PGE enrichment in chromitite layers and the Merensky Reef of the western Bushveld Complex; a Re–Os and Rb–Sr isotope study

Page 1

ELSEVIER Earth and Planetary Science Letters 172 (1999) 49–64www.elsevier.com/locate/epsl

PGE enrichment in chromitite layers and the Merensky Reef of thewestern Bushveld Complex; a Re–Os and Rb–Sr isotope study

Ronny Schoenberg a,Ł, F. Johan Kruger b, Thomas F. Nagler a, Thomas Meisel c,Jan D. Kramers a

a Gruppe Isotopengeologie, Mineralogisch.–Petrographisches Institut, Universitat Bern, Erlachstrasse 9a, CH-3012 Bern, Switzerlandb Hugh Allsopp Laboratory, University of the Witwatersrand, Wits 2050, Johannesburg, Republic of South Africa

c Institut fur Allgemeine und Analytische Chemie, Universitat Leoben, Franz-Josef-Strasse 18, A-8700 Leoben, Austria

Received 5 June 1998; revised version received 7 July 1999; accepted 29 July 1999

Abstract

Four poikilitic pyroxenites of the Bastard Unit, all being relatively low in osmium concentrations, define a Re–Os isochron age of 2043 š 11 Ma (with initial 187Os=188Os D 0.151), which is in accordance with the normallycited crystallization age of the Bushveld Complex. The isochron fit demonstrates an excellent radiogenic Os isotopichomogeneity of this portion of the complex at the time of crystallization and excludes any later disturbance of the Re–Osisotope system. Initial 187Os=188Os ratios of platinum-group element (PGE) enriched sulfides and whole rocks of thedirectly underlying Merensky Reef are highly radiogenic and variable (with values between 0.168 and 0.181). Separatesof chromite grains and interstitial phases of Critical Zone chromitite layers show initial 187Os=188Os signatures rangingfrom the near-chondritic mantle values of lower group chromitites (i.e. ¾0.120) to 0.137 for middle group chromititesand to 0.150 for the upper group 2 chromitite layer. Osmium concentrations of interstitial phases exceed those of theassociated chromite in all layers by up to 30 fold, showing the presence of highly PGE-enriched interstitial sulfides. Equalor very similar initial 187Os=188Os signatures of the interstitial minerals and the cumulus chromite grains together with theircoupled increasing and decreasing trends in this ratio clearly demonstrate that interstitial sulfides and associated cumuluschromite of the different chromitite layers scavenged their osmium from the same or similar sources. Therefore, a PGEenrichment of the interstitial sulfides by a primary magmatic process, such as precipitation of immiscible sulfide dropletsduring magma mixing appears more likely than a scenario involving fluid or melt transport of PGE into these layers. Sucha primary magmatic PGE enrichment process is also preferred for the formation of the Merensky Reef. © 1999 ElsevierScience B.V. All rights reserved.

Keywords: Bushveld Complex; magmas; mixing; mineral deposits, genesis; chromitite; platinum group; Re=Os

1. Introduction

The Critical Zone of the Bushveld Complex ischaracterized by repeated sequences of mafic cumu-

Ł Corresponding author. Tel.: C41-31-631-8533; Fax: C41-31-631-4988; E-mail: [email protected]

late rocks (hereafter referred to as cyclic units, e.g.[1]), ranging from harzburgite to orthopyroxenite inthe lower and from orthopyroxenite over norite up toanorthosites in the upper Critical Zone (Figs. 1 and2). At the bases of these cyclic units massive chromi-tite layers may occur, which, compared to theirsurrounding rocks, are enriched in platinum-group

0012-821X/99/$ – see front matter © 1999 Elsevier Science B.V. All rights reserved.PII: S 0 0 1 2 - 8 2 1 X ( 9 9 ) 0 0 1 9 8 - 3

Page 2

50 R. Schoenberg et al. / Earth and Planetary Science Letters 172 (1999) 49–64

Fig. 1. Simplified map of the western limb of the Bushveld Com-plex. Note the platinum mine sections along the Merensky Reefand UG2 chromitite layer and borehole as well as undergroundsampling localities.

elements (PGE). These PGE are mainly incorporatedin sulfide grains occurring as inclusions in the cumu-lus chromite, in the chromite lattice itself and in theintercumulus sulfides [2,3].

The Merensky and Bastard Units form the tran-sition from the Critical Zone to the overlying MainZone. They are frequently considered to be part ofthe Critical Zone as they contain similar sequences oforthopyroxenite, upwardly grading through melano-and leuconorite into anorthosite [4]. Here they areconsidered separate on geochemical grounds [5].The basal layer of the Merensky Unit is betterknown as the Merensky Reef (Fig. 1) and con-tains a pegmatoidal pyroxenite, in which a smallamount of chromite forms thin stringers at the con-tact of the pegmatoid with the poikilitic pyroxenitecountry rock. Interstitial sulfides in the pegmatoidalpyroxenite are highly enriched in PGE.

Although petrographically the cyclic units seemto follow normal fractional crystallization trends [4],geochemical and isotopic data point to a much more

complicated formation history. In most of the cyclicunits, the lowermost cumulates display lower Mg#(D Mg=(Mg C Fe2C)) than would be expected forearly crystallizing phases of a mantle-derived magma[6,7]. Furthermore, whole rock Mg# as well as Crconcentrations and Ni=Ti ratios in cumulus orthopy-roxenes increase with height in some of the cyclicunits [6–8], trends that are opposite to what would beexpected from fractional crystallization. Highly vari-able initial 87Sr=86Sr ratios [9] accompany these geo-chemical trends and their radiogenic values can beexplained by crustal contamination of the Bushveldmagma(s), as also indicated by other isotope sys-tematics, such as Pb=Pb, Sm–Nd and δ18O [10–13]. Os isotopic heterogeneity has also been demon-strated within the Bushveld Complex [14–16], withmuch higher 187Os=188Os ratios than those observedwithin the contemporaneous chondritic mantle, evenin plumes of very deep-seated origin [17,18] and,therefore, in accord with the crustal contaminationscenario: in mantle melting Os is strongly compati-ble, whereas Re is slightly incompatible, leading tovery high Re=Os ratios in the melt phase. Hence,crustal rocks have high Re=Os ratios and rapidlydevelop very radiogenic Os signatures with time,compared to those of the chondritic mantle model[19,20].

The accumulation of prominent chromitite lay-ers at the bases of the cyclic units is now widelyaccepted to result from mixing of newly injected,primitive melt with residual, evolved magma in thechamber, forcing the composition of the resultinghybrid magma into the stability field of chromite[21]. However, the repeated injection and mixingprocess cannot fully account for all geochemicaland isotopic variations observed in Bushveld rocks.For example, in a drill core from the Atok section(eastern Bushveld lobe), Mathez et al. [22] found el-evated rare earth concentrations with light rare earth-enriched patterns in pyroxenites of the Merenskyand Bastard Units, but not in the directly enclosinganorthosites. From these and textural observations,they suggest that compaction of lower cumulates (i.e.norites of the underlying Critical Zone) and result-ing upward percolation of the collected interstitialmelt through the porous overlying cumulates led tohydration partial melting and recrystallization.

Within the framework of the proposed models in-

Page 3

R. Schoenberg et al. / Earth and Planetary Science Letters 172 (1999) 49–64 51

Fig. 2. Initial Os isotopic signatures of Critical Zone chromitite layers (chromite separates and interstitial gangue phases), Merensky ReefPGE-alloys [14] as well as pyroxenites from the Merensky and Bastard Units (see also Table 1). Uncertainties of initial 187Os=188Osratios are reported at the 2¦ confidence level.

Page 4

52 R. Schoenberg et al. / Earth and Planetary Science Letters 172 (1999) 49–64

volving magma mixing, crustal contamination, crys-tal sedimentation and postcumulus processes, thereare a number of possible mechanisms for the in-troduction and enrichment of PGE in the CriticalZone chromitite layers and the Merensky Reef, basedon the two principles of orthomagmatic enrichment(i.e. scavenging by and accumulation of sulfide meltdroplets [23]) and enrichment by late magmatic flu-ids or postcumulus melt percolation (e.g. [22,24–26]).

