Preparation of high permeable alumina ceramic membrane with good separation performance via UV curing technique

The traditional dip-coating method for preparation of ceramic membranes requires a long drying time and easily produces drying defects. In this work, an improved dip-coating process was proposed. The UV curing technique was utilized to avoid crack formation and agglomeration of ceramic particles, for drying to be completed in a few minutes. Photosensitive resin and a photoinitiator were added into the aqueous ceramic suspension. Under the action of free radicals excited by ultraviolet light, a giant network formed in the green membrane within a short time which limits the migration of membrane particles. Experiments were performed to explore the influence of UV curing process on membrane properties and the optimum preparation conditions were obtained. Following a rapid drying treatment and firing, crack-free membranes were prepared, which exhibited a narrow pore size distribution centered at approximately 65.2 nm and a water permeance of 887 ± 48 L m−2 h−1 bar−1. The largest pore size of the membrane was 85.7 nm while it could filter out 98.2% of the 100 nm monosize PS microsphere and the 60.1% of 60 nm, indicating its potential application in both membrane production efficiency and separation accuracy improvements.


Introduction
In recent decades, porous ceramic membranes have been successfully used in many industry areas, such as waste solution treatment, 1 oil concentration 2 and gas separation process 3 etc. The formation of ceramic membranes through dip-coating on ceramic supports is a very common procedure to prepare microand ultra-ltration membranes. [4][5][6][7] The crucial part of dipcoating process is the preparation of membrane-forming suspension which mainly consists of ceramic powders and other additives such as binders, dispersants and plasticizers. With higher environmental protection criteria implemented, aqueous membrane-forming suspension is becoming increasingly common for being eco-friendly and of low-cost compared to organic solvents. Adversely certain disadvantages are commonly confronted in a water based system, leading to the poor performances of the nal products. These negative consequences include a long drying time and high crack sensitivity, due to the huge surface tension of water. 8 Once a wet membrane containing suspended submicronsized particles is coated on the porous support, the shrinkage occurs with the loss of water by evaporation. Further evaporation exerts compressive capillary force on the particle network under induction of surface tension. 9 The support is free of contraction while the membrane generally binds to the support surface, which gives rise to the transverse tensile stresses. Cracks are formed spontaneously when the magnitude of the tensile stress exceeds a critical value. 10 A commonly utilized method to reduce drying defects is to add membrane forming agents in the membrane-forming suspension. Researchers have studied the drying behavior of wet membranes with polymer binders coated on porous supports. 4,11,12 The results demonstrated that a sufficient amount of reasonable polymer binder (such as PVA, PVP and MC) in the green membrane would contribute to avoid crack formation resulting from drying shrinkage. 4 This was mainly due to the combination of particles with the polymer binder, induced by the hydrogen bonding of polymer binder molecules, 13 which led to the membrane tensile strength improvement. Even though the addition of polymer binders may eventually form a crosslinking framework, the early strength of a wet membrane is not sufficient to resist the cracking under rapid dehydration. Therefore, the solvent removal rate is a fatal factor during drying, which is controlled through the ambient temperature and humidity adjustments. For this reason, a long drying time (almost beyond 12 h) and strict drying conditions are required at the initial stage of drying, [14][15][16] which is a timeconsuming and cumbersome process.
At present, the UV curing technology has been widely used in the coatings industry [17][18][19] and in 3D printing 20 due fact that it promotes fast polymerization and solidication of the prepolymer network. In addition to high production efficiency, the other advantages of light curing technology are high efficiency energy utilization as well as no solvent emission absence, signifying that it is safe, low-cost and pollution-free. Caroline Durif et al. 11 demonstrated the possibility of solvent-free tape casting through UV curable binder. Unfortunately, few studies have been focused on UV curing process used in the preparation of ceramic membranes.
This paper reports a fast membrane process by adding appropriate amount of photosensitive resins and corresponding photoinitiators in the conventional membrane-forming suspension. The selected resins are water dispersible and can also be used as thickening agents. The curing step was performed directly aer dip-coating. Following a short-time irradiation of UV light, the membrane solidication immediately and becomes tough. Subsequently the organic part provided the strength of the green membrane was directly removed through an appropriate thermal treatment prior to sintering. The mechanism of UV curing technique in preparation of alumina ceramic membrane is studied and the effects of content of photosensitive resin, ring temperature and drying method on the pore size distribution, water permeation and microstructure of membrane are also discussed. Compared with the traditional preparation process, this method has considerable advantages that eliminating the high shrinkage and cracking risk caused by solvent evaporation while shortening the preparation cycle greatly. This method appeared promising for the preparation of a high-quality and cost-effective ceramic membrane.

