The Bushveld Complex hosts abundant fine-grained noritic to melanoritic blocks that occur predominantly within the Critical Zone (CZ) and Main Zone (MZ) but likely extend to other stratigraphic levels of the intrusion. These blocks range from centimetre-scale angular fragments to large lenticular and tabular bodies up to ∼0.5–1.0 m thick and ∼ 10 m long. Field relationships show that the blocks truncate, indent, and locally penetrate igneous layering involving anorthosite, orthopyroxenite, and chromitite, and are themselves overlain by younger layers of the same lithologies. These relationships demonstrate that the blocks fell from above through a melt column and impacted an already formed cumulate pile, allowing the blocks to be interpreted as magmatic dropstones. These dropstones are texturally and compositionally distinct from their host orthopyroxenitic and anorthositic cumulates. They are uniformly fine-grained and commonly display textural equilibration with ∼120° triple junctions. Orthopyroxene in the dropstones is characterized by low Cr₂O₃ (<0.6 wt%) and Al₂O₃ (<1.5 wt%) contents, indicating crystallization at low crustal pressures (<2 kbar) within the shallow Bushveld chamber and ruling out derivation of blocks from a deep staging reservoir. At the same time, the Mg# and Cr contents of orthopyroxene as well as the initial 87Sr/86Sr ratios of plagioclase closely match those of CZ and MZ cumulates at the stratigraphic levels where the dropstones occur, demonstrating that the dropstones crystallized from Bushveld parental magmas. These observations demonstrate that the dropstones likely represent fragments of a roof-grown, hornfels-like lithology that was episodically disrupted and collapsed onto the inward-growing chamber floor. We interpret their formation as a two-stage process involving (1) crystallization of melanoritic rocks at the roof from parental magmas locally contaminated by Si-rich partial melts, followed by (2) thermal recrystallization during magma replenishment, producing fine-grained and texturally equilibrated hornfels. The systematic interaction between falling dropstones and floor cumulates indicates that the top of the cumulate pile was a mechanically coherent hardground at the time of dropstone emplacement, with igneous layering (including massive chromitites) already fully developed at the crystal-liquid interface. These results provide direct physical constraints on the growth, internal architecture, and chemical evolution of the Bushveld magma chamber.

The Bushveld Complex in South Africa hosts the lion’s share of the world’s noble metal resources in platinum reefs – thin layers of silicate/chromite rocks containing platinum-rich sulphides. The reefs are widely attributed to multiple replenishments by ore-forming magmas that have been entering the evolving Bushveld chamber through numerous feeder conduits. The replenishment events are marked by regional and local disconformities/unconformities, significant isotopic shifts, and notable reversals in the whole-rock and mineral compositions. Surprisingly, however, so far no single feeder conduit for platinum reefs has been found despite extensive surface and underground mining for over a century. Feeder conduits appear entirely absent from the Bushveld Complex. This paradox has long been known but has never been specifically addressed. Here, we suggest that the absence of feeder channels is a natural consequence of the magma chamber replenishment through a cumulate pile. The fossilization of the feeder channels in the cumulate pile is likely impeded by two principal factors: (a) a cumulate pile is too hot to enable efficient cooling and crystallization of magma flowing through the channels, and (b) the channels are closed by an adjacent elastically deformable pile immediately after cessation of the magma emplacement. The feeding dykes are thus absent because there is little chance for the conduits to get preserved in a hot and deformable cumulate pile of layered intrusions.

