Chemical and Textural Re-equilibration in the UG2 Chromitite Layer of the Bushveld Complex, South Africa

Chemical and Textural Re-equilibration in the UG2 Chromitite Layer of the Bushveld Complex, South... Abstract Variations of mineral chemistry and whole-rock compositions were studied in detail, at millimetre to centimetre intervals, in two vertical drill core profiles through the platiniferous UG2 chromitite layer in the western and eastern limbs of the Bushveld Complex, South Africa. Analytical methods included electron microprobe and LA-ICP-MS analyses of the main rock-forming minerals, orthopyroxene, plagioclase and interstitial clinopyroxene. One profile was also studied by synchrotron-source XRF. Statistical analysis of crystal size distribution of chromite was also performed at different levels in the chromitite layer and in adjacent silicate rocks. The results provide new evidence for chemical and textural late magmatic re-equilibration in the UG2 layer and in the silicate rocks at the contact zones. The chromite crystal size distributions imply extensive coarsening of that mineral within the main chromitite seam, which has erased any textural evidence of primary deposition features such as recharge or mechanical sorting of crystals, if those features originally existed. The mineral compositions in chromitite differ from those in adjacent silicate rocks, in general agreement with predictions of chemical re-equilibration with evolved, residual melt (the trapped liquid shift effect). In detail, the geochemical data imply, however, that the conventional trapped liquid shift model has shortcomings, due to the effects of material transport driven by chemical gradients between modally contrasting layers of crystal mush undergoing re-equilibration reactions. In the presence of such gradients, selective open-system conditions may hold for alkalis and hydrogen because of their higher diffusion rates in silicate melts. Differential mobility of components in the interstitial melt can also sharpen the original modal layering by causing minerals to crystallise in one layer and dissolve in another. Detailed trace element profiles by synchrotron XRF reveal an uneven vertical distribution of incompatible elements which implies that the permeability of the chromitite layer may have been significant, even at the latest stages of interstitial crystallization. INTRODUCTION Layers of chromitite chiefly composed of Cr-rich spinel are among the most spectacular examples of mineral layering in mafic-ultramafic intrusions. The most prominent examples come from Precambrian layered intrusions such as Stillwater in Montana, USA, the Great Dyke in Zimbabwe and the Bushveld Complex in South Africa. In terms of economic geology, chromitite layers constitute major global reserves of chromium and some contain economic concentrations of platinum-group elements (PGE). As common and important as chromitite layers are, their origin remains highly contentious. Nowhere is the economic importance of chromitite layers greater than in the case of the UG2 layer in the Bushveld Complex, which hosts the world’s largest reserves of PGE, has been extensively mined for decades and is the subject of this report. The UG2 is the most studied chromitite layer in the world, yet there is still no consensus on the relative roles of early-, late-, and post-magmatic processes in producing this remarkable rock layer. Opinions are divided on the fundamental question of whether chromite crystallized within the Bushveld magma chamber or was transported as a slurry from a staging magma chamber below. For detailed accounts of the main features of the UG2 layer and its relationship with other units of the layered sequence, readers are referred to an extensive recent literature, e.g. Cawthorn (2005, 2011, 2015); Kinnaird et al. (2002); Mathez & Mey (2005); Mondal & Mathez (2007); Maier & Barnes (2008); Voordouw et al. (2010); Maier et al., 2013; Naldrett et al. (2012); Junge et al. (2014); Godel (2015), Latypov et al. (2017) and Manoochehri et al. (2015). A major and unsolved obstacle to understanding how chromitites form in the Bushveld and other layered mafic intrusions is that the primary crystallization and chromite accumulation are modified by re-equilibration of minerals with interstitial liquid in the ‘crystal mush’ during final solidification of the layered rocks and even subsequently at sub-solidus conditions. This is a long-recognized problem and considerable progress has been made in understanding the chemical and physical processes operating in such crystal mushes. Of particular relevance for this study is the so-called ‘trapped liquid shift effect’ that describes the compositional re-equilibration among minerals and the evolving interstitial melt (trapped liquid). The first geochemical models of this effect were developed for Mg–Fe exchange and for the distribution of some trace elements (Barnes, 1986a; Cawthorn, 1996). In an extension of the trapped liquid calculations, Manoochehri et al. (2015) considered results from compaction experiments and cooling models to assess the development of porosity in the UG2 and UG3 chromitite layers. Quantitative textural studies of chromite grains and the interstitial silicate minerals can also give insights into the processes operating in a crystal mush (see review by Higgins, 2015). The quantitative approach has been previously applied to statistical analysis of the crystal-size distribution of chromites and silicate minerals (e.g. O’Driscoll et al., 2010; Vukmanovic et al., 2013) and the geometry of grain boundaries (Holness et al. 2007,, 2013). Rarely have textural studies and detailed geochemical analyses been undertaken on the same vertical profiles across a chromitite layer. This is the main objective of the study presented here. Results are presented from two complete profiles through the UG2 layer, one from the eastern limb (Nkwe mine) and one from the western limb (Khuseleka mine). We provide petrographic descriptions and report the chemical compositions of the main rock-forming minerals in the UG2 layer (chromite, orthopyroxene, plagioclase, locally clinopyroxene) and in the silicate rocks immediately above and below the layer, using electron microprobe and laser ablation ICP-MS. The geochemical effect of the evolved, trapped liquid is to produce local enrichments in incompatible trace elements, and in some cases crystallization of accessory phases such as zircon or monazite. The spatial distribution of trace element enrichment may reflect the distribution of trapped liquid in the layer, or the paleo-porosity in the sense of Manoochehri et al. (2015). In an effort to quantify this distribution, we used synchrotron-source XRF spectrometry to obtain continuous, millimetre-resolution trace-element profiles across the UG2 drill core from Khuseleka (including Pd and Rh). To our knowledge this is the first application of the method to layered igneous rocks. Finally, we performed crystal size distribution (CSD) analyses of chromite from UG2 and the adjacent silicate layers, providing the first CSD data from the UG2 chromite. GEOLOGIC SETTING: THE BUSHVELD CHROMITITE LAYERS The Paleoproterozoic (2055–2057 Ma: Scoates & Wall, 2015; Zeh et al., 2015) Bushveld Complex in South Africa is the largest layered igneous intrusion on the Earth. Present-day outcrops of the Complex (Fig. 1) cover a surface area of 66 000 km2 and the total thickness of igneous rocks is about 9 km; however, its former extent and thickness are thought to be much greater (see Cawthorn, 2015 for a comprehensive review). The South African Committee for Stratigraphy (SACS, 1980) recognized three major rock suites in the complex: (1) the mafic–ultramafic Rustenburg Layered Suite (RLS) which is of interest in this study, and overlying felsic suites represented by (2) the Lebowa Granite and (3) the Rashoop Granophyre. Fig. 1 View largeDownload slide Schematic geological map of the Bushveld Complex with sampling locations. Fig. 1 View largeDownload slide Schematic geological map of the Bushveld Complex with sampling locations. The Rustenburg layered suite The c.8 km thick Rustenburg Layered Suite (RLS) is subdivided vertically into several lithologic zones, from bottom to top: the Marginal Zone dominated by norite, the Lower Zone composed of pyroxenites and harzburgites, the Critical Zone (chromitites, harzburgites, pyroxenites, norites and minor anorthosites), the Main Zone (norites, gabbronorites and minor anorthosites) and the Upper Zone (gabbronorites and augite diorites with magnetite and nelsonite layers). There is strong evidence that the RLS rocks formed in an open-system magma chamber affected by multiple injections and mixing of different parental magmas, as well as by significant amounts of crustal assimilation (Harris et al., 2004). The open-system nature of magma evolution in the RLS is shown by trace element ratio variations, changes in the initial Sr isotope ratios and by reversals in compositional trends of cumulate minerals (Eales & Cawthorn, 1996; Maier et al., 2013; Cawthorn, 2015 and references therein). Studies of chilled margins (Wilson, 2012) and comagmatic sills (Barnes et al., 2010) revealed three distinct types of parental magma, referred to in recent publications as B1, B2 and B3 (Barnes et al., 2010; Godel et al., 2011; Maier et al., 2013; VanTongeren & Mathez, 2013). For the Critical Zone rocks of interest here, the B1 magma is believed to represent the parental magma composition (Cawthorn, 2007). The estimates of the B1 magma composition published by different authors vary in detail, but in general it corresponds to an Mg-rich basaltic andesite (SiO2 and MgO contents at about 55 and 12 wt %, respectively). The other two magma types are more evolved, with major element characteristics corresponding to tholeiitic basalt (50–51 wt % SiO2; 6–7·5 wt % MgO). UG2 and other chromitite layers of the critical zone The Critical Zone is defined by the presence of prominent chromitite layers, which are sufficiently laterally continuous to be used as stratigraphic markers. Three groups of chromite layers are distinguished, each with several members: Lower (LG1–7), Middle (MG 1–4) and Upper (UG1–3). Individual layers within each group are numbered sequentially from the base upwards; locally some layers split into one or more sublayers, e.g.LG6a, MG2b, etc. The division of the lower and upper Critical Zones is marked by the appearance of cumulus plagioclase and this occurs within the MG group. The thickest of the chromitite layers reaches about 2 m, but 0·7 to 1 m is more common and there are many thinner chromite seams and stringers of only a few centimetres or even millimetres in thickness between the major chromitite layers. There are systematic compositional variations of chromite with stratigraphic height in the sequence of LG to UG groups. A summary from many localities in the complex is provided by Naldrett et al. (2012) who defined two general trends. Chromites from LG1 to LG4 show a trend of increasing Cr/(Cr+Al) ratio and decreasing Mg/(Mg + Fe2+) upward. This was termed trend A. Trend B, displayed by the LG5 to MG2 sequence, shows upward-decreasing Cr/(Cr+Al) and decreasing Mg/(Mg+Fe2+) ratios. Chromite layers from MG3 upwards again follow trend A. Naldrett et al. (2012) explained the trends by a combination of magma recharge and fractional crystallization. The UG2 is the uppermost massive chromitite layer in the Bushveld Complex except for part of the eastern limb where a UG3 is also present. Above the UG2 or UG3, chromite occurs as an accessory mineral dispersed in silicate rocks or as thin, millimetre- to centimetre-thick stringers. Important examples of the latter are chromitites at the basal and top contacts of the Merensky Reef, which lies several tens to hundreds of metres above UG2. The concentrations of platinum-group elements (PGE) increase upward in the Critical Zone chromitites, from substantial, but sub-economic, levels of around 0·5 to 1 ppm in the LG group to concentrations of 5–10 ppm in UG2 and the Merensky chromite stringers where they are actively mined (Scoon & Teigler, 1994; Naldrett et al., 2009). Indeed, UG2 and the Merensky Reef with its chromitite stringers contain the largest PGE resource in the world (Godel, 2015 and references therein). The mechanisms by which PGE are concentrated in chromitites remain unclear and the situation is unlikely to improve without a better understanding of how the chromite layers themselves formed (Kinnaird et al., 2002). The UG2 layer is typically about 0·8–1·0 m thick, but ranges between 0·15 and 2·5 m Its modal abundance of chromite is 60 to 90 wt % (Mathez & Mey, 2005; Naldrett et al., 2012). Laterally, the UG2 layer may split into two or more thinner seams, which merge again further along strike. As noted above, this is characteristic of other Bushveld chromitite layers, e.g.the MG series, LG6 and especially UG1, whose numerous thin seams interlayer with anorthosite to form the famous ‘zebra rock’ of Dwars River (Voordouw et al., 2009). Cawthorn (2011) proposed that UG3, known only from part of the eastern lobe, has elsewhere merged with the top of the UG2 layer. Lee (1996) and Mathez & Mey (2005) documented a textural break in the middle of the UG2 layer between a lower part with fine-grained chromitite and an upper part containing abundant plagioclase and pyroxene oikocrysts. These and other lines of evidence led Cawthorn (2011) to conclude that the UG2 chromitite may be a composite of three distinct layers, between which there are occasional thin screens or ‘partings’ of silicate-rich rock. Variations of chromite composition in vertical profiles across the UG2 layer have been documented by, e. g., Mathez & Mey (2005) and Junge et al. (2014). Whereas Mathez & Mey (2005) found locally strong but non-systematic differences in Cr/Al, Mg/Fe, Ti and Ni contents, Junge et al. (2014) reported subtle, but recurring, systematic trends of an upward increase in Cr/(Cr+Al) values and TiO2 content, coupled with decreasing Mg/(Mg+Fe2+) values (Trend ‘A’ of Naldrett et al., 2012). We will discuss these results in more detail and compare with our profiles across the UG2. SAMPLES AND ANALYTICAL METHODS This study is based on two complete vertical drill core sections through the UG2 layer, one collected at the Khuseleka mine (formerly Townlands shaft) in the western limb near Rustenburg and the other at the Nkwe Platinum mine in the eastern limb of the Bushveld complex (Fig. 1). Schematic stratigraphic columns of the sections are presented in Fig. 2. The drill cores were cut lengthwise and segmented into continuous sets of petrographic thin sections for textural studies and for electron microprobe and LA-ICP-MS analyses of mineral compositions. The remaining half core (Khuseleka only) was used for synchrotron XRF scanning. Fig. 2 View largeDownload slide Simplified lithologic profiles of the studied drill core sections. (a) Khuseleka; (b) Nkwe. Fig. 2 View largeDownload slide Simplified lithologic profiles of the studied drill core sections. (a) Khuseleka; (b) Nkwe. Electron microprobe analyses Mineral compositions were determined in situ on polished, carbon-coated thin sections at GFZ Potsdam using a JEOL JXA-8230 electron microprobe equipped with four wavelength-dispersive spectrometers (WDS) and at the Museum of Natural History (MfN) in Berlin using a JEOL JXA-8500F electron microprobe equipped with a field emission cathode and five WDS. At GFZ Potsdam, the electron microprobe was operated at 15 kV accelerating voltage and a beam current of 15 nA, with a fully-focused (c.1 micron) beam for clinopyroxene and olivine. Plagioclase was analysed using a defocused 5 micron beam and Na was measured first in the element sequence to minimize the effects of alkali loss. Replicate analyses of natural and synthetic standards indicate that the precision of major oxides (>0·2 wt %) is ± 9 % (2σ) or better. Cr-spinel measurements were made with a 20 KV accelerating voltage, beam current of 75 nA and a 1 μm focussed electron beam spot size. The element concentrations were obtained through calibration with natural and synthetic mineral standards (orthoclase for Al and K, rutile for Ti, wollastonite for Si and Ca, albite and jadeite for Na, apatite for P, hematite for Fe, and diopside and periclase for Mg). Analytical uncertainties for the major components are considered to be better than 1–2%. Automated data reduction and correction in both laboratories followed the PAP method (Pouchou & Pichoir, 1985). In calculations of the Mg# [atomic ratio Mg/(Mg+Fe2+)] for pyroxene and olivine, all Fe was assumed to be Fe2+. For spinel, total Fe was divided into Fe2+ and Fe3+ as described, for example, by Barnes & Roeder (2001) based on the assumption of stoichiometry and charge balance with the ideal formula XY2O4 where X = (Mg, Fe2+, Mn) and Y = (Cr, Fe3+, Al). Titanium was assumed to form the ulvöspinel component. Mineral compositions from the Nkwe profile were previously published by Veksler et al. (2015). Laser ablation ICP-MS analyses Laser ablation analyses were performed in two laboratories, at the Vernadsky Institute of Geochemistry and Analytical Chemistry (GEOKHI) in Moscow, Russia and the University of Erlangen in Germany. Analyses at the Vernadsky Institute used the UP-213 laser system and mass spectrometer Element-XR. The gas flow through ablation chamber was optimized for maximum the signal intensity and minimum background noise. For the He carrier gas a flow rate of 0·6–0·8 L/min was used; Ar sampling gas was set at 0·8–1·2 L/min; auxiliary gas at 1·5–2·0 L/min and the flow of the cooling gas was at 16 L/min. Measurements were performed at low resolution (R = 300). The ablation system parameters were as follows: the laser pulse frequency at 4 Hz, power 35%, diameter of the ablation crater 30 µm. Crystal sizes in most samples allowed line scan ablations and, whenever possible, we preferred to move the laser beam at 1 µm/s with a resulting total track length of 180 µm The analyses were carried out in sequences set by the Element XR software package. Standards were measured at the beginning and the end of each sequence. Between analyses we used a wash time pause of 240 s, in order to remove the evaporation products of the previous sample. Synthetic glass NIST-610 was used as the main standard with further control of the quality of measurements against the reference glass ML3B. The concentration values were taken from the GeoRem Database (http://georem.mpch-mainz.gwdg.de). Concentrations of Ca and Al measured by microprobe were used as internal standards for pyroxene and plagioclase, respectively. The obtained data were processed using the program ‘Glitter’ (Van Achterbergh et al., 1999). The laser ablation device at the University of Erlangen included a New Wave Research UP193FX laser and an Agilent 7500i, Plasma power 1320 W ICP-MS. The flow rate for carrier gas I (He) was 0·65 L/min, for carrier gas II (Ar) 1·07 L/min, for plasma gas (Ar) 14·9 L/min. The auxiliary gas (Ar) with flow rate at 0·9 L/min was tuned for maximum sensitivity and ThO+/Th+ <0·5%. We used a single spot ablation mode with a laser pulse rate of 15 Hz and crater diameter of 100 µm. The irradiance laser energy was at 0·67 GW/cm2 with the energy density at 3·37 J/cm2. The ablation time was 30 s for the background and 60 s for the analysis. Integration time for trace element isotopes was 20 ms for each mass, 5 ms each for 26 Mg, 27Al, 57Fe, and 10 ms each for 29Si and 44Ca. Silica was used as the internal standard. NIST SRM 612 glass standard was used for external calibration. Accuracy and reproducibility were checked by ablation of the NIST SRM 614 standard. Data reduction employed the programme LAMTRACE described by Longerich et al. (1996) and van Achterbergh et al. (2001). The programme performs background correction, correction for instrumental drift, internal calibration, choice of integration intervals and calculation of element concentrations using external calibration. The Ca signal and microprobe data were used for internal calibration of trace element concentrations in plagioclase and Mg was used in the same way for pyroxene, olivine and spinel. A summary of the electron microprobe and LA-ICP-MS data is given in Tables 1 and 2. Table 1 Orthopyroxene compositions, average concentrations (Av.) and standard deviations (S. D.) of N microprobe and n LA ICP-MS analyses. Major oxides are in wt %, trace elements are in ppm Profile Khuseleka Nkwe Host rock Footwall harzburgite Pyroxenite parting Hanging wall pyroxenite Footwall pyroxenite UG2 chromitite Hanging wall anorthosite Hanging wall pyroxenite N(n) 8 (7) 36 (3) 25 (3) 28 (11) 44 (10) 3 (2) 40 (10) Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. SiO2 54·84 0·27 54·73 0·92 54·81 0·59 54·88 0·42 56·52 0·24 52·77 0·17 54·96 0·31 TiO2 0·13 0·02 0·16 0·04 0·15 0·02 0·16 0·05 0·08 0·02 0·15 0·01 0·15 0·03 Al2O3 1·74 0·06 1·32 0·16 1·38 0·10 1·24 0·24 1·22 0·11 1·22 0·08 1·17 0·07 Cr2O3 0·55 0·06 0·54 0·07 0·51 0·04 0·35 0·10 0·55 0·15 0·09 0·02 0·33 0·05 MgO 30·37 0·63 30·03 1·70 30·25 0·92 28·67 0·62 33·55 0·37 22·73 0·05 28·96 0·78 MnO 0·26 0·02 0·29 0·03 0·27 0·03 0·29 0·03 0·18 0·02 0·42 0·03 0·27 0·03 FeO 10·29 0·46 11·42 1·80 10·93 0·94 14·41 0·72 8·22 0·26 22·52 0·36 13·61 0·84 CaO 2·07 0·83 1·57 1·38 1·66 1·19 0·94 0·47 0·54 0·15 0·83 0·10 1·10 0·51 Na2O 0·03 0·02 0·03 0·03 0·03 0·03 0·02 0·02 0·01 0·01 0·02 0·03 0·02 0·02 Total 100·26 100·09 100·00 101·08 101·01 100·96 100·72 Mg# 0·84 0·82 0·83 0·78 0·88 0·64 0·79 Trace elements, ppm P 35·2 11·3 34·4 10·0 37·5 8·9 n. a. n. a. n. a. n. a. K n. a. n. a. n. a. 6·7 0·4 3·1 0·3 b. d. l. 1·6 0·2 Sc 28·5 1·9 31·4 1·2 36·9 2·3 29·4 1·4 49·7 2·2 30·1 1·3 29·5 1·2 Ti 682 118 885 139 864 181 1191 62 542 26 1148 56 1062 46 V n. a. n. a. n. a. 155 7 72 3 156 7 177 7 Cr 3448 620 3500 23 3130 261 3055 178 2200 113 2504 120 2919 138 Mn 1905 57 2293 112 2025 106 2408 92 1568 57 2492 91 2309 80 Co 96 5 112 10 81 5 132 5 46 2 129 5 131 4 Ni 746 47 871 15 736 8 834 30 857 29 1754 60 954 31 Cu 0·87 0·23 0·89 0·13 0·82 0·11 n. a. n. a. n. a. n. a. Rb 0·14 0·02 0·26 0·05 0·20 0·03 0·049 0·004 0·019 0·006 0·013 0·001 Sr 0·41 0·27 0·28 0·03 0·26 0·03 0·48 0·02 0·066 0·007 0·132 0·009 0·236 0·010 Y 3·2 1·0 3·2 0·1 4·4 0·8 3·1 0·2 2·4 0·1 2·0 0·1 2·2 0·1 Zr 8·2 5·6 5·5 0·7 4·3 0·6 n. a. n. a. n. a. n. a. Nb 0·092 0·048 0·055 0·005 0·054 0·013 n. a. n. a. n. a. n. a. Ba 0·265 0·089 0·333 0·021 0·263 0·050 0·136 0·011 0·063 0·025 b. d. l. 0·058 0·006 La 0·130 0·067 0·080 0·025 0·050 0·019 0·158 0·007 0·021 0·004 0·007 0·002 0·037 0·003 Ce 0·444 0·290 0·427 0·087 0·247 0·173 0·565 0·022 0·048 0·005 0·025 0·003 0·143 0·007 Pr 0·073 0·050 0·072 0·031 0·056 0·034 0·080 0·004 0·008 0·002 0·007 0·002 0·025 0·002 Nd 0·385 0·252 0·405 0·138 0·377 0·205 0·394 0·020 0·054 0·014 0·050 0·013 0·140 0·014 Sm 0·174 0·112 0·317 0·086 0·266 0·134 0·137 0·010 0·039 0·014 0·028 0·012 0·072 0·011 Eu 0·051 0·035 0·055 0·024 0·056 0·013 0·026 0·002 0·008 0·003 0·010 0·003 0·012 0·002 Gd 0·295 0·114 0·410 0·096 0·460 0·135 0·227 0·016 0·083 0·019 0·084 0·016 0·135 0·015 Tb 0·060 0·028 0·067 0·003 0·094 0·026 0·050 0·003 0·025 0·004 0·022 0·003 0·033 0·003 Dy 0·505 0·216 0·319 0·086 0·718 0·265 0·446 0·030 0·298 0·032 0·217 0·022 0·301 0·023 Ho 0·123 0·046 0·109 0·008 0·147 0·036 0·112 0·007 0·086 0·009 0·066 0·006 0·078 0·006 Er 0·402 0·124 0·386 0·107 0·524 0·039 0·403 0·026 0·357 0·033 0·272 0·023 0·276 0·019 Tm 0·063 0·018 0·065 0·027 0·084 0·031 0·073 0·005 0·065 0·007 0·052 0·005 0·050 0·004 Yb 0·532 0·183 0·478 0·101 0·672 0·063 0·576 0·041 0·558 0·053 0·485 0·040 0·391 0·030 Lu 0·088 0·025 0·075 0·015 0·092 0·032 0·096 0·007 0·100 0·010 0·089 0·008 0·067 0·005 Hf 0·195 0·140 0·164 0·071 0·258 0·151 n. a. n. a. n. a. n. a. Ta 0·029 0·005 0·028 0·007 0·031 0·012 n. a. n. a. n. a. n. a. Pb 0·238 0·132 0·153 0·078 0·171 0·105 0·101 0·007 0·053 0·010 0·042 0·008 0·113 0·009 Th 0·185 0·178 0·078 0·011 0·132 0·108 n. a. n. a. n. a. n. a. U 0·042 0·034 0·026 0·005 0·016 0·003 n. a. n. a. n. a. n. a. Profile Khuseleka Nkwe Host rock Footwall harzburgite Pyroxenite parting Hanging wall pyroxenite Footwall pyroxenite UG2 chromitite Hanging wall anorthosite Hanging wall pyroxenite N(n) 8 (7) 36 (3) 25 (3) 28 (11) 44 (10) 3 (2) 40 (10) Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. SiO2 54·84 0·27 54·73 0·92 54·81 0·59 54·88 0·42 56·52 0·24 52·77 0·17 54·96 0·31 TiO2 0·13 0·02 0·16 0·04 0·15 0·02 0·16 0·05 0·08 0·02 0·15 0·01 0·15 0·03 Al2O3 1·74 0·06 1·32 0·16 1·38 0·10 1·24 0·24 1·22 0·11 1·22 0·08 1·17 0·07 Cr2O3 0·55 0·06 0·54 0·07 0·51 0·04 0·35 0·10 0·55 0·15 0·09 0·02 0·33 0·05 MgO 30·37 0·63 30·03 1·70 30·25 0·92 28·67 0·62 33·55 0·37 22·73 0·05 28·96 0·78 MnO 0·26 0·02 0·29 0·03 0·27 0·03 0·29 0·03 0·18 0·02 0·42 0·03 0·27 0·03 FeO 10·29 0·46 11·42 1·80 10·93 0·94 14·41 0·72 8·22 0·26 22·52 0·36 13·61 0·84 CaO 2·07 0·83 1·57 1·38 1·66 1·19 0·94 0·47 0·54 0·15 0·83 0·10 1·10 0·51 Na2O 0·03 0·02 0·03 0·03 0·03 0·03 0·02 0·02 0·01 0·01 0·02 0·03 0·02 0·02 Total 100·26 100·09 100·00 101·08 101·01 100·96 100·72 Mg# 0·84 0·82 0·83 0·78 0·88 0·64 0·79 Trace elements, ppm P 35·2 11·3 34·4 10·0 37·5 8·9 n. a. n. a. n. a. n. a. K n. a. n. a. n. a. 6·7 0·4 3·1 0·3 b. d. l. 1·6 0·2 Sc 28·5 1·9 31·4 1·2 36·9 2·3 29·4 1·4 49·7 2·2 30·1 1·3 29·5 1·2 Ti 682 118 885 139 864 181 1191 62 542 26 1148 56 1062 46 V n. a. n. a. n. a. 155 7 72 3 156 7 177 7 Cr 3448 620 3500 23 3130 261 3055 178 2200 113 2504 120 2919 138 Mn 1905 57 2293 112 2025 106 2408 92 1568 57 2492 91 2309 80 Co 96 5 112 10 81 5 132 5 46 2 129 5 131 4 Ni 746 47 871 15 736 8 834 30 857 29 1754 60 954 31 Cu 0·87 0·23 0·89 0·13 0·82 0·11 n. a. n. a. n. a. n. a. Rb 0·14 0·02 0·26 0·05 0·20 0·03 0·049 0·004 0·019 0·006 0·013 0·001 Sr 0·41 0·27 0·28 0·03 0·26 0·03 0·48 0·02 0·066 0·007 0·132 0·009 0·236 0·010 Y 3·2 1·0 3·2 0·1 4·4 0·8 3·1 0·2 2·4 0·1 2·0 0·1 2·2 0·1 Zr 8·2 5·6 5·5 0·7 4·3 0·6 n. a. n. a. n. a. n. a. Nb 0·092 0·048 0·055 0·005 0·054 0·013 n. a. n. a. n. a. n. a. Ba 0·265 0·089 0·333 0·021 0·263 0·050 0·136 0·011 0·063 0·025 b. d. l. 0·058 0·006 La 0·130 0·067 0·080 0·025 0·050 0·019 0·158 0·007 0·021 0·004 0·007 0·002 0·037 0·003 Ce 0·444 0·290 0·427 0·087 0·247 0·173 0·565 0·022 0·048 0·005 0·025 0·003 0·143 0·007 Pr 0·073 0·050 0·072 0·031 0·056 0·034 0·080 0·004 0·008 0·002 0·007 0·002 0·025 0·002 Nd 0·385 0·252 0·405 0·138 0·377 0·205 0·394 0·020 0·054 0·014 0·050 0·013 0·140 0·014 Sm 0·174 0·112 0·317 0·086 0·266 0·134 0·137 0·010 0·039 0·014 0·028 0·012 0·072 0·011 Eu 0·051 0·035 0·055 0·024 0·056 0·013 0·026 0·002 0·008 0·003 0·010 0·003 0·012 0·002 Gd 0·295 0·114 0·410 0·096 0·460 0·135 0·227 0·016 0·083 0·019 0·084 0·016 0·135 0·015 Tb 0·060 0·028 0·067 0·003 0·094 0·026 0·050 0·003 0·025 0·004 0·022 0·003 0·033 0·003 Dy 0·505 0·216 0·319 0·086 0·718 0·265 0·446 0·030 0·298 0·032 0·217 0·022 0·301 0·023 Ho 0·123 0·046 0·109 0·008 0·147 0·036 0·112 0·007 0·086 0·009 0·066 0·006 0·078 0·006 Er 0·402 0·124 0·386 0·107 0·524 0·039 0·403 0·026 0·357 0·033 0·272 0·023 0·276 0·019 Tm 0·063 0·018 0·065 0·027 0·084 0·031 0·073 0·005 0·065 0·007 0·052 0·005 0·050 0·004 Yb 0·532 0·183 0·478 0·101 0·672 0·063 0·576 0·041 0·558 0·053 0·485 0·040 0·391 0·030 Lu 0·088 0·025 0·075 0·015 0·092 0·032 0·096 0·007 0·100 0·010 0·089 0·008 0·067 0·005 Hf 0·195 0·140 0·164 0·071 0·258 0·151 n. a. n. a. n. a. n. a. Ta 0·029 0·005 0·028 0·007 0·031 0·012 n. a. n. a. n. a. n. a. Pb 0·238 0·132 0·153 0·078 0·171 0·105 0·101 0·007 0·053 0·010 0·042 0·008 0·113 0·009 Th 0·185 0·178 0·078 0·011 0·132 0·108 n. a. n. a. n. a. n. a. U 0·042 0·034 0·026 0·005 0·016 0·003 n. a. n. a. n. a. n. a. n. a., not analysed Table 1 Orthopyroxene compositions, average concentrations (Av.) and standard deviations (S. D.) of N microprobe and n LA ICP-MS analyses. Major oxides are in wt %, trace elements are in ppm Profile Khuseleka Nkwe Host rock Footwall harzburgite Pyroxenite parting Hanging wall pyroxenite Footwall pyroxenite UG2 chromitite Hanging wall anorthosite Hanging wall pyroxenite N(n) 8 (7) 36 (3) 25 (3) 28 (11) 44 (10) 3 (2) 40 (10) Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. SiO2 54·84 0·27 54·73 0·92 54·81 0·59 54·88 0·42 56·52 0·24 52·77 0·17 54·96 0·31 TiO2 0·13 0·02 0·16 0·04 0·15 0·02 0·16 0·05 0·08 0·02 0·15 0·01 0·15 0·03 Al2O3 1·74 0·06 1·32 0·16 1·38 0·10 1·24 0·24 1·22 0·11 1·22 0·08 1·17 0·07 Cr2O3 0·55 0·06 0·54 0·07 0·51 0·04 0·35 0·10 0·55 0·15 0·09 0·02 0·33 0·05 MgO 30·37 0·63 30·03 1·70 30·25 0·92 28·67 0·62 33·55 0·37 22·73 0·05 28·96 0·78 MnO 0·26 0·02 0·29 0·03 0·27 0·03 0·29 0·03 0·18 0·02 0·42 0·03 0·27 0·03 FeO 10·29 0·46 11·42 1·80 10·93 0·94 14·41 0·72 8·22 0·26 22·52 0·36 13·61 0·84 CaO 2·07 0·83 1·57 1·38 1·66 1·19 0·94 0·47 0·54 0·15 0·83 0·10 1·10 0·51 Na2O 0·03 0·02 0·03 0·03 0·03 0·03 0·02 0·02 0·01 0·01 0·02 0·03 0·02 0·02 Total 100·26 100·09 100·00 101·08 101·01 100·96 100·72 Mg# 0·84 0·82 0·83 0·78 0·88 0·64 0·79 Trace elements, ppm P 35·2 11·3 34·4 10·0 37·5 8·9 n. a. n. a. n. a. n. a. K n. a. n. a. n. a. 6·7 0·4 3·1 0·3 b. d. l. 1·6 0·2 Sc 28·5 1·9 31·4 1·2 36·9 2·3 29·4 1·4 49·7 2·2 30·1 1·3 29·5 1·2 Ti 682 118 885 139 864 181 1191 62 542 26 1148 56 1062 46 V n. a. n. a. n. a. 155 7 72 3 156 7 177 7 Cr 3448 620 3500 23 3130 261 3055 178 2200 113 2504 120 2919 138 Mn 1905 57 2293 112 2025 106 2408 92 1568 57 2492 91 2309 80 Co 96 5 112 10 81 5 132 5 46 2 129 5 131 4 Ni 746 47 871 15 736 8 834 30 857 29 1754 60 954 31 Cu 0·87 0·23 0·89 0·13 0·82 0·11 n. a. n. a. n. a. n. a. Rb 0·14 0·02 0·26 0·05 0·20 0·03 0·049 0·004 0·019 0·006 0·013 0·001 Sr 0·41 0·27 0·28 0·03 0·26 0·03 0·48 0·02 0·066 0·007 0·132 0·009 0·236 0·010 Y 3·2 1·0 3·2 0·1 4·4 0·8 3·1 0·2 2·4 0·1 2·0 0·1 2·2 0·1 Zr 8·2 5·6 5·5 0·7 4·3 0·6 n. a. n. a. n. a. n. a. Nb 0·092 0·048 0·055 0·005 0·054 0·013 n. a. n. a. n. a. n. a. Ba 0·265 0·089 0·333 0·021 0·263 0·050 0·136 0·011 0·063 0·025 b. d. l. 0·058 0·006 La 0·130 0·067 0·080 0·025 0·050 0·019 0·158 0·007 0·021 0·004 0·007 0·002 0·037 0·003 Ce 0·444 0·290 0·427 0·087 0·247 0·173 0·565 0·022 0·048 0·005 0·025 0·003 0·143 0·007 Pr 0·073 0·050 0·072 0·031 0·056 0·034 0·080 0·004 0·008 0·002 0·007 0·002 0·025 0·002 Nd 0·385 0·252 0·405 0·138 0·377 0·205 0·394 0·020 0·054 0·014 0·050 0·013 0·140 0·014 Sm 0·174 0·112 0·317 0·086 0·266 0·134 0·137 0·010 0·039 0·014 0·028 0·012 0·072 0·011 Eu 0·051 0·035 0·055 0·024 0·056 0·013 0·026 0·002 0·008 0·003 0·010 0·003 0·012 0·002 Gd 0·295 0·114 0·410 0·096 0·460 0·135 0·227 0·016 0·083 0·019 0·084 0·016 0·135 0·015 Tb 0·060 0·028 0·067 0·003 0·094 0·026 0·050 0·003 0·025 0·004 0·022 0·003 0·033 0·003 Dy 0·505 0·216 0·319 0·086 0·718 0·265 0·446 0·030 0·298 0·032 0·217 0·022 0·301 0·023 Ho 0·123 0·046 0·109 0·008 0·147 0·036 0·112 0·007 0·086 0·009 0·066 0·006 0·078 0·006 Er 0·402 0·124 0·386 0·107 0·524 0·039 0·403 0·026 0·357 0·033 0·272 0·023 0·276 0·019 Tm 0·063 0·018 0·065 0·027 0·084 0·031 0·073 0·005 0·065 0·007 0·052 0·005 0·050 0·004 Yb 0·532 0·183 0·478 0·101 0·672 0·063 0·576 0·041 0·558 0·053 0·485 0·040 0·391 0·030 Lu 0·088 0·025 0·075 0·015 0·092 0·032 0·096 0·007 0·100 0·010 0·089 0·008 0·067 0·005 Hf 0·195 0·140 0·164 0·071 0·258 0·151 n. a. n. a. n. a. n. a. Ta 0·029 0·005 0·028 0·007 0·031 0·012 n. a. n. a. n. a. n. a. Pb 0·238 0·132 0·153 0·078 0·171 0·105 0·101 0·007 0·053 0·010 0·042 0·008 0·113 0·009 Th 0·185 0·178 0·078 0·011 0·132 0·108 n. a. n. a. n. a. n. a. U 0·042 0·034 0·026 0·005 0·016 0·003 n. a. n. a. n. a. n. a. Profile Khuseleka Nkwe Host rock Footwall harzburgite Pyroxenite parting Hanging wall pyroxenite Footwall pyroxenite UG2 chromitite Hanging wall anorthosite Hanging wall pyroxenite N(n) 8 (7) 36 (3) 25 (3) 28 (11) 44 (10) 3 (2) 40 (10) Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. SiO2 