<<

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 97, NO. B4, PAGES 4521-4540, APRIL 10, 1992

Depth and Degree of Melting of

CLAUDE HERZBERG

Departmentof GeologicalSciences, Rutgers University,New Brunswick,New Jersey PhysicsInstitute, State Universityof New York, StonyBrook, New York

High pressuremelting experimentsßhove ." v ...... new constraintsto be placedon the depthand degreeof of komatiites. Komatiitesfrom GorgonaIsland were formed by relatively low degreesof pseudoinvariantmelting(< 30 %)involving L + O1 + Opx + Cpx + Gt on the solidusat 40 kbar, about 130 km depth. Munro-typekomatiites were separatedfrom a harzburgiteresidue (L + O1 + Opx) at pressuresthat are poorly constrained,but were probablyaround 50 kbar, about 165 km depth;the degreeof partial melting was <40%. Komatiites from the BarbertonMountain Land were formed by high degrees(-50 %) of pseudoinvariantmelting (L + O1 + Gt + Cpx) of fertile mantleperidotitc in the 80- to 100-kbarrange, about 260- to 330- km depth. Secularvariations in the geochemistryof komatiitescould have formed in response to a reductionin the temperatureand pressureof meltingwith time. The 3.5 Ga Barbertonkomatiites and the 2.7 Ga Munro-typekomatiites could have formedin plumesthat were hotterthan the present-daymantle by 500ø and 30(Y',respectively. When excesstemperatures are this size, melting is deeperand volcanismchanges from basalticto komatiitic. The komatiitesfrom Gorgona Island, which are Mesozoic in age, may be representativeof komatiitesthat are predictedto occur in oceanicplateaus of Cretaceousage throughoutthe Pacific [Storeyet al., 1991].

1. INTRODUCTION range of CaO and A1203contents in the 80- to 160-kbar range. A calibration has been made of the effect of pressure on Komatiites are high MgO volcanic rocks that can be CaO/(CaO + A1203)and MgO in komatiiticliquids formed on roughly explained by high degrees of melting of the solidus, and an examinationhas been made of the effect of peridotitc,typically 50 to 100 % [e.g., Vi.ljoenand Vi.ljoen, FeO. It is now possibleto simulatethe melting of peridotitc 1969; Green, 1972; Brooks and Hart, 1974; Cawthorn and from 1 to 100 % and from 40 to 100 kbar, and the results are Strong, 1974; Nesbitt and Sun, 1976; Bickle et al., 1977; Nisbet comparedto the geochemistryof komatiites. This information and Walker, 1982; Hess, 1990; Miller et al., 1991b]. But this is used to addressthe depthsand degreesof melting, and the interpretationis neitherunique nor withoutits difficulties.Arndt thermal state of the during their formation. [1977]pointed out the problemof formingkomatiites in a single These problems gain special significance now because stage becausebasaltic melts are buoyant and they would komatiites have the potential of being probes into the major segregateand drain from the sourceregion before melting elementgeochemical structure of the mantle. If, for example, reached20 to 30 %. Nisbet and Walker [1982] noted that this the Munro-typekomatiites formed by meltingin excessof 50 %, problem disappearsat high pressuresbecause and melting itself would have had to commenceat lower mantle komatiiticmagmas are expectedto have similardensities. The pressures(> 200 kbar [Miller et al., 1991b]), indicating a problem also disappearsin the modelsof O'Hara et al. [1975] peridotiresource region distributedthroughout portions of the andHowells et al. [1975], whopointed out that initial melting at lower mantleduring Archean times. Similarly, if the Barberton high pressureswas likely to yield in a singlestage komatiites komatiiteswere formedby extensivemelting, the sourceregion insteadof . This high pressureorigin has received must have been very differentfrom mantleperidotitc, possibly considerablesupport from experimentaldata in the 100-kbar similar to piclogitein the transitionzone [Anderson,1989]. It range [Herzberg and O'Hara, 1985; Takahashi, 1986; Wei et is demonstratedin this paperthat this kind of informationcannot al., 1990; Herzberg et al., 1990]. be obtainedbecause the geochemistryofkomatiites is determined Nesbitt e! al. [1979] noted that wide variations exist in largely by phaseequilibrium processes rather than by the bulk contentsof CaO, A1203, and TiO2, and subdividedkomatiites compositionof their sourceregion. into alumina-depletedand -undepletedtypes. If they were formedby very high degreesof melting, thesedifferences can 2. EXPERIMENTAL METHOD AND RESULTS onlybe explainedby differencesin the compositionof the source region[Cawthorn and Strong, 1974; Jahn e! al. , 1980;Herzberg With oneexception, all experimentswere carriedout using and Ohtani, 1988; Anderson, 1989]. However, Arndt [1986] the 2000-ton split sphereanvil apparatusat Stony Brook. A and Fujii et al. [1989] noted that these differencesmay not detaileddescription of the press, the sampleassembly, and the reflect the source,but insteadare phaseequilibrium-controlled techniqueshas been given in numerouspapers [Gasparik, 1989, at very high pressures. In order to test these possibilities, 1990; Herzberg et al., 1990; Herzbergand Gasparik, 1991], and meltingexperiments have been run on compositionswith a wide will not be repeatedhere. The startingmaterials were naturallyoccurring komatiites andseveral compositions in CaO-MgO-A1203-SiO2 (CMAS), and Copyright1992 by the AmericanGeophysical Union. theyare listedin Table 1. The compositionsand experiments in Paper number91JB03066. CMAS are identical to the ones reportedby Herzberg and 0148-0227/92/91JB-03066505.00 Gasparik[1991]. In boththis andthe companionstudy, melting

4521 4522 HERZBERG: DEPTH AND DEGREE OF MELTING OF KOMAT•TES

TABLE 1. Compositionsof StartingMaterials and AverageMantle

1 2 3 4 5 6 C67PX• C•oPX•o PX HSS-15 P9-118 Mantle

$iO2 53.14 54.58 58.92 46.77 44.26 44.47 TiO 2 - - - 0.33 0.35 0.10 AI202 3.37 2.53 - 3.42 7.38 2.57 FeeOs - - - 1.88 _ FeO - - - 9.38 13.26 8.31 MnO - - - 0.19 0.26 0.13 MgO 39.12 38.23 35.58 31.51 27.37 40.98 CaO 4.37 4.66 5.50 5.67 6.31 2.46 Na20 - - - 0.12 0.06 0.23 K20 - - - 0.08 0.08 0.05 Cr:O• - - - 0.34 0.48 0.43 NiO - - - 0.27 0.20 0.27

Total 100.00 100.00 100.00 99.96 100.00 100.00 CaO/(CaO + AI20•) 0.56 0.65 1.00 0.62 0.46 0.49

TABLE 2. ExperimentalResults

Starting T Nominal P Time Material oC kbar min Sequence

1 2000 80 4.0 L,Opx,OI,Gt,Cpx [O1 + Cpx + Opx + Gt] 1 2050 90 4.0 L,Opx,OI,Gt,Cpx [O1 + Cpx + Opx + Gt] 1 2070 100 3.0 L,Opx,OI,Gt,Cpx [O1 + Cpx + Opx + Gt] 1 2140 110 4.0 L,Opx,Ol+Gt,Cpx [O1 + Cpx + Opx + Gt]

2 2080 100 4.0 L,Opx,OI,Gt+Cpx [O1 + Cpx + Opx + Gt] 2 2140 110 2.0 L,Opx, OI,Gt,Cpx [O1 + Cpx + Opx + Gt] 2 2220 130 5.0 L,Gt,Cpx2,Cpxl,Ol* [O1 + Cpxl + Cpx2 + Gt] 2 2240 140 4.0 L,Gt,Cpx2,Cpxl,Ol [O1 + Cpxl + Cpx2 + Gt]

3 2200 150 4.0 L,Cpx2,Cpxl [Cpxl + Cpx2] 3 2240 160 4.0 L,Gt,Cpxl [Gt]

4 1920 80 4.0 L,Opx,OI,Cpx,Gt [O1 + Cpx + Gt] 4 1920 100 5.0 L,Opx,Gt,OI,Cpx [O1 + Cpx + Gt]

5 1600 29 60.0 L,OI,Cpx,Opx,(Gt} [ ? ]

[Phasesin squarebrackets] = subsolidusphase assemblage. {Gt} = Garnetinferred, not observed. Cpxl = High Ca Clinopyroxene. * Opx transformsto Cpx2, a low Ca clinopyroxene,between 110 and 120 kbar [Herzbergand Gasparik,1991]. experimentswere performedon compositionsthat containon the assemblageL + O1 + Opx + Cpx + Gt. The purposeof this solidus the phase assemblage + Olivine + paper is to presentthese liquid data in light of the Orthopyroxene+ Clinopyroxene+ Garnet (L + O1 + Opx + problem. Cpx + Gt) in the 80- to 160-kbar range. The purposeof the Startingmaterials were loadedinto rheniumcontainers, and Herzberg and Gasparik [1991] study was to calibrate the 10 mm size assemblies were used. These were then fired at compositionsof coexistinggarnets and pyroxenesthat occur at 1000øCfor 1 hour prior to runningin order to expell all water. all conditionsin the upper mantle and transitionzone. The Run conditionsranged from 80 to 160 kbar, and from 2000 to startingmaterial compositionswere carefully selectedin order 2240øCnominal. Run durationswere typically 3 to 5 min. Thin to provideexcess orthopyroxene at all conditions,and to provide sectionswere made of the run products, and the garnet and the greatestprobability of observingliquids that were multiply compositionswere determinedby electron probe saturated in the greatest number of crystalline phases. In analysis[Herzberg and Gasparik, 1991]. addition to highly premiseequilibrium pyroxene and garnet Of particularinterest here is the crystallizationsequence, compositions, these experimental data provide improved or the order of the phasesas they appearfrom the hot spotto the constraintson the compositionsof the for the solidus cold end. These are listed in Table 2. The nomenclatureL, HERZBERG: DEPTH AND DEGREEOF MELTING OF KOMATIITES 4523

