1. Introduction
Reverse osmosis (RO) technology has remained the most dominant solution to water scarcity issues for the last few decades [
1]. The ease of manufacturing and maturity of the thin-film composite coating (TFC) process have enabled the widespread commercialization [
2] and adoption of RO technology with a comprehensive understanding of energy impact and desalination efficiency [
3]. However, the RO process involves energy-intensive operation due to the high-pressure requirement to exceed the osmotic pressure [
4]. Leon et al. analyzed the operational costs of seawater RO plants and reported that the energy consumed by high-pressure pumps forms about 62% of the total operational costs [
5]. It has also been established that improving RO system efficiency by increasing permeability could reduce specific energy consumption, thereby leading to lower operating costs. Several research advancements have been made to reduce specific energy consumption (SEC) in the RO process for brackish water, as well as seawater applications. Efforts to improve process efficiency and lower SEC have been expended in different areas, including membrane structure, properties, operational process variables, and spiral-wound module components such as feed spacers. Some examples of more recent development towards lowering SEC in RO systems include (a) aquaporin-based biomimetic membranes, reported by Li et al. [
6], which exhibited an 80% improvement in water flux compared to commercial RO membranes; (b) development of ultrahigh permeable membranes to improve module throughput in a given housing volume, as an approach to improve efficiency and reduce SEC [
7]; (c) a seawater RO system (SWRO) in a hybridized operation with pressure-retarded osmosis and nanofiltration processes, in which net energy consumption was reduced to 1.4 kWh/m
3 with a high recovery ratio of 88%, as reported by Touati et al. [
8]; and (c) theoretical and experimental evaluation of commercial feed spacers with different geometries and permeability coefficients, to optimize SEC for brackish water [
9] and seawater applications [
10], as reported by Ruiz-Garcia et al.
Even though several researchers have reported significant developments in improving water permeability, at a given applied pressure, as an approach to achieving lower SEC at the lab-scale coupon level, few developments have been reported on the scaling-up of these membranes to spiral-wound modules for system-level testing or field validation. Spiral-wound modules are the only configuration currently in use in all RO-based water treatment plants. The objective to reduce the gap between the thermodynamic minimum for desalination (~1.1 kWh/m
3) and the actual SEC in an RO process may never be realized, as it is constrained by engineering design limitations such as the efficiency of energy recovery devices or high-pressure pumps, also aggravated by process uncertainties such as fouling, scaling, and concentration polarization effects. Many such technical innovations involving high-permeable membranes, high-efficiency pumps, and energy recovery devices have been studied, as summarized by Voutchkov et al. [
11]. Alternatively, process efficiency can be improved by significantly improving the permeate flux and product water recovery ratios in spiral-wound modules (SWMs). Cohen-Tanugi et al. reported that a threefold increase in permeate flux could effectuate a ~15% reduction in SEC [
12]. Okamoto et al. reiterated a similar theory in their review, based on empirical evidence, that SEC can be lowered either by reducing hydraulic overpressure or by increasing permeate flux at a fixed applied pressure [
13]. Very little has been published on work involving permeate carrier-based modifications to spiral-wound modules for improving module-level productivity as an approach to lowering SEC per unit volume of permeate water produced. She et al. reported the fabrication of non-woven fabric-reinforced membranes for pressure-retarded osmosis applications with water permeabilities up to 20 L m
−2 h
−1 bar
−1 [
14]. In another paper, Sun et al. reported the fabrication of fabric-reinforced aliphatic polyketone TFC membranes for pressure-retarded osmosis applications with a water flux of 24 L m
−2 h
−1 [
15]. However, to the best of our knowledge, there are no reports of TFC PA membranes reinforced with permeate carrier materials for the optimization of permeate recovery to lower SEC in RO applications.