Combined Os isotope work on both chromite andinterstitial material (D gangue, containing sulfide)has the potential to constrain these models further, ascrustal contamination should lead to variations alongthe magmatic stratigraphy of a layered complex. Theinitial isotope ratio of Os within cumulus chromiteis expected to reflect that of the magma from whichit crystallized, as chromite is likely to remain closedwith respect to Re–Os systematics during exposureto a chemically different, evolved interstitial melt[27,28], unless the melt contains a significant portionof vapor [22,29], lowering the liquidus of chromite.In contrast, the signature of interstitial silicate andsulfide material would be affected by postcumuluspercolation of melts or fluids. We have, therefore,measured Re=Os and Os isotope ratios on separatesof cumulus chromites and the associated interstitialmaterial from nine Critical Zone chromitite layers aswell as five whole rock pyroxenites from the Meren-sky and Bastard Units. The results are discussedtogether with published single-grain Os signatures ofPGE alloys from the Merensky Reef pegmatoid itself[14].

2. Sample description

All chromitite samples from the lower group (LG)1 up to the middle group (MG) 4 layers were ob-tained from three boreholes (NG1, 2 and 3) col-lared by the Geological Survey of South Africa onthe farms Nooitgedacht (406 KQ; NG1 and 2) andZwartklip (405 KQ; NG3). Simplified core logs,modal mineralogy and geochemical trends are givenin [6]. Detailed descriptions of petrology, geochem-istry and platinum-group element mineralization ofchromitite layers from these cores were publishedby Scoon and Teigler [30]. The upper group (UG) 1

and 2 chromitite layers were sampled undergroundat the Union section of Rustenberg Platinum MinesLimited. The chromitite layers used in this studyare characterized by chromite as the only cumulusphase, making up more than 50% of the total rock’svolume. The only exception is the LG4 chromititelayer, which also contains minor amounts of cumulusolivine. Postcumulus phases are orthopyroxene andplagioclase and minor amounts of sulfides and bi-otite. The interstitial phase assemblage changes fromorthopyroxene in the lowest (LG) chromitites to or-thopyroxene and plagioclase in the MG chromititesand to dominantly plagioclase with minor orthopy-roxene in the uppermost (UG) chromitite layers.

Samples SKN-19=15 and SKN-19=25 are froma borehole located in the Amandelbult section onthe farm Schildpadnest (385 KQ). The DK1 sam-ples were obtained from a borehole in the Britsarea, collared by the platinum company. BS-1 wassampled underground at the RPM Rustenberg sec-tion in the Brakspruit shaft. All pyroxenites an-alyzed belong to the Bastard cyclic unit, exceptsample SKN-19=26, which belongs to the Meren-sky cyclic unit (uppermost poikilitic pyroxenite ofthe Merensky Reef). The dominant cumulus phasein DK1=52.4, DK1=49.2, SKN-19=26, SKN-19=15and BS-1 is fine- to medium-grained orthopyroxene(usually bronzite), which normally makes up morethan 90% of the rock, whereas chromite is a sub-ordinate cumulus mineral. Plagioclase is the domi-nant and clinopyroxene a subordinate postcumulusphase, both poikilitically enclosing orthopyroxeneand chromite. Clinopyroxene also poikilitically en-closes plagioclase, pointing to a protracted (re-) crys-tallization. Biotite, quartz and calcite occur in minoramounts as intercumulus phases and appear to belate crystallization products of a fluid-rich melt (seealso [24–26]). The rounded appearance of orthopy-roxene grains, enclosed by clinopyroxene oikocrystsas well as the occurrence of (partly kinked) clinopy-roxene exsolution lamellae in some orthopyroxenecrystals point to postcumulus mineral reaction anddeformation. The obliteration of plagioclase zon-ing and polygonization of large plagioclase crystalsinto smaller individuals during recrystallization arealso strain effects due to postcumulus deformation.In the literature, these rocks are generally referredto as poikilitic pyroxenites and, where postcumu-

Page 5

R. Schoenberg et al. / Earth and Planetary Science Letters 172 (1999) 49–64 53

lus plagioclase makes up 10% or more of the rock,as feldspathic poikilitic pyroxenite or as norite, al-though, using IUGS nomenclature [31], they shouldbe named (orthopyroxene-) gabbros. Regarding theoverall ultramafic to mafic nature of the BushveldComplex, we also prefer the normally cited rockname, which is (feldspathic) poikilitic pyroxenite.All samples used in this study are fresh with verylittle if any alteration evident.

3. Results

Re–Os and Rb–Sr data are summarized in Table 1and analytical techniques for isotope determinationsare described in Appendix A.

3.1. The Merensky and Bastard Units

The four poikilitic pyroxenites (DK1=52.4,DK1=49.2, SKN-19=15 and BS-1) belonging to theBastard Unit (Fig. 2), gave a Re–Os isochron (Fig. 3)yielding an age of 2043 š 11 Ma (MSWD D 0.707)and an initial 187Os=188Os ratio of 0.1506. This is anew, independent age determination for the BushveldComplex, which reproduces within uncertainty theRb–Sr age of 2061 š 27 Ma, the latter being recal-culated using 33 whole rock data from the UpperZone (i.e. the uppermost ¾2 km of the Bushveldstratigraphy, consisting of very homogeneous ferro-gabbronorite and ferro-diorite) reported by others[10,32,33]. The time difference between the twoapparent ages is not significant, since both decayconstants are somewhat uncertain. In the calcula-tion of λ187Re (1:666 š 0:017 ð 10�11 year�1 [34])via isochrons of iron meteorites, the uncertainty ofš1.02% is mainly due to spike calibration uncertain-ties. If this uncertainty is included in our isochronage calculation, the resulting upper uncertainty limitoverlaps with the Rb–Sr age.

Concerning the distribution of the different sam-ple localities over the entire western Bushveld lobe(Fig. 1), the isochron fit demonstrates a remarkableOs isotope homogeneity of this portion of the Bas-tard Unit at the time of consolidation. This is allthe more remarkable as the initial 187Os=188Os ratioof 0.1506 is much more radiogenic than that of thecontemporaneous chondritic mantle (i.e. 187Os=188Os

D 0.1128 at 2.04 Ga, see Fig. 2) and this exceptionalhomogeneity must therefore have been achieved af-ter considerable crustal contamination. The initial Osratio of the isochron is nevertheless lower than thevalue of 0.1773 obtained for sample SKN-19=26,which is the poikilitic pyroxenite of the MerenskyReef, and which lies within the range of Meren-sky Reef laurites (RuS2; 187Os=188Os values between0.168 and 0.181) determined by Hart and Kinloch[14].

The Rb–Sr data of the pyroxenites fitting the Re–Os isochron yield an errorchron age of 2027 š 160Ma (MSWD D 67.9) with an initial 87Sr=86Sr ratioof 0.70772 which also clearly indicates crustal con-tamination, as noted before. Sr isotope homogeneityin this portion of the Bushveld stratigraphy appearsto be less good than that of Os (Fig. 4). Heteroge-neous behavior of Sr isotopes within the MerenskyUnit, which directly underlies the Bastard Unit, hasalready been reported by Kruger [35].

3.2. The chromitite layers of the Critical Zone

Initial 187Os=188Os isotopic composition ofchromite and associated interstitial material (silicatesplus sulfides) from the lower group (LG) chromititelayers are relatively constant (¾0.120, Table 1) andonly slightly more radiogenic than the average valueof 0.111 for mantle peridotites from the Premiermine [36], corrected to 2043 Ma, or the chondriticmantle model at this time (i.e. 0.113, Fig. 2). Fromthe lower to the middle group chromitites, a sig-nificant increase in initial Os isotopic compositionfrom ¾0.120 to ¾0.137 can be observed, whereasthe UG1 layer, compared to the ones of the middlegroup, displays a decrease of the initial 187Os=188Osratio. Finally, the UG2 chromitite layer shows a fur-ther increase in initial 187Os=188Os ratios of chromiteas well as interstitials (up to 0.145 and 0.151, respec-tively).

