Regensburg 2019 – wissenschaftliches Programm

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

MM: Fachverband Metall- und Materialphysik

MM 37: Topical session (Symposium MM): Big Data Analytics in Materials Science

MM 37.11: Vortrag

Donnerstag, 4. April 2019, 18:30–18:45, H43

Predicting Reaction Energetics with Machine Learning — •Sina Stocker, Johannes T. Margraf, and Karsten Reuter — Technische Universität München, Germany

Predictive-quality first-principles based microkinetic models are increasingly used to analyze (and subsequently optimize) reaction mechanisms in heterogeneous catalysis. In full rigor such models require the knowledge of all possible elementary reaction steps and their corresponding reaction barriers. Unfortunately, for complex catalytic processes, such as the generation of synthetic fuels out of syngas, the number of such steps becomes so large that an exhaustive first-principles calculation of all barriers becomes prohibitively expensive.

As a remedy, we explore the possibility of machine learning (ML) approaches to the prediction of the reaction energetics. An essential component in such data-driven approaches are efficient molecular representations (descriptors). We test a range of such representations that have been suggested to describe properties of closed-shell molecules and specifically assess their capabilities in describing open-shell systems and consequently reaction energetics. The obtained overall promising performance confirms the potential of ML approaches for a high throughput screening of elementary steps in large reaction networks.