Opx, O1, Gt, Cpx [Opx + O1 + Gt + Cpx] refers to this O1 + Gt at 2140øCand 110 kbar, and Gt + Cpx at 2080øCand crystallizationsequence, which is a short-handnotation for L; L 100 kbar (Table 2). In the latter experiment, olivine was + Opx;L +Opx+ O1;L +Opx+ O1+ Gt;L +Opx+O1 observedseveral tens of degreesup the temperaturegradient, + Gt + Cpx; the subsolidusphase assemblageis given in indicatinga close approachto simultaneouscrystallization of brackets[Opx + O1 + Gt + Cpx]. The nominaltemperatures olivine, clinopyroxene,and garnet for this compositionat the givenin Table 2 are temperaturesrecorded by the thermocouple. solidus. Experimentslike theseyield the mostinformation on The thermocouplejunction is locatedclose to but to one sideof the T-P-X dispositionof corecticsand invariantpoints, and these the hot spot, and the temperatureit readsis valid at one point are discussed below. alongthe temperaturegradient. This is 50øCwithin 0.7 mm of A 29 kbar experimenton a natural komafiite is also the hot spot, increasingto about 250øC to the cold end of the reportedin Table 2. This was done with the piston-cylinder capsule,about 2 mm distant[Herzberg et al., 1990]. In most apparatusin 1980 at the JohnsonSpace Center. The purposeof experimentsthe nominaltemperature is slightlyabove the solidus this and otherslike it was to determinethe partitioningof Fe and temperature[Herzberg and Gasparik,1991], and the crystalline Mg between olivine and liquid; a full reporting of these low phasesof interestare usuallydistributed within 1 mm of the hot pressuredata will be made elsewherein conjunctionwith new spot where the temperaturegradient is minimal. The liquidus data in the 100-kbar range. The low pressureexperiment is and solidustemperatures differ by 100-150øC,and it is within includedin this paper becauseit providesa constrainton the this temperaturerange where the crystallizationsequence can be compositionsof liquidsalong the highpressure solidus. observed. A sourceof someconcern involves the acquisitionof phase 3. DISCUSSION OF RESULTS equilibriuminformation in a temperaturegradient, and again a detaileddiscussion of this problem appearsin Herzberg et al. 3.1. The SystemCaO-MgO-A1203-$iO 2 [1990]. The major effect is the expulsionof someintergranular melt toward the hot spot and solution-precipitationof the The locations of the cotecticsand invariant points are crystallinephases, observations that were originally made and projectedin Figures 1-4. These figuresare very similar to the discussedby Lesher and Walker [1988] from piston-cylinder onespublished previously by Herzberget al. [1990], differingin experiments. The concernis that this processcould affect the thatthe phaseboundaries are muchmore tightly constrained. In liquidusphase and crystallizationsequence. For experiments particular, it has been possible to position accurately the that range from secondsto 10 min., we have observed no importantparameter CaO/(CaO + A1203)for liquids on the changesin the liquidusphase, in agreementwith the treatment solidus(L + O1 + Opx + Cpx + Gt), and this was done by of Lesher and Walker [1988]. However, for liquids that are observingsimultaneous crystallization of elinopyroxene+ garnet multiply saturatedin two or more crystalline phases, these at the solidus. phasesare not alwaysdistributed in the corecticproportions that It was determinedpreviously that the generationof liquids would be expected from crystallizationin the absenceof a coexistingwith the four crystallinephases Ol + Opx + Gt + temperaturegradient. Rather, somephases concentrate in bands Cpx at 30 and40 kbar is peritectieand takes the form L + Opx that parallel the isothermsfor experimentsgreater than 2 - 3 = O1 + Gt + Cpx [Davis, 1964; Davis and Schairer, 1965; minutesin duration[see Figures 7, 10, & 19 in Herzberg et al., O'Hara and Yoder, 1967]. This periteetiereaction was also 1990]. Thesephases are orderedalong the temperaturegradient observedat 50 and 100 kbar [Herzberg et al., 1990], and our in a way that reflects the crystallization sequence. In newresults support this earlier work (Figure1). The extensive experimentscontaining only threecrystalline phases, the liquidus solubilityof enstatitein elinopyroxeneat melting conditions phaseis obvious,the secondphase appears at and below the yieldsa smalltetrahedron of O1 + Opx + Cpx + Gt, thereby solidus,and the third phase is distributedbetween these two restrictingthe stabilizationof orthopyroxene-bearingassemblages boundingphases. A more complicatedcase involves four to compositionsthat are fairly silica-rich,such as ehondrite. crystallinephases. A good example involves the case of Liquidsgenerated by meltingat 50 and 100 kbar are located peritecticmelting in the 80 kbar experimenton naturalkomatiite outsidethis tetrahedron,to the enstatite-poorside of the plane O1 HSS-15 (Table 2). The crystallization sequence is + Cpx + Gr. The experimentson naturalkomafiite HSS-15 at L,Opx,OI,Cpx,Gt, [O1 + Cpx + Gt]. Orthopyroxeneis 80 and 100 kbar illustratethe peritecticmelting behaviouras completelyabsent in the subsolidus,but is abundantand well (Table 2; point #4 in Figure 1). ubiquitousthroughout the crystallizationinterval. Identical Figure 2 showsthat fraetionationof orthopyroxenecan observationswere reportedby O'Hara and Yoder[1967] on the yield liquidswith lower silicacontents by crystallizationdown meltingof a mixtureof olivine + garnet+ clinopyroxeneat 30 the eotectie L + O1 + Gt + Cpx which degeneratesto a kbar by a seriesof isothermalpiston-cylinder experiments, each thermal minimum. The exact location of this thermal minimum with a minimal temperaturegradient. The melting relation hasnot yet beendetermined. However, the evidencepresently proposedby O'Hara and Yoder[1967] is L + Opx = O1 + Cpx availableindicates that it is more siliceousthan mantle peridotitc + Gt, identicalto the one reportedby us at 50 and 100 kbar analoguesin CMAS, similarto the periteetiepoints [Herzberg et [Herzberget al., 1990](seebelow) based on observationsof the al., 1990]. isobaric crystallization sequenceand the compositionsof Presnall and Gasparik [1990] concludedthat the silica coexistingpyroxenes. We have now obtained internally contentsat theseperitecties were still too low, basedon the consistentisobaric phase diagrams from experimentsdone on a locationof their 165 kbar eutectieL + Fo + Opx on the binary wide rangeof differentstarting material compositions, and have Mg2SiO4- MgSiO3(Figure 2). They arguedthat the peritectie encounteredno complicationsarising from the temperature point locatedby us is in error becauseit would require an gradient. implausiblystrong rotation of the L + Fo + Opx corectie,as In two important experiments,a pair of phaseswere indicated also in Figure 2. But this criticism is actually observedto appearsimultaneously along an isotherm. Theseare erroneousbecause it stems from a distortionin the 4524 HERZBERG:DEPTH AND DEGREE OF MELTING OF KOMATnTES

ChondriticCaAI 2O ''- e20 ""1:•F,P

10

[34 Oc

A

+1% MS 50 40 30 20 10 ENSTATITE DIOPSIDE CS Fig.1. Aprojection toand from olivine (M2S) into part of the plane CS-MS-A (mole percent) [O'Hara, 1968], modified at•er Herzbergetal. [1990]. Garnet and pyroxene compositions atthe pressures indicated (kilobars) arefrom Het7berg andGasparik [1991].The crystallization fieldsand cotectics arestrictly valid for the system CaO-MgO-AI:O3-SiO2. DSis 40 kbar invariant pointfrom Davis and Schraier [1965]; F is 54 kbar invariant point from Fujii et al. [1989]; Wis natural komatiite M620 from Weiet al. [1990].K (komatiite),P (peridotitc), C (chondrite), andCP (chondrite-peridotite mixtures) are compositions and experimentalresultsreported inHerzberg etal. [1990]. Compositions 1,2, 3, 4,and 5 arelisted inTable 1.

70

tmBasalt S so \ 417 9.3 el I e14

ß 20 50 MS a•'mATrm Picrite • DSo

Gt PG

Komatiite 40

5O

(::::) lOO 1•12SFORSTERITE _+1%

20 10 CA Fig.2. Aprojection toand from diopside (CMSz) into part of the plane CA-M-S (mole percent) [O'Hara, 1968], modified after Herzbergetal. [1990]. Solid circles are the compositions ofliquids formed byisobaric invariant melting onthe solidus inCaO- MgO-AI203-SiO2:1atmfrom Longhi [1987] and 7 to20 kbar from Presnall etal. [1979]. Other symbols asfor Figure 1. PG iseutectic inMgSiO3-Mg:SiO4 at165 kbar from Presnall and Gasparik [1990]. Crystallization fieldsand cotecftc boundaries arestrictly valid for the system CaO-MgO-AI:O3-SiO2 (kbar),and may or may not be saturated inclinopyroxene.

projection.This point is illustratedbetter in Figure3, a it is absentfor mantleperidotite which will melt to the projectionfrom A1203 into a partof theplane CaO-MgO-SiO2. assemblage L + O1+ Gt + Cpxon the solidus. Initial melting It canbe seen that pressure does indeed rotate the eotectie L + is strictlypseudoinvariant because the liquid is unlikelyto be Fo + Opxrather substantially, andthat the periteeti½ invariant perfectly coincident in composition with the thermal minimum. pointsat 100to 150kbar have contents of silicathat are very Theseinitial liquids and the thermalminimum are probably similar to thoseon the -enstatitejoin. enrichedin SiO:compared to mantleperidotire (Figure 2), and At 50kbar the crystallization field o f orthopyroxeneisvery progressivemelting will cause a reduction inthese silica contents extensive(Figure 1), andorthopyroxene will participatein the by migrationup the eoteetie L + O1+ Gt + Cpx(Figure 2). meltingof mantleperidotitc (L + O1+ Opx+ Cpx+ Gt; Figure4 is a projectionfrom enstatite into a planethat Figure1) [Herzberget al., 1990]. Pressurecauses the containsforsterite - diopside-pyrope. It is geologicallyvery orthopyroxenecrystallization field to contract,and at 100kbar usefulbecause the compositions of most mantle peridotires and HERZBERG:DEPTH AND DEGREEOF MELTING OF KOMATIITES 4525 komatiitesare containedclose to this plane, so that ambiguities a thermodynamic analysis of these results positions it more stemmingfrom distortionsare minimized. The advantageof precisely at 158 kbar [Herzberg and Gasparik, 1991]. At 140 viewing thingsthis way is that we can now concentrateon the kbar, CaO/(CaO + Al203) is constrainedto be greater than contentsof CaO andA1203, and howthese change with pressure. 0.79; this was determinedfrom solidusgarnets occurring without The mostimportant observation is that pressuregreatly expands clinopyroxene,data reportedin Herzberg et al. [1990] together the crystallizationfield of garnetat the expenseof both olivine with the equilibrium constantthat describesthe partitioningof and clinopyroxene(Figure 4), and garnetcan be seento expand CaO and A1203 between garnet and liquid [Herzberg and at the expenseof orthopyroxeneas well (Figure1). This causes Gasparik,1991]. Includedin Figure5 are experimentaldata in a very pronouncedincrease to occurin CaO/(CaO + A1203)for naturalFeO-bearing systems from many sources.The important liquids that are multiply saturatedin Cpx + Gt, and this is observationis that theseadditional components have no effect, •u;)ilk) .....• 11 in Figure 5. a point discussedin more detail below. The experimentaldata in CMAS position (2aO/((2aO + At pressuresgreater than or equalto 158 kbar, A1203)at 0.65 and 100 kbar (Table 2; 2080ø(2and 100 kbar). cannot crystallize in the melting interval for any composition Near 160 kbar, this value increases to 1.0, based on the becauseof the extensivecrystallization field of majoritc garnet. stabilizationof majoriregarnet in the alumina-freesystem (2MS; Alumina is absent in liquids generated on the solidus (i.e., CaO/(CaO + A1203) = 1.0), but it can be present in liquids locatedbetween the liquidusand solidus. This conditionis likely to persist until about 180 kbar where perovskite has Si02 been predictedto becomestabilized relative to majoritc garnet [Gasparik, 1990]. At pressuresgreater than this, liquids on the 90 soliduswill containA1203 once again. A betterunderstanding of these very high pressure phase relations must await the 8O acquisitionof additionalexperimental data. % SiO 2 The phaseequilibrium data reportedhere and in Herzberg 7O et al. [1990] for the systemCMAS have permitted for the first time estimatesof the absolutecompositions of the liquids for the

6O equilibriumL + O1 + Opx + Cpx + Gt in the 50-to 100-kbar range. They were determinedby randomlyvarying the liquid arm compositionat 50 and 100 kbar until the isobaricinvariant points / • X Diopsid. • •o- •Y•50 En.tatit. were internally consistentwith both the high pressuredata (Figures1-4) andthe low pressuredata of Presnallet al. [1979]. Theseliquid compositionsare listed in Table 3, togetherwith a

/ I.,r-o-•• _ 4)•For.terit. rough estimate of the 150 kbar liquid. The latter was estimated ß ' ' ß ' ' ' • 30 70 60 50 40 30 20 10 in part by the value of unity for CaO/((2aO + A1203)at 158 kbar / (Figure 5). CaO % CaO MgO We are now in a position to evaluatethe full range of Fig. 3. A projectionof invariantpoints and the cotecticequilibrium L liquid compositionson the solidusto 150 kbar, and the results + Fo + Px from alumina(A1203) into part of the planeCaO-MgO-SiO2. are shownin Figure 6. The experimentaldata at 1 arm to 20 Data on thejoin MgO-SiO2:1 atm from Andersen[1915]; 15 kbar from kbar [Longhi, 1987; Presnall et al., 1979] providean important Taylor [1973]; 165 kbar from Presnalland Gasparik[1990]. low pressurereference frame for understandingthe highpressure

ChondriticCa/•l '-.• , • • ..-_PYROPE '''-.ß 9.3ß S-• +1%N•2 0 C M

ß . Picrite • A2S3 5ø / , ,oo x 04 P •

e' D I /Ko..t,,t.I •150 ' 'I '"'-....X'' 30 20 10 M2S C2S3 FORSTERITE Fig. 4. A projectionto andfrom enstatite(MS) intopart of theplane M2S - C2S3- A2S3(mole percent) [O'Hara, 1968],modified afterHerzberg et al. [1990]. Symbolsas for Figures1 and2. The crystallizationfields and corectic boundaries are strictlyvalid for the systemCaO-MgO-A1203-SiO2, and may or may not be saturatedin orthopyroxene. 4526 HERZBERG:DEPTH AND DEGREEOF MELTING OF KOMATIITES

data, and demonstratethat discontinuouschanges occur in all kbar. Pressurecauses a continuousincrease in MgO to contents oxide componentsat phaseboundaries. Theseare especially that are typicalof komatiiteanalo•ues in CMAS, and it tendsto evidentat around9 kbar,where peridotitc transforms decreasethe contentof CaO, and most notably A1203,which to spinelperidotitc [Presnall et al., 1979] and at 22 kbar where dropsto zero at 158 kbar. The low contentsof aluminaare spinelperidotitc transforms to garnetperidotitc (Figure 6). The particularlyrelevant to the aluminum-depletedkomatiites from lowestconcentrations of SiO2 are 47 to 48 % at 22 kbar, and at the Barberton Land. higherpressures there is a steadyincrease to about52 % at 150 The concentrationsof these four oxide componentsat pressuresgreater than 22 kbar have been fitted to the polynomial:

1.0 ! I SiO:•(wt%)= a + bP + cP2 + dP•

.9 The coefficients a, b, c, and d are listed for each oxide A Liquids on the Soil componentin Table 4. It is importantto notethat these equations . are only valid for liquidsthat are bufferedby the four crystalline phasesO1 + Opx + Cpx + Gt. Many naturally-occurring peridotitesand peridotite analogues contain orthopyroxene on the ' L+ Cpx • _ ' solidusat 22 to 50 kbar, but loose it at pressuresmuch higher • ß ß ß than 80 kilobars. When this occurs, melting is along a cotectic o• •n •m m m m L + O1 + Cpx + Gt (see above), and the liquids will have contentsof SiO2 that are less and contentsof MgO that are oyO • / T - o. somewhatgreater than thoseshown in Figure 6.

