hydrophilic composite hollow fibers for direct contact membrane distillation

Chemical Engineering Science 137 (2015) 79–90

Contents lists available at ScienceDirect

Chemical Engineering Science journal homepage: www.elsevier.com/locate/ces

New insights into fabrication of hydrophobic/hydrophilic composite hollow fibers for direct contact membrane distillation Jiaming Zhu a, Lanying Jiang a,n, Takeshi Matsuura b a b

School of Metallurgy and Environment, Central South University, Changsha 410083, Hunan, China Department of Chemical and Biological Engineering, University of Ottawa, Ottawa, Ontario, Canada, K1N 6N5

H I G H L I G H T S







Hydrophobic/hydrophilic dual-layer hollow fibers were fabricated for DCMD. Outer surface of the outer layer was independent of the additive types. Hydrophilicity of the inner layer affected the DCMD performance. Pore wetting status was tentatively correlated with the membrane performance.

art ic l e i nf o

a b s t r a c t

Article history: Received 26 January 2015 Received in revised form 26 May 2015 Accepted 27 May 2015 Available online 6 June 2015

Dual-layer composite membrane is a new design for direct contact membrane distillation (DCMD) with membrane performance potentially superior to that of single layer porous membranes. Using dry–wet phase inversion technology, novel dual-layer hollow fiber membranes were fabricated in current research. The outer layer was made from polyvinylidene fluoride (PVDF) with polyvinylpyrolidone (PVP) or glycerol as non-solvent additive, while the inner layer consists of PVDF and polyvinyl alcohol (PVA) blend. The effect of the nonsolvent additive type in the outer layer and that of PVA/PVDF blending ratio in the inner layer on the morphological, mechanical and separation characteristics of the composite membranes was investigated. Membrane performance was further correlated to the physicochemical and morphological characteristics of the membranes. In particular, a thorough investigation of pore wetting in DCMD was attempted for the first time in this work, observing the cross-sectional distribution of EDX chlorine signals as an indication of the penetration of sodium chloride solution into the pore from the feed. & 2015 Elsevier Ltd. All rights reserved.

Keywords: Distillation Dual-layer composite membrane Phase inversion Membrane structure Wetting

1. Introduction Freshwater resource scarcity is a common problem for many countries. With growth in population and expansion of human activities, the problem of water contamination becomes more and more serious due to the huge volume of wastewater discharged and the complicated pollutants. Keeping the balance between freshwater supply and wastewater generation, which is essential to the conservation of water, is already beyond the capability of hydrological cycle involving only natural processes. Environmentally sound measures must be taken to prevent the situation from further deterioration.

n

Corresponding author. Tel.: þ 86 731 88716206; fax: þ86 731 88710171. E-mail address: [email protected] (L. Jiang).

http://dx.doi.org/10.1016/j.ces.2015.05.064 0009-2509/& 2015 Elsevier Ltd. All rights reserved.

Seawater desalination and wastewater reclamation have been incorporated into the memorandum of maintaining global water sustainability. Among the various separation technologies actually utilized or potentially applicable in the above systems, membrane distillation (MD) is commented positively by many researchers dealing with the tasks of removing inorganic salts from water stream (Alkhudhiri et al., 2012; Curcio and Drioli, 2005). Compared with non-membrane separations, the benefits of MD lie in the small foot-print and continuous operation. Relative to other membrane separations (i.e. reverse osmosis) it has the advantages of theoretically 100% ions rejection, high concentration factor, and mild operation temperature and/or pressures (Lawson and Lloyd, 1997). MD is classified into direct contact membrane distillation (DCMD), air gap membrane distillation (AGMD), sweeping gas membrane distillation (SGMD) and vacuum membrane distillation (VMD). DCMD using cold liquid to collect water vapor is the

80

J. Zhu et al. / Chemical Engineering Science 137 (2015) 79–90

simplest configuration. Like other membrane separations, DCMD relies on high flux and efficient rejection as important variables for reducing system investment and operation cost (Criscuoli et al., 2008). Due to the constraint by heat transfer, further improvement in flux might resort to using materials with higher thermal resistance. However, the hydrophobic materials, such as polyvinylidene fluoride (PVDF) and polypropylene (PP), which are already high in ranking of heat resistance, have been widely used in DCMD and the further breakthrough seems difficult (Han and Fina, 2011). On the other hand, fabrication of more porous or thinner membranes is desirable to increase the flux (Bonyadi et al., 2009; Lawson and Lloyd, 1997). Those membranes are, however, not mechanically strong enough for practical applications. Thus, in order to satisfy both requirements of high heat transfer resistance and high mass transport, the concept of hydrophilic/hydrophobic dual-layer membranes was proposed for DCMD (Bonyadi and Chung, 2007; Edwie et al., 2012; Khayet et al., 2006, 2005; Peng and May, 2011; Su et al., 2010). The thin hydrophobic layer facing the feed is not wetted by liquid water, while allowing the transport of vapor through the pores. The relatively thick porous hydrophilic layer is filled with water from the cold permeate side and does not contribute to vapor transfer resistance. At the same time, the hydrophilic layer serves as the mechanical support for the thin hydrophobic layer as well as can provide additional resistance for heat conduction through the membrane. One of the methods to fabricate such dual-layer membranes is the casting of polymer solutions which consists of a surface modifying macromolecule (SMM) and a host polymer, e.g. SMM was incorporated in hydrophilic polyetherimide (PEI) in Khayet et al.'s work for DCMD (Khayet et al., 2005). Khayet's dual-layer membranes showed fluxes at least as high as those of the commercial polytetrafluoroethylene (PTFE) membranes. Another method is relying on addition of the hydrophobic and hydrophilic additives into the outer and inner layers to modify or promote the hydrophobicity and hydrophility respectively. Su et al. fabricated the dual-layer membrane using the co-extrusion/phase inversion method and investigated the influence of changing the thermal conductivity of inner hydrophilic layer on the flux in DCMD (Su et al., 2010). Graphite particles and multiwall carbon nanotubes were used as the filler to make the inner layer more hydrophilic. It was found that the synergistic effect of graphite and multiwall carbon nanotubes has substantially enhanced the thermal conductivity of hydrophilic layer, and consequently the vapor flux. The most ideal design of the dual-layer membrane is such that the hydrophobic outer layer is kept dry to enable desalination, while the inner layer is fully wetted to enhance the flux. These twin requirements are satisfied when the materials for the inner and outer layers are properly chosen and their morphologies in terms of pore size, porosity, tortuosity etc. are properly adjusted (Curcio and Drioli, 2005; Song and Jiang, 2013). However, few researches have been so far performed to reveal material selection and structural design criteria of the dual-layer membrane for DCMD. In light of this, the current work will investigate: (1) the effect of spinning conditions, particularly the composition of the polymer and the additive in the spinning dope on the morphological and physiochemical characteristics of the dual-layer hollow fiber membranes, and (2) the relationship between the DCMD performance (flux and rejection) and above physicochemical and structural properties of the hollow fiber membranes. Hollow fibers are fabricated in the current work making the outer (shell side) and inner (lumen side) layers hydrophobic and hydrophilic, respectively. As the base polymer for both layers, PVDF is used to ensure good adhesion between the two dopes, one for the outer layer and the other for the inner layer to prevent delamination of two layers (Bonyadi and Chung, 2007). PVDF is widely used in the preparation of MD membranes due to its sufficiently high

