Preparation of nanofiltration membranes from polyacrylonitrile ultrafiltration membranes

Preparation of nanofiltration membranes from polyacrylonitrile ultrafiltration membranes

Journal of Membrane Science 286 (2006) 333–341 Preparation of nanofiltration membranes from polyacrylonitrile ultrafiltration membranes Jinwen Wang ∗...

947KB Sizes 0 Downloads 49 Views

Journal of Membrane Science 286 (2006) 333–341

Preparation of nanofiltration membranes from polyacrylonitrile ultrafiltration membranes Jinwen Wang ∗ , Zhongren Yue, Jeffrey Scott Ince, James Economy 1 Center of Advanced Materials for the Purification of Water with Systems, Department of Material Science and Engineering, University of Illinois at Urbana-Champaign, 1304 West Green Street, Urbana, IL, USA Received 26 February 2006; received in revised form 20 September 2006; accepted 9 October 2006 Available online 12 October 2006

Abstract Polyacrylonitrile (PAN) membranes display some unusual features for ultrafiltration (UF). The meso-macropores of PAN UF membranes can be easily reduced into the range of micro-mesopores by taking advantage of surface tension forces within the capillary pores during heat treatment in the presence of ZnCl2 . Asymmetric PAN nanofiltration (NF) membranes with controlled highly dense pore surface functional groups were prepared by hydrolysis of the nitrile groups with NaOH. The combined effects of heat treatment and the presence of ZnCl2 on the formation of nanofiltration membranes were investigated. In addition, membrane post-treatment with NaOH was studied. The effect of counterion species on the membrane performance was also investigated. A simple new method, which utilized univalent alkaline ions as probes, was developed to derive the average pore size of cationic nanofiltration PAN membranes from pure water permeability coefficients. Published by Elsevier B.V. Keywords: Polyacrylonitrile; Nanofiltration; Pore size characterization; Pure water permeability coefficients; Zinc chloride

1. Introduction Nanofiltration (NF), defined by many “as a process between ultrafiltration (UF) and reverse osmosis (RO)”, is a relatively recent technology, largely developed over the past decade. Typically, NF membranes have sodium chloride rejections between 20 and 80% and molecular weight cutoffs for dissolved organic solutes of 200–1000 Da. Their separation mechanisms frequently involve both size and Donnan exclusion effects. NF membranes have found many applications in a variety of industries. In water treatment, for example, NF membranes are promising for the treatment of both organic and inorganic pollutants. Their low-pressure operation (4–14 bar) provides increased energy savings with significantly lower installation and operating costs [1,2]. Polyacrylonitrile (PAN), because of its high degree of solvent resistance, is widely used as an UF membrane. Due to its

Corresponding author. Tel.: +1 2173332088; fax: +1 2173332736. E-mail addresses: [email protected] (J. Wang), [email protected] (J. Economy). 1 Tel.: +1 2173339260; fax: +1 2173332736. 0376-7388/$ – see front matter. Published by Elsevier B.V. doi:10.1016/j.memsci.2006.10.022

highly hydrophilic properties, it has been known as a low fouling membrane for aqueous filtration. Compared to other polymer materials, PAN also has good resistance against chlorine [3]. Usually, PAN UF membrane is prepared via the phase separation technique. Average pore size and effective porosity of PAN membranes can be controlled by the polymer concentration in the casting solution. However, it is not practical to reduce the pore size of PAN membrane into the range of micro-mesopores by phase separation technique due to poor solubility of PAN in various solvents. PAN UF membranes have been modified to improve permeation behavior or create NF membranes by various methods such as chemical conversions of the nitrile groups [4–6], photoinitiation [7,8], heterogeneous [9] or plasma [10,11] graft polymerization of acrylic monomers. Here, we demonstrate a novel technique to conveniently prepare an asymmetric cationic PAN NF membrane from PAN UF membranes. This conversion is achieved by taking advantage of the interaction of surface tension forces at the vapor–liquid interface within the capillary pores during the heat treatment of UF membranes that are saturated with a ZnCl2 aqueous solution. Continuous micro-mesopores with controlled highly dense pore surface functional carboxylic acid groups were formed by post-treatment with 1 M NaOH. We also investigated the effect


