Witwatersrand gold-pyrite-uraninite deposits do not support a reducing Archean atmosphere
Published:January 01, 2006
- PDF LinkChapter PDF
Jonathan Law, Neil Phillips, 2006. "Witwatersrand gold-pyrite-uraninite deposits do not support a reducing Archean atmosphere", Evolution of Early Earth's Atmosphere, Hydrosphere, and Biosphere - Constraints from Ore Deposits, Stephen E. Kesler, Hiroshi Ohmoto
Download citation file:
The first serious suggestion that the Archean atmosphere was reducing was based on the interpretation of round uraninite and pyrite grains in the Witwatersrand Basin in the early 1950s. It was then inferred that these minerals were detrital and that they reflected equilibrium with a reducing Archean atmosphere.
Over the past 20 years the understanding of the Witwatersrand Basin has changed dramatically with more integrated studies of the basin and the recognition of widespread alteration in close spatial association with the mineralization in every goldfield. Post-depositional mobility of gold, sulfur, and uranium during alteration is widespread and supports hydrothermal ore genesis, or at least substantial modification of the original mineral assemblage. Pseudomorphic replacements of pre-existing detrital minerals (e.g., pyrite after titano-magnetite), and precipitation and/or chemical rounding to generate round mineral shapes (e.g., uraninite in carbon seams) have all been documented.
The recognition that the carbon seams formed by the post-depositional introduction and maturation of migrated hydrocarbons is a dramatic departure from earlier models of coalified algal material deposited with the sediments. The enrichment of both gold and uraninite in carbon seams implies that these minerals are hydrothermal and that their shapes do not reflect detrital processes. Uranium mobility in basinal waters may in fact require a relatively oxidizing atmosphere.
None of the existing arguments for the Witwatersrand mineralization unambiguously support a placer or modified placer model for the mineralization. Consequently, round uraninite and pyrite of the Witwatersrand Basin do not provide support for a reducing Archean atmosphere.
Since the early 1950s, it has been suggested that the Archean atmosphere was “reducing,” and that it may not have contained significant amounts of free oxygen until ca. 2 Ga. However, the geological evidence for a reducing Archean atmosphere is both ambiguous and controversial (see reviews by Dimroth and Kimberley, 1976; Clemmey and Badham, 1982; Palmer et al., 1987; Ohmoto, 1996; Phillips et al., 2001).
In fact it was the discovery of widespread uraninite in the Witwatersrand Basin in the 1950s that led to the first serious suggestion of a reducing Archean atmosphere. This hypothesis arose from combining the widely held beliefs that the gold grains in the Witwatersrand conglomerates were detrital in origin, and that round uraninite and pyrite grains spatially associated with the gold probably had a similar origin. As the uraninite grains could not have survived detrital transportation under oxidizing atmospheric conditions, a reducing Archean atmosphere was proposed. Subsequently, paleosols were inferred in and around the Witwatersrand Basin (and elsewhere), and their chemical composition was inferred to support the reducing atmosphere hypothesis (Button and Tyler, 1981).
These models for a reducing Archean atmosphere were developed in an era of overwhelming dominance of the placer and modified placer models for the Witwatersrand gold mineralization. As a result, research focused on the sedimentology of the host rocks, and there was a widespread belief that Witwatersrand rocks were “unaltered” and “unmetamorphosed,” effectively precluding the possibility of a hydrothermal origin for the mineralization. Over the past 20 years this picture has changed dramatically with more integrated studies of the basin and the recognition of widespread alteration in close spatial association with the mineralization in every goldfield. In the light of these developments, it is appropriate to critically review new and existing data for the Witwatersrand to assess their relevance as constraints on the Archean atmosphere. In particular we review the arguments linking mineralization to the Archean atmosphere, i.e., whether or not the gold, pyrite, and uraninite are primary detrital grains.
The possible link between mineralization and the Archean atmosphere has important implications for ore genesis beyond the Witwatersrand. Although a reducing Archean atmosphere is “built in” to many existing genetic models for a variety of orebodies, the alternative of an oxidizing atmosphere would require critical re-evaluation of the origins of many Archean orebodies that assume reducing atmospheric conditions. Also, new exploration opportunities for Witwatersrand-style exploration targets may exist in younger sedimentary basins.
DEFINITION OF ATMOSPHERIC REDOX STATE
Most models for the early atmosphere assume that it was reducing when Earth formed, and subsequently changed to be oxidizing as it is today. A reducing atmosphere in the very earliest Archean is common to all models and, in the absence of geological constraints, is based exclusively on conceptual studies of the early evolution of Earth, Sun, and the solar system. Two basic models have been proposed:
An oxidizing Archean atmosphere: an oxidizing atmosphere evolved very early in the evolution of Earth and was already well developed in the early Archean (i.e., by 3.5 Ga; Palmer et al., 1987, 1989; Ohmoto, 1996).
A reducing Archean atmosphere: Earth's atmosphere was reducing until ca. 2.4–2.0 Ga, when a proliferation of photosynthetic organisms resulted in gradually increasing levels of oxygen in the atmosphere (Krupp et al., 1994; Holland et al., 1994; Kasting, 1993; 2001; Rye and Holland, 1998; Farquhar et al., 2000).
There are several different understandings of what constitute “reducing” and “oxidizing” conditions in the early atmosphere. Holland (1984) has used the composition of inferred “paleosols” levels relative to present atmospheric compositions, levels relative to present atmospheric compositions, of inferred “paleosols” to express O2 levels relative to present atmospheric compositions, whereas others have argued that such calculations are premature before conclusive evidence is presented demonstrating that the paleosols are unaltered, or indeed are paleosols at all (Palmer et al., 1987; 1989).
We base our working definition of atmospheric redox state on the geochemistry of iron, and this definition is implicit in many other studies. For example, supporters of the “reducing” model is too low to stabilize assume that the level of atmospheric O2 significant Fe3+ (e.g., when drawing conclusions on iron mobility in inferred “paleosols”; Holland, 1984; 1994). Iron is one of the most important components in Earth's crust, and knowledge of its behavior and its oxidation states is possibly the most important factor related to the early atmosphere affecting our understanding of early geological processes. The chemical behavior of iron is strongly influenced by whether near-surface environments, including aqueous solutions, had abundant Fe2+ and Fe3+, or simply Fe2+. If the former (i.e., oxidizing atmosphere), then many near-surface processes would be similar to today; if the latter (i.e., reducing atmosphere), then none of the near-surface processes that involve and require abundant iron in two oxidation states (e.g., laterite formation and iron pisolith formation) would be operative. An “oxidizing” environment in terms of this definition does not necessarily imply an atmospheric composition exactly the same as today: it means oxidizing enough to facilitate geological processes that require the presence of two oxidation states of iron. Calculations of the amount of free oxygen necessary to stabilize ferric iron minerals suggest that free oxygen may have been several orders of magnitude less than the present atmospheric level (e.g., Krupp et al., 1994; Holland, 1994). Any more precise definition is premature given uncertainties in interpreting basic geological relationships and then inferring the associated chemical environment. For example, absolute thermodynamic calculations require many assumptions on the origin and evolution of the mineral assemblages, chemical boundary conditions, and kinetic controls on reaction progress.
The stability of detrital pyrite and uraninite are directly linked to atmospheric redox conditions. In a reducing atmosphere, detrital uraninite and pyrite would be meta-stable along with a suite of other sulfide minerals, and iron might be leached from detrital iron oxides. In an oxidizing atmosphere, many iron oxide minerals would be stable and uranium would be highly soluble, destabilizing detrital uraninite. Oxidation of U4+ minerals such as uraninite and brannerite occurs readily under modern atmospheric conditions, making their persistence during detrital transport highly unlikely in an oxidizing atmosphere (Davidson, 1953, 1957; Davidson and Cosgrove, 1955; Holland, 1994), though remotely possible given localized modern occurrences (Robinson and Spooner, 1984; Maynard et al., 1991).
