Lattice Thermal Conductivity of Monolayer InSe Calculated by Machine Learning Potential

Abstract

The two-dimensional post-transition-metal chalcogenides, particularly indium selenide (InSe), exhibit salient carrier transport properties and evince extensive interest for broad applications. A comprehensive understanding of thermal transport is indispensable for thermal management. However, theoretical predictions on thermal transport in the InSe system are found in disagreement with experimental measurements. In this work, we utilize both the Green–Kubo approach with deep potential (GK-DP), together with the phonon Boltzmann transport equation with density functional theory (BTE-DFT) to investigate the thermal conductivity (κ) of InSe monolayer. The κ calculated by GK-DP is 9.52 W/mK at 300 K, which is in good agreement with the experimental value, while the κ predicted by BTE-DFT is 13.08 W/mK. After analyzing the scattering phase space and cumulative κ by mode-decomposed method, we found that, due to the large energy gap between lower and upper optical branches, the exclusion of four-phonon scattering in BTE-DFT underestimates the scattering phase space of lower optical branches due to large group velocities, and thus would overestimate their contribution to κ. The temperature dependence of κ calculated by GK-DP also demonstrates the effect of higher-order phonon scattering, especially at high temperatures. Our results emphasize the significant role of four-phonon scattering in InSe monolayer, suggesting that combining molecular dynamics with machine learning potential is an accurate and efficient approach to predict thermal transport.