hydrophobicity, acid resistance and other suitable chemical and physical properties (Drioli et al., 2011; Lovinger, 1982). Polyvinyl alcohol (PVA) was blended into PVDF to form an inner hydrophilic layer. It is reported that PVA has good dynamic miscibility with PVDF (Li et al., 2010). Glycerol and polyvinyl pyrrolidone (PVP) are the non-solvent additives for improving the porosity of the hydrophobic outer layer (Simone et al., 2010; Song and Jiang, 2013; Wang et al., 2000).

2. Experimental 2.1. Materials Poly (vinylidene fluoride) (PVDF 1300, Mw 350,000) was purchased from Kureha (Japan). Polyvinylalcohol (PVA T-350 Mw 800,000) supplied by Nippon Gohsei (Japan) was rinsed using pure water and dried at either 60 1C or 100 1C for 2 h prior to use. Polyvinyl pyrrolidone (PVP Mw 8000) was purchased from Aladdin Industrial Corporation (Shanghai, China). PVDF and PVP were dried at 100 1C in vacuum for 24 h before being used for dope preparation. N-methyl-2-pyrrolidone (NMP), methanol and hexane were purchased from Sinopharm Chemical Reagent (Shanghai, China) and used as received. 2.2. Preparation of spinning dopes To prepare the outer layer spinning dope, predetermined amounts of PVDF and PVP were added into NMP under vigorous stirring. The resultant mixture was heated to 80 1C and kept stirred for 12 h to form a homogenous solution. When glycerol was used as an additive instead of PVP, a predetermined amount of glycerol was added into NMP first, followed by addition of PVDF. To prepare the inner layer spinning dope, a predetermined amount of PVA was first dissolved in NMP at 95 1C. After complete dissolution of PVA, the temperature was lowered to 80 1C and the predetermined amount of PVDF powder was added into the PVA solution. After 12 h of continuous stirring, a clear tri-component solution was obtained. The dope compositions are summarized in Table 1. The solutions were stored in containers at 60 1C and degassed for 12 h before being subjected to hollow fiber spinning. 2.3. Hollow fiber spinning and module fabrication A triple-orifice-spinneret was used for dual-layer hollow fiber spinning. The polymer solutions were transferred to the spinneret by two gear pumps while the bore fluid was supplied to the center of the spinneret by a syringe pump from their respective storage Table 1 Compositions of dopes and coagulants for the dual-layer hollow fiber membranes preparation. ID

Outer layer dope Composition (wt%)

Inner layer dope

Bore fluid

Composition PVDF/PVA ratio (total polymer wt% 15, NMP wt% (wt%) 85)

External coagulant (wt%)

P5 PVDF: 12 P10 PVP: 10 P15 NMP: 78 P20

95/5 90/10 85/15 80/20

Water:80 Ethanol:20

Water:100

G5 PVDF:12 G10 Glycerol:10 G15 NMP:78 G20

95/5 90/10 85/15 80/10

Water: 80 Ethanol:20

Water:100

J. Zhu et al. / Chemical Engineering Science 137 (2015) 79–90

Table 2 Dual-layer membrane fabrication conditions. Spinning conditions

Value

Humidity (%) Spinneret temperature (1C) Outer layer dope flow rate (ml/min) Inner layer dope flow rate (ml/min) Bore fluid temperature (1C) Bore fluid flow rate (ml/min) External coagulant temperature (1C) Take up speed Air gap (cm)

75 Room temperature 2 6 25 4 25 Free fall 10

tanks. After being extruded from the spinneret, the nascent fiber traveled a certain distance of air-gap before entering into the external coagulant bath by free fall. The details of the spinning conditions are summarized in Table 2. Subsequently, the fibers were immersed in tap water, which was replaced daily, for 3 days to remove the residual solvent. Thereafter, they were cut into pieces of 30 cm long and subjected to solvent-exchange using methanol and hexane (Bonyadi and Chung, 2007; Wang et al., 2008). The fibers were then dried in ambient atmosphere. To fabricate a hollow fiber module, both ends of a plastic tube with an inner diameter of 6 mm and a specific length were joined with a union tee. A piece of fiber was placed in the tube with both ends protruded from the union tee outlets. The space between the hollow fiber and the union tee was then sealed with fast epoxy and dried overnight. Epoxy was applied twice to ensure complete sealing. The effective length of the fibers was 19 cm.

2.4. Hollow fiber characterization 2.4.1. Field emission scanning electron microscopy (FESEM) The FESEM (FEI Electron Optics B.V./Nova Nano SEM 230) was employed to examine the membrane structures including crosssection, inner surface and outer surface. The samples were fractured in liquid nitrogen to observe the cross-section. The current and time duration for spraying gold were 30 s and 20 mA, respectively.

2.4.2. Contact angle measurement The contact angle of the membranes was measured by a goniometer (Wetting Angle Test, Changchun) at room temperature. Porous flat PVDF membranes were prepared by wet phase inversion using the same dopes as those used for the outer layer of hollow fibers, i.e. PVDF/PVP and PVDF/glycerol. The coagulant was water that was the same as the external coagulant of hollow fiber spinning. Hence, the top surface of the flat sheet membranes so fabricated can represent the structure of the outer layer of the dual-layer hollow fibers. For contact angle measurement, the membrane was further dried by solvent exchange using methanol and hexane. Dense films were also prepared using the same casting dopes as those for the flat sheet membranes by solvent evaporation, i.e. cast polymer solution films were dried in air for 3 days at room temperature. For each sample, contact angles were measured at ten different spots and an average value was recorded. It should be noted that both porous membrane and dense film were also prepared from a casting dope, whose composition was PVDF (12 wt%), ethanol (10 wt%) and NMP (78 wt%) for comparison, by wet and dry phase inversion, respectively, for contact angle measurement.