54·84 0·27 54·73 0·92 54·81 0·59 54·88 0·42 56·52 0·24 52·77 0·17 54·96 0·31 TiO2 0·13 0·02 0·16 0·04 0·15 0·02 0·16 0·05 0·08 0·02 0·15 0·01 0·15 0·03 Al2O3 1·74 0·06 1·32 0·16 1·38 0·10 1·24 0·24 1·22 0·11 1·22 0·08 1·17 0·07 Cr2O3 0·55 0·06 0·54 0·07 0·51 0·04 0·35 0·10 0·55 0·15 0·09 0·02 0·33 0·05 MgO 30·37 0·63 30·03 1·70 30·25 0·92 28·67 0·62 33·55 0·37 22·73 0·05 28·96 0·78 MnO 0·26 0·02 0·29 0·03 0·27 0·03 0·29 0·03 0·18 0·02 0·42 0·03 0·27 0·03 FeO 10·29 0·46 11·42 1·80 10·93 0·94 14·41 0·72 8·22 0·26 22·52 0·36 13·61 0·84 CaO 2·07 0·83 1·57 1·38 1·66 1·19 0·94 0·47 0·54 0·15 0·83 0·10 1·10 0·51 Na2O 0·03 0·02 0·03 0·03 0·03 0·03 0·02 0·02 0·01 0·01 0·02 0·03 0·02 0·02 Total 100·26 100·09 100·00 101·08 101·01 100·96 100·72 Mg# 0·84 0·82 0·83 0·78 0·88 0·64 0·79 Trace elements, ppm P 35·2 11·3 34·4 10·0 37·5 8·9 n. a. n. a. n. a. n. a. K n. a. n. a. n. a. 6·7 0·4 3·1 0·3 b. d. l. 1·6 0·2 Sc 28·5 1·9 31·4 1·2 36·9 2·3 29·4 1·4 49·7 2·2 30·1 1·3 29·5 1·2 Ti 682 118 885 139 864 181 1191 62 542 26 1148 56 1062 46 V n. a. n. a. n. a. 155 7 72 3 156 7 177 7 Cr 3448 620 3500 23 3130 261 3055 178 2200 113 2504 120 2919 138 Mn 1905 57 2293 112 2025 106 2408 92 1568 57 2492 91 2309 80 Co 96 5 112 10 81 5 132 5 46 2 129 5 131 4 Ni 746 47 871 15 736 8 834 30 857 29 1754 60 954 31 Cu 0·87 0·23 0·89 0·13 0·82 0·11 n. a. n. a. n. a. n. a. Rb 0·14 0·02 0·26 0·05 0·20 0·03 0·049 0·004 0·019 0·006 0·013 0·001 Sr 0·41 0·27 0·28 0·03 0·26 0·03 0·48 0·02 0·066 0·007 0·132 0·009 0·236 0·010 Y 3·2 1·0 3·2 0·1 4·4 0·8 3·1 0·2 2·4 0·1 2·0 0·1 2·2 0·1 Zr 8·2 5·6 5·5 0·7 4·3 0·6 n. a. n. a. n. a. n. a. Nb 0·092 0·048 0·055 0·005 0·054 0·013 n. a. n. a. n. a. n. a. Ba 0·265 0·089 0·333 0·021 0·263 0·050 0·136 0·011 0·063 0·025 b. d. l. 0·058 0·006 La 0·130 0·067 0·080 0·025 0·050 0·019 0·158 0·007 0·021 0·004 0·007 0·002 0·037 0·003 Ce 0·444 0·290 0·427 0·087 0·247 0·173 0·565 0·022 0·048 0·005 0·025 0·003 0·143 0·007 Pr 0·073 0·050 0·072 0·031 0·056 0·034 0·080 0·004 0·008 0·002 0·007 0·002 0·025 0·002 Nd 0·385 0·252 0·405 0·138 0·377 0·205 0·394 0·020 0·054 0·014 0·050 0·013 0·140 0·014 Sm 0·174 0·112 0·317 0·086 0·266 0·134 0·137 0·010 0·039 0·014 0·028 0·012 0·072 0·011 Eu 0·051 0·035 0·055 0·024 0·056 0·013 0·026 0·002 0·008 0·003 0·010 0·003 0·012 0·002 Gd 0·295 0·114 0·410 0·096 0·460 0·135 0·227 0·016 0·083 0·019 0·084 0·016 0·135 0·015 Tb 0·060 0·028 0·067 0·003 0·094 0·026 0·050 0·003 0·025 0·004 0·022 0·003 0·033 0·003 Dy 0·505 0·216 0·319 0·086 0·718 0·265 0·446 0·030 0·298 0·032 0·217 0·022 0·301 0·023 Ho 0·123 0·046 0·109 0·008 0·147 0·036 0·112 0·007 0·086 0·009 0·066 0·006 0·078 0·006 Er 0·402 0·124 0·386 0·107 0·524 0·039 0·403 0·026 0·357 0·033 0·272 0·023 0·276 0·019 Tm 0·063 0·018 0·065 0·027 0·084 0·031 0·073 0·005 0·065 0·007 0·052 0·005 0·050 0·004 Yb 0·532 0·183 0·478 0·101 0·672 0·063 0·576 0·041 0·558 0·053 0·485 0·040 0·391 0·030 Lu 0·088 0·025 0·075 0·015 0·092 0·032 0·096 0·007 0·100 0·010 0·089 0·008 0·067 0·005 Hf 0·195 0·140 0·164 0·071 0·258 0·151 n. a. n. a. n. a. n. a. Ta 0·029 0·005 0·028 0·007 0·031 0·012 n. a. n. a. n. a. n. a. Pb 0·238 0·132 0·153 0·078 0·171 0·105 0·101 0·007 0·053 0·010 0·042 0·008 0·113 0·009 Th 0·185 0·178 0·078 0·011 0·132 0·108 n. a. n. a. n. a. n. a. U 0·042 0·034 0·026 0·005 0·016 0·003 n. a. n. a. n. a. n. a. n. a., not analysed Table 2 Plagioclase compositions, average concentrations (Av.) and standard deviations (S. D.) of N microprobe and n LA ICP-MS analyses Mine Khuseleka Nkwe Host rock Footwall harzburgite UG2 chromitite Pyroxenite parting Hanging wall pyroxenite Footwall pyroxenite UG2 chromitite Hanging wall anorthosite Hanging wall pyroxenite N(n) 9 (7) 79 (3) 32 (3) 45 (3) 41 (11) 45 (21) 50 (8) 98 (15) Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. SiO2 49·95 1·03 50·39 0·90 50·51 1·09 50·81 1·16 55·10 0·83 50·99 0·77 50·27 0·34 52·55 1·34 Al2O3 31·83 0·73 31·88 0·80 31·75 0·77 31·72 0·78 28·82 0·56 31·83 0·48 32·11 0·24 30·48 0·85 FeO 0·32 0·03 0·27 0·18 0·19 0·03 0·19 0·02 0·21 0·09 0·19 0·08 0·23 0·06 0·22 0·11 CaO 14·63 0·79 14·05 0·88 14·24 0·92 13·94 0·77 11·38 0·62 14·72 0·54 15·18 0·24 13·41 0·96 Na2O 2·95 0·40 3·23 0·40 3·22 0·50 3·18 0·49 4·97 0·40 2·99 0·34 2·63 0·15 3·67 0·60 K2O 0·28 0·08 0·11 0·05 0·20 0·05 0·21 0·05 0·17 0·05 0·02 0·01 0·12 0·02 0·20 0·04 Total 99·96 99·92 100·1 100·1 100·7 100·9 100·6 100·6 An 72·1 70·1 70·2 69·9 55·8 73·0 75·9 66·9 Trace elements, ppm Mg 435 241 18 5 84 13 142 15 n. a. n. a. n. a. n. a. P 217 28 316 20 465 52 435 42 n. a. n. a. n. a. n. a. Sc 0·68 0·40 0·72 0·03 1·16 0·39 0·84 0·22 0·66 0·04 0·88 0·05 0·58 0·03 0·79 0·04 Ti 115 28 11 2 87 47 103 25 138 8 60 3 138 8 125 5 Mn 17·7 15·0 2·4 0·1 8·9 2·1 5·1 1·4 6·9 0·3 3·5 0·2 17·4 0·7 7·2 0·3 Fe 2966 440 280 98 1143 88 1062 42 n. a. n. a. n. a. n. a. Sr 479 24 422 17 421 4 454 49 463 19 527 18 497 22 517 17 Ba 146 12 83 27 113 13 148 50 150 6 80 3 57 3 153 6 Rb 0·935 0·39 0·417 0·015 0·685 0·191 0·455 0·163 0·266 0·016 0·023 0·007 0·119 0·009 0·106 0·008 Y 0·209 0·07 0·433 0·179 0·415 0·134 0·330 0·057 0·183 0·013 0·271 0·021 0·192 0·015 0·322 0·018 La 3·61 0·96 5·94 2·45 5·98 1·07 3·70 0·40 10·83 0·40 5·46 0·20 1·48 0·06 7·08 0·26 Ce 8·36 1·93 10·83 3·87 9·98 1·01 6·04 0·45 13·47 0·48 8·76 0·30 2·65 0·10 10·66 0·36 Pr 0·693 0·18 1·040 0·44 0·882 0·038 0·547 0·016 0·961 0·04 0·80 0·04 0·266 0·013 0·885 0·035 Nd 2·335 0·61 3·477 1·42 2·550 0·424 2·105 0·120 2·514 0·12 2·65 0·147 0·957 0·052 2·706 0·136 Sm 0·246 0·10 0·487 0·12 0·231 0·027 0·430 0·269 0·226 0·02 0·28 0·044 0·132 0·018 0·295 0·029 Eu 0·634 0·11 0·720 0·36 0·733 0·038 0·661 0·194 0·835 0·04 0·62 0·037 0·312 0·017 0·870 0·037 Gd 0·232 0·03 0·663 0·15 0·560 0·113 0·595 0·134 0·114 0·02 0·19 0·036 0·088 0·015 0·190 0·022 Tb 0·035 0·014 0·071 0·029 0·065 0·038 0·034 0·008 0·011 0·002 0·019 0·004 0·008 0·002 0·017 0·002 Dy 0·100 0·014 0·280 0·017 0·254 0·122 0·201 0·001 0·046 0·008 0·079 0·019 0·046 0·008 0·078 0·012 Ho b. d. l. b. d. l. b. d. l. b. d. l. 0·007 0·001 0·013 0·003 0·007 0·001 0·010 0·002 Er b. d. l. b. d. l. b. d. l. b. d. l. 0·013 0·003 0·029 0·010 0·014 0·003 0·022 0·005 Tm b. d. l. b. d. l. b. d. l. b. d. l. 0·002 0·001 0·002 0·001 0·002 0·001 0·004 0·001 Yb b. d. l. b. d. l. b. d. l. b. d. l. 0·008 0·003 0·004 0·002 0·011 0·004 0·013 0·004 Lu b. d. l. b. d. l. b. d. l. b. d. l. b. d. l. b. d. l. 0·005 0·002 0·002 0·001 Pb 2·35 0·21 2·16 1·09 2·56 0·08 3·24 1·24 12·85 0·55 12·58 0·52 0·82 0·04 3·75 0·18 Mine Khuseleka Nkwe Host rock Footwall harzburgite UG2 chromitite Pyroxenite parting Hanging wall pyroxenite Footwall pyroxenite UG2 chromitite Hanging wall anorthosite Hanging wall pyroxenite N(n) 9 (7) 79 (3) 32 (3) 45 (3) 41 (11) 45 (21) 50 (8) 98 (15) Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. SiO2 49·95 1·03 50·39 0·90 50·51 1·09 50·81 1·16 55·10 0·83 50·99 0·77 50·27 0·34 52·55 1·34 Al2O3 31·83 0·73 31·88 0·80 31·75 0·77 31·72 0·78 28·82 0·56 31·83 0·48 32·11 0·24 30·48 0·85 FeO 0·32 0·03 0·27 0·18 0·19 0·03 0·19 0·02 0·21 0·09 0·19 0·08 0·23 0·06 0·22 0·11 CaO 14·63 0·79 14·05 0·88 14·24 0·92 13·94 0·77 11·38 0·62 14·72 0·54 15·18 0·24 13·41 0·96 Na2O 2·95 0·40 3·23 0·40 3·22 0·50 3·18 0·49 4·97 0·40 2·99 0·34 2·63 0·15 3·67 0·60 K2O 0·28 0·08 0·11 0·05 0·20 0·05 0·21 0·05 0·17 0·05 0·02 0·01 0·12 0·02 0·20 0·04 Total 99·96 99·92 100·1 100·1 100·7 100·9 100·6 100·6 An 72·1 70·1 70·2 69·9 55·8 73·0 75·9 66·9 Trace elements, ppm Mg 435 241 18 5 84 13 142 15 n. a. n. a. n. a. n. a. P 217 28 316 20 465 52 435 42 n. a. n. a. n. a. n. a. Sc 0·68 0·40 0·72 0·03 1·16 0·39 0·84 0·22 0·66 0·04 0·88 0·05 0·58 0·03 0·79 0·04 Ti 115 28 11 2 87 47 103 25 138 8 60 3 138 8 125 5 Mn 17·7 15·0 2·4 0·1 8·9 2·1 5·1 1·4 6·9 0·3 3·5 0·2 17·4 0·7 7·2 0·3 Fe 2966 440 280 98 1143 88 1062 42 n. a. n. a. n. a. n. a. Sr 479 24 422 17 421 4 454 49 463 19 527 18 497 22 517 17 Ba 146 12 83 27 113 13 148 50 150 6 80 3 57 3 153 6 Rb 0·935 0·39 0·417 0·015 0·685 0·191 0·455 0·163 0·266 0·016 0·023 0·007 0·119 0·009 0·106 0·008 Y 0·209 0·07 0·433 0·179 0·415 0·134 0·330 0·057 0·183 0·013 0·271 0·021 0·192 0·015 0·322 0·018 La 3·61 0·96 5·94 2·45 5·98 1·07 3·70 0·40 10·83 0·40 5·46 0·20 1·48 0·06 7·08 0·26 Ce 8·36 1·93 10·83 3·87 9·98 1·01 6·04 0·45 13·47 0·48 8·76 0·30 2·65 0·10 10·66 0·36 Pr 0·693 0·18 1·040 0·44 0·882 0·038 0·547 0·016 0·961 0·04 0·80 0·04 0·266 0·013 0·885 0·035 Nd 2·335 0·61 3·477 1·42 2·550 0·424 2·105 0·120 2·514 0·12 2·65 0·147 0·957 0·052 2·706 0·136 Sm 0·246 0·10 0·487 0·12 0·231 0·027 0·430 0·269 0·226 0·02 0·28 0·044 0·132 0·018 0·295 0·029 Eu 0·634 0·11 0·720 0·36 0·733 0·038 0·661 0·194 0·835 0·04 0·62 0·037 0·312 0·017 0·870 0·037 Gd 0·232 0·03 0·663 0·15 0·560 0·113 0·595 0·134 0·114 0·02 0·19 0·036 0·088 0·015 0·190 0·022 Tb 0·035 0·014 0·071 0·029 0·065 0·038 0·034 0·008 0·011 0·002 0·019 0·004 0·008 0·002 0·017 0·002 Dy 0·100 0·014 0·280 0·017 0·254 0·122 0·201 0·001 0·046 0·008 0·079 0·019 0·046 0·008 0·078 0·012 Ho b. d. l. b. d. l. b. d. l. b. d. l. 0·007 0·001 0·013 0·003 0·007 0·001 0·010 0·002 Er b. d. l. b. d. l. b. d. l. b. d. l. 0·013 0·003 0·029 0·010 0·014 0·003 0·022 0·005 Tm b. d. l. b. d. l. b. d. l. b. d. l. 0·002 0·001 0·002 0·001 0·002 0·001 0·004 0·001 Yb b. d. l. b. d. l. b. d. l. b. d. l. 0·008 0·003 0·004 0·002 0·011 0·004 0·013 0·004 Lu b. d. l. b. d. l. b. d. l. b. d. l. b. d. l. b. d. l. 0·005 0·002 0·002 0·001 Pb 2·35 0·21 2·16 1·09 2·56 0·08 3·24 1·24 12·85 0·55 12·58 0·52 0·82 0·04 3·75 0·18 Major oxides are in wt %, trace elements are in ppm. b. d. l., below the detection limit; n. a, not analysed Table 2 Plagioclase compositions, average concentrations (Av.) and standard deviations (S. D.) of N microprobe and n LA ICP-MS analyses Mine Khuseleka Nkwe Host rock Footwall harzburgite UG2 chromitite Pyroxenite parting Hanging wall pyroxenite Footwall pyroxenite UG2 chromitite Hanging wall anorthosite Hanging wall pyroxenite N(n) 9 (7) 79 (3) 32 (3) 45 (3) 41 (11) 45 (21) 50 (8) 98 (15) Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. SiO2 49·95 1·03 50·39 0·90 50·51 1·09 50·81 1·16 55·10 0·83 50·99 0·77 50·27 0·34 52·55 1·34 Al2O3 31·83 0·73 31·88 0·80 31·75 0·77 31·72 0·78 28·82 0·56 31·83 0·48 32·11 0·24 30·48 0·85 FeO 0·32 0·03 0·27 0·18 0·19 0·03 0·19 0·02 0·21 0·09 0·19 0·08 0·23 0·06 0·22 0·11 CaO 14·63 0·79 14·05 0·88 14·24 0·92 13·94 0·77 11·38 0·62 14·72 0·54 15·18 0·24 13·41 0·96 Na2O 2·95 0·40 3·23 0·40 3·22 0·50 3·18 0·49 4·97 0·40 2·99 0·34 2·63 0·15 3·67 0·60 K2O 0·28 0·08 0·11 0·05 0·20 0·05 0·21 0·05 0·17 0·05 0·02 0·01 0·12 0·02 0·20 0·04 Total 99·96 99·92 100·1 100·1 100·7 100·9 100·6 100·6 An 72·1 70·1 70·2 69·9 55·8 73·0 75·9 66·9 Trace elements, ppm Mg 435 241 18 5 84 13 142 15 n. a. n. a. n. a. n. a. P 217 28 316 20 465 52 435 42 n. a. n. a. n. a. n. a. Sc 0·68 0·40 0·72 0·03 1·16 0·39 0·84 0·22 0·66 0·04 0·88 0·05 0·58 0·03 0·79 0·04 Ti 115 28 11 2 87 47 103 25 138 8 60 3 138 8 125 5 Mn 17·7 15·0 2·4 0·1 8·9 2·1 5·1 1·4 6·9 0·3 3·5 0·2 17·4 0·7 7·2 0·3 Fe 2966 440 280 98 1143 88 1062 42 n. a. n. a. n. a. n. a. Sr 479 24 422 17 421 4 454 49 463 19 527 18 497 22 517 17 Ba 146 12 83 27 113 13 148 50 150 6 80 3 57 3 153 6 Rb 0·935 0·39 0·417 0·015 0·685 0·191 0·455 0·163 0·266 0·016 0·023 0·007 0·119 0·009 0·106 0·008 Y 0·209 0·07 0·433 0·179 0·415 0·134 0·330 0·057 0·183 0·013 0·271 0·021 0·192 0·015 0·322 0·018 La 3·61 0·96 5·94 2·45 5·98 1·07 3·70 0·40 10·83 0·40 5·46 0·20 1·48 0·06 7·08 0·26 Ce 8·36 1·93 10·83 3·87 9·98 1·01 6·04 0·45 13·47 0·48 8·76 0·30 2·65 0·10 10·66 0·36 Pr 0·693 0·18 1·040 0·44 0·882 0·038 0·547 0·016 0·961 0·04 0·80 0·04 0·266 0·013 0·885 0·035 Nd 2·335 0·61 3·477 1·42 2·550 0·424 2·105 0·120 2·514 0·12 2·65 0·147 0·957 0·052 2·706 0·136 Sm 0·246 0·10 0·487 0·12 0·231 0·027 0·430 0·269 0·226 0·02 0·28 0·044 0·132 0·018 0·295 0·029 Eu 0·634 0·11 0·720 0·36 0·733 0·038 0·661 0·194 0·835 0·04 0·62 0·037 0·312 0·017 0·870 0·037 Gd 0·232 0·03 0·663 0·15 0·560 0·113 0·595 0·134 0·114 0·02 0·19 0·036 0·088 0·015 0·190 0·022 Tb 0·035 0·014 0·071 0·029 0·065 0·038 0·034 0·008 0·011 0·002 0·019 0·004 0·008 0·002 0·017 0·002 Dy 0·100 0·014 0·280 0·017 0·254 0·122 0·201 0·001 0·046 0·008 0·079 0·019 0·046 0·008 0·078 0·012 Ho b. d. l. b. d. l. b. d. l. b. d. l. 0·007 0·001 0·013 0·003 0·007 0·001 0·010 0·002 Er b. d. l. b. d. l. b. d. l. b. d. l. 0·013 0·003 0·029 0·010 0·014 0·003 0·022 0·005 Tm b. d. l. b. d. l. b. d. l. b. d. l. 0·002 0·001 0·002 0·001 0·002 0·001 0·004 0·001 Yb b. d. l. b. d. l. b. d. l. b. d. l. 0·008 0·003 0·004 0·002 0·011 0·004 0·013 0·004 Lu b. d. l. b. d. l. b. d. l. b. d. l. b. d. l. b. d. l. 0·005 0·002 0·002 0·001 Pb 2·35 0·21 2·16 1·09 2·56 0·08 3·24 1·24 12·85 0·55 12·58 0·52 0·82 0·04 3·75 0·18 Mine Khuseleka Nkwe Host rock Footwall harzburgite UG2 chromitite Pyroxenite parting Hanging wall pyroxenite Footwall pyroxenite UG2 chromitite Hanging wall anorthosite Hanging wall pyroxenite N(n) 9 (7) 79 (3) 32 (3) 45 (3) 41 (11) 45 (21) 50 (8) 98 (15) Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. SiO2 49·95 1·03 50·39 0·90 50·51 1·09 50·81 1·16 55·10 0·83 50·99 0·77 50·27 0·34 52·55 1·34 Al2O3 31·83 0·73 31·88 0·80 31·75 0·77 31·72 0·78 28·82 0·56 31·83 0·48 32·11 0·24 30·48 0·85 FeO 0·32 0·03 0·27 0·18 0·19 0·03 0·19 0·02 0·21 0·09 0·19 0·08 0·23 0·06 0·22 0·11 CaO 14·63 0·79 14·05 0·88 14·24 0·92 13·94 0·77 11·38 0·62 14·72 0·54 15·18 0·24 13·41 0·96 Na2O 2·95 0·40 3·23 0·40 3·22 0·50 3·18 0·49 4·97 0·40 2·99 0·34 2·63 0·15 3·67 0·60 K2O 0·28 0·08 0·11 0·05 0·20 0·05 0·21 0·05 0·17 0·05 0·02 0·01 0·12 0·02 0·20 0·04 Total 99·96 99·92 100·1 100·1 100·7 100·9 100·6 100·6 An 72·1 70·1 70·2 69·9 55·8 73·0 75·9 66·9 Trace elements, ppm Mg 435 241 18 5 84 13 142 15 n. a. n. a. n. a. n. a. P 217 28 316 20 465 52 435 42 n. a. n. a. n. a. n. a. Sc 0·68 0·40 0·72 0·03 1·16 0·39 0·84 0·22 0·66 0·04 0·88 0·05 0·58 0·03 0·79 0·04 Ti 115 28 11 2 87 47 103 25 138 8 60 3 138 8 125 5 Mn 17·7 15·0 2·4 0·1 8·9 2·1 5·1 1·4 6·9 0·3 3·5 0·2 17·4 0·7 7·2 0·3 Fe 2966 440 280 98 1143 88 1062 42 n. a. n. a. n. a. n. a. Sr 479 24 422 17 421 4 454 49 463 19 527 18 497 22 517 17 Ba 146 12 83 27 113 13 148 50 150 6 80 3 57 3 153 6 Rb 0·935 0·39 0·417 0·015 0·685 0·191 0·455 0·163 0·266 0·016 0·023 0·007 0·119 0·009 0·106 0·008 Y 0·209 0·07 0·433 0·179 0·415 0·134 0·330 0·057 0·183 0·013 0·271 0·021 0·192 0·015 0·322 0·018 La 3·61 0·96 5·94 2·45 5·98 1·07 3·70 0·40 10·83 0·40 5·46 0·20 1·48 0·06 7·08 0·26 Ce 8·36 1·93 10·83 3·87 9·98 1·01 6·04 0·45 13·47 0·48 8·76 0·30 2·65 0·10 10·66 0·36 Pr 0·693 0·18 1·040 0·44 0·882 0·038 0·547 0·016 0·961 0·04 0·80 0·04 0·266 0·013 0·885 0·035 Nd 2·335 0·61 3·477 1·42 2·550 0·424 2·105 0·120 2·514 0·12 2·65 0·147 0·957 0·052 2·706 0·136 Sm 0·246 0·10 0·487 0·12 0·231 0·027 0·430 0·269 0·226 0·02 0·28 0·044 0·132 0·018 0·295 0·029 Eu 0·634 0·11 0·720 0·36 0·733 0·038 0·661 0·194 0·835 0·04 0·62 0·037 0·312 0·017 0·870 0·037 Gd 0·232 0·03 0·663 0·15 0·560 0·113 0·595 0·134 0·114 0·02 0·19 0·036 0·088 0·015 0·190 0·022 Tb 0·035 0·014 0·071 0·029 0·065 0·038 0·034 0·008 0·011 0·002 0·019 0·004 0·008 0·002 0·017 0·002 Dy 0·100 0·014 0·280 0·017 0·254 0·122 0·201 0·001 0·046 0·008 0·079 0·019 0·046 0·008 0·078 0·012 Ho b. d. l. b. d. l. b. d. l. b. d. l. 0·007 0·001 0·013 0·003 0·007 0·001 0·010 0·002 Er b. d. l. b. d. l. b. d. l. b. d. l. 0·013 0·003 0·029 0·010 0·014 0·003 0·022 0·005 Tm b. d. l. b. d. l. b. d. l. b. d. l. 0·002 0·001 0·002 0·001 0·002 0·001 0·004 0·001 Yb b. d. l. b. d. l. b. d. l. b. d. l. 0·008 0·003 0·004 0·002 0·011 0·004 0·013 0·004 Lu b. d. l. b. d. l. b. d. l. b. d. l. b. d. l. b. d. l. 0·005 0·002 0·002 0·001 Pb 2·35 0·21 2·16 1·09 2·56 0·08 3·24 1·24 12·85 0·55 12·58 0·52 0·82 0·04 3·75 0·18 Major oxides are in wt %, trace elements are in ppm. b. d. l., below the detection limit; n. a, not analysed Synchrotron radiation X-ray fluorescence microanalysis (SR XRF) X-ray fluorescence microanalysis of the Khuseleka drill core section was carried out at the Institute of Nuclear Physics, Siberian Branch of the Russian Academy of Sciences in Novosibirsk using a synchrotron radiation beam from the electron–positron VEPP-3 storage ring with a perimeter of 74·4 m, injection energy of 350 MeV and maximal energy of 2000 MeV. In the early 1990s, the source was equipped with a one-dimensional scanner for samples up to 300 mm in length, with a minimum step size of 10 μm. The system for collimation and focusing can produce a beam of minimum 100 μm in diameter at the excitation energy of 15–30 keV. The XRF scanner is routinely used for chemical analyses of lacustrine and marine sediments, but this study is the first application to igneous rocks. Darin et al. (2016) describe the results of trial analyses carried out on a layered petrographic thin section of the UG1 chromitite. The samples used in this study were drill core fragments 37 × 20 × 5 mm in size, polished on one side. They were fixed into the scanner and moved at 1·2 mm increments. The size of the collimated beam of excitation radiation was 1 mm along the profile and 20 mm across. Scanning was done twice at different excitation energies of 20 and 30 keV. The acquired spectra were processed using the AXIL software (Van Espen et al., 1986). The algorithm is based on nonlinear optimisation by the least squares method. The fitting functions are a set of Gaussians with the positions and mutual intensities of individual peaks corresponding to the tabulated positions and intensities of fluorescence lines for detected elements. Concentrations of elements were calculated from the intensities of their analytical lines using the external standard method. No suitable international standard was available for chromitite, so concentrations were calibrated against electron microprobe and conventional whole-rock XRF data from the same sample. Reference elements used in the calibration included K, Ca, Ti, Cr, Mn and Fe. The XRF data are given in the Supplementary Data; supplementary data are available for downloading at http://www.petrology.oxfordjournals.org. We acknowledge that absolute concentrations of some trace elements may have systematic errors that are difficult to evaluate. However, the relative variations within the chromitite seams primarily depend on the intensities of the analytical lines and not on the calibration procedure. Crystal size distribution Crystal size distribution (CSD) analysis of chromite crystals in massive chromitite seams and disseminated crystals in the silicate rocks below and above the seams was carried out using reflected light photomicrographs of polished thin sections from the Nkwe and Khuseleka profiles using 2 D image analysis and the 2D-3D conversion method developed by Higgins (2000). In the main chromitite seam, samples for the CSD analysis were taken from different levels, at the top, in the middle and at the bottom. Numerous random areas of polished petrographic thin sections were photographed under a microscope in reflected light and the digital greyscale photographs were transformed into binary images showing chromite crystals as black areas on a white background. Binary images were manually edited by removing artefacts due to polishing defects and highlighting interfaces between touching crystals. The exposed areas of all individual chromite crystals were contoured and measured by maximal Feret diameters using ImageJ software. The resulting tables of Feret diameters were fed into the CSDCorrections 1.50 software (Higgins, 2000), which calculated the conventional CSD curves and histograms. In the software input parameters, rock fabric was defined as massive, the input crystal dimension was set to the maximal size, crystal shapes were defined as equidimensional in all three axes and crystal roundness was set to zero. The minimal number of chromite grains used for the construction of the CSD curves was 520 (in Khuseleka harzburgite) and in all the other cases the number was greater than 1500. RESULTS Petrography and electron microprobe analyses The Khuseleka profile The UG2 chromitite seam is 80 cm thick and consists of euhedral to anhedral chromite crystals with an average size of about 0·2 mm (Fig. 3). The crystals are usually poikilitically enclosed by plagioclase and pyroxene oikocrysts up to 7 mm across, but locally form aggregates of larger, tightly intergrown anhedral crystals (chromitite textures and crystal size distribution are discussed in more detail in a separate section below). Chromite compositions throughout the seam have Mg# [atomic ratio Mg/(Mg+Fe2+)] at 0·45–0·52, atomic Cr/(Cr+Al+Fe3+) at 0·57–0·59 and TiO2 concentrations at about 0·9–1·2 wt %. The orthopyroxene oikocrysts have Mg# of 0·88–0·91, which is significantly more magnesian than orthopyroxene in the footwall harzburgite (0·79–0·85). Plagioclase in the chromitite seam varies within the An60–80 range, except for a narrow interval in the upper part where the anorthite content decreases to An28 (Fig. 4b). Minor interstitial minerals are mostly biotite and high-Ca clinopyroxene. Apart from exsolution lamellae in orthopyroxene, clinopyroxene locally forms isolated interstitial crystals. Fig. 3 View largeDownload slide Photographs of petrographic thin sections in (a–d) transmitted and (e–f) reflected light. (a) Footwall harzburgite, Khuseleka. (b) Sharp contact of the main chromitite seam with overlying pyroxenite parting, Khuseleka. (c) Hanging wall pyroxenite, Khuseleka. (d) A single poikilitic orthopyroxene crystal in the main chromitite seam, Nkwe, cross-polarized light. (e) Transition from poikilitically enclosed chromite to an aggregate of larger chromite crystals, Nkwe. (f) An aggregate of tightly intergrown chromite crystals with sulphides at triple junctions, Nkwe. Abbreviations for minerals: Chr, chromite; Ol, olivine; Opx, orthopyroxene; Pl, plagioclase; S, sulphide. Scale bars below the images correspond to 2 mm. Fig. 3 View largeDownload slide Photographs of petrographic thin sections in (a–d) transmitted and (e–f) reflected light. (a) Footwall harzburgite, Khuseleka. (b) Sharp contact of the main chromitite seam with overlying pyroxenite parting, Khuseleka. (c) Hanging wall pyroxenite, Khuseleka. (d) A single poikilitic orthopyroxene crystal in the main chromitite seam, Nkwe, cross-polarized light. (e) Transition from poikilitically enclosed chromite to an aggregate of larger chromite crystals, Nkwe. (f) An aggregate of tightly intergrown chromite crystals with sulphides at triple junctions, Nkwe. Abbreviations for minerals: Chr, chromite; Ol, olivine; Opx, orthopyroxene; Pl, plagioclase; S, sulphide. Scale bars below the images correspond to 2 mm. Fig. 4 View largeDownload slide Variations of orthopyroxene (a) and plagioclase (b,c) compositions in the Khuseleka drill core profile. (a-c) major element ratios; (d) concentrations of Ti. Background colours correspond to different rock types: white, pyroxenite and harzburgite; grey, chromitite. Fig. 4 View largeDownload slide Variations of orthopyroxene (a) and plagioclase (b,c) compositions in the Khuseleka drill core profile. (a-c) major element ratios; (d) concentrations of Ti. Background colours correspond to different rock types: white, pyroxenite and harzburgite; grey, chromitite. The lower contact of the UG2 main seam is sharp. The footwall rock is a coarse-grained feldspathic harzburgite (Fig. 3a) composed of euhedral olivine crystals (Fo82–83) up to 7 mm in size, euhedral to subhedral orthopyroxene grains up to 15 mm across and interstitial plagioclase (An60–81). The orthopyroxene Mg# gradually increases from 0·79 to 0·84 toward the contact with the main chromitite seam (Fig. 4a). The compositions of disseminated chromite crystals vary broadly (Fig. 5). The values of Mg# and Fe3+/total Fe in them tend to increase toward the chromitite contact, whereas the TiO2 contents decrease in the same direction and Cr/(Cr+Al+Fe3+) ratios are nearly constant (Fig. 5b–d). Biotite, amphibole and clinopyroxene are present in the rock in minor amounts as interstitial phases. Fig. 5 View largeDownload slide Variations of chromite composition in the Khuseleka drill core profile. (a-c) major element ratios; (d) concentrations of Ti. Background colours correspond to different rock types: white, pyroxenite and harzburgite; grey, chromitite. Fig. 5 View largeDownload slide Variations of chromite composition in the Khuseleka drill core profile. (a-c) major element ratios; (d) concentrations of Ti. Background colours correspond to different rock types: white, pyroxenite and harzburgite; grey, chromitite. The upper contact of the UG2 main seam is also sharp (Fig. 3b), and above it is a 23 cm-thick layer (‘parting’) of fine-grained orthopyroxenite with minor interstitial plagioclase and minor disseminated chromite. This parting separates the main chromitite seam from a thin (6 cm) chromite stringer (‘leader’ in mining terminology) above. The rock above the chromite leader is feldspathic pyroxenite with more abundant chromite than the orthopyroxenite in the parting and the chromite grains are arranged in sub-horizontal ‘chains’ (Fig. 3c). Orthopyroxene compositions in the parting and hanging wall pyroxenite are practically identical, with Mg# of 0·79–0·84 and thus less magnesian than the oikocrysts in the massive chromitite (Fig. 4a). Plagioclase compositions in the parting and the hanging wall pyroxenite vary broadly from An35 to An78, but tend to have lower An and higher K2O contents than poikilitic plagioclase in the chromitite seams (Fig. 4b and c). Chromite from the pyroxenite parting, hanging wall pyroxenite, the main seam and the leader define a linear trend of decreasing Cr/(Cr+Al) with increasing Mg/(Mg+Fe2+). This trend correlates with decreasing chromite abundance and corresponds to ‘trend A’ of Naldrett et al. (2012). By contrast, disseminated chromite crystals in the footwall harzburgite plot off the trend at lower Cr/(Cr+Al) values (Fig. 6a) implying that some other factors affected chromite compositions there. Fig. 6 View largeDownload slide Variations in the Cr/(Cr+Al) vs Mg/(Mg+Fe2+) of chromites from massive chromitite seams and silicate rocks at contacts. (a) Khuseleka, (b) Nkwe. Fig. 6 View largeDownload slide Variations in the Cr/(Cr+Al) vs Mg/(Mg+Fe2+) of chromites from massive chromitite seams and silicate rocks at contacts. (a) Khuseleka, (b) Nkwe. The Nkwe profile The UG2 chromitite layer at Nkwe Platinum mine in the eastern limb is a single, 70 cm thick seam but, as typical for the eastern limb, there is a UG3 layer positioned approximately 10 m above it. The UG2 layer is underlain by pegmatoidal feldspathic pyroxenite, whereas the hanging wall rock is a 10 cm thick, chromite-free anorthosite layer which, in turn, is overlain by pyroxenite and UG3 chromitite. Both contacts of the UG2 seam are sharp. Euhedral to subhedral chromite crystals comprise 62–72 vol. % of the seam. Like in the Khuseleka UG2 samples, chromite is either poikilitically enclosed by plagioclase and orthopyroxene oikocrysts or forms clusters of bigger, tightly intergrown anhedral crystals. However, at Nkwe the clusters are notably more abundant and larger. Towards the lower contact with pyroxenite, larger chromite crystals contain numerous rectangular to roundish inclusions, 10–70 µm in size, composed of orthopyroxene, talc, Na-rich biotite (aspidolite) and rutile. Chromite near the upper contact is inclusion-free. Similar inclusions in chromite were described by Spandler et al. (2005) and Li et al. (2005) from chromitite-footwall contacts in the Stillwater Complex and the Bushveld Complex and are interpreted as crystallized melt inclusions. Vukmanovic et al. (2013) noted the presence of chromite crystals with similar inclusions in the chromitite seam at the bottom of the Merensky Reef. The footwall pegmatoidal pyroxenite is composed of euhedral to subhedral coarse grained orthopyroxene and anhedral interstitial plagioclase. The rock also contains abundant biotite and minor disseminated chromite. Typical interstitial phases are hornblende, biotite, pumpellyite, rutile, calcite and sulphide. Similar observations in the footwall pyroxenite were reported in the eastern limb by Cameron (1975) and in the western limb by Penberthy & Merkle (1999) and Hulbert & von Gruenewaldt (1985). Mineral compositions in the Nkwe profile were reported by Veksler et al. (2015), but that work did not present vertical profiles. Briefly, orthopyroxene Mg# in footwall pegmatoidal pyroxenite is at about 0·7 and towards the contact with UG2 the value increases to 0·81. The Mg# of poikilitic orthopyroxene cementing the chromitite seam is 0·88 and the values for orthopyroxene in hanging wall anorthosite and pyroxenite are 0·77 and 0·79–0·8, respectively. Plagioclase in anorthosite is An75–77; in hanging wall pyroxenite the composition is An60–75; poikilitic plagioclase in the chromitite seam is An66–76 and in the footwall pegmatoidal rock plagioclase compositions vary broadly between An36 and An76. The composition of disseminated chromite in the footwall pegmatoid also varies broadly and generally follows ‘trend A’ of Naldrett et al. (2012). However, chromite from the chromitite seam has higher Mg# (0·37–0·53) and does not plot along the same trend (Fig. 6c). The profile shows a subtle trend towards higher Mg# and an increase of Fe3+ towards the top of the seam (Fig. 7). At three levels of the main seam, one near the top and two in the lower third, TiO2 and to some extent V2O3 show a two-fold enrichment, and at these three levels chromite shows more coarsening. Fig. 7 View largeDownload slide Variations of chromite composition in the Nkwe drill core profile. Major element ratios (a-c) and minor element concentrations (d-f). Background colours correspond to different rock types: white, hanging wall anorthosite; dark grey, chromitite; light grey, footwall pegmatitic pyroxenite. Fig. 7 View largeDownload slide Variations of chromite composition in the Nkwe drill core profile. Major element ratios (a-c) and minor element concentrations (d-f). Background colours correspond to different rock types: white, hanging wall anorthosite; dark grey, chromitite; light grey, footwall pegmatitic pyroxenite. Trace element distribution Samples in the Khuseleka profile chosen for LA-ICP-MS trace element analysis were taken from the top, middle and bottom parts of the UG2 layer, and from the footwall and the hanging wall silicate rocks. The results are presented together with a summary of the electron microprobe data in Tables 1 and 2. Trace element compositions of minerals from the Nkwe samples were reported by Veksler et al. (2015), and we use data from that publication in this study. Interstitial clinopyroxene is more abundant in the Nkwe samples and only in that profile were we able to find grains large enough for laser ablation analysis. A significant, if not the predominant, proportion of incompatible trace elements in plutonic rocks resides not in the major minerals but in interstitial or included accessory minerals, many of which form at the latest stages of crystallization and/or in the sub-solidus field. The contribution of these minerals to the trace element budget is not captured by microprobe mineral analyses of the major phases and, therefore, important information on magma evolution may be overlooked. Whole-rock analysis captures the contribution of all minerals present, but has the drawback that samples are crushed to powder and small-scale spatial variations are lost. As an alternative method for bulk analysis we tested the use of synchrotron-source XRF microanalysis (SR XRF), which provides information on a spatial scale intermediate between the in situ microbeam methods and conventional XRF, and has the advantage of being non-destructive. Orthopyroxene, clinopyroxene, and REE geothermometry Concentrations of compatible elements Cr, Co, Ni and Sc, and moderately