Magmatic PGE Sulphide Mineralization in Clinopyroxenite from the Platreef, Bushveld Complex, South Africa

: The Platreef, at the base of the northern limb of the Bushveld Complex in South Africa, hosts platinum-group element (PGE) mineralization in association with base-metal sulphides (BMS) and platinum-group minerals (PGM). However, whilst a magmatic origin of the stratiform mineralization of the upper Platreef has been widely conﬁrmed, the processes responsible for the PGE and BMS mineralization and metasomatism of the host rocks in the Platreef are still under discussion. In order to contribute to the present discussion, we present an integrated petrographical, mineral-chemical, whole-rock trace- and major-element, sulphur- and neodymium-isotope, study of Platreef footwall clinopyroxenite drill core samples from Overysel, which is located in the northern sector of the northern Bushveld limb. A metasomatic transformation of magmatic pyroxenite units to non-magmatic clinopyroxenite is in accordance with the petrography and whole-rock chemical analysis. The whole-rock data display lower SiO 2 , FeO, Na 2 O and Cr ( < 1700 ppm), and higher CaO, concentrations in the here-studied footwall Platreef clinopyroxenite samples than primary magmatic Platreef pyroxenite and norite. The presence of capped globular sulphides in some samples, which display di ﬀ erentiation into pyrrhotite and pentlandite in the lower, and chalcopyrite in the upper part, is attributed to the fractional crystallization of a sulphide liquid, and a downward transport of the blebs. In situ sulphur (V-CDT) isotope BMS data show isotopic signatures ( δ 34 S = 0.9 to 3.1 % (cid:24) ; ∆ 33 S = 0.09 to 0.32% (cid:24) ) close to or within the pristine magmatic range. Elevated (non-zero) ∆ 33 S values are common for Bushveld magmas, indicating contamination by older, presumably crustal sulphur in an early stage chamber, whereas magmatic δ 34 S values suggest the absence of local crustal contamination during emplacement. This is in accordance with the ε Nd (2.06 Ga) (chondritic the IPGE (Os, Ir, Ru) were interpreted to mainly occur in discrete PGM. However, the presence of pentlandite with variable PGE concentrations on the thin section scale may be related to variations in the S content, already at S-saturation during magmatic formation, and / or post-solidiﬁcation mobilization and redistribution.

. The borehole location (OY541) is indicated by the red diamond.
Detailed descriptions of the Platreef lithologies are given in Armitage, McDonald, Edwards and Manby [15]; Harris and Chaumba [3]; Kinnaird, Hutchinson, Schurmann, Nex and de Lange [22]; McDonald, Holwell and Armitage [62]; Holwell, McDonald and Armitage [20]; Kinnaird, Yudovskaya, McCreesh, Huthmann and Botha [58]; McDonald, Harmer, Holwell, Hughes and Boyce [54] and Grobler, Brits, Maier and Crossingham [44]. The footwall lithologies beneath the primary igneous Platreef at Overysel comprise Archean gneiss, banded granulite (local term: granofelses), footwall clinopyroxenite (historically called parapyroxenite), calcsilicate rocks, dolomite and serpentinite. The Platreef contains >4 g/t PGE where the footwall is made of Malmani Dolomite, and 1 to 2 g/t elsewhere, with sporadically elevated values up to 10 g/t [5]. The clinopyroxenites are described to occur in the footwall of the Platreef below the igneous pyroxenites and above the Malmani Dolomite, and comprise diopside and bulk-rock Cr contents of <2000 ppm, while the igneous pyroxenite mainly consists of orthopyroxene and cumulus Ca-poorer clinopyroxene, and contains bulk-rock Cr contents of > 2000 ppm, according to the definitions of Harris and Chaumba [3], Armitage, McDonald, Edwards and Manby [15], and Holwell, McDonald and Armitage [20]. The actual calcsilicate rocks or xenoliths within the Platreef, however, are described as a mixture of dolomite and igneous reef pyroxenites, and have lower SiO 2 Minerals 2020, 10, 570 5 of 28 (<40 wt %), lower FeO/MgO, and (much) higher CaO/MgO than the non-igneous clinopyroxenites [3]. Most of these lithologies underwent, or were the product of, thermal metamorphism and hydrothermal alteration, as well as sediment assimilation, which was thought to have led to a significant change in the bulk-rock chemistry and thus mineralogy of the footwall rock types. However, the exact cause of the post-magmatic processes, and the impact on the primary igneous mineralization, is still a matter of discussion [3,5,15,18,19,49]. In order to understand the nature of these processes and their influence on the sulphide-associated PGE mineralization of the clinopyroxenites, the present study is focused on characterizing the mineralization within drill core clinopyroxenite samples from Overysel.

Samples and Analytical Methods
A total of 54 samples for this study were collected from two drill cores of OY541 at Overysel 815LR (Anglo American Platinum), with corresponding coordinates referring to the WGS84 (South African Grid) Lo.29 grid (X: 50187.18; Y: 13091.78). Drill core OY541 was chosen for the present study, because it covers a typical section of clinopyroxenite between the magmatic Platreef and Malmani Dolomite ( Figure 1) [6]. Preferably high-grade samples were chosen for the present study.
