High-pressure phase equilibria of a high-magnesia basalt and the

American Mineralogist, Volume 69, pages l-15, 1984

High-pressure phase equilibria of a high-magnesia basalt andthe genesis of primary oceanic basalts

DoN Er-rHoN

Department of GeosciencesUniversity of Houston, Houston, Texas 77004

eNo CHRIsToPHER M. Scenpe

Experimental Petrology LaboratoryDepartment of Geology, University of Alberta

Edmonton, Alberta T6G2E3, Canada

Abstract

High-pressure phase equilibria studies on a high-MgO basalt from the Tortuga ophiolitecomplex indicate that olivine + orthopyroxene * clinopyroxene + garnet are its liquidusphases at 25 kbars. At l0 to 20 kbars, olivine is the liquidus phase and is followed by spinel,clinopyroxene, and orthopyroxene as temperature decreases. At <10 kbars, the order ofcrystalization is olivine, spinel, plagioclase, clinopyroxene, and orthopyroxene.

The high-pressure phase equilibria determined in this study, supplemented with theresults obtained on other tholeiitic basalt compositions, are used to construct pseudo-liquidus phase diagrams for evaluating and predicting the nature of primary magmas andmelting of the mantle at high pressures. These pseudo-liquidus phase diagrams suggest that"primitive" oceanic basalts with >9.5Vo MgO are derived from primary oceanic basaltsgenerated at 15 to 25 kbars.

Models for the origin of oceanic basalts at low pressures (=10 kbars) are consideredincapable of generating the most "primitive" oceanic basalts that have >9.57o MeO. It islikely that most oceanic basalts with>9.5Vo MgO are ultimately derived from primary high-MgO basalts with>l4Vo MgO that have been produced by melting at 15 to 25 kbars. High-MgO basalts of this type are found in numerous oceanic or rifting environments (e.g.,Baffin Bay, Gorgona Island, Tortuga ophiolite, Betts Cove ophiolite, Lewis Hills ophiolite,Ungava Peninsula), suggesting that these high-MgO basalts are produced on a world-widescale throughout geologic time. Chemical differentiation processes, however, normallymodify these high-MgO basalts into the more common basalts with 7 to lD%MgO. Becausemost oceanic basalts with <9.5% MgO have equilibrated to low-pressure cotectics,determination of the conditions of origin for the primary magmas from which they arederived is not well constrained. For this reason, melting of the mantle at pressures of =10kbars could produce primary magmas capable of differentiating to form oceanic basalts thatare similar to many of the evolved oceanic basalts that have <9.5% MgO. Untilcrystallization and mixing processes are more fully understood, melting at <10 kbarscannot be eliminated as a possibility for the origin of some primary oceanic basalts.

Introduction

Numerous models for the origin of oceanic basalts havebeen proposed and are being actively debated by petrolo-gists and geochemists (e.g., O'Hara, 1968; Green et al.,1979; Presnall et a|.,1979; Stolper, 1980; and referencestherein). One ofthe critical questions to be resolved in thecontroversy concerning the origin of oceanic basalts isthe nature of primary magmas and their evolution withinthe crust and upper mantle. A broad consensus exists

0003-004)v84l0 l 02-000 l $02. 00

among petrologists and geochemists that oceanic basaltsare ultimately derived by partial melting of the peridotiticsub-oceanic mantle. Primary magmas produced by melt-ing under these conditions would have olivine (OL) +orthopyroxene (OPx)+clinopyroxene (CPX) and eithergarnet (GAR), spinel (SP), or plagioclase (PLAG) asliquidus phases at the pressure offinal equilibration of theprimary magma with the surrounding mantle. Especiallycritical in this regard is the requirement that primarybasalts generated in the sub-oceanic mantle must be

ELTHON AND SCARFE: EOUILIBRIA OF HIGH.MAGNESIA BASALT

multiply-saturated with OL + OPX because these twominerals are known to coexist with the liquid produced bypartial melting of peridotites even at very large incre-ments of fusion (Mysen and Kushiro,1977; Jaques andGreen, 1980; Scarfe et al.,1979; Harrison, 1981, Elthon,l98l). As O'Hara (1968) originally proposed, most ocean-ic basalts do not have olivine as a liquidus phase atpressures >10 kbars and do not have orthopyroxene as aliquidus phase at any pressure. These factors, in conjunc-tion with the observation that the Mg/(Mg + Fe) [themg #l of most oceanic basalts is too low for them to haveequilibrated with mantle olivines of Forss composition,indicate that most oceanic basalts are not primary mag-mas.

The main controversy, however, surrounding the na-ture of primary magmas that are generated in the sub-oceanic mantle is focused on the relatively "primitive"oceanic basalts that are characterized by high Ni contents(>200 ppm), high MgO contents (9.5 to ll%MeO), andgenerally low incompatible trace element abundances.These basalts are believed to have undergone the smallestextent of chemical differentiation prior to eruption, al-though the actual extent of this differentiation is conten-tious.

Previous studies of these "primitive" oceanic basaltsby Green et al. (1979) and Bender et al. (1978) haveshown that olivine disappears as a liquidus phase in"primitive" oceanic basalts at -10 kbars and that ortho-pyroxene is not a liquidus phase in these basalts at anypressure that was experimentally investigated, as predict-ed by O'Hara. In fact, Green et al. (1979) demonstratedthat DSDr basalt 3-18-7-1, one of the most "primitive"oceanic basalts recovered, was significantly displacedfrom the orthopyroxene primary phase volume at 12kbars, requiring the addition of9Vo orthopyroxene to thebasalt before the basalt became saturated with orthopyr-oxene. Green et al. (1979) also demonstrated that DSDIbasalt 3-18-7-1 required the addition of 17% olivine beforeco-saturation with both olivine and orthopyroxene wasachieved at 20 kbars. On the other hand, Kushiro (1973)and Fujii and Kushiro (1977) present phase diagrams oftwo oceanic basalts that do show multiple saturation ofOL + OPX + CPX + PLAG at the liquidus at 7.5 to 8kbars. In the Kushiro (1973) and Fujii and Kushiro (1977)experiments, however, the liquidus orthopyroxenes inthe experimental runs showing OL + OPX + CPX +PLAG saturation were too iron rich (mg# of 84 and 86,respectively) to have been in equilibrium with residualmantle OPX (mg# 88-91.5). In support of the proposalthat some "primitive" oceanic basalts may be primarymagmas, Fujii and Bougault (1983) have shown that OL *OPX + CPX + PLAG + SPINEL of an appropriate high-magnesium composition are on or near the liquidus of a"primitive" basalt from the reuous area at 10 kbars.

One problem with many of these previous experimentalstudies is that the basalts chosen as the starting composi-tions have apparently undergone crystal fractionation

prior to eruption. Attempts to deduce and then to ade-quately compensate for the pre-eruption history of thesebasalts introduced additional ambiguity into the interpre-tation of the origin of oceanic basalts.

In this study, we have chosen an aphyric high-MgObasalt (sample number NT-23) from the Tortuga ophiolitecomplex in southern Chile as our starting composition.This rock was selected because its high Me/(Me+Fe)[75.9] indicates equilibration with residual mantle oliv-ines, the olivine microphenocrysts are appropriate forprimary magmas (ave Fo-ee z), and its composition issimilar to an estimated bulk composition of the oceaniccrust (Elthon, 1979). Basalts from the Tortuga ophiolitecomplex (Elthon, 1980) are very similar to the FAMousbasaltic glasses (Bryan and Moore, 1977) and otheroceanic basalt suites (Melson et al., 1977). High-MgObasalts broadly similar in composition to NT-23 havebeen reported from a number of localities (Baffin Bay:Clarke, 1970; Ungava Peninsula: Francis and Hynes,1979; Betts Cove ophiolite: Upadhyay 1978; GorgonaIsland: Echeverria, 1980; Lewis Hills Massif of the Bay ofIslands ophiolite: Karson et al., 1983) suggesting thatthese high-MgO basalts are produced on a world-widescale throughout geological time, and that chemical dif-ferentiation processes normally modify these high-MgObasalts into the more abundant basalts with 7 to l0%oMeo.

This experimental study shows that a high-MgO basal-tic liquid, similar in composition to NT-23, is multiplysaturated with olivine + orthopyroxene * clinopyroxene+ garnet at 25 kbars. The preliminary results ofthis studyhave been previously described by Elthon and Scarfe0980).

Experimental procedure

Chips of the fine-grained margin of the rock (Table l)were powdered in a tungsten--carbide vial of a Spexmill.The powder was further ground under acetone for 3hours, then dried for 20 hrs. at -ll00"C at the QFMoxygen fugacity buffer to decompose secondary alter-ation products. The powder was stored in a drying ovenat -110'C between runs.

