Contents lists available at ScienceDirect

Lithos

j ourna l homepage wwwe lsev ie r com locate l i thos

Lithos 189 (2014) 2ndash15

Review paper

The world turns over HadeanndashArchean crustndashmantle evolution

WL Griffin a EA Belousova a C ONeill a Suzanne Y OReilly a V Malkovets ab NJ Pearson aS Spetsius ac SA Wilde d

a ARC Centre of Excellence for Core to Crust Fluid Systems (CCFS) and GEMOC Dept Earth and Planetary Sciences Macquarie University NSW 2109 Australiab VS Sobolev Institute of Geology and Mineralogy Siberian Branch Russian Academy of Sciences Novosibirsk 630090 Russiac Scientific Investigation Geology Enterprise ALROSA Co Ltd Mirny Russiad ARC Centre of Excellence for Core to Crust Fluid Systems Dept of Applied Geology Curtin University GPO Box U1987 Perth 6845 WA Australia

Corresponding authorE-mail address billgriffinmqeduau (WL Griffin)

0024-4937$ ndash see front matter copy 2013 Elsevier BV All rihttpdxdoiorg101016jlithos201308018

a b s t r a c t

a r t i c l e i n f o

Article historyReceived 13 April 2013Accepted 19 August 2013Available online 3 September 2013

KeywordsArcheanHadeanCrustndashmantle evolutionLithospheric mantle Hadean zirconsMantle Os-isotopes

We integrate an updatedworldwide compilation of UPb Hf-isotope and trace-element data on zircon and RendashOsmodel ages on sulfides and alloys inmantle-derived rocks and xenocrysts to examine patterns of crustal evolutionand crustndashmantle interaction from 45 Ga to 24 Ga ago The data suggest that during the period from 45 Ga to ca34 Ga Earths crust was essentially stagnant and dominantly mafic in composition Zircon crystallized mainlyfrom intermediatemelts probably generated both bymagmatic differentiation and by impact melting This quies-cent state was broken by pulses of juvenile magmatic activity at ca 42 Ga 38 Ga and 33ndash34 Ga which mayrepresent mantle overturns or plume episodes Between these pulses there is evidence of reworking and resettingof UndashPb ages (by impact) but no further generation of new juvenile crust There is no evidence of plate-tectonic ac-tivity as described for the Phanerozoic Earth before ca 34 Ga and previousmodelling studies indicate that the earlyEarth may have been characterised by an episodic-overturn or even stagnant-lid regime New thermodynamicmodelling confirms that an initially hot Earth could have a stagnant lid for ca 300 Ma and then would experiencea series of massive overturns at intervals on the order of 150 Ma until the end of the EoArchean The subcontinentallithosphericmantle (SCLM) sampledonEarth todaydidnot exist before ca 35 Ga A lull in crustal production around30 Ga coincides with the rapid buildup of a highly depleted buoyant SCLM which peaked around 27ndash28 Ga thispattern is consistent with one or more major mantle overturns The generation of continental crust peaked later intwo main pulses at ca 275 Ga and 25 Ga the latter episode was larger and had a greater juvenile component TheageHf-isotope patterns of the crust generated from 30 to 24 Ga are similar to those in the internal orogens of theGondwana supercontinent and imply the existence of plate tectonics related to the assembly of the Kenorland (ca25 Ga) supercontinent There is a clear link in these data between the generation of the SCLM and the emergence ofmodern plate tectonics we consider this link to be causal as well as temporal The production of both crust andSCLM declined toward a marked low point by ca 24 Ga The data naturally divide the Archean into three periodsPaleoArchean (40ndash36 Ga) MesoArchean (36ndash30 Ga) and NeoArchean (30ndash24 Ga) we suggest that this schemecould usefully replace the current four-fold division of the Archean

copy 2013 Elsevier BV All rights reserved

Contents

1 Introduction 32 Methods and databases 3

21 Zircon data 322 RendashOs data 323 Numerical modelling 3

3 Geochemical data 431 Zircon data UndashPb ages Hf isotopes trace elements 4

311 Hadean zircons 5312 Early- to mid-Archean zircons 5313 NeoArchean zircons 5

ghts reserved

3WL Griffin et al Lithos 189 (2014) 2ndash15

32 Os-isotope data 5321 Sulfides and platinum group minerals 5322 Whole-rock analyses of peridotites 8

4 Discussion 941 Numerical modelling of the early Earth 942 Evolution of the crustndashmantle system 10

421 The Hadean Hf-isotope recordmdash magmatic metamorphic or both 10422 Composition of the earliest crust 12423 Early Archean crustal development 12424 Late Archean mdash the formation of the SCLM (and plate tectonics) 13

43 Revising the geologic time scale 14Acknowledgements 14References 14

1 Introduction

The state of the crustndashmantle systemduring theHadean period is stillunclear the only recognised relics of the Hadean crust are a few zirconswith ages mostly 41ndash43 Ga recovered from much younger rocks(Fig 1) However even these traces have been the subject of severaldifferent tectonic interpretations ranging from the modern to the cata-clysmic Large populations of zircons in younger sediments or datablerocks only begin to appear well into EoArchean time from about37 Ga However the oldest dated rocks from the subcontinental litho-spheric mantle (SCLM) are only about 35 Ga old (Griffin et al 2009)and it is not clear on what type of substrate these earliest crustal rocksmight have rested

The conventional subdivision of the Archean into EoArcheanPaleoArchean MesoArchean and NeoArchean is shown in Fig 1 Thetransition from the Hadean to the EoArchean is conventionally set at40 Ga Ideally such boundaries should be defined by a geologicallyimportant (or at least definable) event in this case none have beenidentified although dynamic modelling (see ONeill et al 2007 inpress) offers some insights

In this report we examine the zircon record (ages Hf and O isotopestrace elements) using a database of N6500 analyses with ages N20 Gaand theOs-isotope data derived fromboth the in situ analysis of sulfidesandOsndashIr phases in peridotite xenoliths andwhole-rock RendashOs analysisof such xenoliths We couple our observationswith dynamic modellingto propose mechanisms for the evolution of the crustndashmantle systemthrough the Hadean and Archean periods Finally we also suggest anew subdivision of the earliest part of the geological time scale basedon currently recognisable patterns in the data inferred to mark signifi-cant tectonic events from 45 Ga to 24 Ga

2 Methods and databases

21 Zircon data

Data on zircon ages and Hf-isotope ratios are taken from the databasedescribed by Belousova et al (2010) supplemented by more recentlypublished analyses (Geng et al 2012 Naeraa et al 2012) and ourunpublished data This database (n = 6699) is built largely on detrital-zircon suites but also includes many zircons separated from igneousrocks All analytical data were recalculated using the same parametersFor the calculation of εHf we have used the chondritic values of Bouvieret al (2008) and the decay constant (1865 times 10minus11 yrminus1) of Schereret al (2001) The Depleted Mantle (DM) curve (Fig 1) is that definedby Griffin et al (2000 176Lu177Hf = 00384)

To examine the distribution of ldquojuvenilerdquo Hf-isotope ratios we havetaken all data within a band corresponding to plusmn075 around theDepleted Mantle curve (Belousova et al 2010 Fig 1) Such ldquojuvenilerdquogeochemical signatures indicate that the source rock for the zircon wasderived directly from the convecting mantle or had only a very short

crustal residence time after formation from mantle-derived magmaThe described procedure reduces the exaggeration of the younger peaksthat would be produced by the Expanding difference through time be-tween the Chondritic Earth (εHf = 0) line and the Depleted Mantle(Fig 1) where trace-element data are available on single zircon grainswe have classified them in terms of rock type using a modified versionof the discrimination scheme outlined by Belousova et al (2002)Where zircons are extracted from igneous rocks the composition ofthose rocks is recorded in the database

Oxygen-isotope data can likeHf-isotope data allow an evaluation ofthe juvenile vs recycled nature of the source region of the zircons hostmagma In nearly all cases the two isotopic systems are in agreementO-isotope data are available for relatively few zircons in the datasetthese are discussed where relevant Analyses of ThU ratios have beenused to exclude clearlymetamorphic zircons and thus the ages reportedhere are considered to reflect the timing of igneous crystallisation exceptas discussed below

