Photocatalytic Degradation of Water Contaminants in Multiple

Oct 25, 2013 - Dolat , D.; Quici , N.; Kusiak-Nejman , E.; Morawski , A. W.; Li Puma , G. .... Andrea Turolla , Domenico Santoro , John R. de Bruyn , ...
0 downloads 0 Views 1MB Size
Article pubs.acs.org/est

Photocatalytic Degradation of Water Contaminants in Multiple Photoreactors and Evaluation of Reaction Kinetic Constants Independent of Photon Absorption, Irradiance, Reactor Geometry, and Hydrodynamics Ivana Grčić† and Gianluca Li Puma*,‡ †

Faculty of Chemical Engineering and Technology, University of Zagreb, Marulicev trg 19, 10000 Zagreb, Croatia Environmental Nanocatalysis and Photoreaction Engineering, Department of Chemical Engineering, Loughborough University, Loughborough LE11 3TU, United Kingdom



S Supporting Information *

ABSTRACT: The literature on photocatalytic oxidation of water pollutants often reports reaction kinetic constants, which cannot be unraveled from photoreactor type and experimental conditions. This study addresses this challenging aspect by presenting a general and simple methodology for the evaluation of fundamental “intrinsic” reaction kinetic constants of photocatalytic degradation of water contaminants, which are independent of photoreactor type, catalyst concentration, irradiance levels, and hydrodynamics. The degradation of the model contaminant, oxalic acid (OA) on titanium dioxide (TiO2) aqueous suspensions, was monitored in two annular photoreactors (PR1 and PR2). The photoreactors with significantly different geometries were operated under different hydrodynamic regimes (turbulent batch mode and laminar flow-through recirculation mode), optical thicknesses, catalyst and OA concentrations, and photon irradiances. The local volumetric rate of photon absorption (LVRPA) was evaluated by the six-flux radiation absorption-scattering model (SFM). The SFM was further combined with a comprehensive kinetic model for the adsorption and photodecomposition of OA on TiO2 to determine local reaction rates and, after integration over the reactor volume, the intrinsic reaction kinetic constants. The model could determine the oxidation of OA in both PR1 and PR2 under a wide range of experimental conditions. This study demonstrates a more meaningful way for determining reaction kinetic constants of photocatalytic degradation of water contaminants.



INTRODUCTION Advanced oxidation processes (AOPs) are characterized by the generation of strongly oxidative radical species, such as singlet oxygen, perhydroxyl, superoxide, sulfate, persulfate, and hydroxyl radicals. Among different AOPs, heterogeneous photocatalysis has emerged as a green and sustainable process for the treatment and purification of water and wastewater. The n-type semiconductor, titanium dioxide (TiO2), and its modifications are the most common photocatalysts due to their strong oxidizing power, relative nontoxicity, and long-term photostability.1 However, despite extensive research, the transfer of this technology to commercial products has been slow, particularly regarding water treatment and purification. One reason has been attributed to the small quantum efficiencies realized with current photocatalytic materials, which implies high treatment cost. The development of advanced photocatalytic materials with wider band-gaps and enhanced charge separation capabilities could yield faster degradation of water contaminants.2−6 Another reason is the unawareness of the process of scaling-up of photocatalytic processes from laboratory to industrial scale, combined with the © 2013 American Chemical Society

lack of standardized and simple methodologies for the design and optimization of photocatalytic reactors. Indeed, the interaction between photons, catalyst, and reactants involve complex physical and chemical phenomena that are difficult to model in simple terms.7−11 Further development of photocatalytic water purification at industrial scale will be driven by new legislation on discharge of environmental contaminants of emerging concern such as disinfection byproducts, pharmaceuticals, estrogens, and personal care products.12−16 The large literature on photocatalytic oxidation of water pollutants also presents significant difficulty in the interpretation of the kinetic data because these have been generated with photoreactors with different geometries, dimensions, radiation sources, irradiances, flow conditions, and hydrodynamics (batch, flow-through, reciculation, laminar, or turbulent flow). For example, the decomposition rate constants of water Received: Revised: Accepted: Published: 13702

