Ultra-strong polymeric hollow fiber membranes for saline dewatering and desalination

Characteristics of PES hollow fiber membrane substrates

For a TFC hollow fiber membrane, the mechanical strength is dominated or controlled by the hollow fiber membrane substrate because the mechanical strength of the ultrathin selective layer of <1 µm is low. Therefore, the membrane substrates are firstly fabricated and optimized by tuning the PES polymer concentration, their wall thickness, and morphologies. Table 1 summarizes their physical and mechanical properties. As the ratio of the dope to bore fluid flow rate (hereafter, designated as the dope/bore fluid ratio) increases from 1 to 10, the outer diameters (OD) of hollow fibers are maintained at about 1000 µm, while the inner diameters (ID) decrease from about 640 to 300 µm. As a result, the wall thicknesses of the hollow fibers increase from ~200 to 360 µm. Consistent with the increases in PES concentration and wall thickness, the burst pressure of the hollow fiber substrates also increases from 17 to 83 bar. The burst pressure is the maximum pressure that a hollow fiber membrane can withstand under an in-to-out testing mode. Once the burst pressure is reached, the gauge pressure of the testing device suddenly decreases as a result of membrane failure.

Overall, the bulk porosity of all substrates varies from about 65 to 75%. Generally, the bulk porosity declines with an increase in PES polymer concentration. Interestingly, even though the dope/bore fluid ratio increases from 1 to 10, the resultant substrates spun from the same dope solution have almost similar bulk porosities. Besides, there is a down-and-up trend for the bulk porosities as a function of the dope/bore fluid ratio. The bulk porosity decreases slightly as the ratio increases from 1 to 4 but increases slightly when the ratio reaches 10. This down-and-up trend possibly arises from the combination of 3 factors. (1) Since the dope viscosity increases exponentially with the PES concentration as shown in Fig. 1, it significantly alters the phase inversion process, die swelling phenomenon, and non-solvent intrusion mechanism^{19,38,39,40,41,42,43,44}. (2) Because an increase in dope/bore fluid ratio results in a thicker fiber wall, this would cause the delay of solvent exchange between the solvent (NMP) and non-solvent (water) that may lead to a less porous structure or a lower bulk porosity^19,39,40. (3) As the dope/bore fluid ratio increases to 10, the flow rate of bore fluid becomes very small (0.2 ml/min). This leads to a rapid change of bore fluid composition from pure water to a solvent (NMP) enriched mixture because of the solvent exchange between the bore fluid and the dope solution. The solvent enriched bore fluid would lead to a different phase inversion path and result in a more porous inner structure and thus a slightly higher bulk porosity.

The above rationales are evidenced by the FESEM morphologies of the hollow fiber substrates. Supplementary Figure 1 shows their overall cross-section morphologies, while Fig. 2 displays their enlarged morphologies as a function of PES concentration in spinning dopes and dope/bore fluid ratio during spinning. Generally, the finger-like macrovoids in the cross-section decrease whilst the sponge-like microstructure increases with an increase in dope/bore fluid ratio and PES concentration. A macrovoid structure tends to appear at the inner region of the hollow fiber substrate possibly due to the non-solvent intrusion from the lumen side and the effect of die swelling^38,39,42,43 while the sponge-like structure is formed at the outer region owing to the delayed demixing. The die swell phenomenon is a relaxation phenomenon when the shear-oriented polymer chains return to their random coil state after exiting from the annular channel of a spinneret. The relaxation rapidly expands the polymer chains moving inward to the lumen side as well as outward as referred to the die swell^42,43.

Figure 3 compares the FESEM morphologies of PES hollow fiber substrates spun from different PES concentrations but with the same dope/bore fluid ratio of 10. Interestingly, the hollow fiber (P22-D) spun from a dope containing a low PES concentration of 22 wt% has a dual-layer structure of finger-like macrovoids, while the others (P26-D and P30-D) possess only one layer of finger-like macrovoids. This phenomenon may result from the combined effects of the aforementioned factors. In other words, the nascent hollow fiber substrate spun from a dope with a low viscosity may experience non-solvent intrusion from both the internal and external coagulants, thus it has a dual-layer structure of finger-like macrovoids³⁹. Once the dope viscosity is increased with the addition of PES, the nascent hollow fiber substrate can resist the non-solvent intrusion from the external coagulant because it has been relaxed from die swell in the air gap region and oriented by gravity and spinning-line stresses^38,39. On the other hand, the inner region still has finger-like macrovoids because (1) the bore fluid and the dope meet immediately after extrusion from the spinneret and (2) the effect of inward die swelling may cause the non-solvent intrusion from the bore fluid.