22 RendashOs data

RendashOs data on sulfide andOsndashIr phases have been obtained by in situLA-MC-ICPMS analysis essentially as described by Pearson et al(2002) and Griffin et al (2004a 2004b) The database includes pub-lished analyses from kimberlite-borne xenoliths in S Africa Siberiaand the Slave Craton (Aulbach et al 2004 2009 Griffin et al 20022004a 2004b 2011 2012 Spetsius et al 2002) as well as newdata on sulfides included in olivine xenocrysts from the Udachnayakimberlite (Yakutia) and sulfides included in chrome-pyrope garnetxenocrysts from the Mir kimberlite (Yakutia) We have added ourunpublished analyses of sulfides in xenoliths from other localities inN America and Asia An overview of whole-rock RendashOs analyses ofmantle-derived peridotite xenoliths is given by Carlson et al (2005)

23 Numerical modelling

To assess the potential behaviour of the plate-mantle system underevolvingHadeanmantle conditionswe employ the visco-plasticmantleconvection codeUnderworld (Moresi et al 2007)We solve the standardconvection equations for conservation of mass momentum and energyunder the Boussinesq approximation with varying Rayleigh number(ie varying basal temperatures) and internal heating The mantle itselfis a modelled as a viscoplastic fluid with an extremely temperature-dependent viscosity that varies using a Frank-Kamenetski approxima-tion from 1 at the base to 3 times 104 at the cold upper boundary (seeONeill et al in press for details) Deformation in the near-surface isaccommodated by plastic yield using a Byerlee criterion for yield stress(scaled so the cohesion is 1 times 105 and the depth-coefficient is 1 times 107

for a Rayleigh number of 1 times 107 see Moresi and Solomatov (1998)for details) We do not consider phase transitions or depth-dependentproperties in these models All our units are non-dimensionalised

(a)

(b)

(c)

Fig 1 εHf vs age for zircons worldwide divided by region The database (n = 6699) is attached as Appendix A1 it builds on the data of Belousova et al (2010) updated with N2000 analysespublished andor produced in our laboratories after 2009 The conventional subdivision of theHadean andArchean is shown along the X axis (a) Zircon age distribution shownas a histogramand a cumulative-probability curve (b) Distribution of Hf-isotope data The ldquojuvenile bandrdquo defined asplusmn075 of the εHf of the DepletedMantle line at any time is shown as a grey envelopeEvolution lines from45 Ga are shown for reservoirs corresponding to typicalmafic rocks (176Lu177Hf = 0024) the average continental crust (176Lu177Hf = 0015) and zircons approximatedby 176Lu177Hf = 0 (c) shows these evolution lines and a simplified explanation (see text)

4 WL Griffin et al Lithos 189 (2014) 2ndash15

before being incorporated into the model in the manner of Moresi andSolomatov (1998)

The initial condition for the simulation is from an equilibrated modelwith an internal heat generation of Q = 8 (non-dimensional present dayQ ~ 1) and a basal temperature of Tbase = 12 (present day Tbase ~ 1)The model initially is in a hot stagnant-lid regime as a result of theseproperties which are meant to represent the hot Hadean mantle withcontributions from high radioactive heat production a hot core and sig-nificant primordial heat We then allow the internal heat generation andbasal temperatures to decay through time While short-lived radioactiveisotopes like 26Al may contribute to the thermal state of the initialmodel theydecay too rapidly to be incorporated into long-termevolutionmodels and we solely consider the decay of 40K 238U 235U and 232Th We

also adopt a simple core-evolution model based on the results of Nimmoet al (2004) If we assume present-day temperatures at the core-mantleboundary are around Tbase ~ 1 then according to Nimmo et al (2004) atthe beginning of the Hadean they would have been around Tbase ~ 12The temperature evolution of the core is pseudo-linear in these modelsand we project this evolution forward along this trend

3 Geochemical data

31 Zircon data UndashPb ages Hf isotopes trace elements

The full zircon dataset (n = 6699) coded by geographical region isshown in Fig 1 and the numbers of zircons with εHf N 0 in each time

Fig 2 Cumulate-probability curve and age histogram for all zirconswith εHf falling within075 of theDepletedMantle curve in Fig 1 Inset uses an expanded scale to show detail inthe oldest datasets

5WL Griffin et al Lithos 189 (2014) 2ndash15

slice are shown in Fig 2 The cumulative-probability plot of Fig 2 showsa small peak around 42 Ga a larger one at ca 38ndash36 Ga another at ca34ndash31 Ga and then a buildup to a major peak at 275 Ga which isclosely followed by a final sharp larger peak at 25 Ga Wewill describethe nature of each ldquoeventrdquo as revealed by the isotopic data in turnspeculations on mechanisms will be deferred to the Discussion

311 Hadean zirconsThe Hadean zircon record is dominated by material from Western

Australia but has recently been supplemented by a few data from Asiaand N America (Fig 1) The most striking aspect of this record is therelative paucity of juvenile signatures the peak at 42ndash41 Ga in Fig 2represents ca 5 of the zircons in that age band and none of the oldestzircons (445ndash425 Ga) has significantly suprachondritic Hf-isotopesThis may suggest that the Depleted Mantle model is not relevant tothe earliest Earth prior to ca 425 Ga but the data are too limited forfirm conclusions The very low εHf of many of the oldest zircons mightsuggest the reworking of older material but this would require thatthe older reservoir had evolved with LuHf = 0 since 455 Ga whichseems unlikely Alternative explanations are that these zircons crystal-lized from magmas derived from a non-chondritic reservoir or thattheir ages have been reset by later metamorphic events which left theHf-isotope signatures unchanged (see Discussion below)

Many of the younger Hadean zircons (Fig 3) lie in a band extendingdownward from the DM line This trend which extends down to ages asyoung as 39 Ga could be interpreted as the result of reworking of the juve-nilematerial represented by the 42-Ga ldquopeakrdquo Thiswould require that the42 Gamaterial wasmore felsic (176Lu177Hf asymp 001) than the present-daymean continental crust (176Lu177Hf asymp 0015 Griffin et al 2000) Grainsfalling in the lowerpart of the bandwould require evenmore felsic sources

Cavosie et al (2005) give trace-element data for 40 zircon grains (38ndash44 Ga) from Jack Hills 75 classify as derived from intermediate rocks(b65 SiO2) Sevendated grains (allwith agesN41 Ga) classify as derivedfrom mafic rocks and three grains all with ages b41 Ga classify as de-rived from felsic granitoids (N70 SiO2) Inclusions of quartz feldsparand muscovite in the zircons have been used to suggest their derivationfrom intermediate to felsic rocks although later studies (Rasmussenet al 2011) suggest thatmany such inclusions aremetamorphic in originA few zircons fromN America andAsiawith ages near 40 Ga come fromfelsic rocks (Geng et al 2012 Pietranik et al 2008)

312 Early- to mid-Archean zirconsThe first major peak in juvenile input occurs at 38 Ga followed by a

series of minor peaks to ca 35 Ga (Fig 2) this is defined by data fromAsia and N America (including Greenland) with a few from Africa

(Fig 3) While most of the zircons with ages 385ndash375 Ga have εHf N0consistent with derivation from a Depleted Mantle source very few ofthe younger ones in this age band have εHf N0 (Fig 3) Most of thezircons with ages 37ndash34 Ga have low εHf lying in a band with a slopecorresponding to 176Lu177Hf asymp 001 this suggests that their hostmagmas could be derived by reworking of the 38 Ga juvenile materialNaeraa et al (2012) have argued that the data for a set of zircons fromW Greenland fit a shallower trend consistent with derivation from aldquoTTG crustrdquo (187Lu188Hf = 0024) corresponding to the 38 Ga juvenilematerial from the same area On a larger scale the N America dataset inFig 3a suggests a more felsic crust or at least remelting of the morefelsic portions Data from other continents scatter more widely butthe mean data from each continent are consistent with those fromN America suggesting a global process

The next evidence of juvenile magmatic activity is the peak at335ndash315 Ga (Fig 2) This is represented by material mainly fromAsia N America and S America (Fig 4a) Unlike the earlier peaksmany of the zircons with juvenile Hf-isotope signatures appear to be de-rived frommafic andmetamorphic rocks (Fig 4b)Many zircons from thistime range fall along a trend extending toprogressively lower εHf possiblycorresponding to the progressive reworking of reservoirs 33ndash335 Ga oldwith 176Lu177Hf asymp 001 Interestingly most of the zircons in the lowerpart of this band are detrital (from metasediments no trace elements)whereas most of those with identified host magmas lie in the upperpart of the band The proportion of zircons from granitoid rocks appearsto increase further in this time slice

