2013 - Preparation of Polyethersulfone_carbon Nanotube Substrate For

Desalination 330 (2013) 70–78

Contents lists available at ScienceDirect

Desalination

j ourna l homepage: www.e lsev ie r .com/ locate /desa l

Preparation of polyethersulfone/carbon nanotube substrate forhigh-performance forward osmosis membrane

Yaqin Wang a,b, Ranwen Ou a, Qianqian Ge a, Huanting Wang c,⁎, Tongwen Xu a,⁎a Functional Membrane Laboratory, School of Chemistry and Material Science, University of Science and Technology of China, Hefei, Anhui 230026, Chinab School of Materials and Chemical Engineering, Anhui Jianzhu University, Hefei, Anhui 230026, Chinac Department of Chemical Engineering, Monash University, Clayton, Victoria 3800, Australia

H I G H L I G H T S

• Forward osmosis (FO) membrane with MWCNT–PES nanocomposite substrate is fabricated.• MWCNT can modify the structure characteristics of substrate.• PES composite substrate containing MWCNT shows excellent FO performance.• The mechanical property of substrate is enhanced by the MWCNT.

⁎ Corresponding authors. Tel.: +86 551 3601587; fax:E-mail addresses: [email protected] (H. W

0011-9164/$ – see front matter © 2013 Elsevier B.V. All rihttp://dx.doi.org/10.1016/j.desal.2013.09.028

a b s t r a c t

Article history:Received 19 June 2013Received in revised form 28 August 2013Accepted 30 September 2013Available online 18 October 2013

Keywords:Forward osmosisPolyethersulfone membraneMultiwall carbon nanotubesNanocompositeThin film composite

Forward osmosis (FO) process has attracted increasing interest because of its potential applications for low-energy desalination. However, the internal concentration polarization (ICP) has been considered as one of thekey issues that can significantly reduce the water flux across the FO membrane. In this paper, we report thepreparation of polyethersulfone (PES)/multiwalled carbon nanotube (MWCNT) substrate for the formation ofa high-performance FOmembrane. NanocompositeMWCNT/PES substrateswere obtained by dispersing carbox-ylated MWCNTs within PES via solution blending, and subsequent phase inversion process; The FO membraneswere then prepared by depositing a polyamide active layer in-situ on the MWCNT/PES substrate with a finger-like macrovoid structure. The influence of addition of MWCNTs on morphology and properties of substratesand final FO membranes was systematically investigated. The results show that the performance of the FOmembranes with MWCNT/PES nanocomposite substrates is better than that of the commercial membrane.Furthermore, the tensile strength of the substrate with MWCNTs is also greater than that of the neat PES. Thiswork indicates that the FO membranes prepared from MWCNT–PES substrates are promising for practical FOapplications.

© 2013 Elsevier B.V. All rights reserved.

1. Introduction

Nowadays, unprecedented growth in demand for fresh water andenergy is a huge challenge for sustainable development of both societyand economy due to population bulge, rapid urbanization and climatechange [1]. Forward osmosis (FO) membrane process featuring lowerenergy consumption, and lower fouling tendency [2] than reverseosmosis (RO) process is an emerging technology for desalination [3,4],waste water treatment [4–6] and many other applications such as inpharmaceutical industry [7,8], agriculture and power generation[9,10]. FO may be a viable alternative to RO as a low-cost and moreenvironmentally friendly desalination technology if it was very wellintegrated with various processes [11]. However, the lack of high-

+86 551 3602171.ang), [email protected] (T. Xu).

ghts reserved.

performance membrane is a bottleneck in the development of the FOprocess. Similar to RO membranes, asymmetric aromatic polyamidemembranes consisting of an active layer, porous substrate layer and afabric layer fabricated by Loeb–Sourirajan process have been widelyinvestigated for FO. However, the porous substrate and fabric layer ofthe FO membrane usually give rise to a severe internal concentrationpolarization (ICP) and thus lower water flux in the FO process[12–18]. A good protocol has been proposed for the development ofFO membranes: firstly, sub-layer-free membranes may be well suitedfor FOmembranes [19]; secondly, the substrate layerwith low structureparameter (S)may be designed tominimize ICP [20]. The S valuemaybecalculated by

S ¼ tτε

ð1Þ

71Y. Wang et al. / Desalination 330 (2013) 70–78

where t, τ and ε are the membrane thickness, tortuosity and porosity,respectively. Seen fromEq. (1), a thinner, less tortuous andmore poroussubstrate will have a lower permeation resistance, and a higher waterpermeability. Recently, some researchers have reported the fabricationof optimal microstructure substrate layer with high porosity and lowtortuosity for FO membranes. Yip et al. [18] produced a substrate layerwith a mix of finger-like and sponge-like microstructure, which couldsignificantly improve water flux. Since ICP problem is mainly responsi-ble for water flux decline, the fabrication of a good FO membranerequires the minimization of ICP in the support layer. A scaffold-likenanofiber substrate fabricated by electrospinning technique successful-ly overcame this obstacle, and this unique structure offers direct pathsfor salt and water diffusion [21].

