Fractionation of oxygen and iron isotopes by partial melting processes: Implications for the interpretation of stable isotope signatures in mafic rocks

Recycling of oceanic crust into the deep mantle via subduction is a widely accepted mechanism for creating compositional heterogeneity in the upper mantle and for explaining the distinct geochemistry of mantle plumes. The oxygen isotope ratios (δ18O) of some ocean island basalts (OIB) span values both above and below that of unmetasomatised upper mantle (5.5 ± 0.4‰) and provide support for this hypothesis, as it is widely assumed that most variations in δ18O are produced by near-surface low-temperature processes. Here we show a significant linear relationship between δ18O and stable iron isotope ratios (δ57Fe) in a suite of pristine eclogite xenoliths. The δ18O values of both bulk samples and garnets range from values within error of normal mantle to significantly lighter values. The observed range and correlation between δ18O and δ57Fe is unlikely to be inherited from oceanic crust, as δ57Fe values determined for samples of hydrothermally altered oceanic crust do not differ significantly from the mantle value and show no correlation with δ18O. It is proposed that the correlated δ57Fe and δ18O variations in this particular eclogite suite are predominantly related to isotopic fractionation by disequilibrium partial melting although modification by melt percolation processes cannot be ruled out. Fractionation of Fe and O isotopes by removal of partial melt enriched in isotopically heavy Fe and O is supported by negative correlations between bulk sample δ57Fe and Cr content and bulk sample and garnet δ18O and Sc contents, as Cr and Sc are elements that become enriched in garnet- and pyroxene-bearing melt residues. Melt extraction could take place either during subduction, where the eclogites represent the residues of melted oceanic lithosphere, or could take place during long-term residence within the lithospheric mantle, in which case the protoliths of the eclogites could be of either crustal or mantle origin. This modification of both δ57Fe and δ18O by melting processes and specifically the production of low-δ18O signatures in mafic rocks implies that some of the isotopically light δ18O values observed in OIB and eclogite xenoliths may not necessarily reflect near-surface processes or components