We report a novel method to optimize the membrane-spacer envelope in a spiral-wound module (SWM) to increase product water throughput at the module scale, to lower specific energy consumption (SEC) per unit volume of product water, in a desalination process. Towards this objective, we developed a novel process for the fabrication of a polysulfone (PSf) UF support layer directly on a permeate carrier mesh, instead of conventional non-woven fabric, followed by interfacial polymerization leading to a TFC PA layer, specifically for RO applications. Spiral-wound modules have been the most common configuration of packing flat-sheet membranes since the early development of polyamide (PA) membranes for water treatment applications. Additionally, the SWM configuration has long replaced the plate-and-frame configuration to become the only module configuration that is currently in use for flat-sheet RO membranes. Since the objective of this study is to demonstrate the feasibility of scaling-up novel reinforced membranes to the module level, we selected the standard spiral-wound module configuration for scale-up testing. A smaller leaf set thickness is critical for the fabrication of low-energy spiral-wound modules (SWMs) for many pressure-driven and thermal-driven membrane applications, such as reverse osmosis [
16], membrane distillation [
17], and pressure-retarded osmosis [
18].
Even though current desalination technologies have matured immensely towards significantly lowering energy consumption from >7 kWh/m
3 to ~3 kWh/m
3, large-scale cost-effective desalination plants are urgently needed in several water-scarce regions. To expand commercial accessibility to SWRO desalination plants, it is necessary that desalination costs are reduced even further to make them affordable for developing nations. One approach to make the technology more accessible and deployable is to lower energy consumption costs, which account for more than 60% of operational expenses (OPEX) in desalination plants. Therefore, we propose a novel method to improve the efficiency of SWM RO modules by packing more membrane area. This results in higher product water per module at the same applied pressure, which translates to lower specific energy consumption per unit volume of produced water. The membrane fabrication method described in the current study involved reinforcing the membrane with a mesh material, such as a permeate carrier, to enable higher water permeability through a thinner support layer matrix, leading to lower resistance to permeate flow, as well as lower potential for concentration polarization [
19]. This method offers another key advantage of replacing the typical non-woven substrate with a permeate carrier while retaining the permeate flow channels on the backside of the membrane in an SWM. This enables the membrane sheet to act as a separation barrier, as well as a permeate carrier, thereby significantly reducing leaf set thickness and facilitating more membrane area to be packed in a standard module. The packing of the larger membrane area in the same volume leads to higher product water output and, consequently, lower energy consumption per unit volume of produced water.
3. Results and Discussions
The conventional membrane-spacer envelope configuration in typical spiral-wound modules is shown in
Figure 5a.
The reinforced membrane was cast onto the permeate carrier using a standard industrial-scale, 1 m wide, phase-inversion coater according to the process shown in
Figure 1. In the current process, the phase-inversion method was slightly modified to pre-wet the permeate carrier, onto which the pre-mixed dope, with polysulfone (PSf) as the base polymer, was deposited using a slot die [
20]. The typical dope formulation and the casting conditions are given in
Table 1. The polymer cast substrate was passed through a coagulation bath maintained at room temperature at a line speed of 2.3 m/min. Polymer–solvent demixing and polymer phase-inversion resulted in UF support layer membrane formation on the substrate. During the preliminary trials in the current work, the reinforced UF layer was coated with a polyamide thin-film composite coating (TFC) using the formulation developed in our laboratory, to evaluate the integrity of the UF layer, as well as to validate the applications of these materials in different pressure-driven separations. Initial attempts at dope deposition and phase inversion directly on the permeate carrier with d (mesh size) > 300 μm produced membranes with pinhole defects and a weak polymer layer, which failed under high-pressure RO testing after interfacial polymerization (IP). Direct casting was successful on the permeate carrier with a smaller mesh size of d < 300 μm, but this method was not repeatable during scale-up to an industrial scale of 1 m width and more than 50 m length [
22]. Additionally, permeate carriers with a smaller mesh size were either significantly more expensive or not readily commercially available. Another challenge encountered was to inhibit the polymer dope from seeping through the permeate carrier mesh and upon phase inversion, blocking the product water flow towards the permeate tube. In order to overcome this limitation, it was essential to pre-wet the mesh with a suitable wetting agent prior to casting the polymer and coagulation. The pre-wetting process involved the deposition of wetting agents such as the DI water, surfactants, and solvents on the permeate carrier as a primer before the PSf polymer coating, either on a hand frame or on the industrial fabrication line. This ensured the partial pre-coagulation of the PSf polymer instantly upon touching the wetted mesh, uniformly anchoring the polymer to the mesh surface, while the rest of the polymer phase separated inside the coagulation bath, resulting in defect-free membranes devoid of pinholes while also preserving the permeate channels for product water flow during RO operations. In the lab-scale hand-frame casting, in addition to ensuring the thin coagulated layer formed instantly after casting onto the wetted mesh, the phase-inversion process was carried out with the polymer surface facing down as the weight of the casted polymer pulled the polymer away from the mesh and prevented the dope from seeping through the holes. The coagulating membrane facing upwards was not as robust, as the weight of the polymer pulled the polymer through the holes and had a high probability of creating a pinhole defect or blocked permeate channels.
