Fouling resistant, high flux nanofiltration membranes from polyacrylonitrile-graft-poly(ethylene oxide) (original) (raw)

Provided for non-commercial research and education use. Not for reproduction, distribution or commercial use.

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues.

Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited.

In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier’s archiving and manuscript policies are encouraged to visit:
http://www.elsevier.com/copyright

Found of Membrane Science

Final of

Founding resistant, high flux nanofiltration membranes from polyacrylonitrile-graft-poly(ethylene oxide)

Ayse Asatekin a,∗{ }^{\mathrm{a}, *}, Elsa A. Olivetti b { }^{\text {b }}, Anne M. Mayes b { }^{\text {b }}
a{ }^{a} Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139, United States
b{ }^{\mathrm{b}} Department of Materials Science and Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139, United States

A R T I C L E I N F O

Article history:
Received 16 September 2008
Received in revised form 20 January 2009
Accepted 21 January 2009
Available online 31 January 2009

Keywords:
Nanofiltration
Membrane
Fouling
Comb copolymer
Self-assembly

A B STR A C T

Membranes possessing both fouling resistance and permeate size cut-offs in the nanometer range would open membrane processes to numerous molecular separations in the biochemical, pharmaceutical and food industries. This study describes the preparation of novel thin film composite (TFC) nanofiltration (NF) membranes with the above properties, where the selective layer is a dense coating of the amphiphilic comb copolymer polyacrylonitrile-graft-poly(ethylene oxide) (PAN-g-PEO). PAN-g-PEO is synthesized by free-radical polymerization using a macromonomer method, and solution coated onto a PAN ultrafiltration (UF) membrane support. Upon coagulation of the coating, microphase separation occurs between the hydrophobic PAN backbone and hydrophilic PEO side chains, as determined by differential scanning calorimetry (DSC). Transmission electron microscopy (TEM) reveals the formation of an interconnected network of PEO domains that act as water-permeable nanochannels and provide the size-based separation capability of the membrane. In filtration studies, the pure water permeability of PAN-g-PEO TFC NF membranes was found to be 85±25 L/m2 h85 \pm 25 \mathrm{~L} / \mathrm{m}^{2} \mathrm{~h} MPa, over 4 times that of a commercial NF membrane control. The permeate size cut-off was determined by the filtration of a series of rigid dye molecules to be less than 1 nm , and the fractionation of a mixture of two anionic dyes, Congo Red and Ethyl Orange, was demonstrated. Further, the membrane showed complete resistance to fouling, defined as flux loss that cannot be recovered by a water rinse, by 1 g/L1 \mathrm{~g} / \mathrm{L} bovine serum albumin (BSA) in a 24hour dead-end filtration experiment, demonstrating the promise of these membranes for biomolecule separations.

Published by Elsevier B.V.

1. Introduction

Nanofiltration (NF) membranes have properties between reverse osmosis (RO) and ultrafiltration (UF). They combine the advantages of relatively high flux and low operational pressure with size cut-off on the molecular scale, in the 0.5−2 nm0.5-2 \mathrm{~nm} range [1]. Commercial NF membranes are generally thin film composite (TFC) membranes consisting of a negatively charged aromatic polyamide interfacially polymerized onto a polysulfone UF membrane support [2,3]. The charged nature of this interfacial layer leaves NF membranes susceptible to fouling by solvated or suspended charged species in feed solutions [4]. Furthermore, the separation obtained by this method is strongly influenced by electrostatics due to this surface charge [4,5][4,5].

[1]Uncharged NF membranes could potentially operate as molecular filters and perform true size-based separations of molecules [6-15]. This could greatly expand the applications of membrane processes in the pharmaceutical, biochemical and food industries [1,8], e.g., for the fractionation or purification of peptides [4], proteins, saccharides [16,17] and fatty acids. Because they involve no phase changes, membrane separation processes have the potential to consume less energy than fractionations performed by crystallization or distillation methods, and with higher throughputs than chromatography.

Martin and coworkers first reported the separation of small molecules [11] and globular proteins [13] using track-etched polycarbonate membranes modified by the electroless deposition of gold to reduce pore diameters to molecular dimensions. Czaplewski et al. [6] demonstrated the separation of charged molecules differing in size by ∼1 nm\sim 1 \mathrm{~nm} using molecular “squares” coated onto a porous substrate to form TFC membranes. Gin and coworkers [9,10,12] more recently used polymerized lyotropic liquid crystal assemblies as the selective layer to manufacture NF membranes with the ability to fractionate charged small molecule probes and neutral polyethylene oxide (PEO) oligomers by size. Yang et al. [14]

1. a Corresponding author at: Department of Chemical Engineering, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Room 66-419, Cambridge, MA 02139, United States. Tel.: +1 617821 9946.
   E-mail address: aysea@mit.edu (A. Asatekin). ↩︎

prepared TFC membranes with a block copolymer selective layer in which oriented cylinder domains were etched to obtain pores ∼15 nm\sim 15 \mathrm{~nm} in diameter. Filtration studies showed complete retention of human rhinovirus but passage of bovine serum albumin.

These proof-of-concept studies demonstrate the promise of membrane processes for molecular-level size-based separations. To be commercially viable, however, new materials systems are needed that (1) can be synthesized inexpensively in large quantities; (2) are readily fabricated into large-area membranes; (3) exhibit permeabilities comparable to or above current NF membranes; and (4) exhibit low susceptibility to fouling by biomolecules.

