Poly(phthalazinone ether sulfone ketone) (PPESK) hollow fiber asymmetric nanofiltration membranes: Preparation, morphologies and properties

Poly(phthalazinone ether sulfone ketone) (PPESK) hollow fiber asymmetric nanofiltration membranes: Preparation, morphologies and properties

Journal of Membrane Science 270 (2006) 1–12 Poly(phthalazinone ether sulfone ketone) (PPESK) hollow fiber asymmetric nanofiltration membranes: Prepar...

879KB Sizes 0 Downloads 36 Views

Journal of Membrane Science 270 (2006) 1–12

Poly(phthalazinone ether sulfone ketone) (PPESK) hollow fiber asymmetric nanofiltration membranes: Preparation, morphologies and properties Yongqiang Yang a , Xigao Jian a,b,∗ , Daling Yang a , Shouhai Zhang a , Longjiang Zou c a College of Chemical Engineering, Dalian University of Technology, Dalian 116012, PR China State Key Laboratory of Fine Chemicals, Dalian University of Technology, Dalian 116012, PR China c Center of Material Testing and Analysis, Dalian University of Technology, Dalian 116012, PR China

b

Received 6 April 2005; received in revised form 8 June 2005; accepted 11 June 2005 Available online 19 September 2005

Abstract Using ethylene glycol methyl ether (EGME) or acetic acid (AA) as additives and N-methyl-2-pyrrolidone (NMP) as a solvent, defect-free skinned poly(phthalazinone ether sulfone ketone) (PPESK) hollow fiber asymmetric nanofiltration membranes were prepared by dry/wet phase inversion technique from 23 wt.% solids of PPESK/additive/NMP solutions. The pure water was used as a coagulant media. The effects of non-solvent additive on the membrane morphologies and separation performance of PPESK hollow fiber membranes were studied. The membrane structures of PPESK hollow fiber membranes including the cross-section (CS) of the inner edge and the outer edge and the external surface were characterized by scanning electron microscopy (SEM). It was found that using EGME, AA or the combination of them as additives can improve the permeation flux and solute rejection by changing the membrane morphology from finger-like shape to the spongelike structure. Based on the experimental results, both the membrane performance and the membrane structures were correlated and the pure water permeation fluxes of PPESK hollow fiber membranes can reach 211 L m−2 h−1 while the molecular weight cut-off is approximately 600 (96.1%), determined by PEG. © 2005 Elsevier B.V. All rights reserved. Keywords: Poly(phthalazinone ether sulfone ketone); Hollow fiber; Asymmetric nanofiltration membrane; Defect-free; Morphology

1. Introduction The development of nanofiltration technology in separation processes is being widely studied because of its advantages, such as low operation pressure, high flux, high retention of multivalent anion salts and organic molecules above 300 MW. Therefore, it has given rise to worldwide interest [1]. Moreover, because the hollow fiber [2] membranes have high packing density (membrane area per unit volume of vessel), hollow fibers would be the best in terms of high area/volume ratio and easier modules assembly. A large variety of nanofiltration membranes have been developed and used in different separation applications [3]. ∗

Corresponding author. Tel.: +86 411 836 39223; fax: +86 411 8363 9223. E-mail address: [email protected] (X. Jian).

0376-7388/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.memsci.2005.06.036

Most commercially available nanofiltration membranes are composite polyamide membranes prepared by interfacial polymerization [4,5]. However, the main drawback of polyamide membranes is their susceptibility to free chlorine and alkaline which causes degradation of the amide group. The choice of membrane materials depends on both chemical and physical compatibility of the selective layer with the substrate, which in turn determines the stability and performance of the resulting composite membranes. A series of poly(phthalazinone ether sulfone ketone) (PPESK) copolymers, containing different ratios of diphenylsulfone and diphenylketone units, were previously synthesized [6–8], which is a series of versatile high temperature resistant polymeric materials. In addition, PPESK containing rigid aromatic rings has shown superior mechanical strength, chemical resistance and very high glass transition

Y. Yang et al. / Journal of Membrane Science 270 (2006) 1–12

2

temperature (Tg ), in the range of 263–305 ◦ C [9]. Flat-sheet membranes made from PPESK have shown good separation and permeation properties for the separation of gases and liquids [10,11]. However, there were no previous reports about fabricating PPESK hollow fiber membranes through dry/wetspinning technique. Therefore, the need for hollow fibers in nanofiltration applications makes it desirable to prepare defect-free skinned nanofiltration membranes from casting solution to overcome the drawbacks of polyamide composite membranes. In this study, defect-free skinned asymmetric nanofiltration membranes were prepared from the solution of PPESK, additives and NMP by dry/wet phase inversion technique. Two different kinds of non-solvent additives (NSA), EGME and AA, were mixed with the polymer solution. The effects of the PPESK casting solutions, including different non-solvent additives, on the membrane morphologies and performance are discussed in this paper. This research is focused on preparing defect-free skinned asymmetric nanofiltration membranes.

