Platinum group elements in Kostomuksha komatiites and ...humayun/7Kosto2000.pdflower absolute PGE...

PII S0016-7037(00)00492-0

Platinum group elements in Kostomuksha komatiites and basalts:Implications for oceanic crust recycling and core-mantle interaction

IGOR PUCHTEL* and MUNIR HUMAYUN

Department of the Geophysical Sciences, The University of Chicago, 5734 S. Ellis Avenue, Chicago, IL 60637, USA

(Received January7, 2000;accepted in revised form June27, 2000)

Abstract—We report precise PGE (Os, Ir, Ru, Pt, and Pd) concentrations for komatiite-basalt lava sequencesof the Kostomuksha greenstone belt, and for the komatiite standards KAL-1 (Abitibi) and WITS-1 (Barber-ton). The flowtop breccia komatiites (MgO5 27.5%) from the Kostomuksha sequence, which represent thecomposition of the primary komatiite liquid, have Os5 1.56 0.1 ppb, chondritic (Os/Ir)N 5 1.066 0.05, andmoderately fractionated PGE patterns, (Pd/Os)N 5 6.56 0.4. Based on the KAL-1 standard, the Alexoprimary liquid had Os5 1.86 0.1 ppb, (Os/Ir)N 5 1.056 0.05, and (Pd/Os)N 5 8.56 1.0. PGEs in theKostomuksha komatiites exhibit incompatible behavior during magmatic differentiation, with bulk D valuesof Os, Ir 5 0.75, Pt5 0.52 and Pd5 0.

PGE abundances from Kostomuksha komatiites and basalts were used to calculate the Pt/Os and Re/Oscomposition of a magnesian Archean oceanic crust. The Pt-Re-Os isotopic evolution of this crust wasmodeled, indicating that coupled186,187Os enrichments observed in some modern plumes require a non-crustalsource. A previously proposed model for core-mantle interaction by physical admixture was evaluated andwas shown to produce mantle sources with high PGE abundances (Os5 6 to 14 ppb, Pd5 140 ppb for 1%outer core addition) and (Os/Ir)N $ 1, depending on the assumptions made. These features were not observedin the Kostomuksha komatiites (g187Os(T) 5 13.6), shown here to be derived from a reservoir with;25%lower absolute PGE abundances than primitive mantle. If the radiogenic initial187Os/188Os in the Kosto-muksha komatiite is an outer core signature, then core-mantle interaction cannot have involvedphysicaladmixtureof outer core material into mantle plume sources. Instead, it must have occurred in the form ofisotopic equilibration between liquid outer core and solid mantle at the core-mantle boundary.Copyright ©2000 Elsevier Science Ltd

1. INTRODUCTION

Due to their highly siderophile and generally compatiblenature, platinum group elements (PGE5 Ru, Rh, Pd, Os, Ir, Pt)are a powerful tool for studying processes of the Earth’s ac-cretion, core-mantle differentiation, mantle evolution andmagma genesis. In addition, the development and applicationof the siderophile element radiogenic systems,187Re-187Os and190Pt-186Os (see review in Shirey and Walker, 1998 and Walkeret al., 1997a) provide both tracer and chronological informationon differentiation processes affecting these elements. However,high-quality studies of PGE concentrations still cover a limitedrange of mantle samples (e.g., Pattou et al., 1996; Rehka¨mperet al., 1997; Rehka¨mper et al., 1999a; Rehka¨mper et al., 1999b)and do not report Os abundances.

The behavior of PGEs during partial melting and crystalfractionation can be studied from PGE compositions of mantle-derived lavas such as basalts and komatiites. Rehka¨mper et al.(1999b) showed that PGE abundances in basalts were drasti-cally modified by extensive PGE fractionation in these sulfide-saturated magmas, rendering basalts unsuitable for determiningsource PGE compositions. We used komatiites as indicators ofmantle source PGE composition since these lavas are sulfide-unsaturated primary melts.

Komatiites formed as a result of extensive melting in hotmantle plumes (e.g., Campbell and Griffiths, 1990; Storey et

al., 1991). The degrees of melting during komatiite formationwere very substantial (30 to 50%) enabling the melts to extractlarge proportions of compatible elements (e.g., PGE). Komati-ites are low-viscosity melts, rapidly erupted, at temperatureswell in excess of their liquidus temperatures, assuring thatbasically no differentiation occurred before eruption (Arndt,1976; Herzberg, 1995). Differentiation that occurs in individualflows during emplacement can be assessed by thorough sam-pling from representative sections, such as afforded by drillcores.

In this study, we report precise PGE data for komatiites fromthe 2.8 Ga Kostomuksha and 2.7 Ga Abitibi greenstone belts.Data for Kostomuksha basalts associated with the komatiitesare also presented. On the basis of field relationships, andlithophile trace element and Nd-Pb isotope characteristics, thesubmarine lava sequences of these greenstone belts have beenproposed to be remnants of upper crustal portions of Archeanplume-derived oceanic plateaux (Desrochers et al., 1993; Pu-chtel et al., 1998); both contain thick sequences of well-pre-served komatiites free of crustal contamination. Kostomukshaand Abitibi komatiites were derived from mantle sources withdistinct187Os/188Os isotope compositions. While the source ofthe Abitibi komatiite was characterized by a chondritic initial187Os/188Os (Walker et al., 1988; Shirey, 1997), Os data for theKostomuksha komatiites indicate that the source of these lavashad a suprachondritic initialg187Os of 13.66 1.0 (Puchtel etal., 2000a). This radiogenic signature, though extremely rare inthe Archean (the only other occurrence being the 3.3 Ga Non-dweni greenstone belt in South Africa, Shirey et al., 1998), is

*Author to whom correspondence should be addressed ([email protected]).

Pergamon

Geochimica et Cosmochimica Acta, Vol. 64, No. 24, pp. 4227–4242, 2000Copyright © 2000 Elsevier Science LtdPrinted in the USA. All rights reserved

0016-7037/00 $20.001 .00

4227

ubiquitous in modern and recent plumes (Pegram and Alle`gre,1992; Hauri and Hart, 1993; Marcantonio et al., 1995; Widomand Shirey, 1996; Lassiter and Hauri, 1998). The enriched187Os/188Os isotopic composition of the Kostomuksha plumehas been recently interpreted (Puchtel et al., 2000a) in terms ofthe core-mantle interaction model of Walker et al. (1995).

Two major mechanisms have been proposed to explain theenriched187Os/188Os isotopic compositions of mantle plumesources. These include (1) recycling of oceanic crust (Walker etal., 1991; Hauri and Hart, 1993; Marcantonio et al., 1995;Roy-Barman and Alle`gre, 1995) and (2) entrainment of smallamounts of outer core material by rising mantle plumes (Walk-er et al., 1995; Walker et al., 1997a). Recently, efforts havebeen made to distinguish the effects of the two processes byexamining the Pt-Re-Os isotope systematics of modern andrecent plume-derived lavas (Walker et al., 1997a; Brandon etal., 1998). The crucial test of the two hypotheses lies in ob-taining a correlation between the186Os/188Os and187Os/188Osinitial ratios. Due to a substantially higher Pt/Re ratio in themodel outer core (Walker et al., 1995; Morgan et al., 1995;Smoliar et al., 1996) compared to the oceanic crust (90 to 100vs. ;7), melts derived from the sources that experienced core-mantle interaction plot along a much steeper trend ing187Os vs.«186Os space (Brandon et al., 1999).

One consequence of such physical addition of outer core tomantle sources is a very significant increase in PGE abun-dances. For example, Widom and Shirey (1996) showed thatOIB that experienced physical addition of small amounts(0.3%) of outer core would have elevated Os, and especiallyRe, abundances, but pointed out that this effect would bedifficult to recognize as compatible Os (and Ir) can be severelydepleted in even relatively primitive basaltic magmas. Theabsence of a correlation between Os abundances and187Os/188Os ratios was not regarded as conclusive evidence againstsuch physical addition by Shirey and Walker (1998) because ofpetrogenetic complexities in relating OIB Os abundances totheir source compositions. Hauri and Hart (1997) comparedRe/Yb ratios of OIB and MORB and argued for less than 0.3%outer core addition in most plumes, with a maximum limit of1.2% for all OIB lavas in their study. Rehka¨mper et al. (1999b)in their study of PGE abundances in OIB and MORB foundevidence of substantial PGE fractionation, limiting the recon-structions of mantle source PGE compositions. In this paper,we determine the PGE abundances in Kostomuksha and Abitibikomatiites and basalts and show how to reconstruct PGE com-positions of mantle sources. This work addresses the questionof PGE fractionation during mantle melting and differentiation;tests the hypothesis that subducted Archean oceanic crust couldbe the source of coupled radiogenic186,187Os in the mantle; andevaluates the potential effects of core-mantle interaction on thePGE abundances in plume-derived magmas.

2. GEOLOGICAL BACKGROUND AND PREVIOUSISOTOPE AND TRACE ELEMENT STUDIES

The geological setting of the Gimola-Kostomuksha green-stone belt in the NW part of the Karelian granite-greenstoneterrane has been described in detail elsewhere (e.g., Puchtel etal., 1997; Puchtel et al., 1998, and references therein) and isonly briefly summarized here. The belt consists of several

conjugate synforms that can be traced over a distance of.400km (Fig. 1) with the Kostomuksha synform being the largest ofthese. The latter includes at least two distinct lithotectonicterranes, one mafic igneous and the other sedimentary, sepa-rated by a major structural unconformity. The mafic terranecontains submarine-erupted komatiite and basalt lavas, volca-niclastic mafic and ultramafic sediments and numerous gabbroand peridotite sills. The sedimentary terrane is composed ofshelf-type rocks and banded iron formations (BIF). The kom-atiites and basalts from the mafic terrane have Sm-Nd andPb-Pb isochron ages of 28436 39 and 28136 78 Ma, respec-tively, and are intruded and overlain by felsic subvolcanic,volcanic and volcaniclastic rocks of dacite-rhyolite composi-tion with a U-Pb zircon age of 28216 1 Ma. Trace elementcharacteristics and Nd-Pb isotope compositions of the sourcereservoirs of the komatiites and basalts resemble those of recentPacific oceanic flood basalts (see review in Puchtel et al., 1998)at their respective time of eruption. Based on these lines ofevidence, the authors argued that the Kostomuksha lava se-quences represent remnants of the upper crustal part of anArchean oceanic plateau derived from partial melting in amantle plume head. Similar conclusions were reached by Des-rochers et al. (1993) for the Abitibi greenstone belt.

