Humic acid degradation by fenton-like process using Fe²⁺ and Mn⁴⁺

In recent years, there has been ever increasing interest concerning information on humic substances (HSs) found in surface water and water supply. These substances are products of natural processes of biological degradation (Stevenson 1994). The interest in researching HSs stems from the fact that their presence can cause various problems, such as the color, odor and taste in the water supply, depending on the compounds present. In addition to organoleptic alterations, the possibility of byproduct formation, even at the lowest concentration could be an additional risk to human health (Goel et al. 1995; Wu et al. 2010a; Kim et al. 2013).

Advanced oxidation processes (AOPs) are characterized by transforming organic contaminants into carbon dioxide, water and inorganic anions, mainly transitory species such as hydroxyl radicals (OH^•), which have a high oxidation-reduction potential. These are non-selective processes, have short half-lives and operate under normal pressures and temperatures. The high standard potential of hydroxyl radical reduction (E_o = 2.7 V in acid solution and 1.8 V in neutral solution), lower only than the fluorine, makes the systems which generates these radicals very efficient in the oxidation of organic contaminants in an aqueous environment (Buxton et al. 1988; Tchobanoglous et al. 2003; Loures et al. 2013).

Based on what was presented, the objective of this study was to optimize the Fenton-Like process using Fe²⁺ and Mn⁴⁺ for degradation of Humic Acid by Central Composite Design (CCD) and to develop a predictive model for degradation involving five independent factors. The effects of the initial pH, [H₂O₂], [Fe²⁺], [Mn⁴⁺] and reaction time were also investigated.

All compounds were of reagent grade. The hydrogen peroxide (H₂O₂) solutions were prepared using a solution of H₂O₂ (30% w/w), [H₂O₂] = 9.7 mol L⁻¹ and density of 1.1 kg L⁻¹. The source of manganese (Mn⁴⁺) added as a catalyst was manganese oxide (MnO₂), and the source of iron (Fe²⁺) was ferrous sulfate heptahydrate (Fe₂SO₄.7H₂O).

The stock solution of Humic Acid (Sigma Aldrich^®) was prepared by diluting 14.7 mmol of HA in a flask of 0.1 L increased volume with sodium hydroxide 0.1 mol L⁻¹, and was agitated for a period of 48 hours, resulting in a stock solution with 147 mmol L⁻¹ of HA (Zouboulis et al. 2004).

There are still no direct methods for determining the HA concentration (Vogel et al. 1999; Mertig et al. 2002; Rodrigues et al. 2009), although in surface water, they can be determined indirectly by ultraviolet light absorbance at 254 nm (Korshin et al. 1997; Wang & Hsieh 2001; Rodrigues et al. 2006; USEPA 2009; APHA et al. 2012). The method used in this experiment was the 5910-B Standard Methods (APHA et al. 2012).

The statistical planning chosen for developing this experiment was the CCD. By using the CCD, response surfaces can be generated, which help develop, improve and optimize processes. They are able to generate enough information to obtain favorable responses to the experiment through dimensional surfaces. CCD is widely used when there is interest in optimizing a process, where there is an influence of several factors in a response variable.

The dependent variable was the HA relative concentration, and the independent variables evaluated the effect of dosage of H₂O₂, the catalysts, the reaction time and the initial pH of the solution to be treated. The direct response was the reading of the HA concentration, converted into HA relative concentration (HA/HA_o), thus leading to making inferences regarding the effect of the independent variables.

The low and high levels (−1 and +1) of the independent factors were the following: 4 and 6 for initial pH, 2.65 and 6.17 mmol L⁻¹ for [H₂O₂], 0.27 and 0.81 mmol L⁻¹ for [Fe²⁺]_, 0.27 and 0.82 mmol L⁻¹ for [Mn⁴⁺] and 60 and 120 minutes for the reaction time, as explained in Table 1. Thus, 50 assays were performed: 32 points of the factorial (2⁵) and 8 central points to estimate the experimental error and 10 axial points to verify the response curvature.

The suitability of the proposed models was evaluated by the analysis of residue, classified as lack of fit and pure error and analyzing the proportion of the variation explained by the model, that is, by the analysis of the coefficient of determination (R²). Once a response-adjusted model was obtained, its optimization was done by defining a restricted desirability function at the interval of 0–1, adopting for the relative concentration of HA 1.00 for the lower limit, 0.50 for the mean value and 0.01 for the upper limit of desirability. The analysis and graphs were made using the trial software Statistica 10.0 (StatSoft 2011).

