Double-blade casting technique for optimizing substrate membrane in thin-film composite forward osmosis membrane fabrication

Journal of Membrane Science 469 (2014) 112–126

Contents lists available at ScienceDirect

Journal of Membrane Science journal homepage: www.elsevier.com/locate/memsci

Double-blade casting technique for optimizing substrate membrane in thin-film composite forward osmosis membrane fabrication Xiao Liu, How Yong Ng n Centre for Water Research, Department of Civil & Environmental Engineering, National University of Singapore, 1 Engineering Drive 2, E1A-07-03, Singapore 117576, Singapore

art ic l e i nf o

a b s t r a c t

Article history: Received 26 March 2014 Received in revised form 8 June 2014 Accepted 18 June 2014 Available online 24 June 2014

The thin-film composite (TFC) methodology has been explored in the last couple of years to fabricate forward osmosis (FO) membranes. Different tactics have been proposed to mitigate the internal concentration polarization (ICP) phenomenon in the substrate membrane layer. However, such modifications on the substrate membranes would likely alter their top surface morphology and this in turn, would profoundly influence the subsequent process of interfacial polymerization (IP) to form the active layer. In the current work, we presented a facile substrate membrane fabrication strategy – double-blade casting technique, to produce substrate membranes with enhanced structural features to mitigate ICP, yet retaining an ideal top surface for the formation of an intact and highly salt-rejecting active layer. A series of standard protocols have been utilized to characterize the substrate membranes and resultant TFC-FO membranes. Overall the resultant TFC-FO membranes based on the double-blade casted polysulfone substrate membranes showed improvement in water flux, J v with reduced apparent structural parameter, S values and retained a relatively low reverse salt flux/water flux, J s =J v ratio. With a 1 M NaCl draw solution and DI water feed, the best TFC-FO membranes achieved a J v of 31.1 LMH and a J s of 8.5 gMH in the FO orientation, and a J v of 60.3 LMH and a J s of 17.6 gMH in the PRO orientation. This concept demonstration study may open up many new platforms for flat-sheet substrate membrane fabrication for FO membranes. & 2014 Elsevier B.V. All rights reserved.

Keywords: Forward osmosis Thin-film composite membrane Double-blade casting Reduced internal concentration polarization Polyamide layer integrity

1. Introduction Forward osmosis (FO) is a remarkable concept that offers a viable prospect as the next generation water purification and seawater desalination technology [1,2]. FO operates based on the osmotic pressure differential across a semi-permeable membrane, rather than hydraulic pressure differential such as in the reverse osmosis (RO) and nano-filtration (NF) process, to extract clean water from the feed into the draw solution [3]. Compared to the traditional pressure driven processes, FO process displays several merits worthy of being further explored and capitalized on: (1) lower energy consumption and equipment costs and higher water recovery [1–3]; (2) less propensity towards fouling [4,5]; and (3) wide-ranging applications, such as power generation [6–9], juice or food concentration [10], and essential products and pharmaceuticals enrichment [11,12]. Nevertheless, FO process does have its inherent drawbacks as its performance is highly sensitive to osmotic pressure of the draw solution and concentration polarization (CP) [13–17]. Concentration

n

Corresponding author. Tel.: þ 65 65164777; fax: þ65 67791635. E-mail address: [email protected] (H.Y. Ng).

http://dx.doi.org/10.1016/j.memsci.2014.06.037 0376-7388/& 2014 Elsevier B.V. All rights reserved.

polarization at the membrane surface (external concentration polarization or ECP) or within the substrate membrane layer (internal concentration polarization or ICP) has been considered as the main reason for the lower-than-expected osmotic efficiency and thus, the poor water flux observed in most reported osmotic experiments. Especially ICP, resulted from the dilution of the draw solution (dilutive ICP) or the concentration of the feed (concentrative ICP) within the porous substrate membrane layer, has been proven to be the most aggravating factor in undermining the osmotic driving force across the membrane [5,18], since it cannot be effectively mitigated with changes in process hydrodynamics [13,14,19]. As a result, FO still remains in a status of laboratory study due to these critical technical challenges. In order for FO to be applied in full-scale application, more conceited effort must be done in sourcing for: (1) suitable draw solutions that exert high osmotic pressure, can be recycled effectively with low energy consumption and display minimal reverse solute fluxes; and (2) high performance membranes with properties deliberated specially for osmotic processes, such as desirable characteristics in reducing the effect of ICP [1–3,14]. Numerous developments of membranes catered to FO processes have been carried out in the forms of asymmetric flat sheet or hollow fiber membranes, mostly involving phase inversion

X. Liu, H.Y. Ng / Journal of Membrane Science 469 (2014) 112–126

technique [3]. The current benchmark flat-sheet asymmetric FO membranes available in the market are made of cellulose triacetate (CTA). Although these membranes have been tested for broad applications, they have relatively low water permeability and salt rejection, particularly for application in seawater desalination [20]. The next generation of FO membranes was motivated by the thinfilm composite (TFC) approach originally industrialized for RO membranes production and can be found in the form of both flatsheets and hollow fibers. These membranes are typically fabricated in a two-step process: (1) a porous substrate membrane is formed by phase inversion technique with special features to enhance solute mass transfer and thus to reduce ICP and a second step where (2) a thin polyamide (PA) active layer is prepared by interfacial polymerization (IP) of amine and acid chloride on the substrate membrane surface [18,20–25]. For flat-sheet TFC-FO membranes, they generally consist of three layers: (1) a thin PA active layer; (2) a porous polymeric sub-layer as the intermediate; and (3) a woven or non-woven fabric support backing [18,20,26– 28]. These TFC membranes are preferred over integral asymmetric membranes formed directly from phase inversion because apparently they possess superior separation properties and allow independent optimization of the active layer and substrate membrane layer [23,28,29]. It has been generally acknowledged among the FO membrane researchers that high performance TFC-FO membranes should possess the following characteristics [18,20,22,26,30]: (1) an active layer with high permeate water flux and low solute permeability; (2) a substrate membrane layer which is thin, highly porous/ permeable and hydrophilic to reduce ICP and boost water/solutes transport; and (3) an overall structure capable of withstanding the mechanical stresses generated by the operating conditions. The substrate membrane layer is of particular importance because it not only provides mechanical strength and flow pathways, but also governs the extent of ICP and hence the effective osmotic gradient for water flux in the final TFC-FO membranes [25]. In order to capitalize on the osmotic driving force more effectively, the improvement of substrate membrane plays a more significant role in TFC-FO membrane fabrication [18,26,27,31,32]. Although alternative fabricating methodologies have been proposed to create substrate membrane layer that reduces ICP [33,34], phase inversion techniques still remain as the conventional and most widely accepted and practiced techniques in the industry. Thus a bulk of current research for FO substrate membranes focuses on the optimization of membrane material property or the conditions during phase inversion process [2,18]. Although TFC-FO membranes allow independent optimization of the active layer and substrate membrane layer as mentioned earlier, the modifications on the substrate membrane would inevitably alter the substrate membrane top surface morphology and may inadvertently undermine the integrity of PA active layer developed during the subsequent IP process [22,23,31,34]. In other words, their individual optimizations cannot be totally de-coupled and a paradox exists where any amendment aiming at improving the substrate membrane structure may potentially compromise the PA active layer integrity. Consequently, it is imperative to find new methods to assure that structural changes to the substrate membrane, which are essential in minimizing ICP, must not compromise its critical role as a suitable surface for the formation of an intact active layer during IP process [33]. Co-casting technique is an effective one-step membrane casting technique which can realize a multilayered configuration in the membrane [35–37]. Pereira et al. has firstly demonstrated the application of double-blade in fabricating flat-sheet membranes with two different polymer solutions [37]. More recently, Hashemifard et al. [38] also presented a study demonstrating the co-casting technique with a dual-blade casting knife to produce

113

dual-layered flat-sheet membranes for gas separation. This simple fabrication strategy can be innovatively adapted in conventional flat-sheet substrate membrane fabrication (with non-woven mesh backing) and may provide an alternative method to produce a more permeable substrate membrane, yet retain its surface morphology suitable for PA active layer deposition. This could eventually bring about an enhancement of osmotic performance for the resultant TFC-FO membranes by maintaining the integrity of the PA active layer while reducing the ICP phenomenon in the substrate membrane. Earlier works have demonstrated that reducing the polymer concentration of casting solution may reduce the substrate membrane thickness and increase its porosity, ensuing an enhancement of water flux in resultant TFC-FO membranes during osmotic performance testing [26,39]. The current study will further investigate the effect of casting solution polymer concentration on substrate membrane surface morphology and performance of resultant TFC-FO membrane in terms of both water flux and reverse salt flux. More importantly, the current study shall demonstrate, with the polymer concentration in the casting solution as the varying parameter, the concept of double-blade casting technique as an innovative adaptation to conventional casting procedure. The protocol strives to produce flat-sheet substrate membrane especially tailored for synthesis of TFC-FO membranes. The outcome of current work aims to validate, with membrane characterization and performance evaluation, that the resultant TFC-FO membranes fabricated based on the doubleblade casted substrate membranes are able to display reduced ICP with enhancement in osmotic water flux yet not retreating on reverse salt flux. To the best of our knowledge, it is the first time that the double-blade casting technique has been employed to fabricate flat-sheet substrate membranes for TFC-FO membranes.

