Optimization of Fenton Technology for Recalcitrant Compounds and Bacteria Inactivation

In this work, the Fenton technology was applied to decolorize methylene blue (MB) and to inactivate Escherichia coli K12, used as recalcitrant compound and bacteria models respectively, in order to provide an approach into single and combinative effects of the main process variables influencing the Fenton technology. First, Box–Behnken design (BBD) was applied to evaluate and optimize the individual and interactive effects of three process parameters, namely Fe2+ concentration (6.0 × 10−4, 8.0 × 10−4 and 1.0 × 10−3 mol/L), molar ratio between H2O2 and Fe2+ (1:1, 2:1 and 3:1) and pH (3.0, 4.0 and 5.0) for Fenton technology. The responses studied in these models were the degree of MB decolorization (D%), rate constant of MB decolorization (kapp) and E. coli K12 inactivation in uLog units (IuLog). According to the results of analysis of variances all of the proposed models were adequate with a high regression coefficient (R2 from 0.9911 to 0.9994). BBD results suggest that [H2O2]/[Fe] values had a significant effect only on D% response, [Fe2+] had a significant effect on all the responses, whereas pH had a significant effect on D% and IuLog. The optimum conditions obtained from response surface methodology for D% ([H2O2]/[Fe] = 2.9, [Fe2+] = 1.0 × 10−3 mol/L and pH = 3.2), kapp ([H2O2]/[Fe] = 1.7, [Fe2+] = 1.0 × 10−3 mol/L and PH = 3.7) and IuLog ([H2O2]/[Fe] = 2.9, [Fe2+] = 7.6 × 10−4 mol/L and pH= 3.2) were in good agreement with the values predicted by the model.


