Order of Authors: Belinda Godel; Sarah-Jane Barnes, PhD; Wolfgang D Maier, PhD

Transcription

1 Elsevier Editorial System(tm) for Lithos Manuscript Draft Manuscript Number: LITHOS2854 Title: Parental magma composition inferred from in situ trace elements in cumulus and intercumulus silicate minerals: example from the Lower and Lower critical Zones of the Bushveld complex (South- Africa) Article Type: Research Paper Keywords: parental magma; LA-ICP-MS; Bushveld Complex Corresponding Author: Dr. Belinda Godel, Corresponding Author's Institution: CSIRO First Author: Belinda Godel Order of Authors: Belinda Godel; Sarah-Jane Barnes, PhD; Wolfgang D Maier, PhD Abstract: Major and trace element concentrations in the whole-rock and cumulus and intercumulus minerals were determined in cumulate rocks from the Lower and the Lower Critical zones of the Bushveld Complex (South-Africa). Together with microtextural observations, these new chemical data are used to provide insights on the formation of the cumulate and to estimate the possible parental liquid composition from which the cumulate rocks crystallized. Cumulus (orthopyroxene and olivine) and intercumulus (clinopyroxene and plagioclase) minerals in the samples considered have a relatively constant composition in term of major and trace element throughout the Lower and the Lower Critical Zones suggesting that the minerals formed from magma with a relatively constant composition and followed a similar crystallization path or evolution history. Evidence of high-temperature deformation in orthopyroxene and plagioclase consistent with compaction of the cumulate pile are present, however the absence of dihedral angle modification and the steady increase in Ti concentrations in intercumulus plagioclase with the An content suggest that no significant intercumulus liquid migration has occurred. Comparison of the whole-rock and mineral chemistry indicate for some samples a decoupling of the major and the trace elements consistent with the uncomplete equilibration of the cumulus phases with the trapped liquid rather than being an effect due to change in magma composition during fractional crystallization. The cumulus and intercumulus minerals are in most cases unzoned in terms of major and trace elements. In constrast to Fe-Mg or Cr, Ti shows a chemical zonation in pyroxene. The observed Ti zonation is consistent with the steady increase in Ti concentrations in interstitial plagioclase with decreasing An and is interpreted to be the results of the disequilibrium crystallization of the trapped melt during closed system fractional crystallization leading to a zonation reflecting differences in element diffusivities in pyroxene. Mass balance between mineral and whole rock data indicates that most of the transition metals (Ti, Sc, V, Mn, Co and Ni) and the HREE are present in orthopyroxene and to a lesser extent in clinopyroxene whereas the LREE are mainly controlled by plagioclase and to a lesser extent accessory phases such as monazite, apatite and zirconium-rich minerals. Calculation, based on whole-rock and in-situ cumulus and intercumulus mineral chemistry, indicate that the parental liquid from which the cumulates of the Lower and the Lower Critical Zones formed from a magma similar in composition to B1-type magma implying that the B1-type sills intruding in the vicinity of the Bushveld Complex are representative of the magma from which the Lower and Lower Critical zones have crystallized as suggested by previous

2 authors. In addition, our study illustrates that the combined analysis of whole-rock chemistry and insitu cumulus and intercumulus minerals give useful results which allow the estimation of parental liquid composition from cumulate rocks for a large range of elements. Moreover, the results illustrate that the simple assumption of equilibrium between a melt and a mineral to calculate a parental magma can only be applied to pure adcumulate rocks.

3 Highlights Research highlights Determination of the trace element concentrations in minerals from the Bushveld Complex Constraints on the formation of the cumulate rocks from the Lower and Lower Critical Zones Methods to calculate parental magma composition from cumulus and intercumulus minerals Calculation of the composition of Bushveld parental magma using the proposed method

4 *Manuscript Click here to download Manuscript: Godel-et-al_ dec-2010.doc Click here to view linked References Parental magma composition inferred from in situ trace elements in cumulus and intercumulus silicate minerals: example from the Lower and Lower Critical Zones of the Bushveld Complex (South-Africa) Bélinda Godel a,b,*, Sarah-Jane Barnes b and Wolfgang D. Maier c a* CSIRO Earth Science and Resource Engineering, Australian Resource Research Centre, Kensington, 6151, WA, Australia b Science de la Terre, Université du Québec à Chicoutimi, G7H 2B1, Chicoutimi, QC, Canada c Department of Geology, University of Oulu, Linnanmaa, 90014, Finland * Corresponding author address: CSIRO Earth Science and Resource Engineering Australian Resources Research Centre 26 Dick Perry Avenue. Kensington, 6151 Western Australia Australia godelbelinda@gmail.com Ph: +61 (0) Fax: +61 (0)

5 Abstract Major and trace element concentrations in the whole-rock and cumulus and intercumulus minerals were determined in cumulate rocks from the Lower and the Lower Critical zones of the Bushveld Complex (South-Africa). Together with microtextural observations, these new chemical data are used to provide insights on the formation of the cumulate and to estimate the possible parental liquid composition from which the cumulate rocks crystallized. Cumulus (orthopyroxene and olivine) and intercumulus (clinopyroxene and plagioclase) minerals in the samples considered have a relatively constant composition in term of major and trace element throughout the Lower and the Lower Critical Zones suggesting that the minerals formed from magma with a relatively constant composition and followed a similar crystallization path or evolution history. Evidence of high-temperature deformation in orthopyroxene and plagioclase consistent with compaction of the cumulate pile are present, however the absence of dihedral angle modification and the steady increase in Ti concentrations in intercumulus plagioclase with the An content suggest that no significant intercumulus liquid migration has occurred. Comparison of the whole-rock and mineral chemistry indicate for some samples a decoupling of the major and the trace elements consistent with the uncomplete equilibration of the cumulus phases with the trapped liquid rather than being an effect due to change in magma composition during fractional crystallization. The cumulus and intercumulus minerals are in most cases unzoned in terms of major and trace elements. In contrast to Fe-Mg or Cr, Ti shows a chemical zonation in pyroxene. The observed Ti zonation is consistent with the steady increase in Ti concentrations in interstitial plagioclase with decreasing An and is interpreted to be the results of the disequilibrium crystallization of the trapped melt during closed system fractional crystallization leading to a zonation reflecting differences in element diffusivities in pyroxene. 2

6 Mass balance between mineral and whole rock data indicates that most of the transition metals (Ti, Sc, V, Mn, Co and Ni) and the HREE are present in orthopyroxene and to a lesser extent in clinopyroxene whereas the LREE are mainly controlled by plagioclase and to a lesser extent accessory phases such as monazite, apatite and zirconium-rich minerals. Calculation, based on whole-rock and in-situ cumulus and intercumulus mineral chemistry, indicate that the parental liquid from which the cumulates of the Lower and the Lower Critical Zones formed from a magma similar in composition to B1-type magma implying that the B1-type sills intruding in the vicinity of the Bushveld Complex are representative of the magma from which the Lower and Lower Critical zones have crystallized as suggested by previous authors. In addition, our study illustrates that the combined analysis of whole-rock chemistry and in-situ cumulus and intercumulus minerals give useful results which allow the estimation of parental liquid composition from cumulate rocks for a large range of elements. Moreover, the results illustrate that the simple assumption of equilibrium between a melt and a mineral to calculate a parental magma can only be applied to pure adcumulate rocks

