Leveraging data mining, active learning, and domain adaptation for efficient discovery of advanced oxygen evolution...

Area of Science:

• Materials Science
• Catalysis
• Computational Chemistry

Background:

• Developing efficient electrocatalysts for the acidic oxygen evolution reaction (OER) is essential for sustainable hydrogen production.
• Traditional methods for catalyst discovery are often time-consuming and rely on trial-and-error.
• Advanced multimetallic catalysts offer potential for improved OER performance.

Purpose of the Study:

• To develop and implement a multistage machine learning (ML) approach for streamlined discovery and optimization of complex multimetallic catalysts for acidic OER.
• To reduce reliance on subjective intuition and accelerate the identification of high-performance electrocatalysts.
• To enhance theoretical simulations for deeper mechanistic insights into catalyst performance.

Main Methods:

• Integration of data mining, active learning, and domain adaptation for catalyst discovery.
• Systematic narrowing of the materials exploration space using domain knowledge.
• Iterative experimental feedback to refine element composition and synthesis conditions.
• Application of domain adaptation to improve theoretical simulations and align with experimental findings.

Main Results:

• Discovery of a promising Ru-Mn-Ca-Pr oxide catalyst for acidic OER.
• Demonstration of an efficient pathway for electrocatalyst discovery and optimization using a data-driven approach.
• Enhanced theoretical simulations providing deeper mechanistic insights aligned with experimental results.

Conclusions:

• The presented multistage ML approach offers an efficient and systematic pathway for electrocatalyst discovery and optimization.
• This data-driven methodology represents a paradigm shift and potential benchmark in electrocatalyst research.
• The developed workflow accelerates the identification of advanced catalysts for sustainable energy applications like hydrogen production.