Photografted Thin Polymer Hydrogel Layers on PES Ultrafiltration

7818

Langmuir 2007, 23, 7818-7830

Photografted Thin Polymer Hydrogel Layers on PES Ultrafiltration Membranes: Characterization, Stability, and Influence on Separation Performance Heru Susanto† and Mathias Ulbricht* Lehrstuhl fu¨r Technische Chemie II, UniVersita¨t Duisburg-Essen, 45117 Essen, Germany ReceiVed February 28, 2007. In Final Form: April 20, 2007 Highly fouling-resistant ultrafiltration (UF) membranes were synthesized by heterogeneous photograft copolymerization of two water-soluble monomers, poly(ethylene glycol) methacrylate (PEGMA) and N,N-dimethyl-N-(2methacryloyloxyethyl-N-(3-sulfopropyl)ammonium betaine (SPE), with and without cross-linker monomer N,N′methylene bisacrylamide (MBAA), onto a polyethersulfone (PES) UF membrane. The characteristics, the stability, and the UF separation performance of the resulting composite membranes were evaluated in detail. The membranes were characterized with respect to membrane chemistry (by ATR-IR spectroscopy and elemental analysis), surface wettability (by contact angle), surface charge (by zeta potential), surface morphology (by scanning electron microscopy), and pure water permeability and rejection of macromolecular test substances (including the “cutoff” value). The surface chemistry and wettability of the composite membranes did not change after incubating in sodium hypochlorite solution (typically used for cleaning UF membranes) for a period of 8 days. Changes in water permeability after static contact with solutions of a model protein (myoglobin) were used as a measure of fouling resistance, and the results suggest that PEGMA- and SPE-based composite membranes at a sufficient degree of graft modification showed much higher adsorptive fouling resistance than unmodified PES membranes of similar or larger nominal cutoff. This was confirmed in UF experiments with myoglobin solutions. Similar results, namely, a very much improved fouling resistance due to the grafted thin polymer hydrogel layer, were also obtained in the UF evaluation using humic acid as another strong foulant. In some cases, the addition of the cross-linker during modification could improve both permeate flux and solute rejection during UF. Overall, composite membranes prepared with an “old generation” nonfouling material, PEGMA, showed better performance than composite membranes prepared with a “new generation” one, the zwitterionic SPE.

Introduction Because fouling significantly reduces the performance of ultrafiltration (UF) membranes, efforts to overcome the fouling problem have drawn more and more attention. Important UF applications include, for example, downstream processing in biotechnology1 and water purification technologies.2 Fouling studies include the identification/characterization of foulants, investigations of fouling mechanisms, and minimizing or control of fouling. Typical foulants for UF membranes are proteins,3 polysaccharides,4 and humic substances.2 Process conditions have been remarkably engineered to achieve better control of membrane fouling, but in most cases, the permeate fluxes are determined by the UF membrane itself. High-performance UF membranes made from polysulfone (PSf) or polyethersulfone (PES) are strongly fouled by all of the substances mentioned above. Therefore, the preparation of low-fouling membranes is strongly needed. Cellulose-based membranes, such as stabilized regenerated cellulose, are the state-of-the-art for low-fouling UF membranes. However, their low chemical stability and relatively low surface porosity are significant limitations. Blending of the membrane polymer (e.g., PES) with a hydrophilic polymeric modifier might in some cases improve the fouling resistance.5 * Corresponding author. E-mail: [email protected]. Tel: +49201-183-3151. Fax: +49-201-183-3147. † Permanent address: Department of Chemical Engineering, Universitas Diponegoro, Indonesia. (1) Charcosset, C. Biotechnol. AdV. 2006, 24, 482. (2) Maartens, A.; Swart, P.; Jacobs, E. P. Water Sci. Technol. 1999, 40, 113. (3) Koehler, J. A.; Ulbricht, M.; Belfort, G. Langmuir 2000, 16, 10419. (4) Ye, Y.; Clech, P. L.; Chen, V.; Fane, A. G. J. Membr. Sci. 2005, 264, 190. (5) Wang, Y. Q.; Wang, T.; Su, Y. L.; Peng, F. B.; Wu, H.; Jiang, Z. Y. Langmuir 2005, 21, 11856.

However, membrane manufacturing from a polymer blend will yield a different pore structure and hence flux and selectivity, and the stability of the modification may be another problem. Therefore, the surface modification of established commercial membranes, while preserving their chemical resistance and mechanical strength, is of great interest for producing lowfouling UF membranes. Two general strategies to reduce the fouling tendency have been proposed depending on the application conditions: the introduction of charged groups to promote electrostatic repulsion and hydrophilization to improve watersurface interaction.6 Surface modifications of UF membranes made of PSf or PES using photoinitiated grafting methods have already been intensively investigated.7-12 The motivation is based on the distinct advantages compared with other technologies (e.g., plasma activation): it can be performed at mild reaction conditions and low temperature, high selectivity is possible by choosing wellsuited photoreactive groups or chromophors and respective excitation wavelengths,12 and it is simple and has a relatively low cost and therefore a wide range of applications.9,11 Earlier work had indicated that photografting may also be a promising technique for drastically reducing the fouling tendency of the base membrane.7-11,13 Nevertheless, to be practically useful, the (6) Ulbricht, M. Polymer 2006, 47, 2217. (7) Susanto, H.; Balakrishnan, M.; Ulbricht, M. J. Membr. Sci. 2007, 288, 157. (8) Taniguchi, M.; Belfort, G. J. Membr. Sci. 2004, 231, 147. (9) Taniguchi, M.; Pieracci, J.; Samsonoff, W. A.; Belfort, G. Chem. Mater. 2003, 15, 3805. (10) Pieracci, J.; Crivello, J. V.; Belfort, G. Chem. Mater. 2002, 15, 256. (11) Pieracci, J.; Crivello, J. V.; Belfort, G. J. Membr. Sci. 2002, 202, 1. (12) Ulbricht, M.; Riedel, M.; Marx, U. J. Membr. Sci. 1996, 120, 239. (13) Ulbricht, M.; Matuschewski, H.; Oechel, A.; Hicke, H. G. J. Membr. Sci. 1996, 115, 31.

10.1021/la700579x CCC: $37.00 © 2007 American Chemical Society Published on Web 06/05/2007

Photografted Hydrogel Layers on PES Membranes

Langmuir, Vol. 23, No. 14, 2007 7819

Scheme 1. Chemical Structure of the Monomers Used for Modification: (a) PEGMA (n ≈ 9) and (b) SPE

