Tuning the structure and performance of polyelectrolyte complexation based aqueous phase separation membranes

Aqueous Phase Separation (APS) provides a new and sustainable platform to fabricate polymeric membranes entirely in water. Still, little is known on how the casting solution and coagulation bath compositions can be used to tune membrane structure and performance. This work comprises a detailed investigation on the tuning pa- rameters avaliable to tailor the morphology, pore size distribution, and water permeability of polyelectrolyte complex membranes prepared from poly(sodium 4-styrenesulfonate) (PSS) and polyallylamine hydrochloride (PAH). To avoid complexation of PAH and PSS in the casting solution, an optimum amount of base (NaOH) must be added to deprotonate PAH. In addition, the monomer mixing ratio of PSS to PAH significantly influences membrane morphology by modulating the interactions between the two polyelectrolytes. Coagulation bath pH can be used to control the driving force for complexation. Decreasing bath pH facilitates the formation of denser membranes, allowing ~97% protein retentions, whereas increasing bath pH leads to more open membrane structures. Changing the concentration of crosslinker in the coagulation bath allows tuning of membrane pore size from ~2 nm to ~46 nm, while simultaneously influencing membrane mechanical properties. Overall, this work highlights several key parameters to tune APS membrane morphology, demonstrating the versatility of APS to prepare optimized sustainable membranes for specific


Introduction
Polymeric membranes are predominantly produced by the nonsolvent induced phase separation (NIPS) approach. The first NIPS membranes were prepared by Loeb and Sourirajan for seawater desalination when they showed that a liquid polymer solution could be precipitated, in a controlled manner, to form solid asymmetric membranes with dense skin layers and porous support structures [1]. From that moment on, the NIPS process gained widespread attention from researchers and industry alike with polymeric membranes widely applied, for example in water treatment, kidney dialysis and gas separation applications [2].
Over the past 60 years, the NIPS approach has been very successful, as it is a simple and scalable approach. First, a polymer is dissolved in an organic solvent, typically N-methyl pyrrolidone (NMP), to obtain a homogeneous solution. For flat sheet membranes, the polymer solution is cast on a substrate and subsequently immersed in a non-solvent bath, which is typically water. The polymer is insoluble in water and hence, after contact it phase separates into a solid porous film, with the speed of phase separation determined by the rate of solvent-non-solvent exchange. Strathman et al. found that membrane structure, including the pore size, greatly depends on both the thermodynamic and kinetic aspects of the phase inversion [3,4]. Their results indicated that faster precipitation rates generally result in membranes with 'finger-like' macro-voids in the structure. On the other hand, slower precipitation rates generally result in membranes with 'sponge-like' morphologies. This degree of control over membrane structure and pore size enabled the NIPS approach to be widely used for the fabrication of microfiltration and ultrafiltration membranes in addition to excellent supports for nanofiltration, reverse osmosis, and dense gas separation membranes [5].
For NIPS there are several key parameters that can be used to tune membrane morphology, especially ones related to the composition of the casting solution and the coagulation bath. For casting solutions, polymer concentration is one such parameter that has been widely studied. Higher polymer concentrations in the casting solution result in membranes with denser structures and narrow pore size distributions [6][7][8]. This is because the solvent and non-solvent exchange is slowed down due to the higher viscosity of the polymer solution at higher concentrations, leading to membranes with thicker and denser top layers. Casting solutions can also be modified by the addition of additives such as polyethyleneglycol (PEG) and polyvinylpyrrolidone (PVP), which function as pore formers, hydrophilic agents, viscosity increasers or phase separation accelerators [9,10]. Generally, the additives leach out of the casting solution into the non-solvent bath during the phase inversion process affecting the rate of polymer precipitation [11]. The kinetics of membrane formation during NIPS can also be controlled by changing the coagulation bath conditions. A high miscibility of solvent and non-solvent generally results in porous membranes with macro-voids in the sub-structure, while a lower miscibility results in asymmetric membranes with dense top layers. Addition of solvent to the coagulation bath is known to slow down the precipitation rate by lowering the non-solvent activity, and generally this reduces macro-void formation [12]. For in-depth descriptions of the NIPS process and the full suite of membrane tuning parameters in NIPS, we guide readers to several review articles [12][13][14][15].
