Effect of Polyethylene Glycol Additive on the Structure and Performance of Fabric-Reinforced Thin Film Composite

Fabric-reinforced thin film composite (TFC) membranes exhibit outstanding mechanical durability over free-standing membranes for commercial applications. In this study, polyethylene glycol (PEG) was incorporated to modify the polysulfone (PSU) supported fabric-reinforced TFC membrane for forward osmosis (FO). The effects of PEG content and molecular weight on the structure, material property and FO performance of the membrane were investigated comprehensively, and the corresponding mechanisms were revealed. The membrane prepared by using 400 g/mol PEG exhibited better FO performances than those of membranes with 1000 and 2000 g/mol PEG, and 20 wt.% was demonstrated to be the optimal PEG content in the casting solution. The permselectivity of the membrane was further improved by reducing the PSU concentration. The optimal TFC-FO membrane had a water flux (Jw) of 25.0 LMH using deionized (DI) water feed and 1 M NaCl draw solution, and the specific reverse salt flux (Js/Jw) was as low as 0.12 g/L. The degree of internal concentration polarization (ICP) was significantly mitigated. The membrane behaved superior to the commercially available fabric-reinforced membranes. This work provides a simple and low-cost approach in the development TFC-FO membrane and shows great potential in the large-scale production for practical applications.


Introduction
FO has been widely regarded as a valuable alternative process for water treatment. It holds several advantages, including mild operating conditions, low requirements for pretreatment, high water recovery, efficient contaminant rejection, low membrane fouling propensity and good fouling reversibility, over the conventional thermally-driven and pressure-driven membrane separation technologies, especially in the applications for seawater desalination, high value-added product concentration and treatment of complex and impaired wastewater, high salinity and high fouling potential streams [1][2][3][4][5][6].
As the core of FO process, the lack of high-performance FO membrane is the major bottleneck restricting the development of FO technology [7]. An ideal FO membrane material should achieve high water permeability, high solutes selectivity and be mechanically robust. Current developed FO membranes mainly include integrally asymmetric, layer-by-layer assembled and TFC membranes [8][9][10]. Among them, TFC is the most attractive FO membrane, owing to its high permselectivity, acid-base resistance and biological stability [11]. The two-layer structured TFC membrane is typically composed of an active layer (i.e., separation layer) and a porous support layer [12,13]. The ultra-thin dense active layer that is mainly prepared by interfacial polymerization contributes the permselectivity of the TFC membrane, while the underneath porous layer acts as an essential mechanical support and provides a physical interface for the formation of the active layer. It has been reported that

Results and Discussion
The intrinsic property of the TFC membrane substrate is one of the key factors that affect its ICP and the formation and performance of the active layer. In this paper, a series of woven fabric-reinforced PSU/PEG membrane substrates were prepared according to the casting solution composition in Table 1, and their physical and chemical properties including porosity, mean surface pore size, surface contact angle, thickness and surface chemical composition, were characterized as shown in Table 2 and Figure 1. The substrate consisted of a porous layer and a fabric layer. Our previous research showed that in the preparation of the additive-free woven fabric-reinforced PSU substrates, high-quality and low-defect coatings on the woven fabrics were achieved when the PSU concentration in the casting solution was 15 wt.% [26]. As a non-solvent additive, the incorporation of PEG increases the viscosity of the casting solution when the PSU concentration is constant [30]. In this study, the casting solution viscosity increased with the increase in PEG content, and the resultant substrates showed good quality when the PEG concentration was up to 20 wt.%. However, excess PEG incorporation impaired the quality of the substrate due to the increased high viscosity that was not conducive to the mixing and spreading of the casting solution. Furthermore, casting solution with PSU concentration of 12 wt.% and 9 wt.% was prepared when PEG content maintained at 20 wt.%. The substrates still exhibited good quality as the viscosity of the casting solution was appropriate. However, the quality significantly decreased when the PSU concentration was below 9 wt.%, where the solution viscosity was too low for the substrate formation. result in a denser surface structure and increase the thickness of the skin layer and th spongy-like microporous structure. Thus, the PSU content at the substrate surface w increased. Furthermore, under a certain PEG4 content at 20 wt.%, the peak intensity d creased with the decrease in the PSU concentration (PEG4-4, PEG4-5, PEG4-6), which ma be relative to the reduced sulfone content and thickness of the surface layer. The surface, cross-section and bottom morphology of the prepared PSU/PEG su strates are shown in Figures 2-4, respectively. The surface morphology ( Figure 2) of th prepared PSU and PSU/PEG substrates was similar, and the surface pore diameter was the nanometer level, which was consistent with the pore size analysis results. However few concavities (darker area) in microscale appeared on the surface of the PEG-0 substrat In the dry state for SEM observation, the surface layer of the substrate tended to shrin and collapse to some extent, which resulted in the formation of the concavities. The thi ner the surface layer was, the greater degree of the collapse was. With the increase in th PEG content, the number of the concavities gradually decreased and disappeared whe the PEG content was over 10 wt.%. This could result from the increased thickness of th The addition of PEG increased the porosity of the PSU substrate, compared with the control substrate (PEG-0), as shown in Table 2. It has been reported that the PEG in the casting solution facilitated the pore formation of the substrate during the NIPS [31,32]. However, the porosity was almost unchanged when the PEG content exceeded 10 wt.%. The increase in PEG content increased the viscosity of the casting solution, which slowed down the rate of the outward diffusion of the solvent and the inward diffusion of the non-solvent in the coagulation bath [33,34]. Thus, the NIPS rate of the casting solution was reduced, which was not conducive to the increase in the membrane porosity. Moreover, the porosity of the porous layer of PEG4-4, PEG10-4 and PEG20-4 was 84.6, 84.7 and 84.6, respectively, which indicated that the molecular weight of PEG did not affect the porosity under a certain PSU concentration. By reducing the PSU content, the porosity was increased as observed in PEG4-4 (PSU: 15 wt.%), PEG4-5 (PSU: 12 wt.%) and PEG4-6 (PSU: 9 wt.%), whose porosity was 84.6, 85.7 and 89.5, respectively. Compared with the porous layer, both the warp and weft yarns of the woven fabric layer did not have porous structures. Therefore, the porosity of the overall substrate was lower than that of its porous layer. The porosity of the fabric-reinforced substrate was affected by both the porosity of its porous layer and the volume proportion of its woven fabric.