Materials
Disc alumina supports with the sizes of 25 mm in diameter and 1.5 mm in thickness were pre-treated in a 5% hydrochloric aid solution and followed by the thermal treatment at 600 C for 1 h. The disc supports were stored in a vacuum dryer and purged with nitrogen before use. The characteristics of the disc alumina supports are listed in Table 1, while the corresponding pore size distribution and microstructure are presented in Fig. 1.

Fabrication of alumina ceramic membrane
In accordance to the formulas in Table 2, the alumina powder, the monomers, the photoinitiator and the dispersant were added to a planetary mill jar and milled for 60 min. Then PVA and glycerol were added to this mixture and stirred for another 30 min. The slurry was treated for 1 min with a homogenous emulsifying machine (IKA T18, Germany) at a speed of 8000 rpm to break the agglomeration of nano-alumina particles. The as-obtained ceramic slurry was degassed at room temperature under the vacuum degree of 3 kPa for beyond 30 min. A uniform coating suspension was obtained.
The forming of the green membrane follows: the support, of which one side had been covered with adhesive tape, was dip coated in the slurry for 16 s at a withdrawal speed of 5 mm s À1 . With a high pressure mercury lamp, the undried membranes were directly exposed to the light source for 30 s (RX 2 kW, 400 mm, Dongguan Ergu Photoelectric Technology Co., Ltd). Subsequently to curing, the sample was placed in an oven for drying at 150 C. Following drying, the green membranes were moved into an electric furnace and red according the temperature schedule: from room temperature to 600 C, the temperature was increased at a rate of 2 C min À1 and soaked at 600 C for 1 h to remove the organic matters, aer that it was raised to the ring temperature at a rate of 5 C min À1 and held for 2 h at ring temperature.

Measurement methods
The hardness of the green membrane was measured with the pencil scratch tester (QHQ-A, China). The weight loss test of the wet membrane is performed by the wet lm prepared on the glass substrate, which was peeled from the porous support, to avoid interference caused by solvent evaporation in the support. The thermal analysis of the wet membrane was done by a thermal analyzer (Netzsch STA 449C, Germany). The absorbance spectra were measured with the Lambda 950/UV/Vis/NIR spectrophotometer (Perkin-Elmer, America). The morphology of membrane was observed by a SEM (ZEISS EVO 18, Germany). Fourier Transform Infrared Spectroscopy (FTIR) spectra were recorded using a VERTEX 70 (Bruker, Germany), in Attenuated Total Reection (ATR) mode. The water permeance of the membrane was measured with the Fully Automated Fluid and Gas Handing Systems OSMO Inspector 2.0 (Poseidon, Convergence, Netherlands). The membrane permeance was determined by collecting the permeation in a mass owmeter of the OSMO Inspector and timing the collection period. To avoid non-stationary transient effects, the membranes were saturated with deionized water (18.2 MU) before the pressure was applied.
The effective ltration area of each sample was 2.64 Â 10 À4 m 2 by measurements and calculations. The pore size distribution of the disc alumina support and the prepared membrane were characterized with a capillary ow porometer (Porolux 500, IB-FT GmbH, Germany). The membranes were fully wetted with the commercial low surface tension liquid Porel (surface tension 16 dyn cm À1 ). The measurements consisted of a wet-run program and a dry run program, all the measurements were carried out following the procedure described in literature. 21 The mean pore radius r min (nm) and maximum pore size r max (nm) were determined with the computer soware.
The rejection performance of the membrane was tested through different sizes of monosize PS microspheres ltering in aqueous dispersion with the concentration of 100 mg L À1 . To drive the ux through the membrane, a tangential ltration described further was applied to the permeated uid to maintain a transmembrane pressure of 1 bar and the cross ow velocity at 0.9 m s À1 . The permeate concentration was determined with the Lambda 950/UV/Vis/NIR spectrophotometer (Perkin-Elmer, America) and compared to the initial concentration, to obtain the rejection rate.
where R (%) is the rejection rate, C p (mg L À1 ) the permeate concentration, and C o (mg L À1 ) the initial concentration.

