A high stability GO nanofiltration membrane preparation by co-deposition and crosslinking polydopamine for rejecting dyes

The nanofiltration (NF) technology has been widely used in industry, including desalination (Bai et al. 2019; Ma et al. 2019), water purification (Kim et al. 2018), textile industry (Xu et al. 2017; Ye et al. 2017), pharmaceutical, and bioengineering (Rathore & Shirke 2011; Zhao et al. 2018). NF membranes should achieve the required retention and permeability for industrial applications while providing excellent mechanical strength and operational stability. To meet this requirement, nanocomposite membranes with selective separation layers deposited on porous supports have been extensively studied in the NF field (Lau et al. 2012). Several coating methods, such as interfacial polymerization, dip coating and layer-by-layer assembly (LbL) were used to build this unique structure (Wei et al. 2013; Han et al. 2014; Joseph et al. 2015). Graphene oxide (GO), a two-dimensional (2D) material, can be easily assembled into layered stacks with a well-defined interlayer distance offering new possibilities for membrane separation (Zheng et al. 2016).

Graphene-polymer nanocomposites display remarkable mechanical properties and thermal stability compared to pure polymer (Awad & Khalaf 2018; Ibrahim et al. 2021). GO is the oxidized form of graphene, making nanocomposites have better thermal and mechanical properties than pristine graphene (Shiu & Tsai 2014). Therefore, GO as a novel two-dimensional nanomaterial has gained much attention to preparing and modifying nanofiltration membrane (Perreault et al. 2015). Methods for introducing GO materials into membranes include merging GO to membrane matrix (Li et al. 2016; Liu et al. 2020a) and modifying membrane surface (Liu et al. 2020b). GO contained abundant oxygen-containing functional groups in epoxy, hydroxyl, and carboxyl. These groups render GO high hydrophilicity and enhance the compatibility with GO and membranes (Dreyer et al. 2010). Indeed, GO membranes showed an enhanced filtration performance and excellent anti-biofouling properties (Liu et al. 2016; Kumar et al. 2019). For example, embedding GO in a polyamide membrane improves the flux, salt removal rate, chlorine resistance and anti-biofouling performance (Bano et al. 2015). The conjugation of polyamide with GO resulted in irreversible binding and inactivated bacteria. This improvement in bactericidal performance did not affect the mass transfer performance of the membrane (Perreault et al. 2014). (Januário et al. 2021) used polyvinyl alcohol (PVA) and GO to prepare the layer-by-layer self-assembly membrane and used it for the removal of sunset yellow (SY) dye. It was found that GO had a great influence on the flux, decolorization ability, surface morphology and hydrophilicity of the modified membrane. Although GO membrane has good separation and purification performance, these membranes have some limitations when used in pressure-driven systems (Becerril et al. 2008; Choi et al. 2017). GO membranes are likely to be damaged after long-term operation in water and high pressure, and the nanosheets are easily separated from each other (Park et al. 2010).

PDA is hydrophilic and can be non-selectively strong deposited on any solid substrate (Lee et al. 2007). PDA coating can improve the stability of the stacked GO layer by enhancing the adhesion between GO and the substrate (Lucas et al. 1995; Xu et al. 2016). According to reports, the PDA coating can cover the surface of GO, hiding some surface characteristics (Yang et al. 2012; Yang et al. 2017). In the oxidation and polymerization of DA, GO may be reduced by dopamine hydrochloride (Xu et al. 2010; Gao et al. 2013). It is worth noting that polydopamine can form a strong adhesive coating on various surfaces including the surface of GO layers. PDA's covalent coupling to the membrane via an aryl-aryl coupling or possibly Michael type addition reaction (Lee et al. 2006; Liu et al. 2014) is much stronger than the electrostatic interaction between GO and membrane surface. In addition, the PDA can be used to crosslink the GO layer chemically, the reaction between the group of PDA and GO can form a strong covalent bond and enhance the structural stability of the membrane. Studies have shown that if a single DA deposition time is too long, it is easy to form self-aggregation on the membrane surface, which will make it difficult for GO nanosheets to deposit on the membrane uniformly (Yang et al. 2014; Lv et al. 2015). It has been shown that incorporating low molecular weight PEI into dopamine solutions can promote the uniform formation of PDA, which can prepare the uniform and charged coating membrane (Chew et al. 2017). In addition, the covalent network between PDA and PEI can improve the stability of the selection layer.

