Linking Archaean climate change with gold metallogeny Open Access

Although gold is known to have become concentrated locally in the Earth’s crust already prior to 3.0 Ga, as exemplified by hydrothermal Au-rich arsenopyrite from ancient seafloor deposits preserved in the Tartoq Greenstone Belt in Greenland (Saintilan et al. 2020) and shear zone-hosted gold deposits in the Barberton Greenstone Belt in South Africa (Dziggel et al. 2010), the overall gold endowment of rocks older than 2.9 Ga appears poor in comparison to younger ones (e.g., so far <400 t Au has been recovered from the Barberton mines). Although this lack of gold in Eo- to Palaeoarchaean strata may be partly an artefact of poor preservation of such ancient rocks, it is likely also a primary feature. Endogenous gold mineralization styles that played the most significant role in the formation of younger gold deposits, such as porphyry/epithermal and orogenic-type gold systems, are with the above few very minor exceptions younger than 2.75 Ga (e.g., Goldfarb et al. 2001), although much older host rocks of suitable metamorphic grade and tectonic setting exist as exemplified by the Archaean geology of Greenland (Saintilan et al. 2020).

The by far largest concentration of gold in crustal rocks took place at around 2.9 Ga, when the richest, synsedimentary ore bodies in the Witwatersrand Basin in South Africa were formed. After decades of intense debate on the origin of the quartz-pebble conglomerate-hosted Witwatersrand gold over the past decade consensus has emerged on these deposits representing former placers (e.g., Frimmel et al. 2005; Robb and Hayward 2014; Frimmel and Nwaila 2020), although some workers still emphasize the role of postdepositional hydrothermal fluids in their formation (e.g., Fuchs et al. 2021). The latter follow previous arguments that gold occurs in most instances in late paragenetic textural positions and is allegedly associated with supposedly basin-wide acid alteration as summarized by Phillips and Powell (2015). While a syngenetic placer model satisfies the huge wealth of analytical data and field observations collected for more than a century of (largely underground) mining, for a long time this model has been plagued by an apparent lack of a suitable gold source as emphasized by Phillips and Powell (2011), who favoured an entirely epigenetic introduction of the gold into its host rocks by metamorphic fluids from outside the Witwatersrand Basin.

The total amount of gold in the Witwatersrand ore province has been estimated as roughly 90 000 t Au, >53 000 t of which has been already extracted (Frimmel 2014; Frimmel and Nwaila 2020). Much of the remaining resource is located at such great depths (>4 km) that its mining does not seem economically feasible in the foreseeable future. Globally, the other two most important settings for gold deposits are orogenic-type deposits and porphyry–epithermal (Cu–)Au systems, accounting for roughly 30% and 28%, respectively, of all gold ever mined (currently estimated at ca. 205 000 t) and suspected to exist in remaining resources (Frimmel 2018). In comparison, all the other genetic types of gold deposit play only a very minor role. The principal endogenous types of gold deposit, porphyry and orogenic ones (including intrusion-related ones), only started to play a significant role in Earth’s history from about 2.75 Ga onwards as illustrated by the ca. 2.7 Ga Boddington Cu–Au deposit in Western Australia (Turner et al. 2020) and the 2.74 Ga Côté Cu–Au deposit in the Abitibi Greenstone Belt, Canada (Katz et al. 2017). Although this could be merely due to a lack of preservation of such ancient rocks, it is most likely that the lack of modern-style plate tectonics and consequently the lack of a supra-subduction gold fertilisation of the sublithospheric mantle prior to that time is the primary reason (Frimmel 2018). Older examples, such as the 3.2 Ga Talga Talga deposit in Western Australia and the 3.08–3.04 Ga deposits in the Barberton Greenstone Belt in South Africa have produced only a few hundred kilograms of gold and thus are of little significance for the global gold budget. This lack of endogenous >2.75 Ga primary gold deposits raises a major problem for the choice of a suitable source of the Witwatersrand placer gold. Apart from the age constraints, for simple mass balance reasons, none of the major endogenous gold deposit types, be it orogenic or porphyry/epithermal, could have served as source of the Witwatersrand gold and similar Mesoarchaean palaeoplacers on other cratons (e.g., Dharwar and Singhbhum cratons in India). A possible solution for this source problem arose from the recognition that Mesoarchaean surface waters most likely had a composition that enabled a very high Au solubility, that is, four orders of magnitude greater than in today’s river water and seawater (Frimmel 2014; Heinrich 2015). Thus, no specific point sources (as in eroded gold deposits) or specific hinterland lithology are required anymore to explain the unique and extraordinary concentration of gold in Mesoarchaean sediments. Instead, the entire Palaeo- to Mesoarchaean land surface becomes a potential gold source, with leaching of background concentrations of Au by meteoric waters having provided the bulk of the Au in those ancient sedimentary environments (Frimmel and Hennigh 2015). Microbial mats acted as first-order trap for the dissolved Au (Horscroft et al. 2011; Frimmel 2014; Heinrich 2015) and their mechanical reworking released the minute gold particles that subsequently concentrated as micronuggets (<0.1 mm in size) into rich placer deposits (which conspicuously lack larger nuggets of >1 mm in size as would appear more typical of younger placer deposits). This syngenetic model of microbial gold fixation elegantly solves the source problem for the Mesoarchaean gold placers but raises a new question pertaining to the timing of such synsedimentary gold mineralization.

