Membraanconservering en databeheer: ontbrekende bouwstenen voor de ontwikkeling van solvent-resistente nanofiltratie

Membrane technology has become a promising separation method, capable of drastically reducing energy- and waste-intensity of industrial processes. Solvent-resistant nanofiltration (SRNF) membranes are applied for the separation of low molecular weight compounds (200 - 1000 g/mol) from solvent-based feed streams encountered in e.g., pharmaceutical and chemical industries. In these applications, SRNF has the potential to replace energy-intensive distillation processes or enable novel processing and synthesis routes. However, several barriers are presently still hindering commercial SRNF adoption, of which two will be specifically addressed in this manuscript. First, the industrial implementation of SRNF is poorly documented, as is the membrane upscaling process. This work aims to facilitate the latter by investigating membrane preservation, a necessary step in industrial membrane production. Preservation entails all treatments necessary to safeguard membrane integrity and performance throughout storage, module preparation, and transport. Chapter 2 provides an overview of membrane preservation practices, emphasizing dry preservation over storing membranes in a wet state. As pore collapse during drying is detrimental to membrane performance, three methods are evaluated to mitigate its effects i.e. solvent exchange (SE), pore preserving agents (PPAs) and freeze-drying. Consulting industry professionals provided insights into challenges and shortcomings of preservation methods, resulting in identification of priorities for future research. SE relies on replacing the pore liquid in the membrane prior to drying. However, no empirically validated relationship between liquid properties and pore collapse is currently available, resulting in SE yielding average and unpredictable results. A fundamental investigation in SE is presented in Chapter 3. Regression analysis indicated that surface tension, viscosity and chemical affinity between liquid and membrane determined permeance and rejection of dried membranes, pointing at their role in driving pore collapse. Additionally, the SE method determined the concentration of trace exchange liquids during drying, affecting pore collapse. Physical characterization showed that pore collapse resulted in a decrease in surface pore size and a simultaneous increase in surface roughness and skin layer thickness. PPAs are the state-of-the-art preservatives used in water purification membranes. They are non-volatile compounds deposited in the membrane pores to inhibit pore collapse after which they are washed out again prior to end-use of the membrane. In Chapter 4, several PPAs are investigated for SRNF applications. The efficiency of PPA preservation depended on the washout efficiency which was related to the solvent used for washout and the affinity of the PPA for the membrane polymer. A decision framework was introduced to guide selection of suitable PPAs and washout procedures based on the chemical affinities between membrane, PPA and filtration solvent. Chapter 5 explored the application of polyethyleneimine (PEI) as a high molecular weight crosslinker for SRNF membranes, enabling greener preparation of crosslinked polyimide membranes. It was shown that addition of the PEI in the coagulation bath during the membrane preparation process influenced morphology and filtration performance of the resultant membrane as it affected the phase inversion during the SIM crosslinking procedure. Drying these membranes made them brittle and resulted in extremely low permeance, suggesting that PEI was unable to provide a preservation effect in contrast with what has been illustrated for other types of polyamine crosslinkers. A second roadblock hindering development of SRNF processes is the scattered and non-standardized nature of available filtration data. A findable, accessible, interoperable, reusable, and open (FAIR/O) dataset was constructed containing over 5000 unique filtrations to facilitate benchmarking, verification, exploration and replication of research data. Curation of this dataset is presented in Chapter 6, outlining general constraints, parameter selection, and the data standardization process. A brief analysis of the collected data is shared to gauge the state of the field and inspire potential use cases for the dataset. Overall, this dissertation aims to lower technical barriers holding back SRNF adoption. The gained insights on membrane preservation are expected to directly translate to pilot-scale membrane production, hence accelerating their challenging scale-up process. The FAIR/O database will enable identification of trade-off and synthesis-structure-property relationships, allow machine actionability, and increase approachability of SRNF to a non-expert audience.