2.1 Reagents

All reagents were of analytical grade: NaCl (99%), yellow 5 colorant (98%), FeSO₄·7H₂O (99%), H₂O₂ (30%), and H₂SO₄ (98%), furnished by Merck, and used without any further purification. The corresponding reagent solutions were prepared using extra pure water (Milli-Q system; 18.0 MΩ·cm resistivity). All studies were performed with simulated yellow 5 wastewater since the received industrial samples were mixtures of different dyes, between them yellow 5 one, with variable concentration depending on the specific industrial activities. Thus, simulated wastewater samples of yellow 5, with concentrations in the range from 200 to 1000 mg/L, were prepared by diluting the corresponding amount of the dye with deionized water. This range of concentration was selected according to the monitoring of the enterprise activity performed during one month.

2.2 Photo-reactor

A schema of the photo-Fenton process is depicted in Fig. 1. The experiments were performed at 25 °C, as described previously in details [8,21]. For each experiment, a desired quantity of reactive solution was freshly prepared. Additionally, the required amounts of FeSO₄·7H₂O and H₂O₂ were added to the obtained solution. Its pH = 3 was fixed using 1 N H₂SO₄. This pH value is highly recommended for Fenton's reaction since it avoids the formation and precipitation of Fe²⁺ salts. At the beginning of each experiment, the dye solution was fed into the reactor. Next, Fe²⁺ and H₂O₂ were charged. The UV lamp (365 nm F6T5 Black Light Hg and 4 W power) was switched on when the 1-L reactor was completely filled and the time started to be monitored. Each time, 5-mL samples were taken at periodic intervals for immediate analysis.

2.3 Analytical methods

In order to follow the efficiency of yellow 5 (E102) decolorization (%DD) and degradation (%COD), the samples were withdrawn at certain time intervals for spectrophotometric analysis (vis double-beam spectrophotometer, Spectronic Genesys 2PC), in the range from 200 to 700 nm. A calibration plot based on Beer–Lambert's law was prepared to correlate absorbance magnitude with dye concentration. The absorbance, measured at 430 nm, corresponds to the dyes’ color. Its changes were used to monitor the decolorization efficiency. The changes in the absorbance measured at 246 nm, representing the aromatic part of yellow 5 (E102), indicated the degradation of the aromatic part of the dye. Standard methods [25] were used for the quantitative analysis of the chemical oxygen demand (COD), the total organic carbon (TOC). The COD analyses were performed following the closed reflux method with colorimetric measurements (method 5220D). Mineralization was monitored by changes in the total organic carbon (TOC) content. TOC measurements were carried out with a Shimadzu analyzer (model TOC-5000A), following the method 5310D. The BOD₅ measurements were performed following the respirometric method (5210B).

The efficiency of the photo-Fenton process was calculated as the dye, COD, and TOC degradation percentages (%DD, %DCOD, and %DTOC, respectively). For example, the %DD was calculated using the following expression:

2.4 Experimental design and statistical analysis

The RSM was chosen and implemented to establish the influence of different operating factors on the decolorization and mineralization of yellow 5 by the photo-Fenton process. A multifactorial BBD was defined to evaluate the interactive effect of process variables and to perform its optimization. The following variables were selected for RSM: H₂O₂ concentration, initial dyestuff concentration, and UV-radiation power (number of lamps). The RSM coupled with BBD was chosen to find the relationship between the response functions (%DD and %COD) and variables using the statistical software tool Statgraphics 5.1 (Statistical Graphics Corp 1999–2004). The three-level second-order design demanded a relatively low number of experiments. Thus, fifteen tests with replica including three center points were randomly made in order to avoid any systematic error. From preliminary experiments (not presented here), three different levels (values) were chosen for each of the four following variables (Table 1). The interaction between the variables and the analysis of variance (ANOVA) has been studied by RSM. The quality of the fit of this model and its prediction capacity is expressed by the coefficient of determination, R². The variables were coded according to Eq. (2):