7 Introduction Layered intrusions are studied to constrain and better understand the evolution of large magmatic systems including magma differentiation and magma chamber processes (e.g. Barnes et al., 2004; Bédard, 1994b; Bédard, 2001; Cawthorn, 1983; Charlier et al., 2005; Duchesne and Charlier, 2005; Maier et al., 2000; McBirney, 1995; Seat et al., Submitted; Tollari et al., 2008; Zingg, 1996). One of the key parameters in the modelling of the crystallization history of an intrusion is the estimation of the composition of its parental magma(s) (Wager and Brown, 1968). Traditionally, the composition of the parental liquid of a layered intrusion has been estimated by using two different approaches: (i) by analysing the whole-rock composition of the fine-grained rocks occurring at the contact of the intrusion as chilled margins, or as dykes or sills spatially associated with the intrusion (Barnes et al., in press; Cawthorn, 2006; Cawthorn et al., 1981; Curl, 2001; Davies and Tredoux, 1985; Godel et al., submitted; Harmer and Sharpe, 1985; Seat et al., Submitted; Sharpe, 1981; Sharpe and Hulbert, 1985) or; (ii) by using the major and trace element compositions of cumulate minerals to back-calculate the parental magma with which they were in equilibrium (Bédard, 1994b; Bédard, 2001; Davies and Cawthorn, 1984; Duchesne and Charlier, 2005; Eales, 2000b; Maier and Barnes, 1998). Several issues arise in the estimation of the parental magma from the chilled margins or the fine-crystallized rocks spatially associated with an intrusion: (i) the rocks may represent melts that were contaminated by interaction with the country rocks; (ii) all the magmas from which a layered intrusion crystallized are not necessarily preserved as dykes or sills (Latypov et al., 2007; Marsh, 1995); (iii) complex mixing between several magma types may have occurred (Godel et al., submitted) or; (iv) the magma may have undergone a complex fractional crystallization history between successive magma chamber replenishments. As a 4

8 result, the fine-grained rocks may have formed from a melt which may not necessarily be representative of the parental magma of the intrusion with which they are associated. The estimation of the parental magma from the analysis of cumulus minerals was until recently limited to major and few trace elements due to the lack of in situ methods to accurately determine the full range of trace elements at low level. Mineral separates are susceptible to containing inclusions of accessory phases often enriched in trace elements. Hence, although the separates may be analysed to determine trace elements at low concentration level, the presence of inclusions may lead to the overestimation of the trace element abundances (Arndt et al., 2005) which in turn will impact on the estimation of the parental magma. In addition, the interpretation of the composition of cumulate minerals remains complicated as the composition of a given mineral depends on several factors including its degree of equilibration with the magma, the potential reaction between minerals and trapped liquid or fractionated melt which may have percolated during compaction, the exsolution of low-temperature phases during cooling, the potential requilibration of the mineral on cooling or metamorphism, and the alteration by late-stage magmatic or nonmagmatic fluids. The Rustenburg Layered Suite (RLS) of the Bushveld Complex forms the largest layered intrusion on Earth and hence is a good natural laboratory to consider and constrain the effects of the above processes in the formation of cumulate rocks. We have determined the trace element concentrations in the whole-rocks and in situ (by laser ablation ICP-MS) in the cumulus (orthopyroxene and olivine) and intercumulus (clinopyroxene and plagioclase) minerals from the Lower and Lower Critical zones of the RLS. Together with micro-textural observations, the chemical data are used to propose a parental magma from which the cumulate crystallized and to provide insights on the formation of the cumulate

9 Geology 2.1.Geological settings The ± 1.3 Ma old Rustenburg Layered Suite of the Bushveld Complex (Harmer and Armstrong, 2000; Scoates and Friedman, 2008) (Fig. 1) is the world s largest layered intrusion (Eales and Cawthorn, 1996) and has undergone little deformation or metamorphism after its solidification. The 6.5 to 8.7 kilometres thick ultramafic and mafic rocks of the RLS are sub-divided into five major zones (South African Committee For Stratigraphy, 1980): the Marginal-, Lower-, Critical-, Main- and Upper Zones (Fig. 1b) Lower and Lower Critical zones The present study focuses on the Lower Zone and the Lower Critical Zone of the Union Section in the northern part of the Western Limb of the Bushveld Complex. The samples studied were taken from boreholes NG-1 to NG-3 drilled by the Geological Survey of South- Africa in 1987 (Fig. 1a). At this locality, the Marginal Zone is absent and the Lower Zone directly overlies the sedimentary floor rocks. The Lower Zone at the Union Section (Fig. 2) has a thickness of approximately 800 metres and consists principally of cyclic units of harzburgite, orthopyroxenite and dunite (Teigler and Eales, 1996). The Lower Critical Zone consists of cyclic units of orthopyroxenite, minor harzburgite, and chromitite (Fig. 2). The stratigraphic position of the samples analysed in the present study is given in Fig. 2. The NG- 1 to NG-3 boreholes have been the subject of a number of petrographic and geochemical studies (Eales, 2000a; Eales et al., 1993b; Eales et al., 1990; Maier and Barnes, 2003; Maier and Barnes, 1999a; Teigler, 1990b) which guided our sampling strategy. In order to provide insights on the parental magma from which the cumulates crystallized, we selected samples covering the whole stratigraphic interval based on two main criteria: (i) the samples should show the least alteration possible and, (ii) the rocks should have formed from the least 6

10 evolved magma at a given stratigraphic height, i.e. based on whole-rock chemistry the samples selected should have the highest Mg number Parental magmas Several different magma types have been recognized intruding in the floor of the Bushveld Complex, mainly in the form of sills (Barnes et al., in press; Cawthorn et al., 1981; Curl, 2001; Davies and Tredoux, 1985; Eales and Cawthorn, 1996; Harmer and Sharpe, 1985; Sharpe, 1981). Based on their stratigraphic positions and their compositions, the marginal rocks, sills and dykes were divided into three groups (Harmer and Sharpe, 1985; Sharpe, 1981; Sharpe and Hulbert, 1985): (i) quench-textured norites referred to as the Bushveld 1 (B1) magma which have compositions of Mg-rich basaltic andesites (Barnes et al., in press) and ultramafic sills formed by a mixture of olivine, chromite and B1 liquids referred to as B1- UM (Barnes et al., in press; Davies and Tredoux, 1985; Sharpe and Hulbert, 1985); (ii) the Bushveld 2 (B2) and the Bushveld 3 (B3) sills consisting of fine-grained gabbronorites with compositions close to tholeiitic basalts. Based on field relationships and whole rock geochemistry, it was suggested that the Lower and Lower Critical Zones formed largely from B1 liquid (Barnes et al., in press; Davies and Tredoux, 1985; Harmer and Sharpe, 1985; Maier and Barnes, 1998). 3. Petrography The orthopyroxenites (Fig. 3 and 4) are mainly composed of cumulus crystals of orthopyroxene (84 to 96 modal%) of various size (0.5 up to 10 mm in length). The orthopyroxenes are generally aligned perpendicular to the paleovertical and define a foliation in the rock (Fig. 4). Clinopyroxene (1 to 6 modal%) and plagioclase (3 to 10 modal%) occur as interstitial phases between the orthopyroxene crystals. The clinopyroxenes are in most 7

11 cases small (0.1 to 0.3 mm), but in some samples they form oikocrysts 3-6 mm in size (Fig. 3) surrounding orthopyroxenes. Chromite and phlogopite are present as minor phases (<1%). The harzburgite (sample NG ) consists of elongate cumulus crystals of (2-5 mm) orthopyroxene (~45% modal%) and subhedral (1-3 mm) olivine (~45 modal%). Small (0.2 to 0.8 mm) crystals of plagioclase (~5 modal%) and clinopyroxene (~3 modal% ) occur as interstitial phases between orthopyroxene or olivine. Minor phlogopite (up to 0.5 modal%) and euhedral crystals of chromite (up to 3 modal% and 1 mm in size) are observed as inclusions in orthopyroxene or at the contact between cumulus silicates. Most of the orthopyroxene crystals exhibit indented contacts and show undulose extinctions. Clinopyroxene exsolution lamellae are common (Fig. 3) and are oriented along the cleavages. Interstitial plagioclase exhibits undulose extinction and deformation twins oriented parallel to the paleovertical direction. These crystal deformation features suggest that deformation occurred at a high temperature by diffusion creep or dissolution, consistent with compaction or flow of the cumulate. Similar crystal deformation features have previously been documented in the Bushveld Complex (Barnes and Maier, 2002; Boorman et al., 2004; Godel et al., 2006). Dihedral angles between orthopyroxene plagioclase or orthopyroxene clinopyroxene are in most cases <50 degrees and do not exhibit visible angle modification due to sub-solidus processes (Fig. 3 and 4). 4. Methodology 4.1. Whole-rock geochemistry Major oxide and trace-element concentrations of the rocks have previously been determined by X-ray fluorescence analysis and instrumental neutron activation analysis (Maier and Barnes, 1998). Trace elements were re-determined for the present study on the same rock powders by ICP-MS analysis at the Ontario Geosciences Laboratories (Sudbury, Canada). The results for most elements were similar to those previously obtained, but with 8

