Effect of internal coagulant on effectiveness of polyvinylidene fluoride membrane for carbon dioxide separation and absorption

Effect of internal coagulant on effectiveness of polyvinylidene fluoride membrane for carbon dioxide separation and absorption

Available online at www.sciencedirect.com Journal of Membrane Science 311 (2008) 153–158 Effect of internal coagulant on effectiveness of polyvinyli...

916KB Sizes 3 Downloads 111 Views

Available online at www.sciencedirect.com

Journal of Membrane Science 311 (2008) 153–158

Effect of internal coagulant on effectiveness of polyvinylidene fluoride membrane for carbon dioxide separation and absorption Aotian Xu a , Aihua Yang b , Stephanie Young a,∗ , David deMontigny b , Paitoon Tontiwachwuthikul b a

Environmental Systems Engineering, Faculty of Engineering, University of Regina, 3737 Wascana Parkway, Regina, Saskatchewan, Canada S4S 0A2 b International Test Centre for CO Capture, Faculty of Engineering, University of Regina, 2 3737 Wascana Parkway, Regina, Saskatchewan, Canada S4S 0A2 Received 1 September 2007; received in revised form 5 December 2007; accepted 6 December 2007 Available online 15 December 2007

Abstract Asymmetric hollow fiber membranes with inner skinless structures are favourable for carbon dioxide (CO2 ) separation and absorption in gas–liquid membrane contactors by reducing gas transport resistance. In this study, polyvinylidene fluoride (PVDF) hollow fiber membranes were fabricated by using different internal coagulants, which include water, ethanol–water, and dimethylacetamide (DMAc)–water solutions. The spinning dope solution was prepared with DMAc as the solvent, lithium chloride and polyvinylpyrrolidone (PVP) as the non-solvent additives. The resultant membranes were examined by scanning electron microscopy (SEM). The examination of membrane cross-sections indicated that the composition of internal coagulants had significant impacts on the thickness and structures of the inner skin layer. Results showed that an inner skinless surface could be formed using a 30 wt.% water–DMAc mixture. Gas permeation tests demonstrated that PVDF hollow fibers with an inner skinless surface had a higher CO2 absorption rate than other PVDF and polypropylene (PP) membranes and attained performance levels equivalent to a polytetrafluoroethylene (PTFE) membrane. PVDF membranes could be a cost-effective alternative for gas–liquid membrane contactors in applications of CO2 separation and absorption. © 2008 Elsevier B.V. All rights reserved. Keywords: PVDF hollow fibers; Carbon dioxide capture; Gas–liquid membrane contactors; Mass-transfer resistance; Skin layer

1. Introduction Greenhouse gas emission reduction, particularly with respect to carbon dioxide (CO2 ), has attracted increasing attention worldwide in recent years. CO2 is considered a major greenhouse gas that contributes to global warming and other environmental concerns [1]. CO2 emissions related to human activities mainly result from fossil fuel combustion [2]. CO2 capture is the main approach to reducing CO2 emissions for coal-fired power plants and other industrial sources. Gas–liquid membrane contactors integrated with chemical absorption are considered one of the most promising techniques Abbreviations: CO2 , Carbon dioxide; PVDF, Polyvinylidene fluoride; DMAc, Dimethylacetamide; PVP, Polyvinylpyrrolidone; SEM, Scanning electron microscopy; PP, Polypropylene; PTFE, Polytetrafluoroethylene. ∗ Corresponding author. Tel.: +1 306 585 4722; fax: +1 306 585 4855. E-mail address: [email protected] (S. Young). 0376-7388/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.memsci.2007.12.008

for CO2 capture [3,4]. A gas–liquid membrane contactor is a gas absorption system with a high mass-transfer rate that is achieved by the use of microporous membranes. Hollow fiber membranes are suitable for use in gas–liquid contactors due to their large specific surface areas and asymmetric structures. During the absorption process, the liquid phase (absorbent) passes through the outer side of the membrane and the gas is fed into the lumen side of the hollow fibers. The gas–liquid interface is located at the mouth of each membrane pore on the solution side by adjusted pressure difference [5]. This process can achieve significantly high absorption efficiencies due to the much larger surface area for gas–liquid interface than conventional gas absorption processes. Other advantages of the system include independence of liquid and gas flowrates, absence of emulsions, and flexibility of operations [6–8]. Hollow fibers with inner skinless structures and porous substrates are favourable for application of gas–liquid membrane contactors due to their low mass-transfer resistance and high