A review of available data on gold grade and tonnage in known Mesoarchaean and Palaeoproterozoic conglomerates worldwide revealed a very punctuated temporal distribution (Frimmel 2018). After a major peak at ca. 2.9 Ga, both with regard to Au grade and tonnage, a roughly exponential decrease in both is noted over geological time in the course of the Neoarchaean and Palaeoproterozoic (Fig. 1). Interestingly, hardly any placer gold deposits are known that are older than 2.9 Ga although in many places, suitable environments seem to have existed. To address this specific enigma is the aim of the current study. The possibility of drastic climate change during the Mesoarchaean Era will be discussed based on a review of field geological/lithological and geochemical proxy data.

Sagan and Mullen (1972) first pointed out the by now well-known “Faint Young Sun Paradoxon”, which highlights that the Earth’s hydrosphere should have been completely frozen in the Archaean due to a lower solar luminosity (70%–80% of the current value) if the concentration of greenhouse gases in the atmosphere had remained at similar levels through Earth’s history. This was not the case as indicated by the geological record, which proves widespread presence of liquid water on the Archaean Earth evidenced by water-laid sedimentary rocks and fossil remnants therein. The solution to this paradoxon has been the assumption of a much higher proportion of greenhouse gases in the Archaean atmosphere, mainly CO₂ and CH₄. Agreement exists that the atmospheric CO₂ level at that time was two to three orders of magnitude higher than today, and those of CH₄ (primarily derived from methanogens) at levels exceeding 1000 ppmv (Kasting 2014; Catling and Zahnle 2020).

Under the above constraints on the Archaean atmosphere, surface temperature should have been even higher than today and O isotope data on cherts and carbonates have been used to infer a hot climate with modelled ocean temperatures in a wide range of 55 °C–85 °C (Knauth and Lowe 2003; Tartèse et al. 2017). Others, however, argued for much cooler conditions with average surface temperatures between 0 °C and 40 °C (Hren et al. 2009; Catling and Zahnle 2020). In any event, warm to hot conditions are contested by the finding of glaciogenic deposits within the Archaean rock record. Grant Young, to whom this paper is dedicated, was the pioneer who described a diamictite from the ca. 2.9 Ga Mozaan Group (Pongola Supergroup) in the eastern Kaapvaal Craton and made a case for it to be glaciogenic (Young et al. 1998). It became the oldest known glacial deposit, although subsequently, De Wit and Furnes (2016) described a supposedly glaciogenic diamictite from the 3.47–3.43 Ga Onverwacht Suite in the Barberton Greenstone Belt in South Africa. All these findings highlight not only a large uncertainty in our understanding of Archaean climate but also open up the possibility of significant climate change in the course of the Archaean. Such climate change could have had far-reaching consequences on the transfer and eventual concentration of gold in the Archaean sedimentary environment.

The economically by far most important area of gold-bearing conglomerates is the Mesoarchaean Witwatersrand Basin in the central Kaapvaal Craton in South Africa. Auriferous quartz-pebble conglomerates occur in a series of stratigraphic positions starting with the Dominion Group, which forms part of one of the oldest intracontinental igneous provinces in the world—the 2.96 Ga Dominion-Nsuze Igneous Province (Paprika et al. 2021). At the base of an up to 2250 m thick sequence, dominated by felsic and mafic volcanic rocks, is a pyritic quartz-pebble conglomerate (Dominion Reef) with a comparatively low Au content of, on average, 0.6 g/t but locally high U concentration with a reported resource of 256 Mt at 0.4 kg/t U₃O₈ (Frimmel 2019).

Unconformably above the Dominion Group follows the almost exclusively siliciclastic Witwatersrand Supergroup, which is divided into the 2.96–2.91 Ga West Rand Group and the 2.90–2.78 Ga Central Rand Group, representing two different stages or types of basin development. The West Rand Group, reaching >5 km in thickness in places, is interpreted as reflecting thermal subsidence after initial drowning of the post-Dominion land surface had led to a passive margin in the south(east) of a landmass north(west) of today’s Witwatersrand. This is suggested by uniform palaeocurrent directions and an upwards increase in the complexity of detrital zircon age spectra and by a facies change from proximal fluvio-deltaic and shoreface to distal offshore environments in a southeasterly direction (Frimmel 2019 and references therein). Correlation with similar aged deposits in the lowermost Mozaan Group (Pongola Supergroup) in the eastern part of the Kaapvaal Craton (Zeh and Wilson 2022) highlights that the extent of this basin was much wider than suggested by outcrops of West Rand Group rocks in the Witwatersrand (Fig. 2). Repeated transgressions and regressions, largely caused by eustatic sea level fluctuations, left behind a shale- and quartzite-dominated sequence, but also several fluvial conglomerates that can contain some gold (Promise, Bonanza, Coronation, Rivas, Government, Buffelsdorn, and Veldschoen reefs, Fig. 3). The total amount of gold extracted from these reefs is, however, with ca. 10 t miniscule in comparison to that from the overlying Central Rand Group (Robb and Robb 1998).