resulting membrane performance must further be improved with respect to hydraulic permeability and solute rejection in relation to the fouling resistance as well as with respect to stability in chemical cleaning solutions. Typically, the permeability of the membranes decreased after modification. Then, in most past studies, it had not been carefully analyzed whether the lower fouling tendency was caused by hydrophilization or smaller pore size as a consequence of the grafting process. (It is important to note that in many cases the larger pore size yields a higher fouling tendency.) Moreover, the performance of modified membranes for long-term applications largely depends on the stability of the grafted polymer layer in the chemical cleaning solution. Even though efforts to reduce the chemical cleaning in UF have been made, chemical cleaning is still required in technical processes.14 For example, the stability of grafted PVP, which is one of the best modifier monomers as reported by Belfort et al.,8-11 has not yet been investigated. Polymeric modifier PVP in PSf or PES UF membranes could be either degraded or removed by cleaning using hypochlorite as reported by many authors.14-16 In the present study, photoinitiation of the heterogeneous graft copolymerization will proceed via direct excitation of PES and subsequent PES chain scission17 as well as the action of the formed PES-based radicals as a starter for polymerization.12,18 Selective UV irradiation (>300 nm) at moderate UV intensity has been chosen to ensure high surface selectivity.12 Two watersoluble monomers, PEGMA and the zwitterionic SPE, were used as the functional monomers. The use of PEGMA is based on the fact that PEG derivatives are well-known as nonfouling materials (cf. refs 6, 13, and 19-22). However, PEG may be susceptible to oxidative degradation and chain cleavage in aqueous systems, especially in the presence of transition-metal ions,23 and it looses its resistance to fouling beyond 35 °C.24 Zwitterionic substances have been reported as a new generation of nonfouling material, which has mainly been demonstrated for the modification of nonporous surfaces.23-26 The use of zwitterionic polymeric additives to prepare low-fouling UF membranes via a phaseseparation method has been proposed recently.27,28 Only very few studies have been reported on the grafting of zwitterionic polymers onto polymer membranes. One early example is the grafting of a phospholipid zwitterionic polymer on a cellulose (14) Rouaix, S.; Causeserand, C.; Aimar, P. J. Membr. Sci. 2006, 277, 137. (15) Wienk, I. M.; Meuleman, E. E. B.; Borneman, Z.; van den Boomgaard, T.; Smolders, C. A. J. Polym. Sci., Part A: Polym. Chem. 1995, 33, 49. (16) Qin, J. J.; Wong, F. S.; Li, Y.; Liu, Y. T. J. Membr. Sci. 2003, 211, 139. (17) Kuroda, S.; Mita, I.; Obata, K.; Tanaka, S. Polym. Degrad. Stab. 1990, 27, 257. (18) Yamagishi, H.; Crivello, J.; Belfort, G. J. Membr. Sci. 1995, 105, 237. (19) Ulbricht M.; Richau, K.; Kamusewitz, H. Colloids Surf., A 1998, 138, 353. (20) Akthakul, A.; Salinaro, R. F.; Mayes, A. M. Macromolecules 2004, 37, 7663. (21) Sharma, S.; Johnson, R. W.; Desai, T. A. Langmuir 2004, 20, 348. (22) Harder, P.; Grunze, G.; Dahint, R.; Whitesides, G. M.; Laibinis, P. E. J. Phys. Chem B 1998, 102, 426. (23) Chang, Y.; Chen, S.; Zhang, Z.; Jiang, S. Langmuir 2006, 22, 2222. (24) Chen, S.; Zheng, J.; Li, L.; Jiang, S. J. Am. Chem. Soc. 2005, 127, 14473. (25) Holmlin, R. E.; Chen, X.; Chapman, R. G.; Takayama, S.; Whitesides, G. M. Langmuir 2001, 17, 2841. (26) Kitano, H.; Kawasaki, A.; Kawasaki, H.; Morokoshi, S. J. Colloid Interface Sci. 2005, 282, 340. (27) Sun, Q.; Su, Y.; Ma, X.; Wang, Y.; Jiang, Z. J. Membr. Sci. 2006, 285, 299. (28) Wang, T.; Wang, Y.; Su, Y.; Jiang, Z. J. Membr Sci. 2006, 280, 343.

hemodialysis membrane.29 Recently, Xu et al.30 reported the photografting of phospholipid analogous polymers using benzophenone onto a polypropylene microfiltration membrane. Therefore, it is interesting to evaluate the feasibility of the grafted zwitterionic polymers for UF membrane modification. Furthermore, the use of hydrophilic cross-linker monomer N,N′methylene bisacrylamide (MBAA) was investigated because cross-linking in a grafted polymer hydrogel could lead to a sizeexclusion effect of solutes (e.g., proteins) from the membrane surface with a further reduced fouling tendency as a result. Yang et al.31 had reported that protein adsorption to hydrophobic surface could be completely suppressed by a highly cross-linked layer obtained by photografting using a reaction mixture containing only MBAA. Therefore, the surface functionalization of PES via photografting was also studied from solutions of hydrophilic monomers PEGMA and SPE with added MBAA as well as from solutions containing only MBAA. The identification of parameters, which can be used to control the surface functionalization via photografting, the surface characterization of the resulting composite membranes, and the evaluation of the stability of the grafted polymer layers in a chemical cleaning solution were the main aims of this work. The key target of this work was to obtain low-fouling composite membranes having a cutoff of ∼10 kg/mol but with better performance than commercial membranes with the same specification. Therefore, an analysis of membrane performance was performed by comparing photografted composite membranes with commercial PES UF membrane having a similar water permeability and cutoff. Experimental Section Materials. Commercial PES UF membranes with nominal molecular weight cutoffs (NMWCOs) of 100, 30, and 10 kg/mol obtained from Sartorius AG (Germany) were used. The commercial membranes were selected to minimize the inconsistency in membrane properties. A membrane with a cutoff of 100 kg/mol was used as the base membrane for modification. Prior to use for experiments, the membrane samples were washed with ethanol by shaking at 100 rpm on a mechanical shaker for 1.5-2 h and were then equilibrated with water. The heterogeneity of UF membrane properties is a common problem, in particular for experiments with a small membrane sample area.32 Therefore, to avoid the effects of initial property variation, only membrane samples that had an initial water permeability (L/m2hkPa) in the range of (15% relative to the average values (i.e., 5.71 ( 0.85) were used for the experiments. Poly(ethylene glycol) methacrylate (PEGMA 400, the number indicating the PEG molar mass in g/mol) from Polysciences Inc. (Warrington, PA) and N,N-dimethyl-N-(2-methacryloyloxyethyl-N-(3-sulfopropyl) ammonium betaine (SPE 279, the number indicating the monomer molar mass in g/mol) from Raschig GmbH (Germany) were used as the functional monomers (Scheme 1). N,N′-Methylene bisacrylamide (MBAA) as a cross-linker monomer, myoglobin from horse skeletal muscle (95-100% purity), humic acid, and sodium hypochlorite were purchased from Sigma-Aldrich Chemie GmbH (29) Ishihara, K.; Miyazaki, H.; Kurosaki, T.; Nakabayashi, N. J. Biomed. Mater. Res. 1995, 29, 181. (30) Xu, Z. K.; Dai, Q. W.; Wu, J.; Huang, X. J.; Yang, Q. Langmuir 2004, 20, 1481. (31) Yang, H.; Lazos, D.; Ulbricht, M. J. Appl. Polym. Sci. 2005, 97, 158. (32) Susanto, H.; Ulbricht, M. J. Membr. Sci. 2005, 266, 132.

7820 Langmuir, Vol. 23, No. 14, 2007

Susanto and Ulbricht

(Steinheim, Germany). Polyethylene glycols (PEG 1.5, PEG 6, PEG 10, and PEG 35 from Fluka Chemica GmbH and PEG 100 and PEG 200 from Acros, Geel, Belgium; the number indicates the molar mass in kg/mol) were used for UF rejection measurements. Potassium dihydrogen phosphate (KH2PO4) and disodium hydrogen phosphate dihydrate (Na2HPO4·2H2O) were purchased from Fluka Chemie AG (Buchs, Germany). Nitrogen gas purchased from Messer Griesheim GmbH (Krefeld, Germany) was ultrahigh purity. Water purified with a Milli-Q system from Millipore was used for all experiments. Membrane Modification by Photografting. A UVA print system (Hoenle AG, Gra¨felfing, Germany) equipped with a high-pressure mercury lamp and a glass filter, providing homogeneous UV illumination (wavelength >300 nm) over an area of up to 100 cm2 with an intensity of 35 ( 5 mW/cm2 (measured with a UVA meter, Hoenle AG), was used. Circular PES membrane samples (cutoff 100 kg/mol) with a diameter of either 25 or 44 mm were immersed in monomer solutions (previously degassed by bubbling with nitrogen for at least 10 min) in a Petri dish. A second smaller glass Petri dish was used to cover the membranes and also as another deep-UV filter. After at least 2 min since the first contact between the membrane and monomer solution, the samples were subjected to UV irradiation for various times. Thereafter, the membranes were taken out, immediately rinsed with water, and then washed with excess water to remove any unreacted monomer or physically adsorbed polymer. The washing was sequentially done at room temperature for 30 min, at 50 ( 2 °C for 2 h, and again at room temperature for 30 min. In a modification using UV irradiation only, water was used instead of monomer solutions. Degree of Grafting (DG). DG was gravimetrically determined as the weight increase per membrane outer surface area as described by eq 1 DG )

