Fast preparation of a polydopamine/ceramic composite nanofiltration membrane with excellent permselectivity

Fabrication of a dense polymer/ceramic composite membrane with high permeability remains a great challenge. In this study, a highly selective polydopamine (PDA)/ceramic composite nanofiltration (NF) membrane was prepared by using an Al2O3 ceramic membrane with pore size of 0.1 μm as the support layer. In order to improve the membrane formation rate, KMnO4 was introduced to oxidize the dopamine to improve the reactivity, and Na2CO3 was used to adjust the pH value of the dopamine solution. When the addition amount of KMnO4 is 0.2 g L−1 and that of Na2CO3 is 1 g L−1, a functional layer can be formed within 10 min. PDA and polyethyleneimine (PEI) were added to the functional layer to adjust the selectivity of the composite membrane. The composite membrane showed a rejection of 99.7% towards Congo red dye with a high flux of 165 L (m2 h bar)−1 at ambient temperature. After 3 h treatment with Congo red, the fouling resistance of the membrane was improved compared with that of the ceramic based membrane. The surface morphology and composition of the composite membrane were also characterized with scanning electron microscopy (SEM) and X-ray photoelectron spectroscopy (XPS), which confirmed the successful preparation of the PDA/ceramic composite membrane.


Introduction
With rapid urbanization, water environment problems are becoming more and more serious. There are many ways to treat waste water, including adsorption, chemical oxidation, precipitation and ltration. [1][2][3] However, waste water treatment with low energy consumption is still urgently needed. Membrane separation technology has the advantages of low energy consumption, small footprint and easy integration, and is widely used in the petrochemical industry, biological medicine industry, waste water treatment and other elds. Nanoltration (NF) with high ux and selectivity is widely used in the precise separation of small molecules during 200-1000 Da and ions with high valence. It is generally believed that the separation mechanism of NF membrane is the synergistic effect of sizesieving and Donnan repulsion. 4 Most commercial NF membranes are polymeric, because the pore size and surface property of polymer membrane are easy to adjust, while ceramic NF membranes are difficult to obtain because preparation of dense functional layers without defect needs special technology. 5 The common functional layer materials of commercial composite NF membrane are polyamide and sulfonated polysulfone, while the supporting layer material is polysulfone, polyvinylidene uoride or polyimide. 6 However, these kinds of polymer membrane oen have some problems, such as low solvent resistance, acid and alkali resistance, oxidation resistance or thermal resistance. In contrast, inorganic membranes such as Al 2 O 3 membrane, zeolite membrane, b-SiAlON membrane and TiO 2 membrane, in many cases have high chemical stability, enhanced mechanical strength and high thermal stability, which is of great value in some special separation process. 7,8 Ceramic ultraltration membrane with large pore size and high ux is relatively mature. Nevertheless, low fouling resistance and loose functional layer are still two great challenges to further application of the ceramic membranes. Therefore, an increasing number of studies have been carried out to fabricate ceramic NF membranes with small pore size and hydrophilic surface. In order to change the permselectivity and surface properties of the ceramic membranes, a lot of materials such as hydrophilic polymer, graphene oxide (GO), MXene and nanoparticles have been used to modify the surface of the polymer or ceramic membranes. 9-13 Chen et al. utilized in situ chemical deposition method to tune pore size and fabricate TiO 2 multichannel nanoltration membrane with ceramic ultraltration substrate. Aer modication, the average membrane pore radius was changed from 2.4 nm to 0.9 nm. 14 MnO 2 nanoparticles were incorporated in the ceramic membranes and the prepared membrane showed high removal efficiency of pchloronitrobenzene. 15 Li et al. used the technique of atomic layer deposition (ALD) to tune the pore size of ceramic membranes. With the increase of ALD cycles, the pore size decreased with a gradient porous structure. 16 However, the modication technique oen needs high energy consumption or complicated preparation process. Polymer coating with high hydrophilicity is common for polymeric composite NF membrane preparation. But polymer/ceramic composite membranes are rarely prepared by depositing polymer on ceramic membranes because of low compatibility between polymer and ceramic materials. There is still great challenge in preparing defect-free polymer/ceramic composite membranes.
Dopamine can spontaneously self-polymerize to form bioinspired polydopamine (PDA) under mild alkaline condition. Recently, PDA has attracted great interest for its wide utilization in surface modication because it can work as a glue on different substrates such as polymer, ceramic and metal. [17][18][19] The PDA coating on polysulfone UF membrane can reduce fouling in oily water ltration because it can enhance surface hydrophilicity. 20 Furthermore, PDA coating has plenty of functional groups which can act as a platform for the postfunctionalization. And the excellent hydrophilicity of PDA is meaningful for the antifouling performance of PDA membrane. However, the base ceramic membrane with ne pore size of 5 nm was needed in this experiment. Also, it needed 2 h to form the polymer coating and additional crosslinking. 26 Recently, KMnO 4 was used as a water-soluble oxidant during PDA coating, which can reduce coating time obviously in a single reaction process and endow glass materials with dense functional layer and durable UV shielding property. 27 This process route has important inspiration to prepare polymer/ceramic composite membrane efficiently.
Herein, we propose a very easy process to modify the ceramic membrane surface with PDA which is of great signicance to promote the industrial application of ceramic NF membrane. The base ceramic membrane was rstly immersed in the PDA solution for several minutes to form skin layer and then dried in oven for 10 min to form dense functional layer. In this design, the KMnO 4 acts as a trigger to accelerate the polymerization, which decreases the functional layer forming time obviously. The PDA/ceramic composite membrane showed dense and thin skin layer with excellent permselectivity.

