A Machine Learning-Boosted High-Throughput Screening of Metal–Organic Frameworks for Ethane/Ethylene Separation: From Molecular Simulation to Process Modeling

Ethane/ethylene (C2H6/C2H4) separation is crucial in the petrochemical industry. This study proposes a multiscale framework combining machine learning and molecular modeling to identify optimal C2H6-selective metal–organic frameworks (MOFs) for this separation. A database of 505,973 experimental/hypothetical MOFs underwent geometric/chemical prescreening (254,500 retained). Grand canonical Monte Carlo (GCMC) simulations on sampled MOFs trained a Random Forest model to predict adsorption behaviors, enabling rapid screening of the remaining candidates. This yielded 14,927 MOFs verified using GCMC, with 10 top performers selected. Process-level evaluation via vacuum pressure swing adsorption modeling confirmed their robustness under industrial conditions (varying temperatures, pressures, and gas compositions). Comparative analysis demonstrated superior performance over literature-reported MOFs, particularly under realistic operational parameters. The integrated approach establishes an efficient pathway for MOF discovery, bridging molecular properties with industrial process requirements while reducing reliance on energy-intensive distillation.