Removal of antibiotics with different charges in water by graphene oxide membranes

As a kind of emerging pollutant, antibiotics in the aquatic environment and their potentially harmful effects on aquatic organisms have aroused widespread concern around the world (Liu et al. 2020). Antibiotics in water may seriously interfere with various physiological functions of the human body, damage the human immune system, bring adverse effects to the aquatic ecosystem, and promote the production of antibiotic resistance of microorganisms in the environment (Wallmann et al. 2021). Therefore, the removal of such pollutants is particularly important. In recent years, nanofiltration technology has been widely welcomed because it can remove trace organics in drinking water or reclaimed water (Nghiem et al. 2004; Kong et al. 2016). However, the traditional nanofiltration membrane has the disadvantages of poor anti-pollution and oxidation resistance. So, it still cannot operate well in treating wastewater with high concentrations of pollutants/salts under adverse conditions (Wang et al. 2020; Yuan et al. 2020).

Graphene oxide (GO) membrane has attracted extensive attention due to its inherent physical and chemical properties, such as adjustable surface function, excellent oxidation resistance, and high thermal stability (Hu & Mi 2013; Dong et al. 2016; Ying et al. 2016; Zhao et al. 2016).

GO is generated from graphene through a series of oxidation reactions. Since there are rich oxygen-containing functional groups (such as –COOH, –OH, etc.) on the edge and base surface, GO nanosheets can be easily modified and adjusted to adapt to different applications (Yu et al. 2015). In addition, these oxygen-containing functional groups enable GO nanosheets to have excellent dispersibility in water, which can further improve the processability of GO nanosheets as membrane materials through different methods (such as vacuum suction filtration, spin coating, and layer-by-layer assembly) (Liang et al. 2016; Nam et al. 2016; Karunakaran et al. 2017). Due to its compact layered structure and controllable layer spacing (the distance between adjacent GO nanosheets), nanofiltration membranes based on GO materials have proved to have excellent performance in the removal of organics in water (Hung et al. 2014; Lecaros et al. 2017).

Similar to the interception mechanism of the polymer nanofiltration membrane, GO-based membranes can become a selective separation membrane through various effects (steric effect, solute GO interaction, electrostatic effect, and adsorption). Previous studies have shown that the removal efficiencies of different antibiotics by the GO membrane vary greatly, and the retention mechanism mainly depends on membrane material characteristics and the physical and chemical properties of target substances (Bellona et al. 2004; Khanzada et al. 2020; Kong et al. 2014, 2015). Many research works showed that the GO membrane had high retention performance for small molecule dyes (such as new carmine, methyl blue, and Congo red) (Zhang et al. 2017a, 2017b; Song et al. 2018; Qi et al. 2020). The good adsorption mechanism of GO enables GO membranes to have high adsorption performance for antibiotic molecules. Since the surface of the GO membrane has negative charges, the removal mechanisms of the membrane for antibiotics with different charges in water may be different. If the performance and mechanism of the GO membrane in antibiotics interception can be fully understood, it will be very helpful for nanofiltration technology to remove antibiotics in the future. Therefore, it is necessary to verify the applicability of the GO membrane in the interception of antibiotics.

In this study, typical antibiotics with different charges in wastewater were selected as the research object. The GO membrane was prepared by the vacuum filtration method, and its removal effect on antibiotics was investigated. Furthermore, the influences of membrane material, coexisting cation, pH, and antibiotic concentration on antibiotics removal performance were analyzed.

Commercial GO powder was purchased from XFNANO (XF002-2, Nanjing XFNANO Materials Tech Co., Ltd, China). Characterized by the manufacturer, GO powder had a film diameter of 0.5–5 μm and a thickness of 0.8–1.2 nm. Polyvinylidene fluoride (PVDF) membranes with a diameter of 63.5 mm and a nominal pore size of 0.22 μm (GVWP06225, Millipore, Billerica, MA) were used as the support layer for the GO membrane fabrication. Three substances represented by antibiotics were purchased from TCI (Chemical Industry Development Co., Ltd, Shanghai). In solution, they are uncharged chloramphenicol (CAP), penicillin (PNC) with a negative charge, and erythromycin (ERY) with a positive charge, and their chemical structures are shown in Supplementary material, Figure S1. Inorganic salts including NaCl, KCl, MgCl₂, CaCl₂, and KOH were of analytical grade. All solutions were prepared using ultrapure (UP) water (Q-Gard^® 1, Millipore, France).

