This study aimed to kinetically discover optimal conditions on characteristics of Reactive Black 5 decolorization/degradation via ferrous (Fe2+)-activated potassium persulfate (PS). Monod-like kinetics and interactive model-based response surface methodology (RSM) were applied to fitting and predict optimize treatment. Biodegradability of the intermediates was also tested by shaking culture with two species (Proteus hauseri ZMd44 and Shewanella sp. WLP72). Results showed that the optimal degradation efficiency was predicted (through RSM) as pH 3.72, (PS) = 0.39 mM, and (Fe2+) = 0.29 mM. The transformation products (dl-4-hydroxymandelic acid, benzoic acid, benzene, formic acid, oxalic acid and acetic acid) were less toxic than the original dye solution. According to those results, clean-up of dye pollutants by the Fe2+/S2O82− process is feasible as a pre-processing for the biodegradation, and the predicted optimal conditions are meaningful for further industry utilization.

The textile dyeing industry generates a considerable amount of dye-bearing wastewater (WW). Thus, residual dyes in WW inevitably released to the environment pose significant threat to living organisms globally (Peng et al. 2008). Azo dyes are the most popularly used textile dyes (>60%), and the removal of these dyes from WW has been of utmost priority under US EPA regulations (Chen et al. 2013). Azo dyes or derived intermediates (e.g., aromatic amines) are potentially cytotoxic and possibly mutagenic to organisms (Stylidi 2004). Conventional physical, chemical, and biological methods are usually not efficient for degrading recalcitrant azo dyes (Rodriguez et al. 2014). Thus, exploring cost-effective methods to degrade textile dyes is important for sustainable development.

Advanced oxidation processes (AOPs) are used to effectively treat dye-bearing WW by destroying recalcitrant organic contaminants, because the hydroxyl radical and/or other radicals are highly active for degradation (Anipsitakis & Dionysiou 2003). An example of an AOP is the Fenton-associated reaction (e.g., electro-Fenton, photo-Fenton, and photoelectro-Fenton) (Khataee et al. 2014; Antonin et al. 2015; Li et al. 2015a, 2015b). TiO2-mediated photocatalysis (Wetchakun et al. 2015) and other oxidative reactions using potassium permanganate (Aleboyeh et al. 2009) and Co2+/oxone reagent (Yu et al. 2013) have been examined for dye degradation. Dye is usually removed at acidic pH levels (ca. 3) in Fenton processes, because the decolorization rates of dyes are relatively slow under alkaline condition. Moreover, Fenton-like systems are not very promising because of the required high levels of H2O2 (Kasiri et al. 2008) or electric energy (Panizza & Oturan 2011; Solano et al. 2015). Methods using TiO2-mediated photocatalysis are still not economically feasible for large-scale treatment (Kansal et al. 2010) and present a potential safety hazard due to the nano-particles (Li et al. 2016).

Persulfate (PS)-catalyzed techniques are used to degrade several pollutants (Liang et al. 2013; Zhao et al. 2013; Ayoub & Ghauch 2014; Monteagudo et al. 2015; Yan et al. 2015). These techniques provide alternatives for dye mineralization (Xu & Li 2010). As one of the strongest oxidants in aqueous phases (E0 = 2.01 V) and compared with H2O2 (E0= 1.76 V), PS is highly soluble, appropriate for numerous reactants, relatively stable, inexpensive, and more persistent in the subsurface than peroxide (Liang et al. 2007). An application based on the activation of PS ion (S2O82−) to generate sulfate radicals (SO4) for reactivity (E0 = 2.5–3.1 V vs. NHE) exhibits great potential for contaminant degradation (Xu & Li 2010). Readily available ferrous ions can effectively activate PS to generate SO4 radicals for oxidation, as shown by the following mechanisms (Li et al. 2015a, 2015b):
formula
1
formula
2
However, if Fe2+ is in excess, Fe2+ would significantly scavenge SO4 and inhibit radical oxidation of contaminants (Equation (2)). Quenching of SO4 by Fe2+ can be a major side reaction (e.g., at higher concentrations of Fe2+) based on the reaction rate constant between SO4 and Fe2+ (Equation (2)). A threshold amount of Fe2+ is required to effectively activate PS, whereas excessive Fe2+ is detrimental to degradation. Thus, the main factors affecting degradation performance are pH, concentration of PS and ferrous ions, and the concentration of target pollutant. The concentration of dyes during dye degradation via AOPs could significantly affect the decontamination efficiency (He et al. 2007; Medien & Khalil 2010; Ghoreishian et al. 2014). Kusic et al. reported the optimal conditions of different PS activation methods (Fe0/PS and Fe2+/PS) through response surface methodology (RSM) (Kusic et al. 2011). The current study attempted to discover the interactions of the main factors with the decolorization performance of reactive black 5 (RBk5) and RSM was used to quantitatively confirm the individual and interactive effects of the operating parameters (i.e., proton and PS and Fe(II) concentration) on the efficiency of treating 0.05 and 0.10 mM RBk5. Additionally, the biodegradability of RBk5 transformation products treated by the Fe2+/PS process is rarely reported. Therefore, the feasibility of the biodegradation of these intermediates was comprehensively assessed.

Materials and chemicals

RBk5 (C26H21N5Na4S6, MW = 991.82 g mol−1) (Zhang et al. 2012) was obtained from Everlight Chemical Ltd (Taiwan). All chemicals were of reagent grade and used without further purification. Milli-Q water, with a resistivity of 18.25 MΩ cm−1 (Millipore Milli-Q system), was used to prepare the solution. Sodium PS (Na2S2O8, 99%) and ferrous sulfate (FeSO4·7H2O) were obtained from Sigma-Aldrich Inc. Moreover, 1,10-phenanthroline, potassium iodide, and ethanol were purchased from the National Chemical Reagent Co. (Xiamen, China). Concentrated sulfuric acid (>98.0%) and sodium hydroxide (>96.0%) were purchased from Sinopharm Chemical Reagent Company (Shanghai, China) and diluted to 1 M for initial pH adjustment.

Batch experiments

RBk5 was degraded in a 100 mL shake flask under mechanical stirring at room temperature (25 ± 1 °C). The initial pH levels were adjusted by 1 M sodium hydroxide (NaOH) or sulfuric acid (H2SO4). The effect of initial pH was determined by inspecting the dye degradation at pH 2.0–12.0 (±0.05), and the effect of initial pH on the shift in absorbance was measured. A pH buffer was not added to the flask to prevent interference from the buffer solution. The results showed no significant differences in the tested pH range. The initial pH was adjusted to 4.0 in the remaining experiments. Seven different mole ratios of Fe2+ to PS, namely, 1:4, 1:2, 1:1, 1.5:1, 2:1, 3:1, and 4:1, were tested to study the effect of Fe2+-activated PS, and (Fe2+) or (PS) was controlled at 0.2 mM. The effect of dye concentration was determined by testing 0.025, 0.050, 0.075, and 0.100 mM RBk5 at pH 4.0, (PS) = 0.2 mM, and (FeSO4) = 0.2 mM.