While NIPS is currently the foremost technique for producing commercial membranes, the most used solvent in this process, NMP, is reprotoxic for humans and therefore not a sustainable choice for membrane fabrication [16]. Indeed, NMP has recently been restricted within the European Union through REACH legislation [17]; leading to increased interest in developing more sustainable approaches to polymeric membrane fabrication. Polyelectrolytes, are one class of polymer that have long exhibited interesting properties as membrane materials [18]. Recent work from our research group has proposed several more sustainable approaches to membrane fabrication, where water is used both as the solvent and non-solvent to prepare polyelectrolytes-based membranes [19][20][21]. In one very promising version of this Aqueous Phase Separation (APS) approach, the phase inversion is achieved by polyelectrolyte complexation of a strong polyelectrolyte, e.g. poly(sodium 4-styrenesulfonate) PSS, and a weak polyelectrolyte, e.g. polyallylamine hydrochloride PAH. For this system, a homogeneous solution of PSS and PAH is prepared at a high pH, where the weak polybase PAH is uncharged and can be mixed with PSS without forming a polyelectrolyte complex. The PSS-PAH solution is cast on a glass plate and immersed in a low pH coagulation bath. The pH gradient from a high pH in the solution to a low pH in the bath allows the PAH to acquire charge and to form a water-insoluble polyelectrolyte complex with PSS. Similar to NIPS, the casting solution in APS is transformed from a liquid phase to a solid phase via phase inversion. We have shown that polyelectrolyte concentration and molecular weight both play an important role in determining membrane morphology and their separation performance [19], with excellent examples of microfiltration, ultrafiltration, and nanofiltration membranes produced.
Beyond our initial work on complexation-induced APS [19], there are many other important parameters that can be used to control membrane formation. In this work, we systematically investigate the effects of various parameters that give control over the final PSS-PAH membrane structure and separation performance. The concentration of sodium hydroxide in the individual PAH solution and the effect of monomer mixing ratio of PSS to PAH was investigated in detail to identify the optimum casting solution composition(s). Moreover, the coagulation bath conditions such as the pH and the amount of crosslinker were varied to better understand their effects on membrane pore sizes and morphology. The findings of this work contribute towards a better understanding of how to optimize the performance of APS membranes towards specific applications.

Preparation and casting of polyelectrolyte solutions
The stock solution of PSS (25 wt% in water) was diluted to 12 wt% using deionized water. Similarly, the stock solution of PAH (40 wt% in water) was diluted to 12 wt% using deionized water and a small volume of a 10 M NaOH solution. The effect of NaOH concentration was studied by varying the amount of NaOH in PAH solution such that the ratio R (C NaOH /C PAH ) varied from 0.1 to 1.25 as shown in Table 1, where C NaOH (wt%) and C PAH (wt%) are the concentrations of NaOH and PAH in the initial PAH solution, respectively. Here, the order of addition of NaOH is vital; i.e. adding NaOH to PSS instead of PAH and then mixing the two solutions does not give homogeneous casting solutions. After preparation of the PSS solution and the pH-adjusted PAH solution, they were mixed to achieve final casting solutions (see Table 1) with monomer ratios of PSS:PAH of 2:1, 1:1, 1:2, 1:3 and 1:4. The homogeneity of the final casting solutions was studied via turbidity measurements using a Turb® 430 IR -WTW portable turbidity meter.
The homogeneous polyelectrolyte casting solutions were cast as thin films on glass plates using a casting bar with a gap height of 0.6 mm. The cast films were immediately immersed in a coagulation bath composed of deionized water at varying pH values (0.25, 0.5, 0.75, 1, 1.25, and 1.5, adjusted using HCl) and different GA concentrations (0 wt%, 0.05 wt%, 0.1 wt%, 0.2 wt%, and 0.5 wt%). After coagulation, the precipitated films (membranes) were removed from the bath and stored in deionized water until further use.

Viscosity measurements
The dynamic viscosity of the polyelectrolyte casting solutions was measured on HAAKE Viscotester 550 Rotational Viscometer (Thermo-Fisher Scientific, USA). Approximately 15 mL of the polyelectrolyte solution was poured into the standard spindle cylinder (SV-DIN). The cylinder was mounted on the viscometer and the dynamic viscosity of the solutions was measured as a function of shear rate (2.5-250 s − 1 ). Two samples of each polyelectrolyte solution were measured and the average value extrapolated to zero shear conditions is reported here along with the standard deviation. All the measurements were performed at 20 • C.

Characterization of the membranes
For scanning electron microscopy (SEM) imaging, membrane samples were first placed in 20 wt% glycerol solution for 2 h followed by drying in air. Afterwards for cross-section SEM imaging, the dried samples were first immersed in liquid N 2 before being carefully fractured. The samples were then dried further in a vacuum oven at 30 • C for 24 h before sputter coating them with a 5 nm layer of Pt/Pd (80%/20%) using Quorum Q150T ES sputter coater (Quorum Technologies Ltd., UK). Electron microscopy images were then taken by SEM (JSM-6010LA, JEOL, Japan). Fourier-transform infrared spectroscopy (FTIR) was performed on Spectrum Two™ (Perkin Elmer, USA) in the wavenumber range 4000 cm − 1 to 600 cm − 1 .

