Clean water is a critical resource for industrial, agricultural, and domestic activities. In industry, water is used as a universal solvent for washing, cooking, cooling, heating, extraction, reaction by-products, and separation processes. 1 As countries develop, water resources are diverted towards industrial applications. If not treated properly, wastewater from these industrial activities will pollute and diminish available freshwater sources. This would further aggravate water scarcity around the globe. Currently, two billion people do not have access to safe drinking water at home.
Nearly four billion people are experiencing severe water scarcity at least one month out of the year. 2,3 Improper disposal of wastewater is also a major contributor to public health issues. 4,5 Therefore, it is vital to develop robust, affordable, and efficient solutions to treat industrial wastewater around the globe. Industrial wastewater should be treated at the source, in a cost-effective manner. This would reduce the overall cost and energy consumption, avoid inappropriate discharge of wastewater, and incentivize reuse. Water reuse is also essential to tackle current water scarcity problems. 6 Many industries struggle with increased water usage as they modernize. This has led to increased interest in water reuse and zero-liquid-discharge streams in several industries. Membranes offer a highly energy efficient, and environmentally friendly method to filter water-based feeds. 7 Membranes offer better and more reliable effluent quality than conventional wastewater treatment processes (e.g., flotation, chemical precipitation). However, the complexity and diversity of wastewater streams is an obstacle to broader and more reliable operation of membrane units in these contexts, especially due to challenges with membrane fouling. This is particularly true for water reuse where the investment and operating costs are normally not amortized by the potential savings. Fouling reduces the efficiency and operability of the overall process and increases the costs associated with membrane cleaning and replacement. 8 Highly fouling resistant membranes reduce costs by up to 46% compared to standard systems in municipal wastewater filtration. 9 Water streams from certain industries are even more challenging to treat due to their high content of biomacromolecules, oil, and even microorganisms. For example, waste (or spent media) streams resulting from large-scale biomanufacturing processes (e.g., biologic drugs, food products, cellular agriculture a.k.a. "cell cultivated meat" applications) are extremely high in organic biomacromolecules and other foulants, making them difficult to treat. 10 Beyond water treatment, membranes could be crucial for enabling the recovery and reuse of valuable media components that mediate the growth of the microorganisms. However, these feeds are highly prone to fouling typical membranes through the adsorption of organic components, clogging membrane pores.
It is important to highlight that for most membrane systems fouling is highly dependent on feed composition. 11 Thus, complex feeds with numerous components will likely alter how fouling occurs. A simple and common example is the substantial increase in alginate and humic acid fouling when Ca 2+ ions are present, particularly when the membrane surface holds fixed anionic charges. 12,13 Therefore, it is important to test membranes using feeds that contain foulants that correctly correspond to the composition of the feed of a particular application. Generally, the complexity of these feeds limits the impact of most common fouling prevention methods (e.g., pre-treatment, surface grafting, coating, and blending). 14 Polymer self-assembly is a powerful and spontaneous method to create functional and easily scalable membrane materials with unique properties. [15][16][17][18] For example, the use of selfassembling polymers as membrane selective layers can create pores with low polydispersity and controlled pore chemistry, controlled by the polymer design and the fabrication process. 19,20 Thin film composite (TFC) membranes prepared by coating random zwitterionic amphiphilic copolymers (r-ZACs) onto a porous support have been shown to be highly fouling-resistant even when faced with high oil and organic contents. [21][22][23] Their effective pore size can be varied within a narrow range by custom synthesis. 24,25 However, these r-ZACs, which combine a hydrophobic monomer with a zwitterionic monomer, have limitations regarding their chemical structure. [21][22][23] There are only a couple of commercially available zwitterionic monomers, severely limiting the number of most easily accessible potential chemistries. Moreover, zwitterionic monomers typically show poor solubility in most common and safe solvents, further hindering their manufacturing and overall scalability. On the other hand, there is an abundant number of commercially available ionic or ionizable monomers that are typically inexpensive and easy to solubilize.
Recently, we reported self-assembled TFC layers composed of a random Polyampholyte Amphiphilic Copolymers (r-PACs), which combine positively and negatively charged monomer subunits with hydrophobic monomers in a random/statistical copolymer chain (Figure 1). 26 Apart from the few instances where polyampholytes were grafted onto commercial membrane materials to enhance fouling resistance, 27,28 their application in membrane systems is largely under-studied. These r-PAC layers showed highly tunable selectivity. 26 r-PACs self-assemble into a disordered bicontinuous structure with an ionic domain and a hydrophobic domain. 26 Electrostatic interactions between charged side groups along the polymer backbone likely control self-assembly as well as the effective pore size. Membranes were prepared by coating a commercial porous support with a thin layer of r-PAC, 26 a highly scalable approach utilized in roll-to-roll systems.
Changing the ratio of anionic and cationic monomers (i.e., net charge) yielded significant changes in effective pore size and membrane selectivity towards charged molecules and salts, with effective pore sizes varying from 1.6 to 2.4 nm and sodium sulfate rejections ranging from 5% to 93%. 26 Our initial study showed that r-PACs were highly resistant to fouling by oil-in-water emulsions.
Although these initial results were promising, there is still a need to characterize how r-PAC-based membranes perform when faced with other foulants (e.g., proteins, natural organic matter) and more complex feeds found in realistic applications. Additionally, it is critical to better characterize and understand how the ionic content of r-PACs influences membrane separation capabilities, fouling resistance, and overall membrane stability.
In this work, we aimed to address these substantial gaps by characterizing the membrane separation capabilities, fouling resistance, and membrane stability of r-PAC membranes with varying ionic content against multiple foulants (oil, organics, proteins, and complex biological media). We synthesized three compositions of r-PACs with net neutral charge and different ratios of hydrophobic and ionic monomers (Table 1). We prepared TFC membranes by coating a commercial porous support with a thin layer of r-PAC. The membrane permeance decreased with decreasing ionic content, but the effective pore sizes were independent of ionic content, at around 1.5 nm. All r-PAC layers behaved as a net-neutral layer, with constant rejection of the same salt at greatly different salt concentrations. All compositions showed excellent fouling resistance properties toward all representative foulants tested (oil, humic acid and alginate, and two proteins).
