Degradation of 2,6-dichloro-1,4-benzoquinone by advanced oxidation with UV, H₂O₂, and O₃: parameter optimization and model building

Trace organic pollutants of natural and human origins are often detected in drinking water sources (Shimabuku et al. 2019). Disinfection by-products (DBPs) are produced by the reaction of dissolved organic matter in water with chlorine (Nikolaou et al. 2004). Because of their confirmed or potential carcinogenicity, many DBPs have received increasing attention, especially in the treatment of slightly polluted water sources.

Halobenzoquinones (HBQs) have been identified as an emerging class of DBPs. HBQs are cytotoxic and may be carcinogenic and genotoxic and cause DNA damage through oxidative stress (Wei et al. 2016). Water containing HBQs should be treated to protect human health. Zhao et al. (2010) detected HBQs in tap water from Canada and found that the compounds with the highest concentrations were 2,6-dichloro-1,4-benzoquinone (DCBQ; 165.1 ± 9.1 ng/L), 2,6-dichloro-3methyl-1,4-benzoquinone (1.3 ± 0.2 ng/L), 2,3,6-trichloro-1,4-benzoquinone (9.1 ± 0.6 ng/L), and 2,6-dibromo-1,4-benzoquinone (0.5 ± 0.1 ng/L). Zuo et al. (2017) evaluated the toxicity of DCBQ using Caenorhabditis elegans, and found that DCBQ had stronger general toxicity than controlled DBPs, and caused some inheritable genetic mutations in vivo.

Advanced oxidation processes remove organic pollutants from water and can greatly reduce the reaction time and make mineralization easier. Various organic pollutants can be completely decomposed in a few hours by the generated free radicals, which are strong oxidants (Chen et al. 2018). Ozonation is used as a treatment technology to remove pollutants from water through direct ozonation and indirect hydroxyl radical (HO^•)-induced oxidation reactions. The main ozone-based advanced oxidation processes are O₃, UV/O₃, and UV/H₂O₂/O₃. Research has shown that the DCBQ removal rate with potassium ferrate (K₂FeO₄) can be as high as 87.12%, but the reaction is greatly affected by the pH with the removal rate decreasing under alkaline conditions (Ding et al. 2018). Advanced oxidation technology with UV utilizes the strong oxidation property of HO^· (E₀ = 2.80 V) to non-selectively degrade organic pollutants (Du et al. 2020) in a green method with high efficiency. This method has been applied in drinking water plants around the world. In addition, as a back-end deep treatment process, this method can control HBQs generated in the process of containing or pre-oxidizing raw water. Further research on the degradation of HBQs by UV advanced oxidation process is required.

The controlled variable method requires a large number of experiments to determine the optimum level for a variable and cannot reflect the comprehensive effect of various factors. Response surface methodology (RSM) can jointly optimize all parameters and overcome the limitations of the controlled variable method (Elibol 2002). RSM can be used to determine the optimum conditions for a multivariable system. It is a collection of mathematical and statistical techniques that can be used to develop, improve, and optimize processes. Using RSM, we can evaluate the relationship between experimental factors and measured responses and the relative influence of factors. The main goal of RSM is to determine the optimum operating conditions of a system to meet the operating specifications. In RSM designs, low-, medium-, and high-level codes are usually expressed as −1, 0, and +1, respectively. In different experiments, the levels of one factor are combined with the levels of other factors. A mathematical model of the relationship between the response value and each factor is generated from the results of the optimization experiments (Schenone et al. 2015).

In this study, DCBQ, which has the highest detection frequency and highest concentrations in water among the HBQs, was selected as the target pollutant. The suitability of the UV/H₂O₂/O₃ process for the removal of DCBQ was evaluated. The main purpose of this study was to study the effects of O₃, H₂O₂, and UV on DCBQ removal. A Box–Behnken design (BBD) was used to study the effects of the UV dose, H₂O₂ concentration, and O₃ concentration on their own and in interaction, and a quadratic model was developed for DCBQ removal. The influence of natural organic matter (NOM) on the UV/H₂O₂/O₃ degradation pathway was also explored.