The powder was loaded into graphite containers, heat-ed until red, and then sealed inside 3mm O.D. platinumcapsules. A pyrex glass sleeve around the heating ele-ment inhibited water access to the sample region andminimized hydrogen diffusion into the capsule. The inter-nal graphite capsules prevented iron loss from the sampleto the platinum capsule and maintained a relatively lowfo2 at the graphite-CO buffer. Because graphite capsuleswere used in these experiments, a minor amount of COmay have dissolved in the melt at high pressures (e.g.,Green et al.. 19791. The amount of CO in the melt ispresumed to be small because the electron microprobeanalyses of the glasses sum to near l00Vo (see Table 3).

All experiments were conducted with a solid-media,high-pressure apparatus with a Yz in. diameter furnace

ELTHON AND SCARFE: EOUILIBRIA OF HIGH-MAGNESIA BASALT

Table l. Major element composition of NT-23

si0"Ti.o:A1,02FeO "

lrtoCaOKr0re"0HndTotal

-7913.649 . 7 7

1 7 . 6 19 . 5 8

.06

.89

. 1 499.73

R r l t R n n k a F i n - - G . o i . o i M . - o i . b

natural pyroxenes and amphiboles for the pyroxene anal-yses. All of the samples are corrected for matrix effectsby the method of Bence and Albee (1968) and Albee andRay (1970). Clinopyroxene crystals in the experimentalcharges are often small and are difficult to analyze withhigh-quality because of overlap with the sulroundingglass. Unless otherwise noted below for a specific run,this problem was not significant for other minerals.

Results

The high-pressure phase equilibria of high-MgO basaltNT-23 are shown in Figure l. The pressure, temperature,duration, and run products of the experiments are listedin Table 2. Olivine (*spinel) is the liquidus phase atpressures of <25 kbar, but the temperature interval forolivine-only (-+spinel) crystallization diminishes with in-creasing pressure from (150'C at l0 kbars to <10" at 25kbar. The sequence in which minerals crystallize in NT-23 changes from OL-SP-PLAG-CPX at I atm to OI-SP-CPX-OPX at 10-20 kbars. At 25 kbar and 1470'C.olivine, orthopyroxene, clinopyroxene, and garnet areco-liquidus phases. This is the only pressure at which alherzolite assemblage is on the liquidus of NT-23. Theresults of I atmosphere runs, which were done on 0.003in. diameter Pt wire loops at the QFM oxygen fugacitybuffer, will be reported elsewhere with the I atm experi-mental phase relationships of other basalts from Tortuga.

10 kbar

Olivine is the liquidus phase at l0 kbars and crystallizesalone (ta trace of spinel) for -150'C; clinopyroxenebegins to crystallize with olivine + spinel at 1230";

4 7 . 7 3 ! . 2 1. 8 5 ! . 0 3

13.42 ! .239 . 5 8 ! . 2 4

1 6 . 8 9 t . 1 89 . 9 4 I . 1 40 . 0 5 1 . 0 21 0 7 1 . 0 3

. 2 0 t . 0 39 9 . 7 3

aFron Elthon (1929)

bAv"..g. of Slass cooposit iooa froD rus 304, 306, 313, 320, 328

assembly (Boyd and England, 1960). The piston-out tech-nique was used with a -4% correction for friction. Thequoted pressures given in Table I are considered accurateto -+l kbar. Temperatures were monitored with Pttgo-Pb,oRhtg thermocouples and were automatically con-trolled (Hadidiacos, 1972) with an accuracy of tl0"C.Most runs were I hour in duration. The observation thatno change occurred in the phase equilibria betweenexperiments of I and l0 hrs. duration of 25 kbars and1470" was taken to indicate that equilibrium was achievedin the shorter runs.

All phase identifications have been verified with theelectron microprobe. The microprobe analyses reportedhere were determined on an ARL-srue nine-channelmicroprobe housed at the Department of Mineral Sci-ences at the Smithsonian Institution using the standardanalytical techniques described by Melson (1978). Stan-dards of natural basaltic glasses were used for glassanalyses, natural olivines for the olivine analyses, and

ooa

Ot

o

o-

Temperolure (cCl

Fig. l. Phase diagram for NT-23.

o L i Q. O l r S p! o t + P l+ O l + C p x + S px Ol+Cpx+Opx+Sp* Ol+Opx+Cpx+Gora Cprv Cpx+Goro Cpr+Opx+Gor

O l * P l + C p x r /{ .L iq

, t/ O l + P l

, ' + L i q

Ol + Cpx +Sp+ L


Table 2. Summary of experimental results

l 01 0l 010101 0101010

Pressure Terp€latu!€ RM Tioe Phases( (bara) ( 'C) I lMber (h r r . ) Present

over the interval from lZ-l1Vo crystallization, and OL +OPX + CPX + SP from 15-17% crvstallization.

20 kbar

Olivine is the liquidus phase but its crystallizationinterval is much smaller (-30'C; -6Vo crystals) than atlower pressures. The isobaric pseudo-invariant point isreached after only 8Vo crystallization.

25 kbar

Olivine, orthopyroxene, clinopyroxene and garnet co-exist with a high-MgO basaltic liquid on the liquidus orwithin 10" of it at 25 kbars. Only a trace amount ofcrystals (<2Vo) was formed in run 164 at 1470'C. Theolivine (Foas s) and the orthopyroxene (mg = 99.4;formed crystals that were large enough to easily analyzewith the microprobe. The clinopyroxene and garnet crys-tals were very small and it is possible that those analysesinclude overlap with some of the adjacent glass. Attemperatures below 1465', olivine has not been observedas a run product. This absence is in agreement with theOL + LIQ c GAR + OPX + CPX reaction of Kushiroand Yoder (1974\.

30 kbar

The phase relationships have been determined on areconaissance basis at 30 kbars. CPX is the liquidus phaseat 30 kbars and garnet is the second mineral to crystallize.The extensive development of quench pyroxene over-growths precluded meaningful analysis of the run prod-ucts at 30 kbars.

Discussion of results

Spatial location of isobaric pseudo-invariantpoints in natural basalt-mantle systems

Coexistence offour mineral phases with a silicate liquidat a specified pressure defines an isobaric invariant pointin a quaternary system. Presnall et al. (1979) have deter-mined the chemical compositions of silicate liquids coex-isting with olivine, orthopyroxene, clinopyroxene, andeither plagioclase (at <9.3 kbar) or spinel (at >9.3 kbar)for 6 isobaric invariant points at I atm to 20 kbar pressurein the quaternary CaO-MgO-Al2O3-SiO2 system. Hoo-ver and Presnall (1981) have extended this earlier workinto the sodium-bearing synthetic system at I atm to 20kbar.

Stolper (1980) attempted to equilibrate basaltic glassfrom the project FAMous area of the Mid-Atlantic Ridgewith an olivine + orthopyroxene I clinopyroxene assem-blage at 10, 15, and 20 kbar pressure in order to constrainthe compositions of liquids saturated with these minerals(OL + OPX * CPX) in the natural multicomponentbasalt-mantle system. The glass compositions reportedby Stolper (1980) were not in equilibrium with spinel and,therefore, they represent compositional points along the

13901380135013201290r26012301200l r 7 0

l 4 t 013901350r325r300

3083r7

3 1 6324323

3 3 3

3203 1 93 1 8325326330

c t + 0 tG t + 0 nG t + 0 1G t + o ! + s P6 1 + o tG L + O L T S PG L + 0 L + S P + C P Xc [ + 0 [ + c P xG L + O L + S P + C P X +OPX + PLAC

CLG L + O [0 L + 0 L + s PG L + O LG L + O L + S P + C P Xc [ + 0 L + s P + c P x + o P x

l 5I Jt )l 5

20202020

2020

2 2 5

25

25252525

303030

1465 3061440 3071425 1631420 310

1420 31041400 309

1460 321

1475 3041470 3281470 1541465 3321463 3141450 3021425 305'1400 303

1500 32214E5 3111465 312

1I

l 0I1

III

I1I

GI,c L + 0 1c L + 0 L + s P + c P xc L + o t + s P + c P x + o P x +quench Pyroxenc6c L + o L + s P + c P x + o P xG L + 0 L + S P + C P X + O P X

CL

GL

G L + O L + C P X + O P X + C A RG T , + 0 t + c P x + o P x + o A RG l + c P x + o P x + G A RCPX + OPX + CAR + quench PyroxeoeaCPX + OPX + CAR + quench PyroxeoeaCPX + OPX + GAR + Quoch PyloxeDes

CPX + Quelcb PyroxenesCgX + Quench PyroxcnedCPX + GAR + Quench Pyroxeaea +Cu, Ni, Fe Sulphide

orthopyroxene, and plagioclase crystallize with olivine,clinopyroxene and spinel at 1170'C. Using a least squaresmixing program (Bryan et al.,1969), we have calculatedthe weight percentages of the various phases in each ofthe experimental runs. The results of these calculationsare given at the bottom of Table 3. Spinel was notincluded as a mineral in these mass balance calculationsbecause spinel occurs in only trace amounts of very smallcrystals that are generally too small to quantitativelya\alyze with the electron microprobe.

