Palladosilicide, Pd2Si, a new mineral from the Kapalagulu Intrusion, Western Tanzania and the Bushveld Complex, South Africa

Abstract Palladosilicide, Pd2Si, is a new mineral (IMA 2014-080) discovered in chromite-rich samples from the Kapalagulu intrusion, western Tanzania (30°03′51′′E 5°53′16′′S and 30°05′37′′E 5°54′26′′S) and from the UG-2 chromitite, Bushveld complex, South Africa. A total of 13 grains of palladosilicide, ranging in size from 0.7 to 39.1 μm (equivalent circle diameters), were found. Synthetic Pd2Si is hexagonal, space group P62m, with a = 6.496(5), c = 3.433(4) Å, V = 125.5(1) Å3, c:a = 0.529 with Z = 3. The strongest lines calculated from the powder pattern (Anderko and Schubert, 1953) are [d in Å (I) (hkl)] 2.3658 100 (111); 2.1263 37 (120); 2.1808 34 (021); 3.240 20 (110); 1.8752 19 (030); 1.7265 12 (002); 1.3403 11 (122); 1.2089 10 (231). The calculated density for three analyses varies from 9.562 to 9.753 g cm–3. Palladosilicide is considered to be equivalent to synthetic Pd2Si based on results from electron backscattered diffraction analyses. Reflectance data in air for the four Commission on Ore Mineralogy wavelengths are [λ nm, R 1 (%) R 2 (%)] 470 49.6 52.7; 546 51.2 53.8; 589 51.6 53.7; 650 51.7 53.3 and the mineral is bright creamy white against chromite, weakly bireflectant and displays no discernible pleochroism or twinning. It is weakly anisotropic, has weak extinction and rotation tints in shades of blue and olive green. Electron probe microanalyses of palladosilicide yield a simplified formula of Pd2Si.

Introduction PALLADOSILICIDE (Pd 2 Si) is a new mineral, discovered recently in two different localities. One was the Platinum-Group Element (PGE)chromite horizon of the Kapalagulu Intrusion near the eastern shore of Lake Tanganyika, western Tanzania, with the mineral being discovered in diamond drill cores KPD 044 (30º03'51''E 5º53'16''S; sample D396) and KPD 024 (30º05'37''E 5º54'26''S; sample 10369), and from the UG-2 chromitite, Bushveld Complex, RSA. In both cases the mineral was discovered in heavy-mineral concentrates collected for detailed quantitative mineralogical investigations associated with mineral processing and optimization of PGE recovery; details regarding these concentrates were reported by Cabri (2004) for the Tanzanian samples and from a flotation tailing for the UG-2 chromitite by Cabri et al. (2008).
Disseminated copper and nickel sulfide in harzburgite near the base of the Kapalagulu Intrusion has been known for over a century. The Kapalagulu Intrusion lies near the eastern shore of Lake Tanganyika in Western Tanzania and forms part of a series of mafic intrusions that include Musongati and Kabanga, known as the Central African Nickel Belt (Fig. 1). In the late 1990s, Broken Hill Proprietary found Pt-Pd mineralization associated with Ni-bearing lateritic regolith. During 2002 and2003, Goldstream Mining, in partnership with Lonmin plc, identified sulfide-chromitite horizons in harzburgite below the lateritic regolith that contain PGE with grades of between 1 and 12 g/t PGE (Wilhelmij and Joseph, 2004). The PGE horizons are is located in the Lower Ultramafic Sequence of the Kapalagulu Intrusion that is preserved in a dyke-like extension to the Upper Mafic Sequence, known as the Lubalisi Zone (Fig. 2).
FIG. 1. Location of the Kapalagulu Intrusion with respect to other intrusions of the Central African nickel belt (after Maier et al., 2008).
The upper Critical Zone of the Bushveld Complex hosts the largest concentration of PGE in the world (Schouwstra et al., 2000). The UG-2 Reef is a PGE-bearing chromitite layer, usually 1 m thick but can vary from~0.4 to 2.5 m, developed some 20À400 m below the betterknown Merensky Reef (Schouwstra et al., 2000). A comprehensive review of the UG-2 may be found in Cawthorn (2002).

Mineral name and type material
The mineral is named after the two essential chemical components, palladium and silicon. The mineral and name were approved by the International Mineralogical Association Commission on New Minerals, Nomenclature and Classification on 3 December, 2014 (IMA 2014-080, Cabri et al., 2015. Part of the holotype has been deposited in the collections of the Canadian Museum of Nature, Gatineau, Québec, Canada, catalogue number CMNMC 86891.