Pure water permeability and retention tests
The membrane was cut into a circular disk of 25 mm diameter and placed in a dead-end Amicon® cell. Deionized water from Milli-Q® water purification system was used as the feed. The cell was pressurized using N 2 gas. A pressure of 4 bar was applied at the beginning of the measurement until a steady state permeate flux was obtained. Permeate mass was measured as a function of time using a weight balance connected to a computer. Pure water permeability was calculated using equation (1).
where P is the pure water permeability (L m − 2 h − 1 bar − 1 ), J W is the water flux (L m − 2 h − 1 ), and Δp is the pressure difference between the feed and permeate side (bar). The separation performance of the membranes was studied using a BSA solution (1 g L − 1 , M W ~ 66 kg mol − 1 , average hydrodynamic diameter of 8.6 nm [22]) prepared in a 0.1 M citrate buffer at pH 5.5. The solution was fed to a dead-end Amicon® cell equipped with a magnetic stirrer positioned above the membrane sample. Retention tests were performed at 3 bar of transmembrane pressure. Permeate and retentate samples were collected and analyzed using UV-vis spectrophotometry (Shimadzu UV-1800, Japan) at λ max = 280 nm (i.e. the maximum absorbance wavelength of BSA). Retention (R in %) was calculated using equation (2): where C p is the concentration of BSA in the permeate and C f is taken as the average of feed and retentate concentrations because in the dead-end configuration, the concentration of feed solution above the membrane surface is not constant.

Results and discussion
Various parameters that influence the formation of PSS-PAH membranes by APS were studied and are discussed in four separate sections. The first two sections focus on the preparation and processing of polyelectrolyte casting solutions, including on the added amount of NaOH and on polyelectrolyte ratios. In contrast, the last two sections describe how coagulation bath conditions such as pH and crosslinker concentration can be used to tune membrane morphology, pore size, and separation performance.

Effect of added NaOH on the polyelectrolyte casting solutions
Two important choices in this pH-triggered APS approach based on the polyelectrolyte pair PSS-PAH are the pH of the casting solution and the total polymer concentration. In our previous work, we reported that the optimum polyelectrolyte solution concentration to give stable PSS-PAH ultrafiltration type membranes was 12 wt% [19], and this was used as the starting point to further showcase the versatility of pH-triggered polyelectrolyte complexation APS.
PSS is a strong polyanion that remains in a charged state over the entire pH range (pH [1][2][3][4][5][6][7][8][9][10][11][12][13][14]. On the other hand, PAH is a weak polycation with a pK a value of ~9 [23], which means that when pH ≫ pK a , PAH will be uncharged. When a PSS solution is mixed with a PAH solution that has not been adjusted to a high pH with NaOH, a polyelectrolyte complex is formed. Here, complexation occurs because the pH of the unadjusted PAH solution is approximately 3, i.e. pH < pK a . At pH 3, PAH is protonated and thus forms a polyelectrolyte complex with PSS upon mixing. In order to obtain a homogeneous casting solution of PSS and PAH, it is necessary to make the weak polyelectrolyte PAH uncharged so that it does not complex with PSS. This was achieved by adding NaOH to the PAH solution. The addition of base deprotonates PAH making it uncharged (i.e. pH ≫ pK a ). A series of PAH solutions (all at 12 wt% PAH) were prepared by adding different amounts of NaOH (C NaOH ) in order to find the optimum conditions to obtain homogeneous solutions when mixed with PSS (see Fig. S1 for photographs of the PAH only solutions). The ratio R = C NaOH /C PAH was varied according to Table 1. This ratio is a vital descriptor for the system as the NaOH concentration controls the deprotonation and thus charge of the weak polybase PAH, moreover for these viscous solutions at high hydroxide ion contents the pH cannot be accurately determined. All the prepared PAH solutions were clear and homogeneous except for R = 1.25 where the solution separates into a polymer rich and polymer lean phase (see Fig. S1). This is because PAH loses its charge at high pH and adopts a coiled conformation. The transition of PAH chains from extended conformation at pH ≪ pK a [24], to coiled conformation at pH ≫ pK a increases the hydrophobic interactions by reducing the repulsion between the PAH chains. At sufficiently large amounts of NaOH in the PAH solution such as R = 1.25, the high ionic strength of the solution may mean that PAH behaves more like a neutral polymer and thus is more prone to phase separation. This phase behavior can also be correlated to the polyelectrolyte "salting-out" effect [25].
