Partial melt composition of enstatite chondritic mantle around the rheological transition at 23 GPa: Implications for the chemical differentiation of the Earth's mantle

Abstract

Terrestrial planets are thought to be made of chondritic materials. However, previous studies have suggested that the Earth's upper mantle is depleted in some incompatible refractory lithophile elements (RLEs), such as U and Th in comparison with chondritic values. To explain the compositional contradiction between the upper mantle and chondrites, a hidden reservoir for incompatible RLEs in the lower mantle has been proposed. These studies have invoked a possible formation of hidden reservoir for incompatible RLEs during magma ocean solidification and by subsequent mantle overturn due to the gravitational instability of Fe-rich dense cumulates. However, the density of partial melt and its fate in a crystallizing deep magma ocean is still debated. Here we report partial melt compositions of enstatite chondritic mantle at 23 GPa as a function of the extent of melting. The results show that bridgmanite crystallized in a chondritic magma ocean becomes enriched in Fe and Ca with decreasing extent of melting and prevents the crystallization of ferropericlase and CaSiO₃ perovskite (Davemaoite). At the rheological transition (the extent of melting of 40% based on mass balance calculation), where the solid-liquid separation efficiently occurs even in the case of equilibrium crystallization, the solid part is mostly composed of bridgmanite with a trace amount of majorite, and the melt becomes enriched in FeO and CaO compared to the initial chondritic composition. The calculated density of the melt at the rheological transition is lighter than bridgmanite, but denser than majorite, suggesting that the melt may have ponded at the bottom of the mantle transition zone. The melt ponded around 660 km depth may have been enriched in incompatible RLEs and formed the hidden reservoir for missing RLEs if this melt-bearing region is stable against subsequent mantle convection.