In this age range there are few of the older zircons with εHf near theLuHf = 0 line However there is a significant populationwith very lowεHf which may be derived from remelting of the ca 38 Ga population

313 NeoArchean zirconsThe NeoArchean zircon record is marked by two major sharply

defined peaks at 275 Ga and 255 Ga (Fig 2) The buildup to the olderpeak is marked by minor shoulders at 29 Ga and 28 Ga only the firstmay be significant The older peak is dominated by material from Nand S America Asia and Europe the younger peak is mainly made upof zircons from Asia and Australia However nearly all regions have con-tributed material to both age peaks The burst of magmatic activity at255 Gamarks the end of the Archean as traditionally defined the overallabundance of zircons in the database decreases markedly from ca245 Ga and zircons with juvenile Hf isotopes disappear almost entirelymarking the start of a worldwide period of quiescence (low mantle-derived magmatic activity) that lasted for about 300 Ma (Fig 1)

The data also appear to reflect a change in themechanisms involved inthe large-scaleNeo-Archeanmagmatism thewhole rangeof εHf increasesin the last 05 Gaof theArchean (Fig 1) The agepeaks at 275 and255 Gaare also peaks of juvenile input to the crust (Fig 2) However while thejuvenile inputs are clearly important the record also suggests a more in-tense reworking of the older crust than is seen in earlier episodes asreflected in the large proportion of zirconswith εHf b0 (Fig 5) The lowestof these are consistent with derivation of their host magmas from ldquomaficcrustrdquo up to 45 Ga in age or younger but more felsic crust (Figs 1 5)

Zircons from granitoid rocks feature prominently in both of themajor NeoArchean magmatic episodes but especially in the youngerone those in the younger peak generally have higher εHf

32 Os-isotope data

321 Sulfides and platinum group mineralsThe Os budget of mantle-derived peridotites (whether sampled as

massifs or xenoliths) resides in minor phases these are typically base-metal sulfides but in some cases especially where the peridotitescarry lenses of chromitite (eg Gonzalez-Jimenez et al 2012a Shiet al 2007) Platinum Group Minerals (PGMs) such as OsndashIr alloysand Platinum Group Element (PGE) sulfides may be abundant (see re-view by Gonzalez-Jimenez et al 2013) The samples included in the

Fig 3 εHf vs age for all zircons in the age range 45ndash35 Ga (a) by region (b) by rock type either estimated from trace-element data or taken from the sample The dark blue linein (b) shows the evolution of a mafic reservoirs with 176Lu177Hf = 0024 (see text for explanation) the solid green line shows the evolution of a reservoir 456 Ga old with LuHf = 0(ie ancient zircons) The ldquounknownrdquo category in (b) corresponds to suites of detrital zircons with no trace-element data (c) as from Fig 1

6 WL Griffin et al Lithos 189 (2014) 2ndash15

following discussion have been screened all have very low ReOs andtheir Os-isotope values therefore are taken as recording the composi-tion of Os in the melts or fluids that deposited them The Re-depletionmodel ages (TRD) calculated from individual mineral analyses can givean indication of the timing of melt depletion in their host rocks or thetiming of metasomatic activity derived from the convecting mantle(Gonzalez-Jimenez et al 2012b Griffin et al 2004a 2004b Shi et al

2007) Our previous studies have shown that most peridotite xenolithsandmanymassif peridotites contain Os-bearing phases with a range ofTRD reflecting a complex history of fluid-related processes (Griffin et al2004a 2004bMarchesi et al 2010) These include bothmelt extractionand multiple overprinting by metasomatic fluids of different origins(OReilly and Griffin 2012) While model-age peaks commonly matchwith events in the overlying crust these fluid-rock interactions

Fig 4 εHf vs age for all zircons in the age range 35ndash30 Ga (a) by region (b) by rock type either estimated from trace-element data or taken from the sample The light red dashed lines in(b) show the evolution of reservoirswith 176Lu177Hf = 001 (see text for explanation) The ldquounknownrdquo category if (b) corresponds to suites of detrital zirconswith no trace-element data

7WL Griffin et al Lithos 189 (2014) 2ndash15

undoubtedly lead to grains of mixed origins withmodel ages that do notcorrespond to any real event

The full xenolith dataset is shown in Fig 6 as noted above it is dom-inated by samples from South Africa Siberia and the Slave Craton inCanada The most striking observation compared with the zircon data

is the absence of Hadean or Eo-Archean model ages we have no evi-dence in the sulfide data for the existence of any SCLM for the first billionyears of Earths history The earliest significant peak in TRD appears at34ndash32 Ga corresponding to the second major peak in TDM from thezircon data (Figs 2 6) This is part of a buildup to the major peak in

Fig 5 εHf vs age for all zircons in the age range 30ndash23 Ga (a) by region (b) by rock type either estimated from trace-element data or taken from the sample The red lines show theevolution of reservoirs with 176Lu177Hf = 001 (see text for explanation) The ldquounknownrdquo category if (b) corresponds to suites of detrital zircons with no trace-element data

8 WL Griffin et al Lithos 189 (2014) 2ndash15

sulfide TRD at ca 28 Ga this peakmay bemildly bimodal with sub-peaksat 285 Ga and 27 Ga coinciding with the Neo-Archean explosion ofmagmatic activity shown by the zircon data This peak drops off sharplytoward younger ages mirroring the ldquoshutdownrdquo in magmatic activityfrom ca 24 Ga as shown by the zircon data

322 Whole-rock analyses of peridotitesThe TRD of whole-rock samples that have experienced no metasoma-

tism following their originalmelt depletionmay record the timing of thatdepletion event The Os-isotope compositions of any samples that havebeen metasomatised by (Os-bearing) melts or fluids must necessarily

0 1 2 3 4

Age Ga

Rel

ativ

e p

rob

abili

ty

TRD ages of low-ReOs sulfide and alloy grains in mantle-derived xenoliths

(n=370)

Whole-rock TRD

Fig 6 RendashOs data Red line shows relative-probability distribution of TRD model ages forsulfide and alloy grains from mantle-derived xenoliths peridotite massifs and diamondsGrey histogram (after Carlson et al 2005) shows distribution of whole-rock TRD agesfrom mantle-derived peridotite xenoliths

9WL Griffin et al Lithos 189 (2014) 2ndash15

represent mixtures and their TRD will not be meaningful in terms ofdating an event Since all of the samples discussed here come fromthe subcontinental lithospheric mantle (SCLM) they provide informa-tion only on the timing of formation andmodification of the SCLM ratherthan the convecting mantle A review by Carlson et al (2005) on thegeochronology of the SCLM shows only three whole-rock xenoliths (allfrom South Africa) that give TRD ages around 36 Ga (Fig 6) Thereis a sharp increase from 3 Ga leading up to a pronounced peak at25ndash275 Ga mirroring the sulfide data As with the sulfidealloydata there is no evidence for the existence of significant amounts ofSCLM prior to 35 Ga A long tail of TRD ages down to ca 1 Ga probablyreflects the abundance of younger sulfides in repeatedlymetasomatisedxenoliths (cf Griffin et al 2004a 2004b)

4 Discussion

The data presented above suggest that juvenile addition to thecontinental crust in the HadeanndashEoArchean as preserved in the zirconrecord occurred primarily as pulses at ca 42 Ga 38 Ga and 33ndash34 GaBetween these pulses of activity zircon appears to have crystallisedprimarily from intermediate melts The zircon data suggest the surfacewas largely stagnant and tectonically quiescent during the interval from45 Ga to 34 Ga interspersed by short bursts of surface tectonicndashmagmatic activity Is this pattern simply an artefact of limited samplingor preservation or could it reflect the real behaviour of the early EarthBefore proceeding with a discussion of the data it is useful to considersome aspects of Earths early evolution as revealed by dynamicmodelling

41 Numerical modelling of the early Earth

Previousmodelling byONeill et al (2007) suggested amechanismbywhich plate activity itself could have been episodic in the Precambrianwithout appealing to internal phase transitions or other mechanismsThis model was based on the idea that the high internal temperaturesof the early Earth would have resulted in lower mantle viscositiesdecreasing the coupling between the mantle and the plates The simula-tions showed first a transition into an episodic overturn mode ndash longperiods of tectonic quiescence interspersed by rapid intervals of subduc-tion spreading and tectonic activity ndash and eventually another transitioninto a stagnantmodewhere the internal stresses are too low to generateany surface deformation These initial models have been largely borneout by later work including the conceptual approach of Silver and

