H2O2 ... - ACS Publications

A kinetic model was developed to predict the removal of aqueous metronidazole utilizing the UV/H2O2 process. The rate constant for the reaction betwee...
7 downloads 0 Views 1MB Size
Ind. Eng. Chem. Res. 2008, 47, 6525–6537

6525

KINETICS, CATALYSIS, AND REACTION ENGINEERING Aqueous Metronidazole Degradation by UV/H2O2 Process in Single-and Multi-Lamp Tubular Photoreactors: Kinetics and Reactor Design Melody Blythe Johnson and Mehrab Mehrvar* Department of Chemical Engineering, Ryerson UniVersity, 350 Victoria Street, Toronto, Ontario, Canada M5B 2K3

A kinetic model was developed to predict the removal of aqueous metronidazole utilizing the UV/H2O2 process. The rate constant for the reaction between metronidazole and hydroxyl radicals was determined to be 1.98 × 109 M-1 s-1. The model was able to predict an optimal initial H2O2 dose and the inhibitory effects of high H2O2 doses and bicarbonate ions in the aqueous solution. Simulations were performed for three different photoreactors treating a 6 µM solution of metronidazole at various influent H2O2 doses and photoreactor radii. The predicted removal rates of metronidazole were 4.9-13% and 14-41% for the single-lamp and multilamp photoreactors, respectively. Selection of a photoreactor radius for maximum metronidazole removal varied with influent H2O2 concentration. The lowest operational cost of $0.05 per mmol removed was projected for the multilamp photoreactor. Operationally, it was cost-effective to utilize higher UV lamp output (36W), while keeping influent H2O2 concentration low (25 mg/L). 1.Introduction Wastewater treatment works have, to date, incorporated biological and chemical removal of harmful constituents from raw wastewater prior to discharge of the treated effluent to a receiving body. Over time, various processes have evolved for the removal of these constituents. Today, with more complex chemicals being discharged to wastewater collection systems, greater emphasis on the removal of small amounts of micropollutants is becoming an expensive necessity. As an example, greater concern is being expressed on a daily basis on the amounts and occurrences of pharmaceuticals and personal care products (PPCPs) in the treated wastewater effluent being discharged to surface water bodies.1–4 Since the early 1980s, several studies have documented the presence of pharmaceuticals and related metabolites in the aquatic environment.1,5 Since then, with better and more sophisticated analytical methods, a large number of chemicals belonging to various classes of drugs, including antibiotics, analgesics and anti-inflammatories, lipid regulator agents, β-blockers, antiepileptics, contraceptives, steroids, and related hormones, have been documented in the aquatic environment.1,5 Pharmaceuticals are unique pollutants in that they are designed to act in a specific way, that is, to target a biological effect. Because of their regular and widespread use for health benefits for both humans and animals, pharmaceuticals and their metabolites are continuously being introduced into the aquatic environment. By their nature, many of these compounds are resistant to conventional forms of biological and chemical treatment, and are therefore considered to be “pseudopersistent compounds” in the environment.5 In recent years, more attention has been directed to drug residues as they most often have similar physicochemical behavior as other harmful chemicals that can be accumulated in living organisms. As a result of the presence of pharmaceuticals and their metabolites in surface and ground waters, investigations into methods for their removal * To whom correspondence should be addressed. E-mail: mmehrvar@ ryerson.ca.

have intensified, with advanced oxidation processes (AOPs) showing much promise for the treatment of these compounds.2,6–11 Metronidazole is an antibiotic used to treat infections caused by anaerobic bacteria and various protozoans.7 This compound was selected as the target compound for this study because it is widely used, highly soluble, nonbiodegradable, and a suspected carcinogen.12 Metronidazole has been detected in concentrations of 0.011 to 0.055 µM in hospital effluent wastewaters.13 Figure 1 presents the chemical structure of metronidazole. One promising AOP for the treatment of PPCPs in aqueous solutions is the UV/H2O2 process. To effectively utilize the UV/H2O2 process, we require a means of evaluating and optimizing photochemical kinetics and photoreactor design. Although evaluation of experimental results compared with reactor-specific models has been previously investigated,7,14–17 little work has been done on models that can be applied to various photoreactor geometries.6,18 The development of a model that can be used to predict the behavior of photoreactors of varying geometries and influent solution composition would provide a valuable tool for the design and optimization of photoreactors using the UV/ H2O2 process. The focus of this study is the development of design equations for the modeling and optimization of UV photoreactors for the treatment of the pharmaceutical compound metronidazole utilizing the UV/H2O2 process. 2. Model Development and Methodology 2.1. Hydrodynamics and Geometry of the Photoreactors.

Figure 1. Chemical structure of metronidazole.

10.1021/ie071637v CCC: $40.75  2008 American Chemical Society Published on Web 08/06/2008

6526 Ind. Eng. Chem. Res., Vol. 47, No. 17, 2008

Figure 2. Schematic diagram of the single UV lamp photoreactor.

As shown in Figure 2, the single-lamp photoreactor is tubular with radius R and a UV lamp down the center axis with radius Rt inside a quartz sleeve with radius Ri. The aqueous solution flows upward along the z-direction in the annular space between the quartz sleeve and the outer shell of the reactor. Cylindrical coordinates were used for the purpose of modeling. Two separate photoreactors, reactors 1 and 2, were modeled such that each configuration had two single-lamp UV photoreactors operating in series. Hence, the overall length of each configuration was equal to the sum of the length of two single-lamp UV photoreactors. Both reactors 1 and 2 have identical dimensions; however, each is equipped with a different lamp: the lamps in reactor 1 each have a UV output of 25.5 W, whereas the lamps in reactor 2 each have a UV output of 36 W. Physical properties and operating conditions of the systems are summarized in Table 1. The multilamp photoreactor is also tubular with radius R and four UV lamps, each with a radius Rt and inside a quartz sleeve of radius Ri, as shown in Figure 3. There are four UV lamps located at any cross section of the photoreactor. The lamps are positioned parallel to the z-axis and equally spaced, each located half the reactor radius from the center of the reactor. The aqueous solution flows upward along the z-direction in the space between the quartz sleeve and the outer shell of the reactor. For modeling purposes, Cartesian coordinates were used for the multilamp UV photoreactor. Other characteristics of this multilamp photoreactor are summarized in Table 1. The multi-lamp photoreactor, reactor 3, was also configured such that it comprised two multilamp UV photoreactors operating in series. It was assumed that plug flow was maintained along the full length of all three reactors. Low-pressure (LP) UV lamps were chosen as they produce nearly monochromatic light at 254 nm. Two commercially available LP UV lamps, Philips model TUV75W HO and Emperor Aquatics G48T6LVHO, were selected. A range of

