Machine Learning for Predicting Activation Energies and Understanding the Brønsted–Evans–Polanyi Relation in MXene-Catalyzed Water Gas Shift Reactions
Description
This dataset supports the publication titled "A Machine Learning Perspective on the Brønsted–Evans–Polanyi Relation in Water Gas Shift Catalysis on MXenes", offering machine learning workflows and curated data for predicting activation energies in catalysis. It includes: Comprehensive Dataset on MXene Catalytic Properties (2025): A curated dataset of 92 MXenes, covering single (M₂C, M₂N) and dual-transition-metal (M'₂M''C₂) compositions. For each MXene, density functional theory (DFT)-derived activation energies are provided for four key reactions in the water-gas shift (WGS) process: water (H₂O) and hydroxyl (OH) dissociation, and hydrogen (H₂) and carbon dioxide (CO₂) formation. Additional structural and physicochemical descriptors are included. Machine Learning Models for Activation Energy Prediction (Python Notebook): A Python notebook implementing multiple ML algorithms—Random Forest, Gradient Boosting, ANN, SVM, Decision Tree, and KNN—to predict activation energies. The Random Forest Regressor (RFR) is identified as the best-performing model based on accuracy and robustness. The notebook includes full data processing, training, testing, and feature importance analysis. Key Insights: Reaction energy and LogP (logarithmic partition coefficient) are the most predictive features. The findings support and quantify the Brønsted–Evans–Polanyi (BEP) relationship, linking thermodynamics and kinetics of catalytic reactions. This dataset and analysis aim to guide the design of efficient MXene-based catalysts using interpretable machine learning approaches.
Files
Institutions
- Universidade de Aveiro CICECO