Activated carbon incorporation on forward osmosis membrane surface for enhanced performance

Cellulose triacetate (CTA) is the first-generation forward osmosis (FO) membrane used for desalination. There have been a few chemical modifications of the CTA membrane surface. This has improved membrane hydrophilicity, water flux, and salt rejection compared with unmodified CTA membranes. Chitosan-containing porous materials as composites have resulted in increased pore characteristics. It has motivated the modification of the surface of the commercial CTA forward osmosis (FO) membrane by surface coating with chitosan (CS)–powdered activated carbon (AC) mix. The membrane morphology was characterized by SEM, FTIR-ATR, contact angle measurement and AFM. Operational conditions for FO such as the orientation of the membrane active layer, feed and draw solution flow rates, and type and concentration of draw salt were optimized with the original CTA membrane. The modified membrane exhibited around a two-fold increase in the water flux and reduced reverse salt flux compared with the original CTA membrane. The improved water flux was attributed to the CSAC coating enhancing water wettability of the membrane surface and the porous AC generating additional water flow channels. Overall, the water flux of the CTA-CS-AC membrane developed in this work was superior to that of CTA and cellulose acetate (CA) membranes reported in the literature.

GRAPHICAL ABSTRACT INTRODUCTION related to the thickness and porosity of the support layer (Gao et al. 2014). Chemical modification approaches especially of the top active surface layer can be a way to decrease ECP (Fu et al. 2021). The first-generation FO cellulose triacetate (CTA) membrane commercially available from Hydration Technology Inc. (HTI), USA, has been extensively studied for desalination. However, it is characterized by the drawbacks of relatively low water permeability, low salt rejection, and poor resistance to biochemical degradation (Suwaileh et al. 2020). Recent studies have reported an improvement in the properties of CTA membrane by incorporating porous materials e.g., carbon nanotubes (CNTs) (Choi et al. 2015), iron nanoparticles (Chi et al. 2017), graphene oxide (GO) , and metal organic frameworks (Wang et al. 2019).
There are several studies on commercial RO and nanofiltration (NF) membranes coated with a variety of polymers to increase their active layer robustness and antifouling properties (Choudhury et al. 2018;Sharma et al. 2021). This is generally achieved by applying neutral or hydrophilic polymers to the existing membrane active layer thereby decreasing the surface roughness and minimizing the surface charge. The method of active layer modification of FO membranes employed by HTI is proprietary. Other similar studies on CTA FO membranes are limited, e.g., CTA FO membrane was modified via interfacial polymerization of m-phenylene diamine (MPD) trimesoyl chloride (TMC) (Wu et al. 2018). The study shows that a hydrophilic CTA porous membrane of thickness 102 + 16 μm and porosity 87% improves the water flux and salt rejection. In another study, dip-coating with polyvinyl alcohol (PVA) resulted in a 20% increase in water flux compared with the unmodified membrane and without any salt leakage (Ahn et al. 2015). Commercial CTA membranes were modified by plasma-grafting polymerization, which resulted in the improvement of water flux and anti-protein-fouling properties of an FO membrane (Khongnakorn et al. 2020); gamma-ray irradiation of commercial CTA membrane improved membrane surface hydrophilicity, and it showed increased water flux and retention of Co and Sr (Liu et al. 2020). Coating of the membrane surface results in an additional layer to achieve the desired performance. Membrane surface coating enhances surface charge and hydrophilicity and decreases surface roughness (Miller et al. 2017). But, it also increases the membrane resistance to water flux (Liu et al. 2021). Thus, it is essential to look for a method that will not decrease the desirable properties of the membrane. Table 1 shows that the addition of porous materials (activated carbon, biochar, carbon nanotubes, and graphene oxide) in CS improves the surface area, total pore volume and average pore size of the composite compared with CS (Hydari et al. 2012;Nitayaphat & Jintakosol 2015;Debnath et al. 2017;Khakpour & Tahermansouri 2018;Banu et al. 2019). Recently, tannic-acid-ferric (TA-Fe) complex coated on top of a commercial CTA-FO membrane purchased from Fluid Technology Solutions, Inc. (FTS) (Albany, OR, USA), showed enhanced water flux, improved antifouling property, and micropollutant removal (Liu et al. 2021). Similar improvement in membrane performance was reported with a polydopamine (PDA) coating on a commercial thin-film composite (TFC) FO membrane obtained from Hydration Technology Innovations (HTI) (Guo et al. 2018).
