| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Journal of Petrology | Volume 44 | Number 6 | Pages 1031-1054 | 2003
© Oxford University Press 2003
Melt Generation and Fluid Flow in the Thermal Aureole of the Bushveld Complex
1 DEPARTMENT OF EARTH SCIENCES, OPEN UNIVERSITY, MILTON KEYNES MK7 6AA, UK
2 DEPARTMENT OF EARTH SCIENCES, UNIVERSITY OF CAMBRIDGE, CAMBRIDGE CB2 3EQ, UK
3 SCHOOL OF GEOLOGICAL AND COMPUTER SCIENCES, UNIVERSITY OF NATAL, DURBAN 4041, SOUTH AFRICA
4 SCOTTISH UNIVERSITIES ENVIRONMENTAL RESEARCH CENTRE, EAST KILBRIDE, GLASGOW G75 0QF, UK
E-mail: n.b.w.harris{at}open.ac.uk
RECEIVED JULY 10, 2002; ACCEPTED DECEMBER 3, 2002
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONTHE AUREOLE OF THE...ANALYTICAL TECHNIQUESCONTACT METAMORPHISM OF QUARTZ...GRANITE SHEETS AND VEINSPEGMATITIC LENSESTIME SCALE OF FLUID...CONCLUSIONSREFERENCES |
|---|
Granite sheets emplaced into the migmatite zone of the eastern contact aureole of the Bushveld Complex resulted from fluid-enhanced, incongruent biotite melting of the underlying Silverton Formation shales during prograde metamorphism. Ba concentrations are extreme in both the sheets (>1000 ppm) and the hornfels (>800 ppm) into which they have been emplaced. We conclude that a Ba-rich, hydrothermal fluid induced melting in the aureole, and that fluid transport of Ba2+, and to a lesser extent, Sr2+ and Eu2+, persisted in the melt zones under subsolidus conditions. Sr-isotope systematics from high-Ba localities define an errorchron of 2161 ± 106 Ma with an initial (87Sr/86Sr) ratio of 0·705 ± 0·001. Metasedimentary rocks unaffected by fluid infiltration were homogenized at the same time but with an increased initial ratio, suggesting that whereas isotope homogenization was achieved between outcrops permeated by fluids, there is no evidence of regional homogenization. Oxygen-isotope compositions of psammitic metasediments in the aureole are uncorrelated with distance from the contact, suggesting the infiltrating fluid equilibrated isotopically with the metasediments. Their elevated
18O values (11·3–12·1
) are consistent with a fluid source from devolatilization of the sedimentary lithologies. Textural analysis at both outcrop and thin-section scale of arkose and psammites in the aureole (Lakenvalei and Magaliesberg Formations) shows that the extent of melting was highly heterogeneous, even on the grain scale, and resulted from heterogeneously distributed infiltration of aqueous fluid on dilatant cracks and grain boundaries. Cathodoluminescence imaging of quartz shows a marked difference in the amount of fine structure, with that in the melted rocks having uniform luminescence in contrast to that in rocks containing little or no melt, which preserve textures inherited from the regionally metamorphosed protolith. Simple finite-difference thermal modelling of the aureole suggests that the width of the melt zone (>500 m) is inconsistent with conductive heat transfer, and hence that the thermal structure has been modified by fluid advection. KEY WORDS: thermal aureole; crustal melting; fluid infiltration; isotope homogenization
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONTHE AUREOLE OF THE...ANALYTICAL TECHNIQUESCONTACT METAMORPHISM OF QUARTZ...GRANITE SHEETS AND VEINSPEGMATITIC LENSESTIME SCALE OF FLUID...CONCLUSIONSREFERENCES |
|---|
Thermal aureoles of mafic intrusions provide natural laboratories in which to study the formation and segregation of crustal melts under conditions of rapid prograde heating in the upper crust. In contrast to crustal melts formed during orogenesis, the high rates of prograde heating in thermal aureoles can lead to disequilibrium features between melt and its restite that carry kinetic information on crustal anatexis (Harris et al., 2000) specific to this mechanism of melting. Equally, as thermal aureoles allow both the heat source and the melt protolith to be precisely identified, they are an ideal environment for assessing the relative importance of protolith lithology, temperature and fluid fluxing in crustal melting.
The Rustenburg Layered Suite (RLS) of the Bushveld Complex of southern Africa is the world's largest reported layered mafic intrusion, ranging in thickness between 7 and 9 km and exposed over an area of 65 000 km2 (Walraven, 1982). The magmas intruded the sediments of the Transvaal Supergroup (Fig. 1) at 2058·9 ± 0·8 Ma (Buick et al., 2001). Strong thermal metamorphism resulted in crustal melting close to the contact with the floor of the intrusion (Fig. 2). The highest temperatures associated with emplacement of the RLS were generated by the intrusion of the Bushveld Lower Zone magmas, which crystallized to form a 1500 m thick cumulate sequence of pyroxenites, norites and peridotites at temperatures between 1100 and 1300°C (Sharpe & Hulbert, 1985; Wallmach et al., 1995). Evidence for crustal melting is best preserved around the Burgersfort Bulge on the eastern contact of the Lower Zone exposed in the eastern lobe of the Bushveld Complex (Fig. 1). This study focuses on the petrogenesis of these wall-rock melts and assesses the role of fluid infiltration within the aureole.
