Adsorption of H2S by metal-organic frameworks based on machine learning

• 1、Integrated Circuits School,Beijing University of Posts and Telecommunications,beijing 100876

Abstract：Hydrogen sulfide (H?S) is a highly toxic and corrosive gas, which poses a serious threat to industrial production and safety. Adsorption has become the mainstream technology for H?S removal due to its high efficiency, economy and ease of operation. Among them, metal organic frameworks (MOFs) show great potential in the field of gas adsorption and separation due to their high specific surface area, adjustable pore structure and chemical characteristics. Traditionally, the selection of MOFs mainly relies on experimental and simulation calculations, such as GCMC and DFT, which are computationally expensive and difficult to process large-scale databases efficiently. In recent years, the rise of machine learning (ML) technology has provided new ideas for high-throughput screening of MOFs materials, which can realize the rapid prediction of material properties by constructing the mapping relationship between MOFs structure and adsorption property. In this study, we explored the application of ML in the selection of MOFs materials. Six machine learning models were used to evaluate the H?S adsorption performance of MOFs in hMOFs and CoRE databases, and the best model was selected: Lightweight Gradient Enhanced model (LGBM). Based on the optimized model, the key structural and chemical characteristics that affect the adsorption of H?S by MOFs are analyzed, and the high-throughput prediction is carried out to identify the MOFs materials with excellent H?S adsorption ability, which provides theoretical guidance for the application of MOFs in natural gas desulfurization and other fields.

Keywords： Metal-organic framework Machine learning Light gradient boosting machine Adsorption Molecular simulation