Introduction
Some of the effluents produced by industries such as textiles, dyes, tanneries, cosmetics and pulp are colored [1]. In the pulp industry, effluents are colored due to the presence of lignin byproducts and other phenolic compounds formed [2]. These compounds are considered dangerous and recalcitrant because of their low biodegradability and resistant to chemical degradation [1,3]. Besides, recalcitrant compounds with biological activity contained in treated effluent discharges are generating a loss of biodiversity in ecosystems. Even more, some of these compounds with benzyl and phenolic structures are considered endocrine disruptors [4][5][6][7]. In addition to the presence of recalcitrant compounds, the bacteria in the effluents of the pulp industry must also be seriously considered. The presence of bacteria in effluents discharged to water bodies are generating humans and animals diseases. In this sense, Escherichia coli and other bacteria have been identified in pulp industry effluent [8]. The presence of these bacteria in the effluents of the pulp industry raises an important concern regarding the current technologies (biological treatment) and regulations that govern the discharge of these effluents [9]. The inactivation of a wide range of pathogens in the cellulose industry effluent is effective by chlorination at a relatively low cost [10,11]. However, despite its effectiveness there is a problem to consider: the formation of organochlorine compounds [12]. Tawabini, et al. [13] states that chlorine has a high reactivity that affects the formation of these byproducts (chlorinated organic compounds) when reacting with organic matter. These byproducts are characterized by high toxicity and mutagenic capacity for the environment.
The low effectiveness of conventional water treatments in the destruction of recalcitrant contaminants and the formation of hazardous byproducts has encouraged the search for treatments with a higher oxidative capacity avoiding the formation of harmful byproducts. In this sense, it has been proposed the use of the so-called advanced oxidation processes (AOP) for the elimination of recalcitrant compounds and disinfection of effluents [3,14,15]. Nevertheless, there are differences between bacterial inactivation and decolorization of recalcitrant organic compounds by AOP [16]. Bacteria with the ability to self-repair and grow again after damage are much more complex than recalcitrant compounds [17].
A common point of the vast majority of AOP is the formation of hydroxyl radicals (·OH). Furthermore, ·OH is considered one of the species with the greatest oxidizing power. For example, chlorinated compounds used in conventional effluent treatments such as Cl 2 and ClO 2 have standard reduction potentials of 1.36 and 1.27 V/SHE respectively, while ·OH has a standard potential of 2.8 V/SHE [18].
Among the AOP, the Fenton technologies has focused a lot of attention for many years [19]. This process involves the reaction of H 2 O 2 as an oxidant agent with Fe 2+ ions as a metal catalyst to produce the degradation agent of ·OH as illustrated in Equation (1). Fe 3+ produced by the Fenton reaction can also oxidize H 2 O 2 to produce perhydroxyl radicals (HO 2 ·; Equation (2)), named the Fenton-like reaction. The ·OH and HO 2 · produced in Fenton and Fenton-like reactions can participate in parallel reactions to produce singlet oxygen ( 1 O 2 ; Equations (3) and (4)).
Andreozzi, et al. [20] highlight the reactivity of ·OH, since this species has adequate properties to attack organic compounds, in addition to reacting 10 6 -10 12 times faster than other oxidants. Additionally, the cellular damage produced by ·OH in the disinfection processes takes place on different macromolecules present in the bacterial membrane causing its inactivation [21].
Many of the parameters that can affect any type of chemical reaction could affect the Fenton reaction, among which the effects of pH and reagent concentration stand out. The pH is one of the most important parameters in the Fenton reaction. However, it is possible to find that the optimum pH varies. One of the reasons why it is possible to find this variety at the optimum pH of the Fenton reaction may be associated with the speciation of Fe 2+ and Fe 3+ [22], changes in redox potentials of the main oxidizing species produced [23], or changes in the type of oxidizing species produced depending on the pH [24][25][26].
Without the presence of Fe 2+ in the Fenton system there is no formation of ·OH, so the presence of Fe 2+ is essential. However, it has been studied that too high Fe 2+ concentrations can cause the Fenton reaction oxidizing capacity to decrease (Equation (5)) [27].
The H 2 O 2 concentration, like Fe 2+ , is also essential in Fenton systems [28]. However, an excess of H 2 O 2 could act as a scavenger of ·OH [27,28] according to the Equation (6).
Accordingly, to minimize Fe 2+ and H 2 O 2 acting as scavengers, but maximizing the production of oxidizing species from these reagents, it is very important to know the optimal [H 2 O 2 ]/[Fe 2+ ] [29].
Therefore, the aim of this work is to evaluate the decolorization of a model recalcitrant compound (methylene blue) and the inactivation of a model bacteria (   Using Design Expert software (version 10), experimental data in Table 1 were analyzed by a second-order linear polynomial regression model (Equation (7)). η = γ 0 + γ 1 A + γ 2 B + γ 3 C + γ 12 AB + γ 13 AC + γ 23 BC + γ 11 A 2 + γ 22 B 2 + γ 33 C 2 (7) in which η is the dependent factor (response), γ 0 is the intercept; Analysis of variances (ANOVAs) and significant test results for the quadratic regression equations are shown in Table 2.  Table 2 listed the results of variance analysis for the MB decolorization and E. coli K12 removal using the Fenton process. The values of the sum of squares demonstrate the contribution of independent variables on responses [30]. The mean squares, which are the sums of squares divided by the degree of freedom. Adequacy of the model parameters in the present study for response variables (D % MB , k app MB and I uLog EC ) was determined by the Fisher value (F-value), obtained by dividing the mean squares of each effect by the mean squares of error [31]. The probability critical level (p-value) of 0.05 was considered to reflect the statistical significance of the parameters of the proposed model. The F-values > 0.001 (975.81, 6.47 and 62.10) and p-values < 0.05 obtained for D % MB , k app MB and I uLog EC responses confirming the qualification of the model to predict the decolorization of MB (D % MB and k app MB ) and the inactivation of E. coli K12 (I uLog EC ) by the Fenton reaction. In addition, the validity of the model is confirmed by the p-value of the lack of fit with values greater than the lowest limit of fit as recommended (>0.05) [32]. As a result, the models developed in this work for predicting the D % MB , k app MB and I uLog EC by the Fenton process were considered adequate. These models can be described as shown in Table 3 with coded three factors. Table 3. Statistical results of the proposed models in terms of the coded factors.