To address these requirements, our research has focused on the development of TFC NF membranes employing amphiphilic comb-shaped graft copolymers as the selective layer [18-21]. In those studies, poly(vinylidene fluoride)-graft-poly(oxyethylene) methacrylate (PVDF- gg-POEM), a copolymer with a hydrophobic backbone of poly(vinylidene fluoride) (PVDF) and hydrophilic PEO side chains, was solution coated onto a UF membrane support, then immersed in a non-solvent. During precipitation, the copolymer undergoes microphase separation, forming interpenetrating networks of PVDF- and PEO-rich nanodomains. The hydrophilic domains allow for water passage, acting as high flux “nanochannels” that enable molecular separations in the 1 nm size range [18-20]. Further, the PEO “brush” layer that forms on the membrane surface and throughout the nanochannels acts as a steric barrier to biomolecule adsorption, endowing these membranes with exceptional fouling resistance [19,21]. The short brush structure appears to have a significant role in preventing irreversible fouling, as a recent study found that membranes made entirely of PEO gel still showed less than complete flux recovery after fouling [22].

The synthesis of PVDF- gg-POEM follows atom transfer radical polymerization (ATRP) protocols, using commercial PVDF as a macroinitiator, and copper (I) chloride and a ligand as the cocatalysts [23-26]. Due to the grafting-from approach, the POEM content of the copolymer produced is strongly dependent on reaction time [26] as well as factors that influence the reaction rate, including the age of reactants. Furthermore, the purification of ATRP-synthesized polymers still poses a challenge, due to the presence of copper complexes as the catalyst. The removal of these complexes is generally done using adsorptive column treatments, which are not easily scalable [27]. An amphiphilic comb copolymer with separation and anti-fouling properties similar to PVDF- gg POEM, but obtained by a more readily scalable synthesis route, would be beneficial in the commercialization of such membranes.

The present article describes the use of polyacrylonitrile-graftpoly(ethylene oxide), PAN-g-PEO, as an alternative selective layer component to PVDF- gg-POEM for TFC NF membranes. The synthesis of PAN-g-PEO is performed using traditional free-radical polymerization, employing macromonomers [28]. This method is simpler than ATRP, readily scaled up, and requires minimal purification of the product. In previous studies, PAN-g-PEO was shown to impart exceptional fouling resistance to UF membranes when used as a surface-segregating additive [28,29]. Here we show that PAN-gPEO can also serve as an effective molecular filter when prepared as the selective layer of TFC NF membranes, allowing size-based separation with subnanometer resolution. The molecular sieving properties are attributed to the microphase separation of PAN-gPEO, which is demonstrated by transmission electron microscopy (TEM) and thermal analysis. The PAN-g-PEO NF membranes are further shown to exhibit pure water permeabilities (PWP) above those of commercial NF membranes, and complete resistance to irreversible fouling by a model protein, defined as the flux loss after physical cleaning by a water rinse. This indicates that the membranes resist adsorptive fouling completely, and suggests their promise in biomolecule separations.

2. Experimental

2.1. Materials

Acrylonitrile, poly(ethylene glycol) methyl ether acrylate (PEGA, Mn∼454 g/mol,8−9\mathrm{M}_{\mathrm{n}} \sim 454 \mathrm{~g} / \mathrm{mol}, 8-9 EO repeat units/monomer), azobisisobutyronitrile (AIBN), Brilliant Blue R, Acid Blue 45, Ethyl Orange and Congo Red were purchased from Sigma-Aldrich (St. Louis, MO). Dimethyl formamide (DMF), toluene, deuterated dimethyl sulfoxide (DMSO- d6\mathrm{d}_{6} ), ethanol, isopropanol, bovine serum albumin (BSA, 66.5kDa)66.5 \mathrm{kDa}), phosphate buffered saline (PBS, pH 7.4 ), and hexane were purchased from VWR (West Chester, PA). All chemicals and solvents were reagent grade, and were used as received. PAN-400 ultrafiltration membranes, purchased from Sepro Membranes, Inc. (Oceanside, CA), were used as the base membrane for the coating process, as well as a control. Deionized water was produced from a Millipore Milli-Q unit.

2.2. Synthesis of PAN-g-PEO

Polyacrylonitrile-graft-poly(ethylene oxide) (PAN-g-PEO) was synthesized by free-radical polymerization. Five grams each of acrylonitrile and PEGA were charged to a round-bottomed flask. Toluene ( 25 mL)25 \mathrm{~mL}) and AIBN (0.01 g)(0.01 \mathrm{~g}) were added. The flask was sealed with a rubber septum, purged with nitrogen gas for 20 min , and placed in an oil bath at 90∘C90^{\circ} \mathrm{C} with stirring. After approximately 20 min , precipitated polymer was observed in the flask. After 20 h the flask was removed from the oil bath. The reaction mixture was diluted with approximately 50 mL of DMF to dissolve the precipitated polymer, and precipitated in a 3:1 mixture of ethanol and hexane. The recovered product was filtered and remaining solvent and monomer were extracted by stirring the polymer in an ethanol bath overnight. The polymer was then dried in a vacuum oven overnight. The yield was 72%72 \%.

