Reduced sediment melting at 7.5–12 GPa: phase relations, geochemical signals and diamond nucleation

Melting of carbonated sediment in the presence of graphite or diamond was experimentally investigated at 7.5–12 GPa and 800–1600 °C in a multianvil apparatus. Two starting materials similar to GLOSS of Plank and Langmuir (Chem Geol 145:325–394, 1998) were prepared from oxides, carbonates, hydroxides and graphite. One mixture (Na-gloss) was identical in major element composition to GLOSS, and the other was poorer in Na and richer in K (K-gloss). Both starting mixtures contained ~6 wt% CO2 and 7 wt% H2O and were doped at a ~100 ppm level with a number of trace elements, including REE, LILE and HFSE. The near-solidus mineral assemblage contained a silica polymorph (coesite or stishovite), garnet, kyanite, clinopyroxene, carbonates (aragonite and magnesite–siderite solid solution), zircon, rutile, bearthite and hydrous phases (phengite and lawsonite at 10 GPa). Hydrous phases disappear at ~900 °C, and carbonates persist up to 1000–1100 °C. At temperatures >1200 °C, the mineral assemblage consists of coesite or stishovite, kyanite and garnet. Clinopyroxene stability depends strongly on the Na content in the starting mixture; it remains in the Na-gloss composition up to 1600 °C at 12 GPa, but was not observed in K-gloss experiments above 1200 °C. The composition of melt or fluid changes gradually with increasing temperature from hydrous carbonate-rich ( 1). Aragonite and Fe–Mg carbonate have very different REE partition coefficients (DMst-Sd/L ~ 0.01 and DArg/L ~ 1). Nb, Ta, Zr and Hf are strongly incompatible in both carbonates. The bearthite/melt partition coefficients are very high for LREE (>10) and decrease to ~1 for HREE. All HFSE are strongly incompatible in bearthite. In contrast, Ta, Nb, Zr and Hf are moderately to strongly compatible in ZrSiO4 and TiO2 phases. Based on the obtained partition coefficients, the composition of a mobile phase derived by sediment melting in deep subduction zones was calculated. This phase is strongly enriched in incompatible elements and displays a pronounced negative Ta–Nb anomaly but no Zr–Hf anomaly. Although all experiments were conducted in the diamond stability field, only graphite was observed in low-temperature experiments. Spontaneous diamond nucleation and the complete transformation of graphite to diamond were observed at temperatures above 1200–1300 °C. We speculate that the observed character of graphite–diamond transformation is controlled by relationships between the kinetics of metastable graphite dissolution and diamond nucleation in a hydrous silicocarbonate melt that is oversaturated in C.