Blank experiments were conducted prior to all experiments (e.g., PS/RBk5, Fe2+/RBk5, and Vis600 nm/RBk5) to ensure that no decolorization (<0.5%) had occurred (data not shown). Decolorization of RBk5 was determined by on-line continuous monitoring with absorbance at λmax (Hach, DR3900), because the decolorization was performed rapidly.

Analysis

Decolorization percentage was established by measuring the absorbance in the visible spectra at the characteristic wavelength of RBk5 (λmax = 600 nm) via a UV-vis spectrophotometer. The UV-vis results were confirmed by high performance liquid chromatography (Waters) (Figures S1–S3, available with the online version of this paper). The concentration of PS anions was determined by spectrophotometric determination with potassium iodide (Liang et al. 2008), whereas total iron and ferrous iron were measured with 1,10-phenanthroline at 510 nm (Rice et al. 2012). The pH level and oxidation reduction potential (ORP, the index of the tendency of a chemical species to donate/acquire electrons and thereby be oxidized or reduced) were monitored using a pH meter (SP-2100, Suntex, Taiwan) equipped with a pH electrode and a silver chloride and saturated calomel electrode, respectively. Chemical oxygen demand (COD) was determined using rapid digestion spectrophotometric determination (Hach, DR3900).

The initial degradation rate of RBk5 was investigated using the first-order kinetics (Equations (3) and (4)). The decolorization percentage DE (%) of RBk5 was calculated using Equation (5) as follows:
formula
3
formula
4
formula
5
where C0 is the initial concentration of RBk5 (mM), C is the RBk5 concentration at time t (s), and k1 is the first-order rate constant (s−1).

Experimental design

A central composite design (CCD) consisting of 17 experimental sets was performed, including the center point in triplicate. CCD combined with RSM was used to determine the optimal conditions of initial pH, and concentration of iron activator (Fe2+) and oxidant (K2S2O8) to develop empirical models to apply Fe2+/S2O82− processes to degrade RBk5.

Experimental data were analyzed using Design-Expert 8.0.6 software and fitted to a second-order polynomial equation. The following function (Equation (6)) was employed to obtain the interactive relationships between dependent and independent variables (Arslan-Alaton et al. 2009):
formula
6
where b0 is the average value of the experimental response, bi is the main effect of the coded variable Xi, bii is the quadratic effect of the coded variable Xi, and bij is the interaction effect between the respective coded variables.

Initial Fe2+ and PS concentrations and initial pH of the WW were considered as the independent process-specific variables, whereas DE and k1 were considered as the dependent variables (i.e., output responses). Values of the independent variables and their variation limits were determined based on the related scientific literature and the experimental results obtained in pretest studies conducted with azo dye. The results are presented in Table 1 for (RBk5)0 = 0.05 mM and 0.10 mM.

Table 1

Experimental design of level setting for three independent variables ((a) (RBk5)0 = 0.05 mM; (b) (RBk5)0 = 0.10 mM)

Input variableSymbolRange and levels
− 1.73− 10+ 1+ 1.73
(a)
Initial pH X1 2.37 5.73 
PS (mM) X2 0.13 0.2 0.3 0.4 0.47 
Fe2+ (mM) X3 0.13 0.2 0.3 0.4 0.47 
(b)
Initial pH X1’ 2.37 5.73 
PS (mM) X2’ 0.26 0.4 0.6 0.8 0.94 
Fe2+ (mM) X3’ 0.26 0.4 0.6 0.8 0.94 
Input variableSymbolRange and levels
− 1.73− 10+ 1+ 1.73
(a)
Initial pH X1 2.37 5.73 
PS (mM) X2 0.13 0.2 0.3 0.4 0.47 
Fe2+ (mM) X3 0.13 0.2 0.3 0.4 0.47 
(b)
Initial pH X1’ 2.37 5.73 
PS (mM) X2’ 0.26 0.4 0.6 0.8 0.94 
Fe2+ (mM) X3’ 0.26 0.4 0.6 0.8 0.94 

Biodegradability test

The acute toxicity of the reaction solution was determined with two species (Proteus hauseri ZMd44 and Shewanella sp. WLP72) which had been cultured in the laboratory for more than five generations. Prior to the tests, 200 mg L−1 RBk5 was decolorized through the Fe2+/PS process under different molar ratios ((PS)/(Fe2+) = 0.5:1, 1:1, 2:1, and 4:1; (Fe2+) = 0.2 mM; pH 4; reaction time = 30 min), and 50 mL of treated RBk5 solution was added into 50 mL of Luria-Bertani medium (LB broth, Miller, Difco). The pH level was then adjusted to pH 7 by phosphate buffer. The acute toxicity experiments were conducted in 150 mL Erlenmeyer flasks using pre-cultured P. hauseri ZMd44 and Shewanella sp. WLP72 species, and the samples were divided into five groups. Synchronous growth activity and maximum metabolic function were ensured in the same growth phase for bioassay by inoculating 5% (v/v) cultured broth on fresh sterile LB medium. A cell culture was harvested at approximately mid-exponential growth phase (ca. 4 h) for further toxicity assessment. Four groups were used for the test, while one group was used for the blank (200 mg L−1 RBk5 not decolorized). All samples were incubated aerobically at 30 °C and centrifuged at 200 rpm. The specific growth rates in the exponential growth phase were determined through the measured OD600, which was determined by removing the supernatant of 1 mL of culture after centrifugation and diluting the microbial sediment to 1 mL using sterile deionized water.