August 5, 2013 October 20, 2013 October 25, 2013 October 25, 2013 dx.doi.org/10.1021/es403472e | Environ. Sci. Technol. 2013, 47, 13702−13711

Environmental Science & Technology

Article

continuously passed through the reactor and were fed back to a well-mixed recirculation vessel.27 PR2 was a batch annular photocatalytic reactor with an irradiated volume of 0.5 L made of borosilicate glass. The reactor was fitted with a sampling port, magnetic stirrer, and water jacket for temperature control. The irradiation source was a low-pressure mercury discharge UVA lamp (PenRay 90-001904, 365 nm) (Figure S2, Supporting Information), which was immersed axially inside a quartz tube located in the center axis of the reactor.28 Experimental Methodology. Aqueous solutions were made with high purity water produced by a Millipore Milli Q water purification system (resistivity = 18 M cm−1, total organic carbon < 1 ppb). Oxalic acid dihydrate (Fluka, TraceSELECT ≥ 99.9999%) was the modeling contaminant (10−40 ppm initial concentration). In a typical experiment, a specified amount of TiO2 (Evonik P25, primary particle size, 20−30 nm by transmission electron microscopy analysis; specific surface area, 52 m2 g−1 by Brunauer−Emmett−Teller surface area analysis; composition, 78% anatase and 22% rutile by X-ray diffraction analysis) was added to an OA aqueous solution. Two liters of the resulting suspension were transferred to a continuously stirred recirculation tank (PR1), where oxygen was constantly bubbled (40 mL min−1) and recirculated passing through the reactor. In contrast, in PR2, 0.5 L of the suspension was transferred directly into the reactor, and mixing was established to obtain a homogeneous suspension. Oxygen was supplied by the headspace (8 ppm in solution) and was never limiting the oxidation of OA. The suspensions, in both PR1 and PR2 setups, were equilibrated in the dark for 30 min before irradiation. After such equilibration time, the concentration of OA in solution measured by HPLC was practically at equilibrium. The systems were operated at ambient temperature (T = 22 °C) at a pH between 3.5 and 4 depending on the concentration of OA used. The pH changed less than 7% during the photocatalytic experiments. Samples were collected at appropriate time intervals during the reaction time lasting between 60 and 240 min. They were filtered through 0.45 μm nylon (Sartorius) or cellulose filters (CHROMAFIL RC 45/ 25) to remove the TiO2 and were promptly analyzed by HPLC. The chromatographs were an Agilent 1100 Series, with SUPELCOSIL LC-8 column and 0.2 M phosphoric acid mobile phase flowing at 1.0 mL min−1, and a Shimadzu, with a SUPELCOGEL H carbohydrate column, length 250 mm, internal diameter 4.6 mm. The flow rate of the mobile phase (0.5% phosphoric acid) was 0.15 mL min−1. The OA peak was monitored at 210 nm. UV Lamp Irradiance Calculation. The radiation fluxes (irradiances) emerging from the lamps’ walls were measured radiometrically for each power supply setting. Readings were taken along the axial direction and circumference of the lamp bulb, and the results were averaged across the axial and radial direction. Because the emission spectrum of the UVA lamps (Figures S1−S2, Supporting Information) spanned over a range wider than the spectral response of the 365 nm UVA sensor, the total UV irradiance at the wall of the lamp, averaged across the useful emission spectrum, was estimated from eq 1

