Transport, structural, and interfacial properties of poly(vinyl alcohol)–polysulfone composite nanofiltration membranes

Journal of Membrane Science 353 (2010) 169–176

Contents lists available at ScienceDirect

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

Transport, structural, and interfacial properties of poly (vinyl alcohol)–polysulfone composite nanofiltration membranes Fubing Peng a,b , Xiaofei Huang a,b , Anna Jawor a,b , Eric M.V. Hoek a,b,∗ a b

Department of Civil & Environmental Engineering, University of California, Los Angeles, CA, USA California NanoSystems Institute, University of California, Los Angeles, CA, USA

a r t i c l e

i n f o

Article history: Received 27 October 2009 Received in revised form 3 February 2010 Accepted 15 February 2010 Available online 20 February 2010 Keywords: Nanofiltration Poly(vinyl alcohol) Hydrogel Cross-linking Water treatment

a b s t r a c t Composite nanofiltration membranes were prepared by coating poly(vinyl alcohol) hydrogels on polysulfone ultrafiltration support membranes. Ultra-thin and defect-free poly(vinyl alcohol) hydrogels were cast using multi-step coating procedure with dilute PVA aqueous solutions and novel in situ cross-linking. The combined Spiegler–Kedem–film theory model was used to extract water permeability, solute permeability, reflection coefficients, and mass transfer coefficients for the composite membranes. Transport of water and salt ions through PVA coating films was dramatically influenced by feed solution pH and counter-ion valence as well as PVA molecular weight, concentration, and extent of cross-linking. Characterization of PVA coating film thickness, extent of cross-linking, surface thermodynamic properties, and crystallinity were used to explain differences in observed transport properties of the composite membranes. © 2010 Elsevier B.V. All rights reserved.

1. Introduction Nanofiltration (NF) membranes provide separation performance between reverse osmosis (RO) and ultrafiltration (UF) membranes, which makes them useful for separating toxic metals, hardness ions, oil emulsions, and dissolved organics from water. More stringent water quality regulations and decreasing availability of pristine fresh water resources demands improved water purification methods. Nanofiltration (NF) has become a popular technology to augment conventional water treatment processes because of its low operating pressure, high retention of multivalent ions and dissolved organic molecules larger than about 300 Da, and relatively low costs [1,2]. In the past several decades, a wide variety of interfacial composite membranes were developed around the world [3]. Currently, two methods are widely used to prepare composite membranes—dip-coating and interfacial polymerization. Polyamide composite nanofiltration membranes have been successfully commercialized with high water flux, good rejection of multivalent ions, and low rejection of monovalent salts. Polyamide interfacial composites are prepared by the interfacial polymerization of multifunctional amine and acyl chloride monomers, that form the amide linkage. Polyamide nanofiltration membranes have

∗ Corresponding author at: University of California, Los Angeles, Department of Civil & Environmental Engineering, 5732-G Boelter Hall, P.O. Box 951593, Los Angeles, CA 90095-1593, USA. Tel.: +1 310 206 3735; fax: +1 310 206 2222. E-mail address: [email protected] (E.M.V. Hoek). 0376-7388/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.memsci.2010.02.044

low chemical stability and poor chlorine tolerance, which restricts their application [16]. Therefore, TFC membranes with high chemical stability, good chlorine tolerance, and low fouling need to be developed [2,16]. In this study, the poly(vinyl alcohol) hydrogels were formed by dip-coating polysulfone ultrafiltration membranes to produce interfacial composite nanofiltration membranes. Poly(vinyl alcohol) (PVA) is used to make membranes for a number of separation process, especially pervaporation membranes [4,5], due to its high water uptake, good physical and chemical stability, low cost, commercial availability, and good membrane-forming properties. Poly(vinyl alcohol) is also used as a protective surface coating on the top of composite membranes because of its hydrophilic character, and hence, fouling resistance [3]. To make stable PVA films or films with a desired salt selectivity, PVA can be cross-linked through a variety of methods [6]. There are many past studies of PVA membranes developed for reverse osmosis [7–15] and nanofiltration [16–18] separations. Most of these membranes give relatively low flux and rejection, which might be caused by unsuitable coating methods, improper cross-linking reaction, or excessive coating layer thickness. These past attempts use the dip-coating method followed by surface cross-linking to prepare PVA membranes, which leads to relatively thick and minimally cross-linked PVA layers [19]. In this study, composite PVA–polysulfone (PSf) nanofiltration membranes were formed by coating and cross-linking ultra-thin, defect-free PVA films on polysulfone ultrafiltration membrane supports. First, PVA layers were prepared by applying multiple coats of dilute PVA solutions onto PSf support membranes. Second, in

170

F. Peng et al. / Journal of Membrane Science 353 (2010) 169–176

situ cross-linking was used to tailor the water and salt permeability of the coating films. Membranes prepared using different coating film formulations and cross-linking conditions were rigorously characterized to elucidate differences in transport, structural, and interfacial properties.



