| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Journal of Petrology Volume 41 Number 10 Pages 1517-1539 2000
© Oxford University Press 2000
The Halogen Geochemistry of the Bushveld Complex, Republic of South Africa: Implications for Chalcophile Element Distribution in the Lower and Critical Zones
1DIVISION OF EARTH AND OCEAN SCIENCES, BOX 90227, DUKE UNIVERSITY, DURHAM, NC 27708, USA
2HUGH ALLSOPP LABORATORY, UNIVERSITY OF THE WITWATERSRAND, WITS 2050, REPUBLIC OF SOUTH AFRICA
Received October 20, 1999; Revised typescript accepted March 24, 2000
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONGEOLOGY AND PETROLOGYMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Halogen-bearing minerals, especially apatite, are minor but ubiquitous phases throughout the Bushveld Complex. Interstitial apatite is near end-member chlorapatite below the Merensky reef (Lower and Critical Zones) and has increasingly fluorian compositions with increasing structural height above the reef (Main and Upper Zones). Cl/F variations in biotite are more limited owing to crystal-chemical controls on halogen substitution, but are also consistent with a decrease in the Cl/F ratio with structural height in the complex. A detailed section of the upper Lower Zone to the Critical Zone is characterized by an upward decrease in sulfide mode from 0·01–0·1% to trace–0·001%. Cu tends to correlate with other incompatible elements in most samples, whereas the platinum-group elements (PGE) can behave independently, particularly in the Critical Zone. The decrease in the Cl/F ratio of apatite in the Main Zone is associated with a shift to more radiogenic Sr isotopic signature, implying that the unusually Cl-rich Lower and Critical Zones are not due to assimilation of crustal rocks. Nor is the Main Zone more Cl rich where it onlaps the country rocks of the floor, suggesting little if any Cl was introduced by infiltrating country rock fluids. Instead, the results are consistent with other studies that suggest Bushveld volatile components are largely magmatic. This is also supported by apatite–biotite geothermometry, which gives typical equilibrium temperatures of 750°C. The increasingly fluorian apatite with height in the Upper Zone can be explained by volatile saturation and exsolved a Cl-rich volatile phase. The high Cl/F ratio inferred for the Lower and Critical Zone magma(s) and the evidence for volatile saturation during crystallization of the Upper Zone indicate the Lower and Critical Zones magma(s) were unusually volatile rich and could easily have separated a Cl-rich fluid phase during solidification of the interstitial liquid. The stratigraphic distribution of S, Cu and the PGE in the Critical Zone cannot readily be explained either by precipitation of sulfide as a cotectic phase or as a function of trapped liquid abundance. Evidence from potholes and the PGE-rich Driekop pipe of the Bushveld Complex imply that migrating Cl-rich fluids mobilized the base and precious metal sulfides. We suggest that the distribution of sulfide minerals and the chalcophile elements in the Lower and Critical Zones reflects a general process of vapor refining and chromatographic separation of these elements during the evolution and migration of a metalliferous, Cl-rich fluid phase.
KEY WORDS: Bushveld Complex; chlorine; platinum-group elements; layered intrusions
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONGEOLOGY AND PETROLOGYMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
The effects of volatile components in igneous systems are well documented. During partial melting, the addition of water lowers the solidus of the source rock (e.g. Burnham & Davis, 1974
), which can change both the major and trace element chemistry of the partial melt (e.g. Hickey & Frey, 1982; Iwanori, 1998). Further, volatile elements can alter the anhydrous crystallization sequence and result in features such as pegmatites by suppressing crystal nucleation (e.g. Hort, 1998).
In layered intrusions, the separation of a volatile-rich phase (or fluid) from evolved magma in the interstices between primary crystals (the interstitial or trapped liquid) can also have dramatic effects on the resultant chemistry and textures of the rocks. For example, separation and migration of a fluid phase can alter the primary igneous chemistry by metasomatism and the promotion of recrystallization of the primary igneous assemblage (e.g. Mathez, 1995; Meurer & Boudreau, 1998a, 1998b; Meurer et al., 1999), and form features such as discordant, pegmatoidal replacement bodies (e.g. Schiffries, 1982) and potholes (e.g. Ballhaus, 1988; Boudreau, 1992). In addition, if the fluid phase contains a strong complexing agent such as Cl, it can redistribute and concentrate base and precious metal sulfides (e.g. Schiffries, 1982; Candela & Holland, 1984; Ballhaus & Stumpfl, 1986; Boudreau & McCallum, 1992; Boudreau & Meurer, 1999).
Previous work has suggested that the Bushveld magma(s) parental to the Lower and Critical Zones contained significant Cl and that the Cl-bearing fluids could have been an important means of redistribution of precious metal sulfides. Boudreau & Kruger (1990) presented analyses of a short stratigraphic section that indicate that the Critical Zone contained Cl-rich interstitial apatite. Further, Schiffries (1982) has shown that Cl-rich fluids were involved in the anomalous concentration of the platinum-group elements (PGE), and particularly Pt and Pd, in the Driekop iron-rich ultramafic pegmatite. This study adds to these previous studies by addressing two major questions: (1) what is the halogen composition of hydrous minerals throughout the Bushveld Complex? (2) Is there evidence of fluid interaction with either the crystal pile or resident magma that could have redistributed base and precious metals and sulfides in the Lower and Critical Zone rocks as suggested by Boudreau & Meurer (1999) and contribute to the formation of the PGE-rich upper Group 2 (UG-2) chromitite, the Merensky reef, and anomalous enrichments in other chromitites of the Critical Zone (e.g. Scoon & Teigler, 1994)?
