Chemical Geology (2012) Contents lists available at SciVerse ScienceDirect. Chemical Geology

Transcription

1 Chemical Geology (2012) Contents lists available at SciVerse ScienceDirect Chemical Geology journal homepage: Research paper The 190 Pt 186 Os decay system applied to dating platinum-group element mineralization of the Bushveld Complex, South Africa J.A. Coggon a,, G.M. Nowell a, D.G. Pearson a, T. Oberthür b, J.-P. Lorand c, F. Melcher b, S.W. Parman d a Northern Centre for Isotopic and Elemental Tracing, Department of Earth Sciences, Durham University, South Road, Durham DH1 3LE, UK b Federal Institute for Geosciences and Natural Resources, Stilleweg 2, 30655, Hannover, Germany c Laboratoire de Minéralogie et Cosmochimie, Muséum National d'histoire Naturelle and CNRS (UMR 7202), 61 rue Buffon, Paris 75005, France d Department of Geological Sciences, Brown University, Providence, Rhode Island 02912, USA article info abstract Article history: Received 10 September 2010 Received in revised form 12 October 2011 Accepted 18 October 2011 Available online 25 October 2011 Keywords: Pt Os Isotopes Dating Geochronology Bushveld PGM We evaluate the 190 Pt 186 Os decay system, measured by laser ablation MC-ICPMS, as a useful geochronometric tool for direct dating of platinum-group minerals (PGM) in large mafic/ultra-mafic intrusions by analysing mineralised rocks from the geochronologically well constrained Bushveld Complex, S. Africa. Mixed PGM (laurite, cooperite, laurite platarsite, sperrylite and Pt Fe alloy) from the Merensky Reef yield a Pt Os isochron age of 1995±50 Ma (2σ, n=48, MSWD=1.16, initial 186 Os/ 188 Os= ± ). While this is 59 Myr younger than the U Pb zircon age for the Merensky Reef (Scoates and Friedman, 2008) it is consistent with recent Re Os ages for the Platreef. Considering the younger age recorded by both Re Os andpt Os systems, we propose that the Pt Os isochron age records a late, low temperature hydrothermal event affecting PGM and sulphides in the Platreef and Merensky Reef. Pt-rich phases may also yield single grain 190 Pt 186 Os model ages since initial 186 Os/ 188 Os can be well constrained. A Pt Os model age of 2024 Ma±101 Ma was calculated for a Merensky cooperite; a Pt Fe alloy from the Onverwacht pipe and a Tweefontein Hill sperrylite give Pt Os model ages of 2113±106 Ma and 2042± 102 Ma, respectively. A Pt Os isochron derived from all Bushveld units gives an age of 2012±47 Ma (2σ, 186 Os/ 188 Os i = ± ) with an MSWD of 1.19, hence PGE mineralisation in different stratigraphic horizons of the complex cannot be distinguished temporally using the Pt Os chronometer. Nonetheless, the potential for the Pt Os system in dating events in PGE mineralised systems is evident and the agreement of single grain model ages in high Pt/Os phases with the Pt Os and Re Os isochron ages indicates their usefulness in the geochronology of these systems Elsevier B.V. All rights reserved. 1. Introduction The Bushveld Complex is the largest known igneous intrusion on Earth (Harney and von Gruenewaldt, 1995). It hosts 75% of the world's platinum resources, as well as the majority of the other accessible platinum-group elements (PGE) plus 16% of global nickel reserves (Naldrett, 2004). As a result of its scientific and economic significance the intrusion has been widely and intensely studied, with a large body of research focused on determining on its age. A variety of geochronometers have been applied to a range of materials from various stratigraphic intervals throughout the complex (Table 1). These studies constrain the age of the Bushveld Complex to Ga, however au Pb zircon age of around Ga from the Merensky Reef is widely accepted as the most recent and accurate age (Scoates and Friedman, 2008). Although various techniques have been used to date a wide Corresponding author at: Mineralogisch-Petrologisches Institut, Universität Bonn, Poppelsdorfer Schloss, Bonn, Germany. address: jude@uni-bonn.de (J.A. Coggon). range of lithologies, few of these, with the