After extrusion of the 2.914 Ga Crown Formation lavas (the only magmatic unit in the entire supergroup), basin inversion led to shoreline propagation and fluvial and braid-plain deposits. Thus the uppermost West Rand Group bears stronger similarities to the Central Rand Group than the lower West Rand Group, that is, foreland basin deposits superimposed on the older passive margin deposits. The Central Rand Group, with a maximum thickness of almost 2.9 km near the basin centre, is dominated by arenitic and conglomeratic fluvio-deltaic deposits from alluvial braid-plains and fans, and it lacks marine deposits except for the shaly Booysens Formation in the middle of the group (Fig. 3). A series of conglomerate units, typically overlying low-angle unconformities, are present and most of them are variably enriched in gold (if constituting an ore body, referred to as reefs). The size and architecture of the Central Rand Basin was probably similar to today’s known distribution of Central Rand Group rocks, ignoring the much younger up-doming of the central portion due to the 2.023 Ga gigantic Vredefort impact (Fig. 2). On a regional scale, the gold is concentrated at the main entry points of complex river systems where they reached the foreland basin as indicated by palaeocurrent directions from all around the basin margins (Minter and Loen 1991). A strong spatial association between Au grade and sedimentary facies highlights the syngenetic nature of the mineralization (Frimmel and Nwaila 2020). Two facies are particularly prone to high Au grade: river channels and wind-blown deflation surfaces. The richest ore bodies existed in the lowermost Central Rand Group (Carbon Leader, Main and South or Nigel Reefs, Figs. 2 and 3). In places, remnants of microbial mats in the form of “carbon seams” are preserved. A biological derivation of these carbon seams is beyond doubt, as evidenced by organic chemical and C isotopic data as well as their stratiform mode of occurrence, and they are not to be confused with evidently hydrothermal pyrobitumen nodules reflecting postsedimentary oil migration (Mossman et al. 2008; Frimmel and Hennigh 2015). The carbon seams can be extremely rich in gold, which has led to the suggestion of this gold having been the principal source of the generally very small micronuggets (typically 100–150 µm in size) in the nearby conglomerates (Frimmel 2014). Larger nuggets sensu stricto are conspicuously absent.

Central Rand Group sedimentation was followed by regional peneplanation at around 2.78 Ga prior to deposition of the largely volcanic Ventersdorp Supergroup at the base of which is another fluvial conglomerate that constitutes a major gold producer, the Ventersdorp Contact Reef (Fig. 3). As with all the reefs in the upper Central Rand Group, it sourced its gold from the erosion of older, slightly tilted reefs that became exposed on palaeosurfaces. The same applies to the youngest reef, the ca. 2.6 Ga Black Reef at the base of the Transvaal Supergroup, which is, however, of only little economic significance. Yet, it highlights the importance of the local availability and erosion of older gold-bearing conglomerates on the Au grade in the comparatively younger placers.

Outside the Kaapvaal Craton, gold-bearing Archaean conglomerates are known from the Pilbara Craton in West Australia, the São Francisco Craton in Brazil and the Dharwar as well as the Singhbhum cratons, both in India. The oldest of these seem to be the Indian examples. In the Singhbhum Craton in eastern India, gold-bearing conglomerates have been reported from two different stratigraphic positions: (i) from the base of the Upper Iron Ore Group in the Badampahar Greenstone Belt in the eastern part of the craton and in the Tomka-Daitari Basin in the southwestern part of the craton and (ii) from the basal Phuljhari Formation at the bottom of the Dhanjori Basin in the northeastern part of the craton (Fig. 4). Their age is only loosely constrained by detrital zircon data as being <3.03 Ga in the Badampahar Greenstone Belt, <2.91 Ga in the Tomka-Daitari Basin, and <3.00 Ga in the case of the Phuljhari Formation (Frimmel et al. 2022). The minimum age is roughly limited by an intrusive ca. 2.8 Ga granite in the Tomka-Daitari Basin (Sreenivas et al. 2019). The overall Au grade in these conglomerates remains unknown but locally up to 2 g/t Au has been reported and widespread artisanal mining attests to some economic potential, in addition to locally elevated U contents due to the presence of detrital uraninite (Frimmel et al. 2022 and references therein). Remnants of microbial mats as in the lower Central Rand Group are absent. Delicate carbonaceous microstructures, initially suspected to be of microbial origin (Chakravarti et al. 2018) have since been recognized as being abiogenic (Chakravarti et al. (unpublished data 2021)).

In southwestern India, quartz-pebble conglomerates at the base of the 2.96–2.72 Ga Bababudan Group in the western Dharwar Craton have long been known to contain elevated Au and U contents due to concentration of detrital gold and uraninite (Viswanath et al. 1988). The oligomictic quartz-pebble conglomerates bear many sedimentological and petrological similarities to the conglomerates in the Witwatersrand (Frimmel et al. 2019). In contrast to the Witwatersrand, the Babadudan Group reflects a very different kind of basin, dominated by volcanism and marine environments. It lacks the numerous intraformational unconformities as can be expected in a foreland basin, and the overall thickness of the only conglomeratic unit at the base is with c. 40 m comparatively small (Frimmel et al. 2019).

In the Pilbara Craton, gold-bearing conglomerates have been the subject of intense exploration and most recently also mining. The target there is the 2.78–2.63 Ga Fortescue Group (Mount Bruce Supergroup), which contains auriferous polymictic, poorly sorted, generally clast-supported, pyrite-rich pebble to boulder conglomerates, and minor conglomeratic sandstone at several stratigraphic positions, that is, at its base and within the 2.766–2.752 Ga Hardey Formation (Fig. 5). The latter is also uraniferous. The sedimentary facies suggested for these deposits are very similar to those in the Witwatersrand, that is, palaeochannel fills, braided fluvial or sheet-wash deposits and alluvial fans (Frimmel 2014). A notable difference to the Witwatersrand ores is the presence of proper nuggets, reaching several millimetres in size. Detrital uraninite particles embedded in bitumen reflect migration of oil, but syngenetic microbial mats as in the Central Rand Group have not been discovered so far. Gold grades at the various deposits are between 2 and 3 g/t, reaching >5 g/t in the underground resource (https://novoresources.com/).