All GO membranes in this study were fabricated by vacuum filtration. For the non-crosslinked membranes, GO suspensions (0.005 mg⁻¹ mL⁻¹) were prepared by sonication for 2 h. Then, 52 mL of the GO suspensions were filtered through the PVDF membrane by vacuum filtration under 0.8 bar at room temperature. The obtained membrane was dried at 40 °C for 40 h to remove water molecules trapped within the GO layers.

For the crosslinked membranes, 5 mL of ethylenediamine (EDA, 99.5 wt%) was added dropwise to 500 mL of GO suspension (0.005 mg mL⁻¹). The mixed solution was stirred for 3 h (temperature 30 °C, rotation speed 200 rpm) and then cooled to room temperature. After that, the filtration process was the same as that of the non-crosslinked membrane described previously, and finally, an EDA covalently crosslinked GO membrane (EDA-GO membrane) is obtained (Supplementary material, Figure S2).

The membrane morphology was characterized by a scanning electron microscope (SEM, HitachiS4800, Japan), X-ray photoelectron spectroscopy (XPS, PHI5000VersaProbe, ULVCA-PHI, USA), a Fourier transform infrared (FTIR) spectrometer (Vertex 70, Bruker, Karlsruhe, Germany), and X-ray diffraction (XRD, Bruker D8 Advance, Karlsruhe, Germany). The d-spacing of the non-crosslinked GO membrane and the EDA-GO membrane under different conditions was calculated by XRD. Scanning was conducted at a speed of 3°/min, for 2θ from 5° to 25°.

The stability of the GO membrane was investigated by placing different GO membranes in UP water, recording the appearance changes of the inner membrane from 0 to 40 days, and observing the shedding of the GO functional layer from the PVDF support layer.

The GO membrane fluxes were tested using a dead-end filtration system at room temperature. In this process, the feed solution was loaded onto a stirred cell (Merck Millipore, Germany) with an effective membrane area of 13.4 cm² from a feed tank of 800 mL (Merck Millipore, Germany). Before the filtration of the solutions, UP water was fed into the cell under 4 bar to compact the GO membrane for about 2 h until the water flux was stable. The feed was then replaced by an antibiotic solution. The filtration was conducted under the same pressure (4 bar) for 5 h. Changes in effluent flux were recorded every 5 min by a computer connected to the electronic balance. Permeate samples were collected every 1 h to evaluate the concentration of antibiotics in the effluent. The antibiotic concentrations of the feed used in this study were 500 μg L⁻¹ and 10 mg L⁻¹.

Antibiotic samples with lower concentrations (<1 mg · L⁻¹) were detected by ultrahigh performance liquid chromatography-mass spectrometry (UPLC-MS); 0.1% formic acid aqueous solution and acetonitrile were used as the mobile phase, the flow rate of the mobile phase was 0.4 mL·min⁻¹, the chromatographic column temperature was 40 °C, and the injection volume was 10 μL (Supplementary material, Table S1).

Antibiotic samples with higher concentrations (1–10 mg·L⁻¹) were determined by the high-performance liquid chromatography (HPLC) method. The CAP solution was detected by an ultrahigh-performance liquid phase ultraviolet detector. Acetonitrile and 0.3% formic acid aqueous solution were used as the mobile phase, the constant ratio was 2:1, the flow rate was 0.1 mL·min⁻¹, the chromatographic column temperature was 35 °C, and the injection volume was 10 μL. The wavelength of the UV detector was 280 nm, and the running time of each sample was 4 min. PNC solution was also detected by an ultrahigh-performance liquid phase ultraviolet detector. Methanol and 0.001 mol·L⁻¹ of potassium dihydrogen phosphate solution was used as the mobile phase, the constant ratio was 60:40, the flow rate was 0.15 mL·min⁻¹, the temperature of the chromatographic column was 35 °C, and the injection volume was 10 μL. The wavelength of the UV detector was 225 nm, and the running time of each sample was 4.5 min. The ERY solution after 20 times dilution was detected by the UPLC-MS method which was used for low-concentration solutions.

