Photocatalytic degradation of azo dye Metanil Yellow: Optimization and kinetic modeling using a chemometric approach Mohamad Sleiman a,b,*, Daniel Vildozo a,b, Corinne Ferronato a,b, Jean-Marc Chovelon a,b a

Institut de recherches sur la catalyse et l’environnement de Lyon (IRCELYON), UMR CNRS 5256, 2 Avenue Albert Einstein, F-69626 Villeurbanne Cedex, France b Universite´ Lyon 1, 43 Bd du 11 Novembre 1918, F-69622 Villeurbanne Cedex, France Received 7 May 2007; received in revised form 17 June 2007; accepted 21 June 2007 Available online 28 June 2007

Abstract The photocatalytic degradation of an azo dye Metanil Yellow was carried out in aqueous solution using TiO2 as photocatalyst under UV irradiation. The decolorization and degradation kinetics were investigated and both followed a pseudo first order kinetic according to Langmuir– Hinshelwood model. Using HPLC/DAD and GC/MS analyses, more than 10 major reaction intermediates were identified and a tentative degradation pathway was proposed. Furthermore, ion chromatography (IC) and TOC measurements revealed a complete mineralization of Metanil Yellow into CO2, N2, H2O and inorganic ions (NH4+, NO3 and SO42). On the other hand, an experimental design based on the surface response methodology was applied to assess the individual and interaction effects of several operating parameters (dye concentration, TiO2 concentration, pH, light flux, etc.) on the treatment efficiency (dye removal time). Based on the experimental design data, a semi-empirical expression was obtained, permitting to predict and optimize the dye removal time. This model was very consistent with experiment results (correlation factor: 99.5%). Moreover, additional experimental results obtained under near optimal conditions were found to be very close to the predicted values. This work demonstrates well the utility and benefits of the experimental design approach for screening and modeling the reaction parameters. Furthermore, it contributes significantly to the improvement and better understanding of photocatalytic processes. # 2007 Elsevier B.V. All rights reserved. Keywords: Azo dye; Metanil Yellow; TiO2 photocatalysis; Experimental design; Optimization

1. Introduction Azo dyes constitute an important class of synthetic, colored, organic compounds, which are characterized by the presence of one or more azo bonds (–N N–). They represent about 50% of the worldwide dye production and are widely used in a number of industries, such as textile dyeing, food, cosmetics, paper printing, with the textile industry as the largest consumer [1,2]. It is reported that, during manufacturing or processing operations, dye loss in wastewaters could vary from 2% for basic dyes to as high as 50% for reactive dyes [3]. This leads to

* Corresponding author at: Institut de recherches sur la catalyse et l’environnement de Lyon (IRCELYON), UMR CNRS 5256, 2 Avenue Albert Einstein, F-69626 Villeurbanne Cedex, France. Tel.: +33 4 72 43 11 50; fax: +33 4 72 44 84 38. E-mail address: [email protected] (M. Sleiman). 0926-3373/$ – see front matter # 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.apcatb.2007.06.015

severe contamination of surface and ground water as many of these compounds are not readily biodegradable, suspected to be carcinogenic and can produce toxic aromatic amines [4–7]. Adequate treatment of these pollutants is therefore of primary concern. A number of physical and chemical techniques has been reported for the removal of dye compounds such as adsorption on activated carbon [8], biodegradation [9], ozonation [10], and advanced oxidation processes (AOPs) such as Fenton and photo-Fenton catalytic reactions [11,12], H2O2/UV processes [13] and TiO2 photocatalysis [14–16]. Among the AOPs, heterogeneous photocatalysis oxidation using TiO2 as photocatalyst has been extensively studied, and proved to be efficient and potentially advantageous, as it leads to a complete and fast mineralization of a wide range of dyes to CO2, water and inorganic ions [17–19]. The TiO2 mediated photocatalysis technique depends on various parameters that can modify the degradation process

such as TiO2 concentration, light flux, initial substrate concentration, pH, oxygen concentration, temperature and photoreactor dimensions. Most of studies dealing with treatment of dyes by photocatalysis report the investigation of some of these operational parameters using a univariate approach [20–22]. In this case, one parameter is varied each time while keeping constant the other variables. However, this may be misleading since it implies a partial exploration of the experimental field and it ignores the possible interactions between the variables. Moreover, it is time consuming and does not necessarily permit an effective optimization of the process. An alternative approach allowing to overcome these problems is the experimental statistical designs. Recently, this methodology, known also as multivariate analysis, has been successfully applied for the investigation of operational parameters of photocatalytic treatment of Orange II dye [23] and Reactive Blue 19 [24]. The aim of this work was to assess the photocatalytic treatment of a mono azo dye Metanil Yellow (see structure, Fig. 1), selected due to its toxicity and carcinogenic nature as well as its presence in wastewater of several industrial such as textile, tannery, soap, cosmetics, polishes, etc. Two main aspects were studied: The first one was the optimization of the operational conditions for the removal of Metanil Yellow by means of a chemometric approach based upon a surface response modelling (RSM). All the possible significant factors were investigated such as dye concentration, pH, TiO2 concentration, light flux, temperature and oxygen concentration. The second aspect was the identification of possible degradation products as well as the evaluation of the total mineralization during the process. For that, powerful analytical techniques such as liquid chromatography coupled to diode array detector (HPLC/DAD), gas chromatography coupled to mass spectrometry (GC/MS) and ion chromatography were employed. 2. Materials and methods 2.1. Chemicals Reactive Metanil Yellow (98% purity) was purchased from Sigma–Aldrich and used as received. Titanium dioxide Degussa P25 was provided by Degussa (Frankfurt, Germany) with a specific BET area of 50 m2 g1 and a mean particle size of 30 nm. Nylon filters (0.45 mm) were purchased from Millipore. Methanol (HPLC grade) was purchased from SDS (Peypin, France). Other reagents were at least of analytical grade.