As in all of the above cases, a major regional unconformity also separates the Meso- to Neoarchaean basement in the São Francisco Craton from the overlying supracrustal succession, the 2.65–2.12 Ga Minas Supergroup (Fig. 6). The oldest stratigraphic unit in the latter, the Caraça Group, records the evolution from fluvio-deltaic continental rift to passive margin platform deposits. It contains near its base the siliciclastic 2.65 Ga Moeda Formation, which comprises an up to 180 m thick basal poly- to oligomictic, pyrite-rich quartz-pebble conglomerate and coarse-grained quartz arenite, an up to 80 m thick very fine-grained quartz arenite, overlain by a ca. 100 m thick poorly sorted pebble conglomerate (Minter et al. 1990). This succession has been interpreted to reflect a change from alluvial plains and braided rivers to shallow marine environments. Of interest here is the basal conglomerate because it contains gold at grades as high as 10 g/t, locally reaching as much as 150 g/t (Frimmel 2014). In contrast to the Witwatersrand Reefs, the gold occurs in the form of proper nuggets, exceeding 1 mm in size. Highly variable degrees of rounding of the gold particles reflects differences in transport distance, with some grains having come from very proximal sources, such as orogenic-type gold deposits in the underlying greenstones of the 2.78–2.72 Ga Rio das Velhas Supergroup (Frimmel 2014).

With regards to Au endowment, all of the above examples of auriferous conglomerates outside of the Kaapvaal Craton are dwarfed by the huge amount of gold in the Witwatersrand Basin. This uniqueness of the Witwatersrand is best explained by the exceptional preservation of such an ancient sediment succession (Frimmel 2014). On the one hand, its location in the middle of one of the oldest and most buoyant cratons prevented these sedimentary rocks from becoming tectonically recycled. On the other hand, emplacement of voluminous sheets of flood basalt (Ventersdorp Supergroup) immediately after Witwatersrand sediment deposition and then a thick impact melt sheet as consequence of the 2.023 Ga Vredefort impact both served as highly protective shields that prevented erosion of the underlying Witwatersrand sediments. On other Archaean cratons, similar gold concentration mechanisms most likely took place as well, but the resulting deposits are hardly preserved anymore or have been tectonically overprinted, metamorphosed, and the gold therein remobilized.

With regards to the mode of gold occurrence, all of the above gold-bearing conglomerates can be grouped into two types: those that contain proper nuggets for which sources in the form of eroded discrete gold anomalies (e.g., in orogenic-type gold deposits) are most plausible, and those in which gold occurs as very fine-grained micronuggets and for which no evidence of specific point sources exists. The former type comprises the younger cases (<2.8 Ga) such as those in the Fortescue Group and the Moeda Formation, the latter type the older cases (2.9–2.8 Ga) as in the Witwatersrand Supergroup and those in India. An intermediate position in this regard is taken by the reefs in the upper Central Rand Group, the Ventersdorp Contact Reef and the Black Reef, for which the principal gold sources were older conglomeratic reefs after they had been tilted and exposed on palaeosurfaces. Of greater interest here are, however, those deposits for which no specific point sources can be identified and whose formation can be explained only by regional leaching of background concentrations of Au in the hinterland. These deposits cluster in age at around 2.9 Ga but are seemingly missing in >2.9 Ga rocks, which might reflect different leaching rates prior and after 2.9 Ga. To test the working hypothesis of a climatic control on this leaching rate, possible proxies for palaeoweathering in the Mesoarchaean will be evaluated in the following.

The Witwatersrand Supergroup is the by far best investigated and documented Mesoarchaean siliciclastic succession, thanks to its enormous economic significance and a wealth of exploration and mining-related drill core, underground observations, and analytical data. It shall serve, therefore, as reference here for the time period from 2.96 to 2.78 Ga.

One key feature of potential palaeoclimatic relevance is the stratigraphic, and thus temporal, distribution of climate-sensitive sediment deposits, such as glaciogenic diamictite. Within the West Rand Group, at the base of the Government Subgroup, a laterally extensive diamictite unit occurs as Witfontein Member of the Promise Formation (Fig. 3). Three diamictite beds are distinguished within this member, separated by magnetic mudstone beds, overlying quartz wacke of the Orange Grove Formation (Fig. 7A, Smith et al. 2013). The mudstone comprises Fe-rich chlorite with subordinate magnetite, siderite, and ankerite. The uppermost diamictite of the Witfontein Member is overlain with a sharp transgressive contact by iron formation, which is a mixed magnetite–carbonate facies laminated ferhythmite. The latter grades upwards into a mudstone consisting of Fe-rich chlorite, quartz, magnetite, ankerite, and siderite. For a detailed description of the iron formation units and the various ferruginous mudstones, see Smith et al. (2013).

The next higher lithostratigraphic unit, the Coronation Formation, also contains iron formation with sharp transgressive contact to an underlying diamictite, the Kensington Member (Fig. 7B). The iron formation is a laminated ferhythmite consisting of alternating laminae of magnetite-chert and actinolite with minor calcite, and it grades upwards into a mudstone of Fe-rich chlorite, quartz, and magnetite (Smith et al. 2013).

A deep-water, below-wave base, Fe-rich depositional environment has been deduced by Smith et al. (2013) based on the finely laminated to microbanded nature of the iron formation and their upwards grading into fine-grained magnetic mudstone. The associated diamictite beds rest on low-angle erosional unconformities and mature inner shelf marine quartzite. Deposition in environments that lacked steep slopes can be inferred from this. Further, they have a laterally extensive, sheet-like geometry, all of which speaks against a debris flow origin. Stratigraphic equivalents of the diamictite units in the Mozaan Group (Pongola Supergroup) further east in the Kaapvaal Craton have been described in greater detail. There the presence of dropstones and faceted or at least highly angular polymictic and partly exotic clasts of granite, gabbro, gneiss, schist, chert, iron formation as well as volcanic and sedimentary rocks has been used to argue for a glacial origin (Von Brunn and Gold 1993; Young et al. 1998).

