Publication Type Journal Article
Title Approach to Potential Energy Surfaces by Neural Networks. A Review of Recent Work
Authors Diogo A. R. S. Latino Rui P. S. Fartaria F Freitas Joao Aires-de-Sousa Fernando M S Silva Fernandes
Groups MTFT
Journal INTERNATIONAL JOURNAL OF QUANTUM CHEMISTRY
Year 2010
Month February
Volume 110
Number 2
Pages 432-445
Abstract In the last years, Neural Networks (NN s) turned out as a suitable approach to map accurate Potential Energy Surfaces (PES) from ab initio/DFT energy data sets. PES are crucial to study reactive and nonreactive chemical systems by Monte Carlo (MC) or Molecular Dynamics (MD) simulations. Here we present a review of (a) the main achievements, from the literature, on the use of NNs to obtain PES and (b) our recent work, analyzing and discussing models to map PES, and adding a few details not reported in our previous publications. Two different models are considered. First, NNs trained to reproduce PES represented by the Lennard-Jones (LJ) potential function. Second, the mapping of multidimensional PES to simulate, by MD or MC, the adsorption and self-assembly of solvated organic molecules on noble-metal electrodes, focusing the ethanol/Au (111) interface. In both cases, it is shown that NNs can be trained to map PES with similar accuracy than analytical representations. The results are relevant in the second case, in which simulations by MC or MD require an extensive screening of the interaction sites at the interface, turning the development of analytical functions a nontrivial task as the complexity of the systems increases. (C) 2009 Wiley Periodicals, Inc. Int J Quantum Chem 110: 432-445, 2010
DOI http://dx.doi.org/10.1002/qua.22198
ISBN
Publisher
Book Title
ISSN 0020-7608
EISSN 1097-461X
Conference Name
Bibtex ID ISI:000272601400018
Observations
Back to Publications List