The surface pore size of the substrate is critical in the formation of the active layer of the TFC-FO membrane. As shown in Table 2, the average surface pore diameter, d p , of the substrate was between 7-24 nm, which was within the pore size range of the ultrafiltration membrane. It was conducive to the formation of the high-quality active layer. The d p of PEG-0 was 19.0 nm. Under a constant PSU concentration, the d p decreased with the increase in PEG content and molecular weight in the casting solution. It resulted from the increased viscosity of the casting solution at higher PEG content or molecular weight. As discussed before, the increase in the viscosity slowed down the NIPS rate. It facilitated the formation of dense and small sized pores. However, at constant PEG content and molecular weight, the d p increased with the decrease in PSU concentration, which could be related to the reduced viscosity at low PSU concentration.
The surface hydrophilicity/hydrophobicity of the substrate is another important factor affecting the formation of the active layer of the TFC-FO membrane, and it can be investigated by means of water contact measurement. For porous materials, it should be Molecules 2023, 28, 2318 5 of 19 pointed out that the measured contact angle is an apparent value, depending on pore size, pore size distribution, roughness and chemical nature of the substrate. It can be seen from Table 2 that for the same substrate, its surface contact angle in the wet state was smaller than that in the dry state. In the dry state, the surface contact angle of the prepared PSU/PEG substrate was similar to that of the PEG-free substrate. In contrast, in the wet state, the contact angle of the PSU/PEG substrate is slightly lower than that of the PEG-0. This could result from the effect of the PEG on the surface roughness of the substrate. In general, the content and molecular weight of PEG did not significantly influence the contact angle of the substrate, because most of the PEG would dissolve into the conjugation bath during the phase conversion course. The hydrophilicity of the prepared PSU/PEG substrates could be comparable.
The ICP of a TFC-FO membrane is positively related to its substrate thickness. Table 2 shows that the overall thickness of the prepared PSU/PEG membranes was 59-67 µm, including the thickness of the woven fabric (about 50 µm). The thickness of the porous layer was only 9-17 µm, approximately. Due to the use of double-blade co-casting method, the ultra-thin and low-defect coating was prepared on the surface of the woven fabric. The thickness of the porous layer was effectively controlled. It should be noted that there was a difference in the thickness of the prepared substrates, especially when PEG20 was incorporated. Under a certain PSU concentration, the thickness of the PSU/PEG substrate increased with the increase in PEG molecular weight. The coating amount of the polymer per unit area tended to be increased with the increase in the viscosity of casting solution under the same casting knife thickness and speed, hence increasing the substrate thickness.
The porous support layer of the substrates exhibits a typical asymmetric structure, which is composed of a dense skin layer, a spongy-like microporous structure and a finger-like macroporous structure. ATR-FTIR was used to analyze the surface chemical composition of the substrates, and the FTIR spectra are shown in Figure 1. A characteristic absorption peak, corresponding to the sulfone group (S=O=S), was observed at~1323 cm −1 in each FTIR spectrum. The peak intensity of the sulfone group was 0.044 in the PEG-0 substrate. At a certain PSU concentration, the intensity increased with the increase in PEG content (PEG-0, PEG4-1, PEG4-2, PEG4-3, PEG4-4) or molecular weight (PEG4-4, PEG10-4, PEG20-4). According to the positive correlation between the peak intensity of the sulfone group and its content, the increase in the PEG content or molecular weight could result in a denser surface structure and increase the thickness of the skin layer and the spongylike microporous structure. Thus, the PSU content at the substrate surface was increased. Furthermore, under a certain PEG4 content at 20 wt.%, the peak intensity decreased with the decrease in the PSU concentration (PEG4-4, PEG4-5, PEG4-6), which may be relative to the reduced sulfone content and thickness of the surface layer.