81

2.4.3. Fourier transform infrared spectroscopy (FTIR) and X-ray photoelectron spectroscopy (XPS) FTIR investigation was carried out using Thermo Nicolet (NEXUS 670) FTIR equipment. The FTIR samples were prepared as follows: Neat PVDF film: PVDF was dissolved in NMP and cast to a thin film before solvent was evaporated at 50 1C in vacuo. Neat PVA film: PVA was dissolved in water and cast to a thin film before solvent was evaporated at 50 1C in vacuo. PVDF/PVA blends: the flat sheet membranes were made via phase inversion method from the PVDF/PVA casting dopes specified in Table 1 (3rd column). The porous membranes so prepared were dried by the solvent exchange method. X-ray Photoelectron Spectroscopy was carried out by using Thermo Fisher-VG Scientific (ESCALAB 250Xi) equipment. The samples for XPS were the hollow fibers prepared in Section 2.3 for DCMD tests. 2.4.4. Mechanical strength The mechanical strength of the hollow fibers was measured by a tensile machine (WDW-100N, Jilin Tanhor Testing Machine). The two ends of a fiber were clamped and fixed without bending or extension of the sample. The effective length of hollow fiber was 10 cm and the test mode involved stretching at a rate of 10 mm/min along the axial direction until fiber failure. The fiber elongation and tensile strength at break were recorded. The tests were carried out at room temperature (25 1C) and 70% relative humidity. 2.4.5. Porosity and pore size measurement Kerosene immersion method was used to determine the total porosity of the membranes (Bonyadi and Chung, 2007). The fibers were immersed in kerosene for 7 days to ensure that the pores were all perfectly wetted and filled with kerosene. The following equation was used to calculate the total porosity of the dual-layer hollow fibers:

εtotal ¼

m2  m1

π Lρker ðR1 2  R3 2 Þ

ð1Þ

where εtotal is the overall porosity of the dual-layer hollow fiber, ρker is the density (g/cm3) of the kerosene, m1 and m2 are the masses (g) of the sample before and after immersion into kerosene, respectively, R1 and R3 are the outer and inner diameters (cm) respectively before immersion into kerosene, and L is the effective fiber length (cm). The liquid–liquid displacement porosimetry method was used for determining the pore size distribution of the hollow fibers (Morison, 2008). The porosimetry setup is shown in Fig. S1. Before the test, DI water and butanol were mixed with a volume ratio of 1:4 and the mixture was stored at room temperature for 12 h for stratification into an upper butanol phase and a bottom water phase. The water phase was poured into the feed tank (5) shown in Fig. S1. In Fig. S1, (6) is the module for the inner layer pore size determination. One end of the fiber was sealed with epoxy resin while water phase could enter from the other end into the lumen side of the fiber under pressure supplied from nitrogen cylinder (1). In Fig. S1, (8) is the module for outer layer pore size determination, where the water phase can enter into the shell side of the module and the bottom end of the hollow was kept open to the air. The effective length of the fiber was 10 cm for both the modules. In both the modules, fibers were kept in contact with the butanol phase for 2 h for the pores to be fully wetted. Thereafter, hydraulic pressure was applied to the fiber lumen channel (module 6) or the shell side channel (module 8) to push the water phase into the pores to replace the butanol phase. The volume of the liquid collected in the volumetric cylinder indicates

82

J. Zhu et al. / Chemical Engineering Science 137 (2015) 79–90

the flow, Qi (m3/s) a function of hydraulic pressure pi (Pa). All the above measurements were performed at room temperature. The pore size ri (m) corresponding to the pressure pi (Pa) was calculated using the following equation: ri ¼

2e cos θ pi

ð2Þ

where e is the interfacial tension between water and butanol (N/m2) and θ is the contact angle of the membrane surface using butanol as testing liquid which is assumed as 01. Based on the pore size, ri, and flux, Qi, mentioned above, the pore size distribution is obtained by the following equation:
 pi pi  1 Q i  pi Q i  1 f ðr Þ ¼ ð3Þ m  P pi
ðr i  1  r i Þpi  1 p Q  p Q i1 i i i1 Pi  1

equation as follows: Cd ¼

C2V 2  C1V 1 V2 V1

ð6Þ

where C1 and C2 are the NaCl concentrations and V1 and V2 are the volumes of liquid in the distillate reservoir at time t1 and t2, respectively. A long term DCMD experiment was also conducted using P5 hollow fiber for more than 7 days. In this experiment, the inlet temperatures of feed and distillate were maintained at 65 1C and 17 1C, respectively, while the feed and distillate flow rates were maintained at 80 L/h and 2.3 L/h, respectively, during the entire period of experiment. 2.6. Wetting test

i¼1

where f(r) is the distribution function.

2.4.6. Viscosity characterization The viscosity of the inner layer dopes was measured by the falling ball method at room temperature. The time for the steel ball with a diameter of 1 cm to travel a vertical distance of 6.5 cm from the air/dope surface to the bottom was recorded and used as an indirect indicator for the viscosity of the blending dopes (Song and Jiang, 2013). 2.5. DCMD test The device for DCMD test contained two pumps to propel the concurrent circulations of the feed and the distillate fluids through the shell and lumen side of the fibers, respectively, unless otherwise stated. The module was placed vertically in the DCMD system. The volume flow rate (feed side: 40–120 L/h; distillate side: 1.4–3.29 L/h) and the pressure of feed and distillate stream were monitored and controlled by the flow meter and the back pressure valve, respectively. The temperature of the feed stream was regulated by a water bath, while that of the distillate stream by a cryostat. The inlet temperatures of the feed and the distillate were monitored by thermometers and maintained at 65 1C and 17 1C, respectively. The feed was aqueous 3 wt% sodium chloride (NaCl) solution while the distillate was DI water at the beginning of DCMD experiment. At least two modules were prepared for each hollow fiber fabrication condition and tested under the same DCMD conditions to insure replication. After the steady state was reached in one hour, the weight of the distillate container was monitored continuously using an electric balance (TP-220H, Changsha Xiangyi Instruments). The vapor flux J (kg/m2h) was then calculated by the following equation: J¼

Δm A Δt

ð4Þ

where Δm is the mass change of distillate (kg) in a selected period of time Δt (h) and A is the effective area of membrane for vapor transport (m2). The conductivity of feed and distillate was measured by a conductivity meter. The rejection coefficient R of inorganic salt (i.e. NaCl) was calculated by the following equation:   C R ¼ 1  d  100 ð5Þ Cf where Cf is the NaCl concentration in feed. Cd is the NaCl concentration of the distillate and was calculated by the balance