The PAH solutions were mixed with PSS solutions in a monomer ratio of 1:2 (PSS:PAH) according to Table 1 (see Fig. 1 for photographs of the mixed solutions). The choice of this specific monomer ratio will become apparent in the next section. Fig. 1 shows that a white complex was formed for solutions where R = 0.1 and 0.25, indicating that the amount of NaOH added was insufficient to prevent polyelectrolyte complexation. A homogeneous solution was, however, obtained when R was 0.5. At R = 0.75, the solution was slightly cloudy. Further increases of R to 1.0 and 1.25 resulted in completely cloudy solutions most likely due to the aggregation of the polyelectrolyte chains due to salting out effects [26]. In addition, the latter two solutions were more gel-like than liquid-like. The behavior from R = 0.75 onwards was confirmed by turbidity measurements where the solution turbidity increased dramatically above R = 0.5 (see Fig. S2). The homogeneous polyelectrolyte solution with R = 0.5 was chosen for further experimentation.

Effect of PSS:PAH monomer mixing ratio
With R fixed at 0.5, a set of polyelectrolyte casting solutions were prepared at different monomer mixing ratios of PSS:PAH i.e. 4:1, 2:1, 1:1, 1:2, 1:3, and 1:4 (see Fig. S3 for photographs of the casting solutions). The casting solution with monomer ratio 4:1 was unstable, with two distinct phases observed a few minutes after stirring making it unsuitable for casting. Casting solutions at the other monomer mixing ratios were all homogeneous. Fig. 2 shows the dynamic viscosities of these casting solutions. Viscosity increases when the monomer ratio is changed from 2:1 to 1:1 and then it increases slightly further for the 1:2 ratio. Increasing the amount of PAH further by changing the monomer mixing ratio to 1:3 and then to 1:4 results in a decrease in the solution viscosity. A maximum in dynamic viscosity (~48 Pa s) is thus observed for the 1:2 PSS/PAH ratio. In comparison, the individual PSS solution at 12 wt% had a dynamic viscosity of 1 Pa s, while the 12 wt% PAH solution (R = 0.5) had an even lower viscosity of 0.7 Pa s. Increasing the monomer ratio of PSS:PAH evidently increases the mutual interaction between the two polyelectrolytes as the dynamic viscosity of the solutions increases up to a monomer ratio of 1:2. The maximum in the dynamic viscosity is likely due to a combination of electrostatic and hydrophobic interactions as reported in the previous studies [27,28]. There is always a balance between these two interactions, and in this case the interactions have a maximum at a monomer mixing ratio of 1:2. Literature shows that for PSS/PAH polyelectrolyte multilayers, PAH charges overcompensate PSS charges resulting in a strong deviation from 1:1 stoichiometry [29], in line with our observations. The shear stress vs shear rate graph in Fig. S4 shows the shear thinning effect of casting solutions prepared at different PSS to PAH monomer ratios. Shear stress increases dramatically with the shear rate for the solution of 1:2 monomer mixing ratio, indicating stronger interactions between PSS and PAH chains. On the other hand, the increase in shear stress with shear rate is less abrupt for the 2:1 mixing ratio solution suggesting much weaker interactions between PSS and PAH. When the PSS:PAH ratio increases beyond 1:2, the dynamic viscosity decreases to 40 Pa s for the 1:3 ratio and 25 Pa s for the 1:4 ratio. This behavior is most likely due to decreased PSS-PAH interactions at these ratios.
The PSS-PAH casting solutions with monomer ratios of 2:1 through to 1:4 were cast on glass plates as 0.6 mm thin films and immediately immersed in coagulation baths at pH 1 with 0.05 wt% GA. GA is a commonly used chemical for the crosslinking of PAH due to its ability to react with the amines in aqueous conditions [30,31]. The aldehyde groups of GA react with the amine groups of PAH to form imine (-C--N-) bonds as shown in the reaction scheme in Fig. S5. The concentration of GA in the coagulation bath and the bath pH are very important parameters that have a significant influence on membrane structure and performance and these factors will be discussed in the later sections. Fig. 3 shows photographs of the polyelectrolyte complex membranes prepared with casting solutions of different monomer mixing ratios. At the monomer mixing ratio of 2:1, the resulting film resembled a membrane, but did not have the mechanical strength to be further processed. At 1:1 monomer mixing ratio, the resulting films were relatively brittle and could not be handled further for pure water permeability measurements without rupturing. While at 1:2 mixing ratio, the resulting films were mechanically stable. Increasing the mixing ratio to 1:3 and to 1:4 resulted in films that were relatively soft with uneven top surfaces. Fig. 4 shows the cross-section and top surface SEM images of the films shown in Fig. 3. Films prepared using a monomer mixing ratio of 2:1 were too weak to be processed for SEM analysis. The cross-section SEM image shown in Fig. 4a reveals that there are no macro-voids in the substructure of the 1:1 PSS-PAH film. At 1:2 monomer mixing ratio (Fig. 4b), the resulting films showed small voids in the substructure. For the monomer mixing ratios of 1:3 and 1:4, the resulting films have finger-like macro-voids that exist throughout the thickness of the film (see Fig. S6). Another observation from the SEM images of Fig. S6 is that the 1:3 and 1:4 membranes are much thicker than 1:1 and 1:2. This could be due to the excess of positively charged PAH in the polyelectrolyte complex membrane which can lead to swelling of the