The objective of this research was to construct a high stability GO nanofiltration membrane by depositing alternately between PDA@PEI solution and GO on PES ultrafiltration membrane, followed by chemically crosslinking with PDA. The surface properties and structures of the pristine and modified membranes were characterized by several analytical techniques. As a result, the membrane has a good removal effect on divalent ions, and better rejection of negatively charged dyes as well as higher molecular weight dyes, high stability and good anti-biofouling performance.

The commercial ultrafiltration membrane (PES flat sheet membrane, MWCO = 30 kDa) purchased from Ande Membrane Separation Technology Engineering Co., Ltd (Beijing, China) was used in this study. Dopamine hydrochloride was obtained from Nanjing Ronghua Scientific Equipment Co., Ltd PEI (Mw = 1,800 Da) was acquired from Nanjing Ronghua Scientific Equipment Co., Ltd. Graphene oxide was provided by Shenzhen Turing Evolution Technology Co., Ltd. Dyes DB56, VB5, AR151, DY142, AG27, DB19, and DY86 were procured from Nanjing Ronghua Scientific Equipment Co., Ltd. BG11 medium was procured from Qingdao High-tech Industrial Park Haibo Biotechnology Co., Ltd. Microalgae was procured from Institute of Hydrobiology, Chinese Academy of Sciences, Tris-HCl buffer solution (pH = 8.5, 50 mM). Potassium chloride, sodium chloride, magnesium chloride hexahydrate, potassium sulfate, sodium sulfate, and magnesium sulfate were from Shanghai Lingfeng Chemical Reagent Co., Ltd, Nanjing Chemical Reagent Co., Ltd and Sinopharm Chemical Reagent Co., Ltd. All aqueous solutions were prepared using ultra-pure water with a resistance of 18.2 MΩ/cm.

The fabrication process of the composite membrane was shown in Figure 1. First of all, a 50 mL Tris-HCl buffer solution (pH = 8.5) was prepared followed by dissolving DA (0.125 g,2.5 g/L) and PEI (0.075 g,1.5 g/L) for 1 h to formulate a blend solution of DA and PEI.

Secondly, the PES membrane named M0 with the area of 12.56 cm² was prewetted by deionized water (30 mL) for 1 h, and immediately transferred into the prepared DA and PEI blended solution for 2 h, then the surface of the membrane was washed with DI water, the obtained (PDA@PEI)/PES membrane named as M1. Then the M1 was soaked in GO solution (30 ml,0.3 g/L) for 1 h to obtain a (PDA@PEI/GO)/PES membrane named M2. Thirdly, the dip-coated composite membrane is placed in the prepared DA crosslinking agent (30 mL, 2.0 g/L) for 30 min. The unbound DA was washed away with deionized water and dried at room temperature to obtain (PDA@PEI/GO/PDA)/PES composite nanofiltration membrane named M3. Finally, the obtained composite membrane was stored in deionized water for 24 h.

The chemical structure of the membrane surface was observed by Fourier transform infrared spectrometer (FTIR/ATR Nicolet iS5). The sample film was obtained in a 4,000–500 cm⁻¹ FTIR spectrum spectral range. In this experiment, a scanning electron microscope (JEOL JSM-6380LV) was used to scan the images of the PES membrane and the GO composite nanofiltration membrane to obtain the micromorphology of the GO composite membrane before and after dip coating and cross-linking. Atomic force microscope (AFM, Dimension310) was utilized to observe the surface structure and roughness of the PES membrane and modified membrane. The DSA30 contact angle system was used to discuss the changes in its hydrophilicity. Additionally, the zeta potential on the surface of the membrane was obtained by the electrokinetic analyzer (SurPASS™ 3, Austria) with KCl (1 mmol/L) solution as the electrolyte solution.

Figure 2(a1)–2(d1) showed the top surface morphology of M0, M1, M2, and M3 at 10 k or 100 k magnification. Compared with M0, the membrane pores of M1 became smaller, and the surface remained smooth without apparent aggregation. After the introduction of GO nanosheets, due to the large surface area of GO nanosheets and the folding on the surface of the composite membrane, some deep folds will form on the surface of the M2 membrane. After the composite membrane was crosslinked, a wrinkled structure was also observed on the surface of the M3 membrane, but the formed wrinkles became fewer and shallower because the PDA formed a coating on the surface of the composite membrane.

