MDI (P84) copolyimide flat sheet and hollow fiber membranes for ultrafiltration: Morphology transition and membrane performance

Desalination 285 (2012) 336–344

Contents lists available at SciVerse ScienceDirect

Desalination journal homepage: www.elsevier.com/locate/desal

Development of asymmetric BTDA-TDI/MDI (P84) copolyimide flat sheet and hollow fiber membranes for ultrafiltration: Morphology transition and membrane performance Jizhong Ren a, b,⁎, Zhansheng Li b a b

National Engineering Research Center of Membrane Technology, Dalian Institute of Chemical Physics, Chinese Academy of Sciences, 457 Zhongshan Road, Dalian 116023, China Institute of Environmental Science and Engineering, Nanyang Technological University, 18 Nanyang Drive, Singapore 637723, Singapore

a r t i c l e

i n f o

Article history: Received 10 April 2011 Received in revised form 8 October 2011 Accepted 18 October 2011 Available online 17 November 2011 Keywords: Membrane morphology Membrane thickness Hollow fiber membrane Ultrafiltration Pore size distribution

a b s t r a c t In this paper, the effects of γ-butyrolactone (GBL) weight ratio (wGBL) and membrane thickness on the formation of asymmetric flat sheet membranes prepared with P84 (BTDA-TDI/MDI co-polyimide)/N-methyl-2-pyrrolidone (NMP)/GBL casting solutions are investigated. With the increase of membrane thickness, the transition of membrane morphology from sponge-like to finger-like structure occurs at critical structure-transition thickness Lc. Lc and the general sponge-like structure thickness (Lgs) increase with wGBL. For 20 wt.% P84/ NMP/GBL casting solution, the membrane morphology changes from finger-like to sponge-like structure at the critical weight ratio of GBL (w⁎ = 0.69). The membrane morphology and performance of hollow fibers spun with various wGBL are observed. Compared with the hollow fiber membranes made of 18 wt.% P84/ NMP/GBL dope solution with wGBL = 0.75, the hollow fiber membranes spun with wGBL = 0.25 present a higher permeation flux and a larger MWCO. As wGBL increases from 0.25 to 0.75, the membrane morphology transfers from finger-like to sponge-like structure. An increase in shear rate shifts the rejection curves towards left, and lowers the MWCO of hollow fiber membranes. For hollow fiber membranes spun with wGBL = 0.75, a relatively high permeation flux and a large MWCO are obtained by the wet spinning process. © 2011 Elsevier B.V. All rights reserved.

1. Introduction Membrane processes have been widely used in a variety of industries such as water and wastewater treatment, the separation and purification in food technology, biotechnology and petrochemical processes, etc. Asymmetric membranes prepared with phase-inversion technique mainly include flat sheet and hollow fiber membranes. For flat sheet membranes, the nascent films are cast onto a moving nonwoven fabric web by the casting blade, and then exposed to a gaseous environment for solvent evaporation prior to entering the coagulant bath, finally precipitated by quenching in the coagulant bath. Hollow fiber membranes are prepared by extruding a polymer solution through an annular spinneret and a bore fluid in the annulus center. The fabrication of asymmetric membranes is a complicated process, which involves the thermodynamics and the rheological properties of the dope solutions, the formation of nascent membranes and solidification/verification of asymmetric membranes [1–10]. The morphology of asymmetric membranes is mainly dominated by the thermodynamics of the casting solution and the kinetics of ⁎ Corresponding author at: National Engineering Research Center of Membrane Technology, Dalian Institute of Chemical Physics, Chinese Academy of Sciences, 457 Zhongshan Road, Dalian 116023, China. Fax: + 86 0411 84379613. E-mail address: [email protected] (J. Ren). 0011-9164/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.desal.2011.10.024

transport process [1–5,10–14]. The thermodynamics of the casting solution is related to the phase equilibrium between the components of the casting solution, while the kinetics of transport process can be described by the mutual diffusion and transport of the components. Delay time and gelation time are two macroscopic time scales in the phase separation by immersion precipitation [1,14–15]. According to delay time, demixing can be streamed into instantaneous demixing and delayed demixing. Instantaneous demixing occurs when the phase separation begins immediately after immersion. But for delayed demixing, the composition of the entire solution remains in the homogeneous region of the phase diagram for a certain period of time. Usually, a finger-like structure is formed in instantaneous demixing process, but a sponge-like structure in delayed demixing process [1,14]. The nascent asymmetric membrane morphology is gradually fixed according to the subsequent solidification within the gelation time [16–17]. For the asymmetric membrane morphology prepared by phaseinversion process, sponge-like and macrovoid (especially finger-like structure) structures are usually observed. The macrovoids inside the membrane usually weaken the membrane mechanical strength, induce the presence of some defects and deteriorate membrane performance. The formation of macrovoids can be described with different mechanisms for different membrane fabricating conditions [5,14,18–27]. The approaching ratio [5,24–25] of the casting solution and the approaching coagulant ratio [26–27] of the coagulant bath,