reactor diameters and initial H2O2 concentrations were chosen such that the effect of varying these operating conditions could be examined. As shown in Table 1, influent H2O2 concentrations of 25, 50, 75, and 100 mg/L and reactor radii of 50, 100, 150, 200, and 250 mm were investigated. Thus, for each reactor, 20 pairs of operating conditions were investigated, comprising combinations of influent H2O2 concentrations and reactor radii. In addition, it was assumed that each photoreactor configuration operates at steady-state conditions, isothermal with constant temperature of 25 °C, and behaves similar to an ideal plug flow reactor. Furthermore, it was assumed that the system is treating an aqueous solution with characteristics similar to that of tap water containing only the target compound, metronidazole, at an inlet concentration of 6 µM, and hardness as carbonate and bicarbonate ions, at a concentration of 100 mg/L as CaCO3. Finally, the effect of reaction intermediates in the system was neglected. For a single UV lamp photoreactor, similar to that presented in Figure 1, Shen and Wang19 found that the variation of temperature in the photoreactor during photodecomposition was negligible, holding steady at around 25 °C, implying that the system is nearly isothermal. As a result, it was assumed that both the single- and multilamp photoreactors were isothermal at a uniform temperature of 25 °C. On the basis of the flow rate of 2.0 L/s, the values of Reynolds number (NR) for reactors 1 and 2 are greater than 5300 for reactor radii up to 250 mm, indicating turbulent flow through the reactor. An empty tubular reactor can often be simulated as a plug flow reactor.20 Shen and Wang19 determined that for a single UV lamp photoreactor similar to the configuration presented in Figure 2, the flow behavior could be assumed to be ideal plug flow. Therefore, ideal plug flow was assumed for all modeled reactors in this work. In ideal plug flow, Vr ) Vθ ) 0, and Vz ) constant at all points in the reactor. As a result, the momentum balance can be neglected, and the residence time in the reactor, τ, is a function of axial position in the reactor, z, only. 2.2. Radiation Energy Balance. Assuming no emissivity of molecules in aqueous solution, the radiation energy balance, for a single UV lamp in a quartz sleeve parallel to the z-axis, centered at the origin, is 1 d(rq) ) -q(2.303µs) r dr

(1)

where q is the radiant energy flux in mol photons/(m2 s) and µs is the extinction coefficient of the aqueous solution (base 10) in m-1. Integration yields: Ri q ) qo e-2.303µs(r-Ri) r

(2)

In this equation, qo is the radiant energy flux at the outer surface of the quartz sleeve. The value of the extinction coefficient of distilled water at 254 nm, µw ) 0.7 m-1, was used to evaluate the experimental results that have been previously published.7 For the purposes of modeling the behavior of the system for treating a potable water supply, the value of µw at 254 nm was taken to be 10 m-1, or that of average tap water.21 The UV absorptive properties of metronidazole and H2O2 can be expressed in terms of their molar absorptivities at 254 nm, RH and H2O2, which are 220 and 1.86 mM-1 m-1, respectively.6,7 All above extinction coefficients and molar absorptivities are to the base 10. The UV absorptive properties of reaction intermediates were neglected because it was assumed that the concentrations of these compounds would be very small. As a result, the local extinction coefficient of the aqueous solution

Ind. Eng. Chem. Res., Vol. 47, No. 17, 2008 6527 Table 1. Physical Properties and Operating Conditions of the Photoreactors parameter

reactor 1

type number length (overall) (m) radius (mm)

single UV lamp 2 2.4 50-250

number model type nominal length (m) nominal radius (Rt) (mm) input current (each) (amps) input watts (each) (W) UV output (each) (W)

2×1 Philips TUV75W HO low pressure 1.2 13 0.84 75 25.5

outer radius (Ri) (mm)

17

flow rate to be treated (L/s) [metronidazole]o (µM) [H2O2]o (mg/L)

2.0 6 25-100

reactor 2

reactor 3

Reactors in Series single UV lamp 2 2.4 50-250

multiple UV lamp 2 2.4 50-250

UV Lamps 2×1 Emperor Aquatics G48T6LVHO low pressure 1.2 8.5 1.20 120 36

2×4 Emperor Aquatics G48T6LVHO low pressure 1.2 8.5 1.20 120 36

Quartz Sleeve 17

17

Operating Conditions 2.0 6 25-100

can be expressed as µs ) µw + εH2O2CH2O2 + εRHCRH

(3)

The local volumetric rate of energy absorption (LVREA), A, in mol photons/(L s), is a function of the radiant energy flux22 A ) µsq

(4)

Using eqs 2 and 4, the LVREA can be expressed in terms of radial position: Ri A ) µsqo e-2.303µs(r-Ri) r

Figure 3. Schematic diagram of the multiple UV lamp photoreactor.

(5)

2.0 6 25-100

For any given compound, i, the fraction of photons absorbed by that compound at any given location within the reactor, fi, can be expressed as a ratio of its local extinction coefficient, in terms of its local concentration, Ci, and molar absorptivity, i, and the local extinction coefficient of the aqueous solution, µs fi )

εiCi µs

(6)

For the purposes of modeling the multilamp photoreactor (reactor 3), the expression for the LVREA was converted to Cartesian coordinates, as shown in eq 7. This expression represents the LVREA from a lamp centered at a point at (a, b).

6528 Ind. Eng. Chem. Res., Vol. 47, No. 17, 2008

A ) µsqo

Ri

√(x - a)2+(y - b)2-Ri)

√(x - a)2 + (y - b)2

e-2.303µs(

(7) The total LVREA for reactor 3 is represented by the sum of the contributions from the four lamps centered at (0.5R, 0), (0, 0.5R), (-0.5R, 0), and (0, -0.5R). 2.3. Mass Balance. The continuity equation in cylindrical coordinates for a compound of constant density, F, and diffusivity, DAB, is shown below

(

)

( ( ) )

∂Ci ∂Ci ∂Ci 1 ∂Ci 1 ∂ ∂Ci + Vr + Vθ + υz ) DAB r + ∂t ∂r r ∂θ ∂z r ∂r ∂r 2 2 1 ∂ Ci ∂ Ci + - Rrxn,i (8) r2 ∂θ2 ∂z2

where Ci is the concentration of species i (M), Vr, Vθ, and Vz are velocities (m/s) in the r-, θ-, and z-directions, respectively. DAB is the diffusivity of the species in the aqueous solution (m2/s), and Rrxn,i is the reaction rate of species i (M/s). By implementing the assumptions mentioned earlier, eq 8 can be simplified to ∂Ci ) -Rrxn,i ∂τ