|
GEOLOGY AND PETROLOGY |
|---|
|
TOPABSTRACTINTRODUCTIONGEOLOGY AND PETROLOGYMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Geologic setting
The Bushveld Complex is a 2060 ± 3 Ma (Kruger et al., 1986
) layered mafic–ultramafic intrusion that covers >66 000 km2 of the northern Kaapvaal craton of the Republic of South Africa (Fig. 1). The magmas forming the Bushveld Complex were intruded into the Pretoria group, a succession of quartzitic to arkosic sandstone and mudstone with interlayered andesitic lavas and volcanoclastic sediments (Eriksson et al., 1991), broadly along the interface with the overlying Rooiberg group felsic volcanic rocks. The Pretoria group was deposited within the half-graben Transvaal basin between c. 2·3 and 2·1 Ga (Schreiber et al., 1992b) in a lacustrine environment. Fresh-water deposition of the Pretoria group is inferred from sedimentary structures (Eriksson et al., 1991) and the boron content of shale (Schreiber et al., 1992a; Eriksson et al., 1996).
|
The eruption and intrusion of the units of the Bushveld magmatic province [which includes the Bushveld Complex (Coetzee & Kruger, 1989)] are not fully understood. However, it is known that the igneous suite is the result of at least three distinct episodes of magmatism. The first was the eruption of the 2061 ± 2 Ma Rooiberg felsites [(Walraven, 1997) zircon evaporation data]. These are composed of >300 000 km3 of silicic ash flows and pyroclastic deposits (Twist, 1985). The felsic volcanic rocks that form the roof of the Bushveld Complex are thought to be genetically related to the calc-alkaline Dullstroom volcanic rocks overlying the Pretoria group (Cheney & Twist, 1991).
The second major igneous event was the intrusion of the mafic Bushveld magmas along the interface between the Pretoria group and the Rooiberg felsites to form the Bushveld Complex. There were at least three and possibly four identifiable mafic liquids that crystallized to form the Bushveld Complex, based upon a Sr isotopic shift between the Critical and Main Zones and the Main and Upper Zones (Kruger & Marsh, 1982; Kruger et al., 1986).
The third and final event was the intrusion of the 2040 ± 8 Ma Bushveld (Nebo) granite (Walraven, 1988). The granite was intruded through the mafic complex and into and along the boundary between the Upper Zone and the overlying Rooiberg felsic volcanic and granophyric rocks. These three Bushveld magmatic events are clearly temporally and spatially related, but their genetic relations are speculative.
Stratigraphy
The rocks of the Bushveld Complex form a funnel-shaped sill, which crosscuts the Pretoria group along the boundary noted above. The intrusion is 
7 km thick, although the thickness varies laterally. The layered series is divided into a stratigraphy based largely on modal mineralogy. The primary divisions are the Lower Zone, Critical Zone, Main Zone, and the Upper Zone.
The Lower Zone is a succession of chemically relatively homogeneous dunite, harzburgite, and orthopyroxenite, with grain sizes on the order of 1 mm. The interstices are dominated by pyroxene and minor plagioclase, but apatite, biotite and sulfides are also present. The base of the overlying Critical Zone is marked by a increase in the modal abundance of interstitial plagioclase from <1 to 4%.
The Lower Critical Zone is composed largely of feldspathic orthopyroxenites, with some harzburgite, whereas the Upper Critical Zone is composed of feldspathic orthopyroxenite, melano- and leuconorite and anorthosite. Chromitite layers occur in both subzones and range between 20 cm and 2 m in thickness, and may contain economic concentrations of base and precious metals. Apatite and biotite continue to be interstitial phases, quartz becomes more abundant, but sulfides are less abundant. Visual estimates suggest an increase in the abundance of accessory rutile in the Critical Zone, both as an interstitial mineral and included in primocryst minerals.
The Merensky cyclic unit has been described in great detail by other workers (e.g. Cousins, 1969; Vermaak, 1976; Kruger & Marsh, 1985; Mathez, 1995; Wilson et al., 1999). The base of the cyclic unit is the Merensky pegmatoid (Merensky reef). The pegmatoid itself is a pegmatitic feldspathic pyroxenite of 0–50 cm thickness, typically sandwiched between thin (1–2 cm) lower and upper chromitite layers. The crystals of the pegmatoid are considerably coarser than the 1 mm typical of most of the complex, being commonly greater than 1–2 cm in long dimension, and typically subhedral to anhedral. The interstitial areas contain abundant PGE-bearing sulfides, biotite, and apatite. The top chromitite stringer is laterally inconsistent, and chromite is in places disseminated. The Merensky orthopyroxenite overlies the chromitite layer, and grades into the Merensky norite, which is itself capped by the Merensky anorthosite. The Merensky cyclic unit is overlain by the Bastard cyclic unit, which has much the same character as the Merensky unit, but is poorer in PGE and is usually much thicker, and pegmatoid is seldom developed.
The boundary between the Critical and Main Zones is conventionally taken as the top of the Giant Mottled Anorthosite, which caps the Bastard cyclic unit (e.g. Eales & Cawthorn, 1996). However, based upon Sr-isotope changes and other evidence, including a major unconformity (see, e.g. Carr et al., 1999), Kruger (1990) suggested placing the boundary at the base of the Merensky cyclic unit, and this convention is adopted here. Above the Bastard cyclic unit the Main Zone is generally uniform, composed of equigranular gabbronorite with anorthositic horizons. The interstitial areas, unlike those of the Lower and Critical Zones, contain only rare biotite and apatite, the interstitial areas being filled dominantly by minor quartz. The Main Zone is also characterized by a more radiogenic initial 87Sr/86Sr ratio (0·7085) than the underlying rocks (0·7064). However, a shift back to less radiogenic initial 87Sr/86Sr ratios (Sharpe, 1985) and an associated Mg/Fe reversal at the pyroxenite marker at the top of the Main Zone imply a major unconformity or replenishment event at this level (Kruger et al., 1986).