i 3.2. Natural ComplexSystems 50 100 150

PRESSURE (kbar) For ukramaficrocks such as komatiitesand ,the major difference between the analogue system CMAS and Fig. 5. CaO/{CaO+ A1203)by weightfor liquidson the solidus as a natural complex systemsis the presenceof . The system function of pressure. Squaresare data in CaO-MgO-A1203-SiO2,circles CaO-MgO-FeO-A1203-SiO2 (CMFAS) thereforerepresents about are data in naturalFeO-bearing systems. Solid symbolsare liquid + garnet;open symbols are liquid + clinopyroxene;half open/halfsolid 98-99 % of the geochemistryof theserocks in nature. Several symbolsare liquid + garnet + clinopyroxene.Squares with crossesare experimentswere run on natural komatiite compositions. One solidusliquids in CaO-MgO-AI20•-SiO2 at 1 atm (L; Longhi [ 1987]) and is P9-118, an alumina-undepletedkomatiite from Munro 7 to 20 kbar (P; Presnall et al. [1979]). Unlabelled symbolsare data Townshipdonated to the authorby NicholasArndt. The other reportedin this paper. Other data sources:A, Amdt [1976]; B, Bertka is HSS-15, an alumina-depletedkomatiite from the Barberton and Holloway [1988]; D, Davis [1964]; E, Elthonand Scarfe[1984]; F, Mountain Land, originally describedby Smith and Erlank Fujii et al. [1989];G, Falloonet al. [1988];GR, Greenand Ringwood [1982]. Melting experimentswere reportedfor HSS-15 by Wei [1967]; H, Herzberget al. [1990]; I, Ito and Kennedy[1967]; M, Mysen et al. [1990], and sampleswere donatedto the authorby Reidar and Kushiro [1976]; O, Howells et al. [1975]; OD, Obata and Dickey Tronnes. The reasonwhy additionalexperiments were run on [1976]; OY, O'Hara and Yoder [1967]; T, Takahashi [1986]; W, Wei the Barberton sample is becauseno distinctionwas reported et al. [1990]. The 40 kbar invariantpoint in CMAS estimatedby Davis and Schairer[1965] coincideswith datapoint A. Data of Kushiroet al. betweenclinopyroxene and orthopyroxenein the experimentsof [1968] and Takahashi and Kushiro [1983] on peridotitc HK66 Wei et al. [1990]. The distinctionis importantbecause it is the [CaO/(CaO + AI20•) = 0.38] are anomalousbecause they show observationof multiple saturationin garnet and clinopyroxene clinopyroxenestable relative to garnet (L + Cpx) from 30 to 50 kbar; that determinesthe parameterCaO/(CaO + A1203)for liquidson this indicatesa possibleerror in their reportedbulk composition. the solidus. The compositionsof these komatiite starting

TABLE 3. Compositionsof Liquids on the Anhydrous Solidus

Pressure, kbar

1atm 7 9.3 11 14 20 50 100 150

SiO: 55.40 50.62 49.57 49.13 48.88 48.23 49.5 51.2 52.4 AI:O• 15.20 18.95 20.54 19.72 18.45 17.47 10.0 4.2 0.5 MgO 12.40 14.66 14.30 15.69 17.47 20.22 30.5 36.8 40.2 CaO 17.00 15.76 15.59 15.47 15.20 14.08 10.0 7.8 6.9

1 atm from Longhi [1987]; 7-20 kbar from Presnall et al. [1979] normalizedto 100%' and 50-150 kbar from this work. HERZBERG: DEPTH AND DEGREE OF MELTING OF KOMATIITES 4527

2O

5O

A 15 4O

O o lO

5 i i o 50 lOO 15o 50 100 150

PRESSURE (kbar) PRESSURE (kbar)

6O I i

2O

DS •, 55

50 -

oo 200 PRESSURE (kbar)

450 50' 100, 150

PRESSURE (kbar) Fig. 6. Compositionsof liqi•idsformed on the solidusin the systemCaO-MgO-A•O3-SiO 2. Liquidsare for the equilibria: L + Fo + Opx + Cpx + (1 atm to 9.3 kbar [Longhi, 1987; Presnallet al., 1979]); L + Fo + Opx + Cpx + (9.3 to 22 kbar [Presnallet al., 1979]); L + O1 + Opx + Cpx + Garnet (22 to 158 kbar; Table 3). DS, Davis and Schairer [1965]; F, Fujii et al. [1989].

TAI•LE4. CoefficientsforLiquids onthe Anhydrous Solidus inthe These results are shown in Figure 5. The pressureat which systemCaO-MgO-A1203-SiO2 and at Pressuresin the Garnet multiplesaturation occurs in garnetand clinopyroxeneis located StabilityField (L + Ol + Opx + Cpx + Gt) betweenthese brackets, and is valid for CaO/(CaO4- A1203)of HSS-15, that being 0.62. This is exactly the vfilue that is predictedby the experimentaldata in the simplesystem CMAS. a b c(l(Yz) d(10-e) Most garnet synthesisexperiments in natural complex systemshave been done on komatiite, peridotite, and ecli•gite SiO2 46.15 0.0913 -0.561 1.54 compositionsrestricted to the 23- to 40-kbar range, and the AI:O• 26.04 -0.4775 3.677 -10.86 results are included in Figure 5. The important observationis MgO 9.21 0.6562 -5.415 16.12 that the addition of FeO neither expandsnor contractsthe CaO 18.60 -0.2699 2.299 -6.80 c,rystallization field of garnetwith respect to clinopyroxbne. This is illustratedby the data of Bertka and Holloway [1988] on Si02(wt%) = a + bP+ cP2 + dP3 ...where P is in kbar a veryiron-rich garnet composition •hat is appropriate for Mars. The minimum melt at 23 kbar has the following materialsare given in Table 1, and the experimentalresults are properties:23.3 % FeO; 12.7 % MgO; CaO/(CaO + A1203)= included in Table 2. 0.44. In the iron-free system CMAS, it is predicted that Iri the 80 kbar experimenton HSS-15,clinopyroxene CaO/(CaO + A1203) = 0.44 at 23 kbar. It can be seenthat the occursabove the solidus,and garnetis restrictedto temperatures parameterCaO/(CaO + A120•) is commonto both CMAS and at andbelow the solidus;the oppositewas observedat 100 kbar. all geochemicalsystems of interest. 4528 HERZBERG: DEPTH AND DEGREE OF MELTING OF KOMATIITES

Addition of FeO doesnot affect the crystallizationfield of The algorithm is simple becauseCMAS can be extendedto garnet with respect to clinopyroxenepresumably because it CMFAS by the introductionof two constantparameters (i.e., lowers the freezing temperaturesof the iron-free end-members 0.46 and 0.69), and two adjustableparameters, the composition CaMgSi20• and Mg3A12Si3012about equally. The major effect of olivinein the sourceand the degreeof partial melting;a copy of FeO is to lower the SiO2 content of a liquid at an isobaric of it can be obtainedfrom the authoras a computerprogram. invariantpoint. In the very FeO-rieh compositionreported by Againthe liquidsso calculatedare valid only for initial melting Bertka and Holloway [1988], SiO2 = 41.9 % for the liquid (L at pressuresin the garnetstability field. + O1 + Opx + Cpx + Gt; 23 kbar). This compareswith about Table 5 shows that the calculated iron-rich initial melt 48 % in the systemCMAS (Figure 6d) and 44.4 % in CMFAS compositionof Bertka and Holloway is in excellentagreement when MgO is replaced by FeO in the appropriateproportions. with the observedcomposition. The meltingof mantleperidotitc This effect is also observedin other ways. For komatiiteHSS- on Earth at 25 kbar will generatepierite, andTable 5 showsthat 15 at 100 kbar (point #4 in Figures 2 & 4), the crystallization the calculatedcomposition is also in very good agreementwith sequencein CMAS is predictedto be L,Opx,O1,Gt,Cpx. This that determinedexperimentally by Elthon and $carfe [1984]. differs from the observed crystallization sequence Initial liquids formed at 50 kbar contain23.2 to 25.3 % L,Opx,Gt,O1,Cpx, demonstrating that FeO expands the MgO and 13.7 to 10.7 % FeO, dependingon the degreeof crystallizationfield of garnetat the expenseof olivine. Liquids partial melting (Table 5). At 100 kbar, MgO rangesfrom 27.4 at isobaricinvariant points in CMAS have slightlyhigher CaO, to 30.4 %. These compare with 24 % MgO at 40 kbar, A1203, and SiO2 than those found in nature. An estimateof determinedexperimentally by Litvin [1989]. Liquids generated these effects is now made. at 50 and 100 kbar invariantpoints are distinctlykomatiitic in As a first stepin extendingCMAS to CMFAS, MgO was composition, and their significance to the Earth will be replaced by FeO using the distribution coefficient Kt• = addressedbelow. These liquid compositionsare now projected (MgO/FeO)i•q•iax (FeO/MgO)mvi•. The value of Kt•is 0.34 at in Figures7a & b in order to demonstratethe effect of FeO on 30 to 40 kbar [Bickel et al. 1977], and remains fairly constant expandingthe crystallizationfields of garnet,clinopyroxene, and in the 100-kbar range [Takahashi, 1986; C. Herzberg, orthopyroxeneat the expenseof olivine. unreporteddata, 1991]. The initial liquid compositionswere Use of liquid compositionsin CMAS are actually very calculatedby massbalance with olivine of a known composition, successfulwhen applied to naturally occurring rocks with a this being Fo•0 for averagemantle peridotite [Boyd, 1989], and terrestrial FeO/MgO. Simple substitutionof MgO for FeO Fo75 for a presumed Martian mantle olivine [Bertka and without accountingfor a contractionof the crystallizationfield Holloway, 1988]. This first stepwill yield a liquid composition of olivine results in contentsof SiO2, A1205, and CaO that are that projectsto the same location as that in CMAS. In the too high by 1.7 %, 0.3 %, and 0.6 %, respectively(absolute), secondstep, the contentsof SiO2, A1203,and CaO were lowered at 100 kbar; MgO and FeO are too low by 2 % and 0.6 %, becauseFeO contracts the crystallization field of olivine as respectively. However, thesesmall differencescan yield very discussedabove. The Bertka and Holloway [1988] data were different crystallizationsequences. For example, the system usedto parameterizethis effect, and the resultsare: CMAS predicts the crystallizationsequence L,Opx,O1,Gt,Cpx for HSS-15 at 100 kbar. This compareswith L,Opx,Gt,O1,Cpx, SiO2 = [a + bP + cP• + dP3 ] x [1- 0.46(FeO/MgO)] a crystallizationsequence that is both experimentallyobserved where the coefficientsa, b, c, and d are valid for the system (Table 2) and predictedin CMFAS (Figure 7). CMAS (Table 4), and FeO/MgO (wt. %) is for the composition At the present time, the largest discrepancybetween of olivinein the sourceregion. Similarly, the reductionsin CaO calculatedand observed liquid compositionsis for the Elthonand and A1203were parameterizedby the following: Scarfe [1984] pierite at 25 kbar (Table 5). Although the disagreementis not understood,accepting it at face value A1203= [a + bP + cP2 + d/a] x [1- 0.69(FeO/MgO)] illustratesthat the calculatedand observedliquid compositions CaO = [a + bP + eP2 q- dP3] x [ 1-0.69(FeO/MgO)] differ by no more thanthe followingrelative amounts: +/- 5 %

TABLE 5. Compositionsof Liquidson the AnhydrousSolidus in the SystemCaO-MgO-FeO-A1203-SiO2 and at Pressuresin the GarnetStability Field (L + O1 + Opx + Cpx + Gt)

,

Pressure 23 25 50 100 kbar

% Melting 1 1 1 20 40 1 20 40 Fo 0.75 0.90 0.90 0.91 0.93 0.90 0.92 0.93

Calc. Obs.* Calc. Obs. # Calc. Calc. Calc. Calc. Calc. Calc.

