Experimental determination of CO2 content at graphite saturation along a natural basalt-peridotite melt join: Implications for the fate of carbon in terrestrial magma oceans

- High P-T CO2 contents of mafic to ultramafic melts at graphite saturation.
- Thermodynamic model of CO2 content for crystallizing magma ocean (MO).
- Peridotite melt CO2 content - Shallow MO should degas CO2 or precipitate graphite at the surface.
- At shallow MO, carbon partitions more strongly to alloy melt over peridotite than previously predicted.

Knowledge of the carbon carrying capacity of peridotite melt at reducing conditions is critical to constrain the mantle budget and planet-scale distribution of carbon set at early stage of differentiation. Yet, neither measurements of CO2 content in reduced peridotite melt nor a reliable model to extrapolate the known solubility of CO2 in basaltic (mafic) melt to solubility in peridotitic (ultramafic) melt exist. There are several reasons for this gap; one reason is due to the unknown relative contributions of individual network modifying cations, such as Ca2+ versus Mg2+, on carbonate dissolution particularly at reducing conditions. Here we conducted high pressure, temperature experiments to estimate the CO2 contents in silicate melts at graphite saturation over a compositional range from natural basalts toward peridotite at a fixed pressure (P) of 1.0 GPa, temperature (T) of 1600 °C, and oxygen fugacity (log⁡fO2∼IW+1.6). We also conducted experiments to determine the relative effects of variable Ca and Mg contents in mafic compositions on the dissolution of carbonate. Carbon in quenched glasses was measured and characterized using Fourier transform infrared spectroscopy (FTIR) and Raman Spectroscopy and was found to be dissolved as carbonate (CO32−). The FTIR spectra showed CO32− doublets that shifted systematically with the MgO and CaO content of silicate melts. Using our data and previous work we constructed a new composition-based model to determine the CO2 content of ultramafic (peridotitic) melt representative of an early Earth, magma ocean composition at graphite saturation. Our data and model suggest that the dissolved CO2 content of reduced, peridotite melt is significantly higher than that of basaltic melt at shallow magma ocean conditions; however, the difference in C content between the basaltic and peridotitic melts may diminish with depth as the more depolymerized peridotite melt is more compressible. Using our model of CO2 content at graphite saturation as a function of P-T-fO2-melt composition, we predict that a superliquidus shallow magma ocean should degas CO2. Whereas if the increase of fO2 with depth is weak, a magma ocean may ingas a modest amount of carbon during crystallization. Further, using the carbon content of peridotite melt at log⁡fO2 of IW and the knowledge of C content of Fe-rich alloy melt, we also consider the core-mantle partitioning of carbon, showing that DCmetal/peridotite of a shallow magma ocean is generally higher than previously estimated.