Olivine (-rspinel) crystallizes alone until -l8Vo olivinecrystals have formed. The composition of the first olivineto crystallize in the l0 kbar experimental runs is Foes.5but, by extrapolation of results from lower temperatures,it is likely that the liquidus olivine is Foeqs_e1.s. Theolivines become more iron-rich during crystallization asthe liquid becomes depleted in MgO and enriched in thoseelements that are incompatible in olivine. Notice that theglass compositions are basaltic in character along theisobaric OL + CPX + SP + LIQ pseudo-univariant curveand at rhe oL + opx + cpx + sp + PLAG + LIQpseudo-invariant point.

15 kbar

Olivine -+ spinel is the liquidus phase at l5 kbars. CpXbegins to crystallize at -1325"(-100o below the liquidus)and OPX begins to crystallize at -1300'. Olivine crystal-lizes alone (-+-spinel) for the first l2Vo, OL + CPX + Sp


OL + OPX + CPX + LIQ isobaric pseudo-univariantcuryes or the OL + OPX + LIQ isobaric pseudo-divariant surface. It is uncertain. however. whether thisabsence of spinel saturation is of much significance,because of the low abundance of spinel in the upper,mantle and its complex solid solution behavior. /

In an effort to evaluate the compositions of liquids atisobaric pseudoinvariant points in the natural basalt-mantle system, we use the compositions of glasses fromruns 164 (25 kbar),310A and 309 (20 kbar),330 (15 kbar),and 333 (10 kbar) as well as the results of high-pressureexperiments on other tholeiitic basalt compositions.These high-pressure phase equilibria are projected fromclinopyroxene onto the olivine-plagioclase-silica plane(Fig. 2) and from anorthite onto the olivine-clinopyrox-ene-silica plane (Fig. 3). For details of the projectionmethod, see Elthon (1983).

Several of the compositions studied show multiple-saturation involving olivine + orthopyroxene+otherphases at high pressures. In addition to the isobaricpseudo-invariant points derived from NT-23 (runs 164,309, 330, and 333) and Fujii and Bongault (1983), theolivine basalt (O.B.) studied by Green and Ringwood(1967) crystallized olivine * orthopyroxene + clinopyr-oxene just below the liquidus at 9 kbar. The oceanictholeiite (3954) studied by Fujii and Kushiro (1977) hasolivine + orthopypoxene * clinopyroxene + plagioclasejust below the liquidus at 7.5 kbar. Kushiro (1973) foundthat olivine tholeiite T-87 has olivine * orthopyroxene *clinopyroxene + plagioclase as liquidus phases at 8 kbar.Basalt 66018 from Skye (Thompson, 1974) has olivine asthe liquidus phase at <16.5 kbars and clinopyroxene athigher pressures. Slightly below the liquidus in the pres-sure interval of -14 to 16 kbar, olivine + orthopyroxene+ clinopyroxene -t spinel(?) are in equilibrium with theliquid, indicating that the composition of this basalt liesclose to the l5 kbar pseudo-invariant point.

The olivine tholeiite (O.T.) of Green and Ringwood(1967) has olivine as a liquidus phase up to 13.6 kbar. Atpressures >13.6 kbars and <22.5 kbar, orthopyroxene isthe liquidus phase with clinopyroxene as the secondphase. At pressures of >22.5 kbar, clinopyroxene is theliquidus phase. These experiments agree with the shift inphase equilibrium boundaries at high pressures as shownin Figures 2 and 3.

The picrite of Green and Ringwood (1967) has olivine +clinopyroxene + garnet within 20" of the liquidus at 22.5kbar and orthopyroxene + clinopyroxene + garnet justbelow the liquidus at 27.5 kbar. These experiments sug-gest that at -25 kbar olivine + orthopyroxene * clino-pyroxene + garnet would be on or near the liquidus forthis composition.

The 1840 picrite of Tilley and Yoder (1964) has olivineas the liquidus phase, orthopyroxene is second, clinopy-roxene is third and garnet is the fourth phase to crystallizeat 20 kbar. These phase equilibria also agree with the shiftin boundaries as shown in Figure 2 and 3.

The phase equilibria boundaries as shown in Figures 2and 3 are also supported by the high-pressure experimen-tal results obtained on two of the most "primitive"oceanic basalts. Green et al. 0979\ found that olivine isthe liquidius phase of pspp basalt 3-18-7-l up to 10 kbarand that olivine is replaced by clinopyroxene as theliquidus phase at higher pressures. Orthopyroxene wasnot observed within the P-T interval studied and evenwith the addition of 5Vo orthopyroxene to the sample, theIiquid did not become saturated with orthopyroxene at l2kbar. It took the addition of 9Vo orthopyroxene to thestarting basalt in order to saturate the resulting liquid withorthopyroxene. Green et al. added 9Vo olivine to thestarting basalt and found that at l5 kbar, this mixture hadolivine * clinopyroxene as the liquidus phases and that athigher pressures, clinopyroxene was the liquidus phase.Orthopyroxene was not found in any of these experi-ments. By mixing lTVo olivine with the starting basalt,Green et a/. found that olivine was the liquidus phase at20 kbar, but that olivine + orthopyroxene were nearIiquidus phases.

Bender et al. (1978) determined the phase relationshipsfor project FAMous basalt 527-l-l at 0 to 15 kbar. Theirdata indicate that olivine is the liquidus phase at pressuressll kbars and clinopyroxene is the liquidus phase athigher pressures. Even though Bender et al. state that527-l-l "is multiply saturated at 10.5 kbar with olivine,c l inopyroxene, and or thopyroxene", or thopy-roxene is not observed as a liquidus phase at any pres-sure. The 15 kbar results of Bender et a/. show thatclinopyroxene is the liquidus phase but is replaced byorthopyroxene at lower temperatures. This elimination ofclinopyroxene and its replacement by orthopyroxenerequires a CPX + LIQ = OPX reaction relationship. Thisreaction does not occur unless a plane tangential to theOPX + CPX + LIQ surface intersects the extension ofthe OPX,,-CPX", join at the orthopyroxene end. Thistangential plane lies close to natural clinopyroxene com-positions and it is possible that a reaction eliminatingorthopyroxene but not clinopyroxene will occur. In aneffort to evaluate the proximity of basalt 527-l-l to theorthopyroxene primary phase volume, Bender (pers.comm., l98l) has tried orthopyroxene addition experi-ments similar to those of Green et al. /197D and foundthat the 521-l-l liquid at 15 kbar and l3l0' is not inequilibrium with orthopyroxene. Presumably, the ortho-pyroxene at 15 kbar reported by Bender et al. (1978)wasa metastable product.

These results from several different laboratories form areasonably consistent data set for the locations of uppermantle phase equilibria boundaries. This arrangement ofphase equilibria boundaries (Figs. 2 and 3) is not veryconsistent with the melting relationships of natural peri-dotites. The results from partial melting of peridotites(not shown) scatter incoherently across these diagrams, afeature that we believe is a consequence of the difficultyin determining the equilibrium composition of the small


Table 3. Electron microprobe analyses of run products

l0 Kbar Ru Products

Ru NuberPhase

3r3 308 308 3r7 317 3r5 3r5 3r6 316Glass Olivine GIa6s Olivine Glass Olivine Glass OlivioeGlass

. 8 3

r3 . 37

9 . 6 2

16.7 |

1 0 . 0 3

1 . 0 8

. 0 9

. 0 89 9 . r 3

/ f . o

1oox

. 8 5

1 3 . 8 4

9 . 6 6

1 5 . 9 4

. 1 6

1 0 . 4 0

1 . 1 0

. 0 599.82

91?