SiO2 42.0 41.9 45.5 47.9 45.7 45.9 46.2 46.1 46.7 47.0 A1203 11.9 11.9 14.5 13.8 8.7 8.8 8.9 3.6 3.6 3.7 FeO 23.4 23.3 10.7 9.9 13.7 12.1 10.7 16.2 14.0 12.1 MgO 13.1 12.7 17.5 16.6 23.2 24.4 25.3 27.4 29.0 30.4 CaO 9.6 9.4 11.8 10.9 8.7 8.8 8.9 6.7 6.7 6.8

* From Bertka and Holloway [1988]. # FromElthon and Scarfe[1984]. HERZBERG: DEPTH AND DEGREE OF MELTING OF KOMATIITES 4529

•,,•GARNE'x .50 MS OPX

• ...... s•.- --"•'•50kbar•': 1•2S350•ClaASkb:r•HSS"5••40

30 20 10 M2S 10 0 CA C2S3 OLIVINE

Fig. 7. Pseudoinvariantpoints and cotecticboundaries in the systemCMFAS comparedto CMAS (Figures 2 and 4). Experimentalwork on KLB-1 [Herzberg et al., 1990] show the crystallizationsequence L,Gt,OI,Cpx at 140 kbar; multiple saturationin olivine + garnet (L + Ol + Gt) has not been determined,but will not be much lower than 130 kbar.

for SiO2, A1203,and MgO; +/-8 % for FeO and CaO (Table 5). and is very close to the average of 384 spincl lhcrzolites The most seriousdifficulty is with SIO2. The value offered by compiledby Maaloe and Aoki [1977]. El&on and Searle [1984] is 47.9% substantiallyhigher than the The komatiitcsare dividedinto two populations. One is calculatedvalue (45.5 %). But this high SiO2 is probably the alumina-depletedtype komatiitc[Nesbitt e! al., 1979] which incorrectbecause it predictsa crystallizationfield of olivine (L is characteristicof the Barberton Mountain Land and Pilbara, + O1) at 27 to 30 kbar for which orthopyroxene(L + Opx) was and are these have ages of about 3.5 Ga [Jahn et al., 1982; observedinstead for the binary diopside-pyropein CMAS, the Gruau et al., 1990a]. The secondis the alumina-undcplctcd FeO-bearingcomposition Cpx2sOpx2sGrs0 (A3/10596) [O'Hara type which is most commonly found in other eratons and is and Yoder, 1967], and the pierite of Green and Ringwood typically younger (e.g., 2.7 Ga [Gruau e! al., 1990a]); these [1967]. Mysen and Kushiro [1977] also reportedvery low silica are well representedby komatiitcsfrom the Munro Townshipin contents at 20 kbar. It is concluded, therefore, that initial the Abitibi greenstonebelt in Canada [Arndt e! al., 1977]. The meltingin the 25 kbar rangeyields picritic liquids (-- 16 - 17 % alumina-undcplctcd character of the Munro komatiitcs is MgO) with low SiO2 (45 - 46 %; Table 5). revealedby their ratio of CaO/(CaO + A1203)(by weight) that The componentNa20 will reducethe amountof MgO that averages0.49 (see appendix;CaO/A1203 = 0.99), essentially is containedin liquids on the solidusbecause it reducesthe identicalto average mantle peridotitc [Her:berg e! al., 1990]. solubilityof olivine in silicateliquids [Herzberg, 1979]. At 15 The contentsof CaO and A1203are indicatedin Figure 8 by the kbar, the addition of 3.5 % Na20 to the system CMAS can coordinatesC2S 3 and A2S3, the chondritic value included for reduceMgO from about 17 % to 10 %, and it also contributes reference. The alumina-depletedcharacter of the Barberton to a lowering of SiO2 [D.C. Presnall, personalcommunication komatiitcsis manifestby high CaO/(CaO + A1203), and this is 1991]. This will causean importantuncertainty in estimating clearly evident in Figure 8b. the MgO contentof initial liquidsin which Na20 is concentrated, Komatiitcsfrom the GorgonaIslands have a major clement a problemthat is likely to be importantfor smallmelt fractions geochemistrythat is similar to Munro [Gansser e! al., 1979; (i.e., 1-2 %). However, this problem is not likely to be Echeverria, 1982; Aitken and Echeverria, 1984], and they are importantin interpretingmost komatiites because they typically important becauseof they are Mesozoic in age [Walker e! al., containless 0.5 % Na20, and most of them were formed by 1991]. In detail, however, they are slightly lower in both degreesof meltingthat exceeded25 % (see below). But the CaO/(CaO + A1203) and MgO compared to Munro-type reader should be cautionedthat liquids on the solidusin the komatiitcs. systemCMFAS (Table 5) will bc too high in MgO, especially An analysis now follows of the depths and degrees of for low melt fractionsthat contain severalpercent Na20. partial meltingthat were involvedin the formationof the Munro- type, Barberton,and Gorgonakomatiitcs. The depthsof melting 4. DEPTH AND DEGREE OF MELTING OF KOMATITES have been estimatedfrom CaO/(CaO + A1203). The degreesof 4.1. GeochemicalTypes of Komatiites partial melting have been estimatedby mass balancing the compositionsof the liquid and crystallinephases with the bulk In Figure 8 arc projected the compositions of 668 compositionbeing melted, so that meltingis strictly equilibrium. pcridotitcsfrom our world data base [Her:berg et al., 1988, The compositionsof garnet, clinopyroxcnc,and orthopyroxcnc 1990] togetherwith 255 komatiitcs. The komatiitc data base is at all temperaturesand pressuresof interestare now very well from too many sourcesto list, but a samplingincludes Nesbit! known, particularlyalong the solidus,and the compositionsused and Sun [1976], Nesbitt et al. [1979], Arndt et al. [1977], in this analysis have been retrieved from the equations of Viljoenand Viljoen [1969], Smithand Erlank [1982], Pyke et al. Her:berg and Gasparik[1991]. The compositionsof the liquids [1973], and Nisbet et al. [1987]. The peridoritosare the so- on the solidusdepend on both pressureand the degreeof partial called "oceanic"types [Boyd, 1989]; theseinclude all samples melting, and these were computedfor the system CMFAS. For with the exceptionof the "continental"types which are the advancedlevels of melting (e.g., L + O1 and L + O1 + Opx), coarse-grainedand low temperaturexcnoliths from sourthorn the compositionsof the liquids were estimated from the Africa. The averageof the 668 peridotitcsis listed in Table 1, dispositionof the cotecries,and these are shown in Figures 9a, 4530 HERZBERG: DEFTH AND DEGREE OF MELTING OF KOMATIITES

ChondriticCa/A/ PYROPE Munro-Type Komatiites 9.3 2O 11 ß 14 2O A•S•

&&

& lO

lOO ß.A '

DIOPSIDE a

M2S

9.3 OLiVINE 2O 7 11 ß 14 Barberton Komatiites A2S3

5o Gt lO

DIOPSIDE

30 20 10 M2S C2Sa OLIVINE Fig. 8. A projectionof liquidusphase boundaries in CMFAS fromFigure 7, mantleperidotitc (dots), and komatiites (triangles).

GARNET 9b, and 9c for the Gorgona, Munro-type, and Barberton • \ . oPx GorgonsKornatiitea komatiitesrespectively• 4.2• Gorgona Komatiites

The parameter CaO/(CaO + A1203)is restrictedto 0.46- 0.47 [Echeverria, 1982; Aitken and Echeverria, 1984], indicating • a ,oo a pressureof meking of about36 kbar. There is someindication OLIVINE that CaO has been removed becauseof alteration (see below and appendix), so the pressureof melting was probably somewhat GARNET higher. MunroKomat • Opx%, At 40 kbar, the partial meltingo f averagemantle peridotite Gt • 44 X on the soliduswill involve the assemblageL + O1 + Opx + 50kbarJ •,-3s•-•---- •, Cpx + Gt, and the degreeof melting will range from about 1 to 30 %. In CMFAS the liquids formed will have 20 to 23 % MgO and a low contentof SiO2 (i.e., 45 - 46 %). The Gorgona CPX / --01 AverageMantlePeridotite N OL"VIN" komatiites are relatively high in Na20 (i.e., about0.8 % [Aitken •oo % •• OLIVINE and Echeverria, 1984]), indicatingthat theseMgO contentsare overestimatedby 1 to 2 %(seeabove).Nevertheless, the Gorgona komatiitesare very similarto theseprimary liquid compositions, GARNET OPX exceptthat they have beenmodified by addkionand subtraction of olivine (Figure 10). BerbertonKomatiite• When the degreeof partial melting exceeds30 %, garnet is the first phaseto be consumed(L + O1 + Opx + Cpx). CaO and A1203will changeslightly until the degreeo f partial melting reaches33 %, at which point CaO/(CaO + A120•) of the liquid CPX'/" cpx-"•Gt'5•rl 1100 •• 'VINEis 0.49, defined by the peridotite source. Progressivemelting OLIVINE will exhaust clinopyroxeneand melting will proceed up the Fig. 9. A projectionof liquids formed by the melt percentagesshown harzburgitecotectic L + O1 + Opx. Orthopyroxenewill remain at (a) 40, (b) 50, and (c) 100 kbar. Average mantleperidotitc is given in the residueuntil the degreeof melting reaches40 %. It is in Table 1. Primitivemantle peridotitc is lb8 of Stoschand Seck[1980]. during the progressivemelting of harzburgitewhere the largest HERZBERG: DEPTH AND DEGREE OF MELTING OF KOMATIITES 4531

ßOlivine 15 I I I 50 'I I I

Gorgona lOO 40 - e•90 Gorgona

10 + _ -10 kbar

O _ 0 50 _ • 30

_

5-- ß 30 40

--

_

--

. 40 kbar 20- t+ lO kbar o. -- 100'•90 80 :.. 40 kbar

--

,.J _! I I I I I I I i I , , , I I I I I I i i i i 10 ' ' ' ! 040b.•ne 45 50 4O 45 50 SiO2 SiO 2

15 , I I I 15 I i i I I I I '1 I I I I I

_ Gorgona

10 10kbar I .•.30

0 lOO

40 40 Gorgona 40 kbar ,o 40 kbar 5 9Oo& 1009•08070 Olivine , I I I I , I I i ,I 0 f I 100I I , I , , I : I I I I I 04O bli;ine 45 50 55 40 45 50 SiO2 SiO2 Fig. 10. Liquidcompositions formed by 1 to 100 % meltingof averagemantle peridotite at 40 kbar, comparedto the geochemistryof Gorgonakomatiites (weight %). Phaserelations are shownin Figure 9a. Liquidsare valid for the system CMFAS; Na20 will lower MgO (see text).

changesoccur in the compositionof the liquid, particularlyfor compositionsradiate to and from olivine, consistentwith the CaO, A1203, and SiO2 (Figure 10). The content of SiO2 additionand removal of olivine from a 50 kbar partial melt. But increasesbecause of the dissolutionof orthopyroxenein the this interpretationis neither unique nor without its difficulties. melt, and reachesa maximum of 49%. Mysen and Kushiro Another possibility is that the Munro komatiites were all [1977] documentedsimilar changes at 20 kbar; liquidsformed by residual liquids formed by extensive (L + O1) or total (L) < 2 % melting are nephelinenormative with 44 to 45 % SiO2; melting of mantle peridotite. Figure 8a showsthat they could liquids formed by 35 % melting and in equilibrium with have formedby fractionationof variableamounts of olivine from (L + O1 + Opx) have 51% SiO2. Large variations a liquid with the compositionof mantleperidotite. Most of the were also reported for CaO and A1203 [Mysen and Kushiro, Munro-typekomatiites have contentsof MgO that exceed24 %, 19771. and these project to the olivine-rich side of the 50 kbar Inspectionof Figure 10 demonstratesthat most of the pseudoinvariantpoint (Figure 8a). Although some can be Gorgonakomatiites were formedby pseudoinvariantmelting that explainedas olivine cumulatesin a 50 kbar partial melt, Bickle did not exceeded30 %. Garnet is expected to have been a [1982] hasdemonstrated that komatiiteshaving 27 to 30 % MgO residual phase for most komatiites(L + O1 + Opx + Cpx + could have been cumulate-freeliquids. Since these highly Gt), and this is consistentwith the observationof depletionsin magnesianliquids are not coincident with a high pressure the heavy rare earth elements (Gd/Yb >1 [Aitken and invariantpoint, the logicalinterpretation is that they formedby Echeverria, 1984]). Again, the pressureof partial melting was very high degreesof melting. For the caseswhere melting is 40 kbar, about 130 km depth. total (L), high (L + O1), or moderate(L + O1 + Opx), the contentsof CaO and A1203in komatiites are inherited from the 4.3. Munro-Type Komatiites peridotite source. And since CaO/(CaO + A1203) for average mantleperidotite and partial melts at 50 kbar are similar purely Inspectionof Figure8a showsthat many of the Munro-type by chance,it is possibleto concludethat komatiitesplot near the komatiitesare similar in compositionto liquids at the 50 kbar 50 kbar invariant point also by chance. pseudoinvariantpoint L + O1 + Opx + Cpx + Gt. The We therefore have two possible interpretationsfor the content of MgO is 23 to 25 % (Table 5), and it is therefore Munro komatiitesbased strictly on CaO/(CaO + A1203). Either possibleto interpretthese as primary . The spectrumof they were partial melts from 50 kbar that had experienced 4532 HERZBERG:DEPTH AND DEGREE OF MEt.TING OF KOMATIITES olivineaddition and subtraction,or theywere all residualliquids At 50 kbar,the partial melting of averagemantle peridotite formedby variableolivine removal from an extensively melted on the soliduswill involve the assemblageL + O1 + Opx + peridotitcsource. This problem is importantto resolvebecause Cpx + Gt andthe degree of meltingwill rangefrom about 1 to the implicationsare enormous. Miller et al. [1991b] 35 %. The geochemicalcharacteristics of these initial liquids demonstratedthat mantle peridotitcwould have to melt at are CaO/(CaO+ A1203)= 0.50; MgO = 23 to 25 %; andSiO2 pressuresin the lower mantle ( > 200kbar) in orderto be totally = 45 - 46 %. Whenthe degreeof partialmelting exceeds 35 or extensivelymolten at the surface.This in turnwould place %, clinopyroxeneis the firstphase to be consumed(L + O1+ profoundconstraints on the compositionof the lower mantleß Opx+ Gt; Figure9b). CaOand A1203 change slightly until the However,a komatiiteformed by lowerdegrees of meltingat 50 degreeof partial meltingreaches 36 %, at which point kbar doesnot constrainthe geochemicalstructure of the source CaO/(CaO+ A120•)of the liquid is 0.49, definedby the region. In orderto evaluatewhich of thesetwo possibilities is peridotitesource. Progressivemelting will proceedup the erroneous,a more detailedexamination is now made of the harzburgitecotectic L + O1 + Opx after garnethas been partialmelting possibility, with degreesof partialmelting that exhausted.This scenariois very similarto the one discussed rangefrom 1% to 100 %. The phasediagram is shownin abovefor the originof the Gorgonakomatiites at 40 kbar. It is Figure9b, andthe resultsare shownin Figure11. duringthe progressivemelting of harzburgitewhere large An additionalcomplication that has madethe Munro4ype changesoccur for CaO, A1203,and SiO 2 in the liquid(Figure komatiites difficult to understand is that there is such a wide 10), eventhough CaO/(CaO + A1203)remains at 0.49, defined rangeof compositions,with SiO2that rangesfrom 44 % to 51 by theperidotite source. Changes in MgO are moremodest, %, andCaO andA1203 that both range from 4% to 10% (Figure increasingfrom 25 % to 27 %. The contentof SiO2increases 11). This variationcannot be explainedsimply by olivine to 49 %, and this increaseoccurs because orthopyroxene is additionand removalfrom a singleparental composition or by beingdissolved into the melt. Furthermelting will involveonly alteration(see appendix). However, Figure 11 showsthat this olivine(L + O1), andits dissolutionwill causeSiO 2 to drop. spectrumof compositionscan be explainedby additionand Inspectionof Figure11 showsthat most Munro-type komatiites subtractionof olivine from a spectrumof liquidsformed by 35 can be describedby additionand subtractionof olivinefrom to 44 % meltingof averagemantle peridotitc. liquidsthat had harzburgite in the residue (i.e., L + O1+ Opx),

© Offvine 50 -, • 10 I I I I I I I i I I . I .I I ø . I .. I . I I I I I ß :35' - _

lOO Munro-Type _ 4O 50 kbar

_

_ : ... 70

O ß "' : '•. 44 0 -5 o• 30 - ' ' '/.:..., ..50.. (,3 44." ø ' ß ø60 ' - ß ß Munro-Type _ - 10 kbar ß .'. . . 2o .. 100ß' •o"ø 50 kbar - jE+_ ß•..:.. ,.::...

