Search Menu
DOE PAGES title logo U.S. Department of Energy
Office of Scientific and Technical Information
Search Scholarly Publications

Title: Density functional theory based neural network force fields from energy decompositions

Journal Article · · Physical Review B
 [1];  [2];  [1];  [2]
  1. California Institute of Technology (CalTech), Pasadena, CA (United States). Joint Center for Artificial Photosynthesis (JCAP)
  2. Lawrence Berkeley National Lab. (LBNL), Berkeley, CA (United States)
+ Show Author Affiliations

In order to develop force fields (FF) for molecular dynamics simulations that retain the accuracy of ab initio density functional theory (DFT), in this work 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.

Research Organization:
California Institute of Technology (CalTech), Pasadena, CA (United States). Joint Center for Artificial Photosynthesis (JCAP); Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States). National Energy Research Scientific Computing Center (NERSC); Oak Ridge National Laboratory (ORNL), Oak Ridge, TN (United States). Oak Ridge Leadership Computing Facility (OLCF)
Sponsoring Organization:
USDOE; USDOE Office of Science (SC), Basic Energy Sciences (BES). Materials Sciences & Engineering Division
Grant/Contract Number:
AC02-05CH11231; SC0004993
OSTI ID:
1544118
Journal Information:
Physical Review B, Journal Name: Physical Review B Journal Issue: 6 Vol. 99; ISSN 2469-9950
Publisher:
American Physical Society (APS)Copyright Statement
Country of Publication:
United States
Language:
English

References (36)

First Principles Neural Network Potentials for Reactive Simulations of Large Molecular and Condensed Systems journal August 2017
Neural network and ReaxFF comparison for Au properties: Comparison of ReaxFF and BPNN Potentials journal March 2016
Shape retrieval using 3D Zernike descriptors journal September 2004
The analysis of a plane wave pseudopotential density functional theory code on a GPU machine journal January 2013
Amp: A modular approach to machine learning in atomistic simulations journal October 2016
Fast plane wave density functional theory molecular dynamics calculations on multi-GPU machines journal October 2013
Intrinsic Bond Energies from a Bonds-in-Molecules Neural Network journal June 2017
Deep Neural Nets as a Method for Quantitative Structure–Activity Relationships journal February 2015
Learning representations by back-propagating errors journal October 1986
Quantum-chemical insights from deep tensor neural networks journal January 2017
Searching for exotic particles in high-energy physics with deep learning journal July 2014
Machine learning phases of matter journal February 2017
Machine learning applications in genetics and genomics journal May 2015
Machine learning in materials informatics: recent applications and prospects journal December 2017
Effects of the c-Si/a-SiO 2 interfacial atomic structure on its band alignment: an ab initio study journal January 2017
Atom-centered symmetry functions for constructing high-dimensional neural network potentials journal February 2011
Structure-activity relationships derived by machine learning: the use of atoms and their bond connectivities to predict mutagenicity by inductive logic programming. journal January 1996
Energy density in density functional theory: Application to crystalline defects and surfaces journal March 2011
High-dimensional neural-network potentials for multicomponent systems: Applications to zinc oxide journal April 2011
On representing chemical environments journal May 2013
Force and heat current formulas for many-body potentials in molecular dynamics simulations with applications to thermal conductivity calculations journal September 2015
Machine learning based interatomic potential for amorphous carbon journal March 2017
Efficient and accurate machine-learning interpolation of atomic energies in compositions with many species journal July 2017
First-principles Green-Kubo method for thermal conductivity calculations journal July 2017
Gaussian Approximation Potentials: The Accuracy of Quantum Mechanics, without the Electrons journal April 2010
Molecular Dynamics with On-the-Fly Machine Learning of Quantum-Mechanical Forces journal March 2015
Ab Initio Green-Kubo Approach for the Thermal Conductivity of Solids journal April 2017
Charge-Density Patching Method for Unconventional Semiconductor Binary Systems journal June 2002
Thermal Conductivity and Specific Heat of Thin-Film Amorphous Silicon journal February 2006
Generalized Neural-Network Representation of High-Dimensional Potential-Energy Surfaces journal April 2007
Thermal conductivity of thin films: Measurements and understanding
  • Cahill, David G.; Fischer, Henry E.; Klitsner, Tom
  • Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and Films, Vol. 7, Issue 3 https://doi.org/10.1116/1.576265
journal May 1989
Machine learning of accurate energy-conserving molecular force fields journal May 2017
Machine learning: Trends, perspectives, and prospects journal July 2015
Thermal Conductivity of Amorphous Silicon journal May 1996
Deep learning for computational biology journal July 2016
Computational Methods in Developing Quantitative Structure-Activity Relationships (QSAR): A Review journal March 2006

Cited By (3)

Design, Parameterization, and Implementation of Atomic Force Fields for Adsorption in Nanoporous Materials journal September 2019
Atomic energy mapping of neural network potential journal September 2019
Atomic energy mapping of neural network potential text January 2019

Similar Records

Related Subjects

75 CONDENSED MATTER PHYSICS
SUPERCONDUCTIVITY AND SUPERFLUIDITY