(9)

where τ (s) is the reactor residence time. Hence, the concentration profile is a function of the residence time, τ, and chemical reaction rate, Rrxn,i. For photolyzed reactions, Rrxn,i is a function of A, which is itself a function of r. Therefore, the concentration profile is a function of residence time, τ, and radial position, r. Similar results are obtained utilizing Cartesian coordinates for multilamp photoreactor. In this case, the concentration profile is a function of residence time, τ, and chemical reaction rate, Rrxn,i. For photolyzed reactions, Rrxn,i is a function of A, which is itself a function of x and y. 2.4. Reaction Model. The UV/H2O2 process involves the formation of hydroxyl radicals from an aqueous solution of H2O2 exposed to UV light with wavelengths varying between 200 and 280 nm. The major reactions involved in the UV/H2O2 process along with reported values of associated quantum yields or rate constants are shown below18,23 •

-1

H2O2 + hV f 2 OH φ10 ) 0.5 mol photon

(10)

H2O2 + •OH f HO•2 + H2O k11 ) 1.4-4.5 × 107 M-1 s-1 (11) 2•OH f H2O2 k12 ) 5.0-8.0 × 109 M-1 s-1

•• 7 -1 -1 HCOs 3 + OH f CO3 + H2O k17 ) 2.0 × 10 M (17) •• CO2k18 ) 3.7 × 108 M-1 s-1 3 + OH f CO3 + OH (18) • 5 -1 -1 CO3•- + H2O2 f HCOs 3 + HO2 k19 ) 8.0 × 10 M (19)

To develop the reaction model, it was assumed that reactions 11–15 and reactions 17–19 were elementary. Reactions 10 and 16, however, were based on the absorption of photons by the reacting species, and thus the chemical reaction rate expressions depend on the energy balance of the system. As an example, for reaction 16, the reaction rate can be written as the product of the photons absorbed by the target compound and the quantum yield for the reaction ∂CRH ) -RRH ) -φ16AfRH (20) ∂τ where RRH is the rate of degradation of the target compound (mM/s). On the basis of chemical reactions 10–19 and eq 9, we may write the kinetic models of the species as follows ∂CRH ) -RRH ) -k15CRHC•OH - φ16AfRH ∂τ ∂CH2O2

2 k13CHO • 2

HO•2 + •OH f H2O + O2 k14 ) 1.4 × 1010 M-1 s-1 (14)

(22)

∂C•OH

) -R•OH ) -k15CRHC•OH + 2φ10AfH2O2 ∂τ 2 k11CH2O2C•OH - 2k12C•OH - k14C•OHCHO2• - k17CHCO3-C•OH k18CCO32-C•OH (23)

∂CHO2• (13)

(21)

2 ) -RH2O2 ) -φ10AfH2O2 - k11CH2O2C•OH + k12C•OH +

∂τ

(12)

2HO•2 f H2O2 + O2 k13 ) 0.8-2.2 × 106 M-1 s-1



photon-1 reported by Shemer et al.,7 k15 needs to be determined using the chemical reaction models. The above list of reactions is by no means a comprehensive list of all the reactions that can take place in a UV/H2O2 process. Competing reactions with radical scavengers, such as carbonate and bicarbonate ions, can reduce the number of hydroxyl radicals in the system, reducing the removal efficiency of target compounds. A measure of the concentration of carbonate and bicarbonate ions in an aqueous solution is known as alkalinity and is reported in mg/L as CaCO3. Some of these reactions are shown in reactions 17–19.18,24

∂τ

2 ) -RHO2• ) + k11CH2O2C•OH - 2k13CHO • 2

k14C•OHCHO2• + k19CCO3•-CH2O2 ∂CHCO3∂τ

k15

(24)

) -RHCO3- ) -k17CHCO3-C•OH + k19CCO3•-CH2O2

OH + RH 98 intermediatesfCO2 + H2O k15varies

(25) ∂CCO32-

(15) φ16

∂τ ∂C•CO3-

RH + hVf intermediatesfCO2 + H2O φ16 varies (16) where φi and ki are the quantum yield (mol photon-1) and reaction rate constant (M-1 s-1) of the reactions listed in reaction i, respectively. Considering the value of φ16 ) 3.3 × 10-3 mol

∂τ

) -RCO32- ) -k18CCO32-C• OH

(26)

) -R•CO3- ) k17CHCO3-C•OH + k18CCO32-C•OH k19CCO3•-CH2O2

(27)

Equations 21–27 thus become the mathematical reaction model for the photoreactor system. As noted earlier, A ) A(r), and

Ind. Eng. Chem. Res., Vol. 47, No. 17, 2008 6529

thus the above set of equations are functions of residence time, τ, and radial position, r, for the single lamp photoreactors. For the multilamp photoreactor, A ) A(x, y), the above set of equations are thus functions of residence time, τ, and the position within the reactor, in terms of x- and y-coordinates. Therefore, on the basis of the developed models, the objectives of this study were as follows: (1) To calibrate the reaction model on the basis of the published experimental data to determine the reaction rate constant for the reaction between metronidazole and the hydroxyl radical (k15). (2) To validate and examining the behavior of the model by introducing inhibitors into the reaction system. (3) To utilize the calibrated model to determine effluent concentration profiles of the target compound and H2O2 for each photoreactor configuration. (4) To determine the optimal initial H2O2 dose, [H2O2]/inf, and optimal reactor radius, R*, for each photoreactor configuration. The optimal parameters are those that provide the lowest operational cost per mmol of the parent compound removed. Matlab, a computer software package, was used to solve the set of ordinary differential equations (ODEs) utilizing a predictor/corrector numerical method, ode15s, based on the numerical differentiation formulas (NDF). This allowed the determination of the effluent concentration profiles for all compounds in the reaction model for each reactor configuration. Average photoreactor effluent concentrations were determined utilizing the trapezoidal rule numerical integration method for reactors 1 and 2. To determine the optimal initial H2O2 dose and outer reactor radius, it is desired to minimize the ratio, F ($/mmol removed), of operational cost in terms of electrical, SE, and chemical costs, SC, to the amount of target compound removed, Q(CRH,inf CRH,eff) F)

SE + Sc Q(CRH,inf - CRH,eff)

(28)

Electrical cost, SE, was based on a cost of $0.11/kWh for electricity. Chemical cost, SC, was based on an assumed cost for H2O2 of $1.10/kg,25 and an assumed cost for a 12% (w/v) solution of sodium hypochlorite of $0.14/L. Sodium hypochlorite is required to quench the residual H2O2 remaining in solution after treatment in the photoreactor based on the following reaction H2O2 + NaOCl f NaCl + H2O + O2

(29)

