The effects of iron-rich ultramafic pegmatite on the composition and mineralogy of the UG2 chromitite: a case study in the western Bushveld Complex, South Africa

Discordant iron-rich ultramafic pegmatites (IRUPs) intersect the UG2 chromitite at many places in the Bushveld Complex. The effects of IRUP interactions on the UG2 ore mineralogy and PGE grade are assessed at the Thaba mine, north-west Bushveld, based on a borehole profile through the UG2 layer and detailed analysis of mineral textures and compositional variations across the UG2-IRUP contacts using micro-XRF element mapping. The UG2-IRUP interaction operated at different scales and probably by different mechanisms. At the local scale (< 10 cm), a thin layer of Fe–Ti–Cr spinel and ilmenite formed on the IRUP side of the contact with UG2, whereas the UG2 chromitite developed grain coarsening, loss of interstitial silicates, and chemical gradients in Cr, Al, Fe, and Ti that extend a few centimeters from the contact into the seam. These local effects are attributed to the intrusion of IRUP melt into the solidified UG2 layer, followed by re-equilibration of the oxide minerals across the contact during cooling. On a larger scale, changes in the ore and gangue mineral assemblages in UG2 took place throughout the entire meter-thick main seam. Compared with regional UG2 compositions, chromite has higher TiO2 and lower Mg#, and there is an anomalously low abundance of interstitial plagioclase. The IRUP-affected UG2 shows relatively abundant secondary hydrous silicates, replacement of PGE sulfides by PGE alloys and PGE-As–Sb–Bi–Te–Pb phases, and formation of secondary Ni–Cu–Fe sulfides after pentlandite and chalcopyrite. These large-scale effects are attributed to hydrothermal fluids derived from IRUP melts. The IRUP bodies at the Thaba mine caused redistribution of PGE within the UG2 layer but did not significantly reduce the overall grade. However, significant changes in the ore mineral assemblage and an increased abundance of secondary silicates can reduce the efficiency of PGE recovery.


Introduction
Discordant bodies of iron-rich ultramafic pegmatite (IRUP) are widespread in the Rustenburg Layered Suite (RLS) of the Bushveld Complex. The name IRUP derives from the distinctly coarse grain size, the iron-rich nature of constituent mafic silicates, and the local abundance of Fe-Ti oxide minerals (Viljoen and Scoon 1985). Field observations and several geochemical-mineralogical studies have demonstrated that IRUP bodies formed from a mobile phase at high temperatures and that they partly replaced existing layered-suite rocks. There is still no consensus on the nature of IRUP-generating melts or fluids, nor on how and where they form (Viljoen and Scoon 1985;Scoon and Mitchell 1994, 2004Cawthorn et al. 2000Cawthorn et al. , 2018Reid and Basson 2002;Scoon et al. 2017). In this paper, we use the term IRUP melt but acknowledge that its nature is undefined. IRUP bodies occur in most levels of the Rustenburg Layered Suite above the Lower Zone, but they are particularly well known in the Upper Critical Zone where they intersect the Merensky Reef and UG2 platinum group element (PGE) orebodies. Several studies of the contact zones of IRUP Editorial handling: G. Beaudoin with Merensky Reef have shown that the interaction caused changes in the ore assemblages of the reef and some remobilization of metals (Viljoen and Scoon 1985;Kinloch 1982;Mitchell 1994, 2011;Viring and Cowell 1999;Reid and Basson 2002;Scoon and Eales 2002). The effects of IRUP's interaction with the UG2 chromitite have been less well studied. Penberthy and Merkle (1999) mentioned the effects of IRUP on the mineralization of the UG2 in a regional overview of the Bushveld Complex, but detailed case studies are lacking. A study by Mitchell et al. (2019a) described the contact zone of UG2 with IRUP from a platinum exploration project at Wilgerspruit in the western limb, but no information on PGE concentrations or mineralogy was reported.
This paper presents a case study of the mineralogical and geochemical effects of IRUP-UG2 interaction as revealed in a drillcore profile from the Thaba chromite mine in the northwestern Bushveld Complex. The borehole MD17 is located at the edge of one of the largest known IRUP bodies in the complex, the 1.5-km-diameter Middellaagte body. This was described as a pipe by Viljoen and Scoon (1985) and is incorrectly marked on Thaba mine records as "dunite pipe," but it has since been characterized as disc-shaped (Scoon et al. 2017). The 645-m borehole contains many IRUP intersections, which we interpret as apophyses of the main body. These include one about 6-m thick at the footwall of the UG2 chromitite and two cm-thick IRUPs within the main seam, which are the subject of this study. We present electron microprobe analyses of the major oxide and silicate phases in the IRUPs, the UG2 layer, and the hanging wall pyroxenite, as well as quantitative ore mineralogy by mineral liberation analysis (MLA) and PGE distribution from a series of samples through the UG2 chromitite. Detailed textural and compositional profiles across the contacts of all three IRUP bodies with UG2 were obtained by micro-Xray fluorescence (micro-XRF) element mapping. These data are used to document the variations in PGE geochemistry and mineralogy of the IRUP-effected UG2 chromitite and to understand how IRUPs interact with UG2.

