Manufacturing supported loose-nanofiltration polymeric membranes with eco-friendly solvents on an R2R System

Membrane separations offer lower energy use, operating flexibility, and tailored selectivity compared to traditional separations. Loose nanofiltration (LNF) membranes, positioned between ultrafiltration (UF) and tight nanofiltration (NF), can remove organic macromolecules with higher permeability than tight NF while maintaining higher selectivity than UF. Scaling up LNF manufacturing while adhering to eco-manufacturing principles is a key challenge because conventional nonsolvent-induced phase separation (NIPS) often uses hazardous, petroleum-derived solvents (e.g., NMP, DMF, DMAc). Eco-friendly solvent alternatives such as Rhodiasolv PolarClean and gamma-valerolactone (GVL) are biodegradable, less toxic, water-miscible, and commercially available, and a 3:1 PolarClean:GVL mixture has produced PSf membranes with performance comparable to traditional solvents. Bench methods like doctor blade extrusion (DBE) are common but are less suitable for roll-to-roll (R2R) scale-up. Slot die coating (SDC) integrates well with R2R and can enable continuous, uniform films, yet its application to polymeric membrane manufacture—especially direct coating onto porous support layers for clean water applications—remains limited. This study aims to demonstrate formation of PSf LNF flat-sheet membranes via SDC on an R2R platform using eco-friendly solvents, determine a coating window from dope solution properties, incorporate a nonwoven PET support layer with pre-wetting to manage penetration and adhesion, and compare the structure–performance of SDC membranes against DBE analogs during BSA and sodium alginate (SA) filtration.

Polymeric membranes are widely produced by NIPS, yielding asymmetric structures with selective skin layers and porous sublayers. DBE, spin coating, and dip coating dominate bench-scale fabrication, but DBE is not ideal for R2R continuity and uniformity. SDC, a widely used thin-film coating method, allows continuous deposition on moving substrates and better control of film thickness within a defined coating window governed by flow rate, substrate speed, viscosity, surface tension, and wettability. Support layers (e.g., nonwoven PET) increase mechanical integrity and affect morphology and performance; support choice can alter solvent–nonsolvent exchange, leading to sponge-like or finger-like morphologies. Eco-manufacturing emphasizes waste reduction and eco-friendly materials. Traditional solvents (NMP, DMF, DMAc) pose health and environmental risks and increase carbon footprint. Green dipolar aprotic solvents like PolarClean (a valorized product from nylon 6,6 byproducts) and GVL (from lignocellulosic biomass) are biodegradable and water-miscible. Prior work shows PSf membranes made with PolarClean–GVL (3:1) can match performance of DMAc/NMP-based membranes and enhance durability versus using either solvent alone. Limited studies exist on SDC-R2R fabrication of membranes with eco-friendly solvents; earlier work demonstrated UF membranes by SDC with properties comparable to DBE on glass substrates, but detailed relationships among dope properties, coating parameters, defect formation, and direct coating onto porous supports remained largely unexplored.

