Thermal conductivity predictions in monolayer MoSi2N4: Integrating neural network potentials with phonon scattering analysis

Two-dimensional (2D) materials, known for their exceptional thermal conductivity and mechanical flexibility, have emerged as promising candidates for thermal management applications. Recently, increasing attention has been given to investigating the lattice thermal conductivity of these materials. While traditional methods combining density functional theory (DFT) with the Boltzmann transport equation (BTE) can produce accurate results, these approaches are computationally expensive and demand substantial resources. To address this challenge, we employed machine learning to successfully model the interatomic potential of monolayer MoSi₂N₄. This neural network potential (NNP), combined with BTE, facilitated the theoretical calculation of MoSi₂N₄′s thermal conductivity. Using NNP, we efficiently and accurately calculated the lattice thermal conductivity of MoSi₂N₄, highlighting the importance of selecting an appropriate interaction cutoff distance to ensure calculation accuracy. Furthermore, using this NNP, we investigated how four-phonon scattering influences the heat conduction properties of MoSi₂N₄, thereby strengthening our comprehension of phonon scattering dynamics. This study not only optimized computational efficiency but also provided fresh perspectives on the heat transfer mechanisms in complex 2D materials.