Petrogenesis of high-MgO lavas of the lower Mull Plateau Group, Scotland: insights from melt inclusions

Published data on Palaeogene flood basalts of the lower Mull Plateau Group (Scotland) show that the most primitive lavas (MgO > 8 wt %) have the greatest extent of crustal assimilation, inconsistent with a simple coupled assimilation–fractional crystallization (AFC) model. We present elemental data on rehomogenized olivine-hosted melt inclusions from four high-MgO flows to investigate the nature of crustal assimilation and melt aggregation processes during the initial stages of flood basalt magmatism on Mull. Whole-rock compositions have been variably modified by hydrothermal alteration associated with the nearby Central Complexes. Nd isotope compositions, which should be insensitive to this alteration, are lower than typical mantle values (εNd + 2·4 to −5·7), indicating variable modification by crustal assimilation in all four samples. Melt inclusions are protected against alteration effects within their host olivine crystals, and provide more robust estimates of magmatic liquid compositions than whole-rocks, particularly for the alkali elements Na, K and Ba. The whole-rock samples show limited variations in Na2O (2·4–2·8 wt %) and K2O (0·23–0·29 wt %), despite a wide range in immobile elements (e.g. Zr 62–126 ppm). In contrast, the melt inclusions show far greater variability in Na2O (1·8–4·0 wt %) and K2O (0·02–0·35 wt %) and positive correlations between K and Na. Melt inclusions from different samples show systematic correlations between alkalis (K + Na) and incompatible element ratios (e.g. Zr/Y), indicating that the inclusions record magmatic values for the fluid-mobile elements. For the two most incompatible-element-enriched samples, the whole-rock analyses are similar to the melt inclusions except for lower Na and higher Ba that are related to alteration. Therefore, the crustal assimilation in these magmas must have taken place prior to growth of the olivines. For the two more depleted samples, the inclusions have less contaminated compositions than the whole-rocks, and show broad trends of increasing K/Ti with decreasing Fo% of the host olivine. For these samples, crustal assimilation must have taken place both during and after growth of the olivines and in an AFC style in which assimilation is linked to magmatic differentiation. Melt inclusions from single samples show limited variability in Zr/Y compared with K/Ti, indicating that aggregation of melts from different parts of the melt column must have occurred at deeper levels prior to growth of the olivines in the samples. Although the whole-rock compositional variations capture the broad details of crustal assimilation and melting histories for the Mull lavas despite the variable effects of hydrothermal alteration, the melt inclusion data more clearly resolve significant details of these magmatic processes. The extent of assimilation and differentiation is linked to the depth of magma stalling: primitive, contaminated magmas in the lower crust vs. evolved, uncontaminated magmas at sub-Moho depths