Machine Learning-Enhanced Design of 2D TM₃(HXBHYB)@MOF-Based Single-Atom Catalysts for Efficient Oxygen...

Area of Science:

• Materials Science
• Computational Chemistry
• Electrochemistry

Background:

• Two-dimensional (2D) metal-organic frameworks (MOFs) are emerging materials for catalysis.
• Investigating their oxygen electrocatalytic activity is crucial for energy applications.
• Systematic screening of various transition metal (TM) and ligand combinations is needed.

Purpose of the Study:

• To integrate machine learning (ML) and density functional theory (DFT) for predicting oxygen electrocatalytic activity.
• To identify promising 2D TM3(HXBHYB)@MOF materials for oxygen reduction reaction (ORR) and oxygen evolution reaction (OER).
• To establish design principles for high-performance electrocatalysts.

Main Methods:

• Construction of stable 2D TM3(HXBHYB)@MOF systems using various transition metals and ligands (HIB, HHB, HTB, HSB).
• Application of ML models, including Random Forest Regression (RFR), to correlate material properties with ORR/OER overpotentials.
• Utilizing SHAP analysis to identify key descriptors influencing catalytic activity.

Main Results:

• The RFR model demonstrated superior performance in predicting electrocatalytic activity.
• Identified Co3(HXBHYB) and Ir3(HXBHYB) as promising candidates.
• Co3(HHBHSB) and Co(HIB)2 showed exceptional ORR (ηORR = 0.276 V) and OER (ηOER = 0.294 V) activities, respectively.
• Valence electron count and atomic radius of TM were identified as critical descriptors.

Conclusions:

• The study provides a high-precision, low-cost method for screening electrocatalysts.
• Established universal design principles for evaluating ORR/OER activities in 2D MOFs.
• Highlights the potential of ML-DFT integration for accelerated materials discovery in catalysis.