Occurrence and associated minerals
Palladosilicide was found in samples from both localities in monolayer polished sections of highgrade gravity concentrates made with an earlymodel manual hydroseparator (HS-02) by one of us (NSR); see for example, Rudashevsky et al. (2004) and McDonald et al. (2015) who report a large number of platinum-group minerals (PGM) from the Coldwell complex (Ontario) found by using the newer computer-operated model hydroseparator (HS-11).
A total of 13 grains of palladosilicide were found, ranging in size from 0.7 to 39.1 mm (equivalent circle diameters, ECD), four of which, found in the Tanzanian samples, were imaged (Fig. 3). They came from two unoxidized samples, both 30 cm long: DDH KPD044 FIG. 2. Geological map outlining the Kapalagulu Intrusion and the Lower Ultramafic Sequence present within a dyke-like body known as the Lubalisi Zone. The two drill holes from which the harzburgite samples containing the platinum-group minerals were taken are shown.
(D-396, at a depth of between 169.60 and 169.90 m) and DDH KPD024 (010369, at a depth of between 53.0 and 53.3 m). The mineralogical study of pristine sulfide-bearing samples as well as oxidized samples from the regolith will form part of a future publication (by Wilhelmij and Cabri). Common heavy minerals in the samples include chromite, pentlandite, pyrrhotite/troilite; minor minerals were chalcopyrite and magnetite; and rare to trace minerals include gudmundite, arsenopyrite, zircon, galena and anglesite. Six of the palladosilicide grains were free; the others were either attached to chromite (n = 5) or locked in chromite (n = 2). The total frequency of precious-metal minerals found in the three samples ranged from 93 for sample D396 to 137 for sample 10369, which included some found during image analysis of unconcentrated sample splits. Table 1 illustrates a typical range of precious-metal minerals found in heavy-mineral concentrates with their relative abundances for sample 10369.
The concentrates of the UG-2 tailing sample are dominated by chromite with a few PGM comprising free laurite (ideally RuS 2 ) particles, a graphic intergrowth of tetraferroplatinum (ideally PtFe) in pyrrhotite, isoferroplatinum (ideally Pt 3 Fe) included in pentlandite, sobolevskite (ideally PdBi 2 ) attached to a larger particle of orthopyroxene, and a smaller (3.8 mm ECD) inclusion of kotulskite (ideally PdTe) included in an undefined Cu-Ag telluride (22.2 mm ECD), which is itself attached to a larger particle of chromite. A single free particle of palladosilicide measuring 39.1 mm in size (ECD) with an unidentified Pd-Sn mineral (taimyrite?) 6.3 mm in size (ECD) and undetermined very small bright metallic inclusions (maximum width~1 to 3 mm) were found after preparing additional concentrations.

Physical and optical properties
Palladosilicide appears bright creamy white in reflected light against the surrounding plastic, grains of chromite, and its inclusions of a Pd-(Cu)-Sn mineral (taimyrite?), unknown 1 (grey) and unknown 2 (bright white), as shown in Fig. 4.
The two unknown inclusions are not visible on the earlier scanning electron microscopy-backscattered electron (SEM-BSE) image ( Fig. 3) but appeared after repolishing, along with a larger area of the Pd-(Cu)-Sn mineral. Palladosilicide has an anhedral to subhedral habit, a metallic lustre, is weakly bireflectant and displays no discernible reflection pleochroism or twinning. It is weakly anisotropic, has a weak extinction and rotation tints in shades of light blue and olive green. Micro-indentation hardness could not be measured due to small grain size. Tenacity and streak were not determined, and cleavage, partings and fracture were not observed in reflected light.
Reflectance spectra were measured relative to a WTiC standard (Zeiss 314) on the grain from UG-2 in air and oil following the methodology of Stanley et al. (2002) and are shown in Fig. 5; the colour values are listed in Table 2 and reflectance data in Table 3. No internal reflections were noted.