A common feature of layered intrusions is the absence of the Upper Border Series (UBS) - a rock sequence that is expected to grow from the roof downwards in large basaltic magma chambers. This is surprising because magma cooling occurs predominantly through the roof of magma chambers. The lack of the UBS is thus indicative of some fundamental reason that prevents basaltic magmas from crystallizing at the chamber roof. Here, we propose that this could be attributed to a layer of anatectic granitic melt that develops on the top of evolving basaltic chambers. The layer may be formed by partial melting of silica -rich crustal rocks in response to the upward transfer of heat from the crystallizing magma. The heat transfer takes place through the initial UBS, resulting in this unit becoming suspended between two buoyant liquids: granitic melt above and basaltic melt below. This unstable position leads to the disruption of the UBS and its collapse onto the temporary chamber fl oor. A two -layer magma system thus forms, with a light granitic melt fl oating atop a denser basaltic magma. The granitic melt is expected to be superheated with respect to its liquidus temperature via heat transfer from the underlying basaltic magma. In the absence of a solid surface at the top of the basaltic magma layer, the growth of the UBS cannot resume due to the lack of a solid place for new crystals to attach. Further crystallization of basaltic magma may therefore only continue from the base upwards, producing a layered intrusion that consists of only fl oor rock sequences.

The controversy over the origin of massive chromitites in layered intrusions has recently become more contentious than ever before. At issue is whether they are produced via gravity settling/in situ crystallization of chromite directly on the chamber floor or by kinetic sieving, metasomatic replacement, or sill-like intrusions beneath the chamber floor, i.e., in deep parts of the cumulate pile. The latter group of models implies that massive chromitites have never been on the chamber floor. In this paper, we show that decisive clues to the ‘chamber floor dilemma’ come from field relations of massive chromitite with magmatic dropstones in the Bushveld Complex, South Africa - the largest fossilized magma chamber in the Earth's continental crust. The roof sequences in such basaltic magma chambers are known to be inherently unstable; therefore, crustal instability results in their breakdown and collapse as angular blocks (i.e., magmatic dropstones) through the resident melt column onto the upward-growing chamber floor. We have discovered such magmatic dropstones in the sequences that host massive UG1, UG2, and MG2 chromitites in the Critical Zone of the Bushveld Complex. The dropstones are easily recognizable in the field because they are composed of fine-grained melanorite/orthopyroxenite, which are texturally dissimilar from and have sharp contacts with adjacent cumulate rocks (chromitite, anorthosite, norite, etc.). The dropstones range in shape from angular to lenticular fragments a few centimetres in size to large tabular dropstones of ~0.5–1.0 m across and ~ 10 m long. The dropstones indent pre-existing layers of chromitite/anorthosite beneath them and are covered by subsequently deposited layers of the same rocks above them. Some dropstones appeared to have knifed into chromitite/anorthosite layers and cut them off, and a few dropstones appeared to have been driven into the floor cumulates and crumpled the layers outwards and upward. Also, some beheaded dropstones indicate the truncation of the floor cumulates by planar erosional surfaces. The physical relationships of dropstones with their host rocks indicate that: (1) there was almost invariably a sharp interface between the top of the inward-growing cumulate pile and the overlying resident melt; (2) the uppermost part of the cumulate pile was coherent and igneous layering, involving chromitites, was generally fully developed right up to the crystal-liquid interface; (3) by the time of dropstones's landing, the chromitite layers were already formed on the chamber floor, with the resident melt immediately overlying them; (4) the existence of chromitite layers at the crystal-liquid interface implies their formation through the deposition of chromite alone directly onto the chamber floor; and (5) such crystal deposition requires the resident melt in the chamber to be chromite-only saturated (i.e., no other liquidus phases in the chamber). The occurrence of magmatic dropstones thus rules out the formation of massive chromitites by processes that operate at some depth in the cumulate pile (e.g., kinetic sieving, metasomatic replacement, or sill-like intrusions) and rather indicate their deposition via in situ crystallization from a chromite-only-saturated magma directly on the upward-growing floor of a Bushveld magma chamber. A thorough review of this concept demonstrates its high potential in explaining most field, textural and chemical features of massive chromitites in the Bushveld Complex.