J. Ren, Z. Li / Desalination 285 (2012) 336–344

2. Experimental

used as coagulation bath for flat sheet membranes. All the reagents are used as received. 2.2. Fabrication of P84 flat sheet membranes P84 co-polyimide is dried at 110 °C in a vacuum oven for 2 days before use. The homogeneous casting solutions are prepared by dissolving P84 co-polyimide in NMP/GBL mixtures with different wGBL, where wGBL is defined as the weight ratio of GBL in the GBL/NMP mixture: wGBL ¼

mGBL mGBL þ mNMP

ð1Þ

mGBL and mNMP are the weight of GBL and NMP in the GBL/NMP mixture. The homogeneous casting solutions are usually degassed before use. The viscosity of the casting solutions is tested with RheoStress 300 rheometer (HAAKE Instruments Inc.) at 25 °C. The flat sheet membranes are cast on glass plates with a casting knife, and the membrane thickness is controlled by adjusting the gap between the glass plate and the casting knife. The prepared flat sheet membranes are immersed in a water bath at 25 °C and then they are washed with water in the following 72 h and dried for another 24 h at room temperature. 2.3. Fabrication of P84 hollow fiber membranes P84 co-polyimide hollow fiber membranes are fabricated by a dryjet wet spinning or a wet spinning process. The dried polyimide is dissolved in the NMP/GBL mixtures (wGBL = 0.25, 0.75) to form two 18 wt.% P84 dope solutions, and then stirred for about 2 days at room temperature. The prepared dope solutions are degassed under vacuum for 3 h and then kept for 1–2 days for use. The spinneret has an outer diameter of 1.0 mm, inner diameter of 0.44 mm, and annual length of 1.0 mm. The coagulation bath is tap water at room temperature. The spun hollow fibers are taken up by a roller and stored in a water bath to remove residual solvent for at least 3 days. The membrane thickness and the outer diameter of hollow fiber membranes are described in Eq. (2) [33]:   δ D−d 1 1 ¼ ¼ 1− pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi D 2D 2 1 þ εω

ð2Þ

where δ, D, d, and ω are the membrane thickness, outer diameter, inner diameter and the volume ratio of the dope solution to the bore fluid, respectively. The revised factor (ε) reflects the shrinkage

Intensity(I)

which reflect the thermodynamic property of the casting solution and kinetic exchanging characteristics of solvent and non-solvent in precipitation process respectively, also strongly influence the membrane morphology. Recently, membranes with different thickness were prepared to investigate the thickness dependence of macrovoids and to explore the origin of macrovoids [19–23]. A critical structure-transition thickness Lc is observed, which describes the transition of membrane morphology from sponge-like to finger-like structure with increasing membrane thickness [21–23]. For hollow fiber spinning, it is very important to determine and optimize the key spinning parameters [10,28–29]. Mckelvey et al. [10] used dominant process parameters (e.g., spinneret, dope extrusion rate, drawing ratio, air gap distance, bore fluid extrusion rate, solvent concentration in bore fluid, vitrification kinetics and spin line guides etc.) to control hollow fiber macroscopic properties and to determine the optimal separation properties. Generally, the dope rheology in the spinneret, the die swell and elongation in the air gap, phase inversion and solidification of nascent hollow fiber membranes in the coagulation bath are the core of the spinning process [28]. The rheological properties of the dope solution in spinning process mainly involve shear flow induced by shear stress within the spinneret and elongation flow induced by elongational stress (gravity and drawing force) in the air gap, which influence chain conformation and induce molecular orientation in the skin layer during the phase-inversion process of hollow fiber formation [6,28,30–35]. Shilton et al. [30] found that both gas permeability and selectivity of hollow fiber membranes increased with dope extrusion rate due to enhanced orientation of the membrane skin. However, according to Chung et al. [6,31–32], a lower gas flux and a higher selectivity could be obtained for hollow fiber membranes spun at high shear rate. Ren [28,33] also found that the increase of shear rate could result in ultrafiltration membranes with low permeation water flux and small molecular weight cut-off (MWCO). The elongational stress of the dope solution also strongly affects the performance of hollow fiber membranes. For wet-spinning process, Paul [34] founded that the molecular orientation was correlated well with the draw-ratio during fiber formation. Ekiner et al. [35] observed that an increase in elongation strain resulted in denser morphology and higher gas selectivity. According to Ren [8], the elongational stress induced by the take-up unit could lower both gas permeance and selectivity. The reduction in permeance is mainly due to the elongational induced molecular orientation, while the decrease in selectivity is probably due to the elongational induced defects. In this paper, BTDA-TDI/MDI co-polyimide (P84) was used to prepare the flat sheet and hollow fiber membranes due to its excellent chemical resistance and thermal stability [36–37]. The effects of membrane thickness and GBL weight ratio (wGBL) on the morphology of asymmetric flat sheet membranes prepared with P84/N-methyl-2pyrrolidone (NMP)/γ-butyrolactone (GBL) casting solutions are investigated. Furthermore, the influence of wGBL on morphology and performance of P84 co-polyimide hollow fiber membranes is discussed. Also, the influence of shear rate of the dope solution within the spinneret on the morphology and performance of the hollow fiber membranes prepared is discussed according to the rejection characteristics of dextran molecules.

337

If

Ip 2.1. Membrane material and chemicals The membrane material, BTDA-TDI/MDI co-polyimide (P84) powder is a commercial polymer produced by Lenzing. NMP is used as a solvent and GBL as a poor solvent, which are supplied by MERCK. Dextrans (C6H10O5)n with different molecular weights (1500 to 200,000) from Fluka and Sigma are used to test the rejection curves and MWCO of hollow fiber membranes. Milli-Q ultrapure water is

1

10

100

1000

10000 100000 1000000

Molecular weight of Dextran Fig. 1. The dextran molecular weight distribution in feed and permeate solutions for one hollow fiber membrane sample (●: feed solution; ○: permeate solution).

