Oxygen isotope constraints on the structure and evolution of the Hawaiian Plume

Abstract

The oxygen isotope stratigraphy of Ko‘olau volcano, Hawaii, is constructed by analyzing olivine phenocrysts from the KSDP drill core and submarine land-slide deposits. Along with those of subaerial (Makapu‘u) Ko‘olau olivines (Eiler and others, 1996a), they span the full range of the δ^(18)OVSMOW variation previously observed in Loa-trend Hawaiian volcanoes (Lō‘ihi, Mauna Loa, Hualalai, and Ko‘olau), vary systematically with the stratigraphic position, and correlate with other geochemical properties of their host lavas (Tanaka and others, 2002; Haskins and Garcia, 2004; Huang and Frey, 2005; Salters and others, 2006; Fekiacova and others, 2007). These observations can be explained if the Loa-trend volcanoes (including Ko‘olau) are constructed of magmas made by mixing peridotite melt with variable proportions of eclogite melt derived from a mafic constituent of the Hawaiian plume having a composition resembling recent mid-ocean-ridge basalts. We present a model of this magma mixing process that simultaneously explains the correlations among oxygen isotopes, major elements, trace elements and radiogenic isotopes. Although a number of models of this kind, differing in several parameters, describe the data equally well, all statistically acceptable models require an component with a MORB-like HREE pattern and enriched oxygen isotope composition (δ^(18)OVSMOW = 7.8-9.7‰), consistent with this component being an upper crustal (layer 1 or 2) basalt or gabbro with a low-temperature alteration history, possibly containing a small amount of sediment. Abundances of some minor elements—Ni and Ti—are not well described by this model; we show that these shortcomings are derived from the compositional assumption and operational difficulties, that is, TiO_2 content is too high in our assumed eclogite end-member, and the inversion of NiO content in the melt by the olivine addition calculation is imprecise due to the sensitivity of D_(NiO)^(olivine/melt) to the melt composition and to crystallization process. Previous studies have advocated that Hawaiian lavas were derived from partial melts of an olivine-free pyroxenite formed by reaction of eclogite-derived melt with peridotite (for example, Sobolev and others, 2005). Our study shows that the peridotite-derived and eclogite-derived melt-mixing model can explain the geochemistry of Hawaiian lavas as well, including high-Ni Ko‘olau olivines. We find that an olivine-free mantle source for Hawaiian lavas is unnecessary, although melt-rock interaction could be important in modifying melt composition. Inverting for mixing proportions and degree of melting, we estimate that the amount of recycled crust in the Hawaiian plume is <24 weight percent. Comparison of the late shield-stage Loa-trend (particularly Ko‘olau lavas) and Kea-trend (particularly Mauna Kea lavas) suggests that the geochemical diversity of Hawaiian lavas is produced by a thermally and chemically-zoned plume.