Regensburg 2013 – wissenschaftliches Programm

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

O: Fachverband Oberflächenphysik

O 58: Poster Session III (Solid-liquid interfaces; Scanning probe and other methods; Electronic structure theory; Spin-orbit interaction)

O 58.71: Poster

Mittwoch, 13. März 2013, 18:15–21:45, Poster B1

Representation of the 42-dimensional DFT Potential-Energy Surface of R,R-Tartaric Acid by Neural Networks — •Sinja Klees and Jörg Behler — Lehrstuhl für Theoretische Chemie, Ruhr-Universität Bochum, D-44780 Bochum, Germany

The development of accurate interatomic potentials is a prerequisite to study complex systems, but often several physical approximations are indispensable to reduce the computational costs. Neural Networks (NNs) are a class of flexible mathematical functions, which allow to construct accurate interatomic potentials for systems with an arbitrary number of atoms without the introduction of any constraints on the functional form. Instead, a systematic construction of the potential-energy surface (PES) is carried out by interpolation of energies and forces obtained from first-principles reference calculations. To date, most NN PESs refer to inorganic solids and molecules up to six atom. We use R,R-Tartaric Acid to demonstrate that the method is equally applicable to organic molecules. We show that the PESs obtained from DFT reference calculations and NN interpolation are in a very good agreement.