|
|
|
THE AUREOLE OF THE BURGERSFORT BULGE |
|---|
|
TOPABSTRACTINTRODUCTIONTHE AUREOLE OF THE...ANALYTICAL TECHNIQUESCONTACT METAMORPHISM OF QUARTZ...GRANITE SHEETS AND VEINSPEGMATITIC LENSESTIME SCALE OF FLUID...CONCLUSIONSREFERENCES |
|---|
The immediate floor of the Lower Zone of the RLS is underlain by quartzites and psammites (Lakenvalei and Magaliesberg formations) and argillaceous metasediments and minor calc-silicates (Vermont and Silverton formations) of the Early Proterozoic Pretoria Group from the Transvaal Supergroup (Fig. 1). The outer margin of the aureole,
3·5 km below the RLS, is marked by the appearance of biotite and andalusite in argillaceous rocks near the lower contact of the basal Pretoria Group, indicating temperatures of 500°C (Sharpe & Chadwick, 1982; Waters & Lovegrove, 2002). As the intrusion is approached, a succession of metamorphic zones have been identified: chiastolite–chloritoid slate, andalusite shale, and andalusite–cordierite hornfels in pelites, and the onset of significant recrystallization of quartzite. The mineral assemblages indicate temperatures of 600°C (Button, 1976; Sharpe & Chadwick, 1982). The innermost zone, between 1 and 2 km wide in map view, is characterized by sillimanite-bearing hornfels, schists and migmatites, indicating temperatures >700°C (Willemse & Viljoen, 1970). The highest-grade pelites in the aureole are defined by the assemblage alkali feldspar–sillimanite–cordierite, consistent with pressures of equilibration between 2·1 ± 0·4 kbar (Kaneko & Miyano, 1990) and 3·0 ± 0·5 kbar [obtained for garnet-bearing assemblages in the outer aureole by Waters & Lovegrove (2002)]. Recent mineral equilibria studies on a pelitic xenolith enclosed in granite overlying the RLS provide a minimum pressure estimate of 1·5 kbar for the base of the RLS (Pitra & De Waal, 2001). During the emplacement of this massive igneous complex, pressures will have increased progressively at the base of the intrusion, and it is possible that peak temperatures in the migmatite zone were experienced before peak pressure. Indeed, the emplacement of such a large igneous body will have imposed a complex and diachronous stress regime on the underlying rocks. Intense disharmonic folding (Fig. 3e) suggests an early period of compression and shear to the NE, probably associated with loading. Boudins within calc-silicate and quartzite horizons and a system of vertical east–west tension gashes within the pelites (Fig. 3f) mark a subsequent period of extension that may relate to sag of the lower contact (Uken & Watkeys, 1997).
|
Estimates of peak temperatures in the migmatite zone of the aureole are 750 ± 50°C (Nell, 1985; Kaneko & Miyano, 1990). Significantly, there is no evidence for the development of ultra-high temperatures in the aureole, in contrast to granulitic assemblages containing corundum, spinel and sapphirine preserved in pelitic xenoliths within the RLS (Willemse & Viljoen, 1970). Stable-isotope studies of the eastern Bushveld Complex suggest a vigorous hydrothermal system at temperatures of 350–700°C in the aureole (Schiffries & Rye, 1990; Buick et al., 2000). A
18O traverse of quartzite at the contact (Schiffries & Rye, 1990) is consistent with a metamorphic origin for the fluids, perhaps from the Transvaal Supergroup. 18O values of carbonates from the aureole (15–25) also suggest infiltration of a fluid derived from devolatilization of metasedimentary country rock (Buick et al., 2000). Melting within the migmatite zone can be recognized at a range of scales from the grain scale in quartzites and psammites to the presence of metre-scale granite veins and sheets (Fig. 3d) that have segregated from their migmatitic pelitic source. The textures, geochemistry and significance of these melts are described below.
|
ANALYTICAL TECHNIQUES |
|---|
|
TOPABSTRACTINTRODUCTIONTHE AUREOLE OF THE...ANALYTICAL TECHNIQUESCONTACT METAMORPHISM OF QUARTZ...GRANITE SHEETS AND VEINSPEGMATITIC LENSESTIME SCALE OF FLUID...CONCLUSIONSREFERENCES |
|---|
Bulk-rock samples were collected from the localities indicated in Table 1 by drilling 2 cm diameter cores (length 3–5 cm). Whole-rock major and selected trace elements (Rb, Sr) were analysed on an ARL Fisons wavelength-dispersive X-ray fluorescence (XRF) spectrometer at the Open University. Major elements were determined using glass discs prepared by fusing powdered samples with Spectroflux 105. Trace elements (Rb, Sr) were determined on pressed powder pellets following the procedures described by Potts et al. (1984).
|
Remaining trace elements were obtained by inductively coupled plasma mass spectrometry (ICP-MS) with an Agilent 7500 s system at the Open University. Fused glass discs were crushed and digested using standard HF–HNO3 digestion techniques. Samples were introduced to the ICP-MS system as 2% HNO3 solutions, with drift monitored by on-line spiking of Be, In, Tm and Re. Calibration was achieved using external and internal standards of known composition.
For Sr isotopic analysis, samples were digested in equal volumes of HF–HNO3–HCl in sealed beakers using a CEM Mars®5 microwave. Samples were dissolved at 210°C and 150 psi for a period of 30–60 min. Sr separates were obtained by standard ion exchange procedures and isotopic ratios were determined statically in multi-collector mode on the Nu Instruments© MC-ICP-MS system at the Open University. Reproducibility of the NBS987 Sr standard over the analysis period was 0·710231 ± 0·000017 (2
). Sr fractionation was corrected to 86Sr/88Sr = 0·1194.
Electron microprobe analyses were obtained with a Cameca SX100, operating at 20 kV accelerating voltage, with a 20 nA beam current and 10 µm beam size.
Samples for stable isotope analysis were crushed and sieved for picking in ethanol under a binocular microscope. Oxygen isotope analyses were carried out at the Scottish Universities Environmental Research Centre by the laser fluorination method of Sharp (1990) as modified by Macaulay et al. (2000). The precision at 1 is ±0·2 or better, and for NBS28 the value measured is +9·60. All data are reported as delta per mil (18O ) relative to V-SMOW.