12 greater precision. Duplicate samples by ICP-MS show relative standard deviation for the trace elements in the 1 to 7 percent range Major and trace elements in silicate minerals Major element compositions of the silicate minerals including orthopyroxene, clinopyroxene, olivine and plagioclase were determined at Laval University (Quebec City, Canada) using a Cameca SX100 electron microprobe. The microprobe was operated in wavelength dispersive mode (WDS) using an acceleration voltage of 15 kv, a beam current of 20A, a beam diameter of 2-5 microns and counting times set to 20 and 10 seconds on the peaks and backgrounds, respectively. Trace element concentrations of the minerals were determined by laser-ablation inductively coupled mass spectrometry (LA-ICP-MS) at the University of Quebec at Chicoutimi (UQAC) using a Thermo X7 ICP-MS coupled with a New Wave Research 213 nm Nd:YAG UV laser ablation system. The analyses were conducted using 80 micron diameter spots, a laser frequency of 10 Hz, a power of 0.8 mj/pulse and, a He-Ar carrier gas. The gas background was collected for 20 seconds followed by 60 seconds of data acquisition. In addition, laser traverses (55 microns diameter beam) along mineral grain boundaries and microcracks were carried out to examine whether the incompatible elements may be concentrated in these particular areas. In both cases, the ablated material was analysed by ICP-MS operated in time-resolved mode using peak jumping. NIST-610 reference material was used for calibration and NIST-612 as a monitor (Table 1). The reduction of the data was computed using PlasmaLab software (ThermoElemental) using 29 Si as internal standard. Si rather than Ca was used as an internal standard because Ca is inhomogeneous in the mineral considered. The fractionation of Si from Ca was checked and is found to be negligible (Table 1). The results of calibration during all analytical sessions and detection limits are summarized in Table 1. 9

13 Geochemistry 5.1. Whole-Rock Chemistry The rocks from the Lower and Lower Critical zones show high concentrations of compatible elements (FeO Total, MnO, MgO, Ni, Cr, Co and V) and low concentrations of incompatible elements (Table 2) consistent with the petrographic observations that the rocks consist largely of cumulus orthopyroxene with minor olivine and chromite (Table 2 and Fig. 5). The whole-rocks have similar, parallel mantle normalized trace element patterns (Fig. 6). The rocks are enriched (up to 10 times mantle values) in large-ion lithophile elements (LILE) and mostly exhibit positive U, Pb and Cr anomalies and negative Th, Nb and Ti anomalies (Fig. 6). The positive Pb anomalies are more pronounced in the rocks from the Lower Zone (Fig. 6). In both zones, the rocks are enriched in light rare earth elements (LREE) relative to the middle and heavy rare earth elements (MREE and HREE) with (La/Sm) CN ratios varying from 3.0 to 3.6 and the (Sm/Yb) CN ranging from 1.0 to 2.2. No systematic changes in trace element concentrations are observed within the two zones, i.e. the samples with lesser or greater incompatible element concentrations are not necessarily located at the bottom and top of the sequence, respectively. Overall, the mantle-normalized patterns of the rocks from both the Lower and Lower Critical Zones are parallel to the B1 magma (Fig. 6) consistent with the presence of a B1 liquid fraction in the cumulate rocks (see discussion below) Olivine Chemistry Olivine (Table 3) was analysed in the harzburgite sample NG where the mineral occurs as a cumulus phase. The major element composition of the olivine is homogenous within the sample, with forsterite content (Fo) varying within a small range of values (85.6 ± 0.1, n= 9). These values are similar to those reported previously for the Lower Zone of the Bushveld Complex (e.g. Cameron, 1978; Eales et al., 1993a; Hulbert and Von Gruenewaldt, 1985; Teigler and Eales, 1996). Nickel, Co and Mn are compatible in olivine 10

14 and their concentrations vary within narrow ranges, i.e 2540 ± 100 ppm, Co 192 ± 10 ppm and 1410 ± 90 ppm for Ni and Mn, respectively. Chromium and Ti concentrations are very low, and only slightly above the detection limits. All of the other trace elements are below the lower detection limit given in Table 1 which is consistent with the published low partition coefficients (Bédard, 2005) of these elements into olivine Orthopyroxene Chemistry Enstatite occurs as a cumulus phase throughout the studied stratigraphic interval. The enstatite has a Mg number (Mg#) ranging from 0.80 to 0.88 (Table 4) which is within the range of previously published orthopyroxene analyses (e.g. Cameron, 1978; Eales et al., 1993a; Eales and Cawthorn, 1996; Teigler and Eales, 1996). The Mg# variation within a sample is less than 0.01 indicating that the orthopyroxene composition remains relatively homogenous at the sample scale (Table 4). The most primitive orthopyroxenes (Mg# from 0.88 ± 0.004) are found in the harzburgite and in orthopyroxenite at the top of the Lower Zone (Fig. 2). The Mg# does not show any systematic variation trends within the Lower Zone (Fig. 2). No difference in the concentration of most of the compatible or moderately incompatible elements (D> 0.1, Bédard, 2007) such as Co, Ni, V, Cr, Sc and Mn are observed between the cores and the margins of the orthopyroxenes (e.g. Fig. 7). However, the margins of some crystals are enriched in Ti relative to the core of the minerals. No discernable difference (Fig. 7) between the cores and the margins of orthopyroxenes are observed for most of the highly incompatible (D<0.1, Bédard, 2007) elements, except for Ti which is systematically enriched in the margins of grains (Fig. 7). The mantle-normalized trace element patterns of Lower Zone orthopyroxenes are slightly depleted in LREE compared to those of the Lower Critical Zone (Fig. 8). All orthopyroxenes are depleted in Ba relative to 11

15 the mantle (0.05 to 0.2 times mantle values) and slightly enriched in Th and U (1 to 4 times mantle values). Overall, the orthopyroxene patterns show a smooth increase from the LREE to the transition metals (Fig. 8) with concentration varying from 0.1 to 1 times mantle, respectively. The general shapes of the mantle-normalized patterns reflect the partition behaviour of the trace elements into orthopyroxene (Bédard, 2007). Depending on the sample considered (Table 5), there are positive correlations between the Ca concentration (as determined with the ICP-MS) and the trace element concentrations (particularly the REE) in the orthopyroxenes. The significance of this observation is discussed in Part Clinopyroxene Chemistry The clinopyroxenes analysed in the present study occur as interstitial phases between orthopyroxene and/or olivine and consist mainly of diopside with minor augite (Table 7). The Mg# of clinopyroxene varies from 0.90 ± 0.02 to 0.87 ± 0.02 in the Lower and Lower Critical Zones, respectively (Table 6). These compositions are similar to the limited published data for clinopyroxene from the Lower and Lower Critical zones (Atkins, 1969; Teigler, 1990a). The Mg# of the clinopyroxenes follows the same trend as the Mg# of the orthopyroxenes (Fig. 2). No significant compositional variations were observed for the major elements in the clinopyroxenes (Fig. 9 and Table 6) either at the sample or mineral scales. There is also no systematic variation between the margins and the cores of the grains suggesting the clinopyroxenes are unzoned in terms of their major element composition. In contrast to the orthopyroxenes, the clinopyroxenes are in most cases small crystals. Hence, it has been difficult to systematically analyse the core and border zones of the minerals by LA-ICP-MS (Figure 9). As observed in orthopyroxene, only Ti and Th concentrations vary from the core to the margins of clinopyroxene (Fig. 9 and Table 6). No 12