R Jw Ro = exp − 1 − Ro 1−R k

Early membrane transport models were based on the principle of irreversible thermodynamics, which assumes the membrane is not far from equilibrium and so fluxes are derived from phenomenological thermodynamic relationships. One of the early models was proposed by Kedem and Katchalsky [20]. The disadvantage of the Kedem–Katchalsky model is that the coefficients are functions of concentration. To avoid concentration dependence of the transport parameters, Kedem and Spiegler [21] developed the nonlinear Spiegler–Kedem model by defining fluxes in terms of local water and solute permeability coefficients as Jw = Lp (p − ) ,

(1)

Js = Ps y

(2)

 dc  dy

+ (1 − ) Jw c,

where Jw and Js are the water and solute fluxes, Lp and Ps are the local water and solute permeability coefficients, p is the transmembrane hydraulic pressure,  is the trans-membrane osmotic pressure,  is the reflection coefficient, c is the solute concentration, and y is the membrane thickness. The reflection coefficient represents the intrinsic rejection by a membrane when an infinite pressure is applied. A reflection coefficient of zero indicates complete solute passage, while  = 1 indicates complete rejection. In the latter case, Eq. (2) reduces to the solution-diffusion model typically applied to RO membrane separations [22]. Therefore, the Spiegler–Kedem model is (for practical purposes) the solution-diffusion model with an additional component to account for convective solute transport through NF membrane pores. Now, Eq. (2) can be integrated across the membrane thickness to yield cp = cm

 1−F  1 − F

 J  w (1 − )

F = exp −

Ps

(3)

Concentration polarization describes the elevated concentration of rejected solutes at the membrane surface relative to that of the bulk feed solution. Based on film theory and the relevant boundary conditions (c = cf at x = 0, c = cm at x = ı, where ı is the boundary layer thickness), a solute mass balance across the membrane gives [24] w

k

,

(6)

 R = (1 − F) , 1−R 1−

(7)

and then, substituting Eqs. (7) and (4) into Eq. (6) to yield  Ro = 1 − Ro 1−



 J  w (1 − )

1 − exp −

Ps

 J  w

exp −

k

.

(8)

Eq. (8) is the working equation for combined Spiegler–Kedem–film theory model. Transport parameters were determined from the experimental flux and rejection data as follows. First, Lp was determined from filtration experiments using pure water at different pressures. Second, , Ps , and k were determined simultaneously via a nonlinear parameter estimation method using the Ro and Jw data measured at different pressures, but at constant feed flow rate and constant feed concentration. The nonlinear parameter estimation, based on nonlinear least-square fitting of the data, was executed using Matlab. 3. Experimental 3.1. Chemicals and materials Mowiol® PVA 4-98, 6-98, and 10-98 with average molecular weights of 27,000, 47,000, and 61,000 g/mol, respectively, 98.0–98.8% hydrolyzed, was purchased from Sigma–Aldrich Company for the formation of active layers of the NF composite membranes. Commercial polysulfone ultrafiltration membranes (NanoH2 O Inc., Los Angeles, CA, USA) were used as supports on which the PVA films were cast. Succinic acid (>99%, Sigma–Aldrich, St. Louis, MO, USA) was used as the cross-linking agent. All membranes were made with PVA 6-98 unless otherwise specified. A commercial nanofiltration membrane (NF270, Dow Water Solutions, Midland, MI, USA) was tested as comparison.

(4)

2.2. Combined Spiegler–Kedem–film theory model

J 

.



