You are on page 1of 46

Earth-Science Reviews 70 (2005) 1 46 www.elsevier.

com/locate/earscirev

Archaean atmospheric evolution: evidence from the Witwatersrand gold fields, South Africa
Hartwig E. Frimmel*
Department of Geological Sciences, University of Cape Town, Rondebosch 7701, South Africa Received 6 July 2004; accepted 12 October 2004

Abstract The Witwatersrand gold fields in South Africa, the worlds largest gold-producing province, play a pivotal role in the reconstruction of the Archaean atmosphere and hydrosphere. Past uncertainties on the genetic model for the gold caused confusion in the debate on Archaean palaeoenvironmental conditions. The majority of Witwatersrand gold occurs together with pyrite, uraninite and locally bitumen, on degradational surfaces of fluvial conglomerates that were laid down between 2.90 and 2.84 Ga in the Central Rand Basin. Although most of the gold appears as a precipitate within, or associated with, postdepositional hydrothermal phases and along microfractures, available microtextural, mineralogical, geochemical and isotopic data all indicate that this hydrothermal gold, analogous to some pyrite and uraninite, was derived from the local mobilisation of detrital particles. Some of the key pieces of evidence are a significant correlation of the gold, pyrite and uraninite with other heavy minerals as well as sedimentary lithofacies, local preservation of in-situ gold micronuggets and abundant rounded forms of pyrite and uraninite, compositional heterogeneity on a microscale of the gold as well as the rounded pyrite and uraninte, and radiometric age data that indicate an age of the gold, pyrite and uraninite that is older than the maximum age of deposition for the host sediment. None of these observations/data is compatible with any of the suggested hydrothermal models, in which auriferous fluids were introduced from an external source into the host rock succession after sediment deposition. In contrast, those arguments, used in favour of hydrothermal models, emphasise the microtextural position of most of the gold, which highlights the undisputed hydrothermal nature of that gold in its present position, but does not explain its ultimate source. Furthermore, the macro-scale setting of the stratiform ore deposits is in stark contrast to any known type of epigenetic, hydrothermal gold deposit. Consequently, the best-fit genetic model involves post-depositional textural and mineralogical modification of original fluvial placer deposits. On the Kaapvaal Craton, Witwatersrand-type mineralisation is recorded over an extended period of time from 3074 to 2642 Ma. Rounded pyrite is common in the coarser grained fractions of the siliciclastic basin fill. A lack of sulphur isotope fractionation and typical magmatic d 34S values support its detrital origin. Together with rounded uraninite, which is particularly abundant in the older beds, it provides important constraints on the redox potential of the Meso- to Neoarchaean (3.12.6 Ga) atmosphere and hydrosphere. In combination with eukaryotic steroids documented from the Pilbara Craton, Australia, the

* Current address: Institute of Mineralogy, University of Wqrzburg, Am Hubland, D-97074 Wqrzburg, Germany. Tel.: +49 931 888 5420; fax: +49 931 888 4620. E-mail addresses: hartwig.Frimmel@mail.uni-wuerzburg.de, hef@geology.uct.ac.za. 0012-8252/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.earscirev.2004.10.003

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

ambient Neoarchaean oxygen fugacity is calculated as having been approximately 103, in equilibrium with a relatively acidic hydrosphere (pH=6). This is in agreement with the preservation of mass-independent S isotope fractionation, which provides independent support for an anoxic atmosphere and which has so far been recorded predominantly from sediments older than 2.3 Ga. An acidic meteoric palaeoenvironment is supported by intense chemical weathering below erosional unconformity surfaces in the Witwatersrand Basin. In contrast to the pyrite-bearing fluvial and near-shore shallow marine deposits, marine shale deposits contain magnetite. This supports the postulated reducing environment but also highlights total sulphur concentrations in the ancient ocean that were orders of magnitude lower than in modern ocean water. D 2004 Elsevier B.V. All rights reserved.
Keywords: Witwatersrand; Palaeoplacer deposits; Anoxic Archaean atmosphere; Gold genesis

1. Introduction The basic conditions for the most important aspects of modern life on Earth have been shaped as early as in Precambrian times. These include most likely the origin of life and definitely the evolution from the origin of the eukaryote cell to hard bodied animals, the growth of continental crust, the oxygenation of the atmosphere as well as the formation of mineral deposits without which modern civilisation would be unthinkable. All of these aspects appear to be interrelated but the various feed-back mechanisms that controlled the relationships between lifes evolution, plate tectonic processes, palaeoclimate and distribution of metals between hydrosphere and geosphere throughout the critical periods of the Precambrian remain highly speculative. In fact, the determination of the above relationships can be regarded as a fundamental problem in earth sciences. One of the more important questions concerns the timing and cause of the rise of atmospheric O2. Photosynthesis was the principal process by which free O2 was produced, though some of it may also be ascribed to hydrogen escape to space after CH4 photolysis (Catling et al., 2001). Consequently, the evolution of the Archaean (3.82.5 Ga) to Proterozoic (2.50.54 Ga) atmosphere must have been strongly influenced by organisms. Production and decomposition of organic matter, in combination with the volcanic degassing of the planet, also influenced the atmospheric CO2 and CH4 concentrations through a carbon cycle that involves chemical weathering and formation of silicates and carbonates. As both CO2 and CH4 are greenhouse gases, the evolution of life and that of Archaean to Eoproterozoic (2.52.05 Ga) palaeoclimate are intimately interlinked.

General agreement seems to exist on substantial O2 levels from around 2.3 Ga onwards, but the preceding atmospheric evolution is a matter of intense debate (Fig. 1). Two competing schools of thought try to explain the environments and mechanisms that controlled the emergence of life: one assumes that life originated under a reducing atmosphere (O2b1 ppm) and that the formation of an oxic atmosphere triggered the emergence of eukarya (Kasting and Siefert, 2002; Knoll, 1992; Rye and Holland, 2000), whereas the other proposes the emergence of oxygenic photo-

Fig. 1. Contrasting models for the evolution of atmospheric chemistry: (A) Models involving a reducing Archaean atmosphere; pO2 evolution curve proposed by (a) Kasting (1987, 2001) and (b) Rye and Holland (2000), remaining curves after Pavlov et al. (2001a) and Kasting (2001); (B) model involving an oxidising atmosphere (Ohmoto, 2004).

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

synthetic organisms as early as 4 Ga with essentially constant atmospheric O2 levels (N10%) since then (Lasaga and Ohmoto, 2002). Towe (1990) argued for a low (0.20.4%) but stable O2 concentration in the Archaean that was maintained by aerobic respiration. This is in contrast to anaerobic carbon cycles that are implicit in any model of a reducing Archaean atmosphere. The debate extends to the question of Archaean palaeoclimate. As the Suns luminosity is estimated to have been about 30% less than today (Newman and Rood, 1977), Earths hydrosphere should have been completely frozen. No geological support exists for such a scenario, however, and elevated greenhouse gas concentrations have therefore been proposed to offset this so-called Faint Young Sun Paradox (Sagan and Mullen, 1972). A dense Archaean atmosphere is thought to have consisted largely of CO2, and the stronger greenhouse gases H2O and CH4. For the time around 3.0 Ga, atmospheric CO2, CH4 and O2 concentrations of approximately 3000, 1000 and b0.1 ppm, respectively, have been suggested by those advocating a reduced Archaean atmosphere (Kasting, 2001; Kasting and Siefert, 2002; Rye and Holland, 2000), which implies that CH4 was the dominant greenhouse gas. In contrast, some workers (Lasaga and Ohmoto, 2002; Ohmoto et al., 1999; Ohmoto, 2004) argue for CO2 having been the principal greenhouse gas at that time at a level of about 10%. As the continents tied up vast amounts of carbon in the form of organic matter, hydrocarbons, and carbonaceous as well as carbonate rocks, their growth must have led to the removal of considerable amounts of carbon from active circulation. In addition to burial of biogenic carbon in sediments, the consumption of CO2 by biota and the oxidation of CH4 by biogenic O2 must have caused a further decrease in the concentration of greenhouse gases throughout the Neoarchaean (2.82.5 Ga) and Proterozoic. This decrease was further exacerbated by a decrease in radiogenic heat production, volcanic activity, and thus a decrease in the emission rate of potential greenhouse gases. The consequential cooling of the Earths surface triggered repeated global glaciations. Owing to their old age, the preservation potential of Archaean glacial deposits is relatively small, but glaciogenic deposits have been reported from Mesoarchaean (3.2 2.8 Ga) sedimentary successions (Young et al., 1998).

They are more common in Proterozoic successions, in which major glaciations are recorded around 2.35 Ga and again repeatedly throughout the Neoproterozoic (1.00.54 Ga). Their close stratigraphic association with marine carbonate successions is suggestive of severe climate fluctuations, which highlight that no equilibrium had been achieved between solar luminosity, greenhouse gas concentrations and global climate. It was by no means only carbon that was removed from active circulation in the biosphere during the Precambrian. Apart from gravitation towards the planets core during the very early stages of Earths history, most heavy metals were removed from biogeochemical circulation by both the growth of continental crust and mineralisation processes. Indications of the latter exist in the form of huge sedimentary Fe, Cu, PbZn, Mn and possibly also U and Au deposits that are characteristic of the Archaean and Proterozoic Aeons (Lambert et al., 1992). To use Fe as an example, the distribution of iron formations, which provide the best direct evidence of Fe-depletion of the ocean water, appears particularly intriguing. Their occurrence in Archaean and Eoproterozoic strata has been explained by lower atmospheric O2 concentrations at those times that made possible a stratified ocean with Fe-rich bottom waters (Trendall, 2002). According to this model, a steady increase in ocean oxygenation towards the end of the Eoproterozoic, as living organisms and photosynthesis became more abundant and effective, would have lowered the interface between Fe-rich anoxic bottom waters and oxygenated waters down to effectively seafloor level and thereby prohibited any further deposition of iron formations. Global ice ages seem to be particularly favourable for the deposition of iron formations when a largely, or possibly completely, ice-covered Earth would have shut down ocean water circulation and led to the development of anoxic, Fe-rich ocean bottom waters (Kirschvink, 1992; Klein and Beukes, 1993). Although a relationship between iron formation, palaeoclimate, palaeolatitude and volcanism is indicated, the nature of this relationship remains elusive and is the subject of on-going debate in the current literature. It appears clear from the above that considerable uncertainty exists regarding the palaeoenvironmental conditions prior to 2.2 Ga. One of the best natural

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

laboratories to study these conditions is the Kaapvaal Craton in South Africa. It hosts, similar to the Pilbara Craton in Australia, an undisturbed sedimentary cover (Transvaal Supergroup) that spans almost the entire Eoproterozoic Era. This sedimentary succession contains prime examples of banded iron formations and glaciogenic diamictite deposits, as well as a number of paleosols that have been studied extensively by numerous researchers and provided most useful constraints on the atmospheric evolution in the Eoproterozoic. One of the more recent results that emanated from these studies is the recognition that already at 2.2 Ga weathering was comparable with modern tropical laterites under an oxic atmosphere (Beukes et al., 2002). The Kaapvaal Craton is also host to some of the worlds finest examples of Palaeoarchaean (3.63.2 Ga) sedimentary successions. These occur in the 3.5 to 3.2 Ga Barberton Supergroup of the Barberton Greenstone Belt as the Fig Tree and Moodies Groups. There the rocks contain evidence of Fe-rich hydrothermal vents forming iron formations on the seafloor (de Ronde et al., 1994), extensive silicification that signals ocean water saturated in SiO2, as well as remnants of early stromatolites (Byerly et al., 1986) and microfossils (Walsh, 1992). Whereas these earliest sedimentary rocks provide minimum constraints on the age of the oldest life forms, and the younger Eoproterozoic successions provide minimum constraints on the timing of an oxic atmosphere covering the planet, the time in between, i.e. the Meso- and Neoarchaean Eras, holds the key for the principal question of atmospheric evolution and its bearing on the further evolution of life following its emergence sometime in the early Archaean. The best available geological record to study the palaeoenvironmental conditions during that crucial time span is the siliciclastic sediment fill of the Mesoarchaean Witwatersrand Basin on the Kaapvaal Craton. It represents the worlds best preserved Archaean sedimentary succession and it contains redox-sensitive minerals that might hold the key for our understanding of the Mesoarchaean atmosphere and hydrosphere. Iron oxides are conspicuously lacking in fluvial sediments of the Witwatersrand, with the principal Febearing phase being pyrite. The coarse-grained fraction of these pyrite-rich, fluvial deposits (con-

glomerate) hosts the worlds largest known accumulation of gold, but also represents its largest unmined inferred uranium resource. More than 49,400 metric tonnes (t) of gold have been produced from these conglomerate beds (reefs) between 1886 and 2003, amounting to almost 40% of all the gold ever mined during recorded history (Frimmel and Minter, 2002; Sanders et al., 1994). South Africa is still the worlds number one gold producer with a share of approximately 16% and, according to the South African Chamber of Mines, the remaining reserves in the Witwatersrand Basin, estimated around 38,000 t (Frimmel and Minter, 2002), amount to 46% of known world reserves. Between 1952 and 1975, as much as 1.5 t of U 3 O 8 were produced from Witwatesrand conglomerates at an average grade of 271 ppm (Frimmel et al., in press). Although poorly exposed, the Witwatersrand is one of the bestdocumented basins of its kind in the world thanks to more than 100 years of underground mining and exploration. Yet, in spite of the enormous economic significance of the Witwatersrand gold deposits, the genesis of these deposits is still a matter of controversy. Comparing the extensive older with the more recent literature (for a review of the older and more recent literature see Pretorius, 1975 and Frimmel and Minter, 2002, respectively), it appears fair to say that the debate around this controversy has not lost anything of its intensity since it had been described by Davidson (1965) as bthe most disputed issue in the history of economic geologyQ. Depending on the preferred genetic model for the Witwatersrand gold, the pyrite and uraninite, both of which occur predominantly as rounded particles, are interpreted either as sedimentary heavy sands or as hydrothermal precipitates. In the latter case, the abundant rounded pyrite is explained as pseudomorphic replacement of detrital FeTi oxides, Fe-pisolite, ferricrete, banded iron formation and Fe-rich shale (Phillips and Law, 2000) and/or product of postdepositional dissolution and re-precipitation mechanisms (Barnicoat et al., 1997; Phillips and Myers, 1989). The different genetic models have equally different ramifications for the inferred redox state of the atmosphere at the time of sediment deposition as they imply either Fe-sulphides or Fe-oxides as having been stable under the Archaean atmosphere. One of the major goals of this paper is therefore, to review

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

and assess the various genetic models that have been proposed for the Witwatersrand gold deposits, and by implication also for the associated pyrite and uraninite. It will be shown that the best-fit genetic model is that of modified palaeoplacer deposits. The significance thereof for the reconstruction of the Meso- to Neoarchaean atmosphere will then be discussed. The siliciclastic successions of the Witwatersrand are characterised by an abundance of erosional unconformity surfaces. Theoretically, geochemical studies across these unconformities should make it possible to gain insight into the extent of chemical weathering at the time. Similarly as with the debate around the genesis of the gold and the associated pyrite and uraninite, a difference in opinion exists with regard to the cause of widespread acidic alteration of the siliciclastic mineral assemblages that has been reported from throughout the Witwatersrand Basin (Barnicoat et al., 1997; Phillips and Law, 1994). Weathering under an aggressive, CO2 and/or CH4rich, acidic atmosphere should lead to considerable chemical change in the weathered rock (loss of alkalies and alkaline earths). Unfortunately, such a chemical change would be difficult to distinguish from acid leaching by post-depositional magmatic or metamorphic fluids. Thus the question arises whether systematic chemical changes observed across the unconformities reflect paleosols or post-depositional hydrothermal infiltration. Those workers who prefer a hydrothermal model for the gold, pyrite and uraninite, postulate basin-wide H+-metasomatism after sediment deposition and link this with the formation of the gold, pyrite and uraninite during hydrothermal infiltration (Barnicoat et al., 1997). Others concluded from geochemical and mineralogical studies of profiles across stratigraphic units that metamorphism was essentially isochemical, except for potassium (Sutton et al., 1990). A further aim of this contribution is, therefore, to assess this apparent discrepancy by providing alteration profiles across various siliciclastic Witwatersrand units and discussing the significance of trends in the calculated chemical index of alteration. Finally, the distribution of redox-sensitive and pHsensitive minerals, such as Fe-sulphides versus Feoxides, uraninite verus brannerite or feldspars versus pyrophyllite will be examined across the various stratigraphic units of the Witwatersrand Basin. Trends

in the distribution of these minerals along stratigraphic directions as well as between different lithofacies will be discussed in terms of their implications for the palaeoenvironmental conditions. In summary, the overall goal of this paper is to provide: (1) an up-todate review of the evolution of the Witwatersrand Basin fill; (2) a best-fit genetic model for the worlds largest gold province and thus also of for the redoxsensitive monitor phases pyrite and uraninite; and (3) constraints on the likely Mesoarchaean to Eoproterozoic atmospheric and hydrospheric conditions based on the spatial and temporal distribution of redoxsensitive minerals.

2. Geological setting The Witwatersrand Basin occupies a central position on the Archaean Kaapvaal Craton (Fig. 2). The causes of its development and early evolution are linked with tectonic processes in and around this craton, whose history is subdivided into two main periods (de Wit et al., 1992). The first period (3.64 3.08 Ga) saw the initial formation of the continental lithosphere of the craton, with a major pulse of accretion around 3.2 Ga. The second period (3.08 2.64 Ga) was dominated by the development of intracontinental basins and likely subduction-related magmatism along the edge of the Kaapvaal Craton. By the end of the Archaean Eon, the Kaapvaal Craton had amalgamated with the Zimbabwe Craton along the Limpopo Belt (Fig. 2). 2.1. Pre-Witwatersrand basement The oldest known crustal fragment of the Kaapvaal Craton is the ~3.64 Ga old Ancient Gneiss Complex in Swaziland (Kroner and Tegtmeyer, 1994). Products of Palaeoarchaean crust formation are well preserved within the Barberton greenstone belt, where dominantly basicultrabasic magmatism between 3.49 and 3.42 Ga followed tonalite emplacement between 3.55 and 3.52 Ga (de Ronde and de Wit, 1994). The mafic to ultramafic rocks have been interpreted as remnants of ocean-like lithosphere (de Wit et al., 1987). The upper clastic part of the greenstone belt is interpreted to comprise synorogenic deposits (Moodies Group) related to 3.2 Ga accretion. A change from compres-

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

Fig. 2. Distribution of the main Archaean stratigraphic units of the Kaapvaal Craton. The Witwatersrand Basin fill comprises the West Rand and Central Rand Groups; also shown is the outline of the three crustal blocks that are believed to have amalgamated by 2.8 Ga to form a single craton (modified from Schmitz et al., 2004).

sional to transtensional tectonic activity around 3.1 Ga was accompanied by widespread orogenic gold mineralisation along late shear zones. This inversion in the overall stress field marked the beginning of intracontinental basin formation that eventually led to the formation of the Witwatersrand Basin. 2.2. Witwatersrand basin development The first stage of sediment deposition in what eventually became the Witwatersrand Basin is recognised in the Dominion Group. This group comprises an up to 2250 m thick bimodal volcanic succession with a thin basal siliciclastic unit for which a continental rift basin of unknown extent is assumed. Maximum and minimum age constraints are given by

UPb zircon ages of 3086F3 (the youngest age of pre-Dominion basement: Robb et al., 1992) and 3074F6 Ma (the age of volcanism: Armstrong et al., 1991), respectively. The basal siliciclastic unit includes a conglomerate bed with abundant uraninite and pyrite but relatively low gold content (Dominion Reef). Palaeocurrent data consistently point to a source area to the north or northeast (Frimmel and Minter, 2002). Thus, a continuation of the original basin to the south of the present distribution of Dominion Group rocks is likely (Fig. 3). The development of the proper Witwatersrand Basin followed with a hiatus of almost 100 million years. This great time gap implies that the tectonic regime for the Dominion Basin is unrelated to that of the subsequent Witwatersrand Basin. No agreement

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

Fig. 3. Simplified surface and subsurface geological map of the Witwatersrand Basin, also showing the distribution of Archaean granitoid domes, the location of the gold fields, major faults and palaeocurrent directions of reefs in the Central Rand Group (from Frimmel and Minter, 2002).

exists on the detailed lithostratigraphic correlation between the various gold fields across the Witwatersrand Basin, but a broad subdivision of the basin fill into the West Rand and Central Rand Groups, both constituting the Witwatersrand Supergroup, has long been recognised (Fig. 4). 2.2.1. West Rand Group The metasedimentary rocks of the West Rand Group (Fig. 3) rest with angular unconformity above the Dominion Group volcanic rocks. The group attains a maximum thickness of 5150 m in the Klerksdorp gold field and thins to the northeast. No information is available from the Welkom gold field. In the most distal section south of the Vredefort Dome, about 2000 m of West Rand Group rocks have

been intersected in a borehole (Stevens and Preston, 1999). Sediment input throughout the West Rand Group was consistently from the north or northeast (Frimmel and Minter, 2002). UraniumPb data from detrital zircon grains provide a maximum age of 2985F14 Ma for West Rand Group sedimentation (Kositcin and Krapez, 2004). A minimum age for most of the group is given by the Crown Formation lava, the only volcanic unit within the succession, for which a UPb single zircon age of 2914F8 Ma has been obtained (Armstrong et al., 1991). Three subgroups are distinguished based on varying shale/sandstone ratios and basin-wide disconformities (Fig. 4). Within the basal Hospital Hill Subgroup, the shale/sandstone ratio decreases up-section, with the sandstone being predominantly quartz arenite,

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

Fig. 4. Generalised stratigraphic column for the Witwatersrand Supergroup (from Frimmel and Minter, 2002); also shown are the stratigraphic positions of the main auriferous conglomerate beds (reefs) and their relative significance as gold producers (insert); average gold grade typical of mined reefs from Frimmel and Minter (2002); SHRIMP UPb or PbPb age data from (1) zircon (Armstrong et al., 1991), (2) zircon and xenotime (Kositcin and Krapez, 2004) and (3) authigenic xenotime (Kositcin et al., 2003); Chemical Index of Alteration from Gartz and Frimmel (1999), Sutton et al. (1990) and H.E. Frimmel (unpubl. data).

which is interpreted as subtidal deposits (Eriksson et al., 1981). Evidence of tidal deposits has been reported from several units. Thickthin pairs of siltstoneshale couplets within the upper Coronation

