Tuning d-spacing of graphene oxide nanofiltration membrane for effective dye/salt separation

PAN with a purity of ≥ 99.5% and molecular mass of 120,000 g·mol⁻¹ was purchased from Shanghai Jinshan Petroleum Chemical Co. Ltd, China. GO powder with a purity of ≥ 99.9% was purchased from Suzhou Hengqiu Graphene Technology Co., Ltd, China. PEI samples with various molecular weights (2000, 10,000, 25,000, and 65,000 g·mol⁻¹) and purities ≥ 99.5% were purchased from Aldrich. N, N-dimethylformamide (DMF) with a purity of ≥ 99.5% was purchased from Xilong Scientific Co., Ltd. Acid Blue 90 (AB90) with a purity of ≥ 99.5%, Direct Red 80 (DR80) with a purity of ≥ 99.0%, Alcian Blue 8GX (AB8GX) with a purity of ≥ 99.5%, and Congo Red (CR) with a purity of ≥ 99.0% were purchased from Sinopharm Chemical Reagent Co., Ltd. Amaranth (AM) with a purity of ≥ 99.5% and Sunset Yellow (SY) with a purity of ≥ 99.0% were purchased from Aldrich. Sodium sulfate (Na₂SO₄) with a purity of ≥ 99.0% and sodium chloride (NaCl) with a purity of ≥ 99.5% were purchased from Sinopharm Chemical Reagent Co., Ltd. All of the experiments in this work used deionized (DI) water.

PAN nanofibrous substrates were prepared via electrospinning, as described in our previous study [15, 16]. Each substrate was cold-pressed for 30 s at 0.6 MPa before use (at 25 °C).

The PEI/GO top layer was prepared as follows. First, a 0.01 mg·ml⁻¹ GO solution was obtained by mixing GO nanosheets and DI water using ultrasound for 60 min. Thereafter, a series of homogeneous PEI/GO composite dispersions were obtained by dissolving PEI into GO solutions at PEI/GO mass ratios of 10:1, 15:1, 20:1, 25:1, and 35:1 and magnetically stirred at room temperature for 3 h. Finally, the resultant PEI/GO dispersions were vacuum filtered through an electrospun PAN substrate (0.1 MPa) to form the PEI/GO top layer on the PAN substrate. It should be noted that during the vacuum deposition process, some PEI molecules in the dispersions passed through the pores of the PAN substrate, and were therefore, lost. Similarly, GO@PAN composite membranes were prepared by depositing GO dispersions onto PAN substrates using vacuum filtration. The resultant PEI/GO@PAN and GO@PAN composite membranes were stored in deionized water. The designations of the membranes are listed in Table 1.

After sputtering a thin layer of gold on the samples, the cross section and surface topography of each membrane were observed using an SU8010 device (Hitachi, Japan) for scanning electron microscopy (SEM). X-ray photoelectron spectroscopy (XPS) was performed using a Thermo Scientific ESCALAB 250Xi platform. Energy-dispersive spectroscopy (EDS) was performed on a JEOL JSM-7500FA platform. The equipment for detecting the Raman spectra was obtained from Tokyo Instrument Co., Osaka, Japan. The surface hydrophilicity of the membranes was estimated using a dynamic contact angle testing instrument (OCA40). The zeta potentials were monitored via an electrokinetic analyzer (SurPASS, Anton Paar, Austria) at different pH values, using a KCl solution as the electrolyte (0.001 mol·L⁻¹). X-ray diffraction (XRD) data were obtained using a Shimadzu 6000 diffractometer with Cu Kα radiation (λ = 0.15406 nm). The membrane samples were precompacted using deionized water at different operating pressures and freeze-dried prior to XRD testing.