The intermediate composition, P02, showed no irreversible flux decline after filtering any of the foulant solutions. In addition, to simulate concentrated waste streams from biomanufacturing, fouling experiments were performed using cell culture media employed in cellular agriculture to produce cellular biomass. Impressively, P02 also showed no irreversible flux decline after filtering this complex biological mixture, while a commercial nanofiltration membrane lost 90% of its initial flux. P02 and P03 layers were stable at high ionic strength solutions (e.g., PBS 1X). Amphiphilic Copolymer (r-PAC) selective layers. (a) r-PACs synthesis scheme via free radical polymerization (FRP), showing (top left) monomers used, (top center) final random copolymer structure after synthesis, (top right) r-PAC chemical structure when immersed in water. (b) Visual representation of the r-PAC structure in solution after FRP, with hydrophobic, anionic, and cationic repeat units. (c) Visual representation of the r-PAC self-assembled structure after membrane casting and water immersion, where ionic domains serve as channels through which permeation occurs. (d) Cross-sectional scanning electron microscope image of r-PAC TFC membrane (P01) with a selective layer ~400 nm in thickness on top of a porous support.
Azobisisobutyronitrile (AIBN), [2-(methacryloyloxy)ethyl]trimethylammonium chloride (TAEMA), 4-methoxyphenol (MEHQ), bovine serum albumin (BSA, 66.5 kDa), lysozyme phosphate buffered saline (PBS) packets (0.01 M, pH 7.4), humic acid (HA), sodium alginate (SA, ultra-low viscosity 5-40 cP), magnesium chloride hexahydrate (≥99.0%), sodium sulfate dodecahydrate (≥99.0%), sucrose (≥99.0%), and copper (II) chloride were purchased from Sigma Aldrich (St. Louis, MO). 2-sulfoethyl methacrylate (SEMA), and 2,2,2 trifluoroethyl methacrylate (TFEMA) were purchased from Scientific Polymer Products Inc. (Ontario, NY). Dimethyl sulfoxide (DMSO), acetone, methanol, isopropyl alcohol (IPA), reagent alcohol, hexane, Dglucose, glycerol (lab grade), isopropyl alcohol (IPA, 99.5%), riboflavin (98%), trifluoroethanol (TFE, ≥99.0%), neutral activated alumina, and basic activated alumina were purchased from Fisher Scientific International Inc. (Pittsburgh, PA). PS35 ultrafiltration membrane was purchased from Solecta (Oceanside, Calif.). Vitamin B12 was purchased from MP Biomedicals. Compressed nitrogen (ultrahigh purity) was purchased from Airgas. Deuterated dimethyl sulfoxide (DMSO-d6) was purchased from Cambridge Isotope Laboratory (Tewksbury, MA). DC193 surfactant and soybean oil were obtained from Dow Chemicals (Providence, RI) and the supermarket, respectively. TFEMA was purified using a column packed with basic activated alumina. SEMA and TAEMA were passed through a column of neutral activated alumina.
All other chemicals and solvents were reagent grade and used as received.
Three r-PACs were synthesized using the same monomers, SEMA, TAEMA, and TFEMA, by free radical polymerization (FRP). The general approach followed procedures described in our previous research study. 26 P01, P02, and P03 each had a 1:1 mole ratio of the anionic SEMA and cationic TAEMA monomers, but varying contents of the hydrophobic TFEMA. In each case, SEMA, TAEMA, and TFEMA, in amounts shown in Table 1, were dissolved in this order in DMSO (40 mL). AIBN (0.01 g) was added to the flask. The flask was sealed with a rubber septum.
Nitrogen was bubbled through the reaction mixture for 30 minutes to purge any dissolved oxygen.
The flask was then kept at 60 °C, while stirring at 300 rpm for 17 hours. After the reaction, 1 g of MEHQ was used to terminate the synthesis. The reaction mixture was viscous at this point. For P01, the same procedure described in our previous study was utilized for the precipitation and washing of the r-PAC polymer. 26 The reaction mixture for P02 was precipitated in 1:1 acetone: hexane mixture by volume, purified by stirring three fresh portions of 1:1 acetone: hexane mixture by volume and kept in the same solvent mixture for a minimum of 5 hours, one fresh portion of acetone for a minimum of 5 hours, and three fresh portions of DI water for a minimum of 5 hours.
The mixture for P03 was precipitated in DI water, purified by stirring two fresh portions of DI water for a minimum of 5 hours, and two fresh portions of IPA for a minimum of 5 hours. All copolymers were dried in a vacuum oven at 60 °C for 48 hours. The composition of the copolymers was calculated from the 1 H-NMR spectra (Figures S1-2 following peak assignments reported in our previous publication. 26 TEM analysis was conducted using a Hitachi 7800 transmission electron microscope, operating in bright-field mode at an acceleration voltage of 100 keV. Copolymer films were fabricated by evaporating the solvent from a solution containing 5 wt% copolymer and TFE in a Teflon dish. To visualize ionic domains using TEM, the films were immersed in a 2 wt% solution of CuCl2 for a duration of 4 hours. Following staining, the films underwent drying and embedding in an epoxy resin, after which they were sectioned to a thickness of 50 nm using an ultramicrotome and transferred onto copper grids. The TEM imaging was conducted at the Harvard Center for Nanoscale Systems (CNS), and subsequent analysis of TEM images was performed utilizing ImageJ software.
