Chromite in komatiites: 3D morphologies with

Contrib Mineral Petrol DOI 10.1007/s00410-012-0804-y

ORIGINAL PAPER

Chromite in komatiites: 3D morphologies with implications for crystallization mechanisms Be´linda Godel • Stephen J. Barnes • Derya Gu¨rer • Peter Austin • Marco L. Fiorentini

Received: 16 February 2012 / Accepted: 27 August 2012 Ó Her Majesty the Queen in Right of Australia 2012

Abstract High-resolution X-ray computed tomography has been carried out on a suite of komatiite samples representing a range of volcanic facies, chromite contents and degrees of alteration and metamorphism, to reveal the wide range of sizes, shapes and degrees of clustering that chromite grains display as a function of cooling history. Dendrites are spectacularly skeletal chromite grains formed during very rapid crystallization of supercooled melt in spinifex zones close to flow tops. At slower cooling rates in the interiors of thick flows, chromite forms predominantly euhedral grains. Large clusters (up to a dozen of grains) are characteristic of liquidus chromite, whereas fine dustings of mostly individual *20-lm grains form by in situ crystallization from trapped intercumulus liquid. Chromite in coarse-grained olivine cumulates from komatiitic dunite bodies occurs in two forms: as clusters or chains of euhedral crystals, developing into ‘‘chicken-wire’’ texture

where chromite is present in supra-cotectic proportions; and as strongly dendritic, semi-poikilitic grains. These dendritic grains are likely to have formed by rapid crescumulate growth from magma that was close to its liquidus temperature but supersaturated with chromite. In some cases, this process seems to have been favoured by nucleation of chromite on the margins of sulphide liquid blebs. This texture is a good evidence for the predominantly cumulus origin of oikocrysts and in situ origin of heteradcumulate textures. Our 3D textural analysis confirms that the morphology of chromite crystals is a distinctive indicator of crystallization environment even in highly altered rocks. Keywords X-ray computed tomography Chromite Nucleation Dendrites

Introduction Communicated by C. Ballhaus. B. Godel (&) S. J. Barnes CSIRO Earth Sciences and Resource Engineering, Australian Resources Research Centre, 26 Dick Perry Avenue, Kensington, WA 6151, Australia e-mail: [email protected] D. Gu¨rer Steinmann Institut fu¨r Geologie, Mineralogie und Pala¨ontologie, Rheinische Friedrich-Wilhelms-Universita¨t Bonn, Poppelsdorfer Schloss, 53115 Bonn, Germany P. Austin CSIRO Process and Science Engineering, Australian Mineral Research Centre, Waterford 6152, WA, Australia M. L. Fiorentini The University of Western Australia, Centre for Exploration Targeting, Crawley, WA 6009, Australia

Interpretation of textural features of igneous rocks is central to the interpretation of crystallization processes in magmas, and is often controversial. A long-standing controversy surrounds the interpretation of cumulate rocks as to whether they form predominantly by mechanical deposition of transported crystals (Wager and Brown 1968; Irvine 1980; Morse 1986) or by in situ nucleation and crystal growth (Campbell 1968, 1978; McBirney and Hunter 1995; Eales and Costin 2012). Most studies on cumulate textures concern large, slowly cooled layered intrusions. Cumulate rocks formed from komatiite lavas are an important component of some komatiite flow fields and provide an alternative natural laboratory for the study of cumulate textures. This study is the first attempt to investigate these rocks and other distinctive komatiite lithologies through quantitative textural analysis in three dimensions.

123

Contrib Mineral Petrol

Conventional petrographic studies of crystal morphologies, shapes and sizes rely upon two-dimensional (2D) observations on rock surfaces or thin sections, and lack information on the third dimension. Quantitative 2D petrographic analysis using crystal-size distribution theory (Cashman and Marsh 1988; Higgins 2006; Marsh 1988, 1998; Vinet and Higgins 2010) and spatial analysis of grain clustering (Jerram and Cheadle 2000; Jerram et al. 1996, 2003) provide useful information, but are potentially considerably more valuable where textures can be quantified in three dimensions (3D). The recent development of highresolution X-ray computed tomography (HRXCT) allows the acquisition of 3D datasets of rock samples in situ and in a non-destructive manner at micrometre scale. Over the past few years, the application of HRXCT to igneous rocks has revealed textural features only recognizable in 3D images (Barnes et al. 2008; Godel et al. 2010, 2012; Jerram et al. 2010). Such features have potential to address some fundamental problems to do with the mechanism of nucleation and crystallization of igneous rocks. Chromite is a widespread mineral in mafic and ultramafic rocks including komatiites and is a sensitive indicator of magmatic environments (Barnes 1998; Irvine 1965, 1967; Page and Barnes 2009; Toriumi 1994; Zhou and Kerrich 1992). In many cases, chromite is the only igneous mineral that survives the extensive hydration, carbonation and metamorphism that affect almost all komatiites. As a result, chromite commonly retains delicate textural features in rocks where all original silicate minerals have been replaced. Consequently, the textural attributes of chromite are just as useful as their chemical compositions as recorders of otherwise cryptic processes. Chromite in komatiitic rocks displays a range of characteristic morphologies, which have been inferred so far from the analysis of 2D thin sections. Chromite in spinifextextured upper zones of komatiite flows displays characteristic dendritic structures associated with coarser spinifex olivine plates (Arndt et al. 1977; Zhou and Kerrich 1992). In the olivine cumulate B zones of spinifex textured flows, chromites typically form clusters of euhedral octahedral grains. In the coarse-grained olivine adcumulate and mesocumulate rocks associated with large-scale flow conduits (e.g. Mount Keith and Yakabindie, Fig. 1), chromite forms a variety of morphologies depending on the composition of associated olivine (Barnes 1998; Barnes and Hill 1995). In cumulates where the olivine is Mg poor (\Fo92), chromite tends to form clusters or chains of euhedral grains, whereas in cumulates whose olivine compositions range from Fo91.5 to Fo93, chromite forms characteristic interstitial lobate or poikilitic crystals (Barnes 1998; Barnes and Hill 1995). In this contribution, we present the first 3D study of chromite morphologies obtained by high-resolution X-ray computed tomography of komatiitic rocks selected from an

123

Fig. 1 Simplified geological map of the east Yilgarn showing various sample localities (modified after Barnes 2006)

extensive collection made during three decades of study of the komatiites in the Yilgarn Craton, Western Australia (Fig. 1). The samples were selected to represent the observed diversity of chromite occurrences, morphologies and textural relationships. Together with chromite chemistry, these new data on 3D morphologies, sizes and clustering characteristics of chromites provide crucial information on crystallization mechanisms and environments and in some cases have broader implications for the nature and crystallization mechanisms of cumulate-textured igneous rocks.


