A coarse-grained study on mechanical behaviors of diamond-like carbon based on machine learning

Diamond-like carbon (DLC) films have broad application potential due to their high hardness, high wear resistance, and self-lubricating properties. However, considering that DLC films are micron-scale, neither finite element methods nor macroscopic experiments can reveal their deformation and failure mechanisms. Here we propose a coarse-grained molecular dynamics (CGMD) approach which expands the capabilities of molecular dynamics simulations to uniaxial tensile behavior of DLC films at a higher scale. The Tersoff potential is modified by high-throughput screening calculations for CGMD. Given this circumstance, machine learning (ML) models are employed to reduce the high-throughput computational cost by 86%, greatly improving the efficiency of parameter optimization in second- and fourth-order CGMD. The final obtained coarse-grained tensile curves fit well with that of the all-atom curves, showing that the ML-based CGMD method can investigate DLC films at higher scales while saving a large number of computational resources, which is important for promoting the research and production of high-performance DLC films.