The initial HA concentration adopted for the assays (carried out under an environmental condition, = 25.6 ± 1.5 °C), after carrying out the literature review (Motheo & Pinhedo 2000), was 0.882 mmol L⁻¹ (30 mg L⁻¹) which, according to Brum & Oliveira (2007), is generally the concentration found in surface water. To prepare the assays, 3 mL of the HA stock solution was pipetted into volumetric flasks of 500 mL, and were then completed with deionized water.

After diluting the stock solution of HA to obtain the initial concentration of the assays, the pH of the sample was corrected, according to the experimental arrangement suggested, using 0.1 mol L⁻¹ of sulfuric acid (H₂SO₄), when a reduction of is needed pH, and sodium hydroxide (NaOH) 0.1 mol L⁻¹ when a pH increase is required.

After the reagent dosages ([Fe²⁺], [Mn⁴⁺] and [H₂O₂]) were carried out, the assays were subjected to the reaction times cited, under agitation of 100 rpm in the Jar test equipment (Kang et al. 2002; Mortazavi et al. 2005; Oliveira et al. 2007; Khalid et al. 2011). After the expected agitation time, sodium sulfite in a stoichiometric ratio of 1.1:1 (3.71 × 1.1 = 4.08 mol of Na₂SO₃ by 1 mol de H₂O₂) was added and it was left for 5 more minutes reacting to completely quench the Fenton reaction (Liu et al. 2003). After that, the membrane sample of 0.45°μm was filtered and the absorbance was read in a spectrometer, at the wavelength of 254 nm.

Table 2 shows the HA relative concentration and the respective treatments suggested by the design and the predicted HA relative concentration using Fe₂SO₄.7H₂O (Fe²⁺) as a source of iron and MnO₂ (Mn⁴⁺) as a source of manganese in the Fenton-Like process.

Table 3 shows the ANOVA for the relative concentration of HA after carrying out the experiments. The complete regression model is significant (p < 0.001) and can be used to predict the data. Despite the lack of fit being significant, it can be considered apparent, because as the responses at the central points were very close and the average square of pure error very low (reasons for F of lack of fit very high), the model can be considered valid for predictive purposes (Souza & Menezes 2008). The coefficient of determination of the model (R²) showed that 84.95% of the response variation was explained by the estimated function.

To make it easier to visualize the variables that had a significant effect (p≤ 0.05), the Pareto chart can be seen in Figure 1, which shows the effects that are statistically important within the independent variables investigated. Those that exceeded the line of significance are considered significant, whereas the factors that did not exceed the line are not significant, and therefore did not influence the HA degradation by the Fenton-Like process. The value next to the bar represents the values of the statistics of test t and the size of the bar is proportional to the estimated effect that the variable in question has on the response variable.

It can be observed (Table 2) that assays 1, 2, 3, 4, 9, 11, 17, 19, 20, 25, 27 and 28 were not efficient and were those that used low [Fe²⁺] (0.27 mmol L⁻¹). The only assays under this low concentration of iron added, in which some efficiency was obtained (between 72.6 and 86.8%) were the assays 10, 12, and 26, which had a high concentration of H₂O₂ (6.17 mmol L⁻¹) and remained under a higher reaction time (120 min).

Among the treatments which presented efficiency higher than 95%, it can be observed that the pH was between 4 and 6, [H₂O₂] between 0.88 and 2.65 mmol L⁻¹, [Fe²⁺] between 0.54 and 1.07 mmol L⁻¹, [Mn⁴⁺] between 0 and 0.82 mmol L⁻¹ and the reaction time between 30 and 120 min. In those assays, the molar ratio [H₂O₂]:[Fe²⁺] = 3.3, [H₂O₂]:[Mn⁴⁺] = 9.8 and [Fe²⁺]:[Mn⁴⁺] = 3.0 presented the greatest efficiencies.

Figure 1 shows the effect of the tested variables, as well as their significance (p≤ 0.05). The variables that were significant in the HA degradation were [Fe²⁺] (linear effect and square), pH (square effect) and the interactions between [Fe²⁺] vs [H₂O₂] and [Fe²⁺] vs reaction time.

In Figures 2(a) and 2(b), the contour plot for the HA/HA_o under the effect of the interaction of the independent variables is presented, which had a significant effect. It was observed (Figure 2(a)) that in [Fe²⁺] of 0.54 mmol L⁻¹ and [H₂O₂] of 4.41 mmol L⁻¹, it is estimated to obtain degradation of practically all HA. The same happens at the concentration of 0.81 mmol L⁻¹ for Fe²⁺ and concentrations of 2.65 mmol L⁻¹ and 6.17 mmol L⁻¹ for H₂O₂.