Samples

Methodology

The samples selected and analysed encompass spinifex A zones and fine-grained B zones from differentiated thin komatiite flow lobes and coarse-grained olivine meso- to adcumulate rocks from thick cumulate-dominated komatiite units from the Yilgarn craton (Fig. 1). Sample descriptions and locations are summarized in Table 1 and Fig. 2.

High-resolution X-ray computed tomography Cores were drilled out of the samples (Table 2) with core diameter optimized to obtain pixel size suitable to image precisely chromite morphology in the different types of samples. Each core was scanned using a Skyscan 1,172 desktop high-resolution scanner at CSIRO Process Science

Table 1 Sample localities and description Deposit

Location/ deposit

Lithology

Chromite morphology

Chromite abundance

Comments

References

Mount Keith

MKD5 orebody, Mount Keith Ultramafic unit

Olivine adcumulate containing up to 5 vol.% base-metal sulphides)

Lobate and poikilitic \0.1 crystals, interstitial to vol.% olivine and/or sulphides

World’s largest accumulation of magmatic sulphide associated with komatiites

1, 2, 3, 4, this study

Goliath

Yakabindie (25 km south of Mount Keith)

Coarse-grained olivine adcumulates

Lobate to poikilitic \0.1 crystals in the central vol.% part of the body

4, 5, this study

Euhedral crystals forming centimetre tick olivine– chromite-sulphide cumulates

Up to 50 vol.%

4, 5, this study

Six Mile

Yakabindie (25 km south of Mount Keith)

Lenticular olivine meso- to adcumulate (olivine composition Fo89-Fo93.6)

Euhedral and poikilitic crystals

\0.1 vol.%

4, 5, this study

Fly Bore

WildaraLeonora Belt

Thin differentiated spinifextextured flows

Dendritic

Up to 5 vol.%

Sullivans

WildaraLeonora Belt

Sheet-like compound flow overlying lenses of coarsegrained olivine mesocumulate (chromite-bearing)

Euhedral

\0.1 vol.%

Black Swan

Gindalbie Domain/ Kalgoorlie Terrane

Coarse-grained ‘‘hopper’’ or skeletal-textured olivine orthocumulate

Euhedral crystals or \0.1 skeletal crystals vol.% surrounding spherical base-metal sulphide blebs

Murphy Well

Eastern Goldfields

Olivine-rich peridotite overlain by a porphyritic peridotite

Octahedral crystals which can form clusters of several crystals

Agnew

Agnew– Wiluna Belt

Basal orthocumulate layer overlain by spinifex-textured horizons with 10 centimetres scale tick chromite-rich layers or pods

Euhedral crystals forming ‘‘chickenwire’’ texture

Chromite crystallizing at high cooling rates about 50 cm below the flow top

6, this study

Interpreted to form as a magma conduit within the flow field

7, 8, 9, this study

\1 vol.%

c.a. 50 m

9, this study

Up to 50 vol.%

Chromite texture similar to Kurrajong and interpreted as indicative of lava lake environments

10, 11, 12, this study

1 Rosengren et al. (2005), 2 Godel et al. (submitted), 3 Grguric (2003), 4 Grguric et al. (2006), 5 Barnes (2006), 6 Thebaud et al. (2012), 7 Barnes et al. (2009), 8 Dowling et al. (2004), Hill et al. (2004), 9 Lewis and Williams (1972), 10 Fiorentini et al. (2004), 11 Hill et al. (1995), 12 Barnes (1998)

123


123

Contrib Mineral Petrol b Fig. 2 Photomicrographs of selected samples showing 2D morphologies of chromite crystals from selected komatiite occurrences in Western Australia. a Lobated and poikilitic chromite associated with base-metal sulphides from Mount Keith. b Chromitite layer from Goliath with interstitial base-metal sulphides. c Poikilitic chromite containing silicate inclusions and associated with base-metal sulphides, Goliath. d Dentritic chromite grains from Fly Bore. e Euhredral to subhedral chromite crystals from Sullivan. f Dentritic chromite forming a shell around base-metal sulphide, Black Swan. g Aggregate of chromite from Agnew forming chicken-wire texture

and Engineering (Perth, Australia). The scanner was set up to 100-kV, 100-lA electron source and an Al–Cu filter. A projection of the sample was recorded at each step (i.e. 0.3 degree rotation of the sample) over 360 degrees. Image (slice) reconstruction was carried out using NRecon software. Beam hardening and ring artefacts were corrected and minimized during the reconstruction. The reconstruction was optimized to differentiate between sulphide, magnetite and chromite. Postprocessing, 3D rendering and quantification of images were carried out using AvizoFireÒ software. Chromite grains were thresholded, separated and analysed following methods similar to those used in Godel et al. (2010, 2012) by using 3D gradient maps and watershed-based algorithms. In some samples, automated segmentation of individual chromite crystals proved to be impossible. In these cases, the chromite grains were segmented manually using 3D region growing. Results of both automated and manual segmentations were verified using either optical or scanning electron microscopy on referenced polished sections cut through the sample. Statistics on crystal-size distribution and shape were then calculated and visualized using CSDToolbox software (Ricard et al. 2012) and in-house MATLAB codes. Whole-rock and mineral chemistry Major and selected (Cr, Co) trace elements of the Mount Keith, Goliath and Six Mile samples were analysed by X-ray

fluorescence (XRF) at Geoscience Laboratory (Ontario, Canada) and UltraTrace Laboratories, Perth. Details of UltraTrace procedures can be found in The´baud et al. (2012). Major and selected trace elements of chromite grains were analysed by wavelength-dispersive electron microprobe analysis using the Cameca SX50 electron microprobe at CSIRO Earth Science and Resource Engineering, Australian Resources Research Centre, Perth (Australia). The elements were analysed under the following conditions: accelerating voltage 15 kV and beam current 30 nA, counting times of 30 s on peak and 15 s on backgrounds and by using PAP matrix correction. Calibration was carried out using mineral standards: diopside for Mg and Ca; pyrope for Al and Si; rutile for Ti; V metal for V; chromite for Cr; rhodonite for Mn, magnetite for Fe; Ni metal for Ni and; gahnite for Zn. Major elements were standardized using JJ100 Chro A31 standards.

