Fabrication of a novel cyanoethyl cellulose substrate for thin-film composite forward osmosis membrane

A

water permeability

B

salt permeability

S

structure parameter

ε

porosity

r_m

average pore size

CEC

cyanoethyl cellulose

DS

draw solution

FS

feed solution

FO

forward osmosis

ICP

internal concentration polarization

PRO

pressure-retarded osmosis

PWP

pure water permeability

PVP

polyvinylpyrrolidone

RSF

reverse salt flux

SRSF

specific reverse salt flux

TFC

thin-film composite

Because of the shortage of freshwater resources, seawater desalination and wastewater treatment have become important issues (Elimelech & Phillip 2011). Thus, forward osmosis (FO) technology, a potential technology for seawater desalination and wastewater treatment, has gained significant attention in recent decades (Akther et al. 2015; Qasim et al. 2015; Chekli et al. 2016; Lu et al. 2016a). Moreover, besides seawater desalination and wastewater treatment, the FO process could have applications in the enrichment of liquid foods or pharmaceuticals and osmotic power generation (Ling & Chung 2011; Altaee et al. 2015; Sant'Anna et al. 2016; Ansari et al. 2017). Unlike reverse osmosis (RO) and nanofiltration (NF), which use high hydraulic pressures to separate the solvents and solutes, FO utilizes the osmotic pressure difference between the low-osmotic-pressure feed solution (FS) and high-osmotic-pressure draw solution (DS) to drive water across the FO membrane (Chekli et al. 2016). Furthermore, the FO process has distinct advantages, such as low energy consumption, excellent separating precision, and reversible membrane fouling (Qasim et al. 2015).

An ideal FO membrane should have a high water flux and limited reverse salt flux (RSF). However, FO performance is severely hindered by the inevitable internal concentration polarization (ICP) that occurs during the FO process (Qasim et al. 2015; Wang & Xu 2015; Bui & McCutcheon 2016; Heikkinen et al. 2017). Meanwhile, structure parameter (S value) is used to reflect the ICP degree in the FO process (Huang et al. 2019). Generally, lower S value reflects slighter ICP, which means better FO performance for the membrane (Zheng et al. 2018a). Compared to asymmetric FO membranes, thin-film composite (TFC) membranes exhibit better performance for the FO process (Chekli et al. 2016). Numerous substrate optimization studies have been published in recent years, and the membrane substrate is known to play an important role in controlling the performance of TFC FO membranes. Consequently, improving the performance of TFC FO membranes remains a hot research topic. The first basic strategy to increase FO performance is to modify the pore morphology of the substrate by adding pore-forming agents (Chen et al. 2017c). LiCl, ZnCl₂, ethylene glycol (EG), ethanol, polyethylene glycol (PEG), and polyvinyl pyrrolidone (PVP) are common pore-forming agents (Emadzadeh et al. 2014; Morales-Torres et al. 2016; Wu et al. 2016; Chen et al. 2017c). Recently, Morales-Torres et al. (2016) illustrated that the pore-forming agent PVP could form larger and elongated finger-like pores across membrane substrates. As such, PVP can significantly improve the pore morphology and hydrophilicity of membrane substrates. Wu et al. (2016) used PEG as a pore-forming agent to optimize the hydrophilicity, porosity, and pore size of the TFC FO membrane substrates. Their results showed that an FO membrane prepared with 6 wt.% PEG-400 yielded the best FO performance. In addition, alternative methods such as blending hydrophilic polymers or nanoparticles with basic membrane substrate materials are becoming attractive to enhance the hydrophilicity of the substrates and reduce the ICP during the FO process (Zhou et al. 2014; Ghanbari et al. 2015a; Lu et al. 2016a, 2016b; Zhang et al. 2016). Zhou et al. (2014) blended sulfonated poly(phenylene oxide) (SPPO) with polysulfone to fabricate hydrophilic substrates. The results demonstrated that SPPO alleviated the ICP during the FO process and improved the membrane performance. Ghanbari et al. (2015a) illustrated that addition of hydrophilic halloysite nanotubes into the substrates improved the overall porosity, mean pore size, and hydrophilicity of the blended substrates and reduced the S value of the substrate. Lu et al. (2016b) reported that layered double hydroxide nanoparticles (LDH-NPs) could be successfully introduced into FO membrane substrates. After adding 2 wt.% LDH-NPs, the water flux became 18.1 LMH at 1 M NaCl as DS and deionized (DI) water as FS under FO mode, which is 42.5% higher than the pristine FO membrane.