In the lower group, chromite and interstitials ofLG6 have identical initial 187Os=188Os, whereas inthe LG5 chromitite layer the interstitial materialis somewhat less radiogenic than the chromite. Incontrast, in the higher chromitite layers MG3 upto UG2, the interstitials all display higher initial187Os=188Os ratios than their associated chromites(mainly those of the UG chromitites), but both follow

Page 6

54 R. Schoenberg et al. / Earth and Planetary Science Letters 172 (1999) 49–64

Tabl

e1

Isot

ope

data

ofch

rom

itese

para

tes

and

inte

rstit

ialp

hase

s(D

gang

ue)

ofC

ritic

alZ

one

chro

miti

tes,

aM

eren

sky

Ree

fpy

roxe

nite

and

four

pyro

xeni

tes

ofth

eB

asta

rdU

nit

Sam

ple

Uni

tTy

peR

eO

s18

7O

s=18

8O

s18

7R

e=18

8O

s18

7O

s=18

8O

sR

bSr

87Sr

=86

Sr87

Rb=

86Sr

87Sr

=86

Sr(p

pb)

(ppb

)(m

easu

red)

(at2

043

Ma)

a(p

pm)

(ppm

)(m

easu

red)

(at2

061

Ma)

BS-

1B

Upx

t1.

774.

020:

2223

š6

2.08

0:15

03š

152.

7355

.50:

7120

30š

200.

142

0:70

780

š16

SKN

-19=

15B

Upx

t1.

560.

797

0:46

96š

269.

240:

1496

š18

6.26

65.2

0:71

5440

š78

0.27

80:

7071

8š

20D

K1=

49.2

BU

pxt

0.41

10.

730

0:23

36š

132.

410:

1503

š27

1.06

41.1

0:71

0040

š12

0.07

490:

7078

2š

14D

K1=

52.4

BU

pxt

0.33

60.

275

0:28

77š

414.

000:

1492

š18

9.06

31.9

0:73

1870

š10

0.82

40:

7074

1š

36SK

N-1

9=26

MR

pxt

4.91

50.1

0:19

36š

10.

472

0:17

73š

184.

3458

.50:

7136

85š

580.

215

0:70

731

š18

H23

5=8-

Top

UG

2ch

r0.

534

22.6

0:14

83š

40.

0916

0:14

51š

25ch

r–

33.0

0:14

89š

7–

–ga

2.37

880

0:15

17š

20.

0128

0:15

13š

1713

.438

50:

7097

39š

160.

100

0:70

676

š15

UG

1-lo

wer

sect

ion

UG

1ch

r2.

4125

.20:

1416

š4

0.44

40:

1262

š16

chr

–36

.70:

1401

š3

––

ga3.

0114

70:

1399

š2

0.09

760:

1365

š18

1.00

84.7

0:70

9293

š15

0.03

410:

7082

8š

12

NG

3=15

2.1

MG

400

chr

0.82

717

.90:

1433

š5

0.19

50:

1366

š20

chr

–19

.60:

1464

š3

––

ga11

.933

30:

1458

š2

0.17

20:

1399

š25

6.08

18.4

0:70

7456

š19

0.07

760:

7051

5š

30

NG

2=21

1.5

MG

3ch

r0.

277

17.5

0:13

85š

50.

0467

0:13

69š

31ch

r–

18.5

0:13

92š

17–

–ga

0.66

213

90:

1389

š2

0.02

040:

1382

š17

1.06

90.3

0:71

2727

š34

0.03

400:

7112

6š

12

NG

1=25

6.9

LG

6ch

r0.

425

13.3

0:12

54š

60.

117

0:12

13š

16ch

r1.

0748

.20:

1265

š4

0.10

10:

1230

š13

ga9.

7528

20:

1276

š2

0.16

70:

1218

š13

26.5

71.7

0:73

9210

š46

1.07

0:70

743

š44

NG

1=33

1.5

LG

5ch

r0.

190

20.3

0:13

27š

50.

0307

0:13

16š

17ch

r–

23.6

0:13

29š

5–

–ga

3.69

46.5

0:13

27š

30.

365

0:12

01š

122.

029.

970:

7239

83š

370.

587

0:70

655

š29

NG

1=42

9.5

LG

4ch

r0.

369

8.76

0:12

51š

40.

143

0:12

02š

17ch

r–

8.11

0:12

24š

9–

–ch

r–

17.9

0:12

49š

6–

–ga

–49

.70:

1249

š4

––

0.58

022

.20:

7171

75š

620.

0760

0:71

492

š14

NG

1=47

5.3

LG

3ch

r0.

865

13.0

0:12

52š

100.

292

0:11

51š

15ch

r0.

389

3.87

0:12

80š

90.

354

0:11

57š

15ch

r–

6.76

0:12

87š

11–

–ga

–17

.00:

1283

š5

––

0.80

115

.90:

7255

37š

400.

147

0:72

118

š18

NG

1=56

0.1

LG

1ch

r0.

382

13.7

0:12

50š

90.

116

0:12

10š

15ch

r–

6.73

0:12

56š

7–

–ch

r–

10.9

0:12

35š

8–

–ga

–13

.90:

1259

š19

––

2.30

3.67

0:75

9294

š36

1.82

0:70

537

š66

NG

1,N

G2,

NG

3,D

K1

and

SKN

-19

Dbo

reho

les

(loc

aliti

esgi

ven

inFi

g.1;

for

mor

ede

tails

see

text

);L

GD

low

ergr

oup,

MG

Dm

iddl

egr

oup,

UG

Dup

per

grou

p,M

RD

Mer

ensk

yR

eef,

BU

DB

asta

rdU

nit;

pxtD

pyro

xeni

te,c

hrD

chro

mite

,ga

Dga

ngue

.All

unce

rtai

ntie

sar

ere

port

edat

the

2¦co

nfide

nce

leve

l.B

ack-

calc

ulat

ion

for

initi

al18

7O

s=18

8O

sra

tios

with

½18

7R

eD

1:66

6ð

10�1

1ye

ar�1

[34]

and

187O

s=18

8O

sun

cert

aint

ies

calc

ulat

edby

ast

atis

tical

erro

radd

ition

met

hod

incl

udin

gun

cert

aint

ies

on(1

)R

ean

dO

sm

easu

rem

ents

,(2)

wei

ghin

g,an

d(3

)rep

rodu

cibi

litie

sof

Re

and

Os

stan

dard

s.U

ncer

tain

ties

onth

ein

itial

87Sr

=86

Srra

tioar

eca

lcul

ated

usin

ga

1st

anda

rdde

viat

ion

onth

e87

Rb=

86Sr

ratio

ofš0

.5%

prev

ious

lyde

term

ined

on14

repl

icat

esof

aw

ell

hom

ogen

ized

nori

tesa

mpl

efr

omth

eB

ushv

eld

Com

plex

.aN

ewin

depe

nden

tRe–

Os

isoc

hron

age

(see

Fig.

3).

Page 7

R. Schoenberg et al. / Earth and Planetary Science Letters 172 (1999) 49–64 55

Fig. 3. Re–Os isochron plot of four pyroxenites of the BastardUnit. Regression was performed in ISOPLOT [64] and the agewas calculated using λ187Re D 1:666 ð 10�11 year�1 [34]. Alluncertainties are reported at the 2¦ confidence level.

increasing and decreasing trends with stratigraphicheight.

