Data‐Driven Screening of Pivotal Subunits in Edge‐Anchored Single Atom Catalysts for Oxygen Reactions

• Published in: Advanced functional materials Vol. 34; no. 28
• Main Authors: Ye, Qitong, Yi, Xingxing, Wang, Cai‐Zhuang, Zhang, Tao, Liu, Yipu, Lin, Shiwei, Fan, Hong Jin
• Format: Journal Article
• Language: English
• Published: Hoboken Wiley Subscription Services, Inc 01.07.2024
• Subjects: Atomic structure; Catalytic activity; Catalytic converters; Chemical reduction; Configurations; data‐driven screening; Edge effect; edge‐anchored single‐atom catalysts; Electronic properties; Electronic structure; Energy conversion; high‐throughput calculations; Machine learning; Oxygen evolution reactions; oxygen reaction; Oxygen reduction reactions; Regression models; Screening; Single atom catalysts; Stability criteria; Structural analysis; Structural stability; Transition metals
• DOI: 10.1002/adfm.202400107
• ISSN: 1616-301X

Oxygen reduction reaction (ORR) and the oxygen evolution reaction (OER) are key reactions in diverse energy conversion devices, highlighting the importance of efficient catalysts. Edge‐anchored single atom catalysts (E‐SACs) emerge as a special class of atomic structure, but the detailed configuration and its correlation with catalytic activity remain little explored. Herein, a total of 78 E‐SACs (E‐TM‐Nx‐C) have been constructed based on 26 transition metal (TM) species with three coordination patterns. Using structural stability and ORR/OER catalytic activity as the evaluation criteria, a few catalytic structures comparable to Pt (111) for ORR and IrO2 (110) for OER are screened based on high‐throughput calculations. The screening results unveil that the E‐Rh‐N4‐C configuration exhibits most efficient bifunctional activity for both ORR and OER with an overpotential of 0.38 and 0.61 V, respectively. Electronic structure analysis confirms the distinctive edge effects on the electronic properties of TM and N species, and the feature importance derived from machine learning illustrates the efficacy of E‐TM‐Nx subunit configuration in determining the catalytic activity of E‐SACs. Finally, the trained Gradient Boosting Regression (GBR) model exhibits acceptable accuracy in predicting the OH intermediates adsorption strength for E‐SACs, thereby paving the way for expanding catalytic structures based on E‐SACs. A data‐driven strategy is formulated to accelerate the discovery of Edge‐SACs by predicting the reactivity trends and structure‐activity relationships. A few outstanding E‐SACs structures on graphene for ORR and OER are screened out. It is shown that edge effects together with coordinated patterns govern their catalytic activity.