Effect of UV–visible spectral changes

Time courses of spectral profiles were monitored (Figure 1(a)) to discover whether decolorized intermediates (e.g., aromatic amines) were generated. RBk5 was decolorized by ca. 90% within 10 min (Figure 1(b)). Despite the complexity, RBk5 concentrations at any time t following a single impulse and RBk5 addition can be described by , where λi and Ci denote eigenvalues and concentrations of reactant and intermediate(s) species, respectively. Initial rates at all intermediate concentrations were determined to be zero (Notari 1987). RBk5 concentration is known to behave according to linear kinetics, because the equation is empirically adequate. The RBk5 profiles followed the model of bioexponential disposition (Chen 2002). That is, the degradation of RBk5 might have occurred in two steps, namely, fast decolorization of RBk5 in the first 2 min and then slow degradation for a longer period of ca. 30 min. Intermediary products were likely more inhibitory of chemical stability than raw compound RBk5. Thus, the first step was the predominant overall reaction, and its rate constant may follow Monod-like kinetics. That is, the rate of reaction in RBk5 during the first step is a first-order reaction. The reaction order in RBk5 diminished continuously from maximum to zero as intermediate concentrations continually increased.
Figure 1

(a) Changes in the UV-vis spectra of RBk5; (b) time-series profile of RBk5 measured at 600 nm; (c) COD changes of 0.05 mM RBk5 during Fe2+/PS treatment; (d) changes in the UV-vis spectra of real wastewater ((RBk5)0 = 0.05 mM; (PS)0 = 0.2 mM; (Fe2+)0 = 0.2 mM; pH0 = 5).

Figure 1

(a) Changes in the UV-vis spectra of RBk5; (b) time-series profile of RBk5 measured at 600 nm; (c) COD changes of 0.05 mM RBk5 during Fe2+/PS treatment; (d) changes in the UV-vis spectra of real wastewater ((RBk5)0 = 0.05 mM; (PS)0 = 0.2 mM; (Fe2+)0 = 0.2 mM; pH0 = 5).

The spectral profiles of RBk5 before and after oxidation could be detected at the visible wavelength regions (e.g., 600 nm as signal wavelength of azo linkage in RBk5). The characteristic peak at 600 nm decreased rapidly within 2 min, and then the absorbance of peaks was attenuated to minimal values (Figure 1). This result implied that the conjugate structure (–N=N–) of RBk5 was destroyed, leading to rapid color removal. Two peaks at 336 and 322 nm occurred after 30 s of decolorization reaction, and the absorbance decreased slowly afterward. Hence, RBk5 was degradation, and derived intermediates formed. The time-series profile of COD was measured (Figure 1(c)) to investigate the mineralization of RBk5 by Fe2+/PS treatment. COD was rapidly reduced in the first 2 min and then decreased slowly in the next 30 min. This phenomenon was in accordance with the trend of RBk5 decolorization. The results indicated that the decolorization and mineralization of RBk5 were synchronous. By comparing with other research, the advantages of the proposed approach were obvious since high efficiency can be achieved in short time (as shown in Table S1, available with the online version of this paper). In the research of Zhang and coworkers, 90% efficiency can be achieved in 120 min by the photocatalytic process (Zhang et al. 2012); however, lower efficiency of biodegradation was achieved by a graphene-TiO2 triggered reaction (Hsueh et al. 2009; Li et al. 2013; Zhang et al. 2015). In contrast, high decolorization percentage of 98% could be obtained by the Fe2+/S2O82− process in 30 min in this research. As known, the rate of oxidation and reduction for RBK5 dyes is strongly related to the degradation pathway of azo dyes. To deeply understand the mechanism of reaction, the byproducts and intermediate products were analyzed by gas chromatography–mass spectrometry (GC-MS, Varian 240-MS); the results are shown in Figure S4 (available online). Three compounds were elucidated as byproducts and intermediate products formed during the degradation of RBK5 (Table S2, available online), after 10 min reaction. DL-4-hydroxymandelic acid (retention time (RT) = 55.05 min) is identified as a major product in the Fe2+/PS process, which supports the fission of the –N=N– group. Then the DL-4-hydroxymandelic acid could be further degraded to benzoic acid (RT = 43.74 min) and benzene (RT = 33.68 min). However, after 40 min reaction, all the benzodiazepine material was completely decomposed. The cleavage of the benzene ring formed some intermediates of formic acid, oxalic acid and acetic acid (as shown in Table S2). Finally, all the byproducts will be mineralized to CO2 and water. By comparing with the results of Zhang et al. (2015), smaller molecules can be achieved in this research, which indicates the complete degradation. The photodegradation results in the desulfonation of dyes (Zhang et al. 2015) rather than the fission of the –N=N– group (this research). The difference in the degradation pathway explains the difference in efficiency.

Real textile WW (CODCr = 307–1,120 ppm; biochemical oxygen demand = 66.0–199 ppm; suspended solids = 104–114 ppm; pH = 10.62–11.16; chromaticity = 200–1,000; RBk5 = 200–1,000 ppm) from Eastern Asia (Xiamen, China) Corporation Ltd was also analyzed to further demonstrate the effect of treatment using Fe-activated PS. Chemical characterization of WW is shown in the inserted table in Figure 1(d). Besides RBK5, Reactive Blue 171 (RB171) and Reactive Green 19 (RG19) are the main organics in the WW. The concentration of all the dyes changed with the operation process. According to the result in Figure 1(d), the main adsorption band of real textile WW was located at 600 nm and increasing the reaction time led to decreased intensity of the adsorption band. This phenomenon indicated the strong activity of Fe-activated PS for oxidation, regardless of the complexity of the real textile WW.

Optimal operation variables

Effect of pH

The decolorization profile followed the bioexponential model (Figure 2). RBk5 could be decolorized rapidly in the first 2 min once ferrous sulfate was added, and then decolorization tended to be steady (Notari 1987; Zhang et al. 2012). The effect of initial pH for the Fe2+/PS process is presented in Figure 2(a). The most favorable pH levels for RBk5 degradation were at ca. pH 3–4. Alkaline pH levels (ca. pH 9–12) were not favorable for the reactions involved in RBk5 degradation. High DE (DE > 80% in 10 min) was observed at pH 2–7, indicating that the Fe2+/S2O82− process could still be performed within a wide range of acidic pHs. The fastest k1 of RBk5 was observed at ca. pH 4.0, and increasing or decreasing pH levels would lower decolorization performance. However, second-stage RBk5 DE (t > 2 min) was slightly slower at pH 4.0 than at pH 3.0. This phenomenon is probably caused by the fact that additional SO4 could be formed with acid-catalyzation in acidic pH levels (Equations (7) and (8)), as follows (Saien et al. 2011):
formula
7
formula
8
Figure 2

Effect of (a) pH; (b) PS; and (c) Fe2+ on the efficiency of the Fe2+/PS-catalyzed decolorization process.

Figure 2

Effect of (a) pH; (b) PS; and (c) Fe2+ on the efficiency of the Fe2+/PS-catalyzed decolorization process.