Osmium concentrations in the gangue fractionsof LG1 and LG3 are equal to their correspondingchromite, whereas in all the overlying chromititelayers, the Os concentrations of interstitial phasesexceed those of their corresponding chromite sepa-rates by up to 30 fold (Fig. 5). The only interstitialphases, able to enrich osmium sufficiently to causesuch high concentrations are the various sulfides,since osmium, like all PGE, is strongly chalcophile.Os concentrations of dry chromite splits from thesame sample may vary as much as 300%. This can-not be attributed to incomplete sample dissolution,since Re then would not be expected to behave simi-lar to Os and initial ratios would not be reproducible,as they are for LG3 and LG6 (Table 1, Fig. 2). Thishuge variability in Os concentrations of chromiteseparates from single layers might reflect a nuggeteffect due to Os being mainly (although not ex-clusively) hosted in sulfide microinclusions withinchromite (e.g. [2,29]).

Overall, Re–Os results of chromitites of this studyagree well with the data reported by McCandless etal. [16] on chromitites from the Brits Graben andthe Winterveld area, both regarding Os concentra-tions and trends in Os isotopic compositions withstratigraphic height.

The interstitial silicates of chromitite layers LG1,

LG5 and MG4 have initial Sr signatures withinthe range of 0.7045–0.7065, which generally char-acterizes silicate rocks from the Critical Zone [9].The interstitial silicates of all other chromitite lay-ers have significantly more radiogenic values thanthose of any whole rock or mineral separate anal-yses determined within the Critical Zone, with87Sr=86Sr-spikes of up to 0.721 (Fig. 4). Thus nocorrelations of Sr concentrations and=or isotope sig-natures with stratigraphic height or mineralogicalobservations within these chromitite horizons areapparent.

4. Discussion

4.1. Source of the radiogenic Os signatures

The 187Os=188Os ratios of the lowest chromititehorizons are close to the ‘chondritic’ mantle evo-lution, which is 0.1128 at 2043 Ma, as are the187Os=188Os ratios of 0.1131 and 0.1114 found fortwo erlichmanite grains (OsS2) from the MerenskyReef [14], demonstrating that effectively uncontam-inated mantle melt of chondritic Re–Os evolutionwas involved in the Bushveld formation, even at theMerensky Reef level. Incorporation of crustal ma-terial into the Bushveld magmas is apparent fromradiogenic signatures of other isotope systems [10–13], and this is also by far the most likely mechanismfor the introduction of radiogenic Os at all levelsabove the lowermost horizons. However, the isotopedata of the chromitite layers do not reveal if crustalcontamination occurred during ascent through thelower crust [37] or by assimilation of country rocksat the roof and sides of the magma chamber.

4.2. The Critical Zone chromitite layers

Conditions for chromite saturation and subse-quent crystallization in basic to ultrabasic meltshave been investigated experimentally by Murckand Campbell [38], who stress the fact that mix-ing of primitive and fractionated magma favors theformation of chromite-rich cumulates, but accumu-lation of massive chromitite layers will only occurif the composition of the mixed magma lies withinthe primary field of chromite [21]. In most of the

Page 8

56 R. Schoenberg et al. / Earth and Planetary Science Letters 172 (1999) 49–64

Fig. 4. Initial Sr data of interstitial phases of Critical Zone chromitite layers, a poikilitic pyroxenite of the Merensky Reef and poikiliticpyroxenites of the Bastard Unit. Sr stratigraphy (shaded area) and the range of initial Sr signatures throughout the whole Critical Zone(shaded, dotted area [9]) has been included to demonstrate that most of the gangue samples are more radiogenic than directly overlyingand underlying silicate rocks.

Critical Zone chromitite layers, euhedral (and to aminor extent subhedral) chromite crystals develop agrain-supported cumulate fabric, clearly demonstrat-ing that chromite growth was not confined and mostlikely occurred in a melt before accumulation intothese layers. Thus experimental and petrographic ev-idence reveals that mixing between a primitive and afractionated magma indeed is the most likely processleading to the formation of the massive chromititelayers of the Bushveld Critical Zone. Because of the

textural evidence for cumulus nature of the chromite,we consider that the initial osmium isotope signatureof the chromite separates (see Table 1) should reflectthose of the hybrid magma from which they crystal-lized, irrespective of whether this was chiefly incor-porated as sulfide microinclusion or into the chromitelattice itself. The radiogenic initial 187Os=188Os val-ues of chromite separates in all chromitite layersthen would indicate that crustal contamination of themagma occurred before chromite crystallization.

Page 9

R. Schoenberg et al. / Earth and Planetary Science Letters 172 (1999) 49–64 57

Fig. 5. Diagram showing the osmium concentrations ofchromite–gangue pairs of several chromitite layers from theCritical Zone of the Bushveld Complex.

The origin of PGE collected in interstitial sulfides(Fig. 5) of the chromitite layers is variously ascribedto (1) introduction of PGE related to metasomatismby upward percolation of either late magmatic flu-ids or interstitial melt, the latter being collected bycompaction of lower cumulates (e.g. [22,24–26], seealso [39,40] for the Merensky Reef), and (2) PGEconcentration by immiscible sulfide droplets, whichformed in the overlying magma and settled togetherwith chromite into the accumulation layer at thechambers floor [23].

If PGE were introduced into the interstitial sul-fides of the chromitite layers by an upward-perco-lating fluid=melt or the interstitial sulfides them-selves precipitated from such a melt, their initial187Os=188Os ratio would represent that of the melt,thus that of the cumulates underlying the chromi-tite, where the melt has been collected. In this case,some difference between the initial osmium isotopesignatures of chromite and interstitial phases withinchromitite layers would be expected [27], since ini-tial 187Os=188Os ratios vary significantly with strati-graphic height. Magma mixing, on the other hand,may not only be responsible for chromite crystal-lization, but could also result in S oversaturationof the hybrid magma, from which, subsequently,immiscible sulfide droplets segregate, scavengingPGE from the hybrid magma they pass throughduring settling. If both, cumulus chromite crystalsand interstitial sulfides derived their PGE from thesame hybrid magma, it is expected that they would

have broadly similar initial Os isotopic composi-tions, as is observed. Although initial 187Os=188Osratios of interstitial phases from the upper CriticalZone are more radiogenic than those of their associ-ated chromites (whereas those of the middle groupstill overlap within the 2¦ uncertainty), the parallelincreasing and decreasing trends of osmium isotopesignatures of both clearly demonstrates a link be-tween the reservoirs they scavenged their osmiumfrom. This observation points strongly to an accu-mulative magmatic deposition of the PGE-enrichedsulfides into the chromitite horizons. The offset ininitial 187Os=188Os ratios between interstitial sulfidesand cumulus chromites apparent in some chromi-tite layers could reflect changing mixing proportionsin the hybrid magma during segregation of thesephases.

The following mechanism is proposed to explainthe highly radiogenic initial Sr signatures of someinterstitial phases of chromitites compared to theirenclosing silicate rocks (see Fig. 4). Fountains ofinjected melt [41] penetrated through the residualmagma to the roof of the chamber, where the foun-tain top entrained the overlying granophyric roofmelt. This crustally derived melt would be charac-terized by low osmium but high strontium concen-trations. Since the roof rocks of the Bushveld Com-plex are derived from the approximately 3.0 Ga oldKaapvaal Craton both Os and Sr would have radio-genic isotope signatures compared to mantle meltsat Bushveld times. Contamination of the injectedmelt with the residual one in the chamber causedCr saturation in the resulting hybrid magma, leadingto crystallization of chromite, which settled to thechamber floor to form the massive chromitite lay-ers. Thereby, the blending with the granophyric roofmelt could have led to the highly radiogenic initial87Sr=86Sr ratios typically found for most interstitialphases (i.e. the gangue phases) of the chromitite lay-ers. The rapid density changes caused by chromitecrystallization [42] led to a compositional overturnof this hybrid layer with a neighboring one or evenwith the residual magma (e.g. [41,43]). Such furthermixing could lead to a new hybrid magma undersatu-rated in chromite and with less radiogenic Sr isotoperatios such as are reflected in the directly overlyingchromite-free cumulates (Fig. 4).