Cathodoluminescence (CL) imaging of quartz was carried out using a JEOL JSM-820 scanning electron microscope running at an accelerating voltage of 15 kV, fitted with a retractable MonoCL detector supplied by Gatan UK. A parabolic mirror placed under the pole-piece 3 mm above the sample allows a high proportion of the photon flux to be collected and directed to a photomultiplier tube for panchromatic imaging. Such a system, together with the scanning technique, allows images to be obtained using a current of 3 nA, which minimizes electron beam damage. The minimum magnification for the MonoCL system is 100x and is constrained by the aperture of the paraboloidal mirror. The resultant panchromatic images show variations in luminescence intensity as shades of grey. Black indicates no (or very weak) lumines-cence, and white indicates very strong luminescence.
|
CONTACT METAMORPHISM OF QUARTZ-RICH METASEDIMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONTHE AUREOLE OF THE...ANALYTICAL TECHNIQUESCONTACT METAMORPHISM OF QUARTZ...GRANITE SHEETS AND VEINSPEGMATITIC LENSESTIME SCALE OF FLUID...CONCLUSIONSREFERENCES |
|---|
Two psammite horizons occur in the high-grade metamorphic aureole in the Burgesfort area: the Lakenvalei Formation, which marks the contact itself at one outcrop (SA19), and the 500 m thick Magaliesberg Formation (Fig. 1), which is coarse grained and recrystallized over a distance of 1–2 km perpendicular to the contact (Button, 1976). Field observations from some outcrops within 100 m of the contact around the Burgesfort Bulge provide evidence for a high degree of extension-related deformation of psammite including flowage into the low-strain necks of extended pelitic horizons (Fig. 3b) and loss of sedimentary structures. However, loss of sedimentary structures such as ripple marks and cross bedding is not ubiquitous within the Magaliesberg Formation, even within a few metres of the contact. Indeed, sedimentary structures coexist with evidence of extreme deformation within a few metres in the same outcrop (SA22), illustrating the localized nature of psammite mobility. A related phenomenon is the presence of coarse fingers of virtually feldspar-free, coarse-grained quartzite within finer-grained psammite (Fig. 3a). These are structurally similar to the pipe-like fingers of coarse peridotite within a generally fine-grained peridotite massif that have been attributed to localized recrystallization related to fluid infiltration (Kostenko et al., 2002).
Samples of Magaliesberg and Lakenvalei Formation psammites and quartzites were collected within the aureole, at the contact with the RLS (SA19, SA23), and at 100–120 m from the contact (SA21, SA18). Magaliesberg Formation psammite was collected 3–4 km from the contact (SA05). Samples within the aureole containing relict sedimentary structures are SA21b and c, and SA23. Samples SA19a and SA21d show outcrop evidence for deformation and contact metamorphism-related destruction of sedimentary structures. Sample SA21d contains the coarse-grained feldspar-poor fingers. Table 2 summarizes these observations.
|
The psammites contain up to 20 vol. % feldspar (dominated by alkali feldspar) and muscovite. Significant post-peak grain growth and recrystallization has occurred, with abundant evidence for quartz grain-growth inhibition by feldspar (Fig. 4a). The feldspar grains have turbid, commonly albite-rich, margins, and the quartz contains abundant planar aggregates of fluid inclusions, suggestive of late-stage infiltration of aqueous fluids (e.g. Saigal et al., 1988; Lee & Parsons, 1997).
|
Petrographic observations
Regardless of composition, feldspar grains have two distinct forms within samples collected in the aureole. The first is attenuated and cuspate, with elongate extensions on quartz grain boundaries or into individual quartz grains (Fig. 4a–c). These shapes may be poly- or monocrystalline, and are similar to those observed in partially melted feldspar-bearing quartzites (e.g. Holness & Clemens, 1999).
The second form is more compact and equant, with grains either on quartz grain boundaries (Fig. 4d) or enclosed entirely within single quartz grains. The latter tend to be small and well rounded (Fig. 4a and c). The relative proportion of the two forms varies between samples, with SA19a containing predominantly attenuated grains and SA23 dominated by the more compact form. The significance of these two textural types is illustrated by measurement of the quartz–quartz–feldspar (qqf) dihedral angle. Under conditions of textural equilibrium, the minimization of internal energies results in the development of a characteristic angle subtended at the junction of a grain of one phase against two grains of another. This angle, the dihedral angle, is a function of the relative magnitudes of the two interfacial energies (Smith, 1964). The equilibrium value of the quartz–quartz–plagioclase angle is in the range 105°–110° (Vernon, 1968).
Although a 2-D cross-section through a three-grain junction will not necessarily give the true 3-D value of dihedral angle, the median of a large (>25) population of measured 2-D angles is within 1° of the true 3-D angle (Riegger & Van Vlack, 1960). For a single value of 3-D dihedral angle, a unimodal population of 2-D angles is expected, whereas a bi- or poly-modal distribution indicates the presence of two or more 3-D angles. Optical measurement of the qqf dihedral angle in the unmetamorphosed sample SA05 gives a median value of 105° ± 3° (Fig. 5). Errors are calculated using the method of Wilks (1962) as reported by Stickels & Hucke (1964). The median of the population measured for SA05 is within the range expected for the equilibrium qqf angle (Vernon, 1968), although the distribution of measured angles is slightly skewed from that expected for a rock with only a single 3-D value of dihedral angle (Fig. 5). The qqf angle in a late granitic vein from sample MW15a is shown in Fig. 5 for comparison. The distribution of angles in this sample has a unimodal peak and a median value of 72° ± 3°. As values of dihedral angle much lower than 100° are associated only with fluid-bearing geological systems [see review by Holness (1997)], low values of the qqf angle in this sample must have been inherited from a precursor melt phase (e.g. Laporte, 1994; Laporte & Watson, 1995).
|
The populations of observed angles in samples SA19a and SA21c show a peak at around 105° but also a significant proportion of low angles (see Fig. 5), with median values of 87° ± 3° and 95° ± 3°, respectively. In these samples, the low angles are associated with the attenuated forms, and the higher angles with the more compact feldspar forms.