Formic acid and methanol (chromatography grade) were used in this study. DCBQ was purchased from Sigma–Aldrich (St. Louis, MO). Humic acid was purchased from Tianjin Guangfu Fine Chemical Research Institute (Tianjin, China), formic acid was purchased from Tianjin Kermel Chemical Reagent Co., Ltd (Tianjin, China), and methanol was purchased from Xilong Chemical Engineering Co., Ltd (Guangdong, China). The solutions were prepared with ultrapure water, and all equipment used was washed with ultrapure water at 105 °C for 24 h before use. Hydrogen peroxide (H₂O₂, 30 wt%, analytical reagent grade) was purchased from Sinopharm Chemical Reagents Co., Ltd (Shanghai, China).

Experiments were carried out at the pilot plant of the Quehua Water Plant in Jinan, China. The test device (Figure 1) included a UV reactor with a volume of 2.4 L, which contained an ultraviolet lamp with intensity of 420 μW/cm². The DCBQ solution was pressurized and transferred to the UV reactor by a centrifugal pump. The main pipeline was connected to a H₂O₂ solution feeding facility and an O₃ gas feeding facility. The H₂O₂ solution feeding facility contained H₂O₂ solution that was transferred to the main pipeline by a metering pump. The O₃ gas feeding facility contained O₂, which was used to convert a certain concentration of O₃ gas by an O₃ generator. The generated O₃ gas was then transferred to the main pipeline through a gas flow meter. A gas concentration detector was used to detect the O₃ concentration.

Before the experiments, the equipment was cleaned with purified water, which was removed before the test. A stock solution of DCBQ in methanol (50 mg/L) was diluted with water to prepare a DCBQ solution with a concentration of 1 μg/L. This solution was placed in the DCBQ solution tank. Next, a H₂O₂ solution was prepared and placed in the H₂O₂ solution tank. A centrifugal pump was used to control the liquid flow rate (Q_L) in the main circuit and adjust the residence time of the solution in the reactor. The O₃ and H₂O₂ concentrations in the gas flow meter and the metering pump control reactor were adjusted using Equations (1) and (2). Samples (500 mL) were collected at set intervals (1, 3, 5, and 7 min) for testing. Appropriate volumes of formic acid and 0.1 mol/L sodium thiosulfate aqueous solution were added for solid-phase extraction (SPE) on-machine detection.

The HBQ concentrations in the water samples after the advanced oxidation were determined by SPE ultra-performance liquid chromatography (UPLC) tandem MS. The detection limit of this method was 0.01 μg/L. The recoveries were 80.0–85.0%, and the relative standard error is 3.8–4.9%.

Water samples were collected in brown bottles, and an appropriate volume of formic acid was added to adjust the pH to 2.6–2.8. Pretreatment by SPE was used to concentrate and enrich the target substances in the water samples. Before adding the sample solution, the Water Oasis HLB SPE column (Waters; Shanghai, China) was activated with a methanol solution containing 0.25% formic acid. And the column was washed twice with an aqueous solution containing 0.25% formic acid. During loading, each sample (500 mL) passed through the column at a flow rate of 8 mL/min. After loading the sample, the column was washed with an aqueous solution containing 0.25% formic acid and a solution containing methanol/water (50:50, v/v) containing 1% formic acid, and dried in vacuum for 10 min. The analyte from the extraction column is then eluted into a collection tube with a methanol solution containing 0.25% formic acid. The collected liquid was reduced dryness under a stream of nitrogen, and the residue was reconstituted in 1 mL of a water/methanol solution containing 0.25% formic acid (80:20, v/v) and placed in a sample bottle for analysis.

Chromatography was performed with an Acquity UPLC BEH C18 analytical column (2.1 mm i.d., 50 mm long, 1.7 μm particle diameter, Waters, Shanghai, China). The mobile phase was a mixture of an aqueous 0.25% formic acid solution (A) and methanol 0.25% formic acid solution with a flow rate of 0.4 mL/min. A gradient elution was performed as follows: 0–1 min, 20% B; 1–12 min, increased from 20% to 90% B; 12–14 min, 90% B; reduced to 20% B within 0.01 min; and 14.01–18 min, 20% B. The sample volume was 10 μL.

The MS was performed using electrospray ionization, negative ion scanning, and multiple reaction monitoring with the following parameters: negative ionization voltage, −4,500 V; solvent removal temperature, 350 °C; curtain air pressure, 207 kPa; spray gas pressure, 345 kPa; and auxiliary heating air pressure, 345 kPa.

