Monomineralic layers in mafic intrusions are critical sources of rare metals; it is estimated that the Bushveld's magnetitite layers host 45% of global reserves of vanadium (Latypov et al., 2024). The prevailing theory is that magnetitite layers form by in situ crystallisation on a ‘hardground' chamber floor. However, this requires unrealistically frequent magma recharge. We hypothesise that the magnetitite layers formed via percolation of buoyant reactive fluids across sharp chemical potential gradients within a late-stage crystal-dominated mush. We performed experiments to explore mineral growth across a sharp chemical potential gradient in a simplified MgO-FeO-SiO2 system. Our results recreate textures observed in the Bushveld's magnetitite layers and suggest that small degree melting is required for mineral growth.

We performed petrological experiments, nominally dry and with 1 wt% H2O at 800-1000˚C, 0.5-1.5 GPa, in which reaction rims of orthopyroxene (opx) grow between olivine and quartz. We expand upon previous work in this system by Incel et al., 2022, and references therein. We observed that all olivine-quartz interfaces grew opx rims at 1000˚C but no opx growth was observed at 800˚C, suggesting incipient melting of quartz is required for mineral growth to occur. We present a unified framework for understanding the formation of major element zoning across opx rims via advection in a dissolution-precipitation mechanism (e.g. Borg et al, 2014). We observed that opx growth is not confined to the immediate contact between olivine and quartz; opx also grew along quartz grain boundaries and within pore space in the olivine, exhibiting similar textures to disseminated magnetite in the silicate rocks overlying the Bushveld's magnetitite layers.