Geological setting
The Bushveld Complex, summarized by Cawthorn (2015) and Viljoen (2016), comprises the approximately 8-km-thick Rustenburg Layered Series (RLS) of mafic to ultramafic rocks, the overlying 3.5-km-thick volcanic and subvolcanic roof rocks known as the Rooiberg Group, and granitic intrusions of the Lebowa Suite that were emplaced between the layered series and roof rocks. Radiometric dating suggests that all igneous components of the Bushveld Complex, including the felsic units at the roof were emplaced over a period of ~ 5 Ma from 2060 to 2055 Ma (Scoates et al. 2021). The focus of this paper is on the Critical Zone of the RLS, which hosts world-class PGE and Cr deposits. The Critical Zone contains about a dozen layers of chromitite, each on the order of tens of centimeters to 1-m thick, and many more centimeter-thick or thinner chromitite stringers. The chromitite layers are grouped into a lower (LG), middle (MG), and upper group (UG), each of which contains several layers numbered consecutively upwards from the base (inset in Fig. 1). The UG2 is stratigraphically the second-highest chromitite layer of the upper group. There has been abundant research on the geology and mineralogy of the UG2 and other chromitite layers since the 1970s (Von Gruenewaldt et al. 1986;Naldrett et al. 2012), but debate continues on how they formed (Kinnaird et al. 2002;Voordouw et al. 2009;Latypov et al. 2017Latypov et al. , 2018Cawthorn et al. 2018;Mitchell et al. 2019b;Veksler and Hou 2020;Zhou et al. 2021).
Discordant IRUP bodies in the RLS are widespread on the eastern and western limbs of the Bushveld Complex, occurring in the Critical, Main, and Upper Zones. The IRUP bodies vary considerably in composition and form, depending partly on the stratigraphic interval in which they occur (e.g., Scoon and Mitchell 1994). The rocks are dominantly made up of Fe-rich clinopyroxene and olivine, with variable amounts of Fe-Ti oxides and accessory sulfide minerals occurring locally. Typical features that set IRUPs apart from their host-rock pyroxenites and norites are as follows: (1) a coarser grain size, (2) higher Fe/(Fe + Mg) ratios of ferromagnesian minerals, (3) a paucity of plagioclase and orthopyroxene, and (4) a higher abundance of hydrous phases (accessory amphibole and mica). Field observations demonstrate a spatial association of IRUP with zones of structural disturbance in the RLS (i.e., faults, dome structures), highly variable shapes of the IRUP bodies, and a differential replacement of the layered series rocks by IRUP, which is especially strong for plagioclase-rich lithologies and weak for chromitite and ultramafic layers (Cameron and Desborough 1964;Viljoen and Scoon 1985;Reid and Basson 2002). Mineral compositions of pyroxene and plagioclase (Scoon and Mitchell 1994), as well as O-isotope exchange thermometry (Cawthorn et al. 2000), attest to high temperatures of formation and a magmatic affinity, but the rock composition represents mineral accumulation, not quenched magma. Initial Sr isotope ratios of IRUP in the Critical Zone contrast with those of the host rocks but resemble those of the Upper Zone (Scoon and Mitchell 1994;Reid and Basson 2002). This, and the similar Ti-magnetite compositions in IRUP and in magnetite layers of the Upper Zone suggest that IRUP may have derived from residual or exsolved melts from the Upper Zone, an idea supported by field evidence for downward-directed emplacement of IRUP (Scoon and Mitchell 2021). On the other hand, Cawthorn et al. (2018) argued for an external source of the IRUP melts, so many fundamental questions about the origin of IRUPs remain unanswered.