338

J. Ren, Z. Li / Desalination 285 (2012) 336–344

Table 1 The solubility parameters of NMP, GBL, H2O and their difference to P84. Solubility parameter (MPa)0.5

P84 NMP GBL H2O

δd

δp

δh

Δδ

20.4 18.0 19.0 15.6

20.4 12.3 16.5 16.0

10.3 7.2 7.3 42.3

3.9 5.3 32.6

where, Cf and Cp are the dextran concentration in the feed and permeate solutions, which are represented by the GPC spectrum intensity If and Ip of dextran in feed and permeate solutions. Fig. 1 shows the dextran molecular weight distribution intensity (If, Ip) in feed and permeate solutions for one hollow fiber membrane sample. So the rejection coefficient of hollow fiber membranes for different dextran molecular weights can be calculated. The MWCO of hollow fiber membranes is defined as the molecular weight at 90% rejection. 2.5. Membrane pore size distribution

degree of hollow fiber membranes due to the exchange of solvent and non-solvent in the coagulant bath. 2.4. Performance of hollow fiber membranes

f ðdÞ ¼

The testing modules are prepared with 8–10 fibers in glass tubes and their effective length is 25 cm. The pure water flux of hollow fiber membranes is tested with Milli-Q ultrapure at 1 atm. The membrane module is run for 1 h before sampling. The pure water flux of the hollow fiber membranes is calculated according to: J¼

Q ΔP ⋅A

ð3Þ

where, J is the pure water flux of the hollow fiber membrane (L. m − 2.atm − 1.hr − 1); Q is the water flux reading (L.hr − 1); ΔP is the pressure difference between the feed side and the permeation side of the membrane (atm); A is the effective membrane surface area (m 2). In order to test the rejection curves and MWCO of hollow fiber membranes, a 1500 ppm dilute solution with 1500 to 200,000 Da broad dextran molecular weight distribution is used. The dextran molecular weight distribution in the feed and permeate solutions is measured by gel permeation chromatography (GPC). The rejection of hollow fiber membranes is calculated according to R¼

The two-parameter log-normal distribution function for ultrafiltration membranes is written as follows [38]:

C f −C p I f −Ip ¼ Cf If

ð4Þ

"   # 1 lnðd=D Þ 2 pffiffiffiffiffiffi exp − 2 lnðσ Þ lnðσ Þd 2π 1

ð5Þ

where, d, D⁎, and σ are the pore diameter, the geometric mean diameter and the geometric standard deviation, respectively. The geometric mean diameter D⁎ and the geometric standard deviation σ is determined by Eq. (6) [38]: D ¼ ds @R ¼ 50% d @R ¼ 84:13% σ¼ s ds @R ¼ 50%

ð6Þ

where ds is the solute diameter. The correlation between the radius (ds, in nm) and the molecular weight (MW, in g/mol) of dextran was as follows [28,39]:

0:46

ds ¼ 0:066ðMW Þ

ð7Þ

2.6. Scanning electron microscopy The dried flat sheet and hollow fiber membrane samples are broken in liquid nitrogen and then dried at 80 °C in a vacuum oven for one day. The membrane samples are sputtered with a thin layer of gold using a SPI-Module sputter coater. The cross-section and outer surface of

wGBL=0

wGBL=0.25

wGBL=0.50

wGBL=0.56

wGBL=0.63

wGBL=0.69

wGBL=0.75

wGBL=1

Fig. 2. Influence of GBL weight ratio (wGBL) on the cross-section morphology of flat sheet membranes prepared with 20 wt.% P84/NMP/GBL casting solutions (coagulant: water; casting temperature: 25 °C).

J. Ren, Z. Li / Desalination 285 (2012) 336–344

339

20000

viscosity (mPa.s)

4.2 μm 15000

7.3 μ m

10000

12.3 μm

7.7 μm

5000

11 .2

64 . 3μ

μm

0

0

0.2

0.4

0.6

0.8

1

5 .8

m

μ

GBL weight ratio (wGBL) in solvent mixture

m

17

asymmetric membranes are examined using a Jeol JSM-5310LV Scanning Electron Microscope (SEM).

.2

m

μ

Fig. 3. Influence of GBL weight ratio (wGBL) on the viscosity of 20 wt.% P84/NMP/GBL casting solutions (shear rate: 10 s− 1; testing temperature: 25 °C).

Fig. 5. The cross-section morphology of different thickness flat sheet membranes prepared with 20 wt.% P84/NMP/GBL casting solutions by wet phase-inversion (coagulant: water; casting temperature: 25 °C; wGBL =0).

3. Results and discussion

The solubility parameters of P84, NMP, GBL, H2O are shown in Table 1. According to the solubility parameter difference (Δδ), NMP is a strong solvent and GBL is a relative weak solvent for P84 polymer. Water is a strong coagulant for P84/NMP/GBL casting solutions. Compared with pure NMP solvent, the presence of GBL will decrease the solvent power of NMP/GBL mixture for P84 polymer and influence the morphology of flat sheet membranes. Fig. 2 shows the influence of wGBL on the cross-section morphology of flat sheet membranes prepared with 20 wt.% P84/NMP/GBL casting solutions. With the increase of wGBL, the transition of membrane morphology from fingerlike to sponge-like structure occurs. For the flat sheet membranes prepared with low wGBL (such as wGBL = 0, 0.25, 0.5), a finger-like structure is obtained. With the increase of wGBL (such as wGBL = 0.56, 0.63), the quantity of macrovoids per unit area decreases and the formation of finger-like macrovoids is suppressed partially. Meanwhile, some sponge-like structure is formed underneath the membrane surface. For the flat sheet membranes prepared with high wGBL (such as wGBL = 0.69, 0.75, 1.0), the finger-like macrovoids cannot

100

Sponge-like structure

Lgs(μm)

80 60 40 20 0 0

0.2

0.4

0.6

w

0.8

1

GBL weight ratio (wGBL) in solvent mixture Fig. 4. Influence of GBL weight ratio (wGBL) on the general sponge-like structure thickness (Lgs) of flat sheet membranes prepared with 20 wt.% P84/NMP/GBL casting solutions at different GBL weight ratios (coagulant: water; casting temperature: 25 °C).

