Bereiche | Tage | Auswahl | Suche | Downloads | Hilfe

MM: Fachverband Metall- und Materialphysik

MM 39: Phase Transitions II

MM 39.4: Vortrag

Donnerstag, 26. März 2009, 12:45–13:00, IFW B

High-Dimensional Neural Network Potential-Energy Surfaces: From Elemental Solids to Multicomponent Systems — •Jörg Behler and Nongnuch Artrith — Lehrstuhl für Theoretische Chemie, Ruhr-Universität Bochum, D-44780 Bochum, Germany

Recently, artificial Neural Networks (NN) have been shown to provide accurate high-dimensional potential-energy surfaces for condensed systems. The evaluation of these NN potentials is several orders of magnitude faster than the underlying electronic structure calculations, which enables a routine application in molecular dynamics and metadynamics simulations of large systems. However, so far the applicability of the NN potentials has been limited to elemental systems. By combining the flexibility of the NN methodology with physically motivated terms we are now able to include long-range interactions. This is a necessary prerequisite for studying binary systems like oxides and general multicomponent systems with significant charge transfer. The capabilities of the method are demonstrated for zinc oxide as a benchmark system. We show that structural and energetic properties of various phases are in excellent agreement with reference density-functional theory calculations.