In this study, we collected the effluents of secondary biological treatment and ultrafiltration processes from a wastewater treatment plant to study the removal of antibiotics under the condition of actual wastewater quality by the EDA-GO membrane. Antibiotics were added into actual wastewater samples, in which the concentration of these three antibiotics was lower than the detection limit (1–10 μg L⁻¹), and the samples containing 10 mg·L⁻¹ of CAP, PNC, and ERY were prepared, respectively. The water quality of the two samples collected is shown in Table 1.

All data were analyzed using SPSS 27.0 software. The significant differences were evaluated by ANOVA. The P values of <0.05 were considered significant and expressed as ‘*’.

SEM images showed that the surface morphology of the non-crosslinked GO membrane and the EDA-GO membrane show a certain degree of wrinkles, the wrinkles of the EDA-GO membrane are more uniform, and there are no obvious defects in both membranes (Supplementary material, Figure S3). XPS results showed that the GO membrane contains C–O, C–C and C = C bonds. The strength of C–O and C–C bonds in the EDA-GO membrane decreased and C–N bonds appeared, which further indicated the crosslinking reaction between EDA and GO during the fabrication of the EDA-GO membrane (Supplementary material, Figure S4). In the FTIR spectra, the C = O absorption peak at 1,740 cm⁻¹, the C–O tensile vibration peak in the epoxy group near 1,620 cm⁻¹, and the –OH absorption peak at 1,000 cm⁻¹, indicated that there were many hydrophilic oxygen-containing functional groups on the surfaces of the two GO membranes. Also, the stretching vibration peak of the O = C–NH group of the EDA-GO membrane at 1,574 cm⁻¹ indicated that the COOH of GO interacted with NH₂ of EDA and generated the O = C–NH group (Supplementary material, Figure S6) (Meng et al. 2018; Kong et al. 2020).

The prepared non-crosslinked GO membrane and the EDA-GO membrane were immersed in UP water, and the stability of the two membranes was judged by investigating the peeling of the GO functional layer from the PVDF support layer. The results showed that the non-crosslinked GO membrane had an obvious swelling effect when immersed in water, and that the stable chemical bond formed in the EDA-GO membrane can reduce the swelling effect. As a result, the EDA-GO membrane is much more stable than the non-crosslinked GO membrane (Supplementary material, Figure S7).

Figure 2(b) shows that the average removal efficiencies of CAP, PNC, and ERY solutions by the non-crosslinked GO membrane were 52, 84, and 72%, respectively, while those were 57, 92, and 81% by the EDA-GO membrane. Due to better stability and smaller d-spacing of the EDA-GO membrane, the removal of PNC and ERY was higher than that of the non-crosslinked membrane, while the removal of CAP had no significant difference.

The removal mechanism of the GO membrane for organic substances with different charges is different. For electrically neutral CAP, the removal mechanism was mainly the steric effect. The removal mechanism of negatively charged PNC was mainly electrostatic repulsion and the steric effect. The positively charged ERY had a large molecule, and the steric hindrance effect was obvious. In addition, ERY molecules entering the membrane nano channel have an electrostatic attraction with the negatively charged GO nanosheets and stay inside the membrane.

The performance of the nanofiltration membrane was related not only to the material of the membrane itself but also to the load of the membrane material. In this study, the EDA-GO membranes with a load of 50, 100, and 150 mg·m⁻² were used to filter CAP, PNC, and ERY solutions with a concentration of 10 mg·L⁻¹.

The removal efficiencies of the three antibiotics by the EDA-GO membrane with a load of 50 mg·m⁻² were the lowest (Figure 3(b)). Meanwhile, the removal increased with the increase in load, but it did not change significantly when the load was greater than 100 mg·m⁻².

When the load of the membrane was too low, there had been uneven coverage or damage on the membrane surface during the membrane fabrication process, and the GO layer was too thin to intercept antibiotic molecules. As the membrane thickness increased, the ability of the EDA-GO membrane to intercept antibiotic molecules was enhanced. When the load was greater than 100 mg·m⁻², membrane thickness was no longer the influencing factor of antibiotic removal.

The adsorption performance of the EDA-GO membrane on antibiotics was investigated, so as to further explore the removal mechanism of antibiotics. The EDA-GO membranes were put into solutions of 10 mg·L⁻¹ CAP, PNC, and ERY for 24 h, respectively. So, the changes in antibiotic concentrations were measured to explore the adsorption performance of the EDA-GO membrane for the three antibiotics.

