Separation of Fluoride from Other Monovalent Anions Using Multilayer


Jan 9, 2007 - Seong Uk Hong,Ramamoorthy Malaisamy, andMerlin L. Bruening*. Department of Chemical Engineering, Hanbat National University, Daejeon 305...
0 downloads 7 Views 82KB Size


1716

Langmuir 2007, 23, 1716-1722

Separation of Fluoride from Other Monovalent Anions Using Multilayer Polyelectrolyte Nanofiltration Membranes Seong Uk Hong,† Ramamoorthy Malaisamy,‡ and Merlin L. Bruening‡,* Department of Chemical Engineering, Hanbat National UniVersity, Daejeon 305-719, South Korea, and Department of Chemistry, Michigan State UniVersity, East Lansing, Michigan 48824 ReceiVed June 13, 2006. In Final Form: NoVember 13, 2006 Nanofiltration (NF) is an attractive technique for reducing F- concentrations to acceptable levels in drinking water, but commercial NF membranes such as NF 270 and NF 90 show minimal Cl-/F- selectivity. In contrast, simple layer-by-layer deposition of 4.5-bilayer poly(styrene sulfonate) (PSS)/poly(diallyldimethylammonium chloride) (PDADMAC) films on porous alumina supports yields NF membranes that exhibit Cl-/F- and Br-/F- selectivities >3 along with solution fluxes that are >3-fold higher than those of the commercial membranes. Fluoride rejection by (PSS/PDADMAC)4PSS membranes, which is >70%, is independent of pressure over a range of 3.6 to 6.0 bar, suggesting that the primary transport mechanism in these films is convection. Moreover, the fact that Br-/F- selectivity is 12% higher than Cl-/F- selectivity suggests that discrimination among the monovalent ions is based on size (Stokes radius). Chloride/fluoride selectivities are essentially constant over Cl-/F- feed ratios from 1 to 60, so these separations will be viable over a range of conditions. Interestingly, PSS/protonated poly(allylamine) films show little Cl-/Fselectivity, and the selectivity of PSS/PDADMAC membranes is a strong function of the number of deposited layers, indicating that NF properties are very sensitive to film structure.

Introduction Regular drinking of water with F- concentrations higher than 2 mg/L can cause dental and bone fluorosis and possibly other adverse health effects.1-4 Nevertheless, in the United States alone, about 200,000 people drink water with natural F- concentrations of >4 mg/mL,1 and cost-effective means for reducing F- levels in water supplies are needed. Removal of F- from drinking water using strong ion-exchange resins is not very effective because other ions such as Cl- will be more strongly adsorbed to the resin than F-.2,3 Thus, reduction of F- concentrations is typically accomplished using adsorption on activated alumina, but the binding capacities of alumina are relatively low (around 5 g Fper kg of alumina, depending on the composition of the solution).3 Nanofiltration (NF) is an attractive alternative for F- removal. This technique is similar to reverse osmosis (RO) in that pressure drives a solvent (usually water) across a membrane against a concentration gradient. However, NF membranes allow higher salt passage than RO membranes, which decreases osmotic pressure and avoids the need for remineralization.5 Moreover, the high permeability of NF membranes also lowers the pressure needed for water purification. Hence in applications such as water softening or F- removal, where high rejections of NaCl are not required, NF is preferable to RO. Development of NF for F- removal will require the synthesis of highly permeable membranes that selectively reject F- while * Author to whom correspondence should be addressed. E-mail: [email protected], Phone: (517) 355-9715 ext. 237, Fax: (517) 353-1793. † Hanbat National University. ‡ Michigan State University. (1) Hileman, B. Chem. Eng. News 2006, 84, 11. (2) Lhassani, A.; Rumeau, M.; Benjelloun, D.; Pontie, M. Water Res. 2001, 35, 3260-3264. (3) Veressinina, Y.; Trapido, M.; Ahelik, V.; Munter, R. Proc. Estonian Acad. Sci. Chem. 2001, 50, 81-88. (4) Chidambaram, S.; Ramanathan, A. L.; Vasudevam, S. Water SA 2003, 29, 339-343. (5) Petersen, R. J. J. Membr. Sci. 1993, 83, 81-150. (6) Leva¨salmi, J. M.; McCarthy, T. J. Macromolecules 1997, 30, 1752-1757. (7) Kotov, N. A.; Magonov, S.; Tropsha, E. Chem. Mater. 1998, 10, 886-895.