Influent H2O2 dose, [H2O2]inf, was varied from the lower and upper operating bounds, as shown in Table 1, in 5 mg/L intervals. Reactor radius, R, was varied from the lower and upper operating bounds, as shown in Table 1, in 50 mm intervals. / The optimal influent H2O2 dose,[H2O2]inf , and reactor radius, R*, were determined by trial and error by calculating the value of F for each pair of set of operating conditions. 3. Results and Discussion 3.1. Reaction Rate Constant for Metronidazole: Model Calibration and Validation. The chemical reaction model was calibrated utilizing the experimental data published in the open literature, including the value for φ16 ) 3.3 × 10-3 mol photon-1.7 The calibration involved determining the value of the rate constant between metronidazole and the hydroxyl radical, k15, which was the only value undefined in the reaction eqs 10–19. This data was based on the UV/H2O2 oxidation of

a 6 µM solution of metronidazole in deionized water, with a pH of 6.0, under a collimated beam apparatus emitting 1.5 mW/ cm2 of UV light at 254 nm onto the surface of liquid held in a square Petri dish 7 cm × 5 cm by 2.9 cm deep. Two initial H2O2 doses, 25 and 50 mg/L, were investigated. Because deionized water was used, it was possible to assume that the solution used by Shemer et al.7 did not contain any inhibitory compounds, and thus the extinction coefficient for distilled water was used in the determination of the solution extinction coefficient. These conditions were simulated utilizing the chemical reaction model developed. Unlike the model developed for the single and multilamp photoreactors, the LVREA (A), was not calculated for various positions within the Petri dish. Instead, an average value for the LVREA was calculated over the volume of the Petri dish. It was assumed that the solution in the Petri dish was perfectly mixed, thus the extinction coefficient of the solution, µs, varied with respect to time because of the changing concentrations of the target compound and H2O2. On the basis of reported reaction rate constants for eq 15 for chemically similar compounds, it was assumed that the value for the reaction rate constant, k15, would likely lie within the range 0.5 to 10.0 × 109 M-1 s-1.18 The value of k15 was set to 1.0 × 109 M-1 s-1, which is within the range of expected values, and a simulation run was conducted with an initial H2O2 dose of 25 mg/L. After this run, simulation results were compared to the experimental data reported by Shemer et al.7 If the model overestimated the rate of removal of metronidazole, the value of k15 was reduced; if it under-estimated the rate of removal of metronidazole, the value of k15 increased. This was continued until the absolute relative error between the model prediction and any experimental data point was less than 10%. Using the above technique, the value of k15 was determined to be 1.98 × 109 M-1 s-1, which was in the expected range, and resulted in an average absolute relative error of 3.4% between the model predictions and the reported experimental data from Shemer et al.7 To test this result, a simulation was run utilizing the determined value of k15 with an initial H2O2 concentration of 50 mg/L. The average absolute relative error between the model prediction and the experimental data was 1.7%. On the basis of the low average absolute relative errors (e3.4%), the calibration of k15 to the value of 1.98 × 109 M-1.s-1 was selected for all subsequent model simulations. As shown in Figure 4, the calibrated model was able to predict the concentration of metronidazole versus time for both initial H2O2 concentrations. In addition, the model predicted a decreasing H2O2 concentration with increasing time, because H2O2 is consumed in the oxidation process. The concentration of hydroxyl radicals, which react directly with the target compound (see eq 15), at any time instant was quite low (on the order of 1 × 10-12 M), because this highly reactive species tends to be consumed almost as quickly as it is produced. In a separate figure (not shown), the residuals were plotted for error difference between the experimental and simulated results. It was obvious from the residual plots that there was no specific pattern of error for this model; therefore, it verified a good agreement between the dynamic model and the experimental data by a random gun shot pattern of the residuals.26 Furthermore, the reaction models were examined for the removal of clofibric acid, a metabolite of the lipid regulating drug clofibrate, utilizing the UV/H2O2 AOP to ensure the model developed would be valid for target compounds other than metronidazole. Experimental data from the open literature were used.6 The experimental data were based on the UV/H2O2

6530 Ind. Eng. Chem. Res., Vol. 47, No. 17, 2008

Figure 4. Metronidazole concentration vs time, calibrated model. Model predictions obtained utilizing a value for the reaction rate constant between metronidazole and the hydroxyl radical, k15, of 1.98 × 109 M-1 s-1, with experimental data from the literature.7

oxidation of a 0.024-0.0255 mM solution of clofibric acid in deionized water, with a pH of 5.5, in a batch cylindrical reactor with an outer diameter of 9.5 cm and a height of 28 cm. The reactor had an effective optical path length of 2.01 cm, volume of 0.42 L, and a LP UV lamp with an output of 2.7 × 10-6 mol photon.s-1 at 254 nm. Because deionized water was used, it was possible to assume that the solution used by Andreozzi et al.6 did not contain any inhibitory compounds, and thus the extinction coefficient for distilled water was used in the determination of the solution extinction coefficient. The reaction rate constant for the reaction between clofibric acid and the hydroxyl radical, k15, and the quantum yield for the direct photolysis of clofibric acid, φ16, were reported by Andreozzi et al.6 to be (2.38 ( 0.18) × 109 M-1 s-1 and 1.08 × 10-4 mol photon-1, respectively. These values were implemented into the calibrated model, and all other reaction rate constants and quantum yield values were kept constant. It was shown that (figure not shown) there is a good agreement between the modeled results and the experimental data for an initial H2O2 dose of 34 mg/L. In fact, the absolute relative error remained less than 10% for all data points up to 50% removal of the target compound (0-40 s of reaction time). Also, for the simulation done for an initial H2O2 dose of 340 mg/L, there was good agreement between the modeled results and the experimental data, with an absolute relative error less than 15% for all data points up to 67% removal of the target compound (0-40 s of reaction time). At higher removal percentages (>67%), the absolute relative error remained below 43%. 3.2. Effect of H2O2 as an Inhibitor. Above the optimal dose of H2O2, the rate of removal of organics decreases because of the hydroxyl radical scavenging behavior of H2O2.19,24 Model simulations were run with increasing initial H2O2 doses to determine if the model would express a similar behavior. The extinction coefficient for distilled water was used to determine the extinction coefficient of the aqueous solution for these simulations. As depicted in Figure 5, increasing the initial H2O2 dose increased the predicted removal of metronidazole up to an initial dose of approximately 50 mg/L. At dosages greater