2. Experimental and methods 2.1. Materials PPESK in powder form as membrane material was obtained from Dalian Polymer New Materials Co. Ltd. (PR China). Its chemical structure is shown in Fig. 1. Reagent grade N-methyl-2-pyrrolidone (NMP 98%) was used as solvent and EGME, AA used as NSA. PEG600 MW was used as a solute obtained from Shenyang Chemical Agent Company (PR China). Deionized water was used to make all solutions. All reagents used in the experiments are analytical grade chemicals. TOC (VAMAX1200, Japan) was employed for measuring the concentration of PEG600 solution. 2.2. Solubility parameters of solvent/non-solvent and binary mixture The solubility parameter (δi,s ) of a binary mixture (solvent/NSA mixture) is calculated based on the following equation [12,13]:

δi,s =

X1 V1 δi,1 + X2 V2 δi,2 , X 1 V1 + X 2 V 2

i = d, p, h

(1)

The solubility parameters (δsp ) of the solvent/non-solvent and the binary mixture can be calculated by Eq. (2): 0.5

δsp = (δ2d + δ2p + δ2h )

(2)

where X is the molar fraction, V represents molar volume, and subscripts 1 and 2 refer to the solvent NMP and NSA, respectively; subscripts d, p and h represent dispersion, polar and hydrogen bonding components of the solubility parameter of pure component, respectively. Based on Eqs. (1) and (2), solubility parameters of the solvent and the mixture of NMP/NSA are obtained in Table 1. 2.3. Phase separation diagram The phase diagram of the ternary membrane forming system, PPESK/NMP/NSA, was determined by the means of turbidity measurements and theoretical calculations, which were based on experimentally, determined parameters of Boom’s [14] linearized cloud point relation for this system. The cloud point was determined by a titrimetric method. The certain content of PPESK casting solution was prepared and placed in a glass bottle. The solution temperature was stabilized at room temperature (25 ◦ C) by a water thermostat and the cloud point of the system at this temperature was measured. For example, EGME or AA was slowly added to the polymer solution with a burette under vigorous stirring, respectively. This process lasted until the solution became visually turbid. Then the solution temperature was quickly elevated from 25 to 70 ◦ C for 30 min to make the casting solution clear. If the solution was turbid again when the polymer solution was naturally cooled to room temperature, the cloud point was determined as well as the amount of the NSA needed. Then, the composition of the system at the cloud point was calculated as the ratio of NSA to polymer solution plus NSA. The determination of the cloud points of the casting solution at high polymer concentration was carried out on the basis of the cloud point correlation. The binodal lines in ternary phase diagram were obtained by combining the experimental data of the cloud point measurement in a wide concentration range. The cloud point curves (bimodal lines) using those experimentally measured for a

Fig. 1. Chemical structure of poly(phthalazinone ether sulfone ketone) (PPESK, S/K=1/1).

Y. Yang et al. / Journal of Membrane Science 270 (2006) 1–12

3

Table 1 The solubility parameters of PPESK, solvent, non-solvents and the mixture of NMP/NSA [12,13] Solvent, binary mixture solvent and PPESK

δd (MPa)1/2

δp (MPa)1/2

δh (MPa)1/2

δsp (MPa)1/2

NMP EGME AA H2 O PPESK (S/K=1:1) NMP/EGME 70:7 NMP/EGME 64:13 NMP/EGME 59:18 NMP/EGME 55:22 NMP/EGME 51.5:25.5 NMP/AA 75:2 NMP/AA 72:5 NMP/AA 69:8 NMP/AA 67:10 NMP//EGME/AA 58.1//17.4/1.5 NMP//EGME/AA 56.9//17.1/3.0 NMP//EGME/AA 55.9//16.7/4.4 NMP//EGME/AA 54.8//16.5/5.7

18.0 16.9 20.2 15.5 18.8 17.89 17.80 17.73 17.67 17.62 18.06 18.14 18.23 18.29 17.78 17.83 17.87 17.91

12.3 8.5 8.9 16.0 11.3 11.93 11.62 11.36 11.16 10.98 12.21 12.08 11.95 11.86 11.33 11.28 11.24 11.19

7.2 17.5 14.6 42.4 8.3 8.20 9.04 9.74 10.29 10.77 7.39 7.68 7.97 8.17 9.80 9.90 9.97 10.07

22.9 25.7 26.5 47.9 23.3 23.02 23.10 23.20 23.29 23.39 23.02 23.11 23.21 23.27 23.25 23.30 23.35 23.40

PPESK/NMP/NSA casting system in this work are plotted in Fig. 2. 2.4. Measurement of coagulation values Coagulation values were measured to evaluate NSA tolerance. A polymer solution (100 g) with a mass ratio of mixture of PPESK and NMP was prepared. The polymer solution was placed in a constant temperature bath and titrated with deionized water at 25 ◦ C until the clear polymer solution became cloudy. The results of the coagulation values are shown in Table 2. 2.5. Hollow fiber membrane and membrane modules preparation Dried PPESK powder was added into a mixture of additive and NMP in the glass bottle and mixed until the solution became homogeneous. Hollow fiber membranes were spun

at room temperature employing the dry/wet phase inversion method, which has been described elsewhere [15–18]. Table 3 gives the spinning conditions of the prepared hollow fiber nanofiltration membranes in this paper. The nascent hollow fiber membranes emerged from the tip of the spinneret and passed through an air gap (20 mm) before entering the external coagulant (water). All hollow fiber membranes were not drawn (hence, no extension), which means that the take-up velocity was nearly the same as the free falling velocity of the nascent hollow fiber velocity in the coagulation bath. The coagulation bath and bore fluid were maintained at room temperature. The prepared hollow fiber membranes were stored in the water bath for 72 h at room temperature to remove the residual additive and NMP, then kept in an aqueous glycerin solution (30 wt.%) and Towen80 (1.5 wt.%) for 12 h to prevent collapse of its porous structure and dried in air at room temperature for making test membrane modules. Table 4 summarizes the composition of the casting solution and the outer diameter (o.d.)/inner diameter (i.d.) dimensions of the spinneret and the fabricated hollow fiber membranes. Membrane modules were prepared to test hollow fiber separation performance in terms of pure water permeation flux and solute rejection quantitatively. Each module consisted of 10 fibers with a length of 270 mm. The shell sides of the two ends of the bundles were glued onto two glass tube tees Table 2 The coagulation value of H2 O in PPESK/NMP/NSA systems Ratio of NMP to NSA (NMP/NSA) 10/0 9/1 7/3 5/5

Fig. 2. Binodal lines in the ternary phase diagram of PPESK/NMP/NSA system at 25 ◦ C.

