Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

O: Fachverband Oberflächenphysik

O 35: Poster Session II (Polymeric biomolecular films; Nanostructures; Electronic structure; Spin-orbit interaction; Phase transitions; Surface chemical reactions; Heterogeneous catalysis; Particles and clusters; Surface magnetism; Electron and spin dynamics; Surface dynamics; Methods; Electronic structure theory; Functional molecules)

O 35.138: Poster

Dienstag, 27. März 2012, 18:15–21:45, Poster B

Systematic Construction of High-Dimensional Potential-Energy Surfaces by Neural Networks — Tobias Morawietz, Nongnuch Artrith, and •Jörg Behler — Lehrstuhl für Theoretische Chemie, Ruhr-Universität Bochum, D-44780 Bochum, Germany

Artificial neural networks (NNs) have become a promising method for the development of reliable potential-energy surfaces (PESs) for a wide range of systems including molecules, solids and surfaces [1,2]. Due to their high flexibility NNs are able to accurately represent energies and forces obtained from quantum chemical calculations, which makes them ideal tools to extend the length and time scale of molecular dynamics simulations. An important aspect for the constructing of high-dimensional NN PESs is the choice of the reference configurations. Here we present a systematic approach to build up NN potentials in an iterative fashion by identifying poorly represented parts of the configuration space.

[1] J. Behler, PCCP 13, 17930 (2011).

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