A framework for machine‐learning‐augmented multiscale atomistic simulations on parallel supercomputers
- 6 June 2015
- journal article
- research article
- Published by Wiley in International Journal of Quantum Chemistry
- Vol. 115 (16) , 1129-1139
Abstract
Recent advances in quantum mechanical (QM)‐based molecular dynamics (MD) simulations have used machine‐learning (ML) to predict, rather than recalculate, QM‐accurate forces in atomic configurations sufficiently similar to previously encountered ones. Here, we discuss how ML approaches can be deployed within large‐scale QM/MM materials simulations on massively parallel supercomputers, making QM zones ofatoms routinely attainable. We argue that the ML approach allows computational effort to be concentrated on the most chemically active subregions of the QM zone, significantly improving the overall efficiency of the simulation. We thus propose a novel method to partition large QM regions into multiple subregions, which can be computed in parallel to achieve optimal scaling. Then we review a recently proposed QM/ML MD scheme (Z. Li, J.R. Kermode, A. De Vita Phys. Rev. Lett., 2015, 114, 096405), discussing how this could be efficiently combined with QM‐zone partitioning.Keywords
Funding Information
- The Engineering and Physical Sciences Research Council (EP/L014742/1, EP/L027682/1)
- Argonne National Laboratory through a collaboration with the Thomas Young Centre and by the Rio Tinto Centre for Advanced Mineral Recovery based at Imperial College London
This publication has 36 references indexed in Scilit:
- Macroscopic scattering of cracks initiated at single impurity atomsNature Communications, 2013
- Atomic mechanism and prediction of hydrogen embrittlement in ironNature Materials, 2012
- Hybrid atomistic simulation methods for materials systemsReports on Progress in Physics, 2009
- Low-speed fracture instabilities in a brittle crystalNature, 2008
- Generalized Neural-Network Representation of High-Dimensional Potential-Energy SurfacesPhysical Review Letters, 2007
- Quickstep: Fast and accurate density functional calculations using a mixed Gaussian and plane waves approachComputer Physics Communications, 2005
- Introducing ONETEP: Linear-scaling density functional simulations on parallel computersThe Journal of Chemical Physics, 2005
- Efficient Quantum Monte Carlo Energies for Molecular Dynamics SimulationsPhysical Review Letters, 2005
- “Learn on the Fly”: A Hybrid Classical and Quantum-Mechanical Molecular Dynamics SimulationPhysical Review Letters, 2004
- The SIESTA method forab initioorder-Nmaterials simulationJournal of Physics: Condensed Matter, 2002