Figure 2(a2)–2(d2) showed the cross-sectional morphology of M0, M1, M2, and M3 at 5 k magnification. All the membrane sections in the picture have similar asymmetric morphology, including a finger-like porous support layer and dense epidermis layer. With the progress of the dipping process of the PES membrane, the thickness of the separation layer of the prepared composite membrane gradually increased. Compared with M2, the thickness of M3 does not change significantly because the polymerization of DA occurs in the outermost layer, which increased the thickness of the composite membrane; also enhanced the cross-linking of GO, which would reduce the distance between GO nanosheet. Considering these two factors, the difference in composite membrane thickness is unclear. The characteristic AFM 3D images and the averaged root-mean-squares (RMS) roughness of different membrane samples were shown in Figure 2(a3)–2(d3) and Table 2. It can be seen that M0 has a relatively smooth surface with an RMS value of about 2.78 nm. It can be seen from Figure 2(b3) that when the PDA/PEI is co-deposited on the surface of M0, the roughness of the film surface increases, and the RMS value slightly increases to 3.19 nm, which also means that the deposition of the PDA@PEI layer on the film surface is uniform. It can be seen from the Figure 2(c3) that after introducing GO nanosheets on the surface of M1, many rough ridges and valleys appear on the surface of M2, and the RMS value has increased to 5.54 nm, which means that the GO layer is undistributed on the surface of M1. As can be seen from the Figure 2(c3) and 2(d3), after cross-linking of the composite film, the obtained M3 surface becomes tighter and smoother than the M2 surface, and the RMS is reduced to 3.67 nm, which is the same as that of the M2 surface in the SEM image. After the PDA cross-linking, the wrinkles on the M3 surface will be less consistent. All in all, the immersion of the PDA helps build a smoother and denser modified layer composed of PDA@PEI, GO, and PDA.

Through Figure 3(a), when PDA/PEI was co-deposited on the surface of the PES membrane, a new peak appeared at about 3,400 cm⁻¹, which is attributed to the O-H and N-H stretching vibrations in PDA and PEI. After the introduction of GO on the surface of M1, weak new absorption peaks appeared at 1,650 cm⁻¹ and 1,170 cm⁻¹, which is attributed to the stretching vibration of C = C and C-O-C in GO. In addition, the peak strength of M2 at 1,715 cm⁻¹ increases, which is caused by the stretching vibration of the C = O bond in GO. Due to the further increase of the thickness of M2, it significantly masks the PDA and PEI stretching vibration caused by O-H and N-H. In addition, the absorption peaks caused by the stretching vibrations of the C-H on the saturated carbon in -CH₃ and -CH₂- have weakened at about 2,980 cm⁻¹ and about 2,890 cm⁻¹ due to the GO layer containing a large number of unsaturated bonds. These results indicate that GO nanosheets are successfully introduced to the M1 surface. The positions and intensities of the peaks of M3 and M2 are unchanged because PDA cross-linking layer is thin, these PDA characteristic peaks cannot be distinguished well.

The visual diagram of the contact angle of the composite membrane is shown in Figure 3(b). Compared to M0 with a contact angle of 67.8°, the hydrophilicity of all modified membranes is increased. Among them, the contact angle of M1 decreased to 58.4°, which is due to the hydrophilic amine groups in PDA and PEI. After the introduction of GO nanosheets, the contact angle of the composite film further decreased to 47.7°, indicating that the hydrophilicity continued to increase, which was due to the increase of the oxygen-containing functional groups and roughness of M2 due to the introduction of GO nanosheets. The contact angle of M3 increased slightly to 54.2°; due to the chemical reaction between GO and PDA, some hydrophilic groups of GO will be masked, and the surface roughness will also have reduced.

For the zeta potential of the composite membrane, it can be seen from the Figure 3(b) that compared to M0, the zeta potential value of M1 has increased to 0.6 mV, which is due to the amine groups in PDA and PEI, and the amine groups will generate protons in water, so the negative charge on the surface of the composite film will be greatly reduced. After introducing GO nanosheets, the zeta potential of M2 increased to −17.8 mV due to the deprotonation reaction of the carboxyl and hydroxyl groups in GO on the surface of the composite membrane in water, which will significantly increase the negative charge of M2. The zeta potential of M3 increased to −18.4 mV because the phenolic hydroxyl group in the PDA cross-linked layer also undergoes a deprotonation reaction in water. However, the hydrolysis of the phenolic hydroxyl group is much weaker than the carboxyl and hydroxyl groups in GO. Therefore, the negative charge of the surface of the obtained composite film is only slightly enhanced.

