Journal of Environmental Management 155 (2015) 177e183

Contents lists available at ScienceDirect

Journal of Environmental Management

journal homepage: www.elsevier.com/locate/jenvman

Modeling and optimization of photo-Fenton degradation of 2,4-D

using ferrioxalate complex and response surface methodology (RSM)
Agustina V. Schenone a, Leandro O. Conte a, María A. Botta b, Orlando M. Alfano a, *
a
gico para la Industria Química (INTEC), Consejo Nacional de Investigaciones Científicas y T

Instituto de Desarrollo Tecnolo ecnicas (CONICET) and
Universidad Nacional del Litoral (UNL), Ruta Nacional N
168, 3000, Santa Fe, Argentina
b
Facultad de Bioquímica y Ciencias Biolo
gicas, Universidad Nacional del Litoral (UNL), Santa Fe, Argentina

a r t i c l e i n f o a b s t r a c t

Article history: This study reports the application of the photo-Fenton process for the degradation of the herbicide 2,4-
Received 27 November 2014 dichlorophenoxyacetic (2,4-D). The objective of this research was the evaluation of the procedure at
Received in revised form natural pH (pH ¼ 5) using the ferrioxalate complex as iron source at two incident irradiation levels. For
13 March 2015
this purpose, different combinations of attenuation filters from a solar simulator were tested. Since the
Accepted 18 March 2015
Available online 26 March 2015
process depends on several parameters, the influence of the temperature (T) and peroxide to 2,4-D initial
concentration ratio (R) were investigated and optimized by the application of a three-level factorial
experimental design combined with the Response Surface Methodology (RSM). The significance of
Keywords:
Photo-Fenton
models and their coefficients were assessed with the analysis of variance (ANOVA). The found optimal
2,4-D conditions were: T ¼ 50
C and R ¼ 46.3 and T ¼ 41.53
C and R ¼ 41.46, achieving experimental con-
Response surface methodology versions of 91.4 and 95.9% for the low and high radiation levels, respectively. The obtained results are
Ferrioxalate very close to the values predicted by the quadratic models (93.8 and 100.0%). It was concluded that
temperature and hydrogen peroxyde concentration have different influences on the response factor
depending on the incident irradiation level. It was demonstrated that RSM is a good tool for studying the
effects of different variables and their interactions on 2,4-D conversion percentage in the photo-Fenton
process. In addition, solution acute toxicity was also evaluated during the treatments under optimum
conditions, since some degradation by-products of 2,4-D can be more toxic than the parent compound.
For this purpose, the commonly used Microtox® test based on the bacteria Vibrio fischeri was employed.
© 2015 Elsevier Ltd. All rights reserved.

1. Introduction equipment. In Argentina, the triple washing technique is used to

treat empty chemical containers (Allevato and Po
rfido, 2002;
The great extension of the cultivated area in Argentina, mainly Casafe, 2004). The resulting water contains residues of the agro-
soybeans, wheat and maize, has led to an increase in the use of chemical and needs to be treated properly. Furthermore, due to its
pesticides. Glyphosate, 2,4-dichlorophenoxyacetic acid (2,4-D) and relative high solubility in water, 2,4-D was detected in river and
atrazine are the most utilized herbicides. 2,4-D is an auxin-like drainage water (Lagan a et al., 2002). For these reasons, it is
herbicide most commonly used for the control of broadleaf weed necessary to develop methods for the reduction or elimination of
species and is widely applied because of its low cost and high ef- agrochemicals from water resources.
ficiency. However, the knowledge of the toxicity of this compound Biological systems are generally preferred due to their low cost
(EPA, 2005), has given rise to public concern and to the inclusion of and because they are environmentally friendly, but pesticide and
2,4-D in several legislations. Because of their application in crops, herbicide elimination efficiencies are not always possible or satis-
pesticides can enter watercourses from different practices, such as: factory. On the other hand, advanced oxidation processes (AOPs)
wastewater from food processing industries and from plants pro- are effective remediation methods based on the generation of
ducing pesticides, washing water from containers and spraying reactive species, such as hydroxyl radicals (HO), which react
rapidly and with low selectivity with a variety of toxic and
nonbiodegradable compounds, yielding CO2 and inorganic ions
(Pignatello et al., 2006). Among the AOPs, processes using Fenton
* Corresponding author.
E-mail address: alfano@intec.unl.edu.ar (O.M. Alfano). type reagent are relatively inexpensive and the oxidizing agent is

http://dx.doi.org/10.1016/j.jenvman.2015.03.028
0301-4797/© 2015 Elsevier Ltd. All rights reserved.
178 A.V. Schenone et al. / Journal of Environmental Management 155 (2015) 177e183

