Phases Transition Mechanism in the Growth of WS2 and MoS2 Layers: Ab Initio Data Driving Machine Learning Molecular Dynamics

Accurate and large-scale atomic simulation is crucial for understanding growth mechanism and precise synthesis control of crystals. However, it is beyond the capacities of both ab initio and classical molecular dynamics (MD). This study takes a feasible way by developing ab initio data to accurate machine learning interatomic potentials (MLIPs) to explore the complex growth processes. It successfully simulated the growth of monolayer, bilayer MoS₂/WS₂ and MoS₂/WS₂ heterostructures under variable conditions. Importantly, a SMMS (M is Mo/W atom) is disclosed as a new 2D structure with high stability during the growth. The two-step vapor-deposition can be regarded as the alternating between MS₂ and SMMS structures. Besides, the intermediate SMMS structure easy forms alloys can result in production of MoₓW₁₋ₓS₂ alloy. Furthermore, metallic SMMS is an ideal self-intercalating electrode for transition metal dichalcogenides based FET. This research provides a mode that ab initio data driving MLIPs to efficient and accurate atomic growth simulation and design of materials.