passing other monovalent anions such as Cl- and Br-. As shown below, typical NF membranes exhibit about equal rejections of Cl- and F-. This study examines whether layer-by-layer adsorption of polycations and polyanions on porous supports can provide selective NF membranes for removal of F- in the presence of Cl- and Br-. Multilayer polyelectrolyte membranes are especially attractive for such separations because the layerby-layer technique affords control over the skin thickness to maximize flux and also allows formation of a wide range of membranes with a variety of permeation properties.6-12 In general, ion transport through NF membranes is a function of the charge and pore size of the skin layer, which is usually highly charged with an effective pore size between 1 and 5 nm.13-15 The layerby-layer adsorption of polyelectrolytes naturally affords a charged surface, and effective pore size can be controlled through variation of the polyelectrolytes deposited, the number of adsorbed layers, and deposition conditions such as pH, supporting salt concentration, and adsorption time.16-33 (8) Dubas, S. T.; Farhat, T. R.; Schlenoff, J. B. J. Am. Chem. Soc. 2001, 123, 5368-5369. (9) Liu, X.; Bruening, M. L. Chem. Mater. 2004, 16, 351-357. (10) Miller, M. D.; Bruening, M. L. Langmuir 2004, 20, 11545-11551. (11) Krasemann, L.; Tieke, B. Mater. Sci. Eng. 1999, C8-9, 513-518. (12) Ball, V.; Voegel, J. C.; Schaaf, P. Langmuir 2005, 21, 4129-4137. (13) Bowen, W. R.; Mohammad, A. W.; Hilal, N. J. Membr. Sci. 1997, 126, 91-105. (14) Afonso, M. D.; Hagmeyer, G.; Gimbel, R. Sep. Purif. Technol. 2001, 22-23, 529-541. (15) Shibata, M.; Kobayashi, T.; Fujii, N. J. Appl. Polym. Sci. 2000, 75, 15461553. (16) Berg, M. C.; Yang, S. Y.; Hammond, P. T.; Rubner, M. F. Langmuir 2004, 20, 1362-1368. (17) Harris, J. J.; DeRose, P. M.; Bruening, M. L. J. Am. Chem. Soc. 1999, 121, 1978-1979. (18) Schlenoff, J. B.; Dubas, S. T. Macromolecules 2001, 34, 592-598. (19) Yoo, D.; Shiratori, S. S.; Rubner, M. F. Macromolecules 1998, 31, 43094318. (20) Wong, J. E.; Rehfeldt, F.; Ha¨nni, P.; Tanaka, M.; Klitzing, R. v. Macromolecules 2004, 37, 7285-7289. (21) Lo¨sche, M.; Schmitt, J.; Decher, D.; Bouwman, W.; Kjaer, K. Macromolecules 1998, 31, 8893-8906. (22) Dubas, S. T.; Schlenoff, J. B. Macromolecules 1999, 32, 8153-8160. (23) Scholer, B.; Kumaraswamy, G.; Caruso, F. Macromolecules 2002, 35, 889-897.

10.1021/la061701y CCC: $37.00 © 2007 American Chemical Society Published on Web 01/09/2007

Multilayer Polyelectrolyte Nanofiltration Membranes

Langmuir, Vol. 23, No. 4, 2007 1717

Table 1. Molecular Weights (Mw), Stokes Radii (rs), and Aqueous Diffusion Coefficients (D)43 of the Anions Used in Transport Studies anions

Mw (g mol-1)

rs (nm)

D (m2/s)