exception of U Pb, date mineralized horizons and none have directly dated the platinum minerals themselves. The 190 Pt 186 Os decay system, measured by laser ablation-multicollector-inductively coupled plasma mass spectrometry (LA-MC- ICPMS), has recently been demonstrated as a useful geochronometric tool for detrital PGM (platinum-group minerals), particularly PGE alloys (Lapland Greenstone Belt and Ural Mountains, Nowell et al., 2008b; southeast Borneo, Coggon et al., 2011). An advantage of this technique is that it allows simultaneous measurement of 186 Os/ 188 Os and 187 Os/ 188 Os, facilitating the exploitation of the Pt Os chronometer coupled with the Re Os isotopic tracer. In this contribution we present and discuss Pt Os ages for PGM extracted from (i) Merensky Reef ore samples from Rustenburg, (ii) a dunite pipe cross-cutting the layered series (Onverwacht), and (iii) late Bushveld hydrothermal veins cross cutting the Banded Ironstone Formation directly beneath the layered series in the Northern Limb (Tweefontein Hill). The wide variation in mineralogy offers considerable potential for Pt/Os and Re/Os fractionation and hence is appropriate for Pt Os and Re Os isochron studies. In terms of the Pt Os chronometer this potential for a wide range in parent/ /$ see front matter 2011 Elsevier B.V. All rights reserved. doi: /j.chemgeo

2 J.A. Coggon et al. / Chemical Geology (2012) Table 1 Published Bushveld ages. Stratigraphic interval Age (Ma) System Material dated Event dated Reference Reference # Bushveld Granites 2099±3 U Pb Cassiterite Bushveld granites Gulson and Jones, Rustenburg Layered Suite ±0.8 U Pb Titanite Hydrothermal activity associated Buick et al., (RLS) with final cooling of the RLS Platreef 2011 ±51 Ma Re Os Pyroxenite Platreef Reisberg et al., Merensky Reef ±1.3 U Pb Zircon Late Merensky Scoates and Friedman, Merensky Reef ±3.9 U Pb Rutile UG2 Chromitite ±3.2 Ar Ar Biotite UG2 Chromitite Nomade et al., Bastard Unit 2043±11 Re Os Poikilitic pyroxene Bastard Unit Schoenberg et al., Dullstroom and Damwal / 65 Rb Sr Whole-rock Earliest Bushveld magmatism Buchanan et al., Formations Merensky Reef 1995±50 Pt Os (isochron) Mixed PGM This study Merensky Reef 2024±101 Pt Os (single grain model age) Cooperite This study Onverwacht pipe (LG6?) 2113±106 Pt Os (single grain model age) Pt Fe nugget This study Rustenburg Layered Suite (RLS) 2012±47 Pt Os (isochron) Mixed PGM This study N.B. Reference # relates to Fig. 5. daughter isotope ratios acts to counter-balance the long half-life of the 190 Pt parent isotope (Begemann et al., 2001). 2. Samples A selection of Bushveld PGM was obtained from collections at the Federal Institute for Geosciences and Natural Resources (BGR), Hannover, and Muséum National d'histoire Naturelle (MNHN), Paris. BVD 8801 and BVD 8801-R2 are polished mounts of metallic heavy mineral concentrates from the Merensky Reef, similar to the material used by Hart and Kinloch (1989). The material was collected in the 1970s at the Frank Shaft, Rustenburg Platinum Mine (AngloPlatinum; S, E) (Fig. 1). Concentrates were produced prior to fine milling and flotation of the ore using density separation methods. Grains of laurite (RuS 2, n=23), cooperite (PtS, n=10), laurite platarsite (n=1), sperrylite (PtAs 2, n=13) and Pt Fe alloy (n=1) were selected for this study. PGM compositions (Appendices A and B) were confirmed by electron probe microanalysis following isotopic analyses with the exception of those grains that were entirely consumed during ablation. A total of 47 PGM grains with lengths ranging from approximately 150 to 500 μm were analysed; some examples are shown in Fig. 2. Sample BVD is a polycrystalline nugget of Pt Fe on chromite that measures approximately mm. It was collected from the Onverwacht Mine, Eastern Bushveld (24 40 S, E; Fig. 1) and was given to the MNHN, Paris, in 1955 by Colonel Jean Paul Louis Vésigné. The Onverwacht dunite pipe intrudes perpendicular to the typical igneous layering of the Bushveld Complex and cuts through