Formation (Fig. 4) are a particularly good example (Eriksson and Simpson, 2004). Higher up in the succession, feldspathic sandstone and quartz wacke become more abundant (Law et al., 1990). The

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

sediment record of the West Rand Group reflects fluctuations between distal fluvio-deltaic and shoreface to offshore environments, ascribed by some workers to eustatic sea level changes (Stanistreet and McCarthy, 1991). Indirect evidence of eustatic sea level changes comes from the presence of two diamictite beds within the Government Subgroup, which are possibly correlatives of the oldest known glacial deposit on Earth in the Pongola Supergroup (Young et al., 1998). Of particular significance is also the presence of several magnetic shale beds, because they contain magnetite as principal Fe-phase and not Fe-sulphides as in the coarser grained units. These shale beds have been described from the northern and eastern parts of the basin, as well as from the Klerksdorp gold field (Tankard et al., 1982), where they occur at a number of stratigraphic levels throughout the West Rand Group (Fig. 4). Geophysical maps showing aeromagnetic anomalies (Corner and Wilshire, 1989) that are caused by the strong response of up-turned magnetiterich shale beds along the basin margin indicate a basin-wide distribution of this lithotype. 2.2.2. Central Rand Group The Central Rand Group lies unconformably above the West Rand Group and reaches a maximum thickness of 2880 m near the centre of the basin. Similar to the West Rand Group, a series of cycles can be distinguished, each of which comprises fluvially dominated coarse-grained siliciclastic metasedimentary rocks above an erosion surface, but in contrast to the West Rand Group, shale (mudstone) plays a subordinate role. Fluvio-deltaic processes dominated sediment deposition, but tidal reworking has been suggested for the interfaces between fluvial and shallow marine systems (Els, 1998). A particularly fine example of siltstone-shale couplets, that have been interpreted as dominant and subordinate semidiurnal tidal currents, respectively (Eriksson and Simpson, 2004), occur in the uppermost Randfontein Formation (Fig. 4). During deposition of the Central Rand Group, the palaeoslope direction changed from a consistent dip to the south or southwest to east and northeast along the western and southwestern margins and to the southeast and south along the northwestern and northern margins (Minter and Loen, 1991). Two minor, locally developed, amygdaloidal basalt units

(Bird lava) in the Krugersdorp Formation provide the only evidence of volcanism in the Central Rand Group. The Central Rand Group is subdivided into the lower Johannesburg and the upper Turffontein Subgroups. The maximum age of deposition for the sediments of the group is constrained by the youngest age obtained on detrital zircon from the bottom of the Johannesburg Subgroup (i.e. 2902F13 Ma: Kositcin and Krapez, 2004). The youngest detrital zircon age of 2872F6 Ma from the Krugersdorp Formation sets the best available constraint for the age of the top of this subgroup, whereas 2849F18 Ma based on detrital zircon, or 2840F3 Ma based on detrital xenotime, represents the maximum age for the top of the Central Rand Group (Kositcin and Krapez, 2004). A minimum age constraint on deposition of the Central Rand Group is 2780F3 Ma, which is the oldest age obtained on any authigenic mineral (xenotime) to date (Kositcin et al., 2003). 2.2.3. Tectonic setting of the Witwatersrand Basin General agreement seems to exist on the lower part of the West Rand Group having been deposited in a passive continental margin setting, facing an open ocean to the south. Continental rift-related rhyolite at the bottom of the Pongola Supergroup (Nsuze Formation, Fig. 1), considered a correlative of the Witwatersrand Supergroup (Beukes and Cairncross, 1991), has an age of 2985F11 Ma (Hegner et al., 1994), equivalent to that of the lower West Rand Group. The inferred passive margin setting is therefore explained by post-rift thermal subsidence. A change from upward-deepening to upwardshallowing at the Hospital Hill/Government Subgroup boundary has been used to suggest a change from a passive margin to a foreland basin setting (Coward et al., 1995). Others prefer a passive margin setting for the entire group (de Wit et al., 1992). In a recent SHRIMP detrital zircon study, a general decrease in the complexity of zircon provenance age spectra is recorded up-section through the West Rand Group (Kositcin and Krapez, 2004). This is interpreted as reflecting an increasing maturity of the hinterland as is expected for an evolving passive margin, thus supporting a thermal subsidence origin for the entire West Rand Group.

10

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

The variation in palaeoslope directions in the lower Central Rand Group, together with the increase in continental to marine sediment ratio up-section, indicates a change to a progressively shrinking continental basin. This is well reflected by the detrital zircon age spectra, which increase in complexity upsection through the Central Rand Group. A progressively greater variety of source rocks must have been eroded through Central Rand Group times, reflecting tectonic loading and thus a foreland basin setting for this group. The same study also suggested that granitoids were contemporaneously unroofed. Such a provenance is more consistent with a back-arc foldthrust belt than with a foreland fold-thrust belt, which led Kositcin and Krapez (2004) to postulate a retroarc basin setting for the Central Rand Group. With considerably more and better age data available, more reliable integrated sedimentation rates can be calculated. Interestingly, these indicate a minimum of 63 m/my for the West Rand Group, considerably higher than 18 m/my calculated for the Central Rand Group. Such a difference seems to be inconsistent with a simple retroarc foreland basin and would point to a significant additional strike-slip component, as suggested by Maynard and Klein (1995). Stanistreet and McCarthy (1991) have already emphasised both sinistral and dextral strike-slip along the northern and western margins, respectively, and postulated a southeast-directed tectonic escape basin. However, some of the faults referred to by these authors are post-Witwatersrand in age. Alternatively, the lower integrated sedimentation rate calculated for the Central Rand Group might simply reflect a higher degree of sediment re-working, and thus less accommodation space, and not necessarily lower actual sedimentation rates. By analogy with the Pongola Supergroup, a major folding event in upper Central Rand Group times between 2837F5 and 2824F6 Ma, defined by preand post-folding felsic intrusive rocks, has been suggested (Gutzmer et al., 1999). Along the western basin margin, the angles of unconformity increase upsection in the Central Rand Group, which reflects progressive uplift of the hinterland to the west (Frimmel and Minter, 2002). From the above tectonic synthesis, it becomes apparent that the generally used, deeply entrenched term bWitwatersrand BasinQ can be misleading, because it embraces at least two different

basin types, with a major sequence boundary between the West Rand and Central Rand Groups. Therefore the Witwatersrand Basin is best described as a successor basin that contains erosional remnants of tectonically stacked sediments originally deposited in at least two fundamentally different tectonic settings. The two mafic volcanic units within the Witwatersrand Supergroup remain somewhat enigmatic, as neither passive margins nor foreland basins are particularly favourable settings for such volcanism. Basaltic volcanism on passive margins can be caused by reactivation of fundamental faults as a far-field response to changes in the rate or direction of plate tectonic processes or plume magmatism. By implication the Crown Formation lavas might record a change from thermal to reactivated subsidence. In contrast, the Bird volcanic interval in the Upper Central Rand Group could be impactogenic, implying a change from a subduction- to collision-related, retroarc foreland domain (Stanistreet and McCarthy, 1991). Based on the palaeoslope directions, the main tectonic domains in the hinterland controlling the style of sedimentation in the Witwatersrand Basin are to the north and west, although palaeoslope directions in the Evander gold field are more locally controlled. Along the northern margin of the Mesoarchean Kaapvaal Craton, east- to northeast-trending greenstone belts (Murchison, Pietersberg and Giyani greenstone belts, Fig. 2), surrounded by granitoids, yielded single zircon age data that overlap with the age of Witwatersrand sediment deposition (Brandl et al., 1996; Kroner et al., 2000; Poujol, 2001; Poujol and Robb, 1999; Poujol et al., 1996). All of these belts seem to have a 3.23.3 Ga basement. In the Murchison granitoidgreenstone terrain, metamorphosed granite, tonalite and rhyolite are dated between 3.02 and 3.09 Ga, overlapping with deposition of the Dominion Group rocks. Felsic volcanic rocks and granite from the same terrain, dated between 2971 and 2966 Ma, might reflect crustal thinning related to early Witwatersrand rifting. Younger felsic intrusions in the Giyani and Pietersberg granitoidgreenstone terrains are dated at 2874 Ma, whereas, in the Murchison terrain, granitoid emplacement is indicated around 2901 and 2820 Ma. Shear-zone hosted gold deposits occur in the mafic to ultramafic rocks of these greenstone belts. Only the Pietersberg Belt hosts

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

11

a sequence of predominantly coarse-grained siliciclastic metasedimentary rocks with minor palaeoplacer gold occurrences (Uitkyk Formation) that are possibly correlatives of the Witwatersrand Supergroup (de Wit et al., 1992). A series of intrusive events has also been identified in the AmaliaKraaipan granitoidgreenstone terrain along the western margin of the craton (Anhaeusser and Walraven, 1999; Poujol et al., 2002; Robb et al., 1992; Schmitz et al., 2004). A minimum age for the supracrustal successions is given by a UPb zircon age of 3033F1 Ma obtained for a mafic intrusive body. Ages between 2932 and 2926 Ma for metamorphic and anatectic zircon date both the time of accretion of a ca. 2930 Ma volcanic arc and continental collision between the so-called Kimberley and Witwatersrand crustal blocks (Schmitz et al., 2004). Deep seismic reflection profiles through the west-central Kaapvaal Craton indicate a westwarddipping subduction (de Wit and Tinker, 2004). Subsequent crustal thickening and uplift led to variable exhumation of, and decompression melting in, the Kimberley block, with resulting high-level granitoids as young as 2727 Ma, while the Witwatersrand block underwent subsidence with sedimentation in the Central Rand Basin. The age of the youngest granitoids in the Kraaipan Belt (around 2790 Ma) is similar to the age of the Gaborone Suite rapakivi-type granite and related volcanic rocks in southern Botswana (2783 Ma; Grobler and Walraven, 1993; Moore et al., 1993). Considering the inferred retroarc foreland setting and above age constraints for the Central Rand Group, the younger granitoids in the hinterland to the north, northwest and west might reflect the corresponding magmatic arc, possibly related to a southward-dipping subduction zone. The ocean that was closing at that time probably separated the amalgamated WitwatersrandKimberley block from the Pietersburg block (Fig. 2). Such a combination of foreland basin, related to westward subduction, and retroarc basin, related to the closure of an ocean further north, would help explain the apparent inconsistency between integrated sedimentation rates and a simple retroarc basin mentioned above. The Limpopo orogeny, often used as an explanation for the inferred foreland position of the Central Rand Basin in the past (e.g. Burke et al., 1986), took place more than 100 million years after

the Witwatersrand Supergroup rocks had been laid down. Apart from great ambiguity about the existence of a Himalayan-style Limpopo orogeny, the available age data rule out any relationship between tectonic events that shaped the Limpopo Belt and the Witwatersrand Basin. 2.3. Post-Witwatersrand evolution of the Kaapvaal Craton Mild deformation in the form of block faulting and subordinate folding and thrusting along the western and northern basin margin, and low-grade regional metamorphism of the Witwatersrand Basin fill, are testimony to post-depositional alteration related to a series of tectono-thermal events initiated during, and succeeding, Witwatersrand sediment deposition. Stratigraphically above the Witwatersrand Supergroup follows the Ventersdorp Supergroup, which attains a thickness of more than 3600 m in places. In the northern parts of the basin, the two supergroups are separated by a regional angular unconformity that is overlain by a thin fluvial conglomerate (Venterspost Conglomerate Formation). In places, the conglomerates are highly auriferous and form important ore bodies (Ventersdorp Contact Reef) that are similar to other Witwatersrand reefs. Further south, in the Welkom gold field, the Ventersdorp Supergroup rests paraconformably above the Witwatersrand Supergroup and the Ventersdorp Contact Reef is not developed (Minter et al., 1986). This is explained by a lack of re-working of the older, auriferous Witwatersrand sediments along that contact there. The thin siliciclastic sequence of the Venterspost Conglomerate Formation is conformably overlain by a thick succession of flood basalt (Klipriviersberg Group), though localised erosional scours exist and have been explained by contemporaneous eruption of channelised lava flows over unconsolidated sediment (Hall, 1997). A UPb zircon age of 2714F8 Ma is so far the best constraint on the timing of eruption (Armstrong et al., 1991), which implies a hiatus of at least 50 million years for the underlying unconformity. This is supported by the youngest detrital xenotime age obtained on the Ventersdorp Contact Reef, i.e. 2729F19 Ma (Kositcin and Krapez, 2004). The Klipriviersberg Group basalt layers are overlain by siliciclastic and volcanic, predominantly felsic

12

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

and minor mafic, deposits of the Platberg Group. The siliciclastic sediments reflect alluvial fan deposits that prograded into lacustrine environments. A UPb zircon age of 2709F4 Ma for Platberg Group basalt (Armstrong et al., 1991) shows that the bimodal volcanic activity recorded in the Ventersdorp Supergroup was short-lived. Deposition in a continental rift that evolved into an Atlantic-type continental margin is inferred from seismic profiles that indicate considerable thickening of the Ventersdorp Supergroup towards the west of the craton (Tinker et al., 2002). A link between Ventersdorp rifting and collision between the Kaapvaal and Zimbabwe Cratons, as suggested previously (Burke et al., 1985), is unlikely. Granulite facies metamorphism associated with southward thrusting of the Southern Marginal Zone of the Limpopo Belt on to the Kaapvaal Craton is dated at 2691F7 Ma Ma (Kreissig et al., 2001) and syntectonic granites from the Central Zone of that belt range in age from 2664 to 2572 Ma (McCourt and Armstrong, 1998), later than onset of Ventersdorp flood basalt extrusion. However, the age of the Klipriviersberg Group overlaps with an earlier northward thrusting phase, dated at 2729F19 Ma (Passeraub et al., 1999), that affected greenstone belts along the northern Kaapvaal Craton. The Klipriviersberg extension could thus be explained by southwarddirected subduction beneath the Kaapvaal Craton prior to continental collision. Alternatively, the ultimate cause of Ventersdorp rifting might not reside in crustal plate tectonics but as a result of a mantle plume (Hatton, 1995). A second thrust event in the Witwatersrand Basin, post-Platberg and pre-Transvaal in age (Roering, 1990), is most likely a distant expression of the collisional tectonic processes in the Southern Marginal Zone of the Limpopo Belt. At that stage, lower greenschist facies metamorphic conditions were attained for the first time in the Witwatersrand rock column, at least along the northern margin of the basin (Coetzee et al., 1995). This phase of compression and uplift was followed by regional peneplanation. The corresponding unconformity represents a major sequence boundary, separating the Ventersdorp from the overlying Transvaal Supergroup. Of significance to the debate on the genesis of the Witwatersrand gold and of Archaean atmospheric evolution is the presence of a thin, but laterally

extensive, basal conglomerate and sandstone unit (Black Reef Quartzite Formation) that rests on the post-Ventersdorp erosion surface. This basal siliciclastic unit contains a conglomerate (Black Reef) that is almost indistinguishable from the auriferous, uraniferous and pyrite-rich Witwatersrand reefs, except for a lower metamorphic grade. A synWitwatersrand or syn-Ventersdorp age for the Black Reef, as suggested by Phillips et al. (1997) is untenable, because of field relationships (Els et al., 1995) and geochronological data. The Black Reef Quartzite Formation reflects fluvial sedimentation in channels on a locally deeply incised palaeosurface, braided fluvial and braid delta deposits that grade into shallow marine deposits. A locally transitional and conformable upper contact with the overlying predominantly chemical, marine sedimentary successions of the Chuniespoort Group is unequivocal evidence of an early Transvaal Supergroup age for this genetically important auriferous formation (SACS, 1980). An indirect age constraint for the Black Reef Quartzite Formation exists in the form of a PbPb single zircon age of 2642F2 Ma (Walraven and Martini, 1995), obtained for a lava in a stratigraphic correlative (Vryberg Formation). In agreement with the above timing of tectonic activity in the Limpopo Belt, the sediment sources were located to the north (Barton and Hallbauer, 1996; Els et al., 1995). The overlying Chuniespoort Group starts with a transgressive black shale, followed by thick platform carbonates (Malmani Subgroup), banded iron formation (Penge Formation) and, erosive into the older strata and reflecting a eustatic sea level fall, lacustrine deposits including a glaciogenic diamictite (Duitschland Formation). An upper age constraint is given by a PbPb single zircon age of 2550F3 Ma for a tuff layer in the lowermost formation of the Malmani Subgroup (Walraven and Martini, 1995), whereas UPb single zircon ages of 2432F31, 2465F7 and 2480F6 Ma reported for the stratigraphic equivalent of the Penge Formation (Nelson et al., 1999; Walraven and Martini, 1995) set some constraints on the minimum age of the group. Subsidence rates calculated from the thickness of the group are in good agreement with thermal subsidence due to lithospheric cooling after the Ventersdorp thermal anomaly (Tinker et al., 2002). A further major sequence boundary separates the Chuniespoort Group from the overlying Pretoria

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

13

Group. These two groups, though unified as Transvaal Supergroup in the literature, have little in common. The boundary is marked by a strongly weathered erosional unconformity reflecting a hiatus of ~80 million years, above which lies a volcano-sedimentary succession with alluvial fan and fan delta deposits with minor marine influence. This hiatus is likely to reflect a glacio-eustatic sea level fall, because the unconformity is overlain by glaciogenic diamictite. Alluvial to lacustrine deposits grade into shallow to deep marine deposits before a further erosional unconformity at the top of the Timeball Hill Formation marks another sequence boundary. The unconformity above that formation contains the paleosol examples that have been used to constrain the rise in atmospheric O2 during the Eoproterozoic (Beukes et al., 2002). Deposition of the Pretoria Group was followed after another hiatus by the contemporaneous extrusion of felsic volcanic rocks (Rooiberg Group), and emplacement of the mafic to ultramafic 2059F1 Ma (Buick et al., 2001) Rustenberg Suite and 2054F2 Ma (Walraven and Hattingh, 1993) granitic Lebowa Suite, both of the Bushveld Igneous Complex, to the north of the Witwatersrand Basin. An elevated crustal geotherm related to large-scale magmatic underplating and intraplating also probably resulted in the intrusion of alkali granite and mafic plutons within the Witwatersrand Basin, some dated at 2078F12 Ma (Moser, 1997). The last major disturbance experienced by the Witwatersrand strata is reflected by a 30km-wide domal structure, the Vredefort Dome (Fig. 3), that could well represent the oldest (2023F2 Ma; Kamo et al., 1996; Moser, 1997) and largest (250 to 280 km diameter; Henkel and Reimold, 1998) known terrestrial impact structure. 2.4. Post-depositional alteration of the Witwatersrand Supergroup The multitude of tectono-thermal events that affected the Kaapvaal Craton after the deposition of the Witwatersrand Supergroup sediments is reflected by recrystallisation, formation of metamorphic mineral assemblages, and likely also metasomatic reactions, thus masking geochemical and mineralogical characteristics of the sedimentary protolith. A series of detailed petrological, geochronological and fluid

inclusion studies made it possible to distinguish between several stages of post-depositional alteration throughout the basin. Initial fluidrock interaction most probably involved leaching by basin-wide penetration of meteoric water shortly after sediment deposition. A prime Phanerozoic analogue of this process has been documented for the Parana Basin of Brazil (Franca et al., 2003) and a similar scenario is likely to have affected the Witwatesrand Basin fill (Phillips et al., 1990), when uplift of at least one basin margin provided a steep hydraulic gradient for groundwater to flow towards the basin centre. A decrease in pH with increasing burial is predicted from the maturation of organic material and the release of H+ from dehydration reactions. This, in turn, could have led to the dissolution of detrital feldspars, thus generating a secondary porosity for further diagenetic fluid flow. Following diagenesis, dated between 2776 and 2780 Ma (Kositcin and Krapez, 2004), a first stage of lower greenschist facies metamorphism was attained along the northern basin margin, coeval with highgrade metamorphism and tectonism in the Southern Marginal Zone of the Limpopo Belt. In most parts of the basin, regional peak metamorphic conditions of 300 to 350 8C at 3 kbar were likely achieved during deposition of the Pretoria Group (Frimmel and Minter, 2002). Only around the Vredefort Dome were metamorphic grades up to amphibolite facies reached (Gibson and Wallmach, 1995). There the peak of metamorphism is ascribed to the emplacement of the Bushveld Igneous Complex, which is supported by an anomalously high geothermal gradient (Frimmel, 1997). Several stages of enhanced fluid circulation through the Witwatersrand Basin are recognised. These range from diagenetic basin dewatering to a number of hydrothermal fluid infiltration events. Hydrothermal rutile, zircon and xenotime age data are considered most robust, whereas RbSr, UPb and PbPb ages of various hydrothermal precipitates have only limited geochronological meaning (Zartman and Frimmel, 1999). The most reliable of these age data (Armstrong et al., 1995; Kositcin and Krapez, 2004; Robb et al., 1990) cluster around 2720 Ma (Ventersdorp extension), 2580 Ma (early Transvaal thermal subsidence), possibly around 2200 Ma (Pretoria extension), and 2060 Ma (Bushveld event). A further

14

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

hydrothermal infiltration event has been ascribed to the Vredefort impact, based on cross-cutting relationships with dated pseudotachylyte, fluid inclusion and gold chemistry data (Frimmel et al., 1999). Although there is some evidence for post-2.0 Ga hydrothermal alteration from KAr age spectra on very fine-grained white mica (Zhao et al., 1999), none of these younger events, which are far field effects of various stages of accretion along the craton margin, are of great significance for the post-depositional alteration history of the Witwatersrand sediments.