The base of the Upper Zone is conventionally defined by the first appearance of euhedral (cumulus) magnetite, but, as shown by Kruger (1990), the pyroxenite marker is the petrological base of the Upper Zone and is also associated with an unconformity related to magma injection [see Cawthorn et al. (1991)]. Hence, the section between the pyroxenite marker and the appearance of magnetite is included in Subzone A of the Upper Zone. The overlying stratigraphy is a series of gabbronorite, magnetite gabbronorite and economic (vanadiferous) magnetitite layers. Biotite returns as a relatively abundant interstitial phase, and apatite becomes a cumulus phase in Upper Zone C and is particularly abundant in the lower part of this subzone. Although variable, apatite mode decreases in abundance toward the top of the intrusion, consistent with a decrease in the bulk P contents, which may exceed 4 wt % P2O5 near the base of Upper Zone C, to typically <0·4% just under the roof [compare summary of Eales & Cawthorn (1996)].
The stratigraphic units of the layered series can have a complex association of marginal rocks and sills (the Marginal Zone) where the layered series onlaps the country rocks along the floor of the intrusion. For example, the B1 and B2 sills intrude into the floor rocks adjacent to the Lower and Critical Zones, respectively. The sills vary greatly in thickness, from tens of meters to >800 m, and range in composition from homogeneous peridotite to differentiated bodies of peridotite, harzburgite, orthopyroxenite and norite (Sharpe & Hulbert, 1985). There are also a small number of quench-textured micropyroxenites. Many of the sills have quenched margins that contain devitrified glass, and smaller sills can have devitrified glass throughout (Sharpe & Hulbert, 1985). For additional description of the Marginal Zone rocks, the reader is referred to Cawthorn et al. (1981) and Harmer & Sharpe (1985). Sill equivalents to the Main Zone magmas have not yet been unequivocally identified.
Sample location and selection
Most samples of this study are typical rocks from the layered suite, with the addition of a Lower Zone ‘chilled’ marginal sample (No. 1581), and two samples from the heterogeneous Marginal Zone where the lower Main Zone transgresses the floor (Nos 1544 and 1548). The latter two are both nominally B2 (Critical Zone equivalents) as defined by Sharpe (1981), although G. Cawthorn (personal communication, 1996) believes they are more likely to be related to the Main Zone, based on texture, low Cr contents, and proximity of Main Zone cumulates.
An extensive set of samples were taken from the ‘Cameron section’ of the Eastern Bushveld (Cameron, 1978) (Fig. 1) for additional whole-rock analysis. The samples cover a stratigraphic interval of 1900 m from above the lowest orthopyroxenite layer to the anorthosite below the Merensky reef. Merensky pegmatoid samples were taken from drill core provided by the Rustenburg Platinum Mines. All field samples were taken from locations away from obvious fractures and alteration veins, and were selected based on apparent freshness in hand specimen.
Petrology of late-crystallizing minerals
Chromite, plagioclase, quartz, and clinopyroxene (cpx) are the most common minor and interstitial phases in the Lower and Critical Zones. Except for chromite, these minerals are typically anhedral to subhedral with crystal shapes controlled by the primocryst minerals.
Apatite is the principal mineral of interest as it is the most commonly occurring halogen-bearing mineral and, unlike the micas and amphiboles, halogen substitution is ideal at high temperatures (e.g. Volfinger et al., 1985; Tacker & Stormer, 1989). Apatite is a minor but relatively common interstitial and chadocrystic mineral even in the Lower and Critical Zones. Its low modal abundance (a few small grains observed in a typical thin section) is consistent with low whole-rock P2O5 concentrations of 0·01–0·02 wt % that are characteristic of much of the lower 80% of the complex below the Upper Zone (e.g. Cawthorn, 1983). As noted above, where apatite first appears as a primocryst mineral in Upper Zone C, apatite mode is observed to locally exceed 6%, but then decreases in abundance irregularly with structural height until it is again only a minor to trace mineral in a ferrodiorite sample from the top of the complex.
The majority of apatite grains analyzed in this study are interstitial to the primocrysts, and many are associated with late crystallized phases such as quartz and biotite (Fig. 2a and b). Apatite grains in this environment are typically subhedral to anhedral with crystal faces controlled, in part, by the primocryst minerals. Interstitial apatite grains vary in size from <20 to >300 µm.
|
Chadocryst apatite grains, although rare, are typically small,
20 µm in long dimension (Fig. 3). Many of the apatite chadocrysts appear to have formed along previous grain boundaries that have been overgrown by adcumulus crystal growth or during crystal aging.
|
Biotite is a locally abundant interstitial mineral, particularly in the Lower Zone, but overall not as common as apatite. Where biotite is abundant, apatite is usually present as well. In contrast, only a few of the many apatite-bearing samples contain biotite. Amphibole is also locally present as a minor mineral, although in most instances it appears in reaction relationship with pyroxene. Because of its ambiguous textural relationships as to it being a primary (igneous) or secondary mineral, amphibole was not a focus of this study.
Fe–Ni–Cu sulfides are minor to trace phases and are commonly associated with other late-crystallizing minerals. Interstitial sulfides range in size from <10 to >200 µm (Fig. 2). The interstitial sulfides have irregular grain boundaries controlled by the grain boundaries of the silicate phases. Sulfides also occurs as chadocrysts in silicate minerals, where they are typically small, <50 µm in long dimension, and rounded. Petrographic observations suggest that chalcopyrite forms a larger proportion of the Critical Zone sulfide assemblage as compared with the Lower Zone.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONGEOLOGY AND PETROLOGYMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Microprobe analyses were performed on polished thin sections using the Cameca Camebax electron microprobe at Duke University. Typical analytical conditions were 15 keV acceleration voltage, 15 nA beam current, and 5–10 µm beam diameter. Peak and background counting times for Cl were in the range of 20–30 s and 10–15 s, respectively. Other elements were counted on the peak for 30 s and background for 15 s. Standards included natural chlorapatite (RM-1, Morton & Catanzaro, 1964
) and fluorapatite (Wilburforce) for Ca, P and the halogens, allanite (102522, No. 42; Frondel, 1964) for the rare earth elements (REE), and a variety of silicate mineral standards. F and Cl in the apatite standards were checked against fluorite, topaz and halite. Fluorine was analyzed using a synthetic Si–W layered diffracting crystal. OH- and H2O were calculated by difference based on hydroxyl site occupancy.
