Generation And Screening Of MOFs With High Methane Adsorption Rate Based On Generation Models

Porous organic materials such as metal organic frameworks(MOFs)and covalent organic frameworks(COFs)have huge internal surface area,high porosity and thermal stability,and their excellent adsorption performance has been proved to be expected to solve the problems of industrial gas storage and transportation.Due to its modular design and universal chemical properties,millions of possible nanoporous materials can be synthesized at present.It is very difficult to tailor or fine-tune materials to meet specific adsorbent molecules and conditions in such a vast chemical space.It is not feasible to screen out high-performance gas storage materials,and even use molecular simulation to calculate the material's gas adsorption capacity.Therefore,it is necessary to find other methods that can identify high-performance materials in a reasonable time.In this article,the research combines existing molecular simulation calculations and machine learning methods,and use machine learning to assist in accelerating the discovery of new materials for designated gas storage under specific conditions.In the first part,the research studied the adsorption of acetylene(C₂H₂)by MOFs.Due to the low-pressure storage limitation of C₂H₂,it has become an urgent need for the industry to store C₂H₂ efficiently in a conventional environment.Because MOFs containing open metal sites are important materials for C₂H₂ storage,the study first synthesized more than 7,000 MOFs containing Cu metal using tobacco software,and then used machine learning to explore the relationship between C₂H₂ storage capacity and material characteristics.The classification model is used to obtain the design rules for producing the best material,and the regression model can accurately predict the adsorption capacity of MOFs for C₂H₂ based on the structural characteristics of MOFs.The second part of this article mainly introduces the research of COFs'working ability on methane(CH₄).The main obstacle of machine learning algorithms is that the performance of the algorithm is affected by many design decisions.For non-professionals,finding the right algorithm and model parameters is a challenge.This study uses Automatic Machine Learning(AutoML)to analyze its ability to adsorb CH₄ based on 403,959 COFs.Use the tree-based pipeline optimization tool(TPOT)in AutoML to compare with the optimal model obtained by parameter adjustment in traditional machine learning.It is found that TPOT can not only save complex data preprocessing and model parameter adjustment,but also obtain higher performance than traditional ML models.Compared with traditional large-standard Monte Carlo simulation,it can save a lot of time.AutoML enables non-professionals to use machine learning to conduct experiments and obtain highly accurate and easyto-understand results,which is of great significance to promote research in the field of materials.