The removal of single salts (0.5 g/L) was determined to evaluate the separation performance of the modified membrane under 6 bar. Figure 4 indicate that the rejections of the modified membrane to different single salts follow the order of Na₂SO₄ > K₂SO₄ > MgSO₄ > NaCl > KCl > MgCl₂, and the flux to all inorganic salt solutions is close to 11.5 L·m⁻²·h⁻¹, which conforms to the properties of the negatively charged membrane. The rejection rate of Na₂SO₄ and K₂SO₄ can reach 83.26% and 78.56%, respectively. In comparison, the rejection rate of NaCl and KCl decreases to 61.47% and 58.44%, respectively, and the retention rate of MgCl₂ is further reduced to 54.43%. The phenomenon can be explained by the Donnan repulsion effect and steric hindrance effect: due to the negative charge of the composite membrane, the repulsive force of the negatively charged SO₄²⁻ and Cl⁻ is much larger than that of the positively charged Mg²⁺, and the valence of SO₄²⁻ is higher than that of Cl⁻. Therefore, the rejection rate of the former should be higher than the latter (Wang et al. 2016). Because the hydrated ionic radius of Na⁺ (0.358 nm) is larger than the hydrated ionic radius of K⁺ (0.331 nm) (Tansel 2012), the rejection rate of the former should also be higher than the latter.

Table 3 summarizes the properties of NF membranes reported in the recent literature. Compared with the previously reported membranes, the membranes prepared in the laboratory have higher Na₂SO₄ rejection and pure water flux.

Figure 5(a) reflects the effect of feed solution concentration (Na₂SO₄) on the separation performance of M3. As can be seen from the figure, when the concentration of Na₂SO₄ increased from 0.5 g/L to 2.5 g/L, the rejection and flux of the composite membrane decreased. This can be explained by the Donnan equilibrium theory (Nam et al. 2016); the electrostatic repulsion between ions and membrane is the main reason for the desalination of charged nanofiltration membrane. With the increase of Na₂SO₄ concentration, the shielding effect of Na⁺ on the negatively charged composite membrane enhance. Therefore, the surface charge density of the composite membrane is lower; this results in reduced electrostatic repulsion to SO₄²⁻ and salt repulsion. In addition, an increase in the feed concentration also causes an increase in the osmotic pressure Δπ, which lowers the driving force of the composite membrane (ΔP-Δπ) and leads to a decrease in flux (Spiegler & Kedem 1966).

Figure 5(b) reflects the effect of operating pressure on the separation performance of the M3. As can be seen from the figure, the flux of the composite membrane linearly increases as the operating pressure increases from 0.2 to 1 MPa. The Spiegler-Kedem model can explain this trend. When the feed concentration is low enough (as in the study of 0.5 g/L), the osmotic pressure Δπ is negligible. The amount increases linearly with increasing operating pressure. Therefore, when the operating pressure increases from 0.2 MPa to 0.6 MPa, the salt rejection rate first increases due to increased flux. However, as the operating pressure further increases, a significant increase in water flux will result in a membrane feed side concentration that is much higher than the solution body concentration, resulting in strong concentration polarization. Therefore, the concentration polarization effect balances the increasing water flux and maintains a stable salt removal rate under high pressure. Based on the above phenomenon, in the nanofiltration process, the operating pressure of 0.6 MPa can enhance the separation efficiency of the GO composite membrane.

Figure 6 reflects the separation performance of M3 for different dye solutions. It can be seen from the figure that the retention order of M3 for different dyes is DY86 > DB19 > AG27 > DY142 > DB56 > AR151 > VB5, and the flux to all dye solutions is close to 10.5 L·m⁻²·h⁻¹. Among them, the aqueous solutions of DB56, VB5, and AR151 are sequentially positively charged, uncharged, and negatively charged. The molecular weights of these three dyes sequentially increase. The removal rates of these three dyes by M3 can reach 94.82%, 84.74%, and 93.33%, respectively, which shows that among the removal effects of M3 on the dye, the adsorption effect is dominant, followed by the Donnan repulsion effect and steric hindrance effect. In addition, it can be concluded from the figure that under the same changeability of the dye, the larger the relative molecular weight of the dye molecule, the more significant the steric hindrance effect of the composite film, and the greater the removal rate of the dye.

