Predicting catalytic pathways for Thiophenol decomposition on TM-doped MoS₂: a comparative machine learning study

Area of Science:

• Materials Science
• Catalysis
• Computational Chemistry
• Machine Learning

Background:

• Thiophenol (TP) is a toxic industrial and pharmaceutical compound requiring efficient catalytic degradation.
• Two-dimensional molybdenum disulfide (MoS₂) has potential for catalysis but suffers from low thiophenol adsorption.
• Single-atom catalysts offer enhanced activity but require strategic design.

Purpose of the Study:

• To design and investigate single-atom catalysts based on transition metal (TM)-doped MoS₂ for thiophenol decomposition.
• To enhance the adsorption and catalytic activity of MoS₂ for toxic organic pollutant removal.
• To develop and validate machine learning models for predicting catalytic performance.

Main Methods:

• First-principles calculations were employed to study the electronic structure and catalytic properties of TM-doped MoS₂.
• Density Functional Theory (DFT) was used to calculate adsorption energies and activation barriers for TP decomposition.
• Four machine learning models (Linear Regression, K-Nearest Neighbors, Random Forest, Gradient Boosting Regression Trees) were evaluated for predicting key reaction parameters.

Main Results:

• Transition metal doping significantly alters MoS₂'s local charge density, enhancing thiophenol adsorption and catalytic activity.
• Nickel (Ni)-doped MoS₂ was found to be kinetically favored, while Cobalt (Co)-doped MoS₂ was thermodynamically favored for TP decomposition into H₂ and H₂S.
• Random Forest regression demonstrated the highest accuracy in predicting activation barriers and reaction energies among the evaluated machine learning models.

Conclusions:

• Transition metal doping is an effective strategy to create highly active single-atom catalysts for toxic organic pollutant decomposition.
• Ni- and Co-doped MoS₂ show significant promise for efficient thiophenol degradation.
• Machine learning, specifically Random Forest, offers a powerful tool for accelerating the screening and design of novel catalysts.