3.2. Membrane preparation .

Here, R is the intrinsic rejection by the membrane, cp is the solute concentration at the permeate side of the membrane–solution interface, and cm is the solute concentration at the feed side of the membrane–solution interface. Assuming the intrinsic rejection can be estimated from the observed permeate concentration and a computed membrane concentration, the values of  and Ps can be determined [23].

cm − cp = exp cf − cp



Here Ro = 1 − cp /cf is the observed (measurable) rejection. The intrinsic transport properties of a membrane are determined by re-arranging Eq. (3) to

2.1. Spiegler–Kedem membrane transport model

and





2. Theory

R=1−



where k = Dı is the solute mass transfer coefficient. Upon rearrangement, Eq. (5) relates the observed rejection and the true rejection to the flux and mass transfer coefficient via

(5)

PVA powder was dissolved in DI water at 90 ◦ C using mechanical stirring for about 60 min to make PVA aqueous solutions. Unless otherwise specified, the PVA molecular weight was 47 kDa and the PVA concentration was 0.10 wt.%. Next, PVA solutions were cooled to room temperature and the cross-linking agent was added along with 2 M HCl as catalyst under continuous stirring to produce the PVA casting solution. Succinic acid concentration was selected to produce a theoretical cross-linking degree of 20% unless otherwise specified. The theoretical cross-linking degree was defined by CL [%] =

WCL × MWPVAunit × 2 × 100, WPVA × MWCL

(9)

where WCL , WPVA , MWPVAunit , and MWCL represented the weight of cross-linking agent, the weight of PVA, the molecular weight of one PVA unit (–CHOH–CH2 –), and the molecular weight of the crosslinking agent, respectively. The polysulfone support membranes were taped onto the glass plate, and only the membrane surface side was contacted with PVA solution in the dip-coating process. Poly(vinyl alcohol) casting

F. Peng et al. / Journal of Membrane Science 353 (2010) 169–176

171

solutions were coated onto polysulfone ultrafiltration membranes three times. First, the casting solution was poured onto the PSf support and let sit for 10 min. Then, the solute was drained and the remaining water was allowed to evaporate at room temperature over night (24 h). Next, the coated membrane was contacted with the same PVA solution for 10 s and air-dried for 24 h again. The 10 s coating and drying was repeated to produce defect-free ultra-thin PVA coating layers. The PVA coated polysulfone membranes were then cured at 100 ◦ C for 10 min. 3.3. Membrane characterization The morphology and thickness of the PVA active layers of the composite membranes were characterized with Nova 600 DualBeamTM FIB-SEM (FEI Company, Hillsboro, Oregon). For some PVA–PSf composite membrane samples, cross-sectional SEM images were used to usually estimate PVA film layer thickness. Using the SEM scale bar, we measured the distance between the surface and the top of the first visible pore in the PSf layer at 10 different locations. The slope from the plot of measured water permeability versus measured film thickness provided the thickness independent pure water permeability of each PVA film composition. The extent of cross-linking of PVA coating layers was confirmed by attenuated total reflection infrared spectroscopy (ATR-IR) performed on a Jasco FTIR 670 plus with variable angle ATR attachment coupled to a germanium crystal operated at a 45◦ . Prior to the ATR-IR measurement, the samples were dried in a desiccator for a minimum of 24 h. Crystallinity of PVA coating films were observed using X-ray diffraction, XRD (Brüker AXS D8 diffractometer, Germany, using Cu-K␣ radiation). The membrane surface hydrophilicity, surface tensions, and interfacial free energies were determined from measured contact angles using an automated contact angle goniometer (DSA0 KRÜSS GmbH, Hamburg, Germany). At least 12 equilibrium contact angles were measured for each sample. The highest and lowest values were discarded before taking the average and standard deviation. Contact angle measurements for deionized water (polar liquid), diiodomethane (apolar liquid), and glycerol (polar liquid) enable determination of interfacial tension parameters using the extended Young–Dupre equation [25]. 3.4. Separation performance tests The separation performance of PVA–PSf composite membranes was evaluated in a bench scale cross-flow membrane filtration system equipped with six parallel membrane cells (effective membrane area is 12.9 cm2 for each membrane cell). The system was described in detail elsewhere [26]. Pure water flux of polysulfone and PVA–PSf membranes were determined using 18 M laboratory deionized water at 25 ◦ C and applied pressures of 173 and 1034 kPa (25 and 150 psi), respectively. The cross-flow Reynolds number was maintained at 312 without no mesh spacer in the feed channel. Flux was measured by a digital flow meter. Nanofiltration membrane selectivity for NaCl or Na2 SO4 was characterized by evaluating the conductivity rejection of 2000 ppm NaCl or Na2 SO4 solutions individually. Conductivity calibration curves were linear for concentration between 0 and 2000 ppm of these salts; hence, observed rejections calculated directly from feed and permeate conductivities. All reported flux and rejection data represent the averages of at least three separate tests of membranes hand-cast on three different days using independently prepared PVA coating solutions.

