Machine learning screening of high-performance single-atom electrocatalysts for two-electron oxygen reduction...

Area of Science:

• Materials Science
• Electrochemistry
• Computational Chemistry

Background:

• Conventional hydrogen peroxide (H2O2) production is energy-intensive and pollutive.
• Electrocatalysis offers a greener alternative for H2O2 synthesis.
• The oxygen reduction reaction (ORR) can produce H2O2 via a 2e- pathway, but competes with H2O production via a 4e- pathway.

Purpose of the Study:

• To develop machine learning (ML) models for predicting the performance of single-atom catalysts (SACs) in the 2e- oxygen reduction reaction (ORR).
• To identify highly selective and active SACs for efficient H2O2 electroproduction.
• To establish a rapid and cost-effective method for discovering novel electrocatalysts.

Main Methods:

• Utilized density functional theory (DFT) calculations for adsorption free energy of O* (ΔG(O*)) and limiting potential (UL) of 149 and 31 SACs, respectively.
• Developed five ML models based on DFT data to identify key descriptors for SAC performance.
• Predicted the 2e- ORR catalytic performance for 690 unknown SACs.

Main Results:

• Identified four promising SACs (Zn@Pc-N3C1, Au@Pd-N4, Au@Pd-N1C3, Au@Py-N3C1) with high selectivity and activity for 2e- ORR.
• DFT calculations validated the ML predictions, showing favorable limiting potentials for the screened SACs.
• The ML-based approach demonstrated high accuracy in predicting material properties for electrocatalyst discovery.

Conclusions:

• Machine learning provides an efficient, rapid, and low-cost strategy for discovering and designing SACs for H2O2 electroproduction.
• The developed ML models and identified SACs pave the way for greener and more sustainable H2O2 synthesis.
• This work highlights the potential of integrating ML with DFT for accelerated materials discovery in electrocatalysis.