Whole-rock compositions Whole-rock Cr concentrations in komatiites have a distinctive relationship with MgO concentrations governed by three components: (1) liquid (having variable Cr concentration determined by chromite solubility considerations), (2) olivine (having Cr concentration typically close to 1,000 ppm) and (3) chromite. Different proportions of these components result in samples plotting in distinct fields on plots of Cr versus MgO concentrations (Fig. 3a). The sample from Sullivans falls within the field of olivine–chromite orthocumulates, having a chromite content somewhat less than the expected cotectic proportion. The chromite-rich Agnew and Goliath samples plot well above the olivine–chromite cotectic line and represent excess accumulation of chromite. All the remaining samples plot within the broadly linear field (Fig. 3a) labelled ‘‘liquid-olivine mixtures’’, implying that the proportion of cumulus chromite in these samples is very low. In this case, the Cr whole-rock budget is dominated by the liquid and olivine components. However, the highly adcumulate nature

Table 2 Characteristics of the samples analysed using high-resolution X-ray computed tomography Sample name

Deposit

AGNEW-1

Agnew

AGNEW-2

Agnew

BSD155-89-5

Black Swan

FBL7

Fly Bore

MKD153-669.5

Mount Keith

YEX178-325.2a

Goliath

MW3

Murphy’s Well

SUD3 YEX192-573.7

Sullivan Six Mile

a

Sample volume (lm3)

Chromite texture

Chromite (volume %)

8.320

3.19E ? 11

Chicken-wire

2.64

7.982

3.16E ? 11

Chicken-wire

6.43

8.996

1.35E ? 12

Shell around sulphide or euhedral

0.04

3.565

7.78E ? 10

Dentritic in spinifex A zone

4.37

Voxel size (lm)

11.886

2.33E ? 12

Poecilitic

0.07

7.982

6.98E ? 10

Euhedral, layer

2.54

2.547

4.92E ? 10

Euhedral (B zone)

0.10

2.547 11.886

2.05E ? 10 3.07E ? 12

Euhedral (B zone) Euhedral (B zone)

0.34 0.15

Volume of the part considered for the crystal-size analysis

123

Contrib Mineral Petrol b Fig. 3 Whole-rock and mineral chemistry of the samples studied. a Cr versus MgO concentration plot showing theoretical curves of temperature-dependant chromite saturation surface, olivine liquid mixing lines, chromite–olivine cotectic cumulates and olivine– chromite mixing (modified after Barnes 1998). b Cr versus Al2O3 concentration plot illustrating the amount of trapped liquid present in the different samples. c Plot of Cr/(Cr ? Al) versus Mg/(Mg ? Fe2?) showing the composition of chromite cores from the studied samples. The dashed and filled dot lines represent the 50th and 90th percentiles of chromite cores from dunitic sheets and channels from komatiite (greenschist to lower amphibolites facies) in Western Australia (Barnes and Roeder 2001 and CSIRO unpublished data)

of the Goliath, Perseverance and Mount Keith samples precludes the possibility that chromite in these rocks crystallized from a trapped liquid component (Fig. 3b).

Chromite chemistry Chromite compositions from most samples studied (Fig. 3c) fall within the typical range for Munro-type (Al-undepleted) komatiites at subamphibolite metamorphic grade (Barnes 1998; Barnes and Roeder 2001). Exceptions are some analyses from Perseverance tending towards a very high Cr number (Cr# defined as molar Cr/(Cr ? Al)) and some analyses from Goliath with anomalously low Cr number. The range of Mg number (Mg# defined as Mg/(Mg ? Fe2?) is very wide due to varying degrees of metamorphic and/or subsolidus Mg–Fe exchange between chromite and coexisting igneous and metamorphic silicate phases, as discussed in detail by Barnes (1998, 2000). The extent of the Fe–Mg interchange is controlled primarily by cooling rate and trapped liquid content in the weakly metamorphosed rocks and by metamorphic exchange under amphibolite facies metamorphic conditions. The highest Mg numbers are found in adcumulate rocks at the lowest metamorphic grades (e.g. Mount Keith). Trivalent ion proportions are relatively unaffected by postcumulus exchanges, with the exception of the highly metamorphosed samples from Perseverance, where extensive Cr–Al reequilibration has taken place between chromite and chlorite (Barnes 2000). Excluding Perseverance, there is a systematic contrast in chromium numbers (Fig. 3c) between the poikilitic and lobate chromites from the dunite localities (Mount Keith and Goliath), and the higher values are seen in euhedral chromites from the other localities. 3D chromite morphology Dendritic chromite in spinifex A zones Dentritic chromites have been described in spinifex lavas (Arndt et al. 2008 and references therein; Zhou and Kerrich

123


Fig. 4 Example of 3D morphology of dendritic chromite from Fly Bore (Western Australia). a Volume rendering of the Fly Bore sample showing variable dentritic chromite morphologies (green) and olivine (blue). b ‘‘Hedge-trimmer’’-like or ‘‘Christmas-tree’’-like dentritic chromite (red arrow). c Examples of linear (red arrow) and planar

(purple arrow) dentritic morphologies. d The size of dentritic chromites is represented by the maximum and minimum lengths of a 3D-bounding box around each chromite grain. The aspect ratio (AR) is defined by maximal length/minimal length

1992). In our Fly Bore sample, chromite represents 4.4 vol.% of the sample, forms over 60,000 different chromite grains and exhibits various 3D shapes, all of which are strongly dendritic (Fig. 4). HRXCT results highlight the variability and heterogeneity of chromite shapes at small scale (Fig. 4). In 3D, chromites occur under 4 main different topologies as: (a) ‘‘planar’’ snow-flake-like structures up to 1 mm across; (b) as elongated tubes which may form millimetre scale planar and parallel structures; (c) ‘‘hedge-trimmer’’ or ‘‘Christmas-tree’’-like structures; or (d) small (\50 lm across) rounded grains. In few cases, chromite forms kinked tubes with an angle varying from 100 to 120 degrees.