Fig. 1. Pure water permeability and salt rejections by PVA–PSf composite membranes at pH 7.0 and 25 ◦ C.

4. Results and discussion 4.1. Evaluating nanofiltration-like separation performance Fig. 1 presents permeability and rejection data for pure water, NaCl, and Na2 SO4 solution with feed pressure through PVA–PSf composite membranes. Pure water flux and solute rejection were measured after the PVA–PSf composite membrane compacted at 1724 kPa (250 psi) for 3 h. Both pure water flux and solute rejection were relatively stable over the range of applied pressures considered. In principle, flux is proportional to feed pressure and inversely proportional to membrane thickness in membrane (nanofiltration). Any operating condition that produces higher flux will generally increase the observed solute rejection—this is the “dilution effect”. However, the pure water permeability was relatively constant with pressure. The commercial nanofiltration membrane (Dow NF270) was tested in our cross-flow membrane filtration system. The pure water permeability was 31 ␮m MPa s−1 and the rejections of NaCl and Na2 SO4 were 51% and 94%, respectively. For the PVA–PSf composite membrane used to test the effect of pressure, the pure water permeability was only 10.4 ␮m MPa s−1 with NaCl and Na2 SO4 rejections of 37.4% and 90.0%, respectively. Here the lower flux of the PVA–PSf composite might be compensated by the larger differential in NaCl/Na2 SO4 separation, in addition to the better stability expected for PVA over polyamides. In addition, previous studies of PVA membranes have reported fluxes and rejections of 40 l/m2 h and 95% for Na2 SO4 at 200 psi [16], 8 l/m2 h and 25% for NaCl at 150 psi [17], and 6 l/m2 h and >60% for NaCl at 150 psi [18]. Therefore, it is clear that the polyamide composite membranes gave higher flux and excellent differential salt rejection, but the PVA membranes reported here are among the highest flux membranes ever reported. 4.2. Membrane transport parameters and mass transfer coefficients Experimental results from permeation tests described above were used to estimate the membrane transport and mass transfer coefficients. The model-fitted data are presented in Fig. 2 and transport parameters estimated from these data are given in Table 1. Solute permeability coefficient, reflection coefficient, and mass transfer coefficient were different for each solute. Model predictions agreed reasonably well with the experimental results. Model-fitting results offer insight into the key mechanism of sep-

172

F. Peng et al. / Journal of Membrane Science 353 (2010) 169–176

9.0). Here, the increase in rejection was apparently due to greater Donnan exclusion at high pH, rather than structural changes in the film layer, which would have also significantly altered the water flux. Incomplete PVA cross-linking reaction with succinic acid leaves some carboxylic residues on the PVA membrane surface. Therefore, at higher pH values PVA membranes were more negatively charged at their surfaces due to the dissociation of pendent (unreacted) carboxylic acid groups. Therefore, higher rejection occurred at higher pH by Donnan exclusion. 4.4. Effect of PVA coating layer thickness on separation performance

Fig. 2. Comparison between the theoretical and experimental results of combined Spiegler–Kedem–film theory model for NaCl and Na2 SO4 solutions at pH 7.0 and 25 ◦ C. Table 1 Transport coefficients from the combined Spiegler–Kedem–film theory model.

 Ps , ␮m/s k, ␮m/s Lp , ␮m/(MPa s)

DI water

NaCl

Na2 SO4

– – – 10.41

0.972 8.19 15.1 –

0.996 0.394 10.4 –

aration, that is, the solute permeability coefficient which is more than an order of magnitude smaller for the divalent salt. 4.3. Effect of feed solution pH on separation performance The water flux and salt rejection of PVA–PSf composite NF membranes were investigated for different feed solution pHs (Fig. 3). The pH was adjusted by NaOH addition for all solutions and HCl or H2 SO4 addition for NaCl and Na2 SO4 solutions, respectively. The investigated pH values were 5, 7, and 9. Pure water permeability did not change significantly with pH (9.4 ␮m ± 0.4 MPa s−1 ), but the rejection of both NaCl and Na2 SO4 significantly increased with pH. For example, NaCl rejection increased from 24% (pH 5.0) to 47% (pH 9.0), while Na2 SO4 rejection increase from 77% (pH 5.0) to 92% (pH

Fig. 3. Effect of pH value of feed solution on the permeability and solute rejection of PVA–PSf composite membranes at 1034 kPa and 25 ◦ C.