PEO content of the copolymer was determined by 1H{ }^{1} \mathrm{H} Nuclear Magnetic Resonance (NMR) spectroscopy in DMSO- d6\mathrm{d}_{6} using a Bruker DPX 400 spectrometer to obtain the ratio of the total backbone protons ( 1.5−2.5ppm1.5-2.5 \mathrm{ppm} ) to the COOCH2\mathrm{COOCH}_{2} protons of PEGA (4−4.5ppm)(4-4.5 \mathrm{ppm}). The polymer was determined to contain 53wt%53 \mathrm{wt} \% PEGA, equivalent to 45wt%45 \mathrm{wt} \% PEO. The molecular weight distribution of the polymer was determined by gel permeation chromatography in DMF. The number-average molecular weight was measured to be 93 kg/mol93 \mathrm{~kg} / \mathrm{mol} based on polystyrene standards. The high polydispersity index (PDI of 13) is attributed to the precipitation of the copolymer during the synthesis.

2.3. Characterization of polymer phase separation

Modulated differential scanning calorimetry (MDSC) was performed with a Q100 TA Instruments differential scanning calorimeter (DSC). The sample was prepared by placing approximately 11.5 mg of PAN-g-PEO in a hermetic aluminum pan. A regular heat-cool-heat cycle was run before the MDSC to equilibrate the polymer. The scan rate for MDSC was 2∘C/min2^{\circ} \mathrm{C} / \mathrm{min}, with modulations of 1.25∘C1.25^{\circ} \mathrm{C} every 60 s . Reversible heat flow rate was calculated from this data using TA Universal Analysis software to eliminate kinetic effects on the curve and isolate glass transition temperatures.

Microstructural characterization was carried out using a JEOL 2010 CX transmission electron microscope (TEM) in bright field mode at 200 keV . Films of PAN-g-PEO cast from a 20%20 \% (w/v) DMF solution were mounted onto posts with epoxy and ultrathin ∼15 nm\sim 15 \mathrm{~nm} sections were cryomicrotomed using a diamond knife (Diatome) and placed on copper grids ( 400 mesh, Ted Pella). PEO domains were preferentially stained with RuO4\mathrm{RuO}_{4} to enhance contrast. The samples were stained by placing grids onto a glass slide and load-

ing them, for 1 h , into a chamber containing RuO4\mathrm{RuO}_{4}-vapor freshly added from ampoules (EMS Acquisition Corp.). The grids were then coated with 10 nm of carbon through thermal evaporation. Fast Fourier Transform (FFT) analysis was performed on the obtained TEM images using ImageJ image analysis software [30].

2.4. Preparation of coated membranes

The polymer solution was prepared by dissolving 1 g of PAN-gPEO copolymer in 4 mL DMF at approximately 50∘C50^{\circ} \mathrm{C}. This solution was passed through a 1−μm1-\mu \mathrm{m} syringe filter (Whatman) and degassed in an oven at approximately 90∘C90^{\circ} \mathrm{C} for approximately 1 h , until no air bubbles could be seen. Membranes were coated using a control coater (Testing Machines Inc., Ronkonkoma, NY). The PAN-400 base membrane was fixed onto the coater, and the coating blade, adjusted to a nominal film thickness of 40μ m40 \mu \mathrm{~m}, was inserted. The coating solution was poured onto the base membrane to form a thin line about 0.5 cm from the blade, and the coater was used to move the blade at a constant reproducible speed (speed level 4 on the instrument). After 5 min , the membrane was immersed in a bath of isopropanol for 30 min , and then moved into a water bath. The membranes were kept wet at all times due to the poor drying resistance of PAN [31].

Membrane morphology was characterized using a JEOL 5910 scanning electron microscope (SEM) operating at 5 kV . The membranes were fractured in liquid nitrogen while still hydrated for cross-sectional observation, and allowed to dry. The sample was sputter coated with gold-palladium for SEM imaging.

2.5. Filtration experiments

Circular pieces were cut from coated membranes, keeping them moist at all times. Experiments to determine the permeate size cut-off of the membrane were performed on 43 mm diameter membranes using an Amicon 8050 stirred, dead-end filtration cell (Millipore) with a cell volume of 50 mL and an effective filtration area of 13.4 cm213.4 \mathrm{~cm}^{2}. The cell was stirred at 500 rpm using a stir plate to minimize concentration polarization, and a pressure of 30 psi ( 0.21 MPa ) was used. The membrane was first allowed to stabilize by passing DI water through for 1−3 h1-3 \mathrm{~h}, until the flux stabilized to a constant value retained for over half an hour. The pure water permeability (PWP) of the membrane was then measured by collecting and weighing the permeate over a 5 min interval. The PWP values reported were averaged over five experiments performed on separate membrane samples. The cell was next filled with a 100mg/L100 \mathrm{mg} / \mathrm{L} solution of a dye of known size. The pressure was set to 30 psi ( 0.21 MPa ), and a ∼2 mL\sim 2 \mathrm{~mL} sample was collected once the color of the filtrate was stable, after a minimum equilibration period of an hour. During the filtration, dye solution was added as necessary to maintain the solution level in the cell above 60%60 \% of the initial level and avoid a gradual increase in feed concentration. The flux through the membrane did not change appreciably from the pure water flux during the filtration of the dye solution. Upon the collection of the sample, the cell was rinsed with DI water, and DI water was filtered through the membrane for at least an hour until the filtrate was clear. The above procedure was repeated for each dye. In addition, a diafiltration experiment
was run, using the procedure described above, with a feed containing 100mg/L100 \mathrm{mg} / \mathrm{L} each of Ethyl Orange and Congo Red dyes. All of the reported dye filtration experiments were performed on the same membrane sample. The order of dye filtration did not affect dye retention, which indicated that the selectivity of the membrane was not altered during these experiments. Comparable results were obtained from other samples cut from the same casting.