m m - mo A

(1)

where mo is the initial membrane sample weight, mm is the membrane weight after modification, and A is the outer surface area of the membrane used. All membranes used for DG determination were not used for flux, fouling, and sieving experiments. Control experiments for the washing process as well as the gravimetric method were also performed. Characterization. Chemistry. The membrane surface chemistry was analyzed by attenuated total reflection infrared (ATR-IR) spectroscopy using a Bruker Equinox 55 instrument (Bruker Optics Inc., Billerica, MA) equipped with a liquid-nitrogen detector. A total of 64 scans were performed at a resolution of 4 cm-1 using a diamond crystal; the temperature was 21 ( 1°C. A program written for Opus software from Bruker was used to record the different spectra versus the corresponding background spectra. The elemental analysis was also used to analyze the chemical composition of the membranes. Contact Angle (CA). CA was measured using an optical contact angle measurement system (OCA 15 Plus; Dataphysics GmbH, Filderstadt, Germany). A static captive bubble method, which is preferred for porous membrane surfaces,32 was chosen. Membranes were inverted (active layer to the bottom) in pure water at a temperature of 21 + 1°C. An air bubble (5-10 µL) was injected from a syringe with a stainless steel needle onto the sample surface under water. At least seven measurements of bubbles at different locations were averaged to obtain CA for one membrane sample. Zeta Potential (ZP). The membrane surface charge was investigated via an outer surface streaming potential measurement. Experiments were carried out in a flat-sheet tangential flow module described in detail in a previous study.32 Before measurement, the membrane was equilibrated by soaking in a 0.001 mol/L KCl solution over night. The streaming potentials of membranes were measured using a 0.001 mol/L KCl solution in the range of pH 3-10 at a temperature of 25 ( 1°C. The ZP, ζ, was calculated using the Helmholtz-Smoluchowski equation. Surface Morphology. The top surface morphology of the unmodified and modified membranes was observed by using a Quanta 400

FEG (FEI) environmental scanning electron microscope (ESEM) at standard high-vacuum conditions. A K 550 sputter coater (Emitech, U.K.) was used to coat the outer surface of the sample with gold/ palladium. Membrane Rejection. Rejection tests were conducted with a sixcomponent mixture of PEGs with molar mass ranging from 1.5 to 200 kg/mol at a total concentration of 1 g/L. Experiments were performed with a dead-end stirred filtration system (Amicon cell model 8050, Millipore; cf. membrane performance test) at pressure of 100 kPa and a stirring rate of 300 rpm. Around 10 mL of permeate was collected. The compositions of PEG mixtures in the permeate (Cdownstream) and feed/retentate sides of membrane (Cupstream) were analyzed using gel permeation chromatography (GPC). The apparent rejection for each molar mass was calculated using eq 2.

(

R) 1-

)

Cdownstream × 100% Cupstream

(2)

Membrane Performance Test. The performance of the membrane was investigated with respect to the hydraulic permeability, static adsorption, and ultrafiltration. All experiments were conducted with a dead-end stirred-cell filtration system. The system consisted of a filtration cell (Amicon model 8010, Millipore) connected to a reservoir (∼450 or 1850 mL). It was pressurized by nitrogen. To avoid the effects of membrane compaction on the interpretation of modification and fouling data, each sample was first compacted by the filtration of pure water at a pressure of 450 kPa for at least 0.5 h. Thereafter, the pressure was reduced to the desired pressure for the water flux measurement. To know the effect of modification on membrane hydraulic permeability, the water flux was measured before and after modification at a constant pressure of 300 kPa. All pure water fluxes were measured (by the gravimetric method) until the consecutively recorded values (for periods of 5 min) were considered to be constant (i.e., they differed by less than 4%). The evaluation of membrane performance was expressed in terms of the hydraulic permeability ratio (eq 3). For static adsorption experiments, the water flux was initially measured, and then a test solution of myoglobin (1 g/L; pH 7 in phosphate buffer, prefiltered through a 0.45 µm microfilter, Sartorius, Germany, to remove undissolved material) was added to the cell. Thereafter, the outer membrane surface was exposed for 2 h without any flux at a stirring rate of 300 rpm. Our previous study has shown that 2 h of adsorption was sufficient to achieve saturation of the surface adsorption capacity for this protein.33 Then the test solution was removed, and the membrane surface was rinsed two times by filling the cell with pure water (5 mL) and shaking it for 30 s. Water fluxes before and after exposure were compared. The evaluation of membrane performance was expressed in terms of the relative flux reduction, RFR (eq 4). In addition, the fouling resistance term, Rf, was also introduced to evaluate the membrane performance during adsorptive fouling resistancehydraulic permeability analysis (eq 5). A fouling resistance of 1 means no adsorptive fouling has occurred. hydraulic permeability ratio )

RFR )

Jam - Jads Jam

R f ) 1 - RFR

Lpo Lpm

(3)

(4) (5)

where Lpo and Lpm are the membrane hydraulic permeabilities before and after modification, respectively, and Jam and Jads are the water fluxes of the modified membrane before and after exposure to the protein solution test, respectively. In the RFR calculation for the unmodified membrane, the initial water flux (Jo) was used instead of Jam. Ultrafiltration experiments at a constant transmembrane pressure (100 kPa) were conducted using a myoglobin solution (33) Susanto, H.; Franzka, S.; Ulbricht, M. J. Membr. Sci. 2007, 296, 147.

Photografted Hydrogel Layers on PES Membranes (1 g/L; pH 7 in phosphate buffer) as the feed. The use of constant pressure mode was due to relatively short operation. In addition, all membranes tested (except the unmodified PES membrane with a 100 kg/mol NMWCO) had similar initial water fluxes. The balance was connected to the PC, the weight of the permeate was recorded online, and the flux was calculated. Myoglobin concentrations were determined by measuring UV absorbance at 230 nm. Profiles of permeate flux and apparent myoglobin rejection over time were investigated. Ultrafiltrations of humic acid solutions (50 mg/L, pH 7.2, 1 mM Ca2+ added, conductivity 1100 µS/cm, prefiltered through a 0.45 µm microfilter, Sartorius, Germany) were performed in the same way, but with a similar initial water flux instead of an identical transmembrane pressure; humic acid concentrations were determined by measuring UV absorbance at 255 nm. Stability Test. Sodium hypochlorite was chosen because it is usually used for membrane cleaning.14-16 The stability of the grafted polymer layer was examined by incubating membrane samples in sodium hypochlorite solution (active chlorine concentration 500 mg/ L) for up to 8 days. This time seemed sufficient to achieve the maximal degradation effect.14 Samples were rinsed with pure water before measurements. ATR-IR spectroscopy, supported by contact angle measurements and elemental analyses, were used to identify changes in the surface or bulk composition of the modified membranes. Degree of Swelling of Functional Polymer Hydrogels. The method used to determine the degree of swelling used in this supporting experiment was adapted from the synthesis of functional polymer hydrogels reported in detail elsewhere.34 Briefly, PEGMAor SPE-based hydrogels were prepared by free radical polymerization in aqueous solution using the following composition: 6.211 g of either PEGMA or SPE, 0.311 g of MBAA, and 47.478 g of water were mixed in the glass vessel. Thereafter, 43.29 mg of ammonium peroxodisulfate and 173.2 mg of N,N,N′,N′-tetramethyl ethylene diamine were added to initiate the polymerization. The moderate monomer/cross-linker ratio has been selected so that swelling experiments could be performed and yield relevant information about differences in the water uptake of the two different (homo)polymers. The glass vessel was then closed and kept for 24 h at room temperature. The gels were cut to a small size (∼1 cm × 1 cm × 1 cm) and washed by repeated transferring into pure water. The gels were weighed, then dried at 80 °C for 24 h in a vacuum oven, and then weighed again. The degree of swelling of the gel is defined as the mass of the swollen gel relative to the mass of the dried gel.

