Niche membrane technologies, such as organic solvent nanofiltration (OSN), offer considerable energy and operation cost reduction compared with conventional separation methods. However, despite their many advantages, their industrial implementation is hindered by small and specialized datasets, which hinders the development of more advanced prediction methods. In this study, we developed a medium-throughput system (MTS) for OSN with high robustness and low error. The MTS was used to generate a dataset containing 336 different molecules, and their rejection values were measured at two different pressures using three commercial DuraMem polyimide membranes with different molecular weight cut-off values in methanol. The diversity of the generated dataset was compared with the diversity values of other relevant datasets using 26 different chemometric molecular descriptors, including the heteroatom count, topological surface area, different shape descriptors, Van der Waals volume, logP, and logS. The rejection was found to be weakly dependent on the functional group and molecular weight at the lower end of the nanofiltration range. We proposed the use of a novel structural similarity-based indexing method for comparing solutes. Also, we established the first open-access and searchable dataset for OSN rejection values. The newly established www.osndatabase.com pilot website acts as the foundation of the dataset.
ASJC Scopus subject areas
- Filtration and Separation
- Materials Science(all)
- Physical and Theoretical Chemistry