Data-driven discovery of biaxially strained single atoms array for hydrogen production

Area of Science:

• Materials Science
• Catalysis
• Computational Chemistry

Background:

• The structure-performance relationship in single-atom catalysts is often obscured by averaged coordination data.
• Periodic arrays of single atoms offer a potential solution for more accurate structural analysis.

Purpose of the Study:

• To develop a data-driven method for screening single-atom catalysts using high-throughput computations and machine learning.
• To identify precise structural features governing catalyst performance.

Main Methods:

• Employed high-throughput density functional theory (DFT) computations.
• Utilized machine learning (ML) to analyze a library of 1248 single-atom sites on strained transition metal dichalcogenides.
• Validated predictions through experimental testing of hydrogen evolution reactions.

Main Results:

• Identified a gold (Au) atom anchored on biaxial-strained molybdenum diselenide (MoSe2) via Au-Se3 bonds as a promising candidate.
• Machine learning analysis revealed four key structural features influencing hydrogen adsorption energy (ΔG_H*).
• The average band center of adsorption sites emerged as a predictor for hydrogen adsorption energy.

Conclusions:

• The study validates the predictive power of the average band center for hydrogen adsorption energy.
• Experimental results demonstrate exceptional 1000-hour stability at 800 mA cm⁻² for Au single-atom arrays on strained MoSe2 in acidic hydrogen evolution.
• An active hotspot involving Au atoms and neighboring Se atoms was identified as crucial for enhanced catalytic activity.