fluoride chloride bromide

19.0 35.5 79.9

0.166 0.121 0.118

1.48 × 10-9 2.03 × 10-9 2.08 × 10-9

Several recent papers examined the abilities of different multilayer polyelectrolyte membranes to separate monovalent and divalent ions such as Cl- and SO42-.34-42 Fluxes through poly(styrene sulfonate) (PSS)/poly(diallyldimethylammonium chloride) (PDADMAC) membranes were as high as 4.2 m3/ m2-day at 4.8 bar, and Cl-/SO42- selectivities reached values >30.39 Those separations were based primarily on the charge of the analytes, however, and discrimination among monovalent ions will be more difficult. Nevertheless, the effective size (Stokes radius) of aqueous F- is significantly higher than that of Cl- and Br- (Table 1), so size-based separations of these anions should in principle be possible. This work probes the potential of both PSS/PDADMAC and PSS/protonated poly(allylamine) (PAH) NF membranes for the separation of monovalent anions, with an emphasis on the selective rejection of F-. Remarkably, deposition of 4.5-bilayer PSS/PDADMAC films on porous alumina supports yields membranes that exhibit Cl-/F- and Br-/F- selectivities >3 with minimal Cl- and Br- rejections, and a solution flux of 3.5 m3/m2-day at 4.8 bar. In contrast, the selectivities of the commercial NF membranes that we tested (NF 90 and NF 270 from DOW) and PSS/PAH membranes were about 1 under the same experimental conditions, and the highest flux through the commercial membranes was 1.1 m3/m2-day at 4.8 bar. Experimental Section Materials. Poly(styrene sulfonate) (PSS, Mw ) 70 000 Da), poly(allylamine hydrochloride) (PAH, Mw ) 70 000 Da), poly(diallyldimethylammonium chloride) (PDADMAC, Mw ) 100 000-200 000 Da, 20 wt % in water), and NaF (99+%) were obtained from SigmaAldrich and used as received. NaCl and NaBr (Jade Scientific, ACS reagent grade) were also used as received. The structures of the polyelectrolytes are shown in Figure 1. Porous alumina supports (0.02-µm Whatman Anodisc filters) were UV/O3 cleaned (Boekel UV-Clean Model 135500) with the feed side up for 15 min before film deposition, and deionized water (Milli-Q, 18.2 MΩcm) was used for membrane rinsing and preparation of the polyelectrolyte solutions. (24) Scholer, B.; Poptoshev, E.; Caruso, F. Macromolecules 2003, 36, 52585264. (25) Tjipto, E.; Quinn, J. F.; Caruso, F. Langmuir 2005, 21, 8785-8792. (26) Blomberg, E.; Poptoshev, E.; Caruso, F. Langmuir 2006, 22, 4153-4157. (27) Dubas, S. T.; Schlenoff, J. B. Macromolecules 2001, 34, 3736-3740. (28) Sui, Z.; Salloum, D.; Schlenoff, J. B. Langmuir 2003, 19, 2491-2495. (29) Shiratori, S. S.; Rubner, M. F. Macromolecules 2000, 33, 4213-4219. (30) Choi, J.; Rubner, M. F. Macromolecules 2005, 38, 116-124. (31) Mermut, O.; Barrett, C. J. J. Phys. Chem. B 2003, 107, 2525-2530. (32) Porcel, C. H.; Izquierdo, A.; Ball, V.; Decher, G.; Voegel, J. C.; Schaaf, P. Langmuir 2005, 21, 800-802. (33) Izquierdo, A.; Ono, S. S.; Voegel, J. C.; Schaaf, P.; Decher, G. Langmuir 2005, 21, 7558-7567. (34) Stanton, B. W.; Harris, J. J.; Miller, M. D.; Bruening, M. L. Langmuir 2003, 19, 7038-7042. (35) Harris, J. J.; Stair, J. L.; Bruening, M. L. Chem. Mater. 2000, 12, 19411946. (36) Hollman, A. M.; Bhattacharyya, D. Langmuir 2004, 20, 5418-5424. (37) Hollman, A. M.; Bhattacharyya, D. Langmuir 2002, 18, 5946-5952. (38) Balachandra, A. M.; Dai, J.; Bruening, M. L. Macromolecules 2002, 35, 3171-3178. (39) Hong, S. U.; Malaisamy, R.; Bruening, M. L. J. Membr. Sci. 2006, 283, 366-372. (40) Tieke, B.; Toutianoush, A.; Jin, W. AdV. Colloid Interf. Sci. 2005, 116, 121-131. (41) Krasemann, L.; Tieke, B. Langmuir 2000, 16, 287-290. (42) Jin, W.; Toutianoush, A.; Tieke, B. Langmuir 2003, 19, 2550-2553.

Figure 1. Structures of the polyelectrolytes used in this study. Film Deposition. Polyelectrolyte films were made using a literature procedure.10,39 Deposition of these films on alumina supports with 0.02 µm-diameter surface pores allows the polyelectrolyte multilayers to cover the support without filling underlying pores, as seen in previouslyreportedscanningelectronmicroscopy(SEM)images.10,35,44-46 The alumina provides the mechanical strength of the membrane while presenting a minimal resistance to mass transport. For PSS/ PDADMAC, film deposition started with exposure of the top of the alumina support to an aqueous solution containing 0.02 M PSS in 0.5 M NaCl for 3 min (concentrations of polyelectrolytes are always given with respect to the repeating unit). The alumina support was then rinsed with deionized water for 1 min before exposure to 0.02 M PDADMAC in 0.5 M NaCl for 3 min, followed by a second 1-min water rinse. Additional bilayers were deposited similarly. For PSS/PAH films, the PSS solution contained 0.02 M PSS and 0.5 M NaCl at a pH of 2.1, whereas the PAH solution contained 0.02 M PAH and 0.5 M NaCl at a pH of 2.3. The adsorption times were 2 and 5 min for PSS and PAH, respectively, and 1-min water rinses were performed after each adsorption. In some cases, after deposition of three or four bilayers of PSS/PDADMAC or PSS/PAH, membranes were exposed to PSS solutions that contained 1 M NaCl for adsorption of the terminating layer. After deposition of the desired number of polyelectrolyte layers, membranes were stored in water until use. Deposition times were selected based on previous studies or on a modest optimization of film properties.10,34 The immersion times may not be sufficient to achieve the thickest films.22 Ellipsometry. Ellipsometry studies were performed to determine the thicknesses of films on Au-coated Si wafers using a literature procedure.39 To allow layer-by-layer deposition to commence with a polyanion, the Au surface was first coated with a monolayer of an amine precursor by immersing the cleaned wafer in 5 mM cystamine dihydrochloride in water for at least 1 h and rinsing with water. Subsequently, the polyanion and polycation layers were deposited on Au using the same deposition conditions applied with the alumina supports. After deposition, films were dried with a stream of nitrogen, and thicknesses were determined at three different points per wafer on at least two different wafers. Transport Studies. NF experiments were performed with a homebuilt cross-flow apparatus described previously.34 This system was pressurized with Ar, and a centrifugal pump circulated the analyte solution through the apparatus and across the membrane, which had (43) Handbook of Chemistry and Physics, CRC Press: Boca Raton, Florida, 2000, page 5-91. (44) Stair, J. L.; Harris, J. J.; Bruening, M. L. Chem. Mater. 2001, 13, 26412648. (45) Sullivan, D. M.; Bruening, M. L. Chem. Mater. 2003, 15, 281-287. (46) Hong, S. U.; Miller, M. D.; Bruening, M. L. Ind. Eng. Chem. Res. 2006, 45, 6284-6288.