incompatible Ti in orthopyroxene from the Khuseleka and Nkwe samples are similar and consistent with crystallization of the mineral from B1 parental magma, assuming element abundances in B1 as proposed by Barnes et al. (2010) and experimentally-determined orthopyroxene-melt D values compiled and reviewed by Bédard [(2007); see also Veksler et al. (2015) for more discussion]. Orthopyroxene oikocrysts in UG2 chromitite are depleted in Cr and Co, probably because of reaction with chromite, the same process which resulted in the higher Mg# of the oikocrysts (Figs 5 and 7). Experimental data on REE distribution between orthopyroxene, clinopyroxene and melt have been reviewed by Yao et al. (2012), Sun & Liang (2012) and Liang et al. (2013), and in Fig. 8 we compare theoretical predictions from their studies with the observed REE concentrations from UG2. If REE concentrations in the B1 parental magma proposed by Barnes et al. (2010) were correct and if the orthopyroxene retained liquidus compositions, the observed ratio of REE concentrations in orthopyroxene to the B1 values would have agreed with the orthopyroxene-melt D values. In fact, this is not the case. For the observed REE orthopyroxene B1 ratios in all rock types and both locations exceed the experimental D values by approximately an order of magnitude, and the excess is greatest for the light REE from La to Nd. Experimental D values according to Yao et al. (2012) are plotted for two temperatures, 890 and 1170°C, corresponding to the minimal and maximal estimates obtained by the two-pyroxene thermometer explained below. Notably, the mismatch between the observed concentrations and the theoretical predictions for light REE becomes greater with falling temperature. Fig. 8 View largeDownload slide REE concentrations in orthopyroxene from different lithologies at Khusleka (a) and Nkwe (b) (Table 1) normalized to the concentrations in the model parental liquid B1 (Barnes et al., 2010). (a) Khuseleka; (b) Nkwe. The elements are in the order of decreasing orthopyroxene-melt distribution coefficient. Experimentally determined orthopyroxene-melt D values from Yao et al. (2012) depend on temperature and are shown for 890 and 1170°C by dashed and solid lines. Fig. 8 View largeDownload slide REE concentrations in orthopyroxene from different lithologies at Khusleka (a) and Nkwe (b) (Table 1) normalized to the concentrations in the model parental liquid B1 (Barnes et al., 2010). (a) Khuseleka; (b) Nkwe. The elements are in the order of decreasing orthopyroxene-melt distribution coefficient. Experimentally determined orthopyroxene-melt D values from Yao et al. (2012) depend on temperature and are shown for 890 and 1170°C by dashed and solid lines. Liang et al. (2013) used experimental data on REE partitioning between coexisting orthopyroxene, clinopyroxene and melt to calibrate a two-pyroxene REE thermometer. They applied the thermometer to seven samples of harzburgites and pyroxenites from the Lower and Lower Critical zones of the Bushveld Complex analysed by Godel et al. (2011) and obtained equilibration temperatures between 1080 and 1230°C. They noted that these values were much higher than those calculated using the two-pyroxene Ca exchange thermometer of Brey & Köhler (1990), and attributed the difference to the relatively fast cooling rate of the Bushveld rocks and lower closure temperature for Ca diffusion than for REE. Our application of the REE and Ca exchange thermometers to the UG2 samples produced similar results to Liang et al. (2013) (Fig. 9). Unfortunately, clinopyroxene crystals large enough for laser ablation are sparse in our samples of the UG2 layer and we were able to analyse only three samples from Nkwe. Equilibration temperature in the hanging wall pyroxenite at 1174 ± 19°C is in a good agreement with the crystallization temperature of the Upper Critical Zone according to experimental studies (Cawthorn & Davies, 1983). Equilibration temperature in footwall pyroxentite at 1288 ± 12°C appears to be too high for the Upper Critical Zone, whereas the equilibration temperature calculated for the UG2 chromitite (889 ± 29°C) is low. Furthermore, the two-pyroxene distribution of light REE elements (La to Pr) deviates from the predicted equilibrium trend towards greater enrichment of orthopyroxene. Similar deviations were observed in Bushveld samples by Liang et al. (2013) and those elements were excluded from the temperature calculations, as in our example. Equilibration temperatures according to the Ca exchange thermometer give consistently low values in the 840–885°C interval for all samples, which is also in agreement with Liang et al. (2013). Fig. 9 View largeDownload slide REE distribution between orthopyroxene and clinopyroxene and the results of the two-pyroxene thermometry for hanging wall pyroxenite (a), chromitite (b) and footwall pyroxenite (c) at Nkwe. The observed orthopyroxene–clinopyroxene concentration ratios (D Opx/Cpx) are plotted as open circles; the predicted values according to Yao et al. (2012) and Sun & Liang (2012) are plotted as grey solid lines. Temperature estimates TREE are according to the two-pyroxene REE thermometer (Liang et al., 2013); TBKN are according to the thermometer based on Ca distribution (Brey & Köhler, 1990). Fig. 9 View largeDownload slide REE distribution between orthopyroxene and clinopyroxene and the results of the two-pyroxene thermometry for hanging wall pyroxenite (a), chromitite (b) and footwall pyroxenite (c) at Nkwe. The observed orthopyroxene–clinopyroxene concentration ratios (D Opx/Cpx) are plotted as open circles; the predicted values according to Yao et al. (2012) and Sun & Liang (2012) are plotted as grey solid lines. Temperature estimates TREE are according to the two-pyroxene REE thermometer (Liang et al., 2013); TBKN are according to the thermometer based on Ca distribution (Brey & Köhler, 1990). Plagioclase In Fig. 10, we plot trace element concentrations in plagioclase normalized to the element abundances in the hypothetical B1 parental melt (Barnes et al., 2010) and compare the normalized values with experimental plagioclase-melt distribution coefficients compiled and reviewed by Bédard (2006). The comparison is aimed at testing whether the observed trace element concentrations are primary magmatic and consistent with plagioclase crystallization from the B1 liquid. A close match between the B1-normalized concentrations and experimental plagioclase-melt D values is observed for all analysed trace elements, with the exception of Pb (Veksler et al., 2015), in plagioclase from the hanging wall anorthosite at Nkwe (Fig. 10b). Good consistency with magmatic crystallization from the B1 parental magma is also observed for all the studied samples, with remarkably constant concentrations of the most compatible elements in plagioclase (Sr and Eu). In contrast, REE concentrations in interstitial plagioclase from pyroxenite, chromitite and harzburgite show enrichment relative to the predicted curve, consistent with REE accumulation in residual intercumulus liquid (trapped liquid shift effect) and possibly also an increase of plagioclase-melt DREE values with falling temperature (Bédard, 2006). Poikilitic plagioclase from the main UG2 seam is strongly depleted in K and Ti relative to the mineral in the silicate rocks. The depletion in Ti is probably due to diffusion of that element to chromite. The depletion in K, which is evident in Fig. 4c, was first described in detail by Veksler et al. (2015) and is attributed to alkali migration out of the chromitite layer during postcumulus crystallization (see discussion). Fig. 10 View largeDownload slide Trace element concentrations in plagioclase from different lithologies at Khusleka (a) and Nkwe (b) normalized to the concentrations in the model parental liquid B1 (Barnes et al., 2010). The elements are in the order of decreasing plagioclase-melt distribution coefficient. Grey solid line in (a) shows the pattern of experimentally determined plagioclase-melt D values for the plagioclase composition An70. Grey field in (b) shows the range of the D values increasing from An76 to An55. All the D values are from Bédard (2006). Fig. 10 View largeDownload slide Trace element concentrations in plagioclase from different lithologies at Khusleka (a) and Nkwe (b) normalized to the concentrations in the model parental liquid B1 (Barnes et al., 2010). The elements are in the order of decreasing plagioclase-melt distribution coefficient. Grey solid line in (a) shows the pattern of experimentally determined plagioclase-melt D values for the plagioclase composition An70. Grey field in (b) shows the range of the D values increasing from An76 to An55. All the D values are from Bédard (2006). Synchrotron radiation XRF results Concentration data collected during the SR XRF scanning were averaged for each 10 analyses, so each point on the vertical profiles in Fig. 11 and the Supplementary Data represents integration over an area of approximately 2 cm2, 10 mm along the profile and 20 mm across. The elements V, Zn and Ga are compatible in chromite and the distribution of these elements in the profile show a very high covariance and a strong correlation with Cr2O3 as a proxy for the modal amounts of chomite (Fig. 11a–d). However, a closer look at the V variations shows an increase in the upper part of the main seam and even higher concentrations of V in the chromitite leader. We have no synchrotron data from the Nkwe profile, but note that the microprobe data for chromite also shows an abrupt increase in V (and Ti) in the upper section (Fig. 7e). An increase of V concentrations at the top of the UG2 layer was also noted by Naldrett et al. (2012). Fig. 11 View largeDownload slide Concentrations of selected trace elements (a-n) in the Khuseleka profile analysed by synchrotron XRF. The background colours correspond to different rock types: white, pyroxenite and harzburgite; grey, chromitite. Fig. 11 View largeDownload slide Concentrations of selected trace elements (a-n) in the Khuseleka profile analysed by synchrotron XRF. The background colours correspond to different rock types: white, pyroxenite and harzburgite; grey, chromitite. The pattern of incompatible Rb, Y and Zr variations is of special interest because of their potential to indicate the distribution of interstitial residual liquid. In the silicate layers, these elements have consistently low concentrations in footwall harzburgite, and higher but variable contents in the pyroxenite parting and the hanging wall pyroxenite (Fig. 11e–g). Notably, Zr and Y show a close covariance throughout the profile, probably reflecting a specific mineral host (zircon is present as an accessory phase). Rubidium correlates generally with Zr and Y, but less strongly since it is likely associated with interstitial mica or amphibole. Inside the main UG2 chromitite, all three elements show heterogeneous and correlated variations, with peaks at the bottom and near the top of the layer, but as in the wall-rocks, Rb has a more independent behaviour than Zr and Y. In view of their different mineral hosts, the correlation between Rb and Zr or Rb and Y is remarkable and probably implies a greater proportion of trapped interstitial liquid near the contacts of the main seam. Two other incompatible elements, Nb and Mo, show a different distribution pattern. Both have low and highly variable concentrations, but they show a strong covariance. The correlation and erratic concentrations suggest that both elements are hosted by the same accessory phase or phases. Rutile is a logical host for Nb. It occurs in our samples as small needles and anhedral grains within, and interstitial to, chromite crystals, and has been described in association with chromite elsewhere in the Upper Critical Zone (e.g.in the UG2: Junge et al., 2014; Merensky chromitite: Vukmanovic et al., 2013). Notably, Mo shows no correlation with Cu, Pd or As, so it is apparently not concentrated in sulphides, and given the strong correlation with Nb we suggest it may also be hosted in rutile. We found no information on Mo partitioning in natural rutile, but the idea is supported by the existence of Mo-doped synthetic TiO2 polymorphs, which are used in photocatalysis (e.g.Devi & Murthy, 2008). The highest concentrations of Ni are observed in the footwall harzburgite, where it mostly resides in olivine, and in a sulphide-rich zone near the top of the main chromitite seam, where Ni probably resides in pentlandite (Fig. 11j). Sulphide enrichment in that part of the profile is confirmed by reflected light microscopy and by the correlated strong peaks of Cu, Pd and As concentrations (Fig. 11k, l and n). Note that the distribution patterns of Pd and Ru are significantly different. In particular, Ru shows no spike in the sulphide zone. Ruthenium is known to form laurite (RuS2) inclusions in chromite (e.g.Maier et al., 1999), and it does show a correlation with As in much of the main seam and the chromite leader. However, it is not concentrated with Cu and Pd in the sulphide zone. The decoupling of Pd and Ru was described in the larger study of PGE in Bushveld chromitites by Naldrett et al. (2012), who concluded that they are controlled by different processes, Ru being incorporated during chromite crystallization and Pd being concentrated in an interstitial sulphide melt. Textural types of chromite and crystal size distribution Following Cameron (1969, 1975) and earlier works cited therein, we distinguish three types of textural relationships between chromite and major rock-forming silicates in the UG2 layer. The first type is what Cameron (1975) called inter-grain chromite. This texture is observed in the rocks above and below the main chromite seams where chromite is interstitial to silicate crystals. In our samples the ‘inter-grain’ chromite forms networks surrounding silicate crystals or is aligned in sub-horizontal chains (Fig. 3c). The textural types 2 and 3 are characteristic of the main UG2 seam and (in Khuseleka) the upper leader. The most common of the two types comprises chromite grains enclosed by orthopyroxene and plagioclase oikocrysts reaching a few centimetres in size (Fig. 3d and e). In the oikocrysts, size and the volume proportion of chromite tend to increase from the centre to the rim and the centre may be completely inclusion-free. In textural type 3, chromite forms aggregates of tightly intergrown, anhedral crystals practically devoid of silicates. The size of individual chromite crystals in the aggregates is notably larger than in the other two textural types and the grain boundaries are curved (Fig. 3e and f), which is a good indication for textural annealing, as discussed in detail below. The chromite aggregates commonly surround orthopyroxene oikocrysts with chromite-free cores. Sulphides are often found in the aggregates at triple junctions between anhedral chromite crystals (Fig. 3f). The coarse chromite aggregates are present in samples from both studied profiles, but they are much more abundant at Nkwe. Their distribution is uneven in the profiles, and the proportion of chromite forming aggregates vs the poikilitic texture may change, even within the area of a standard thin section. However, we do not observe major changes in texture and modal composition between the upper and lower parts of the UG2 layer like those documented by Lee (1996) and Mathez & Mey (2005). The three textural types of chromite need consideration in the analysis of crystal size distribution because the ability of chromite grains to adjust in shape and size during post-cumulus processes differs depending on whether or not they are enclosed in silicate minerals (see discussion). We determined CSD curves for chromite in the main UG2 chromitite seam and adjacent wall-rocks at Khuseleka, but the wall-rocks at Nkwe contained too few crystals for a statistically valid analysis. The number of crystals processed in the samples varied from 500 in footwall harzburgite at Khuseleka to around 8000 in the main seam from both profiles. Note that the CSD measurements were made on several thin sections from all parts of the UG2 seam. In both profiles, we found no variations in the shapes of the CSD curves from the lower and upper parts of the seam and, therefore, have pooled the data in the curves on Fig. 12. All data from the silicate layers refer to texture type 1 (inter-grain). The distinction between textural types 2 and 3 (poikilitic vs aggregate) in the chromitite seams was made by the presence or absence of triple junctions. Thus, all chromite grains in chromitite forming triple junctions were selected as belonging to the aggregates and all others were attributed to the poikilitic group. At Khuseleka, there were not enough grains in the aggregate type to calculate statistically valid CSD curves and, therefore, all chromite grains were analysed together. In the Nkwe chromitite the proportion of crystal aggregates is high and we generated separate CSD curves for the poikilitic and aggregate chromite types (Fig. 12b). Fig. 12 View largeDownload slide Crystal size distribution diagrams for chromite in massive chromitite seams and silicate rocks. (a) Khuseleka; (b) Nkwe. See text for discussion. Fig. 12 View largeDownload slide Crystal size distribution diagrams for chromite in massive chromitite seams and silicate rocks. (a) Khuseleka; (b) Nkwe. See text for discussion. For reference, the ideal CSD in magma due to nucleation and growth would be a straight line with a negative slope intersecting the vertical axis at a point corresponding to the nucleation density (Marsh, 1988). For UG2 chromite, the CSD curves show the maximum population density at about 0·1–0·15 mm and a relative deficiency of smaller crystals. The Khuseleka CSD results (Fig. 12a) show differences in the position and slope of curves from the wall-rocks vs the main UG2 seam. Both the footwall and hanging wall curves have a deficiency of grain sizes less than 0·1 mm, but the CSD curve for larger grains is linear as in the ideal kinetic distribution mentioned above. By contrast, the CSD curve for UG2 has a non-linear slope with an upwards-concave shape, the curvature increasing in the grain size range above 0·6 mm. The Nkwe profile lacks data for the wall-rocks because of the paucity of chromite. The CSD curve for chromites of type 2 (poikilitic) is very similar to that of the Khuseleka chromite (also dominantly type 2), whereas the curve for chromite aggregates from Nkwe is quite different, reflecting the much larger proportion of coarse-grained chromite in the aggregates. Notably, all CSD curves from the UG2 main seam are not linear, but show an upward-concave shape which suggests textural annealing as discussed in a separate section below. DISCUSSION In the following discussion we address some of the current questions concerning the origin and evolution of the UG2 chromitite, especially the presence or absence of multiple magma injections and the ability of the trapped liquid shift effect to account for the chemical composition and variability in the layer. First, it is worth pointing out that the two profiles presented here are typical for the UG2 layer in the Bushveld Complex, including such macroscopic features as one or more leaders above the main seam (Khuseleka), sharp contacts and abrupt changes in mineral chemistry between chromitite and adjacent silicate rocks (both profiles), coarse pegmatoidal rocks in the footwall with abundant amphibole and mica (both profiles), thin anorthosite layers in immediate contact with chromitite (Nkwe) and poikilitic texture in the main seam (both profiles). Other detailed studies of UG2 reported similar findings (Hiemstra, 1985, 1986; Cawthorn & Barry, 1992; Mathez & Mey, 2005; Mondal & Mathez, 2007; Maier & Barnes, 2008; Voordouw & Beukes 2009; Voordouw et al., 2009; Junge et al. 2014; Mungall et al., 2016). There are differences between the Khuseleka and Nkwe profiles that have been described above. These include the lack of an upper leader and the presence of a thin hanging-wall anorthosite at Nkwe. Low-temperature interstitial minerals such as clinopyroxene, amphibole and mica are more abundant at Nkwe, the maximum Mg# of chromite is lower (but Cr# is the same) and the concentrations of incompatible trace elements in orthopyroxene and plagioclase are more variable than in Khuseleka samples (Figs 8 and 10). All of these features suggest that the UG2 layer at Nkwe underwent more chemical re-equilibration than at Khuseleka and the chromite CSD results also support a greater degree of textural evolution. The UG2 as a composite layer Cawthorn (2011) proposed, on the basis of observations by Lee (1996) and the pattern of PGE distribution, that the UG2 chromitite appeared to be a composite layer comprizing at least 3 parts. Junge et al. (2014) described rhythmic variations in Cr/(Cr + Al) ratios and TiO2 contents in a vertical profile of UG2 in the Karee mine on the western limb, and interpreted these as cryptic layering resulting from multiple episodes of magma replenishment and fractional crystallization during the formation of the seam. The eight sub-layers suggested by Junge et al. (2014) can possibly be questioned because although average values do vary, the range of individual samples between cycles overlap. Before our study, there were no comparable fine-scale microprobe profiles through the full UG2 layer to confirm these findings. Our results from the Khuseleka profile and from Nkwe on the eastern limb show no fine-scale rhythmic variations of the Mg# and Cr# (Figs 5a and b and 7a and b) so they are certainly not a universal feature. Indeed, the lower 25 cm of the Khuseleka profile shows remarkably stable chromite compositions (Fig. 5). However, we note greater compositional variations in chromite and poikilitic silicate crystals, well beyond analytical scatter, in the upper part of the main seam in both the Khuseleka and Nkwe profiles. These differences in mineral compositions between the upper and lower portions of the main seam and frequent presence of hanging wall leaders, e.g.at Khuseleka, appear to support the inference of Cawthorn (2011) that UG2 represents a composite layer. However, it is not clear that these variations can be explained by multiple episodes of chromite crystallization and accumulation, or different degrees of low-temperature re-equilibration. As discussed in the next section, we believe the evidence for the latter is very strong. The trapped liquid shift effect The concept of trapped liquid shift effect emerged from the fact that, as a rule, mineral compositions in crystallized plutonic rocks differ from the original compositions because of re-equilibration with intercumulus liquid or indeed in the subsolidus if cooling is slow enough. In a densely packed crystal mush where the volume proportion of crystals exceeds that of the liquid, movement of liquid relative to crystals practically stops and, with further cooling, the mush is expected to crystallise as a nearly closed system. Mineral compositions in the final product should depend on the kinetics of mineral-melt reaction and the modal proportions of the crystal and liquid phases. Since most UG2 minerals are solid solutions, slow exchange rates would produce, at a given cooling rate, compositionally zoned crystals that may retain original compositions in their cores, whereas fast exchange rates would result in homogeneous grains whose compositions differ from the original cumulus compositions (Cawthorn, 2015). The magnitude of the trapped liquid shift effect should increase with increasing mass proportion of the trapped liquid to the crystal phases. A quantitative model proposed by Barnes (1986a) emphasizes this relationship, and application of the model to the layered rocks of the Bushveld Complex showed that major and trace element variations in the main rock-forming silicates in various zones of the Complex could be explained by the trapped liquid shift (Cawthorn, 1996, 2015). With respect to UG2, the trapped liquid shift effect has been invoked for explaining why chromites in massive chromitite seams have a constant and generally less evolved composition than in the adjacent silicate rocks (Cameron, 1977). Our results support this general idea and many observations fit the simple model. However there are several features, especially in the trace element contents of major minerals, which are inconsistent with the closed-system trapped liquid scenario. It is important to recall that mineral–melt reactions in a multiply-saturated magma involve all phases in the assemblage. In a previous study of the Upper Critical Zone, Veksler et al. (2015) discussed chromite–melt reactions in detail and showed that the reactions actively involve plagioclase, orthopyroxene and olivine, when the latter is present. Naldrett et al. (2012) also recognized the importance of other crystal phases and pointed out the key role of plagioclase buffering the activity of alumina to produce their trends A and B in the chemical evolution of Bushveld chromite. Figures 5–7 show that the compositions of disseminated chromite in silicate rocks are shifted from those in the main seam and in the leader. Moreover, the compositions are shifted to a different extent and in different directions, with no apparent relationship to the modal mineralogy of the layers. Thus, the presence or absence of plagioclase and the chromite: melt mass ratio may be important factors (Naldrett et al., 2012), but they cannot be the only ones controlling chromite re-equilibration. The observation of different chromite trends in different lithologies is important because it means that the interstitial liquid in those parts of the crystal mush may also have differed in composition, which implies development of chemical gradients. Chemical gradients are driving forces for diffusion and kinetic effects related to differences in diffusion rates for different components in different phases may come into play in their presence. Veksler et al. (2015) discussed connections between redox gradients and alkali diffusion in the interstitial liquid between modally-contrasting parts of the crystal mush. The results of the present study confirm an anomalous mobility of the alkali elements as evidenced, for example, by a strong depletion of K in plagioclase from the chromitite relative to the wall-rocks (Figs 4c, 7c and 10). Chemical gradients and differential mobility of components during evolution of a layered crystal mush can lead to the situation where a mineral dissolves in one layer but continues to crystallize in another, thus sharpening the boundary between them (see Veksler et al. 2015). If dissolution keeps pace with crystallization in adjacent layers, the overall process may be nearly neutral in terms of thermal energy. These processes of migration of chemical components in the crystal mush and across its boundaries violate the closed-system assumption of the conventional trapped liquid shift model proposed by Barnes (1986a). In real systems, modal layering is not a fixed starting point, but an evolving and growing feature, partly shaped by post-cumulus mineral-melt reactions. The behaviour of trace elements The trapped liquid shift effect has predictable effects on the incompatible element contents of crystals in the evolving mush and these are examined in the REE plots of Figs 8 and 10. The point of reference for this discussion is the composition of the parental melt from which the Upper Critical Zone crystallized. In this discussion we use the composition of the B1 magma from Barnes et al. (2010) and assume that this is appropriate for both the Nkwe (eastern limb) and Khuseleka (western limb) localities. There is some debate about the parental magma for UG2, but many workers consider B1 to be a good approximation for the initial Critical Zone magma (e.g.see Cawthorn, 2007). The results of this study show that the REE distribution between orthopyroxene, clinopyroxene and plagioclase deviates from that predicted by experimentally determined mineral-melt distribution coefficients (D) from Bédard (2006, 2007), Sun & Liang (2012) and Yao et al. (2012), and the B1 parental melt composition. In general agreement with the trapped liquid shift effect, interstitial plagioclase in pyroxenite, harzburgite and chromitite is enriched in moderately incompatible REE from La to Nd, but contrary to expectations, it is not enriched or, locally, even depleted in the strongly incompatible element Y (Fig. 10). Orthopyroxene also has higher concentrations of light REE from La to Pr than expected for crystallization from B1 parental magma (Fig. 8), in qualitative agreement with the trapped liquid shift. However, there is a discrepancy between the light REE contents of orthopyroxene and coexisting clinopyroxene (Fig. 9). Orthopyroxene and plagioclase should both be equally affected by re-equilibration with residual liquid so the discrepancy in REE distribution between them might be due to changes of the plagioclase-melt and orthopyroxene-melt DREE values with temperature and, or, melt composition. Chutas et al. (2012) reported a ‘disequilibrium’ distribution of Sr and Pb between plagioclase and orthopyroxene throughout the Lower and Critical Zones, and it is possible that the ‘odd’ REE distribution between plagioclase and pyroxenes resulted from the same processes. Notably, Sr is compatible with plagioclase, but strongly incompatible with orthopyroxene, and, therefore, the trapped liquid shift should have a much greater effect on Sr concentration and isotope composition in the latter mineral. Permeability of the UG2 layer The permeability of the UG2 crystal mush and its changes during late-stage crystallization is obviously important for assessing the possibility of material transport inside and out of the layers. Mondal & Mathez (2007) postulated a low permeability of the UG2 chromitite and proposed that upwards-migrating volatile-rich melt was trapped below UG2 to form the footwall pegmatoid. Conventional models of the trapped liquid shift effect also consider the chromitite and adjacent layers as closed, effectively separate systems, which implies a very low permeability (Barnes, 1986a; Cawthorn, 1996). Manoochehri et al. (2015) used whole-rock concentrations of strongly incompatible elements and micro-X-ray tomography to evaluate the paleo-porosity and extent of compaction in a 20 m section of the Upper Critical Zone including the UG2 and UG3 chromitites. Their results indicate a rather complex picture. The ‘final porosity’ (trapped melt fraction) in UG2 and UG3 chromitite based on the average of six incompatible elements was 0.12 – 0.36, generally about the same or slightly higher than the final porosity of typical pyroxenite. However, the authors pointed out that the permeability of chromitite was possibly lower, but hard to determine precisely because it strongly depends on the crystallization of poikilitic orthopyroxene and plagioclase, which could block permeability and hamper compaction. Our synchrotron XRF results for the full UG2 profile at Khuseleka (Fig. 11) show a discrete clustering of the incompatible elements Zr, Y Rb, Nb and Mo near the bottom and, especially, the top of the seam. Manoocherhi et al. (2015) noted a clustering of accessory apatite in chromitite at a centimetre scale, which resulted in strong variations of P content from one sample to another. Phosphorous was not measured in our SR-XRF profile, but the clustering of other incompatible elements probably also reflects the presence of particular accessory minerals (e.g.zircon, mica, rutile) in locations where the final stage of interstitial crystallization took place. Note that the Nb and Mo concentration peaks, which are interpreted