On the role of the porous support membrane in seawater reverse osmosis membrane synthesis, properties and performance

Derrick S. Dlamini, Javier A. Quezada-Renteria, Jishan Wu, Minhao Xiao, Mackenzie Anderson, Richard B. Kaner, Arian Edalat, Nikolay Voutchkov, Ahmed Al- Ahmoudi, Eric M.V. Hoek

Abstract

In this study, we explore the role of the polysulfone (PSU) support membrane skin-layer and whole-body pore morphology on the physical-chemical properties and separation performance of hand-cast polyamide-PSU (PA- PSU) composite seawater reverse osmosis (SWRO) membranes. We tailor the support membrane pore morphology by varying the PSU concentration and the coagulation bath temperature. In order to isolate the impacts of the PSU support, all of the porous supports are coated using identical m-phenylene diamine (MPD) and trimesoyl chloride (TMC) solution compositions, reaction conditions, and post-treatments. As PSU concentration increases, PSU support membrane pore size, surface and body porosity, water permeability, MPD mass uptake (by the PSU support), and PA-PSU composite membrane water permeability decrease, while MPD and TMC conversion, PA film mass and thickness, crosslinking degree (XD), and PA-PSU composite membrane NaCl rejection increase. As PSU support membrane coagulation bath temperature increases, support membrane pore size, surface and body porosity, MPD uptake, PA film mass, thickness and XD, MPD and TMC conversion, and NaCl rejection increase, while composite PA-PSU membrane water and NaCl permeability decrease. Composite membrane permeabilities were 70–90 percent of the (theoretical) unsupported PA film permeabilities due to the high XD and low PA coating film thickness. Composite PA-PSU membrane water/salt permeabilities both correlate most strongly with PA coating film mass/thickness and XD, which in turn correlate most strongly with TMC conversion. The key to increasing water permeability is lower PA film mass/thickness and XD with higher support membrane surface pore size and porosity. In contrast, the key to increasing NaCl rejection is higher PA film mass/thickness and XD with lower support membrane surface pore size and porosity.

1. Introduction

Reverse osmosis (RO) membranes most often comprise a “thin film composite” (TFC) structure, where a polyamide (PA) selective layer is synthesized in situ to form an ultra-thin and highly-crosslinked micro porous coating film on top of a mesoporous (at the top surface) poly sulfone (PSU) support [1]. The selective layer is often synthesized by interfacial polymerization of 1,3-diaminobenzene (or “meta-- phenylenediamine,” MPD) and 1,3,5-benzenetricarbonyl trichloride (or “trimesoyl chloride,” TMC) [2]. The separation performance of an MPD-TMC coating film is governed by its mass/thickness, morphology, crosslinking degree (XD), and charge density, which give rise to tunable water and salt permeability ranging from “ultra-low pressure” RO membranes for water recycling, “low-pressure” RO membranes for brackish water desalination and “high-pressure” RO membranes for seawater desalination [3]. Understanding the fundamental mechanisms governing the interfacial polymerization reaction is essential for tailoring PA membrane separation performance. The MPD and TMC monomer concentrations and their ratio determine the mass/thickness and the XD of the microporous PA film; these, in turn, govern the  separation performance of the resultant PA coating film [4–8].

As depicted at the top of Fig. 1, the XD indicates the proportion of acid chloride groups (Cl–C–– O) of TMC that react with the amine (–NH ) groups of MPD forming the amide bonds and the ultimate polyamide film structure. Unreacted acid chloride groups hydrolyze in water to form pendant carboxylic acid groups, which give rise to the net negative charge of PA membrane surfaces [9]. In the lower left image of Fig. 1, the three layers of a typical PA composite membrane are shown: on top, an ultra-thin microporous PA coating film over a PSU mesoporous support membrane previously phase inverted on top of a microporous polyester nonwoven fabric. In the lower right image of Fig. 1, the formation of a PA film via interfacial polymerization is shown (from bottom to top): (1) a bare porous PSU support membrane, (2) saturation of the aqueous MPD solution into the PSU support, (3) contacting the aqueous MPD saturated support with an immiscible organic solvent containing TMC, and (4) a fully formed PA film. Not shown is the process of removing excess aqueous amine solution from the surface of the PSU support membrane, which occurs between steps (2) and (3). At the top of Fig. 1, the theoretical structure of MPD-TMC films when fully cross linked (m) and fully linear (n) are provided. Theoretically, as shown by Kim et al. [8], the elemental N/O ratio of a MPD-TMC coating film that is 100% X-linked will be 1, while a completely linear film will have N/O ratio of 0.5.

