Revealing the effect of structure curations on the simulated CO2 separation performances of MOFs

Abstract

Experimentally reported metal organic frameworks (MOFs) may have structural issues such as the presence of solvent molecules in their pores, missing hydrogen atoms in the frameworks and/or absence of charge balancing ions, which all require curation of structures before using them in molecular simulations. The development of computation-ready MOF databases significantly accelerated the assessment of CO2 adsorption by providing directly usable, curated crystal structures for molecular simulations. Each database followed different methods to curate MOFs which caused the same material to be reported with different structural features in databases. In order to understand the role of curated computation-ready MOF databases in the predicted CO2 separation performances of MOFs, we studied various MOFs commonly existing in databases but curated differently in terms of (i) removal of bound solvents, (ii) treatment of missing hydrogens, and (iii) retention of charge balancing ions (CBIs). We used molecular simulations to compute CO2/CH4, CO2/H2, and CO2/N2 mixture adsorption and predicted various separation performance metrics such as selectivity, regenerability (R%), and the adsorbent performance score (APS) for the curated computation-ready MOFs. Our results showed that the CO2 separation performances of MOFs and the identity of the best performing MOFs significantly change depending on the structure curation. For example, removal of coordinated solvents from MOFs resulted in higher simulated CO2 uptakes, selectivities, and APSs compared to the structures having solvents. On the other hand, the absence of CBIs in the frameworks resulted in overestimated CO2 uptakes, APSs, and R%, and underestimated CO2 selectivities compared to MOFs having CBIs. Based on these results, we suggested a path showing how to use the curated, computation-ready MOF structures in highthroughput molecular simulations.