Page 10

58 R. Schoenberg et al. / Earth and Planetary Science Letters 172 (1999) 49–64

4.3. The Merensky Reef

Merensky Reef laurites display highly variable Ossignatures [14], which are also much more radio-genic than that of any under- or overlying layersdetermined so far (Fig. 2). Furthermore, the Re–Osisochron fit of the overlying pyroxenites from theBastard Unit, reproducing the normally cited crys-tallization age of the Bushveld Complex, reveals astriking initial osmium isotope homogeneity at thislevel in the Complex, although Os concentrationsare low (Table 1). Therefore, introduction of to-tal or radiogenic osmium into the Merensky Reefby a postcumulus, upwardly percolating fluid=meltseems very unlikely, since either the fluid=melt musthave undergone significant changes in its Os isotopiccomposition during penetration through the reef orvery radiogenic osmium must have been emplacedjust within the high-PGE reef without affecting theoverlying low-PGE pyroxenites of the Bastard Unit.These findings suggest that primary enrichment ofPGE into the sulfide phases of the Merensky Reefwas of magmatic origin, most probably by a pro-cess very similar to the one leading to the sulfidedeposition into the chromitite layers, which is bysegregation of an immiscible sulfide melt duringmagma mixing.

The mantle 187Os=188Os ratios of the erlichman-ites analyzed by Hart and Kinloch [14] (Fig. 2) showthat the sulfide melt from which they crystallizedsegregated from a mantle-derived magma before anycrustal contamination. This implies that an injec-tion of new primitive magma at the Merensky Reeflevel must have occurred, since the residual magmaat that time was already crustally contaminated, asdemonstrated by radiogenic signatures of differentisotope systems in directly underlying rocks andthe reef itself. Further, it is unlikely that this seg-regation happened at mantle depth and the sulfidethen was entrained upwards: Naldrett [44], Naldrettand Cabri [45] and Mavrogenes and O’Neill [46]experimentally investigated factors controlling S ca-pacity at sulfide saturation in silicate melts (i.e. fS2 ,fO2 , temperature, pressure and melt composition).Mavrogenes and O’Neill [46] showed a strong in-crease in S solubility for mafic silicate melts withdecreasing pressure (e.g. 200 ppm at 40 kbar to 1200ppm at 0 kbar for a basaltic melt with a tempera-

ture of 1400ºC). They concluded that no sulfide meltcan segregate from an ascending basaltic magma,unless a significant compositional change occurs,which can happen through fractionation, assimila-tion or mixing. Thus sulfide droplets segregated atmantle depth from a basaltic magma would mostlikely be redissolved during ascent through the crust.However, mixing of a primitive ascending melt witha more evolved melt resident in the magma cham-ber may produce a hybrid magma body which washeterogeneous with respect to mixing proportionsand, therefore, Os and Sr isotopes (e.g. [41]). Seg-regating sulfide droplets could have scavenged PGEfrom different parts of such an isotopically hetero-geneous magma. This process can readily accountfor the range in radiogenic 187Os=188Os ratios of theMerensky Reef laurites as well as the erlichman-ites with mantle signatures, which all accumulatedwithin the same horizon. Together with these erlich-manites, the near-mantle Os signatures of the lowergroup chromites provide the best direct evidencethat newly arriving magma batches underwent nosignificant crustal contamination during ascent.

It has been well documented that late and post-magmatic processes significantly affected the geo-chemistry of some (if not most) horizons of theBushveld Complex including the Merensky Reef(e.g. [22,39,40]). Nevertheless, our conclusion thatthe primary PGE enrichment was magmatic is inagreement with results of Peach and Mathez [47–49] who showed that Pd=Pt, Pd=Cu and Pd=Au ra-tios were within the range expected for magmaticdeposits.

4.4. Amount and mode of crustal contamination

In this section mixing of osmium and strontiumisotopes between a mantle melt reservoir and anevolved continental crust (i.e. the Kaapvaal Craton)is simulated to allow appropriate estimations of thecrustal contamination needed to explain the initialisotope signatures observed in Critical Zone rocksused in this study.

Based on the mantle Os signatures of the erlich-manites, the near-mantle initial Os composition ofthe lower group chromitites and geochemical indica-tions, primitive end-members for modeling were de-fined as juvenile mantle melts [50,51]. Melt fractions

Page 11

R. Schoenberg et al. / Earth and Planetary Science Letters 172 (1999) 49–64 59

Fig. 6. (A) Binary mixing between mantle melts with different melt fractions and continental crusts of different isotopic signatures inthe initial 187Os=188Os vs. initial 87Sr=86Sr space. Gangues of the lower Critical Zone ✧, gangues of the upper Critical Zone diamf,a poikilitic pyroxenite of the uppermost Merensky Reef ✚, and pyroxenites of the Bastard Unit ✩. Present-day 187Os=188Os ratios,indicated by arrows, are given for some lower Critical Zone samples, due to failure of Re measurement. However, the stippled linerepresents the mantle Os signature at the time of the Bushveld intrusion and thus is the minimum value for the back correction. Thisalso demonstrates the small time corrections for present-day 187Os=188Os ratios needed, which is a result of the low Re concentrationscompared to Os in the gangue phases. (B) Close up of (A) displaying the evolution of the two isotope systems in more detail. Legend isidentical to (A). For explanation of the step numbers 1 to 3 see text.

of 2% (¾alkali-basalt) and 10% (¾MOR-basalt)were chosen (Fig. 6A). The Bushveld Complexmagma chamber is located in Transvaal sediments,derived from an Archean crustal segment, the Kaap-vaal Craton, formed around 3000 Ma. Evolution of87Sr=86Sr from 0.7009 at 3000 Ma (value of the uni-form reservoir, UR, at that time) up to 0.7195 at 2061Ma is realistic for high Rb=Sr (¾1.3) reservoirs, suchas upper continental crust. Other, coeval intrusionsin this region of the Kaapvaal Craton also show vari-able contamination with crustal material in their Sr,Nd and Pb signatures. Barton et al. [52,53] found

variably elevated Sr and Pb isotope ratios and unra-diogenic (crustal) Nd signatures in the Schiel com-plex in the northern Transvaal and in the Sand Rivergneisses of the Limpopo Mobile Belt. Further, Kam-ber et al. [54] found that 2.57 Ga sedimentary rocksin the northeastern Kaapvaal Craton had elevatedinitial 87Sr=86Sr ratios ranging between 0.720 and0.750. Nevertheless, a second mixing curve has beenincluded which used an average 87Rb=86Sr value forcontinental crust of 0.66, yielding a 87Sr=86Sr ratioof 0.710 at 2061 Ma. All modeling parameters andrespective references are listed in Appendix B.

Page 12

60 R. Schoenberg et al. / Earth and Planetary Science Letters 172 (1999) 49–64

The fact that gangue samples do not directly ploton the calculated mixing curves in the 187Os=188Osvs. 87Sr=86Sr space (Fig. 6A) does not mean thatcrustal contamination is not a valid mechanism toproduce the radiogenic values observed in Bushveldrocks. It probably reflects a decoupling of Re–Osand Rb–Sr isotope systematics during the processesof magma mixing and sulfide segregation. Therefore,the formation history of the massive chromitites andthe Merensky Reef cannot be portrayed by a simplebinary mixing model.

The high Os concentrations and the subchondritic187Re=188Os of Critical Zone samples of interstitialmaterial (mantle melts are expected to be supra-chondritic in this ratio) demonstrate that their Re–Os system is strongly dominated by sulfides andPGE-alloys and thus represents processes which areconnected with osmium scavenging into the segre-gated sulfide melt. Thereby, Os from the crustallycontaminated magma must have been incorporatedinto the sulfide melt. It probably reflects a decouplingof initial 187Os=188Os ratios of the gangues, whichare minimum values for the Os isotopic compositionof the residual magma at the moment of new meltinjections. Initial 187Os=188Os ratios can therefore beused to estimate minimal crustal contamination pro-portions at the different levels of the Critical Zone,which is about 5% for lower group gangues and upto 20% for the Merensky Reef level (steps 1 and 2 inFig. 6B). These amounts are merely the proportionsof modeled, evolved crust needed to bring the initial187Os=188Os signature of a 10% mantle melt at 2060Ma b.p. up to the values observed and do not entailinterpretations about the magma mixing processesduring the Critical Zone formation, as those are notunequivocal using the trends in Os and Sr isotopesignatures.

