Introduction
This study examines a section of South Africa’s Bushveld Complex that hosts the principal PGE orebodies, the UG2 chromitite and Merensky Reef, as well as the low-grade UG3 and UG3a–b chromitites and the uneconomic Bastard Reef. These PGE-bearing chromitites and pyroxenites are important stratigraphic markers in the Rustenburg Layered Suite, especially in the well-developed Eastern Limb (Fig. 1). In the study area, the UG2–Merensky interval is unusually thick, reaching ∼400 m (South African Committee for Stratigraphy, Reference Kent1980). Chromitites and pyroxenites typify the Critical Zone (Fig. 2), whose subdivisions display pronounced metre-scale layering (Hall, Reference Hall1932). In the Lower Critical Zone (LCZ), they are interlayered with subordinate peridotite (Cameron, Reference Cameron1980; Scoon et al., Reference Scoon, Bristow, Reichhardt and Mitchell2024); in the Upper Critical Zone (UCZ), they occur within a noritic succession containing laterally persistent anorthosite marker layers (Vermaak, Reference Vermaak1976; Cameron, Reference Cameron1982; Gain, Reference Gain1985). The MG3 anorthosite marks the LCZ–UCZ boundary, while the top of the Giant Mottled anorthosite defines the UCZ–Lower Main Zone (LMZ) boundary.
A regional map illustrates the three main limbs of the Rustenburg Layered Suite, detailing sectors, a schematic outcrop of the Merensky Reef, and the position of Winnaarshoek. In the Northern Limb, the Platreef serves as the regional equivalent of the Merensky Reef. Additional sites identified within the Eastern Limb are: (1) Maandagshoek; (2) Hackney; (3) Middelpunt and Zeekoegat; (4) Doornvlei; (5) Dwars River; and (6) Helena and Richmond.
Figure 1 Long description
The Rustenburg Layered Suite in South Africa, highlighting different sectors and limbs. The legend identifies sectors: Western, Northern, Southern, Zwartklip and Rustenburg, along with the Northern and Western Limbs. Key locations include Rustenburg, Marikana, Pretoria, Mokopane, Winnaarshoek and Steelpoort. The Merensky Reef is marked in red near Rustenburg and Steelpoort. The Platreef is noted near Mokopane. The map includes a scale bar and compass for orientation, with coordinates ranging from 24 degrees S to 25 degrees S latitude and 27 degrees E to 30 degrees E longitude. Dip direction is indicated by arrows.
A generalised stratigraphic column for the northern sector of the Eastern Limb displays the three main components of the Bushveld Complex. The older Rooiberg Group and the younger Lebowa Granite Suite are found mainly in the roof structure. The Rustenburg Layered Suite was intruded after formation of the felsites into the upper levels of the Transvaal Supergroup, which makes up the floor rocks. The irregularity of the contact with the floor rocks is schematic. The divisions into zones, subzones, and the placement of marker layers are simplified. The section studied is restricted to part of the UCZ containing the UG2 and UG3 chromitites, as well as the Merensky Reef, and extends up into the LMZ. The Bastard Reef (not shown) is omnipresent in the stratigraphy above the Merensky Reef.
Figure 2 Long description
A stratigraphic column of the Bushveld Complex, divided into zones, marker layers and lithology. On the left, the column is bordered by the Rooiberg Group and the Transvaal Supergroup, labeled as the roof and floor, respectively. The Rustenburg Layered Suite is central, divided into zones: Upper Zone, Upper Main Zone, Lower Main Zone, Upper Critical Zone, Lower Critical Zone and Lower Zone. Each zone is numbered from one to six, with a study section highlighted in the Upper Critical Zone. Marker layers are listed in the middle, including Upper Magnetite Layers, Pyroxenite Marker and others. On the right, the lithology column details rock types like diorite, gabbro and pyroxenite. The diagram includes a scale in kilometers at the top and the contact with the floor rocks is shown as irregular.
This study extends the lithostratigraphic and geochemical section of Mitchell and Scoon (Reference Mitchell and Scoon2007) and Scoon et al. (Reference Scoon, Costin and Mitchell2025a) for Winnaarshoek, Marula platinum mine. Here we present whole-rock Sr and Nd isotope data in support of the earlier electron microprobe and whole-rock analyses. The studied section lies near the centre of the best-characterised of the three sectors in the Eastern Limb, where igneous layering is unlikely to reflect floor-rock irregularities. The UG2–UG3 stratigraphy changes little along 70 km of strike, whereas the Merensky–Bastard Reef interval is markedly more variable, as Wagner (Reference Wagner1929) observed. Scoon et al. (Reference Scoon, Costin and Mitchell2026) compared the Winnaarshoek succession with the Merensky and Bastard Reefs at other localities, including Maandagshoek and Zeekoegat.
We assess regional Sr-isotope variation across the Merensky–Bastard Reef interval. This comparison is significant because the Sr-isotopic inflexion at the Merensky Reef, first identified in the Western Limb by Kruger and Marsh (Reference Kruger and Marsh1982) and later reassessed at the Atok or Bokoni mine by Lee and Butcher (Reference Lee and Butcher1990), has no counterpart at the UG2 in either the Western Limb (Kruger, Reference Kruger1994; Schoenberg et al., Reference Schoenberg, Kruger, Nägler, Meisel and Kramers1999) or Eastern Limb (Scoon et al., Reference Scoon, Costin and Mitchell2025a). We then assess whether the isotope data support repeated intrusion of thin magma sheets into crystalline strata as the origin of the igneous layering, rather than the widely accepted upper-level magma-chamber model.
Overview
The Bushveld Complex consists of three main groups of igneous rocks: Rooiberg Group; Rustenburg Layered Suite; and Lebowa Granite Suite (Hall, Reference Hall1932; Wager and Brown, Reference Wager and Brown1968; Willemse, Reference Willemse1969; Tankard et al., Reference Tankard, Jackson, Eriksson, Hobday, Hunter and Minter1982; Von Gruenewaldt et al., Reference Von Gruenewaldt, Sharpe and Hatton1985; Eales, Reference Eales2014; Cawthorn, Reference Cawthorn, Charlier, Namur, Latypov and Tegner2015). The felsites and granophyres, together with the layered mafic-ultramafic rocks and the granites, constitute a chronological sequence attributable to a single thermal event with an age of ∼2.055 Ga (Walraven et al., Reference Walraven, Armstrong and Kruger1990; Hatton and Schweitzer, Reference Hatton and Schweitzer1995; Zeh et al., Reference Zeh, Ovtcharova, Wilson and Schaltegger2015; Mungall et al., Reference Mungall, Kamo and McQuade2016).
The Rustenburg Layered Suite intruded the Transvaal Supergroup beneath the Rooiberg Group felsites and granophyres, which form part of a roof structure that also includes younger granite and granophyre sheets. Its mafic–ultramafic rocks occur mainly in three tabular limbs, up to 10 km in thickness, that dip towards the centre of the Transvaal Basin (Fig. 1). The Eastern Limb is particularly well known because of its extensive outcrops and was the basis for Hall’s (Reference Hall1932) formal definition of the stratigraphic zones. Correlation of these zones and marker layers across the suite is essential for interpreting its igneous layering (Fig. 2). However, chromitites and pyroxenitic reefs vary markedly between limbs (Lee, Reference Lee and Cawthorn1996; Viljoen, Reference Viljoen2016) and among sectors, the latter defined by irregularities in the floor rocks.
The UG2 chromitite is the principal component of the Upper Group of chromitites which, together with the Lower and Middle Groups, forms the threefold subdivision of Cousins and Feringa (Reference Cousins, Feringa and Haughton1964) (Fig. 2). The chromitite occurs between metre-thick layers of feldspathic orthopyroxenite (‘pyroxenite’) and minor feldspathic harzburgite (‘harzburgite’), both locally pegmatitic, and in places rests on a noritic–anorthositic footwall. It is the only chromitite mined as a primary PGE source and is exploited over hundreds of kilometres of strike in the Eastern and Western Limbs (Vermaak, Reference Vermaak1985; Schürmann et al., Reference Schürmann, Grabe, Steenkamp, Wilson and Anhaeusser1998; Viljoen and Schürmann, Reference Viljoen, Schürmann, Wilson and Anhaeusser1998). The orebody has relatively uniform PGE grades, is typically 0.6–1.4 m in thickness and is consistently overlain by thin Leader chromitites (Eales, Reference Eales and Stowe1987; Von Gruenewaldt et al., Reference Von Gruenewaldt, Dicks, De Wet and Horsch1990; Lea, Reference Lea1996; Schoenberg et al., Reference Schoenberg, Kruger, Nägler, Meisel and Kramers1999; Maier and Barnes, Reference Maier and Barnes2008; Viljoen, Reference Viljoen2016; Scoon et al., Reference Scoon, Costin and Mitchell2025a). Plagioclase is generally the dominant silicate mineral in the main UG2, although orthopyroxene occurs with plagioclase in the Leader seams. The UG2 is stratiform because PGE are confined to the chromitite layers. At Winnaarshoek, mineralisation is limited to the main chromitite and a single centimetre-thick Leader seam; the footwall and hangingwall silicates, together with silicate partings and xenoliths, are invariably unmineralised (Fig. 3).
The UG2 is a stratiform ore deposit at Winnaarshoek in which PGE are hosted mainly by a 0.7 m-thick chromitite layer and the overlying centimetre-thick Leader chromitite, exposed here in a trial stope. These laterally persistent, parallel chromitites can be traced across Marula mine. The main chromitite overlies the regional UG2 Footwall anorthosite marker and the younger UG2 Footwall pegmatite, which thins leftward in the photograph and may locally be absent. The Leader chromitite occurs within the UG2 Hangingwall pyroxenite. The vertical PGE distribution is based on a nearby drill hole.
The UG3 chromitite is weakly mineralised and, at 0.2–0.25 m thick, is too thin to constitute an orebody (Gain, Reference Gain1985; Mossom, Reference Mossom, Anhaeusser and Maske1986; Scoon et al., Reference Scoon, Costin and Mitchell2025a). The UG3 lies between an anorthositic footwall and a pyroxenitic hangingwall. At Winnaarshoek, the UG3a–b seams are 0.18 m and 0.13 m thick, respectively, and are distinguished by their association with harzburgite. Restricted to the northern sector (Eastern Limb), the UG3 chromitites have been interpreted as lateral equivalents of the UG2 Leader seams. The UG3 silicate mineralogy broadly resembles that of the UG2, with plagioclase dominant; however, in the UG3a–b seams, orthopyroxene and olivine are present and plagioclase is distinctly subordinate.