Behn (2008) and the numerical models of van Hunen and Moyen(2012) and Gerya (2012) which focus on different aspects of themech-anism but reach the same conclusion They are largely consistent withmore recent geological (Condie and ONeill 2010 Condie et al 2009)and paleomagnetic constraints (Piper 2013) Further complexities suchas the dehydration of the depleted residuum after melt extraction havebeen suggested to assist (Korenaga 2011) or hinder (Davies 2006) Ha-dean tectonics and given the lack of constraints we do not consider itfurther

Here we expand these earlier models to consider the temporal evolu-tion of the Hadean Earth from its earliest thermal state to the end of theEoArchean including the effects of waning radioactive decay loss of pri-mordial heat and cooling of the coreWe adopt the approach outlined byONeill et al (in revision) expanding those models to high-aspect-ratiosimulations to examine how the Earth system would adjust to coolingthrough time

A representative model of Hadean to EoArchean evolution is shownin Fig 7 The four time slices show an evolution from a hot initial state(a) characterised by a hot thin stagnant lid and vigorous convectionbeneath it through an overturn event where the initial lid is largelyrecycled and a large proportion of the initial heat is expelled (b c) toanother stagnant lid (d) that develops and stops the surface activityThis timeline extends from 455 Ga to roughly the end of Eo-Archeantime at 35ndash36 Ga

An important component to thesemodels is the interaction betweenthe evolving basal temperature and waning heat production throughtime High basal temperatures result in greater buoyancy forces mdasheffectively a higher basal Rayleigh number This enhances the tendencytowards lid mobilisation However greater heat production within themantle also tends to promote stagnation of the lid The balance betweenthese two forces as a planet evolves largely controls its tectonic evolu-tion We have adopted a simple linear core-evolution model and anexponential decay of heat production and the tectonic scenario inFig 7 is insensitive to minor perturbations in these factors (ONeillet al in revision)

A second critical factor is the initial condition ONeill et al (inrevision) noted that the initial conditions in evolutionarymantlemodelscan fundamentally affect the apparent evolution of a model planet Theyshowed that for the same evolving core temperatures and heat produc-tion curves a planet may follow widely divergent tectonic pathsdepending on the amount of initial (primordial) heating Models with aldquohotrdquo initial condition (Q ~ 8 similar to Fig 7) show transitions fromstagnant to episodic overturn to a mobile-lid regime and eventually toa cold stagnant-lid regimeModels starting from a ldquocoldrdquo initial condition(Q b 4 roughly in equilibriumwith radioactive heat productionwithoutany significant primordial heat) may begin in a plate tectonic regimeand then never leave it until the internal temperatures decay to thepoint that the lid becomes too thick to mobilise and the model entersa cold stagnant-lidmode However it is highly probable that heat gener-ation during Earths accretion was significant mdash the sum of accretionalimpact energy gravitational energy from core segregation and heatfrom the decay of short-lived isotopes such as 60Fe and 26Al was enoughto raise the temperature of the entire Earth by thousands of degrees(ONeill et al in revision) Therefore the initial conditions shown inFig 7a are probably the most relevant and it is likely that the earliestEarth was largely stagnant

The curve in Fig 8b illustrates the evolution of the Nusselt number(convective heat flux) through time for the simulation shown inFig 7 The initial mode of heat loss is stagnant-lid ndash characterised by in-efficient heat loss through the conductive lid (Fig 7a cf Debaille et al2013) It is likely that voluminousmaficndashultramafic volcanismwould beassociated with such a regime with a thin hot stagnant lid Eventuallyongoing convergence triggers a convective instability and thefirst over-turn event happens (Fig 7b) This is associated with an enormous spikein the heat flux (Fig 8b) as much of the trapped primordial heat andgenerated radioactive heat is dumped during the active-lid episode

Fig 7 Simulation of the evolution of a mantle convection model from a hot initial condition in a stagnant-lid regime (a) through an overturn event where the majority of the crust isrecycled (b and c) and into a subsequent stagnant lid again (d) The colours in the plots reflect the temperature field from an invariant surface temperature of 0 (290 K) to a basalcore-mantle boundary temperature of 12 which varies as outlined in the text Black arrows denote velocities and their length velocity magnitude

10 WL Griffin et al Lithos 189 (2014) 2ndash15

The episode is shortmdash on the order of 30 Ma though the activity probablyis localised in different areas at different times and the surface velocitiesare extremely high It seems likely that most of the proto-lid would berecycled into the convecting mantle

After this initial recycling of the proto-lid the new lithosphere isthin hot and comparatively buoyant and subduction ceases (Fig 7d)The lid stagnates convection re-establishes itself beneath the immobilelithosphere and heat flux subsides to low (stagnant-lid) levels As theconvective planform evolves convection thins some portions of thelithosphere and thickens others until a critical imbalance is onceagain attained and a subduction event ensues The second event ismarginally more subdued as a large fraction of the initial heat waslost in the first episode However extremely high heat fluxes areseen throughout the 900 million years of evolution covered in thismodel

It should be emphasised that thismodel is not attempting to replicateEarths precise evolution but rather to provide some insights into theplausible range of dynamics of an Earth-like system under a thermalevolution similar to Earths during its earliest history However it isinteresting to note that the time lag between the initialization of themodel and its first overturn is ~300 Myr and the overturn events inthis example are spaced at ~150 Myr intervals This is of the samemagnitude of the recurrence rate observed in the zircon data (seebelow) but it should be noted this spacing is very sensitive to rheologyand mantle structure which are not explored in detail here

42 Evolution of the crustndashmantle system

The distribution of zirconUndashPb ages from the ArcheanndashHadean periodis reasonably well-known As shown in Fig 1 the ca 6700 availableanalyses worldwide (Table A1) define a small peak in the late Hadean(probably over-represented due to its inherent interest) anotherbroad low peak in the EoArchean and a third in the PaleoArchean be-fore a steady climb from ca 30 Ga to the major peaks at 27 and25 Ga When compared with the much smaller dataset of zircons

with juvenile Hf-isotope compositions (Fig 2) there are two striking as-pects the low proportion of juvenile input (45 of all analyses 20have εHf N 0) prior to 3 Ga and the dramatic rise in that proportionduring the youngest Archean activity These two observations in them-selves underline the stark differences between HadeanArchean andmodern crustndashmantle relationships and the evolution of those processesfrom the Hadean through to the end of the Archean However there areseveral questions that must be asked about the nature of the zirconrecord and the sort of events that it may define

421 The Hadean Hf-isotope record mdash magmatic metamorphic or bothThere are essentially no zircons with juvenile Hf-isotope signatures

before the small ldquobloomrdquo at 42 Ga Older zircons have very low εHf thiswould commonly be interpreted as suggesting that they were generatedby remelting of much older material (back to 45 Ga) However manycan only be modelled on the basis of a 454 Ga source with LuHf = 0The small peak of juvenile input at 42 Ga is followed by a large groupof zircons with εHf b 0 extending down to ca 40 Ga (Fig 3) Againsome can be modelled as produced by reworking of mafic to felsicmaterial produced during the 42 Ga ldquoeventrdquo but some of these42ndash40 Ga zircons also would require a source that retained a very lowLuHf since at least 42 Ga

This appears unlikely the most obvious alternative is that the UndashPbages ofmanyof these ancient zirconshavebeen resetwithoutmeasurabledisturbance of the Hf-isotope system leading to artificially low εHf Beyeret al (2012) demonstrated that zircons from the Archean (N3 Ga)Almklovdalen peridotite in western Norway give a range of UndashPb agesstretching from396 to 2760 Ma but all have the sameHf-isotope compo-sition corresponding to a TDM of 32 Ga (similar to whole-rock RendashOsmodel ages) the lowest εHf is minus49 They interpreted this pattern asreflecting metamorphic resetting (probably through several episodes) ofthe primary 32 Ga ages without affecting the Hf-isotope composition

Nemchin et al (2006) have argued thatmany of the Jack Hills zirconsno longer retain truemagmatic compositions either elementally or isoto-pically due to weathering metamorphism or other causes Rasmussen

