Berlin 2024 – scientific programme

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

O: Fachverband Oberflächenphysik

O 111: Heterogeneous Catalysis II

O 111.5: Talk

Friday, March 22, 2024, 11:30–11:45, TC 006

Exploring Unsupervised Learning for Analysis of Adsorption Energy Distributions in CO2 to Methanol Synthesis — •Prajwal Pisal, Ondrej Krejci, and Patrick Rinke — Department of Applied Physics, Aalto University, P.O.Box 11100, FI-00076 AALTO, Finland

Synthesis of methanol from carbon dioxide using heterogeneous catalysts is a reaction of great relevance from an industrial and environmental perspective, underscoring the need for extensive catalytic exploration. Understanding the adsorption energy distributions (AE) of reactants on catalytic surfaces is crucial for evaluating material reactivity, as catalyst reactivity is often linked to AE.

In this work, we leverage our extensive database of AE for more than 100 catalytic materials using the Open Catalyst Project infrastructure [1]. We perform statistical analysis and unsupervised machine learning, like dimensionality reduction and clustering, on the dataset for recognition of features and patterns in the AE distributions. Utilizing these tools, we classify the materials under study and compare with experimental and literature data (e.g. [2]). This analysis provides a deeper understanding of the key properties of the materials that enhance the catalytic activity. The data obtained will further aid us in building predictive models to maximize methanol yield as a function of AE distributions and experimental conditions.

[1] L. Chanussot et al., ACS Catal. 11, 6059-6072 (2021); R. Tran et al., ACS Catal. 13, 3066-3084 (2023)

[2] A. J. Medford et al, J. Catal. 328, 36-42 (2015).

Keywords: Catalysis; Adsorption energy distributions; Unsupervised learning; ML force fields