Norfloxacin removal efficiency by a carbon filtration column under the influence of nanoplastics: mechanistic analysis and prediction model Open Access

Plastic products are widely used in various aspects of human life and social production, due to their low production cost and excellent properties, such as shopping bags, agricultural films, and medical components (Zhang et al. 2022). According to statistics, global plastic production has increased dramatically from 1.7 million tons in the 1950s to 368 million tons in 2019 (Yee et al. 2021). The huge plastic production has led to more than 13 million tons of plastic waste being discharged into rivers and oceans every year, and this quantity will continue to rise (Enfrin et al. 2019). Due to the difficulty in the degradability of materials, these plastic wastes usually accumulate in water bodies for decades. The larger plastic blocks are subjected to various physical, chemical, and biological effects, gradually breaking down into microplastics (MPs) with particle sizes less than 5 mm and then into nanoplastics (NPs) with particle sizes less than 1 μm (Peller et al. 2022). In comparison to MPs, NPs have smaller particle sizes and larger specific surface areas, which gives them higher surface energy and facilitates the adsorption of heavy metal ions and organic pollutants (Feng et al. 2022). Wang et al. (2019) investigated the adsorption capacity of polystyrene (PS) from 170 μm to 50 nm for phenanthrene and nitrobenzene. The results showed that the adsorption coefficients of both phenanthrene and nitrobenzene increased with decreasing PS particle size and reached the maximum at 235 nm. It indicated that NPs have stronger adsorption capacity than MPs and are more likely to be potential transport carriers of organic pollutants in water bodies. Pivokonsky et al. (2018) found considerable levels of particles smaller than 1 μm in size in raw water samples, with abundances ranging from 111 to 2,181 L⁻¹ particles. Although such small particles could not be analyzed for proper material composition, it can be expected that an unknown percentage of them could be of plastic origin. In addition, once entering the organism, NPs may release contaminants that lead to the Trojan horse effect with increased toxicity (Binelli et al. 2017). It indicates that the adsorption capacity of NPs for contaminants and the potential mobility of NPs can influence the transport behavior of contaminants in the aquatic environment, which may pose a threat to drinking water quality safety.

Antibiotics are metabolites produced by bacteria, molds, and other microorganisms or analogs synthesized by artificial means (Inyinbor et al. 2018). They are widely used in human and veterinary medicine because of their excellent antibacterial effects (Kovalakova et al. 2020). While the use of antibiotics has brought many benefits to humans, excessive use has also caused a number of environmental problems. The most important of these problems is the generation of antibiotic-resistant genes (ARGs) and antibiotic-resistant bacteria (ARB), which negatively affect the effectiveness of antibiotics in treating human and animal pathogens. According to the chemical structures, antibiotics are mainly classified into β-lactams, macrolides, quinolones, sulfonamides, and tetracyclines (Kümmerer 2009). Among them, quinolone antibiotics are commonly used to treat infectious diseases such as urinary tract infections, respiratory tract infections, and gastrointestinal and abdominal infections caused by Escherichia coli or other gram-negative pathogens (Naeem et al. 2016). In China, the scale of production and consumption of quinolone antibiotics is huge, taking norfloxacin (NOR) as an example, the use of NOR in the whole country was up to 5,440 tons in 2013 (Zhang et al. 2015). On the other hand, since antibiotics are not completely absorbed by humans and animals, for example, 70% of ofloxacin and 30% of NOR are excreted unchanged in urine (Zhang et al. 2013), the residual antibiotics may be discharged through the municipal sewerage network into municipal wastewater treatment plants and eventually into the natural water bodies. However, the conventional drinking water treatment processes are also not effective in removing antibiotics and have been reported to remove less than 30% of NOR (He et al. 2015). It would lead to the increasing resistance of bacteria to quinolone antibiotics and the increased abundance of ARGs and ARB has been demonstrated in water distribution pipeline systems (Xu et al. 2016), which makes humans and animals more vulnerable to these microorganisms. Therefore, it is particularly important to control the content of quinolone antibiotics in drinking water treatment.