Materials and instruments
Flat sheet Al 2 O 3 ceramic membrane (Jiangxi Bocent Advanced Ceramic, average pore diameter: 0.1 mm) with inner ow channel and dense outer layer has an active membrane area of 842 cm 2 . It was used as supporting membrane to prepare the composite membrane while PDA was used as the functional layer material. Dopamine hydrochloride (AR, Aladdin, shanghai, China) was used as the monomer of PDA. Polyethyleneimine (PEI, MW: 10 000 Da), KMnO 4 , Na 2 CO 3 and Congo red used in the experiments were all analytical purity grade (AR, Aladdin, shanghai, China) without further purication. The performance of the at-sheet composite membrane was tested with a dead-end ltration setup. UV-Vis Spectrometer (752N, Shanghai Youke, shanghai, China) was used to characterize the dye concentration while the conductivity meter (DDS-11A, Leici, shanghai, China) was used to test the salt concentration.

Membrane preparation
In this experiment, PDA functional layer was prepared on the top surface of ceramic membrane by oxidative self-polymerization of dopamine. Firstly, 1 g Na 2 CO 3 and 0.2 g KMnO 4 were added into 1 L deionized water with strong stirring, and then 2 g dopamine hydrochloride (DA) was added into the solution and the purple solution turned quickly into black during 20 min stirring. The clean ceramic membrane was immersed into the solution with xed time to adsorb polymer and form PDA membrane. Then the membrane was taken out and dried at 80°C for 10 min in oven. At last, the product was cleaned with deionized water for further characterization. For the membrane with PEI, the procedures are all the same except a certain amount of PEI was added in the solution during stirring.

Membrane characterization
The morphology and element composition of the membrane were analyzed by SEM equipped with Energy Dispersive X-ray (EDX) (FEI Nova NanoSEM 450, USA). The samples were sputtered with gold before detection to observe the images of the membranes at the voltage of 15 kV. The elemental composition of membrane surface was analyzed with Al K Alpha Gun with a spot size of 500 mm and 10 scans by XPS (ESCALAB™ 250Xi, ThermoFisher).
The performances including ux (F) and rejection (R) of the membrane were characterized by a vacuum membrane ltration set-up. In order to decrease the concentration polarization, an aeration system was xed above the membrane surface. The membranes were tested under 0.1 MPa at room temperature. The concentration of Congo red solution was xed 0.05 g L −1 . The permeation ux (F) was calculated as follows: where W is the total volume of the water or solution permeated during ltration process; A is the valid membrane area; and t is the operation time. Rejection, R, was calculated using the following equation: where C p and C f are the concentrations of the permeate solution and the feed solution, respectively. All the experiments on ux and rejection were tested on three membranes with the average data shown in the paper. For the fouling resistance experiment, the data was tested with one membrane without repetition.

