d="p-2">Drill core from a fence of five boreholes located in the vicinity of Roossenekal, in the eastern limb of the Bushveld Complex, South Africa, provides complete coverage of what is probably one of the thickest and most well-developed sections of the Upper Zone. The lowermost portions of the Rooiberg Group felsites that overlie the Upper Zone, and the uppermost portion of the Upper Main Zone, which forms the footwall, were also intersected. Field relationships from floor-attached domes (eastern limb) and gap areas (western limb) reveal a marked angular unconformity at the base of the Upper Zone. The Upper Zone represents a discrete intrusive event and should not be interpreted as a continuum, or the fractionated residue of the underlying components of the Rustenburg Layered Suite (RLS). A new subdivision of the Upper Zone into five subunits (A through E) is proposed on the basis of the cumulus mineralogy and whole-rock geochemistry, the latter based on samples at an average vertical spacing of 20 m throughout the column at Roossenekal. The separation of subzones A and B is based on our geochemical data, as is the identification of subzone E, but recognition of subzones C (olivine) and D (apatite) is broadly consistent with the detailed mapping and stratigraphic column presented by d="xref-ref-55-1" class="xref-bibr">Von Gruenewaldt (1973). Subunits A through D reveal macro-layering for which the intrusion is so well known. More than 40 individual Ti-magnetite layers were identified in the drill-core, which together with the numerous anorthosite layers, are the most prominent feature of this part of the column. The principal lithology is a somewhat monotonous sequence of ferrogabbro or magnetite-olivine gabbro. The absence of Ti-magnetite layers and anorthosite from subzone E, which is dominated by diorite with <5 modal % Ti-magnetite is significant.
d="p-3">The whole-rock geochemical data presented here is not consistent with a closed system fractionation hypothesis for subunits A through D, despite this being widely accepted in recent contributions. An upward differentiation trend is preserved only in subunit E. The incremental build up of cumulates from successive pulse of magma is not only consistent with our views on the underlying layered cumulates (periods of closed system differentiation are rare) but is also consistent with field relationships, including each of the principal limbs representing individual sill-like features. The Upper Zone crystallized from batches of iron-rich tholeiitic magma derived from a deep staging chamber. Magma was injected as near-lateral flows, probably in a southward direction in the eastern limb. The introduction of evolved magma into a shallow crustal environment resulted in the early crystallization of Fe-Ti oxides. Ti-magnetite layers crystallized from oxide-rich slurries intruded near the floor of the chamber. Fe-Ti oxide pipes formed from slurries forced downward in areas of structural weakness. The regular, albeit step-like, upward change in composition of the Ti-magnetite layers identified by d="xref-ref-28-1" class="xref-bibr">Molyneux (1974) and others is ascribed to mixing of crystal-rich magma slurries with increasingly differentiated resident magma: the lowermost layers (and pipes) contain the highest concentrations of V2O5. The spatial association of Ti-magnetite layers and anorthosite is emphasized and we suggest the latter formed as a consequence of disequilibrium partial melting of earlier formed silicate cumulates due to heat associated with new pulses of magma.