easily handled. The combination of the Fenton process with (RSM) were proposed to study the influence of two factors (tem-
UVeVisible radiation results in photo-Fenton reaction, which has perature and H2O2 concentration) as well as their interactive in-
several advantages. Some of them are the increase of degradation fluence. Several experimental designs have been used for photo-
rates and the possibility of using alternative energy sources, such as Fenton optimization in different applications (Ay et al., 2009;
solar radiation. Thus, the photo-Fenton process has gained atten- Colombo et al., 2013; Dopar et al., 2011; Prieto-Rodriguez et al.,
tion due to the possibility of reducing the demand of energy (Bauer 2013; Trovo
et al., 2013). Finally, experiments were performed at
and Fallmann, 1997; Bauer et al., 1999; Malato et al., 2002). In the optimized found conditions for each filter combination, in order
addition, this AOP has been successfully applied in order to degrade to evaluate the results obtained with the predicted model.
several residues of agrochemicals present in liquid wastes (Ahmed
et al., 2011; Al Momani et al., 2007; Carra et al., 2013; Hincapie
2. Experimental
Pe
rez et al., 2006; Huston and Pignatello, 1999; Ikehata and
Gamma El-Din, 2006; Janin et al., 2013; Oliveira et al., 2014; 2.1. Chemicals and reagents
Zapata et al., 2010).
Generally, ferric sulfate is employed as iron source in photo- Analytical reagent-grade chemicals and ultrapure water were
Fenton process and the reaction is commonly carried out at pH used. 2,4-dichlorophenoxiacetic acid (C8H6Cl2O3, 98%) was pur-
2.8. In this condition, photo-active iron species predominate. At chased from Aldrich. Potassium ferrioxalate solution was prepared
higher values, the concentration of the catalyst decreases as a according to the methodology described by Murov et al. (1993),
consequence of the precipitation of Fe(III) as hydroxides (Malato using potassium oxalate monohydrate (Carlo Erba, 99.5%) and
et al., 2009; Pignatello et al., 2006). On the other hand, the need FeCl3.6H2O (Merck). Hydrogen peroxide (H2O2, 30%, reagent-
for acidification could be circumvented with the presence of com- grade), sodium carbonate and sodium bicarbonate were obtained
plexing agents. The resulting iron complexes typically have higher from Cicarelli. NaOH (Cicarelli) was employed for pH adjustment
molar absorption coefficients in the UV and visible regions than the and HPLC grade methanol (Sintorgan) was used to stop the Fenton
aqueous complexes (Pignatello et al., 2006). Nogueira et al. (2005) reaction. Acetonitrile and anhydrous acetic acid were purchased
studied the influence of ferrioxalate as iron source on the degra- from Sintorgan and Anedra, respectively.
dation efficiency of several organic compounds. This ferric complex
strongly absorbs up to 550 nm, allowing that a higher portion of 2.2. Experimental procedure
solar radiation be absorbed by the photo-Fenton process. Further-
more, Conte et al. (2014) investigated the degradation of 2,4-D Experiments were performed in a lab-scale flat plate reactor of
employing different iron sources, demonstrating the effectiveness borosilicate glass with external recycle. A solar simulator (Oriel
of the process by employing ferrioxalate complexes under condi- 9600) was used as the irradiation source. The storage tank features
tions of pH close to neutrality. a pH-meter, a thermometer and a liquid sample valve. The exper-
Photo-Fenton treatment is affected by several parameters such imental devise was connected to a thermostatic bath to ensure
as the temperature, iron concentration, hydrogen peroxide con- constant temperature during the reaction and a centrifugal pump
centration and pH of the reaction medium. Moreover, operating to achieve a high recirculation flow rate and good mixing condi-
conditions need to be carefully determined since they are specific tions. The total volume of the treated solution was 3.00 L, while the
for each contaminant. The aim of the present investigation was to irradiated volume of the reactor was 69.94  103 L.
optimize the procedure to obtain the maximum conversion and The next steps were followed to carry out the experimental
mineralization of 2,4-D in aqueous medium. It is known that the runs: first, the temperature was set at the specified working con-
drawback of the traditional photo-Fenton process is the operating dition. Then, the solutions of ferric complex
1
pH at 2.8. Hence, the objective of this research was the evaluation of 0
(CFe 3þ ¼ 0.054 mmol L
0
) and 2,4-D (C2;4D ¼ 0.136 mmol L1) were
the procedure at pH ¼ 5, which is the natural value of 2,4-D active added to the reactor, followed by the pH adjustment to 5 using a
compound at the concentration under study, thus reducing the concentrated NaOH solution. Next, the first sample was taken
operating costs. In order to accomplish this, ferrioxalate complex (defining the reaction time equal to zero) after the addition of the
was employed as iron source. It is a photoactive complex capable of hydrogen peroxide to the storage tank. Finally, the lamp shutter
expanding the usage of the solar spectrum range, which improves was removed to start the run. During the photo-Fenton reaction,
the oxidation efficiency of the process. Furthermore, it allows the samples were withdrawn at different times.
use of iron concentrations below the discharge limit, avoiding its The solar simulator was equipped with an ozone free xenon
subsequent removal. lamp of 150 W, which produces a collimated beam of 0.033 m
Another way of reducing the degradation process costs is to use diameter. In order to simulate various solar conditions, it is feasible
solar technology. As a preliminary step to the study of the system in to couple different filters to the device, such as air, liquid and
a solar pilot plant reactor, an artificial sunlight source was applied attenuation filters, and redirect the beam with a set of reflecting
to analyze the viability of the process, considering the solar radia- mirrors. In the present study, two filter combinations were
tion in Santa Fe city (31
390 S, 60
430 W, 25 m above sea level), analyzed: (a) water filter, air mass filter AM0 and air mass filter
Argentina. Thus, two types of incident irradiation levels of a solar AM1.5 Direct, and (b) water filter, air mass filter AM0 and air mass
simulator were tested by different combinations of attenuation filter AM1 Direct, which will be called F-1.5D and F-1D, respectively.
filters in order to match the output to the solar spectrum. The selection was made regarding the radiation level in a pilot
In addition, the effect of temperature on the homogeneous plant-solar reactor quantified in previous essays, considering the
photo-Fenton degradation was investigated since it is a parameter period of maximum incident solar radiation (noon) for clear sum-
which influences the kinetic of the reaction, generally causing an mer days and standard relative humidity conditions (50e60%), in
increase in the reaction rate at higher temperatures. Furthermore, Santa Fe city, Argentina (Conte et al., 2012).
the hydrogen peroxide concentration effect was taken into account Radiation fluxes were measured with a modular USB spec-
because, according to the reaction temperature, different amounts trometer (USB2000, Ocean Optics). From the experimental results,
of oxidizing agent could be consumed to reach the same level of the local radiation fluxes averaged over the reactor window (qW)
2,4-D conversion. were determined: qW ¼ 3.64  108 E cm2 s1 and
Three level factorial design and response surface methodology qW ¼ 6.58  108 E cm2 s1, for F-1.5D and F-1D, respectively, in
 A.V. Schenone et al. / Journal of Environmental Management 155 (2015) 177e183 179