to reflect accumulations of rutile, are in different parts of the profile from the Zr, Y and Rb peaks, which are likely to mark zircon and mica accumulations. This implies that material transport was still active within the UG2 layer at the time when the accessory minerals crystallized. In the above context, it may be significant that the two-pyroxene REE thermometry in UG2 from Nkwe records a lower equilibration temperature within UG2 compared to the adjacent silicate layers (Fig. 9), implying that interstitial crystallization and equilibration proceeded longer in the chromitite layers than in the wall-rocks. The few samples analysed makes this a preliminary finding and it would be important for future studies to check if the low closure temperature is a general feature of the Bushveld chromitite layers. The mechanisms of textural equilibration In theory, the primary size distribution of chromite crystals due to the nucleation and growth kinetics could have been modified by mechanical processes (e.g.compaction and crystal sorting by gravity) and processes of textural equilibration. The latter are driven by the topological requirements of space-filling and by the need to minimise the total surface energy at grain boundaries. Because the surface energy at chromite–silicate interfaces is higher than at chromite–chromite and silicate–silicate interfaces, the formation of monomineralic crystal aggregates is beneficial. In the context of chromite textural evolution one should keep in mind that the very low contents of Cr in silicate melts [0·05 wt % Cr2O3 in the B1 magma at the time of Upper Critical Zone formation, according to Barnes (1986b)] effectively rules out net addition by new in situ nucleation and growth, so that the total mass of chromite in the rock during textural equilibration should stay practically constant. According to theoretical models, the crystal size distribution due to nucleation and growth is log-linear and such a distribution would be represented on Fig. 12 by a straight line with a negative slope (Marsh, 1988). The Y-intercept corresponds to the nucleation density and the slope is defined by the reciprocal product of the growth rate and the growth time, so that the greater the product, the shallower the slope. The UG2 CSD curves are clearly not straight and imply significant modification of the primary distribution. All curves from both profiles show a depletion of the smallest crystals in the chromite population relative to the ideal distribution. Pressure–solution, compaction and accumulation by gravity could explain the lack of smaller crystals but those processes operating alone would produce convex-upward curvature in the size distribution of larger crystals (Higgins, 2015), whereas we observe concave upward curves in both profiles. Also, in the case of accumulation and crystal sorting by gravity, we would expect to see differences in the crystal size distribution from the top to the bottom of the UG2 layer, but no significant changes were observed. The concave-upward curvature of the CSD curves, which is especially well developed in chromite forming aggregates at Nkwe (Fig. 12b), implies that the growth rate, growth time, or both increased with the grain size. This, together with the deficit of small crystals, implies textural adjustment by crystal coarsening, whereby larger crystals grow at the expense of smaller ones. Apart from grain coarsening, textural equilibration would also minimise, where possible, the area of high-energy interfaces between chromite and silicates. This will be ineffective in poikilitic grains, but should enhance the formation and growth of chromite aggregates. The aggregates of intergrown chromite crystals (Fig. 3f), thus appear to represent an advanced stage of textural equilibration. Their large size is in agreement with the theory of so-called ‘normal’ grain growth that applies to single phase polycrystalline systems, whereas ‘abnormal’ growth refers to large grains enclosing a matrix of much smaller crystals, the situation of pokilitic chromitite texture (Fig. 3d). Theories of normal and abnormal grain growth have been actively developed in material science and especially in metallurgy where annealing is important and carefully controlled (e.g. Harker & Parker, 1945; Smith, 1964; Atkinson, 1988). The characteristic features of crystal aggregates shaped by normal growth are curved grain boundaries intersecting in triple junctions with constant dihedral angles of 120°. The uniform dihedral angles arise from the requirement of minimal surface energy, whereas the curvature of crystal faces is a consequence of topological restrictions and the fact that there is no regular polyhedron with plane faces satisfying the optimal dihedral angles in 3D. Normal crystal growth in the aggregates proceeds by the movement of curved grain boundaries in the direction towards the centre of the curvature. Thus, crystals with concave faces expand, while neighbours with convex faces shrink and eventually disappear. Therefore, theory implies three different growth rates for chromite crystals in a chromitite layer when interstitial melt is still present. The growth rate is practically zero for chromite crystals in silicate oikocrysts. Crystals that are still in contact with the melt may grow at a rate limited by material transport in the melt phase and/or the kinetics of dissolution-precipitation at crystal-melt interfaces. In view of the very low amount of Cr in the melt, dissolution of smaller chromite crystals will be required for significant coarsening of the crystal population because the total chromite amount in the layer will stay constant. The third and fastest growth rate corresponds to the merger of crystals into one by migration of chromite–chromite grain boundaries (normal grain growth). The concave upward slopes of CSD curves for the poikilitic chromite inclusions may thus be attributed to the variable growth times depending on their position in the host crystals, with the longer growth time for the larger crystals at the rims of the oikocrysts than in their centres (Fig. 3d). As expected, the highest growth rate is for the intergrown chromite aggregates (Fig. 12b). The concave CSD curvature in the aggregates means that the product of growth rate and growth time increases with crystal size. Since the growth rate should increase as soon as two chromite crystals form a shared interface and undergo ‘normal growth’ mode, the concave curvature may indicate that largest crystals in the aggregates were in the mode of normal grain growth for a longer time. Comparative CSD studies of chromitite occurrences in or outside the Bushveld are sparse and, unfortunately, the calculation methodologies are not standardized, making direct comparison of CSD curves difficult. The closest comparison is the study by Vukmanovic et al. (2013) of thin chromitite stringers at the upper and lower contacts of the Merensky Reef. They found two crystal populations in the lower stringer and a single one in the upper chromitite. Their analysis of the CSD curves used a different input parameter (equivalent circle diameter vs the Feret diameter), which may affect the absolute values, but not the shapes of the CSD curves. The CSD curve for the lower Merensky stringer, is complex because of the two populations inferred by Vukmanovic et al. (2013), but the CSD curve for the upper stringer is concave-upwards much like our results. O’Driscoll et al. (2010) performed CSD analysis on chromite in the Rum layered intrusion in Scotland using the same measurement, conversion and calculation methods as in our study. In comparison with Bushveld, the population density maxima for Rum are at finer grain sizes, between 0·013 and 0·05 mm, and the right-hand slopes are much steeper. The latter are probably due to shorter time available for textural coarsening in Rum than in Bushveld. Jackson (1961) and Waters & Boudreau (1996) measured chromite size distributions in the Stillwater intrusion (Montana, USA), but their methods differ from those used in this study and cannot be directly compared. Qualitatively, however, the size distribution of chromite from Stillwater published by Waters & Boudreau (1996) looks similar to our results, i.e.coarser and more texturally evolved than in the case of Rum. To conclude, the processes of textural equilibration seem to account for the characteristic features of all observed CSD curves within the chromitite layer. The implication is that even if crystal settling and gravity sorting played a role in UG2 formation, as suggested, for example, by Kinnaird et al. (2002), or Maier et al. (2013), the textural evidence for these primary processes was overprinted in both profiles by subsequent recrystallization and crystal coarsening. Overall, the textural equilibration is more advanced at Nkwe than at Khuseleka, and this appears to correlate with the greater abundance at Nkwe of low-temperature interstitial minerals (e.g.clinopyroxene, amphibole, biotite) and with the slightly more evolved compositions of the major phases (e.g.the lower maximal Mg# of chromite). CONCLUSIONS There is an ongoing debate about how the chromitite layers in the Upper Critical Zone of the Bushveld complex formed, but no-one would argue that the mineral compositions and textures within the layered rocks are primary magmatic. Major adjustments took place by a combination of compaction, textural annealing and the reaction of early-formed crystals with the interstitial, residual melt. To better understand these late-stage processes, the c.1 metre-thick UG2 chromitite layer and its immediately surrounding wall-rocks were studied with millimetre-scale resolution in two continuous profiles from drill core samples: one is from the western limb at Khuseleka mine and the other is from the Nkwe mine on the eastern limb. In both cores, the major and minor element composition of chromite and the main silicates were determined by microprobe across the entire profile and chromite textures were analysed in terms of their crystal size distribution (CSD) in the same contiguous sets of thin sections. In addition, a new application of synchrotron XRF scanning was used to provide continuous trace element profiles of the Khuseleka core. Chromite is an accessory mineral in the silicate rocks above and below UG2, forming interstitial crystals with a considerable range of Mg#, Cr# and TiO2 contents, which contrasts with the rather constant compositions in the chromitite layer. Most of the chromite in the UG2 seam is included in anhedral, centimetre-sized orthopyroxene and plagioclase oikocryts. Chromite also occurs as tightly-intergrown aggregates with curved grain boundaries and 120° triple junctions. The grain size in the aggregates is larger than in the oikocrysts and we attribute this to late-stage annealing, which is supported by the shape of the CSD curves for the different textural types. The annealing texture is more pronounced in the Nkwe samples than at Khuseleka, but the chemical composition of the two chromite types are the same. Vertical microprobe profiles show nearly constant Mg# and Cr# in the lower half of the layer and more variable compositions in the upper part. The mineral compositions in the upper part of these profiles are variable but erratic, and we attribute the heterogeneity to local re-equilibration effects rather than to multiple charges of magma or chromite slurries. The trace element data for orthopyroxene, clinopyroxene and plagioclase from within the UG2 layer differ from those in the adjacent silicate layers. The latter are generally consistent with an origin from a B1-like parental magma, but the compositions within the chromitite layer have higher incompatible element contents. This discrepancy can be explained by equilibration with a more evolved, interstitial melt (the trapped liquid shift effect), but the shift is not constant for all incompatible elements and the premise of a closed-system evolution of the crystal mush is not entirely valid. Anomalously low concentrations of K, in particular, suggest selective migration of alkali elements out of the chromite layer in response to chemical potential gradients across the boundary with adjacent layers (see Veksler et al., 2015). An important and hitherto under-used application of trace element data from Bushveld rocks is the two-pyroxene-melt REE geothermometer (Liang et al., 2013), which yielded equilibration temperatures of 1170–1290°C for the silicate layers in the Nkwe profile, but only 890°C for the UG2 itself. This suggests that REE equilibration in the chromitite continued to a much lower temperature and possibly into the sub-solidus stage. Our synchrotron radiation XRF results prove the value of this method for non-destructive and high-resolution profiles of trace element distribution in drill cores. Correlated peaks in the concentration of incompatible elements Zr, Y, Rb, Nb and Mo occur near the bottom and the top of the UG2 layer. We attribute these to local concentrations of accessory zircon, mica and rutile, which may indicate the distribution of the last pockets of residual melt in the chromite crystal mush. The distribution of ore elements Ni, Cu, Pd, Ru and As is also imaged by this method. The concentration of Ni, Cu and Pd in a single zone near the top of UG2 is explained by hosting of these elements in pentlandite, whereas Ru shows a more diffuse distribution that reflects its distribution in primary Ru sulphide inclusions in chromite. Acknowledgements We are grateful to Dieter Rhede, Oona Appelt (GFZ Potsdam) and Helene Braetz (University of Erlangen) for their help with electron microprobe and laser ablation analyses. Tawanda Manyeruke and others at Nkwe Platinum are thanked for help and access to sampling the drill core, and permission to publish the results. Thoughtful and detailed reviews by Marian Holness, Michael Higgins, Ben Hayes and Jim Mungall led to a major improvement of the original version of the paper. FUNDING Field work and electron microprobe analyses were funded by the German Science Foundation (DFG), grant VE 619/2–1. Whole-rock analyses, including SR XRF, and studies of rock textures were funded by the Russian Science Foundation (RSF) grant No. 14–17-00200 and the AMREP project funded by German Ministry of Education and Research (BMBF). REFERENCES Atkinson H. V. ( 1988 ). Overview no. 65: theories of normal grain growth in pure single phase systems . Acta Metallurgica 36 , 469 – 491 . Google Scholar Crossref Search ADS Barnes S. J. ( 1986a ). The effect of trapped liquid crystallization on cumulus mineral compositions in layered intrusions . Contributions to Mineralogy and Petrology 93 , 524 – 531 . Google Scholar Crossref Search ADS Barnes S. J. ( 1986b ). The distribution of chromium among orthopyroxene, spinel and silicate liquid at atmospheric pressure . Geochimica et Cosmochimica Acta 50 , 1889 – 1909 . Google Scholar Crossref Search ADS Barnes S. J. , Roeder P. L. ( 2001 ). The range of spinel compositions in terrestrial mafic and ultramafic rocks . Journal of Petrology 42 , 2279 – 2302 . Google Scholar Crossref Search ADS Barnes S.-J. , Maier W. D. , Curl E. A. ( 2010 ). Composition of the marginal rocks and sills of the Rustenburg Layered Suite, Bushveld Complex, South Africa: implications for the formation of the platimum-group element deposits . Economic Geology 105 , 1491 – 1511 . Google Scholar Crossref Search ADS Bédard J. H. ( 2006 ). Trace element partitioning in plagioclase feldspar . Geochimica et Cosmochimica Acta 70 , 3717 – 3742 . Google Scholar Crossref Search ADS Bédard J. H. ( 2007 ). Trace element partitioning coefficients between silicate melts and orthopyroxene: parameterization of D variations . Chemical Geology 244 , 263 – 303 . Google Scholar Crossref Search ADS Brey G. P. , Köhler T. ( 1990 ). Geothermobarometry in four-phase lherzolites II. New thermobarometers, and practical assessment of existing thermobarometers . Journal of Petrology 31 , 1353 – 1378 . Google Scholar Crossref Search ADS Cameron E. N. ( 1969 ). Postcumulus changes in the Eastern Bushveld Complex . American Mineralogist 54 , 754 – 779 . Cameron E. N. ( 1975 ). Postcumulus and subsolidus equilibration of chromite and coexisting silicates in the Eastern Bushveld Complex . Geochimica et Cosmochimica Acta 39 , 1021 – 1033 . Google Scholar Crossref Search ADS Cameron E. N. ( 1977 ). Chromite in the central sector of Eastern Bushveld Complex, South Africa . American Mineralogist 62 , 1082 – 1087 . Cawthorn R. G. ( 1996 ). Re-evaluation of magma compositions and processes in the uppermost Critical Zone of the Bushveld Complex . Mineralogical Magazine 60 , 131 – 148 . Google Scholar Crossref Search ADS Cawthorn R. G. ( 2005 ). Stratiform platinum-group element deposits in layered intrusions. In: Mungall J. E. (ed.) Exploration for Platinum-Group Element Deposits. Mineralogical Association of Canada, Ottawa, Short Course Series Volume 35 , pp. 57 – 74 . Cawthorn R. G. ( 2007 ). Cr and Sr: keys to parental magmas and processes in the Bushveld Complex, South Africa . Lithos 95 , 381 – 398 . Google Scholar Crossref Search ADS Cawthorn R. G. ( 2011 ). Geological interpretations from the PGE distribution in the Bushveld Merensky and UG2 chromitite reefs . The Journal of the Southern African Institute of Mining and Metallurgy 111 , 67 – 79 . Cawthorn R. G. ( 2015 ). The Bushveld Complex, South Africa. In: Charlier B. , Namur O , Latypov R. , Tegner C. (eds) Layered Intrusions . Springer Geology , Dordrecht: Springer , pp. 517 – 588 . Cawthorn R. G. , Barry S. D. ( 1992 ). The role of intercumulus residua in the formation of pegmatoid associated with the UG2 chromitite, Bushveld Complex . Australian Journal of Earth Sciences 39 , 263 – 276 . Google Scholar Crossref Search ADS Cawthorn R. G. , Davies G. ( 1983 ). Experimental data at 3 kbars pressure on parental magma to the Bushveld Complex . Contributions to Mineralogy and Petrology 83 , 128 – 135 . Google Scholar Crossref Search ADS Chutas N. I. , Bates E. , Prevec S. A. , Coleman D. S. , Boudreau A. E. ( 2012 ). Sr and Pb isotopic deisequilibrium between coexisting plagioclase and orthopyroxene in the Bushveld Complex, South Africa: microdrilling and progressive leaching evidence for sub-liquidus contamination within a crystal mush . Contributions to Mineralogy and Petrology 163 , 653 – 668 . Google Scholar Crossref Search ADS Darin A. V. , Veksler I. V. , Rakshun Y. V. ( 2016 ). First results of the application of scanning XRF analysis with synchrotron-radiation beams from the VEPP-3 to study the spatial distribution of trace elements in samples of stratiform chromite ores . Journal of Surface Investigation. X-ray, Synchrotron and Neutron Techniques 10 , 88 – 91 . Google Scholar Crossref Search ADS Devi L. G. , Murthy B. N. ( 2008 ). Characterization of Mo doped TiO2 and its enhanced photo catalytic activity under visible light . Catalysis Letters 125 , 320 – 330 . Google Scholar Crossref Search ADS Eales H. V. , Cawthorn R. G. ( 1996 ). The Bushveld Complex. In: Cawthorn R. G. (ed.) Layered Intrusions . Amsterdam, New York : Elsevier , pp. 181 – 229 . Godel B. ( 2015 ). Platinum-group element deposits in layered intrusions: recent advances in the understanding of the ore forming processes. In: Charlier B. , Namur O. , Latypov R. , Tegner C. (eds) Layered Intrusions . Springer , Dordrecht , pp. 379 – 432 . Godel B. , Barnes S.-J. , Maier W. D. ( 2011 ). Parental magma composition inferred from trace element in cumulus and intercumulus silicate minerals: an example from the lower and lower critical zones of the Bushveld Complex, South Africa . Lithos 125 , 537 – 552 . Google Scholar Crossref Search ADS Harker D. , Parker E. R. ( 1945 ). Grain shape and grain growth . Transactions of the American Society of Metals 34 , 156 – 195 . Harris C. , Pronost J. J. M. , Ashwal L. D. , Cawthorn R. G. ( 2004 ). Oxygen and hydrogen isotope stratigraphy of the Rustenburg Layered Suite, Bushveld Complex: Constraints on crustal contamination . Journal of Petrology 46 , 579 – 601 . Google Scholar Crossref Search ADS Hiemstra S. A. ( 1985 ). The distribution of some platinum-group elements in the UG-2 chromitite layer of the Bushveld Complex . Economic Geology 80 , 944 – 957 . Google Scholar Crossref Search ADS Hiemstra S. A. ( 1986 ). The distribution of chalcophile and platinum-group elements in the UG-2 chromitite layer of the Bushveld Complex . Economic Geology 81 , 1080 – 1086 . Google Scholar Crossref Search ADS Higgins M. D. ( 2000 ). Measurement of crystal size distributions . American Mineralogist 85 , 1105 – 1116 . Google Scholar Crossref Search ADS Higgins M. D. ( 2015 ). Quantitative textural analysis of rocks in layered mafic intrusions. In: Charlier B. , Namur O , Latypov R. , Tegner C. (eds) Layered Intrusions . Springer Geology, Dordrecht : Springer , pp. 153 – 181 . Holness M. B. , Namur O. , Cawthorn R. G. ( 2013 ). Disequilibrium dihedral angles in layered intrusions: a microstructural record of fractionation . Journal of Petrology 54 , 2067 – 2093 . Google Scholar Crossref Search ADS Holness M. B. , Tegner C. , Nielsen T. F. D. , Stripp G. , Morse S. A. ( 2007 ). A textural record of solidification and cooling in the Skaergaard intrusion, East Greenland . Journal of Petrology 48 , 2359 – 2377 . Google Scholar Crossref Search ADS Hulbert L. J. , von Gruenewaldt G. ( 1985 ). Textural and compositional features of chromite in the lower and critical zones of the Bushveld Complex south of Potgietersrus . Economic Geology 80 , 872 – 895 . Google Scholar Crossref Search ADS Jackson E. D. ( 1961 ). Primary textures and mineral associations in the ultramafic zone of the Stillwater complex, Montana . United States Geological Survey Professional Paper 358 . Junge M. , Oberthur T. , Melcher F. ( 2014 ). Cryptic variation of chromite chemistry, platinum group element and platinum group mineral distribution in the UG-2 chromitite: an example from the Karee Mine, western Bushveld Complex, South Africa . Economic Geology 109 , 795 – 810 . Google Scholar Crossref Search ADS Kinnaird J. A. , Kruger F. J. , Nex P. A. M. , Cawthorn R. G. ( 2002 ). Chromitite formation—a key to understanding processes of platinum enrichment . Applied Earth Science: Transactions of the Institutions of Mining and Metallurgy: Section B 111 , 23 – 35 . Latypov R. , Chistyakova S. , Mukherjee R. ( 2017 ). A Novel Hypothesis for Origin of Massive Chromitites in the Bushveld Igneous Complex . Journal of Petrology 58 , 1899 – 1940 . Google Scholar Crossref Search ADS Lee C. A. ( 1996 ). A review of mineralization in the Bushveld Complex and some other layered mafic intrusions. In: Cawthorn R. G. (ed.) Layered Intrusions . Amsterdam : Elsevier , pp. 103 – 146 . Li C. , Ripley E. M. , Sarkar A. , Shin D. , Maier W. D. ( 2005 ). Origin of phlogopite-orthopyroxene inclusions in chromites from the Merensky Reef of the Bushveld Complex, South Africa . Contributions to Mineralogy and Petrology 150 , 119 – 130 . Google Scholar Crossref Search ADS Liang Y. , Sun C. , Yao L. ( 2013 ). A REE-in-two-pyroxene thermometer for mafic and ultramafic rocks . Geochimica et Cosmochimica Acta 102 , 246 – 260 . Google Scholar Crossref Search ADS Longerich H. P. , Jackson S. E. , Günther D. ( 1996 ). Laser ablation inductively coupled plasma-mass spectrometric transient signal data acquisition and analyte concentration calculation . Journal of Analytical Atomic Spectrometry 11 , 899 – 904 . Google Scholar Crossref Search ADS Maier W. D. , Barnes S.-J. ( 2008 ). Platinum-group elements in the UG1 and UG2 chromitites, and the Bastard reef, at Impala platimum mine, western Bushveld Complex, South Africa: evidence for late magmatic cumulate instability and reef constitution . South African Journal of Geology 111 , 159 – 176 . Google Scholar Crossref Search ADS Maier W. D. , Barnes S.-J. , Groves D. I. ( 2013 ). The Bushveld Complex, South Africa: formation of platinum-palladium, chrome- and vanadium-rich layers via hydrodynamic sorting of a mobilized cumulate slurry in a large, relatively slowly cooling, subsiding magma chamber . Mineralium Deposita 48 , 1 – 56 . Google Scholar Crossref Search ADS Maier W.D. , Prichard H.M. , Barnes S.J. , Fisher P.C. ( 1999 ). Compositional variation of laurite at Union Section in the Western Bushveld Complex . South African Journal of Geology 102 , 286 – 292 . Manoochehri S. , Schmidt M. W. , Britz W. ( 2015 ). Trapped liquid, paleo-porosity and formation time scale of a chromitite–(ortho) pyroxenite cumulate section, Bushveld, South Africa . Journal of Petrology 56 , 2195 – 2222 . Google Scholar Crossref Search ADS Marsh B. D. ( 1988 ). Crystal size distribution (CSD) in rocks and the kinetics and dynamics of crystallisation . Contributions to Mineralogy and Petrology 99 , 277 – 291 . Google Scholar Crossref Search ADS Mathez E. A. , Mey J. L. ( 2005 ). Character of the UG2 chromitite and host rocks and petrogenesis of its pegmatoidal footwall, northeastern Bushveld Complex . Economic Geology 100 , 1617 – 1630 . Google Scholar Crossref Search ADS Mondal S. K. , Mathez E. A. ( 2007 ). Origin of the UG2 chromitite layer, Bushveld Complex . Journal of Petrology 48 , 495 – 510 . Google Scholar Crossref Search ADS Mungall J. E. , Kamo S. L. , McQuade S. ( 2016 ). U–Pb geochronology documents out-of-sequence emplacement of ultramafic layers in the Bushveld Igneous Complex of South Africa . Nature Communications 7 , 13385 . Google Scholar Crossref Search ADS PubMed Naldrett A. J. , Wilson A. , Kinnaird J. , Yudovskaya M. , Chunnett G. ( 2012 ). The origin of chromitites and related PGE mineralization in the Bushveld Complex: new mineralogical and petrological constraints . Mineralium Deposita 47 , 209 – 232 . Google Scholar Crossref Search ADS Naldrett A. J. , Kinnaird J. , Wilson A. , Yudovskaya M. , McQuade S. , Chunnett G. , Stanley C. ( 2009 ). Chromite composition and PGE content of Bushveld chromitites: part 1—the lower and middle groups . Transactions of the Institute of Mineralogy and Metallurgy 118 , 131 – 161 . O’Driscoll B. , Emeleus C. H. , Donaldson C. H. , Daly J. S. ( 2010 ). Cr-spinel seam petrogenesis in the Rum layered suite, NW Scotland: assimilation and in situ crystallisation in a deforming crystal mush . Journal of Petrology 51 , 1171 – 1201 . Google Scholar Crossref Search ADS Penberthy C. J. , Merkle R. K. W. ( 1999 ). Lateral variations in the platinum-group element content and mineralogy of the UG2 chromitite . South African Journal of Geology 102 , 240 – 250 . Pouchou J. L. , Pichoir F. ( 1985 ). “PAP” (ϕ-ρ-Z) procedure for improved microanalysis. In: Armstrong J. T. (ed.) Microbeam Analysis . San Francisco : San Francisco Press , pp. 104 – 106 Scoates J. S. , Wall C. J. ( 2015 ). Geochronology of layered intusions. In: Charlier B. , Namur O , Latypov R. , Tegner C. (eds) Layered Intrusions . Springer Geology, Dordrecht : Springer , pp. 3 – 74 . Scoon R. N. , Teigler B. ( 1994 ). Platinum-Group element mineralization in the Critical Zone of the Western Bushveld Complex. I. Sulfide poor-chromitites below the UG-2 . Economic Geology 89 , 1094 – 1121 . Google Scholar Crossref Search ADS Smith C. S. ( 1964 ). Some elementary principles of polycrystalline microstructure . Metallurgical Reviews 9 , 1 – 48 . Google Scholar Crossref Search ADS South African Committee for Stratigraphy (SACS) ( 1980 ). Lithostratigraphy of South Africa, South West Africa/Namibia, and the Republics of Boputhatswana, Transkei, and Venda. In: Stratigraphy of Southern Africa, Handbook 8. Part 1. Geological Survey of South Africa, Government Printer, pp. 690. Spandler C. , Mavrogenes J. , Arculus R. ( 2005 ). Origin of chromitites in layered intrusions: Evidence from chromite- hosted melt inclusions from the Stillwater Complex . Geology 33 , 893 – 896 . Google Scholar Crossref Search ADS Sun C. , Liang Y. ( 2012 ). Distribution of REE between clinopyroxene and basaltic melt along a mantle adiabat: Effects of major element composition, water, and temperature . Contributions to Mineralogy and Petrology 163 , 807 – 823 . Google Scholar Crossref Search ADS Van Achterbergh E. , Ryan C. G. , Griffin W. L. ( 1999 ). GLITTER: On-line interactive data reduction for the laser ablation ICP-MS microprobe. Proceedings of the 9th V.M. Goldschmidt Conference. Cambridge, MA: Lunar and Planetary Institute, Houston. pp. 305. Van Achterbergh E. , Ryan C.G. , Jackson S.E. , Griffin W.L. ( 2001 ). Data reduction software for LA-ICP-MS. In: Sylvester, P. (ed.) Laser-Ablation-ICPMS in the Earth Sciences—Principles and Applications, Mineralogical Association of Canada (short course series), Ottawa, Vol. 29, pp. 239–243. Van Espen P. , Janssens K. , Nobels J. ( 1986 ). AXIL-PC: software for the analysis of complex X-ray spectra . Chemometrics and Intelligent Laboratory Systems 1 , 109 – 114 . Google Scholar Crossref Search ADS VanTongeren J. A. , Mathez E. A. ( 2013 ). Incoming magma composition and style of recharge below the Pyroxenite Marker, Eastern Bushveld Complex, South Africa . Journal of Petrology 54 , 1585 – 1605 . Google Scholar Crossref Search ADS Veksler I. V. , Reid D. L. , Dulski P. , Keiding J. K. , Schannor M. , Lutz H. , Trumbull R. B. ( 2015 ). Electrochemical processes in a crystal mush: cyclic units in the Upper Critical Zone of the Bushveld Complex, South Africa . Journal of Petrology 56 , 1229 – 1250 . Google Scholar Crossref Search ADS Voordouw R. J. , Beukes N. J. ( 2009 ). Alteration and metasomatism of the UG2 melanorite and its stratiform pegmatoids, Bushveld Complex, South Africa–characteristics, timing and origins . South African Journal of Geology 112 , 47 – 64 . Google Scholar Crossref Search ADS Voordouw R. , Gutzmer J. , Beukes N. J. ( 2009 ). Intrusive origin for Upper Group (UG1, UG2) stratiform chromitite seams in the Dwars River area, Bushveld Complex, South Africa . Mineralogy and Petrology 97 , 75 – 94 . Google Scholar Crossref Search ADS Voordouw R. , Gutzmer J. , Beukes N. J. ( 2010 ). Zoning of platinum group mineral assemblages in the UG2 chromitite determined through in situ SEM-EDS-based image analysis . Mineralium Deposita 45 , 147 – 159 . Google Scholar Crossref Search ADS Vukmanovic Z. , Barnes S. J. , Reddy S. M. , Godel B. , Fiorentini M. L. ( 2013 ). Morphology and microstructure of chromite crystals in chromitites from the Merensky Reef (Bushveld Complex, South Africa) . Contributions to Mineralogy and Petrology 165 , 1031 – 1050 . Google Scholar Crossref Search ADS Waters C. , Boudreau A. E. ( 1996 ). A re-evaluation of crystal size distribution in chromite cumulates . American Mineralogist 81 , 1452 – 1459 . Google Scholar Crossref Search ADS Wilson A. H. ( 2012 ). A chill sequence to the Bushveld Complex: insight into the first stage of magma emplacement and implications for parental magmas . Journal of Petrology 53 , 1123 – 1168 . Google Scholar Crossref Search ADS Yao L. , Sun C. , Liang Y. ( 2012 ). A parameterized model for REE partitioning between low-Ca pyroxene and basaltic melts with applications to adiabatic mantle melting and pyroxenite-derived melt and peridotite interaction . Contributions to Mineralogy and Petrology 164 , 261 – 280 . Google Scholar Crossref Search ADS Zeh A. , Ovtcharova M. , Wilson A. H. , Schaltegger U. ( 2015 ). The Bushveld Complex was emplaced and cooled in less than one million years–results of zirconology, and geotectonic implications . Earth Planetary Science Letters 418 , 103 – 114 . Google Scholar Crossref Search ADS © The Author(s) 2018. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Petrology Oxford University Press

Chemical and Textural Re-equilibration in the UG2 Chromitite Layer of the Bushveld Complex, South Africa

Loading next page...
