Application of artificial neural networks for rigid lattice kinetic\n Monte Carlo studies of Cu surface diffusion

Abstract

Kinetic Monte Carlo (KMC) is a powerful method for simulation of diffusion\nprocesses in various systems. The accuracy of the method, however, relies on\nthe extent of details used for the parameterization of the model. Migration\nbarriers are often used to describe diffusion on atomic scale, but the full set\nof these barriers may become easily unmanageable in materials with increased\nchemical complexity or a large number of defects. This work is a feasibility\nstudy for applying a machine learning approach for Cu surface diffusion. We\ntrain an artificial neural network on a subset of the large set of $2^{26}$\nbarriers needed to correctly describe the surface diffusion in Cu. Our KMC\nsimulations using the obtained barrier predictor show sufficient accuracy in\nmodelling processes on the low-index surfaces and display the correct\nthermodynamical stability of these surfaces.\n