Liquidus phase relations in the system MgO–MgSiO 3 at pressures up to 25 GPa—constraints on crystallization of a molten Hadean mantle

Abstract

An understanding of the details of the crystallization history of a Hadean magma ocean requires a knowledge of liquidus phase relations of the mantle at very high pressures. The system MgO–MgSiO 3 is a good simplified chemical model of the mantle and provides a foundation for study of more complex systems that approximate the composition of the mantle more closely. We present a determination of the pressure–temperature univariant curve for the reaction Mg 2SiO 4+MgSiO 3=Liquid at pressures up to 16.5 GPa, new data on the change in composition of the eutectic liquid with pressure, and a pressure–temperature projection of univariant and invariant equilibria in the system MgO–MgSiO 3 at pressures up to 25 GPa. With increasing pressure, the eutectic curve between Mg 2SiO 4 and MgSiO 3 encounters five invariant points as follows: orthoenstatite+clinoenstatite+forsterite+liquid, 11.6 GPa, 2150°C; clinoenstatite+majorite+forsterite+liquid, 16.5 GPa, 2240°C; majorite+forsterite+modified spinel+liquid, 16.6 GPa, 2245°C; majorite+perovskite+modified spinel+liquid, 22.4 GPa, 2430°C; and perovskite+modified spinel+periclase+liquid, 22.6 GPa, 2440°C (last two points from data of [Gasparik, T., 1990a. Phase relations in the transition zone. J. Geophys. Res. 95, 15751–15769]). Above 22.6 GPa, no form of Mg 2SiO 4 is stable at liquidus temperatures, and the melting reaction changes to periclase+perovskite=liquid. The composition of the eutectic liquid, in wt.%, varies with pressure in a nearly linear fashion from 21% Mg 2SiO 4, 79% MgSiO 3 at 2 GPa to 32% Mg 2SiO 4, 68% MgSiO 3 at 16.5 GPa, and reaches its maximum enrichment in Mg (45% Mg 2SiO 4, 55% MgSiO 3) at 22.6 GPa. These data are consistent with experimental data on natural peridotite compositions indicating that perovskite and magnesiowüstite would be the main phases to crystallize in the deeper parts of a mantle magma ocean. Published partition coefficient data show that fractional crystallization of these two phases in the lower mantle would produce an upper mantle with C1 chondrite normalized Ca/Al and Ca/Ti weight ratios of 2.0–2.3, far higher than primitive upper mantle estimates of 1.1–1.25 and 0.86–1.06, respectively. However, Ca-perovskite, which would crystallize in small amounts in the lower mantle, is such a powerful sink for Ca that Ca/Al and Ca/Ti enrichment of the upper mantle could be suppressed. We conclude that extensive fractional crystallization of a deep magma ocean is not at present proscribed by element partitioning arguments.