Geology of the Thaba mine and borehole MD17
The Thaba mine in the northwestern Bushveld Complex produces chromite from the chromitite layers LG-6, LG-6A, and MG-1 to MG-4. This rock package dips southeast at 15° to 27° and is cut by steep, NW-SE trending normal faults (Fig. 2). Mafic dikes with the same orientation as the faults also occur. Information about the mineralogy and ore grades of the LG and MG chromitite layers at Thaba, including their PGE concentrations, is given by Bachmann et al. (2018) and Kaufmann et al. (2019). The Thaba mining lease is adjacent and parallel to the Amandelbult section (Dishaba and Tumela platinum mines), which exploits the UG2 and the Merensky Reef. South-east of the Amandelbult section is the Northam platinum mine, which also exploits UG2 and Merensky Reef ores (Fig. 1). The UG2 chromitite is generally not exposed in the Thaba mine, with the exception of a graben that is also the locus of the Middellaagte IRUP body. Many smaller IRUPs in the vicinity of the Middellaagte body are known from outcrops and magnetic surveys (Viljoen and Scoon 1985). Borehole logs from the Thaba mine compiled by Bachmann et al. (2019) show that IRUP intersections are widespread in the stratigraphic interval of the LG and MG series chromitites ( Fig. 2A). Viljoen and Scoon (1985) and Scoon and Mitchell (1994) described the geology of the Middellaagte body and smaller IRUPs in the region in detail, including observations from underground exposures in the Amandelbult mine to the west of the main body. They also provided bulk-rock and mineral geochemical data from IRUP and host rocks and discussed the possible origin of IRUP, including comparisons with occurrences in the eastern limb of the Bushveld Complex. Three clusters of IRUPs were recognized by Viljoen and Scoon (1985), all of which are related to NW-SE trending faults. The Middellaagte body, with a 1.5-km-diameter subcircular outcrop pattern, is located in a graben ( Fig. 2A) with about 500 m of vertical offset. Underground exposures in the Amandelbult mine demonstrate that IRUP extensively replaced pre-existing rocks of the layered suite, especially those rich in plagioclase (Fig. 2B). In places, IRUP has largely replaced the Merensky Reef feldspathic pyroxenite, leaving the chromitite stringers partly intact (Scoon and Mitchell 2011). An interesting feature of the replaced Merensky Reef described by Scoon and Eales (2002) is the transformation of chromite to an intermediate Fe-Ti-Cr spinel and the local thickening of relict chromitite stringers by the formation of a new, oxide-rich layer. Similar oxide-rich Fig. 1 Map of the NW sector of the Bushveld Complex with the Amandelbult, Union, and Pilanesberg "enclaves" separated by gap areas (modified from Mitchell et al. 2019a). Note the association of IRUP bodies with faults (dashed lines) and the location of the Middellaagte graben and IRUP study area. Insets show the location relative to the Bushveld Complex (modified from Kinnaird et al. 2005) and simplified stratigraphy of chromitite layers in the Critical Zone of the Rustenburg Layered Suite (MR, Merensky Reef; UG, Upper Group; MG, Middle Group; LG, Lower Group) contact zones between IRUP and chromitite have been found elsewhere in the Bushveld Complex (e.g., Cameron and Glover 1973;Mitchell et al. 2019a), and this is also a prominent feature of the IRUP-UG2 contact at Thaba mine as described below.

Petrography of the UG2 and IRUP in MD17
The MD17 borehole was drilled to a depth of 645 m and intersects Critical Zone chromitite layers from UG2 at 143 m below the collar to LG1 at 633 m. Core logs show that the upper 130 m of MD17 consists almost entirely of IRUP, interrupted by a few intervals of pyroxenite and norite 2 to 5 m thick. Thinner sheets of IRUP occur at and below the UG2, which we interpret as apophyses of the Middellaagte body.
The drillcore profile investigated in this study comprises a 2.3-m interval across the UG2 chromitite, extending from the pyroxenite hanging wall through two chromitite leaders and the main seam (1.5-m thick), and ending in the IRUP footwall (Fig. 3). The core interval shows that the footwall IRUP is 6.5-m thick (referred to herein as IRUP-1). Two additional, thin IRUP apophyses (IRUP-2 and IRUP-3, about 4-cm thick each) occur within the main UG2 seam.
The footwall IRUP-1 is made up of approximately 40 vol.% clinopyroxene (up to 20 mm), 40% olivine (up to 10 mm), 10% Ti-Fe oxides, 6% hornblende, 2% sulfides, and 2% combined orthopyroxene, plagioclase, and phlogopite. Hornblende occurs as independent grains (up to 5 mm) and locally as inclusions (50-500 μm) in clinopyroxene. Euhedral phlogopite is present as inclusions in hornblende. Plagioclase (1-2 mm), sulfide, and oxide minerals are sparsely disseminated in the interstitial network. The oxide assemblage comprises titanomagnetite and ilmenite with Cr-Fe-Ti spinel forming at the UG2 contact. The sulfide minerals are dominated by pyrrhotite, with minor pentlandite, chalcopyrite, and cubanite. Similar sulfide mineral assemblages have been reported in the IRUP from the Middellaagte body and elsewhere in the Upper Critical Zone Mitchell 1994, 2011;Reid and Basson 2002).
The IRUP-2 and IRUP-3 bodies occur within the main chromitite seam at 37 and 84 cm above its base, respectively. They mainly consist of olivine and clinopyroxene, in modal proportions of 35% vs. 65%, respectively, in IRUP-2 and 75% vs. 25% in IRUP-3. Accessory sulfide and oxide minerals have similar proportions and characteristics as described for IRUP-1.
Two chromitite "leaders" (20 and 15 cm in thickness) occur in the hanging wall pyroxenite. Both are characterized by a poikilitic texture with orthopyroxene oikocrysts. Mitchell et al. (2019a) also noted a poikilitic texture in the UG2 leaders at Winnarshoek. Only the upper leader in MD17 was analyzed for quantitative mineralogy, and the results are similar to those of poikilitic samples from the main seam.
Finally, the pyroxenite hanging wall is made up mainly of orthopyroxene, with interstitial plagioclase and minor hornblende, disseminated chromite, and olivine.