Tio2

A12O3

Cr203

FeO

ItSO

ltn0

Nio

Ca0

Na2O

Pzo5

("0Sfn

100 x lr8l(H8+fe)

Ilt ft of Phase

_ . 8 9 . 92 r . 0 4

1 6 . 0 4 . 1 4

- . 1 7

9.45 7r .24

r r .65 46 .68

- . 1 5

l 1 . 7 0 . 3 9

1 . 1 7

. 1 0

.40 .o2100.40 98 .E0

68.7 88 . r

8s%' 1sI,

. 0 0 1 4 . 2 t . o 4 1 5 . 2 0 . o 4

. 1 5 - . 1 8 - . 1 9

9 . 2 8 9 . 6 E 9 . 4 4 9 . 1 4 1 0 . 5 9

49.45 15 .02 49 .07 12 .85 47 .39

. 1 3 . 1 5 . t 6 . 1 5 . 1 3

. 2 4 - . 2 2 - . 2 3

. 3 1 1 0 . 5 9 . 3 7 1 r . 3 7 . 4 r

- 1 . 1 5 - r . 1 6

- . 0 6 - . l l

. o 7 . 0 5 . 0 2 . 0 5 . 0 1100. 63 99 .76 100 .28 100.69 99 .62

9 0 . s 7 3 . 4 9 0 . 3 7 0 . 2 8 8 . 9

3% 94% 61, 87% 12?

PhaseRun Nwber 323 323 329 329 329

.07 76 .84

. t 6 . 1 5

1 0 . 6 8 9 . 0 7

4 7 . 8 3 9 . 4 7

. 1 5 . 1 3

. 4 J

.38 12.20

- r . 4 2

33r 331 331Glass Olivine CPX

333 333 333Glass Olivine CPXGlass Olivire Glass olivine CpX

Ti02

Ar203

ct2o3

Fe0

ugo

ltnO

Ni0

CaO

Naro

P^0-Z J

Kzo

Su

I . 0 7 - . 7 4 1 . 1 8 - . 8 9 l . 4 1

16 .27

9 . 2 2

9 . 9 7

12.30

I . 3 1

. 0 7

'.06

98.62

. 0 2

100.00

E 8 . 9

18%

l- 08

. 10 5 .11 17 .20 . r 3 5 .87 17 .92 . 20 5 .66

. 13 . 20 - . r 4 . 17 . 27

1 r . 6E 7 .77 9 .43 12 .55 8 .17 rO .O l 15 .98 7 .4s

45.90 16.40 8.92 46.13 15.44 8.62 42.69 15.33

. l I . r 7 . 1 8 . 2 0 . 2 0 . 2 0 . 2 3 . 2 1

.20

.39 19.24 17.72 .31 19.77 rr .2r .70 2r .o4

- . 29 1 .71 - . 31 2 .O2 - . 3 r

l lg-nmber 65.7

Ht fl, of Phase 82%

. 0 1 . 0 0

99.28 99 .96 99 .03

. 0 1

99.10 99.97

86 .7 77 . r

20x s%

. 0 6 . 0 3

100.44 99 .83 r00 .63

60.6 82 .6 78 .6

322 loX, r81

65 . 0

80x

E7 . 7

20\

79 . 0

r?

-05

99.25

62.8

7sx

pockets of glass that are produced during melting ofperidotites.

Melting of the sub-oceanic mantle

These high-pressure phase diagrams can be used forevaluating melting in the mantle and the production ofprimary oceanic basalts. We emphasize that these arelimitations to the application of these pseudo-liquidusphase diagrams. For example, all of the phase relationshere are for anhydrous systems and are strictly applicable

only to melt production in an anhydrous environment.For oceanic basalts, this is probably not a major concernbecause of the low water contents of oceanic basalts(Moore, 1970; Delaney et al., 1978). For very smallincrements of melting of natural peridotites, where acces-sory phases such as phlogopite, apatite, amphibole, orcarbonate may be present and the influence of CO2 andH2O may be important, it is likely that these phasediagrams do not adequately portray the true variations inliquid compositions. Finally, it should be noted that there

ELTHON AND SCARFE: EQUILIBRIA OF HIGH-MAGNESIA BASALT

Table 3. (cont.)

15 Kbar Ru Products

Ru l{uber 320Phase Gless

319 319Class olivitre

325 326 326olivine Gless Olivi[e

330 330 330 330GI6ss oliviE CPX OPX

3r8Glass

3r8 325olivine GIaa6

!26cPx

si02 47,.93

Tio2 .85

r 3 . 7 9

9 . 9 8

r 6 . 8 8

4r203

c.203

Fe0

xro

l'lo

Ni0

Ca0

Na20

P^0-

Kzo

Su

!18-trMber

l{r x of Phase

48.31

. 9 2

r 5 . 0 6

9 . 7 t

13 .50

r 1 . 0 8

l . l 7

.06

.03

99 9r

7 l . l

90x

47.ar 40.99 4 0 . 1 0 4 8 , 5 7 3 9 . 8 2 4 8 . 6 5 3 9 . 8 7 5 2 . 9 0 4 6 . 5 ) 4 0 . 1 4 ) J . o J ) q ' u r

.95

14.72

9 . 9 7

13.92

09

.09

49.20

. 1 3

44

. 3 r

00

99 72

9 r . 2

8X

.09

. 1 6

9 . 1 6

49.00

, l J

. 3 4

.05

99.47

9 0 . 5

. 9 5

9 1 7

12 -59

1 1 2 9

r . 23

, 1 I

.01

100-14

69.7

88X

. 1 0

- t 2

9 . 4 7

48.93

.40

-34

. 9 5

1 5 . 8 5

9 . 5 7

12.19

. 1 5

I 1 . 6 5

l . l E

.09

.09

r0 .56

48.65

. 1 7

. 3 1

4 . 7 1

. 1 6

6 . t l

l 9 7 8

1 5 . 6 9

. 2 0

1 . 1 9

15 95

.o7

9 . 4 8

r 1 . 9 9

. 1 1

1 r . 7 5

1 . 3 1

.20 .62 -22

. 1 1 4 . 9 3 3 . 1 9

.08 . l0 .40

1 3 . 4 8 1 . 3 2 1 1 . 7 9

46.22 18 .19 21 .79

. 1 6 . r 9 . 2 r

-29 15 .91 2 .14

.00 .20 .06

9.95 10 92

1 . 1 2 r . 2 3

. 0 9 . 1 0

.04 .05

100.62 99 67

7 5 . r 7 r . 3

1001 927

.01 .05 .01

99.34 r00 -26 r00 .14

90 2 69 .4 89 . r

r2I 8s1 13X

____L .02 .00

1or .04 100.43 100.58 r00 .49

8 3 . 7 5 9 3 8 5 . 9 8 r . 6

2',, 831 r24 2l

1 0 0 . 4 1

8 0 . t

31

20 Kbar Rus Products

Ru Nuber 306Phase Class

307 307 3104Class olivine class

310A 309 309oPx Glass ol iv iae

3r0AOl iv ine

309 309cPx oPx

3104cPx

sio2 47 65 47-42

T i02 .8 r ' 88

AL203 13.42 14.31

3 9 . 9 8 4 8 . 0 2 3 9 . 5 3 5 0 . 5 4 54.53 48 . l l 40 .U)

.29 97 . l0

4 61 14 .55 .15

25 21

I 1 2 9 4 0 1 1 . 6 9

28.26 14 50 47.09

. r 8 1 6 . 1 7

3 10 11 .08 .30

. 0 8 r . 2 7 . 0 0

ct2o3

Fe0

Mto

tln0

I io

Ca0

ila20

Pzos

K20

Su

llt-Duber

9t 1 of Phase

9 . 5 1 9 . 5 5

17.09 14 .85

9 - 7 1 1 0 . 6 7

1 . 0 6 r . 2 3

- . t 0

.05 .04

9 9 . 5 2 9 9 . 5 5

76.2 73-3

1001 94%

l 6

r6

10 20

47 .99

. 1 2

35

9 2

1 4 . 4 8

. l l

9 . 3 5

. 1 3

10 98

1 -32

. 1 2

14

11 .20

. t 5

33

85

6 . 8 6

. l E

7 . 1 2

1 8 . 0 5

-20

1 5 . 0 7

4v.6J )J . Z f ,

. 7 5 . 2 0

7 . 5 9 4 . 8 6

.29 .24

7.2a 8-65

1 8 . 7 r 2 8 . 8 6

. 2 0 . 1 9

15.16 3 .44

. 0 1

9 9 . 4 3

86 1

.02

99.48

8 8 3

. 0 2

9 9 . 3 6

8 9 . 3

67"

. 0 1

9 9 . 9 9

7 3 - 7

92X

.00

9 9 . 1 3

8 r . 9

t race

.03 00 .00 .00

ro0 38 99.55 100-20 99.42

1 3 . 3 8 7 . 8 a 2 . r 8 6 . 6

901 5X trace 51

is a finite, but poorly defined, width of the phase equilibri-um boundaries in the diagrams as a consequence ofanalytical uncertainty and variations in the compositionof the basalts (such as Fe/Mg) that are not compensatedfor in the formulation of the end member minerals.

