2020/08/31 by Yu Xie, Jonathan Vandermause, Lixin Sun +2 · 57 citations
Computer Science · Engineering · Materials Science · Physics and Astronomy · #Block Copolymer Self-Assembly #Force field (fiction) #Gaussian #Gaussian process #Machine Learning in Materials Science #Molecular dynamics #Nanopore and Nanochannel Transport Studies #Nucleation #Phase (matter) #Stability (learning theory) #Transformation (genetics) #cond-mat.mtrl-sci #cs.LG #physics.comp-ph
paper · pdf · doi:10.1038/s41524-021-00510-y
published in npj Computational Materials 7(1) (Nature Portfolio) · 30 pages of main text, 14 pages of supplementary materials, 9 figures in total
arxiv created 2021/02/18 · openalex created_date 2021/03/01 · openalex publication_date 2021/03/19 · arxiv updated 2021/03/23 · openalex updated_date 2026/08/05
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 for automated active learning of models combining near-quantum accuracy, built-in uncertainty, and constant cost of evaluation that is comparable to classical analytical models, capable of simulating millions of atoms. Using this approach, we perform large-scale molecular dynamics simulations of the stability of the stanene monolayer. We discover an unusual phase transformation mechanism of 2D stanene, where ripples lead to nucleation of bilayer defects, densification into a disordered multilayer structure, followed by formation of bulk liquid at high temperature or nucleation and growth of the 3D bcc crystal at low temperature. The presented method opens possibilities for rapid development of fast accurate uncertainty-aware models for simulating long-time large-scale dynamics of complex materials.