Results and Discussion Degree of Grafting (DG). The amount of grafted polymer on the membrane was measured for modifications without and with cross-linker (Figures 1-3). All of the data were normalized to the outer surface area of the membranes. However, as also discussed on the basis of all results of this study (cf. below), this should definitely not imply that the functionalization has taken place only on the outer membrane surface. The DG increased with increasing monomer concentration. However, at a concentration beyond 40 g/L, the increase in DG was only very slight with further increases in monomer concentration (Figure 1) implying that the “optimal” monomer concentration was 40 g/L (cf. also below). Therefore, the following experiments were mainly conducted using a monomer concentration of 40 g/L. More pronounced was the influence of UV irradiation time. With increasing UV time, the DG would rise for both modifications without and with cross-linker (Figures 1 and 2). These results suggest that DG was influenced by the amount of monomer as well as free radicals in the grafting zone, which increased upon increasing the monomer concentration in the bulk and the UV irradiation time, respectively. However, for modifications without cross-linker the increase in the grafted chain length could diminish (34) Fa¨nger, C.; Wack, H.; Ulbricht, M. Macromol. Biosci. 2006, 6, 393.

Langmuir, Vol. 23, No. 14, 2007 7821

Figure 1. Effect of UV irradiation time and monomer concentration (without cross-linker) on DG. The numbers in the legend indicate monomer concentrations (g/L). Control experiment performed by washing the unmodified membrane via a similar procedure used for the modification did not significantly change the membrane weight (-3.5 µg/cm2). The variation of DG data for modifications where two or three independent experiments have been performed was smaller than 10 µg/cm2.

Figure 2. Effect of UV irradiation time on DG for modification using PEGMA and SPE (solid marker) at various cross-linker concentrations. M and C in the legend are the monomer and crosslinker, respectively, whereas the numbers indicate the concentrations (g/L). For information about the experimental error, see Figure 1.

the reactivity of the growing polymer radicals and slow down the diffusive transport of monomer into the grafting zone as evidenced by the decreasing slope of the plots at longer irradiation times. Moreover, for the same concentration and UV irradiation time, modification using PEGMA yielded greater DG values than modification using SPE. The data as shown in Figures 2 and 3 and also after converting the DG values and the monomer concentrations from µg/cm2 to µmol/cm2 and µg/L to µmol/L, respectively, revealed that the grafting rate was larger for PEGMA than SPE. With the used initiation method and conditions, it is reasonable that termination by recombination of growing grafted chains with macroradicals in solution (as evoked, for example, by Ruckert and Geuskens35) will have a very low probability because, with selective UV irradiation, radicals are generated only from the excited PES backbone in the solid surface and not from monomer or solvent. The apparently higher reactivity of the larger monomer PEGMA as compared with SPE was (35) Ruckert, D.; Geuskens, G. Eur. Polym. J. 1996, 32, 201.

7822 Langmuir, Vol. 23, No. 14, 2007

Figure 3. Effect of cross-linker content in the reaction mixture on DG (for grafting at 5 min of UV irradiation). For information about the experimental error, see Figure 1.

somewhat surprising (cf. refs 13 and 36 for monomer characteristics that influence the grafting reaction). An explanation may be based on the better solvation of polyPEGMA in water as indicated by the swelling data (cf. below). Overall, an increasing grafting density and longer average length of grafted chains are expected with increasing UV time. Figure 2 shows that the addition of cross-linker could either decrease (for PEGMA) or increase (for SPE) the DG relative to the modification without cross-linker. The reduction of DG for PEGMA seemed to be a function of cross-linker concentration in the bulk solution. Indeed, further experiments showed a similar trend for both monomers (i.e., as the cross-linker content was increased, the DG would first fall and then rise beyond the value for the uncross-linked grafted layers (Figure 3)). The phenomenon of decreasing DG with cross-linker addition had also been observed by other authors.37,38 By contrast, enhancing the DG via addition of cross-linker has also been found in the literature.39,40 Overall, these phenomena could be explained as follows: the presence of cross-linker could, on the one hand, restrict the diffusion of monomer into the grafting zone leading to a decrease in DG. On the other hand, the presence of cross-linker will also lead to side groups with double bonds in the grafted chain, and then branching or even cross-linking reactions can occur, leading to an increase in DG. The net effect of cross-linker on grafting will also depend on the monomer characteristics. As seen in Figure 3, the effects upon reducing the grafting efficiency seem to dominate, but the mass-based effects on DG are relatively small. Only for grafted polySPE at high cross-linker content is the DG strongly increased, and this may indicate a different mechanism for grafted layer growth (presumably involving phase separation; see below). Membrane Characterization. Chemistry. Significant changes in IR spectra after the modifications confirmed that the functional polymer has been photochemically grafted onto the PES base membrane. Both grafted polymers were identified by additional IR bands at ∼1725 and ∼1727 cm-1 for PEGMA and SPE, respectively, which are due to CdO vibration of the ester group of the methacrylates. In addition, IR bands due to aliphatic C-H (36) Yang, W.; Rånby, B. J. Appl. Polym. Sci. 1996, 62, 545. (37) Kai. T.; Goto, H.; Shimizu, Y.; Yamaguchi, T.; Nakao, S.; Kimura, S. J. Membr. Sci. 2005, 265, 101. (38) Gupta, B.; Bu¨chi, F.; Scherer, G.; Chapiro´, A. J. Membr. Sci. 1996, 118, 231. (39) Dworjanyn, P. A.; Garnett, J. L. J. Polym. Sci., Part C: Polym. Lett. 1988, 26, 135. (40) Mika, A. M.; Childs, R. F.; Dickson, J. M.; McCarry, B. E.; Gagnon, D. R. J. Membr. Sci. 1997, 135, 81.