A total of 16 representative clinopyroxenite drill core samples were analyzed for bulk-rock Pt, Pd, Au and S, as well as major and trace-element contents ( Table 1). All analyses were done by ACTLABS (Canada), using fusion coupled ICP-MS and combined fire assay ICP-MS methods. Major elements were determined with analytical uncertainties of less than 5%. Trace elements were analyzed with analytical uncertainties of mostly less than 10%.
Scanning electron microscope (SEM) investigations, which were primarily used for the detection and semi-quantitative analyses of PGM, were undertaken at the Bundesanstalt für Geowissenschaften und Rohstoffe (BGR) in Hannover, Germany, using a MLA 650F SEM (FEI Company, Hillsboro, OR, USA).
Quantitative mineral chemical analyses of sulphides and pyroxenes (Tables 2 and 3) were carried out on three polished thin sections and five polished ore sections, using a JEOL JXN-8200 electron microprobe (EMP) at the GeoZentrum Nordbayern, University of Erlangen. The JEOL JXN-8200 was operated with an acceleration voltage of 20 kV, a 3-µm beam diameter and a probe current of 20 nA (for further analytical details see [13]).
Neodymium isotope (CHUR) measurements and Sm and Nd concentrations were determined in Class 1000 metal-free laboratories at the GeoZentrum Nordbayern. For Nd isotope measurements (Table 4), approximately 100 mg of sample powder was digested in 0.5 mL 15 M HNO 3 and 2 mL 12 M HF in sealed Teflon beakers overnight at 80 • C. A quantity of 0.2 mL of HClO 4 was then added to the sample solution, which was evaporated to near dryness at 130 • C. The sample was taken up in 2 mL 15 M HNO 3 , and evaporated to complete dryness at 150 • C. This step was repeated 3 times, until the sample was completely in solution, before redissolving the sample in 2 mL 3.5 M HNO 3 . The rare-earth elements were separated from the rock matrix using 0.2 mL of Eichrom TRU-spec resin. The sample in 3.5 M HNO 3 was loaded onto the resin, which was washed 3 times with 2 mL 3.5 M HNO 3 to remove major elements, and the rare earth elements (REE) were selectively eluted with 2 mL 2.5 M HCl. This solution was evaporated, the residue redissolved in 0.5 mL 0.25 M HCl, and Nd was then separated from the other REE using 1.5 mL Eichron LN-spec resin in 0.25 M HCl. All reagents used were Teflon distilled, and the Nd blank was below 20 pg.
Nd isotope measurements were carried out using a Thermo Triton thermal ionization mass spectrometer in static mode. Nd was loaded in 3 µL 0.1 M H 3 PO 4 onto the Ta filament of a double Ta-Re filament assembly and analyzed as the metal. Interference of 144 Sm on 144 Nd was corrected for by measuring 147 Sm, but was negligible for all samples. Instrumental mass fractionation was corrected for, assuming a 144 Nd/ 146 Nd ratio of 0.7219. An in-house Nd standard yielded 143 Nd/ 144 Nd = 0.511541 ± 0.000008 (n = 6), equivalent to 0.511852 for the La Jolla Nd standard.  ppm  2710  230  750  1210  2580  2170  60  230  1990  BDL  290  60  4260  240  3780  30  Cu  ppm  1260  167  136  1050  1330  1910  9  35  1450  BDL  3  96  2920  BDL  2560  882  Mn  ppm  1020  1280  1380  1280  1790  1380  1850  1860  1490  972  1850  1500  3710  4070  2290  2190  Pd  ppb  1730  1  617  1550  2310  3640  2  32  2480  1  51  3  2890  4  5940  4090  Pt  ppb  1360  BDL  438  1120  1550  10200  4  17  1550  BDL  14  3  1440  3  3750  2370  Au  ppb  302  BDL  32  100  221  600  3  12  214  7  9  BDL  9  3  496  117  Ba  ppm  BDL  BDL  BDL  BDL  4  BDL  BDL  BDL  BDL  BDL  9    Sm and Nd concentrations were measured on a separate dissolution of rock powder using a Thermo Scientific X-Series 2 quadrupole inductively coupled plasma mass spectrometer. Approximately 0.05 g of sample was accurately weighed into a Teflon beaker, and digested in 1 mL 15 M HNO 3 and 3 mL 12 M HF for 12 h in sealed beakers on a hotplate at 80 • C. After cooling, 0.2 mL of HClO 4 was added to the sample, and the solution evaporated to incipient dryness at 120 • C. 2 mL of 15 M HNO 3 was added to the sample, and evaporated to near dryness, and this step was repeated twice before increasing the hotplate temperature to 150 • C and fuming off excess HClO 4 . The sample was then redissolved in 4 mL 15 M HNO 3 and 4 mL H 2 O, 2 drops of 12 M HF were added, and the sealed beakers left on a hotplate at 80 • C for 12 h. The samples were then placed in an ultrasonic bath for 30 min, before heating at 80 • C for another 12 h. At this stage, all samples were completely in solution. The sample solutions were then quantitatively transferred to 250 mL HDPE bottles and diluted to 200 g with MQ water to obtain a final solution of 2% HNO 3 + 0.002 M HF with a sample dilution factor of about 4000 and total dissolved solids of 250 µg/mL. All reagents used were distilled in Teflon stills, and diluted with MQ 18.2 MΩ water. Sample solutions were introduced into the ICP-MS through a Cetac Aridus 2 desolvating nebulizer system in order to reduce molecular interferences. An ESI SC-2 DX FAST autosampler was used to reduce washout times between samples. The instrument was tuned using a 5-ppb solution of Be, In and U; the typical sensitivity for 238 U was 2 × 106 counts per second for a sample uptake rate