[S-IT14] Deep Earth Sciences

[SIT14-17] Immiscibility of liquid Fe-Light-Element mixtures under high pressures using moleculer dynamics simulations based on machine-learning interatomic potential

*Satoshi Ohmura¹, Taku Tsuchiya² (1.Hiroshima Institute of Technology, 2.Ehime University)

Keywords:Immiscibility, Molecular dynamics, Machine learning interatomic potential, First principle calculation, Liquid Fe mixtures

Light elements (LE) such as hydrogen, carbon, oxygen, silicon, and sulfur are expected to exist in the Earth’s core because the core density is approximately 10 % smaller than that of pure iron [1,2]. For the liquid Fe-LE-O ternary systems, several experimental and theoretical studies have suggested the immiscibility of liquid Fe-LE-O [3,4]. However, besides thermodynamic phase equilibria, it is unclear the detailed local structures in liquid Fe–LE–O ternary systems, which are the origins of various properties of liquids. To investigate the immiscibility of these liquid systems, in addition to the local structures, larger scale MD simulation than DFT-based calculation is needed. For this reason, in this study, we investigated the structural and bonding properties of liquid iron-light-element systems, such as Fe-C-O, Fe-Si-O and Fe-S-O under high pressure using molecular-dynamics simulations based on density functional theory and machine learning interatomic (ANN) potentials. MD simulations based on ANN potentials reproduced behavior similar to that seen in the DFT-based simulations in the sense that H, C and O show “interstitial-type” behavior while Si and S show “substitutional” type properties in the liquid Fe-LE-O ternary systems.

[1] F. Birch, J. Geophy. Res. 69, 4377-4388 (1964)
[2] O.L Anderson, and D.G. Isaak, Phys. Earth Planet. Inter. 131, 19-27 (2002).
[3] S.M. Arveson, J. Deng, B.B. Karki, and K.K.M. Lee, Proc. Natl. Acad. Sci. 116, 10238 (2019).
[4] D. Huang, J.Badro, J. Brodholt, and Y. Li, Geophys. Res. Lett. 46, 6397-6405 (2019).