Each r-PAC was dissolved in TFE to form a 5 wt% solution. The resulting polymer solutions were filtered using a 0.45 μm syringe filter (Whatman, glass fiber with PP casing) and subsequently degassed using a vacuum oven for 1 h. The thin film composite membranes were fabricated by coating a thin layer of polymer solution on an ultrafiltration (UF) membrane (PS35, Solecta, Oceanside, CA) utilizing a wire-wound metering rod (Gardco, #8). Membranes were submerged in a deionized water bath 30 s after coating. Finally, membranes were stored in deionized water for a minimum of four hours prior to testing.
TFC membranes were imaged using a field emission scanning electron microscope (FESEM) Ultra Plus (Carl Zeiss, Inc.) with 5 kV and 4.6 mm working distance. Dry membranes were submerged into liquid nitrogen for 1 min and fracture with a razor blade for cross-sectional analysis. All samples were sputter coated (Cressington 108 manual, Ted Pella Inc., CA) with Au/Pd (60/40) in argon atmosphere. A Rame-Hart contact angle goniometer (Succasunna, ŃJ) was utilized to measure the water contact angle on dried membranes cut and taped onto glass plates.
A 10 mL Amicon 8010 dead-end stirred filtration cell (Millipore) with an effective filtration area of 4.1 cm 2 was used for all filtration experiments. Membrane swatches 25 mm in diameter were placed in the cells, which were attached to a 1-gallon reservoir. The cells were continuously stirred at 500 rpm to minimize concentration polarization. To calculate membrane permeance, DI water was filtered through each membrane at a trans-membrane pressure of 40 psi (~2.8 bar) until the flux remained stable for at least 1 hour. Permeate mass was tracked using a Scout Pro SP401 balance connected to a computer, which recorded measurements at 30 s intervals using TWedge 2.4 software (TEC-IT, Austria).
Membrane flux is defined as the flow rate through the membrane normalized by membrane area. Permeance is calculated by normalizing the flux to account for operating conditions
where Lp is the membrane permeance (L m -2 h -1 bar -1 ), J is the membrane water flux (L m -2 h -1 ),
and Δp is the transmembrane pressure (bar).
Membrane size-based selectivity was determined by filtering different organic solutes (sugars, dyes, and small molecules, Table S1) at a transmembrane pressure of 40 psi. Sugar solutions were prepared at a concentration of 4,000 ppm; permeate concentration was determined utilizing chemical oxygen demand (COD) kits (K-7365, CHEMetrics). Dyes were filtered using a feed concentration of 0.1 mM; permeate concentration was quantified using a UV-vis spectrometer (Genesys10, ThermoScientific). Charge-based selectivity and salt rejections were evaluated by filtering salts at two different concentrations, 1 mM and 10 mM. Salt concentrations were determined using a conductivity meter (high range, VWR). When performing rejection tests, the first milliliter was discarded and the following 1 mL was collected and used for measuring rejection. 21 To calculate rejection the following equation was used
where R is rejection (%) of the solute, Cf is the concentration (mg/L) of the feed, and Cp is the concentration (mg/L) of the permeate.
Utilizing a feed solution comprising a 1500 ppm oil-in-water emulsion (9:1 ratio of soybean oil:DC 193 surfactant) (Dow-Corning) we tested the fouling resistance against oily water.
The oil-in-water emulsion was made by mixing water, oil, and surfactant utilizing a blender operated at high rpm for 3 min. Fouling tests were then conducted utilizing the previously mentioned dead-end filtration cell (Millipore). Initially, deionized water was filtered through the membrane until the flux stabilized at 1.2 L m -2 h -1 . Adjustments were made to the transmembrane pressure for each membrane to achieve this initial flux, ensuring uniform hydrodynamics during the experiment. The cell and reservoir were then filled with the prepared oil-in-water emulsion, which underwent filtration through the membrane for a duration of 25 hours. Prior to measuring the final flux, membranes were gently rinsed with DI water.
As a control, we performed the same fouling tests on a commercial membrane with similar membrane performance characteristics, NP030, a poly(ether sulfone) (PES) nanofiltration (NF) membrane manufactured by Microdyn Nadir, purchased from Sterlitech (Auburn, WA). For all fouling tests, the Flux Recovery Ratio (FRR) was calculated using
where Jf is the final DI water or 1X PBS flux after the fouling test and a gentle rinse, and Ji is the initial DI water or 1X PBS flux before the fouling test.
A feed solution consisting of 500 ppm humic acid, 500 ppm sodium alginate, and 10 mM CaCl2 was utilized. CaCl2 has been shown to increase the fouling propensity of anionic polymeric organics by forming complexes. 12,13 These fouling experiments were performed using the deadend filtration equipment described above using the same procedure as for the oil-in-water emulsions.
Protein fouling experiments were performed using two different proteins, lysozyme (LYZ), and bovine serum albumin (BSA). In each case, the feed solution was prepared by dissolving 1 g/L of the protein in 1X phosphate buffered saline (PBS, pH 7.4). These fouling experiments were performed using the dead-end filtration equipment described above. 1X PBS, which was the background matrix used to dissolve the proteins without aggregation, was filtered through the membrane until the flux stabilized. Transmembrane pressure was adjusted to achieve a 1X PBS flux of 1 L m -2 h -1 for each membrane. The cell and reservoir were filled with one of the prepared protein-containing fouling solutions, which was filtered through the membrane for 25 hours.
Membranes were gently rinsed with 1X PBS before measuring the final flux.
This experiment aimed to document the fouling resistance of membranes with a realistic, complex feed with high bioorganic content, similar to waste products from the most demanding bioprocessing operations. For this purpose, we used two streams associated with the suspension culture of chicken cells, a small-scale simulation of a bioprocess for cellular agriculture.
Passage 36 suspension-adapted DF-1 embryonic chicken fibroblast cells (ATCC® #CRL-12203™) were seeded onto 6-well ultra-low attachment plates (Corning™ #3471) at an initial seeding density of 250,000 viable cells/mL. The working volume of each well was 4 mL of DF-1 cell culture medium, consisting of DMEM supplemented with Glutamax (Gibco™ #10566016), 10% FBS (Gibco™#12662011), 1% Antibiotic-Antimycotic (100X) (Gibco™#15240062), and 0.01% Poloxamer 188 Non-ionic Surfactant (100X) (Gibco™#24040032). To mitigate aggregation, cells were incubated at 39°C with 5% CO2 supplementation for 96 hours before enzymatic dissociation using 0.25% Trypsin-EDTA dissociation reagent (Gibco™#25200056).