J. Wang et al. / Journal of Membrane Science 286 (2006) 333–341

of different counter ions on the permeability properties of the cationic PAN NF membranes and developed a simple method, which uses univalent alkaline ions as probes, to estimate the average pore radius of the cationic NF membranes. 2. Experimental 2.1. Materials PAN (homopolymer, Tg = 85 ◦ C, average Mw 150,000), polyvinylpyrrolidone (PVP, Mw = 29,000) and DMF were received from Aldrich. All the salts were used as received from Fisher Scientific. 2.2. Preparation of asymmetric PAN UF membrane The UF membranes were prepared by the phase separation technique using water as a coagulant. PAN was used as a membrane material and PVP as an additive to make the membrane more porous. PAN and PVP powder were dissolved at 80–90 ◦ C with stirring in DMF to form a 15:5 wt.% PAN:PVP casting solution. The solution was cast onto a Hollytex® polyester nonwoven fabric using a laboratory membrane-casting machine (Separation Systems Technology, USA). The nascent membrane was immersed in a room temperature tap water coagulation bath without evaporation of solvents in the air. After precipitation, the membrane was kept in a water bath for several days and then washed with deionized water before further experiments. 2.3. Preparation of asymmetric PAN NF membrane The PAN UF membrane was soaked in ZnCl2 aqueous solution for several days. After saturation with the ZnCl2 solution, the membrane was heat treated in air at varying temperatures and times. Following heating, the membrane was allowed to cool to room temperature in air. By soaking the membrane in a very dilute HCl aqueous solution (pH = 3–4), ZnCl2 was removed and a half-transparent PAN membrane that was two to three times thinner was formed. The membrane was then hydrolyzed with 1 M NaOH at room temperature for a controlled time. After the reaction was complete, the hydrolyzed membrane was treated with 1 M HCl at room temperature overnight. The color of the hydrolyzed yellowish red PAN membrane turned to yellowish white. Following that, the membrane was infiltrated with a dilute NaOH solution (pH = 8–9) to convert it into a NF membrane with highly dense pore surface functional groups (–COONa, –CN and –CONH2 ). The preparation process was described in Fig. 4. 2.4. Membrane performance measurement The membranes were cut into disks appropriate for use in a filtration cell (SterlitechTM HP4750 Stirred Cell). A standard magnetic stirrer (Corning Stirrer/Hot Plate, Model PC-420) was used and the stirrer speed was set to achieve a reasonable rate of stirring. Rejection was determined using a NaCl solution. The salt solution flux and salt rejection were measured at 13.79 bar

and room temperature. The feed concentration was typically 2000 mg/L in pure DI water. The permeated samples were collected for a few minutes and the concentration of permeates were determined, using a Corning pH/ion analyzer 455. Every point of membrane performance including flux and salt rejection was measured at least three times to get an average value. 2.5. Measurement of pure water permeability coefficients of NF membranes with different counterions One molar solutions of the chloride salts in Table 5 were passed through the membrane at 13.79 bar for 1 h to form different carboxylic acid salts on the membrane pore surfaces. The membrane was then washed with DI water for at least half an hour to flush the membrane and permeate flow channel. Pure water flux experiments were performed by applying pressures of 6.89, 10.34 and 13.79 bar. The purpose of this experimental technique was to track any changes to the water flux that would indicate changes in the membrane porous structure. 2.6. Elemental analysis, scanning electron microscope, attenuated total reflectance Fourier transform infrared spectroscopy (ATR/FTIR) and nitrogen adsorption A Model CE440 Elemental Analyzer was used to directly determine the C, H, and N (wt.%) in the samples. The morphology of membranes (cross-section and surface) was observed with a scanning electron microscope (Hitachi S-4700). Membranes were pretreated by the solvent-exchange method to prevent the structure from collapsing upon drying. Water in the membrane was replaced first with iso-propanol and then with n-hexane. Samples were cryogenically fractured in liquid nitrogen to form a suitable cross-section and then coated with gold. ATR/FTIR spectra were collected in the range 4000– 600 cm−1 , on a Nexus 670 FT-IR (Thermo Electron Corporation, Madison, WI) with a Golden GateTM MKII Single Reflectance ATR (Specac Inc., Woodstock, GA). The spectrometer was installed with a deuterated triglycine sulfate–potassium bromide (DTGS–KBr) detector and KBr beamsplitter. Spectra collection was performed using FT-IR software (OMNIC, Thermo Electron Corporation, Madison, WI) and analyzed using spectrum software (KnowItAll Informatics System 5.0 Academic Edition, Bio-Rad Laboratories, Inc.). Spectra were recorded by positioning the samples on a cell platform operating at room temperature (64 scans, 4 cm−1 resolution). The adsorption was carried out with an Autosorb-1 volumetric sorption analyzer controlled by Autosorb-1 for Windows 1.19 software (Quantachrome Corp.). All samples were pretreated by the solvent-exchange method and outgassed at 80 ◦ C until the test of outgas pressure rise was below 5 ␮m Hg min−1 prior to their analysis. N2 isotherm data from the appropriate relative pressure ranges were used for subsequent calculations. The surface areas were determined using the standard BET equation for N2 adsorption at 77 K with correlation coefficients R > 0.99 in all cases. The DR method was used to estimate the micropore volume from nitrogen adsorption isotherms. The volume of

