A record of assimilation preserved by exotic minerals in the lowermost platinum-group element deposit of the Bushveld Complex: The Volspruit Sulphide Zone

Abstract Low-gradeplatinum-group element mineralisation in the Volspruit Sulphide Zone is sulphide-poor ( Here, we present a petrological investigation on the distribution of PGEs and chalcophile metals in mineralised pyroxenite cumulates from the Volspruit Sulphide Zone, to determine the origin of the PGE mineralisation in ultramafic cumulates and evaluate whether Volspruit-style mineralisation could occur in the stratigraphically lowest, ultramafic portions of other layered intrusions. Electron microscopy of pyroxenite cumulates revealed (1) chromite inclusions containing dolomite, albite, monazite, Pb-chlorides, base metal sulphides and Pt-As minerals, (2) the presence of exotic microxenocrysts ( Systematic mapping of high-density mineral assemblages in pyroxenite cumulates across the Volspruit Sulphide Zone identified 196 precious metal mineral grains (Pt-, Pd-, Rh-, Au- or Ag-minerals), 98 Pb-sulphide grains (± Se, Cl), 27 Pb-chloride grains (± K, Se, Te, S), as well as 1 grain of Pb-telluride, 1 monazite grain and 1 grain of U–Pb-Th oxide. Trace element analyses of base metal sulphides reveal the highest S/Se values in pyrrhotite and chalcopyrite yet recorded in the Bushveld Complex. While some base metal sulphides are enriched in PGEs, the overall low-grade of the deposit and inferred fertile ultramafic magma(s) require relatively low R-factors (mass of silicate to sulphide melt) compared to other sulphide-poor PGE deposits, with a calculated R factor of ~500–3000. We consider that the presence of exotic inclusions in chromite, exotic microxenocrysts, and Pb/Zn/Cl grains enclosed within primary base metal sulphide assemblages provide strong evidence for crustal contamination in the Volspruit Sulphide Zone. The Malmani dolomite and the Black Reef quartzite within the lower Chuniespoort Group (2.2–2.4 Ga) are the most likely source of xenocrysts, assimilated in a staging chamber beneath the main Grasvally chamber, in which the Volspruit Sulphide Zone developed. It is possible that the Malmani dolomite contained an enrichment of Pb, Zn, Cl, and S minerals prior to assimilation. The assimilation of dolomite and limestone would locally increase the fO2 of the magma, triggering chromite crystallisation. The sudden removal of Fe from the melt, coupled with the addition of external sulfur triggered saturation of an immiscible sulphide melt in the ultramafic Volspruit magma. Chromite and base metal sulphides were subsequently emplaced into the main Grasvally magma chamber as a crystal-bearing slurry. Therefore, we consider it is possible for PGE mineralisation to occur in the ultramafic portion of any layered intrusion intruding in the vicinity of carbonate units. Even if this style of mineralisation in the lowermost portions of layered intrusions is sub-economic, it may reduce the grade or opportunity for PGE mineralisation higher up in the local magmatic stratigraphy, or in later magma emplacement events sourced from the same reservoir. The technique of specifically searching for microxenocrysts could be applied beyond layered intrusion research, to identify the range of crustal contaminants in other magmatic systems where macro-scale xenoliths are neither sampled nor preserved.


Introduction
Platinum-group element (PGE) mineralisation in layered intrusions typically occurs as stratiform reefs in the lower to central portions of an intrusive body, following some degree of magmatic differentiation (e.g. Naldrett, 2004;Maier, 2005). Stratiform PGE deposits hosted exclusively in the lowermost, least evolved peridotitic or pyroxenitic portions 1 of a layered intrusion are relatively uncommon (Maier, 2005). Here, we investigate the distribution of precious metals and chalcophile elements in the Volspruit Sulphide Zone -the lowermost and least evolved PGE deposit of the Bushveld Complex -hosted entirely within pyroxenite. These results will be used to: (1) identify ore-forming processes that concentrate PGEs exclusively within ultramafic cumulates, and (2) evaluate whether Volspruit-style PGE mineralisation could occur elsewhere in the Bushveld Complex, or in less-explored layered intrusions.

Geological setting
The Bushveld Complex comprises three suites of plutonic rocks: (1) the Rustenburg Layered Suite (layered ultramafic to mafic cumulates), crosscut by (2) the Rashoop Granophyre Suite and (3) the Lebowa Granite Suite (von Gruenewaldt and Walraven, 1980). The focus of this study is the mineralised Volspruit Sulphide Zone in the Lower Zone of the ca. 2056 Ma Rustenburg Layered Suite, the largest known layered mafic intrusion on Earth (Cawthorn, 2015;Zeh et al., 2015). The Rustenburg Layered Suite is divided into five lobe-shaped limbs ( Figure 1a) which span an area of >40,000 km 2 (Cawthorn, 2015). Recently, a northern extension of the Rustenburg Layered Suite has been described at the Waterberg project (Huthmann et al., 2018;Kinnaird et al., 2017).
Insert Figure 1 Insert Figure 2 The Rustenburg Layered Suite comprises a series of layered cumulate horizons ≤7 km thick (Cawthorn, 2015) (Figure 2), subdivided into five informal subzones 2 : the noritic Marginal Zone, the ultramafic Lower Zone, the mafic-ultramafic, chromite-bearing Critical Zone, the mafic Main Zone and the mafic, magnetite-bearing Upper Zone. Recently, additional informal subzones have been defined: the Basal Ultramafic Sequence in the eastern limb (Wilson, 2015) and subzones within the Waterberg project (Kinnaird et al., 2017). The stratigraphic correlation of the northern limb is summarised in Figure  2 The (South African Committee for Stratigraphy, 1980) do not officially recognise these informal subzones and would prefer workers to refer to the names of local subsuites. For example, the term for the Lower Zone of the northern limb of the Bushveld Complex is the Zoetveld Subsuite. In this paper we have followed the precedent to use the names of informal subzones instead, to reduce confusion for readers.
Informally, the Lower Zone in the eastern and western limbs of the Bushveld Complex is defined by either the increase from £ 2% to ³ 6% vol. % intercumulus plagioclase (Cameron, 1978), or the top of the olivine-rich interval ~200 m above the increase to ³ 6% vol. % intercumulus plagioclase (Teigler and Eales, 1996).
The mineralised Volspruit Sulphide Zone is hosted within the Volspruit subzone, in the Lower Zone of the northern limb ( Figure 2). Figure 3 demonstrates that there is £ 2 vol. % plagioclase throughout the Volspruit subzone, including the interval of the Volspruit Sulphide Zone. While accessory chromite is present (Figure 3), no chromitite seams occur within the Volspruit subzone (Hulbert, 1983). So despite the presence of cumulus chromite, we consider the Lower Zone of the northern limb cannot be analogous to the Lower Critical Zone in the eastern and western limbs of the Bushveld Complex.
As observed by van der Merwe (1976), mineral compositions in the Lower Zone of the northern limb (En77-En91; Fo85-Fo95; Hulbert, 1983;Wilson, 2015;Yudovskaya et al., 2013), contain greater Mg content than minerals from the Lower Zone of the eastern limb (En67-En88; Fo84-Fo87; Wilson, 2015;Yudovskaya et al., 2013) (Figure 2). Based on the relatively unevolved mineral compositions, we propose that the Lower Zone of the northern limb is more akin to the Basal Ultramafic Sequence of the eastern limb (En71-En92; Fo82-Fo92; Wilson, 2015;Yudovskaya et al., 2013) than the Lower Zone of the eastern limb. Suite. In 2014, 163 PGE resources 3 were defined in the Bushveld Complex (Zientek et al., 2014). Of these 163 resources, the Volspruit (Ni-PGE) Sulphide Zone is the only deposit known to occur in an exclusively ultramafic sequence (i.e. in cumulates where mafic minerals constitute >90% of the rock).