Murphy Well, where they exhibit various sizes and degrees of clustering (Fig. 6). At Murphy Well, chromite occurs as small (10 to 15 lm) euhedral octahedral crystals. These crystals are present either as isolated grains or as clusters of up to 100 crystals. The remaining chromite crystals are widely and randomly dispersed as individual octahedra with grain diameters of less than 15 lm, while the much smaller population of larger grains extends to about 50 lm. These larger crystals, although representing a limited number of grains (20 %), account for up to 90 % of the total chromite volume. The Sullivans sample contains a similar distribution of small isolated octahedra and a second mode of larger and mostly isolated single octahedra with a size mode at around 120 lm diameter (Fig. 6). Grains larger than 80 lm represent only 32 % of the total number of grains, but account for 96.5 % of the total volume of chromite in the sample.

Euhedral chromite in B-zone cumulates Euhedral to subhedral chromite crystals (Fig. 5) have been observed in samples from Sullivans, Black Swan and

123


towards lower abundance at the top of the layer (*2.6 vol.%, Agnew 1, Table 1). Chromite crystal morphologies and sizes (Figs. 7, 8) are similar between the two parts of the layer. The length of crystals ranges from *10 to 490 lm, and more than 80 % of crystals are smaller than 275 lm (Fig. 8). Chromite crystals have an octahedral shape with several ‘‘hollow’’ crystals observed in the bottom part of the layer (Fig. 7). In both layers, the chromite crystals form 3D complex clusters with the number of crystals per cluster varying between 10 and 1,275 (Fig. 8d). As with chromite abundance, cluster size varies depending on the position in the layer. At the bottom, 35 % of the clusters contain more than 40 crystals with a large cluster containing 1,275 grains. In contrast, at the top of the layer, more than 85 % of the clusters contain less than 40 chromite crystals (Fig. 8d). Chromite clusters in the Agnew sample have very delicate open internal structures with strongly aligned centres and with a strong preference for solid angles of around 120° between lines joining centres of adjoining crystals (Fig. 7b). At Goliath, chromite crystals locally form thin (few centimetres thick) chromitite layers associated with finely disseminated base-metal sulphides consisting mainly of pyrrhotite, pentlandite and chalcopyrite (Fig. 9). These layers are associated with the stratigraphically upper (western) portion of the complex, interpreted to have formed after a change from dynamic channel flow to static crystallization associated with stagnant or weakly convecting lava (Hill et al. 1995). Despite the high abundance of chromite in the sample, the degree of clustering of chromite grains is not particularly high and is much less than in the more chromite-rich Agnew samples. Lobate interstitial and poikilitic chromite in adcumulates Fig. 5 Example of 3D morphologies of euhedral chromite grains from Sullivan, Back Swan and Murphy Well (Western Australia). a Isolated euhedral chromite crystals and euhedral chromite clusters from Murphy’s Well. b, c Euhedral chromite from Black Swan. d Euhedral chromite crystals from Six Mile. e Euhedral chromite crystals from Sullivan

Chicken-wire textures in layered cumulates Chicken-wire networks of euhedral crystal of chromites moulding olivine crystals are usually found in chromite-rich rocks in lava lakes or sills (Barnes 1998) and strongly resemble the characteristic textures of disseminated chromite–olivine cumulates found in layered intrusions (e.g. Jackson 1961). At Agnew (Fig. 1), chromite network morphologies vary at the millimetre scale depending on their position within a distinctive chromite-rich layer a few centimetres thick. Chromite is more abundant (*6.4 vol.%, Agnew 2, Table 1) at the bottom of the layer and grades

123

Lobate and poikilitic chromites have been identified in most of the known komatiite dunite bodies in the Yilgarn Craton of Western Australia, including both unmineralized dunite bodies at Mount Clifford, Marshall Pool and Wildara, and host units of low-grade disseminated nickel sulphide mineralization at Mount Keith, Yakabindie and Honeymoon Well (Barnes 1998, 2006; Barnes and Hill 1995). Examples have also been found in the dunite bodies of the Forrestania Greenstone Belt (Perring et al. 1995), although they are obscured by extensive metamorphic replacement of chromite by chromian magnetite under mid-amphibolite conditions (Barnes 2000). Barnes (1998) demonstrated that the lobate to poikilitic chromite grains characteristically have low Cr/(Cr ? Al) and low Fe3? contents, relative to typical euhedral chromites from lithologies other than dunite. At Mount Keith and Goliath (Fig. 2), chromites in thin section display apparently interstitial morphologies with

Contrib Mineral Petrol Fig. 6 Sizes of euhedral chromites from selected localities. a Murphy Well, the sample was divided into two parts with euhedral crystals (top) and euhedral to clustered crystals (bottom), b Sullivan, c Six Mile and, d Black Swan

low dihedral angles against olivine, partially or completely enclosing olivine crystals and associated or not with basemetal sulphides. In 3D, the chromites occur as three different morphologies (Fig. 10): 1.

2.

3.

Dendritic plates of several millimetres in size (up to 10 mm) where 3D angles between the branches or plates vary between 55°–65° and 110°–125°; this consistent angle implies the plates are the [111] faces of the spinel lattice; As single-grain dendritic networks or moulds around olivine and base-metal sulphides where chromite may partially or entirely enclose the base-metal sulphides; or As ‘‘skeletal’’ tetrahedra up to 2 mm across which in planar view may correspond to the ‘‘skeletal triangle’’ described by Zhou and Kerrich (1992) in spinifex zone from the Belingwe komatiites.

Chromite shells around sulphides Dendritic or skeletal chromites have been observed at the margins of komatiite-hosted massive sulphide orebodies at Black Swan (Dowling et al. 2004) and in a number of Kambalda ore deposits (Frost and Grove 1989; Groves et al. 1974), but hitherto have only very rarely been found within olivine-rich cumulates. In the disseminated sulphide ore from Black Swan, chromite occurs in 2D as apparently dendritic chromite grains forming ‘‘shells’’ around subspherical sulphide globules sitting within segregation vesicles in coarse-grained mineralized olivine orthocumulates (Barnes, et al. 2009). Our 3D results show that chromites

within the disseminated sulphide blebs occur either as partial shells around base-metal sulphides (Fig. 11) or as single euhedral crystals (Fig. 6). In 3D, the skeletal chromite forms a tetrahedron-like ‘‘cup’’ around about half of the sulphide globule and is directly attached to an euhedral chromite (Fig. 11b, c), suggesting the two morphologies are closely related.