Sampling and analytical methods
The drillcore through UG2 was cut into 60 segments of 3-5-cm length, numbered upwards from MD17-01 to MD17-60. Thin sections for the petrographic study were prepared from all segments, and a subset of these ( Fig. 3) was chosen for analysis.
Whole-rock PGE analyses were performed by fire assay with nickel sulfide collection at Mintek Analytical Services Division in Johannesburg, South Africa. After crushing and milling, the samples were fused with nickel sulfide, the sulfide bead was dissolved in aqua regia, and the solution was analyzed for five PGEs (Pt, Pd, Ir, Rh, Ru) and Au using inductively coupled plasma-optical emission spectrometry (ICP-OES). A minimum of 40 g of sample mass was needed for each analysis, and this required combining two of the 3-5-cm drillcore segments as one assay sample. Due to the small thickness of IRUP layers 2 and 3, only the footwall IRUP-1 could be analyzed for PGE. Four samples from the UG2 chromitite were analyzed (UG2-1 to UG2-4), one from the upper chromitite leader (UG2-5), and one from the pyroxenite hanging wall (Fig. 3). Reference Material SARM-64 was analyzed with the samples to monitor analytical accuracy, and the results are included with the UG2 analyses in Table 1.
MLA was used to determine the identity and distribution of ore and gangue minerals in polished thin sections. The MLA instrumentation at the Helmholtz Institute Freiberg for Resource Technology (Bachmann et al. 2017) combines an FEI Quanta 650F scanning electron microscope (SEM) with two Bruker Quantax X-Flash 5030 energy-dispersive X-ray spectrometers (EDS). Automated data acquisition and analysis were controlled by the MLA 3.1.4 software suite using the sparse phase liberation (SPL-Lt) mode (Fandrich et al. 2007). The MLA study was done on 14 polished thin sections, 10 of which correspond to the sections used for whole-rock PGE analyses of the main seam and upper leader. An additional four samples were chosen to include the base of UG2 above the footwall IRUP-1 (MD17-04, MD17-07) and the middle of UG2 below IRUP-3 (MD17-23, MD17-24). The position of the thin sections studied and their correspondence with the fire-assay samples are shown in Fig. 3.
Electron probe microanalysis (EPMA) on polished, carbon-coated thin sections was carried out using a JEOL super probe JXA-8230 instrument at the German Research Centre for Geosciences (GFZ), Potsdam. An accelerated voltage of 15 kV, a beam current of 20 nA, and beam sizes of 3 μm (for Fe-Ti-Cr oxides), 5 μm (for olivine, pyroxenes, hornblende, and phlogopite), and 15 μm (for plagioclase) were used for measurement. Analytical details about the standards, X-ray lines used, counting time on peak and background, and detection limits are listed in Zhou et al. (2023).
Micro-XRF mapping of element distribution across the contact zones between IRUP and chromitite was obtained by a Bruker Tornado M4 micro-XRF instrument at GFZ Potsdam. The instrument is equipped with an Rh X-ray source that operates at 50 kV and 600 µA and with poly-capillary X-ray optics that irradiates a spot of 20 µm for 50 ms. Dual 30 mm 2 Si-drift detectors were used for spectra acquisition. Element intensities were normalized by an initial spectrum deconvolution and visualized by relative element abundance in 2D maps with a spatial resolution of 50 µm.

Whole-rock analyses
The PGE concentrations of IRUP, UG2 chromitite, and hanging wall pyroxenite are listed in Table 1. A summary of total 5PGE + Au concentrations (Pt, Pd, Ir, Rh, Ru, Au), and the distribution of individual elements and element ratios in the profile are shown in Fig. 4A. The values are compared with the average UG2 from Amandelbult (McLaren and De Villiers 1982;Von Gruenewaldt et al. 1986) and with a regional UG2 average (Barnes and Maier 2002) on a chondrite-normalized plot in Fig. 4B. Values of the 5PGE + Au in the UG2 main seam are between 2.8 and 11.4 ppm, with an average of 5.8 ± 3.4 ppm (n = 5). The highest value occurs near the top (sample UG2-4, with 11.4 ppm), which is twice that of the next-highest value of 5.9 ppm in sample UG2-1 from the base. The upper leader (sample UG2-5) yielded a value of 5.3 ppm. The values in the footwall IRUP-1 and the hanging wall pyroxenite are 0.16 and 0.26 ppm, respectively.
Platinum-group element ratios also vary with sample height in the UG2 profile (Fig. 4A). The Pt/Pd ratios are between 1.8 and 2.5 in all but one UG2 sample from MD17. These values agree with the average Pt/Pd ratio of 2.5 for UG2 in the western Bushveld according to McLaren and Villiers (1982) and 1.8 for the Amandelbult mine (Von Gruenewaldt et al. 1986). However, the PGE-rich sample UG2-4 from the top of the main seam has a high Pt/Pd ratio of 3.9. Another discriminant variable is the ratio of "palladium group elements" (PPGE: Pd + Pt + Rh) to "iridium-group elements" (IPGE: Ir + Os + Ru). Our 5PGE + Au assay data do not include Os, but data from the adjacent Amandelbult mine show that Os accounts for just 11 wt.% of the IPGE sum (Von Gruenewaldt et al. 1986), so its absence should not affect the PPGE/IPGE ratio by much. The PPGE/IPGE ratio in the Amandelbult UG2 averages 4.2 (Von Gruenewaldt et al. 1986). In comparison, the MD17 samples have considerably lower PPGE/IPGE ratios in the bottom and middle  Fig. 4B).