In common with earlier studies (Platten, 1982; Hunter, 1987; Holness & Clemens, 1999; Clemens & Holness, 2000; Rosenberg & Riller, 2000), we suggest that the highly cuspate feldspar grains crystallized during the final stages of solidification of partially melted psammite, whereas the more compact forms represent unmelted feldspar. The field evidence for disappearance of sedimentary structures and the evidence for localized extreme deformation in the quartz-rich metasediments are also consistent with partial melting. The coexistence of the two forms of feldspar in any one sample, and the juxtaposition of sedimentary structures with structurally homogeneous rock in a single outcrop are thus indicative of incomplete melting that was heterogeneous on a range of scales from grain to outcrop.
The evidence for partial melting given above, together with the generally low temperatures within the aureole (Nell, 1985; Kaneko & Miyano, 1990; Johnson, 2001), points to the necessity for melting to have occurred in the presence of water, as dry quartz–feldspar rocks require extremely high temperatures to melt. Given the likelihood that the equilibrium fluid–quartz dihedral angle was >60° at the metamorphic peak (Holness, 1993, 1998), we suggest that fluid infiltration occurred along dilatant cracks and grain boundaries.
The amount of melt in the psammites at the metamorphic peak was likely to have been small, with the maximum possible melt volume a function of the amount of feldspar in the rock and the maximum temperature. Given that most of the samples contain <20 vol. % feldspar, we consider the maximum amount of melt was generally <50 vol. %. However, loss of sedimentary structures is likely to occur at low (<10 vol. %) melt fractions during deformation.
Several psammites were analysed for oxygen isotopes to compare unmetamorphosed arkose with partially melted psammite. In contrast to the study of Schiffries & Rye (1990) we found no pattern of decreasing 18O as the contact is approached. The quartz in all samples has 18O in the range 11·3–12·3 regardless of location or extent of melting (Table 2). This is consistent with small amounts of infiltrated fluid. Assuming fractionation values for 
18Oqtz-px at 700–800°C from Matthews et al. (1983) and pyroxene data for the RLS from Harris & Chaumba (2001), fluids derived from the RLS would reset quartz values to 8. The 18O data in this study therefore preclude the RLS as a fluid source and support the interpretation that devolatilization of Transvaal Supergroup sediments is the likely source of the convecting fluids (Buick et al., 2000).
Cathodoluminescence of psammites
Details of recrystallization and fluid infiltration can be inferred from CL images of quartz (e.g. Seyedolali et al., 1997; Holness & Watt, 2001). The causes of luminescence in quartz are not well understood but result from a complex interplay of factors such as lattice defects, trapped electron–hole pairs, oxygen vacancies, and traces of impurities such as Al, Ti, Li, Na and OH (e.g. Marshall, 1988; Waychunas, 1988; Fisher et al., 1990; Fabre et al., 2002). In general, quartz grown at low temperatures (<200°C) has low luminescence (D'Lemos et al., 1997), possibly because the defect density and impurity concentration are lower than those in high-temperature quartz (Watt et al., 2000). The extreme sensitivity of CL to chemical heterogeneities and lattice defects makes it highly superior to conventional optical microscopy for distinguishing different quartz growth generations.
Three samples were chosen for detailed investigation. SA21c was sampled close to the contact but contains well-preserved sedimentary structures and has mainly compact feldspar grains. Sample SA19a has no sedimentary structures and contains attenuated feldspar grains with a low dihedral angle. Both characteristics are indicative of relatively voluminous partial melting. Sample SA21d was collected from near SA21c but has no relict sedimentary features, a significant population of low qqf dihedral angles (Fig. 5), and coarse-grained fingers of virtually feldspar-free quartzite.
Sample SA21c shows a great complexity of quartz textures under CL with at least six distinct generations, which can be assigned an age order based on cross-cutting relationships (Fig. 6). The youngest of these generations (generation 1) is dark and non-luminescent, forming cross-cutting planar features up to 20 µm wide, often with a dominant orientation, associated with fluid inclusion arrays and non-luminescent low-temperature feldspar alteration (Fig. 6a). The same dark, non-luminescent quartz is present at many of the present-day grain boundaries. Such features are commonly observed in both igneous and metamorphic quartz (Valley & Graham, 1996; D'Lemos et al., 1997; Seyedolali et al., 1997; Watt et al., 1997; Holness & Watt, 2001) and are believed to form during late-stage, low-temperature fluid infiltration along fractures and dilatant grain boundaries.
|
The next youngest quartz generation (generation 2) is weakly luminescent and is closely associated with generation 1 quartz on grain boundaries. Generation 2 quartz alternates on either side of irregular grain boundaries and represents quartz recrystallized during the later stages of grain growth (Fig. 6a and b).
These two youngest generations cut sets of oscillatory zoned quartz (generation 3) (Fig. 6b). Such oscillatory zoning is frequently present in phenocrystic magmatic quartz (Seyedolali et al., 1997; Watt et al., 1997; Peppard et al., 2001), in granite plutons (Valley & Graham, 1996; D'Lemos et al., 1997; Seyedolali et al., 1997) and in quartz grown in open vein systems (Wilkinson & Johnston, 1996; Penniston-Dorland, 2001). A similar occurrence found in contact metamorphosed quartzite has been ascribed to grain growth during episodic fluid infiltration on grain boundaries (Yardley et al., 1991; Holness & Watt, 2001).
