al 2003) in NorthAmerican kimberlites and compositions of low-temperature eclogites from Diavik and of Lac de Gras A154 South kimberlite fromSchmidberger et al (2007)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1864

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

one low-Mg eclogite has eNd similar to the depleted mantle(Fig 12b)Clinopyroxene in the phlogopitized eclogite YK1911

has unradiogenic Nd and Hf and strongly radiogenicSr whereas the coexisting garnet has radiogenic Nd andHf Although this sample contains 50 vol phlogopiteits whole-rock isotopic composition was calculated

from garnet and cpx alone because the contributionfrom phlogopite to the Hf and Nd budget is 52 and51 respectively (unpublished data) The whole-rockhas radiogenic Nd but has lower eHf than anyother sample analysed in accord with the extreme Hfabundance (53 ppm) and low LuHf of the cpx

DISCUSS IONOrigins of eclogites beneath the SlaveCratonHigh-Mg and high-Ca eclogites

With regard to their major- and trace-element contentsthe high-Mg and high-Ca eclogites from Lac de Gras arebroadly similar to variably evolved oceanic gabbros(Fig 13b and c) which may indicate an origin as subductedoceanic crust Compatible trace-element abundancesare useful to deduce potential crustal protoliths as theseelements are less susceptible to post-formation modification(Jacob 2004) The flat HREE patterns in the high-Mg andhigh-Ca eclogites are compatible with a low-pressureorigin as the oceanic crust is a product of moderate degreesof partial melting mainly within the spinel stability field(eg Presnall et al 2002) whereas melting with residualgarnet would qualitatively be expected to produce stronglyfractionated REE patterns Positive Sr and Pb anomaliesalthough compatible with a plagioclase signature couldalso be due to cryptic metasomatism (Dawson 1984) ofthe eclogites during their residence in the lithosphericmantle These anomalies are probably not related to kim-berlite infiltration as the whole-rock eclogite compositionswere reconstructed from analyses of fresh mineral coresLow ZrSm similar to that in primitive gabbros is

observed for many of the high-Mg and high-Caeclogites (Fig 13b and c) and is independent of the ratioof garnetcpx and whether or not the whole-rock wasreconstructed with rutile (Electronic Appendix 10) Thenegative trend of Mg-number against Yb in garnet(Fig 14) argues against igneous processes involving garnetaccumulation (Taylor amp Neal 1989) as suggested forexample for some high-Mg eclogites from West Africa(Barth et al 2002b) Rather the anti-correlation suggeststhat olivine fractionation played a role in the formation ofthe protoliths Fractionating olivine would excludeYb (Dolivinebasalt

frac14 002 Nikogosian amp Sobolev 1997)and incorporate Mg relative to Fe (Dfrac1415 and 05 respec-tively Taura et al1998) leading to progressive depletion ofMg and enrichment of Yb in the residual melt and henceanti-correlated Mg and Yb in the whole-rock This isinherited after eclogitization All eclogite types (exceptVR43452) plot along the olivine fractionation trend sug-gesting that they formed by a similar processHigh-Mg eclogites have lower

PREE than high-Ca

eclogites and some appear to have positive Eu anomalies

high-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

gabbrosSE Indian ridge

gabbrosOman ophiolite

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

high-Mg

RockNMORB(a)

VR43452

YK1949

low-Mg

boninite-like melts

01

001

1

10

(b)

01

001

1

10

(c)

01

001

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

Fig 13 Trace-element patterns of (a) low-Mg (b) high-Mg and(c) high-Ca eclogites compared with boninite-like volcanic rocksfrom the Isua greenstone belt (Polat et al 2002) gabbros from theOman ophiolite (Benoit et al 1996) and the SE Indian Ridge (Hartet al 1999) normalized to N-MORB of Sun amp McDonough (1989)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1865

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(EuEu up to 14) similar to more primitive cumulates(troctolite gabbros and olivine gabbros) reported in the lit-erature (Fig 13b) Although the sizes of these Eu anomaliesare not outside the analytical uncertainty their correlationwith CeYb in reconstructed eclogites supports their signif-icance (r2frac14066 nfrac14 4) The more primitive nature ofhigh-Mg eclogites is consistent with higher Mg-numbersrelative to high-Ca eclogites and may be due to olivineaccumulation during formation of their protolithsThe two samples with the lowest HREE abundanceshave the highest EuEu which is compatible with varia-tions in plagioclase modesMore evolved protoliths with higher pyroxeneplagio-

clase plus trapped melt could account for small or absentpositive Eu anomalies and higher

PHREE of high-Ca

eclogites (Fig 13c) Although the reason for the frequentassociation of graphite or diamond with high-Ca eclogitesis unclear (Aulbach et al 2003) the frequent occurrence ofkyanite in high-Ca eclogites also observed at Ekati justnorth of Lac de Gras has been argued to support a crustalprotolith (Jacob 2004) Kyanite andor corundum havebeen suggested to exsolve from high-temperature high-Alclinopyroxenes (Caporuscio amp Smyth 1990) Eclogitesfrom Lac de Gras show no exsolution textures suggestiveof such an origin although recrystallization can obscureor destroy microstructural evidence for exsolution(eg Griffin et al 1984)Modelling shows that qualitatively plagioclase accumu-

lation from a low-pressure melt derived from MORBmantle can account for positive Pb Sr and Eu and negativeZr anomalies relative to N-MORB whereas cpx accumula-tion or trapped melt increases Zr Y MREE and HREE

abundances (Fig 15) Some melt addition (for exampleby entrapment) seems to be required to increase all trace-element abundances to those observed for the eclogitesAlthough high-Ca and high-Mg eclogites could representshallow and deep portions of the same oceanic crust thehigher average temperatures obtained for the high-Caeclogites the frequent occurrence of carbon in high-Caeclogites and its absence in high-Mg eclogites the strongLREEHREE fractionation observed for six of seven high-Ca eclogites and the narrow range of compositions of thehigh-Mg eclogites suggest that they may have protolithsthat are not cogenetic Some discrepancies in particularaffecting the incompatible elements may be due to post-crystallization processes acting upon the protoliths suchas dehydration melting and later metasomatism that isevident in peridotites from the same locality (Griffin et al1999)Low K2OTh for high-Ca and high-Mg eclogites

(average 08 Table 4) relative to assumed crustal

0

2

4

6

8

10

30 40 50 60 70 80

Mg

VR43452

Yb

GARNET

pyroxenites

low-Mghigh-Mghigh-Ca

eclogites

VR19673ecl-g2

Fig 14 Garnet Mg-number vs Yb (ppm including 2s uncertainty)Lines qualitatively show separate correlations for eclogites (exceptoutlierVR43452) and pyroxenites (except outlierVR19673ecl-g2)

high-Ca

high-Mg

001

01

1

10

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

melt08 melt02 cryst cpx01 melt09 cryst

plag

(a)

001

01

1

10(b)

RockNMORB

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 15 Reconstructed (a) high-Ca and (b) high-Mg eclogitescompared with compositions of plagioclase and cpx crystallizingfrom a MORB melt modelled using distribution coefficients ofNorman et al (2005)The MORB was calculated as a 10 equilibriummelt of MORB pyrolite [composition from Kelemen et al (1993)except Pb Pr Sr Y and Lu which were calculated as residues from1 fractional melting of pyrolite from McDonough amp Sun (1995)using melt modes of Kinzler (1997) and distribution coefficients ofNorman et al (2005)] Also shown are mixtures of crystals (consistingof 50 plagioclase plus 50 cpx) and melt in the proportions 0901and 0208

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1866

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

precursors (eg gabbros SE Indian Ridge average 22Hart et al 1999) may reflect the preferred loss of fluid-mobile elements during dehydration (Becker et al 2000)although large uncertainties on these low-abundance ele-ments in the constituent minerals and reconstructedwhole-rocks makes this interpretation less than robustThe REE patterns of most high-Mg eclogites are too flatto represent melting residues as partial melting shouldproduce low LREEHREE (Electronic Appendix 10) Incontrast the marked LREE depletion of five of sevenhigh-Ca eclogites (LaLufrac14 02^08) contrasts with that ofassumed evolved gabbroic protoliths (LaLufrac1415^32Hart et al 1999) and can be modelled as a consequence ofpartial melting during subduction after eclogitization (iein the presence of garnet which retains HREE relative toLREE eg Barth et al 2002a) has taken place (Aulbachet al 2003)Our results are compatible with previous studies on

eclogite xenoliths from the Slave Craton includingthose from the same kimberlite that have linked thesesamples or some of these samples to Paleoproterozoicsubduction events (the 195^191Ga formation of theHottah terrane and the 18Ga formation of the GreatBear magmatic arc) in the Slave craton (Griffin et al1999 Kopylova et al 1999 Heaman et al 2002 2006Schmidberger et al 2005 2007)The compositional features discussed above although

compatible with crustal protoliths are not necessarilydefinitive of a subduction origin However if direct crystal-lization from basaltic melts at high pressure (ie mantledepths) is invoked instead specific conditions would haveto be met to satisfy the observations The Mg-number ineclogites from Lac de Gras are lower than those expectedof primary high-pressure mantle melts (eg Walter 1998OrsquoHara amp Herzberg 2002) requiring crystallization athigh pressures after olivine fractionation This originwould further require shielding from peridotitic wallrockwith which the fractionated melts would no longer be inequilibrium (Barth et al 2001) In such a setting the lowrelative abundances of fluid-mobile elements comparedwith those of less fluid-mobile elements and of LREErelative to HREE (Table 4) would reflect an expulsion offluids during crystallization of the protoliths and latermelt mobilization respectively The eclogites clearly donot represent garnet plus cpx cumulates which wouldrequire a positive correlation of Mg-number with Yb thatis at variance with the observed negative correlation ineclogitic garnets from this study (Fig 14)

Low-Mg eclogites

Garnets in low-Mg eclogites compositionally overlap withthose in massive eclogites from Jericho (NW of the Lacde Gras kimberlites) that were suggested to have a high-pressure intrusive origin based on the similarity of theirmajor-element compositions to those of Group A eclogites

fromsouthernAfrica forwhichamantleoriginhasbeenpro-posed (Kopylova et al1999) Garnets in low-Mg eclogites inthe present study have major-element compositions moresimilar to garnets in groupB rather thangroupAeclogitesThe positive slope in N-MORB-normalized HREE

abundances displayed by the low-Mg eclogites is clearlydistinct from other eclogite types and is similar to those ofboninite-like second-stage melts that are closely associatedwith ultramafic volcanics (Smithies et al 2004) (Fig 13a)The range of Mg-numbers in the low-Mg eclogites (55^66)overlaps with that of Archaean second-stage melts fromIsua (58^75) and Abitibi (52^82 Smithies et al 2004)These similarities might indicate that the low-Mg eclogiteshave an origin in a previously depleted mantle source suchas supra-subduction zone mantle Subchondritic ZrHf ofthe low-Mg eclogites contrasts with the supra-chondriticZrHf of the high-Mg and high-Ca eclogites (Aulbach et

al in preparation) and is consistent with previous cpx-controlled depletion in an arc mantle source This sourcecould have been subsequently remelted to generate theboninite-like precursor to the low-Mg eclogites as sug-gested for the protoliths of eclogites from Udachnaya(Jacob amp Foley 1999)Although some second-stage melts to which the low-Mg

eclogites are similar are closely associated with komatiitesand could contain a plume component the radiogenic ordepleted mantle-like Nd and Hf and unradiogenic Sr inthe low-Mg eclogites do not support plume involvement

Trace-element and Sr^Nd^Hf isotopeconstraints on post-formationmetasomatismSeveral low-Mg and high-Ca eclogites show negativeslopes in the LREE which may be indicative of interactionwith small amounts of LREE-enriched melt (Navon ampStolper 1987) amounting to cryptic metasomatism(ie not accompanied by crystallization of new phases) inthe nomenclature of Dawson (1984) High Nb abundancesrelative to elements of similar compatibility displayed bysome low-Mg and high-Ca eclogites (Fig 10) could resultfrom overestimation of rutile modes However high-CaeclogiteVR40345 is rutile-free and has been reconstructedwithout rutile its high Nb abundance is accompanied byLREE enrichment indicating that metasomatic additionmay be responsibleLow-Mg eclogites show a positive correlation of La with

Nb (Electronic Appendix 11a) indicating concomitantaddition of LREE and Nb This suggests that the metaso-matic agent was a solute-rich fluid or silicate silicic melt ashydrous fluids and carbonatites do not carry enough HFSEto affect the budget of eclogites (Green amp Wallace 1988Klemme et al 1995 Adam et al 1997 Stalder et al 1998)Of the samples that are enriched in Nb all but one (high-Ca eclogiteVR50909) show a simultaneous enrichment in

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1867

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Hf (Electronic Appendix 11b) Together these trace-element relationships suggest that both SmNd and LuHfwere lowered in these samples Their low parent^daughterratios contrast with higher ratios in unenriched samplesand resulted in the ingrowth of a range of Nd and Hf iso-tope compositions Compared with the low-Mg and high-Ca eclogites the high-Mg eclogites have a narrow range ofLa and Nb concentrations which correlates with a narrowrange in Nd and Hf isotope compositions (Fig 12)Precise model ages cannot be calculated because parent^

daughter ratios are available only from in situ analyseswhich have relatively large uncertainties Howeverevidence for long-term enrichment and depletion may bediscerned Low-Mg eclogites even those that are enrichedin LREE have relatively high SmNd as a result of havingoverall positive slopes in their REE patterns relative toMORB whereas the high-Mg eclogites have relativelyflat REE patterns and therefore lower SmNd (Fig 11)This translates into lower eNd for high-Mg eclogites thanfor low-Mg eclogites In contrast they show no systematicdifference with respect to 176Hf177Hf Because the low- andhigh-Mg eclogites are proposed to have different origins asdiscussed above and therefore did not necessarily have thesame initial isotopic composition and formation age theseisotopic differences cannot be used to estimate their agesA long-term coupling of trace-element and isotopic

abundances is also indicated by the fact that low-Mg eclo-gite YK1943 which has experienced Nb and LREE addi-tion has lower eNd (0) and eHf (thorn13) than low-Mg eclogiteVR43469 with higher LuHf and SmNd (eNdfrac14thorn47eHffrac14thorn128) If low-Mg eclogites YK1943 and VR43469originally had identical initial Nd and Hf isotope composi-tions and if the parent^daughter ratios are accurate andwere modified only once during their evolution shortlyafter metamorphism (when their isotopic compositionwould have been reset) this overprint can be calculatedfor both isotopic systems to have occurred at 17GaThis would suggest that Hf and Nd were added duringthe same event In contrast high-Mg eclogites VR67360and YK1949 would have had identical Ndisotope compositions at 10Ga whereas their Hf isotopecompositions are disturbed in that the sample with thehigher parent^daughter ratio has a lower 176Hf177Hf thanthe sample with the lower parent^daughter ratio Theseresults suggest that the LREE and HFSE enrichmentobserved in different samples occurred during separateevents that did not affect the eclogites equallyExtreme HFSE enrichment leading to the growth of

zircon and rutile and LREE addition leading to thegrowth of apatite in eclogites from the Jericho kimberlitein the northern Slave craton have been ascribed to twoseparate metasomatic events at around 18Ga (quasi-coincident with the age of formation and metamorphism)and 13^10Ga respectively (Heaman et al 2002 2006

Schmidberger et al 2005) The authors associated theseevents with Paleoproterozoic subduction and aMesoproterozoic thermal disturbance in the Slave litho-spheric mantle possibly linked to the Mackenzie igneousevent Mesoproterozoic Nd model ages mostly between13 and 14Ga were reported for eclogites from thenearby Ekati kimberlites (Jacob et al 2003) Although wedo not observe zircon or apatite in volatile-free eclogitesfrom Lac de Gras and the age constraints obtained fromour data are weak they support a link between theseevents described for other eclogite populations and themetasomatic enrichment observed in our studyThe Sr isotope data in the present study are similar to

those of Schmidberger et al (2007) who suggested that thedepleted-mantle-like values are consistent with derivationof the protoliths from an oceanic mantle source However87Sr86Sr in volatile-free eclogites in the present study isremarkably constant considering their Nd and Hf isotopevariability (Fig 12) and the variability of trace-elementpatterns (Fig 10) Given the evidence for dehydration andmelting in many of the eclogites fractionation of highlyincompatible Rb from Sr and differential ingrowth ofradiogenic Sr would be expectedThis may indicate a rela-tively recent overprint and considering the similarityof 87Sr86Sr to present-day depleted mantle which is signif-icantly lower than 87Sr86Sr of North American kimber-lites a juvenile (asthenospheric) source

PyroxenitesPyroxenites from Lac de Gras often show relationshipsin their major-element contents that are different fromthose observed in the eclogites For example pyroxeniticgarnets show a negative correlation between CaO andNa2O contents whereas these oxides are positively corre-lated for garnets in eclogites (Fig 5) CaO is well correlatedwith C2O3 in pyroxenitic garnets but not in eclogitic gar-nets (Fig 4) Mg-number is positively correlated withYb abundance for pyroxenitic garnets but anti-correlatedfor eclogitic garnets (Fig 14) This suggests that the pyrox-enites in the present study are not related to the eclogitesby fractionation or accumulation processes as proposedfor the origin of some pyroxenites (Foley et al 2003)Pyroxenites have also been interpreted as the products

of crystal segregation in magma channels with alteredoceanic crust as the magma source this would explaintheir Eu anomalies and stable and radiogenic isotopecompositions (Pearson et al 1993 Becker 1996) As thehigh-pressure liquidus phases of most terrestrial basaltsare cpx and garnet (OrsquoHara amp Herzberg 2002) partialmelts of eclogites with basaltic composition wouldcrystallize cpx and garnet but not opx More than half ofthe pyroxenites from Lac de Gras contain opx suggestingthat they probably did not crystallize directly fromeclogite-derived melts

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1868

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Low-Cr websterites and pyroxenites could be crystalli-zation products of mafic melts that interacted with perido-titic mantle thus becoming more refractory The positivecorrelation of Yb and Mg-number in garnet would allowan origin by fractional crystallization from such meltsAlternatively they formed by reaction of silicic (possiblyeclogite-derived) melts with mantle peridotites(eg Kelemen et al 1993 Aulbach et al 2002) This agreeswith the similarity of hypothetical melts in equilibriumwith pyroxenitic garnets in most samples to eclogite-derived melts (Electronic Appendix 12) Niobium concen-trations in all hypothetical melts are much higher than ineclogite-derived melts regardless of whether or not rutileis assumed to be in the residue The same is true for LaCe and Zr concentrations in some samples which mayindicate that NbZr and LREE were added to thepyroxenites some time after their formation as discussedin the previous section for some eclogites Because mini-mum ages for pyroxenitic rutile with an unradiogenicHf isotope composition (calculated by projecting the176Hf177Hf to the model depleted mantle isotopic evolutioncurve assuming zero ingrowth) range to 178066Ga(Aulbach et al in preparation) it appears likely that thepyroxenites were not formed recently (eg during kimber-lite-related or precursory melt migration) and that some ofthe samples are at least MesoproterozoicGarnet pyroxeniteYK1952 consists of schlieren of equili-

brated garnet pyroxenite in a lsquomatrixrsquo of spongy cpxfine cpx^opx intergrowths embayed garnet and meltpatches (Fig 2f) Garnet in the pyroxenite schlieren hasvery low LREE and HFSE abundances Orthopyroxenein the matrix has distinctly higher Mg-number andCr-number and contains more CaO than opx from thepyroxenite schlieren whereas opx rims in the schlierenare compositionally similar to opx in the matrix The con-siderable compositional and textural disequilibrium inthis sample suggests that this is a young feature causedby reaction with fluids shortly before or at the time ofentrainmentAs a consequence of the small sample sizes of most of

the pyroxenites studied here a comprehensive isotopicdataset is not available Clearly more such data areneeded in addition to stable isotope data to obtain betterconstraints on the origins of this rock type Consideringthe microstructural and compositional variability ofthe pyroxenites and major-element relationships oppositeto those observed for eclogites it appears certain that pyr-oxenites form by a variety of processes some of which areunrelated to the formation of eclogites and their protoliths