structure. In addition, the top surface of 1:3 and 1:4 membranes is uneven and shows signs of inhomogeneity ( Fig. 4c and d), which can also be seen in the photographs shown in Fig. 3. The existence or absence of finger-like macro-voids is usually correlated with the viscosity of the casting solution, which affects the phase inversion kinetics, as explained in detail by Smolders et al. [32]. The finger-like macro-void morphology is usually obtained when the casting solution viscosity is low, favoring instantaneous demixing (rapid phase inversion). However, in our case, the differences in the viscosity of casting solutions at 1:1, 1:2, and 1:3 monomer mixing ratios is small, but still the morphology of the resultant  Table 1. Here R = C NaOH /C PAH , where C NaOH is the concentration of NaOH in PAH solution (wt%) and C PAH is the concentration of PAH (wt%). All the solutions contain a total of 12 wt% polyelectrolyte. membranes is significantly different. This suggests that the aggregation behavior of the polyelectrolyte complex varies with different monomer ratios. We expect that as the monomer mixing ratio is increased from 1:1 to 1:4, the polyelectrolyte chains tend to form more loosely packed structures. A possible reason for such behavior could be that at larger excess amounts of PAH, such as for the 1:2 and especially the 1:3 and 1:4 mixing ratios, there is also an excess amount of charge present in the polyelectrolyte complex. In water such excess charge would result in swelling and thus a much more open material. Moreover, the excess charge could hinder phase separation, only allowing the formation of a much more swollen and open material. Indeed, the SEM images shown in Figs. 4 and S6 demonstrate that the 1:1 and 1:2 mixing ratio membranes have denser structures than 1:3 and 1:4 membranes. Furthermore, the 1:3 and 1:4 membranes were physically more soft as will be shown by the permeability data. In case of 1:4 mixing ratio, the existence of finger-like macro-voids could likely also be due to the lower solution viscosity.
The pure water permeability of the films shown in Figs. 3 and 4 are presented in Fig. 5. The permeabilities were expected to be low, as SEM images clearly showed membranes with relatively dense top layers. The 1:2 monomer mixing ratio membrane had a pure water permeability of ~15 L m − 2 h − 1 bar − 1 . When the mixing ratio is increased to 1:3 and then to 1:4, the permeability of the resulting membrane did not change significantly. However, the variance between the water permeability values measured for three different samples of the 1:3 and 1:4 membranes was higher confirming that these membrane structures were not homogeneous (see the SEM images in Fig. 4c and d). Another major difference here is that the 1:4 mixing ratio membranes undergo extensive structural compaction at 4 bar of applied water pressure compared to the 1:2 and 1:3 membranes (see Fig. S7). Structural (or pore) compaction manifests as a decrease in water permeability over time, which typically stabilizes after some time period. The compaction shown in Fig. S7 for the 1:4 membrane signifies that it is physically   softer and more porous. Due to the high variance in water permeability values and their inhomogeneous structure, 1:3 and 1:4 membranes could not be used for further processing. As the membranes prepared from solutions with a monomer ratio of 1:2 showed stable and reproducible pure water fluxes (see Fig. S7), this ratio was selected for further investigations.

Effect of pH of the coagulation bath
In this APS approach, the membrane formation is driven by a pH gradient from the casting solution (high pH) to the coagulation bath (low pH). So naturally, the pH of the coagulation bath will have a significant impact on the phase inversion kinetics and resulting membrane morphology. To understand the influence of this parameter, a solution with monomer ratio of 1:2 was cast, as before, and immersed in coagulation baths at different pH values (with 0.05 wt% GA). The pH of the bath was varied from 0.25 to 1.5 with increments of 0.25 (adjusted using HCl). Photographs of the cast membranes are shown in Fig. S8. The membranes prepared at pH = 0.25 were relatively rigid as compared to the others. As the pH of the coagulation bath increases, the membranes lose some of their rigidity but they remain mechanically stable and can be operated up to 20 bar of applied water pressure (Fig. S12). The membranes prepared at pH 1.25 and 1.5 visibly showed rougher surfaces than the other membranes.
Cross-section and top surface SEM images of these membranes are presented in Fig. 6. Cross-section images indicate that all the membranes have a skin layer on top with a porous substructure. It is important to mention that the top layer is not fully dense, but rather contains small pores as will be revealed later from the BSA retention measurements. The substructure directly beneath the top layer in the cross-section SEM images of Fig. 6 changes drastically when the pH of the coagulation bath is increased from 0.25 through to 1.5. At pH 0.25 (Fig. 6a), the substructure does not show clear pores at the shown magnification. But when the coagulation bath pH is increased to 0.5, small pores start to appear in the substructure (Fig. 6b). As the pH of the bath is increased further, the size and density of these pores also increases. The cross- section images in Fig. 6c-f shows a skin layer with a porous substructure. Due to the higher polyelectrolyte concentrations and high viscosities of the PSS-PAH casting solutions, no finger-like macro-voids were observed in any of these membranes [19].