than 50 mg/L, there was a reduction in the removal efficiency, indicating the inhibitory effect of H2O2 at concentrations above the optimal dose. As a result, the model was able to predict an optimal initial H2O2 dose, and the inhibitory effect of H2O2 at dosages greater than the optimal level. 3.3. Effect of Alkalinity as an Inhibitor. For practical application, the UV/H2O2 process could be used for the treatment of potable water supplies. Because surface and groundwater potable water sources can have varying levels of alkalinity, carbonate and bicarbonate ions naturally present can have a significant impact on the target compound removal. The Ontario Drinking Water Standards, Objectives, and Guidelines state an operational guideline for alkalinity of 30-500 mg/L.27 It was desired to model the impact of alkalinity levels within that range. Simulations were performed using the same calibrated model. Alkalinity concentrations of 75, 150, and 225 mg/L as CaCO3 were used, and it was assumed that the alkalinity was due to the presence of bicarbonate ions only, because at the solution pH of 6.0, the concentration of carbonate ions would be negligible. The extinction coefficient for typical tap water (10 m-1) was used to determine the overall extinction coefficient of the aqueous solutions for these simulations. All simulations were performed for an initial H2O2 concentration at 50 mg/L. All other parameters used were equal to those used in the model calibration. As illustrated in Figure 6, increasing alkalinity levels have an inhibitory effect on the model predicted removal of metronidazole. As the concentration of alkalinity was increased, there was a decrease in the predicted removal of metronidazole. 3.4. Effect of Alkalinity on the Optimal H2O2 Dose. Simulations were run in order to determine if the presence of alkalinity would impact the model predicted optimal H2O2 dose. An alkalinity concentration of 75 mg/L as CaCO3 was used for each run, and initial H2O2 doses were increased to determine the predicted optimal dose. As shown in Figure 7, increasing the initial H2O2 dose increased the predicted removal of metronidazole up to an initial dose of approximately 200 mg/ L. At dosages greater than 200 mg/L, there was a reduction in removal efficiency, indicating the inhibitory effect of H2O2 at

Ind. Eng. Chem. Res., Vol. 47, No. 17, 2008 6531

Figure 5. Effect of [H2O2]0 on model predicted removal of metronidazole, optimal H2O2 dose. Thicker lines represent conditions at or below the optimal H2O2 dose; thinner lines represent conditions above the optimal H2O2 dose.

Figure 6. Effect of alkalinity as CaCO3 on model predicted removal of metronidazole with [H2O2]o ) 50 mg/L.

concentrations above the optimal dose. It was shown that the optimal H2O2 dose for the system without the presence of alkalinity was approximately 50 mg/L, and that with the

presence of alkalinity the optimal dose was approximately 200 mg/L. On the basis of these results, it appears that the presence of alkalinity increases the model predicted optimal H2O2 dose.

6532 Ind. Eng. Chem. Res., Vol. 47, No. 17, 2008

Figure 7. Effect of [H2O2]o on model predicted removal of metronidazole with alkalinity of 75 mg/L as CaCO3, optimal H2O2 dose. Thicker lines represent conditions at or below the optimal H2O2 dose; thinner lines represent conditions above the optimal H2O2 dose.

3.5. Modeling Metronidazole Removal in Single Lamp Photoreactors, Reactors 1 and 2. The calibrated reaction model was used to predict the removal of metronidazole in reactors 1 and 2. For all modeling runs in the photoreactors, the initial concentration of metronidazole was 6 µM in an aqueous solution with an alkalinity of 75 mg/L as CaCO3. The extinction coefficient of the aqueous solution was assumed to be constant within the photoreactors, and was calculated on the basis of the extinction coefficient of tap water and in the influent concentration of H2O2. The UV absorptive properties of metronidazole were neglected because of its low influent concentration, and the influent concentration of H2O2 was used because it was assumed that very little H2O2 would be consumed, resulting in an almost uniform concentration of H2O2 throughout the reactor. As a preliminary investigation, the concentration profiles of metronidazole along the reactor radius at various retention times were developed for both reactor 1 and reactor 2. The initial H2O2 dose was kept constant at 100 mg/L. The results presented in Figure 8 predict that the removal of metronidazole increases as the residence time within reactor 1 increases. In addition, for any given time instant greater than 0 s, the concentration of metronidazole increases with increasing radial position. This is due to the fact that UV irradiance decreases with increasing radial position, because of increased distance from the UV lamp and the UV radiation adsorptive properties of the aqueous solution. Similar trends are observed for reactor 2 as shown in Figure 9. However, it can be seen that the concentration profiles for Reactor 2 show lower concentrations at any given time instant greater than 0 s and

radial position that those shown for reactor 1. Average effluent metronidazole concentrations for reactor 1 and reactor 2 were determined for each set of operating conditions and are summarized in Table 2. On the basis of the average effluent concentrations presented in Table 2, 4.9-9.8% of the metronidazole was removed in reactor 1, and 6.6-13% was removed in reactor 2. Little target compound was removed; however, removal was greater in reactor 2 for each set of operating conditions (22-39% more removal in reactor 2 than in reactor 1). This is due to the fact that the UV output of the lamp used in reactor 2 was 41% greater than that in reactor 1. In addition, the low removal of metronidazole implies that very few reaction intermediates were formed (see eq 16), thus the assumption that the effect of these intermediates were negligible appears to be valid. On the basis of the detailed modeling results for reactor 1 and reactor 2, at all points in the reactor, the concentration of H2O2 was greater than 80% of the influent H2O2 concentration. As a result, the assumption that very little H2O2 would be consumed, resulting in an almost uniform concentration of H2O2 throughout the reactor, also appears to be valid. For both reactors, the lowest average effluent metronidazole concentration was obtained with an influent H2O2 dose of 100 mg/L and a reactor radius of 100 mm (5.41 µM for reactor 1, and 5.22 µM for reactor 2). It should be noted, however, that for reactor 1 with an influent H2O2 concentration of 25 mg/L, the average effluent metronidazole concentration was lowest for two reactor radii: 100 and 150 mm. This implies that at lower influent H2O2 concentrations, the most efficient use of UV light energy may be provided with a larger reactor diameter. To test this result,

Ind. Eng. Chem. Res., Vol. 47, No. 17, 2008 6533

Figure 8. Concentration profile of metronidazole in reactor 1 at various residence times with [H2O2]inf ) 100 mg/L. Residence time of 0 s equivalent to the inlet of the reactor, 37 s equivalent to an axial position of 2.4 m, 84 s equivalent to an axial position of 5.5 m, and 150 s equivalent to an axial position of 9.7 m.