3.1 Development of the regression model equation

A set of preliminary experiments (not shown here) allowed defining the variables that considerably affect decolorization, %DD, and mineralization, %DCOD, efficiency of yellow 5 by the photo-Fenton process. Thus, dyestuff concentration, H₂O₂ concentration and UV radiation, at 365 nm (number of lamps), were chosen for optimization using RSM and BBD. To examine the combined effect of three independent process parameters on its efficiency, 15 experiments were performed. Furthermore, Fe²⁺ concentration (1 mM), pH (of 3) and UV radiation (365 nm) were kept constant. The experimental design is given in Table 2, together with the corresponding experimental data. As shown in Table 2, the removal efficiencies range from 91% to 100% for %DD and 4–75% for %DCOD.

Regression analysis was performed to fit the response function (%DCOD). The developed second-order polynomial equation represents responses as functions of dyestuff concentration (A), number of lamps (B), and H₂O₂ concentration (C). An empirical relationship between the response and the input test variables in coded units can be expressed by the following equation:

Equation (4) describes how %DCOD was affected by the individual variables and/or their double interactions. %DCOD was linear and also quadratic with respect to dyestuff concentration, number of lamps, and H₂O₂ concentration; and it is also quadratic with respect to dyestuff concentration, number of lamps, and H₂O₂ concentration. This indicates that there are the following interactions: dyestuff concentration–dyestuff concentration, dyestuff concentration–number of lamps, dyestuff concentration–H₂O₂ concentration, number of lamps–H₂O₂ concentration, which can affect %DCOD.

3.2 Statistical analysis of photodegradation tests (ANOVA)

For all experiments, a reaction time of three hours was established and %DCOD was chosen as the response variable. Analysis of variance, ANOVA, was employed to determine the significant main and interaction effects of factors influencing %DCOD. The ANOVA results are presented in Table 3.

ANOVA consists in classifying and cross-classifying statistical results. The Fisher F-test, based on the ratio of the respective mean-square effect to the mean-square error, was used to evaluate the presence of a significant difference from the control's response and to calculate standard errors. The bigger the magnitude of the F value, more significant is the corresponding coefficient. The P values were used to identify experimental parameters that present a statistically significant influence on a particular response. If the P value is lower than 0.05, it is statistically significant at the 95% confidence level [8]. According to ANOVA results, one can see that all terms in the regression model are not equally important. Only two of them (initial dye concentration [A:Ci] and A–A interactions) presented P values lower than 0.05 (Table 2), which implies that they have a truthfully effect on %DCOD, with a confidence interval of 95% [8,20]. Similar results were presented elsewhere [21,26]. On the other hand, the effect of H₂O₂ concentration on %DCOD removal efficiency is not as significant (in the analyzed range). However, it can present some importance and should not be ignored.

The quality of the developed model was evaluated based on the variation coefficient (R²) and the standard deviation value. The closer the R² value to unity and the lower the value of the standard deviation, more accurately the response can be predicted by the model. The R² value was found to be 0.9878, indicating that 98.78% of the total variation in %DCOD was attributed to the studied experimental variables. Moreover, the value of the predicted R² (0.9878) was in very good agreement with the value of the adjusted R² (0.9659). Since the R² value is close to unity, there is a good agreement between the experimental data and those predicted by the model. The highest %DCOD (99) and %DD (100) were reached for the following experimental conditions: Fe²⁺ concentration = 1.00 mM; H₂O₂ concentration = 1.75 mL/L; one lamp UV radiation power (wavelength = 365 nm, 4 W); and initial dye concentration = 200 mg/L. The probability plots are one of the most widely used analysis tools to identify data reliability. Therefore, Fig. 2, a normal probability plot, presents the effect of each variable on %DCOD during the photodegradation process. Again, one can see that the initial dye concentration presents the most significant effect on %DCOD.

