Dehydration melting of nominally anhydrous mantle: The primacy of partitioning

Abstract

The onset of dehydration melting of nominally anhydrous peridotite can be calculated by combination of appropriate mineral/melt partition coefficients for H 2O, D H min / liq , and a parameterization of the influence of the H 2O content of melt on the solidus of peridotite. Thermodynamic models predict that olivine/melt partitioning, D H ol / liq , should increase with pressure, and though direct experimental determinations of D H ol / liq from 0.5 to 3 GPa do not show the predicted pressure dependence, storage capacity experiments suggest increases in D H ol / liq at pressures above 8 GPa and particularly at 12–14 GPa, near the base of the upper mantle. Calculations using experimental values of D H min / liq and ignoring the likely effect of pressure on D H ol / liq indicate that D H perid / liq increases from 0.006 at 1 GPa up to 0.009 at the onset of garnet stability at 2.8 GPa and then diminishes with further increases in pressure owing to decreasing pyroxene mode and decreasing Al in pyroxene. Because these calculations ignore the likely pressure effect on D H ol / liq , they represent minima. Incipient partial melts of mantle with 100 ppm H 2O have 1–2 wt.% H 2O from 1 to 5 GPa, and this modest H 2O concentration limits the stability of hydrous partial melts to temperatures approaching the dry solidus. The influence of H 2O on the melting behavior of peridotite can be quantified using a simple cryoscopic approach benchmarked against experiments on hydrous peridotite. Along a mantle adiabat with a potential temperature of 1323 °C, calculations indicate that dehydration partial melting of peridotite with 100 ppm H 2O begins at 80 km, or about 15 km deeper than would be the case for truly dry peridotite. However, decreases in D H perid / liq related to the onset of the stability of garnet mean that mantle modestly enriched in H 2O will begin melting significantly deeper, i.e., at 104 km for 200 ppm H 2O. In the low velocity zone (LVZ) beneath mature (50 Ma) oceanic lithosphere, incipient partial melting at 110 km requires 300 ppm H 2O and generation of small finite (≥0.1%) melt fractions across the entire LVZ from 90 to 200 km requires 600 ppm H 2O. The minimum concentration, 300 ppm H 2O, is 2–3 times that of typical convecting oceanic (MORB-source) mantle, so it is not likely that pervasive hydrous partial melting is responsible for the seismic properties of the LVZ. Extrapolation of low pressure partition coefficients to the base of the upper mantle indicates that at least 500 ppm H 2O is required to induce partial melting at depths of 300–400 km along a normal mantle geotherm. This argues that typical upper mantle with ∼100 ppm H 2O is not produced by partial melting above the 410 km discontinuity. Furthermore, the 500 ppm H 2O concentration is likely to be an underestimate, as it does not take into account probable enhancement in D H min / liq at high pressure.