The Rustenburg Layered Suite formed as a stack of mush with transient magma chambers

Layered mafic intrusions such as the Rustenburg Layered Suite (RLS) have traditionally been interpreted as products of differentiation within large, melt-dominated magma chambers via fractional crystallization (FC) and assimilation–fractional crystallization (AFC). An emerging view instead emphasizes transcrustal plumbing systems dominated by crystal mush with only transient, localized melt-rich zones. The key question addressed is whether the RLS formed primarily through processes in melt-dominated chambers or through assimilation coupled with batch (equilibrium) crystallization (ABC) during dynamic magma transport through the crust. The study proposes and tests ABC-based, mush-dominated models for the RLS, contrasting them with classical AFC models, to explain the stratigraphy, geochemistry, isotopes, and mineralogy of the suite.

Classical igneous petrology since Bowen emphasized FC, later incorporating crustal assimilation (AFC) and deep processing in MASH zones. Layered intrusions have often been modeled as crystallization sequences in large chambers (e.g., Skaergaard), a view challenged by recent evidence for mush-dominated systems and out-of-sequence layering. For the RLS, previous work invoked U-type (~12–14 wt% MgO) and A-type (~7–8 wt% MgO) parental magmas preserved as marginal sills (B1, B2, B3). However, primitive minerals and the Basal Ultramafic Sequence require a komatiitic parent (>~19 wt% MgO), with B1–B3 likely representing liquids complementary to cumulates rather than true parents. Prior studies suggested extensive upper- and mid-crustal contamination of komatiite (>~40%) to produce B1 and B2–B3 characteristics, and isotope data (Sr–Nd–O) support significant crustal input, while alternative models invoking subcontinental lithospheric mantle (SCLM) contributions have been questioned by newer Hf–O constraints. Thermal and geochronological studies also support transient magmatism and mush processes rather than long-lived chambers.

The authors used alphaMELTS thermodynamic modeling, supplemented by isotopic mass-balance calculations, to evaluate transcrustal assimilation scenarios for komatiitic parent magma interacting with crustal assimilants. Endmembers (komatiite; upper-, middle-, lower-crust compositions) were set from regional data (Pretoria Supergroup, Vredefort Dome, Limpopo Belt; global analogs for lower crust). Three scenarios were tested: (1) One-stage, isenthalpic ABC in the upper crust at 0.2 GPa, producing crystal–liquid suspensions that remain at internal equilibrium during forced convection/turbulent flow and then separate into cumulates and supernatant liquid upon emplacement; (2) Two-stage ABC: initial ABC in the mid-crust at 0.45 GPa (∼390 °C crust), with most solids retained at depth, and ascent of the supernatant liquid that undergoes further batch crystallization at upper-crustal levels to yield mafic cumulates and complementary liquids; (3) Lower-crustal AFC at 1 GPa (∼770 °C crust) to generate an Fe-rich parent for the Upper and Upper Main Zone (UUMZ), followed by closed-system fractional crystallization in an upper-crustal sill-like chamber with minor recharge. Modeling included isenthalpic assimilation increments, equilibrium between melts and solids, and options to retain or discard solids (ABC vs AFC). Subsequent isobaric crystallization at 0.2 GPa simulated ascent/emplacement cooling. Trace elements were treated using variable partition coefficients (lattice strain for REE in cpx/feldspar; literature constants otherwise). Isotopic models tracked Sr–Nd–O systematics based on mixing proportions and alphaMELTS outputs. Key parameters and compositions are in Supplementary Tables/Figures. Pressures/temperatures: upper crust 0.2 GPa (~300 °C assimilant), mid-crust 0.45 GPa (~390 °C), lower crust 1 GPa (~770 °C). Oxygen fugacity during upper-crustal cooling used the FMQ buffer.

