Berlin 2024 – wissenschaftliches Programm

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

MM: Fachverband Metall- und Materialphysik

MM 29: Poster II

MM 29.9: Poster

Dienstag, 19. März 2024, 17:00–19:00, Poster B

Computational modeling of mesoporous aluminosilicates via ab-initio based Machine Learning Interatomic Potentials — •Tom Schächtel, Jong-Hyun Jung, Konstantin Gubaev, and Blazej Grabowski — Institute for Materials Science, Department of Materials Design, University of Stuttgart, 70569 Stuttgart, Germany

Mesoporous silica are commonly used as catalyst supports for heterogeneous catalysis. To further enhance the properties of these materials the introduction of small amounts of metal atoms into the amorphous silica matrix was proposed. To better understand these mesoscale materials at the atomistic level a dual approach is suggested: The structure is obtained via Molecular Dynamics and Monte Carlo methods based on Machine Learned Interatomic Potentials, namely Moment Tensor potentials (MTPs), which are trained on Density Functional Theory data, while the electronic properties, specifically the electronic Density of States, is calculated with the Density Functional Tight Binding method. As a proof of concept a first trial MTP is trained to simulate the structure of a single mesopore contained in an amorphous aluminosilicate matrix. The accuracy of the trained MTP is investigated by analyzing the aluminum distribution and other properties of interest in a trial non-porous aluminosilicate bulk system containing additional hydrogen atoms.

Keywords: Machine learning potentials; catalysis; Density Functional Theory; Molecular Dynamics