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

CPP: Fachverband Chemische Physik und Polymerphysik

CPP 58: Computational Physics of Soft Matter II

CPP 58.5: Vortrag

Mittwoch, 18. März 2015, 17:00–17:15, PC 203

Assessing the Quality of Ab Initio Molecular Dynamics Simulations of Water by High-Dimensional Neural Network Potentials — Tobias Morawietz¹, Andreas Singraber², Christoph Dellago², and •Jörg Behler¹ — ¹Lehrstuhl für Theoretische Chemie, Ruhr-Universität Bochum, D-44780 Bochum, Germany — ²Computational Physics, Universität Wien, Sensengasse 8/9, A-1090 Vienna, Austria

Ab initio molecular dynamics simulations have become a standard tool to study the properties of water, but they are severely hampered by the computational costs if long simulations of large systems are required. In recent years, interatomic potentials based on artificial neural networks (NNs) have been shown to provide very efficient and accurate potential-energy surfaces close to the reliability of electronic structure methods. Here we present a set of NN potentials for water based on density-functional theory. We demonstrate that these potentials can be used to assess the quality of different exchange-correlation functionals and the importance of van der Waals interactions in the description of liquid water.