• Most of the RLS below the Upper Main Zone (~two-thirds of total thickness) can be reproduced by one- or two-stage ABC processes forming crystal-rich mushes; marginal sills record complementary supernatant liquids.
• Scenario 1 (upper-crust ABC, 0.2 GPa): After assimilation of upper crust, modeled cumulates match Lower Zone (LZ) and Lower Critical Zone (LCZ) ultramafics and the B1 sill compositions. Quantitatively: LZ dunite after ~17.4% assimilation (adcumulate with ~4.5% trapped liquid); LZ harzburgite after ~22.5% assimilation (mesocumulate with ~17.8% trapped liquid); LZ/LCZ pyroxenite after ~27–34% assimilation (~15% trapped liquid). B1 magma modeled as ~78% liquid + ~22% solids equivalent to LCZ pyroxenite. Trace element patterns for modeled cumulates and B1 match observations. Under equivalent settings, AFC fails to reproduce common olivine+orthopyroxene coexistence (granular harzburgites) and whole-rock compositions.
• Scenario 2 (two-stage ABC): Komatiite plus mid-crust assimilation at 0.45 GPa yields Bulk-1 (∼21% assimilation for Upper Critical Zone norites; ∼24% for Lower Main Zone gabbronorites), cooling to ~1240–1250 °C. Retaining ~90–97% of solids at depth creates hidden ultramafic cumulates akin to LZ pyroxenites; the remaining solids+liquid (Bulk-2) ascend and at 0.2 GPa separate into mafic cumulates and liquids resembling B2 (for UCZ norites) and B3 (for LMZ gabbronorites). Modeled crystallization: UCZ norite after ~40% crystallization at 1181 °C (adcumulate with ~5% trapped liquid); ejected liquid (~5% solids) matches B2 (except some Rb–Th depletion). LMZ gabbronorite after ~63.9% crystallization at 1130 °C (mesocumulate with ~3% trapped liquid); B3 ejected at ~34.2% crystallization with ~42% solids, consistent with coarser grains and lower trace-element abundances.
• Scenario 3 (lower-crust AFC + FC): An Fe-rich parent for the UUMZ is produced by ~43.5% AFC contamination at 1 GPa, followed by ~24% fractional crystallization during ascent. Subsequent closed-system fractional crystallization in an upper-crustal chamber reproduces the observed paragenetic sequence, mineral modes, and mineral compositional trends down to ~21% residual melt. Cyclic mineral composition reversals (e.g., between cycles V–VI) can be matched by a small (~1.2%) recharge of initial parent magma. Simple FC does not reproduce the abundant magnetitite layers; additional processes (e.g., double-diffusive convection or liquid immiscibility) are likely required.
• Isotopes: Modeled transcrustal assimilation trends reproduce the observed inverse correlation between initial (87Sr/86Sr) and εNd for RLS cumulates and marginal sills. High δ18O (~7.1‰ average) matches crustal assimilant signatures and is inconsistent with an eclogite-bearing lithospheric mantle source alone for the observed signatures. Restricted isotopic ranges in B1–B3 sills are consistent with their being complementary liquids to deposited cumulates.
• Field/stratigraphic coherence: The volumes and distribution of B1 sills are consistent with their role as supernatant liquids to the LZ–CZ pyroxenites; out-of-sequence layering, chromitite transgressive contacts, and local variability are naturally accommodated by emplacement of independent slurry batches.
• Ore implications: Chromitite and PGE-bearing sulfide reefs can form from crystal-laden mushes without a large chamber. Modeled Critical Zone pyroxenitic mushes include ~2% chromite, predicting chromitite thicknesses broadly consistent with observations; small degrees of sulfide oversaturation during crustal assimilation of B1-type slurries can account for Merensky Reef grade–thickness. Viscous segregation in laterally flowing crystal slurries explains chromitite concentration and associated anorthosite/norite formation.

The modeling demonstrates that much of the RLS can be generated by assimilation–batch crystallization during dynamic, forced-convective magma transport through a transcrustal network of dikes and sills, rather than by prolonged differentiation in a melt-dominated chamber. Hot, low-viscosity komatiitic magmas can assimilate large proportions of crust, precipitating abundant suspended crystals that remain in equilibrium with the melt and are later deposited as mushy macrolayers. This framework explains: (a) coexistence of olivine and orthopyroxene cumulates forbidden by AFC peritectic relations, (b) local to regional stratigraphic variability and out-of-sequence layers, (c) isotopic heterogeneity at mineral scales, (d) the complementary nature of marginal sills, and (e) density-filtered vertical arrangement of ultramafic to mafic units resembling a classic stratigraphy without requiring progressive in-situ FC. For the UUMZ, lower-crustal AFC followed by FC within a transient, relatively melt-rich pocket accounts for mineral modes and cyclic compositional reversals, with minor recharge events. This mush-dominated perspective broadens the potential settings for forming stratiform chromitite and PGE reefs, implying that similar deposits could develop in smaller or irregular intrusions without large, open magma chambers.

This study provides a quantitative, thermodynamically constrained framework in which most of the RLS formed by one- or two-stage ABC processes that produced crystal-rich mushes emplaced as independent macrolayers; the marginal sills represent complementary supernatant liquids. Only the uppermost third (Upper and Upper Main Zone) requires lower-crustal AFC to generate an Fe-rich parent, followed by fractional crystallization with minor recharge in a transient melt-rich pocket. The results challenge the necessity of large, open magma chambers for layered intrusion formation and for associated chromitite and PGE ore genesis, expanding viable geometries and scales of magmatic systems that can produce such deposits. Future work should further resolve processes forming the abundant magnetitite layers (e.g., testing double-diffusive convection or immiscibility), refine endmember compositions and thermal regimes of the lower crust beneath Bushveld, obtain higher-resolution geochronology to test out-of-sequence emplacement, and apply the ABC-based approach to other layered intrusions and smaller ore-bearing intrusions.

The fractional crystallization model for the UUMZ does not reproduce the numerous magnetitite layers, implying additional processes not fully constrained here (e.g., double-diffusive convection, liquid immiscibility). Some isotopic comparisons (e.g., sparse B3 data) are limited by sample availability. Compositions and temperatures of crustal assimilants, particularly for the lower crust, are inferred from regional analogs and global datasets, introducing uncertainty. Certain partition coefficients and phase identifications required adjustments (e.g., distinguishing high-Ca clinopyroxene) that may affect trace-element fits. The linkage between the final residual liquids and Rooiberg felsites remains unresolved and is outside the modeling scope.