the LG6 chromitite layer. It is possible that PGM hosted in the pipe are derived from PGE from disrupted chromitites (Zaccarini et al., 2002). BVD is an isolated, euhedral sperrylite crystal measuring ~1 mm across. The crystal was liberated from a hand specimen of vein material measuring approximately cm. The host rock is composed predominantly of iron oxide, with limonite and malachite. Museum records show that it was collected at Tweefontein Hill, on the Northern Limb of the Bushveld (24 00 S, E; Fig. 1). Large sperrylite crystals occur at this locality and are associated with hydrothermal veins in breccia zones within banded ironstone close to the Platreef (Spencer, 1926; Nex, 2005). Visual identification of samples BVD and was confirmed by energy dispersive X-ray spectrometer (EDS) spectra using a Low-Vacuum, T Scan Scanning Electron Microscope at the MNHN. BVD 8801 and 8801-R2 were provided as polished blocks therefore no further preparation was needed for these grains. BVD and were mounted on adhesive carbon SEM tabs on a glass slide. 3. Methods 3.1. Electron Microprobe Compositions of Merensky PGM were analyzed at the Federal Institute for Geosciences and Natural Resources (BGR) using a CAMECA SX100 EMP with an attached energy-dispersion analytical system, via 26 E 27 E 28 E 29 E 30 E N Northern limb TWEEFONTEIN HILL Mokopane 24 S Eastern & Western limbs Upper Zone Northern limb Western limb RUSTENBURG 8801 & 8801-R2 ONVERWACHT Steelpoort Bela Bela Pretoria Eastern limb 25 S 8801 & 8801-R (?) MR UG 2 UG 1 MG LG Main Zone Critical Zone Platreef km 26 S Johannesburg Lower Zone Fig. 1. Extent of exposure of the Rustenburg Layered Suite of the Bushveld Complex (after Eales and Cawthorn, 1996) and simplified stratigraphic column (after Cawthorn et al., 2002). Black stars show sample locations and relative stratigraphic positions. LG, Lower Group; MG, Middle Group; UG, Upper Group; MR, Merensky Reef.

3 50 J.A. Coggon et al. / Chemical Geology (2012) a b c 200 um 200 um 200 um Fig. 2. BSE images of some Merensky laurites (a: 8801_9, platarsite; b: 8801_13, laurite; and c: 8801_15, laurite) following laser ablation and re-polishing. The grain centres appear unzoned, however concentric magmatic zoning is seen within ~50 μm of the rims of some grains. Laser spots were positioned to analyse the centres of grains. the method described by Oberthür et al. (2004). An accelerating voltage of 20 kv and specimen current of 30 na were applied and counting times were 10 and 20 s on background and peak positions, respectively. Details of detection limits and X-ray lines and standards used are given in Appendix C. The PAP programme supplied by CAMECA was used to correct the raw data. Concentration determinations of Rh, Pd, Ag, Cu, As and Sb were enhanced by secondary lines; further corrections were applied in the cases of these elements Mass spectrometry Isotopic analyses were performed at Durham University Northern Centre for Isotopic and Elemental Tracing (NCIET) using a New Wave UP 213 nm laser and Thermo Fisher Neptune multi-collector ICPMS (MC-ICPMS) via the method described in detail by Nowell et al. (2008b). Bushveld PGM were analysed during six sessions between July 2009 and March All analyses consisted of one block of 40 cycles, with a cycle integration time of four seconds for standards and one second for samples. The measurement routine of Nowell et al. (2008b) for standard solutions mimicked that for laser ablation analyses, with 1 s integrations. The integration time has been increased to 4 s (total measurement time of 160 s) for standard solutions to more accurately measure 190 Os/ 188 Os, as this ratio is used to correct for Os interference on 190 Pt. The measurement routine for samples remains necessarily short to accommodate commonly small sample size. Laser spot size was selected based on the size and Os content of each grain. The ablation was started a few seconds prior to measurement. During each spot ablation the laser power was tailored to the Os content of