Integrating Machine Learning and DFT Descriptors for Screening Dual Metal-Site Catalysts for CO₂ Reduction to C₂...

Area of Science:

• Electrochemistry and Catalysis
• Materials Science
• Computational Chemistry

Background:

• Electrochemical reduction of carbon dioxide (CO2RR) is a promising sustainable route for producing valuable multi-carbon (C2) products.
• Achieving high activity and selectivity in CO2RR remains a significant challenge, particularly for C2 product formation.
• Dual metal-site catalysts (DMSCs) offer unique active centers for enhanced CO adsorption, a crucial step for C-C coupling.

Purpose of the Study:

• To systematically investigate a library of 156 dual metal-site catalysts (DMSCs) supported on nitrogen-doped carbon for CO2RR.
• To identify promising DMSCs with high activity and selectivity for C2 product generation.
• To establish predictive models for CO2 dimerization energy and C2 production overpotential based on electronic descriptors.

Main Methods:

• Computational screening of 156 DMSCs for stability and hydrogen binding energy.
• Calculation of CO dimerization energies (∆G*CO dimer-2*CO) for promising candidates.
• Linear correlation analysis and machine learning (Random Forest Regressor) to identify key electronic descriptors and predict performance.

Main Results:

• 33 DMSCs showed favorable CO dimerization energies (< 0.75 eV), indicating potential for C2 product formation.
• The occupancy of the dz²↓ orbital (O-dz²↓) was identified as a key descriptor correlating with CO dimerization energy (R² = 0.69).
• A Random Forest Regressor model achieved high accuracy (R² = 0.96) in predicting CO dimerization energy, and O-dz²↓ strongly correlated with C2 overpotential (R² = 0.87).

Conclusions:

• Dual metal-site catalysts show significant promise for selective electrochemical CO2 reduction to C2 products.
• Electronic descriptors, particularly the occupancy of the dz²↓ orbital, are crucial for predicting catalyst performance.
• Machine learning models provide a powerful framework for rapid screening and design of efficient CO2RR catalysts.