Coagulation value (g)a EGME

AA

8.4 6.0 2.1 1.5

8.4 4.3 1.7 –

a Polymer solution: 1 g PPESK +99 g NMP, room temperature; coagulation: H2 O.

Y. Yang et al. / Journal of Membrane Science 270 (2006) 1–12

4

Table 3 The spinning conditions of the prepared hollow fiber membranes Parameter

Casting flow rate (mL/min)

Fluid flow rate (mL/s)

Tcoagulant (◦ C)

Troom (◦ C)

RH (%)

M1–M5 M6–M9 M10–M13

3.3 2.8 2.5

0.024 0.034 0.037

11.7 16.3 17.6

14.5 16.9 20.0

61.2 65.6 54.1

Table 4 The composition of the casting solution and outer diameter/inner diameter dimensions of the spinneret and the prepared PPESK hollow fiber membranes Membrane no.

PPESK/NSA/NMP ratio

o.d. (␮m)

i.d. (␮m)

i.d./o.d. ratio

Thickness (␮m)

Spinneret



1149

753

0.66

198

M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 M11 M12 M13

23/7/70 23/13/64 23/18/59 23/22/55 23/25.5/51.5 23/2/75 23/5/72 23/8/69 23/10/67 23/17.4//1.5/58.1 23/17.1//3.0/56.9 23/16.7//4.4/55.9 23/16.5//5.7/54.8

1115 1138 1160 1173 1054 1014 1042 1145 1071 1092 1086 1060 1099

757 777 780 796 685 632 629 742 727 694 691 704 769

0.68 0.68 0.67 0.68 0.65 0.62 0.60 0.65 0.68 0.64 0.64 0.66 0.70

179 181 190 188 185 189 207 202 173 199 198 178 165

o.d., outer diameter; i.d., inner diameter.

using a normal-setting epoxy resin. These modules were left overnight for curing before tested. To eliminate the effect of the residual glycerol and Towen80 on module performance, each module was immersed in water for 24 h and run in the test system for 30 min under a pressure of 0.4 MPa before any sample collection. 2.6. Membrane performance measurement Fig. 3 shows a schematic diagram of solute–water separation membrane unit designed for hollow fiber membrane characterization in our laboratory. Pure water permeation flux

and rejection measurements for deioned water and PEG600 were performed for each membrane sample, three modules were parallel tested and the average of their performance was reported. At the transmembrane pressure 0.4 MPa and room temperature for 30 min, the pure water permeation flux and the rejection of PEG600 were measured at the pressure 0.3 MPa and room temperature. Water was fed at a constant pressure of 0.3 MPa from the inner lumen to the outer surface of hollow fiber membranes and was collected in a measuring cylinder and measured. Pure water permeation fluxes (F), were obtained as follows: F=

Q At

(3)

where Q is the total volume of the permeation water during the experiment; A represents the virtual membrane area; t denotes the measuring time. The solute rejection of the membrane was determined with an aqueous solution containing 100 ppm PEG600, which was fed at constant flow from the lumen to the outer surface of the membranes. The membrane solute rejection R (%) was calculated by Eq. (4):   Cp R (%) = 1 − × 100 (4) Cf Fig. 3. Schematic diagram of the properties evaluation of the hollow fiber membrane: (1) feed reservoir, (2) tank, (3) recirculation pump, (4) pressure gauge, (5) hollow fiber membranes module, (6) rotor flowmeter, (7) circumfluence pressure valve.

where Cf and Cp represent the solute concentration in feed and separated solution, respectively. The PEG600 concentration in the feed and the permeation were analyzed with TOC. All results were the average of three parallel modules.

Y. Yang et al. / Journal of Membrane Science 270 (2006) 1–12

2.7. Structure of asymmetric membrane The morphologies of the cross-section and the external surface of the asymmetric hollow fiber membranes were observed with a scanning electron microscope (SEM; JSM5600L, JEOL, Japan). After the ethanol–hexane solvent exchange step, the hollow fiber membrane samples were cryogenically fractured in liquid nitrogen and then coated with gold to obtain an adequate contrast of the membrane fracture.

3. Results and discussion 3.1. Ternary phase diagram analysis of PPESK/NMP/NSA system In order to characterize the hollow fiber membrane forming system with respect to physico-chemical parameters, the cloud points of the casting solution system were determined. Cloud point measurement is a useful way to interpret the thermodynamic properties of a membrane-forming system because cloud point curves can approximately represent the bimodal lines in the phase diagrams [19]. Through the bimodal lines, one can distinguish the thermodynamic difference among membrane-forming systems and the corresponding coagulation power of NSA. Recent studies [20,21] have also revealed that this is a good and simple solution for the establishment of dope and coagulant composition. It can be seen from Fig. 2 that, when adding NMP to the NSA, the binodal line of the casting solutuin system shifted to the polymer/NSA axis and the homogeneous phase region was broadened. For the system containing higher amounts of NMP in the NSA, the demixing gap became smaller. This ternary phase diagram indicated that, with the addition of NMP in the NSA, the PPESK/NMP/NSA system became thermodynamically more stable and needed more NSA to induce a liquid–liquid phase separation. These results were useful for interpreting the morphology and properties of the PPESK hollow fiber nanofiltration membranes. It is also found from Table 2 that the coagulation values of water increase as follows: EGME > AA. In fact, when nonsolvents are added to the PPESK/NMP solutions, the thermodynamic status in the PPESK/NMP solutions is changed because the solvent power is weakened. The higher the solubility parameter disparity of solvent and non-solvent, the more intense the interaction of solvent with non-solvent and the faster the solubility of solvent to polymer will decline [16]. If NSA has a strong effect on PPESK casting solution, the precipitation of PPESK will occur more easily. 3.2. Morphological studies of PPESK hollow fiber membranes It is known that the compositions of the casting solution influence the structure of the asymmetric membrane strongly