Moreover, the amount of free Fe2+ declined because of the formation of Fe2+ complexes at pH > 4.0 (Equation (9)) (Rao et al. 2014):
formula
9
The degradation efficiency drastically decreased from 89.79% to 53.83% in 10 min (inserted box in Figure 2(a)) when pH was increased from 4.0 to 9.0. Then, decolorization was almost terminated at alkaline pH levels (e.g., pH 12). Further increase in pH levels led to the precipitation of Fe(OH)2. Both FeOH+ and Fe(OH)2 had much lower catalytic activity toward PS ions, thereby producing fewer sulfate radicals (Rao et al. 2014). The k1 value of RBk5 was lower at pH 2.0 than at pH 3.0, which can be explained by two points. First, complex (Fe(H2O))2+ formed at low pH levels (Masomboon et al. 2009), leading to a deficiency of free Fe2+ to activate PS ions in the solution and fewer sulfate radicals. Second, since PS could be acid-activated, higher SO4 was generated in highly acidic conditions, and this process could favor radical scavenging reactions, instead of radical with RBk5 reactions (Peyton 1993). Possible chemical scavenging reactions (i.e., Equations (10)–(13)) to compete with SO4/RBk5 interaction are as follows (Saien et al. 2011):
formula
10
formula
11
formula
12
formula
13
The pH levels may decline, because H+ was propagated from the reaction shown in Equation (10), and OH was consumed via Equation (11). The pH level and ORP changed rapidly during the first stage (<2 min) when pH0 was between pH 4 and pH 9, and the final pH (pHf) was ca. 3.5 (data not shown). The pHf was almost identical with pH0 at pH0 < 3. This result does not imply that the Equation (10) reaction does not occur in strong acidic media, but rather confirms that the amount of released H+ is so low such that significant reduction is appropriate under alkaline and natural pH conditions only (pH 4–9) (Saien et al. 2011). However, when initial pH0 was much higher than 7, Fe2+inevitably precipitated as Fe(OH)2 ↓ , and apparently Fe2+/S2O82− AOP was terminated. Given that the pKa value for bisulfate ion (HSO4) was 2.0 (Ghoreishian et al. 2014), at pH ≤ 2.0, the released H+ via Equation (10) could react with SO42− and form HSO4. That is, pH did not significantly change when initial pH0 was 2.0 probably because of the significant amounts of protons provided for the reaction.

Effect of PS dosage

Increasing the concentration of PS from 0.05 mM to 0.20 mM enhanced the degradation process (i.e., decolorization percentage increased from 41.29% to 88.87% in 10 min; Figure 2(b)). However, further increase in PS dosage did not significantly increase RBk5 decolorization. This phenomenon was similar to the finding when trichloroethane was degraded using a ferrous-activated PS process (Liang et al. 2004). Increasing the PS concentration from 0.2 mM to 0.8 mM resulted in the slight enhancement of RBk5 decolorization from 88.87% to 93.02% (within 10 min). Thus, the optimal concentration of PS was chosen at 0.2 mM for cost-effective consideration.

PS was activated by Fe2+ to generate sulfate radicals via Equation (1) to decolorize RBk5. Higher PS concentration would result in more collision with Fe2+ to generate more SO4. Then, the more opportunities to attack the RBk5 molecules, the higher the DE. Moreover, higher probability of collision between PS and Fe2+ caused significant reductions in Fe2+ (Jiang et al. 2013). Consequently, less SO4 was quenched by Fe2+, enhancing DE as the concentration of PS increased. Meanwhile, higher SO4 generation led to higher radical concentrations, which could favor the reaction with radicals (e.g., •OH in Equations (10)–(13)) instead of radical with RBk5. However, adding more S2O82− ions led to a low variation in DE and k1, probably because of the recombination of excessive generated •OH radicals to form less-reactive H2O2 molecules (Saien et al. 2011).

Effect of Fe2+

The DE of RBk5 increased from 76.91% to 88.61% with an increase in Fe2+ concentration from 0.05 mM to 0.2 mM (Figure 2(c)). Further increase in Fe2+ to 0.8 mM decreased the DE of RBk5 to 60%.

High concentration of Fe2+ had advantages and disadvantages in the generation and existence of SO4. Higher dosage of Fe2+ ions improved the generation rate of SO4, and this condition accelerated the decomposition of RBk5. Dye degradation improved as Fe2+ dosage increased. However, Fe2+ can also act as an effective scavenger of SO4, significantly reducing DE (Miao et al. 2015). Therefore, the optimal concentration of Fe2+ was chosen at 0.2 mM.

The variation in (Fe2+) was monitored during the reaction with different initial (Fe2+)0 ranging from 0.05 mM to 0.8 mM. Fe2+ was consumed rapidly, and its concentration was reduced sharply in the first 2 min. The PS degradation rate was enhanced from 10.45% to 88.47% with increased Fe2+ concentration (data not shown). This result was consistent with the finding of Liang et al. (2004), who showed that further increase in Fe2+ ions could have proportional augmentations to PS decomposition.

The kinetic model of k1

Thus, for the operation variables, the effect of pH on k1 could be modeled via diprotic acid-base equilibria (Equation (14)), as described in the Appendix (available with the online version of this paper):
formula
14
and the optimal pH at
The effects of PS and Fe2+ on k1 are determined using Michaelis–Menten or Monod-like kinetics, as follows (Equation (15)):
formula
15
where (A) represents the concentration of PS or Fe2+, and Ks is the half-velocity constant (i.e., the value (S) when k/kmax = 0.5).
The curves of the model predictions (Figure 3) show that pH > 7.0 is apparently not appropriate, possibly because of precipitate formation (e.g., Fe(OH)2 ↓). Kinetic modeling of Equations (14) and (15) could be simulated as follows:
formula
16
Figure 3

Comparison of experimental and predicted values via proposed kinetic models.

Figure 3

Comparison of experimental and predicted values via proposed kinetic models.

The optimal concentration of H+ for maximal k1 was as follows:
formula
Moreover, according to the aforementioned mechanisms of PS catalyst–multiple species systems (Berrington de Gonzalez & Cox 2005), the plausible model of k1 based upon PS and Fe2+ effect is shown in Equations (17) and (18):
formula
17
formula
18
In summary, the interactive model with a correlation factor of ϕ = 0.74 could be presented as shown in Equation (19), as follows:
formula
19

Effect of RBk5 concentration

Apparently, initial dye concentration could crucially affect the efficiency of decolorization, because pollutant concentration is of great importance to WW treatment (Ghoreishian et al. 2014). The DE gradually decreased with increased initial dye concentrations (Figure 4). After 10 min of reaction, decolorization of 0.025, 0.050, 0.075, and 0.10 mM RBk5 reached 92.32%, 88.87%, 86.70%, and 77.32%, respectively. The k1 values sharply decreased with increased RBk5 concentration as indicated in the inserted box in Figure 4. The efficiency of the process based on the amount of dye decolorized generally increased with increased initial concentrations (He et al. 2007).
Figure 4

Effect of RBk5 concentration on decolorization rate ((Fe2+)0 = 0.2 mM; (S2O82−)0 = 0.2 mM; initial pH 4.0).