The concentration of MPD that participates in the interfacial polymerization reaction is related to the mass of MPD solution taken up by the PSU support membrane through its surface pores and retained within its body porosity. Therefore, the resultant PA coating film thickness and XD are not determined by the MPD concentration of the as-prepared aqueous MPD solution, but rather by the MPD mass held within the support membrane’s body porosity. The mass of PA film and the XD are also influenced by the TMC concentration, but more specifically, the mass of TMC in the reaction zone above the porous support membrane [10]. These concepts are conceptually illustrated in Fig. 2. The support membrane solid phase is shown in solid blue, while the void volume is filled with MPD solution with MPD represented by the red dots. In Fig. 2a/b we see two idealized support membranes with different porosities where it is obvious that support (a) can take up 2X more amine monomer solution than support (b). Next, in Fig. 2c/d, we see both MPD saturated supports now in contact with the immiscible organic TMC solution shown with red dots in a green background. Now, in Fig. 2e/f, we see that support (a) delivers ~2X the amine monomer mass forming a much thicker reaction zone relative to support (b); while the aqueous MPD concentrations are identical in both supports, 2X mass of MPD is available in support (a) to partition into and diffuse through the organic TMC solution above. Finally, in Fig. 2g/h, the amount of polyamide formed over the PSU support (a) is more than that formed over PSU support (b) because there is more mass of amine monomer available to react with the relatively infinite source of TMC. Since sup port (a) contains and delivers ~2X the mass of MPD into the reaction zone, the mass and mole ratios of MPD:TMC will be proportionately higher than the situation involving support (b), and hence, the cross linking degree may be significantly different.

As first pointed out by Lonsdale et al. [11], the porous support membrane skin-layer pores play a role in water and solute permeation through a composite membrane. The apparent permeability (for both water and salt) of the composite is lower than it would be through the thin film alone because part of the membrane film is effectively blocked by the solid phase (in between pores) of the support membrane. This is illustrated in Fig. 3. Thus, a composite membrane has an “effective thickness,” δeff , accounting for the increased path length for permeation (see the arrows in Fig. 3a/b), where water and solutes must transport both across and laterally through the PA film until they find an under lying pore through which they exit. Decades later, Ramon et al. [12] extended this concept showing that higher porosity and smaller pores produced higher water and solute permeability, whereas lower porosity and larger pores produced lower water and solute permeability. They further elucidated the combined effects of support membrane pore morphology and coating film morphology (roughness and thickness). For coating films of constant thickness, Ramon et al. [13] provided an analytical expression to approximately model the impacts of support membrane pore size and porosity on composite membrane transport.

The porous PSU supports are formed via nonsolvent induced phase separation (NIPS), in which, (1) the PSU polymer is dissolved into a good solvent (e.g., NMP, DMF, DMAC, DMSO, etc.), (2) spread over a nonwoven fabric to a certain thickness, and (3) immersed into a “precipitation bath” containing a very poor, but miscible solvent or “non solvent” [14]. This generally produces an asymmetric structure with fewer, smaller skin-layer pores with increasing pore size and porosity moving towards the bottom layer in contact with the fabric; in some cases, the PSU body may also contain significantly larger macrovoids throughout [15]. Approaches to tailor PSU membrane skin-layer pore size and porosity in addition to the macrovoid morphology include: (1) changing the PSU molecular weight and/or polydispersity index, (2) use of different solvents and co-solvents, (3) use of pore forming agents, e.g., polyethylene glycol (PEG) or polyvinylpyrrolidone (PVP) of varied molecular weights, (4) addition of solvent, coagulant salts and/or surfactants to the nonsolvent precipitation bath, and (5) altering the temperature of the polymer casting solution and/or the nonsolvent precipitation bath [16]. These approaches are well known and have been reviewed previously [17].

Fig. 1. (Lower left) Illustration of three layers of an interfacial composite membrane, typically, comprising an ultra-thin microporous PA coating film over a PSU mesoporous support membrane previously cast on top of a polyester nonwoven fabric. (lower right) Illustration of the steps involved in the interfacial polymerization reaction including: (from bottom to top) saturation of the aqueous amine solution in the PSU support, contacting the aqueous amine saturated support with a nonpolar organic solvent containing the acyl halide monomer, and a fully formed PA film. At top, the theoretical structure of MPD-TMC films when fully crosslinked (m) and fully linear (n). Note the PA structure was reproduced from Ref. [8].

 Fig. 2. Illustration of the role of support membrane porosity in interfacial polymerization.

Fig. 3. Graphical illustration of the longest and shortest permeation path lengths (black arrows) for water and solutes through composite membranes with (a) high porosity and (b) low porosity. The red layer represents a polyamide coating over the blue PSU support solid phase with columns representing the PSU skin-layer pores. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

We know of only two publications that report on the impacts of support membrane pore morphology on the performance of a uniformly synthesized MPD-TMC coating film [16,18]. However, in both publications they produced PSU supports with large finger-like macrovoids, which are thought to lead to compaction and loss of performance over time in high pressure applications like SWRO, and in one [16], they did not differentiate between pore morphology and hydrophilicity. Herein, our hypothesis is that each combination of PSU support membrane skin-layer pore size, skin-layer porosity, and whole-body porosity will govern the amounts of MPD uptake and release, and hence, will deter mine the MPD:TMC ratio in the reaction zone producing PA films of different mass/thickness, XD, and separation performance. Moreover, PSU support membrane skin-layer pore size and porosity contribute significantly to composite membrane separation performance by altering the effective permeation path length through the composite structure. In this study, we vary the PSU support membrane skin-layer and whole-body pore morphology by varying the PSU concentration (in DMF) as well as the temperature of the coagulation bath. We characterize PSU support membrane and composite PA-PSU membrane SWRO separation performance and physical-chemical properties. Finally, we demonstrate that by varying the PSU support pore morphology, one can fine-tune the PA film mass/thickness, XD and separation performance to achieve SWRO performance on par with commercial SWRO membranes.