J. Wang et al. / Journal of Membrane Science 286 (2006) 333–341


mesopores was calculated by subtracting the volume of micropores from the total pore volume at a relative pressure of 0.95. 3. Results and discussion 3.1. Effect of ZnCl2 In the preparation of an asymmetric PAN UF membrane, the hydrophilic nature of the nitrile groups directs their aggregation on the pore surface during the phase separation step. This aggregation produces a continuous asymmetric meso-macropore with nitrile groups dangling from the surface. Since Lewis acids are able to create coordinate bonds with strongly electronegative groups (e.g. electronegative nitrogen atoms in the nitrile groups) [12], they anchor the nitrile groups on the pore surface even at elevated temperatures. In the absence of ZnCl2 , some nitrile groups would migrate from the pore surface and imbed in the bulk of the membrane when the temperature is higher than the PAN glass transition point. This migration will result in fewer functional groups (–COONa, –CN and –CONH2 ) on the pore surface after the hydrolysis of nitrile groups. The effect of this migration on flux and rejection is one of the reasons that the salt rejection of Membrane VI is much higher and the flux lower than those of Membrane IV and VII in Table 1. In addition, the complexation of ZnCl2 causes more severe reduction of pore size upon heating. When compared with Membrane III and VII in Table 1, for example, the flux of Membrane V is reduced, i.e. the pore size is smaller. The other function of ZnCl2 is to act as a filler to prevent complete collapse of the pore upon heating. In order to investigate the effect of ZnCl2 , porous membranes, VIII–X, were prepared from UF membranes saturated with aqueous solution containing 15, 30 and 60 wt.% ZnCl2 , respectively. Each of these membranes was then heated at 110 ◦ C for 10 min. Fig. 1 depicts the hydrolysis reaction of these membranes with 1 M NaOH. At the beginning of the reaction, the fluxes of Membranes VIII and X are almost identical and both much larger than that of Membrane IX. Because the amount of filler, i.e. ZnCl2 , in Membrane X is much more than the amount in Membrane IX, the space ZnCl2 occupies is greater during the membrane collapse and the resultant pore is larger after ZnCl2 is removed. While the complexation from ZnCl2 on the pore walls in Membrane IX is Table 1 Permeation properties of PAN membranes via different treatments

stronger than in Membrane VIII because of higher ZnCl2 concentration, the pore in Membrane IX is smaller than in the latter. Also the presence of ZnCl2 can actually increase the number of nitrile groups on the pore surface and result in a higher reaction rate during the post-treatment of the membrane with NaOH. As shown in Fig. 1, the more concentrated the ZnCl2 solution used, the more rapidly the rejection ratio increased and the flux decreased. As expected, when using MgCl2 instead of ZnCl2 , the increase of the reaction rate and the sharp shrinkage of the membrane thickness and pores was not observed because of the lack of the interaction between MgCl2 and nitrile groups. 3.2. Effect of heat treatment