Investigation of UV curing process
The polymerization of the green membrane under UV irradiation is illustrated in Fig. 2. The energy produced by the UV light could activate the photoinitiator to generate the free radical and consequently break the C]C bond of the photosensitive resin, for the resin to polymerize rapidly with chain reactions, forming a strong network within a few seconds to a few tens of seconds in a general system. 19 This polymerization led to a rapid hardening of the green membrane.
In order to investigate the change of membrane composition during UV curing, FTIR was performed and the results are presented in Fig. 3a. The evolution of the FTIR spectra was recorded before and aer the irradiation under UV light. The carbon-carbon double bond of the resin was characterized by the peaks in the region of 1600-1660 cm À1 , relevant for the C]C stretching as well as the region of 1400-1430 cm À1 , representative of the twisting. The absorbance value of those two peaks obviously decreased aer UV treatment, which indicates a sharp reduction in the number of double bonds aer an exposure. This indirectly reects that the crosslinking reaction did occur in the membrane. Corresponding to the transformation of structure, the contact angle also demonstrated the great difference before and aer UV treatment, shiing from 15.3 to 65.8 , as is shown in the Fig. 3b and c.
Another intuitive property change is hardness of the membrane measured by standard pencil hardness tester. As is shown in Table 3, within 10 s of the initial exposure, the hardness of the membrane using formula S25 did not have a signicant change. At this point, the photoinitiator was not able to generate sufficient radicals to break the C]C bond in a relatively short irradiation time, which greatly inuences the crosslinking degree of green membrane. Actually, in the 10-20 s, the hardness has a rapid improvement, from 1H directly up to nearly 3H. Aer 20 s, an increase in hardness gradually slowed down and eventually stayed within 3H. Therefore the curing time was set to 30 s.
In order to evaluate the effect of thermal radiation during UV irradiation on the consolidation of wet membrane, the membrane without photoinitiator (formula S25 0 ) was also treated under the same process as the one mentioned above. The results showed that the wet membrane without photoinitiator did not have any strength aer UV irradiation of 30 s, which indicated that UV curing played a key role in the hardening of the wet membrane.
Drying behavior of the UV curing membrane was studied. As for the supported membrane, drying occurs from the top surface and there is no noticeable difference in the drying behavior of wet membrane coated on porous support or nonporous substrate. 22,23 Therefore it was characterized by taping casting membrane on glass substrate. As is shown in Fig. 4 shows, the initial weight loss of solvent was 44.56%. This part of weight loss is due to the rapid absorption of solvent in the slurry by the pore of support under capillary force. Subsequently, 14.67% of solvent was lost in the process of UV curing. Aer 30 minutes of drying at 150 C, almost all the solvent were removed from the membrane.
As is shown in Fig. 5, the exothermic loss of about 2.0 wt% in the range of 150-250 C was attributed to the decomposition of polyvinyl alcohol. The drastic weight loss occurred at temperatures from 300 C to 500 C was mainly due to the decomposition of photosensitive resin. Besides, thermal analysis showed that organic matter was almost removed before 650 C.