From these pseudo-liquidus phase diagrams it is possi-ble to obtain information on the nature of primary meltsand the effects of crystallization at high and low pres-sures. Information on the types of residual ultramaficrocks in equilibrium with primary magmas can also beobtained from pseudo-liquidus phase diagrams. In thismanner, melts projecting at the isobaric pseudo-invariantpoints would be in equilibrium with an aluminous (garnet/spineUplagioclase) lherzolite whereas liquids projecting

along an OL + OPX + SP + LIQ isobaric pseudo-univariant curve would be in equilibrium with a spinelharzburgite residuum. In principle, the study of theseresidual ultramafic rocks will provide critical informationon the nature of primary magmas.

Information on residual ultramafic mantle rocks relatedto oceanic basalt petrogenesis is obtained from ophiolitesand oceanic fracture zones. Bonatti and Hamlyn (1981)conclude, in a survey of oceanic ultramafic rocks, thatharzburgite is slightly more common than lherzolite in therift zones of the Indian Ocean Ridge (Dmitriev, 1969) andwithin fracture zones of the equatorial Atlantic Ocean(Bonatti et al., 1970; Melson and Thompson, l97l). Insome of the major fracture zones of the western Indian


Table 3. (cont.)

25 Kbar Bun Products

Run NuberPhase

304 328 164 164 164Glass Glass Glass olivine CPX

t o 4 l o 4

OPX GARNET

si02

Tio2

Ar203

CtrO,

Fe0

Mgo

Ca0

Na20

Pzo5

("obM

l{g-nuber

. 2 1

9 . 9 4

1 . 0 4

. 0 19 9 . 9 4

r 0 . 0 6

1 . 0 7

. 0 69 9 . 4 L

7 6 . r

. 1 7

1 0 . 8 7

o <

. 0 61 0 1 . 0 9

7 5 . 0

l 6 . 0 1

48.62

. 8 9 . 8 5 . 9 0 - . 6 2 . 1 5 2 . 0 8

1 3 . 1 6 1 3 . 3 8 1 3 . 8 2

, 2 0

. 0 7 7 . 3 9 6 . 6 4 1 7 . 6 7

. t 2 . 2 1 - . 7 9

9 . 4 r 9 . 3 7 9 . 8 5 1 0 . 1 0 5 . 8 1 6 . 4 3 9 . 2 7

r7 .o5 16 .72 76 .51 48 .31 18 .31 30 .53 17 .10

. 1 4 . 1 r . 2 2t n O

Ni0

. r 7

.32 2 . 4 3 8 . 7 2

. t 2 . 7 3. 0 0

. 0 199.24

8 9 . 5

. 00 . 02 . o298 .94 99 . 72 97 .8 r

84 .9 89 . 4 76 .7

Ocean (Engel and Fisher, 1975; Bonatti and Hamlyn,1978; Hamlyn and Bonatti, 1980) lherzolites are morecommon than harzburgites. In the major ophiolite bodies,residual mantle harzburgite is much more abundant thanlherzolite (Samail: Boudier and Coleman, l98l; Bay ofIslands: Casey et al., l98l; Papua-New Guinea: Davies,l97l; Troodos: Wilson 1959; Vourinos: Jackson e/ a/..1975) although residual mantle lherzolites are reportedfrom some alpine ophiolites.

Although published mineral chemistry data on oceanicor ophiolitic residual mantle rocks are limited, these dataindicate that there are only minor diferences in themineral compositions in orthopyroxenes and olivinesfrom harzburgites and lherzolites (Hamlyn and Bonatti,1980; H. Dick, pers. comm.). During progressive meltingof mantle assemblages, the Mg/(Mg+Fe) of residual min-erals should gradually increase (e.g., Jaques and Green,l9E0). Because of the similarity of olivine and orthopyr-oxene compositions in harzburgites and lherzolites, it islikely that many of the harzburgites are the residuum of apartial melting event in which the degree of meltingprogressed just slightly beyond the disappearance of thefinal residual clinopyroxene crystals.

The above data on residual mantle rocks from ophio-lites and oceanic fracture zones would suggest that arange of melting conditions probably occurs within thesuboceanic mantle, in some cases leaving a lherzoliteresiduum and in other instances, a harzburgite residuum.For the genesis of oceanic basalts, a continuum of modelsranging from a lherzolite to a harzburgite residuum shouldbe considered.

Pressure of origin for primary oceanic basalts

O'Hara (1968) and Walker et al. (1979) have shown thatmost oceanic basalts cluster near the I atm olivine *plagioclase * clinopyroxene + liquid pseudo-univariantcurve and presumably have been afected by low-pres-sure crystal fractionation. As noted above, however, themain controversy in the origin of primary oceanic basaltsconcerns those rare relatively "primitive" basalts with>9.5% MgO, which are interpreted to have undergonethe least amount of differentiation prior to eruption.

In the discussion that follows it is assumed that themajority of oceanic basalts, those with <9.5% MgO, areultimately derived by crystallization and magma mixingprocesses from parental liquids whose major elementcompositions are very similar to these "primitive" ocean-ic basalts shown in Figures 2 and 3. This type of assump-tion, that the basaltic liquids with the highest Mgocontents (or Mg number) are parental to the more evolvedbasaltic liquids with lower MgO contents (or Mg number),is very commonly made in modeling crystal fractionationand other evolutionary processes in oceanic basalt petro-genesis. Because we believe that this assumption is validin most suites of oceanic basalts, we initially discuss ourdata within this context. In the final section, we note thatthis assumption may not always be valid, but that someoceanic basalts might be derived from "primitive" mag-mas that are distinctly diferent from those with >9.5%MgO that are shown in Figures 2 and 3.

Figure 4, which is a plot of the MgO content of basalticglasses versus the amount of normative silica (data from


OLIVINE SILICAFig. 2. PseudoJiquidus phase diagram projected from clinopyroxene. See text for discussion. The apices ofthe triangle are: (top)

AN6sOLroSIL2s (left) OLeoAN,sSILzs (right) SIL75ANr5OLr6. Details of the projection method are given in Elthon (1983). I atm phase

equilibria are from Walker et al. (1979).

PLAG.

C l i n o P y 7 6 1 9 1 " CPX

OLIVINEFig. 3. Pseudo-liquidus phase diagram projected from anorthite.

cPX6oOLroSIL3o (left) OL6oCPXroSIL3o (right) sILsocPXloOLro. I

slLlcASee text for discussion. The apices of the triangle are: (top)

atm phase equilibria are from Walker et al. (1979).

l 0 ELTHON AND SCARFE: EQUILIBRIA OF HIGH.MAGNESIA BASALT

, . , ' . ' i

. i i } - M 6 l n c l u s l e r o f- - . : i 1 - \ \ o c e a n i c b a s a l l g

40 44 4A 52 56 60

% S i l i ca i n t h€ o l / cpx / s i l p l ane

Fig. 4. MgO content of oceanic basalt glasses versus thepercentage of normative silica in the olivine-<linopyroxene-silica plane. Sources of data are listed in the text. Note thedecrease of MgO with increasing normative silica.

In order to avoid biasing the data to those locations that areheavily represented in the listing by Melson et al. (1977), onlyone sample per location is included on these plots andprojections unless there was a significant variation incompositions at a site.

Melson e t al., 1977 ; Bryan and Moore, 1977 ; W ood e t al.,1979 and others) within the olivine-clinopyroxenFsilicaplane, shows that the most "primitive" oceanic basalts(i.e., those basalts with >9.5Vo MgO) lie at the silica-poorend ofthe basalt cluster. The MgO content decreases andthe normative silica content increases during differentia-tion. For this reason, it is critical to look at the silica-poor(least diferentiated) end ofthe basalt cluster when evalu-ating the compositions of primary oceanic basalts.

The processes of melt production, magma ascent,crystallization diferentiation, and magma mixing proba-bly operate in a very complex manner to produce efectsthat are, at present, difficult to distinguish from eachother in oceanic basalt petrogenesis. In order to evaluatethe ditrerent models for the genesis of primary oceanicbasalts, first consider melting of the mantle at - l0 kbar toproduce primary oceanic basalts as suggested by Kushiro(1973), Fujii and Kushiro (1977), Presnall et al. (1979) andFujii and Bougault (1983). The primary melt in equilibri-um with oL + oPX + cPX + SP/PLAG at l0kbarwouldproject at or near the l0 kbar pseudo-invariant pointshown in Figures 5 and 6.