Quantitative mineralogy
The MLA results of ore and gangue mineral assemblages are summarized in a bar chart in Fig. 5 in terms of the PGE, BMS, and gangue mineral groups. The MLA data are reported in Zhou et al. (2023). The typical PGE mineral assemblage of UG2 ore regionally is dominated (> 80 modal %) by sulfide minerals, comprising Pt and Pd sulfides, laurite (RuS 2 ), and PGE-Cu sulfides (Kinloch 1982;Gain 1985;Penberthy et al. 2000;Voordouw et al. 2010;Fig. 5 In this sense, the leader also stands in stark contrast to "normal" UG2 from the Amandelbult mine and elsewhere, but the assay values of PGE concentration and ratios (sample UG2-5) are unaffected by this change in the mineral assemblage. The distribution of base-metal sulfide (BMS) minerals in the MD17 drillcore interval also varies within the UG2 seam as well as between it and the upper leader. Pentlandite and chalcopyrite dominate the BMS assemblage in UG2 regionally (McLaren and Villiers 1982;Voordouw et al. 2010;Langa et al. 2021), and this is the case for most samples of the main seam and upper leader of MD17, but the middle part of the main seam (samples MD17-23, MD17-24, MD17-30 and MD17-31), near IRUP-2 and IRUP-3, shows a local abundance of millerite, violarite, and secondary CuFeS phases at the expense of pentlandite and chalcopyrite. Notably, these are the same samples with the highest proportion of PGE sulfarsenide and Sb-Bi-Te-Pb-bearing phases in the main seam, as well as abundant alteration silicates and amphibole at the expense of pyroxene.
It is noted that the lowermost sample, MD17-04, is mostly within the chromitite but contains the IRUP contact at its base.  (Barnes and Maier 2002). Chondrite values from McDonough and Sun (1995) Because of the negligible PGE contents in IRUP-1 (0.16 ppm; Table 1), the PGE mineralogy of this sample represents that of the chromitite layer alone, but the abundance and proportion of BMS and silicate phases reflect a mixture of both rock types and cannot be usefully interpreted.

Plagioclase
Plagioclase is rare in IRUP generally (e.g., Viljoen and Scoon 1985), although Reid and Basson (2002) found ubiquitous interstitial plagioclase in IRUP from the Northam mine, with up to 15 vol.% locally. In our samples plagioclase was found only in the footwall IRUP-1. Plagioclase in this sample is surrounded by clinopyroxene and likely represents a relict phase. This is supported by the fact that the average anorthite component (An) in this sample is 80 ± 3 (n = 8), essentially the same as in the hanging wall pyroxenite (An = 83 ± 1, n = 5) and significantly lower than reported from other IRUP bodies nearby. Viljoen and Scoon (1985), Scoon and Mitchell (1994), and Reid and Basson (2002) found An values of 90 and above in plagioclase from the Middellaagte IRUP and in IRUP from Northam, respectively.

Fe-Ti-Cr oxides
Ilmenite was found only in the IRUP bodies, not in UG2, and its composition is similar throughout except that MgO contents are lower in footwall IRUP-1 (average 1.6 wt.%) than in IRUP-2 and IRUP-3 that intrude the main seam (2.5 and 2.8 wt.% MgO, respectively). The contents of Cr 2 O 3 in ilmenite are uniformly low (0.1-0.2 wt.%) in all samples, independent of their location.
The comparison of the chromite Mg# and TiO 2 contents in UG2 from this study and from other localities in the complex (Fig. 8) shows that TiO 2 concentrations in the main seam and upper leader at Thaba mine reach values typical for UG2 regionally, but the Mg# is 5 to 10% lower than typical UG2. We interpret the pervasively low Mg# in chromite as  Fig. 7). Comparisons are shown for IRUP-unaffected UG2 in the Khuseleka (western Bushveld; Veksler et al. 2018) and the Middelpunt mines (eastern Bushveld; Mathez and Mey 2005). The star represents the average UG2 value from McLaren and De Villiers (1982) an effect of IRUP interaction, as noted above for the interstitial olivine, pyroxene, hornblende, and phlogopite in UG2.

