A preparation method for the highly permeable ceramic microfiltration membrane – precursor film firing method

A method called the precursor film firing method is proposed to improve the permeance of ceramic microfiltration membranes by avoiding intermediate layer and dip-coating process, and efficiently control the thickness of the separation layer. In this method a precursor film is prepared independently of the support by a film coating machine. This precursor film consists of two layers: Al2O3/PVA layer made of Al2O3 powder and polyvinyl alcohol (PVA), and polyvinyl butyral (PVB) layer. The precursor film is pasted on the support and fired to obtain membranes. Because the intermediate layer and dip-coating process are avoided in this method, the fabricated membranes show high permeance. For the fabricated membrane with average pore size = 0.18 μm and separation layer thickness = 10.7 μm, the permeance is 3890 L m−2 h−1 bar−1. Also the separation layer thickness can be controlled efficiently. And the precursor film can be used on fabricate curved membranes such as cylindrical membranes.


Introduction
In comparison with other membranes, such as polymeric membranes, ceramic membranes show higher cost because of their high sintering temperatures and raw materials. However ceramic membranes normally show considerable mechanical strength, thermal stability and chemical compatibility. 1,2 Due to these properties, they are promising in some ltration processes under harsh conditions, such as high pressure and high temperatures.
Ceramic microltration membranes normally consist of three layers: support layer, intermediate layer and separation layer. 3,4 The support layer provides mechanical strength, and the separation layer gives selectivity efficiency. 4 The intermediate layer that bridges the support layer and the separation layer plays several roles: 3 usually the precursor system of the separation layer such as ceramic slurry [5][6][7][8][9][10][11] and sol [12][13][14][15] are coated on supports by dip-coating technique. In the absence of the intermediate layer, the small particles in the precursor system for the separation layer will probably penetrate to the large pores of the support and decrease the permeance of the membrane. Moreover, coating a separation layer directly on a support with large pores normally may make the separation layer peel off easily. 3 However, the intermediate layer will increase the total thickness of membrane system, which will reduce the permeance of membrane. Moreover, the fabrication of the intermediate layer will increase the total fabrication costs of membrane system. Hence, the ideal ceramic microltration membrane should consist of a support with large pores and a very thin and defectfree separation layer with controllable pore sizes. 3 The permeance of membrane is sensitive to the thickness of the separation layer. The separation layer of commercial microltration membranes is generally in the range of 10-20 mm. 3 Controlling the thickness of the separation layer is an important issue in the membrane fabrication process. In traditional dip-coating process, the liquid precursor is absorbed towards the substrate under the driving force of capillary suction pressure of the support or the external applied pressure. 5 The coating thickness of the separation layer is affected by many parameters, including viscosity, density and surface tension of the precursor system; porosity, surface roughness and radius of the support; the contact time; and the withdrawal speed, etc. 5 Hence, it is complex to control the thickness of the separation layer by adjusting so many parameters in traditional dip-coating technique.
Recently, Qin et al. 16,17 attempted to fabricate membranes without intermediate layer based on one-step coating method and sacricial-interlayer method. The obtained membranes show relatively high permeance in comparison with the membranes fabricated using traditional dip-coating process. However, because the dip-coating technique was also used in these two methods, 16,17 the small particles in separation layer might penetrate to the big pores of support. There is still some room to improve these two methods. Moreover it is inconvenient to control the separation layer thickness using these two methods.
This study aims to propose a novel ceramic microltration membrane preparation method to improve the permeance of membrane by avoiding intermediate layer and dip-coating process, and efficiently control the separation layer thickness. In this method a precursor lm is prepared independently of the supports by a lm coating machine. This precursor lm consists of two layers: Al 2 O 3 /PVA layer made of Al 2 O 3 powder and polyvinyl alcohol (PVA), and polyvinyl butyral (PVB) layer. The precursor lm is pasted on the support and red to obtain membrane with high permeance. Because of the important role of precursor lm in this method, this novel method is called precursor lm ring method. It is noted that the pore size of microltration membrane is in a relatively wide range: 0.05-10 mm. 4 To illustrate the mechanism of precursor lm ring method, this study will concentrate on the membranes with pore size around 0.1-0.5 mm. However the applicability of the proposed precursor lm ring method is not limited in this pore size range.
2.1.2 For the fabrication of precursor lms. Commercially available a-Al 2 O 3 powder with mean particle size ¼ 0.52 mm (99.5% purity, Showa Denko Co., Ltd., Japan) was used as main material. PVA (1799, Shanghai Aladdin Biochemical Technology Co., Ltd., China) was used as polymeric binder. PAA (solid mass fraction ¼ 30%, Tianjin Kermel Chemical Reagent Co., Ltd., China) was used as dispersant. Silicone-based defoamer (Nanjing Qicheng New Type Material Co., Ltd, China) was used to reduce air bubbles. Fluorocarbon surfactant (Capstone FS-60, DuPont, USA) was used to reduce the surface tension of membrane slurry. Deionized water was used as solvent for membrane slurry. Polyvinyl butyral (PVB) powder (Shanghai Aladdin Biochemical Technology Co., Ltd., China) was used as main raw material for preparing PVB layer. Ethanol (99.9% purity, Tianjin Fuyu Huagong Co., Ltd, China) was used as solvent for dissolving PVB powder.