Membranes from different treatmentsa

Fluxb (m3 m−2 day−1 )

Rejection ratiob (%)

(I) Without any treatment (II) With NaOH (III) With heat (IV) With heat and NaOH (V) With ZnCl2 and heat (VI) With ZnCl2 , heat and NaOH (VII) With MgCl2 , heat and NaOH

24.8 17.8 2.7 1.2 2.1 0.2 10.0

0 0 0 6.0 0 34.0 0

Heat: 110 ◦ C for 10 min; NaOH: 1 M NaOH for 24 h; ZnCl2 or MgCl2 : 30 wt.% aqueous solution for 24 h. b 13.79 bar at room temperature; 2000 mg/L NaCl as the feed solution. a

Fig. 1. Effects of the concentration of ZnCl2 and reaction time with NaOH on: (a) flux and (b) rejection of PAN membranes. Square: 15 wt.% ZnCl2 , Membrane VIII; circle: 30 wt.% ZnCl2 , Membrane IX; triangle: 60 wt.% ZnCl2 , Membrane X.

Once the heating temperature is high enough (e.g. >100 ◦ C), water vapor will attempt to evacuate the meso-macropores and capillary forces begin to play their role. When the capillary force (described by the well-known Young–Laplace relationship: p = 2γ/r) is higher than the modulus of the membrane material (in the swollen state), the meso-macropores will collapse [13]. At the same time, as the amount of water is reduced, ZnCl2 precipitates to act as a template and exerts an attracting force on the membrane pore walls. While the membrane is confined in the horizontal direction by the non-woven polyester web, these forces cause the membrane to shrink sharply in the vertical


J. Wang et al. / Journal of Membrane Science 286 (2006) 333–341

Fig. 2. SEM micrographs of PAN membrane surfaces: (a) prepared by phase inversion technique and (b) prepared by heat treatment at 110 ◦ C for 10 min with 30 wt.% ZnCl2 from (a).

Fig. 3. SEM micrographs of PAN membrane cross-sections: (a) prepared by phase inversion technique and (b) prepared by heat treatment at 110 ◦ C for 10 min with 30 wt.% ZnCl2 from (a).

direction. Thus, the pore surface density of nitrile groups increases dramatically. With the complexation of ZnCl2 and the effect of capillary force, the pore size is reduced greatly from the meso-macroporosity range into the micro-mesoporosity range. In Fig. 2, pores (in the range of 10–20 nm) in the image (a) shrink after the treatment and only a few of those pores remain in the image (b). In Fig. 3, after the heat treatment, the

finger-like structure disappears and the pore size underneath the skin is greatly reduced. Results of N2 adsorption are listed in Table 2. After heat treatment, all the pore volumes including micropores and mesopores increase. The increase of mesopore volume is the most notable. This means many macropores are shrunk into mesopores. All the properties in Table 2 are based on the mass

Table 2 Surface areas and pore volumes of PAN membranes with and without heat treatment Sample

BET surface area (m2 g−1 )

Total pore volume (P/P0 = 0.95) (×10−2 cm3 g−1 )

Micropore volume (×10−2 cm3 g−1 )

Mesopore volume (×10−2 cm3 g−1 )


128.0 129.0

36.1 49.3

7.4 (20.5%) 7.6 (15.4%)

28.7 (79.5%) 41.7 (84.6%)

Sample A: PAN membrane prepared by phase inversion technique; sample B: PAN membrane prepared by heat treatment at 110 ◦ C for 10 min with 30 wt.% ZnCl2 from sample A.

