Screening Of Platinum-based Bimetallic Catalysts For Propane Dehydrogenation By First-Principles Calculation And Machine Learning

Platinum-based catalysts are widely used in the process of propane dehydrogenation.Sintering and coking are the two main adverse factors affecting their catalytic performance.The second metal promoters can effectively improve the catalytic performance of the Pt-based catalyst and prevent the occurrence of particle sintering and coke deposition to a certain extent.In this paper,the selection of Pt-based bimetallic catalysts for propane dehydrogenation were studied in detail by means of first-principles calculation,machine learning and microkinetic simulation:(1)Six kinds of catalyst structures with different proportions and configurations were established to fully describe the bimetallic catalyst structures.After adding metal promoters,the adsorption energy of propane and propene on the catalyst surface and the C-H bond activation barrier of propane and propene were calculated.The data of adsorption energy and barrier from first-principles calculation were streamed into five machine learning methods including Gradient Boosting Regressor(GBR),K-Neighbors Regressor(KBR),Random Forest Regressor(RFR),Adaboost Regressor(ABR),the Sure Independence Screening and Sparsifying Operator(SISSO).Root mean square error and coefficient of determination indicated that Gradient Boosting Regressor and SISSO had the most optimal performance.Furthermore,it was found that descriptors which derived from intrinsic properties of metal promoters could determine their properties.The optimal machine learning model was used to predict the adsorption energy and C-H bond activation energy barrier of uncalculated doping configurations.The adsorption energy and energy barrier on the surface of Pt-based catalysts were used as screening thresholds,and all the predicted results were preliminatively screened.The predicted results were verified by turnover frequency calculation and electronic structure analysis,and it was confirmed that the platinum-based catalysts doped with Mo atoms in 3:1 ratio had the best catalytic performance.(2)The kinetic properties of the Pt-based catalytic system before and after Mo doping were studied by first-principles calculation and microkinetic simulation,and the effects of Mo addition on the activity,selectivity and side reactions of propane dehydrogenation were investigated.By density functional theory calculation,the corresponding data of reaction energy and transition state barriers were obtained,which were used as input data for the microkinetic simulation.The analysis of turnover frequency,reaction order,degree of rate control and coverage of the catalytic system confirmed the positive effect of Mo on the catalytic system,and Mo effectively improved the catalytic activity and selectivity.The accuracy of Mo selected by machine learning above was verified from the perspective of kinetics,which provided a more sufficient theoretical basis for the design of platinum-based catalysts.