A fundamental understanding of the kinetics of heterogeneous catalytic systems is difficult due to the multisite complexity of the catalyst and the array of competing elementary reaction steps. In this aspect, single-atom alloys (SAAs) are emerging catalytic materials that have attracted significant interests in recent years because of its great simplicity in site structure and flexibility in tuning site properties. The formation of the free-atom-like d-states that resemble molecular metal complexes is a unique phenomenon that bridges the gap between heterogeneous and homogeneous catalysis. Indeed, the well-defined structure of a single-atom site with noble metal ligands has some similarities to the metal complexes with organic ligands in terms of the localized nature of metal d-states. However, the lack of detailed understanding of orbital interactions at single-atom alloys limits its application in catalysis, particularly for driving selective chemistry that is not possible on traditional alloys due to adsorption-energy scaling relations. Here, we develop a Bayeschem learning framework to directly learn interaction parameters from data. From the theory of chemisorption, adsorption energies of an adsorbate at metal surfaces can be decomposed to electrostatic interactions and one-electron interactions. Bayeschem provides a general platform to interpret the nature of chemical bonding and its relation to the underlying electronic structure. The rapid screening approach using the Bayeschem framework not only provides conceptual insights into the physical factors that control surface reactivity, but also yields implementable strategies for tailoring catalytic properties of single-atom sites.

Bayesian learning reveals the nature of chemical bonding at single-atom alloys

ACS Fall 2021

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