16 discernable chemical zonation was detected for the other compatible elements (e.g. V, Cr, Mn, Ni) or the incompatible elements. The mantle-normalized trace element patterns of the clinopyroxenes (Fig. 10) have a similar shape for the rocks from the Lower Zone and the Lower Critical Zone. The clinopyroxenes are depleted in Ba and enriched Th and U. The clinopyroxene patterns show a smooth decrease from the LREE (10 to 20 times mantle) to the HREE (2 to 3 times mantle) and the transition metals (1 to 3 times mantle) and are depleted in Co and Ni (Fig. 10). There is a strong negative anomaly in Ti for all the samples. These results are consistent with the published partition data (Bédard, 2001) Plagioclase Chemistry The plagioclase observed in the Lower and Lower Critical zones is interstitial to orthopyroxene and/or olivine. The anorthite content (An) of the plagioclase shows a wide range of values from 0.63 to 0.85 (Table 7) where the lowest value (average 0.63 ± 0.06) is recorded in the basal portion (147 meter height) of the Lower Zone (Fig. 2). The plagioclase composition varies at both sample and mineral scales. Although the An contents vary within a large range of composition, and most of the plagioclases exhibit zonations in terms of major elements, there is no relationship between the An content and the trace element concentrations (Table 8). The trace element composition remains constant throughout the entire Lower and Lower Critical zones (Fig. 11). The plagioclases are enriched in Ba and LREE relative to the HREE and the transition metals (Fig. 11). The shape of the mantle- normalized trace element patterns reflects the partitioning behaviour of the elements in plagioclase (Bedard, 2006). 6. Determining the host of trace-elements 6.1. Principle and method 13

17 In order to determine the proportion of each of the individual trace elements hosted by the various minerals analyzed, we calculated the percentage (P i Min ) of each element (i) in a given mineral (Min) as follows: P i Min = (F Min C i Min / C i WR ) x where F Min is the weight fraction of the mineral considered and C Min i and C WR i are the concentrations of the element i in the mineral and whole-rock, respectively. The weight fraction of each mineral was calculated from whole-rock major element concentrations using the least square method described in Albarède (1995) and based on mineral and whole-rock compositions. The results obtained are in agreement with the CIPW norm (Table 9) and the petrographic observations (recalculated to weight fraction) Results In the harzburgite, 35 to 60 wt% of the Ti, V, Cr and Mn are present in orthopyroxene (Fig 12a). The remainder of the Ti, V and Cr budget is balanced by the chromite (up to 1 wt%) present in the sample whereas the Mn is present in the olivine. Nickel and cobalt are compatible in both orthopyroxene and olivine (Bédard, 2005, 2007). Up to 70 wt% of Ni and Co are hosted in olivine and orthopyroxene (Fig. 12a). Plagioclase and clinopyroxene control about half of the LREE and MREE budget (Fig. 12a). The HREE are mainly controlled by orthopyroxene and to a lesser extent clinopyroxene and olivine (e.g. Yb and Lu). Only 10 to 30 wt% of Th and U is controlled by orthopyroxene, clinopyroxene and minor plagioclase (Fig. 12a). In the pyroxenite (Fig. 12b), over 60 wt% of the transition metals (Ti, Sc, V, Mn, Co and Ni) are present in orthopyroxene. The balance is hosted in minor chromite representing 0.5 to 1.5 wt% of the rocks. Almost all the HREE are controlled by orthopyroxene which also hosts 20 to 50 % of Th and U. As observed in the harzburgite, the remainder of the elements are balanced by accessories phases (Fig. 13) such as apatite or 14

18 zircon, or as fine alteration products observed within cracks or along crystal boundaries, as indicated on the time-resolved spectras during the laser ablation analysis. 7. Role of clinopyroxene exsolution Small clinopyroxene exsolution lamellaes or blebs are common in the samples from the Lower and Lower Critical zones (Fig. 3). Clinopyroxene (Table 6) is enriched in trace elements relative to the orthopyroxene (Table 4). Hence, the ablation of tiny clinopyroxene exsolution lamellae or blebs during the analysis would result in: (i) a higher Ca and trace element concentration of the orthopyroxene and; (ii) a positive correlation between the contents of Ca and the other trace elements. In some samples, the Ca concentration in orthopyroxene (as determined by LA-ICP-MS) is positively correlated with trace element concentrations (Table 5) and is usually greater (up to 20%) than the average value obtained by microprobe (Table 4) suggesting a clinopyroxene-component in the orthopyroxene. However, our calculations of the parental liquid by using the average trace element concentrations of the orthopyroxene as detailed below are consistent for all the samples (Table 10). This would suggest that the laser beam averages the orthopyroxene host and clinopyroxene exsolution lamellae compositions resulting in average measured trace element concentrations similar to that in orthopyroxene prior clinopyroxene exsolution upon cooling. 8. Estimation of the parental liquid composition 8.1. Method 1: Equilibrium between cumulus mineral-silicate liquid Cumulate rocks are composed of a mixture of cumulus minerals and interstitial liquid. In many publications, the trace element concentrations of the parental liquid of an intrusion have been estimated using mineral analysis (e.g. Cawthorn, 1996; Lambert and Simmons, 1987; Mathez, 1995a). In these cases, the concentration (C i Liq) of an element i in the liquid 15

19 from which a cumulus mineral formed was calculated by assuming chemical equilibrium between the mineral (Min) and the liquid (Liq) by using the following equation. C i Liq = C i Min / D i Min/Liq (1) where C i Min is the concentration of the element i in a cumulus mineral and D i Cum/Liq is the partition coefficient between the mineral and the liquid. This equation (1) assumes that the composition of the mineral has not changed after its accumulation. However, several studies (e.g. Barnes, 1986b; Bédard, 1994b; Bédard, 2001; Cawthorn, 1996) have shown that the cumulus minerals may react with the trapped intercumulus liquid ( trapped liquid shift effect, Barnes (1986)). The trapped liquid crystallizes to produce overgrowths on cumulus crystals and interstitial phases which are enriched in incompatible elements (Fig. 14). In addition, the minerals may re-equibrate during sub-solidus processes. As a result, the final concentration of an element in the cumulus minerals will be a weighted average of liquidus and post-cumulus composition (Barnes, 1986b) and hence estimation of the parental liquid composition from equation (1) will be affected (see discussion below). In order to take the above processes into account, we used two additional methods of calculation Method 2: Estimation from whole-rock chemistry The calculation is in many ways comparable to that of proposed by Bédard (1994a) based on the hypothesis that the whole-rock chemistry represents the composition of cumulus phases (orthopyroxene and/or olivine) and the composition of interstitial melt (crystallized as clinopyroxene, plagioclase and accessory phases) in a closed system. In that case, the cumulate minerals are assumed to be in equilibrium with the trapped liquid present and no significant melt migration has occurred. The overgrowths on cumulus minerals due to trapped liquid is taken into account by assuming that the initial orthopyroxene overgrowth crystallized from a liquid with similar composition as the trapped liquid. In this case, the concentration 16

20 (C i Liq) of the element i in the parental liquid may be calculated using the following mass balance equation for olivine-orthopyroxene cumulates: C i WR = F Opx(Initial). C i Opx + F Ol. C i Ol + F Liq. C i Liq (2) with C i Opx = D Opx/Liq. C i Liq and C i Ol = D Ol/Liq. C i Liq (3) by substituting equation (3) in equation (2), this gives: C i Liq = C i WR / ( F Opx(Initial). D Opx/Liq + F Ol.D Ol/Liq. + F Liq ) (4) C i WR, C i Opx and C i Ol are the concentrations of the element i in the whole rock and the cumulus orthopyroxene (core) and olivine, respectively. F Opx(Initial) is the fraction of cumulus orthopyroxene prior to post-cumulus overgrowth and F Ol is the fraction of olivine (for the pyroxenite F Ol =0 ). D Opx/liq and D Ol/Liq are the partition coefficients of the element i between the orthopyroxene or olivine and the liquid (see details below). In order to calculate the initial fraction of orthopyroxene, we assume the proportions of post-cumulus orthopyroxene equal to the proportion of the clinopyroxene which crystallized from the trapped liquid as proposed by Cawthorn (2006). The results of parental magma calculations are detailed and discussed below Method 3: Estimation from cumulus and intercumulus mineral chemistry In an alternative approach, we used the proportion and composition of the cumulus and major intercumulus phases to evaluate the composition of the parental liquid from which the cumulate formed. In that case, the concentration (C i Liq) of the element i in the parental liquid may be calculated using the following mass balance equations for olivine-orthopyroxene cumulate: F Opx (Initial). D Opx/Liq. C i Liq + F Ol. C i Ol + F Liq. C i Liq = C i Opx (Final). F Opx (Final) + F Ol. C i Ol + C i Cpx(Inter). F Cpx(Inter) + C i Plg(Inter). F Plg(Inter) (5) which may be solved for C Liq 17