be formed and a fully sponge-like structure is obtained. It seems like that the critical weight ratio of GBL (w⁎) for flat sheet membranes prepared with 20 wt.% P84/NMP/GBL casting solutions is about 0.69, which reflects the transition of membrane morphology from fingerlike to sponge-like structure [23]. It is obvious that wGBL of the casting solutions changes the thermodynamics of casting solutions, which in turn alters the membrane morphology. From Fig. 3, the viscosities of 20 wt.% P84/NMP/GBL casting solutions increase with wGBL, as the including of weak solvent GBL usually leads into a tighter conformation of the polymer molecule and finally increases the viscosity of the casting solutions [40]. The viscosity increase of the casting solution usually lowers diffusion rates of solvents and nonsolvents in the casting solution, and restraints the tendency to macrovoid formation [1,23]. For asymmetric membranes, the general sponge-like structure thickness (Lgs) [22] underneath the membrane surface, is also an important parameter for membrane morphology. The influence of wGBL on the general sponge-like structure thickness (Lgs) of flat sheet membranes prepared with 20 wt.% P84/NMP/GBL casting solutions is shown in Fig. 4, the general sponge-like structure thickness (Lgs) increases with wGBL. And a fully sponge-like structure is obtained as the wGBL is greater than the critical weight ratio of GBL (w⁎ = 0.69).

Thickness of the sponge-like structure(μm)

3.1. Effect of GBL weight ratio on morphology of flat sheet membranes

15 LC

12

I

II

III

9 6 3 Lgs

0 0

LC

30

60

90

120

150

Entire membrane thickness(μm) Fig. 6. Influence of membrane thickness on the structure-transition parameters (Lc, Lgs) of flat sheet membranes prepared with 20 wt.% P84/NMP/GBL casting solutions by wet phase-inversion (coagulant: water; casting temperature: 25 °C; wGBL = 0).

340

J. Ren, Z. Li / Desalination 285 (2012) 336–344 Table 2 Spinning conditions of Batches I and II.

2 9.

11.5 μm

2 μm

Spinning dope composition (P84/NMP/GBL)

Spinning dope temperature Bore temperature Bore fluid

89.3 μm 19.0 μm

20.2 μm

Fig. 7. The cross-section morphology of different thickness flat sheet membranes prepared with 20 wt.% P84/NMP/GBL casting solutions by wet phase-inversion (coagulant: water; casting temperature: 25 °C; wGBL =0.50).

3.2. The thickness dependence of asymmetric membrane morphology

Thickness of sponge-like structure(μm)

Recently, in order to investigate the morphology transition of asymmetric membranes, a new method based on the thickness dependence of macrovoid evolution has been reported [20–21]. In this paper, the influence of membrane thickness on the morphology of flat sheet membranes prepared with 20 wt.% P84/NMP casting solution is investigated. As shown in Fig. 5, for the thin membrane thickness such as 4.2 and 12.3 μm, a fully sponge-like structure is formed. For the flat sheet membranes with thickness of 17.2 μm, some macrovoids appear underneath the membrane surface, which induces the sharp decrease of sponge-like structure thickness. A completely finger-like structure is formed for the flat sheet membranes with thickness of 64.3 μm. It is clear that the increase of the membrane thickness induces the transition of membrane morphology from sponge-like to finger-like structure. The relationship between the thickness of sponge-like structure and entire membrane thickness is shown in Fig. 6. From Fig. 6, the critical structure-transition thickness Lc for the flat sheet membranes prepared with 20 wt.% P84/NMP casting solutions (wGBL = 0) is about 12.3 μm. In region I, a completely sponge-like structure is obtained with the membrane thickness less than Lc. In region II, as the membrane

Air gap External coagulant bath External coagulant temperature Power-law equation

Batch I

Batch II

P84:18 wt.%; NMP/GBL:75/25 (wt./wt.); w GBL = 0.25 25 °C

P84: 18 wt.%; NMP/GBL:25/75 (wt./wt.); w GBL = 0.75 The same

25 °C H2O/NMP:15/85 (wt./wt.) 0–5 cm Tap water Room temperature

The same H2O/NMP:15/85 (wt./wt.) 0–3 cm The same The same

σ = 2.15γ0.977

σ = 3.57γ0.979

thickness is slightly greater than Lc, the transition of membrane morphology from sponge-like to finger-like structure occurs, and the thickness of sponge-like structure decreases quickly. In region III, the entire membrane thickness is relatively larger, a finger-like structure is obtained. Meanwhile, the general sponge-like structure thickness (Lgs) underneath the membrane surface is about 2.3 μm, which is not affected by the membrane thickness. For the flat sheet membranes prepared with 20 wt.% P84/NMP/GBL (wGBL = 0.50), the influence of membrane thickness on the morphology of flat sheet membranes is shown in Figs. 7 and 8. From Fig. 8, the critical structure-transition thickness Lc of flat sheet membranes prepared with 20 wt.% P84/NMP/GBL (wGBL = 0.50) is about 33 μm, which is greater than that for 20 wt.% P84/NMP (wGBL = 0). That is to say, the presence of GBL in the casting solution changes the exchange process of NMP, GBL and H2O, and retards the transition of membrane morphology from sponge-like to finger-like structure. Meanwhile, the general sponge-like structure thickness (Lgs) is also increased to 19 μm in region III. It is clear that the increasing wGBL induces the increase of critical structure-transition thickness Lc and general sponge-like structure thickness (Lgs). The thickness dependence of asymmetric membrane morphology is mainly due to the change of boundary condition of the casting solution

50

40

I

II

III

LC

A (cross-section 350)

30

A (cross-section 1000)

20

10 Lgs

0 0

LC

50

100

150

200

Entire membrane thickness(μm) Fig. 8. Influence of membrane thickness on the structure-transition parameters (Lc, Lgs) of flat sheet membranes prepared with 20 wt.% P84/NMP/GBL casting solutions by wet phase-inversion (coagulant: water; casting temperature: 25 °C; wGBL = 0.50).