The nanofiltration performances of the GO@PAN and PEI/GO@PAN membranes were evaluated using a laboratory-scale cross-flow device. Before the filtration experiment, the membranes were pre-compacted with deionized water at 0.3 MPa. To investigate the separation performance, the feeding solutions were mixed with aqueous dye solutions (AM, CR, DR80, AB90, SY, and AB8GX) and salt solutions (Na₂SO₄ and NaCl) at various concentrations. The feeding solutions were stirred continuously during filtration to minimize concentration polarization. The solute rejection (R) rate of each membrane was calculated via Eq. (1), and the permeate flux (J_p) of the membrane was calculated via Eq. (2):

where C_p represents the permeate concentration; C_f represents the feeding solution concentration; the effective membrane area (A) is 2.9 cm²; Δt represents the test time; and V represents the volume of permeate solution used for the test. The salt concentrations were detected using a Mettler Toledo conductivity meter, and the dye concentrations were detected using an ultraviolet–visible (UV–Vis) spectrometer. A mixture of DR80 dye (100 mg·L⁻¹) and NaCl (6 g·L⁻¹) was used to assess the antifouling properties and reusability of the membranes. A three-cycle filtration process tested the reusability at 0.3 MPa, with a filtration time of 240 min for every cycle. Between successive cycles, the membranes were flushed with distilled water for 30 min. The flux recovery ratio (FRR) and total flux decline ratio (FDR) were used to evaluate the antifouling properties of the membranes, with a higher FRR and lower FDR indicating better antifouling properties. The FRR of each membrane was calculated via Eq. (3), and the FDR of each membrane was calculated via Eq. (4):

where J_pi represents the starting permeate flux value for the i cycle, J_p1 represents the starting permeate flux value for the first (i = 1) cycle, and J_pi,6 represents the final flux value of the i cycle after 60 min of filtration (i ranges from 1 to 3).

A flat-top GO layer is often required in nanofiltration membranes because it offers a better filtration performance than an uneven performance. As the supporting layer for the nanofiltration membrane, the flatness and density of the PAN substrates are crucial for the flat deposition of the GO top layers. Nevertheless, conventional PAN substrates always have large drops between fibers, making the surface rough and uneven, and resulting in a loose substrate formation (Figure S1a). This study used a PAN substrate with a smoother surface, where the drops between the fibers became significantly smaller after cold pressing (Fig. S1b). With a much smoother surface, the PAN substrate was more conducive to forming an integrated and flat GO layer on the membrane, while simultaneously reducing the quantity of GO waste during the deposition process.

The thickness of the top membrane is another crucial factor that influences the permeability of a nanofiltration membrane. In general, a thicker membrane results in a lower permeability. Therefore, to obtain the thinnest and most optimized integrated top layer, we used the smallest possible quantity of GO during the deposition process. The quantity of GO used for deposition was very small (approximately 16 mg·m⁻²), and we noticed the presence of visible pores on the membrane surface (as highlighted in red in Fig. S2a). Such pores were highly undesirable because they allowed dye molecules and other solutes to pass through the membrane, resulting in an unsatisfactory rejection rate for the GO@PAN membrane (i.e., 16.5% ± 0.3%). To improve the condition of the membrane surface, the GO@PAN membrane was further modified with PEI solutions to ensure a high separation performance of the ultrathin top membrane.

To repair the surface pores and improve the rejection performance, various PEI solutions with different molecular weights were used to modify the GO membranes at a PEI/GO mass ratio of 15:1. The rejection performances of the modified membranes for a 100 mg·L⁻¹ DR80 solution were evaluated, and the results are shown in Fig. 1a. The figure shows that except for PEI with a molecular weight of 2000 g·mol⁻¹, which gave a slightly lower rejection rate of approximately 94.9% (still much better than the bare GO@PAN membrane), the PEI/GO(15:1)@PAN membranes comprising PEI with molecular weights larger than 10,000 g·mol⁻¹ exhibited excellent rejection performances that were greater than 99%. We also found that when the PEI molecular weight was increased from 10,000 to 65,000 g·mol⁻¹, the membrane demonstrated nearly linear growth in the permeate flux, from 88.2 to 173.8 L·m⁻²·h⁻¹ at 0.3 MPa, as shown in Fig. 1a. PEI with a molecular weight of 65,000 g·mol⁻¹ offered the best performance in terms of the permeate flux; therefore, we chose PEI with this particular weight to tune the top GO layer in our membrane modification strategy.

