https://link.springer.com/article/10.1007/s40242-022-2218-3

For a practical high-loading single-atom catalyst, it is prone to forming diverse metal species owing to either the synthesis inhomogeneity or the reaction induced aggregation. The diversity of this metal species challenges the discerning about the contributions of specific metal species to the catalytic performance, and thus hampers the rational catalyst design. In this paper, a distinct solution of dispersion analysis based on transmission electron microscopy imaging specialized for metal-supported catalysts has been proposed in the capability of full-metal-species quantification(FMSQ) from single atoms to nanoparticles, including dispersion densities, shape geometry, and crystallographic surface exposure. This solution integrates two image-recognition algorithms including the electron microscopy-based atom recognition statistics (EMARS) for single atoms and U-Net type deep learning network for nanoparticles in different shapes. When applied to the C3N4- and nitrogen-doped carbon-supported catalysts, the FMSQ method successfully identifies the specific activity contributions of Au single atoms and particles in butadiene hydrogenation, which presents remarkable variation with the metal species constitution. This work demonstrates a promising value of our FMSQ strategy for identifying the activity origin of heterogeneous catalysis.