UPLC–MS was used to identify DCBQ degradation intermediates. The MS was operated in electron impact ionization mode with an ion source temperature of 350 °C. Data were acquired 5 min after injection, and the m/z scan range was 30–500. The carrier gas pressure was kept constant at 100 kPa. The ionization energy was 70 eV, the MS resolution was 60,000, and all other parameters are set to default values.

The BBD and Central Composite Design are the most commonly used designs in Design-Expert. Compared with the BBD, the Central Composite Design requires more experiments, takes longer, and is more expensive for establishing model equations (Ferreira et al. 2007). Therefore, we used the BBD to evaluate the effects of the UV dose, O₃ concentration, and H₂O₂ concentration on the DCBQ removal rate. The values of the independent variables are shown in Table 1.

Advanced oxidation processes with O₃ (2 mg/L), H₂O₂ (2 mg/L), and UV (420 μW/cm²) were investigated for DCBQ removal (average concentration of 1 μg/L) (Figure 2).

The effect of H₂O₂ oxidation alone on DCBQ was not ideal. Because H₂O₂ itself is not a strong oxidant, Baciogliu & Otker (2003) used a H₂O₂ solution with a maximum concentration of 100 mmol/L to treat wastewater for 1 h. No changes in the chemical oxygen demand or UV₂₅₄ were observed during the experiment.

Although DCBQ can be removed by HO^· in UV indirect photolysis, UV photolysis of water molecules cannot generate sufficient HO^· for this process (Yan 2019). Consequently, DCBQ removal by UV indirect photolysis is negligible. Therefore, DCBQ removal using UV radiation on its own can take a long time.

To overcome these deficiencies, we used a BBD to optimize the parameters of the UV/H₂O₂/O₃ process for DCBQ removal.

Experiments were carried out using a three-factor, three-level BBD, and 17 groups of experiments with five center points were performed, which made the mathematical model statistically consistent. The dosage of UV, H₂O₂, and O₃ in the test was determined according to the above test. The results are shown in Table 2.

The model was statistically evaluated using the determination coefficient (R²) and analysis of variance results. The analysis of model variance is shown in Table 3. The P-value of ≤0.0001 was taken as highly significant, P ≤ 0.05 as significant, and P > 0.05 as not significant.

The F-value of the model was 44.80 (i.e., >0.001) (Table 3) and indicated that the model was highly significant. The P-value of the model was <0.0001 and indicated that the total variance that the model could not explain because of errors was only 0.01%. Mismatch terms were used to test the adequacy of model fitting according to data changes in the model. The P-value of the mismatch term was greater than 0.05, indicating that the F-value of the mismatch term (6.48) was not statistically significant. In addition, the mismatch term was not significant, which indicated that there was a significant correlation between the variable and response value. The P-values of X₁, X₂, X₃, X₁X₂, X₁X₃, X₂X₃, and were all less than 0.05, indicating that these terms were significant in the model. The coefficient of determination (R²) is the ratio of the sum of regression squares and the sum of total squares and can be used to evaluate model fitting. In our experiments, R² = 0.9829 showed that the model could fit 98.29% of the total variables, so the model could fit the test data well (Sahoo & Gupta 2012). The corrected R² (0.9610) also showed that the model had good fit. For R², an increase will be observed as the independent variable increases. By contrast, for the corrected R², an increase will only be observed as the independent variable that has a significant impact on the model increases; otherwise, the corrected R² will decrease. Therefore, a small difference between R² and the corrected R² indicates that the model has good fit.

The coefficient of variation (CV) is the ratio of the standard error of the model to the average value of the actual response. When the CV value is less than 10%, the model is considered repeatable (Biglarijoo et al. 2016). The AP (Adeq Precision) value represents the range of predicted response values and their associated errors. The CV value of the present model was 6.04% (i.e., <10%) and the AP value was 21.322 (i.e., >4), which indicated that the model had high applicability. The F-value indicates the importance of the influence of an independent variable on the model. In this case, the order of influence of the independent variables on the model was O₃ concentration > UV dose > H₂O₂ concentration (Table 3).

The relationship between the measured value and the model's predicted response is shown in Figure 3(a), and the residual distribution of the model is shown in Figure 3(b). This relationship was almost linear, which indicated that the model had good fit.

A three-dimensional response surface map and a two-dimensional contour map of the regression model were obtained by fixing one variable and changing the other two variables. The interaction of various factors affected the curvature of the response surface map and the contour map. Figure 4 shows the response surface and contour plots of the effects of various factors on the DCBQ removal rate.