5

[4]. Generally, an asymmetric membrane with a dense skin supported by a porous substructure is obtained when the casting solution is immersed directly into a non-solvent bath after experience certain air distance. This skin, formed due to the differences in solution rates and diffusion rates of the components in the membrane formation, plays an essential and crucial role in many membrane separation processes such as reverse osmosis, gas separation and pervaporation. During the phase separation process, a one-phase casting solution is converted into two-phase system consisting of a solid phase (PPESK-rich) that forms the membrane structure and a liquid phase (PPESK-poor) that forms the pores in the final membrane. For instance, the factor to be considered is the presence of an additive, which changes the thermodynamic and kinetic conditions in which the phase separate process occurs during the immersion in the coagulation bath. Therefore, PPESK hollow fiber nanofiltration membranes are prepared from the casting solution containing PPESK and EGME or AA. We examined the hollow fiber membrane morphology using SEM to investigate the effect of the kind and the concentration of NSA on the structure formation. The hollow fiber membrane thickness values that were observed under the SEM were 165–207 ␮m whatever the casting solution composition. The hollow fiber membranes exhibited the classical asymmetric structure with a top skin layer supported by a finger-like structure or sponge-shape structure. The effects of NSA on the membrane structures of the cross-section (CS) and the external surface (ES) of PPESK hollow fiber membranes dry/wet spinning are shown in Figs. 4, 6 and 8. It can be seen from those figures that the PPESK hollow fiber membrane prepared from lower NSA concentration have similar, two-layer finger-like structure extended to the middle of the cross-section and there is a layer sponge-like structure at the middle of the cross-section because the pure water was used as both the inner and outer precipitant. Moreover, with the content of NSA increasing in the casting solution, the structures of the hollow fiber membranes change from a finger-like shape to a sponge-like structure. The active skin layer was denser for the higher NSA concentration than that observed for lower NSA concentration whereas a more open structure was apparent beneath. From the solubility parameters (δsp ) of NSA listed in Table 1, it is also found that the δsp values of NSA have the following order: AA > EGME. This might explain the formation of a denser skin layer. Moreover, as EGME is more hydrophobic than AA, the skin layer formed during the demixing process slowed down the water penetration thereby allowing the growth of polymer-lean droplets by coalescence before the membrane solidification. The experimental results of SEM of the hollow fiber membranes illustrate that the precipitation of PPESK forming the hollow fiber membrane external surfaces and the exchange of solvents with non-solvents may be affected by the diffusion of the water-soluble additives (EGME, AA and of two) from the casting solution to the water bath. Since the separation

6

Y. Yang et al. / Journal of Membrane Science 270 (2006) 1–12

performance of hollow fiber membranes is largely determined by the surface structure, this can explain why membranes produced with different additives resulted in different separation performance. 3.3. Effect of EGME on the morphologies and properties of PPESK hollow fiber membrane PPESK hollow fiber membranes were prepared from a 23 wt.% PPESK casting solution using water as internal and external coagulant. The other spinning parameters are spec-

ified in Table 3. In the dry/wet phase inversion process, the hollow fiber membrane experiences air coagulation of about 20 mm after extrusion from the spinneret. The effects of the EGME content in the 23 wt.% PPESK casting solution on the morphologies of the resultant membranes are shown in Fig. 4. Fig. 4 (M1-IE to M5-IE and M1-OE to M5-OE) illustrates the effect of EGME content on the inner and outer edge structure of the hollow fiber membrane. With an increase of EGME content in the casting solution, it can be seen from Fig. 4 that the finger-like structures decrease both in size and number, which are replaced by a sponge-like structure

Fig. 4. Effect of different EGME content on the morphologies (IE: inner edge; OE: outer edge; ES: external surface; PM: part magnified of the cross-section) of PPESK HF membranes prepared from the casting solution of 23 wt.% PPESK (1, 2, 3, 4 and 5 refer to the casting solution containing 7, 13, 18, 22 and 25.5 wt.% EGME, respectively).