Dye permeation values were obtained by measuring the dye concentration in this sample by UV-visible spectroscopy using a Cary 500i UV-Vis-NIR dual-beam spectrophotometer. Dye solutions with concentrations above 20mg/L20 \mathrm{mg} / \mathrm{L} were diluted to one-fifth of their initial concentration, to ensure a linear concentration-absorbance relationship and accurate calculation of concentration. The percentage of each dye passing through the membrane was defined as the dye concentration in the filtrate divided by that in the retentate. The molecular size of each dye was calculated using Molecular Modeling software by ChemSW (Fairfield, CA). The molecular model of the dye was uploaded, and the molecular volume was calculated using the software. Then the molecule was assumed to be a sphere of equivalent volume, and its diameter was reported as the size. The characteristics of the dyes used in this study are given in Table 1.

The amount of each dye adsorbed by the PAN base and the PAN-g-PEO TFC NF membranes was determined by immersing a membrane swatch 2.85 cm22.85 \mathrm{~cm}^{2} in area into 10 mL of 100mg/L100 \mathrm{mg} / \mathrm{L} dye solution for a 2-day period. This area-to-volume ratio matched that used in the filtration experiments closely. The initial and final dye concentrations were compared by UV-visible spectrometry. The amount of dye adsorbed was below 3%3 \% for all cases except Brilliant Blue R, which was adsorbed 8.6±0.4%8.6 \pm 0.4 \% and 7.1±1.9%7.1 \pm 1.9 \% by the base UF and PAN-g-PEO TFC NF membranes respectively. Importantly, there was no statistical difference between the amount of dye adsorbed by the base and coated membranes. These results suggest that the observed NF membrane selectivity was not a result of dye adsorption.

Fouling experiments were performed on 25 mm diameter membranes using an Amicon 8010 stirred, dead-end filtration cell (Millipore) with a cell volume of 10 mL and an effective filtration area of 4.1 cm24.1 \mathrm{~cm}^{2}, attached to a 3.5 L dispensing vessel. Permeate was collected at fixed time intervals using a FRAC-100 fraction collector (Pharmacia) and weighed to determine trans-membrane flux.

Filtration cells were stirred at 500 rpm using a stir plate to minimize concentration polarization. Deionized (DI) water was first passed through the membrane until the flux remained stable over at least a half hour ( 2 h minimum DI water filtration). The end of the stabilization period was taken to be the zero time point in the filtration plots. The cell was then emptied and refilled with the model foulant solution. Protein solutions comprised 1000mg/L1000 \mathrm{mg} / \mathrm{L} BSA in PBS. A sample of permeate was collected after 1 h of filtration. BSA retention values were obtained by measuring the foulant concentration in this sample by UV-visible spectroscopy using a Cary 500i UV-Vis-NIR dual-beam spectrophotometer. The BSA concentrations were quantified using UV absorbance at 280 nm . After each fouling test, the filtration cell was rinsed 5−75-7 times with DI water and then refilled with DI water as a feed to determine the reversibility of fouling.

Table 1
Characteristics of molecular dyes.

3. Results and discussion

3.1. Synthesis and characterization of PAN-g-PEO

For a graft/comb copolymer to be used in the manufacture of NF membranes described in this study, it must satisfy one important condition: the backbone and side chains of the copolymer must undergo microphase separation, forming a bicontinuous network structure. The PEO brush-lined nanochannels of this structure provide the membrane with its size-based molecular sieving properties [18−20][18-20].

To verify that microphase separation occurs in PAN-g-PEO, MDSC was performed. Fig. 1 shows the reversible heat flow versus temperature plot for a sample of PAN-g-PEO obtained by MDSC analysis. The reversible heat flow was chosen to minimize the observation of kinetic effects, which are especially significant in systems with very small-scale phase separation. Three transitions were observed. A large glass transition was observed at approximately −49∘C-49^{\circ} \mathrm{C}. This value is consistent with the glass transition temperature of low molecular weight, non-crystalline PEO [32], and indicates the presence of a separate PEO phase. There is a second transition observed at approximately 70∘C70^{\circ} \mathrm{C}. This value corresponds to the second amorphous relaxation observed for PAN cast from DMF [33], indicating the presence of a separate PAN phase. Finally, a third transition is observed at 153∘C153^{\circ} \mathrm{C}, which corresponds to backbone rotation in PAN [33]. The presence of separate transition temperatures for PEO and PAN verify that microphase separation occurs in PAN-g-PEO, and suggests the promise of this copolymer for the preparation of TFC NF membranes.