The surface, cross-section and bottom morphology of the prepared PSU/PEG substrates are shown in Figures 2-4, respectively. The surface morphology (Figure 2) of the prepared PSU and PSU/PEG substrates was similar, and the surface pore diameter was at the nanometer level, which was consistent with the pore size analysis results. However, a few concavities (darker area) in microscale appeared on the surface of the PEG-0 substrate. In the dry state for SEM observation, the surface layer of the substrate tended to shrink and collapse to some extent, which resulted in the formation of the concavities. The thinner the surface layer was, the greater degree of the collapse was. With the increase in the PEG content, the number of the concavities gradually decreased and disappeared when the PEG content was over 10 wt.%. This could result from the increased thickness of the surface layer, including the skin layer and the spongy-like microporous structure, at higher PEG incorporation. surface layer, including the skin layer and the spongy-like microporous structure, at higher PEG incorporation. It can be observed in Figure 3 that the woven fabrics were embedded in the porous support layers instead of being cushioned at their bottom for the PSU/PEG substrates. Some of the finger-like macropores were cut off by the warp and weft yarns of the woven fabrics, and only the macropores located in the openings of fabrics penetrated through to the bottom of the substrate, which may increase the pore curvature of the substrate. It is worth noting that the cross-section morphology of the porous support layer under different PEG content or molecular weight was different. It can be clearly observed from the insets in Figure 3 that at the same PSU concentration, the thickness of the spongy-like micropore structure increased with the increase in the PEG content or molecular weight, while the number of the finger-like macropore structure decreased, and the macropore channel became wider. For the substrates with 20 wt.% PEG4, the thickness decreased with the decrease in the PSU content. They were consistent with the FTIR results. Compared with the PEG-0 substrate, the porous support layer in the PSU/PEG substrate exhibited better macropore channel connectivity with fewer closed capsule-like macropores. The macropore channels were uniform for most of the PSU/PEG substrates. According to the NIPS mechanism, the cross-section morphology of the porous support layer is determined by the thermodynamic and kinetic properties during the NIPS course. When the primary film was immersed in the coagulation bath, the casting solution solvent and coagulation agent inter-diffused. With the increasing amount of coagulation agent into the casting solution, liquid-liquid phase separation occurred in the casting solution, forming a polymer-poor phase and a polymer-rich phase. Delayed liquid-liquid phase separation was often conducive to the formation of a sponge-like micropore structure, while instantaneous liquid-liquid phase separation was often conducive to the formation of a large macropore structure [17,35,36]. After the liquid-liquid phase separation of the casting solution, the mutual diffusion between the casting solution solvent and the coagulation agent continued. The polymer-rich phase was solidified to form a film during crystallization, gelation or vitrification. When the growth rate of the polymer-poor phase was lower than the solidification rate of the polymer-rich phase, the formation of the macroporous structure was inhibited, otherwise it was conducive to the macro pore formation [17,34]. Due to the poor compatibility between PSU and PEG, the addition of PEG in the casting solution reduced the thermodynamic stability of the casting solution, and the solubility of PEG in NMP decreased with the increase in its molecular weight [34]. Therefore, when the PSU concentration was constant, the thermodynamic stability of PSU/PEG/NMP casting solution decreased with the increase in PEG content or molecular weight, which was conducive to the liquid-liquid phase separation. The growth rate of the polymer-rich phase was slower than the solidification rate of the polymer-rich phase. However, the viscosity of the casting solution increased with the increase in PEG content or molecular weight. The increased viscosity could slow down the rate of the outward diffusion of the casting solution solvent and the inward diffusion of the coagulation agent, which was unfavorable to the liquid-liquid phase separation and reduced the growth rate of the polymer-poor phase. Figure 4 shows that there were numerous macro size pores on the bottom surface of each substrate, and the presence of these pores could facilitate the solution to enter and exit the substrate. By means of statistical analysis, it was found that the number of the macropores increased with the increase in the PEG content but decreased with the increase in its molecular weight. PEG4 substrates may exhibit better permeability than PEG10 and PEG20 substrates. Moreover, at a certain PEG4 content of 20 wt.%, the number of the pores first increased and then decreased with the decrease in the PSU concentration. PEG4-5 possessed the largest number of pores, compared with PEG4-4 and PEG4-6. It can be seen from Figure 4 that most of the woven fabrics, excluding the intersections of warp and weft yarns, were embedded in the porous support layer, which ensured the firm combination of the woven fabric and the porous support layer.   