3. Sedimentological and mineralogical characteristics of the orebodies 3.1. Host lithofacies The Witwatersrand gold orebodies typically occur in conglomerate beds (reefs) on unconformities within major stratigraphic sequences (Pretorius, 1981). Gold has been mined from at least 30 such reefs, the most important of which are shown in Fig. 4, with those from the Central Rand Group accounting for 90% of total production. In addition, the base of the younger Ventersdorp and Transvaal Supergroups also host, in places, economic to subeconomic, orebodies of comparable lithology and mineralogy. The orebodies comprise various fluvial lithofacies that range from clast-supported oligomictic conglomerate to loosely packed conglomerate, pebbly arenite, or just pebble lag surfaces associated with trough cross-bedded quartz arenite. In exceptional circumstances, ore is associated with immature debris flow lithofacies. These include black argillite in the Beatrix Reef (Minter et al., 1988), where overlying playa facies covered eroded older ores that occurred as eroded subcrops, and risers between terraces of the Ventersdorp Contact Reef (Henning et al., 1994), where

underlying immature sedimentary rocks were eroded and incorporated into the alluvium during entrenchment. The Witwatersrand reef rocks are generally described as quartz pebble conglomerates. Exceptions exist, however, such as the Steyn Reef in the Welkom gold field, which is polymictic with as much as 30 vol.% yellow quartz porphyry in both the pebble and sand fractions. The Kimberley Reef in the Evander gold field contains as much as 50% chert pebbles (Tweedie, 1986). The average pebble assemblage is 85 vol.% vein quartz, 12 vol.% chert, 2 vol.% quartz porphyry, and 1 vol.% metamorphic clasts (Frimmel and Minter, 2002). The orebodies range in thickness from decimetres to a few metres and are confined between a basal degradation surface, usually an angular unconformity, and an upper planar bedding surface that separates it from overlying quartz wacke or siltstone (Fig. 5A,B; Minter, 1991). They have a lenslike geometry and define fluvial bar-and-channel bedforms that display unimodal palaeocurrent directions (Smith and Minter, 1980). Multichannel sequences of conglomerate and quartz arenite, deposited by repeated flood- and waning-stage flows, make up the thicker orebodies (Fig. 5D). Depositional environments that have been reconstructed from the geometry of the basal degradation surface, coarseness of the sediment, and the distribution of lithofacies, range from proximal alluvial fan (e.g. EA Reefs in the Eldorado Formation), to terraced fluvial deposits (e.g. Ventersdorp Contact Reef; Henning et al., 1994) to braid plains (Composite Reef; Tucker, 1980), to braid deltas that merge with shoreline environments. In some places, aeolian deflation has been suggested to account for the planar top of the orebodies (Minter, 1999), whereas in others, shoreline encroachment and associated wave action, followed by burial beneath fine-grained sediment is

Fig. 5. (A) Oligomictic pebbly quartz arenite (reef), Vaal Reef, Stilfontein mine, Klerksdorp gold field; note the veneer of bitumen on the basal unconformity and 3 cm above the base on a bedding plane defining the top of pebbly layer; cross-bedding (so) in hanging wall, and an in-situ ventifact; scale bar=1 cm. (B) same as A, but highlighting the position of fine- to coarse-grained rounded pyrite. (C) Upward fining imbricate pyrite pebble lags between quartz pebble lags, Basal Reef, Welkom gold field; scale bar=1 cm. (D) Underground exposure of upper half of Ventersdorp Contact Reef, Tau Lekoa mine, Klerksdorp gold field, showing multichannel sequences of conglomerate and quartz arenite; top contact with Klipriviersberg Group mafic lava flow is erosive into originally unconsolidated conglomerate; scale bar=50 cm. (E) Truncated, well rounded, oolitic pyrite (arrow) within pyrite pebble lag shown in C (white rectangle), scale bar=0.5 cm. (F) Concentration of gold particles along cross beds in pebbly quartz arenite at the base of the Basal Reef, Welkom gold field. Three foresets converge with the bottomset toward the right (see Minter et al., 1993); scale bar=1 cm.

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

15

16

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

indicated as in the Basal Reef (Buck, 1983) and Carbon Leader Reef (Buck and Minter, 1985). 3.2. Textural, chemical and isotopic features of the main ore components The bulk of gold, uraninite and pyrite, whose origin is debated, occur together with undoubted allogenic detrital minerals, such as mechanically abraded zircon and chromite grains, and more rarely PGE-minerals and diamond (for a complete list of minerals identified to date in Witwatesrand reefs see Phillips and Law, 2000), on degradation surfaces marked by pebble lags or the base of clast-supported conglomerate beds, cross-bedded foresets, bottomsets, and coset boundaries (Fig. 5F). In thick (12 m) conglomerate units, representing multichannel sequences, the allogenic minerals are concentrated along the basal degradation surface of each graded bed. The highest concentrations of allogenic minerals are found above unconformity surfaces, which reflect a direct relationship to the amount of erosion. Cutand-fill structures, such a trough cross beds, contain greater allogenic mineral concentrations than aggradational structures, such as planar cross beds. Experiments (James and Minter, 1999) have confirmed that the dominant concentration mechanism was by selective entrainment of coarser, less dense particles during bedload transport (Slingerland and Smith, 1986). Average gold grades of mined reefs range from 3 to 25 g/t (Fig. 4). Case studies on the element distribution within reefs, such as the Kimberley Reef (Rasmussen and Fesq, 1973), the Steyn Reef (Frimmel and Minter, 2002), the Vaal Reef (Fox, 2002), and the Ventersdorp Contact Reef (H.E. Frimmel, unpubl. data) illustrate a relatively good positive correlation between U and Au, but only poor correlation between Au and Zr, and Crelements that are controlled by detrital phases (zircon, chromite) that were not susceptible to hydrothermal mobilisation. All data sets show gold enrichment being linked with Zr enrichment, but samples rich in Zr typically do not contain elevated Au contents. A detailed study of different lithofacies in a channel sidebar of the Steyn Reef, Welkom gold field, where 13 aggradation events can be separated (Frimmel and Minter, 2002) showed that degradation surfaces were preferentially mineral-

ised with an average of 38 ppm Au, 410 ppm Zr, 1750 ppm U, and 300 ppm Cr. Higher concentrations of all these elements were recorded in the coarser sediment fraction, reflecting higher flow velocities. Similar observations were also made quantitatively on other reefs, such as the Leader Reef (Smith and Minter, 1980) and the Ventersdorp Contact Reef (H.E. Frimmel, unpubl. data). The Witwatersrand reefs are not only extremely rich in gold but also are one of the worlds largest uranium depositories. Between 1952 and 1975, as much as 1.5106 t U3O8 (Frimmel et al., in press) was produced at an average grade of 271 ppm (CamisaniCalzolari et al., 1984). One of the richest reefs was the Monarch Reef, a small-pebble distal placer, mined at a mean grade of 2860 ppm U3O8. The principal Uminerals are uraninite, brannerite and leucoxene, with a systematic decrease in uraninite/brannerite ratio upsection. For example, in the Welkom gold field, the uraninite/brannerite ratio is 8.7 in the Steyn Reef, whereas in the younger Beatrix Reef it is zero (Minter et al., 1988). Similarly, all the U in the Black Reef, from which a specimen with as much as 3350 ppm U3O8 has been reported (Bourret, 1975), occurs as brannerite. There is a broad systematic trend in the U/Au ratio up-section. In contrast to the Central Rand Group reefs, the Dominion Reef is highly uraniferous but does not contain significant amounts of gold. Within a given stratigraphic unit, both Au and U concentrations decrease from the basin margin towards its centre, but at different rates. A systematic increase in the U/Au ratio down the paleoslope from 103 to 10 was noted in the Welkom gold field, with uraninite being enriched in the more distal facies, probably as a result of hydraulic mineral sorting (Minter et al., 1986). This is also illustrated by the coarse-grained Main Reef, which has an average pebble size of 37 mm and a U/ Au ratio of less than 25 as opposed to the relatively finer grained, more distal Monarch Reef, whose average pebble size is 16 mm and U/Au ratio is 128 (Vennemann et al., 1995). It has been noted in many studies that Witwatersrand gold is intimately associated with carbonaceous matter, which occurs as stratiform seams and as spherical, glassy globules. Both forms have been recognised as metamorphosed solidified hydrocarbons, i.e. pyrobitumen (Gray et al., 1998). The

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

17

bcarbon seamsQ occur preferentially in deposits reflecting distal environments, but are conspicuously absent from proximal, high-energy deposits. Hydrocarbon derivation from original algal mats is suggested by the distribution of the bcarbon seamsQ on palaeosurfaces, on sedimentary accumulation surfaces and on trough cross-beds and ripple surfaces (Buck, 1983; Minter, 1981). In those reefs that contain such carbonaceous matter, such as the Carbon Leader, Basal and Vaal Reefs, the highest Au and U concentrations are located in the zones that are particularly rich in pyrobitumen. Nagy (1993) estimated that about 40% of all mined Witwatersrand gold was hosted by such bitumen seams. On a microscale, the pyrobitumen-filled microfractures typically contain some gold. Oil-migration, and thus by implication gold transport, has been suggested both prior (England et al., 2002a) and during (Jolley et al., 2004) fracturing. It should be noted, however, that many highly auriferous reefs contain little or no carbonaceous matter; a number of economic reefs, such as the Main and Kimberley reefs in the Central and West Rand gold fields do not contain any noteworthy amounts of bitumen. The bitumen is interpreted to have formed by the polymerisation and crosslinking of liquid hydrocarbons around irradiating grains, predominantly uraninite, in the host sedimentary rock (Schidlowski, 1981). Oil-bearing fluid inclusions provide direct evidence of oil migration through the Witwatersrand sedimentary rocks (Drennan et al., 1999; England et al., 2002a). Bitumen derivation from a variety of biomass in a reducing environment, with subsequent short-range hydrothermal mobilisation is indicated by organogeochemical, bulk and molecular C isotopic studies (Spangenberg and Frimmel, 2001). The phases associated with the gold, which are most important for the topic of this paper are pyrite and uraninite. The textural, mineral chemical and isotopic characteristics of these phases will, therefore, be discussed in somewhat greater detail below. 3.2.1. Pyrite Pyrite is the most common heavy mineral in all of the fluvial deposits of the Witwatersrand. Only in marine sedimentary rocks, such as shales in the West Rand Group, are Fe-oxides (predominantly magnetite) found instead of pyrite (Frimmel, 1996). Locally, such

as in parts of the Ventersdorp Contact Reef, Klerksdorp gold field, pyrrhotite is the stable Fe-sulphide instead of pyrite. There the distribution of the two Fesulphides seems to be controlled by locally variable oxygen fugacity of post-depositional fluids. Generally, the pyrite occurs in a number of different textural forms (England et al., 2002b; Hallbauer, 1986; Ramdohr, 1958) that are grouped into (1) rounded, compact, (2) rounded, porous, and (3) euhedral. The rounded compact variety is by far the most abundant form of pyrite in all reefs (Fig. 5C) except for the Ventersdorp Contact Reef, where the pyrite grains are predominantly euhedral. However, etching of these euhedral grains reveals that most have one or more rounded cores, with the euhedral outline being an artifact of secondary, authigenic/hydrothermal overgrowth around pre-existing rounded, compact pyrite cores (Fig. 6I; England et al., 2002b; Frimmel and Minter, 2002). Evidence of mechanical abrasion of the rounded grains is given by truncation of oscillatory growth zonation, defined by variable As contents at grain boundaries (McLean and Fleet, 1989). A crystallographic study (Fleet, 1998) revealed that some of these rounded grains are single crystals and not the polycrystalline or twinned crystals one would expect if they were pseudomorphs after bblack sandsQ as suggested by Phillips and Myers (1989) and Phillips and Law (2000). A further argument against basin-wide sulphidation of bblack sandsQ is that the latter typically comprise titanomagnetite. Pyritisation of such a precursor characteristically leads to intergrowths of minute rutile needles. Such rutile-bearing pyrite pseudomorphs are the exception and not the rule in the Witwatersrand reefs (Ramdohr, 1958). The rounded pyrite grains from the Witwatersrand are, however, devoid of Ti. Associated with this form of pyrite are similarly rounded, compact arsenopyrite and cobaltite particles (Saager and Oberthu 1984; England et al., 2002b). Genetr, ically significant mineral inclusions in rounded pyrite are feldspar (Fig. 6F), calcite, corundum and spessartine, which are absent in the metamorphic mineral assemblage of the metasedimentary host rocks. Examples of several centimetre-thick, almost monomineralic, fining-upward pyrite beds, displaying imbrication with the same orientation as intercalated quartz pebble lags, occur within the Basal Reef of the Welkom gold field (Fig. 5C).

18

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

19

The porous pyrite displays a variety of internal textures, ranging from laminated aggregates, and rounded concretions to oolitic-colloform and dendritic forms. Many concretionary and colloform varieties are fragmented, broken and have their internal structures truncated, from which mechanical transport is inferred (Fig. 5E). Post-sedimentary, euhedral to subhedral pyrite occurs preferentially adjacent to zones of hydrothermal alteration, such as veins and faults. This form is typically associated with other authigenic/hydrothermal sulphides (chalcopyrite, cobaltitegersdorffite, pyrrhotite, galena and arsenopyrite) and pyrobitumen (Gartz and Frimmel, 1999). Euhedral pyrite overgrowths are common and, in places, are contiguous with pyrite that fills fracture or pore spaces. A laser-ablation sulphur isotope study (England et al., 2002b) revealed that the rounded pyrite forms have a wide range in d 34S values (5.0x to +6.7x), not only at the mine and stope-face scale but even at the sample scale over less than 1.5 cm2. In a previous SHRIMP study (Eldrige et al., 1993), large heterogeneities in d 34S (7x to +32x) were noted in single pyrite grains and between different morphological types. Such heterogeneity is difficult to reconcile with precipitation from a geochemically homogeneous hydrothermal fluid, and more likely reflects variation in pyrite from the eroded source rock and/or microbial sulphate reduction in the depositional environment. The heterogeneity in rounded pyrite is in contrast with the narrow range in d 34S values obtained for authigenic/euhedral pyrite (0.5x to +2.5x), which is also distinguished from rounded pyrite by higher Ni and As contents as well as gold inclusions. Of particular significance are two texturally adjacent but isotopically contrasting ooid-like

pyrite grains from the Ventersdorp Contact Reef, which display a strong isotopic zonation but of opposite signs (England et al., 2002b). The d 34S ratios in one of the two grains increase systematically from 4.1x in the core to +1.4x in the rim, whereas those in the other grain decrease from 0.8x in the core to 4.5x in the rim. The former has been interpreted by these authors as indicative of sulphidation of an original sulphate grain. The latter grain reflects a different provenance and highlights that the isotopic differences must be pre-depositional and cannot be due to fluctuations in redox potential of a hydrothermal, sulphidising fluid that mixed with local meteoric formation water during diagenesis as proposed by Phillips and Law (2000). Attempts to date the various forms of pyrite using the UThPb isotope systems (Barton and Hallbauer, 1996; Poujol et al., 1999; Zartman and Frimmel, 1999) were met with mixed success. Authigenic and hydrothermal pyrite is typically enriched in uranogenic Pb, whereas the rounded forms have a less radiogenic isotopic signature. However, absolute Pb Pb ages need to be viewed with caution as the 238U and 235U decay schemes were likely decoupled, presumably by the selective diffusion of 222Rn from uraninite and its subsequent capture in hydrothermal precipitates, leading to erroneous ages (Zartman and Frimmel, 1999). More reliable are ReOs data obtained on rounded, compact pyrite from the Vaal Reef, which yielded an age of 2.99F0.11 Ga (Kirk et al., 2001). This is older than the time of sediment deposition, which provides a strong argument for the detrital origin of much of the pyrite. 3.2.2. Uraninite and leucoxene Most of the uraninite particles are well rounded and enclosed, or partially replaced, by bitumen, thus

Fig. 6. Photomicrographs illustrating morphological and textural features of Witwatersrand gold orebodies (combined transmitted and reflected light, scale bars=0.2 mm, except for scanning electron microscope images C and D: 0.1 mm): (A) Contrasting morphological types of gold particles occurring together on a mm-scale. (B) Gold micro-nugget (Au) with overfolded rims next to rounded pyrite (Py) from same sample as shown in A. (C) Spheroidal gold micro-nugget (Au I) with secondary, well crystallised gold overgrowth (Au II). (D) Same as in C, but secondary gold filling a fracture within a detrital zircon grain that pierces a rounded gold grain. (E) In-situ gold micro-nugget (Au I) with overfolded rims next to rounded pyrite; all of the above from the same hand specimen of Basal Reef, Welkom gold field, shown in Fig. 5B. (F) Sericite pseudomorphs after K-feldspar (FspPsm) as inclusions within rounded pyrite from the Kimberley Reef, South Roodepoort mine, West Rand gold field; matrix silicates are chloritoid (Ctd) and muscovite (Mus). (G) Hydrothermal gold with chlorite in matrix between lithic fragment and quartz pebble, Ventersdorp Contact Reef, Klerksdorp gold field. (H) Gold inclusions within secondary, hydrothermal pyrite (Py II), B-Reef, Free State Geduld mine, Welkom gold field. (I) Euhedral, secondary pyrite overgrowths around older, rounded cores (Py I), Ventersdorp Contact Reef, Klerksdorp gold field.

20

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

explaining the good correlation between bitumen seams and U content (Minter, 1978). Considering the relationship between bitumen formation and locally available radioactivity, at least some of the uraninite must be older than the hydrocarbons, and hence detrital or early diagenetic. Secondary, postdepositional uraninite and other U-bearing minerals, mainly brannerite and uraniferous leucoxene, are considered to be products of partial mobilisation of the earlier, rounded uraninite (England et al., 2001a). A detrital origin of the rounded uraninite grains is suggested by their mineral chemistry and UPb geochronology. Even on a thin section scale, adjacent grains show a great variation in Th/U ratio, which reflects provenance from a variety of source rocks. Many uraninite grains are rich in Th (average 3.9 wt.%), which is inconsistent with a low-temperature hydrothermal origin but indicative of granitic to pegmatitic sources (Feather and Glathaar, 1987; Grandstaff, 1981). Direct dating of uraninite grains from the Dominion Reef yielded an UPb age of 3050F50 Ma (Rundle and Snelling, 1977). This age overlaps with that of Dominion Group sedimentation but is distinctly older than the Witwatersrand sediments. Another important U-mineral in the Witwatersrand assemblages is brannerite with a composition close to UTi2O6. It is characteristically of secondary origin, derived from the oxidation of uraninite in the presence of rutile, which, in turn, is an alteration product derived from original detrital ilmenite and minor titanomagnetite particles. Form relics of rounded ilmenite occur concentrated on all scour surfaces throughout the Witwatersrand Supergroup. Most of it is, however, altered to leucoxene that is in many cases highly uraniferous. Typical TiO2 concentrations in the siliciclastic metasedimentary rocks are between 0.1 and 1 wt.%. The very fine-grained nature of the rounded leucoxene pseudomorphs after ilmenite points towards them being the result from weathering. This is in contrast to the presence of distinct, euhedral to subhedral rutile and brannerite grains, both of which are related to the hydrothermal oxidation of original ilmenite and uraninite. Direct dating of such rutile from the West Rand Group (Robb et al., 1990) confirms such an age relationship as it yielded a UPb age of 2578F34 Ma, which is markedly younger than the age of sediment deposition.

3.2.3. Gold It has been noted by many workers that the Witwatersrand gold appears late in the paragenetic sequence (Ramdohr, 1958), with most of it occurring in textural association with bitumen, hydrothermal/ metamorphic chlorite (Gartz and Frimmel, 1999) or pyrophyllite (Barnicoat et al., 1997), along microfractures (Jolley et al., 2004), and as inclusion within euhedral, secondary pyrite (Fig. 6H). In contrast to the latter, rounded pyrite is typically devoid of gold inclusions. This generation of gold displays either euhedral crystals or dendritic or otherwise irregularly shaped habit. In a few instances, however, gold particles have a completely different morphology. In contrast to the above, they display rounded, spheroidal, disc-like and toroidal forms. Of particular importance is that both morphological types, the rounded to torroidal and the dendritic to euhedral, secondary gold, can occur together on a micro-scale (Fig. 6A,C,D), i.e. within millimetres in the same thin section (Minter et al., 1993). This provides strong evidence for a polyphase gold entrapment history, with the rounded particles derived by mechanical (fluvial) transport and secondary gold by precipitation from a hydrothermal fluid. Considerable compositional variability with respect of Au:Ag:Hg ratios in gold particles exists between reefs, within a given reef, and in some cases even on a micro-scale within a given thin section (Frimmel and Gartz, 1997; Reid et al., 1988). Individual gold particles are, however, homogeneous, which is readily explained by the diffusion rates of Ag and Hg through gold at the temperatures to which the Witwatersrand rocks were subjected during burial and metamorphism (Frimmel et al., 1993). Only gold particles in quartz veins that have been ascribed to the Vredefort impact, based on field relationships with impact-related pseudotachylite, have internal compositional variability. This is readily explained by the short duration of that event that did not permit diffusional homogenisation (Frimmel and Gartz, 1997). First attempts to date the gold directly by the Re Os method yielded ages that are older than that of sediment deposition (Kirk et al., 2002). Four gold samples from the Vaal Reef define an isochron that corresponds to an age of 3016F110 Ma and, when combined with rounded pyrite from the same hand-

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

21

sample, define an age of 3033F21 Ma. Exceedingly high Os concentrations, of as much as 4.16 ppm, reported for gold from the Vaal Reef (Kirk et al., 2001), are possibly an artifact of contamination by minute inclusions of platinum-group element minerals. However, a series of subsequent analyses of gold from the Vaal Reef (Kirk et al., 2002) and from the Basal Reef, both of toroidal micro-nuggets and secondary, hydrothermal gold crystals (Frimmel et al., in press) show consistent Re concentrations between 4 and 37 ppb and Os concentrations between 2 and 15 ppb. These values are one to four orders of magnitude greater than those for younger gold deposits as well as for average continental crust (Kirk et al., 2002). 3.3. Sediment provenance Derivation of the pebbles from mainly Archaean granite and pegmatite (55%) as well as mesothermal quartz veins and marine chert (45%) is indicated by oxygen isotope data (Vennemann et al., 1992, 1995). A granitic and/or pegmatitic source is further indicated by rare grains of detrital cassiterite, molybdenite and columbite (Feather and Koen, 1975) and by bright cathodoluminescence of many detrital quartz grains (Gartz, 1996). Furthermore, the concentrations of granitophile elements, such as Zr, Ta, Th, and rare earth elements, show a very good correlation (rN0.9) with each other in the conglomerates. By comparison with Archaean greenstone terrains in the Kaapvaal Craton, which typically contain, apart from the dominant mafic to ultramafic volcanic rocks, also highly siliceous chemical sedimentary rocks and felsic volcanic rocks, an equivalent to such terrains is inferred as source region from the abundance of chert and locally quartz porphyry pebbles in the Witwatersrand conglomerates. A mafic to ultramafic component in the source area, as expected for an Archaean granitoidgreenstone terrane, is indicated by the abundance of detrital chromite and subordinate platinum group elements (PGE)-bearing minerals in the conglomerates, but also by elevated Cr, Co and Ni concentrations in all Witwatersrand shale units (Wronkiewicz and Condie, 1987). The ratios between the different PGE is surprisingly consistent throughout the Witwatersrand gold mines with (Os+Ir)/ (Os+Ir+Pt+Ru) around 0.7, but significantly different

from younger deposits, such as the Rustenberg Layered Suite, dunite and kimberlite, for which this ratio is around 0.1 (De Waal, 1982). This difference may reflect a high maturity of the placer (Cousins, 1973). However, the consistency in both the PGE mineralogy and the PGE ratios along strike and downslope, and the relative proximal position of the Witwatersrand placer deposits, may be an indication of a specific source area characteristic, i.e. that of a chondritic to subchondritic mantle, as recently proposed for PGE alloys in the Evander gold field (Malitch and Merkle, 2004). Furthermore, the prior existence of a relatively stable cratonic block is implied from the rare presence of diamond in some reefs (Feather and Koen, 1975; Ramdohr, 1958), probably related to the presence of kimberlite pipes in the source area. Detrital zircon age spectra (Kositcin and Krapez, 2004) indicate the following ages of significant felsic rocks in the source area: 33103300, 30903060, 29902980, 29502940, and 29202910 Ma. The zircon provenance age spectrum for the Central Rand Group is considerably more complex and spans a wider range (3450 to 2870 Ma) than that for the West Rand Group (3300 to 2960 Ma). This confirms the inferred tectonic setting of a passive margin for the West Rand Group, with sediment supply from fewer sources and no tectonic rejuvenation, and that of a foreland basin for the Central Rand Group, with increasingly more varied source rocks, continuous tectonic rejuvenation and erosion to older stratigraphic levels. Corresponding counterparts for all of the observed detrital zircon age modes are known from the surroundings of the Witwatersrand Basin (Frimmel et al., in press). Detrital xenotime ages ranging from 2840 to 2813 Ma are not represented among the detrital zircon ages, most likely because they were derived from high-U granitoids, whose zircon grains would have been metamict (Kositcin and Krapez, 2004). High-U granite and pegmatite bodies of comparable age are known from the southern Murchison Belt (Poujol, 2001; Poujol and Robb, 1999), near the Giyani Belt (Kroner et al., 2000) and the Barberton Belt (Meyer et al., 1994). Almost 3 billion years of erosion would make any of the currently exposed tectonic units improbable source areas for the Witwatersrand sediments. A nearly complete overlap of detrital zircon and

22

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

xenotime age spectra with ages from Palaeo- to Neoarchaean granitoidgreenstone terranes surrounding the Witwatersrand Basin is, however, evident and source areas that correspond to at least some of these terranes are therefore possible. Ages that correspond to the oldest detrital zircons from the Central Rand Group are reported only from the Barberton Belt. Equivalents to the detrital zircon age mode of 3310 3300 Ma (West and Central Rand Groups) are known from the Giyani and Barberton Belts. The minor detrital zircon age modes between 3210 and 3090 Ma reflect various granitoid bodies that form the basement to the Witwatesrand Basin. Most of the detrital zircon grains in both the West Rand and Central Rand Groups represent the age modes between 3060 and 3080 Ma, which correspond to the time of felsic volcanism in the Dominion rift. Comparable ages are also known from felsic volcanic rocks and granitoids in the Murchison Belt (Poujol and Robb, 1999; Poujol et al., 1996). All younger detrital zircon grains could have been sourced, based on sediment transport directions and age correlations, from higher crustal level equivalents of the Mesoarchaean AmaliaKraaipan, Murchison and Giyani granitoidgreenstone terranes.

