CaltechAUTHORS
  A Caltech Library Service

Density functional theory based neural network force fields from energy decompositions

Huang, Yufeng and Kang, Jun and Goddard, William A., III and Wang, Lin-Wang (2019) Density functional theory based neural network force fields from energy decompositions. Physical Review B, 99 (6). Art. No. 064103. ISSN 2469-9950. doi:10.1103/physrevb.99.064103. https://resolver.caltech.edu/CaltechAUTHORS:20190221-074413398

[img] PDF - Published Version
See Usage Policy.

2MB
[img] PDF (information on the locality of the energy decomposition scheme [S1], the machine learning code used in the manuscript [S2], the convergence of the neural network potential with respect to the size of the neural network [S3], how the DFT data are...) - Supplemental Material
See Usage Policy.

675kB

Use this Persistent URL to link to this item: https://resolver.caltech.edu/CaltechAUTHORS:20190221-074413398

Abstract

In order to develop force fields (FF) for molecular dynamics simulations that retain the accuracy of ab initio density functional theory (DFT), we developed a machine learning protocol based on an energy decomposition scheme that extracts atomic energies from DFT calculations. Our DFT to FF (DFT2FF) approach provides almost hundreds of times more data for the DFT energies, which dramatically improves accuracy with less DFT calculations. In addition, we use piecewise cosine basis functions to systematically construct symmetry invariant features into the neural network model. We illustrate this DFT2FF approach for amorphous silicon where only 800 DFT configurations are sufficient to achieve an accuracy of 1 meV/atom for energy and 0.1 eV/A for forces. We then use the resulting FF model to calculate the thermal conductivity of amorphous Si based on long molecular dynamics simulations. The dramatic speedup in training in our DFT2FF protocol allows the adoption of a simulation paradigm where an accurate and problem specific FF for a given physics phenomenon is trained on-the-spot through a quick DFT precalculation and FF training.


Item Type:Article
Related URLs:
URLURL TypeDescription
https://doi.org/10.1103/physrevb.99.064103DOIArticle
ORCID:
AuthorORCID
Huang, Yufeng0000-0002-0373-2210
Goddard, William A., III0000-0003-0097-5716
Additional Information:© 2019 American Physical Society. Received 24 September 2018; revised manuscript received 14 December 2018; published 6 February 2019. We thank Dr. Ling Miao for the help in evaluating the force field energies in Fig. 6(d). This work was supported by the Director, Office of Science, Office of Basic Energy Science, Materials Science and Engineering Division, of the U.S. Department of Energy under Contract No. DE-AC02-05CH11231, through the Material Theory (KC2301) program in Lawrence Berkeley National Laboratory. The work performed by Y.H. and W.A.G. was supported by the Joint Center for Artificial Photosynthesis, a Department of Energy (DOE) Energy Innovation Hub, supported through the Office of Science of the U.S. Department of Energy under Award No. DE-SC0004993. This work uses the resource of the National Energy Research Scientific Computing Center (NERSC) as well as the Oak Ridge Leadership Computing Facility through the INCITE project.
Group:JCAP
Funders:
Funding AgencyGrant Number
Department of Energy (DOE)DE-AC02-05CH11231
Joint Center for Artificial Photosynthesis (JCAP)UNSPECIFIED
Department of Energy (DOE)DE-SC0004993
Other Numbering System:
Other Numbering System NameOther Numbering System ID
WAG1315
Issue or Number:6
DOI:10.1103/physrevb.99.064103
Record Number:CaltechAUTHORS:20190221-074413398
Persistent URL:https://resolver.caltech.edu/CaltechAUTHORS:20190221-074413398
Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:93013
Collection:CaltechAUTHORS
Deposited By: Tony Diaz
Deposited On:21 Feb 2019 18:35
Last Modified:16 Nov 2021 16:55

Repository Staff Only: item control page