The rational design of high-performance graphene-based single-atom electrocatalysts for the ORR using machine...

Area of Science:

• Materials Science
• Computational Chemistry
• Electrochemistry

Background:

• Developing efficient electrocatalysts for the oxygen reduction reaction (ORR) is crucial for energy conversion technologies.
• Two-dimensional (2D) graphene-based single-atom electrocatalysts offer promising alternatives to traditional catalysts.
• Accurate and rapid screening methods are needed to identify high-performance catalysts from a vast chemical space.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for fast screening of 2D graphene-based single-atom electrocatalysts for ORR.
• To introduce novel descriptors, valence electron correction (VEc) and degree of construction differences (DC), to enhance ML model prediction accuracy.
• To propose new evaluation criteria, high-performance catalyst retention rate (rR) and occupancy rate (rO), for assessing ML model performance in catalyst screening.

Main Methods:

• Machine learning (ML) model construction for predicting electrocatalyst performance.
• Integration of valence electron correction (VEc) and degree of construction differences (DC) descriptors into the ML model.
• Utilizing Density Functional Theory (DFT) for validation of ML-screened candidate catalysts.

Main Results:

• The ML model incorporating VEc and DC descriptors showed improved accuracy, reducing Mean Absolute Error (MAE_test) from 0.334 V to 0.271 V and increasing the coefficient of determination (R^2_test) from 0.683 to 0.774.
• The proposed evaluation criteria, rO and rR, demonstrated significant enhancement, increasing from 0.222 and 0.360 to 0.421 and 0.671, respectively.
• DFT calculations confirmed the accuracy of the ML model, yielding MAE = 0.157 V and R^2 = 0.821 for screened catalysts like ZZ-CoN4 and ZZ-CoN3C1.

Conclusions:

• The developed ML model with novel descriptors and evaluation metrics provides an accurate and efficient approach for screening high-performance 2D graphene-based single-atom electrocatalysts for ORR.
• The findings suggest that ML-driven screening can accelerate the discovery of advanced materials for electrochemical applications.
• The validated approach paves the way for rapid identification of promising electrocatalyst candidates, reducing experimental costs and time.