stretching at ∼2925 and ∼2875 cm-1 were also found for both grafted polymers. It is interesting that the intensity of those bands increased linearly with increasing DG. This was similar to the results of previous studies.7,9,13,19 However, the characteristic absorptions of the PEG side group (νCO ∼1100 cm-1) in PEGMA and of the SO32- group (∼1220 and ∼1040-1070 cm-1) in SPE could not be detected in the IR spectra. Thus, overlapping bands with the base membrane polymer as had also been found in a previous study by Belfer et al.41 could be the reason. The elemental analyses support the assertion that a chemical modification has indeed occurred (cf. Table S1, Supporting Information). Overall, the C and H contents increased after modification. A slight increase in S content was observed for membranes modified using SPE. As expected considering the contents in the reaction mixtures, the addition of cross-linker did not lead to observable differences in IR spectra or elemental analyses. Contact Angle. PEGMA-modified membranes had CA values ranging from 37.8 ( 4.2° to 42.5 ( 4.3° depending on the degree of grafting (cf. Figure S1, Supporting Information). These values were somewhat smaller than for the unmodified base membrane (100 kg/mol NMWCO: 44.8 ( 4.2°) but were much lower than for the unmodified membrane with a similar cutoff (10 kg/mol NMWCO: 61.7 ( 2.5°; see below). These results agree well with previously reported CA data for PEGMA-modified PAN membranes19 and are somewhat smaller than in our recent study for the grafting-from functionalization of a different PES UF membrane (∼47° for DG ) 168 µg/cm2).7 Relatively constant values of CA were observed for SPE-modified membranes (ranging from 44.1 ( 3.8° to 45.3 ( 3.0°). These data are in good agreement with the CA for PAN-based zwitterionic (sulfobetaine) polymer blend membranes.27 Care should be taken to interpret these results. The CA of the modified membrane is affected by the base membrane (e.g., surface porosity), the degree of surface coverage (related to DG), and the structure of the grafted polymer (e.g., chain length). If the membrane surface has been completely covered by the grafted polymer, then the contact angle will mainly depend on the hydrophilicity/ hydrophobicity balance of this polymer. The water CA for polyPEGMA had been found in the literature to be between 35 and 50°.42-44 For a membrane with incomplete surface coverage, both the base material and grafted polymer will contribute to CA. We believe that the relatively high surface porosity of the base PES membrane (100 kg/mol NMWCO; Figure 5a) contributed to contact angles (45°) that were much lower than typically measured for nonporous PES (76°33), whereas the CA data for the modified membranes are mainly due to the hydrophilic grafted polymer. This is supported by the much higher CA of the unmodified PES membrane (10 kg/mol NMWCO; CA ) 62°). Also, during modification a decrease in the effective PES surface porosity through pore narrowing and pore blocking could not be avoided (see SEM images, rejection, and water permeability reduction below). Furthermore, the slightly lower CA data for the polyPEGMA-modified than for the polySPE-modified membranes were most probably due to the higher degree of swelling of the grafted layer in water (see below). The reduction of swelling in the polymer hydrogel layers caused by chemical cross-linking with MBAA might then explain the slightly higher CA values for all modified cross-linked membranes compared to their uncross-linked counterparts. (41) Belfer, S.; Fainchtain, R.; Purinson, Y.; Kedem, O. J. Membr. Sci. 2000, 172, 113. (42) Singh, N.; Cui, X.; Boland, T.; Husson, S. M. Biomaterials 2007, 28, 763. (43) Yu, W. H.; Kang, T.; Neoh, K. G.; Zhu, S. J. Phys. Chem. B 2003, 107, 10198. (44) Uchida, E.; Uyama, Y.; Ikada, Y. Langmuir 1994, 10, 481.


Langmuir, Vol. 23, No. 14, 2007 7823

Figure 4. Zeta potential, ζ, of unmodified (PES-U) and modified membranes with varying pH calculated from the tangential streaming potential of the outer surface (using 0.001 mol/L KCl).

Zeta Potential. As clearly seen in Figure 4, the surface of both unmodified membranes had a negative charge over the entire pH range studied, and the absolute values decreased to acidic pH values. (The reasons for this phenomenon have been discussed in our previous study.32) Indeed, the effective surface charge of the PES base membrane has been decreased toward neutral by the grafted polymer hydrogel layers, even though all composite membranes were not fully neutral. However, it is important to note that a linear decrease in ZP with increasing pH, observed for both different grafted polymers, reflects the typical behavior of a surface that has a negative charge due to anion adsorption.45-47 For grafted polyPEGMA, this interpretation is supported by results of previous studies by Ulbricht et al.19 and by Uchida et al.44 In addition, the ZP data also support the interpretation of CA data with respect to coverage of the outer membrane surface (i.e., PES-g-polyPEGMA membranes with different DGs (e.g., ∼170 and ∼320 µg/cm2) showed significantly different ZPs (Figure 4) even though they had almost the same CA (see above). Sequestering of mobile ions in the dipolar layer (e.g., positive inner surface, negative outer surface) had been discussed in previous ZP studies for zwitterionic micelles.48,49 However, considering the shape of the zeta potential as a function of pH in our study, the dissociation of fixed ions should not be involved. Therefore, the ZP data for grafted polySPE was mainly dependent on the adsorption of anions from the bulk solution. Another reasonable explanation is that the charged groups of the zwitterionic polymer might be evenly distributed over the entire membrane surface, generating an interface that is overall electrically neutral. This hypothesis had also been suggested by Holmlin et al.25 Then, the reduction of effective surface charge was larger with increasing DG. This phenomenon is similar to the results of previously reported ZP data.7,45 Furthermore, for (45) Chan, Y. H. M.; Schweiss, R.; Werner, C.; Grunze, M. Langmuir 2003, 19, 7380. (46) Schwarz, S.; Jacobasch, H.-J.; Wyszynski, D.; Staude, E. Angew. Makromol. Chem. 1994, 221, 165. (47) Werner, C.; Jacobasch, H. J.; Reichelt, G. J. Biomater. Sci., Polym. Ed. 1995, 7, 61. (48) Baptista, M. D.; Cuccovia, I.; Chaimovich, H.; Politi, M. J.; Reed, W. F. J. Phys. Chem. 1992, 96, 6442. (49) Iso, K.; Okada, T. Langmuir 2000, 16, 9199.

similar DG, the membrane modified using SPE showed a less negative ZP than the membrane modified using PEGMA. Surface Morphology. As presented in Figure 5, SEM images support the preceding characterization results and the rejection behavior (see below). The unmodified membrane had a smoother surface and apparently a larger surface porosity than all modified membranes. The entire membrane surface seemed to be covered by a thin polymer layer (the only exception was the membrane after UV irradiation alone without monomer, which seemed to be unchanged), indicating that the membrane surface has been evenly modified. Both pore narrowing and pore blocking were observed as consequences of the surface modification. These effects were more pronounced for the modifications using SPE. Only for modifications performed in the presence of the crosslinker monomer can a trend to less homogeneous, presumably phase-separated morphologies be observed (Figure 5c,d). Especially for SPE, the addition of MBAA seemed to lead to a more compact and heterogeneous morphology of the grafted polymer. For functionalizations performed with solutions of only MBAA, this phenomenon was even more pronounced (see Figure S2, Supporting Information). It should be kept in mind that all of the SEM images were obtained under dry conditions. Therefore, the actual surface morphology during ultrafiltration would be even more different from that one of the unmodified PES membrane as a result of the pronounced hydrogel character of both grafted polymers (see below). Nevertheless, the layers are apparently much thinner than what would be expected for the deposition of all grafted polymers on the outer membrane surface. (Note that a DG of 250 µg/cm2 would yield a dry layer thickness of about 2.5 µm.) Membrane Rejection. The rejection of polydisperse macromolecular test substances (PEGs) was measured to investigate in more detail the effect of modification on active-layer pore structure. As clearly seen in Figure 6, UV irradiation alone had no effect on the pore structure of the membrane (see also Figure 5). It should be noted that this had not been achieved in many similar previous attempts to functionalize the PES UF membrane via UV-initiated grafting.8-11 By contrast, the modification significantly shifted the membranes’ nominal cutoff, and hence

7824 Langmuir, Vol. 23, No. 14, 2007


Figure 5. SEM images of the top surface of the unmodified and the modified membranes: (a) unmodified membrane PES UF (100 kg/mol), (b) 5 min of UV irradiation (only), (c) PES-g-polyPEGMA (40 g/L, 5 min), (d) PES-g-polyPEGMA/MBAA (40/2 g/L, 5 min), (e) PESg-polySPE (40/2 g/L, 5 min), and (f) PES-g-polySPE/MBAA (40/2 g/L, 5 min).

