Petrogenesis of magnetite-rich layers in the Bushveld Complex, South Africa

The Upper Zone (UZ) of the Bushveld Complex, South Africa, hosts 21 prominent magnetite (mt) ± ilmenite (ilm) and nelsonite (mt + ilm + apatite) layers, and at least five thinner (<10 cm) discrete layers. Layer thickness ranges from 10’s of cm to two meters in the Main Magnetite Layer and nearly ten meters in Layer 21. The magnetite layers almost always exhibit a sharp planar to undulatory contact with underlying anorthosite and may be overlain by gabbro or anorthosite, typically with a graded contact marked by an increased abundance of plagioclase. Several models have been proposed for the formation of the magnetite layers within the silicate packages throughout the stratigraphy, including crystal settling and sorting by density, liquid immiscibility, fractional crystallization, changes in oxygen fugacity and/or pressure, and magma recharge. Additionally, nucleation and growth of individual magnetite layers may occur either by in situ crystallization or mineral accumulation. The models applied to the larger-scale occurrence of the magnetite layers may also be attributed to individual layer formation, including rapid disequilibrium crystallization, fractional crystallization, and magma recharge, as well as possible intermittent convection events. Fifteen discrete titanomagnetite layers and adjacent silicate layers and a cm-scale profile through a single 24 cm thick magnetite layer from the Eastern Limb were studied texturally and by trace element analysis by LA-ICP-MS. Compatible elements in magnetite including Cr, V, Ni, Co, and Mg decrease with stratigraphic height relative to the Main Magnetite Layer and incompatible elements including Ga, Sc, Nb, Ta, Hf, and Mo increase with stratigraphic height. These trends suggest that fractional crystallization contributed to the overall geochemical trends in the Upper Zone. However, reversals to more primitive abundances of Cr, the most highly compatible element in magnetite, at Layers 6, 7, 13, and 14, indicate a magma mixing event. These mixing events do not necessarily suggest an open-system magma recharge event and may indicate the breakdown of double-diffusive boundary layers produced by closed-system fractional crystallization. Other reversals marked by relative increases in compatible elements and relative decreases in incompatible elements at Layers 3, 8, 11, and 12 are not associated with reversals in Cr. The layers associated with reversals in Cr may represent magma recharge events, while reversals at other layers that are not associated with Cr may represent the breakdown of boundary layers produced by fractional crystallization. At Layers 4-5, 6-7, and 10-12, there are no reversals in most elements; these may represent closed-system fractional crystallization. These observations suggest at least four recharge events in the Upper Zone but do not indicate that each magnetite layer formed through the addition of primitive magma. Replacive symplectites (plag + opx ± cpx ± olivine, mt primocryst) in Subzone A (UZb) and at the transition into Subzone C (UZc) likely represent the onset of liquid immiscibility. Additionally, plagioclase-hosted melt inclusions are present in all gabbroic samples in the UZc, after apatite appears in the mineral assemblage. While liquid immiscibility may have produced the magnetite layers in UZc, geochemical data, including the absence of a relative increase of HFSE, Ti, Mn, and P in the magnetite layers in UZa suggest that these magnetite layers did not form through immiscibility. Instead, the geochemical trends expected in magnetite formed through immiscibility are present in the silicate rocks associated with Layers 5, 6, 8. The vertical profile through Layer 10 suggests a model of in situ crystallization with intermittent convection events may be the dominant process of individual magnetite layer formation. A model of magma replenishment may also accommodate the geochemical variation through the layer, though a single model does not fit all of the inflection points in the most compatible elements. Overall, systematic variation throughout the vertical profile in many elements demonstrates crystallization processes that cannot be wholly explained by crystal settling and sorting by density.