Y. Yang et al. / Journal of Membrane Science 270 (2006) 1–12

formed under the inner and outer skin layer that eventually disappeared when NMP/EGME (51.5/25.5) was used as the casting solvent system. It is network porous from the magnified picture of the cross-section of Fig. 4 (M5-PM) when a 25.5 wt.% EGME was used as NSA. The reason is that by adding NSA, the polymer solution becomes unstable, which indicates that the aggregation of polymer molecules is due to entanglement of polymer chains promoted. Therefore, solvents located between polymer chains can be easily diffused out the coagulation bath. In addition to this aggregation of polymer chains by adding NSA in casting solutions, a phenomenon forming a sharp interface between the casting solution and the water should be investigated. However, the outer coagulation and the inner coagulation will meet each other and form sponge structure in the middle of the cross-section. In the other words, it can be seen from Fig. 4 that the morphology change in the inner edge of the fiber wall was smaller than that in the outer edge. This could be due to the less water amount in the fiber lumen compared with the mass volume of water in the outer coagulant, which reached the precipitation rate of the polymer solution. The main route of the phase inversion process involves two different types of phase transition [22]. These are: (1) liquid–liquid phase separation, in which the completely miscible solution crosses the bimodal boundary to enter the two-phase region, and (2) solidification; Since the viscosity of the casting solution increases to a certain assumed value, the motion of polymer chains assumed value, the motion of polymer chains will be limited and the system can be regarded as a solid to fix the membrane structure. Therefore, the solidification is the major factor determining pore wall structure and the rate for the liquid– liquid phase separation is essentially dependent on NSA content. Fig. 4 (M1-ES to M5-ES) shows the SEM images of the external surface structures of hollow fiber membranes. From M1-ES to M5-ES, it can be seen that the external surface of PPESK hollow fiber membrane looks more smooth, especially membrane M5-ES with higher EGME content. This was the result of lower coagulation rate at the inner and the outer fiber wall. With the increase of EGME content in the casting solution, its coagulation rate became lower than that of water. Like a report of Strathmann et al. [23] it seems that the NSA outflow rate into the coagulation bath is faster than that of NMP. Moreover, it can be regarded that the water inflow rate into the casting solution is slower compared with the outflow rate of solvent mixture. Thus, the sponge layer in the fiber wall moved to the inner and outer surface. Fig. 2 shows the binodal lines of different membraneforming systems. It can be seen from Fig. 2 that, with the increase of EGME content in casting solution, the binodal line of the phase separation system shifted away from the PPESK/NMP axis. The addition of EGME appears to reduce the solvent power for the polymer, and lower the intermolecular interaction between polymer and good solvent (NMP). Since polymer molecules are clustered together to exclude

7

solvent from within their domains, a small pore rapidly grows into a large pore. The concentration of the solution that surrounds the pore continually increases until it enters the solidification region to form the pore wall. In addition, at high affinity between solvent and NSA, the pore wall solidifies late so it has adequate time to grow or even to form macrovoids in the ultimate membrane. The above conditions are advantageous to a finger-like structure formation requires that both solvent and NSA can rapidly and consecutively diffuse into a younger pore before the solidification of the pore wall [24]. In another way, in order to observe the effect of EGME on the casting solution properties, the coagulation value was also determined. Table 3 shows that with the increase of NSA content in casting solution, the coagulation value decreased. Compared to AA, EGME required more coagulation, which means that the solvent power of NMP/EGME solvent mixture is greater than that of NMP/AA solvent mixture. In other words, strong NSA appears to reduce the solvent power for polymer, and lower the polymer–solvent interaction, and permit the occurrence of the polymer-polymer intermolecular interaction [5]. As a result, EGME acts as a weak NSA and AA a strong NSA. Increasing the weak or strong NSA content can result in decreasing solution compatibility and making the polymer solution closer to the point of incipient gelation. Table 1 also gives the solubility parameters of PPESK, solvent, non-solvents and the mixture of NMP/NSA. The ratio of NMP to EGME in the casting solution is an important factor for reducing the pore size in the dry/wet phase separation process. Hachisuka et al. [25] reported that polar solvent such as NMP, which dissolves in water easily, could not form a sharp interface between the solvent and the water. Therefore, by using a polar solvent, the porous skin layer can be formed. However, EGME can form a sharp interface with water because of the low miscibility between EGME and water, which is due to larger dipole moment difference of EGME and water than that of NMP and water. This is attributed to slower exchange rate of a polar solvent into the coagulation bath (water) than the ingress rate of water to polymer solution. This interface affects the exchange rate of NMP and EGME with water and suggests that EGME can inhibit water from entering into the polymer solution. These two phenomena (sharp interface formation and aggregation of the polymer chains) can explain the formation of the dense skin layer. The separation characteristics (such as the pure water permeation and the solute separation) of PPESK/EGME/NMP hollow fiber nanofiltration membrane spun from 23 wt.% PPESK concentration are given in Fig. 5. It appears that with the increase of EGME content in casting solution, the pure water permeation flux decreases from 299 to 93 L m−2 h−1 and the PEG600 rejection is increased from 39.5% to 98.5%. The reason is the tighter membrane structure for higher EGME content. The SEMs in Fig. 4 give a better explanation that the change of morphology leads to different membrane performances.

8

Y. Yang et al. / Journal of Membrane Science 270 (2006) 1–12

Fig. 5. Effect of EGME content in casting solution on the performance of the 23 wt.% PPESK hollow fiber membranes.

3.4. Effect of AA on the morphologies and properties of PPESK hollow fiber membrane The acetic acid content for the demixing of the PPESK casting solution was calculated according to the binodal lines in the ternary phase diagram of PPESK/NMP/AA system based on the cloud point measurement [26]. PPESK casting solutions with different AA content were used to prepare