0 _,_, , • I I • • I I I I • 10 I I I I I , , I I 40 •//;lne 45 50 55 40 45 50 Si02 Si02

10 15 ' ' ' ' , la

_ ß . '.'. .• ß ..•-• ...., : ; ß 10 kbar f++,O kbar""•'::i'"'I I I I ]I I I + 10 : • '• ß '"".'": 44 _ , ' ø70: . o --'1•F09o':ø '.'"_;'0•44 -.• 5 ß • J.' 5o • - 34•40•F094 100 Munro-Type Munro-Type ß /d ø - 80 50 kbar ß•oo 50 kbar - 100 - Off vine

_ 0 _.l:, , , I , , , , I t t t , • I ; I I I 1 I I i I 40 Offvine 45 50 55 40 45 50 55 SiO2 Si02 Fig. 11. Liquidcompositions formed by 1 to 100 % meltingof averagemantle peridotite at 50 kbar,compared to the geochemistryof Munro-type komatiites (weight %). Phaserelations are shown in Figure9b. Liquidsare valid for thesystem CMFAS. HERZBERG: DEPTH AND DEGREE OF MELTING OF KOMATIITES 4533

and that the degreeof partial meltingranged from about35 to 44 Barbertonkomatiites, characteristicallyhigh in CaO/(CaO + %. The absenceof garnet in the residueis consistentwith the A1203). But before we can proceed with an igneous rare earthevidence, which showsno depletionsin the heavyrare interpretationof theserocks, an assessmentmust be madeof the earth elements [Arth et al., 1977; Jahn et al., 1980]. effect of alteration on CaO/(CaO + A1203) [Smith and Erlank, Although this analysishas been restricted to the partial 1982]. This hasbeen the subjectof considerablyscrutiny by de melting of averagemantle peridotite, it is also possiblethat the Wit et al. [1987], and an expansionof their discussionis given sourceregion containeda considerableamount of harzburgite in the appendix. Alteration by circulating hydrothermal sensustricto (O1 + Opx) in additionto lherzolite(O1 + Opx + solutionstends to strip from a significantamounts of CaO Cpx + Gt)[D. Canil, personal communication,1991]. If and SiO2, a processthat appearsto have affectedthe Barberton correct,the spectrumof Munro komatiitescould have formed by komatiitesmore than most Munro-types. Removal of CaO is degrees of melting that were lower than the estimatesgiven usuallynot accompaniedby a local gain in CaO, so the system above, that is <40%. is open on an outcropscale; alteration therefore tends to lower So far we have considered melting at 50 kbar; but SiO2 and CaO/(CaO + A120•). Pressuresestimated from inspectionof Figure 10 illustratesthat this scenariocould have CaO/(CaO + A120•) will therefore be minimum possible operatedalso at 40 kbar. For liquids that have a harzburgite bounds. The samples projected in Figure 8b are the least residue(L + O1 + Opx), the parameterCaO/(CaO + A1203)is altered, about 50 % of the Barbertondata base (appendix). definedlargely by the compositionof the sourceregion. In this Figure 8b demonstratesthat many Barbertonkomatiites case, pressure information is lost becauseMunro komatiites have the compositionsof liquids at a pseudoinvariantpoint could have had harzburgiteresidues from 25 to 80 kbar. The involving L + O1 + Gt + Cpx in the 80- to 100-kbar range; the actual pressureof melting is therefore subjectto considerable spectrumof komatiitesradiate to and from olivine, indicating uncertainty. However, a number of lines of evidence indicate that someare liquidsthat have gainedand lost olivine. Partial that 50 kbar is probablyrealistic. If the pressurewas as low as meltingat 100 kbar hasbeen modeled, and the resultsare shown 25 kbar or so, initial liquidsand thosefor harzburgiteresidues in Figure 9c and Figure 12. Melting involvesthe assemblageL will be too low in MgO. Conversely,if the pressureof melting + O1 + Gt + Cpx and, although orthopyroxenedoes not was near 80 kbar, the SiO2 contentwould be too high (Table 5, participate,the modelingwas simplifiedby usingan intial liquid Figure 12). compositionfor the peritectic equilibrium L + O1 + Opx + It is alsopossible to explainthe geochemistryof the Munro Cpx + Gt. As explain above, use of this compositionwill result komatiitesas residualliquids from highdegree (L + O1)or from in MgO that is too low and SiO2 that is too high, a problem that total melts (L), but it requires some special circumstances. will not seriouslyaffect the conclusionsthat follow. Although not explicity shown in Figure 11, residual liquids A 1% melt fractionwill have about3.5 % A1205and 27 formed by fractional crystallizationof olivine from a liquid % MgO (Table 5), but doesnot quite matchthe geochemistryof having the compositionof average mantle peridotite would Barbertonkomatiites because SiO 2 is too low (Table 5; Figure definea track similarto, but not identicalwith, the equilibrium 12). To increase SiO2 to the levels observed, the degree of L + O1 track (44 - 100 %; Figure 11). There would be only partial melting must have been high. But becauseorthopyroxene one way to explaina komatiitethat did not have a composition is not involved, the only way this can be achieved is by high on thistrack. For a komatiitethat is low in bothMgO and SiO2, degreesof pseudoinvariantmelting on the solidus (L + O1 + Gt that is one with the characteristics of an initial melt at 50 kbar + Cpx) involving a sourceregion that was also fairly high in (i.e., 24 % MgO and 46 % SiO2), the sourceregion would have SiO2. Although the compositionof the sourceregion does not to be lower in SiO2 than averagemantle peridotite. directly affect the compositionof liquids formed on the solidus, Our world database of 668 peridotitesyields an averageof it does affect the degree of partial meking that is possible. 44.47 +/- 2 % (two standarddeviations), so there exist real Higher degree initial melts will be higher in MgO, MgO/FeO, variationsin silica in the mantle. If thesehigh degreemelting and SiO2 (Table 5), even for liquids that are strictly argumentsare applied to the Gorgona komatiites, which are pseudoinvariant.Average mantle peridotite is not a good source Munro-like in that CaO/(CaO + A1203) is similar to mantle regionbecause pseudoinvariant melting could not have exceeded peridotite, then we may anticipate wide variations in their 35 %. But for the primitive mantleperidotite chosen, which is geochemistryalso. But inspectionof Figure 10 showsthat the Ib8 of Stoschand Seck [1980], initial melting can reach up to 52 geochemistryof the Gorgonakomatiites is quite restricted. This % (Figure 9c). The liquid producedhas about 31% MgO and means that if indeed they were formed by high degreesof 47 % SiO2 (Figure 12), and successfullyreproduces the meltinginstead of by pseudoinvariantmelting, the sourceregion geochemistryof manyBarberton komatiites. Again the spectrum must have been very homogeneouswith respectto the major of komatiites can be explained by olivine addition and elements. substractionfrom this composition. We are left with the unfortunate conclusion that the Melting in excessof 52 % is very unusual. Clinopyroxene geochemistryof the Munro-typekomatiites is compatiblewith will be first phase to be consumed,and advancedmelting will both very high (e.g., 100 %) and moderatedegrees (35 - 44 %) proceedup the cotecticL + O1 + Gt (Figure 9c). When this of partial melting. But the essentialdifficulty with the high melt occurs,there will actually be an increasein A1205in the liquid fractionscenario is that it would have requireda uniquesource becausethe A1203 content of the source is region to explain each Munro komatiiteand fails, therefore,on higher than initial Barberton liquids (Figure 12). Barberton groundsof improbability. komatiitescould not haveformed from thesehigh degree liquids. The next phaseto be consumedis predictedto be olivine, and 4.4. Barberton Komatiites garnetwill be the liquidusphase for Ib8 at 100 kbar (Figure 9c). When initial melting occursat pressuresthat are greater However, the crystallizationinterval of garnet only (L + Gt) is than 50 kbar, the liquidswill begin to have affinitieswith the likely to be only sometens of degrees,and the liquid trajectory 4534 HERZBERG: DEPTH AND DEGREE OF MELTING OF KOMATIITES

c Oilvine lO I I I I i i 50 ., , , I I I I I I I .I. I. I ' I ß . Barberton ß

ß ß , 40 - + +2o kbir ß ..lOO ß ß

_

ß o 0 ß •ø52 - 5 m 30 - -•- +_ 2okbir o ß .

ß ß loo ß Barberton o ß

100 kbar 20 100 kbar 10 I I t , I i I I I [ I i I I %0•ii•n• 45 50 55 40 45 50 55 SiO2 SiO 2

1 10 I I I - (50kb•r• Barberton 5f-I-I + 20] kbnr[ I ][ I I I [

100 . ß ß ' 10f••Ooo ' 400 Barberton _ 20 kbir ' '" .

_ 100 kbar ß . . ß 100 kbar -- h 100

_ Oilvine -- -, _, , , I , , , , ! i i I i 40 Oilvine 45 50 55 40 45 50 55 SiO2 Si02 •ig. 12. Liquidcompositions formed by 1 to 100 % meltingof primitivemantle peridotitc [Ib8 Stoschand SecA, 1980], comparedto the geochemistryof Barbertonkomatiites (weight %). Phaserelations are shownin Figure9c. Liquidsare valid for the systemCMFAS. from 52 to 100 % meltingin Figure 12 was treatedas if Ib8 was 5. EFFECTSOF DECOMPRESSIONON THE GEOCHEMISTRYOF multiply saturatedin olivine and garnet (L + O1 + Gt) at all KOMATIITE temperaturesbetween the liquidusand the solidus. A Barbertonkomatiite formed as a partial melt at 100 kbar Althoughisobaric phase diagrams like Figure 9 are useful would have had garnetin the sourceregion (i.e., L + O1 + Gt for examininghow the geochemistryof a komatiiteis affectedby + Cpx). The importanceof garnet was discussedin the advanced melting above the solidus, they are geophysically pioneeringarticles of Green [1975], Sun and Nesbitt [1978], unrealistic because the melting of mantle peridotitc cannot Nesbittet al., [1979], and Jahn et al., [1982], andthis was proceedfrom initialto totalisobarically. It is triggeredwhen basedon the observationthat high CaO/(CaO + A1203)is the mantleadvects adiabatically and intersects the solidus,and accompaniedby depletionsin the heavyrare earthelements the parcel of magmawill followa newadiabatic T-P pathwhich (e.g., highGd/Yb). However,there has been less agreement connects the temperaturei and pressures of meltinitiation on the has to the mechanismby which gai'netwas removed. The solidusto eruptiontemperature at the surface.These concepts generalpOssiblities that have been discussed are (15 garnet was were appliedto the generationof basalticmagmas at low retainedas the residualphase or fractionatedsbmehow during pressures[McKenzie, 1984; McKenzie and Bickle,1988], and the time of melting[Green, 1975; Sun and Nesbitt,1978; more recentlyto the generationof komatiitesat elevated Ohtani,1984; Arndt, 1986],(2) the sourceregion had inherited pressures[Miller et al., 1991a,b]. thegeochbmical signature ofa priorstage of garnet removal by, AdiabaticT-P trajectories forkomatiite magmas are shown for example,removal in anearly ma•,m•. ocban, and garnet did in Figure•3. Thekomatiites formed on the solidus at 40to 100 not necessarilyplay a role duringmagma genesis [Jahn et al., kbarhave been connected to eruptiontemperatures by meansof 1982; Cawthornand Strong,1974; Ohtaniet al., 1989]. The the adiabatictrajectories given in Miller et al. [1991b].Figure nearlychondritic initial isotopi6, ratios reported for bothhafnium 13 differsfrom the Work ofMiller et al. [1991b]only in thatthe [Gruauet al., 1990a]and neodymium [Gruau et al., 1990b]in meltingtemperatures for KLB1 determinedat StonyBrook Barberton komatiites is strong support for the residuum [Herzberget al., 1990;McFarlane et al., 1990] are higherthan interpretation. This is in complete accord with our phase thosereported by Takahashi[1986]. The importantpoirit is that equilibriumresults which show that garnetwas fractionatedas the adiabaticdT/dP trajectoriesapproximately parallel dT/dP for a residualphase (L + O1 + Gt + Cpx) duringpartial melting. the olivine liquidusat pressuresbelow about50 kbar. If they HERZBERG: DEPTH AND DEGREE OF MELTING OF KOMATIITES 4535