In terms of mechanical properties and morphologies, Table 1 shows that the hollow fiber substrates with a dual-layer structure of finger-like macrovoids tend to have weaker mechanical properties (e.g., lower Young’s modulus and the maximal tensile stress) than the ones with a single layer of macrovoids. This is because finger-like macrovoids are the weak points of hollow fibers^19,38. However, the dual macrovoid structure is favorable in reducing the internal concentration polarization (ICP), as discussed later.

Supplementary Figure 2 shows the inner and outer surface morphologies of the substrates as a function of PES concentration spun at the dope/bore fluid ratio of 10. Both the inner and outer surfaces have relatively dense skin because water is used as the bore fluid and external coagulant. Figure 4 describes the pure water permeance (PWP) of these hollow fiber substrates as a function of the dope to bore fluid ratio. PWP varies in the range of approximately 230 to 30 litre/(m² h bar) or LMH/bar, and decreases with the rise of the dope/bore fluid ratio. In addition, the substrate spun from a higher PES concentration has a lower PWP. The decline of PWP is because a higher dope/bore fluid ratio leads to a thicker membrane wall and a higher PES concentration results in a denser membrane structure. Both effects increase the mass transport resistance and lower the water permeance across the membranes.

Characteristics of PES-TFC hollow fiber membranes

By performing the interfacial polymerization at the lumen side of the hollow fiber substrates, inner selective TFC hollow fiber membranes were fabricated. Table 2 depicts their characteristics including pure water permeance (A), NaCl rejection (R), salt permeability (B) and burst pressures. Since these PES-TFC membranes have different mechanical properties, they were tested under different conditions. For PWP and salt rejection tests, the PT22-A TFC membrane was conditioned and tested at 10 bar; PT26-A and PT30-A were conditioned and tested at 20 bar; all other TFC membranes were conditioned at 30 bar and tested at 20 bar. Most PWP values of TFC membranes are about 3 LMH/bar while their salt rejections are around 98%. Interestingly, although PWP of the hollow fiber substrates declines with an increase in PES concentration and dope/bore fluid ratio (Fig. 4), the PWP values of TFC membranes do not show a similar trend. This arises from the fact that the major transport resistance across the TFC membranes comes from the selective polyamide layer formed by the interfacial polymerization.

Polyamide layer

Figure 5 illustrates the FESEM images of the inner surfaces and inner edges of the PES-TFC membranes as a function of PES concentration and dope/bore fluid ratio. Generally, all polyamide selective layers possess a typical ridge-and-valley morphology due to the rapid interfacial polymerization reaction between MPD and TMC monomers^27,34,35,36. The apparent thickness of the polyamide layers ranges from about 300 to 400 nm. Interestingly, the TFC membrane spun from a higher dope/bore fluid ratio tends to have a thinner polyamide layer. The decline in the polyamide thickness might be ascribed to the combined effects of capillary pressure and convective flow. Since the MPD solution is firstly circulated in the lumen side and absorbed near the inner surface region for the subsequent interfacial polymerization, the capillary pressure induced by the pores near the region and finger-like macrovoids would pull (suck) the MPD solution into the bulk membrane. According to the Young–Laplace equation, the capillary pressure is reciprocally proportional to the radius of a capillary^45,46. A higher PES concentration would result in hollow fiber substrates with a smaller pore radius, while a higher dope/bore fluid ratio would produce substrates with a thicker fiber wall. The former would pull the MPD solution into the bulk substrate more effectively, whereas the latter would significantly retard the diffusion of MPD molecules back to the lumen surface to react with TMC. Thus, it leads to form a thinner TFC layer with an increase in PES concentration and dope/bore fluid ratio.