each grain to optimize signal size and, where possible, maintain a steady beam intensity. The ranges of power and spot sizes used, along with fixed value laser parameters, are given in Table 2. A 2 sigma rejection was applied to all analyses of standards and samples with the exception of BVD 8801_19. In this case the grain exhibits extreme internal heterogeneity and appears to define two distinct domains. For further details see Section 4.2. Table 2 UP213 operating conditions for laser ablation Os isotopic analysis. Parameter Value Wavelength 213 nm Ablation cell Std New Wave cell Carrier gas Ar Carrier gas flow rate 1.5 l min 1 Wash out time >60 s Spot diameter μm depending on Os content of sample Crater depths ~40 80 μm Repetition rate 20 Hz Laser power % depending on Os content of sample Laser power density ~4.3 7Jcm DROsS standard solution At the start of each analytical session the Neptune was tuned using a 1 μgml 1 Durham Romil Osmium Standard (DROsS) standard solution to achieve maximum sensitivity and optimum peak shape. Baseline and gain calibrations were then carried out, followed by up to 15 analyses of a DROsS standard solution (Appendix D). Mean 187 Os/ 188 Os and 186 Os/ 188 Os values of ± and ± , respectively (2 SD, n=72), are identical within uncertainty to the values of ± and ± (2 SD, n=5) reported for this standard by Nowell et al. (2008b). Reproducibility over the six sessions for a total of 72 analyses of DROsS solutions was 120 ppm for 187 Os/ 188 Os and 82 ppm for 186 Os/ 188 Os Mass bias, interfering element corrections and error propagation When preparing samples for N-TIMS or solution-mode MC-ICPMS analyses non-molecular isobaric interferences can be eliminated or minimised via column chemistry and purification of the analyte. During laser ablation the whole sample matrix is introduced into the mass spectrometer, resulting in a more complex mass spectrum and a greater potential for non-molecular and molecular isobaric interferences on the analyte element. Despite not being able to resolve molecular interferences within the analytical resolution of the present LA-MC-ICPMS technique we see no clear evidence (such as sporadic data points) for molecular isobaric interferences, although this does not exclude their presence. Non-molecular elemental isobaric interferences are more problematic during laser ablation analysis of PGM and must be corrected for algebraically; this is accomplished via measurement of interference-free monitor isotopes. The Pt-rich nature of the minerals analysed here precludes the use of 192 Os or 190 Os for mass bias correction as these isotopes are subject to direct interference from 192 Pt and 190 Pt, respectively. Therefore we employ the interference-free 189 Os/ 188 Os ratio for both solution and ablation analyses for this paper. A 189 Os/ 188 Os value of was assumed for this correction (Shirey and Walker, 1998). All processing of raw data was performed offline, standard and sample data were treated identically. 184 W, 186 Wand 187 Re are elemental isobaric interferences on 184 Os, 186 Os and 187 Os during laser ablation. To correct for these interferences we monitored 182 Wand 185 Re during analyses. Analyses of a 1 μgml 1 DROsS solution doped with varying concentrations of W and Re were used to derive values for the stable 182 W/ 184 W, 182 W/ 186 Wand 185 Re/ 187 Re (see Appendix E), which were in turn used to calculate the interference for each 1 s integration. Interferences were subtracted to give the corrected Os isotope ratios. Interferences on 190 Os and 192 Os by 190 Pt and 192 Pt cannot be corrected in the same way as Re and W interferences as the Faraday cup configuration used does not provide a Pt monitor isotope. Instead Os must be treated as the interfering element, with 188 Os taken as the monitor isotope. Mean 190 Os/ 188 Os was determined from measurements of the DROsS standard for each session. These values were then used to correct for Os overlap on mass 190, allowing 190 Pt intensities to be obtained.