OVERVIEW OF PYRITE, GOLD AND URANIUM IN CONGLOMERATES
Geological Overview and Distribution
Gold, pyrite, and uraninite in fluvial sediments of continental basins have been variably described as Witwatersrand-type, conglomerate-hosted, quartz pebble-associated, and quartz arenite-associated mineralization. Each of these descriptors focuses on the nature of the host rocks and each implicitly relies on a placer origin of the mineralization to provide useful discrimination from other deposit styles. In fact, the genesis of conglomerate-hosted pyrite and uraninite ore bodies was once thought to be so well understood that the most commonly used name for the orebodies included the genetic descriptor “paleoplacer.” Numerous occurrences of gold (with or without pyrite and uraninite) in conglomerates are known from around the world but only a few have been significant economic producers (Fig. 1). Of all these, the Witwatersrand Basin in South Africa is by far the largest producer of both gold and uranium. All-time Witwatersrand production now exceeds 50,000 t of gold from seven goldfields. In each goldfield there is typically one reef horizon that has produced the majority of the gold (e.g., Kimberley Reef at Evander, Main Reef Leader in the Rand goldfields, Carbon Leader Reef at Carletonville, Vaal Reef at Klerksdorp, and Basal Reef at Welkom). With the exception of Evander, all goldfields also contain one or more subordinate reef horizons that are economic over smaller areas, and in the case of the West Rand goldfield there are at least ten such reefs (Phillips and Law, 2000).
Both placer and hydrothermal models have been proposed for Witwatersrand-type pyrite, uraninite, and gold mineralization. Each of these models acknowledges the relationship between gold distribution and rock type but provides differing explanations. According to the unmodified placer model gold and uranium were introduced as placer concentrations during sedimentation. The model assumes that detrital gold, pyrite, and uraninite grains have the same shape, composition, and location now as they had at their time of burial and hence can be used to constrain source terranes and their compositions. The modified placer model is similar but assumes that although gold and related minerals were detrital in origin they have been locally remobilized and recrystallized after burial to account for secondary grain shapes and compositions. A scale of movement in the millimeter to centimeter range is generally inferred. The hydrothermal replacement model proposes that the gold and uranium mineralization is not detrital and was introduced by the circulation of hydrothermal fluids in the basin after burial (see Robb and Meyer, 1995 and Phillips and Law, 2000 for opposing views on the merits of these models).
The placer and modified placer models have dominated thinking throughout much of the twentieth century and have been used to support the reducing Archean atmosphere hypothesis.
Time Distribution of “Oxidized” and “Reduced” Witwatersrand-type Deposits
All pyrite-, gold-, and uraninite-bearing conglomerates are older than 1900 Ma and together span the transition from a reducing to oxidizing atmosphere proposed by the reducing Archean atmosphere hypothesis (i.e., between 2.4 and 2.0 Ga; Fig. 2). In South Africa, Witwatersrand-type gold, pyrite, and uranium mineralization is sporadically represented over a considerable period of time from local occurrences in the Archean greenstone basement, through the Dominion Reef, Witwatersrand Supergroup, and Black Reef Formation of the Transvaal Sequence (Fig. 3).
The Jacobina deposits are the youngest economically significant “reduced” conglomerate-hosted gold-pyrite-uranium deposits. Sedimentation at Jacobina is bracketed between 2086 Ma (age of the youngest detrital zircon) and 1883 Ma (age of intrusive post-tectonic granite) (Milesi et al., 2002). In contrast, the oxidized conglomerate-hosted gold ores such as those at Tarkwa have been inferred to postdate the oxygenation of the atmosphere and have been dated between 2124 ± 9 Ma (the age of the youngest detrital zircon) and 1991 ± 12 Ma (the age of intruding granites) by Bossière et al. (1996). Uraninite and pyrite are absent from the Tarkwaian ores and iron oxides dominate the heavy mineral suite. Available geochronological constraints for oxidized mineralization at Tarkwa and reduced mineralization at Jacobina indicate that they may overlap in time. The transition between reduced and oxidized styles of conglomerate-hosted gold mineralization has been used to support a transition from a reduced to oxidizing atmosphere; however, the available age data indicate that they are broadly coeval and cannot both reflect global atmospheric composition.
Common Features of Reduced Deposits
In spite of the variety in terms of age and geological setting, some aspects of the gold-pyrite conglomerates listed in Table 1 (from South Africa and around the world) are remarkably consistent and imply that similar processes have operated in each:
All deposits are located in clastic continental sediments closely associated with active uplift during sedimentation and associated unconformity development.
All contain elevated gold and uranium, although the relative abundance of each varies dramatically both within and among deposits.
All contain round pyrite grains.
Detrital magnetite and ilmenite are virtually absent in all deposits.
All contain uraninite and a variety of U-Ti minerals, including brannerite and leucoxene, that at least in part reflect post-depositional modification of pre-existing uraniferous minerals.
All deposits have an overprint of diagenetic and hydrothermal alteration that modifies the primary detrital mineralogy.
All contain migrated hydrocarbons intimately associated with uranium minerals and inferred to be syn- or post-uranium mineralization in age.
All contain gold that is mostly paragenetically late in the mineralization sequence and is variably interpreted either as hydrothermal in origin or as remobilized detrital gold.
Although the Witwatersrand gold production is more than 100 times larger than any of the other reduced deposits known from elsewhere in the world, the genetic models proposed for each are similar.
THE WITWATERSRAND BASIN
Structural and Sedimentological Setting
The Witwatersrand orebodies are confined to a tectonically preserved remnant of a more extensive sedimentary basin with a complex syn- and post-depositional history (Phillips and Law, 2000). The Witwatersrand sedimentary sequence (Witwatersrand Supergroup) was deposited on the stable granite-greenstone crust of the Kaapvaal Craton. The preserved structural basin is elongate to the NE and is ∼350 km long, 150 km wide, and up to 8 km thick. In a general sense, the stratigraphic record preserves a progressively upward-coarsening depositional sequence dominated by marine sedimentary rocks toward the base and clastic continental sedimentary rocks toward the top (Fig. 3). The lower marine sequence was considerably more extensive than the currently preserved structural basin. The upper parts of the stratigraphic sequence were deposited in response to local syn-depositional deformation near the currently preserved margin (McCarthy, 1994; Coward et al., 1995). Most of the gold mineralization is in the upper sequence of continental sedimentary rocks.
The Witwatersrand Supergroup unconformably overlies the Dominion Group and surrounding Archean granite-greenstone basement and is itself overlain unconformably by the Ventersdorp Supergroup. All three units are Archean in age. The Ventersdorp is overlain by variable thicknesses of the Archean to Proterozoic Transvaal Sequence and generally flat-lying Mesozoic sedimentary rocks of the Karoo Sequence. Rapid lateral variations in the thickness of the cover sequence around the Witwatersrand Basin result from regional unconformities and local structural complexity. Dikes and sills of Ventersdorp, Bushveld, and Karoo age are common throughout the Basin.
The basement granite-greenstone terranes are mostly from 3.3 Ga to 3.1 Ga in age (Robb et al., 1990a, 1990b), and volcanic rocks from the Dominion Group have been dated at 3.07 Ga (Armstrong et al., 1991). By synthesizing available age constraints, Robb et al. (1990a) inferred that that Witwatersrand sedimentation commenced ca. 3.0 Ga and was terminated by extrusion of the Klipriviersberg volcanic rocks ca. 2714 Ma. Within the Witwatersrand Supergroup, the only reliable age is 2914 Ma from the Crown metabasalt toward the top of the West Rand Group (Armstrong et al., 1991).