Solutions of 0.05, 0.10, 0.20, 0.30 and 0.50 wt.% (PVA powder weight percentage) were used to cast PVA coating films. Representative SEM images of PVA–PSf composite membranes made from different PVA concentrations are provided in Fig. 4. The polysulfone support membrane (Fig. 4a) had a very thin skin layer of about 10–50 nm in thickness between the top surface and the tops of the first visible pores through the cross-section. In fact, these nanopores were also observed at the surface (surface SEM images shown previously by Ghosh and Hoek [27]). The PVA layers appeared non-porous, but were hard to discriminate from the polysulfone skin layer suggesting a good bond formed between the PSf support and PVA coating film. From the SEM images, the thicknesses of PVA coatings in Fig. 4b–f were estimated usually to be about 86 ± 43, 230 ± 28, 320 ± 41, 415 ± 50, 512 ± 67 nm for PVA membranes made from 0.05, 0.10, 0.20, 0.30, 0.50 wt.% PVA in the casting solution, respectively. The pure water permeability of PVA–PSf composite membranes decreased, while solute rejection (both sodium chloride and sodium sulfate) increased as PVA solution concentration in the casting solution increased (Fig. 5a). When PVA concentration in the casting solution was higher than 0.10 wt.%, the rejection of sodium sulfate was about 90%, but it was below 80% for 0.05 wt.% PVA casting solutions. For sodium chloride, the rejection was 35–45% for PVA casting solutions with more than 0.10 wt.% PVA, but the rejection was below 20% for 0.05 wt.% PVA concentrations in the casting solution. The pure water permeability of the PVA membrane with 0.05 wt.% PVA concentration was 17.5 ␮m MPa s−1 , but reduced in proportion to the PVA casting solution concentration. The film thickness and permeation results produced a correlation between pure water permeability and PVA layer thickness of Lp = (18.72–0.032) × ım in PVA–PSf composite membranes (Fig. 5b), where ım is the PVA layer thickness in nm. The membrane transport model described above assumed solvent and solute permeability were proportional to a characteristic diffusivity and solubility for each within the polymer phase, and inversely proportional to the polymer film thickness (i.e., P ∼ DK/ım ). We have already experimentally demonstrated the validity of this proportionality for solvent permeability. Here we present an analysis of film thickness and solute permeability. The  and k values determined for the 0.1 wt.% PVA film were assumed independent of film thickness. Next, the Ps value for 0.1 wt.% PVA film was multiplied by the film thickness. This thickness independent permeability was divided by the film thickness determined for each PVA film concentration. Finally, the observed rejection is predicted for each film thickness using  and k from the 0.1 wt.% film, plus ım and Jw observed during the filtration experiment. In Fig. 5b, the predicted rejections agree reasonably with observed rejections; hence, these PVA films exhibited selectivity that was inversely dependent on film thickness. These results illustrate one of the mechanisms responsible for the classic tradeoff between permeability and selectivity that is often observed for solution-diffusion membranes.

F. Peng et al. / Journal of Membrane Science 353 (2010) 169–176

173

Fig. 4. The FIB–SEM graphs of the cross-section structure of PVA membranes with different PVA concentrations in the casting solution: (a) polysulfone support membranes, PVA–PSf composite membranes with PVA concentrations of (b) 0.05%, (c) 0.10%, (d) 0.20%, (e) 0.30%, and (f) 0.50% in the casting solution.