2. Materials and methods

2.1. Chemicals and reagents

All of the following chemicals were purchased from Sigma Aldrich (St. Louis, Missouri, USA) and used as received: dimethylformamide (DMF), meta-phenylenediamine (MPD, 99.5 %), tri-mesoyl chloride (TMC) ≥98 %, (+)-Camphor-10-sulfonic acid (CSA) ≥98 %, triethyl amine (TEA) ≥99 %, sodium dodecyl sulfate (SDS), sodium chloride (NaCl), sodium hydroxide (NaOH) and citric acid. Isoparᵀᴹ G was pro cured from (Gallade Chemical Inc. (Santa Ana, CA, USA) and 35 kDa Ultrason PSU was generously supplied by BASF (Charlotte, North Carolina, USA).

2.2. PSU support membrane preparation

Porous, flat-sheet ultrafiltration (UF) membranes were prepared by NIPS. The casting solution comprised of PSU dissolved in DMF was prepared by continuously stirring at 150 rpm at 80◦C with an MXD All Purpose Mixer (MXD Process, Jeffersonville, Indiana, USA) until the solution became homogeneous. The PSU concentrations used were 16, 17 and 18 wt%. After mixing, PSU casting solutions were cooled to room temperature (23◦C) and stored in a vacuum desiccator overnight for degassing. The viscosity of selected casting solutions was determined using a viscometer (USS-DVT6 Viscometer, US SOLID) at 12 rpm and room temperature.

Thin PSU films were cast over a polyester terephthalate (PET) non- woven fabric (Excester 75AX, Hirose, Japan) with a weight of 75 g/ m². The PET fabric was taped to a glass plate. The polymer films were drawn out using a casting knife (Gardco Co., Pompano Beach, Florida, USA) set at blade height of 200-μm, unless otherwise stated. The ensemble was immersed in the coagulation bath 5 seconds after drawing out the film. The coagulation bath consisted of 2 wt% DMF in 18 MΩ ultra-pure water (UPW) at temperatures of 23, 35 and 45°C. The phase inverted membrane remained in the coagulation bath for 10 min. Next, the PSU support membranes were removed from the coagulation bath and immersed into a bath containing NaOH in UPW at pH 10 and 50°C for 10 min, then stored in UPW water in light-proof containers at 5 °C until use.

2.3. PA coating film formation

A matrix of experiments was designed to produce flat sheet mem branes with the target water permeability >1 LMH/bar and 32 g/L NaCl rejection >99.5 %. In each case, 3.4g of MPD was dissolved in 93.45g of UPW. To enhance uptake by the support membrane, 0.15 wt% SDS was added to the MPD-DI solution in all cases. In addition, 2.0g of CSA was added into the MPD-DI mixture, which dropped the pH to 3 to 4. Then, 1.0g of TEA was added into the CSA-MPD-DI mixture to bring the pH back up to 8 to 9. TMC was dissolved in Isopar G at a concentration of 0.17 wt% in all experiments.

A 17cm x 11.3cm dry PSU support membrane was placed on a clean, dry glass plate and clamped down under a 15 cm x 10 cm wooden picture frame. The 100 g MPD-SDS-TEA-CSA-UPW mixture was poured onto the PSU support within the frame. The amine monomer solution coating time was fixed at 120s. Next the solution was poured out and the frame removed. Excess solution on the surface of the membrane was blown off with dry, clean air using an air knife at an angle below 45° in one slow pass at 1 bar. After making sure no aqueous droplets remained on the membrane surface, the membrane was then dipped into the TMC- Isopar G solution for 10s unless otherwise stated. The membrane was put in an oven at 60°C for 10 min to accelerate cross-linking without removing the Isopar G, and then 115°C for 2 min to evaporate the Isopar G solvent.

Next, the cured seawater RO membranes were subjected to two stages of chemical treatment. First, the membranes were soaked in a 1g/L aqueous solution of citric acid at 50°C for 10 min. The was removed from the citric acid solution and immersed into UPW at 50°C for 10 min. Next, the membranes were immersed into a 1 g/L sodium bicarbonate solution (pH ~11) at 50°C for 10 min, and then immersed into UPW at 50°C for 10 min. Finally, membranes were immersed into containers of fresh UPW at room temperature twice, and then stored in fresh UPW in light-proof containers in a laboratory refrigerator at 5°C until use.