The Merensky Reef consists mainly of metre-thick layers of pyroxenite and harzburgite, together with thin chromitite layers or chromite stringers (Wagner, Reference Wagner1929; Cousins, Reference Cousins1969; Vermaak, Reference Vermaak1976; Viljoen, Reference Viljoen1999). The pyroxenites and harzburgites can be exceptionally coarse grained and pegmatitic. Cryptic PGE layering makes the Merensky Reef a stratabound orebody that is more difficult to mine than the UG2 because mineralisation extends across several rock types, including footwall anorthosite, although peak grades consistently occur in the chromitites. Wagner (Reference Wagner1929) recognised about a dozen Merensky Reef facies. At Winnaarshoek, the reef belongs to the Maandagshoek facies and comprises between two and four pyroxenite layers with two chromite stringers (Fig. 4). PGE are concentrated in lower and upper reefs separated by a barren middling (Mitchell and Scoon, Reference Mitchell and Scoon2007). In Wagner’s Zeekoegat facies (Atok mine), the reef includes harzburgite as well as pyroxenite, together with two or more chromitites (Mossom, Reference Mossom, Anhaeusser and Maske1986; Lee and Butcher, Reference Lee and Butcher1990). The Dwars River and Helena facies consist of pyroxenites enclosed by anorthosites tens of metres thick (Wagner, Reference Wagner1929). The Doornvlei facies is unusual in comprising four pyroxenite layers, together with chromitites, which are separated from one another by norites and anorthosites (Von Gruenewaldt et al., Reference Von Gruenewaldt, Dicks, De Wet and Horsch1990; Lea, Reference Lea1996). There are further variations in the Western Limb where the Merensky chromitites tend to be thicker and more persistent. In the Rustenburg sector, several facies are recognised, including the thin Kroondal reef (Wagner, Reference Wagner1929; Viljoen and Hieber, Reference Viljoen, Hieber, Anhaeusser and Maske1986) and the A-B-C reefs, which occur at different stratigraphic levels at the Impala mine (Leeb-Du Toit, Reference Leeb–Du Toit, Anhaeusser and Maske1986). In the Zwartklip sector, harzburgite is a major reef component (Wagner, Reference Wagner1929; Scoon and de Klerk, Reference Scoon and De Klerk1987; Viljoen, Reference Viljoen1999; Smith et al., Reference Smith, Basson and Reid2004; Mitchell et al., Reference Mitchell, Henckel and Mason-Apps2019a). This facies is also associated with the Pseudoreef harzburgites, which form additional, poorly mineralised layers between the UG2 and the Merensky Reef.
The Merensky Reef at Winnaarshoek, exposed at the entrance to Decline 21, is a stratabound orebody comprising two chromite stringers and two reef intervals. The upper reef is more continuous, has higher PGE grades, and is capped by the upper chromite, whereas the lower reef may be restricted to the lower chromite. The two stringers are separated vertically by an average of 1.8 m. The weakly mineralised Merensky pegmatite differs from equivalent reef facies in the Western Limb. Four pyroxenites, including a limited exposure of the Merensky Footwall pyroxenite, define separate marker layers. Base-metal sulfides occur in the upper reef and Merensky Hangingwall pyroxenite and have oxidised to form a surface gossan that aided exploration. A thin leuconorite layer separates the pyroxenite from the Merensky anorthosite.
Figure 4 Long description
The entrance to Decline 21 at the Merensky Reef in Winnaarshoek, featuring a stratabound orebody. The central part of the image shows a tunnel entrance surrounded by rock layers. Two chromite stringers, labeled as ′Upper Cr′ and ′Lower Cr′, are marked with dashed red lines, separated by approximately 1.8 meters. On the left, a vertical stratigraphic column lists layers from top to bottom: Merensky anorthosite, leuconorite, Merensky Hangingwall pyroxenite, Merensky pyroxenite and Merensky pegmatite. On the right, a color-coded legend indicates different rock types and their corresponding layers: surface gossan, leuconorite, Merensky Hangingwall pyroxenite, Merensky pyroxenite and Merensky Lower chromite stringer. The diagram provides a visual representation of the geological structure and composition of the Merensky Reef, highlighting the distinct layers and their spatial arrangement.
The Bastard Reef is consistently present above the Merensky Reef in the stratigraphy but, despite its similarities, is only weakly mineralised (Wagner, Reference Wagner1929; Vermaak, Reference Vermaak1976). It consists of pyroxenite, which may be pegmatitic, and a basal chromitite. At Atok mine, the Bastard Reef comprises a single pyroxenite layer, as is typical of the Western Limb, whereas two or more pyroxenite layers occur at Maandagshoek and Winnaarshoek.
The potholing of the chromitites and the pyroxenitic reefs, in which saucer-shaped, near circular depressions extend several metres into the footwall (Lee, Reference Lee and Cawthorn1996; Viljoen, Reference Viljoen2016), helps determine the order of emplacement of rock layers (Latypov et al., Reference Latypov, Chistyakova and Kramers2017). Potholes can be regarded as part of a continuum that includes dimples and protuberances located at the base of reefs, as discussed by Campbell (Reference Campbell1986) and Eales et al. (Reference Eales, Field, De Klerk and Scoon1988). The UG2 and Merensky Reef potholes have been studied extensively at the platinum mines in the Zwartklip sector (Lomberg et al., Reference Lomberg, Martin, Patterson and Venter1999; Viljoen, Reference Viljoen1999; Viring and Cowell, Reference Viring and Cowell1999; Smith and Basson, Reference Smith and Basson2006). Merensky Reef transgression is especially important because the upper chromitite and hangingwall pyroxenite first cut downward through the pegmatite and then into, and locally displace, the footwall stratigraphy. As the pothole extends further into the footwall, the Merensky pegmatite may be completely displaced, thereby constraining the relative ages of the reef layers and footwall rocks. Merensky Reef potholes at Winnaarshoek are similar in that the two pyroxenites containing the upper chromite cut downward and displace the lower chromite and pegmatite before impacting the footwall anorthosite.
Section at Winnaarshoek
Sampling
Half- and quarter-core samples were collected from three drill holes at Winnaarshoek (Fig. 5). Two holes, WH–1 and WH–14 (deflection 2), were collared in the UCZ and intersected the Merensky Reef; WH–1 continued down to the UG2. LMZ samples were obtained from drill hole WH–27. Mitchell and Scoon (Reference Mitchell and Scoon2007) give a more complete LMZ section. Drill-hole locations are shown on the geological map in Scoon et al. (Reference Scoon, Costin and Mitchell2026). Samples are identified by drill-hole depth and by elevation relative to the Merensky upper chromite and height above the UG2 chromitite (Table 1).
Vertical columns showing lithological units, chromitite and reef positions, and sample locations in the Winnaarshoek study section. (a) Lower part of drill hole WH–1. (b) Upper part of drill hole WH–1 and drill hole WH–27. (c) Drill hole WH–14/Deflection 2. Scales differ between columns. The seven anorthosite markers are: (1) UG2 Footwall anorthosite; (2) UG3 Footwall anorthosite; (3) UG3 Hangingwall anorthosite; (4) Merensky Footwall anorthosites; (5) Merensky anorthosite; (6) Giant Mottled anorthosite; and (7) Lower Mottled anorthosite.
Figure 5 Long description
A vertical drill hole log titled, WH dash 1, with a vertical scale labeled, m and tick labels at 0, minus 100, minus 200, minus 300, minus 370, minus 380, minus 390, minus 400, minus 410 and minus 420. At the top of the log a rounded cap is labeled, Overburden. Below the cap the log is divided into stacked patterned blocks, with several thin horizontal bands and multiple small boxed depth labels aligned along the right side of the log. A tall bracket label to the right reads, Merensky Footwall unit, spanning much of the upper and middle part of the column. Lower down, a bracket label reads, UG3 Unit and below that another bracket label reads, UG2 Unit. Near the lower part of the column, several thin horizontal bands are labeled, Cr, with one Cr label near the transition into the UG3 Unit and additional Cr labels lower down near the UG2 Unit. The image B showing two adjacent vertical drill hole logs titled, WH dash 27 and, WH dash 1, each with a vertical scale labeled, m. The WH dash 27 log shows tick labels at 120, 100, 80, 60, 40, 20 and 0 and contains stacked patterned blocks with several thin horizontal bands and multiple small boxed depth labels along the right side. The WH dash 1 log in this image shows tick labels at 40, 20 and 0 and also contains stacked patterned blocks with thin horizontal bands and multiple small boxed depth labels along the right side. Between and beside the logs are bracket labels reading, Bastard Unit and, Merensky Footwall unit and within the WH dash 1 log are text labels including, Bastard pyroxenites, Merensky anorthosite and, Merensky pyroxenites. Several thin horizontal bands are labeled, Cr, including one near the level of the Merensky labels and additional Cr labels lower down. A slanted label near the lower right reads, Merensky Footwall unit. The image C showing a shorter vertical drill hole log titled, WH dash 14 slash D2, with a vertical scale labeled, m and a tick label at 0. The log contains stacked patterned blocks and thin horizontal bands, with multiple small boxed depth labels along the right side. Two thin horizontal bands near the upper part are labeled, Cr. A bracket label to the right reads, Merensky Reef upper and lower. To the far right of the full figure is a legend box with a heading, Legend and a vertical list of labeled swatches and symbols reading, PGE Reef, Norite, Gabbronorite, Spotted and mottled, Leuconorite, Chaotic layering pyroxenite, norite, anorthosite, Pegmatite, Feldspathic orthopyroxenite and, Chromitite layers or stringer.