Platinum-group element (PGE) mineralisation in the Rustenberg Layered Suite
Ni-PGE mineralisation also occurs within the Ultramafic Sequence of the Waterberg project, an extension to the northern limb of the Bushveld Complex (Huthmann et al., 2018;Kinnaird et al., 2017). However, this Ultramafic Sequence is not ultramafic (c.f. Le Maître et al., 2002) as "the amount of interstitial plagioclase in the Waterberg ultramafic rocks commonly exceeds 10-15 vol. % …" (page 1383, Kinnaird et al., 2017). Similarly, PGE mineralisation is documented in Lower Zone lithologies beneath the Platreef at Turfspruit, in the northern limb of the Bushveld Complex (Yudovskaya et al., 2014;, yet again, this mineralisation is not hosted exclusively within ultramafic lithologies. Where modal abundance data are presented for the drill core UMT006 at Turfspruit (Yudovskaya et al., 2014), it is evident that the only portions of the Lower Zone exhibiting elevated PGE concentrations correspond with plagioclase pyroxenite, where plagioclase content exceeds 15 vol. %.

Geology of the Volspruit Sulphide Zone
The Volspruit Sulphide Zone 4 is a stratiform horizon enriched in Ni and PGEs (with an average grade of 0.14 wt% Ni and 1.25 ppm Pt+Pd+Au; Table 1), hosted within orthopyroxenechromite cumulates of the Volspruit subzone in the Lower Zone of the northern limb ( Figure   3). The Volspruit Sulphide Zone constitutes the stratigraphically lowest occurrence of potentially economic PGE mineralisation in the magmatic stratigraphy of the Rustenburg Layered Suite, and occurs in the least evolved cumulates of any Bushveld orebody ( Figure 2).
Insert Table 1 Petrologic field relationships in this region ( Figure 1c) were most comprehensively established by van der Merwe (1976van der Merwe ( , 2008 and Hulbert (1983). The Volspruit Sulphide Zone is located at least ~500 m above the base of the Rustenburg Layered Suite (Fig. 2 in Hulbert and von Gruenewaldt, 1982), however drill cores in the Volspruit region have not intersected the basement rocks beneath the Lower Zone (Hulbert, 1983;van der Merwe, 2008;Venmyn, 2010), so the depth of the Lower Zone remains unknown.
Two economic mineral resources are defined in the Volspruit Sulphide Zone: a northern and a southern orebody, both planned as open cut mines (the pit outlines of both orebodies are shown in Figure 1b-c) (Sylvania Resources Ltd., 2010;Venmyn, 2010;Zientek et al., 2014).
The northern orebody of the Volspruit Sulphide Zone (Figure 1b-c) contains two-thirds of the defined resource. This orebody is described as a "flat lying" zone of PGE mineralisation with "an average vertical width of 59 m and strike length of approximately 1,800 m" (page 5, Venmyn, 2010), gently dipping ~15 o to the NW (Hulbert, 1983). The southern pit comprises the remainder of the resource: an orebody with an average vertical width of 47 m, dipping ~45 o NW for 1 km along strike (Venmyn, 2010).
Faulting is evident in both orebodies, although detailed structural information is not available for this region (Venmyn, 2010). Based on the map of Hulbert (1983) (Figure 1c), we suppose that the northern and southern orebodies are the same Volspruit Sulphide Zone that have since experienced ~1.5 km of dextral displacement along a N-S striking fault.
Partially digested country rock xenoliths and associated partial melts of floor rocks are observed in a prominent ~100 m-thick harzburgite horizon that occurs a few tens of metres beneath the Volspruit Sulphide Zone; these features are less common in the pyroxenites of the  (Hulbert, 1983;Hulbert andvon Gruenewaldt, 1985, 1982;Paktunc et al., 1990;this study).
They attribute Ni-PGE mineralisation in the Volspruit Sulphide Zone to a combination of (1) gradually increasing sulfur in the magma via fractional crystallisation of ~500 m of ultramafic cumulates (i.e. the cyclic units beneath the Volspruit Sulphide Zone), with (2) an influx of a volumetrically small component of less primitive (i.e. basaltic) magma, that upon mixing with the resident magma, created a decrease in temperature, resulting in sulfur saturation and chromite precipitation (Hulbert and von Gruenewaldt, 1982).

Sulphide-poor PGE mineralisation in the ultramafic portions of layered intrusions
Of 115 PGE deposits or occurrences in layered intrusions worldwide, Maier (2005) documented 18 PGE occurrences hosted within ultramafic silicate rocks from the lower portion of a layered intrusion. However, literature on cumulate rocks is fraught with petrologic inconsistencies caused by the range of proposed nomenclature schemes: the IUGS system (Le Bas and Streckeisen, 1991;Le Maître et al., 2002;Streckeisen, 1976), systems based on cumulus and intercumulus texture (Hunter, 1996;Irvine, 1980) and occasionally terms inherited from local mine geologists. In some cases, the lack of petrologic information makes it difficult to determine whether mineralisation was exclusively hosted within ultramafic cumulates (>90% mafic minerals), or in a mixture of mafic and ultramafic lithologies.
Insert Table 2 The eight sulphide-poor PGE occurrences within the ultramafic sequences of layered intrusions are documented in Table 2, in order of increasing sulphide content. We have excluded examples of PGE mineralisation in lamprophyric cumulates (Barnes et al., 2008), discordant dunite pipes (e.g. McDonald, 2008), Alaskan-type deposits (e.g. Thakurta et al., 2008) and similar concentrically-zoned intrusions (e.g. Helmy and Mogessie, 2001) from the compilation presented in Table 2.

Mineralisation in the lower portions of layered mafic intrusions
While sulphide-poor PGE mineralisation in ultramafic cumulates constitutes ~7% of known PGE deposit styles, any economic or subeconomic mineralisation could deplete a fertile magma in PGEs. Thus ultramafic-hosted deposits have the potential to destroy the opportunity for other styles of PGE mineralisation higher up in the magmatic stratigraphy (c.f., Latypov et al., 2017), or in subsequent magma emplacement events sourced from the same reservoir (c.f., Mungall et al., 2016). A few detailed analytical studies of PGE distribution in sulphide-poor, ultramafic-hosted PGE deposits exist (e.g. Barnes et al., 2011;Diella et al., 1995;Gervilla and Kojonen, 2002;Grokhovskaya et al., 2012;Knight et al., 2011;Teixeira et al., 2015), but do not contain coupled studies comparing platinum-group mineral assemblages with the trace element chemistry of sulphides. Until now, few data have been available on the distribution of PGE and chalcophile trace elements from the Volspruit Sulphide Zone (e.g. Hulbert, 1983;Paktunc et al., 1990). By filling this knowledge gap, we hope to refine the genetic model for mineralisation in the lowermost portion of the Bushveld Complex. Understanding how and why fertile, primitive magmas achieve sulphide saturation in layered intrusions is crucial to understanding the formation of layered intrusions and guiding future mineral exploration.

