ZURUECK HOCH VOR INHALT SUCHEN

» Back to overview
Proposing Institution

Theoretische Chemie, Universität Göttingen
Project Manager

Dr. Jörg Behler
Tammanstrasse 6
37077 Göttingen
Abstract
In our group we use neural network potentials (NNP) trained to large data sets of energies and forces obtained from electronic structure calculations to perform large scale molecular dynamics (MD) simulations. Due to their efficiency, NNPs allow to address even complex systems, which are beyond the scope of conventional ab initio MD, with no significant loss of accuracy with respect to the underlying first principles data.NNPs have already been applied successfully to a variety of systems.In this project, we make use of NNPs to investigate copper clusters supported on zinc oxide, which are used as industrial catalysts for methanol synthesis. Although a variety of experimental structural and reactivity data is available, detailed information on the atomic level properties of the catalyst and the oxide/copper interface is required. Due to the large size of the nanoparticles, it is not possible to perform ab initio simulations that can cover the full system. We therefore aim to perform NNP-based simulations of realistic structural models.This project is a continuation of a Summer of Simulation 2016 project at the LRZ. During that project, the initial data set for copper, ZnO, and the ternary component system has been generated. This data set was used to train the ternary NNPs and to obtain the first results for the formation of copper clusters on zinc oxide slabs with the basing hopping Monte Carlo. The extra computation time afforded by a new project would be used to extend the data set to copper and ZnO nanoparticles, and to add atomic hydrogen to the potential, which is omnipresent in the catalytic process.

Impressum, Conny Wendler