The structural evolution of the basin has been reviewed by Coward et al. (1995), who have linked basin formation to the progressive tectonic evolution of the surrounding region during the Archean:
Pre-Witwatersrand rifting and deposition of the Dominion Group comprising dominantly basaltic and felsic volcanic rocks and minor sedimentary rocks.
Post-Dominion thermal subsidence and deposition of the dominantly clastic marine lower Witwatersrand succession.
A progressive change to a compressional tectonic regime and the development of an emergent fold-thrust belt along the northern and western basin margins. The upper Witwatersrand succession thus comprises a generally upward-coarsening succession of dominantly fluvial sedimentary rocks forming an asymmetrical foredeep in the north and west, thinning progressively toward the south and east.
Extrusion of the Klipriviersberg flood basalts during the waning stages of compression that effectively terminated Witwatersrand sedimentation at 2714 Ma.
Post-Klipriviersberg extension and deposition of the Platberg Group sedimentary and bimodal volcanic rocks in asymmetric grabens related to northwest-southeast extension ca. 2709 Ma.
The overlying Transvaal Sequence covers an area much greater than the Witwatersrand Basin and comprises extensive clastic and chemical sedimentary rocks and minor volcanic rocks. At the base of the Transvaal, the Black Reef Formation unconformably overlies the Ventersdorp Supergroup and a range of older stratigraphic units, and is developed over a wide area of the Kaapvaal Craton both within and around the Witwatersrand Basin. It comprises a basal conglomerate overlain by quartzite and black shale of fluvial and marine origin, respectively (Tankard et al., 1982; Els et al., 1995). Platform carbonates of the Chuniespoort or Ghaap Groups and fine-grained clastic rocks of the Pretoria Group successively overlie the Black Reef. A Pb-Pb whole rock age from the Chuniespoort Group of the Transvaal sequence of 2557+49 Ma suggests that the base of the Transvaal may be much earlier and that the lower Transvaal rocks are of late Archean age (Jahn et al., 1990). The stratigraphically highest gold mineralization in the Witwatersrand goldfields has been attributed to the mineralized conglomerate at the base of the Black Reef Formation.
Metamorphism and Alteration
Metamorphic assemblages indicate that greenschist facies conditions of 300–400 °C and 2–3 kbars were reached in each goldfield (Phillips and Law, 1994). Metamorphism was accompanied by widespread alteration that affected virtually the entire upper Witwatersrand succession in every goldfield and resulted in the progressive conversion of pre-existing detrital and diagenetic assemblages to muscovite, chlorite, pyrophyllite, and chloritoid. Pyrite is very widespread in samples in the Central Rand Group. Fluid flow has been channeled along bedding-subparallel brittle-ductile faults particularly in heterogeneous sedimentary units overlying unconformity surfaces (Phillips, 1988; Barnicoat et al., 1997). Alteration crosscuts stratigraphy, is locally focused by thick, regionally persistent shales, and is spatially related to mineralization (Phillips and Law, 2000). Isotopic ages from alteration assemblages typically reflect younger resetting events, but a minimum age on alteration is inferred from large-scale extensional faults of Platberg age that displace both alteration and mineralization on all scales (i.e., alteration and mineralization predate 2709 ± 4 Ma; Phillips and Law, 2000). Mineralization in the Black Reef and specifically pyrophyllite alteration and related deformation in mineralized Black Reef at South Deep Mine are inconsistent with a pre-Platberg age. This evidence has been used to infer hydrothermal gold introduction with at least some of this gold post–Black Reef in age (Wall et al., 2004).
In addition to the regional-scale alteration, mesoscopic chlorite veinlets, retrogression around shear zones, and late micas and calcite are all common small-scale features in the Witwatersrand and not unlike what is found in other sedimentary basins. These small-scale alteration systems typically overprint the regional alteration and are not related to mineralization.
Alteration has resulted in the loss of mobile cations, including Na, Ca, and locally K in quartzites and conglomerates, to stabilize the highly aluminous alteration assemblages. Given the similarity between cation loss during alteration and that during weathering, several authors have argued that the unusual bulk-rock compositions reflect weathering at the source, during transport, or after deposition of the sediment (Sutton et al., 1990; Reimer, 1985). However, recent studies have shown that alteration zones cut stratigraphy on a local and regional scale and are spatially related to bedding-subparallel shear zones and brittle fractures that channeled fluid flow (Barnicoat et al., 1997; Phillips and Law, 1997; 2000). The large-scale addition of K is a feature of alteration, but less likely during weathering.
Alteration of Witwatersrand sedimentary rocks has even greater implications for the interpretation of the Archean atmosphere in that many of the early examples of reduced “paleosols” from the Witwatersrand are now recognized as alteration zones (Palmer et al., 1987; 1989). Similarly, studies of arenite textures and unaltered sedimentary rocks show extensive evidence of hydrothermal alteration superimposed on preexisting weathering and/or diagenetic assemblages (Phillips et al., 1990; Law et al., 1990; Law, 1991; Phillips and Law, 2000)
Mineralization and Paragenesis
Gold, uranium, pyrite, and hydrocarbons are important components of all “Witwatersrand-style” deposits worldwide although their relative abundance varies dramatically both within and among deposits (Table 1). In the Witwatersrand, paragenetic studies on the timing of gold grains and related minerals have yielded diametrically opposing conclusions: that some (but not all) of the gold is texturally early (i.e., detrital; Minter, 1999) or that the gold is late (i.e., hydrothermal or remobilized; Davidson, 1960; Ramdohr, 1958). Nevertheless, there is now a broad consensus in the literature that the bulk of the gold is texturally late with (Frimmel et al., 1993; Minter, 1999) or without (Phillips and Law, 1997; Barnicoat et al., 1997) a significant placer gold contribution. Differences still exist as to the timing and paragenetic relationships among the most closely related ore components, i.e., the gold, carbon, uraninite and pyrite; in spite of local complexities, however, most studies indicate a similar temporal sequence of early uraninite followed by hydrocarbons followed by gold (see review in Phillips and Law, 2000; see also Barnicoat et al., 1997; England, 1999; England et al., 2001).
Phillips and Law (2000) have suggested that this paragenetic sequence is compatible with the inferred evolution of the basin with uranium introduction by meteoric water during syn–Central Rand Group tectonic uplift, followed by diagenetic maturation and migration of hydrocarbons, and final introduction of gold by hydrothermal processes. In contrast, Barnicoat et al. (1997) and Jolley et al. (1999) also support a hydrothermal model but suggest that the gold, uranium, and hydrocarbon relate to the one fluid event.
On a larger scale, detailed structural studies of alteration and deformation show a network of bedding-subparallel shears, faults, quartz veins, and fractures that control alteration (Law and Spencer, 1992; Jolley et al., 1999; 2004). Individual fractures within these deformation zones host carbon seams that contain a substantial proportion of the mineralization in many orebodies. In the case of the Carbon Leader Reef, the mineralization is locally hosted entirely within the carbon seam. More commonly, mineralization is hosted by a variety of rock types but hydrocarbon nodules and seams are commonly important. A complete progression from discrete individual nodules to aggregated nodules and finally continuous seams along the margins of shear zones suggests that the hydrocarbons were emplaced during deformation (Law and Spencer, 1992).
The recognition that the carbon seams formed by the post-depositional introduction and maturation of migrated hydrocarbons is a dramatic departure from earlier models that interpreted the seams as coalified material derived from the burial of algal material interbedded with the sediments. The enrichment of both gold and uraninite in carbon seams implies that these minerals are hydrothermal in origin and that their round shapes do not reflect detrital rounding. U mobility in basinal waters may in fact require a relatively oxidizing atmosphere.