2.4. Membrane separation performance

2.4.1. Ultrafiltration membrane separation performance

The performance of the UF membranes was evaluated by measuring pure water flux and rejection of 1 g/L of 100 kDa polyethylene glycol (PEG). The experiments were conducted using a dead-end filtration cell (Stirred Cell, Sterlitech Corp., Auburn, WA, USA). The filter cell accommodates 40 mm diameter membrane samples, which are supported by a stainless-steel porous frit. All membranes were tested at 1 bar, unless otherwise stated. The pressure was supplied by N2 gas cylinder (Airgas USA, Radnor, Pennsylvania, USA). The membrane was compacted at 1 bar for 30 min, after which, the DI water flow rate was measured as volume over time. The flow rate was used to calculate the water flux using Eq. (1) with Δπ =0 (see section 2.4.2 below). The separation performance was tested by the rejection of PEO at 1 bar. The concentration of the PEO was measured via total organic carbon (TOC) using a TOC analyzer (Shimadzu TOC-LCSN).

2.4.2. SWRO membrane separation performance

The separation performance of the SWRO membranes was evaluated in a bench scale Sterlitech crossflow membrane filtration system equipped with three (3) membrane cells at 4 cm wide by 8.5 cm long and 1 mm channel height giving an active membrane area of 34 cm2 (5.27 in2) in each cell, the crossflow rate was 0.333 gpm. The permeate flow rate (Qp) was directly measured with a digital flow meter (Optiflow 1000; Agilent Technology, Foster City, California, USA). The feed water temperature was held constant 23 ±1°C using a recirculating heater/ chiller with an immersion coil in the feed tank. The saltwater flux and salt rejection were tested using 32 g/L NaCl at an applied pressure of 55.2 bar (800 psi). Electrical conductivity (EC) calibration curves were linear for concentration between 0 and 2000 ppm; hence, observed rejection was calculated directly from feed and permeate EC values below 2000 ppm. The feed solution was diluted 100X and permeate samples above 2000 ppm were diluted 10X to stay within the EC meter calibration range, and then converted back to the real concentration by multiplying by the dilution ratio. The volumetric water flux was calculated from

where A is the water permeance, Δp is the applied feed pressure, Δπ [ = fos(cm – cp)] is the trans-membrane osmotic pressure, Jv (=Qp/Am) is the volumetric permeate flux and Am is the area of the membrane inside the cell. The solute flux is calculated from

where B is the solute permeance, cm is the membrane surface solute concentration and cp is the permeate salt concentration. Eq. (2) can be re-arranged with Js =Jvcp to produce

And from the definition of the intrinsic salt rejection (rs =1-cp/cm) one obtains:

Eq. (4) reveals that the intrinsic rejection by a membrane is a function of an intrinsic parameter, B, and at least one operating condition, Jv. Further re-arranging the right side of Eq. (4) produces a practical equation to determine the B-value of a membrane according to:

The intrinsic rejection cannot be measured directly because cm cannot be measured directly; however, the observed solute rejection (Rs) can be measured and is described by

where cf is the feed solute concentration.

In RO membrane separations, rejected solutes accumulate at the surface of the membrane and diffuse back out into the bulk of the crossflow. The ratio of the solute concentration at the membrane surface to that in the bulk of the flowing feed stream is known as the concentration polarization (CP) modulus, which can be experimentally deter mined from

where ks is the solute mass transfer coefficient. According to boundary layer theory, the mass transfer coefficient is Ds/ δbl where Ds is the solute diffusivity and δbl is the hydrodynamic boundary layer thickness. The boundary layer thickness is a function of the crossflow velocity and gives rise to a form of equation like,

For a thin rectangular crossflow channel filled with a mesh feed spacer [19]. Here, Sh is the Sherwood number, Re is the Reynolds number (=ρudH/μ), Sc is the Schmidt number (=μ/ρDs), dH (=2H, where H is the channel height) is the hydrodynamic radius, ρ is the solution density, u is the crossflow velocity and μ is the solution dynamic viscosity. In all experiments, the cross-flow rate is 1.249 L/min and ks from Eq. (8) is 7.95x10-5 m/s. Using this value for ks in combination with Eq. (7), one can calculate the CP modulus and the membrane surface solute concentration (cm), which is then used to determine the intrinsic rejection, rs, and B value.

2.4.3. Composite membrane permeability

As first pointed out by Lonsdale et al. [11], when the porous support skin-layer pores are considered, water and solute permeation through a thin film coated composite membrane will be lower than through the isolated thin film since part of the membrane film is effectively blocked by the solid portions (between pores) of the support membrane. Lonsdale et al. used computational fluid dynamics (CFD) simulations to illustrate an “effective thickness,” δeff, accounting for the increased path length for diffusion where water and solutes must move laterally through the film until they find an underlying pore. Later work by Ramon et al. [12,13] and Wong et al. [20] extended this concept using CFD simulations to explore the role of coating film thickness, support membrane pore size and porosity as well as coating film roughness and the presence of liquid filled voids within PA films on composite membrane transport. For simplified composite membranes with uniform coating film thickness, δeff can be approximated from:

This simple expression shows the dependence of δm,eff on the coating film thickness (δm), porosity (ε) and pore-size (rp) of the support mem brane. The reduced permeability, Preal/Pideal, is determined from the ratio of δm/δm,eff.