2. Materials and methods 2.1. Chemicals and membrane materials Unless otherwise specified, all chemicals were of analytical grade with purity over 99% and were used as received. Chemicals used for membrane substrate formation included Polysulfone (Psf) beads (Mn: 22,000 Da, Sigma-Aldrich Pte. Ltd, Singapore) and N,Ndimethylformamide (DMF, anhydrous, 99.8%, Fisher Scientific, Aik Moh Pt. Ltd Singapore). Chemicals used for interfacial polymerization included M-phenylenediamine (MPD, 499%, ACROS, Aik Moh Pte. Ltd, Singapore), Trimesoyl Chloride (TMC, 98%, Sigma-Aldrich Pte. Ltd, Singapore) and n-Hexane (Fisher Scientific, Aik Moh Pte. Ltd, Singapore). Polyethylene oxides with molecular weight ranging from 200,000 to 1,000,000 g mol  1 (PEO, Sigma-Aldrich Pte. Ltd, and Singapore) were used as solutes in the single-solute rejection test. Ultrapure deionized (DI) water was supplied from a Milli-Q ultrapure water system (Millipore Singapore Pte Ltd) with a resistivity of 18.2 MΩ cm. Sodium chloride (NaCl, ACS, Aik Moh Pte. Ltd, Singapore) was used as the draw solutes in osmosis performance tests and salt permeability tests. A commercial polyester mesh (PET, 90 μm (T)  1002 μm (S), see Fig. 2) was used as a backing layer for the Psf substrate membrane. Commercial FO membranes were cut from a Hydrowell module procured from Hydration Technologies Inc (HTI, US) and were used for comparison purpose. 2.2. Synthesis of flat-sheet TFC-FO membranes The TFC-FO membranes were prepared via two steps: (1) a phase inversion step to form the substrate membrane; and (2) an interfacial polymerization step to form the polyamide active layer.

114

X. Liu, H.Y. Ng / Journal of Membrane Science 469 (2014) 112–126

Fig. 1. Conceptual drawing of substrate membranes casting using (a) single-blade casting knife and (b) double-blade casting knife.

2.2.1. Casting of polysulfone substrate membrane To prepare a casting solution for substrate membrane, Psf beads were dried firstly in air circulation oven (Lang Shang LC108, Nan Tong Hu Nan Ltd, China) at 110 1C for 12 h. A certain amount of dried Psf beads was dissolved in DMF and well-mixed using an orbital shaker (MRC, Inov Solutions Pte Ltd, Singapore) for at least 24 h at 25 1C until the solution became homogeneous and transparent. The solution was stored in a desiccator for at least 24 h before use. The PET fabric was attached to a clean glass plate using laboratory adhesive tape. For single-blade casting, a casting knife (RK Print Coat Instruments Ltd, UK), set at the gate height of 90 μm above the polyester mesh top surface, was used to spread the casting solution evenly onto the PET fabric to form a uniform film. The polysulfone substrate membranes fabricated using singleblade casting knife were designated as S#. For double-blade casting, the casting knife was modified to include a second blade (blade 2) in front of the original blade (blade 1). Blade 1 was maintained at the same gate height of 90 μm to spread casting solution A, while the second blade had a zero clearance and spread casting solution B into the open spacing of PET mesh. The polysulfone substrate membranes fabricated using double-blade casting knife were designated as D#. The compositions of S# and D# substrate membranes are listed in Table 1. The details of substrate membrane casting using both single-blade and doubleblade casting knife are illustrated in Fig. 1. After casting, the film was then quickly and smoothly immersed with the glass plate into a coagulant bath where 23 1C tap water was used to initiate the phase separation. The substrate membranes were allowed to sit in the coagulation bath for 15 min and were then transferred into a

flowing water bath for 12 h to remove residual solvent and subsequently stored in DI water before use. 2.2.2. Interfacial polymerization of polyamide active layer The active layer of TFC-FO membrane was prepared by interfacial polymerization on the surface of a Psf substrate membrane. The preparation was carried out at ambient temperature of 24 1C. The pre-casted Psf substrate membrane was firstly soaked in an aqueous solution of MPD (4.0 wt %) for 150 s, and the excessive MPD solution on the substrate surface was carefully removed with an air knife supplied with pure compressed nitrogen gas. Subsequently, an n-hexane solution of TMC (0.1 wt%) was gently poured onto the MPD-soaked substrate and was allowed to react with the residual MPD for 60 s to form the polyamide (PA) rejection layer. Excess TMC solution was then drained off. The resultant TFC composite membrane was air-cured for 20 min, rinsed thoroughly with tap water and then DI water to remove the residual monomers and then stored in DI water before performance evaluation. The TFC-FO membranes formed from S# substrate membranes were designated as TFC-S#; the TFC-FO membranes formed from D# substrate membranes were designated as TFC-D#. 2.3. Substrate membrane characterization 2.3.1. Thickness measurement The thickness of substrate membrane was measured using a digital microscope (Keyence VHX-500) with built-in measurement software. Two separate thickness measurements were made for

X. Liu, H.Y. Ng / Journal of Membrane Science 469 (2014) 112–126

each membrane sample. Thickness 1, designated as T1, was measured from the top surface of the substrate membrane to the top point of the PET mesh; Thickness 2, designated as T2, was

115

measured from the top surface of the substrate membrane to the bottom surface of the Psf portion, as illustrated in Fig. 2c. Each membrane sample was measured randomly at 15 different locations minimally and at least three replicated samples were observed for each designated S# and D# substrate membrane. 2.3.2. Pure water permeability Pure water permeability (PWP) of the substrate membranes was determined in a cross-flow filtration set-up described in Mo et al. [40]. Briefly, the dimension of the stainless steel testing cell was 30  100  2 mm3, giving an effective membrane area of 30 cm2. DI water as feed water was circulated into the membrane cell by a low pressure pump (GE A-C Motor 4805, Mexico). Desired pressure and feed flow rate were achieved by adjusting the bypass needle valve and back-pressure regulator. The applied pressure and retentate flow rate were monitored by a digital pressure gauge (PSI-Tronix, Inc., Tulane, CA) and a variable area flow meter (BlueWhite industries, Ltd., Huntington Beach, CA), respectively. All tests were conducted at 14 psi and cross-flow rate of 0.9 LPM (25 cm/s). The feed water temperature was maintained constant at 257 0.5 1C by a temperature control unit (Model CWA-12PTS, Wexten Precise Industries Co., Taiwan). The pure water flux was calculated based on gravimetric measurement of permeate water collected within a specific time duration and normalized over the effective membrane area. The pure water permeability was then calculated by normalizing the water flux over the applied pressure. The reported values were the average of at least three replicates. A simple pore-flow model, the Hagen–Poiseuille equation, was used as an aid to qualitatively elucidate the variation of PWP for different substrate membranes: 4

Q¼

ΔPεdpore 32ηlpore

ð1Þ

2.3.3. Contact angle measurement The contact angles of freeze-dried (Christ Alpha 1-2 LD plus, Germany) substrate membranes were measured by a contact angle goniometer (VCA-Optima, AST product Inc) using DI water as the probe liquid at 23 1C. The contact angle was randomly measured at 10 different locations for each sample, and the average of at least three samples was reported for each designated substrate membrane.

Fig. 2. Microscopic images of (a) PET mesh spacing, (b) PET mesh thickness and (c) thickness measurement of substrate membrane using digital microscope.

2.3.4. Mean pore size and pore size distribution The pore size and pore-size distribution of S# and D# substrate membrane were characterized by solute rejection experiments described in an earlier work [41], with modifications in the protocol. The rationale for such amendments was explained in detail in Section 3.1. Briefly, single solute rejections against poly (ethylene oxide) (PEO)s with different molecular weights (ranging from 200,000 to 1,000,000 g mol  1) were tested. During each

Table 1 Summary of substrate membrane compositions, thickness, PWP and contact angle characteristics. Membrane ID

Solution A, Psf wt%a

Solution B, Psf wt%

T1, μm

T2, μm

Pure water permeability, LMH/Bar

Contact angle, deg

S1 S2 S3 S4 D1 D2 D3 D4 D5

10 9 8 7 10 10 10 10 10

– – – – 9 8 7 6 5

46 7 3 417 3 337 5 317 3 42 7 3 40 7 4 32 7 3 28 7 4 217 3

132 76 119 711 110 719 95 78 131 74 114 78 97 715 78 717 66 713

2496 7 63 2545 7 242 2781 7 282 38127 455 2522 7 253 26127 285 30977 217 33847 199 4042 7 247

83.3 7 1.5 83.17 1.8 82.3 7 2.1 80.7 7 2.1 83.4 7 1.5 83.17 1.6 83.0 7 1.6 83.2 7 1.7 83.0 7 2.4

a For single-blade casted substrate membranes, the polymer concentration was decreased to 7%. Further reduction did not permit formation of free-standing and viable substrate membranes.