List of Samples, Winnaarshoek
Table 1 Long description
Columns list drill hole, depth in meters, height in meters, zone or unit, marker layer, and lithology. Drill hole, WH-27 has six Lower Main zone entries at depths 94.0 to 144.0 m, all with lithology, Norite, and marker layer shown as a hyphen. Drill hole, WH-1 begins in the Bastard unit at depths 25.3 to 42.5 m, including Giant Mottled anorthosite marker layers with mottled anorthosite at 25.3 to 31.3 m, then leuconorite at 33.3 to 36.2 m, and multiple Bastard Reef marker layers with pyroxenite and leuconorite between 37.3 and 42.5 m. WH-1 continues in the Merensky unit from 43.3 to 55.0 m with repeated Merensky anorthosite marker layers and mottled or spotted anorthosite, followed by Merensky hangingwall, main, and footwall pyroxenite marker layers with pyroxenite from 52.9 to 55.0 m. WH-1 then lists the Merensky Footwall unit from 60.5 to 419.1 m, with lithologies including melanorite norite noted as chaotic layering, pyroxenite anorth noted as chaotic layering, mottled or spotted anorthosite at marker layers including Merensky footwall anorthosite and UG3 hangingwall anorthosite, and deeper entries of leuconorite and gabbronorite. WH-1 includes UG3 unit entries from 422.9 to 434.6 m with blank marker layer cells, including harzburgite at 427.0 m, pyroxenite at 431.4 m, and mottled anorthosite at the UG3 footwall anorthosite marker layer at 434.6 m. WH-1 includes UG2 unit entries from 436.4 to 469.3 m with mostly blank marker layer cells, lithologies of pyroxenite and one melanorite at 456.7 m, ending with spotted anorthosite at the UG2 footwall anorthosite marker layer at 469.3 m, then one UG1 unit entry at 472.3 m with pyroxenite. Drill hole, WH-14 D2 lists Merensky unit entries from 55.2 to 59.3 m with marker layers Merensky hangingwall pyroxenite, Merensky pyroxenite, Merensky pegmatite, and Merensky footwall pyroxenite, and lithologies pyroxenite except pegmatite at 57.8 m.
* Height relative to Merensky Upper chromite; hw is hangingwall; fw is footwall; cl denotes chaotic layering
Analytical Techniques
Rock powders for whole-rock and isotopic analyses were prepared at Rhodes University. Major- and trace-element analyses were conducted at the Laboratoire Magmas et Volcans, St-Etienne; analytical procedures are summarised by Scoon et al. (Reference Scoon, Costin and Mitchell2026). For Sr and Nd isotope separation, samples were processed through four columns following the methods of Pin et al. (Reference Pin, Briot, Bassin and Poitrasson1994) and Pin and Zalduegui (Reference Pin and Zalduegui1997). After evaporation and dissolution in 2M HNO3, Sr and rare earth elements were separated by coupled passage through Sr-spec and TRU-spec columns. Neodymium was then isolated using Ln-spec columns. Sr isotopic measurements were performed by TIMS (TRITON) at the Laboratoire Magmas et Volcans, Clermont–Ferrand, and 87Sr/86Sr ratios were normalised to 86Sr/88Sr = 0.1194. Nd isotopic and Sm–Nd determinations were made by isotope dilution using a mixed 149Sm–150Nd tracer and MC–ICP–MS (Neptune Plus coupled to an Aridus II system), also at the Laboratoire Magmas et Volcans, Clermont–Ferrand. Strontium and Nd isotopic ratios were calculated using normalised values of 87Sr/86Sr = 0.710230 ±11 for NBS SRM987 and 143Nd/144Nd = 0.512115 ±16 for JNdi–1. Total procedural blanks were <1 ng for both Sr and Nd.
Lithologies
The Winnaarshoek section contains five principal rock types: norite–leuconorite, gabbronorite, anorthosite, pyroxenite and chromitite. Cameron (Reference Cameron1982) and Gain (Reference Gain1985) provide detailed petrographic descriptions of comparable Eastern Limb sequences. Silicate rocks are classified by the relative abundances of orthopyroxene, plagioclase and augite, together with their characteristic textures, and can be grouped into feldspathic and pyroxenitic components. In the UCZ, feldspathic rocks consist mainly of leuconorite and anorthosite; norite is scarce, and gabbronorite occurs in only one interval. Norite, leuconorite and gabbronorite in the LMZ are distinctly coarser grained than their UCZ counterparts. Anorthosites occur in both the UCZ and LMZ, ranging from spotted to mottled varieties with large, irregular oikocrysts. In some layers, these oikocrysts decrease upward in abundance, producing progressively more feldspathic rocks. Disseminated chromite is rare in UCZ feldspathic rocks and absent from the LMZ.
In the pyroxenites, orthopyroxene is the only primocryst silicate phase, although interstitial and oikocrystic plagioclase may exceed 10 modal% and locally reach 25 modal%. Large emerald-green augite oikocrysts are ubiquitous. Disseminated chromite is an important accessory phase, whereas biotite is minor and quartz–perthite intergrowths occur interstitially. Many pyroxenites contain multiple generations of orthopyroxene and/or plagioclase which have been interpreted as xenocrysts derived at depth (Cameron, Reference Cameron1982; Eales et al., Reference Eales, Maier and Teigler1991). Olivine is rare in the pyroxenites, and harzburgite at Winnaarshoek is confined to thin layers between the UG3a and b chromitites.
Chromitite layers contain more than 50 modal% chromite, whereas silicate-rich chromite stringers, including those in drill hole WH–1 (Fig. 5), are dominated by plagioclase (Scoon and Costin, Reference Scoon and Costin2018). Chromitite layers are generally associated with pyroxenites, whereas chromite stringers occur either within pyroxenite layers or along pyroxenite–anorthosite contacts.
Marker layers
The UG2–UG3 chromitites and pyroxenitic reefs serve as marker horizons in the Eastern Limb because they can be traced laterally for tens of kilometres, from Steelpoort to Zeekoegat (Fig. 1). Anorthosites are also useful markers because of their lateral continuity, uniformity, and distinctive magnetic signatures (Mitchell and Scoon, Reference Mitchell and Scoon2007). Seven anorthosite markers are recognised in the study section (Fig. 5).
Lithological units
Following Cameron (Reference Cameron1982), we use anorthosite markers to define the bases of the UG2, UG3 and Merensky Footwall units. In contrast, the Merensky and Bastard units follow the conventional definitions of the South African Committee for Stratigraphy (Reference Kent1980): basal pyroxenites and chromitites overlain by anorthosites. Because the lithostratigraphy is distinctly bimodal and feldspathic lithologies alternate with pyroxenite–chromitite without a consistent pattern, these units are descriptive rather than genetic.
Gain (Reference Gain1985) and Scoon et al. (Reference Scoon, Costin and Mitchell2025a) describe the UG2–UG3 units in detail. The 350 m-thick Merensky Footwall unit consists mainly of feldspathic rocks (Fig. 5). This exceptional thickness is characteristic of the sector; equivalent intervals elsewhere in the Eastern Limb and in the Western Limb’s Zwartklip facies may be less than 30 m thick (South African Committee for Stratigraphy, Reference Kent1980; Eales et al., Reference Eales, Field, De Klerk and Scoon1988; Mitchell et al., Reference Mitchell, Henckel and Mason-Apps2019a). In the upper part of the unit, 1–5 m-thick anorthosite and pyroxenite layers form poorly defined markers. The fine-grained Merensky Footwall pyroxenites contain little disseminated chromite, although thin chromite stringers may occur at contacts with the Merensky Footwall anorthosites. Vermaak (Reference Vermaak1976) described these pyroxenites as discontinuous layers that merge and bifurcate laterally. The anorthosites are similarly rather poorly developed. Immediately beneath the Merensky Reef, the stratigraphy is compositionally variable and includes both chaotic layering and well-defined anorthosite layers.
Bimodality is most pronounced in the Merensky unit, where 3–4 m of basal pyroxenite is sharply overlain by 8–10 m of anorthosite (Fig. 5). The Merensky pyroxenites merge and bifurcate laterally, and the lowest of the four layers may be absent. Norite–leuconorite is nearly absent, except for a discontinuous 1–10 cm layer between the pyroxenite and anorthosite. The laterally persistent Merensky anorthosite forms a prominent marker that caps a low ridge traceable for several kilometres. In the overlying Bastard unit, two or more pyroxenite layers are separated by thin leuconorite in the lower part and by norite–leuconorite centrally; the unit is capped by the Giant Mottled anorthosite. This 40–50 m-thick marker has a more complex internal stratigraphy than the Merensky anorthosite, including thin leuconorite and spotted-anorthosite layers (Vermaak, Reference Vermaak1976).
Although the LMZ lacks well-defined internal units, the Porphyritic Gabbro marker divides it into a lower noritic section and an upper-central gabbronoritic section. Thick mottled-anorthosite layers are intercalated with both norite and gabbronorite.
Chaotic layering
Localised chaotic igneous layering is a distinctive feature of the UCZ in the Eastern Limb. It occurs at several stratigraphic levels, including the K unit between the UG1 and MG4 chromitites near Zeekoegat (Cameron and Emerson, Reference Cameron and Emerson1959; Cameron, Reference Cameron1982) and the UG2 and Merensky Footwall units at Winnaarshoek (Fig. 5). Unlike metre-scale layering, it comprises 1–25 cm-thick pyroxenite, norite and anorthosite layers that extend laterally for only a few metres and are disrupted by small lenses of the same lithologies. Disseminated chromite is abundant in the pyroxenites but scarce in the norites and anorthosites.