The present study aims to improve the properties of commercial CTA membrane by modifying its active surface using hydrophilic chitosan (CS), an organic biopolymer and porous powdered activated carbon (AC). CS, an abundant biomaterial obtained from crustacean seafood, is inexpensive, non-toxic, biodegradable, and environmentally friendly. Its molecular structure contains a good number of amino (-NH 2 ) and hydroxyl (-OH) functional groups which can enhance the wettability of AC (Shanmugam et al. 2016). Powdered AC is characterized by small particle size (,100 μm), and a high degree of microporosity and surface area (800-1,000 m 2 /g). However, the behaviour of CS-AC coated CTA membrane is unknown.
In this study, commercial CTA membranes modified with CS-AC mix were prepared and characterized. Operation conditions for FO such as flow rate, membrane orientation, draw solution concentration and types of draw salts were initially optimized with the original CTA membrane and the modified membranes' performance tested under optimal conditions. The performance of the membranes developed in this work was also compared with reported literature on commercial CTA and cellulose acetate (CA) FO membranes.

Chemicals and reagents
CTA FO membrane was purchased from Tech Inc. Incubating Technologies (Tech Inc.), Chennai, India. As per the supplier specifications, the 180-200 μm thick membrane was stable at pH 4-8, operating temperature of 20-50°C and had a chlorine tolerance of 0.1 g/L. It was characterized by a pure water flux (J w ) of 15-20 L/m 2 h (with 2 M NaCl as draw solution) and reverse salt flux (J s ) of less than 0.20 g/m 2 h. The membranes were rinsed repeatedly with deionized (DI) water and stored at 4°C before use. Chemicals and reagents were purchased from various local suppliers: Loba Chemie Pvt Ltd (NaCl, MgCl 2-·6H 2 O); Merck Life Science Pvt Ltd (Na 2 SO 4 , H 2 SO 4 , glutaraldehyde 25% aqueous solution, glacial acetic acid, CS flakes from shrimp shells 75% deacetylated); Himedia Lab Pvt Ltd (powdered activated carbon); Merck Specialities Pvt Ltd (MgSO 4 ·7H 2 O). All chemicals were used as received without further purification.

Preparation of CTA-CS-AC FO membranes
The CS-AC mix was prepared at room temperature as follows. CS (1.0 g) was dissolved in glacial acetic acid (2 g) and DI water (100 mL) for 8 h. Then AC (0, 0.05, 0.1, 0.2 w/v %) was added in the CS solution and the mixture stirred using a magnetic stirrer (REMI 2ML, India) for one day. Thereafter, the solution was kept idle for another one day before being used for membrane modification. The CTA membrane was fixed in a rectangular wooden frame and the CS-AC mix was poured on top of it to coat the surface. The ensemble was kept horizontal for 3 min and then tilted to an angle of 60°to drain off the excess solution. The coated membrane was removed from the frame and kept for 1 h in the cross-linking solution (mixture of 25 mL of 25% glutaraldehyde solution, 0.11 g Na 2 SO 4 and 0.2 mL conc. H 2 SO 4 , made-up to 250 mL with DI water). The modified membrane was then repeatedly washed five times to remove any unreacted CS. The membrane samples were kept soaked in water and tested on the same day. Figure 1 shows the schematic representation of the chemical reaction between CS and AC and the synthesis of the CS-AC mix.