The surface chemical composition of the prepared TFC-FO membranes was analyzed by ATR-FTIR, and the results are shown in Figure 5. Two characteristic peaks at ~1663 and ~1541 cm −1 , corresponding to the amide I (C=O stretching vibration) and amide II (C-N stretching and in-plane N-H bending vibration), respectively, were observed. This indicated that PA active layers were successfully synthesized on the surface of the prepared PSU/PEG substrates through interfacial polymerization. Since the detection depth of ATR-FTIR measurement was greater than the thickness of the PA active layer, the characteristic peak at ~1323 cm −1 for the sulfone groups (S=O=S) in the substrate also appeared. The thickness of the PA active layer could be analyzed by using the characteristic absorption peak intensity ratio of amide and sulfone groups in the FTIR spectrum of each membrane. In general, a higher ratio of I (C=O)/I (S=O=S) indicates a thicker PA layer [11,37]. It can be seen from Figure 5 that the intensity ratio for TFC/PEG-0 was 0.64, which was higher than that for the TFC/PEG membranes with 15 wt.% PSU. It implied that the incorporation of PEG could reduce the thickness of the PA layer. A thinner PA layer could result in lower resistance to salt and water transport. However, there was no big difference of the ratio for the membranes with varied PEG content or molecular weight. The PEG content and molecular weight may not significantly affect the formation and the thickness of the active layer. In contrast, for the membranes with 12 and 9 wt.% PSU, the intensity ratio increased dramatically. In this case, the analysis of PA layer thickness by intensity ratio could be invalid, because the intensity of sulfone groups reduced notably with the decrease in PSU concentration. It can be observed in Figure 3 that the woven fabrics were embedded in the porous support layers instead of being cushioned at their bottom for the PSU/PEG substrates. Some of the finger-like macropores were cut off by the warp and weft yarns of the woven fabrics, and only the macropores located in the openings of fabrics penetrated through to the bottom of the substrate, which may increase the pore curvature of the substrate. It is worth noting that the cross-section morphology of the porous support layer under different PEG content or molecular weight was different. It can be clearly observed from the insets in Figure 3 that at the same PSU concentration, the thickness of the spongy-like micropore structure increased with the increase in the PEG content or molecular weight, while the number of the finger-like macropore structure decreased, and the macropore channel became wider. For the substrates with 20 wt.% PEG4, the thickness decreased with the decrease in the PSU content. They were consistent with the FTIR results. Compared with the PEG-0 substrate, the porous support layer in the PSU/PEG substrate exhibited better macropore channel connectivity with fewer closed capsule-like macropores. The macropore channels were uniform for most of the PSU/PEG substrates. According to the NIPS mechanism, the cross-section morphology of the porous support layer is determined by the thermodynamic and kinetic properties during the NIPS course. When the primary film was immersed in the coagulation bath, the casting solution solvent and coagulation agent inter-diffused. With the increasing amount of coagulation agent into the casting solution, liquid-liquid phase separation occurred in the casting solution, forming a polymerpoor phase and a polymer-rich phase. Delayed liquid-liquid phase separation was often conducive to the formation of a sponge-like micropore structure, while instantaneous liquid-liquid phase separation was often conducive to the formation of a large macropore structure [17,35,36]. After the liquid-liquid phase separation of the casting solution, the mutual diffusion between the casting solution solvent and the coagulation agent continued. The polymer-rich phase was solidified to form a film during crystallization, gelation or vitrification. When the growth rate of the polymer-poor phase was lower than the solidification rate of the polymer-rich phase, the formation of the macroporous structure was inhibited, otherwise it was conducive to the macro pore formation [17,34]. Due to the poor compatibility between PSU and PEG, the addition of PEG in the casting solution reduced the thermodynamic stability of the casting solution, and the solubility of PEG in NMP decreased with the increase in its molecular weight [34]. Therefore, when the PSU concentration was constant, the thermodynamic stability of PSU/PEG/NMP casting solution decreased with the increase in PEG content or molecular weight, which was conducive to the liquid-liquid phase separation. The growth rate of the polymer-rich phase was slower than the solidification rate of the polymer-rich phase. However, the viscosity of the casting solution increased with the increase in PEG content or molecular weight. The increased viscosity could slow down the rate of the outward diffusion of the casting solution solvent and the inward diffusion of the coagulation agent, which was unfavorable to the liquid-liquid phase separation and reduced the growth rate of the polymer-poor phase. Figure 4 shows that there were numerous macro size pores on the bottom surface of each substrate, and the presence of these pores could facilitate the solution to enter and exit the substrate. By means of statistical analysis, it was found that the number of the macropores increased with the increase in the PEG content but decreased with the increase in its molecular weight. PEG4 substrates may exhibit better permeability than PEG10 and PEG20 substrates. Moreover, at a certain PEG4 content of 20 wt.%, the number of the pores first increased and then decreased with the decrease in the PSU concentration. PEG4-5 possessed the largest number of pores, compared with PEG4-4 and PEG4-6. It can be seen from Figure 4 that most of the woven fabrics, excluding the intersections of warp and weft yarns, were embedded in the porous support layer, which ensured the firm combination of the woven fabric and the porous support layer.