Examination about the wetting of the hollow fibers by the feed and the distillate liquids has also been carried out. For this purpose, two types of wetting experiments were carried out as follows: (1) Static wetting: a fiber was cut into halves along the axis and one half was immersed into 3 wt% aqueous NaCl solution at 25 1C for 8 h. (2) Dynamic wetting: this is further classified into two different modes. ● Test A: the hollow fiber module was subjected to DCMD for 5 h with 3 wt% NaCl aqueous solution flowing through the shell channel as feed and DI water flowing through the lumen channel as distillate. ● Test B: 3 wt% NaCl solution flowed in the lumen channel. The shell side was not contacted with liquid. Under these conditions, wetting of the inner layer can be independently characterized. The fibers were taken out of the module after experiment was over and dried in vacuo before being subjected to line scanning Energy Dispersive X-Ray (EDX). The distribution of Chlorine (Cl) elements across the cross-section of the hollow fiber was considered as indicators of pore wetting. The signal of Carbon (C) is also shown for comparison.

3. Results and discussion 3.1. Membrane characterizations 3.1.1. SEM images and FTIR Figs. 1 and 2 reveal the cross-sectional morphologies of the P and G series dual-layer hollow fibers. The magnification increases progressively from A to C. The overall dimensions (inner and outer diameters) of the fibers seem to be similar. A dense skin is formed at the inner and outer surfaces and pyriform-like macrovoids appear near the inner and outer skin surfaces, which are followed by a sponge-like structure. There is a clear boundary between the sponge-like structure of the inner and outer layers, as shown in C images. As for the outer layers, they are similar from P5 to P20 as well as from G5 to G20. The macrovoids of G-series fibers are larger than those of P-series. As for the inner layer, increasing the PVA content in the dopes (from P5 to P20 and G5 to G20) leads to the size increment of both pyriform-like macrovoids and spongelike cavities. Membrane structure is fixed during gelation caused by increased viscosity, micro-crystals formation, or glass transition of the polymer rich phase during and after solvent removal (Gaides and McHugh, 1989; Guenet et al., 1985; Tan et al., 1983; Wijmans et al., 1985). The hydrophilic nature of PVA may delay the gelation of polymer blending network inside water coagulant and

J. Zhu et al. / Chemical Engineering Science 137 (2015) 79–90

P10-A

P5-A

500μm

500μm

P5-B

100μm

P10-B

P5-C

10μm

P15-A

500μm

100μm

P20-A

500μm

P15-B

100μm

P10-C

10μm

83

P20-B

100μm

P15-C

10μm

P20-C

10μm

Fig. 1. Cross section morphology of P-series hollow fibers.

G5-A

G10-A

500μm

G15-A

500μm

100μm

G5-C

100μm

G10-C

10μm

500μm

G15-B

G10-B

G5-B

G20-A

G20-B

100μm

100μm

G20-C

G15-C

10μm

500μm

10μm

10μm

Fig. 2. Cross section morphology of G-series hollow fibers.

allows full development of the nucleus (Tohge et al., 1988). The dope viscosity is also increased by an increase in PVA content, which also enhances the effect of delayed demixing. As shown by the falling ball testing results in Fig. S2, the falling time of the small ball increases with the PVA content. It indicates that higher PVA content leads to higher viscosity of the inner layer dope.

The FTIR spectra of the dense PVDF and PVA membrane and the porous PVDF/PVA are shown in Fig. 3. Clearly, the spectra of PVDF and PVA experience no shift by blending, indicating weak interaction between the two polymers. Therefore, it seems possible that phase separation between two polymers take place after the solidification of the polymer blend, leaving pores when PVA is

84

J. Zhu et al. / Chemical Engineering Science 137 (2015) 79–90

1072 878

1172

1404

2943 2941

Transmittance (%)

3380

dissolved into coagulant water (Li et al., 2010). Thus, the higher the PVA loading the larger the pores become. As mentioned earlier, a clear boundary is observed between the inner and the outer layers. The two dopes did not mix with each other at the interface due to the difference in their properties. They also follow dissimilar phase inversion routes and the resultant spongy-like structure changes drastically where they meet. The void size of the sponge-like structure is smaller in the outer layer than in the inner layer for all the hollow fibers. The major reasons may include: the phase inversion initiates from the inner surface earlier than the outer surface; the solid content in the inner layer dope is lower; and the hydraulic pressure of bore fluid constrained in a limited space facilitates the macrovoids' growth by the mechanism of bulky coagulant penetration (Peng et al., 2008).

4000

3500

3000

2500

2000

1500

1000

500

Fig. 3. FTIR spectrum for films with different PVA/PVDF blending ratios.

Fig. 4 shows the surface morphologies of the shell side (outer) and the lumen side (inner) skin of the P and G series dual-layer hollow fibers. As for the shell side surface, G series looks denser and rougher. That the Mw of glycerol (92.09 g/mol) is much lower than that of PVP (8000 g/mol) is the major cause for the differences observed (Ohya et al., 2009). As for the lumen side surface, they are smooth and more porous than the shell side. Comparing P and G series, they look alike when the same dopes (e.g. P5 VS G5) are used for spinning. In fact, the change in PVA content affected the lumen side surface only little. The pore size distributions of P- and G-series hollow fibers measured by the liquid–liquid displacement porosimetry method are illustrated in Fig. 5. For the test of examining the outer surface, the fibers collapsed when the pressure reached the values which were shown in Fig. 5. The maximum pressure with no fiber collapse was also shown with each curve. Therefore, the distribution of pore size could only be obtained for relatively bigger pores, as shown in the upper rows of Fig. 5A and B. The biggest pore size is about 50–60 nm. As for the inner surface, the distribution curves exhibit no obvious difference for all the fibers. They are centralized around 16–20 nm, which confirms the findings in Fig. 4 regarding the SEM characterizations.

3.1.2. Mechanical strength The tensile strength was approximately 2.3 MPa for all the Pseries hollow fibers, having no obvious dependence on the inner layer polymer composition. As shown in Fig. S3, the elongation increased from 44 to 73% as PVA content increased from P5 to P15 but when PVA content was further increased to P20 the elongation decreased to 44%. PVA has hydroxyl groups that cause strong intermolecular interactions by hydrogen bonding. This helps improve the mechanical strength of its blend with PVDF (Riaz

Fig. 4. Shell (A) and lumen (B) skins' morphology of the dual-layer composite hollow fibers.

