Multivariate Metal Organic Frameworks for High-Efficiency H2S/CO2/CH4 Adsorption and Separation: A Combined Molecular Simulation and Machine Learning Study.

Natural gas with high sulfur content frequently contains harmful impurities, such as CO2 and H2S, which poses a critical demand for the development of highly efficient and selective adsorbent materials to achieve effective purification. In this study, we have systematically investigated and evaluated the adsorption and separation capabilities of metal-organic frameworks (MTV-MOFs) for H2S/CO2/CH4 ternary mixture by combining molecular simulation with machine learning techniques. Grand Canonical Monte Carlo (GCMC) simulations reveals that the framework functionalized with the NH2-F-F group not only exhibits excellent H2S adsorption capacity, but also has excellent separation ability for ternary mixed gas. The cuf_6586 containing functional group NH2-F-F can achieve an adsorption capacity of 18.12 mmol/g under the conditions of 298 K and 1 bar. Based on Henry's constant and adsorption heat, the interaction mechanism between the gas and the framework was deeply analyzed. For separation, cuf_10289, which also contains functional group NH2-F-F, is identified as the most effective separation adsorbent. Within the pressure range of 0.1-5 bar, its selectivity for H2S and CO2 is higher than 45. Furthermore, machine learning models were constructed using structural, chemical, and thermodynamic descriptors to predict adsorption capacity and selectivity. Among them, the XGBoost model achieved excellent performance, with an R2 of 0.93 for regression tasks and a precision of 0.90 for high-performance material classification. SHAP analysis further confirmed the dominant role of thermodynamic descriptors in driving model predictions. This study can provide some theoretical references and an efficient strategy for the high-throughput screening and rational design of advanced adsorbents targeting the purification of high-sulfur natural gas.