2.5. Physical-chemical membrane properties

2.5.1. Scanning electron microscopy

Membrane surface pore morphology was determined by scanning electron microscopy (SEM) (Zeiss Supra 40 VP, Carl Zeiss Microscopy, LLC, NY). Membrane samples were secured on SEM stubs using double sided carbon tape and sputter-coated (Ion beam sputtering/etching system, South Bay Technology, San Clemente, CA) with platinum for 90s at 15mA prior to analysis. Scanning was performed at 10.0 kV.

2.5.2. Surface porosity determination by SEM-ImageJ analysis

SEM images were analyzed by open-source ImageJ software (v1.8.0) to evaluate the surface porosity of membranes. For ImageJ analysis, the SEM images were converted to 8-bit mode to present. Brightness and contrast adjustment was used to remove the noise in the background of the images. Automatic thresholding was then used to convert images to binary (black and white), in which the black areas represent the pores and white areas represent the membrane surface as previously described [21].

2.5.3. MPD solution uptake

To understand the role of support membrane pore morphology on the MPD-TMC interfacial polymerization reaction, we developed a method to determine the amount of MPD taken up by support membranes. The amount of MPD in the soaked support was determined by measuring the volume of aqueous amine solution taken up and multiplying it by the MPD concentration in the solution. The mass per unit area was determined by dividing the mass by the area of the membrane coupon. Uptake experiments were performed in the stirred-cell described earlier without any applied pressure. A 13.80 cm² coupon was weighed and then clamped in the cell. Then, 50 ml of the amine solution used to make the RO membranes was poured into the cell and exposed only to the membrane surface. After 2 min, the solution was drained off and the excess amine solution was removed using an air knife as described above. The MPD solution-soaked coupon was weighed and the mass of solution taken up was calculated by subtracting the dry weight from the wet weight. The volume of MPD solution was determined using a density of 1003 kg/m3 based on the mass of water and each of the constituents in the amine solution. The volume of amine solution taken up divided by the volume of PSU (length x width x height) is assumed to represent the body porosity of the support membrane.

2.5.4. PA film mass

Free films of PA were obtained as follows. The nonwoven fabric was carefully peeled off so that only two layers, PSU +PA, remained. The two-layered membrane was pressed gently onto a small piece of polished Si wafer in such a way that the PA side stuck to the substrate. If the sample did not stick due to curling of the membrane, a drop of 2-propanol, which slightly softened the support, was placed on the membrane and then allowed to dry. The porous PSU support was then removed as follows. A few drops of DMF were placed on the membrane. After a few minutes, the PSU/DMF solution was carefully removed by slightly tilting the substrate and wiping off the solvent with tissue paper without touching the membrane. This procedure was repeated a few times to get rid of most of the PSU. After dissolution of the PSU and evaporation of the solvent, the free PA film became tightly fixed to the solid surface by adhesion. For complete removal of the PSU, the substrate with the attached membrane was thoroughly rinsed in clean DMF and dichloromethane. Afterwards, the film was weighed.

2.5.5. Elemental analysis and coating film XD

As depicted in Fig. 1, the cross-linking degree (XD) indicates the proportion of fully cross-linked structure in the PA layer, where each acid chloride group (Cl–C––O) of TMC reacts with the –NH2 group of an amine monomer forming an amide structure; unreacted acid chloride groups eventually become hydrolyzed to form a carboxylic acid. To calculate the XD, we used the following,

where m and n are the cross-linked and linear portions, respectively. Elemental N/O ratios can be determined via XPS whereby a coating film that is 100 % X-linked will have N/O =1.0, while a completely linear film will have N/O =0.5. The N/O ratio is obtained from X-ray photo electronic spectroscopy (XPS) analysis. The experiments were performed in a Kratos Axis ULTRA system from Kratos Analytical, using monochromatic Al-Ka (1486.6 eV) radiation and a charge neutralizing system. The spectra of dried samples were recorded using a 500x800 μm beam size, operating at 25 W and 15 KV. Survey spectra were obtained using a pass energy of 160 eV and a 0.2 eV step size for 3 sweeps.

2.5.6. MPD and TMC monomer conversion

The molar concentration of the PA film (mol/m²) was determined from a weighted average of crosslinked and linear film structures ac cording to:

Here, MPA is the mass of the PA film (g/m²), Mw,MPD is the molecular weight of MPD in the film, Mw,TMC is the molecular weight of TMC in the film, m is the fraction of the PA film that is crosslinked, and n is the fraction of PA film that is linear. The molecular weights of MPD and TMC in solution and in the film are given below in Table 1.

The molecular weights of MPD and TMC in the PA film are reduced by the amount of H⁺atoms that MPD loses when it reacts with TMC and the amount of Cl⁻ atoms lost when TMC reacts with MPD, respectively, plus the amount of –OH moieties gained by unreacted acid chlorides when they react with water. The mole fractions of MPD and TMC were determined from

and

where RMPD/TMC is the ratio of MPD to TMC in the PA films as determined by XPS from

Here, RN/O is the nitrogen to oxygen ratio determined from Eq. (10). The masses of MPD and TMC in the bulk solutions that ended up in the PA film were determined by multiplying their respective molar concentrations by their respective molecular weights in solution. Finally, the conversion of MPD was calculated from the ratio of MPD mass taken up by the PSU porous support to MPD mass in the PA film (both in g/m²); the conversion of TMC was calculated from the mass in the PA film (g/ m²) to the mass in solution divided by the membrane area, the latter of which was 45.9 g/m² for all membranes in this study.