4.5. Effect of PVA molecular weight on separation performance PVA–PSf composite membranes were prepared using PVA with molecular weights of 27, 47, and 61 kDa at 0.10 wt.% PVA concentrations in the casting solution. The pure water permeability and solute rejections for PVA–PSf composite membranes are shown in Fig. 6. The 27 kDa PVA composite membranes had the highest pure water permeability of 12.5 ␮m MPa−1 s−1 and rejections of 13.5% (NaCl) and 60.6% (Na2 SO4 ). The membranes made from PVA with molecular weight of 47 kDa showed the highest rejections of 37.5% (NaCl) and 90.5% (Na2 SO4 ) with nearly the lowest permeability. Composite nanofiltration membranes made from different PVA molecular weights exhibited different contact angles, and wettability and hydrophilicity as shown in Table 2. The contact angle of DI water for the polysulfone support membrane was about

74◦ , but the contact angles of DI water for all PVA composite membranes were between 25◦ and 32◦ . The solid–liquid interfacial free energy (−G13 ) calculated from the measured contact angles and known liquid surface tension of water is a more fundamental property for describing the wettability of solid surfaces. Typically, a condensed-phase material is considered “wetting” if −G13 > 72.8 mJ/m2 , which corresponds to a contact of 90◦ for pure water at 20 ◦ C. Apparently, lower molecular weight PVA produced slightly more hydrophilic surfaces. Both the LW and the AB components of surface tensions increase with PVA molecular weights. The higher total surface tension made wetting less favorable, while the increased electron acceptor functionality enhanced PVA self-attraction (i.e., decreased hydrophilicity). The interfacial free energy of cohesion, G131 , offers a quantitative description of the “hydrophilicity” of solid sur-

174

F. Peng et al. / Journal of Membrane Science 353 (2010) 169–176

Fig. 6. Effect of PVA molecular weight on the permeability and solute rejection of PVA–PSf composite membranes at 1034 kPa, pH 7.0 and 25 ◦ C.

Fig. 5. Effect of PVA concentrations in the casting solution (a) and PVA layer thickness (b) on the permeability and solute rejection of PVA–PSf composite membranes at 1034 kPa pH 7.0 and 25 ◦ C.

face [26,28]. As shown in Table 2, G131 followed the same trend. This is not always the case; we previously reported that alginate is wetting, but not hydrophilic according to this classification scheme [26]. The functionality responsible for the wettability and hydrophilicity of PVA was elucidated by ATR-IR spectroscopy. In Fig. 7, the absorbance at 3000–3600 cm−1 represents the –OH stretch associated with –OH groups in the PVA polymer chain and pendent –COOH groups from incomplete cross-linking reaction. The membrane made from PVA with molecular weight of 47 kDa

Fig. 7. FTIR spectra of polysulfone support membrane and PVA nanofiltration membranes with different PVA molecular weights.

Table 2 PVA layer contact angles and surface tensions. Liquid/membrane

Diiodomethane Glycerol Water Polysulfone SepRO PVA (MW = 27,000) membrane PVA (MW = 47,000) membrane PVA (MW = 61,000) membrane

Contact angle

 LW (mJ/m2 )

 + (mJ/m2 )

 − (mJ/m2 )

 AB (mJ/m2 )

 TOT (mJ/m2 )

G13 (mJ/m2 )

G131 (mJ/m2 )

Diiodomethane

Glycerol

DI water

n/a n/a n/a 14.5 ± 0.9 28.9 ± 1.8

n/a n/a n/a 63.2 ± 0.7 52 ± 2.9

n/a n/a n/a 74.3 ± 0.3 27.5 ± 1.9

50.8 34.0 21.8 49.2 44.7

0.0 3.9 25.5 0.00 0.13

0.0 57.4 25.5 7.0 60.5

0.0 29.9 51.0 0.2 5.5

50.8 63.9 72.8 49.4 50.2

n/a n/a n/a −92.5 −137.4

n/a n/a n/a −59.4 43.1

24.4 ± 0.4

40 ± 1.2

28.6 ± 2.1

46.4

0.19

46.3

5.9

52.3

−136.7

36.2

21.9 ± 2.5

56.7 ± 4.2

32.1 ± 2.7

47.2

0.59

59.8

11.9

59.1

−134.5

23.2

Cross-linking agent: succinic acid, PVA concentration in the casting solution: 0.10 wt.%, PVA hydrolysis degree: 98.0–98.8%, and cross-linking degree: 20%.