FO operations
The schematic representation of the experimental set-up is shown in Figure 2. The FO laboratory-scale cross-flow filtration unit used in this study was purchased from Tech Inc., Chennai, India. This test cell had an effective filtration area of 0.016275 m 2 (0.105 m Â 0.155 m). The experiments were performed at 20, 40, 60 LPH (L/h) flow rate of feed solution (FS) and draw solution (DS) pumped in a counter-current flow arrangement using two peristaltic pumps (Rivotek 50171, Mumbai, India). Flow rate was controlled by adjusting the valve settings and was measured using in-line rotameters (CVG Technocrafts, Mumbai, India) on both FS and DS sides. Two overhead stirrers (Remi RQ 5, Elektrotechnick Ltd, India, operated continuously at 400 rpm) in the feed and draw containers were used to ensure solution homogeneity. A digital balance (A&D Company Ltd FX 5000 GD, Delhi, India) connected to a digital logging system (WinCT software) was used on the feed corrected Proof side to monitor the change in the weight of the FS every 5 min. The change in weight of the FS was converted to water flux (J w , L/m 2 h or LMH) using Equation (1): where W i is the initial weight (g) of the FS, W t is the weight (g) of the FS at time t, A is the membrane cross-section area (m 2 ), and t is the filtration time (h). All experiments were conducted at an ambient temperature of 25 + 2°C. Change in the conductivity of the FS was recorded using a conductivity meter (Labindia, Mumbai, India). This data was used to calculate the reverse salt flux (J s , g/m 2 h or GMH) as given in Equation (2): where V 0 and V t are volume (L) of feed, C 0 and C t are the salt concentration (mg/L) of FS; the subscripts 0 and t (h) are time at 0 and t. From the J w and J s values obtained from Equations (1) and (2), the specific salt ratio was calculated by Equation (3):

Experimental methodology and performance analysis
The wet membrane was mounted in the FO cell and was initially flushed with DI water for an hour to ensure stable permeate flux. Then 1 L FS (DI water) and 1 L DS (NaCl, MgCl 2 ·6H 2 O, MgSO 4 ·7H 2 O or Na 2 SO 4 ) was circulated on the feed and draw sides, respectively. The draw salts were selected for high osmotic pressure (MgCl 2 ), moderate osmotic pressure (NaCl) and low osmotic pressure (Na 2 SO 4 , MgSO 4 ) . Each set of experiments, for optimizing a particular parameter, was done in triplicate and a fresh FO membrane was used for each set. Between the replicates, the membrane was cleaned by recirculating 0.5 L deionized water on both sides of the membrane for 30 min followed by flushing under tap water for 10 min. The water recovery (F r ) was calculated by Equation (4): where ΔV is the difference in initial and final volume of FS, and V 0 is the initial volume. The performance of the original CTA membrane was evaluated at different flow rates, active layer (AL) orientation, concentrations of DS and draw salts. The optimal operation conditions thus identified were then used to test the modified membrane samples. In all FO experiments, the water flux (J w ), reverse salt flux (J s ) and water recovery (F r ) was determined. The membranes were tested in FO mode (AL facing FS) and PRO mode (pressure retarded osmosis; AL facing DS). Water flux was measured in both the studies.

Membrane characterization
Fourier-transform infrared-attenuated total reflection (FTIR-ATR) spectroscopy (IR-Prestige 21, Shimadzu Corp., Japan) was used to analyze the membrane functional groups. Scanning electron microscopy (SEM, JSM-6390LV, Jeol, Japan) was used to investigate the morphologies of the membranes and AC. The membrane samples were prepared by freeze-fracturing under liquid nitrogen; both membrane and AC samples were coated with Au in an Ar atmosphere using a vacuum evaporator before SEM analysis. Atomic force microscopy (AFM, Solver Pro-47, NT-MDT, Russia) was used to evaluate the surface roughness of the membranes. The BET values and pore structure of AC was obtained from nitrogen adsorption isotherms at 77 K using a surface area analyzer (Quantachrome Instruments, USA). The water contact angle was measured by the sessile drop method using the set-up from Tech Inc., Chennai, India. Figure 3 shows the spectrum of the CTA (Figure 3(a)) and CTA-CS-AC (1% CS-0.2% AC) (Figure 3(b)) membranes. The amino and the hydroxyl groups are important functional groups of CS structure. The wide band detected in the 3,500-3,100 cm À1 region is ascribed to the O-H bond stretching modes. The peaks at 1,380 cm À1 indicate the -NH vibration in -NH 2 . It also represents the interaction between the NH 2 groups in CS and OH groups of the CTA. The characteristic adsorption bands are located around 1,500 cm À1 and 1,755 cm À1 , due to stretching vibrations of the carbonyl group. The bands at 1,219 cm À1 and 1,055 cm À1 correspond to the stretching of C-O single bonds. The weak bands at 2,935 cm À1 and 2,888 cm À1 are attributed to C-H bonds. After modification of the membrane (Figure 3(b)), the carbonaceous content of AC shows intense bands between 3,200 cm À1 and 3,500 cm À1 (O-H and N-H stretching vibrations) (Amara et al. 2009;Queiroz et al. 2015;Shu et al. 2017). The peaks at 1,600 cm À1 and 1,420 cm À1 show stretching vibrations of C ¼ O in the carbonyl group and deformation of C-O and O-H in carboxylic acid, respectively (Figure 1) (Nowruzi et al. 2020). The presence of the functional groups improves the membrane surface hydrophilicity, which eventually improves the membrane selectivity (Shakeri et al. 2017). The surface roughness of the membrane was studied by AFM (Figure 3(c) and 3(d)) and the roughness parameters calculated based on the AFM scanning area of 10 μm Â 10 μm. The dark areas show the pores of the membrane, and the bright areas show the highest sites of the membrane surface. The R a value of the CTA membrane is 15.45 nm, while the R a value of the CTA-CS-AC (1% CS-0.2% AC) is 20.99 nm. The increased R a value may result in improved water flux, followed by an increase in the area available for water transport (Hirose et al. 1996;Rana & Matsuura 2010). The SEM image of the CTA membrane surface (Figure 3(e)) shows a dense, smooth surface without any cracks. In the modified CTA-CS-AC (1% CS-0.2% AC) membrane (Figure 3(f)), the presence of powdered AC (Figure 3(g)) scattered on the surface is observed. The modified membrane surface seems to be non-uniform and irregular, as evident from the increased surface roughness value. It is interesting to note that the AC are completely wrapped in the CS matrix, and the AC are not visible in Figure 3(f), even at higher magnification, compared with Figure 3(g).

Performance of CTA membrane
The FO operational conditions were optimized with the original CTA membrane. Figure 4 shows the water flux and specific salt ratio of the original CTA membrane at different flow velocities (20 LPH, 40 LPH and 60 LPH) with 1M NaCl as DS. The active layer (AL) was facing the FS. Increasing the cross-flow rate from 20 LPH to 60 LPH results in marginal increase in water flux (from 8.27 LMH to 8.56 LMH) with a corresponding increase in the specific salt ratio (from 1.55 g/L to 1.70 g/L). This indicates that the water flux and specific salt ratio for this membrane are independent of the flow rate in the range tested, and there are no defects on the membrane separation layer. High flow velocities intensify the scouring on the membrane surface thereby lowering the chances of ECP, which is an undesirable phenomenon contributing to poor water flux (Qasim et al. 2015). Further, high flow rates of FS and DS enhance the mass transfer which, in turn, improves  Table 2 shows a comparative study of the performance of commercial CTA membranes on changing the orientation of the membrane active layer (AL) towards FS (FO mode) and DS (PRO mode), respectively. The study shows that the water flux in PRO mode is more than in FO mode. It could be due to the internal concentration polarization (ICP). ICP occurs in a FO process when the water flows in the opposite direction to the solute flux. This can lead to either (i) a concentrative ICP, or (ii) a dilutive ICP. When the active layer is facing the draw solution (AL-DS), the solutes from the FS accumulate in the porous support layer due to their rejection by the active separation layer. This is called concentrative ICP. On the other hand, when the active layer is facing the feed solution (AL-FS), it results in dilution of the draw solution inside the support layer, and this is called dilutive ICP. However, in both cases, the effective osmotic pressure is reduced, decreasing the water flux (McCutcheon & Elimelech 2006;Tang et al. 2010). In PRO mode, the foulants deposited in the porous support result in membrane fouling and reduce the membrane cleaning efficiency (Qi et al. 2012). In the present study, in AL-FS (DI water) no ICP occurs. As it operates under low hydraulic pressure, no solute accumulates at the AL and the effect of ECP on the water flux is insignificant. Thus, all further FO experiments were conducted with AL oriented towards FS.