Fig. 1. Chemical structure of azo dye Metanil Yellow.

2.2. Photoreactor and light source The irradiation experiments were carried out in an open borosilicate (Pyrex) glass cell (cut-off at 295 nm, 4 cm diameter, 9 cm height) equipped with a magnetic stirring bar and water circulating jacket. The light source was a HPK 125 W Philips mercury lamp with main emission wavelength at 365 nm, cooled with a water circulation. The light intensity was regulated by including calibrated grids between the lamp and the reactor. The radiant flux entering the irradiation cell was measured by uranyloxalate actinometry. For all experiences, before and during irradiation, the suspensions were magnetically stirred and cell temperature was adjusted to the required value. 2.3. Photocatalytic degradation procedure Stock solutions of Metanil Yellow were prepared using Millipore Milli-Q deionized water and diluted according to the working conditions (40–100 mg L1). The required pH was adjusted by adding diluted aqueous solution of HCl or NaOH. A 25 mL of dye solution was then taken into the photoreactor and required amount of TiO2 powder was added according to the experimental design (0.5, 1.5 or 2.5 g L1). Before irradiation was turned on, the suspension was stirred for 30 min in the dark to reach the adsorption–desorption quasi-equilibrium. One first sample was taken out at the end of the dark adsorption period just before turning on the irradiation, in order to determine the bulk dye concentration. This value was taken as the initial concentration for the photocatalytic experiment, denoted hereafter as Ceq (in addition to C0, the initial bulk concentration of dye before dark adsorption). During irradiation, samples were withdrawn regularly from the reactor and filtered immediately through 0.45 mm Nylon membrane filters to remove TiO2 particles. For some experiments, oxygen was bubbled to the reactor at a regulated pressure in order to assess its effect on the process, according to the experimental design. 2.4. Chemical analysis 2.4.1. UV–vis spectroscopic analysis The decolorization kinetics of Metanil Yellow irradiated solutions were followed using a UVIKON 930 UV–vis spectrophotometer (double monochromator and double beam optical system). For each irradiation experiment, one sample (3 mL) was taken from the reactor and divided into two vials and the final absorbance value was determined by calculating the average over the two measurements. The absorption spectrum of each sample was recorded over the range 200– 700 nm and the absorbance at the lmax (434 nm) was registered. 2.4.2. HPLC/DAD The concentration profile of Metanil Yellow during irradiation and the formation of organic intermediates were

monitored using Shimadzu VP series HPLC system consisting of LC-10AT binary pump, a SPD-M10A DAD and Shimadzu Class-VP software (Version 5.0). Analytical separation was performed using a column Hypersil BDS C18 (5 mm, 125 mm long  4 mm i.d.), mobile phase, 45:55 (v/v) methanol/water (pH set to 3.0 with phosphoric acid) and flow rate of 1 mL min1. The detection wavelength was 434 nm and injection volume, 20 mL.

2.4.6. Experimental design data analysis The chemometric approach was performed using a central composite design face-centered (CCF). Analysis of the experimental data was supported by the statistical graphics software system STATGRAPHICS Centurion Version XV (STSC, Rockville, MD, USA).

2.4.3. SPE–GC/MS The separation and identification of the intermediate products formed during the photocatalytic degradation of Metanil Yellow solutions were performed by using solid-phase extraction (SPE) method followed by gas chromatography– mass spectrometry (GC–MS). By these means, a mixture of Metanil Yellow solutions irradiated at different time intervals was extracted using 6 mL cartridges packed with 200 mg of Isolute ENV+ (International Sorbent Technology, Cambridge, UK). Before extraction, the pH of samples mixture was adjusted to pH 2 using sulfuric acid. The SPE cartridges were conditioned with 6 mL of methanol followed by 6 mL of acidified water (pH 2) at a flow rate of 6 mL min1 using a Varian vaccum manifold. Subsequently 50 mL of sample were passed through the cartridge at a flow rate of 3 mL min1. The analytes were eluted using two aliquots of 1 mL of methanol at a flow rate of 1 mL min1. GC–MS analyses were performed on a Perkin-Elmer (PE) Clarus 500 instrument, equipped with Elite 5 MS column (5% diphenyl–95% dimethylsiloxane) of 60 m length, 0.25 mm i.d. and 1 mm film thickness. Separation of byproducts was conducted under the following chromatographic conditions: injector temperature: 250 8C; oven temperature program: 60 8C held for 2 min, ramped at 15 8C min1 to 270 8C (held for 30 min). Helium was used as carrier gas at a flow of 1.6 mL min1 and sample injections of 1 mL were accomplished in split mode. The temperatures of ion source and interface were set at 250 8C. The MS operated in electron ionization mode with a potential of 70 eV and the spectra were obtained at a scan range from m/z 33 to 500 with a scan time of 0.3 s and interscan of 0.01 s.