Sampling and Analytical Techniques
To determine the mechanism(s) for ore genesis, we used nine pyroxenite samples from the Volspruit Sulphide Zone to (1) characterise high-density mineral assemblages (including platinum-group minerals), and (2) quantify the trace element chemistry of base metal sulphides.
Our methodology for identification of platinum-group minerals in sulphide ore differs from previous studies because: (1) we systematically recorded all high-density minerals present in each sample (not just platinum-group minerals & electrum), and (2) the texture of every high-density phase and the adjacent mineral assemblage were compiled and are provided as a visual atlas (Appendix A) and spreadsheet (Appendix B).

Samples
Nine samples of mineralised pyroxene ± chromite cumulates from the northern orebody of the Volspruit Sulphide Zone were taken from seven intervals between 67.88-121.88 m depth in the drill core GVN-042. Each sample is ~ 4 cm in length. By georeferencing the figure on page 8 of Pan Palladium Limited, (2002), we estimate that the core GVN-042 was collared at 24°20'49.66"S, 28°56'54.13"E (the star shown in Figure 1c) on the farm Volspruit 326KP. The location of centre-pivot irrigation systems around the drill-hole permit us to locate the drill core collar with reasonable accuracy (within a twenty-metre radius).
The nine drill core samples were impregnated with resin and prepared as rectangular The mineralised pyroxene ± chromite cumulates used in this study were sampled across 53 m of drillcore. As the northern orebody is ~59 m thick in this area (Venmyn, 2010), these samples span from the top to the of the base of the Volspruit Sulphide Zone. Each thin section billet used in this study was only located within a 1 m interval of core GVN-042, corresponding to bulk geochemical assay data provided by Pan Palladium (Table 1). Therefore, the depth range of each sample is reported as a range (e.g. sample 7B = 67.88-68.88 m), but for convenience in figures, values are plotted as the mean of each depth range (e.g. sample 7B = 68.38 m).

Bulk rock geochemistry
The bulk rock geochemistry of one-metre intervals of drillcore was provided to us by Pan Palladium. The bulk geochemical data was determined by X-ray fluorescence for Ni, Cu, Cr and Sr; Rh, Pd, Pt and Au were determined by fire assay and inductively coupled plasma mass spectrometry.

Reflected light microscopy
Reflected light microscopy was used to record petrographic relationships in each polished block at 2.5x and 10x magnification, using a Nikon Optishot reflected light microscope at Cardiff University.

Electron microscopy: identification and characterisation of high-density minerals
A polished block from each stratigraphic level in the deposit was carbon-coated and examined for high-density minerals under the FEI XL30 field emission gun environmental scanning electron microscope (SEM) at Cardiff University. The perimeter of each polished rock surface was outlined using FEI software to facilitate sample navigation in the SEM. Analyses and back-scattered electron images were acquired at 20 kV with a nominal beam current of ~2 nA. Once the sample was in focus, the brightness and contrast of the back-scattered electron image were adjusted so that silicate, oxide and base metal sulphide minerals were black, and only minerals with a greater molar mass (and therefore, greater density) were visible (e.g. platinum-group minerals and Pb-minerals). Each sample was then systematically searched for high-density minerals at 350x magnification by conducting manual y-axis traverses across the stage at a moderate scanning speed (1.68 ms per line with 968 lines per frame). Once a highdensity mineral was found, it was imaged and analysed semi-quantitatively using an Oxford Instruments X-act energy dispersive spectrometer (EDS) and Inca X-ray analysis system. The brightness and contrast were recorded and briefly readjusted to image and analyse the adjacent mineral assemblage surrounding each high-density mineral. All visible minerals observed using this method were recorded and presented in Appendixes A and B.
To collect additional analyses and back-scattered electron images, we used a bench-top Phenom XL scanning electron microscope with a built-in EDS detector at the University of Wollongong.

Trace element chemistry of base metal sulphides
Trace element contents of base metal sulphide minerals were analysed in situ, in each of the nine samples using laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS). Relative abundances of trace elements were determined by ablating traverses (40 μm diameter wide; 300 μm long) at 6 μm/sec across base metal sulphide grains. We used a New Wave Research UP213 UV laser system coupled to a Thermo X series 2 ICP-MS housed at Cardiff University, following the methods described by Prichard et al. (2013) and Smith et al. (2014).
Given the fine grain size of sulphides analysed in this study, regions of the traverse containing mixed peaks from co-ablation of silicate, oxide, other sulphide or platinum-group minerals were excluded.

Bulk rock chemistry
The bulk rock chemistry of one-metre drillcore intervals from the northern orebody of the Volspruit Sulphide Zone is provided in Insert Figure 4

Petrologic features of orthopyroxenites in the Volspruit Sulphide Zone
The mineralised cumulates from the Volspruit Sulphide Zone are medium-grained As well as identifying the mineralogy of high-density mineral assemblages in the Volspruit Sulphide Zone, electron microscopy of petrological relationships surrounding the high-density assemblages revealed: (1) sub-rounded to spherical inclusions hosting exotic 5 phases within chromite, (2) the presence of exotic microxenocrysts (<300 μm diameter) in the pyroxenite matrix, and (3) base metal sulphide assemblages atypical of sulphide-poor PGE mineralisation.

Chromite-hosted inclusions
Silicate, carbonate and sulphide inclusions are observed within chromite crystals, with differences in the inclusion density between samples (Figure 6a (Table 1).
In the sulphide-only inclusions, pyrrhotite is the dominant mineral, with pentlandite, magnetite and minor chalcopyrite. In one sulphide inclusion, flames of Fe-rich pyrrhotite were observed amongst Fe-poor pyrrhotite ( Figure 6h). Three-quarters of sulphide-rich inclusions contain sperrylite (PtAs2) too fine to observe using systematic high-density mineral identification at 350x magnification. The largest high-density mineral within a chromite-hosted sulphide inclusion was an elongate 0.9 x 0.2 μm sperrylite inclusion ( Figure 6f). Sperrylite occurs at the contact between sulphides (pyrrhotite, pentlandite and chalcopyrite) and enclosing chromite ( Figure 6). In only one case, sperrylite was observed away from the inclusion margin, at a pentlandite-pyrrhotite grain boundary.
In the three sulphide-silicate inclusions, Pb-chlorides and Pb-sulphides were observed in association with pyrrhotite, magnetite and hydrated Fe-Mg silicates such as phlogopite. In both the silicate-sulphide inclusions and the sulphide-only inclusions, magnetite is only observed at the boundary between chromite and pyrrhotite, so it likely formed post-entrapment.
Figure 9a-f demonstrates that base metal sulphide assemblages include accessory sphalerite, Pb-sulphide minerals (± minor Se, Cl) and a suite of Pb-chloride minerals (± K, S, Se, Te). While sphalerite and Pb-minerals are not exclusively hosted within base metal sulphide assemblages, it is significant for the genesis of the Volspruit Sulphide Zone that many are, as they appear to be primary inclusions.
Unlike the Pb-minerals, sphalerite is not a high-density phase that is easily distinguished under back-scattered electron imaging; only two occurrences are observed -although they may be more prevalent than we have documented. In one instance at 106.88-107.88 m depth (the sample containing the highest number of included Pb-sulphides and Pb-chlorides), sphalerite is included in the centre of a base metal sulphide bleb. Figure 9b shows the ~125 μm 2 sphalerite grain with curvilinear to cuspate margins bounded by pyrrhotite and chalcopyrite. We interpret this image as primary sphalerite, crystallising in the remaining space against the already solidified pyrrhotite. In the same sample, there is a second occurrence of sphalerite with a more ambiguous origin, with relict sphalerite (<4 μm 2 ), chalcopyrite and pentlandite trapped within the cleavage of a blocky Se-bearing galena crystal ( Figure 9c).  Figure 9a). Figure 8a shows both Pb-chloride and Pb-sulphide included within the same sulphide bleb. As Pb-minerals are high density, a more complete analysis of Pb-sulphide and Pb-chloride minerals is presented below.