Whole-rock geochemistry
Whole-rock data for the Winnaarshoek samples were presented and discussed by Scoon et al. (Reference Scoon, Costin and Mitchell2026). In summary, selected variation diagrams show strong bimodality among the lithological units. Pyroxenites define a restricted field, whereas norites–leuconorites, gabbronorites and anorthosites form trends controlled mainly by increasing plagioclase content. This modal control is evident on the Al2O3–MgO plot (Fig. 6). Limited fractionation across the 400 m interval is indicated by similar whole-rock Mg numbers (Mg#), calculated as [(Mg × 100)/ (Mg + Fe2+)] for pyroxenites and norites. This is shown by plots of Mg# against Al2O3 and MgO (Fig. 7). The narrow Mg# range of the pyroxenites (0.81–0.76) is independent of stratigraphic height, consistent with electron-microprobe data from Mitchell and Scoon (Reference Mitchell and Scoon2007) and Scoon et al. (Reference Scoon, Costin and Mitchell2025a). The Mg# values of LMZ norites are similar to those of UCZ samples, indicating that these higher stratigraphic levels are not more fractionated. Mg# is a poor proxy for fractionation in anorthosites because pyroxene is commonly restricted to mottles that crystallised from trapped melt (Vermaak, Reference Vermaak1976; Eales et al., Reference Eales, Marsh, Mitchell, De Klerk, Kruger and Field1986). Compositional variation in UCZ and LMZ anorthosite may result from trapped-liquid shifts, as defined by Barnes (Reference Barnes1986), and subsequent partial melting and metasomatism (Scoon and Mitchell, Reference Scoon and Mitchell1994; Mitchell and Manthree, Reference Mitchell and Manthree2002; Mitchell et al., Reference Mitchell, Henckel and Mason-Apps2019a). These processes are unlikely to have affected the bulk interstitial melt or resident magma. Among the trace elements, pyroxenites are relatively enriched in Cr because they contain abundant disseminated chromite and because Cr partitions into both orthopyroxene and augite. By contrast, norites–leuconorites, gabbronorites and anorthosites have very low Cr contents. Anorthosites have the highest Sr concentrations, consistent with preferential Sr partitioning into plagioclase. Pyroxenites consistently contain more Zr, which may reflect greater proportions of trapped melt than in the norites and anorthosites.
Plot of Al2O3 against MgO for whole-rock data of samples from the three Winnaarshoek drill holes. Samples are classified by stratigraphic unit or zone and lithology. The near-linear trends are interpreted to reflect modal variation rather than fractionation. Sample WH–1/64.0 is not representative of the Merensky Footwall pyroxenites as it is part of a chaotically layered sequence.
Figure 6 Long description
The scatter plot displays the relationship between MgO percentage on the vertical axis, ranging from 0 to 30 percent and Al subscript 2 O subscript 3 percent on the horizontal axis, ranging from 0 to 35 percent. The plot includes data points categorized into different rock types: Pyroxenites, Norites-leuconorites-gabbronorites and Anorthosites. Pyroxenites are clustered in the upper left region with higher MgO values. Norites-leuconorites-gabbronorites are spread in the middle region, while Anorthosites are located in the lower right with higher Al subscript 2 O subscript 3 values. A legend at the bottom identifies various symbols and colors representing different units and rock types, such as LMZ, norites, Bastard unit, Merensky anorthosite and others. The sample WH to 1/64.0 is marked within the Pyroxenites cluster. The plot illustrates the modal variation among the lithological units.
(a) and (b) Plots of Al2O3 and MgO against the Mg-number for whole-rock analyses of samples from the three Winnaarshoek drill holes. The data define groups based on lithology. Sample WH–1/64.0 is not representative of the Merensky Footwall pyroxenites as it is part of a chaotically layered sequence.
Figure 7 Long description
Two scatter plots labelled a and b with a shared y-axis label, Mg number. Plot a: The x-axis label is Al2O3 percent, ranging from 0 to 35 with tick marks at 5, 10, 15, 20, 25, 30 and 35. The y-axis label is Mg number, ranging from 0.50 to 0.85 with tick marks at 0.55, 0.60, 0.65, 0.70, 0.75, 0.80 and 0.85. Points form a curved band across the plot. At Al2O3 percent about 0 to 5, points lie around Mg number about 0.75 to 0.82. From Al2O3 percent about 10 to 25, points lie mostly around Mg number about 0.72 to 0.80. From Al2O3 percent about 25 to 33, points extend downwards from about Mg number 0.75 to about 0.52. Two dashed ellipses mark clusters: one near Al2O3 percent about 2 to 6 and Mg number about 0.76 to 0.82 and one near Al2O3 percent about 26 to 30 and Mg number about 0.74 to 0.80. Text inside the plot reads, Pyroxenites and, Norites leuconorites gabbronorites anorthosites. The label, WH 1 64.0, appears near the middle right. Plot b: The x-axis label is MgO percent, ranging from 0 to 30 with tick marks at 5, 10, 15, 20, 25 and 30. The y-axis label is Mg number, ranging from 0.50 to 0.85 with tick marks at 0.55, 0.60, 0.65, 0.70, 0.75, 0.80 and 0.85. Points form a curved band. At MgO percent about 0 to 5, points lie around Mg number about 0.52 to 0.70. From MgO percent about 5 to 20, points lie mostly around Mg number about 0.70 to 0.80. From MgO percent about 20 to 30, points lie around Mg number about 0.74 to 0.82. Two dashed ellipses mark clusters: one near MgO percent about 2 to 6 and Mg number about 0.70 to 0.80 and one near MgO percent about 22 to 28 and Mg number about 0.75 to 0.82. Text inside the plot reads, Pyroxenites and, Norites leuconorites gabbronorites anorthosites. The label, WH 1 64.0, appears near the center. A legend lists multiple sample groups using different marker shapes and colors, including entries beginning with, UCZ and, LMZ and several entries containing the word, Merensky, plus entries containing, pyroxenites, norites and, anorthosites.
Strontium and neodymium isotopes
The Sr and Nd isotope data are summarised in Table 2. Initial Sr ratios (SrR0) and 143Nd/144Nd ratios, expressed as epsilon Nd (ƐNd), were normalised to 2.055 Ga. Strongly negative ƐNd values from 17 samples were excluded and reported as no data, while the anomalous Sr result for sample WH–1/37.3 was omitted from the Table 3 averages. Some lithology- and unit-based averages in Table 3 should be treated cautiously because SrR0 varies with stratigraphic height in parts of the section. SrR0 is partly decoupled between feldspathic lithologies and pyroxenites, whereas Nd isotope values remain broadly uniform throughout the profile. This contrast contributes to the scatter in plots of SrR0 and 147Sm/144Nd against Zr (Fig. 8a–b), which also highlight the comparatively high Zr contents of pyroxenites. The SrR0–147Sm/144Nd plot (Fig. 8c) broadly separates UCZ pyroxenites and LMZ norites from UCZ feldspathic rocks, although the fields overlap and all three groups span wide 147Sm/144Nd ranges. By contrast, ƐNd plotted against 147Sm/144Nd (Fig. 8d) is highly scattered and shows no systematic relationship with lithology or stratigraphic height.
(a) and (b) Plots of Zr (ppm) against the Sr initial ratio (SrR0) and the 147Sm/144Nd isotopic ratio. (c) Plot of the 147Sm/144Nd isotopic ratio against SrR0. (d) Plot of the 147Sm/144Nd isotopic ratio against the ƐNd value. Whole-rock analyses of samples from the three Winnaarshoek drill holes. The data are age corrected and define broad groups based on lithology and yet show a broad scatter.
Figure 8 Long description
A scatter plot with the x-axis labeled, Zr (ppm) and the y-axis labeled, SrR0. The x-axis runs from 0 to 30 with labeled ticks at 0, 5, 10, 15, 20, 25 and 30. The y-axis runs from 0.700 to 0.714 with labeled ticks at 0.700, 0.702, 0.704, 0.706, 0.708, 0.710, 0.712 and 0.714. Most points lie between Zr 0 and 10 and SrR0 about 0.704 to 0.708. A smaller set of points extends to about Zr 25 to 30 with SrR0 around 0.703 to 0.706. Two higher points appear near Zr about 14 to 16 with SrR0 about 0.712 to 0.713. The image B showing a scatter plot with the x-axis labeled, Zr (ppm) and the y-axis labeled, 147Sm/144Nd. The x-axis runs from 0 to 30 with labeled ticks at 0, 5, 10, 15, 20, 25 and 30. The y-axis runs from 0.08 to 0.20 with labeled ticks at 0.08, 0.10, 0.12, 0.14, 0.16, 0.18 and 0.20. Most points lie between Zr 0 and 10 and 147Sm/144Nd about 0.10 to 0.18. Several points extend to Zr about 12 to 18 with 147Sm/144Nd about 0.13 to 0.16. A few points extend to about Zr 25 to 30 with 147Sm/144Nd about 0.12 to 0.14. The image C showing a scatter plot with the x-axis labeled, 147Sm/144Nd and the y-axis labeled, SrR0. The x-axis runs from 0.08 to 0.20 with labeled ticks at 0.08, 0.10, 0.12, 0.14, 0.16, 0.18 and 0.20. The y-axis runs from 0.700 to 0.714 with labeled ticks at 0.700, 0.702, 0.704, 0.706, 0.708, 0.710, 0.712 and 0.714. Most points lie between 147Sm/144Nd 0.09 and 0.14 and SrR0 about 0.704 to 0.708. Points extend to about 147Sm/144Nd 0.18 to 0.20 with SrR0 around 0.703 to 0.706. Two higher points appear near 147Sm/144Nd about 0.13 to 0.15 with SrR0 about 0.712 to 0.713. The image D showing a scatter plot with the x-axis labeled, 147Sm/144Nd and the y-axis labeled, epsilon Nd. The x-axis runs from 0.08 to 0.20 with labeled ticks at 0.08, 0.10, 0.12, 0.14, 0.16, 0.18 and 0.20. The y-axis runs from minus 14 to minus 4 with labeled ticks at minus 14, minus 12, minus 10, minus 8, minus 6 and minus 4. Most points lie between 147Sm/144Nd 0.09 and 0.15 and epsilon Nd about minus 9 to minus 5. A smaller set of points lies between 147Sm/144Nd about 0.16 and 0.19 with epsilon Nd about minus 8 to minus 6. A legend below the plots lists multiple categories, each shown with a distinct marker symbol.
