Thermomechanical Properties of Transition Metal Dichalcogenides Predicted by a Machine Learning Parameterized Force Field.

Abstract

The mechanical and thermal properties of transition metal dichalcogenides (TMDs) are directly relevant to their applications in electronics, thermoelectric devices, and heat management systems. In this study, we use a machine learning (ML) approach to parametrize molecular dynamics (MD) force fields to predict the mechanical and thermal transport properties of a library of monolayered TMDs (MoS2, MoTe2, WSe2, WS2, and ReS2). The ML-trained force fields were then employed in equilibrium MD simulations to calculate the lattice thermal conductivities of the foregoing TMDs and to investigate how they are affected by small and large mechanical strains. Furthermore, using nonequilibrium MD, we studied thermal transport across grain boundaries. The presented approach provides a fast albeit accurate methodology to compute both mechanical and thermal properties of TMDs, especially for relatively large systems and spatially complex structures, where density functional theory computational cost is prohibitive.