4. Neoarchaean weathering One of the pillars, on which recent hydrothermal models rest, is the apparently large-scale, acidic hydrothermal alteration of the Witwatersrand Basin fill (Barnicoat et al., 1997). As weathering under an acidic Archaean atmosphere would lead to similar bulk rock chemical changes as acid leaching by postdepositional fluids, the question arises whether systematic chemical changes observed in the siliciclastic successions across unconformities reflect paleosol horizons or whether they are related entirely to post-depositional fluidrock interaction. Most of the Witwatersrand siliciclastic rocks did not achieve thermodynamic equilibrium during postdepositional alteration, including burial and regional metamorphism. This is indicated by the widespread survival of detrital clasts that are now embedded in a metamorphic matrix. Detrital quartz is distinguished from secondary quartz by displaying highly variable cathodoluminescence (Gartz and Frimmel, 1999),

variable degrees of strain and different fluid inclusion populations (Frimmel et al., 1993). Similarly, detrital white mica can be distinguished from metamorphic mica both on textural and compositional grounds (Frimmel et al., 1993; Sutton et al., 1990). The most common metamorphic silicates in these rocks are muscovite, pyrophyllite, chlorite, sudoite and chloritoid. In the coarser grained rocks these are randomly orientated, whereas in the argillitic units they define a slaty cleavage. Based on the silicate equilibrium assemblages, the peak metamorphic temperatures achieved throughout the Witwatersrand Basin range between 300 and 400 8C (Phillips and Law, 1994), except for the area around the Vredefort dome, where up to amphibolite facies conditions are recorded in the upturned, lowermost parts of the basin fill, i.e. the West Rand Group (Gibson and Wallmach, 1995). Away from the Vredefort dome, the presence of kyanite and the mineral assemblage chlorite+sudoite+muscovite (and/or pyrophyllite) in the middle Central Rand Group constrain peak metamorphic conditions at approximately 300 8C and 3 kbar (Frimmel, 1997). Sericite pseudomorphs after andalusite at the bottom of the Transvaal Supergroup and in the Ventersdorp Supergoup (McCarthy et al., 1986) reflect a lower pressure, i.e. an expected lower overburden for the higher stratigraphic levels. Metamorphic biotite is rare but has been described from the bottom of the Central Rand Group (Phillips et al., 1988) and also from the West Rand Group, together with stilpnomelane, relics of K-feldspar and chlorite (Frimmel, 1994). Feldspar is extremely rare in the Central Rand Group (Fig 6F), but few units, particularly within the West Rand Group contain as much as 30 vol.% detrital K-feldspar that can be related to a granitic source due to its perthitic textures (Fuller, 1958). Albite is very rare in the metasedimentary rocks but has been reported as metamorphic phase from an argillite in the uppermost Johannesburg Subgroup (Booysens Formation; Frimmel, 1994). Pyrophyllite is widespread and becomes particularly important towards the basin margin. Crosscutting relationships between pyrophyllite distribution and stratigraphic boundaries have led Barnicoat et al. (1997) to postulate basin-wide acidic hydrothermal alteration. These workers reported intense pyrophyllitic alteration within mineralised conglomerate horizons, as well as for more than 500 m above and below

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

23

surrounding quartzite. A postulated correlation of gold distribution with this alteration (Phillips and Law, 1994) that postdates the hydrothermal/metamorphic phyllosilicates forms the fundamental argument for a hydrothermal origin of the gold. Considering the multi-stage tectono-metamorphic history of the Witwatersrand Basin, it appears unlikely that all the post-depositional alteration phenomena in the basin are ascribable to a single hydrothermal infiltration event. Although there is little doubt that H+-metasomatism has caused the formation of some of the pyrophyllite at the expense of mica, as evidenced by cross-cutting pyrophyllite veinlets, it might be difficult, in other places, to distinguish such hydrothermal acid leaching from the effects of weathering under acidic conditions. The latter can be expected to occur along erosional unconformities, where kaolinite would have been the starting material for pyrophyllite formation during prograde metamorphism. The findings of Barnicoat et al. (1997) are at variance with those of Sutton et al. (1990) who found, by studying the compositional and mineralogical changes across stratigraphic units, a strong stratigraphic control on the chemistry and mineralogy of Witwatersrand arenites and concluded that, except for potassium, metamorphism was essentially isochemical. 4.1. Geochemical alteration profiles across stratigraphic units The shape of geochemical alteration profiles across individual unconformities can be used to test whether observed alteration patterns are related to palaeoweathering or to post-depositional metasomatism. In the former case, a systematic change in bulk rock chemistry is expected towards the top of the footwall, but not in the hanging wall. In the latter case, both footwall and hanging wall should show an alteration halo around the unconformity, which is presumed to be the principal post-depositional fluid pathway, with dispersion causing alteration to comparable extent both above and below that pathway. To this effect, the chemical index of alteration (Nesbitt and Young, 1982), CIA Al2 O3 =Al2 O3 CaO4 Na2 O K2 O 100

in which the oxides are expressed as molar proportions and CaO* is CaO in silicates, as opposed to carbonates and phosphates, was applied to analyses of siliciclastic rocks across a number of reefs throughout the Central Rand Group and the upper West Rand Group (Figs. 4 and 7). The concentration of CaO in most samples studied is less than 0.1 wt.% but typically above the lower limit of detection (0.004%). Small amounts of Na (Na2O contents are typically around 0.2 wt.%) are bound largely to white mica as paragonite intergrowths (Frimmel, 1994), except for albite-bearing shale in the West Rand Group. Consequently, the CIA reflects essentially the distribution of K-feldspar, muscovite and pyrophyllite, i.e. the extent to which K was leached out of the rock during lateritic weathering and/or post-depositional reaction with an acidic fluid. Comparison of average CIA values across the stratigraphic units (Fig. 4) shows significantly lower indices for the few analyses available from the West Rand Group. Analyses for most of the West Rand Group are lacking, but data from a likely stratigraphic equivalent, the Mozaan Group, a ca. 5000 m thick siliciclastic succession within the Pongola Supergroup (Beukes and Cairncross, 1991), may serve as proxies. There, the matrix of diamictite units and associated mudstones have significantly lower CIA values (on average 66) than other mudstones in the Pongola Supergroup, and these have been interpreted as indicative of a glacial origin for the diamictite beds (Young et al., 1998). Most of the available data for the West Rand Group (Sutton et al., 1990) cluster between 50 and 60, whereby 50 is typical of unweathered material, representing the composition of fresh feldspar. It should be noted that these data were obtained on arenitic samples and thus reflect sediment that has been transported over shorter distance. They are therefore not directly comparable with those obtained on mudstone for which the calculated CIA values across the Witwatersrand Supergroup are variable (7098; Wronkiewicz and Condie, 1987). Nevertheless, a marked difference in the chemical weathering intensity between West Rand and most of the Central Rand Group sediments seems to be indicated, similarly as noted for the Pongola Supergroup. This is well shown by a systematic increase in the CIA to as much as 85 towards the sequence boundary with the Central Rand Group.

24 H.E. Frimmel / Earth-Science Reviews 70 (2005) 146 Fig. 7. Variation in Chemical Index of Alteration (CIA) and Fe/Al ratio with distance from the Dennys, Crystalkop (Frimmel and Minter, 2002; note that the length scale in the original source paper is incorrect) and Ventersdorp Contact Reefs (Gartz and Frimmel, 1999) at various mines in the Klerksdorp gold field (AG) and the Welkom gold field (H.E. Frimmel, unpubl. data; H); data are for argillitic to arenitic siliciclastic rocks, except for the hanging wall of the Ventersdorp Contact Reef, which consists of metabasalt and thus has a lower CIA.

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

25

For most of the Central Rand Goup, the CIA is around 80, which reflects above average (Nesbitt and Young, 1996) extent of chemical weathering. Only in the uppermost formation do the CIA values decrease to around 70, which is in agreement with average chemical weathering. The internal variability in CIA within the group, particularly across unconformities, is, however substantial (Fig. 4). A closer inspection of this variability, using examples from the footwall and hanging wall of the Crystalkop and Dennys Reefs at Vaal Reefs mine, Klerksdorp gold field (Frimmel and Minter, 2002) and the B-Reef at Freestate No. 3 mine, Welkom gold field (H.E. Frimmel, unpubl. data) reveals a gradual upward increase in CIA in the footwall towards the reef contacts (Fig. 7). This trend is exemplified by two profiles through the Crystalkop Reef (Fig. 7A,B). In some areas, this trend can be traced over more than 10 m (Fig. 7A,B), whereas in others, a systematic increase in CIA towards the reef was found over a distance of only a few metres below the reef (Fig. 7C,D,E,F). In the hanging wall, this trend is reversed with an abrupt decrease in the CIA away from the reefs (Fig. 7A,B,C,D,E,H). Note that the very low CIA in the hanging wall of the Ventersdorp Contact Reef (Fig. 7G) is an artifact of the lithology as it consists of metabasalt and not of arenitic metasedimentary rocks as in all other cases. Superimposed on the above relatively large-scale trends, with the highest CIA values reaching a maximum of 95 in the footwall near the respective reef contact, is a decrease in CIA on a smaller scale along the reef horizons. This is best exemplified by two profiles through the Crystalkop Reef (Fig. 7B,D), which show a marked drop in the CIA within a few metres both above and below the reef. Note that this trend affects both the footwall and the hanging wall, whereas the former, larger-scale trend appears asymmetrical below and above the reef horizons. Similarly, corresponding chemical changes, both above and below the reef, have also been observed at the Ventersdorp Contact Reef (Fig. 7G). There an earlier potassic alteration (Gartz and Frimmel, 1999) affected the pyrophyllite-bearing footwall arenite as well as the metabasaltic hanging wall, causing a decrease in CIA over the top few metres in the footwall and an increase in CIA over the bottom few metres in the hangingwall. This was followed by chloritisation only along the reef and its immediate contact zones over a few

centi- to decimetres, which is reflected by a sharp but very local increase in CIA. 4.2. Interpretation of trends in CIA Trends in CIA are observed on different scales, i.e. on the hundred metres, few metres and decimetrescale. An example for the former is the systematic increase in CIA towards the West Rand/Central Rand Group unconformity over more than 100 m. This is too long to be explained by palaeoweathering along that unconformity, but may point to a systematic change in the environmental conditions (increase in temperature and/or acidity) towards that major stratigraphic boundary. The upwardly increasing CIA trends on the scale of a few metres are typical of, but more extreme than, those of both modern and Eoproterozoic paleosols (Nesbitt and Markovics, 1997). With one exception (Fig. 7F), the very high CIA values are confined to the footwall, which is not the expected result if they were related to reef-parallel acidic hydrothermal infiltration. The very high CIA values are, therefore, ascribed to intense chemical weathering. Whereas Ca and Na are removed during the initial stages of development of a weathering profile, K is removed only during the latest stages (Nesbitt and Markovics, 1997; Nesbitt and Young, 1984). Consequently, in Al2O3(CaO+Na2O)K2O (ACNK) space (Fig. 8), a typical weathering trend will follow a line parallel to the ACN join until it reaches the AK join from where it will continue towards the A apex. Plotting a large data set of arenitic bulk rock analyses, predominantly from the Central Rand Group with a few data from the West Rand Group, in that space (Fig. 8) reveals that the bulk of the non-mineralised arenitic rocks from various stratrigraphic levels between individual reef (unconformity) horizons follows a weathering trend of progressive removal of CaO and Na2O prior to removal of K2O, with a starting point close to the average composition of Neoarchaean upper continental crust. A considerable number of samples are displaced towards the K apex, thus indicating variable and, in places, considerable K-metasomatism. Comparison between footwall and hanging wall analyses (from a distance of up to 1 m below or above a given reef), shows that especially hanging wall samples show evidence of this K-

26

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

Fig. 8. ACNK (Al2O3[CaO+Na2O]K2O) diagram showing alteration trends in siliciclastic metasedimentary rocks from auriferous reefs (conglomerate), immediate footwall and hangingwall of reefs, and from arenite of various stratigraphic positions in the Witwatersrand Supergroup (n=226); data from Frimmel and Minter (2002), Gartz and Frimmel (1999), Sutton et al. (1990), Q. Hawes (unpubl. data), and H.E. Frimmel (unpubl. data); small triangular diagram in top left corner shows general alteration trends: 1chemical weathering of average Neoarchaean upper continental crust ( ) from Condie (1993), 2Ca/Na-metasomatism, 3K-metasomatism.

enrichment, which consequently has to be inferred as post-depositional. Such metasomatism would explain the local negative excursion in CIA along the Crystalkop Reef as shown in Fig. 7D. In contrast, most footwall samples plot very close to the A apex (very high CIA values)a feature that is absent in the hanging wall samples and therefore regarded as reflecting palaeo-weathering. Analyses from within reef beds display a wide spread close to the ACN line (Fig. 8). Such a trend cannot be explained by any weathering process but clearly illustrates the effects of post-depositional alteration. As most of the plotted analyses come from the Ventersdorp Contact Reef, the enrichment in Ca, and to a lesser extent Na, reflected by the trend towards the CN apex, can be explained by interaction of a hydrothermal fluid with the overlying Ca- and Na-rich metabasalt of the Klipriviersberg Group. Alteration patterns around this reef (Fig. 7G) also illustrate well the effects of K-metasomatism as described in more detail by Gartz and Frimmel (1999). Using Ti as the least mobile reference element, a sharp increase in K is observed with increasing proximity to the reef, which reflects

sericitisation. Only in the immediate reef environment, within less than one metre distance, the rocks are depleted in K and enriched in Fe, reflecting chloritisation. As the potassic alteration over several metres and the smaller-scale ferric alteration over a few decimetres affected both the footwall and the hanging wall, they cannot be related to weathering on a palaeo-surface but must be explained by reefparallel post-depositional hydrothermal fluid flow as postulated on structural grounds by Jolley et al. (1999). This type of alteration might also be responsible for the halo of elevated CIA values observed around the Dennys Reef (Fig. 7F) and a sericitisation similar to that in the VCR might explain the local sharp drop in the CIA values around the Crystalkop Reef as shown in Fig. 7D. In summary, large-scale trends in CIA may be the result of an overall change in climate and/or reflect different degrees of sediment re-working. Systematic increases in CIA to very high values in the footwall beneath erosional unconformities on the scale of several metres is ascribed to deep chemical weathering along these palaeo-surfaces and thus provide indirect information on the contemporaneous atmospheric composition. Small-scale variations in CIA in the immediate vicinity of reefs, typically over centi- to decimetres, reflect dispersive metasomatism caused by reef-parallel fluid flow.

5. Pre- or post-depositional age of the ore? A sedimentary model for the Witwatersrand gold deposits, originally proposed by Mellor (1916), was first challenged by Graton (1930), who suggested a magmatichydrothermal model. Since then, workers who have studied the ore and host rocks on a microscopic scale have emphasised a hydrothermal model, because gold is typically late in the paragenetic sequence as micro-fracture fills and inclusions in secondary, clearly hydrothermal mineral grains (Feather and Koen, 1975; Ramdohr, 1958). In contrast, those who have studied the rocks on a macroscale noted a strong sedimentological control on gold grade, which has been used highly successfully throughout the history of Witwatersrand exploration and day-to-day mining, and prompted them to advocate a sedimentary palaeoplacer model (Minter,

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

27

1978). Following the generally accepted recognition of regional metamorphism and post-depositional fluid flow throughout the basin (Phillips and Law, 1994), a palaeoplacer model sensu stricto, in which all gold particles are perceived as detrital, has been abandoned. Today there is general agreement that the majority of gold particles that appear late in the paragenetic sequence are indeed hydrothermal precipitates. Thus, the debate has shifted to the question of whether the source of that hydrothermal gold was proximal, fluvially deposited, detrital gold within the conglomerate beds (modified palaeoplacer model), or external to the host rocks (hydrothermal model). These two possibilities represent end-member models, which are, each with variations, the focus of current debate: (1) In the modified palaeoplacer model, transport of detrital gold particles into the host sediments is assumed to have taken place by fluvial processes with subsequent short-range mobilisation of the gold by infiltrating hydrothermal fluids and/or degradation of in situ hydrocarbon or hydrous phases. Gold mobilisation and recrystallisation induced by hydrothermal fluid infiltration has been ascribed to the emplacement of the Ventersdorp Supergroup lavas (Pretorius, 1991), to burial metamorphism in lower Transvaal Supergroup times and the Vredefort impact event (Frimmel et al., 1999), as well as to Pretoria Group deposition and the emplacement of the Bushveld Igneous Complex (Robb et al., 1997). (2) Hydrothermal models explain the presence of gold as the result of post-depositional hydrothermal fluids from an external source. The presence of gold in the conglomeratic host rocks is inferred as consequence of long-range, basin-wide fluid flow combined with chemical and structural controls (Phillips and Law, 2000). One hydrothermal model infers a separate origin for gold, uraninite and hydrocarbons (Phillips and Law, 2000), whereas another seeks to explain all of them as cogenetic (Barnicoat et al., 1997). The infiltration of the inferred auriferous hydrothermal fluids has been linked with several different events that range from Ventersdorp volcanism (Phillips et al., 1997), regional metamorphism (Phillips and Myers, 1989), or the emplacement of the Bushveld Igneous Complex (Stevens et al., 1997).