A primary magma generated at l0 kbar and subsequent-

1 2

1 1

oo 8

7

6

5

< 9 . 5 %

> 9 . 5 %MsoMso

OLIVINE SILICAFig. 5. Projection of oceanic basalt glasses from anorthite onto a part of the olivine-clinopyroxene-silica plane. phase equilibria

boundaries and triangle apices are from Fig. 3.

'$ftl)-

cPx

. B a s a l t sa B a s a l t s


OLIVINE SIL ICAFig. 6. Projection of oceanic basalt glasses from clinopyroxene onto the olivine-plagioclase-silica plane. Phase equilibria

boundaries and triangle apices are from Fig. 2.

l l

ly transported to a near-surface magma chamber willundergo olivine fractionation followed by olivine + pla-gioclase, olivine + plagioclase * clinopyroxene, andfinally olivine + plagioclase * clinopyroxene * orthopyr-oxene as the liquid composition changes from the l0-kbarpseudoinvariant point to points a and b (Fig. 5) duringcrystallization. Even with open-system fractionation(O'Hara, 1977; Rhodes and Dungan, 1979) in which thecrustal magma chamber is periodically refilled with analiquot of primary magma, periodic injection of the pri-mary magma generated at l0 kbar into the fractionatingmagma chamber (a-b) will produce mixed basaltic compo-sitions that project to the silica-rich (right-hand) side ofthe olivine subtraction line drawn from the l0-kbar pseu-do-invariant point (Fig. 5). Similarly, fractionation at highpressures within the mantle will produce basalts that liefurther to the silica-rich side of this olivine subtractionline because combinations of the fractionating phases(OL, OPX, CPX, GAR, SP) lie to the silica-poor side ofthe pseudo-invariant point.

More advanced melting of the mantle at l0 kbar toleave a spinel harzburgite residuum rather than the spinellherzolite residuum model discussed above would resultin even higher normative silica in the derivative basaltsthan in the lherzolite residuum model. If melting of themantle at l0 kbar proceeds until a spinel harzburgite

residuum remains, producing a primary liquid such as"2" (Fie.S), the basalts derived from"Z" would crystal-lize orthopyroxene before clinopyroxene and would beunlike most oceanic basalts.

It is unlikely that melting at such low pressures (10 kbaror less) is a major process in the production of "primi-tive" oceanic basalts with >9.5Vo MgO because the vastmajority of these basalts lie significantly to the silica poor(lefrhand) side of the olivine subtraction line drawn fromthe l0 kbar pseudo-invariant point. If it is assumed thatthose basalts with <9.5% MgO are derived from the"primitive" basalts shown in Figs. 5 and 6 with >9.5%MgO by crystallization and mixing processes, then melt-ing at =10 kbar is considered less preferable than modelsinvolving primary magma generation at higher pressures(15 to 25 kbar).

O'Hara (1968) has suggested that the primary magmasfrom which most oceanic basalts are ultimately derivedare high-MgO basalts generated at 25 to 30 kbar. Theexperimental data on NT-23, which has olivine + ortho-pyroxene + clinopyroxene + garnet on the liquidus at 25kbar, agrees with O'Hara's proposal. In fact, if thesimplest model involving only olivine fractionation be-tween genesis and eruption is considered, the majority ofoceanic basalts with >9 .5Vo MgO lie on or near the olivinesubtraction line drawn from the 25 kbar pseudo-invariant

PLAG

. Basa l t s w i t h < 9 .5% MgO6Basa l t s w i t h >9 .5% MgO

t 2 ELTHON AND SCARFE: EQUILIBRIA OF HIGH-MAGNESIA BASALT

point. Open-system fractionation within the crust, inwhich pulses of a high-MgO basalt similar to the composi-tion of the 25 kbar pseudo-invariant point glass areperiodically injected into a magma chamber (O'Hara,1977) will generate complex mixing-fractionation cyclessimilar to those shown in Fig. 5 that result in a spectrumof oceanic basalt compositions that cluster near the I atmolivine + plagioclase * clinopyroxene cotectics. Anadditional process that needs to be considered, the highpressure fractionation of wehrlite or clinopyroxenite fromprimary oceanic basalts (Bender et al.,1978;Bence et al.,1979; Elthon et al., 1982), will also result in increasingnormative silica in the derivative basalts, driving thecompositions of basalts towards the right-hand (silica-rich) side of the basalt cluster. Thus, magma mixing orcrystal fractionation processes will clearly have the ten-dency to produce evolved basalts whose compositionsare displaced toward normative silica enrichment relativeto the primary basalt from which they are derived. Inexamining the petrogenesis of a suite of cogenetic oceanicbasalts derived from a primary liquid by these processes,those basalts that lie at the left-hand (silica-poor) end ofagiven suite would be the most indicative of the conditionsof origin of the primary magma. We interpret the observa-tion that most of the "primitive" oceanic basalts with>9.5% MgO are close to olivine-controlled liquid lines ofdescent from the 15 to 25 kbar isobaric pseudo-invariantpoints to indicate that those basalts have been derivedfrom primary magmas generated within this pressureinterval.

Other types of evidence from oceanic rocks may indi-cate that the final pressure of equilibration for someprimary oceanic basalts with the upper mantle may be at>20 kbar pressure. Hamlyn and Bonatti (1980) describeequant pyroxene + spinel clusters in the Owen F. Z.ultramafics that, because of the high proportion of spinelto pyroxene, they attributed to the breakdown of garnetvia the reaction: olivine + garnet €r pyroxene * spinel.The presence of garnet in these peridotites would imply apressure of =25 kbar. Further evidence on the equilibra-tion of some of the Owen F. Z. ultramafics with a primarymagma at high pressures is indicated by the comparative-ly high abundances of Al2O3 in orthopyroxenes from theharzburgites. Elthon (1981) determined the melting rela-tionships of two oceanic peridotites, one from the OwenF. Z. and the other from the Atlantic Ocean, at highpressure and found that the Al2O3 content ofboth ortho-pyroxenes and clinopyroxenes decreases with increasingdegree ofmelting at l0 and 20 kbar. At any percentage ofmelting, the aluminum content of both clinopyroxene andorthopyroxene increases with increasing pressure. Ortho-pyroxenes in the Owen F. Z. harzburgite contain 4.8-5.2Vo Al2O1 The data of Elthon (1981 and in prep.) andJaques and Green (1980) indicate that at <20 kbar, -3.5VoAl2O3 is found in harzburgite orthopyroxenes, comparedto the 5.0+0.2Vo Al2O3 in the Owens F. Z. harzburgiteorthopyroxenes. Because these harzburgites contain 50

to 100 times as much orthopyroxene as spinel, it isunlikely that the Al2O3 contents of the harzburgite ortho-pyroxenes have been significantly changed due to subso-lidus equilibration with the spinels, even though thespinels may have changed significantly. These featuressuggest that the final equilibration of a melt with theseorthopyroxenes was at pressures greater than 20 kbar.

The olivine composition in equilibrium with the pseu-do-invariant point liquid at 25 kbar (run 164) is Fose.5which falls within the range (Fose.e-e6.a) reported from theOwen Fracture Zone ultramafics (Hamlyn and Bonatti,1980). This agreement between the liquidus olivine com-positions in our experiments at 25 kbar and the olivinecompositions in the Owen F. Z. ultramafics, when cou-pled with the Al2O3 contents of orthopyroxenes in harz-burgites, supports our inferences based on the projectionof oceanic basalts onto high-pressure pseudo-liquidusphase diagrams that many of the "primitive" oceanicbasalts are ultimately derived from primary high-MgObasalts generated at pressures >15 kbar.

Concluding remarks

The model described above, in which prirnary oceanicbasalts are generated by partial melting at 15 to 25 kbar, isin agreement with previous models by O'Hara (1968),Green et al. (1979), and Stolper (1980), who propose thatmost oceanic basalts are derived from primary magmasthat are generated at 15 to 30 kbar pressure. Theseprevious pressure estimates are derived, for the mostpart, by assuming that simple olivine fractionation hasbeen the process that has occurred between melting anderuption. If it is assumed that only olivine fractionationoccurs between melting and eruption, olivine additionlines drawn from primitive oceanic basalts (>9.5%MgO)in an effort to reconstruct the composition of the primarymagma prior to olivine fractionation result in intersec-tions with the isobaric pseudoinvariant points at 15 to 25kbar (see Figs. 5 and 6). This indicates the generalagreement between the models of O'Hara, Green et al.,Stolper, and Elthon and Scarfe, in which primary oceanicbasalts are generated between 15 and 25 kbar.