 
/lp/ou_press/chemical-and-textural-re-equilibration-in-the-ug2-chromitite-layer-of-OZK5KrEZOu
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
ISSN
0022-3530
eISSN
1460-2415
D.O.I.
10.1093/petrology/egy058
Publisher site
See Article on Publisher Site

Abstract

Abstract Variations of mineral chemistry and whole-rock compositions were studied in detail, at millimetre to centimetre intervals, in two vertical drill core profiles through the platiniferous UG2 chromitite layer in the western and eastern limbs of the Bushveld Complex, South Africa. Analytical methods included electron microprobe and LA-ICP-MS analyses of the main rock-forming minerals, orthopyroxene, plagioclase and interstitial clinopyroxene. One profile was also studied by synchrotron-source XRF. Statistical analysis of crystal size distribution of chromite was also performed at different levels in the chromitite layer and in adjacent silicate rocks. The results provide new evidence for chemical and textural late magmatic re-equilibration in the UG2 layer and in the silicate rocks at the contact zones. The chromite crystal size distributions imply extensive coarsening of that mineral within the main chromitite seam, which has erased any textural evidence of primary deposition features such as recharge or mechanical sorting of crystals, if those features originally existed. The mineral compositions in chromitite differ from those in adjacent silicate rocks, in general agreement with predictions of chemical re-equilibration with evolved, residual melt (the trapped liquid shift effect). In detail, the geochemical data imply, however, that the conventional trapped liquid shift model has shortcomings, due to the effects of material transport driven by chemical gradients between modally contrasting layers of crystal mush undergoing re-equilibration reactions. In the presence of such gradients, selective open-system conditions may hold for alkalis and hydrogen because of their higher diffusion rates in silicate melts. Differential mobility of components in the interstitial melt can also sharpen the original modal layering by causing minerals to crystallise in one layer and dissolve in another. Detailed trace element profiles by synchrotron XRF reveal an uneven vertical distribution of incompatible elements which implies that the permeability of the chromitite layer may have been significant, even at the latest stages of interstitial crystallization. INTRODUCTION Layers of chromitite chiefly composed of Cr-rich spinel are among the most spectacular examples of mineral layering in mafic-ultramafic intrusions. The most prominent examples come from Precambrian layered intrusions such as Stillwater in Montana, USA, the Great Dyke in Zimbabwe and the Bushveld Complex in South Africa. In terms of economic geology, chromitite layers constitute major global reserves of chromium and some contain economic concentrations of platinum-group elements (PGE). As common and important as chromitite layers are, their origin remains highly contentious. Nowhere is the economic importance of chromitite layers greater than in the case of the UG2 layer in the Bushveld Complex, which hosts the world’s largest reserves of PGE, has been extensively mined for decades and is the subject of this report. The UG2 is the most studied chromitite layer in the world, yet there is still no consensus on the relative roles of early-, late-, and post-magmatic processes in producing this remarkable rock layer. Opinions are divided on the fundamental question of whether chromite crystallized within the Bushveld magma chamber or was transported as a slurry from a staging magma chamber below. For detailed accounts of the main features of the UG2 layer and its relationship with other units of the layered sequence, readers are referred to an extensive recent literature, e.g. Cawthorn (2005, 2011, 2015); Kinnaird et al. (2002); Mathez & Mey (2005); Mondal & Mathez (2007); Maier & Barnes (2008); Voordouw et al. (2010); Maier et al., 2013; Naldrett et al. (2012); Junge et al. (2014); Godel (2015), Latypov et al. (2017) and Manoochehri et al. (2015). A major and unsolved obstacle to understanding how chromitites form in the Bushveld and other layered mafic intrusions is that the primary crystallization and chromite accumulation are modified by re-equilibration of minerals with interstitial liquid in the ‘crystal mush’ during final solidification of the layered rocks and even subsequently at sub-solidus conditions. This is a long-recognized problem and considerable progress has been made in understanding the chemical and physical processes operating in such crystal mushes. Of particular relevance for this study is the so-called ‘trapped liquid shift effect’ that describes the compositional re-equilibration among minerals and the evolving interstitial melt (trapped liquid). The first geochemical models of this effect were developed for Mg–Fe exchange and for the distribution of some trace elements (Barnes, 1986a; Cawthorn, 1996). In an extension of the trapped liquid calculations, Manoochehri et al. (2015) considered results from compaction experiments and cooling models to assess the development of porosity in the UG2 and UG3 chromitite layers. Quantitative textural studies of chromite grains and the interstitial silicate minerals can also give insights into the processes operating in a crystal mush (see review by Higgins, 2015). The quantitative approach has been previously applied to statistical analysis of the crystal-size distribution of chromites and silicate minerals (e.g. O’Driscoll et al., 2010; Vukmanovic et al., 2013) and the geometry of grain boundaries (Holness et al. 2007,, 2013). Rarely have textural studies and detailed geochemical analyses been undertaken on the same vertical profiles across a chromitite layer. This is the main objective of the study presented here. Results are presented from two complete profiles through the UG2 layer, one from the eastern limb (Nkwe mine) and one from the western limb (Khuseleka mine). We provide petrographic descriptions and report the chemical compositions of the main rock-forming minerals in the UG2 layer (chromite, orthopyroxene, plagioclase, locally clinopyroxene) and in the silicate rocks immediately above and below the layer, using electron microprobe and laser ablation ICP-MS. The geochemical effect of the evolved, trapped liquid is to produce local enrichments in incompatible trace elements, and in some cases crystallization of accessory phases such as zircon or monazite. The spatial distribution of trace element enrichment may reflect the distribution of trapped liquid in the layer, or the paleo-porosity in the sense of Manoochehri et al. (2015). In an effort to quantify this distribution, we used synchrotron-source XRF spectrometry to obtain continuous, millimetre-resolution trace-element profiles across the UG2 drill core from Khuseleka (including Pd and Rh). To our knowledge this is the first application of the method to layered igneous rocks. Finally, we performed crystal size distribution (CSD) analyses of chromite from UG2 and the adjacent silicate layers, providing the first CSD data from the UG2 chromite. GEOLOGIC SETTING: THE BUSHVELD CHROMITITE LAYERS The Paleoproterozoic (2055–2057 Ma: Scoates & Wall, 2015; Zeh et al., 2015) Bushveld Complex in South Africa is the largest layered igneous intrusion on the Earth. Present-day outcrops of the Complex (Fig. 1) cover a surface area of 66 000 km2 and the total thickness of igneous rocks is about 9 km; however, its former extent and thickness are thought to be much greater (see Cawthorn, 2015 for a comprehensive review). The South African Committee for Stratigraphy (SACS, 1980) recognized three major rock suites in the complex: (1) the mafic–ultramafic Rustenburg Layered Suite (RLS) which is of interest in this study, and overlying felsic suites represented by (2) the Lebowa Granite and (3) the Rashoop Granophyre. Fig. 1 View largeDownload slide Schematic geological map of the Bushveld Complex with sampling locations. Fig. 1 View largeDownload slide Schematic geological map of the Bushveld Complex with sampling locations. The Rustenburg layered suite The c.8 km thick Rustenburg Layered Suite (RLS) is subdivided vertically into several lithologic zones, from bottom to top: the Marginal Zone dominated by norite, the Lower Zone composed of pyroxenites and harzburgites, the Critical Zone (chromitites, harzburgites, pyroxenites, norites and minor anorthosites), the Main Zone (norites, gabbronorites and minor anorthosites) and the Upper Zone (gabbronorites and augite diorites with magnetite and nelsonite layers). There is strong evidence that the RLS rocks formed in an open-system magma chamber affected by multiple injections and mixing of different parental magmas, as well as by significant amounts of crustal assimilation (Harris et al., 2004). The open-system nature of magma evolution in the RLS is shown by trace element ratio variations, changes in the initial Sr isotope ratios and by reversals in compositional trends of cumulate minerals (Eales & Cawthorn, 1996; Maier et al., 2013; Cawthorn, 2015 and references therein). Studies of chilled margins (Wilson, 2012) and comagmatic sills (Barnes et al., 2010) revealed three distinct types of parental magma, referred to in recent publications as B1, B2 and B3 (Barnes et al., 2010; Godel et al., 2011; Maier et al., 2013; VanTongeren & Mathez, 2013). For the Critical Zone rocks of interest here, the B1 magma is believed to represent the parental magma composition (Cawthorn, 2007). The estimates of the B1 magma composition published by different authors vary in detail, but in general it corresponds to an Mg-rich basaltic andesite (SiO2 and MgO contents at about 55 and 12 wt %, respectively). The other two magma types are more evolved, with major element characteristics corresponding to tholeiitic basalt (50–51 wt % SiO2; 6–7·5 wt % MgO). UG2 and other chromitite layers of the critical zone The Critical Zone is defined by the presence of prominent chromitite layers, which are sufficiently laterally continuous to be used as stratigraphic markers. Three groups of chromite layers are distinguished, each with several members: Lower (LG1–7), Middle (MG 1–4) and Upper (UG1–3). Individual layers within each group are numbered sequentially from the base upwards; locally some layers split into one or more sublayers, e.g.LG6a, MG2b, etc. The division of the lower and upper Critical Zones is marked by the appearance of cumulus plagioclase and this occurs within the MG group. The thickest of the chromitite layers reaches about 2 m, but 0·7 to 1 m is more common and there are many thinner chromite seams and stringers of only a few centimetres or even millimetres in thickness between the major chromitite layers. There are systematic compositional variations of chromite with stratigraphic height in the sequence of LG to UG groups. A summary from many localities in the complex is provided by Naldrett et al. (2012) who defined two general trends. Chromites from LG1 to LG4 show a trend of increasing Cr/(Cr+Al) ratio and decreasing Mg/(Mg + Fe2+) upward. This was termed trend A. Trend B, displayed by the LG5 to MG2 sequence, shows upward-decreasing Cr/(Cr+Al) and decreasing Mg/(Mg+Fe2+) ratios. Chromite layers from MG3 upwards again follow trend A. Naldrett et al. (2012) explained the trends by a combination of magma recharge and fractional crystallization. The UG2 is the uppermost massive chromitite layer in the Bushveld Complex except for part of the eastern limb where a UG3 is also present. Above the UG2 or UG3, chromite occurs as an accessory mineral dispersed in silicate rocks or as thin, millimetre- to centimetre-thick stringers. Important examples of the latter are chromitites at the basal and top contacts of the Merensky Reef, which lies several tens to hundreds of metres above UG2. The concentrations of platinum-group elements (PGE) increase upward in the Critical Zone chromitites, from substantial, but sub-economic, levels of around 0·5 to 1 ppm in the LG group to concentrations of 5–10 ppm in UG2 and the Merensky chromite stringers where they are actively mined (Scoon & Teigler, 1994; Naldrett et al., 2009). Indeed, UG2 and the Merensky Reef with its chromitite stringers contain the largest PGE resource in the world (Godel, 2015 and references therein). The mechanisms by which PGE are concentrated in chromitites remain unclear and the situation is unlikely to improve without a better understanding of how the chromite layers themselves formed (Kinnaird et al., 2002). The UG2 layer is typically about 0·8–1·0 m thick, but ranges between 0·15 and 2·5 m Its modal abundance of chromite is 60 to 90 wt % (Mathez & Mey, 2005; Naldrett et al., 2012). Laterally, the UG2 layer may split into two or more thinner seams, which merge again further along strike. As noted above, this is characteristic of other Bushveld chromitite layers, e.g.the MG series, LG6 and especially UG1, whose numerous thin seams interlayer with anorthosite to form the famous ‘zebra rock’ of Dwars River (Voordouw et al., 2009). Cawthorn (2011) proposed that UG3, known only from part of the eastern lobe, has elsewhere merged with the top of the UG2 layer. Lee (1996) and Mathez & Mey (2005) documented a textural break in the middle of the UG2 layer between a lower part with fine-grained chromitite and an upper part containing abundant plagioclase and pyroxene oikocrysts. These and other lines of evidence led Cawthorn (2011) to conclude that the UG2 chromitite may be a composite of three distinct layers, between which there are occasional thin screens or ‘partings’ of silicate-rich rock. Variations of chromite composition in vertical profiles across the UG2 layer have been documented by, e. g., Mathez & Mey (2005) and Junge et al. (2014). Whereas Mathez & Mey (2005) found locally strong but non-systematic differences in Cr/Al, Mg/Fe, Ti and Ni contents, Junge et al. (2014) reported subtle, but recurring, systematic trends of an upward increase in Cr/(Cr+Al) values and TiO2 content, coupled with decreasing Mg/(Mg+Fe2+) values (Trend ‘A’ of Naldrett et al., 2012). We will discuss these results in more detail and compare with our profiles across the UG2. SAMPLES AND ANALYTICAL METHODS This study is based on two complete vertical drill core sections through the UG2 layer, one collected at the Khuseleka mine (formerly Townlands shaft) in the western limb near Rustenburg and the other at the Nkwe Platinum mine in the eastern limb of the Bushveld complex (Fig. 1). Schematic stratigraphic columns of the sections are presented in Fig. 2. The drill cores were cut lengthwise and segmented into continuous sets of petrographic thin sections for textural studies and for electron microprobe and LA-ICP-MS analyses of mineral compositions. The remaining half core (Khuseleka only) was used for synchrotron XRF scanning. Fig. 2 View largeDownload slide Simplified lithologic profiles of the studied drill core sections. (a) Khuseleka; (b) Nkwe. Fig. 2 View largeDownload slide Simplified lithologic profiles of the studied drill core sections. (a) Khuseleka; (b) Nkwe. Electron microprobe analyses Mineral compositions were determined in situ on polished, carbon-coated thin sections at GFZ Potsdam using a JEOL JXA-8230 electron microprobe equipped with four wavelength-dispersive spectrometers (WDS) and at the Museum of Natural History (MfN) in Berlin using a JEOL JXA-8500F electron microprobe equipped with a field emission cathode and five WDS. At GFZ Potsdam, the electron microprobe was operated at 15 kV accelerating voltage and a beam current of 15 nA, with a fully-focused (c.1 micron) beam for clinopyroxene and olivine. Plagioclase was analysed using a defocused 5 micron beam and Na was measured first in the element sequence to minimize the effects of alkali loss. Replicate analyses of natural and synthetic standards indicate that the precision of major oxides (>0·2 wt %) is ± 9 % (2σ) or better. Cr-spinel measurements were made with a 20 KV accelerating voltage, beam current of 75 nA and a 1 μm focussed electron beam spot size. The element concentrations were obtained through calibration with natural and synthetic mineral standards (orthoclase for Al and K, rutile for Ti, wollastonite for Si and Ca, albite and jadeite for Na, apatite for P, hematite for Fe, and diopside and periclase for Mg). Analytical uncertainties for the major components are considered to be better than 1–2%. Automated data reduction and correction in both laboratories followed the PAP method (Pouchou & Pichoir, 1985). In calculations of the Mg# [atomic ratio Mg/(Mg+Fe2+)] for pyroxene and olivine, all Fe was assumed to be Fe2+. For spinel, total Fe was divided into Fe2+ and Fe3+ as described, for example, by Barnes & Roeder (2001) based on the assumption of stoichiometry and charge balance with the ideal formula XY2O4 where X = (Mg, Fe2+, Mn) and Y = (Cr, Fe3+, Al). Titanium was assumed to form the ulvöspinel component. Mineral compositions from the Nkwe profile were previously published by Veksler et al. (2015). Laser ablation ICP-MS analyses Laser ablation analyses were performed in two laboratories, at the Vernadsky Institute of Geochemistry and Analytical Chemistry (GEOKHI) in Moscow, Russia and the University of Erlangen in Germany. Analyses at the Vernadsky Institute used the UP-213 laser system and mass spectrometer Element-XR. The gas flow through ablation chamber was optimized for maximum the signal intensity and minimum background noise. For the He carrier gas a flow rate of 0·6–0·8 L/min was used; Ar sampling gas was set at 0·8–1·2 L/min; auxiliary gas at 1·5–2·0 L/min and the flow of the cooling gas was at 16 L/min. Measurements were performed at low resolution (R = 300). The ablation system parameters were as follows: the laser pulse frequency at 4 Hz, power 35%, diameter of the ablation crater 30 µm. Crystal sizes in most samples allowed line scan ablations and, whenever possible, we preferred to move the laser beam at 1 µm/s with a resulting total track length of 180 µm The analyses were carried out in sequences set by the Element XR software package. Standards were measured at the beginning and the end of each sequence. Between analyses we used a wash time pause of 240 s, in order to remove the evaporation products of the previous sample. Synthetic glass NIST-610 was used as the main standard with further control of the quality of measurements against the reference glass ML3B. The concentration values were taken from the GeoRem Database (http://georem.mpch-mainz.gwdg.de). Concentrations of Ca and Al measured by microprobe were used as internal standards for pyroxene and plagioclase, respectively. The obtained data were processed using the program ‘Glitter’ (Van Achterbergh et al., 1999). The laser ablation device at the University of Erlangen included a New Wave Research UP193FX laser and an Agilent 7500i, Plasma power 1320 W ICP-MS. The flow rate for carrier gas I (He) was 0·65 L/min, for carrier gas II (Ar) 1·07 L/min, for plasma gas (Ar) 14·9 L/min. The auxiliary gas (Ar) with flow rate at 0·9 L/min was tuned for maximum sensitivity and ThO+/Th+ <0·5%. We used a single spot ablation mode with a laser pulse rate of 15 Hz and crater diameter of 100 µm. The irradiance laser energy was at 0·67 GW/cm2 with the energy density at 3·37 J/cm2. The ablation time was 30 s for the background and 60 s for the analysis. Integration time for trace element isotopes was 20 ms for each mass, 5 ms each for 26 Mg, 27Al, 57Fe, and 10 ms each for 29Si and 44Ca. Silica was used as the internal standard. NIST SRM 612 glass standard was used for external calibration. Accuracy and reproducibility were checked by ablation of the NIST SRM 614 standard. Data reduction employed the programme LAMTRACE described by Longerich et al. (1996) and van Achterbergh et al. (2001). The programme performs background correction, correction for instrumental drift, internal calibration, choice of integration intervals and calculation of element concentrations using external calibration. The Ca signal and microprobe data were used for internal calibration of trace element concentrations in plagioclase and Mg was used in the same way for pyroxene, olivine and spinel. A summary of the electron microprobe and LA-ICP-MS data is given in Tables 1 and 2. Table 1 Orthopyroxene compositions, average concentrations (Av.) and standard deviations (S. D.) of N microprobe and n LA ICP-MS analyses. Major oxides are in wt %, trace elements are in ppm Profile Khuseleka Nkwe Host rock Footwall harzburgite Pyroxenite parting Hanging wall pyroxenite Footwall pyroxenite UG2 chromitite Hanging wall anorthosite Hanging wall pyroxenite N(n) 8 (7) 36 (3) 25 (3) 28 (11) 44 (10) 3 (2) 40 (10) Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. SiO2 54·84 0·27 54·73 0·92 54·81 0·59 54·88 0·42 56·52 0·24 52·77 0·17 54·96 0·31 TiO2 0·13 0·02 0·16 0·04 0·15 0·02 0·16 0·05 0·08 0·02 0·15 0·01 0·15 0·03 Al2O3 1·74 0·06 1·32 0·16 1·38 0·10 1·24 0·24 1·22 0·11 1·22 0·08 1·17 0·07 Cr2O3 0·55 0·06 0·54 0·07 0·51 0·04 0·35 0·10 0·55 0·15 0·09 0·02 0·33 0·05 MgO 30·37 0·63 30·03 1·70 30·25 0·92 28·67 0·62 33·55 0·37 22·73 0·05 28·96 0·78 MnO 0·26 0·02 0·29 0·03 0·27 0·03 0·29 0·03 0·18 0·02 0·42 0·03 0·27 0·03 FeO 10·29 0·46 11·42 1·80 10·93 0·94 14·41 0·72 8·22 0·26 22·52 0·36 13·61 0·84 CaO 2·07 0·83 1·57 1·38 1·66 1·19 0·94 0·47 0·54 0·15 0·83 0·10 1·10 0·51 Na2O 0·03 0·02 0·03 0·03 0·03 0·03 0·02 0·02 0·01 0·01 0·02 0·03 0·02 0·02 Total 100·26 100·09 100·00 101·08 101·01 100·96 100·72 Mg# 0·84 0·82 0·83 0·78 0·88 0·64 0·79 Trace elements, ppm P 35·2 11·3 34·4 10·0 37·5 8·9 n. a. n. a. n. a. n. a. K n. a. n. a. n. a. 6·7 0·4 3·1 0·3 b. d. l. 1·6 0·2 Sc 28·5 1·9 31·4 1·2 36·9 2·3 29·4 1·4 49·7 2·2 30·1 1·3 29·5 1·2 Ti 682 118 885 139 864 181 1191 62 542 26 1148 56 1062 46 V n. a. n. a. n. a. 155 7 72 3 156 7 177 7 Cr 3448 620 3500 23 3130 261 3055 178 2200 113 2504 120 2919 138 Mn 1905 57 2293 112 2025 106 2408 92 1568 57 2492 91 2309 80 Co 96 5 112 10 81 5 132 5 46 2 129 5 131 4 Ni 746 47 871 15 736 8 834 30 857 29 1754 60 954 31 Cu 0·87 0·23 0·89 0·13 0·82 0·11 n. a. n. a. n. a. n. a. Rb 0·14 0·02 0·26 0·05 0·20 0·03 0·049 0·004 0·019 0·006 0·013 0·001 Sr 0·41 0·27 0·28 0·03 0·26 0·03 0·48 0·02 0·066 0·007 0·132 0·009 0·236 0·010 Y 3·2 1·0 3·2 0·1 4·4 0·8 3·1 0·2 2·4 0·1 2·0 0·1 2·2 0·1 Zr 8·2 5·6 5·5 0·7 4·3 0·6 n. a. n. a. n. a. n. a. Nb 0·092 0·048 0·055 0·005 0·054 0·013 n. a. n. a. n. a. n. a. Ba 0·265 0·089 0·333 0·021 0·263 0·050 0·136 0·011 0·063 0·025 b. d. l. 0·058 0·006 La 0·130 0·067 0·080 0·025 0·050 0·019 0·158 0·007 0·021 0·004 0·007 0·002 0·037 0·003 Ce 0·444 0·290 0·427 0·087 0·247 0·173 0·565 0·022 0·048 0·005 0·025 0·003 0·143 0·007 Pr 0·073 0·050 0·072 0·031 0·056 0·034 0·080 0·004 0·008 0·002 0·007 0·002 0·025 0·002 Nd 0·385 0·252 0·405 0·138 0·377 0·205 0·394 0·020 0·054 0·014 0·050 0·013 0·140 0·014 Sm 0·174 0·112 0·317 0·086 0·266 0·134 0·137 0·010 0·039 0·014 0·028 0·012 0·072 0·011 Eu 0·051 0·035 0·055 0·024 0·056 0·013 0·026 0·002 0·008 0·003 0·010 0·003 0·012 0·002 Gd 0·295 0·114 0·410 0·096 0·460 0·135 0·227 0·016 0·083 0·019 0·084 0·016 0·135 0·015 Tb 0·060 0·028 0·067 0·003 0·094 0·026 0·050 0·003 0·025 0·004 0·022 0·003 0·033 0·003 Dy 0·505 0·216 0·319 0·086 0·718 0·265 0·446 0·030 0·298 0·032 0·217 0·022 0·301 0·023 Ho 0·123 0·046 0·109 0·008 0·147 0·036 0·112 0·007 0·086 0·009 0·066 0·006 0·078 0·006 Er 0·402 0·124 0·386 0·107 0·524 0·039 0·403 0·026 0·357 0·033 0·272 0·023 0·276 0·019 Tm 0·063 0·018 0·065 0·027 0·084 0·031 0·073 0·005 0·065 0·007 0·052 0·005 0·050 0·004 Yb 0·532 0·183 0·478 0·101 0·672 0·063 0·576 0·041 0·558 0·053 0·485 0·040 0·391 0·030 Lu 0·088 0·025 0·075 0·015 0·092 0·032 0·096 0·007 0·100 0·010 0·089 0·008 0·067 0·005 Hf 0·195 0·140 0·164 0·071 0·258 0·151 n. a. n. a. n. a. n. a. Ta 0·029 0·005 0·028 0·007 0·031 0·012 n. a. n. a. n. a. n. a. Pb 0·238 0·132 0·153 0·078 0·171 0·105 0·101 0·007 0·053 0·010 0·042 0·008 0·113 0·009 Th 0·185 0·178 0·078 0·011 0·132 0·108 n. a. n. a. n. a. n. a. U 0·042 0·034 0·026 0·005 0·016 0·003 n. a. n. a. n. a. n. a. Profile Khuseleka Nkwe Host rock Footwall harzburgite Pyroxenite parting Hanging wall pyroxenite Footwall pyroxenite UG2 chromitite Hanging wall anorthosite Hanging wall pyroxenite N(n) 8 (7) 36 (3) 25 (3) 28 (11) 44 (10) 3 (2) 40 (10) Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. SiO2 54·84 0·27 54·73 0·92 54·81 0·59 54·88 0·42 56·52 0·24 52·77 0·17 54·96 0·31 TiO2 0·13 0·02 0·16 0·04 0·15 0·02 0·16 0·05 0·08 0·02 0·15 0·01 0·15 0·03 Al2O3 1·74 0·06 1·32 0·16 1·38 0·10 1·24 0·24 1·22 0·11 1·22 0·08 1·17 0·07 Cr2O3 0·55 0·06 0·54 0·07 0·51 0·04 0·35 0·10 0·55 0·15 0·09 0·02 0·33 0·05 MgO 30·37 0·63 30·03 1·70 30·25 0·92 28·67 0·62 33·55 0·37 22·73 0·05 28·96 0·78 MnO 0·26 0·02 0·29 0·03 0·27 0·03 0·29 0·03 0·18 0·02 0·42 0·03 0·27 0·03 FeO 10·29 0·46 11·42 1·80 10·93 0·94 14·41 0·72 8·22 0·26 22·52 0·36 13·61 0·84 CaO 2·07 0·83 1·57 1·38 1·66 1·19 0·94 0·47 0·54 0·15 0·83 0·10 1·10 0·51 Na2O 0·03 0·02 0·03 0·03 0·03 0·03 0·02 0·02 0·01 0·01 0·02 0·03 0·02 0·02 Total 100·26 100·09 100·00 101·08 101·01 100·96 100·72 Mg# 0·84 0·82 0·83 0·78 0·88 0·64 0·79 Trace elements, ppm P 35·2 11·3 34·4 10·0 37·5 8·9 n. a. n. a. n. a. n. a. K n. a. n. a. n. a. 6·7 0·4 3·1 0·3 b. d. l. 1·6 0·2 Sc 28·5 1·9 31·4 1·2 36·9 2·3 29·4 1·4 49·7 2·2 30·1 1·3 29·5 1·2 Ti 682 118 885 139 864 181 1191 62 542 26 1148 56 1062 46 V n. a. n. a. n. a. 155 7 72 3 156 7 177 7 Cr 3448 620 3500 23 3130 261 3055 178 2200 113 2504 120 2919 138 Mn 1905 57 2293 112 2025 106 2408 92 1568 57 2492 91 2309 80 Co 96 5 112 10 81 5 132 5 46 2 129 5 131 4 Ni 746 47 871 15 736 8 834 30 857 29 1754 60 954 31 Cu 0·87 0·23 0·89 0·13 0·82 0·11 n. a. n. a. n. a. n. a. Rb 0·14 0·02 0·26 0·05 0·20 0·03 0·049 0·004 0·019 0·006 0·013 0·001 Sr 0·41 0·27 0·28 0·03 0·26 0·03 0·48 0·02 0·066 0·007 0·132 0·009 0·236 0·010 Y 3·2 1·0 3·2 0·1 4·4 0·8 3·1 0·2 2·4 0·1 2·0 0·1 2·2 0·1 Zr 8·2 5·6 5·5 0·7 4·3 0·6 n. a. n. a. n. a. n. a. Nb 0·092 0·048 0·055 0·005 0·054 0·013 n. a. n. a. n. a. n. a. Ba 0·265 0·089 0·333 0·021 0·263 0·050 0·136 0·011 0·063 0·025 b. d. l. 0·058 0·006 La 0·130 0·067 0·080 0·025 0·050 0·019 0·158 0·007 0·021 0·004 0·007 0·002 0·037 0·003 Ce 0·444 0·290 0·427 0·087 0·247 0·173 0·565 0·022 0·048 0·005 0·025 0·003 0·143 0·007 Pr 0·073 0·050 0·072 0·031 0·056 0·034 0·080 0·004 0·008 0·002 0·007 0·002 0·025 0·002 Nd 0·385 0·252 0·405 0·138 0·377 0·205 0·394 0·020 0·054 0·014 0·050 0·013 0·140 0·014 Sm 0·174 0·112 0·317 0·086 0·266 0·134 0·137 0·010 0·039 0·014 0·028 0·012 0·072 0·011 Eu 0·051 0·035 0·055 0·024 0·056 0·013 0·026 0·002 0·008 0·003 0·010 0·003 0·012 0·002 Gd 0·295 0·114 0·410 0·096 0·460 0·135 0·227 0·016 0·083 0·019 0·084 0·016 0·135 0·015 Tb 0·060 0·028 0·067 0·003 0·094 0·026 0·050 0·003 0·025 0·004 0·022 0·003 0·033 0·003 Dy 0·505 0·216 0·319 0·086 0·718 0·265 0·446 0·030 0·298 0·032 0·217 0·022 0·301 0·023 Ho 0·123 0·046 0·109 0·008 0·147 0·036 0·112 0·007 0·086 0·009 0·066 0·006 0·078 0·006 Er 0·402 0·124 0·386 0·107 0·524 0·039 0·403 0·026 0·357 0·033 0·272 0·023 0·276 0·019 Tm 0·063 0·018 0·065 0·027 0·084 0·031 0·073 0·005 0·065 0·007 0·052 0·005 0·050 0·004 Yb 0·532 0·183 0·478 0·101 0·672 0·063 0·576 0·041 0·558 0·053 0·485 0·040 0·391 0·030 Lu 0·088 0·025 0·075 0·015 0·092 0·032 0·096 0·007 0·100 0·010 0·089 0·008 0·067 0·005 Hf 0·195 0·140 0·164 0·071 0·258 0·151 n. a. n. a. n. a. n. a. Ta 0·029 0·005 0·028 0·007 0·031 0·012 n. a. n. a. n. a. n. a. Pb 0·238 0·132 0·153 0·078 0·171 0·105 0·101 0·007 0·053 0·010 0·042 0·008 0·113 0·009 Th 0·185 0·178 0·078 0·011 0·132 0·108 n. a. n. a. n. a. n. a. U 0·042 0·034 0·026 0·005 0·016 0·003 n. a. n. a. n. a. n. a. n. a., not analysed Table 1 Orthopyroxene compositions, average concentrations (Av.) and standard deviations (S. D.) of N microprobe and n LA ICP-MS analyses. Major oxides are in wt %, trace elements are in ppm Profile Khuseleka Nkwe Host rock Footwall harzburgite Pyroxenite parting Hanging wall pyroxenite Footwall pyroxenite UG2 chromitite Hanging wall anorthosite Hanging wall pyroxenite N(n) 8 (7) 36 (3) 25 (3) 28 (11) 44 (10) 3 (2) 40 (10) Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. SiO2 54·84 0·27 54·73 0·92 54·81 0·59 54·88 0·42 56·52 0·24 52·77 0·17 54·96 0·31 TiO2 0·13 0·02 0·16 0·04 0·15 0·02 0·16 0·05 0·08 0·02 0·15 0·01 0·15 0·03 Al2O3 1·74 0·06 1·32 0·16 1·38 0·10 1·24 0·24 1·22 0·11 1·22 0·08 1·17 0·07 Cr2O3 0·55 0·06 0·54 0·07 0·51 0·04 0·35 0·10 0·55 0·15 0·09 0·02 0·33 0·05 MgO 30·37 0·63 30·03 1·70 30·25 0·92 28·67 0·62 33·55 0·37 22·73 0·05 28·96 0·78 MnO 0·26 0·02 0·29 0·03 0·27 0·03 0·29 0·03 0·18 0·02 0·42 0·03 0·27 0·03 FeO 10·29 0·46 11·42 1·80 10·93 0·94 14·41 0·72 8·22 0·26 22·52 0·36 13·61 0·84 CaO 2·07 0·83 1·57 1·38 1·66 1·19 0·94 0·47 0·54 0·15 0·83 0·10 1·10 0·51 Na2O 0·03 0·02 0·03 0·03 0·03 0·03 0·02 0·02 0·01 0·01 0·02 0·03 0·02 0·02 Total 100·26 100·09 100·00 101·08 101·01 100·96 100·72 Mg# 0·84 0·82 0·83 0·78 0·88 0·64 0·79 Trace elements, ppm P 35·2 11·3 34·4 10·0 37·5 8·9 n. a. n. a. n. a. n. a. K n. a. n. a. n. a. 6·7 0·4 3·1 0·3 b. d. l. 1·6 0·2 Sc 28·5 1·9 31·4 1·2 36·9 2·3 29·4 1·4 49·7 2·2 30·1 1·3 29·5 1·2 Ti 682 118 885 139 864 181 1191 62 542 26 1148 56 1062 46 V n. a. n. a. n. a. 155 7 72 3 156 7 177 7 Cr 3448 620 3500 23 3130 261 3055 178 2200 113 2504 120 2919 138 Mn 1905 57 2293 112 2025 106 2408 92 1568 57 2492 91 2309 80 Co 96 5 112 10 81 5 132 5 46 2 129 5 131 4 Ni 746 47 871 15 736 8 834 30 857 29 1754 60 954 31 Cu 0·87 0·23 0·89 0·13 0·82 0·11 n. a. n. a. n. a. n. a. Rb 0·14 0·02 0·26 0·05 0·20 0·03 0·049 0·004 0·019 0·006 0·013 0·001 Sr 0·41 0·27 0·28 0·03 0·26 0·03 0·48 0·02 0·066 0·007 0·132 0·009 0·236 0·010 Y 3·2 1·0 3·2 0·1 4·4 0·8 3·1 0·2 2·4 0·1 2·0 0·1 2·2 0·1 Zr 8·2 5·6 5·5 0·7 4·3 0·6 n. a. n. a. n. a. n. a. Nb 0·092 0·048 0·055 0·005 0·054 0·013 n. a. n. a. n. a. n. a. Ba 0·265 0·089 0·333 0·021 0·263 0·050 0·136 0·011 0·063 0·025 b. d. l. 0·058 0·006 La 0·130 0·067 0·080 0·025 0·050 0·019 0·158 0·007 0·021 0·004 0·007 0·002 0·037 0·003 Ce 0·444 0·290 0·427 0·087 0·247 0·173 0·565 0·022 0·048 0·005 0·025 0·003 0·143 0·007 Pr 0·073 0·050 0·072 0·031 0·056 0·034 0·080 0·004 0·008 0·002 0·007 0·002 0·025 0·002 Nd 0·385 0·252 0·405 0·138 0·377 0·205 0·394 0·020 0·054 0·014 0·050 0·013 0·140 0·014 Sm 0·174 0·112 0·317 0·086 0·266 0·134 0·137 0·010 0·039 0·014 0·028 0·012 0·072 0·011 Eu 0·051 0·035 0·055 0·024 0·056 0·013 0·026 0·002 0·008 0·003 0·010 0·003 0·012 0·002 Gd 0·295 0·114 0·410 0·096 0·460 0·135 0·227 0·016 0·083 0·019 0·084 0·016 0·135 0·015 Tb 0·060 0·028 0·067 0·003 0·094 0·026 0·050 0·003 0·025 0·004 0·022 0·003 0·033 0·003 Dy 0·505 0·216 0·319 0·086 0·718 0·265 0·446 0·030 0·298 0·032 0·217 0·022 0·301 0·023 Ho 0·123 0·046 0·109 0·008 0·147 0·036 0·112 0·007 0·086 0·009 0·066 0·006 0·078 0·006 Er 0·402 0·124 0·386 0·107 0·524 0·039 0·403 0·026 0·357 0·033 0·272 0·023 0·276 0·019 Tm 0·063 0·018 0·065 0·027 0·084 0·031 0·073 0·005 0·065 0·007 0·052 0·005 0·050 0·004 Yb 0·532 0·183 0·478 0·101 0·672 0·063 0·576 0·041 0·558 0·053 0·485 0·040 0·391 0·030 Lu 0·088 0·025 0·075 0·015 0·092 0·032 0·096 0·007 0·100 0·010 0·089 0·008 0·067 0·005 Hf 0·195 0·140 0·164 0·071 0·258 0·151 n. a. n. a. n. a. n. a. Ta 0·029 0·005 0·028 0·007 0·031 0·012 n. a. n. a. n. a. n. a. Pb 0·238 0·132 0·153 0·078 0·171 0·105 0·101 0·007 0·053 0·010 0·042 0·008 0·113 0·009 Th 0·185 0·178 0·078 0·011 0·132 0·108 n. a. n. a. n. a. n. a. U 0·042 0·034 0·026 0·005 0·016 0·003 n. a. n. a. n. a. n. a. Profile Khuseleka Nkwe Host rock Footwall harzburgite Pyroxenite parting Hanging wall pyroxenite Footwall pyroxenite UG2 chromitite Hanging wall anorthosite Hanging wall pyroxenite N(n) 8 (7) 36 (3) 25 (3) 28 (11) 44 (10) 