Graphene oxide/Al₂O₃ membrane with efficient salt rejection for water purification

The increasing scarcity of fresh water sources across the globe has urged the need to develop alternative water supplies, including seawater desalination, reuse and recycling of wastewater and storm water (Mahmoud et al. 2015). Desalination is one of the most important and promising methods for fresh water augmentation (Elimelech & Phillip 2011). In recent times, numerous considerable efforts have been put towards obtaining fresh water from abundantly available saline seawater. Many water treatment technologies, such as ion exchange, electrodialysis, reverse osmosis and distillation, have been investigated over the past few decades (Kheriji et al. 2015; Bergquist et al. 2016; Gabarrón et al. 2016; Wu et al. 2017). Among these technologies, the pressure-driven membrane-based desalination process has attracted lots of attention due to its advantages of excellent separation rate against salts, energy savings, and being environmentally friendly (Hegab & Zou 2015; Werber et al. 2016; Qin et al. 2017).

Recently, graphene oxide (GO), which is prepared by the chemical exfoliation of graphite using strong oxidants, has been demonstrated to be a potential membrane material in the field of water purification applications due to its distinctive structural characteristics, excellent antifouling properties, high mechanical strength and negligible thickness (You et al. 2016; Fathizadeh et al. 2017; Ma et al. 2017). In particular, GO contains hydroxyl and epoxy functional groups on the basal planes, in addition to carbonyl and carboxyl groups located at the sheet edges (Dreyer et al. 2010). These oxygen-containing functional groups maintain a relatively large intersheet distance and empty spaces between nonoxidized regions, resulting in the formation of nanochannels between GO sheets that can be used for highly selective molecular and ionic separation of the GO membrane (Han et al. 2015; Xu et al. 2017). In addition, the low production cost of GO from exfoliating graphite provides a greater competitive edge in terms of cost for scaled-up fabrication of such membranes when compared with other more expensive nanomaterials (Goh et al. 2015).

In recent years, numerous studies have adopted GO as the separation layer on the polymer or ceramic matrix in the design of filtration and desalination materials (Aba et al. 2015; Liu et al. 2015). However, studies with reference to the desalting properties of the GO membrane on the asymmetric ceramic membrane support are rarely reported.

Herein, we report the preparation of the GO top membrane on the asymmetric double-layer Al₂O₃ ceramic membrane support, and use the GO/Al₂O₃ membrane to remove salt from the different aqueous salt solutions. The results demonstrate the membrane has an efficient salt rejection toward NaCl, Cu(NO₃)₂ and MgSO₄. Our work presents a simple and low-cost method for preparing a GO/Al₂O₃ membrane. The membrane exhibits many fascinating advantages such as a high rejection of salt, simple and scalable synthesis procedure and technical readiness for scaling up membrane production. All the superior performances of the membrane could open the door to opportunities to overcome the challenges in making clean water easily available.

Flake graphite powders (99.9% purity) were purchased from Qingdao Sanyuan Graphite Co., Ltd. Polyacrylic acid (Dolapix CE 64) was purchased from Zschimmer & Schwarz Co., Germany. Commercial α-Al₂O₃ powders (99.9%, average particle size of 5 μm) were purchased from Zhengzhou Baige Abrasives Co., Ltd. Nano α-Al₂O₃ powders (99.99%, average particle size of 200 nm) were purchased from Shanghai Aladdin Biochemical Technology Co., Ltd.

The synthesis of GO has been reported in our previous works (Hu et al. 2013). In this work, the prepared GO was diluted with deionized water to make a homogeneous GO suspension (0.5 mg·L⁻¹) for storing.

The preparation process of the Al₂O₃ membrane support can be described as follows: firstly, the Al₂O₃ powders were mixed with 1 wt% PVA solution. After sieving, the obtained powders were aged for 24 h. Secondly, 1 g aged Al₂O₃ powders were uniaxially pressed in a disk (Φ24 mm) at 10 MPa. After drying at 120 °C for 3 h, these disks were sintered at 1,500 °C for 2 h. Finally, after cleaning and drying, the support disks (porosity = 40% and thickness = 1.5 mm) with an average pore size of about 400 nm were ready for use.