The surface chemical composition of the prepared TFC-FO membranes was analyzed by ATR-FTIR, and the results are shown in Figure 5. Two characteristic peaks at~1663 and~1541 cm −1 , corresponding to the amide I (C=O stretching vibration) and amide II (C-N stretching and in-plane N-H bending vibration), respectively, were observed. This indicated that PA active layers were successfully synthesized on the surface of the prepared PSU/PEG substrates through interfacial polymerization. Since the detection depth of ATR-FTIR measurement was greater than the thickness of the PA active layer, the characteristic peak at~1323 cm −1 for the sulfone groups (S=O=S) in the substrate also appeared. The thickness of the PA active layer could be analyzed by using the characteristic absorption peak intensity ratio of amide and sulfone groups in the FTIR spectrum of each membrane. In general, a higher ratio of I (C=O)/I (S=O=S) indicates a thicker PA layer [11,37]. It can be seen from Figure 5 that the intensity ratio for TFC/PEG-0 was 0.64, which was higher than that for the TFC/PEG membranes with 15 wt.% PSU. It implied that the incorporation of PEG could reduce the thickness of the PA layer. A thinner PA layer could result in lower resistance to salt and water transport. However, there was no big difference of the ratio for the membranes with varied PEG content or molecular weight. The PEG content and molecular weight may not significantly affect the formation and the thickness of the active layer. In contrast, for the membranes with 12 and 9 wt.% PSU, the intensity ratio increased dramatically. In this case, the analysis of PA layer thickness by intensity ratio could be invalid, because the intensity of sulfone groups reduced notably with the decrease in PSU concentration.
The PA active layer of the TFC membrane was synthesized by MPD and TMC. It was composed of a linear portion that contained free pendant -COOH hydrolyzed from the unreacted -COCl and a cross-linked portion where -COCl was all involved in a cross-linking reaction with -NH 2 . The cross-linking degree was important in determining the water flux and salt rejection of TFC membranes [11,38]. Since the detection depth of XPS was less than the thickness of the active layer, XPS was used to investigate the surface element composition, content and cross-linking degree of the active layer. The results are shown in Figure 6 and Table 3. It can be seen from Table 3 that the active layer was mainly composed of carbon, oxygen and nitrogen, which was consistent with the element composition of PA. The oxygen atoms were from -CONH-(N-C=O: O I,~531.6 ev) and -COOH (O-C=O: O II,~533.0 ev), while nitrogen atoms were from -CONH-and -NH 2 . A higher O I/O II content ratio corresponded to a smaller O/N content ratio. Theoretically, the O/N ratios of fully cross-linked and fully linear PA were 1.0 and 2.0 respectively. Thus, the cross-linking degree of the PA layer can be estimated by the O/N ratio and O I/O II ratio. The closer the ratio was to 1.0, the higher the cross-linking degree of the PA layer was. In Table 3, the O/N ratio of TFC/PEG-0 was 1.18, indicating a highly cross-linked active layer. The ratio increased with the increase in the PEG content and its molecular weight. This could mainly result from the pore size of the substrate. As shown in Table 2, the pore diameter of the substrate decreased with the increase in the PEG content and its molecular weight. During the IP course, the smaller pores on the substrate surface absorbed less amount of MPD solution that migrated to reaction interface for the following IP with TMC. Thus, a less cross-linked PA layer was favored with higher O/N ratio. TFC/PEG20-4 exhibited the highest O/N ratio among the membranes. In contrast, with the decrement in PSU content from 15 to 12 wt.%, the pore size of the substrate increased. It facilitated sufficient MPD supply to the reaction zone, resulting in a notable increase in the cross-linking degree. The O/N ratio of TFC/PEG4-5 was as low as 1.13, compared with that of TFC/PEG4-4. However, the ratio of TFC/PEG4-6 was slightly raised with further decline in the PSU content to 9 wt.%. The excess MPD absorbed in its big pores could adversely affect the active layer formation [39].  The PA active layer of the TFC membrane was synthesized by MPD and TMC. It w composed of a linear portion that contained free pendant -COOH hydrolyzed from th unreacted -COCl and a cross-linked portion where -COCl was all involved in a cross-lin ing reaction with -NH2. The cross-linking degree was important in determining the wat flux and salt rejection of TFC membranes [11,38]. Since the detection depth of XPS w less than the thickness of the active layer, XPS was used to investigate the surface eleme composition, content and cross-linking degree of the active layer. The results are show in Figure 6 and Table 3. It can be seen from Table 3   The surface and cross-section morphologies of the TFC-FO membranes were investigated by SEM as shown in Figures 7 and 8, respectively. The surface of each membrane presented a typical ridge-and-valley structure. There was no significant difference in terms of the surface morphologies of the membranes prepared at different PEG content or molecular weight. The active layer was firmly combined with its porous support layer. The ridge-and-valley structure of the active layer was further confirmed from AFM images as displayed in Figure 9. The corresponding surface roughness (R q and R a ) of each membrane was obtained and listed in Table 4. In general, the fouling resistance of a TFC membrane is negatively correlated with its surface roughness, while the water permeability coefficient is positively correlated with the roughness. The R q and R a of TFC/PEG-0 was 93.6 and 75.2, respectively. The incorporation of PEG in the casting solution of the substrate did not significantly influence the surface roughness of the active layer under the same interfacial polymerization conditions.  