154

A. Xu et al. / Journal of Membrane Science 311 (2008) 153–158

permeability. Since most of the mass-transfer resistance in the membrane is credited to the dense layer [9], inner skinless structures can reduce the gas transport resistance of the hollow fiber. However, most commercial hollow fiber membranes are fabricated with double skin layers, which may not fully meet the desired characteristics of gas–liquid contactors. The aim of this work was to evaluate the effect of internal coagulant composition on structures of hollow fiber membranes and to investigate manufacturing methods to eliminate the inner skin layer. Unlike membranes used in gas separation applications, the hollow fiber membranes suitable for CO2 absorption must have a distinct inner surface structure. For conventional gas separation, hollow fiber membranes were fabricated with a thin and dense inner skin layer, which was characterized by high selectivity. The formation of a defect-free skin layer has been studied by many researchers [10]. However, for CO2 separation and absorption using gas–liquid membrane contactors, high selectivity is not necessary and the inner skin layer is not expected. Li’s study [11] showed that the composition of water coagulation bath had significant effects on morphology of PVDF hollow membranes. This study focused on the fabrication of hollow fiber membranes with an inner skinless surface aimed at the applications of CO2 separation and absorption. A hollow fiber membrane with an inner skinless surface will result in high process efficiency, compact process systems, and low operation costs. PVDF was employed as the membrane material combined with five different internal coagulant compositions to prepare the hollow fiber membranes. PVDF is a hydrophobic polymer material with excellent chemical and thermal resistance performance. The PVDF hollow fiber membranes were fabricated using a dry/wet phase inversion method. Five compositions of internal coagulation were applied to the fabrications, including water, ethanol–water, and DMAc–water solutions. Lithium chloride and PVP were used as non-solvent additives. DMAc was the solvent used in the spinning dope solution. The manufactured hollow fiber membranes were examined with SEM to investigate both porosity and morphology. Moreover, gas permeability of the resultant hollow fiber membranes was tested in a gas–liquid membrane contactor. 2. Experimental procedures and set-up

Table 1 Bore fluid compositions No.

Composition

1 2 3 4 5

Distilled water (100 wt.%) Distilled water (50 wt.%), ethanol (50 wt.%) Distilled water (90 wt.%), DMAc (10 wt.%) Distilled water (80 wt.%), DMAc (20 wt.%) Distilled water (70 wt.%), DMAc (30 wt.%)

The hollow fiber membranes were fabricated using a dry/wet phase inversion process. The hollow fiber membrane spinning system, DKN-02, was supplied by Kana Science Technique Co., China. A tube-in-orifice spinneret, with an outer diameter of 1.0 mm and an inner diameter of 0.6 mm, was used to form hollow fibers. The degassed dope solution was extruded to a gear pump by a high-pressure nitrogen gas and was then fed to the spinneret by the gear pump. The extrusion rate of the dope solution was controlled at 2.0 ml/min. The bore fluid was fed to the center tube of the spinneret by a syringe pump. The injection rate of the bore fluid was controlled at 1.2 ml/min to form a circular lumen in the hollow fiber. Once extruded from the spinneret nozzle, the nascent hollow fiber passed through a 7 cm air gap before being immersed in the external coagulant. After spinning, the hollow fibers were flushed by deionised water for 1 week to remove any residual solvent and additives. The wet membranes were immersed into 50% ethanol, pure ethanol, and then an ethanol–hexane mixture (1:1) in sequence for solvent exchange. The membranes were then dried in air at room temperature to minimize shrinkage. 2.2. SEM examination The membrane morphology was examined using a SEM (JSM-5600, JEOL Ltd.). The hollow fiber membrane specimens were immersed in liquid nitrogen and carefully fractured to present cross-sections of the structures. The specimens were dried in a vacuum oven for 24 h and then coated by sputtering gold/platinum before testing. SEM images of the membrane fiber cross-section and surface were taken at 50, 500, and 1500× magnifications.