THE DILEMMA OF WITWATERSRAND MINERALOGY
All of the economically mineralized Witwatersrand conglomerates accumulated on long-lived regional unconformities and some are moderately enriched in durable minerals such as zircon and chromite. On the basis of their grain shape, chemical and mechanical character, and textural setting, these two minerals almost certainly represent detrital placer accumulations. By inference, many other heavy minerals should have been accumulated at the same time, including magnetite, garnet, ilmenite, and possibly hematite. Most of these minerals exist in significant quantities in potential source rocks in southern Africa (e.g., magnetite in banded iron formations) and they are well represented in the overlying and underlying stratigraphic units spatially removed from mineralization (Fig. 4) (Fuller, 1958; Phillips and Law, 2000; Meyer et al., 1990). However, any visual examination shows the reefs to be devoid of iron oxides (except as magnetite preserved within BIF clasts; e.g., Hirdes and Saager, 1983) whereas U and Ti oxides including rutile, brannerite, leucoxene, and uraninite are common. Iron-bearing minerals are common, but Fe-oxides are virtually absent.
Another dilemma comes from the implications for the source area if the sulfides are inferred to be detrital. Some very common sulfides (especially of the base metals) are under-represented, and some unusual ones are over-represented (e.g., gersdorffite and cobaltite), despite no evidence to suggest this relationship in any reasonable Archean source area. This discrepancy between the predicted placer assemblage and the observed Witwatersrand Reef assemblage must be examined in light of the post-depositional history of the sedimentary sequence but even with local remobilization, the base metal sulfides should be better represented given likely source areas.
Despite the emphasis on sedimentological aspects of the reefs in the modern literature, there are relatively few published whole rock geochemical data for the orebodies and especially for the surrounding lithologies. One consequence is that marked chemical differences between reefs and their enclosing lithologies are not widely reported. A comparison of compositional differences between the Leader Reef in the Welkom goldfield and the surrounding upper Witwatersrand arenites (Fig. 5) illustrates several key relationships with implications for ore genesis:
There is a dramatic enrichment in the orebody of gold and uranium relative to all other elements.
Major elements constituting the silicate assemblage are similar in both units with the exception of Mg and Fe, reflecting the increased abundance of chlorite and iron sulfides in the ore assemblage.
Cr and Zr, which reflect detrital concentrations of chromite and zircon, respectively, are slightly enriched in the Leader Reef. However, the concentrations do not support substantial enrichment of the heavy mineral suite, especially not enough to explain the enormous enrichment of gold and uranium. Fe and Ti also reflect a substantial detrital contribution either as detrital pyrite or as the missing “black sands” component of the reef, including magnetite and ilmenite (prior to sulfidation of the reefs).
Chalcophile metals including Ni, Co, As, and Sb are considerably enriched in the Leader Reef. Although As and Sb are common in many hydrothermal gold deposits, Ni and Co are rare, although they are well represented in hydrothermal uranium deposits, notably the unconformity-associated deposits in Australia and Canada. Cu and Zn are moderately enriched in the Leader Reef but their absolute abundances are relatively low in both the reef and enclosing quartzites.
Enrichments in iron sulfides and associated chalcophile metals similar to those in the Leader Reef in Welkom have been reported from many other Witwatersrand reefs (e.g., Tucker, 1980; Pretorius, 1976a, 1976b; Minter 1978).
Ore Mineralogy: Pyrite, Gold and Uraninite
One of the most striking features of the Witwatersrand conglomerates is the concentration of pyrite. The relationship between pyrite and gold is clear in a stratigraphic sense; most reef horizons have significantly more pyrite than average for the upper Witwatersrand although not all pyrite rich zones are auriferous. In a lateral sense, there is also a general correlation between modal pyrite abundance and gold grades within some reefs but this is less clear when data from many reefs are plotted together (Phillips and Law, 2000). Pyrite takes many forms and has been extensively studied and morphologically classified (e.g., Saager, 1970; see summary in Hallbauer, 1986).
An overwhelming proportion of round pyrite grains are spatially related to scours and other sedimentological features (Fig. 6A) and to heavy minerals that almost certainly form part of the detrital suite (e.g., zircon, chromite). There is thus broad agreement that round pyrite grains are located in positions controlled by the sedimentary dynamics during deposition of the sediments. The atmosphere debate has thus centered on whether the grains were deposited as pyrite or as detrital iron oxide minerals that were pseudomorphously replaced by pyrite.
Three main types of round (“buckshot”) pyrite are common although other textural varieties are also present (see summary in Hallbauer, 1986):
Compact round pyrites (Fig. 6B) comprising single pyrite grains with round shapes and diameters of less than 1 mm that have been widely interpreted as detrital. Alternatively they could be rounded magnetite and titano-magnetite grains that were sulfidized after burial. Round pyrite grains commonly host gold in crosscutting, post-depositional fractures (Fig. 6C).
Porous round or “mudball” pyrite comprising round aggregates of numerous smaller pyrite grains that are 1 mm to 1 cm in diameter and commonly display radial, concentric, or oolitic textures (Figs. 7A [lighter colored nodules to the right of scale bar], 7C) (Ramdohr, 1958; Hallbauer, 1986). An interesting feature of these pyrite grains is that they are considerably larger than other heavy minerals, such as zircon, chromite, and normal (compact round) pyrite found in the conglomerates (e.g., Coetzee, 1965). Porous round pyrites could represent syn-sedimentary growth of sulfides at or near the depositional site (e.g., Hallbauer, 1986) or post-depositional sulfidation of iron oxide pisoliths (Phillips and Law, 2000).
Laminated round pyrites (Fig. 6D) comprising banded layers of fine-grained pyrite giving the appearance of rounded fragments of pyrite. These may represent banded sulfide clasts or replacement of other iron-bearing minerals in a banded precursor (e.g., ferruginous chert).
Euhedral pyrite grains (Fig. 6E) and euhedral overgrowths on round grains (Fig. 6F) are very common in samples from Witwatersrand mines. There is also abundant evidence for sulfidation of a variety of other minerals throughout the Witwatersrand goldfields (Ramdohr, 1958; Phillips and Dong, 1993; Myers et al., 1993) (Figs. 8A–C). In some round pyrite grains, remnants of unsulfidized ilmenite are preserved (Fig. 8D) (Phillips and Dong, 1993). In other cases, exsolution lamellae of rutile are preserved within round pyrite grains indicating their origin by sulfidation of titano-magnetite (Fig. 9D). Rare examples of leached titano-magnetite and ilmenite grains have also been described (Fig. 8E). MacLean and Fleet (1989 have described some compact round pyrite grains with growth zones truncated by round grain boundaries (Figs. 9A, 9B).
Sulfur isotope studies of pyrite grains from the Witwatersrand provide conflicting evidence partly because of conflicting data sets depending on the scale of observation (Fig. 10). Conventional bulk grain analyses show little departure from 0‰ (see review in Phillips and Law, 2000, p. 478). In contrast, SHRIMP (sensitive high-resolution ion microprobe) ion microprobe analysis of micron-sized spots on individual grains show variations of 6‰ within single pyrite grains, 9‰ between adjacent touching grains, and 11‰ in single samples (Eldridge et al., 1993; Phillips and Law, 2000), and 21‰ in round porous pyrite (England, 1999).
There is now widespread agreement that the majority of Witwatersrand gold grains have secondary grain shapes and/or occur in textural sites that postdate deposition of the sediments (e.g., Frimmel, 1997; Frimmel and Gartz, 1997; Barnicoat et al., 1997; Phillips and Law, 2000). Some possible exceptions have been noted by Minter (1999) but the interpretation of these grain shapes remains controversial (Barnicoat et al., 2001).