J. Zhu et al. / Chemical Engineering Science 137 (2015) 79–90

85

Hollow fiber shell surface 0.8 0.6

1.2

(0.137)

P5

8.0

(0.183)

P10

0.8

1.6

(0.183)

P15

6.0

0.4

4.0

0.8

0.2

0.0

2.0

0.4

0.0

(0.4)

0.0

0.0

(0.2)

(0.8)

(2.0)

(0.4)

f (r)

0.4

12

20

28

36

44

12 14 16 18 20 22 24 26

52

12

14

Pore size (nm)

Pore size (nm)

16

18

20

12

22

(0.176)

P20

1.2

14

16

18

20

22

Pore size (nm)

Pore size (nm)

Hollow fiber lumen surface 0.5

f (r)

0.4

1.5

P5

0.5

P10

1.2

0.5

P15

0.4

0.9

0.3

0.3

0.2

0.6

0.2

0.2

0.1

0.3

0.1

0.1

0.0

(0.0)

0.0

12

14

16

18

20

12

22

14

16

18

20

22

0.0 12

Pore size (nm)

Pore size (nm)

P20

0.4

0.3

14

16

18

20

12

22

14

16

18

20

22

Pore size (nm)

Pore size (nm)

Hollow fiber shell surface 0.4

f (r)

0.3

2.0

(0.097)

G5

0.2

0.4

(0.124)

G10

0.3

1.5

0.1

0.5

0.0

0.0 10

20

30

40

50

10

60

(0.5)

(0.1)

(2.0)

(0.1)

20

30

40

50

12

60

16

20

24

28

10

32

20

30

40

50

60

Pore size (nm)

Pore size (nm)

Pore size (nm)

Pore size (nm)

(0.154)

G20

2.5

0.2

0.0

0.1

3.5

(0.117)

G15

Hollow fiber lumen surface 0.5

f (r)

0.4

0.5

G5

0.4

0.5

G10

0.4

0.5

G15

0.3

0.3

0.3

0.3

0.2

0.2

0.2

0.2

0.1

0.1

0.1

0.1

0.0

0.0

0.0

18

20

22

24

26

Pore size (nm)

28

18

20

22

24

26

28

G20

0.4

0.0 18

20

22

24

26

28

18

20

Pore size (nm)

Pore size (nm)

22

24

26

28

Pore size (nm)

Fig. 5. (A) The pore size distribution of P-series fibers. (B) The pore size distribution of G-series fibers. (the data in the bracket is the maximum pressure in Mpa with no hollow fiber collapse).

et al., 2011). On the other hand, higher PVA content leads to larger cavities and higher porosity, which deteriorates tensile strength (Wang et al., 1996). As will be shown later, the porosity for P-series hollow fiber increases from ca. 85% to 93% with the increasing PVA content in the inner layer. In addition, PVDF membrane with low concentration of PVA forms spherulitic globule structure (see P5-C in Fig. 2 and G5-C in Fig. 3) with weaker mechanical properties (Sukitpaneenit and Chung, 2009). The patterns of the tensile strength and elongation variation are attributed to the trade-off in magnitude of these factors.

Table 3 XPS analysis of the inner layer of hollow fiber.

3.1.3. XPS The results of XPS analysis of the inner layer surface were performed and the results are shown in Table 3. The table also

includes the values calculated on the assumption that PVDF and PVA are blended uniformly throughout the inner layer and PVA did not leach out into the coagulant (water). The table shows that PVA

Blending ratio of PVDF/PVA

95/5 90/10 85/15 80/20

Measured values

Calculated values

O (wt%)

F (wt%)

Ratio of O/F

O (wt%)

F (wt%)

Ratio of O/F

4.44 7.56 9.24 14.53

54.05 47.71 46.87 37.22

0.0821 0.1585 0.1971 0.3904

1.88 3.78 5.68 7.60

58.41 55.50 52.58 49.64

0.0321 0.0681 0.1080 0.1531

86

J. Zhu et al. / Chemical Engineering Science 137 (2015) 79–90

remains in the inner layer surface despite the fact that PVA can be dissolved in water during the coagulation process. Surprisingly, PVA content at the inner surface, indicated by the O/F ratio, is higher than the theoretically expected value. The XPS is an evidence of the surface migration of PVA from inside the polymer dope to the spinning dope/coagulant interface during the phase inversion process, owing to its favorable interaction with water coagulant (Liu et al., 2013). 3.1.4. Contact angle measurement Fig. 6 shows that all the three flat sheet PVDF membranes have contact angles nearly equal to 801 regardless of nonsolvent additives. Fig. 6 also shows the contact angles of dense films. Interestingly, the contact angles are as low as 37.71 and 53.91, respectively, when nonvolatile PVP or glycerol was added as a pore former. On the other hand, the contact angle was as high as 117.71 when volatile ethanol was added as the pore former, indicating almost complete removal of ethanol together with the solvent NMP. These results indicate, unlike the dense PVDF film where solvent and pore former are nearly completely removed, the porous PVDF membranes, which represent the outer layers of the dual-layer hollow fibers, are not as hydrophobic as we think. 3.2. DCMD performance The DCMD performance in terms of flux and salt rejection is summarized in Fig. 7 for both P and G series membranes. With increasing PVA content in PVDF/PVA, the flux declines progressively from 7.6 kg/m2 h to 4.4 kg/m2 h for P series membranes and from 6.1 kg/m2 h to 2.4 kg/m2 h for G series membranes. The salt rejection is nearly equal to 100% when PVDF/PVP in the inner layer

is 95/5 (P5 and G5). However, the salt rejection decreased with increasing PVA content to the lowest 80 and 60% of P and G series, respectively, due to the increase in pore wetting. 3.3. Wetting tests There are several common factors that can lead to porous membrane wetting, a potential cause for declination of salt rejection (Ge et al., 2014). The first is associated with the defects at the membrane surface that allow free bulk liquid flow into the membrane; the second is associated with the DCMD operating condition and it occurs when the hydraulic pressure is higher than the liquid entry pressure and the third is due to the lowering of contact angle of the membrane that allows salt deposition or crystallization to proceed from the pore inlet deep into the interior of the pore. 3.3.1. Static wetting tests Fig. 8 shows the EDX analysis of the P- and G-series hollow fibers in static wetting investigation. In the figure intense Cl signals appear near the inner surface of all the fibers. As for P5 and G5, of which the PVA content of the inner layer is the lowest, an intense Cl signal appears only near the inner surface, indicating penetration of salt solution only from the lumen side. As PVA content increases, intense Cl signals appear deeper inside the hollow fiber and the distribution of the signals becomes more uniform throughout the cross-section. In other words, pore wetting started from the lumen side and penetrated deeper inside as the PVA content in the inner layer increased. This is ascribed to the increased water absorption capacity of the inner layer due to an increase in hydrophility as the PVA content increases. Moreover, as observed by SEM (Figs. 1 and

140

Contact angle (°)

120

Porous membranes

Dense membranes

79.5

82.5

117.7

100

80

81.5

53.9

60 37.7

40 20 0

PVP

Glycerol

Ethanol

Additive to the polymer solution Fig. 6. The contact angle of three types of outer layer of different additives in PVDF.