Quantification and characterisation of high-density mineral assemblages
A total area of 32.895 cm 2 of pyroxenite cumulates were searched for high-density minerals, identifying 0.00027 area % high density phases. The high-density mineral assemblage in the Volspruit Sulphide Zone contains: 196 precious metal mineral grains (Pt-, Pd-, Rh-, Auor Ag-minerals; 24 area %), 126 Pb-mineral grains (76 area %), and 2 exotic mineral grains (<0.1 area %) ( Figure 10). These data are available to download electronically in Appendix B.
Below, we report the results of the precious metal mineral assemblages and Pb-minerals separately. Precious metal minerals are only observed in close association with Pb-minerals in three instances (Figure 9d-f). The two exotic high-density minerals, (La-Ce) monazite and U-Th-Pb oxide ( Figure 8) are both interpreted as microxenocrysts, as discussed above.

Characterisation of precious metal minerals
The 196  and Pd-sulphides are notably absent from this assemblage.
Insert Figure 10 The abundance of precious metal minerals varies with height in the Volspruit Sulphide Zone ( Figure 10). However, there are no systematic mineralogical trends with depth in the core.
The Pt/Pd of precious metal minerals oscillates above and below 1 throughout the Volspruit Precious metal minerals were observed in a variety of host assemblages ( Figure 11), with textures ranging from spongy, symplectic to euhedral. Of the 196 precious metal mineral grains identified, 39% are enclosed within serpentinite near 7 altered base metal sulphide(s), 29% are at the altered grain boundary between base metal sulphide(s) and silicate(s), 10% are enclosed within base metal sulphide, and 8% are enclosed within serpentinite. The remainder of assemblages occur at altered grain boundaries between base metal sulphide, silicate and chromite (4%), grain boundaries between serpentine and pyroxene near altered base metal sulphide (4%) and in the absence of base metal sulphide (4%), altered silicate-chromite grain boundaries (1%) and enclosed within silicates in fractured chromite (0.5%).

Insert Figure 11
Precious metal minerals either occur as individual phases or as intergrowths with more than one mineral (Figure 7). Two assemblages contain clear core-rim relationships, with Pd(Bi,Te,Sb) overgrowing a cylindrical core of PtAs2 at the margin of a pentlandite grain ( Figure 7b)

Characterisation of Pb-minerals
Pb-minerals were observed at each of the seven intervals from the Volspruit Sulphide Zone. Of the 126 Pb-mineral grains observed in this study, 98 grains (92 area%) were Pbsulphides (± Se, Cl) and 27 grains (7 area% Zone, where Pb is dominantly hosted by Pb-sulphides. Figure 10 shows that while Pb-sulphides are more abundant overall, they were only observed at four out of seven depth intervals. While Pb-chlorides were less abundant, they were observed at six out of seven depth intervals. Pb mineral assemblages exhibit a range of textural associations ( Figure 11). Of the 125 Pb-minerals, 50% of these are enclosed within serpentinite. 20% occur within serpentinite near altered base metal sulphide assemblages. 8% are enclosed in base metal sulphide assemblages, while 5% occur at altered grain boundaries between base metal sulphide(s) and silicate(s). The remainder occur at serpentine-pyroxene grain boundaries (sometimes near altered sulphide), enclosed within chromite-hosted silicate inclusions, enclosed within altered base metal sulphide assemblages in fractured chromite, enclosed within silicate(s) in fractured chromite, enclosed within pyroxene and at altered triple junctions between silicate-chromite-base metal sulphide(s).
Half the Pb-sulphide minerals observed in this study have blocky/graphic textures. Nine out of the ten K-Pb-chlorides, as well as two Pb-chloride grains and four Pb-sulphide grains exhibit a spongy texture. Otherwise, Pb-minerals are typically homogenous. and PbCl2 (e.g. Figure 6a-b).

Trace element chemistry of base metal sulphides
LA-ICP-MS traverses across pyrrhotite, pentlandite, chalcopyrite and cubanite demonstrate that a range of precious metals and semimetals are concentrated within base metal sulphides in the Volspruit Sulphide Zone. These data are available to download in Appendix E, are summarised in Table 3 and are plotted against depth in the Volspruit Sulphide Zone in Figure 12. Due to the small size of base metal sulphides and fine intergrowths between different sulphide phases in the Volspruit Sulphide Zone, pure trace element analyses 8 could not be obtained for each mineral at every depth. A lack of suitable reference materials meant that some trace metals such as Pb could not be measured. Nevertheless, we measured a diverse range of trace metals in sulphides from the Volspruit Sulphide Zone, which reveal critical information about the distribution of platinum-group elements and semi-metals between minerals in base metal sulphide assemblages.

Concentration of precious metals in base metal sulphides
Most notably, magmatic sulphide assemblages in the Volspruit Sulphide Zone achieve correct for the internal standard ( 33 S), thus producing inaccurate trace element concentrations. Therefore, mixed analyses are not considered in this study.
Insert Figure 12 Insert

Discussion
The Volspruit Sulphide Zone is a rare occurrence of sulphide-poor PGE mineralisation in the ultramafic portion of a layered intrusion, with low-grade ore distributed over a ~60 m horizon. It is the only known deposit of this kind in the spectrum of Bushveld PGE mineralisation. Hulbert & von Gruenewaldt (1982;1985) proposed that the Volspruit Sulphide Zone formed as a consequence of sulphide melt saturation in response to S enrichment in the magma during fractionation. They propose that this ultramafic magma mixed with a sudden influx of denser, cooler and less primitive basaltic magma, which triggered saturation of chromite and sulphides to form the Volspruit Sulphide Zone. In light of the new data presented above, and recent advances in our knowledge of ore-forming processes, we consider that the current genetic models for mineralisation of the Volspruit Sulphide Zone (Hulbert & von Gruenewaldt, 1982;1985) require significant re-evaluation.