the wavelength range 305e435 nm. behavior of the system is represented by the following quadratic
equation:
2.3. Analytical determinations
Y ¼ b0 þ b1 X1 þ b2 X2 þ b12 X1 X2 þ b11 X12 þ b22 X22 þ ε (1)
The concentrations of 2,4-D and 2,4-dichlorofenol (2,4-DCP)
were monitored by HPLC (Waters) equipped with a dual UVeVis in which Y is the response (dependent variable), X1 and X2 the
detector. The maximum global errors were 0.16 and 0.09 mg L1, studied factors (R and T, respectively), bi the model coefficients and
respectively (Miller and Miller, 2002). A LC-18 Supelcosil reverse ε the experimental error.
phase column (Supelco) was employed and detection was done at
236 and 280 nm. A mixture of ultrapure water, acetonitrile and 3. Results and discussion
anhydrous acetic acid (49:50:1) was applied as the mobile phase,
with a flow rate of 1.0 mL min1. The concentration of oxalate was All the experiments were performed with the same initial 2,4-D
determined by ion exchange chromatography, using a Water 0
concentration (C2;4D ¼ 0.136 mmol L1, 30 mg L1). The herbicide
equipment and IonPac AS4A-SC analytical column. The eluent was a concentration corresponds to the obtained quantity after the triple
mixture of sodium carbonate and sodium bicarbonate (1:8), flow- washing technique, which is widely used to treat empty chemical
ing at 0.7 mL min1. A modified iodometric technique was applied containers (Allevato and Po
rfido, 2002; Casafe, 2004). The initial
to determine hydrogen peroxide at 350 nm and the content of iron concentration was 0.054 mmol L1 (3 mg L1); therefore, the
ferrous ions was quantified by absorbance measurements of the treated effluent could be discharged to a surface watercourse (Res.
formed complex by means of 1,10-phenanthroline at 510 nm. Total N
1089/82) or redirected to a biological treatment. The pH ¼ 5 was
organic carbon content was measured employing a Shimadzu TOC- selected because it is close to the value of an effluent with
5000A analyzer. 30 mg L1 of 2,4-D. Moreover, when the process is complete, the
samples can be discharged without the need to modify this value,
2.4. Toxicity assay as it is in the case of the processes carried out at pH ¼ 2.8. An ox-
alate/iron molar ratio of 10:1 was chosen in order to avoid iron
Microtox Model 500 Toxicity Analyser (Strategic Diagnostic Inc.) compounds precipitation and to keep them in solution along the
was used in order to carry out the acute bioluminescence assay and photo-Fenton process at pH ¼ 5. Consequently, the initial oxalate
evaluate the toxicity of the samples during the photo-Fenton pro- ion concentration was 0.54 mmol L1. Note that the speciation of
cess. Toxicity was measured as the percentage of inhibition of light the complexes is influenced by the system pH value. Under these
emission of the bacteria strain Vibrio fischeri NRRL-B-11177, after conditions, the dominant ferrioxalate complex is FeIII(C2O4)3 3 ,
5 min of incubation (I %). Sample pH was adjusted between 6 and 8 although FeIII(C2O4) 2 is also present. Compared with the aqueous
and hydrogen peroxide was removed prior to analysis using cata- complexes, these oxalate complexes have higher molar radiation
lase (1500 mg L1 of >2000 U mg1 bovine liver) acquired from absorption coefficients in the UVeVis region (Conte et al., 2014;
Fluka. Pignatello et al., 2006).