Individual Witwatersrand gold grains are compositionally homogenous with respect to Ag and Hg (Utter, 1979; Hirdes and Saager, 1983; von Gehlen, 1983; Hallbauer, 1986; Oberthür and Saager, 1986; Reid et al., 1986; Frimmel et al., 1993). On the scale of a single hand specimen, greater variability has been described. Some samples show significant between-grain variability whereas all grains in other samples are compositionally homogeneous. In a study of the Basal Reef, Frimmel and Gartz (1997) demonstrated substantial within-sample variability and inferred a detrital origin for the gold. However, there is a strong mineralogical control on gold grain composition (Fig. 11) with relatively restricted compositional variability in grains spatially associated with chlorite and pyrite and far greater variability within grains associated with quartz grains. This pattern implies an in situ control on gold compositions.
A recent study on within-sample gold homogeneity has described grain shapes from a single sample of the Basal Reef in the Welkom goldfield (Minter et al., 1993; Frimmel et al., 1993). These authors have inferred that 75% of the gold preserves primary detrital gold shapes and that the remaining 25% has typical hydrothermal characteristics. Other workers studying exactly the same sample inferred that all of the gold grains are hydrothermal in origin (Barnicoat et al., 2001), that locations of the grains are controlled by fractures, and that their shapes reflect intergrowths with other secondary minerals. Irrespective of the origin of the gold, the compositions and compositional ranges of all grains are similar, requiring either a remarkably homogeneous source terrane (if detrital) or a post-depositional control on gold compositions (Fig. 12).
On a regional scale, there are substantial variations in the compositions of gold grains among samples. For example, gold grains in two samples from the Ventersdorp Contact Reef (VCR) at East Driefontein mine are compositionally distinct from those at West Driefontein mine some 10 km away on the same stratigraphic horizon (Fig. 12). The degree of variability at each of these locations is similar to that reported by Frimmel et al. (1993) and discussed previously.
Round to “muffin” shaped uraninite grains are common in the Witwatersrand reefs (see summaries in Liebenberg, 1955, Saager, 1968, and England et al., 2001). They typically occur as rounded to subrounded grains although rare euhedral grains have been reported. Grains are typically less than 250 µm in diameter and occur in two dominant associations: (1) as isolated grains often spatially associated with, and partially disaggregated and replaced by, hydrocarbons (Fig. 13); or (2) as concentrated clusters within bedding-subparallel fractures known as carbon seams (e.g., Fig. 8F).
IMPLICATIONS OF WITWATERSRAND MINERALOGY FOR THE ARCHEAN ATMOSPHERE
The placer model for Witwatersrand gold has been the mainstay of the reducing Archean atmosphere hypothesis since the 1950s. Over the past 20 years the geological framework for the Witwatersrand has changed significantly and many of the assumptions underpinning the placer model are no longer valid. In particular, the recognition of post-depositional alteration in and around each of the orebodies requires that the placer model and its implications for the Archean atmosphere be carefully reassessed. Three questions are critical: (1) Are the round pyrite grains detrital? (2) Do gold compositions reflect atmospheric composition? (3) Are round uraninite grains detrital?
Are the Round Pyrite Grains Detrital?
In addition to weathering and mechanical abrasion, post-depositional mechanisms that have modified Witwatersrand detrital assemblages include chemical rounding, grain dissolution/replacement and sulfidation. The significance of these processes for interpretation of round pyrite grains is discussed below.
Chemical Rounding of Iron Sulfides
Many of the textural varieties described above display complex internal structure that is locally truncated by round grain margins (e.g., Ramdohr, 1958). Several recent studies of the Witwatersrand have highlighted the aggressive nature of hydrothermal alteration in the reefs, especially in association with hydrocarbons, suggesting the potential for widespread dissolution and modification of original grain shapes (Gray et al., 1998; Jolley et al., 1999; Phillips and Law, 2000; England et al., 2001). Some compact round pyrite grains have euhedral growth zones truncated by round grain boundaries (Figs. 9A, 9B; MacLean and Fleet, 1989). Similar grains, with textures similar to the Witwatersrand, have been described by the same authors (Fleet et al., 1989) from the greenstone hosted hydrothermal ores at the Agnico-Eagle Mine of the Abitibi greenstone belt of Canada (Fig. 9C) where hydrothermal pyrite grains have been modified by later hydrothermal processes. The Witwatersrand pyrites may thus have formed by similar hydrothermal processes, but MacLean and Fleet have argued that these textures demonstrate that the pyrites were rounded by detrital processes and deposited as a sulfide phase. In the case of the hydrothermal Agnico-Eagle mineralization, this conclusion is demonstrably wrong and similar textures cannot be used to constrain the origin of Witwatersrand sulfides.
Dissolution of Iron Oxides
Iron is highly mobile in reduced, low-sulfur fluids near Earth's surface. It is thus possible that iron could be leached from detrital iron oxides during transport under reducing atmospheric conditions. If such a process occurred, it would be difficult to constrain the extent of the dissolution. However, in some cases textural evidence suggests that some titano-magnetite and ilmenite grains did survive transport to the basin and were subsequently leached in situ to leave skeletal titanium oxides that are probably too delicate to have survived detrital transport (Fig. 8E).
The dissolution of heavy minerals by reduced basinal fluids is a common process in sedimentary basins (Morton and Hallsworth, 1999) and has been important in the Witwatersrand. However, the relative importance of this process and the location of the dissolution process (i.e., in transport versus in situ) remain effectively unconstrained.
Sulfidation of Detrital Iron Oxides
Reduced Witwatersrand-type pyrite and uraninite mineralization is invariably associated with round pyrite and the virtual absence of the “black sand” components typical of modern placers, including magnetite, ilmenite, and hematite, all of which are common components of any likely Witwatersrand source area. Several authors have suggested that these minerals were sulfidized, dissolved, or replaced prior to deposition (e.g., Reimer and Mossman, 1990). Others have argued that because detrital magnetite and ilmenite are common detrital phases in stratigraphy above and below the Witwatersrand orebodies, post-depositional processes must be responsible for their removal (Dimroth and Kimberley, 1976; Clemmey and Badham, 1982).
There is abundant supporting evidence of in situ sulfidation of a variety of other minerals throughout the Witwatersrand goldfields (Figs. 8A–C). Furthermore, euhedral pyrite grains indicate the presence of a fluid capable of mobilizing gold and sulfur and the potential to dissolve and/or re-precipitate pyrite during alteration. This process is widely accepted, but the location and timing of the sulfidation are unconstrained and could have occurred at source, during transport into the basin, or after deposition. Diagnostic textural relationships such as replacement parallel to clast margins are the exception rather than the rule, and the extent of the sulfidation is thus unconstrained.
Perhaps the most telling argument is the regional distribution of the round sulfide assemblages. In each example of reduced Witwatersrand-type pyrite and uraninite mineralization worldwide (Table 1), round pyrites are common, black sand components are virtually absent, and the evidence for sulfidation, in the form of secondary euhedral pyrites and partially sulfidized mineral grains, is widespread. If sulfidation occurred outside the basin, the detrital assemblage (moving up the sequence) must have changed from oxide-dominated to sulfide-only, and then back again, to account for the stratigraphic distribution of assemblages noted above. Furthermore, this process must have been repeated in several localities throughout the world. In situ sulfidation by hydrothermal processes appears to be more likely.
If the current isotopic distribution has not been reset during retrogression, the wide range of sulfur isotopic values for pyrite (Fig. 10) reflects either a source area that supplied heterogeneous detrital pyrite or sulfidation by solutions with variable sulfide-sulfate ratios. The detrital pyrite model (with the variable δ34S thus reflecting processes in the source terrane) has the conflicting requirement of a reducing atmosphere to stabilize pyrite during transport, and a quite oxidizing atmosphere to provide the wide range of δ34S values. If such grains were not the result of an oxidizing atmosphere, a suitable mechanism to generate the observed variability in the absence of an oxygenated atmosphere must be defined.