The peak in initial 187Os=188Os ratios of theMerensky Reef compared to the underlying UG2chromitite layer and the range of Merensky Reeflaurites in this ratio from 0.168 to 0.181, point to anevent leading to a drastic change in the compositionof the crystallizing magma during the formation ofthis layer. The highly radiogenic 187Os=188Os sig-natures of the laurites indicate a strong influenceof a highly crustally contaminated magma, whichis also suggested by the sudden increase of initial87Sr=86Sr ratios of Merensky Reef rocks compared

to underlying layers. However, initial 187Os=188Osratios return to a constant, but less radiogenic valuein the overlying Bastard Unit, whereas initial Sr sig-natures continuously increase throughout this levelof the complex [35].

Of all possible scenarios, which might explain thetrends of the Re–Os and Rb–Sr isotope systematicsat the Merensky Reef level, we prefer a plume-likeinjection of a new primitive magma into a stratifiedmagma chamber, where the plume mixed with highlycontaminated granophyric roof melt at the top of thechamber. The resulting hybrid magma was charac-terized by high 187Os=188Os and 87Sr=86Sr ratios,explaining the sudden increase of initial isotope val-ues of both systems compared to underlying rocks.Further, it is suggested that the decrease of initial187Os=188Os ratios immediately above the MerenskyReef (step 3 in Fig. 6B) demonstrates a replenish-ment of the hybrid magma in PGE by segregatedsulfide droplets during this injection event, whichcontinued and thus, after magma homogenization,forced the Os signatures down to more primitivevalues. In contrast to Os, the radiogenic Sr in theresidual magma and the granophyric roof melt isnot affected by the sulfide melt; its concentration re-mains high and the 87Sr=86Sr ratio may even increaseafter the formation of the Merensky Reef.

5. Summary and conclusions

Our preferred interpretation of the primary enrich-ment of PGE into cumulus chromite and intercum-ulus sulfide phases of the chromitite layers and theMerensky Reef is a magmatic process of the kinddescribed in earlier work (e.g. [23]). Introduction ofPGE into these layers by upward migration of postcu-mulus fluid=melt cannot be reconciled with the newRe–Os data. This does not exclude a postcumulusor postcrystallization redistribution of PGE on a lo-cal scale by late magmatic fluids and=or upwardlymigrating intercumulus melt, collected during com-paction of lower cumulates (e.g. [22,24–26]).

Offsets in initial 187Os=188Os signatures betweenchromite–gangue pairs of the upper Critical Zone, thehighly radiogenic initial 87Sr=86Sr ratios of interstitialphases of almost all chromitite layers as well as thehighly variable initial osmium isotope signatures of

Page 13

R. Schoenberg et al. / Earth and Planetary Science Letters 172 (1999) 49–64 61

Merensky Reef laurites indicate a highly transient en-vironment undergoing rapid compositional changesduring the formation of these layers [41,43].

The near mantle values of initial Os signatures oflower group chromitites, inverse geochemical frac-tionation trends after injections [6] as well as themantle Os signatures of the erlichmanites foundin the Merensky Reef [14] indicate that the newmagma injections throughout the Critical Zone mostprobably consisted of juvenile mantle melt with nosignificant contamination during ascent.

The similar Os concentrations of mantle melts(¾95 ppt) and continental crust (¾50 ppt) and thecompatible behavior of this element make Os sys-tematics a viable tool to estimate average crustalcontamination in mafic intrusions, which lies be-tween 5% for the lower and up to 20% for theupper Critical Zone of the Bushveld Complex. Fur-ther, they allow direct insight in the evolution ofore-forming minerals such as chromite or platinum-group minerals and thus give unique informationabout the genesis of ore deposits in layered com-plexes. The Sr isotopic system, on the other hand, isa good indicator for local processes during crystal-lization and accumulation of mineral layers and lateto postmagmatic fluid=melt percolation.

Acknowledgements

We thank E.A. Mathez, I.H. Campbell and J.E.Snow for detailed reviews which significantly helpedto improve the quality of this paper. We gratefullyacknowledge financial support by Swiss NationalFonds grant number 20-47157.96. F.J. Kruger waspartly supported by the FRD, De Beers and WitsUniversity during the time that this work was done.[CL]

Appendix A. Analytical techniques

Samples from chromitite layers were crushed and chromiteconcentrates were obtained using magnetic and heavy liquidseparation techniques. Any composite grains and chromite withvisible ongrowths or inclusions were removed from the chromiteseparates by careful hand-picking under a binocular microscopewith a magnification of 50 times. The ‘gangues’ are intersti-tial silicate (orthopyroxene š clinopyroxene š plagioclase) and

sulfide, from which the chromite was largely (>98%) but notcompletely removed after magnetic separation.

For Re–Os isotopic determinations, 100–350 mg of the sam-ple powders and appreciable amounts of 185Re and 190Os spikeswere weighed into Carius tubes [55] and wetted with 10 mlof inverse aqua regia. Sample dissolution, oxidation of Re andOs, as well as sample–spike homogenization was achieved byheating the closed Carius tubes in an oven at 225ºC for 7 days.After cracking the Carius tubes, Os was directly distilled inone step from the digestion solution into HBr, using the tech-nique described by Nagler et al. [56], and purified, using themicrodistillation described by Roy-Barman et al. [57]. Re wasseparated from the residual solution of the first Os distillationstep by solvent extraction [58] and purified by anion-exchangein microcolumns (80 μl of Dowex AG1-X8 resin) using eitherHBr–HCl or conventional HNO3 eluants. For the Rb–Sr isotopicwork, 50–100 mg of sample powder (gangue) was weighed intopre-spiked (87Rb and 84Sr) Teflon beakers and dissolved on a hotplate using a mixture of HF and HNO3. After drying down, theresidue was wetted with HCl for conversion to chloride. Rb andSr were separated by standard cation-exchange techniques.

Re and Os isotopic compositions were determined byN-TIMS [59,60] on a modified single-collector AVCO 9000mass spectrometer equipped with a low-noise secondary elec-tron multiplier at the Mineralogisch.–Petrographisches Institut,Gruppe Isotopengeologie, Universitat Bern (Switzerland). Rewas measured as ReO�

4 complex from Ta-filaments [61]. Os wasmeasured from Pt-filaments using Ba(NO3)2 as electron donor asOsO�

3 with ionization enhanced by bleeding O2 into the source(to ¾ 1 ð 10�6 Torr). Fractionation corrections for Os weremade using 192Os=188Os D 3.082614 [62] for normalization.Oxygen correction was performed using O isotope abundancesdescribed by Nier [63]. The in-run precision of the machineduring this work was assessed from the stable isotope ratio189Os=188Os of the samples themselves. This value was found tobe 1:2230 š 0:0029 (uncertainty is 2¦ of the population; n D 32.It is slightly higher (0.26%), but within uncertainties identical to189Os=188Os D 1:2198 š 0:0018 (n D 40) achieved by measuringstandards to control the long-term reproducibility of the instru-ment. Total procedure blanks were <20 pg for Re and <1 pg forOs.

Rb and Sr isotope determinations were done at the HughAllsopp Laboratory, Bernard Price Institute, University of Wit-watersrand (South Africa) by TIMS on a VG MM-30 equippedwith a Keithley electrometer and a VG-354 multi-collector in-strument using Pyramid TIMS software. Rb was transferred inchloride form on outgassed double Ta-filaments and Sr wasloaded with phosphoric acid on single Ta-filaments. Correctionsfor fractionation and spike contribution on the 87Sr=86Sr ratiowere made using an exponential law. Runs of nine spiked andunspiked NIST SRM-987 Sr standards yielded a 87Sr=86Sr ratioof 0:71023 š 4 (2¦ ) during the period of this work. Total pro-cedure blanks have with <100 pg for Rb and <1 ng for Sr noinfluence on the results and are ignored.

