Spatial-adaptive active learning identifies ultra-durable and highly active catalysts for acidic oxygen evolution...

Area of Science:

• Electrochemistry
• Materials Science
• Catalysis

Background:

• Acidic water electrolysis for hydrogen production faces challenges with oxygen evolution reaction (OER) catalyst activity and stability.
• Current OER catalyst development relies heavily on trial-and-error, hindering progress.

Purpose of the Study:

• To develop an AI-driven, target-oriented approach for optimizing acidic OER catalysts.
• To accelerate the discovery of highly active and stable OER catalysts for efficient hydrogen production.

Main Methods:

• Implemented a two-stage spatial-adaptive active learning strategy with closed-loop experimentation.
• Utilized Bayesian optimization and a conditional variational autoencoder for catalyst screening and subspace generation.
• Employed active learning in the second stage to identify the most stable catalyst within the generated subspace.

Main Results:

• Discovered a novel copper-ruthenium oxide (Cu-RuO2) catalyst with exceptional stability (625 hours).
• Achieved a low overpotential of 177 mV at 10 mA cm-2 for the Cu-RuO2 catalyst.
• Provided detailed characterization and mechanistic insights into the new catalyst's performance.

Conclusions:

• The developed AI strategy significantly accelerates the design of stable acidic OER catalysts.
• This approach enhances the feasibility of large-scale green hydrogen production via acidic water electrolysis.
• The novel Cu-RuO2 catalyst represents a significant advancement in OER catalysis.