Effect of reaction time on the morphology and performance of composite membranes
The morphologies of ceramic membrane and composite membrane were analyzed by SEM with a magnication of 20 000×. The top surface morphologies of the membranes were shown in Fig. 1. The base ceramic membrane showed Al 2 O 3 particles and large pores in the top surface. When PDA was introduced in the membrane surface at room temperature, part of the membrane was covered by polymer. With the increase of reaction time, denser membrane surface was observed. When the reaction time was increased to 15 min, the surface of the membrane was completely covered by nanoakes. However, the density of the nanoakes is not sufficient to plug the pores of the base membrane. So PDA provides a facile and fast method for the surface modication of ceramic membranes with the existence of oxidant at room temperature. The catechol moieties are used to anchor the ceramic membrane and self-polymerization formed highly cross-linked PDA coatings. In order to demonstrate the success of the membrane forming in the top surface, XPS test was performed to verify the success membrane formation on the support membrane. Fig. 2 showed the electron region of the XPS spectra of the membrane surface and high resolution spectra of C 1s and N 1s. The peaks at 284.8 eV and 399 eV were attributable to C 1s and N 1s, which were generated from PDA membrane. The N1s spectrum of the membrane can be tted into two peaks centered at 398.4 and 399.8 eV, which corresponded to pyridinic-and pyrrolic-like nitrogen, respectively. The small peaks at 74 eV and 120 eV were indexed to Al 2p and Al 2s because the functional layer was very thin and Al 2 O 3 support membrane was detected. The sharp peak located at 531 eV was attributable to O 1s, which may came from Al 2 O 3 and PDA. The tiny peak located at 641 eV was attributable to Mn 2p. As part of KMnO 4 was reduced by dopamine hydrochloride, MnO x was formed on the membrane surface, and the peak value of Mn element was very small.
The pure water ux of the composite membranes showed in  reaction time was 10 min, the composite membrane showed reasonable ux and morphology. So the reaction time was xed at 10 min in the following experiments.

Effect of reaction temperature on the membrane morphology
In order to investigate the effect of reaction temperature on the morphology of the composite membrane, the ceramic substrate membrane was placed in a constant temperature water bath containing 2 g L −1 dopamine solution. The results showed that the temperature had a signicant effect on the membrane formation. As shown in Fig. 4, the surface of the ceramic membrane was covered by PDA and had some small bumps. At 20°C, some very small cracks appeared on the upper surface of the composite membrane, which limited the selectivity of the composite membrane. When the reaction temperature was raised to 40°C, the ceramic membrane was covered with a dense PDA coating. The membrane surface was very smooth without any cracks or pores observed even if the SEM image was magnied 50 000 times. When further heated to 60°C, long cracks appeared on the surface of the membrane, and the alumina particles of the ceramic base membrane were revealed. The cross-section morphologies of the composite membrane were shown in Fig. 5. It is evident that the dense skin layer of the membrane gradually was thickened from 28.6 nm to 60.7 nm. Higher temperatures led to thicker layers but larger defects, which may be caused by low compatibility and big difference in thermal expansion coefficient between the polymer and ceramic materials.

