Discovering Electron-Sponge Behavior at Organic-Metal Interfaces for CO₂ Electroreduction via Machine Learning.

Area of Science:

• Electrochemistry and Materials Science
• Computational Chemistry and Machine Learning

Background:

• Molecular regulation at organic-metal interfaces is key for C─C coupling in CO2 electroreduction, impacting multi-carbon (C2+) product selectivity.
• Establishing predictive quantitative structure-activity relationships (QSAR) is challenging due to complex molecular descriptor interplay, limiting mechanistic understanding.

Purpose of the Study:

• To develop an interpretable machine learning (ML)-QSAR framework to correlate molecular features with C─C coupling free energy barriers (ΔG‡) on copper (Cu) surfaces.
• To elucidate the dominant interfacial 'electron-sponge' mechanism governing CO2 electroreduction.

Main Methods:

• An interpretable ML-QSAR framework was employed to link molecular descriptors with the C─C coupling free energy barrier (ΔG‡) on Cu surfaces.
• Shapley Additive Explanations (SHAP) analysis identified key electronic descriptors: minimal local electron affinity (LEAmin), HOMO-LUMO gap, and HOMO energy.
• A representative molecule, 3,4-diaminofurazan (DAF), was selected and synthesized.

Main Results:

• The study uncovered an 'electron-sponge' mechanism where modifier molecules donate electrons to Cu, facilitating intermediate stabilization and C2+ product formation.
• Key electronic descriptors (low LEAmin, narrow HOMO-LUMO gap, elevated HOMO energy) were identified as crucial for reducing ΔG‡.
• Experimental validation with DAF showed a significant increase in C2+ Faradaic efficiency from 42% to 77%.

Conclusions:

• The developed ML-QSAR framework effectively predicts molecular performance in CO2 electroreduction, driven by the electron-sponge mechanism.
• The descriptor-driven approach offers a scalable pathway for designing efficient next-generation electrocatalysts for CO2 conversion on Cu, Au, and Ag surfaces.