2500 buoyancybetween olivine and magmas[Agee and Walker, 1988; Miller et al., 1991a]. It is within this high pressureregion where melting couldbe mostly equilibriumrather than fractional [Herzberg, 1984]. The density contrast will increase upon adiabaticdecompression, and this will create opportunitiesfor 2ooo the melt to fractionateby segregationinto pools [Arndt, 1986]. The 80 to 100 kbar inferred from the geochemistryof the Barbertonkomatiites could be recordingonly the final stagesin equilibriummelting prior to separationof the liquid from its ':!!ii iii' • •Aunro , and the actualpressure of melt initiation may have been even greater. •::////• / •AO•e - 6. FORMATION OF KOMATIITES IN PLUMES 1500.• ...Mantle Peridotite Thermal models point to an Archcan mantle with a mean temperaturethat was only 100 to 150øChigher than at present [Turcott, 1980; Jarvis and Campbell, 1983; Calnpbell alwl Jarvis, 1984], and komatiitescould have formed by melting in 1000 0 100 200 300 hot rising plumes [Jarvis and Campbell, 1983' Arndt, 1986' Campbell et al., 1989; Miller et al., 1991a,b]. These would PRESSURE (kbar) have differed from present-dayhot spotsto the extent that the Fig. 13. Adiabatictrajectories and depthsof melting inferred for plumes intersected the solidus greater depths. An obvious Gorgona,Munro, and Barbertonkomatiites. The solidusand liquidus co•rollaryto this model is that the Earth was largely solidified temperaturesfor KLB1 are basedon experimentaldata of Takahashi throughoutthe Archcan, andthis hasconsiderable support in the [1986] adjustedto higher temperaturesobserved by Herzbet• et al. isotopiccompositions o f the earliestArchcan basalts, komatiites, [1990] andMcFarlane et al. [1990] andunreported data. Brokenlines and continentalcrust. The mantle sourceregions of theserocks are extrapolatedsolidus and liquidus locations. Adiabatic trajectories are based on the work of Miller et al. [1991b]. MORB adiabat is bounded had high Sm/Nd [Shirey and Hanson, 1986; McCulloch and betweenthe 1280øCpotential temperature of McKenzieand Bickle [1988] Compston,1981], and this is most simply interpretedas a andslightly higher temperatures indicated by garnetin thesource region depletionevent caused by approximatel.y10 to 20 % partial [Saltersand Hart, 1989]. Deflectionsin the adiabatsdue to •theolivine- melting of a fully solidifiedmantle soon al•er the Earth first modified spinel transformationare not shown. formed [Arndt, 1986; Chase and Patcherr,1988; Galer and GoMstein,1991; Smith and Ludden, 1989]. •lthoughthere is are preciselyparallel then it follows that olivine can be neither considerableevidence for a magmaocean stage of Earth history meltednor crystallized,and there can be no changein the melt [Wetherill,1990, Stevenson, 1987; Herzber• and Gaspar&, fractionwith decompression.Additionally, the compositionof 1991],the isotopic data and thermal modeli•{g [Miller et al., the eruptedmagma would. be identicalto the primarymagma 1991b]indicate that it solidifiedearly, well l•fore the onset of formedon the solidus.This may explain why manyGorgona the Archcan [see also Arndt, 1986]. and Barbertonkomatiites are so geochemicallysimilar to their The formation of the Barbertonkomatiites at pressures primarymagmas formed at 40 and 100 kbar respectively.The equal to or greaterthan the 80 to 90 kbar pressureof neutral principal modificationshave been in terms of olivine addition buoyancybetween olivine and magmas[Agee and Walker, 1988; and subtraction,a processthat couldhave occurred by settling Miller et al., 1991a] demonstratesthat there existedno density at•er eruption on the surface. barrier to the eruptionof komatiiticmagmas in the Archcan. If A reductionin pressuresimultaneously lowers dT/dP of the komatiites formed by percolationin a passive matrix like adiabats[McKenzie and Bickle, 1988; Miller et al., 1991b] and present-dayMORB [e.g., Takahashi,1990], it is raisesdT/dP of the olivine liquidus(Figure 13). This combined difficult to understandhow they could have erupted to the effect can result in an increased melt fraction with surface from 300 km depth. Arndt [1986] and Miller et al. decompression,the magnitudeo f whichwill dependon a precise [1991a,b] addressedthis dilema by consideringan origin in location of the adiabat; these are difficult to determine with mantle plumesor . Thermal buoyancydrives a plume precisionfor trajectoriesthat are locatedbetween the liquidus becausethe temperaturein it is higher than that of ambient and solidus[Miller et al., •1991b]. This kind of informationis mantle outside, and the plume will continue to rise even if a ß needed in order to understand how the Munro komatiites could parcel of komatiite liquid within it is denserthan a coexisting have evolved from initial melts to larger melt fractions in olivi.ne . The work of Miller et al. [1991a]indicates that equilibriumwith harzburgite;at 50 kbar the melt fractionmust komatiiteand peridotitcliquids have lower densitiesthan the increasefrom 35 % to 44 % (Figure 11). But at 40 kbar, the present-daybulk mantleto 400 kbar, a possibleexception being meltfraction needed to formsimilar high SiC 2 komatiite magmas a similarity in densities at around 133 kbar. Barberton dropsto about40 % (Figure10). The wide spectrumof Munro komatiitescontained within a hot plume would have been less komatiitecompositions can then be explainedby meltingthat dense than the ambient mantle at 100 kbar. variedfrom 35 % at 50 kbarto 40 % at 40 kbar. The important Comparedto the present-dayœarth, excess temperatures in point is that small increasesin melting attendingadiabatic a 3.5 Ga Barbertonplume would have been about 500øC, and decompressioncan causelarge changes in the geochemistryof excesstemperatures in a 2.7 Ga Munro plume about 250 to komatiiteswhen harzburgiteis eitherthe sourceor the residue 300øC (Figure 13). However, if the Archcan mantle was 100 to from a more fertile source. 150øC hotter than at present [Jarvis and Campbell, 1983; The 80 to 100 kbar pressuresof meltingfor the Barberton Turcotte, 1980], excesstemperatures in Archcan plumes would komatiitesare •imilar to the80 to 90kbar pressure of neutral have been correspondinglylower, about 150 to 350øC above 4536 HERZBERG: DEPTH AND DEGREE OF MELTING OF KOMATIITES ambientmantle temperatures.These are very similar to 200 to demonstratingthat it is a good indicatorof depth of melting. 300øCexcess temperatures that have been recommendedfor the Addition of FeO doesnot affect CaO/(CaO + A1203)of liquids present-dayHawaiian and Icelandic plumes [Sleep, 1990; at an isobaricinvariant point, but it lowersthe contentsof CaO, McKenzie, 1984; Griffiths and Campbell, 1991; Watsonand A1203,and SiO2. The MgO contentsof liquids on the solidus McKenzie, 1991; Liu and Chase, 1989]. are about 24 % at 50 kbar and 30 % at 100 kbar. Melt initiationwill occurat 40 kbar if present-dayplumes The meking of peridotite from 1 to 100 % has been havean excesstemperature that is 200ø above the 1280øCMORB simulated at a range of high pressures,and the liquids so potentialtemperature [McKenzie and Bickle, 1988; Figure 13], producedhave been compared to the geochemistryof komatiites. and the primary magmas produced will be Gorgona-type Komatiitesfrom GorgonaIsland can be understoodas relatively komatiiteswith about 20 % MgO. But the volcanicrocks from low degree partial melts (< 30 %), the products of Hawaii are typically basaltic instead. Clague et al. [1991] pseudoinvariantmelting involving L + O1 + Opx + Cpx + Gt reported 15 % MgO in Hawaiian picritic glasses, and on the solidusat 40 kilobars, about 130 km depth. recommendeda primary with 17 % MgO. These have Munro-type komatiitesare more difficult to understand. CaO/(CaO + A1203) and a major element chemistrythat is They have CaO/(CaO + A1203)that is similar to low degree similar to liquids formed at 25 to 30 kbar [Elthon and Scarfe, melts at 50 kbar (L + O1 + Opx + Cpx + Gt), moderate 1984; Table 5], indicatingan excesstemperature of only about degree melts (L + O1 + Opx), or high degree melts (L; L + 100-150 ø for . O1). Becauseof this problem, the depth of melting for the Attention is drawn to the possibility that the Gorgona Munro-type komatiites cannot be rigorously determined. And komatiiteswere formedin plumesthat rivaled the Munro plume unlike the Gorgonakomatiites, they contain a wide range of in depth of melting. This is of specialinterest because Re-Os values for CaO, A1203, and SiO2, variations that cannot be age dating [Walker et al., 1991] constrainsthe Gorgona attributed simply to olivine addition and subtraction. This komatiites to be 155 +/- 43 Ma, late Jurassic to early spectrumof compositionscan be understoodas liquids coexisting Cretaceous. Walker et al. [1991] noted the coincidencein this with a harzburgiteresidue (L + O1 + Opx) formedby < 40 % age with the breakupof Pangea,and suggestedthat this event melting of either average mantle peridotite or of harzburgite could have been causedby large-scalemantle upwelling. The sensu stricto. Although the pressure of melting cannot be error bars permit an equally intriguing possibility;that the rigorouslyconstrained, it was probably around 50 kbar, about Gorgonakomatiites are early to mid Cretaceousin age (120 165 km depth. The difficulty with the high degree melting Ma). If correct, they may have been a part of the Caribbean scenario(L + O1) is that it would haverequired a uniquesource [Storey et al., 1991] formed by the Galapagoshot spot region to explain each Munro komatiite and fails, therefore, on [Duncanand Hargraves, 1984]. An early Cretaceousage is also groundsof improbability. the time of formation of the gigantic oceanicplateaus in the Komatiitesfrom the BarbertonMountain Land have high Pacific [Larson, 1991; Tarduno et al., 1991]. At the present CaO/(CaO + A1203)and are similar to liquids formedby about time, there is no information on the compositionof the oceanic 50 % pseudoinvariantmelting (L + O1 + Gt + Cpx)of fertile crustbelow suchplateaus as OntongJava. However, in addition mantleperidotite in the 80- to 100-kbarrange (260- to 330-km). to basalts,komatiites of Cretaceousage may be fairly abundant Komatiiteshave been reasonably interpreted as the products in oceanicplateaus throughout the Pacific, and may provide a of extensiveor total melting of the mantle becausethey have a good analogue for understandingkomatiites in Archean geochemistrythat can be roughly explainedby the removal of greenstonebelts [Storeyel a/.,1991]. olivine from peridotite. But the resultsof this work point to Attentionhas alsobeen drawn to possiblesecular variations problemswith this interpretation. Instead, they can be more in the geochemistryof komatiites [Jahn et al., 1982]. High easily explained as partial melts (< 50 % melting) with a CaO/(CaO + A1203) and Gd/Yb characterizethe early Archean geochemistrythat was determined by highpressure crystal-liquid komatiites from Barberton and the Pilbara block, and these have phase equilibria. An important corollary is that komatiites agesof about3.5 Ga [e.g., Gruau et al., 1990a]. Munro-type cannotbe usedto place constraintson the geochemistryof their komatiites with mantle-like CaO/(CaO + A1203) and Gd/Yb source region, except in the most rudimentaryway. This is were formed in the late Archean, about 2.7 Ga ago [Nesbittet unfortunatebecause it meansthat they cannotbe usedto probe al., 1979; Jahn et al., 1982; Gruau et al., 1990a], and the the structure of the mantle, and address important questions Gorgonakomatiites are Mesozoic in age. The similarity in concerningthe geochemicalproperties of seismicdiscontinuities. conditions of melting between the Gorgona and Munro Secularvariations in the geochemistryof komatiitescould komatiites (Figure 13), and the possibility of Munro-like have formed in responseto a reductionin the temperatureand volcanismin the Pacific of Cretaceousage, point to an interior pressureof meltingwith time. The 3.5 Ga Barbertonkomatiites that cooled marginally over the last 2.5 billion years. Secular and the 2.7 Ga Munro-type komatiitescould have formed in coolingappears to have been expressedmainly in the early part plumesthat were hotterthan the present-daymantle by 500ø and of the Archean. 300ø, respectively. When excesstemperatures are this size, melting is deeper and volcanism changes from basaltic to 7. CONCLUSIONS komatiitic. The komatiites from Gorgona Island, which are High pressuremelting experiments on compositionsin the Mesozoic in age, may be representativeof komatiites that are systemCaO-MgO-A1203-SiO2 demonstrate that liquids on the predicted to occur in oceanic of Cretaceous age solidusare komatikic in the 50-to 150-kilobar pressurerange. throughoutthe Pacific [Storeyet al., 1991]. The effectof pressureis to expandthe stabilityfield of garnetat APPENDIX the expenseof olivine, clinopyroxene,and orthopyroxene.This causesa very pronouncedincrease to occur in CaO/(CaO + With the exceptionof the extremely fresh komatiitesfrom A1203) for liquids that are multiply saturatedin Cpx + Gt, [Nisbet et al., 1987], most have been HERZBERG: DEPTH AND DEGREE OF MELqlNG OF KOMAIIlT ,ES 4537