Filtration is one of the core processes in drinking water treatment, and the quality of the filtrated water is affected by various aspects such as filtration methods, pre-filtration water quality, and filtration media. For example, Liang et al. (2021) used ultrafiltration and two-stage reverse osmosis-integrated membrane filtration technology to treat pig wastewater, and the results showed that integrated membrane filtration could reduce more than 99.0% of conventional pollutants and more than 99.79% of ARGs, but it was not suitable for municipal drinking water treatment due to its high cost. Recent research also reported that sand filtration was superior to other conventional drinking water processes in terms of absolute abundance of ARGs removal (Tan et al. 2019), but its removal of antibiotics was not satisfactory, and furthermore, the filtered water did not directly meet the requirements of drinking water. Granular activated carbon (GAC) is an adsorbent material and filter material with a well-developed pore structure, high surface activity, and good mechanical and chemical stability, which is commonly used in the advanced treatment process of drinking water treatment to adsorb and remove fine particulate matter and antibiotics from water. The removal of quinolones by batch adsorption using GAC alone can achieve removal rates of more than 90% (Liu et al. 2017). However, when the water coexists with NPs, it has a strong adsorption effect on antibiotics due to its large specific surface area, which may increase the risk of antibiotic migration and diffusion in drinking water, and moreover, the static adsorption effect does not represent a treatment effect of the carbon filtration. The removal efficiency of carbon filter columns for quinolones in the presence of NPs was hardly reported, and the effect of NPs on the removal of quinolones by carbon filtration columns needs further study.

In this study, polystyrene nanoplastics (PSNPs) and NOR were selected as experimental materials and the orthogonal design method was used to design the experiment. The effects of four orthogonal factors, namely, PSNPs concentration, NOR concentration, filtration rate, and ionic strength, on NOR removal by the carbon filtration column were investigated. The possible influence mechanism was explored with characterization and transport curves. In addition, an orthogonal design and back-propagation (BP) network model were used to establish a prediction model for the removal effect of carbon filtration column on NOR under the influence of PSNPs. The model can predict the removal effect of carbon porous media on NOR under different conditions, and this model can greatly reduce the experimental volumes. The results contribute to further understand the adsorption characteristics of PSNPs on NOR on the surface of activated carbon and provide new ideas to predict the removal efficiency of quinolone antibiotics in the porous media of activated carbon when NPs and quinolones are coexisting. It also provides a theoretical basis for environmental and risk management of NPs and quinolone antibiotics.

PSNP microspheres were purchased from Big Goose Technology Co., LTD. (Tianjin, China). Using a transmission electron microscopy (TEM), the diameter of the PSNP microspheres was determined to be 98 ± 9 nm, with a density of 1.05 g/cm³ at 20 °C, and a specific surface area of 67 m²/g. Stock suspensions of 10 g/L were prepared from the original solution, and diluted with ultrapure water. PS stock suspension was sonicated for 3–5 min with a sonication bath before each extraction. The stock solution was stored in a dark place at a constant temperature of 4 °C and used for further experiments.

GAC of diameter 1.0–2.0 mm and pore volume of 0.357 cm³/g was purchased from Henan Huanyu Activated Carbon Co., LTD. (Henan, China). Then, the GAC was cleaned thrice by ultrasound for 5 min, and then rinsed with pure water for 3–5 times until the supernatant was clear.

NOR (98%) was purchased from Shanghai Aladdin Biochemical Technology Co., Ltd. (Shanghai China). NOR was dissolved in 4% sodium hydroxide solution, and then the NOR storage solution of 10 g/L was prepared with ultrapure water and stored in dark at 4 °C.

Scanning electron microscopy (SEM, FEI-F50) was used to observe the surface morphology of GAC before and after the experiment to help explain the mechanism of NOR removal in the carbon filtration column. The pore characteristics of the GAC were obtained by recording the adsorption isotherms of N₂ on ASAP 2460 (Micrometrics, USA) at the boiling temperature of liquid nitrogen (77 K). The Brunauer–Emmett–Teller (BET) method was used to calculate the surface area and pore volume of GAC samples. The NOR concentration was measured by UV-5200 spectrophotometer produced by Shanghai Yuanxi Instrument Co., Ltd. (Shanghai, China). Standard curves are shown in Supplementary Material, Figure S1.

Orthogonal experiments are able to reflect the influence of various factors on the removal of NOR from carbon filtration columns more comprehensively, and can greatly reduce the number of experiments that would otherwise be required. In this study, PSNPs concentration, NOR concentration, flow rate, and ionic strength were selected as orthogonal design factors, and four levels were determined for each factor. The orthogonal factor levels are shown in Table 1. Standard orthogonal tables L1₆(5⁴) were used to design the experiments according to the determined four factors and four levels. Also, the concentration of NaCl was used uniformly to control the ionic strength in each experiment.

The mixture solutions containing 3 mg of PSNPs and different NOR concentrations (2, 4, 6, 8, 10, 12, 14, and 16 mg/L) were prepared in 50-mL sample bottles in sequence. The sample bottles were placed in a thermostatic shaker (150 rpm, 25 °C) and shaken for 16 h against light before sampling. Then, the sampled solutions were filtered through a pinhole filter membrane (0.22 μm) in time, and the remaining concentrations of NOR were determined by visible-UV spectrophotometer. Blank samples without PSNPs and two parallel samples were set up for each group of experiments.