B (cross-section 500)

B (cross-section 1000)

Fig. 9. The membrane morphology of hollow fiber membranes spun with Batch I and II at the same dope flow rate. (A: Batch I; B: Batch II; Dope flow rate in spinneret: 6.4 ml/ min; air gap: 1 cm).

J. Ren, Z. Li / Desalination 285 (2012) 336–344

341

Table 3 Spinning parameters, performance of hollow fiber membranes at different shear rates in the spinneret for Batch I. Module

Batch Batch Batch Batch

I-1 I-2 I-3 I-4

Spinning condition Dope shear rate within spinneret (s− 1)

Dope flow rate (ml/min)

Bore flow rate (ml/min)

Ratio of dope solution to bore flow rate (ω)

1632 3266 4897 6529

3.2 6.4 9.6 12.8

2 4 6 8

1.6 1.6 1.6 1.6

in the coagulant bath [23,41]. When the membrane thickness L is less than Lc, the boundary condition of the casting solution at the thickness L is described as: ∂ðΦi Þ ¼ 0; l ¼ L ∂l

ð8Þ

where ϕi is the volume fraction of component i in the casting solution, l the thickness coordinate. The transport of nonsolvent is retarded at the thickness L, the concentration profile of nonsolvent inside the casting solution becomes higher. As a result, a fully sponge-like structure is formed before the macrovoids appear. When the membrane thickness L is greater than Lc, the boundary condition of the casting solution at the thickness L is assumed to be the same as L = ∞: ∂ðΦi Þ ¼ 0; l ¼ ∞: ∂l

ð9Þ

The membrane thickness cannot influence the concentration profiles of solvent and nonsolvent inside the membrane. When the macrovoids are formed, the sponge-like structure thickness underneath the membrane surface does not change. 3.3. Effect of the shear rate on the performance of hollow fiber membranes According to above analysis, the GBL weight ratio (wGBL) strongly influences the membrane morphology of flat sheet membranes. In this paper, two 18 wt.% P84/NMP/GBL dope solutions with different GBL weight ratios (wGBL = 0.25, 0.75) are prepared for the spinning of hollow fiber membranes, respectively. The formulas of Batches I and II are listed in Table 2. The relationship between the shear stress and the shear rate (from 0.1 s − 1 to 100 s− 1) for the two dope solutions at 25 °C is described as: BatchI : σ ¼ 2:15γ

0:979

BatchII : σ ¼ 3:57γ

ð10Þ

0:979

ð11Þ

where, σ is the shear stress (Pa) and γ is the shear rate (s− 1). The two dope solutions behave like a shear thinning power-law fluid, which are

D (μm)

d (μm)

Flux (L.m− 2. atm− 1.hr− 1)

Rejection (10,000 Da)

659 847 971 1018

429 547 618 671

51.6 56.9 107.5 90.3

49% 52% 54% 61%

assumed to be applicable at higher shear rates. The shear rate of the dope solutions inside the annular region within the spinneret is calculated as described elsewhere [42]. For the spinning of hollow fiber membranes with Batch I (18 wt.% P84; wGBL = 0.25) and Batch II (18 wt.% P84; wGBL = 0.75), the volume ratio of dope solution to bore fluid is controlled at 1.6 (vol./vol.) and the spinning speed is kept at the free-falling velocity of the nascent hollow fiber membranes in the water bath. The flow rate of dope solution in the spinneret increases from 3.2 to 12.8 ml/min to yield different shear rate within the spinneret. The membrane morphology of hollow fiber membranes spun with Batches I and II at the same dope flow rate is shown in Fig. 9. For Batch I (18 wt.%P84; wGBL = 0.25), hollow fiber membranes with finger-like structure are obtained, but for Batch II (18 wt.% P84; wGBL = 0.75), sponge-like hollow fiber membranes are formed. As wGBL increases from 0.25 to 0.75, the transition of membrane morphology from finger-like to sponge-like structure also occurs for hollow fiber membranes. The spinning parameters, performance and dimensions of the hollow fiber membranes spun at different shear rates within the spinneret using Batch I and II dopes are listed in Tables 3 and 4, respectively. The outer diameter (D) and inner diameter (d) of the hollow fibers increases with the shear rate of the dope solution within the spinneret. The relationship between the membrane thickness (δ) and hollow fiber diameter (D) for different hollow fiber membranes spun with same volume ratio of the dope solution to the bore fluid (ω = 1.6), different shear rates and different air gaps (0–5 cm for Batch I and 0–3 cm for Batch II) is shown in Fig.10, which can be described by Eq. (2). From Fig. 10, the revised factors (ε) are regressed to be 0.89 and 0.70 for the hollow fiber membranes spun from Batch I and Batch II, respectively. The revised factor (ε), which reflects the shrinkage degree of nascent hollow fiber membranes, is less than 1, as the exchange of solvent and non-solvent between dope solution and bore fluid/coagulant bath usually favors the outflow of solvent in the nascent hollow fiber membranes. For hollow fiber membranes spun with Batch I (18 wt.% P84; wGBL = 0.25), the high revised factor (ε = 0.89) reflects the relative weak shrinkage degree of nascent hollow fiber membranes, which means a certain amount of solvent is still present inside the nascent hollow fiber membranes, hence relatively loose finger-like macrovoids are formed. Compared with Batch I, the including of more GBL in Batch II (18 wt.% P84; wGBL = 0.75) delays the phase-inversion process and

Table 4 Spinning parameters, performance of hollow fiber membranes at different shear rates in the spinneret for Batch II. Module

