| Abstract Scope |
It is known that chemical short-range order (SRO) has impacts on various material behaviors, including mechanical deformation, radiation tolerance. In simulations, chemical SRO is typically generated from the hybrid Monte Carlo and molecular dynamics (MCMD) method, which merely considers system energy change associated with random atom swapping. Here, we aim to propose an alternative approach, which combines deep learning technique and kinetic Monte Carlo (KMC) method, to study SRO formation mediated by diffusion kinetics. It is shown that the proposed deep learning model can efficiently predict diffusion barriers in any local chemical configuration. With the energy barriers, atomic diffusion and the resulting SRO nucleation and growth are achieved using KMC. We expect that the chemical SRO generated from diffusion is different from that produced by the hybrid MCMD method. Meanwhile, KMC provides the diffusion timescale, which is helpful in evaluating the time needed for various degrees of chemical SRO formation. |