Potential problems associated with electron microprobe analyses of F and Cl in apatite have been documented by Stormer et al. (1993). Specifically, F and Cl counts are affected by grain orientation, with grains cut parallel to the c-axis showing significant variations with time. Standards are mounted such that count rates on freshly polished surfaces remain approximately constant, and restandardization over the course of 1–2 weeks of a typical analysis run shows no significant change in standard count rates with time.
To test the effect of grain orientation on analyzed F and Cl abundance, 17 grains from an apatite-rich sample from the Upper Zone of the Bushveld Complex, nine cut perpendicular to the c-axis and eight cut parallel to the c-axis, were analyzed by the methods described above. Analysis of nine c-normal grains had an average composition of 2·66 ± 0·14 wt % F and 0·55 ± 0·05 wt % Cl, and eight c-parallel grains had an average composition of 2·52 ± 0·09 wt % F and 0·53 ± 0·04 wt % Cl (Fig. 4). Assuming that the grains are homogeneous (see Discussion), these results suggest that the errors associated with all causes (volatile loss, machine drift, and counting statistics) are approximately twice the expected counting statistic error alone.
|
On a few later analyses, a software upgrade to the Cameca microprobe operating system allowed extrapolation of Cl and F count rate to zero time by fitting a line to 10 consecutive 3 s counting periods. Results using this technique in most cases were within error of those obtained by integrating over a 30 s counting period.
Whole-rock major element compositions were determined by direct-current-plasma optical emission spectroscopy (DCP). The sample preparation and analysis conditions are as described by Klein et al. (1991). Sample solutions were run along with an in-house drift monitor, a procedural blank, and known standards on the ARL Fisons Spectrospan 7 instrument at Duke University. Drift was corrected by forcing the blank analyses to zero and the drift monitor to a constant composition.
Whole-rock trace element compositions were determined by inductively coupled plasma mass spectrometry (ICP-MS) on powdered samples using a VG-Elemental Quadrapole-3 instrument at Duke University. One-step digestion (Meurer et al., 1999) did not dissolve chromite in the samples. Therefore, 40 mg of sample was dissolved by a two-step digestion in a sealed Teflon vial. The powder was initially digested in 1:1:1 solution of HF, HNO3, and HCl. The products of the first digestion were dried and taken up in a 3:10 mixture of HF–HNO3. The products of the final digestion were dried and taken up in a 4:1 mixture of H2O–HNO3. Dilutions of 1:1000 and 1:10 000 were analyzed. Machine operating conditions and data manipulations are as reported by Meurer et al. (1999).
The abundance of sulfide was determined as follows: sulfide grains were located in back-scattered electron images and their setting relative to the cumulus phases was determined, i.e. either interstitial or as chadocrysts in other minerals. The areas of the sulfide grains were then measured from the back-scattered images using the image processing program NIH IMAGE.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONGEOLOGY AND PETROLOGYMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Apatite composition and stratigraphic trends
The composition of apatite with respect to stratigraphic height in the intrusion is shown in Fig. 5. This plot illustrates the trend of increasing F/Cl from Lower, through Critical and Main, to Upper Zone apatite. Calculated OH is generally low (<0·35 mol %) throughout the intrusion. The calculated OH end-member component is negligible where apatite approaches end-member chlorapatite in the lower third of the complex or where apatite is essentially end-member fluorapatite at the top of the complex (Fig. 6).
|
|
In general, apatite below the level of the Merensky unit is uniformly rich in Cl with compositions approaching end-member chlorapatite. The data cluster near 1·0 mole fraction Cl and extend to 0·4 mole fraction OH and 0·6 mole fraction F, with the more F-rich compositions preferentially occurring in the Critical Zone. Because of the variations seen in different grains in individual samples, suggestions of trends in halogens on any finer scale within the Lower and Critical Zones with height need a more detailed sampling interval to be confirmed.
Within the Main Zone, Cl contents of apatite are highest in the pegmatoidal rocks of both the Merensky and Bastard units at the base of the zone. Above these units the Cl/F ratio of apatite falls sharply to more fluorian compositions (Boudreau & Kruger, 1990). The offset in the trends in halogens across the Critical Zone–Main Zone boundary parallels changes observed in major element and Sr isotopic compositions across this boundary (e.g. Kruger & Marsh, 1982; Sharpe, 1985; Kruger, 1994). Apatite composition in the Main Zone is nearly constant, which mimics the relatively homogeneous character of the gabbronorite and norite that make up much of the Main Zone (von Gruenewaldt, 1973; Molyneux, 1974). This isotopic break associated with the pyroxenite marker at the top of the Main Zone is not strongly reflected in a change in apatite compositional trends, if at all.
In the Upper Zone, the highest Cl concentrations occur in interstitial apatite in troctolite of Upper Zone B (Table 1, analysis 7). Once apatite becomes a cumulus phase in the Upper Zone C, Cl begins to decrease and F increases sharply such that the most F-rich, Cl-poor apatite occurs in the apatite-poor ferrodiorite just beneath the roof of the intrusion (Table 1, analysis 9). Indeed, in these rocks there is generally no detectable Cl.
|
Single apatite crystals tend to be homogeneous and unzoned. However, different grains in a sample may show a range in composition, typically expressed as inverse variations in F and Cl (an extreme example is shown in Fig. 7). In these and other rocks, apatite compositions do not appear to be noticeably affected by the presence or absence of biotite, either within individual samples or comparing nearby biotite-bearing and biotite-absent samples.
|
Although the rocks used in this study would be generally described as fresh, some patches of alteration are observed in some samples. Apatite analyses were generally performed on crystals without obvious evidence of alteration such as a mottled appearance in back-scattered electron images, ragged grain boundaries or alteration in adjacent silicates (e.g. epidote or chlorite). However, a few analyses were carried out on obviously affected grains and compared with an analysis of an optically fresh region of the same crystal. In most cases the altered regions have a lower Cl concentration and elevated calculated water content compared with the optically fresh region, a feature also observed in analyses of altered Stillwater apatite (Boudreau & McCallum, 1990). An exception is a single, partially altered grain from the Upper Zone, which was anomalously enriched in Cl (Table 1, analysis 6).