Discussion Chromite abundance and cumulus or postcumulus origin In komatiites, whole-rock Cr concentration is controlled by the relative abundance of chromite, olivine and silicate melt. The solubility of Cr in komatiitic melts is dependent on temperature and oxygen fugacity with Cr solubility decreasing with decreasing temperature and increasing with decreasing fO2 (Barnes and Roeder 2001; Murck and Campbell 1986). The details of the theoretical basis of the crystallization of chromite in komatiitic magma are presented in Barnes (1998), and expected fields for chromites of different origins are shown in Fig. 3a. As noted by Barnes and Fiorentini (2012), komatiitic cumulates commonly plot below the expected linear trend on Cr–MgO plots for cotectic precipitation of olivine and chromite, even when chromite is present as a cumulus phase, including most of the samples represented here. In that case, the bulk rock composition is almost indistinguishable from chromite-free olivine liquid mixing lines. The simplest explanation for this

123


Fig. 7 Example of 3D morphologies of chicken-wire textures in samples from Agnew. a Isosurface showing the extent of chickenwire texture (sample Agnew-1). b Skeletonization of a interconnected chromite network showing the orientation of grains within the cluster.

c, d Isosurface showing the extent of strongly interconnected chromite clusters. e Example of ‘‘hollow’’ crystal observed in Agnew 2 sample

is that most or all of the chromite in the rock is not a cumulus phase, but crystallized in situ from the trapped liquid component. Evidence for a trapped liquid chromite component is seen in the orthocumulate sample from Sullivans, where the majority of chromite grains (although only a small proportion of the total volume of chromite) occur as small isolated octahedra between cumulus olivines. The small proportion

of larger cumulus grains accounts for the elevated Cr concentration in the Sullivans sample. This bimodal population is best interpreted as representing a sparse population of mechanically accumulated large cumulus crystals, and the smaller as the result of crystallization of small isolated grains from stagnant trapped liquid during in situ solidification. Occurence as individual crystals, as opposed to

123


clusters of connected grains, is suggested as a diagnostic criterion for recognizing genuinely postcumulus, subliquidus crystals. The Murphy Well and Fly Bore samples are likely to represent close approximations to initially chromite-undersaturated liquid compositions. The Murphy Well sample contains predominantly isolated fine chromite octahedra similar to the Sullivans intercumulus component, but it lacks the extreme dendritic development observed in the Fly Bore spinifex sample. The Murphy Well flow shows a number of other anomalous features, which may derive from rapid crystallization following sudden degassing (Siebel et al., in review). The inferred very high MgO content, around 34%, of the Murphy Well lava (Arndt et al. 2008) makes it highly unlikely that the lava was chromite saturated on eruption, and we conclude that the isolated chromites are postcumulus in origin. Interpretation of the whole-rock chemistry in relation to the chromite morphology is a challenge for the samples from Mount Keith, Perseverance and Goliath where chromite is present as large dendrites. The bulk rock compositions of these samples plot close to the olivine liquid line. Based on the whole-rock MgO–Cr data, it can be concluded that the chromite is derived by in situ crystallization of trapped liquid, but this is impossible to reconcile with the very strongly adcumulate (strictly, heteradcumulate) character of the rock (Barnes and Hill 1995). The answer is probably that the chromite is indeed cumulus in origin, but the modal proportion of chromite in these rocks is actually extremely low (Table 1). Origin of dendritic chromite morphologies in adcumulates

Fig. 8 Size and shape of Agnew chromites. a Histogram showing the size of individual chromite grains within each sample from Agnew. b, c Chromite shape diagram based on the short/intermediate/long axes of ellipsoid fitted around each chromite grains. d Histogram showing the number of chromite crystals per cluster and the overall number of clusters within each sample

The dendritic morphologies observed in the adcumulate dunite bodies at Mount Keith and Goliath (Fig. 10) are an entirely unexpected observation and are completely unrecognizable in two dimensions, where these grains appear to be isolated interstitial lobes or in some cases poikilitic grains enclosing small olivine grains. The 3D images clearly reveal that these grains are in fact extensive dendrites, with enhanced growth of idiomorphic spinel crystal faces growing to enclose olivine crystals. Previous discussion of apparently interstitial to poikilitic chromite textures (Barnes and Hill 1995) pointed out the essentially heteradcumulate nature of the rocks. These rocks are bimineralic (leaving aside the sulphide component), and no crystallization products of trapped intercumulus liquid are present. The solubility of Cr in komatiite magmas is limited to about 3,000 ppm or less, so that the proportion of chromite in the rock could not be accounted for by crystallization of intercumulus trapped liquid. Chromite must therefore have crystallized in a freely advecting or convecting body of olivine–chromite-saturated magma.

123

Contrib Mineral Petrol Fig. 9 Chromitite from Goliath. a 3D distribution of chromite (grey) and sulphide (yellow) in the chromitite. b, c Example of disseminated euhedral crystals at the top of the chromitite

The observed poikilitic textures are interpreted as the result of in situ nucleation and growth, with high rates of growth of chromite relative to olivine causing grains to form casts around olivine crystals. This model of competing simultaneous growth is essentially the same as that proposed for heteradcumulate textures in the Skaergaard (McBirney and Noyes 1979), Jimberlana (Campbell 1968) and Windimurra (Mathison and Booth 1990) layered intrusions, but it manifests in much smaller cumulate bodies and involves a high degree of concentration of a minor melt component. Barnes and Hill (1995) further pointed out that the distinctive range of olivine compositions in rocks with this texture corresponds to the onset of chromite crystallization. This is analogous to a similar relationship commonly observed in layered intrusions, whereby a cumulus phase makes its first appearance as oikocrysts just below the first appearance of conventionally ‘‘cumulus’’ grains of that phase. The new observations of dendritic morphologies presented here bear out the conclusions of Barnes and Hill (1995) and strongly support a model of in situ competitive

123

growth (Fig. 12) between olivine and chromite. Sparse nucleation of chromite from supersaturated magma gives rise to spurts of rapid growth from sparse scattered nuclei, the characteristic combination of factors required to form dendrites. The melt is supersaturated with chromite but saturated with olivine, such that olivine crystals grow simultaneously into the space between the chromite dendrites, but at slower rates despite their much smaller degree of enrichment relative to the liquid component (Fig. 12). Critical to this argument is the observation that the plates that comprise the dendritic grains are developed at solid angles characteristic of the cubic crystal system. In other words, the morphology of the chromite grains is controlled by intrinsic crystal growth kinetics, not simply by the availability of space between pre-existing crystals, as it would be the case if the chromites were growing in available intercumulus space. This model implies that the dendritic chromites are equivalent to crescumulate harrisites in layered intrusions such as the Rum intrusion (Donaldson 1982; O’Driscoll et al. 2006).