All of the currently debated genetic models include the presence of gold derived from the movement of hydrothermal fluids, but they differ principally in the inferred distances of hydrothermal gold transport and the composition of that goldtransporting fluid. The composition of the goldbearing fluid has been suggested to be similar to that inferred for Archaean orogenic gold deposits (Phillips and Law, 2000), in which gold is assumed to have been transported as bisulphide complex in an H2OCO2 dominated, relatively reducing, low-sulphur, low salinity fluid. In that model, all rounded pyrite is assumed to be the product of post-depositional hydrothermal sulphidation of black sands that existed originally of various FeTi oxides and Feoxyhydroxide pisolites (Phillips and Myers, 1989). This contrasts with models that suggest gold transport in highly acidic, oxidising fluids (Barnicoat et al., 1997), and models which prefer gold transport as a hydroxy-complex (Gray et al., 1998). Reduction of aqueous gold species to elemental gold by interaction with pre-existing bitumen plays an important role in all hydrothermal models and is used to explain the strong association of gold and bitumen and the textural position of gold grains within microfractures in bitumen. A major argument in favour of an external, hydrothermal source of gold is based on mineral and chemical zonation patterns at the deposit to handspecimen scale as well as elemental correlations (Fox, 2002). The recorded zonation patterns clearly demonstrate hydrothermal fluidrock interaction, but do not contribute to solving the question of the gold source. Hydrothermal versus detrital element correlations are more crucial in this regard. Excellent correlations between Au, U and Ag, a poor but positive correlation of Au with Cr and (Co+Ni), and only a very weak correlation of Au with Zr have been noted in several studies on the Kimberley Reef (Rasmussen and Fesq, 1973), the Vaal Reef (Fox, 2002), the Steyn Reef (Frimmel and Minter, 2002), and the B- as well as the Ventersdorp Contact Reef (H.E. Frimmel, unpubl. data). The good correlation between Au and Ag is expected, because gold is the principal sink for Ag, with Au and Ag being alloyed in variable proportions with each other and Hg (Frimmel and Gartz, 1997; Reid et al., 1988; Utter, 1979). Similarly, the positive

28

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

correlation with Co and Ni is readily explained by precipitation of cobaltitegersdorffite together with hydrothermal gold particles (Fox, 2002; Frimmel et al., 1993) and Co as well as Ni enrichment in coexisting pyrite. The good correlation between Au and U as well as Cr can be ascribed to a spatial association with heavy sands that contained detrital uraninite and chromite particles. As the formation of bitumen is genetically and spatially related to detrital uraninite, hydrothermal gold precipitation by a locally available reductant would have taken place preferentially near, or within, fractures of uraninite, thus further exacerbating the positive AuU correlation. The poor correlation between Au and Zr has been used as evidence against a detrital origin of the gold, because the distribution of Zr is controlled by zircon, almost all of which is detrital (Fox, 2002). Apart from gold and zircon being derived from different source rock types, the poor correlation could be a function of differences in the hydraulic behaviour during repeated sediment re-working (Smith and Minter, 1980). It is noteworthy that many Zr-rich samples are devoid of Au, simply indicating a gold-poor source area, whereas only very few samples show elevated Au but low Zr contents and these typically contain auriferous veinlets indicative of local gold mobilisation. If the gold had been introduced into the host sedimentary rocks by hydrothermal fluids, both zircon-rich and zircon-poor domains should have been affected to a similar degree. This is not the case. A less than perfect correlation between Au and other elements concentrated in detrital minerals, such as Zr, is likely the result of dispersion of the gold by shortrange hydrothermal mobilisation. Other arguments for a hydrothermal origin include the observation that the bulk of gold grains are located within or near micro-fractures, filled with bitumen and uraninite/brannerite, which post-date early, beddingparallel pyritepyrrhotitequartz filled fractures that contain no gold (Fox, 2002). While this observation demonstrates the undisputed hydrothermal nature of the bulk of the gold particles, it does not clarify the distance of hydrothermal gold transport. The same applies to effectively all other arguments that have been brought forward in favour of a hydrothermal model (Table 1). In contrast, a number of observations and analytical data can hardly be explained other than in terms of a modified palaeoplacer model.

Some of the most important lines of evidence for a modified placer model are the observation of gold micro-nuggets that are spatially associated with secondary, locally remobilised, hydrothermal gold. Although these spheroidal to torroidal micro-nuggets are very rare and have so far only been found in samples from the Basal Reef, Vaal Reef, B-Reef and Crystalkop Reef (Frimmel and Minter, 2002; Minter et al., 1993), their existence gives a clear clue as to the primary process of gold enrichment in the Witwatersrand sediments. Strong support for such a sedimentary gold enrichment process also comes from the ReOs isotope data, which indicate an age for the gold that is clearly older than that of host sediment deposition, but the same as for rounded pyrite (3033F21 Ma; Kirk et al., 2001, 2002). Notwithstanding local evidence of sulphidation of FeTi oxides (Ramdohr, 1958), chert and iron formation pebbles (Hallbauer, 1986; Hirdes and Saager, 1983), the ReOs data imply a detrital origin of the most abundant, rounded form of pyrite. All of the above examples of sulphidation can be related to the same fluids that caused the formation of secondary pyrite at various stages throughout the complex post-depositional alteration history of the Witwatersrand sedimentary rocks. The noted textural association of gold as inclusions within secondary pyrite and the lack of gold inclusions in rounded pyrite further indicate that hydrothermal gold transport cannot be linked with a postulated sulphidation event that supposedly formed all the pyrite. A similar argument can be applied to uraninite. The UPb age of 3050F50 Ma obtained for uraninite from the Dominion Reef (Rundle and Snelling, 1977) is older than the age of Witwatersrand sediment deposition and, though subject to a considerable uncertainty, effectively rules out a post-depositional introduction of significant amounts of U into the basin fill. In particular, the interpretation of the ReOs data appears to be very robust as the only alternative to the interpretation given above would be that of isotope mixing between two or more different ad-hoc end members. Any mixing model would be difficult to reconcile with the excellent agreement between the ReOs isochron ages of gold and rounded pyrite from various localities, the high precision of the isochron ages, and the initial Os isotopic compositions that are identical within error to the Os isotopic composition of the mantle at 3 Ga.

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146 Table 1 Main arguments for a hydrothermal and modified palaeoplacer model, respectively, for the Witwatersrand gold Hydrothermal model Gold is late in paragenetic sequence (Barnicoat et al., 1997; Feather and Koen, 1975) Gold is associated with acid metamorphic alteration (Phillips and Myers, 1989; Barnicoat et al., 1997) Basin-wide distribution of pyrophyllite related to large-scale H+-metasomatism (Barnicoat et al., 1997) Abundant rounded pyrite and uraninite particles associated with the gold ore are of post-depositional hydrothermal origin (Barnicoat et al., 1997) Rounded pyrite derived from basin-wide sulphidation of dblack sandsT (Phillips and Myers, 1989) Strong correlation between gold and hydrothermal pyrobitumen (Nagy, 1993; Gray et al., 1998) Modified palaeoplacer model

29

Conglomerate beds were preferred channels for infiltration of gold-bearing hydrothermal fluids (Barnicoat et al., 1997)

Rare co-existence of rounded gold micro-nuggets with secondary, hydrothermal gold on mm-scale (Minter et al., 1993) Composition of fluid inclusions in auriferous hydrothermal quartz indicate neutral to basic pH (Frimmel et al., 1999) Increase in chemical index of alteration in footwall towards reef related to chemical weathering under acid atmosphere (Sutton et al., 1990; Frimmel and Minter, 2002) Isotopic data for rounded sulphides and uraninite yield ages older than time of sedimentation (Rundle and Snelling, 1977; Kirk et al., 2001) Pyrite morphology, cyrstallography, and truncated growth zonation patterns indicate detrital nature of rounded grains (McLean and Fleet, 1989; Fleet, 1998; England et al., 2002b) Derivation of hydrothermal pyrobitumen from local intrinsic oils, based on bulk and molecular y13C data (Spangenberg and Frimmel, 2001) and fluid inclusion studies (England et al., 2001a) Local variability in gold composition within a given reef reflects differences in source areas (Frimmel and Gartz, 1997) Sedimentological control on gold distribution (Minter, 1978); Negative correlation between authigenic/hydrothermal xenotime and ore bodies (Kositcin et al., 2003) ReOs ages of the gold are older than time of sedimentation (Kirk et al., 2002) Lack of significant secondary permeability after N600 my of burial, diagenesis and low-grade metamorphism (Frimmel, 1997) Sedimentary reworking of Witwatersrand gold ore in late-Ventersdorp diamictite was followed by Witwatersrand-style gold mineralization in the post-Ventersdorp Black Reef Calculated background value for eroded source area corresponds to mean Au content of Archaean granitoidgreenstone crust (Loen, 1992)

Gold was introduced into the Witwatersrand Basin after sediment deposition (Phillips and Myers, 1989; Barnicoat et al., 1997) Hydrothermal introduction of gold into the basin during peak metamorphism, coeval with 2.06 Ga Bushveld event (Phillips and Law, 1994) Hydrothermal introduction of gold into the Witwatersrand basin during global 2.72.6 Ga gold-forming thermal event, coeval with Ventersdorp Supergroup volcanism (Phillips et al., 1997) Lack of suitable source area for placer gold (Phillips and Myers, 1989)

If the gold were brought into the Witwatersrand Basin during post-depositional fluid infiltration, large fluid/rock ratios would be expected. The auriferous fluids would have had to flow preferentially along the conglomerate beds in order to explain the apparent sedimentological control on the basin-wide ore distribution. Although some evidence exists for beddingparallel fluid flow (Jolley et al., 1999), mass balance calculations (Gartz and Frimmel, 1999) point to rather limited external fluid infiltration into the reefs. Only if all the pyrophyllite is explained by post-depositional H+-metasomatism (Barnicoat et al., 1997), can a case be made for large-scale fluid infiltration. Bearing in mind that the loss of alkalies on the scale of tens of metres can be attributed to palaeoweathering, as outline above, it is more likely that a large proportion

of pyrophyllite in the Witwatersrand metasedimentary rocks is derived from the prograde metamorphism of a kaolinite-bearing protolith, thus revoking the necessity for significant post-depositional fluid infiltration. However, even if all the pyrophyllite were hydrothermalmetasomatic, it has been shown from fluid inclusion analyses that in those cases studied, hydrothermally mobilised gold was transported by a fluid of a composition that is incompatible with the stability of pyrophyllite (Frimmel et al., 1999). Only limited interaction between potentially auriferous fluids and the host conglomerate beds is also supported by studies on the chemistry and age distribution of xenotime, which is a particularly useful monitor for hydrothermal infiltration (England et al., 2001b; Kositcin et al., 2003). These studies revealed that

30

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

diagenetic xenotime is relatively abundant in the reefs compared to their surroundings, but hydrothermal xenotime is present in much lesser amounts within the reefs. This implies that the reefs were more permeable during diagenesis but less permeable during postdiagenetic hydrothermal events, which is in disagreement with proposed hydrothermal models. Apart from the textural, geochemical and isotopic evidence, there is also important geological evidence that speaks against a hydrothermal model not only for the Witwatersrand gold, but also for the associated pyrite and uraninite. The Witwatersrand ore bodies must have formed prior to deposition of the Platberg Group sediments as the latter contain pebbles of mineralised Witwatersrand reef material (Phillips et al., 1997). Yet, Witwatersrand-style mineralisation is evident in the Black Reef at the base of the Transvaal Supergroup that is some 70 million years younger than the Platberg Group. Formation of a placer deposit under similar environmental conditions repeatedly through time is to be expected, but the same style of hydrothermal metal introduction into the basin at different stages of basin development in different tectonic settings is unlikely. If there had been a second major hydrothermal oreforming event in the central Kaapvaal Craton, including the conglomerates at the base of the Transvaal Supergroup, it would have affected the Witwatersrand rocks after they had undergone diagenesis and first low-grade metamorphism. Very little permeability would have remained at that stage in the Witwatersrand strata, and basin-wide, largescale, predominantly bedding-parallel fluid flow, as postulated by the various hydrothermal models, would have been effectively impossible. Problems with the hydrothermal models extend from the micro- to the macroscale. On a tectonic scale, the Witwatersrand deposits have often been compared with orogenic gold deposits by those who favour a hydrothermal model. This is mainly for the similarity in the paragenetic sequence and in the common goldpyritehydrocarbon association (Phillips and Myers, 1989). There are, however, a number of significant differences between the two styles of deposit (Frimmel et al., in press; Groves et al., 2003). The rounded, sub-spherical morphology of most of the Witwatersrand pyrite and its highly variably geochemical and S isotopic composition (England

et al., 2002b) are in contrast to the typically subhedral to euhedral, compositionally restricted pyrite found in orogenic deposits. Wallrock alteration, inferred to have taken place over several hundreds of metres across stratigraphic boundaries throughout the Witwatersrand Basin (Barnicoat et al., 1997), is orders of magnitude more extensive than known from any orogenic gold deposit. The latter are characterised by the abundance of auriferous quartz veins, but the basin that hosts the by far greatest known concentration of Au is characteristically devoid of a plentitude of such veins. A foreland/retroarc basin setting is indicated for the bulk of Witwatersrand deposits. This is analogous with modern placer gold deposits but in stark contrast to orogenic gold deposits, which typically occur in near-arc or arc settings. Finally, the geometry of the Witwatersrand orebodies (gently dipping decimetre to metre thick, laterally extensive sheets) is unlike the overall shape of most known epigenetic or orogenic deposit. 5.1. Best-fit genetic model A modified palaeoplacer models accounts best for all the available data and observations. Clastic sediments, first laid down in the 3074 Ma Dominion rift graben, were largely derived from felsic sources, which led to enrichment in detrital uraninite but only low gold contents. During subsequent West Rand Group sedimentation (29852914 Ma), progressively more mafic rocks from Mesoarchaen greenstone belts in the hinterland were eroded, but only during Central Rand Group times (29022780 Ma) were the high levels of gold concentrations in the placer sediments reached, for which the Witwatersrand has become so famous. These extraordinary gold concentrations are explained by a combination of factors that range from fertile source regions, to tectonic setting and to palaeoenvironmental conditions. Initially, the Central Rand Basin took a foreland position relative to the overriding Kimberley block in the west, with the intervening AmaliaKraaipan greenstone belt representing an obducted slice of former oceanic crust. A subsequent change in the continental stress field led to the accretion of the Murchison greenstone belt to the north, with the Central Rand Basin taking a retroarc position (Frimmel et al., in press). Gold and chromite, predominantly from the surrounding greenstone belts,

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

31

together with uraninite and zircon from the associated granitoids, were brought into this basin by fluvial transport and concentrated by mineral sorting. Aeolian transport and re-sedimentation on a number of erosional unconformities led to further up-grading of gold along the basal degradation surfaces. Sedimentary re-working in braided stream systems and effective wind sorting were particularly vigorous because of intense acid weathering and a lack of vegetation and widespread organisms. The complex post-Witwatersrand tectono-thermal evolution of the central Kaapvaal Craton affected the gold to a variable extent. The bulk of fluid flow through the Witwatersrand Basin must have taken place due to diagenetic dewatering. Different parts of the basin experienced further fluid flow to various degrees as consequence of the following events: (1) low-grade burial metamorphism; (2) the syn-Ventersdorp thermal anomaly, including syn-Platberg rifting; (3) dynamic metamorphism in a compressional stress field during thrusting of the Southern Marginal Zone of the Limpopo Belt on to the Kaapvaal Craton; (4) lower Transvaal thermal subsidence-induced extension; (5) the thermal anomaly related to the Bushveld magmatic event; and (6) pervasive fracturing due to the Vredefort impact. Some of these events caused the mobilisation of the gold, together with other detrital phases, such as pyrite and uraninite. Hydrocarbons, derived from oil/bitumen, played an important role in the re-precipitation of the mobilised gold by acting as reductants. The distances over which gold, pyrite and uraninite were mobilised, in general were in the order of millimetres to centimetres. Locally, fracture-controlled fluid flow allowed transport of Au over longer distances. Most of the hydrocarbons, and thus by implication gold, was first mobilised during diagenesis, but meteoric waters that percolated through Vredefort impact-related secondary interconnected (micro-) fracture space were also capable of transporting hydrocarbons, sulphides and gold.

6. Neoarchaean palaeoenvironment 6.1. Neoarchaean atmosphere It has long been recognised that attempts to constrain the Archaean atmospheric composition

hinge essentially on four lines of evidence: (1) presence of detrital uraninite; (2) presence of detrital pyrite; (3) composition of detrital gold particles; and (4) the presence and composition of paleosols (Holland, 1984). Since then the debate has shifted from a previously favoured palaeoplacer model to various hydrothermal models for the Witwatersrand gold, thus invalidating the reliability of the above pieces of evidence (Holland, 1994). The recognition that the best-fit model for the Witwatersrand deposits is that of a modified palaeoplacer, reaffirms however the usefulness of the above pieces of evidence for the debate on the Archaean atmospheric composition. The significance of the Witwatersrand in this regard is obvious, considering that all four lines of evidence can be tested there. However, other Mesoarchaean to Eoproterozoic deposits that bear strong similarities with those of the Witwatersrand outside the Kaapvaal Craton should not be ignored and can contribute useful additional information. These include the 2.13 to 2.10 Ga Tarkwaian System (Ghana), the 2.09 to 1.88 Ga Jacobina and the poorly dated 2.8 to 2.2 Ga Moeda deposits (Brazil), as well as the 1.90 Ga Roraima Supergroup (northern South America). Similar styles of mineralisation with abundant detrital pyrite and uraninite, but a conspicuous lack of gold, are known from the 2.9 to 2.6 Ga Bababudan Group (India) and the 2.45 Ga lower Elliot Lake Group (Huronian Supergoup) in Canada. All of these deposits have in common that a case for a palaeoplacer origin can be made (Frimmel et al., in press), and that the fluvial siliciclastic host sediments were laid down in foreland/retroarc basins. An Archaean to Palaeoproterozoic greenstone terrain is suggested as the most likely source area for all of the above gold palaeoplacer deposits. In addition, detrital pyrite, uraninite, and signficantly also siderite, have been reported from effectively unmetamorphosed fluvial siliciclastic sedimentary rocks from the Pilbara Craton (Rasmussen and Buick, 1999). The first piece of evidence to be assessed is the presence of detrital uraninite. Based on thermodynamic grounds U4+-minerals, such as uraninite, are not stable under modern atmospheric conditions, as they would oxidise rapidly (Fig. 9). The dependence of the uraninite stability on oxygen fugacity is almost independent of pH and the fugacities of other critical species, such as CO2 and CH4. Consequently, these

32

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

Fig. 9. Oxygen fugacity versus temperature diagram showing the conditions at which oxidation of pyrite to goethite (solid lines) takes place at variable pH, as well as that of uraninite to a dissolved oxyhyroxide (dashed line); calculated using PHREEQC (Parkhurst and Appelo, 1999).

reduced U-minerals are unlikely to survive mechanical transport in the fluvial environment under an oxidising atmosphere (e.g. Holland, 1994). Localised modern occurrences of detrital uraninite, such as in sand from the Indus river, are confined to domains that have experienced virtually no chemical weathering and can, therefore, not be used as analogues for the Witwatersrand occurrences (Maynard et al., 1991), for which intense chemical weathering is indicated. Detrital uraninite is abundant from the oldest siliciclastic rocks, the Dominion Reef, throughout the entire Witwatersrand Supergroup, to the base of the Ventersdorp Supergroup. Its occurrence thus spans a period from 3074 to 2714 Ma. The uraninite/ brannerite ratio varies greatly between reefs. Brannerite is not important in the Dominion Reef. A systematic decrease in the uraninite/brannerite ratio towards younger stratigraphic levels has been noted in the Welkom gold field, where this ratio changes from 8.7 in the Steyn Reef to zero in the Beatrix Reef (Minter et al., 1988). In the Klerksdorp gold field, the younger Ventersdorp Contact Reef does contain uraninite, but its textural relationships (typically as inclusions within bitumen) do not permit to distinguish between a detrital and a hydrothermal derivation. No uraninite is known from the 2642 Ma Black Reef, in which all U occurs as brannerite. This trend towards lower uraninite/brannerite ratios up-section might be interpreted as reflecting repeated sediment re-working under an atmosphere that became slightly more oxidising in the course of the Neoarchaean Aera. As

the brannerite is secondary, such an interpretation is, however, not imperative and this trend might equally reflect a more oxidising hydrothermal fluid composition at shallower crustal levels. The latter explanation is preferred, because of the abundance of detrital uraninite in the Elliot Lake Group, which is significantly younger than the Black Reef. The second piece of evidence concerns rounded pyrite, whose detrital nature has been established. The stability of pyrite requires even lower oxygen fugacity than that of uraninite (Fig. 9) thus supporting a reducing environment. Kinetic limitations to the solubility of pyrite are unlikely to explain the preservation of pyrite in the fluvial to fluvio-deltaic sedimentary rocks of the Witwatersrand. In particular, along those unconformities that reflect repeated reworking of the underlying sediment, pyrite must have been exposed to the meteoric environment over sufficient lengths of time to equilibrate with its immediate environment. The rounded pyrite type includes a variety of textural forms, ranging from compact rounded to ooid-like particles, all of which have formed prior to sediment deposition. However, as the mentioned petrographic and S isotope study (England et al., 2002b) revealed, some varieties represent pseudomorphs after other minerals, including sulphates, with replacement having taken place before erosion of the source rocks. It must be emphasised that rounded, detrital pyrite is not restricted to some small, localised occurrences, but is a characteristic feature of all fluvial deposits that range in time from the Dominion Group to the bottom of the Transvaal Supergroup (3074 2642 Ma), and in space over several hundred square kilometres. The common occurrence of detrital pyrite in fluvial sediments is not a peculiarity of the Kaapvaal Craton, but also typical of all other known Neoarchaean to early Palaeoproterozoic fluvial deposits, such as those of the Pilbara Craton in Australia, the Elliot Lake Group in Canada, the Bababudan Group in India and the Moeda deposits of Brazil, all of which are older than 2.2 Ga. The third piece of evidence, the composition of detrital gold particles is more problematic and less conclusive. Modern placer gold tends to be depleted in Ag, owing to an increased mobility of Ag relative to Au in oxidising waters (Morrison et al., 1991). Modern placer gold typically shows compositional