In addition, since the flow rate of the TMC solution is fixed during the IP process and the hollow fiber substrate spun from a higher dope/bore fluid ratio has a smaller inner diameter, this leads to a higher velocity of the TMC solution flowing through the smaller lumen (or smaller inner diameter) of the hollow fiber substrate. The fast-flowing TMC solution may lead to a shorter reaction time and constrains the growth of the polyamide layer on top of the inner surface. Besides, a fast TMC solution flowing through the lumen may result in interfacial polymerization not only on top of the inner surface but also within the pores near the inner-surface region. Thus, an apparent thinner polyamide layer is formed on top of the inner surface of the substrate.

Burst pressure

As shown in Table 2, the burst pressures of the PES-TFC membranes range from 22 to 110 bar. Consistent with the burst pressures of their PES substrates as tabulated in Table 1, the PES-TFC membranes spun from the same PES concentration possess a higher burst pressure if the dope/bore fluid ratio is increased (e.g., due to a thicker wall). For PES-TFC membranes with a similar wall thickness, the one fabricated from a higher PES concentration has a higher burst pressure. Interestingly, even though the polyamide layer typically has a thickness of less than 1 µm, a comparison of Tables 1 and 2 indicates that the deposit of a polyamide layer on top of PES substrates remarkably enhances their burst pressures. The PES-TFC hollow fiber membranes have burst pressures about 10–40% higher than their corresponding substrates.

The surprising increase in burst pressure may result from (1) the enhanced mechanical properties of the inner-layer region due to the interfacial polymerization and (2) different structure failure mechanisms for these two kinds of membranes during burst tests. As aforementioned, during the interfacial polymerization, capillary pressure and a fast-flowing TMC solution flowing stream occur simultaneously. These may render the interfacial polymerization not only on top of the inner surface but also within the pores near the inner-surface region, and thus strengthen the mechanical properties of the inner-layer region and improve the overall mechanical properties of PES-TFC hollow fiber membranes. As a result, the reinforced inner-layer region has a higher burst pressure because it can dissipate the burst stresses more uniformly and effectively than the original PES hollow fiber substrates. In addition, the inner bulk structure of the original PES substrate may deform partly from within due to the combined effects of (1) the weak macrovoid bulk structures and (2) the build-up of the hydraulic pressure within the bulk structure because its outer surface is denser than the inner surface (or the influx water from inner surface is larger than the outflux water from the outer surface). By contrast, the PES-TFC membranes might not have such deformation because the inner polyamide selective layer is much denser than the outer surface of the TFC membrane (or the influx water from the inner surface is much smaller than the outflux water from the outer surface).

Comparison of burst pressures and benchmark

Figure 6 compares the burst pressures obtained from experiments (Tables 1 and 2) and calculated by Barlow’s equation (Eq. 9). The calculated burst pressures are lower than those obtained experimentally for both PES-TFC membranes and PES substrates. The calculated burst pressures are comparable to the experimental results at low burst pressures (e.g., <50 bar). However, when the dope/bore fluid ratio increases, the asymmetry of the resultant membranes increases. It leads to a great deviation between the calculated and experimental burst pressures as high as 100% at a very high burst pressure (e.g., 110 bar). Since Barlow’s equation is typically applied to estimate the burst pressure of isotropic tubes²⁹, one may need to adjust the safety factor of the Barlow’s equation to make a more accurate estimation for the burst pressure of asymmetric hollow fiber membrane. Figure 6d shows the FESEM images of the representative burst PES-TFC hollow fiber (PT22-D). A narrow and long opening can be observed at the outer and inner surfaces of the burst (broken) fiber. The enlarged inner surface shows that the PES substrate is exposed and the polyamide selective layer is displaced due to the outward expansion of the fiber under a high pressure or burst pressure (e.g., ~73 bar).

A benchmark of burst pressures for various membranes is presented in Supplementary Table 3. To the best of our knowledge, the as-developed PES-TFC hollow fiber membranes are the strongest hollow fiber ever reported with the highest burst pressure of up to 110 bar. This value is very comparable to the strongest commercial flat sheet membranes with a burst pressure of 120 bar made of a special module design and spacer³. In particular, the newly designed TFC hollow fiber membranes are inner selective hollow fibers, which are favorable for scaling-up and large production.

