SurfaceScience21 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

O: Fachverband Oberflächenphysik

O 109: Poster Session VIII: Poster to Mini-Symposium: Machine learning applications in surface science III

O 109.1: Poster

Thursday, March 4, 2021, 13:30–15:30, P

Development of a Neural Network Potential for Metal-Organic Frameworks — •Marius Herbold and Jörg Behler — Georg-August Universität Göttingen, Institut für Physikalische Chemie, Theoretische Chemie, Tammannstraße 6, 37077 Göttingen, Germany

Metal-organic frameworks (MOFs) are crystalline porous materials with many applications in chemistry and materials science, from gas separation to heterogeneous catalysis. Computer simulations of chemical processes in MOFs are severely limited by the use of classical force fields (FFs), because most FFs are unable to describe bond formation and breaking. In principle, electronic structure methods, like density-functional theory (DFT), can overcome this problem, but often the required systems are too large for routine applications of DFT. A high-dimensional neural network potential (NNP) combines the advantages of both worlds - the accuracy of first principle methods with the efficiency of simple empirical potentials. Here we present a method to construct a NNP for MOFs using size-converged fragments only.