From the Kostomuksha belt, samples of flowtop breccias,and olivine spinifex-textured and cumulate komatiites from anumber of differentiated lava flows recovered from a 300 mdeep diamond drill hole (#15) were selected. In addition, sam-ples of pillow and massive basalts from this and two other drillholes (#79 and #124), and from surface outcrops, were studied.Most of these samples have been previously analyzed for majorand trace elements and Pb-Nd isotopes. Sample location andmore detailed geological information can be found in Puchtel etal. (1998). For a selected set of komatiite samples, Puchtel et al.(2000a) obtained a Re-Os isochron with an age of 27956 40Ma and a radiogenic initial187Os/188Os isotopic composition(g187Os(T) 5 13.66 1.0). From the Abitibi belt, we studiedan olivine spinifex-textured komatiite sample, KAL-1, from asingle Alexo lava flow (Arndt 1986) developed as a standard byN. T. Arndt, C. M. Lesher, and S. B. Shirey (personal commu-nication). This sample is stratigraphically equivalent to sampleM663 from the same flow.

3. ANALYTICAL TECHNIQUE

The NiS fusion procedure (Robert et al., 1971; Hoffman etal., 1978) employed here was modified from that used by Hauriand Hart (1993) for Os-isotope analysis, by the substitution ofLi-tetraborate as the flux. Ravizza and Pyle (1997) developed aNiS fusion procedure similar to that employed here, to analyzeboth Os-isotope composition and PGE abundances usingN-TIMS and quadrupole ICP-MS. We utilize many of their steps,but obtain PGE (Ru, Pd, Os, Ir, Pt) abundances by magneticsector ICP-MS. Two to four grams of sample powder werespiked with a mixed99Ru 1 110Pd 1 190Os 1 191Ir 1 198Ptspike, mixed with 10 g of 99.9% pure Alfa Aesart lithiumtetraborate, 1 g of 99.99% pure Aldricht Ni-powder (2100mesh) and 0.5 g of J.T. Baker sublimed sulfur powder, thenfused in porcelain Coors™ crucibles at temperatures of 1050 to1100°C for 75 to 90 min. to form a homogenous silicoboratemelt with immiscible NiS liquid. The melt was quenched and

4228 I. Puchtel and M. Humayun

the NiS beads were extracted and dissolved in 6N HCl. Double-distilled acids from Seastar™, and 18.2 MV deionized water,were used throughout the chemical procedures. Upon dissolu-tion, Ni forms a NiCl2 solution and the PGEs react with the

released H2S gas to form insoluble PGE sulfides. The solutionswere filtered through a cellulose membrane to recover theprecipitates and to remove most of the Ni. The cellulose mem-brane was then dissolved in 2 mL of hot (;100°C) aqua regia

Fig. 1. Geological sketch map of the Karelian granite-greenstone terrane, Baltic Shield (after Sokolov; 1987, with somemodifications). The study area in the inset map includes the Kostomuksha synform of the Gimola-Kostomuksha greenstonebelt.

4229PGEs on Kostomuksha komatiites

for 3 to 4 h to ensure the complete digestion of the PGEsulfides. This oxidation proved insufficient to completely re-move the dissolved cellulose nitrate, which tended to clog thenebulizer during sample introduction. Like Ravizza and Pyle(1997), we found that Os was only partially lost during the aquaregia digestion, so an aliquot of this liquid was taken up in 1NHCl and used for Os, Ir and Pt abundance determinations. Theremaining solution was dried down and the residue was sub-jected to aggressive oxidation, including a 5-h perchloric acidattack in sealed Savillex™ vials at 160°C, followed by dry-down at;200°C, to remove the cellulose nitrate. The residuewas processed through a cation exchange column to separateanionic PGEs from cationic impurities such as residual Ni, andMo and Cd, which were probably concentrated from the sampleduring the NiS fusion. A procedure modified from Jarvis et al.(1997) was followed. The sample was loaded in 0.5 mL of0.15N HCl on a quartz glass column filled with 2.0 mL ofcleaned, equilibrated AG 50WX8 (100 to 200 mesh) cationexchange resin. Anionic PGE chlorocomplexes were eluted inthe first 1.5 mL of 0.15N HCl, while Ni, Cd and Mo wereretained on the column. The column yield was;97% for allPGEs. From this aliquot, Ru, Pd, Ir and Pt were determined.Isotopic compositions of Ir and Pt were found to be identicalwithin mass-spectrometric error to those determined from thefirst aliquot, but significant improvements were obtained for Ru(interfered at masses 98, 100, 101 and 102 by Mo1 and NiAr1)and Pd (110Cd interference on110Pd spike isotope). Separationof the PGEs employed in several published procedures (Reh-kamper and Halliday, 1997; Pearson and Woodland, 2000) wasnot found to be essential.

Isotopic measurements were performed on a single-collector,magnetic sector, high-resolution ICP-MS, the Finnigan Ele-ment™. Sample solutions were introduced with a CETACMCN6000 nebulizer. Three sets of scans were carried out foreach analysis: an initial set of 20 scans of the masses: 7, 64, 97,98, 98, 100, 101, 102, 105, 106, 108, 110, 112, 190, 191, 192,193, 195, 198, 200, to assess the Ni signal, and to drive themagnet through the mass range used in the centering routine.The magnet was then parked at mass 97, and the Mo-Ru-Pd-Cdpeaks (97, 98, 99, 100, 101, 102, 105, 106, 108, 110, 112) werescanned 200 times by varying the accelerating potential

(EScan). Isobaric interferences from102Pd on102Ru were cor-rected, and those from NiAr1 and Cd were monitored at masses98 and 112. These corrections were generally,1%. The mag-net was then moved to mass 190, and 200 scans of the Os-Ir-Pt-Hg peaks (190, 191, 192, 193, 194, 195, 198, 200) wereobtained. Isobaric corrections of the minor Pt isotopes (190 and192) on both measured Os isotopes, and of198Hg (,1000 cps)on 198Pt, were performed. Although the Hg corrections wereminor (,,1%), the Pt interferences on Os isotopes were sub-stantial (30 to 100%) because of high Pt/Os ratios in thesolutions. The use of separate EScan routines for each of thePGE mass ranges allowed rapid scanning with 1 ms settlingtimes facilitating the collection of precise isotope ratios. Typ-ical count rates were 105 2 106 cps (except for Os;5 to9 3 104 cps), internal precisions were;0.2 to 0.5%, andexternal reproducibility of the ratios was comparable.

Total procedural blanks were determined using a reversestandard addition technique by analyzing two sets of low-PGEbasalts, 4 to 5 aliquots in each set (Table 1, Fig. 2). Blanks weredetermined from the intercepts of the linear regressions of eachset and were (average for the standard mass of reagents, in pg):Os, Ir 11, Ru 170, Pt 325, Pd 655 (Table 1). Blank correctionsapplied were estimated as,5% for the komatiites and 10 to20% for the basalts. Precision of the procedure, determined byduplicate analyses of 10 samples from this study, and byquadruplicate analyses of two komatiite standards KAL-1 andWITS-1 (Tables 2 and 4), was 2 to 5% (except for WITS-1),which included uncertainties from both analytical errors andsample powder heterogeneity. A few Ru analyses were irrepro-ducibly high, and are not considered further. Our data for theKAL-1 standard agree within;10% with those reported byRehkamper et al. (1999b), although we note that their Ru, Pd,and Pt abundances are systematically up to 12% higher. TheBarberton komatiite standard, WITS-1, shows poorer reproduc-ibility and substantial interlaboratory biases (Pearson andWoodland 2000; Tredoux and Mcdonald 1996). Hoffman et al.(1978) obtained yields for the NiS procedure that were 95% foreach of the PGEs. No attempt was made here to determinefusion yields. The accuracy of the analysis is guaranteed byspiking of the sample before fusion, attaining complete sample-

Table 1. PGE abundances in the analytical blank.

Sample Sample wt. (g)

Amount of analyte (pg)

Os Ir Ru Pt Pd

9436 0.5576 29.3 40.8 170.7 1018 14411.4920 58.6 99.1 412.3 2145 31862.0037 81.8 130.2 538.3 3119 31104.4032 155.0 258.4 1229 6041 6931

Blank (pg) 11 14 310 704

9443 0.5648 38.6 323.6 696.8 761.70.9181 59.6 434.0 879.9 10281.5500 88.8 504.6 1179 12242.2140 133.1 743.3 1929 17133.5049 201.5 1074 2429 1997

Blank (pg) 7.0 170 339 606

Average (pg) 11 11 170 325 655


spike equilibrium, and accurate calibration of the separatedisotope spikes using high-purity metals as standards.

4. RESULTS

PGE data for the komatiites and basalts are listed in Tables3 and 4 and are shown in Figures 3 and 4. Initial results werereported by Puchtel and Humayun (1999), and since that timewe have significantly improved the Ru and Pd results by cationcolumn clean-up.