Fig 8Upper) The evolution of core temperatures (dashed line) and heat production Q forthe model shown in Fig 7 Both heat production and core temperatures wane throughtime according to the evolution shown affecting the thermal state of the evolving simula-tion Time is non-dimensional in these simulations andwe keep our results in that formatto facilitate comparison between codesmdash here a time interval of 01 is equivalent to ~1Gaof real evolution lower) Evolution of Nusselt number (convective heat flow) for themodel shown in Fig 7 Four overturn events are apparent where the lid was rapidlyrecycled into the mantle resulting in the creation of new lithosphere and high resultantheat fluxes

11WL Griffin et al Lithos 189 (2014) 2ndash15

et al (2010) used xenotime-monazite geochronology and Wilde (2010)used zircon geochronology to document several episodes of depositionvolcanism and low-grade metamorphism through the Jack Hills beltRasmussen et al (2011) have shown that many of the silicate(quartz + muscovite) inclusions in the ancient Jack Hills zircons aremetamorphic in origin and probably have replaced primary inclusionsof apatite and feldspar Some inclusions with such metamorphic assem-blages are connected to the grain surface by microcracks but others arenot suggesting small-scale permeation of the zircons by fluids Monaziteand xenotime in the zircons give ages of 26 or 08 Ga and temperaturesof 350ndash490 degC If fluids have been able to penetrate the zircon grains tothis extent it is unlikely that the UndashPb systemswould remain unaffectedKusiak et al (2013) have recently demonstrated that Hadean ages aswell as anomalously young ages in some Antarctic zircons are artefactsof the redistribution of Pb within single grains probably in response tometamorphic heating and the activity of fluids

An obvious question is whether this process also could modify theO-isotope composition of the zircons Ushikubo et al (2008) documentedextremely high Li contents and highly fractionated Li isotopes in theHadean zircons from Jack Hills They identified this as the signature ofextreme weathering but then argued for remelting of highly weatheredmaterial to produce the parental magmas of the zircons However it

must be questioned whether in fact the high Li represents a secondaryeffect on the zircons themselves as documented by eg Gao et al(2011)

Wilde et al (2001) noted correlations between LREE enrichment in-creased δ18O and lower UndashPb age within a single grain although thiswas interpreted in terms of evolvingmagma composition it is also con-sistent with the infiltration of pre-existing zircon by crustal fluidsCavosie et al (2005) extended the inverse age-δ18O correlation to thewhole population of grains with ages of 44ndash42 Ga The δ18O values ofgrains that retain zoning progressively depart from the mantle valueas apparent age decreases and even larger dispersions of δ18O occurin grains with disturbed CL patterns However Cavosie et al (2006)still argued that most of the Jack Hills zircons have magmatic REEpatterns with only a small proportion showing LREE-enrichmentAlthough these were also grainswith disturbed age patterns the authorsappeared reluctant to consider them as altered by fluid processes

Valley et al (2002 2006) and Pietranik et al (2008) summarisedpublished O-isotope analyses of the Jack Hills zircons A comparison ofthe data distributions of O-isotope and Hf-isotope data suggests thatthe zircon populations dominated by low εHf values also showa predom-inance of heavy O-isotopes (δ18O N 6 permil) Lower values appearmainly in time-slices that contain more zircons with more juvenile HfHowever in the dataset reported by Kemp et al (2010) zircons withδ18O N 6 permil appear to be present in most time slices

Valley et al (2006) Pietranik et al (2008) and Kemp et al (2010)interpreted the high values of δ18O as reflecting the subduction or burialleading to remelting of felsic or mafic rocks that had been altered bysurface water However the question remains whether burial andmelting were necessarily involved Instead the scatter of high-δ18Ovalues may represent alteration of zircons by surface (or shallow-crustal) processes either throughweathering or during themetamorphicdisturbances documented by Rasmussen et al (2011) Another questionis whether the inferred high P and T represent long-term burial or theeffects of large meteorite impacts producing felsic melts like thosefound at the Sudbury impact crater (eg Zeit and Marsh 2006)

Many of the above studies demonstrate the confusion that can arisefrom a focus on εHf in isolation A more nuanced viewmay be gained byconsidering the trends in the ldquoraw datardquo ie the simple variation of Hfiwith time Fig 9a shows that the data for the Jack Hills zircons groupinto two populations each with a very small range of Hfi but a widerange (ca 300 Ma) in age This effect also has been demonstrated insingle zoned grains in which younger rims have 177Hf176Hf similar tothat of the core (Kemp et al 2010) Fig 9b shows a similar trendextending horizontally from ca 38 Ga Another but less well-definedtrend may extend horizontally from ca 335 Ga These trends are similarto though less extended in terms of age to those described byBeyer et al(2012) in zircons from the Almklovdalen (Norway) peridotites

We suggest that each of the trends in Fig 9 represents a singlecrustal-formation event (45 Ga 42ndash43 Ga 38ndash39 Ga 33ndash34 Ga)and that most of the age spread in each population reflects (possibly re-peated) metamorphic disturbance of the UndashPb systems In contrastreworking of a crustal volume with a non-zero LuHf would produce anoticeable rise in the overall Hfi through time no such rise is obvious ineither of the two populations outlined in Fig 9a The partial resetting ofthe UndashPb ages may have occurred during the Proterozoic events docu-mented by Grange et al (2010) and Rasmussen et al (2011) but mayalso reflect the events (eg impact melting) that grew rims on somezircons at 39ndash34 Ga (Cavosie et al 2004) Kempet al (2010) recognizedmany of these problems and attempted to filter their UPbndashHf isotopedata on Jack Hills zircons to account for them They concluded that theεHf-age correlation in their filtered data is consistent with the evolutionof a mafic crustal reservoir a similar trend can be seen in the largerdataset (Fig 3b) This raises the possibility that further high-precisiondatasets with multiple isotopic systems measured on the same spots ofcarefully selected zircons will be able to define real crustal-evolutiontrends in specific areas At present the data shown in Fig 9 do not require

Fig 9 Plots of UndashPb age vs 177Hf176Hf in ancient zircons Ovals enclose single populations of zirconswith similar Hf-isotope ratios but a range in ageWe suggest that each of these trendsrepresents a single original population of zircons (ca 45 Ga 42 Ga and38 Ga) inwhich the ages have been reset by surficial ormetamorphic processes withoutmodifying theHf-isotopecompositions (cf Beyer et al 2012) More typical plate-tectonic processes appear to begin around 34 Ga

12 WL Griffin et al Lithos 189 (2014) 2ndash15

real crustal reworking in the sense of burial and remelting to haveoccurred prior to ca 35 Ga

422 Composition of the earliest crustNumerous students of the Jack Hills zircons have argued that they

document the presence of a felsic crust and a cool planet early in theHadean period (Harrison et al 2005 2008 Pietranik et al 2008Valley et al 2002 2006) However as noted above some of the miner-alogical evidence for this idea has been questioned The data of Kempet al (2010) suggest a generally mafic composition for the Hadeancrust The trace-element data of Cavosie et al (2006) indicate thatb10 of the analysed grains came from felsic rocks while 75 of the40 zircons studied crystallized from intermediate (SiO2 b 65) meltswhich could represent differentiates of large mafic complexes Anotherpossibility is the generation of felsic melts by large-scale meteoriteimpact followedbymagmatic fractionation (eg the Sudbury complex)

A ldquoLate Heavy Bombardmentrdquo around 40ndash38 Ga (Fig 10a) has beenwidely accepted as a mechanism for supplying the upper Earth with arange of elements (such as the PGE) that are overabundant relative tothe levels expected after the separation of Earths core (Maier et al2010) However Bottke et al (2012) have argued that the heaviestbombardment would have been earlier during Hadean time it wouldhave begun tapering off by ca 4 Ga but would have remained a

prominent feature of the surface environment down to ca 3 Ga(Fig 10a) andwould have produced large volumes of shallow relativelyfelsic melts especially during the Hadean and early Archean periods

423 Early Archean crustal developmentFrom thediscussion abovewewould argue that the existingHadean

zircon data carry no information requiring ldquocrustal reworkingrdquo (otherthan by impact melting) and cannot be used to infer the presence ofsubduction or othermanifestations of plate tectonics similar conclusionshave been reached by other recent studies (eg Kemp et al 2010)