Figure 2 shows the optical microscope images of all the permeate carriers that were used in this study, along with the mesh sizes. Permeate carriers with a variety of mesh hole sizes ranging from 100 microns to 400 microns were employed to demonstrate the versatility of the fabrication method. This lab-scale coating method with pre-wetting was scaled-up and translated successfully onto the industrial-scale coating line for fabricating the reinforced membrane. On the industrial-scale coating line of 1 m width and length > 50 m, the PSf polymer was cast using a slot die positioned perpendicular to the substrate (
Figure 1). Due to the higher porosity of the mesh as compared to a non-woven polyester backing, the casting process was challenging due to significant shrinkages and creasing along the cast membranes, resulting from in-line stress due to continuous operation of the line. The process was modified to adjust the line speed and, consequently, the line tensions so as to mitigate the creasing and the in-line bending of the web. This ensured that the process was stabilized and optimized over several rounds to achieve a good substrate with no pinhole defects across the 1 m width.
Figure 6 shows the front and back view of the conventional flat-sheet RO membrane (
Figure 6 a,b) and the newly fabricated low-energy membrane (
Figure 6 c,d) with a polyamide selective layer. It can be easily seen that, unlike the regular flat-sheet membrane, the newly fabricated membrane had clearly defined water channels on the back, as also confirmed by the FESEM images of the membrane cross-section (
Figure 6e), as well as the backside of the membrane (
Figure 6f). Therefore, the reinforced membranes have the potential to replace the typical non-woven substrate with the permeate carrier and enable single-membrane sheet element rolling with feed spacer, which then could significantly reduce the cost and labor for element rolling. The membrane when rolled into a standard module would require lower energy for the same output as more membrane could be rolled within the standard housing.
The reinforced PSf support layer was coated with the polyamide TFC coating through IP on an industrial coater to make the dense selective polyamide (PA) layer.
Figure 7 depicts the FESEM images of the reinforced membrane after the TFC PA layer coating. A continuous nanoporous “ridge-and-valley” morphology, which is typical of the TFC PA layer [
23], is clearly seen on the surface of the membrane, as shown in
Figure 7a–c, which confirms the formation of the TFC selective layer. The cross-sectional images (
Figure 7d–f) showcase the presence of the permeate carrier material impregnated in the polymeric matrix (
Figure 7d), while the UF support layer filled with finger-like macro void morphologies is shown to be preserved. The membrane was then wound into spiral-wound modules (
Figure 3), with the feed spacer facing the active selective layer side and the permeate carrier facing the backside of the membrane. The method here originated from the key objectives of fabricating a pressure-driven reinforced membrane for pressure-driven application and with a thin leaf set with permeate carrier and membrane combined in a single sheet. This thin leaf set was critical for the fabrication of low-energy spiral-wound modules (SWMs) for all pressure-driven membrane applications and enabled two-layered membrane sheet element rolling as compared to the typical three-layer rolling (
Figure 5).
TJ-30, with the largest average mesh hole size of 400 microns, was selected to demonstrate the effect of pre-wetting, as well as the pre-wetting solution, on membrane performance (
Table 2). The pre-wetting of the permeate carrier was conducted prior to the casting of polymer solution to prepare the substrate membrane. Subsequently, a thin-film composite (TFC) polyamide selective layer formed on the resultant substrate membrane via interfacial polymerization. It can be seen from
Table 1 (Entries 5 and 6) that without any pre-wetting with water or other agents, the substrate membrane was highly defective, and it was highly difficult to form a defect-free polyamide selective layer. This was presumably because, without pre-wetting, the PSf polymer solution intruded into the mesh holes upon deposition, leading to non-uniform surface coverage on the mesh, which resulted in a defective PA layer, as seen in
Table 1, likely because the pre-wetting method helps to fill up the mesh holes of the permeate carrier prior to coating. The pre-wetting process temporarily sealed the mesh pores with pre-coagulation of the PSf polymer and eliminated the problem of solution intrusion. As a result, the flux and rejection of the TFC membrane could be further improved (
Table 2).