A third diamictite unit within the West Rand Group, the so-called Blue Grit Member in the Afrikander Formation (Government Subgroup), is present in the eastern part of the Witwatersrand Basin (Fig. 3). It might have been also overlain by iron formation but a major hiatus between the Government and overlying Jeppestown subgroups indicates deep erosion, which is likely to have removed such iron formation.

The repeated occurrence of glacial diamictite (tillite) and associated iron formation strongly points at cold climatic conditions during at least parts of West Rand Group times. In contrast, no glaciogenic diamictite units are known from the Central Rand Group. Diamictite described from the Johannesburg Subgroup is a debris flow deposit in which argillaceous sediment of the exposed Jeppestown Subgroup was reworked into channels that cut into the Main Reef (Martin et al. 1989).

A further key feature of potential palaeoclimate relevance is the temporal distribution of detrital feldspar. Detrital feldspar, predominantly alkali feldspar for which a granitic source is indicated by its perthitic appearance, is common in arenites of the West Rand Group but conspicuously absent in those of the Central Rand Group (Law et al. 1990). This lack of feldspar in the latter has been attributed by those who favoured an epigenetic, hydrothermal model for the Witwatersrand gold to H⁺-metasomatism, that is, to basin-wide leaching of the arenitic sedimentary rocks by acidic, supposedly gold-mineralizing fluids (Barnicoat et al. 1997). Acidic conditions during low-grade metamorphic overprint are indicated by the widespread occurrence of pyrophyllite, which has been equally ascribed to gold-mineralizing fluids of metamorphic origin (e.g., Phillips and Powell 2015). Such epigenetic models contradict, however, the geological evidence that undoubtedly supports a primarily placer gold concentration (for a review of arguments, see Frimmel and Nwaila 2020). The abundance of pyrophyllite in the quartz arenites and the matrix of the conglomerates of the Central Rand Group can be explained alternatively by intense chemical weathering and the formation of kaolinite in the sedimentary environment and its subsequent dehydration to pyrophyllite during low-grade metamorphic overprint (Frimmel 1994). A similar reasoning can be applied to other low-pH minerals, such as chloritoid. The importance of kaolinite as precursor to the metamorphic pyrophyllite becomes particularly evident from geochemical data that suggest deep palaeoweathering profiles beneath erosional unconformities (detailed below).

Since the pioneering work by Nesbitt and Young (1982), the chemical index of alteration (CIA) has become a standard tool in the assessment of chemical weathering intensity in the geological past, whereby CIA = 100 × Al₂O₃/(Al₂O₃ + CaO* + Na₂O + K₂O) and CaO* refers to CaO in silicates only. The choice of element oxides considered makes it possible to describe the extent of weathering of feldspars to clay minerals. Although several other indices of weathering and modifications to the CIA have been proposed to account also for the weathering of mafic minerals and to incorporate source rock variability as in the index of compositional variability (Cox et al. 1995), the CIA withstood the test of time as useful proxy for the degree of chemical weathering of felsic rocks. As shown by Wang et al. (2020), it remains an excellent monitor of climate change during glacial to interglacial transitions throughout Earth’s history.

Considering the detrital mineralogy of the arenitic sediment fraction in the Witwatersrand strata, the CIA is well suited to assess the extent of feldspar destruction. A correction for CaO in nonsilicates (essentially carbonates and apatite) is unnecessary because of negligible amounts of such nonsilicate minerals in the siliciclastic metasedimentary rocks of the Witwatersrand and overall low CaO concentrations (typically <0.1 wt%). The amount of Na₂O is also small (ca. 0.2 wt%) and mainly bound to paragonite intergrowths in white mica (Frimmel 1994), except for albite-bearing metapelite in the upper West Rand Group. Thus, CIA values of Witwatersrand siliciclastic rocks effectively reflect the proportions of K-feldspar, muscovite, and pyrophyllite, that is, the extent of leaching of K during chemical weathering and (or) postdepositional reaction with a low-pH fluid.

Several studies have been conducted on the chemistry of Witwatersrand metasedimentary rocks, specifically on arenites and the arenitic matrix of conglomerates (Sutton et al. 1990; Frimmel and Minter 2002; Frimmel 2005) as well as shales (Fuller et al. 1981; Wronkiewicz and Condie 1987; Nwaila et al. 2017). These studies revealed a very wide range in CIA from ca. 50—corresponding to unweathered rock—to as much as 95, which indicates extreme K-leaching. This huge range is, however, by no means arbitrary, but the CIA is controlled by stratigraphic position. Sutton et al. (1990) had concluded already that the CIA systematically increases not only from older to younger units across the Witwatersrand Supergroup but also within individual formations. This finding is of great importance for two reasons: (i) it implies a lower rate of chemical weathering at West Rand Group times, with CIA values typically between 50 and 60 and rising to 85 near the boundary to the Central Rand Group. In the latter, CIA values are typically between 75 and 95, only to decrease again to values around 70 in the uppermost formation (Frimmel 2005). The, on average, much higher CIA values for Central Rand Group arenites are, of course, reflected by the absence of detrital feldspars in Central Rand Group rocks as opposed to those in the West Rand Group; (ii) the increase in CIA within stratigraphic units was confirmed in subsequent studies (Frimmel and Minter 2002; Frimmel 2005), who noted a systematic increase in CIA from values around 80 to values of 90 and higher in the arenitic footwall of gold-bearing conglomerates in the Upper Central Rand Group (Fig. 8). This increase in CIA over distances of a few tens of metres towards erosional unconformities is best explained by deep palaeoweathering profiles beneath old land surfaces. The same studies revealed, however, also a second kind of alteration that affected the CIA, but over much shorter distances, that is, only a few decimetres to up to 1 m above and below a given conglomeratic reef. This second kind of alteration is clearly post-depositional in age because it affected both the footwall and the hangingwall of a given conglomeratic reef. It is also mirrored by a drastic increase in Fe/Al, reflecting chloritisation of muscovite and (or) pyrophyllite by fluids that percolated along the conglomerate beds. It is most intensely developed along the Ventersdorp Contact Reef, which was a focus of bedding-parallel fluid flow because of impermeable thick flood basalt overlying it (Gartz and Frimmel 1999). To interpret CIA values obtained on arenitic rocks in the Witwatersrand Supergroup (and related rocks elsewhere) correctly, it is, therefore, paramount to bear in mind the scale (stratigraphic distance) of alteration and textural observations on the replacement of previous alteration minerals. Changes in CIA are observed on three different scales: (i) over several hundreds of metres, (ii) over metres to tens of metres, and (iii) over decimetres. The first, as expressed by the difference in CIA between West Rand and Central Rand groups, is best explained by changes in atmospheric conditions, e.g., an increase in temperature, and (or) humidity, and (or) acidity. The second is restricted to arenites below erosional unconformities, with a maximum CIA in the immediate footwall and a decrease in CIA towards greater depths below that contact. This type of trend in CIA, as exemplified by Fig. 8A, is best explained by palaeoweathering. The third type, affecting both footwall and hangingwall, thus describing an alteration halo above and below a given conglomerate bed, clearly represents a superimposed postdepositional hydrothermal fluid-induced dispersive alteration as exemplified by the Ventersdorp Contact Reef (Fig. 8G).