Polyethersulfone (PES) has beenwidely used asmembranematerial,especially for the fabrication of FOmembranes, due to its excellent ther-mal stability, mechanical properties, chemical resistance and wide pHtolerance [22,23]. On the other hand, multi-walled carbon nanotubes(MWCNTs) have superior separation capability as well as excellentphysical properties including high tensile moduli and strength, whichcan be used as potential fillers in the fabrication of nanocompositemembranes [24]. It was reported that a high loading of MWCNTs likelyresulted in a pore network in the membrane, and the hollownanochannels of MWCNTs and their interspaces could provide newtransport channels for water [25]. Also, it was hypothesized thatMWCNTs could improve water flow by disrupting polymer chain pack-ing, and creating external nanoscale channels in the membrane andadditional internal channels from opening nanotubes [25,26]. Important-ly, the rejection of contaminants using themembraneswithMWCNTfillermay be improved by forming a special porous substrate interface for po-lymerization of the active layer. In addition, the special structure ofMWCNTs makes them particularly attractive for reinforcement of com-posite materials [27]. However, the uniform dispersion of CNTs in a poly-mer matrix has been identified as a critical issue that must be addressedin the fabrication of high-performance membranes [28].

Herein,we report our attempt formodifying themicrostructure of PESsubstrate using MWCNTs as filler for the purpose for preparing FOmem-branes. The formation of such nanocomposites is expected to not onlystrengthen the substrate but also provide favorable microstructure forpromoting the separation properties in the FO process. In particular PESor MWCNT/PES composite substrate membranes with different loadingsof carboxylated MWCNTs were fabricated by solution blending andphase inversion method; m-phenylenediamine and 1,3,5-trime-soylchloridewere used as themonomers for the in situ polycondensationreaction to form a thin aromatic polyamide selective layer on thesubstrate surface. The effects of MWCNTs on the microstructure and FOperformance were systematically investigated.

2. Experimental

2.1. Materials

Multiwalled carbon nanotubes (MWCNTs, diameters of 10–20 nm,lengths of 1–5 μm, 95% purity) were purchased from SkySpringNanomaterials Inc. (USA). Polyethersulfone was acquired from BASF(Ultrason E3010, Germany). 1,3,5-Benzenetricarbonyle trichloride(TMC) with a purity of all over 98% was purchased from Qingdao SanliChemical Engineering Co., Ltd (Qingdao, China). N,N-Dimethylacetamide(DMAc,), polyvinylpyrrolidone (PVP K-30), 1,3-phenylenediamine(MPD) and all other reagents were of analytical grade and the productsof China National Pharmaceutical Group Industry Co., Ltd (Beijing,China) and used without further purification. Deionized water (DI) wasprepared using a Elix 5 pure water system (Millipore, Ltd., USA). Com-mercial asymmetric cellulose triacetate (HTI-M) FO membrane waskindly provided by Hydration Technology Innovations Inc. (HTI Co.USA), and it was used for comparison with the membranes prepared inthis study.

2.2. Modification of MWCNTs

Briefly, 0.1 g of MWCNT was suspended in a mixture of concentratedsulfuric acid and nitric acid (3:1 v/v), and then sonicated for 2 h in an ul-trasonic bath. The suspensionwas stirred at 60 °C for 6 h and subsequent-ly cooled to room temperature. The oxidizedMWCNTs (MWCNTs-COOH)were washed repeatedly with deionized water until the residual acidwas completely removed, followed by drying in a vacuum oven at60 °C for 4 h. The resulting MWCNTs were denoted as carboxylatedMWCNTs. More details about functional MWCNT characterization areshown in the Supporting Information.

2.3. Preparation of MWCNT/PES substrates

MWCNT/PES substrates were synthesized using the solution blend-ing and phase inversion technique. Carboxylated MWCNTs were dis-persed in DMAc. A homogeneous PES solution was obtained bydissolving PES in DMAc undermagnetic stirring for 4 h at room temper-ature. Subsequently, the suspension containing carboxylated MWCNTswas mixed with the PES solution. The mixture was then ultrasonicatedfor 1 h, and stirred for another 24 h to form a casting solution. Theresulting casting solution composed of 15 wt.% PES, 5 wt.% PVP, and0–2.5 wt.% carboxylated MWCNTs (on the basis of PES) in DMAc keptstill for 24 h for degassing, and then cast on a clean glass plate. Thesolvent was allowed to evaporate in air for 10 s. The glass plate wasthen immersed into a coagulant bath filled with a mixture of DMAcand deionized water (30/70, v/v) for 10 min at room temperature to in-duce phase separation. The resultant MWCNT/PES substrate waswashed thoroughly in a flowing water bath for 24 h, and stored in adistilled water bath. The substrates containing different amounts ofMWCNTs (0, 0.5, 1, 1.5, 2 and 2.5, wt.%) were denoted as S-0, S-0.5,S-1, S-1.5, S-2, and S-2.5, respectively.

2.4. Preparation of thin film composite (TFC) membrane

A rejection layer of FO membrane was formed by interfacialpolymerization of MPD and TMC on the dense side of the substratesS-0 ~ 2.5. Specially, the substrate was at first immersed into a 3.4 wt.%aqueous MPD solution for 2 min. And then, the excessive MPD solutionwas removed from the substrate membrane using tissue paper. Themembrane was finally dipped into a 0.15 wt.% TMC solution in hexanesolution for 1 min to form an ultrathin polyamide rejection layer onthe substrate. The resultant membrane was washed and stored in DIwater before use. The membranes prepared with S-0, S-0.5, S-1, S-1.5,S-2 and S-2.5 substrates were denoted as M-0, M-0.5, M-1, M-1.5, M-2and M-2.5, respectively.

2.5. Substrate and membrane characterization

The gravimetry was carried out to determine the volume porosity ofthe substrates, and water was usually selected as the wetting solvent.Generally, the total porosity ε (%) of the substrate was obtained bydigital subtraction of theweight of thewet (m1, g) anddry (m2, g)mem-branes, which was written as:

ε ¼ m1−m2ð Þ=ρH2O

m1−m2ð Þ=ρH2O þm2=ρpð2Þ

where ρH2O is the density of water, 1.0 g/cm3, and ρp is the density ofPES, 1.37 g/cm3.