116

X. Liu, H.Y. Ng / Journal of Membrane Science 469 (2014) 112–126

single-solute rejection test, 100 ml of the solution containing 50 ppm of the solute was filtered through the membrane (3.5 cm diameter) under 100 mbar vacuum (Vacuumbrand diaphragm vacuum pump, Germany). 20 ml of the permeate was collected and the solute concentrations in both feed and permeate were then determined by a total organic carbon analyzer (TOC) (Shimazu TOC-L). The solute separation factor R, was calculated by using the following equation:
 Cp R ¼ 1  100 ð2Þ Cf The relationship between the Stokes radius (r s , nm) of PEO and its molecular weight (M w , g mol  1) can be described in the following equation [41]: r s ¼ 10:44  10  3  M 0:587 w

ð3Þ

When the solute separation factor, R, is plotted against solute diameter, ds , on log-normal probability paper, a straight line is yielded. By ignoring the influences of the steric and hydrodynamic interaction between solute and membrane pores, the mean effective pore diameter, μp ; and the pore size distribution geometric standard deviation, σ p ; can be assumed to be the same as the geometric mean diameter of solute at R ¼ 50% and the solute rejection geometric standard deviation, defined as the ratio of the ds at R¼ 84.13% over that at R¼50%. Therefore, based on μp and σ p , the pore size distribution, dp ; of the membrane can be estimated from the following probability density function [12,42]: " # ðln dp  ln μp Þ2 dRðdp Þ 1 pffiffiffiffiffiffiexp  ¼ ð4Þ ddp dp ln σ p 2π 2ðln σ p Þ2

2.3.5. Porometer study The bubble-point pressure and bubble-point pore diameter of the substrate membranes were measured using a capillary flow porometer (PMI, CFP-1200A). The bubble-point pore is defined as the largest pore on the skin surface of an asymmetric membrane, and can be inversely related to the applied pressure with the bubble point equation: P bp ¼

4δ cos θ dcap

2.4. TFC-FO membrane performance evaluation 2.4.1. Evaluation of TFC-FO membranes in osmotic process The TFC-FO membrane performance (i.e., water flux and reverse salt flux) was evaluated with a bench-scale FO setup as described in Parida et al. [5]. Briefly, the cross-flow cell unit contained two channels with each having dimensions of 120 mm long, 50 mm wide and 2 mm deep. The solution flow velocities in both the draw and feed channels were maintained at 1.5 LPM (25 cm/s) in co-current flow. The temperature of the feed and draw solutions were maintained at 2470.5 1C. DI water and 1.0 M NaCl solution were used as the feed solution and draw solution, respectively. The calculated osmotic pressure of the draw solution was 48.4 bar (702 psi) (OLI Systems, Inc., Morris Plains, NJ). Each solution reservoir had an initial volume of at least 5.0 L. All membranes were tested under two different modes: (1) FO mode where the feed water flows against the dense active layer; and (2) PRO mode where the draw solution faces against the dense active layer. The water flux through the membrane was measured by the change in mass of the feed solution reservoir, which was placed on a digital balance (MS 16001LE Mettler Toledo, Switzerland). The initial flux was allowed to stabilize for 10 min. After this, weight change data for 0.5 h was taken and water flux, J v ; was determined via the following equation: Jv ¼

Δmdraw =ρdraw Am Δt

ð6Þ

The change of draw solution concentration was ignored because the ratio of water permeation to the volume of the draw solution was less than 5% during the testing. The reverse salt flux, J s ; of the FO membrane was characterized by calculating the change of salt content in the feed solution based on conductivity measurement with a portable conductivity meter (Orion 4 star, Thermo scientific) at regular time intervals: Js ¼

V t Ct  V oCo Am Δt

ð7Þ

The change in the feed solution volume was ignored because of the large volume of feed solution reservoir and it was assumed that V t ffi V o .

ð5Þ

Prior to the measurement, the substrate membrane samples were frozen at  80 1C for 12 h and freeze-dried at  56 1C and 0.01 mbar for 7–8 h (Christ Alpha 1-2 LD plus, Germany). The dry samples were then wetted with a proprietary wetting solution (Galwick) with surface tension of 15.9 dynes/cm. The samples were then loaded and tested in both wet/up and dry/up conditions. 99.99% pure N2 was used as the pressurizing gas with pressure incremented at 0.1 psi/s from 0 psi up to 90 psi. The results were reported as average of three replicates for each designated S# and D# substrate membrane.

2.3.6. Morphology study The top surface morphology of the substrate membranes was observed with a field emission scanning electron microscope (FESEM JEOL JSM-6700LV). The cross-section and bottom surface morphology of the substrate membranes were observed using a scanning electron microscope (SEM JEOL JSM-5600LV). Prior to taking the SEM images, the vacuum dried samples were stored in liquid nitrogen and then carefully sliced with a surgical knife. The samples were then coated with platinum using a sputtering coater (JEOL LFC-1300) before taking the SEM images.

2.4.2. Evaluation of TFC membranes in pressure driven process After evaluating the membrane performance in osmotic process, the pure water permeability and salt permeability of the TFCFO membrane were then determined by testing the membranes using the pressure-driven cross-flow filtration set-up described earlier in Section 2.3.2. The pure water permeation flux with DI water feed was measured under an applied trans-membrane pressure of 125 psi (8.62 bar). The feed water temperature was maintained constant at 257 0.5 1C and the cross-flow rate was maintained at 0.9 LPM (25 cm/s). The pure water flux, J w was measured by weighing the amount of permeate water collected within a specific time duration and normalized over the effective membrane area. The water permeability coefficient of the TFC-FO membranes, A was acquired as J w ¼ AðΔP  ΔπÞ

ð8Þ

DI water was used as feed during the acquisition of A. Subsequently, the salt rejection, Rs ; was determined from the measured conductivities of permeate and feed by using feed water containing 2,000 ppm NaCl at 125 psi (8.62 bar) according to Eq. (9). The operating conditions remained identical with runs using DI as the feed water, whereby the temperature was maintained constant at 25 70.5 1C and the cross-flow rate at

X. Liu, H.Y. Ng / Journal of Membrane Science 469 (2014) 112–126

0.9 LPM (25 cm/s). Cf  Cp Rs ¼  100% Cf

ð9Þ

The salt permeability coefficient, B, an intrinsic property of membrane skin layer, was calculated based on the solutiondiffusion theory [6,20]: 1  Rs B ¼ AðΔP  ΔπÞ Rs

ð10Þ

117

2.4.3. Determination of substrate membrane structural parameter The structural parameter, S, of the FO membrane is of critical importance, and it is defined as the product of the support layer thickness, l, and tortuosity, τ, over its porosity ε as shown in Eq. (11) [44]. A smaller S value is generally favored for the FO membrane as the effect of ICP would be minimized.

S¼

lτ ε

ð11Þ

Fig. 3. FESEM and SEM images of S# substrate membranes: (a) S1 top surface, (b) S1 cross section, (c) S1 bottom surface, (d) S2 top surface, (e) S2 cross section, (f) S2 bottom surface, (g) S3 top surface, (h) S3 cross section, (i) S3 bottom surface, (j) S4 top surface, (k) S4 cross section and (l) S4 bottom section.

118

X. Liu, H.Y. Ng / Journal of Membrane Science 469 (2014) 112–126

Fig. 4. FESEM and SEM images of D# substrate membranes: (a) D1 top surface, (b) D1 cross section, (c) D1 bottom surface, (d) D2 top surface, (e) D2 cross section, (f) D2 bottom surface, (g) D3 top surface, (h) D3 cross section, (i) D3 bottom surface, (j) D4 top surface, (k) D4 cross section, (l) D4 bottom surface, (m) D5 top surface, (n) D5 cross section, (o) D5 bottom surface.

X. Liu, H.Y. Ng / Journal of Membrane Science 469 (2014) 112–126

An apparent S value can be estimated using the flux-fitting method with experimentally determined FO water flux [43–45]:   Ds Aπ draw þ B ln ð12Þ Jv ¼ Aπ f eed þ J v þ B S where π draw is the osmotic pressure of the bulk draw solution and π f eed is the osmotic pressure of the feed solution at the membrane surface. Since DI water was used as the feed solution and the membranes were highly salt rejecting, π f eed was approximated as zero.

