Machine learning-driven discovery of high-performance Co/Sr-free air electrodes for protonic ceramic electrolysis cells

作者：Ning Wang, Qingwen Su, Huanxin Xiang, Biyao Zhou, Jiaxin Huang, Fangyuan Zheng, Ruijie Zhu, Yoshitaka Aoki, Ling Meng, Feng Jiao, Baoyin Yuan, Chunmei Tang, Siyu Ye · 发表于：eScience · 年份：2025 · DOI：10.1016/j.esci.2025.100486 · 被引用次数：5 · 研究领域：Advancements in Solid Oxide Fuel Cells、Fuel Cells and Related Materials、Gas Sensing Nanomaterials and Sensors

Protonic ceramic electrolysis cells (PCECs) have emerged as a promising solid-state ion device, attracting considerable attention for efficient hydrogen generation. However, PCECs face multiple constraints, most notably the lack of high-performance Co/Sr-free air electrodes. Because the current widely used air electrodes generally contain Co and Sr elements, they encounter serious problems of thermochemical expansion and Sr segregation. Traditional air electrode development primarily relies on experience-guided experiments and trial-and-error methods, which is time-consuming and inefficient in exploring vast material compositional spaces. The introduction of machine learning (ML)-driven discovery of air electrodes provides a transformative new approach to the fast development of PCECs. Here, based on the well-constructed ML models, high-performance novel Co/Sr-free Ca 0.5 La 0.5 Fe 1- x Sc x O 3- δ oxides are successfully designed as the air electrodes of PCECs. Among them, Ca 0.5 La 0.5 Fe 0.8 Sc 0.2 O 3- δ (CLFS0.2) oxide demonstrates exceptional properties, including the high hydrated proton defects amount (0.143 mol unit −1 at 550 °C), high catalytic activity, and low thermal expansion coefficient. As a result, PCEC with CLFS0.2 air electrode achieves a current density of 1.58 A cm −2 at 1.3 V at 650 °C, which is higher than the most of Co-free air electrodes and rivals mainstream Co/Sr-containing air electrodes. • A robust ML framework was employed to discover high-perfo...