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

33

zonation with an increase in Au/Ag ratio towards the rim of an individual particle (Groen et al., 1990). Such a zonation pattern is absent in all of the thousands of studied gold particles from the Witwatersrand. This is readily explained by diffusional homogenisation with respect to Ag (and Hg) in gold particles at the temperatures to which the Witwatersrand gold was exposed (Frimmel et al., 1993). In spite of this intragrain homogenisation, inter-grain homogenisation was not achieved as evidenced by considerable differences in gold composition between reefs (Utter, 1979), between different domains within a given reef (Frimmel and Gartz, 1997) and even between individual grains on a hand-specimen scale (Reid et al., 1988). These differences are most likely a function of provenance and reflect gold sources of variable composition. Mobilisation of detrital gold particles during postdepositional hydrothermal alteration may have led to some modification of the composition, because of different mobility of Au- and Ag-bearing dissolved species. Consequently, individual gold particles that formed as hydrothermal precipitates should not be used as reference for deciphering contemporaneous atmospheric oxidation potential. However, the compositions of preserved detrital gold particles may be more illuminating in this regard. The only example of well studied detrital gold particles from the Witwatersrand concerns the Basal Reef in the Welkom gold field, for which average Ag and Hg concentrations of 8.9 and 1.2 wt.%, respectively, have been obtained (Frimmel et al., 1993). Such a composition is in very good agreement with Archaean greenstone-hosted gold (for compilation see Morrison et al., 1991). The elevated Ag contents could thus be used as argument against an oxidising environment during fluvial transport, but it may be argued that Agdepleted rims that had developed were subsequently mechanically eroded during fluvial transport. Furthermore, Au-rich rim formation in placer gold is probably related to a combination of self-electrorefining and cementation instead of selective leaching of Ag and intra-grain diffusion (Groen et al., 1990), in which case little inferences can be made regarding atmospheric conditions. Last but not least, paleosols can provide some of the most reliable information on the composition of the atmosphere at the time of surface exposure. In an

extensive review of paleosols described from the time period of interest here (Rye and Holland, 1998), it was concluded that all paleosols older than 2.2 Ga are characterised by significant Fe-loss. This includes examples from the Witwatersrand, but they are all problematic because of their complex post-depositional alteration history. Metamorphic chloritisation or formation of pyrophyllite at the expense of mica would cause enrichment or depletion in Fe, respectively. In many cases, it is not clear, whether an observed loss in Fe is related to palaeo-weathering or hydrothermal alteration. At least some of the inferred paleosols from the Witwatersrand, previously used to make inferences regarding Archaean atmospheric conditions, appear to reflect hydrothermally altered zones (Palmer, 1986), as documented, for example, for the so-called Deelkraal paleosol below the Ventersdorp Contact Reef (Jolley et al., 1999). From the geochemical changes across stratigraphic boundaries, outlined above, chemical weathering over several metres below unconformity surfaces can be deduced and distinguished from hydrothermal alteration that took place, in many cases, over only a very limited distance away from a given reef. Assuming that Al behaved conservatively, the extent of Feenrichment or depletion may be illustrated by the total Fe/Al ratio. A sympathetic relationship between Fe/Al ratio and CIA can reflect either hydrothermal chloritisation, as exemplified by the Ventersdorp Contact Reef (Fig. 7G) or a stratigraphic difference in sediment grain size and thus clay content of the protolith, as is the case in the footwall of the B-Reef (Fig. 7H). The former is typically found over very short distances (Fig. 7F,G), whereas the latter is found over longer distances across stratigraphic boundaries (Fig. 7H). In contrast, an antipathetic relationship may be explained by weathering under an acidic, reducing atmosphere. Indeed, some Witwatersrand reefs show a distinct depletion in Fe in their immediate footwall that is not linked to a corresponding decrease in CIA (Fig. 7B,C,D,F). Even in profiles in which no distinct trend is recognisable, the overall Fe/Al ratios tend to be very low, i.e. less than 0.2. This compares well with numerous examples of Meso- to Neoarchaean paleosols from other areas, for which Fe depletion has been described (Rye and Holland, 1998). The exact acidity of the contemporaneous atmosphere is difficult to constrain, but an upper limit on pH

34

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

can be placed from the lack of detrital feldspar in otherwise relatively immature siliciclastic sediment. The principal chemical weathering product was probably kaolinite, which subsequently gave rise to the abundant metamorphic pyrophyllite in the succession. In a number of samples, detrital white mica can still be recognised (Frimmel et al., 1993; Sutton et al., 1990). A pH close to the boundary between the kaolinite and muscovite (illite) stability fields, calculated as 6.2 at a temperature of 25 8C (for aK+=0.1), is therefore likely. Such acid weathering agrees well with the very high CIA values obtained for most footwall sections beneath unconformities, where they exceed 80 (for comparison, the CIA of modern tropical stream sediments is around 75, Maynard et al., 1991). Further evidence for a reducing sedimentary environment during upper Witwatersrand times comes from the bulk and molecular isotopic composition of the indigenous kerogen component in bitumen (Spangenberg and Frimmel, 2001). The C isotopic composition and distribution of n-alkane in stratiform bcarbon seamsQ within reefs point to considerable input from autochthonous algal-bacterial lipids. It may be argued that the spatial association between these hydrocarbons and all the other evidence provided above for a reducing Mesoarchaean atmosphere may render this evidence inconclusive. The argument could be that localised areas that were covered by terrestrial biomass, provided islands in which reducing conditions prevailed under an overall oxidising atmosphere. However, the distribution of detrital pyrite and uraninite, and even gold, is not at all restricted to domains rich in hydrocarbons. The bulk of siliciclastic fluvial to fluvio-deltaic sedimentary rocks of the Witwatersrand do not contain bcarbon seamsQ (they are restricted to localised lithofacies on unconformities) but contain abundant rounded pyrite as well as elevated uraninite concentrations. A reducing Meso- to Neoarchaean atmosphere, inferred here from multiple aspects of the nature of the Witwatesrand placer deposits, is also in agreement with totally different types of data from massindependent fractionation of S isotopes. This phenomenon, which refers to the deviation from the massdependent relationship between S isotopes typical of most processes in aqueous solution or solid phase (d 33Si0.515d 34S, d 36Si1.91d 34S), has so far been reported from Archean to Eoproterozoic sedimentary

sulphate and sulphide minerals, volcanic beds in ice cores and modern sulphate aerosols (Farquhar and Wing, 2003). A large scatter in d 33S, which does not obey mass-dependent relationships between S isotopes, is evident in samples older than 2.45 Ga, with a transition for the loss of this peculiar isotopic behaviour spnning from 2.45 to 2.09 Ga (Farquhar et al., 2000). The phenomenon, which has since been verified with data from a number of older cratons (Farquhar and Wing, 2003; Mojzsis et al., 2003), is explained by photochemical reactions, such as SO2 photolysis. As SO2 and SO photolysis are caused by intense ultraviolet radiation, the existence of such photolytic reactions in the Archaean atmosphere implies a lack of ozone and oxygen, which are the principal atmospheric components that absorb ultraviolet radiation. According to the photochemical model of Pavlov and Kasting (2002), the preservation of the observed mass-independent S isotope fractionation is only possible in an atmosphere with O2 concentrations less than 105 times the present atmospheric level. Furthermore, the preservation of the mass-independent S isotopic signatures is only possible in the absence of large-scale, homogenising, mass-dependent bacterial S processing in a marine, sulphate-rich reservoir (Farquhar et al., 2000). Thus, the mass-independent S isotope fractionation that characterises Archaean to Eoproterozoic sediments provides a very strong, independent argument for an anoxic atmosphere at those times. Reducing atmospheric conditions must have prevailed until at least 2.64 Ga as indicated by the pyriterich Black Reef at the bottom of the Transvaal Supergroup and probably lasted until at least 2.45 Ga, taking into consideration the uraninite-bearing conglomerates of the Elliot Lake Group. A minimum age for a reducing atmosphere is given by the 2.2 Ga Hekpoort paleosol, for which lateritic weathering with Fe-enrichment in the top zone has been documented (Beukes et al., 2002)in agreement with paleosols of similar age in other areas, for which weathering under a highly oxidising atmosphere is indicated (Holland and Rye, 1997). 6.2. Neoarchaean hydrosphere Having established that the Archaean atmosphere was most likely reducing, one might intuitively jump

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

35

to the conclusion that the contemporaneous hydrosphere was equally reduced. Such a conclusion seems not to be justified, however, because syngenetic barite deposits as old as Palaeoarchaean, with good examples known from the 3.53.2 Ga Barberton Supergroup in the Barberton greenstone belt (for a compilation of all known occurrences see Huston and Logan, 2004), clearly evidence that sulphur was present in its oxidised state in the form of sulphate as dominant hydrous S-species at least in some parts of the Archaean ocean. The main reductants supplied to the Archaean hydrosphere by hydrothermal discharge as well as metamorphic and weathering fluxes involve Fe2+, H2, CO, H2S, and SO2. In this context it is particularly interesting to scrutinise the distribution of Fe-oxides, principally magnetite, Fe-sulphides, essentially pyrite, and sulphates in the various sedimentary environments. The Witwatersrand rock record provides some pivotal information to this effect that can be summarised as follows: (1) Pyrite is stable in all fluvial to fluvio-deltaic deposits, even in those that experienced extensive sedimentary re-working and thus prolonged exposure to the meteoric environment; (2) various textural forms of evidently detrital pyrite show a complex S isotopic composition; and (3) magnetite is present instead of pyrite in most marine shale deposits (Fig. 4). The significance of the first criterion has already been discussed in the previous section and it suffices to state here that the above constraints on the Neoarchaean atmosphere are equally applicable to the meteoric hydrosphere and even the oceanic top waters. Sulphur isotopic composition has been used repeatedly to constrain Archaean O2 distribution (for review of available data see Strauss, 2003). Palaeoarchaean barite, at least some of which is supposedly sedimentary, has a fairly uniform S isotopic composition (d 34S=+2.7x to +8.7x) that matches the composition of Palaeoarchaean pyrite from the Barberton Supergroup (d 34S=3.1x to 8.8x), which, in turn, corresponds to that of magmatic sulphur. Larger variations in, and deviation from 0x, of d 34S is typical of younger sedimentary sulphides and is usually ascribed to enhanced S isotope fractionation between seawater sulphate and reduced sulphide, accomplished by biological S recycling. However, some authors pointed out relatively large variations in d 34S already in Archaean sediments (e.g. Ohmoto et

al., 1993; Shen et al., 2001), which prompted them to postulate microbial sulphate-reduction to have taken place as early as in Palaeoarchaean times. The S isotopic composition of the rounded, evidently detrital pyrite from the Witwatersrand corresponds to the limited range typical of Palaeoarchaean pyrite. This does not seem to support a model of extensive microbial sulphate reduction, but in the absence of a good control on the difference between the isotopic composition of the S source (marine sulphate) and the product (pyrite), no definitive conclusions on the role of microbial sulphate reduction in the Archaean hydrosphere can be drawn. The third of the above criteria highlights an apparent stratification with regard to deep ocean waters and near surface or freshwater environments during the Neoarchaean. Although the magnetite in its current textural position is clearly metamorphic, it is implausible to derive it from the oxidation of original sulphide, bearing in mind the comparatively overwhelming proportion of sulphide in the entire stratigraphic column. Any post-depositional fluid that percolated through the Witwatersrand Basin fill is more likely to have been enriched in S-species from the partial dissolution of the abundant pyrite in the stratigraphic succession than to have been capable of selectively oxidising a hypothetical primary sulphide in the intercalated shale beds. This is clearly evidenced by the common observation of secondary pyrite overgrowths at many stratigraphic levels. Furthermore, the shale beds were most likely those with the lowest permeability and thus least likely to have been affected by chemical change due to fluid circulation across stratigraphic boundaries. Derivation of the magnetite in these marine shale deposits from an oxide precursor is therefore assumed (Frimmel, 1996). As illustrated by Huston and Logan (2004), Fe solubility in the system FeSO is highest and lowest in the magnetite and pyrite stability fields, respectively (Fig. 10). Magnetite is only stable under reduced conditions (ASO4/AH2Sb102.5) and very low total sulphur concentrations, whereas pyrite can be stable even under oxidising conditions (ASO4/ AH2Sb108), provided the total sulphur concentration is high, i.e. close to that of modern seawater. In the presence of Ba, the pyrite stability field decreases to reducing conditions (ASO4bAH2S) as barite precipitates under relatively oxidising conditions (ASO4/

36

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

Fig. 10. Phase relationships and Fe solubilities in the system Fe BaSO as a function of redox potential (shown as ASO4/AH2S) and total sulphur concentration normalised to modern ocean water composition at a temperature of 25 8C and pH=7.8 (from Huston and Logan, 2004).

AH2SN102). In contrast to barite, Ca-sulphates are highly soluble and a considerable degree of evaporation is needed to precipitate gypsum. No occurrence of sedimentary sulphate is known to date from the Witwatersrand. However, based on S isotopic evidence, England et al. (2002b) concluded that some of the rare ooid-like pyrite grains found in the Ventersdorp Contact Reef represent the product of microbial sulphate reduction. This might reflect sulphate in the source area, such as barite that has been described from the Palaeoarchaean Barberton greenstone belt (see compilation by Huston and Logan, 2004). The apparent lack of sulphates in the Witwatersrand rock record may be explained in various ways. From a thermodynamic standpoint it should reflect a decrease in redox potential and/or total S concentration, or a lack of Ba. The latter possibility is not considered further as the principal Ba source is hydrothermal discharge, which is unlikely to have been shut down over several hundred million years during Witwatersrand sediment deposition. Whether the lack of sulphate in the Witwatersrand rocks in particular, and in Meso- to Neoarchaean rocks in general, reflects a combined increase in oceanic Fe and decrease in oceanic sulphate as well as total S concentrations relative to the Palaeoarchaean ocean, remains contentious and dependent on the interpretation of Palaeoarchaean sulphate deposits. An abrupt change in ocean water chemistry from a stratified ocean with sulphate-bearing top waters to a sulphate-

free, Fe-rich ocean has been suggested to have occurred around 3.2 Ga (Huston and Logan, 2004) and has been ascribed to heavy bombardment of Earths surface by meteorites (Glikson, 2001). Palaeoarchaean sulphate precipitation may well have been restricted to isolated oases of marine evaporative ponds (Pavlov and Kasting, 2002; Shen et al., 2001) and not representative of the world ocean. The lack of comparable deposits in the Witwatersrand succession could be merely a function of a lack of suitable environments for such oases in the Witwatersrand Basin at that time, and of preservation as evaporite deposits would be prone to erosion especially in a tectonically active foreland/retroarc setting. The presence of magnetite in many of the Witwatersrand shales mirrors a global surge in iron formation occurrences during the Meso- to Neoarchaean (Trendall, 2002), which reflects a reduced ocean, in which high concentrations of Fe2+ were possible in the bottom waters. The total S concentration in that ocean must have been less than 105 that of modern ocean water (Fig. 10). In contrast, the prevalence of pyrite in fluvial to shallow marine Witwatersrand deposits points to relatively higher total S concentration and/or higher O2 levels (Fig. 10) in the oceanic top waters and the meteoric realm. One of the magnetite-rich shale beds in the West Rand Group rests above diamictite deposits (Coronation Formation, Fig. 4). A causal link between iron formation and global ice age, as suggested for the younger Neoproterozoic iron formations, which are viewed as result of isolation of the oceans from the atmosphere by global, or near-global, ice cover (Kirschvink, 1992; Klein and Beukes, 1993), may therefore be applicable also to the magnetite-rich shale beds of the Witwatersrand. According to that model, melting of the ice cover would have triggered Feoxide precipitation following hydrothermal Fe-enrichment during glaciation. It should be noted, however, that a number of magnetite shale beds in the West Rand Group are not associated with diamictite. Furthermore, the above model is not universally accepted, not even for the Neoproterozoic deposits (for critical assessment of existing evidence see Young, 2004). Consequently, glaciation might have further facilitated the deposition of iron formation (magnetite shale) in the Meso- to Neoarchaean, but was probably not the primary cause of it.

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

37

6.3. Quantifying Neoarchaean oxygen levels Apart from the mentioned preservation of massindependent S isotope fractionation in pre-2.3 Ga sediments, the two main constraints on the O2 concentration in the palaeo-atmosphere between 3.0 and 2.6 Ga are based on the detrital mineralogy of placer deposits and on biochemical data on early microfossils. As pyrite requires even lower oxygen fugacity to be stable than uraninite, it is the preferred phase for setting an upper limit on ancient O2 levels. A lower limit is given by lipid biomarker data from sedimentary rocks of the Pilbara Craton in Western Australia. They provide evidence of oxygenic photosynthesis at least as early as 2.7 Ga (Brocks et al., 1999), and a lower limit of at least 1% of present atmospheric O2 is set by the presence of steranes, found in these rocks, because eukaryotic steroids require free oxygen (Jahnke and Klein, 1983). Although some workers have suggested the presence of aerobic bacteria already in Palaeoarchaean times, the evidence for that is problematic and a question of debate (Canfield and Raiswell, 1999).

In order to quantify the redox conditions that are required to explain the presence of pyrite in the given environments, the physico-chemical conditions for, and the rate of, oxidation of divalent to trivalent iron were calculated (Fig. 11). Considering that pyrite was also stable in Meso- to Neoarchaean shallow marine environments, a modern seawater composition (pH=8.22) was used for the initial calculations. Under these conditions, oxidation of Fe2+ to Fe3+ would be effectively instantaneous at f O2 of 103, very slow at f O2 of 107 and impossible at f O2 of less than 108 at a temperature of 25 8C, with two orders of magnitude lower f O2 required for comparable reaction rates at a temperature of 50 8C (Fig. 11A). This result is incompatible with the above biochemical limit. As indicated by the geochemical data, the atmosphere at the time of interest was likely acidic. A lower pH of the contemporaneous seawater would be more in line with the combined evidence from detrital pyrite and eukaryotic steroids. At a pH of 6.0, which would correspond to the inferred silicate weathering to kaolinite and the partial survival of detrital muscovite, oxidation of Fe2+ to Fe3+ would be very rapid

Fig. 11. Rate of oxidation of Fe2+ to Fe3+ in a solution of modern seawater composition as a function of time, in dependence of atmospheric oxygen fugacity ( f O2), pH and temperature (T): (A) pH of modern seawater, (B) pH=6; left panelT=25 8C, right panelT=50 8C; note that modern atmospheric O2 pressure corresponds to f O2=100.67; calculated using data generated from PHREEQC (Parkhurst and Appelo, 1999).

38

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

at f O2 of 100.67the modern atmospheric O2 leveland extremely slow at f O2 of 103the biochemical limit (Fig. 11B). Consequently, the biochemical f O2 limit of 1% present atmospheric level cannot have been exceeded significantly.

7. Conclusions The Witwatersrand gold fields in South Africa, which are the worlds largest gold producing province, hold important keys for understanding Archaean atmospheric and hydrospheric evolution. Crucial to the debate around atmospheric O2 levels at that time is the genesis of redox-sensitive minerals that are associated with the gold. Although most of the gold appears as a precipitate within, or associated with, post-depositional hydrothermal phases and along microfractures, available microtextural, mineralogical, geochemical and isotopic data, as well as the macro-scale stratiform distribution of the ore bodies and its strong sedimentological control, all indicate that this hydrothermal gold, analogous to the associated pyrite and uraninite, was derived from the local mobilisation of detrital particles. Some of the key pieces of evidence are a significant correlation between gold and other heavy minerals as well as sedimentary lithofacies, local preservation of in-situ micro-nuggets with well preserved delicate textures indicative of aeolian abrasion, compositional heterogeneity on a microscale, and radiometric age data that indicate an age of the gold, pyrite and uraninite (3.03 Ga) that is older than the maximum sedimentation age for the host sediment (2.90 Ga). None of these observations/data is compatible with a hydrothermal model, in which auriferous, potentially sulphidising fluids were introduced from an external source into the host rocks after sediment deposition. In contrast, those arguments, used in favour of hydrothermal models, emphasise the microtextural position of most of the gold, which highlights the undisputed hydrothermal nature of much of the gold in its present position, but does not explain the ultimate source of that gold. Similarly, microtextural features, S isotopic heterogeneity within and between grains, as well as direct dating by the ReOs method indicate that

the by far most abundant morphological variety of pyrite in the Witwatersrand deposits, i.e. rounded pyrite, is detrital. The same applies to rounded uraninite, which is responsible for the Witwatersrand, together with the siliciclastic deposits of the 2.45 Ga Elliot Lake Group, representing the worlds largest inferred U resource. Mineral chemical characteristics and particularly variability between grains, together with direct dating by the UPb method, provide independent evidence of its detrital nature. As with the gold, both pyrite and uraninite were mobilised during post-depositional fluidrock interaction to variable degree, whereby partial to complete oxidation to brannerite affected the detrital uraninite, especially at higher stratigraphic levels. The currently available data point to mechanical transport of gold, pyrite and uraninite from eroded source regions that bear similarities to granitoid greenstone terranes currently exposed to the north and west of the Witwatersrand Basin. The proportion between felsic and mafic/ultramafic source rocks varies strongly between and within reefs, and this explains the only poor correlation between Au and elements that are concentrated in detrital minerals. Sediment deposition took place initially in a continental rift (Dominion Group), followed by a passive margin (West Rand Group), and then in a foreland setting relative to the collision between the Witwatersrand and Kimberley crustal blocks, with a subsequent change to a retroarc position in consequence of oceanic basin closure to the north of the craton (Central Rand Group). Most of the gold accumulated in the foreland/retroarc setting as that setting favoured particularly intense sedimentreworking on a series of erosional unconformities. The recognition that rounded pyrite and uraninite, both of which are characteristic of 3.07 to 2.64 Ga fluvial to fluvio-deltaic siliciclastic sediments on the Kaapvaal Craton are not the products of postdepositional hydrothermal fluid infiltration, as suggested repeatedly in the past (e.g. Barnicoat et al., 1997; Phillips and Law, 2000; Phillips and Myers, 1989), but detrital components of numerous placer deposits re-affirms their pivotal importance for constraining the evolution of O2 concentrations in the Archaean atmosphere and hydrosphere. The lack of Fe-oxides/hydroxides and the survival of pyrite

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146

39

and uraninite during fluvial transport strongly support a reducing Meso- to Neoarchaean atmosphere. In contrast to the fluvial to near-shore coarser grained deposits, marine shale deposits contain magnetite instead of pyrite. Comparably reducing conditions are therefore inferred also for the oceanic bottom waters, but with orders of magnitude lower total sulphur concentrations than in modern ocean water. Thus the observed distribution of pyrite and magnetite in the Witwatersrand strata are in support of the hypothesis, largely based on sedimentological and S isotopic evidence, that Fe2+ was the principal oceanic redox buffer prior to 2.4 Ga, whereas after 1.8 Ga, following an intervening transition period, sulphate took over that role. During the Meso- to Neoarchaean Aeras, total S levels in the oceans must have been extremely low. Sulphate-reducing bacteria have been inferred from as early as Palaeoarchaean times (Shen et al., 2001). Any microbial sulphate reduction, together with hydrothermal inorganic sulphate reduction, would have removed sulphate rapidly to precipitate Fe-sulphides from Fe-rich ocean waters. Consequently, the oceans during Witwatersrand times would have been essentially free of sulphate. Furthermore, the lack of oxidative weathering of terrestrial pyrite would have prevented the supply of sulphate to the oceans at a rate that was greater than sulphate removal by Fe-sulphide precipitation. Most other evidence used for constraining Archaean atmospheric O2 concentrations, such as mass-dependent S isotope fractionation between Archaean sulphate and sulphide, chemical composition of gold grains, and geochemical characteristics of inferred paleosols, are not as conclusive as the abundant occurrence of detrital pyrite and uraninite in sediments that were laid down over extensive areas on several cratons. Independent support for a reducing Archaean atmosphere comes from mass-independent S isotope fractionation. An acid atmosphere, inferred from geochemical data, was probably in equilibrium with a correspondingly acid ocean. Kinetic calculations of the oxidation from Fe2+ to Fe3+ show that a pH of 6 is required in order to explain both the survival of detrital pyrite and the presence of eukaryotic steroids in Neoarchaean sediments. The calculated f O2 of 103 is in good agreement with the atmospheric evolu-

tion suggested previously by Kasting (1987, 2001; Fig. 1A: curve a). In summary, the available data endorse an acidic hydrosphere beneath a reducing atmosphere during the Archaean and early Palaeoproterozoic. Such an acid environment requires elevated concentrations of greenhouse gases, predominantly CO2. In addition, the postulated anoxic atmosphere would have favoured methanogenic bacteria that could contribute to elevated atmospheric CH4 concentrations, in agreement with the available C isotope record for the Archaean (Pavlov et al., 2001b). Thus the Archaean atmosphere was likely to be enriched in effective greenhouse gases that would have efficiently offset the lower solar luminosity in the early history of Earth as suggested by Walker et al. (1983). Such a palaeoclimate model explains the mineralogy of ancient placer deposits. Variably modified palaeoplacer deposits are known from the Mesoarchaean to the Palaeoproterozoic from a number of cratons, but only those older than about 2.4 Ga contain detrital pyrite and uraninite, whereas in the younger deposits, detrital sulphides and uraninite are conspicuously lacking and Fe-oxides occur instead.