3. Results and discussion 3.1. Characteristics and performance of substrate membrane Table 1 shows the thickness, PWP and top surface contact angle for the designated substrate membranes. For S# substrate membrane, both T1 and T2 decreased as the polymer concentrations in the casting solutions were decreased from 10 to 7 wt%. Membrane thickness shrinkage is a common phenomenon in immersion precipitation process especially when the solvent outflow outweighs the non-solvent inflow, resulting in a large difference between the casting thickness and the ultimate membrane thickness [39]. As the polymer concentration was decreased from S1 to S4 substrate membranes, the content of solvent in the casting solution increased. This promoted the solvent outflow during the formation process. As a result, the boundary between the nonsolvent bath and the casting solution shrinked more during the demixing process, producing thinner final membrane thickness [26,39,46]. For D# substrate membranes, contrary to initial expectation of T1 remaining constant while the overall thickness of T2 decreasing, both T1 and T2 showed appreciable reduction as the polymer concentrations of solution B was decreased from 9 to 5 wt %. This might be attributed to the partial “sinking” of solution A into solution B since the polymer content of solution A was always higher than that of solution B, as explained by Hashemifard et al. [38]. Nevertheless, such membrane shrinkage is desirable to produce thin substrate membranes for high performance TFC-FO membranes. As depicted in Table 1, using the same casting knife clearance (90 μm), the thinnest functional (capable of forming TFC-FO membrane) substrate membrane from single-blade casting was S4, which stood at an overall thickness of 95 78 μm. This is substantially thicker than the substrate membrane D5, the thinnest among the double-blade casted substrate membranes, with an overall thickness of 667 13 μm. Such membrane thickness reduction in the D# substrate membrane might bring about potential enhancement in osmotic performance in D#-based TFC membranes, as we discuss in more details in Section 3.2. The pure water permeability (PWP) of S# and D# substrate membranes is also presented in Table 1. For S# membranes, PWP increased from 2496 LMH/Bar for S1 substrate membrane to 3812 LMH/Bar for S4 substrate membrane, because the membrane thickness decreased as explained earlier; the membranes were also likely to be more porous [26,39]. The FESEM images of top surfaces of S# substrate membranes are presented in Fig. 3. The surface pore size was observed to increase tremendously from S1 to S4. According to the Hagen–Poiseuille pore flow model (see Eq. (1)), water permeability increases with pore size squared and linearly with porosity, but decreases with effective pore length, which could be roughly represented by the effective substrate membrane thickness [47,48]. This gives a good correlation between the experimental PWP variation with the observed changes in pore size and thickness. A similar trend was also observed for D# membranes, where PWP increased from 2522 LMH/Bar for D1 substrate membrane to 4042 LMH/Bar for

119

D5 substrate membrane. It is interesting to note that even though the top surface morphology and pore sizes remained similar for all D# membranes, as portrayed in Table 1 and Fig. 4, the PWP still invariably increased. This suggests that the entire membrane structure, rather than merely the top surface pores, had contributed to the resistance against water flow in the substrate membranes [39]. The decrease in substrate membrane thickness from D1 to D5 might also contribute to the increase in PWP. The water contact angles of the substrate membrane top surface are also summarized in Table 1. In general, the contact angles for most S# and D# membrane remained at around 831, since all of the substrate membranes were made of the same material of Psf. However, a slight decrease to 801 was observed for S4 substrate membrane, probably because of the presence of large, breach-like pores on the membrane surface [39]. Table 2 and Fig. 5 display the pore-size characteristics of S# and D# substrate membranes. It was mentioned earlier in Section 2.3.4 that the mean pore sizes and pore size distributions of substrate membranes were determined with a modified protocol compared to the conventional single-solute rejection test described in Singh et al. [41]. During the initial solute-rejection tests following the conventional protocol, 200 ppm equivalent of solute was spiked in the feed and the substrate membranes were tested in the crossflow set-up described in Section 2.3.2 under 14 psi at 20 cm/s cross-flow velocity. However, rejection of solutes with smaller molecular weights/sizes was observed to be higher than the rejection of solutes with higher molecular weights for some of the substrate membranes, especially for the S# membranes. The permeate fluxes of the substrate membranes with these solutes present in the feed were also observed to be significantly lower than pure water flux, indicating that the substrate membranes had been considerably fouled by the solutes [41]. In retrospect, since the substrate membranes fabricated in the present work had rather large nominal pores sizes (cross-over between UF and MF size range), the solutes would penetrate the membrane pores and become lodged in the interior of the membranes [49]. Such poreclogging likely increased the apparent rejections of smaller molecular weight solutes [39,41,49]. In order to minimize the pore clogging effect by the solutes and ensuring the validity of assumption of minimal steric and hydrodynamic interaction between solute and membrane pores, the test protocol was thus modified to filter a small fixed volume of feed with reduced solute concentration under very low vacuum pressure as described in Section 2.3.4. For S# membranes, the mean pore size increased from 41.7 nm for S1 to 59.2 nm for S4 membranes. It is interesting to note that S4 has a rather wide pore size distribution with a long right-sided tail (Fig. 5a), suggesting the presence of very large pores/breaches on the surface of the membrane. For D# Table 2 Summary of substrate membrane mean pore sizes and bubble point pore size. Membrane ID

Mean pore Standard sizea μp , nm deviation, σp

Bubble point pressure Pbp, psi

Bubble point pore diameterb dcap, nm

S1 S2 S3 S4 D1 D2 D3 D4 D5

41.7 44.8 48.2 59.2 40.9 42.0 42.2 45.6 42.7

887 1 577 3 477 5 207 3 857 0 897 3 837 3 817 5 797 5

757 1 1167 6 1427 15 3357 48 797 0 767 2 807 3 827 5 847 6

1.42 1.45 1.39 1.42 1.45 1.42 1.43 1.38 1.42

a Determined by solute rejection test with 50 ppm PEOs under 100 mbar vacuum. b Tested using capillary-flow porometer (PMI).

120

X. Liu, H.Y. Ng / Journal of Membrane Science 469 (2014) 112–126

Probability density distribution, nm-1

0.035 S1 S2 S3 S4

0.030 0.025 0.020 0.015 0.010 0.005 0.000

0

50

100

150

200

250

Pore size, nm

Probability density distribution, nm-1

0.035 D1 D2 D3 D4 D5

0.030 0.025

3.2. Characteristics and performance of TFC-FO membranes

0.020 0.015 0.010 0.005 0.000

0

50

100

150

200

and pore size distribution, as an indication of suitability of the substrate membrane surface morphology for subsequent interfacial polymerization, since it has been recognized that the largest pore may represent the defect-prone area where integrity of PA active layer is likely to be compromised in the IP process [21,31]. For S# substrate membranes, P bp decreased from S1 to S4, corresponding to an increase in the average dcap of 75 nm for S1 to 335 nm in S4. This is confirmed in the FESEM observation of the top surfaces of the S# substrate membranes as shown in Fig. 3. Such dramatic increase in the largest pore size might impede the formation of an integral PA layer during IP and thus performance of the resultant TFC membranes could be compromised, as we shall further elaborate in Section 3.2. For D# substrate membrane, the P bp remained relatively constant from D1 to D5, maintaining the average dcap of less than 90 nm across the board. This is also confirmed in the FESEM images presented in Fig. 4, where no pores larger than 100 nm had been observed on the top surfaces of the D# substrate membranes.

250

Pore size, nm

Fig. 5. Pore size distribution of (a) S# substrate membranes and (b) D# substrate membranes.

membranes, the estimated mean pore sizes remained around 43 nm for all membranes in the series and the pore sizes distributions closely resembled each other as seen in Fig. 5b. Such an observation was expected because the skin surface porosity is chiefly determined by the interface composition between the polymer solution film and the non-solvent water bath at the instance of immersion [46,50]. In the current study, S# substrate membranes were casted from casting solutions with decreasing polymer concentration. As a result, the interface polymer concentration at the onset of precipitation decreased, resulting in greater skin surface porosity and larger pores at the skin surface [46]. On the contrary, for D# substrate membranes, although the polymer concentration of casting solution B was decreased from 9 to 5 wt%, the polymer content in casting solution A remained at 10wt% for all D# membrane. Assuming there was no substantial mixing between solution A and solution B during casting process [38], the interface polymer concentration at the instance of precipitation would thus remain the same for all D# substrate membranes, producing similar pores size at the skin surface. Bubble-point pressure and bubble-point pore diameter measurements are presented in Table 2 for S# and D# substrate membranes as well. Bubble point pressure is defined as the threshold gas pressure required to displace liquid from the pores of a fully wetted membrane. Eq. (5) describes the inverse relationship between the applied pressure and pore diameter, and the pressure at which bubbles are first detected in a fully wetted membranes, P bp , can be used to calculate the diameter of the largest pore, the bubble-point pore diameter, dcap [51]. Bubblepoint test is one of the most widely practiced methods for pore size determination and bubble-point pore is a good surrogate to represent the largest pore on the substrate membrane top surface [51]. It is used here, together with the estimated mean pore sizes