To determine the microphase-separated morphology, TEM was performed. To obtain contrast, PEO domains were stained preferentially with RuO4\mathrm{RuO}_{4}. Fig. 2 presents a typical micrograph negative in which the PEO domains appear as light regions. Both PAN (dark) and PEO domains were observed to form continuous networks, as needed for this application. The PEO-rich nanochannels appear to be around 1 nm in width-comparable to those observed for PVDF- gg-POEM [19,24], which showed selectivity in the NF range. The domain periodicity was estimated from the inset FFT image, which shows a scattering peak (light circle) corresponding to a characteristic period of the PAN-g-PEO morphology of ∼1.4 nm\sim 1.4 \mathrm{~nm}. Quantitative analysis by TEM is limited by the fact that it is performed under high vacuum, and thus does not account for the effects of water absorption on channel width. Moreover, staining procedures and electron beam damage are known to influence TEM estimates of copolymer domain widths [34]. Hence the TEM results are presented here primarily to confirm the formation of a bicon-

Fig. 1. MDSC thermogram of PAN-g-PEO.

Fig. 2. TEM bright-field negative of PAN-g-PEO microphase-separated morphology, showing interconnected nanodomains of PAN (dark) and PEO (light), and FFT of the image (inset).
tinuous microphase-separated morphology on a molecular length scale.

3.2. Membrane characterization

The membranes were prepared by coating a commercial PAN UF base membrane by a thin layer of copolymer solution (PAN-gPEO in DMF) using a control coater, followed by immersion into a non-solvent to precipitate out the copolymer. The non-solvent, isopropanol, was selected due to its low diffusivity with DMF, to prevent the formation of a porous coating [3]. Cross-sectional SEM micrographs of the base UF membrane and the coated thin film composite (TFC) NF membrane are shown in Fig. 3. The coating layer was observed to be approximately 2μ m2 \mu \mathrm{~m} in thickness, and had no visible pores at the magnification used.

PAN-g-PEO TFC NF membranes also showed high pure water permeabilities (PWPs). The average PWP of PAN-g-PEO TFC NF membranes was 85±25 L/m2 hMPa85 \pm 25 \mathrm{~L} / \mathrm{m}^{2} \mathrm{~h} \mathrm{MPa}, averaged over five samples. This value was compared with the measured PWPs of two commercial NF membranes by GE Osmonics, selected for exhibiting the highest nominal flux among the NF membranes offered by the manufacturer. PAN-g-PEO TFC NF membranes exhibited an average PWP over 4 times that measured for GE Osmonics DS-5-DL ( 19±3 L/m2 hMPa19 \pm 3 \mathrm{~L} / \mathrm{m}^{2} \mathrm{~h} \mathrm{MPa}, averaged over five samples), and approximately 16 times that measured for Osmonics DS-5-HL ( 5.1 L/m2 hMPa5.1 \mathrm{~L} / \mathrm{m}^{2} \mathrm{~h} \mathrm{MPa}, only one test performed). The average PWP of PAN-g-PEO TFC NF membranes was also twice that of the TFC NF membranes based on PVDF- gg-POEM graft copolymers investigated in our previous work (39±19 L/m2 hMPa)\left(39 \pm 19 \mathrm{~L} / \mathrm{m}^{2} \mathrm{~h} \mathrm{MPa}\right) [21]. Optimization of the coating process should allow for membranes with thinner selective layer coatings and even higher pure water permeabilities.

3.3. Subnanometer size selectivity

Commercial NF membranes generally have a negative surface charge, resulting in separation characteristics that are based on a combination of sieving and electrostatics. TFC NF membranes fabricated with PAN-g-PEO copolymer as the selective layer, by contrast, are expected to operate more purely as size-based molecular filters, due to the neutral nature of PEO. Our previous studies on similar

Fig. 3. Cross-sectional SEM micrographs of (a) uncoated Sepro PAN-400 base membrane and (b) PAN-g-PEO TFC NF membrane.
membranes employing PVDF- gg-POEM as a selective layer demonstrated the ability to fractionate two dyes of like charge based on size, irrespective of whether the charge was positive or negative [19]. Sharp size cut-offs were further shown in filtrations of Au nanoparticles [18].

To determine the permeate size cut-off of the PAN-g-PEO TFC NF membranes prepared in this study, rigid dyes were used as probes. Characteristics of the dyes that were used are listed in Table 1. The percentage of each dye that passed through the PAN-g-PEO TFC NF membrane is shown in Fig. 4. It was observed that the two larger dye molecules, Brilliant Blue R ( 11.1A˚11.1 \AA ) and Congo Red ( 10.1A˚10.1 \AA ), were retained completely by the membrane. In contrast, 81%81 \% of Ethyl Orange, a molecule 8.2A˚8.2 \AA in size, permeated through the membrane, suggesting that the nanochannel width was above 8A˚8 \AA. When the hydration sphere around the dye molecule, which was not considered in the molecular size calculations, is taken into account, the effective pore size of the membrane is estimated to be around 1 nm . This value is consistent with the width and periodicity of the microphase-separated morphology observed by TEM in Fig. 2. The passage of 81%81 \% Ethyl Orange reflects the restricted mobility of the dye (vs. water molecules) through the PAN-g-PEO coating. This might be attributed to the dynamic nature of the PEO chain dimensions in the brush layer lining the nanochannels, which presents a steric barrier to the passage of molecules of size comparable to the channel width.