contaminants determined from kinetic models that neglect the effects of irradiance, photon absorption, and scattering cannot be compared among studies. Furthermore, these kinetic constants cannot be used for the rational design of photocatalytic treatment processes or for the evaluation of intrinsic catalyst activity and quantum yield. Such apparent limitations can be addressed through the development of simple mathematical tools that can characterize the degradation of water contaminants in photoreactors and that can represent the transport of photons through the photocatalytic reactor. Using this approach, a number of studies have been proposed to model slurry photocatalytic reactors for water purification of thin-film and tubular geometry.11,17−20 The photoreactor radiation field can be modeled with the two-f lux21 or the sixf lux22 absorption-scattering model or more rigorously by solving the radiative transfer equation.23 However, these studies have used specific photoreactors and have not been extended to the modeling of the removal of contaminants from photoreactors with different geometries or operational modes.17 In this study, a comprehensive kinetic model for the adsorption and photodecomposition of the model contaminant, oxalic acid (OA), on TiO2 aqueous suspensions24−26 was combined with a validated radiation model (six-flux absorptionscattering model (SFM)) to predict the degradation of oxalic acid in two annular photoreactors. To demonstrate the general nature of the modeling approach, the reactor diameters and optical thicknesses of the two photoreactors were purposely significantly diverse, as well as their operating hydrodynamic regimes, which were turbulent batch mode and laminar flowthrough recirculation mode. The methodology proposed yields more meaningful intrinsic reaction rate constants of photocatalytic degradation of water contaminants, which are independent of radiation field, hydrodynamic conditions, and reactor geometry.



EXPERIMENTAL SECTION Photocatalytic Reactor Setups. Two photoreactor setups (PR1 and PR2) were constructed. The reactors and lamps specifications are summarized in Table S1 of the Supporting Information. PR1 was a batch recirculation system consisting of an annular photoreactor, a pulse-free peristaltic pump, and a well-mixed recirculation vessel. The outer wall of the photoreactor was a QVF borosilicate glass tube (internal diameter 0.038 m, wall thickness 0.0045 m), while the inner wall was a Pyrex tube (external diameter 0.026 m) mounted at the axial center of the reactor. The longitudinal length of the irradiated zone in PR1 was 0.255 m. The irradiated volume of suspension (Vreactor) was 0.134 L. A Philips blacklight-blue TL 8W/08 F8 T5/BLB lamp (nominal power 8 W, bulb length 0.213 m, bulb radius 0.007525 m) emitting radiation between 323 and 406 nm with a peak at 365 nm (Figure S1, Supporting Information) was mounted axially inside the Pyrex tube (cutoff wavelength 300 nm) and centered on the longitudinal length of the reactor. The lamps’ ends were covered with Teflon to expose a 0.213 m long section with an approximately constant photon irradiance (5% error). The lamp’s photon irradiance was modulated with a variable power supply unit and monitored with a radiometer (Cole-Parmer) and a 365 nm UVA sensor (range 355−375 nm, sensor accuracy ±5%, display accuracy ±0.2% at full scale). The photocatalytically useful UVA irradiance (323 nm < λ < 384 nm) at the lamp wall and at the maximum power supply setting reached 90 W m−2. PR1 was operated in a recirculating batch mode, whereby reactants

384nm

I w = Isensor

13703

∫λ

min

384nm

∫λ

min

Wλdλ

WλPλdλ

(1)

dx.doi.org/10.1021/es403472e | Environ. Sci. Technol. 2013, 47, 13702−13711

Environmental Science & Technology

Article

Table 1. Parameters of Six Flux Absorption Scattering Model parameter

equation

a = 1 − ωpf −

b

b = ωpb +

τ

τ = (σ + κ )ccatR(1 − η)

τapp

ω=

ωcorr

γ=

where Isensor is the incident radiation measured with the radiometer, Pλ the relative spectral response of the 365 nm sensor, Wλ the radiant power of the lamp at wavelength λ, with the integration limits of λmin as the minimum wavelength emitted by the lamp, and λ = 384 nm as the highest wavelength that produces electron−hole couples in the TiO2 P25 photocatalyst. For example, in the PR1 setup, the integral quotient in eq 1 was equal to 1.815 for the specific UV lamp and sensor used here. Such corrections are encouraged in literature studies employing radiometric readings.