Mafic layered intrusions show abundant evidence for the formation through the progressive inwards growth of a mushy solidification layer, with a central body of nearly crystal-free resident melt. The thickness and permeability of the mushy layer play a vital role in operation of several petrogenetic processes but remain poorly constrained in layered intrusions. To address this issue, we have re-examined the cumulate sequence in the Upper and Critical Zones of the Bushveld Complex. Although some observations (e.g., soft-sediment deformation) are indicative of cumulate rocks being mushy and having experienced interstitial melt percolation (e.g., metasomatic anorthosites), other field and textural evidence suggest the contrary. Some of the clearest evidence comes from sections through chromitite and magnetitite layers as well as through the Merensky Reef, layers that were deposited on the eroded rocks of the temporary chamber floor. Below the erosional surface, the layered rocks (1) were tilted and truncated without losing their internal coherence, (2) have knife-sharp contacts with the overlying rocks, (3) show minimal evidence of penetration by downward-percolating melt, (4) reveal sharp truncation of individual crystals, particularly beheading of pyroxene oikocrysts and intersecting of polysynthetic twinning in plagioclase, and (5) show resumed growth on the truncated crystals into the overlying rocks. These observations indicate that the cumulates of the inward-growing chamber floor were almost entirely solid at the time of the erosion and deposition of these layers. To reconcile these two sets of contrasting observations, we propose that only a few metres of the uppermost portion of the cumulate pile of the Upper and Critical Zones of Bushveld magma chamber were mushy, with other underlying rocks being almost totally solidified. The mushy layer tends to be thin and locally even non-existent (e.g., in magnetitite layers) due to high efficiency of primary adcumulus growth at the crystal-liquid interface of the temporary chamber floor. The solid chamber floor results from thermochemical erosion of the entire mushy layer by magmas replenishing the evolving chamber. The study implies that such petrogenetic processes as the hydrodynamic sorting of crystals and reactive porous flow may only operate within this very thin mushy portion of the cumulate pile and cannot therefore be responsible for any large-scale petrological features (e.g., platinum reefs, massive chromitites, stratiform anorthosite layers) of the Bushveld Complex.

Several recent petrogenetic models propose that layers of stratiform chromitites in the Bushveld Complex are produced by mechanical separation of chromite from co-existing silicate minerals (e.g., olivine, orthopyroxene, plagioclase) within crystal-rich slurries. One feature in common for the slurry models is that they imply that interstitial liquid in stratiform chromitites solidifies underneath a crystal pile ranging from 20 to 100 m thick. This means that the interstitial liquid would not be able to chemically communicate with the resident melt overlying the crystal pile. Equilibrium crystallization of interstitial melt within chromitites will, therefore, produce interstitial plagioclase with an average Ca/Na ratio similar to that of the interstitial liquid itself, i.e., plagioclase must be evolved in composition (-40-50% An-content). In contrast to this prediction, extensive microprobe data show that interstitial plagioclase in Bushveld chromitites has a rather primitive composition (-60-80% An-content) that is close to that of cumulus plagioclase (-70-80% An-content) from adjacent norites and anorthosites that are produced by fractional crystallization. This implies that interstitial plagioclase is not a product of equilibrium crystallization within a deeply buried chromite-rich pile. Modification of interstitial plagioclase composition by various additional processes (fractional crystallization, selective diffusion, reactive melt flow, etc.) are considered but shown to be inconsistent with field, textural or chemical observations. We conclude that either slurry models for the origin of stratiform chromitites are incorrect and should be abandoned or must be substantially modified to address the issue of the high An-content of interstitial plagioclase in stratiform chromitites.

Several recent studies have argued that large, long-lived and molten magma chambers may not occur in the shallow Earth’s crust. Here we present, however, field-based observations from the Bushveld Complex that provide evidence to the contrary. In the eastern part of the complex, the magmatic layering continuously drapes across a ~ 4-km-high sloping step in the chamber floor. Such deposition of magmatic layering implies that the resident melt column was thicker than the stepped relief of the chamber floor. Prolonged internal differentiation within this thick magma column is further supported by evolutionary trends in crystallization sequence and mineral compositions through the sequence. The resident melt column in the Bushveld chamber during this period is estimated at > 5-km in thickness and > 380,000 km 3 in volume. This volume of magma is three orders of magnitude larger than any known super-eruption in the Earth’s history and is only comparable to the extrusive volumes of some of Earth’s large igneous provinces. This suggests that super-large, entirely molten, and long-lived magma chambers occur, at least occasionally, in the geological history of our planet. Therefore, the classical view of magma chambers as ‘big magma tanks’ remains a viable research concept for some of Earth’s magmatic provinces.