Figure 4

Effect of RBk5 concentration on decolorization rate ((Fe2+)0 = 0.2 mM; (S2O82−)0 = 0.2 mM; initial pH 4.0).

As the concentration of RBk5 increased, the radical provided for degradation approached saturated levels. Increasing dye concentration only increased the number of dye molecules, but not the radicals available for reaction concentration. Therefore, the removal rate should be reduced (Lucas & Peres 2006). Moreover, the competition between dye intermediates and RBk5 for SO4 became intense because of the high oxidation capability of SO4. These factors contributed to such a decreased rate constant with an increase in initial concentration.

RSM analysis

Decolorization performance via the Fe2+/PS system was dependent on the initial concentration of RBk5 (Figure 4). However, a few studies mentioned optimal dye decolorization at numerous operation concentrations. In this study, the optimal color removal of RBk5 was investigated through RSM. Table 2 summarizes two sets of experimental designs for 0.05 and 0.10 mM RBk5, and the response results of DE for 0.05 and 0.10 mM RBk5 (Y′ and Y′′, respectively) are discussed. The predicted values were evidently consistent with the experimental data.

Table 2

Experimental design matrix and response based on experimental runs and predicted values on decolorization rate (DE; %) proposed by CCD design

RunDE (%)
pHX1PSX2Fe2+X3Y′(0.05 mM)
Y′′(0.10 mM)
LevelLevelLevelExperimentalPredictedExperimentalPredicted
−1 −1 −1 89.86 89.58 89.19 87.72 
−1 −1 +1 87.98 87.51 85.05 84.70 
−1 +1 −1 92.86 91.79 91.05 91.22 
−1 +1 +1 90.25 89.90 88.44 87.62 
+1 −1 −1 88.12 87.26 87.55 86.98 
+1 −1 +1 87.72 87.57 88.56 87.00 
+1 +1 −1 90.78 90.05 92.25 91.21 
+1 +1 +1 91.46 88.99 90.59 90.67 
−1.73 90.58 90.53 88.28 89.03 
10 +1.73 88.84 89.79 89.91 91.02 
11 −1.73 87.76 88.19 82.94 84.55 
12 +1.73 91.48 92.67 90.49 90.74 
13 −1.73 86.48 87.59 88.07 89.07 
14 +1.73 85.71 86.22 85.12 85.98 
15 92.59 92.53 91.64 91.51 
16 92.34 92.53 91.52 91.51 
17 92.68 92.53 91.38 91.51 
RunDE (%)
pHX1PSX2Fe2+X3Y′(0.05 mM)
Y′′(0.10 mM)
LevelLevelLevelExperimentalPredictedExperimentalPredicted
−1 −1 −1 89.86 89.58 89.19 87.72 
−1 −1 +1 87.98 87.51 85.05 84.70 
−1 +1 −1 92.86 91.79 91.05 91.22 
−1 +1 +1 90.25 89.90 88.44 87.62 
+1 −1 −1 88.12 87.26 87.55 86.98 
+1 −1 +1 87.72 87.57 88.56 87.00 
+1 +1 −1 90.78 90.05 92.25 91.21 
+1 +1 +1 91.46 88.99 90.59 90.67 
−1.73 90.58 90.53 88.28 89.03 
10 +1.73 88.84 89.79 89.91 91.02 
11 −1.73 87.76 88.19 82.94 84.55 
12 +1.73 91.48 92.67 90.49 90.74 
13 −1.73 86.48 87.59 88.07 89.07 
14 +1.73 85.71 86.22 85.12 85.98 
15 92.59 92.53 91.64 91.51 
16 92.34 92.53 91.52 91.51 
17 92.68 92.53 91.38 91.51 

A second-order polynomial model (Equation (6)) was fitted according to the 17 experimental sets of data presented in Table 2. The approximating function was obtained in terms of the coded factors, and this function described the DE for 0.05 and 0.10 mM initial RBk5 concentration as indicated in Equations (20) and (21), respectively.
formula
20
formula
21

Checking the adequacy of a model is an integral part of overall data analysis because the approximating model functions would yield poor or misleading results if the fit is inadequate (Korbahti 2007). Analysis of variance (ANOVA) results of the established model for DE are shown in Table 3 (DE′ and DE′′ for 0.05 and 0.10 mM RBk5, respectively). The mean squares were obtained by dividing the sum of the squares of each of the two sources of variation, that is, the model and the error (residual) variance, by the respective degrees of freedom (DF). The model F-value was calculated by dividing the model mean square by the residual mean square.

Table 3

ANOVA results of the quadratic model for DE

SourcesDFSum of squares
Mean square
F-value
Prob > F
DE′DE′′DE′DE′′DE′DE′′DE′DE′′
Model 74.30 97.66 8.26 10.85 7.05 5.84 0.0087 0.0149 
X1 2.47 4.62 2.47 4.62 2.11 2.49 0.1894 0.1589 
X2 23.44 44.84 23.44 44.84 20.03 24.11 0.0029 0.0017 
X3 2.20 11.18 2.20 11.18 1.88 6.01 0.2131 0.0440 
X1X2 0.16 0.27 0.16 0.27 0.14 0.15 0.7228 0.7126 
X1X3 2.84 4.65 2.84 4.65 2.43 2.50 0.1630 0.1578 
X2X3 0.015 0.16 0.015 0.16 0.013 0.087 0.9121 0.7761 
X12 5.51 2.99 5.51 2.99 4.71 1.61 0.0666 0.2452 
X22 6.01 20.27 6.01 20.27 5.14 10.90 0.0577 0.0131 
X32 43.02 21.55 43.02 21.55 36.75 11.59 0.0005 0.0114 
Residual 8.19 13.02 1.17 1.86 – – – – 
Lack of fit 8.13 12.98 1.63 2.60 52.40 153.35 0.0188 0.0065 
Pure error 0.062 0.034 0.031 0.017 – – – – 
Cor total 16 82.49 110.68 – – – – – – 
SourcesDFSum of squares
Mean square
F-value
Prob > F
DE′DE′′DE′DE′′DE′DE′′DE′DE′′
Model 74.30 97.66 8.26 10.85 7.05 5.84 0.0087 0.0149 
X1 2.47 4.62 2.47 4.62 2.11 2.49 0.1894 0.1589 
X2 23.44 44.84 23.44 44.84 20.03 24.11 0.0029 0.0017 
X3 2.20 11.18 2.20 11.18 1.88 6.01 0.2131 0.0440 
X1X2 0.16 0.27 0.16 0.27 0.14 0.15 0.7228 0.7126 
X1X3 2.84 4.65 2.84 4.65 2.43 2.50 0.1630 0.1578 
X2X3 0.015 0.16 0.015 0.16 0.013 0.087 0.9121 0.7761 
X12 5.51 2.99 5.51 2.99 4.71 1.61 0.0666 0.2452 
X22 6.01 20.27 6.01 20.27 5.14 10.90 0.0577 0.0131 
X32 43.02 21.55 43.02 21.55 36.75 11.59 0.0005 0.0114 
Residual 8.19 13.02 1.17 1.86 – – – – 
Lack of fit 8.13 12.98 1.63 2.60 52.40 153.35 0.0188 0.0065 
Pure error 0.062 0.034 0.031 0.017 – – – – 
Cor total 16 82.49 110.68 – – – – – – 