2.5. Experimental design

3.1. Process optimization under low radiation level (F-1.5D)
A three level factorial experimental design was employed to
optimize the reaction conditions. The effect of the variables: Previous publications have demonstrated that the photo-Fenton
hydrogen peroxide to 2,4-D initial concentration ratio (R) and process is highly influenced by some operating variables, such as
temperature (T) on the herbicide conversion was assessed by using temperature and hydrogen peroxide to pollutant initial concen-
the RSM (Montgomery, 2005). For two variables and three levels, 12 tration ratio (Benzaquen et al., 2014; Conte et al., 2012, 2014; Farias
experiments were performed including four central points in order et al., 2009). Therefore, the RSM was applied in order to find the
to give statistical consistency to the mathematical model. The optimal conditions that maximizes the 2,4-D conversion. For this
applied ranges of process parameters were: 25e50
C and 7e50 for purpose, two solar simulator filter combinations were studied. In
T and R, respectively. The percentage of 2,4-D conversion after this section, the low radiation level (qW ¼ 3.64  108 E cm2 s1)
180 min of photo-Fenton reaction was chosen as the response will be analyzed.
factor. The same procedure was employed to each of the filter A two factor with three levels design was carried out. The design
combinations: F-1.5D and F-1D, giving a total of 24 experiments. consisted of 12 experiments with independent variables combi-
The temperature range was chosen due to the fact that the po- nations in the following ranges: R between 7 and 50, temperature
tential samples would be at room temperature (25
C) and could from 25 to 50
C, including three central points replicates (Table 1).
reach almost 50
C in the case that the sample were treated in a All experiments were performed in random order to avoid intro-
pilot-plant solar reactor (Conte et al., 2012; Farias et al., 2009). The ducing a bias on the measurements produced by uncontrolled (or
stoichiometric H2O2 to 2,4-D initial concentration ratio required for external) factors.
the 2,4-D complete mineralization is 7. However, Pignatello et al. After removing the outliers (Table 1) by applying the Cook's test
(1992) have demonstrated that better conversion levels are (Myers and Montgomery, 2002), the herbicide conversions after
reached with higher R values in a shorter time. 180 min for the experiments were fitted to a quadratic model (Q1).
The RSM is useful for evaluating the relationship between the Multiple regression analysis was used to calculate the model co-
experimental factors and the measured response according to one efficients and the analysis of variance (ANOVA) with 95% confi-
or more selected criteria (Montgomery, 2005; Myers and dence level to validate them, resulting in the following expression:
Montgomery, 2002). In the 3k design is common to denote low,
2;4D
intermediate and high levels as 1, 0 y þ1, respectively. Each level X180 ¼ 49:771 þ 2:015T þ 3:435R  0:032T  R  0:020R2
of a factor is combined in the different experiments with all levels (2)
of the other factors. Once the results are obtained for the optimi-
2;4D
zation experiments, a mathematical model is created to describe in which X180 is the 2,4-D conversion at 180 min of photo-Fenton
the relationship between the response and the studied factors. The reaction.
180 A.V. Schenone et al. / Journal of Environmental Management 155 (2015) 177e183

Table 1 catalytic activity (see Equation (3)) (Pignatello et al., 2006).

Results of 2,4-D conversion percentage at 180 min. 3k factorial design with 2 inde-
pendent variables for F-1.5D and F-1D.
H2O2 þ OH / H2O þ HO2 (3)
2;4D c
Run T (
C)a,b Rb X180

F1.5D F1D The interaction plot between T and R is presented in Fig. 1. As can
be observed, keeping the temperature at 25
C and varying the
1 35 (0) 7 (1) 36.9 78.3
2 50 (þ1) 50 (þ1) 95.6 100 H2O2/2,4-D ratio, a larger amplitude of 2,4-D conversion is obtained
3 50 (þ1) 7 (1) 63.8 86 (difference of 67% in the 2,4-D conversion) in comparison with the
4 35 (0) 28.5 (0) 72.4 94.1 produced effect when the temperature is set at 50
C (difference of
5 25 (1) 50 (þ1) 83.0 96.5 32% in the 2,4-D conversion). On the other hand, a high operating
6 25 (1) 28.5 (0) 73.0 92.6
7 50 (þ1) 28.5 (0) 83.4 100
temperature would compensate a lower hydrogen peroxide con-
8 35 (0) 28.5 (0) 71.8 93.9 centration, which is important from an economical point of view.
9 35 (0) 28.5 (0) 72.9 94.5 Accordingly, at R ¼ 7 the difference in the conversion is 47% when
10 35 (0) 50 (þ1) 84.8 100 the temperature is varied between 25 and 50
C.
11 25 (1) 7 (1) 16.4 53.7
The overall effect can be observed in Fig. 2A, which shows the
12 35 (0) 28.5 (0) 70.0 94.8
response surface that represents the Q1 model for the photo-
a
The calculated value for the T central point by the design was 37.5
C, but it was Fenton reaction with F-1.5D (Equation (2)). According to the
set at 35
C because of operational issues.
b
Codified values for variables in parentheses.
shape of the surface, the H2O2/2,4-D ratio is more important when
c
Outliers: F1.5D: #2, F1D: #5 and 11. the low radiation level is used in the process at 25
C. Furthermore,
experimental conversion results are in accordance with the model
predicted by applying multiple regression analysis.
The ANOVA results are shown in Table 2. The p value represents The highest herbicide conversion (95.6%) was achieved for run
the probability that F occurred due to noise; the smaller the value of #2 (R ¼ 50 and T ¼ 50
C) after 180 min of reaction. On the other
p, the more significant is the corresponding parameter in the model hand, the minimum conversion was 16.4%, obtained when the
(Montgomery, 2005). According to the high calculated values of F lowest values of each parameter, i.e. T ¼ 25
C and R ¼ 7 (run #11),
(198.41) and low values of p (<0.0001), Q1 model can be charac- were used. At this T, an important increment of the conversion is
terized as statistically significant. As can be appreciated, the model observed when R is increased (83.0% in run #5). The same can be
terms are also significant (p < 0.05), while the lack of fit is non- seen when R is kept at its lowest value and T is raised (63.8% in run
significant (p ¼ 0.0704). T2 term was not taken into account #3), but the percentage of conversion is lower. This is in accordance
because it was non-significant. Furthermore, the model was with the F values gathered by the ANOVA of the model for R and T
selected due to the low standard deviation (2.55) and the R2 and (518.37 and 153.51, respectively), showing the higher influence of
adjusted R2 satisfactory values (0.9925 and 0.9875, respectively). the former in the model predictions.
These results and the information on the model performance given The total organic carbon (TOC) was monitored as well. The
in Supplementary material (Fig. S1) indicate that Q1 properly de- lowest TOC after 180 min of operation was observed in run #2, with
scribes the system behavior in the studied conditions. a final value of 9.86 mg L1, disregarding the contribution of oxa-
The significance and contribution of each process parameter late. Furthermore, 2,4-DCP (principal reaction intermediate) was
was evaluated. For this purpose, F and p values for each linear, detected in this run at t ¼ 30 min. The concentration of this organic
interaction or quadratic model components were compared. The compound also decreased up to 0.62 mg L1 at the end of the
results presented in Table 2 indicate that the most important term reaction.
is R (F ¼ 518.37), followed by the other linear term, T, the interaction The optimal conditions for the highest 2,4-D conversion, pre-
(T  R) and finally the quadratic term corresponding to R2. More- dicted by Q1, are found to be: T ¼ 50
C and R ¼ 46.3. The criteria
over, the signs of these terms in Equation (2) provide physical which were followed for the optimization consisted on maximizing
significance to the obtained results. For instance, larger T values the response and, in the case of R and T, the goal was to maintain
would result in a higher herbicide conversion, demonstrating the these factors in the range of study (7e50 and 25e50
C, respec-
importance of its role in the photo-Fenton reaction. On the other tively). Under these conditions, the herbicide conversion predicted
hand, the influence of the quadratic term is negative due to the fact
that an excess of H2O2 produces a scavenging effect, reducing the