The removal of antibiotics mainly depended on the characteristics of the membrane and its physical and chemical properties (Kong et al. 2016; Mahlangu et al. 2016; Khanzada et al. 2020). CAP was electrically neutral in water, thus the mechanism of removing CAP by the EDA-GO membrane was mainly the steric effect (Han et al. 2016; Hao et al. 2017). PNC was negatively charged in water, while the EDA-GO membrane contained rich oxygen-containing functional groups (such as –COOH, –OH, epoxy group, etc.) with a negative charge (Chu et al. 2017; Zhao et al. 2019). Therefore, the mechanism of the EDA-GO membrane to remove PNC mainly included the steric effect and electrostatic repulsion. ERY was positively charged in water. When ERY molecules entered the inner nanochannel of the EDA-GO membrane, they had an electrostatic attraction with GO nanosheets with a negative charge, resulting in a part of ERY molecules staying inside the EDA-GO membrane (Xu et al. 2018). Therefore, the mechanism of the EDA-GO membrane to remove ERY mainly included the steric effect and the electrostatic effect. In addition, the size of the ERY molecule was larger than the other two antibiotics, and its steric effect was more significant.

Inorganic ions in actual wastewater can adjust the d-spacing between independent GO layers during the filtration process, thus changing the structure of the EDA-GO membrane and its separation performance (Kang et al. 2018). In this study, the changes of flux and removal of antibiotics during filtration were investigated in the presence of coexisting cations (10, 100, and 200 mmol·L⁻¹ Na⁺, 10 mmol·L⁻¹ Mg²⁺, and 10 mmol·L⁻¹ Ca²⁺, respectively).

The previous study showed that Na⁺ could cause the increase of d-spacing of the GO membrane, which reduced the van der Waals force between GO nanosheets, and the electrostatic attraction between Na⁺ and GO nanosheets was too weak to form a bridge between them to compensate for the decrease of van der Waals force, so the flux increased when the solution contained Na⁺ (Xu et al. 2018). When Ca²⁺ was present, a strong complexation reaction occurred between Ca²⁺ and GO. Ca²⁺ was closely combined with –COOH at the edge of the GO nanosheet and acted as a crosslinking agent, which crosslinked the single GO nanosheets together, increased the stability of the membrane, and led to the reduction of d-spacing of the GO layer. As a result, J/J_UP was reduced.

For the removal of antibiotics, when CAP and ERY solutions were filtered, the highest removal efficiencies were obtained in the presence of Ca²⁺ (Figure 5(b)). Meanwhile, the removal of PNC did not change significantly with the coexisting cations and was higher than those of CAP and ERY.

When filtering CAP and ERY solutions, Ca²⁺ had a strong crosslinking effect with –COOH in GO, resulting in a decrease in d-spacing, thus increasing the removal of antibiotics due to the steric effect. For PNC, the removal was generally higher than 90% by the EDA-GO membrane, so the influence of Ca²⁺ was not as obvious as that for CAP and ERY.

Previous studies have shown that inorganic ions can significantly change the d-spacing of the non-crosslinked GO membrane, increase the effluent flux, and reduce the removal of target pollutants (Mi 2014; Mo et al. 2016). This study showed that coexisting cations had a small impact on membrane flux and removal performance. It demonstrated that the stability of the EDA-GO membrane was higher.

pH can affect the structure of the EDA-GO membrane, thus affecting its pollutant removal performance (Huang et al. 2013; Yeh et al. 2015). This study investigated the effect of pH (6–9) on the flux and antibiotic removal performance of the EDA-GO membrane.

The removal of the three antibiotics by the EDA-GO membrane under different pH conditions was compared. The results showed that with the increase of pH, the removal of CAP gradually decreased, the removal of PNC gradually increased, while the removal of ERY did not change significantly (Figure 6(b)). When filtering the CAP solution, the increase of pH led to the increase of d-spacing of the EDA-GO membrane. As CAP was electrically neutral in water, the main interception mechanism for removing CAP was the steric effect. The larger the d-spacing was, the weaker the steric effect, resulting in the lower removal efficiency of CAP. PNC molecules were negatively charged in water, and the electrostatic repulsion between GO and PNC was the main removal mechanism. The increase in pH led to more negative charges on the GO membrane, and a stronger electrostatic repulsion between GO and PNC increased the removal of PNC. ERY molecules were positively charged. On the one hand, an increase in pH led to an increase in the d-spacing of the EDA-GO membrane. On the other hand, the electrostatic attraction between GO and ERY molecules increased, resulting in a decrease in the d-spacing of the membrane. The effects of the two actions on the removal performance were opposite, so the removal of ERY had no obvious change.