4.1. Komatiites

The PGE abundances in the chilled samples (flowtop brec-cias and spinifex-textured komatiites) with MgO5 24 to 28%vary in a narrow range (by;10% about the mean). Relative tothe primitive mantle, they are depleted in Os-Ir and enriched inPt and Pd (Fig. 3). The olivine cumulates (MgO5 33 to 40%)are distinct in having slightly lower (except for sample 9490)Os and Ir abundances and substantially lower Pt and Pd con-centrations compared to the chilled samples. Both chilled sam-ples and the cumulates show unfractionated (Os/Ir)N 5

Fig. 2. The reverse standard addition procedure for assessing the total analytical blank using two low-PGE basalts (Table1). Symbols: open squares5 9443, filled diamonds5 9436.

Table 2. PGE abundances in komatiite standards (ppb).

Sample Os Ir Ru Pt Pd

KAL-1a 1.86 1.65 4.33 13.5 20.2KAL-1b 1.73 1.68 3.89 12.5 17.6KAL-1c 1.81 1.79 5.41 12.1 18.2KAL-1d* 1.87 1.71 4.73 12.1 15.1KAL-1 average 1.82 1.71 4.59 12.5 17.861 s 0.07 0.06 0.65 0.69 2.1KAL-1 [1] 1.71 5.10 13.9 20.061 s 0.03 0.37 0.60 0.40WITS-1a 0.846 1.21 3.79 6.31 6.51WITS-1b 0.718 1.14 2.96 5.12 4.97WITS-1c* 0.827 1.25 5.04 6.00 5.35WITS-1d* 0.779 1.31 5.40 5.90 5.21WITS-1 average 0.792 1.23 4.30 5.83 5.5161 s 0.057 0.07 1.12 0.51 0.68WITS-1 [2] 1.0–1.5 1.6 5.1 11.0 7.961 s — 0.2 1.2 2.0 0.9WITS-1 [3] 1.09 1.44 5.29 6.26 5.5961 s 0.04 0.09 0.23 0.61 1.18

Note. Data from [1] Rehka¨mper et al. (1999b), [2] Tredoux andMcDonald (1996), and [3] Pearson and Woodland (2000).

* Samples not processed through ion exchange columns.


1.006 0.05 and 0.986 0.05 and, with the exception of thecumulate sample 9490, are depleted in Os relative to Pd ((Pd/Os)N 5 6.46 0.6 and 4.26 0.4, respectively). The Alexokomatiite, KAL-1, has higher PGE abundances than the Kos-tomuksha chilled samples, with higher (Pd/Os)N 5 8.56 1.0,and chondritic (Os/Ir)N 5 1.056 0.05 (Table 2). The Barber-ton komatiite, WITS-1, has a distinctly non-chondritic (Os/Ir)N 5 0.656 0.04, the origin of which is unclear. The lack ofreproducibility for WITS-1 indicates the presence of PGEnuggets. The loss of iridosmine nuggets would explain thesubchondritic Os/Ir ratio, and the lower Os and Ir abundances

in our batch of powder relative to the consensus WITS-1composition of Tredoux and Mcdonald (1996).

4.2. Basalts

The basalts analyzed have MgO abundances varying be-tween 6.5 to 8.4%, a range typical of Archean tholeiites. Theyare strongly depleted in Os-Ir relative to the primitive mantle(B/PM 5 0.01 to 0.03), are characterized by highly fractionatedPGE patterns, (Pd/Os)N 5 616 10, and variable but basicallysubchondritic (Os/Ir)N 5 0.696 0.17. One sample (#9443) has

Table 3. PGE abundances in Kostomuksha basalts (ppb).

Sample Location MgO Os Ir Ru Pt Pd Ni Cr

91144 SW 7.42 0.063 0.076 0.150 2.15 3.78 134 26191145 SW 8.37 0.092 0.093 0.449 3.76 5.72 145 2909436 124/225.0 6.46 0.039 0.065 0.216 2.05 2.53 129 2619437 124/250.0 6.83 0.038 0.061 0.157 1.72 2.48 120 2539443 79/2.0 7.10 0.022 0.028 0.061 0.35 0.40 108 14494127 15/16.0 7.75 0.037 0.063 0.246 2.06 3.34 118 27694128 15/208.0 7.28 0.038 0.069 0.219 2.23 2.92 114 270

Note: Here, and in Table 4, analyses were recalculated on an anhydrous basis; MgO in wt%, Cr and Ni in ppm. For sample location, numeratorcorresponds to the drill hole number, and denominator—the depth. SW—surface outcrops in the SW part of the synform. For drill hole and outcroplocations see Figure 1 in Puchtel et al. (1998). Location of the drill hole #15 corresponds to the SK tick (Southern Komatiite section) in the same figure.

Table 4. PGE abundances in Kostomuksha komatiites (ppb).

Sample Location MgO Os Ir Ru Pt Pd Ni Cr

Komatiite flowtop breccia:9493 15/248.0 26.4 1.50 1.43 4.70 9.94 11.5 1244 31209493* 15/248.0 26.4 1.35 3.93 8.31 10.6 1244 312094111 15/261.2 27.4 1.65 1.46 4.49 9.45 12.2 1401 300994111* 15/261.2 27.4 1.45 1.38 4.24 7.84 10.9 1401 300994121 15/290.0 28.7 1.53 1.38 3.72 10.3 10.4 1352 315194121* 15/290.0 28.7 1.27 1.27 3.08 8.52 10.1 1352 3151Average 27.5 1.45 1.38 3.86 8.22 10.6 1333 3093Spinifex-textured komatiite:9469 15/197.7 25.8 1.53 1.59 3.97 9.19 12.0 1138 31279479 15/225.0 27.8 1.83 1.78 4.12 9.15 12.2 1232 30839479* 15/225.0 27.8 1.71 1.63 3.95 7.86 10.1 1232 30839495 15/250.0 26.6 1.29 1.27 3.77 8.62 10.6 1325 30969496 15/251.0 25.4 1.52 1.52 3.43 8.74 12.6 1103 29689496* 15/251.0 25.4 1.54 1.53 3.95 8.49 9.86 1103 29689497 15/252.0 24.8 1.43 1.51 3.83 8.38 11.7 1147 30229498 15/252.5 24.1 1.50 1.61 3.94 8.62 11.0 1097 32449498* 15/252.5 24.1 1.65 3.90 8.27 11.4 1097 324494104 15/257.3 27.9 1.56 1.46 3.79 8.11 9.97 1378 294694114 15/263.0 28.3 1.43 1.32 3.63 8.32 9.54 1456 289594114* 15/263.0 28.3 1.34 1.33 3.71 7.24 9.67 1456 289594123 15/292.5 25.5 1.52 1.47 3.40 9.42 11.4 1175 3250Average 26.6 1.48 1.46 3.80 8.46 10.8 1254 3076Olivine-cumulate komatiite:9490 15/237.5 37.3 6.19 6.15 4.93 6.82 5.37 2119 246694100 15/254.7 34.8 1.19 1.25 4.28 7.48 5.83 1889 261294100* 15/254.7 34.8 1.28 4.09 7.26 6.01 1889 261294116 15/265.9 32.7 1.48 1.41 11.2 7.34 7.33 1797 274394116* 15/265.9 32.7 1.31 1.33 3.94 7.99 7.62 1797 274394117 15/266.4 40.2 1.16 1.20 4.01 5.89 5.11 2413 227394118 15/267.0 37.1 1.37 1.28 3.80 6.20 6.05 2145 243094126 15/296.5 33.6 1.43 1.53 3.58 7.76 6.59 1861 276394126* 15/296.5 33.6 1.56 3.48 7.50 6.70 1861 2763

* Different fusions of the same sample.


Os and Ir abundances;50% lower, and Pt and Pd concentra-tions about an order of magnitude lower compared to themajority of the basalt samples.

5. DISCUSSION

5.1. Assessing PGE Mobility During PostmagmaticProcesses

PGE mobility can potentially be an important issue in studiesof metamorphosed Precambrian rocks. Kostomuksha se-quences were subjected to seafloor alteration and greenschistfacies metamorphism, which resulted in almost complete re-placement of igneous mineralogy by serpentine1chlorite as-

semblages, though primary volcanic textures and structuresremained well preserved. To assess the potential PGE mobilityin the Kostomuksha lavas, the approach developed for litho-phile elements by Arndt (1986) and Barnes et al. (1988) wasutilized here. These workers recognized that komatiite samplescontaining magmatic elemental abundances must follow oli-vine control lines. The latter are mixing lines for suites ofsamples derived from a single liquid containing variable pro-portions of olivine. Figure 4 shows PGE, Ni and Cr abundancesplotted against MgO, with olivine control lines drawn forcomparison. Note that even olivine-rich, and hence highlyserpentinized, cumulate samples follow the olivine control linesfor MgO, Ni, and Cr, indicating that these elements wereimmobile during secondary alteration. In Figure 4, the PGEsexhibit regular trends relative to MgO, that are interpreted hereas being magmatic in origin and representing olivine controllines for these elements. It is noteworthy that the application ofthis approach to lithophile elements in Kostomuksha had indi-cated that only the alkalis, Sr, Ba, Pb and U, were mobilizedduring alteration (Puchtel et al., 1998).

