Physically transparent and predictive models that quantify structure-property relationships of materials are important in numerous fields, including catalysis. Most applications of machine learning (ML) in materials science thus far have focused on using black-box models such as neural networks for the fast and accurate prediction of target properties such as material stability or adsorbate binding energies.[1] Although there is no denying the utility of these black-box models, interpretable models are desirable because they can lead to theories and hypotheses that advance our scientific knowledge and accelerate progress beyond simply enabling high-throughput screening. In this talk, I will discuss various “transparent box” ML applications with the goal of interpretability in mind. I will first present on the use of the SISSO algorithm to find a descriptor that predicts the stability of perovskite (ABX<sub>3</sub>) compounds. SISSO finds a physically meaningful descriptor that predicts the stability of perovskite oxide and halide materials with superior performance compared to the well-known Goldschmidt tolerance factor.[2] Second, I will present the use of generalized additive models (GAMs) to clarify geometric structure-property relationships for chemisorption on alloys. By comparing the GAM-derived chemisorption models to previously established electronic-structure models, we clarify the critical physical parameters that control the chemisorption process on metal surfaces.[3] Third, I will show that unsupervised ML using principal-component analysis provides a pathway to develop accurate and interpretable electronic-structure descriptors that describe chemisorption on materials.[4]

Interpretable “transparent box” machine learning to understand materials and catalytic properties

ACS Spring 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.