Table 2
ANOVA results for Q1 model.

Factors Statistics

SSa dfb MSc F pd

Model 5175.43 4 1293.86 198.41 <0.0001

T 1001.07 1 1001.07 153.51 <0.0001
R 3380.35 1 3380.35 518.37 <0.0001
TR 295.81 1 295.81 45.36 0.0005
R2 229.67 1 229.67 35.22 0.0010
Residual 39.13 6 6.52
Lack of fit 34.32 3 11.44 7.14 0.0704
Pure error 4.81 3 1.6
Total 5214.56 10
a
Sum of squares.
b
Degree of freedom.
c
Mean square.
d
Considered significant when p < 0.05. Fig. 1. Interaction plot between T and R for Q1 model.
 A.V. Schenone et al. / Journal of Environmental Management 155 (2015) 177e183 181

Fig. 2. Response surface and experimental values (circles) for 2,4-D conversions after 180 min versus reaction temperature and H2O2/2,4-D inicial molar ratio. (A) F-1.5D and (B) F-
1D.

by the model was 93.8%. In order to corroborate this value, an term, lineal T term and finally, the interaction between the two
additional experiment was carried out under optimal conditions. studied variable terms. All of them have a p value lower than 0.05
Results are shown in a subsequent section of the paper. and a high F value. The impact of each independent variable in the
oxidation process is attributable directly to its coefficient. From
Equation (4), the positive contribution of T and R is observed, while
3.2. Process optimization under high radiation level (F-1D) the negative coefficient value of R2 suggests that higher H2O2
concentrations reduce the herbicide conversion, such as in Q1
In order to evaluate the influence of two variables (T and R) and model. Fig. S2 in Supplementary Material also demonstrates the
to optimize 2,4-D conversion through the application of the photo- good performance of the Q2 model.
Fenton process with the F-1D filter combination Predicted and experimental results for photo-Fenton reaction
(qW ¼ 3.64  108 E cm2 s1), the same methodology employed using the high radiation level is presented in Fig. 2B, which shows
above for F-1.5D was used. The application of multiple regression the 3D plot of 2,4-D conversion at t ¼ 180 min as a function of R and
analysis on the experimental conditions and the chosen response T. Notice that higher conversions are obtained in comparison with
described in Table 1, resulted in the quadratic model Q2: those achieved with the F-1.5D filter combination. Moreover, the
influence of R in Q2 model is considered moderate, meanwhile the
2;4D
X180 ¼ þ48:163 þ 0:628T þ 1:641R  0:012T  R  0:013R2 effect of T is lower.
(4) Fig. 3 displays the interaction plot between T and R. In this case,
when R is kept constant at 50, the change in temperature does not
The ANOVA indicated that Q2 model is significant (p < 0.0001), modify the 2,4-D conversion significantly. On the other hand, at the
while the lack of fit is non-significant (p ¼ 0.0716). The results are lowest R value (R ¼ 7), a moderate effect is observed when the
displayed in Table 3. temperature is varied, increasing the conversion when going from
In addition, high regression coefficients were obtained: 25 to 50
C (difference of 32% in the 2,4-D conversion).
R2 ¼ 0.9932 and adjusted R2 ¼ 0.9878, as well as a low standard In this case, almost the total 2,4-D conversion is achieved in
deviation value: 0.75. As in the previous section, the term T2 three experiments (runs #2, 7 and 10 from Table 1). Furthermore,
resulted non-significant; thus, it was not taken into account in the nearly complete conversion of the main reaction intermediate, 2,4-
final model.
Given the results summarized in Table 3, the terms of the model
are significant in the following order: lineal R term, quadratic R

Table 3
ANOVA results for Q2 model.

