Norfloxacin removal by ultraviolet-activated sodium percarbonate and sodium hypochlorite: process optimization and anion effect

Most antibiotics are expelled in the urine and feces in non-metabolized forms, or leftover antibiotics are disposed of straight into sewers, posing a severe threat of pollution to the aquatic ecosystems, and other water-related ecosystems risk severe contamination (Lien et al. 2016; Wang et al. 2020; Ao et al. 2021). This eco-toxicological effect may not only develop antibiotic-resistant genes and bacteria, but it may also harm soil organisms and plant growth (Wang et al. 2020, 2021a). One of these drugs, the fluoroquinolone antibiotic Norfloxacin (Norf), is often prescribed for both human and veterinary usage. By limiting DNA gyrase, it has strong antibacterial activity against both Gram-positive and Gram-negative bacteria. Norf antibiotics may be found in agricultural settings, in hospital effluent, and in municipal treatment plants (Yang et al. 2012; Sui et al. 2012; Dou et al. 2019). It has been hypothesized that fluoroquinolones have negative impacts on the microbiota, including the modification of the structure and diversity of bacterial populations (Näslund et al. 2008; Cui et al. 2014). Additionally, fluoroquinolones give long-term stability in the environment by building complexes with a variety of ions, such as Ca²⁺, Mg²⁺, Fe³⁺, and Al³⁺; hence, it is vital to find efficient strategies for reducing the amount of Norf found in waters (Seifrtová et al. 2009).

The removal of antibiotics, particularly, which may not be effectively achieved by the use of conventional treatment, has been the subject of a lot of research and experimentation recently (Xu et al. 2015). Even if only a little quantity of elimination is accomplished by biological treatment, the antimicrobial activity of Norf and other antibiotics kinds may destroy the microbiota in the treatment facilities. Even though adsorption techniques appear to be suitable for the degradation of Norf due to their low cost, simple operation, lack of byproducts, and strong applicability, the adsorbents to be used for Norf adsorption have challenges like low adsorption capacities and regeneration issues (Liu et al. 2011; Sui et al. 2012; Yang et al. 2012). Antibiotics like Norf may also be removed from waters using another treatment method, membrane processes; however, these high-pressure membrane technologies have drawbacks like high energy consumption and rapid fouling tendencies (Watkinson et al. 2007; Greenlee et al. 2009; De Souza et al. 2018). Apart from the mentioned methods, another method that has been working frequently recently is advanced oxidation processes (AOPs), a state-of-the-art alternative treatment method that can break down organic pollutants into nontoxic small molecules.

AOPs include several processes that create excess reactive oxygen species (ROS) such as hydroxyl and sulfate radicals, which may be responsible for the fast destruction of wastewater contaminants through Fenton-like oxidation, ozonation, electrochemical oxidation, and photocatalytic oxidation (Brillas & Martínez-Huitle 2015; Yuan et al. 2020; Ghanbari et al. 2021). Many potential byproducts might develop in an oxidation system because of the many reactions that take place throughout the process, in addition to the primary interactions with the pollutants of interest (Nawaz & Ahsan 2014; Brillas & Martínez-Huitle 2015; Fedorov et al. 2020, 2022; Bilińska & Gmurek 2021). There are several limitations to the use of commonly used oxidants such as hydrogen peroxide (HP), persulfate (PS), and peroxymonosulfate (PMS). Sulfate radicals are selective, while pH and transport are the limiting variables in the application of HP (Fischbacher et al. 2017; Li et al. 2019). As a result of the limiting impact that the oxidants indicated above have, recent research has focused on testing novel oxidants. Sodium percarbonate (SPC) is an oxidant that forms hydroxyl and carbonate radicals and stands out from other oxidants owing to its stability during transit and storage, pH buffering capabilities, affordable price, and availability via techniques that are kind to the environment (Viisimaa & Goi 2014; Miao et al. 2015). Green oxidants like SPC are preferable to traditional oxidants since they have fewer negative effects on the environment. The extremely reactive hydroxyl radicals created by ultraviolet (UV) activation allow SPC to destroy a broad spectrum of pollutants, including organic chemicals and microorganisms. When compared to other oxidants, SPC is rather affordable and easy to use (Viisimaa & Goi 2014; Miao et al. 2015).