Volatile-rich eclogitesVolatile-rich eclogites from Lac de Gras display the mostextreme Hf^Nd isotope systematics of all the analysedsamples Garnets in phlogopitized (YK1911) and in cal-cite-bearing eclogite (VR67112b) have the most radiogenic

Nd (eNdfrac14 2486 and 1927 respectively) Garnet inVR67112b has extremely high eHf (2516) Clinopyroxene inYK1911 has the lowest eHf measured for any mineralseparate from Lac de Gras (53) in accord with the veryhigh Hf abundance in its cpx (5 ppm) In contrastgarnet in this sample has radiogenic Hf (eHffrac14 136) Thisdiscrepancy is at least in part due to their relatively lowequilibration temperatures (7698C and 8558C respec-tively) which allowed the isotopic compositions of cpxand garnet to evolve separately at their low andhigh parent^daughter element ratios respectively forextended periods of time and possibly since formationAccordingly the cpx in YK1911 gives a minimum Hf ageof 27GaSlab-derived fluids could have been introduced during

the 27Ga collision of the western Slave Craton domainincluding the thick plume-derived deep lithosphericmantle layer and the eastern domain with its highlydepleted lithosphere This would also be consistent withthe observation that volatile-rich eclogites occur at shallowdepths in the mantle (TKroghfrac14 685^8558C correspondingto depths of 85^115 km along a 40 mWm2 geotherm)where the bulk of dehydration and production of solute-rich melts in a subducting slab might be expected to occur(Becker et al 2000 Hermann et al 2006) Their negativeEu anomalies might indicate that the fluid was derivedfrom the shallow portion of the slab that experiencedplagioclase fractionation The presence of carbonate inVR67112b requires a carbonated sourceHigh 87Sr86Sr in cpx in YK1911 is probably the conse-

quence of isotopic equilibration with high-RbSr phlogo-pite after long-term radiogenic ingrowth while Hfaddition led to retarded ingrowth of radiogenic HfThese systematics combined with radiogenic Nd appear

inconsistent with the involvement of sediments in the pet-rogenesis of this sample (eg Perfit et al 1980 Vervoortet al 1999) or a hydrous fluid such as those responsible forMARID metasomatism in the Kaapvaal craton (Konzettet al 1998)Whatever the origin of the fluids the apparentHf mobility suggests that they were not in equilibriumwith a rutile-bearing source and also that they weresolute-rich in nature (ie hydrous melts able to dissolveHFSE) and the relatively high HREE abundances of thevolatile-rich eclogites indicate a garnet-free source suchas amphibolite This allowed evolution of the sample atrelatively high time-integrated SmNd

SUMMARY AND CONCLUSIONS

(1) The Lac de Gras kimberlites carried abundant eclo-gite and pyroxenite xenoliths to the surface Threemain eclogite types are distinguished in terms ofmajor-element composition and trace element

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1869

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

patterns high-Mg high-Ca and low-Mg eclogitesThe garnets of all eclogite types have anti-correlatedMg-number and Yb pointing to olivine fractionationas a controlling factor in the formation of their pre-cursor rocks Their flat or positively sloped HREEpatterns qualitatively indicate that the melt parentalto the precursor rocks was not in equilibrium with agarnet-bearing residue

(2) Reconstructed high-Mg eclogites and high-Caeclogites have flat HREE patterns positive Pb andSr anomalies and low ZrSm similar to oceanic gab-bros reported in the literature High-Mg eclogiteshave variably low HREE contents and some havesmall positive Eu anomalies possibly suggestive ofmore primitive relatively plagioclase-rich protolithswhereas high-Ca eclogites have higher HREE abun-dances and small or absent Eu anomalies that requiremore cpx- andor melt-rich protoliths if a subductionorigin is assumed

(3) Low-Mg eclogites have distinctive trace-element pat-terns that show a marked positive slope in the HREEand subchondritic ZrHf They resemble someArchaean second-stage melts (lsquoboninitesrsquo) and theirprotoliths appear to have formed from a previouslydepleted source

(4) After formation the eclogites were subject to a varietyof processes Most have low ratios of fluid-mobile vsless fluid-mobile elements possibly as a result of dehy-dration of the protoliths In addition some of the low-Mg and high-Ca eclogites have low LREEHREEconsistent with partial melt lossWhereas the low-Mgeclogites show concomitant enrichment in La Nb andZr some high-Mg and high-Ca eclogites show evi-dence for variable and unsystematic (decoupled)LREE and Nb addition

(5) Isotopically the trace-element characteristics of thedifferent eclogite types translate into lower eNd forhigh-Mg eclogites (negative slope in MREEfrac14 lowerSmNd) than for low-Mg eclogites (positive slope inMREEfrac14higher SmNd) Within the low-Mg groupsamples that show evidence for metasomatism (lowLuHf and SmNd) have lower eNd and eHf than asample that was apparently not metasomatized BothNd and Hf addition can be dated to 17Ga if identi-cal initial isotope compsitions are assumed for thelow-Mg eclogites In contrast the relatively constantand depleted mantle-like 87Sr86Sr of eclogitic cpxdespite variable Nd and Hf isotope compositions andparent^daughter ratios suggests recent input of juve-nile material

(6) Garnet and cpx in pyroxenites show major-elementrelationships different from those in eclogitesTherefore these two rock types are probably notrelated by fractionation processes The presence of

opx precludes direct crystallization from slab-derivedmelts Pyroxenites may be the product of (slab-derived) melts hybridized during assimilation of peri-dotitic mantle

(7) Garnets in phlogopite- and carbonate-bearing eclo-gites have extreme isotopic compositions includingthe most radiogenic Nd and Hf in the datasetNegative Eu anomalies may point to the involvementof a crustal protolith Highly radiogenic Sr radiogenicNd and unradiogenic Hf in the reconstructedphlogopite-bearing sample combined with highHREE abundances and evidence for Hf mobilitymay indicate involvement of solute-rich fluids origi-nating in a garnet- and rutile-free source

ACKNOWLEDGEMENTSHelp in the clean laboratories and with the analyticalfacilities by S Elhlou C Lawson A Sharma P Wielandand M Zhang is greatly appreciated Bill Griffin isthanked for his support and for agreeing to reviewseveral versions of the manuscript despite disagreeingwith many of the conclusions presented here The paperbenefited greatly from the detailed and constructivereviews provided by the editors Gareth Davies andMarjorie Wilson and by Matthias Barth Larry Heamanand an anonymous reviewer This work was funded by aMacquarie University International PostgraduateAward and Postgraduate Research Fund (SA) by anARC SPIRT grant sponsored by Kennecott Canada Incand by ARC Large and Discovery Grants to SYOrsquoRAnalytical data were obtained using instrumentationfunded by ARC LIEF and DEST Systemic InfrastructureGrants industry partners and Macquarie UniversityThis is Publication 481 from the ARC National KeyCentre for Geochemical Evolution and Metallogeny ofContinents (wwwesmqeduauGEMOC)

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online

REFERENCESAdam J GreenT H Sie S H amp Ryan C G (1997) Trace element

partitioning between aqueous fluids silicate melts and mineralsEuropeanJournal of Mineralogy 9 569^584

Aulbach S Stachel T Viljoen K S Brey G P amp Harris J W(2002) Eclogitic and websteritic diamond sources beneath theLimpopo Beltccedilis slab-melting the link Contributions to Mineralogy

and Petrology 143 56^70Aulbach S GriffinW L Pearson N J OrsquoReilly S Y Kivi K amp

Doyle B (2003) Origins of eclogites beneath the central SlaveCraton Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA-0011

Aulbach S Griffin W L Pearson N J OrsquoReilly S Y amp Kivi K(2004a) Mantle formation and evolution Slave Craton constraints

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1870

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

from HSE abundances and Re^Os isotope systematics of sulfideinclusions in mantle xenocrysts Chemical Geology 208 61^88

Aulbach S Griffin W L OrsquoReilly S Y amp McCandless T E(2004b) Genesis and evolution of the lithospheric mantle beneaththe Buffalo HeadTerrane Alberta (Canada) Lithos 77 413^451

Aulbach S GriffinW L Pearson N J OrsquoReilly S Y amp Doyle B(2007) Lithosphere formation in the central Slave Craton(Canada) plume subcretion or lithosphere accretion Contributionsto Mineralogy and Petrology (in press) doi 101007s00410-007-0200-1

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2001)Geochemistry of xenolithic eclogites from West Africa Part IA link between low MgO eclogites and Archean crust formationGeochimica et Cosmochimica Acta 65 1499^1527

Barth M G Foley S F amp Horn I (2002a) Partial melting inArchean subduction zones constraints from experimentally deter-mined trace element partition coefficients between eclogitic miner-als and tonalitic melts under upper mantle conditions PrecambrianResearch 113 323^340

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2002b)Geochemistry of xenolithic eclogites from West Africa part 2Origins of the high MgO eclogites Geochimica et Cosmochimica Acta

66 4325^4345Beard B L Fraracci K NTaylor L A Snyder G A Clayton R

A Mayeda T K amp Sobolev NV (1996) Petrography andgeochemistry of eclogites from the Mir kimberliteYakutia RussiaContributions to Mineralogy and Petrology 125 293^310

Becker H (1996) Geochemistry of garnet peridotite massifs fromlower Austria and the composition of deep lithosphere beneatha Palaeozoic convergent plate margin Chemical Geology 134 49^65

Becker H Jochum K P amp Carlson R W (2000) Trace elementfractionation during dehydration of eclogites from high-pressureterranes and the implications for element fluxes in subductionzones Chemical Geology 163 65^99

Bell K amp Tilton G R (2002) Probing the mantle The story fromcarbonatites EOS Transactions American Geophysical Union 83275^277

Benoit M Polvecurren M amp Ceuleneer G (1996) Trace element andisotopic characterisation of mafic cumulates in a fossil mantlediapir (Oman ophiolite) Chemical Geology 134 199^214

BleekerW (2003) The late Archean record a puzzle in ca 35 piecesLithos 71 99^134

BleekerW Ketchum JW F JacksonV A amp Villeneuve M (1999)The central Slave Basement Complex Part I Its structural topol-ogy and autochthonous core Canadian Journal of Earth Sciences 361083^1109

Blichert-Toft J amp Albare de F (1997) The Lu^Hf isotope geochemis-try of chondrites and the evolution of the mantle^crust systemEarth and Planetary Science Letters 148 243^258

Blichert-Toft J Chauvel C amp Albare de F (1997) Separation of Hfand Lu for high precision isotope analysis of rock samples by mag-netic sector-multiple collector ICP-MS Contributions to Mineralogy

and Petrology 127 248^260Brey G P amp Kolaquo hler T (1990) Geothermobarometry in four-phase

lherzolites II New thermobarometers and practical assessment ofexisting thermobarometers Journal of Petrology 31 1353^1378

Brey G P Kolaquo hlerT amp Nickel K G (1990) Geothermobarometry infour-phase lherzolites I Experimental results from 10 to 60 kbJournal of Petrology 31 1313^1352

Caporuscio F amp Smyth J (1990) Trace element crystal chemistryof mantle eclogites Contributions to Mineralogy and Petrology 105550^561

Carbno G B amp Canil D (2002) Mantle structure beneath the SWSlave Craton Canada constraints from garnet geochemistry in theDrybones Bay Kimberlite Journal of Petrology 43 129^142

Coleman R G Lee E D Beatty L B amp Brannock W W (1965)Eclogites and eclogites their differences and similaritiesGeological Society of America Bulletin 76 483^508

Creaser R A Grulaquo tter H Carlson J amp Crawford B (2004)Macrocrystal phlogopite Rb^Sr dates for the Ekati propertykimberlites Slave Province Canada evidence for multiple intru-sive episodes in the Paleocene and Eocene Lithos 76 399^414

Davies R M GriffinW L Pearson N J Andrew A S Doyle BJ amp OrsquoReilly S Y (1999) Diamonds from the deep Pipe DO-27Slave Craton Canada In Gurney J J Gurney J L Pascoe MD amp Richardson S H (eds) Proceedings of the 7th International

Kimberlite Conference CapeTown Red Roof Design pp 148^155Davies R M Griffin W L OrsquoReilly S Y amp Doyle B J (2004)

Mineral inclusions and geochemical characteristics of microdia-monds from the DO27 A154 A21 A418 DO18 DD17 and RanchLake kimberlites at Lac de Gras Slave Craton Canada Lithos 7739^55

DavisW J amp Hegner E (1992) Neodymium isotopic evidence for thetectonic assembly of Late Archean crust in the Slave Provincenorthwest Canada Contributions to Mineralogy and Petrology 111493^504

Davis W J Jones A G Bleeker W amp Grutter H (2003)Lithosphere development in the Slave craton a linked crustal andmantle perspective Lithos 71 575^589

Dawson J B (1984) Contrasting types of upper-mantle metasoma-tism In Kornprobst J (ed) Kimberlites II The Mantle and

Crust^Mantle Relationships Amsterdam Elsevier pp 289^294DePaolo D J (1981) Neodymium isotopes in the Colorado Front

Range and crust^mantle evolution in the Proterozoic Nature 291193^196

Erlank A J amp Kushiro I (1970) Potassium contents of syntheticpyroxenes at high temperatures and pressures Carnegie Institution ofWashingtonYearbook 68 433^439

Foley S F Barth M G amp Jenner G A (2000) Rutilemelt partitioncoefficients for trace elements and an assessment of the influence ofrutile on the trace element characteristics of subduction zonemagmas Geochimica et Cosmochimica Acta 64 933^938

Foley S F Buhre S amp Jacob D E (2003) Evolution of the Archaeancrust by delamination and shallow subduction Nature 421 249^252

Graham I Burgess J L Bryan D Ravenscroft P J Thomas EDoyle B J Hopkins R amp Armstrong K A (1999) Explorationhistory and geology of the Diavik Kimberlites Lac de GrasNorthwest Territories Canada In Gurney J J Gurney J LPascoe M D amp Richardson S H (eds) Proceedings of the 7th

International Kimberlite Conference Cape Town Red Roof Designpp 262^279

Green D H amp Wallace M E (1988) Mantle metasomatism byephemeral carbonatite melts Nature 336 459^462

Green T H Blundy J D Adam J amp Yaxley G M (2000) SIMSdetermination of trace element partition coefficients betweengarnet clinopyroxene and hydrous basaltic liquids at 2^75 GPaand 1080^12008C Lithos 53 165^187

GriffinW L amp OrsquoReilly SY (2007) Cratonic lithospheric mantle Isanything subducted Episodes 30(1) 43^53

Griffin W L Wass S Y amp Hollis J D (1984) Ultramafic xenolithsfrom Bullenmerri and Gnotuk maars Victoria Australia petrol-ogy of a sub-continental crust^mantle transition Journal of

Petrology 25 53^87GriffinW L Doyle B J Ryan C G Pearson N J OrsquoReilly SY

Davies R Kivi K van Achterbergh E amp Natapov LM (1999)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1871

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Layered mantle lithosphere in the Lac de Gras Area SlaveCraton composition structure and origin Journal of Petrology 40705^727

Griffin W L Pearson N J Belousova E Jackson S E vanAchterbergh E OrsquoReilly SY amp Shee S R (2000)The Hf isotopecomposition of cratonic mantle LAM-MC-ICPMS analysis ofzircon megacrysts in kimberlites Geochimica et Cosmochimica Acta

64 133^147Griffin W L OrsquoReilly S Y Doyle B J Pearson N J

Coopersmith H Kivi K Malkovets V amp Pokhilenko N(2004) Lithosphere mapping beneath the North American plateLithos 77 873^922

Grulaquo tter H S Apter D B amp Kong J (1999) Crust^mantle couplingevidence from mantle-derived xenocrystic garnets In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 307^312

Hart S R Blusztajn J Dick H J B Meyer P S ampMuehlenbachs K (1999) The fingerprint of seawater circulationin a 500-meter section of ocean crust gabbros Geochimica et

Cosmochimica Acta 63 4059^4080Harte B amp Kirkley M B (1997) Partitioning of trace elements

between clinopyroxene and garnet Data from mantle eclogitesChemical Geology 136 1^24

Heaman L M Creaser R A amp Cookenboo H O (2002) Extremeenrichment of high field strength elements in Jericho eclogitexenoliths A cryptic record of Paleoproterozoic subduction partialmelting and metasomatism beneath the Slave craton CanadaGeology 30 507^510

Heaman L M Kjarsgaard B A amp Creaser R A (2004) The tem-poral evolution of North American kimberlites Lithos 76 377^397

Heaman L M Creaser R A Cookenboo H O amp Chacko T(2006) Multi-stage modification of the northern Slave mantlelithosphere Evidence from zircon- and diamond-bearing eclogitexenoliths entrained in Jericho Kimberlite Journal of Petrology 47821^858

Hermann J Cpandler C Hack A amp Korsakov A (2006) Aqueousfluids and hydrous melts in high-pressure and ultra-high pressurerocks implications for element transfer in subduction zones Lithos92 399^417

Hills DV amp Haggerty S E (1989) Petrochemistry of eclogites fromthe Koidu Kimberlite Complex Sierra Leone Contributions to

Mineralogy and Petrology 103 397^422Hoffman P F (1989) Precambrian geology and tectonic history of

North America In Bally A amp Palmer A (eds) The Geology of

North AmericaccedilAn Overview Boulder CO Geological Society ofAmerica pp 447^512

IrelandT R Rudnick R L amp Spetsius Z (1994) Trace-elements indiamond inclusions from eclogites reveal link to Archean granitesEarth and Planetary Science Letters 128 199^213

Irvine G J Pearson D G Kjarsgaard B A Carlson R WKopylova M G amp Dreibus G (2003) Evolution of the litho-spheric mantle beneath Northern Canada a Re^Os isotope andPGE study of kimberlite-derived peridotite xenoliths fromSomerset Island and a comparison to the Slave and Kaapvaalcratons Lithos 71 461^488

Isachsen C E amp Bowring S A (1994) Evolution of the Slave cratonGeology 22 917^920

Jacob D E (2004) Nature and origin of eclogite xenoliths from kim-berlites Lithos 77 295^316

Jacob D E amp Foley S F (1999) Evidence for Archean ocean crustwith low high field strength element signature from diamondifer-ous eclogite xenoliths Lithos 48 317^336