1718 Langmuir, Vol. 23, No. 4, 2007

Hong et al.

Table 2. Rejections, Solution Fluxes, and Selectivities from NF Experiments with (PSS/PDADMAC)nPSS-Coated Alumina Membranes and Solutions Containing NaF (1 mM) and NaCl (1 mM). NF Was Performed at 4.8 Bar with a Cross-Flow Rate of 18 mL/min number of bilayers, n

thickness, nmb

solution flux, m3/m2-day

chloride rejection, %

fluoride rejection,

chloride/ fluoride selectivity

3a 4 4a 5 6

16.4 ( 1.8 26.7 ( 2.3c 28.5 ( 0.8 32.9 ( 1.8c 40.4 ( 1.3

5.7 ( 0.2 3.5 ( 0.2 3.4 ( 0.1 2.6 ( 0.1 2.0 ( 0.1

26.6 ( 1.3 9.5 ( 1.1 40.9 ( 2.6 7.3 ( 1.2 52.1 ( 1.2

55.2 ( 4.8 73.1 ( 0.9 78.1 ( 1.0 50.2 ( 0.5 57.5 ( 1.1

1.7 ( 0.2 3.4 ( 0.2 2.7 ( 0.1 1.9 ( 0.0 1.1 ( 0.1

a The top layer of PSS was deposited from a solution containing 1.0 M NaCl. b Thicknesses were determined ellipsometrically with films on Au-coated wafers, and our previous research suggests that these thicknesses are similar to those of films deposited on porous alumina.10,35,45 To estimate the mass of polyelectrolyte deposited in g of polyelectrolyte per cm2 of wafer or membrane, multiply the thickness by 10-7. c Data from reference 39.

Table 3. Comparison of the Performance of (PSS/PDADMAC)4PSS-Coated Alumina Membranes in NF of Solutions Containing 1 mM NaCl, 1 mM NaF or 1 mM NaBr, 1 mM NaFa NF membrane

solution flux, m3/ m2-day

chloride rejection, %

bromide rejection, %

fluoride rejection, %

anion/fluoride selectivity

(PSS/PDADMAC)4PSSb (PSS/PDADMAC)4PSSc (PSS/PDADMAC)4PSSb (PSS/PDADMAC)4PSSc

3.5 ( 0.2 3.4 ( 0.1 3.7 ( 0.3 3.4 ( 0.1

9.5 ( 1.1 40.9 ( 2.6 -

-3.9 ( 2.6 28.8 ( 1.3

73.1 ( 0.9 78.1 ( 1.0 73.0 ( 1.4 79.5 ( 1.1

3.4 ( 0.2 2.7 ( 0.1 3.8 ( 0.1 3.5 ( 0.2

a NF experiments were performed at 4.8 bar with a cross-flow rate of 18 mL/min. b Deposition of the top layer of PSS occurred from a solution containing 0.5 M NaCl. c Deposition of the top layer of PSS occurred from a solution containing 1 M NaCl.