The average pore radius (rm) of the substrate membrane wasestimated from the pure water volume flux (JW) and porosity data (ε)via the Guerout–Elford–Ferry equation (Eq. (3)):

rm ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2:9−1:75εð Þ � 8ηt JW

εΔP

rð3Þ

72 Y. Wang et al. / Desalination 330 (2013) 70–78

where η is the water viscosity (1.002 × 10−3 Pa s), t is the membranethickness (m) and ΔP is the operation pressure (0.12 MPa) [29]. Thepure water volume flux, JW (m/s) wasmeasured using a dead end filtra-tion device, where the feed pressure was controlled by compressed air.The compressed air was generated using an air compressor (Shanghaidynamic industry Co., Ltd, CHN). The area for water permeation in thisdevice is 15.9 cm2. Here, the unit of JW is m/s.

In order to examine variations in the surface wetting characteristicsof themembranes as a function ofMWCNT concentration,water contactangle (CA) between water and membrane surface measurement wasconducted using a contact angle meter (SL200B, Solon Tech Co., Ltd,CHN). Prior to performing the current and following characterization,all membranes were dried in the oven at 50 °C for 4 h. The value wasaveraged from 15 measurements at different parts of every sample.

Surface morphology and cross-section structure of the PES andMWCNT/PES composite substrate and thin film composite (TFC) mem-branes with polyamide active layers were qualitatively evaluated by afield emission scanning electron microscope (FE-SEM, Sirion200, FEICompany, USA) and an atomic force microscopy (AFM, Innova SPM,Veeco Company, USA).

The tensile strength of the substrates was characterized by DynamicMechanical Thermal Analyses (DMA, Q800, TA Company, USA) underisothermal conditions. All samples were tested with ramp force at anincreasing rate of 0.2500 N/min to 18,0000 N.

The thickness of the membrane was measured using a digitalmicrometer at four different locations for each sample.

2.6. FO membrane evaluation

2.6.1. Determination of purewater permeability (A) and solute permeabilitycoefficient (B)

Pure water permeability (A) was measured at different feedpressures (2, 3, 4 bar), and then determined by dividing the purewater volume flux (JW) by the applied pressure (ΔP).

A ¼ JWΔP

ð4Þ

The active layer NaCl permeability coefficient, B, was determinedusing a diaphragm cell experiment. 1000 ppm NaCl and DI water wereused as draw solution and feed respectively, and the experiment wascarried out in PRO mode. Assuming that the draw solution concentra-tion was constant, the B value was calculated by Eq. (5) [30]

B ¼ 1

Amt1

Vdrawþ 1V feed

� � lnC0draw−C0

feed

Ctdraw−Ct

feed

!ð5Þ

where Am is the membrane area, Vdraw and Vfeed are the volume of thedraw and the feed solution, respectively, and the concentrations ofNaCl in the draw (Cdraw) and feed (Cfeed) as a function of time (t) weremonitored using a conductivity meter.

2.6.2. Evaluation of osmotic water flux (JW)The FO performance (water flux and NaCl rejection) was evaluated

using a bench-scale FO setup with an effective membrane area of9.6 cm2 as described in Fig. S3. Both feed solution and draw solutionwere circulated at a fixed cross flow rate of 60 mL/min (2 cm/s). Themembrane active layerwas in contactwith the draw solution. The valuesof JW (L/m2 h, LMH) and R are the average values for 4 h. 0.01 mol/L NaClaqueous solutionwith about 0.49 bar osmotic pressure was as feed solu-tion and 2 mol/L glucose aqueous solution as draw solution with about73 bar osmotic pressure. The FO water flux JW was determined by

JW ¼ ΔVdraw

Am � Δtð6Þ

where ΔVdraw is the volume change of draw solution, and Δt is the mea-suring time interval.

2.6.3. Determination of structural parameter (S)Because the membrane orientation is feed solution against the po-

rous skin of membrane and draw solution against the dense skin side(PRO mode), S value was calculated by the following equation (Eq. (7))[31].

S ¼ DJW

lnBþ AπD;b− JW

Bþ Aπ F;bð7Þ

where πD,b and πF,b are the osmotic pressures of the bulk draw solutionand bulk feed solution, respectively, neglecting external polarizationeffects. D is the diffusion coefficient of the solute. For sodium chloride, Dis equal to 1.61 × 10−9 m2/s [32].While the thickness of themembraneswas measured by a digital micrometer, the ratio of τ/ε can be calculatedby

τε¼ S

t: ð8Þ

2.6.4. Determination of salt rejection (R)The NaCl rejection R in PROmodewas determined bymeasuring the

electrical conductivity of the feed and permeate collected from the FOsetup.

R ¼ C f−Cp

C f� 100% ð9Þ

Cp ¼ Ct � Vt−C0 � V0

Vt−V0ð10Þ

where Cf and Cp are the salt concentrations in the feed and permeate, re-spectively. C0 and Ct are the initial and final salt concentrations of drawsolution, respectively. V0 and Vt are the initial and final volumes of drawsolution, respectively.

3. Results and discussion

A systematic analysis of the relationship between the observed sub-strate structure and MWCNT concentration in the substrate will bepresented below. The results from the FO experiments will be used torelate the membrane performance to membrane structure. On theother hand, membrane mechanical strength was evaluated to furtherunderstand the effect of MWCNT addition on membrane property.