3. Results and discussion

3.1. Support membrane pore morphology and performance

The SEM images in Fig. 4 illustrate the effects of PSU concentration and coagulation bath temperature on the surface pore morphology of the phase inverted PSU membranes. Visually, the pore size increases as the temperature increases for each PSU concentration. At higher temperature, the non-solvent molecules diffuse faster enhancing the rate of solvent and non-solvent exchange, which appears to increase pore size and porosity. As the PSU concentration increases, the pore size and porosity visually decrease at a given coagulation bath temperature.

Table 2 provides pure water permeability and 100 kDa PEG rejection by the uncoated PSU membranes, along with their body porosity, surface porosity and surface pore diameter of PSU membranes made at different PSU concentrations and coagulation bath temperatures. First, we see the viscosity increase with PSU concentration as expected. Next, the PSU film thickness and body porosity derived from analysis of cross-sectional SEM images (Fig. S2, Table S2) both increase with precipitation bath temperature while porosity decreases, but thickness increases with increasing PSU concentration. Further, the average pore size and porosity both decrease with increasing PSU concentration. In contrast, pore size and porosity increase with coagulation bath temperature for a given PSU concentration. A linear correlation coefficient analysis reveals that for a given PSU concentration, the support membrane skin-layer pore size and porosity correlate 95–100 % with coagulation bath temperature. A similar correlation between membrane pore size and water flux was reported by Gekas et al. [21]. In contrast, the rejection decreases with coagulation bath temperature and increases with PSU concentration. Kamal et al. [22] found similar results for phase inverted mesoporous membranes.

3.2. Composite RO membrane separation performance

In Table 3, we provide the MPD mass uptake, PA film mass, and XD along with the seawater RO separation performance of PA composite membranes formed over the PSU supports depicted in Fig. 4 with properties described in Table 2. Moreover, the PA layer thickness determined from the SEM images in Fig. 5 with thickness values Table 2. Generally, PA film mass, thickness and XD increase with PSU concentration and precipitation bath temperature. We include experimentally determined A, B and Rs values for three commercial SWRO membranes to illustrate that handcast membranes exhibit separation performance on par with the commercial standards. Composite PA membrane A and B values generally decrease with increasing PSU concentration, which produces PSU supports with decreasing pore sizes and porosities. For a given PSU concentration, as the coagulation bath temperature increases, NaCl rejection increases while A and B values both decrease. The underlying PSU support membranes all have increasing body porosity and MPD mass uptake as well as increasing surface porosity and pore size. All of the above results agree with predictions by the model of Ramon et al. [23]. Raw XPS spectra (Fig. S3) and elemental percentages (Table S3) are provided in the supplemental materials. The PSU body porosity governs the amount of MPD solution taken up and the amount of MPD available to partition into the reaction zone, and hence, both the PA film mass and XD. More MPD in the supports generally produces higher PA film mass (i.e., thickness) and XD, and hence, higher solute rejection and lower water permeability.

3.3. Mechanisms governing composite membrane XD, thickness and performance

A cross-correlation matrix provided in Table S1 reveals key insights about the mechanisms governing the interfacial polymerization reaction between MPD and TMC. First, PA-PSU composite membrane A and B values correlate most strongly and inversely with PA film mass and XD;

Fig. 4. Effect of coagulation bath T and different casting solution concentration. a-16, b-17 and c-18 wt% PSU. The numbers 1, 2 and 3 stand for coagulation bath T =23, 35 and 45 ◦C respectively. The coagulation bath contained 2 wt% DMF in all cases.

This is no surprise, and it strengthens our confidence in the experimental results presented above. Second, there are weakly negative correlations between A and B values and PSU support membrane surface porosity and pore size ( 41 % to 9%). This is not expected given our under standing of composite membrane transport, a la, Lonsdale et al. [11] and Ramon et al. [12,13,20]; it will be discussed further in Section 3.4. Next, MPD uptake is moderately correlated with PSU support membrane body porosity (69.3 %), but nothing correlates strongly with MPD uptake except rejection, which may be a mathematical coincidence. Further, PA film mass is weakly correlated with MPD uptake by the support membrane, but is perfectly correlated with TMC conversion, while XD is strongly correlated with both MPD (85 %) and TMC (92 %) conversion. Finally, conversion of MPD and TMC is 2.7 ±0.9 % and 0.21 ±0.07 %, respectively. These very small conversions seem to explain (at least partially) the weak to moderate correlation between MPD uptake and composite RO membrane properties.