The relatively sharp size cut-off derived from the microphaseseparated morphology can be used to fractionate two molecules of like charge and similar dimension. As a demonstration of the diafiltration ability of the PAN-g-PEO TFC NF membrane, a solution containing 100mg/L100 \mathrm{mg} / \mathrm{L} each of Congo Red and Ethyl Orange, both

Fig. 4. Percentage of dyes of different sizes passing through the PAN-g-PEO TFC NF membrane, defined as the dye concentration in the filtrate divided by that in the retentate. Experiments performed at 30 psi ( 0.21 MPa ), using 100mg/L100 \mathrm{mg} / \mathrm{L} dye solution as feed.
negatively charged, was filtered. In Fig. 5, the UV-visible spectrum of the filtrate (solid line) is compared with the spectra of the two components of the feed, Ethyl Orange (dashed line) and Congo Red (dotted line). All solutions were diluted to one-fifth of their initial concentration to avoid non-linear behavior at high concentrations. It can be seen that the filtrate spectrum follows that of Ethyl Orange, at a slightly lower concentration, calculated to be 82%82 \% of the initial concentration. The characteristic peaks of Congo Red, at approximately 344 and 500 nm , are not observed in the filtrate, indicating that the dye was retained completely.

The capability of the PAN-g-PEO NF membranes to separate small molecules predominantly by size could open new applications in the biotechnology [35] and food industries [8]. Current commercial NF membranes operate on a solution-diffusion type mechanism and have rejections that are strongly dependent on charge in addition to size. This is exemplified by the comparable retentions reported for the Osmonics DS-5-DL for two negatively charged dyes of different size: Congo Red, with a molecular size of 10.1A˚10.1 \AA, was retained by 93−96%93-96 \% depending on pH ; the retention of the much smaller Methyl Orange ( 7.9A˚7.9 \AA ) similarly ranged between 90−97%90-97 \% [36]. The PAN-g-PEO TFC NF membranes thus enable molecular separations not possible with the NF membranes now on the market.

3.4. Fouling resistance

Previous research on PVDF- gg-POEM TFC NF membranes demonstrated the exceptional fouling resistance of NF systems based on amphiphilic comb copolymers with PEO side chains [19,21].

Fig. 5. UV-visible spectrum of permeate in the diafiltration experiment (solid line) compared with the spectra of Ethyl Orange (dashed) and Congo Red (dash-dot). Feed solution: 100mg/L100 \mathrm{mg} / \mathrm{L} each of Congo Red and Ethyl Orange in water. All samples were diluted to 1/51 / 5 their initial concentration to avoid high-concentration non-linearities in absorption. Experiments performed at 30 psi ( 0.21 MPa ).

Fig. 6. Dead-end filtration of model protein solution with PAN-g-PEO TFC NF membrane (filled symbols) and Sepro PAN-400 UF membrane (empty symbols). ∙.⊙\bullet . \odot Milli-Q water, ↓.⊙1000mg/L\downarrow . \odot 1000 \mathrm{mg} / \mathrm{L} BSA in PBS. Tests performed at 30 psi ( 0.21 MPa ) and 10 psi ( 0.07 MPa ) for the NF and UF membranes, respectively.

These membranes were shown to completely resist irreversible fouling due to the adsorption of organic molecules by oily water mixtures [19], protein, humic acid and polysaccharide solutions, and activated sludge from a membrane bioreactor [21]. The similar structure and nanochannel chemistry of PAN-g-PEO TFC NF membranes might be expected to yield equivalently high fouling resistance.

Fig. 6 shows the 24-hour dead-end filtration results from a PAN-gg-PEO NF membrane (filled symbols) for a 1000mg/L1000 \mathrm{mg} / \mathrm{L} solution of BSA in PBS, performed at 30 psi ( 0.21 MPa ), and plotted as a function of normalized flux (flux/pure water flux) vs. time. Milli-Q water was passed through the membrane until the flux stabilized. The flux decline during this initial period is due to membrane compaction. With the application of pressure the base membrane is compressed, resulting in partial blocking of pores. This effect is less notable for the NF membranes because the flux is largely determined by the non-porous coating layer, which does not deform substantially under pressure. There is only a small decline in flux over the course of the filtration of the protein solution ( ∼15%\sim 15 \% after 24 h ). This loss is fully recovered when the cell is rinsed and the foulant solution is replaced with deionized water, indicating complete resistance to adsorptive fouling by this protein. This small flux loss, as well as its complete reversibility, makes these membranes promising for feeds with large fouling potential. BSA was completely retained by the PAN-g-PEO membrane. This is consistent with the reported globular dimensions of this protein (a heart-shaped molecule with 8 nm sides and 3 nm width [37]), relative to the pore size data obtained using the dye filtrations and the size scale of microphase separation observed by TEM.

As a control, a similar filtration experiment was conducted using the Sepro PAN-400 base UF membrane (PWP 4580 L/m2 h4580 \mathrm{~L} / \mathrm{m}^{2} \mathrm{~h} MPa), at an operating pressure of 10 psi ( 0.07 MPa ); the results are also shown in Fig. 6 (open symbols). The base UF membrane lost 81%81 \% of its flux irreversibly during the same 24 h time period. BSA retention for the UF base membrane after 1 h filtration was 73%73 \%.