apparent pore sizes, to lower values. Because the water permeabilities were also reduced (see below), both pore blocking and pore narrowing have apparently occurred during modification. All modified membranes (except M2) had rejection curves with similar slopes and an NMWCO within the range 10 kg/mol. The rejections of the modified membranes were slightly smaller than for unmodified 10 kg/mol but much larger than for unmodified 100 kg/mol original PES UF membranes. These observations indicate that the grafting influenced not only the outer membrane surface but also the membrane pores. Hence, it is proposed that the following steps occur during modification: Considering the high values of obtained DGs, photografted polymer is formed as a thin layer not only on the outer surface of the membrane but also in the pores of the UF membrane. The UV intensity at the bottom membrane surface, measured with a UVA meter

(Experimental Section), was only about 5% compared to the actual excitation intensity. Hence, because of the strong UV absorbance of PES (which is the basis for controlled photodegradation12,17) and light scattering, there will be a gradient of the amount of grafted polymer over membrane thickness, and most of the grafted polymer will be located in the upper several micrometers below the (irradiated) top surface of the membrane (i.e., also “behind” the UF active layer, which has a thickness of less than 1 µm6,32). As DG is further increased (e.g., by increasing the UV irradiation time), the grafted chains would be stretched away from the solid surface, and the interior structure of the porous active layer of the base membrane (rejection curve) would be influenced more significantly. This is analogous to results and discussions in our earlier publication.7


Langmuir, Vol. 23, No. 14, 2007 7825

Figure 6. Rejection curve of unmodified and modified membranes determined using a PEG mixture (1 g/L) at a pressure of 100 kPa. U1 and U2 are 100 and 10 kg/mol unmodified membranes, respectively. M1 is a modified membrane using UV (only); M2 is PES-g-polyPEGMA (40 g/L, 2 min); M3 is PES-g-polyPEGMA (40 g/L, 5 min); M4 is PES-g-polyPEGMA/MBAA (40/0.4 g/L, 5 min); M5 is PES-g-polyPEGMA/ MBAA (40/2 g/L, 5 min); M6 is PES-g-polySPE (40 g/L, 6 min); and M7 is PES-g-polySPE/MBAA (40/2 g/L, 5 min).

Figure 7. Normalized absorbance intensity (by subtracting the absorbance of the base membrane) at wavenumbers of ∼1725 cm-1 for PEGMA- and 1727 cm-1 for SPE-modified PES membranes after soaking in sodium hypochlorite solution.

Furthermore, for modified membranes with similar (relatively small) DG values (M2 and M6), grafted polySPE (M6) yielded a much higher rejection but a lower hydraulic permeability than grafted polyPEGMA (M2). The addition of the cross-linker during photografting led to significant reductions of the apparent pore size as compared to that of the respective uncross-linked hydrogels (cf. M3 and M4; M5; M6 and M7; Figure 6). The difference in the grafted layer structure was the most probable reason for these observations (see below). Because the grafted polymer hydrogel had a strong impact on the membrane separation function, the modified membranes should be considered to be PES-based composite UF membranes.6 Measurements of the Degree of Swelling of Functional Polymers. To help explain the characteristics of the effects of a hydrogel grafted onto a membrane, a supporting experiment

to measure the water uptake of both polymers was performed. It was found that the degree of swelling of polyPEGMA was 12.3 ( 0.3 whereas the value for polySPE was 7.7 ( 0.2. This implies that the PEGMA-based hydrogel was significantly more swollen than SPE-based one. Stability Test Study. The stability of the grafted polymer layer in sodium hypochlorite solution was evaluated. As presented in Figure 7, ATR-IR absorbance for the introduced ester groups, relative to groups in the PES backbone, did not change after incubating for increasing times. (It should be noted that the ATRIR absorbance ratio was well correlated with the DG.) These results are supported by CA measurements (data not shown) and by elemental analyses (Table S1, Supporting Information); both types of data did not change significantly with increasing incubation time. Even though many previous reports had shown that sodium hypochlorite could degrade the PSf or PES membrane polymers, this study clearly indicates that no loss or hydrolysis of the (very thin) grafted layers occurred within 8 days, proving the chemical stability of the active layer of the photografted composite membranes. Membrane Performance Based on Permeabilitys Adsorptive Fouling Resistance Analysis. Hydraulic permeability and adsorptive fouling resistance (eq 5) were used to evaluate the performance of modified membranes and then finally to select those that perform significantly better than the original membranes. Surface modification caused simultaneous membrane pore narrowing and even blocking, which decreased the water permeability, and surface hydrophilization, which increased the membrane fouling resistance (Figure 8). However, it is important to note that the reduction of pore size could also contribute to the reduced effects of the interactions of the membrane with the solute. The influence of membrane pore structure on solute adsorption on/in the PES UF membrane was confirmed by our recent study.33 Nevertheless, the tradeoff analysis shows very clearly the dominating effect of the grafted thin polymer hydrogel layers on the achieved high fouling resistance. The preparation

7826 Langmuir, Vol. 23, No. 14, 2007


Figure 8. Hydraulic permeability-fouling resistance analysis: (a) PEGMA-modified and cross-linker (alone)-modified membranes and (b) unmodified and SPE-modified membranes. PES-U10, PES-U30, and PES-U100 are 10, 30, and 100 kg/mol unmodified membranes, respectively. Membrane-fouling resistance was evaluated using myoglobin (1 g/L, pH 7, 2 h of exposure). The numbers inside the plot indicate the UV irradiation time whereas those in the legend indicate the monomer (M) and cross-linker (C) concentrations (g/L).

conditions of the composite membrane can be used to fine tune the performance (Figures 1 and 2 and Supporting Information). Figure 8b reveals clearly that all original PES membranes had high hydraulic permeability but showed strong fouling. Hence, modified membranes should have similar water permeability compared to that of the PES UF membrane with 10 kg/mol NMWCO, but they should have higher fouling resistance (which was the key target of this work). The performance indicators were then defined (i.e., the hydraulic permeability should be more than 0.75 L/m2 h kPa, and the fouling resistance should be higher than 0.9). As seen in Figure 8, all composite membranes with an uncrosslinked grafted hydrogel layer displayed a tradeoff relationship with respect to fouling resistance and membrane permeability (i.e., membranes with higher fouling resistance had lower hydraulic permeability; see also Figures S3 and S4, Supporting Information). Interestingly, many composite membranes with an uncross-linked polyPEGMA layer have fully achieved both performance criteria defined above (Figure 8a). For example,

composite membranes prepared at a PEGMA concentration of 40 g/L with UV irradiation time ranging from 3 to 7 min yielded hydraulic permeabilities between 2.0 and 0.9 L/m2 h kPa and fouling resistance values ranging from 0.91 to 0.94. Different behavior was observed for composite membranes with a crosslinked hydrogel layer (i.e., as the degree of cross-linking was increased by increasing the UV irradiation time, the water permeability would continuously decrease whereas the membrane fouling resistance would first increase and then decrease even though a more arbitrary trend was observed for grafted polySPE; see also Figures S5 and S6, Supporting Information). Somewhat surprisingly, no PES-g-polySPE composite membrane (uncrosslinked and cross-linked) would achieve these criteria (Figure 8b). However, when compared to the unmodified membranes, some of them had at least promising performance. Therefore, it was still reasonable to include some of the PES-g-polySPE composite membranes in the further evaluation. Overall, it is observed that composite membranes prepared under different conditions (concentration, UV irradiation time,


