Identifying Active Sites for CO₂ Reduction on Dealloyed Gold Surfaces by Combining Machine Learning with Multiscale...

Area of Science:

• Computational materials science
• Electrocatalysis
• Nanotechnology

Background:

• Gold nanoparticles (AuNPs) and dealloyed Au3Fe core-shell nanoparticles show enhanced performance for CO2 reduction to CO (CO2RR).
• Identifying specific active surface sites responsible for this improved CO2RR performance is challenging due to the large number of surface atoms and limitations of traditional experimental and quantum mechanical methods.
• Current methods struggle to analyze the vast surface area of nanoparticles (e.g., 10 nm NPs have ~10,000 surface sites).

Purpose of the Study:

• To develop and apply a computational approach combining machine learning, multiscale simulations, and quantum mechanics (QM) to predict the catalytic performance of individual surface sites on AuNPs and dealloyed Au surfaces.
• To efficiently identify optimal active sites for CO2RR on dealloyed gold surfaces.
• To provide a tool for visualizing the catalytic activity across the entire nanoparticle surface.

Main Methods:

• Integration of machine learning algorithms with multiscale simulations and QM calculations.
• Prediction of the performance (a-value) for thousands of surface sites on gold nanoparticles and dealloyed surfaces.
• Identification of optimal active sites for CO2 reduction reaction (CO2RR) through computational analysis.

Main Results:

• Successfully predicted the performance of individual surface sites on gold nanoparticles and dealloyed gold surfaces.
• Identified optimal active sites for CO2RR on dealloyed gold surfaces with significantly reduced computational cost.
• Developed a method to visualize the catalytic activity distribution over the entire nanoparticle surface.

Conclusions:

• The combined machine learning, multiscale simulation, and QM approach is a powerful tool for understanding and predicting electrocatalyst performance at the site level.
• This methodology enables efficient identification of optimal active sites for CO2RR, accelerating the design of high-performance electrocatalysts.
• Comparing predicted a-values with experimental or theoretical descriptors offers new avenues for guiding the development of advanced electrocatalysts for clean energy conversion.