The same is true of the next juvenile peak at ca 385 Ga in this casemost of the zircons with ages from 38 to 355 Ga and some youngerones have been modelled as reflecting the reworking of felsic materialproduced around 385 Ga with little other juvenile input (eg Pietraniket al (2008) However a large proportion of the zircons with ages from385 to 34 Ga again define a population (Fig 9b) that has essentiallyconstant Hfi over that range of ages rather than the rising Hfi expectedof an extended crustal-reworking process Zircons from the 34 to36 Ga age range which represent the ldquotailrdquo of the population mostlyhave δ18O N6 (Valley et al 2002) This pattern also is perhaps bestinterpreted as the crustal modification of a single population of juvenilezircons rather than reworking of large crustal volumes through burialand melting

Fig 10 Summary of crustndashmantle evolution Grey histogram shows distribution of all zircons in the database for the time period shown red line shows the probability distribution of theseages Green line shows distribution of RendashOs TRDmodel ages for sulfides inmantle-derived xenoliths and xenocrysts The revised distribution ofmeteorite bombardment intensity (Bottkeet al 2012) and a generalized summary of older interpretations of the ldquoLate Heavy Bombardmentrdquo are shown as LHB The homogenization of PGE contents in komatiites from 35 to 29 Ga(Maier et al 2010) may mark the major mantle-overturncirculation events discussed in the text The O-isotope shift noted by Dhuime et al (2012) marks the beginning of a quiescentperiod or perhaps the destruction of the crust during the major overturns from 35 to 29 Ga

13WL Griffin et al Lithos 189 (2014) 2ndash15

The situation appears to change significantly beginning with thejuvenile peak at 335ndash34 Ga (Figs 5 9b) There are still populationsthat can be defined by constant Hfi over a range of ages but there alsois a clear rise in the mean values of Hfi from ca 33 to 31 Ga This periodalso includes many more zircons derived from truly felsic melts(Fig 4b) in contrast to the earlier periods The pattern suggests a pulseof juvenile input followed by 200ndash300 Ma of real crustal reworkingwith continuing minor juvenile contributions

Overall the HadeanndashMesoArchean Hf-isotope pattern defined byseveral peaks in juvenile material followed by long tails of resetting (orreworking) is not that of Phanerozoic subduction systems (eg Collinset al 2011) We suggest that the pattern of the data from 45 Ga to ca34 Ga (Fig 1) represents an essentially stagnant outer Earth (cf Kempet al 2010) affected by two interacting processes One is large-scalemeteorite bombardment producing localised reworking at each majorimpact site as observed in the Sudbury complex but probably on evenlarger scales The second is the periodic overturn of the asthenosphericmantle beneath the stagnant lid bringing up bursts of new melt todisrupt recycle and replenish the lid at intervals of 150ndash200 Ma as illus-trated in Figs 7 and 8

424 Late Archean mdash the formation of the SCLM (and plate tectonics)In contrast the markedly different pattern shown by the NeoArchean

zircons with a continuous wide range of εHf over 100ndash200 Ma is moreconsistent with a subuction-type environment

The original crust now repeatedly reworked is still evident in thedata until ca 34 Ga (εHf values down to minus15 asymp minus20 TDM model agesge38 Ga) but is less obvious after that In several ways the 34 Ga ldquoover-turnrdquo may mark the beginning of a different tectonic style The sulfidedata presented above (Figs 6 10a) suggest that most of the SCLM wasformed between 33 and 27 Ga This is consistent with the results ofthe Global Lithosphere Architecture Mapping project which has con-cluded that at least 70 of all existing SCLM had formed before 27 Ga(Begg et al 2009 2010 Griffin et al 2009 2011) The Archean SCLM isunique it differs clearly from later-formed SCLM in having anomalouslylow Fe contents (relative to Mg) This is a signature of very high-degree melting at high P (4-6 GPa Griffin et al 2009 Herzberg andRudnick 2012) which is consistent with the melting of deep-seatedperidotitic mantle during rapid upwelling

As noted above there is a gap in the zircon data as a whole (Fig 1)and in the degree of juvenile input (Fig 2) around 31ndash29 Ga coincidingwith the rapid buildup of the SCLM recorded in the sulfide data (Fig 10a)A recent summary of most existing compositional data on crustal rocks(Keller and Schoen 2012) shows amarked change in crustal compositionat ca 3 Ga MgO Ni Cr Na and NaK are high in the most ancient rocksand drop sharply at ca 3 Ga from 30 to 25 Ga indicators of more felsiccrust increase steadily (Fig 10b) Dhuime et al (2012) have defined a sig-nificant change in the O-isotope compositions of crustal zircons at 3 Gacorresponding to a ldquomarked decrease in production of juvenile crustrdquo(Fig 10) This time period also coincides with the late stages of mixing

14 WL Griffin et al Lithos 189 (2014) 2ndash15

between older (mantle) and newer (surficial from meteorite bombard-ment) PGE reservoirs as recorded in komatiite data (Maier et al 2010Fig 10a)

We suggest that the apparent gap in the zircon record around thistime reflects larger more rapid mantle overturns during which mostof the SCLM was produced This could cause very chaotic conditions atthe surface resulting in the loss of contemporaneous crust In theupper mantle it could lead to the mixing observed in the komatiitePGE record and to the homogenization and loss of the Hadean mantlereservoir as a source of crustal magmas (Figs 1 and 10)

We also infer from these data that the intense magmatism of the29ndash25 Ga period may reflect not only continued mantle overturnsbut the advent of a modern form of plate tectonics A comparison ofFigs 1 and 2 shows a marked difference in the relative heights of the275 Ga and 25 Ga peaks the younger episode of magmatism containsa much higher proportion of juvenile material In contrast to the earlierpattern of growth spurts followed by long periods of quiescence theoverall pattern of the data in this time period shows both intensereworking of older material (low εHf) and increased input of juvenilematerial (high εHf) In the Phanerozoic record this pattern is seen in theorogens internal to the assembly of Gondwana (Collins et al 2011)and contrasts with the upward trend of the data from external orogens(ie Pacific rim-type settings) These patterns may be related to the29ndash25 Ga assembly of a supercontinent (Kenorland) and imply thatmuch of the surviving crust from that period represents NeoArchean in-ternal orogens involved in that assembly

Naeraa et al (2012) used zircon data (included here) from theancient rocks of W Greenland to argue for a major change in thedynamics of crustal growth at this time a suggestion confirmed bydata from many other regions (Fig 1) The rapid appearance of largevolumes of highly depleted (and thus both rigid and buoyant) SCLMmoving up to and about the surface of the planet would provide amechanism for the initiation of modern-style plate tectonics Thesevolumes of buoyant SCLM also would provide ldquolife raftsrdquo that wouldenhance the possibility of preserving newly formed (and some older)crust

In summary the combined zircon and sulfide data suggest that duringthe Hadean and earliest Archean periods Earths surface was essentiallystagnant and probably dominated by mafic rocks This crust probablywas continually reworked by meteorite bombardment to produce arange of mafic to felsic melts analogous to those exposed in the Sudburyintrusion The Hadean and early Archean were marked by a seriesof mantle overturns or major plume episodes each followed by150ndash300 Ma of quiescence This regime may have continued up to atleast 35 Ga after which one or more large melting events (mantleoverturns) finally produced the buoyant highly depleted residues thatformed the first stable SCLM This activity peaked in the Neo-Archeanand then ceased abruptly by 24 Ga The end of this activity may havemarked the final assembly of the Kenorland supercontinent Piper(2013) using paleomagnetic data has argued that most of the crustexisting by25 Ga accumulated into this landmass resulting in a dramaticdrop inmeanplate velocitieswhich could in turn be linked to themarkeddecrease in magmatic activity from 24 to 22 Ga (Fig 2)

43 Revising the geologic time scale

The International Geologic Time Scale sets the HadeanndashArcheanboundary at 40 Ga and divides the Archean into four approximatelyequal time slices (Fig 1) Eo-Archean (40ndash36 Ga) Paleo-Archean (36ndash32 Ga) Meso-Archean (32ndash28 Ga) and Neo-Archean (28ndash25 GaThere is no obvious basis in the present datasets for these divisions andwe suggest a simplified subdivision based on the apparent changes intectonic style (Fig 8) at identifiable times