Langmuir, Vol. 23, No. 14, 2007 7827

Figure 9. Flux profile as a function of time for various modified membranes during the ultrafiltration of myoglobin solution (1 g/L, pH 7) at a transmembrane pressure of 100 kPa. Jo is the initial water flux (L/m2 h). PES-U10 and PES-U100 are 10 and 100 kg/mol unmodified membranes. Table 1. Initial Water Flux, Water Flux after Ultrafiltration and External Cleaning, and Apparent Protein (Myoglobin) Rejection during Ultrafiltration

a

no.

membrane

initial water flux (Jo), L/m2 h

1 2 3 4 5 6 7

PES-U10 PES-U100 PES-g-PEGMA PES-g-PEGMA/MBAA (40/0.4) PES-g-PEGMA/MBAA (40/2) PES-g-SPE PES-g-SPE/MBAA (40/2)

95.2 525.0 93.5 84.3 71.3 87.2 80.1

water flux after filtration external cleaning, L/m2 h

myoglobin rejection (%)a

42.6 248.8 83.2 76.8 56.5 68.0 57.7

69.4 ( 1.2 15.7 ( 1.3 57.5 ( 1.0 61.3 ( 0.9 66.0 ( 1.1 59.1 ( 1.3 64.6 ( 1.4

Average value from filtration times of 20, 60, 90, and 120 min.

monomer type, and cross-linker content) showed different permeability and fouling resistance. Differences in grafting density and grafted chain length leading to different chain conformations in the grafted layer (mushroom vs brush) are the most probable reason for these differences in membrane performance and membrane surface characteristics (see above). Even though determinations of grafting density and chain lengths have been reported in previous publications (e.g. refs 50 and 51), in the context of our study (with PES as the substrate and covalently grafted polyacrylates) such analyses cannot be performed. With respect to the influence of the main modification parameters (UV time and monomer concentration), it is difficult to discuss separately the effects on grafting density and chain length. However, on a qualitative level this is possible for the results of this study. An increase in fouling resistance with increasing degree of grafting but also with increasing UV irradiation time and monomer concentration was convincingly observed (Figures 1 and 8). Comparing the membranes with similar DG and fouling resistance but prepared at different UV irradiation times (e.g., PEGMA-modified membranes from 20 g/L at 15 min irradiation and from 40 g/L at 4 min irradiation; Figure 8a), it is reasonable that membranes modified at longer irradiation time (but lower monomer concentration) had a higher grafting density but a shorter chain length than (50) Sofia, S. J.; Premnath, V.; Merrill, E. W. Macromolecules 1998, 31, 5059. (51) Zdyrko, B.; Klep, V.; Luzinov, I. Langmuir 2003, 19, 10179.

modified membranes at shorter irradiation time (but higher monomer concentration). This suggests that there is no need to have a high grafting density to obtain high fouling resistance when a large enough chain length has already been achieved. Above a critical value, the grafting density did not seem to influence the fouling resistance anymore. This is supported by the fact that no significant difference in fouling resistance was observed between membranes obtained from 50 and 40 g/L at the same irradiation time (Figure 8a). The identification of such trends is somewhat more complicated for the SPE-modified membrane (Figure 8b), but overall similar influences can be discussed. The presence of cross-linker in the reaction mixture seemed not to improve the adsorptive fouling resistance and the hydraulic permeability relative to those of the respective composite membranes with uncross-linked polyPEGMA. Nevertheless, two membranes modified using monomer/cross-linker concentrations of 40/0.4 g/L (at 5 min irradiation) and 40/2 g/L (at 4 min irradiation) could fulfill both performance criteria. Increasing the degree of cross-linking should increase the size exclusion of macromolecular solutes from the membrane surface. However, phase separation might occur at too a high degree of crosslinking, and thus a heterogeneous structure of the grafted layer might be formed (see above). These heterogeneities may increase the membrane-solute interactions. Such insoluble grafted layers were also identified during other photoinitiated surface-grafting

7828 Langmuir, Vol. 23, No. 14, 2007


Figure 10. Flux profile as a function of time during the ultrafiltration of humic acid (50 mg/L, pH 7.2, 1 mM Ca2+, ionic strength (total) 1100 µS/cm) through composite membranes M3 and M4 and the unmodified PES membrane (cutoff 10 kg/mol) at similar initial water fluxes (∼92 L/m2 h). The images on the right show photographs of the fouled membrane surface after external cleaning with water.

studies.35 In addition, stronger solute interactions with grafted MBAA segments might also be possible (see data for PES membranes photografted from pure MBAA solutions; Figure 8a and Figure S6, Supporting Information). Overall, in all cases, a significant contribution of the photografted thin polymer hydrogel layer to the flux and selectivity of the resulting composite membranes have been observed. The interaction of the photografted membranes with sufficiently high DG (approximately >300 µg/cm2 for polyPEGMA and >170 µg/cm2 for polySPE, along with this base membrane) with solutes, including adsorptive fouling, should be controlled by the structure of the grafted layer, which is much more hydrophilic and less prone to anion adsorption than the surface of the membrane polymer PES. Ultimately, the composite membranes already characterized with respect to rejection curves (see Figure 6 for preparation conditions) were selected for the stability and UF performance tests. Membrane Performance Based on Ultrafiltration. UF experiments were conducted to determine the performance of the novel composite membranes from an application point of view. Indeed, a common phenomenon was also observed in this study: the membrane with larger pore size leading to high flux yielded more severe fouling than the membrane with smaller pore size even though it had a smaller flux loss in the beginning of operation (Figure 9, Table 1). All composite membranes had a much higher flux ratio than both unmodified membranes. Even in cases where the addition of cross-linker did not improve the adsorptive fouling resistance relative to the uncross-linked composite membranes (Figure 8a), it improved both the permeate flux and rejection in the UF experiments. The unmodified membranes had permeate fluxes of only ∼20% (for 100 kg/mol) and ∼30% (for 10 kg/mol) relative to the initial water flux whereas the composite membranes had more than 60% for SPE-based types and more than 80% for PEGMAbased types. The unmodified membrane with a 10 kg/mol NMWCO had the highest protein rejection whereas the unmodified membrane with a 100 kg/mol NMWCO had the lowest

rejection. All composite membranes had a slightly lower protein rejection than the 10 kg/mol unmodified membrane. Furthermore, composite membranes with a cross-linked hydrogel layer displayed a higher rejection than the respective uncross-linked types. These protein rejection results agree well with rejection curves obtained with PEGs (Figure 6). PES-g-polyPEGMA and PES-g-polyPEGMA/MBAA composite membranes were also examined for UF of solutions of humic acid, which is well known as a strong foulant in the filtration treatment of surface water. As presented in Figure 10, it is obvious that both composite membranes had higher permeate fluxes than the unmodified membrane. The apparent humic acid rejections for all three membranes (93 ( 3%) were identical within the range of error. The photographs show the membrane surfaces after external cleaning by rinsing with water, and significant differences in the appearance of fouling were observed: a darkbrown layer on the original membrane indicates strong fouling whereas the outer surfaces of the composite membranes seemed almost clean. The fouling tendency seemed slightly larger for the membrane with a cross-linked hydrogel layer. These observations correlated quite well with the water fluxes after external rinsing (unmodified PES, 59 L/m2 h; PES-g-polyPEGMA/ MBAA, 79 L/m2 h; PES-g-polyPEGMA, 83 L/m2 h; note that the initial water fluxes were identical, i.e., 92 L/m2 h). All UF results show very convincingly that the grafted polymer hydrogel layers completely changed the strength of interactions with macromolecular solutes and hence lead to a very pronounced improvement of the fouling resistance and cleanability of PESbased UF membranes. Influence of Thin Grafted Hydrogel Layers on Membrane Flux and Selectivity and Mechanism of the Fouling Reduction of Composite Membranes. For the linear grafted copolymers (g-polyPEGMA and g-polySPE) at DG values above the critical value (approximately >300 µg/cm2 for polyPEGMA and >170 µg/cm2 for polySPE, see above), the PES surface in contact with the feed solution is completely covered by the grafted polymer layer, so the resistance to fouling by biomacromolecules is high.


