Machine Learning-Assisted Spray Pyrolysis for the Synthesis of Single-Atom Fe–N–C Porous Hollow Microspheres for Zinc–Air Batteries

Non-noble metal single-atom catalysts with high catalytic activity have garnered considerable attention from researchers in recent years. Yet, their synthesis is affected by various factors, making process optimization a challenging and systematic task. In this study, single-atom Fe–N–C porous hollow microspheres were successfully synthesized via ultrasonic spray pyrolysis, with machine learning employed to optimize the fabrication process. Machine learning models, trained on pre-experimental data, identified the key factors influencing material structure and oxygen reduction reaction (ORR) performance. The resulting Fe–N–C (600–900) material demonstrated excellent ORR activity with a half-wave potential of 0.865 V, along with high stability and methanol tolerance. When applied to traditional liquid zinc–air batteries (ZABs), it achieved an open-circuit voltage of 1.56 V and a maximum power density of 313.4 mW cm–2, with a discharge capacity of 806.5 mAh gZn–1 at 10 mA cm–2, outperforming commercial noble metal catalysts. This work offers valuable insights into the application of machine learning for optimizing ORR catalysts and designing high-performance materials for energy conversion devices.