3.1. Porosity, pore size of substrate

The structure and physicochemical property of the substrate notonly play a key role in the structure parameter but also impact thepermeability and separation properties of the polyamide barrier layersubsequently formed via interfacial polymerization [13,33]. The porosityand average surface pore radius after the phase inversion are presentedin Table 1. It shows that with increasing the amount of MWCNTs, the po-rosity of substrate slightly increases in the beginning and then decreases.The optimal MWCNT concentration is 2 wt.%. This result indicates thatthe τ/ε value of PES membranes could be minimized by the addition ofMWCNTswith appropriate concentration, and therefore FO performanceof membranes might be enhanced.

As also shown in Table 1, the average pore radius of dense side sur-face of substrates gradually decreases with increasingMWCNT concen-tration. This result has a further implication that the water flux of thesubstrate membrane also has the similar tendency since pore size isproportional to the square root of water flux according to Eq. (3)

Table 1Porosity, surface mean pore size as a function of MWCNT concentration.

Substrates MWCNT concentration (%) Porosity (%) Pore size (nm)

S-0 0 85.6 ± 0.35 11.38 ± 0.16S-0.5 0.5 85.8 ± 0.4 11.32 ± 0.03S-1 1.0 86.1 ± 0.45 10.28 ± 1.92S-1.5 1.5 86.9 ± 0.1 10.07 ± 3.75S-2 2.0 87.7 ± 0.55 9.77± 0.07S-2.5 2.5 87.1 ± 0.32 9.14± 1.6

73Y. Wang et al. / Desalination 330 (2013) 70–78

when the other parameters of this equation are fixed. The experimentalresults show that the water permeation rate for the substrate mem-brane is about 0.24 mL/s when MWCNT concentration in PES is below1%. However, the rate begins to decrease with further increasing theMWCNTs to above 1%, i.e., the water permeation rates for S-2 andS-2.5 are 0.21 mL/s and 0.14 mL/s, respectively.

The smaller pores of the substrate presumably produced polyamidecomposites with smaller salt permeability [34]. Therefore, MWCNT/PEScomposite substrates could contribute to the permeation selectivity ofmembranes. According to themechanism of polyamide layer formation,the pore size affects MPD adsorption and diffusion into the interactionzone. The less porous substrate surface results in less absorption of thediamine monomer at the same adsorption time for MPD solution. Theacid chloride solution diffuses react with diamine monomer from theabsorbed aqueous layer, leading to formation of a thin polyamide. In ad-dition, the MPD diffusion rate can be limited with decreasing the poresize of the substrate. It should be noted that the polyamide selectivelayer seems to easily delaminate from the neat PES substrate layer. Onthe other hand, the resulting polyamide layer formed on MWCNT/PESsubstrate is better integrated with the substrate, indicating the betteradhesion with the substrate.

3.2. CA of substrate and membrane

Fig. 1 shows the CA of both sides of the final membrane as a functionof the MWCNT concentration in the substrate. In comparison with theCA of the dense side and the loose side, there is a striking differencebetween them. More porous surface of the substrate bottom likely re-sults in a decreased CA. The rougher polyamide formed on the densesurface of the substrate membrane is more hydrophilic, and it wouldfavor water permeability of the membrane.

However, the effect of addition of MWCNTs on the wetting perfor-mance is usually small in this study despite the fact that the MWCNTfiller has hydrophilic groups on the surface of nanotubes. The

0.0 0.5 1.0 1.5 2.0 2.50

20

40

60

80

MWNTs concentration (wt %)

Con

tact

ang

le(o )

Dense skin of substratePolyamide layer of membraneLoose skin of substrate

Fig. 1. Comparison of water contact angles of top layer, bottom layer and polyamideformed on the top layer between the different MWCNT loading contents.

phenomenon would be explained by comparing the surface SEMimage Fig. 2(A) of the PES membrane with that of the MWCNT/PESmembrane Fig. 2(B). These SEMmicrographs show thatMWCNTs almostcannot be found on the surfaces though the external feature of theseMWCNT/PES membranes is uniformly black. Therefore, the hydrophilic-ity of the membrane surface is still mainly determined by the polymermatrix property and surface morphology, rather than the MWCNTs; itmay be reasonable to consider that MWCNTs were wrapped by PVP orPES macromolecules.

In fact, PVP as an amphiphilic polymer was often used for dispersionof nanotubes because PVP with a hydrophobic alkyl backbone and hy-drophilic pendant groups can be envisaged to coil around the nanotubeso that its backbone is in contact with the nanotube surface and pyrrol-idone groups are exposed to water. In addition, non-covalent interac-tion between MWCNTs and PES can also contribute to the nanotubesbeingwrapped with PES [35]. When the dense surface layer was peeledoff from the substrate S-2 using a tape, Fig. 2(C) and (D) confirms thatMWCNTs wrapped in the polymer matrix. However, it is difficult toclarify howmuch PVP or PES on the surface of MWCNTs by the currentanalysis. Meanwhile, it can be seen that more pores were induced be-tween polymer chains and carbon nanotubes, and the microstructurecharacter also gave the cause for why porosity of the substrate mem-branes increased with increasing MWCNT fraction below 2 wt.%MWCNTs in PES. The interaction of PES with MWCNTs was confirmedby the mechanical strength in the following section.