PPESK hollow fiber nanofiltration membranes. All other spinning parameters can be seen from Table 3. In making the defect-free asymmetric membranes, many researchers emphasized that it is necessary to add a volatile NSA in the casting solution [4,27–30]. It is possible to make the defect-free skinned asymmetric nanofiltration membrane by controlling the ratio of solvent to NSA. This paper used AA as NSA. The addition of a small amount of strong NSA like water, AA and ethanol into the casting solution brings the initial composition of the casting solution to a very unstable state [31]. As can be seen in Table 2, AA is a stronger NSA than EGME. This implies that the addition of a small amount of AA is enough to make the composition of the casting solution close to the precipitation point. Fig. 6 shows that the morphologies of hollow fiber nanofiltration membranes fabricated from PPESK casting solutions with different AA content. With the content of NSA increased, the ternary system becomes unstable, the solvent diffuses out rapidly from the surface casting solution, and then the top layer is dense. As a result, the composition of the sublayer does not change as rapidly, and the solution is within the initial homogeneous solution region. In the ternary phase diagram the composition path of the top layer directly enters the solidification region. The sublayer structures are influenced by the top layer since the dense

Fig. 6. Effect of different AA content on the morphologies (CS: cross-section; IE: inner edge; OE: outer edge; ES: external surface; PM: part magnified of the cross-section) of PPESK HF membranes prepared from the casting solution of 23 wt.% PPESK (6, 7, 8, and 9 refer to the casting solution containing 2, 5, 8 and 10 wt.% AA, respectively).

Y. Yang et al. / Journal of Membrane Science 270 (2006) 1–12

top layer will increase the barrier for diffusion of solvent from and NSA into the interior casting solution. Therefore, the dense top layer controls the diffusion ratio of solvent to NSA [24]. The SEM of M6 and M7 reveals the long finger-like structures are formed near the inner wall and the outer wall. Between the inner and the outer layers sponge-shape structure appears. However, there is formation of smaller cavities when the AA content increases to 8 wt.%. The morphology of the cross-section (M8) consists of large voids in the shape of spheres or ellipsoids, and between these voids there is a spongy structure, while the figure M9 shows that the sponge-like and the smaller cavities have been eliminated when AA content is 10 wt.% and a uniform sponge-like structure extends over the entire membrane cross-section. These changes in the morphology are different to the EGME as the NSA. The reason for these phenomena may be that the presence of AA changed the phase separation path of the PPESK casting solution. Upon the addition of AA into the PPESK casting solution, the binodal line of the membrane-forming system shifted to the NMP/PPESK axis, less AA was needed for the phase separation. Therefore, the PPESK casting solution composition would be located in the unstable region of the phase diagram before the phase separation occurred in the metastable region. According to the literature [24] mentioned in liquid–liquid phase separation, the polymer concentration of the polymer-rich phase will increase very greatly at low mutual affinity between solvent and non-solvent so that a pore wall will immediately reach solidification. Furthermore, when the affinity between solvent and non-solvent is relatively low, the diffusion process will be quite slow so as to result in the slow growing rate of the pore; the pore size is therefore comparatively small. Rapid exchange between solvent and non-solvent would render the PPESK casting solution supersaturated, and spinodal demixing would take place [32]. As shown in Fig. 6 (M9-PM), the nodular structure was eventually formed in the membrane cross-section. This makes the radius of gyration of the polymer coil smaller. The resulting intermolecular physical cross-links restrict chain mobility. In other words, by increasing the amount of EGME, the coils of the polymer chains are crowded together and as a result become smaller, indicating that the interaction between the polymer and the solvent mixture is apparently greater than that between the polymer molecules. However, in case of AA, the size of the polymer network was increased by enhancing the interaction between the polymer molecules at the expense of the interaction between polymer and solvent, which suggests that polymer chains are interpenetrated, indicating that the strength of interaction of polymer and solvent molecule is slightly greater than that of polymer molecules. The increasing size of the polymer network owing to chain interpenetration increases the viscosity of the casting solution [31]. In other words the pure water flux is lower for increasing the rejection rate of PEG600. By the introduction of NSA additive, solvent mixture becomes

9

Fig. 7. Effect of AA content in casting solution on the performance of the 23 wt.% PPESK hollow fiber membranes.

readily diffused from the surface casting solution to the coagulation bath, then the dense top layer is formed. Fig. 7 illustrates the pure water permeation and separation properties of PPESK hollow fiber membrane, which was fabricated from different AA content. It is shown from Fig. 7 that the pure water flux decreased from 306 to 221 L m−2 h−1 and the PEG600 rejection rate increased from 47.1% to 88.7%. It can be seen from Fig. 6 that with increasing the content of AA, which implies the coagulant tolerance is lowered due to highly entangled solution conformation. The SEM photographs in Fig. 6 also show that with increasing of AA content, the large finger-like macrovoids gradually disappear. Instead, sponge-shape cross-sections appear. Even though we could not precisely see the structure of skin layer, it seems that the nodular skin structure was changed to closely connect cellular dense skin structure. This can also explain the reason why the pure water flux decreased and the solute rejection rate increased. 3.5. Effect of AA/EGME on the morphologies and properties of PPESK hollow fiber membrane As shown in Figs. 5 and 7 the solvent systems of NMP/EGME and NMP/AA were not suitable for a good performance nanofiltration membrane. Instead, the solvent system of NMP//AA/EGME was chosen. It is reasonable to suspect that the addition of EGME can induce a sharp interface between the nascent membrane and the coagulant. Moreover, the addition of AA in the PPESK/NMP/EGME casting solution can phase separate by shifting the polymer solution composition into two-phase separate gap. Therefore, it is very difficult to control the AA content. When the amount of EGME is too small a sharp interface between the polymer solution and the coagulant cannot be formed. Solvent mixture ratio of NMP to EGME was fixed to 10/3 and the content of AA changed. It can be seen from Fig. 8, the effect of NSA contents on the morphology of the hollow fiber membrane is great, and when the NSA made by the combination the ratio of AA to EGME

