LAUSR

Abstract Single-layer transition metal dichalcogenides (TMDs) and their alloys have attracted intensive attention due to their potential in optoelectronics and energy conversion. Understanding the thermal transport in these two-dimensional materials is crucial for designing reliable devices where these materials are integrated. Molecular dynamics simulation is a commonly employed approach to investigate phonon transport and thermal conductivity in solids, but interatomic potentials that could satisfactorily characterize the phonon properties of multiple TMDs simultaneously are not available at present. In this paper, a machine-learning-driven interatomic potential for MoS2-MoSe2 system based on the spectral neighbor analysis approach is parameterized by learning from a large amount of data generated from first-principles. The phonon dispersion and mode-specific Gruneisen parameters of MoS2 and MoSe2, as well as the phonon dispersion of the MoS2-MoSe2 superlattice, obtained from first-principles are well reproduced by the potential parameterized. By performing equilibrium molecular dynamics simulations, the lattice thermal conductivity of MoS2(1−x)Se2x alloys are calculated and a ten-fold reduction compared with MoS2 is found when x  = 50%. Moreover, the roles of mass disorder and force-field disorder on the low thermal conductivity is identified. The parameterized interatomic potential could be utilized to study phonon transport in other MoS2-MoSe2 nanostructures.