Apatite in marginal rocks has compositions broadly similar to those in the immediately associated layered rocks. Thus, apatite grains from the Lower Zone chill sample are rich in Cl, as is typical of apatite from the Lower and Critical Zones (compare analyses 1 and 2 of Table 1). Apatite grains from the Marginal Zone where the Main Zone onlaps the floor are similar to the more intermediate compositions typical of the Main Zone (compare analyses 3 and 10 of Table 1).
The higher values for the REE (specifically Ce) in apatite are associated with the most evolved rocks that occur in the uppermost parts of Upper Zone C (Table 1, analysis 9). In contrast, there are modestly higher values for the interstitial apatite from the Lower and Lower Critical Zones as compared with those from the Upper Critical and Main Zones. REE are most variable in REE-rich fluorapatite from ferrodiorite just beneath the roof of the complex, and REE–Si correlation (Fig. 8) is consistent with the coupled substitution of REE3+–Si4+ for Ca2+–P5+ as suggested by Rønsbo (1989).
|
Biotite compositions and apatite–biotite geothermometry
Biotite compositions tend to be uniformly poor in both Cl and F throughout the complex, containing <10 mol % of these two components combined (Table 2 and Fig. 9). This is attributed to the strong Fe–F and Mg–Cl exclusions that characterize the micas and amphiboles in general (Munoz & Swenson, 1981; Volfinger et al., 1985). Thus, in the Lower and Critical Zones, where the solids assemblages are relatively Mg rich and the associated apatite is characterized by a high Cl concentration, phlogopitic biotite contains only modest amounts of Cl. Similarly, in the more iron-rich rocks near the top of the intrusion, where apatite approaches end-member fluorapatite compositions, the locally associated annite-rich biotite contains only modest amounts of F.
|
|
Although strongly damped by these avoidance rules, biotite compositional trends nonetheless mimic the decreasing Cl/F ratio with structural position observed in the apatite trends. Thus, biotite associated with chlorapatite from a marginal sample from the Lower Zone is the most Cl rich observed, even though the mole fraction of Cl is only 0·06 (Table 2, analysis 1). Biotite compositions from the Lower and Critical Zones are similar to those reported from the Ultramafic series of the Stillwater Complex (Page & Zientek, 1987). Finally, biotite from individual samples may contain widely variable amounts of Ti (Table 2, analyses 5 and 6).
A formulation of the apatite–biotite F–OH geothermometer that takes account of the Mg/Fe dependence on the incorporation of fluorine in biotite was presented by Zhu & Sverjensky (1992). F–OH exchange between apatite and the phlogopite end-member of biotite can be expressed in the following exchange reaction:
 |
The partition coefficient for the exchange reaction is defined as follows:
 |
into the graphical geothermometer shown in Fig. 10. It should be noted that even with a wide range in halogen compositions in apatite and Fe/Mg ratio of biotite, most samples cluster around 750°C.
|
Whole-rock geochemistry of the Lower and Critical Zones
Whole-rock analyses, pyroxene analyses, and sulfide abundance for the upper part of the Lower Zone and through the Critical Zone to just below the Merensky reef are listed in Table 3 and plotted in Fig. 11. Figure 11a shows the mg-number [molar MgO/(MgO + FeO)] trend with height of orthopyroxene primocryst from this section. In the Critical Zone, the mg-number generally decreases from 0·75 at the base of the Critical Zone to 0·65 below the Merensky reef.
|
|
Most of the major and compatible trace element data correlate positively with the modal mineralogy of the Lower and Critical Zones (Fig. 12). However, P, U, Pb, Th, Cs, Ba, Zr and Y are incompatible elements in the major Bushveld silicate minerals, and as such, are not strongly controlled by the silicate mineralogy. Because incompatible element concentrations are correlated (Fig. 13), only the concentration of P with height is shown in Fig. 11. The concentrations of P and other incompatible elements are low in the Lower Zone and uniformly higher in the Critical Zone.
|
|
Sulfide and the chalcophile elements in the Lower and Critical Zones
Sulfide abundance is highest in the upper Lower Zone and the base of the Critical Zone, and decreases to trace levels through the remainder of the Critical Zone (Fig. 11d). This is consistent with the findings of Maier et al. (1996) and Maier & Barnes (1999), who looked at a more complete section for the Lower Zone from the western Bushveld (Fig. 14). In addition, the data of Maier et al. and Lee & Tredoux (1986) (from a section in the eastern Bushveld, 50 km south of the Cameron section of this study) show an increase by several orders of magnitude in the background concentration of Pd and Pt with stratigraphic height, the highest values occurring in the upper parts of the Critical Zone. No noticeable Pt–Pd enrichment is associated with the S-rich sections of the lower half of the Lower Zone from either the eastern or western Bushveld sections (Fig. 14).
|
There is not a strong correlation between the modal abundance of sulfide minerals and the concentration of Cu in the section covered in this study. Qualitatively, this is supported by visual modal estimates that chalcopyrite is a much more abundant fraction of the Critical Zone sulfide assemblage than in the upper half of the Lower Zone. However, this contrasts with the findings of Maier et al. (1996), who showed that the lower half of the Lower Zone has a higher Cu concentration than the upper Lower and Critical Zones and that Cu and S correlate positively (Fig. 14) in this part of the section.