Fabrication of supports
Qin et al. 18 lowered the sintering temperature for fabricating high-purity porous alumina support by using boehmite. Similar to Qin et al., 18 the present study also used boehmite as a sintering additive to fabricate the supports. There are three main steps.
2.2.1 Preparation of a-Al 2 O 3 slurry. 200 g PVA solution (PVA mass content ¼ 2%), 180 g a-Al 2 O 3 powder (D 50 ¼ 9 mm), 45 g rice starch and 1.8 g PAA were mixed evenly. Meanwhile, the pH of the solution was kept at approximately 5.0 for achieving the electrostatic stabilization.
2.2.2 Preparation of boehmite sol. 12 g boehmite and 0.6 g PEI were added in 150 mL deionized water while stirring. The pH of the mixture was also adjusted to around 5.0 for achieving the electrostatic stabilization.
2.2.3 Preparation of support. The prepared a-Al 2 O 3 slurry was gradually added into the prepared boehmite sol by titration and stirring. Aer drying and sieving, the powders were placed in a disk-shaped module with diameter of 27.0 mm and thickness of 3.5 mm. The green support body was obtained by applying 50 MPa pressure on the disk-shaped module. The green support body was then sintered in air with the schedule that from room temperature to 600 C with heating rate ¼ 2 C min À1 and soaked for 1 hour, subsequently heated at 10 C min À1 to 1550 C and soaked for 3 hours. Aer being sintered at 1550 C for 1 hour and polished by diamond polishing paste (45 mm), the supports were obtained.

Preparation of Al 2 O 3 /PVA slurry and PVB solution.
PVA solution was obtained by dissolving 2.4 g PVA in 21.6 g deionized water. Then, 10.2 g a-Al 2 O 3 powder (D 50 ¼ 0.52 mm), 25.8 g deionised water, 0.36 g PAA, 0.12 g silicone-based defoamer and 0.18 g uorocarbon surfactant were mixed in a ball mill for 15 minutes. Aer that, the prepared PVA solution was added in the ball mill and mixed for another 30 minutes. The air bubbles were removed from the Al 2 O 3 /PVA slurry with a vacuum pump. PVB solution was obtained by dissolving 5 g PVB powder in 95 g ethanol.
2.3.2 Preparation of precursor lm. As shown in Fig. 1a, the PVB lm was prepared by coating the PVB solution on a glass plate using a lm coating machine (Automatic Film Application, BEVS Industrial Co., Ltd., China). The thickness of the PVB lm was kept at approximately 15 mm with a lm thickness controller (the accuracy of this lm thickness controller ¼ AE2 mm). Aer that, the PVB lm was dried at room temperature for 30 minutes. As shown in Fig. 1b, the Al 2 O 3 /PVA slurry was then coated on the dried PVB lm with the lm coating machine. Using the lm thickness controller, the thickness of the Al 2 O 3 /PVA layer was also controlled to be about 15 mm, 20 mm, 25 mm, 30 mm, 35 mm and 40 mm, respectively. The Al 2 O 3 /PVA layer was dried at room temperature for 12 hours. Subsequently, the precursor lm that consists of the Al 2 O 3 /PVA layer and the PVB layer was separated from the glass plate.
2.3.3 Coating the precursor lm on the support. Soluble polymer is oen mixed in inorganic particle suspension to provide mechanical strength and adhesiveness to particulate coatings. 19 In the precursor lm ring method, PVA was used as a binder to pasting the precursor lm and the support. The precursor lm (Al 2 O 3 /PVA and PVB composite lm) was coated on the support as schematically shown in Fig. 2. The support was moistened ( Fig. 2a and b) using water spray.
Then, the precursor lm was pasted on the support ( Fig. 2b  and c). When the lm contacted the wet surface of the support the PVA in the Al 2 O 3 /PVA layer would partially become wetting, which provided an adhesive force for pasting the precursor lm on the support. Aer drying at room temperature for 24 hours the precursor lm with the support was calcined in air with the schedule that from room temperature to 600 C with heating rate ¼ 2 C min À1 and soaked for 1 hour, subsequently heated at 5 C min À1 to 1280 C and soaked for 1 hour. During the calcination process PVA, PVB and other organic components in the precursor lm were oxidized and volatilized. The remaining Al 2 O 3 was sintered and the separation layer was formed (Fig. 2d).