Because cellulose is the most abundant natural polymer in the world, cellulose and its derivatives are widely used (Teramoto 2015). In particular, cellulose acetate (CA) and cellulose triacetate (CTA) are suitable materials for membrane fabrication because of their moderate flux, excellent separating precision, mechanical robustness, and nontoxicity (Qasim et al. 2015; Li et al. 2016; Chen et al. 2017a; Liu et al. 2017). However, CA and CTA are being gradually replaced by other artificial polymers, such as polysulfone (PSF), polyether sulfone (PES), polyacrylonitrile (PAN), and polyvinylidene fluoride (PVDF), mainly because both CA and CTA undergo hydrolysis and biodegradation (Qasim et al. 2015). However, another cellulose derivative, cyanoethyl cellulose (CEC), exhibits excellent physical and chemical properties, such as increased thermal resistance, excellent mechanical properties, and, especially, microbiological resistance (Zhou et al. 2010; Joshi et al. 2017). Therefore, CEC is an excellent candidate membrane material.

In this study, the potential of CEC for the fabrication of substrates for TFC FO membranes was investigated. PVP was used as the pore-forming agent and added to the casting solutions. The influence of PVP on the CEC substrates properties in terms of pure water permeability, contact angle, membrane thickness, porosity, average pore size, mechanical strength, and membrane morphology was systematically investigated. In addition, the mechanism of the effect of the pore-forming agent on the active layer properties is demonstrated. The FO performances of CEC-substrate FO membranes with different PVP contents were studied, and the FO performances of the CEC-substrate FO membranes were compared to different FO membrane substrates. This is the first report concerning the application of CEC to the fabrication of FO membranes substrates.

Cyanoethyl cellulose (CEC) powders (substitution degree of 2.6, the chemical structure is shown in Figure 1), polyvinyl pyrrolidone K30 (PVP K30), and N,N-dimethylacetamide (DMAc) were purchased from Aladdin and used to fabricate the membrane substrates. m-Phenylenediamine (MPD) with >99% purity was purchased from Sigma-Aldrich, and trimesoyl chloride (TMC) of >98% purity was purchased from Alfa Aesar; these were used as the monomers for interfacial polymerization (IP). n-Hexane with >99% purity from Aladdin was utilized as the solvent for TMC. Reagent grade sodium chloride (NaCl) supplied by Aladdin was used to prepare the DS.

The casting solutions were composed of 16 wt.% CEC and various amounts of PVP K30, as shown in Table 1. The CEC and PVP were dissolved in DMAc and stirred by an orbital shaker at 60 °C for 24 h to obtain a homogeneous and transparent solution. Before casting, the prepared solution was left for degassing at room temperature for 24 h. About 10 mL casting solution was poured on a pre-cleaned glass plate, and was cast by a stainless-steel film applicator with a gap of 150 μm to form a 150 mm × 200 mm liquid membrane. After evaporation of the solvent for 30 s, the glass plate with the liquid membrane was immersed in a deionized water coagulation bath at room temperature to initiate phase inversion and form a solid substrate membrane. After sitting in the coagulation bath for 10 min, the obtained substrate was transferred into a flowing deionized water bath for 48 h to remove residual solvent before further use.