Page 14

62 R. Schoenberg et al. / Earth and Planetary Science Letters 172 (1999) 49–64

Appendix B

Parameters used for modeling Os–Sr mixing (Fig. 6)

Reservoir Os conc. 187Re=188Os 187Os=188Os Sr conc. 87Rb=86Sr 87Sr=86Sr(ppb) (at 2061 Ma) (ppm) (at 2061 Ma)

Mantle melt 2% 0.090 – 0.111 872 – 0.702Mantle melt 10% 0.098 – 0.111 238 – 0.7023 Ga average crust 0.053 39 0.717 333 0.66 0.713 Ga upper crust 0.053 39 0.717 333 1.3 0.719

187Os=188Os mantle melt average of peridotites from the Premier mine [36].187Os=188Os crust back-calculated from CHUR value at 3 Ga (chondritic 187Re=188Os D 0.40076 and 187Os=188Os D 0.12672

[65]) using crustal 187Re=188Os D 39.Os conc. mantle melt calculated, using average Os conc. [36] and a KD Os D 50 [66].Os conc. crust average value of continental crust [19].87Sr=86Sr mantle melt UR evolution at 2.061 Ga.87Sr=86Sr crust back-calculated from UR value at 3 Ga with 87Rb=86Sr D 0.66 for average and 1.3 for upper crust and a Sr

concentration of 333 ppm.Sr conc. mantle melt calculated using KD Sr D 0.01.

References

[1] R.N. Scoon, W.J. de Klerk, The relationship of olivinecumulates and mineralization to cyclic units in part of theupper Critical Zone of the western Bushveld Complex, Can.Mineral. 25 (1987) 51–77.

[2] S.A. Hiemstra, The role of collectors in the formation of theplatinum deposits in the Bushveld Complex, Can. Mineral.17 (1979) 469–482.

[3] G. von Gruenewaldt, C.J. Hatton, R.K.W. Merkle, S.B.Gain, Platinum-group element — chromitite associations inthe Bushveld Complex, Econ. Geol. 81 (1986) 1067–1079.

[4] J. Willemse, The geology of the Bushveld igneous complex,the largest repository of magmatic ore deposits in the world,in: H.D.B. Wilson (Ed.), Magmatic Ore Deposits, Econ.Geol. Monogr. 4 (1969) 1–22.

[5] F.J. Kruger, The stratigraphy of the Bushveld Complex: areappraisal and the relocation of the Main Zone boundaries,S. Afr. J. Geol. 93 (1990) 376–381.

[6] H.V. Eales, W.J. de Klerk, B. Teigler, Evidence for magmamixing processes within the Critical and Lower Zones ofthe northwestern Bushveld Complex, South Africa, Chem.Geol. 88 (1990) 261–278.

[7] H.V. Eales, W.J. de Klerk, A.R. Butcher, F.J. Kruger, Thecyclic unit beneath the UG1 chromitite (UG1FW unit) atRPM Union Section Platinum Mine-Rosetta Stone of theBushveld upper Critical Zone?, Mineral. Mag. 54 (1990)23–43.

[8] B. Teigler, Mineralogy, Petrology and Geochemistry of theLower and Lower Critical Zones, Northwestern BushveldComplex, Unpublished PhD Thesis, Rhodes University,1990.

[9] F.J. Kruger, The Sr-stratigraphy of the western BushveldComplex, S. Afr. J. Geol. 97 (1994) 393–398.

[10] J. Hamilton, Isotope and trace element studies of the GreatDike and Bushveld mafic phase, J. Petrol. 18 (1977) 24–52.

[11] R.E. Harmer, J.M. Auret, B.M. Eglington, Lead isotopevariations within the Bushveld Complex, Southern Africa; areconnaissance study, J. Afr. Earth Sci. 21 (1995) 595–606.

[12] C.M. Schiffries, D.M. Rye, Stable isotope systematics ofthe Bushveld Complex, I, Constraints of magmatic pro-cesses in layered intrusions, Am. J. Sci. 289 (1989) 841–873.

[13] C.M. Schiffries, D.M. Rye, Stable isotopic systematics ofthe Bushveld Complex, II, Constraints on hydrothermalprocesses in layered intrusions, Am. J. Sci. (1990) 209–245.

[14] S.R. Hart, E.D. Kinloch, Osmium isotope systematics inWitwatersrand and Bushveld ore deposits, Econ. Geol. 84(1989) 1651–1655.

[15] T.E. McCandless, J. Ruiz, Osmium isotopes and crustalsources for platinum-group mineralization in the BushveldComplex, South Africa, Geology 19 (1991) 1225–1228.

[16] T.E. McCandless, J. Ruiz, B.I. Adair, Magma and platinum-group-element sources in the Bushveld Complex inferredfrom rhenium–osmium, rubidium–strontium and platinum-group- element systematics, in: 7th Annual V.M. Gold-schmidt Conference, Lunar and Planetary Institute, Hous-ton, 1997, p. 134.

[17] R.J. Walker, J.W. Morgan, M.F. Horan, Osmium-187 en-richment in some plumes: evidence for core–mantle inter-action?, Science 269 (1995) 819–822.

[18] R.J. Walker, J.W. Morgan, E.S. Beary, M.I. Smoliar,G.K. Czamanske, M.F. Horan, Applications of the 190Pt–186Os isotope system to geochemistry and cosmochemistry,Geochim. Cosmochim. Acta 61 (1997) 4799–4807.

[19] B.K. Esser, K.K. Turekian, The osmium isotopic composi-tion of the continental crust, Geochim. Cosmochim. Acta57 (1993) 3093–3104.

[20] C.J. Allegre, J.M. Luck, Osmium isotopes as petrogeneticand geological tracers, Earth Planet. Sci. Lett. 48 (1980)148–154.

Page 15

R. Schoenberg et al. / Earth and Planetary Science Letters 172 (1999) 49–64 63

[21] T.N. Irvine, Origin of chromitite layers in the Muskoxintrusion, Geology 5 (1977) 273–277.

[22] E.A. Mathez, R.H. Hunter, R. Kinzler, Petrologic evolu-tion of partially molten cumulate: the Atok section of theBushveld Complex, Contrib. Mineral. Petrol. 129 (1997)20–34.

[23] I.H. Campbell, A.J. Naldrett, S.J. Barnes, A model for theorigin of the platinum-rich sulfide horizons in the Bushveldand Stillwater Complexes, J. Petrol. 24 (1983) 133–165.

[24] C.G. Ballhaus, E.F. Stumpfl, Sulfide and platinum min-eralization in the Merensky Reef: evidence from hydroussilicates and fluid inclusions, Contrib. Mineral. Petrol. 94(1986) 193–204.

[25] A.E. Boudreau, E.A. Mathez, I.S. McCallum, Halogen geo-chemistry of the Stillwater and Bushveld Complexes: ev-idence for transport of the platinum-group elements byCl-rich fluids, J. Petrol. 27 (1986) 967–986.

[26] A.E. Boudreau, I.S. McCallum, Concentration of platinum-group elements by magmatic fluids in layered intrusions,Econ. Geol. 87 (1992) 1830–1848.

[27] C.E. Martin, Re–Os investigation of the Stillwater Com-plex, Montana, Earth Planet. Sci. Lett. 93 (1989) 336–344.

[28] D.D. Lambert, R.J. Walker, J.W. Morgan, S.B. Shirey,R.W. Carlson, M.L. Zientek, B.R. Lipin, M.S. Koski, R.L.Cooper, Re–Os and Sm–Nd isotope geochemistry of theStillwater Complex, Montana: implications for the petroge-nesis of the J-M Reef, J. Petrol. 35 (1994) 1717–1753.

[29] F. Marcantonio, A. Zindler, L. Reisberg, E.A. Mathez, Re–Os systematics in chromitites from the Stillwater Complex,Montana, USA, Geochim. Cosmochim. Acta 57 (1993)4029–4037.