4 Table 3 Os isotope compositions of Bushveld PGM grains analyzed by laser ablation-multi-collector-icpms at Durham University. Mineralogy 188 Os (V) ±2SE 187 Os/ 188 Os ±2SE 186 Os/ 188 Os Total abs. error 184 Os/ 188 Os ±2SE 187 Re/ 188 Os ±2SE 185 Re/ 188 Os ±2SE 182 W/ 188 Os ±2SE 190 Pt/ 188 Os Total abs. error Initial 187 Os/ 188 Os Grain 8801_1 Laurite _2 Laurite _3 Laurite _4 Laurite _5 Laurite _6 Laurite _7 Laurite _8 Laurite _9 Platarsite _11 Laurite/ platarsite 8801_12 Laurite _13 Laurite _14 Laurite _15 Laurite _16 Laurite _17 Laurite _18 Sperrylite _19 Sperrylite _19a* Sperrylite _19b* Sperrylite _20 Cooperite _22* Cooperite _24* Sperrylite? _25 Sperrylite _26 Sperrylite? _27 Sperrylite R2_1* Laurite? R2_2 Pt Fe alloy? n.r R2_3 Laurite? R2_4 Cooperite? n.r R2_5* Cooperite n.r R2_6 Cooperite n.r R2_7 Laurite R2_8* Laurite? R2_9 Sperrylite R2_10 Sperrylite n.r R2_11 Cooperite n.r R2_12 Cooperite n.r R2_13 Sperrylite R2_14 Cooperite R2_15 Laurite R2_16 Laurite? R2_17* Cooperite n.r R2_19 Sperrylite? n.r R2_20 Laurite R2_21 Cooperite? n.r R2_23 Laurite? R2_24 Laurite R2_25* Laurite? Sperrylite n.r _ mean Pt Fe alloy (continued on next page) J.A. Coggon et al. / Chemical Geology (2012)

5 Table 3 (continued) Mineralogy 188 Os (V) ±2SE 187 Os/ 188 Os ±2SE 186 Os/ 188 Os Total abs. error 184 Os/ 188 Os ±2SE 187 Re/ 188 Os ±2SE 185 Re/ 188 Os ±2SE 182 W/ 188 Os ±2SE 190 Pt/ 188 Os Total abs. error Grain/spot _01 Pt Fe alloy _02 Pt Fe alloy _03 Pt Fe alloy _04 Pt Fe alloy _05 Pt Fe alloy _06 Pt Fe alloy _07 Pt Fe alloy n.r _08 Pt Fe alloy n.r _09 Pt Fe alloy _10 Pt Fe alloy _11 Pt Fe alloy _12 Pt Fe alloy _13 Pt Fe alloy _14 Pt Fe alloy _15 Pt Fe alloy _16 Pt Fe alloy _17 Pt Fe alloy _18 Pt Fe alloy _19 Pt Fe alloy _20 Pt Fe alloy _21 Pt Fe alloy _22 Pt Fe alloy _23 Pt Fe alloy _24 Pt Fe alloy _25 Pt Fe alloy _26 Pt Fe alloy _27 Pt Fe alloy _28 Pt Fe alloy _29 Pt Fe alloy _30 Pt Fe alloy _31 Pt Fe alloy n.r _32 Pt Fe alloy _33 Pt Fe alloy _34 Pt Fe alloy _35 Pt Fe alloy Initial 187 Os/ 188 Os 52 J.A. Coggon et al. / Chemical Geology (2012) 48 60

6 Table 3 (continued) Mineralogy 188 Os (V) ±2SE 187 Os/ 188 Os ±2SE 186 Os/ 188 Os Total abs. error 184 Os/ 188 Os ±2SE 187 Re/ 188 Os ±2SE 185 Re/ 188 Os ±2SE 182 W/ 188 Os ±2SE 190 Pt/ 188 Os Total abs. error Initial 187 Os/ 188 Os _36 Pt Fe alloy _37 Pt Fe alloy n.r _38 Pt Fe alloy n.r _39 Pt Fe alloy _40 Pt Fe alloy _41 Pt Fe alloy _42 Pt Fe alloy n.r _43 Pt Fe alloy _44 Pt Fe alloy _45 Pt Fe alloy _46 Pt Fe alloy _47 Pt Fe alloy n.r _48 Pt Fe alloy _49 Pt Fe alloy _50 Pt Fe alloy _51 Pt Fe alloy _52 Pt Fe alloy _53 Pt Fe alloy _54 Pt Fe alloy _55 Pt Fe alloy _56 Pt Fe alloy _57 Pt Fe alloy _58 Pt Fe alloy _59 Pt Fe alloy _60 Pt Fe alloy _61 Pt Fe alloy _62 Pt Fe alloy _63 Pt Fe alloy _64 Pt Fe alloy _65 Pt Fe alloy _66 Pt Fe alloy Within run uncertainties are quoted as 2SE. Total absolute error for 186 Os/ 188 Os ratios incorporates an estimate of external reproducibility based on repeat analyses, over a period of ~one year, of an in house LA standard (Urals Os-rich PGE alloy G1). Merensky (8801 and 8801-R2) sample compositions were confirmed by EMP (see Appendix B) following isotopic analyses. Where insufficient material remained for EMP analysis (shown as? ) identification is based on previous EDS mapping. Onverwacht and Tweefontein Hill sample compositions were confirmed by EDS. Initial 187 Os/ 188 Os was calculated assuming an age of 1995 Ma and a decay constant of for 187 Re (Smoliar et al., 1996). Negative ratios occur where an overcorrection has been made for an interference on the numerator isotope. The necessarily short measurement times for laser ablation of these samples result in higher levels of noise in the background signals measured for samples than standards; hence it is sometimes possible that very slight overcorrection will occur on very low abundance isotopes. Negative values in this table have been highlighted in grey text and should be interpreted as being equal to zero. n.r. = not reported. Samples with very low 188 Os