J. Wang et al. / Journal of Membrane Science 286 (2006) 333–341 Table 3 Elemental analysis of PAN membranes before and after hydrolysis Sample

C (wt.%)

N (wt.%)

H (wt.%)

O (wt.%)a


65.7 62.9

26.3 19.6

5.7 5.9

2.2 11.6

Sample A: PAN membrane prepared by phase inversion technique; sample C: PAN membrane prepared by heat treatment at 110 ◦ C for 10 min with 30 wt.% ZnCl2 from sample A and hydrolyzed by 1 M NaOH for 24 h and washed by dilute HCl solution. a Calculated from the C, N, H content.

of the membrane. However, sample B is about two to three times denser than Sample A. Thus, the effect of heat treatment on the surface area or the pore volume will be enormous on the scale of volume instead of mass. After the membrane was soaked in 1 M NaOH, negative charge is introduced by the hydrolysis of nitrile groups. This also reduces the pore size further because some of the small nitrile groups on the pore surface are replaced with the relatively bulky carboxylic groups [6]. Elemental analysis of the membrane treated with 1 M NaOH for 24 h is listed in Table 3. The oxygen content increased about 9 wt.% because of the introduction of –COOH and –CONH2 groups. Around 20 mol% nitrile groups were hydrolyzed to yield a NF membrane. The general process used is summarized in Fig. 4. Fig. 5 shows the ATR/FTIR spectra of the PAN membranes modified according to the process in Fig. 4. There was no change between Fig. 5(a) and (b). This indicates heat treatment does not affect the chemical composition of the PAN membranes. In Fig. 5(c), the broad adsorption band at 3415 cm−1 shifted


to 3330 cm−1 , with the intensity of the peak increased significantly. This band corresponds to the stretching vibration of the –NH2 group. The new peak at 1561 cm−1 corresponds to the –C O stretching vibration. These indicate the introduction of the –COONa and –CONH2 groups. Since the peak at 2243 cm−1 for –C N did not change significantly, only part of the –C N groups were hydrolyzed. In order to find the optimum heat treatment conditions, the heating time and temperature were varied. Table 4 lists the effect of heating time on membrane permeation properties. The highest rejection is obtained at heating time of 10 min while no rejection is observed at 6 min. The reason for this is that the water within pores is not heated enough to evaporate and hence no surface tension, no complexation from the precipitation of ZnCl2 and no shrinkage of pores. Once the heating time is longer than 15 min, because PAN might crystallize to make membranes brittle or the force from surface tension and ZnCl2 is too fierce, some defects or cracks might form and salt rejection is small or even unstable. Fig. 6 shows the hydrolysis reaction curves of membranes heated at 110 and 125 ◦ C for 10 min. The performance of the membrane heated at 110 ◦ C is much better than that heated at 125 ◦ C. When heated at 125 ◦ C, the water in pores evaporates too rapidly and damages the membrane. 3.3. Effect of operating conditions on NF membrane performance Figs. 7 and 8 demonstrate, respectively, the effect of operating conditions, i.e. feed NaCl concentration and operation pressure on the NaCl rejection. The reason for the decrease of the rejection

Fig. 4. Preparation process of nanofiltration membranes.


J. Wang et al. / Journal of Membrane Science 286 (2006) 333–341

Fig. 5. ATR/FTIR of PAN membranes. Table 4 Effect of heating time in air on permeation properties of PAN membranesa Properties



Heating time (min)

day−1 )

Flux Reaction time with 1 M NaOH (h) Flux after hydrolysis (m3 m−2 day−1 ) Rejection (%) a b







24.8 24.0 17.8 0

30.0 43.0 10.0 0

1.8 25.5 0.1 36.0

1.1 10.0 0.5 17.0

1.2 24.0 10.0–0.3b 0–10.0

1.0 4.0 5.9–0.7b 0–12.0

Saturated with 30 wt.% ZnCl2 and heated at 110 ◦ C. Flux was not stable and decreased quickly.

ratio with increasing concentration of salts is that effective area of the membrane pore becomes larger due to reduction of the thickness of the electrical double layer [14]. The rejection versus pressure relationship can provide insight into the nature of the separation mechanisms of the membrane. Typically, with RO membranes the rejection increases with increasing pressure, which can be interpreted by the ‘nonporous’ solution–diffusion model, where the salt flux is essentially independent of pressure and the water flux increases with applied pressure [15]. For microporous UF membranes, the

Fig. 6. Effects of heating temperature on: (a) flux and (b) rejection of PAN membranes. Saturated with 30 wt.% ZnCl2 and heated for 10 min.

Fig. 7. NaCl rejection of NF PAN membrane as a function of feed NaCl concentration.