Batch Batch Batch Batch

II-1 II-2 II-3 II-4

Spinning condition Dope shear rate within spinneret (s− 1)

Dope flow rate (ml/min)

Bore flow rate (ml/min)

Ratio of dope solution to bore flow rate (ω)

1631 3263 4894 6525

3.2 6.4 9.6 12.8

2 4 6 8

1.6 1.6 1.6 1.6

D (μm)

d (μm)

Flux (L.m− 2. atm− 1.hr− 1)

Rejection (10,000 Da)

612 741 847 944

418 500 582 664

28.0 19.9 13.9 7.6

81% 89% 94% 94%

J. Ren, Z. Li / Desalination 285 (2012) 336–344

250

Flux(L/m2.h.atm)

δ(μm)

200 150 100 50 0

0

200

400

600

800

1000

80

20000

60

15000

40

10000

20

5000

0

1200

0

D(μm)

2000

4000

6000

8000

MWCO

342

0 10000

Shear rate (s-1)

Fig. 10. The relationship between membrane thickness and outer diameter (D) of hollow fiber membranes spun for two different dope solutions (● Batch I, ε = 0.89; ○ Batch II, ε = 0.70; ω = 1.6).

induces outflow of more GBL/NMP before the occurence of phaseinversion, resulting in lower revised factor (ε = 0.70). Because of the strong shrinkage of nascent hollow fiber membranes, a relatively tight sponge-like structure is obtained. The big difference of revised factor between Batches I and II is reflected clearly from the membrane morphology. Fig. 11 and Fig. 12 show the influence of the shear rate on the performance of hollow fiber membranes spun with Batch I (18 wt.% P84; wGBL = 0.25) and Batch II (18 wt.% P84; wGBL = 0.75). Due to strong solvent NMP of the dope solution and strong coagulant water, the demixing process occurs more quickly for hollow fiber membranes spun from Batch I (18 wt.% P84; wGBL = 0.25) than that from Batch II (18 wt.% P84; wGBL = 0.75), resulting in a thinner skin layer and finger-like structure. Compared with Batch II (18 wt.% P84; wGBL = 0.75), the hollow fiber membranes spun from Batch I (18 wt.% P84; wGBL = 0.25) have a high permeation flux and a large MWCO. With an increase in the shear rate of the dope solution, the permeation flux and rejection increases, resulting in a low MWCO. The increase of permeation flux due to the increase of shear rate was also investigated by Shilton et al. [30]. They found that the both gas permeability and selectivity of polysulfone and polyacrylonitrile hollow fiber membranes increased with dope extrusion rate due to the enhanced orientation of the membrane skin. From Fig. 12, for the hollow fiber membranes spun from Batch II (18 wt.% P84; wGBL =0.75), both the permeation flux and MWCO decreases with increasing shear rate due to the shear induced

Fig. 12. The influence of shear rate of dope solution within the spinneret on the performance of hollow fiber membranes spun with Batch II.

molecular orientation as describe by Ren et al. [33]. The rejection curves of the hollow fiber membranes spun at different shear rates for Batches I and II are shown in Fig. 13. Compared with Batch I, the rejection curves of the hollow fiber membranes spun with Batch II shift toward left, which implies that hollow fiber membranes spun with Batch II could have a better rejection property. For the hollow fiber membranes spun with Batch I and Batch II, the rejection curves shift towards left with the increase of shear rate, indicating that the MWCO can be decreased due to the shear induced molecular orientation. The shear rate of the dope solution within the spinneret also influences the membrane pore size distribution. The influence of shear rate on the pore size distribution of hollow fiber membranes spun from Batch I and Batch II is shown in Fig. 14. Compared with Batch I, the pore size distribution curves of hollow fiber membranes spun with Batch II are relatively narrow, and their geometric mean diameter is also relatively less than that for Batch I. That is to say, with the increase of GBL concentration in the dope solution, the geometric mean diameter of the spun hollow fiber membranes gets smaller and their pore size distribution curve becomes sharper. From Fig. 14, with the increase of shear rate, the pore size distribution curves for hollow fiber membranes spun from Batches I and II shifted towards left, resulting in sharp peaks and slightly low MWCO. The rheological characteristics of the dope solution within the spinneret affect the membrane pore size distribution.

1 300

50000 0.8 40000 0.6 30000

150 20000

R

200

MWCO

Flux(L/m2.h.atm)

250

0.4

100 0.2 10000

50 0 0

2000

4000

6000

8000

0 10000

Shear rate (s-1) Fig. 11. The influence of shear rate of dope solution within the spinneret on the performance of hollow fiber membranes spun with Batch I.

0 100

1000

10000

100000

Molecular weight of Dextran Fig. 13. Rejection curves of hollow fiber membranes spun with Batch I and Batch II with different shear rates (Batch I: ○ 1632 s− 1; Δ 3266 s− 1; ◊ 4897 s− 1; □ 6529 s− 1; Batch II: ■ 1631 s− 1; ▲ 3263 s− 1; ♦ 4894 s− 1; ■ 6525 s− 1).

J. Ren, Z. Li / Desalination 285 (2012) 336–344

0.5

343

1

Batch II

6525s-1 4894s-1 3263s-1

0.4

0.8

1631s-1 6529s-1 4897s-1

Batch I

0.6

3266s-1 1632s-1

R

f(d)

0.3 0.2

0.4

0.1 0

0.2 0

5

10

15

0 100

Pore diameter (nm)

1000

10000

100000

1000000

Molecular weight of Dextran Fig. 14. Influence of shear rate on the pore size distribution of hollow fiber membranes spun with Batch I and Batch II.

