Silicate-Metal Partitioning of Trace Elements: An Exploratory Ab Initio Molecular Dynamics Study

Wednesday, 17 December 2014
Daniela Künzel, Deutsches GeoForschungsZentrum GFZ, Potsdam, Germany and Sandro Jahn, Deutsches GeoForschungsZentrum, Potsdam, Germany
Partition coefficients of trace elements are very important for the reconstruction of Earth formation and evolution processes. As such processes typically take place at extreme conditions of pressure and temperature, it is still challenging to obtain experimental data. However, recent developments in super-computing facilities and in computational methods have made it possible to obtain supplementing information on melts and element partitioning from ab initio atomistic calculations.

The model system used in this pilot study consists of two different melts: a simple Fe-Ni alloy representing the metal and a silicate phase with a varying ratio of Fe and Mg ((Fe,Mg)2SiO4). Traces of Ni or Cr are added to each system. Molecular dynamics simulations based on density functional theory are initially run at 2500 K and ambient pressure, using the CPMD and CP2K software packages. Conditions are chosen so that the results can be compared to available experimental data in order to assess the feasibility of the approach and the quality of its results. However, preliminary results at increased pressure and temperature conditions that are more relevant for core formation will be presented as well.

The results of the calculations include information on the melt structure, such as coordination environment, nearest neighbor distance and x-ray diffraction structure factors. The calculations at ambient pressure show that the behavior of Ni atoms in the silicate melt shows similarities to Mg and differs clearly from Fe. First results with a Cr trace show that it resembles Fe rather than Ni.

Furthermore, thermodynamic integration is able to provide thermodynamic information about the exchange of trace elements between both phases. It is possible to obtain the energy difference connected to this exchange, which can then be used to estimate partition factors. First results on the Ni partitioning will be presented.