Factors Statistics

SSa dfb MSc F pd

Model 415.17 4 103.79 183.50 <0.0001

T 45.06 1 45.06 79.67 0.0003
R 288.58 1 288.58 510.19 <0.0001
TR 14.82 1 14.82 26.20 0.0037
R2 68.09 1 68.09 120.39 0.0001
Residual 2.83 5 0.57
Lack of fit 2.34 2 1.17 7.20 0.0716
Pure error 0.49 3 0.16
Total 418.00 9
a
Sum of squares.
b
Degree of freedom.
c
Mean square.
d
Considered significant when p < 0.05. Fig. 3. Interaction plot between T and R for Q2 model.
182 A.V. Schenone et al. / Journal of Environmental Management 155 (2015) 177e183

DCP, was observed when the operating conditions were R ¼ 50 and F-1.5D and F-1D, during photo-Fenton experiments.
T ¼ 50
C (data not shown). In addition, TOC values at t ¼ 180 min Comparing the results of both processes, once again it can be
were 6.88 and 6.84 mg L1, disregarding the contribution of oxalate, seen that F-1D is more effective for the treatment of aqueous
under the operating conditions R ¼ 50 and T ¼ 50
C and R ¼ 28.5 samples with 2,4-D, which is in part due to the difference in the
and T ¼ 50
C, respectively. These TOC values were the lowest for all local radiation fluxes averaged over the reactor window. Further-
the experiments. more, lower values of T and R were obtained as optimum in the
The experimental conditions corresponding to the maximum higher radiation level, confirming that the influence of the pa-
2,4-D conversion predicted by Q2, are found to be T ¼ 41.53
C and rameters T and R is more pronounced in the case of F-1.5D.
R ¼ 41.46. As in the previous case for Q1, the criteria were to The results showed that experimental design strategy and
maximize the herbicide conversion as well as to maintain T and R in response surface methodology can be successfully employed to
2;4D
the evaluated range. The predicted result (X180 ¼ 100.0%) is model the relationship between photo-Fenton variables and 2,4-D
checked and the results are shown in the following section. conversion. Furthermore, the optimum parameter values were
found and corroborated.
Because of the possibility of transient formation of more toxic
3.3. Herbicide conversion under optimized conditions
pollutants during AOP application, a toxicity test was carried out. It
is based on the inhibition of bioluminescence of V. fischeri bacteria
The predicted values obtained in the multivariate optimization
strains after 5 min of incubation.
were confirmed for each filter combination. As previously
Despite the nearly complete conversion of 2,4-D, a percentage of
explained, the aim was to maximize the herbicide conversion as
inhibition remained in solution at t ¼ 180 min in both experiments,
well as to maintain T and R in the evaluated range. Accordingly, 2,4-
being I % ¼ 60 and 47% for F1.5D and F-1D, respectively. This fact
D experimental conversions of 91.4 and 95.9% were obtained for F-
indicates the existence of other toxic intermediates.
1.5D and F-1D, respectively. Both results are very close to the values
EC50 values (concentration which causes 50% reduction of the
predicted by Q1 and Q2 models (93.8 and 100.0%), demonstrating
bioluminescence) can be found in the literature. 2,4-D is appointed
satisfactory performances. Fig. 4 shows the evolution of 2,4-D, 2,4-
as moderately toxic (EC50 ¼ 59 mg L1) (Zona and Solar, 2003) and
DCP and H2O2 as a function of time for both optimized conditions,
2,4-DCP as toxic (EC50 ¼ 1.2e6.1 mg L1) (Kaiser and Palabrica,
1991). These values are in conformity with the bacteria growth
inhibition. However, the observed concentration of 2,4-DCP in so-
lutions is low to be the only reason of the total toxicity at
t ¼ 180 min. Therefore, residual toxicity could be also attributed to
other compounds present in solution along with 2,4-DCP. Thus,
formation of other intermediates during the herbicide degradation,
such as chlorohydroquinones or chlorobenzoquinones, should
focus larger attention. Fig. 5 shows a simplified degradation
pathway of 2,4-D, taking into account the intermediates identified
by Kwan and Chu (2004).
Finally, further work is under way in order to develop a kinetic
model based on a detailed reaction scheme to study the photo-
Fenton degradation of 2,4-D at natural pH, using the ferrioxalate
complex as iron source. Moreover, the system will be analyzed in a
pilot plant solar reactor in order to better assess the applicability of
this methodology.