4. Conclusion

In this study, novel NF TFC membranes were developed based on the microphase separation of the backbone and side chains of a PAN-g-PEO comb copolymer. These membranes were shown to exhibit high pure water permeability, an ability to fractionate small molecules by size, and complete resistance to fouling by a BSA solution. This membrane system appears promising for separations in the pharmaceutical, biochemical and food industries, where the use of membrane processes for molecular fractionations is currently limited by severe fouling due to the high organic content of the
feed, and the limited size selectivity of commercial NF membranes. They also show potential for wastewater treatment applications, for example in membrane bioreactors, due to their fouling resistance and high permeability coupled with their ability to retain smaller contaminants [21,38,39]. Furthermore, the scalable synthesis of PAN-g-PEO, as well as the simplicity of the TFC manufacturing process, offer easy scale-up of their production.

One potentially significant drawback of the macromonomer chemistry employed in this study is the ester bond linkage of the PEO side chain to the comb polymer backbone. This functional group is susceptible to hydrolysis in the presence of acidic or basic media, which might limit the applications suitable for these membranes to ones that occur near neutral pH . In preliminary studies, we found that UF membranes containing PAN-g-PEO were stable between pH 5 and 8 , but exhibited degradation with more acidic or basic conditions [40]. PEO brushes have also been reported to degrade and lose their resistance to cell attachment when exposed to cells over a time scale measured by days [41,42]. A study addressing the long-term stability of the PAN-g-PEO TFC NF membranes under various feed conditions would serve to better delineate the types of applications for which these novel membranes might be suitable.

Acknowledgements

Financial support for this work was provided by the WaterCAMPWS, a Science and Technology Center of Advanced Materials for the Purification of Water with Systems under the National Science Foundation agreement number CTS-0120978, and by the U.S. Office of Naval Research under Award N00014-02-1-0343. This work made use of MRSEC Shared Experimental Facilities supported by the National Science Foundation under Award DMR-0213282. AA thanks Prof. Michael F. Rubner for helpful discussions.

References

[1] N. Hilal, H. Al-Zoubi, N.A. Darwish, A.W. Mohammad, M. Abu Arabi, A comprehensive review of nanofiltration membranes: treatment, pretreatment, modelling, and atomic force microscopy, Desalination 170 (2004) 281-308.
[2] R.J. Petersen, Composite reverse osmosis and nanofiltration membranes, Journal of Membrane Science 83 (1993) 81-150.
[3] R.W. Baker, Membrane Technology and Applications, John Wiley, Chichester, New York, 2004.
[4] J.F. Lapointe, S.F. Gauthier, Y. Pouliot, C. Bouchard, Fouling of a nanofiltration membrane by a beta-lactoglobulin tryptic hydrolysate: impact on the membrane sieving and electrostatic properties, Journal of Membrane Science 253 (2005) 89-102.
[5] Y.K. Shim, S. Chellam, Steric and electrostatic interactions govern nanofiltration of amino acids, Biotechnology and Bioengineering 98 (2007) 451-461.
[6] K.F. Czaplewski, J.T. Hupp, R.Q. Sourr, Molecular squares as molecular sieves: size-selective transport through porous-membrane-supported thin-film materials, Advanced Materials 13 (2001) 1895-1897.
[7] J.Y. Han, J.P. Fu, R.B. Schoch, Molecular sieving using nanofilters: past, present and future, Lab on a Chip 8 (2008) 23-33.
[8] A. Kalbasi, L. Cisneros-Zevallos, Fractionation of monomeric and polymeric anthocyanins from concord grape (Vitis labruscu L.) juice by membrane ultrafiltration, Journal of Agricultural and Food Chemistry 55 (2007) 7036-7042.
[9] D.L. Gin, J.E. Rara, R.D. Noble, B.J. Elliott, Polymerized lyotropic liquid crystal assemblies for membrane applications, Macromolecular Rapid Communications 29 (2008) 367-389.
[10] M. Zhou, T.J. Kidd, R.D. Noble, D.L. Gin, Supported lyotropic liquid-crystal polymer membranes: promising materials for molecular-size-selective aqueous nanofiltration, Advanced Materials 17 (2005) 1850-1853.
[11] K.R. Jirage, J.C. Holteen, C.R. Martin, Nanotubule-based molecular-filtration membranes, Science 278 (1997) 655-658.
[12] D.L. Gin, W.Q. Gu, B.A. Pindzola, W.J. Zhou, Polymerized lyotropic liquid crystal assemblies for materials applications, Acrounts of Chemical Research 34 (2001) 973−980973-980.
[13] S.F. Yu, S.B. Lee, M. Kang, C.R. Martin, Size-based protein separations in poly(ethylene glycol)-derivatized gold nanotubule membranes, Nano Letters 1 (2001) 495-498.
[14] S.Y. Yang, I. Ryu, H.Y. Kim, J.K. Kim, S.K. Jang, T.P. Russell, Nanoporous membranes with ultrahigh selectivity and flux for the filtration of viruses, Advanced Materials 18 (2006) 709-712.