Figure 7 reflects the effect of different ion species on the separation performance of M3. It can be seen from the figure that Na⁺ and Ca²⁺ with the same cation strength and Cl⁻ and SO₄²⁻ with the same anion strength have a significant difference in the rejection rate of the composite membrane but have a slight difference in the flux of the composite membrane.

In the dye solution with 500 mg/L NaCl, the retention rates of M3 for DB56, VB5, and AR151 reduce to 92.41%, 84.22%, and 90.23%, respectively. In the dye solution with 500 mg/L MgCl₂, the retention rates for DB56, VB5, and AR151 reduce to 91.82%, 83.81%, and 87.64%, respectively. In the presence of 500 mg/L Na₂SO₄ dye solution, the retention rates for DB56, VB5, and AR151 reduce 88.53%, 82.33%, and 77.82%, respectively. It can be concluded that when the cations are present in the dye solution, the rejection rate in the presence of Na⁺ is larger than that of Mg²⁺. The reason may be that Mg²⁺ has more charges than Na⁺. The larger the number of charges masks more of the charge on the film surface; this makes Donnan repulsion and adsorption effects weaken. When anions are present in the dye solution, for the same reason, the rejection rate in the presence of Cl⁻ is larger than that in of SO₄²⁻.

Figure 8 reflects the effect of the operating time and ethanol immersion time on the separation performance of GO composite nanofiltration membranes before and after crosslinking. Comparing Figure 8(a1), 8(b1), 8(a2) and 8(b2), in the figure, it can be seen that during the continuous operation time of 60 h and the soaking time of ethanol, the salt removal rate of M3 still retains more than 90% of the original salt removal rate, but the salt removal rate of M2 only keep 80 ∼ 90%. In addition, the degree of flux change of M3 is lower than that of M2, because the GO layer in M2 relies more on non-covalent interactions between the layers (intermolecular forces, π-π conjugate forces) to maintain the stability of the GO layer, too long running time of the membrane or long-term immersion in organic solvents will cause more GO nanosheets to fall off the GO layer, resulting in a large reduction in salt removal rate. However, in M3, dopamine has completed the cross-linking of the GO layer. There is non-covalent interaction between GO layers and covalent effects to maintain the stability of the GO layer (Li et al. 2018). Therefore, the membrane runs for too long or immerses in organic for a long time in the solvent; only a small amount of GO nanosheets is removed due to insufficient crosslinking, so the effect on the salt removal rate is slight. In addition, because the surface structure of the membrane is destroyed to a certain extent, the flux of the membrane will increase to varying degrees. The above results show that M3 has good stability and swelling resistance.

Figure 9 reflects the effect of NaClO soaking time under different pH conditions on the separation performance of the composite membrane before and after GO dip-coating. Compared with Figure 9(a1) and 9(b1), the salt removal rate of the composite nanofiltration membrane coated with GO can retain 80 ∼ 90% after 10 h of continuous immersion under acidic conditions, but the salt removal rate of the composite nanofiltration membrane without GO only retains 60 ∼ 70%. In Figure 9(a2) and 9(b2), after 10 h of continuous immersion under neutral conditions, the salt removal rate of the composite nanofiltration membrane coated with GO can retain 70 ∼ 80%. However, the salt removal rate of the composite nanofiltration membrane without GO dip-coating retains 60 ∼ 70%. It can be seen in Figure 9(a3) and 9(b3), after 10 h of continuous immersion under alkaline conditions, the salt removal rate of the composite nanofiltration membrane coated with GO can retain 70 ∼ 80%. However, the salt removal rate of the composite nanofiltration membrane without GO dip-coating only retains 40 ∼ 50% of the original salt removal rate. In addition, under different pH conditions, the flux change degree of the composite nanofiltration membrane dipped with PDA was lower than that without PDA.

Under acidic conditions, the main components in the solution are HClO and Cl₂. At the beginning of immersion, the amide group in the active layer is chlorinated; therefore, the composite membrane hydrophilicity and flux increase slightly, but with the immersion time prolonged, the chlorinated amide group is likely to undergo Orton rearrangement with the aromatic ring, causing the amide structure to be partially destroyed, thus increasing the membrane flux and reducing the rejection rate. Under neutral and alkaline conditions, the main component in the solution is ClO⁻, which has strong oxidizing properties and will seriously damage the composite membrane structure. In addition, the oxidizing property of ClO⁻ under alkaline conditions is greater than that under neutral conditions, which will immensely destroy the composite membrane structure, so under alkaline conditions, the damage of composite membrane performance is more pronounced (Powell et al. 2014).

