Trace element partitioning in Earth’s lower mantle and implications for geochemical consequences of partial melting at the core–mantle boundary

Trace element partitioning data between CaSiO 3-perovskite (CaPv), MgSiO 3-perovskite (MgPv), calcium–aluminum silicate (CAS-phase), and coexisting melts in peridotite and mid-ocean ridge basalt (MORB) compositions were obtained at 25–27 GPa and 2400–2530 °C using multi-anvil apparatus and ion microprobe. Results clearly show that CaPv is the predominant host for large ion lithophile elements (LILE) in the lower mantle. Because of the overwhelmingly high CaPv/melt partition coefficients (>10 for many of the LILE), partial melting in the lower mantle causes strong enrichment of LILE in the CaPv-bearing solid phase residue. CaPv has the following partitioning characteristics: (1) uniformly high partition coefficients for heavy rare earth elements (HREE) (e.g. 15 for Yb), decreasing toward light REE (e.g. 7 for La), (2) systematically lower partition coefficients for high field strength elements (Nb, Zr, Ti) and Sr relative to neighboring REE, (3) high Th and U, and systematically low Pb partition coefficients. Previous high-pressure studies have shown that the stability field of CaPv above solidus temperature is much wider in basaltic composition than in peridotite, indicating that melting of subducted oceanic crust in the lower mantle could produce significant geochemical CaPv signatures. Strong enrichment in Th and U relative to Pb in CaPv would result in radiogenic Pb isotopic compositions of the CaPv-bearing solid residue. Some clinopyroxenes in plume mantle peridotite xenoliths possess trace element patterns closely resembling those of natural CaPv found in diamonds and CaPv from the present experiments, suggesting that they were inherited from the CaPv-bearing precursor. In contrast, CaPv is the first phase to disappear during partial melting of peridotite above 24 GPa, and its geochemical signature may not be observable in nature. (MgPv + CaPv) fractional crystallization from a magma ocean has previously been put forward as a mechanism for Si depletion of the upper mantle. However, our data show that the magma ocean Sm/Ba ratio would deviate significantly from observed chondritic values when fractionation exceeds only 6%, if the crystallizing mixture were 10% CaPv and 90% MgPv. A significant hidden perovskite reservoir in the lower mantle can therefore be excluded.