A Combined Theoretical Calculations And Machine Learning Study On The Structure And Surface Reaction Of Cobalt-Based Fischer-Tropsch Catalysts Under Reaction Conditions

Fischer-Tropsch synthesis（FTS）is a technology which converts syngas（CO and H₂）into liquid fuels and high-value-added chemicals,offering an effective solution to the energy resource.Cobalt-based FTS catalysts,with superior catalytic activity,excellent stability and low water gas shift activity,have been widely used in industrial FTS production.However,despite years of research on cobalt-based FTS catalysts,the"structure-activity"relationships remain controversial,and criticle scientific questions,such as the structure of active site,reaction mechanism,and the essential factors affecting product selectivity,are yet to be fully understood.This is mainly because the reaction atmosphere and conditions have a significant impact on the structure of cobalt-based catalysts in real reactions.Restricted characterization methods also make it difficult to probe the structure and state of cobalt-based catalysts under real FTS reaction conditions.Consequently,most of the literature establishing the"structure-activity relationships"on cobalt-based FTS catalysts based on ex situ characterization results or guessed results.Therefore,studying the structure and catalytic properties of cobalt-based FTS catalysts under real reaction conditions and revealing their inherent connections is immensely significant.Based on the previous experiment on cobalt-based FTS catalysts in our research group,A combined density functional theory（DFT）calculations with machine learning（ML）methods were applied,and followed a research idea that focuses on"establishing the surface structure of the catalyst under reaction conditions,to study the FT reaction mechanism on the catalyst surface under reaction conditions,and analyze the essential factors of reaction conditions on cobalt-based FTS reactions".The surface structure of cobalt catalysts under reaction atmosphere conditions,the state of surface adsorbates and the surface reactions were investigated.The surface structure of the catalyst under real reaction conditions was established,and the adsorption state and coverage of reactant species on the catalyst surface were identified.The influence of surface adsorbates on critical elementary steps of the FT reaction was revealed,such as the dissociation of C-O bonds and the coupling of C-C bonds,and the essential factors affecting the catalytic performance of cobalt-based catalysts under reaction conditions were clarified.The results of this study are expected to provide new insights into the"structure-activity relationships"of cobalt-based FTS catalysts under in situ reaction conditions,and provide theoretical guide for designing highly active and selective FT catalysts.The main research contents of this paper are as follows:（1）Firstly,a combined DFT calculations,thermodynamic phase diagram calculations and molecular dynamics simulations were used to determine the most reasonable surface structure of the catalyst under reaction conditions.By analyzing the effect of reaction temperature and water vapor pressure on the bulk and surface structures of cobalt-based catalysts using DFT calculations and thermodynamic phase diagram analysis,hcp metallic Co was confirmed to be the most stable phase of cobalt-based catalysts under FTS reaction conditions,and the most stable surface structure was the hcp Co（0001）surface.Moreover,the Co（0001）surface remained the most stable surface structure under different reactant coverage conditions.Further research explored the reactant coverage of Co（0001）surface under the reaction atmosphere.241 different models of reactant adsorption at different coverages were constructed,and a thermodynamic phase diagram of reactant coverage was established.Combining with the global optimization method,the thermodynamically most stable adsorption structure of reactant species on the catalyst surface under reaction conditions was determined（CO coverageθ=0.583）.Molecular dynamics simulations were carried out to further verify the stability of the surface adsorption structure ofθ=0.583 under FTS reaction temperature conditions.Electronic structure calculations showed that the surface adsorption layer formed under reaction conditions had a significant impact on the surface electronic properties of cobalt catalysts.Part of the surface electrons of Co atoms were transferred to the CO molecule,leading to a decrease in the charge density on the Co surface.The calculation of the density of states also revealed that the number of empty orbitals on the Co surface increased after the formation of the surface adsorption layer.（2）Based on the already established surface structure of the catalyst under reaction conditions,the surface FTS reaction mechanism was investigated.The effects of surface reactant coverage on the CO direct dissociation,H-assisted CO dissociation,and C-C coupling reaction steps on the Co（0001）surface was calculated.The results revealed that,as the CO coverage increased,the CO direct dissociation process on the Co（0001）surface became more difficult.The increase of CO coverage during the H-assisted CO dissociation and direct dissociation processes was also not conducive to the breaking of C-O bonds.In the process of C-C coupling,as the CO coverage increased,the coupling energy barrier showed an increasing and then decreasing trend.When a reactant adsorption layer with a coverage of 0.583 was formed on the Co（0001）surface under reaction conditions,compared with the clean Co（0001）surface,the surface electron-donating ability and CO adsorption ability weakened,which were not favorable for the breaking of C-O bonds but were conducive to the C-C coupling process.This indicates that the surface reactant coverage layer formed on the cobalt-based catalyst under real reaction conditions has an important impact on the elementary steps of the FTS reaction.（3）Finally,machine learning method was applied to analyze the essential factors affecting cobalt-based FTS reactions under the surface reactant coverage formed under reaction conditions.Linear regression,ridge regression,LASSO regression,support vector regression,and random forest regression models were trained to predict the C-C coupling energy barrier under different CO coverages.The results showed that the support vector regression model had the best predictive performance.Feature analysis was performed on this model,and partial dependence plots were generated.The results showed that higher formation energy of intermediate species and CO adsorption energy lead to lower coupling energy barrier.In addition,farther distance between Co and the C or O atoms in the intermediate species lead to lower coupling energy barrier.The values of the formation energy of intermediate species,CO adsorption energy,and the distance between Co and the C atom in the intermediate species were found to be well correlated with C-C coupling energy barrier,with an R² value of 0.73.The C-C coupling energy barriers under the conditions of 0.583 and 0.028 coverages were distributed at the two ends of the fitted line,indicating that high surface coverage was conducive to surface coupling.Combining theoretical calculations with machine learning analysis,the results could be concluded that,as the surface CO coverage increased,the surface adsorption of CO weakened and the stability of intermediate species on the surface decreased,resulting in a decrease in the surface coupling energy barrier.