Table 4. Rejections, Solution Fluxes, and Selectivities from NF Experiments with (PSS/PAH)nPSS-Coated Alumina Membranes and Solutions Containing NaF (1 mM) and NaCl (1 mM). NF Was Performed at 4.8 Bar with a Cross-Flow Rate of 18 mL/min number of bilayers, n 4 4a 5 6

thickness, nmb

solution flux, m3/m2-day

chloride rejection, %

fluoride rejection, %

chloride/ fluoride selectivity

13.7 ( 0.5c 14.3 ( 1.0 19.4 ( 0.4c 23.5 ( 1.6

4.7 ( 0.3 4.4 ( 0.2 3.5 ( 0.2 2.6 ( 0.2

16.2 ( 1.0 16.7 ( 0.7 19.4 ( 1.0 27.3 ( 0.6

21.8 ( 1.8 23.9 ( 0.7 24.2 ( 1.4 30.7 ( 0.8

1.1 ( 0.0 1.1 ( 0.0 1.1 ( 0.0 1.1 ( 0.0

a Deposition of the top layer occurred from a solution containing 1 M NaCl. b Thicknesses were determined ellipsometrically with films on Au-coated wafers, and our previous research suggests that these thicknesses are similar to those of films deposited on porous alumina.10,35,45 To estimate the mass of polyelectrolyte deposited in g of polyelectrolyte per cm2 of wafer or membrane, multiply the thickness by 10-7. c Data from reference 39.

an exposed area of 1.5 cm2. The flow rate across the membrane was controlled by a flowmeter located between the pump and membrane cells. After 18-h of filtration, four permeate samples were collected using a graduated cylinder for time periods ranging from 10 to 60 min each, depending on the flux of the membrane, and the feed was analyzed at the end of the experiment. The flux measurements reported are the steady-state solution flux after the initial 18 h of filtration. Anion concentrations were determined using ion chromatography (Dionex 600 Ion Chromatograph with an Ionpac (AS14A) column) with conductivity detection, and all reported transport results are the averages of experiments with at least two different membranes.

Results and Discussion This section first presents NF results with PSS/PDADMAC membranes and then compares these data with the NF performance of PSS/PAH and commercial NF 270 and NF 90 membranes. Because (PSS/PDADMAC)4PSS membranes exhibited much higher Cl-/F- selectivities than the commercial or PSS/PAH systems, subsequent sections of this paper examine the NF properties of (PSS/PDADMAC)4PSS films on porous alumina as a function of pressure, feed flow rate (cross-flow), and feed composition in an effort to understand the mechanism and limitations of these separations. For all of the membranes, percent rejections, R, are defined by eq 1 where Cperm and Cfeed are the solute concentrations in the permeate and feed, respectively. The selectivity, S, for solute A over solute B is defined by eq 2, which can be conveniently expressed in terms of rejections as shown. Percent rejection and selectivity were determined after

18 h of filtration to achieve steady-state permeate concentrations.

(

R) 1S)

)

Cperm × 100% Cfeed

CA,perm CB,feed 100 - RA ) CA,feed CB,perm 100 - RB

(1)

(2)

NF with PSS/PDADMAC Membranes. Table 2 contains percent rejection values, selectivities, and solution fluxes from NF experiments with several PSS/PDADMAC membranes. Dry thicknesses of films on Au are also included to help in understanding the NF experimental results. Previous studies showed that the ellipsometric thicknesses of polyelectrolyte films deposited on metal-coated Si wafers are similar to SEM-derived thickness of films on porous alumina.10,35,45 All of the membranes were terminated with a PSS layer because the negative charge of the predominantly polyanionic surface may increase selectivity among anions.34 (PSS/PDADMAC)3PSS films are very thin, and the flux through these membranes (deposited entirely from solutions containing 0.5 M NaCl) was too high for NF experiments because the volume of the feed solution (2 liters per membrane) was insufficient for an 18-h filtration. The high flux through these films suggests that they would show minimal selectivity because they likely do not fully cover membrane pores.35 However, when the top layer of (PSS/PDADMAC)3PSS membranes was de-

Multilayer Polyelectrolyte Nanofiltration Membranes

Langmuir, Vol. 23, No. 4, 2007 1719

Table 5. Comparison of the Performance of Several Membranesa in NF of Solutions Containing NaCl and NaF NF membrane

solution flux, m3/m2-day

chloride rejection, %

fluoride rejection, %

chloride/fluoride selectivity

(PSS/PDADMAC)4PSSb (PSS/PAH)4PSSb NF 90 NF 270

3.5 ( 0.2 4.7 ( 0.3 0.3 ( 0.1 1.11 ( 0.02

9.5 ( 1.1 16.2 ( 1.0 81.6 ( 3.4 73.9 ( 1.3

73.1 ( 0.9 21.8 ( 1.8 80.6 ( 5.8 63.0 ( 1.7

3.4 ( 0.2 1.1 ( 0.0 1.0 ( 0.1 0.7 ( 0.1

a NF experiments were performed at 4.8 bar with a cross-flow rate of 18 mL/min using a 1 mM NaCl, 1 mM NaF solution. b The top layer was deposited from a solution containing 0.5 M NaCl.