The active layer of the TFC FO membrane was formed by IP on the surface of the CEC substrate (Zheng et al. 2018b). First, the pre-cast substrate membrane was soaked in an aqueous solution of 2 wt.% MPD for 120 s, and the excess MPD solution on the substrate surface was removed by blowing pure compressed nitrogen gas through an air knife. Subsequently, an 0.10 wt.% n-hexane solution of TMC was gently poured onto the MPD-soaked substrate surface and allowed to react with the residual MPD for 60 s, forming the PA active layer. After the TMC solution had drained off, the obtained TFC FO membrane was air-cured for 5 min. The resultant TFC FO membrane was rinsed with DI water to remove the residual solution and then stored in DI water before characterization and performance testing.

The mechanical properties of the CEC substrates were tested using a dynamic mechanical analyzer (DMA, Q800DE, TA, USA). The membrane samples were cut into strips of 8 mm width and 30 mm length. All tests were run at a ramp rate of 1 N/min at 25 °C.

The membrane morphologies were studied using a high-resolution field emission scanning electron microscope (HR-FESEM, Merlin, Carl Zeiss, Germany). The membrane substrates and TFC FO membranes were first dried by vacuum freeze drying for 48 h. The top surfaces, bottom surfaces, and cross sections of the membrane substrates and the top surfaces of TFC FO membranes were scanned. In particular, to observe cross sections of the membrane substrates, the samples were flash-frozen and cracked in liquid nitrogen.

FO experiments were conducted on a lab-scale cross-flow filtration setup. The effective membrane area in the FO cell was 12 cm² (2 cm × 6 cm). The FS and DS were circulated at the same fixed cross-flow rate of 480 mL/min on both sides of the FO membrane. A 1.0 M NaCl aqueous solution was used as the DS, while DI water was used as the FS. All FO performance experiments were evaluated under two operating modes: FO mode, where the active layer faced the FS, and pressure-retarded osmosis (PRO) mode, where the active layer faced the DS. Water flux was calculated by the weight changes of the DS recorded by a data-logging digital balance (AX4202ZH, Ohaus, USA). Meanwhile, the RSF was calculated from the change in concentration of the NaCl FS, as measured by a data-logging conductivity meter (DDSJ-308F, Rex, China). All tests were carried out at ambient temperature.

PVP was applied as a pore-foaming agent in the CEC substrate casting solution. Figure 2 summarizes the PWP values, contact angles, thicknesses, porosities, and average pore sizes of the CEC substrates with different PVP contents.

Figure 2(a) shows the thickness change of the CEC substrates with different PVP loadings. Although the gap height of casting applicator was 150 μm for all the CEC substrates, the thickness of the membranes was less than the gap height, ranging from 40 to 95 μm, after the phase inversion process. The thickness of the CEC substrates showed obvious growth with increasing PVP content from 0 to 4.0 wt.%. At higher PVP contents, a viscous casting solution with a low cloud point was formed. In other words, PVP increased the thermodynamic instability of the casting solution, which is considered to be a major contributor to the thickness of the phase inverted membrane (Han & Nam 2002).

The surface wetting characteristics of the substrate membranes were investigated by contact angle measurements. A lower value of contact angle indicates a more hydrophilic membrane surface. Figure 2(b) shows the change in the contact angle of the CEC substrates. Compared to the M0 substrate membrane, which had an average contact angle of 78.55°, the contact angles of the hydrophilic modified substrate membranes M0.5 to M4 declined from 72.82 to 64.37°. The result indicates that PVP improved the wettability of the CEC substrates because of improved porous top surfaces which is discussed in the following morphology study part.

Figure 2(c) illustrates the effect of PVP loading on the permeability of CEC substrates. The M0 substrate membrane, which contains no PVP, had a low PWP of 57.21 LMH/bar. On increasing the PVP loading to 2.0 wt.%, the PWP improved significantly to 362.22 LMH/bar. As the PVP loading increased, the water permeability of the CEC substrates improved markedly, which is ascribed to the PVP loading enhancing the porosity and hydrophilicity of the substrate membrane. However, on further increasing the PVP loading, the PWP rose only slightly from 362.22 LMH/bar for M2 to 431.83 LMH/bar for M4. The smaller increase in PWP on increasing the PVP loading could be a consequence of the continued thickening of the CEC substrates.