21 C Liq = (C i Opx (Final). F Opx (Final) + C i Cpx(Inter). F Cpx(Inter) + C i Plg(Inter). F Plg(Inter) ) / (F Opx (Initial). D Opx/Liq + F Ol.D Ol/Liq + F Liq ) (6) F Opx(Initial) is the fraction of orthopyroxene prior to post-cumulus overgrowth crystallization from the trapped melt, F Ol is the fraction of olivine (for pyroxenite F Ol =0 ), F Cpx(Inter) and F Plg(Inter) are the fractions of clinopyroxene and plagioclase which crystallized as interstitial phases from the trapped liquid, and F Opx(Final) is the proportion of orthopyroxene in the cumulate after post-cumulus overgrowth (i.e. cumulus + post-cumulus proportion). C i Opx, C i Cpx and C i Plg are the concentration of the element i in the orthopyroxene, clinopyroxene and plagioclase. D Opx/liq and D Ol/Liq are the partition coefficient of the element i between the orthopyroxene or olivine and the liquid. The results of parental magma calculations are detailed and discussed below Determination of partition coefficients One of the key parameters in the modelling of magmatic processes is the determination of the Nernst partition coefficient (D Min/Liq ) between the mineral and the liquid considered to be in equilibrium with each other. Partition coefficients may vary as a function of several intensive parameters (Bédard, 2005, 2007; Blundy and Wood, 2003; Gaetani and Grove, 1997). Our calculation of the parental liquid from which the rocks of the Lower and Lower Critical zones formed (cf. equation (4) and (5)) require the estimation of partition coefficients for a range of elements between orthopyroxene or olivine and silicate melts (D Opx/Liq and D Ol/Liq ). Values of D Opx/Liq have been determined based on experimental data or natural minerals (McDade et al., 2003; Norman et al., 2005). Recently, Bédard (2005, 2007) examined the variation of D Opx/Liq and D Ol/Liq as a function of pressure, temperature and mineral and melt composition and proposed algebraic parameterization using least square 18

22 regression of D Opx/Liq and D Ol/Liq values. In order to calculate realistic D Opx/Liq for our samples, we used the method and the equations given in detail in Bédard (2007) and compared the results obtained with available published values (Bédard, 2001; McDade et al., 2003; Norman et al., 2005). The D Opx/Liq values used in the present work are given in Table 10. The D Ol/Liq calculated using Bédard (2005) Calculated parental liquid compositions Simple consideration of equilibrium between a cumulus mineral and melt (Method 1, equation 1) results in a estimation of parental magma with trace element concentrations for most of the elements (Fig. 15a and b) an order of magnitude higher than those obtained using either the whole-rock chemistry (Method 2, Fig. 15 c and d) or the cumulus and intercumulus mineral chemistry (method 3, Fig. 15e). This significant difference is due to the fact the Method 1 does not take into account the potential effect of the trapped liquid on the cumulus mineral chemistry suggesting that this method can only be applied to pure adcumulate rocks where trapped liquid is absent. In addition, the parental magmas (Fig. 15 and Table 11) calculated from the whole-rock data (Method 2, Fig. 15c and d) are significantly different for Ba, Th, U and the LREE than those calculated from the intercumulus and cumulus mineral data (Method 3, Fig. 15c). These elements are highly incompatible into the crystallizing cumulus minerals and mass balance results (Fig. 12) indicate that these elements are mainly hosted by accessory phases (e.g. apatite or zircon) which may crystallize during the late stage of crystallization forming interstitial phases along cumulus crystal boundaries (Fig. 13). These minor phases are not taken into account in Model 3 and hence the concentration in the parental magma inferred for these elements will be underestimated. In contrast, calculations based on whole-rock chemistry(method 2) is taking into account these minor phases and hence will give from all the methods considered the closest estimate of the parental magma. The transition metals (Ti, Sc, V, Mn, Co and Ni) and the other rare-earth elements are mainly 19

23 hosted by either cumulus orthopyroxene or intercumulus clinopyroxene and plagioclase. Thus, for these elements the results from Model 3 (taking into account only the in situ mineral analysis) give a good estimate of the parental liquid. By considering the above results, we propose an estimate of the parental liquid from which the Lower and Lower Critical Zones formed (Table 11 and Fig. 15d) by using a combination of results from both Method 2 (for the most incompatible elements) and 3 (for the remainder). By considering this, the parental magma inferred from our study (Fig. 15f and Table 10) is similar to the average values obtained from the analysis of B1-rocks (Barnes et al., in press; Curl, 2001) and consistent with previous suggestions (Davies and Tredoux, 1985; Harmer and Sharpe, 1985; Maier and Barnes, 1998). 9. Role of trapped liquid and post-cumulus processes Cumulate rocks are composed of a mixture of cumulus minerals and various proportion of interstitial liquid trapped in the interstices of the crystal framework. The trapped liquid equilibrate to various degree with the cumulus phases to produce postcumulus overgrowths and to crystallize as intercumulus phases. If one consider a perfect closed-system where equilibrium crystallization occurs, the cumulus minerals will perfectly equilibrate with the trapped liquid to produce unzoned minerals. In contrast, if the trapped liquid undergoes closed system fractional crystallization, the cumulus minerals will not be in equilibrium with the trapped melt resulting in the formation of zoned minerals with homogenous cores and zoned post-cumulus overgrowths. In addition, intercumulus liquid may migrate through the crystal pile during compositional convection (Morse, 1986; Tait et al., 1984), infiltration metasomatism (Irvine, 1980), compaction of the partially molten rocks (Boorman et al., 2004; Mathez, 1995b; Mathez et al., 1997; McKenzie, 1984; Meurer and Boudreau, 1998), or magma chamber instabilities. As a result, cumulus minerals may be in contact with more fractionated melt (expected to be enriched in incompatible elements) which in turn will 20

24 modify their initial composition and results in the formation of chemical zonation in minerals reflecting the variation in element diffusivities. Overall, our results show that cumulus and intercumulus minerals have relatively constant compositions in terms of major and trace elements over the entire Lower and Lower Critical zones suggesting they follow a similar crystallization path or evolution history. This observation is consistent with the constant parental liquid composition (similar to B1-magma) calculated from the different whole-rock and in-situ cumulus and intercumulus compositions throughout the stratigraphic interval examined (Table 11 and Fig.15). The presence of high-temperature deformation features in orthopyroxene, and deformation twins in interstitial plagioclase oriented parallel to the paleovertical direction indicate that compaction has affected the rocks within a range of temperature on cooling. The presence of spindle twins in plagioclase indicates that the deformation occurred in a solid state and that compaction may have not been efficient enough to trigger extensive interstitial melt migration. Dihedral angles between orthopyroxene orthopyroxene-plagioclase or orthopyroxene clinopyroxene are in most cases <50 degree (Fig. 3). No angle modification due to sub-solidus processes were observed (Holness et al., 2006; Holness, 2006; Holness et al., 2007; Holness et al., 2005). The steady increases in Ti concentrations in interstitial plagioclase with decreasing An content observed in all the samples considered is consistent with in situ fractional crystallization of the intercumulus minerals without significant melt migration (Humphreys, 2009). Consequently, all the above chemical and textural arguments indicate that although evidence of compaction is present, no significant intercumulus liquid migration occurred in our samples suggesting that the assumption of a closed system is correct as proposed in previous studies (Cawthorn, 1999; Cawthorn et al., 1992). Comparison of whole-rock chemistry with mineral chemistry indicate for some samples decoupling of the major and trace elements(fig. 16a). The whole-rock Mg # is lower than the 21