Jacob D Jagoutz E Lowry D Mattey D amp Kudrjavtseva G(1994) Diamondiferous eclogites from Siberiaccedilremnants ofArchean oceanic-crust Geochimica et Cosmochimica Acta 585191^5207

Jacob D E Fung A Jagoutz E amp Pearson D G (2003) Petrologyand geochemistry of eclogite xenoliths from the Ekati kimberlitesarea Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA_239

Jacob D E Bizimis M amp Salters V J M (2005) Lu^Hf and geo-chemical systematics of recycled ancient oceanic crust evidencefrom Roberts Victor eclogites Contributions to Mineralogy and

Petrology 148 707^720Jerde E A Taylor L A Crozaz G Sobolev NV amp SobolevV N

(1993) Diamondiferous eclogites from Yakutia Siberiaccedilevidencefor a diversity of protoliths Contributions to Mineralogy and Petrology

114 189^202Jones A G Ferguson I J Chave A D Evans R L amp

McNeice GW (2001) The electric lithosphere of the Slave cratonGeology 29 423^426

Kelemen P B Shimizu N amp Dunn T (1993) Relative depletionof niobium in some arc magmas and the continental-crust-partitioning of K Nb La and Ce during meltrockreaction in the upper-mantle Earth and Planetary Science Letters 120111^134

Kinzler R J (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition Application to mid-ocean ridge basalt petrogenesis Journal of Geophysical Research 102853^874

Klemme S van der Laan S R Foley S F amp Guenther D (1995)Experimentally determined trace and minor element partitioningbetween clinopyroxene and carbonatite melt under upper mantleconditions Earth and Planetary Science Letters 133 439^448

Konzett J Armstrong R A Sweeney R J amp CompstonW (1998)The timing of MARID metasomatism in the Kaapvaal mantle anion probe study of zircons from MARID xenoliths Earth and

Planetary Science Letters 160 133^145Kopylova M G Russell J K amp Cookenboo H (1999) Mapping the

lithopshere beneath the north central Slave Craton In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 468^479

Krogh E (1988) The garnet^clinopyroxene iron^magnesiumgeothermometerccedila reinterpretation of existing experimentaldata Contributions to Mineralogy and Petrology 99 44^48

Kusky T M (1989) Accretion of Archean Slave Province Geology 1763^67

LeCheminant A N amp Heaman L M (1989) Mackenzie igneousevents Canada Middle Proterozoic hotspot magmatismassociated with ocean opening Earth and Planetary Science Letters

96 38^48LeCheminant A N Richardson D G DiLabio R N W amp

Richardson K A (1996) La recherche de diamants au CanadaGeological Survey of Canada Open File 3228 268pp

MacGregor I D amp Carter J L (1970) The chemistry of clinopyrox-enes and garnets of eclogite and peridotite xenoliths from theRoberts Victor mine South Africa Physics of the Earth and PlanetaryInteriors 3 391^397

MacGregor I D amp Manton W I (1986) Roberts Victor eclogitesancient oceanic crust Journal of Geophysical Research 91 14063^14079

MacKenzie J M amp Canil D (1999) Composition and thermal evo-lution of cratonic mantle beneath the central Archean SlaveProvince NWT Canada Contributions to Mineralogy and Petrology

134 313^324

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1872

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

McCandlessT E amp Gurney J J (1989) Sodium in garnet and potas-sum in clinopyroxene criteria for classifying mantle eclogites InRoss J (ed) Kimberlites and Related Rocks Carlton Blackwellpp 827^832

McDonoughW F amp Sun S-S (1995) The composition of the EarthChemical Geology 120 223^253

Navon O amp Stolper E (1987) Geochemical consequences of meltpercolation the upper mantle as a chromatographic columnJournal of Geology 95 285^307

Nikogosian I K amp Sobolev AV (1997) Ion-microprobe analysis ofmelt inclusions in olivine experience in estimating the olivine^meltpartition coefficients of trace elements Geochemistry International 35119^126

Norman M D Griffin W L Pearson N J Garcia M O ampOrsquoReilly S Y (1998) Quantitative analysis of trace element abun-dances in glasses and minerals a comparison of laser ablationinductively coupled plasma mass spectrometry solution inductivelycoupled plasma mass spectrometry proton microprobe and elec-tron microprobe data Journal of Analytical Atomic Spectrometry 13477^482

Norman M D Garcia M O amp Pietruszka A J (2005) Trace-element distribution coefficients for pyroxenes plagioclase andolivine in evolved tholeiites from the 1955 eruption of KilaueaVolcano Hawairsquoi and petrogenesis of differentiated rift-zonelavas American Mineralogist 90 888^899

OrsquoHara M J amp Herzberg C (2002) Interpretation of trace elementand isotope features of basalts Relevance of field relations petrol-ogy major element data phase equilibria and magma chambermodeling in basalt petrogenesis Geochimica et Cosmochimica Acta 662167^2191

PadghamW A amp FysonW K (1992) The Slave Province a distinctcraton CanadianJournal of Earth Sciences 29 2072^2086

Pearson D G Davies G R amp Nixon P H (1993) Geochemical con-straints on the petrogenesis of diamond facies pyroxenites from theBeni Bousera peridotite massif North Morocco Journal of Petrology34 125^172

Pearson N J Griffin W L Doyle B J OrsquoReilly S Y vanAchterbergh E amp Kivi K (1999) Xenoliths from kimberlitepipes of the Lac de gras area Slave Craton Canada In GurneyJ J Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite ConferenceCapeTown Red Roof Design pp 644^658

Perfit N R Gust D A Bence A E Arculus R J amp Taylor S R(1980) Chemical characteristics of island-arc basalts implicationsfor mantle sources Chemical Geology 30 227^256

Polat A Hofmann A W amp Rosing M T (2002) Boninite-likevolcanic rocks in the 37^38 Ga Isua greenstone belt WestGreenland geochemical evidence for intra-oceanic subductionzone processes in the early Earth Chemical Geology 184 231^254

Poudjom DjomaniY H GriffinW L OrsquoReilly SY amp Doyle B J(2005) Lithospheric domains and controls on kimberlite emplace-ment Slave Province Canada evidence from elastic thickness andupper mantle composition Geochemistry Geophysics Geosystems 6doi1010292005GC0009978

Presnall D C Gudfinnsson G H amp Walter M J (2002)Generation of mid-ocean ridge basalts at pressures from 1 to 7GPa Geochimica et Cosmochimica Acta 66 2073^2090

Schmidberger S S Heaman L M Simonetti A Creaser R A ampCookenboo H O (2005) Formation of Paleoproterozoiceclogitic mantle Slave Province (Canada) Insights from in-situHf and U^Pb isotopic analyses of mantle zircons Earth and

Planetary Science Letters 240 621^633

Schmidberger S S Simonetti A Heaman L M Creaser R A ampWhiteford S (2007) Lu^Hf in-situ Sr and Pb isotope and traceelement systematics for mantle eclogites from the Diavik diamondmine Evidence for Paleoproterozoic subduction beneath the Slavecraton Canada Earth and Planetary Science Letters 254 55^68

Schulze D J amp Helmstaedt H (1988) Coesite-sanidine eclogitesfrom kimberlite products of mantle fractionation or subductionJournal of Geology 96 435^443

Smithies R H Champion D C amp Sun S S (2004) The case forArchaean boninites Contributions to Mineralogy and Petrology 147705^721

Smyth J R Caporuscio F A amp McCormick T C (1989) Mantleeclogites evidence of igneous fractionation in the mantle Earth andPlanetary Science Letters 93 133^141

Sobolev NV amp Lavrentrsquoyev Y G (1971) Isomorphic sodium admix-ture in garnets formed at high pressure Contributions to Mineralogy

and Petrology 31 1^12Stalder R Foley S F Brey G P amp Horn I (1998) Mineral aqueous

fluid partitioning of trace elements at 900^12008C and 30^57GPa New experimental data for garnet clinopyroxene and rutileand implications for mantle metasomatism Geochimica et

Cosmochimica Acta 62 1781^1801Sun S-S amp McDonough W F (1989) Chemical and isotopic sys-

tematics of oceanic basalts implications for mantle compositionsand processes In Saunders A D amp Norris M J (eds)Magmatism in the Ocean Basins Geological Society London Special

Publications 42 313^345Taura H Yurimoto H Kurita K amp Sueno S (1998) Pressure

dependence on partition coefficients for trace elements between oli-vine and the coexisting melts Physics and Chemistry of Minerals 25469^484

Taylor L A amp Neal C R (1989) Eclogites with oceanic crustal andmantle signatures from the Bellsbank kimberlite South AfricaPart I Mineralogy petrography and whole rock chemistryJournal of Geology 97 551^567

van Achterbergh E Ryan C G amp Griffin W L (1999) Glitter online intensity reduction for the laser ablation inductively coupledplasma spectrometry Abstract 9th V M GoldschmidtConference LPI Boston MA 905 pp Abstract no 7215

van Breemen O DavisW J amp King J E (1992) Temporal distribu-tion of granitoid plutonic rocks in the Archean Slave ProvinceNorthwest Canadian Shield Canadian Journal of Earth Sciences 292186^2199

Vervoort J D Patchett P J Blichert-Toft J amp Albare de F (1999)Relationship between Lu^Hf and Sm^Nd isotopic systems in theglobal sedimentary system Earth and Planetary Science Letters 16879^99

Viljoen K S Smith C B amp Sharp Z D (1996) Stable and radio-genic isotope study of eclogite xenoliths from the Orapa kimberliteBotswana Chemical Geology 131 235^255

Walter M J (1998) Melting of garnet peridotite and theorigin of komatiite and depleted lithosphere Journal of Petrology 3929^60

Wasserburg G J Jacobsen S B DePaolo D J McCulloch M Tamp Wen T (1981) Precise determination of SmNd ratios Sm andNd isotopic abundances in standard solutions Geochimica et

Cosmochimica Acta 45 2311^2323Westerlund K J Shirey S B Richardson S H Carlson R W

Gurney J J amp Harris JW (2006) A subduction wedge origin forPaleoarchean peridotitic diamonds and harzburgites from thePanda kimberlite Slave craton evidence from Re^Os isotope sys-tematics Contributions to Mineralogy and Petrology 152 275^294

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1873

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Page 21: Origins of Xenolithic Eclogites and Pyroxenites from the Central

Table 5 Individual and averaged Sr Nd and Hf isotope ratios obtained from garnet and cpx solutions

Sample 87Sr86Sr 2SE2SD 143Nd144Nd 2SE2SD eNd 2SE2SD 176Hf177Hf 2SE2SD eHf 2SE2SD

Low-Mg eclogites

VR43469 cp 0702661 0000010 0514844 0000006 43 01 0286130 0000012 119 04

VR43469 gt 0515508 0000014 56 03 0286506 0000012 132 04

VR43469 gt r 0515453 0000010 55 02 0286548 0000011 134 04

VR43469 gt r 0515384 0000007 54 01 0287244 0000010 158 04

VR43469 gt r 0515487 0000014 56 03 0286996 0000014 149 05

VR43469avg gt 0515458 0000109 55 21 0286823 0000715 143 253

WR VR43469 0515059 0000273 47 5 0286405 0001079 128 38

YK1926 cp 0703546 0000011 0513043 0000007 8 0 0283154 0000009 13 03

YK1926 gt 0513364 0000008 14 01 0295718 0000005 458 02

WR YK1926 0513120 0000026 9 1 0285980 0000716 113 25

YK1943 cp 0512143 0000007 10 01 0282213 0000005 20 02

YK1943 gt 0514461 0000012 36 02 0287510 0000008 168 03

YK1943 gt r 0514491 0000009 36 02 0287332 0000005 161 02

YK1943 gt r 0514517 0000010 37 02 0287290 0000007 160 02

YK1943avg gt 0514489 0000056 36 11 0287377 0000233 163 82

WR YK1943 0512625 0000228 0 5 0283138 0000879 13 31

High-Mg eclogites

VR43480 gt 0520965 0000017 162 03 0284950 0000010 77 04

VR67360 cp 0283509 0000074 26 26

VR67360 cp r 0283433 0000031 23 11

VR67360avg cp 0701662 0000018 0511660 0000010 19 02 0283471 0000108 25 38

VR67360 gt 0512984 0000101 7 20 0287429 0000180 165 64

WR VR67360 0511761 0000132 17 3 0284734 0000469 69 17

YK1949 cp 0703019 0000013 0512231 0000009 8 02 0282513 0000009 9 03

YK1949 gt 0513511 0000010 17 02 0290297 0000042 266 15

YK1949 gt r 0513508 0000012 17 02 0289947 0000015 254 05

YK1949 gt r 0513552 0000014 18 03

YK1949avg gt 0513523 0000049 17 10 0291569 0000494 311 175

WR YK1949 0512446 0000103 4 2 0284622 0000847 65 30

Volatile-rich eclogites

VR67112b gt 0522790 0000008 198 02 0354367 0000036 2532 13

YK1911 cp 0751186 0000009 0512165 0000007 9 01 0281263 0000009 53 03

YK1911 gt 0525721 0000013 255 03 0288051 0000012 187 04

WR YK1911 0513940 0000342 25 7 0281414 0000018 48 1

Pyroxenites

YK1914 cp 0703015 0000034 0513283 0000041 13 08

YK1914 cp r 0702956 0000005 0513200 0000027 11 05

YK1914 cp r 0513527 0000059 17 11

YK1914 cp r 0513045 0000010 8 02

YK1914avg cp 0702986 0000083 0513264 0000402 12 78

YK1915 gt 0514003 0000011 27 02 0283958 0000021 42

Analyses of full procedural replicates are designated with an lsquorrsquo after sample name SE standard error (internal errorsingle analyses) SD standard deviation (external error averaged multiple analyses) Italics used for reconstructed whole-rocks (mineral isotopic compositions and concentrations weighted by modes) Uncertainties for reconstructed whole-rocksinclude 2SE or 2SD uncertainties on each mineralrsquos isotope composition combined with 20 uncertainty in the amountof Nd or Hf contributed by the constituent minerals eNd and eHf are per 10 000 deviation from model chondritic mantle[model of Wasserburg et al (1981) and Blichert-Toft amp Albarede (1997) respectively]

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1863

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

εHf

εNd

Ekati eclogitesminus100

0

100

200

300

400

500

minus40 0 40 80 120 160 200 240 280

gt in YK1915

gt in VR67112bεNd=198 εHf=2532

minus40

0

40

80

120

160

minus40 minus20 0 20 40 60

minus60

minus40

minus20

0

20

40

070 071 072 073 074 075 076

North Americankimberlite field

εNd

87Sr86Sr

cpx in YK1914

Ekatieclogites

DM

DM

DM

DM

(a)

(b)

pyroxenites

low-Mg

high-Mg

volatile-rich

eclogites

Diavik low-TA154 S kimberlite

cpx in YK1911

Fig 12 (a) Sr^Nd for cpx and (b) Nd-Hf isotope diagram for garnet cpx and reconstructed whole-rocks (symbols with bold outline) Finedashed lines in (b) connect data points from the same sample with garnets having more radiogenic isotopic compositions Uncertainties forreconstructed whole-rocks combine 20 modal uncertainty with uncertainties on each mineralrsquos isotope composition Error bars on individualminerals may be smaller than the symbols Also shown are the isotopic compositions of the depleted mantle (DM DePaolo 1981 Griffin et al2000 Bell amp Tilton 2002) Range of 87Sr86Sr in eclogites from Ekati (Lac de Gras area 100 km north of Diavik Jacob et al 2003) in NorthAmerican kimberlites and compositions of low-temperature eclogites from Diavik and of Lac de Gras A154 South kimberlite fromSchmidberger et al (2007)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1864

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

one low-Mg eclogite has eNd similar to the depleted mantle(Fig 12b)Clinopyroxene in the phlogopitized eclogite YK1911

has unradiogenic Nd and Hf and strongly radiogenicSr whereas the coexisting garnet has radiogenic Nd andHf Although this sample contains 50 vol phlogopiteits whole-rock isotopic composition was calculated

from garnet and cpx alone because the contributionfrom phlogopite to the Hf and Nd budget is 52 and51 respectively (unpublished data) The whole-rockhas radiogenic Nd but has lower eHf than anyother sample analysed in accord with the extreme Hfabundance (53 ppm) and low LuHf of the cpx

DISCUSS IONOrigins of eclogites beneath the SlaveCratonHigh-Mg and high-Ca eclogites

With regard to their major- and trace-element contentsthe high-Mg and high-Ca eclogites from Lac de Gras arebroadly similar to variably evolved oceanic gabbros(Fig 13b and c) which may indicate an origin as subductedoceanic crust Compatible trace-element abundancesare useful to deduce potential crustal protoliths as theseelements are less susceptible to post-formation modification(Jacob 2004) The flat HREE patterns in the high-Mg andhigh-Ca eclogites are compatible with a low-pressureorigin as the oceanic crust is a product of moderate degreesof partial melting mainly within the spinel stability field(eg Presnall et al 2002) whereas melting with residualgarnet would qualitatively be expected to produce stronglyfractionated REE patterns Positive Sr and Pb anomaliesalthough compatible with a plagioclase signature couldalso be due to cryptic metasomatism (Dawson 1984) ofthe eclogites during their residence in the lithosphericmantle These anomalies are probably not related to kim-berlite infiltration as the whole-rock eclogite compositionswere reconstructed from analyses of fresh mineral coresLow ZrSm similar to that in primitive gabbros is

observed for many of the high-Mg and high-Caeclogites (Fig 13b and c) and is independent of the ratioof garnetcpx and whether or not the whole-rock wasreconstructed with rutile (Electronic Appendix 10) Thenegative trend of Mg-number against Yb in garnet(Fig 14) argues against igneous processes involving garnetaccumulation (Taylor amp Neal 1989) as suggested forexample for some high-Mg eclogites from West Africa(Barth et al 2002b) Rather the anti-correlation suggeststhat olivine fractionation played a role in the formation ofthe protoliths Fractionating olivine would excludeYb (Dolivinebasalt

frac14 002 Nikogosian amp Sobolev 1997)and incorporate Mg relative to Fe (Dfrac1415 and 05 respec-tively Taura et al1998) leading to progressive depletion ofMg and enrichment of Yb in the residual melt and henceanti-correlated Mg and Yb in the whole-rock This isinherited after eclogitization All eclogite types (exceptVR43452) plot along the olivine fractionation trend sug-gesting that they formed by a similar processHigh-Mg eclogites have lower

PREE than high-Ca

eclogites and some appear to have positive Eu anomalies

high-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

gabbrosSE Indian ridge

gabbrosOman ophiolite

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

high-Mg

RockNMORB(a)

VR43452

YK1949

low-Mg

boninite-like melts

01

001

1

10

(b)

01

001

1

10

(c)

01

001

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

Fig 13 Trace-element patterns of (a) low-Mg (b) high-Mg and(c) high-Ca eclogites compared with boninite-like volcanic rocksfrom the Isua greenstone belt (Polat et al 2002) gabbros from theOman ophiolite (Benoit et al 1996) and the SE Indian Ridge (Hartet al 1999) normalized to N-MORB of Sun amp McDonough (1989)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1865

