Partitioning of elements between peridotite and H2O at 2.0–3.0 GPa and 900–1100°C, and application to models of subduction zone processes

Abstract

Partition coefficients of Rb, Nb, La, Sr, Sm, Zr, Tm, and Y have been measured between a natural peridotite ± (TiO2 or ZrO2), its constituent minerals (olivine, clinopyroxene, orthopyroxene and accessories), and H2O-rich fluids at 2.0–3.0 GPa, 900–1100°C, and fO2 = NNO. Selected starting materials were doped with TiO2 or ZrO2 to measure the saturation concentrations of these oxides and to evaluate the effect of saturation in Zr-rich and Ti-rich minerals on measured values of the trace element partition coefficients Diperidotite/fluid. At 2.0 GPa the stable subsolidus assemblage is spinel lherzolite; ilmenite and rutile were stabilized by addition of TiO2, and zircon was stabilized by addition of ZrO2. At a nominal pressure of 3.0 GPa the stable subsolidus assemblage without added TiO2 or ZrO2 is garnet peridotite + zircon. A melt was present at 1100°C at 2.0 GPa and at 1000°C at 3.0 GPa. At 2.0 GPa spinel lherzolite contains rutile when the concentration of TiO2 is ≥ 0.11 wt% at 900°C, 0.13 wt% at 1000°C, and 0.31 wt% at 1100°C. The high solubility of TiO2 in spinel lherzolite at high temperatures implies that rutile is unlikely to be present in the source regions of Island Arc Basalts (IAB) during melting.Estimates of the mass fraction of dissolved solids in fluids in equilibrium with spinel lherzolite at 2.0 GPa range from 3.6–8.0 wt%, from the mass of quenched solute collected, to 8.7–24.8 wt% from mass balance of rock and solute compositions, showing a weak positive correlation with increasing temperature. Quenched solutes have high silica concentrations and high normative amounts of feldspars at 2.0 GPa, but are more Mg-rich and silica-poor in equilibrium with garnet peridotite at 3.0 GPa.Measured mineral/fluid partition coefficients do not show a strong dependence on temperature, nor are they affected by the presence of a melt. Rutile/fluid and ilmenite/fluid partition coefficients are similar, with high values for Nb (∼ 102–103) and Zr (∼ 102), intermediate values for REE and low values for Rb (∼ 10−3–10−2) and Sr (10−2). Clinopyroxene/fluid values decrease from Tm, Y, and Sm (∼ 10–100) to La (1–10), Zr and Sr (∼ 1), and Nb and Rb (0.01–0.1). Olivine/fluid values range from Tm (∼ 0.1–1), Y (0.01–1), and Sm, Zr and Nb (∼ 0.001–0.1) to La and Sr (0.001–0.01). Orthopyroxene/fluid values decrease from Tm (∼ 1–10) to Y (∼ 1) to Nb, La, Sm and Zr (∼ 0.1). Finally, pyrope garnet/fluid values show that the elements Tm (∼ 10000), Y (> 1000), Sm (< 1000), and Zr (∼ 100) strongly prefer garnet over fluid, while the elements Nb and La (∼ 1), Rb (∼ 0.1), and Sr (∼ 0.01) are less compatible.Peridotite/fluid bulk partition coefficients measured at 2.0 GPa range from ∼ 100 for Tm, with patterns on an incompatible element diagram being relatively smooth. The presence of rutile increases Dperidotite/fluid of Nb and Zr, but not by more than a factor of 10 because of its low modal abundance. At 3.0 GPa the presence of garnet in the stable assemblage increases the compatibility of heavy REE so that Digarnetperidotite/fluid values range from 1 for Nb to 235 for Tm. Aqueous fluids in equilibrium with spinel lherzolite have trace element compositions similar to those predicted for the metasomatic agent added to the source region of IAB, while fluids in equilibrium with garnet peridotite (measured) and eclogite (estimated from measured mineral/fluid partition coefficients) do not, suggesting that fluids released from the subducted slab equilibrate with peridotite in the mantle wedge before reaching the IAB source.