of 50 µL/min. The Ce/CeO ratio was typically > 4500. The instrument was calibrated using multielement solutions covering the relevant concentration range. A mixed Be, Rh, In and Bi solution (30, 10, 10 and 5 ppb) was mixed with the sample online and these elements used as internal standards to correct for instrumental drift. Procedural blanks analyzed during this work were negligible for all elements measured. The uncertainty for Sm and Nd concentrations is typically 2-5%, and the uncertainty for the Sm/Nd ratio approximately 1% based on repeated analysis of standards. The initial Nd isotope ratios in Table 4 were calculated assuming an age of 2.061 Ga, using the 147 Sm/ 144 Nd ratios for individual samples determined by quadrupole ICP-MS. Errors on the initial Nd isotope ratios in Table 4 are calculated by propagating errors on both 143 Nd/ 144 Nd and 147 Sm/ 144 Nd through the age correction. For more analytical details see [63].  59 Co, 60 Ni, 61 Ni, and 63 Cu in BMS were determined by in situ spot analyses on sulphides in three polished thin sections and five polished ore sections. He and Ar functioned as carrier gas with a flux of 0.9 l/min. The plasma power was 1100 W and Ar was used as plasma gas with a flux of 14.9 l/min. Measurements were undertaken with a repetition rate of 19 Hz for 25 s after a background scanning for 20 s. The laser energy settings are configured at 0.45 GW/cm 2 (fluence: 2.24 J/cm 2 ). Beam diameters were varied between 15 µm and 50 µm depending on grain size. External standard materials include PO724 B2 SRM (Sulphide-standard Memorial University Newfoundland), (Fe,Ni)1-xS [64] and Mass1 (USGS) for BMS. S was analyzed by EMP and was used as internal standard for sulphide analysis. Reproducibility for SRM was <9% and the accuracy, tested by ablating the PGE SRM FeNiS standard, was <15%. The data were processed using GLITTER 4.4.4 software (Macquarie Research Ltd., Sydney, Australia, 2000). Detection limits for the analyses of sulphide minerals are given in Table 5 68 Zn 40 Ar. 108 Pd is additionally interfered by 108 Cd. Due to the extremely low relative isotope abundance of 36 Ar (0.337%) and 38 Ar (0.063%), their effect on 99 Ru, 101 Ru and 103 Rh is negligible. Only a few counts per second for Zn were recorded for the sulphides in this study, thus excluding a significant influence. However, a remarkable impact occurs for CuAr on 103 Rh and 105 Pd when analyzing chalcopyrite. The argide-unaffected 108 Pd for chalcopyrite was used, which nonetheless required a correction for the elemental interference from 108 Cd. A PGE-free hydrothermal chalcopyrite was analyzed to check the 63 Cu 40 Ar and the 65 Cu 40 Ar production. Since the interference of 61 Ni 40 Ar lead to an overestimation of 101 Ru in pentlandite, 99 Ru was used instead to obtain the correct concentration. Numerous measurements of pentlandite yielded 0.5 ppm as the lowest concentration for Ru, indicating that this value solely originated from the Ni-argide. This concentration was always within the 1σ error margin for every analysis.
In situ quadruple sulphur isotope ratios (V-CDT) were measured using a CAMECA IMS1280 large-geometry ion microprobe at the Center for Microscopy, Characterization and Analysis (CMCA), The University of Western Australia ( Table 6). The ion microprobe was operated in multicollection mode using a Cs+ primary beam with an intensity of~3-4 nA in Gaussian mode that interacted with the sample at 20 keV during the quadruple sulphur isotope analyses. Each analysis consists of a 25 four-second cycles acquisition. The analytical session was monitored in terms of drift using two bracketing standards every 6 sample analyses. Instrumental mass fractionation (IMF) was corrected using the matrix matched reference materials for pentlandite (VMSO), pyrrhotite (Alexo), and chalcopyrite (Nifty-b); reference values can be found in [65]. IMF correction follows the procedure described in [66].

Petrography and Mineral Chemistry
The drill core OY541 at Overysel comprises mineralized, serpentinized and carbonated clinopyroxenites ( Figure 2). The least altered granoblastic clinopyroxenite (samples OY541-2, -3a) contains 90-95 vol % idioblastic diopside ( Figure 3A; Table 2). Clinopyroxene of the uppermost part of the sampled succession contains enstatite relic inclusions ( Figure 3B; Table 3). The compositions of clinopyroxene are chemically not perfectly homogeneous, but do not reveal any consistent pattern from core to rim ( Table 2). The rim and core compositions, and the standard deviations, for three samples are given in Table 2 and Figure 4. Diopside from sample 3b additionally contains exsolution lamellae of augite and/or pigeonite ( Figure 4). Some clinopyroxenite samples also display slight ductile and/or brittle deformation overprints, as is reflected by kinked clinopyroxene bands and subsequent crosscutting fractures filled with serpentine and carbonate. Pyroxene grains display varying degrees of patchy serpentinization and chloritization, which are in places associated with actinolite. Accessory phases comprise biotite, relic olivine as well as secondary calcite and dolomite usually occurs with serpentine and talc along crosscutting veinlets.