inqetamorphosedto greenschist and amphibolitefacies and have I I I I I I I I I I I I I I suffered variable amounts of alteration by hydrothermal solutions. An evaluation of these effects is required before Barberton Komatiites confidencecan be placed in any igneousinterpretation. Smith and Erlank [1982] documented some possible problems in understandingCaO and A1203in the Barbertonkomatiites, and de Wit et al. [1987] have documentedimportant evidence for widespreadmobilization of Ca andSi, amongstothers elements. The effects of alteration on the geochemistryof volcanic I _ rockshave been examinedby laboratoryexperimentation and by ALTERATION carefully controlledfield studies,and the reader is referred to the excellent discussionof this work in de Wit et al. [1987]. Experimentalstudies have monitoredgeochemical changes in 0 seawaterbefore and after reactionwith , andesitc,, 40 45 50 55 and peridotitc at elevated temperatures and pressures [e.g., Si02 Hajash and Chandler, 1981; Bischoff and Dickson, 1975; Fig. A1. CaO/AI:O3 (weight % ratio) versus SiO: for Barberton Hajash, 1975]. The resultsdemonstrate that importantquantities komatiites. of Ca and Si can be removed from basalts and peridotitesand Mg can be added. Studieshave also been made on samplesof fresh and hydrothermallyaltered pillow basaltsfrom the Mid- Atlantic Ridge [Humphrisand Thompson,1978], and the results [ [ [ I I [ [ [ I I I I [ I agree with the experimentalwork in that Si and Ca can be leachedfrom the rock while Mg can be taken up. One of the Munro-Type Komatiites important findings of the field studiesis that mobilization of these elementsis not confinedto a hand specimenor outcrop scale. While leachingof Ca can be accompaniedby localized .. precipitation,in generalthe systemis open on an outcropscale. This processhas addedimportant quantities of Si and Ca to the I _ world's oceans, and resulted in the depositionof Fe, Mn, Cu, ß ß and Zn on the floor [Von Datum et al., 1985; Hart, 1973; ß

Reed, 1983]. ALTERATION These studiesagree that Ca and Si are more mobile than A1 I I I I I I I I I I I I I I in hydrothermalsolutions. A positivecorrelation of CaO/A1203 40 45 $0 55 and SiO2 is therefore an indication of alteration. This is observedfor the Barberton komatiites in Figure A1, and this SiO2 supportsthe suggestionof de Wit et al. [1987] that Ca and Si Fig. A2. CaO/AI203 (weight % ratio) versusSiO 2 for Munro-type have indeed been removed from the komatiites. The Munro komatiites. komatiitesare notablefor having CaO/A1203equal to 1.0 [e.g., Jahn et al., 1982] and, unlike the Barbertonkomatiites, Figure A2 shows that it is largely independentof SiO2. It can be 40 I I I concludedfrom thesecomparisons that the Barbertonkomatiites were more severly affectedby alterationthan the Munro-types. Additionand subtractionof olivine from a komatiiticmagma 30 Barberton Komatiites cannotchange CaO/A1203. Pseudoinvariantmelting can yield high pressureliquids with CaO/A1203 and SiO2 that are higher thanliquids generated at lower pressures.But to accountfor the spectrumof compositionsin Figure A1 would requirea rangeof pressuresthan span over 100 kbar, and is unreasonableon geological grounds. One other igneous process than can 10 potentiallyexplain the data is by advancedmelting through the sequenceL + O1 + Cpx + Gt --> L + O1 + Gt --> L + O1. In projection, highly magnesiankomatiites from 0 i i I Barbertonwould differ from the others in having Munro-like 0 I 2 3 CaO/A1203. This observationwas originallymade by Smithand Erlank [1982], and correctly dismissedas another artifact of CEIO•1203 alterationbecause no changewas observedfor A1203/TiO2. This Fig. A3. AI203/TiO2 versusCaO/AI203 (weight % ratios) for Barberton parameteris shown also in Figure A3. It can be seenthat for komatiites. Barberton komatiites having CaO/A1203 > 1.5, the ratio A1203/TiO2 is typically around11, and this contrastswith about over 20. It is this part of the Barbertonpopulation that is 20 for Munro-typekomatiites [Jahn et al., 1982; Gruau et al., suspectof being altered, and the highly scattereddata have no 1990a]. But for Barbertonkomatiites having CaO/A1203< 1.5, obvious igneous explanation. Barberton komatiites with the ratio A1203/TiO2 stayseither unchangedor it increasesto unusuallyhigh A1203/TiO2 may be bad analyses[No T. Arndt, 4538 HERZBERG: DEYITI AND DEGREE OF MELTING OF KOMATIITES

personalcommunication, 1991] or theymay be an artifactof Ti Boyd,F.R., Compositionaldistinction between oceanic and cratonic removalduring hydrothermal alteration. Barbertonkomatiites ,Earth Planet. Sci. Lett., 96, 15-26, 1989. havingCaO/A1203 > than1.5 areabout 50 % of thepopulation, Brooks,C., andS.R. Hart, On the signficanceof komatiite,Geology, 2, and are probablythe leastaltered of the database. The high 107-110, 1974. pressureigneous interpretation offered above has been based on Campbell,I.H., andG.T. Jarvis, and early crustal evolution,Precambrian Res., 26, 15-56, 1984. these "clean" Barberton samples. Campbell,I.H., R.W. Griffiths,and R.I. Hill, Meltingin an Archaean Inspectionof Figure 12 showsthat the cleanBarberton mantleplume: heads it's basalts,tails it's komatiites,Nature, 339, komhtiitescan be explainas liquids formedby about52 % 697-699, 1989. pseudoinvariantmelting at around100 kbar. But a detailed Ca-wthorn,R.G., and D.F. Strong,The petrogenesisof komatiitesand examinationof these plots shows that while this adequately relatedrocks as evidencefor a layeredupper mantle, Earth. Planet. explainsMgO, A1203,and SiO• the contentsof CaO that are Sci. Lett., 23, 369-375, 1974. predictedare mostly higher than those that are observed even for Chase,C.G., and P.J. Patchett,Stored /ultramafic and early the clean Barbertondata set. Although these differencesare Archeanmantle depletion, Earth Planet. Sci. Lett., 91, 66-72, 1988. certainlywithin the experimentalerror of thehigh pressure data Clague,D.A., W.S. Weber, and J.E. Dixon, Picriticglasses from base,they may indicate that even the cleanBarberton komatiites Ha-waii,Nature, 353, 553-556, 1991. have lost some CaO. Davis, B.T.C., The systemDiopside-Forsterite-Pyrope at 40 kilobars, YearBook Carnegie Inst. Washington,63, 165-171,1964. Davis, B.T.C, and J.F. Schairer, Melting relationsin the join Acknowledgments. This researchwas supportedby NSF diopside-forsterite-pyropeat 40 kilobars and at oneatmosphere, Year grantEAR 89-16836 to C. Herzberg. The resultson garnetand BookCarnegie Inst. Washington,64, 123-126,1965. pyroxenewere reportedby Hetzberg and Gasparik[1991], and de Wit, M.J., R.A. Hart, and R.J. Hart, The Jamesto-wnophiolite supportedin part by NSF EAR 90-03748 to T. Gasparik. The complex,Barberton mountain belt: a sectionthrough 3.5 Ga oceanic high pressureexperiments reported in thispaper were performed crust,J. AfricanEarth Sci., 6, 681-730, 1987. in the Stony Brook High PressureLaboratory which is jointly Duncan,R.A., and R.B. Hargraves,Plate tectonicevolution of the supportedby the National ScienceFoundation (EAR 89-17563) Caribbeanregion in the mantlereference frame, The Caribbean- and the State University of New York at Stony Brook. Thanks SouthAmerican plate boundary and regional , Mem. Geol. are extended to Nick Arndt and Dan McKenzie for critical Soc. Am., 162, 81-93, 1984. commentson a first draft of this paper, and to Tibor Gasparik Echeverria, L.M., Komatiites from Gorgona Island, Colombia, in: and Paul Hess for discussions. Koman'ites,edited by N.T. Arndt and E.G. Nisbet, pp. 199-209, GeorgeAllen and Un-win,London, 1982. Elthon, D., and C.M. Scarfe, High-pressurephase equilibria of a high- REFERENCES magnesiabasalt and the genesis of primary oceanic basalts, Am. Agee,C.B., andD. Walker,Static compression and olivine floatation Mineral., 69, 1-15, 1984. in ultrabasicsilicate liquid, J. Geophys.Res., 93, 3437-3449,1988. Falloon, T.J., D.H. Green, C.J. Hatton, and K.L. Harris, Anhydrous Aitken,B.G., and L.M. Echeverria, and geochemistryof partialmelting of a fertile and depletedperidotite from 2 to 30 kb and komatiitesand tholeiitesfrom GorgonaIsland, Colombia, Contrib. applicationto basaltpetrogenesis, J. Petrol., 29, 1257-1282,1988. Mineral. Petrol., 86, 94-105, 1984. Fujii, T., M. Tachikara, and K. Kurita, Melting experimentsin the Anderson,D.L., Compositionof the Earth, Science,243, 367-370, systemCaO-MgO-AI203-SiO 2 to 8 GPa: Constraintsto the origin of 1989o komatiites,Eos Trans. AGU, 70, 483, 1989. Andersen,O., The systemanorthite-forsterite-silica, Am. J. Sci., 4th Galer, S.J.G., and S.L. Goldstein,Early mantle differentiationand its ser., 39, 407-454, 1915. thermal consequences,Geochim. Cosmochim.Acta, 55, 227-240, Arndt,N.T., Meltingrelations of ultramaficlavas (komatiites) at 1 atm 1991. andhigh pressure, Year Book Carnegie Inst. Washington,75, 555- Gansser,A., V.J. Dietrich, and W.E. Cameron, Palaeogenekomatiites 562, 1976. from GorgonaIsland, Nature, 278, 545-546, 1979. Arndt,N.T., Ultrabasicmagmas and high-degree melting of themantle, Gasparik, T., Transformation of enstatite-diopside-j adeite Contrib. Mineral. Petrol. 64, 205-221, 1977. pyroxenesto garnet,Contrib. Mineral. Petrol., 102, 389-405, 1989. Arndt,N.T., Komatiites:a dirty -windo-wto the Archeanmantle, Terra Gasparik,T., Phaserelations in the transitionzone, J. Geophys.Res., Cognita,6, 59-66, 1986. 95, 15,751-15,769, 1990. Arndt,N.T., A.J. Naldrett,and D.R. Pyke, Komatiiticand iron-rich Green, D.H., Archean greenstone belts may include terrestrial tholeiiticlavas of MunroTownship, Northeast Ontario, J. Petrol.,18, equivalentsof lunar maria?,Earth Planet. Sci. Lett., 15, 263-270, 319-369, 1977. 1972. Arth, J.G., N.T. Arndt, and A.J. Naldrett,Genesis of Archean Green, D.H., Genesisof Archeanperidotitic magmas and constraintson komatiitesfrom Munro Township,Ontario: traceelement evidence, Archeangeothermal gradients and tectonics, Geology, 3, 15-18, 1975. Geology,5, 590-594, 1977. Green, D.H., and A.E. Ringwood, The genesisof basalticmagmas, Bertka, C., and J.R. Hollo-way,Martian mantleprimary melts:an Contrib. Mineral. Petrol., 15, 103-190, 1967. experimentalstudy of iron-richgarnet lherzolite minimum melt Griffiths, R.W., and I.H. Campbell, On the dynamicsof long-lived composition,Proc. Lunar Planet. Sci. Conf., 18, 723-739,1988. plume conduitsin the convectingmantle, Earth. Planet. Sci. Lett., Bickle, M.J., The magnesiumcontents of komatiiticliquids, In: 103, 214-227, 1991. Komatiites,edited by N.T. Arndt, and E.G. Nisbet,pp. 479-494, Gruau, G., C. Chauvel, N.T. Arndt, and J. Cornichet, Aluminum GeorgeAllen and Unwin, London,1982. depletionin komatiitesand garnetfractionation in the early Archean Bickle, M.J., C.E. Ford, and E.G. Nisbet, The petrogenesisof mantle: isotopicconstraints, Geochirn. Cosmochim. Acta, peridotitic komatiites: evidence from high-pressuremelting $4, 3095-3101, 1990a. experiments,Earth Planet.Sci. Lett., 37, 97-106, 1977. Gruau, G., C. Chauvel,and B.M. Jahn,Anomalous Sm-Nd agesfor the Bischoff,J.L., and F.W. Dickson,Seawater-basalt interaction at 200'C early Archean OnverwachtGroup Volcanics: Significanceand and 500 bars: implicationsfor origin of sea-floorheavy-metal petrogeneticimplications, Contrib. Mineral. Petrol., 104, 27-34, depositsand regulationof seawaterchemistry, Earth Planet.Sci. 1990b. Lett., 25, 385-397, 1975. Hajash, A., Hydrothermalprocesses along mid-ocean-ridges:an HERZBERG: DEP_• AND DEGREE OF MELTING OF KOMATIITES 4539