Figure 9. Concentration profile of metronidazole in reactor 2 at various residence times with [H2O2]inf ) 100 mg/L. Residence time of 0 s equivalent to the inlet of the reactor, 37 s equivalent to an axial position of 2.4 m, 84 s equivalent to an axial position of 5.5 m, and 150 s equivalent to an axial position of 9.7 m.

were performed simulations for reactor 1 with an influent H2O2 concentration of 10 mg/L. On the basis of these simulations, the lowest average effluent metronidazole concentration was obtained with two reactor radii: 150 and 200 mm. Thus, for a given influent H2O2 concentration, the reactor radius resulting in the lowest effluent metronidazole concentration varies with respect to the influent H2O2 concentration; as the influent H2O2 concentration is decreased, the

reactor radius resulting in the lowest average effluent metronidazole concentration is increased. 3.6. Modeling Metronidazole Removal in Multi-Lamp Photoreactor, Reactor 3. The calibrated reaction model was used to predict the removal of metronidazole in reactor 3. For all modeling runs in the photoreactor, the initial concentration of metronidazole was 6 µM in an aqueous solution with an alkalinity of 75 mg/L as CaCO3. As discussed earlier, the

6534 Ind. Eng. Chem. Res., Vol. 47, No. 17, 2008 Table 2. Average Effluent Metronidazole Concentrations for All Three Photoreactors at Various Operating Conditionsa influent H2O2 concentration (mg/L) residence time (s)

25

50

75

100

radius (mm)

R1 & R2

R3

R1

R2

R3

R1

R2

R3

R1

R2

R3

R1

R2

R3

50 100 150 200 250

8.3 37 84 150 235

5.1 33 81 146 231

5.71 5.58 5.58 5.61 5.65

5.60 5.43 5.44 5.50 5.56

5.18 4.21 4.06 4.15 4.31

5.60 5.46 5.48 5.54 5.60

5.45 5.27 5.33 5.42 5.50

4.88 3.75 3.68 3.87 4.10

5.55 5.42 5.47 5.54 5.60

5.38 5.23 5.31 5.42 5.51

4.74 3.60 3.61 3.85 4.12

5.53 5.41 5.48 5.56 5.62

5.35 5.22 5.33 5.45 5.54

5.35 5.22 5.33 5.45 5.54

a All reported average effluent metronidazole concentrations are in µM. Concentration in bold represents the lowest average effluent concentration observed for each reactor. R1, R2, and R3 represent reactors 1, 2, and 3, respectively.

LVREA at any point in reactor 3, A, is the sum of the contributions of all four lamps. The shielding of photons by the lamps and quartz tubes was neglected since the diameter of the UV tubes and also the distance of the UV penetration are small. Due to the symmetry of reactor 3, it was only necessary to model one quadrant of the reactor. As with reactors 1 and 2, the extinction coefficient of the aqueous solution was assumed to be constant within the photoreactor, and was calculated on the basis of the extinction coefficient of tap water and in the influent concentration of H2O2. The UV absorptive properties of metronidazole were neglected because of its low influent concentration, and the influent concentration of H2O2 was used because it was assumed that very little H2O2 would be consumed, resulting in an almost uniform concentration of H2O2 throughout the reactor. As a preliminary investigation, the concentration profile of metronidazole for a quadrant of reactor 3 was developed at the midpoint of reactor 3, and the effluent of reactor 3 with an influent H2O2 dose of 100 mg/L and a reactor radius of 100 mm. The results presented in Figure 10 predict increased removal of metronidazole in the regions closest to the UV lamps. A similar result was obtained for reactor 1 and reactor 2. In addition, removal was enhanced in regions of the reactor that have high LVREA values because of the additive contribution of multiple UV lamps. For example, in Figure 10b, the effluent concentration in the region around the coordinates (22 mm, 22 mm) is approximately 2.5 µM, and approximately 3.5 µM around the coordinates (78 mm, 22 mm). These two regions are a similar distance away from the UV lamp; however, the LVREA around the coordinates (22 mm, 22 mm) is enhanced because of the additive effect of the two UV lamps, resulting in a 40% increase in the removal of metronidazole as compared to around the coordinates (78 mm, 22 mm). Average effluent metronidazole concentrations for Reactor 3 were determined for various sets of operating conditions, and are summarized in Table 2. Simulations were performed for a given influent H2O2 concentration with increasing outer reactor radius until it was ascertained that the average effluent concentration reached a minimum. On the basis of the average effluent concentrations presented in Table 2, 14-41% of the metronidazole was removed in reactor 3. The removal was greater in reactor 3 than in either reactor 1 or reactor 2. This is due to the fact that reactor 3 is equipped with four UV lamps, each emitting 36 W of UV output, resulting in a greater UV output than that in either reactor 1 or reactor 2. Ninety percent removal of metronidazole was predicted for seven reactors of the same configuration as reactor 3 operated in series. As with reactor 1 and reactor 2, the lowest average effluent metronidazole concentration for reactor 3 for a given reactor radius was obtained at the highest modeled influent H2O2 concentration of 100 mg/L. The lowest average effluent concentrations of metronidazole attained for reactor 3, 3.54 µM, was with a radius of 100 mm and an influent H2O2 concentration of 100 mg/L. In addition, it

can be seen that, for a given influent H2O2 concentration, the reactor radius resulting in the lowest effluent metronidazole concentration varies with respect to the influent H2O2 concentration. This implies that selection of a reactor outer radius for maximum metronidazole removal varies with influent H2O2 concentration for the multiple UV lamp reactor. As presented in section 2.4, the reaction rates for the photoreactors are functions of not only residence time, but also the position within the reactor. The formation of hydroxyl radicals is dependent on the absorption of UV radiation by the hydrogen peroxide molecule. As one moves further from the UV lamp, the radiant energy flux decays exponentially (see eq 2). As a result, as Table 2 shows, the higher residence time provided by reactors with larger radii is offset by the decrease in UV radiation being absorbed by the peroxide molecules. This “offset” is what results in there being an optimal reactor radius for a given set of influent conditions. The increase in hydrogen peroxide concentration also results in an increase in the UV absorptive nature of the solution (see eq 3). This will impact the radiant energy flux distribution throughout the reactor. 3.7. Local Optimal Initial H2O2 Dose and Photoreactor Radius. It has been determined in this work that the average effluent metronidazole concentration for a given reactor radius could be lowered by increasing the UV lamp output, and increasing the influent H2O2 concentration. Increasing these operational parameters also increases the operational cost, hence it was desired to investigate the operational cost per mmol of metronidazole removed to determine the optimal operating conditions for each reactor studied. Using eq 28, operational costs per mmol of metronidazole removed were determined for various operating conditions for each reactor studied. These costs were based on average influent and effluent H2O2 concentrations, as well as the number of UV lamps in operation and their total wattage. As shown in Table 3, the operational costs for each reactor for different influent H2O2 concentrations at various radii were calculated. Table 4 presents a summary of the local optimal operating conditions determined for each reactor. It was found that the optimal influent H2O2 concentration, [H2O2]/inf, was the same for all three reactors, at 25 mg/L. The lowest operational cost per mmol of metronidazole removed was projected for reactor 3, whereas the highest was projected for reactor 1. On the basis of the results presented in Table 4, for a given set of operating conditions, the operational cost per mmol of metronidazole removed was higher for reactor 1 than for reactor 2. Because the only difference between reactor 1 and reactor 2 is the UV output of the lamp, this implies that reduced operational costs can be achieved by increasing the UV output of the lamp, in spite of the increase in electricity costs. A similar trend was observed when comparing the results for reactor 2 and reactor 3. For all reactors, increasing the influent H2O2 concentration while keeping the reactor radius constant resulted in an increase in operational costs per mmol of metronidazole removed. Con-