beam intensities (i.e. very low common Os contents, in other words high Pt/Os) yield unreliable 187 Os/ 188 Os values due to an unexplained overcorrection on mass 187. Analyses with mean 188 Os beam intensities lower than 0.02 V are not reported here. Those with mean 188 Os beams of V are given in italics. Sample 8801-R2_2 is also strongly influenced by this apparent overcorrection (since 35 of the 40 integrations that make up this analysis were measured at 188 Os beams of ~0.01 V), the 187 Os/ 188 Os value for this sample is therefore not reported. Please see Appendix F, section F3 for further discussion of these data. J.A. Coggon et al. / Chemical Geology (2012)

7 54 J.A. Coggon et al. / Chemical Geology (2012) It was observed in this study that samples with low common Os yielded unreliable 187 Os/ 188 Os data. Analyses with mean 188 Os beam intensities of less than 0.02 V were subject to large, systematic errors on mass 187 leading to unrealistically low 187 Os/ 188 Os values, thus these ratios are not reported. In addition, modelling shows that the mass bias and Re-corrected 187 Os/ 188 Os ratios of samples with mean 188 Os beam intensities between 0.02 and 0.1 V are likely to be underestimated by more than 1%. Sample 8801-R2_2 is also affected by this apparent overcorrection on mass 187; the 187 Os/ 188 Os ratio for this sample is not reported. It should be noted that there is no evidence for systematic errors on any of the other masses. For more details and further discussion please see Appendix F, section F3. If W corrections are systematically inaccurate then any correlation between W/Os and Pt/Os could cause rotation of a Pt Os isochron, as discussed by Nowell et al. (2008b). The Merensky data do show a general trend towards higher W/Os ratios in Pt-rich minerals, but no linear relationship is seen between W/Os and Pt/Os. Therefore any inaccuracies in the interfering element corrections will not cause systematic rotation of the isochrons constructed using these data. The W correction is small for the majority of samples and the magnitude of the Os correction decreases as Pt/Os ratio increases, thus the correction is smaller for the samples that have the greatest influence on the Pt Os isochron age. Total errors on 186 Os/ 188 Os ratios were calculated to incorporate external reproducibility. A value of 176 ppm was used for grains with 188 Os beams of 1 V or more; grains with 188 Os beams b1v were assigned external reproducibility values of 352 ppm. These values were derived from repeat analyses over ~1 year of an in-house standard (Urals Os-rich PGE alloy G1, Nowell et al., 2008b). Similar data are not available for 190 Pt/ 188 Os as there is no homogeneous Pt-rich ablation standard (see Appendix F, section F4 for further discussion). Total uncertainties on the 190 Pt/ 188 Os ratio include the within-run error plus a conservative estimate of 5% uncertainty to account for external reproducibility on the measurement of this ratio and elemental fractionation that is likely to occur at the ablation site (Nowell et al., 2008b). 4. Results 4.1. Re Os isotopes Analyses of Merensky PGM (see Table 3) yielded measured 187 Os/ 188 Os ratios of ± to ± , and 187 Re/ 188 Os ratios of b to ± Os/ 188 Os ranging from ± to ± (mean= ) and 187 Re/ 188 Os values from ± to ± (mean = ) were measured in the heterogeneous Onverwacht Pt Fe nugget (Table 3, Appendix G). The Tweefontein Hill sperrylite contains extremely low concentrations of common osmium, with measured 187 Os/ 188 Os and 187 Re/ 188 Os ratios such that 187 Os/ 188 Os cannot be accurately corrected. No isochronous relationships exist within this Re Os dataset. Whole rock samples of the Bushveld Complex (poikilitic pyroxenite from the Merensky Reef at the Amandebult section of the western limb; chromites and pyroxenites of the Rustenburg Layered Series (RLS) from various locations across the western limb) display measured 187 Os/ 188 Os ranging to much higher values ( to ) than the PGM presented in this study (Schoenberg et al., 1999). This difference