Crystal structure prediction starting from a liquid using machine-learning potentials

Recent experimental measurements have discovered a number of temperature-driven phase transitions at high pressure, underlying the importance of including lattice dynamics in theoretical predictions of phase stability. However, direct structure prediction at finite temperature remains a challenge. In this talk we will discuss the feasibility of finite-temperature structure prediction by quenching liquids in large-scale simulations with machine learning (ML) potentials. The approach is based on solid-liquid structural similarities for efficiently training ML potentials and predicting crystal structures at target densities. To showcase the methodology, we have considered dense Li, where we have been able to simulate its rich phase diagram over a large pressure range.



This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344.