Machine Learning-Driven Prediction of Electrochemical Promotion in the Reverse Water Gas Shift Reaction

Area of Science:

• Catalysis
• Materials Science
• Chemical Engineering

Background:

• Electrochemical promotion of catalysis (EPOC) enhances catalytic reactions like the reverse water-gas shift (RWGS).
• Predicting EPOC performance for new materials and conditions is currently challenging.
• RWGS reaction is crucial for efficient CO2 hydrogenation to syngas.

Purpose of the Study:

• Develop a machine learning framework to predict EPOC behavior and rate enhancement ratios (ρ).
• Utilize catalyst, reaction, and electrochemical descriptors for prediction.
• Facilitate the discovery of novel catalytic materials for EPOC in RWGS.

Main Methods:

• Compiled a dataset from existing literature on EPOC systems.
• Trained and tested various classification and regression models.
• Validated the best models (Random Forest and XGBoost) with new experimental data.

Main Results:

• Achieved high prediction accuracy with R² of 0.97 and MSE of 0.01 using RF and XGBoost models.
• Successfully validated the machine learning framework with experimental data from LLTO electrolyte and Pt-ZnO catalysts.
• Demonstrated the models' ability to predict rate enhancement ratios (ρ).

Conclusions:

• The developed data-driven machine learning approach is interpretable and generalizable.
• This framework accelerates the development of advanced catalytic materials for EPOC in RWGS reactions.
• Offers a powerful tool for optimizing catalytic processes under milder conditions.