3.1. Preliminary experiments: adsorption, hydrolysis, photolysis and photocatalysis of Metanil Yellow

3. Results and discussion

Preliminary experiments were carried out, before the development of experimental design, to evaluate the extent of hydrolysis, adsorption and photolysis processes on the Metanil Yellow transformation. Fig. 2 illustrates the kinetic evolution of Metanil Yellow concentration (%) under the following conditions: (i) in the dark (hydrolysis), (ii) UV irradiation without TiO2 (photolysis), (iii) in the dark with TiO2 (adsorption) and (iv) UV irradiation with TiO2 (photocatalysis). As can be seen, the concentration of Metanil Yellow remained unchanged after 2 h in dark without TiO2. Similarly, irradiation in absence of TiO2 showed no significant photodegradation, because after 90 min, direct photolysis caused only 5% decrease in the concentration of Metanil Yellow. This demonstrates that the photocatalytic experiments occurred in a pure photocatalytic regime where hydrolysis and photochemical processes can be neglected. On the other hand, in the dark with the presence of TiO2, a rapid and an important decrease of the dye concentration (about 50%) was observed indicating a strong adsorption of the dye on TiO2 surface. This result could be explained by the high electrostatic interaction between the dye molecules negatively charged and the surface of TiO2 which exists mainly in the positive form TiOH2+ at the pH of suspension (pH 5.5 < pHpzc: 6.25). It should be pointed out that at lower pH, particularly at

2.4.4. Ion chromatography A Dionex DX-120 instrument has been employed equipped with a Dionex AS 40 autosampler and a conductimeter detector. The determination of ammonium ions has been performed by adopting a CS12 A (4 mm  250 mm) column and 20 mM sulfuric acid as eluent, flow rate of 1 mL min1. The anions have been analyzed by using a AS14A (4 mm  250 mm) anion exchange column and a mixture of Na2CO3/NaHCO3 (8 mM/1 mM) at a flow rate 1 mL min1. 2.4.5. Total organic carbon analysis In order to determine the extent of mineralization, total organic carbon (TOC) measurements were performed on filtered suspensions samples using a Bioritech TOC analyzer.

Fig. 2. Comparison of hydrolysis, photolysis, adsorption and UV/TiO2 photocatalysis kinetics of Metanil Yellow. [TiO2] = 2 g L1, [MY] = 40 mg L1, pH 5.5  0.1.

pH < 3, a quasi-total adsorption of the initial dye concentration (40 mg L1) occurred along with a color change from yellow to red (pKa dye: 2.3). As consequence, all photocatalytic experiments were carried out in the pH range 4–8 to avoid these problems. 3.2. Decolorization and degradation kinetics Fig. 3 shows the time course of dye concentration and sample absorbance at 434 nm during irradiation of Metanil Yellow solution (40 mg L1) using 2.5 g L1 TiO2 concentration and a light flux of 4  1016 photons s1 cm2. As it can be observed, dye disappearance is faster than the sample decolorization, which means that the degradation of Metanil Yellow produces some colored compounds that took longer time to be removed. This was further confirmed by characterizing several colored intermediates using HPLC/ DAD. On the other hand, both process profiles followed an apparent pseudo-first order kinetic (r = kapp C), with a good correlation (kdec: 0.025 min1, R2: 0.996; kdeg: 0.073 min1, R2: 0.993). This behavior could be in agreement with a generally observed Langmuir–Hinshelwood (LH) kinetic model (Eq. (1)) when the concentration is low enough and no catalyst saturation occurs. However, several experimental observations have been reported in previous studies which call into question the validity of the LH model in interpreting the results of heterogeneous photocatalysis [25]. Furthermore, the present kinetic data are not sufficient to conclude that the LH mechanism is the most suitable one to describe the photocatalytic process of Metanil Yellow. r¼

K LH C eq dC ¼ kLH u ¼ kLH dt 1 þ K LH C eq

(1)

in the literature [15,19,26], the photocatalytic process can be affected by a large number of parameters such as TiO2 loading (1), initial substrate concentration (2), pH (3), light intensity (4), oxygen concentration (5), temperature (6), etc. However, it is too difficult to realize an experimental design including all these factors, because of the excessive number of experiments required and the complexity of data analysis. Therefore, preliminary experiments were conducted to detect the variables having the most significant impact on the reaction rate of Metanil Yellow. Among the six parameters investigated (1–6) it was found that the main affecting factors were TiO2 concentration, light flux (F), initial dye concentration and pH. The remaining parameters such as O2 and temperature were found to be less important and therefore were not considered as variables in the central composite design. Their values were kept constant during all the experiments (temperature 20 8C, constant O2 concentration 8 mg L1). The experimental measured response (y) was the dye removal time which corresponds to the disappearance of the initial substrate. This variable was selected for two reasons: the first is its fundamental and application importance, as one of the main goals of the photocatalytic treatment is to eliminate the initial toxic substrate as fast as possible. One can say that the degradation rate constant could be an alternative way; nevertheless, kapp is not an experimental measured value but a calculated one using a specific kinetic model (e.g. first order kinetic) which might change with the experimental conditions. Hence, the error can be greater. Moreover, the degradation rate could be sometimes a misleading parameter because it depends mainly on the reactant concentration used and therefore a high reaction rate does not mean certainly that the reaction time will be short. The second reason is that the removal time is not strongly related to the formation of reaction intermediates comparing to other variables used in previous recent studies such as decolorization efficiency (%) or TOC removal.

3.3. Experimental design methodology 3.3.1. Choice of factors and response The best performance of an experimental design depends on some knowledge about the system to being studied. As reported

3.3.2. Central composite design A central composite design face-centered (CCF) was employed in order to optimize the dye removal time. Four factors were considered: TiO2 concentration, light flux (F), initial dye concentration and pH. Table 1 summarizes the levels for each factor involved in the design strategy. These levels were chosen according to previous experiments carried out at our lab and also from data in the literature for similar laboratory experiments [26,27]. The design consists of three series of experiments: (i) a twolevel full factorial design 24 (all possible combinations of codified values +1 and 1); (ii) four central, replicates of the Table 1 Range of variation of the parameters used in the central composite design Parameter X1 X2 X3

Fig. 3. Kinetic profiles of Metanil Yellow degradation and sample decolorization.

