Causes of the Compositional Variability among Ocean Floor Basalts

The chemistry of ocean floor basalts (OFB) has been much studied for the insights that the compositions of their parental magmas may give into the thermal structure of the mantle, and its compositional heterogeneity. But as Mike O’Hara pointed out, OFBs are not primary magmas, and their extensive low-pressure differentiation ‘must be quantified before chemical features of lavas are ascribed to the nature of the underlying upper mantle’. O’Hara also emphasized that individual magmatic eruptions are products of complex magmatic systems. Volcanoes are not formed by isolated events, but are sites of long-term activity at which eruptions are caused by the arrival of replenishing magma that mixes with earlier magma through repeated cycles. In the intervals between replenishment events, part of the magma crystallizes, concentrating incompatible elements in the remaining magma. If the compositions of OFB are to be used to infer differences in the conditions of melting owing to mantle thermal structure, first it is necessary to disentangle the effects of this low-pressure evolution. It is a remarkable feature of OFB magmatism that these potentially complex replenish–mix–tap–crystallize (RMTX) cycles give rise to a simple pattern of chemical evolution when averaged at the global scale. Here we look at the deviations of 45 minor and trace element concentrations from these global average trends. The global trends are taken as the logarithms of the concentrations of the elements as a function of MgO concentration (log [M] vs [MgO]), and the deviations from these trends are defined as ‘variabilities’. An element’s variability is expected to result from a combination of source heterogeneity, partial melting in the mantle and melt extraction, and from the local variations in the RMTX process during crustal evolution. There is a strong correlation of an element’s variability with its slope in the global average trend, where the slope is a proxy for the element’s incompatibility (bulk partitioning) during global average RMTX. This correlation implies that the chemical controls on source heterogeneity and partial melting and melt extraction are similar to those during RMTX. Striking exceptions are the variabilities in plagioclase-hosted Na and Sr, indicating that a substantial part of the variability of Na in OFBs is due to variations in RMTX, rather than reflecting differences in extent of partial melting. If Iceland is excluded, there is no evidence from OFB chemistry to suppose that the distribution of potential temperatures in the sub-ridge mantle has a standard deviation, σ(Tp), larger than about 10°C, and if variability of the mantle source composition is allowed for, σ(Tp) could be even less. A principal component analysis of the variabilities reveals two more-or-less equally important principal components, suggesting that two distinct factors have affected incompatible trace element concentrations in global OFB. We suggest that these two factors are source heterogeneity and the RMTX process rather than the extent of partial melting or shape of the melting regime. An unexpected finding is the decoupling of the variabilities of very incompatible elements (VICE) such as Ba, Th and Nb, from those of the highly incompatible elements (HICE), such as La and the other light rare earth elements, Zr, P and Be