We consider it likely that the evolutionary processesinvolved in oceanic basalt genesis are considerably morecomplex than these simple olivine fractionation modelsmay imply. For example, evidence on the role of high-pressure crystal fractionation in the evolution of oceanicbasalts is obtained from Elthon et al. (1982) who interpretthe basal ultramafic cumulates from the North ArmMountain massif of the Bay of Islands to have formed byhigh-pressure crystallization. On the basis ofother ophio-Iite massifs that have similar cumulate ultramafic mineralassemblages and mineral compositions, as well as datafrom oceanic basalts, Elthon et al. propose that high-pressure crystal fractionation may be a common processin the evolution of oceanic basalts.

Mineral assemblages within the ultramafic cumulates ofNorth Arm Mountain indicate that basaltic liquids have a

ELTHON AND SCARFE: EOUILIBRIA OF HIGH.MAGNESIA BASALT l 3

strong tendency to fractionate to lower pressure isobaricpseudoinvariant points during magma ascent (Elthon etal.,1982). As an example, a primary magma generated inequilibrium with a harzburgite residuum at 20 kbar thatsubsequently undergoes high-pressure crystallization (du-nite --+ wehrlite + lherzolite) at 10 kbar until the liquid isin equilibrium with OL + OPX + CPX + SP wouldproject at the l0 kbar pseudo-invariant point and mightappear (erroneously) to represent a primary magma gen-erated at 10 kbar. If high-pressure fractionation of thistype occurs within the sub-oceanic mantle, estimates ofthe pressure of melting in the mantle based on pseudo-liquidus phase diagrams must represent minimum pres-sures. It is possible that the oceanic basalts studiedexperimentally by Kushiro (1973) and Fujii and Kushiro(1977), which crystallize a comparatively Fe-rich plagio-clase lherzolite assemblage at 7 .5 to 8 kbar, have experi-enced a similar high-pressure crystallization history inwhich the liquid has been modified to the 7.5 to 8 kbarpseudo-invariant point.

It is also interesting to note that most "primitive"oceanic basalts (>9.5% MgO) lie along or adjacent to thel0 kbar OL + CPX + SP + LIQ isobaric pseudounivar-iant curve (see Figs. 5 and 6). One possible cause of thisalignment of "primitive" oceanic basalts along the iso-baric pseudo-univariant curve is that these basalts mayhave undergone crystal fractionation at high pressures,with those basalts that project closest to the isobaricpseudo-invariant point having undergone the greatestdegree of crystal fractionation.

On the basis of the above discussion, we believe thatoceanic basalts may commonly undergo high-pressurecrystallization prior to eruption. As briefly noted above,the efiect of high-pressure crystal fractionation on thesepseudo-liquidus phase diagrams is to drive residual liq-uids towards silica enrichmeflt, i.e., towards the right onFigures 5 and 6. Because of the difficulty in quantitativelyevaluating the magnitude of the effect that high-pressurefractionation has had on individual "primitive" oceanicbasalts, the extent of enrichment in normative silica as aconsequence of this process is uncertain.

In evaluating overall scenarios for the petrogenesis ofoceanic basalts, it is considered likely that a combinationof high pressure fractionation and/or magma mixing pro-cesses may be important, even in those primitive basaltswith >9.5Vo MgO. Recognizing that both of these pro-cesses tend to produce an increase in normative silicawithin the melt, we prefer a model for the origin ofprimary oceanic basalts at 15 to 25 kbar: In this model,these primary magmas are modified by a combination ofhigh-pressure crystal fractionation and magma mixing toproduce the more abundant oceanic basalts with 7.5 to95%MgO.

The discussion above has been based on the assump-tion that the majority of oceanic basalts with <9.5Vo MgOare diferentiates of the "primitive" basalts with >9.5%MgO, which are shown in Figures 5 and 6. It should be

noted that for many of these basalts, which have equili-

brated to the low-pressure cotectics, this lineage is not

readily constrained. In fact, it is possible that basalts thatproject to the center of the group with <9.5Vo MgO are

derived by the crystallization of a primary magma that, byvirtue of having been produced at 5 to 10 kbar was

chemically different from those basalts with >9.5% MgO

shown in Figures 5 and 6. Until the processes of crystalli-zation and magma mixing are more adequately quantified,

it will not be possible to unambiguously ascertain thepressure of origin for the primary magmas related to

many of the evolved oceanic basalts with <9.5% MgO.

For this reason, we do not propose a lower-most pressure

limitation for the production of primary oceanic basalts.

Rather, we suggest that if primary oceanic basalts areproduced at slO kbar, these basalts would not be like

most of those basalts with>9.5Vo MgO that are recovered

from the present-day oceanic basins.

Acknowledgments

The experimental work was performed at the GeophysicalLaboratory of the Carnegie Institution of Washington while theauthors were a predoctoral fellow (DE) and guest investigator(CMS). Dr. H. S. Yoder Jr. is thanked for his continued interestand for giving us the opportunity to work at the GeophysicalLaboratory. The Carnegie Institution of Washington is thankedfor financial support.

Most of the microprobe analyses and follow-up studies wereperformed while Elthon was a Research Associate at the Smith-sonian Institution in Washington, D. C., and Scade had returnedto the University of Alberta. W. G. Melson, E. Jarosewich, andJ. Nelen are thanked for their assistance with the analyticalaspects ofthis project and for discussions concerning the results.

Scarfe acknowledges support from Canadian NsERc grantA8394. Elthon acknowledges support from the United StatesNational Science Foundation through grants EARE0-26445 andOCE 8l-21232. Drs. T. Fujii, D. Walker, and an anonymousreviewer are thanked for their critical reviews.

References

Albee, A. L. and Ray, L. (1970) Correction factors for electronmicroprobe microanalysis of silicates, oxides, carbonates,phosphates, and sulfates. Analytical Chemistry, 42, 1408-1414.

Bence, A. E. and Albee, A. (1968) Empirical colrection factorsfor the electron microanalysis of silicates and oxides. Journalof Geology. 76, 382-401.

Bence, A. E., Baylis, D. M., Bender, J. F. and Grove, T. L.(1979) Controls on the major and minor element chemistry ofmidocean ridge basalts and glasses. In M. Talwani, C. G.Harrison and D. E. Hayes, Eds., Deep Drilling Results in theAtlantic Ocean: Oceanic Crust, p. 331-341. American Geo-physical Union, Washington, D. C.

Bender, J. F., Hodges, F. N. and Bence, A. E. (1978) Petrogene-sis of basalts from the project FAMous area: Experimentalstudy from 0 to 15 kbars. Earth and Planetary Science Letters,41,277-302.

Bonatti, E. and Hamlyn, P. R. (1978) Mantle uplifted block in theWestern Indian Ocean. Science. 201.249-251.

Bonatti, E. and Hamlyn, P. R. (1981) Oceanic ultramafic rocks.

l 4 ELTHON AND SCARFE: EOUILIBRIA OF HIGH.MAGNESIA BASALT

In C. Emiliani, Ed., The Sea, vol. 7, p. 241-283. Wiley, NewYork.

Bonatti, E., Honnorez, J. and Ferrara, G. (1970) Equatorial Mid_Atlantic Ridge: Petrologic and Sr-isotopic evidence for analpine-type rock assemblage. Earth and Planetary ScienceLetters. 9. 247-256.

Boudier, F. and Coleman, R. G. (1981) Cross section through theperidotite in the Samail Ophiolite, southeastern Oman Moun-tains. Journal of Geophysical Research, 86,2573-2592.

Boyd, F. R. and England, J. L. (1960) Apparatus for phaseequilibrium measurements at pressures up to 50 kilobars andtemperatures up to 1750'C. Journal of Geophysical Research,65, 741-748.

Bryan, W. B., Finger, L. W. and Chayes, F. (1969) Estimatingproportions in petrographic mixing equations by least-squaresapproximation. Science, 163, 926-927.

Bryan, W. B. and Moore, J. G. (1977) Compositional variation ofyoung basalts in the Mid-Atlantic Ridge Rift valley near 36"49,N. Geological Society of America Bulletin, 88, 556-570.

Casey, J. F., Dewey, J. F., Fox, P. J., Karson, J. A. andRosencrantz, E. (1981) Heterogeneous nature ofoceanic crustand upper mantle: A perspective from the Bay of Islandsophiolite complex. In C. Emiliani, Ed., The Sea, vol. 7,p. 305-338. Wiley, New York.

Clarke, D. B. (1970) Tertiary basalts of Baffin Bay: possibleprimary magma from the mantle. Contributions to Mineralogyand Petrology, 25, 203-224.

Davies, H. L. (1971) Peridotite-gabbro-basalt complex in East-ern Papua: An overthrust plate of oceanic mantle and crust.Australian Bureau of Mines Resources Bulletin, 128, l-48.