2.1. Fabrication of hollow fiber membrane 2.3. Gas permeation test The PVDF Kynar® K-760 in pellet form (Arkema Inc.) was used for the fabrication of the hollow fiber membranes. DMAc (99%+,) (Fisher Scientific) was used as the solvent. Lithium chloride (Alfa Aesar) and PVP (M.W. 1,300,000) (Acros Organics) were used as non-solvent additives. The spinning dope solution was prepared by mixing PVDF, DMAc, and additives under continuous mixing for 48 h. After obtaining a homogenous solution, the spinning dope was degassed for several days under vacuum pressure (0.02 MPa). The composition of the polymer solution was 18 wt.% PVDF, 80 wt.% DMAc and 2 wt.% PVP. Tap water was used as an external coagulant at room temperature (22.5 ◦ C ± 1). The five different bore fluids (internal coagulants) used for the fabrications are presented in Table 1.

The resultant hollow fibers were tested in a gas–liquid membrane contactor to study the permeation performance. A total of 200 hollow fibers were mounted in an acrylic membrane module with an effective length of 120 mm. The inside diameter of the acrylic module was 28 mm with a wall thickness of 6 mm. Pure CO2 was employed as the feed gas, while distilled water was used as the absorbing solvent. In this study, counter-current flow was used, with the liquid absorbent being fed to the membrane contactor on the shell side and the CO2 gas passing through the lumen side of fibers. The influent gas flowrate was controlled by a mass flow controller, while the effluent gas flowrate was measured by a bubble flow meter.

A. Xu et al. / Journal of Membrane Science 311 (2008) 153–158

3. Results and discussion 3.1. Morphology of hollow fiber membranes The PVDF hollow fiber membranes were fabricated using a dry/wet phase inversion process with five different internal coagulants. The morphologies of the membranes were studied by SEM to present the cross-section and internal surface micrographs. The resultant membranes had outer diameters ranging from 1000 to 1100 ␮m, inner diameters ranging from 750 to 850 ␮m, and hollow fiber wall thicknesses ranging from 110 to 125 ␮m. It was found that the composition of the internal coagulants has significant impacts on the structures and properties of hollow fiber membranes. The morphologies of the membranes fabricated with internal coagulants of water and ethanol solution (50 wt.%) were observed using SEM. The cross-sectional structures of these membranes are presented in Fig. 1. Fig. 1(a) reveals that skin layers are present on both inner and outer sides of the hollow fiber wall. Macrovoids and finger-like structures are formed underneath both of the inner and outer skin layers with sponge-like structures between the finger-like structures. Near the inner side, the cavities in the sponge-

Fig. 1. SEM morphologies of cross-sectional structures. Internal coagulants: (a) water (100 wt.%); and (b) water–ethanol (50 wt.%).

155

like structures are apparently denser than those of the outer side. In addition, the inner finger-like structures, with a thickness of 35–39 ␮m, are thicker than those of the outer side. This was mainly attributed to the drying process during the membrane fabrication. In this study, a dry/wet phase inversion process was adopted with an air gap of 7 cm. Phase separation of the drying process plays an important role in the formation of skin layers [12]. A nascent outer skin layer was formed due to solvent evaporation in air. Compared to the phase separation on the lumen side, a lower amount of water penetrated through the nascent outer layer to form droplets, and eventually this resulted in fewer macrovoids on the outer side. This agrees with the findings of other researchers. It was reported that an increase in air gap distance resulted in a reduced layer of finger-like voids and a lower permeance [13]. Fig. 1(b) shows that inner finger-like structures were eliminated with the use of a 50% ethanol solution. The combined thickness of the inner skin layer and the sub-layer structures was approximately 10 ␮m. The finger-like structures on the outer side, which were still present, were longer than those of Fig. 1(a). The sponge-like structures were located on the lumen side of the hollow fiber wall, and the thickness was about half of the wall. Since the formation of sponge layer depends on the relative powers of the internal and external coagulants, and other factors [14], a weak non-solvent coagulant changes the position of the sponge-like structures. Therefore, it can be concluded that the use of the water–ethanol solution as an internal coagulant has significant effects on the structures of PVDF hollow fiber membranes. The presence of the inner skin layer indicated that using the 50% ethanol solution as an internal coagulant could not form an inner skinless surface for PVDF hollow fiber membranes. Other researchers reported that the outer skin layer was eliminated when an aqueous solution containing more than 50% ethanol was applied as the external coagulant [15]. However, it was found that the lumen of the hollow fiber could not be formed when a higher concentration of ethanol solution was used as the internal coagulant. PVDF hollow fiber membranes prepared with three other internal coagulants, containing different concentrations of DMAc, were examined and are presented in Fig. 2. As shown in Figs. 1(a) and 2(a and b), the macrovoid structures were formed under the inner skin layer when 0, 10, and 20 wt.% water–DMAc mixtures were used as the internal coagulants. The thickness of the finger-like structures on the lumen side decreases with the increasing of DMAc concentration. The thickness ranges between 29 and 34 ␮m when 10 wt.% water–DMAc mixture was used and decreases to 26 and 31 ␮m when the DMAc concentration was 20 wt.%. Correspondingly, the longer finger-like structures on the outer side presented when a higher concentration of the water–DMAc mixture was used. The locations of the sponge-like structures change as a result of the variations of the macrovoid structures on both sides of the membrane walls. When the concentration of internal coagulant varied from 0 to 20%, the thinness of the sponge-like structure was maintained approximately constant in the range of between 32 and 35 ␮m.