J. Wang et al. / Journal of Membrane Science 286 (2006) 333–341


Fig. 8. NaCl rejection of NF PAN membrane as a function of operation pressure.

effect of increasing applied pressure is often a decrease in apparent rejection due to convective solute transport and concentration polarization effects. Fig. 8 shows that PAN NF behaves more like a ‘non-porous’ RO membrane over the experimental pressure range, i.e. rejection increases with increasing pressure, although pores definitely exist within the membrane. This can be interpreted by the Donnan–steric–pore model, which is described by the extended Nernst–Planck equation with equilibrium partitioning due to a combination of Donnan and steric mechanisms [16].

Fig. 9. Effects of pressure on pure water flux of NF PAN membranes.

3.4. Effect of counterion species on NF membrane performance The simplest model to describe membrane performance considers a membrane as a number of parallel cylindrical pores, the length of each pore being equal or almost equal to the thickness of the top layer of the membrane. This model describes the flux through micropores in an ideal situation, i.e. with uniformly distributed and evenly sized pores in the membrane, with no fouling, negligible concentration polarization, hydrophobicity, surface tension, etc. The volume flux through these pores may be described by the Hagen–Poiseuille equation for viscous flow [17]: J = L0p P, pure water permeability coefficient : L0p =

εr2 8Lμτ

where J is the solvent flux through the membrane; ε the surface porosity of the membrane; r the average pore radius; P the applied transmembrane pressure; μ the viscosity of the liquid

Fig. 10. Effects of hydrated radii on pure water permeability coefficient.

permeating the membrane; L the membrane thickness; τ is the tortuosity factor. In Fig. 9, a typical curve of the pure water flux versus the applied pressure is given. According to the Hagen–Poiseuille equation, L0p can be determined from the slope of the plot. The L0p of cationic NF PAN membranes with different species of counterions are calculated and listed in Table 5. For univalent alkaline counterions, the variation of (L0p )0.5 is approximately proportional to the hydrated radii, rH , of these ions as shown in Fig. 10. For divalent counterions, Mg2+ and Ca2+ , the L0p is inversely related to the hydrated radii, i.e. the smaller the rH , the higher L0p . Furthermore, the L0p of membranes with divalent counterions is greater than that of membranes with univalent

Table 5 Effect of counterion species on pure water permeability coefficients of NF PAN membranes Properties

L0p (×10−3 m3 m−2 day−1 bar−1 ) ˚ a Hydrated radii, rH (A) ˚ Average pore radii (A) a

From Ref. [18].

Membranes H+








102.1 2.82 34.5

27.1 3.82 17.8

28.4 3.58 18.2

30.3 3.31 18.8

37.0 4.28 20.8

40.2 4.12 21.7

102.4 4.30 34.6

119.7 4.75 37.4


J. Wang et al. / Journal of Membrane Science 286 (2006) 333–341

membrane will not change. This assumption is acceptable considering the similar effect of the univalent alkaline ions on these variables. Therefore, ε/(8Lμτ) in the Hagen–Poiseuille equation, k and d in Fig. 11 can be regarded as constants. Combining the Hagen–Poiseuille equation with the relationship of r = k − 2rH results in the following equation:   0.5 0.5 ε ε 0 0.5 −2 × rH (Lp ) = k 8Lμτ 8Lμτ Fig. 11. Scheme of radii within a nanopore.