Sr and Nd isotopic data for whole-rock samples, Winnaarshoek
Table 2 Long description
Columns report drill hole, depth in meters, height in meters, Rb and Sr in parts per million, 87Rb 86Sr, 87Sr 86Sr, 2 S. D., SrR0 at 2.055 Ga, Sm and Nd in parts per million, 147Sm 144Nd, 143Nd 144Nd, 2 S. D., 143Nd 144Nd at 2.055 Ga, and epsilon Nd. Hole WH 27 has six rows from 94.0 to 144.0 m depth, with Sr from 175.8 to 271.2 ppm and 87Sr 86Sr from 0.70855 to 0.70875; epsilon Nd is reported for three rows, from minus 8.16 to minus 6.19. Hole WH 1 spans 25.3 to 472.3 m depth, with heights from 27.2 m down to minus 419.8 m; Sr ranges from 47.9 to 541.6 ppm and Rb from 0.1 to 17.3 ppm. In WH 1, 87Sr 86Sr values include 0.70638 at 434.6 m, 0.70781 at 31.3 m, 0.71555 at 37.3 m, and 0.72294 at 431.4 m; corresponding 87Rb 86Sr values include 0.00289 at 434.6 m and 0.56738 at 431.4 m. WH 1 epsilon Nd values are listed for many rows and range from minus 12.47 at 64.0 m to minus 4.79 at 456.7 m, with several rows between minus 6 and minus 8. Hole WH 14 D2 includes seven rows from 55.2 to 59.3 m depth, with Sr from 48.1 to 71.5 ppm and 87Sr 86Sr from 0.70779 to 0.71431; epsilon Nd is reported for four rows, from minus 8.60 to minus 6.29. Several rows across all holes show a dash in the Nd isotope ratio columns and epsilon Nd, indicating no data for those fields.
Notes: S.D. = standard deviation; ‘-’ denotes no data
Averages of the isotopic data
Table 3 Long description
Columns report lithology and zone or unit, plus SrR0 at 2.055 Ga with standard deviation, correlation coefficient and sample count, and epsilon Nd at 2.055 Ga with standard deviation and sample count. Lower Main Zone, all samples lists SrR0 0.70853 with standard deviation 0.00016, correlation coefficient noted as no trend, n 6, and epsilon Nd minus 7.3 with standard deviation 1 and n 3. Upper Critical Zone 1 has three rows with SrR0, standard deviation, correlation coefficient and n left blank, while epsilon Nd is minus 7.2 with standard deviation 1.6 and n 54 for all samples, and minus 7.2 with standard deviation 1.7 and n 32 for pyroxenites. Upper Critical Zone 2, all pyroxenites lists SrR0 0.70709 with standard deviation 0.00116, correlation coefficient no trend, n 29, and epsilon Nd minus 7.3 with standard deviation 1.6 and n 23. Bastard unit, leuconorites and anorthosites gives an SrR0 range from 0.70701 to 0.70794 with correlation coefficient 0.81 and n 8, and epsilon Nd minus 7.9 with standard deviation 2.5 and n 7. Merensky unit, anorthosites gives an SrR0 range from 0.70650 to 0.70729 with correlation coefficient 0.75 and n 11, and epsilon Nd minus 7.1 with standard deviation 1.4 and n 11; Merensky and Bastard units, leuconorites and anorthosites gives an SrR0 range from 0.70650 to 0.70794 with correlation coefficient 0.85 and n 19, and epsilon Nd minus 7.4 with standard deviation 1.9 and n 18. UG2 UG3 Merensky Footwall units, leuconorites gabbronorites and anorthosites lists SrR0 0.70648 with standard deviation 0.00013, correlation coefficient no trend, n 17, and epsilon Nd minus 7.1 with standard deviation 1.6 and n 13; the pyroxenites row for the same units lists SrR0 0.70643 with standard deviation 0.00049, correlation coefficient left blank, n 13, and epsilon Nd minus 7.3 with standard deviation 1.9 and n 12. Bastard Reef pyroxenites lists SrR0 0.70763 with standard deviation 0.00071, correlation coefficient no trend, n 5, and epsilon Nd minus 6.4 with standard deviation 0.5 and n 4; Merensky Reef pyroxenites lists SrR0 0.70723 with standard deviation 0.00023, correlation coefficient no trend, n 9, and epsilon Nd minus 7.4 with standard deviation 0.9 and n 7; UG2 and UG3 unit pyroxenites lists SrR0 0.70627 with standard deviation 0.00029, correlation coefficient no trend, n 11, and epsilon Nd minus 6.8 with standard deviation 0.9 and n 10.
Notes: S.D. is the standard deviation; r2 is the correlation coefficient for linear regressions; n is the number of samples. 1 – Averages not shown due to upward progressions; 2 – excluding samples WH-1/37.3 and WH-14/D2/57.8; 3 – excluding Merensky Footwall pyroxenites; 4 – excluding Bastard middling; 5 – excluding Merensky pegmatite
When plotted against stratigraphic height, SrR0 values from the Winnaarshoek section define five broad patterns (Fig. 9): (1) UG2–UG3 pyroxenites show minor SrR0 scatter (average of 0.70627 ±0.00029). (2) Feldspathic lithologies in the UG2–UG3–Merensky Footwall units maintain a nearly constant SrR0 (average 0.70648 ±0.00013). (3) Merensky and Bastard Reef pyroxenites have higher SrR0 values than those of UG2–UG3 pyroxenites (respective averages of 0.70723 ±0.00023 and 0.70763 ±0.00071) and exhibit considerable scatter. (4) SrR0 increases stepwise upward through the Merensky–Bastard anorthosites with both intervals defining linear regressions (correlation coefficients of 0.75 and 0.81, respectively). (5) SrR0 at the base of the Merensky anorthosite matches values in the Merensky Footwall unit, whereas SrR0 at the top of the Bastard unit approaches the higher, near-invariant values of LMZ norites (average 0.70853 ±0.00016).
Sr initial ratio (SrR0) plotted against stratigraphic height for whole-rock analyses from drill holes WH–1 and WH–27 at Winnaarshoek. Data are age corrected; WH–14/D2 data are not shown. Three trends are recognised: scattered, near-invariant, and progressively increasing upward.
Figure 9 Long description
The diagram consists of two main sections labeled a and b, each representing data from different drill holes. In section a, the left side shows a stratigraphic column for drill hole WH-1, with various colored layers representing different geological units, such as PGE Reef, pyroxenite and anorthosite. Each layer is marked with specific depth values and labels like UG2 and UG3. To the right, a graph plots the Sr initial ratio against stratigraphic height, with values ranging from 0.706 to 0.708. The graph shows a trend line with data points indicating changes in the Sr initial ratio across different depths. Section b displays similar information for drill hole WH-27, with a stratigraphic column on the left and a graph on the right. The column is divided into units like the Lower Main Zone, Bastard unit and Merensky unit, each with corresponding depth values. The graph plots the Sr initial ratio, showing trends such as scattered, near-invariant and upward progression. The diagram includes a legend at the bottom, explaining the color coding and symbols used for different geological features and trends.
Regardless of the interval’s substantial thickness, the absence of SrR0 inflexions across the UG2–UG3–Merensky Footwall units may reflect a broader regional pattern. Within this interval, feldspathic lithologies and pyroxenites have broadly similar SrR0 values, although the pyroxenites show greater variability. Comparable metre-scale and finer variation occurs in pyroxenites associated with the Middle Group chromitites across the LCZ–UCZ boundary, where both unit means and individual samples have large standard deviations (Scoon et al., Reference Scoon, Reichhardt, Costin and Mitchell2025b).
The Merensky–Bastard SrR0 profile at Winnaarshoek closely resembles that at Atok mine, including substantial variability within the pyroxenites (Lee and Butcher, Reference Lee and Butcher1990; Fig. 10). Bastard Reef pyroxenite values return to Merensky-level compositions, as also reported at Hackney (Magson et al., Reference Magson, Roelofse, Bybee and Bolhar2024). By contrast, the pronounced intra-reef step at the Merensky level in the Western Limb, which persists into the hangingwall (Kruger and Marsh, Reference Kruger and Marsh1982; Eales et al., Reference Eales, Marsh, Mitchell, De Klerk, Kruger and Field1986; Kruger, Reference Kruger1990, Reference Kruger1992), and in the Platreef—the Northern Limb equivalent of the Merensky Reef (Cawthorn et al., Reference Cawthorn, Barton and Viljoen1985; Kruger, Reference Kruger1990; Harris and Chaumba, Reference Harris and Chaumba2001; Pronost et al., Reference Pronost, Harris and Pin2008; Scoon et al., Reference Scoon, Costin, Mitchell and Moine2020)—is absent from the Eastern Limb. Here, feldspathic rocks immediately below and above the Merensky Reef pyroxenites have similar SrR0 values (compare WH–1/60.5 and 62.0 with WH1–1/51.1 and 51.2; Fig. 9), indicating that the inflexion is restricted to the pyroxenites. SrR0 is strongly decoupled between lithologies in the Merensky–Bastard interval. Notwithstanding their interruption by the Bastard Reef, the metre-scale upward SrR0 trends in the Merensky and Bastard anorthosites define a single linear regression (correlation coefficient 0.85), suggesting that the two may originally have formed a continuous sequence. The sole Merensky pegmatite sample (WH–14/D2/57.8) has a markedly higher SrR0 value, comparable to those reported for reef pegmatites at Bafokeng mine, Rustenburg (Hayes et al., Reference Hayes, Maghdour-Mashhour, Ashwal, Smith, Ueckermann and Vermeulen2025).
Elevated LMZ SrR0 values at Winnaarshoek broadly accord with data from all three limbs, although values are still higher in the Northern Limb (Hamilton, Reference Hamilton1977; Cawthorn et al., Reference Cawthorn, Barton and Viljoen1985; Sharpe, Reference Sharpe1985; Eales et al., Reference Eales, Marsh, Mitchell, De Klerk, Kruger and Field1986; Kruger, Reference Kruger1994; Mitchell et al., Reference Mitchell, Eales and Kruger1998; Scoon et al., Reference Scoon, Costin, Mitchell and Moine2020; Magson et al., Reference Magson, Roelofse, Bybee and Bolhar2024). The near-constant trend in the basal norites is probably regional, whereas the much thicker overlying gabbronorites exhibit isotopic scatter and complex trends that vary among profiles.
(a) Sr initial ratio (SrR0) plotted against stratigraphic height through the Merensky and Bastard units at Winnaarshoek, using data from drill holes WH–1 and WH–14/D2. (b) Equivalent SrR0 profile for Atok (Bokoni) mine, based on Lee and Butcher (Reference Lee and Butcher1990). The profiles are broadly similar and show a progressive upward increase in SrR0 through the norites and anorthosites that can be represented by a single linear regression. All values are from whole-rock analyses and were age-corrected to 2.055 Ga for Winnaarshoek and 2.05 Ga for Atok.