Ore Petrography
According to the terminology of Barnes, Mungall, Le Vaillant, Godel, Lesher, Holwell, Lightfoot, Krivolutskaya and Wei [67], sulphide mineralization within the footwall clinopyroxenites is heterogeneous in its distribution, and occurs as "disseminated" (up to about 5 vol % sulphides), poly-phased interstitial millimetre-to centimetre-sized grains, which locally grade into patchy "net-textured" ore (between 30 and 70 vol % sulphides), displaying sulphide-rich and sulphide-poor domains on the cm-scale. The main ore minerals are pentlandite, pyrrhotite and chalcopyrite, with accessory (titano) magnetite, cubanite and mackinawite. Pyrrhotite shows rare pentlandite exsolution lamellae, while chalcopyrite shows cubanite exsolution in places. Interstitial blebs and accumulations of the disseminated and patchy net-textured sulphides usually range from 200 to 600 µm, but may locally reach several centimetres in size ( Figure 5A). Almost all BMS throughout the drill core are subto anhedral, with serrated grain boundaries, indicating post-solidification dissolution and replacement by secondary silicates and magnetite (Figure 5 A,B).        In addition, rare globular sulphides, which to our knowledge have not been reported from the Platreef sulphide ore before, occur as rounded, interstitial aggregates, which are 100 to ca. 1000 µm in diameter ( Figure 5C,D). The differentiated globules display chalcopyrite in its upper part, separated by a smooth meniscus from pentlandite and pyrrhotite occupying the lower portion of the sulphide blebs. The globules may or may not have silicate caps, which comprise variable proportions of clinopyroxene, serpentine and magnetite, above the sulphide aggregates ( Figure 5C,D). Moreover, the BMS infrequently show a crimped angular habit due to the recrystallization of the surrounding clinopyroxene. The BMS assemblage of the deeper core section is dominated by pentlandite, while pyrrhotite is rather rare. The pentlandite in the strongly altered samples is partly replaced by secondary alteration phases, such as serpentine and carbonates, and also occurs embedded along the cleavage planes of clinopyroxene, and along veins associated with secondary serpentine and carbonates ( Figure 5E). Veinlet-controlled mineralization is dominated by chalcopyrite, and is usually associated with carbonate minerals, magnetite or secondary serpentine ( Figure 5F). Thus, these textures reflect a complex multistage mineralization history.

Bulk-Rock Chemistry
The major-and trace-element compositions of the Platreef clinopyroxenite were described in detail by Harris [20]; and Kinnaird, Hutchinson, Schurmann, Nex and de Lange [22], and thus only the characteristics relevant for this study are discussed here.
Binary variation diagrams of CaO vs. MgO, and SiO2 vs. MgO (Figure 6), are consistent with the clinopyroxene-dominated mineralogy and distinct from the "magmatic trend", representing magmatic variations due to varying plagioclase and orthopyroxene proportions, commonly found in the pyroxenitic units ( Figure 6A) [6,48]. The clinopyroxenite samples investigated in this study plot The identified PGM were bismuthotellurides, sperrylite and stibiopalladinite, which were found to be almost always associated or enclosed within the BMS blebs ( Figure 5E) and, in addition, one grain of gold was identified on a late fracture filled with chalcopyrite ( Figure 5F).

Bulk-Rock Chemistry
The major-and trace-element compositions of the Platreef clinopyroxenite were described in detail by Harris and Chaumba [3]; Armitage, McDonald, Edwards and Manby [15]; McDonald, Holwell and Armitage [62]; Holwell, McDonald and Armitage [20]; and Kinnaird, Hutchinson, Schurmann, Nex and de Lange [22], and thus only the characteristics relevant for this study are discussed here.
Binary variation diagrams of CaO vs. MgO, and SiO 2 vs. MgO (Figure 6), are consistent with the clinopyroxene-dominated mineralogy and distinct from the "magmatic trend", representing magmatic variations due to varying plagioclase and orthopyroxene proportions, commonly found in the pyroxenitic units ( Figure 6A) [6,48]. The clinopyroxenite samples investigated in this study plot off the "magmatic trend" and define a steeper slope in the CaO vs. MgO diagram ( Figure 6A). Furthermore, the low Cr (<1700 ppm; Table 1) contents are in accordance with a non-magmatic origin of the clinopyroxenites [3] (see also Figure 4 in [6]). This is further supported by their lower SiO 2 , FeO, Na 2 O and Cr, but higher CaO concentrations, than those of primary magmatic Platreef pyroxenites and norite (see [6]).
Minerals 2020, 10, x FOR PEER REVIEW 15 of 29 off the "magmatic trend" and define a steeper slope in the CaO vs. MgO diagram ( Figure 6A). Furthermore, the low Cr (<1700 ppm; Table 1) contents are in accordance with a non-magmatic origin of the clinopyroxenites [3] (see also Figure 4 in [6]). This is further supported by their lower SiO2, FeO, Na2O and Cr, but higher CaO concentrations, than those of primary magmatic Platreef pyroxenites and norite (see [6]).  [28]; and Deer, Howie and Zussman [69]. The "magmatic trend" is a correlation between CaO and MgO, displaying a magmatic variation due to varying plagioclase and orthopyroxene proportions cf. [3,6].