3 (2) 40 (10) Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. SiO2 54·84 0·27 54·73 0·92 54·81 0·59 54·88 0·42 56·52 0·24 52·77 0·17 54·96 0·31 TiO2 0·13 0·02 0·16 0·04 0·15 0·02 0·16 0·05 0·08 0·02 0·15 0·01 0·15 0·03 Al2O3 1·74 0·06 1·32 0·16 1·38 0·10 1·24 0·24 1·22 0·11 1·22 0·08 1·17 0·07 Cr2O3 0·55 0·06 0·54 0·07 0·51 0·04 0·35 0·10 0·55 0·15 0·09 0·02 0·33 0·05 MgO 30·37 0·63 30·03 1·70 30·25 0·92 28·67 0·62 33·55 0·37 22·73 0·05 28·96 0·78 MnO 0·26 0·02 0·29 0·03 0·27 0·03 0·29 0·03 0·18 0·02 0·42 0·03 0·27 0·03 FeO 10·29 0·46 11·42 1·80 10·93 0·94 14·41 0·72 8·22 0·26 22·52 0·36 13·61 0·84 CaO 2·07 0·83 1·57 1·38 1·66 1·19 0·94 0·47 0·54 0·15 0·83 0·10 1·10 0·51 Na2O 0·03 0·02 0·03 0·03 0·03 0·03 0·02 0·02 0·01 0·01 0·02 0·03 0·02 0·02 Total 100·26 100·09 100·00 101·08 101·01 100·96 100·72 Mg# 0·84 0·82 0·83 0·78 0·88 0·64 0·79 Trace elements, ppm P 35·2 11·3 34·4 10·0 37·5 8·9 n. a. n. a. n. a. n. a. K n. a. n. a. n. a. 6·7 0·4 3·1 0·3 b. d. l. 1·6 0·2 Sc 28·5 1·9 31·4 1·2 36·9 2·3 29·4 1·4 49·7 2·2 30·1 1·3 29·5 1·2 Ti 682 118 885 139 864 181 1191 62 542 26 1148 56 1062 46 V n. a. n. a. n. a. 155 7 72 3 156 7 177 7 Cr 3448 620 3500 23 3130 261 3055 178 2200 113 2504 120 2919 138 Mn 1905 57 2293 112 2025 106 2408 92 1568 57 2492 91 2309 80 Co 96 5 112 10 81 5 132 5 46 2 129 5 131 4 Ni 746 47 871 15 736 8 834 30 857 29 1754 60 954 31 Cu 0·87 0·23 0·89 0·13 0·82 0·11 n. a. n. a. n. a. n. a. Rb 0·14 0·02 0·26 0·05 0·20 0·03 0·049 0·004 0·019 0·006 0·013 0·001 Sr 0·41 0·27 0·28 0·03 0·26 0·03 0·48 0·02 0·066 0·007 0·132 0·009 0·236 0·010 Y 3·2 1·0 3·2 0·1 4·4 0·8 3·1 0·2 2·4 0·1 2·0 0·1 2·2 0·1 Zr 8·2 5·6 5·5 0·7 4·3 0·6 n. a. n. a. n. a. n. a. Nb 0·092 0·048 0·055 0·005 0·054 0·013 n. a. n. a. n. a. n. a. Ba 0·265 0·089 0·333 0·021 0·263 0·050 0·136 0·011 0·063 0·025 b. d. l. 0·058 0·006 La 0·130 0·067 0·080 0·025 0·050 0·019 0·158 0·007 0·021 0·004 0·007 0·002 0·037 0·003 Ce 0·444 0·290 0·427 0·087 0·247 0·173 0·565 0·022 0·048 0·005 0·025 0·003 0·143 0·007 Pr 0·073 0·050 0·072 0·031 0·056 0·034 0·080 0·004 0·008 0·002 0·007 0·002 0·025 0·002 Nd 0·385 0·252 0·405 0·138 0·377 0·205 0·394 0·020 0·054 0·014 0·050 0·013 0·140 0·014 Sm 0·174 0·112 0·317 0·086 0·266 0·134 0·137 0·010 0·039 0·014 0·028 0·012 0·072 0·011 Eu 0·051 0·035 0·055 0·024 0·056 0·013 0·026 0·002 0·008 0·003 0·010 0·003 0·012 0·002 Gd 0·295 0·114 0·410 0·096 0·460 0·135 0·227 0·016 0·083 0·019 0·084 0·016 0·135 0·015 Tb 0·060 0·028 0·067 0·003 0·094 0·026 0·050 0·003 0·025 0·004 0·022 0·003 0·033 0·003 Dy 0·505 0·216 0·319 0·086 0·718 0·265 0·446 0·030 0·298 0·032 0·217 0·022 0·301 0·023 Ho 0·123 0·046 0·109 0·008 0·147 0·036 0·112 0·007 0·086 0·009 0·066 0·006 0·078 0·006 Er 0·402 0·124 0·386 0·107 0·524 0·039 0·403 0·026 0·357 0·033 0·272 0·023 0·276 0·019 Tm 0·063 0·018 0·065 0·027 0·084 0·031 0·073 0·005 0·065 0·007 0·052 0·005 0·050 0·004 Yb 0·532 0·183 0·478 0·101 0·672 0·063 0·576 0·041 0·558 0·053 0·485 0·040 0·391 0·030 Lu 0·088 0·025 0·075 0·015 0·092 0·032 0·096 0·007 0·100 0·010 0·089 0·008 0·067 0·005 Hf 0·195 0·140 0·164 0·071 0·258 0·151 n. a. n. a. n. a. n. a. Ta 0·029 0·005 0·028 0·007 0·031 0·012 n. a. n. a. n. a. n. a. Pb 0·238 0·132 0·153 0·078 0·171 0·105 0·101 0·007 0·053 0·010 0·042 0·008 0·113 0·009 Th 0·185 0·178 0·078 0·011 0·132 0·108 n. a. n. a. n. a. n. a. U 0·042 0·034 0·026 0·005 0·016 0·003 n. a. n. a. n. a. n. a. n. a., not analysed Table 2 Plagioclase compositions, average concentrations (Av.) and standard deviations (S. D.) of N microprobe and n LA ICP-MS analyses Mine Khuseleka Nkwe Host rock Footwall harzburgite UG2 chromitite Pyroxenite parting Hanging wall pyroxenite Footwall pyroxenite UG2 chromitite Hanging wall anorthosite Hanging wall pyroxenite N(n) 9 (7) 79 (3) 32 (3) 45 (3) 41 (11) 45 (21) 50 (8) 98 (15) Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. SiO2 49·95 1·03 50·39 0·90 50·51 1·09 50·81 1·16 55·10 0·83 50·99 0·77 50·27 0·34 52·55 1·34 Al2O3 31·83 0·73 31·88 0·80 31·75 0·77 31·72 0·78 28·82 0·56 31·83 0·48 32·11 0·24 30·48 0·85 FeO 0·32 0·03 0·27 0·18 0·19 0·03 0·19 0·02 0·21 0·09 0·19 0·08 0·23 0·06 0·22 0·11 CaO 14·63 0·79 14·05 0·88 14·24 0·92 13·94 0·77 11·38 0·62 14·72 0·54 15·18 0·24 13·41 0·96 Na2O 2·95 0·40 3·23 0·40 3·22 0·50 3·18 0·49 4·97 0·40 2·99 0·34 2·63 0·15 3·67 0·60 K2O 0·28 0·08 0·11 0·05 0·20 0·05 0·21 0·05 0·17 0·05 0·02 0·01 0·12 0·02 0·20 0·04 Total 99·96 99·92 100·1 100·1 100·7 100·9 100·6 100·6 An 72·1 70·1 70·2 69·9 55·8 73·0 75·9 66·9 Trace elements, ppm Mg 435 241 18 5 84 13 142 15 n. a. n. a. n. a. n. a. P 217 28 316 20 465 52 435 42 n. a. n. a. n. a. n. a. Sc 0·68 0·40 0·72 0·03 1·16 0·39 0·84 0·22 0·66 0·04 0·88 0·05 0·58 0·03 0·79 0·04 Ti 115 28 11 2 87 47 103 25 138 8 60 3 138 8 125 5 Mn 17·7 15·0 2·4 0·1 8·9 2·1 5·1 1·4 6·9 0·3 3·5 0·2 17·4 0·7 7·2 0·3 Fe 2966 440 280 98 1143 88 1062 42 n. a. n. a. n. a. n. a. Sr 479 24 422 17 421 4 454 49 463 19 527 18 497 22 517 17 Ba 146 12 83 27 113 13 148 50 150 6 80 3 57 3 153 6 Rb 0·935 0·39 0·417 0·015 0·685 0·191 0·455 0·163 0·266 0·016 0·023 0·007 0·119 0·009 0·106 0·008 Y 0·209 0·07 0·433 0·179 0·415 0·134 0·330 0·057 0·183 0·013 0·271 0·021 0·192 0·015 0·322 0·018 La 3·61 0·96 5·94 2·45 5·98 1·07 3·70 0·40 10·83 0·40 5·46 0·20 1·48 0·06 7·08 0·26 Ce 8·36 1·93 10·83 3·87 9·98 1·01 6·04 0·45 13·47 0·48 8·76 0·30 2·65 0·10 10·66 0·36 Pr 0·693 0·18 1·040 0·44 0·882 0·038 0·547 0·016 0·961 0·04 0·80 0·04 0·266 0·013 0·885 0·035 Nd 2·335 0·61 3·477 1·42 2·550 0·424 2·105 0·120 2·514 0·12 2·65 0·147 0·957 0·052 2·706 0·136 Sm 0·246 0·10 0·487 0·12 0·231 0·027 0·430 0·269 0·226 0·02 0·28 0·044 0·132 0·018 0·295 0·029 Eu 0·634 0·11 0·720 0·36 0·733 0·038 0·661 0·194 0·835 0·04 0·62 0·037 0·312 0·017 0·870 0·037 Gd 0·232 0·03 0·663 0·15 0·560 0·113 0·595 0·134 0·114 0·02 0·19 0·036 0·088 0·015 0·190 0·022 Tb 0·035 0·014 0·071 0·029 0·065 0·038 0·034 0·008 0·011 0·002 0·019 0·004 0·008 0·002 0·017 0·002 Dy 0·100 0·014 0·280 0·017 0·254 0·122 0·201 0·001 0·046 0·008 0·079 0·019 0·046 0·008 0·078 0·012 Ho b. d. l. b. d. l. b. d. l. b. d. l. 0·007 0·001 0·013 0·003 0·007 0·001 0·010 0·002 Er b. d. l. b. d. l. b. d. l. b. d. l. 0·013 0·003 0·029 0·010 0·014 0·003 0·022 0·005 Tm b. d. l. b. d. l. b. d. l. b. d. l. 0·002 0·001 0·002 0·001 0·002 0·001 0·004 0·001 Yb b. d. l. b. d. l. b. d. l. b. d. l. 0·008 0·003 0·004 0·002 0·011 0·004 0·013 0·004 Lu b. d. l. b. d. l. b. d. l. b. d. l. b. d. l. b. d. l. 0·005 0·002 0·002 0·001 Pb 2·35 0·21 2·16 1·09 2·56 0·08 3·24 1·24 12·85 0·55 12·58 0·52 0·82 0·04 3·75 0·18 Mine Khuseleka Nkwe Host rock Footwall harzburgite UG2 chromitite Pyroxenite parting Hanging wall pyroxenite Footwall pyroxenite UG2 chromitite Hanging wall anorthosite Hanging wall pyroxenite N(n) 9 (7) 79 (3) 32 (3) 45 (3) 41 (11) 45 (21) 50 (8) 98 (15) Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. SiO2 49·95 1·03 50·39 0·90 50·51 1·09 50·81 1·16 55·10 0·83 50·99 0·77 50·27 0·34 52·55 1·34 Al2O3 31·83 0·73 31·88 0·80 31·75 0·77 31·72 0·78 28·82 0·56 31·83 0·48 32·11 0·24 30·48 0·85 FeO 0·32 0·03 0·27 0·18 0·19 0·03 0·19 0·02 0·21 0·09 0·19 0·08 0·23 0·06 0·22 0·11 CaO 14·63 0·79 14·05 0·88 14·24 0·92 13·94 0·77 11·38 0·62 14·72 0·54 15·18 0·24 13·41 0·96 Na2O 2·95 0·40 3·23 0·40 3·22 0·50 3·18 0·49 4·97 0·40 2·99 0·34 2·63 0·15 3·67 0·60 K2O 0·28 0·08 0·11 0·05 0·20 0·05 0·21 0·05 0·17 0·05 0·02 0·01 0·12 0·02 0·20 0·04 Total 99·96 99·92 100·1 100·1 100·7 100·9 100·6 100·6 An 72·1 70·1 70·2 69·9 55·8 73·0 75·9 66·9 Trace elements, ppm Mg 435 241 18 5 84 13 142 15 n. a. n. a. n. a. n. a. P 217 28 316 20 465 52 435 42 n. a. n. a. n. a. n. a. Sc 0·68 0·40 0·72 0·03 1·16 0·39 0·84 0·22 0·66 0·04 0·88 0·05 0·58 0·03 0·79 0·04 Ti 115 28 11 2 87 47 103 25 138 8 60 3 138 8 125 5 Mn 17·7 15·0 2·4 0·1 8·9 2·1 5·1 1·4 6·9 0·3 3·5 0·2 17·4 0·7 7·2 0·3 Fe 2966 440 280 98 1143 88 1062 42 n. a. n. a. n. a. n. a. Sr 479 24 422 17 421 4 454 49 463 19 527 18 497 22 517 17 Ba 146 12 83 27 113 13 148 50 150 6 80 3 57 3 153 6 Rb 0·935 0·39 0·417 0·015 0·685 0·191 0·455 0·163 0·266 0·016 0·023 0·007 0·119 0·009 0·106 0·008 Y 0·209 0·07 0·433 0·179 0·415 0·134 0·330 0·057 0·183 0·013 0·271 0·021 0·192 0·015 0·322 0·018 La 3·61 0·96 5·94 2·45 5·98 1·07 3·70 0·40 10·83 0·40 5·46 0·20 1·48 0·06 7·08 0·26 Ce 8·36 1·93 10·83 3·87 9·98 1·01 6·04 0·45 13·47 0·48 8·76 0·30 2·65 0·10 10·66 0·36 Pr 0·693 0·18 1·040 0·44 0·882 0·038 0·547 0·016 0·961 0·04 0·80 0·04 0·266 0·013 0·885 0·035 Nd 2·335 0·61 3·477 1·42 2·550 0·424 2·105 0·120 2·514 0·12 2·65 0·147 0·957 0·052 2·706 0·136 Sm 0·246 0·10 0·487 0·12 0·231 0·027 0·430 0·269 0·226 0·02 0·28 0·044 0·132 0·018 0·295 0·029 Eu 0·634 0·11 0·720 0·36 0·733 0·038 0·661 0·194 0·835 0·04 0·62 0·037 0·312 0·017 0·870 0·037 Gd 0·232 0·03 0·663 0·15 0·560 0·113 0·595 0·134 0·114 0·02 0·19 0·036 0·088 0·015 0·190 0·022 Tb 0·035 0·014 0·071 0·029 0·065 0·038 0·034 0·008 0·011 0·002 0·019 0·004 0·008 0·002 0·017 0·002 Dy 0·100 0·014 0·280 0·017 0·254 0·122 0·201 0·001 0·046 0·008 0·079 0·019 0·046 0·008 0·078 0·012 Ho b. d. l. b. d. l. b. d. l. b. d. l. 0·007 0·001 0·013 0·003 0·007 0·001 0·010 0·002 Er b. d. l. b. d. l. b. d. l. b. d. l. 0·013 0·003 0·029 0·010 0·014 0·003 0·022 0·005 Tm b. d. l. b. d. l. b. d. l. b. d. l. 0·002 0·001 0·002 0·001 0·002 0·001 0·004 0·001 Yb b. d. l. b. d. l. b. d. l. b. d. l. 0·008 0·003 0·004 0·002 0·011 0·004 0·013 0·004 Lu b. d. l. b. d. l. b. d. l. b. d. l. b. d. l. b. d. l. 0·005 0·002 0·002 0·001 Pb 2·35 0·21 2·16 1·09 2·56 0·08 3·24 1·24 12·85 0·55 12·58 0·52 0·82 0·04 3·75 0·18 Major oxides are in wt %, trace elements are in ppm. b. d. l., below the detection limit; n. a, not analysed Table 2 Plagioclase compositions, average concentrations (Av.) and standard deviations (S. D.) of N microprobe and n LA ICP-MS analyses Mine Khuseleka Nkwe Host rock Footwall harzburgite UG2 chromitite Pyroxenite parting Hanging wall pyroxenite Footwall pyroxenite UG2 chromitite Hanging wall anorthosite Hanging wall pyroxenite N(n) 9 (7) 79 (3) 32 (3) 45 (3) 41 (11) 45 (21) 50 (8) 98 (15) Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. SiO2 49·95 1·03 50·39 0·90 50·51 1·09 50·81 1·16 55·10 0·83 50·99 0·77 50·27 0·34 52·55 1·34 Al2O3 31·83 0·73 31·88 0·80 31·75 0·77 31·72 0·78 28·82 0·56 31·83 0·48 32·11 0·24 30·48 0·85 FeO 0·32 0·03 0·27 0·18 0·19 0·03 0·19 0·02 0·21 0·09 0·19 0·08 0·23 0·06 0·22 0·11 CaO 14·63 0·79 14·05 0·88 14·24 0·92 13·94 0·77 11·38 0·62 14·72 0·54 15·18 0·24 13·41 0·96 Na2O 2·95 0·40 3·23 0·40 3·22 0·50 3·18 0·49 4·97 0·40 2·99 0·34 2·63 0·15 3·67 0·60 K2O 0·28 0·08 0·11 0·05 0·20 0·05 0·21 0·05 0·17 0·05 0·02 0·01 0·12 0·02 0·20 0·04 Total 99·96 99·92 100·1 100·1 100·7 100·9 100·6 100·6 An 72·1 70·1 70·2 69·9 55·8 73·0 75·9 66·9 Trace elements, ppm Mg 435 241 18 5 84 13 142 15 n. a. n. a. n. a. n. a. P 217 28 316 20 465 52 435 42 n. a. n. a. n. a. n. a. Sc 0·68 0·40 0·72 0·03 1·16 0·39 0·84 0·22 0·66 0·04 0·88 0·05 0·58 0·03 0·79 0·04 Ti 115 28 11 2 87 47 103 25 138 8 60 3 138 8 125 5 Mn 17·7 15·0 2·4 0·1 8·9 2·1 5·1 1·4 6·9 0·3 3·5 0·2 17·4 0·7 7·2 0·3 Fe 2966 440 280 98 1143 88 1062 42 n. a. n. a. n. a. n. a. Sr 479 24 422 17 421 4 454 49 463 19 527 18 497 22 517 17 Ba 146 12 83 27 113 13 148 50 150 6 80 3 57 3 153 6 Rb 0·935 0·39 0·417 0·015 0·685 0·191 0·455 0·163 0·266 0·016 0·023 0·007 0·119 0·009 0·106 0·008 Y 0·209 0·07 0·433 0·179 0·415 0·134 0·330 0·057 0·183 0·013 0·271 0·021 0·192 0·015 0·322 0·018 La 3·61 0·96 5·94 2·45 5·98 1·07 3·70 0·40 10·83 0·40 5·46 0·20 1·48 0·06 7·08 0·26 Ce 8·36 1·93 10·83 3·87 9·98 1·01 6·04 0·45 13·47 0·48 8·76 0·30 2·65 0·10 10·66 0·36 Pr 0·693 0·18 1·040 0·44 0·882 0·038 0·547 0·016 0·961 0·04 0·80 0·04 0·266 0·013 0·885 0·035 Nd 2·335 0·61 3·477 1·42 2·550 0·424 2·105 0·120 2·514 0·12 2·65 0·147 0·957 0·052 2·706 0·136 Sm 0·246 0·10 0·487 0·12 0·231 0·027 0·430 0·269 0·226 0·02 0·28 0·044 0·132 0·018 0·295 0·029 Eu 0·634 0·11 0·720 0·36 0·733 0·038 0·661 0·194 0·835 0·04 0·62 0·037 0·312 0·017 0·870 0·037 Gd 0·232 0·03 0·663 0·15 0·560 0·113 0·595 0·134 0·114 0·02 0·19 0·036 0·088 0·015 0·190 0·022 Tb 0·035 0·014 0·071 0·029 0·065 0·038 0·034 0·008 0·011 0·002 0·019 0·004 0·008 0·002 0·017 0·002 Dy 0·100 0·014 0·280 0·017 0·254 0·122 0·201 0·001 0·046 0·008 0·079 0·019 0·046 0·008 0·078 0·012 Ho b. d. l. b. d. l. b. d. l. b. d. l. 0·007 0·001 0·013 0·003 0·007 0·001 0·010 0·002 Er b. d. l. b. d. l. b. d. l. b. d. l. 0·013 0·003 0·029 0·010 0·014 0·003 0·022 0·005 Tm b. d. l. b. d. l. b. d. l. b. d. l. 0·002 0·001 0·002 0·001 0·002 0·001 0·004 0·001 Yb b. d. l. b. d. l. b. d. l. b. d. l. 0·008 0·003 0·004 0·002 0·011 0·004 0·013 0·004 Lu b. d. l. b. d. l. b. d. l. b. d. l. b. d. l. b. d. l. 0·005 0·002 0·002 0·001 Pb 2·35 0·21 2·16 1·09 2·56 0·08 3·24 1·24 12·85 0·55 12·58 0·52 0·82 0·04 3·75 0·18 Mine Khuseleka Nkwe Host rock Footwall harzburgite UG2 chromitite Pyroxenite parting Hanging wall pyroxenite Footwall pyroxenite UG2 chromitite Hanging wall anorthosite Hanging wall pyroxenite N(n) 9 (7) 79 (3) 32 (3) 45 (3) 41 (11) 45 (21) 50 (8) 98 (15) Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. Av. S. D. SiO2 49·95 1·03 50·39 0·90 50·51 1·09 50·81 1·16 55·10 0·83 50·99 0·77 50·27 0·34 52·55 1·34 Al2O3 31·83 0·73 31·88 0·80 31·75 0·77 31·72 0·78 28·82 0·56 31·83 0·48 32·11 0·24 30·48 0·85 FeO 0·32 0·03 0·27 0·18 0·19 0·03 0·19 0·02 0·21 0·09 0·19 0·08 0·23 0·06 0·22 0·11 CaO 14·63 0·79 14·05 0·88 14·24 0·92 13·94 0·77 11·38 0·62 14·72 0·54 15·18 0·24 13·41 0·96 Na2O 2·95 0·40 3·23 0·40 3·22 0·50 3·18 0·49 4·97 0·40 2·99 0·34 2·63 0·15 3·67 0·60 K2O 0·28 0·08 0·11 0·05 0·20 0·05 0·21 0·05 0·17 0·05 0·02 0·01 0·12 0·02 0·20 0·04 Total 99·96 99·92 100·1 100·1 100·7 100·9 100·6 100·6 An 72·1 70·1 70·2 69·9 55·8 73·0 75·9 66·9 Trace elements, ppm Mg 435 241 18 5 84 13 142 15 n. a. n. a. n. a. n. a. P 217 28 316 20 465 52 435 42 n. a. n. a. n. a. n. a. Sc 0·68 0·40 0·72 0·03 1·16 0·39 0·84 0·22 0·66 0·04 0·88 0·05 0·58 0·03 0·79 0·04 Ti 115 28 11 2 87 47 103 25 138 8 60 3 138 8 125 5 Mn 17·7 15·0 2·4 0·1 8·9 2·1 5·1 1·4 6·9 0·3 3·5 0·2 17·4 0·7 7·2 0·3 Fe 2966 440 280 98 1143 88 1062 42 n. a. n. a. n. a. n. a. Sr 479 24 422 17 421 4 454 49 463 19 527 18 497 22 517 17 Ba 146 12 83 27 113 13 148 50 150 6 80 3 57 3 153 6 Rb 0·935 0·39 0·417 0·015 0·685 0·191 0·455 0·163 0·266 0·016 0·023 0·007 0·119 0·009 0·106 0·008 Y 0·209 0·07 0·433 0·179 0·415 0·134 0·330 0·057 0·183 0·013 0·271 0·021 0·192 0·015 0·322 0·018 La 3·61 0·96 5·94 2·45 5·98 1·07 3·70 0·40 10·83 0·40 5·46 0·20 1·48 0·06 7·08 0·26 Ce 8·36 1·93 10·83 3·87 9·98 1·01 6·04 0·45 13·47 0·48 8·76 0·30 2·65 0·10 10·66 0·36 Pr 0·693 0·18 1·040 0·44 0·882 0·038 0·547 0·016 0·961 0·04 0·80 0·04 0·266 0·013 0·885 0·035 Nd 2·335 0·61 3·477 1·42 2·550 0·424 2·105 0·120 2·514 0·12 2·65 0·147 0·957 0·052 2·706 0·136 Sm 0·246 0·10 0·487 0·12 0·231 0·027 0·430 0·269 0·226 0·02 0·28 0·044 0·132 0·018 0·295 0·029 Eu 0·634 0·11 0·720 0·36 0·733 0·038 0·661 0·194 0·835 0·04 0·62 0·037 0·312 0·017 0·870 0·037 Gd 0·232 0·03 0·663 0·15 0·560 0·113 0·595 0·134 0·114 0·02 0·19 0·036 0·088 0·015 0·190 0·022 Tb 0·035 0·014 0·071 0·029 0·065 0·038 0·034 0·008 0·011 0·002 0·019 0·004 0·008 0·002 0·017 0·002 Dy 0·100 0·014 0·280 0·017 0·254 0·122 0·201 0·001 0·046 0·008 0·079 0·019 0·046 0·008 0·078 0·012 Ho b. d. l. b. d. l. b. d. l. b. d. l. 0·007 0·001 0·013 0·003 0·007 0·001 0·010 0·002 Er b. d. l. b. d. l. b. d. l. b. d. l. 0·013 0·003 0·029 0·010 0·014 0·003 0·022 0·005 Tm b. d. l. b. d. l. b. d. l. b. d. l. 0·002 0·001 0·002 0·001 0·002 0·001 0·004 0·001 Yb b. d. l. b. d. l. b. d. l. b. d. l. 0·008 0·003 0·004 0·002 0·011 0·004 0·013 0·004 Lu b. d. l. b. d. l. b. d. l. b. d. l. b. d. l. b. d. l. 0·005 0·002 0·002 0·001 Pb 2·35 0·21 2·16 1·09 2·56 0·08 3·24 1·24 12·85 0·55 12·58 0·52 0·82 0·04 3·75 0·18 Major oxides are in wt %, trace elements are in ppm. b. d. l., below the detection limit; n. a, not analysed Synchrotron radiation X-ray fluorescence microanalysis (SR XRF) X-ray fluorescence microanalysis of the Khuseleka drill core section was carried out at the Institute of Nuclear Physics, Siberian Branch of the Russian Academy of Sciences in Novosibirsk using a synchrotron radiation beam from the electron–positron VEPP-3 storage ring with a perimeter of 74·4 m, injection energy of 350 MeV and maximal energy of 2000 MeV. In the early 1990s, the source was equipped with a one-dimensional scanner for samples up to 300 mm in length, with a minimum step size of 10 μm. The system for collimation and focusing can produce a beam of minimum 100 μm in diameter at the excitation energy of 15–30 keV. The XRF scanner is routinely used for chemical analyses of lacustrine and marine sediments, but this study is the first application to igneous rocks. Darin et al. (2016) describe the results of trial analyses carried out on a layered petrographic thin section of the UG1 chromitite. The samples used in this study were drill core fragments 37 × 20 × 5 mm in size, polished on one side. They were fixed into the scanner and moved at 1·2 mm increments. The size of the collimated beam of excitation radiation was 1 mm along the profile and 20 mm across. Scanning was done twice at different excitation energies of 20 and 30 keV. The acquired spectra were processed using the AXIL software (Van Espen et al., 1986). The algorithm is based on nonlinear optimisation by the least squares method. The fitting functions are a set of Gaussians with the positions and mutual intensities of individual peaks corresponding to the tabulated positions and intensities of fluorescence lines for detected elements. Concentrations of elements were calculated from the intensities of their analytical lines using the external standard method. No suitable international standard was available for chromitite, so concentrations were calibrated against electron microprobe and conventional whole-rock XRF data from the same sample. Reference elements used in the calibration included K, Ca, Ti, Cr, Mn and Fe. The XRF data are given in the Supplementary Data; supplementary data are available for downloading at http://www.petrology.oxfordjournals.org. We acknowledge that absolute concentrations of some trace elements may have systematic errors that are difficult to evaluate. However, the relative variations within the chromitite seams primarily depend on the intensities of the analytical lines and not on the calibration procedure. Crystal size distribution Crystal size distribution (CSD) analysis of chromite crystals in massive chromitite seams and disseminated crystals in the silicate rocks below and above the seams was carried out using reflected light photomicrographs of polished thin sections from the Nkwe and Khuseleka profiles using 2 D image analysis and the 2D-3D conversion method developed by Higgins (2000). In the main chromitite seam, samples for the CSD analysis were taken from different levels, at the top, in the middle and at the bottom. Numerous random areas of polished petrographic thin sections were photographed under a microscope in reflected light and the digital greyscale photographs were transformed into binary images showing chromite crystals as black areas on a white background. Binary images were manually edited by removing artefacts due to polishing defects and highlighting interfaces between touching crystals. The exposed areas of all individual chromite crystals were contoured and measured by maximal Feret diameters using ImageJ software. The resulting tables of Feret diameters were fed into the CSDCorrections 1.50 software (Higgins, 2000), which calculated the conventional CSD curves and histograms. In the software input parameters, rock fabric was defined as massive, the input crystal dimension was set to the maximal size, crystal shapes were defined as equidimensional in all three axes and crystal roundness was set to zero. The minimal number of chromite grains used for the construction of the CSD curves was 520 (in Khuseleka harzburgite) and in all the other cases the number was greater than 1500. RESULTS Petrography and electron microprobe analyses The Khuseleka profile The UG2 chromitite seam is 80 cm thick and consists of euhedral to anhedral chromite crystals with an average size of about 0·2 mm (Fig. 3). The crystals are usually poikilitically enclosed by plagioclase and pyroxene oikocrysts up to 7 mm across, but locally form aggregates of larger, tightly intergrown anhedral crystals (chromitite textures and crystal size distribution are discussed in more detail in a separate section below). Chromite compositions throughout the seam have Mg# [atomic ratio Mg/(Mg+Fe2+)] at 0·45–0·52, atomic Cr/(Cr+Al+Fe3+) at 0·57–0·59 and TiO2 concentrations at about 0·9–1·2 wt %. The orthopyroxene oikocrysts have Mg# of 0·88–0·91, which is significantly more magnesian than orthopyroxene in the footwall harzburgite (0·79–0·85). Plagioclase in the chromitite seam varies within the An60–80 range, except for a narrow interval in the upper part where the anorthite content decreases to An28 (Fig. 4b). Minor interstitial minerals are mostly biotite and high-Ca clinopyroxene. Apart from exsolution lamellae in orthopyroxene, clinopyroxene locally forms isolated interstitial crystals. Fig. 3 View largeDownload slide Photographs of petrographic thin sections in (a–d) transmitted and (e–f) reflected light. (a) Footwall harzburgite, Khuseleka. (b) Sharp contact of the main chromitite seam with overlying pyroxenite parting, Khuseleka. (c) Hanging wall pyroxenite, Khuseleka. (d) A single poikilitic orthopyroxene crystal in the main chromitite seam, Nkwe, cross-polarized light. (e) Transition from poikilitically enclosed chromite to an aggregate of larger chromite crystals, Nkwe. (f) An aggregate of tightly intergrown chromite crystals with sulphides at triple junctions, Nkwe. Abbreviations for minerals: Chr, chromite; Ol, olivine; Opx, orthopyroxene; Pl, plagioclase; S, sulphide. Scale bars below the images correspond to 2 mm. Fig. 3 View largeDownload slide Photographs of petrographic thin sections in (a–d) transmitted and (e–f) reflected light. (a) Footwall harzburgite, Khuseleka. (b) Sharp contact of the main chromitite seam with overlying pyroxenite parting, Khuseleka. (c) Hanging wall pyroxenite, Khuseleka. (d) A single poikilitic orthopyroxene crystal in the main chromitite seam, Nkwe, cross-polarized light. (e) Transition from poikilitically enclosed chromite to an aggregate of larger chromite crystals, Nkwe. (f) An aggregate of tightly intergrown chromite crystals with sulphides at triple junctions, Nkwe. Abbreviations for minerals: Chr, chromite; Ol, olivine; Opx, orthopyroxene; Pl, plagioclase; S, sulphide. Scale bars below the images correspond to 2 mm. Fig. 4 View largeDownload slide Variations of orthopyroxene (a) and plagioclase (b,c) compositions in the Khuseleka drill core profile. (a-c) major element ratios; (d) concentrations of Ti. Background colours correspond to different rock types: white, pyroxenite and harzburgite; grey, chromitite. Fig. 4 View largeDownload slide Variations of orthopyroxene (a) and plagioclase (b,c) compositions in the Khuseleka drill core profile. (a-c) major element ratios; (d) concentrations of Ti. Background colours correspond to different rock types: white, pyroxenite and harzburgite; grey, chromitite. The lower contact of the UG2 main seam is sharp. The footwall rock is a coarse-grained feldspathic harzburgite (Fig. 3a) composed of euhedral olivine crystals (Fo82–83) up to 7 mm in size, euhedral to subhedral orthopyroxene grains up to 15 mm across and interstitial plagioclase (An60–81). The orthopyroxene Mg# gradually increases from 0·79 to 0·84 toward the contact with the main chromitite seam (Fig. 4a). The compositions of disseminated chromite crystals vary broadly (Fig. 5). The values of Mg# and Fe3+/total Fe in them tend to increase toward the chromitite contact, whereas the TiO2 contents decrease in the same direction and Cr/(Cr+Al+Fe3+) ratios are nearly constant (Fig. 5b–d). Biotite, amphibole and clinopyroxene are present in the rock in minor amounts as interstitial phases. Fig. 5 View largeDownload slide Variations of chromite composition in the Khuseleka drill core profile. (a-c) major element ratios; (d) concentrations of Ti. Background colours correspond to different rock types: white, pyroxenite and harzburgite; grey, chromitite. Fig. 5 View largeDownload slide Variations of chromite composition in the Khuseleka drill core profile. (a-c) major element ratios; (d) concentrations of Ti. Background colours correspond to different rock types: white, pyroxenite and harzburgite; grey, chromitite. The upper contact of the UG2 main seam is also sharp (Fig. 3b), and above it is a 23 cm-thick layer (‘parting’) of fine-grained orthopyroxenite with minor interstitial plagioclase and minor disseminated chromite. This parting separates the main chromitite seam from a thin (6 cm) chromite stringer (‘leader’ in mining terminology) above. The rock above the chromite leader is feldspathic pyroxenite with more abundant chromite than the orthopyroxenite in the parting and the chromite grains are arranged in sub-horizontal ‘chains’ (Fig. 3c). Orthopyroxene compositions in the parting and hanging wall pyroxenite are practically identical, with Mg# of 0·79–0·84 and thus less magnesian than the oikocrysts in the massive chromitite (Fig. 4a). Plagioclase compositions in the parting and the hanging wall pyroxenite vary broadly from An35 to An78, but tend to have lower An and higher K2O contents than poikilitic plagioclase in the chromitite seams (Fig. 4b and c). Chromite from the pyroxenite parting, hanging wall pyroxenite, the main seam and the leader define a linear trend of decreasing Cr/(Cr+Al) with increasing Mg/(Mg+Fe2+). This trend correlates with decreasing chromite abundance and corresponds to ‘trend A’ of Naldrett et al. (2012). By contrast, disseminated chromite crystals in the footwall harzburgite plot off the trend at lower Cr/(Cr+Al) values (Fig. 6a) implying that some other factors affected chromite compositions there. Fig. 6 View largeDownload slide Variations in the Cr/(Cr+Al) vs Mg/(Mg+Fe2+) of chromites from massive chromitite seams and silicate rocks at contacts. (a) Khuseleka, (b) Nkwe. Fig. 6 View largeDownload slide Variations in the Cr/(Cr+Al) vs Mg/(Mg+Fe2+) of chromites from massive chromitite seams and silicate rocks at contacts. (a) Khuseleka, (b) Nkwe. The Nkwe profile The UG2 chromitite layer at Nkwe Platinum mine in the eastern limb is a single, 70 cm thick seam but, as typical for the eastern limb, there is a UG3 layer positioned approximately 10 m above it. The UG2 layer is underlain by pegmatoidal feldspathic pyroxenite, whereas the hanging wall rock is a 10 cm thick, chromite-free anorthosite layer which, in turn, is overlain by pyroxenite and UG3 chromitite. Both contacts of the UG2 seam are sharp. Euhedral to subhedral chromite crystals comprise 62–72 vol. % of the seam. Like in the Khuseleka UG2 samples, chromite is either poikilitically enclosed by plagioclase and orthopyroxene oikocrysts or forms clusters of bigger, tightly intergrown anhedral crystals. However, at Nkwe the clusters are notably more abundant and larger. Towards the lower contact with pyroxenite, larger chromite crystals contain numerous rectangular to roundish inclusions, 10–70 µm in size, composed of orthopyroxene, talc, Na-rich biotite (aspidolite) and rutile. Chromite near the upper contact is inclusion-free. Similar inclusions in chromite were described by Spandler et al. (2005) and Li et al. (2005) from chromitite-footwall contacts in the Stillwater Complex and the Bushveld Complex and are interpreted as crystallized melt inclusions. Vukmanovic et al. (2013) noted the presence of chromite crystals with similar inclusions in the chromitite seam at the bottom of the Merensky Reef. The footwall pegmatoidal pyroxenite is composed of euhedral to subhedral coarse grained orthopyroxene and anhedral interstitial plagioclase. The rock also contains abundant biotite and minor disseminated chromite. Typical interstitial phases are hornblende, biotite, pumpellyite, rutile, calcite and sulphide. Similar observations in the footwall pyroxenite were reported in the eastern limb by Cameron (1975) and in the western limb by Penberthy & Merkle (1999) and Hulbert & von Gruenewaldt (1985). Mineral compositions in the Nkwe profile were reported by Veksler et al. (2015), but that work did not present vertical profiles. Briefly, orthopyroxene Mg# in footwall pegmatoidal pyroxenite is at about 0·7 and towards the contact with UG2 the value increases to 0·81. The Mg# of poikilitic orthopyroxene cementing the chromitite seam is 0·88 and the values for orthopyroxene in hanging wall anorthosite and pyroxenite are 0·77 and 0·79–0·8, respectively. Plagioclase in anorthosite is An75–77; in hanging wall pyroxenite the composition is An60–75; poikilitic plagioclase in the chromitite seam is An66–76 and in the footwall pegmatoidal rock plagioclase compositions vary broadly between An36 and An76. The composition of disseminated chromite in the footwall pegmatoid also varies broadly and generally follows ‘trend A’ of Naldrett et al. (2012). However, chromite from the chromitite seam has higher Mg# (0·37–0·53) and does not plot along the same trend (Fig. 6c). The profile shows a subtle trend towards higher Mg# and an increase of Fe3+ towards the top of the seam (Fig. 7). At three levels of the main seam, one near the top and two in the lower third, TiO2 and to some extent V2O3 show a two-fold enrichment, and at these three levels chromite shows more coarsening. Fig. 7 View largeDownload slide Variations of chromite composition in the Nkwe drill core profile. Major element ratios (a-c) and minor element concentrations (d-f). Background colours correspond to different rock types: white, hanging wall anorthosite; dark grey, chromitite; light grey, footwall pegmatitic pyroxenite. Fig. 7 View largeDownload slide Variations of chromite composition in the Nkwe drill core profile. Major element ratios (a-c) and minor element concentrations (d-f). Background colours correspond to different rock types: white, hanging wall anorthosite; dark grey, chromitite; light grey, footwall pegmatitic pyroxenite. Trace element distribution Samples in the Khuseleka profile chosen for LA-ICP-MS trace element analysis were taken from the top, middle and bottom parts of the UG2 layer, and from the footwall and the hanging wall silicate rocks. The results are presented together with a summary of the electron microprobe data in Tables 1 and 2. Trace element compositions of minerals from the Nkwe samples were reported by Veksler et al. (2015), and we use data from that publication in this study. Interstitial clinopyroxene is more abundant in the Nkwe samples and only in that profile were we able to find grains large enough for laser ablation analysis. A significant, if not the predominant, proportion of incompatible trace elements in plutonic rocks resides not in the major minerals but in interstitial or included accessory minerals, many of which form at the latest stages of crystallization and/or in the sub-solidus field. The contribution of these minerals to the trace element budget is not captured by microprobe mineral analyses of the major phases and, therefore, important information on magma evolution may be overlooked. Whole-rock analysis captures the contribution of all minerals present, but has the drawback that samples are crushed to powder and small-scale spatial variations are lost. As an alternative method for bulk analysis we tested the use of synchrotron-source XRF microanalysis (SR XRF), which provides information on a spatial scale intermediate between the in situ microbeam methods and conventional XRF, and has the advantage of being non-destructive. Orthopyroxene, clinopyroxene, and REE geothermometry Concentrations