Figure 11. Schematic drawing of grafted polyPEGMA and polySPE chains on the outer surface of a PES UF membrane and interactions with macromolecular solutes (gray circles).

Even when the polymer chains will not necessarily be in the brush regime, the internal layer structure will have an influence on flux, rejection, and fouling resistance. We believe that the degree of swelling of this layer is an important parameter: At the same (mass-based) degree of modification, polyPEG-based layers will be more swollen than polySPE-based ones, leading to a thicker but less dense structure. This assumption can then help us to understand the higher permeability and lower rejection as well as the higher fouling resistance of PEGMA-based composite membranes (Figure 11; Figures 6 and 8 and Supporting Information). The MBAA-cross-linked grafted layer structure, causing a reduction in swelling, leads to increased membrane rejection at similar DG values (observed for PEGs and the protein myoglobin; Figure 6 and Table 1). This shows that the internal structure of the grafted polymeric hydrogel has a significant influence on membrane selectivity and can be explained by solute sieving through a swollen hydrogel network with a mesh size that is adjustable by the cross-linker content (e.g., ref 34). Overall, the increase in hydrophilicity and the reduced tendency for ion adsorption imparted by the thin grafted neutral hydrogel layers on the PES UF membrane surface improved the proteinresistant character because the main driving force for adsorption has been very much reduced. The same property is also very useful in reducing the adsorption of many other foulants (here shown for humic acid, where the aromatic groups also promote binding to hydrophobic surfaces). Therefore, the PEGMA- and SPE-based composite membranes were much less prone to fouling than the unmodified membranes with respect to adsorptive fouling as well as dynamic fouling. However, the detailed mechanism has not been exactly understood, and possible explanations can be drawn from existing theories,52-54 supported by recent experimental studies (e.g., refs 55-57). The adsorption of solutes to hydrophobic solid surfaces is mainly driven by increasing the system’s entropy via replacement of water molecules at the surface (surface dehydration).53 That is the reason for much less attractive interactions between biomolecules and hydrophilized surfaces. Recently, it has been proposed that “kosmotropes” as functional groups at surfaces form a general basis for protein-resistant surfaces.54 Both oligoethyleneglycol (as in PEGMA) and zwitterionic groups (as in SPE) are nonionic kosmotropes. However, when linked to polyacrylate chains, the 3D structure of the layer (content of other groups, thickness, and flexibility) cannot be ignored. The protein resistance of a PEGylated surface is often explained by the steric stabilization force, which is a general phenomenon for neutral, hydrophilic polymers in water (52) Lee, J. H.; Lee, H. B.; Andrade, J. D. Prog. Polym. Sci. 1995, 20, 1043. (53) Vogler, E. A. J. Biomater. Sci. 1999, 10, 1015. (54) Kane, R. S.; Deschatelets, P.; Whitesides, G. M. Langmuir 2003, 19, 2388. (55) Ostuni, E.; Chapman, R. G.; Holmlin, R. E.; Takayama, S.; Whitesides, G. M. Langmuir 2001, 17, 5605. (56) Sethuraman, A.; Han, M.; Kane, R. S.; Belfort, G. Langmuir 2004, 20, 7779. (57) Unsworth, L. D.; Sheardown, H.; Brash, J. L. Langmuir 2005, 21, 1036.

Langmuir, Vol. 23, No. 14, 2007 7829

and the chain mobility effect.21,52 When a solute molecule (e.g., a protein) approaches a PEGylated surface in water, the available volume for a solute to approach the surface is reduced by the large excluded volume of the PEG chains. Consequently, a repulsive force is generated by the loss of conformational entropy of the PEG chains. From this point of view, the differences in protein resistance, water permeability, and rejection between PEGMA- and SPE-based hydrogel layers would be supported by the results of the swelling experiments. The excluded volume and, consequently, the repulsion force should be larger for grafted polyPEGMA than for grafted polySPE (at the same mass-based surface coverage of PES). Finally, the results indicate that the 3D structure of the photografted polymer hydrogel layers with suited functional groups is of main importance for the lowfouling properties of the composite membranes.

Conclusions Highly protein-resistant thin-layer hydrogel composite membranes have been prepared by photograft copolymerization of neutral hydrophilic monomers (PEGMA and SPE) onto PES UF membranes. The conditions were chosen so that no pore degradation occurred, and the grafting efficiency was influenced by UV irradiation time, monomer concentration, and monomer characteristics. The addition of small amounts of cross-linker monomer (MBAA) could either increase (for PEGMA) or decrease (for SPE) the DG relative to the uncross-linked modification. The modification changed the membrane characteristics with respect to the surface chemistry, surface morphology, and pore structure. All modified membranes beyond a critical degree of grafting were more hydrophilic and less negatively charged than unmodified PES UF membranes, and pore narrowing and partial pore blocking were identified. The surface chemistry as well as surface hydrophilicity imparted by the grafted polymer on the PES membrane surface did not change after incubation in a chemical cleaning solution (sodium hypochlorite) for 8 days. The characteristics of the composite membranes would be influenced by the properties of the base membrane itself (polymer and pore structure), the properties of the grafted polymer (hydrophilicity and charge), and the grafted chain structure (incomplete vs complete surface coverage, mushroom vs brushlike grafted chain conformation, layer swelling, and thickness). Importantly, the grafted polymer hydrogel layers had significant influences on the flux and selectivity (for PEGs and for protein) of the composite membrane, and the functional monomer structure (hydrophilicity) and chemical cross-linking (MBAA content) have been identified as important parameters. All composite membranes showed much higher adsorptive fouling resistance than the unmodified PES UF membrane with a similar cutoff (10 kg/mol) and similar water permeability. Consequently, much higher permeate fluxes at similar solute rejections were obtained for ultrafiltrations of solutions of a model protein and of humic acid. Overall, composite membranes prepared with an old generation nonfouling material, PEGMA, showed better performance than membranes prepared with a new generation material, the zwitterionic SPE. The differences between PEGMA- and SPE-based composite membranes in terms of separation characteristics (flux and rejection) and of antifouling performance could be well explained by the difference in the degree of swelling of the grafted layers and the related excluded volume of the grafted chains. Acknowledgment. H.S. is indebted to the DAAD, Germany, for his Ph.D. scholarship. We thank Dieter Jacobi (Technische Chemie, Universita¨t Duisburg Essen) for the GPC analysis and

7830 Langmuir, Vol. 23, No. 14, 2007

Tim Leimer (Physikalische Chemie, Universita¨t Duisburg Essen) for her contribution to the ATR-IR analysis. We also thank Sartorius AG and Raschig GmbH (both in Germany) for supplying the membranes and the SPE monomer, respectively. Supporting Information Available: Elemental analysis, contact angles, SEM micrograph of a membrane photografted

Susanto and Ulbricht from a solution containing only cross-linker monomer (MBAA), and data demonstrating the effect of modification on membrane water permeability and membrane-solute interactions (adsorptive fouling). This material is available free of charge via the Internet at http://pubs.acs.org. LA700579X

Photografted Thin Polymer Hydrogel Layers on PES Ultrafiltration

Recommend Documents