X4

Notation 1

CMY (mg L ) pH F (flux, photon s1 cm2) CTiO2 (g L1)

Low (1)

Center (0)

High (+1)

40 4 0.6  10 16

70 6 2.3  1016

100 8 4.0  1016

0.5

1.5

2.5

central point (0); (iii) eight axial or star points (a = +1) located at the center and both extreme levels of the experimental models. Experiments were undertaken in random order to provide protection against the effects of lurking variables. Table 2 shows the CCF design matrix applied and experimental results for the response (y) corresponding to the removal time of the azo dye. Data analysis using the statgraphics software at 95% of confidence level permitted to obtain a semi-empirical expression which consists of 15 statistically significant coefficients having absolute value greater than zero, with a probability of 95% ( p < 0.05): Fig. 4. Pareto chart: standardized effects of individual factors and interactions on removal time of Metanil Yellow. The vertical line represents the 95% confidence interval for a significant effect.  and + signs indicate negative and positive effects, respectively.

YðminÞ ¼ 205:8ð50:9Þ þ 166:1X 1 ð4:1Þ  12:0X 2 ð4:1Þ  184:5X 3 ð4:1Þ  98:5X 4 ð4:1Þ  62:4X22 ð10:3Þ þ 55:6X32 ð10:3Þ

the measured values against the predicted responses by the model for the removal time (Table 2). For visualization of the calculated factor effects, the Pareto chart is presented in Fig. 4. This chart gives a graphical presentation in which the effects are presented in chart form with the causes depicted in rank order. All the standardized effects were in absolute values and those which overpass the significance line (vertical line), exert a statistically significant influence on the removal time of Metanil Yellow. As it can be seen, the most important parameters of the overall treatment

þ 72:1X42 ð10:3Þ  17:4X 1 X 2 ð4:3Þ  108:2X 1 X 3 ð4:3Þ  63:9X 1 X 4 ð4:3Þ þ 25:6X 2 X 4 ð4:3Þ þ 52:0X 3 X 4 ð4:3Þ þ 18:2X 1 X 2 X 4 ð4:3Þ þ 31:1X 1 X 3 X 4 This model explains perfectly the experimental range studied (R2 adjusted = 0.9989). This can be seen by comparing Table 2 Experimental data of the CCF design Experiment number

Variable levels (codified values) 1

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28

16

1

CMY (mg L )

pH

F (flux, 10 photon s

100 100 100 100 40 40 100 40 100 40 40 40 100 40 100 40 70 70 70 70 40 100 70 70 70 70 70 70

8 8 4 4 4 8 8 8 4 4 4 4 4 8 8 8 6 6 6 6 6 6 4 8 6 6 6 6

4 (+1) 4 (+1) 0.6 (1) 0.6 (1) 4 (+1) 4 (+1) 0.6 (1) 0.6 (1) 4 (+1) 0.6 (1) 4 (+1) 0.6 (1) 4 (+1) 4 (+1) 0.6 (1) 0.6 (1) 2.3 (0) 2.3 (0) 2.3 (0) 2.3 (0) 2.3 (0) 2.3 (0) 2.3 (0) 2.3 (0) 0.6 (1) 4 (+1) 2.3 (0) 2.3 (0)

(+1) (+1) (+1) (+1) (1) (1) (+1) (1) (+1) (1) (1) (1) (+1) (1) (+1) (1) (0) (0) (0) (0) (1) (+1) (0) (0) (0) (0) (0) (0)

(+1) (+1) (1) (1) (1) (+1) (+1) (+1) (1) (1) (1) (1) (1) (+1) (+1) (+1) (0) (0) (0) (0) (0) (0) (1) (+1) (0) (0) (0) (0)

2

cm )

Yexp (min)

Ycalc (min)

153 76 465 1050 36 48 505 145 296 105 10 245 58 21 900 225 202 213 204 208 44 365 156 130 450 72 380 175

150 79 470 1049 45 40 499 138 297 113 3 239 51 28 902 235 206 206 206 206 40 372 155 131 446 77 376 179

1

CTiO2 (g L ) 0.5 2.5 2.5 0.5 0.5 0.5 2.5 2.5 0.5 2.5 2.5 0.5 2.5 2.5 0.5 0.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 0.5 2.5

(1) (+1) (+1) (1) (1) (1) (+1) (+1) (1) (+1) (+1) (1) (+1) (+1) (1) (1) (0) (0) (0) (0) (0) (0) (0) (0) (0) (0) (1) (+1)

procedure were the light flux (C) followed by the initial dye concentration (A) whereas TiO2 concentration (D) has less important effect and the pH appears to cause a weak direct effect but significant for the confidence level selected (95%). On the other hand, the dye removal time was also affected by several interrelated variables such as two factor interactions (AC, AD, BD, CD) and even three factor interactions (ACD, ABD). Fig. 5a and b display the plots of individual and two-factor interactions effects, respectively, on the removal time. In Fig. 5a, the curve slope is proportional to the effect size whereas the line direction specifies a positive or negative influence of the effect. In Fig. 5b, an interaction between variables occurs when the change in response from the low level to the high level of one variable is not the same as the change in response at the same two levels of a second variable. That is, the effect of one variable is dependent upon a second variable. Any discrepancy between the two lines (lines marked with + and ) can be attributed to a significant interaction between the factors in question. 3.4. Screening of main effects and factor interactions The first step of the chemometric approach was to establish the relative influence of the considered factors and their possible interactions. 3.4.1. Initial concentration of Metanil Yellow As it can be seen the variation of removal time is linear as function of the concentration and no quadratic effect is associated with it. Three possible reasons can be invoked to