The surface and cross-section morphologies of the TFC-FO membranes were investigated by SEM as shown in Figures 7 and 8, respectively. The surface of each membrane presented a typical ridge-and-valley structure. There was no significant difference in terms of the surface morphologies of the membranes prepared at different PEG content or  The FO performances, including J w , J s and J s /J w , were measured at different draw solution concentrations, and the results are presented in Figures 10, 11 and S1-S4. The transport properties, including A, B, A/B and S, of the prepared TFC-FO membranes were calculated by fitting the measured J w and J s . The values were listed in Tables 5 and S1. Generally, the A/B and J s /J w can reflect the permselectivity of FO membrane. A/B could be mainly affected by the active layer thickness and cross-linking degree. Higher A/B and lower J s /J w means better permselectivity of the formed PA layer and less draw solution loss in FO. The S value quantifies the ICP extent in the membrane, which is critical for membrane design. The S value of TFC/PEG-0 was 363 µm. Incorporation of PEG in the casting solution reduced the structural parameter. It further reduced with the increase in PEG content, which was mainly related to the changes in the properties of the membrane substrate. As discussed in the SEM morphology, under the same PSU concentration, the thickness of spongy-like microporous structure in the substrate increased with the increase in PEG content, while that of the finger-like macroporous structure decreased relatively. Previous studies have indicated that sponge-like micropore structure is more likely to cause ICP than finger-like macropore structure [17,27]. Additionally, the curvature of sponge pore is normally greater than that of finger pore. Nevertheless, we believe that the curvature of sponge pore could be reduced in the case of high interpore connectivity. Feng et al. [40] demonstrated that the addition of PEG in the casting solution could increase the interpore connectivity in the sponge structure of the substrate that was formed during NIPS. It contributed to the decrease in structural parameter.
substrate did not significantly influence the surface roughness of the active layer under the same interfacial polymerization conditions. The FO performances, including , and / , were measured at different draw solution concentrations, and the results are presented in Figures 10 and 11 and S1-S4. The transport properties, including A, B, A/B and S, of the prepared TFC-FO membranes were calculated by fitting the measured and . The values were listed in Tables 5 and S1. Generally, the A/B and / can reflect the permselectivity of FO membrane. A/B could be mainly affected by the active layer thickness and cross-linking degree. Higher A/B and lower / means better permselectivity of the formed PA layer and less draw solution loss in FO. The S value quantifies the ICP extent in the membrane, which is critical for membrane design. The S value of TFC/PEG-0 was 363 μm. Incorporation of PEG in the casting solution reduced the structural parameter. It further reduced with the increase in PEG content, which was mainly related to the changes in the properties of the membrane substrate. As discussed in the SEM morphology, under the same PSU concentration, the thickness of spongy-like microporous structure in the substrate increased with the increase in PEG content, while that of the finger-like macroporous structure decreased relatively. Previous studies have indicated that sponge-like micropore structure is more likely to cause ICP than finger-like macropore structure [17,27]. Additionally, the curvature of sponge pore is normally greater than that of finger pore. Nevertheless, we believe that the curvature of sponge pore could be reduced in the case of high interpore connectivity. Feng et al. [40] demonstrated that the addition of PEG in the casting solution could increase the interpore connectivity in the sponge structure of the substrate that was formed during NIPS. It contributed to the decrease in structural parameter.         the ICP of the FO membrane. The S value is lower than those of previous reported fabricreinforced TFC FO membranes [18,29,[41][42][43].    the ICP of the FO membrane. The S value is lower than those of previous reported fabri reinforced TFC FO membranes [18,29,[41][42][43].    In can be observed in Figure 10 that the water flux of all TFC-FO membranes increased with the concentration of the draw solution, because the driving force for water transfer in the FO process was positively correlated with the draw solution concentration. In contrast, the specific reverse salt flux of the membranes was hardly affected by the draw solution concentration as shown in Figure 11. The prepared FO membranes exhibited good integrity. Under the same operation condition, the water flux of the TFC-FO membrane decreased with the increase in PEG molecular weight in the casting solution, which may be mainly related to the molecular composition of the active layer. The PA layer may contain ester groups, especially when the PEG with high molecular weight was used in the casting solution. The increase in the number of hydrophobic ester groups in the active layer was not conducive to the solution-diffusion behavior of water molecules, resulting in the decrease in the water permeability coefficient. Thus, the A value of the membranes decreased with the increase in PEG molecular weight in the casting solution. Figures 10 and S1 shows that when the molecular weight of PEG was 400 and 1000 g/mol, the water flux of the membranes increased with the increase in PEG content, which was due to the improvement of their water permeability coefficient and structural parameter. The J w of TFC/PEG4-4, containing 20 wt.% PEG in casting solution, was 22.4 LMH when using 1 M NaCl as draw solution. However, when the molecular weight of PEG was 2000 g/mol, the water flux of the membranes first increased and then decreased with the increase in PEG content. All the PEG20 membranes showed lower water flux compared with that of TFC/PEG-0. Although their S value was lower than that of the control, their A value was quite low. As indicated in Table 3, the O/N ratio of PEG20-4 was significantly higher than that of TFC/PEG-0, indicating a low cross-linking degree. The insufficient cross-linked PA layer could result in a low osmotic pressure and thus a low water permeation under FO operation. In terms of specific reverse salt flux, overall, the J s /J w value of TFC/PEG4 was slightly lower than that of TFC/PEG10 and TFC/PEG20 under the same operation conditions. This could result from the influences of both PA layer thickness and cross-linking degree. Table 5. Intrinsic properties of PSU/PEG supported TFC-FO membranes.