In the absence of better data we accept the HadeanndashArcheanboundary at 40 Ga although we suggest that the stagnant-lid regimemay have continued for another 500ndash800 Ma We propose that the

term Eo-Archean be dropped since there is little preserved evidence ofmagmatic activity between 42 Ga and 38 Ga and that PaleoArcheanbe used for the period 40ndash36 Ga which contains the oldest preservedcrust and perhaps evidence for one major overturn MesoArcheancould be applied to the period 36ndash30 Ga during which overturnsbecame more prominent ending with the buildup towards the major28ndash255 Ga magmatism that accompanied the building of the bulk ofthe Archean SCLM NeoArchean can be usefully applied to the period30ndash24 Ga whichmarks a new tectonic stylewith frequent plume activ-ity the beginning of some form of plate tectonics and the preservationof large volumes of continental crust The zircon data suggest that thisactivity continued up to ca 24 Ga and then ceased rather abruptlymarking a natural end to the Neo-Archean period

This suggested revision appears to be a natural one in terms of thedatasets summarised in Fig 10 It will be interesting to see how well itstands the test of time as more data sets emerge

Supplementary data to this article can be found online at httpdxdoiorg101016jlithos201308018

Acknowledgements

This study was funded by Australian Research Council (ARC) grantsprior to 2011 and subsequent funding to the ARC Centre of Excellencefor Core to Crust Fluid Systems (CCFS) and used instrumentation fundedby the Australian Research Council ARC LIEF grants NCRIS DEST SIIgrants Macquarie University and industry sources CON acknowledgesARC support through FT100100717 DP110104145 and the CCFS ARCNational Key Centre This is publication 344 from the ARC Centre of Ex-cellence for Core to Crust Fluid Systems and 903 from the Arc Centre ofExcellence for Geochemical Evolution and Metallogeny of ContinentsVladMalkovets was supported by Russian Foundation for Basic Researchgrants 12-05-33035 12-05-01043 and 13-05-01051

References

Aulbach S Griffin WL Pearson NJ OReilly SY Kivi K Doyle BJ 2004 Mantle for-mation and evolution Slave Craton constraints from HSE abundances and RendashOssystematics of sulfide inclusions in mantle xenocrysts Chemical Geology 208 61ndash88

Aulbach S Creaser RA Pearson NJ Simonetti SS Heaman LM Griffin WL StachelT 2009 Sulfide and whole rock RendashOs systematics of eclogite and pyroxenite xeno-liths from the Slave Craton Canada Earth and Planetary Science Letters 283 48ndash58

Begg GC Griffin WL Natapov LM OReilly SY Grand SP ONeill CJ HronskyJMA Poudjom Djomani Y Swain CJ Deen T Bowden P 2009 The lithosphericarchitecture of Africa Seismic tomography mantle petrology and tectonic evolutionGeosphere 5 23ndash50

Begg GC Hronsky JAM Arndt N Griffin WL OReilly SY Hayward N 2010 Litho-spheric cratonic and geodynamic setting of Ni-Cu-PGE sulfide deposits EconomicGeology 105 1057ndash1070

Belousova EA Griffin WL OReilly SY Fisher NI 2002 Igneous zircon trace elementcomposition as an indicator of host rock type Contributions to Mineralogy andPetrology 143 602ndash622

Belousova EA Kostitsyn YA Griffin WL Begg GC OReilly SY Pearson NJ 2010The growth of the continental crust constraints from zircon Hf-isotope data Lithos119 457ndash466

Beyer EE Brueckner HK Griffin WL OReilly SY 2012 Laurentian provenanceof Archean mantle fragments in Proterozoic Baltic crust of the Norwegian CaledonidesJournal of Petrology 53 1357ndash1383

Bottke WF Vokrouhlicky D Minton D Nesvorny D Morbidelli A Brasser RSimonson B Levison HF 2012 An Archean heavy bombardment from a destabilizedextension of the asteroid belt Nature 485 78ndash81

Bouvier A Vervoort JD Patchett PJ 2008 The LundashHf and SmndashNd isotopic compositionof CHUR constraints from unequilibrated chondrites and implications for the bulkcomposition of terrestrial planets Earth and Planetary Science Letters 273 48ndash57

Carlson RW Pearson DG James DE 2005 Physical chemical and chronological charac-teristics of continental mantle Reviews of Geophysics 43 RG1001 httpdxdoiorg1010292004RG000156

Cavosie AJ Wilde SA Liu D Weiblen PW Valley JW 2004 Internal zoning and U-Th-Pb chemistry of Jack Hills detrital zircons a mineral record of early Archean toMesoproterozoic (4348-1576 Ma) magmatism Precambrian Research 135 251ndash279

Cavosie AJ Valley JW Wilde SA EIMF 2005 Magmatic δ18O in 4400ndash3900 Ma detri-tal zircons a record of the alteration and recycling of crust in the Early Archean Earthand Planetary Science Letters 235 663ndash681

Cavosie AJ Valley JW Wilde SA EIMF 2006 Correlated microanalysis of zircon traceelement δ18O and UndashThndashPb isotopic constraints on the igneous origin of complexN3900 Ma detrital grains Geochimica et Cosmochimica Acta 70 5601ndash5616

15WL Griffin et al Lithos 189 (2014) 2ndash15

CollinsWJ Belousova EA Kemp AIS Murphy JB 2011 Two contrasting Phanerozoicorogenic systems revealed by hafnium isotope data Nature Geoscience 4 334ndash337

Condie K ONeill C 2010 The ArcheanndashProterozoic boundary 500 My of tectonic tran-sition in Earth history American Journal of Science 310 775ndash790

Condie KC ONeill C Aster R 2009 Evidence and implications of a widespread magmaticshut down for 250 Myr on Earth Earth and Planetary Science Letters 282 294ndash298

Davies GF 2006 Gravitational depletion of the early Earths uppermantle and the viabilityof early plate tectonics Earth and Planetary Science Letters 243 (3) 376ndash382

Debaille V ONeill C Brandon AD Haenecour P Yin Q-Z Mattielli N Treiman AH2013 Stagnant-lid tectonics in early Earth revealed by 142Nd variations in lateArchean rocks Earth and Planetary Science Letters 373 83ndash92

Dhuime B Hawkesworth CJ Cawood PA Storey CD 2012 A change in thegeodynamics of continental growth 3 billion years ago Science 335 1334ndash1336

Gao Y Snow JE Casey JF YJ 2011 Cooling-induced fractionation of mantle Li iso-topes from the ultraslow-spreading Gakkel Ridge Earth and Planetary Science Letters301 231ndash240

Geng Y Du L Ren L 2012 Growth and reworking of the early Precambrian continentalcrust in the North China Craton constraints from zircon Hf isotopes GondwanaResearch 21 517ndash529

Gerya T 2012 Precambrian geodynamics concepts and models Gondwana Researchhttpdxdoiorg101016jgr201211008

Gonzalez-Jimenez JM Gervilla F Griffin WL Proenza JA Auge T OReilly SYPearson NJ 2012a Os-isotope variability within sulfides from podiform chromititesChemical Geology 291 224ndash235

Gonzalez-Jimenez JM Griffin WL Gervilla F Kerestedjian TN OReilly SY ProenzaJA Pearson NJ Sergeeva I 2012b Metamorphism disturbs the RendashOs signatures ofplatinum-group minerals in ophiolite chromitites Geology 40 659ndash662

Gonzalez-Jimenez JM Griffin WL Gervilla F Proenza JA OReilly SY Pearson NJ2013 Chromitites in ophiolites How where when and why Part I A review andnew ideas on the origin and significance of platinum-group minerals Lithos 189127ndash139

Grange ML Wilde SA Nemchin AA Pidgeon RT 2010 Proterozoic events recordedin quartzite cobbles at Jack Hills Western Australia new constraints on sedimenta-tion and source of N4 Ga zircons Earth and Planetary Science Letters 292 158ndash169

Griffin WL Pearson NJ Belousova E Jackson SE OReilly SY van Achterberg E SheeSR 2000 The Hf isotope composition of cratonic mantle LAM-MC-ICPMS analysis ofzircon megacrysts in kimberlites Geochimica et Cosmochimica Acta 64 133ndash147

GriffinWL Spetsius ZV Pearson NJ OReilly SY 2002 In-situ RendashOs analysis of sulfideinclusions in kimberlitic olivine new constraints on depletion events in the Siberianlithospheric mantle Geochemistry Geophysics Geosystems 3 1069