Methodology, photomicrographs of major rock units of the Sudbury Igneous Complex, and a data table.

A recent re-interpretation of the Bushveld Complex and other layered intrusions as stacks of randomly emplaced, amalgamated sills is mostly fuelled by finding of zircon ages that are not getting progressively younger from the base upwards, as expected from a classical model for the formation of layered intrusions. Rather, they display several reversals from older to younger ages and vice-versa with moving up-section through the layered intrusions. Here, we show that the reported zircon ages are at odds with the relative ages of rocks as defined by cross-cutting relations in potholes of the Bushveld Complex. This indicates that interpretation of the zircon isotopic data as the emplacement age of the studied rocks/units is incorrect, making a new emplacement model for layered intrusions baseless. This conclusion is further buttressed by the phase equilibria analysis showing that regular cumulate sequences of layered intrusions are not reconcilable with a model of randomly emplaced sills. In this model, the late sills are free to intrude at any stratigraphic position of the pre-existing rocks, producing magmatic bodies with chaotic crystallization sequences and mineral compositional trends that are never observed in layered intrusions. There are thus no valid justifications for the re-evaluation of the current petrological model of the Bushveld Complex and other layered intrusions as large, long-lived and largely molten magma chambers. A fundamental implication of this analysis is that the current high-precision U-Pb TIMS ages from layered intrusions are inherently unreliable on the scale of several million years and cannot therefore be used for rigorous estimations of the timing of crystallization, duration of magmatism, and cooling of these intrusions.

The magmatic stratigraphy of the Sudbury Igneous Complex (Canada) is thought to have resulted from closed-system differentiation of an initially homogeneous impact melt sheet. The topography of its upward-growing chamber floor is therefore thought to have been planar and subhorizontal. However, we report on the discovery of a large pothole-like depression (~300 m in depth and ~550 m in width) in the chamber floor of this complex. The depression has been revealed through two-dimensional mapping of igneous layering that is defined by systematic vertical changes in cumulus assemblages and bulk rock chemistry. Although the formation of the depression by syn- to post-magmatic folding and/or slumping of chamber floor cumulates cannot be completely excluded, we favor an alternative explanation that follows from the recent recognition that the Sudbury Igneous Complex melt sheet crystallized concurrently from the floor and roof inward. The roof sequence was subsequently disrupted and collapsed as large discrete blocks onto the floor sequence. This may have resulted in local irregularities in topography of the upward-growing chamber floor so that crystal deposition onto and between the neighboring blocks produced pothole-like depressions. The phenomenon of physical disruption of roof sequences appears to provide a reasonable explanation for the common lack of the rocks that grew from the roof downward in layered intrusions.

The classical paradigm of the 'big magma tank' chambers in which the melt differentiates, is replenished, and occasionally feeds the overlying volcanos has recently been challenged on various grounds. An alternative school of thought is that such large, long-lived and largely molten magma chambers are transient to non-existent in Earth's history. Our study of stratiform chromitites in the Bushveld Complex-the largest magmatic body in the Earth's continental crust-tells, however, a different story. Several chromitites in this complex occur as layers up to 2 m in thickness and more than 400 kms in lateral extent, implying that chromitite-forming events were chamber-wide phenomena. Field relations and microtextural data, specifically the relationship of 3D coordination number, porosity and grain size, indicate that the chromitites grew as a 3D framework of touching chromite grains directly at the chamber floor from a basaltic melt saturated in chromite only. Mass-balance estimates imply that a few km thick column of this melt is required to form each of these chromitite layers. Therefore, an enormous volume of melt appears to have been involved in the generation of all the Bushveld chromitite layers, with half of this melt being expelled from the magma chamber. We suggest that the existence of thick and laterally extensive chromitite layers in the Bushveld and other layered intrusions supports the classical paradigm of big, albeit rare, 'magma tank' chambers.

