Oxygen isotope evidence for slab melting in modern and ancient subduction zones

Abstract

We measured oxygen isotope compositions of 34 adakites, high-Mg andesites, and lavas suspected to contain abundant slab and sediment melts from the Western and Central Aleutians, the Andes, Panama, Fiji, Kamchatka, Setouchi (Japan), and the Cascades. This suite covers much of the diversity of arc lavas previously hypothesized to contain abundant slab melts. Measured and calculated values of delta 18 O for olivine phenocrysts in these samples vary between 4.88 per mil and 6.78 per mil, corresponding to calculated melt values of 6.36 per mil to 8.17 per mil. Values of delta 18 O for these samples are correlated with other geochemical parameters having petrogenetic significance, including Sr/Y, La/Yb, 87Sr/86Sr, and 143Nd/144Nd. Archetypical adakites from Adak Island (Central Aleutian) and Cook Island (Andean Austral zone), previously interpreted to be nearly pure melts of basaltic and gabbroic rocks in subducting slabs, have values of delta 18 O slightly higher than those of normal mid-oceanridge basalts, and in oxygen isotope equilibrium with typical mantle peridotite (i.e., their subtle 18O enrichment reflects their Sirich compositions and low liquidus temperatures, not 18 O-rich sources). Other primitive adakites from Panama and Fiji show only subtle sub-per mil enrichments in the source. This finding appears to rule out the hypothesis that end-member adakites are unmodified partial melts of basaltic rocks and/or sediments in the top (upper 1–2 km) of the subducted slab, which typically have delta 18 O values of ca. 9–20 per mil, and also appears to rule out them being partial melts of hydrothermally altered gabbros from the slab interior, which typically have delta 18 O values of ca. 2–5 per mil. One explanation of this result is that adakites are mixtures of partial melts from several different parts of the slab, so that higher- and lower- delta 18 O components average out to have no net difference from average mantle. Alternatively, adakites might be initially generated with more extreme delta 18 O values, but undergoisotopic exchange with the mantle wedge before eruption. Finally, adakites might not be slab melts at all, and instead come from differentation and/or partial melting processes near the base of the arc crust in the over-riding plate. High-Mg andesites and Setouchi lavas are commonly higher in delta 18 O than equilibrium with the mantle, consistent with their containing variable amounts of partial melts of subducted sediments (as we conclude for Setouchi lavas), slab-derived aqueous fluid (as we conclude for the Cascades) and/or crustal contaminants from the over-riding plate (as we conclude for Kamchatka).