Next, the effect of the type of pre-wetting solution on the membrane performance was explored. Generally, the permeate carriers are made of polyester, which tends to be hydrophobic. Thus, for the hydrophilic pre-wetting solution of water or aqueous agents to fill the mesh holes more easily and completely, the weight percentages of sodium laureth sulfate (SLS) or N-methyl-2-pyrrolidone (NMP) in the pre-wetting solution were increased. Interestingly, it was found that increasing the SLS content to 0.2 wt.% or the NMP content to 20 wt.% had a negligible impact on the performance of the membrane. When further increasing the SLS content to 2.5 wt.% or the NMP content to 80 wt.%, the membrane performance deteriorated drastically, leading to a significant loss of salt rejection. This was likely because the highly hydrophilic pre-wetting solution may have interfered with the phase-inversion process of the polymer dope and prevented the formation of a defect-free substrate.
In general, as the mesh size increases, the MWCO is also expected to increase as the PSf solution intrusion into the mesh hinders the formation of a homogenous UF support layer during the phase-inversion process. However, in our study, no clear trend was observed for the relationship between the MWCO data and the mesh size (
Table 3). In comparison to the control membrane (non-woven), TF800, with the smallest mesh size, gave the smallest MWCO of 37 kDa, presumably due to lower solution intrusion and the formation of a thicker substrate layer. On the other hand, the MWCO increased to 96 kDa for the P16 material with a mesh size of 300 μm, which reduced to 54 kDa for the TJ030 material. Given the smaller scale of experiments, several factors, including mass transfer through the mesh during phase inversion, as well as other extraneous factors such as variations in dope viscosity through solvent evaporation or moisture, may have contributed to this observation.
Table 3 demonstrates the effect of the average mesh hole size on the membrane performance. The water fluxes for the UF support layer membranes prepared using conventional non-woven fabric, TJ-030, P16, and TF800 were measured as 130 ± 18 L m
−2 h
−1, 136 ± 36 L m
−2 h
−1, 115 ± 36 L m
−2 h
−1, and 186 ± 44 L m
−2 h
−1, respectively. It can be observed that as the average mesh hole size increases from 100 microns to 400 microns, the average flux of the resultant TFC membrane decreases from 48.65 L m
−2 h
−1 to 21.95 L m
−2 h
−1, while the NaCl rejection remains relatively constant at 95%, for all the pre-wetted substrates. This was likely because as the average mesh hole size increased, the amount of solution intrusion into the mesh hole increased as well. Hence, a greater amount of polymer needed to be deposited to achieve a defect-free substrate, and as a result, the permeate transport resistance of the resultant membrane increases. In addition, it was also observed that there was an optimum mesh hole size of about 300 microns (P16), which enabled the greatest reduction in thickness compared with the non-woven membrane backing while having comparable membrane performance. Thus, Permeate P16, with a ~9% reduction in thickness, shall be used for further scale-up on the industrial fabrication line.
Table 4 demonstrates the successful translation of the fabrication method onto the industrial line, which was 1 m in width with ~7.4% reduction in thickness compared to the conventional membrane. The performance of the line run membrane with a flux of 51.1 L m
−2 h
−1 and 2000 ppm NaCl rejection of 95.9% is comparable to that of the hand-frame membrane, which has a flux of 42.1 L m
−2 h
−1 and 2000 ppm NaCl rejection of 94.8% Subsequently, the 1812 (1.8-inch diameter and 12-inch length) and 2514 (2.5-inch diameter and 14-inch length) elements were fabricated from the above-mentioned membrane. The 1812 element has a flux of 0.032 L per min (LPM) and a 500 ppm NaCl rejection of 79.45% under the operating pressure of 3.4 bar, while the 2514 element has a flux of 0.325 LPM and a 2000 ppm NaCl rejection of 87.25% under the operating pressure of 15 bar. Successful industrial fabrication and demonstration of next-generation low-energy membranes have been achieved.