Berlin 2018 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
O: Fachverband Oberflächenphysik
O 37: Electronic-Structure Theory: General II
O 37.2: Vortrag
Dienstag, 13. März 2018, 14:30–14:45, MA 141
Representing energy landscapes by combining neural networks and the empirical valence bond method — •Sinja Klees1, Eckhard Spohr2, and Jörg Behler1,3 — 1Lehrstuhl für Theoretische Chemie, Ruhr-Universität Bochum, D-44780 Bochum, Germany — 2Lehrstuhl für Theoretische Chemie, Universität Duisburg-Essen, D-45141 Essen, Germany — 3Theoretische Chemie, Georg-August-Universität Göttingen, D-37077 Göttingen, Germany
Computer simulations of aqueous electrolyte solutions can be challenging for several reasons: (i) there can be huge variations in the ratio of solvent molecules and ions; (ii) large-scale simulations are needed to avoid artificial periodicity and (iii) reactive potentials are necessary to take omnipresent proton transfer reactions into account. Artificial neural networks (NNs) are a powerful method to construct reliable and unbiased interatomic potentials for a wide range of systems. However, the construction of NN potentials can become computationally very demanding due to the high dimensionality of the configuration space, which needs to be mapped. Combining NN potentials of different system fragments with the empirical valence bond (EVB) method offers a promising approach to derive the potential energy surface of complex systems with substantially reduced effort. Preliminary results will be discussed and compared to density functional theory data.