Figure 2. Solution flux and Cl-/F- selectivity as a function of applied pressure in NF of 1 mM NaF, 1 mM NaCl solutions through membranes composed of (PSS/PDADMAC)4PSS on alumina. The top layer of the membranes was deposited from a solution containing 0.5 M NaCl. Two freshly prepared membranes were employed at each pressure. Error bars represent standard deviations.

posited using a PSS solution containing 1 M NaCl instead of 0.5 M NaCl, the feed volume was sufficient for the experiment, and the Cl-/F- selectivity was about 1.7. (Higher salt concentrations in deposition solutions generally result in adsorption of thicker layers.10,22,39) Remarkably, flux through this membrane was 5.7 m3/m2-day, or about 5-fold greater than that through commercial membranes (see below). Although this flux is very attractive, the F- rejection of 55% is a little lower than one would like for Fremoval. (Typical drinking water applications will involve decreasing the amount of F- in water from 50%, and flux, though not reported, was likely low.)2 Deposition of the top PSS layer of (PSS/ PDADMAC)4PSS membranes from 1 M (rather than 0.5 M) NaCl did not greatly change flux or F- rejection, but it did increase Cl- rejection from 10% to 41% to reduce Cl-/F- selectivity by 20%. When the number of PSS/PDADMAC bilayers was increased from 4.5 to 5.5, F- rejection decreased from 73% to 50% and Cl-/F- selectivity dropped to 1.9. Addition of another bilayer to form (PSS/PDADMAC)6PSS films resulted in a significant increase in Cl- rejection to give essentially no Cl-/F- selectivity. Moreover, flux decreased monotonically with an increasing number of bilayers because of increasing film thickness, so 4.5bilayer membranes will obviously be preferred for F- removal. (47) Choi, S.; Yun, Z.; Hong, S.; Ahn, K. Desalination 2001, 133, 53-64.

Although decreases in flux with increasing film thickness were expected, the decreasing selectivity with deposition of more than 4.5 bilayers was not. However, our previous data for NF of sugars and ions hinted at decreasing rejections when the number of PSS/PDADMAC bilayers on alumina supports increased from 4.5 to 5.5, although experimental uncertainties did not allow firm conclusions.10,39 Assuming that the separation of F- and Cl- is due to sieving of F-, which has the larger Stokes radius, trends in F- rejection with the number of PSS/PDADMAC bilayers suggest that the effective pore size of PSS/PDADMAC films increases upon going from 4.5 to 5.5 bilayers. Several studies indicate that polyelectrolyte films are not homogeneous and that the first few bilayers have properties different from those in the bulk, including lower hydration.48-51 Moreover, in thick films, the film-solution interface is more strongly hydrated than the bulk.48,51 One previous study of anion-exchange membranes suggests that more hydrated materials show significantly higher F-/Cl- mobility ratios.52 Thus, in the regime of Cl- > F-.31 Because of such interactions, PSS/PDADMAC films prepared in 0.5 M NaBr can be 4-fold thicker than films prepared in the presence of 0.5 M NaF.53 Moreover the stiffness of films depends on the composition of the electrolyte solution from which they were deposited, although after film formation, stiffness does not seem to change upon immersion in a different electrolyte solution.48 The fact that Frejection is the same in solutions containing Cl- and Br- confirms that at low ion concentrations (1 mM), film properties are (53) Saloma¨ki, M.; Tervasma¨ki, P.; Areva, S.; Kankare, J. Langmuir 2004, 20, 3679-3683.

Hong et al.

Figure 5. Effect of feed cross-flow rate on solution flux and Cl-/Fselectivity in NF of 1 mM NaF, 1 mM NaCl solutions through membranes composed of (PSS/PDADMAC)4PSS deposited on alumina. The top layer of the membranes was adsorbed from a solution containing 0.5 M NaCl.