Chemical composition
Chemical analyses were made using energy dispersive spectroscopy (EDS) and a Camscan Microspec-4DV SEM with a Link AN-10000 detector. The operating conditions included an accelerating beam voltage of 30 kV, a beam current of 1À2 nA, a beam diameter of 1 mm and   Fig. 2), as well as on one grain from the UG-2 sample (Pd 1.61 Ni 0.22 Cu 0.08 Sn 0.05 Pt 0.03 ) S1.99 (Si 0.90 As 0.12 ) S1.02 (shown in Fig. 4). Although Ag, Se, Sb, Te, Au, Pb and Bi were sought in sample 10369 (grains 6 and 18) and sample D396 (grain 21), these elements were not detected. For the UG-2 grain, Sb was sought and not detected.
Electron-microprobe analyses (wavelength dispersive spectroscopy mode) were later carried out on two grains with a JEOL 8900L electron microprobe at CANMET-MMSL in Ottawa (Table 4). Operating conditions were, 20 kV accelerating current, 40 nA beam current, 1 mm spot size, raw data were corrected using a ZAF matrix correction. The following standards and analytical lines were used: SiKa, Si metal; CrKa, chromite; PdLa, Pd metal; AgLb, Au 60 Ag 40 ; SKa, pyrite; SeLa, Se metal; NiKa, NiSb; TeLa, Te metal; PKa, apatite; SbLa, NiSb; AsLa, FeAs 2 ; FeKa, pyrite; PtMa, Pt metal; SnLa, Sn metal; PbMa, galena; CuKa, chalcopyrite; RhLa, Rh metal. Counting times were 20 s peak and 10 s background on both sides, except for Pd and Pb which were measured for 50 s peak and 25 s FIG. 5. Reflectance data in air and oil for Pd 2 Si, measured on a grain from UG-2. background on both sides. The results, calculated on the basis of three atoms per formula unit, show that the totals for the two sites are close to ideal values with minor deviations, possibly indicating that some of the elements reside in both sites. In this regard, the presence of significant Si $ As substitution is noteworthy (Table 5).

Crystallography
The small grain size and the presence of inclusions precluded the analysis of palladosilicide by standard powder X-ray diffraction (XRD) methods. The crystallographic properties of the mineral were thus studied by electron backscattered diffraction (EBSD). Preliminary SEM-EDS analyses of the grain selected for EBSD analysis (Grain 6, 10369 45-1) confirm major Pd and Si, along with trace Ni, Pt, As and Sn. The grain studied was found to contain submillimeter inclusions, representing perhaps 5% of the total area, of an unidentified Sn-bearing Ag telluride.  First analysis. [2] Average of 8 analyses. After the first analysis the epoxy softened, the grain tilted and became covered with epoxy resulting in low totals.
LOD À Limits of detection; St. dev À standard deviation.

PALLADOSILICIDE, A NEW MINERAL FROM AFRICA
The EBSD analyses were carried out on grain 6 (10369 45-1) from Kapalagulu with a JEOL 6400 SEM equipped with an HKL EBSD system (HKL Technology Inc., Oxford Instruments Group), an accelerating voltage of 20 kV, a beam current of 2.4 nA and a sample-to-camera working distance of 130 mm. The system was calibrated using a crystal of synthetic Si. Care was taken to avoid collecting EBSD (Kikuchi) patterns from those areas proximal to the Ag-Te inclusion (Fig. 6). Frames were collected for 20 ms, with 64 frames per image, both being selected so as to mitigate degradation of the epoxy surrounding the grain, this being found to be quite unstable under the electron beam. Channel5 software (Oxford Instruments; Day and Trimby, 2004) was used to collect Kikuchi (EBS) patterns along with pattern matching and interpretation (Fig. 7). Note that the Kikuchi patterns from three subareas were found to be near identical, indicating that all three areas are nearly identical in crystallographic orientation and thus that the grain being studied is presumably close to being single.
From each of the Kikuchi patterns obtained (n = 3), seven strong bands were selected manually and used for matching purposes.  , 1969). The crystal structure of synthetic, metal-rich Pd 2 Si (Fe 2 P type) has been solved and refined to R = 9.2% (Nylund, 1966). It has Pd in the 3f and 3g positions, with Si in the 2c and 1b sites.