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(EuEu up to 14) similar to more primitive cumulates(troctolite gabbros and olivine gabbros) reported in the lit-erature (Fig 13b) Although the sizes of these Eu anomaliesare not outside the analytical uncertainty their correlationwith CeYb in reconstructed eclogites supports their signif-icance (r2frac14066 nfrac14 4) The more primitive nature ofhigh-Mg eclogites is consistent with higher Mg-numbersrelative to high-Ca eclogites and may be due to olivineaccumulation during formation of their protolithsThe two samples with the lowest HREE abundanceshave the highest EuEu which is compatible with varia-tions in plagioclase modesMore evolved protoliths with higher pyroxeneplagio-

clase plus trapped melt could account for small or absentpositive Eu anomalies and higher

PHREE of high-Ca

eclogites (Fig 13c) Although the reason for the frequentassociation of graphite or diamond with high-Ca eclogitesis unclear (Aulbach et al 2003) the frequent occurrence ofkyanite in high-Ca eclogites also observed at Ekati justnorth of Lac de Gras has been argued to support a crustalprotolith (Jacob 2004) Kyanite andor corundum havebeen suggested to exsolve from high-temperature high-Alclinopyroxenes (Caporuscio amp Smyth 1990) Eclogitesfrom Lac de Gras show no exsolution textures suggestiveof such an origin although recrystallization can obscureor destroy microstructural evidence for exsolution(eg Griffin et al 1984)Modelling shows that qualitatively plagioclase accumu-

lation from a low-pressure melt derived from MORBmantle can account for positive Pb Sr and Eu and negativeZr anomalies relative to N-MORB whereas cpx accumula-tion or trapped melt increases Zr Y MREE and HREE

abundances (Fig 15) Some melt addition (for exampleby entrapment) seems to be required to increase all trace-element abundances to those observed for the eclogitesAlthough high-Ca and high-Mg eclogites could representshallow and deep portions of the same oceanic crust thehigher average temperatures obtained for the high-Caeclogites the frequent occurrence of carbon in high-Caeclogites and its absence in high-Mg eclogites the strongLREEHREE fractionation observed for six of seven high-Ca eclogites and the narrow range of compositions of thehigh-Mg eclogites suggest that they may have protolithsthat are not cogenetic Some discrepancies in particularaffecting the incompatible elements may be due to post-crystallization processes acting upon the protoliths suchas dehydration melting and later metasomatism that isevident in peridotites from the same locality (Griffin et al1999)Low K2OTh for high-Ca and high-Mg eclogites

(average 08 Table 4) relative to assumed crustal

0

2

4

6

8

10

30 40 50 60 70 80

Mg

VR43452

Yb

GARNET

pyroxenites

low-Mghigh-Mghigh-Ca

eclogites

VR19673ecl-g2

Fig 14 Garnet Mg-number vs Yb (ppm including 2s uncertainty)Lines qualitatively show separate correlations for eclogites (exceptoutlierVR43452) and pyroxenites (except outlierVR19673ecl-g2)

high-Ca

high-Mg

001

01

1

10

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

melt08 melt02 cryst cpx01 melt09 cryst

plag

(a)

001

01

1

10(b)

RockNMORB

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 15 Reconstructed (a) high-Ca and (b) high-Mg eclogitescompared with compositions of plagioclase and cpx crystallizingfrom a MORB melt modelled using distribution coefficients ofNorman et al (2005)The MORB was calculated as a 10 equilibriummelt of MORB pyrolite [composition from Kelemen et al (1993)except Pb Pr Sr Y and Lu which were calculated as residues from1 fractional melting of pyrolite from McDonough amp Sun (1995)using melt modes of Kinzler (1997) and distribution coefficients ofNorman et al (2005)] Also shown are mixtures of crystals (consistingof 50 plagioclase plus 50 cpx) and melt in the proportions 0901and 0208

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1866

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

precursors (eg gabbros SE Indian Ridge average 22Hart et al 1999) may reflect the preferred loss of fluid-mobile elements during dehydration (Becker et al 2000)although large uncertainties on these low-abundance ele-ments in the constituent minerals and reconstructedwhole-rocks makes this interpretation less than robustThe REE patterns of most high-Mg eclogites are too flatto represent melting residues as partial melting shouldproduce low LREEHREE (Electronic Appendix 10) Incontrast the marked LREE depletion of five of sevenhigh-Ca eclogites (LaLufrac14 02^08) contrasts with that ofassumed evolved gabbroic protoliths (LaLufrac1415^32Hart et al 1999) and can be modelled as a consequence ofpartial melting during subduction after eclogitization (iein the presence of garnet which retains HREE relative toLREE eg Barth et al 2002a) has taken place (Aulbachet al 2003)Our results are compatible with previous studies on

eclogite xenoliths from the Slave Craton includingthose from the same kimberlite that have linked thesesamples or some of these samples to Paleoproterozoicsubduction events (the 195^191Ga formation of theHottah terrane and the 18Ga formation of the GreatBear magmatic arc) in the Slave craton (Griffin et al1999 Kopylova et al 1999 Heaman et al 2002 2006Schmidberger et al 2005 2007)The compositional features discussed above although

compatible with crustal protoliths are not necessarilydefinitive of a subduction origin However if direct crystal-lization from basaltic melts at high pressure (ie mantledepths) is invoked instead specific conditions would haveto be met to satisfy the observations The Mg-number ineclogites from Lac de Gras are lower than those expectedof primary high-pressure mantle melts (eg Walter 1998OrsquoHara amp Herzberg 2002) requiring crystallization athigh pressures after olivine fractionation This originwould further require shielding from peridotitic wallrockwith which the fractionated melts would no longer be inequilibrium (Barth et al 2001) In such a setting the lowrelative abundances of fluid-mobile elements comparedwith those of less fluid-mobile elements and of LREErelative to HREE (Table 4) would reflect an expulsion offluids during crystallization of the protoliths and latermelt mobilization respectively The eclogites clearly donot represent garnet plus cpx cumulates which wouldrequire a positive correlation of Mg-number with Yb thatis at variance with the observed negative correlation ineclogitic garnets from this study (Fig 14)

Low-Mg eclogites

Garnets in low-Mg eclogites compositionally overlap withthose in massive eclogites from Jericho (NW of the Lacde Gras kimberlites) that were suggested to have a high-pressure intrusive origin based on the similarity of theirmajor-element compositions to those of Group A eclogites

fromsouthernAfrica forwhichamantleoriginhasbeenpro-posed (Kopylova et al1999) Garnets in low-Mg eclogites inthe present study have major-element compositions moresimilar to garnets in groupB rather thangroupAeclogitesThe positive slope in N-MORB-normalized HREE

abundances displayed by the low-Mg eclogites is clearlydistinct from other eclogite types and is similar to those ofboninite-like second-stage melts that are closely associatedwith ultramafic volcanics (Smithies et al 2004) (Fig 13a)The range of Mg-numbers in the low-Mg eclogites (55^66)overlaps with that of Archaean second-stage melts fromIsua (58^75) and Abitibi (52^82 Smithies et al 2004)These similarities might indicate that the low-Mg eclogiteshave an origin in a previously depleted mantle source suchas supra-subduction zone mantle Subchondritic ZrHf ofthe low-Mg eclogites contrasts with the supra-chondriticZrHf of the high-Mg and high-Ca eclogites (Aulbach et

al in preparation) and is consistent with previous cpx-controlled depletion in an arc mantle source This sourcecould have been subsequently remelted to generate theboninite-like precursor to the low-Mg eclogites as sug-gested for the protoliths of eclogites from Udachnaya(Jacob amp Foley 1999)Although some second-stage melts to which the low-Mg

eclogites are similar are closely associated with komatiitesand could contain a plume component the radiogenic ordepleted mantle-like Nd and Hf and unradiogenic Sr inthe low-Mg eclogites do not support plume involvement

Trace-element and Sr^Nd^Hf isotopeconstraints on post-formationmetasomatismSeveral low-Mg and high-Ca eclogites show negativeslopes in the LREE which may be indicative of interactionwith small amounts of LREE-enriched melt (Navon ampStolper 1987) amounting to cryptic metasomatism(ie not accompanied by crystallization of new phases) inthe nomenclature of Dawson (1984) High Nb abundancesrelative to elements of similar compatibility displayed bysome low-Mg and high-Ca eclogites (Fig 10) could resultfrom overestimation of rutile modes However high-CaeclogiteVR40345 is rutile-free and has been reconstructedwithout rutile its high Nb abundance is accompanied byLREE enrichment indicating that metasomatic additionmay be responsibleLow-Mg eclogites show a positive correlation of La with

Nb (Electronic Appendix 11a) indicating concomitantaddition of LREE and Nb This suggests that the metaso-matic agent was a solute-rich fluid or silicate silicic melt ashydrous fluids and carbonatites do not carry enough HFSEto affect the budget of eclogites (Green amp Wallace 1988Klemme et al 1995 Adam et al 1997 Stalder et al 1998)Of the samples that are enriched in Nb all but one (high-Ca eclogiteVR50909) show a simultaneous enrichment in

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1867

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Hf (Electronic Appendix 11b) Together these trace-element relationships suggest that both SmNd and LuHfwere lowered in these samples Their low parent^daughterratios contrast with higher ratios in unenriched samplesand resulted in the ingrowth of a range of Nd and Hf iso-tope compositions Compared with the low-Mg and high-Ca eclogites the high-Mg eclogites have a narrow range ofLa and Nb concentrations which correlates with a narrowrange in Nd and Hf isotope compositions (Fig 12)Precise model ages cannot be calculated because parent^

daughter ratios are available only from in situ analyseswhich have relatively large uncertainties Howeverevidence for long-term enrichment and depletion may bediscerned Low-Mg eclogites even those that are enrichedin LREE have relatively high SmNd as a result of havingoverall positive slopes in their REE patterns relative toMORB whereas the high-Mg eclogites have relativelyflat REE patterns and therefore lower SmNd (Fig 11)This translates into lower eNd for high-Mg eclogites thanfor low-Mg eclogites In contrast they show no systematicdifference with respect to 176Hf177Hf Because the low- andhigh-Mg eclogites are proposed to have different origins asdiscussed above and therefore did not necessarily have thesame initial isotopic composition and formation age theseisotopic differences cannot be used to estimate their agesA long-term coupling of trace-element and isotopic

abundances is also indicated by the fact that low-Mg eclo-gite YK1943 which has experienced Nb and LREE addi-tion has lower eNd (0) and eHf (thorn13) than low-Mg eclogiteVR43469 with higher LuHf and SmNd (eNdfrac14thorn47eHffrac14thorn128) If low-Mg eclogites YK1943 and VR43469originally had identical initial Nd and Hf isotope composi-tions and if the parent^daughter ratios are accurate andwere modified only once during their evolution shortlyafter metamorphism (when their isotopic compositionwould have been reset) this overprint can be calculatedfor both isotopic systems to have occurred at 17GaThis would suggest that Hf and Nd were added duringthe same event In contrast high-Mg eclogites VR67360and YK1949 would have had identical Ndisotope compositions at 10Ga whereas their Hf isotopecompositions are disturbed in that the sample with thehigher parent^daughter ratio has a lower 176Hf177Hf thanthe sample with the lower parent^daughter ratio Theseresults suggest that the LREE and HFSE enrichmentobserved in different samples occurred during separateevents that did not affect the eclogites equallyExtreme HFSE enrichment leading to the growth of

zircon and rutile and LREE addition leading to thegrowth of apatite in eclogites from the Jericho kimberlitein the northern Slave craton have been ascribed to twoseparate metasomatic events at around 18Ga (quasi-coincident with the age of formation and metamorphism)and 13^10Ga respectively (Heaman et al 2002 2006

Schmidberger et al 2005) The authors associated theseevents with Paleoproterozoic subduction and aMesoproterozoic thermal disturbance in the Slave litho-spheric mantle possibly linked to the Mackenzie igneousevent Mesoproterozoic Nd model ages mostly between13 and 14Ga were reported for eclogites from thenearby Ekati kimberlites (Jacob et al 2003) Although wedo not observe zircon or apatite in volatile-free eclogitesfrom Lac de Gras and the age constraints obtained fromour data are weak they support a link between theseevents described for other eclogite populations and themetasomatic enrichment observed in our studyThe Sr isotope data in the present study are similar to

those of Schmidberger et al (2007) who suggested that thedepleted-mantle-like values are consistent with derivationof the protoliths from an oceanic mantle source However87Sr86Sr in volatile-free eclogites in the present study isremarkably constant considering their Nd and Hf isotopevariability (Fig 12) and the variability of trace-elementpatterns (Fig 10) Given the evidence for dehydration andmelting in many of the eclogites fractionation of highlyincompatible Rb from Sr and differential ingrowth ofradiogenic Sr would be expectedThis may indicate a rela-tively recent overprint and considering the similarityof 87Sr86Sr to present-day depleted mantle which is signif-icantly lower than 87Sr86Sr of North American kimber-lites a juvenile (asthenospheric) source

PyroxenitesPyroxenites from Lac de Gras often show relationshipsin their major-element contents that are different fromthose observed in the eclogites For example pyroxeniticgarnets show a negative correlation between CaO andNa2O contents whereas these oxides are positively corre-lated for garnets in eclogites (Fig 5) CaO is well correlatedwith C2O3 in pyroxenitic garnets but not in eclogitic gar-nets (Fig 4) Mg-number is positively correlated withYb abundance for pyroxenitic garnets but anti-correlatedfor eclogitic garnets (Fig 14) This suggests that the pyrox-enites in the present study are not related to the eclogitesby fractionation or accumulation processes as proposedfor the origin of some pyroxenites (Foley et al 2003)Pyroxenites have also been interpreted as the products

of crystal segregation in magma channels with alteredoceanic crust as the magma source this would explaintheir Eu anomalies and stable and radiogenic isotopecompositions (Pearson et al 1993 Becker 1996) As thehigh-pressure liquidus phases of most terrestrial basaltsare cpx and garnet (OrsquoHara amp Herzberg 2002) partialmelts of eclogites with basaltic composition wouldcrystallize cpx and garnet but not opx More than half ofthe pyroxenites from Lac de Gras contain opx suggestingthat they probably did not crystallize directly fromeclogite-derived melts

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1868

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Low-Cr websterites and pyroxenites could be crystalli-zation products of mafic melts that interacted with perido-titic mantle thus becoming more refractory The positivecorrelation of Yb and Mg-number in garnet would allowan origin by fractional crystallization from such meltsAlternatively they formed by reaction of silicic (possiblyeclogite-derived) melts with mantle peridotites(eg Kelemen et al 1993 Aulbach et al 2002) This agreeswith the similarity of hypothetical melts in equilibriumwith pyroxenitic garnets in most samples to eclogite-derived melts (Electronic Appendix 12) Niobium concen-trations in all hypothetical melts are much higher than ineclogite-derived melts regardless of whether or not rutileis assumed to be in the residue The same is true for LaCe and Zr concentrations in some samples which mayindicate that NbZr and LREE were added to thepyroxenites some time after their formation as discussedin the previous section for some eclogites Because mini-mum ages for pyroxenitic rutile with an unradiogenicHf isotope composition (calculated by projecting the176Hf177Hf to the model depleted mantle isotopic evolutioncurve assuming zero ingrowth) range to 178066Ga(Aulbach et al in preparation) it appears likely that thepyroxenites were not formed recently (eg during kimber-lite-related or precursory melt migration) and that some ofthe samples are at least MesoproterozoicGarnet pyroxeniteYK1952 consists of schlieren of equili-

brated garnet pyroxenite in a lsquomatrixrsquo of spongy cpxfine cpx^opx intergrowths embayed garnet and meltpatches (Fig 2f) Garnet in the pyroxenite schlieren hasvery low LREE and HFSE abundances Orthopyroxenein the matrix has distinctly higher Mg-number andCr-number and contains more CaO than opx from thepyroxenite schlieren whereas opx rims in the schlierenare compositionally similar to opx in the matrix The con-siderable compositional and textural disequilibrium inthis sample suggests that this is a young feature causedby reaction with fluids shortly before or at the time ofentrainmentAs a consequence of the small sample sizes of most of

the pyroxenites studied here a comprehensive isotopicdataset is not available Clearly more such data areneeded in addition to stable isotope data to obtain betterconstraints on the origins of this rock type Consideringthe microstructural and compositional variability ofthe pyroxenites and major-element relationships oppositeto those observed for eclogites it appears certain that pyr-oxenites form by a variety of processes some of which areunrelated to the formation of eclogites and their protoliths

Volatile-rich eclogitesVolatile-rich eclogites from Lac de Gras display the mostextreme Hf^Nd isotope systematics of all the analysedsamples Garnets in phlogopitized (YK1911) and in cal-cite-bearing eclogite (VR67112b) have the most radiogenic

Nd (eNdfrac14 2486 and 1927 respectively) Garnet inVR67112b has extremely high eHf (2516) Clinopyroxene inYK1911 has the lowest eHf measured for any mineralseparate from Lac de Gras (53) in accord with the veryhigh Hf abundance in its cpx (5 ppm) In contrastgarnet in this sample has radiogenic Hf (eHffrac14 136) Thisdiscrepancy is at least in part due to their relatively lowequilibration temperatures (7698C and 8558C respec-tively) which allowed the isotopic compositions of cpxand garnet to evolve separately at their low andhigh parent^daughter element ratios respectively forextended periods of time and possibly since formationAccordingly the cpx in YK1911 gives a minimum Hf ageof 27GaSlab-derived fluids could have been introduced during

the 27Ga collision of the western Slave Craton domainincluding the thick plume-derived deep lithosphericmantle layer and the eastern domain with its highlydepleted lithosphere This would also be consistent withthe observation that volatile-rich eclogites occur at shallowdepths in the mantle (TKroghfrac14 685^8558C correspondingto depths of 85^115 km along a 40 mWm2 geotherm)where the bulk of dehydration and production of solute-rich melts in a subducting slab might be expected to occur(Becker et al 2000 Hermann et al 2006) Their negativeEu anomalies might indicate that the fluid was derivedfrom the shallow portion of the slab that experiencedplagioclase fractionation The presence of carbonate inVR67112b requires a carbonated sourceHigh 87Sr86Sr in cpx in YK1911 is probably the conse-

quence of isotopic equilibration with high-RbSr phlogo-pite after long-term radiogenic ingrowth while Hfaddition led to retarded ingrowth of radiogenic HfThese systematics combined with radiogenic Nd appear

inconsistent with the involvement of sediments in the pet-rogenesis of this sample (eg Perfit et al 1980 Vervoortet al 1999) or a hydrous fluid such as those responsible forMARID metasomatism in the Kaapvaal craton (Konzettet al 1998)Whatever the origin of the fluids the apparentHf mobility suggests that they were not in equilibriumwith a rutile-bearing source and also that they weresolute-rich in nature (ie hydrous melts able to dissolveHFSE) and the relatively high HREE abundances of thevolatile-rich eclogites indicate a garnet-free source suchas amphibolite This allowed evolution of the sample atrelatively high time-integrated SmNd

SUMMARY AND CONCLUSIONS

(1) The Lac de Gras kimberlites carried abundant eclo-gite and pyroxenite xenoliths to the surface Threemain eclogite types are distinguished in terms ofmajor-element composition and trace element

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1869

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

patterns high-Mg high-Ca and low-Mg eclogitesThe garnets of all eclogite types have anti-correlatedMg-number and Yb pointing to olivine fractionationas a controlling factor in the formation of their pre-cursor rocks Their flat or positively sloped HREEpatterns qualitatively indicate that the melt parentalto the precursor rocks was not in equilibrium with agarnet-bearing residue

