Magnesium isotope heterogeneity in oceanic island basalts

We analysed the Mg isotope compositions of fifty-seven ocean island basalt (OIB) samples from eleven ocean islands by critical mixture double-spiking. Our results show variable δ26Mg ranging from -0.15 ‰ to -0.33 ‰, compared to a uniform upper mantle value of -0.236±0.006 ‰ and analytical reproducibility of ±0.027 ‰. While equilibrium partial melting and magma differentiation can fractionate Mg isotope compositions, the isotope fractionations caused by these processes cannot account for much of the observed Mg isotope variability in the OIB samples we analysed. Equilibrium melting models predict most OIB should have δ26Mg higher than the mantle value, but the majority have δ26Mg within error of or lower than the mantle datum. We argue that such low δ26Mg values are the result of neither pyroxenite nor recycled carbonate in OIB sources, in contrast to previous studies. We additionally analysed a sub-set of samples for δ57Fe, which strikingly correlate negatively with δ26Mg, as previously observed in mid-ocean ridge basalts. This trend can be readily explained by diffusion-controlled Mg-Fe exchange between melt and mantle olivine, which decreases δ26Mg and increases δ57Fe in the melt and likely occurs during melt transport through the mantle. Likewise, such kinetic fractionation is required to account for the Δ′25Mg JP-1 > 0, measured in two selected samples with low δ26Mg. The widespread occurance of diffusional effects in Mg and Fe isotope ratios of mantle-derived melts adds complexity in using them to investigate mantle melt sources, but provides a novel means to account for the much-debated, high δ57Fe of such melts.