The different types of alteration can be also visualized using the Al₂O₃–(CaO* + Na₂O)–K₂O space (A–CN–K diagram in Fig. 9A) as originally devised by Nesbitt and Young (1989). Starting from point X, the average composition of Archaean continental crust, chemical weathering will first lead to a trend from X upwards along a line parallel to the CN–A join, resulting from destruction of plagioclase, until it reaches the A–K join from where it will continue towards the A apex, now reflecting the progressive destruction of first K-feldspar and then muscovite (line 1 in Fig. 9B). For a granitic source composition, this trend will be offset to the right, that is, starting with a rock with a higher K/CN, reflecting a higher K-feldspar/plagioclase ratio. In contrast, Ca–Na-metasomatism will lead to a trend towards the CN apex (e.g., line 2 in Fig. 9B), K-metasomatism to a trend towards the K apex (e.g., line 3 in Fig. 9B). A data set of 226 samples from arenites and the arenitic matrix of conglomerates from various stratigraphic positions within the Witwatersrand Supergroup (Fig. 9B, Frimmel 2005) illustrates that most arenites describe chemical weathering trends of granitic to tonalitic source rocks whereas the arenitic matrix of conglomerates has, in most cases, a composition that is incompatible with any typical weathering trend. Most of the latter are strongly depleted in K and correspondingly enriched in Ca + Na. As the majority of these samples are from the Ventersdorp Contact Reef (Frimmel 2005), this type of alteration is easily explained by interaction of a post-depositional fluid with the overlying metabasalt of the Klipriviersberg Group (Ventersdorp Supergroup). Consequently, these data have little bearing on past chemical weathering rates.

Although the CIA was designed primarily for the arenitic sediment fraction, its application to shale units can also provide useful information (e.g., Fedo et al. 1996). The CIA should increase with increased chemical weathering in the source area and (or) the site of sediment deposition as well as with increased sediment recycling. Plotting available data on shale geochemistry (Nwaila et al. 2017) onto an A–CN–K diagram (Fig. 10A) reveals the full range of different degrees of chemical weathering of felsic to intermediate source rocks. Their CIA values cover a wide spectrum from 50 to 96. Although there is little stratigraphic control within formations and the CIA for a given shale unit varies widely between different goldfields (Nwaila et al. 2017), an overall trend can be observed when comparing data for the West Rand Group with those for the Central Rand Group (Fig. 10A): whereas the West Rand Group shales cover the full range of CIA, including very low values, those in the Central Rand Group are largely limited to high CIA values.

For comparison, the CIA obtained for the matrix of the glaciogenic diamictite in the lower Mozaan Group (Pongola Supergroup) ranges from 58 to 76 with an average of 68, associated Fe-rich shale has an average CIA of 65 (Young et al. 1998). The low CIA values have been used by these authors, together with the relatively high Fe content of the shales mainly in the form of fine-grained magnetite, as arguments in favour of a glacial origin or at least glacial influence. The low CIA values are in stark contrast to data from shales in the upper Mozaan Group (McLennan et al. 1983; Wronkiewicz and Condie 1987). They have distinctly higher CIA, which is also reflected by them plotting higher up in the A–CN–K diagram (Fig. 10B).