Characterizations
The surface tension of the Al 2 O 3 /PVA slurry was measured using a contact angle measuring instrument (FM40Mk2 Easy Drop, KRÜSS GmbH, Germany) at 25 C. The morphology of the Al 2 O 3 / PVA lm was observed using a stereo microscope (L330-M1000, AOSVI, China). The thickness of the membrane and the morphology of the membrane and the support were determined with a scanning electron microscope (SEM) (ZEISS EVO 18, Germany). The pore size distribution and the porosity of the support were tested applying a mercury porosimetry (MIP) (Micrometrics AutoPore 9510, USA). The contact angle of mercury for MIP was set as 130 .
Water treatment like drinking water product, municipal wastewater treatment, and food and beverage industries is an important application of microltration membranes. 1 The present study concentrates on measuring the water permeance of membranes. The water permeance of membrane and support was measured using a fully automated uid and gas handling systems (OSMO Inspector 2.0, Poseidon, Convergence, Netherlands) at 25 C. As the pressure of the deionized water was gradually increased, the corresponding ow through the membrane was recorded. The water permeance was calculated based on the relationship between the transmembrane pressure and the ow. To avoid non-stationary transient effects, the membrane and the support were saturated with deionized water before the pressure was applied. The pore size distribution, largest pore size, and average pore size of membranes were tested and calculated according to ASTM: F316-03. 20 The exural strength of support is determined by three-point bending test according to ASTM: C1161-13. 21 In three-point bending test, the support is cut into rectangle-shaped bar with size of 26 (length) Â 4 (width) Â 3 (thickness) mm 3 . The span of three-point bending test is xed at 20 mm.

Results and discussions
3.1 Performance of the support Table 1 lists the exural strength of supports in this work and literature. The strength of ceramic supports are normally related to the porosity and the pore size. The support in this   work shows higher strength than that in Qin et al. 18 This is because the support in this work has smaller porosity than that in Qin et al. 18 The support in this work shows lower strength than that in Dong et al. 22 This is probably because the support in this work has larger average pore size than that in Dong et al. 22 Fig. 3 shows the morphology of the support surface viewed using SEM. It can be seen that the Al 2 O 3 grains are spherical and with the size of around 10 mm. Fig. 4 shows the pore size distribution and the porosity of the support measured using MIP. As shown in Fig. 4a, the pore diameter ranges from 0.82 mm to 9 mm, and the mean diameter is approximate 4.6 mm. The support shows a narrow pore size distribution (Fig. 4a) and the porosity of the support is 0.31 (Fig. 4b).
As shown in Fig. 5, the ux and the transmembrane pressure through the support exhibit a linear relationship (eqn (1)), which is similar to Darcy's law for the ow through a porous material (eqn (2)). 12,[23][24][25] The slope (17 895 L m À2 h À1 bar À1 ) in eqn (1) represents the water permeance of the support. With J ¼ 0 L m À2 h À1 , eqn (1) gives DP ¼ 0.093 bar. This pressure (0.093 bar) is called threshold pressure gradient (TPG). In some cases for the low-velocity ow through a porous material, the ow might not occur until the transmembrane pressure increases to the value of TPG. 26,27 where, J is the ux through support (L m À2 h À1 ), DP is the transmembrane pressure through support (bar).
where, K m is the permeability of the membrane (m 2 ), m is the uid viscosity (Pa s), L is the support thickness (m), DP is the transmembrane pressure (Pa). Table 2 lists the water permeance of the supports for microltration membrane in this work and literature. The water permeance of support in this work shows same order of magnitude in comparison with the water permeance of supports in literature. In general the average pore size, the porosity and the support thickness are important factors which inuence the water permeance. The support in this work shows similar average pore size, porosity and support thickness to that in Dong et al. 8 Accordingly, the support in this work shows similar water permeance to that in Dong et al. 8