Materials: PSf (Mw ~35,000), PolarClean, GVL (3:1 v/v), BSA, sodium alginate (SA), Na2SO4, Hollytex 3265 nonwoven PET support, DI water (18.2 MΩ·cm). Dope formulations: 17 wt% PSf, 83 wt% solvent; mixed at 80 °C, 200 rpm, 72 h; cooled to room temperature before casting. Dope characterization: Dynamic viscosity measured on a rheometer (0–90 s−1) showing initial ~4.85 Pa·s decreasing to ~3.39 Pa·s above 30 s−1; capillary number computed at each shear rate; surface tension measured by pendant drop (31.5 ± 1.3 mN/m). Contact angles of dope on glass 20.7° ± 1.7°, aluminum 28.8° ± 1.9°, stainless steel 31.5° ± 1.8°; stainless steel slot die selected. Washburn equation used with measured γ and known support pore data estimated penetration length of 0.15 mm, exceeding the 0.12 mm support thickness, indicating complete penetration without pre-wetting. DBE fabrication (bench scale): Four configurations (see Table 1 description). DBE-1/2: dope cast on glass at 0.2 mm gap (200 cm/min); DBE-1 used filter paper support during filtration, DBE-2 used PET support under membrane during filtration. DBE-3: dope cast directly onto taped PET support at 0.2 mm gap. DBE-4: PET support taped; a pre-wetting layer cast at 0.05 mm gap, followed by a 0.2 mm layer. After casting, ~20 s evaporation, then immersion in DI water nonsolvent bath (NIPS); membranes stored in DI ≥24 h and air-dried before characterization. SDC fabrication (R2R): Single-cavity slot die on R2R line; PET carrier on feed roller; syringe pump controls flow rate (3.5–90 mL/min), substrate speed adjustable (5.4–40.2 mm/s); coating gap set via dial indicator (nominally 0.1 mm for window determination; 0.2 mm later to reach thickness). Substrates: glass plates (20.3×25.4 cm) for unsupported; PET supports (17.8×22.9 cm) taped to glass for supported runs. Coating window assessment at 0.1 mm gap via video microscopy under transparent platen to detect air entrainment and dripping. Acoustic and visual cues defined lower/upper boundaries. For production, parameters chosen to avoid instability and wrinkling: flow rate 20 mL/min, substrate speed 17.1 mm/s; wet thickness ~0.156 mm; dry thickness ~0.09 mm. Evaporation time ~20 s at 17.1 mm/s prior to immersion (~1 min) in DI water. Support pre-conditioning in SDC: Because predicted penetration exceeded support thickness, a pre-spray of the same dope (5.52 bar/80 psi for 3 min) was used on the PET support; two variants prepared: uncompressed (SDC-3) and compressed (SDC-4; rolling pin, 20 passes, ~0.01 mm reduction in support thickness). SDC-2 (direct coat on support without pre-coat) experienced penetration and was not further characterized. Characterization: SEM (surface and cross-sections) after freeze-fracture in liquid N2; 2 nm Pt sputter coat; interface imaging to assess microcavities. Surface pore size by ImageJ from 50,000× SEM images. Total porosity by helium gas pycnometry (Accupyc 1330) with 1.25 cm radius coupons; validated by gravimetric method with Silwick oil. Adhesion assessed by ASTM D3359 cross-cut tape test (classification 0B–5B). Coating-region theory referenced to interpret minimum wet thickness versus gap in viscous-force-dominant regime. Filtration testing: Dead-end stirred cell (Amicon 50 mL), constant pressure 4.137 bar (60 psi). Pre-compaction with DI water for 10 intervals; then 100 ppm BSA for 10 intervals; permeate BSA quantified by UV–Vis at 277 nm. For select samples, 100 ppm SA tested similarly; SA quantified at 216 nm. Membrane area in figure summaries 3.14 cm²; operating conditions consistent for all samples.

• Dope rheology and wetting: Shear-thinning PSf–PolarClean–GVL dope viscosity decreased from ~4.85 Pa·s to ~3.39 Pa·s by 30 s−1; capillary number >>0.1 across shear rates indicating viscous forces dominate; surface tension 31.5 ± 1.3 mN/m; contact angles: glass 20.7° ± 1.7°, Al 28.8° ± 1.9°, stainless steel 31.5° ± 1.8°.
• Support penetration prediction: Washburn penetration length 0.15 mm versus PET support thickness 0.12 mm indicated complete penetration without pre-treatment; motivated pre-wetting with the dope as wetting agent to avoid dilution and improve adhesion.
• SDC coating window and optimal parameters: Partial coating window established at 0.1 mm gap; air entrainment and dripping defined boundaries. To achieve target thickness and avoid wrinkling, selected flow rate 20 mL/min and substrate speed 17.1 mm/s (viscous-force-dominant region); wet thickness ~0.156 mm, dry thickness ~0.09 mm. Increasing gap to 0.2 mm used to meet thickness target without shifting window materially.
• Morphology: All membranes exhibited sponge-like cross-sectional structures with thin selective skin layers, consistent with slower demixing from higher-viscosity eco-solvent mixture. Mean surface pore sizes (nm): DBE-1/2 28.4 ± 6.3; DBE-3 43.2 ± 9.1; DBE-4 40.1 ± 7.7; SDC-1 40.5 ± 10.5; SDC-3 34.6 ± 10.5; SDC-4 41.1 ± 9.45. SDC-3/4 displayed spherical microcavities near the membrane–support interface, attributed to dripping-induced trapped microbubbles during coating; selective layers remained intact.
• Porosity (gas pycnometry, %): Support 68.8 ± 3.0; DBE-1/2 67.7 ± 9.1; DBE-3 53.0 ± 2.4; DBE-4 50.4 ± 2.8; SDC-1 79.1 ± 2.7; SDC-3 57.4 ± 2.7; SDC-4 54.8 ± 1.4. Support presence reduced total porosity primarily via larger dry volume and minimized lateral shrinkage.
• Adhesion (ASTM D3359): Pre-conditioned supports improved adhesion. Classifications: DBE-3 0B, DBE-4 0B, SDC-2 1B, SDC-3 3B, SDC-4 2B. SDC samples exhibited stronger adhesion versus DBE, possibly influenced by slightly different ambient conditions during fabrication.
• Filtration performance (dead-end, 4.137 bar): All membranes reached >110 LMH/bar at end of pre-compaction. During BSA filtration, DBE final permeabilities: DBE-1 38.9 ± 5.09 LMH/bar; DBE-2 32.4 ± 2.53; DBE-3 33.1 ± 4.19; DBE-4 32.0 ± 2.92. SDC final permeabilities: SDC-1 52.1 ± 5.33; SDC-3 51.5 ± 14.3; SDC-4 75.9 ± 7.22 LMH/bar. SDC-3 achieved highest BSA rejection 99.2% ± 1.31% (with adequate mean permeability 70.5 ± 8.33 LMH/bar). SDC-1 also high BSA rejection 99.0% ± 1.70%. DBE-3/DBE-4 BSA rejections ~90.6% ± 3.86% and 92.9% ± 3.68%.
• SA filtration: SDC-3 and SDC-4 showed the highest mean SA rejections 52.0% ± 3.8% and 55.9% ± 6.1%, respectively; SA fouling caused sharp permeability declines and later stabilization; pre-wetted support configurations improved SA selectivity.
• Overall, SDC membranes matched DBE analogs in morphology, porosity, permeability, and rejection, validating SDC–R2R scale-up with eco-friendly solvents and pre-conditioning of supports.