Based on the interaction of [Fe²⁺] with reaction time (Figure 2(b)), it is possible to estimate that for [Fe²⁺] of 0.54 mmol L⁻¹, in a reaction time of 90 minutes the maximum efficiencies of degradation (HA/HA_o ≤ 0.10) will be obtained. Although for shorter reaction times in higher [Fe²⁺] high HA degradation can also be observed.

In the Fenton-Like process, the initial pH has a significant importance in the oxidation of organic compounds as these systems are based on the Fe²⁺ speciation used in the catalysis of the reaction. The acid pH is paramount for the Fenton reaction, as the presence of H⁺ ions favor the decomposition of H₂O₂ and shifts the equilibrium of the reaction for HO⁻ and HO^• generation.

A higher HA degradation was observed when pH was more acid, although it was also observed that, at pH 6, there were also lower HA/HA_o (0.10–0.03), in cases where there was greater availability of Fe²⁺ for the reaction. This fact corroborates the idea of some authors (Barbusiński 2009; Lee & Sedlak 2009) of not only obtaining considerable degradations in more acid pHs.

The square effect of the pH variable, adjusted to the model, decreasing the HA/HA_o of pH 3–5 and increasing from pH 5–7 was also observed by other authors. Wu et al. (2010a), analyzing the degradation of HSs of the landfill leachate, using the Fenton process with a molar ratio [H₂O₂]:[Fe²⁺] of 6:1 and reaction time of 120 min, found that the degradation efficiency (≅60%, pH 1) was higher according to the increase in the initial pH of the reactive medium until pH 4 (≅85%) and returned to decrease until pH 8, when it reached the maximum value close to 17%. The same authors observed that under reaction times similar to the ones tested in this experiment, the maximum efficiency reached was 85.2%, which is lower than the ones obtained in this study. It is believed that this lower efficiency may be related to the high initial concentration of HSs (around 3980 mg L⁻¹) presented in the leachate.

There was a negative effect of the increase in the concentration of H₂O₂ at the HA degradation (an example is assay 36 – Table 2) because in this condition the peroxide starts to change the hydroxyl radical into another less reactive radical (production of the radical peroxide).

As observed in this experiment and also in the literature (Park et al. 2006), [H₂O₂]:[Fe²⁺], to achieve better efficiencies, the molar ratio should be 3:1. The molar ratio of [H₂O₂]:[Fe²⁺] of 6:1, observed in the experiment performed by Wu et al. (2010a), may also be another factor which helps to explain this lower efficiency that the authors observed, as the excess of oxidant acts as a scavenger or quencher, transforming the free radical into another less reactive one, or even, completely neutralizing the reaction.

In an experiment carried out by the same authors mentioned above (Wu et al. 2010a), behavior similar to the one obtained in this experiment regarding the elevation of [H₂O₂] was observed. As the [H₂O₂] rises, there is an increase in the efficiency of the HA degradation, up to a certain value and after that it becomes constant or with a slight tendency to decrease (within the concentrations analyzed).

In high [Fe²⁺], the rate of HA degradation increased until reaching a maximum value, and from this point on there were no more significant changes. Supposedly the increase of [Fe²⁺] accelerated the generation of hydroxyl radicals, which brought about an increase in the HA degradation rate, however the increase of [Fe²⁺] in the solution increased the HA degradation up to a determined concentration. Very high concentrations of Fe²⁺ did not prove what was expected (higher HA degradation), and may even lead to a decrease in the efficiency. The excess of Fe²⁺ ions may have been harmful to the efficiency of the process, as it can generate competition between them, by the hydroxyl radicals and the degraded organic compound (HA), also reported by Souza et al. (2013).

Considering the Pareto chart in the effects tested (Figure 1), we can clearly see the aforementioned statement, because if we consider the linear effect of [Fe²⁺] at the HA/HA_o, there is a significant effect. The higher the [Fe²⁺], the lower the HA/HA_o, however there was also a significant effect for the square effect of [Fe²⁺], which means that due to the coefficient of the regressor being positive, the concavity of the parabola is turned upwards, which shows that by increasing [Fe²⁺], there is a lower HA relative concentration (higher efficiency), but from a certain [Fe²⁺], the HA degradation decreases. Similar behavior was observed in other experiments (Wu et al. 2010a; Jung et al. 2013).