15 G series

Rejection (%)

Flux (kg/h -m2)

P series 12 9 6

100

60

3 P series 0

95/5

90/10

85/15

80/20

PVDF/PVA ratio

20

95/5

90/10

85/15

G series

80/20

PVDF/PVA ratio

Fig. 7. DCMD performance of the P- and G-series dual-layer hollow fibers membranes. Operation conditions: The inlet temperatures of bulk feed and distillate are 65oC and 17oC, respectively; the feed and distillate flow rates are 80L/h and 2.3L/h, respectively.

J. Zhu et al. / Chemical Engineering Science 137 (2015) 79–90

Intensity

300

P5

C

P10

Cl

87

P15

P20

200 100 0

0

50

100

150

200

250 0

P5 C Na

50

100

150

200

250 0

50

P10

P15

C

C Na

Na Cl

Cl

100

150

200

250 0

50

100

150

200

250

P20 C Na

Cl

Cl

300

Intensity

G5

C

G10

Cl

G15

G20

200 100

0

50

0

G5

100

150

200

250 0

50

100

150

200

250 0

G10

C Na Cl

50

100

150

200

G15

C

250 0

50

100

150

G20

C

250

C Na

Na

Na Cl

200

Cl

Cl

Fig. 8. EDX line scanning analysis for the static wetting test of P- and G-series hollow fibers (the dotted lines are the two layers' boundary).

Intensity

400

A-P5

300

C

A-P20

B-P20

B-P5

Cl

200 100 0 0

50

100 150 200 250 300 0

50

C Na

C Na

Cl

Cl

100

150

200

250

300 0

50

100

C Na Cl

150

200

250

300 0

50

100

150

200

250

300

C Na Cl

Fig. 9. Line scanning analysis of elements for the dynamic wetting tests A (1 and 2) and B (3 and 4) for P5 and P20 hollow fibers (the dotted lines indicate the boundaries of the two layers).

2C images), the spongy structures become bigger and macrovoids become more open with an increase in PVA content. Consequently, salt solution flows in the pores more readily from the inner surface toward the outer surface until the edge of the solution reaches the boundary between inner and outer layers. The dope compositions of the outer layers were unchanged within either P- or G-series. Therefore, it is expected that the outer layer will behave in the same manner, regardless of the change in PVA content in the inner layer. Notwithstanding, they behaved quite differently in these static wetting experiments. The outer layers of fibers P5 and G5, with least PVA content in the inner layer, showed practically no Cl signals, indicating no wetting of the outer layer, whereas, in the fibers of the highest PVA content in the inner layer (e.g. P-20 and G-20) intense Cl signals appeared in the outer layer, indicating severer wetting of the outer layer. This unexpected behavior of the outer layer wetting can probably be explained by the relatively low hydrophobicity of the porous PVDF membrane, particularly when

they are fabricated using a casting dope containing nonvolatile hydrophilic pore former such as PVP (P-series) and glycerol (Gseries), as indicated by the contact angle measurement. Therefore, the salt solution could enter the pores of the outer layer from the inner layer/outer layer boundary when the solution was drawn into the hollow fiber from the inner layer. Thus, the outer layer is not necessarily immune to water wetting. It is interesting to note however that the salt solution was not drawn into the outer layer from the shell side surface in this static wetting experiment. The outer layer pore wetting could take place only from the inner layer/outer layer boundary.

3.3.2. Dynamic wetting tests Fig. 9 shows the EDX results after the P-series hollow fibers (P5 and P20) were subjected for DCMD experiments. It is reminded that the outer surface (left side) was contacted with the feed salt

88

J. Zhu et al. / Chemical Engineering Science 137 (2015) 79–90

solution of high temperature in Test A, while the inner surface (right side) was contacted with the feed salt solution in Test B. It is interesting to note that, unlike in the static wetting experiments, A-P5 exhibits an intense Cl signal not only at the outer layer surface but also near the outer layer/inner layer boundary. Most likely, this was caused by internal concentration polarization, i.e. salt solution flowed through defective large pores of the outer layer, where the local liquid entry pressure is lower than the hydraulic pressure as predicted by the Laplace equation (Peña et al., 1993), by convection to reach the outer layer/inner layer boundary, then diffused slowly through the stagnant liquid water trapped in the sponge-like structure of the inner layer. The highest salt concentration hence arises at the outer layer/inner layer boundary. On the other hand, the intense Cl signal appears in B-P5 only at the inner surface that was brought into contact with hot salt solution in Test B. In both Test A and B, the middle part of P5 hollow fiber cross-section did not allow the intrusion of salt solution, at least during the testing period, due to the relatively small pore size and the low hydrophilicity of the inner layer. This is the reason for the nearly complete salt separation observed in DCMD experiments when P5 hollow fiber is used (see Fig. 7). As for P20, deep penetration of salt solution is evidenced by the distribution of intense Cl signals throughout the cross-section of the duallayer hollow fiber in both A-P20 and B-P20. As there is no liquid flow in the external channel of the DCMD module in dynamic test B, the wetting of the outer layer is solely attributed to the saline water coming from the internal channel of the module. 3.4. Effect of feed and distillate flow rates

12

P5

The long term DCMD performance of P5 was investigated and the results displayed in Fig. 12. In the figure, the flux of P5 remained at 16

P20

8 4 0 20

3.5. Long term performance of the dual-layer hollow fibers

Flux (kg/h -m2)

Flux (kg/h-m 2)

