Machine Learning-Assisted Design Framework of Carbon Edge-Dominated Dual-Atom Catalysts for Urea Electrosynthesis

Area of Science:

• Electrochemistry
• Materials Science
• Computational Chemistry

Background:

• Direct electrosynthesis of urea is a key goal but is hindered by complex proton-coupled electron transfer and competing side reactions.
• Designing efficient catalysts requires understanding intricate reaction networks and overcoming limitations of traditional descriptors.

Purpose of the Study:

• To develop a data-driven strategy for designing edge-anchored dual-atom carbon-based catalysts for urea electrosynthesis.
• To identify a universal descriptor for predicting catalyst performance under coadsorption conditions.
• To screen a large chemical space for superior catalysts that favor urea formation over competing reactions.

Main Methods:

• High-throughput density functional theory (DFT) calculations were used to decode reaction networks for 90 heteroatomic metal pairs.
• Machine learning (ML), specifically an XGBoost regression model, was employed to screen candidate catalysts based on intrinsic atomic features.
• A quantitative selectivity phase diagram was constructed using coadsorption energy (Eads(*CO_NO)) as a descriptor.

Main Results:

• Coadsorption energy (Eads(*CO_NO)) was identified as a robust descriptor, outperforming single-molecule adsorption descriptors.
• A narrow thermodynamic window (-3.57 to -3.08 eV) was found to favor the C-N coupling pathway for urea synthesis.
• Zr_Pd@A and Zn_Pd@Z were identified as superior catalysts with downhill thermodynamic pathways, attributed to optimal Pd d-electron density and controlled CO binding.

Conclusions:

• A scalable machine learning-assisted paradigm was established for designing catalysts in complex electrocatalytic systems.
• The developed strategy effectively decouples competitive reaction mechanisms, paving the way for efficient urea electrosynthesis.
• The identified catalysts demonstrate a promising approach for selective urea production via C-N coupling.