Models and Regression Analysis
The ANOVA results of three parameters (D % MB , k app MB and I uLog EC ) showed that the significant (p < 0.05) response surface models with high R 2 value (0.9210-0.9994) were obtained as shown in Table 3, ensuring a satisfactory adjustment of the quadratic models to the experimental data. The R adj 2 values (0.7787-0.9984) obtained suggests that the three proposed models had an adequate predictive capacity. Even more, plots comparing the experimental and predicted values for D % MB , k app MB and I uLog EC indicated a good agreement between experimental and predicted data from the model ( Figure 1). Therefore, this finding indicates high correlation and adequacy of the proposed model to predict performance of the Fenton process (D % MB , k app MB and I uLog EC ).     The 3D surface and contour plots in Figure S1 show      The perturbation plots (Figure 4b) Figure S3 show

Analysis of Optimization and Model Validation
The optimal conditions obtained for MB decolorization and E. coli K12 inactivation are different for each of the responses studied. These results indicate that although some authors have suggested that it is possible to analyze the bacteria inactivation of AOP by extrapolating from dye decolorization [42], these processes have differences. The results in the present study (Table 1) [43].
To validate the model obtained by the Box-Behnken optimization technique, experiments were carried out with the suggested optimum values of independent variables. Table 5 shows the optimal conditions predicted by the models, the predicted response value and the response value obtained experimentally (Table 5). value (studied at 6.0 × 10 −4 , 8.0 × 10 −4 and 1.0 × 10 −3 mol/L) exhibited a significant positive effect on these responses, while for the I uLog EC analysis this parameter showed a significant negative effect.

Effect of pH on Responses
The pH value (studied at pH 3.0, 4.0 and 5.0) showed a significant effect on the D % MB and I uLog EC responses but did not show a significant effect on the k app MB response. The Fenton reaction (Fe 2+ and H 2 O 2 ), with a rate constant 76 L·mol −1 s −1 [48], form ·OH quickly, consume Fe 2+ and produce Fe 3+ . The Fenton-like reaction (Fe 3+ and H 2 O 2 ) has a much slower rate constant (0.01 L·mol −1 s −1 ) than the Fenton reaction [48] and it only produces O 2 · − , a much less reactive radical than ·OH. Additionally it has been established that the species of Fe(II) that prevails in the working pH range (3.0-5.0) is Fe 2+ [49], while in the same pH range the speciation of Fe(III) demonstrates the formation of Fe(OH) 2+ and Fe(OH) 2 + species, species that are less reactive than Fe 3+ [22]. Based on this information, it is expected that the rate of ·OH formation from the Fenton reaction, at least in the first minutes of reaction that directly influence the determination of k app MB , will not be greatly altered when changing system pH between 3.0 and 5.0. However, D % MB and I uLog EC , which are obtained in a final time of 15 min, will be influenced by both the Fenton reaction and the subsequent Fenton-like reaction. Considering this, the participation of the Fenton-like reaction implies that the pH and its effect on Fe(III) speciation have a greater influence on the D % MB and I uLog EC responses, as observed in this investigation.

Fenton Experiments
Methylene blue, a dye that does not generate toxic byproducts when reacting with ·OH [50][51][52], was used as a model of recalcitrant compound, while E. coli K12, a non-pathogenic E. coli [53], was used as a model of bacteria. Experiments were performed in 20 mL glass reactors containing the MB solution (5.0 × 10 −5 mol/L) or E. coli K12 (10 6 CFU), kept under magnetic stirring at room temperature (25 • C) [43,44]. First, FeSO 4 ·7H 2 O solution was added to each sample according to the experimental design. The pH of each sample was adjusted by using NaOH (0.25 mol/L) or HCl (0.10 mol/L) solutions. Reactions were started by adding an aliquot of H 2 O 2 solution. After the experimental time elapsed (15 min), for E. coli K12 analysis, 0.2 mL of each sample was collected for its enumeration. The decolorization of MB was studied by determining its kinetic constants of color decay and the degree of decolorization. After 2 min of maintaining the reaction under constant agitation, samples (3.0 mL) were withdrawn, and immediately injected into a cuvette for analysis at time intervals of 3, 6, 9, 12 and 15 min. The analyses in samples were performed spectrophotometrically by UV-Vis spectrophotometry (Shimadzu UV-1800, Shimadzu Inc., Kyoto, Japan) at 668 nm using quartz cells with path lengths of 1 cm. A calibration curve was constructed (5.30 × 10 −7 -1.30 × 10 −5 mol/L; R 2 = 0.999). Fitting decolorization kinetics and the rate constant was obtained by Sigma Plot 11.0 software (Systat Software, Inc., San Jose, CA, USA).