The fourth type of luminescence (generation 4) has not been previously described in the literature. It resembles a pleated fabric, with a fine network of interconnected pairs of bright and dark lines in two dimensions (Fig. 6c). Quartz of generation 4 is predominantly found at the edges of the present-day grains. Rarely, grains are cut by a fine network of bright features (generation 5) similar to the Bright Vein Networks (BVN) of Watt et al. (2000) (Fig. 6b). These are thought to be healed micro-cracks formed during internally generated overpressure related to reaction.
The oldest generation (6) of quartz is mottled or homogeneous and occurs predominantly at the centres of grains (Fig. 6c and d). Following Lind (1996) and Holness & Watt (2001) we suggest it represents quartz that retained its regional metamorphic signature (Seyedolali et al., 1997), remaining unrecrystallized during contact metamorphism. Adjacent grains show varying levels of luminescence (Fig. 5a)—this is related to the variation of luminescence intensity with crystallographic orientation (Walderhaug & Rykkje, 2000).
In contrast to the detail visible in sample SA 21c, SA19a shows only three distinct quartz generations (Fig. 7). Much of the sample comprises completely homogeneous, weakly luminescent quartz, cut by the non-luminescent late-stage fractures (generation 1) and Bright Vein Networks (generation 5). The latter appear to be slightly more common in the vicinity of feldspar (Fig. 7c). Very rarely the centres of large quartz grains contain a fine tracery of bright irregular planar features (Fig. 7d) that are highly reminiscent of the ‘mesh texture’ of Lind (1996), interpreted as unrecrystallized regional metamorphic relicts (generation 6; Lind, 1996; Holness & Watt, 2001).
|
Sample SA21d shows similarities to both SA21c and SA19a. The regions between the coarse-grained quartz fingers are indistinguishable from SA21c, whereas the fingers themselves show a reduced degree of detail reminiscent of SA19a. The fingers display the dark non-luminescent fracture network of generation 1, occasionally associated with the growth of weakly luminescent generation 2 quartz on grain boundaries (Fig. 7e). Pleating is rare and poorly developed, although much of the quartz shows a mottled texture. Faint oscillatory zoning is present, with zones parallel to the present-day grain boundaries (Fig. 7f).
The above interpretation of the CL images is necessarily tentative. It is notable that the sample in which sedimentary features are preserved displays the greatest amount of detail and the greatest number of distinguishable quartz generations, whereas the sample that shows melting-related deformation in outcrop has very little record of a complex recrystallization history.
The significant proportion of low measured dihedral angles in SA21c (Fig. 5) demonstrates that this sample did contain some melt at the metamorphic peak. A similar study of partially melted arkose has shown that solidification results in the formation of uniformly bright quartz at the present-day grain boundaries (Holness & Watt, 2001). This is not apparent in any of the Bushveld samples imaged. It is possible that the previously undescribed pleated texture is related to melt solidification but this is far from certain. It is perhaps more likely that the amount of melt formed in SA21c was small and is not distinguishable using CL. This is consistent with the lack of disruption of sedimentary structures such as cross-bedding and points to spatially highly variable amounts of melting associated with well-focused fluid infiltration.
The loss of fine CL structure in the fingers of SA21d is possibly attributable to recrystallization at high temperatures. Low-temperature growth would probably have resulted in loss of CL intensity (D'Lemos et al., 1997). We suggest the fingers represent the pathways for focused sub-solidus fluid infiltration closely following the metamorphic peak (see Kostenko et al., 2002). This is consistent with a slightly lower oxygen isotope composition for the fingers relative to their matrix (Table 2).
|
GRANITE SHEETS AND VEINS |
|---|
|
TOPABSTRACTINTRODUCTIONTHE AUREOLE OF THE...ANALYTICAL TECHNIQUESCONTACT METAMORPHISM OF QUARTZ...GRANITE SHEETS AND VEINSPEGMATITIC LENSESTIME SCALE OF FLUID...CONCLUSIONSREFERENCES |
|---|
The largest granite body recognized in the aureole is a biotite granite sheet,
50 cm wide, intruded into migmatized Silverton Formation shales containing quartz–biotite–alkali feldspar–plagioclase–cordierite assemblages (Fig. 3c and d). The granite was emplaced 100 m from the contact with the RLS, separated from the latter by the Magaliesberg quartzite (Fig. 1). The sheet has been slightly boudinaged. Twenty metres downslope of the sheet, nebulitic leucosomes enclose undisturbed pelitic schlieren suggesting in situ melting (Fig. 3c). The leucosomes have not been analysed because of their marked inhomogeneity resulting from entrainment of restite (biotite, quartz, sillimanite) and peritectic (cordierite and possibly alkali feldspar) phases. In addition, migmatite leusosomes often incorporate cumulate phases from an extracted melt (Sawyer, 1987; Ellis & Obata, 1992; Otamendi & Patiño Douce, 2001) and back-reaction between melt and host rock is probable during crystallization (Fourcade et al., 1992; Kriegsman & Hensen, 1998). These factors conspire to make the interpretation of leucosome compositions equivocal and so this study focuses on the extracted melt, represented by the sheet, which provides a homogeneous mineralogy.