10

Y. Yang et al. / Journal of Membrane Science 270 (2006) 1–12

Fig. 8. Effect of different AA content on the morphologies (CS: cross-section; IE: inner edge; OE: outer edge; ES: external surface; PM: part magnified of the cross-section) of PPESK HF membranes prepared from the casting solution of 23 wt.% PPESK and the ratio of the EGME/NMP is 3/10 (10, 11, 12, and 13 refer to the casting solution containing1.5, 3.0, 4.4 and 5.7 wt.% AA, respectively).

was added in the casting solution. With an increase in the ratio of AA//EGME, finger-type macrovoids disappear and the skin layer is denser. The structure of these membranes differs from that of porous and dense skinned membranes. As reported by Kesting [5], generally the large finger-like macrovoids and the cavity-like structures are formed when the coagulation process is fast, whereas the slow coagulation rate results in a porous sponge-like structure. This indicates that the addition of AA//EGME will delay the coagulation process and a long finger-like structure will change to a short finger-like structure and farther to a sponge-like structure. This is to say, the smaller the mutual affinity of solvent and NSA, the more NSA in the polymer-poor phase and the slower the mixing tendency of solvent and NSA, the more induction time it takes for the NSA into the casting solution to form nuclei, and the slower the liquid–liquid phase separation of the system. Thus nucleation starts until a certain amount of

NSA has diffused into the casting solution. Consequently, many nuclei are initiated later but at the same time. Hence, the growth of every nucleus will be limited by other nearby nuclei because every nucleus consumes solvent. In this way the growth of macrovoids is impossible, and a sponge-like structure is formed [24]. It also can be seen from the photographs of the part magnified of cross section (M12-PM and M13-PM) and the external skin that the skin layer was very well developed and supported by porous sponge layer with micropore. It appears that during the dry phase separation process, liquid–liquid phase separation takes place, which is of benefit in preparing defect-free skinned asymmetric nanofiltration membranes. Though it is difficult to measure the outflow rate of solvent and NSA from the polymer solution, the diffusion rates of EGME and AA in water should be adjusted by varying the solvent mixture ratio during the coagulation process.

Y. Yang et al. / Journal of Membrane Science 270 (2006) 1–12

11

nanofiltration membranes have the pure water permeation fluxes of 211 L m−2 h−1 , and the molecular weight cut-off of the membranes is approximately 600 MW (96.1%).

Acknowledgements The authors are grateful for financial supports of the National “973” Foundation R&D Plan Project of China (Grant No. 2003CB615700) and the National “863” High-tech R&D Plan Project of China (Grant No. 2003AA33G030). Fig. 9. Effect of AA content in the casting system of AA//EGME/NMP on the performance of the 23 wt.% PPESK hollow fiber membranes.

The separation performances of PPESK/EGME//AA/ NMP hollow fiber membrane are given in Fig. 9. It can be concluded from Fig. 9 that this system of the casting solution has the higher pure water permeation fluxes (211 L m−2 h−1 ) and the rejection of PEG600 (96.1%) than that of EGME or AA as NSA. The membrane morphologies in Fig. 8 also can explain these experimental results. The present results show that the mixed solvent system containing AA/EGME is suitable for preparing defect-free asymmetric hollow fiber nanofiltration membranes with good performance.

4. Conclusion In order to prepare defect-free skinned hollow fiber PPESK nanofiltration membranes having high flux and high solute rejection, EGME and AA were used as additives, and the asymmetric membranes were dry-wet spun from 23 wt.% polymer solutions. Because the composition of the casting solution containing appropriate NSA was very important, a weak NSA (ethylene glycol methyl ether) was chosen forming a sharp interface between the polymer solution and the coagulant (water). Simultaneously, adding acetic acid that is a strong NSA in the casting solution can render the casting solution close to the binodal composition. The effects of different additives on the membrane structure, pure water permeation fluxes and PEG600-water separation characteristics of PPESK hollow fiber membranes were studied. The SEM images illustrated that the presence of EGME or AA as an additives could result in the change of the hollow fiber membrane structure from the finger-shape structure to the sponge-like structure. PPESK hollow fiber asymmetric nanofiltration membrane exhibiting both appropriate nanofiltration performance characteristics (having the higher pure water permeation fluxes and higher solution rejections rate for PEG600) and ideal membrane morphology (the denser skin and the sublayer of sponge-like structure) could be prepared by the combination the ratio of EGME to AA. From the experimental results obtained, the defect-free skinned asymmetric