[15] W.A. Phillip, J. Rzayev, M.A. Hillmyer, E.L. Cussler, Gas and water liquid transport through nanoporous block copolymer membranes, Journal of Membrane Science 286 (2006) 144-152.
[16] I. Catarino, M. Minhalma, L.L. Bela, M. Mateus, M.N. de Pinho, Assessment of saccharide fractionation by ultrafiltration and nanofiltration, Journal of Membrane Science 312 (2008) 34-40.
[17] A.K. Goulas, P.G. Kapasakalidis, H.R. Sinclair, R.A. Rastall, A.S. Grandison, Purification of oligosaccharides by nanofiltration, Journal of Membrane Science 209 (2002) 321-335.
[18] A. Akthakul, A.I. Hochbaum, F. Stellacci, A.M. Mayes, Size fractionation of metal nanoparticles by membrane filtration, Advanced Materials 17 (2005) 532-535.
[19] A. Akthakul, A.F. Salinaro, A.M. Mayes, Antifouling polymer membranes with subnanometer size selectivity, Macromolecules 37 (2004) 7663-7668.
[20] A. Asatekin, A.M. Mayes, Responsive pore size properties of composite NP membranes based on PVDF graft copolymers, Separation Science and Technology (submitted for publication).
[21] A. Asatekin, A. Meonitti, S. Kang, M. Elimelech, E. Morgenroth, A.M. Mayes, Antifouling nanofiltration membranes for membrane bioreactors from selfassembling graft copolymers, Journal of Membrane Science 285 (2006) 81-89.
[22] G.D. Kang, Y.M. Cao, H.Y. Zhao, Q. Yuan, Preparation and characterization of crosslinked poly(ethylene glycol) diacrylate membranes with excellent antifouling and solvent-resistant properties, Journal of Membrane Science 318 (2008) 227-232.
[23] J.F. Hester, P. Banerjee, A.M. Mayes, Preparation of protein-resistant surfaces on poly(vinylidene fluoride) membranes via surface segregation, Macromolecules 32 (1999) 1643-1650.
[24] J.F. Hester, P. Banerjee, Y.Y. Won, A. Akthakul, M.H. Acar, A.M. Mayes, ATRP of amphiphilic graft copolymers based on PVDF and their use as membrane additives, Macromolecules 35 (2002) 7652-7661.
[25] J.F. Hester, A.M. Mayes, Design and performance of foul-resistant poly(vinylidene fluoride) membranes prepared in a single-step by surface segregation, Journal of Membrane Science 202 (2002) 119-135.
[26] S. Inceoglu, S.C. Olugebefola, M.H. Acar, A.M. Mayes, Atom transfer radical polymerization using poly(vinylidene fluoride) as macroinitiator, Designed Monomers and Polymers 7 (2004) 181-189.
[27] K. Matyjaszewski, J.H. Xia, Atom transfer radical polymerization, Chemical Reviews 101 (2001) 2921-2990.
[28] A. Asatekin, S. Kang, M. Elimelech, A.M. Mayes, Anti-fouling ultrafiltration membranes containing polyacrylonitrile-graft-poly(ethylene oxide) as an additive, Journal of Membrane Science 298 (2007) 136-146.
[29] S. Kang, A. Asatekin, A.M. Mayes, M. Elimelech, Protein antifouling mechanisms of PAN UP membranes incorporating PAN-g-PEO, Journal of Membrane Science 296 (2007) 42-50.
[30] ImageJ software, http://rsbweb.nih.gov/ij/index.html, 2008 (accessed 14.10.08).
[31] N. Scharnagl, H. Buschatz, Polyacrylonitrile (PAN) membranes for ultra- and microfiltration, Desalination 139 (2001) 191-198.
[32] B.E. Read, Mechanical relaxation in some oxide polymers, Polymer 3 (1962) 529−542529-542.
[33] A.S. Kenyon, M.J. Rayford, Mechanical relaxation processes in polyacrylonitrile polymers and copolymers, Journal of Applied Polymer Science 23 (1979) 717−725717-725.
[34] H. Ott, V. Abetz, V. Alstrstadt, A. Pfau, Comparative study of a block copolymer morphology by transmission electron microscopy and scanning force microscopy, Journal of Microscopy-Oxford 205 (2002) 106-108.
[35] R. van Reis, A. Zydney, Membrane separations in biotechnology, Current Opinion in Biotechnology 12 (2001) 208-211.
[36] K. Boussu, C. Vandecasteele, B. Van der Bruggen, Relation between membrane characteristics and performance in nanofiltration, Journal of Membrane Science 310 (2008) 51-65.
[37] M.L. Ferrer, R. Duchowicz, B. Carrasco, J.G. de la Torre, Acuna A.U., The conformation of serum albumin in solution: a combined phosphorescence depolarization-hydrodynamic modeling study, Biophysical Journal 80 (2001) 2422−24302422-2430.
[38] J.H. Choi, S. Dockko, K. Fukushi, K. Yamamoto, A novel application of a submerged nanofiltration membrane bioreactor (NF MBR) for wastewater treatment, Desalination 146 (2002) 413-420.
[39] J.H. Choi, K. Fukushi, K. Yamamoto, Comparison of treatment efficiency of submerged nanofiltration membrane bioreactors using celulose treacetate and polyamide membrane, Water Science and Technology 51 (6-7) (2005) 305−312305-312.
[40] A. Asatekin Alexiou, Improved filtration membranes through self-organizing amphiphilic comb copolymers, Ph.D. thesis, MIT, Cambridge, MA, 2009.
[41] G.N. Whitesides, E. Oelsoi, R. Chapman, M. Grunze, X.Y. Jiang, Sams and biofunctional surfaces: the “inert surface” problem, Abstracts of Papers of the American Chemical Society 227 (2004) U354-U1354.
[42] Y.Y. Luk, M. Kato, M. Miksich, Self-assembled monolayers of alkanethiolates presenting mannitol groups are inert to protein adsorption and cell attachment, Langmuir 16 (2000) 9604-9608.