5.2. PGE Fractionation During Magmatic Differentiationand Partial Melting

The hosts of PGEs in the mantle are most likely not silicates,but sulfides and PGE minerals and alloys (Mitchell and Keays,1981; Hart and Ravizza, 1996; Humayun, 1998; Burton et al.,1999; Handler and Bennett, 1999; Rehka¨mper et al., 1999b).Oceanic basaltic magmas are thought to be sulfide-saturated(McGoldrick et al., 1979; Hertogen et al., 1980; Wallace andCarmichael, 1992), and experimentally determined Ni-Cu sul-fide melt-silicate melt partition coefficients for PGEs are on theorder of 104 (Peach et al., 1990; Fleet et al., 1996). Thissuggests that Ni-Cu sulfide fractionation is the main cause ofPGE depletion in basaltic magmas during differentiation,though this latter process cannot be responsible for fraction-ation of Os from Pd as both have very similar sulfide melt-silicate melt partition coefficients (Peach et al., 1994). How-ever, since differentiation of basaltic magmas is usuallyaccompanied by olivine and chromite fractionation, it has beenassumed that these phases controlled PGE fractionation inbasalts and, by analogy, in komatiites (Crocket and MacRae,1986; Brugmann et al., 1987; Barnes and Picard, 1993).

The available PGE data for olivines and chromites are quiteambiguous. Ross and Keays (1979) observed elevated concen-trations of Ir in Kambalda olivines (5.7 ppb vs. 1.3 ppb in hostlava) and Bru¨gmann et al. (1987) estimated a bulk DOsof ;1.8for the olivines from Gorgona and Alexo komatiites. Similarresults were reported by Barnes and Picard (1993) for CapeSmith Fold Belt lavas and by Crocket and MacRae (1986) forPyke Hill komatiites. This is in contrast with direct measure-ments of Os concentrations in komatiitic and picritic olivinesfrom Gorgona Island (Walker et al., 1999), Pechenga (Walkeret al., 1997b) and Vetreny belt (Puchtel et al., 2000b), indicat-ing that Os was incompatible with olivine.

Chromite is usually substantially enriched in Os and depletedin Pd relative to the melt from which it crystallized. Forinstance, several workers obtained Os concentrations of 6 to 51ppb in komatiitic chromites from Gorgona (Walker et al., 1999)

Fig. 3. PGE concentrations in the Kostomuksha basalts (invertedtriangles) and komatiites (diamonds5 spinifex-textured and flowtopbreccias, circles5 cumulate) and the komatiite standard KAL-1 (graysquares) normalized to CI-chondrite abundances of Anders andGrevesse (1989). Primitive mantle (PM) abundances are shown forcomparison.


and Kostomuksha and Vetreny belt (Puchtel et al., 2000a;Puchtel et al., 2000 b), compared to 0.2 to 2 ppb in respectiveprimary magmas. Direct measurements of PGE abundances inchromites from ultramafic residues also indicate that the formerare about two orders of magnitude lower in Pd compared to Osor Ir (Page and Talkington 1984; Walker et al., 1996), thoughit is not yet clear if the IPGEs are incorporated into the spinelcrystal lattice or are carried by the mineral species that areincluded into spinel (e.g., sulfide or Os-Ir alloy). Therefore, itis possible that chromite under some circumstances (e.g., whenthe amount of the fractionating phase exceeds 1 to 2%) maycause IPGE depletions and fractionation of Os from Pd duringdifferentiation of basaltic magmas.

In the Kostomuksha basalts, there is a strong positive corre-lation between MgO, Ni, Cr, and PGE, which attests to the

compatible behavior of these components in the whole range ofbasaltic compositions (Fig. 4). However, due to the advanceddifferentiation that these basaltic melts have undergone underconditions of sulfide saturation, it is difficult to draw anydefinite conclusions as to the ultimate cause of the PGE frac-tionation. A similar problem has been recently addressed byRehkamper et al. (1999b), who concluded that “the PGE are notwell suited as geochemical tracers of basalt source composi-tions, because the PGE systematics of melts are so readilymodified during petrogenesis.” Though the Kostomuksha ba-salts plot within the upper end of the concentration range forthe basalt samples studied by Rehka¨mper et al. (1999b), i.e.,have PGE abundances similar to those of plume-derived Ice-landic tholeiites, yet these by no means represent primarymantle melts. We will, therefore, concentrate our efforts mostly

Fig. 4. Siderophile element variation diagrams for the Kostomuksha basalts and komatiites.


on the studies of PGE abundances in the komatiites, whichapparently do not show sulfide saturation.

In the Kostomuksha komatiites, the flowtop breccias formedduring magma emplacement and did not experience flow dif-ferentiation. Therefore, the erupted Kostomuksha komatiitemagma PGE composition is constrained by the average com-position of the flowtop breccias collected from three flows(Table 4). In the spinifex-textured and cumulate komatiitesamples, the variations in major and trace element composi-tions during differentiation were controlled by a single majormineral phase, olivine (Puchtel et al., 1998). Indeed, on thevariation diagrams in Figure 4, a positive correlation betweenMgO and Ni and a negative correlation between MgO and Crindicates that the low-pressure fractionating assemblage in-cluded only olivine; no chromite was precipitated over theentire compositional range of the komatiite lavas. From litho-phile trace element variations, Puchtel et al. (1998) estimatedthe MgO content of the liquidus olivine and the erupted ko-matiite liquid to be 51 and 27%, respectively. Furthermore,from the regressions in Figure 4, the Ni and Cr contents of thisolivine were estimated to be 3210 and 1715 ppm, respectively.These estimates are very close to direct microprobe analyses ofolivine compositions from komatiites with a similar MgO range(e.g., Arndt et al., 1977; Arndt, 1986). Thus, the Ni-MgOvariations in Figure 4 can be accounted for solely by olivinefractionation, and no Ni-Cu sulfide liquid fractionation is re-quired.

Figure 4 shows that a reverse correlation exists betweenPGEs and MgO for both cumulate (excluding Os and Ir data for#9490, which on both plots is well above the regression lines)and noncumulate komatiites indicating incompatible behaviorof PGEs during olivine fractionation. From these correlations,apparent bulk partition coefficients between the fractionatingassemblage (Ol6 PGE-phases) and komatiitic melt were cal-culated to be DOs 5 0.70, DIr 5 0.75, DPt 5 0.52 and DPd ;0.Thus, the degree of incompatibility varies substantially from Osto Pd, the latter behaving as a highly incompatible elementduring olivine fractionation. These calculations show that crys-tal fractionation during differentiation of the Kostomukshakomatiite magma tends to concentrate PGEs in the residualliquid (represented by the spinifex-textured lavas), though theeffect is relatively small. From the MgO vs. Pd and Nicorrelations in Figure 4, it is also evident that no Ni-Cusulfide fractionated during differentiation, implying that theinitial komatiite magma was sulfide-unsaturated, and re-mained so during olivine fractionation. Note from Table 4that the average composition of the flowtop breccias and theaverage composition of the spinifex-textured komatiites arepractically identical.

One cumulate sample (#9490) is substantially enriched in Osand Ir compared to the primary komatiite melt, and this enrich-ment is not supported by variations in either Pt-Pd or Ni-Cr-MgO concentrations. We attribute this enrichment to the pres-ence of tiny Os-Ir nuggets as an additional cumulate phase inthis komatiite. However, as this enrichment was only found ina single cumulate sample, we conclude that the effect of Os-Irnugget fractionation in the Kostomuksha komatiites was verysporadic.

5.3. PGE Fractionation During Mantle Melting and PGECompositions of Mantle Sources

To estimate the PGE composition of a mantle source, thefollowing are required: (1) the composition of a primary liquidderived from this source; and (2) the relationships between thecomposition of the primary liquid and the source composition(degree of melting F, partition coefficients D, and meltingmodel). There are no experimentally determined PGE partitioncoefficients for melt extraction from the mantle. The PGE hostphases in the mantle, the melting reactions of which determinethe amounts of PGE entering the melt, are also not well known.Some workers have adopted the assumption that PGEs reside insulfide (Os, Ir, Ru, Pt, Pd) and/or in alloy (Os, Ir, Ru), and thatthe relative fractions of these phases involved in the meltingcan be modeled to describe PGE behavior during melting(Barnes et al., 1985; Barnes and Picard 1993; Rehka¨mper et al.,1999b). We will assume for the sake of simplicity that thecarrier phases of PGEs in the mantle were non-silicates, thatthese were the same for all komatiite source regions, and thatthese phases were not affected by phase transformations in thepressure range of interest, i.e., constant D values during partialmelting were assumed. The approach adopted here involved thederivation of mantle-melt apparent partition coefficients for areference komatiite primary liquid, followed by the applicationof these apparent partition coefficients to other komatiite suites.The composition of the Alexo komatiite was chosen as areference since arguably both its primary liquid compositionand its source composition can be constrained.

The primary melt PGE composition of the Alexo komatiitewas estimated from the mean of four measurements of KAL-1(Table 2), an olivine spinifex-textured komatiite from a singleAlexo lava flow, the major element composition of whichclosely approximates the Alexo liquid composition estimatedby Arndt (1986). The Munro Township komatiites have achondritic initial 187Os/188Os implying derivation from asource with a long-term chondritic Re/Os ratio (Walker et al.,1988; Shirey, 1997). We assume that this is also true for theAlexo komatiite (for which Os-isotope data are presently notavailable), since both come from spatially closely associatedparts of the Abitibi greenstone belt. As Re is one of the mostincompatible highly siderophile elements, and Os is one of themost compatible ones, the chondritic Re/Os ratio implies thatthe Munro-Alexo source was also characterized by chondriticrelative abundances of other PGEs. For the absolute PGEabundances in this source, we then assume a primitive mantle(PM) composition (Table 5), and will examine the ramificationsof that assumption later.