We further performed an SEM analysis to determine whether a larger molecular weight PEI had any significant impact on the structural morphology of the PEI/GO(15:1)@PAN membrane. This SEM analysis was performed on the modified membrane, and the results were compared with those of the bare GO@PAN membrane. The results of this analysis are presented in Fig. 1c, d. A cross-sectional SEM image showed that the thickness of the PEI/GO(15:1)@PAN membrane did not change significantly when PEI with a high molecular weight was incorporated into the GO layer; the thickness was still similar to that of the GO@PAN membrane. Both were approximately 60 nm, thus proving that incorporating PEI with a high molecular weight did not affect the thickness of the GO layer, and the ultrathin feature remained.

The concentration of the PEI solution also played a crucial role in modifying the PEI/GO@PAN membranes. Compared to the bare GO@PAN membrane (Fig. S2a), the PEI/GO@PAN membranes (Fig. S2b–f) became more intact after the PEI deposition, proving that the PEI molecules could effectively regulate the GO deposition process to obtain a flat-top membrane. Additionally, we found that changing the concentration of the PEI solution had a negligible effect on the surface morphology of the top membrane. None of the PEI/GO@PAN membranes (Fig. S2b–f) had visible pores, and they were all extremely thin. The outline of the supporting layer fibers could be observed through the top GO layer.

Furthermore, as the concentration of PEI changed, we found that the rejection rates of all the PEI/GO@PAN membranes were still greater than 99% (Fig. 1b). Interestingly, with an increase in the PEI/GO mass ratio from 0:1 to 35:1, the included angle of the PEI/GO@PAN membrane decreased from 72.6° to 32.0° (Fig. S2g–l), indicating an improvement in the surface hydrophilicity of the PEI/GO@PAN membrane. This is an interesting finding, as an improvement in the surface hydrophilicity can contribute to an increase in permeate flux. The lower flux performance of the PEI/GO(10:1)@PAN membrane compared with that of the PEI/GO(15:1)@PAN membrane verified this hypothesis (Fig. 1b).

As for the PEI/GO(35:1)@PAN membrane, we found that the decrease in the permeate flux rate from 173.8 to 59.1 L·m⁻²·h⁻¹ at 0.3 MPa was slightly unusual. This could have been because an excessive number of PEI molecules on the GO surface blocked the water channels. More importantly, excess PEI molecules can provide redundant positive charges and create strong electrostatic attractions with the negatively charged GO molecules, further reducing the d-spacing of the GO layer.

To further explore the changes in the intrinsic properties of the membranes, we performed several tests to determine the characteristics of the modified PEI/GO@PAN. The PEI/GO(15:1)@PAN membrane was used in all the analyses described below. The chemical composition of the PEI/GO(15:1)@PAN membrane was analyzed using EDS microanalysis. The EDS spectra (Fig. S3) of the PEI/GO(15:1)@PAN membrane show the presence of nitrogen, proving the successful introduction of PEI into the GO membrane. The XPS results also exhibit a pronounced nitrogen peak (Fig. 2a and Table S1), which further confirms the presence of PEI in the membrane. The prominent nitrogen peak in the XPS spectrum appears to be further divided into three secondary peaks, which correspond to three types of nitrogen bonds in PEI (i.e., –NH₂ (398.6 eV), –NH– (399.5 eV), and –NH₃⁺ (400.9 eV)) (Fig. 2b) [44, 45]. Furthermore, we observed a peak shift of C=O in C 1s (Fig. S4a) and O 1s (Fig. S4b) in the XPS spectra. This shift indicates that the negatively charged GO laminates are physically cross-linked to the positively charged PEI molecules via electrostatic coupling, thus stabilizing the GO top layer.

Stress–strain tests were also performed on the three membranes (i.e., the PEI/GO(15:1)@PAN, GO@PAN, and PAN membranes). The results in Fig. 2c show that the PEI/GO(15:1)@PAN membrane has the highest yield (i.e., 17.2 MPa) and breaking strengths (i.e., 23.4 MPa), revealing that it is the most robust among the three membranes. The structural stabilities of the various GO-based membranes were verified via ultrasonic treatment in water, and the results are shown in Fig. S5. After 120 s of ultrasonic treatment, the integrated barriers of the PEI/GO@PAN membranes are still visible, regardless of the PEI quantity. Conversely, the macroscopic structure of the GO@PAN membrane collapses completely within 60 s of ultrasonic treatment. Hence, we could reasonably conclude that the structure of the PEI/GO@PAN membranes was more stable than that of the other bare membranes owing to the introduction of PEI.

