Discovering descriptors linking catalyst film uniformity to oxygen reduction reaction performance

The uniformity of catalyst films is widely recognized as crucial for the oxygen reduction reaction (ORR), yet it has traditionally been assessed only qualitatively. In this study, 105 catalyst films of varying uniformity were prepared using the same Pt/C catalyst. About 42 features were then extracted from their microscope images. These features capture film uniformity from multiple aspects and together provide an objective means to quantify characteristics such as coffee-ring formation and catalyst agglomeration, which are beyond the reach of visual labeling. Principal component analysis (PCA) revealed that these features captured uniformity differences that were consistent with visual inspection, while also identifying quantitative variations that were beyond human perception. The Mann – Whitney U test and decision tree classification were then employed to identify the feature sets that most strongly influence three electrochemical parameters: ECSA, E1/2 and jk. Although film uniformity was generally important for all parameters, the most influential feature sets differed among them, suggesting that each parameter is sensitive to distinct aspects of uniformity. This study proposes descriptors that link catalyst film uniformity with ORR performance, paving the way for more reproducible and reliable catalyst evaluations. A quantitative, interpretable image-descriptor framework (42 microscope-derived features) links catalyst-film uniformity to ORR performance and provides parameter-specific criteria to improve reproducibility and reporting consistency.