Fig. 10 Example of 3D morphologies of poikilitic chromite from Mount Keith and Goliath deposits

Relationship between dendritic morphology and chromite chemistry Barnes (1998) pointed out that the lobate and poikilitic chromite in adcumulate dunite bodies characteristically have a distinctly lower Cr number and apparently a lower Fe3? content than more typical euhedral chromites from orthocumulates in B zones of thin flows. A variety of possible explanations were considered, including that the dunites may have crystallized under unusually reducing conditions. This interpretation was never supported by independent evidence and is invalidated by subsequent observations on whole-rock V–Ti trends that are incompatible with low oxygen fugacities (Barnes and Fiorentini 2012).

The new textural observation presented here provides a likely explanation. We would now interpret much if not all of the lobate to poikilitic chromite identified by Barnes (1998) as being dendritic. Dendritic growth is a disequilibrium process in which mineral chemistry is modified as a result of nutrient depletion in boundary layers around growing crystals. Given the very strong concentration factor of several hundred between komatiite magma and chromite, it is highly likely that dendritic grains would have lower Cr contents than they should have at equilibrium. The Fe3? deficiency cannot be explained in the same way, but may be an artefact of the calculation used to determine it. All of the published data on Fe3? in komatiitic chromites are derived by stoichiometry-based calculation, not direct measurement, and may be incorrectly

123


Fig. 12 Sketch illustrating the postulated mode of growth of dendritic chromite in olivine–chromite heteradcumulates in 2D. a Chromite dendrite nucleates and grows rapidly from komatiite magma flowing over bed of growing olivine crystals. The magma at this point is supersaturated in chromite, but exactly saturated with olivine. b Ongoing growth of both chromite and olivine. Fast-growing chromite overgrows and moulds around slowly growing equant olivine crystals

Fig. 11 Example of 3D morphology of chromite from Black Swan forming a shell around base-metal sulphide

estimated if the dendritic grains grew with a high proportion of point defects in their crystal lattices. The dendritic chromite: sulphide association Much (but not all) of the dendritic chromite reported in this study is present in samples that also contain disseminated sulphide blebs, and in the case of the Black Swan sample (Fig. 11), the chromite has apparently nucleated on a

123

sulphide bleb. This suggests that the degassing-oxidation mechanism proposed by Dowling et al. (2004) and by Fonseca et al. (2008) for chromite rinds on massive sulphide bodies may also operate at finer scale on disseminated blebs. The suggestion is that oxygen originally dissolved in the sulphide melt is driven back out into the silicate melt as the sulphide melt begins to crystallize and causes nucleation of chromite at a redox interface with the silicate magma. However, this mechanism is unlikely to work for the very magnesian adcumulates here, as these rocks must have formed close to the liquidus temperature of the komatiite magma, and this temperature would have


been several hundred degrees higher than the liquidus temperature of the sulphide melt. Some sulphide blebs in the Black Swan deposit contain what appear to be unusually high proportions of primary magnetite for disseminated ores (Barnes et al. 2009). This may have been inherited from the assimilation stage of ore genesis; the sulphidic contaminant that gave rise to sulphide liquid formation may itself have had a high oxide content, giving rise to anomalously oxygen-rich sulphide liquids that lost oxygen back to the silicate melt above their liquidus temperature. This could have resulted in nucleation of the chromite rinds. An alternative mechanism is that the highly reactive surface of the sulphide bleb served as a preferential nucleation site for chromite for purely kinetic reasons. However, poikilitic chromite, likely also to be of dendritic origin, is also commonly found in komatiitic olivine adcumulates in the absence of sulphide, so the relationship is apparently not critical.

between grain centres having strongly preferred distributions (Fig. 7b). Such structures would have to survive transport, settling and accumulation, and would have to accumulate in such a way as to mould uniformly around probably equally open olivine crystal frameworks. The likely consequence of gravitational accumulation of open clusters of chromite and olivine grains would result strongly orthocumulate textures, and heterogeneous random distributions of chromite-rich and olivine-rich domains or bands. The high degree of regularity of the ‘‘chicken-wire’’ texture, where a 3D chain of chromite crystals surrounds each individual olivine grain, is inconsistent with an essentially random accumulation process. In contrast, this texture suggests a crystallographic control and a degree of grain-scale self-organization consistent with in situ nucleation and growth.

Chicken-wire and clustered chromite formation

Chromite morphology is strongly controlled by the kinetics of crystal nucleation and growth and shows a close relationship with the environment of crystallization. A number of distinct textural-genetic associations have been identified. Dendrites as seen at Fly Bore are spectacularly skeletal grains formed during very rapid crystallization of supercooled melt in spinifex zones close to flow tops. At slower cooling rates in interiors of thick flows, as at Murphy Well and Sullivans, chromite forms predominantly euhedral grains: large clusters of up to a dozen grains are characteristic of liquidus chromite, while fine dustings of mostly individual *20-lm grains form by in situ crystallization of trapped intercumulus liquid between cumulus olivine grains. Chromite in coarse-grained olivine cumulates from komatiitic dunite bodies occurs in two forms: as clusters or chains of euhedral grains, developing into ‘‘chicken-wire’’ texture in unusually chromite-rich cumulates as at Agnew and Goliath; and as spectacularly dendritic and semipoikilitic grains. These dendritic grains could not have formed in intercumulus space, as the rocks they are found in are essentially adcumulates; instead, they are likely to have formed by rapid growth from chromite-supersaturated magma, which was close to its liquidus temperature with respect to olivine. High growth rates of chromite from sparse nuclei relative to those of neighbouring olivine produced dendritic poikilitic textures, where the abundance of chromite in the rock as a whole is very low. In some cases, this process seems to have been favoured by nucleation of chromite on the margins of sulphide liquid blebs. This texture is good evidence for the predominantly in situ origin of heteradcumulate textures and the crescumulate origin of oikocrysts originally proposed by Campbell (1968).

