Learning a force field for the martensitic phase transformations in Zr

丁向东

西安交通大学金属材料强度国家重点实验室

Abstract

Atomic simulations provide an effective means to understand the underlying physics of martensitic transformations under extreme conditions. However, this is still a challenge for certain allotropic metals due to the lack of an accurate classical force field. Based on machine learning (ML) techniques, we develop a hybrid ML-QMD method in which interatomic potentials describing martensitic transformations can be learned with a  high degree of fidelity from quantum molecular dynamics simulations (QMD). Using Zirconium as a model system, we demonstrate the feasibility and effectiveness of our approach. Specifically, the ML-QMD potential correctly captures energetic and structural properties of zirconium as verified by comparison to experimental and density-functional data for phonons, elastic constants, surface and stacking fault energies. Molecular-dynamics simulations successfully map out the pressure-temperature phase diagram of Zirconium. Furthermore, the potential is promising for simulations of phase transitions under shock compression.

DOI: 10.12110/firstfmge.20171121.410