Multi-element diagrams of the clinopyroxenite samples display a slight negative slope, with negative anomalies of Ba, Nb, Ta, Sr, Ti and Y, while an Eu anomaly is absent (Figure 7). All investigated clinopyroxenite samples display a similar REE pattern, involving a slight enrichment of light REE (LREE) relative to heavy REE (HREE) (LaN/YbN ranging from 1.6 to 7.0; DyN/YbN ranging from 0.9 to 2.4). These patterns are similar to those of other clinopyroxenite and norite samples from Overysel and Sandsloot, as reported by Harris and Chaumba [3], Pronost, Harris and Pin [6] and McDonald and Holwell [28].   [3]). The plagioclase, dolomite, calcite, serpentine and calcsilicate compositions are compiled from Pronost, Harris and Pin [6]; McDonald and Holwell [28]; and Deer, Howie and Zussman [69]. The "magmatic trend" is a correlation between CaO and MgO, displaying a magmatic variation due to varying plagioclase and orthopyroxene proportions cf. [3,6].
Multi-element diagrams of the clinopyroxenite samples display a slight negative slope, with negative anomalies of Ba, Nb, Ta, Sr, Ti and Y, while an Eu anomaly is absent (Figure 7). All investigated clinopyroxenite samples display a similar REE pattern, involving a slight enrichment of light REE (LREE) relative to heavy REE (HREE) (La N /Yb N ranging from 1.6 to 7.0; Dy N /Yb N ranging from 0.9 to 2.4). These patterns are similar to those of other clinopyroxenite and norite samples from Overysel and Sandsloot, as reported by Harris and Chaumba [3], Pronost, Harris and Pin [6] and McDonald and Holwell [28].
Minerals 2020, 10, x FOR PEER REVIEW 15 of 29 off the "magmatic trend" and define a steeper slope in the CaO vs. MgO diagram ( Figure 6A). Furthermore, the low Cr (<1700 ppm; Table 1) contents are in accordance with a non-magmatic origin of the clinopyroxenites [3] (see also Figure 4 in [6]). This is further supported by their lower SiO2, FeO, Na2O and Cr, but higher CaO concentrations, than those of primary magmatic Platreef pyroxenites and norite (see [6]).  [28]; and Deer, Howie and Zussman [69]. The "magmatic trend" is a correlation between CaO and MgO, displaying a magmatic variation due to varying plagioclase and orthopyroxene proportions cf. [3,6].
Multi-element diagrams of the clinopyroxenite samples display a slight negative slope, with negative anomalies of Ba, Nb, Ta, Sr, Ti and Y, while an Eu anomaly is absent (Figure 7). All investigated clinopyroxenite samples display a similar REE pattern, involving a slight enrichment of light REE (LREE) relative to heavy REE (HREE) (LaN/YbN ranging from 1.6 to 7.0; DyN/YbN ranging from 0.9 to 2.4). These patterns are similar to those of other clinopyroxenite and norite samples from Overysel and Sandsloot, as reported by Harris and Chaumba [3], Pronost, Harris and Pin [6] and McDonald and Holwell [28].

PGE Mineralization
The variable PGE distribution throughout the clinopyroxenite samples is related to the heterogeneous distribution of BMS. Furthermore, the Pt/Pd of samples with bulk PGE concentrations > 400 ppm are between 0.5 and 0.8, and match the known Platreef ratios of 0.7 to 0.8 at Overysel [72]. Samples (OY541-7, 8) from the deepest part of the core yield Pt/Pd ratios from 0.5 to 0.7, whereas sample OY541-4a1 displays an unusually high ratio of 2.8.

PGE in BMS
A total of 450 spots on 72 sulphide grains (pyrrhotite, pentlandite, chalcopyrite) were analyzed by LA-ICP-MS (Table 5). In the following, the distributions of Os, Ir, Ru, Rh, Pt and Pd in pentlandite, pyrrhotite and chalcopyrite are described using median values of the different rock samples from the Overysel drill core (Table 5; Figure 8). The here-investigated pentlandite grains display several compositional populations (I, II, III) with different PGE contents at the thin section scale (Table 5), while replicate analysis of individual grains revealed a homogeneous trace element distribution. These chemically defined populations do not correlate with the textural occurrences of the pentlandite (see discussion below). Table 7   n. p.: not present; b. d. l.: below the detection limit  [70]. The grey polygon represents the bulk Pt, Pd and Au range from Sandsloot, the dashed black line represents the bulk Pt, Pd and Au range from Overysel reef pyroxenites [72], both normalized [73].  [70]. The grey polygon represents the bulk Pt, Pd and Au range from Sandsloot, the dashed black line represents the bulk Pt, Pd and Au range from Overysel reef pyroxenites [72], both normalized [73].  (Figure 9), whereas chalcopyrite and pyrrhotite display only slightly elevated Ru (~4 ppm) contents and low Os, Ir, Rh, Pt and Pd contents, similar to previous datasets (cf. [9,12,21]). The PGE patterns of the pentlandite from Overysel are similar to those from Sandsloot, and are characterized by Pt concentrations < 1 ppm, and concentrations of Ru and Rh ranging from 28 ppm to 19 ppm, respectively (cf. [9,12]). In contrast, pentlandite PGE concentrations from Turfspruit are commonly below the detection limit, except for Pd, which often has concentrations of~10 ppm [22]. The pentlandite PGE concentrations from Aurora display values below tens of ppm for Pt, Rh, Ru, Os and Ir, and 30 ppm for Pd [29], whereas Pd concentrations in pentlandite for the Flatreef can range up to 500 ppm [19]. The pentlandite Pd concentrations display a broad range concerning the location (i.e., Overysel, Sandsloot and Aurora) and the host rock type (cf. [9,12]). The clinopyroxenite and calcsilicate rock samples from Sandsloot show high average values of ca. 360 ppm, whereas the clinopyroxenite at Overysel displays lower average values of ca. 218 ppm (this study) and 120 ppm [12].    [70]. The grey polygon represents the bulk Pt, Pd and Au range from Sandsloot, the dashed black line represents the bulk Pt, Pd and Au range from Overysel reef pyroxenites [72], both normalized [73].