The Pareto analysis was used to identify factors that present the greatest cumulative effect on %DCOD, and thus to screen out the less significant ones. A Pareto diagram is a series of bars whose heights reflect the frequency or impact of each factor. The bars are arranged in descending order of heights, from left to right. Therefore, the factors represented by the tall bars are relatively more significant. Here, the Pareto analysis was also carried out to determine the percentage effect of each factor, according to the following equation [19,20]:

Thus, statistically important factors correspond to all these whose values overpass the inner vertical line (Fig. 3). The vertical line corresponds to the t value in the t-Student distribution, with a 95% confidence and for 14 degrees of freedom. Next, this value is compared with the values of each effect and interaction of the analyzed factor. The comparison defines the statistical significance of each factor in the analyzed process. Therefore, the factor that presents an influence on the degradation process is the initial dye concentration (A) and its interactions: A–A.

One can see that %DCOD is inversely proportional (–) to the initial dye concentration (A) and directly proportional (+) to the A–A interaction. Moreover, the initial dye concentration can be considered as the most dominating factor during COD degradation (Fig. 3). In fact, the higher the dye concentration is, the lower the degradation rate that can be reached (considering also that the ^•OH radical concentration was constant). Similar results were reported by Modirshahla et al. [13].

The surface response plot was used to study the effect of all factors on the response variable (Fig. 4). This type of plots summarizes the effect of two variables on %DCOD, considering that the third one was kept constant. One can see that %DCOD depends on the initial dye concentration (with a maximum at 200 ppm). Moreover, no significant effect of the UV-radiation power (number of lamps) on %DCOD was observed. Modirshala et al. [13] and Gupta et al. [27] reported contradictory results mainly due to differences in the reactor's configuration.

Fig. 5 shows the representative UV-vis spectra of the E102 degradation process 1.0 mM Fe²⁺ concentration. One can see that before treatment (t = 0 min), its UV-vis spectrum consists of two main characteristic absorption bands: the first one, in the UV region, with a maximum at 246 nm, corresponds to the benzenic rings of E102; and the second one, in the visible region, with a maximum at 430 nm, is a feature of the two-azo group (–N = N–), which is responsible for the bright lemon yellow dye. After 15 min of the degradation process, a significant decrease in the intensity of the absorption band in the UV region was observed. Moreover, the absorption band observed in the visible region disappears and, in its place, two new bands, at 220 and 330 nm, appear. The first band is related to the presence of unreacted and/or in situ formed H₂O₂ (by ^•OH) radical recombination (Eq. (6)), the second one can be attributed to the π system delocalization [28]:

After 45 min of reaction, the band at 430 nm disappeared and the color of the solution was completely degraded. However, the other two bands (at 220 and 330 nm) still remained, indicating that some intermediate degradation compounds are reformed. It is known that the reaction intermediates (mainly oxalic and fumaric acid, highly dependent on the initial dye concentration) can be formed as a result of the oxidation of the azo dyes. Moreover, some of them are characterized by stability and toxicity much higher than these of their parent compounds [9]. Generally, up to a certain level, the higher ferrous salt concentration can generate a higher degradation rate of organic compounds. However, further addition of ferrous ions becomes inefficient, probably due to the consumption of the ^•OH by the ion excess. In this work, no significant change in %DCOD was detected in the studied range of Fe²⁺ concentration (not shown here). Therefore, a 1.0 mM Fe²⁺ concentration was selected for the final experimental design. This value is consistent with those reported elsewhere [8,29].