alkaline counterions in spite of the larger rH of divalent counterions. In the case of divalent counterions, the ions interact with two –COO− groups and bend them towards each other, which make –COO− groups incline tighter to the pore surface. This effect is so strong that the influence from the increase in rH is negligible. Also, two –COO− groups attract Mg2+ and Ca2+ , which further reduces the distance, i.e. d, between the divalent ions and –COO– groups. Therefore, effective pore diameters, as well as the L0p increase when compared to membranes with univalent alkaline counterions. The relatively larger increase in L0p of the membrane with Zn2+ as counterions might be due to the complexation between Zn2+ and the nitrile groups. Nitrile groups are much smaller than –COO− , in other words, nitrile groups are much closer to the pore wall. These nitrile groups can drag Zn2+ and hence the –COO− groups combined with Zn2+ towards the pore surface, which results in much larger effective pore diameters. The L0p of membranes loaded with Al3+ counterions shows an additional increase over those loaded with Zn2+ . As with Zn2+ , Al3+ can complex with nitrile groups, but the trivalent Al3+ combines with three –COO− groups further reducing the effective pore diameter. This might be the explanation for the highest L0p despite the largest rH for Al3+ . For the H+ type membrane, in addition to the smallest rH of + H , there are two other possible reasons for the large L0p . The first is that hydrogen bonds can form through the –COOH groups where hydrogen ions play a role as divalent ions. Secondly, as a weak acid, –COOH does not dissociate completely, which makes the –COOH groups smaller than those completely dissociated salt forms. Interestingly, the L0p of membranes with Li+ , K+ , Mg2+ , Ca2+ and Zn2+ as counterions can return to the value of the membrane with Na+ as the counterion after soaking in 1 M NaCl overnight, while the L0p of those membrane with H+ and Al3+ can only return partly when treated in this manner. If carefully treated with dilute NaOH, the L0p could also return to the value of the Na+ type membrane. This additional step is required by the stronger combination of –COO− with H+ and Al3+ . From the slope of the plot in Fig. 10, the average pore radius of a PAN NF membrane can be estimated based on the simple model in Fig. 11. This model assumes that univalent alkaline counterions, Li+ , Na+ and K+ , will only affect the pore size through their physical size, i.e. hydrated radii. Other properties of the membrane such as the surface porosity, the tortuosity factor, the thickness and the viscosity of water permeating the

Thus, from the linear fit function in Fig. 10, the constants can be determined:   7.437 ˚ , k=2 = 25.426 A 0.585 ε = 0.08556 (×10−3 m3 m−4 day−1 bar−1 ) 8Lμτ The average pore radii of membranes in Table 5 are calculated from the value of ε/(8Lμτ) and the corresponding L0p according to the Hagen–Poiseuille equation. Those nanopore radii are ˚ beyond which, according to the simmuch larger than the 6 A ulation by Lynden-Bell and Rasaiah [19] the water coordinate number and solvation energy of sodium ions are similar to the value in the bulk solution. Thus, the ions in these nanopores are hydrated and the use of hydrated radii is reasonable. This simple method using univalent alkaline counterions as probes might be used to conveniently estimate the average pore radius of other cationic NF membranes. 4. Conclusions We have shown that by using the PAN/ZnCl2 system one can easily prepare NF membranes with highly dense pore surface functional groups. The phenomenon of pore collapse during the drying of UF membranes could be used to prepare NF membranes when the right template was chosen. This method could be extended to other polymer systems such as polystyrene or poly (styrene-co-acrylonitrile). In these systems, UF membranes could be precipitated from non-solvents besides water by the phase separation technique. These non-solvents can cause functional groups of the polymers to migrate onto the mesomacro continuous pore surface as in the case of water and nitrile groups of PAN. Also, the large pore size in the UF membrane allows easier modification of the pore surface chemistry. With the help of templates that interact with the functional groups on the meso-macropore surface, as in the case of ZnCl2 with nitrile groups, continuous micropores with highly dense pore surface functional groups could be formed during the heating and evaporation of the solvent of the template, e.g. water of ZnCl2 , once the surface tension force exceeds the modulus of the polymers. The effect of counterions on the L0p of cationic NF PAN membranes was also studied. A simple method based on the Hagen–Poiseuille equation was developed to estimate the average pore radius of the cationic NF membranes. Only by using the change of L0p with univalent alkaline counterions, can the average pore radius be conveniently calculated. This method