The distribution of Cu with height in the Lower and Critical Zones (Fig. 11c) has a trend similar to the lithophile incompatible trace elements; the Critical Zone samples are modestly enriched in Cu relative to the Lower Zone. In most samples, there is a good correlation between Cu abundance and incompatible element concentration (Fig. 15).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONGEOLOGY AND PETROLOGYMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Possible causes of halogen variations
Bushveld apatite and biotite show compositional variations both on the scale of an individual thin section and as a function of structural height. We attribute the local scale compositional variation in late crystallizing interstitial minerals to be a function of both the result of small pockets of differentially evolved interstitial liquid produced once the interstitial liquid became saturated in a halogen-bearing phase and a later, local re-equilibration during cooling. The grain-to-grain variability seen in individual samples in halogens and REE concentrations can arise if the interstitial liquid did not crystallize as a bulk equilibrium system but instead underwent significant micro-scale fractionation as individual grains of apatite are isolated from remaining pore liquid at different times.
For example, the Cl–F–Ce correlation seen in Fig. 7 illustrates this process on the thin-section scale. In this case, pore-scale fractionation is interpreted to have led to increases in Cl and the REE as apatite preferentially incorporated the smaller F ion. (It is also noted that if this evolved liquid is displaced upward, perhaps as a result of compaction, or by migration of Cl-rich fluids degassed from these evolved liquids, it will result in even higher Cl/F ratios of overlying rocks. This can explain why the Lower and Critical Zones typically show a more restricted compositional range than is seen in the Marginal Zone apatite of Fig. 7.) In contrast, those samples with relatively abundant apatite primocryst near the base of Upper Zone C are more uniform in composition, consistent with crystallization from a much more homogeneous environment.
The typical temperature calculated by the Zhu & Sverjensky geothermometer, 750°C, is lower than estimated crystallization temperatures of 1000–1050°C for liquidus temperatures at apatite saturation in the Upper Zone (Cawthorn & Walsh, 1988). This implies that that there has been some degree of sub-solidus mobility of the halogens, and also suggests that any compositional zoning, if initially present, could have been lost during re-equilibration.
However, halogen exchange with other halogen-bearing phases during cooling is not the primary control on apatite compositions for several reasons. First, as noted previously, apatite compositional trends do not appear to be a strong function of the presence or absence of biotite either within individual samples or comparing biotite-bearing and biotite-absent samples from the same unit. This is consistent with the observation that apatite is the only mineral that accepts significant F and Cl. Second, apatite compositions are not random but correlate with the major rock types or stratigraphic units of the complex. Third, for temperatures in the range observed, minor adjustment of the F–OH partitioning can give significant variation in the calculated equilibration temperature (although the effect on either mineral is strongly affected by modal proportions).
Equilibration temperatures in the range of 750°C indicate that Bushveld apatite grains were not significantly affected by low-temperature (<750°C) fluids, except those closely associated with local patches of alteration noted previously. This and the broad correlation of halogen compositions with the major stratigraphic units of the Bushveld Complex imply that halogen contents are a primary (igneous) feature of the Bushveld system. Furthermore, the observation that the Main Zone assemblage is not more Cl rich where it onlaps the country rock implies little, if any, Cl was introduced by infiltrating country fluids. This is also consistent with previous studies (e.g. Reid et al., 1993; Mathez et al., 1994) and Cl isotopic data (Boudreau et al., 1999) that imply that Bushveld volatile components are largely magmatic.
Given that the apatite is indeed magmatic, the most prominent features of the Bushveld apatite data are the unusually high Cl/F ratios in the Lower and Critical Zones and the decreasing Cl/F ratio with height above the Merensky reef. Whereas a trend of decreasing Cl/F is common to many intrusions, such as the Munni Munni Complex, Western Australia (Boudreau et al., 1993), Cl-rich apatite is typical of only a few intrusions, including the Stillwater Complex, Montana (Boudreau & McCallum, 1989) (see Fig. 6). Modern magmas with high Cl/F ratios are typically rich in other volatile components such as water (see summary by Boudreau et al. (1997)]. On the basis of analogy to modern boninitic magmas with chemistry similar to the inferred high-silica, high-magnesian magma that formed the Lower and Critical Zones, this magma was probably enriched in volatile components and had a Cl/F ratio of >1·0 (Sharpe & Hulbert, 1985; Hatton & Sharpe, 1989; Hickey & Frey, 1989; Sobolev & Chaussidon, 1996; Boudreau et al., 1997).
It is generally agreed that the Main Zone represents a chemically distinct magma from the Lower and Critical Zone, probably involving some interaction with crustal rocks. The coincidence of the jump in initial 87Sr/86Sr ratio (e.g. Kruger & Marsh, 1982) with an abrupt decrease in the mole fraction chlorapatite end-member from an average of 0·8 to 0·4 across the Merensky unit suggests that the Main Zone magma had a low Cl/F ratio. This Cl/F ratio is still high compared with many intrusions. For example, apatite analyses from the Munni Munni and Great Dyke typically do not contain >10% chlorapatite component (Boudreau et al., 1993, 1995). Also, a noticeable decrease in the abundance of hydrous minerals in the Main Zone as compared with the Lower and Critical Zones is consistent with a lower volatile concentration in the Main Zone magma.
The association of lower Cl/F ratios in the Main Zone with a more radiogenic Sr isotopic signature implies that the jump to more F-rich compositions in the Main Zone could have been the result of assimilation of F-rich crustal rocks at source. This also implies that the unusually high Cl/F ratio seen in the Lower and Critical Zone apatite compositions are not simply the result of assimilation of Cl-rich crustal rocks. However, given that the Main Zone is not as primitive as the Lower and Critical Zones, it may be that the Main Zone magma degassed and lost Cl before emplacement.