On the basis of the abundance of pyrite, the absence of iron and iron-titanium oxides, and the textural evidence for pseudomorphous replacement, the most reasonable conclusion is that the origin of any specific round pyrite grain in the Witwatersrand and other altered sedimentary sequences is likely to be ambiguous on the basis of shape alone. This ambiguity currently prevents any definite conclusions to be made about the original proportions of pyrite in the conglomerates and precludes the use of pyrite to constrain the composition of the Archean atmosphere.
This does not necessarily preclude the presence of some detrital pyrite grains—it is possible that uplift of older basin sediments exposed local sources of detrital pyrite that was formed during diagenesis, to erosion and redeposition. The extent of this process and its implications for the composition of the Archean atmosphere cannot be determined in the absence of data on other key variables such as the duration of transport and the physical and chemical nature of the depositional environment.
Redox gradients in the near-surface are commonly extreme, and many modern near-surface waters are not in equilibrium with the atmosphere. Even within modern weathering profiles, the redox potential changes significantly with both time and position in the profile, often in proximity to the water table. Inferences of atmospheric composition based on mineral assemblages and paleosols thus require a clear demonstration of equilibrium with the atmosphere, and that remains a challenge for most studies.
Do Gold Compositions Reflect Atmospheric Composition?
Detrital gold grains in equilibrium with the modern atmosphere are commonly (but not universally) depleted in silver relative to their inferred source, reflecting the preferential leaching of silver in an oxygenated atmosphere (Morrison et al., 1991). In contrast, Witwatersrand gold grains typically contain significant silver (around 10 wt%; range 0–30 wt%). It has thus been suggested that compositional differences between gold in modern placers and in the Witwatersrand may reflect differences in atmospheric composition. This hypothesis is based on two key assumptions: (1) Witwatersrand gold is detrital, and (2) there has been no secondary remobilization and/or post-depositional alteration of gold chemistry (i.e., assuming and unmodified placer model).
There is now a broad consensus in the literature that compositions of gold grains in the Witwatersrand reflect post-depositional processes either by local remobilization and/or re-equilibration of detrital gold (e.g., Frimmel et al., 1993; Frimmel and Gartz, 1997) or by a hydrothermal origin for the gold (e.g., Barnicoat et al., 1997; Phillips and Law, 2000). Gold grain compositions (e.g., Figs. 10, 11) show limited variability, with a range less than that observed in gold grains from auriferous quartz veins cutting the orebodies. As a result, within-sample variability cannot be used to suggest that the grains are derived from a variety of different source rocks and is compatible with a hydrothermal genesis.
In summary, textural studies indicate that the overwhelming majority of gold grains are in structural sites that were not present at the time of deposition and cannot be detrital (unmodified) in origin. Local remobilization of any detrital grains may be possible but gold compositions now reflect post-depositional processes and cannot be used to constrain atmospheric compositions at the time of sedimentation or to date the age of inferred source rocks for detrital gold.
Are Round Uraninite Grains Detrital?
The interpretation of round uraninite grain shapes is controversial. Davidson (1960, p. 155) drew attention to round pitchblende grains in vein deposits from Freiberg, Germany. More recently Phillipe et al. (1993) showed that epigenetic uraninite grains in unconformity uranium deposits of the Athabasca Basin are round and have compositions similar to those in the Witwatersrand. These grains demonstrate that roundness is not necessarily a function of detrital transport.
The textural interpretation of isolated uraninite grains in the Witwatersrand is particularly difficult because of the intimate association of uraninite and carbon in the reefs. This association has variably been ascribed to sedimentological (Minter, 1976; Hallbauer, 1986) or chemical processes (Phillips and Law, 1997; 2000), and it is now widely accepted that the carbon is derived from the thermal maturation of migrated hydrocarbons (e.g., Phillips et al., 1990; Gray et al., 1998; Jolley et al., 2004). Round nodules of carbon, known as “flyspeck carbon,” are not volumetrically abundant in the Witwatersrand, but they are found in virtually every reef and are intimately related to the distribution of uranium. Although these blebs have been interpreted as detrital grains (e.g., Hallbauer, 1986), it is now believed that they are genetically related to the carbon seams and were precipitated around uraniferous minerals by radiolytic polymerization (McCready and Parnell, 1998). Most Witwatersrand carbon nodules and seams contain high uranium concentrations disseminated throughout and imply substantial post-depositional mobility of uranium.
There are at least two post-depositional processes that could account for the intimate association of uraninite with migrated hydrocarbon material. If the uraninite pre-dates the hydrocarbons, then it could precipitate migrating hydrocarbons by radiolytic polymerization (McCready and Parnell, 1998; England et al., 2001). Importantly, this process requires that hydrocarbons postdate the uranium minerals, but does not differentiate between detrital and early hydrothermal uranium mineralization. Alternatively, if the uraninite postdates or is synchronous with the hydrocarbon-bearing fluid, the carbon may represent a localized site of reduction capable of precipitating uraninite from solution.
Criteria to differentiate these processes are not clear-cut. In some cases, the textural evidence suggests that carbon replaced and progressively dismembered pre-existing uraninite grains (e.g., Liebenberg, 1955; Smits, 1984; England et al., 2001) (Figs. 13A, 13B). However, mesoscopic relationships indicate that carbon seams in fractures subparallel to stratigraphy host substantial uraninite and are thus incompatible with a simple detrital origin for at least this part of the mineralization (cf. England et al., 2001). In other cases, uraninite and other uraniferous minerals are finely disseminated within the hydrocarbons and are probably epigenetic precipitates (Fig. 8F). The process of nodule formation results in round, discrete uraninite grains that reflect either precipitation of uraninite or disaggregation of pre-existing grains. In the latter case, rounding of pre-existing grains may be an inevitable consequence of the disaggregation process. England et al. (2001) have argued that the hydrocarbons are precipitated around detrital uraninite grains. Other workers argue that uraninite and hydrocarbon are both localized along structurally controlled fractures (e.g., Law and Spencer, 1992; Barnicoat et al., 1997; Jolley et al., 1999; 2004; Phillips and Law, 2000).
In addition to uraninite, brannerite and uraniferous leucoxene are important uranium-bearing minerals in many Witwatersrand reefs (Hallbauer, 1986), but the concentrations and proportions of these minerals vary greatly on local and regional scales. Textural studies show that interaction of Ti-bearing detrital minerals on unconformity surfaces with uranium in solution has formed widespread secondary brannerite and uraniferous leucoxene (Davidson, 1953; 1957; 1960; Liebenberg, 1955; Ramdohr, 1958; Hallbauer, 1986). The migration of uranium after deposition of the Witwatersrand sediments provides a clear indication of dissolution from surrounding grains or the introduction of uranium by hydrothermal solutions. Similarly, the widespread distribution of round uraninite grains in hydrocarbon seams that postdate the deposition of the sediments demonstrates that round shapes are not related to sedimentary rounding and that uranium has been mobile after deposition of the sediments.
In summary, round uraninite grains in hydrocarbon seams in brittle fractures and along shear zones cannot reflect detrital processes and imply substantial uranium mobility after deposition. Aggressive dissolution of uraninite by hydrocarbons together with widespread uranium mobility casts further doubt on the significance of any grain shapes. Round uraniferous hydrocarbon nodules in granitic rocks surrounding the basin probably reflect the same alteration and have been dated between 2.7 Ga and 2.0 Ga and thus postdate Witwatersrand sedimentation (Klemd, 1999).