Effect of PEI content on the membrane morphology and performance
The composite membrane with 2 g L −1 PEI was analyzed by SEM as shown in Fig. 6. The top surface of the composite membrane was very dense without pores detected at the high magnication of 50 000. The ultrathin PDA layer was detected on the Al 2 O 3 base membrane with a thickness about 43 nm. With the aid of PEI, the membrane without defects was obtained, which is critical to the high rejection of the composite membrane. And the functional layer is ultrathin which is benecial to the membrane ux. Meanwhile, the elemental mapping images in Fig. 7 showed that N is evenly distributed within the membrane surface with an atomic ratio of 5% which conrmed the effective coating of PDA. The atoms of Al (37%) and O (47%) were observed in the membrane surface because of the Al 2 O 3 base membrane. Trace element Mn was also found because KMnO 4  was reduced to Manganese oxide and which was stuck to the surface of the membrane by PDA.
The effect of PEI content on the performance was studied with different weight ratios of PEI in the PDA solution. As shown in Fig. 8, the blank ceramic membrane has a rejection of 47.7% to Congo red with high water ux of 524 L m −2 h −1 at 0.1 MPa. When PDA was coated on the ceramic membrane, the dye rejection was improved to 80.4% dye while membrane ux was decreased to 429 L (m 2 h bar) −1 . The large pores in the membrane may be blocked by PDA, resulting in a smaller mean pore size which was in accordance with the SEM images. PEI has a signicant effect on the membrane properties. When PEI content was increased to 2 g L −1 , the membrane showed a 99.7% rejection to Congo red with a reasonable ux of 165 L m −2 h −1 . The high dye rejection is due to the very dense layer of the membrane.
As shown in Table 1, PDA coating was time consuming, although the composite membrane showed good permselectivity. 21 In order to improve the membrane forming rate, oxidants such as CuSO 4 /H 2 O 2 was used which can accelerate the reaction obviously. The reaction time was shortened to 45 min. [24][25][26] The surface modication of ceramic membrane with inorganic particles usually required higher temperature and longer time. 28 In our work, the modication time was shortened to 10 min and the modication temperature was very low (40°C), which is very conducive to the production of composite membrane. At the same time, the composite membranes prepared in our work showed high rejection and high ux to Congo red.
The fouling resistance of the composite membrane was also tested by continuous operation with Congo red solution for 180 min. The control experiment was conducted with base ceramic membrane at the same condition. As shown in Fig. 9, the ux of base membrane decreased quickly during the rst 30 min, and the membrane ux decreased to 24% of the initial ux aer 180 min operation. When PDA and PEI were introduced in the functional layer, the membrane ux also decreased with extension of operation time because dye molecule also   adsorbed into the polymer layer. However, the ux decrease velocity was slower obviously compared with the membrane without modication. Aer 180 min operation, the ux was decreased to 57% and 55% of initial ux, respectively. For the PDA/ceramic membrane without PEI also showed good fouling resistance because PDA is also hydrophilic material. The membrane ux was declined to the 49% of initial ux aer continuous 3 h operation.

Conclusions
In this study, the PDA/Al 2 O 3 composite membrane was successfully prepared on the Al 2 O 3 at membrane with pore size of 0.1 mm. The composite membrane was prepared with dipcoating method and PDA worked as the functional layer material. KMnO 4 was introduced to accelerate the rate of polymerization and membrane forming. The membrane preparation time was reduced to 10 min, which signicantly improved the preparation efficiency. The functional layer of the composite membrane was very dense and thin, with a thickness of about 43 nm. The results of XPS and EDX mapping conrmed that the PDA functional layer was successfully prepared on the Al 2 O 3 ceramic membrane, because not only Al and O elements but also N element were observed in the composite membrane surface. The rejection to Congo red of the PDA/ceramic composite membrane was signicantly improved from 47.7% to 80.4% while the ux was decreased from 524 to 429 L m −2 h −1 . When 2 g L −1 PEI were added in the coating solution, the rejection was increased to 99.7% while the ux decreased to 165 L m −2 h −1 . Aer continuous operation of 180 min, the Al 2 O 3 ceramic membrane without modication showed only 24% of initial ux. In contrast, the ux of PDA/ceramic composite membrane with 2 g L −1 PEI decreased to 55% of initial ux. The PDA/ceramic composite membrane without PEI also showed 49% of initial ux. The results conrmed the formation of a hydrophilic coating on the surface of the ceramic membrane.

Data availability statement
The data presented in this study are available on request from the corresponding author.

Conflicts of interest
The authors declare no conict of interest.