3.4. Role of support membrane surface pore size and porosity

Theoretically, it is expected that the porous support increases the effective path length for water and solutes to permeate through composite RO membranes. The reduced permeability (Preal/Pideal) of hand cast membranes, calculated from the experimentally determined pore radii, surface porosity and film thicknesses (estimated from the mass and

Fig. 5. Cross-section and surface SEM of the PA membranes. (a, b) 16 % PSU at 25°C, (c, d) 16 % PSU at 45°C, (e, f) 17 % PSU at 25°C, (g, h) 17 % PSU at 45°C, (i, j) 18 % PSU at 25°C, (k, l) 18 % PSU at 45°C.

 Fig. 5. (continued).

an assumed density of ~1.4 g/cm 3 ) are plotted in Fig. 6 (symbols). Reference model predictions of reduced permeability in Fig. 6 (lines) assume a coating film thickness of 200 nm along with the minimum, average and maximum pore radii of approximately 7, 12 and 17 nm provide a frame of reference over a wider range of porosities. The experimental and model estimates are in reasonable agreement with all of the membranes exhibiting reduced permeability values between about 70 and 90 % (6 of 9 are between 80 and 90 %), which suggests the coating films were sufficiently thin and dense that the support membrane surface pore morphology had a small impact on the overall observed permeability. This is reflected by the weakly negative correlations between PSU support layer pore size and porosity discussed in section 3.3. However, in Fig. 6 we see that as support pore size and porosity increase the reduced permeability does move closer to unity.

4. Conclusions

Herein, we establish the role of PSU support membrane pore

Fig. 6. Plot of normalized permeability, which is the reduced value due to the support membrane pore radius and porosity relative to an unsupported coating film of the same structure. The lines are model calculations using Eq. (9) and assuming a typical SWRO PA film thickness of 200 nm.

morphology – whole body porosity, skin-layer pore size and porosity – on the thickness, XD and separation performance of hand-cast PA-PSU composite SWRO membranes. As PSU concentration increases, PSU support membrane pore size, surface porosity, bulk porosity, water permeability, MPD mass uptake, and PA-PSU composite membrane water and salt permeability generally decrease, while MPD and TMC conversion, PA film mass, XD, and PA-PSU composite membrane NaCl rejection increase. As PSU support membrane coagulation bath temperature increases, support membrane pore size, surface porosity, bulk porosity, MPD uptake, PA film mass, XD, MPD and TMC conversion, and NaCl rejection increase, while composite PA-PSU membrane water and NaCl permeability decrease. It is clear that support membrane body porosity dictates the uptake of the amine monomer solution, which governs the MPD concentration and MPD:TMC ratio in the reaction zone, but not directly the final PA film mass or crosslinking degree because the monomer conversions are quite low (on average 2.7 % for MPD and 0.21 % for TMC). Without varying monomer chemistry, monomer concentrations, interfacial polymerization conditions, or post- treatments, we demonstrate the critical role the PSU support membrane pore morphology – both skin-layer and whole body – plays in PA composite SWRO membrane performance.

CRediT authorship contribution statement

Derrick S. Dlamini: Writing – review & editing, Writing – original draft, Methodology, Investigation, Data curation. Javier A. Quezada- Renteria: Writing – review & editing, Investigation, Data curation. Jishan Wu: Writing – review & editing, Investigation, Data curation. Minhao Xiao: Writing – review & editing, Investigation, Data curation. Mackenzie Anderson: Writing – review & editing, Investigation, Data curation. Richard B. Kaner: Writing – review & editing, Supervision. Arian Edalat: Writing – review & editing, Project administration, Funding acquisition, Conceptualization. Nikolay Voutchkov: Writing – review & editing, Project administration, Funding acquisition. Ahmed Al-Ahmoudi: Writing – review & editing, Funding acquisition, Conceptualization. Eric M.V. Hoek: Writing – review & editing, Visualization, Supervision, Resources, Project administration, Methodology, Funding acquisition, Formal analysis, Conceptualization.

Declaration of competing interest

The authors declare they have no competing interests.

Data availability

Data will be made available on request.

Acknowledgements

The authors, in particular E.M.V.H., A.E., J.W. and D.D., are grateful for many valuable technical discussions with Dr. Hsiao Hachisuka at Memstar USA on various aspects of the “art of making RO membranes.” This material is based upon work supported by the National Alliance for Water Innovation (NAWI), funded by the U.S. Department of Energy, Advanced Manufacturing Office under Funding Opportunity Announcement DE-FOA-0001905. J.W. is grateful for the fellowships awarded by the National Water Research Institute (NWRI), the Southern California Salinity Coalition (SCSC), the American Membrane Technology Association (AMTA), and the North American Membrane Society (NAMS). The authors would like to acknowledge financial support for D. D. from SLAGC, support for A.E., J.W. and E.M.V.H. from SWCC, support for J.A.R.Q. from the UC Presidential Postdoctoral Fellowship program, support from the Lubrizol Corporation for M.X., and additional financial support from the UCLA Samueli School of Engineering.

References

[1] E.M.V. Hoek, D. Jassby, R.B. Kaner, J. Wu, J. Wang, Y. Liu, U. Rao, Sustainable desalination and water reuse, Synth. Lect. Sustain. Dev. 2 (2021) 1–204. Morgan & Claypool Publishers.