Cell suspensions from each of two wells of a 6-well ultra-low attachment plate were aliquoted into two independent conical tubes. One conical tube was centrifuged at 1,640 rpm for 5 minutes at 22°C to separate the viable cells from the spent media, while the other tube served as the "spent media with cells" sample. The spent media from the centrifuged tube was subsequently aliquoted into a separate conical tube to serve as the "spent media only" sample (no cells).
Fouling experiments were performed using the dead-end filtration equipment and spent cell culture media with and without cells as the feed. First, deionized water was filtered through the membrane until the flux stabilized at 1.2 L m -2 h -1 for all membranes. The cell and reservoir were filled with the spent cell culture media, either with or without cells. Membranes were exposed to the fouling solution for 20 hours. Membranes were gently rinsed with DI water before measuring the final flux using deionized water.
We synthesized r-PACs using a simple and scalable synthesis method, FRP (Figure 1). To avoid the dissolution or excessive swelling of the selective layers in water, 2,2,2-trifluoroethyl methacrylate (TFEMA) was chosen as the hydrophobic monomer unit for all r-PACs. The high hydrophobicity of TFEMA likely also aids in the formation of a phase separated bicontinous nanostructure. 26 Additionally, previous research studies utilizing amphiphilic copolymers [20][21][22][23] showed that TFEMA-containing copolymers are easy to solubilize in different solvents for the synthesis and membrane casting. While there is growing concern regarding perfluoroalkyl substances (PFAS), TFEMA only contains one CF3 group and is not included in most definitions of PFAS. 29 Nonetheless, we expect that many other hydrophobic monomers (e.g., alkyl methacrylates, acrylonitrile, styrenics) can be used in the synthesis of r-PACs. In the future, we expect to further explore such non-fluorinated hydrophobic monomer units to replace TFEMA.
We selected the same polyampholyte ionic pair from our previous original study. 26 This pair of charged moieties mimics those in the zwitterionic monomer sulfobetaine methacrylate (SBMA), which comprises a quaternary amine group covalently linked to a sulfonate group. 22,23 .
SBMA is commercially available, and its fouling resistance and interactions are well-researched.
To synthesize a polyampholyte with similar properties, we selected a methacrylate monomer that contains a quaternary amine group, [2-(methacryloyloxy)ethyl]trimethylammonium chloride (TAEMA). The quarternary amine remains cationic over a wide pH range. We selected 2sulfoethyl methacrylate (SEMA) as the anionic monomer due to its sulfonate group, which exhibits an anionic charge in water over a broad pH range.
We synthesized r-PACs with different mass fractions of the hydrophobic repeat unit while keeping the molar ratios of the ionic monomers constant at 1:1 (Table 1). Three distinct r-PAC compositions were synthesized to study how the membrane performance and self-assembled nanostructure might be influenced by these compositional changes. The first polymer, P01, had close to a 1:1 mass ratio of ionic to hydrophobic repeat units (TFEMA). While exhibiting good oil-in-water fouling resistance, this composition had poor stability in saline media (e.g., PBS) due to the anti-polyelectrolyte effect. When immersed in PBS, the membrane permeance would steadily increase over time. This behavior has been reported to be more significant in random copolymers with a lower hydrophobic mass fraction. 25 While higher hydrophobic monomer contents would yield better stability, it could also make the surface more hydrophobic and possibly more prone to fouling. Therefore, we studied two other compositions with higher hydrophobic monomer content. P02 contained close to a 1:2 mass ratio of ionic to hydrophobic groups, whereas the mass ratio of ionic to hydrophobic groups in P03 was roughly 1:3. We expected all membranes to closely mimic zwitterionic copolymers with net-neutral charge characteristics, exhibiting sizebased selectivity. To avoid highly unstable layers in water, we did not synthesize r-PACs with a larger ionic monomer content than hydrophobic monomer content. For all copolymers, 1 H-NMR analysis was used to determine the chemical composition (Tables 12, Figure S1-2). Copolymer yields were between 40% -70%. These copolymers are best described as statistical copolymers, in which the monomer order is determined by the reactivity ratios of each monomer pair. r-PACs described here demonstrated a close match between the reaction solution compositions and the compositions of the final copolymers (Table 2). This suggests that the copolymers are probably close to random, with monomer reactivity ratios near 1, particularly considering that the copolymer composition is measured far from full conversion.
Hence, we utilize the term "random copolymer" in this work. However, additional in-depth characterization of the copolymerization could provide further insight.
To understand the behavior of our copolymers during membrane operation in an aqueous environment, we performed water uptake measurements. We expect the water to largely partition into and swell the ionic domains, with P01 swelling significantly more than P03. Therefore, the degree of swelling will highly influence membrane performance. As predicted, water uptake measurements (Table 2) show that copolymers with higher ionic content swell significantly more, with P01, P02, and P03 having a 39%, 15%, 8% water uptake, respectively. Interestingly, contact angle measurements showed similar wettability characteristics and overall hydrophilic surface behavior for all r-PAC membrane surfaces. The low contact angle measurements, at around 65° for all r-PACs, were promising indications of hydrophilic and fouling resistant membrane surfaces.
TEM analysis was utilized to characterize the self-assembled morphology of r-PACs. The immersion in 2 % aqueous CuCl2 solution was used to positively stain the ionic domains. The Cu 2+
ions selectively partition into the water-permeable ionic domains. Staining is enhanced by the formation of sulfonate-copper complexes. 30 Figure S3 shows brightfield TEM images of r-PAC copolymers. All copolymers display a disordered bicontinuous structure, with a network of ionic nanochannels (dark regions due to staining) surrounded by the hydrophobic phase (light regions).