4. Conclusions

The photo-Fenton degradation of the herbicide 2,4-D at pH ¼ 5,

using the ferrioxalate complex as iron source and a solar simulator
with different attenuation filters (F-1.5D and F-1D) was optimized.
With the aim of studying the effects of the process parameters:
temperature (T) and hydrogen peroxide to 2,4-D initial concentra-
tion ratio (R), a three-level factorial experimental design combined
with response surface methodology was applied. It was shown that
the investigated parameters strongly influence the herbicide con-
version percentage under the studied conditions. It was also found
that R over 50 should be avoided, since it causes a decrease in the
process efficiency.
The developed models, Q1 and Q2, are found to be highly

Fig. 4. Experimental relative concentrations versus time for 2,4-D (circles), 2,4-DCP
(squares) and hydrogen peroxide (diamonds) at the optimum conditions for (A) F- Fig. 5. Proposed pathway for 2,4-D degradation (CHQ: chlorohydroquinones, CBQ:
1.5D and (B) F-1D. chlorobenzoquinones).
 A.V. Schenone et al. / Journal of Environmental Management 155 (2015) 177e183 183

accurate and predictive, according to the obtained statistical pa- efficiencies for the homogeneous photo-Fenton degradation of herbicide 2,4-D
using different iron complexes. J. Chem. Technol. Biotechnol. http://dx.doi.org/
rameters. The conversion achieved by the F-1D filter at the optimal
10.1002/jctb.4284.
2;4D
conditions (T ¼ 41.53
C and R ¼ 41.46, X180 ¼ 95.9%) was higher Dopar, M., Kusic, H., Koprivanac, N., 2011. Treatment of simulated industrial
than that of F-1.5D filter at the optimal conditions (T ¼ 50
C and wastewater by photo-Fenton process. Part I: the optimization of process pa-
2;4D rameters using design of experiments (DOE). Chem. Eng. J. 173, 267e279.
R ¼ 46.3, X180 ¼ 91.4%). The obtained results are very close to the
Farias, J., Albizzatti, E.D., Alfano, O.M., 2009. Kinetic study of the photo-Fenton
values predicted by the quadratic models (93.8 and 100.0%, degradation of formic acid. Combined effects of temperature and iron con-
respectively). In addition, lower values of T and R were obtained as centration. Catal. Today 144, 117e123.
Hincapie
Pe
rez, M., Pen ~ uela, G., Maldonado, M.I., Malato, O., Fen

n
andez-Iba ~ ez, P.,
optimum with the higher radiation level, confirming that the in-
Oller, I., Gernjak, W., Malato, S., 2006. Degradation of pesticides in water using
fluence of the parameters T and R is more pronounced in the case of solar advanced oxidation proceses. Appl. Catal. B Environ. 64, 272e281.
F-1.5D. Huston, P.L., Pignatello, J.J., 1999. Degradation of selected pesticide active in-
The obtained mathematical models were able to explain the gredients and commercial formulations in water by the photo-assisted Fenton
reaction. Water Res. 33, 1238e1246.
influence of the parameters under study in the degradation of 2,4-D Ikehata, K., Gamal El-Din, M., 2006. Aqueous pesticide degradation by hydrogen
by the photo-Fenton process at natural pH. peroxide/ultraviolet irradiation and Fenton-type advanced oxidation processes:
A significant percentage of inhibition remained in solution at a review. J. Environ. Eng. Sci. 5, 81e135.
Janin, T., Goetz, V., Brosillon, S., Plantard, G., 2013. Solar photocatalytic minerali-
t ¼ 180 min in both experiments at the optimal conditions despite zation of 2,4-dichlorophenol and mixtures of pesticides: kinetic model of
the nearly complete conversion of 2,4-D and the low 2,4-DCP mineralization. Sol. Energy 87, 127e135.
concentration, indicating the existence of other toxic compounds Kaiser, K.L.E., Palabrica, V.S., 1991. Photobacterian phosphoreum. Toxicity data In-
dex. Water Pollut. Res. J. Can. 26, 361e431.
along with the principal intermediate.
Kwan, C.Y., Chu, W., 2004. A study of the reaction mechanisms of the degradation of
2,4-dichlorophenoxyacetic acid by oxalate-mediated photooxidation. Water
Acknowledgments Res. 38, 4213e4221.
Lagana , A., Bacaloni, A., Leva, I.D., Faberi, A., Fago, G., Marino, A., 2002. Occurrence
and determination of herbicides and their major transformation products in
The authors are grateful to Universidad Nacional del Litoral (UNL environmental waters. Anal. Chim. Acta 462, 187e198.
e PI 85-192), Consejo Nacional de Investigaciones Científicas y Malato, S., Blanco, J., Vidal, A., Richter, C., 2002. Photocatalysis with solar energy at a

cnicas (CONICET e PIP 112-201101-00163), and Agencia Nacional
Te pilot-plant scale: an overview. Appl. Catal. B Environ. 37, 1e15.
Malato, S., Ferna
ndez-Iban ~ ez, P., Maldonado, M.I., Blanco, J., Gernjak, W., 2009.
de Promocio
n Científica y Tecnolo