The pH switch in APS can be considered analogous to the "solventnon-solvent" transition in traditional NIPS. In this version of APS, the polyelectrolytes in the casting solution are soluble and miscible with each other in water at pH ~14. Decreases in pH, as seen earlier in Fig. 1, results in complexation and phase separation. This is because as the pH of the solution is lowered (towards and then below the pK a of PAH), PAH becomes charged and therefore can undergo complexation with the fully charged PSS. In other words, at high pH conditions, water acts as a solvent while at low pH conditions, water acts as a non-solvent for the PSS-PAH solution.
At very low pH conditions in the coagulation bath, such as pH 0.25, the difference between the pH of polyelectrolyte casting solution (~13-14) and the coagulation bath (pH 0.25) is large, resulting in a strong driving force for the H + ions in the bath to diffuse into the polyelectrolyte solution. Rapid exchange of H + ions results in immediate polyelectrolyte complexation between PSS and PAH and hence faster precipitation. It is known from literature on the NIPS process that a stronger interaction between the solvent and the non-solvent results in denser membranes [33]. The same is also true for APS membranes prepared in a coagulation bath at pH 0.25, where the abundance of H + ions results in stronger interactions with the polyelectrolyte casting solution, favoring denser complexes. This effect is also seen in the SEM images of Fig. 6(a, g) where the membranes do not contain any visible pores at the shown magnification. However, there is a stark difference between the substructures of APS and NIPS membranes when the precipitation rate is fast. As mentioned earlier, instantaneous demixing in NIPS typically results in membranes that contain macro-voids in the substructure. But in case of APS, the H + ions continue to diffuse through the skin layer because of their small size. In other words, there is little mass transfer hindrance for the H + ions and therefore, the precipitation below the skin layer continues and this acts to prevent the formation of macro-voids. In APS, the limiting parameter then is the concentration of H + ions (or pH) in the coagulation bath, and this determines the porosity of the substructure. With increases in bath pH (less H + ions), the driving force for complexation gradually decreases and this results in a slower precipitation rate and hence results in membranes with more porous substructures. Fig. 7 shows the pure water permeability and BSA retention of the membranes prepared in coagulation baths with varying pH values. The membranes prepared at pH 0.25 had a pure water permeability of ~5 L m − 2 h − 1 bar − 1 with ~98% BSA retention. The pure water permeability and BSA retention remained similar (~15 L m − 2 h − 1 bar − 1 and 97%) for the membranes prepared in coagulation baths with pH 0.5 through to 1. This behavior indicates that the membranes are ultrafiltration type membranes that are well suited for the concentration of protein solutions. When the pH of the coagulation bath was increased to 1.25 and 1.5, the pure water permeability of the resultant membranes increased to ~30 L m − 2 h − 1 bar − 1 and 110 L m − 2 h − 1 bar − 1 , respectively. The membranes prepared at these pH conditions had defects in the structure due to the inhomogeneous surface (see Fig. S8 for photographs of the membranes) and therefore, as expected, these membranes did not show any BSA retention.
The water permeability of the membranes prepared at pH 1 and 1.5 coagulation bath were also measured at 20 bar of applied pressure in order to further explore the mechanical strength of these membranes (see Fig. S12 in the Supplementary information). Membranes prepared using the pH 1 coagulation bath showed stable permeability (~15 L m − 2 h − 1 bar − 1 ) up to 20 bar of applied water pressure indicating that the membrane structure remains intact. However, the membranes prepared in pH 1.5 coagulation bath showed a decrease in water permeability with increasing water pressure, which shows that these membrane undergo slight structural compaction at increased pressures.
It is clear that the pH of the coagulation bath is an important parameter to tune the properties of APS membranes because it directly affects the rate of phase inversion. The membranes prepared in pH 1 coagulation bath were chosen for further experiments because of their high BSA retention and higher water permeability.