The experimental setup consists of magnetic agitator, peristaltic pump, carbon filtration column, and collector. Each part is connected by silicone tubes, and the carbon filtration column used in the experiment was made of Plexiglas acrylic column with an inner diameter of 15 mm and a height of 80 mm. The experimental setup sketch and more details of the setup are described in Supplemental Material, Figure S2. Before each experiment, pure water was pumped through the carbon filtration column using a peristaltic pump at a certain flow rate for 6 h to rinse impurities out of the column to reduce errors in NOR measurement. Also, prior to the transport experiments, 10 PV of background salt solution (5 mmol/L NaCl solution) was used to stabilize the experimental conditions, and the mixture of PSNPs (5 and 10 mg/L) and NOR (5 and 10 mg/L) was first equilibrated by dark adsorption in a constant temperature water bath shaker for 16 h. The experimental conditions are shown in Table 2. The transport experiments were performed by injecting 5 PV of prepared NOR solution or mixed solution of PSMPs and NOR, followed by 5 PV of a background solution containing the same ionic strength. The concentration of NOR in the effluent was measured for every 0.5 PV. A magnetic stirrer was used to continuously stir the mixed solution while the experiment was in progress to ensure that the PSNPs were uniformly dispersed in the solution. All solutions were injected into the column at 3 mL/min with empty bed contact time (EBCT) of 4.7 min and a hydraulic loading rate of 1.70 mL/min. Since this experiment is a dynamic process and the removal efficiency of GAC on NOR is high, the influence of PSNPs on the transport of NOR cannot be visualized by the outflow ratio of NOR, so the transport curve of NOR in the activated carbon filter column was plotted using the discharge concentration of NOR as the vertical coordinate and PV as the horizontal coordinate.

The PSNPs concentration, NOR concentration, flow rate, and ionic strength were selected as input variables, and the NOR removal rate at the relative saturation of the carbon filter column was selected as the output variable. Experimental groups 1–16 in Table 3 were used as training samples to train the neural network, and groups 17–18 were used as prediction samples to test the reliability of the prediction model.

Analysis of variance was performed with SPSS software based on the experimental results in Table 3. The results are shown in Table 6, and it can be seen that the degree of influence of each factor on the retention of NOR by the carbon filtration column was flow rate, NOR concentration, ionic strength, and PSNPs concentration, in that order. In addition, flow rate was a significant influence on NOR removal by carbon filtration column (p ≤ 0.05), and there was no significant difference in the effect of different levels of other orthogonal factors on NOR removal by the carbon filtration column (i.e., p > 0.05).

As shown in Figure 4, the lower the PSNPs concentration, the more favorable the removal of NOR by the carbon filtration column (i.e., k-value decreased as the PSNPs concentration increases from level 1 to level 4). On the one hand, when the carbon filter column is relatively saturated, some of the adsorption sites that should be occupied by NOR are occupied or obstructed by PSNPs, resulting in an increase in NOR effluent concentration. Also because of the concentration gradient, the higher the concentration of PSNPs, the more adsorption sites are occupied or obstructed, which is why the concentration of NOR leakage increases with PSNPs concentration when the carbon filter column is relatively saturated. On the other hand, a part of NOR might be adsorbed on the PSNPs and removed out of the carbon filtration column along with the water flow. This was confirmed in NOR transport experiments as well. The value of k gradually increased with the increase in the NOR concentration to the peak at level 2, and subsequently the value of k decreased with the further increase in the concentration. It may be due to the fact that in the carbon filtration column, the weights of the filled activated carbon were almost equal, and the numbers of their active sites were also close to each other. When the NOR concentration was low, most of the NOR could be adsorbed on the active sites of the activated carbon. But when the NOR concentration exceeded a certain limit, the NOR generated by the concentration exceeding the limit could not be retained in the carbon filtration column because the active sites were saturated, resulting in a decrease in the removal rate of NOR with the increase in the NOR concentration.