25 Mg# of the orthopyroxene present in the same sample suggesting that the whole-rock Mg# is primarily controlled by the trapped liquid and uncomplete equilibration of cumulus and postcumulus component occured. Chromium concentrations ranges in orthopyroxenes (Fig. 16b) are relatively constant with decreasing Mg# and may reflect the temperature dependence of the Cr partition coefficient into orthopyroxene (Barnes, 1986a). However, in some cases (samples NG or NG , Fig. 16b), the Cr concentrations in orthopyroxene for a given Mg# is lower than that expected by considering the fractional crystallization of the B1- magma. This observations is consistent with the decoupling of major elements illustrated in Figure 16a for these samples and would suggest the chemical variability reflect a trapped liquid shift rather than a change in magma composition upon fractional crystallization Orthopyroxene and clinopyroxene compositions are homogeneous at the mineral scale in terms of major elements and most trace elements. However, in both minerals Ti concentrations are higher at the margins of the crystals (Fig. 7 and 9). The Ti zonations in pyroxenes may be the result of several processes, including: (i) changes in magma composition; (ii) interaction with relatively evolved intercumulus melt; or (iii) subsolidus processes. Textural and chemical data indicate that no significant interstitial melt migration has occurred implying that the high Ti concentrations observed at the margins of pyroxene crystals are not due to the percolation of fractionated melt during cumulate crystallization. Comparison of the Ti concentrations at the margins of orthopyroxenes in contact with various cumulus (i.e. other orthopyroxene or olivine) and intercumulus (i.e clinopyroxene and plagioclase) minerals indicates that there is no correlation between the Ti concentrations and the type of mineral in contact with them. These observations suggest that Ti did not diffuse from one particular mineral to the other and rather would be due to crystallization processes. Experimental work (Cherniak and Liang, 2008) has shown that Ti diffusivities in enstatite are slower than those of Fe-Mg and Cr (Ganguly et al., 2007). This diffusivity variation may 22

26 reflect the interplay of cation size and charge, or the substitution of Ti in the tetrahedral site. The lack of Fe-Mg and Cr zonation in the pyroxenes analysed in the present study would suggest that the Ti zonation is due to due disequilibrium crystallization of the trapped melt during closed system fractional crystallization. This conclusion is consistent with the steady increases in Ti concentrations in interstitial plagioclase with decreasing An content as described above Conclusions Major and trace element concentrations in the whole-rock and cumulus and intercumulus minerals from the Lower and the Lower Critical zones of the Bushveld Complex provide insight on processes and parental liquids from which the cumulates formed. Whole-rock trace element concentration patterns are parallel to B1-magma consistent with the presence of a B1 liquid fraction in the cumulate. Cumulus olivine and orthopyroxene are unzoned in terms of major and trace elements. Only Ti shows a chemical zonation in pyroxene which is interpreted to be the result of mineral scale subsolidus re-equilibration on cooling. The absence of significant changes in major and trace element compositions of cumulus and intercumulus minerals throughout the Lower and the Lower Critical Zones indicates that the minerals formed from magma with a relatively constant composition, and without significant melt migration, as confirmed by textural evidence. The mass balance between mineral and whole rock data indicates that most of the transition metals (Ti, Sc, V, Mn, Co and Ni) and the HREE are present in orthopyroxene and to a lesser extent in clinopyroxene whereas the LREE are mainly controlled by plagioclase. The balance of the elements is present in accessory phases such as monazite, apatite and zirconium-rich minerals. Calculation, based whole-rock and in-situ cumulus and intercumulus mineral chemistry, of the trace element concentrations of the parental liquid from which the cumulates of the Lower Zone and the Lower Critical 23

27 Zone have formed indicate that the rocks crystallized from a magma similar in composition to B1-type magma. This conclusion is similar to that drawn by Barnes et al. (in press) based on the whole-rock chemistry of the chilled margins, sills and dykes within the footwall of the Bushveld complex therefore suggesting that the analysed B1-type sills are representative of the magma from which the Lower and Lower Critical zones have crystallized. In addition, our results illustrate that the simple calculation of the parental magma by considering a cumulus mineral in equilibrium with a magma (as defined by partition coefficient relationship) is solely applicable to pure adcumulate rocks where trapped liquid is absent Acknowledgments Dr. Richard Cox, Nancy Lafrance and Dany Savard are acknowledged for their assistance with LA-ICP-MS analysis. Dr. Marc Choquette is thanked for his assistance with microprobe analysis. Drs. Brian O Driscoll and Edmond Mathez are acknowledged for comments on an early version of the present work. Stephen Barnes is gratefully thanked for his useful discussion which improve the quality and focus of this manuscript. The analytical work was funded by a Discovery Grant from Natural Science and Engineering Research Council of Canada and the Canadian Research Chair in Magmatic Metallogeny to SJB. Bélinda Godel is funded by the CSIRO Office of the Chief Executive Post-Doctoral Fellowship scheme. This paper is a contribution from the CSIRO Minerals Down Under National Research Flagship. References Albarède, F., Introduction to geochemical modelling. Cambridge University Press. 24

28 Arndt, N., Jenner, G., Ohnenstetter, M., Deloule, E., Wilson, A., Trace elements in the Merensky Reef and adjacent norites Bushveld Complex South Africa. Mineralium Deposita 40, Atkins, F.B., Pyroxenes of the Bushveld Intrusion, South Africa. J. Petrology 10, Barnes, S.-J., Maier, W.D., Platinum-group elements and microstructures of Normal Merensky Reef from Impala Platinum Mines, Bushveld Complex. Journal of Petrology 43, Barnes, S.-J., Maier, W.D., Curl, E.A., in press. Composition of the marginal rocks and sills of the Rustenburg Layered Suite, Bushveld Complex, South Africa: implications for the formation of the Platinum-Group Element Deposits. Economic Geology. Barnes, S.J., 1986a. The distribution of chromium among orthopyroxene, spinel and silicate liquid at atmospheric pressure. Geochimica et Cosmochimica Acta 50, Barnes, S.J., 1986b. The effect of trapped liquid crystallization on cumulus mineral compositions in layered intrusions. Contributions to Mineralogy and Petrology 93, Barnes, S.J., Maier, W.D., Ashwal, L.D., Platinum-group element distribution in the main zone and upper zone of the Bushveld Complex, South Africa. Chemical Geology 208, Bedard, J.H., Trace element partitioning in plagioclase feldspar. Geochimica and Cosmochimica Acta 70, Bédard, J.H., 1994a. A procedure for calculating the equilibrium distribution of trace elements among minerals of cumulate rocks and the concentration of trace elements in the coexisting liquids. Chemical Geology 118,

29 Bédard, J.H., 1994b. A procedure for calculating the equilibrium distribution of trace elements among the minerals of cumulate rocks, and the concentration of trace elements in the coexisting liquids. Chemical Geology 118, Bédard, J.H., Parental magmas of the Nain Plutonic Suite anorthosites and mafic cumulates: a trace element modelling approach. Contributions to Mineralogy and Petrology 141, Bédard, J.H., Partitioning coefficients between olivine and silicate melts. Lithos 83, Bédard, J.H., Trace element partitioning coefficients between silicate melts and orthopyroxene: Parameterizations of D variations. Chemical Geology 244, Blundy, J.D., Wood, B.J., Partitionning of trace elements between crystals and melts. Earth and Planetary Sciences Letters 210, Boorman, S., Boudreau, A., Kruger, F.J., The Lower Zone-Critical Zone transition of the Bushveld Complex: A quantitative textural study. Journal of Petrology 45, Cameron, E.M., The Lower Zone of the eastern Bushveld complex in the Olifants River trough. Journal of Petrology 19, Cawthorn, R.G., Magma addition and possible decoupling of major- and trace-element behaviour in the Bushveld Complex, South Africa. Chemical Geology 39, Cawthorn, R.G., Models for incompatible trace-element abundances in cumulus minerals and their application to plagioclase and pyroxene compositions in the Bushveld Complex. Contributions to Mineralogy and Petrology 123, Cawthorn, R.G., Permeability of the footwall cumulates to the Merensky Reef, Bushveld Complex. South African Journal of Geology 102, Cawthorn, R.G., Cr and Sr: Keys to parental magmas and processes in the Bushveld Complex, South Africa. Lithos. 26