Mass-Balance Calculation
The relative quantities of BMS-hosted PGE in individual bulk samples can be constrained using a mass-balance based on BMS and whole-rock Pd, Pt, S, Cu and Ni concentrations (Table 7) [8,12,74]. The weight fraction (in %) of the PGE in BMS (P sul i ) is calculated using: where F sul denotes the BMS weight fraction per sample, C sul and C wr are the median concentrations of the element i in the BMS and whole rock, respectively. The weight fractions of pentlandite, chalcopyrite and pyrrhotite are calculated based on the assumption that Ni is exclusively hosted by pentlandite, Cu is exclusively hosted by chalcopyrite, and that residual S represents the pyrrhotite content [8]: Cubanite and mackinawite were excluded from the calculation due to their low modal abundance. The whole-rock Ni content was adjusted to account for the Ni hosted in clinopyroxene. Further, mass-balance results are still error-prone, considering their dependency on (1) the assumption that all Ni is as pentlandite and clinopyroxene, neglecting other Ni-bearing phases such as serpentine, relic olivine, pyrrhotite, Ni-alloys and millerite, (2) uncertainties of the analytical chemistry, (3) the nugget effect and (4) a calculation based on median values for the PGE concentrations, which show clustering.
The calculated sulphide weight fractions and the petrography of five representative clinopyroxenite samples revealed a relatively constant pentlandite content throughout the drill core samples ( Figure 10, Table 8). Chalcopyrite contents show minor variations, while pyrrhotite contents distinctly vary down-hole.

Mass-Balance Calculation
The relative quantities of BMS-hosted PGE in individual bulk samples can be constrained using a mass-balance based on BMS and whole-rock Pd, Pt, S, Cu and Ni concentrations (Table 6) [8,12,74]. The weight fraction (in %) of the PGE in BMS (Psul i ) is calculated using: where Fsul denotes the BMS weight fraction per sample, Csul and Cwr are the median concentrations of the element i in the BMS and whole rock, respectively. The weight fractions of pentlandite, chalcopyrite and pyrrhotite are calculated based on the assumption that Ni is exclusively hosted by pentlandite, Cu is exclusively hosted by chalcopyrite, and that residual S represents the pyrrhotite content [8]: Cubanite and mackinawite were excluded from the calculation due to their low modal abundance. The whole-rock Ni content was adjusted to account for the Ni hosted in clinopyroxene. Further, mass-balance results are still error-prone, considering their dependency on (1) the assumption that all Ni is as pentlandite and clinopyroxene, neglecting other Ni-bearing phases such as serpentine, relic olivine, pyrrhotite, Ni-alloys and millerite, (2) uncertainties of the analytical chemistry, (3) the nugget effect and (4) a calculation based on median values for the PGE concentrations, which show clustering.
The calculated sulphide weight fractions and the petrography of five representative clinopyroxenite samples revealed a relatively constant pentlandite content throughout the drill core samples ( Figure 10, Table 7). Chalcopyrite contents show minor variations, while pyrrhotite contents distinctly vary down-hole.   Table 8). The calculated proportions (%) of Pt and Pd hosted by pentlandite are given in Table 8. Due to the clustering of the Pd contents in individual pentlandite grains, different populations were calculated to account for the possible range of the PGE distribution. Furthermore, the clustering of the Pd concentrations also results in high standard deviations for each population ( Table 8). The mass-balance indicates that the pentlandite hosts variable fractions of the bulk-rock Pd-between~110% (sample OY 541-1) and~21% (OY 541-8a) throughout the drill core ( Figure 11). In addition, chalcopyrite and pyrrhotite host only minor Pd contents, with maximum median values of 0.72 ppm each (Table 7).  The calculated proportions (%) of Pt and Pd hosted by pentlandite are given in Table 7. Due to the clustering of the Pd contents in individual pentlandite grains, different populations were calculated to account for the possible range of the PGE distribution. Furthermore, the clustering of the Pd concentrations also results in high standard deviations for each population ( Table 7). The mass-balance indicates that the pentlandite hosts variable fractions of the bulk-rock Pd-between ~110% (sample OY 541-1) and ~21% (OY 541-8a) throughout the drill core ( Figure 11). In addition, chalcopyrite and pyrrhotite host only minor Pd contents, with maximum median values of 0.72 ppm each (Table 6).