Effect of GA concentration in the coagulation bath
In the described APS approach, a small amount of crosslinker (GA) is added to the coagulation bath to improve the mechanical properties of the membranes. Up until now, the GA concentration in the coagulation bath was kept constant at 0.05 wt%. However, the concentration of GA also affects the properties of the resultant membranes as it determines the crosslinking density causing the densification of the structure [34]. To investigate this, membranes were prepared in baths at pH 1 with varying GA concentrations i.e. from no GA to 0.5 wt% GA (see Fig. S9 for photographs of the resultant membranes). Fig. 8 shows the SEM images of the membranes prepared with varying concentrations of GA in the coagulation bath. It appears from the cross-section SEM images that the thickness of the top layer increases with increasing amount of GA. The membranes prepared at GA>0.15 wt % were physically more rigid, while at 0.5 wt% GA, the resulting membranes were too brittle to be used for the water permeability measurements. Cross-section SEM images reveal that the pore size of the substructure reduces significantly with the addition of GA in the bath until 0.15 wt%. It is important to mention here that the polyelectrolyte complexation/precipitation process occurs at the same time as the crosslinking reaction. These two kinetic parameters are likely linked, which results in different membrane structures.
FTIR spectra of the membranes is shown in Fig. S11. After crosslinking, imine bonds will be formed and this peak should appear between 1635 cm − 1 and 1645 cm − 1 [35,36]. However, amine (N-H bending) groups from the non-crosslinked moieties also give rise to a characteristic peak between 1580 and 1627 cm − 1 [35]. As these two regions are very close to each other, it is very difficult to draw any conclusions on the effect of GA concentration from the FTIR spectra.
The average pore size of these membranes was estimated by analyzing the top surface SEM images shown in Fig. 8 using ImageJ software. The pore size distributions are shown in Fig. S10. The average pore sizes are summarized in Fig. 9 where the error bars are an indication of pore size distribution. The average pore sizes measured using the SEM images may not reflect the true pore size of the membranes during operation, but they do provide insight into the relative differences in pore size between the different membranes. The membranes prepared without the GA in the bath had an average surface pore size of ~46 nm.
Increasing the GA to 0.05 wt% resulted in membranes with an average pore size of ~2 nm. Further increases in the amount of GA crosslinker in the coagulation bath resulted in an increased surface pore size of the membrane to ~18 nm at 0.1 wt% GA, and then to ~46 nm at 0.2 wt% GA. At 0.5 wt% GA, the average pore size was estimated to be ~450 nm (not shown in figure). It is important to note here that the crosslinking reaction of PAH with GA occurs at the same time as the polyelectrolyte complexation of PAH with PSS. Therefore, as the concentration of GA is increased, the crosslinking reaction becomes more likely and this may compete more with the polyelectrolyte complexation. In turn, this would result in less PAH being available for complexation with PSS and may lead to the formation of larger voids in the resultant complex and hence the greater measured pore sizes of the membranes. The overall behavior indicates that there exists an optimum concentration of GA, which results in membranes with the most compact structure. A similar optimum concentration of GA has been reported by Duong et al. for the PSS-PAH multilayer based membranes [37]. For APS with PSS/PAH the GA concentration can be seen as an additional control parameter to control the eventual membrane pore size, a parameter not available in traditional NIPS. Fig. 9 shows the effect of GA concentration present in the coagulation bath on the pure water permeability of the resulting membranes. Following the same trend as the membrane pore sizes, the pure water permeability shows a minimum at 0.05 wt% GA, in line with the smallest observed pores. Membranes prepared without GA in the bath showed a pure water permeability of ~45 L m − 2 h − 1 bar − 1 , while the pure water permeability decreases to ~15 L m − 2 h − 1 bar − 1 for the membranes prepared with 0.05 wt% GA. This is the same membrane that was able to  retain ~97% BSA discussed in the previous section. Further increases in the amount of GA in the bath, results in membranes with a higher water permeability, which is due to the increase in the average pore size of the membranes.
The water permeability of the membranes prepared in coagulation baths containing different amounts of GA was also measured at 20 bar of applied pressure to further evaluate their mechanical stability (see Fig. S13 in the Supplementary information). The results show that the pure water permeability decreases as the applied pressure is increased to 20 bar, suggesting that a degree of structural compaction occurs. At 0.05 wt% GA in the bath, the resultant membranes exhibited stable water permeability even up to 20 bar (see Fig. S12). Upon increasing the concentration of GA in the bath to 0.2 wt%, the membrane pore size increases to ~46 nm and the membrane shows compaction under pressure, but importantly the membrane does not rupture and operation can continue accounting for the structural compaction.
The results reveal that the concentration of GA is a key parameter that acts to improve the mechanical properties and tune the separation performance of the PSS/PAH based APS membranes.