Recently, Oancea and Meltzer [30] have reported on the photo-Fenton efficiency in yellow 5 degradation. After 20 min of irradiation, (initial tartrazine concentration = 5.5 ppm, Fe²⁺ = 0.0828 mM, H₂O₂ = 0.062 mL/L), 98% of degradation was obtained and the decrease in TOC was up to 43%. The prolongation of the irradiation time (120 min) led only to the partial degradation of these small molecules and consequently TOC did not exceed 80% [31]. In this work, 100% of dye degradation, 99% of COD degradation and 85% of mineralization (TOC) were reached in 180 min. We believe that the observed discrepancies in the obtained results can be related to the difference between the initial dye concentration, reactor type and concentration of the oxidizing agent used. Gupta et al. [27] have studied the removal of tartrazine by photodegradation on titanium oxide (TiO₂ 0.18 g/L and pH 11, UV lamp wavelength of 254 nm – heterogeneous photo-Fenton process). They have found that the photodegradation rate decreases with increasing dye concentration (from 10 to 32 ppm) and then it became almost constant up to 42 ppm. The increase in dye concentration from 10 to 42 ppm led to the decrease in the degradation efficiency from 96.12% to 41.89%, during 40 min of reaction. They found that TiO₂ concentration, light intensity and H₂O₂ concentration were the main factors that affected the decolorization process. The authors concluded that the maximum degradation efficiency was achieved with the combination of UV + H₂O₂ + TiO₂ (H₂O₂ 1.5 mM) [30]. On the other hand, Thiama et al. [31] have investigated a traditional technology such as electrocoagulation (EC) and some emerging ones like electrochemical advanced oxidation processes (EAOPs). After 15 min of the treatment, the dye (278 ppm) degradation reached 100% and 60% TOC.

3.3 Mineralization study

In order to evaluate appropriately the degradation level that can be reached using the photo-Fenton process, it is necessary to understand the details of E102 dye mineralization. Fig. 6 presents the degradation percentages of COD and TOC achieved in this study.

As it could be seen, the change in COD degradation is more significant than that of TOC. It could be related to the fact that, in the earlier stages of wastewater oxidation, some partially oxidized intermediate compounds can be formed, which are unable to absorb UV radiation. As the process runs, the created intermediate compounds increase their oxidation state, finally achieving their partial mineralization. This suggests that oxidation would preferably occur on the chromophore structure rather than on the dye molecule skeleton. We believe that in the case of yellow 5 degradation, the ^•OH radicals can react by hydrogen abstraction from or addition to double bonds [32]. Thus, the ^•OH radicals first attack the azo groups and open the N = N bonds, destructing the long conjugated π systems, causing decolorization. Considering that C–C and C = C bonds, present in the aromatic ring structures, are stronger than N = N bonds, ring opening needs more time [33]. Consequently, the aromatic fragments remaining after the oxidation process can be considered as intermediate compounds. Under optimized conditions, a kinetic analysis was developed by monitoring %DCOD and %DTOC as a function of time. All experiments were performed during 180 min, at pH = 3, and at 25 °C. It was found that %DCOD dye degradation reached 99.8% at an initial dye concentration equal to 100 ppm (Fig. 4). The correlation between ln(C_i/C_i0) and the irradiation time was linear, as in the case of a typical first-order plot [34]. The kinetic constant equaled 0.0791 min⁻¹ (R² = 0.952).

3.4 Effect of the presence of inorganic ions

The presence of inorganic anions in textile wastewaters plays an important role in the photo-oxidation kinetics of different dyes. Some of them (e.g., the Cl⁻ anion) may be added to facilitate the dyeing process. However, the inorganic anions may induce or reduce the rate of photooxidation. It has been reported that the “common ion effect” due to the presence of inorganic anions led to a decreased π-electron (electrostatic) repulsion between two ionic dyes, increasing the degree of aggregation [35]. It is known that solubilization and ionization of dyes, which affect their attack by ^•OH radicals, decreased by increasing the aggregation degree. Thus the tendency for reaction between ^•OH radicals and the dye is expected to decrease as the dye undergoes greater aggregation [36]. A series of experiments were performed to study the effect of sodium chloride (NaCl), with concentrations similar to those of the operational conditions, on the photodegradation process. Fig. 7 presents the effect of NaCl on %DCOD during the photo-Fenton process.

One can see that %DCOD decreased dramatically in the presence of NaCl. This behavior may be attributed to ^•OH radical scavenging by the Cl⁻ ion, which is known to react with the ^•OH radical (forming an OHCl^•− radical) at pH 2–3 [37]. The OHCl^•− radical generated is less reactive than the ^•OH one, leading to a lower degradation rate of yellow 5 dye.