experimentalinvestigation, Contrib. Mineral. Petrol., 53, 205-226, the Earth, (abstract), Proc. Lunar Planet Sci. Co•, 22, 1990. 1975. McKenzie, D.P., The generationand compactionof partial molten rock, Hajash,A., andGoW. Chandler, An experimentalinvestigation of high- J. Petrol., 25, 713-765, 1984. temperatureinteractions between seawater and rhyolite, , McKenzie, D., and M.J. Bickle, The volume and compositionof melt basaltand peridotite, Contrib. Mineral Petrol., 78, 240-254, 1981. generatedby extensionof the lithosphere,J. Petrol., 29, 625-679, Hart, R.A., Geochemicaland geophysicalimplications of the reaction 1988. between seawater and the , Nature, 243, 76-78, 1973. Miller, G.H., E.M. Stolper, and T.J. Ahrens, The equationof stateof Herzberg, C.T., The solubilityof olivine in basalticliquids: an ionic a molten komatiite, 1, Shock wave compressionto 36 GPa, J. model, Geochim. Cosmochim.Acta, 43, 1241-1251, 1979. Geophys.Res., 96, 11,831-11,848, 1991a. Herzberg, C.T., Chemical stratificationin the Earth, Earth Miller, G.H., E.M. Stolper, and T.J. Ahrens, The equationof stateof Planet. Sci. Lett., 67, 249-260, 1984. a molten komatiite, 2, Application to komatiite petrogenesisand the Herzberg, C., and T. Gasparik,Garnet and pyroxenesin the mantle:a Hadean mantle, J. Geophys.Res., 96, 11,849-11,864, 1991b. testof themajorite fractionation hypothesis, J. Geophys.Res., 96, 16, Mysen, B.O., and I. Kushiro, Compositionalrelations of coexisting 263-16,274, 1991. phaseswith degree of melting of peridotite in the upper mantle, Am. Herzberg, C.T., and M.J. O'Hara, Origin of mantle peridotiteand Mineral., 62, 843-865, 1977. komatiiteby partialmelting, Geophys. Res. Lett., 12, 541-544, 1985. Nesbitt, R.W., and S.-S. Sun, Geochemistryof Archaean spinifex- Herzberg,C.T., and E. Ohtani, Origin of komatiiteat high pressures, textured peridotitesand magnesianand 1ow-magnesiantholelites, Earth Planet. Sci. Lett., 88, 321-329, 1988. Earth Planet. Sci.. Lett., 31,433-453, 1976. Herzberg, C.T., M. Feigenson,C. Skuba, and E. Ohtani, Majoritc Nesbitt, R.W., S.-S. Sun, and A.C. Purvis, Komatiites: geochemistry fractionationrecorded in the geochemistryof peridotitesfrom South andgenesis, Can. Mineral., 17, 165-186,1979. Africa, Nature, 332, 823-826, 1988. Nisbet, E.G., N.T. Arndt, MJ. Bicide, W.E. Cameron, C. Chavel, M. Herzberg, C., T. Gasparik,and H. Sawamoto,Origin of mantle Cheadle, E. Hegner, T.K. Ky•er, A. Martin, R. Rennet, and E. peridotite:constraints from meltingexperiments to 16.5 GPa, J. Roetidier, Uniquely fresh 2.7 Ga komatiitezfrom the Belingwe Geophys.Res., 95, 15,779-15,803, 1990. greenstonebelt, Zimbabwe, Geology,I5, 1147-1150, 1987. Hess, P.C., Komatiites and the Archean mantle (abstract), Proc. Lunar Nisbet, E.G., and D. Walker, Komatiite• and the •tructure of the Planet. Sci. Conf., 21,501-502, 1990. Archaean mantle, Earth Planet. Sci. Lett., 60, 105-113, 1982. Howells, S.,C. Begg, and M.J. O'Hara, Crystallizationof somenatural Obata, M., and J.S. Dickey Jr., Phase relationsof mafic layers in the and garnetiferous ultrabasic rocks at high pressure and RondaPeridotite, Year Book CarnegieInst. Washington,75, 562-566, temperature,Phys. Chem.Earth, 9, 895-902, 1975. 1976. Humphris, S.E., and G. Thompson,Hydrothermal alteration of oceanic O'Hara, M.J., The bearing of phaseequilibria studiesin syntheticand basaltsby seawater,Geochim. Cosmochim. Acta, 42, 107-125, 1978. natural systemson the origin and evolution of basic and ultrabasic Ito, K., and G.C. Kennedy, Melting and phase relations in a natural rocks, Earth Sci. Rev., 4, 69-133, 1968. peridotiteto 40 kilobars,Am. J. Sci., 265, 519-538, 1967. O'Hara, M.J., and H.S. Yoder, Jr., Formation and fractionation of basic Jahn, B.-m., P. Vidal, and G.R. Tilton, Archean mantle heterogeneity: magmasat high pressures,Scott. J. Geol., 3, 67-117, 1967. evidence from chemical and isotopic abundancesin Archean igneous O'Hara, M.J., M.J. Saunders, and E.P.L. Mercy, Garnet-peridotite, rocks, Philos. Trans. R. Soc. London A, 297, 353-364, 1980. primary ultrabasic magma and eclogites; interpretation of upper Jahn, B.-m., G. Gruau, and A.Y. Glikson, Komatiites of the Onverwacht mantleprocesses in ,Phys. Chem.Earth, 9, 571-604, 1975. Group, S. Africa: REE geochemistry, Sm/Nd age and mantle Ohtani, E., Generation of komatiite magma and gravitational evolution, Contrib. Mineral. Petrol., 80, 25-40, 1982. differentiationin the deepupper mantle,Earth Planet. Sci. Lett, 67, Jarvis, G.T., and I.H. Campbell, Archean komatiites and geotherms: 261-272, 1984. solutionto an apparentcontradiction, Geophys. Res. Lett., 10, 1133- Ohtani, E., I. Kawabe, J. Moriyama, and Y. Nagata, Partitioning of 1136, 1983. elementsbetween garnet and melt and implicationsfor Kushiro, I., Y.. Syono, and S. Akimoto, Melting of a peridotirenodule petrogenesisof komatiite, Contrib. Mineral. Petrol., 103, 263-269, at high pressuresand high water pressures,J. Geophys. Res., 73, 1989. 6023-6029, 1968. Presnall, D.C., and T. Gasparik, Melting of enstatite0VIgSiO3) from 10 Larson, R.L., Latest pulse of Earth: evidence for a mid-Cretaceous to 16.5 GPa andthe forsterite(Mg2SiO4) - majorite(MgSiO•) eutectic superplume,Geology, 19, 547-550, 1991. at 16.5 GPa: implicationsfor the origin of the mantle,J. Geophys. Lesher, C.E., and D. Walker, Cumulate maturationand melt migration Res., 95, 15,771-15,778, 1990. in a temperaturegradient, J. Geophys.Res., 93 10, 295-10, 311, Presnall, D.C., J.R. Dixon, T.H. O'Donnell, and S.A. Dixon, 1988. Generationof mid-oceanridge tholeiites,J. Petrol., 20, 3-35, 1979. Litvin, Y.A., High pressuremelting relations in the garnet-lherzolitic Pyke, D.R., A.J. Naldrett, and O.R. Eckstrand, Arcbean ultramafic system olivine-orthopyroxene-clinopyroxene-garnet,Phase Trans- flows in Munro Township,Ontario, Geol. Soc. Ant. Bull., 84, 955- formationsat high pressuresand high temperatures: applicationsto 978, 1973. geophysicaland petrological problems, Abstracts of the secondJapan- Reed, M.H., Seawater-basaltreaction and the origin of greenstonesand USSRSytnposium, Misasa, 48-50, 1989. relatedore deposits,Econ. Geol., 78, 466-485, 1983. Liu, M., and C.G. Chase, Evolution of midplate hot spot swells: Salters, V.J.M., and S.R. Hart, The hafnium paradox and the role of numericalsolutions, J. Geophys.Res., 94, 5571-5584, 1989. garnet in the sourceof mid-ocean-ridgebasalts, Nature, 342, 420- Longhi, J., Liquidus equilibria and solid solution in the system- 422, 1989. CaA12Si208- Mg2SiO4 - CaSiO3- SiO2 at low pressure,Am. J. Sci., Shirey, S.B., and G.N. Hanson, Mantle heterogeneityand crustal 287, 265-331, 1987. recycling in Arcbean -greenstonebelts: evidence t'rom Nd Maaloe, S., and K. Aoki, The major elementcomposition of the upper isotopes and trace elements in the Rainy Lake area, Superior mantle estimated from the composition of , Contrib. Province, Ontario, Canada, Geochim. Cosrnochim.Acta, 50, 2631- Mineral. Petrol., 63, 161-173, 1977. 2651, 1986. McCulloch, M.T., and W. Compston, Sm-Nd age of Kambalda and Sleep, N., Hotspots and mantle plumes: some phenomenology,J. Kanowna greenstonesand heterogeneityin the Archaean mantle, Geophys.Res., 95, 6715-6736, 1990. Nature, 294, 322-327, 1981. Smith, A.D, and J.N. Ludden, Nd isotopicevolution of the precambrian McFarlane, E.A., M.J. Drake, and C. T. Herzberg, Magnesiowustite/ mantle, Earth Planet. Sci. Lett., 93, 14-22, 1989. melt and majorire/meltpartitioning and the early thermalhistory of Smith, H.S., and A.J. Erlank, Geochemistryand petrogenesisof 4540 HERZBERG: DEPTH AND DEGREE OF MELTING OF KOMATIITES

komatiites from the Barberton greenstonebelt, South At¾ica, In: Turcotte, D.L., On the thermal evolution of the Earth, Earth Planet. Sci. Komatiites, edited by N.T. Arndt and E.G. Bickle, pp. 347-397, Lett., 48, 53-58, 1980. GeorgeAllen and Unwin, London, 1982. Viljoen, R.P., and M.I. Viljoen, Evidence for the compositionof the Stevenson,D.J., Origin of the moon - the collisionhypothesis, Annu. primitive mantle and its productsof partial melting from a studyof Rev. Earth Planet. Sci. 15, 271-315, 1987. the rocks of the BarbertonMountain Land, Geol. Soc. S. Afr. Spec. Storey, M., J.J. Mahoney, L.W. Kroenke, and A.D. Saunders,Are Publ., 2, 275-296, 1969. oceanicplateaus sites of komatiiteformation?, Geology, 19, 376-379, Von Damm, K.L., J.M. Edmond, B. Grant, C.I. Measures, B. Walden, 1991. and R.F. Weiss, Chemistry of submarinehydrothermal solutions at Stosch,H.-G., and H.A. Seck, Geochemistryand mineralogyof two 21øN, East Pacific Rise, Geochim. Cosmochim.Acta, 49, 2197-2220, spinel peridotitc suites from Dreiser Weiher, West Germany, 1985. Geochim. Cosmochim.Acta, 44, 457-470, 1980. Walker, R.J., L.M. Echeverria,S.B. Shirey, and M.F. Horan, Re-Os Sun, S.-S., and R.W. Nesbitt, Petrogenesisof Archaeanultrabasic and isotopicconstraints on the origin of volcanicrocks, GorgonaIsland, basic volcanics:evidence from rare earth elements,Conttib. Mineral. Colombia: Os isotopic evidence for ancient heterogeneitiesin the Petrol., 65, 301-325, 1978. mantle, Contrib. Mineral. Petrol., 107, 150-162, 1991. Takahashi, E., Melting of a dry peridotitc KLB-1 up to 14 GPa: Watson, S., and D. McKenzie, Melt generationby plumes:a study of Implicationson the origin of peridotiticupper mantle, J. Geophys. Hawaiian volcanism,J. Petrol., 32, 501-537, 1991. Res., 91, 9367-9382, 1986. Wei, K., R.G. Tronnes, and C.M. Scarfe, Phase relationsof aluminum- Takahashi,E., Speculationson the Archcanmantle: missing link between undepletedand aluminum-depletedkomatiites at pressuresof 4-12 komatiiteand depletedgarnet peridotitc,J. Geophys.Res., 95, 15, GPa, J. Geophys.Res., 95, 15,817-15,828, 1990. 941-15,954, 1990. Wetherill, G., Formation of the Earth, Annu. Rev. Earth Planet. Sci., Takahashi, E., and I. Kushiro, Melting of a dry peridotitc at high 18, 205-256, 1990. pressuresand basalt magmagenesis, Ant. Mineral., 68, 859-879, 1983. C. Herzberg,Department of GeologicalSciences, Rutgers University, Tarduno,J.h., W.V. Sliter,L. Kroenke,M. Leckie,H. Mayer,J.J. Wright Laboratory,Busch Campus, New Brunswick,NJ 08903. Mahoney,R. Musgrave,M. Storey,and E.L. Winterer,Rapid formationof OntongJava plateau by Aptianmantle plume volcanism, Science, 254, 399-403, 1991. (ReceivedMarch 31, 1991; Taylor, H.J.C., Melting in the systemMgO-A1203-SiO 2 at 15 kb, Geol. revised November 18, 1991 Soc. Am. Bull., 84, 1335-1348, 1973. acceptedDecember 10, 1991.)