The interaction between different organic substances may have an impact on their removal during the membrane filtration process. A mixed solution of CAP, PNC, and ERY was prepared and was filtered by the EDA-GO membrane. The results were compared with the single antibiotic solution. The results showed that the average J/J_UP when filtering the mixed solution was 104%, while the average J/J_UP of the single antibiotic solution was 113, 118, and 99% for CAP, PNC, and ERY, respectively (Supplementary material, Figure S9(a)). Also, the removal efficiencies of the three antibiotics in the mixed solution were similar to those in the single antibiotic solution (Supplementary material, Figure S9(b)). It indicated that the interaction between CAP, PNC, and ERY did not affect their rejection performances by the EDA-GO membrane.

For the removal of antibiotics by the EDA-GO membrane, the results showed that the removal efficiencies of antibiotics in low-concentration solutions were 76, 97, and 88% for CAP, PNC, and ERY, respectively (Figure 7(b)). When the antibiotic concentration was 10 mg·L⁻¹, the removal efficiencies of CAP, PNC, and ERY by the EDA-GO membrane were 57, 92, and 81%, respectively. In general, the removal efficiencies of the three antibiotics with lower concentrations were higher than those with higher concentrations.

The composition of organic and inorganic substances in actual wastewater is an extremely complex system, where some pollutants affect the separation performance of the EDA-GO membrane. This study investigated the removal of antibiotics under the condition of wastewater quality by adding antibiotics to actual wastewater samples.

The removal of antibiotics by the EDA-GO membrane was compared under different water quality conditions (Figure 8(b)). It showed that the removal efficiencies of antibiotics by the EDA-GO membrane in the solutions prepared by UP water were 57–90%, while the removal efficiencies of antibiotics in secondary and ultrafiltration effluents were 85–98%. The removal efficiencies of antibiotics in secondary and ultrafiltration effluents were higher than those in the solutions prepared by UP water. This was because the blocking effect of organic matters in wastewater samples increased the steric resistance effect of the EDA-GO membrane to intercept antibiotics (Xu et al. 2018). The interception mechanism of the EDA-GO membrane to intercept CAP was mainly the steric effect. The removal of CAP in secondary and ultrafiltration effluents was much higher than that in the solution prepared by UP water. Meanwhile, the interception mechanism of the EDA-GO membrane to PNC and ERY included the electrostatic effect and the steric effect, so the removal performances of the EDA-GO membrane to PNC and ERY were not quite different under three different water quality conditions.

This study investigated the removal performance of antibiotics with different charges in water by the GO membrane, explored the influence of water quality conditions (coexisting inorganic ions, pH, organic matter concentration, etc.) on the removal performance, and explained the removal mechanism of the GO membrane on different antibiotics. The crosslinked EDA-GO membrane exhibited better stability and better removal performances on antibiotics than the non-crosslinked GO membrane. Na⁺ in water could increase the flux of the EDA-GO membrane slightly but had no significant effect on the removal of antibiotics. Ca²⁺ could reduce the membrane flux and improve the removal of antibiotics. Mg²⁺ had little effect on the membrane flux and antibiotic removal. In the range of pH 6–9, with the increase of pH, the membrane flux of CAP and PNC solution increased, and the removal of CAP decreased, while the removal of PNC increased. The membrane flux and removal of ERY were not significantly affected by pH change. The removal efficiencies of low-concentration (500 μg·L⁻¹) antibiotic solutions were higher than those of high-concentration (10 mg·L⁻¹) antibiotic solutions. Furthermore, compared with the antibiotic solutions prepared with UP water, when filtering actual wastewater samples, J/J_UP of the membrane was lower and the removal of antibiotics was higher. In short, the EDA-GO membrane has a good application prospect in the removal of trace organic matters in water.

This work was supported by the National Natural Science Foundation of China (No. 21707100).

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

The authors declare there is no conflict.