These disordered bicontinous structures are consistent with morphologies of previously reported random copolymers with highly incompatible repeat units. 22,24 The characteristic length scale given by the FFT was between 3.1 and 4.4 nm for all copolymers, similar to previously studied copolymer with pore sizes between 1 and 2 nm. 22,24,26 These results imply that all r-PACs in this study have similar ionic domain sizes in dry state. In water, the ionic domains are expected to swell to some extent, as indicated by the water uptake measurements. These swollen ionic "channels" are partly filled by the ionic polymer backbone and side groups. Hence, the available space (pore size) for solute permeation is likely smaller than the swollen domain size.
All r-PAC layers were manufactured by coating a 5 wt% polymer solution utilizing a wirewound metering rod (Gardco, #8) on top of a commercial ultrafiltration membrane (PS35, Solecta).
2,2,2-trifluoroethanol (TFE) was the casting solvent. The membranes were immersed in DI water 30 s after casting. All TFC r-PAC-based membranes showed a selective layer on top of the support membrane, with thicknesses around 400 nm (Figure 1d, Figure S4). This thickness can potentially be further decreased by optimizing and modifying the coating process, both in lab scale systems and during scale-up. For example, previous work has demonstrated that using an ionic liquid additive in the coating solution 21 or electrospray deposition 31 can create TFC membranes with extremely thin r-ZAC selective layers along with very high permeances.
The r-PAC TFC membrane surfaces were quite smooth, as shown be SEM images of samples placed on a 45 o holder so both the cross-section and the surface are visible (Figure S5). These TFC membranes showed no signs of delamination in SEM images (Figure 1) or during filtration experiments, indicating good integration of the selective layer onto the UF support membrane.
This layer is held in place by intermolecular interactions between the copolymer layer and the support. Furthermore, we expect that a small amount of copolymer penetrates under the topmost pores of the support membrane before solidifying as the solvent evaporates. This creates an additional, mechanical means of adhesion between the layer and the support. This membrane manufacturing approach could be easily adapted to roll-to-roll production, utilizing readily available machinery already used in these common processes, eliminating the need for additional or costly steps.
To characterize the performance of r-PAC membranes, dead-end stirred cell filtration experiments were completed utilizing three separate membrane samples for each copolymer composition. The water permeances of P01, P02, and P03 membranes were 2.1 ± 0.4 L m -2 h -1 bar - 1 , 0.9 ± 0.1 L m -2 h -1 bar -1 , and 0.4 ± 0.1 L m -2 h -1 bar -1 , respectively. Membrane permeance decreased with hydrophobic monomer content. This is consistent with decreasing water uptake values, which would be proportional to the volume fraction and overall connectivity of permeable ionic "nanochannels" in the membrane selective layer. It is also consistent with trends observed for zwitterionic amphiphilic copolymer membranes. 23 Commercial NF membranes with similar pore size usually have permeances in the 1-10 L m -2 h -1 bar -1 range. 32 The permeances of r-PAC membranes described in this study are in the lower end when compared to commercially available membranes for this pore size range, though the control membrane, NP030 by Microdyn Nadir, has the same permeance as P02, at ~1 L m -2 h -1 bar -1 . We expect that further development of casting methodologies will lead to improved permeances. For example, a past publication has shown that the use of an ionic liquid co-solvent can increase the permeance of zwitterionic amphiphilic copolymer membranes by up to 10 times. 21 The fouling resistance of these membranes, described below, will also lead to higher permeance during operation under realistic conditions.
To estimate the effective pore sizes of r-PAC membranes, different neutral organic solutes with known Stokes radii (vitamin B12, sucrose, glucose, and glycerol; see Table S1) were filtered by the membranes (Figure 2a). The average rejection for every membrane composition was fitted to a Donnan Steric Pore Model with single pore diameter to determine the effective pore size. 24 All r-PAC membranes had similar neutral solute rejections, with effective pore diameter in the 1.5-1.6 nm range, with P02 and P03 with the smaller effective pore diameters at 1.5 nm, and P01 with the largest effective pore diameter at 1.6 nm. These values are consistent with the trends in effective domain sizes documented by TEM (Figure S3). TEM domain sizes and effective pore sizes from filtration experiments consistently supports our hypothesis that transport occurs through the ionic domains, whose size is determined by microphase separation of the copolymer and by the anioncation interactions. We also observe minimal influence of the vol% of the ionic domains in the effective pore sizes of the different r-PAC membranes. These observations are consistent with previously reported random copolymers with highly incompatible repeat units, including r-ZACs, 22,24 and r-PACs. 26 Exploratory experiments showed membranes prepared with shorter (5 s) or longer (overnight) evaporation times before water immersion also show identical effective pore sizes, indicating membrane selectivity is not measurably affected by this manufacturing step. for uniform effective pore diameters (Table 2). (b) Sodium sulphate rejection at different salt molar concentrations for P01, P02, and P03 TFC membranes. (c) Magnesium chloride rejection at different salt molar concentrations for P01, P02, and P03 TFC membranes.
In addition to quantifying effective pore size, we tested the rejections of two divalent salts, magnesium chloride and sodium sulfate, by r-PAC membranes (Figure 2b and 2c). These salts were selected as nanofiltration (NF) membranes, whose effective pore size is similar to the r-PAC membranes, are often used in water softening applications and divalent ion removal. All three membranes exhibited relatively high Na2SO4 rejections, between 75-85%. P01 exhibited the lowest rejection, in agreement with its marginally higher effective pore size as well as its potential stability issues; P02 and P03 showed comparable rejections ~85%. Interestingly, the MgCl2 rejection varied significantly among the three membranes despite similar pore sizes and Na2SO4 rejections, from ~30% for P01 to ~85% for P02. P02 had the highest rejection of both salts. It is possible that these differences arise from a combination of slightly different pore sizes combined with specific solute transport mechanisms. Previous studies on zwitterion-containing membranes have shown ion-zwitterion interactions may affect ion selectivity. 24,34 Further in-depth studies on the different solutes transport mechanisms may reveal further insight.