The granite sheet is medium grained (1 mm grain size), comprising quartz, oligoclase (An12), microcline and subsidiary muscovite and biotite with an interlocking texture. The bulk of the feldspar in the adjacent psammitic hornfels has attenuated and cuspate shapes indicative of in situ melting (e.g. Fig. 4c). In addition to this body, examples of small-scale granitic intrusions include early granitic veins (2–10 cm wide) that have been strongly deformed resulting in disharmonic ptygmatic folding (SA13a, Table 3), and an undeformed post-deformation vein (SA22e, Table 3).
|
Trace-element abundances of both the granite sheet and the veins show broadly similar trends. Compared with a peraluminous orogenic granite from the Himalaya, Rb is depleted but Sr and Ba are enriched (Fig. 8). High field strength elements are similar for both granite types, with the notable exception of Zr and Hf, which may reflect an inherited zircon component in the granite sheet.
|
Modelling trace-element abundances in the granite melt
At the minimum pressure of 2–3 kbar (Kaneko & Miyano, 1990; Waters & Lovegrove, 2002), melting of pelites in the presence of aqueous fluid can occur at temperatures of
650°C (Le Breton & Thompson, 1988). However, the mineral assemblages of the metasediments in the aureole indicate that the granites formed at higher temperatures (>700°C). Under these conditions, melts may form in pelitic or semi-pelitic assemblages by fluid-absent melting of muscovite and/or biotite, as well as by fluid fluxing (Fig. 9). Major-element compositions, when plotted in the granite ternary system (Fig. 10), suggest that compositions are depleted in albite compared with granites formed by the muscovite-dehydration reaction.
|
|
Coarse poikiloblastic cordierite is widespread in migmatites, hornfels and schists and is likely to result from subsolidus conditions. However, fine-grained euhedral cordierite is restricted to the anatectic migmatites, suggesting that it is a solid product of the melt reaction. As H2O-saturated melting does not generate peritectic phases, the migmatite assemblages suggest that biotite breakdown, under fluid-absent conditions, is the primary mechanism for melt formation.
The most fertile assemblage at the onset of meltingwill be biotite–quartz–alkali feldspar–plagioclase–sillimanite, found in metagreywacke or semi-pelitic lithologies, resulting in a melt reaction of the general form
 |
2% H2O. The precise value of added H2O required to generate these melt compositions will depend on the mineral compositions and modal abundances of biotite and feldspar in the protolith. However, it should be noted that such low percentages of fluid infiltration are insufficient to allow an aqueous fluid and melt to coexist.
To assess whether such a melt reaction can generate the observed trace-element composition of the granite it is necessary to establish the reaction stoichiometry. For typical feldspar, biotite and cordierite compositions in the hornfels (Table 4), and the bulk analyses of the granite sheet (Table 3), the following melt reaction is obtained by mass balance:
 | (1) |
 | (2) |
i XiKdi, where Kdi is the partition coefficient between mineral i and melt.
|
Mass-balance constraints for melting of the Timeball Hill Formation shales (basal Pretoria Group) from the Bushveld contact aureole in the Potgietersus region to the NW of the study area suggest a value of F = 0·65 (Nell, 1985). This is probably too high for the metasedimentary rocks and granite compositions observed in this study as there is insufficient alkali feldspar in the unmelted rocks to allow a melt fraction >0·3. This compares with melt fractions of 0·20–0·35 from the experimental melting of a pelitic gneiss at
750°C, 3 kbar and 2% added H2O (Fig. 9). A larger melt fraction will result in depletion of the source in alkali feldspar, thereby restricting the K2O of the melt to <4% (Holtz & Johannes, 1991) in contrast to the observed values of 5% in the granite sheet. However, this may explain the origin of the late vein SA22e, which has low K2O but high Zr, possibly indicative of higher temperatures than the granite sheet.
Rb, Sr and Ba geochemistry
The geochemistry of Rb, Sr and Ba is highly influenced by the behaviour of protolith feldspar and mica, into which they are strongly partitioned. Mica and feldspar are the main reactant phases during melting, and if we assume a melt fraction of 0·3, the metasediments will generate a modal restite assemblage similar to that in Table 5.
|
For quantitative modelling of trace-element behaviour during equilibrium melting, the appropriate partition coefficients (Kd) must be known. The uncertainties in the equilibrium partition coefficients of Rb, Sr and Ba between micas or feldspars and granite melts propagate significant uncertainties in modelled melt compositions. Published data have been derived from experimental studies (Blundy & Wood, 1991) and empirical studies of volcanic glasses (Mahood & Hildreth, 1983; Nash & Crecraft, 1985; Ewart & Griffin, 1994; Icenhower & London, 1995, 1996) and migmatites (Bea et al., 1994). The latter are the most problematical in that diffusion of Rb, Sr and Ba is likely during both crystallization and subsolidus cooling of the melt, leading to lower-temperature re-equilibration (Nabelek, 1999). Variation in published Kd values is partly a result of the sensitivity of partitioning to volatile content, to the degree of polymerization in the melt (Ewart & Griffin, 1994) and to the crystal chemistry of the restite phases (Blundy & Wood, 1991). Ideally, partition coefficients should be determined for a melt of similar composition to the granite, that contains phenocrysts of the mineral of interest and that has been rapidly cooled from temperatures equivalent to the solidus temperature of the granite. Also important is the role of H2O, which can act as a network modifier or behave as a charge-balancing cation (Kohn, 2000).
For modelling melting under disequilibrium conditions, such that the melt inherits the trace-element characteristics of the reactant minerals, we have assumed the equation of Harris et al. (1993):
 | (3) |
For F = 0·3 and the range of partition coefficients given in Table 6, the values for Cl/C0 under both equilibrium and disequilibrium conditions are given in Table 5. To calculate abundances of trace elements in the melt a protolith composition (C0) must be assumed. For a model protolith we have assumed the composition of MW16 with Rb, Sr and Ba abundances characteristic of metasedimentary values throughout the aureole where melting is not observed.
|
The resulting concentrations of Rb, Sr and Ba in the modelled melt, assuming a melt fraction of 0·3, are given in Table 5. For the pelites and psammites sampled within 10 cm of the granite sheet, or in the proximal migmatites that source similar granites, anomalous concentrations of both Ba and Sr are observed, exceeding 800 and 200 ppm, respectively (Fig. 11). As there are no indications that these lithologies have unusual abundances of detrital feldspar, this suggests metasomatic enrichment of these elements.
