A framework for machine-learning-augmented multiscale atomistic simulations on parallel supercomputers

摘要: Recent advances in quantum mechanical (QM)-based molecular dynamics (MD) simulations have used machine-learning (ML) to predict, rather than recalculate, QM-accurate forces atomic configurations sufficiently similar previously encountered ones. Here, we discuss how ML approaches can be deployed within large-scale QM/MM materials on massively parallel supercomputers, making QM zones of ≳1000 atoms routinely attainable. We argue that the approach allows computational effort concentrated most chemically active subregions zone, significantly improving overall efficiency simulation. thus propose a novel method partition large regions into multiple subregions, which computed achieve optimal scaling. Then review recently proposed QM/ML MD scheme (Z. Li, J.R. Kermode, A. De Vita Phys. Rev. Lett., 2015, 114, 096405), discussing this could efficiently combined with QM-zone partitioning.