Figure 2(d) shows the porosity (ε) of the CEC substrates. Compared to the neat CEC membrane, PVP increased the porosity of the membranes. According to the experimental results, the value of ε increased from 73.48% for M0 to 83.40% for M2. However, the porosity barely increased when the PVP content increased from 2.0 to 4.0 wt.%. This phenomenon could be caused by two major factors. On the one hand, PVP, as a hydrophilic additive with non-solvent properties, could hasten the phase inversion and increase its intensity, resulting in the formation of more pores. On the other hand, excess PVP content could notably increase the viscosity of the casting solution, which could slow down the phase inversion rate and prevent the formation of pores (Han & Nam 2002; Anvari et al. 2016). Thus, the porosity surged when PVP content increased from 0 to 2.0 wt%, and stopped increasing when PVP content was over 2.0 wt%. For the same reason, a similar trend was observed in the average pore size (r_m) study of CEC substrates, which is shown in Figure 2(e). The value of r_m rose to 33.44 nm for M2 from 13.09 nm for M0 and increased to 39.26 nm for M4.

The mechanical properties are not crucial for the FO substrates because the FO process operates under normal pressures. However, the membrane must possess a reasonable mechanical strength for the practical use of the membrane in FO. The stress–strain curves of CEC substrates are shown in Figure 3. The stress–strain curves demonstrate that substrate M0 has the best tensile strength of the CEC substrates studied. The tensile strength declined obviously as the PVP content increased from 0 to 2%. The increasing porosity of the CEC substrates may be the cause of the reduction in strength. However, the tensile strength recovered slightly as the PVP content increased. There are two possible reasons for this result. First, the growth of the porosity slowed down when the PVP content was greater than 2%. Second, the thickness of CEC substrates increased with increasing PVP content.

Figure 4 shows the FESEM images of the cross sections, top surfaces, and bottom surfaces of the CEC substrates. According to the cross-section images, all CEC substrates exhibit typical asymmetric structures with finger-like pores formed under a thin sponge-like skin layer. However, the neat CEC substrate has a remarkably different morphology to those of the PVP-modified substrates. First, the finger-like pores of the neat CEC substrate membrane are stumpier, even drop-like in shape. Meanwhile, a two-sublayer structure was observed in the modified substrates: a typical sublayer containing finger-like pores in the top part and macrovoids and a sponge-like wall sublayer in the bottom part. This results in the modified membranes having a more porous structure, which corresponds to the experimental results concerning the CEC substrate porosity. Finally, these cross-section images show the changes in the thickness of the substrates. The membrane thickness increases as the PVP content increases, which matches the thickness measurement results. As shown in the top surface images, the top surfaces have a dense morphology. This originates from the diffusion of the solvent from the cast-substrate membrane into the non-solvent, which could result in the polymer concentration at the top surface being higher than that at the underside, thus resulting in a denser top surface during the phase inversion process (Sahebi et al. 2016). However, the FESEM images do not show a difference in the top surfaces with different PVP contents. In contrast, the images of the bottom surfaces demonstrate the remarkably different character of the bottom surface compared to the top surface. As shown in the images, submicrometer pores were distributed on the bottom surface randomly. Furthermore, the density distribution of the submicrometer pores increases significantly with increasing PVP content.

TFC-FO membranes were prepared via interfacial polymerization to synthesize an active layer on the CEC substrates. The resultant TFC-FO membranes are labeled TFC-M0 to TFC-M4.