(2) Reconstructed high-Mg eclogites and high-Caeclogites have flat HREE patterns positive Pb andSr anomalies and low ZrSm similar to oceanic gab-bros reported in the literature High-Mg eclogiteshave variably low HREE contents and some havesmall positive Eu anomalies possibly suggestive ofmore primitive relatively plagioclase-rich protolithswhereas high-Ca eclogites have higher HREE abun-dances and small or absent Eu anomalies that requiremore cpx- andor melt-rich protoliths if a subductionorigin is assumed

(3) Low-Mg eclogites have distinctive trace-element pat-terns that show a marked positive slope in the HREEand subchondritic ZrHf They resemble someArchaean second-stage melts (lsquoboninitesrsquo) and theirprotoliths appear to have formed from a previouslydepleted source

(4) After formation the eclogites were subject to a varietyof processes Most have low ratios of fluid-mobile vsless fluid-mobile elements possibly as a result of dehy-dration of the protoliths In addition some of the low-Mg and high-Ca eclogites have low LREEHREEconsistent with partial melt lossWhereas the low-Mgeclogites show concomitant enrichment in La Nb andZr some high-Mg and high-Ca eclogites show evi-dence for variable and unsystematic (decoupled)LREE and Nb addition

(5) Isotopically the trace-element characteristics of thedifferent eclogite types translate into lower eNd forhigh-Mg eclogites (negative slope in MREEfrac14 lowerSmNd) than for low-Mg eclogites (positive slope inMREEfrac14higher SmNd) Within the low-Mg groupsamples that show evidence for metasomatism (lowLuHf and SmNd) have lower eNd and eHf than asample that was apparently not metasomatized BothNd and Hf addition can be dated to 17Ga if identi-cal initial isotope compsitions are assumed for thelow-Mg eclogites In contrast the relatively constantand depleted mantle-like 87Sr86Sr of eclogitic cpxdespite variable Nd and Hf isotope compositions andparent^daughter ratios suggests recent input of juve-nile material

(6) Garnet and cpx in pyroxenites show major-elementrelationships different from those in eclogitesTherefore these two rock types are probably notrelated by fractionation processes The presence of

opx precludes direct crystallization from slab-derivedmelts Pyroxenites may be the product of (slab-derived) melts hybridized during assimilation of peri-dotitic mantle

(7) Garnets in phlogopite- and carbonate-bearing eclo-gites have extreme isotopic compositions includingthe most radiogenic Nd and Hf in the datasetNegative Eu anomalies may point to the involvementof a crustal protolith Highly radiogenic Sr radiogenicNd and unradiogenic Hf in the reconstructedphlogopite-bearing sample combined with highHREE abundances and evidence for Hf mobilitymay indicate involvement of solute-rich fluids origi-nating in a garnet- and rutile-free source

ACKNOWLEDGEMENTSHelp in the clean laboratories and with the analyticalfacilities by S Elhlou C Lawson A Sharma P Wielandand M Zhang is greatly appreciated Bill Griffin isthanked for his support and for agreeing to reviewseveral versions of the manuscript despite disagreeingwith many of the conclusions presented here The paperbenefited greatly from the detailed and constructivereviews provided by the editors Gareth Davies andMarjorie Wilson and by Matthias Barth Larry Heamanand an anonymous reviewer This work was funded by aMacquarie University International PostgraduateAward and Postgraduate Research Fund (SA) by anARC SPIRT grant sponsored by Kennecott Canada Incand by ARC Large and Discovery Grants to SYOrsquoRAnalytical data were obtained using instrumentationfunded by ARC LIEF and DEST Systemic InfrastructureGrants industry partners and Macquarie UniversityThis is Publication 481 from the ARC National KeyCentre for Geochemical Evolution and Metallogeny ofContinents (wwwesmqeduauGEMOC)

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online

REFERENCESAdam J GreenT H Sie S H amp Ryan C G (1997) Trace element

partitioning between aqueous fluids silicate melts and mineralsEuropeanJournal of Mineralogy 9 569^584

Aulbach S Stachel T Viljoen K S Brey G P amp Harris J W(2002) Eclogitic and websteritic diamond sources beneath theLimpopo Beltccedilis slab-melting the link Contributions to Mineralogy

and Petrology 143 56^70Aulbach S GriffinW L Pearson N J OrsquoReilly S Y Kivi K amp

Doyle B (2003) Origins of eclogites beneath the central SlaveCraton Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA-0011

Aulbach S Griffin W L Pearson N J OrsquoReilly S Y amp Kivi K(2004a) Mantle formation and evolution Slave Craton constraints

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1870

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

from HSE abundances and Re^Os isotope systematics of sulfideinclusions in mantle xenocrysts Chemical Geology 208 61^88

Aulbach S Griffin W L OrsquoReilly S Y amp McCandless T E(2004b) Genesis and evolution of the lithospheric mantle beneaththe Buffalo HeadTerrane Alberta (Canada) Lithos 77 413^451

Aulbach S GriffinW L Pearson N J OrsquoReilly S Y amp Doyle B(2007) Lithosphere formation in the central Slave Craton(Canada) plume subcretion or lithosphere accretion Contributionsto Mineralogy and Petrology (in press) doi 101007s00410-007-0200-1

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2001)Geochemistry of xenolithic eclogites from West Africa Part IA link between low MgO eclogites and Archean crust formationGeochimica et Cosmochimica Acta 65 1499^1527

Barth M G Foley S F amp Horn I (2002a) Partial melting inArchean subduction zones constraints from experimentally deter-mined trace element partition coefficients between eclogitic miner-als and tonalitic melts under upper mantle conditions PrecambrianResearch 113 323^340

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2002b)Geochemistry of xenolithic eclogites from West Africa part 2Origins of the high MgO eclogites Geochimica et Cosmochimica Acta

66 4325^4345Beard B L Fraracci K NTaylor L A Snyder G A Clayton R

A Mayeda T K amp Sobolev NV (1996) Petrography andgeochemistry of eclogites from the Mir kimberliteYakutia RussiaContributions to Mineralogy and Petrology 125 293^310

Becker H (1996) Geochemistry of garnet peridotite massifs fromlower Austria and the composition of deep lithosphere beneatha Palaeozoic convergent plate margin Chemical Geology 134 49^65

Becker H Jochum K P amp Carlson R W (2000) Trace elementfractionation during dehydration of eclogites from high-pressureterranes and the implications for element fluxes in subductionzones Chemical Geology 163 65^99

Bell K amp Tilton G R (2002) Probing the mantle The story fromcarbonatites EOS Transactions American Geophysical Union 83275^277

Benoit M Polvecurren M amp Ceuleneer G (1996) Trace element andisotopic characterisation of mafic cumulates in a fossil mantlediapir (Oman ophiolite) Chemical Geology 134 199^214

BleekerW (2003) The late Archean record a puzzle in ca 35 piecesLithos 71 99^134

BleekerW Ketchum JW F JacksonV A amp Villeneuve M (1999)The central Slave Basement Complex Part I Its structural topol-ogy and autochthonous core Canadian Journal of Earth Sciences 361083^1109

Blichert-Toft J amp Albare de F (1997) The Lu^Hf isotope geochemis-try of chondrites and the evolution of the mantle^crust systemEarth and Planetary Science Letters 148 243^258

Blichert-Toft J Chauvel C amp Albare de F (1997) Separation of Hfand Lu for high precision isotope analysis of rock samples by mag-netic sector-multiple collector ICP-MS Contributions to Mineralogy

and Petrology 127 248^260Brey G P amp Kolaquo hler T (1990) Geothermobarometry in four-phase

lherzolites II New thermobarometers and practical assessment ofexisting thermobarometers Journal of Petrology 31 1353^1378

Brey G P Kolaquo hlerT amp Nickel K G (1990) Geothermobarometry infour-phase lherzolites I Experimental results from 10 to 60 kbJournal of Petrology 31 1313^1352

Caporuscio F amp Smyth J (1990) Trace element crystal chemistryof mantle eclogites Contributions to Mineralogy and Petrology 105550^561

Carbno G B amp Canil D (2002) Mantle structure beneath the SWSlave Craton Canada constraints from garnet geochemistry in theDrybones Bay Kimberlite Journal of Petrology 43 129^142

Coleman R G Lee E D Beatty L B amp Brannock W W (1965)Eclogites and eclogites their differences and similaritiesGeological Society of America Bulletin 76 483^508

Creaser R A Grulaquo tter H Carlson J amp Crawford B (2004)Macrocrystal phlogopite Rb^Sr dates for the Ekati propertykimberlites Slave Province Canada evidence for multiple intru-sive episodes in the Paleocene and Eocene Lithos 76 399^414

Davies R M GriffinW L Pearson N J Andrew A S Doyle BJ amp OrsquoReilly S Y (1999) Diamonds from the deep Pipe DO-27Slave Craton Canada In Gurney J J Gurney J L Pascoe MD amp Richardson S H (eds) Proceedings of the 7th International

Kimberlite Conference CapeTown Red Roof Design pp 148^155Davies R M Griffin W L OrsquoReilly S Y amp Doyle B J (2004)

Mineral inclusions and geochemical characteristics of microdia-monds from the DO27 A154 A21 A418 DO18 DD17 and RanchLake kimberlites at Lac de Gras Slave Craton Canada Lithos 7739^55

DavisW J amp Hegner E (1992) Neodymium isotopic evidence for thetectonic assembly of Late Archean crust in the Slave Provincenorthwest Canada Contributions to Mineralogy and Petrology 111493^504

Davis W J Jones A G Bleeker W amp Grutter H (2003)Lithosphere development in the Slave craton a linked crustal andmantle perspective Lithos 71 575^589

Dawson J B (1984) Contrasting types of upper-mantle metasoma-tism In Kornprobst J (ed) Kimberlites II The Mantle and

Crust^Mantle Relationships Amsterdam Elsevier pp 289^294DePaolo D J (1981) Neodymium isotopes in the Colorado Front

Range and crust^mantle evolution in the Proterozoic Nature 291193^196

Erlank A J amp Kushiro I (1970) Potassium contents of syntheticpyroxenes at high temperatures and pressures Carnegie Institution ofWashingtonYearbook 68 433^439

Foley S F Barth M G amp Jenner G A (2000) Rutilemelt partitioncoefficients for trace elements and an assessment of the influence ofrutile on the trace element characteristics of subduction zonemagmas Geochimica et Cosmochimica Acta 64 933^938

Foley S F Buhre S amp Jacob D E (2003) Evolution of the Archaeancrust by delamination and shallow subduction Nature 421 249^252

Graham I Burgess J L Bryan D Ravenscroft P J Thomas EDoyle B J Hopkins R amp Armstrong K A (1999) Explorationhistory and geology of the Diavik Kimberlites Lac de GrasNorthwest Territories Canada In Gurney J J Gurney J LPascoe M D amp Richardson S H (eds) Proceedings of the 7th

International Kimberlite Conference Cape Town Red Roof Designpp 262^279

Green D H amp Wallace M E (1988) Mantle metasomatism byephemeral carbonatite melts Nature 336 459^462

Green T H Blundy J D Adam J amp Yaxley G M (2000) SIMSdetermination of trace element partition coefficients betweengarnet clinopyroxene and hydrous basaltic liquids at 2^75 GPaand 1080^12008C Lithos 53 165^187

GriffinW L amp OrsquoReilly SY (2007) Cratonic lithospheric mantle Isanything subducted Episodes 30(1) 43^53

Griffin W L Wass S Y amp Hollis J D (1984) Ultramafic xenolithsfrom Bullenmerri and Gnotuk maars Victoria Australia petrol-ogy of a sub-continental crust^mantle transition Journal of

Petrology 25 53^87GriffinW L Doyle B J Ryan C G Pearson N J OrsquoReilly SY

Davies R Kivi K van Achterbergh E amp Natapov LM (1999)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1871

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Layered mantle lithosphere in the Lac de Gras Area SlaveCraton composition structure and origin Journal of Petrology 40705^727

Griffin W L Pearson N J Belousova E Jackson S E vanAchterbergh E OrsquoReilly SY amp Shee S R (2000)The Hf isotopecomposition of cratonic mantle LAM-MC-ICPMS analysis ofzircon megacrysts in kimberlites Geochimica et Cosmochimica Acta

64 133^147Griffin W L OrsquoReilly S Y Doyle B J Pearson N J

Coopersmith H Kivi K Malkovets V amp Pokhilenko N(2004) Lithosphere mapping beneath the North American plateLithos 77 873^922

Grulaquo tter H S Apter D B amp Kong J (1999) Crust^mantle couplingevidence from mantle-derived xenocrystic garnets In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 307^312

Hart S R Blusztajn J Dick H J B Meyer P S ampMuehlenbachs K (1999) The fingerprint of seawater circulationin a 500-meter section of ocean crust gabbros Geochimica et

Cosmochimica Acta 63 4059^4080Harte B amp Kirkley M B (1997) Partitioning of trace elements

between clinopyroxene and garnet Data from mantle eclogitesChemical Geology 136 1^24

Heaman L M Creaser R A amp Cookenboo H O (2002) Extremeenrichment of high field strength elements in Jericho eclogitexenoliths A cryptic record of Paleoproterozoic subduction partialmelting and metasomatism beneath the Slave craton CanadaGeology 30 507^510

Heaman L M Kjarsgaard B A amp Creaser R A (2004) The tem-poral evolution of North American kimberlites Lithos 76 377^397

Heaman L M Creaser R A Cookenboo H O amp Chacko T(2006) Multi-stage modification of the northern Slave mantlelithosphere Evidence from zircon- and diamond-bearing eclogitexenoliths entrained in Jericho Kimberlite Journal of Petrology 47821^858

Hermann J Cpandler C Hack A amp Korsakov A (2006) Aqueousfluids and hydrous melts in high-pressure and ultra-high pressurerocks implications for element transfer in subduction zones Lithos92 399^417

Hills DV amp Haggerty S E (1989) Petrochemistry of eclogites fromthe Koidu Kimberlite Complex Sierra Leone Contributions to

Mineralogy and Petrology 103 397^422Hoffman P F (1989) Precambrian geology and tectonic history of

North America In Bally A amp Palmer A (eds) The Geology of

North AmericaccedilAn Overview Boulder CO Geological Society ofAmerica pp 447^512

IrelandT R Rudnick R L amp Spetsius Z (1994) Trace-elements indiamond inclusions from eclogites reveal link to Archean granitesEarth and Planetary Science Letters 128 199^213

Irvine G J Pearson D G Kjarsgaard B A Carlson R WKopylova M G amp Dreibus G (2003) Evolution of the litho-spheric mantle beneath Northern Canada a Re^Os isotope andPGE study of kimberlite-derived peridotite xenoliths fromSomerset Island and a comparison to the Slave and Kaapvaalcratons Lithos 71 461^488

Isachsen C E amp Bowring S A (1994) Evolution of the Slave cratonGeology 22 917^920

Jacob D E (2004) Nature and origin of eclogite xenoliths from kim-berlites Lithos 77 295^316

Jacob D E amp Foley S F (1999) Evidence for Archean ocean crustwith low high field strength element signature from diamondifer-ous eclogite xenoliths Lithos 48 317^336

Jacob D Jagoutz E Lowry D Mattey D amp Kudrjavtseva G(1994) Diamondiferous eclogites from Siberiaccedilremnants ofArchean oceanic-crust Geochimica et Cosmochimica Acta 585191^5207

Jacob D E Fung A Jagoutz E amp Pearson D G (2003) Petrologyand geochemistry of eclogite xenoliths from the Ekati kimberlitesarea Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA_239

Jacob D E Bizimis M amp Salters V J M (2005) Lu^Hf and geo-chemical systematics of recycled ancient oceanic crust evidencefrom Roberts Victor eclogites Contributions to Mineralogy and

Petrology 148 707^720Jerde E A Taylor L A Crozaz G Sobolev NV amp SobolevV N

(1993) Diamondiferous eclogites from Yakutia Siberiaccedilevidencefor a diversity of protoliths Contributions to Mineralogy and Petrology

114 189^202Jones A G Ferguson I J Chave A D Evans R L amp

McNeice GW (2001) The electric lithosphere of the Slave cratonGeology 29 423^426

Kelemen P B Shimizu N amp Dunn T (1993) Relative depletionof niobium in some arc magmas and the continental-crust-partitioning of K Nb La and Ce during meltrockreaction in the upper-mantle Earth and Planetary Science Letters 120111^134

Kinzler R J (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition Application to mid-ocean ridge basalt petrogenesis Journal of Geophysical Research 102853^874

Klemme S van der Laan S R Foley S F amp Guenther D (1995)Experimentally determined trace and minor element partitioningbetween clinopyroxene and carbonatite melt under upper mantleconditions Earth and Planetary Science Letters 133 439^448

Konzett J Armstrong R A Sweeney R J amp CompstonW (1998)The timing of MARID metasomatism in the Kaapvaal mantle anion probe study of zircons from MARID xenoliths Earth and

Planetary Science Letters 160 133^145Kopylova M G Russell J K amp Cookenboo H (1999) Mapping the

lithopshere beneath the north central Slave Craton In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 468^479

Krogh E (1988) The garnet^clinopyroxene iron^magnesiumgeothermometerccedila reinterpretation of existing experimentaldata Contributions to Mineralogy and Petrology 99 44^48

Kusky T M (1989) Accretion of Archean Slave Province Geology 1763^67

LeCheminant A N amp Heaman L M (1989) Mackenzie igneousevents Canada Middle Proterozoic hotspot magmatismassociated with ocean opening Earth and Planetary Science Letters

96 38^48LeCheminant A N Richardson D G DiLabio R N W amp

Richardson K A (1996) La recherche de diamants au CanadaGeological Survey of Canada Open File 3228 268pp

MacGregor I D amp Carter J L (1970) The chemistry of clinopyrox-enes and garnets of eclogite and peridotite xenoliths from theRoberts Victor mine South Africa Physics of the Earth and PlanetaryInteriors 3 391^397

MacGregor I D amp Manton W I (1986) Roberts Victor eclogitesancient oceanic crust Journal of Geophysical Research 91 14063^14079

MacKenzie J M amp Canil D (1999) Composition and thermal evo-lution of cratonic mantle beneath the central Archean SlaveProvince NWT Canada Contributions to Mineralogy and Petrology

134 313^324

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1872

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

McCandlessT E amp Gurney J J (1989) Sodium in garnet and potas-sum in clinopyroxene criteria for classifying mantle eclogites InRoss J (ed) Kimberlites and Related Rocks Carlton Blackwellpp 827^832

McDonoughW F amp Sun S-S (1995) The composition of the EarthChemical Geology 120 223^253

Navon O amp Stolper E (1987) Geochemical consequences of meltpercolation the upper mantle as a chromatographic columnJournal of Geology 95 285^307

Nikogosian I K amp Sobolev AV (1997) Ion-microprobe analysis ofmelt inclusions in olivine experience in estimating the olivine^meltpartition coefficients of trace elements Geochemistry International 35119^126