|
Although modelled trace-element concentrations are relatively insensitive to variations in melt fractions between 0·2 and 0·4, the uncertainties in the relevant partition coefficients (Table 6) generate large variations. Nevertheless, the following conclusions can be drawn. For Rb, trace-element behaviour is consistent with equilibrium melting, but the abundances of Sr and Ba are inconsistent with either equilibrium or disequilibrium melting, suggesting open-system behaviour for these elements. In other thermal aureoles, Ba anomalies mimic oxygen-isotope patterns (Nabelek & Labotka, 1993), indicating that Ba is particularly mobile in hydrothermal fluids and thus providing evidence for high fluid/rock ratios. Under subsolidus conditions, Ba will partition strongly into alkali feldspars from a hydrothermal fluid at 2 kbar and 700–800°C, similar to conditions that prevailed in the Bushveld aureole (Carron & Lagache, 1980). However, it is unlikely that the limited fluid required for melting (2–3%) was solely responsible for the observed Ba enrichment, as this suggests unfeasibly high Ba concentrations in the fluid (3–4% Ba). It is, therefore, probable that metasomatism persisted under sub-solidus conditions during cooling.
Sr has similar distribution coefficients to Ba between fluid, magma and feldspars (Carron & Lagache, 1980). Sr abundances in the metasedimentary rocks are partially determined by the abundance of plagioclase, but a plot of Sr vs Ba suggests that Sr is also enriched in schists close to the granites (Fig. 11). Anomalously high levels of Ba and Sr in the melt are suggested by their strong positive anomalies relative to orogenic granites formed under fluid-absent conditions (Fig. 8). We suggest that advection of Ba- and Sr-bearing fluids promoted melting. The country rock was metasomatized both during crystallization of the Ba-rich granite by late-stage fluid exsolution and during subsolidus metasomatism. The system was therefore open to both elements, overprinting any disequilibrium signature that may result from rapid melting.
Rare earth element geochemistry
Chondrite-normalized rare earth element (REE) plots of Silverton shales from outside the aureole (Fig. 12a) display a negative Eu anomaly, typical of post-Archaean shales, and are indistinguishable from pelites in the aureole (Fig. 12b), suggesting that solid-state metamorphism has no appreciable effect on the REE concentration. However, those metasediments sampled from the locality of the granite sheet display a reduced negative Eu anomaly (Fig. 12c). REE concentrations in the granite are lower than in their protoliths, and are characterized by a positive Eu anomaly in contrast to the marked negative anomaly of Himalayan granites that result from dehydration of metapelitic lithologies (Fig. 12d).
|
Although REE concentrations in basaltic melts are controlled by partitioning between major phases and melt, for felsic magmas, accessory phase dissolution is the dominant control (Ayres & Harris, 1997). This can be illustrated by extending the modelling of Rb, Sr and Ba to the REE to generate REE abundances in the melt from a semi-pelitic source, assuming a melt fraction of 0·3 and the partition coefficients given in Table 6. The results (Fig. 13) demonstrate that if REE concentrations result from major-phase reaction and subsequent equilibration, then, compared with the granites in the aureole, the predicted REE patterns (1) have much higher REE abundances than are observed and (2) have a negative, rather than a positive Eu anomaly. The underlying cause for the discrepancy between modelled and observed data is that a significant proportion of the REE in the metasedimentary source is sequestered in refractory accessory phases that retain these elements throughout melting and melt extraction. Thus only a small proportion of the total REE budget is available to the melt through dissolution of these phases.
|
The positive Eu anomaly in the granite sheet might be explained by rapid melting and melt extraction, as the REE profile resulting from disequilibrium melting of a plagioclase-bearing source has a strong positive Eu anomaly derived from the reacting plagioclase (Prince et al., 2001). The extent of the anomaly depends on the proportion of the Eu budget in the source that is in accessory phases, but for a typical pelitic schist
25% of the Eu occurs in major phases (principally feldspars). Disequilibrium melting of such a source would yield the REE profile illustrated in Fig. 13. Hence the REE abundances in the granite could result from the combination of disequilibrium melting of major phases and accessory phase dissolution. However, it is also possible that melting did not occur in a closed system for REE. We have shown that the system was open to fluids transporting divalent ions such as Sr2+ and Ba2+. Given the geochemical similarities between Sr2+ and Eu2+, it is likely that these fluids also transported Eu2+. Evidence for Eu mobility may be derived by comparing the Eu/Eu* ratio for metasediments that have experienced Ba and Sr metasomatism (0·62–1·03, Fig. 12c) with the ratio from other aureole metasediments (0·54–0·74, Fig. 12b). |
PEGMATITIC LENSES |
|---|
|
TOPABSTRACTINTRODUCTIONTHE AUREOLE OF THE...ANALYTICAL TECHNIQUESCONTACT METAMORPHISM OF QUARTZ...GRANITE SHEETS AND VEINSPEGMATITIC LENSESTIME SCALE OF FLUID...CONCLUSIONSREFERENCES |
|---|
The most widespread evidence for possible melt formation in the aureole is provided by coarse-grained quartz–fibrolite lenses that develop within low-strain areas of the deformed schists (Fig. 3e), or fill tension gashes connecting anastomosing veins of fibrolite (Fig. 3f). The lenses are made up of quartz and fibrolite, tourmaline, muscovite ± biotite, cordierite, alkali feldspar or plagioclase. Although previously described as tourmaline-rich granites representing melt-fractions expelled from residual stromatic leucosomes (Johnson, 2001), the absence of either alkali feldspar or plagioclase in many lenses indicates that they are not granites. The lenses are undeformed, and frequently zoned with tourmaline-rich cores. Quartz and tourmaline display eutectic intergrowths in some lenses. Fibrolite is abundant in all lenses, concentrated along the grain boundaries of the quartz. Although their grain size (
1 cm) is insufficient to conform to many definitions of a pegmatite sensu stricto these assemblages share many of the textures and composition of pegmatites and so are termed ‘pegmatitic’ in this study. The obvious chemical features of these lenses are their highly siliceous nature and strong variability in major elements inconsistent with being a minimum equilibrium melt in the granite system (Table 3, Fig. 10). The alumina saturation indices (ASI) vary from 1·5 to >20, comparable with the normal range of 1–1·4 for peraluminous granites. The coarse grain size, wide range in major-element chemistry and mineral modal zoning suggest that lens formation may be similar to that of granitic pegmatites. Trace-element concentrations in the pegmatites are highly variable (Table 2), and both strongly negative and positive Eu anomalies are observed in the analysed samples. Positive Eu anomalies and low overall REE abundances are found in feldspar-bearing pegmatites, and negative Eu anomalies and higher overall REE abundances are found in tourmaline-rich samples.