Figure 10 Long description
Two adjacent plots and a stratigraphic column. The stratigraphic column is labeled Winnaarshoek and has depth ticks from minus 4 to 28 with the unit, m. The column contains stacked rock units with internal labels including 5 and 6 and several thin horizontal marker lines. To the right is a plot with the top axis labeled, Sr R subscript 0, with tick labels 0.706, 0.707 and 0.708. The left axis is labeled, m, with tick labels from minus 4 to 28. A vertical text label reads, Vertical Trends. A legend shows, Linear Regression and, Whole Rock. Black circular points are connected by a green line. The connected points start near 0.7065 at about 0 m, rise to about 0.7067 at about 4 m, then to about 0.7068 at about 6 m. The series continues upward with points near 0.7069 at about 8 m, about 0.7070 at about 10 m and about 0.7071 at about 12 m. Above this, points lie near 0.7070 to 0.7071 between about 14 m and 18 m, then increase to about 0.7072 at about 20 m, about 0.7073 at about 22 m, about 0.7074 at about 24 m and about 0.7076 to 0.7077 near 26 m to 28 m. A second set of points appears around 10 m to 14 m, with black points connected by a pink line, spanning roughly 0.7066 to 0.7072. Several small open square markers appear near the lower part of the plot around minus 2 m to 2 m, near 0.7070 to 0.7076. The image B showing a stratigraphic column and a plot. The stratigraphic column is labeled Atok or Bokoni Mine and has depth ticks from minus 4 to 24 with the unit, m. The column contains stacked rock units with internal labels including 5 and 6 and a thin basal unit labeled Pegmatite. To the right is a plot with the top axis labeled, Sr R subscript 0, with tick labels 0.706, 0.707 and 0.708. The left axis is labeled, m, with tick labels from minus 4 to 24. A legend shows, Vertical Trends, Linear Regression and, Whole Rock. Black circular points are connected by a green line. The connected points begin near 0.7067 at about 0 m, rise to about 0.7069 at about 4 m, about 0.7070 at about 6 m, about 0.7071 at about 8 m and about 0.7072 at about 10 m. The series continues upward to about 0.7071 at about 12 m, about 0.7070 at about 14 m, then increases to about 0.7072 at about 16 m, about 0.7074 at about 18 m, about 0.7075 at about 20 m and about 0.7076 near 22 m to 24 m. A cluster of black points appears near the bottom around minus 2 m to 0 m, spread horizontally between about 0.7066 and 0.7074.
ƐNd shows no consistent relationship with lithology or stratigraphic height. Mean values for the UCZ (–7.2 ±1.6) and LMZ (–7.3 ±1.0) overlap within analytical uncertainty (Table 3). Data from the Merensky–Bastard units are similar to those from Hackney (Magson et al., Reference Magson, Roelofse, Bybee and Bolhar2024), but do not reproduce the Merensky Reef ƐNd inflexion reported by Prevec et al. (Reference Prevec, Scoon and Raines2016) from the same Winnaarshoek drill hole, WH–1. Overall, the Winnaarshoek section broadly matches regional Western Limb values reported by Maier et al. (Reference Maier, Arndt and Curl2000).
Although SrR0 plotted against ƐNd defines several broad groups, the data are widely scattered (Fig. 11), indicating partial decoupling of the Sr and Nd isotope systems. By contrast, Northern Limb Platreef samples show higher SrR0 and, in some cases, more negative ƐNd. Whole-rock Sr and Nd isotope data should nevertheless be interpreted cautiously. In the Western Limb Merensky Reef, Prevec et al. (Reference Prevec, Ashwal and Mkaza2005) documented mineral-scale Nd-isotopic disequilibrium, with orthopyroxene ƐNd values of –7 to –8 and plagioclase values of –1 to –3. Seabrook et al. (Reference Seabrook, Cawthorn and Kruger2005) similarly identified SrR0 differences between orthopyroxene and plagioclase, and among individual plagioclase grains, in the Merensky Reef at Richmond.
Sr initial ratio (SrR0) plotted against ƐNd for whole-rock analyses from the three Winnaarshoek drill holes. Three Winnaarshoek samples outside the main fields are highlighted. Winnaarshoek UCZ data are compared with the more widely scattered Platreef data from Akanani (Scoon et al., Reference Scoon, Costin, Mitchell and Moine2020). LMZ samples from the two localities define distinct fields. With the exception of the Main Mineralised Reef at Akanani, data from both localities indicate partial decoupling of the Sr and Nd isotope systems.
Figure 11 Long description
The horizontal axis is labeled, Sr R subscript 0, with tick labels from 0.705 to 0.713. The vertical axis is labeled, epsilon subscript Nd, with tick labels from minus 14 to minus 4. Points are plotted with multiple marker styles and a legend. The legend lists these groups: LMZ norites; Bastard unit leucnorites and anorthosites; Bastard pyroxenites; Merensky anorthosites; Merensky pyroxenites; Merensky Footwall pyroxenites; Merensky Footwall unit gabbronorite leucnorite anorthosite; UCZ UQ3 norites and anorthosites; UCZ UQ3 units pyroxenites and harzburgite. A dashed outline encloses a dense cluster of points around Sr R subscript 0 near 0.706 to 0.708 and epsilon subscript Nd near minus 6 to minus 10. Text near this cluster includes, UCZ UQ3 Winnaarshoek and, Merensky Footwall, Merensky and Bastard units Winnaarshoek. A large filled field spans much of the middle to higher Sr R subscript 0 range, from about 0.708 to 0.713 and epsilon subscript Nd from about minus 10 to minus 5. Text within or near this field includes, Main Zone Winnaarshoek, Main Zone Akanani and, Platreef Unit Akanani. A separate hatched field is shown around Sr R subscript 0 about 0.710 to 0.712 and epsilon subscript Nd about minus 13 to minus 8, with nearby text, Main Mineralised Reef Akanani. Several individual points are labeled: WH 164.0; WH 140.0 267.8; WH 107.3. Across the plotted fields and clusters, higher Sr R subscript 0 values coincide with epsilon subscript Nd values that remain within the labeled range from about minus 10 to minus 5 in the large filled field, while the dashed outlined cluster occupies lower Sr R subscript 0 values with epsilon subscript Nd extending down to about minus 10. The hatched field occupies a narrower Sr R subscript 0 band near 0.710 to 0.712 with epsilon subscript Nd extending to about minus 13.
Discussion
Alternative hypothesis
Despite more than a century of research, the petrogenesis of the Rustenburg Layered Suite remains unresolved. Our lithostratigraphic and geochemical studies at Winnaarshoek, supported by regional field relationships, provide little support for conventional upper-level magma-chamber models. We instead propose that thin, pressurised magma sheets intruded pre-existing crystalline rocks, in some cases as out-of-sequence sills. We favour this hypothesis over models involving gravitational crystal settling (Wager and Brown, Reference Wager and Brown1968), double-diffusive convection (Irvine et al., Reference Irvine, Keith and Todd1983), or repeated replenishment and progressive hybridisation with resident magma (Eales et al., Reference Eales, Marsh, Mitchell, De Klerk, Kruger and Field1986; Eales, Reference Eales2014). Initially applied to thin LMZ pyroxenite layers (Mitchell, Reference Mitchell1996), the model was later extended to the Merensky and Bastard Reefs at Winnaarshoek (Mitchell and Scoon, Reference Mitchell and Scoon2007) and Sedibelo in the Western Limb (Mitchell et al., Reference Mitchell, Henckel and Mason-Apps2019a, Reference Mitchell, Scoon and Sharpe2019b). Consistent with Marsh’s (Reference Marsh and Watts2007) theoretical reasoning, convection, diffusion and fractionation were probably limited within individual sheets. The metre-scale layers formed mainly by batch crystallisation, with little evidence that fractionated residues were removed. At Winnaarshoek, abrupt lithological and geochemical discontinuities—especially between anorthosites and pyroxenites—therefore reflect changes in parental magma composition rather than repeated fractionation cycles.
Incremental construction of the Rustenburg Layered Suite, mainly through stacked magma sheets, accounts for the overlapping stratigraphic relationships in the Eastern Limb shown in Fig. 2 and in the longitudinal section of Scoon et al. (Reference Scoon, Reichhardt, Costin and Mitchell2025b). Early doming of the floor rocks produced discrete sectors with distinct stratigraphy. Field relationships show no evidence that fractionated residues from lower zones supplied overlying units. For example, the Lower Zone could not have supplied the LCZ or UCZ because these higher zones locally rest unconformably on marginal rocks or directly on the floor. Similarly, the UCZ is unlikely to contain LCZ residues, and although the LMZ commonly overlies the UCZ, it may also rest directly on marginal rocks or the floor. This interpretation rests on two principles. First, low intrusion rates and lateral spreading of magma sheets prevented a magma chamber from forming. Second, the sheets were emplaced beneath the Rooiberg Group acting as structural wedges that progressively raised the roof while preserving a broadly planar contact. The absence of an upper chill zone supports this model. Stepwise emplacement separated by brief hiatuses therefore explains the tabular geometry of the limbs more readily than a few large magma injections.
Our thin-sheet intrusion model contrasts with interpretations that invoke major replenishment of an upper-level magma chamber at the base of the Merensky Reef, including the fountains and plumes of Campbell et al. (Reference Campbell, Naldrett and Barnes1983) and the basal flows of Eales et al. (Reference Eales, Marsh, Mitchell, De Klerk, Kruger and Field1986). In the Western Limb, Sr-isotopic inflexions have been explained by initial emplacement of LMZ magma, followed by underplating by sheets that crystallised the Merensky and Bastard units (Eales et al., Reference Eales, Marsh, Mitchell, De Klerk, Kruger and Field1986). Out-of-sequence sill emplacement more simply accounts for the isotopic decoupling between anorthosites and pyroxenites in the Merensky–Bastard interval. The return of Bastard Reef pyroxenite values to Merensky-level compositions strongly supports this interpretation and provides a consistent framework for all UCZ pyroxenites, including those in the UG2–UG3 units.