F. Peng et al. / Journal of Membrane Science 353 (2010) 169–176

175

PVA composite membrane permeability decreased with increasing PVA molecular weight. 5. Conclusions

Fig. 8. XRD spectra for cross-linked PVA films comprising different PVA molecular weights.

showed strongest peaks at 3000–3600 cm−1 , which may explain the relatively low  − functionality of this membrane. The peaks at 1630–1760 cm−1 are the –C O– in –C O–O–C–, which reflect the extent of cross-linking. As shown in Fig. 7, films made from PVA with molecular weight of 27 kDa showed the highest peak at both wavenumbers, films made from 61 kDa PVA showed the lowest peaks. We suggest that the actual extent of cross-linking was highest for PVA coating films made from 27 kDa polymer even though the theoretical cross-linking degrees were designed to be the same (20%). The extent of cross-linking controls the crystallinity of PVA films, which impacts permeability and selectivity. The crystalline properties of PVA films are described by XRD results in Fig. 8. Composite membranes made from PVA with molecular weight of 27,000 exhibit the highest extent of cross-linking based on FTIR results. The higher extent of cross-linking of PVA destroyed more crystalline areas of the PVA films, which resulted in looser polymer chain packing or aggregate structure. As shown in Fig. 9, the 27 kDa crosslinked PVA had lower crystallinity (11.6%) than uncross-linked PVA (15.4%). The XRD results confirmed that the crystallinity of PVA films increased with increasing PVA molecular weight. The degrees of crystallinity were 11.6%, 15.2%, and 15.9% for PVA with molecular weight of 27, 47, and 61 kDa, respectively. This explained why

Fig. 9. The comparison of XRD results between uncross-linked PVA and cross-linked PVA with molecular weight of 27,000 Da.

Ultra-thin and defect-free poly(vinyl alcohol) hydrogels were successfully coated onto polysulfone ultrafiltration support membranes to make PVA–PSf composite nanofiltration membranes. The coating film formation relied on multiple coatings with dilute PVA aqueous solution or in situ cross-linking. PVA active layer thickness ranged from 30 to 350 nm according to FIB–SEM images; pure water permeability correlated strongly with PVA film thickness. Infrared spectroscopy suggested the PVA cross-linking reaction formed –C O–O–C– groups. The PVA membranes were all very hydrophilic with water contact angles ranging between about 25◦ and 32◦ depending on PVA molecular weight and extent of crosslinking. Crystallinity of PVA films obtained by XRD helped to explain the permeation properties of PVA membranes—more cross-linking decreased crystallinity, which increased permeability. Predicted fluxes and rejections were in good agreement with the experimental results and nearly all of the PVA films gave the differential separation of monovalent and divalent salts—a characteristic of nanofiltration membranes. Disclosure statement The corresponding author has a financial interest in one of the project co-sponsors, NanoH2 O Inc., through stock ownership and a consulting agreement. Acknowledgements The authors are grateful for financial support for this research provided by a UC Discovery Grant (#GCP07-10239A) from the UC Industry-University Cooperative Research Program with industrial matching funds provided by NanoH2 O Inc. We are also very grateful to Dr. Bruce Dunn (UCLA Materials Science & Engineering Department) for providing access to the FTIR instrument. References [1] N. Hilal, H. Al-Zoubi, N.A. Darwish, A.W. Mohamma, M. Abu Arabi, A comprehensive review of nanofiltration membranes: treatment, pretreatment, modelling, and atomic force microscopy, Desalination 170 (2004) 281–308. [2] B. Van der Bruggen, M. Mänttäri, M. Nyström, Drawbacks of applying nanofiltration and how to avoid them: a review, Sep. Purif. Technol. 63 (2008) 251–263. [3] R.J. Petersen, Composite reverse osmosis and nanofiltration membranes, J. Membr. Sci. 83 (1993) 81–150. [4] F. Peng, L. Lu, H. Sun, Y. Wang, J. Liu, Z. Jiang, Hybrid organic–inorganic membranes: solving the trade-off between permeability and selectivity, Chem. Mater. 17 (2005) 6790–6796. [5] F. Peng, L. Lu, C. Hu, H. Wu, Z. Jiang, Significant increase of permeation flux and selectivity of poly(vinyl alcohol) membranes by incorporation of crystalline flake graphite, J. Membr. Sci. 259 (2005) 65–73. [6] B. Bolto, et al., Crosslinked poly(vinyl alcohol) membranes, Prog. Polym. Sci. (2009), doi:10.1016/j.progpolymsci.2009.05.003. [7] D. Bezuidenhout, M.J. Hurndall, R.D. Sanderson, et al., Reverse osmosis membranes prepared from potassium peroxydisulphate-modified poly(vinyl alcohol), Desalination 116 (1998) 35–43. [8] K. Lang, S. Sourirajan, T. Matsuura, et al., A study on the preparation of polyvinyl alcohol thin-film composite membranes and reverse osmosis testing, Desalination 104 (1996) 185–196. [9] K. Lang, T. Matsuura, G. Chowdhury, et al., Preparation and testing of polyvinylalcohol composite membranes for reverse-osmosis, Can. J. Chem. Eng. 73 (1995) 686–692. [10] M.H. Yang, Hydraulic permeation of NaCl solution through sulfonated polysulfone–polyvinyl alcohol/polysulfone composite reverse-osmosis membrane, Polym. Test. 14 (1995) 415–424. [11] K. Lang, G. Chowdhury, T. Matsuura, et al., Reverse-osmosis performance of modified polyvinyl-alcohol thin-film composite membranes, J. Colloid Interface Sci. 166 (1994) 239–244. [12] E. Immelman, R.D. Sanderson, E.P. Jacobs, et al., Poly(vinyl alcohol) gel sublayers for reverse-osmosis membranes. 1. Insolubilization by acid-catalyzed dehydration, J. Appl. Polym. Sci. 50 (1993) 1013–1034.