Membrane
A Overall, TFC/PEG4 exhibited better FO performances over TFC/PEG10 and TFC/PEG20. A 20 wt.% was demonstrated to be the optimal PEG content in the casting solution. Moreover, the water flux of the membrane was improved by reducing the PSU concentration from 15 wt.% to 12 wt.%. The J w of TFC/PEG4-5 was 25.0 LMH using 1 M NaCl as draw solution, and the J s /J w was as low as 0.12 g/L. However, further decrease in PSU concentration (9 wt.%) reduced the J w . This could mainly result from the improvement in the substrate structure. As observed in the bottom surface of the substrates in Figure 4, TFC/PEG4-5 possessed a larger number of macro pores that penetrated throughout the substrate, compared with that of TFC/PEG4-4 and TFC/PEG4-6. The pores played a key role in the water permeation. The J w and J s /J w of TFC/PEG4-5 in this work is better than that of the reported fabric-reinforced TFC FO membranes [29,[41][42][43]. The S value of TFC/PEG4-5 and TFC/PEG4-6 membranes were 248.8 and 242.5 µm, respectively ( Table 5). The decrease in PSU concentration in the casting solution was beneficial to reducing the ICP of the FO membrane. The S value is lower than those of previous reported fabric-reinforced TFC FO membranes [18,29,[41][42][43].

Preparation of TFC Membrane
The fabric reinforced PSU/PEG substrate was fabricated by means of non-solvent induced phase separation (NIPS). PSU and PEG were stirred in NMP solvent at room temperature until it was completely dissolved. The mixture was then placed in a desiccator for degassing for about 24 h, in order to obtain homogeneous and transparent casting solution. The PET woven fabric was flatted and taped on a clean glass plate. The casting solution was uniformly spread on the fabric by a 45 µm casting knife. Then, the primary film was immediately immersed into a 20 • C coagulation bath (DI water) for phase transformation for 10 min. The prepared membrane substrate was rinsed by DI water to completely remove the solvent. The polyamide (PA) active layer was synthesized on the substrate by interfacial polymerization (IP): The substrate was immersed by 2.0 wt.% MPD aqueous solution and soaked for 2 min. After removing the excess MPD solution, 0.1 wt.% TMC/n-hexane solution was placed on the substrate surface for 1 min IP reaction, followed by hexane rinsing to remove the unreacted TMC. The resultant TFC membrane was cured at 60 • C for 3 min, and then stored in DI water at 4 • C for further tests. A schematic of the preparation process of the TFC membrane was depicted in Figure 12, and the images recording the process were presented in Figure S5. The prepared TFC-FO membranes were denoted as TFC/PEG-0, TFC/PEG4, TFC/PEG10 and TFC/PEG20, corresponding to the membrane substrates PEG-0, PEG4, PEG10 and PEG20, respectively.

Preparation of TFC Membrane
The fabric reinforced PSU/PEG substrate was fabricated by means of non-solvent induced phase separation (NIPS). PSU and PEG were stirred in NMP solvent at room temperature until it was completely dissolved. The mixture was then placed in a desiccator for degassing for about 24 h, in order to obtain homogeneous and transparent casting solution. The PET woven fabric was flatted and taped on a clean glass plate. The casting solution was uniformly spread on the fabric by a 45 μm casting knife. Then, the primary film was immediately immersed into a 20 °C coagulation bath (DI water) for phase transformation for 10 min. The prepared membrane substrate was rinsed by DI water to completely remove the solvent. The polyamide (PA) active layer was synthesized on the substrate by interfacial polymerization (IP): The substrate was immersed by 2.0 wt.% MPD aqueous solution and soaked for 2 min. After removing the excess MPD solution, 0.1 wt.% TMC/n-hexane solution was placed on the substrate surface for 1 min IP reaction, followed by hexane rinsing to remove the unreacted TMC. The resultant TFC membrane was cured at 60 °C for 3 min, and then stored in DI water at 4 °C for further tests. A schematic of the preparation process of the TFC membrane was depicted in Figure 12, and the images recording the process were presented in Figure S5. The prepared TFC-FO membranes were denoted as TFC/PEG-0, TFC/PEG4, TFC/PEG10 and TFC/PEG20, corresponding to the membrane substrates PEG-0, PEG4, PEG10 and PEG20, respectively.