of compatible elements Cr, Co, Ni and Sc, and moderately incompatible Ti in orthopyroxene from the Khuseleka and Nkwe samples are similar and consistent with crystallization of the mineral from B1 parental magma, assuming element abundances in B1 as proposed by Barnes et al. (2010) and experimentally-determined orthopyroxene-melt D values compiled and reviewed by Bédard [(2007); see also Veksler et al. (2015) for more discussion]. Orthopyroxene oikocrysts in UG2 chromitite are depleted in Cr and Co, probably because of reaction with chromite, the same process which resulted in the higher Mg# of the oikocrysts (Figs 5 and 7). Experimental data on REE distribution between orthopyroxene, clinopyroxene and melt have been reviewed by Yao et al. (2012), Sun & Liang (2012) and Liang et al. (2013), and in Fig. 8 we compare theoretical predictions from their studies with the observed REE concentrations from UG2. If REE concentrations in the B1 parental magma proposed by Barnes et al. (2010) were correct and if the orthopyroxene retained liquidus compositions, the observed ratio of REE concentrations in orthopyroxene to the B1 values would have agreed with the orthopyroxene-melt D values. In fact, this is not the case. For the observed REE orthopyroxene B1 ratios in all rock types and both locations exceed the experimental D values by approximately an order of magnitude, and the excess is greatest for the light REE from La to Nd. Experimental D values according to Yao et al. (2012) are plotted for two temperatures, 890 and 1170°C, corresponding to the minimal and maximal estimates obtained by the two-pyroxene thermometer explained below. Notably, the mismatch between the observed concentrations and the theoretical predictions for light REE becomes greater with falling temperature. Fig. 8 View largeDownload slide REE concentrations in orthopyroxene from different lithologies at Khusleka (a) and Nkwe (b) (Table 1) normalized to the concentrations in the model parental liquid B1 (Barnes et al., 2010). (a) Khuseleka; (b) Nkwe. The elements are in the order of decreasing orthopyroxene-melt distribution coefficient. Experimentally determined orthopyroxene-melt D values from Yao et al. (2012) depend on temperature and are shown for 890 and 1170°C by dashed and solid lines. Fig. 8 View largeDownload slide REE concentrations in orthopyroxene from different lithologies at Khusleka (a) and Nkwe (b) (Table 1) normalized to the concentrations in the model parental liquid B1 (Barnes et al., 2010). (a) Khuseleka; (b) Nkwe. The elements are in the order of decreasing orthopyroxene-melt distribution coefficient. Experimentally determined orthopyroxene-melt D values from Yao et al. (2012) depend on temperature and are shown for 890 and 1170°C by dashed and solid lines. Liang et al. (2013) used experimental data on REE partitioning between coexisting orthopyroxene, clinopyroxene and melt to calibrate a two-pyroxene REE thermometer. They applied the thermometer to seven samples of harzburgites and pyroxenites from the Lower and Lower Critical zones of the Bushveld Complex analysed by Godel et al. (2011) and obtained equilibration temperatures between 1080 and 1230°C. They noted that these values were much higher than those calculated using the two-pyroxene Ca exchange thermometer of Brey & Köhler (1990), and attributed the difference to the relatively fast cooling rate of the Bushveld rocks and lower closure temperature for Ca diffusion than for REE. Our application of the REE and Ca exchange thermometers to the UG2 samples produced similar results to Liang et al. (2013) (Fig. 9). Unfortunately, clinopyroxene crystals large enough for laser ablation are sparse in our samples of the UG2 layer and we were able to analyse only three samples from Nkwe. Equilibration temperature in the hanging wall pyroxenite at 1174 ± 19°C is in a good agreement with the crystallization temperature of the Upper Critical Zone according to experimental studies (Cawthorn & Davies, 1983). Equilibration temperature in footwall pyroxentite at 1288 ± 12°C appears to be too high for the Upper Critical Zone, whereas the equilibration temperature calculated for the UG2 chromitite (889 ± 29°C) is low. Furthermore, the two-pyroxene distribution of light REE elements (La to Pr) deviates from the predicted equilibrium trend towards greater enrichment of orthopyroxene. Similar deviations were observed in Bushveld samples by Liang et al. (2013) and those elements were excluded from the temperature calculations, as in our example. Equilibration temperatures according to the Ca exchange thermometer give consistently low values in the 840–885°C interval for all samples, which is also in agreement with Liang et al. (2013). Fig. 9 View largeDownload slide REE distribution between orthopyroxene and clinopyroxene and the results of the two-pyroxene thermometry for hanging wall pyroxenite (a), chromitite (b) and footwall pyroxenite (c) at Nkwe. The observed orthopyroxene–clinopyroxene concentration ratios (D Opx/Cpx) are plotted as open circles; the predicted values according to Yao et al. (2012) and Sun & Liang (2012) are plotted as grey solid lines. Temperature estimates TREE are according to the two-pyroxene REE thermometer (Liang et al., 2013); TBKN are according to the thermometer based on Ca distribution (Brey & Köhler, 1990). Fig. 9 View largeDownload slide REE distribution between orthopyroxene and clinopyroxene and the results of the two-pyroxene thermometry for hanging wall pyroxenite (a), chromitite (b) and footwall pyroxenite (c) at Nkwe. The observed orthopyroxene–clinopyroxene concentration ratios (D Opx/Cpx) are plotted as open circles; the predicted values according to Yao et al. (2012) and Sun & Liang (2012) are plotted as grey solid lines. Temperature estimates TREE are according to the two-pyroxene REE thermometer (Liang et al., 2013); TBKN are according to the thermometer based on Ca distribution (Brey & Köhler, 1990). Plagioclase In Fig. 10, we plot trace element concentrations in plagioclase normalized to the element abundances in the hypothetical B1 parental melt (Barnes et al., 2010) and compare the normalized values with experimental plagioclase-melt distribution coefficients compiled and reviewed by Bédard (2006). The comparison is aimed at testing whether the observed trace element concentrations are primary magmatic and consistent with plagioclase crystallization from the B1 liquid. A close match between the B1-normalized concentrations and experimental plagioclase-melt D values is observed for all analysed trace elements, with the exception of Pb (Veksler et al., 2015), in plagioclase from the hanging wall anorthosite at Nkwe (Fig. 10b). Good consistency with magmatic crystallization from the B1 parental magma is also observed for all the studied samples, with remarkably constant concentrations of the most compatible elements in plagioclase (Sr and Eu). In contrast, REE concentrations in interstitial plagioclase from pyroxenite, chromitite and harzburgite show enrichment relative to the predicted curve, consistent with REE accumulation in residual intercumulus liquid (trapped liquid shift effect) and possibly also an increase of plagioclase-melt DREE values with falling temperature (Bédard, 2006). Poikilitic plagioclase from the main UG2 seam is strongly depleted in K and Ti relative to the mineral in the silicate rocks. The depletion in Ti is probably due to diffusion of that element to chromite. The depletion in K, which is evident in Fig. 4c, was first described in detail by Veksler et al. (2015) and is attributed to alkali migration out of the chromitite layer during postcumulus crystallization (see discussion). Fig. 10 View largeDownload slide Trace element concentrations in plagioclase from different lithologies at Khusleka (a) and Nkwe (b) normalized to the concentrations in the model parental liquid B1 (Barnes et al., 2010). The elements are in the order of decreasing plagioclase-melt distribution coefficient. Grey solid line in (a) shows the pattern of experimentally determined plagioclase-melt D values for the plagioclase composition An70. Grey field in (b) shows the range of the D values increasing from An76 to An55. All the D values are from Bédard (2006). Fig. 10 View largeDownload slide Trace element concentrations in plagioclase from different lithologies at Khusleka (a) and Nkwe (b) normalized to the concentrations in the model parental liquid B1 (Barnes et al., 2010). The elements are in the order of decreasing plagioclase-melt distribution coefficient. Grey solid line in (a) shows the pattern of experimentally determined plagioclase-melt D values for the plagioclase composition An70. Grey field in (b) shows the range of the D values increasing from An76 to An55. All the D values are from Bédard (2006). Synchrotron radiation XRF results Concentration data collected during the SR XRF scanning were averaged for each 10 analyses, so each point on the vertical profiles in Fig. 11 and the Supplementary Data represents integration over an area of approximately 2 cm2, 10 mm along the profile and 20 mm across. The elements V, Zn and Ga are compatible in chromite and the distribution of these elements in the profile show a very high covariance and a strong correlation with Cr2O3 as a proxy for the modal amounts of chomite (Fig. 11a–d). However, a closer look at the V variations shows an increase in the upper part of the main seam and even higher concentrations of V in the chromitite leader. We have no synchrotron data from the Nkwe profile, but note that the microprobe data for chromite also shows an abrupt increase in V (and Ti) in the upper section (Fig. 7e). An increase of V concentrations at the top of the UG2 layer was also noted by Naldrett et al. (2012). Fig. 11 View largeDownload slide Concentrations of selected trace elements (a-n) in the Khuseleka profile analysed by synchrotron XRF. The background colours correspond to different rock types: white, pyroxenite and harzburgite; grey, chromitite. Fig. 11 View largeDownload slide Concentrations of selected trace elements (a-n) in the Khuseleka profile analysed by synchrotron XRF. The background colours correspond to different rock types: white, pyroxenite and harzburgite; grey, chromitite. The pattern of incompatible Rb, Y and Zr variations is of special interest because of their potential to indicate the distribution of interstitial residual liquid. In the silicate layers, these elements have consistently low concentrations in footwall harzburgite, and higher but variable contents in the pyroxenite parting and the hanging wall pyroxenite (Fig. 11e–g). Notably, Zr and Y show a close covariance throughout the profile, probably reflecting a specific mineral host (zircon is present as an accessory phase). Rubidium correlates generally with Zr and Y, but less strongly since it is likely associated with interstitial mica or amphibole. Inside the main UG2 chromitite, all three elements show heterogeneous and correlated variations, with peaks at the bottom and near the top of the layer, but as in the wall-rocks, Rb has a more independent behaviour than Zr and Y. In view of their different mineral hosts, the correlation between Rb and Zr or Rb and Y is remarkable and probably implies a greater proportion of trapped interstitial liquid near the contacts of the main seam. Two other incompatible elements, Nb and Mo, show a different distribution pattern. Both have low and highly variable concentrations, but they show a strong covariance. The correlation and erratic concentrations suggest that both elements are hosted by the same accessory phase or phases. Rutile is a logical host for Nb. It occurs in our samples as small needles and anhedral grains within, and interstitial to, chromite crystals, and has been described in association with chromite elsewhere in the Upper Critical Zone (e.g.in the UG2: Junge et al., 2014; Merensky chromitite: Vukmanovic et al., 2013). Notably, Mo shows no correlation with Cu, Pd or As, so it is apparently not concentrated in sulphides, and given the strong correlation with Nb we suggest it may also be hosted in rutile. We found no information on Mo partitioning in natural rutile, but the idea is supported by the existence of Mo-doped synthetic TiO2 polymorphs, which are used in photocatalysis (e.g.Devi & Murthy, 2008). The highest concentrations of Ni are observed in the footwall harzburgite, where it mostly resides in olivine, and in a sulphide-rich zone near the top of the main chromitite seam, where Ni probably resides in pentlandite (Fig. 11j). Sulphide enrichment in that part of the profile is confirmed by reflected light microscopy and by the correlated strong peaks of Cu, Pd and As concentrations (Fig. 11k, l and n). Note that the distribution patterns of Pd and Ru are significantly different. In particular, Ru shows no spike in the sulphide zone. Ruthenium is known to form laurite (RuS2) inclusions in chromite (e.g.Maier et al., 1999), and it does show a correlation with As in much of the main seam and the chromite leader. However, it is not concentrated with Cu and Pd in the sulphide zone. The decoupling of Pd and Ru was described in the larger study of PGE in Bushveld chromitites by Naldrett et al. (2012), who concluded that they are controlled by different processes, Ru being incorporated during chromite crystallization and Pd being concentrated in an interstitial sulphide melt. Textural types of chromite and crystal size distribution Following Cameron (1969, 1975) and earlier works cited therein, we distinguish three types of textural relationships between chromite and major rock-forming silicates in the UG2 layer. The first type is what Cameron (1975) called inter-grain chromite. This texture is observed in the rocks above and below the main chromite seams where chromite is interstitial to silicate crystals. In our samples the ‘inter-grain’ chromite forms networks surrounding silicate crystals or is aligned in sub-horizontal chains (Fig. 3c). The textural types 2 and 3 are characteristic of the main UG2 seam and (in Khuseleka) the upper leader. The most common of the two types comprises chromite grains enclosed by orthopyroxene and plagioclase oikocrysts reaching a few centimetres in size (Fig. 3d and e). In the oikocrysts, size and the volume proportion of chromite tend to increase from the centre to the rim and the centre may be completely inclusion-free. In textural type 3, chromite forms aggregates of tightly intergrown, anhedral crystals practically devoid of silicates. The size of individual chromite crystals in the aggregates is notably larger than in the other two textural types and the grain boundaries are curved (Fig. 3e and f), which is a good indication for textural annealing, as discussed in detail below. The chromite aggregates commonly surround orthopyroxene oikocrysts with chromite-free cores. Sulphides are often found in the aggregates at triple junctions between anhedral chromite crystals (Fig. 3f). The coarse chromite aggregates are present in samples from both studied profiles, but they are much more abundant at Nkwe. Their distribution is uneven in the profiles, and the proportion of chromite forming aggregates vs the poikilitic texture may change, even within the area of a standard thin section. However, we do not observe major changes in texture and modal composition between the upper and lower parts of the UG2 layer like those documented by Lee (1996) and Mathez & Mey (2005). The three textural types of chromite need consideration in the analysis of crystal size distribution because the ability of chromite grains to adjust in shape and size during post-cumulus processes differs depending on whether or not they are enclosed in silicate minerals (see discussion). We determined CSD curves for chromite in the main UG2 chromitite seam and adjacent wall-rocks at Khuseleka, but the wall-rocks at Nkwe contained too few crystals for a statistically valid analysis. The number of crystals processed in the samples varied from 500 in footwall harzburgite at Khuseleka to around 8000 in the main seam from both profiles. Note that the CSD measurements were made on several thin sections from all parts of the UG2 seam. In both profiles, we found no variations in the shapes of the CSD curves from the lower and upper parts of the seam and, therefore, have pooled the data in the curves on Fig. 12. All data from the silicate layers refer to texture type 1 (inter-grain). The distinction between textural types 2 and 3 (poikilitic vs aggregate) in the chromitite seams was made by the presence or absence of triple junctions. Thus, all chromite grains in chromitite forming triple junctions were selected as belonging to the aggregates and all others were attributed to the poikilitic group. At Khuseleka, there were not enough grains in the aggregate type to calculate statistically valid CSD curves and, therefore, all chromite grains were analysed together. In the Nkwe chromitite the proportion of crystal aggregates is high and we generated separate CSD curves for the poikilitic and aggregate chromite types (Fig. 12b). Fig. 12 View largeDownload slide Crystal size distribution diagrams for chromite in massive chromitite seams and silicate rocks. (a) Khuseleka; (b) Nkwe. See text for discussion. Fig. 12 View largeDownload slide Crystal size distribution diagrams for chromite in massive chromitite seams and silicate rocks. (a) Khuseleka; (b) Nkwe. See text for discussion. For reference, the ideal CSD in magma due to nucleation and growth would be a straight line with a negative slope intersecting the vertical axis at a point corresponding to the nucleation density (Marsh, 1988). For UG2 chromite, the CSD curves show the maximum population density at about 0·1–0·15 mm and a relative deficiency of smaller crystals. The Khuseleka CSD results (Fig. 12a) show differences in the position and slope of curves from the wall-rocks vs the main UG2 seam. Both the footwall and hanging wall curves have a deficiency of grain sizes less than 0·1 mm, but the CSD curve for larger grains is linear as in the ideal kinetic distribution mentioned above. By contrast, the CSD curve for UG2 has a non-linear slope with an upwards-concave shape, the curvature increasing in the grain size range above 0·6 mm. The Nkwe profile lacks data for the wall-rocks because of the paucity of chromite. The CSD curve for chromites of type 2 (poikilitic) is very similar to that of the Khuseleka chromite (also dominantly type 2), whereas the curve for chromite aggregates from Nkwe is quite different, reflecting the much larger proportion of coarse-grained chromite in the aggregates. Notably, all CSD curves from the UG2 main seam are not linear, but show an upward-concave shape which suggests textural annealing as discussed in a separate section below. DISCUSSION In the following discussion we address some of the current questions concerning the origin and evolution of the UG2 chromitite, especially the presence or absence of multiple magma injections and the ability of the trapped liquid shift effect to account for the chemical composition and variability in the layer. First, it is worth pointing out that the two profiles presented here are typical for the UG2 layer in the Bushveld Complex, including such macroscopic features as one or more leaders above the main seam (Khuseleka), sharp contacts and abrupt changes in mineral chemistry between chromitite and adjacent silicate rocks (both profiles), coarse pegmatoidal rocks in the footwall with abundant amphibole and mica (both profiles), thin anorthosite layers in immediate contact with chromitite (Nkwe) and poikilitic texture in the main seam (both profiles). Other detailed studies of UG2 reported similar findings (Hiemstra, 1985, 1986; Cawthorn & Barry, 1992; Mathez & Mey, 2005; Mondal & Mathez, 2007; Maier & Barnes, 2008; Voordouw & Beukes 2009; Voordouw et al., 2009; Junge et al. 2014; Mungall et al., 2016). There are differences between the Khuseleka and Nkwe profiles that have been described above. These include the lack of an upper leader and the presence of a thin hanging-wall anorthosite at Nkwe. Low-temperature interstitial minerals such as clinopyroxene, amphibole and mica are more abundant at Nkwe, the maximum Mg# of chromite is lower (but Cr# is the same) and the concentrations of incompatible trace elements in orthopyroxene and plagioclase are more variable than in Khuseleka samples (Figs 8 and 10). All of these features suggest that the UG2 layer at Nkwe underwent more chemical re-equilibration than at Khuseleka and the chromite CSD results also support a greater degree of textural evolution. The UG2 as a composite layer Cawthorn (2011) proposed, on the basis of observations by Lee (1996) and the pattern of PGE distribution, that the UG2 chromitite appeared to be a composite layer comprizing at least 3 parts. Junge et al. (2014) described rhythmic variations in Cr/(Cr + Al) ratios and TiO2 contents in a vertical profile of UG2 in the Karee mine on the western limb, and interpreted these as cryptic layering resulting from multiple episodes of magma replenishment and fractional crystallization during the formation of the seam. The eight sub-layers suggested by Junge et al. (2014) can possibly be questioned because although average values do vary, the range of individual samples between cycles overlap. Before our study, there were no comparable fine-scale microprobe profiles through the full UG2 layer to confirm these findings. Our results from the Khuseleka profile and from Nkwe on the eastern limb show no fine-scale rhythmic variations of the Mg# and Cr# (Figs 5a and b and 7a and b) so they are certainly not a universal feature. Indeed, the lower 25 cm of the Khuseleka profile shows remarkably stable chromite compositions (Fig. 5). However, we note greater compositional variations in chromite and poikilitic silicate crystals, well beyond analytical scatter, in the upper part of the main seam in both the Khuseleka and Nkwe profiles. These differences in mineral compositions between the upper and lower portions of the main seam and frequent presence of hanging wall leaders, e.g.at Khuseleka, appear to support the inference of Cawthorn (2011) that UG2 represents a composite layer. However, it is not clear that these variations can be explained by multiple episodes of chromite crystallization and accumulation, or different degrees of low-temperature re-equilibration. As discussed in the next section, we believe the evidence for the latter is very strong. The trapped liquid shift effect The concept of trapped liquid shift effect emerged from the fact that, as a rule, mineral compositions in crystallized plutonic rocks differ from the original compositions because of re-equilibration with intercumulus liquid or indeed in the subsolidus if cooling is slow enough. In a densely packed crystal mush where the volume proportion of crystals exceeds that of the liquid, movement of liquid relative to crystals practically stops and, with further cooling, the mush is expected to crystallise as a nearly closed system. Mineral compositions in the final product should depend on the kinetics of mineral-melt reaction and the modal proportions of the crystal and liquid phases. Since most UG2 minerals are solid solutions, slow exchange rates would produce, at a given cooling rate, compositionally zoned crystals that may retain original compositions in their cores, whereas fast exchange rates would result in homogeneous grains whose compositions differ from the original cumulus compositions (Cawthorn, 2015). The magnitude of the trapped liquid shift effect should increase with increasing mass proportion of the trapped liquid to the crystal phases. A quantitative model proposed by Barnes (1986a) emphasizes this relationship, and application of the model to the layered rocks of the Bushveld Complex showed that major and trace element variations in the main rock-forming silicates in various zones of the Complex could be explained by the trapped liquid shift (Cawthorn, 1996, 2015). With respect to UG2, the trapped liquid shift effect has been invoked for explaining why chromites in massive chromitite seams have a constant and generally less evolved composition than in the adjacent silicate rocks (Cameron, 1977). Our results support this general idea and many observations fit the simple model. However there are several features, especially in the trace element contents of major minerals, which are inconsistent with the closed-system trapped liquid scenario. It is important to recall that mineral–melt reactions in a multiply-saturated magma involve all phases in the assemblage. In a previous study of the Upper Critical Zone, Veksler et al. (2015) discussed chromite–melt reactions in detail and showed that the reactions actively involve plagioclase, orthopyroxene and olivine, when the latter is present. Naldrett et al. (2012) also recognized the importance of other crystal phases and pointed out the key role of plagioclase buffering the activity of alumina to produce their trends A and B in the chemical evolution of Bushveld chromite. Figures 5–7 show that the compositions of disseminated chromite in silicate rocks are shifted from those in the main seam and in the leader. Moreover, the compositions are shifted to a different extent and in different directions, with no apparent relationship to the modal mineralogy of the layers. Thus, the presence or absence of plagioclase and the chromite: melt mass ratio may be important factors (Naldrett et al., 2012), but they cannot be the only ones controlling chromite re-equilibration. The observation of different chromite trends in different lithologies is important because it means that the interstitial liquid in those parts of the crystal mush may also have differed in composition, which implies development of chemical gradients. Chemical gradients are driving forces for diffusion and kinetic effects related to differences in diffusion rates for different components in different phases may come into play in their presence. Veksler et al. (2015) discussed connections between redox gradients and alkali diffusion in the interstitial liquid between modally-contrasting parts of the crystal mush. The results of the present study confirm an anomalous mobility of the alkali elements as evidenced, for example, by a strong depletion of K in plagioclase from the chromitite relative to the wall-rocks (Figs 4c, 7c and 10). Chemical gradients and differential mobility of components during evolution of a layered crystal mush can lead to the situation where a mineral dissolves in one layer but continues to crystallize in another, thus sharpening the boundary between them (see Veksler et al. 2015). If dissolution keeps pace with crystallization in adjacent layers, the overall process may be nearly neutral in terms of thermal energy. These processes of migration of chemical components in the crystal mush and across its boundaries violate the closed-system assumption of the conventional trapped liquid shift model proposed by Barnes (1986a). In real systems, modal layering is not a fixed starting point, but an evolving and growing feature, partly shaped by post-cumulus mineral-melt reactions. The behaviour of trace elements The trapped liquid shift effect has predictable effects on the incompatible element contents of crystals in the evolving mush and these are examined in the REE plots of Figs 8 and 10. The point of reference for this discussion is the composition of the parental melt from which the Upper Critical Zone crystallized. In this discussion we use the composition of the B1 magma from Barnes et al. (2010) and assume that this is appropriate for both the Nkwe (eastern limb) and Khuseleka (western limb) localities. There is some debate about the parental magma for UG2, but many workers consider B1 to be a good approximation for the initial Critical Zone magma (e.g.see Cawthorn, 2007). The results of this study show that the REE distribution between orthopyroxene, clinopyroxene and plagioclase deviates from that predicted by experimentally determined mineral-melt distribution coefficients (D) from Bédard (2006, 2007), Sun & Liang (2012) and Yao et al. (2012), and the B1 parental melt composition. In general agreement with the trapped liquid shift effect, interstitial plagioclase in pyroxenite, harzburgite and chromitite is enriched in moderately incompatible REE from La to Nd, but contrary to expectations, it is not enriched or, locally, even depleted in the strongly incompatible element Y (Fig. 10). Orthopyroxene also has higher concentrations of light REE from La to Pr than expected for crystallization from B1 parental magma (Fig. 8), in qualitative agreement with the trapped liquid shift. However, there is a discrepancy between the light REE contents of orthopyroxene and coexisting clinopyroxene (Fig. 9). Orthopyroxene and plagioclase should both be equally affected by re-equilibration with residual liquid so the discrepancy in REE distribution between them might be due to changes of the plagioclase-melt and orthopyroxene-melt DREE values with temperature and, or, melt composition. Chutas et al. (2012) reported a ‘disequilibrium’ distribution of Sr and Pb between plagioclase and orthopyroxene throughout the Lower and Critical Zones, and it is possible that the ‘odd’ REE distribution between plagioclase and pyroxenes resulted from the same processes. Notably, Sr is compatible with plagioclase, but strongly incompatible with orthopyroxene, and, therefore, the trapped liquid shift should have a much greater effect on Sr concentration and isotope composition in the latter mineral. Permeability of the UG2 layer The permeability of the UG2 crystal mush and its changes during late-stage crystallization is obviously important for assessing the possibility of material transport inside and out of the layers. Mondal & Mathez (2007) postulated a low permeability of the UG2 chromitite and proposed that upwards-migrating volatile-rich melt was trapped below UG2 to form the footwall pegmatoid. Conventional models of the trapped liquid shift effect also consider the chromitite and adjacent layers as closed, effectively separate systems, which implies a very low permeability (Barnes, 1986a; Cawthorn, 1996). Manoochehri et al. (2015) used whole-rock concentrations of strongly incompatible elements and micro-X-ray tomography to evaluate the paleo-porosity and extent of compaction in a 20 m section of the Upper Critical Zone including the UG2 and UG3 chromitites. Their results indicate a rather complex picture. The ‘final porosity’ (trapped melt fraction) in UG2 and UG3 chromitite based on the average of six incompatible elements was 0.12 – 0.36, generally about the same or slightly higher than the final porosity of typical pyroxenite. However, the authors pointed out that the permeability of chromitite was possibly lower, but hard to determine precisely because it strongly depends on the crystallization of poikilitic orthopyroxene and plagioclase, which could block permeability and hamper compaction. Our synchrotron XRF results for the full UG2 profile at Khuseleka (Fig. 11) show a discrete clustering of the incompatible elements Zr, Y Rb, Nb and Mo near the bottom and, especially, the top of the seam. Manoocherhi et al. (2015) noted a clustering of accessory apatite in chromitite at a centimetre scale, which resulted in strong variations of P content from one sample to another. Phosphorous was not measured in our SR-XRF profile, but the clustering of other incompatible elements probably also reflects the presence of particular accessory minerals (e.g.zircon, mica, rutile) in locations where the final stage of interstitial crystallization took place. Note that the Nb