Values of Prob > F (p value) less than 0.050 imply that the model is statistically significant, whereas values >0.1000 usually imply non-significance. In this case, Prob > F-values of 0.0087 and 0.0149 denote that the DE of the employed models was significant. The model F-values of 7.05 and 5.84 imply that the model was significant (Table 3). Generally, a smaller p value of a parameter implies higher significance in the model (Yetilmezsoy et al. 2009). Obtained p values of the model components indicated that the quadratic effects of each parameter are more significant than the effects of their interactions.

Three-dimensional response surfaces were also provided to investigate the interactive effect of two variables on DE as shown in Figure 5(a) and 5(b) for 0.05 and 0.10 mM RBk5, respectively. The maximum predicted value is illustrated as the surface confined in the smallest ellipse in the contour diagram. Elliptical contours were obtained in the case of perfect interaction between the independent variables.
Figure 5

(a) Surface plot of the empirical model for 0.05 mM RBK5 decolorization percentage (DE; %) by Fe2+/S2O82− process (reaction time:10 min). (b) Surface plot of the empirical model for 0.10 mM RBK5 decolorization percentage (DE; %) by Fe2+/S2O82− process (reaction time:10 min).

Figure 5

(a) Surface plot of the empirical model for 0.05 mM RBK5 decolorization percentage (DE; %) by Fe2+/S2O82− process (reaction time:10 min). (b) Surface plot of the empirical model for 0.10 mM RBK5 decolorization percentage (DE; %) by Fe2+/S2O82− process (reaction time:10 min).

The normal probability plot and internally standardized residuals are shown in the Figure 6(a) and 6(c) for 0.05 mM RBk5 and Figure 6(b) and 6(d) for 0.10 mM RBk5. The normal probability plot is an important diagnostic tool used to detect and describe systematic deviations from assumptions that errors are normally distributed and independent of each other and that error variances are homogeneous (Yetilmezsoy et al. 2009). No serious violations were found in the assumptions underlying the analysis, and the residuals were independent (Figure 6(a) and 6(b)). Further check of model adequacy was performed by correlating the internally studentized residuals against the predicted values (Figure 6(c) and 6(d)). This plot tests the assumption of constant variance that points are randomly scattered, lying within the top and bottom limits for outlier detection. Thus, the developed DE was successful in capturing the correlation between the influencing parameters of the Fe2+/S2O82− process to predict the degradation of RBk5 model WW.
Figure 6

Model for DE diagnostic: (a) and (b) the internally studentized residuals and normal % probability plot for DE of 0.05 mM and 0.10 mM RBk5; respectively. (c) and (d) Predicted and plot of studentized residuals for DE of 0.05 mM and 0.10 mM RBk5; respectively.

Figure 6

Model for DE diagnostic: (a) and (b) the internally studentized residuals and normal % probability plot for DE of 0.05 mM and 0.10 mM RBk5; respectively. (c) and (d) Predicted and plot of studentized residuals for DE of 0.05 mM and 0.10 mM RBk5; respectively.

This model could be obtained using Taylor's expansion, and the polynomial model may describe the kinetic model in the domain near operation conditions. Here, the function (DE) could be written as follows:
formula
22
where , , and denote the deviations of variables X1, X2, and X3, respectively. Thus, the empirical polynomial model (i.e., contour analysis) could technically describe the system behavior (not just curve-fitting). According to Equation (22), Equation (6) could be presented as follows:
formula
23
The optimum conditions obtained according to the model in Equations (20) and (21) are shown in Table 4. Random experiments were also conducted to test the feasibility of the RSM model, and the results are shown in Table 5. Good agreement was found between the experimental data and predicted results.
Table 4

Optimal Fe2+/S2O82− process conditions for RBk5 decolorization

(RBk5)X1X2 (mM)X3 (mM)PredictedActual
0.05 mM 3.72 0.39 0.29 93.19 93.28 
0.10 mM 4.55 0.72 0.53 92.43 92.56 
(RBk5)X1X2 (mM)X3 (mM)PredictedActual
0.05 mM 3.72 0.39 0.29 93.19 93.28 
0.10 mM 4.55 0.72 0.53 92.43 92.56 
Table 5

The random experimental results for the feasibility test of RSM

(RBk5)X1X2 (mM)X3 (mM)PredictedActualDeviation
0.05 mM 4.0 0.20 0.20 89.09 88.87 −0.25 
0.05 mM 4.0 0.20 0.30 90.53 87.34 −1.34 
0.05 mM 5.0 0.30 0.40 89.75 88.86 −0.99 
0.05 mM 3.0 0.30 0.20 91.39 90.36 −1.13 
0.10 mM 4.0 0.40 0.20 91.59 91.26 −0.36 
0.10 mM 3.5 0.7 0.5 91.96 92.48 +0.57 
0.10 mM 0.60 0.80 87.45 86.54 −1.04 
0.10 mM 4.5 0.50 0.70 89.85 89.28 0.63 
0.10 mM 0.40 0.60 88.32 86.92 −1.59 
0.10 mM 4.0 0.80 0.60 92.00 91.26 −0.80 
(RBk5)X1X2 (mM)X3 (mM)PredictedActualDeviation
0.05 mM 4.0 0.20 0.20 89.09 88.87 −0.25 
0.05 mM 4.0 0.20 0.30 90.53 87.34 −1.34 
0.05 mM 5.0 0.30 0.40 89.75 88.86 −0.99 
0.05 mM 3.0 0.30 0.20 91.39 90.36 −1.13 
0.10 mM 4.0 0.40 0.20 91.59 91.26 −0.36 
0.10 mM 3.5 0.7 0.5 91.96 92.48 +0.57 
0.10 mM 0.60 0.80 87.45 86.54 −1.04 
0.10 mM 4.5 0.50 0.70 89.85 89.28 0.63 
0.10 mM 0.40 0.60 88.32 86.92 −1.59 
0.10 mM 4.0 0.80 0.60 92.00 91.26 −0.80 

Table 4 shows that the optimal conditions of RBk5 decolorization were dependent on concentration of RBk5. Highly acidic conditions were favorable to decolorize 0.05 mM RBk5. However, the optimal decolorization of 0.10 mM RBk5 was performed at relatively high pH level, where (PS) and (Fe2+) were just less than twice the results at 0.05 mM RBk5, indicating that the RBk5 could be more readily and efficiently decolorized at high concentration. That is, more SO4 could be consumed by RBk5 when at high RBk5 concentration. Consequently, less SO4 was quenched by Fe2+ (Jiang et al. 2013). Thus, the degradation and utilization efficiency of PS and Fe2+ both increased. This finding is of great importance when considering the industrial application to WW decolorization.

