AI-Assisted Electron Microscopy in Structure-Performance Analysis of Advanced Catalysts: From Atomic Resolution to...

Electron microscopy provides direct real-space access to the structural heterogeneity of catalysts across multiple length scales. However, its application to practical catalytic systems is often constrained by imaging artifacts, low signal-to-noise ratios, narrow fields of view, and the difficulty of extracting statistically representative information from large and complex data sets. Recent advances in artificial intelligence (AI), particularly deep learning, are transforming electron microscopy into a quantitative, high-throughput, and increasingly standardized analytical platform. This review highlights how AI enables large-scale, statistically grounded analysis of electron microscopy data, making it possible to extract structurally meaningful descriptors from catalytic materials across atomic, nanoscale, and dynamic regimes and to connect these descriptors more rigorously with structure-performance relationships. Current challenges, including data scarcity, model transferability, interpretability, and integration with spectroscopy, theory, and autonomous microscopy, are also discussed.