These experiments were also useful for probing the rejection mechanisms observed in these novel membrane materials. Donnan exclusion in charged polymers or pores predicts that when the ionic strength of a solution increases, charge screening effects increase as well. 35 Consequently, for a polymer selective layer of non-zero net charge, the rejection of salts is lower at higher salt concentrations. In contrast, rejection should not change with ionic strength when using a selective layer that is net neutral (i.e., charge balanced). For all three membranes, we see no change in rejection when using different ionic strength solutions. These results indicate that all three membranes behave as layers with net neutral electrostatic charge.
Net neutral membranes with pore sizes in the 1-2 nm range are valuable in a variety of biomedical and chemical applications where proteins, macromolecules, and bigger solutes need to be retained while salts, metabolic by-products, and other small molecules (i.e., glucose, lactic acid, urea, charged dyes) are allowed to permeate through the membrane. Due to these characteristics, our membranes are great candidates to be used in not only water treatment and reuse, but also potentially the recovery and reuse of valuable components during biofabrication, like nutrients, hormones, and growth factors. 36 For example, in applications including cellular agriculture for "cell cultivated meat", these membranes can potentially enable the recovery and reuse of expensive components like growth factors, which account for a significant portion of final product costs, while allowing salts and metabolites (e.g. lactate) to pass through, preventing their accumulation.
Alternatively, they can be used to recover and concentrate nutritious proteins and polysaccharides in food waste digestate for use in animal feed. However, all these applications necessitates membranes that are extremely resistant to fouling by these complex feeds.
Membranes are often exposed to harsh chemical environments due to fluctuations in feed quality, for sterilization of the filtration system, and for chemical cleaning operations. Therefore, novel membrane materials should remain stable when exposed to acidic and basic media, and to chlorine. In fact, one of the major shortcomings of current reverse osmosis and NF membrane materials is their limited chlorine resistance. 37,38 Therefore, we evaluated the chemical stability of P02 membranes upon exposure to acidic (pH 3) and basic (pH 12) solutions for 24 hours, as well as a 24-hour chlorine tolerance test using a 2000 ppm sodium hypochlorite solution (pH 8.5).
Membrane swatches were tested before and after exposure for their permeance and vitamin B12 rejection. We also performed FTIR analysis to determine any chemical changes. Permeance and rejection of these membranes remained stable after exposure (Table S5). No changes in FTIR spectra were detected, demonstrating their stability (Figure S7). These findings indicate the overall chemical stability of these membranes, in line with past findings on similar zwitterionic amphiphilic copolymer membranes. 11, 39
A key objective of this work was to characterize the fouling behavior of r-PAC membranes with different foulant classes. Performing different membrane fouling tests, using oil, organics, and proteins, is crucial to gain a comprehensive understanding of r-PAC's overall fouling resistance. This approach provides valuable insights into the membrane's performance under various conditions, enhancing its potential use in different applications. For this purpose, we prepared and tested four representative foulant solutions with the three r-PAC membranes with varying hydrophobic fractions. A commercial poly(ether sulfone) (PES) membrane with similar permeance (0.9 ± 0.2 L m -2 h -1 bar -1 ) and solute rejection capabilities, NP030 manufactured by Microdyn Nadir, was selected as a control membrane for all fouling tests in this study. The four solutions comprised an oil-in-water emulsion mimicking oily wastewater, 40,41 a solution of common organic foulants found in surface water and wastewater effluent streams from secondary treatment, [11][12][13] and two proteins with varying surface charge (anionic BSA, cationic lysozyme). All fouling experiments were performed in dead-end stirred cell mode. While this is not closely representative of realistic operation, it is a worst-case scenario for fouling as the foulant concentration in the feed is progressively concentrated in the cell during the experiment. Therefore, membranes that exhibit high resistance to fouling, especially irreversible fouling, in such experiments are likely to perform even better in more realistic cross-flow systems.
To simulate oily wastewater, we utilized a 1500 ppm oil-in-water emulsion prepared using 9:1 ratio of soybean oil to DC193 surfactant. 41 After stabilizing the membrane by filtering DI water, the feed was replaced with the oil emulsion. The transmembrane pressure was adjusted to achieve identical initial water fluxes for all membranes to limit the effect of hydrodynamics and different permeances during the experiments. After 25 hours of filtering this solution, the membranes were rinsed with deionized water to simulate a physical cleaning, and pure water flux was measured once again. All r-PAC membranes, P01, P02, and P03 exhibited extraordinary fouling resistance with no flux decline during the fouling test or after water rinse. The commercial PES membrane (NP030) irreversibly lost about half of its original flux (Figure 3), in agreement with previous studies with several commercial membranes. 11,42,43 It is remarkable that even at very high hydrophobic content (P03) the membrane still shows no flux reduction during and after exposure to the fouling solution. This shows that even at very low ionic polymer content is possible to create a fouling resistant layer using r-PACs. These results show that self-assembled polyampholytes are highly promising materials to filter oily water. One possible hypothesis is that the affinity of water molecules to the ionic nanochannels is what drives the transport, and this decreases the influence of hydrophobic regions and concentration polarization on flux even as the feed concentration increases. However, this hypothesis needs to be tested by other experiments. nanofiltration membrane of similar pore size and permeance. The dashed lines are for visual orientation only. The horizontal line tracks the original water flux, while the data points between the vertical lines represent the time in which the membranes were exposed to the fouling solution.