3.3. Surface and cross-sectional morphologies of membranes

As shown in Fig. 3, SEM images indicate that the morphologies of allmembranes have an asymmetric structure, which consists of a densetop layer, a thick porous sublayer occupied by closed cells within thepolymer matrix and a layer with finger-like macro-voids. Moreover, amore open pore structure (macro-voids and large elongated pores)over thewhole cross section of theMWCNT/PESmembranewas obtain-ed at a higher concentration of MWCNTs. At a higher magnification, thenear top dense layer of the membrane, an improvement in the poreamount, a better connectivity between up-layer and sub-layer, anincrement in the porosity of the pore walls and a looser the sublayermembrane structure were observed. This desired microstructurewould decrease the structural parameter of the membrane by shorten-ing the transport path of the solute in the FO separation process.Probably, the presence of \OH and \COOH groups in MWCNTs(FT-IR spectra provided the evidence about these groups in MWCNTs,Fig. S2) and PVP molecules wrapping around MWCNTs affected thehydrophilicity property of the casting solution and changed the mem-brane formation during the phase inversion process. It is believed thatthe increase of MWCNTs affects the rates of non-solvent inflow andsolvent outflow. The rapid demixing in the MWCNT/PVP/PES blendand water is greater than that of PVP/PES and water due to the higheraffinity of water with the MWCNT/PVP/PES. This leads to the quick for-mation of a skin layer, creating an additional resistance tomass transfer,and results in a longer time for the entire exchange to take placebetween the non-solvent bath and the polymer casting film [36].Experimentally, the required time for the formation of S-1.5 and its de-tachment from the glass plate was about 34 s; whereas it took about30 s to form the PES membrane with the same dimensions of 10 cm ×10 cm. Increasing the exchange time between the solvent (DMAc) andnon-solvent (water) results in more development of the growth and co-alescence of the polymer-lean phase. Therefore, more open pore interiorstructure and porosity cell wall were obtained. As shown in previousstudies [23], nanoparticles as casting solution additives always affectthemembrane structure, for example, clay addition significantly affectedthemembrane internal and surface pore structures while it had no effecton porosity [37]. MWCNTs also modified the structure of the substratemembrane with more interconnecting pores as seen from the aboveSEM images.

Fig. 2. SEM micrographs displaying the top surface of S-0 (A) and S-2 (B), and interior of S-2 cross section (C) and top layer (D).

74 Y. Wang et al. / Desalination 330 (2013) 70–78

To isolate the effects of substrate structure on FO water flux and saltrejection, the samepolyamide formation protocolwas used in the fabrica-tion of all the membranes base on different substrates. Polyamide layerswere also found near the top surface of themembranes in these SEM im-ages. Their roughness measured by AFM was used for morphologicalcharacterization of polyamide layer formed on different substrates.

Fig. 4 shows the three-dimensional height of the membranes con-taining different amounts of MWCNTs. It shows that the mean squareroughness of the membranes firstly increases and then drops asMWCNTs are increased. For M-0, M-1, M-2 and M-2.5 membranes, themean square roughness is found to be 55 nm, 126 nm, 46.6 nm and46.5 nm, respectively, revealing the differences in the polyamide layersurface of the different substrates. When the concentration of MWCNTsin the casting solution is above 2 wt.%, the polyamide layer onthe surface of the MWCNT/PES substrate becomes smoother than thatof neat PES substrate. The decreased surface roughness was shown toimprove the antifouling properties by enhancing the hydrophilicity[38]. However, further improvement of smoothness through increasingthe amount of theMWCNTs in the substrate is not effective because theroughness only slightly varies with more addition of MWCNTs.

3.4. Performance of membranes

3.4.1. Pure water and salt permeability coefficient of membranesThe pure water and salt permeability coefficients of the synthesized

FO membranes in this study in comparison with a commercial CTA-Mmembrane are presented in Table 2. The results show that the waterpermeability of the TFC membranes is higher than that of the commer-cial membrane. Comparing the membranes with MWCNTs to the M-0,the M-1 shows the highest water permeability while the M-2.5 showsthe lowest water permeability. Even though the higher A value for theTFC membrane may result from low water diffusion resistance in thesubstrate [21], it demonstrates the water permeability under hydraulicpressure while it cannot appropriately reflect the FO performance of

membranes since the substrate support layer was an important pointfor optimization [13]. To further understand the impact of MWCNTaddition on the performance, osmosis experiment would be performed.In addition, except for M-0, other membranes yield lower salt perme-ability than CTA-M, which indicates that the TFC layer of these modifiedmembranes can serve well as a selective layer.

3.4.2. Osmotic flux performance and salt rejection of membranesIn order to evaluate osmosis performance of membranes with

MWCNTs in regard to brine water treatment, cross-flow experimentswere completed with simulated surface water. The influence of differ-ent substrates on the osmotic water flux and salt rejection in the FOprocess were investigated. Fig. 5 presents significant differencesamong the performances of the fabricated TFCmembranes. The averagewater flux of M-2.0 was 47% higher than that of M-0, and 33% higherthan that of HTI-M. Except for M-0, our other TFC membranes achievedhigher salt rejection than that of HTI-M, which further demonstratedthat the addition of MWCNTs into the PES substrate can lead to betterperformance of the FO membrane.

Compared to the commercial membrane, higher water flux of mem-branes with MWCNT–PES substrates may mainly attribute to threereasons in the present study. Firstly, ICP as an unfavorable phenomenonin the osmotic process was mitigated due to smaller structure parame-ters, whichwill be discussed in the next section. Secondly, reverse drawsolute flux that can not only waste draw agent but also make ICP moresevere was reduced because the TFC membranes have greater soluterejection as presented in Table 2 and Fig. 5. The last but not least, asshowed in SEM images, M-2 andM-2.5 have a much smoother polyam-ide layer than the others, and this characteristic contributes to lessmembrane fouling with the active layer against the high concentrationglucose solution.