Evidence to support a high degree of sediment assimilation
We contend that the Volspruit Sulphide Zone experienced a high degree of country rock assimilation. Our interpretation is supported by quantitative geochemical data: (1) (Eckstrand and Hulbert, 1987). In mineralised GNPA samples from the Rooipoort area, the highest S/Se values (up to 8,900) occur within secondary pyrite, with S/Se up to 5,600 in primary pyrrhotite (Smith, 2014;Smith et al., 2016). In the Platreef at Turfspruit, the highest S/Se values (up to 10,800) are also recorded in pyrite, with S/Se up to 7,500 in pyrrhotite.
Magmatic sulphide assemblages in the Volspruit Sulphide Zone contain the highest S/Se values yet recorded in the Bushveld Complex, with S/Se up to 21,566 in pyrrhotite and 6,804 in chalcopyrite. We return to a more in depth discussion on the origin of extremely high S/Se values after reviewing all the evidence for assimilation in the Volspruit Sulphide Zone.

Qualitative evidence for assimilation
Multiple processes can be invoked to explain the range of δ 34 S and S/Se values in the Volspruit Sulphide Zone (c.f. Queffurus and Barnes, 2015;Smith et al., 2016). Fortunately, the regional geology and petrology of samples provide strong, if qualitative evidence in support of sedimentary assimilation, and allow us to identify some of the assimilated sediments.
While ~500 m of the Volspruit subzone is documented beneath the Volspruit Sulphide Zone (Figure 3), country rock beneath the Volspruit subzone has never been intersected by  (1) pyroxene is absent and not abundant within silicate inclusions, and (2) relict dolomite is preserved. While post-entrapment equilibration likely alters the distribution of incompatible elements between chromite and included silicate minerals (Spandler et al., 2007), this process cannot explain the exotic mineralogy observed within chromite grains in this study.
In order to assess the end-members for assimilation, we have compiled a list of xenocrystic phases and contaminants, and compared their potential sources (Table 4). While this style of analysis is qualitative, it provides strong evidence in support of carbonate assimilation. The source of albite and U-Th oxide could be from assimilation of a felsic intrusion, or clastic sediments.
Insert Table 4 The combination of primary Pb-sulphides, Pb-chlorides and sphalerite in base metal sulphide assemblages is rather unusual in magmatic sulphide deposits. It is possible that their presence is under-reported in some studies, because they are (1) accessory minerals, (2) considered as secondary assemblages, and (3) do not concentrate precious metals other than Ag. Sphalerite, galena and Pb-chlorides associated with magmatic sulphides are also documented both Cu-rich ores at Sudbury, Ontario (Dare et al., 2014), and in Cu-Ni ores from the Minnamax deposit, at the base of the Duluth Complex, Minnesota (McSwiggen, 1999). In extreme circumstances (e.g. metamorphosed sulphide deposits), galena and sphalerite can fractionate from a sulphide liquid (Mavrogenes et al., 2013), rather than exsolve from a crystalline sulphide.
The association of primary Pb-sulphides, Pb-chlorides and sphalerite in base metal sulphide assemblages could be explained by either (1) crystallisation from residual liquids, or (2) addition from an external source. While some studies suggest that this mineral assemblage is consistent with a magmatic origin (e.g. Mungall and Brenan, 2003), reviews of sulphide liquid evolution in natural systems (Holwell and McDonald, 2010;Naldrett, 2004) do not include these minerals in the resultant mineral assemblage. As some of the Pb-sulphide inclusions in the Volspruit Sulphide Zone are up to 50 µm across (e.g. the sample at 106.88-107.88 m depth), these accessory minerals should be readily and routinely identified under reflected light if they are typical products of sulphide liquid fractionation.
Instead, the rare association of primary Pb-sulphides, Pb-chlorides and sphalerite within base metal sulphide assemblages imply that Pb, Zn and Cl were added from an external reservoir, and incorporated into the sulphide liquid, as halogens can be dissolved in sulphide melts (c.f. Mungall and Brenan, 2003). Hulbert (1983) notes two localities in Pretoria Group metasediments southeast of the Volspruit orebodies where sulphides are visible (shown with asterisks in Figure 1c). Unfortunately, no further information on these localities was given.

Massive sulphides have been documented in metasedimentary sequences from the Silverton
Formation and Timeball Hill Formation, within the Pretoria Group (Reczko et al., 1995).  (Martini et al., 1995). An alternative hypothesis is that basinal brines (i.e. CO2-H2O-Cl ± metal-rich fluids) were introduced to the Volspruit magmatic system at the time of, or prior to chromite mineralisation.
From the range of xenocrysts in Table 4, we infer that assimilation of Mississipi Valley Type deposits hosted within dolomites of the Malmani subgroup is the best explanation for the presence of Pb-chloride minerals in the Volspruit Sulphide Zone. As Mississipi Valley Type deposits are patchy and structurally controlled, there was some serendipity involved creating the Volspruit Sulphide Zone. However, if Pb-chlorides are observed more widely in layered mafic intrusions elsewhere, this interpretation will require revision.
The addition of a Pb and Zn-bearing source did not result in ores with a high proportion of base metal sulphides, as observed in Ni-Cu deposits, where 20-90 vol. % sulphide is typical (Naldrett, 2004). It is possible that assimilation initially created a high volume of sulphide liquid, which was subsequently reduced via sulphide resorbtion (Kerr and Leitch, 2005) or by mechanical sorting during hydrodynamic processes . In either case, Ni-PGE mineralisation in the Volspruit Sulphide Zone is highly unusual in the spectrum of magmatic sulphide deposits, as it contains strong evidence for assimilation of sediments yet only contains 2-5 vol. % sulphide.

Evidence to explain chromite formation
Our study has documented a diverse assemblage of chromite-hosted inclusions ( Figure   6). Hulbert & von Gruenewaldt (1985) observed similar chromite inclusions. They inferred that euhedral, inclusion-free chromite co-crystallised with ultramafic cumulates; in situ postcumulus sintering of chromite produced amoeboidal textures and rounded inclusions (viz. Figure 6 in Hulbert & von Gruenewaldt, 1985). They attribute the onset of chromite saturation to increased fO2 caused by fractional crystallisation of mafic minerals. Thus, Hulbert & von Gruenewaldt (1985) consider that chromites are post-cumulus oikocrysts, overgrowing pyroxene chadacrysts from the adjacent assemblage of stratiform cumulates. (1985)  This is because we observe cumulate assemblages comprised of orthopyroxene and serpentinised mafic silicates, yet a diverse array of mineral inclusions in chromite. Dolomite, sulphides, hydrous and fractionated silicates as well as exotic high-density phases are all present in sub-rounded inclusions. As chromite inclusions are mineralogically distinct from their cumulate matrix (i.e. inclusions are dominated by "exotic" mineralogy rather than pyroxene), we contend that chromite inclusions observed in this study were trapped during chromite crystallisation in an earlier, highly contaminated magma, prior to emplacement at the current level in the Volspruit subzone.

The Hulbert & von Gruenewaldt
The morphology of observed chromite inclusions range from negative crystal shapes to sub-spherical. Inclusions often occur within chains of interconnected chromite clusters, similar to those described in komatiitic cumulates (e.g., Vukmanovic et al., 2013). We envision that negative crystal shapes and subrounded-rounded inclusions in chromite cores result from dendritic growth and subsequent recrystallisation, as described by Vukmanovic et al. (2013).
Dendritic chromite is observed in komatiitic dunites and spinifex zones at or near the flow tops of komatiites, as a result of Cr supersaturation in the melt (e.g., Barnes, 1998).