Norman M D Griffin W L Pearson N J Garcia M O ampOrsquoReilly S Y (1998) Quantitative analysis of trace element abun-dances in glasses and minerals a comparison of laser ablationinductively coupled plasma mass spectrometry solution inductivelycoupled plasma mass spectrometry proton microprobe and elec-tron microprobe data Journal of Analytical Atomic Spectrometry 13477^482

Norman M D Garcia M O amp Pietruszka A J (2005) Trace-element distribution coefficients for pyroxenes plagioclase andolivine in evolved tholeiites from the 1955 eruption of KilaueaVolcano Hawairsquoi and petrogenesis of differentiated rift-zonelavas American Mineralogist 90 888^899

OrsquoHara M J amp Herzberg C (2002) Interpretation of trace elementand isotope features of basalts Relevance of field relations petrol-ogy major element data phase equilibria and magma chambermodeling in basalt petrogenesis Geochimica et Cosmochimica Acta 662167^2191

PadghamW A amp FysonW K (1992) The Slave Province a distinctcraton CanadianJournal of Earth Sciences 29 2072^2086

Pearson D G Davies G R amp Nixon P H (1993) Geochemical con-straints on the petrogenesis of diamond facies pyroxenites from theBeni Bousera peridotite massif North Morocco Journal of Petrology34 125^172

Pearson N J Griffin W L Doyle B J OrsquoReilly S Y vanAchterbergh E amp Kivi K (1999) Xenoliths from kimberlitepipes of the Lac de gras area Slave Craton Canada In GurneyJ J Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite ConferenceCapeTown Red Roof Design pp 644^658

Perfit N R Gust D A Bence A E Arculus R J amp Taylor S R(1980) Chemical characteristics of island-arc basalts implicationsfor mantle sources Chemical Geology 30 227^256

Polat A Hofmann A W amp Rosing M T (2002) Boninite-likevolcanic rocks in the 37^38 Ga Isua greenstone belt WestGreenland geochemical evidence for intra-oceanic subductionzone processes in the early Earth Chemical Geology 184 231^254

Poudjom DjomaniY H GriffinW L OrsquoReilly SY amp Doyle B J(2005) Lithospheric domains and controls on kimberlite emplace-ment Slave Province Canada evidence from elastic thickness andupper mantle composition Geochemistry Geophysics Geosystems 6doi1010292005GC0009978

Presnall D C Gudfinnsson G H amp Walter M J (2002)Generation of mid-ocean ridge basalts at pressures from 1 to 7GPa Geochimica et Cosmochimica Acta 66 2073^2090

Schmidberger S S Heaman L M Simonetti A Creaser R A ampCookenboo H O (2005) Formation of Paleoproterozoiceclogitic mantle Slave Province (Canada) Insights from in-situHf and U^Pb isotopic analyses of mantle zircons Earth and

Planetary Science Letters 240 621^633

Schmidberger S S Simonetti A Heaman L M Creaser R A ampWhiteford S (2007) Lu^Hf in-situ Sr and Pb isotope and traceelement systematics for mantle eclogites from the Diavik diamondmine Evidence for Paleoproterozoic subduction beneath the Slavecraton Canada Earth and Planetary Science Letters 254 55^68

Schulze D J amp Helmstaedt H (1988) Coesite-sanidine eclogitesfrom kimberlite products of mantle fractionation or subductionJournal of Geology 96 435^443

Smithies R H Champion D C amp Sun S S (2004) The case forArchaean boninites Contributions to Mineralogy and Petrology 147705^721

Smyth J R Caporuscio F A amp McCormick T C (1989) Mantleeclogites evidence of igneous fractionation in the mantle Earth andPlanetary Science Letters 93 133^141

Sobolev NV amp Lavrentrsquoyev Y G (1971) Isomorphic sodium admix-ture in garnets formed at high pressure Contributions to Mineralogy

and Petrology 31 1^12Stalder R Foley S F Brey G P amp Horn I (1998) Mineral aqueous

fluid partitioning of trace elements at 900^12008C and 30^57GPa New experimental data for garnet clinopyroxene and rutileand implications for mantle metasomatism Geochimica et

Cosmochimica Acta 62 1781^1801Sun S-S amp McDonough W F (1989) Chemical and isotopic sys-

tematics of oceanic basalts implications for mantle compositionsand processes In Saunders A D amp Norris M J (eds)Magmatism in the Ocean Basins Geological Society London Special

Publications 42 313^345Taura H Yurimoto H Kurita K amp Sueno S (1998) Pressure

dependence on partition coefficients for trace elements between oli-vine and the coexisting melts Physics and Chemistry of Minerals 25469^484

Taylor L A amp Neal C R (1989) Eclogites with oceanic crustal andmantle signatures from the Bellsbank kimberlite South AfricaPart I Mineralogy petrography and whole rock chemistryJournal of Geology 97 551^567

van Achterbergh E Ryan C G amp Griffin W L (1999) Glitter online intensity reduction for the laser ablation inductively coupledplasma spectrometry Abstract 9th V M GoldschmidtConference LPI Boston MA 905 pp Abstract no 7215

van Breemen O DavisW J amp King J E (1992) Temporal distribu-tion of granitoid plutonic rocks in the Archean Slave ProvinceNorthwest Canadian Shield Canadian Journal of Earth Sciences 292186^2199

Vervoort J D Patchett P J Blichert-Toft J amp Albare de F (1999)Relationship between Lu^Hf and Sm^Nd isotopic systems in theglobal sedimentary system Earth and Planetary Science Letters 16879^99

Viljoen K S Smith C B amp Sharp Z D (1996) Stable and radio-genic isotope study of eclogite xenoliths from the Orapa kimberliteBotswana Chemical Geology 131 235^255

Walter M J (1998) Melting of garnet peridotite and theorigin of komatiite and depleted lithosphere Journal of Petrology 3929^60

Wasserburg G J Jacobsen S B DePaolo D J McCulloch M Tamp Wen T (1981) Precise determination of SmNd ratios Sm andNd isotopic abundances in standard solutions Geochimica et

Cosmochimica Acta 45 2311^2323Westerlund K J Shirey S B Richardson S H Carlson R W

Gurney J J amp Harris JW (2006) A subduction wedge origin forPaleoarchean peridotitic diamonds and harzburgites from thePanda kimberlite Slave craton evidence from Re^Os isotope sys-tematics Contributions to Mineralogy and Petrology 152 275^294

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1873

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Page 22: Origins of Xenolithic Eclogites and Pyroxenites from the Central

εHf

εNd

Ekati eclogitesminus100

0

100

200

300

400

500

minus40 0 40 80 120 160 200 240 280

gt in YK1915

gt in VR67112bεNd=198 εHf=2532

minus40

0

40

80

120

160

minus40 minus20 0 20 40 60

minus60

minus40

minus20

0

20

40

070 071 072 073 074 075 076

North Americankimberlite field

εNd

87Sr86Sr

cpx in YK1914

Ekatieclogites

DM

DM

DM

DM

(a)

(b)

pyroxenites

low-Mg

high-Mg

volatile-rich

eclogites

Diavik low-TA154 S kimberlite

cpx in YK1911

Fig 12 (a) Sr^Nd for cpx and (b) Nd-Hf isotope diagram for garnet cpx and reconstructed whole-rocks (symbols with bold outline) Finedashed lines in (b) connect data points from the same sample with garnets having more radiogenic isotopic compositions Uncertainties forreconstructed whole-rocks combine 20 modal uncertainty with uncertainties on each mineralrsquos isotope composition Error bars on individualminerals may be smaller than the symbols Also shown are the isotopic compositions of the depleted mantle (DM DePaolo 1981 Griffin et al2000 Bell amp Tilton 2002) Range of 87Sr86Sr in eclogites from Ekati (Lac de Gras area 100 km north of Diavik Jacob et al 2003) in NorthAmerican kimberlites and compositions of low-temperature eclogites from Diavik and of Lac de Gras A154 South kimberlite fromSchmidberger et al (2007)

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1864

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

one low-Mg eclogite has eNd similar to the depleted mantle(Fig 12b)Clinopyroxene in the phlogopitized eclogite YK1911

has unradiogenic Nd and Hf and strongly radiogenicSr whereas the coexisting garnet has radiogenic Nd andHf Although this sample contains 50 vol phlogopiteits whole-rock isotopic composition was calculated

from garnet and cpx alone because the contributionfrom phlogopite to the Hf and Nd budget is 52 and51 respectively (unpublished data) The whole-rockhas radiogenic Nd but has lower eHf than anyother sample analysed in accord with the extreme Hfabundance (53 ppm) and low LuHf of the cpx

DISCUSS IONOrigins of eclogites beneath the SlaveCratonHigh-Mg and high-Ca eclogites

With regard to their major- and trace-element contentsthe high-Mg and high-Ca eclogites from Lac de Gras arebroadly similar to variably evolved oceanic gabbros(Fig 13b and c) which may indicate an origin as subductedoceanic crust Compatible trace-element abundancesare useful to deduce potential crustal protoliths as theseelements are less susceptible to post-formation modification(Jacob 2004) The flat HREE patterns in the high-Mg andhigh-Ca eclogites are compatible with a low-pressureorigin as the oceanic crust is a product of moderate degreesof partial melting mainly within the spinel stability field(eg Presnall et al 2002) whereas melting with residualgarnet would qualitatively be expected to produce stronglyfractionated REE patterns Positive Sr and Pb anomaliesalthough compatible with a plagioclase signature couldalso be due to cryptic metasomatism (Dawson 1984) ofthe eclogites during their residence in the lithosphericmantle These anomalies are probably not related to kim-berlite infiltration as the whole-rock eclogite compositionswere reconstructed from analyses of fresh mineral coresLow ZrSm similar to that in primitive gabbros is

observed for many of the high-Mg and high-Caeclogites (Fig 13b and c) and is independent of the ratioof garnetcpx and whether or not the whole-rock wasreconstructed with rutile (Electronic Appendix 10) Thenegative trend of Mg-number against Yb in garnet(Fig 14) argues against igneous processes involving garnetaccumulation (Taylor amp Neal 1989) as suggested forexample for some high-Mg eclogites from West Africa(Barth et al 2002b) Rather the anti-correlation suggeststhat olivine fractionation played a role in the formation ofthe protoliths Fractionating olivine would excludeYb (Dolivinebasalt

frac14 002 Nikogosian amp Sobolev 1997)and incorporate Mg relative to Fe (Dfrac1415 and 05 respec-tively Taura et al1998) leading to progressive depletion ofMg and enrichment of Yb in the residual melt and henceanti-correlated Mg and Yb in the whole-rock This isinherited after eclogitization All eclogite types (exceptVR43452) plot along the olivine fractionation trend sug-gesting that they formed by a similar processHigh-Mg eclogites have lower

PREE than high-Ca

eclogites and some appear to have positive Eu anomalies

high-Ca

Th Nb La Ce Pb Pr Sr P Nd Zr Sm Eu Ti Dy Y Yb Lu

gabbrosSE Indian ridge

gabbrosOman ophiolite

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

high-Mg

RockNMORB(a)

VR43452

YK1949

low-Mg

boninite-like melts

01

001

1

10

(b)

01

001

1

10

(c)

01

001

1

10

Th Nb La Ce Pb Pr Sr P Nd Zr SmEu Ti Dy Y Yb Lu

Fig 13 Trace-element patterns of (a) low-Mg (b) high-Mg and(c) high-Ca eclogites compared with boninite-like volcanic rocksfrom the Isua greenstone belt (Polat et al 2002) gabbros from theOman ophiolite (Benoit et al 1996) and the SE Indian Ridge (Hartet al 1999) normalized to N-MORB of Sun amp McDonough (1989)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1865

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

(EuEu up to 14) similar to more primitive cumulates(troctolite gabbros and olivine gabbros) reported in the lit-erature (Fig 13b) Although the sizes of these Eu anomaliesare not outside the analytical uncertainty their correlationwith CeYb in reconstructed eclogites supports their signif-icance (r2frac14066 nfrac14 4) The more primitive nature ofhigh-Mg eclogites is consistent with higher Mg-numbersrelative to high-Ca eclogites and may be due to olivineaccumulation during formation of their protolithsThe two samples with the lowest HREE abundanceshave the highest EuEu which is compatible with varia-tions in plagioclase modesMore evolved protoliths with higher pyroxeneplagio-

clase plus trapped melt could account for small or absentpositive Eu anomalies and higher

PHREE of high-Ca

eclogites (Fig 13c) Although the reason for the frequentassociation of graphite or diamond with high-Ca eclogitesis unclear (Aulbach et al 2003) the frequent occurrence ofkyanite in high-Ca eclogites also observed at Ekati justnorth of Lac de Gras has been argued to support a crustalprotolith (Jacob 2004) Kyanite andor corundum havebeen suggested to exsolve from high-temperature high-Alclinopyroxenes (Caporuscio amp Smyth 1990) Eclogitesfrom Lac de Gras show no exsolution textures suggestiveof such an origin although recrystallization can obscureor destroy microstructural evidence for exsolution(eg Griffin et al 1984)Modelling shows that qualitatively plagioclase accumu-

lation from a low-pressure melt derived from MORBmantle can account for positive Pb Sr and Eu and negativeZr anomalies relative to N-MORB whereas cpx accumula-tion or trapped melt increases Zr Y MREE and HREE

abundances (Fig 15) Some melt addition (for exampleby entrapment) seems to be required to increase all trace-element abundances to those observed for the eclogitesAlthough high-Ca and high-Mg eclogites could representshallow and deep portions of the same oceanic crust thehigher average temperatures obtained for the high-Caeclogites the frequent occurrence of carbon in high-Caeclogites and its absence in high-Mg eclogites the strongLREEHREE fractionation observed for six of seven high-Ca eclogites and the narrow range of compositions of thehigh-Mg eclogites suggest that they may have protolithsthat are not cogenetic Some discrepancies in particularaffecting the incompatible elements may be due to post-crystallization processes acting upon the protoliths suchas dehydration melting and later metasomatism that isevident in peridotites from the same locality (Griffin et al1999)Low K2OTh for high-Ca and high-Mg eclogites

(average 08 Table 4) relative to assumed crustal

0

2

4

6

8

10

30 40 50 60 70 80

Mg

VR43452

Yb

GARNET

pyroxenites

low-Mghigh-Mghigh-Ca

eclogites

VR19673ecl-g2

Fig 14 Garnet Mg-number vs Yb (ppm including 2s uncertainty)Lines qualitatively show separate correlations for eclogites (exceptoutlierVR43452) and pyroxenites (except outlierVR19673ecl-g2)

high-Ca

high-Mg

001

01

1

10

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

melt08 melt02 cryst cpx01 melt09 cryst

plag

(a)

001

01

1

10(b)

RockNMORB

La Ce Pb Pr Sr Nd Zr Sm Eu Ti Dy Y Yb Lu

Fig 15 Reconstructed (a) high-Ca and (b) high-Mg eclogitescompared with compositions of plagioclase and cpx crystallizingfrom a MORB melt modelled using distribution coefficients ofNorman et al (2005)The MORB was calculated as a 10 equilibriummelt of MORB pyrolite [composition from Kelemen et al (1993)except Pb Pr Sr Y and Lu which were calculated as residues from1 fractional melting of pyrolite from McDonough amp Sun (1995)using melt modes of Kinzler (1997) and distribution coefficients ofNorman et al (2005)] Also shown are mixtures of crystals (consistingof 50 plagioclase plus 50 cpx) and melt in the proportions 0901and 0208

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1866

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

precursors (eg gabbros SE Indian Ridge average 22Hart et al 1999) may reflect the preferred loss of fluid-mobile elements during dehydration (Becker et al 2000)although large uncertainties on these low-abundance ele-ments in the constituent minerals and reconstructedwhole-rocks makes this interpretation less than robustThe REE patterns of most high-Mg eclogites are too flatto represent melting residues as partial melting shouldproduce low LREEHREE (Electronic Appendix 10) Incontrast the marked LREE depletion of five of sevenhigh-Ca eclogites (LaLufrac14 02^08) contrasts with that ofassumed evolved gabbroic protoliths (LaLufrac1415^32Hart et al 1999) and can be modelled as a consequence ofpartial melting during subduction after eclogitization (iein the presence of garnet which retains HREE relative toLREE eg Barth et al 2002a) has taken place (Aulbachet al 2003)Our results are compatible with previous studies on

eclogite xenoliths from the Slave Craton includingthose from the same kimberlite that have linked thesesamples or some of these samples to Paleoproterozoicsubduction events (the 195^191Ga formation of theHottah terrane and the 18Ga formation of the GreatBear magmatic arc) in the Slave craton (Griffin et al1999 Kopylova et al 1999 Heaman et al 2002 2006Schmidberger et al 2005 2007)The compositional features discussed above although

compatible with crustal protoliths are not necessarilydefinitive of a subduction origin However if direct crystal-lization from basaltic melts at high pressure (ie mantledepths) is invoked instead specific conditions would haveto be met to satisfy the observations The Mg-number ineclogites from Lac de Gras are lower than those expectedof primary high-pressure mantle melts (eg Walter 1998OrsquoHara amp Herzberg 2002) requiring crystallization athigh pressures after olivine fractionation This originwould further require shielding from peridotitic wallrockwith which the fractionated melts would no longer be inequilibrium (Barth et al 2001) In such a setting the lowrelative abundances of fluid-mobile elements comparedwith those of less fluid-mobile elements and of LREErelative to HREE (Table 4) would reflect an expulsion offluids during crystallization of the protoliths and latermelt mobilization respectively The eclogites clearly donot represent garnet plus cpx cumulates which wouldrequire a positive correlation of Mg-number with Yb thatis at variance with the observed negative correlation ineclogitic garnets from this study (Fig 14)

Low-Mg eclogites

Garnets in low-Mg eclogites compositionally overlap withthose in massive eclogites from Jericho (NW of the Lacde Gras kimberlites) that were suggested to have a high-pressure intrusive origin based on the similarity of theirmajor-element compositions to those of Group A eclogites

fromsouthernAfrica forwhichamantleoriginhasbeenpro-posed (Kopylova et al1999) Garnets in low-Mg eclogites inthe present study have major-element compositions moresimilar to garnets in groupB rather thangroupAeclogitesThe positive slope in N-MORB-normalized HREE

abundances displayed by the low-Mg eclogites is clearlydistinct from other eclogite types and is similar to those ofboninite-like second-stage melts that are closely associatedwith ultramafic volcanics (Smithies et al 2004) (Fig 13a)The range of Mg-numbers in the low-Mg eclogites (55^66)overlaps with that of Archaean second-stage melts fromIsua (58^75) and Abitibi (52^82 Smithies et al 2004)These similarities might indicate that the low-Mg eclogiteshave an origin in a previously depleted mantle source suchas supra-subduction zone mantle Subchondritic ZrHf ofthe low-Mg eclogites contrasts with the supra-chondriticZrHf of the high-Mg and high-Ca eclogites (Aulbach et

al in preparation) and is consistent with previous cpx-controlled depletion in an arc mantle source This sourcecould have been subsequently remelted to generate theboninite-like precursor to the low-Mg eclogites as sug-gested for the protoliths of eclogites from Udachnaya(Jacob amp Foley 1999)Although some second-stage melts to which the low-Mg

eclogites are similar are closely associated with komatiitesand could contain a plume component the radiogenic ordepleted mantle-like Nd and Hf and unradiogenic Sr inthe low-Mg eclogites do not support plume involvement