These fabricated composite membranes except for M-0 exhibitedexcellent NaCl rejections. For example, a maximum value of 97% wasachieved for M-1.5 while the 89% salt rejection was measured for the

20 µm

20 µm

20 µm

20 µm

Cross section of M-0

Cross section of M-1

Cross section of M-2

Cross section of M-2.5

Near top surface of M-0

Near top surface of M-1

Near top surface of M-2

Near top surface of M-2.5

1 µm

1 µm

1 µm

1 µm

Fig. 3. SEM images showing cross section morphologies of different substrates. Left is the cross section of whole membranes, and right is the cross section of near top surface.

75Y. Wang et al. / Desalination 330 (2013) 70–78

HTI-M membrane. As for M-0, 78% was achieved at the same experi-mental conditions. As shown in SEM images, the substrate S-0 had awide pore-size, and it may induce the salt leakage to the draw solutionacross M-0 because the polyamide layer does not cover completely on

the surface of the substrate. It seemed that MWCNT–PES supported FOmembranes could overcome the trade-off between the water flux im-provement and increased salt retention formembrane used in the desa-lination process.

M-0 M-1

M-2 M-2.5

SEM image of M-0

Fig. 4. Tapping-mode AFM three-dimension height images of polyamide layers of different membranes and SEM image of membrane M-0. The fields of AFM view are 5 × 5 μm.

80

100

120

10

12

14

16 (

%)

MH

)

water flux salt rejection

76 Y. Wang et al. / Desalination 330 (2013) 70–78

3.4.3. Structure parameters of membranesThe structural parameter calculated from osmotic flux tests implied

how severe the ICP effect and it should also be as low aspossible tomax-imize the water flux [39]. Fig. 6 shows that the S of our TFC membranesgradually decreased from 3939 μm to 2042 μm with increasing theMWCNT loading in the substrate. By contrast CTA-M had an S of3074 μm at the same FO experimental conditions. This suggests thatthe structure of the MWCNT/PES substrate is more favorable for higherwater flux performance when the concentration of MWCNTs in castingsolution is under 2 wt.%whilemoreMWCNT additionwould play a neg-ative role in the FO performance. The ratio of τ/ε (S was divided by the

Table 2Thickness, pure water permeability coefficient (A) and salt permeability coefficient (B) ofTFC membrane based on different substrates.

Membranes Thickness (×10−6 m) A (LMH/bar) B (×10−7 m/s)

M-0 85.3 ± 15.2 2.42 ± 0.29 4.2 ± 1.4M-0.5 95.1 ± 13.9 2.69 ± 0.16 2.3 ± 0.7M-1 79.5 ± 0.7 2.54 ± 0.07 2.1 ± 0.1M-1.5 85.5 ± 10.6 2.37 ± 0.33 0.48 ± 0.7M-2 90.1 ± 2.8 2.31 ± 0.18 2.2 ± 0.9M-2.5 86.3 ± 10.9 1.92 ± 0.24 1.4 ± 0.6HTI-M 89.7 ± 2 0.64 ± 0.09 2.6 ± 0.6

membrane thickness, t, that is, the ratio of τ/ε was instead of S) maybe a better indicator of the inherent resistance to diffusion providedby the structure [13], since it can directly reflect the effect of membrane

0

20

40

60

0

2

4

6

8

HTI-M M-0 M-0.5 M-1 M-1.5 M-2 M-2.5

Salt

rej

ecti

on

Wat

er f

lux

(L

Fig. 5. Comparison of FO water flux and salt rejection of membrane HTI-M, neat PES andMWCNT/PES based membranes.

0 5 10 15 20 25 300

1

2

3

4

5

Stre

ss (

MP

a)

Strain (%)

0.0

1.5

2.5

Fig. 7. Stress–strain profiles of MWCNT/PES substrates at different MWCNT loadings. Thecurves are labeled with the percentage of MWCNT in the PES matrix.

77Y. Wang et al. / Desalination 330 (2013) 70–78

intrinsic microstructure on the FO performancewhile the thickness wasfeasibly controlled by adjusting the height of the casting knife. Similarlyincreasing the MWCNT concentration in the substrate caused the ratioto decline due to more open pore structure of the substrate and muchsmoother polyamide surface; then the ratio of τ/ε for M-2.5 was slightlyincreased due to the reduction of porosity and dramatic decline of poresize of the substrate.

In general, the FOperformance ofmembranes based onMWCNT/PESmembranes was consistentwith their microstructure characteristic andmorphologies. Higher porosity, more open interior pores and smoothermorphology of polyamide layerwould result in less structure parameterand water flux enhancement. The ratio of τ/ε also proved that themicrostructure of M-2 is optimal for the FO process in terms of osmoticwater flux.

3.5. Mechanical property of substrates

The mechanical property of the substrate was seldom discussed inthe literature because of a conventional membrane with a fabric layeras a mechanical support, and the porous substrate only functions as amedium for the formation of active selective layer. The tensile strengthof the substrates with different MWCNT loadings is shown in Fig. 7. Itcan be observed that the strength generally increased with MWCNTloading. For example, the strength of the substrate with 2.5 wt.% ofMWCNT loading was two times higher than that of the neat PES, andtherefore, the incorporation of MWCNTs as a reinforcing agent in PESis helpful for fabricating fabric-free substrate support. The substratewith higher MWCNT concentration tends to have higher strength.However, as revealed by FO performance measurement experiments,it is unfavorable for the osmotic flux of the membranes. Accordingly, abalancemust bemade between the strength and the FO flux in practicalapplications.

