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

Title: An atomistic fingerprint algorithm for learning ab initio molecular force fields

ORCiD logo [1];  [1];  [1]
  1. Division of Applied Mathematics, Brown University, Providence, Rhode Island 02912, USA
Publication Date:
Sponsoring Org.:
OSTI Identifier:
Grant/Contract Number:
Collaboratory on Mathematics for Mesoscopic Modeling of Materials
Resource Type:
Journal Article: Publisher's Accepted Manuscript
Journal Name:
Journal of Chemical Physics
Additional Journal Information:
Journal Volume: 148; Journal Issue: 3; Related Information: CHORUS Timestamp: 2018-02-14 18:32:18; Journal ID: ISSN 0021-9606
American Institute of Physics
Country of Publication:
United States

Citation Formats

Tang, Yu-Hang, Zhang, Dongkun, and Karniadakis, George Em. An atomistic fingerprint algorithm for learning ab initio molecular force fields. United States: N. p., 2018. Web. doi:10.1063/1.5008630.
Tang, Yu-Hang, Zhang, Dongkun, & Karniadakis, George Em. An atomistic fingerprint algorithm for learning ab initio molecular force fields. United States. doi:10.1063/1.5008630.
Tang, Yu-Hang, Zhang, Dongkun, and Karniadakis, George Em. 2018. "An atomistic fingerprint algorithm for learning ab initio molecular force fields". United States. doi:10.1063/1.5008630.
title = {An atomistic fingerprint algorithm for learning ab initio molecular force fields},
author = {Tang, Yu-Hang and Zhang, Dongkun and Karniadakis, George Em},
abstractNote = {},
doi = {10.1063/1.5008630},
journal = {Journal of Chemical Physics},
number = 3,
volume = 148,
place = {United States},
year = 2018,
month = 1

Journal Article:
Free Publicly Available Full Text
This content will become publicly available on January 16, 2019
Publisher's Accepted Manuscript

Save / Share:
  • A class II atomistic force field with Lennard-Jones 6-9 nonbond interactions is used to investigate equations of state (EOS) for important high explosive detonation products N{sub 2} and H{sub 2}O in the temperature range 700-2500 K and pressure range 0.1-10 GPa. A standard 6th order parameter-mixing scheme is then employed to study a 2:1 (molar) H{sub 2}O:N{sub 2} mixture, to investigate in particular the possibility of phase-separation under detonation conditions. The simulations demonstrate several important results, including: (1) the accuracy of computed EOS for both N{sub 2} and H{sub 2}O over the entire range of temperature and pressure considered; (2)more » accurate mixing-demixing phase boundary as compared to experimental data; and (3) the departure of mixing free energy from that predicted by ideal mixing law. The results provide comparison and guidance to state-of-the-art chemical kinetic models.« less
  • 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
  • Cited by 1
  • Machine learning (ML) techniques with the genetic algorithm (GA) have been applied to determine a polarizable force field parameters using only ab initio data from quantum mechanics (QM) calculations of molecular clusters at the MP2/6-31G(d,p), DFMP2(fc)/jul-cc-pVDZ, and DFMP2(fc)/jul-cc-pVTZ levels to predict experimental condensed phase properties (i.e., density and heat of vaporization). The performance of this ML/GA approach is demonstrated on 4943 dimer electrostatic potentials and 1250 cluster interaction energies for methanol. Excellent agreement between the training data set from QM calculations and the optimized force field model was achieved. The results were further improved by introducing an offset factor duringmore » the machine learning process to compensate for the discrepancy between the QM calculated energy and the energy reproduced by optimized force field, while maintaining the local “shape” of the QM energy surface. Throughout the machine learning process, experimental observables were not involved in the objective function, but were only used for model validation. The best model, optimized from the QM data at the DFMP2(fc)/jul-cc-pVTZ level, appears to perform even better than the original AMOEBA force field (amoeba09.prm), which was optimized empirically to match liquid properties. The present effort shows the possibility of using machine learning techniques to develop descriptive polarizable force field using only QM data. The ML/GA strategy to optimize force fields parameters described here could easily be extended to other molecular systems.« less
  • Using first principles density functional theory in combination with the nonequilibrium Green's function formalism, we study the effect of derivatization on the electronic and transport properties of C{sub 60} fullerene. As a typical example, we consider [6,6]-phenyl-C{sub 61}-butyric acid methyl ester (PCBM), which forms one of the most efficient organic photovoltaic materials in combination with electron donating polymers. Extra peaks are observed in the density of states (DOS) due to the formation of new electronic states localized at/near the attached molecule. Despite such peculiar behavior in the DOS of an isolated molecule, derivatization does not have a pronounced effect onmore » the electronic transport properties of the fullerene molecular junctions. Both C{sub 60} and PCBM show the same response to finite voltage biasing with new features in the transmission spectrum due to voltage induced delocalization of some electronic states. We also study the diffusive motion of molecular fullerenes in ethanol solvent and inside poly(3-hexylthiophene) lamella using reactive molecular dynamics simulations. We found that the mobility of the fullerene reduces considerably due to derivatization; the diffusion coefficient of C{sub 60} is an order of magnitude larger than the one for PCBM.« less