The Bushveld Complex in South Africa shows spectacular examples of regional and local magmatic erosion of the floor cumulates by new melt batches that replenished the evolving magma chamber. Field observations indicate that, at some stratigraphic levels, at least, 15–20 m of pre-existing floor cumulates, often nearly monomineralic in composition (e.g., anorthosite or orthopyroxenite) were completely removed on a regional scale. What was the major agent of this erosion – (partial) melting or dissolution of the floor cumulates – remains poorly understood. Thermal melting appears to be a poor candidate because normal basaltic melts (∼1220–1260°C) cannot heat up the cumulates up to their melting temperature (∼1400–1500°C). We explored, therefore, the possibility of dissolution of these high-temperature-melting cumulates by slightly superheated (15°C above the liquidus) basaltic-andesitic melts that recharged the chamber and spread out laterally along its floor as basal flows. This was done using a freely available thermodynamic tool for phase equilibria modeling of open magmatic systems – the Magma Chamber Simulator. Our thermodynamic modeling shows that the superheated melts can digest up to 4.5–8.0 wt% of the bulk floor cumulates without inducing crystallization of the melts, despite them being much colder than the liquidus temperatures of these cumulates. This is equivalent to regional erosion of 15–24 m of the floor cumulates, given a basal melt layer of about 350 m thick. We conclude that the regional erosion of the high-temperature-melting floor cumulates in the Bushveld chamber has been mostly controlled by their chemical dissolution by replenishing superheated melts.

The vertical growth rate of basaltic magma chambers remains largely unknown with available estimates being highly uncertain. Here, we propose a novel approach to address this issue using the classical Skaergaard intrusion that started crystallizing from all margins inward only after it had been completely filled with magma. Our numerical simulations indicate that to keep the growing Skaergaard magma chamber completely molten, the vertical growth rate must have been on the order of several hundreds to a few thousands of meters per year, corresponding to volumetric flow rates of tens to hundreds of cubic kilometers per year. These rates are several orders of magnitude higher than current estimates and were likely achieved by rapid subsidence of the floor rocks along faults. We propose that the Skaergaard is a plutonic equivalent of supereruptions or intrusions that grow via catastrophically rapid magma emplacement into the crust, producing totally molten magma chambers in a matter of a few months to dozens of years.

Orthopyroxenite cumulates throughout the Critical Zone of the Bushveld Complex commonly contain prominent euhedral crystals of bottle-green augite, typically around 1 cm in size and surrounded by “haloes” of nearly pure plagioclase. On close examination, the augite grains can be seen to be oikocrysts, containing extensively resorbed chadacrysts of orthopyroxene, indicating an effective peritectic reaction relationship between the two pyroxenes. A detailed textural study of one such layer from the UG3 Unit in the Eastern Bushveld reveals some distinctive grain-scale features, including the presence of extensively 3D-interconnected chains of thousands of cumulus chromite grains that surround the orthopyroxene grains and extend through the augite oikocrysts. The high degree of interconnectivity is remarkable in view of the relatively low (5 vol.%) modal proportion of chromite in the sample. Other noteworthy features include zoning of the oikocrysts towards higher incompatible element contents in the rims and the presence of the plagioclase haloes around the augite oikocrysts. These haloes are essentially narrow zones devoid of orthopyroxene inclusions developed within large plagioclase oikocrysts that overgrow all the other phases. We propose a mechanism whereby the plagioclase oikocrysts grow early in communication with the main magma body within an initial crystal mush of orthopyroxene and chain-textured chromite. The clinopyroxene oikocrysts grew in the remaining pore space in such a way that dissolution of orthopyroxene occurred within a narrow few-mm wide chemical boundary layer ahead of the advancing oikocryst margin. The relative rates of crystallisation and dissolution were controlled by limited chemical diffusion through the boundary layer such that orthopyroxene grains more than a few grain diameters away showed no reaction at all. The plagioclase haloes developed as a result of continuing growth of the plagioclase oikocrysts to overtake the growing pyroxene oikocrysts, locking in the orthopyroxene-depleted boundary layer, and preserving euhedral cumulus morphologies away from the pyroxene oikocrysts. This texture represents a circumstance whereby oikocrysts in the same rock develop at different overlapping stages: plagioclase oikocrysts forming early in diffusive connection with the magma column, and clinopyroxene forming peritectic poikilitic textures within residual liquid pockets chemically isolated from the overlying magma body.