It was observed (Figure 1) that [Mn⁴⁺], although not showing a significant effect, tends to follow linear and square behavior regarding the efficiency of HA degradation. At the data interval of the experimental design, when [Mn⁴⁺] was increased, smaller HA degradation was obtained. This shows that the use of Mn⁴⁺ as a catalyst of the Fenton reaction was not a good alternative as an additive in the Fenton process in the conditions tested. Deguillaume et al. (2004) showed that the H₂O₂ + Mn⁴⁺reaction would generate as a product Mn²⁺ + 2H⁺ + O₂ and, according to Huang et al. (2009), Mn²⁺ in acid medium could react with H₂O₂ resulting in Mn³⁺ + OH^• + OH⁻, however, apparently, this reaction was not observed in this experiment.

The reaction time was directly related to the concentration of iron (Figure 2(b)). By analyzing the global averages of HA degradation, the longer the reaction time, the greater the efficiencies, and these results are within the expected, which is corroborated by other authors, such as Kavitha & Palanivelu (2004), Babuponnusami & Muthukumar (2012) and Wang et al. (2013).

The addition of Mn⁴⁺ (MnO₂) to the Fenton process was also not susceptible to success, which can be verified both by the linear and square effects, and the higher the concentration of Mn⁴⁺, the lower the process degradation efficiency.

In the Fenton-Like process in which the significance of the regression model was obtained, the analysis of the optimization conditions of the tested variables can be performed, obtaining the optimal values for these variables, through the desirability function, as shown in Figure 3.

In Figure 3, the five first profiles are related to the variation of the predictive response of each factor, analyzing each factor individually, keeping a level fixed within a factor and varying the other factors and levels. The sixth upper profile shows the range of the desirability response (0 < D < 1). The greater the value of D, the more convenient the system response is, that is, the maximum value of D is the optimal condition of the system. In the five lower profiles, the individual desirability for each factor can be seen, as well as the global desirability equal to 1, which is to obtain the optimization. The vertical dotted lines in the graphs correspond to the optimal values of the studied parameters. For this case study, it can be observed that in the absence of [Fe²⁺], in the other variables there is not only a value for the maximum desirability condition, thus it is possible to obtain better degradation efficiencies with pH in 4 and 5, [H₂O₂] between 6.17 and 7.94 mmol L⁻¹, [Fe²⁺] in 0.27 mmol L⁻¹, [Mn⁴⁺] between 0 and 0.54 mmol L⁻¹ and t between 120 and 150 min.

The most economical condition, considering a complementary step to a water treatment, would be pH equal to 5, where there would be no or less need for acidification of the effluent; [H₂O₂], [Fe²⁺] and [Mn⁴⁺] in 6.17, 0.27 and 0 mmol L⁻¹, for lower input expenditure and t at 120 for a shorter operating time.

The variables that had a significant effect on the efficiency of HA degradation by the Fenton-Like process were by order of influence: [Fe²⁺] linear and square effect, pH square effect, interaction [H₂O₂] vs [Fe²⁺] and interaction [Fe²⁺] vs t.

The Fenton-Like process modified by adding Mn⁴⁺ (MnO₂) was not a good alternative for HA degradation, whereas, using only Fe²⁺ (Fe₂SO₄.7H₂O), it can be considered good for HA degradation.

It is estimated that by operating the Fenton-Like process with pH in 4 and 5, [H₂O₂] in 6.17 and 7.94 mmol L⁻¹, [Fe²⁺] in 0.27 mmol L⁻¹, [Mn⁴⁺] between 0 and 0.54 mmol L⁻¹ and t in 120 and 150 min, better efficiencies in the HA degradation can be obtained.

The most economical condition would be pH 5, [H₂O₂] 6.17 mmol L⁻¹, [Fe²⁺] 0.27 mmol L⁻¹, [Mn⁴⁺] 0 mmol L⁻¹ and a reaction time of 120 min.

We would like to thank the Fundação de Amparo à Pesquisa do Estado de Goiás (FAPEG) for the funding of this project (FAPEG-Universal n° 05 2012), the Federal University of Goiás for the institutional PhD scholarship awarded to the first author (Edital PRPPG/PROAD/PRODIRH n° 13, 2012), the Agricultural Engineer, Marta Cristina Carvalho, the Chemist, Ana Paula Aparecida Borges for the support during the experimental procedure and the company specialized in academic translation, AudioText, for translating this manuscript.