Beyond the Kaapvaal Craton, geochemical data exist for variably metamorphosed Archaean siliciclastic rocks in the adjacent Limpopo Belt (Eriksson et al. 1988) and the Zimbabwe Craton (Fedo et al. 1996). There, intense weathering at around 3.0 Ga has been suggested based on quartzite chemistry. Similar data are available also for the Dharwar and the Singbhum cratons in India. The latter are of particular interest here because they contain conglomerates with Witwatersrand-type gold mineralization, though orders of magnitude lower Au endowment, and they are roughly of similar age (3.0–2.8 Ga, Frimmel et al. 2022). For the arenitic matrix of conglomerates in the Bababudan Group, calculated CIA values range from 71 to 99, on average 80, and are somewhat higher than those for the overlying quartzite, which are between 67 and 80, on average 74 (Frimmel et al. 2019). For the Mesoarchaean conglomerates in the Singhbhum Craton, data are available for the basal conglomerate in the upper Iron Ore Group in the Badampahar Greenstone Belt and in the Phuljhari Formation (Frimmel et al. 2022). Some of these have a relatively low matrix proportion, and complete physical separation of quartz-rich clasts could not be achieved, which resulted in high SiO₂ contents of >96 wt.%. These are not further considered for CIA calculations. The remaining analyses gave CIA values between 74 and 96, on average 82. Unfortunately, the geochronological control on these deposits is not good enough to establish a firm correlation with either the West Rand or the Central Rand groups in the Kaapvaal Craton, but it is noted that all of them are marked by relatively high CIA, indicative of an above-average extent of chemical weathering. Evidence of elevated levels of chemical weathering in the Archaean hinterland of the Singhbhum Craton has been documented also in arenitic siliciclastic rocks in the Jamda-Koira Basin (western Iron Ore Group), the base of the Mahagiri Quartzite in the Upper Iron Ore Group in the south of the Singhbhum Craton and its correlative, the Keonjhar Quartzite (Mukhopadhyay et al. 2014; Ghosh et al. 2016; Kumar et al. 2017). For these deposits, an age between ca. 2.91 and 2.8 Ga is suggested (Sreenivas et al. 2019; Frimmel et al. 2022), which would roughly overlap with that of the Central Rand Group. Interestingly, the CIA values obtained for these Indian examples are all similar to those for the Central Rand Group arenites.

The Mesoarchaean Era was without doubt the most important time for the concentration of gold in Earth’s crust with a punctuated temporal peak at around 2.9 Ga. Microbial gold fixation at that time seems to have played the pivotal role in this regard, and microbial mats probably acted as principal source of fine-grained gold in more or less coeval placer deposits. Repeated mechanical reworking of the 2.9 Ga placers led to further placers in younger fluvial to fluvio-deltaic deposits but with decreasing Au grade and endowment (see Fig. 1). Although there are some gold placers that are older than 2.9 Ga (2.96 Dominion Reef, reefs in upper West Rand Group), the total amount of gold bound therein is negligible in a global context, which raises the question as to the likely reasons for this abrupt change in crustal gold concentration at around 2.9 Ga. Climate change has been presented here as one possible hypothesis for this. Unfortunately, direct evidence of climatic conditions over the critical period in the Mesoarchaean is not available. We, therefore, can only speculate based upon indirect evidence as reviewed in this paper. All of this evidence, be it from the lithological and petrological makeup or from geochemical data, specifically CIA, speaks for a broad subdivision of Mesoarchaean strata into those deposited under relatively cool, occasionally even glacial, conditions (represented by much of the West Rand Group in the Witwatersrand Basin and the lower Mozaan Group in the Pongola Basin) and those evidently deposited under more humid conditions that enabled deep chemical weathering of the old land surface (represented by the Central Rand Group and the upper Mozaan Group).

As discussed by Young et al. (1998), a glacial origin of the diamictites in the lower Mozaan Group is far more likely than their formation as debris flows. The main arguments for such a genetic interpretation have been the following: (i) presence of dropstones in a laminated to finely bedded muddy Fe-rich matrix, (ii) deflected bedding around dropstones typical of vertical emplacement of the dropstones into unconsolidated mud, (iii) presence of striated and facetted clasts, (iv) the polymictic clast population of largely extrabasinal clasts reflecting the lithology of the pre-Pongola basement, (v) a sharp lower contact of the diamictite beds, and (vi) the Fe-rich composition and low CIA of the matrix. The same arguments can be extended to the diamictite units in the West Rand Group, where the spatial association with iron formation adds further evidence of cold conditions.

Such a link between iron formation as well as Fe-rich argillaceous diamictite matrix and glacial events is well established for younger glaciations, such as those in the Palaeoproterozoic and Cryogenian (e.g., Cox et al. 2016; Hoffman et al. 2017) and has been suggested, by analogy, also for the Mesoarchaean glaciogenic deposits (Young et al. 1998). The latter authors followed the reasoning that iron formation formed at the end of a glacial event by upwelling of bottom waters rich in hydrothermally derived Fe²⁺ and its mixing with subglacial meltwater—a hypothesis also suggested to explain Cryogenian Snowball Earth features (e.g., Lechte et al. 2019). This notion fails, however, to explain why iron formation is, in places, overlain by, or interbedded with, glaciogenic diamictite, as is the case in the West Rand Group and lower Mozaan Group. Recently, a possible solution to this enigma has been presented by Mitchell et al. (2021), who made a case for orbital forcing, reflected by Milankovitch Cycles, as a plausible explanation of the sedimentological cyclicity that marks banded iron formation in the Cyrogenian Period. This might as well be applicable to the Mesoarchaean iron formation of interest here as they also display rhythmic banding that reflects a similar cyclicity (Smith et al. 2013).

While a reasonably strong case can be made for climate change at around 2.9 Ga as shown here for sedimentary deposits of that age in the Kaapvaal Craton, our understanding of climatic conditions at even older times is very poor. The oldest well-preserved siliciclastic sedimentary deposits reflecting fluvial to fluvio-deltaic environments are from the 3.22–3.21 Ga Moodies Group of the Barberton Greenstone Belt in the Kaapvaal Craton (Heubeck 2019). If microbial gold fixation was key to the first large-scale concentration of gold not only in the sedimentary environment but in the Earth’s crust in general, one would expect considerable gold contents in the Moodies Group because it even contains microbial remnants (Homann et al. 2018). Yet, no evidence of elevated gold concentrations in these ancient sedimentary rocks is known.