Optimization of membrane structure and properties
Based on the method of UV curing, the content of photosensitive resin cured components (hereinaer referred to as prepolymer) and the sintering process were both controlled in order to prepare alumina micro ltration membranes with better performances (Table 4).  Fig. 4 Weight loss as a function of drying time for the wet membrane.  The hardness of the lm reects the integrity of the crosslinking network, and the results show that the content of prepolymer increases as gradually and obviously as the hardness does. However, there is a critical value above which the lm hardness no longer depends on the prepolymer content. From the morphology of the cured green membrane characterized by SEM (Fig. 6), cracks in the surface of membranes can be observed when the prepolymer content is 5 wt% (Fig. 6a). While smooth and crack-free membrane is obtained with an increase of prepolymer content, which improves the hardness of the membrane. Hence, the amount of prepolymer has a great inuence on UV curing process.
The content of prepolymer also affects the viscosity of the membrane-forming suspension and the thickness of the membrane. As shown in Fig. 7a, the membrane thickness increases from 15 mm to 26 mm when the prepolymer/Al 2 O 3 ratio increases from 5 wt% to 30 wt%.
It is generally known that although the membrane permeation is greatly affected by porosity and thickness, while the viscosity of membrane forming suspension could change both properties in the green membrane. Besides, the higher viscous suspension prevents ne particles from suction into supports, which leads to increase of permeation. 24 As shown in Fig. 7b, permeance of the membrane increases with weight ratio of prepolymer/Al 2 O 3 at the range from 5 wt% to 25 wt%, which could be attributed to increment of pore size and porosity. However, there exists a turning point in the curve, presenting that when the ratio was up to 30 wt%, the permeance saw a modest reduce. This is mainly due to a surge in membrane thickness, which surpasses the effect of porosity increment, and the pore size increases slightly.
As shown in the Fig. 8, the most frequent pore size of the membranes increases with elevation of the relative amount of prepolymer, ranging from 80.9 nm to 113.5 nm and the pore  Paper size distribution narrows down gradually. It concludes that UV curable assisted ceramic membranes have a relatively narrow pore size distribution. This is mainly due to the rapid polymerization of prepolymer under UV irradiation. According to Gonzalez, 25 the adsorption of PVA molecules onto the alumina particles surface reduces their migration in the suspension, which leads to a well dispersion of alumina particles in green membrane before UV cured. The prepolymer in the green membrane is to form a strong network aer polymerization, which helps to prevent crack formation from drying shrinkage.
Based on the above analysis, prepolymer content is a key factor in adjusting the suspension viscosity and the thickness of the membrane.
The morphology of the sintered membrane characterized by SEM is shown in Fig. 9a, smooth and defect-free membrane with uniformly distributed pores was obtained. The thickness of the membrane is about 20 mm (Fig. 9b). The interface between the separation layer and the support is clearly visible, and the membrane and the support are tightly bonded.
Under the condition of setting the prepolymer/Al 2 O 3 ratio to 25 wt%, the inuence of ring temperature on the pore size distribution was investigated. The pore size distribution of the membrane red at 1100 C for 2 h was measured and calculated through the gas bubble pressure method at 25 C according to the ASTM F316-03(2011) standard. The relationship between the nitrogen ow and the trans-membrane pressure is shown in Fig. 10a. The gas ow of the wet membrane occurred at the rst   bubble point of 5.3 bar, which corresponds to the largest membrane pore size of 85.7 nm. As the trans-membrane pressure increased, more pores were opened and the gas ow increased nonlinearly. When the trans-membrane pressure is increased to 7.0 bar, the gas ow increased sharply, indicating that the most frequent pore size of the membrane is about 65.2 nm. Aer all of the membrane pores are opened, the gas ow increased linearly with the trans-membrane pressure according to the Hagen Poiseuille equation. The calculated pore size distributions of the membranes are shown in Fig. 10b. The largest pore size increases from 85.7 nm to 222.3 nm as the temperature increased from 1100 C to 1300 C. Meanwhile the most frequent pore size rises from 65.2 nm to 204.6 nm while the pore size distribution broadens at the same time. The main reason is that the growth of grains leads to the formation of large pores and the elimination of small pores when the membranes are red at higher temperature. Fig. 11 demonstrates that there is no obvious change in the grain sizes of the membranes red below 1150 C (Fig. 11b and c), compared with the size of raw powder (Fig. 11a). However, as shown in Fig. 11d, the obvious grain growth in the membrane surface red at 1200 C has been observed, the size of most grains is located in the range from 300 nm to 500 nm. The change law of membrane pore size is in agreement with the theory of constrained sintering 26,27 (Fig. 11b-f). Fig. 12 presents the effects of ring temperature on the membrane thickness, the most frequent pore size and the water permeance of the membrane. As is shown in the Fig. 12a, no great changes occur in the thickness of the membranes at different ring temperature ranging from 20 mm to 23 mm. While the change in ux is still a good illustration of the change in pore size distribution (Fig. 12b).  Based on the above analysis, the ring temperature controlling is an effective method to regulate the pore size distribution and water permeance of the membrane on largescale. It is suitable to prepare narrow pore size distribution and high ux alumina ceramic membrane at a relative low ring temperature.