The analyzed pentlandite and pyrrhotite show δ 34 S signatures within or close to the pristine magmatic range of −1.8% to +3.2% [16,[75][76][77]. The ∆ 33 S range, from +0.11% to +0.38% , may indicate the assimilation of sedimentary sulphur in an early stage chamber prior to emplacement ( Figures 12B and 13) [78]. origin may be derived from sulphur isotope ratios. The isotope ratios of the three analyzed sulphide phases in OY541 (chalcopyrite, pentlandite, pyrrhotite) are indicative of the mass-independent fraction of sulphur (MIFS), expressed as Δ 33 S ( Figure 12B). MIFS is almost exclusively found in rocks older than the great oxidation event, i.e., older than 2.45 Ga [19,[92][93][94]. This implies that the sulphide phases crystallized from pristine Bushveld magmas (i.e., 2.054 Ga) are not expected to show MIFS, unless sulphur was assimilated from a significantly older source.   origin may be derived from sulphur isotope ratios. The isotope ratios of the three analyzed sulphide phases in OY541 (chalcopyrite, pentlandite, pyrrhotite) are indicative of the mass-independent fraction of sulphur (MIFS), expressed as Δ 33 S ( Figure 12B). MIFS is almost exclusively found in rocks older than the great oxidation event, i.e., older than 2.45 Ga [19,[92][93][94]. This implies that the sulphide phases crystallized from pristine Bushveld magmas (i.e., 2.054 Ga) are not expected to show MIFS, unless sulphur was assimilated from a significantly older source.
The here-studied clinopyroxenite samples from hole OY541 contain a less diverse mineral assemblage than that documented at Sandsloot and Tweefontein (cf. [3,15,20,36,81]), and mainly comprise clinopyroxene, orthopyroxene and serpentine, and additional carbonates in the deeper part of the drill core. The occurrence of orthopyroxene inclusions and pigeonite exsolution lamellae in diopside supports a metasomatic formation of clinopyroxene at the expense of magmatic orthopyroxene, consistent with the observation at Sandsloot [20,82]. In addition, preliminary oxygenand carbon-isotope ratios that were determined for the dolomite grains separated from two pervasively altered clinopyroxenite samples (OY541-7b and 7c), revealing δ 13 C values of −4.96% and −5.88% , and δ 18 O values of 20.35% and 21.88% , respectively (unpublished data). This stable isotope data is consistent with late hydrothermal alteration of these Platreef clinopyroxenite samples (cf. [3,6]).
The clinopyroxenite paragenesis is in accordance with the whole-rock chemical composition. The clinopyroxenite samples studied here, like those from Sandsloot, plot off the "magmatic trend" ( Figure 6A) due to higher CaO/MgO, and lower SiO 2 , FeO and Na 2 O contents, compared to the typical magmatic Platreef pyroxenite and norite from Sandsloot, further supporting the non-magmatic origin of the clinopyroxenite (cf. [3,6,28]). This is also consistent with an overall relatively low Cr concentration (<1700 ppm). Moreover, clinopyroxenite εNd(2.06 Ga) values between −6.16 and −6.94 are similar to those of the magmatic pyroxenite and norite of the Platreef at Sandsloot and Overysel (average εNd = −6.8 ± 1.25; [6]), and also the Main Zone [average εNd(2.06 Ga) = −6.76 ± 0.47; [71]], indicating that clinopyroxenite formation at Overysel is unlikely to be related to the emplacement of a compositionally distinct melt. The clinopyroxenite was previously interpreted as indicating the assimilation of country rock, such as Malmani Dolomite (cf. [3,6,28]). However, dolostone assimilation and Ca-enrichment during ongoing pyroxene crystallization would have promoted the co-crystallization of orthopyroxene and clinopyroxene, and pyroxene-pyroxene exsolution, upon cooling, instead of the formation of clinopyroxene containing orthopyroxene relics (cf. [54]). Thus, these results rather support a post-magmatic metasomatic formation of clinopyroxenite, at the expense of former cumulate orthopyroxene.

Sulphide Mineralization and Distribution of PGEs
Even though there is broad agreement on the magmatic origin of the Platreef, its sulphide mineralization mechanism is still at the centre of the scientific debate [1,2,19,44]. In particular, the composition of the Platreef footwall's clinopyroxenite (and thus the associated sulphide mineralization) fails to reconcile with a purely magmatic evolution, and suggests modification by country rock assimilation and/or interaction with hydrothermal fluids. Crustal contamination may have modified (cf. [3,6,56]), or even triggered, mineralization by increasing the S-saturation of the magma, via the assimilation of S-rich country rocks [36,75,83], while fluid infiltration may have redistributed mineralization (e.g., [20]).
We contribute to this ongoing discussion by presenting sulphide and silicate textural observations in combination with novel geochemical data. The described globular sulphides display a characteristic distribution of chalcopyrite, pyrrhotite and pentlandite in relation to the consistently flat base of the globules (Figure 4C,D). These observations are consistent with recrystallization from the former monosulphide solid solution, which crystallized from a sulphide liquid during downward transport of the blebs (e.g., [38,84]). The silicate caps that overlie the sulphide globules comprise variable proportions of clinopyroxene, and secondary magnetite and serpentine.