3.4. Effect of the air gap on the hollow fiber performance In this paper, the spun hollow fiber membranes are collected with free-falling speed. As bore fluid is 85 wt.% NMP, the extruded polymer solution in the air gap cannot be vitrified immediately, thus it will be elongated by its gravity. The influence of air gap on the performance of the hollow fiber membranes spun with Batch II is shown in Figs. 15 and 16. From Fig. 15, when the air gap decreases from 3 cm to 0.5 cm, the permeation flux and MWCO of the spun hollow fiber membranes almost do not change. Compared with the dry-jet wet spinning, the permeation flux and MWCO of the hollow fiber membranes spun in the wet spinning process increases significantly. Without the air gap, the extruded dope solution immerses into the strong water coagulant bath directly, the demixing process occurs instantaneously. The accumulated stress induced by die-swell phenomena cannot be released quickly, which causes some defects on the membrane surface. Therefore, the hollow fiber membranes spun in wet spinning process exhibit a very poor rejection (shown in Fig.16).

Fig. 16. Rejection curves of hollow fiber membranes spun with Batch II at different air gap (shear rate: 4894 s− 1; air gap: ○ 0 cm; □ 0.5 cm; Δ 1 cm; ◊ 3 cm).

1) For 20 wt.% P84/NMP/GBL casting solutions, the transition of membrane morphology from finger-like to sponge-like structure occurs at the critical weight ratio of GBL (w⁎ = 0.69). 2) With the increase of membrane thickness, the transition of membrane morphology from sponge-like to finger-like structure occurs at critical structure-transition thickness Lc, and the increase of wGBL induces the increase of Lc and the general sponge-like structure thickness (Lgs). 3) Compared with the dope solution with wGBL = 0.75, the hollow fiber membranes spun with wGBL = 0.25 present a high permeation flux and a large MWCO. With the increase of shear rate, the rejection curves shift towards left, resulting in slightly low MWCO. 4) For the hollow fiber membrane spun with Batch II, the permeation flux and MWCO almost do not change with the air gap. But for the wet spinning process, a relatively high permeation flux and a large MWCO are obtained.

References 4. Conclusion For asymmetric membranes prepared with P84/NMP/GBL casting solutions, the transition of membrane morphology is investigated with flat sheet membranes. The morphology and performance of spun hollow fiber membranes are observed according to the thermodynamic and rheological properties of the dope solutions.

50

1000000

100000

30 10000 20

MWCO

Flux (L/m2.h.atm)

40

1000

10 0 0

1

2

3

4

100

Air gap (cm) Fig. 15. Influence of air gap on the permeation flux and MWCO of hollow fiber membranes spun with Batch II (shear rate: 4894 s− 1).

[1] P. van de Witte, P.J. Dijkstra, J.W.A. van den Berg, J. Feijen, Review, phase separation processes in polymer solutions in relation to membrane formation, J. Membr. Sci. 117 (1996) 1. [2] A.J. Reuvers, J.W.A. van den Berg, C.A. Smolders, Formation of membranes by means of immersion precipitation. Part I: a model to describe mass transfer during immersion precipitation, J. Membr. Sci. 34 (1987) 45. [3] F.W. Altena, C.A. Smolders, Calculation of liquid–liquid phase separation in a ternary system of a polymer in a mixture of a solvent and a nonsolvent, Macromolecules 15 (1982) 1491. [4] A.J. Reuvers, C.A. Smolders, Formation of membranes by means of immersion precipitation. Part II: the mechanism of formation of membranes prepared from the system cellulose acetate–acetone–water, J. Membr. Sci. 34 (1987) 67. [5] J. Ren, Z. Li, F.S. Wong, Membrane structure control of BTDA-TDI/MDI (P84) co-polyimide asymmetric membranes by wet-phase inversion process, J. Membr. Sci. 241 (2004) 305. [6] T.S. Chung, W.H. Lin, R.H. Vora, The effect of shear rates on gas separation performance of 6FDA-durene polyimide hollow fibers, J. Membr. Sci. 167 (2000) 55. [7] A. Idris, M.Y. Noordin, A.F. Ismail, S.J. Shilton, Study of shear rate influence on the performance of cellulose acetate reverse osmosis hollow fiber membranes, J. Membr. Sci. 202 (2002) 205. [8] J. Ren, T.S. Chung, D. Li, R. Wang, Y. Liu, Development of asymmetric 6FDA-2,6DAT hollow fiber membranes for CO2/CH4 separation 1. The influence of dope composition and rheology on membrane morphology and separation performance, J. Membr. Sci. 207 (2002) 227. [9] T.S. Chung, X. Hu, Effects of air gap distance on the morphology and thermal properties of polyethersulfone hollow fiber, J. Appl. Polym. Sci. 66 (1997) 1067. [10] S.A. Mckelvey, D.T. Clausi, W.J. Koros, A guide to establishing hollow fiber macroscopic properties for membrane applications, J. Membr. Sci. 124 (1997) 223. [11] L. Yilmaz, A.J. Mchung, Analysis of nonsolvent-solvent-polymer phase diagrams and their relevance to membrane formation modeling, J. Appl. Polym. Sci. 31 (1986) 997. [12] C. Cohen, G.B. Tanny, S. Prager, Diffusion-controlled formation of porous structures in ternary polymer systems, J. Polym. Sci., Part B. Polym. Phys. Ed. 17 (1979) 477. [13] A.J. McHugh, L. Yilmaz, The diffusion equation for polymer membrane formation in ternary systems, J. Polym. Sci., Part B: Polym. Phys. Ed. 23 (1985) 1271.