The effect of varying DS concentrations (0.5, 1.0, and 2.0 M NaCl) is shown in Figure 5. As the concentration of DS increases from 0.5 M to 2.0 M, the corresponding osmotic pressure of the DS increases from 24.6 atm to 98.4 atm. The increased differences in the pressure gradient across the membrane result in a rise in water flux from 4.33 LMH to 13.27 LMH, and a decrease in specific salt ratio from 1.8 to 1.18. The increase in water flux with an increasing osmotic pressure of the DS is reported in other works (Achilli et al. 2010;Liu et al. 2021). The specific salt ratio indicates the selectivity of the FO membrane. The decrease in specific salt ratio with increased DS concentration indicates low draw solute migration during the FO process and a highly efficient FO membrane (Fu et al. 2021).
In general, a DS with high osmotic pressure can generate high water flux (Achilli et al. 2010). Typically, selection of a suitable DS resulting in (i) high water flux, (ii) low reverse salt diffusion and (iii) easy recovery of the diluted DS is important for attaining high FO system performance (Chekli et al. 2012). Figure 6 shows the FO performance with different types of DS   132) is also lower than that of NaCl (1.537). The difference in water flux with different DS is due to ICP effects that reduce the effective osmotic pressure difference across the membrane (Achilli et al. 2010). The variation could also be due to the variation in the hydrated diameter of the anion (Cl À : 300 Â 10 À12 m, SO 4 2À : 400 Â 10 À12 m) and cation (Mg þ2 : 300 Â 10 À12 m, Na þ : 450 Â 10 À12 m) and its interaction with the negatively charged CTA membrane (Achilli et al. 2010;Phillip et al. 2010). In a similar attempt to identify suitable DS for desalination, different water flux and salt flux were reported for different salts at the same osmotic pressure but at different DS concentrations (Achilli et al. 2010). Although MgCl 2 ·6H 2 O outperforms the other DS in this study, the next best option of NaCl was chosen for further experiments because it is cheaper (9 $/kg) than MgCl 2 ·6H 2 O (26 $/kg) (using a conversion of 1$ ¼ 73.51 INR).

Performance of CTA-CS-AC membranes
The properties of the powdered AC used in modifying the CTA membrane surface are summarized in Table 3. In this work, the active layer of the commercial CTA membrane was modified with CS containing different concentrations of powdered AC. A positron annihilation lifetime spectroscopy (PALS) study shows that the active layer of CTA membranes is composed of uniform pores of size 0.29-0.30 nm (Kim et al. 2017). Further, Table 1 shows that the pore diameter of CS-AC increases to 7.32 nm. Thus, the addition of AC increases the size of the flow channels of the separation layers. In a separate study, we found that coating the CTA membrane with 1% CS produces higher water flux than coating with 0.5% or 2% CS ( Supplementary Information, Table S1). Figure 7 shows the effect of varying AC loading on the performance of CTA-CS-AC membrane. The membrane with 0.1% AC loading (M-IV) displayed maximum water flux of 24.6 LMH compared with 13.2 LMH for the original membrane. Correspondingly the specific salt ratio decreased with AC loading, from 1.05 (original membrane) to 0.9 (0.1% AC). Surface modification decreased the membrane contact angle (Table 4) and improved the water wettability. The contact angle decrease was significant from 73.9°(original membrane) to around 10°(M-IV and M-V  Water Supply Vol 00 No 0, 9 corrected Proof membranes with 0.1% and 0.2% AC loading respectively). Further, the high surface area and the well-defined porous structure of AC (Table 3) creates new water channels within the membrane surface resulting in higher water flux. The pore diameter of 2.37 nm readily allows passage of water molecules of diameter 0.275 nm while restricting salt migration from DS to FS. This observation is also supported by the increased R a value of the CTA-CS-AC (0.2%) membrane (Figure 3(c)). However, there is a decrease in water flux with increased AC content (0.2%), possibly due to the formation of a denser CS-AC matrix. Figure 8 displays the water recovery for the various membranes. The water recovery of the original CTA membrane was 30%, which increased marginally to 33% upon surface modification with CS. Incorporation of AC in CS resulted in further increase in water