Understanding the Nature and Activity of Supported Platinum Catalysts for the Water–Gas Shift Reaction: From Metallic Nanoclusters to Alkali-Stabilized Single-Atom Cations

Identifying the nature of an active site and understanding the reaction mechanism at the atomic level are of critical importance in the design of efficient catalysts for targeted applications. While extended metal surfaces have been studied extensively for various catalytic processes and relative activities of different metals were predicted using volcano relationships, much less is known with regard to catalysis at metal/oxide interface sites. This Perspective focuses on recent computational studies that were aimed at understanding catalysis at such metal/oxide interface sites. The water–gas shift (WGS) reaction catalyzed by supported Pt catalysts has been chosen as a model system for such an analysis, since extensive computational studies with varying sizes of Pt clusters and supports are available and, thus, by comparison to experimental data, a deeper understanding of the active sites can be attained. Pt catalysts with different sizes and shapes stabilized on various supports were found to be active for the WGS. However, the identification of the exact nature and function of the active site in these catalysts is still a matter of debate. Here, we analyzed the computational studies performed using different active site models, namely, Pt surface models, reducible oxide supported Pt cluster models, and supported single Pt site models. The focus of this Perspective is not to summarize computational or experimental studies of the WGS but to highlight the importance of choosing appropriate active site models and methods in the computational studies such that the experimental behavior of these catalysts and the specific roles of the metal and support can be understood. Furthermore, results obtained for the WGS mechanism catalyzed by Na-stabilized single Pt cations supported on TiO 2 are discussed and the promotional role of alkali cations are identified. Finally, we touch upon the importance of uncertainty quantification to account for the inexact nature of DFT in the correlation of computational predictions with experimental observations.