The Upper Zone of the Bushveld Complex has long been known to have formed from a major influx of magma into the chamber that caused large-scale erosion of the chamber floor cumulates. The most dramatic manifestations of this process are two major gap areas (Northern and Southern) in the western Bushveld Complex in which the Upper Zone appears to have eroded away the underlying cumulates down to the very base of the Complex. However, due to almost complete lack of outcrops in the gap areas, no direct field observations have ever been reported to confirm the transgressive nature of the Upper Zone. Here, we present for the first time such observations from the Kameelhoek chromite mine located at the margin of the Northern Gap. In the open pit we have documented several transgressive depressions (up to 40 m in width) in the orthopyroxenite and chromitites of the Lower Critical Zone that are filled in with magnetite gabbro of the Upper Zone. The magnetite gabbro is chilled against the sidewalls of the depressions, forming glassy and fine-grained textured rocks with plagioclase laths arranged in radial clusters. Mineralogically and chemically, the magnetite gabbro correlates with cumulates from the lowermost part of the Upper Zone at its normal position in the complex. Three major points that have emerged from this study are: (1) the Critical Zone has been eroded away by magma that was parental to the Upper Zone, (2) this eroding magma was not the one that initiated formation of the Pyroxenite Marker, but rather the evolved melt that replenished the chamber at some later stage, and (3) the melt was phenocryst-free and likely derived from a deep-seated staging chamber. Our study thus supports a recent notion that even during the formation of the Upper Zone, the Bushveld chamber had still been operating as an open system that was replenished by melts from deeper magma sources.

The formation of some Earth’s monomineralic igneous rocks appears to be prohibited by constraints imposed by liquidus phase-equilibria on evolution of mantle-derived magmas. Yet, these rocks exist as stratiform layers in many mafic-ultramafic intrusions. One conspicuous example is monomineralic anorthosites in the Bushveld Complex that occur as stratiform layers up to hundreds of kilometres in length. Such monomineralic anorthosites appear to require parental melts saturated in plagioclase only but where and how to produce these melts remains a contentious issue. Here we argue that they are likely sourced from deep-seated magma reservoirs. In response to pressure reduction, these ascending melts become first superheated and then saturated in plagioclase after stalling and cooling in shallow-level chambers. Adcumulus growth of plagioclase from such melts at the chamber floor results in the formation of monomineralic anorthosites. We propose that stratiform layers of monomineralic anorthosites in layered intrusions are products of the chamber replenishment by melts whose saturation in plagioclase as a single liquidus phase is triggered by their transcrustal ascent towards the Earth’s surface.