Element mapping of the UG2-IRUP contact
Back-scatter electron (BSE) micrographs and micro-XRF element maps of thin sections across the contacts of all three IRUP bodies with UG2 show important details of their interaction. The precise original contact is not easy to locate because of a silicate-free zone of Fe-Ti oxide minerals (ilmenite and Fe-Ti-Cr spinel) that appears to grade into the chromitite of UG2 (Cameron and Glover 1973;Scoon and Eales 2002). Since ilmenite is not known to form in UG2 except for rare, fine-grained exsolutions in chromite (e.g., Junge et al. 2014), we defined the original contact at the limit of abundant ilmenite grains, which is a few millimeters beyond the limit of the IRUP silicate minerals. Figure 9 shows element maps and electron microprobe profiles of spinel composition across the UG2 contact with IRUP-1. Corresponding profiles across the IRUP-2 and IRUP-3 contacts are presented in Zhou et al. (2023). All contacts examined show similar features.
The IRUP side is characterized by ilmenite and Fe-Ti-Cr spinel intergrowths, whereby the spinel phase shows strong variations in Cr, Ti, Fe, and Mg# (Figs. 7 and 8). In detail, there are complex microtextural relations between the newly formed ilmenite and Fe-Ti-Cr spinel in this replacement layer (Fig. 10). Locally, the two phases appear to form symplectic intergrowths, and ilmenite is commonly rimmed by These dark rims are 1-10 μm thick and too small for quantitative analysis, but EPMA element mapping shows that they have lower Fe and Ti and higher Cr and Mg contents than the associated Fe-Ti-Cr spinel grains, while qualitative analysis by EDS indicates that Al is also a major constituent. Cameron and Glover (1973) and Scoon and Eales (2002) described similar microtextures in the oxide layer formed in IRUP at contact with chromitite. The latter study attributed the symplectic intergrowths of ilmenite and Fe-Ti-Cr spinel to coarsening of an exsolution texture and also identified the dark rims on ilmenite as Al-spinel.
The UG2 side of the contact is characterized by the larger grain size of chromitite grains, the formation of a lesser amount of Fe-Ti-Cr spinel, and the loss of interstitial silicates and sulfide minerals, which is particularly visible in the BSE images and micro-XRF maps of Mg and S distribution (Figs. 9B and 10A). Also, there are smooth concentration gradients in Cr, Al, Fe, and Ti within the chromitite that are parallel to the contact (Fig. 9B). The element maps suggest that the gradients extend a few millimeters from the contact, but the EPMA profiles demonstrate that the chemical effects of IRUP interaction extend into the chromitite for up to about 5 cm beyond the contact (Fig. 9A). Similar concentration gradients of Cr and Ti in chromite were documented by Mitchell et al. (2019a) by micro-XRF scanning of the UG2-IRUP contact at the Wilgerspruit locality to the SW of our study area (Fig. 1). Mitchell et al. (2019a) also described coarsening of chromite grains and loss of interstitial phases ("sintering" in their terminology). These centimeter-scale effects are referred to as "local," but it is important to note that throughout most of the main UG2 seam in drillcore MD17, chromite compositions show higher TiO 2 contents than normal for UG2 (< 2 wt.% as shown in Fig. 7).

Transformation of chromite to Fe-Ti-Cr spinel
Several studies in the Bushveld Complex noted the presence of unusual spinel compositions in places where IRUP intrudes chromitite and interpreted them as recrystallization or replacement of former chromite grains (Cameron and Glover 1973;Viljoen and Scoon 1985;Scoon and Mitchell 1994;Scoon and Eales 2002). This is also the case in our study, as shown by the textural changes and concentration gradients in UG2 chromitite (Fig. 9). Importantly, Fe-Ti-Cr spinel formed on the IRUP side of the contract where chromite was absent, and the modal amount of oxide minerals in this zone is much greater than can be explained by replacement of Ti-magnetite and ilmenite in normal IRUP. Furthermore, an oxide layer formed at UG2 contacts in all three IRUP bodies in MD17, regardless of their thickness ( Fig. 10; Zhou et al. 2023). We suggest that the oxide layer crystallized in situ from the IRUP melt. Possibly, as suggested by Scoon and Eales (2002), the refractory chromitite layers acted to channel and pool the IRUP melt at their contacts, where Fe-Ti-Cr spinel and ilmenite crystallized at the chromitite interface. Subsequent cooling caused recrystallization and/or equilibration within and across the oxide layers, resulting in exsolution textures, grain coarsening, chemical gradients in chromitite, and the "unusual" Fe-Ti-Cr spinel compositions that seem unique to the chromitite-IRUP contact zones in Bushveld. Indeed, the range of spinel compositions reported by Scoon and Eales (2002) coincides remarkably with our results from borehole MD17 (Fig. 7), and they also reported similar spinel-ilmenite intergrowths in association with Fe-Ti-Cr spinel compositions at chromitite contacts in the Merensky Reef as found in our study of UG2 (Fig. 10C).