Trace-element and Sr^Nd^Hf isotopeconstraints on post-formationmetasomatismSeveral low-Mg and high-Ca eclogites show negativeslopes in the LREE which may be indicative of interactionwith small amounts of LREE-enriched melt (Navon ampStolper 1987) amounting to cryptic metasomatism(ie not accompanied by crystallization of new phases) inthe nomenclature of Dawson (1984) High Nb abundancesrelative to elements of similar compatibility displayed bysome low-Mg and high-Ca eclogites (Fig 10) could resultfrom overestimation of rutile modes However high-CaeclogiteVR40345 is rutile-free and has been reconstructedwithout rutile its high Nb abundance is accompanied byLREE enrichment indicating that metasomatic additionmay be responsibleLow-Mg eclogites show a positive correlation of La with

Nb (Electronic Appendix 11a) indicating concomitantaddition of LREE and Nb This suggests that the metaso-matic agent was a solute-rich fluid or silicate silicic melt ashydrous fluids and carbonatites do not carry enough HFSEto affect the budget of eclogites (Green amp Wallace 1988Klemme et al 1995 Adam et al 1997 Stalder et al 1998)Of the samples that are enriched in Nb all but one (high-Ca eclogiteVR50909) show a simultaneous enrichment in

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1867

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Hf (Electronic Appendix 11b) Together these trace-element relationships suggest that both SmNd and LuHfwere lowered in these samples Their low parent^daughterratios contrast with higher ratios in unenriched samplesand resulted in the ingrowth of a range of Nd and Hf iso-tope compositions Compared with the low-Mg and high-Ca eclogites the high-Mg eclogites have a narrow range ofLa and Nb concentrations which correlates with a narrowrange in Nd and Hf isotope compositions (Fig 12)Precise model ages cannot be calculated because parent^

daughter ratios are available only from in situ analyseswhich have relatively large uncertainties Howeverevidence for long-term enrichment and depletion may bediscerned Low-Mg eclogites even those that are enrichedin LREE have relatively high SmNd as a result of havingoverall positive slopes in their REE patterns relative toMORB whereas the high-Mg eclogites have relativelyflat REE patterns and therefore lower SmNd (Fig 11)This translates into lower eNd for high-Mg eclogites thanfor low-Mg eclogites In contrast they show no systematicdifference with respect to 176Hf177Hf Because the low- andhigh-Mg eclogites are proposed to have different origins asdiscussed above and therefore did not necessarily have thesame initial isotopic composition and formation age theseisotopic differences cannot be used to estimate their agesA long-term coupling of trace-element and isotopic

abundances is also indicated by the fact that low-Mg eclo-gite YK1943 which has experienced Nb and LREE addi-tion has lower eNd (0) and eHf (thorn13) than low-Mg eclogiteVR43469 with higher LuHf and SmNd (eNdfrac14thorn47eHffrac14thorn128) If low-Mg eclogites YK1943 and VR43469originally had identical initial Nd and Hf isotope composi-tions and if the parent^daughter ratios are accurate andwere modified only once during their evolution shortlyafter metamorphism (when their isotopic compositionwould have been reset) this overprint can be calculatedfor both isotopic systems to have occurred at 17GaThis would suggest that Hf and Nd were added duringthe same event In contrast high-Mg eclogites VR67360and YK1949 would have had identical Ndisotope compositions at 10Ga whereas their Hf isotopecompositions are disturbed in that the sample with thehigher parent^daughter ratio has a lower 176Hf177Hf thanthe sample with the lower parent^daughter ratio Theseresults suggest that the LREE and HFSE enrichmentobserved in different samples occurred during separateevents that did not affect the eclogites equallyExtreme HFSE enrichment leading to the growth of

zircon and rutile and LREE addition leading to thegrowth of apatite in eclogites from the Jericho kimberlitein the northern Slave craton have been ascribed to twoseparate metasomatic events at around 18Ga (quasi-coincident with the age of formation and metamorphism)and 13^10Ga respectively (Heaman et al 2002 2006

Schmidberger et al 2005) The authors associated theseevents with Paleoproterozoic subduction and aMesoproterozoic thermal disturbance in the Slave litho-spheric mantle possibly linked to the Mackenzie igneousevent Mesoproterozoic Nd model ages mostly between13 and 14Ga were reported for eclogites from thenearby Ekati kimberlites (Jacob et al 2003) Although wedo not observe zircon or apatite in volatile-free eclogitesfrom Lac de Gras and the age constraints obtained fromour data are weak they support a link between theseevents described for other eclogite populations and themetasomatic enrichment observed in our studyThe Sr isotope data in the present study are similar to

those of Schmidberger et al (2007) who suggested that thedepleted-mantle-like values are consistent with derivationof the protoliths from an oceanic mantle source However87Sr86Sr in volatile-free eclogites in the present study isremarkably constant considering their Nd and Hf isotopevariability (Fig 12) and the variability of trace-elementpatterns (Fig 10) Given the evidence for dehydration andmelting in many of the eclogites fractionation of highlyincompatible Rb from Sr and differential ingrowth ofradiogenic Sr would be expectedThis may indicate a rela-tively recent overprint and considering the similarityof 87Sr86Sr to present-day depleted mantle which is signif-icantly lower than 87Sr86Sr of North American kimber-lites a juvenile (asthenospheric) source

PyroxenitesPyroxenites from Lac de Gras often show relationshipsin their major-element contents that are different fromthose observed in the eclogites For example pyroxeniticgarnets show a negative correlation between CaO andNa2O contents whereas these oxides are positively corre-lated for garnets in eclogites (Fig 5) CaO is well correlatedwith C2O3 in pyroxenitic garnets but not in eclogitic gar-nets (Fig 4) Mg-number is positively correlated withYb abundance for pyroxenitic garnets but anti-correlatedfor eclogitic garnets (Fig 14) This suggests that the pyrox-enites in the present study are not related to the eclogitesby fractionation or accumulation processes as proposedfor the origin of some pyroxenites (Foley et al 2003)Pyroxenites have also been interpreted as the products

of crystal segregation in magma channels with alteredoceanic crust as the magma source this would explaintheir Eu anomalies and stable and radiogenic isotopecompositions (Pearson et al 1993 Becker 1996) As thehigh-pressure liquidus phases of most terrestrial basaltsare cpx and garnet (OrsquoHara amp Herzberg 2002) partialmelts of eclogites with basaltic composition wouldcrystallize cpx and garnet but not opx More than half ofthe pyroxenites from Lac de Gras contain opx suggestingthat they probably did not crystallize directly fromeclogite-derived melts

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1868

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Low-Cr websterites and pyroxenites could be crystalli-zation products of mafic melts that interacted with perido-titic mantle thus becoming more refractory The positivecorrelation of Yb and Mg-number in garnet would allowan origin by fractional crystallization from such meltsAlternatively they formed by reaction of silicic (possiblyeclogite-derived) melts with mantle peridotites(eg Kelemen et al 1993 Aulbach et al 2002) This agreeswith the similarity of hypothetical melts in equilibriumwith pyroxenitic garnets in most samples to eclogite-derived melts (Electronic Appendix 12) Niobium concen-trations in all hypothetical melts are much higher than ineclogite-derived melts regardless of whether or not rutileis assumed to be in the residue The same is true for LaCe and Zr concentrations in some samples which mayindicate that NbZr and LREE were added to thepyroxenites some time after their formation as discussedin the previous section for some eclogites Because mini-mum ages for pyroxenitic rutile with an unradiogenicHf isotope composition (calculated by projecting the176Hf177Hf to the model depleted mantle isotopic evolutioncurve assuming zero ingrowth) range to 178066Ga(Aulbach et al in preparation) it appears likely that thepyroxenites were not formed recently (eg during kimber-lite-related or precursory melt migration) and that some ofthe samples are at least MesoproterozoicGarnet pyroxeniteYK1952 consists of schlieren of equili-

brated garnet pyroxenite in a lsquomatrixrsquo of spongy cpxfine cpx^opx intergrowths embayed garnet and meltpatches (Fig 2f) Garnet in the pyroxenite schlieren hasvery low LREE and HFSE abundances Orthopyroxenein the matrix has distinctly higher Mg-number andCr-number and contains more CaO than opx from thepyroxenite schlieren whereas opx rims in the schlierenare compositionally similar to opx in the matrix The con-siderable compositional and textural disequilibrium inthis sample suggests that this is a young feature causedby reaction with fluids shortly before or at the time ofentrainmentAs a consequence of the small sample sizes of most of

the pyroxenites studied here a comprehensive isotopicdataset is not available Clearly more such data areneeded in addition to stable isotope data to obtain betterconstraints on the origins of this rock type Consideringthe microstructural and compositional variability ofthe pyroxenites and major-element relationships oppositeto those observed for eclogites it appears certain that pyr-oxenites form by a variety of processes some of which areunrelated to the formation of eclogites and their protoliths

Volatile-rich eclogitesVolatile-rich eclogites from Lac de Gras display the mostextreme Hf^Nd isotope systematics of all the analysedsamples Garnets in phlogopitized (YK1911) and in cal-cite-bearing eclogite (VR67112b) have the most radiogenic

Nd (eNdfrac14 2486 and 1927 respectively) Garnet inVR67112b has extremely high eHf (2516) Clinopyroxene inYK1911 has the lowest eHf measured for any mineralseparate from Lac de Gras (53) in accord with the veryhigh Hf abundance in its cpx (5 ppm) In contrastgarnet in this sample has radiogenic Hf (eHffrac14 136) Thisdiscrepancy is at least in part due to their relatively lowequilibration temperatures (7698C and 8558C respec-tively) which allowed the isotopic compositions of cpxand garnet to evolve separately at their low andhigh parent^daughter element ratios respectively forextended periods of time and possibly since formationAccordingly the cpx in YK1911 gives a minimum Hf ageof 27GaSlab-derived fluids could have been introduced during

the 27Ga collision of the western Slave Craton domainincluding the thick plume-derived deep lithosphericmantle layer and the eastern domain with its highlydepleted lithosphere This would also be consistent withthe observation that volatile-rich eclogites occur at shallowdepths in the mantle (TKroghfrac14 685^8558C correspondingto depths of 85^115 km along a 40 mWm2 geotherm)where the bulk of dehydration and production of solute-rich melts in a subducting slab might be expected to occur(Becker et al 2000 Hermann et al 2006) Their negativeEu anomalies might indicate that the fluid was derivedfrom the shallow portion of the slab that experiencedplagioclase fractionation The presence of carbonate inVR67112b requires a carbonated sourceHigh 87Sr86Sr in cpx in YK1911 is probably the conse-

quence of isotopic equilibration with high-RbSr phlogo-pite after long-term radiogenic ingrowth while Hfaddition led to retarded ingrowth of radiogenic HfThese systematics combined with radiogenic Nd appear

inconsistent with the involvement of sediments in the pet-rogenesis of this sample (eg Perfit et al 1980 Vervoortet al 1999) or a hydrous fluid such as those responsible forMARID metasomatism in the Kaapvaal craton (Konzettet al 1998)Whatever the origin of the fluids the apparentHf mobility suggests that they were not in equilibriumwith a rutile-bearing source and also that they weresolute-rich in nature (ie hydrous melts able to dissolveHFSE) and the relatively high HREE abundances of thevolatile-rich eclogites indicate a garnet-free source suchas amphibolite This allowed evolution of the sample atrelatively high time-integrated SmNd

SUMMARY AND CONCLUSIONS

(1) The Lac de Gras kimberlites carried abundant eclo-gite and pyroxenite xenoliths to the surface Threemain eclogite types are distinguished in terms ofmajor-element composition and trace element

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1869

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

patterns high-Mg high-Ca and low-Mg eclogitesThe garnets of all eclogite types have anti-correlatedMg-number and Yb pointing to olivine fractionationas a controlling factor in the formation of their pre-cursor rocks Their flat or positively sloped HREEpatterns qualitatively indicate that the melt parentalto the precursor rocks was not in equilibrium with agarnet-bearing residue

(2) Reconstructed high-Mg eclogites and high-Caeclogites have flat HREE patterns positive Pb andSr anomalies and low ZrSm similar to oceanic gab-bros reported in the literature High-Mg eclogiteshave variably low HREE contents and some havesmall positive Eu anomalies possibly suggestive ofmore primitive relatively plagioclase-rich protolithswhereas high-Ca eclogites have higher HREE abun-dances and small or absent Eu anomalies that requiremore cpx- andor melt-rich protoliths if a subductionorigin is assumed

(3) Low-Mg eclogites have distinctive trace-element pat-terns that show a marked positive slope in the HREEand subchondritic ZrHf They resemble someArchaean second-stage melts (lsquoboninitesrsquo) and theirprotoliths appear to have formed from a previouslydepleted source

(4) After formation the eclogites were subject to a varietyof processes Most have low ratios of fluid-mobile vsless fluid-mobile elements possibly as a result of dehy-dration of the protoliths In addition some of the low-Mg and high-Ca eclogites have low LREEHREEconsistent with partial melt lossWhereas the low-Mgeclogites show concomitant enrichment in La Nb andZr some high-Mg and high-Ca eclogites show evi-dence for variable and unsystematic (decoupled)LREE and Nb addition

(5) Isotopically the trace-element characteristics of thedifferent eclogite types translate into lower eNd forhigh-Mg eclogites (negative slope in MREEfrac14 lowerSmNd) than for low-Mg eclogites (positive slope inMREEfrac14higher SmNd) Within the low-Mg groupsamples that show evidence for metasomatism (lowLuHf and SmNd) have lower eNd and eHf than asample that was apparently not metasomatized BothNd and Hf addition can be dated to 17Ga if identi-cal initial isotope compsitions are assumed for thelow-Mg eclogites In contrast the relatively constantand depleted mantle-like 87Sr86Sr of eclogitic cpxdespite variable Nd and Hf isotope compositions andparent^daughter ratios suggests recent input of juve-nile material

(6) Garnet and cpx in pyroxenites show major-elementrelationships different from those in eclogitesTherefore these two rock types are probably notrelated by fractionation processes The presence of

opx precludes direct crystallization from slab-derivedmelts Pyroxenites may be the product of (slab-derived) melts hybridized during assimilation of peri-dotitic mantle

(7) Garnets in phlogopite- and carbonate-bearing eclo-gites have extreme isotopic compositions includingthe most radiogenic Nd and Hf in the datasetNegative Eu anomalies may point to the involvementof a crustal protolith Highly radiogenic Sr radiogenicNd and unradiogenic Hf in the reconstructedphlogopite-bearing sample combined with highHREE abundances and evidence for Hf mobilitymay indicate involvement of solute-rich fluids origi-nating in a garnet- and rutile-free source

ACKNOWLEDGEMENTSHelp in the clean laboratories and with the analyticalfacilities by S Elhlou C Lawson A Sharma P Wielandand M Zhang is greatly appreciated Bill Griffin isthanked for his support and for agreeing to reviewseveral versions of the manuscript despite disagreeingwith many of the conclusions presented here The paperbenefited greatly from the detailed and constructivereviews provided by the editors Gareth Davies andMarjorie Wilson and by Matthias Barth Larry Heamanand an anonymous reviewer This work was funded by aMacquarie University International PostgraduateAward and Postgraduate Research Fund (SA) by anARC SPIRT grant sponsored by Kennecott Canada Incand by ARC Large and Discovery Grants to SYOrsquoRAnalytical data were obtained using instrumentationfunded by ARC LIEF and DEST Systemic InfrastructureGrants industry partners and Macquarie UniversityThis is Publication 481 from the ARC National KeyCentre for Geochemical Evolution and Metallogeny ofContinents (wwwesmqeduauGEMOC)

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online

REFERENCESAdam J GreenT H Sie S H amp Ryan C G (1997) Trace element

partitioning between aqueous fluids silicate melts and mineralsEuropeanJournal of Mineralogy 9 569^584

Aulbach S Stachel T Viljoen K S Brey G P amp Harris J W(2002) Eclogitic and websteritic diamond sources beneath theLimpopo Beltccedilis slab-melting the link Contributions to Mineralogy

and Petrology 143 56^70Aulbach S GriffinW L Pearson N J OrsquoReilly S Y Kivi K amp

Doyle B (2003) Origins of eclogites beneath the central SlaveCraton Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA-0011

Aulbach S Griffin W L Pearson N J OrsquoReilly S Y amp Kivi K(2004a) Mantle formation and evolution Slave Craton constraints

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1870

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

from HSE abundances and Re^Os isotope systematics of sulfideinclusions in mantle xenocrysts Chemical Geology 208 61^88

Aulbach S Griffin W L OrsquoReilly S Y amp McCandless T E(2004b) Genesis and evolution of the lithospheric mantle beneaththe Buffalo HeadTerrane Alberta (Canada) Lithos 77 413^451

Aulbach S GriffinW L Pearson N J OrsquoReilly S Y amp Doyle B(2007) Lithosphere formation in the central Slave Craton(Canada) plume subcretion or lithosphere accretion Contributionsto Mineralogy and Petrology (in press) doi 101007s00410-007-0200-1

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2001)Geochemistry of xenolithic eclogites from West Africa Part IA link between low MgO eclogites and Archean crust formationGeochimica et Cosmochimica Acta 65 1499^1527

Barth M G Foley S F amp Horn I (2002a) Partial melting inArchean subduction zones constraints from experimentally deter-mined trace element partition coefficients between eclogitic miner-als and tonalitic melts under upper mantle conditions PrecambrianResearch 113 323^340

Barth M G Rudnick R L Horn I McDonough W FSpicuzza M J Valley J W amp Haggerty S E (2002b)Geochemistry of xenolithic eclogites from West Africa part 2Origins of the high MgO eclogites Geochimica et Cosmochimica Acta

66 4325^4345Beard B L Fraracci K NTaylor L A Snyder G A Clayton R

A Mayeda T K amp Sobolev NV (1996) Petrography andgeochemistry of eclogites from the Mir kimberliteYakutia RussiaContributions to Mineralogy and Petrology 125 293^310

Becker H (1996) Geochemistry of garnet peridotite massifs fromlower Austria and the composition of deep lithosphere beneatha Palaeozoic convergent plate margin Chemical Geology 134 49^65

Becker H Jochum K P amp Carlson R W (2000) Trace elementfractionation during dehydration of eclogites from high-pressureterranes and the implications for element fluxes in subductionzones Chemical Geology 163 65^99

Bell K amp Tilton G R (2002) Probing the mantle The story fromcarbonatites EOS Transactions American Geophysical Union 83275^277

Benoit M Polvecurren M amp Ceuleneer G (1996) Trace element andisotopic characterisation of mafic cumulates in a fossil mantlediapir (Oman ophiolite) Chemical Geology 134 199^214

BleekerW (2003) The late Archean record a puzzle in ca 35 piecesLithos 71 99^134

BleekerW Ketchum JW F JacksonV A amp Villeneuve M (1999)The central Slave Basement Complex Part I Its structural topol-ogy and autochthonous core Canadian Journal of Earth Sciences 361083^1109

Blichert-Toft J amp Albare de F (1997) The Lu^Hf isotope geochemis-try of chondrites and the evolution of the mantle^crust systemEarth and Planetary Science Letters 148 243^258