and Mo concentration peaks, which are interpreted to reflect accumulations of rutile, are in different parts of the profile from the Zr, Y and Rb peaks, which are likely to mark zircon and mica accumulations. This implies that material transport was still active within the UG2 layer at the time when the accessory minerals crystallized. In the above context, it may be significant that the two-pyroxene REE thermometry in UG2 from Nkwe records a lower equilibration temperature within UG2 compared to the adjacent silicate layers (Fig. 9), implying that interstitial crystallization and equilibration proceeded longer in the chromitite layers than in the wall-rocks. The few samples analysed makes this a preliminary finding and it would be important for future studies to check if the low closure temperature is a general feature of the Bushveld chromitite layers. The mechanisms of textural equilibration In theory, the primary size distribution of chromite crystals due to the nucleation and growth kinetics could have been modified by mechanical processes (e.g.compaction and crystal sorting by gravity) and processes of textural equilibration. The latter are driven by the topological requirements of space-filling and by the need to minimise the total surface energy at grain boundaries. Because the surface energy at chromite–silicate interfaces is higher than at chromite–chromite and silicate–silicate interfaces, the formation of monomineralic crystal aggregates is beneficial. In the context of chromite textural evolution one should keep in mind that the very low contents of Cr in silicate melts [0·05 wt % Cr2O3 in the B1 magma at the time of Upper Critical Zone formation, according to Barnes (1986b)] effectively rules out net addition by new in situ nucleation and growth, so that the total mass of chromite in the rock during textural equilibration should stay practically constant. According to theoretical models, the crystal size distribution due to nucleation and growth is log-linear and such a distribution would be represented on Fig. 12 by a straight line with a negative slope (Marsh, 1988). The Y-intercept corresponds to the nucleation density and the slope is defined by the reciprocal product of the growth rate and the growth time, so that the greater the product, the shallower the slope. The UG2 CSD curves are clearly not straight and imply significant modification of the primary distribution. All curves from both profiles show a depletion of the smallest crystals in the chromite population relative to the ideal distribution. Pressure–solution, compaction and accumulation by gravity could explain the lack of smaller crystals but those processes operating alone would produce convex-upward curvature in the size distribution of larger crystals (Higgins, 2015), whereas we observe concave upward curves in both profiles. Also, in the case of accumulation and crystal sorting by gravity, we would expect to see differences in the crystal size distribution from the top to the bottom of the UG2 layer, but no significant changes were observed. The concave-upward curvature of the CSD curves, which is especially well developed in chromite forming aggregates at Nkwe (Fig. 12b), implies that the growth rate, growth time, or both increased with the grain size. This, together with the deficit of small crystals, implies textural adjustment by crystal coarsening, whereby larger crystals grow at the expense of smaller ones. Apart from grain coarsening, textural equilibration would also minimise, where possible, the area of high-energy interfaces between chromite and silicates. This will be ineffective in poikilitic grains, but should enhance the formation and growth of chromite aggregates. The aggregates of intergrown chromite crystals (Fig. 3f), thus appear to represent an advanced stage of textural equilibration. Their large size is in agreement with the theory of so-called ‘normal’ grain growth that applies to single phase polycrystalline systems, whereas ‘abnormal’ growth refers to large grains enclosing a matrix of much smaller crystals, the situation of pokilitic chromitite texture (Fig. 3d). Theories of normal and abnormal grain growth have been actively developed in material science and especially in metallurgy where annealing is important and carefully controlled (e.g. Harker & Parker, 1945; Smith, 1964; Atkinson, 1988). The characteristic features of crystal aggregates shaped by normal growth are curved grain boundaries intersecting in triple junctions with constant dihedral angles of 120°. The uniform dihedral angles arise from the requirement of minimal surface energy, whereas the curvature of crystal faces is a consequence of topological restrictions and the fact that there is no regular polyhedron with plane faces satisfying the optimal dihedral angles in 3D. Normal crystal growth in the aggregates proceeds by the movement of curved grain boundaries in the direction towards the centre of the curvature. Thus, crystals with concave faces expand, while neighbours with convex faces shrink and eventually disappear. Therefore, theory implies three different growth rates for chromite crystals in a chromitite layer when interstitial melt is still present. The growth rate is practically zero for chromite crystals in silicate oikocrysts. Crystals that are still in contact with the melt may grow at a rate limited by material transport in the melt phase and/or the kinetics of dissolution-precipitation at crystal-melt interfaces. In view of the very low amount of Cr in the melt, dissolution of smaller chromite crystals will be required for significant coarsening of the crystal population because the total chromite amount in the layer will stay constant. The third and fastest growth rate corresponds to the merger of crystals into one by migration of chromite–chromite grain boundaries (normal grain growth). The concave upward slopes of CSD curves for the poikilitic chromite inclusions may thus be attributed to the variable growth times depending on their position in the host crystals, with the longer growth time for the larger crystals at the rims of the oikocrysts than in their centres (Fig. 3d). As expected, the highest growth rate is for the intergrown chromite aggregates (Fig. 12b). The concave CSD curvature in the aggregates means that the product of growth rate and growth time increases with crystal size. Since the growth rate should increase as soon as two chromite crystals form a shared interface and undergo ‘normal growth’ mode, the concave curvature may indicate that largest crystals in the aggregates were in the mode of normal grain growth for a longer time. Comparative CSD studies of chromitite occurrences in or outside the Bushveld are sparse and, unfortunately, the calculation methodologies are not standardized, making direct comparison of CSD curves difficult. The closest comparison is the study by Vukmanovic et al. (2013) of thin chromitite stringers at the upper and lower contacts of the Merensky Reef. They found two crystal populations in the lower stringer and a single one in the upper chromitite. Their analysis of the CSD curves used a different input parameter (equivalent circle diameter vs the Feret diameter), which may affect the absolute values, but not the shapes of the CSD curves. The CSD curve for the lower Merensky stringer, is complex because of the two populations inferred by Vukmanovic et al. (2013), but the CSD curve for the upper stringer is concave-upwards much like our results. O’Driscoll et al. (2010) performed CSD analysis on chromite in the Rum layered intrusion in Scotland using the same measurement, conversion and calculation methods as in our study. In comparison with Bushveld, the population density maxima for Rum are at finer grain sizes, between 0·013 and 0·05 mm, and the right-hand slopes are much steeper. The latter are probably due to shorter time available for textural coarsening in Rum than in Bushveld. Jackson (1961) and Waters & Boudreau (1996) measured chromite size distributions in the Stillwater intrusion (Montana, USA), but their methods differ from those used in this study and cannot be directly compared. Qualitatively, however, the size distribution of chromite from Stillwater published by Waters & Boudreau (1996) looks similar to our results, i.e.coarser and more texturally evolved than in the case of Rum. To conclude, the processes of textural equilibration seem to account for the characteristic features of all observed CSD curves within the chromitite layer. The implication is that even if crystal settling and gravity sorting played a role in UG2 formation, as suggested, for example, by Kinnaird et al. (2002), or Maier et al. (2013), the textural evidence for these primary processes was overprinted in both profiles by subsequent recrystallization and crystal coarsening. Overall, the textural equilibration is more advanced at Nkwe than at Khuseleka, and this appears to correlate with the greater abundance at Nkwe of low-temperature interstitial minerals (e.g.clinopyroxene, amphibole, biotite) and with the slightly more evolved compositions of the major phases (e.g.the lower maximal Mg# of chromite). CONCLUSIONS There is an ongoing debate about how the chromitite layers in the Upper Critical Zone of the Bushveld complex formed, but no-one would argue that the mineral compositions and textures within the layered rocks are primary magmatic. Major adjustments took place by a combination of compaction, textural annealing and the reaction of early-formed crystals with the interstitial, residual melt. To better understand these late-stage processes, the c.1 metre-thick UG2 chromitite layer and its immediately surrounding wall-rocks were studied with millimetre-scale resolution in two continuous profiles from drill core samples: one is from the western limb at Khuseleka mine and the other is from the Nkwe mine on the eastern limb. In both cores, the major and minor element composition of chromite and the main silicates were determined by microprobe across the entire profile and chromite textures were analysed in terms of their crystal size distribution (CSD) in the same contiguous sets of thin sections. In addition, a new application of synchrotron XRF scanning was used to provide continuous trace element profiles of the Khuseleka core. Chromite is an accessory mineral in the silicate rocks above and below UG2, forming interstitial crystals with a considerable range of Mg#, Cr# and TiO2 contents, which contrasts with the rather constant compositions in the chromitite layer. Most of the chromite in the UG2 seam is included in anhedral, centimetre-sized orthopyroxene and plagioclase oikocryts. Chromite also occurs as tightly-intergrown aggregates with curved grain boundaries and 120° triple junctions. The grain size in the aggregates is larger than in the oikocrysts and we attribute this to late-stage annealing, which is supported by the shape of the CSD curves for the different textural types. The annealing texture is more pronounced in the Nkwe samples than at Khuseleka, but the chemical composition of the two chromite types are the same. Vertical microprobe profiles show nearly constant Mg# and Cr# in the lower half of the layer and more variable compositions in the upper part. The mineral compositions in the upper part of these profiles are variable but erratic, and we attribute the heterogeneity to local re-equilibration effects rather than to multiple charges of magma or chromite slurries. The trace element data for orthopyroxene, clinopyroxene and plagioclase from within the UG2 layer differ from those in the adjacent silicate layers. The latter are generally consistent with an origin from a B1-like parental magma, but the compositions within the chromitite layer have higher incompatible element contents. This discrepancy can be explained by equilibration with a more evolved, interstitial melt (the trapped liquid shift effect), but the shift is not constant for all incompatible elements and the premise of a closed-system evolution of the crystal mush is not entirely valid. Anomalously low concentrations of K, in particular, suggest selective migration of alkali elements out of the chromite layer in response to chemical potential gradients across the boundary with adjacent layers (see Veksler et al., 2015). An important and hitherto under-used application of trace element data from Bushveld rocks is the two-pyroxene-melt REE geothermometer (Liang et al., 2013), which yielded equilibration temperatures of 1170–1290°C for the silicate layers in the Nkwe profile, but only 890°C for the UG2 itself. This suggests that REE equilibration in the chromitite continued to a much lower temperature and possibly into the sub-solidus stage. Our synchrotron radiation XRF results prove the value of this method for non-destructive and high-resolution profiles of trace element distribution in drill cores. Correlated peaks in the concentration of incompatible elements Zr, Y, Rb, Nb and Mo occur near the bottom and the top of the UG2 layer. We attribute these to local concentrations of accessory zircon, mica and rutile, which may indicate the distribution of the last pockets of residual melt in the chromite crystal mush. The distribution of ore elements Ni, Cu, Pd, Ru and As is also imaged by this method. The concentration of Ni, Cu and Pd in a single zone near the top of UG2 is explained by hosting of these elements in pentlandite, whereas Ru shows a more diffuse distribution that reflects its distribution in primary Ru sulphide inclusions in chromite. Acknowledgements We are grateful to Dieter Rhede, Oona Appelt (GFZ Potsdam) and Helene Braetz (University of Erlangen) for their help with electron microprobe and laser ablation analyses. Tawanda Manyeruke and others at Nkwe Platinum are thanked for help and access to sampling the drill core, and permission to publish the results. Thoughtful and detailed reviews by Marian Holness, Michael Higgins, Ben Hayes and Jim Mungall led to a major improvement of the original version of the paper. FUNDING Field work and electron microprobe analyses were funded by the German Science Foundation (DFG), grant VE 619/2–1. Whole-rock analyses, including SR XRF, and studies of rock textures were funded by the Russian Science Foundation (RSF) grant No. 14–17-00200 and the AMREP project funded by German Ministry of Education and Research (BMBF). REFERENCES Atkinson H. V. ( 1988 ). Overview no. 65: theories of normal grain growth in pure single phase systems . Acta Metallurgica 36 , 469 – 491 . Google Scholar Crossref Search ADS Barnes S. J. ( 1986a ). The effect of trapped liquid crystallization on cumulus mineral compositions in layered intrusions . Contributions to Mineralogy and Petrology 93 , 524 – 531 . Google Scholar Crossref Search ADS Barnes S. J. ( 1986b ). The distribution of chromium among orthopyroxene, spinel and silicate liquid at atmospheric pressure . Geochimica et Cosmochimica Acta 50 , 1889 – 1909 . Google Scholar Crossref Search ADS Barnes S. J. , Roeder P. L. ( 2001 ). The range of spinel compositions in terrestrial mafic and ultramafic rocks . Journal of Petrology 42 , 2279 – 2302 . Google Scholar Crossref Search ADS Barnes S.-J. , Maier W. D. , Curl E. A. ( 2010 ). Composition of the marginal rocks and sills of the Rustenburg Layered Suite, Bushveld Complex, South Africa: implications for the formation of the platimum-group element deposits . Economic Geology 105 , 1491 – 1511 . Google Scholar Crossref Search ADS Bédard J. H. ( 2006 ). Trace element partitioning in plagioclase feldspar . Geochimica et Cosmochimica Acta 70 , 3717 – 3742 . Google Scholar Crossref Search ADS Bédard J. H. ( 2007 ). Trace element partitioning coefficients between silicate melts and orthopyroxene: parameterization of D variations . Chemical Geology 244 , 263 – 303 . Google Scholar Crossref Search ADS Brey G. P. , Köhler T. ( 1990 ). Geothermobarometry in four-phase lherzolites II. New thermobarometers, and practical assessment of existing thermobarometers . Journal of Petrology 31 , 1353 – 1378 . Google Scholar Crossref Search ADS Cameron E. N. ( 1969 ). Postcumulus changes in the Eastern Bushveld Complex . American Mineralogist 54 , 754 – 779 . Cameron E. N. ( 1975 ). Postcumulus and subsolidus equilibration of chromite and coexisting silicates in the Eastern Bushveld Complex . Geochimica et Cosmochimica Acta 39 , 1021 – 1033 . Google Scholar Crossref Search ADS Cameron E. N. ( 1977 ). Chromite in the central sector of Eastern Bushveld Complex, South Africa . American Mineralogist 62 , 1082 – 1087 . Cawthorn R. G. ( 1996 ). Re-evaluation of magma compositions and processes in the uppermost Critical Zone of the Bushveld Complex . Mineralogical Magazine 60 , 131 – 148 . Google Scholar Crossref Search ADS Cawthorn R. G. ( 2005 ). Stratiform platinum-group element deposits in layered intrusions. In: Mungall J. E. (ed.) Exploration for Platinum-Group Element Deposits. Mineralogical Association of Canada, Ottawa, Short Course Series Volume 35 , pp. 57 – 74 . Cawthorn R. G. ( 2007 ). Cr and Sr: keys to parental magmas and processes in the Bushveld Complex, South Africa . Lithos 95 , 381 – 398 . Google Scholar Crossref Search ADS Cawthorn R. G. ( 2011 ). Geological interpretations from the PGE distribution in the Bushveld Merensky and UG2 chromitite reefs . The Journal of the Southern African Institute of Mining and Metallurgy 111 , 67 – 79 . Cawthorn R. G. ( 2015 ). The Bushveld Complex, South Africa. In: Charlier B. , Namur O , Latypov R. , Tegner C. (eds) Layered Intrusions . Springer Geology , Dordrecht: Springer , pp. 517 – 588 . Cawthorn R. G. , Barry S. D. ( 1992 ). The role of intercumulus residua in the formation of pegmatoid associated with the UG2 chromitite, Bushveld Complex . Australian Journal of Earth Sciences 39 , 263 – 276 . Google Scholar Crossref Search ADS Cawthorn R. G. , Davies G. ( 1983 ). Experimental data at 3 kbars pressure on parental magma to the Bushveld Complex . Contributions to Mineralogy and Petrology 83 , 128 – 135 . Google Scholar Crossref Search ADS Chutas N. I. , Bates E. , Prevec S. A. , Coleman D. S. , Boudreau A. E. ( 2012 ). Sr and Pb isotopic deisequilibrium between coexisting plagioclase and orthopyroxene in the Bushveld Complex, South Africa: microdrilling and progressive leaching evidence for sub-liquidus contamination within a crystal mush . Contributions to Mineralogy and Petrology 163 , 653 – 668 . Google Scholar Crossref Search ADS Darin A. V. , Veksler I. V. , Rakshun Y. V. ( 2016 ). First results of the application of scanning XRF analysis with synchrotron-radiation beams from the VEPP-3 to study the spatial distribution of trace elements in samples of stratiform chromite ores . Journal of Surface Investigation. X-ray, Synchrotron and Neutron Techniques 10 , 88 – 91 . Google Scholar Crossref Search ADS Devi L. G. , Murthy B. N. ( 2008 ). Characterization of Mo doped TiO2 and its enhanced photo catalytic activity under visible light . Catalysis Letters 125 , 320 – 330 . Google Scholar Crossref Search ADS Eales H. V. , Cawthorn R. G. ( 1996 ). The Bushveld Complex. In: Cawthorn R. G. (ed.) Layered Intrusions . Amsterdam, New York : Elsevier , pp. 181 – 229 . Godel B. ( 2015 ). Platinum-group element deposits in layered intrusions: recent advances in the understanding of the ore forming processes. In: Charlier B. , Namur O. , Latypov R. , Tegner C. (eds) Layered Intrusions . Springer , Dordrecht , pp. 379 – 432 . Godel B. , Barnes S.-J. , Maier W. D. ( 2011 ). Parental magma composition inferred from trace element in cumulus and intercumulus silicate minerals: an example from the lower and lower critical zones of the Bushveld Complex, South Africa . Lithos 125 , 537 – 552 . Google Scholar Crossref Search ADS Harker D. , Parker E. R. ( 1945 ). Grain shape and grain growth . Transactions of the American Society of Metals 34 , 156 – 195 . Harris C. , Pronost J. J. M. , Ashwal L. D. , Cawthorn R. G. ( 2004 ). Oxygen and hydrogen isotope stratigraphy of the Rustenburg Layered Suite, Bushveld Complex: Constraints on crustal contamination . Journal of Petrology 46 , 579 – 601 . Google Scholar Crossref Search ADS Hiemstra S. A. ( 1985 ). The distribution of some platinum-group elements in the UG-2 chromitite layer of the Bushveld Complex . Economic Geology 80 , 944 – 957 . Google Scholar Crossref Search ADS Hiemstra S. A. ( 1986 ). The distribution of chalcophile and platinum-group elements in the UG-2 chromitite layer of the Bushveld Complex . Economic Geology 81 , 1080 – 1086 . Google Scholar Crossref Search ADS Higgins M. D. ( 2000 ). Measurement of crystal size distributions . American Mineralogist 85 , 1105 – 1116 . Google Scholar Crossref Search ADS Higgins M. D. ( 2015 ). Quantitative textural analysis of rocks in layered mafic intrusions. In: Charlier B. , Namur O , Latypov R. , Tegner C. (eds) Layered Intrusions . Springer Geology, Dordrecht : Springer , pp. 153 – 181 . Holness M. B. , Namur O. , Cawthorn R. G. ( 2013 ). Disequilibrium dihedral angles in layered intrusions: a microstructural record of fractionation . Journal of Petrology 54 , 2067 – 2093 . Google Scholar Crossref Search ADS Holness M. B. , Tegner C. , Nielsen T. F. D. , Stripp G. , Morse S. A. ( 2007 ). A textural record of solidification and cooling in the Skaergaard intrusion, East Greenland . Journal of Petrology 48 , 2359 – 2377 . Google Scholar Crossref Search ADS Hulbert L. J. , von Gruenewaldt G. ( 1985 ). Textural and compositional features of chromite in the lower and critical zones of the Bushveld Complex south of Potgietersrus . Economic Geology 80 , 872 – 895 . Google Scholar Crossref Search ADS Jackson E. D. ( 1961 ). Primary textures and mineral associations in the ultramafic zone of the Stillwater complex, Montana . United States Geological Survey Professional Paper 358 . Junge M. , Oberthur T. , Melcher F. ( 2014 ). Cryptic variation of chromite chemistry, platinum group element and platinum group mineral distribution in the UG-2 chromitite: an example from the Karee Mine, western Bushveld Complex, South Africa . Economic Geology 109 , 795 – 810 . Google Scholar Crossref Search ADS Kinnaird J. A. , Kruger F. J. , Nex P. A. M. , Cawthorn R. G. ( 2002 ). Chromitite formation—a key to understanding processes of platinum enrichment . Applied Earth Science: Transactions of the Institutions of Mining and Metallurgy: Section B 111 , 23 – 35 . Latypov R. , Chistyakova S. , Mukherjee R. ( 2017 ). A Novel Hypothesis for Origin of Massive Chromitites in the Bushveld Igneous Complex . Journal of Petrology 58 , 1899 – 1940 . Google Scholar Crossref Search ADS Lee C. A. ( 1996 ). A review of mineralization in the Bushveld Complex and some other layered mafic intrusions. In: Cawthorn R. G. (ed.) Layered Intrusions . Amsterdam : Elsevier , pp. 103 – 146 . Li C. , Ripley E. M. , Sarkar A. , Shin D. , Maier W. D. ( 2005 ). Origin of phlogopite-orthopyroxene inclusions in chromites from the Merensky Reef of the Bushveld Complex, South Africa . Contributions to Mineralogy and Petrology 150 , 119 – 130 . Google Scholar Crossref Search ADS Liang Y. , Sun C. , Yao L. ( 2013 ). A REE-in-two-pyroxene thermometer for mafic and ultramafic rocks . Geochimica et Cosmochimica Acta 102 , 246 – 260 . Google Scholar Crossref Search ADS Longerich H. P. , Jackson S. E. , Günther D. ( 1996 ). Laser ablation inductively coupled plasma-mass spectrometric transient signal data acquisition and analyte concentration calculation . Journal of Analytical Atomic Spectrometry 11 , 899 – 904 . Google Scholar Crossref Search ADS Maier W. D. , Barnes S.-J. ( 2008 ). Platinum-group elements in the UG1 and UG2 chromitites, and the Bastard reef, at Impala platimum mine, western Bushveld Complex, South Africa: evidence for late magmatic cumulate instability and reef constitution . South African Journal of Geology 111 , 159 – 176 . Google Scholar Crossref Search ADS Maier W. D. , Barnes S.-J. , Groves D. I. ( 2013 ). The Bushveld Complex, South Africa: formation of platinum-palladium, chrome- and vanadium-rich layers via hydrodynamic sorting of a mobilized cumulate slurry in a large, relatively slowly cooling, subsiding magma chamber . Mineralium Deposita 48 , 1 – 56 . Google Scholar Crossref Search ADS Maier W.D. , Prichard H.M. , Barnes S.J. , Fisher P.C. ( 1999 ). Compositional variation of laurite at Union Section in the Western Bushveld Complex . South African Journal of Geology 102 , 286 – 292 . Manoochehri S. , Schmidt M. W. , Britz W. ( 2015 ). Trapped liquid, paleo-porosity and formation time scale of a chromitite–(ortho) pyroxenite cumulate section, Bushveld, South Africa . Journal of Petrology 56 , 2195 – 2222 . Google Scholar Crossref Search ADS Marsh B. D. ( 1988 ). Crystal size distribution (CSD) in rocks and the kinetics and dynamics of crystallisation . Contributions to Mineralogy and Petrology 99 , 277 – 291 . Google Scholar Crossref Search ADS Mathez E. A. , Mey J. L. ( 2005 ). Character of the UG2 chromitite and host rocks and petrogenesis of its pegmatoidal footwall, northeastern Bushveld Complex . Economic Geology 100 , 1617 – 1630 . Google Scholar Crossref Search ADS Mondal S. K. , Mathez E. A. ( 2007 ). Origin of the UG2 chromitite layer, Bushveld Complex . Journal of Petrology 48 , 495 – 510 . Google Scholar Crossref Search ADS Mungall J. E. , Kamo S. L. , McQuade S. ( 2016 ). U–Pb geochronology documents out-of-sequence emplacement of ultramafic layers in the Bushveld Igneous Complex of South Africa . Nature Communications 7 , 13385 . Google Scholar Crossref Search ADS PubMed Naldrett A. J. , Wilson A. , Kinnaird J. , Yudovskaya M. , Chunnett G. ( 2012 ). The origin of chromitites and related PGE mineralization in the Bushveld Complex: new mineralogical and petrological constraints . Mineralium Deposita 47 , 209 – 232 . Google Scholar Crossref Search ADS Naldrett A. J. , Kinnaird J. , Wilson A. , Yudovskaya M. , McQuade S. , Chunnett G. , Stanley C. ( 2009 ). Chromite composition and PGE content of Bushveld chromitites: part 1—the lower and middle groups . Transactions of the Institute of Mineralogy and Metallurgy 118 , 131 – 161 . O’Driscoll B. , Emeleus C. H. , Donaldson C. H. , Daly J. S. ( 2010 ). Cr-spinel seam petrogenesis in the Rum layered suite, NW Scotland: assimilation and in situ crystallisation in a deforming crystal mush . Journal of Petrology 51 , 1171 – 1201 . Google Scholar Crossref Search ADS Penberthy C. J. , Merkle R. K. W. ( 1999 ). Lateral variations in the platinum-group element content and mineralogy of the UG2 chromitite . South African Journal of Geology 102 , 240 – 250 . Pouchou J. L. , Pichoir F. ( 1985 ). “PAP” (ϕ-ρ-Z) procedure for improved microanalysis. In: Armstrong J. T. (ed.) Microbeam Analysis . San Francisco : San Francisco Press , pp. 104 – 106 Scoates J. S. , Wall C. J. ( 2015 ). Geochronology of layered intusions. In: Charlier B. , Namur O , Latypov R. , Tegner C. (eds) Layered Intrusions . Springer Geology, Dordrecht : Springer , pp. 3 – 74 . Scoon R. N. , Teigler B. ( 1994 ). Platinum-Group element mineralization in the Critical Zone of the Western Bushveld Complex. I. Sulfide poor-chromitites below the UG-2 . Economic Geology 89 , 1094 – 1121 . Google Scholar Crossref Search ADS Smith C. S. ( 1964 ). Some elementary principles of polycrystalline microstructure . Metallurgical Reviews 9 , 1 – 48 . Google Scholar Crossref Search ADS South African Committee for Stratigraphy (SACS) ( 1980 ). Lithostratigraphy of South Africa, South West Africa/Namibia, and the Republics of Boputhatswana, Transkei, and Venda. In: Stratigraphy of Southern Africa, Handbook 8. Part 1. Geological Survey of South Africa, Government Printer, pp. 690. Spandler C. , Mavrogenes J. , Arculus R. ( 2005 ). Origin of chromitites in layered intrusions: Evidence from chromite- hosted melt inclusions from the Stillwater Complex . Geology 33 , 893 – 896 . Google Scholar Crossref Search ADS Sun C. , Liang Y. ( 2012 ). Distribution of REE between clinopyroxene and basaltic melt along a mantle adiabat: Effects of major element composition, water, and temperature . Contributions to Mineralogy and Petrology 163 , 807 – 823 . Google Scholar Crossref Search ADS Van Achterbergh E. , Ryan C. G. , Griffin W. L. ( 1999 ). GLITTER: On-line interactive data reduction for the laser ablation ICP-MS microprobe. Proceedings of the 9th V.M. Goldschmidt Conference. Cambridge, MA: Lunar and Planetary Institute, Houston. pp. 305. Van Achterbergh E. , Ryan C.G. , Jackson S.E. , Griffin W.L. ( 2001 ). Data reduction software for LA-ICP-MS. In: Sylvester, P. (ed.) Laser-Ablation-ICPMS in the Earth Sciences—Principles and Applications, Mineralogical Association of Canada (short course series), Ottawa, Vol. 29, pp. 239–243. Van Espen P. , Janssens K. , Nobels J. ( 1986 ). AXIL-PC: software for the analysis of complex X-ray spectra . Chemometrics and Intelligent Laboratory Systems 1 , 109 – 114 . Google Scholar Crossref Search ADS VanTongeren J. A. , Mathez E. A. ( 2013 ). Incoming magma composition and style of recharge below the Pyroxenite Marker, Eastern Bushveld Complex, South Africa . Journal of Petrology 54 , 1585 – 1605 . Google Scholar Crossref Search ADS Veksler I. V. , Reid D. L. , Dulski P. , Keiding J. K. , Schannor M. , Lutz H. , Trumbull R. B. ( 2015 ). Electrochemical processes in a crystal mush: cyclic units in the Upper Critical Zone of the Bushveld Complex, South Africa . Journal of Petrology 56 , 1229 – 1250 . Google Scholar Crossref Search ADS Voordouw R. J. , Beukes N. J. ( 2009 ). Alteration and metasomatism of the UG2 melanorite and its stratiform pegmatoids, Bushveld Complex, South Africa–characteristics, timing and origins . South African Journal of Geology 112 , 47 – 64 . Google Scholar Crossref Search ADS Voordouw R. , Gutzmer J. , Beukes N. J. ( 2009 ). Intrusive origin for Upper Group (UG1, UG2) stratiform chromitite seams in the Dwars River area, Bushveld Complex, South Africa . Mineralogy and Petrology 97 , 75 – 94 . Google Scholar Crossref Search ADS Voordouw R. , Gutzmer J. , Beukes N. J. ( 2010 ). Zoning of platinum group mineral assemblages in the UG2 chromitite determined through in situ SEM-EDS-based image analysis . Mineralium Deposita 45 , 147 – 159 . Google Scholar Crossref Search ADS Vukmanovic Z. , Barnes S. J. , Reddy S. M. , Godel B. , Fiorentini M. L. ( 2013 ). Morphology and microstructure of chromite crystals in chromitites from the Merensky Reef (Bushveld Complex, South Africa) . Contributions to Mineralogy and Petrology 165 , 1031 – 1050 . Google Scholar Crossref Search ADS Waters C. , Boudreau A. E. ( 1996 ). A re-evaluation of crystal size distribution in chromite cumulates . American Mineralogist 81 , 1452 – 1459 . Google Scholar Crossref Search ADS Wilson A. H. ( 2012 ). A chill sequence to the Bushveld Complex: insight into the first stage of magma emplacement and implications for parental magmas . Journal of Petrology 53 , 1123 – 1168 . Google Scholar Crossref Search ADS Yao L. , Sun C. , Liang Y. ( 2012 ). A parameterized model for REE partitioning between low-Ca pyroxene and basaltic melts with applications to adiabatic mantle melting and pyroxenite-derived melt and peridotite interaction . Contributions to Mineralogy and Petrology 164 , 261 – 280 . Google Scholar Crossref Search ADS Zeh A. , Ovtcharova M. , Wilson A. H. , Schaltegger U. ( 2015 ). The Bushveld Complex was emplaced and cooled in less than one million years–results of zirconology, and geotectonic implications . Earth Planetary Science Letters 418 , 103 – 114 . Google Scholar Crossref Search ADS © The Author(s) 2018. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

Journal of PetrologyOxford University Press

Published: Jun 1, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off