The relationship between source composition and melt com-position is given by the batch melting equation which requiresknowledge of a set of partition coefficients for the PGEs and anestimate of the degree of melting. From the relative abundancesof moderately incompatible lithophile trace elements Ti, Zr,Gd, Y, and Yb compiled from Arndt (1986), the degree ofpartial melting for the Alexo komatiite was estimated to be;45%, which is only slightly higher than that calculated in thesame manner for the Kostomuksha komatiite (;40%) from thedata of Puchtel et al. (1998). The apparent partition coefficientsfor the PGEs applicable to extraction of the Alexo komatiitefrom a PM source were obtained using a batch partial melting


model. The observed Pd enrichment in the Alexo komatiite istoo large for a single-stage batch melting model, i.e., the Pdabundance implies a much lower degree of melting than othertrace elements. This discrepancy, typical of highly incompati-ble elements, was resolved here by applying a multistage batchmelting model, in which the melt fraction,F(zi), is specified asan exponential function of depth:

F~ zi! 5 F~0!e2kzi. . . . (1)

where 0, zi , 100 km was used for Alexo,F(0) is the meltfraction near the surface, andk is the reciprocal e-foldingdistance. The batch partial melting equation is solved for asingle set of partition coefficients (Table 5) for each depth, in1 km increments,

Ci 5C0

@F~ zi! 1 D~1 2 F~ zi!!#(1a)

whereCi is the concentration of the element in the melt at depthzi and C0 is the concentration of the element in the source(assumed to be uniform before melting). The melts are aver-aged according to Eqn. 2:

CL 5(@F~ zi!Ci#

(F~ zi!(2)

whereCL is the concentration of the element in the aggregatedmelt. The functional dependence of melt fraction on depth wastaken as a tunable parameter and the net melt compositionderived from this procedure was matched with the average meltcomposition of the Alexo liquid. The calculated melting curvefor the Alexo komatiite is shown in Figure 5.

The resulting partition coefficients for Os, Ir and Ru arerelatively insensitive to the assumptions of the model, with theexception that DRu ; 1 because of the choice of PM as thesource composition. The partition coefficient for Pd is highlysensitive to the multistage-stage melting model, but the behav-ior of Pd as a highly incompatible element with DPd, 0.1appears to be certain. This model was then applied as a peda-gogical tool to examine which features of the Kostomukshamelt composition are related to source differences and whichaspects can be modeled as variations in the degree of meltingbetween Alexo and Kostomuksha. Relative differences insource composition were resolvable, while absolute abun-

Table 5. Parameters Used in Modeling, and Model Results for Hybrid Sources (ppb).

Parameter Os Ir Ru Pt Pd

Primitive mantle (PM)1 3.33 3.30 4.88 6.79 3.84CI-chondrite2 486 481 712 990 560D (solid metal-liquid metal)3 12.0 7.50 2.10 2.24 0.68D (solid metal-liquid metal)4 19.0 14.0 4.20 3.80 0.49Bulk core5 2280 2260 3340 4640 2630Outer core-I (OC-I)6 2280 2260 3340 4640 2630Outer core-II (OC-II)7 1490 1560 3140 4330 2670Outer core-III (OC-III)8 823 1080 2790 3960 2700Hybrid source-I (PM11%OC-I)9 26.1 25.8 38.2 53.1 30.1Hybrid source-II (PM11%OC-II)10 18.1 18.8 36.1 49.8 30.4Hybrid source-III (PM11%OC-III)11 11.4 14.0 32.5 46.1 30.7D (silicate liquid-mantle residue)12 2.12 2.25 1.09 0.41 0.04Komatiite melt13 from hybrid source-I 14.2 13.4 35.8 97.3 138Komatiite melt14 from hybrid source-II 9.86 9.73 33.8 91.2 140Komattie melt15 from hybrid source-III 6.23 7.24 30.5 84.5 141Alexo primary melt16 1.82 1.71 4.59 12.5 17.8Calculated Alexo batch melt 1.82 1.71 4.57 12.4 17.7Kostomuksha primary melt17 1.45 1.38 3.86 8.22 10.6Calculated Kostomuksha batch melt 1.41 1.33 3.45 8.81 10.5Archean oceanic crust (I)18: 0.197 0.204 0.627 3.06 4.24

1 Calculated assuming an Ir content of 3.3 ppb (Morgan, 1986) and CI-chondrite relative abundances of Anders and Grevesse (1989).2 CI-chondrite data from Anders and Grevesse (1989).3 Partition coefficients from Campbell and Humayun (2000).4 Partition coefficients from Morganet al. (1995), Smoliaret al. (1996), Walkeret al. (1997a).5 Calculated assuming complete extraction of PGEs into liquid iron from a CI-chondritic bulk Earth composition.6 Calculated assuming no differentiation has occurred during the core crystallization.7 Calculated using the partition coefficients (3) and assuming 5.5% crystallization of the bulk core (5).8 Calculated using the partition coefficients (4) and assuming 5.5% crystallization of the bulk core (5).9 Calculated as a physical admixture of 1% of the outer core-I to the primitive mantle source (PM).10 Calculated as a physical admixture of 1% of the outer core-II to the primitive mantle source (PM).11 Calculated as a physical admixture of 1% of the outer core-III to the primitive mantle source (PM).12 Partition coefficients calculated from the multistage batch melting model for the Alexo komatiite.13 Calculated as a multistage batch melt from the hybrid source-I using the partition coefficients (12).14 Calculated as a batch melt from the hybrid source-II using the partition coefficients (12).15 Calculated as a batch melt from the hybrid source-III using the partition coefficients (12).16 Average of four KAL-1 analyses.17 Average of three flowtop breccia samples representing the best estimate of the Kostomuksha komatiite liquid composition.18 Calculated as 10% Kostomuksha komatiite primary melt and 90% average basalt (from Table 3).


dances were dependent on the validity of the choice of PM asthe Alexo source.

Compared to the Alexo komatiite melt, the Kostomukshaprimary komatiite liquid was distinct in having lower abun-dances of all PGE, lower Pd/Ir ratio (Fig. 3), and in havingformed at a slightly lower degree of partial melting. Further-more, the petrological data indicate that the Kostomuksha ko-matiite primary melt is distinct from its Alexo counterpart inthat it was separated from the melting source region at fargreater depths in the mantle, while still in the garnet peridotitestability field, thus acquiring the HREE- and Al-depleted sig-nature. Using a constant set of partition coefficients, an F(0)540% (based on lithophile trace elements), and allowingF(zi) tobe a tunable parameter, we applied the melting model to cal-culate the source of the Kostomuksha primary komatiite liquid.Using PM as the source, the model produces solutions thatyield lower Os and Ir only at lower degrees of melting, givingmuch higher Pd/Ir ratios than observed. To produce the ob-served Os, Ir and Ru abundances, the use of a source withabsolute abundances of 0.753 PM (or 25% lower than theAlexo source, regardless of its assumed composition), is re-quired. The Kostomuksha PGE pattern could then be fitted withthe F(z) curve shown in Figure 5. Two specific features of thiscurve require further comment. Firstly, the presence of a regionof melting where F(z)5 F(0) is required by the lower Pd/Ir andPt/Ir of the Kostomuksha komatiite relative to the Alexo ko-matiite. Such a feature could also be produced by depleting the

Kostomuksha source in Pt and Pd before the melt extraction.There is presently no way of distinguishing between thesealternatives. Secondly, the Al-depleted nature of the Kosto-muksha komatiite was interpreted to require melt removal atdepths corresponding to the majorite stability field (Puchtel etal., 1998). Thus, the curve was modified by starting the meltingat a depth of 500 km (Fig. 5) and proceeding all the way up tothe;260 km level until the degree of melting attains 40%, afterwhich it remains constant till the melt separates from the sourceas the plume reaches the 180 km level. This correspondsapproximately to the depth where garnet becomes unstable andAl-depleted komatiitic melts are replaced by Al-undepletedliquids (e.g., Herzberg, 1995).

Because of the strong fractionation of Pt and Pd from Ir bymelting, variations in the Pt/Ir and Pd/Ir ratios of partial meltscan always be accounted for by manipulation of the parametersof the melting model. Factors that determine the composition ofthe source, such as the nature of accreted chondritic material oraddition of outer core, produce source compositions that havePd/Ir ratios varying within only 10 to 20% of PM (althoughdifferent absolute abundances of Pd and Ir). Thus, the large andcorrelated variations in Pt/Ir and Pd/Ir seen both in our data andin those of others (Fig. 6) can be attributed to the effects ofpartial melting. If the partition coefficients derived here areapproximately correct, the PGE composition of residual mantlecan also be determined. In Figure 6, the calculated residuecompositions are compared with the data on mantle peridotitesand xenoliths. Most of the published data clusters near thechondritic value, partly because of a concerted effort by mostworkers to analyze “fertile” samples, and partly because ofmetasomatic processes in the upper portion of any meltingcolumn which act to enrich residual peridotites in Pd and Pt, aneffect seen most strikingly in abyssal peridotites (Rehka¨mper etal., 1999a). A group of very depleted samples from the studiesof both Lorand et al. (1999) and Handler and Bennett (1999)fall along the calculated melt depletion trend. Samples from theOlmani harzburgite and two E. African xenoliths plot along adepletion trend featuring constant Pt/Pd ratios (Rehka¨mper etal., 1997). The complementary melts of such residues have notbeen observed.

5.4. Subducted Archean Oceanic Crust as the Source ofCoupled Radiogenic186,187Os in the Mantle

Lithophile trace element and isotope studies (e.g., Hofmannand White, 1982; Galer and Goldstein, 1991; Hofmann, 1997)have established recycled oceanic crust to be an importantcomponent of certain plume sources, and this conclusion hasbeen extended to explain the systematically radiogenic187Os/188Os isotopic compositions observed in many plumes (e.g.,Walker et al., 1991; Hauri and Hart, 1993; Roy-Barman andAllegre, 1995; Lassiter and Hauri, 1998; Widom et al., 1999).