Figure 5 shows the top surface images of the TFC-FO membranes. However, both FESEM images and subsequent FO performance tests confirm that the active layer of the TFC-M0 membrane was not tightly bonded to the substrate membrane. At a low magnification of 100× , even by eye, obvious exfoliation of the active layer could be seen. On the other hand, TFC-M0.5 to TFC-M4 exhibited the typical ‘ridge-and-valley’-like polyamide active layer at a magnification of 5,000×. Several previous investigations have tried to explain this phenomenon (Singh et al. 2006; Ghosh & Hoek 2009; Misdan et al. 2013; Liu & Ng 2015). First, according to the contact angle and porosity tests, the neat CEC membrane had poor hydrophilicity and wettability, which indicates that the neat CEC membrane cannot hold sufficient MPD aqueous solution to synthesize a tight active layer. Second, during the IP process, the MPD aqueous solution was kept inside the pores of the CEC substrates after the substrates were dried in the compressed nitrogen gas stream. In comparison with the neat CEC-substrate membrane, the diffusion of residual MPD from the pores of the hydrophilic blend substrates was more difficult during the polymerization reaction. Thus, the polyamide formed inside the pores for the CEC/PVP-blend CEC substrates. However, the polyamide should have grown on top of the pores. Hence, the polyamide active layer was combined more tightly to the CEC/PVP substrates than the neat CEC substrate. This hypothesis reveals why the active layer synthesized on the neat CEC substrate membrane separated easily. Furthermore, because less MPD aqueous solution diffuses from the pores of the substrates as the PVP content increases, a relatively thinner and less cross-linked polyamide active layer was formed. This phenomenon leads to higher flux and salt passage as the membrane PVP content increases.

Water permeability (A), salt permeability (B), and structural parameter (S) are the intrinsic parameters to characterize the FO membranes, and the results are tabulated in Table 2. In particular, because of the obvious exfoliation of the active layer, the TFC-M0 cannot separate solute of draw solution and will not exhibit FO performance (Suwaileh et al. 2018). Consequently, the intrinsic parameters and FO performance of TFC-M0 will not be further discussed. According to the results, the A value and B value significantly improved with increasing PVP content. The reason for the improved hydrophilicity of the CEC substrates is because the thinner and less cross-linked active layer led to higher water flux and salt passage (Misdan et al. 2013). The lower value of the B/A ratio indicates a higher efficiency in reducing the RSF in the FO process. The B/A ratio increased slowly when PVP concentration was below 2.0 wt%, and then climbed appreciably. The S value reflects the membrane ICP during the FO process and has a positive correlation with the ICP (Wang et al. 2015). The results show that the S value significantly declined to 922 μm for TFC-M2 from 1,727 μm for TFC-M0.5, then slightly decreasing to 784 μm for TFC-M4. Taking comprehensive consideration of the B/A ratio and S value, TFC-M2 membrane seems to be the best candidate for FO process, and this will be further discussed in the following FO performance section.

The water flux and reverse salt flux were taken as two parameters to evaluate the performance of TFC-FO membranes. In particular, the water flux and reverse salt flux were extremely unstable arising from the obvious exfoliation of the active layer of TCF-M0. Consequently, the FO performance of TFC-M0 will not be discussed. Figure 6(a) shows the water fluxes of the TFC-M0.5 to TFC-M4 membranes under both FO mode and PRO mode using 1 M NaCl as the DS and DI water as the FS. The water flux is positively correlated with the PVP content in both FO and PRO modes. As shown in Figure 6(a), water flux under FO mode increased to 9.09 LMH for TFC-M2 and then slightly increased to 11.20 LHM for TFC-M4 from 4.35 LMH for TFC-M0.5. A similar trend was observed in PRO mode. The water flux under PRO mode soared to 20.67 LMH for TFC-M2 and then increased to 25.19 LHM for TFC-M4 from 9.87 LMH for TFC-M0.5. Furthermore, the water flux in PRO mode was significantly better than that in FO mode. This difference in performance is attributed to a different type and degree of ICP. In general, a highly dilutive ICP would reduce the osmosis driving force across the FO membrane and decrease the water flux in FO mode, while a slightly concentrative ICP occurs in PRO mode (Liu & Ng 2015).