explain this observation: first, at high dye concentrations the generation of OH radicals on the surface of catalyst is reduced since the active sites are covered by dye ions. Second, with the increase in the dye concentration, less photons reach the photocatalyst surface (UV screening effect), resulting in slower production of hydroxyl radicals (OH) [28,29]. Consequently, the photocatalytic activity is decreased, since the fewer available OH radicals are required to oxidize more dye molecules. The third possible cause is the interference of intermediates formed upon the degradation of Metanil Yellow. Such competition would be more pronounced in the presence of a high concentration level of reaction intermediates produced by the degradation of an increased initial dye concentration. On the other hand, significant interactions were observed between the concentration and other factors, mainly with the concentration of TiO2 and the light flux. This finding shows well that according to the initial dye concentration, the effect of these two parameters changes and therefore the optimal conditions for the degradation can be different. Such information would not be acquired in a univariate study of the photocatalytic process and thus, the use of experimental design of experiments over the conventional univariate optimization is necessary. 3.4.2. pH Although the first order effect of pH is weak, the quadratic effect is significant and results in a slight improvement of the degradation rate at low or high pH. This result indicates that the process efficiency is not considerably affected over a wide range of pH (4–8), which is quite satisfactory in view of applications. The interpretation of pH effect can be principally explained by a modification of the electrical double layer of the solid– electrolyte interface, which consequently affects the sorption– desorption processes and the separation of the photogenerated electron–hole pairs at the surface of the semiconductor particles. For pH values higher than the pHpzc of titania (pH 6.25), the surface becomes negatively charged and it is the opposite for pH < pHpzc, according to the following equilibria: pH < pHpzc : TiOH þ Hþ ! TiOH2 þ pH > pHpzc : TiOH þ OH ! TiO þ H2 O

Fig. 5. Graphical presentation of the statistical evaluation of the effects of individual factors (a) and interactions of two factors (b) on the removal time of Metanil Yellow.

Since Metanil Yellow is an anionic dye and has a sulfonate group, its adsorption is favored at low pH (the extent of adsorption is almost twofold at pH 4.0 compared to that at neutral pH), thus the reaction rate should be improved and the removal time decrease with decrease in pH as reported elsewhere [30,31]. However, it must be noted that this justification is valid only for acid pH greater or equal to 4, because at lower pH, a strong adsorption can lead to a major decrease of the active centers on the catalyst surface, which means that the absorption of the light quanta by the catalyst is decreased as well. Furthermore, several acidic reaction

intermediates can be strongly adsorbed and could inhibit the degradation of the Metanil Yellow. On the other hand, at pH between 6 and 8, the adsorption of Metanil Yellow is disfavored because of the electrostatic repulsion between the surface and the dye molecules, which seems to conflict with the experimental results. However, as it has been stated in previous works [32,33], the degradation can proceed via reaction of the dye with hydroxyl radicals, generated after adsorption of hydroxyl ions onto catalyst surface followed by reaction with positive holes (h+). Thus the efficiency of the process is reasonably enhanced. 3.4.3. Light flux (F) The effect of light flux (F) caused the greatest impact on dye removal time. Its effect profile can be divided in two parts: for low fluxes (F < 2  1016 photons s1 cm2), the reaction rate is proportional to F. It means that the process works in a good photocatalytic regime: the incident photons are efficiently converted into active species that act in the degradation mechanism [19]. For higher fluxes (F > 2  1016 photons s1 cm2), the rate varies as fractional order between zero and one [34]. This behavior is related to the significant quadratic effect (CC) and also the antagonist effect with TiO2

concentrations (CD) which limit the efficiency of the process and slow down the reaction rate. These two effects originate mainly from the electron–hole recombination which becomes predominant. On the other hand, it should be pointed out that a very strong interaction exists between the light flux and dye initial concentration (see Fig. 5b, AC). This indicates an important interdependence of reaction rates on photon flux and on dye concentration. At low dye concentrations, the increase of light flux induced much more effect on dye removal time (or reaction rate) than that at high concentrations. As a consequence, according to the concentration of substrate molecules, the dependence of reaction rate on light flux might vary from zeroorder kinetics at low concentrations of reactant to first-order kinetics at sufficiently high concentrations in the same range of light flux (not stronger and not weaker). 3.4.4. TiO2 concentration The effect of TiO2 concentration on the dye removal time was also significant, as expected, confirming the positive influence of the increased number of TiO2 active sites on the process kinetics. As the amount of TiO2 increased, the photocatalytic efficiency increased proportionally. However,

Fig. 6. Contour plots of the removal time for the four most important pair of factors.