The gradual decrease of the Cl/F ratio through the Upper Zone is not associated with evidence for magma mixing events involving isotopically distinct magmas (e.g. Kruger et al., 1987). The trend of decreasing Cl/F ratio continues once apatite becomes a cotectic phase in Upper Zone C. This trend is the opposite of the expected trend if apatite were the only phase fractionating the halogens from the magma owing to the preferential incorporation of the smaller F ion relative to Cl (e.g. Boudreau & McCallum, 1989; Cawthorn, 1994).
Cawthorn (1994) suggested that significant Cl/F variability of apatite can result if cumulus apatite evolves in closed systems with different volumes of interstitial liquid. This model is similar to the way in which mafic minerals can become more iron rich through interactions with trapped liquid (e.g. Barnes, 1986). To explain the decreasing Cl/F trend in the Upper Zone by this model would imply that the proportion of cumulus apatite to interstitial liquid increases toward the top. Further, the effect must also counter the expected fractionation to more Cl-rich compositions as explained above. However, apatite mode decreases from a significant modal component of the rocks toward the top of the intrusions until it is only a minor phase in the most fractionated rocks at the top of the intrusion.
In addition, the estimated trapped liquid component tends to be uniformly low in these rocks (1–6%, Cawthorn & Walsh, 1988). If closed-system equilibration were the only significant mechanism controlling apatite composition, one would expect the most F-rich compositions to occur at the base of Upper Zone C where apatite mode may exceed 5% and effects of any trapped liquid shift would be lowest. This is not observed.
In other intrusions, we have suggested that a decrease in Cl/F ratio can result from preferential loss of Cl to a separating fluid phase during crystallization (e.g. Boudreau & McCallum, 1989). This is supported by the virtual absence of Cl in both apatite and biotite in the most evolved rocks of the complex, implying a complete loss of Cl at the end stages of crystallization. If true, this implies that the Upper Zone magma became fluid saturated at least by the point of apatite saturation.
In summary, compositional trends from the Upper Zone imply that the magma was fluid saturated at least by the time apatite became a cotectic phase and suggest that an igneous fluid separated at a high temperature (>1000°C). For the more volatile-rich Lower and Critical Zones, fluid separation may have begun even earlier. The lower equilibration temperature and non-uniform range of apatite compositions seen in most samples are interpreted to reflect both pore-scale fractionation and migration of interstitial liquid (concurrent with fluid saturation) and a later (possibly subsolidus) modest redistribution of halogens down to 750°C.
The problem of PGE and S distribution in the Lower and Critical Zones
The problem of PGE concentration mechanism in layered intrusions in general and the Bushveld Complex in particular has been discussed by a number of workers, most recently by Cawthorn (1999). As summarized by Cawthorn, precipitation of an immiscible sulfide liquid can act as a collector for the PGE and is the most conventionally accepted model, but there is very little preserved evidence that the magma was saturated in a sulfide phase. It contains no basal sulfide ores, and (as shown in Fig. 14) background sulfur contents below the Merensky reef are typically <200 ppm in the Lower Zone and less than 50–100 ppm for much of the Critical Zone. As noted by Cawthorn, these values are well below expected S-saturation concentrations.
The generally good correlation between incompatible trace elements and Cu in most samples would suggest that most sulfides can be readily explained as having precipitated from the residual (trapped) liquid. The association of sulfide with other late crystallizing minerals supports this. On the basis of an average P concentration of 0·07 wt % for the B1 sills (Sharpe & Hulbert, 1985), the Lower Zone contains 2–5% crystallized interstitial liquid and the Critical Zone roughly twice this amount. Similarly, the typical Cu concentrations of 5–10 ppm generally are not higher than expected from a trapped liquid component alone, assuming an initial magma concentration of 56 ppm Cu [equal to the average B1 quench textured micropyroxenite composition from Sharpe & Hulbert (1985)]. These low residual liquid fractions are broadly in line with other estimates (e.g. Maier & Barnes, 1999).
In contrast, the PGE concentrations are too high, especially in the Critical Zone, to be simply a function of trapped liquid fraction. For example, estimated parent magmas contain only 30 ppb Pt + Pd (Davies & Tredoux, 1985). The whole-rock Pt + Pd concentrations in silicate rocks from the Lower and Critical Zones exceed the expected trapped liquid proportion by a factor of from two to 100 (Maier & Barnes, 1999), with higher values occurring in rocks of the Critical Zone, in which there may be either no or very little increase in bulk S concentration.
McCarthy et al. (1984) suggested that the Bushveld magma was sulfide saturated throughout the Lower and Critical Zones. They explained the decrease in the modal amount of S with height as a result of increased S solubility in the liquid as it fractionates. However, most existing work suggests that sulfur saturation should actually decrease, not increase with fractionation [see the summary by Naldrett (1989)]. In this case, the amount of ‘cumulus’ sulfide should actually increase with height as the resident magma in the chamber precipitates excess sulfide in response to lower saturation levels. Further, one should expect the highest Pd concentrations to occur in the Lower Zone, where sulfide mode is highest, not the Critical Zone. To produce the observed trend, the bulk distribution for Pd would be required to increase significantly even as the modal amount of sulfide decreases with stratigraphic height.
Maier et al. (1996) also explained metal ratios as a result of cumulus sulfide precipitation, with variations largely resulting from changing silicate liquid/sulfide liquid mass ratio [the ‘R factor’ of Campbell & Naldrett (1979)]. This requires that the magma be saturated in sulfide at least during crystallization of the Critical Zone. Although Maier et al. suggested that the metal ratios are consistent with sulfide saturation, they did not address the low Cu and sulfide abundance of the Lower and especially the Critical Zone.