30 Cawthorn, R.G., Davies, G., Clubley-Armstrong, A., McCarthy, T.S., Sills associated with the Bushveld Complex, South Africa: an estimate of the parental magma composition. Lithos 14, Cawthorn, R.G., Sander, B.K., Jones, I.M., Evidence for the trapped liquid shift effect in the Mount Ayliff Intrusion, South Africa. Contributions to Mineralogy and Petrology 111, Charlier, B., Vander Auwera, J., Duchesne, J.-C., Geochemistry of cumulates from the Bjerkreim-Sokndal layered intrusion (S. Norway): Part II. REE and the trapped liquid fraction. Lithos 83, Cherniak, D.J., Liang, Y., Ti diffusion in pyroxene, AGU Fall meeting, San Fransico. Curl, E.A., Parental magmas of the Bushveld Complex, South Africa. Monash University, p Davies, G., Cawthorn, R.G., Mineralogical data on a multiple intrusion in the Rustenburg Layered Suite of the Bushveld Complex. mineralogical Magazine 48, Davies, G., Tredoux, M., The platinum-group element and gold contents of the marginal rocks and sills of the Bushveld Complex. Economic Geology 80, Duchesne, J.C., Charlier, B., Geochemistry of cumulates from the Bjerkreim-Sokndal layered intrusion (S. Norway). Part I: Constraints from major elements on the mechanism of cumulate formation and on the jotunite liquid line of descent. Lithos 83, Eales, H.V., 2000a. Implications of the chromium budget of the Western Limb of the Bushveld Complex. South African Journal of Geology 103, 141. Eales, H.V., 2000b. Implications of the chromium budget of the Western Limb of the Bushveld Complex. South African Journal of Geology 103,

31 Eales, H.V., B., T., Maier, W.D., 1993a. Cryptic variations of minor elements Al, Cr, Ti and Mn in Lower and Critical Zone orthopyroxenes of the Western Bushveld Complex. Mineralogical Magazine 57, Eales, H.V., Botha, W.J., Hattingh, P.J., de Klerk, W.J., Maier, W.D., Odgers, A.T.R., 1993b. The mafic rocks of the Bushveld complex: a review of emplacement and crystallization history, and mineralization, in the light of recent data. Journal of African Earth Sciences 16, Eales, H.V., Cawthorn, R.G., The Bushveld Complex, in: Cawthorn, R.G. (Ed.), Layered Intrusions. Elsevier, Amsterdam, The Nederlands, pp Eales, H.V., de Klerk, W.J., Teigler, B., Evidence for magma mixing processes within the Critical and Lower Zones of the northwestern Bushveld Complex, South Africa. Chemical Geology 88, Gaetani, G.A., Grove, T.L., Partitioning of moderately siderophile elements among olivine, silicate melt, and sulfide melt: Constraints on core formation in the Earth and Mars. Geochimica et Cosmochimica Acta 61, Ganguly, J., Ito, M., Zhang, X., Cr diffusion in orthopyroxene: Experimental 745 determination, 53 Mn 53 Cr thermochronology, and planetary applications. Geochemica Cosmochimica Acta 71, Godel, B., Barnes, S.-J., Maier, W.D., D distribution of sulphide minerals in the Merensky Reef (Bushveld Complex, South Africa) and the J-M Reef (Stillwater Complex, USA) and their relationship to microstructures using X-Ray Computed Tomography. Journal of Petrology 47, Godel, B., Seat, Z., Maier, W.D., Barnes, S.-J., submitted. The Nebo-Babel Ni-Cu-PGE Sulfide Deposit (West Musgrave Block, Australia): Part 2- Mafic dyke chemistry with Constraints on Parental Magmas and Processes. 28

32 Harmer, R.E., Armstrong, R.A., Duration of Bushveld Complex (seno lato) magmatism: Constraints from new SHRIMP zircon chronology., Workshop on the Bushveld Complex, Gethane Lodge, Burgersfort. Harmer, R.E., Sharpe, M.R., Field relations and strontium isotope systematic of the marginal rocks of the eastern Bushveld Complex. Economic Geology 80, Holness, M., Nielsen, T.F.D., Tegner, C., Textural maturity of cumulates: a record of chamber filling, liquidus assemblage, cooling rate and large-scale convection in mafic layered intrusions. Journal of Petrology. Holness, M.B., Melt-Solid Dihedral Angles of Common Minerals in Natural Rocks. Journal of Petrology, Advance Access. Holness, M.B., Andesron, A.T., Martin, V.M., MacLennan, J., Passmore, E., Schwindinger, K., Textures in partially solidified crystalline nodules: a window into pore structure of slowly cooled mafic intrusions. Journal of Petrology. Holness, M.B., Cheadle, M.J., McKenzie, D., On the use of changes in dihedral angle to decode late-stage textural evolution in cumulates. Journal of Petrology 46, Hulbert, L.J., Von Gruenewaldt, G., Textural and compositional features of chromite in the lower and critical zones of the Bushveld Complex south of Potgietersrus. Economic Geology 80, Humphreys, M.C.S., Chemical Evolution of Intercumulus Liquid, as Recorded in Plagioclase Overgrowth Rims from the Skaergaard Intrusion. Journal of Petrology 50, Irvine, T.N., Magmatic infiltration metasomatism, double diffusive fractional crystallization, and adcumulus growth in the Muskox intrusion and other layered intrusions, in: Hargraves, R.B. (Ed.). Princeton University Press, pp

33 Lambert, D.D., Simmons, E.C., Magma evolution in the Stillwater Complex, Montana: I. Rare-earth element evidence for the formation of the ultramafic series. American Journal Of Science 287, Latypov, R., Chistyakova, S., Alapieti, T., Revisiting problem of chilled margins associated with marginal reversals in mafic-ultramafic intrusive bodies. Lithos 99, Maier, W., Barnes, S.-J., Platinum-group elements in the Boulder Bed, western Bushveld Complex, South Africa. Mineralium Deposita 38, Maier, W.D., Arndt, N.T., Curl, E.A., Progressive crustal contamination of the Bushveld Complex: evidence from Nd isotopic analyses of the cumulate rocks. Contributions to Mineralogy and Petrology 140, Maier, W.D., Barnes, S.-J., 1999a. Platinum-group elements in silicate rocks of the lower, critical, and main zones at Union Section, western Bushveld Complex. Journal of Petrology 40, Maier, W.D., Barnes, S.J., Concentrations of rare earth elements in silicate rocks of the Lower, Critical and Main Zones of the Bushveld Complex. Chemical Geology 150, Maier, W.D., Barnes, S.J., 1999b. Platinum-group elements in silicate rocks of the Lower, Critical and Main Zones at union section, western Bushveld Complex. Journal of Petrology 40, Marsh, B.D., Solidification fronts ans magmatic evolution. Mineralogical Magazine 60, Mathez, E.A., 1995a. Magmatic metasomatism and formation of the Merensky Reef, Bushveld Complex. Contributions to Mineralogy and Petrology 119, Mathez, E.A., 1995b. Magmatic metasomatism and formation of the Merensky reef, Bushveld Complex. Contributions to Mineralogy and Petrology 119,

34 Mathez, E.A., Hunter, R.H., Kinzler, R., Petrologic evolution of partially molten cumulate: the Atok section of the Bushveld Complex. Contribution to Mineralogy and Petrology 129, McBirney, A.R., Mechanisms of differentiation in the Skaergaard Intrusion. Journal of the Geological Society of London 152, McDade, P., Blundy, J.D., Wood, B.J., Trace element partitioning on the TinaquilloLherzolite solidus at 1,5 GPa. Physics of the Earth and Planetary Interiors 139. McDonough, W.F., Sun, S.S., The composition of the Earth. Chemical Geology 120, McKenzie, D., The generation and compaction of partially molten rock. Journal of Petrology 25, Meurer, W.P., Boudreau, A.E., Compaction of igneous cumulates Part I: Geochemical consequences for cumulates and liquid fractionation trends. Journal of Geology 106, Morse, S.A., Convection in aid of adcumulus growth. Journal of Petrology 27, Norman, M., Garcia, M.O., Pietruszka, A.J., Trace-element distribution coefficients for pyroxenes, plagioclase, and olivine in evolved tholeiites from the 1955 eruption of Kilauea Volcano, Hawai'i, and petrogenesis of differentiated rift-zone lavas. American Mineralogist 90, Scoates, J.S., Friedman, R.M., Precise age of the platiniferous Merensky Reef, Bushveld Complex, South Africa, by the U-Pb zircon chemical ID-Tims abrasion technique. Economic Geology 103, Seat, Z., Gee, M.A.M., Grguric, B.A., Beresford, S.W., Grassineau, N.V., Submitted. The Nebo-Babel Ni-Cu-PGE Sulfide Deposit (Western Musgrave, Australia): Part 1- U/Pb Zircon 31

