Accelerating Sustainable Energy Catalysis through Machine Learning: High-Throughput Screening and Machine Learning-Based Potential Energy Surface

Over the past few decades, the research community has witnessed an explosive growth in machine learning (ML) technologies, driven largely by the integration of diverse complex data and advancements in image classification tasks. Among these, High-Throughput Screening (HTS) and Potential Energy Surface (PES) fitting methods have garnered significant attention in the field of energy catalysis due to their high processing efficiency and economic feasibility. Unfortunately, catalytic experiments still largely rely on slow and inefficient trial-and-error approaches, which contribute substantially to exorbitant costs and environmental pollution associated with experimental characterization. In response to this situation, this review focuses on the application of ML-based HTS and PES approaches in catalysis, with particular emphasis on their roles in catalyst design and reaction mechanism studies. We further evaluate available strategies to provide research paradigms. Finally, we tentatively outline the current bottlenecks facing HTS and PES theories, aiming to facilitate their broader practical application.