In a later work, Maier & Barnes (1999) plotted S against a computed ‘trapped liquid’ fraction [based on La content of rock as compared with La contents of the B1 sills of Harmer & Sharpe (1985)]. They suggested that sulfur and metal contents that exceed the expected trapped liquid fraction have a cumulus sulfide component. However, not discussed by them is the fact that S concentrations are still below most estimates of what sulfide-saturated assemblages should contain (e.g. Cawthorn, 1999). Further, a significant number of the analyses reported by Maier & Barnes have S contents that are markedly lower than that expected from a trapped melt component alone. This would suggest a loss of sulfur from these rocks.
None of these existing models take more than a passing look at the possibility that chalcophile elements may have been affected by degassing during crystallization, despite strong field, petrographic and geochemical evidence. As noted in the Introduction, a number of workers including Schiffries (1982) have presented evidence from the platiniferous Driekop pipe from the eastern Bushveld Complex, which indicates that the focused flow of a Cl-rich fluid altered the original norite to a hortonolite–dunite and strongly enriched the core of the pipe in PGE–Fe alloys. Also, whereas normal Merensky reef is composed of PGE–sulfides, it has long been recognized that PGE–Fe alloys commonly characterize ‘pothole’ reef. This has been attributed to sulfur loss to fluids fluxing through the pothole regions (e.g. Kinloch, 1982; Kinloch & Peyerl, 1990). Both features imply the significant redistribution of chalcophile elements during degassing of the Lower and Critical Zones.
In the Stillwater complex of Montana, we have previously noted the presence of discordant podiform and pipelike bodies containing several percent disseminated sulfide in the Middle Banded Zone (Boudreau & McCallum, 1986). Here, the association of the sulfide with other late-crystallizing minerals such as apatite and quartz was interpreted to imply that mineralizing fluids migrated upward through the crystal pile via pockets of residual interstitial liquid. Similarly, associated with the platiniferous J-M reef of the Stillwater complex are half-meter-sized bodies of massive sulfide bodies composed of pyrrhotite–chalcopyrite–pentlandite that form the core of discordant silicate pegmatoids (Boudreau, 1999). Again, this strongly suggests concentration of the chalcophile elements by fluids.
A number of workers have discussed the possible role of fluids in both influencing crystallization of the magma and being a possible transport agent for the PGE (e.g. Ballhaus & Stumpfl, 1986; Nicholson & Mathez, 1991). We have previously suggested that degassing of interstitial liquids can influence PGE, sulfur and base metals in a number of ways (e.g. Boudreau & McCallum, 1989; Boudreau & Meurer, 1999). These models are illustrated in Fig. 16 and are discussed below.
|
Vapor refining model
A process of vapor refining of the crystal pile can occur as interstitial liquid reaches fluid saturation and transports sulfur, base metals and the PGE (the chalcophile elements) upward. The actual behavior of metal transport during degassing of a solidifying and compacting crystal pile can be complex. For example, the crystal pile can act as a chromatographic column to separate the PGE from sulfur and the base metals during fluid migration (Boudreau & Meurer, 1999; Meurer et al., 1999).
In numeric models such as those of Boudreau & Meurer (1999), degassing can lead to 100% loss of ore elements. However, in natural systems, some sulfide grains may become isolated as the interstitial liquid crystallizes and small residual pockets of sulfide and silicate liquid become isolated. The observation that PGE enrichments of the Critical Zone occur above the Lower Zone would be consistent with this, the Lower Zone being the source of fluids and some of the metals now present in the Critical Zone.
Because there is little evidence for the significant input of fluids from underlying country rocks, the fluids that moved through the Lower Zone must have separated within the Lower Zone. Low in the complex the interstitial liquid would contain approximately the volatile content of the parent magma. As a consequence, the Lower Zone should have evolved a relatively small volume of fluid phase.
The Critical Zone, however, experienced chalcophile and especially volatile element enrichment from degassing of the underlying Lower Zone. An upward enrichment in the volatile content of the interstitial liquid would lead to earlier and more extensive degassing. The lower whole-rock S of the Critical Zone is consistent with more extensive sulfur loss. Overall, one would expect an upward displacement in the ore elements by such a vapor refining process.
Fluid-induced sulfide saturation model
In a variation of the above model, the degassing of interstitial liquids leads to fluid migration up into overlying hotter, less evolved interstitial liquids. This leads to a progressive increase in the volatile concentration (including sulfur) of the overlying interstitial liquids, causing them to degas earlier in their crystallization. Eventually, fluid may actually escape the crystal pile and mix with fluid-undersaturated magma in the chamber. The mixing of an S-bearing fluid into a fluid-undersaturated silicate magma leads to an increase in the S concentration in the magma and induces S saturation and PGE collection in addition to whatever may be carried by the fluid itself (Boudreau, 1999; Boudreau & Meurer, 1999). The model is similar to magma mixing schemes but uses an S-bearing fluid to induce sulfide saturation. Potholes have been suggested as channelways by which such fluids may have entered the chamber (e.g. Ballhaus, 1988).
The interaction of S-bearing fluids with resident magma at the top of the crystal pile can induce sporadic sulfide saturation in the magma that may have been otherwise far from sulfide saturation (e.g. Boudreau & Meurer, 1999). Because only part of the magma near the floor need reach sulfide saturation, only small amounts of sulfide need be produced. This sulfide precipitation could then collect the PGE of the resident magma (plus ore metals transported by the fluid), with much of the sulfur eventually lost to later degassing as the crystal pile continues to grow. This could explain the generally high PGE but overall low S content of the Critical Zone.
There have been a number of objections to fluid transport models, most of which focus on rather restricted possibilities of how such models must operate. For example, it is often claimed that the rocks below the various reefs cannot be a source for the PGE because the trapped liquid fraction does not contain enough PGE (e.g. Maier & Barnes, 1999). These arguments make two unwarranted simplifications. First, the amount of interstitial liquid can be substantially reduced by compaction by an order of magnitude or more (e.g. Shirley, 1986