Trace Element Partitioning at High Pressures: New Ion Probe Results and Implications for Differentiation of a Terrestrial Magma Ocean

Erik H. Hauri DTM, 5241 Broad Branch Rd., NW, Washington, DC 20015, USA

J. Zhang CHiPR, SUNY Stony Brook, NY, USA

C. Herzberg CHiPR and Dept. Geol., Rutgers University, New Brunswick, NJ, USA

Y. Fei Geophysical Lab, 5251 Broad Branch Rd NW, Washington, DC, USA

We present preliminary trace element partitioning data between mantle minerals (pyrope garnet, majorite garnet, and magnesiowüstite) and ultramafic melts. Three experiments are reported at 3.3, 7, and 21.7 GPa. The experiments at 3.3 and 7 Gpa were performed at the Geophysical Laboratory on a natural alkali picrite composition (20% MgO, 3.5% Na2O+K2O), prepared by mixing a natural trace-element rich alkali basalt with San Carlos olivine; pyrope garnet (Mg# 88) was the liquidus phase in both experiments. The 21.7 GPa experiment was performed at Stony Brook on KLB-1 peridotite doped with 1000 ppm of 15 trace elements; majorite garnet and magnesiowüstite were the liquidus phases, as predicted by previous experiments on the undoped KLB-1. All the trace element analyses were measured by ion microprobe on crystals as small as 15 µm without phase overlap, by minimizing beam spot size and using extended counting times.

We have obtained the first set of partitioning data for REE and other lithophile trace elements between magnesiowüstite and melt, at 21.7 GPa. Most of the investigated trace elements are incompatible in magnesiowüstite (Mw), the exceptions being Cr and Ni. Mw shows a LREE-depleted partitioning pattern, with positive anomalies (compared to adjacent REE) at Ti, Hf, and Sr. U, Th and Pb also appear to be moderately incompatible in Mw. The garnet partitioning data suggest a systematic effect of pressure and temperature on garnet/melt partitioning for trace elements. The largest shift is observed for the heavy-rare-earth elements (HREE); Yb appears to be incompatible in majorite, whereas it is compatible in pyrope. These effects appear to be systematic with pressure and temperature; as garnet transforms to majorite with increasing P and T, the trace element partitioning pattern appears to rotate about Nd, with the partition coefficients increasing for the most incompatible elements and decreasing for the HREE. As the garnet becomes richer in pyroxene component, the partitioning pattern rotates and becomes more similar to the partitioning pattern for pyroxene/melt. These majorite results, including the incompatible nature of the HREE in majorite, are consistent with all previous majorite partitioning data except those of Kato et al. (1988).

The new data place further trace element constraints on differentiation of a terrestrial magma ocean. There appears to be no sensitive tracer for Mw fractionation among the lithophile incompatible elements. The results for garnet strongly suggest that previously inferred Lu/Hf shifts due to majorite fractionation are too high; the exact Lu/Hf fractionation may strongly depend on the composition of the majorite (i.e. pyroxene component). Accumulation of pyrope garnet will create a high Lu/Hf reservoir, but accumulation of majorite will raise the Lu/Hf ratio only mildly, since both elements are incompatible in majorite. Similarly, removal of majorite will have little effect on the Lu/Hf systematics of a majorite-depleted reservoir compared to a chondritic reservoir. Majorite accumulation/removal has some influence on of all the parent/daughter fractionations in the major isotope systems, but the present-day isotopic characteristics of hypothetical majorite-enriched and majorite-depleted reservoirs is not extremely different from mantle reservoirs known and inferred from studies of hotspot basalts.

There appears to be systematic differences between the data of Kato et al. (1988) and other studies for majorite/melt partitioning. These differences require further experimentation and a reevaluation of existing constraints imposed by trace element partitioning studies relevant to all mantle phases.