Thermal Transport in Mosi 2 N 4  Monolayer: A Molecular Dynamics Study Based on Machine Learning

Abstract

With the continuous miniaturization and integration of nanoelectronic devices, efficient thermal management has become increasingly critical. Two-dimensional (2D) materials have emerged as promising thermal management candidates due to their high thermal conductivity, excellent mechanical properties, and controllable growth characteristics. Among these, monolayer MoSi2N4, a novel 2D semiconductor material, has attracted significant attention for its unique structural configuration and exceptional physical properties. In this study, we developed a high-precision machine learning interatomic potential based on the neuroevolution potential (NEP) framework to systematically investigate the intrinsic thermal transport properties and modulation mechanisms of this 2D material. Through homogeneous nonequilibrium molecular dynamics (HNEMD) simulations, we obtained a room-temperature (300 K) thermal conductivity of 317 W·m-1·K-1, with reliability verified by spectral heat current (SHC) decomposition analysis. Our research further elucidates the size-dependent thermal conductivity behavior, providing theoretical insights into nanoscale thermal transport mechanisms. Notably, we discovered that 2%-4% biaxial tensile strain induces a significant thermal conductivity reduction of 24-39%. This phenomenon originates from strain-induced modifications in phonon dynamics, characterized by a leftward shift and peak suppression in the phonon density of states, which collectively enhance phonon scattering and reduce group velocities. These findings demonstrate that strain engineering serves as an effective strategy for thermal conductivity modulation in 2D materials, offering new perspectives for optimizing thermal management in nanoelectronic devices. This work combines machine learning potentials with advanced thermal transport computational methods, laying a theoretical foundation for the thermophysical properties research of monolayer MoSi2N4.

Keywords: machine learning potentials, molecular dynamics, thermal conductivity, phonon thermal transport, strain.