Bayesian force fields from active learning for simulation of inter-dimensional transformation of stanene

Abstract We present a way to dramatically accelerate Gaussian process models for interatomic force fields based on many-body kernels by mapping both forces and uncertainties onto functions of low-dimensional features. This allows automated active learning combining near-quantum accuracy, built-in uncertainty, constant cost evaluation that is comparable classical analytical models, capable simulating millions atoms. Using this approach, we perform large-scale molecular dynamics simulations the stability stanene monolayer. discover an unusual phase transformation mechanism 2D stanene, where ripples lead nucleation bilayer defects, densification into disordered multilayer structure, followed formation bulk liquid at high temperature or growth 3D bcc crystal low temperature. The presented method opens possibilities rapid development fast accurate uncertainty-aware long-time complex materials.