gica (ANPCyT e PICT-2010-1434) Decontamination and disinfection of water by solar photocatalysis: recent
for financial support. overview and trends. Catal. Today 147, 1e59.
Miller, J.N., Miller, J.C., 2002. Statistics and Chemometrics for Analytical Chemistry.
Pearson-Prentice Hall, New York, USA.
Appendix A. Supplementary data Montgomery, D.C., 2005. Desing and Analysis of Experiments. John Wiley and Sons,
New York, USA.
Supplementary data related to this article can be found at http:// Murov, S., Carmichael, I., Hug, G., 1993. Handbook of Photochemistry, second ed.
Marcel Dekker, New York, pp. 299e305.
dx.doi.org/10.1016/j.jenvman.2015.03.028. Myers, R.H., Montgomery, D.C., 2002. Response Surface Methodology: Process and
Product Optimization Using Designed Experiments, second ed. John Wiley &
References Sons, New York, USA.
Nogueira, R.F.P., Silva, M.R.A., Trovo, A.G., 2005. Influence of the iron source on the
solar photo-Fenton degradation of different classes of organic compounds. Sol.
Ahmed, B., Limem, E., Abdel-Wahab, A., Bensalah, N., 2011. Photo-Fenton treatment
Energy 79, 384e392.
of actual agro-industrial wastewaters. Ind. Eng. Chem. Res. 50, 6673e6680.
Oliveira, C., Alves, A., Madeira, L.M., 2014. Treatment of water networks (waters and
Al Momani, F.A., Shawaqfeh, A.T., Shawaqfeh, M.S., 2007. Solar wastewater treat-
deposits) contaminated with chlorfenvinphos by oxidation with Fenton's re-
ment plant for aqueous solution of pesticide. Sol. Energy 81, 1213e1218.

rfido, D., 2002. Report available on: www.msal.gov.ar/agroquimicos/ agent. Chem. Eng. J. 241, 190e199.
Allevato, H., Po
Pignatello, J.J., 1992. Dark and photoassisted Fe3þ catalyzed degradation of chlor-
pdf/EnvaAgro_CEPIS-OPS.pdf (accessed October 2014).
ophenoxy herbicides by hydroxen peroxy. Environ. Sci. Technol. 26, 944e951.
Ay, F., Catalkaya, E.C., Kargi, F.A., 2009. Statistical Experiment design approach for
Pignatello, J.J., Oliveros, E., MacKay, A., 2006. Advanced oxidation processes for
advanced oxidation of direct Red Azo Dye by photo-Fenton treatment. J. Hazard.
organic contaminant destruction based on the Fenton reaction and related
Mater. 162, 230e236.
chemistry. Crit. Rev. Env. Sci. Technol. 36, 1e84.
Bauer, R., Fallmann, H., 1997. The photo-fenton oxidation e a cheap and efficient
Prieto-Rodriguez, L., Spasiano, D., Oller, I., Fernandez-Calderero, I., Aguera, A.,
wastewater treatment method. Res. Chem. Intermed. 23, 341e354.
Malato, S., 2013. Solar photo-Fenton optimization for the treatment of MWTP
Bauer, R., Waldner, G., Fallmann, H., Hager, S., Klare, M., Krutzler, T., Malato, S.,
effluents containing emerging contaminants. Catal. Today 209, 188e194.
Maletzky, P., 1999. The photo-fenton reaction and the TiO2/UV process for
Resolution N
1089/82. Santa Fe, Argentina. Available on http://www.santafe.gov.ar/
waste water treatment e novel developments. Catal. Today 53, 131e144.
index.php/web/content/view/full/28842/(subtema)/112857 (accessed October
Benzaquen, T.B., Isla, M.A., Alfano, O.M., 2014. Fenton and photo-Fenton processes
2014).
for the degradation of atrazine: a kinetic study. J. Chem. Technol. Biotechnol.
Trovo
, A.G., Silva, T.F.S., Gomes Jr., O., Machado, A.E.H., Neto, W.B., Muller Jr., P.S.,
http://dx.doi.org/10.1002/jctb.4324.
Daniel, D., 2013. Degradation of caffeine by photo-Fenton process: optimization
CASAFE. Program AgroLimpio, 2004. Available on: http://www.casafe.org/quienes-
of treatment conditions using experimental design. Chemosphere 90, 170e175.
somos/agrolimpio/ (accessed October 2014).

pez, J.L., Santos-Juanes, L., Malato, S., S

rez, J.A., 2013. Iron United States Prevention Agency (EPA), 2005. Prevention, Pesticides and Toxic
Carra, I., Casas Lo anchez Pe
Substances (7508C). Reregistration Eligibility Decision for 2,4-D. http://www.
dosage as a strategy to operate the photo-Fenton process at initial neutral pH.
epa.gov/oppsrrd1/REDs/24d_red.pdf (accessed October 2014).
Chem. Eng. J. 224, 67e74.
Zapata, A., Oller, I., Rizzo, L., Hilgert, S., Maldonado, M.I., S
anchez-Pe
rez, J.A.,
Colombo, R., Ferreira, T.C.R., Alves, S.A., Carneiro, R.L., Lanza, M.R.V., 2013. Appli-
Malato, S., 2010. Evaluation of operating parameters involved in solar photo-
cation of the response surface and desirability design to the Lambda-
Fenton treatment of wastewater: interdependence of initial pollutant concen-
cyhalothrin degradation using photo-Fenton reaction. J. Environ. Manag. 118,
tration, temperature and iron concentration. Appl. Catal. B Environ. 97,
32e39.
292e298.
Conte, L.O., Farias, J., Albizzati, E.D., Alfano, O.M., 2012. Photo-Fenton degradation of
Zona, R., Solar, S., 2003. Oxidation of 2,4-dichlorophenoxyacetic acid by ionizing
the herbicide 2,4-dichlorophenoxyacetic acid in laboratory and solar pilot-plant
radiation: degradation, detoxification and mineralization. Radiat. Phys. Chem.
reactors. Ind. Eng. Chem. Res. 51, 4181e4191.
66, 137e143.
Conte, L.O., Querini, P., Albizzati, E.D., y Alfano, O.M., 2014. Photonic and quantum