To synthesize TFC-FO membranes, a thin polyamide film was deposited on each aforementioned S# and D# substrate membrane via interfacial polymerization as described in Section 2.2.2. The resultant TFC-FO membranes were denoted as TFC-S# and TFC-D#, respectively. For example, the TFC-FO membranes made from S1 substrate membrane would be denoted as TFC-S1. Table 3 summarizes the performance of TFC-S#, TFC-D# and HTI membranes in pressure driven tests. For the TFC-S# membranes, the water permeability, A increased from 1.59 LMH/Bar for TFC-S1 to 2.60 LMH/Bar for TFC-S4, while salt permeability, B increased more drastically from 0.19 LMH for TFC-S1 to 4.19 LMH for TFC-S4, demonstrating a strong trade-off relationship between the water permeability and salt rejection of the active layer as elucidated in earlier work [52,53]. Previous works indicate that while high performance PA active layer can be reliably formed on Psf-substrate membrane surfaces, these similarly formed dense layers are also highly sensitive to variation in the structure of the Psf-substrate membrane top surface, especially pore sizes [18,23,52,54]. The permeability results for TFC-S# membranes in the currently study were in good agreement with such prior understanding, and pinpointed the fact that as the surface pore size increases on the substrate membranes, the subsequent deposition of PA active layer would be strongly influenced during the interfacial polymerization step. The integrity of the PA active layer could be compromised especially in the regions of large Table 3 Summary of performance of TFC-S#, TFC-D# and HTI membranes in pressuredriven tests. Membrane ID A, LMH/Bar B, LMH TFC-S1 TFC-S2 TFC-S3 TFC-S4 TFC-D1 TFC-D2 TFC-D3 TFC-D4 TFC-D5c HTI a

1.59 7 0.13 1.88 7 0.07 1.94 7 0.12 2.60 7 0.50 1.60 7 0.09 1.66 70.13 1.66 70.20 1.82 7 0.11 – 0.65 7 0.04

0.197 0.08 0.34 7 0.05 0.62 7 0.15 4.197 3.65 0.25 7 0.08 0.247 0.05 0.277 0.10 0.28 7 0.10 – 0.57 70.05

Rejection Rs, %a B/A, Bar S, μmb 98.4 7 0.5 97.5 70.5 95.8 7 0.7 83.4 7 8.7 97.8 70.5 98.17 0.3 97.8 70.6 97.9 70.8 – 89.17 0.3

0.12 0.18 0.32 1.49 0.15 0.14 0.16 0.15 – 0.88

281 7 30 2617 32 265 7 40 382 7 42 234 7 25 210 718 1847 31 1957 39 – 587 7 65

Tested at 125 psi (8.62 bar) with 2000 ppm NaCl solution. Structural parameters were calculated based on experiments under the FO mode using 1 M NaCl as the draw solution and DI water as the feed using Eq. (9). c Inconsistent permeability and rejection results due to the mechanical weakness of the support membrane. b

X. Liu, H.Y. Ng / Journal of Membrane Science 469 (2014) 112–126

pores, producing less dense active layer with greater water and salt permeability. By co-analyzing the pore size distributions in Fig. 5, bubble-point pore diameters in Table 2 of S# substrate membranes and the salt rejection results of resultant TFC-S# membrane presented in Table 3, the current study suggests that a dense, integral PA layer with high salt rejection and moderate water permeability can be formed if the substrate membrane has surface pore sizes well controlled below 100 nm. Fig. 6 shows the top surface of TFC-S1, TFC-S4 and TFC-D3 membranes. All three membranes depicted the typical ridge-valley structures of polyamide layer. However, compared to the more uniform topology of TFC-S1, TFC-S4 had displayed a more mosaic surface. Nevertheless, it should be pointed out that the TFC-S4 was still a fully functional TFC-FO membrane with a full-fledged active layer exhibiting reasonable salt rejections (ranging from 67% to 88% of NaCl rejections for all the TFC-S4 membranes samples tested).

121

Interfacial polymerization is intrinsically self-sealing and our results suggest that, with the IP monomer concentrations stated in Section 2.2.2, the PA active layer could still be successfully developed even on the S4 substrate membranes, albeit having an apparently looser structure with higher water and salt permeability. Similar discussions were provided by Shi et al. [18] and Singh et al. [54] as well. It is worth mentioning that in the current study, a high concentration of MPD solution (4 wt%) was used in conjunction with a high MPD/ TMC ratio during the interfacial polymerization process. This produced a dense, highly cross-linked polyamide layer which might aid in covering up large breach-like pores in the S4 substrate membrane [29,55], and the rationale for such monomer concentrations formulae will be further elaborated in Section 3.3. Fig. 7 further compares the FO and PRO performance of TFC-S# membranes using 1 M NaCl as the draw solution and DI water as the feed solution. Overall, the PRO configuration showed much

Fig. 6. FESEM images of (a) TFC-S1 top surface, (b) TFC-S1 cross-section polyamide layer close up, (c) TFC-S4 top surface, (d) TFC-S4 cross-section polyamide layer close up, (e) TFC-D3 top surface and (f) TFC-D3 cross-section polyamide layer close up.

122

X. Liu, H.Y. Ng / Journal of Membrane Science 469 (2014) 112–126

higher water flux than the FO configuration for all the membranes. This trend is attributed mainly to the extent of concentrative/ dilutive ICPs in the two membrane orientations, which has been well characterized in earlier works [14,56,57]. Briefly, more significant ICP occurs in the FO mode in the porous substrate membrane layer, leading to a decreased net osmotic driving force and thus, reduced water flux. Nevertheless, water flux in both FO and PRO orientations increased slightly from TFC-S1 to TFC-S3 membranes, probably because of the reduced ICP effect delivered by the thinner, more permeable substrate membrane layer down the series. In addition, conventional wisdom also suggests that higher A of the active layer down the series might also contribute to the slight increase in the water flux [18]. The reverse salt flux increased more rapidly as compared to the water flux from TFC-S1 to TFC-S3, which corresponded to a less dense active layer with higher salt permeability as discussed earlier. However, it is interesting to note that despite having the thinnest and most permeable substrate membrane as the support among the TFC-S# membranes, TFC-S4 bucked the trend by showing a decreased water flux in both the FO and PRO mode. This might be attributed to the less dense or a slightly “leaky” active layer as discussed earlier. While the higher water permeability of a less cross-linked PA active layer might enhance the FO water flux due to reduced transport resistance, its lower salt rejection might simultaneously cause a more severe ICP due to a surging solute reverse diffusion, termed J s -induced ICP, to decrease the FO water flux [29,58]. As depicted in Fig. 7b, TFC-S4 showed a much higher reverse salt flux in both the PRO and FO modes, a direct consequence of an active layer with high B value. In the PRO mode, the reverse-diffused salt accumulates in the porous substrate membrane, compounding the

70 60

Water flux, LMH

50 40 30 20 10 0 TFC-S1

TFC-S2

TFC-S3

TFC-S4

TFC-S1

TFC-S2

TFC-S3

TFC-S4

200 180

Reverse salt flux, gMH

160 140 120 100 80 60 40 20 0

FO orientation PRO orientation

Fig. 7. Performance of TFC-S# membranes in osmotic tests: (a) water flux and (b) reverse salt flux in both FO and PRO orientations with 1 M NaCl as draw solution and DI water as the feed.