Not only a high water flux but also a low RSF is an essential parameter to evaluate the performance of FO membranes. The RSF values of the TFC-M0.5 to TFC-M4 membranes in both FO mode and PRO mode are shown in Figure 6(b). The RSF also increases with increasing PVP content, but its increasing trend is distinct from that of water flux. The RSF increased slightly to 1.35 gMH for TFC-M2 from 1.06 gMH for TFC-M0.5 and increased significantly, to 2.27 gMH, for TFC-M4. Likewise, a similar trend was observed in PRO mode. The RSF increased to 2.24 gMH for TFC-M2 from 1.79 gMH for TFC-M0.5 and rose to 4.65 gMH for TFC-M4. These results could be attributed to the thinner and less cross-linked active layer with higher PVP content, as discussed earlier. Furthermore, the high RSF could intensify the ICP, resulting in reduced water flux across the membranes (Sahebi et al. 2016). Hence, the increase in the water flux slows down with increasing PVP content.

The SRSF (ratio of RSF to water flux) has been identified as a metric to evaluate the osmotic process efficiency from the amount of draw solute loss per unit of water passed and to compare membrane performances for different membranes or draw solutes (Ren & McCutcheon 2015; Lim et al. 2017). High-performance FO membranes require high water flux and low RSF, i.e., a low SRSF value. Figure 7 shows the SRSF of the prepared TFC membranes in FO mode and PRO mode. The SRSF results show similar concave trends in both FO and PRO mode, reaching a minimum of 0.15 g/L in FO mode and 0.11 g/L in PRO mode at 2% PVP content. According to the earlier mentioned water flux and RSF data, the significantly increased water flux and slightly increased RSF at 0.5 to 2 wt.% PVP content contributed to the decline in SRSF, and the contrasting trends in water flux and RSF for 2 to 4 wt.% PVP content contributed to the increase in SRSF. The SRSF results are in good agreement with membrane intrinsic parameters' results. Taking comprehensive consideration of FO performance and membrane intrinsic parameters' results, TFC-M2 should be the optimal FO membrane.

Table 3 lists the FO performances of various lab-made flat-sheet TFC or TFN FO membranes and commercial HTI TFC FO membranes. Recently, PSF, PES, PVDF, and PAN have become common materials for the fabrication of substrates for TFC FO membranes (Widjojo et al. 2011; Ma et al. 2013; Tian et al. 2013; Shen et al. 2016). Because these polymers are hydrophobic, the membrane casting solution must be added to the pore-forming agent, such as PVP, LiCl, PEG, ethylene glycol, and ethanol (like P-1 to P-5), or blended with hydrophilic polymers and nanoparticles (like B-1 to B-4). According to the table, TFC-M2 and TFC-M4 have comparable levels of water flux and notably lower RSF compared to those of membranes P-1 to P-5 and the commercial HTI TFC membrane. Thus, all CEC-substrate TFC membranes exhibit better SRSF performance compared to the other TFC membranes.

In the study, CEC and PVP were chosen as the membranes' material and pore-forming agent to prepare substrates for TFC FO membranes, respectively. With increasing PVP content, the PWP, hydrophilicity, porosity, and average pore size of the substrates significantly improved. Visualization of the substrate morphologies showed that a two-sublayer substrate, a fringe-like top sublayer and a sponge-like bottom sublayer with macrovoids, was formed after the introduction of PVP. The optimization of the substrate properties improved the CEC membranes, making them more suitable TFC membrane substrates. However, the membrane thickness increased and the mechanical strength decreased after the addition of PVP. The surface images of the TFC membranes illustrate that the IP-formed polyamide active layer was exfoliated from the substrate containing no PVP. However, the FO membranes formed of PVP-modified substrates exhibited good FO performance. The TFC-M2 membrane exhibited a water flux of 9.10/20.67 LMH, 1.35/2.24 gMH RSF, and 0.15/0.11 g/L SRSF under FO/PRO mode while using 1 M NaCl as the DS and DI water as the FS. A comparison of FO performance with other FO membranes showed that the FO membranes with CEC substrates exhibited excellent SRSF performance. The results of this investigation clearly show that the PVP-optimized membranes can be used as substrates for TFC FO membranes.

Financial supports from the Ministry of Science and Technology of China for State Key Research and Development Project (2016YFC0400702), and the Department of Science and Technology of Guizhou Province (2017-5409) are gratefully acknowledged.