unaffected by the electrolyte composition. Moreover, results presented below demonstrate that F- rejections do not vary significantly with Cl- concentrations up to 15 mM. The different interaction energies between polycations and monovalent anions also suggest an alternative mechanism to sieving for explaining the Cl-/F- selectivities of PSS/PDADMAC membranes. Schlenoff and co-workers proposed a model of transport through multilayer polyelectrolyte films in which ions move through the membrane by hopping between ion-exchange sites.54,55 The hopping rate of various monovalent anions depends on the kinetics of their interactions with polycations, which should correlate with the strength of anion-polycation interactions. However, in the case of nanofiltration, the primary transport mechanism is likely convection, not hopping, as shown below. Another challenge to a size-based explanation for Cl-/F- and Br-/F- selectivities is that larger neutral molecules such as glycerol and glucose exhibit lower rejections in NF with PSS/ PDADMAC membranes than does F-.10,56 The rejection of glycerol by (PSS/PDADMAC)4PSS membranes, for example, is less than 20% in spite of the fact that the Stokes radius of glycerol (0.26 nm) is 40% larger than that of F-, which shows a > 70% rejection. We speculate that entrance into and passage through membrane pores is a function of both the size and charge of the species. If charges on pore walls reduce effective pore diameters for ions due to electrostatic effects, then size-based selectivity will be achieved at lower Stokes radii for ions than for neutral molecules.57 Schlenoff and co-workers previously showed orderof-magnitude differences in the transport rates of neutral and ionized ascorbic acid through polyelectrolyte multilayers.55 Those results are consistent with effective pore size being smaller for ions than neutral molecules, although in that case, transport occurred through diffusion, which is well-described by the ionhopping model. Differences between (PSS/PDADMAC)4PSS films and (PSS/ PDADMAC)nPSS films with n ) 5 or 6 could be due to architectural variations that lead to changes in the charge density or size of pores. Comparison with Other NF Membranes. Table 4 shows results from NF and ellipsometric studies with PSS/PAH films. In this case, we replaced the strong polycation PDADMAC with (54) Farhat, T. R.; Schlenoff, J. B. J. Am. Chem. Soc. 2003, 125, 4627-4636. (55) Rmaile, H. H.; Farhat, T. R.; Schlenoff, J. B. J. Phys. Chem. B 2003, 107, 14401-14406. (56) Malaisamy, R.; Bruening, M. L. Langmuir 2005, 21, 10587-10592. (57) Siwy, Z.; Kosin´ska, I. D.; Fulin´ski, A.; Martin, C. R. Phys. ReV. Lett. 2005, 94, 048102.

Multilayer Polyelectrolyte Nanofiltration Membranes

Langmuir, Vol. 23, No. 4, 2007 1721

Table 6. Effect of Feed Composition on Nanofiltration Properties of (PSS/PDADMAC)4PSS Membranesa

a

fluoride concentration (mM)

chloride concentration (mM)

solution flux, m3/m2-day

chloride rejection, %

fluoride rejection, %

chloride/fluoride selectivity

0.25 1 0.25

1 1 15

3.7 ( 0.1 3.5 ( 0.2 3.8 ( 0.1

7.8 ( 1.7 9.5 ( 1.1 9.3 ( 0.7

69.8 ( 0.9 73.1 ( 0.9 69.8 ( 1.9

3.1 ( 0.1 3.4 ( 0.2 3.0 ( 0.2

NF experiments were performed at 4.8 bar and a cross-flow rate of 18 mL/min using membranes prepared from solutions containing 0.5 M NaCl.

the weakly basic PAH. Unlike (PSS/PDADMAC)4PSS, rejections of both Cl- and F- by (PSS/PAH)4PSS films are low (16% and 22%, respectively), resulting in a Cl-/F-selectivity of just 1.1. These results demonstrate that the composition of the entire film (not just the capping layer) is important in determining rejection. The fluxes through the (PSS/PAH)4PSS membranes are 30% higher than those through the (PSS/PDADMAC)4PSS membranes, due in part to the lower thickness of the PSS/PAH films. The fluxes presented here are also higher than those we previously achieved with (PSS/PAH)4PSS films prepared under different conditions that yield thicker films.34 We should note that the presence of F- also appears to increase flux, perhaps because it reacts with the alumina substrate to somewhat increase pore size. Regardless of the reason, pure water flux was about 17% lower than the flux of solutions containing 1 mM NaF and 1 mM NaCl through (PSS/PDADMAC)4PSS films. When the number of PSS/PAH bilayers was increased from 4.5 to 5.5 or 6.5, both F- and Cl- rejections increased slightly to maintain the minimal selectivity, and the solution flux decreased due to higher film thicknesses. Since we could not increase the Cl-/F- selectivity by simply increasing the number of bilayers, we tried depositing the top layer of (PSS/PAH)4PSS membranes using the PSS solution containing 1 M NaCl instead of 0.5 M NaCl. However, Cl-/F- selectivity was still only 1.1. Because of the lack of selectivity of PSS/PAH membranes, only (PSS/ PDADMAC)4PSS films were employed in further studies with multilayer polyelectrolyte membranes. We also investigated the NF performance of two commercial membranes, NF 270 and NF 90 from Dow Chemical, and Table 5 compares the NF properties of these membranes with those of the polyelectrolyte films. In the case of NF 90 membranes, rejections of both F- and Cl- were about 80%, resulting in a selectivity of 1. Additionally, the flux through these membranes was only 0.3 m3/m2-day. With the NF 270 system, the flux was 3.7-fold higher than that through NF 90 membranes, but the Clrejection was higher than F- rejection, resulting in a Cl-/Fselectivity of 0.7. Obviously such membranes will not prove useful in selective removal of F-. The F- rejection of (PSS/ PDADMAC)4PSS membranes is about the same as that of the NF 90 and NF 270 membranes, but low Cl- rejection and a 3-foldhigherfluxthanwithNF270makethe(PSS/PDADMAC)4PSS system much more attractive than the commercial membranes for removal of F- from drinking water. Effect of Applied Pressure on Nanofiltration. Variation of the applied pressure can provide insight into the contributions of diffusion and convection to ion transport. If transport is predominantly due to convection, then we would expect rejection to be essentially constant with variations in pressure, but if diffusion is an important transport mechanism at low pressures, we would expect to see rejection increase at higher pressures.58 As seen in Figure 2, the flux through (PSS/PDADMAC)4PSS membranes increased essentially linearly with increasing pressures as would be expected given the small osmotic pressure (58) Zhao, Y.; Taylor, J. S.; Chellam, S. J. Membr. Sci. 2005, 263, 38-46.