at higher concentrations, this beneficial effect tends to level off and then decreases because of the screening effect of excess particles which mask a part of the photosensitive surface active sites and hinder the light penetration. This is clearly represented by the important positive quadratic effect (DD) assigned to the TiO2 concentration (see Fig. 4). On the other hand, three synergistic effects were observed with the concentration of Metanil Yellow (AD), light flux (CD) and pH (BD). The most significant is the first one (AD) which indicates that at low concentrations of TiO2 the removal time is highly correlated with the initial concentration of Metanil Yellow whereas at high TiO2 concentrations, this correlation is less visible. This means that the optimal TiO2 concentration is strongly affected by the working conditions. 3.5. Response surface (contour) plots and optimization conditions After performing a screening of factors and their interactions, the response surface analysis was carried out, in order to find the optimal conditions for the degradation of Metanil Yellow. Response surface plots provide a method to predict the time necessary to achieve a complete disappearance of Metanil Yellow (removal time) for different values of the test variables. In addition, the contours of the plots help in identification of the type of interactions between the selected variables [35]. Each contour curve represents an infinite number of combinations of the two selected variables with the other maintained at their respective zero coded level. A circular contour of response surfaces indicates that the interaction between the corresponding variables is negligible. An elliptical or saddle nature of the contour plots indicates that the interaction between the corresponding variables is significant. In Fig. 6, are displayed the contour plots of the removal time for the four most important pairs of factors: initial concentration versus light flux (Fig. 6a), initial concentration versus TiO2 concentration (Fig. 6b), light flux versus TiO2 concentration (Fig. 6c) and pH versus TiO2 concentration (Fig. 6d). The optimal conditions for fast removal of Metanil Yellow are depicted by the shaded zones:  Fig. 6a: the most suitable conditions are: low dye concentration (40–60 mg L1) and high light flux (F > 2  1016 photons s1 cm2). When dye concentration increases, a higher light flux is necessary to reduce the removal time of Metanil Yellow.  Fig. 6b: a fast removal time ( 3.3  1016 photons

s1 cm2) and a moderate concentration of TiO2 (1.2– 2.3 g L1). Working at low light flux increases considerably the removal time, confirming the strong effect of this parameter on the degradation rate.  Fig. 6d: As can be seen, two optimal zones were found: the first one (at left of the figure) is observed at low pH (around 4) and high TiO2 concentrations (2–2.5 g L1), whereas the second one is more restricted and characterized by a high pH (around 8) and moderate concentration of TiO2 (1.6– 2.2 g L1). However, it should be noted that changing pH did not permit to achieve a very fast reaction rate since the best removal time observed was around 2 h and thus, it is more convenient to modify other parameters such as light flux, TiO2 concentration to enhance the degradation rate of Metanil Yellow. From the polynomial expression, the most suitable conditions for the degradation of Metanil Yellow were calculated for three different initial dye concentrations: 50, 80 and 100 mg L1. The results are listed in Table 3. To validate the model, experiments were performed under the optimal conditions and the observed values for the removal time are presented in Table 3, last column. As showed, the experimental values were reasonably close to the predicted values, indicating the adequacy of the obtained model for modeling the removal time of Metanil Yellow by photocatalysis. On the other hand, using a univariate approach, the optimal conditions were, for the different dye concentrations, as follow: TiO2 concentration: 2.5 g L1, pH: 4.0, light flux: 4  1016 photons s1 cm2. Comparing these results with that found by chemometric approach, it can be seen that univariate approach cannot permit an appropriate optimization of reaction kinetic, since it ignores the interdependence of the reaction parameters. As consequent, the use of experimental design strategy over the conventional univariate one seems to be indispensable to realize a successful investigation and an adequate modeling of photocatalytic reaction processes. 3.6. Characterization of degradation intermediates During the photocatalytic oxidation of Metanil Yellow, various organic intermediates were produced. Using HPLC/ DAD and GC/EI-MS techniques, 11 by-products were identified as possible degradation products. Table 4 summarizes the GC–MS retention times, the molecular weights and the characteristic fragmentation patterns ions by-products. Based Table 3 Optimal conditions for the removal of Metanil Yellow by TiO2 photocatalysis Light flux (F) Concentration pH Concentration TiO2 (g L1) Metanil Yellow 1016 photon 1 1 2 s cm (mg L ) 50 80 100

3.8 4.0 4.0

1.6 1.7 1.7

Removal time (min) Predicted Observed

4.6 5.0 8.0 13.6 8.0 38.5

6.0 15.0 36.0

Table 4 Mass spectra data and structures of identified compounds by GC/MS analysis for an irradiated sample of Metanil Yellow (MY) Compound

Retention time (min)

Molecular weight (g mol1)

Main fragments m/z

MY 1,2 3 4 5 6 7 8 9 10

34.6 38.2 15.8 20.8 25.1 25.3 6.3 10.7 11.6 17.6

353 369 158 169 185 185 78 93 94 110

353a, 196, 169, 158, 93, 77, 66, 51, 39 369a, 212, 185, 158, 108, 94, 93, 77, 66, 51, 39, 158, 141, 94, 77, 66, 51, 39 169, 154, 115, 84, 77, 66, 51, 39 185, 156, 129, 115, 108, 91, 77, 66, 51, 39

a

78, 77, 63, 51, 39 93, 78, 66, 52, 39 94, 78, 66, 63, 55, 52, 39 108, 82, 80, 54, 52

Mass detected using SIM (selected ion monitoring) mode.

on the identification results, a tentative reaction pathway is proposed, as schematically depicted in Fig. 7. The first steps of the degradation most probably involve: (i) Hydroxylation of aromatic ring of Metanil Yellow leading to several isomer intermediates among which only the two major ones were unambiguously identified (1, 2). The UV– vis spectra of these two products (not showed) were similar to that of Metanil Yellow (lmax around 450 nm), which explains the slow kinetic of solution decolorization (Section 3.2) compared to that of Metanil Yellow degradation. It should be pointed out that the attack of hydroxyl group occurs principally on the aromatic ring with

amino group, due to the electron withdrawing effect of sulfonate group which inhibits the reactivity of sulfonated aromatic ring towards OH radicals. (ii) Cleavage of the Metanil Yellow molecule at the level of the very reactive azo group bonds (–N N–), giving arise to benzene sulfonic acid (3) and diphenyl amine (4). The primary hydroxylated by-products (1, 2) can in turn undergo a breakdown of azo group and form benzene sulfonic acid and two hydroxyl-diphenyl amine isomers (5, 6), which could also result from the hydroxylation of diphenylamine (4). Only the ortho- (5) and para- (6) isomers were detected and identified by GC/MS whereas the meta isomer was not