To further clarify the structural characteristics of the membranes, we performed Raman spectroscopy analyses on both the GO@PAN and PEI/GO(15:1)@PAN membranes. The Raman spectra of both membranes are presented in Fig. 2d. With the addition of PEI, an increase in the I_D/I_G value from 0.98 to 1.01 is observed, which corroborates the formation of a looser top GO layer [46, 47]. This result implies that more water channels were created in the PEI/GO(15:1)@PAN with the introduction of PEI, resulting in an improved performance of the membrane flux.

The dye rejection performance is a crucial indicator for evaluating membrane quality. The PEI/GO(15:1)@PAN membrane exhibits the largest flux (Fig. 1b). Hence, we decided to use the PEI/GO(15:1)@PAN membrane in subsequent systematic studies. Typically, during filtration, one of the critical factors that can significantly affect membrane performance is the operating pressure. To obtain the optimal operating pressure for the PEI/GO(15:1)@PAN membrane, the rejection rates and flux performances of the membrane for the DR80 dye (as a dye representative) were evaluated at different pressures.

As shown in Fig. 3a, the PEI/GO(15:1)@PAN membrane demonstrates excellent rejection rates of greater than 99% for the DR80 dye molecules under all the operating pressures. However, the permeate flux decreases from 173.8 to 109.4 L·m⁻²·h⁻¹ with increase in operating pressure (i.e., from 0.3 to 0.8 MPa). This behavior is significantly different from the characteristics of normal membranes, where the flux rate typically increases with the pressure. This may have been related to the interlayer spacing (i.e., d-spacing) of the PEI/GO(15:1)@PAN membrane, which could have been severely compressed as the pressure increased, leading to the collapse of the membrane nanochannels, indicating that PEI/GO(15:1)@PAN is suitable for low-pressure processes [48, 49].

A long-cycle experiment was also performed to evaluate the stability and durability of the membrane structure at 0.3 MPa. The dye rejection rate remains high (over 97.5%) during the nonstop 750-min test (Fig. 3b). Although the normalized flux declines slowly, the drop ratio does not decrease significantly, but is only 15.0% less than that of the initial permeate flux after the 750 min operation. This result shows that the PEI/GO(15:1)@PAN membrane could maintain high-efficiency filtration for a long time.

In some industrial fields (e.g., textiles), the wastewater typically contains a high concentration of dye; therefore, the rejection rates and water permeabilities of the PEI/GO(15:1)@PAN membrane to dyes with various concentrations were also evaluated. The experiments were performed at 0.3 MPa, and the results are shown in Fig. 3c. All the rejection rates for DR80 at concentrations varying from 10 to 800 mg·L⁻¹ remain high, that is, at greater than 97.5%. Although there is a slight decrease in the permeate flux rate of the membrane as the dye concentration increases, the flux rate is still reasonably high; the high flux rate is probably due to an increase in the concentration polarization of the membrane surface during filtration. When the concentration of the DR80 dye is 800 mg·L⁻¹, the flux rate of PEI/GO(15:1)@PAN is still greater than 150 L·m⁻²·h⁻¹, indicating that the PEI/GO(15:1)@PAN membrane could effectively reject high concentrations of DR80 solutes.

In an actual textile wastewater environment, apart from a wide range of dye concentrations, dye solutes can also be present in different forms with different sizes and molecular weights. To explore the universality of the performance of the PEI/GO(15:1)@PAN membrane, rejection experiments were performed with various types of dyes. As shown in Fig. 3d, the PEI/GO(15:1)@PAN membrane exhibits similar rejection percentages for three anionic dyes: DR80, AB90, and CR. All the rejection rates are greater than 99.0%, which are higher than the rejection rate for AM (85.6%) as a result of the additional sieving effect. It is worth noting that the rejection of AB8GX (94.3%) is slightly lower than that of DR80, although the AB8GX cationic dye has a molecular weight similar to that of DR80. This could be attributed to the negative charge of the PEI/GO(15:1)@PAN membrane in a weak acid or base solution, which facilitates the rejection of anionic dyes through the Donnan effect (Fig. 4a).