ICP phenomenon in the substrate membrane; while in the FO mode, the salts from the draw solution pass through the active layer quickly, deleteriously aggravating the establishment of an effective osmotic gradient [6]. As a result, this J s -induced phenomenon also partially negated the effective osmotic driving across the membrane and consequently ensued in a dramatically lower water flux. For the TFC-D# membranes, the A values remained relatively constant from 1.60 LMH/Bar for TFC-D1 to 1.82 LHM/Bar for TFCD4, while B values followed the same trend with a minimal increase from 0.25 LMH for TFC-D1 to 0.28 LMH for TFC-D4. This is not surprising because the surface of D1 to D4 substrate membranes had remained relatively constant, as underlined earlier. This will provide a similar base structure for the later IP process and producing similar PA active layer with similar water permeability and salt permeability. Moreover, this is also in accordance with our previous deduction that substrate membranes with surface pore sizes of less than 100 nm would be suitable for the formation of a dense intact polyamide active layer during the subsequent interfacial polymerization. Fig. 6 also shows the top surface morphology of the TFC-D3. TFC-D3 was used here for comparison because its substrate membrane, D3, is co-casted with two different polymer solutions that produced S1 and S4. As clearly shown in Fig. 6, the top surface of the TFC-D3 resembled that of TFC-S1, further confirming that the PA active layer formed on the D3 substrate membrane would be similar to the PA active layer formed on the S1 substrate membrane. The results for the TFC-D5 are not reported here since variability of the rejection results is too great (with Rs ranging from 98% to 27%). This could be attributed to the deformation of the Psf layer (average T1 of 21 μm) and its easy detachment from the PET fabric as seen in Fig. 4, creating instability that negatively affected the performance of the membranes. Fig. 8 summarizes the FO and PRO performance of the TFC-D# membranes using 1 M NaCl as the draw solution and DI water as the feed solution. Similarly, the PRO configuration showed much higher water flux than the FO configuration for all the membranes. The water flux continually increased for the TFC-D# membranes down the series, with the FO water flux peaking at 31.3 LMH and PRO water flux at around 60 LMH. Fig. 8b shows the reverse salt flux of the TFC-D# membranes in both the FO and PRO modes. The figure clearly manifests the advantage of the TFC-D# over the TFCS# membranes as the reverse salt flux increased only slightly down the series, maintaining a high rejection of salts. This also agrees with the high rejection of salts of the membrane tested in the pressure driven test as discussed earlier. The structural parameter, S, is defined as the product of the support layer thickness, l and tortuosity, τ over its porosity ε [43], and provides an indication of the diffusion path length for the solutes across the substrate membrane layer. It has been used as an aiding parameter to quantify the extent to which the substrate layer of the FO membranes will cause ICP [33]. The calculated S values for the TFC-FO membranes, based on the FO water flux fitting model using Eq. (12), are also tabulated in Table 3. The HTI membranes tested in the current study showed an S value of 587 765 μm, which is in good agreement with earlier studies [27,30,44]. All the TFC-FO membranes showed lower S values. For the TFC-S# membranes, the S values decreased from 281 μm to around 260 μm for the TFC-S1 to TFC-S3. Again, the TFC-S4 membranes bucked the trend by showing a larger S value of 382 μm. It should be pointed out that S parameter discussed here is the apparent value instead of the intrinsic structure parameter of porous substrate. Therefore the apparent S value, as calculated from the flux-fitting method here, is also dependent on the skin layer properties, i.e., A and B values of the skin layer. For the TFCS4 membranes, its A and B values were much higher than those of

X. Liu, H.Y. Ng / Journal of Membrane Science 469 (2014) 112–126

70

123

5

60

4

Js/Jv ratio, g/L

Water flux, LMH

50 40 30

3 2

20

1

10

0

0 TFC-D1

TFC-D2

TFC-D3

TFC-D4

TFC-D5

TFC-S1

TFC-S2

TFC-S3

TFC-S4

50 0.7 0.6

30

Js/Jv ratio, g/L

Reverse salt flux, gMH

40

20

0.5 0.4 0.3

10 0.2

0 TFC-D1

TFC-D2

TFC-D3

TFC-D4

TFC-D5

FO orientation PRO orientation

Fig. 8. Performance of TFC-D# membranes in osmotic tests: (a) water flux and (b) reverse salt flux in both FO and PRO orientations with 1 M NaCl as the draw solution and DI water as the feed.

the other TFC-S# membranes, thus ensuing a larger S value. Conceptually, using FO orientation for illustration, the high A value of the skin layer would produce a higher permeation drag to push the salts in the draw solution from trying to reach the active layer, resulting in a seemingly longer diffusion path length across the porous substrate membrane. For the TFC-D# membranes, the S values also followed a decreasing trend, from 234 μm to around 190 μm down the series. All apparent S values for the TFC-D membranes were lower than those of the TFC-S# membranes, suggesting the TFC-D# membranes have an overall enhanced performance with reduced ICP effect. However, it should be noted that the reduction in the S values for the TFC-D# membranes were not as prominent as the reduction in the membrane thickness for the respective D# substrate membranes. Such an observation is in lieu with prior works [30,31], which hypothesized that the overall S value of the substrate membrane can be considered as a combined contribution from the sponge skin layer and the sublayer of the substrate membrane. The sponge skin layer of the substrate normally commands the dominant contribution to the overall S value because of small porosity and large turtosity. In the current work, the sponge skin layers of the co-casted D# substrates stayed similar while their sub-layers thickness decreased, as indicated in aforementioned discussion. This subsequently led to the less-than-expected S value reduction in the TFC-D# membranes. Lastly, as compared to the TFC-S# membranes, where the A and B values of the PA active layer changed significantly down the series, the A and B values of the TFC-D# membranes remained relatively constant. Consequently, the apparent S values might be a more meaningful tool for substrate membrane structural comparison in TFC-D# membranes.

0.1 TFC-D1

TFC-D2

TFC-D3

TFC-D4

TFC-D5

Ratio in FO orientation Ratio in PRO orientation

Fig. 9. Reverse salt flux to water flux ratio, Js/Jv for: (a) TFC-S# membranes and (b) TFC-D# membranes in both FO and PRO orientations.

Fig. 9 depicts the reverse salt flux/water flux ratio for the TFCS# and TFC-D# membranes. J s =J v ratio is an important performance indicator in osmotic process, and an ideal FO membrane should possess high water flux, J v and low reverse salt flux, J s [28]. An excessively large J s =J v ratio is forbidden due to [59]: (1) the inducement of severe ICP from severe reverse solute diffusion and accumulation in the substrate membrane layer; (2) the poor retention against possible contaminants in the feed solution; and (3) the requirement for costly draw solute replenishment [15]. Thus a lower J s =J v ratio for the fabricated membranes would be much favored. For the TFC-S# membranes synthesized in the current study, J s =J v ratio increased substantially across the board. This implies that although lowering the polymer concentration in the casting solutions might make the substrate membrane thinner and more permeable, the conventional single-blade casting technique produced substrate membrane with top surface morphology unfit for polyamide active layer deposition, adversely affecting the active layer integrity and causing significant increase in J s =J v ratios down the resultant TFC-S# membranes series. For the TFC-D# membranes, however, J s =J v ratios remained low and relatively constant down the series. Thus it implies that the TFC-D# membranes developed in the current study could achieve significantly higher water fluxes while maintaining relatively low J s =J v ratios. As discussed previously, the high water flux was contributed by their optimized substrate membrane structure and the low and relatively stable J s =J v ratios that could be attributed to the excellent selectivity of the uncompromised active layer. The results support the hypothesis that double-blade co-casting technique can produce an optimized substrate membrane that is thinner and more permeable to reduce the ICP, while maintaining a top surface

124

X. Liu, H.Y. Ng / Journal of Membrane Science 469 (2014) 112–126

B must be carefully evaluated for successful substrate membrane formation [12,20,21].

structure desirable for highly selective PA active layer formation during interfacial polymerization.

4. Conclusion 3.3. Implication In osmotic membrane processes, the membrane performance is evaluated in terms of water flux and reverse salt flux. The reverse salt flux of the TFC-FO membranes is chiefly correlated to the saltrejecting capability of the active layer (indicated by the B value). On the other hand, the water fluxes dependent on both the water permeability (indicated by the A value) of the active layer and the ICP-mitigating capability (indicated by the S value) of the porous substrate layer [25]. In a nutshell, the conventional wisdom suggests that a large A value with a small B value and a small S value would be ideal for high performance TFC-FO membranes [18,58]. Analysis of the results from the current study reached similar conclusions but underscored that a small B value is in fact critical in producing high water flux as well, due to J s -induced-ICP in highly permeable membrane. A bulk of earlier research studies focused on the modification of active layer of TFC-FO membranes [2]. Such modification attempts to improve the water permeability of the active layer, producing highly water permeable active layer. However, because of the strong trade-off between the water permeability and salt selectivity, the salt selectivity of the membrane active layer could be compromised (a seemingly high salt rejection in the RO test could be misleading because of the dilution effect of increasingly high water permeability). This will likely lead to excessive reverse salt flux and decrease the TFC-FO membrane performance. Thus optimization of FO water flux appeals for the careful deliberation of the various competing mechanisms. Analysis of the results obtained from our current study seems to support the notion that the active layer of TFC-FO membranes should be enhanced to achieve a high salt rejection even if it could be at the expense of reducing the water permeability, since ICP within the substrate membranes rather than the water permeability of the active layer is likely the limiting factor in controlling water flux in osmotic processes, as evident from Figs. 7 and 8 and a number of references [9,14,18,30,58]. The current study focused primarily on changing the polymer concentrations in the casting solution B as a varying parameter to make the substrate membrane thinner and more permeable to reduce ICP; the polymer material employed in this study, polysulfone, is common and easily available. Nevertheless, the doubleblade co-casting technique has demonstrated encouraging outcomes for flat-sheet substrate membrane fabrication and should be extended to incorporate other changes in casting solutions. This may open up many new frontiers and platforms for substrate membrane fabrication for FO membranes. Further research could be carried out with:

 Incorporating organic pore-formers such as polyethylene glycol





(PEG) and polyvinylpyrrolidone (PVP) etc. or into casting solution B to make the substrate membrane more porous yet maintain the top surface pore size [21]. Blending of microporous and mesoporous particles such as zeolite particles, carbon/silica-based nano-particles into casting solution B to produce more porous substrate membranes without compromising the surface integrity of substrate membrane [31]. Using more hydrophilic materials as the polymer content in casting solution B to boost water transport in the substrate membrane and reduce ICP. A dual-layer structure could be realized for flat-sheet substrate membrane. However, possible delamination must be further investigated and compatibility of the polymer materials in casting solution A and casting solution

In the current concept demonstration study, both single-blade casting technique and double-blade casting technique were employed to synthesize substrate membranes for TFC-FO membrane fabrication. The results have demonstrated that:

 Substrate membrane surface morphology, especially surface 

pore size, was pivotal in polyamide active layer formation during interfacial polymerization for TFC-FO membranes. Substrate membranes with surface pore size o100 nm functioned as a desirable base for the formation of a dense and highly salt-rejecting active layer.