Carbonate assimilation and its affect on oxygen fugacity
Carbonate assimilation is often considered a contributing factor to ore-forming processes in magmatic sulphide deposits (e.g., Harris and Chaumba, 2001;Lehmann et al., 2007;Maier et al., 2007). However, the significance of this contribution is unclear, as there are competing effects of oxidation and reduction during carbonate assimilation and/or following the addition of CO2 to a magma. To some extent, the assimilation of carbonate will reduce the fO2 of a melt, because (1) the creation of skarns promote oxidation of Fe, thus reducing the ferric iron content and fO2 of the magma (Spandler et al., 2012), and (2) increasing the CO2 content of a melt reduces H2O solubility, increasing the CO2/H2O of the exsolved fluid phase (Mollo et al., 2010). Conversely, the experimental work of Simakin et al., (2012) demonstrates that the addition of CO2 oxidises Fe in the melt and promotes spinel nucleation. The oxidative capacity of CO2 addition is also supported by modelling (Wenzel et al., 2002). More recent experiments suggest that carbonate assimilation produces sharp fO2 gradients within in a magma reservoir (Mollo and Vona, 2014). In natural systems, this process is likely to be rapid ( In layered intrusions, there is evidence for locally increased fO2 at the margins of magmatic reservoirs in contact with carbonates. For example, thin chromite stringers are observed between the contact of serpentinized harzburgite and a calc-silicate xenolith in the Lower Chromite zone of the Platreef (Yudovskaya and Kinnaird, 2010). Dolomite xenoliths in the Ioko-Dovyren Intrusion are often mantled by forsterite and spinel rims, and some olivinespinel skarns contain accessory base metal sulphide assemblages (Wenzel et al., 2002). In this way, it is possible to locally nucleate spinel, base metal sulphide, Mg-rich silicates and Ca-rich pyroxene (e.g. pigeonite), which are diluted during transport from the initial magmatic reservoir.

How enriched are PGEs in the Volspruit Sulphide Zone?
As platinum-group minerals were directly observed within chromite-hosted sulphide inclusions (Figure 6), it is unlikely that late-magmatic fluids migrating through the crystal pile (c.f. Boudreau and McCallum, 1992) played a significant role in the PGE enrichment of the Volspruit Sulphide Zone. Instead, the texture and high tenor of base metal sulphides in the Volspruit Sulphide Zone indicate that PGEs were partitioned into an immiscible sulphide liquid, which scavenged PGEs from a larger body of silicate magma (i.e. high R factor).
An inherent problem in reconciling the chalcophile element budget of layered mafic intrusions is that it is difficult to normalise the concentration of elements to 100% sulphide liquid -a calculation necessary for estimating the degree of precious metal enrichment in a sulphide liquid. Recalculating the tenor of base metal sulphides to 100% sulphide in sulphidepoor PGE deposits can create large errors (Barnes and Ripley, 2016). The reasons for this are two-fold: (1) the calculation assumes that chalcophile elements in magmatic sulphide assemblages are only controlled by the minerals pyrrhotite, pentlandite and chalcopyrite, and (2) calculations at such low sulphide concentrations magnify errors. The data in Table 1 is acquired over one-metre drillcore intervals, while the modal abundance of base metal sulphides was variable within each sample and estimated from an area of only a few cm 2 . Therefore, these data cannot be reconciled meaningfully. Instead, we use published PGE data with corresponding S analyses from von Gruenewaldt et al. (1989), to estimate the degree of precious metal enrichment required to form the Volspruit Sulphide Zone (i.e., the R factor, where the mass of silicate liquid:sulphide liquid = R:1; Campbell and Naldrett, 1979).
Mineralisation of the Volspruit Sulphide Zone formed from a relatively unfractionated sulphide liquid. Figure 13 shows the range of R-factors required to concentrate PGEs in the Volspruit Sulphide Zone are between 500:1 to 3000:1, assuming an ultramafic magma (the B1-UM CD-005 magma 9 ; Barnes et al., 2010), a range of published partition coeofficents (Barnes and Ripley, 2016) and bulk PGE and S analyses from von Gruenewaldt et al. (1989). However, the R-factor may also be controlled by kinetic, rather than equilibrium processes such as the nucleation to growth rate of immiscible sulphide liquid, the rate of chalcophile element diffusion, and the rate of melt migration (Mungall, 2002 (Naldrett, 2004).

Insert Figure 13
The range of Pd measured in pentlandite from the Volspruit Sulphide Zone (39.7-236.4 ppm Pd) is similar to the range of Pd in pentlandite from the GNPA at Rooipoort (below detection-386 ppm Pd; Smith et al., 2014), and the Platreef at Overysel (68.6-183 ppm Pd;  and Sandsloot (67.7-170 ppm Pd; . From these data it is evident that PGEs are not homogenously distributed amongst sulphide minerals within a given ore deposit. R-factor (< S/Se with > silicate:sulphide and PGE tenor) and (4) crystallisation (S/Se enriched in MSS relative to ISS) (Queffurus and Barnes, 2015). Smith et al. (2016) proposed that 9 The B1-UM chill CD-005 was chosen as ECBV050 (the only other analysis with PGE values) contained less PGEs than the more evolved B1 magma.

Processes controlling S/Se values in the Volspruit Sulphide Zone
sulphide resorbtion (viz. Kerr and Leitch, 2005) lowers the S/Se of the sulphide melt, decoupling proxies for contamination (S/Se and δ 34 S) in magmatic sulphide assemblages.
Secondary alteration from metamorphism, metasomatism from magmatic fluids, and lowtemperature processes also preferentially remove S and thus decrease S/Se values (Queffurus and Barnes, 2015). Complex. Secondly, if S-addition from assimilation were the primary control on S/Se, then this should dilute the tenor of the sulphide minerals. That is, we would expect to see PGE grade of sulphides decrease with S/Se, which we do not observe in pyrrhotite (Figure 14).

