Abstract
The authors developed a Behler-Parrinello-type neural network (NN) to improve the density-functional tight-binding (DFTB) energy and force prediction. The Δ-machine learning approach was adopted and the NN was designed to predict the energy differences between the density functional theory (DFT) quantum chemical potential and DFTB for a given molecular structure. Most notably, the DFTB-NN method is capable of improving the energetics of intramolecular hydrogen bonds and torsional potentials without modifying the framework of DFTB itself. This improvement enables considerably larger simulations of complex chemical systems that currently could not easily been accomplished using DFT or higher level ab initio quantum chemistry methods alone.
| Original language | English |
|---|---|
| Pages (from-to) | 867-873 |
| Number of pages | 7 |
| Journal | MRS Communications |
| Volume | 9 |
| Issue number | 3 |
| DOIs | |
| State | Published - Sep 1 2019 |
Funding
The authors thank Dr. Andreas Goetz (SDSC) and Dr. Thuong Nguyen (UCSD) for providing us their initial BPNN code for water clusters. We also thank Dr Jacek Jakowski, Dr. Panchapakesan Ganesh, and Dr. Maxim Ziatdinov (all ORNL) for helpful discussions. J.Z. acknowledges support under the Oak Ridge Science Semester (ORSS) grant from the Oak Ridge Institute for Science and Education (ORISE). V.Q.V. acknowledges support by an Energy Science and Engineering Fellowship of the Bredesen Center for Interdisciplinary Research and Graduate Education at the University of Tennessee, Knoxville. S.I. acknowledges support by the Fluid Interface Reactions, Structures and Transport (FIRST) Center, an Energy Frontier Research Center funded by the U.S. DOE Office of Science. Calculations were performed at the Center for Nanophase Materials Sciences, which is a U.S. Department of Energy Office of Science User Facility.