Bird's-Eye View of the Activity Distribution on a Catalyst Surface via a Machine Learning-Driven Adequate Sampling...

Area of Science:

• Catalysis
• Materials Science
• Computational Chemistry

Background:

• Understanding catalyst active centers is crucial for rational catalyst design.
• Characterizing the structure and activity distribution of active sites remains a challenge in catalysis.
• Accurate theoretical and experimental investigations are needed to address these challenges.

Purpose of the Study:

• To develop a machine learning-driven adequate sampling (MLAS) framework.
• To obtain a statistical understanding of the chemical environment near catalyst active sites.
• To apply the MLAS framework to the N2 activation process in ammonia synthesis.

Main Methods:

• Implemented combined strategies for efficient sampling: decomposition of degrees of freedom, stratified sampling, Gaussian process regression, and constraint optimization.
• Developed a machine learning-driven adequate sampling (MLAS) framework.
• Utilized computational methods to analyze the N2 activation step.

Main Results:

• Calculated the population function, PA, providing a comprehensive understanding of active centers.
• Demonstrated the MLAS framework's ability to statistically characterize catalyst active sites.
• Successfully applied the framework to the rate-determining step of ammonia synthesis.

Conclusions:

• The MLAS framework offers a novel approach for statistically understanding catalyst active centers.
• This method provides intuitive insights into the distribution and nature of active sites.
• The MLAS framework shows broad applicability to complex catalytic materials and reaction networks.