Ind. Eng. Chem. Res., Vol. 47, No. 17, 2008 6535

Figure 10. Metronidazole concentration profile for a quadrant of reactor 3 with [H2O2]inf ) 100 mg/L and R ) 100 mm at two axial positions within the reactor. Treatment of a 6 µM metronidazole solution with an alkalinity of 75 mg/L and pH of 6, at a flow rate of 2.0 L/s: (a) Shown above at a distance of 1.2 m from inlet of reactor 3, at an equivalent residence time of 17 s, with an average metronidazole concentration of 4.56 µM; and (b) shown above at the outlet of reactor 3, a distance of 2.4 m from the inlet, at an equivalent residence time of 33 s, with an average metronidazole concentration of 3.54 µM.

versely, it has been shown that an increase in influent H2O2 concentration resulted in an increase in removal of metronidazole. These results imply that the chemical costs associated with increasing influent H2O2 concentration are much greater than the electrical costs associated with increasing UV lamp output. As a result, from an operational perspective, it would be costeffective to increase the UV lamp output while keeping influent

H2O2 concentration low. Such a strategy, however, would have impacts on the capital costs for such an installation. 4. Conclusions A dynamic kinetic model for the aqueous degradation of metronidazole by UV/H2O2 process was developed and using

6536 Ind. Eng. Chem. Res., Vol. 47, No. 17, 2008 Table 3. Operational Cost Per Amount of Metronidazole Removed in All Three Photoreactors at Various Operating Conditionsa influent H2O2 concentration (mg/L) residence time (s) radius (mm) 50 100 150 200 250

25

50

75

100

R1 & R2

R3

R1

R2

R3

R1

R2

R3

R1

R2

R3

R1

R2

R3

8.3 37 84 150 235

5.1 33 81 146 231

0.33 0.22 0.22 0.24 0.27

0.24 0.17 0.17 0.19 0.21

0.13 0.06 0.05 0.06 0.06

0.47 0.34 0.36 0.40 0.46

0.34 0.26 0.28 0.32 0.37

0.18 0.09 0.08 0.09 0.10

0.61 0.47 0.52 0.60 0.70

0.45 0.36 0.40 0.48 0.57

0.23 0.12 0.12 0.13 0.15

0.77 0.63 0.70 0.83 0.98

0.57 0.47 0.55 0.67 0.80

0.29 0.15 0.16 0.18 0.21

a All reported operational costs are in $/mmol metronidazole removed. Concentrations in bold represent the lowest operational cost for each reactor. Metronidazole initial concentration was 6 µM, and an alkalinity was 75 mg/L at a flow rate of 2.0 L/s.

Table 4. Local optimal Operating Conditions in Terms of Minimum Operational Cost Per Mmole of Metronidazole Removeda local optimal parameter influent H2O2 concentration, * [H2O2]inf (mg/L) reactor radius (R*) (mm) average effluent metronidazole concentration (µmol/L) operational cost per mmol of metronidazole removed ($/mmol)

reactor 1

reactor 2

25

25

100-150 5.58

100-150 5.43-5.44

0.22

0.17

reactor 3 25 150 4.06 0.05

a Metronidazole initial concentration was 6 µM, and alkalinity was 75 mg/L at a flow rate of 2.0 L/s.

the experimental data, the reaction rate constant for the reaction between metronidazole and hydroxyl radicals was determined to be 1.98 × 109 M-1 s-1. The calibrated reaction model was able to predict an optimal initial H2O2 dose, the inhibitory effect of H2O2 doses above the optimal value, and the inhibitory effect due to the presence of alkalinity as bicarbonate ions. In addition, the calibrated reaction model predicted an increase in the optimal initial H2O2 dose with the presence of alkalinity in the solution. It was also shown that the reaction model was able to accurately predict the removal of clofibric acid on the basis of the experimental results from the open literature. As a result, the calibrated reaction model was used to predict the removal of metronidazole in the single and multiple UV lamp photoreactors. It was shown that 4.9-9.8, 6.6-13, and 14-41% of the metronidazole was removed in reactors 1, 2, and 3, respectively. Keeping the reactor radius constant, the average effluent metronidazole concentration could be lowered by increasing the UV lamp output, and increasing the influent H2O2 concentration. In addition, the selection of a reactor radius for maximum metronidazole removal varied with influent H2O2 concentration. Furthermore, the optimal influent H2O2 concentration, [H2O2]/inf, and outer reactor radius, R*, resulting in the lowest operational cost per mmol of metronidazole removed, were determined for each photoreactor. The lowest operational cost per mmol of metronidazole removed was projected for reactor 3, at $0.05/ mmol, while the highest was projected for reactor 1, at $0.22/ / mmol. The optimal influent H2O2 concentration, [H2O2]inf , was the same for all three reactors, at 25 mg/L. Acknowledgment The support of Natural Sciences and Engineering Research Council of Canada (NSERC) is greatly appreciated. Nomenclature / [H2O2]inf ) local optimal influent H2O2 concentration (mg/L) [H2O2]inf ) influent H2O2 concentration (mg/L) [H2O2]o ) initial H2O2 concentration (mg/L) A ) local volumetric rate of energy absorption (mol photons/(L

s)) j ) average local volumetric rate of energy absorption (mol A photons/(L s)) Ci ) concentration of species i (mM) Ci,eff ) effluent concentration of species i (mM) DAB ) diffusivity of the species in the aqueous solution (m2/s) fi ) fraction of photons absorbed by species i (dimensionless) F ) operational cost per amount of target compound ($/mmol) k ) reaction rate constant (M-1 s-1) L ) reactor length (m) Q ) flow rate (L/s) q ) radiant energy flux (mol photons/(m2 s)) qo ) radiant energy flux at liquid surface (mol photons/(m2 s)) r ) radial position (m) R ) outer reactor radius (m) R* ) local optimal outer reactor radius (m) Ri ) inner reactor radius (m) RH ) hydraulic radius (m) Rrxn,i ) reaction rate for species i (mM/s) S ) cross-sectional area (m2) SC ) chemical operating costs ($/s) SE ) electrical operating costs ($/s) t ) time (s) V ) velocity (m/s) V ) volume (m3) x ) position on x-axis for multilamp reactor (m) y ) position on y-axis for multilamp reactor (m) z ) axial position (m) Greek Letters τ ) reactor residence time (s) φ ) quantum yield (mol photon-1) F ) density (kg/m3) εRH ) molar absorptivity of the target compound at 254 nm (mM-1 m-1) εH2O2 ) molar absorptivity of H2O2 at 254 nm (mM-1 m-1) µs ) solution extinction coefficient at 254 nm (m-1) µw ) extinction coefficient of water at 254 nm (m-1)