[30] R.N. Scoon, B. Teigler, Platinum-group element mineraliza-tion in the Critical Zone of the western Bushveld Complex,I, sulfide poor-chromitites below the UG-2, Econ. Geol. 89(1994) 1094–1121.

[31] A. Streckeisen, To each plutonic rock its proper name,Earth-Sci. Rev. 12 (1976) 1–33.

[32] M.R. Sharpe, Strontium isotope evidence for preserveddensity stratification in the Main Zone of the BushveldComplex, South Africa, Nature 316 (1985) 119–126.

[33] F.J. Kruger, R.G. Cawthorn, K.L. Walsh, Strontium isotopicevidence against magma addition in the Upper Zone of theBushveld Complex, Earth Planet. Sci. Lett. 84 (1987) 51–58.

[34] M.I. Smoliar, R.J. Walker, J.W. Morgan, Re–Os ages ofgroup IIA, IIIA, IVA, and IVB meteorites, Science 271(1996) 1099–1102.

[35] F.J. Kruger, The origin of the Merensky Cyclic Unit: Sr-isotopic and mineralogical evidence for an alternative or-thomagmatic model, Aust. J. Earth Sci. 39 (1992) 255–261.

[36] D.G. Pearson, R.W. Carlson, S.B. Shirey, F.R. Boyd, P.H.Nixon, Stabilisation of Archean lithospheric mantle: a Re–Os isotope study of peridotite xenoliths from the Kaapvaalcraton, Earth Planet. Sci. Lett. 134 (1995) 341–357.

[37] H.E. Huppert, R.S.J. Sparks, Cooling and contaminationof mafic and ultramafic magmas during ascent through

continental crust, Earth Planet. Sci. Lett. 74 (1985) 371–386.

[38] B.W. Murck, I.H. Campbell, The effects of temperature,oxygen fugacity and melt composition on the behaviourof chromium in basic and ultrabasic melts, Geochim. Cos-mochim. Acta 50 (1986) 1871–1887.

[39] D.M. Nicholson, E.A. Mathez, Petrogenesis of the Meren-sky Reef in the Rustenberg section of the Bushveld Com-plex, Contrib. Mineral. Petrol. (1991) 293–309.

[40] E.A. Mathez, Magmatic metasomatism and formation ofthe Merensky Reef, Bushveld Complex, Contrib. Mineral.Petrol. (1995) 277–286.

[41] I.H. Campbell, J.S. Turner, Fountains in magma chambers,J. Petrol. 30 (1989) 885–923.

[42] R.S.J. Sparks, H.E. Huppert, Density changes during thefractional crystallization of basaltic magmas: fluid dynamicimplications, Contrib. Mineral. Petrol. 85 (1984) 300–309.

[43] H.E. Huppert, R.S.J. Sparks, The fluid dynamics of abasaltic magma chamber replenished by influx of hot, denseultrabasic magma, Contrib. Mineral. Petrol. 75 (1980) 279–289.

[44] A.J. Naldrett, Nickel sulfide deposits — their classificationand genesis, with special emphasis on deposits of volcanicassociation, Can. Inst. Min. Metall. Trans. 76 (1973) 183–201.

[45] A.J. Naldrett, L.J. Cabri, Ultramafic and related maficrocks: their classification and genesis with special refer-ence to concentration of nickel sulfides and platinum-groupelements, Econ. Geol. 71 (1976) 1131–1158.

[46] J.A. Mavrogenes, H.St.C. O’Neill, Pressure and temper-ature dependence of sulfide solubility in mafic magmas,in: 7th Annual V.M. Goldschmidt Conference, Lunar andPlanetary Institute, Houston, 1997, p. 134.

[47] C.L. Peach, E.A. Mathez, Constraints on the formation ofplatinum-group element deposits in igneous rocks, Econ.Geol. 91 (1996) 439–450.

[48] C.L. Peach, E.A. Mathez, R.R. Keays, Sulfide melt–silicatemelt distribution coefficients for noble metals and otherchalcophile elements as deduced from MORB: implicationsfor partial melting, Geochim. Cosmochim. Acta 54 (1990)3379–3389.

[49] C.L. Peach, E.A. Mathez, R.R. Keays, S.J. Reeves, Ex-perimentally determined sulfide melt–silicate melt partitioncoefficients for iridium and palladium, Chem. Geol. 117(1994) 361–377.

[50] S.A. Hiemstra, The distribution of chalcophile and plat-inum-group elements in the UG-2 chromitite layer of theBushveld Complex, Econ. Geol. 81 (1986) 1080–1086.

[51] C.A. Lee, Trace and platinum-group element geochemistryand the development of the Merensky Unit of the westernBushveld Complex, Miner. Deposita 18 (1983) 173–190.

[52] J.M. Barton Jr., E.S. Barton, C.B. Smith, Petrography,age and origin of the Schiel alkaline complex, northernTransvaal, South Africa, J. Afr. Earth Sci. 22 (1996) 135–145.

[53] J.M. Barton Jr., B. Ryan, R.E.P. Fripp, Rb–Sr and U–Th–Pb studies of the Sand River gneisses, Central Zone,

Page 16

64 R. Schoenberg et al. / Earth and Planetary Science Letters 172 (1999) 49–64

Limpopo Mobile Belt, in: W.J. van Biljon, J.H. Legg (Eds.),The Limpopo Belt, Spec. Publ. Geol. Soc. S. Afr., Johan-nesburg, 1983, pp. 9–18.

[54] B.S. Kamber, M. Bau, N.J. Beukes, R.K. O’Nions, R.M.Corfield, Deposition of banded-iron formation by isolationof continental basins, in: 7th Annual V.M. GoldschmidtConference, Lunar and Planetary Institute, Houston, 1997,p. 110.

[55] S.B. Shirey, R.J. Walker, Carius tube digestions for low-blank rhenium–osmium analysis, Anal. Chem. 67 (1995)2136–2141.

[56] T.F. Nagler, R. Frei, ‘Plug in’ Os distillation, Schweiz.Mineral. Petrogr. Mitt. 77 (1997) 123–127.

[57] M. Roy-Barman, C.J. Allegre, 187Os=186Os in oceanic is-land basalts: tracing oceanic crust recycling in the mantle,Earth Planet. Sci. Lett. 129 (1995) 145–161.

[58] R.J. Walker, Low-blank chemical separation of rheniumand osmium from gram quantities of silicate rock for mea-surement by resonance ionization mass spectrometry, Anal.Chem. 60 (1988) 1231–1234.

[59] R.A. Craeser, D.A. Papanastassiou, G.J. Wasserburg, Nega-tive thermal ion mass spectrometry of osmium, rhenium andiridium, Geochim. Cosmochim. Acta 55 (1991) 397–401.

[60] J. Voelkening, T. Walczyk, K.G. Heumann, Osmiumisotope ratio determinations by negative thermal ionizationmass spectrometry, Int. J. Mass. Spectrom. Ion Process.105 (1991) 147–159.

[61] R. Frei, Th.N. Nagler, Th. Meisel, Efficient N-TIMSrhenium isotope measurements on outgassed tantalumfilaments: very low filament blanks determined by a‘standard addition’ approach, Int. J. Mass Spectrom. 153(1996) 7–10.

[62] A.O. Nier, The isotopic constitution of osmium, Phys. Rev.52 (1937) 885.

[63] A.O. Nier, A redetermination of the relative abundancesof the isotopes of carbon, nitrogen, oxygen, argon andpotassium, Phys. Rev. 77 (1950) 789–793.

[64] K.R. Ludwig, ISOPLOT, a plotting and regression programfor radiogenic isotope data, USGS, Open-File Rep. OF91-445 (1994).

[65] J.J. Shen, D.A. Papanastassiou, G.J. Wasserburg, PreciseRe–Os determinations and systematics of iron meteorites,Geochim. Cosmochim. Acta 60 (1996) 2887–2900.

[66] J.H. Jones, M.J. Drake, Geochemical constraints on coreformation in the Earth, Nature 322 (1986) 221–228.