recovery up to 78% (M-IV with 0.1% AC loading). The increase in water recovery of over two-fold is due to 1,032 g/L 0.57 cm 3 /g 990 m 2 /g 1,183 m 2 /g 2.37 nm    Water Supply Vol 00 No 0, 10 corrected Proof the increased hydrophilicity and the presence of porous AC which generates well-defined water channels within the membrane surface layer, resulting in high water flux and reduced ECP. Secondly, the hydrophilic properties are desirable in FO membranes as they allow water molecules to readily pass through the internal fine pores of the membranes, improving overall performance. The hydrophilic CS polymer matrix has good compatibility with the AC. Finally, the increased surface roughness provides a higher filtration area for water transportation. These characteristics are expected to result in high stability of the coating during long-term operations. Unlike this work that targeted the addition of a porous material layer on top of the CTA membrane surface, other literature studies involve the inclusion of porous materials in the CTA-FO membrane matrix itself. Particle content in the polymer matrix reportedly enhanced the water flux while reducing the specific salt ratio and ICP. For example, boehmite particles added to a cellulose acetate/cellulose triacetate mixture augmented the water flux three-fold (up to 21 LMH) (Zirehpour et al. 2015). Inclusion of acid-functionalized CNTs in cellulose acetate resulted in a 1.5 times increase in water flux (up to 14 LMH) (Choi et al. 2015). Metal organic framework (MIL-53) in cellulose acetate enhanced water flux 1.3 times (up to 34.9 LMH) (Wang et al. 2019). The results are comparable to that obtained in the present study with a water flux enhancement of 1.85 times (up to 24.6 LMH) with activated carbon incorporation on the CTA membrane surface. The present membrane showed better performance than polyvinyl alcohol/polydopamine-coated zeolitic imidazolate framework (PVA/PDA@ZIF-8)-coated polyether sulfone (PES) substrate (J w 13 LMH) FO (Fu et al. 2021). Furthermore, the dip-coating procedure to modify the membrane surface used in this work is simple and replicable with other FO membranes; it can also be easily reproduced in the membrane manufacturing process. Wu et al. (2018) reported that increased CTA membrane thickness on coating with MPD-TMC solution affects the performance of the membrane. The study reported that the decreased membrane thickness (198 to 98 μm) improved the water flux by two times in FO and PRO mode. The best result was obtained with a thickness of 98 + 15 μm. Figure 9 shows the SEM image of the modified membrane. The membrane shows an asymmetric structure with a support layer and a separation layer. The total thickness of the modified membrane was 90 + 3 μm. Thus, appropriate membrane thickness and the improved surface properties (Table 1), with porous AC, resulted in the high water flux of the modified membranes. Figure 10 shows the benefit of surface modification of the commercial FO membrane to reduce the effect of ICP. The modified membrane (M-IV) water flux and the three commercial CTA membranes were studied for both FO and PRO modes. The increased ratio of water flux (FO/PRO) of M-IV (94.6%) in comparison with the commercial CTA membranes (52% + 2%) indicates that the modified membrane has a low degree of ICP Wang et al. 2016). Table 5 presents a compilation of CA and CTA membranes tested for desalination. The comparison shows that the surface modification of the commercial CTA membrane by incorporating porous AC largely enhances the water flux compared with

CONCLUSIONS
Under the conditions of this study, the following optimal conditions for FO with commercial CTA membranes were identified: (i) membrane active layer oriented towards feed solution; (ii) feed and draw solution flow rate of 40 LPH; and  corrected Proof (iii) draw of NaCl at 2M concentration. The performance, in terms of higher water flux and lower specific salt ratio, can be improved by membrane surface modification. This was successfully accomplished by coating the CTA membrane with a mix of chitosan and powdered activated carbon. CTA coated with 1% chitosan loaded with 0.1% activated carbon exhibited the best performance. Surface modification of the membrane surface with a porous coating is thus a simple and promising approach to improve FO membrane performance.