The effects of IRUP on UG2 mineralization
Our MLA results and whole-rock assay values from the UG2 profile in borehole MD17, in comparison with IRUPfree localities, show significant changes in ore mineralogy (Fig. 5) and redistribution of PGE within the layer, which we attribute to the effects of IRUP interaction. However, the overall grade of UG2 appears not to have been affected. Thus, the average 5PGE + Au concentration in 5 samples from the main seam is 5.9 ppm, which is similar to that of the upper leader (5.3 ppm) and close to the 6.5 ppm reported from the nearby Amandelbult mine (Von Gruenewaldt et al. 1986). Previous work indicates a range of PGE grades from 4 to 10 ppm for UG2 regionally (Mclaren and De Villiers 1982;Gain 1985;Von Gruenewaldt et al. 1986;Penberthy and Merkle 1999;Cawthorn et al. 2002;Maier and Barnes 2008;Voordouw et al. 2009;Junge et al. 2014;Osbahr et al. 2014). On the other hand, a typical feature of UG2 in the Bushveld Complex, including Amandelbult, is a concentration of PGE in the lower part, where contents typically exceed 10 ppm (Fig. 11). This feature is missing at MD17 from the Thaba mine, where the basal sample UG2-1 yielded 5.9 ppm 5PGE + Au, about half of the maximum 11.4 ppm found near the top of the main seam. We cannot rule out that the base of UG2 at Thaba mine was originally less mineralized than in other localities, but it seems more likely that interaction with the footwall IRUP caused a loss of PGE from the base. This is supported by the textural and mineralogical changes at the IRUP-UG2 contacts described above and the MLA results that show correlated changes in the ore mineral assemblages (PGE and BMS minerals) as well as the interstitial silicate phases (Fig. 5).
The regional PGE mineralogy of UG2 is dominated by Pt and Pd sulfides, Ru sulfide (laurite), and PGE-Cu sulfides (McLaren and De Villiers 1982;Osbahr et al. 2014). Indeed, these phases make up 85% of the PGE assemblage at Amandelbult, according to Kinloch (1982). In contrast, the only abundant PGE sulfide in UG2 at Thaba mine is laurite, while most PGEs are contained in PGE alloys, sulfarsenides, and PGE-bearing Bi-Te-Sb phases. In this respect the Thaba UG2 resembles the IRUP-"affected" UG2 ore from the Brits-Marikana area, southwestern Bushveld (Penberthy and Merkle 1999;Penberthy et al. 2000;Fig. 5). The low proportion of primary PGE sulfide minerals near IRUP in MD17 can be caused by the grain coarsening and loss of interstitial phases in the chromitite (e.g., Fig. 9), as noted in other studies (Scoon and Eales 2002;Mitchell et al. 2019a). Interestingly, the proportion of laurite in "normal" UG2 vs. "affected" UG2 (related to IRUP, faults, and potholes) from comparative studies is about the same (20-30% of the PGE assemblage), and this is also true for the base of UG2 at Thaba mine (Fig. 5). The survival of laurite probably owes to its presence as inclusions in chromite (e.g., Kinloch 1982;McLaren and De Villiers 1982).
The vertical distribution of PGE, BMS, and gangue mineral assemblages within the UG2 layer (Fig. 5) shows correlations that shed light on how the layer was affected by IRUP. The middle part of the main seam, near IRUP-2 and IRUP-3, is enriched in interstitial amphibole and "alteration silicates" (chlorite, serpentine, talc) at the expense of pyroxene. This abundance of hydrous silicates in the chromitite interstitial network is a sign that UG2 was metasomatized by hydrothermal fluid. Also, there is an almost complete lack of interstitial plagioclase in the whole UG2 layer, whereas plagioclase is important or dominant in the interstitial silicates of "normal" UG2 (Mathez and Mey 2005;Voordouw et al. 2010;Veksler et al. 2018;Mitchell et al. 2019a). Destruction of plagioclase in rocks partially replaced by IRUP is common (Viljoen and Scoon 1985;Reid and Basson 2002), and we suspect that the anomalously low plagioclase abundance in UG2 from MD17 is an effect of IRUP interaction. Pentlandite and chalcopyrite are the dominant primary BMS minerals in normal UG2 elsewhere (Fig. 5), but samples near IRUP-2 and IRUP-3 contain very little chalcopyrite and significant amounts of secondary Ni-Fe sulfides (millerite, violarite) at the expense of pentlandite. We consider it significant that the part of the main seam showing an abundance of secondary Ni-FE and Cu-Fe sulfide minerals is also rich in hydrous silicates and in PGE sulfarsenides and Sb-Bi-Te-Pb phases (samples MD17-23 to MD17-41; Fig. 5). Although there is no IRUP intersection near the upper leader in borehole MD17, the PGE mineralogy there is dominated by secondary PGE sulfarsenides and Sb-Bi-Te-Pb phases. Considering the location of borehole MD17 with respect to the Middellaagte IRUP body and the abundance of smaller IRUPs in the region ( Fig. 2A), we can speculate that other IRUP bodies may be close by but not present in the drillcore. On the other hand, it must be noted that secondary ore assemblages and hydrous silicates in UG2 similar to those described here are found in places that are apparently free of IRUP but affected by faults and potholes (McLaren and De Villiers 1982;Kinloch and Peyerl 1990;Penberthy and Merkle 1999;Voordouw et al. 2010). What these settings have in common is that they enhanced the access of hydrothermal fluids to the UG2 layer. A comparison can also be drawn to the "UG2 equivalent" (UG2E) layer on the Bushveld northern limb described by Langa et al. (2021). Here, too, the PGE assemblages are dominated by alloys and As-Bi-Te phases, and the BMS assemblage contains significant amounts of pyrrhotite at the expense of the otherwise typical pentlandite-chalcopyrite association. IRUP bodies are unknown from this locality, and the authors suggested that the particular features of the UG2E chromitite were caused by the magmatic assimilation of Transvaal Group country rocks.