Dissolved/natural organic matter is likely the most significant contributor to fouling among foulants present in natural water streams (e.g., surface water, seawater) as well as wastewater. 44 Fouling by these organic biomacromolecules affects all membrane categories spanning pore sizes from Angstroms to μms, and applications from seawater desalination to municipal wastewater treatment. [44][45][46] Humic acid and sodium alginate are two major biomacromolecules responsible for these fouling phenomena. The addition of CaCl2 has been shown to increase membrane fouling by the formation of complexes between calcium ions and anionic polymeric organics. 12,13 Based on these criteria and typical concentrations of these foulants in relevant feeds (e.g., surface water, littoral seawater to be pretreated for desalination, wastewater secondary treatment effluent) we created an organic mixture containing 500 ppm humic acid, 500 ppm sodium alginate, and 10 mM CaCl2. After 25 hours of filtering this organic mixture, the membranes were rinsed with deionized water to simulate a physical cleaning, and flux was measured once again. We observed a 10-15% decline of the water flux for all r-PAC membranes when filtering the organic mixture (Figure 4), likely associated with cake fouling by the small particulates formed by the aggregation of humic acid with calcium ions. 47 All r-PAC membranes, P01, P02, and P03 exhibited extraordinary fouling resistance with minimal flux decline after deionized water rinse, indicating that there was no irreversible/adsorptive fouling by this foulant mixture. In contrast, the commercial PES membrane (NP030) lost about half of its original flux upon exposure to the organic mixture (Figure 4). Unlike the r-PAC membranes, physical cleaning did not lead to a significant flux recovery. This implies that the foulant layer is strongly bound to the membrane surface. This is consistent with previous research, which has shown that Ca 2+ ions can bridge humic acid and alginate chains and attach them to a membrane surface with carboxylate and sulfonate groups 48 common in commercial NF membranes to enhance salt rejection. nanofiltration membrane of similar pore size and permeance (NP 030). The dashed lines are for visual orientation only, the horizontal line tracks the original water flux, while the data points between the vertical lines represent the time in which the membranes were exposed to the fouling solution.
Proteins are common foulants in wastewater treatment processes, especially food and beverages, agricultural, or municipal wastewater streams. Additionally, with the growth in the biotechnology and pharmaceutical industry, the efficient and cost-effective separation of proteins and other biological ingredients is of key importance to advance medical outcomes. 49 Proteins are a challenging foulant due to their inherently amphiphilic characteristics, and pH dependent net charge. Therefore, the complexity of available protein structures and properties requires the development of membranes and processes that can withstand continuous exposure to these biological components, but also be easily tailored to the specific requirements of a given final application. [50][51][52] Hence, it is vital to develop robust membrane technologies that can deliver manufacturing processes that are efficient and cost-effective to produce biomolecules used as clinical therapeutics. 53 To evaluate fouling resistance towards protein containing solutions, we tested our membranes using two distinct proteins. Many commercial membranes rely on a net anionic surface to improve hydrophilicity and slow down fouling. While they perform well with solutions containing a single protein with a net anionic surface charge (e.g., BSA, a common protein foulant for membrane characterization), they can foul when exposed to cationic proteins (e.g., lysozyme)
or complex media containing a variety of proteins. Therefore, we tested the resistance of r-PAC membranes to fouling by both anionic and cationic proteins. Protein fouling experiments were conducted with 1X PBS as the matrix as many proteins tend to denature and/or aggregate in deionized water. Therefore, the first protein solution contained 1 g/L of lysozyme (LYZ) in 1X PBS (pH 7.4). At this pH, LYZ has a net positive charge. The second solution was a 1 g/L of BSA in 1X PBS (pH 7.4), where BSA has a net negative charge. The membranes were equilibrated by filtering 1X PBS. Then, the feed was switched to the protein solution. After 25 hours of filtering these protein solutions, the membranes were rinsed with 1X PBS to simulate a physical cleaning, and 1X PBS water flux was measured once again.
It is widely known that the swelling behavior of polyampholyte and polyelectrolyte layers is strongly affected by the ionic strength of the solution. [54][55][56] Charged hydrogels with a high charge imbalance show a polyelectrolyte effect, in which, the polymeric layers decrease swelling with increasing ionic strength solutions. [54][55][56] Inversely, net neutral polyampholyte layers exhibit an increase in swelling with increasing ionic strength solutions, also known as the anti-polyelectrolyte effect. [54][55][56] P01, which had the highest ionic content, exhibited poor stability when exposed to 1X PBS. While the membrane permeance and selectivity was stable when filtering DI water and solutions with relatively low salinity, the flux exhibited a gradual increase when filtering 1X PBS (Figure S6). This behavior is likely due to the anti-polyelectrolyte effect. [54][55][56] P01 was the only membrane that exhibited this increased permeance when filtering a high ionic strength solution, this is likely due to the lower hydrophobic content of P01 when compared with P02 and P03. 25 Due to this lack of stability of P01 in high ionic strength solutions, only P02 and P03 membranes were tested when utilizing 1X PBS.
P02 exhibited extraordinary fouling resistance, with no flux decline upon exposure to the protein solution, during the fouling test, or after 1X PBS rinse for both LYZ and BSA solutions (Figure 5, Figure 6). Impressively, these results show that P02 can perform protein filtrations, independent of protein charge, and retain 100% of the initial flux while filtering the fouling solution. This performance showcases how self-assembled layers that are carefully chemically designed have a superior performance when compared with commercial membranes with relatively fixed chemistries that often require surface modifications to improve overall fouling resistance.
