Structure and Dynamics of Supercooled Liquid Ge <sub>2</sub> Sb <sub>2</sub> Te <sub>5</sub> from Machine‐Learning‐Driven Simulations

Yu-Xing Zhou,Han-Yi Zhang,Volker L. Deringer,Wei Zhang
DOI: https://doi.org/10.1002/pssr.202000403
2020-01-01
Abstract:Studies of supercooled liquid phase‐change materials are important for the development of phase‐change memory and neuromorphic computing devices. Herein, a machine‐learning (ML)‐based interatomic potential for Ge 2 Sb 2 Te 5 (GST) to conduct large‐scale molecular dynamics simulations of liquid and supercooled liquid GST is used. A pronounced effect of the thermostat parameters on the simulation results is demonstrated, and it is shown how using a Langevin thermostat with optimized damping values can lead to excellent agreement with reference ab initio molecular dynamics (AIMD) simulations. Structural and dynamical analyses are presented, including the studies of radial and angular distributions, homopolar bonds, and the temperature‐dependent diffusivity. Herein, the usefulness of ML‐driven molecular dynamics for further studies of supercooled liquid GST, with length and timescales far exceeding those that are accessible to AIMD is demonstrated.
What problem does this paper attempt to address?