References [1] R.J. Peterson, Composite reverse osmosis and nanofiltration membranes, J. Membr. Sci. 83 (1993) 81–150. [2] Y. Kusuki, H. Shimazaki, N. Tanihara, et al., Gas permeation properties and characterization of asymmetric carbon membranes prepared by pyrolyzing asymmetric polyimide hollow fiber membrane, J. Membr. Sci. 134 (1997) 245–253. [3] Y. Magara, S. Kunikane, M. Itoh, Advanced membrane technology for application to water treatment, Water Sci. Technol. 37 (1998) 91–99. [4] M. Mulder, Basic Principles of Membrane Technology, 2nd ed., Kluwer Academic Publishers, London, 1996. [5] R.E. Kesting, Synthetic Polymeric Membranes, Wiley, New York, 1985. [6] X.G. Jian, Y.Z. Meng, H.B. Zheng, Polyethersulfone containing diazonaphthalene structure and preparing process, Chinese Patent 93 109 180 (February 1993). [7] X.G. Jian, Y.Z. Meng, H.B. Zheng, Polyetherketone containing diazonaphthalene structure and preparing process, Chinese Patent 93 109 179 (September 1993). [8] X.G. Jian, Y. Dai, L. Zheng, R.X. Xu, Application of poly(phthalazinone ether sulfone ketone)s to gas membrane separation, J. Appl. Polym. Sci. 71 (1999) 2385–2390. [9] Y.Z. Meng, A.S. Hay, X.G. Jian, S.C. Tjong, Synthesis of novel poly(phthalazinone ether sulfone ketone)s and improvement of their melt flow properties, J. Appl. Polym. Sci. 66 (1997) 1425–1432. [10] X.G. Jian, Y. Dai, G.H. He, G.H. Chen, Preparation of UF and NF poly(phthalazinone ether sulfone ketone) membranes for high temperature application, J. Membr. Sci. 161 (1999) 185–191. [11] Y. Dai, X.G. Jian, S.H. Zhang, M.D. Guiver, Thermostable ultrafiltration and nanofiltration membranes from sulfonated poly(phthalazinone ether sulfone ketone), J. Membr. Sci. 188 (2001) 195–203. [12] J. Brandrup, E.H. Immergut, Polymer Handbook [M], 3rd ed., Wiley, New York, 1989. [13] J.Y. Lai, F.C. Lin, C.C. Wang, et al., Effect of non-solvent additives on the porosity and morphology of asymmetric TPX membranes, J. Membr. Sci. 118 (1996) 49–61. [14] R.M. Boom, Th. van den Boomgaard, J.W.A. van den Berg, et al., Linearized cloudpoint curve correlation for ternary systems consisting of one polymer, one solvent and one non-solvent, Polymer 34 (1993) 2348–2356. [15] Z.L. Xu, T.S. Chung, Y. Huang, Effect of polyvinylpyrrolidone molecular weights on morphology, oil/water separation, mechanical and thermal properties of polyetherimide/polyvinylpyrrolidone hollow fiber membranes, J. Appl. Polym. Sci. 74 (1999) 2220–2233. [16] J. Xu, Z.L. Xu, Poly(vinyl chloride) (PVC) hollow fiber ultrafiltration membranes prepared from PVC/additives/solvent, J. Membr. Sci. 208 (2002) 203–212.

12

Y. Yang et al. / Journal of Membrane Science 270 (2006) 1–12

[17] T.S. Chung, Z.L. Xu, Asymmetric hollow fiber membranes prepared from miscible polybenzimidazole and polyetherimide blends, J. Membr. Sci. 147 (1998) 35–47. [18] T.S. Chung, S.K. Teoh, X.D. Hu, Formation of ultrathin highperformance polyethersulfone hollow-fiber membranes, J. Membr. Sci. 133 (1997) 161–175. [19] F.Q. Nie, et al., Preparation and characterization of polyacrylonitrile-based membranes: effects of internal coagulant on poly(acrylonitrileco-malefic acid) ultrafiltration hollow fiber membranes, Desalination 160 (2004) 43–50. [20] W. Albrecht, Th. Weigel, et al., Formation of hollow fiber membranes from poly(ether imide) at wet phase inversion using binary mixtures of solvents for the preparation of the dope, J. Membr. Sci. 192 (2001) 217–230. [21] W.F.C. Koops, Th. van den Boomgaard, et al., Considerations and restrictions on the theoretical validity of the linearized cloudpoint correlation, Polymer 39 (1998) 4835–4840. [22] T.H. Young, L.W. Chen, A two step mechanism of diffusion controlled ethylene vinyl alcohol membrane formation, J. Membr. Sci. 57 (1991) 69–81. [23] H. Strathmann, P. Scheible, R.W. Baker, A relationable for the preparation of Loeb-Sourirajan type cellulose acetate membrane, J. Appl. Polym. Sci. 15 (1971) 811–818. [24] T.H. Young, L.W. Chen, Pore formation mechanism of membranes from phase inversion process, Desalination 103 (1995) 233–247.

[25] H. Hachisuka, T. Ohara, K. Ikeda, New type asymmetric membranes having almost defect free hyper-thin skin layer and sponge-like porous matrix, J. Membr. Sci. 116 (1996) 265–272. [26] L.Q. Shen, Z.K. Xu, Y.Y. Xu, Phase separation behavior of poly(ether imide)/N,N-dimethyl acetamide/nonsolvent systems, J. Appl. Polym. Sci. 89 (2003) 875–881. [27] I. Pinnau, W.J. Koros, Structures and gas separation properties of asymmetric polysulfone membranes made by dry, wet and dry/wet phase inversion, J. Appl. Polym. Sci. 43 (1991) 1491– 1502. [28] H. Kawakami, M. Mikawa, S. Nagaoka, Gas permeability and selectivity through asymmetric polyimide membranes, J. Appl. Polym. Sci. 62 (1996) 965–971. [29] S.C. Pesek, W.J. Koros, Aqueous quenched asymmetric polysulfone membranes prepared by dry/wet phase separation, J. Membr. Sci. 81 (1993) 71–88. [30] Y.S. Kang, H.J. Kim, U.Y. Kim, Asymmetric membranes formation via immersion precipitation method. I. Kinetic effect, J. Membr. Sci. 60 (1991) 219–232. [31] I.C. Kim, K.H. Lee, T.M. Tak, Preparation and characterization of integrally skinned uncharged polyetherimide asymmetric nanofiltration membrane, J. Membr. Sci. 183 (2001) 235–247. [32] I.M. Wienk, Th. van den Boomgaard, C.A. Smolders, Formation of nodular structure in the top layer of ultrafiltration membranes, J. Appl. Polym. Sci. 53 (1994) 1011–1023.