The resultant TFC-D# membranes based on double-blade casted D# substrate membranes showed increased water flux, without compromising on the reverse salt flux. With a 1 M NaCl draw solution and DI water feed solution, the best TFC-D# membranes achieved J v of 31.1 LMH and J s of 8.5 gMH in FO orientation, and J v of 60.3 LMH and J s of 17.6 gMH in the PRO orientation. The resultant TFC-D# membranes showed reduced apparent S values and retained a relatively low J s =J v ratio. This could be attributed to the fact that the double-blade casted D# substrate membrane having a thinner and more permeable structure to reduce ICP, while retaining an ideal surface pore structures for the formation of an intact, dense and highly salt rejecting polyamide active layer. Acknowledgments The authors would like to thank the Environment and Water Industry Programme Office (EWI) as the first author is supported by the National Research Foundation Singapore under its National Research Foundation (NRF) Environmental and Water Technologies (EWT) PhD Scholarship Programme. Dr. Wei Duan's guidance on TFC membrane fabrication is greatly appreciated.

Nomenclature A Am B Co Ct Cf Cp Ct dcap dp dpore ds Ds

DI DMF ECP FO ICP

water permeability, L m  2 h  1 bar  1 effective membrane surface area, m2 salt permeability, L m  2 h  1 initial salt concentration of feed solution, mol L  1 final salt concentration of feed solution, mol L  1 feed concentration, mol L  1 permeate concentration, mol L  1 salt concentration, mol L  1 bubble point diameter, nm pore diameter, nm cylindrical pore size, m solute diameter, nm diffusion coefficient of salt in the membrane substrate (assumed to be constant across the total thickness of the membrane), m2 s  1 deionized N,N-dimethylformamide external concentration polarization forward osmosis internal concentration polarization

X. Liu, H.Y. Ng / Journal of Membrane Science 469 (2014) 112–126

IP Js Jw Jv l lpore m MPD P P bp PA PET PEO PRO PWP Psf R Rs rs S T1 T2 TFC TMC Vo Vt Δmdraw Δt ΔP Δπ μp ε π ρ ρdraw δ θ τ η σp

interfacial polymerization reverse salt flux, g m  2 h  1 water flux in pressure driven process, L m  2 h  1 water flux in osmotic process, L m  2 h  1 thickness, μm effective pore length, μm membrane weight, g M-phenylenediamine applied pressure, bar bubble point pressure, psi polyamide polyester polyethylene oxide pressure-retarded osmosis pure water permeability, L m  2 h  1 bar  1 polysulfone solute separation factor, % salt rejection, % stoke radius, nm membrane structural parameter, μm thickness from substrate membrane top surface to PET mesh top surface, μm thickness from substrate membrane top surface to polysulfone bottom surface, μm thin-film-composite Trimesoyl Chloride initial volume of feed solution, L final volumes of feed solution, L weight change of the draw solution reservoir, g operation time interval, h trans-membrane pressure, pascal osmotic pressure difference between the feed and permeate, bar mean pore size, nm membrane porosity, % osmotic pressure, bar material density, g cm  3 density of draw solution, g cm  3 surface tension, dynes/cm contact angle, deg tortuosity, % viscosity, pascal s pore size distribution standard deviation

References [1] T.S. Chung, S. Zhang, K.Y. Wang, J.C. Su, M.M. Ling, Forward osmosis processes: yesterday, today and tomorrow, Desalination 287 (2012) 78–81. [2] S. Zhao, L. Zou, C.Y. Tang, D. Mulcahy, Recent developments in forward osmosis: opportunities and challenges, J. Membr. Sci. 396 (2012) 1–21. [3] T.Y. Cath, A.E. Childress, M. Elimelech, Forward osmosis: principles, applications, and recent developments, J. Membr. Sci. 281 (1–2) (2006) 70–87. [4] B. Mi, M. Elimelech, Chemical and physical aspects of organic fouling of forward osmosis membranes, J. Membr. Sci. 320 (1–2) (2008) 292–302. [5] V. Parida, H.Y. Ng, Forward osmosis organic fouling: effects of organic loading, calcium and membrane orientation, Desalination 312 (2013) 88–98. [6] K.L. Lee, R.W. Baker, H.K. Lonsdale, Membranes for power generation by pressure retarded osmosis, J. Membr. Sci. 8 (2) (1981) 141–171. [7] K. Gerstandt, K.V. Peinemann, S.E. Skilhagen, T. Thorsen, T. Holt, Membrane processes in energy supply for an osmotic power plant, Desalination 224 (1–3) (2008) 64–70. [8] S. Loeb, Production of energy from concentrated brines by pressure retarded osmosis. I. Preliminary technical and economic correlations, J. Membr. Sci. 1 (1) (1976) 49–63. [9] N.Y. Yip, A. Tiraferri, W.A. Phillip, J.D. Schiffman, L.A. Hoover, Y.C. Kim, et al., Thin-film composite pressure retarded osmosis membranes for sustainable power generation from salinity gradients, Environ. Sci. Technol. 45 (10) (2011) 4360–4369.