Disinfecting using sodium hypochlorite (SHC) is a common practice. Having a high oxidation capacity, it quickly oxidizes contaminants in addition to being able to kill germs efficiently (Bottone et al. 2022; Xu et al. 2022). SHC is a potent disinfectant that can eradicate many different types of bacteria, viruses, and protozoa. Since SHC is capable of quickly oxidizing both organic and inorganic molecules, it is a viable option for treating a wide variety of water pollutants. Like SPC, SHC does not need any special apparatus to use and may be poured straight into water (Bottone et al. 2022; Xu et al. 2022). Although oxidants should be ecologically safe, have a high oxidant capacity, and are cost-effective, they must also be carefully selected for activation techniques. Discharge plasma, photocatalysis, and metal ions are all examples of homogeneous activation techniques, whereas metal oxides and nano-sized zerovalent iron are examples of heterogeneous activation methods. Full contact between reactants is guaranteed by homogeneous activation, and the produced radical is powerful enough to destroy most contaminants (Guo et al. 2021; Wu et al. 2018; Liu et al. 2021). UV irradiation offers several benefits, including the capacity to efficiently remove organic contaminants in the presence of powerful oxidants, no sludge formation, ease of operation, and is eco-friendly (Fernandes et al. 2019, 2020; Lin et al. 2020; Li et al. 2021; Ribeiro & Nunes 2021; Wang et al. 2021b).

Ao et al. (2021) attempted Norf removal using medium-pressure UV/peracetic acid and achieved 97.2% removal at pH 9, while Chen et al. (2021) used ionizing radiation combined with a Fenton-like oxidation process and reached 98% removal after 2 kGy radiation (Ao et al. 2021; Chen et al. 2021). Jin et al. (2019) synthesized spherical N-doped TiO₂ to degrade Norf under visible light irradiation and reported that reached 99.5% removal after 30 min (Jin et al. 2019). Kim et al. (2009) evaluated the effectiveness of UV + HP in removing Norf from secondary sedimentation and sand filter effluent and found that at 7.8 mg/L HP concentration, 69% of the antibiotic was degraded (Kim et al. 2009). While Xue et al. (2019) used UV/PS, Norf degradation reached 100%, but Guo et al. (2017) achieved 62.5% degradation using UV/PS (Guo et al. 2017; Xue et al. 2019). Zhao et al. (2022) reported that MIL-101(Fe)-NH₂@Al₂O₃ (MA) catalysts were successfully synthesized by the reactive seeding method on α-Al₂O₃ substrate, which demonstrated excellent (>90%) photo-Fenton degradation performance toward Norf.

Numerous AOPs have been attempted in the literature for Norf removal from water solutions. No studies on the efficacy of UV-activated SPC and SHC in removing Norf from water solutions have been undertaken, as far as we are aware. The effectiveness of the UV-SPC process was measured and compared to that of the UV alone, the SHC process, and the UV-SHC process in this study. Interactions between process parameters and their effects on the UV-SHC and UV-SPC processes were analyzed. Cost analysis and water matrix effect on the UV-SPC and UV-SHC processes were also evaluated.

Highly pure Norf (C₂₂H₂₄N₂O₉.HCl, MW: 496.89, purity > 95%), sodium hypochlorite (NaOCl), sodium percarbonate (Na₂CO₃.1.5H₂O₂) was obtained from Sigma-Aldrich, while, sodium hydroxide (NaOH) and sulfuric acid (H₂SO₄) were obtained from Merck Co. (Germany). Methanol was used to produce stock Norf solutions. At a wavelength of 274 nm, a UV-visible spectrophotometer (WTW, Photo lab 6600 UV-vis, Turkey) was used to determine Norf concentrations. The samples' pH levels were determined using a WTW Multi 9620 IDS. Water matrix chemicals (sodium chloride (NaCl), sodium bicarbonate (NaHCO₃), sodium sulfate (Na₂SO₄), and sodium nitrate (NaNO₃)) were also bought from Sigma-Aldrich.

In which the concentration of Norf at time t in the reaction is denoted by C_t. C_i is the initial Norf concentration, k (1/min) is the first-order rate constant, and t (min) is the reaction time.

The results obtained in the positive range show the synergistic impact, while the negative range represents the antagonistic effect.

Detailed information about response surface methodology (RSM) and Central Composite Design (CCD) is given in Text S1. Initial Norf concentration, SPC/SHC dosage, and degradation time, which are operational parameters of UV-SPC and UV-SHC processes, were selected as independent variables, with Norf degradation efficiency as the model response. Table 1 provides the value range and coded level for each component as established by the exploratory research, and Table 2 details the 17 experimental sets generated using the CCD for three variables. After the experimental investigation, validation analysis was conducted to assess the applicability of the model.