Synthesis
The earliest account of the synthesis of Pd 2 Si is by Buddery and Welsh (1951), which was subsequently determined to be isotypic with Fe 2 P (C22 type) by Anderko and Schubert (1953). These findings were confirmed by Grigorev et al. (1952). The Pd 2 Si phase is reported to melt congruently at 1330ºC, based on thermal analyses (Hansen and Anderko, 1958). The crystal structure of metal-rich Pd 2 Si (revised C22 type) was reported, refined, to be hexagonal, space group P62m, a = 6.496, c = 3.433 Å and that of Si-rich Pd 2 Si to have a superstructure [a = 15.05(5) c = 27.49(0) Å ] by Nylund (1966). However, Nylund's study was undertaken before electron microprobes became readily available. Phase equilibrium investigations usually require detailed metallographic studies together with XRD analyses in order to determine phase boundaries, tie lines and related information (e.g. Cabri, 1965). Note that Nylund reported that it was extremely difficult to achieve thermodynamic equilibrium, but this aspect was not discussed further.  7. (a) The EBSD pattern for palladosilicide grain 6 (sample 10369); (b) the EBSD pattern for palladosilicide, indexed using data calculated from the crystal structure of synthetic Pd 2 Si (Nylund, 1966).
The strongest lines are given in bold. Natural examples of metal silicides are rare, the first (suessite) having been reported from extraterrestrial objects (an olivine pigeonite achondrite, also known as 'ureilite'). Gupeiite and xifengite were found as cores of spheres 0.1À0.5 mm in diameter, in placers from the Yanshan area, Hebei Province, People's Republic of China (Yu, 1984). They were found as heterogeneous grains, with outer shells consisting of magnetite, wüstite and maghemite, an inner shell of kamacite and taenite, and the cores being either gupeiite or xifengite. The minerals present and the morphology of the spheres were thought to be extraterrestrial in origin. Rudashevskii et al. (2001) gives the first account of several PGE-bearing, Fe-and Cusilicides that were found in ferromanganese crusts that formed on the ocean floor in areas that were almost completely sediment-free. They ascribed formation of the base-and precious-metal silicides to be related to emanations of highly reducing fluids that accompanied basalt formation.
Iron silicides have also been characterized as new minerals from podiform ophiolitic chromitites at Luobusha, Tibet (Bai et al., 2006;Li et al. 2012a,b). Two of these minerals were first reported by Gevork'yan (1969) and Gevork'yan et al. (1969) in heavy-mineral concentrates from placers and drill-core samples in sandstones from the Ukraine and described as two new alloy minerals (FeSi and FeSi 2 ), but without approval by the then named Commission on New Minerals and Mineral Names. Owing to their chemical properties and mineral assemblage, the Luobusha Fe silicides were considered to be xenocrysts from the mantle, transported to shallow depths by a rising plume and then captured by the melts from which the Luobusha chromitites crystallized (Bai et al., 2000;Yang et al., 2014).
Silicides may also form due to very high temperatures (e.g. 1800ºC) such as in 'fulgurites', formed as a result of lightning strikes (e.g. Myers and Peck, 1925).

Genetic implications and discussion
In the case of palladosilicide, which appears to be the first report of a PGM silicide, there are a few clues regarding its origin based on mineral associations. A few of the palladosilicide grains had some chromite attached and two were included within chromite, strongly suggestive that the mineral crystallized in association with or within chromite. The chemistry and zoning of the chromitite has not been studied in this work. The effect of highly reducing fluids during formation of mantle-derived ultramafic rocks was discussed by Rudashevskiy (1983), Rudashevskii (1984) and Rudashevskiy and Yertseva (1987). The formation and evolution of different chromitite deposits was recently reviewed and discussed by Mungall (2014); this includes the questioning by Spandler et al. (2007) of the popular interpretation that anomalous melt inclusions represent samples of unmodified mantle melts. In light of the ease with which some chemical components can diffuse through chromite at magmatic temperatures, it is possible that chromite may serve as a semipermeable membrane allowing some constituents of the melt inclusions, but not others, to reach equilibrium with ambient conditions (James Mungall, pers. comm., 2014). As chromite itself contains both Fe 3+ and Fe 2+ , diffusive exchanges involving both of these species driven by counterfluxes of other cations might lead to extreme perturbations in f O 2 within melt inclusions. Some grains of palladosilicide were found associated with, and others included in, chromite, suggesting that the mineral is paragenetically earlier than chromite, probably forming under conditions of very low f O 2 . Similarly extremely reducing mineral assemblages including moissanite (SiC), FeÀSi and FeÀC phases have been reported from other chromitites (e.g. Bai et al., 2006;Yang et al., 2014).
We anticipate that this first report of a palladium silicide found in chromitite-rich facies of layered PGE-bearing intrusions will lead to further studies focused on understanding the chemical-physical and thermodynamic conditions directed to constraining the specific conditions under which such minerals crystallize. The presence of inclusions of a Ag telluride in one case and of a Pd-(Cu)-Sn mineral plus much smaller unidentified minerals in a second grain also call for a better understanding of phase equilibria in the PdÀSiÀAgÀTe, PdÀSiÀSn and PdÀSiÀAs systems, as well as re-investigation of the PdÀSi binary. results in 2005, when this investigation was first started. They acknowledge W.D. Maier for providing the location map used for Fig. 1, Martine Wilhelmij for drafting Fig. 2, Heinz-Juergen Bernhardt for help with German abbreviations, Jim Mungall for discussions pertaining to genetic aspects and Peter Williams, Chairman of the CNMNC and its members for helpful comments on the submitted data. They also acknowledge the helpful comments of the reviewers: Kari Kojonen, Heinz-Juergen Bernhardt and David Vaughan.