Machine Learning-Guided Coordination Engineering of M-N-C Single-Atom Electrocatalysts for Superior Oxygen Reduction

Area of Science:

• Materials Science
• Catalysis
• Computational Chemistry

Background:

• Tailoring metal-nitrogen-carbon (M-N-C) single-atom catalysts (SACs) for high activity and selectivity is crucial but challenging.
• A key limitation is the absence of a descriptor linking catalytic properties to coordination geometries.

Purpose of the Study:

• To develop an interpretable descriptor for M-N-C SACs' electrocatalytic oxygen reduction reaction (ORR) activity.
• To guide the rational design of SACs with optimized catalytic performance.

Main Methods:

• Integration of density functional theory (DFT) calculations and machine learning (ML) algorithms.
• Correlation of theoretical overpotentials with structural features (coordination number, metal-support interaction).

Main Results:

• An interpretable descriptor was identified, accurately predicting ORR activity across various metal centers.
• The descriptor successfully guided the synthesis of a Cu-SAC with low-coordinated Cu-N3 sites.
• The synthesized Cu-N3 SAC exhibited superior ORR activity and stability compared to Cu-N4 sites.

Conclusions:

• The study establishes a powerful structure-property relationship for M-N-C SACs.
• Interpretable ML combined with theoretical and experimental methods accelerates the development of advanced electrocatalysts.
• This approach offers a pathway for designing next-generation catalysts for sustainable energy conversion.