Basaltic magma chambers are best exemplified by layered intrusions – fossilized natural laboratories that historically constrain most fundamental principles of igneous petrology. Progressive fractional crystallization of basaltic melts in layered intrusions results in successive appearance of 5-10 liquidus minerals. Most minerals appear in the stratigraphy of layered intrusions via two distinct stages. They first emerge as ‘oikocrysts’ – large, irregularly-shaped and interstitial crystals that form in a crystal-liquid framework and then as ‘primocrysts’ – relatively small, idiomorphic and cumulus crystals that crystallize from a main magma body. The transition from oikocrysts to primocrysts is commonly marked by the following textural and chemical features: (a) a similarity in chemical composition of the last-forming oikocrysts and the first-forming primocrysts, (b) slow arrival, modal overabundance and finer-than-normal size of the first primocrysts, (c) a low abundance of trapped liquid in rocks with the first-forming primocrysts, (d) a step-like increase in mineral dihedral angle in rocks slightly below the level of the primocryst arrival, and (e) the rims on plagioclase primocrysts that are identical in composition to cores of plagioclase primocrysts at the first appearance of new liquidus phases higher up in the section. The phase equilibria control on the cumulus stratigraphy of layered intrusions require large, long-lived and largely molten magma chambers and, therefore, their origin is not compatible with several recent concepts that deny the existence of such reservoirs in the Earth’s crust. We therefore suggest that the systematic changes associated with the arrival of cumulus phases in layered intrusions are best explained by a traditional concept that considers layered intrusions as large initially crystal-free melt bodies that gradually lose heat and crystallize from margins inwards through thin solidification fronts. The entire magma body during its internal evolution is thought to be kinetically supercooled (i.e. kept slightly below its liquidus temperature) with respect to cumulus phases that crystallize at the margins. Each new cumulus phase arrives on the liquidus with some kinetic delay and nucleates heterogeneously against pre-existing crystals at the floor, roof and sidewalls of basaltic magma chambers. This implies that internal differentiation of basaltic magma chambers mostly occurs through convective separation of evolved liquid from in situ growing crystals and its mixing with the main magma body.

Plutonic mafic complexes are composed of cumulates in which minerals mostly occur in cotectic proportions. This is consistent with a concept that basaltic magma chambers predominantly crystallize in situ from margins inward. However, cumulates with two (or more) minerals in proportions that are at odds with those expected from liquidus phase equilibria also locally occur in these complexes. Such non-cotectic cumulates are commonly attributed to either mechanical separation of minerals crystallizing from the same parental magma or mechanical mixing of minerals originating from different parental magmas. Here we introduce a novel concept that does not require any of these processes to produce non-cotectic cumulates. The model involves melts that start crystallizing upon their cooling, while ascending along feeder conduits from deep staging reservoirs toward the Earth’s surface. Depending on the degree of cooling, the melts become successively saturated in one, two, and more liquidus phases. Given that most crystals are kept in suspension, the resulting magmas would contain a cargo of equilibrium phenocrysts in notably non-cotectic proportions. The replenishment of basaltic chambers developing through in situ crystallization by such magmas is likely responsible for the occasional formation of non-cotectic cumulates in plutonic mafic complexes.

We describe an impressive ~55 m high outcrop from the Pilanesberg Platinum Mine open pit, located in the North-Western Bushveld Complex. The outcrop exposes the complete two-dimensional structure of three Merensky Unit potholes that cut several metres down into the underlying footwall anorthosites. The transgressive field relationships are interpreted to have resulted from thermochemical erosion of the footwall rocks by new pulses of magma replenishing the chamber and resulting in incremental growth of the Bushveld Complex.

The Merensky Reef (MR) of the Bushveld Complex is known for containing the lion’s share of the world’s reserves of platinum group elements (PGE). It is commonly described as a single, relatively thin layer (or package) of chromite- and sulphide-bearing orthopyroxenite hosted by almost barren norite and/or anorthosite. Here, we present a drill-core section that contains; however, four separate MR orthopyroxenite layers hosted by mottled anorthosites. Each of these orthopyroxenites is distinguished through occurrence/absence of marginal chromitite seams, grain size, petrographic composition, and PGE abundance. We interpret them as multiple undercutting MR that occurs as sill-like apophyses extending laterally from a pothole margin into the footwall. The undercutting MR is attributed to in situ crystallization within sheet-like melt-filled cavities at potholes margins, which are produced through thermal and/or chemical erosion of the footwall rocks by replenishing magmas. This finding indicates that the MR-forming event involved several distinct pulses of replenishing magmas whose crystallization products were mostly eroded away in the stratiform portions, but remained locally preserved in apophyses within footwall rocks. The mineralized MR formed a few meters below a temporary chamber floor is indicative of strong thermal/compositional convection in the basal layer of the Bushveld’s chamber. The convection was able to deliver large volumes of fresh magma for the scavenging of PGE by sulphide and chromite growing within the sheet-like cavities of the MR potholes.