Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
O: Fachverband Oberflächenphysik
O 85: Molecular Simulations
O 85.2: Vortrag
Donnerstag, 3. April 2014, 18:00–18:15, WIL B321
Accelerating ab initio molecular dynamics simulations of water by artificial neural networks — •Tobias Morawietz1, Andreas Singraber2, Christoph Dellago2, and Jörg Behler1 — 1Lehrstuhl für Theoretische Chemie, Ruhr-Universität Bochum, 44780 Bochum, Germany — 2Faculty of Physics, University of Vienna, Boltzmanngasse 5, 1090 Vienna, Austria
Ab initio molecular dynamics (AIMD) simulations based on density-functional theory (DFT) are a valuable tool to study processes involving water at the atomic level. However, many important properties of ab initio water are unknown because the costly evaluation of the forces restricts the simulation time. Artificial neural network (NNs) trained to DFT calculations provide an unbiased way to construct accurate and efficient interatomic potentials [1]. Here, we show that NN potentials closely reproduce the DFT potential-energy surface of water under a wide range of conditions, thus enabling us to study properties of water so far not accessible to AIMD simulations [2].
[1] J. Behler, PCCP 13, 17930 (2011).
[2] T. Morawietz, A. Singraber, C. Dellago, and J. Behler, in preparation.