Performances of RO and structural parameters

The optimal PES-TFC hollow fibers (PT22-D, PT26-D, and PT30-D) were selected to further evaluate their performances for saline dewatering via RO operation mode and investigate their structural parameters (S parameter) by the pressure retarded osmosis (PRO) mode. The results are presented in Fig. 7. As shown in Fig. 7a, b, under the RO mode using a 0.3 mol/L salt solution as the feed at 30 bar, the permeate water permeances of the three hollow fiber membranes are about 1.5 LMH/bar, which is ~50% of their corresponding pure water permeance (PWP) as presented in Table 2. Nevertheless, their salt rejections remain high and are around 97%, which is about 1% less than the rejections evaluated by the standard method using a feed solution of 2000 ppm or 0.034 mol/L (see Table 2). The decreases in permeate water permeance and salt rejection are due to the increase in salt concentration of the RO feed. When using a 0.3 mol/L salt solution as the RO feed, considerable external concentration polarization (ECP) and the internal concentration polarization (ICP) occur, thus reducing the effective driving force across the membrane. Therefore, the water permeances decrease significantly.

The structural parameter of a semi-permeable membrane represents characteristics of the thickness and tortuosity of the membrane support^27,47,48, and can be estimated by performing the PRO tests^27,48. As described in Fig. 7c, under the PRO mode using a 1.2 mol/L NaCl solution as the draw solution at the lumen side and DI water as the feed at the shell side, the water permeates to the lumen side while the salt reversely transports from the draw solution in the lumen side to the shell side. The structural parameters of PT22-D, PT26-D, and PT30-D membranes are about 650, 1000, and 3500 μm, respectively. The significant increase in structural parameter, particularly for the PT30-D membrane, is majorly ascribed to the denser PES substrates (P26-D and P30-D) as presented in Fig. 3. The denser substrate has more tortuous diffusion paths, thus leads to have a higher structural parameter. The increases in structural parameters of PT26-D and PT30-D membranes are consistent with the lower PWP values of their substrates (see Fig. 4).

Performances of OARO

The optimal PES-TFC hollow fiber membranes (PT22-D, PT26-D, and PT30-D) that possess the highest burst pressure at their specific PES concentrations were selected for OARO tests. All TFC membranes were conditioned and stabilized at 30 bar for at least 30 min prior to the OARO measurements. Although they can withstand a pressure of at least 70 bar, 30 bar is the maximum operating pressure used in this study because of the limitations of the pump and the experimental setup.

The OARO performances of the optimal PES-TFC membranes are shown in Fig. 8. Generally, as described in Fig. 8a, b, both water flux and water permeability decrease dramatically as the NaCl concentration increases. For example, the water flux of the PT22-D membrane declines from 57.2 to 8.6 LMH while the corresponding water permeance declines from 1.91 to 0.29 LMH/bar. The water flux of the PT26-D membrane declines from 48.8 to 2.8 LMH while its water permeance declines from 1.91 to 0.10 LMH/bar. The water flux of the PT30-D membrane decreases from 45.4 to 1.7 LMH while its water permeability decreases from 1.51 to 0.06 LMH/bar. Such decreases in water flux and permeance could be ascribed to two reasons: (1) the increase of osmotic pressure in the feed solution from 14.7 to 58.7 bar as the NaCl concentration rises from 0.3 to 1.2 mol/L; (2) the ICP and ECP reduce the effective driving force across the membrane. On the inner surface (or lumen side) of the TFC hollow fiber, the salts are rejected by the polyamide selective layer and accumulate in its vicinity. Thus, the salt concentration near the inner surface is higher than its bulk concentration in the lumen. Such a phenomenon is known as ECP^49,50. On the outer surface (or shell side), the water permeates through the selective layer, diluting the salt concentration in the cross-section of the membrane substrate. As a result, the salt concentration behind the polyamide selective layer is lower than its bulk concentration in the sweep solution. Such a phenomenon is known as ICP^50,51,52. Both ECP and ICP increase the osmotic pressure difference (Δπ) and thus reduce the effective driving force across the membrane under the OARO operation mode. In particular, ICP is believed to be the dominant factor that causes the decrease in water flux and permeance. This is because (1) all ECPs are possibly similar because the flow rates at the shell and lumen sides are kept the same for all cases and (2) ECP declines as the permeate water flux decreases.