most likely results from high concentrations of Re in whole rock samples relative to negligible Re measured in PGM of this study. Initial 187 Os/ 188 Os ( 187 Os/ 188 Os i ) calculated for all PGM analysed in this study, using the Pt Os isochron age of 1995 Ma, fall within the published range ( to ) for a variety of silicate mineral, PGE sulphide and whole-rock studies (Hart and Kinloch, 1989; McCandless et al., 1999; Schoenberg et al., 1999; Reisberg et al., 2011) (Fig. 3). Initial 187 Os/ 188 Os values of BVD 8801 and 8801-R2 Merensky PGM analysed in this study are consistent with those presented by Hart and Kinloch (1989), for PGM taken from the same original Frank metallics sample. The published 187 Os/ 188 Os i value of exhibited by a Merensky poikilitic pyroxenite (Schoenberg et al., 1999) is also consistent with the Merensky data presented here (Fig. 4). No published 187 Os/ 188 Os i data from Onverwacht were available for comparison, however the Onverwacht data of this study are similar to the Merensky BVD Os/ 188 Os i values Pt Os isotopes and age The 47 Merensky PGM grains of polished mounts BVD 8801 and BVD 8801-R2 display a wide range of 186 Os/ 188 Os ratios from to and 190 Pt/ 188 Os values from b to The Onverwacht Pt Fe nugget BVD was analysed 66 times and yielded 186 Os/ 188 Os ratios ranging from to (mean= ) and 190 Pt/ 188 Os ratios of (mean=5.95). The Tweefontein sperrylite BVD yielded a UG2 chromitites MG chromitites a LG chromitites a Merensky Reef poikilitic pyroxenite a b Platreef pyroxenites c Eastern Bushveld Western Bushveld c d Frank Metallics Other Bushveld laurites d Chondrite 59 Onverwacht { Pt-Fe 39 Merensky { 8801 mixed PGM 8801-R2 mixed PGM Os/ Os i a Schoenberg et al. (1999) b Reisberg et al. (2011) c McCandless et al. (1999) d Hart and Kinloch (1989) Fig. 3. Variation in 187 Os/ 188 Os i values of Rustenburg (BVD 8801 and 8801-R2) and Onverwacht (BVD ) PGM compared to published data for Bushveld rocks and minerals.

8 J.A. Coggon et al. / Chemical Geology (2012) a Os/ Os 0 60 b Os/ Os 1995 ± 50 Ma Os/ Os i = ± MSWD = 1.16 n = ± 47 Ma Os/ Os i = ± MSWD = 1.19 n = A Pt/ Os Pt/ Os 2 σ errors Os/ 188 Os and 190 Pt/ 188 Os ratios of and 184 respectively, (Table 3, Appendix G) Merensky Reef Each of the 24 PGM grains from polished mount BVD 8801 is shown on the isochrons in Fig. 4a and b as a single point representing a 40 second analysis, with three exceptions. An initial ablation showed the sperrylite grain 8801_19 to be isotopically heterogeneous, with two distinct domains. It was analysed in two parts (8801_19a and _19b). Cooperite grain 8801_22 and sperrylite grain 8801_24 were also found to be (isotopically) two-component grains. However, in both cases the more Pt-rich components contained negligible common Os. Consequently, accurate measurement of the 189 Os/ 188 Os ratio for the Pt-rich components was impossible, thus mass bias corrections are unreliable and the measurements of Pt-rich components were rejected. Grains 8801-R2_1, _5, _8, _17 and _25.were small; there was insufficient material to sustain an entire 40 s ablation for these samples, therefore some integrations were rejected. For further details of manual data rejection see Appendix F. Merensky PGM yield a Pt Os isochron age of 1995±50 Ma (2σ; n=48; MSWD=1.16; probability of fit=0.21) with an initial 186 Os/ 188 Os ( 186 Os/ 188 Os i ) ratio of ± (Fig. 4a). The uncertainty on the age incorporates the 1% uncertainty on the decay constant A BVD (Onverwacht) 2 σ errors BVD (Tweefontein Hill) n=40 (including 24 laurites) Fig. 4. Pt Os isochron diagrams: a) Merensky Reef PGM; b) All Bushveld data. Fill colour of error ellipse indicates mineralogy: black, laurite (all plot at 190 Pt/ 188 Os values of ~0.0); dark grey, cooperite; light grey, sperrylite; white, Pt Fe alloy. Errors on 186 Os/ 188 Os incorporate within run uncertainties and long term external reproducibility on this ratio based on repeat analyses of an in-house standard grain (Urals Os-rich PGE alloy G1, Nowell et al., 2008b). 