Some plumes, in addition to high187Os/188Os ratios, are alsocharacterized by elevated186Os/188Os ratios (Walker et al.,1997a; Brandon et al., 1998). Brandon et al. (1999) found apositive correlation (r2 5 0.94) between187Os/188Os and186Os/188Os isotopic ratios in Hawaiian picrites and interpretedthis to be the result of incorporation of outer core material intothese plume sources. To account for the observed correlation,Brandon et al. (1999) showed that the recycled oceanic crust

Fig. 5. The F(z) vs. depth used in the melting models. Note thedifference in melting initiation depths and distinct melting curves forthe two plumes.


must have had Pt/Re;80 to 100, while most modern crustalmaterials are characterized by a Pt/Re of,30, generally;7.However, crustal materials recycled into the sources of modernplumes, have been derived from the mantle billions of yearsago and might have included more magnesian lavas (e.g.,komatiites). Therefore, the evolution of the Pt-Re-Os system insuch ancient crust could also be distinct from that of its moderncounterpart. Here, the Pt-Re-Os evolution of a model Archeanoceanic crust is evaluated using the PGE abundances deter-mined for basalts and komatiites from Kostomuksha (Table 5);Re concentrations were from Puchtel et al. (2000a).

Such oceanic crust consisting of komatiite and basalt in theproportion 1:10 (the proportion typical of the majority of lateArchean greenstone belts, DeWit and Ashwal, 1997, includingGimola-Kostomuksha), had the following parameters: Re51.0 ppb, Pt5 3.1 ppb, Os5 0.20 ppb,187Re/188Os 5 23,190Pt/188Os 5 0.014, and Pt/Re5 3. The results of the modelcalculations, illustrated in Figure 7, show that 80 to 100% ofsuch crust with a residence time of 2.0 Ga are needed toaccount for the«186Os(T) values of11.0 to11.3, typical of themajority of picrites from Mauna Loa, Loihi and Hualalai,where,

«186Os~T! 5 10,0003 3S186Os188OsD

i

sample

S 186Os188OsD

i

PM 2 14 (3)

There are major element constraints arguing against such apossibility. A large amount of basaltic crust in the mantle

source of the Hawaiian plume (at that depth converted intoeclogite) is unlikely as its melting products will mostly consistof tonalite-trondhjemite magmas (Drummond et al., 1996;Rapp et al., 1991) and not picrites as observed (Norman andGarcia, 1999). Moreover, such a crust will develop very radio-genic g187Os(T) values of.700, while the most radiogenicHawaiian picrites haveg187Os(T) of;9 (Brandon et al., 1999).

We conclude that although recycledArcheanoceanic crust isa viable source of radiogenic187Os/188Os in the mantle, theorigin of the coupled187Os/188Os and186Os/188Os enrichmentsobserved by Brandon et al. (1999) still requires a non-crustalsource. Such a source should have low lithophile elementconcentrations in order not to destroy the observed trace ele-ment and Pb-Nd-Sr isotope relationships (Hofmann, 1997). Itshould also have suprachondritic Pt/Os ratios, and Os abun-dances substantially exceeding those of the oceanic crust andthe primitive mantle. The only known major terrestrial reser-voir which meets these requirements is the outer core.

5.5. The Effect of Core Metal Addition on the PGEBudget of Plume Magmas and the Mechanisms ofCore-Mantle Interaction

The introduction of radiogenic Os derived from the differ-entiated outer core into silicate mantle at the core-mantleboundary (CMB) was first proposed by Walker et al. (1995).These authors suggested that the crystallization of the outercore to produce the solid inner core, which accounts for 5.5%of the total mass of the bulk core, will fractionate both Pt andRe from Os assuming that apparent Dsm-lm derived from the

Fig. 6. Pt/Ir vs. Pd/Ir diagram for the Kostomuksha lavas, calculated melt residues (dotted curve), and literature data:upper mantle orogenic lherzolites (Lorand et al., 1999), mantle xenoliths (Rehka¨mper et al., 1997; Handler and Bennett,1999), komatiites and basalts (Rehka¨mper et al., 1999b).


magmatic Group IIAB iron meteorites (Morgan et al., 1995;Smoliar et al., 1996) are applicable to the Earth’s core. Thepartitioning behavior during iron solidification isDOs. DRe.. DPt, with Os preferentially entering the solidmetal of the inner core, and Pt/Os, Re/Os and Pt/Re thusincreasing to suprachondritic values during differentiation ofthe liquid outer core. As a result, with time the latter developsradiogenic187Os/188Os and186Os/188Os ratios. Walker et al.(1995) also argued that, because Os abundances are more thantwo orders of magnitude higher in the core compared to theprimitive mantle, physical addition of even small amounts (1 to2%) of outer core material to the plume source will dominate itsOs isotopic composition, as is evident from Figure 7. It shouldbe emphasized that the magnitude of186,187Os-enrichment inthe resulting outer core strongly depends on the choice ofDsm-lm for Re, Os and Pt, and the timing of differentiation.Using partition coefficients from Morgan et al. (1995), Smoliaret al. (1996) and Walker et al. (1997a), and assuming (as didWalker et al., 1997a) for the convenience of this calculationthat core differentiation was completed within the first 100 Maof Earth’s history, the present-day«186Os of the outer core canbe shown to attain the value of11.7 (Fig. 7). We have recentlycarried out siderophile element analyses of the Group IIABirons and have derived a refined set of internally consistentapparent partition coefficients (Campbell and Humayun, 2000),which are smaller than those of previous workers. Calculationsusing the Dsm-lm values of Campbell and Humayun (2000)produce a present-day«186Os in the outer core of10.6. MostHawaiian picrites with radiogenic186Os/188Os plot betweenthese two outer core estimates (Fig. 7). It appears that GroupIIAB irons are at best imperfect analogues for the Earth’s outer

core and that appropriate DOs/DRe and DOs/DPt at the innercore-outer core boundary must be larger than values derivedfrom magmatic iron meteorites if this hypothesis is to be viable.

The mechanism proposed by Walker et al. (1995) for core-mantle interaction is that of physical admixture of outer corematerial into primitive mantle to form hybrid mantle sources,with the metal getting entrained into the mantle as the plumestarts its ascent from the CMB. This process must have aprofound effect on the PGE abundances in the plume source,raising these by about an order of magnitude relative to mantlesource regions not affected by core addition. Below, we quan-titatively model the potential effect of this physical admixtureon the PGE abundances in hybrid sources and, therefore, on theplume magmas derived from such sources.

Three outer core model compositions were used here: un-fractionated [Outer Core-I] and after 5.5% crystallization withDsm-lm from Campbell and Humayun (2000) [Outer Core-II]and Smoliar et al. (1996) [Outer Core-III]. The effects ofphysical admixture of 1% of the outer core material into amantle plume source, here taken as the primitive mantle (Table5), are shown in Figure 8. The multistage batch melting model,described above, was used to calculate the composition ofhypothetical komatiite melts derived from hybrid mantlesources, which were then compared with the Kostomuksha andAlexo komatiite liquid compositions (Fig. 8).

The first conclusion is that independent of which set ofDsm-lm is used to estimate the outer core composition, theresulting komatiite melts derived from such hybrid sourceswould have had PGE abundances substantially higher thankomatiites derived from mantle sources not affected by coreaddition (e.g., Pd5 140 ppb vs. 10 to 20 ppb). As can be seen

Fig. 7. Results of mixing calculations between a plume source with a primitive mantle composition and 2.0 Ga subductedKostomuksha-type oceanic crust (oceanic crust-I and II) or outer core II and III (Table 5). Oceanic crust-I consists of 10%komatiite1 90% basalt, and oceanic crust-II is composed of 100% komatiite.«186Os data for Hawaii (Brandon et al., 1999)and Noril’sk (Walker et al., 1997b) are shown for comparison. Note that, to account for the radiogenic«186Os observed inthese plumes, addition of 80 to 100% model Archean crust is required.


in Figure 8, the Kostomuksha komatiite does not show suchelevated PGE concentrations. The second conclusion concernsthe fractionated Os/Ir ratio in the model komatiite magmasderived from hybrid sources, most obvious for those mixtureswith outer core-III composition (Fig. 8). This, too, is notobserved in the Kostomuksha komatiites, which have a chon-dritic Os/Ir ratio. A chondritic Os/Ir ratio is typical of mantleperidotites (Humayun et al., 1997), komatiites (e.g., this study;Brugmann et al., 1987; Humayun et al., 1997), and somebasalts (Ravizza and Pyle, 1997), indicative of the similarity ofOs and Ir partitioning into mantle melts.

From the discussion above, our data do not support physicaladmixture of any substantial amount (.0.01%) of outer corematerial into the Kostomuksha source, as this would increasethe PGE abundances in the resulting partial melts to levels notseen in this komatiite. We conclude that if the enriched Os-isotope signature in the Kostomuksha komatiite was indeedderived from the outer core, then core-mantle interaction can-not involvesimple physical admixtureof the outer core mate-rial into the plume source. A reconciliation of these facts wouldbe that the enriched187Os isotope signature in the Kostomuk-sha komatiite was acquired from the base of the lower mantle,which had experienced isotopic equilibration with the liquidouter core at the CMB. Small, but distinct, changes in the PGErelative abundances are expected to accompany the Os-isotopicexchange, which would include the development of suprachon-dritic Re/Os and Pt/Os ratios, and slightly subchondritic Os/Irratios and fractionated PGE patterns (Pd.Pt 5 Ru.Ir.Os).These patterns resemble those resulting from the effects ofpartial melting (Pd.Pt.Ru.Ir 5 Os), but the differencesshould be resolvable by precise analyses of PGEs in high-degree partial melts from mantle plumes.