4. Conclusions

The correlations among the surfacemorphology, surface and interiormicrostructure of the different substrates, and themembrane FOperfor-mance of the final corresponding membranes were presented in thisstudy. The controlling factor for the performance and the structure ofMWCNT/PES based FO membranes was the MWCNT concentration incasting solution. When the MWCNT/PES composite substrate mem-brane was used for desalination in the FO process, the improvementsin both salt rejection and water permeability were demonstrated; this

1

2

3

4

Stru

ctur

e pa

ram

eter

S(

10-3

m)

Rat

io o

f to

rtuo

sity

to

poro

sity

s

Membranes

25

30

35

40

45

50

HTI-M M-0 M-0.5 M-1 M-1.5 M-2 M-2.5

Fig. 6. Comparison of structure parameters of membrane HTI-M, neat PES and MWCNT/PES based membranes. The data from commercial membrane and the optimal TFC mem-brane were emphasized by pink and yellow bars, respectively.

wasmainly because themore open interior pore structure and smooth-er selective layer resulted in less ICP.

(1) The membranes with appropriate amounts of MWCNTs weremuch porous than that without MWCNTs, resulting in significantly de-creased internal concentration and higher osmotic water flux; (2) Allmembranes with MWCNTs showed NaCl rejection rate over 90%, andthe maximum value was 97%; their NaCl rejection rates were greaterthan those of neat PES (78%) and the commercial FO membrane (89%)at the same testing condition; and (3) The tensile strength of themem-branes was improved by loadingMWCNTs. Therefore, theMWCNT–PESnanocomposite membranes are very promising for the development ofhigh-performance FO membranes for practical applications.

Acknowledgments

This work was supported by the National Natural Science Founda-tion of China (Nos. 21128004 and 21025626). H. W. thanks theAustralian Research Council for a Future Fellowship.

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.desal.2013.09.028.

References

[1] M. Hightower, The energy challenge, Nature 452 (2008) 285–286.[2] C. Boo, S. Lee, M. Elimelech, Z.Y. Meng, S. Hong, Colloidal fouling in forward osmosis:

role of reverse salt diffusion, J. Membr. Sci. 390 (2012) 277–284.[3] B. Zhang, Z. He, Integrated salinity reduction and water recovery in an osmotic

microbial desalination cell, Rsc Adv. 2 (2012) 3265–3269.[4] T.Y. Cath, D. Adams, A.E. Childress, Membrane contactor processes for wastewater

reclamation in space II. Combined direct osmosis, osmotic distillation, andmembrane distillation for treatment of metabolic wastewater, J. Membr. Sci. 257(2005) 111–119.

[5] T.Y. Cath, S. Gormly, E.G. Beaudry, M.T. Flynn, V.D. Adams, A.E. Childress, Membranecontactor processes for wastewater reclamation in space Part I. Direct osmotic con-centration as pretreatment for reverse osmosis, J. Membr. Sci. 257 (2005) 85–98.

[6] R.W. Holloway, A.E. Childress, K.E. Dennett, T.Y. Cath, Forward osmosis for concen-tration of anaerobic digester centrate, Water Res. 41 (2007) 4005–4014.

[7] K.M. Talaat, Forward osmosis process for dialysis fluid regeneration, Artif. Organs 33(2009) 1133–1135.

[8] K.Y. Wang, M.M. Teoh, A. Nugroho, T.-S. Chung, Integrated forward osmosis–membrane distillation (FO–MD) hybrid system for the concentration of proteinsolutions, Chem. Eng. Sci. 66 (2011) 2421–2430.

[9] A. Achilli, T.Y. Cath, A.E. Childress, Power generation with pressure retardedosmosis: an experimental and theoretical investigation, J. Membr. Sci. 343 (2009)42–52.

[10] K. Lee, R. Baker, H. Lonsdale, Membranes for power generation by pressure-retardedosmosis, J. Membr. Sci. 8 (1981) 141–171.

78 Y. Wang et al. / Desalination 330 (2013) 70–78

[11] J.R. McCutcheon, R.L. McGinnis, M. Elimelech, Desalination by ammonia–carbondioxide forward osmosis: influence of draw and feed solution concentrations onprocess performance, J. Membr. Sci. 278 (2006) 114–123.

[12] S. Loeb, L. Titelman, E. Korngold, J. Freiman, Effect of porous support fabric on osmo-sis through a Loeb–Sourirajan type asymmetric membrane, J. Membr. Sci. 129(1997) 243–249.

[13] A. Tiraferri, N.Y. Yip, W.A. Phillip, J.D. Schiffman, M. Elimelech, Relating performanceof thin-film composite forward osmosis membranes to support layer formation andstructure, J. Membr. Sci. 367 (2011) 340–352.

[14] C.Y.Y. Tang, Q.H. She, W.C.L. Lay, R. Wang, R. Field, A.G. Fane, Modelingdouble-skinned FO membranes, Desalination 283 (2011) 178–186.

[15] K.Y. Wang, R.C. Ong, T.S. Chung, Double-skinned forward osmosis membranes forreducing internal concentration polarization within the porous sublayer, Ind. Eng.Chem. Res. 49 (2010) 4824–4831.

[16] C.H. Tan, H.Y. Ng, Modified models to predict flux behavior in forward osmosis inconsideration of external and internal concentration polarizations, J. Membr. Sci.324 (2008) 209–219.