The Al₂O₃ membrane interlayer was prepared by a dip-coating method. Briefly, 3 wt% PVA as a binder, 0.15 wt% Dolapix CE 64 as a dispersant and 10 wt% nano Al₂O₃ powders were mixed and stirred until the slurry dispersed homogeneously. The above-mentioned Al₂O₃ support disk was placed on the surface of the slurry for 5 s and dried at 50 °C for 3 h. Thereafter, the support with the nano Al₂O₃ layer was sintered at 1,500 °C for 2 h. Finally, the nano Al₂O₃ membrane interlayer with an average pore size of about 80 nm and a thickness of about 10 μm was obtained.

The GO membrane top layer was deposited on the Al₂O₃ membrane interlayer by a spin-coating process. Briefly, the GO suspension was dropped on the above-mentioned membrane interlayer and spun at 200 rpm for 5 s and 3,600 rpm for 5 s. To enhance the mechanical property of the GO membrane top layer, it was dried at 60 °C for 3 h. In order to obtain the definite thickness of the GO membrane, the process was repeated three times.

The pore size distribution of the membrane support and the membrane interlayer was measured with a mercury porosimeter (AutoPore IV 9500, Micromeritics, USA). X-ray diffraction (XRD, D8 ADVANCE, Bruker Corporation, Germany) was recorded on a D/max 2550 V diffractometer with Cu Kα radiation (λ = 0.1542 nm). The Fourier transform infrared spectroscopy (FTIR, Nicolet 5700, Thermo Electron Corporation, USA) spectra were recorded in the range 3,000–500 cm⁻¹ using the KBr pressed pellet technique. Raman spectra were obtained using a LABRAM Raman microscope spectrometer (HR800, Jobin Yvon Horiba, France). The excitation wavelength was 514 nm, supplied by a He–Ne laser. The microstructure of the membrane sample was investigated by field emission scanning electron microscopy (FE-SEM, SU8220, Hitachi Group, Japan). The zeta potential of GO was determined using a zeta potential analyser (ZetaPALS, Brookhaven Instruments Corporation, USA).

The XRD pattern of GO is shown in Figure 1(a). From Figure 1(a), GO has a single intensity peak at 11.2°, which is attributed to the (001) reflection of GO (Hu et al. 2018). This result indicates the flake graphite was fully oxidized after the modified Hummers treatment. According to the Bragg equation, the interlayer distance of GO is about 0.7 nm, which provides possible channels for water molecules to pass through it as a membrane filtration material.

The FTIR spectrum of GO is presented in Figure 1(b). From Figure 1(b), the intense peaks at 1,051, 1,168, 1,726 and 3,382 cm⁻¹ are assigned to the epoxy (C-O-C), the phenol/hydroxyl (C-OH), the carbonyl (C = O) in carboxylic acid, and the hydroxyl (-OH), respectively (Konios et al. 2014). It reveals there are many oxygen functionalities in GO. These oxygen-containing polar groups cause GO to have an excellent hydrophilic property, which will be beneficial for water molecules to pass easily through the GO membrane channels (Han et al. 2015).

Figure 1(c) shows the Raman spectrum of GO. Two main peaks can be observed, and are assigned to the D and G peaks. The D peak is the fingerprint peak in graphite material with a concentration of defects. The G peak associates with the E_2g phonon at the Brillouin zone centre, which indicates the graphitic nature of the material (Eda & Chhowalla 2010). Figure 1(c) shows GO has a high intensity of the D peak indicating the oxidation increases the structural defects of GO. As a result, nanopores are formed in the GO.

The microstructure of the GO/Al₂O₃ membrane is displayed in Figure 2. As displayed in Figure 2(a) and 2(b), it is observed that the surface of the GO membrane top layer has no obvious defects and only presents some corrugations that possibly arise from partial roll up of the edges and the random overlap of individual GO sheets (Hu et al. 2016).

Figure 2(c) demonstrates the finally obtained membrane can be categorized as a typical asymmetric multilayer membrane. Figure 2(c) and 2(d) display the cross-section of the membrane consisting of an Al₂O₃ membrane support, an Al₂O₃ membrane interlayer with a thickness of about 10 μm, and a GO membrane top layer with a thickness of about 800 nm. Figure 2(c) and 2(d) also show the well-layered lamellar GO membrane top layer is obtained successfully. SEM images of the membrane also verify that the GO membrane top layer has been tightly coated on the surface of the Al₂O₃ membrane interlayer, which results from the chemical interaction between hydroxyl and carboxyl groups in the interface of the GO and the Al₂O₃ membrane interlayer (Hu et al. 2015).