156

A. Xu et al. / Journal of Membrane Science 311 (2008) 153–158

Fig. 2. SEM morphologies of cross-sectional structures. Internal coagulants: (a) water–DMAc (10 wt.%); (b) water/DMA (20 wt.%); and (c) water–DMAc (30 wt.%).

In this study, the alterations of membrane structures are mainly attributed to the polymer phase inversion process, with different internal coagulants involved. The macrovoids were formed during the solidification process, also known as the nucleation process [16,17]. A thin surface layer was formed on the lumen side after spinning. The internal coagulant solution diffused through the surface layer into the membrane wall and accumulated as droplets underneath the surface layer. The solvent from the surrounding polymer solutions dissolved into the droplets and contributed to the growth of the droplets in size. Phase inversion occurred during the solvent extraction process from the polymer solutions, which led to solidification of the surrounding polymer. Consequently, the space filled with internal coagulant and solvent was solidified and macrovoids were formed. The formation of macrovoids was determined by the relative powers of the internal coagulant and the solvent. A weaker internal coagulant, containing a higher DMAc concentration, resulted in a lower flux of solvent extraction and a slower phase inversion process. Eventually, smaller macrovoids were generated due to the slower phase inversion. The skin layer was eliminated when a bore fluid containing 30% DMAc was employed as the internal coagulant. In Fig. 2(c), the length of finger-like structures ranges between 52 and 64 ␮m, and this occupies approximately half of the thinness of the membrane wall. The volumes of the finger-like structures

were significantly larger than those of other hollow fiber membranes prepared with stronger internal coagulants. The existence of the outer dense skin layer is clearly present in the SEM micrograph of Fig. 2(c). Remarkably, the skin layer was absent and a microporous surface was formed on the inner side. Some dropshaped macrovoids with lengths of 26–29 ␮m were distributed in the sub-layer. The inner skinless surface of the hollow fiber membrane was examined using SEM, as shown in Fig. 3. From Fig. 3, it can be seen that an open microporous structure was formed on the inner surface of the hollow fiber membrane. DMAc was used as the solvent of polymer solution, and a 30% water–DMAc solution was used as the internal coagulant, which was considered a very weak non-solvent. The inner dense skin layer was not able to form due to the relative high homogeneity between the internal coagulant and the solvent. The low influx of DMAc from polymer solution to internal coagulant led to very slow phase inversion and polymer solidification processes on the inner surface. The open microporous structure was formed, rather than a skin layer, during the solidification processes. Similarly, the formation of drop-shaped macrovoids in the sub-layer was mainly attributed to the slow phase inversion process. For membranes prepared with a strong internal coagulant, such as water, after the internal coagulant penetrated into the membrane and the droplets were formed, the walls

A. Xu et al. / Journal of Membrane Science 311 (2008) 153–158

157

Fig. 4. Results of CO2 absorption test results using a gas–liquid membrane contactor. Fig. 3. SEM morphologies of inner surface structures of the hollow fiber membrane. Internal coagulant: water–DMAc (30 wt.%).

