The effect of tetrahedral Al3+ on the partitioning of water between clinopyroxene and silicate melt

This experimental study examines the influence of tetrahedrally coordinated Al3+ (IVAl3+) on the partitioning of H+ between high-Ca clinopyroxene and silicate melt. Experiments were carried out at 1.5 GPa and 1275 to 1350 °C on a natural high-alumina basalt and a compositionally similar synthetic basalt that is nominally alumina free. The results extend the compositional range of clinopyroxene for which partitioning has been determined experimentally to both higher and lower IVAl3+, thereby clarifying its role in maintaining charge neutrality during H+ incorporation. Clinopyroxene-melt partition coefficients for H+ (DH2OCpx − Melt) determined for the high-alumina basalt are among largest ever reported (DH2OCpx − Melt = 0.0228 to 0.0477), while those determined for the nominally alumina-free starting composition are the smallest (DH2OCpx − Melt = 0.00445 to 0.0071). Our results confirm that H+ is incorporated into clinopyroxene through two independent mechanisms: (1) the creation of metal vacancies and coupled hydroxyl defects and (2) a coupled substitution involving IVAl3+ that creates isolated hydroxyl defects. The relative importance of each of these incorporation mechanisms was quantified by formulating DH2OCpx − Melt as the sum of a metal vacancy-related partition coefficient (DH2OVMe) and an Al-coupled substitution-related partition coefficient (DH2OAl) and fitting the resulting expression to all available experimental data. The contribution of DH2OVMe to DH2OCpx − Melt is relatively constant at 0.006 ± 0.002, but is sensitive to pressure, temperature, H2O fugacity and pyroxene composition. The contribution of DH2OAl varies from ∼ 0 to 0.038 and is strongly dependent on phase composition, but insensitive to pressure and temperature. Two parameterizations of DH2OCpx − Melt that account for the potential effects of temperature, pressure and phase composition were calibrated using our data and experimentally determined partition coefficients from the literature. The first parameterization considers only compositional effects and can be used to estimate DH2OCpx − Melt in cases where pressure and/or temperature are unknown. This equation was used to demonstrate that the pre-eruptive H2O contents of arc lavas calculated on the basis of H+ in clinopyroxene phenocrysts are systematically higher than those inferred from olivine-hosted melt inclusions, so that the former provide a record that is closer to primary (undegassed) values. The second parameterization considers the effects of pressure and temperature in addition to phase composition and provides a moderately better fit to the experimental data. This equation was used to re-evaluate the influence of H2O on the depth at which peridotite partial melting begins beneath oceanic spreading centers. Our calculations indicate that partial melting begins at depths that are shallower than suggested by previous estimates. For a potential temperature of 1350 °C, peridotite containing 50, 100, 150, and 200 ppm H2O dissolved in nominally anhydrous minerals begins melting at depths of 75, 79, 83, and 87 km, respectively.