Figs. 10 and 11 illustrate the effect of feed flow rate on the DCMD performance of P and G series hollow fibers, respectively. In the figures, (A) is for the effect of the feed flow rate and (B) for the effect of the distillate flow rate. As shown in Figs. 10(A) and 11(A), the flux increases with increasing feed flow rate except for P20 at 120 L/h. For example, the flux of G5 increases from around 4.3 to 8.2 kg/m2 h. The enhancement of the feed circulation rate reduces the thickness of the temperature and mass boundary layer adjacent to the membrane surface, which in turn diminishes the degree of temperature and concentration polarization (TP and CP)

effects and improves the actual driving force between the two membrane surfaces (Bouguecha et al., 2003; Schofield et al., 1990). Figs. 10(B) and 11(B) show the effect of the distillate flow rate. All studied hollow fibers showed a maximum in flux as the flow rate was increased. The initial flux increase is attributed to the decrease in TP. The possible explanation for the flux decrease at high distillate flow rates is the hydrostatic pressure on the permeate side. The increase in the flow velocity in a narrow flow channel of the hollow fiber lumen causes an increase in hydrostatic pressure (Knudsen and Katz, 1980), as evidenced by the data shown in Table 4, i.e. the enhancement of the lumen side pressure becomes quite apparent when the flow rate reaches 3.5 L/h. High hydraulic pressure promotes the liquid water entry from the distillate side to enhance the pore wetting, resulting in decrease in flux. Probably, the exceptional flux data of P20 at the feed flow rate of 120 L/h was also caused by the pressure increase in the feed channel. As for the salt rejection, both P5 (Fig. 10) and G5 (Fig. 11) show nearly 100% salt rejection in the entire range of feed flow rate (A) and distillate flow rate (B). The salt rejections by P20 and G20 are much lower than P5 and G5, which was attributed to the more penetration of the salt solution into the membrane, as evidenced by the dynamic wetting test. Interestingly, the trend in the salt rejection parallels the trend in the flux, i.e. both flux and salt rejection show maximum as the feed or permeate flow rate increases and the maximum occurs at the same flow rate, with notable exception of G20 (B) where the position of the maximum shifted from 2.3 L/h of flux to 3.6 L/h of salt rejection. The explanation for the appearance of the maximum in salt rejection is the same as for the maximum of flux, i.e. at the lower flow rate region, increase in flow rate decreases the effect of TC, which leads to the water vapor flux increase without being accompanied by ion transport. As a result the selectivity increases. On the other hand, at higher circulation rates, the more severe pore wetting due to the enhanced hydraulic pressure provides more channels for ion transport and the rejection is reduced. For the negative flux, no rejection coefficients can be reported.

40

60

80

4

0 0

60

20 20

40

60

80

P5 P20 100 120 140

Feed flow rate (L/h)

1

2

3

4

5

6

7

Distillate flow rate (L/h)

Rejection (%)

Rejection (%)

Feed flow rate (L/h)

100

P20

8

4-

100 120 140

P5

12

100

60

20

P5 0

1

2

3

4

5

P20 6

7

Distillate flow rate (L/h)

Fig. 10. Effect of feed flow rate (A) and distillate flow rate (B) on the DCMD performance of the P series dual-layer hollow fiber membranes.

J. Zhu et al. / Chemical Engineering Science 137 (2015) 79–90

G5

G20

Flux (kg/h -m2)

Flux (kg/h-m 2)

12 8 4

0 20

40

60

80

15

5 0 -5 0

100

60 G5 40

60

80

1

2

3

4

5

6

7

Distillate flow rate (L/h)

Rejection (%)

Rejection (%)

Feed flow rate (L/h)

20 20

G20

10

-10

100 120 140

G5

89

G20

100

60

20

100 120 140

Feed flow rate (L/h)

G5 0

1

2

3

4

5

G20 6

7

Distillate flow rate (L/h)

Fig. 11. Effect of feed flow rate (A) and distillate flow rate (B) on the DCMD performance of the G series dual-layer hollow fiber membranes.

40

Table 4 The hydraulic pressure (MPa gauge) at different flow rates on the lumen (permeate) side. P20

G5

G20

1.06 2.3 3.54 4.78 6.03 7.27

0 0 0 0.04 0.08 –

0 0 0 0.04 0.08 0.1

0 0 0.02 0.06 0.08 0.1

0 0 0.02 0.04 0.07 –

80 20

60 Flux of P5

Rejection of P5

40

Rejection (%)

P5

100 30

Flux (kg/h-m2)

Flow rate (L/h)

120

10 20

2

around 7.5 kg/m h during the entire period of more than seven days and the salt rejection was always higher than 99%. Thus the membrane could be successfully used for desalination purpose by DCMD.

4. Conclusions Dual-layer composite hollow fibers were fabricated for DCMD. PVA was used to modify the hydrophility of inner layer, while PVP and glycerol were used as the pore forming agents in outer layer. Some conclusions that can be drawn from the experiments conducted in this work are highlighted below. (1) Presence of a clear boundary is observed between the inner and outer layers of the dual-layer hollow fiber. This boundary seems to play a role in the pore wetting. (2) Pore wetting becomes severer as PVA content in the inner layer increases. (3) As pore wetting becomes severer, both flux and salt rejection deteriorate. (4) The effect of flow rate is not only to reduce TP and CP. At a high flow rate, pore wetting may become severer due to increased

0

0

40

80

120

0 160

Time (hr) Fig. 12. The long term DCMD performance of P5 hollow fiber. Operation conditions: the temperatures of bulk feed and distillate are 65 ˚C and 17 ˚C, respectively; the flow rates of feed and distillate are 80 L/h and 2.3 L/h, respectively.

hydraulic pressure, resulting in deterioration of both flux and salt separation. This trend is more obvious when the lumen side flow rate is increased. (5) With the membrane of the least PVA content in the inner layer, there was no indication of pore wetting and high flux and salt rejection could be maintained during a long term experiment of more than 7 days.

Acknowledgments The authors would like to thank the National Natural Science Foundation of China, China (Project no. 21176265), the Hunan Provincial Science and Technology Plan (Project no. 2014GK3106)

90

J. Zhu et al. / Chemical Engineering Science 137 (2015) 79–90

and the Chinese National Project for Overseas Experts in Culture, Education and Public Health (Account no. 140010001).