Griffin WL Graham S OReilly SY Pearson NJ 2004a Lithosphere evolution beneaththe Kaapvaal Craton RendashOs systematics of sulfides in mantle-derived peridotitesChemical Geology 208 89ndash118

Griffin WL OReilly SY Doyle BJ Pearson NJ Coopersmith H Kivi K Malkovets VPokhilenko NV 2004b Lithosphere mapping beneath the North American PlateLithos 77 873ndash902

Griffin WL OReilly SY Afonso JC Begg G 2009 The composition and evolution oflithospheric mantle a re-evaluation and its tectonic implications Journal of Petrology50 1185ndash1204

GriffinWL Begg GC Dunn D OReilly SY Natapov LM Karlstrom K 2011 Archeanlithospheric mantle beneath Arkansas continental growth by microcontinent accre-tion Bulletin of the Geological Society of America 123 1763ndash1775

Griffin WL Nikolic N OReilly SY Pearson NJ 2012 Coupling decoupling and meta-somatism evolution of crustndashmantle relationships beneath NW Spitsbergen Lithos149 115ndash139

Harrison TM Blichert-Toft J Muumlller W Albarede F Holden P Mojzsis SJ 2005Heterogeneous Hadean hafnium evidence of continental crust at 44 to 45 Ga Science310 1947ndash1950

Harrison TM Schmitt AK McCulloch MT Lovera OM 2008 Early (ge45 Ga) forma-tion of terrestrial crust LundashHf d18 O and Ti thermometry results for Hadean zirconsEarth and Planetary Science Letters 268 476ndash486

Herzberg C Rudnick R 2012 Formation of cratonic lithosphere an integrated thermaland petrological model Lithos 149 4ndash14

Keller CB Schoen B 2012 Statistical geochemistry reveals disruption in secularlithospheric evolution about 25 Ga ago Nature 485 490ndash495

Kemp AIS Wilde SA Hawkesworth CJ Coath CD Nemchin A Pidgeon RTVervoort JD DuFrame SA 2010 Hadean crustal evolution revisited new con-straints from PbndashHf isotope systematics of the Jack Hills zircons Earth and PlanetaryScience Letters 296 45ndash56

Korenaga J 2011 Thermal evolution with a hydrating mantle and the initiation of platetectonics in the early Earth Journal of Geophysical Research Solid Earth (1978ndash2012)116 (B12)

Kusiak MA Whitehouse MJ Wilde SA Nemchin AA Clark C 2013 Mobilization ofradiogenic Pb in zircon revealed by ion imaging implications for early Earth geochro-nology Geology httpdxdoiorg101130G339201 |

Maier WD Barnes SJ Campbell IH Fiorentini ML Peltonen P Barnes S-J SmithiesH 2010 Progressive mixing of meteoritic veneer into the early Earths deep mantleNature 460 620ndash623

Marchesi C Griffin WL Garrido CJ Bodinier J-L OReilly SY Pearson NJ 2010Persistence of mantle lithospheric RendashOs signature during asthenospherization ofsubcontinental lithospheric mantle insights from in situ analysis of sulfides from theRonda peridotite (S Spain) Contributions to Mineralogy and Petrology 159 315ndash330

Moresi L Quenette S Lemiale V Meriaux C Appelbe B Muumlhlhaus HB 2007 Com-putational approaches to studying non-linear dynamics of the crust and mantlePhysics of the Earth and Planetary Interiors 163 (1) 69ndash82

Moresi L Solomatov V 1998 Mantle convection with a brittle lithosphere thoughts onthe global tectonic styles of the Earth and Venus Geophysical Journal International133 (3) 669ndash682

Naeraa T Scherstein A Rosing MT Kemp AIS Hofmann JE Koldelt TFWhitehouse MJ 2012 Harnium isotope evidence for a transition in the dynamicsof continental growth 32 Ga ago Nature 485 627ndash630

Nemchin AA Pidgeon RT whitehouse MJ 2006 Re-evaluation of the origin and evo-lution of N42 Ga zircons from the Jack Hills metasedimentary rocks Earth and Plan-etary Science Letters 244 218ndash233

Nimmo F Price GD Brodholt J Gubbins D 2004 The influence of potassium on coreand geodynamo evolution Geophysical Journal International 156 (2) 363ndash376

ONeill C Debaille V Griffin WL 2013 Deep Earth recycling in the Hadean andconstraints on surface tectonics American Journal of Science (in press)

ONeill C Lenardic A Weller M Moresi L Quenette S 2013 A window for platetectonics in terrestrial planet evolution Geophysical Journal International (in revision)

ONeill C Lenardic A Moresi L Torsvik T Lee CA 2007 Episodic Precambriansubduction Earth Planetary Science Letters 262 552ndash562

OReilly SY Griffin WL 2012 Mantle metasomatism In Harlov DE Austrheim H(Eds) Metasomatism and the chemical transformation of rock Lecture Notes in EarthSystem Sciences Springer-Verlag Berlin Heidelberg pp 467ndash528 httpdxdoiorg101007978-3-642-28394-9_12

Pearson NJ Alard O Griffin WL Jackson SE OReilly SY 2002 In situ measurementof RendashOs isotopes in mantle sulfides by laser ablation multi-collector inductively-coupled mass spectrometry analytical methods and preliminary results Geochimicaet Cosmochimica Acta 66 1037ndash1050

Pietranik AB Hawkesworth CJ Storey CD Kemp AIS Sircombe KN WhitehouseMJ BleekerW 2008 Episodicmafic crust formation from 45ndash28 Ga new evidencefrom detrital zircons Slave Craton Canada Geology 36 875ndash878

Piper JDA 2013 A planetary perspective on Earth evolution lid tectonics before platetectonics Tectonophysics httpdxdoiorg101016jtecto201212042

Rasmussen B Fletcher IR Muhling JR Wilde SA 2010 In situ UndashThndashPb geo-chronology of monazite and xenotime from the Jack Hills belt implications forthe age of deposition and metamorphism of Hadean zircons Precambrian Re-search 180 26ndash46

Rasmussen B Fletcher IR Muhling JR Gregory CJ Wilde SA 2011 Metamorphicreplacement of mineral inclusions in detrital zircon from Jack Hills Australia impli-cations for the Hadean Earth Geology 39 1143ndash1146

Scherer E Munker C Metzger K 2001 Calibration of the LutetiumndashHarnium clockScience 293 683ndash687

Shi R Alard O Zhi X OReilly SY Pearson NJ Griffin WL Zhang M Chen X 2007Multiple events in the Neo-Tethyan coeanic upper mantle evidence from RundashOsndashIralloys in the Luobusa and Dongquao ophiolitic podiform chromitites Tibet Earthand Planetary Science Letters 261 33ndash48

Silver PG Behn MD 2008 Intermittent plate tectonics Science 319 85ndash88Spetsius ZV Belousova EA Griffin WL OReilly SY Pearson NJ 2002 Archean

sulfide inclusions in Paleozoic zircon megacrysts from the Mir kimberlite Yakutiaimplications for the dating of diamonds Earth and Planetary Science Letters 199111ndash126

Ushikubo T Kita NT Cavosie AJ Wilde SA Rudnick RL Valley JW 2008 Lithium inJack Hills zircons Evidence for extensive weathering of Earths earliest crust Earthand Planetary Science Letters 272 666ndash676

Valley JW PeckWH King EMWilde SA 2002 A cool early Earth Geology 30 351ndash354Valley JW Cavoisie AJ Wilde SA 2006What havewe learned from pre-4 Ga zircons

Goldschmidt Conference Abstracts p A664van Hunen J Moyen J-F 2012 Archean subduction fact or fiction Annual Review of

Earth and Planetary Sciences 40 195ndash219Wilde SA 2010 Proterozoic volcanism in the Jack Hills Belt Western Australia some

implications and consequences for the worlds oldest zircon population PrecambrianResearch 183 9ndash24

Wilde SA Valley JW Peck WH Graham CM 2001 Evidence from detrital zir-cons for the existence of continental crust and oceans on the Earth 44 Ga agoNature 269 175ndash178

Zeit MJ Marsh BD 2006 The Sudbury igneous complex viscous emulsion differentia-tion of a superheated impact melt sheet Bulletin of the Geological Society of America177 1427ndash1450