190 Pt/ 188 Os errors include 5% uncertainty to account for elemental (Pt/Os) fractionation that occurs at the ablation site and external reproducibility on the measurement of this ratio (Nowell et al., 2008b). The precision quoted for the isochron age incorporates an uncertainty of 1% on the decay constant of 190 Pt (Begemann et al., 2001). Plotted using Isoplot version 3.1 (Ludwig, 2003). of 190 Pt as estimated by Begemann et al. (2001). The initial 186 Os/ 188 Os ratio is close to estimates for primitive upper mantle (PUM) at 2.05 Ga ( ± , calculated assuming a present day PUM Os isotope composition of ± , and Pt/Os ratio of 2.0±0.2, Brandon et al., 2006). The slightly lower value obtained by LA-MC-ICPMS is consistent with previous MC-ICPMS studies which show systematically lower 186 Os/ 188 Os values for standards when determined by this method compared to ID N-TIMS (Luguet et al., 2008a; Nowell et al., 2008a, 2008b). The most Pt-rich Merensky cooperite grain (8801-R2_11) yields a Pt Os model age of 2024 Ma (a conservative estimate of a 5%, i.e. ±101 Ma, uncertainty is assumed, to account for uncertainty on the measurement of the 190 Pt/ 188 Os ratio). The Pt Os isochron age is 55 Myr younger than precise U Pb zircon age reported for the Merensky Reef (Scoates and Friedman, 2008). However, both Pt Os Merensky PGM ages are within error of the Re Os isochron age for Platreef pyroxenites of 2011±51 Ma presented by Reisberg et al. (2011), which is also considerably younger than the U Pb zircon age Onverwacht pipe BVD is a polycrystalline Pt Fe nugget that exhibits internal heterogeneity. A total of 66 individual spot ablations were performed on this sample; the measured range of Pt/Os ratios is narrow, hence no reliable single-grain isochron was produced for this sample. A Pt Os model age of 2113±106 Ma (assuming an estimated 5% uncertainty, see Section 4.2.1) was calculated for the mean of the 66 analyses, assuming an initial 186 Os/ 188 Os value ( ) of modern day primitive upper mantle (PUM). This is within error of the published ages (Fig. 5, Table 1). An identical age was achieved when the initial 186 Os/ 188 Os value was assumed to be the same as PUM at 2054 Ma ( ), hence the model age is insensitive to the assumed common Os and the composition of the grain is dominated by radiogenic 186 Os Tweefontein Hill BVD is a heterogeneous crystal of sperrylite. Four analyses were attempted on this sample, however only one ablation sampled a region of the crystal with sufficient Os to provide an adequate Os signal for reliable mass bias and interfering element corrections. A Pt Os model age for the sperrylite crystal can be calculated using an initial 186 Os/ 188 Os value of PUM at 2054 Ma. This gives an estimated age for the Tweefontein Hill hydrothermal PGE mineralization of 2042± 102 Ma (assuming an estimated 5% uncertainty, see Section 4.2.1) Combined Bushveld Pt Os isochron age When combined, the entire dataset defines a Pt Os isochron age of 2012±47 Ma (2σ, n=50) with an MSWD of 1.19 and probability of fit of 0.18, and 186 Os/ 188 Os i value of ± (Fig. 4b). The uncertainty on the age incorporates the 1% uncertainty on the decay constant of 190 Pt as estimated by Begemann et al. (2001). This age is within analytical uncertainty of recent high-precision U Pb zircon dating for the Merensky Reef (Scoates and Friedman, 2008) and other published ages for Bushveld samples (Fig. 5, Table 1). The Pt Os isochron was regressed using Merensky, Onverwacht (pipe) and Tweefontein Hill (hydrothermal mineralization) data. The isochron shown in Fig. 4b presents the combined Bushveld dataset in the most consistent way possible, with each point on the isochron representing asinglegrain.initial 186 Os/ 188 Os is identical within uncertainty to the value calculated for the Merensky isochron (Fig. 4). 5. Discussion 5.1. Source variation versus crustal contamination effects on the Re Os and Pt Os systems Bushveld rocks and minerals display a wide range of 187 Os/ 188 Os i values ( to , Fig. 3) (Hart and Kinloch, 1989; McCandless