J. Wang et al. / Journal of Membrane Science 286 (2006) 333–341

might be extended to all the other charged NF membranes once suitable probes have been identified. Acknowledgements This work was supported 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. The ATR/FTIR were measured with the help of David Ladner in Professor Timm Strathmann’s group in University of Illinois at UrbanaChampaign. SEM analysis was carried out in the Center for Microanalysis of Materials, University of Illinois, which is partially supported by the U.S. Department of Energy under grant DEFG02-91-ER45439. References [1] A.I. Sch¨afer, A.G. Fane, T.D. Waite (Eds.), Nanofiltration—Principles and Applications, Elsevier Ltd., 2005. [2] R.W. Baker (Ed.), Membrane Technology and Applications, John Wiley & Sons, Ltd., 2004. [3] B. Jung, J.K. Yoon, B. Kim, H.W. Rhee, Effect of molecular weight of polymeric additives on formation, permeation properties and hypochlorite treatment of asymmetric polyacrylonitrile membranes, J. Membr. Sci. 243 (2004) 45–57. [4] E.H. Silbermann, Reactions of nitrile containing polymers, Usp. Khim. 55 (1983) 62–78. [5] I.C. Kim, H.G. Yun, K.H. Lee, Preparation of asymmetric polyacrylonitrile membrane with small pore size by phase inversion and post-treatment process, J. Membr. Sci. 199 (2002) 75–84. [6] M. Bryjak, H. Hodge, B. Dach, Modification of porous polyacrylonitrile membrane, Die Angew. Makromol. Chem. 260 (1998) 25–29.


[7] M. Ulbricht, A. Oechei, C. Lehmann, G. Tomaschewski, H.G. Hicke, Gas phase photoinduced graft polymerization of acrylic acid onto polyacrylonitrile ultrafiltration membranes, J. Appl. Polym. Sci. 56 (1995) 1707–1723. [8] M. Ulbricht, A. Oechei, Photo-bromination and photo-induced graft polymerization as a two-step approach for surface modification of polyacrylonitrile ultrafiltration membranes, Eur. Polym. J. 32 (1996) 1045–1054. [9] Y. Maeda, M. Tsuyumoto, H. Karakane, H. Tsungata, Separation of water–ethanol mixture by pervaporation through hydrolyzed polyacrylonitrile hollow fiber membranes, Polym. J. 23 (1991) 505–511. [10] M. Ulbricht, G. Belfort, Surface modification of ultrafiltration membranes by low temperature plasma. II: graft polymerization onto polyacrylonitrile and polysulfone, J. Membr. Sci. 111 (1996) 193–215. [11] Z.P. Zhao, J.D. Li, D.X. Zhang, C.X. Chen, Nanofiltration membrane prepared from polyacrylonitrile ultrafiltration membrane by low-temperature plasma. I: graft of acrylic acid in gas, J. Membr. Sci. 232 (2004) 1–8. [12] J. Jang, J. Kim, J. Bae, Effects of Lewis acid-type transition metal chloride additives on the thermal degradation of ABS, Polym. Degrad. Stab. 88 (2005) 324–332. [13] Membrane formation and modification, in: I. Pinnau, B.D. Freeman (Eds.), Proceedings of the ACS Symposium Series 744, vol. 18, 2000, pp. 18–19. [14] D.X. Wang, M. Su, Z.Y. Yu, X.L. Wang, M. Ando, T. Shintani, Separation performance of a nanofiltration membrane influenced by species and concentration of ions, Desalination 175 (2005) 219–225. [15] Membrane transport theory, in: R.W. Baker (Ed.), Membrane Technology and Applications, John Wiley & Sons, Ltd., 2004, pp. 15–87. [16] W.R. Bowen, J.S. Welfoot, Modelling the performance of nanofiltration membranes, in: A.I. Sch¨afer, A.G. Fane, T.D. Waite (Eds.), Nanofiltration—Principles and Applications, Elsevier Ltd., 2005, pp. 120–146. ´ ´ [17] A. Garc´ıa, S. Alvarez, F. Riera, R. Alvarez, J. Coca, Water and hexane permeate flux through organic and ceramic membranes: effect of pretreatment on hexane permeate flux, J. Membr. Sci. 253 (2005) 139–147. [18] E.R. Nightingale Jr., Phenomenological theory of ion solvation: effective radii of hydrated ions, J. Phys. Chem. 63 (1959) 1381–1387. [19] R.M. Lynden-Bell, J.C. Rasaiah, Mobility and solvation of ions in channels, J. Chem. Phys. 105 (1996) 9266–9280.