176

F. Peng et al. / Journal of Membrane Science 353 (2010) 169–176

[13] R.D. Sanderson, Polyvinyl-alcohol and modified polyvinyl-alcohol reverseosmosis membranes, Desalination 91 (1993) 114–1114. [14] R.D. Sanderson, E. Immerman, D. Bezuidenhout, et al., Polyvinyl-alcohol and modified polyvinyl-alcohol reverse-osmosis membranes, Desalination 90 (1993) 15–29. [15] T.J. Chu, M.H. Yang, Hydraulic permeation of water through heat-treated PVA membrane, J. Chem. Eng. Jpn. 25 (1992) 660–665. [16] J. Jegal, N.W. Oh, D.S. Park, et al., Characteristics of the nanofiltration composite membranes based on PVA and sodium alginate, J. Appl. Polym. Sci. 79 (2001) 2471–2479. [17] J. Jegal, N.W. Oh, K.H. Lee, Preparation and characterization of PVA/SA composite nanofiltration membranes, J. Appl. Polym. Sci. 77 (2000) 347–354. [18] J. Jegal, K.H. Lee, Nanofiltration membranes based on poly(vinyl alcohol) and ionic polymers, J. Appl. Polym. Sci. 72 (1999) 1755–1762. [19] J.M. Gohil, P. Ray, Polyvinyl alcohol as the barrier layer in thin film composite nanofiltration membranes: preparation, characterization, and performance evaluation, J. Colloid Interface Sci. 338 (2009) 121–127. [20] O. Kedem, A. Katchalsky, Thermodynamic analysis of the permeability of biological membranes to non-electrolytes, Biochim. Biophys. Acta 27 (1958) 229–246.

[21] O. Kedem, K. Spiegler, Thermodynamics of hyperfiltration (reverse osmosis): criteria for efficient membranes, Desalination 1 (1966) 311–326. [22] J.G. Wijmans, R. Baker, The solution-diffusion model: a review, J. Membr. Sci. 107 (1995) 1–7. [23] S. Koter, Determination of the parameters of the Spiegler–Kedem–Katchalsky model for nanofiltration of single electrolyte solutions, Desalination 198 (2006) 335–345. [24] S. Kim, E.M.V. Hoek, Modeling concentration polarization in reverse osmosis processes, Desalination 186 (2005) 111–128. [25] J.A. Brant, A.E. Childress, Assessing short-range membrane–colloid interactions using surface energetics, J. Membr. Sci. 203 (2002) 257–273. [26] X. Jin, E.M.V. Hoek, Role of specific ion interactions in seawater RO membrane fouling by alginic acid, Environ. Sci. Technol. 43 (2009) 3580– 3587. [27] A.K. Ghosh, E.M.V. Hoek, Impacts of support membrane structure and chemistry on polyamide–polysulfone interfacial composite membranes, J. Membr. Sci. 336 (2009) 140–148. [28] A. Subramani, E.M.V. Hoek, Direct observation of initial microbial deposition onto reverse osmosis and nanofiltration membranes, J. Membr. Sci. 319 (2008) 111–125.