In part, these lenses probably owe their origin to a high F activity in the granitic melt which promoted crystallization of sillimanite over alkali feldspar (Nabelek, 1999). The quartz–fibrolite seams (Fig. 3f) therefore represent melt pathways where early crystallized quartz and sillimanite have been left behind. Recent experimental work has indicated that many pegmatite features can be explained by the disequilibrium crystallization of a volatile-rich peraluminous melt of minimum-melt composition (London, 1992). Volatiles lower the solidus temperature of peraluminous melts, and also decrease nucleation rates, resulting in undercooling. Crystallization is then determined by nucleation kinetics (London, 1996). In tourmaline-rich systems, B forms stable, soluble borosilicate complexes with alkalis and Al, inhibiting nucleation of feldspars. During quartz-dominated crystallization from the undercooled melt, an aqueous fluid exsolves and transports these elements out of the system. The pegmatitic lenses therefore may result from the crystallization of peraluminous granitic melts, as do the granite sheets. However, a high volatile activity, as indicated by the presence of F-bearing tourmaline, allowed the magma to persist to lower temperatures before crystallizing under disequilibrium conditions. Fibrolite, muscovite and quartz coexist in many lenses, suggesting temperatures of <680°C at 3 kbar (Fig. 9). The absence of a Ba anomaly in the host rocks indicates that the lenses result from a phase of fluid fluxing distinct from that associated with the formation of the granite sheet. The siting of the pegmatites in vertical east–west tension gashes suggests that the host melt may have accumulated within sag fractures in the intrusion floor.
|
TIME SCALE OF FLUID CIRCULATION |
|---|
|
TOPABSTRACTINTRODUCTIONTHE AUREOLE OF THE...ANALYTICAL TECHNIQUESCONTACT METAMORPHISM OF QUARTZ...GRANITE SHEETS AND VEINSPEGMATITIC LENSESTIME SCALE OF FLUID...CONCLUSIONSREFERENCES |
|---|
Partial melting within a thermal aureole provides potential for constraining the rates of melting of crustal rocks from determinations of the degree of isotopic disequilibrium between melt and source at the time of melting (Knesel & Davidson, 1999). For the granite sheet, an initial Sr-isotope ratio (87Sr/86Sr)0 of 0·7053–0·7058 is obtained at 2059 Ma. This is within error of calculated values for the pelitic protolith (Table 7), indicating that no discernible disequilibrium has been preserved. The Sr-isotope compositions of five samples from Ba-rich localities within the migmatite zone, including granites, migmatites and hornfels, define an errorchron of 2161 ± 106 Ma, with an initial ratio of 0·705 ± 0·001 Ma (Fig. 14), suggesting that Sr isotopes have approached equilibrium between these diverse rock types both at outcrop scale and between outcrops affected by the high-Ba fluids.
|
|
For locality SA14, which has not been infiltrated by these fluids but where pegmatitic lenses are well developed, the Sr isotope composition approached homogenization at 2090 ± 51 Ma (within error of the previous age) at an (87Sr/86Sr)0 value of 0·715 ± 0·001. This represents a significant increase in the initial ratio, which suggests that whereas both sample sets underwent isotope homogenization at outcrop scale, there was no regional homogenization between rocks that had been affected by Ba metasomatism and those that had not.
For homogenization to occur over a length scale of several metres, volume diffusion within individual grains must first homogenize contiguous grains. Volume diffusion of Sr at 750°C in oligoclase would require 30–80 kyr to homogenize the Sr isotope composition across 1 mm, assuming the diffusivities of Giletti & Casserly (1994). Once homogenization had occurred over this length scale, homogenization at outcrop scale would be accomplished by grain boundary diffusion, accelerated by fluid advection. We conclude that the migmatite zone was held at temperatures >750°C for at least 30 kyr.
Published thermal models of the Bushveld Complex have simulated the cooling history of the RLS rather than the aureole itself. Such modelling is complicated by the fact that the RLS was not emplaced as a single sheet of magma, but as a succession of pulses. Walraven (1982) modelled three pulses, the first being 2500 m thick, intruded at 1400°C. He concluded that 400 kyr was required for complete solidification of the RLS. More recently, Cawthorne & Walraven (1998) modelled the Bushveld as a chamber filled in 14 stages, the first stage (part of the Lower Zone) being 1 km thick (leaving cumulates 0·25 km thick) at 1300°C, emplaced 3 km below the surface. Their model predicts that this layer would have taken 0·25 kyr to crystallize; by their calculations the entire complex was emplaced over 75 kyr and took 200 kyr to crystallize.
We constructed a simple finite-difference model to predict heat flow, assuming conductive transfer, from the Lower Zone of the RLS into the aureole, based on the model of Wartho et al. (2001), using MATHCAD 2001 software. We assumed that a mafic magma sheet, 1500 m thick, was intruded into country rock initially at 125°C over a period of 1 Myr, and that, during this time