SKM 2023 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
CPP: Fachverband Chemische Physik und Polymerphysik
CPP 45: Emerging Topics in Chemical and Polymer Physics, New Instruments and Methods
CPP 45.6: Vortrag
Donnerstag, 30. März 2023, 10:45–11:00, ZEU 255
Structural Descriptors for Constructing High-Dimensional Neural Network Potentials — •Moritz R. Schäfer1,2, Jonas A. Finkler3, Stefan Goedecker3, and Jörg Behler1,2 — 1Lehrstuhl für Theoretische Chemie II, Ruhr-Universität Bochum, 44780Bochum, Germany — 2Research Center Chemical Sciences and Sustainability, Research AllianceRuhr, 44780 Bochum, Germany — 3Basel University, Department of Physics, Klingelbergstrasse 82, 4056 Basel, Switzerland
High-dimensional neural network potentials (HDNNPs) are a well established method to efficiently compute close-to ab initio-quality energies and forces for performing large-scale molecular dynamics simulations of complex systems. In this method, the total energy is constructed as a sum of environment-dependent atomic energy contributions. Also electrostatic interactions based on flexible atomic charges can be included. Both components crucially depend on the quality of the structural descriptors employed to characterize the local atomic environments. Here we investigate the combination of atom-centered symmetry functions with the recently proposed overlap matrix descriptor. Moreover, the advantages and disadvantages of both descriptors are discussed and illustrated for benchmark systems.