Effects of H2O-CO2 Fluids, Temperature, and Peridotite Fertility on Partial Melting in Mantle Wedges and Generation of Primary Arc Basalts

Many lines of evidence from high P-T experiments, thermodynamic models, and natural observations suggest that slab-derived aqueous fluids, which flux mantle wedges contain variable amounts of dissolved carbon. However, constraints on the effects of H2O-CO2 fluids on mantle melting, particularly at mantle wedge P-T conditions, are limited. Here we present new piston cylinder experiments on fertile and depleted peridotite compositions with 3.5 wt.% H2O and XCO2 [= molar CO2 / (CO2 + H2O)] of 0.04-0.17. Experiments were performed at 2-3 GPa and 1350 °C to assess how temperature, peridotite fertility, and XCO2 of slab-derived fluid affects partial melting in mantle wedges. All experiments produce olivine + orthopyroxene + 7 to 41 wt.% partial melt. Our new data, along with previous lower temperature data, show that as mantle wedge temperature increases, primary melts become richer in SiO2, FeO* and MgO and poorer CaO, Al2O3 and alkalis when influenced by H2O-CO2 fluids. At constant P-T and bulk H2O content, the extent of melting in the mantle wedge is largely controlled by peridotite fertility and XCO2 of slab-fluid. High XCO2 depleted compositions generate ~ 7 wt.% melt whereas, at identical P-T, low XCO2 fertile compositions generate ~ 30-40 wt.% melt. Additionally, peridotite fertility and XCO2 have significant effects on peridotite partial melt compositions. At a constant P-T-XCO2, fertile peridotites generate melts richer in CaO and Al2O3 and poorer in SiO2, MgO+FeO, and alkalis. Similar to previous experimental studies, at a constant P-T-fertility condition, as XCO2 increases, SiO2 and CaO of melts systematically decrease and increase, respectively. Such distinctive effects of oxidized form of dissolved carbon on peridotite partial melt compositions are not observed if the carbon-bearing fluid is reduced, such as CH4-bearing. Considering the large effect of XCO2 on melt SiO2 and CaO concentrations and the relatively oxidized nature of arc magmas, we compare the SiO2/CaO of our experimental melts and melts from previous peridotite + H2O ± CO2 studies to the SiO2/CaO systematics of primitive arc basalts and ultra-calcic, silica-undersaturated arc melt inclusions. From this comparison, we demonstrate that across most P-T-fertility conditions predicted for mantle wedges, partial melts from bulk compositions with XCO2 ≥ 0.11 have lower SiO2/CaO than all primitive arc melts found globally, even when correcting for olivine fractionation, whereas partial melts from bulk compositions with XCO2 = 0.04 overlap the lower end of the SiO2/CaO field defined by natural data. These results suggest that the upper XCO2 limit of slab-fluids influencing primary arc magma formation is 0.04 < XCO2 < 0.11, and this upper limit is likely to apply globally. Lastly, we show that the anomalous SiO2/CaO and CaO/Al2O3 signatures observed in ultra-calcic arc melt inclusions can be reproduced by partial melting of either CO2-bearing hydrous fertile and depleted peridotites with 0 < XCO2 < 0.11 at 2-3 GPa, or from nominally CO2-free hydrous fertile peridotites at P > 3 GPa.