[17] J.R. McCutcheon, R.L. McGinnis, M. Elimelech, A novel ammonia–carbon dioxideforward (direct) osmosis desalination process, Desalination 174 (2005) 1–11.

[18] N.Y. Yip, A. Tiraferri, W.A. Phillip, J.D. Schiffman, M. Elimelech, High performancethin-film composite forward osmosis membrane, Environ. Sci. Technol. 44 (2010)3812–3818.

[19] S. Zhang, R.W. Zhang, Y.C. Jean, D.R. Paul, T.S. Chung, Cellulose esters for forwardosmosis: characterization of water and salt transport properties and free volume,Polymer 53 (2012) 2664–2672.

[20] N.-N. Bui, J.R. McCutcheon, Hydrophilic nanofibers as new supports for thin filmcomposite membranes for engineered osmosis, Environ. Sci. Technol. 47 (2013)1761–1769.

[21] X. Song, Z. Liu, D.D. Sun, Nano gives the answer: breaking the bottleneck of internalconcentration polarization with a nanofiber composite forward osmosis membranefor a high water production rate, Adv. Mater. 23 (2011) 3256–3260.

[22] E. Celik, H. Park, H. Choi, H. Choi, Carbon nanotube blended polyethersulfonemembranes for fouling control in water treatment, Water Res. 45 (2011) 274–282.

[23] A.L. Ahmad, A.A. Abdulkarim, B.S. Ooi, S. Ismail, Recent development in additivesmodifications of polyethersulfone membrane for flux enhancement, Chem. Eng. J.223 (2013) 246–267.

[24] Y. Shirazi, M.A. Tofighy, T. Mohammadi, Synthesis and characterization of carbonnanotubes/poly vinyl alcohol nanocomposite membranes for dehydration ofisopropanol, J. Membr. Sci. 378 (2011) 551–561.

[25] M.S. Mauter, M. Elimelech, Environmental applications of carbon-based nan-omaterials, Environ. Sci. Technol. 42 (2008) 5843–5859.

[26] X. Wang, X. Chen, K. Yoon, D. Fang, B.S. Hsiao, B. Chu, High flux filtration mediumbased on nanofibrous substrate with hydrophilic nanocomposite coating, Environ.Sci. Technol. 39 (2005) 7684–7691.

[27] M. Moniruzzaman, K.I. Winey, Polymer nanocomposites containing carbonnanotubes, Macromolecules 39 (2006) 5194–5205.

[28] X.L. Xie, Y.W. Mai, X.P. Zhou, Dispersion and alignment of carbon nanotubes in poly-mer matrix: a review, Mater. Sci. Eng. R 49 (2005) 89–112.

[29] J. Zhang, Z. Xu, W. Mai, C. Min, B. Zhou, M. Shan, Y. Li, C. Yang, Z. Wang, X. Qian, Im-proved hydrophilicity, permeability, antifouling and mechanical performance ofPVDF composite ultrafiltration membranes tailored by oxidized low-dimensionalcarbon nanomaterials, J. Mater. Chem. A 1 (2013) 3101–3111.

[30] J.S. Yong, W.A. Phillip, M. Elimelech, Coupled reverse draw solute permeation andwater flux in forward osmosis with neutral draw solutes, J. Membr. Sci. 392(2012) 9–17.

[31] T.Y. Cath, A.E. Childress, M. Elimelech, Forward osmosis: principles, applications, andrecent developments, J. Membr. Sci. 281 (2006) 70–87.

[32] G.T. Gray, J.R. McCutcheon, M. Elimelech, Internal concentration polarization in for-ward osmosis: role of membrane orientation, Desalination 197 (2006) 1–8.

[33] K.Y. Wang, T.-S. Chung, G. Amy, Developing thin-film-composite forward osmosismembranes on the PES/SPSf substrate through interfacial polymerization, AIChE J58 (2012) 770–781.

[34] M. Fathizadeh, A. Aroujalian, A. Raisi, Effect of lag time in interfacial polymerizationon polyamide composite membrane with different hydrophilic sub layers, Desalina-tion 284 (2012) 32–41.

[35] H. Wang, G. Wang, W. Li, Q. Wang, W. Wei, Z. Jiang, S. Zhang, A material with highelectromagnetic radiation shielding effectiveness fabricated using multi-walled car-bon nanotubes wrapped with poly (ether sulfone) in a poly (ether ether ketone)matrix, J. Mater. Chem. 22 (2012) 21232–21237.

[36] Y. Mansourpanah, S.S. Madaeni, A. Rahimpour, M. Adeli, M.Y. Hashemi, M.R.Moradian, Fabrication new PES-based mixed matrix nanocomposite membranesusing polycaprolactone modified carbon nanotubes as the additive: propertychanges and morphological studies, Desalination 277 (2011) 171–177.

[37] J.C. Mierzwa, V. Arieta, M. Verlage, J. Carvalho, C.D. Vecitis, Effect of claynanoparticles on the structure and performance of polyethersulfone ultrafiltrationmembranes, Desalination 314 (2013) 147–158.

[38] Y.-N. Kwon, S. Hong, H. Choi, T. Tak, Surface modification of a polyamidereverse osmosis membrane for chlorine resistance improvement, J. Membr. Sci.415–416 (2012) 192–198.

[39] L. Huang, N.-N. Bui, M.T. Meyering, T.J. Hamlin, J.R. McCutcheon, Novel hydrophilicnylon 6,6 microfiltration membrane supported thin film composite membranesfor engineered osmosis, J. Membr. Sci. 437 (2013) 141–149.