Controlling of pore size distribution of membrane
In order to illustrate the advantage of UV curing process in preparation of alumina ceramic membrane. The green membrane without photoinitiator was used as contrast samples (formula S25 0 ) which had been prepared under the same process of UV curing membrane. The surface morphology of sintered membrane with and without photoinitiator was presented in Fig. 13, the agglomeration of particles clearly occurs in the contrast sample which leads to the appearance of pinholes. On the contrary, smooth and crack free membrane surface with uniformly distributed pores can be observed in the red UV curing membrane, which indicates that the crosslinked networks play a critical role in resistance to alumina agglomeration and crack formation during the drying process.
The pore size distributions of the membranes prepared with and without photoinitiator are shown in Fig. 14, which illustrates the apparent difference between them. The pore size of the membrane prepared without photoinitiator presents a wide distributions with several peaks, but the membrane prepared with UV curing technology presents a sharp pore distribution curve.
Hence, we believe that the macromolecular network formed by rapid crosslinking under UV curing provides toughness to green membrane which effectively resists the tress resulting from dehydration shrinkage. Different from the traditional method, this is a new one that a crosslinking network is formed    to prevent the volume effect aer drying. Therefore, the drying process of UV curing membrane is much safer and faster than the traditional process.

Characterization of ltration properties
To evaluate the separation performance of the alumina ceramic membrane prepared by UV cured process, the TiO 2 (80 nm, rutile) slurry with concentration of 100 mg L À1 and the membrane with a water permeance of 887 AE 48 L m À2 h À1 bar À1 and the most frequent pore size of 65.2 nm were used. Filtration tests of TiO 2 slurry were performed under the conditions of temperature at 25 C, trans-membrane pressure of 1 bar and cross-ow velocity of 0.9 m s À1 .
As shown in Fig. 15, the permeate ux of the membrane maintained at a stable value of 300 L m À2 h À1 bar À1 aer 100 minutes running. The curve of permeate ux to ltration time might be explained by the formation of a steady cake layer during ltration process which caused the constant resistance permeation of the TiO 2 slurry. 28 Meanwhile, the rejection rate remained almost the same (over 99.2%) through the whole ltration experiment. The TiO 2 contents in the suspension of permeate side were less than 1 mg L À1 , which indicated well separation performances for the suspension formed by nanoparticles with the size of $100 nm.
Monodisperse microspheres through dispersion display an extremely narrow particle size distribution, for the microsphere rejection to present a good correspondence with the membrane pore size. For this reason, we can indirectly describe the actual pore size distribution of the membrane by its rejection rate. A 100 mg L À1 concentration of microsphere dispersions was prepared with 60 nm and 100 nm in diameters of particles respectively. The rejection performance of the membrane red at 1100 C is illustrated in Fig. 16. This membrane could lter out 98.2% of the monosize PS microspheres of 100 nm and 60.1% of the 60 nm monosize PS microspheres, which meant that the membrane had a relative narrow pore size distribution, in accordance with the testing results of the gas bubble pressure method.
To sum up, the UV curing assisted drying method has shown great advantages in the preparation of ceramic membranes. This approach can greatly reduce the drying time and shorten the preparation period while maintaining the superior performance of the ceramic membrane. Its performances are listed with those of other literature as shown in Table 5.

Conclusion
In this work, a rapid preparation of high permeable and reliable alumina membrane has been realized by UV curable technique. The network formed by cross-linking of photocurable resins resulted in rapid solidication of the wet membrane. In this way, the green membrane might obtain its strength at the initial stage of drying and eliminate the drying defects signally. The membrane prepared by this method had a permeance of 887 AE 48 L m À2 h À1 bar À1 when its most frequent pore size was 65.2 nm. The membrane could lter out 98.2% of the monosize PS microspheres of 100 nm and 60.1% of the 60 nm monosize PS microspheres, which meant that a narrow pore size distribution was achieved, corresponding well with the testing results of the gas bubble pressure method. Compared with the uncured method, the UV curing process contributes greatly to avoid particle aggregation and drying defects in drying process. This energy saving and environmental approach could highly reduce the drying time and shorten the preparation period while maintaining the superior performance of the ceramic membrane, which could be universally applied in more industries in the future.