Parts | Days | Selection | Search | Updates | Downloads | Help
TT: Fachverband Tiefe Temperaturen
TT 38: Frontiers of Electronic Structure Theory: Nuclear Dynamics, Methods (jointly with O, HL)
TT 38.9: Talk
Tuesday, March 17, 2015, 12:45–13:00, MA 004
Representing Complex Potential Energy Surfaces by Artificial Neural Networks — •Christopher Handley and Jörg Behler — Lehrstuhl für Theoretische Chemie, Ruhr-Universit*at Bochum, D-44780 Bochum, Germany
Computer simulations of large systems are computationally costly, and in many cases intractable, when using ab initio methods. More efficient potentials are typically based on approximations specific for particular atomic interactions, and the fitting of these potentials is not straightforward. Neural Networks (NNs) can provide interatomic potentials that are comparable to the accuracy of quantum mechanical calculations [1,2]. They are flexible enough to fit complex functions to quantum mechanical training data and yield accurate energies and forces. Here, we present our recent work towards more transferable NN potentials. [1] C. M. Handley and P. L. A. Poplier, J. Phys. Chem. A, 114, 3371- 3383, (2010). [2] J. Behler, PCCP, 13, 17901-18232 (2011).