Exploiting HAADF-STEM to Determine the Surface Coverage and Distribution of Immobilized Molecular Complexes

The immobilization of molecular transition metal catalysts on solid supports, particularly semiconductors like silicon (Si), combines the advantages of homogeneous and heterogeneous catalysis for applications such as photo-electrocatalytic CO2 reduction. While spectroscopy and other methods provide averaged information about surface structures, they lack insight into catalyst distribution and coverage. Aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) offers potential for locating immobilized molecular catalysts with sub-Ångstrom resolution, by exploiting the strong scattering of the single metal atoms in the catalysts. However, applying HAADF-STEM to molecular catalysts on Si presents challenges due to organic ligands' vulnerability to electron beam. Furthermore, the high magnifications used to form the images leads to a small sampling area, hindering quantification. To overcome the limitation of small image regions, we use a convolutional neural network (CNN) to analyze numerous images quickly, enabling statistical analysis of a representative surface percentage. Our study characterizes molecular catalysts immobilized on Si using HAADF-STEM combined with a CNN model, focusing on catalysts used in either CO2 reduction or hydrogen evolution reaction (HER). Once trained, the CNN model can readily detect the single atoms, allowing detailed statistical analysis of surface coverage, catalyst distribution, and exploring how electron irradiation effects different catalytic systems. This approach explored clustering and dispersion behaviors of immobilized molecular catalysts, showing that surface distribution and coverage vary depending on attachment group, ligand type, and reaction conditions. The work demonstrates that HAADF-STEM, in conjunction with CNN models, is an optimal tool for understanding the distribution of molecular catalysts on surfaces, providing unprecedented opportunities to connect linker types, coverage/dispersion, and catalytic activity.