The aluminum conundrum in Hawaiian shield-building lavas: An argument for a deep, garnet-bearing, mantle source

In contrast with other major elements (e.g. SiO2, CaO, TiO2), isotopic ratios (e.g.Sr, Nd, Pb), and some trace element ratios (e.g. Zr/Nb, Zr/Y, La/Nb), the Al2O3 content of Hawaiian shield lavas, at a given MgO content, is remarkably uniform within and between volcanoes. The overall range at 10% MgO is about 0.8% from Loihi (12.1%) to Koolau (12.9%). There is substantial overlap between the “Loa” and “Kea” trend volcanoes, but the regression data for “Loa” trend volcanoes are slightly higher (0.2%) in Al2O3 than “Kea” trend volcanoes. Post-shield lavas, from both “Kea” and “Loa” trends are higher in Al2O3 at a given MgO content than corresponding shield lavas. These observations are completely at odds with what we think we know about melt production in a heterogeneous, thermally zoned, mantle plume. Peridotite melting models and experiments, at varying pressures and temperatures, result in melts with significantly different Al2O3 contents at a given MgO content. The almost constant Al2O3 at a given MgO content is analogous to the almost constant normalized Yb abundances in Hawaiian tholeiites. Both require melting in the presence of a garnet residue. The aluminum conundrum can be explained if Hawaiian shield lavas result from melting garnet peridotite at depths corresponding to around 120 km. However, this explanation is at odds with the variable, but relatively high SiO2 contents of most of these lavas, which imply melt production at relatively shallow depths corresponding to less than 90 km. This paradox can be resolved if low SiO2 parental magmas derived from garnet peridotite react with a depleted harzburgite residue from prior melting as the magma ascends through the upper regions of the plume.