A Machine-Learning Interatomic Potential for Iron Under High Pressure and its Application to Shock Response

Abstract

Iron exhibits complex coupling between plastic deformation and phase transition under shock loading. We develop a machine learning interatomic potential within the moment tensor potential (MTP) framework to capture plasticity and phase transition. Our potential successfully addresses three limitations of previous potentials, including the description of plasticity before phase transformation, eliminating the appearance of unphysical FCC phase in transformation products, and reproducing the pressure dependence of melting temperature. The large-scale molecular dynamics simulations of shock response in single crystal Fe indicate that the distinct dislocation-mediated plasticity before phase transition only occurs in [110] direction shock. The primary deformation modes of the HCP phase were identified as 1/3〈1-100〉 dislocation slip and {10-12}〈10-1-1〉 twinning, while at higher impact velocities, amorphization suppresses the development of twins and dislocations. These results provide an understanding of the response of Fe under extreme conditions.

Keywords: Moment tensor potential, molecular dynamic, shock loading, Fe, phase transition, plasticity