skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Molecular Dynamics with On-the-Fly Machine Learning of Quantum-Mechanical Forces

; ;
Publication Date:
Sponsoring Org.:
OSTI Identifier:
Grant/Contract Number:
Resource Type:
Journal Article: Published Article
Journal Name:
Physical Review Letters
Additional Journal Information:
Journal Volume: 114; Journal Issue: 9; Journal ID: ISSN 0031-9007
American Physical Society
Country of Publication:
United States

Citation Formats

Li, Zhenwei, Kermode, James R., and De Vita, Alessandro. Molecular Dynamics with On-the-Fly Machine Learning of Quantum-Mechanical Forces. United States: N. p., 2015. Web. doi:10.1103/PhysRevLett.114.096405.
Li, Zhenwei, Kermode, James R., & De Vita, Alessandro. Molecular Dynamics with On-the-Fly Machine Learning of Quantum-Mechanical Forces. United States. doi:10.1103/PhysRevLett.114.096405.
Li, Zhenwei, Kermode, James R., and De Vita, Alessandro. 2015. "Molecular Dynamics with On-the-Fly Machine Learning of Quantum-Mechanical Forces". United States. doi:10.1103/PhysRevLett.114.096405.
title = {Molecular Dynamics with On-the-Fly Machine Learning of Quantum-Mechanical Forces},
author = {Li, Zhenwei and Kermode, James R. and De Vita, Alessandro},
abstractNote = {},
doi = {10.1103/PhysRevLett.114.096405},
journal = {Physical Review Letters},
number = 9,
volume = 114,
place = {United States},
year = 2015,
month = 3

Journal Article:
Free Publicly Available Full Text
Publisher's Version of Record at 10.1103/PhysRevLett.114.096405

Citation Metrics:
Cited by: 54works
Citation information provided by
Web of Science

Save / Share:
  • Understanding the relationship between the structure of light-harvesting systems and their excitation energy transfer properties is of fundamental importance in many applications including the development of next generation photovoltaics.
  • We introduce a fingerprint representation of molecules based on a Fourier series of atomic radial distribution functions. This fingerprint is unique (except for chirality), continuous, and differentiable with respect to atomic coordinates and nuclear charges. It is invariant with respect to translation, rotation, and nuclear permutation, and requires no preconceived knowledge about chemical bonding, topology, or electronic orbitals. As such, it meets many important criteria for a good molecular representation, suggesting its usefulness for machine learning models of molecular properties trained across chemical compound space. To assess the performance of this new descriptor, we have trained machine learning models ofmore » molecular enthalpies of atomization for training sets with up to 10 k organic molecules, drawn at random from a published set of 134 k organic molecules with an average atomization enthalpy of over 1770 kcal/mol. We validate the descriptor on all remaining molecules of the 134 k set. For a training set of 10 k molecules, the fingerprint descriptor achieves a mean absolute error of 8.0 kcal/mol. This is slightly worse than the performance attained using the Coulomb matrix, another popular alternative, reaching 6.2 kcal/mol for the same training and test sets. (c) 2015 Wiley Periodicals, Inc.« less
  • Ab initio quantum mechanical charge field molecular dynamics (QMCF-MD) were successfully applied to Cu(II) embedded in water to elucidate structure and to understand dynamics of ligand exchange mechanism. From the simulation studies, it was found that using an extended large quantum mechanical region including two shells of hydration is required for a better description of the dynamics of exchanging water molecules. The structural features characterized by radial distribution function, angular distribution function and other analytical parameters were consistent with experimental data. The major outcome of this study was the dynamics of exchange mechanism and reactions in the first hydration shellmore » that could not be studied so far. The dynamical data such as mean residence time of the first shell water molecules and other relevant data from the simulations are close to the results determined experimentally. Another major characteristic of hydrated Cu(II) is the Jahn-Teller distortion which was also successfully reproduced, leading to the final conclusion that the dominating aqua complex is a 6-coordinated species. The ab initio QMCF-MD formalism proved again its capabilities of unraveling even ambiguous properties of hydrated species that are far difficult to explore by any conventional quantum mechanics/molecular mechanics (QM/MM) approach or experiment.« less
  • We present a systematic study of a recently developed ab initio simulation scheme based on molecular dynamics and quantum Monte Carlo. In this approach, a damped Langevin molecular dynamics is employed by using a statistical evaluation of the forces acting on each atom by means of quantum Monte Carlo. This allows the use of an highly correlated wave function parametrized by several variational parameters and describing quite accurately the Born-Oppenheimer energy surface, as long as these parameters are determined at the minimum energy condition. However, in a statistical method both the minimization method and the evaluation of the atomic forcesmore » are affected by the statistical noise. In this work, we study systematically the accuracy and reliability of this scheme by targeting the vibrational frequencies of simple molecules such as the water monomer, hydrogen sulfide, sulfur dioxide, ammonia, and phosphine. We show that all sources of systematic errors can be controlled and reliable frequencies can be obtained with a reasonable computational effort. This work provides convincing evidence that this molecular dynamics scheme can be safely applied also to realistic systems containing several atoms.« less