Appendix A. Supporting information Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.ces.2015.05.064. References Alkhudhiri, A., Darwish, N., Hilal, N., 2012. Membrane distillation: a comprehensive review. Desalination 287, 2–18. Bonyadi, S., Chung, T.S., 2007. Flux enhancement in membrane distillation by fabrication of dual layer hydrophilic–hydrophobic hollow fiber membranes. J. Membr. Sci. 306, 134–146. Bonyadi, S., Chung, T.S., Rajagopalan, R., 2009. A novel approach to fabricate macrovoid-free and highly permeable PVDF hollow fiber membranes for membrane distillation. AIChE J. 55, 828–833. Bouguecha, S., Chouikh, R., Dhahbi, M., 2003. Numerical study of the coupled heat and mass transfer in membrane distillation. Desalination 152, 245–252. Criscuoli, A., Carnevale, M.C., Drioli, E., 2008. Evaluation of energy requirements in membrane distillation. Chem. Eng. Process.: Process Intensif. 47, 1098–1105. Curcio, E., Drioli, E., 2005. Membrane distillation and related operations—a review. Sep. Purif. Rev. 34, 35–86. Drioli, E., Criscuoli, A., Curcio, E., 2011. Membrane Contactors: Fundamentals, Applications and Potentialities: Fundamentals, Applications and Potentialities. Membrane Science and Technology, vol. 11. Elsevier. Edwie, F., Teoh, M.M., Chung, T., 2012. Effects of additives on dual-layer hydrophobic–hydrophilic PVDF hollow fiber membranes for membrane distillation and continuous performance. Chem. Eng. Sci. 68, 567–578. Gaides, G.E., McHugh, A.J., 1989. Gelation in an amorphous polymer: a discussion of its relation to membrane formation. Polymer 30, 2118–2123. Ge, J., Peng, Y., Li, Z., Chen, P., Wang, S., 2014. Membrane fouling and wetting in a DCMD process for RO brine concentration. Desalination 344, 97–107. Guenet, J.M., Wittmann, J.C., Lotz, B., 1985. Thermodynamic aspects and morphology of physical gels from isotactic polystyrene. Macromolecules 18, 420–427. Han, Z., Fina, A., 2011. Thermal conductivity of carbon nanotubes and their polymer nanocomposites: a review. Prog. Polym. Sci. 36, 914–944. Khayet, M., Matsuura, T., Mengual, J.I., Qtaishat, M., 2006. Design of novel direct contact membrane distillation membranes. Desalination 192, 105–111. Khayet, M., Mengual, J.I., Matsuura, T., 2005. Porous hydrophobic/hydrophilic composite membranes: application in desalination using direct contact membrane distillation. J. Membr. Sci. 252, 101–113. Knudsen, J.G., Katz, D.L., 1980. Fluid Dynamics and Heat Transfer. McGraw-Hill, New York. Lawson, K.W., Lloyd, D.R., 1997. Membrane distillation. J. Membr. Sci. 124, 1–25.

Li, N., Xiao, C., An, S., Hu, X., 2010. Preparation and properties of PVDF/PVA hollow fiber membranes. Desalination 250, 530–537. Liu, B., Chen, C., Li, T., Crittenden, J., Chen, Y., 2013. High performance ultrafiltration membrane composed of PVDF blended with its derivative copolymer PVDF-gPEGMA. J. Membr. Sci. 445, 66–75. Lovinger, A.J., 1982. Poly (vinylidene fluoride). Springer, Netherlands, pp. 195–273. Morison, K.R., 2008. A comparison of liquid–liquid porosimetry equations for evaluation of pore size distribution. J. Membr. Sci. 325, 301–310. Ohya, H., Shiki, S., Kawakami, H., 2009. Fabrication study of polysulfone hollowfiber microfiltration membranes: optimal dope viscosity for nucleation and growth. J. Membr. Sci. 326, 293–302. Peña, L., de Zárate, J.M.O., Mengual, J.I., 1993. Steady states in membrane distillation: influence of membrane wetting. J. Chem. Soc., Faraday Trans. 89, 4333–4338. Peng, N., Chung, T., Wang, K.Y., 2008. Macrovoid evolution and critical factors to form macrovoid-free hollow fiber membranes. J. Membr. Sci. 318, 363–372. Peng, W., May, M.T.A.T., 2011. Morphological architecture of dual-layer hollow fiber for membrane distillation with higher desalination performance. Water Res., 5489–5500. Riaz, U., Ashraf, S.M., Sharma, H.O., 2011. Mechanical, morphological and biodegradation studies of microwave processed nanostructured blends of some biobased oil epoxies with poly (vinyl alcohol). Polym. Degrad. Stab. 96, 33–42. Schofield, R.W., Fane, A.G., Fell, C.J.D., Macoun, R., 1990. Factors affecting flux in membrane distillation. Desalination 77, 279–294. Simone, S., Figoli, A., Criscuoli, A., Carnevale, M.C., Rosselli, A., Drioli, E., 2010. Preparation of hollow fibre membranes from PVDF/PVP blends and their application in VMD. J. Membr. Sci. 364, 219–232. Song, Z.W., Jiang, L.Y., 2013. Optimization of morphology and performance of PVDF hollow fiber for direct contact membrane distillation using experimental design. Chem. Eng. Sci. 101, 130–143. Su, M., Teoh, M.M., Wang, K.Y., Su, J., Chung, T., 2010. Effect of inner-layer thermal conductivity on flux enhancement of dual-layer hollow fiber membranes in direct contact membrane distillation. J. Membr. Sci. 364, 278–289. Sukitpaneenit, P., Chung, T., 2009. Molecular elucidation of morphology and mechanical properties of PVDF hollow fiber membranes from aspects of phase inversion, crystallization and rheology. J. Membr. Sci. 340, 192–205. Tan, H., Moet, A., Hiltner, A., Baer, E., 1983. Thermoreversible gelation of atactic polystyrene solutions. Macromolecules 16, 28–34. Tohge, N., et al., 1988. Fine-patterning on glass substrates by the sol–gel method. J. Non-cryst. Solids 100, 501–505. Wang, D., Li, K., Teo, W.K., 1996. Polyethersulfone hollow fiber gas separation membranes prepared from NMP/alcohol solvent systems. J. Membr. Sci. 115, 85–108. Wang, D., Li, K., Teo, W.K., 2000. Porous PVDF asymmetric hollow fiber membranes prepared with the use of small molecular additives. J. Membr. Sci. 178, 13–23. Wijmans, J.G., Kant, J., Mulder, M., Smolders, C.A., 1985. Phase separation phenomena in solutions of polysulfone in mixtures of a solvent and a nonsolvent: relationship with membrane formation. Polymer 26, 1539–1545. Wang, K.Y., Chung, T., Gryta, M., 2008. Hydrophobic PVDF hollow fiber membranes with narrow pore size distribution and ultra-thin skin for the fresh water production through membrane distillation. Chem. Eng. Sci. 63, 2587–2594.