Blichert-Toft J Chauvel C amp Albare de F (1997) Separation of Hfand Lu for high precision isotope analysis of rock samples by mag-netic sector-multiple collector ICP-MS Contributions to Mineralogy

and Petrology 127 248^260Brey G P amp Kolaquo hler T (1990) Geothermobarometry in four-phase

lherzolites II New thermobarometers and practical assessment ofexisting thermobarometers Journal of Petrology 31 1353^1378

Brey G P Kolaquo hlerT amp Nickel K G (1990) Geothermobarometry infour-phase lherzolites I Experimental results from 10 to 60 kbJournal of Petrology 31 1313^1352

Caporuscio F amp Smyth J (1990) Trace element crystal chemistryof mantle eclogites Contributions to Mineralogy and Petrology 105550^561

Carbno G B amp Canil D (2002) Mantle structure beneath the SWSlave Craton Canada constraints from garnet geochemistry in theDrybones Bay Kimberlite Journal of Petrology 43 129^142

Coleman R G Lee E D Beatty L B amp Brannock W W (1965)Eclogites and eclogites their differences and similaritiesGeological Society of America Bulletin 76 483^508

Creaser R A Grulaquo tter H Carlson J amp Crawford B (2004)Macrocrystal phlogopite Rb^Sr dates for the Ekati propertykimberlites Slave Province Canada evidence for multiple intru-sive episodes in the Paleocene and Eocene Lithos 76 399^414

Davies R M GriffinW L Pearson N J Andrew A S Doyle BJ amp OrsquoReilly S Y (1999) Diamonds from the deep Pipe DO-27Slave Craton Canada In Gurney J J Gurney J L Pascoe MD amp Richardson S H (eds) Proceedings of the 7th International

Kimberlite Conference CapeTown Red Roof Design pp 148^155Davies R M Griffin W L OrsquoReilly S Y amp Doyle B J (2004)

Mineral inclusions and geochemical characteristics of microdia-monds from the DO27 A154 A21 A418 DO18 DD17 and RanchLake kimberlites at Lac de Gras Slave Craton Canada Lithos 7739^55

DavisW J amp Hegner E (1992) Neodymium isotopic evidence for thetectonic assembly of Late Archean crust in the Slave Provincenorthwest Canada Contributions to Mineralogy and Petrology 111493^504

Davis W J Jones A G Bleeker W amp Grutter H (2003)Lithosphere development in the Slave craton a linked crustal andmantle perspective Lithos 71 575^589

Dawson J B (1984) Contrasting types of upper-mantle metasoma-tism In Kornprobst J (ed) Kimberlites II The Mantle and

Crust^Mantle Relationships Amsterdam Elsevier pp 289^294DePaolo D J (1981) Neodymium isotopes in the Colorado Front

Range and crust^mantle evolution in the Proterozoic Nature 291193^196

Erlank A J amp Kushiro I (1970) Potassium contents of syntheticpyroxenes at high temperatures and pressures Carnegie Institution ofWashingtonYearbook 68 433^439

Foley S F Barth M G amp Jenner G A (2000) Rutilemelt partitioncoefficients for trace elements and an assessment of the influence ofrutile on the trace element characteristics of subduction zonemagmas Geochimica et Cosmochimica Acta 64 933^938

Foley S F Buhre S amp Jacob D E (2003) Evolution of the Archaeancrust by delamination and shallow subduction Nature 421 249^252

Graham I Burgess J L Bryan D Ravenscroft P J Thomas EDoyle B J Hopkins R amp Armstrong K A (1999) Explorationhistory and geology of the Diavik Kimberlites Lac de GrasNorthwest Territories Canada In Gurney J J Gurney J LPascoe M D amp Richardson S H (eds) Proceedings of the 7th

International Kimberlite Conference Cape Town Red Roof Designpp 262^279

Green D H amp Wallace M E (1988) Mantle metasomatism byephemeral carbonatite melts Nature 336 459^462

Green T H Blundy J D Adam J amp Yaxley G M (2000) SIMSdetermination of trace element partition coefficients betweengarnet clinopyroxene and hydrous basaltic liquids at 2^75 GPaand 1080^12008C Lithos 53 165^187

GriffinW L amp OrsquoReilly SY (2007) Cratonic lithospheric mantle Isanything subducted Episodes 30(1) 43^53

Griffin W L Wass S Y amp Hollis J D (1984) Ultramafic xenolithsfrom Bullenmerri and Gnotuk maars Victoria Australia petrol-ogy of a sub-continental crust^mantle transition Journal of

Petrology 25 53^87GriffinW L Doyle B J Ryan C G Pearson N J OrsquoReilly SY

Davies R Kivi K van Achterbergh E amp Natapov LM (1999)

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1871

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Layered mantle lithosphere in the Lac de Gras Area SlaveCraton composition structure and origin Journal of Petrology 40705^727

Griffin W L Pearson N J Belousova E Jackson S E vanAchterbergh E OrsquoReilly SY amp Shee S R (2000)The Hf isotopecomposition of cratonic mantle LAM-MC-ICPMS analysis ofzircon megacrysts in kimberlites Geochimica et Cosmochimica Acta

64 133^147Griffin W L OrsquoReilly S Y Doyle B J Pearson N J

Coopersmith H Kivi K Malkovets V amp Pokhilenko N(2004) Lithosphere mapping beneath the North American plateLithos 77 873^922

Grulaquo tter H S Apter D B amp Kong J (1999) Crust^mantle couplingevidence from mantle-derived xenocrystic garnets In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 307^312

Hart S R Blusztajn J Dick H J B Meyer P S ampMuehlenbachs K (1999) The fingerprint of seawater circulationin a 500-meter section of ocean crust gabbros Geochimica et

Cosmochimica Acta 63 4059^4080Harte B amp Kirkley M B (1997) Partitioning of trace elements

between clinopyroxene and garnet Data from mantle eclogitesChemical Geology 136 1^24

Heaman L M Creaser R A amp Cookenboo H O (2002) Extremeenrichment of high field strength elements in Jericho eclogitexenoliths A cryptic record of Paleoproterozoic subduction partialmelting and metasomatism beneath the Slave craton CanadaGeology 30 507^510

Heaman L M Kjarsgaard B A amp Creaser R A (2004) The tem-poral evolution of North American kimberlites Lithos 76 377^397

Heaman L M Creaser R A Cookenboo H O amp Chacko T(2006) Multi-stage modification of the northern Slave mantlelithosphere Evidence from zircon- and diamond-bearing eclogitexenoliths entrained in Jericho Kimberlite Journal of Petrology 47821^858

Hermann J Cpandler C Hack A amp Korsakov A (2006) Aqueousfluids and hydrous melts in high-pressure and ultra-high pressurerocks implications for element transfer in subduction zones Lithos92 399^417

Hills DV amp Haggerty S E (1989) Petrochemistry of eclogites fromthe Koidu Kimberlite Complex Sierra Leone Contributions to

Mineralogy and Petrology 103 397^422Hoffman P F (1989) Precambrian geology and tectonic history of

North America In Bally A amp Palmer A (eds) The Geology of

North AmericaccedilAn Overview Boulder CO Geological Society ofAmerica pp 447^512

IrelandT R Rudnick R L amp Spetsius Z (1994) Trace-elements indiamond inclusions from eclogites reveal link to Archean granitesEarth and Planetary Science Letters 128 199^213

Irvine G J Pearson D G Kjarsgaard B A Carlson R WKopylova M G amp Dreibus G (2003) Evolution of the litho-spheric mantle beneath Northern Canada a Re^Os isotope andPGE study of kimberlite-derived peridotite xenoliths fromSomerset Island and a comparison to the Slave and Kaapvaalcratons Lithos 71 461^488

Isachsen C E amp Bowring S A (1994) Evolution of the Slave cratonGeology 22 917^920

Jacob D E (2004) Nature and origin of eclogite xenoliths from kim-berlites Lithos 77 295^316

Jacob D E amp Foley S F (1999) Evidence for Archean ocean crustwith low high field strength element signature from diamondifer-ous eclogite xenoliths Lithos 48 317^336

Jacob D Jagoutz E Lowry D Mattey D amp Kudrjavtseva G(1994) Diamondiferous eclogites from Siberiaccedilremnants ofArchean oceanic-crust Geochimica et Cosmochimica Acta 585191^5207

Jacob D E Fung A Jagoutz E amp Pearson D G (2003) Petrologyand geochemistry of eclogite xenoliths from the Ekati kimberlitesarea Extended Abstract 8th International KimberliteConferenceVictoria BC Abstract no FLA_239

Jacob D E Bizimis M amp Salters V J M (2005) Lu^Hf and geo-chemical systematics of recycled ancient oceanic crust evidencefrom Roberts Victor eclogites Contributions to Mineralogy and

Petrology 148 707^720Jerde E A Taylor L A Crozaz G Sobolev NV amp SobolevV N

(1993) Diamondiferous eclogites from Yakutia Siberiaccedilevidencefor a diversity of protoliths Contributions to Mineralogy and Petrology

114 189^202Jones A G Ferguson I J Chave A D Evans R L amp

McNeice GW (2001) The electric lithosphere of the Slave cratonGeology 29 423^426

Kelemen P B Shimizu N amp Dunn T (1993) Relative depletionof niobium in some arc magmas and the continental-crust-partitioning of K Nb La and Ce during meltrockreaction in the upper-mantle Earth and Planetary Science Letters 120111^134

Kinzler R J (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition Application to mid-ocean ridge basalt petrogenesis Journal of Geophysical Research 102853^874

Klemme S van der Laan S R Foley S F amp Guenther D (1995)Experimentally determined trace and minor element partitioningbetween clinopyroxene and carbonatite melt under upper mantleconditions Earth and Planetary Science Letters 133 439^448

Konzett J Armstrong R A Sweeney R J amp CompstonW (1998)The timing of MARID metasomatism in the Kaapvaal mantle anion probe study of zircons from MARID xenoliths Earth and

Planetary Science Letters 160 133^145Kopylova M G Russell J K amp Cookenboo H (1999) Mapping the

lithopshere beneath the north central Slave Craton In Gurney JJ Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite Conference Cape TownRed Roof Design pp 468^479

Krogh E (1988) The garnet^clinopyroxene iron^magnesiumgeothermometerccedila reinterpretation of existing experimentaldata Contributions to Mineralogy and Petrology 99 44^48

Kusky T M (1989) Accretion of Archean Slave Province Geology 1763^67

LeCheminant A N amp Heaman L M (1989) Mackenzie igneousevents Canada Middle Proterozoic hotspot magmatismassociated with ocean opening Earth and Planetary Science Letters

96 38^48LeCheminant A N Richardson D G DiLabio R N W amp

Richardson K A (1996) La recherche de diamants au CanadaGeological Survey of Canada Open File 3228 268pp

MacGregor I D amp Carter J L (1970) The chemistry of clinopyrox-enes and garnets of eclogite and peridotite xenoliths from theRoberts Victor mine South Africa Physics of the Earth and PlanetaryInteriors 3 391^397

MacGregor I D amp Manton W I (1986) Roberts Victor eclogitesancient oceanic crust Journal of Geophysical Research 91 14063^14079

MacKenzie J M amp Canil D (1999) Composition and thermal evo-lution of cratonic mantle beneath the central Archean SlaveProvince NWT Canada Contributions to Mineralogy and Petrology

134 313^324

JOURNAL OF PETROLOGY VOLUME 48 NUMBER 10 OCTOBER 2007

1872

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

McCandlessT E amp Gurney J J (1989) Sodium in garnet and potas-sum in clinopyroxene criteria for classifying mantle eclogites InRoss J (ed) Kimberlites and Related Rocks Carlton Blackwellpp 827^832

McDonoughW F amp Sun S-S (1995) The composition of the EarthChemical Geology 120 223^253

Navon O amp Stolper E (1987) Geochemical consequences of meltpercolation the upper mantle as a chromatographic columnJournal of Geology 95 285^307

Nikogosian I K amp Sobolev AV (1997) Ion-microprobe analysis ofmelt inclusions in olivine experience in estimating the olivine^meltpartition coefficients of trace elements Geochemistry International 35119^126

Norman M D Griffin W L Pearson N J Garcia M O ampOrsquoReilly S Y (1998) Quantitative analysis of trace element abun-dances in glasses and minerals a comparison of laser ablationinductively coupled plasma mass spectrometry solution inductivelycoupled plasma mass spectrometry proton microprobe and elec-tron microprobe data Journal of Analytical Atomic Spectrometry 13477^482

Norman M D Garcia M O amp Pietruszka A J (2005) Trace-element distribution coefficients for pyroxenes plagioclase andolivine in evolved tholeiites from the 1955 eruption of KilaueaVolcano Hawairsquoi and petrogenesis of differentiated rift-zonelavas American Mineralogist 90 888^899

OrsquoHara M J amp Herzberg C (2002) Interpretation of trace elementand isotope features of basalts Relevance of field relations petrol-ogy major element data phase equilibria and magma chambermodeling in basalt petrogenesis Geochimica et Cosmochimica Acta 662167^2191

PadghamW A amp FysonW K (1992) The Slave Province a distinctcraton CanadianJournal of Earth Sciences 29 2072^2086

Pearson D G Davies G R amp Nixon P H (1993) Geochemical con-straints on the petrogenesis of diamond facies pyroxenites from theBeni Bousera peridotite massif North Morocco Journal of Petrology34 125^172

Pearson N J Griffin W L Doyle B J OrsquoReilly S Y vanAchterbergh E amp Kivi K (1999) Xenoliths from kimberlitepipes of the Lac de gras area Slave Craton Canada In GurneyJ J Gurney J L Pascoe M D amp Richardson S H (eds)Proceedings of the 7th International Kimberlite ConferenceCapeTown Red Roof Design pp 644^658

Perfit N R Gust D A Bence A E Arculus R J amp Taylor S R(1980) Chemical characteristics of island-arc basalts implicationsfor mantle sources Chemical Geology 30 227^256

Polat A Hofmann A W amp Rosing M T (2002) Boninite-likevolcanic rocks in the 37^38 Ga Isua greenstone belt WestGreenland geochemical evidence for intra-oceanic subductionzone processes in the early Earth Chemical Geology 184 231^254

Poudjom DjomaniY H GriffinW L OrsquoReilly SY amp Doyle B J(2005) Lithospheric domains and controls on kimberlite emplace-ment Slave Province Canada evidence from elastic thickness andupper mantle composition Geochemistry Geophysics Geosystems 6doi1010292005GC0009978

Presnall D C Gudfinnsson G H amp Walter M J (2002)Generation of mid-ocean ridge basalts at pressures from 1 to 7GPa Geochimica et Cosmochimica Acta 66 2073^2090

Schmidberger S S Heaman L M Simonetti A Creaser R A ampCookenboo H O (2005) Formation of Paleoproterozoiceclogitic mantle Slave Province (Canada) Insights from in-situHf and U^Pb isotopic analyses of mantle zircons Earth and

Planetary Science Letters 240 621^633

Schmidberger S S Simonetti A Heaman L M Creaser R A ampWhiteford S (2007) Lu^Hf in-situ Sr and Pb isotope and traceelement systematics for mantle eclogites from the Diavik diamondmine Evidence for Paleoproterozoic subduction beneath the Slavecraton Canada Earth and Planetary Science Letters 254 55^68

Schulze D J amp Helmstaedt H (1988) Coesite-sanidine eclogitesfrom kimberlite products of mantle fractionation or subductionJournal of Geology 96 435^443

Smithies R H Champion D C amp Sun S S (2004) The case forArchaean boninites Contributions to Mineralogy and Petrology 147705^721

Smyth J R Caporuscio F A amp McCormick T C (1989) Mantleeclogites evidence of igneous fractionation in the mantle Earth andPlanetary Science Letters 93 133^141

Sobolev NV amp Lavrentrsquoyev Y G (1971) Isomorphic sodium admix-ture in garnets formed at high pressure Contributions to Mineralogy

and Petrology 31 1^12Stalder R Foley S F Brey G P amp Horn I (1998) Mineral aqueous

fluid partitioning of trace elements at 900^12008C and 30^57GPa New experimental data for garnet clinopyroxene and rutileand implications for mantle metasomatism Geochimica et

Cosmochimica Acta 62 1781^1801Sun S-S amp McDonough W F (1989) Chemical and isotopic sys-

tematics of oceanic basalts implications for mantle compositionsand processes In Saunders A D amp Norris M J (eds)Magmatism in the Ocean Basins Geological Society London Special

Publications 42 313^345Taura H Yurimoto H Kurita K amp Sueno S (1998) Pressure

dependence on partition coefficients for trace elements between oli-vine and the coexisting melts Physics and Chemistry of Minerals 25469^484

Taylor L A amp Neal C R (1989) Eclogites with oceanic crustal andmantle signatures from the Bellsbank kimberlite South AfricaPart I Mineralogy petrography and whole rock chemistryJournal of Geology 97 551^567

van Achterbergh E Ryan C G amp Griffin W L (1999) Glitter online intensity reduction for the laser ablation inductively coupledplasma spectrometry Abstract 9th V M GoldschmidtConference LPI Boston MA 905 pp Abstract no 7215

van Breemen O DavisW J amp King J E (1992) Temporal distribu-tion of granitoid plutonic rocks in the Archean Slave ProvinceNorthwest Canadian Shield Canadian Journal of Earth Sciences 292186^2199

Vervoort J D Patchett P J Blichert-Toft J amp Albare de F (1999)Relationship between Lu^Hf and Sm^Nd isotopic systems in theglobal sedimentary system Earth and Planetary Science Letters 16879^99

Viljoen K S Smith C B amp Sharp Z D (1996) Stable and radio-genic isotope study of eclogite xenoliths from the Orapa kimberliteBotswana Chemical Geology 131 235^255

Walter M J (1998) Melting of garnet peridotite and theorigin of komatiite and depleted lithosphere Journal of Petrology 3929^60

Wasserburg G J Jacobsen S B DePaolo D J McCulloch M Tamp Wen T (1981) Precise determination of SmNd ratios Sm andNd isotopic abundances in standard solutions Geochimica et

Cosmochimica Acta 45 2311^2323Westerlund K J Shirey S B Richardson S H Carlson R W

Gurney J J amp Harris JW (2006) A subduction wedge origin forPaleoarchean peridotitic diamonds and harzburgites from thePanda kimberlite Slave craton evidence from Re^Os isotope sys-tematics Contributions to Mineralogy and Petrology 152 275^294

AULBACH et al ECLOGITE XENOLITHS SLAVE CRATON

1873

Dow

nloaded from httpsacadem

icoupcompetrologyarticle481018431512585 by guest on 01 January 2022

Page 23: Origins of Xenolithic Eclogites and Pyroxenites from the Central
Page 24: Origins of Xenolithic Eclogites and Pyroxenites from the Central
Page 25: Origins of Xenolithic Eclogites and Pyroxenites from the Central
Page 26: Origins of Xenolithic Eclogites and Pyroxenites from the Central
Page 27: Origins of Xenolithic Eclogites and Pyroxenites from the Central
Page 28: Origins of Xenolithic Eclogites and Pyroxenites from the Central
Page 29: Origins of Xenolithic Eclogites and Pyroxenites from the Central
Page 30: Origins of Xenolithic Eclogites and Pyroxenites from the Central
Page 31: Origins of Xenolithic Eclogites and Pyroxenites from the Central