A further decrease in the anionic dye size leads to a lower rejection rate. This can be seen clearly in the rejection rate for SY, which is only 54.8%. From digital photographs taken before and after filtration (Fig. 4b), it can be seen that the membrane does not have a good rejection effect on the AM or SY solutions, and the color of the dyes only become lighter; these results are different from the results obtained for the other four dye solutions, which become clear. Based on the above analysis, we deduced that the PEI/GO(15:1)@PAN membrane is more suitable for rejecting anionic dyes, larger than CR.

The antifouling and reusability of membranes are crucial for practical applications, especially in large-scale processes. Therefore, we performed the relevant tests to evaluate the antifouling and reusability of the PEI/GO(15:1) membrane. In the antifouling test, the FRR value is greater than 94%, and the FDR value is less than 9% during the three filtration cycles, indicating the excellent stability and antifouling properties of the membrane (Fig. 4c). In the reusability test, three continuous filtration cycles were performed, where the filtration time of each cycle was extended to 240 min. As shown in Fig. 4d, the normalized flux declines slightly by approximately 10% toward the end of each cycle. However, the permeate flux can recover to approximately its initial value after the membrane is flushed with water, which confirms the reusability of the membrane. Undoubtedly, this could be ascribed to the introduction of PEI, which improved the hydrophilicity and mechanical strength of the membrane.

In the dye/salt separation process, a nanofiltration membrane should exhibit a high rejection rate for dye molecules and a high brine flux [50]. In the subsequent experiments, 100 mg·L⁻¹ DR80 anionic dye and salt mixture solutions (NaCl or Na₂SO₄ at various concentrations) were used to test the dye/salt separation performance of the PEI/GO@PAN membrane. As shown in Fig. 4e, the PEI/GO(15:1)@PAN membrane demonstrates an excellent rejection rate for the DR80 dye (greater than 99%), while exhibiting a low rejection rate (< 5%) for all concentrations of Na₂SO₄. For instance, when the concentration of the Na₂SO₄ solution is 0.5 g·L⁻¹, the rejection rate for Na₂SO₄ by the PEI/GO(15:1)@PAN membrane (4.7%) is much lower than that of a commercial PES 2 K membrane (9.4%) [51]. More importantly, because the PEI/GO(15:1) top layer is negatively charged in a neutral solution (Fig. 4a), increasing the concentration of SO₄²⁻ further increases co-ion sorption [52, 53], which reduces the rejection rate from 4.7% to approximately 1.7%. The permeate flux declines slightly from 174.3 to 117.8 L·m⁻²·h⁻¹ at 0.3 MPa with an increase in the Na₂SO₄ concentration, but the flux is still high enough for the separation of high salinity dye/salt. According to the XRD results (Fig. 4f), the interlayer spacing of PEI/GO(15:1)@PAN is tuned and fixed at approximately 0.79 nm (i.e., the 2-theta value of the PEI/GO(15:1)@PAN membrane is 11.28°, or 1.12° smaller than that of the GO@PAN membrane). This interlayer spacing is larger than the diameters of water molecules and hydrated ions but smaller than those of most hydrated dyes, thereby explaining the excellent separation performance. More information on the sizes of the hydrated ions and hydrated dye molecules can be obtained from Tables S2 and S3. Similar dye/salt separation performances for NaCl solutions of various concentrations are shown in Fig. S6.

Compared with other reported filtration membranes, the as-prepared PEI/GO(15:1)@PAN membrane exhibits the best results for all the properties detailed in Table 2. However, the reported nanofiltration membranes have shortcomings. For instance, the water permeability of pure GO membranes [10, 54, 55] is generally low because the GO layer spacing is too dense and cannot be adjusted to a proper distance. Although loose polymer nanofiltration membranes [56] exhibit high water permeability, their use in a filtration operation requires a large amount of energy and a large operating pressure. In addition, the salt rejection of most of the reported membranes is high, making them unsuitable for the separation of dye/salt mixtures. Overall, we can say that the PEI/GO(15:1)@PAN membrane presents a much better alternative for nanofiltration technology because it exhibits higher water permeability, higher dye rejection, and lower salt rejection, indicating its excellent potential for dye/salt separation.