(45c) Theory-Guided, Interpretable Machine Learning Finds Predictive Geometric Structure-Property Relationships for Chemisorption on Alloys

Chemisorption, the direct chemical bonding of an atomic or molecular species to a solid-state material, is central in the fields of catalysis, corrosion, and electrochemistry, among many others. Despite the fundamental nature and importance of chemisorption, linking the geometry (i.e., the structure and composition) of different materials to their chemisorption properties remains a critical challenge. Developing physically transparent and quantitatively accurate models that can relate the chemisorption strength between an adsorbate and a solid surface to the adsorption site’s geometry is critical to advance our understanding of chemisorption. In this talk, we discuss our efforts to use a theory-guided machine learning approach, which uses an interpretable class of machine learning models called generalized additive models (iGAM models),¹ to discover predictive structure-property models that can quantify the chemisorption strength of O, OH, S, and Cl on Pt-metal and Au-metal alloy surfaces subject to various strain- and ligand-induced changes in the local geometric structure. The iGAM models show a strong degree of predictive accuracy, with an average root-mean-square-error (RMSE) of 0.046 eV for samples in the test set. Through quantification of the relative importance of the features used to construct the models, we identified three important geometric features of the adsorption site that impact the relative chemisorption strength on metal alloys: the strain in the surface layer, the number of d-electrons in the ligand metal, and the size of the ligand atom. The interpretable functional form of the iGAM models allows us to analyze the model behavior with respect to each of these geometric features. Comparison between the iGAM chemisorption models and established electronic-structure models shed light on the critical physical concepts that control the chemisorption process on metal surfaces.

References

(1) Lou, Y.; Caruana, R.; Gehrke, J. Intelligible Models for Classification and Regression. In Proceedings of the 18th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining; KDD ’12; ACM: New York, NY, USA, 2012; pp 150–158. https://doi.org/10.1145/2339530.2339556.