P03 had a ~20% decline in flux upon exposure to each protein and during the fouling experiments for both LYZ and BSA, but the flux was fully recovered after 1X PBS rinse (Figure 5, Figure 6). Therefore, any adsorptive fouling on P03 was easily cleaned without the need for any harsh chemicals. It is important to highlight that the P03 flux recovery after the 1X PBS rinse is not immediate, similar to the organic mixture fouling test (Figure 4). This suggests that foulants weakly adsorb to the surface of the membrane, when present at high concentrations in solution, but desorb when present at low concentrations (after rinse). The commercial PES membrane (NP030) irreversibly lost a significant amount of its original flux in both protein fouling tests (Figure 5, Figure 6). using a 1 g/L of lysozyme in 1X PBS. (b) P02 fouling test using a 1 g/L of lysozyme in 1X PBS. (c) P03 fouling test using a 1 g/L of lysozyme in 1X PBS. (d) NP030 fouling test using a 1 g/L of BSA in 1X PBS. (e) P02 fouling test using a 1 g/L of BSA in 1X PBS. (f) P03 fouling test using a 1 g/L of BSA in 1X PBS. For all figures (a-f), the dashed lines are for visual orientation only, the horizontal lines track the original water flux, while the data points between the vertical lines represent the time in which the membranes were exposed to the fouling solution. of similar pore size and permeance. Our r-PAC membrane performs much better in all fouling test conditions, with P02 showing no irreversible flux loss in all conditions. membrane (NP030) in all fouling tests performed by this point in the manuscript. All r-PAC membranes exhibited full flux recovery with a simple physical cleaning procedure for every foulant they contacted, as long as the selective layer was stable in the matrix. This was true even for P03, which had a comparatively high mass fraction of hydrophobic groups. Beyond FRR values, P02 retained its high permeance even during the filtration of increasingly concentrated foulant solutions throughout 25-hour dead-end experiments.
While r-PAC membranes exhibited exceptional fouling resistance with multiple foulants, simplified feeds such as those described in Section 3.7 may not necessarily predict fouling resistance where membranes are faced with complex mixtures often encountered in challenging applications, from wastewater treatment to nutrient recovery to food and beverage applications.
Therefore, P02 was selected to test the ability of r-PAC membranes to resist fouling with a realistic, complex, high-organic-content feed. Among various possible applications, we specifically selected the filtration spent culture media from cellular agriculture for this demonstration. "Cell cultivated meat", a component of cellular agriculture, involves the growth of animal cells in nutrient-rich cell culture media to produce food. While it adapts biomanufacturing techniques used commonly for the synthesis of biologic drugs for food production, it must operate at a much larger scale and with tighter economics. Media fed to bioreactors for cellular agriculture contains a mixture of proteins (growth factors, hormones, other proteins), sugars, amino acids, other small molecule compounds, and salts. 57 The most expensive components in this mixture are purified biomacromolecules such as growth factors and proteins. Membrane filtration of spent media from these bioreactors can potentially enable the retention and recycling these expensive components, which mediate cell growth but are not consumed, while allowing undesired small molecule metabolites to be removed. However, commercial membranes are quickly fouled by such complex feeds, preventing reliable long term operation. Furthermore, the small effective pore size of r-PAC membranes promises to retain even smaller valuable biomacromolecules in spent media (e.g. insulin, ~6 kDa), while the net neutral charge of the r-PACs described in this study allow the passage of small ions such as salts and undesired metabolites (e.g. acetate). This enables the completion of this desired separation process in a single step that would not be otherwise achievable by commercial membranes.
Motivated by this potential, we studied the fouling of r-PAC and commercial membranes with spent cell culture media designed to grow chicken cells as a worst-case scenario in terms of membrane fouling potential. The unspent cell culture media contained buffer, nutrients, antibiotics and antimycotics, surfactants, and 10% fetal bovine serum (FBS), a solution rich in proteins and other molecules. Cells were seeded in this medium and left to grow for 96 hours. After the culture period, spent media with cells was filtered "as is". In addition, a "cell-free" spent media was prepared by isolating the supernatant from the cell suspensions using centrifugation. After 20 hours of filtering these complex media, feed and permeate were collected for COD and conductivity analysis, membranes were rinsed with deionized water to simulate a physical cleaning, and deionized water flux was measured once again.
P02 exhibited extraordinary fouling resistance with no flux decline after deionized water rinse for both mixtures (Figure 7). It is important to note that P02 had an immediate decline in flux during the fouling experiments in both tests. This sharp drop in the flux was due to osmotic pressure differences between the feed and the retentate due to the retention of ions and other solutes (Table 3). Flux was fully recovered when the feed was replaced with deionized water. In contrast, the commercial PES NF membrane (NP030) irreversibly lost 90% of its original flux in both tests (Figure 7). Physical cleaning did not lead to any flux recovery for the commercial membrane.
Interestingly, the filtration flux profiles were essentially identical for both "as-is" (with cells) and "cell-free" (without cells) spent media scenarios. The presence of cells seemed to not affect the performance of the membranes, indicating that adsorptive fouling with biomolecules was the dominant fouling mechanism, at least for these time scales. and commercial nanofiltration membrane (NP030) of similar pore size and permeance. (a) Fouling test utilizing spent media without cells. (b) Fouling test utilizing spent media with cells. For all figures (a-b), the dashed lines are for visual orientation only, the horizontal lines track the original water flux, while the data points between the vertical lines represent the time in which the membranes were exposed to the fouling solution.
While the r-PAC membrane, P02, was unaffected by the 20-hour filtration of these complex biological mixtures and had a FRR of 100%, the commercial nanofiltration membrane had a final FRR of 10% for both tests, with and without cells (Table 3). To quantify the rejection properties for each membrane, we measured the chemical oxygen demand (COD) and conductivity of each commercial membranes lost from 30% to 90% of their flux instantly and irreversibly. The composition of r-PACs was improved to create a membrane that is stable at high ionic strength solutions and has water permeance comparable to commercial membranes. The improved membrane, P02, showed no irreversible flux decline in all fouling tests. This membrane was chosen as the ideal r-PAC membrane for further evaluation with a cellular agriculture medium fouling test. This is, to our knowledge the first-time cellular agriculture cell culture medium was filtered with a self-assembled TFC layer. P02 showed no irreversible fouling after 20 hours of filtering cellular agriculture cell culture medium while a commercial membrane lost 90% of its initial flux. These results are extremely promising and demonstrate the great potential of this class of self-assembled membranes to perform highly challenging biological and chemical separations.
license from Tufts University to commercialize the technology described in this manuscript. Luca Mazzaferro, Kirsten Trinidad, and David L. Kaplan declare no conflict of interest.