Insert Figure 14
So while we consider that S-addition is an important control on S/Se in the Volspruit Sulphide Zone (evidenced by the addition of Pb and Zn into base metal sulphides), another process must be invoked to explain the high and heterogenous S/Se values. Exploring processes whereby Se is lost from sulphide is the only remaining option.
Initially, we considered that the high and heterogenous S/Se values may be caused by precipitation of Se-bearing Pb-sulphides and Se-bearing Pb-chlorides from a fractionating sulphide melt. Se is known to be highly compatible in the structure of galena, along with Ag and other semi-metals such as Te and Bi (George et al., 2015(George et al., , 2016. Thus, we can test the influence of galena and Pb-chlorides on the S/Se of sulphides during sulphide melt evolution by assessing the depletion of Ag, Te and Bi with S/Se values ( Figure 14). If the high S/Se of pyrrhotite is inherited from fractionation of galena (i.e. Se removal) from a sulphide meltnoting that galena may not necessarily be visible in the exposed, polished sulphide bleb -then high S/Se analyses should correlate with the lowest values of Ag, Te and Bi, elements also highly compatible in galena. However, this is not observed in pyrrhotite (Figure 14), indicating that high S/Se in pyrrhotite is not inherited from fractionation of galena. Ag decreases with increasing S/Se in chalcopyrite, but not Te and Bi. It is more likely that Ag varies with Au and PGEs, rather than being controlled by the S/Se (and thus galena fractionation) in chalcopyrite.
These data are also supported by the observation that Se-bearing Pb-chlorides or Pb-sulphides In order to fully exclude this hypothesis, we calculated the influence of accessory Sebearing sulphides (galena and Pb-chlorides) on the S/Se budget of the sulphide liquid. While petrologically significant, the volume of Pb-sulphide and Pb chlorides in the Volspruit is extremely small. Of the 33 cm 2 searched for high-density minerals in this study, Pb sulphides account for 0.0002 area % and Pb-chlorides account for 0.00002 area %. So assuming mean modal proportions of base metal sulphides from Hulbert (1983) and normalising these to 3 area %, we considered the effect of adding or removing the mean proportion of galena and Pbchloride measured in this study, assuming a very generous estimate of 2 wt% Se in each Pbphase. As these phases are volumetrically insignificant, they only increased the S/Se of the sulphide liquid by 0.01%. Thus, at such a low volume, crystallisation of Se-rich Pb-sulphides and Pb-chlorides is unlikely to control the S/Se budget of individual magmatic sulphide blebs.
Secondly, we considered the possibility that Se is lost during degassing. In basaltic systems, the proportion of sulfate in a melt (S 6+ /ΣS) increases above ~FMQ-1 e.g., Jugo et al. (2005). So above ~FMQ-1, sulfur becomes increasingly volatile and may be lost during degassing. While the speciation of Se 6+ /ΣSe with oxidation state in basaltic systems is not defined, the transition from immobile selenide to volatile selenate is estimated to occur at much higher oxidation states. This is why S/Se is used as a proxy for degassing in magmatic systems e.g., (Jenner et al. 2010). Figure 10 in Jenner et al. (2010) suggests the transition from immobile selenite to volatile selenate may begin above ~FMQ+2. Regardless of the absolute values of the selenite-selenate transition, it is likely that assimilating carbonate country rocks into an ultramafic magma will produce extreme gradients in fO2. Locally, portions of magma in close contact with country rock may become extremely oxidised, accompanied by extensive degassing from surrounding carbonates. In such an environment, we consider that it might be possible to oxidise Se to a volatile selenate species -so that both S and Se would be lost during degassing. Small changes in the oxidation state could change the proportion of Se lost. Thus, assimilation of a sulphide-rich carbonate would alter the S/Se of magmas in the Volspruit subzone by adding crustal S, and by local removal of S and Se via degassing of an oxidised magma. We consider this process the best available explanation to explain high and heterogenous S/Se values in the Volspruit Sulphide Zone.
An alternative hypothesis to explain the high S/Se values is that Se was removed from sulphides in the Volspruit Sulphide Zone during oxidation by high-pH postmagmatic fluids, as observed in the Jinchuan Ni-Cu-PGE deposit (Prichard et al., 2013). However, this hypothesis is not supported by either (1) the range of Se concentrations observed in sulphides, or (2) the absence of an alteration overprint in BMS assemblages. As pentlandite contains up to 237 ppm Se, this hypothesis would require Se to be selectively leached from pyrrhotite, but retained in pentlandite. As Se is less mobile than S (Prichard et al., 2013), postmagmatic alteration of BMS should create secondary mineral assemblages. For example, alteration with no sulfur loss would create an assemblage of millerite, magnetite, pyrite, violarite and/or cubanite, or abundant magnetite with rare millerite, violarite or cubanite if sulfur is lost from the system (e.g. Holwell et al., 2017;Prichard et al., 2013;Ripley et al., 2005;Smith et al., 2016). These secondary mineral assemblages were not observed in the Volspruit Sulphide Zone.
Therefore, we think the data in this study best supports the model for S-addition and subsequent Se-and S-degassing during assimilation as an explanation for the high S/Se values.

Origin of the Volspruit Sulphide Zone
Here, we summarise the paragenetic sequence of events leading to the creation of the Volspruit Sulphide Zone (Figure 15).
Insert Figure 15 4.6.1. Source magma: Cumulates from the Volspruit subzone cannot have formed from a B1 magma under reasonable conditions (Yudovskaya et al., 2013), as they contain relatively unevolved compositions: up to 47.52 wt% Cr2O3 in chromite, up to En90 in orthopyroxene and up to Fo90 in olivine. Yudovskaya et al. (2013) propose that chills from the Basal Ultramafic Sequence (Wilson, 2015), or analyses of B1-UM chills CD-005 and ECBV050 (Barnes et al., 2010) would be suitable candidates for parental magmas. Of these, PGE analyses of the B1-UM chills CD-005 are most appropriate. Contrary to Hulbert and von Gruenewaldt's (1982) model, we found no evidence to support the contribution of a more evolved B1, B2, or B3 magma (Barnes et al., 2010), a "denser, cooler and less primitive basaltic liquid" (pp. 306, Hulbert, 1983), or northern limb Main Zone-style magma .
Given the Pt/Pd of suitable parental magmas is 1.08-2.13 (Barnes et al., 2010), while the Pt/Pd of the modelled PGE resources in the northern limb are consistently <1.0 (Table 1), it is likely that the Lower Zone of the northern limb had a separate ultramafic source with distinct highly siderophile element chemistry from magmas in the eastern limb of the Bushveld Complex. While the B1 magma contains 33 ppb Pt+Pd, the ultramafic chill B1-UM CD-005 contains 50 ppb Pt+Pd (Barnes et al., 2010). Therefore, less upgrading is required to form a viable ore deposit.

Magma intruded assimilant(s) beneath the Rustenburg Layered Suite
We observed sulphide inclusions, silicate inclusions and trapped crustal assimilants within chromite. As these inclusions typically occur in the centre of chromite grains, we infer that inclusions were trapped soon after chromite crystals nucleated in the parent magma. Thus, these trapped inclusions record the earliest magmatic processes during the formation of the Volspruit Sulphide Zone. As these chromite inclusions (1) occur in the centre of chromite grains, (2) do not match the composition of minerals typically observed in the Volspruit subzone, and (3) contain dolomite, albite, hydrous phases, monazite and Pb-chlorides -we consider that these inclusions record the earliest history of magmatic emplacement into sedimentary rocks in a staging chamber/sill beneath the main Grasvally magma chamber (i.e. beneath the Rustenburg Layered Suite).
We envisage that primitive, ultramafic magmas intruded along a sedimentary contact beneath the main Grasvally magma chamber. The record of assimilation from microxenocrysts (e.g., calcium carbonate, U-Th oxide, Mn-rich ilmenite), exotic chromite inclusions (e.g., dolomite, albite, monazite) and Pb and Zn-rich magmatic sulphide assemblages indicate that limestone and dolomite were assimilated along at least one margin of this contact. The most voluminous local source of calcite and dolomite is the Malmani dolomite, which is also a documented host of Pb-Zn Mississippi Valley-type deposits (Martini et al., 1995). As the Malmani dolomite rests above the Black Reef Quartzite, we postulate that this contact may be the location of the sub-Grasvally staging chamber.

