Machine-Learning-Assisted Optimization of a Single-Atom Coordination Environment for Accelerated Fenton Catalysis

Area of Science:

• Catalysis Science and Technology
• Materials Science and Engineering
• Computational Chemistry and Materials Informatics

Background:

• Traditional materials optimization methods are often time-consuming and inefficient.
• Machine learning (ML) offers a powerful approach to accelerate materials discovery and optimization.
• Single-atom catalysts (SACs) show great promise in various chemical reactions, including Fenton-like processes.

Purpose of the Study:

• To develop and apply a machine learning methodology for assisting the construction of Fenton-like single-atom catalysts (SACs).
• To identify critical synthesis parameters influencing the Fenton activity of SACs.
• To accurately predict the catalytic performance of SACs for phenol degradation.

Main Methods:

• Development of a machine learning framework encompassing model building, training, and prediction.
• Extraction of influential synthesis parameters affecting Fenton activity.
• Prediction of phenol degradation rate (k) using ML models with a mean error of ±0.018 min⁻¹.

Main Results:

• Identified heating temperature during SAC synthesis as a significant factor influencing the Fe-N coordination number and catalytic performance.
• Achieved accelerated learning and an extended synthesis window through ML-guided optimization.
• Discovered a highly efficient SAC, dominated by Fe-N₅ sites, exhibiting exceptional Fenton activity (k = 0.158 min⁻¹).

Conclusions:

• Machine learning provides an effective strategy for optimizing single-atom coordination environments in catalysts.
• The developed ML approach significantly accelerates the development of high-performance Fenton-like SACs.
• This work demonstrates the feasibility of ML in advancing catalyst design and discovery.