(0.05 bar) of the 2 mM salt solutions. Moreover selectivity was essentially independent of pressure. Ion rejections (Figure 3) showed no detectable variation with pressure when the applied pressure was >3.6 bar. The slightly lower rejections and selectivity at 2.4 bar could be due to diffusion effects, but the differences between selectivities and F- rejections at 2.4 bar and at other pressures are small. Moreover, it is possible that membrane compaction was less at the lower pressure.59 Overall, the constancy of rejection despite a 2.5-fold increase in solution flux suggests that convection dominates these separations.60-62 Variation of NF Properties with Changing Feed CrossFlow Rates. We investigated the effect of feed cross-flow rate on NF properties of (PSS/ PDADMAC)4PSS membranes to verify that concentration polarization at the membrane/solution interface was not affecting the separation. As seen in Figure 4, the Frejection did increase significantly (from 45 to 73%) on going from 5 to 18 mL/min, but further increases in flow rate did not enhance rejection. (At a cross-flow rate of 5 mL/min, the stage cut, the ratio of permeate flow rate to feed flow rate, was about 0.07, whereas at a cross-flow rate of 18 mL/min, the stage cut was 0.02). The flux through the membrane increased by 8% on going from a cross-flow of 18 mL/min to 42 mL/min (Figure 5), but this is essentially the error in the measurement. Given the constant rejections at flow rates above 18 mL/min, we can conclude that a cross-flow rate of 18 mL/min was sufficient to prevent significant concentration polarization. Effect of Feed Composition on Nanofiltration. For purification of drinking water, it is important that membranes are selective over a range of Cl- and F- concentrations and especially when there is a large excess of Cl- with respect to F-. Table 6 shows that rejections of F- and Cl- by (PSS/PDADMAC)4PSS membranes are essentially independent of solution composition when the ionic strength is below 15 mM. Even when the Cl-/Fratio was about 60, the Cl-/ F- selectivity was still 3. Fluxes were also essentially constant with the various concentrations employed because the osmotic pressure was negligible compared to the driving pressure. These results indicate that (PSS/ PDADMAC)4PSS membranes can effectively separate F- and Cl- under a reasonably wide range of conditions.

Conclusions The simple layer-by-layer deposition of PSS/PDADMAC films on porous alumina substrates results in high-flux membranes capable of selective removal of F- in the presence of either Clor Br-. In the case of (PSS/PDADMAC)4PSS membranes, the selectivity of Cl- or Br- over F- was above 3, and the Br-/Fselectivity was about 12% higher than that for Cl- over F-, suggesting that selectivity is based on the size (Stokes radii) of these ions. However, Cl-/F- selectivity is also very sensitive to (59) Rautenbach, R.; Linn, T.; Eilers, L. J. Membr. Sci. 2000, 174, 231-241. (60) Anne, C. O.; Tre´bouet, D.; Jaouen, P.; Que´me´neur, F. Desalination 2001, 140, 67-77. (61) Lajimi, R. H.; Abdallah, A. B.; Ferjani, E.; Roudesli, M. S.; Deratani, A. Desalination 2004, 163, 193-202. (62) Hu, K.; Dickson, J. M. J. Membr. Sci. 2006, 279, 529-538.

1722 Langmuir, Vol. 23, No. 4, 2007

film architecture and varies with the number of layers in the film. Interestingly, PSS/PAH membranes showed minimal Cl-/Fselectivities regardless of the number of deposited layers. The flux through (PSS/ PDADMAC)4PSS films (∼3.5 m3/m2-day at 4.8 bar) is at least 3-fold higher than that through commercial NF 270 and NF 90 membranes, which show minimal selectivity,

Hong et al.

and (PSS/PDADMAC)4PSS membranes effectively separate Cland F- over a broad range of pressures and concentrations. Acknowledgment. We thank the Department of Energy Office of Basic Energy Sciences and IACM at Hanbat National University (SUH) for financial support. LA061701Y