Moreover, Yates's algorithm was applied to consider the main and interactive effects. The effects of 23 factorial designed are shown in Table 6. The pH level showed a remarkable negative impact on the Fe2+/S2O82− process. By contrast, the effect of pH level was positive for 0.10 mM RBk5 treatment, suggesting that the treatment of 0.10 mM RBk5 had high efficiency at wider pH ranges. Taking all the parameters into consideration, the best operating conditions were defined as pH 3.72, (PS) = 0.39 mM, and (Fe2+) = 0.29 mM, which resulted in the maximum decolorization percentage and possibility for industrial application.

Table 6

Estimated effects from 23 factorial design via Yates's algorithm

Estimated effectX1X2X3X1X2X1X3X2X3X1X2X3Standard error
DE′ −0.7175 2.9175 −1.0525 0.2825 1.1925 0.0875 0.4525 0.6547 
DE′′ 1.305 2.995 −1.85 0.37 1.525 −0.285 −1.05 0.9548 
Estimated effectX1X2X3X1X2X1X3X2X3X1X2X3Standard error
DE′ −0.7175 2.9175 −1.0525 0.2825 1.1925 0.0875 0.4525 0.6547 
DE′′ 1.305 2.995 −1.85 0.37 1.525 −0.285 −1.05 0.9548 

Biodegradability test

Considering aquatic ecological security, the toxicity of transformation products should be tested because of possible increased biotoxicity (Duarte et al. 2013). As the results of GC-MS show (Figure S4 and Table S2), the transformation products of the PS process were dl-4-hydroxymandelic acid, benzoic acid, benzene, formic acid, oxalic acid and acetic acid, respectively. Considering the toxicity of benzene, the toxicity of the treated dye solution was monitored by inhibiting the growth of two species (P. hauseri ZMd44 and Shewanella sp. WLP72). The specific growth rates (μ, h−1) in the exponential growth phase were calculated using the value of OD600. Higher specific growth rates of the microbes indicated lower acute toxicity of the solution. The lowest growth rates of ZMd44 and WLP72 were found in the raw RBk5 solution (Figure 7). The μ value of ZMd44 and WLP72 under the different degradation conditions are shown in Table 7. The results showed that the microbes grew much better in the degraded solution than in the raw solution, implying that all generated products were less toxic than the original dye, since benzene degraded to organic acid quickly and correspondingly had less negative impact on microbes. De Luna et al. also suggested that all generated products of Orange II were less toxic than the original dye during the homogeneous photo-Fenton process (de Luna et al. 2014). Moreover, the biggest μ indicated that the best operation ratio of PS/Fe2+ was 2:1. The COD removal rate was also determined after decolorization for further verification. COD removal rates of 38.62%, 63.51%, 72.15%, and 56.47% were achieved when the molar ratios of PS/Fe2+ were 0.5:1, 1:1, 2:1, and 4:1, respectively. The toxicity test shows that PS/Fe2+ treatment is a suitable method for complete decolorization and detoxification. Moreover, the biodegradation process progressed better at PS/Fe2+molar ratio of 2:1.
Table 7

The comparison of specific growth rate (μ; h−1) of ZMd44 and WLP72 after supplementation with the decolorization intermediates of RBk5

(PS)/(Fe2+)Blank0.5:11:12:14:1
ZMd44 0.04016 0.04619 0.05052 0.1058 0.06479 
WLP72 0.02766 0.03427 0.03717 0.07037 0.04478 
(PS)/(Fe2+)Blank0.5:11:12:14:1
ZMd44 0.04016 0.04619 0.05052 0.1058 0.06479 
WLP72 0.02766 0.03427 0.03717 0.07037 0.04478 
Figure 7

The biodegradability of RBk5 decolorization intermediates under different molar ratio of PS/Fe2+.

Figure 7

The biodegradability of RBk5 decolorization intermediates under different molar ratio of PS/Fe2+.

Decolorization of azo dye RBk5 by Fe2+-activated PS has been optimized using interactive model-assisted RSM. Three variables (i.e., pH, PS concentration, and ferrous concentration) were defined as significant factors for decolorization. However, the decolorization was more favorable at pH 2–7, and the effect of (PS)/(Fe2+) was highly significant. The oxidative reaction apparently followed Monod-like kinetic characteristics. The effects of independent variables and their interactions on decolorization percentage of RBk5 were clearly elucidated using the RSM. Comparison of decolorizations of 0.05 and 0.10 mM RBk5 indicated that decolorization could be performed in moderate pH levels under high concentrations of RBk5. The biodegradability test showed that the decolorization intermediates were less toxic than the original dye solution, and the optimal molar ratio of PS/Fe2+ for detoxification and mineralization was 2:1.

This work was financially supported by Fujian province Science and Technology project Foundation (2014Y4006), Xiamen Technology project Foundation (3502Z20140057, 3502Z20153025, 3502Z20151256), Quanzhou Technology project Foundation (2016Z074), the Fundamental Research Funds for the Central Universities (ZQN-PY217) and China Postdoctoral Science Foundation (2015M582032).

