Machine learning-assisted computational exploration of the optimal loading of IL in IL/COF composites for carbon dioxide capture

Abstract

Ionic liquid/covalent organic framework (IL/COF) composites have emerged as a promising type of material for CO₂ capture. The IL loading ratio is a key parameter for the design of IL/COF composites to achieve outstanding separation performance. Presently, the optimal IL gravimetric loading ratio (wt%) is determined by laborious trial-and-error experiments. In this study, we propose to use the volumetric loading ratio (vol%) instead of the wt% as the descriptor of IL loading. The vol% is defined as the ratio of the number of actually inserted IL molecules to the maximum number of IL that can be incorporated into COFs considering the pore volume of COFs. Hence, the first merit of using vol% is that it can avoid pore blocking by excessive wt% loading. The high-throughput computational screening and machine learning analysis of 15 410 ([MMIM][BF₄])/COF composites (covering 18 kinds of vol% and 18 kinds of wt%) show that the composites are most favorable for CO₂/N₂ separation at a uniform vol% of 35%, while no definite optimal wt% can be identified. Hence, the second merit of using vol% is the uniformity of the optimal vol% at 35% in various COFs, which lies in the fact that IL forms continuous, alternating cation–anion interconnected IL nanowires in COFs. It is further revealed that IL and COFs can form CO₂-favorable “wire-tube” and “wall-arm” type structures in COFs with pore sizes <10 Å and ≥10 Å, respectively. The findings obtained in this study will provide valuable guidance for the regulation of IL amounts in IL/COF composites to achieve improved CO₂/N₂ separation performance.