Machine learning aided high-throughput prediction of ionic liquid@MOF composites for membrane-based CO2 capture

Abstract

Ionic liquid encapsulated metal-organic framework ([email protected]) composites as promising filler used for mixed matrix membranes (MMMs) fabrication to break the trade-off limitation. However, discovering appropriate [email protected] composites effectively and cost-efficiently still faces a great challenge. In this study, we first construct the filler database consisting of 8167 [email protected] composites by inserting [NH2-Pmim][Tf2N] molecule into computation-ready, experimental metal-organic frameworks (CoRE MOFs). Using molecular simulation, we identified the best [email protected] composites based on different metrics and revealed gas separation mechanism. Working with RF model (R2 > 0.72), we uncover that the AV and gASA are key factors in predicting the membrane selectivity and CO2 permeability, respectively. The [NH2-Pmim][Tf2N]@ZIF-67 predicted can be as one of candidates for MMMs fabrication. The experimental results show that CO2 permeability (9536 Barrer) and CO2/N2 selectivity (31.1) of [NH2-Pmim][Tf2N]@ZIF-67/PIM-1 have 121.3% (37.6%) and 32.6% (38.8%) enhancements compared with unfilled PIM-1 (ZIF-67/PIM-1), surpassing the updated CO2/N2 Jansen/McKeown upper bound. Our computational study could offer effective prediction and may trigger experimental efforts to accelerate development of novel [email protected] composites used for fabricating MMMs with excellent performance.