Magnesium isotope heterogeneity in oceanic island basalts

• Ocean island basalts exhibit negatively correlated δ 26 Mg and δ 57 Fe . • Mg and Fe isotope variations in OIB are not explained by variable melt generation processes nor mantle source heterogeneities. • Mg and Fe isotope variations in OIBresult from Mg-Fe inter-diffusion between melt and mantle during melt migration . • Diffusion-driven Mg and Fe fractionations in OIB add complexity in using these isotope ratios to probe mantle compositions. 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 δ 26 Mg 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 δ 26 Mg higher than the mantle value, but the majority have δ 26 Mg within error of or lower than the mantle datum. We argue that such low δ 26 Mg 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 δ 57 Fe, which strikingly correlate negatively with δ 26 Mg, 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 δ 26 Mg and increases δ 57 Fe in the melt and likely occurs during melt transport through the mantle. Likewise, such kinetic fractionation is required to account for the Δ′ 25 Mg JP-1 > 0, measured in two selected samples with low δ 26 Mg. 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 δ 57 Fe of such melts.