Selective semihydrogenation of acetylene for the production of polymer-grade ethylene is a significant chemical industrial process. Facile activization of acetylene and weak adsorption of ethylene are critical requirements for high-performance catalysis. Single-atom alloys (SAAs) have strong binding effect on acetylene and weak effect on ethylene, which have been regarded as the superior catalysts for acetylene semihydrogenation. Herein, we established a pioneering machine learning (ML) assisted approach to investigate the reaction activity and selectivity of 70 SAA catalysts for acetylene semihydrogenation. As the most desirable ML model, the gradient boosting regression (GBR) algorithm has been extended to predict the energy barrier of *C 2 H n ( n = 2-4) hydrogenation with a root-mean-square error (RMSE) of only 0.02 eV. Notably, five candidate SAAs with excellent activity and selectivity for acetylene semihydrogenation are screened out via accessible descriptors. These data of ML prediction have been verified by DFT calculation with a high-accuracy (error less than 0.07 eV). This work demonstrates the potential of ML-assisted approach for predicting the energy barrier of transition state and simultaneously provides a convenient approach for the rational design of efficient catalysts.
Keyphrases
- machine learning
- highly efficient
- big data
- artificial intelligence
- molecular dynamics
- metal organic framework
- transition metal
- density functional theory
- risk assessment
- gold nanoparticles
- molecular docking
- electron transfer
- crystal structure
- reduced graphene oxide
- quantum dots
- aqueous solution
- data analysis
- neural network
- monte carlo