Learning local equivariant representations for large-scale atomistic dynamics

A simultaneously accurate and computationally efficient parametrization of the potential energy surface of molecules and materials is a long-standing goal in the natural sciences. While atom-centered message passing neural networks (MPNNs) have shown remarkable accuracy, their information propagation has limited the accessible length-scales. Local methods, conversely, scale to large simulations but have suffered from inferior accuracy. This work introduces Allegro, a strictly local equivariant deep neural network interatomic potential architecture that simultaneously exhibits excellent accuracy and scalability. Allegro represents a many-body potential using iterated tensor products of learned equivariant…

• NS
  National Science Foundation

  Awards: DGE1745303, DE-AC05-00OR22725, DE-AC02-06CH11357, 2011754, DMR-2011754
• UD
  U.S. Department of Energy

  Awards: AC05-00OR22725, AC02-06CH11357, DE-AR0000775, DE-SC0022199, DE-SC0021110, DE-AC02, 06CH11357, under Contract DE-AC05-00OR22725, SC0021110, DE-AC02-06CH11357, DE-AC05, DE-SC0012573, 00OR22725, DE-AC02-
• HU
  Harvard University

  Awards: DMR-2011754, DGE1745303
• UO
  University of Texas at Austin
• AR
  Advanced Research Projects Agency - Energy

  Awards: DE-AC05-00OR22725, DE-AR0000775, DE-AC02-06CH11357, DE-SC0021110
• OO
  Office of Science

  Awards: DE-AC05-00OR22725, DE-AC02-06CH11357, DE-SC0022199, DE-AC02, DE-SC0021110, 06CH11357, DE-SC0012573, AC02-06CH11357, AC05-00OR22725
• AR
  Advanced Research Projects Agency
• MR
  Materials Research Science and Engineering Center, Harvard University

  Awards: DGE1745303, DMR-2011754, 2011754
• DO
  Division of Materials Research

  Awards: DMR-2011754, 2011754, DE-AC02-06CH11357
• BE
  Basic Energy Sciences

  Awards: DE-AC02, -SC0012573, DE-AC05-00OR22725, AC05-00OR22725, DE-SC0021110, DE-SC0012573, DE-SC0022199, AC02-06CH11357, 06CH11357
• AS
  Advanced Scientific Computing Research

  Awards: DE-AC05-00OR22725, DE-SC0021110, Contract DE-AC05-00OR22725
• FD
  FAS Division of Science, Harvard University
• AN
  Argonne National Laboratory

  Awards: DE-AC02, Contract DE-AC02-06CH11357, 06CH11357, AC02-06CH11357
• OR
  Oak Ridge National Laboratory

  Awards: AC05-00OR22725, DE-AC02-06CH11357