Even if it is accepted that some or all of the round uraninite grains in the Witwatersrand, in Elliot Lake, and in the Pilbara Craton are primary detrital minerals, the presence of uraninite as a detrital mineral in the Indus River (Maynard et al., 1991) begs the question of how definitive this criterion can really be in constraining the Archean atmosphere. If the Archean atmosphere truly stabilized uraninite, it is surprising how rare uraninite is in the Archean geological record.
Given the evidence for post-depositional alteration in the Witwatersrand, Rasmussen and Buick (1999) completely discard the Witwatersrand ores as a convincing site of detrital pyrite and uraninite, and focus on heavy mineral assemblages in other largely unmineralized Archean sediments such as those of the Pilbara Craton of Western Australia. These authors have identified pyrite, uraninite and siderite in Archean sedimentary rocks of inferred detrital origin and argue that because the sediments are believed to be unaltered, they are more likely to reflect the composition of the Archean atmosphere. However, the presence of sericite in many of these samples (typically a hydrothermal phase in the Witwatersrand; Law, 1991) and migrated hydrocarbons suggests that they have not escaped post-depositional alteration processes, and this study suffers from the same limitations as those outlined above for the Witwatersrand.
Possible Evidence for an Oxidizing Archean Atmosphere
The atmosphere debate on the Witwatersrand has focused largely on the current heavy mineral assemblages in the belief that they are detrital. Given the evidence for post-depositional changes after burial, inferences of minerals existing prior to alteration may be more relevant to constraining atmospheric composition. For example, there is good reason to believe that vital information may be contained within the sediments immediately overlying the unconformity. These sediments may contain material derived from truncated soil profiles and specifically from the B-horizon in which Fe is concentrated in modern profiles (Anand, 1995). Fe-oxyhydroxide concretions (e.g., pisoliths) formed in areas of seasonal rainfall, for example, are durable, coarse-grained, and have a distinctive physical appearance (Brewer, 1964). Therefore, they should be identifiable even after considerable transport, metamorphism, and alteration (Fig. 7).
The genesis of the porous, round pyrite can be subdivided into models postulating original pyrite and those postulating sulfidation of another mineral. The sulfidation origin for Witwatersrand porous round pyrites was initially proposed by Clemmey (1981) and developed by Phillips and Myers (1989). There is a striking similarity in both morphology and size between the porous round pyrite and modern Fe-concretions (Fig. 7). Furthermore, the distribution of the concretions in residual conglomerates on unconformities is exactly as predicted from modern analogues (Smith and Perdrix, 1982). The mechanical migration and accumulation of these pisoliths can be predicted from their density by analogy with modern sedimentological sorting processes (Anand, 1995). Oxyhydroxide minerals in the pisoliths would be highly susceptible to sulfidation after burial because of their high Fe content. In contrast, ferruginous chert has a lower Fe content and generally undergoes only partial alteration to pyrite (Phillips and Dong, 1993). Textural evidence for sulfidation to form these pyrite grains is equivocal. The distribution of porous round pyrite grains is entirely restricted to unconformity surfaces near the proximal basin margin, and thus they lie within the phyllosilicate-plus-pyrite alteration envelope associated with Witwatersrand mineralization (Phillips and Law, 2000). Unconformity surfaces outside the alteration envelope would be required to determine the pre-alteration mineralogy. Comparable sulfide nodules are uncommon elsewhere in the Archean except where associated with similar mineralization and are thus unlikely to reflect a general process related to the Archean atmosphere.
Traditionally, the origin of the porous round pyrite grains has been linked to sulfidic muds accumulated in the intertidal channel areas of fluvial systems, hence the colloquial name “mud ball pyrite.” These sulfidic muds have been attributed to a reducing atmosphere (Hallbauer, 1986) or to exhalative activity around the margins of the Witwatersrand Basin (Hutchinson and Viljoen, 1988). The former explanation requires an atmosphere different from the present to preserve the pyrite during transport, and fails to explain the association of the “mud balls” with unconformities. The hydrothermal activity required for the latter model appears unlikely given the prevalence of the porous round pyrites in the Witwatersrand Supergroup, the Ventersdorp Contact Reef, and the Black Reef at the base of the Transvaal Sequence. Both ideas are built on the premise that the Witwatersrand pyrite is detrital. As shown above, this premise may not be sound and is not independently verified.
This concept of original Fe-rich pisoliths requires a Precambrian atmosphere oxidizing enough to stabilize two forms of iron in different parts of the Archean soil horizon with at least some zone where ferric iron is abundant. No simple relationship between the observed mineralogy and the composition of the atmosphere is predicted for two reasons. First, atmospheric redox state is only one of several thermodynamic and kinetic constraints on the stability of pyrite and uraninite, and simple assumptions regarding other variables and the whole chain of weathering, erosion, transport, and depositional effects on individual grains are unlikely to reflect any single process reliably (Robinson and Spooner, 1984). For example, detrital uraninite and pyrite grains are known from modern sediments, albeit in minor quantities, and round uraninite and pyrite grains are also known from epigenetic deposits. Second, all sedimentary basins are subject to complex post-depositional alteration and diagenesis of the original detrital suite by basinal fluids that are commonly, but not universally, reducing (Phillips et al., 1990; Morton and Hallsworth, 1999). The impact of these fluids must be recognized and understood before the redox state of the atmosphere at the time of deposition can be interpreted.
Given the abundant evidence for post-depositional modification of the Witwatersrand detrital assemblage and the evidence for post-depositional alteration, sulfidation, and widespread mobility of gold and uranium, we conclude that the mineral assemblage of the Witwatersrand is not a reliable indicator of a reducing atmosphere. In fact, reinterpretation of some porous round pyrites as sulfidized iron pisoliths may require a relatively oxidizing environment to stabilize iron in both its ferric and ferrous forms. The widespread inferred mobility of uranium in the Witwatersrand may also require an oxygenated fluid to stabilize U6+ in solution. Similar sedimentary environments in the younger rock record are frequently altered by inflowing oxygenated meteoric waters that reflect prevailing atmospheric conditions.
Significance of “Old” Isotopic Ages
A large range of mineral and whole rock ages have been published for the Witwatersrand that reflect pre-, syn-, and post-depositional events in the basin (see summaries in Robb and Meyer, 1995 and Robb et al., 1990a). Of particular interest to the atmosphere debate are “old” ages for gold, pyrite, and uraninite that purportedly pre-date sedimentation and thus imply a detrital origin for the mineralization.
Re-Os Isochron “Ages” for Gold and Pyrite
Kirk et al. (2001; 2002) have recently used the Re-Os isochron technique in an attempt to directly date Witwatersrand gold and pyrite. These authors assumed that pyrite and gold form part of the heavy mineral assemblage, and used randomly selected grains from a single sample of the Vaal Reef to construct an isochron with an implied age of 3010 ± 20 Ma.
The Vaal Reef sample is described as “fine-grained quartzitic conglomerate with angular to subrounded horizontally fractured quartz clasts 5-10 mm in size. The conglomerate is clast supported and is approximately 10% matrix and 90% clasts. The matrix consists of sericite, fine-grain quartz and carbon seams/patches. The majority of the pyrite/arsenopyrite, and ∼10% of the visible gold are within the quartz and sericite matrix, while ∼90% of the gold, the uraninite and some of the pyrite are confined to the carbonaceous material” (Kirk et al., 2002).