Detection and Enumeration of E. coli K12
Strain samples were stored in cryo-vials containing 20% glycerol at −20 • C. To prepare the bacterial pellet for the experiments, one colony was picked from the precultures and loop-inoculated into a 50 mL sterile PE eppendorf flask containing the Luria Bertani (LB) medium. The flask was then incubated aerobically at 37 • C and 150 rpm in a shaker incubator (Gerhardt THO500, Gerhardt GmbH & Co., Königswinter, Germany) until the stationary physiological phase was reached. After 24 h, cells were centrifuged (SIGMA 2-16P, Sigma Laborzentrifugen GmbH, Steinheim, Germany) and diluted until optical density 0.5 a.u. (i.e., 10 6 CFU/mL) at 600 nm [43,44]. Component of LB medium included sodium chloride (10 g), tryptone (10 g) and yeast extract (5 g) in 1 L of deionized water; this solution was then sterilized by autoclaving for 20 min at 121 • C. The bacterial pellet was resuspended and washed three times with a saline solution (NaCl/KCl). The final pellet was resuspended in saline solution. This procedure resulted in a cell density of approximately 10 9 colony forming units (CFU) per milliliter. The pH of the solution was adjusted to 7.0 and the solution was then sterilized by autoclaving for 30 min at 121 • C. The bacterial solution was diluted in reactors to the required cell density corresponding to 10 6 CFU/mL [43,44].
CFU were performed by plating on plates (PCA method). Of the samples 0.2 mL was withdrawn. Samples were diluted (10% v/v) and 0.1 mL poured on plates. Plates were aerobic incubated for 24 h at 37 • C (Heraeus B6, Kendro, Langenselbold, Germany) and the CFU were counted manually. All experiments were performed in triplicates. The enumeration of colonies was expressed as CFU (colony forming units) per 100 mL of sample. These concentrations were transformed to log 10 and the removal of bacteria, uLog = log(N t /N 0 ), was calculated from the initial bacteria concentration (N 0 ) and the remaining bacteria population at "t" time (N t ).

Experimental Design
To determine the optimal experimental conditions for the decolorization of MB and the inactivation of E. coli K12 by Fenton technology, a Box-Behnken design was performed. pH, Fe 2+ concentration ([Fe 2+ ], mol/L) and molar concentration ratio of Fe 2+ and H 2 O 2 ([H 2 O 2 ]/[Fe 2+ ]) were selected as independent variables in the experimental design (Table 6). Three replicates were performed at the central point, with 15 runs performed for each study. The chosen levels of the independent variables were based on literature reports [54]. The experimental responses were the degree of MB decolorization (D % MB ), the apparent kinetic constant of MB decolorization (k app MB ), and removal of bacteria in uLog units (I uLog EC ) for variables showed in Table 6. A second-order linear polynomial regression model (Equation (7)) was obtained to analyze the data. Data were statistically evaluated and an analysis of variance (ANOVA) was applied at with a confidence level of 95% using software Design Expert version 10 (Stat-Ease Inc., Minneapolis, MN, USA). Responses of the experimental tests were compared to the estimated values, and the fit of model was assessed. Experimental tests, performed under optimal conditions, were performed to achieve maximal D % MB , k app MB and I uLog EC .

Conclusions
The present study provided a comprehensive description regarding the application of the Fenton technology as a process for MB decolorization and E. coli  suggest that different oxidizing species are involved in these processes. Thus, considering that bacteria are larger than dye molecules, the complex self-repair mechanisms of bacteria and the different external structures of bacteria compared to the dyes structure, the E. coli inactivation proved to be less effective than MB decolorization by Fenton processes.