Strain Engineering of Single-layer MoS₂ on SiO₂ Substrate by Developing a Neural Network Interatomic Potential based on Density Functional Theory.

  Transition metal dichalcogenides (TMDCs) are promising materials for nanoelectronics, energy storage applications, and photonics. However, one of the existing challenges for these materials is the low fracture toughness resulting in rapid crack growth. CVD-grown monolayer MoS₂ (ML-MoS₂) on SiO₂ substrate is susceptible to fracture since there are interatomic forces at the interface, which may not be purely vdW force, leading to crack formation at the edge of MoS₂ triangular crystal layer. Moreover, there are restrictions in molecular dynamics (MD) to study this phenomenon due to the unavailability of a customized interatomic potential (IP) for this interfacial system. In this work, we aim to develop an IP for ML-MoS₂ on SiO₂ substrate using a deep neural network (DNN) based on ab-initio molecular dynamics (AIMD) to study fracture mechanisms. The trained model could recreate interatomic forces and energies of a random frame in the test sets. Furthermore, it could accurately describe the structural properties of the system, such as radial distribution function (RDF) and stress-strain response. Consequently, this model enables MD simulations to study fracture in ML-MoS₂ efficiently with high accuracy.