The study addressed whether eco-friendly solvent-based PSf LNF membranes can be manufactured via SDC on an R2R system and achieve performance comparable to bench-scale DBE membranes, particularly when directly coated onto porous supports. By characterizing dope rheology and interfacial properties to define a coating window, the team identified stable SDC parameters (20 mL/min, 17.1 mm/s) that yielded uniform, defect-free selective layers. Washburn-based prediction of complete penetration into the PET support guided a pre-wetting strategy using the dope itself, which minimized thickness loss and significantly enhanced adhesion. Structural analyses showed consistent sponge-like morphologies and similar mean surface pore sizes across DBE and SDC membranes, indicating that dope composition, viscosity, and evaporation time dominated pore formation. Microcavities at the membrane–support interface in SDC-3/4 implicated dripping-induced trapped bubbles, but these localized features did not compromise selective layer integrity or solute rejection. Filtration with BSA and SA demonstrated that SDC membranes, especially pre-wetted support variants, delivered high BSA rejection (~99% for SDC-1/3) with competitive permeability and improved SA rejection (>50% for SDC-3/4), underscoring that SDC can translate bench-developed membranes to scalable manufacture without sacrificing performance. The results substantiate SDC–R2R as a practical, sustainable route for LNF membrane production using eco-friendly solvents.

This work demonstrates successful scale-up of PSf loose-NF flat-sheet membranes using eco-friendly PolarClean–GVL solvents via slot die coating on an R2R system. A dope-informed coating window enabled selection of stable parameters (20 mL/min, 17.1 mm/s) to produce uniform membranes with thin selective layers. Incorporation of a nonwoven PET support with a dope pre-wetting step prevented solution penetration, reduced shrinkage, and enhanced membrane–support adhesion. SDC membranes matched DBE analogs in morphology, porosity, and filtration performance, with SDC-3 achieving 99.2% ± 1.31% BSA rejection and adequate permeability, and SDC-3/4 exhibiting improved SA rejection (>50%). Localized interface microcavities arising from dripping did not degrade performance. Using eco-friendly solvents preserved performance while reducing environmental impact of fabrication. Future work should optimize coating conditions (e.g., pressure balance and gap) to eliminate interfacial microcavities, further tailor the support pre-treatment for adhesion and penetration control, explore broader polymer/solvent systems, assess long-term fouling/cleaning behavior, and extend to continuous R2R manufacturing of larger formats and other separation applications.

• Coating window determination was partial and conducted at a 0.1 mm gap on glass; translation to supported coating required adjustments and may not capture full parameter space.
• No vacuum was applied at the die; the pressure balance may have contributed to microcavity formation at the interface; the potential influence of coating gap on microcavities was not systematically studied.
• SDC-2 (direct coat on support without pre-coating) suffered penetration; further characterization was not pursued, limiting comparison of unsupported versus supported coatings without pre-treatment.
• Filtration tests were short-term dead-end experiments; long-term stability, fouling/cleaning cycles, and cross-flow behavior were not evaluated.
• Results are specific to PSf with 3:1 PolarClean–GVL at 17 wt% polymer; generalization to other polymers/solvent systems and different support architectures requires further study.
• Ambient condition differences between DBE and SDC (humidity, temperature) may confound adhesion comparisons.