Artificial-intelligence-assisted design principle for developing high-performance single-atom catalysts

Area of Science:

• Materials Science
• Catalysis
• Computational Chemistry

Background:

• Artificial intelligence (AI)-assisted approaches accelerate novel catalyst development.
• Lack of mechanistic understanding in AI approaches can lead to unreliable results.
• Elucidating underlying mechanisms is crucial for transparent and dependable AI-driven discoveries.

Purpose of the Study:

• To develop an AI strategy combining machine learning (ML) and data mining (DM) for identifying high-performance catalysts.
• To elucidate key factors governing catalytic performance in complex reactions.
• To enhance the transparency and reliability of AI-assisted catalyst design.

Main Methods:

• An AI strategy integrating ML and DM was developed.
• The strategy was applied to evaluate 10,179 single-atom catalysts (SACs) for electrocatalytic oxygen reduction.
• Experimental validation was performed to confirm the AI strategy's effectiveness.

Main Results:

• Several high-performance single-atom catalysts (SACs) were identified for oxygen reduction.
• Critical factors influencing SAC activity were determined.
• The optimal Co-S₂N₂/g-SAC achieved a high half-wave potential of 0.92 V.

Conclusions:

• The combined ML and DM AI strategy effectively identifies high-performance catalysts and elucidates performance-governing factors.
• This approach enhances transparency and reliability in data-driven catalyst discovery.
• The findings provide valuable insights for the rational design of advanced materials.