Density structure of the cratonic mantle in Southern Africa: 2. Correlations with kimberlite distribution, seismic velocities, and Moho sharpness

We present a new regional model for the depth-averaged density structure of the cratonic lithospheric mantle in southern Africa constrained on a 30' x 30' grid and discuss it in relation to regional seismic models for the crust and upper mantle, geochemical data on kimberlite-hosted mantle xenoliths, and data on kimberlite ages and distribution. Our calculations of mantle density are based on free-board constraints, account for mantle contribution to surface topography of ca. 0.5-1.0 km, and have uncertainty ranging from ca. 0.01 g/cm(3) for the Archean terrains to ca. 0.03 g/cm3 for the adjacent fold belts. We demonstrate that in southern Africa, the lithospheric mantle has a general trend in mantle density increase from Archean to younger lithospheric terranes. Density of the Kaapvaal mantle is typically cratonic, with a subtle difference between the eastern, more depleted, (3.31-3.33 g/cm(3)) and the western (3.32-3.34 g/cm(3)) blocks. The Witwatersrand basin and the Bushveld Intrusion Complex appear as distinct blocks with an increased mantle density (3.34-3.35 g/cm(3)) with values typical of Proterozoic rather than Archean mantle. We attribute a significantly increased mantle density in these tectonic units and beneath the Archean Limpopo belt (3.34-3.37 g/cm(3)) to melt-metasomatism with an addition of a basaltic component. The Proterozoic Kheis, Okwa, and Namaqua-Natal belts and the Western Cape Fold Belt with the late Proterozoic basement have an overall fertile mantle (ca. 337 g/cm(3)) with local (100-300 km across) low density (down to 3.34 g/cm(3)) and high-density (up to 3.41 g/cm(3)) anomalies. High (3.40-3.42 g/cm(3)) mantle densities beneath the Eastern Cape Fold belt require the presence of a significant amount of edogite in the mantle, such as associated with subducted oceanic slabs. We find a strong correlation between the calculated density of the lithospheric mantle, the crustal structure, the spatial pattern of kimberlites, and their emplacement ages. (1) Blocks with the lowest values of mantle density (ca. 3.30 g/cm(3)) are not sampled by kimberlites and may represent the "pristine" Archean mantle. (2) Young (<90 Ma) Group I kimberlites sample mantle with higher density (3.35 +/- 0.03 g/cm(3)) than the older Group II kimberlites (3.33 +/- 0.01 g/cm(3)), but the results may be biased by incomplete information on kimberlite ages. (3) Diamondiferous kimberlites are characteristic of regions with a low-density cratonic mantle (3.32-3.35 g/cm(3)), while non-diamondiferous kimberlites sample mantle with a broad range of density values. (4) Kimberlite-rich regions have a strong seismic velocity contrast at the Moho, thin crust (35-40 km) and low-density (3.32-3.33 g/cm(3)) mantle, while kimberlite-poor regions have a transitional Moho, thick crust (40-50 km), and denser mantle (3.34-3.36 g/cm(3)). We explain this pattern by a lithosphere-scale (presumably, pre-kimberlite) magmatic event in kimberlite-poor regions, which affected the Moho sharpness and the crustal thickness through magmatic underplating and modified the composition and rheology of the lithospheric mantle to make it unfavorable for consequent kimberlite eruptions. (5) Density anomalies in the lithospheric mantle show inverse correlation with seismic Vp, Vs velocities at 100-150 km depth. However, this correlation is weaker than reported in experimental studies and indicates that density-velocity relationship in the cratonic mantle is strongly non-unique. (C) 2016 International Association for Gondwana Research. Published by Elsevier B.V. All rights reserved.