Chemical stability of the membranes
The chemical stability of membranes prepared using the coagulation bath at pH 1 with 0.05 wt% GA were studied by immersing them in aqueous solutions at different pH values, specifically 1, 7, 8, 10, and 12 for 7 days and subsequently measuring their water permeability. The results shown in Fig. S14 of the Supplementary information reveal that the pure water permeability of the treated membranes remains the same at ~15 L m − 2 ⋅h − 1 ⋅bar − 1 when immersed in pH 1-8 solutions. However, the membranes immersed in pH 10 and 12 solutions exhibited increased pure water permeability i.e. ~300 and ~3400 L m − 2 h − 1 bar − 1 , respectively. At pH > 8 (greater than the apparent pK a of PAH), the PAH becomes deprotonated, losing its charge. It is possible that at pH 10 and 12 some of the PSS-PAH complex is broken because of the uncharging of PAH leading to a more open membrane structure. Re-washing this complex membrane with deionized water does not fully recover the complex to its original, untreated state, suggesting irreversible structural changes.
Sodium hypochlorite (NaClO) is a common oxidant used for cleaning polymeric membranes [38]. To evaluate the stability of the PSS-PAH membranes against NaClO treatment, again the membranes prepared in pH 1 bath with 0.05 wt% GA were immersed in different concentrations of NaClO aqueous solutions. The solutions were prepared by diluting NaClO to 300 ppm, 800 ppm, and 1200 ppm with deionized water. The pH of the solutions was adjusted to 8 using 0.1 M NaOH or 0.1 M HCl solution. The membranes were immersed in the NaClO solutions for 1 h, subsequently removed and then washed with deionized water before their pure water permeability was re-measured. The results as shown in Fig. S15 of the Supplementary information and reveal that the pure water permeability of the membranes does not change after exposure to NaClO solutions even up to 1200 ppm h showing that the PSS-PAH membranes are stable against a NaClO cleaning treatment.

Conclusions
Just like NIPS, the fabrication of Aqueous Phase Separation (APS) membranes via polyelectrolyte complexation of PSS and PAH is a versatile approach involving many different tuning parameters. All these parameters influence the thermodynamics and kinetics of phase inversion in one way or the other. In this study, we have systematically investigated the effect of polyelectrolyte casting solution composition and the coagulation bath conditions on the structure, morphology, and properties of PSS-PAH membranes. The results indicate that the casting solution composition is one crucial parameter that controls the mutual interactions between PSS and PAH. Here, an optimum amount of base (R = NaOH/PAH = 0.5) was added to the PAH solution to obtain homogeneous casting solutions when mixed with PSS. R values < 0.5 resulted in polyelectrolyte complexation in the casting solution while R > 0.5 resulted in a difficult to process viscous gel-like phase. Similarly, the monomer mixing ratio of PSS to PAH in the casting solution is a critical parameter that greatly influences the membrane formation process. It was found that the monomer mixing ratio of 1:2 (PSS:PAH) resulted in casting solutions with desired high dynamic viscosities (~50 Pa s), due to stronger interactions between the two polyelectrolytes. A clear optimum in the casting solution composition was thus observed, but it was found that the conditions of the coagulation bath had a greater influence on the membrane pore size and separation performance. The kinetics of phase inversion could be controlled by varying the pH of the coagulation bath. Results revealed that at pH 0.25-1, the membranes showed stable water permeability and high BSA retentions (~97%), with lower pH values leading to denser membranes. Increasing the bath pH beyond pH 1, resulted in the slowing down of the precipitation rate and this induced inhomogeneity in the resulting membrane structures. At a coagulation bath pH > 1, the resulting membranes showed higher water permeabilities and much lower protein retentions. The concentration of glutaraldehyde (GA) in the bath, which was used to crosslink the PAH chains, was varied from 0 to 0.5 wt%. The presence of the crosslinking agent improved the mechanical stability and density (lower pore size) of the membranes. The surface pore size of the membranes could be tuned from 2 nm (without GA) to 46 nm (with 0.2 wt% GA). At 0.05 wt% GA in bath, the resulting ultrafiltration type membranes showed water permeability of ~15 L m − 2 h − 1 bar − 1 and ~97% BSA retention. Moreover, the membranes show chemical stability upon treatment with NaClO solutions till 1200 ppm h.
Overall in this work, and to a lesser extent in our earlier work [19], we have established that membrane fabrication by the APS approach has a lot of similarities with the traditional NIPS process. From the molecular weights of the polyelectrolytes used, to the composition of the casting solutions, and also the coagulation bath conditions, each parameter/step is crucial in determining the morphology and separation performance of the resulting membranes. This study based on the complexation of PSS and PAH highlights the many parameters that are available to fine tune the properties of polyelectrolyte complex membranes prepared by APS. With these tuning parameters sustainable membranes could be optimized for specific applications. Additional parameters such as the addition of additives to the polyelectrolyte casting solution as well as moving to less extreme pH conditions, facilitated by using different polyelectrolyte pairs, will be the focus of our future research to further highlight the significant potential of APS as a sustainable approach to membrane production. Moreover, polyelectrolytes with easily cross-linkable side groups could be used to avoid the use of glutaraldehyde in the coagulation bath.

Declaration of competing interest
The authors declare no conflicts of Interest.