6. CONCLUDING REMARKS

The following conclusions were drawn from this study:

1. Platinum group elements (Os, Ir, Ru, Pt and Pd) were foundto be immobile during seafloor alteration and upper green-schist facies metamorphism of the Kostomuksha komatiitesand basalts.

2. During high-degree partial melting, Os and Ir are compati-ble and mostly retained in the mantle residue. Platinum andPd concentrate in the melt, Pt being moderately and Pdhighly incompatible with the mantle residue. During komat-iite low-pressure differentiation, the PGEs are incompatiblewith olivine. The calculated bulk partition coefficients rangefrom ;0.70 for Os and Ir to 0.50 for Pt and nearly zero forPd.

3. Unlike most basalts, komatiites are primary mantle meltsthat faithfully retain their original PGE compositions. Thesecompositions were used to determine relative PGE abun-dances between the Alexo and Kostomuksha sources, whichwere shown to differ by 25%.

4. Substantial Pt/Os enrichments combined with the relativelyhigh Os abundances estimated for Archean komatiite-basaltcrust, which could potentially become subducted and incor-porated into the source of mantle plumes, are insufficient toaccount for the coupled radiogenic186,187Os signature ob-served in some modern plume magmas. The origin of thissignature requires a non-crustal source.

5. If the radiogenic initial187Os/188Os in the Kostomukshakomatiite is an outer core signature, then core-mantle inter-action cannot involvephysical admixtureof the outer corematerial into mantle plume sources. Instead, this type ofinteraction must occur in the form of Os-isotope exchange atthe core-mantle boundary.

Fig. 8. Diagram illustrating the effect of physical addition of core material on the PGE abundances in the hybrid sourcesand calculated komatiite melts, from Table 5. Note the high abundances of PGEs in model komatiitic melts from sourceswith 1% core addition, contrasted with the distribution of PGEs in the Alexo (chondritic initial187Os/188Os) andKostomuksha (radiogenic initial187Os/188Os) primary komatiite melts. The PGE abundances of the Kostomuksha komatiitedo not supportphysical additionof outer core (regardless of its state of differentiation) to the mantle source region.


This study demonstrates that komatiites are important probesof possible chemical exchange processes at the core-mantleboundary and that PGE abundances in high-degree, sulfide-unsaturated mantle melts provide information on mantle sourcePGE compositions, that is complementary to that derived from186,187Os isotopic studies.

Acknowledgments—Komatiite standards, KAL-1 and WITS-1, weregenerously provided by Steve Shirey. Comments from Alan Brandon,Kevin Righter, Andy Davis, and Larry Grossman were appreciated.Very constructive comments from an anonymous but identified re-viewer helped to substantially improve the original draft of the ms. Wethank Horton Newsom for his efforts as Associate Editor, and discus-sions on core-mantle interaction. We thank Andy Campbell for assis-tance with the ICP-MS measurements and many discussions on PGEpartitioning in metallic systems. The acquisition of the Elementoˆ™ wassupported by NSF EAR 9601478 to MH. This study was also madepossible by funding from the Block Fund at The University of Chicago.

Associate editor:H. E. Newsom

REFERENCES

Anders E. and Grevesse N. (1989) Abundances of the elements: Me-teoritic and solar.Geochim. Cosmochim. Acta53, 197–214.

Arndt N. T. (1976) Melting relations of ultramafic lavas (komatiites) atone atmosphere and high pressure.Yb. Carnegie Inst. Wash.75,555–562.

Arndt N. T. (1986) Differentiation of komatiite flows.J. Petrol. 27,279–301.

Arndt N. T., Naldrett A. J., and Pyke D. R. (1977) Komatiitic andiron-rich tholeiitic lavas of Munro Township, northeast Ontario.J.Petrol. 18, 319–369.

Barnes S.-J., Naldrett A. J., and Gorton M. P. (1985) The origin of thefractionation of platinum-group elements in terrestrial magmas.Chem. Geol.53, 303–323.

Barnes S.-J., and Picard C. P. (1993) The behaviour of platinum-groupelements during partial melting, crystal fractionation, and sulphidesegregation: An example from the Cape Smith Fold Belt, northernQuebec.Geochim. Cosmochim. Acta57, 79–87.

Barnes S. J., Hill R. E. T., and Gole M. J. (1988) The Perseveranceultramafic complex, Western Australia: the product of a komatiitelava river.J. Petrol.29, 305–331.

Brugmann G. E., Arndt N. T., Hofmann A. W., and Tobschall H. J.(1987) Noble metal abundances in komatiite suites from Alexo,Ontario, and Gorgona Island, Colombia.Geochim. Cosmochim. Acta51, 2159–2169.

Brandon, A. D., Norman M. D., Walker R. J., and Morgan J. W. (1999)186Os-187Os systematics of Hawaiian picrites.Earth Planet. Sci. Lett.174,25–42.

Brandon A.D., Walker R. J., Morgan J. W., Norman M. D., andPrichard H. M. (1998) Coupled186Os and187Os evidence for core-mantle interaction.Science280,1570–1573.

Burton K. W., Schiano P., Birck J.-L., and Alle`gre C. J. (1999)Osmium isotope disequilibrium between mantle minerals in a spinel-lherzolite.Earth Planet. Sci. Lett.172,311–322.

Campbell A., and Humayun M. (2000). Platinum group element mi-croanalysis of iron meteorites.Geochim. Cosmochim. Acta(submit-ted).

Campbell I. H., and Griffiths R. W. (1990) Implications of mantleplume structure for the evolution of flood basalts.Earth Planet. Sci.Lett. 99, 79–93.

Crocket J. H., and MacRae W. E. (1986) Platinum-group elementdistribution in komatiitic and tholeiitic volcanic rocks from MunroTownship, Ontario.Econ. Geol.81, 1242–1251.

Desrochers J.-P., Hubert C., Ludden J. N., and Pilote P. (1993) Accre-tion of Archaean oceanic plateau fragments in the Abitibi greenstonebelt, Canada.Geology21, 451–454.

DeWit M. J. and Ashwal L. D. (1997) Greenstone Belts, Oxford,Clarendon Press.

Drummond M. S., Defant M. J., and Kepezhinskas P. K. (1996)

Petrogenesis of slab-derived trondhjemite-tonalite-dacite/adakitemagmas.Trans. R. Soc. Edinburgh.87, 205–215.

Fleet M. E., Crocket J. H., and Stone W. E. (1996) Partitioning ofplatinum-group elements (Os, Ir, Ru, Pt, Pd) and gold betweensulfide liquid and basalt melt.Geochim. Cosmochim. Acta60,2397–2412.

Galer S. J. G. and Goldstein S. L. (1991) Early mantle differentiationand its thermal consequences.Geochim. Cosmochim. Acta55, 227–239.

Handler M. R. and Bennett V. C. (1999) Behaviour of platinum-groupelements in the subcontinental mantle of eastern Australia duringvariable metasomatism and melt depletion.Geochim. Cosmochim.Acta 63, 3597–3618.

Hart S. R. and Ravizza G. E. (1996) Os partitioning between phases inlherzolite and basalt. InEarth Processes: Reading the Isotopic Code.(ed. A. Basu and S. R. Hart). Washington, D.C., Am. Geophys.Union, pp. 123–134.

Hauri E. and Hart S. R. (1993) Re-Os isotope systematics of HIMU andEMII oceanic island basalts from the south Pacific Ocean.EarthPlanet. Sci. Lett.114,353–371.

Hauri E. H. and Hart S. R. (1997) Rhenium abundances and system-atics in oceanic basalts.Chem. Geol.139,185–205.

Hertogen J., Janssens M.-J., and Palme H. (1980) Trace elements inoceanic ridge basalts: Implications for fractionations during mantleevolution and petrogenesis.Geochim. Cosmochim. Acta.44, 2125–2143.

Herzberg C. (1995) Generation of plume magmas through time: anexperimental perspective.Chem. Geol.126,1–16.

Hoffman E. L., Naldrett A. J., and Van Loon J. C. (1978) The deter-mination of all the platinum group elements and gold in rocks andore by neutron activation analysis after preconcentration by a nickelsulfide fire-assay technique on large samples.Anal. Chim. Acta102,157–166.

Hofmann A. W. (1997) Mantle geochemistry: the message from oce-anic volcanism.Nature385,219–229.

Hofmann A. W. and White W. M. (1982) Mantle plumes from ancientoceanic crust.Earth Planet. Sci. Lett.57, 421–436.

Humayun M. (1998) Platinum group element microanalysis of singlesulfides by ICPMS:Am. Geophys. Union, Fall ’98 Meeting.

Humayun M., Brandon A. D., Dick H. J. B., and Shirey S. B. (1997)Ir/Os constraints on terrestrial accretion and core formation:LPSXXVIII, 613–614.

Jarvis I., Totland M. M., and Jarvis K. E. (1997) Determination of theplatinum-group elements in geological materials by ICP-MS usingmicrowave digestion, alkali fusion and cation-exchange chromatog-raphy.Chem. Geol.143,27–42.

Lassiter J. C. and Hauri E. H. (1998) Osmium-isotope variations inHawaiian lavas: Evidence for recycled oceanic lithosphere in theHawaiian plume.Earth Planet. Sci. Lett.164,483–496.

Lorand J.-P., Pattou L., and Gros M. (1999) Fractionation of platinum-group elements and gold in the upper mantle: A detailed study inPyrenean Orogenic lherzolites.J. Petrol.40, 957–981.

Marcantonio F., Zindler A., Elliott T., and Staudigel H. (1995) Osisotope systematics of La Palma, Canary Islands: Evidence forrecycled crust in the mantle source of HIMU ocean islands.EarthPlanet. Sci. Lett.133,397–410.