between droplets solidified rapidly. During the growth process of a droplet, the solidified wall forced the droplet to extend along the depth direction, and a finger-like cavity was formed. When a weak non-solvent was employed as the internal coagulant, the walls between small droplets were formed with difficulty due to the slow solidification. Several small droplets combined to form a large droplet, which, then, expanded along all directions. As a result, a large drop-shaped cavity was created in the sub-layer, as shown in Fig. 2(c). The mechanism of formation of dropshaped macrovoids under the inner skinless surface of a hollow fiber membrane needs to be further studied. 3.2. CO2 absorption using a gas–liquid membrane contactor Three types of PVDF hollow fiber membranes fabricated in this study were compared with two types of commercial hollow fiber membranes. The three resultant hollow fiber membranes (PVDF 1, 2 and 3) were fabricated with water, ethanol solution (50 wt.%), and DMAc solution (30 wt.%), respectively, as the internal coagulants. Comparative testing was conducted on a PTFE membrane and a PP membrane. The characteristics of the commercial hollow fiber membranes are given in Table 2 and the effects of absorbent flowrate on CO2 flux are shown in Fig. 4. As shown in Fig. 4, the approximately parallel curves of PVDF 1, 2 and 3 illustrate that the absorbent flowrate had an effect on the CO2 flux of the three membranes in the same pattern. The CO2 flux increased with an increase in the absorbent flowrate and asymptotically approached a constant value. It also demonstrates that the CO2 flux of the membrane prepared with ethanol solution (50 wt.%) as the internal coagulant

was slightly lower than that of the membrane prepared with water. SEM exanimation demonstrated that ethanol solution resulted in significant alterations with respect to membrane type to the membrane structure; however, it did not have any evidentiary effect on the CO2 flux. The skin layer still existed on the inner side of PVDF 2 and contributed to the mass-transfer resistance. It also indicates that the changes of the sub-layer structures of a hollow fiber membrane could not effectively improve CO2 absorption performance. In conclusion, a crucial approach to improving membrane structures for CO2 absorption is to eliminate the inner skin layer and, consequently, the related mass-transfer resistance. In Fig. 4, the hollow fiber membrane (PVDF 3) prepared with DMAc solution (30 wt.%) as the internal coagulant shows a significantly higher CO2 flux than the other PVDF membranes. The curve PVDF 3 is approximately parallel with PVDF 1, and the CO2 flux is higher than PVDF 1 by about 0.66 × 10−4 mol/m2 s at all experimental operations. The increased CO2 absorption rate of this membrane is attributed to the elimination of the inner skin layer, which was verified by SEM examination. The test results of PTFE and PP membranes are also presented in Fig. 4, which shows that PVDF 3 had a much higher CO2 absorption rate than the PP membrane. The CO2 absorption rate of the PTFE membrane decreased faster with a greater decrease in the absorbent flowrate than that of PVDF 3, and they are equal when the absorbent flowrate is 130 ml/min. This indicates that PVDF 3 had acceptable performance under relatively low absorbent flowrate conditions, which might result in considerable savings on capital and operation costs. In this study, the PVDF hollow fiber membrane with a skinless inner surface had a competitive performance when compared to the PTFE membrane. PTFE membranes are mainly prepared by a thermal induced phase separation method that results in high manufacturing cost and relatively large physical size. By comparison,

Table 2 Characteristics of commercial hollow fiber membranes Membrane

Supplier

Outer diameter (mm)

Inner diameter (mm)

Wall thickness (mm)