Acknowledgements The author is indebted to Lawrie Minter who has never hesitated in sharing his enormous experience on the sedimentology and economic geology of the Witwatersrand and related deposits and who has given unrestricted access to his valuable sample collection, some of which has historic value as it contains material from mined out areas that are not accessible anymore. The work presented is based on numerous visits to underground mines and core yards, which would have been impossible without the cooperation of a number of mining companies. Of particular importance for the conclusions presented in this paper was the logistic support granted by Anglogold and its staff. J.B. Maynard is thanked for a constructive review of the manuscript. Parts of the work were funded through grants from the South African National Research Foundation and the University of Cape Town.

40

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146 Burke, K., Kidd, W.S.F., Kusky, T.M., 1986. Archean foreland basin tectonics in the Witwatersrand, South Africa. Tectonics 5, 439 456. Byerly, G.R., Lowe, D.R., Walsh, M.M., 1986. Stromatolites from the 3,3003,500 Myr Swaziland Supergroup, Barberton mountain land, South Africa. Nature 319, 489 491. Camisani-Calzolari, F.A.G.M., de Klerk, W.J., van der Merwe, P.J., 1984. Assessment of South African Uranium Resources: Methods and Results. Nuclear Development of South Africa (Pty), 39 pp. Canfield, D.E., Raiswell, R., 1999. The evolution of the sulfur cycle. American Journal of Science 299, 697 723. Catling, D.C., Zahnle, K.J., McKay, C.P., 2001. Biogenic methane, hydrogen escape, and the irreversible oxidation of early Earth. Science 293, 839. Coetzee, D.S., van Reenen, D.D., Roering, C., 1995. Quartz vein formation, metamorphism, and fluid inclusions associated with thrusting and bedding-parallel shear in Witwatersrand quartzites, South Africa. South African Journal of Geology 98, 371 381. Condie, K.C., 1993. Chemical composition and evolution of the upper continental crust: contrasting results from surface samples and shales. Chemical Geology 104, 1 37. Corner, B., Wilshire, W.A., 1989. Structure of the Witwatersrand Basin derived from interpretation of aeromagnetic and gravity data. In: Garland, G.D. (Ed.), Exploration 87: 3rd Decennial International Conference on Geophysical and Geochemical Exploration for Minerals and Groundwater. Geological Survey of Canada, Ontario, pp. 532 546. Cousins, C.A., 1973. Platinoids in the Witwatersrand system. Journal of the South African Institute of Mining and Metallurgy 73, 184 199. Coward, M.P., Spencer, R.M., Spencer, C.E., 1995. Development of the Witwatersrand Basin, South Africa. In: Coward, M.P., Ries, A.C. (Eds.), Early Precambrian Processes. Special Publication Geological Society London, 95, 243 269. Davidson, C.F., 1965. The mode and origin of banket orebodies. Institute of Mining and Metallurgy, London, Transcripts 74, 319 338. de Ronde, C.E.J., de Wit, M.J., 1994. The tectonothermal evolution of the Barberton Greenstone Belt, South Africa: 490 million years of crustal evolution. Tectonics 13, 983 1005. de Ronde, C.E.J., de Wit, M.J., Spooner, E.T.C., 1994. Early Archaean (3.2 Ga) Fe-oxide-rich, hydrothermal discharge vents in the Barberton greenstone belt, South Africa. Geological Society of America Bulletin 106, 86 104. De Waal, S.A., 1982. A literature survey of the metallurgical aspectes of minerals in Witwatersrand gold ores. Report No. M37, Mintek, Council for Mineral Technology, Johannesburg, 41 pp. de Wit, M.J., Tinker, J., 2004. Crustal structures across the central Kaapvaal Craton from deep-seismic reflection data. South African Journal of Geology 107, 185 206. de Wit, M.J., Hart, R.A., Hart, R.J., 1987. The Jamestown Ophiolite Complex, Barberton Mountain Land: a section through 3.5 Ga oceanic crust. Journal of African Earth Sciences 5, 681 730.

References
Anhaeusser, C.R., Walraven, F., 1999. Episodic granitoid emplacement in the western Kaapvaal Craton: evidence from the Archaean Kraaipan granitegreenstone terrane, South Africa. Journal of African Earth Sciences 28, 289 309. Armstrong, R.A., Compston, W., Retief, E.A., William, L.S., Welke, H.J., 1991. Zircon ion microprobe studies bearing on the age and evolution of the Witwatersrand triad. Precambrian Research 53, 243 266. Armstrong, R.A., Fanning, C.M., Eldridge, C.S., Frimmel, H.E., 1995. Geochronological and isotopic constraints on provenance and post-depositional alteration of the Witwatersrand supergroup. In: Barton, J.M.J., Copperthwaite, Y.E. (Eds.), Centennial Geocongress (1995). Geological Society of South Africa, Johannesburg, p. 1086. Barnicoat, A.C., et al., 1997. Hydrothermal gold mineralization in the Witwatersrand basin. Nature 386, 820 824. Barton, E.S., Hallbauer, D.K., 1996. Trace-element and UPb isotope compositions of pyrite types in the Proterozoic Black reef, Transvaal sequence, South Africa: implications on genesis and age. Chemical Geology 133, 173 199. Beukes, N.J., Cairncross, B., 1991. A lithostratigraphicsedimentological reference profile for the late Mozaan Group, Pongola Sequence: application to sequence stratigraphy and correlation with the Witwatersrand Supergroup. South African Journal of Geology 94, 44 69. Beukes, N.J., Dorland, H., Gutzmer, J., 2002. Tropical laterites, life on land, and the history of atmospheric oxygen in the Paleoproterozoic. Geology 30, 491 494. Bourret, W., 1975. Investigation of Witwatersrand uranium-bearing quartzpebble conglomerates in 194445, Genesis of uraniumand gold-bearing Precambrian quartzpebble conglomerates. Geological Survey Professional Paper 1161-A, A1 A6. Brandl, G., Jaeckel, P., Krfner, A., 1996. Single zircon age for the felsic Rubberdale Formation, Murchison greenstone belt, South Africa. South African Journal of Geology 99, 229 234. Brocks, J.J., Logan, G.A., Buick, R., Summons, R.E., 1999. Archean molecular fossils and the early rise of eukaryotes. Science 285, 1033 1036. Buck, S.G., 1983. The Saaiplaas Quartzite member: a braided system of gold- and uranium-bearing channel placers within the Proterozoic Witwatersrand Supergroup of South Africa. Special Publication of the International Association of Sedimentologists 6, 549 562. Buck, S.G., Minter, W.E.L., 1985. Placer formation by fluvial degradation of an alluvial fan sequence: the Proterozoic Carbon Leader placer, Witwatersrand Supergroup, South Africa. Journal of the Geological Society (London) 142, 757 764. Buick, I.S., Maas, R., Gibson, R., 2001. Precise UPb titanite age constraints on the emplacement of the Bushveld Complex, South Africa. Journal of the Geological Society (London) 158, 3 6. Burke, K., Kidd, W.S.F., Kusky, T.M., 1985. Is the Ventersdorp Rift system of Southern Africa related to a continental collision between the Kaapvaal and Zimbabwe cratons at 2.64 Ga? Tectonophysics 115, 1 24.

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146 de Wit, M.J., et al., 1992. Formation of an Archean continent. Nature 357, 553 562. Drennan, G.R., Boiron, M.-C., Cathelineau, M., Robb, L.J., 1999. Characteristics of post-depositional fluids in the Witwatersrand Basin. Mineralogy and Petrology 66, 83 109. Eldrige, C.S., Phillips, G.N., Myers, R.E., 1993. Sulfides in the Witwatersrand gold fields: new perspectives on old sediments via SHRIMP. Abstracts with Programs-Geological Society of America, vol. 25, A278. Els, B.G., 1998. The question of alluvial fans in the auriferous Archaean and Proterozoic successions of South Africa. South African Journal of Geology 101, 17 25. Els, B.G., van den Berg, W.A., Mayer, J.J., 1995. The Black Reef Quartzite Formation in the western Transvaal: sedimentological and economic aspects, and significance for basin evolution. Mineralium Deposita 30, 112 123. England, G.L., Rasmussen, B., Krapez, B., Groves, D.I., 2001a. The origin of uraninite, bitumen nodules and carbon seams in Witwatersrand golduraniumpyrite ore deposits, based on a Permo-Triassic analog. Economic Geology 96, 1907 1920. England, G.L., Rasmussen, B., McNaughton, N.J., 2001b. SHRIMP UPb ages of diagenetic and hydrothermal xenotime from the Archean Witwatersrand Supergroup of South Africa. Terra Nova 13, 360 367. England, G.L., Rasmussen, B., Krapez, B., Groves, D.I., 2002a. Archaean oil migration in the Witwatersrand Basin of South Africa. Journal of the Geological Society (London) 159, 189 201. England, G.L., Rasmussen, B., Krapez, B., Groves, D.I., 2002b. Palaeoenvironmental significance of rounded pyrite in siliciclastic sequences of the Late Archaean Witwatersrand Basin: oxygen-deficient atmosphere or hydrothermal evolution. Sedimentology 49, 1122 1156. Eriksson, K.A., Simpson, E.L., 2004. Precambrian tidalites: recognition and significance. In: Eriksson, P.G., Altermann, W., Nelson, D.R., Mueller, W.U., Catuneanu, O. (Eds.), The Precambrian Earth: Tempos and Events. Developments in Precambrian Geology, vol. 12. Elsevier, Amsterdam, pp. 631 642. Eriksson, K.A., Turner, B.R., Vos, R.G., 1981. Evidence of tidal processes from the lower part of the Witwatersrand Supergroup, South Africa. Sedimentary Geology 29, 309 325. Farquhar, J., Wing, B.A., 2003. Multiple sulfur isotopes and the evolution of the atmosphere. Earth and Planetary Science Letters 213, 1 13. Farquhar, J., Bao, H., Thiemans, M., 2000. Atmospheric influence of Earths earliest sulfur cycle. Science 289, 756 758. Feather, C.E., Glathaar, C.W., 1987. A review of uranium-bearing minerals in the Dominion and Witwatersrand placers. In: Pretorius, D.A. (Ed.), Uranium Deposits in Proterozoic Quartz pebble Conglomerates. International Atomic Energy Agency, Vienna, pp. 355 386. Feather, C., Koen, G.M., 1975. The mineralogy of the Witwatersrand reefs. Minerals Science and Engineering 7, 189 224. Fleet, M.E., 1998. Detrital pyrite in Witwatersrand gold reefs: X-ray diffraction evidence and implications for atmospheric evolution. Terra Nova 10, 302 306.

41

Fox, N., 2002. Exploration for Witwatersrand gold deposits, and analogs. In: Cooke, D.R., Pongratz, J. (Eds.), Giant Ore Deposits: Characteristics, Genesis and Exploration, CODES Special Publication vol. 4, University of Tasmania, pp. 243 269. Franca, A.B., Araujo, L.M., Maynard, J.B., Potter, P.E., 2003. Secondary porosity formed by deep meteoric leaching: Botucatu eolinite, southern South America. American Association of Petroleum Geologists Bulletin 87, 1073 1082. Frimmel, H.E., 1994. Metamorphism of Witwatersrand gold. Exploration and Mining Geology 3, 357 370. Frimmel, H.E., 1996. Witwatersrand iron formations and their significance to gold genesis and the composition limits of orthoamphibole. Mineralogy and Petrology 56, 273 295. Frimmel, H.E., 1997. Chlorite thermometry in the Witwatersrand basin: constraints on the Paleoproterozoic geotherm in the Kaapvaal Craton, South Africa. The Journal of Geology 105, 601 615. Frimmel, H.E., Gartz, V.H., 1997. Witwatersrand gold particle chemistry matches model of metamorphosed, hydrothermally altered placer deposits. Mineralium Deposita 32, 523 530. Frimmel, H.E., Minter, W.E.L., 2002. Recent developments concerning the geological history and genesis of the Witwatersrand gold deposits, South Africa. In: Goldfarb, R.J., Nielsen, R.L. (Eds.), Integrated Methods for Discovery: Global Exploration in the Twenty-First Century. Special Publication vol. 9. Society of Economic Geologists, Littleton, pp. 17 45. Frimmel, H.E., Le Roex, A.P., Knight, J., Minter, W.E.L., 1993. A case study of the postdepositional alteration of the Witwatersrand Basal reef gold placer. Economic Geology 88, 249 265. Frimmel, H.E., Hallbauer, D.K., Gartz, V.H., 1999. Gold mobilizing fluids in the Witwatersrand Basin: composition and possible sources. Mineralogy and Petrology 66, 55 81. Frimmel, H.E., et al., 2005. The formation and preservation of the Witwatersrand goldfields, the largest gold province in the world. Economic Geology 100 (in press). Fuller, A.O., 1958. A contribution to the petrology of the Witwatersrand Supergroup. Transactions of the Geological Society of South Africa 61, 19 45. Gartz, V.H., 1996. Post-depositional alteration of the Vetersdorp contact reef at Vaal Reefs no. 10 shaft, Klerksdorp Goldfield. Unpublished MSc Thesis, University of Cape Town, Cape Town, 167 pp. Gartz, V.H., Frimmel, H.E., 1999. Complex metasomatism of an Archean placer in the Witwatersrand Basin, South Africa: the Ventersdorp Contact Reef - A Hydrothermal Aquifer? Economic Geology 94, 689 706. Gibson, R.L., Wallmach, T., 1995. Low pressurehigh temperature metamorphism in the Vredefort Dome, South Africaanticlockwise pressuretemperature path followed by rapid decompression. Geological Journal 30, 121 135. Glikson, A.Y., 2001. The astronomical connection of terrestrial evolution: crustal effects of post-3.8 Ga mega-impact clusters and evidence for major 3.2F0.1 Ga bombardment of the Earth Moon system. Journal of Geodynamics 32, 205 229. Grandstaff, D.E., 1981. Microprobe analyses of uranium and thorium in uraninite from the Witwatersrand, South Africa, and Blind River, Ontario, Canada. In: Armstrong, F.C. (Ed.),

42

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146 Holland, H.D., 1994. Early Proterozoic atmospheric change. In: Bengtson, S. (Ed.), Early Life on Earth. Columbia University Press, New York, pp. 237 244. Holland, H.D., Rye, R., 1997. Evidence in pre-2.2 Ga paleosols for the early evolution of atmospheric oxygen and terrestrial biota: comment and reply. Geology 25, 857 858. Huston, D.L., Logan, G.A., 2004. Barite, BIFs and bugs: evidence for the evolution of the Earths early hydrosphere. Earth and Planetary Science Letters 220, 41 55. Jahnke, L.L., Klein, H.P., 1983. Oxygen requirement for formation and activity of the squalene epoxidase in Saccharomyces cerevisiae. Journal of Bacteriology 155, 488 492. James, C.S., Minter, W.E.L., 1999. Experimental flume study of the deposition of heavy minerals in simulated Witwatersrand sandstone unconformity. Economic Geology 94, 671 688. Jolley, S.J., Henderson, H.C., Barnicoat, A.C., Fox, N.P.C., 1999. Thrust-fracture network and hydrothermal gold mineralization: Witwatersrand Basin, South Africa. In: McCaffrey, K.J.W., Lonergan, L., Wilkinson, J.J. (Eds.), Fractures, Fluid Flow and Mineralization. Special Publications Geological Society London, 155, 153 165. Jolley, S.J., et al., 2004. Structural controls on Witwatersrand gold mineralisation. Journal of Structural Geology 26, 1067 1086. Kamo, S.L., Reimold, W.U., Krogh, T.E., Colliston, W.P., 1996. A 2.023 Ga age for the Vredefort impact event and a first report of shock metamorphosed zircons in pseudotachylitic breccias and granophyre. Earth and Planetary Science Letters 144, 369 387. Kasting, J.F., 1987. Theoretical constraints on oxygen and carbon dioxide concentrations in the Precambrian atmosphere. Precambrian Research 34, 205 229. Kasting, J.F., 2001. Earth history: the rise of atmospheric oxygen. Science 293, 819 820. Kasting, J.F., Siefert, J.L., 2002. Life and the evolution of Earths atmosphere. Science 296, 1066 1068. Kirk, J., Ruiz, J., Chesley, J., Titley, S., Walshe, J., 2001. A detrital model for the origin of gold and sulfides in the Witwatersrand basin based on ReOs isotopes. Geochimica et Cosmochimica Acta 65, 2149 2159. Kirk, J., Ruiz, J., Chesley, J., Walshe, J., England, G., 2002. A major Archean gold and crust-forming event in the Kaapvaal Craton, South Africa. Science 297, 1856 1858. Kirschvink, J.L., 1992. Late Proterozoic low-latitude global glaciation: the snowball earth. In: Schopf, J.W., Klein, C. (Eds.), The Proterozoic Biosphere: A Multidisciplinary Study. Cambridge University Press, Cambridge, pp. 51 52. Klein, C., Beukes, N.J., 1993. Time distribution, stratigraphy, and sedimentological setting and geochemistry of Precambrian iron formations. In: Schopf, J.W., Klein, C. (Eds.), The Proterozoic Biosphere: A Multidisciplinary Study. Cambridge University Press, Cambridge, pp. 139 146. Knoll, A.H., 1992. The early evolution of eukaryotes: a geological perspective. Science 256, 622 627. Kositcin, N., Krapez, B., 2004. SHRIMP UPb detrital zircon geochronology of the Late Archaean Witwatersrand Basin of South Africa: relation between zircon provenance age spectra and basin evolution. Precambrian Research 129, 141 168.

Genesis of Uranium- and Gold-Bearing Precambrian Quartz Pebble Conglomerates. U.S. Geol. Survey Professional Paper, pp. J1 J5. Graton, L.C., 1930. Hydrothermal origin of the Rand gold deposits: Part I. Testimony of the conglomerates. Economic Geology 25 (Suppl. 3), 1 185. Gray, G.J., Lawrence, S.R., Kenyon, K., Cornford, C., 1998. Nature and origin of carbon in the Archean Witwatersrand Basin, South Africa. Journal of the Geological Society (London) 155, 39 59. Grobler, D.F., Walraven, F., 1993. Geochronology of Gaborone granite complex extensions in the area north of Mafeking, South Africa. Chemical Geology 105, 319 337. Groen, J.C., Craig, J.R., Rimstidt, J.D., 1990. Gold-rich rim formation on electrum grains in placers. Canadian Mineralogist 28, 207 228. Groves, D.I., Rasmussen, B., Krapez, B., 2003. A comparison between the Witwatersrand and orogenic gold systems: a test of the hydrothermal Witwatersrand model. In: Eliopoulos, D.G., et al., (Eds.), Mineral Exploration and Sustainable Development. Proceedings of the 7th Biennial SGA Meeting, 2428 August 2003. Millpress, Rotterdam, pp. 727 730. Gutzmer, J., Nhleko, N., Beukes, N.J., Pickard, A., Barley, M.E., 1999. Geochemistry and ion microprobe (SHRIMP) age of a quartz porphyry sill in the Mozaan Group of the Pongola Supergroup: implications for the Pongola and Witwatersrand Supergroups. South African Journal of Geology 102, 139 146. Hall, R.C.B., 1997. The Ventersdorp contact reef: Final phase of the Witwatersrand basin, independent formation, or precursor to the Ventersdorp Supergroup? South African Journal of Geology 100, 213 222. Hallbauer, D.K., 1986. The mineralogy and geochemistry of Witwatersrand pyrite, gold, uranium, and carbonaceous matter. In: Anhaeusser, C.R., Maske, S. (Eds.), Mineral Deposits of Southern Africa. Geological Society of South Africa, Johannesburg, pp. 731 752. Hatton, C.J., 1995. Mantle plume origin for the Bushveld and Ventersdorp magmatic provinces. Journal of African Earth Sciences 21, 571 577. Hegner, E., Krfner, A., Hunt, P., 1994. A precise UPb zircon age for the Archaean Pongola Supergroup volcanics in Swaziland. Journal of African Earth Sciences 18, 339 341. Henkel, H., Reimold, W.U., 1998. Integrated geophysical modelling of a giant, complex impact structure: anatomy of the Vredefort structure, South Africa. Tectonophysics 287, 1 20. Henning, L.T., Els, B.G., Mayer, J.J., 1994. The Ventersdorp Contact Placer at Western Deep Levels gold minean ancient terraced fluvial system. South African Journal of Geology 97, 308 318. Hirdes, W., Saager, R., 1983. The Proterozoic Kimberley reef placer in the Evander gold field, Witwatersrand, South Africa. Monograph Series on Mineral Deposits vol. 20. Gebrqder Borntr7ger, Berlin, pp. 1 101. Holland, H.D., 1984. Chemical Evolution of the Atmosphere and Oceans. Princeton University Press, Princeton, 582 pp.

