Numerical and Experimental Constraints on Trace Element Fractionation During Lunar Magma Ocean Solidification

Traditional lunar magma ocean (LMO) theory suggests the anorthositic lunar crust formed by floating of buoyant plagioclase, one of the major products of LMO solidification, on the surface of LMO. To test this model, after applying experimentally determined LMO solidification sequences, we compared calculated Rare Earth Element (REE) concentrations of lunar plagioclase to plagioclase in Apollo ferroan anorthosites and lunar anorthositic meteorites. Several initial LMO compositions, and lunar-relevant mineral-melt partition coefficient predictive models were tested (where the partition coefficient is the concentration ratio of a specific element between two phases). Compared with plagioclase in the natural samples, modeled plagioclase directly crystallized from the LMO exhibit incongruent REE patterns, making the relationship between the plagioclase and LMO uncertain. After testing the impact of a series of possible secondary processes on trace element fractionation, we found that subsolidus reequilibration after adding a small amount of KREEP (potassium, rare earth elements and phosphorus) component can explain the trace element concentrations of the lunar plagioclase. In these models, we constrained the oxygen fugacity and bulk trace element abundance of the LMO, and proposed a serial processing model to explain the physical evolution of the lunar crust (Chapter 1).

In the process of modeling LMO solidification, we found that there was no trace element partitioning model developed for apatite. Apatite has important petrological significance due to its capacity to substitute trace elements into its mineral structure in abundance. Accurate prediction of trace element concentrations of apatite can help determine the trace elemental fractionation during LMO evolution. We conducted high-temperature and high-pressure experiments with lunar-relevant starting compositions. Compiling our 12 new experiments together with published data, we developed predictive models for apatite-silicate melt trace element partitioning, and an Eu in apatite-plagioclase oxybarometer. Using the aforementioned models, we recovered anticipated oxygen fugacities for one of our piston-cylinder experiments, winonaite HaH 193, and samples from Sept Iles layered intrusion. Using the new models, we additionally investigated the petrogenesis of lunar KREEP basalts, and found that LMO cumulates with fluorapatite or hydroxylapatite in the source are consistent with their formation, suggesting relatively Cl depleted late-stage LMO cumulates (Chapter 2).