As a result of control experiments performed with SPC (pH 3, 2.5 mg/L initial Norf concentration, 3 mM SPC dose, and 45 min reaction time), Norf removal yields were 8.2, 10.8, and 53.5% with UV, SPC, and UV-SPC, respectively. The calculated first-order reaction kinetic rate constants were 0.0019, 0.0025, and 0.0171 L/min for UV, SPC, and UV-SPC, respectively. The synergistic effect of the UV-SPC process was determined as 2.89. The positive SE values obtained for both processes show the synergistic effect. The synergistic effect of the UV-SPC process is higher than that of the UV-SHC process, which can be explained by the fact that SPC activation is more effective with UV application or a single SHC has a higher oxidant effect compared to a single SPC. Furthermore, the reaction time of the UV-SHC process is much shorter than the UV-SPC process. Similar Norf degradation efficiencies were obtained in a much shorter time.

A, B, and C represent Norf concentration, SPC/SHC dosage, and degradation time, respectively. Table 2 shows the model-estimated operating conditions and design matrix data from experiments. The operational parameters affect the UV-SPC process more than the UV-SHC process, and model predictions and experimental data match.

Testing the appropriateness of the model to the data is perhaps the most crucial stage of the investigation. Inadequate model fitting may lead to inaccurate inferences about the true value of the modeled variables. To test the model's fit, certain coefficients must be established. In this regard, the analysis of variance (ANOVA) was conducted, and Tables 3 and 4 display the ANOVA findings. According to the results of the ANOVA, there are two distinct sources of the total variance: the model itself and experimental errors. The model's relevance is calculated by contrasting these two forms (Aleboyeh et al. 2008). The model F-value is obtained by dividing the model mean square value by the residual mean square value and is used for comparison (Arslan-Alaton et al. 2009). When the F-value is large, the majority of the observed variance in the response is explained by the model; when the p-value is less than 0.05, the model is significant; and when the p-value is less than 0.0001, the model is highly significant. If the computed F-value is larger than the tabulated F-value, then the Fisher's variance ratio is big enough to suggest a high degree of compatibility between the quadratic model and the data, and the degradation combinations are very significant (Yetilmezsoy et al. 2009). Furthermore, the calculated F-value was found to be greater than the tabulated F-value at the 5% level. It may be concluded with 95% confidence that the regression model adequately explains the variance in the independent variable if F_cal > F_tab, as determined by Fisher's F test. Table 3 demonstrates that the calculated F-values for the models exceed the theoretical value. Model F-values of 42.18 and 9.69 for UV-SPC and UV-SHC processes, respectively, might be considered as being sufficiently broad. Similarly, p-values for both replies were determined to be less than 0.05. Large F-values and small p-values reflect the models' relevance (Table 3).

The coefficient of determination (R²) is the initial measure of model fitness. Ideally, the R² value of a model with high predictive ability would be close to 1 (Davarnejad et al. 2014). However, the R² value is insufficient for judging the model's predictive power on its own, since it increases monotonically with the number of items in the model, independent of statistical significance. The R² value and the adjusted R² value (which accounts for the total number of factors in the experiment) should be compared (Montgomery & Runger 2010). A considerable discrepancy between R² and adjusted R² shows that the model includes insignificant terms. The gap between the estimated and real values is referred to as the difference, and the values of the different play a significant part in determining model compatibility. As the number of independent factors that have a substantial impact on the dependent variable grows, the adjusted R² value goes up, and as the number of non-significant variables is added to the model, the adjusted R² value decreases. Again, regardless of whether it is significant, as the number of variables increases, so does the R² value. Consequently, as the number of insignificant factors increases, so does the disparity between the R² value and the adjusted R² value (Hassani et al. 2016). Table 3 demonstrates that the difference between R² and adjusted R² values for all responses is small. Both models' high correlation coefficients indicate that regression adequately describes the relationship between independent variables and the outcomes of the experiments. The coefficient of variation (CV) may be calculated by dividing the standard error of the estimate by the mean value of the observed response. To ensure the model's reproducibility, the CV value should not exceed 10% (Davarnejad et al. 2014; Biglarijoo et al. 2016). The signal-to-noise ratio metric is used to determine an acceptable level of accuracy. The average calculated error is compared to the range of possible adequate precision (AP) design point values. It might be suggested that the quadratic model could be used to investigate the design space when the AP value is more than 4, which indicates an adequate signal for all responses (Zinatizadeh et al. 2006; Darvishmotevalli et al. 2019). Both models have CV values less than 10 and AP values larger than 4. This result may be considered acceptable for CV values since both processes can be controlled by easy procedures. The UV-SPC and UV-SHC procedures yielded AP ratios of 22.72 and 11.00, respectively. Since these ratios are more than the necessary 4 for AP, the model provides an adequate signal for the response.