Delaney, J. R., Muenow, D. W. and Graham, D. G. (1978)Abundance and distribution of water, carbon, and sulfur in theglassy rims of submarine pillow lavas. Geochimica et Cosmo-chimica Acta, 42, 581-594.

Dmitriev, L. V. (1969) Origin of the ultrabasic rocks in the riftzones ofthe Indian Ocean Ridge. Geochemistry International,6, 923-931.

Echeverria, L. M. (1980) Tertiary or Mesozoic komatiites fromGorgona Island, Columbia: field relations and geochemistry.Contributions to Mineralogy and Petrology, 73,253-266.

Engel, C. G. and Fisher, R. L. (1975) Granitic to ultramafic rockcomplexes of the Indian Ocean ridge system, Western lndianOcean. Geological Society of America Bulletin, 86, 1553-1578.

Elthon, D. (1979) High magnesia liquids as the parental magmafor ocean floor basalts. Nature. 278. 514-518.

Elthon, D. (1980) The Petrology of the Tortuga Ophiolire Com-plex, Southern Chile: Implications for igneous and metamor-phic processes at oceanic spreading centers. Ph. D. thesis,Columbia University, New York.

Elthon, D. (1981) Melting relationships of two oceanic perido-tites at high pressures. Abstracts volume for AGU ChapmanConference on the Generation of the Oceanic Lithosphere.

Elthon, D. (1983) Isomolar and Isostructural pseudoJiquidusphase diagrams for oceanic basalts. American Mineralogist,68, 506-51 l .

Elthon, D., Casey, J. and Komor, S. (1982) Minerdl chemistry ofultramafic cumulates from the North Arm Mountain massif ofthe Bay of Islands ophiolite: Evidence for high-pressurecrystal fractionation of oceanic basalts. Journal of Geophysi-cal Research, 87, 8717-8734.

Elthon, D. and Scarfe, C. M. (1980) High-pressure phase equilib-ria of a high-magnesia basalt: Implications for the origin of

mid-ocean ridge basalts. Carnegie Institution of WashingtonYearbook. 79. 277-281.

Francis, D. M. and Hynes, A. J. (1979) Komatiite-derivedtholeiites in the proterozoic of New Quebec. Earth andPlanetary Science Letters, 44, 473-481.

Fujii, T. and Bougault, H (1983) Melting Relations of a magne-sian abyssal tholeiite and the origin of Monss. Earth andPlanetary Science Letters, 62, 283--295.

Fujii, T. and Kushiro, l. (197'l) Melting relations and viscosity ofan abyssal tholeiite. Carnegie Institution of Washington Year-book.76. 461-465.

Green, D. H., Hibberson, W. D. and Jacques, A. L. (1979)Petrogenesis of mid-ocean ridge basalts. In M. W. McElhin-ney, Ed., The Earth: Its origin, structure, and evolution, p.265-299. Academic Press, London.

Green, D. H. and Ringwood, A. E. (1967) The genesis of basalticmagmas. Contributions to Mineralogy and Petrology, 15, 103-190.

Hadidiacos, C. (1972) Temperature controller for high-pressureapparatus. Carnegie Institution of Washington Yearbook, 71,620-622.

Hamlyn, P. R. and Bonatti, E. (1980) Petrology of mantle-derived ultramafics from the Owen Fracture Zone, northwestIndian Ocean: Implications for the nature ofthe oceanic uppermantle. Earth and Planetary Science Letters, 48,65-79.

Harrison, W. J. (1981) Partitioning of REE between minerals andcoexisting melts during partial melting of a garnet lherzolite.American Mineralogist, 66, 242-259.

Hoover, J. D. and Presnall, D. C. (1981) Partial melting ofsimplified lherzolite in the system CaO-MgO-AlzO3-SiO2-Na2O from I ATM to 20 kb. Abstract volume of the AGUChapman Conference on the Generation of the Oceanic Litho-sphere.

Jackson, E. D., Green, H. W. and Moores, E. M. (1975) TheVourinos ophiolite, Greece: cyclic units of lineated cumulatesoverlying harzburgite tectonite. Geological Society of AmericaBulletin. 86. 390-398.

Jaques, A. L. and Green, D. H. (1980) Anhydrous melting ofperidotite at 0-15 Kb pressure and the genesis of tholeiiticbasalts. Contributions to Mineralogy and Petrology, 73,278-3 10 .

Karson, J. A., Elthon, D. and DeLonC, S. E. (1983) Ultramaficintrusions in the Lewis Hills Massif, Bay of Islands ophiolite,Newfoundland: Implications for igneous processes at oceanicfracture zones. Geological Society of America Bulletin, 94,15-29.

Kushiro, l. (1973) Origin of some magmas in oceanic andcircumoceanic regions. Tectonophysics, 17, 2l l-222.

Melson, W. G. (1978) Chemical stratigraphy of Leg 45 basalts:Electron probe analyses of glasses. In Melson, W. G. Rabino-witz, P. D. et al., Initial reports of the Deep Sea DrillingProject, v. 45, p. 507-512. U. S. Government Printing Office,Washington, D. C.

Melson, W. G., Byerly, G. R., Nelen, J. A., O'Hearn, T.,Wright, T. L. and Vallier, '1. (1977) A catalog of the majorelement chemistry of abyssal volcanic glasses. Mineral Sci-ences Investigations: Smithsonian Contributions to the EarthSciences, 19, 31-60.

Melson, W. G. and Thompson, G. (1971) Petrology of a trans-form fault zone and adjacent ridge segments. PhilosophicalTransactions of the Royal Society of London, A268, 423441.

Moore, J. G. (1970) Water content of basalt erupted on the ocean

ELTHON AND SCARFE: EOUILIBRIA OF HIGH.MAGNESIA BASALT l 5

floor. Contributions to Mineralogy and Petrology, 28,272-279.Mysen, B. O. and Kushiro, I. (1977) Compositional variations of

coexisting phases with degree of melting of peridotite in theupper mantle. American Mineralogist, 62, 843-865.

O'Hara, M. J. (1968) The bearing of phase equilibria studies onthe origin and evolution of basic and ultrabasic rocks. EarthScience Reviews. 4. 69-133.

O'Hara, M. J. (1977) Geochemical evolution during fractionalcrystallization of a periodically refilled magma chamber. Na-tu re .266.503-507.

Presnall , D. C., Dixon, J. R., O'Donnell , T. H. and Dixon, S. A.(1979) Generation of mid-ocean ridge tholeiites. Journal ofPetrology, 20,3-15.

Rhodes, J. M. and Dungan, M. A. (1979) The evolution of ocean-floor basaltic magmas. In M. Talwani, C. G. Harrison and D.E. Hayes, Eds., Deep Drilling Results in the Atlantic Ocean:Oceanic Cmst, p. 262-272. American Geophysical Union,Washington, D.C.

Scarfe, C. M., Mysen, B. O. and Rai, C. S. (1979) Invariantmelting behavior of mantle material: Partial melting of twolherzolite nodules. Carnegie Institution of Washington Year-book. 78. 498-501.

Stolper, E. (1980) A phase diagram for mid-ocean ridge basalts:Preliminary results and implications for petrogenesis. Contri-butions to Mineralogy and Petrology, 74, lJ-27.

Thompson, R. N. (1974) Primary basalts and magma genesis. I.Skye, northwest Scotland. Contributions to Mineralogy andPetrology, 45, 317-341.

Tilley, C. E. and Yoder, H. S., Jr. (1964) Pyroxene fractionationin mafic magma at high pressures and its bearing on basaltgenesis. Carnegie Institution of Washington Yearbook, 63,rr4-t21.

Upadhyay, H. D. (1978) Phanerozoic peridotitic and pyroxenitickomatiites from Newfoundland. Science, 202, ll92-1 195.

Walker, D., Shibata, T. and DeLonC, S. E. (1979). Abyssaltholeiites from the Oceanographer Fracture Zone. II. Phaseequilibria and mixing. Contributions to Mineralogy and Petrol-ogy,'70, lll-126.

Wilson, R. A. M. (1959) The geology of the Xeros-Trodoos Area'Geological Survey Department of Cyprus, Memoir I, l-184.

Wood, D. A., Tarney, J., Varet, J., Saunders, A' D., Bougault,

H., Joron, J. L., Treui l , M., and Cann, J. R' (1979) Geochem-istry of basalts drilled in the North Atlantic by IPOD leg 49:

Implications for mantle heterogeneity. Earth and PlanetaryScience Letters. 42. 77-97.

Manuscript received, December 23, 1981;acceptedfor publication, July 5, 1983.

Documents

High-pressure phase equilibria of a high-magnesia basalt and the