A two-stage model for development of the Winnaarshoek section as a stacked sill complex is illustrated in Fig. 12. First, successive magma sheets established a primary stratigraphy of norite–leuconorite, gabbronorite and anorthosite. Pyroxenitic sills then intruded at multiple stratigraphic levels, commonly along contacts above or below anorthosite. In this model, chromitites accumulated from chromite slurries entrained at depth (Eales, Reference Eales2002; Eales and Costin, Reference Eales and Costin2012) and transported by pyroxenitic sill magmas (Scoon et al., Reference Scoon, Costin and Mitchell2025a), rather than from basal flows injected into an upper-level chamber. This interpretation may resolve the chromium-budget problem (Cameron, Reference Cameron1977; Eales, Reference Eales and Stowe1987), namely the large quantity of chromium required to explain anomalous concentrations across the LCZ–UCZ. The close association of chromitites and pyroxenites characterises all three chromitite groups. Decoupling pyroxenites–chromitites from norites–anorthosites, a central element of our alternative hypothesis, explains UCZ layering more simply than repeated chamber replenishment followed by hybridisation with resident magma. The conventional interpretation of lithological units as fractionation cycles should therefore be reconsidered.
Schematic model for the development of igneous layering in the Winnaarshoek study interval. (a) Thin basaltic to basaltic-andesitic magma sheets formed the primary noritic–anorthositic stratigraphy of the UG2–UG3–Merensky Footwall units beneath the Rooiberg felsites. Although magma is shown ponded at the roof contact, the model excludes a large upper-level magma chamber. Near-invariant Sr initial ratios (SrR0) indicate isotopically similar magma batches. (b) and (c) Continued intrusion of similar sheets, some crystallising as noritic sills, progressively raised the roof while preserving broadly planar contacts and near-invariant SrR0 values. Partial melting and metasomatism of pre-existing norites and gabbronorites produced laterally persistent mottled anorthosite. (d) UG2 and UG3 pyroxenites intruded the feldspathic stratigraphy as composite sills carrying chromite slurries from depth. Their SrR0 values overlap those of the feldspathic rocks but show greater scatter, reflecting local heterogeneity during emplacement. Chaotic layering formed where thin sheets disrupted the norite. (e) Renewed emplacement of basaltic–andesitic magma sheets above the Merensky Footwall unit formed the Merensky–Bastard feldspathic rocks. Their upward increase in SrR0 records changing compositions through time. (f) Merensky and Bastard pyroxenites were subsequently emplaced as composite sills, followed by development of the LMZ above the Giant Mottled anorthosite. Pyroxenite SrR0 values remain scattered, with Bastard Reef values returning toward Merensky-level compositions. LMZ norites, shown as postdating both reefs, define a higher, near-invariant SrRs0 trend.
Figure 12 Long description
The development of igneous layering in the Winnaarshoek study interval through six stages, labeled a to f, is illustrated. Each stage is represented by a vertical column with various colored layers, indicating different rock types and processes. Stage a shows the initial formation of noritic and anorthositic stratigraphy beneath Rooiberg felsites, with a thin basaltic magma sheet at the roof. Stage b depicts continued intrusion of similar sheets, raising the roof while maintaining planar contacts. Stage c introduces UG2 and UG3 pyroxenites as composite sills with chromite slurries, causing chaotic layering. Stage d shows renewed emplacement of magma sheets forming Merensky to Bastard feldspathic rocks with increasing Sr initial ratios. Stage e illustrates the emplacement of Merensky and Bastard pyroxenites as composite sills, followed by the development of the LMZ above the Giant Mottled anorthosite. Stage f concludes with LMZ norites postdating both reefs, defining a higher, near-invariant Sr initial ratio trend. The diagram includes a legend indicating rock types and processes, such as norite, chaotic layering, PGE reef, harzburgite, pyroxenite, anorthosite, gabbronorite, leuconorite, noritic magma sheet, transient magma sheet, Rooiberg Group with uplift, roof, LMZ, composite sills, primary stratigraphy and UCZ. Arrows and lines indicate vertical trends of Sr initial ratios, with equilibrium, upward progression and scattered trends marked.
Repeated bifurcation and coalescence of the Merensky and Bastard pyroxenites support their interpretation as composite sills (Scoon et al., Reference Scoon, Costin and Mitchell2026). Chromite stringers along internal sill contacts may be reaction features formed during pauses in magma emplacement (Scoon and Costin, Reference Scoon and Costin2018). Merensky Reef potholes likewise support incremental development of the pyroxenite and chromitite layers: upper reef layers cut downward through lower pyroxenites and chromitites before penetrating the underlying norites and anorthosites.
Early regional mapping of the Bushveld Complex anticipated several elements of our hypothesis, notably doubts about a high-level magma chamber and recognition of out-of-sequence sill emplacement (e.g. Schwellnus et al., Reference Schwellnus, Engelbrecht, Coertze, Russell, Maherbe, Van Rooyen and Cooke1962; Coertze, Reference Coertze1970; reviewed by Willemse, Reference Willemse1969). Our interpretation of UCZ pyroxenites, with subordinate harzburgites, and chromitites as out-of-sequence sills was influenced by studies of peridotite sills in the Rum intrusion, northwest Scotland (Harker and Clough, Reference Harker and Clough1904; Bédard et al., Reference Bédard, Sparks, Renner, Cheadle and Hallworth1988), and by Lee and Butcher’s (Reference Lee and Butcher1990) analysis of the Merensky–Bastard interval. Similar associations between chromitites and out-of-sequence sills were proposed by Mondal and Mathez (Reference Mondal and Mathez2007), Maier and Barnes (Reference Maier and Barnes2008) and Voordouw et al. (Reference Voordouw, Gutzmer and Beukes2009). Mungall et al. (Reference Mungall, Kamo and McQuade2016) and Scoates et al. (Reference Scoates, Wall, Friedman, Weis, Mathez and VanTongeren2021) subsequently developed this model using zircon geochronology, providing a basis for thermal models of parts of the Rustenburg Layered Suite as a sill complex (Robb and Mungall, Reference Robb and Mungall2020; Yao et al., Reference Yao, Mungall and Jenkins2021). We nevertheless question some zircon-based relative ages for mafic–ultramafic sills because field relationships indicate that these layers are sequential relative to one another but non-sequential relative to feldspathic strata. Zircon age determination also lacks the resolution to distinguish closely spaced layers such as the Merensky and Bastard Reefs (Latypov et al., Reference Latypov, Chistyakova and Kramers2017). Because the Bastard Reef forms potholes into the Merensky Reef, it must postdate rather than predate it (Cousins, Reference Cousins and Haughton1964; Viljoen and Hieber, Reference Viljoen, Hieber, Anhaeusser and Maske1986). We identify four additional limitations of the thermal models: (1) our model involves crystal-bearing silicate melts, not silicate crystal slurries or mushes; only chromite was transported as a slurry; (2) modelling magma-sheet dynamics and emplacement hiatuses across the Rustenburg Layered Suite is impractical if tens of thousands of variably thick sheets were involved; (3) because we extend the hypothesis to the Main and Upper Zones, we reject their derivation from fractionated resident magma supplied by underlying zones; and (4) our interpretation of anorthosite formation differs fundamentally from that assumed in these models.
Anorthosites
Early anorthosite formation within the primary feldspathic stratigraphy is central to our hypothesis (Mitchell et al., Reference Mitchell, Scoon and Sharpe2019b). This view differs fundamentally from models involving plagioclase flotation in a magma chamber (Wager and Brown, Reference Wager and Brown1968; Vermaak, Reference Vermaak1976; Eales et al., Reference Eales, Marsh, Mitchell, De Klerk, Kruger and Field1986) or emplacement of plagioclase-rich crystal mushes as sill-like bodies (Ashwal et al., Reference Ashwal, Webb and Knoper2005; Cawthorn and Ashwal, Reference Cawthorn and Ashwal2009; Roelofse and Ashwal, Reference Roelofse and Ashwal2012; Bourdeau et al., Reference Bourdeau, Zhang, Hayes and Logue2020). The first requires a magma chamber, whereas the second is difficult to reconcile with the rheology of plagioclase-rich magma (Philpotts et al., Reference Philpotts, Shi and Brustman1998, Reference Philpotts, Brustman, Shi, Carlson and Denison1999; Arai and Maruyama, Reference Arai and Maruyama2005). Mitchell and Manthree (Reference Mitchell and Manthree2002) instead proposed that mottles in many anorthosites formed when repeated magma-sheet intrusions partially melted and metasomatised leuconorites and leucogabbronorites. Supporting field relationships include centimetre-thick anorthosite layers within thick norites and metre-thick noritic layers enclosed by anorthosites more than 50 m thick. Later pyroxenitic sills may have caused further melting and metasomatism but were not the primary cause because anorthosites also occur independently of pyroxenites, notably the Giant Mottled anorthosite at the UCZ–LMZ boundary and the Lower, Main and Upper Mottled anorthosites in the LMZ (Fig. 2). This independence is fundamental to the Mitchell and Manthree (Reference Mitchell and Manthree2002) model. The complex UCZ feldspathic host stratigraphy records repeated, including out-of-sequence, intrusion of thin magma sheets. Evidence for repeated melting includes removal of sodic plagioclase rims, producing anomalously calcic residual compositions, and development of iron-rich pyroxene containing minute Ti-magnetite grains in the mottles (Scoon and Mitchell, Reference Scoon and Mitchell1994; Mitchell et al., Reference Mitchell, Scoon and Sharpe2019b). In some cases, low-degree melting may have nearly dissolved the primary pyroxene grains, as Naslund (Reference Naslund1986) discussed and demonstrated experimentally.
Previously reported isotopic disequilibrium in chromitites and the Merensky Reef may reflect sill-emplacement processes and the early formation of anorthosites rather than contrasting magma compositions. For example, the accessory plagioclase in the UG2 and UG3 chromitites is liable to have been incorporated from anorthositic wall rocks disrupted by chromite slurries during sill emplacement (Mitchell et al., Reference Mitchell, Scoon and Sharpe2019b; Scoon et al., Reference Scoon, Costin and Mitchell2025a; Reference Scoon, Reichhardt, Costin and Mitchell2025b).
Mechanical incorporation of plagioclase from anorthositic wall rocks, rather than magma mixing, could explain the mineral-scale isotopic disequilibrium reported in the Merensky Reef (Prevec et al., Reference Prevec, Ashwal and Mkaza2005). Seabrook et al. (Reference Seabrook, Cawthorn and Kruger2005) attributed isotopically distinct plagioclase populations in the Richmond Merensky Reef to gravitational settling in a magma chamber. However, this interpretation conflicts both with experimental evidence that plagioclase is buoyant in basaltic magma (Bottinga and Weill, Reference Bottinga and Weill1970; McBirney and Noyes, Reference McBirney and Noyes1979) and with our chamber-free model. A more plausible explanation is that the contrasting plagioclase populations comprise crystals formed from the intruding sill and crystals incorporated from the wall rocks.