Lack of fit may also be used to assess model fit. The lack of fit test is used to measure the inaccuracy of the model in terms of how well it represents data points within the constraints of the experimental run. The test is based on the residual error-to-pure error ratio, which quantifies model error by using repeated experimental design points, the majority of which are the center points. The fit value should be negligible. The lack of fit test is a good indicator that the system works as intended and that the model can be used to predict Norf degradation yields in the specified variable range (Ozturk & Yilmaz 2019). A significant lack of fit suggests that there may be unexplained systematic fluctuations in the model. The lack of fit F-values were 2.89 and 1.89 for UV-SPC and UV-SHC processes, respectively, while the lack of fit p-values were 0.28 and 0.38 (Table 4).

Using a Pareto chart, the standardized effects of independent factors on dependent variables and their interactions were established (Figure 3). The length of each bar on the Pareto chart indicates the magnitude of each factor's standardized influence on the response (Yetilmezsoy et al. 2009). It may be understood that the contribution of parameters with short bars to the degradation efficiency is low, while the contribution of parameters with long bars is high. All linear factors have a substantial influence on Norf degradation, as shown in Figure 3. For the UV-SPC process, the relevance of the linear parameters is as follows: degradation time > SPC dose > Norf concentration, while for the UV-SHC process, it is degradation time > Norf concentration > SHC dose. Quadratic parameters are ineffective for both procedures. Only the degradation time has an impact on the UV-SPC process, but it does not influence the UV-SHC process. The findings of the Pareto analysis are consistent with those of the ANOVA analysis.

The solution proposed using numerical optimization methods was tested experimentally using the optimal values of the independent variables to verify and assess the solution's correctness. The validation studies yielded findings that were in close accord with the degradation of Norf predicted by software taking into consideration the standard deviation of the experimental responses. The optimum operating conditions determined by the applied models, the predicted Norf degradation efficiency, and the results of the validation experiments were given in Table 5. The optimum conditions for the UV-SPC process were 1 mg/L initial Norf concentration, 4 mM SPC dose, pH 3, and 50 min degradation time. For the UV-SHC process, optimum process variables were 1 mg/L initial Norf concentration, 1 mM SHC dose, pH 7, and 8 min degradation time. For the UV-SPC and UV-SHC procedures, the model predicted a 72.9 and 73.3% degradation efficiency under optimal conditions, whereas the degradation efficiencies achieved from the validation trials were 71.8 and 72.1%, respectively. Higher Norf removal efficiencies compared to our results were reported in the literature by the UV/peracetic acid (97.2%) (Ao et al. 2021), ionizing radiation combined with a Fenton-like oxidation (98.0%) (Chen et al. 2021), and N-doped TiO₂ under visible light irradiation (99.5%) (Jin et al. 2019) processes. On the other hand, lower removal efficiencies were reported by Guo et al. (2017) with the UV/PS (62.5%) and by Kim et al. (2009) with the UV/HP (69.0%) processes.

The EE/O of the UV-SPC and UV-SHC processes for the optimized process variables were determined as 485 and 77 kWh/m³, respectively. The reason for the huge difference between the EE/O values of these processes is because of the longer reaction time required in the UV-SPC process. Since less electricity is required to attain the desired level of treatment, lower EE/O values indicate a more efficient process. Considered in this respect the UV-SHC process is more energy efficient with similar Norf removal efficiencies.

For the cost calculation, the optimum oxidant doses given in Table 5 were used and chemical costs were found as 28.8 and 1.7 US$/m³ for UV-SPC and UV-SHC processes, respectively. Adding up with the energy costs, the operating costs were calculated as 65.6 and 7.5 US$/m³ for the UV-SPC and UV-SHC processes, respectively. The operating cost of the UV-SPC process was higher due to both the longer reaction time and the higher oxidant dose.

The Norf removal performance of UV-SPC and UV-SHC processes from aqueous solution was evaluated within this study. The effect of process variables (initial Norf concentration, oxidant dose, and degradation time) on Norf removal by the UV-SPC and UV-SHC processes was determined by central composite design. According to the control experiments, the Norf removal efficiency was higher in UV-activated SPC and SHC processes than the alone UV, SHC, and SPC processes. Based on ANOVA results, the applied models were found to be significant in Norf removal for both processes. Norf removal efficiency increased with an increase in oxidant dose and degradation time while it decreased with an increase in initial Norf concentration. Adding anions to the solution decreased the Norf removal efficiency of both processes. As a result, UV-SPC and UV-SHC processes were both effective in Norf removal from aqueous solution. However, the similar removal efficiency was achieved in a much shorter time and much little cost with the UV-SHC process. Considering the advantages and disadvantages of both processes, it should be evaluated in terms of economic and environmental sustainability, and the choice for the final purpose should be made.

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

The authors declare there is no conflict.