Figure 10 shows the attachment growth of microalgae on M0, M1, M2, and M3 surfaces. It can be seen in Figure 10 that due to the strong hydrophobicity of the M0 surface, microalgae can easily grow on its surface. After co-deposition of PDA and PEI on the M0 surface, although the hydrophilicity of the M1 surface improved slightly, its negative charge significantly reduced, so there a large number of microalgae attached to the surface. It can be seen from Figure 10(c) and 10(d) that the number of microalgae attached to M2 and M3 greatly reduced. The algae-resistant mechanism may be due to the GO nanoparticles producing active oxygen (ROS) after direct contact with the algae cell, and release of hydrogen peroxide, hydroxide, and superoxide anion, which destroyed the algae's cell integrity, thereby inhibiting the growth of microalgae on the surface of GO composite nanofiltration membrane (Lim et al. 2017). In addition, GO nanosheets greatly enhance the hydrophilicity of the composite membrane surface, so the binding force of microalgae to the composite membrane surface will also decrease. It is worth noting that after cross-linking the GO layer with DA, DA will partially reduce the GO, so that partial reduced graphene oxide (rGO) will be obtained on the surface of the composite membrane. It is speculated that the difference in surface charge and aggregation state of these two nanomaterials is the key factor to cause the difference in cytotoxicity (Liao et al. 2011; Du et al. 2016). GO has a higher zeta potential and oxygen content, and is more biotoxic than rGO. Therefore, the number of microalgae attached to the surface of M3 should be less than M2, but the actual situation is exactly the opposite. Because the stability of the M2 is less than that of the M3, the algae resistance of the M2 greatly decreased, in addition, the roughness of the M3 is lower, the microalgae will not easily grow on the membrane. The above results show that GO composite nanofiltration membrane has good anti-biofouling performance.

In order to quantitatively evaluate the pollution degree to the membrane, the FIRR after the microalgae adhered to the surface is measured, and the results are shown in Figure 11. It can be seen from the figure that the FIRR of M0 is 98.84%, Rt is 8.03 × 10¹⁴ m⁻¹; the FIRR of M1 is 91.13%, Rt is 6.84 × 10¹⁴ m⁻¹; the FIRR of M2 is 76.75%, Rt is 3.93 × 10¹⁴ m⁻¹; the FIRR of M3 is 59.58%, and the Rt is 4.44 × 10¹⁴ m⁻¹. These results quantitatively reflect that after GO is introduced into the membrane, the growth rate of microalgae attached to the membrane is reduced, which greatly reduces the microalgae clogging of the membrane pores.

In this work, we successfully prepared a (PDA@PEI/GO/PDA)/PES anti-biofouling membrane by dip coating GO and chemical crosslinking PDA for improving rejection and stability performance. The obtained membrane has a highly hydrophilicity and negatively charged surface due to the deprotonation of the carboxyl and hydroxyl group in GO and the phenolic hydroxyl group in PDA, and therefore exhibited better rejection performance to divalent anion and higher molecular weight dyes as well as negatively charged dyes with the rejection rate for DY86 (94.82%) and AR151 (93.33%), respectively. During ethanol and hypochlorite resistance tests, the salt removal rate of (PDA@PEI/GO/PDA)/PES membrane can retain more than 90% of the original salt removal rate, implying the desirable durability, swelling resistance and favorable chlorine resistance under different pH conditions in regards to salt rejection properties. Moreover, the flux irreversible recovery rate of (PDA@PEI/GO/PDA)/PES membrane decreased from 91.13% to 59.58%, suggesting that the microalgae would be difficult to attach and grow on GO composite membrane due to improved stability and lower roughness after loading GO in the composite membrane.

The National Natural Science Foundation of China (Grant No. 51208259) financially supported this study. The facility was mainly supported by Jiangsu Key Laboratory of New Membrane Materials, Key Laboratory of New Membrane Materials, Ministry of Industry and Information Technology, School of Environmental and Biological Engineering, Nanjing University of Science and Technology (Grant No. 30920140122008).

All relevant data are included in the paper or its Supplementary Information.