As shown in Figure 4, the value of k decreased as the flow rate increased. It may be at a low flow rate, the proportion of horizontal diffusion of the mixed solution was higher than that at a high flow rate, which increased the chance of NOR contacting with the activated carbon media, so more NOR was adsorbed on the activated carbon media. While at a high flow rate, the shear force was enhanced and the mixed solution passed through the carbon filtration column in a shorter time, which made it easier for NOR to penetrate the carbon filtration column. As shown in Figure 4, the value of k decreased, then increased, and again decreased with the increase of ionic strength, forming an inverted ‘N’ trend. According to previous studies, the penetration of PSNPs in the carbon filtration column decreased with the increase in the ionic strength (Ji et al. 2023), and the presence of Na⁺ would form a competitive adsorption relationship with NOR. Therefore, when the ionic strength increased from level 1 to level 2, the value of k decreased slightly, probably because the amount of NOR penetration due to competitive adsorption of Na⁺ was greater than the amount of NOR carried by PSNPs and trapped in the carbon filtration column. Also, when the ionic strength increased from level 2 to level 3, the PSNPs were further trapped on the surface of the activated carbon and the number of NORs carried by PSNPs and trapped was dominant, thus showing an increase in k-value. However, when the ionic strength was further increased from level 3 to level 4, the effect of the increase in the ionic strength on the retention of PSNPs by activated carbon was weakened, and the competitive adsorption of Na⁺ with NOR played a dominant role, which eventually led to the decrease in the k-value. The exact relationship of action needs to be determined by further experiments.

In order to further test the reliability of the prediction model, the experimental conditions of groups 17 and 18 were set by random combination for each orthogonal factor and each level to test the prediction performance of the model. The sim function was called to make model predictions for the 17 and 18 groups of experiments and to inverse normalize the results. Table 7 shows the comparison of experimental and predicted values for groups 17 and 18. The predicted values of the model in groups 17 and 18 were not significantly different from the experimental values, and the relative errors were 3.32 and 6.22%, respectively, which indicated that the network model had a good prediction effect.

Activated carbon filtration process is one of the advanced common drinking water treatment processes used to remove antibiotics. Due to the environmental risks of quinolone antibiotics and the transport properties of NPs as carriers, it is very important to investigate the removal efficacy of quinolone antibiotics under the influence of NPs. In this study, the isothermal adsorption lines of NOR by PSNPs and the effect of PSNPs on the transport of NOR in the carbon filter column were investigated. The PSNPs concentration, NOR concentration, flow rate, and ionic strength were selected as four orthogonal factors to design orthogonal experiments. Analysis of variance using SPSS software was used to investigate the effect of each orthogonal factor on NOR removal from carbon filtration columns and to speculate on the possible mechanism. The transport curves of PSNPs synergistic NOR were combined with characterization to verify the possible mechanism of PSNPs in NOR removal from carbon filtration columns. In addition, a prediction model with the four orthogonal factors as input variables and the removal efficiency of NOR at the relative saturation of the carbon filter column as output variables was also developed by the BP neural network, and the model was evaluated by two sets of prediction samples.

The results showed that the adsorption curves of PSNPs to NOR were more consistent with the Langmuir model. The increase of PSNPs concentration might hinder the binding of NOR to the active site of activated carbon, and the PSNPs might also carry a portion of NOR through the filtration column, resulting in a decrease in the removal rate of NOR. The degree of influence on NOR removal from the carbon filtration column was in the order of flow rate, NOR concentration, ionic strength, and PSNPs concentration, and the flow rate was a significant influence factor. The effects of other orthogonal factors were speculated by the analysis of variance. High flow rates lead to faster passage of the mixed solution through the carbon filtration column compared to low flow rates, resulting in a lower probability of NOR contact with the active site. Also, there was an optimum concentration for the NOR concentration on the removal of NOR by a carbon filtration column, lower or higher than this concentration, the removal rate of NOR decreased. However, the increase of ionic strength would promote the aggregation of PSNPs with each other and the formation of new microporous or mesoporous structures with activated carbon, which would promote the removal rate of NOR; on the other hand, the increase of ionic strength would also strengthen the competitive adsorption of Na⁺ with NOR, which would decrease the removal rate of NOR. These two aspects work together to make the NOR removal rate show an inverse ‘N’ effect, i.e., the NOR removal efficiency of the carbon filtration column showed a trend of decreasing, then increasing, and then decreasing with the increase of ionic strength. Moreover, the overall model fit was 0.9643, and the relative errors between the model predicted and sample experimental values were 3.37 and 6.62%, respectively, for the two sets of predicted samples, which indicated that the model prediction was good. This also provided a theoretical calculation basis and reference for the removal of NPs and quinolone antibiotic NOR in drinking water treatment plants using the activated carbon filtration process.

H.J., Z.L., X.X., W.J., S.W., B.W., and X.X. contributed to the study conception and design. Material preparation, data collection, and analysis were performed by H.J. and W.J. The first draft of the manuscript was written by H.J. and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

This work is supported by the Key Laboratory of Poyang Lake Environment and Resource Utilization, Ministry of Education, Nanchang University (No. 2022Y04).

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

The authors declare there is no conflict.