Inuence of surfactant on the formulation of Al 2 O 3 /PVA lm
As shown in Fig. 6, if the surfactant is added in the Al 2 O 3 /PVA slurry a smooth Al 2 O 3 /PVA lm can be obtained (Fig. 6a), while if the surfactant is not added in the Al 2 O 3 /PVA slurry the prepared lm is uneven (Fig. 6b). This is because the addition of surfactant changes the surface tension of the Al 2 O 3 /PVA slurry. As a result the viscosity and the dispersion degree of Al 2 O 3 powder are affected which makes the alumina powder disperse

Function of the PVB layer
The precursor lm consists of Al 2 O 3 /PVA layer and PVB layer. The Al 2 O 3 /PVA layer is calcined as the separation layer of membrane, while the PVB layer is used to inhibit the possible crinkles in the precursor lm when the precursor lm is pasted on the support. As mentioned in Section 2.3, during the process of pasting the precursor lm on the support, the Al 2 O 3 /PVA layer of the precursor lm will absorb the moisture on the support surface. The PVA in the Al 2 O 3 /PVA layer will partially become wetting, which provides an adhesive force for pasting the precursor lm on the support. However during this process the PVA will also become swelling. As a result, crinkles might form in the precursor lm. Aer calcination, the crinkles will be more severe. Because the PVB layer has good toughness and ductility, and is insoluble in water, it can provide a horizontal restraining force to inhibit the formation of crinkles. Fig. 7 shows that the membrane fabricated with PVB layer exhibits no signicant defects (Fig. 7a), and the membrane fabricated without PVB layer has obvious crinkles (Fig. 7b).     exible. The exibility of the precursor lm mainly depends on the PVA in the Al 2 O 3 /PVA layer and the PVB layer, because both the PVB and the PVA are exible. The exibility of the precursor lm provides a potential for applying the precursor lm ring method in fabricating curved membranes e.g. cylindrical membranes in addition to at membranes. Fig. 9 shows the morphology of the membrane surface aer calcination. The membrane is smooth and no signicant defects can be observed.
3.4.2 Pore size distribution. The pore size distributions of the membranes were measured and determined according to ASTM: F316-03. 20 Fig. 10 shows the measured pore size distribution of the membrane. The membrane shows a narrow pore size distribution, which is benecial to improve the selectivity efficiency.
3.4.3 The thickness of the separation layer. Fig. 11 shows the cross section of the membrane (thickness of the Al 2 O 3 /PVA layer of the precursor lm is around 15 mm). The small particles in the separation layer do not penetrate to the pores of the support. Fig. 12 shows the relationship between the thickness of the separation layer and the thickness of the Al 2 O 3 /PVA layer of the precursor lm. With increasing the Al 2 O 3 /PVA layer thickness from 20 mm to 40 mm, the separation layer thickness increases from 10.7 mm to 30 mm. The separation layer of commercial microltration membranes is in normally in the range of 10-20 mm. 5 In other words, the thickness of the separation layer fabricated with the precursor lm ring method meets the commercial requirement. In addition, it is found that the thickness of the separation layer (L s , mm) linearly increases with the thickness of the Al 2 O 3 /PVA layer of the precursor lm (L p , mm) (eqn (3)). The gradient in eqn (3) represents the volume deformation of Al 2 O 3 /PVA layer during the drying at room temperature and calcination at 1280 C.    This linear relationship illustrates that it will be convenient to control the thickness of the separation layer with the precursor lm ring method proposed in this study. As mentioned earlier, the thickness of the separation layer prepared by the traditional dip-coating method is affected by a variety of parameters, including viscosity, density and surface tension of the suspension; porosity, surface roughness and radius of the support; the contact time; and the withdrawal speed, etc. 5 Hence it will be complex and difficult to control the thickness by adjusting so many parameters.
From the viewpoint of controlling the thickness of the separation layer, the precursor lm ring method proposed in this study is more attractive than the traditional dip-coating technique.
3.4.4 Water permeance. The water permeance of membrane is calculated as: where, P M ¼ the water permeance of membrane, L m À2 h À1 bar À1 ; J ¼ the ux through the membrane, L m À2 h À1 ; DP ¼ the transmembrane pressure through the membrane, bar. Fig. 13 shows the calculated water permeance of the membrane with different separation layer thickness (DP ¼ 1 bar). It can be seen that with increasing the separation layer thickness from 10.7 mm to 30.0 mm, the water permeance of the membrane decreases from 3890 L m À2 h À1 bar À1 to 701 L m À2 h À1 bar À1 . This trend is also similar to the Darcy's law for the ow through a porous material (eqn (2)): the water permeance of the membrane (J/DP) is reciprocal to the membrane thickness (L). Table 3 lists the permeance of membranes for micro-ltration in this work and literature. The average pore sizes of these membranes are at the same order of magnitude (0.1-0.5 mm). The membrane in this work shows relatively high water permeance in comparison with other reports. For example, for the membrane with average pore size ¼ 0.18 mm and separation layer thickness ¼ 10.7 mm, the water permeance arrives at 3890 L m À2 h À1 bar À1 . While Qin et al. 16 reported that for a membrane with average pore size ¼ 0.24 mm and separation layer thickness ¼ 10 mm, the water permeance was 1410 L m À2 h À1 bar À1 . In Dong et al., 8 the membrane consisted of three layers: support layer, intermediate layer and separation layer. The average pore size of the intermediate layer was 0.85 mm which was close to the average pore size of the separation layer (0.54 mm). This intermediate layer reduced the water permeance of the membrane system. For the membrane systems in Wang et al. 10 and Qin et al., 16 intermediate layer was avoided. However because dip-coating process was used in Wang et al. 10 and Qin et al., 16 the water permeance might be reduced. For example, in Wang et al. 10 the size of the titania powder for fabricating separation layer was 0.37 mm, and the average pore size of the support layer was 2.1 mm. The small titania powder would probably penetrate to the big pores of the support layer during the dipcoating process. In the proposed precursor lm ring method, the membrane system was obtained without intermediate layer. What's more, because the dip-coating process was avoided the small grains in the separation layer didn't access to the big pores of the support. Because of above two reasons, the obtained membrane system shows relatively high water permeance. Table 4 lists the properties of support and separation layer. It is noted the porosity of the serration layer was not measured in this study.