We interpret the sericitic matrix as part of the phyllosilicate alteration assemblage typical of Witwatersrand orebodies. Much of the gold is located in the carbon seam, whereas the sulfides are from the quartzite matrix. Recent studies by several different authors (Parnell, 1996; Gray et al., 1998; Jolley et al., 1999; 2004; Phillips and Law, 2000; England et al., 2001) concluded that the carbon seams postdate deposition of the sediments and reflect migration of hydrocarbons during post-depositional alteration. Although the precise paragenetic setting of grains used to construct the isochron remains unknown, it is likely that the grains are from different host lithologies, with gold from the carbon seam and pyrite from the host conglomerate. In the case of hydrocarbon-hosted gold, textural evidence invariably shows that the gold was introduced after deposition of the hydrocarbon (Ramdohr, 1958; Saager, 1968; Barnicoat et al., 1997; Phillips and Law, 1997) and is thus not co-genetic with the round pyrite grains as inferred by Kirk et al. (2002).
An important challenge for the Re-Os studies is thus to demonstrate that all grains were part of an isotopically homogeneous initial Os187-Os188 reservoir, which is the foundation of isochron-based dating techniques. In practice this generally means that all grains need to be sourced from a grain population from a single sample all sharing a common genesis. Re-Os dating of round pyrite grains from Steyn Reef in the Welkom goldfield by the same authors clearly reflects this problem and yields imprecise “isochron” ages of 3490 ± 900 Ma. There is no reason to believe that the individual pyrite grains analyzed are derived from an isotopically homogeneous source and therefore we assign no age significance to the “isochron.” In any event, the 3490 Ma age fails to discriminate between detrital and hydrothermal mineralization given the 1800 Ma error.
Post-depositional changes in the composition of gold and pyrite grains are now well established for the Witwatersrand and make the use of gold and pyrite geochemistry for dating of dubious value. In particular, homogenization and/or secondary introduction of gold grains, and crosscutting pyrite veins and overgrowths on round pyrites, are likely to affect isotopic systematics.
U-Pb “Ages” for Uraninite
The pioneering work of Rundle and Snelling (1977) on the geochronology of uraniferous minerals in the Witwatersrand has been widely quoted in the literature in support of the placer model. In their 1977 review of their own and available existing data, these authors infer two discrete age groups: (1) an older population at 3050 ± 50 Ma that predates deposition of the Witwatersrand Basin; and (2) a younger population at 2040 ± 100 Ma that postdates the deposition of the Witwatersrand.
Rundle and Snelling argue that the data reflect detrital uraninite derived from a 3050 Ma source followed by a resetting of the U-Pb system at 2040 Ma within a closed system with limited secondary introduction of uranium. However, they also note that “it would be virtually impossible to distinguish between disturbed detrital systems and disturbed systems with both detrital and authigenic components.” These ages have been widely reported as “uraninite” ages (e.g., Robb and Meyer, 1995); however, a review of the original paper highlights some fundamental shortcomings in the approach used by Rundle and Snelling (1977) and especially the interpretation of the data by other authors.
First, the “ages” were obtained on “portions of the matrix, avoiding as far as possible the pebbles, cut in the form of cubes ∼2cm in size” and thus effectively reflect “bulk rock” analyses rather than individual mineral grains. Second, each age group reported by Rundle and Snelling has been estimated from a series of samples based on their distribution on the Concordia diagram, with many samples showing evidence for both lead and uranium loss. As a result, individual age estimates are highly variable. Third, no paragenetic information on uranium minerals is reported to provide independent geological constraints on the likely validity of the ages. For example, it has now been well documented that many uraninite grains, particularly those hosted by carbon seams filling structural dilation zones, cannot be of detrital origin and must reflect substantial mobility of uranium after burial. Galena is also a common secondary mineral in the reefs and associated quartz veins, suggesting an open system on the scale of the samples discussed above. Similarly, many secondary uranium minerals are now well documented in the reefs and form an unknown proportion of the material sampled. Many of these minerals are poor hosts for lead that was presumably mobilized during alteration. It seems improbable in the light of this petrographic evidence that a closed U-Pb system has operated on either a mineral or whole rock scale.
Rundle and Snelling also report data for “uraninite” and “thucolite” that are apparently from individual mineral separates. Thucolite refers to “thorium-uranium-carbon-oxygen”-bearing material or “carbon nodules” to use the terminology in this paper. These ages range from 2760 Ma for a sample from the Dominion Group to 2540 Ma through 2000 Ma for Witwatersrand samples and are consistently younger than the host sedimentary rocks (although they were thought to be older at the time of the analyses).
In light of these uncertainties and limitations, it is unlikely that the reported ages accurately reflect the age of uraninite or other detrital uranium-bearing minerals. In fact, the younger population reported by Rundle and Snelling could equally have been used to support a hydrothermal origin for the mineralization (although these ages also suffer from the limitations outlined above).
Lithological and Chemical Controls on Witwatersrand Mineralization
Given the close spatial association between gold and uranium in Witwatersrand orebodies, it is possible that there is a genetic link between the two metals. A common assumption has been that both minerals are concentrated in response to sedimentary sorting processes (e.g., Smith and Minter, 1980; Hallbauer, 1986; Smits, 1984; Roscoe and Minter, 1993). These authors point out that concentrations of gold and uranium are commonly correlated over several orders of magnitude, and that metal concentrations are also broadly correlated with sedimentary facies (Figs. 14A, 14B). For example, Smith and Minter (1980) describe better grades in quartzites and conglomerates with pyritic foresets and a general increase in grade from sandstone through conglomerate to carbon seams. Similar associations are widely reported from the Witwatersrand and the implied link between sedimentary facies and gold grade has been used to support a placer model for the mineralization.
From a chemical standpoint these data can also be interpreted to reflect post-depositional mineralizing processes (Phillips and Law, 2000):
an association of better gold grades with round pyrite, reflecting sulfidation of detrital Fe-oxide grains and precipitation of gold transported as sulfur complexes;
an association of better uranium grades with detrital titanium minerals, reflecting precipitation of uranium by reaction with Ti to form brannerite.
The increases in grade could thus reflect either differences in depositional concentration processes between sandstones and conglomerates during sedimentation (placer model) or elevated iron and titanium concentrations as detrital phases during conglomerate deposition (hydrothermal model). However, given the secondary origin of the carbon, the elevated grades in carbon-rich lithologies are more likely to reflect reduction by organic material and hence hydrothermal ore genesis. These mineralogical associations provide strong chemical reasons for the association of mineralization with specific host lithologies.
A major criticism of the placer model has been its inability to explain the variety of mineralized sedimentary rock types either on theoretical grounds or by analogy with young placer deposits (Phillips and Law, 2000). This limitation is most obvious in the case of carbon seams that are locally major host lithologies for gold and uranium. In contrast, the chemical associations of gold with iron (pyrite) and carbon, and uranium with carbon and titanium, are clear in all the major orebodies and crosscut sedimentary facies.
We are grateful to Steve Kesler, Hiroshi Ohmoto, and Michael Kimberley for the invitation to attend the 2002 symposium on the “Evolution of the early atmosphere, hydrosphere, and biosphere.” We would also like to thank coworkers in this field who have been instrumental in shaping our ideas, especially Russell Myers, Judy Palmer, Guoyi Dong, and Martin Hughes. Richard Spencer was a key person in recognizing the structural controls on hydrocarbon seams during mapping at Kinross Mine. Bill Fyfe has provided insightful comments on the early Earth and its atmosphere. We also thank Rob Hough, Katy Evans, and Martin Hughes for their constructive comments on an early draft of this manuscript. Holly Stein and Judy Hannah are thanked for discussions on Re-Os and their early review of this section of the manuscript. We are grateful to Steve Kesler and two anonymous reviewers for insightful and constructive reviews that greatly improved the paper. JL acknowledges an Honorary Research Fellowship at the School of Geosciences, Monash University. NP acknowledges the support of the School of Geosciences, Monash University, and the School of Earth Sciences, University of Melbourne.