344

J. Ren, Z. Li / Desalination 285 (2012) 336–344

[14] C.A. Smolders, A.J. Reuvers, R.M. Boom, I.M. Wienk, Microstructures in phaseinversion membranes. Part 1: formation of macrovoids, J. Membr. Sci. 73 (1992) 259. [15] M.H.V. Mulder, Basic Principals of Membrane Technology, Dordrecht, Kluwer Academic Press, 1991. [16] S.G. Li, T. van den Boomgaard, C.A. Smolders, H. Strathmann, Physical gelation of amorphous polymers in a mixture of solvent and nonsolvent, Macromolecules 29 (1996). [17] J. Arnauts, H. Berghmans, Amorphous thermoreversible gels of atactic polystyrene, Polym. Commun. 28 (1987) 66. [18] M.A. Frommer, F.M. Messalem, Mechanism of membrane formation. VI. Convective flows and large void formation during membrane precipitation, Ind. Eng. Chem. Prod. Res. Dev. 12 (1973) 328. [19] A. Conesa, T. Gumi, C. Palet, Membrane thickness and preparation temperature as key parameters for controlling the macrovoid structure of chiral activated membranes (CAM), J. Membr. Sci. 287 (2007) 29. [20] N. Vogrin, C. Stropnik, V. Musil, M. Brumen, The wet phase separation: the effect of cast solution thickness on the appearance of macrovoids in the membrane forming ternary cellulose acetate/acetone/water system, J. Membr. Sci. 207 (2002) 139. [21] D. Li, T.S. Chung, J. Ren, R. Wang, Thickness dependence of macrovoid evolution in wet phase-inversion asymmetric membranes, Ind. Eng. Chem. Res. 43 (2004) 1553. [22] J. Zhou, J. Ren, L. Lin, M. Deng, Morphology evolution of thickness-gradient membranes prepared by wet phase-inversion process, Sep. Pur. Technol. 63 (2008) 484. [23] J. Ren, J. Zhou, M. Deng, Morphology transition of asymmetric flat sheet and thickness-gradient membranes by wet phase-inversion process, Desalination 253 (2010) 1. [24] T. He, C. Jiang, Effects of nonsolvent additives on performance of poly(ether sulfone) microporous membranes, Membr. Sci. Technol. 18 (1998) 43. [25] Z. Li, C. Jiang, Investigation into the rheological properties of PES/NMP/nonsolvent membrane-forming systems, J. Appl. Polym. Sci. 82 (2001) 283. [26] L. Wang, Studies on gelation time and formation mechanism of support layer of PES membrane-making system, Master thesis, Dalian Institute of Chemical Physics, Chinese Academy of Sciences, China, 1999. [27] L. Wang, Z. Li, J. Ren, S.G. Li, C. Jiang, Preliminary studies on the gelation time of poly(ether sulfones) membrane-forming system with an elongation method, J. Membr. Sci. 275 (2006) 46.

[28] J. Ren, Z. Li, R. Wang, Effects of the thermodynamics and rheology of BTDA-TDI/ MDI co-polyimide (P84) dope solutions on the performance and morphology of hollow fiber UF membranes, J. Membr. Sci. 309 (2008) 196–208. [29] A. Ifris, M.Y. Noordin, A.F. Ismail, S.J. Shilton, Study of shear rate influence on the performance of cellulose acetate reverse osmosis hollow fiber membranes, J. Membr. Sci. 202 (2002) 205. [30] S.J. Shilton, G. Bell, J. Ferguson, The rheology of fibre spinning and the properties of hollow-fibre membranes for gas separation, Polym. 35 (1994) 5327. [31] T.S. Chung, S.K. Teoh, W.W.Y. Lau, M.P. Srinivasan, Effect of shear stress within the spinneret on hollow fiber: membrane morphology and separation performance, Ind. Eng. Chem. Res. 37 (1998) 3930. [32] T.S. Chung, J.J. Qin, J. Gu, Effect of shear rate within the spinneret on morphology, separation performance and mechanical properties of ultrafiltration polyethersulfone hollow fiber membranes, Chem. Eng. Sci. 55 (2000) 1077. [33] J. Ren, Z. Li, F.S. Wong, D. Li, Development of asymmetric BTDA-TDI/MDI (P84) co-polyimide hollow fiber membranes for ultrafiltration: the influence of shear rate and approaching ratio on membrane morphology and performance, J. Membr. Sci. 248 (2005) 177–188. [34] D.R. Paul, Spin orientation during acrylic fiber formation, J. Appl. Polym. Sci. 13 (1969) 817. [35] O.M. Ekiner, G. Vassilatos, Polyaramide hollow fibers for H2/CH4 separation II. Spinning and properties, J. Membr. Sci. 186 (2001) 71. [36] L.S. White, Polyimide membranes for hyperfiltration recovery of aromatic solvents, US patent 6180008, W. R. Grace & Co.-Conn., (2001). [37] L.S. White, Transport properties of a polyimide solvent resistant nanofiltration membrane, J. Membr. Sci. 205 (2002) 191. [38] D.B. Mosqueda-Jimenez, R.M. Narbaitz, T. Matsuura, G. Chowdhury, G. Pleizier, J.P. Santerre, Influence of processing conditions on the properties of ultrafiltration membranes, J. Membr. Sci. 231 (2004) 209. [39] P. Aimar, M. Meireles, V. Sanchez, A contribution to the translation of retention curves into pore size distributions for sieving membranes, J. Membr. Sci. 54 (1990) 321. [40] S. Chou, The preparation of polyetherimide membranes and characterization for gas separation, Master thesis, Dalian Institute of Chemical Physics, 2009. [41] J. zhou, J. Ren, L. Lin, M. Deng, Influence of γ-butyrolactone on the morphology of polyetherimide membranes, Membr. Sci. Technol. 29 (2009) 34. [42] J.J. Qin, R. Wang, T.S. Chung, Investigation of shear stress effect within a spinneret on flux, separation and thermomechanical properties of hollow fiber ultrafiltration membranes, J. Membr. Sci. 175 (2001) 197.