Characterization
The porosity of the substrate was obtained by a wet-dry weight method. The preserved substrate in DI water was weighed (m1) after absorbing the excess water on the surface by filter paper. The substrate was then dried under vacuum at 60 °C for 12h and weighed (m2). Then the substrate was dissolved in NMP to remove PSU. The remaining

Characterization
The porosity of the substrate was obtained by a wet-dry weight method. The preserved substrate in DI water was weighed (m 1 ) after absorbing the excess water on the surface by filter paper. The substrate was then dried under vacuum at 60 • C for 12h and weighed (m 2 ). Then the substrate was dissolved in NMP to remove PSU. The remaining fabric was cleaned with DI water and further dried under vacuum at 60 • C for 12 h. The weight (m 3 ) was measured. The porosity, ε(%), was calculated: where ρ w , ρ p and ρ f represent the density of water, PSU and fabric, respectively. A droplet shape analyzer (DSA100, KRUSS, Germany) was used to measure the water contact angles on the surface of dry and wet substrate. The droplet volume was 3 µL, and the contact angles were measured by sessile drop method. The thickness of the substrate was measured with a bench thickness meter (CH-12.7-STSX, Shanghai Liuling Instrument Factory, Shanghai, China) in wet condition. The mean pore diameter, d p , of the substrate was measured by a liquid-liquid displacement method (PRM-1200GL, Benelux Scientific, Ede, Netherlands). The surface tension of the wetting solution was 1.7 mN/m. Atomic force microscopy (AFM, Dimension Icon, Bruker, Santa Barbara, CA, USA) was used to characterize the surface morphology and roughness of the TFC-FO membranes. The tap mode was selected, and the scan range was 5.0 µm × 5.0 µm. Scanning electron microscopy (SEM, JSM-7800F, JEOL) was used to characterize the surface, cross-section and bottom morphology of the substrates and TFC-FO membranes. The samples for cross-sectional view were placed in liquid nitrogen and carefully cut by a sharp razor knife. All the samples were fully dried and coated by gold before the SEM observation. Attenuated total reflection Fourier transform infrared spectroscopy (ATR-FTIR, Cary630, Agilent) was used to characterize the surface chemistry of the substrates and TFC-FO membranes with resolution of 2 cm −1 . X-ray photoelectron spectroscopy (XPS, Esca Lab 250Xi, Thermo Fisher Scientific, Waltham, MA, USA) was used to analyze the elements on the surface of the TFC-FO membranes. The water flux (J w ), reverse salt flux (J s ) and specific reverse salt flux (J s /J w ) of the prepared TFC-FO membrane was obtained by a custom cross-flow FO system. The effective test membrane area of the FO unit was 42 cm 2 . During the test, the FO mode (active layer against feed solution) was selected, where the cross-flow velocity of the feed solution (DI water) and the draw solution (NaCl solution) was 6.4 cm/s. The concentration of the NaCl draw solution included 0.5 mol/L, 1.0 mol/L, 1.5 mol/L and 2.0 mol/L. All the tests were conducted at 25 ± 1 • C for 30 min, and the changes of the conductivity of the feed solution and the weight of the draw solution were recorded every 2 min by a conductivity meter (DDSJ-308A, INESA Scientific Instrument, Shanghai, China) and balance (BSA6202S-CW, Sartorius), respectively. The J w (Lm −2 h −1 or LMH) was calculated according to Equation (2): J w = </mo>V (2) where ∆V (L) is the volume of permeate water during ∆t (h), and A m (m 2 ) is the effective membrane test area. The J s (gm −2 h −1 or gMH) was calculated via Equation (3): where C t (g/L) and V t (L) are the concentration of the NaCl draw solute and the volume of the feed solution over ∆t, respectively. The C t was obtained according to a standard draw solution concentration-conductivity curve. According to the method reported by Tiraferri et al. [41], the water permeability coefficient (A), salt permeability coefficient (B) and structural parameter (S) of the prepared TFC-FO membrane were calculated by fitting the measured J w and J s at four draw solution concentrations.

Conclusions
In this work, a series of PEG modified fabric-reinforced PSU TFC-FO membranes were prepared. The incorporation of the non-solvent additive PEG could change the thermodynamic and gel kinetic properties of PSU/NMP casting solution, thus affecting the porosity, pore structure, hydrophilicity, thickness, cross-linking degree and roughness of the membrane. The influences of the PEG content and molecular weight as well as PSU concentration on the FO performances and transport properties were systematically investigated. Appropriate PEG addition effectively decreased ICP and optimized the permselectivity of the membrane. J w of TFC/PEG4-5 was 25.0 LMH using 1 M NaCl as draw solution, and the J s /J w was as low as 0.12 g/L.