Parts | Days | Selection | Search | Updates | Downloads | Help

O: Fachverband Oberflächenphysik

O 55: Theoretical methods II

O 55.9: Talk

Wednesday, March 28, 2012, 18:45–19:00, MA 043

A High-Dimensional Neural Network Potential for Water: First Applications to Water Clusters — •Tobias Morawietz, Vikas Sharma, and Jörg Behler — Lehrstuhl für Theoretische Chemie, Ruhr-Universität Bochum, D-44780 Bochum, Germany

Large-scale atomistic simulations using molecular dynamics or Monte Carlo methods require a reliable but efficient representation of the interatomic potential. In recent years, artificial neural networks (NNs) have been shown to provide accurate potential-energy surfaces (PESs) for complex systems. NNs are flexible functions, which allow to interpolate a set of energies and forces obtained from electronic structure data but can be evaluated several orders of magnitude faster. Here we present an application of our generalized NN scheme for high-dimensional systems [1], which incorporates long-range electrostatic interactions based on environment-dependent atomic charges [2], to water clusters [3]. We show that binding energies and vibrational frequencies for several stationary points on the NN-PES are in excellent agreement with reference DFT data.

[1] J. Behler, and M. Parrinello, PRL 98, 146401 (2007).

[2] N. Artrith, T. Morawietz, and J. Behler, PRB 83, 153101 (2011).

[3] T. Morawietz, V. Sharma, and J. Behler, submitted (2011).