To evaluate the separation performance of the GO/Al₂O₃ membrane for the treatment of different salt concentration aqueous solutions, measurements of the permeate flux and salt rejection of the membrane were carried out. As shown in Figure 3(a), in the process of permeation, the permeate flux of the membrane almost keeps a relatively steady value (about 1.254 L·m⁻²·h⁻¹·bar⁻¹) for the different concentrations of aqueous NaCl, Cu(NO₃)₂ and MgSO₄ solutions, meaning that the membrane has excellent and stable filtration performance. This phenomenon should be due to the low feed concentration resulting in less membrane fouling, so that the membrane flux almost keeps constant.

The salt rejection of the membrane with varying feed is presented in Figure 3(b). From Figure 3(b), it can be found that the salt rejection increases as the salt concentration decreases for the different salt solutions, indicating the salt concentration has a great effect on the salt rejection of the membrane and the lower salt concentration is beneficial for obtaining superior salt rejection for the membrane. It is worth pointing out that at 0.01 mol·L⁻¹ feed salt concentration, the salt rejection of the membrane reaches 28.66%, 39.24% and 43.52% for NaCl, Cu(NO₃)₂ and MgSO₄, respectively. Such an efficient salt rejection of the GO/Al₂O₃ membrane can be mainly explained as follows.

Initially, the salt rejection mechanism is perhaps mainly considered to be the physical sieving and Donnan effect (Wang et al. 2016). In general, there are many nanoscale structural defects (e.g. nanopores) that can be introduced in the basal plane of the GO sheets after the modified Hummers treatment (Dave et al. 2016), and these nanopores can allow for water molecules' permeation while rejecting other ions smaller in size than the nanopore. Considering the GO membrane is an assembly of GO sheets stacked together into a network of nanochannels, there are three conditions constituting the water and ion pathway in the GO membrane: nanopores, adjacent sheet edge areas and inner nanochannels (Xu et al. 2015). These pathways will increase the possibility of the ions passing through the membrane so that the physical sieving may have a limited effect of salt rejection in this study. As is known, the negative hydrophilic functional groups such as hydroxyl and carboxylic in GO can induce a negative charge and cause a high zeta potential on the surface of the GO membrane (Hu et al. 2013). The zeta potential of the GO is shown in the inset of Figure 3(b). During salt solution filtration, according to the Donnan effect, the ionic concentrations at the membrane surface are not equal to those in the bulk solution. The counter-ions concentration is higher at the membrane surface and the co-ion concentration is lower at the membrane surface compared with that of the bulk solution. When an external pressure is applied to the membrane, water can pass through the membrane, whereas co-ions are rejected due to the Donnan potential. At the same time, counter-ions are also rejected because of the requirements of electroneutrality. All these results in the GO membrane exhibit excellent rejection of salt ions.

Equation (4) shows the increasing of the salt concentration will improve the diffusion rate (J) for the nanochannels present inside the GO membrane. Namely, with the increasing of the salt concentration in the feed, the rate of the salt ions' permeation through the membrane nanopores will enlarge due to the high salt concentration resulting in the strengthening salt diffusion effect, thereby leading to low salt rejection during the filtration process. The illustration for the desalination of the GO membrane top layer is shown in Figure 4. All these results demonstrate the GO/Al₂O₃ membrane would represent a next generation of the low-cost and energy-efficient membrane for ionic sieving in aqueous solution, with applications in numerous important fields.

An asymmetric GO/Al₂O₃ membrane was successfully fabricated by a spin-coating method. The microstructure measurement shows the GO membrane top layer with a thickness of about 800 nm has been coated on the surface of the Al₂O₃ membrane interlayer homogeneously. During the treatment of different aqueous salt solutions, the membrane preserves a superior separation performance. Especially, at 0.01 mol·L⁻¹ feed salt concentration, the permeate flux of the membrane keeps a relatively steady value and the salt rejection of the membrane reaches 28.66%, 39.24% and 43.52% for NaCl, Cu(NO₃)₂ and MgSO₄, respectively. The stable permeate flux and high salt rejection could ensure the GO/Al₂O₃ membrane would be a very competitive candidate for water purification.

The authors gratefully acknowledge the support of this research by the National Science Foundation of China (Grant No. 51662019) and the Foundation of Jiangxi Science and Technology Committee (Grant No. 20161ACB21008). The project also was funded by the Foundation of Jiangxi Educational Committee (Grant No. GJJ150929).