In summary, we distinguish two sets of phenomena caused by UG2-IRUP interaction at the Thaba mine, which operated at different scales. The small-scale effects are limited to the immediate IRUP-UG2 contact and include crystallization of a layer with Fe-Ti-Cr spinel intergrown with ilmenite on the IRUP side and coarsening, recrystallization, and chemical gradients within chromitite on the UG2 side. We propose that these effects were caused by the interaction of IRUP melt with the solidified UG2, possibly aided by ponding of the melt at the chromitite contact as envisioned by Scoon and Eales (2002). Grain coarsening and "sintering" during this process probably caused a loss of interstitial sulfides and PGE grade from the basal part of UG2, which is typically the richest zone in "normal" UG2. On a larger scale, the IRUP interaction affected most of the main seam and the upper leader. Evidence includes the anomalous chromite composition (elevated TiO 2 contents, lower Mg#), the unusually low abundance of interstitial plagioclase, the Fig. 11 Comparison of 5PGE + Au concentrations from the UG2 profile in the MD17 drillcore with other localities from the Bushveld Complex compiled by Maier and Barnes (2008). The typical PGE enrichment at the base of UG2 is poorly developed in MD17 formation of hydrous silicates and secondary PGE-and Ni-Cu-Fe sulfide minerals, and the lowering of the Mg# of the mafic silicates (e.g., phlogopite). We suggest that these large-scale effects were caused by hydrothermal fluids derived from IRUP melts.
From an economic standpoint, the IRUP interaction with UG2 at the Thaba mine did not significantly reduce the overall PGE grade, causing only a redistribution of grade within the layer. However, significant changes in the ore mineralogy of IRUP-affected UG2 (replacement of PGE sulfides by alloys and PGE-bearing As-Sb-Bi-Te-Pb phases) and in the gangue assemblage (secondary talc, serpentinite) will decrease the efficiency of PGE recovery when using conventional flotation techniques (Kinloch 1982;Peyerl 1982;Penberthy et al. 2000;Bachmann et al. 2018).

Conclusions
Three IRUP bodies intersect the UG2 chromitite in borehole MD17 from the Thaba mine in the northwestern Bushveld Complex: one at the footwall and two within the main chromitite seam. These IRUPs are interpreted as the apophyses of the large Middellaagte body, a disc-shaped IRUP some 1.5 km in diameter. The studied IRUPs consist of clinopyroxene (Mg# = 78-80) and olivine (Fo 54-58 ) with minor hornblende, ilmenite, Ti-magnetite, and accessory iron sulfides. All contacts of IRUP and UG2 in this study display an oxide layer a few millimeters thick that formed on the IRUP side and consists of intergrowths of ilmenite and compositionally diverse Fe-Ti-Cr spinel, with microtextures suggesting exsolution during cooling. On the UG2 side of the contact, there is an increase in the grain size of chromite with concomitant destruction of interstitial phases, including sulfide minerals. Additionally, the chromite composition shows contact-parallel gradients in Ti, Fe, Cr, and Al that extend a few centimeters into the UG2 layer. We attribute these features to the crystallization of an oxide layer from the IRUP melt onto the solidified chromitite interface, followed by re-equilibration and chemical exchange on both sides during cooling.
Quantitative mineral assemblage analysis by MLA and bulk-rock PGE assays show strong effects of IRUP interaction on the PGE mineralogy and redistribution of grade within the UG2 seam with little or no net loss. The average grade of 5PGE + Au (5.9 ppm) is similar to that of unaffected UG2 in the nearby Amandelbult mine and regionally. However, while typical UG2 ore is dominated by PGE sulfide minerals, pentlandite and chalcopyrite, the ore assemblage at Thaba mine is dominated by PGE alloys, sulfarsenides, and Sb-Bi-Te-Pb phases, and there is a significant proportion of secondary Ni-sulfide phases at the expense of pyroxene. These changes in the ore assemblage of UG2 are associated with an abundance of interstitial hornblende and secondary hydrous silicates at the expense of pyroxene. We attribute these meter-scale effects to hydrothermal fluids released from the IRUP melt.
Our results demonstrate that IRUP interaction with UG2 has not destroyed the overall PGE grade, but there is likely to be a significant loss of value because of the replacement of primary PGE sulfides by alloys and PGE-bearing As-Sb-Bi-Te-Pb minerals, along with the formation of alteration silicates. These will have a negative impact on the PGE recovery by flotation.