PTFE PP

Sumitomo Electric Fine Polymer Mitsubishi Rayon

2.0 0.300

1.0 0.244

0.5 0.028

158

A. Xu et al. / Journal of Membrane Science 311 (2008) 153–158

due to its good chemical and physical performance and the low manufacturing/operation costs, the PVDF hollow fiber membrane is considered a less costly alternative for CO2 separation and absorption in gas–liquid membrane contactors. 4. Conclusions PVDF hollow fiber membranes were successfully fabricated with different internal coagulants using a dry/wet phase inversion process. SEM exanimations and gas–liquid membrane contacting tests lead to several significant conclusions. • The concentrations of DMAc solutions had significant impact on the membrane structures, and a hollow fiber membrane with an inner skinless surface was obtained when a DMAc solution (30 wt.%) was used. • The resultant inner skinless hollow fiber membrane showed a higher CO2 absorption rate than other PVDF and PP membranes and a rate equivalent to a PTFE membrane, especially under low absorbent flowrate conditions. • The application of ethanol solution (50 wt.%) resulted in sublayer structure transformation but was not able to remove the inner skin layer. The resultant hollow fiber membrane had a similar CO2 absorption rate to the membrane prepared with water as an internal coagulant. • In addition, the inner skinless hollow fiber has the advantages of high CO2 absorption efficiency, low operation costs, and compact design of facilities. The inner skinless hollow fiber is, thus, a cost-effective alternative membrane for CO2 separation and absorption in gas–liquid membrane contactors. Acknowledgements This study was supported by the Natural Sciences and Engineering Research Council of Canada (NSERC) through a Discovery Grant to Dr. Stephanie Young. The authors would also like to thank Dr. Phillip Y.K. Choi, Dr. Roderick Facey and Dr. D.W. Smith for their reviews, comments and guidance. References [1] D.J. Wuebbles, A.K. Jain, Concerns about climate change and the role of fossil fuel use, Fuel Process Technol. 71 (2001) 99–119.

[2] W. Bartok, A.F. Sarofim, Fossil Fuel Combustion: A Source Book, WileyInterscience, New York, 1991. [3] Q. Zhang, E.L. Cussler, Microporous hollow fibers for gas absorption: I. Mass transfer across the membrane, J. Membr. Sci. 23 (1985) 333–345. [4] R.W. Baker, Membrane Technology and Applications, McGraw-Hill, New York, 2000. [5] M. Mavroudi, S.P. Kaldis, G.P. Sakellaropoulos, Reduction of CO2 emissions by a membrane contacting process, Fuel 82 (15–17) (2003) 2153–2159. [6] V.Y. Dindore, D.W.F. Brilman, P.H.M. Feron, G.F. Versteeg, CO2 absorption at elevated pressures using a hollow fiber membrane contactor, J. Membr. Sci. 235 (2004) 99–109. [7] D. Bhaumik, S. Majumdar, K.K. Sirkar, Absorption of CO2 in a transverse flow hollow fiber membrane module having a few wraps of the fiber mat, J. Membr. Sci. 138 (1998) 77–82. [8] P.H.M. Feron, A.E. Jansen, CO2 separation with polyolefin membrane contactors and dedicated absorption liquids: performances and prospects, Separation Purif. Technol. 27 (2002) 231–242. [9] S. Atchariyawut, C. Feng, R. Wang, R. Jiraratananon, D.T. Liang, Effect of membrane structure on mass-transfer in the membrane gas–liquid contacting process using microporous PVDF hollow fibers, J. Membr. Sci. 285 (2006) 272–281. [10] A.F. Ismail, N. Ridzuan, S.A. Rahman, Latest development on the membrane formation for gas separation, J. Sci. Technol. 24 (2002) 1025–1043. [11] S.P. Deshmukh, K. Li, Effect of ethanol composition in water coagulation bath on morphology of PVDF hollow fiber membranes, J. Membr. Sci. 150 (1998) 75–85. [12] M. Niwa, H. Kawakami, S. Nagaoka, T. Kanamori, T. Shinbo, Fabrication of an asymmetric polyimide hollow fiber with a defect-free surface skin layer, J. Membr. Sci. 171 (2000) 253–261. [13] T.S. Chung, X. Hu, Effect of air-gap distance on the morphology and thermal properties of polyethersulfone hollow fibers, J. Appl. Polym. Sci. 66 (1997) 1067–1077. [14] M. Wang, L. Wu, J. Mo, C. Gao, Preparation and characterization of polyacrylonitrile-based membranes: effects of internal coagulant on poly (acrylonitrileco-maleic acid) ultrafiltration hollow fiber membranes, J. Membr. Sci. 274 (2006) 200–208. [15] H.J. Kim, R.K. Tyagi, A.E. Fouda, K. Jonasson, The kinetic study for asymmetric membrane formation via phase-inversion process, J. Appl. Polym. Sci. 62 (1996) 62–629. [16] D.M. Koenhen, M.H.V. Mulder, C.A. Smolders, Phase separation phenomena during the formation of asymmetric membranes, J. Appl. Polym. Sci. 21 (1977) 199–215. [17] C.A. Smolders, A.J. Reuvers, R.M. Boom, I.M. Wienk, Microstructures in phase-inversion membranes. Part 1. Formation of macrovoids, J. Membr. Sci. 73 (1992) 259–275.