125

[10] B. Jiao, A. Cassano, E. Drioli, Recent advances on membrane processes for the concentration of fruit juices: a review, J. Food Eng. 63 (3) (2004) 303–324. [11] K.B. Petrotos, H.N. Lazarides, Osmotic concentration of liquid foods, J. Food Eng. 49 (2–3) (2001) 201–206. [12] Q. Yang, K.Y. Wang, T.S. Chung, A novel dual-layer forward osmosis membrane for protein enrichment and concentration, Sep. Purif. Technol. 69 (3) (2009) 269–274. [13] C.Y.Y. Tang, Q.H. She, W.C.L. Lay, R. Wang, A.G. Fane, Coupled effects of internal concentration polarization and fouling on flux behavior of forward osmosis membranes during humic acid filtration, J. Membr. Sci. 354 (1–2) (2010) 123–133. [14] J.R. McCutcheon, M. Elimelech, Influence of concentrative and dilutive internal concentration polarization on flux behavior in forward osmosis, J. Membr. Sci. 284 (1–2) (2006) 237–247. [15] J.J. Qin, S.J. Chen, M.H. Oo, K.A. Kekre, E.R. Cornelissen, C.J. Ruiken, Experimental studies and modeling on concentration polarization in forward osmosis, Water Sci. Technol. 61 (11) (2010) 2897–2904. [16] J.C. Su, T.S. Chung, Sublayer structure and reflection coefficient and their effects on concentration polarization and membrane performance in FO processes, J. Membr. Sci. 376 (1–2) (2011) 214–224. [17] C.H. Tan, H.Y. Ng, Modeling of external and internal concentration polarization effect on flux behaviour of forward osmosis, Water Sci. Technol. 8 (2010) 533–539. [18] L. Shi, S.R. Chou, R. Wang, W.X. Fang, C.Y. Tang, A.G. Fane, Effect of substrate structure on the performance of thin-film composite forward osmosis hollow fiber membranes, J. Membr. Sci. 382 (1–2) (2011) 116–123. [19] W.C.L. Lay, J.S. Zhang, C.Y. Tang, R. Wang, Y. Liu, A.G. Fane, Factors affecting flux performance of forward osmosis systems, J. Membr. Sci. 394 (2012) 151–168. [20] N. Widjojo, T.S. Chung, M. Weber, C. Maletzko, V. Warzelhan, A sulfonated polyphenylenesulfone (sPPSU) as the supporting substrate in thin film composite (TFC) membranes with enhanced performance for forward osmosis (FO), Chem. Eng. J. 220 (2013) 15–23. [21] K.Y. Wang, T.S. Chung, G. Amy, Developing thin-film-composite forward osmosis membranes on the PES/SPSf substrate through interfacial polymerization, AIChE J. 58 (3) (2012) 770–781. [22] P. Sukitpaneenit, T.S. Chung, High performance thin-film composite forward osmosis hollow fiber membranes with macrovoid-free and highly porous structure for sustainable water production, Environ. Sci. Technol. 46 (13) (2012) 7358–7365. [23] A.K. Ghosh, E.M.V. Hoek, Impacts of support membrane structure and chemistry on polyamide-polysulfone interfacial composite membranes, J. Membr. Sci. 336 (1–2) (2009) 140–148. [24] G. Han, S. Zhang, X. Li, N. Widjojo, T.S. Chung, Thin film composite forward osmosis membranes based on polydopamine modified polysulfone substrates with enhancements in both water flux and salt rejection, Chem. Eng. Sci. 80 (2012) 219–231. [25] X. Li, K.Y. Wang, B. Helmer, T.S. Chung, Thin-film composite membranes and formation mechanism of thin-film layers on hydrophilic cellulose acetate propionate substrates for forward osmosis processes, Ind. Eng. Chem. Res. 51 (30) (2012) 10039–10050. [26] A. Tiraferri, N.Y. Yip, W.A. Phillip, J.D. Schiffman, M. Elimelech, Relating performance of thin-film composite forward osmosis membranes to support layer formation and structure, J. Membr. Sci. 367 (1–2) (2011) 340–352. [27] N.Y. Yip, A. Tiraferri, W.A. Phillip, J.D. Schiffman, M. Elimelech, High performance thin-film composite forward osmosis membrane, Environ. Sci. Technol. 44 (10) (2010) 3812–3818. [28] J. Wei, C.Q. Qiu, C.Y.Y. Tang, R. Wang, A.G. Fane, Synthesis and characterization of flat-sheet thin film composite forward osmosis membranes, J. Membr. Sci. 372 (1–2) (2011) 292–302. [29] J. Wei, X. Liu, C.Q. Qiu, R. Wang, C.Y.Y. Tang, Influence of monomer concentrations on the performance of polyamide-based thin film composite forward osmosis membranes, J. Membr. Sci. 381 (1–2) (2011) 110–117. [30] R. Wang, L. Shi, C.Y.Y. Tang, S.R. Chou, C. Qiu, A.G. Fane, Characterization of novel forward osmosis hollow fiber membranes, J. Membr. Sci. 355 (1–2) (2010) 158–167. [31] N. Ma, J. Wei, S. Qi, Y. Zhao, Y. Gao, C.Y. Tang, Nanocomposite substrates for controlling internal concentration polarization in forward osmosis membranes, J. Membr. Sci. 441 (2013) 54–62. [32] G. Han, T.S. Chung, M. Toriida, S. Tamai, Thin-film composite forward osmosis membranes with novel hydrophilic supports for desalination, J. Membr. Sci. 423 (2012) 543–555. [33] L.A. Hoover, J.D. Schiffman, M. Elimelech, Nanofibers in thin-film composite membrane support layers: enabling expanded application of forward and pressure retarded osmosis, Desalination 308 (2013) 73–81. [34] L. Dumee, J. Lee, K. Sears, B. Tardy, M. Duke, S. Gray, Fabrication of thin film composite poly(amide)-carbon-nanotube supported membranes for enhanced performance in osmotically driven desalination systems, J. Membr. Sci. 427 (2013) 422–430. [35] T. He, M.H.V. Mulder, H. Strathmann, M. Wessling, Preparation of composite hollow fiber membranes: co-extrusion of hydrophilic coatings onto porous hydrophobic support structures, J. Membr. Sci. 207 (2) (2002) 143–156. [36] X.M. Li, Y. Ji, Y. Yin, Y.Y. Zhang, Y. Wang, T. He, Origin of delamination/adhesion in polyetherimide/polysulfone co-cast membranes, J. Membr. Sci. 352 (1–2) (2010) 173–179. [37] C.C. Pereira, R. Nobrega, C.P. Borges, Membranes obtained by simultaneous casting of two polymer solutions, J. Membr. Sci. 192 (1–2) (2001) 11–26.

126

X. Liu, H.Y. Ng / Journal of Membrane Science 469 (2014) 112–126

[38] S.A. Hashemifard, A.F. Ismail, T. Matsuura, Co-casting technique for fabricating dual-layer flat sheet membranes for gas separation, J. Membr. Sci. 375 (1–2) (2011) 258–267. [39] M. Mulder (Ed.), Basic Principles of Membrane Technology, Kluwer Academic Publishers Group, The Netherlands, 1996. [40] H. Mo, K.G. Tay, H.Y. Ng, Fouling of reverse osmosis membrane by protein (BSA): Effects of pH, calcium, magnesium, ionic strength and temperature, J. Membr. Sci. 315 (1–2) (2008) 28–35. [41] S. Singh, K.C. Khulbe, T. Matsuura, P. Ramamurthy, Membrane characterization by solute transport and atomic force microscopy, J. Membr. Sci. 142 (1) (1998) 111–127. [42] Q. Yang, T.-S. Chung, Y.E. Santoso, Tailoring pore size and pore size distribution of kidney dialysis hollow fiber membranes via dual-bath coagulation approach, J. Membr. Sci. 290 (1–2) (2007) 153–163. [43] S. Loeb, L. Titelman, E. Korngold, J. Freiman, Effect of porous support fabric on osmosis through a Loeb-Sourirajan type asymmetric membrane, J. Membr. Sci. 129 (2) (1997) 243–249. [44] T.Y. Cath, M. Elimelech, J.R. McCutcheon, R.L. McGinnis, A. Achilli, D. Anastasio, et al., Standard methodology for evaluating membrane performance in osmotically driven membrane processes, Desalination 312 (2013) 31–38. [45] C.H. Tan, H.Y. Ng, Modified models to predict flux behavior in forward osmosis in consideration of external and internal concentration polarizations, J. Membr. Sci. 324 (1–2) (2008) 209–219. [46] A.J. McHugh, C.S. Tsay, Dynamics of the phase inversion process, J. Appl. Polym. Sci. 46 (11) (1992) 2011–2021. [47] K.-V.P. Suzana Pereira Nunes, Membrane Technology: in the Chemical Industry, 2nd ed, Wiley-VCH, Federal Republic of Germany, 2006. [48] M. Jurgawezynski, Predicting Absolute and Relative Permeabilities of Carbonate Rocks using Image Analysis and Effective Medium Theory, University of London, London, 2007.

[49] R.W. Baker (Ed.), Membrane Technology and Applications, 2nd ed.,John Wiley & Sons, England, 2004. [50] C.A. Smolders, A.J. Reuvers, R.M. Boom, I.M. Wienk, Microstructures in phaseinversion membranes.1. Formation of microvoids, J. Membr. Sci. 73 (2-3) (1992) 259–275. [51] Membrane Filtration Guidance Manual, Office of Water, USEPA, 2003. [52] R.J. Petersen, Composite reverse-osmosis and nanofiltration membranes, J. Membr. Sci. 83 (1) (1993) 81–150. [53] G.M. Geise, H.B. Park, A.C. Sagle, B.D. Freeman, J.E. McGrath, Water permeability and water/salt selectivity tradeoff in polymers for desalination, J. Membr. Sci. 369 (1–2) (2011) 130–138. [54] P.S. Singh, S.V. Joshi, J.J. Trivedi, C.V. Devmurari, A.P. Rao, P.K. Ghosh, Probing the structural variations of thin film composite RO membranes obtained by coating polyamide over polysulfone membranes of different pore dimensions, J. Membr. Sci. 278 (1–2) (2006) 19–25. [55] Y.J. Song, P. Sun, L.L. Henry, B.H. Sun, Mechanisms of structure and performance controlled thin film composite membrane formation via interfacial polymerization process, J. Membr. Sci. 251 (1–2) (2005) 67–79. [56] G.T. Gray, J.R. McCutcheon, M. Elimelech, Internal concentration polarization in forward osmosis: role of membrane orientation, Desalination 197 (1–3) (2006) 1–8. [57] E.R. Cornelissen, D. Harmsen, K.F. de Korte, C.J. Ruiken, J.J. Qin, H. Oo, et al., Membrane fouling and process performance of forward osmosis membranes on activated sludge, J. Membr. Sci. 319 (1–2) (2008) 158–168. [58] W.A. Phillip, J.S. Yong, M. Elimelech, Reverse draw solute permeation in forward osmosis: modeling and experiments, Environ. Sci. Technol. 44 (13) (2010) 5170–5176. [59] A. Achilli, T.Y. Cath, A.E. Childress, Selection of inorganic-based draw solutions for forward osmosis applications, J. Membr. Sci. 364 (1–2) (2010) 233–241.