Aleboyeh
A.
Olya
M. E.
Aleboyeh
H.
2009
Oxidative treatment of azo dyes in aqueous solution by potassium permanganate
.
Journal of Hazardous Materials
162
,
1530
1535
.
Antonin
V. S.
Garcia-Segura
S.
Santos
M. C.
Brillas
E.
2015
Degradation of Evans Blue diazo dye by electrochemical processes based on Fenton's reaction chemistry
.
Journal of Electroanalytical Chemistry
747
,
1
11
.
Arslan-Alaton
I.
Tureli
G.
Olmez-Hanci
T.
2009
Treatment of azo dye production wastewaters using photo-Fenton-like advanced oxidation processes: optimization by response surface methodology
.
Journal of Photochemistry and Photobiology A: Chemistry
202
,
142
153
.
Chen
B. Y.
Hong
J. M.
Ng
I. S.
Wang
Y. M.
Liu
S. Q.
Lin
B.
Ni
C.
2013
Deciphering simultaneous bioelectricity generation and reductive decolorization using mixed-culture microbial fuel cells in salty media
.
Journal of the Taiwan Institute of Chemical Engineers
44
,
446
453
.
de Luna
L. A. V.
da Silva
T. H. G.
Pupo Nogueira
R. F.
Kummrow
F.
Umbuzeiro
G. A.
2014
Aquatic toxicity of dyes before and after photo-Fenton treatment
.
Journal of Hazardous Materials
276
,
332
338
.
Duarte
F.
Maldonado-Hódar
F. J.
Madeira
L. M.
2013
New insight about orange II elimination by characterization of spent activated carbon/Fe Fenton-like catalysts
.
Applied Catalysis B: Environmental
129
,
264
272
.
Ghoreishian
S. M.
Badii
K.
Norouzi
M.
Rashidi
A.
Montazer
M.
Sadeghi
M.
2014
Decolorization and mineralization of an azo reactive dye using loaded nano-photocatalysts on spacer fabric: kinetic study and operational factors
.
Journal of the Taiwan Institute of Chemical Engineers
45
,
2436
2446
.
He
Z.
Song
S.
Zhou
H.
Ying
H.
Chen
J. C. I.
2007
Reactive Black 5 decolorization by combined sonolysis and ozonation
.
Ultrasonics Sonochemistry
14
,
298
304
.
Jiang
X.
Wu
Y.
Wang
P.
Li
H.
Dong
W.
2013
Degradation of bisphenol A in aqueous solution by persulfate activated with ferrous ion
.
Environmental Science and Pollution Research
20
,
4947
4953
.
Kasiri
M. B.
Aleboyeh
H.
Aleboyeh
A.
2008
Degradation of Acid Blue 74 using Fe-ZSM5 zeolite as a heterogeneous photo-Fenton catalyst
.
Applied Catalysis B: Environmental
84
,
9
15
.
Kusic
H.
Peternel
I.
Koprivanac
N.
Bozic
A. L.
2011
Iron-activated persulfate oxidation of an azo dye in model wastewater: influence of iron activator type on process optimization
.
Journal of Environmental Engineering
137
,
454
463
.
Li
R.
Jin
X.
Megharaj
M.
Naidu
R.
Chen
Z.
2015b
Heterogeneous Fenton oxidation of 2,4-dichlorophenol using iron-based nanoparticles and persulfate system
.
Chemical Engineering Journal
264
,
587
594
.
Li
Y.
Li
P.
Yu
H.
Bian
Y.
2016
Recent advances (2010–2015) in studies of cerium oxide nanoparticles’ health effects
.
Environmental Toxicology and Pharmacology
44
,
25
29
.
Liang
C.
Huang
C. F.
Mohanty
N.
Kurakalva
R. M.
2008
A rapid spectrophotometric determination of persulfate anion in ISCO
.
Chemosphere
73
,
1540
1543
.
Liang
H. Y.
Zhang
Y. Q.
Huang
S. B.
Hussain
I.
2013
Oxidative degradation of p-chloroaniline by copper oxidate activated persulfate
.
Chemical Engineering Journal
218
,
384
391
.
Masomboon
N.
Ratanatamskul
C.
Lu
M. C.
2009
Chemical oxidation of 2,6-dimethylaniline in the Fenton Process
.
Environmental Science & Technology
43
,
8629
8634
.
Medien
H. A. A.
Khalil
S. M. E.
2010
Kinetics of the oxidative decolorization of some organic dyes utilizing Fenton-like reaction in water
.
Journal of King Saud University – Science
22
,
147
153
.
Monteagudo
J. M.
Durán
A.
González
R.
Expósito
A. J.
2015
In situ chemical oxidation of carbamazepine solutions using persulfate simultaneously activated by heat energy, UV light, Fe2+ ions, and H2O2
.
Applied Catalysis B: Environmental
176–177
,
120
129
.
Notari
R. E.
1987
Biopharmaceutics and Chemical Pharmacokinetics
. 4th edn.
Taylor & Francis
,
New York
.
Peng
Y.
Fu
D.
Liu
R.
Zhang
F.
Liang
X.
2008
NaNO2/FeCl3 catalyzed wet oxidation of the azo dye Acid Orange 7
.
Chemosphere
71
,
990
997
.
Rao
Y. F.
Qu
L.
Yang
H.
Chu
W.
2014
Degradation of carbamazepine by Fe(II)-activated persulfate process
.
Journal of Hazardous Materials
268
,
23
32
.
Rice
E. W.
Bridgewater
L.
,
APHA, AWWA & WEF
2012
Standard Methods for the Examination of Water and Wastewater
.
American Public Health Association/American Water Works Association/Water Environment Federation
,
Washington, DC
.
Solano
A. M. S.
Garcia-Segura
S.
Martínez-Huitle
C. A.
Brillas
E.
2015
Degradation of acidic aqueous solutions of the diazo dye Congo Red by photo-assisted electrochemical processes based on Fenton's reaction chemistry
.
Applied Catalysis B: Environmental
168–169
,
559
571
.
Wetchakun
N.
Chainet
S.
Phanichphant
S.
Wetchakun
K.
2015
Efficient photocatalytic degradation of methylene blue over BiVO4/TiO2 nanocomposites
.
Ceramics International
41
,
5999
6004
.
Xu
X. R.
Li
X. Z.
2010
Degradation of azo dye Orange G in aqueous solutions by persulfate with ferrous ion
.
Separation and Purification Technology
72
,
105
111
.
Yu
Z. Y.
Wang
W. H.
Song
L.
Lu
L. Q.
Wang
Z. Y.
Jiang
X. F.
2013
Acceleration comparison between Fe2+/H2O2 and Co2+/oxone for decolouration of azo dyes in homogeneous systems
.
Chemical Engineering Journal
234
,
475
483
.
Zhang
Q.
Jing
Y. H.
Shiue
A.
Chang
C. T.
Chen
B. Y.
Hsueh
C. C.
2012
Deciphering effects of chemical structure on azo dye decolorization/degradation characteristics: bacterial vs. photocatalytic method
.
Journal of the Taiwan Institute of Chemical Engineers
43
,
760
766
.
Zhao
D.
Liao
X.
Yan
X.
Huling
S. G.
Chai
T.
Tao
H.
2013
Effect and mechanism of persulfate activated by different methods for PAHs removal in soil
.
Journal of Hazardous Materials
254–255
,
228
235
.

Supplementary data