As shown in Fig. 8a, b, at a given salt concentration, the water flux and water permeability of the three optimal membranes follow the order of PT22-D > PT26-D > PT30-D. This is because the structural parameters of the three membranes are in the order of PT22-D < PT26-D < PT30-D (see Fig. 7d). Since a larger structural parameter leads to a lower water flux or permeance for a membrane (e.g., evidenced by the PWP values of their substrates in Fig. 4), it may take a longer time for the water permeated from the lumen side to mix with the sweep solution diffused from the shell side. In other words, for the membrane with a larger structural parameter, it may suffer from a severer ICP effect. Therefore, the detrimental ICP effect of the optimal membranes is in the order of PT22-D < PT26-D < PT30-D. The TFC membrane with a smaller structural parameter or made from a weaker substrate with a higher PWP suffers from a less diluting or ICP effect, thereby having a higher effective driving, and thus a higher water flux and water permeance.

The effects of the operating pressure on water flux and water permeance are depicted in Fig. 8c, d. To mimic seawater, a NaCl solution of 0.6 mol/L was chosen to study the effects of operating pressure. The water flux increases from the lowest value of 4.4 LMH at 10 bar for PT30-D to the highest one of 24.9 LMH at 30 bar for PT22-D. The monotonic augment of water flux is due to the increase in operating pressure or driving force. Interestingly, as illustrated in Fig. 8d, the water permeance displays a V-shape trend. The decline-and-increase trend in water permeance may arise from the combined effects of ICP and membrane expansion at high pressure. As the operating pressure increases, the water flux increases, leading to a severer ICP diluting effect. As a result, the water permeability declines. On the other hand, the TFC hollow fiber is subject to a certain degree of expansion because a high pressure (e.g., >20 bar) is applied at the lumen side. It not only radially expands the elastic hollow fiber membrane but also thins the polyamide selective layer, leading to a higher water permeance while being able to maintain the salt rejection^53,54,55. Thus, the increase in water permeance overtakes the decrease induced by ICP at 30 bar.

In the OARO mode, the sweep stream can mitigate the osmotic pressure of the feed across the membrane. As a result, the osmotic pressure difference (Δπ) is reduced. Therefore, it is clear that water can be recovered from saline water even though the operating pressure is lower than the osmotic pressure of salt water by applying the OARO strategy. The water flux can be further increased by increasing the operating pressure because the as-developed TFC hollow fiber membranes can withstand a hydraulic pressure of much more than 30 bar. It is worth noticing that a higher applied hydraulic pressure (e.g., > 30 bar) is needed when the solution/feed is more concentrated or a higher water recovery is required. Although the PT30-D membrane has the highest burst pressure, its structure parameter is much higher than the PT22-D and PT26-D ones. In real applications, one may choose the PT26-D membrane for the OARO process because it has a decent burst pressure but a relatively low structure parameter.

In conclusion, a series of strong TFC hollow fiber membranes have been developed in this study for saline dewatering via reverse osmosis (RO) and osmotically assisted reverse osmosis (OARO) processes. By adopting the strategies of tuning the hollow fiber dimension and adjusting the host polymer (PES) concentration, an ultra-strong hollow fiber membrane with a burst pressure of about 110 bar is demonstrated. Besides the dope composition and membrane spinning conditions, dope rheology such as die swell and chain relaxation also play important roles in determining phase inversion mechanism, membrane morphology, and mechanical properties. The increase in mechanic strength of the hollow fibers does come with the price of the increase in structural parameter or internal concentration polarization (ICP). Nevertheless, this work may pave a new venue for developing ultra-strong polymeric TFC hollow fiber membranes for RO, OARO, and other applications.

Source: Resources - nature.com