Phonon transport properties of two-dimensional materials can play a crucial role in the thermal management of low-dimensional electronic devices and thermoelectric applications. In this study, both the empirical Stillinger–Weber (SW) and machine learning interatomic potentials are employed to investigate the lattice thermal conductivity of monolayer GeS and SnS through solving the phonon Boltzmann transport equation. The accuracy of the two types of interatomic potentials and their performance for the evaluation of thermal conductivity are verified by analyzing phonon harmonic and anharmonic properties. Our results indicate that the thermal conductivity can be predicted more accurately with a machine learning approach, while the SW potential gives rise to an overestimated value for both monolayers. In addition, the in-plane anisotropy of thermal transport properties existing in these monolayers can be confirmed by both potential models. Moreover, the origins of the deviation existing in calculated thermal conductivities, including both the effects of interatomic potential models and monolayer compositions, are elucidated through uncovering the underlying phonon transport mechanisms. This study highlights that in contrast to the machine learning approach, more careful verification is required for the simulation of thermal transport properties when empirical interatomic potential models are employed.
Read full abstract