Fig. 7. Proposed photocatalytic degradation pathway of Metanil Yellow in aqueous suspensions of TiO2.

detected, which indicates that hydroxylation take places preferentially on ortho and para positions of the amino group. This finding let us suppose that hydroxylation by OH radicals is selective and depends on the electron density of carbon sites of the aromatic ring, as it has been demonstrated in our previous studies on the photocatalytic degradation of iodosulfuron [36,37] and also in a recent work on the oxidation of aromatic compounds by heterogeneous photocatalysis [38]. The reaction proceeds by de-sulfonation of benzenesulfonic acid to give benzene and cleavage of –NH– bonds of hydoxylated diphenyl amine compounds (5–6) resulting in the formation of benzene (7), aniline (8) and phenol (9). Other possible routes for the formation of phenol might be hydroxylation of benzene by OH radicals and oxidation of aniline by holes (h+) as it has been previously reported during the degradation of aniline derivatives [39]. Phenol is subsequently attacked by OH radicals to form hydroquinone (10) which can submit an oxidative opening of the aromatic ring leading to low molecular weight aliphatic carboxylic acids, such as formic and acetic acid. The final steps involve the oxidation of carboxylic acids to CO2 and water. 3.7. Formation of mineralization products Complete mineralization is of great importance in water treatment process of organic pollutants. Thus, the monitoring of inorganic ions formation and TOC removal during the photocatalytic reaction was achieved to assess the mineralization extent of Metanil Yellow (experimental conditions are identical to those reported in Sections 3.1 and 3.2). Fig. 8 shows the time course of TOC disappearance represented as the number of carbon atoms per molecule of Metanil Yellow (40 mg L1, 18 carbon atoms). Although Metanil Yellow degradation was attained in a few minutes, TOC removal followed much slower rates. Only 40% of the carbon atoms were transformed into CO2 after total decolorization of the azo dye solution (150 min) whereas complete TOC removal was accomplished after long irradiation time (60 h), because of the formation of more stable intermediates towards oxidation.

Analysis of inorganic ions revealed as expected the formation of NH4+, NO3 and SO42 in solution. It was observed that the concentration of sulfate ions increases rapidly at the first 3 h and tends to reach a plateau of ca. 58 mmol L1 after 5 h. This can be attributed to the fast elimination of the –SO3 moiety in solution. The total amount of sulfate ions observed corresponds to 55% of the expected amount (107 mmol L1) assuming complete mineralization of the dye solution. This behavior can be ascribed to a partially irreversible adsorption of sulfate ions on TiO2, as it has been already reported for the degradation of other sulfurcontaining dyes [40–42]. Regarding the mineralization of nitrogen atoms, the total mineralization of 40 mg L1 (107 mmol L1) of the azo dye would produce about 320 mmol L1 (NH4+ + NO3) ions. However, at the end of the experiment, less than one of third (1/ 3) of the expected stoichometric amount of nitrogen-containing ions was produced. NH4+ ions are released first in solution and reach a maximum concentration of 65 mmol L1 within 10 h whereas NO3 concentration was about 35 mmol L1 at the same time. Later on, nitrate increases slowly to reach a plateau after 50 h. This result indicates that the –N N– azo group was very likely transformed with 100% selectivity into gaseous N2 as reported previously [43] and only the –NH– amino group has been transformed to ammonium which is subsequently oxidized to nitrate. Moreover, this finding corroborates the reaction pathway proposed in Fig. 7. 4. Conclusions Once more time, heterogeneous photocatalysis proves to be a very efficient method for water pollution remediation as a complete mineralization of Metanil Yellow was achieved. The identification of reaction intermediates allowed us to propose a reaction pathway of the degradation which involves breakdown of the azo bond, hydroxylation of the aromatic ring leading at the end to ring opening and formation of carboxylic acids which undergoes oxidation to form CO2 and water. On the other hand, the chemometric approach used in this study has showed to be a valuable tool for the identification and interpretation of the photocatalytic reaction parameters and their possible interactions. The high correlation of the model with the experimental results indicates that this analytical procedure could be a general method to describe any similar TiO2 photocatalytic system and to predict its behavior. This should be of great importance for the optimization of the experimental conditions for technological applications such as scale up of photocatalytic reactors, taking into account the role of the different parameters. Further experiments on other compounds with different chemical structures will help prove, or disprove, the general applicability of this approach for optimization of TiO2 photocatalysis processes. References

Fig. 8. Total organic carbon (TOC) disappearance during the photocatalytic degradation of Metanil Yellow.

[1] H. Ollgaard, L. Frost, J. Gastler, O.C. Hensen, Survey of Azo-Colorants on Denmark: Milgo Project 509, Danish Environmental Protection Agency, 1999.