As the O₃ concentration and UV dose increased, the response surface graph and the contour map darkened (Figure 4(a1) and 4(a2)), indicating that the DCBQ removal rate increased. These results reflected the interaction between the UV dose (A) and O₃ concentration (B). When the H₂O₂ concentration was fixed (1 mg/L), no O₃ was added, and the UV dose was 100 mJ/cm², the DCBQ removal rate was only 52.5%. However, as the O₃ concentration increased, the DCBQ removal rate increased to 95.4%. In addition, when the O₃ concentration was fixed at 1 mg/L and the UV dose was increased from 0 to 200 mJ/cm², the DCBQ removal rate increased from 68.9 to 98.08%. When considering the DCBQ removal rate, the contribution of the O₃ concentration was more important than that of the UV dose.

When the O₃ concentration was fixed, as the UV dose and H₂O₂ concentrations increased, the response surface graph and contour map darkened and the DCBQ removal rate increased (Figure 4(b1) and 4(b2)). These results reflected the interaction between the UV dose (A) and H₂O₂ concentration (C) on the DCBQ removal rate. The changes observed showed that the order of importance of the variables for their influence on the DCBQ removal rate was UV dosage > H₂O₂ concentration.

When the UV dose was fixed, as the O₃ and H₂O₂ concentrations increased, the color of the response surface map and the contour map darkened. The DCBQ removal rate changed (Figure 4(c1) and 4(c2)). These results reflected the interaction of the O₃ concentration (B) and H₂O₂ concentration (C) on the DCBQ removal rate. The changes observed showed that the degree of influence of the variables DCBQ removal rate was in the order O₃ concentration > H₂O₂ concentration. Therefore, the order of influence of the three variables on the DCBQ removal rate was O₃ concentration > UV dose > H₂O₂ concentration.

The numerical optimization function in Design-Expert 10.0 was used to optimize the conditions for DCBQ removal. The measured values were compared with predicted values from the model (Table 4). Under the optimum conditions, the actual DCBQ removal rate was 94.30% and the predicted DCBQ removal rate was 95.34%. The relative error between the measured value and the predicted value was 1.1%, which indicated that the model had good applicability and reliability. The model has a certain guiding significance for the treatment of actual water body by advanced oxidation processes (AOPs). The HO^· produced by UV/H₂O₂/O₃ has a good effect on the disinfection and sterilization of drinking water (Li 2013).

Research has shown (Jia 2019) that NOM in water will affect the efficiency of advanced oxidation. Humic acid (HA) is a typical type of NOM in wastewater and surface water; therefore, we studied the influence of HA as a representative NOM on DCBQ removal.

During the experiment, the initial concentration of DCBQ was 1 μg/L, the UV intensity was 420 μW/cm², the average concentration of H₂O₂ was 1.36 mg/L, and the average concentration of O₃ was 0.51 mg/L. The HA concentrations were set at 0, 1, 5, and 10 mg/L, and the solutions were sampled at 1, 3, 5, and 7 min. The influence of HA concentration on the removal of DCBQ is shown in Figure 5.

DCBQ was removed from water using UV, H₂O₂, and O₃ advanced oxidation processes with removal rates of 36.1, 32.0, and 57.9%, respectively. These removal rates were not ideal, and these processes are not suitable for this application. Consequently, we used a BBD to optimize the operating parameters for the combined UV/H₂O₂/O₃ advanced oxidation process, and a quadratic model was established. The test data agreed well with the model and the correlation coefficient was greater than 0.98. The order of influence of the three factors for their effects on the DCBQ removal rate was O₃ concentration > UV dose > H₂O₂ concentration. The model we used had good applicability and reliability. We found that NOM in the water matrix affected the degradation of DCBQ in the UV/H₂O₂/O₃ process. When the concentration of HA was 1 mg/L, it had a positive effect on DCBQ removal, while higher HA concentrations inhibited DCBQ degradation.

The research was completed at the pilot scientific research base of the National Science and Technology Major Project of Jinan Water Supply and Drainage Monitoring Center under the Evaluation and Standardization of Key Technologies for Urban Water Supply System Operation Management project (2017ZX07501002). We thank the Shandong Province Water Supply and Drainage Monitoring Center for assistance with the experiments and valuable discussions.

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