An updated predictive model for garnet-melt trace element partitioning

We report a new set of predictive relations for trivalent trace element partitioning between garnet and anhydrous silicate melt that updates the formulation of van Westrenen et al. (2001, CMP 142:219, WvW hereafter), adding literature data appearing since initial publication and new experiments performed in part to enhance the applicability of the model. The original WvW formulations parameterized fits of garnet-melt partitioning data to the lattice strain model in terms of garnet composition, pressure, and D-Mg. Our new experiments broaden the compositional coverage into more Fe-rich systems relevant to lunar and martian petrogenetic processes, and contribute data for majoritic garnets. In addition to updated thermodynamic rationales like those in the original WvW model, we evaluated statistical relationships between lattice-strain fit parameters (E, D0, and r0) and compositions of garnets and coexisting melts, D-Mg, pressure, and temperature. We find that there is a statistically significant relationship between E and temperature, D-Mg, and garnet pyrope content; and between D0 and temperature, D-Mg, and melt MgO/SiO2 ratio. The r0 model was updated to incorporate a new term for temperature dependence and an adjusted term for pressure dependence. Hence, the new model covers a wider range of composition, pressure, and temperature, and explicitly includes terms relating to melt composition. Our new model reproduces literature D-values not included in the parameterizations reasonably well, for example to within ~35-40% for Sc and Yb and to within ~55% for Sm. Our ongoing work will continue to broaden compositional coverage, and in particular will focus on the effects of the onset of the garnet-to-majorite transition on element partitioning. Ultimately we hope to be able to predict garnet-liquid partitioning for the full range of magmatic garnet compositions in planetary magmas.