Assimilation of oxidised sediments triggered chromite saturation
Assimilation of carbonates with a minor contribution from other sediments locally increased the fO2 of the ultramafic magma and triggered chromite saturation, evidenced by the presence of the exotic chromite inclusions. Assimilation of carbonates may be accompanied by extensive volatilisation of CO2, released from assimilants and adjacent country rock, creating a turbulent magmatic environment. Despite a potentially turbulent environment, the heterogeneity of assimilants in the magma could still create sharp gradients in fO2, and thus fronts of chromite crystallisation in the magma.

Chromite crystallisation and addition of external sulfur triggered sulphide saturation
As sulphide inclusions are not consistently observed in chromites, we propose that chromite saturation preceded sulphide saturation, but there was at least one period where both processes were simultaneous. It is likely that saturation of an immiscible sulphide liquid was triggered by (1) addition of external sulfur from sedimentary contaminants, and (2) a decrease in Fe following chromite crystallisation (lowering the amount of sulfur required to achieve sulphide saturation). It is possible that sulphide saturation was triggered by a sudden influx of sedimentary sulfur, such as assimilating pre-existing Pb-Zn Mississippi Valley-type mineralisation, basinal brines, or a large quantity of sedimentary sulfur. Continued carbonate assimilation would create sharp fO2 gradients surrounding large xenoliths and roof pendants, accompanied by degassing of S and probably Se. During this process, remaining immiscible sulphide liquid would be stirred amongst silicate liquid, at a ratio of <3000:1. Periods of quiescence with less assimilation and magma replenishment would allow cumulates to settle, forming chains of chromite and sintered chromite textures.

Emplacement of crystal-rich slurries into the main Grasvally magma chamber.
A turbulent pulse of magma or a sudden fault rupture moves crystal-rich slurries containing entrained sulphide droplets, chromite crystals and pyroxene grains out of the staging chamber and into the growing Volspruit subzone in the main Grasvally magma chamber. The chromite-sulphide-silicate slurry is emplaced as a stratiform horizon -the full extent of this layer remains unknown -it is unlikely to be as contiguous as the Merensky Reef. We estimate that the Volspruit Sulphide Zone is restricted to a Grasvally sub-chamber ~60 m thick and continuous ~3 km along strike ( Figure 1C). During and following this event, emplacement of the Volspruit subzone continues.

Crystallisation of the Volspruit Sulphide Zone
There is very little plagioclase present in the Volspruit subzone, so the volume of trapped intercumulus melt is very small. Most of the interstitial liquid must have been squeezed out from the weight of overlying cumulates, and mixed with magmas above it. As the cumulates cooled, interstitial sulphide melt droplets fractionated, exsolving a copper-rich liquid, crystallising Fe-Ni-rich monosulphide solution and Cu-rich intermediate solid solution, before crystallising the observed base metal sulphide and precious metal mineral assemblages (e.g. Holwell and McDonald, 2010;Naldrett, 2004).

Post-cumulus serpentinisation of Volspruit subzone -fluid flow during reheating?
The metamorphic aureole surrounding the northern limb of the Bushveld Complex records two stages of contact metamorphism. These are attributed to (1) emplacement of the Lower Zone and (2) emplacement of layered rocks above Lower Zone cumulates (Nell, 1985). McDonald et al. (2005) contend that emplacement of the GNPA and overlying magmatic stratigraphy occurred while there was still melt replenishment occuring in the Grasvally (Lower Zone) magma chamber, but replenishment of Lower Zone magmas north of Grasvally had ceased. However van der Merwe, (2008) assert that there was a major hiatus between emplacement of Lower Zone and Main Zone magmas. The data from this study can neither confirm nor deny these hypotheses.
If van der Merwe (2008)'s hypothesis is correct, alteration and serpentinisation of the Volspruit deposit could have occurred during the emplacement of overlying magmas. However, it is most likely that alteration and serpentinisation of the Volspruit cumulates occurred during low-grade regional metamorphism associated with gentle folding and/or one of the four episodes of faulting recorded in the region (Hulbert, 1983;van der Merwe, 2008). We postulate that the regional metamorphism must have been low grade, as we observed no evidence for melting and remobilisation of sulphides. However, it is difficult to provide evidence for remelting of magmatic sulphide assemblages, as magmatic assemblages are texturally similar.
Serpentinisation affected base metal sulphides more or less equally -precious metal minerals are the only phases that are not altered.

Would we expect another Volspruit-style deposit in the Bushveld Complex?
To date, there are no other ore deposits in the Bushveld Complex similar to the Volspruit Sulphide Zone, implying that these ore-forming processes are unique. We propose that assimilation of a carbonate unit containing significant sulfur and subsequent crystallisation of chromite was crucial to the formation of an immiscible sulphide liquid in the initial sulphideundersaturated ultramafic magma. The presence of Pb-sulphides, Pb-chlorides and sphalerite suggests that this magma may have assimilated pre-existing Pb-Zn mineralisation or basinal brines. We consider that the Malmani Dolomite is the most likely assimilant. As the Malmani Dolomite extends across the Kaapvaal craton, it is possible that other staging chambers could have assimilated this horizon, but the resulting mineralisation in the main Bushveld chamber is likely to be localised and not continuous across great distances.
It is possible that the occurrence of even sub-economic Volspruit-style mineralisation in ultramafic cumulates could reduce the grade and opportunity for PGE mineralisation events higher up in the local magmatic stratigraphy of a layered intrusion (c.f., Latypov et al., 2017), or in or in subsequent magma emplacement events sourced from the same reservoir (c.f., Mungall et al., 2016). However, given the low R-factors required to form Volspruit-style mineralisation, the presence of this type of mineralisation does not necessarily preclude further opportunities for creating economically viable concentrations of PGEs.

Conclusion
This study demonstrates that qualitative petrological evidence such as detailed inclusion and microxenocryst studies provide additional lines of evidence for assimilation, complementing existing quantitative geochemical proxies such as the S/Se ratio and sulfur isotope compositions. Identification of microxenocrysts is particularly useful when the endmember compositions of assimilants is not known. These new qualitative techniques could be applied beyond layered intrusion research, to identify the range of crustal contaminants in other magmatic systems where macro-xenoliths are neither sampled nor preserved.
We consider that the Volspruit Sulphide Zone formed as a result of the following processes: 1. Parental magmas are derived from a fertile ultramafic source, similar to B1-UM in composition.   to scale, based on stratigraphic summaries of Joubert and Johnson (1998) and Kinnaird et al. (2017), with additional information from Eales and Cawthorn (1996), Kinnaird and McDonald (2005) and McDonald et al. (2016). Stratigraphic logs are modified from Hulbert (1983) with information from Kinnaird and McDonald (2005) and McDonald et al. (2016). Adjacent to the stratigraphic log, we present data on the mineralogy (modal percentage of minerals), the Mg# of orthopyroxene and the Mg# of olivine against stratigraphic depth in the Volspruit subzone, using data from Hulbert (1983).

The ultramafic source intruded into
The interval containing mineralisation, the Volspruit Sulphide Zone, has been highlighted in pink shading to aid visual correlation between the three plots.                 Zientek et al. (2014); **North Pit corresponds to the northern pit outline of the Volspruit Sulphide Zone shown in Figure 1; § bulk assay data provided by Pan Palladium Ltd.