Literature Cited (1) Larsen, T. A.; Lienert, J.; Joss, A.; Siegrist, H. How to avoid pharmaceuticals in the acquatic environment. J. Biotechnol. 2004, 113, 295. (2) Zwiener, C.; Frimmel, F. H. Oxidative treatment of pharmaceuticals in water. Water Res. 2000, 34, 1181. (3) Gagne, F.; Blaise, C.; Andre, C. Occurrence of pharmaceutical products in a municipal effluent and toxicity to rainbow trough (Oncorhynchus mykiss) hepatocytes. Exotoxicol. EnViron. Saf. 2006, 64, 329. (4) Carballa, M.; Omil, F.; Lema, J. M.; Llompart, M.; Garcia-Jares, C.; Rodgriguez, I.; Gomez, M.; Ternes, T. Behaviour of pharmaceuticals, cosmetics and hormones in a sewage treatment plant. Water Res. 2004, 38, 2918. (5) Hernando, M. D.; Mezcua, M.; Fernandez-Alba, A. R.; Barcelo, D. Environmental risk assessment of pharmaceutical residues in wastewater effluents, surface waters and sediments. Talanta 2006, 69, 334.

Ind. Eng. Chem. Res., Vol. 47, No. 17, 2008 6537 (6) Andreozzi, R.; Caprio, V.; Marotta, R.; Radovnikovic, A. Ozonation and H2O2/UV treatment of clofibric acid in water: a kinetic investigation. J. Hazard. Mater., B 2003, 103, 233. (7) Shemer, H.; Kunukcu, Y. K.; Linden, K. G. Degradation of the pharmaceutical Metronidazole via UV, Fenton, and photo-Fenton processes. Chemosphere 2006, 63, 269. (8) Gernjak, W.; Maldonado, M. I.; Malato, S.; Caceres, J.; Krutzler, T.; Glaser, A.; Bauer, R. Pilot-plant treatment of olive mill wastewater (OMW) by solar TiO2 photocatalysis and solar photo-Fenton. Sol. Energy 2004, 77, 567. (9) Andreozzi, R.; Marotta, R.; Pinto, G.; Pollio, A. Carbamazepine in water: persistence in the environment, ozonation treatment and preliminary assessment on algal toxicity. Water Res. 2002, 36, 2869. (10) Tabrizi, G. B.; Mehrvar, M. Pilot-plant study for the photochemical treatment of aqueous linear alkylbenzene sulfonate. Sep. Purif. Technol. 2006, 49, 115. (11) Asadi, A.; Mehrvar, M. Degradation of aqueous methyl tert-butyl ether by photochemical, biological, and their combined processes. Int. J. Photoenergy 2006, Article ID 19790, 1. (12) Kummerer, K.; Al-Ahmad, A.; Mersch-Sundermann, V. Biodegradability of some antibiotics, elimination of the genotoxicity and affection of wastewater bacteria in a simple test. Chemosphere 2000, 40, 701. (13) Gomez, M. J.; Petrovic, M.; Fernandez-Alba, A. R.; Barcelo, D. Determination of pharmaceuticals of various therapeutic classes by solidphase extraction and liquid chormoatography - tandem mass spectrometry analysis in hospital effluent wastewaters. J. Chromatogr., A 2006, 1114, 224. (14) Behnajady, M. A.; Modirshahla, N. Kinetic modelin on photooxidative degradation of C.I. Acid Orange 7 in a tubular continuous-flow photoreactor. Chemosphere 2006, 62, 1543. (15) Aleboyeh, A.; Moussa, Y.; Aleboyeh, H. The effect of operational parameters on UV/H2O2 decolourisation of Acid Blue 74. Dyes Pigm. 2005, 66, 129. (16) Shu, H.-Y.; Chang, M.-C. Pilot scale annular plug flow photoreactor by UV/H2O2 for the decolorization of azo dye wastewater. J. Hazard. Mater., B 2005, 125, 244.

(17) Shu, H.-Y.; Fna, H.-J.; Chang, M.-C.; Hsieh, W.-P. Treatment of MSW landfill leachate by a thin gap annular UV/H2O2 photoreactor with multi-UV lamps. J. Hazard. Mater., B 2006, 129, 73. (18) Beltran, J.; Rivas, J.; Alvarez, P. M.; Alonso, M. A.; Acedo, B. (1999). A kinetic model for advanced oxidation processes of aromatic hydrocarbons in water: application to phenanthrene and nitrobenzene. Ind. Eng. Chem. Res. 1999, 38, 4189. (19) Shen, Y.-S.; Wang, D. K. Development of photoreactor design equation for the treatment of dye wastewater by UV/H2O2 process. J Hazard. Mater., B 2002, 89, 267. (20) Perry, R. H.; Green, D. W. Perry’s Chemical Engineers’ Handbook, 7th ed.; McGraw-Hill: New York, 1997. (21) Gadgil, A. UV Waterworks 2.0-Answers to Ten Commonly Asked Questions about the Design, Operation, And Economics; Energy & Environment Division, Lawrence Berkeley National Laboratory: Berkeley, CA, 1995. (22) Mehrvar, M.; Anderson, W. A.; Moo-Young, M. Comparison of the photoactivities of two commercial titanium dioxide powders in the degradation of 1,4-dioxane. Int. J. Photoenergy 2002, 4, 41. (23) Gernjak, W. Solar photo-Fenton Treatment of EU Priority Substancessprocess Parameters and Control Strategies. Ph.D. Thesis, Universitat fur Bodenkultur Wien, Vienna, Austria, 2006. (24) Mehrvar, M.; Anderson, W. A.; Moo-Young, M. Photocatalytic degradation of aqueous organic solvents in the presence of hydroxyl radical scavengers. Int. J. Photoenergy 2001, 3, 187. (25) Hancu, D.; Beckman, E. J. Generation of hydrogen peroxide directly from H2 and O2 using CO2 as the solvent. Green Chem. 2001, 3, 80. (26) Mehrvar, M.; Anderson, W. A.; Moo-Young, M.; Reilly, P. M. Non-linear parameter estimation for a dynamic model in photocatalytic reaction engineering. Chem. Eng. Sci. 2000, 55, 4885. (27) Ontario Ministry of the Environment. Technical Support Document forOntarioDrinkingWaterStandards,ObjectivesandGuidelines.www.ene.gov.on.ca (Accessed September 5, 2007).

ReceiVed for reView December 2, 2007 ReVised manuscript receiVed May 28, 2008 Accepted July 1, 2008 IE071637V