The chicken-wire textures observed in the Agnew sample, and in an intermediate stage of development in the chromite-rich sample from Goliath, provide useful constraints for considering rival models of in situ crystallization as opposed to gravitational accumulation mechanisms for olivine–chromite cumulates. These arguments have been discussed over decades in the layered intrusion literature (Campbell 1977, 1978; Jackson 1961). Despite compelling evidence has been produced for in situ nucleation mechanisms, this still remains an area of controversy, particularly in discussions over the formation of chromite layers in the Bushveld Complex (Eales 2000; Eales and Costin 2012; Naldrett et al. 2011). Chromite crystals forming at the liquidus of the parent magma (as opposed to those crystallizing in situ from trapped liquid) have a strong tendency to grow as extensive clusters, interpreted as the result of heterogeneous selfnucleation. This feature is commonly observed in experimental studies of chromite growth (Barnes 1986; Campbell et al. 1978; Murck and Campbell 1986), and the same is also true to a lesser degree for olivine. The formation of chicken-wire textures (e.g. Agnew, Fig. 7) by a crystal settling mechanism would require that clusters of olivine grains, and clusters of chromite grains, fortuitously accumulated in such a way that the chromite clusters formed an almost complete framework around individual olivine crystals. This possibility is difficult to envisage given that settling rates for chromite and olivine clusters would be highly variable due to variability in size, density and shape factor affecting Stokes law settling velocities. The 3D results show that chromite clusters in the Agnew sample have very delicate open internal structures, with angles

Conclusions

123


In more static environments, lower degrees of supersaturation with chromite produced ‘‘chicken-wire’’ textures very similar to the disseminated chromite textures associated with massive chromitites in layered intrusions. These rocks are composed of extensive delicate 3D chain-like frameworks of thousands of chromite grains forming continuous rims around individual olivine crystals. This texture implies a high degree of local self-organization not consistent with a mechanical crystal settling origin. These results confirm that textural characteristics of chromite grains are distinctive indicators of crystallization environment and that imaging these textures in 3D reveals significant textural attributes not evident in conventional 2D thin sections. Preservation of chromite textures in highly altered rocks whose igneous textures have been overprinted by alteration provides a new opportunity for investigating facies variations in komatiite sequences in altered greenstone terranes. Furthermore, 3D textural observations on komatiitic chromite–olivine cumulates provide evidence for an in situ nucleation-growth crescumulate origin for some distinctive textures commonly encountered in large-layered intrusions. Acknowledgments Be´linda Godel is funded by the CSIRO Office of the Chief Executive Post-Doctoral Fellowship scheme. Analytical costs were partially funded by a ‘‘Pump Priming Grant’’ from the Faculty of Natural and Agricultural Sciences at The University of Western Australia. Accessibility to supercomputing facilities was provided by iVEC at the Australian Resources Research Centre (Perth). Dr Greg Hitchen provided assistance with electron microprobe analyses. This paper is an output from the CSIRO Minerals Down Under National Research Flagship. Associate editor Chris Ballhaus and two anonymous referees are acknowledged for their comments on the manuscript.

References Arndt NT, Naldrett AJ, Pyke DR (1977) Komatiitic and iron-rich tholeiitic lavas of Munro Township, northeast Ontario. J Petrol 18:319–369 Arndt NT, Barnes SJ, Lesher CM (2008) Komatiite. Cambridge University Press, Cambridge Barnes SJ (1986) The distribution of chromium among orthopyroxene, spinel and silicate liquid at atmospheric pressure. Geochim Cosmochim Ac 50:1889–1909 Barnes SJ (1998) Chromite in komatiites, 1. Magmatic controls on crystallization and composition. J Petrol 39:1689–1720. doi: 10.1093/petroj/39.10.1689 Barnes SJ (2000) Chromite in komatiites, II. Modification during greenschist to mid-amphibolite facies metamorphism. J Petrol 41:387–409 Barnes SJ (2006) Komatiite-hosted nickel sulfide deposits: geology, geochemistry, and genesis. Soc Econ Geol Spec Pub 13:51–118 Barnes SJ, Fiorentini ML (2012) Komatiite magmas and nickel sulfide deposits: a comparison of variably endowed Archean terranes. Econ Geol 107:755–780. doi:10.2113/econgeo.107.5.755 Barnes SJ, Hill RET (1995) Poikilitic chromfite in komatiitic cumulates. Miner Petrol 54:85–92. doi:10.1007/bf01162760

123

Barnes SJ, Roeder PL (2001) The range of spinel compositions in terrestrial mafic and ultramafic rocks. J Petrol 42:2279–2302. doi:10.1093/petrology/42.12.2279 Barnes SJ, Fiorentini ML, Austin P, Gessner K, Hough RM, Squelch AP (2008) Three-dimensional morphology of magmatic sulfides sheds light on ore formation and sulfide melt migration. Geology 36:655–658. doi:10.1130/g24779a.1 Barnes SJ, Wells MA, Verrall MR (2009) Effects of magmatic processes, serpentinization, and talc-carbonate alteration on sulfide mineralogy and ore textures in the Black Swan disseminated nickel sulfide deposit, Yilgarn Craton. Econ Geol 104:539–562. doi:10.2113/gsecongeo.104.4.539 Campbell IH (1968) The origin of heteradcumulate and adcumulate textures in the Jimberlana Norite. Geol Mag 105:378–383 Campbell IH (1977) A study of macrorhythmic layering and cumulate processes in the Jimberlana Intrusion, Western Australia, part I: the upper layered series. J Petrol 18:183–215 Campbell IH (1978) Some problems with the cumulus theory. Lithos 11:311–323 Campbell IH, Roeder PL, Dixon JM (1978) Crystal buoyancy in basaltic liquids and other experiments with a centrifuge furnace. Contrib Mineral Petrol 67:369–377 Cashman KV, Marsh BD (1988) Crystal size distribution (CSD) in rocks and the kinetics and dynamics of crystallization II: Makaopuhi lava lake. Contrib Mineral Petrol 99:292–305 Donaldson CH (1982) Origin of some Rhum harrisite by segregation of intercumulus liquid. Min Mag 45:201–209 Dowling SE, Barnes SJ, Hill RET, Hicks JD (2004) Komatiites and nickel sulfide ores of the Black Swan area, Yilgarn Craton, Western Australia. 2: geology and genesis of the orebodies. Min Deposita 39:707–728 Eales HV (2000) Implications of the chromium budget of the Western Limb of the Bushveld Complex. S Afr J Geol 103:141 Eales HV, Costin G (2012) Crustally contaminated komatiite: primary source of the chromitites and marginal, lower, and critical zone magmas in a staging chamber beneath the Bushveld Complex. Econ Geol 107:645–665. doi:10.2113/econgeo.107.4.645 Fonseca ROC, Campbell IH, O’Neill HSC, Fitzgerald JD (2008) Oxygen solubility and speciation in sulphide-rich mattes. Geochim Cosmochim Acta 72:2619–2635 Frost KM, Grove DI (1989) Ocellar units in the Kambalda– Widgiemooltha komatiite sequence: evidence for sediment assimilation by komatiite lavas. In: Prendrergast MJ, Jones MJ (eds) Magmatic sulphides - The Zimbabwe volume, vol Institution., of Mining and MetallurgyLondon, U.K., pp 204–214 Godel B, Barnes SJ, Barnes S-J, Maier WD (2010) Platinum ore in 3D: insights from high-resolution X-ray computed tomography. Geology 38:1127–1130 Godel B, Gonzalez-Alvarez I, Barnes SJ, Barnes S-J, Parker P, Day J (2012) Sulfides and sulfarsenides from the Rosie Nickel Prospect, Duketon Greenstone Belt, Western Australia. Econ Geol 107:275–294 Groves DI, Hudson DR, Hack TB (1974) Modification of iron-nickel sulfides during serpentinization and talc-carbonate alteration at Black Swan, Western Australia. Econ Geol 69:1265–1281. doi: 10.2113/gsecongeo.69.8.1265 Higgins MD (2006) Quantitative textural measurements in igneous and metamorphic petrology. Cambridge University Press, Cambridge 265 pp Hill RET, Barnes SJ, Gole MJ, Dowling SE (1995) The volcanology of komatiites as deduced from field relationships in the Norseman-Wiluna greenstone belt, Western Australia. Lithos 34:159–188 Hill RET, Barnes SJ, Dowling SE, Thordarson T (2004) Komatiites and nickel sulphide orebodies of the Black Swan area, Yilgarn

