Explorative First-Principles Simulation Study of Mineral - Melt Trace Element Partitioning Behavior

Monday, 15 December 2014
Joahnnes Wagner and Sandro Jahn, Deutsches GeoForschungsZentrum, Potsdam, Germany
Knowledge of trace element partition coefficients is crucial for a range of geoscientific applications. Obtaining the necessary experimental data is still a challenging and time consuming task, as the relevant processes typically take place under extreme P/T conditions. In this study, we explore a molecular scale simulation approach to predict mineral-melt partitioning. We use first principles molecular dynamics to investigate the rare earth element Y in the system garnet - melt, with focus on the influence of the melt. To predict the free energy change of the exchange reaction when Y is distributed between two phases, the method of thermodynamic integration is employed. Here we use an alchemical transmutation by which the identity (here expressed by its pseudopotential parameters) of a major element is gradually changed, in our case from 100% Al to 100% Y. The free energy change in turn enables us to predict the phase Y will partition into, as has been done previously for a CaO-Al2O3-SiO2-Y2O3 model system, employing classical force field simulations1. A major advantage of a molecular dynamics approach is that simulations contain information about the melt structure itself, thus enabling us to link observations to structural changes, e.g. a shift of average Y-O coordination number as a function of network connectivity.

We choose a garnet-rich peridotite and a mid-ocean ridge basalt as starting compositions. After both melts are equilibrated at 3000 K and ambient pressure, we perform the thermodynamic integration and compare the free energy of the exchange reaction. The first results suggest that it is possible to at least qualitatively predict the behavior of Y in the two model systems, as compared to experimental findings. We will discuss the potential of the method to make quantitative predictions and how the effect of P and T can be evaluated. Potentially, this new tool may allow us to make predictions for almost any composition and condition available to computer simulations, e.g. fluid – mineral element distribution in subduction zones and their P/T dependence or partitioning behavior for melt compositions, not or hardly accessible in experiments.

1. Haigis, V., Salanne, M., Simon, S., Wilke, M., Jahn, S., Chem Geol 346, 14-21 (2013).