Conclusions
This study proposes a novel method to fabricate ceramic microltration membranes. A precursor lm is prepared independently of the supports by a lm coating machine. This precursor lm consists of two layers: Al 2 O 3 /PVA layer and PVB layer. The precursor lm is pasted on the support and red to obtain the microltration membrane. Based on the experimental results, the following conclusion can be drawn: The precursor lm (Al 2 O 3 /PVA and PVB composite lm) is exible, which provides a potential for applying the precursor lm ring method in fabricating curved membranes e.g. cylindrical membranes in addition to at membranes.
The separation layer thickness (L s , mm) and the thickness of the Al 2 O 3 /PVA layer (L p , mm) follow the linear relationship: L s ¼ 0.76L p . This shows that it is convenient to control the thickness of the separation layer by using the precursor lm ring method. With increasing the Al 2 O 3 /PVA layer thickness from 20 mm to 40 mm, the separation layer thickness increases from 10.7 mm to 30 mm. The separation layer of commercial micro-ltration membranes is in normally in the range of 10-20 mm. 5 In other words, the thickness of the separation layer fabricated with the precursor lm ring method meets the commercial requirement.
Because intermediate layer and dip-coating process are avoided in the precursor lm ring method, the fabricated membrane shows high permeance. For the fabricated membrane with average pore size ¼ 0.18 mm and separation layer thickness ¼ 10.7 mm, the water permeance reaches 3890 L m À2 h À1 bar À1 .
In view of the above advantages, the proposed precursor lm ring method is attractive for fabricating ceramic micro-ltration membranes.

Conflicts of interest
There are no conicts to declare.