Contrib Mineral Petrol Craton, Western Australia. 1. Petrology and volcanology of host rocks. Miner Deposita 39:684–706 Irvine TN (1965) Chromian spinel as a petrogenetic indicator: part 1 theory. Can J Earth Sci 2:648–672. doi:10.1139/e65-046 Irvine TN (1967) Chromian spinel as a petrogenetic indicator: part 2 petrologic applications. Can J Earth Sci 4:71–103. doi:10.1139/ e67-004 Irvine TN (1980) Magmatic density currents and cumulus processes. Am J Sci 280A:1–58 Jackson ED (1961) Primary textures and mineral associations in the ultramafic zone of the Stillwater Complex. Montana 358:1–106 Jerram DA, Cheadle MJ (2000) On the cluster analysis of grains and crystals in rocks. Am Mineral 85:47–67 Jerram DA, Cheadle MJ, Hunter RH, Elliott MT (1996) The spatial distribution of grains and crystals in rocks. Contrib Mineral Petrol 125:60–74 Jerram DA, Cheadle MJ, Philpotts AR (2003) Quantifying the building blocks of igneous rocks; are clustered crystal frameworks the foundation? J Petrol 44:2033–2051 Jerram DA, Davis GR, Mock A, Charrier A, Marsh BD (2010) Quantifying 3D crystal populations, packing and layering in shallow intrusions: a case study from the Basement Sill, Dry Valleys, Antarctica. Geosphere 6:537–548. doi:10.1130/ges00538.1 Marsh BD (1988) Crystal size distribution (CSD) in rocks and the kinetics and dynamics of crystallization. Contrib Mineral Petrol 99:277–291. doi:10.1007/bf00375362 Marsh BD (1998) On the interpretation of crystal size distributions in magmatic systems. J Petrol 39:553–599. doi:10.1093/petroj/ 39.4.553 Mathison CI, Booth RA (1990) Macrorhythmically layered gabbronorites in the Windimurra gabbroid complex, Western Australia. Lithos 24:171–180 McBirney AR, Hunter RH (1995) The cumulate paradigm reconsidered. J Geol 103:114–122 McBirney AR, Noyes RM (1979) Crystallization and layering in the Skaergaard Intrusion. J Petrol 20:487–554 Morse SA (1986) Convection in aid of adcumulus growth. J Petrol 27:1183–1214

Murck BW, Campbell IH (1986) The effects of temperature, oxygen fugacity and melt composition on the behaviour of chromium in basic and ultrabasic melts. Geochim Cosmochim Acta 50:1871– 1887 Naldrett A, Wilson A, Kinnaird J, Yudovskaya M, Chunnett G (2011) The origin of chromitites and related PGE mineralization in the Bushveld Complex: new mineralogical and petrological constraints. Miner Deposita 47:209–232. doi:10.1007/s00126-0110366-3 O’Driscoll B, Troll VR, Donaldson CH, Jerram DA (2006) A crystal size distribution study of harrisitic and granular olivine from the Rum Layered Suite, NW Scotland. J Petrol 48:253–270 Page P, Barnes S-J (2009) Using trace elements in chromites to constrain the origin of podiform chromitites in the Thetford Mines Ophiolite, Quebec, Canada. Econ Geol 104:997–1018. doi:10.2113/gsecongeo.104.7.997 Perring CS, Barnes SJ, Hill RET (1995) The physical volcanology of komatiite sequences from Forrestania, Southern Cross Province, Western Australia. Lithos 34:189–207 Ricard LP, Godel BM, Chanu J-B (2012) CSDToolbox 1.0: a MATLAB program for the analysis of crystal-size distribution of large datasets. Comput Geosci. doi:10.1016/j.cageo.2012.01.010 The´baud N, Barnes SJ, Fiorentini ML (2012) Komatiites of the Wildara Leonora Belt, Yilgarn Craton, WA: the missing link in the Kalgoorlie Terrane? Precambrian Res 196–197:234–246 Toriumi M (1994) Cluster formation of Cr-spinel during magmatic differentiation. In: Takaki R (ed) Research of pattern formation, Chapter 4. KTK Scientific Publishers, Tokyo, pp 239–257 Vinet N, Higgins MD (2010) Magma solidification processes beneath Kilauea Volcano, Hawaii: a quantitative textural and geochemical study of the 1969–1974 Mauna Ulu Lavas. J Petrol 51:1297–1332. doi:10.1093/petrology/egq020 Wager LR, Brown GM (eds) (1968) Layered igneous rocks. Oliver and Boyd, Edinburgh Zhou M-F, Kerrich R (1992) Morphology and composition of chromite in komatiites from the Belingwe greenstone belt, Zimbabwe. Can Mineral 30:303–317

123