35 Ages, Whole-Rock Chemistry, mineral and O-Sm/Nd-Rb/Sr isotopic compositions of the intrusion with constraints on petrogenesis. Sharpe, M.R., The chronology of magmas influxes to the eastern compartment of the Bushveld Complex, as exemplified by its marginal border group. Journal of the Geological Society 138, Sharpe, M.R., Hulbert, L.J., Ultramafic sills beneath the eastern Bushveld Complex; mobilized suspensions of early lower zone cumulates in a parental magma with boninitic affinities. Economic Geology 80, South African Committee For Stratigraphy, Stratigraphy of Southern Africa. Part 1. Lithostratigraphy of South Africa, South West / Namibia, and the Republics of the Boputhatswana, Trankei, and Venda., Geological Survey of South Africa Handbook, p Tait, S.R., Hupper, H.E., Sparks, R.S.J., The role of compositional convection in the formation of adcumulate rocks.. Lithos 17, Teigler, B., 1990a. Mineralogy, petrology and geochemistry of the Lower and Lower Critical Zoens, Northwestern Bushveld Complex., Rhodes University, p Teigler, B., 1990b. Mineralogy, petrology and geochemistry of the Lower and Lower Critical Zones, Northwestern Bushveld Complex. Unpublish, Ph.D. Thesis. Rhodes University, Grahamstown, p. 247 pp. Teigler, B., Eales, H.V., The Lower and Critical Zones of the western limb of the Bushveld Complex as intersected by the Nooitgedacht boreholes. Geological Survey of South Africa Bulletin 111, 126. Tollari, N., Barnes, S.J., Cox, R.A., Nabil, H., Trace element concentrations in apatites from the Sept-Îles Intrusive Suite, Canada -- Implications for the genesis of nelsonites. Chemical Geology 252,

36 Von Gruenewaldt, G., Platinum-group element-chromitite associations in the Bushveld Complex. Economic Geology 81, Wager, L.R., Brown, G.M., Layered igneous rocks. Oliver and Boyd, Edinburgh. Zingg, A.J., Recrystallization and the origin of layering in the Bushveld Complex. Lithos 37,

37 875 Figure captions Fig. 1. Simplified geology of the Western Limb of the Bushveld Complex (modified after Godel et al., 2007) showing the locations of the drillholes NG1, NG2 and NG3. The locations of the drillholes are from Maier et al. (1999). Fig. 2. Simplified stratigraphy of the Lower Zone and Lower Critical Zones of the Bushveld Complex. (a) Stratigraphic position of the sample analysed in the present study. (b) Whole-rock Mg#. The diamonds represent the samples from this study whereas the circles represent values given in Maier et al. (submitted). (c) Mg# of orthopyroxenes. (d) Mg# of clinopyroxenes (e) An# of plagioclases. (f) Modal percentage of intercumulus minerals (%IM). Abbreviations: Harz. harzburgite, Pyro. pyroxenite, Dun. Dunite and Chr. chromitite Fig. 3. Photomicrographs showing rock textures of cumulates from the Lower and Lower Critical Zones of the Bushveld Complex. (a) Pyroxenite (sample NG ). (b), (c) and (d) Pyroxenite with examples of undeformed interstitial plagioclase (sample NG1-25, NG and NG ). (e) Example of orthopyroxene crystals enclosed by a second generation of orthopyroxene (both with similar major and trace element compositions) in pyroxenite (sample NG ). Clinopyroxene occur as oikocryst between orthopyroxene. 34

38 (f) Euhedral chromite crystals located at the contacts between silicate crystals (sample NG ) (g) Compositional zonation in plagioclase. (h) Small clinopyroxene (CPX) exsolutions along cleavages in orthopyroxene (OPX). Abbreviations: OPX orthopyroxene, CPX clinopyroxene, PLG plagioclase, OL olivine and, CHR chromite. Fig. 4. Digitalized thin section showing the rock texture of pyroxenites from the Lower and Lower Critical Zones of the Bushveld Complex. (a) Example (samplesng ) showing the alignment of large orthopyroxene crystals perpendicular to the direction of the paleovertical (black arrow). (b) Example (samplesng1-25) showing clinopyroxene oikocrysts distribution. (c) Rose diagrams showing the orientation of the long axis of orthopyroxene in the samples considered. The line defines the paleovertical direction. n is the number of crystals considered to compute the statistics. Fig. 5. Plots of (a) MgO vs SiO 2 and (b) Al 2 O 3 vs SiO 2 for the whole-rocks of the Lower and Lower Critical Zones from the Bushveld Complex. The mineral compositions are average values from the microprobe data. The average B1- magma value is from Curl (2001). Fig. 6. Whole-rock mantle-normalized trace-element patterns of the cumulate rocks from the Lower Critical Zone (A) and the Lower Zone(b). The ranges of the composition for the B1 and B2 magmas are from Curl (2001)

39 Fig. 7. Variation between core and margin major and trace element compositions within cumulus orthopyroxene s from the Lower and Lower Critical Zones from the Bushveld Complex Fig. 8. Mantle-normalized trace element patterns of the cumulus orthopyroxenes analysed from cumulate rocks from the Lower Zone (A) and the Lower Critical Zone (B) from the Bushveld Complex. The mantle concentrations are from McDonough and Sun (1995) Fig. 9. Variation between core and margin major and trace element compositions within intercumulus clinopyroxenes from the Lower and Lower Critical Zones from the Bushveld Complex Fig. 10. Mantle-normalized trace element patterns of the intercumulus clinopyroxenes analysed from cumulate rocks from the Lower Zone (A) and the Lower Critical Zone (B) from the Bushveld Complex. The mantle concentrations are from McDonough and Sun (1995) Fig. 11. Mantle-normalized trace element patterns of the intercumulus plagioclases analysed from cumulate rocks from the Lower Zone (A) and the Lower Critical Zone (B) from the Bushveld Complex. The mantle concentrations are from McDonough and Sun (1995)

40 Fig. 12. Histograms summarizing the proportion of each of the individual trace elements hosted by the various minerals analysed in harzburgite (A) and pyroxenite (B) from the Lower and Lower Critical Zones from the Bushveld Complex. Fig. 13. Time-resolved laser ablation spectras showing the presence of trace element-rich micro-inclusions Fig. 14. Sketch showing the evolution of cumulate rock. (a) Accumulation of cumulus orthopyroxene and crystallization of intercumulus minerals. The Ti compositional zonation in pyroxenes is interpreted to be the result of small scale mineral reequilibration upon cooling rather than being due to the percolation of fractionted (Ti-rich) melt in the intercumulus space. (b) Clinopyroxene (CPX) exsolution from orthopyroxene (OPX) on cooling leading to trace element diffusion from orthopyroxene to clinopyroxene. Our results indicate these redistribution as negligible effect on parental magma calculation from mineral composition Fig. 15: Mantle normalized trace element patterns of the inferred parental magma from which the cumulate rocks from the Lower Zone (A and C) and Lower Critical Zone (B and C) crystallized. Our results indicate that the rocks crystallized from a magma similar in composition to B1-type magma therefore suggesting that the analysed B1-type sills are representative of the magma from which the Lower and Lower Critical zones have crystallized. The mantle values are from McDonough and Sun (1995) and the average B1-magma composition is from Curl (2001)

41 Fig. 15: a) Mg# in the whole-rock versus average Mg# in orthopyroxene and b) Cr concentration in orthopyroxene versus Mg# plots illustrating the decoupling of major and trace elements as a result of trapped liquid interaction. 38

42 Figure-1 South Africa 25 00' Pretoria A N Union Section NG2 NG1 NG3 B Km Upper Zone Pilanesberg Complex 5 4 Main Zone ' km Impala Pt Rustenburg Pt Rustenburg Pt Marinaka 2 1 Upper Lower Merensky Reef Critical Zone Lower Zone 27 00' 27 30' 0 Marginal Zone

43 Figure A B C D E Chr. NG1-25 F Height (m) L ower Zone Lower Critical Zon e Dun. Pyro. NG NG NG NG NG NG Harz. NG Mg# WR Mg# Mg# An# OPX CPX PLG 0 10 % IM 20

44 Figure-3