McGoldrick P. J., Keays R. R., and Scott B. B. (1979) Thallium: Asensitive indicator of rock/seawater interaction and of sulfur satura-tion of silicate melts.Geochim. Cosmochim. Acta43, 1303–1311.

Mitchell R. H. and Keays R. R. (1981) Abundance and distribution ofgold, palladium and iridium in some spinel and garnet lherzolites:Implications for the nature and origin of precious metal-rich inter-granular components in the upper mantle.Geochim. Cosmochim.Acta 45, 2425–2442.

Morgan J. W. (1986) Ultramafic xenoliths: Clues to Earth’s late accre-tionary history.J. Geophys. Res.91, 12375–12387.

Morgan J. W., Horan M. F., Walker R. J., and Grossman J. N. (1995)Rhenium-osmium concentration and isotope systematics in groupIIAB iron meteorites.Geochim. Cosmochim. Acta59, 2331–2344.

Norman M. D. and Garcia M. O. (1999) Primitive magmas and sourcecharacteristics of the Hawaiian plume: petrology and geochemistryof shield picrites.Earth Planet. Sci. Lett.168,27–44.

Page N. J. and Talkington R. W. (1984) Palladium, platinum, rhodium,


ruthenium and iridium in peridotites and chromitites from ophiolitecomplexes in Newfoundland.Can. Mineral.22, 137–149.

Pattou L., Lorand J. P., and Gros M. (1996) Non-chondritic platinum-group element ratios in the Earth’s mantle.Nature379,712–715.

Peach C. L., Mathez E. A., and Keays R. R. (1990) Sulfide melt-silicatemelt distribution coefficients for noble metals and other chalcophileelements as deduced from MORB; implications for partial melting.Geochim. Cosmochim. Acta54, 3379–3389.

Peach C. L., Mathez E. A., Keays R. R., and Reeves S. J. (1994)Experimentally determined sulfide melt-silicate melt partition coef-ficients for iridium and palladium.Chem. Geol.117,361–377.

Pearson D. G. and Woodland S. J. (2000). Solvent extraction/anionexchange separation and determination of PGEs (Os, Ir, Pt, Pd, Ru)and Re-Os isotopes in geological samples by isotope dilution ICP-MS. Chem. Geol.165,87–107.

Pegram W. J. and Alle`gre C.-J. (1992) Osmium isotopic compositionsfrom oceanic basalts.Earth Planet. Sci. Lett.111,59–68.

Puchtel I. S., Bru¨gmann G. E., and Hofmann A. W. (2000a)187Os-enriched domain in an Archaean mantle plume: Evidence from 2.8Ga komatiites of the Kostomuksha greenstone belt, NW BalticShield.Earth Planet. Sci. Lett.(submitted).

Puchtel I. S., Bru¨gmann G. E., Hofmann A. W., Kulikov V. S., andKulikova V. V. (2000b). Os-isotope systematics of remarkably freshkomatiitic basalts from the Vetreny belt, Baltic Shield: Evidence fora chondritic source of a 2.45 Ga plume.Contrib. Mineral. Petrol.(inpress).

Puchtel I. S., Hofmann A. W., Jochum K. P., Mezger K., ShchipanskyA. A., and Samsonov A. V. (1997) The Kostomuksha greenstonebelt, NW Baltic Shield: Remnant of a Late Archean oceanic plateau?Terra Nova9, 87–90.

Puchtel I. S., Hofmann A. W., Mezger K., Jochum K. P., ShchipanskyA. A., and Samsonov A. V. (1998) Oceanic plateau model forcontinental crustal growth in the Archaean: A case study from theKostomuksha greenstone belt, NW Baltic Shield.Earth Planet. Sci.Lett. 155,57–74.

Puchtel I. S. and Humayun M. (1999) Platinum group elements in the2.8 Ga Kostomuksha greenstone belt.9th V.M. Goldschmidt Con-ference, p. 234–235.

Rapp R. P., Watson E. B., and Miller C. F. (1991) Partial melting ofamphibolite/eclogite and the origin of Archean trondhjemites andtonalites.Precambrian Res.51, 1–25.

Ravizza G. and Pyle D. (1997) PGE and Os isotopic analyses of singlesample aliquots with NiS fire assay preconcentration.Chem. Geol.141,251–268.

Rehkamper M. and Halliday A. N. (1997) Development and applicationof new ion-exchange techniques for the separation of the platinum-group and other siderophile elements from geological samples.Ta-lanta. 44, 663–672.

Rehkamper M., Halliday A. N., Alt J., Fitton J. G., Zipfel J., andTakazawa E. (1999a) Non-chondritic platinum-group element ratiosin oceanic mantle lithosphere: Petrogenetic signature of melt perco-lation?Earth Planet. Sci. Lett.172,65–81.

Rehkamper M., Halliday A. N., Barfod D., and Fitton J. G. (1997)Platinum-group element abundance patterns in different mantle en-vironments.Science278,1595–1598.

Rehkamper M., Halliday A. N., Fitton J. G., Lee D.-C., Wieneke M.,and Arndt N. T. (1999b) Ir, Ru, Pt and Pd in basalts and komatiites:New constraints for the geochemical behavior of the platinum groupelements in the mantle.Geochim. Cosmochim. Acta63, 3915–3934.

Robert R. V. D., Van Wyk E., and Palmer R. (1971) Concentration ofthe noble metals by a fire assay technique using nickel sulfide as acollector.Nat. Inst. Metall. Rep.1371,1–14.

Ross J. R. and Keays R. R. (1979) Precious metals in volcanic-type

nickel sulfide deposits in Western Australia I. Relationship with thecomposition of the ores and their host rocks.Can. Mineral. 17,417–435.

Roy-Barman M. and Alle`gre C. J. (1995)187Os/186Os in oceanic islandbasalts: Tracing oceanic crust recycling in the mantle.Earth Planet.Sci. Lett.129,145–161.

Shirey S. B. (1997) Initial Os isotopic composition of Munro Town-ship, Ontario komatiites revisited: Additional evidence for near-chondritic, late-Archaean convecting mantle beneath the SuperiorProvince.7th V.M. Goldschmidt Conference,p. 193.

Shirey S. B. and Walker R. J. (1998) The Re-Os isotope system incosmochemistry and high-temperature geochemistry.Ann. Rev.Earth Planet. Sci.26, 423–500.

Shirey S. B., Wilson A. H., and Carlson R. W. (1998) Re-Os isotopicsystematics of the 3300 Ma Nondweni greenstone belt, South Africa:Implications for komatiite formation at a craton edge.Eos79,S372.

Smoliar M. I., Walker R. J., and Morgan J. W. (1996) Re-Os ages ofGroup IIA, IIIA, IVA, and IVB iron meteorites.Science271,1099–1102.

Sokolov V. A. (1987) Geology of Karelia. Leningrad, Nauka.Storey M., Mahoney J. J., Kroenke L. W., and Saunders A. D. (1991)

Are oceanic plateaus sites of komatiite formation?Geology 19,376–379.

Tredoux M. and Mcdonald I. (1996) Komatiite WITS-1, low concen-tration noble metal standard for the analysis of non-mineralizedsamples.Geostandards Newsletter20, 267–276.

Walker R. J., Echeverria L. M., Shirey S. B., and Horan M. F. (1991)Re-Os isotopic constraints on the origin of volcanic rocks, GorgonaIsland, Colombia: Os-isotopic evidence for ancient heterogeneitiesin the mantle.Contrib. Mineral. Petrol.107,150–162.

Walker R. J., Hanski E., Vuollo J., and Liipo J. (1996) The Os isotopiccomposition of Proterozoic upper mantle: Evidence for chondriticupper mantle from the Outokumpu ophiolite, Finland.Earth Planet.Sci. Lett.141,161–173.

Walker R. J., Morgan J. W., Beary E. S., Smoliar M. I., CzamanskeG. K., and Horan M. F. (1997a) Applications of the190Pt-186Osisotope system to geochemistry and cosmochemistry.Geochim. Cos-mochim. Acta61, 4799–4807.

Walker R. J., Morgan J. W., Hanski E., and Smolkin V. F. (1997b)Re-Os systematics of Early Proterozoic picrites, Pechenga complex,NW Russia: Evidence for ancient187Os-enriched plumes.Geochim.Cosmochim. Acta61, 3145–3160.

Walker R. J., Morgan J. W., and Horan M. F. (1995)187Os enrichmentin some plumes: evidence for core-mantle interaction.Science269,819–822.

Walker R. J., Shirey S. B., and Stecher O. (1988) Comparative Re-Os,Sm-Nd and Rb-Sr isotope and trace element systematics for Archeankomatiite flows from Munro Township, Abitibi belt, Ontario.EarthPlanet. Sci. Lett.87, 1–12.

Walker R. J., Storey M., Kerr A. C., Tarney J., and Arndt N. T. (1999)Implications of 187Os isotopic heterogeneities in a mantle plume:Evidence from Gorgona Island and Curac¸ao.Geochim. Cosmochim.Acta 63, 713–728.

Wallace P. and Carmichael I. S. E. (1992) Sulfur in basaltic magmas.Geochim. Cosmochim. Acta.56, 1863–1874.

Widom E., Hoernle K. A., Shirey S. B., and Schmincke H. U. (1999)Os isotope systematics in the Canary Islands and Madeira: Litho-spheric contamination and mantle plume signatures.J. Petrol. 40,279–296.

Widom E. and Shirey S. B. (1996) Os isotope systematics in theAzores: Implications for mantle plume sources.Earth and Planet.Sci. Lett.142,451–465.


Platinum group elements in Kostomuksha komatiites and ...humayun/7Kosto2000.pdflower absolute PGE...

Documents

Transcript of Platinum group elements in Kostomuksha komatiites and ...humayun/7Kosto2000.pdflower absolute PGE...