[2] A. Stolz, Appl. Microbiol. Biotechnol. 56 (2001) 69. [3] C. O’Neill, F.R. Hawkes, D.L. Hawkes, N.D. Lourenco, H.M. Pinheiro, W. Delee, J. Chem. Technol. Biotechnol. 74 (1999) 1009. [4] K.T. Chung, C.E. Cerniglia, Mutat. Res. 277 (1992) 201. [5] M.A. Brown, S.C. De Vito, Crit. Rev. Environ. Sci. Technol. 23 (1993) 249. [6] A. Bianco-Prevot, C. Baiochi, M.C. Brussino, E. Pramauro, P. Savarino, V. Augugliaro, G. Marci, L. Palmisano, Environ. Sci. Technol. 35 (2001) 971. [7] B. Neppolian, H.C. Choi, S. Sackthivel, B. Arabindoo, V. Murugesan, Chemosphere 46 (2002) 971. [8] I. Safarik, K. Nymburska, M. Safaikova, J. Chem. Tech. Biotechnol. 69 (1997) 1. [9] A.T. More, A. Vira, S. Fogel, Environ. Sci. Technol. 23 (1989) 403. [10] Y.M. Slokar, A.M. Le Marechal, Dyes Pigments 37 (1998) 1160. [11] W.G. Kuo, Wat. Res. 26 (1992) 881. [12] C. Morisson, J. Bandara, J. Kiwi, A. Lopez, J. Adv. Oxid. Technol. 1 (1996) 160. [13] I. Arslan, T. Akmehmet, T. Tuhkamen, Environ. Technol. 20 (1999) 921. [14] H. Lachheb, E. Puzenat, A. Houas, M. Ksibi, E. Elaloui, C. Guillard, J.M. Herrmann, Appl. Catal. B: Environ. 39 (2002) 75. [15] I.K. Konstantinou, T.A. Albanis, Appl. Catal. B: Environ. 49 (2004) 1. [16] F. Zhang, J. Zhao, T. Shen, H. Hidaka, E. Pellizzetti, N. Serpone, Appl. Catal. B: Environ. 15 (1998) 147. [17] C.G. Silva, W. Wang, J.L. Faria, J. Photochem. Photobiol. A: Chem. 181 (2006) 314. [18] M. Stylidi, D.I. Kondarides, X.E. Verykios, Appl. Catal. B: Environ. 47 (2004) 189. [19] J.M. Herrmann, Catal. Today 53 (1999) 115. [20] C. Guillard, H. Lachheb, A. Houas, M. Ksibi, E. Elalouli, J.M. Herrmann, J. Photochem. Photobiol. A: Chem. 158 (2003) 27. [21] C. Hu, J.C. Yu, Z. Hao, P.K. Wong, Appl. Catal. B: Environ. 46 (2003) 35. [22] F. Kiriakidou, D.I. Kondarides, X.E. Verykios, Catal. Today 54 (1999) 119. [23] J. Ferna´ndez, J. Kiwi, J. Baeza, J. Freer, C. Lizama, H.D. Mansilla, Appl. Catal. B: Environ. 48 (2004) 205. [24] C. Lizama, J. Freer, J. Baeza, H.D. Mansilla, Catal. Today 76 (2002) 235.

[25] A.V. Emeline, V.K. Ryabchuk, N. Serpone, J. Phys. Chem. B 109 (2005) 18515. [26] E. Vulliet, J.M. Chovelon, C. Guillard, J.M. Herrmann, J. Photochem. Photobiol. A: Chem. 159 (2003) 71. [27] I. Bouzaida, C. Ferronato, J.M. Chovelon, M.E. Rammah, J.M. Herrmann, J. Photochem. Photobiol. A: Chem. 168 (2004) 23. [28] J. Grzechulska, A.W. Morawski, Appl. Catal. B: Environ. 36 (2002) 45. [29] I. Poulios, I. Aetopoulou, Environ. Technol. 20 (1999) 479. [30] S. Naskar, S. Arumugam Pillay, M. Chanda, J. Photochem. Photobiol. A: Chem. 113 (1998) 254. [31] B. Neppolian, S. Sakthivel, M. Palanichamy, B. Arabindoo, V. Murugesan, Stud. Sur. Sci. Catal. 113 (1998) 329. [32] Z. Shourong, H. Qingguo, Z. Jun, W. Bingkun, J. Photochem. Photobiol. A: Chem. 108 (1997) 235. [33] M.S.T. Conc¸alves, A.M.F. Oliveira-Campos, M.M.S. Pinto, P.M.S. Plasencia, M.J.R.P. Queiroz, Chemosphere 39 (1999) 781. [34] D.F. Ollis, E. Pelizzetti, N. Serpone, Environ. Sci. Technol. 25 (1991) 1523. [35] D.C. Montgomry, Design and Analysis of Experiments, 5th ed., John Wiley and Sons, New York, 2001. [36] M. Sleiman, P. Conchon, C. Ferronato, J.M. Chovelon, Appl. Catal. B: Environ. 71 (2007) 279. [37] M. Sleiman, C. Ferronato, B. Fenet, R. Baudot, F. Jaber, J.M. Chovelon, Anal. Chem. 78 (2006) 2957. [38] G. Palmisano, M. Addamo, V. Augugliaro, T. Caronna, A. Di Paola, E.G. Lopez, V. Loddo, G. Marci, L. Palmisano, M. Schiavello, Catal. Today 122 (2007) 118. [39] M. Canle, J.A. Santaballa, E. Vulliet, J. Photochem. Photobiol. A: Chem. 175 (2005) 192. [40] J.M. Herrmann, C. Guillard, P. Pichat, Catal. Today 17 (1993) 7. [41] M. Vautier, C. Guillard, J.M. Herrmann, J. Catal. 201 (2001) 46. [42] M. Tamimi, S. Qourzal, A. Assabane, J.M. Chovelon, C. Ferronato, Y. AitIchou, Photochem. Photobiol. Sci. 5 (2006) 477. [43] M. Karkmaz, E. Puzenat, C. Guillard, J.M. Herrmann, Appl. Catal. B: Environ. 51 (2004) 183.