Parental magma lineages
Interpreting layered igneous rocks depends on identifying parental magma compositions, which are difficult to reconstruct from complex cumulates (e.g. Cox et al., Reference Cox, Bell and Pankhurst1979; Naslund and McBirney, Reference Naslund, McBirney and Cawthorn1996; Marsh, Reference Marsh2013). As chilled margins are rare or absent in the Rustenburg Layered Suite, linking marginal-rock magmas to the UCZ is problematic (Eales, Reference Eales2000). Scoon and Mitchell (Reference Scoon and Mitchell2023) interpreted the marginal rocks not as the chilled carapace of an upper-level chamber, but as stacked sills emplaced between the Transvaal floor rocks and the Rooiberg Group felsites. Micropyroxenite and skeletal olivine locally record rapid cooling at some contacts, although most marginal rocks are medium- to coarse-grained. Their similarity to fully enclosed sills in the Transvaal Supergroup (Sharpe, Reference Sharpe1981; Harmer and Sharpe, Reference Sharpe1985) supports grouping them into a continuous Marginal Sill Phase (MSP; Fig. 2). Although the parental magmas may have been connected at depth, its noritic and pyroxenitic sills and younger peridotite intrusions could not have fractionated in situ. Instead, they form compositionally and isotopically diverse sills and stocks with widely varying SrR0 values (Harmer and Sharpe, Reference Sharpe1985).
Our UCZ model therefore recognises three parental magma lineages (Scoon et al., Reference Scoon, Costin and Mitchell2025a): (1) basaltic–andesitic magmas that formed norites–leuconorites, gabbronorites, and anorthosites; (2) basaltic magmas that produced feldspathic orthopyroxenites; and (3) subordinate picritic magmas associated with feldspathic harzburgites. Their dominant liquidus phases—plagioclase, orthopyroxene and olivine, respectively—distinguish the lineages. The B1–B3 classification of Harmer and Sharpe (Reference Sharpe1985) and the U–A scheme of Sharpe and Irvine (Reference Sharpe and Irvine1983) are too simplified and contextually incompatible for this study. Recognising these three lineages provides the framework for interpreting Sr–Nd isotope systematics. The magmas evolved in a deep-crustal staging chamber through repeated replenishment, remelting and fractionation, consistent with Cox et al. (Reference Cox, Bell and Pankhurst1979). Many UCZ intrusive batches probably contained similar melt fractions, as indicated by limited fractionation and consistent orthopyroxene and plagioclase compositions; crystal load was commonly their main difference. This interpretation accords with the limited Sr and Nd isotope variation at the Middle Group chromitites (Scoon et al., Reference Scoon, Reichhardt, Costin and Mitchell2025b) and throughout the UG2–UG3 units (Scoon et al., Reference Scoon, Costin and Mitchell2025a). Layering in these units and the Merensky Footwall records younger pyroxenitic sills intruding an older noritic–anorthositic sequence whose parental magmas had similar isotopic compositions. By contrast, isotope variation in the Merensky–Bastard interval records substantial changes in deep-crustal parental-magma signatures despite lithological similarity to the underlying UCZ. Like the UCZ, the noritic LMZ probably formed through successive intrusion of thin magma sheets; limited fractionation argues against crystallisation in a large chamber. Moreover, similar major-element compositions in UCZ and LMZ norites challenge the widely accepted abrupt transition from B2 to B3 magmas. Their trace-element and Sr-isotope differences are better explained by deep evolution within a single basaltic–andesitic lineage. The Merensky–Bastard units therefore need not be assigned to either a ‘transitional zone’ (Vermaak, Reference Vermaak1976) or the LMZ (Kruger, Reference Kruger1990), as discussed by Mitchell and Scoon (Reference Mitchell and Scoon1991). Their norites and anorthosites may record the deep-seated UCZ–LMZ transition, while the pyroxenitic magmas retained UCZ characteristics.
Strontium and Nd isotope systematics indicate variable crustal contamination of most parental magmas in the Rustenburg Layered Suite, particularly those of the UCZ and LMZ (e.g. Harmer and Sharpe, Reference Sharpe1985; Walraven et al., Reference Walraven, Armstrong and Kruger1990; Kruger, Reference Kruger1994; Mitchell et al., Reference Mitchell, Eales and Kruger1998; Eales, Reference Eales2000; Maier et al., Reference Maier, Arndt and Curl2000; Pronost et al., Reference Pronost, Harris and Pin2008; Zeh et al., Reference Zeh, Ovtcharova, Wilson and Schaltegger2015). At Winnaarshoek, however, the two isotope systems are partly decoupled, perhaps through preferential addition of radiogenic Sr and a more juvenile Nd component (Fig. 11). SrR0 is higher in Merensky–Bastard pyroxenites than in UG2–UG3 pyroxenites, rises progressively through the Merensky and Bastard leuconorites–anorthosites, and reaches still higher values in the LMZ. These trends indicate that each magma sheet retained isotopic heterogeneity acquired as its parental magma evolved at depth and are consistent with increasing crustal contamination or local assimilation of an isotopically enriched contaminant. The deep processes that modified the parental magmas—including the relative contributions of mantle-derived melts, internal fractionation, and lower-crustal assimilation—are beyond the scope of this study. Interpretation is further complicated because variably contaminated magmas transported silicate crystals with potentially distinct isotopic signatures from depth and incorporated crystals in situ from earlier magma batches.
PGE Reefs
Our igneous-layering model contrasts with conventional explanations for PGE enrichment in upper-level magma chambers, including double-diffusive convection (Irvine et al., Reference Irvine, Keith and Todd1983), fountain- or plume-driven mixing of PGE-enriched resident magma (Campbell et al., Reference Campbell, Naldrett and Barnes1983), lateral mixing induced by basal flows (Scoon and Teigler, Reference Scoon and Teigler1994; Naldrett et al., Reference Naldrett, Kinnaird, Wilson, Yudovskaya and Chunnett2012), and in situ crystallisation from superheated magmas (Latypov et al., Reference Latypov, Chistyakova, Page and Hornsey2015). In our thin-sheet intrusion model, PGE mineralisation was concentrated at depth and transported upward by pyroxenitic magma sheets carrying chromite slurries (Mitchell and Scoon, Reference Mitchell and Scoon2007; Eales and Costin, Reference Eales and Costin2012; Scoon et al., Reference Scoon, Costin and Mitchell2025a). The PGE occurred partly as microxenocrysts, possibly attached to chromite xenocrysts, and partly in immiscible sulfide melts (Scoon and Teigler, Reference Scoon and Teigler1994). Repeated sill emplacement separated by hiatuses can explain both the layered PGE distribution in the UG2 (Fig. 3) and the multiple mineralised zones of the Merensky Reef (Fig. 4). The lateral continuity of PGE-rich layers, including the thin UG2 Leader seams, and their alternation with weakly mineralised layers are therefore best explained by successive, lithologically similar magma sheets with markedly different PGE contents.
Conclusions
The UCZ section studied at Winnaarshoek has a distinctly bimodal lithostratigraphy, in which feldspathic rocks alternate irregularly with subordinate pyroxenites and chromitites. Whole-rock geochemistry distinguishes two principal silicate groups—noritic, gabbronoritic and anorthositic rocks, and pyroxenites—showing that the lithological units are descriptive rather than genetic. The isotope data indicate partial decoupling of the Sr and Nd systems: SrR0 varies systematically with lithology and stratigraphic height in parts of the section, whereas ƐNd (mean −7.2 ±1.6) shows no consistent relationship with either. Below the Merensky Reef, feldspathic rocks maintain nearly constant SrR0 over 400 m (mean 0.70648 ±0.00013), broadly overlapping the more variable UG2–UG3 pyroxenites (mean 0.70627 ±0.00029). Merensky and Bastard Reef pyroxenites have similar but higher and more scattered SrR0 values than the underlying pyroxenites (means 0.70723 ±0.00023 and 0.70763 ±0.00071, respectively). SrR0 increases progressively through the Merensky anorthosite and the leuconorites–anorthosites above the Bastard Reef (0.70650–0.70794). These values define a single linear regression (correlation coefficient 0.85) that may extend to the base of the LMZ, where SrR0 reaches a higher, near-invariant value (0.70853 ±0.00016). This upward increase in the anorthosites above the Merensky Reef, together with the return of Bastard Reef values to Merensky-level compositions, distinguishes the Eastern Limb profile from that of the Western Limb.
Sr-isotopic decoupling between anorthosites and pyroxenites in parts of the succession supports emplacement as thin magma sheets. The pyroxenites and chromitites, locally accompanied by subordinate harzburgites, formed composite sills intruded into an older, largely crystalline feldspathic succession, commonly along contacts between pre-existing noritic and anorthositic strata. Each sheet retained isotopic heterogeneity acquired during evolution of its parental magma in a deep-crustal staging chamber. Stratigraphic isotope trends indicate increasing crustal contamination or local assimilation of an isotopically enriched contaminant, although their complexity and magnitude vary among magma types, implying more complex addition patterns than previously recognised. PGE layering in the UG2 chromitite and Merensky Reef is best explained by mineralisation concentrated at depth and transported upward in pyroxenitic or picritic sills carrying chromite slurries. The non-sequential layering associated with thin magma sheets supports interpreting the Rustenburg Layered Suite as a stacked sill complex rather than the product of a high-level magma chamber.
Acknowledgements
This study was funded by Rhodes University, Grahamstown, and Rice University, Houston. RNS logged and sampled the drill holes reported here and gratefully acknowledges the staff of Marula mine (Impala Platinum), and in particular Johannes du Plessis and Sifuso Mthethwa. We thank Bertrand Moine for assisting with the geochemical data acquisition and Eva Hancox for preparing the diagrams. We also appreciate the thoughtful reviews provided by Marina Yudovskaya and two anonymous referees, as well as the editorial guidance and oversight of Roger Mitchell and William Smith.
Competing interests
The authors declare none.
Open access