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146 Kositcin, N., et al., 2003. Textural and geochemical discrimination between xenotime of different origin in the Archaean Witwatersrand Basin, South Africa. Geochimica et Cosmochimica Acta 67, 709 731. Kreissig, K., et al., 2001. Geochronology of the Hout River Shear Zone and the metamorphism in the Southern Marginal Zone of the Limpopo Belt, Southern Africa. Precambrian Research 109, 145 173. Krfner, A., Tegtmeyer, A., 1994. Gneissgreenstone relationships in the Ancient Gneiss Complex of southwestern Swaziland, southern Africa, and implications for early crustal evolution. Precambrian Research 67, 109 139. Krfner, A., Jaeckel, P., Brandl, G., 2000. Single zircon ages for felsic to intermediate rocks from the Pietersburg and Giyani greenstone belts and bordering granitoid orthogneisses, northern Kaapvaal Craton, South Africa. Journal of African Earth Sciences 30, 773 793. Lambert, I.B., Beukes, N.J., Klein, C., Veizer, J., 1992. Proterozoic mineral deposits through time. In: Schopf, J.W., Klein, C. (Eds.), The Proterozoic Biosphere: A Multidisciplinary Study. Cambridge University Press, Cambridge, pp. 59 62. Lasaga, A.C., Ohmoto, H., 2002. The oxygen geochemical cycle: dynamics and stability. Geochimica et Cosmochimica Acta 66, 361 381. Law, J.D.M., Bailey, A.C., Cadle, A.B., Phillips, G.N., Stanistreet, I.G., 1990. Reconstructive approach to the classification of Witwatersrand dquartzitesT. South African Journal of Geology 93, 83 92. Loen, J.S., 1992. Mass balance constraints on gold placers: possible solutions to bsource area problemsQ. Economic Geology 87, 1624 1634. Malitch, K.N., Merkle, R.K.W., 2004. RuOsIrPt and PtFe alloys from the Evander goldfield, Witwatersrand Basin, South Africa: detrital origin inferred from compositional and osmiumisotope data. Canadian Mineralogist 42, 631 650. Maynard, J.B., Klein, G.D., 1995. Tectonic subsidence analyses in the characterization of sedimentary ore deposits: examples from the Witwatersrand (Au), White Pine (Cu), and Molango (Mn). Economic Geology 90, 37 50. Maynard, J.B., Ritger, S.D., Sutton, S.J., 1991. Chemistry of sands from the modern Indus river and the Archean Witwatersrand basin: implications for the composition of the Archean atmosphere. Geology 19, 265 268. McCarthy, T.S., Charlesworth, E.G., Stanistreet, I.G., 1986. PostTransvaal structural features of the northern portion of the Witwatersrand Basin. Transactions of the Geological Society of South Africa 89, 311 323. McCourt, S., Armstrong, R.A., 1998. SHRIMP UPb zircon geochronology of granites from the Central Zone, Limpopo Belt, southern Africa: implications for the age of the Limpopo Orogeny. South African Journal of Geology 101, 329 338. McLean, P.J., Fleet, M.E., 1989. Detrital pyrite in the Witwatersrand gold fields of South Africa: evidence from truncated growth banding. Economic Geology 84, 2008 2011. Mellor, E.T., 1916. The conglomerates of the Witwatersrand. Transactions of the Institute of Mining and Metallurgy 25, 226 348.

43

Meyer, F.M., Robb, L.J., Reimold, W.U., De Bruiyn, H., 1994. Contrasting low- and high-Ca granites in the Archaean Barberton Mountainland, southern Africa. Lithos 32, 63 76. Minter, W.E.L., 1978. A sedimentological synthesis of placer gold, uranium and pyrite concentrations in Proterozoic Witwatersrand sediments. In: Miall, A.D. (Ed.), Fluvial Sedimentology. Canadian Society of Petroleum Geologists, pp. 801 829. Minter, W.E.L., 1981. The distribution and sedimentary arrangement of carbon in South African Proterozoic placer deposits. United States Geological Survey Professional Paper 1161, P1 P4. Minter, W.E.L., 1991. Ancient placer gold deposits. In: Foster, R.P. (Ed.), Gold Metallogeny and Exploration. Blackie, London, pp. 283 308. Minter, W.E.L., 1999. Irrefutable detrital origin of Witwatersrand gold and evidence of eolian signatures. Economic Geology 94, 665 670. Minter, W.E.L., Loen, J.S., 1991. Palaeocurrent dispersal patterns of Witwatersrand gold placers. South African Journal of Geology 94 (1), 70 85. Minter, W.E.L., Hill, W.C.N., Kidger, R.J., Kingsley, R.J., Snowden, P.A., 1986. The Welkom goldfield. In: Anhaeusser, C.R., Maske, S. (Eds.), Mineral Deposits of Southern Africa. Geological Society of South Africa, Johannesburg, pp. 497 539. Minter, W.E.L., Feather, C.E., Glathaar, C.W., 1988. Sedimentological and mineralogical aspects of the newly discovered Witwatersrand placer deposit that reflect Proterozoic weathering, Welkom gold field, South Africa. Economic Geology 83, 481 491. Minter, W.E.L., Goedhart, M.L., Knight, J., Frimmel, H.E., 1993. Morphology of Witwatersrand gold grains from the Basal reef: evidence for their detrital origin. Economic Geology 88, 237 248. Mojzsis, S.J., Coath, C.D., Greenwood, J.P., McKeegan, K.D., Harrison, T.M., 2003. Mass-independent isotope effects in Archean (2.5 to 3.8 Ga) sedimentary sulfides determined by ion-microprobe analysis. Geochimica et Cosmochimica Acta 67, 1635 1658. Moore, M., Davis, D.W., Robb, L.J., Jackson, M.C., Grobler, D.F., 1993. Archean rapakivi graniteanorthositerhyolite complex in the Witwatersrand basin hinterland, southern Africa. Geology 21, 1031 1034. Morrison, G.W., Rose, W.J., Jaireth, S., 1991. Geological and geochemical controls on the silver content (fineness) of gold in goldsilver deposits. Ore Geology Reviews 6, 333 364. Moser, D.E., 1997. Dating the shock wave and thermal imprint of the giant Vredefort impact, South Africa. Geology 25, 7 10. Nagy, B., 1993. Kerogens and bitumens in Precambrian uraniferous ore deposits: Witwatersrand, South Africa, Elliot Lake, Canada and the natural fission track reactors, Oklo, Gabon. In: Parnell, J. (Ed.), Bitumens in Ore Deposits. Special Publication Society for Geology Applied to Mineral Deposits. Springer, Berlin, pp. 287 333. Nelson, D.R., Trendall, A.F., Altermann, W., 1999. Chronological correlations between the Pilbara and Kaapvaal Cratons. Precambrian Research 97, 165 189.

44

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146 D.I. (Eds.), The Geology of Gold Deposits: The Perspective in 1988. Economic Geology Monograph vol. 6, Society of Economic Geologists, pp. 598 608. Phillips, G.N., Klemd, R., Robertson, N.S., 1988. Summary of some fluid inclusion data from the Witwatersrand Basin and surrounding granitoids. Memoirs of the Geological Society of India 11, 59 65. Phillips, G.N., Law, J.D.M., Myers, R.E., 1990. The role of fluids in the evolution of the Witwatersrand Basin. South African Journal of Geology 93, 54 69. Phillips, G.N., Law, J.D.M., Stevens, G., 1997. Alteration, heat, and Witwatersrand gold: 111 years after Harrison and Langlaagte. South African Journal of Geology 100, 377 392. Poujol, M., 2001. UPb isotopic evidence for episodic granitoid emplacement in the Murchison greenstone belt, South Africa. Journal of African Earth Sciences 33, 155 163. Poujol, M., Robb, L.J., 1999. New UPb zircon ages on gneisses and pegmatite from south of the Murchison greenstone belt, South Africa. South African Journal of Geology 102, 93 97. Poujol, M., Robb, L.J., Respaut, J.P., Anhaeusser, C.R., 1996. 3.07 2.97 Ga greenstone belt formation in the northeastern Kaapvaal Craton: Implications for the origin of the Witwatersrand Basin. Economic Geology 91, 1455 1461. Poujol, M., Robb, L.J., Respaut, J.P., 1999. UPb and PbPb isotopic studies relating to the origin of gold mineralization in the Evander Goldfield, Witwatersrand Basin, South Africa. Precambrian Research 95, 167 185. Poujol, M., Anhaeusser, C.R., Armstrong, R.A., 2002. Episodic granitoid emplacement in the Archaean AmaliaKraaipan terrane, South Africa: confirmation from single zircon U Pb geochronology. Journal of African Earth Sciences 35, 147 161. Pretorius, D.A., 1975. The depositional environment of the Witwatersrand goldfields: a chronological review of speculations and observations. Minerals Science and Engineering 7, 18 47. Pretorius, D.A., 1981. Gold and uranium in quartzpebble conglomerates. Economic Geology, 75th Anniversary Volume, 117 138. Pretorius, D.A., 1991. The sources of Witwatersrand gold and uranium: a continued difference of opinion. In: Hutchinson, R.W., Grauch, R.I. (Eds.), Historical Perspectives of Genetic Concepts and Case Histories of Famous Discoveries. Economic Geology Monographs vol. 8, Society of Economic Geologists, pp. 139 163. Ramdohr, P., 1958. New observations on the ores of the Witwatersrand in South Africa and their genetic significance. Transactions of the Geological Society of South Africa 61, 1 50. (Annexure). Rasmussen, B., Buick, R., 1999. Redox state of the Archean atmosphere: evidence from detrital heavy minerals in ca. 32502750 Ma sandstones from the Pilbara Craton, Australia. Geology 27, 115 118. Rasmussen, S.E., Fesq, H.W., 1973. Neutron activation analysis of samples from the Kimberley reef conglomerate. Report No. 1563, National Institute for Metallurgy, Johannesburg, 241 pp.

Nesbitt, H.W., Markovics, G., 1997. Weathering of granodioritic crust, long-term storage of elements in weathering profiles, and petrogenesis of siliciclastic sediments. Geochimica et Cosmochimica Acta 61, 1653 1670. Nesbitt, H.W., Young, G.M., 1982. Early Proterozoic climates and plate motions inferred from major element chemistry of lutites. Nature 299, 715 717. Nesbitt, H.W., Young, G.M., 1984. Prediction of some weathering trends of plutonic and volcanic rocks based on thermodynamic and kinetic considerations. Geochimica et Cosmochimica Acta 48, 1523 1534. Nesbitt, H.W., Young, G.M., 1996. Petrogenesis of sediments in the absence of chemical weathering: effects of abrasion and sorting on bulk rock composition and mineralogy. Sedimentology 43, 341 358. Newman, M.J., Rood, R.T., 1977. Implications of solar evolution for the Earths early atmosphere. Science 198, 1035 1067. Ohmoto, H., 2004. The Archaean atmosphere, hydrosphere and biosphere. In: Eriksson, P.G., Altermann, W., Nelson, D.R., Mueller, W.U., Catuneanu, O. (Eds.), The Precambrian Earth: Tempos and Events. Developments in Precambrian Geology vol. 12. Elsevier, Amsterdam, pp. 361 388. Ohmoto, H., Kakegawa, T., Lowe, D.R., 1993. 3.4-billion-year-old biogenic pyrite from Barberton, South Africa: sulphur isotope evidence. Science 267, 555 557. Ohmoto, H., Rasmussen, B., Buick, R., Holland, H.D., 1999. Redox state of the Archean atmosphere: evidence from detrital heavy minerals in ca. 32502750 Ma sandstones from the Pilbara Craton, Australia: comment and reply. Geology 27, 1151 1152. Palmer, J.A., 1986. Paleoweathering in the Witwatersrand and Ventersdorp Supergroups. Unpubl. MSc thesis, Univ. Witwatersrand, Johannesburg, 166 pp. Parkhurst, D.L., Appelo, C.A.J., 1999. Users guide to PHREEQC (Version 2)a computer program for speciation, batch-reaction, one-dimensional transport, and inverse geochemical calculations. Water Resources Investigations Report vol. 99-4259. U.S. Geological Survey, Denver, 312 pp. Passeraub, M., Wqst, T., Kreissig, K., Smit, C.A., Kramers, J.D., 1999. Structure, metamorphism, and geochronology of the Rhenosterkoppies Greenstone Belt, South Africa. South African Journal of Geology 102, 323 334. Pavlov, A.A., Kasting, J.F., 2002. Mass-independent fractionation of sulfur isotopes in Archean sediments: strong evidence for an anoxic Archean atmosphere. Astrobiology 2, 27 41. Pavlov, A.A., Brown, L.L., Kasting, J.F., 2001a. UB shielding of NH3 and O2 by organic hazes in the Archean atmosphere. Journal of Geophysical Research 106, 23267 23287. Pavlov, A.A., Kasting, J.F., Eigenbrode, J.L., Feeman, K.H., 2001b. Organic haze in Earths early atmosphere: source of low-13C late Archean kerogens? Geology 29, 1003 1006. Phillips, G.N., Law, J.D.M., 1994. Metamorphism of the Witwatersrand gold fields: a review. Ore Geology Reviews 9, 1 31. Phillips, G.N., Law, J.D.M., 2000. Witwatersrand gold fields: Geology, genesis and exploration. SEG Reviews 13, 439 500. Phillips, G.N., Myers, R.E., 1989. Witwatersrand gold fields: Part II. An origin for Witwatersrand gold during metamorphism and associated alteration. In: Keays, R.R., Ramsay, W.R.H., Groves,

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146 Reid, A.M., le Roex, A.P., Minter, W.E.L., 1988. Composition of gold grains in the Vaal Placer, Klerksdorp, South Africa. Mineralium Deposita 23, 211 217. Robb, L.J., Davis, D.W., Kamo, S.L., 1990. UPb ages on single detrital zircon grains from the Witwatersrand Basin, South Africa: constraints on the age of sedimentation and on the evolution of granites adjacent to the basin. The Journal of Geology 98, 311 328. Robb, L.J., Davis, D., Kamo, S.L., Meyer, F.M., 1992. Ages of altered granites adjoining the Witwatersrand Basin with implications for the origin of gold and uranium. Nature 357, 677 680. Robb, L.J., Charlesworth, E.G., Drennan, G.R., Gibson, R.L., Tongu, E.L., 1997. Tectono-metamorphic setting and paragenetic sequence of AuU mineralisation in the Archaean Witwatersrand Basin, South Africa. Australian Journal of Earth Sciences 44, 353 371. Roering, C., 1990. The Vredefort structure: a perspective with regard to new tectonic data from adjoining terranes. Tectonophysics 171, 7 22. Rundle, C.C., Snelling, N.J., 1977. The geochronology of uraniferous minerals in the Witwatersrand Triad: an interpretation of new and existing UPb data on rocks and minerals from the Dominion Reef, Witwatersrand and Ventersdorp Supergroups. Philosophical Transactions of the Royal Society of South Africa 286, 567 583. Rye, R., Holland, H.D., 2000. Life associated with a 2.76 Ga ephemeral pond? Evidence from Mount Roe #2 paleosol. Geology 28, 483 486. Rye, R.O., Holland, H.D., 1998. Paleosols and the evolution of atmospheric oxygen: a critical review. American Journal of Science 298, 621 672. Saager, R., Oberthqr, T., 1984. Nickelcobalt sulfides in Precambrian gold and uranium-bearing quartzpebble conglomerates of South Africa. In: Wauschkuhn, A., Kluth, C., Zimmermann, R.A. (Eds.), Syngenesis and Epigenesis in the Formation of Mineral Deposits. Springer, Berlin, pp. 37 246. SACS, South African Committee for Stratigraphy, 1980. Part 1: Lithostratigraphy of the Republic of South Africa, South West Africa/Namibia, the Republics of Bophuthatswana, Transkei and Venda. Pretoria, Department of Mineral and Energy Affairs, Geological Survey, 690 pp. Sagan, C., Mullen, G., 1972. Earth and Mars: evolution of atmospheres and surface temperatures. Science 177, 52 56. Sanders, J.W., Rowland, T.W., Mellody, M., 1994. Formation related gold production from the Central Rand Group and the Ventersdorp Supergroup, South Africa. 15th CMMI Congress. Institute of Mining and Metallurgy, Johannesburg, pp. 47 53. Schidlowski, M., 1981. Uraniferous constituents of the Witwatersrand conglomerates: ore-microscopic observations and implications for Witwatersrand metallogeny. United States Geological Survey Professional Paper 1161, N1 N29. Schmitz, M.D., Bowring, S.A., de Wit, M.J., Gartz, V., 2004. Subduction and terrane collision stabilize the western Kaapvaal craton tectosphere 2.9 billion years ago. Earth and Planetary Science Letters 222, 363 376.

45

Shen, Y., Buick, R., Canfield, D.E., 2001. Isotopic evidence for microbial sulphate reduction in the early Archaean era. Nature 410, 77 81. Slingerland, R., Smith, N.D., 1986. Occurrence and formation of water-laid placers. Annual Review of Earth and Planetary Sciences 14, 113 147. Smith, N.D., Minter, W.E.L., 1980. Sedimentological controls of gold and uranium in two Witwatersrand paleoplacers. Economic Geology 75, 1 14. Spangenberg, J., Frimmel, H.E., 2001. Basin-internal derivation of hydrocarbons in the Witwatersrand Basin, South Africa: evidence from bulk and molecular y13C data. Chemical Geology 173, 339 355. Stanistreet, I.G., McCarthy, T.S., 1991. Changing tectono-sedimentary scenarios relevant to the development of the Late Archaean Witwatersrand Basin. Journal of African Earth Sciences 13, 65 81. Stevens, G., Preston, R.F., 1999. The metamorphic and alteration history of West Rand Group shales from distal portions of the Witwatersrand Basin. Mineralogy and Petrology 66, 123 147. Stevens, G., Gibson, R.L., Droop, G.T.R., 1997. Mid-crustal granulite facies metamorphism in the central Kaapvaal craton: the Bushveld Complex connection. Precambrian Research 82, 113 132. Strauss, H., 2003. Sulphur isotopes and the early Archaean sulphur cycle. Precambrian Research 126, 349 361. Sutton, S.J., Ritger, S.D., Maynard, J.B., 1990. Stratigraphic control of chemistry and mineralogy in metamorphosed Witwatersrand quartzites. The Journal of Geology 98, 329 341. Tankard, A.J., et al., 1982. Crustal Evolution of Southern Africa. Springer, New York, 523 pp. Tinker, J., de Wit, M., Grotzinger, J., 2002. Seismic stratigraphic constraints on NeoarcheanPaleoproterozoic evolution of the western margin of the Kaapvaal Craton, South Africa. South African Journal of Geology 105, 107 134. Towe, K.M., 1990. Aerobic respiration in the Archaean? Nature 348, 54 56. Trendall, A.F., 2002. The significance or iron-formation in the Precambrian stratigraphic record. In: Altermann, W., Corcoran, P.L. (Eds.), Precambrian Sedimentary Environments: A Modern Approach to Depositional Systems. Special Publication International Association of Sedimentologist vol. 44, Blackwell, Oxford, pp. 33 66. Tucker, R.F., 1980. The sedimentology and mineralogy of the Composite Reef on Cooke Section, Randfontein Estates Gold Mine, Witwatersrand, South Africa. Unpubl. MSc Thesis, University of the Witwatersrand, Johannesburg, 355 pp. Tweedie, E.B., 1986. The Evander goldfield. In: Anhaeusser, C.R., Maske, S. (Eds.), Mineral Deposits of Southern Africa. Geological Society of South Africa, Johannesburg, pp. 705 730. Utter, T., 1979. The morphology and silver content of gold from the Upper Witwatersrand and Ventersdorp systems of the Klerksdorp gold field, South Africa. Economic Geology 74, 27 44. Vennemann, T.W., Kesler, S.E., ONeil, J.R., 1992. Stable isotope compositions of quartz pebbles and their fluid inclusions as tracers of sediment provenance: Implications for gold- and

46

H.E. Frimmel / Earth-Science Reviews 70 (2005) 146 Zartman, R.E., Frimmel, H.E., 1999. Radon-generated 206Pb in hydrothermal sulphide minerals and bitumen from the Ventersdorp Contact Reef, South Africa. Mineralogy and Petrology 66, 171 191. Zhao, B., Clauer, N., Robb, L.J., Zwingmann, H., Toulkeridis, T., 1999. KAr dating of white micas from the Ventersdorp Contact Reef of the Witwatersrand Basin, South Africa: timing of postdepositional alteration. Mineralogy and Petrology 66, 149 170.

uranium-bearing quartz pebble conglomerates. Geology 20, 837 840. Vennemann, T.W., Kesler, S.E., Frederickson, G.C., Minter, W.E.L., Heine, R.R., 1995. Oxygen isotope sedimentology of gold- and uranium-bearing Witwatersrand and Huronian Supergroup quartzpebble conglomerates. Economic Geology 91, 322 342. Walker, J.C.G., et al., 1983. Environmental evolution of the ArcheanEarly Proterozoic Earth. In: Schopf, J.W. (Eds.), Earths Earliest Biosphere: Its Origin and Evolution. Princeton University Press, Princeton, pp. 260 290. Walraven, F., Hattingh, E., 1993. Geochronology of the Nebo Granite, Bushveld Complex. South African Journal of Geology 96, 31 41. Walraven, F., Martini, J., 1995. Zircon Pb-evaporation age determinations of the Oak Tree Formation, Chuniespoort Group, Transvaal Sequence: implications for TransvaalGriqualand West basin correlations. South African Journal of Geology 98, 58 67. Walsh, M.M., 1992. Microfossils and possible microfossils from the Early Archean Onverwacht Group, Barberton Mountain land, South Africa. Precambrian Research 54, 271 293. Wronkiewicz, D.J., Condie, K.C., 1987. Geochemistry of Archean shales from the Witwatersrand Supergroup, South Africa: Source-area weathering and provenance. Geochimica et Cosmochimica Acta 51, 2401 2416. Young, G.M., 2004. Earths two great Precambrian glaciations: aftermath of the bsnowball Earth hypothesisQ. In: Eriksson, P.G., Altermann, W., Nelson, D.R., Mueller, W.U., Catuneanu, O. (Eds.), The Precambrian Earth: Tempos and Events. Developments in Precambrian Geology vol. 12. Elsevier, Amsterdam, pp. 440 448. Young, G.M., von Brunn, V., Gold, D.J.C., Minter, W.E.L., 1998. Earths oldest reported glaciation: physical and chemical evidence from the Archean Mozaan Group (2.9 Ga) of South Africa. The Journal of Geology 106, 523 538.

Hartwig E. Frimmel, PhD, is Professor at the University of Wqrzburg, Germany, where he heads the Institute of Mineralogy. He is also Professor at the Department of Geological Sciences, University of Cape Town, where for the past 15 years he has worked inter alia on the genesis of the Witwatersrand gold deposits. Other major research interests include the relationship between plate tectonics, palaeo-climate and syn-sedimentary ore-forming processes, with particular focus on the Neoproterozoic Aera, and the geodynamic evolution of Precambrian supercontinents, with regional emphasis on Antarctica, Africa and South America. Since 1998 he has been the leader of the Earth Science subprogramme within the South African National Antarctic Programme. He has served on several editorial boards, supervised numerous post-graduate students and has over 80 publications in international journals and books to his credit.

You might also like