[2] J. He, J.R. McCutcheon, Y. Li, Effect of different manufacturing methods on polyamide reverse-osmosis membranes for desalination: insights from molecular dynamics simulations, Desalination 547 (2023) 116204.

[3] J. He, T. Arbaugh, D. Nguyen, W. Xian, E.M.V. Hoek, J.R. McCutcheon, Y. Li, Molecular mechanisms of thickness-dependent water desalination in polyamide reverse-osmosis membranes, J. Membr. Sci. 674 (2023) 121498.

[4] A.K. Ghosh, E.M. Hoek, Impacts of support membrane structure and chemistry on polyamide–polysulfone interfacial composite membranes, J. Membr. Sci. 336 (2009) 140–148.

[5] X. Lu, M. Elimelech, Fabrication of desalination membranes by interfacial polymerization: history, current efforts, and future directions, Chem. Soc. Rev. 50 (2021) 6290–6307.

[6] R.J. Petersen, Composite reverse osmosis and nanofiltration membranes, J. Membr. Sci. 83 (1993) 81–150.

[7] A.K. Ghosh, B.-H. Jeong, X. Huang, E.M.V. Hoek, Impacts of reaction and curing conditions on polyamide reverse osmosis membrane properties, J. Membr. Sci. 311 (2008) 34–45.

[8] S.H. Kim, S.-Y. Kwak, T. Suzuki, Pasitron annihilation spectroscopic evidence to demonstrate the flux-enhancement mechanism in morphology-controlled thin-film composite (TFC) membrane, Environ. Sci. Technol. 39 (2005) 1764–1770.

[9] A.E. Childress, M. Elimelech, Effect of solution chemistry on the surface charge of polymeric reverse osmosis and nanofiltration membranes, J. Membr. Sci. 119 (1996) 253–268.

[10] J. Ji, J.M. Dickson, R.F. Childs, B.E. McCarry, Mathematical model for the formation of thin-film composite membranes by interfacial polymerization: porous and dense films, Macromolecules 33 (2000) 624–633.

[11] H.K. Lonsdale, R.L. Riley, C.R. Lyons, D.P. Carosella Jr., Transport in composite reverse osmosis membranes, in: M. Bier (Ed.), Membrane Processes in Industry and Biomedicine, 1971.

[12] Guy Z. Ramon, Mavis C.Y. Wong, Eric M.V. Hoek, Transport through composite membrane, part 1: is there an optimal support membrane? J. Membr. Sci. 415–416 (2012) 298–305.

[13] Guy Z. Ramon, Eric M.V. Hoek, Transport through composite membranes, part 2: impacts of roughness on permeability and fouling, J. Membr. Sci. 425–426 (2013) 141–148.

[14] G.R. Guillen, Y. Pan, M. Li, E.M.V. Hoek, Preparation and characterization of membranes formed by nonsolvent induced phase separation: a review, Ind. Eng. Chem. Res. 50 (2011) 3798–3817.

[15] G.R. Gullen, G.Z. Ramon, H.P. Kavehpour, R.B. Kaner, E.M.V. Hoek, Direct microscopic observation of membrane formation by nonsolvent induced phase separation, J. Membr. Sci. 431 (2013) 212–220.

[16] A.K. Ghosh, E.M.V. Hoek, Impacts of membrane structure and chemistry on polyamide-polysulfone interfacial composite membranes, J. Membr. Sci. 336 (2009) 140–148.

[17] G.R. Guillen, Y. Pan, M. Li, E.M.V. Hoek, Preparation and characterization of membranes formed by nonsolvent induced phase separation: a review, Ind. Eng. Chem. Res. 50 (2011) 3798–3817.

[18] Y.J. Lim, J. Lee, T.-H. Bae, J. Torres, R. Wang, Feasibility and performance of a thin-film composite seawater reverse osmosis membrane fabricated on a highly porous microstructured support, Ind. Eng. Chem. Res. 611 (2020) 118407.

[19] G. Gullen, E.M.V. Hoek, Modelling the impacts of feed spacer geometry on reverse osmosis and nanofiltration membranes, Chem. Eng. J. 149 (2009) 221–231.

[20] Mavis C.Y. Wong, Lin Lin, Orlando Coronell, Eric M.V. Hoek, Guy Z. Ramon, Impact of liquid-filled voids within the active layer on transport through thin-film composite membranes, J. Membr. Sci. 500 (2016) 124–135.

[21] V. Gekas, K.M. Persson, M. Wahlgren, B. Sivik, Contact angles of ultrafiltration membranes and their possible correlation to membrane performance, J. Membr. Sci. 72 (1992) 293–302.

[22] N. Kamal, S. Ahzi, V. Kochkodan, Polysulfone/halloysite composite membranes with low fouling properties and enhanced compaction resistance, Appl. Clay Sci. 199 (2020) 105873.

[23] M. Xiao, F. Yang, Sungju Im, D.S. Dlamini, D. Jassby, S. Mahendra, R. Honda, E.M. V. Hoek, Characterizing surface porosity of porous membranes via contact angle measurements, J. of Membr. Sci. Letters 2 (2022) 100022.