LA-ICP-MS results reveal pentlandite as the major carrier of Pd (Table 5; Figure 9), whereas chalcopyrite and pyrrhotite host almost no PGE [8,11,12,16,21,29,85]. High Pt and IPGE contents were interpreted to be due to discrete micro-inclusions in pentlandite, or due to the substitution of PGM for Ni and/or Fe in BMS (cf. Figure 4E) [11,12,21]. In case of the detection of micro-inclusions of PGM, individual analyses were removed from the median values. However, the PGE substitution for Ni and/or Fe is heterogeneous at the scale of a single sample (Table 8), but homogeneous at the single-grain scale. The variability of PGE concentrations may be related to variations in the S content at S-saturation, resulting in relatively diluted PGE concentrations during the abundant formation of BMS at high sulphur saturation (cf. [38]) ( Figure 10). Therefore, the variations of pentlandite grain PGE concentrations on a thin section scale could be due to the presence of different generations of pentlandite. We cannot rule out that this variation in pentlandite PGE contents reflects a complex multistage mineralization, even though the chemically defined populations do not correlate with observed distinct pentlandite textures. Detailed 3D textural analysis by X-ray-computed tomography would be necessary to gain unambiguous textural evidence (cf. [67]). Furthermore, as indicated by the sulphide reaction-textures (i.e., their replacement by the late secondary carbonates and serpentine, as well as dissolution along the grain boundaries), post-solidification mobilization and redistribution may have contributed to the generation of the different pentlandite grain PGE populations.
S/Se ratios and S isotopes often provide important constraints for the source of S and the role of crustal contamination in triggering sulphide saturation in magmatic Ni-Cu-PGE sulphide deposits ( Figures 12B and 13). Most Ni-Cu-PGE deposits show S/Se within or close to the mantle range of 2859 to 4350, whereas S/Se in crustal rocks is typically above 3500 [36,[86][87][88]. Recent investigations on the Aurora limb show a wide range of BMS-S/Se ratios between 774 and 23384 (cf. [29]). However, the here-investigated bulk-samples show S/Se between 454 and 2500, i.e., lower than typical values of mantle and crustal rocks ( Figures 12A and 13). Thus, S/Se ratios provide no evidence for a significant effect of crustal contamination on the mineralization, but contamination by sedimentary country rocks (for which we have no evidence in the investigated drill core) may be masked by syn-or post-magmatic processes. However, the BMS reaction-textures, and their frequent association with hydrothermal alteration products in the studied clinopyroxenite, suggest that the low S/Se results from secondary S-loss (cf. [29,88,89]). This is supported by a S/Se and S isotope study of primary and secondary sulphides from the Grasvally Norite-Pyroxenite-Anorthosite (GNPA) member of the northern Bushveld limb [72,76]. These data revealed minor to no assimilation of crustal rock during the emplacement (based on δ 34 S values), and S-loss due to secondary processes, such as hydrothermal alteration, metamorphism and serpentinization [76,90,91]. Additional constraints on the sulphur origin may be derived from sulphur isotope ratios. The isotope ratios of the three analyzed sulphide phases in OY541 (chalcopyrite, pentlandite, pyrrhotite) are indicative of the mass-independent fraction of sulphur (MIFS), expressed as ∆ 33 S ( Figure 12B). MIFS is almost exclusively found in rocks older than the great oxidation event, i.e., older than 2.45 Ga [19,[92][93][94]. This implies that the sulphide phases crystallized from pristine Bushveld magmas (i.e., 2.054 Ga) are not expected to show MIFS, unless sulphur was assimilated from a significantly older source.

Conclusions
The major-and trace-element characteristics of this clinopyroxenite, in conjunction with higher CaO/MgO ratios and lower SiO 2 , FeO and Na 2 O contents in comparison to those of typical igneous Platreef pyroxenites and norite from elsewhere, are not in accordance with a magmatic trend, and rather appear to be the result of a metasomatic origin, which is in accordance with the findings of Harris and Chaumba [3]; Pronost, Harris and Pin [6]; and McDonald and Holwell [28]. Fluid-precipitated dolomite, and (pervasive) serpentinization and chloritization, suggest a subsolidus (hydrothermal) fluid alteration of the clinopyroxenite. As revealed by the petrographical and sulphur and Sm-Nd isotope evidence, the sulphur-saturation, as well as the formation of an orthopyroxene-bearing melt, was established before emplacement, and subsequently formed a heterogeneous orthomagmatic PGE-BMS-assemblage with evidence of crustal contamination in an early stage chamber [3,6,56,76]. This is further supported by the PPGE (Rh, Pt, Pd) concentrations in the BMS and mass-balance calculations, which are in general agreement with those of the Merensky reef and the UG-2, in both of which large proportions of the whole-rock Pd and Rh are hosted by pentlandite, whereas Pt and the IPGE were interpreted to mainly occur in discrete PGM, rather than being hosted by BMS. Pentlandite is a major carrier of Pd in the Platreef clinopyroxenite, which is in agreement with the observations of Junge, Oberthür, Kraemer, Melcher, Piña, Derrey, Manyeruke and Strauss [17] and Klemd, Herderich, Junge, Oberthür, Schouwstra and Roberts [12]. However, the presence of pentlandite with variable PGE concentrations on the thin section scale may be related to variations in the S content at S-saturation, resulting in relatively diluted PGE concentrations during abundant formation of BMS at high sulphur saturation (cf. [38]). Thus, the variations of pentlandite grain PGE concentrations on a thin section scale could be due to the presence of different generations of pentlandite. In addition, as indicated by the sulphide reaction-textures (i.e., their replacement by the late secondary carbonate and serpentine, as well as dissolution along the grain boundaries), mobilization and redistribution may have contributed to the generation of the different pentlandite grain PGE populations.