One possible answer to this apparent discrepancy could be that the shorelines at that time had not been widely colonized yet and that life there developed only at special places of hydrothermal fluid expulsion related to a coeval magmatic event as alluded to by Heubeck (2019), and (or) the biology of the microbes was not suitable for gold fixation. The potential source area could simply have been too small at that time when the overall ratio of land/ocean area was still very low. In addition, climate could have played a critical role. Simpson et al. (2012) suggested that the high maturity of sandstone in the Moodies Group is due to aeolian mechanical abrasion rather than intense chemical weathering, thus at least not precluding the possibility of relatively cold climatic conditions not conducive to chemical weathering at Moodies Group times. A third possible explanation for the paucity of gold in crustal rocks prior to 2.9 Ga could be a lack of plate tectonics at that time. Subduction is well-known as critical process in the enrichment of the supra-subduction subcontinental lithospheric mantle wedge in a number of metals, including Au (e.g., Wilkinson 2013). A lack of porphyry Au (and Cu-Au), associated epithermal deposits, intrusion-related deposits as well as orogenic-type Au deposits older than 2.75 Ga is a strong indication of modern style subduction not having been operating yet at that time, and it may be speculated that in the absence of such deposits, rich gold placers could not have formed from the erosion of endogenous gold deposits (Frimmel 2018). Thus, the Mesoarchaean gold placers differ fundamentally from their younger equivalents, which are typically derived from the erosion of discrete gold deposits in the hinterland, a prime example of which would be the 2.65 Ga Moeda Formation.

As mentioned at the beginning, the currently only plausible source for the huge amounts of gold that accumulated in conglomerates at around 2.9 Ga is the leaching of background concentrations of Au from the Archaean land surface—a process that is thermodynamically highly feasible under the atmospheric and hydrospheric conditions envisaged for that time (Frimmel 2014; Frimmel and Hennigh 2015; Heinrich 2015). Such leaching would have been driven by chemical weathering under an acidic atmosphere with rain of a pH of approximately 4. The observed punctuated temporal distribution of Mesoarchaean gold placers would then signal a drastic change in the chemical weathering rate, be it due to climate change or changing atmospheric and hydrospheric acidity. The latter is dictated, in the first instance, by volcanic emission of CO₂ and sulphuric gases into the atmosphere. A shift from submarine towards more subaerial volcanic degassing with increasing landmasses emerging in the course of the Archaean and an associated change in the H₂S/SO₂ ratio (Huston and Logan 2004) might well have influenced the atmosphere’s pH and even its oxygenation (Gaillard et al. 2011). This shift took place, however, only in the Neoarchaean—too late to explain the 2.9 Ga gold mega-event—and was most likely not abrupt but gradual. This leaves climate change at around 2.9 Ga as plausible candidate.

Even if a change from cold to more humid, temperate climate from 2.9 Ga onwards facilitated chemical weathering and thus the amount of Au dissolved in river waters draining the Mesoarchaean land, major gold deposits would not have formed without an effective trap. The geological evidence speaks for microbial colonies on riverine wetlands and shorelines to have provided this trap, but the nature of these microbes remains elusive. The trapping reaction could have been reduction as suggested by Heinrich (2015) or oxidation as suggested by Frimmel (2014). The latter could have taken the form of O₂ release by early photosynthesizing microbes, such as cyanobacteria, or by oxidation in the sense of electron donation. Acidophile microbes could have played the decisive role in this regard as suggested by first Cu isotope data on Witwatersrand gold and associated carbonaceous material: gold in carbon seams has δ⁶⁵Cu = +0.02 ± 0.19‰ and thus distinctly higher δ⁶⁵Cu than pyrite in the main reefs of the Central Rand Group (δ⁶⁵Cu= −0.48 ± 0.40‰) and even more so than pyrite in the younger Ventersdorp Contact Reef on top of the Witwatersrand Supergroup, whose δ⁶⁵Cu is −1.34 ± 0.08 ‰ (R. Mathur (unpublished data, 2022)). The comparatively higher δ⁶⁵Cu of the microbially fixed gold is best explained by oxidation. Irrespective of the biology of these ancient microbes, they could have trapped large amounts of gold only if they existed in large colonies and if meteoric waters had been enriched in Au by several orders of magnitude. This could only be possible in a relatively warm and wet climate, and the conclusion can be drawn that such a climate existed for most of the time between 2.9 and 2.8 Ga, at least on the Kaapvaal Craton, possibly also the Singhbhum Craton. Thus the genesis of Witwatersrand-type gold deposits can be linked directly to the fortuitous interplay of atmospheric and biologic evolution coupled to a change from cold to warmer and (or) more humid climate at around 2.9 Ga that promoted microbial growth and deep chemical weathering of palaeosurfaces. This conclusion has implications on future exploration for Witwatersrand-type gold deposits. Rather than searching for a specific hinterland, it seems more advantageous to look for fluvial to fluvio-deltaic conglomerates whose age falls into the “golden window” of 2.9–2.8 Ga and which rest on palaeosurfaces exhibiting deep chemical weathering profiles underneath.

This paper is dedicated to the late Grant Young, who was not only an inspiring teacher on how to recognize glacial deposits in the ancient rock record but also an unforgettable companion in the field. His pioneering work on diamictites from the Pongola Supergroup laid, unknowingly at the time of his studies, the foundation for our current understanding of the metallogenesis of one of the world’s most enigmatic ore provinces, the Witwatersrand goldfields. Helpful comments by an anonymous reviewer and R. Goldfarb on the original manuscript are greatly appreciated.

All data used for this study were taken from published sources as cited in the text.

Conceptualization: HF

Data curation: HF

Formal analysis: HF

Investigation: HF

Methodology: HF

Project administration: HF

Resources: HF

Validation: HF

Visualization: HF

Writing – original draft: HF

Writing – review & editing: HF