=

Ir , z

∫(H−L)/2

1−

2 1 − ωcorr

1+

2 1 − ωcorr

exp(− 2τapp)

⎤ ⎡β SL ⎧ ⎨arctan⎢ (2αz* − α + 1)⎥ ⎦ ⎣ 4πηR ⎩ 2 ⎤⎫ ⎡β − arctan⎢ (2αz* − α − 1)⎥⎬ ⎦⎭ ⎣2

LVRPA = (2)

(4)

τappI0

η ωcorr(1 − γ ) (1 − η)R [(ωcorr − 1 +

where SL is the radiation emission of the lamp per unit time and unit length of the lamp

SL = 2πrLI w

b a

where η is the ratio of internal to external radius of the annulus, α = H/L and β = L/ηR are the geometrical design parameters of an annular photoreactor, and z* = z/H is the dimensionless axial coordinate. Radiation Absorption and Scattering Model. The radiation field inside the aqueous suspension of catalyst was modeled by the six-flux radiation absorption-scattering model (SFM).22,27,30 Although more complex or simpler radiation models could be employed, the SFM strikes a balance between the simplicity of the modeling equations, which allows an integral analytical estimation of the local volumetric rate of photon absorption (LVRPA) at each location in the reactor, and the accuracy of representation of the radiation field. The assumptions of the SFM are as follows: (1) The photocatalytic particles are distributed uniformly within the reaction space. (2) There is negligible absorption of radiation by the fluid and by the dissolved species. (3) Photons are either absorbed or scattered upon colliding with a photocatalytic particle. (4) Scattering follows the route of one of the six directions of the Cartesian coordinates. (5) Geometric optics applies. For an infinitely long annular photocatalytic reactor, the LVRPA at a point (r, z) in the reaction space calculated with the SFM in cylindrical coordinate is31

Four models of radiation emission, radiation absorptionscattering, fluid-dynamics, and reaction kinetics were combined to the reactor material and energy balances and integrated to evaluate the degradation kinetics of OA. Radiation Emission Model. The model by Li Puma et al.27 for an annular photocatalytic reactor was adopted in this study. The cylindrical lamp, which is located in the center axis of the annulus, is modeled as a line source with each point emitting radiation in every direction and isotropically.29 Furthermore, the radiation emitted by each point of the lamp is assumed constant along the axial direction of the lamp. Such assumptions produce simple mathematical equations describing the physical behavior of radiation transfer in the tri-dimensional space. Indeed, the lamps employed had a high length to radius ratio. Under these assumptions, eq 2 models the radiation field in the volumetric space between the lamp wall and the inner wall of the annulus (Figure S3, Supporting Information) 1 dx′ 2 r + (z − x′)2

(1 − ωpf − ωpb − 2ωps )

I0 = I(ηR), z*

PHOTOCATALYTIC REACTOR MODELING

(H + L)/2

4ω2ps2

σ σ+κ

ωcorr =

γ

S = L 4π

(1 − ωpf − ωpb − 2ωps )

2 τapp = aτ 1 − ωcorr

ω



4ω2ps2

a

2 1 − ωcorr )exp( −τappξ∗)

+ γ(ωcorr − 1 −

2 1 − ωcorr )exp( −τappξ∗)]

(5)

where I0 is described by eq 4, τapp is the apparent optical thickness, ωcorr is the corrected scattering albedo, γ is a SFM dimensionless parameter, and ξ* is the dimensionless reaction space radial coordinate (Figure S1, Supporting Information). The model parameters are shown in Table 1. σ and κ are the

(3)

and Iw is the irradiance measured at the lamp wall (eq 1). The incident photon flux (irradiance) at the inner wall of the reactor can then be written using dimensionless parameters:27 13704

dx.doi.org/10.1021/es403472e | Environ. Sci. Technol. 2013, 47, 13702−13711

Environmental Science & Technology

Article

Table 2. Optical and Geometrical Parameters Used To Estimate LVRPA in Photoreactors with SFM photoreactor model parameter scattering albedo, ω specific absorption coefficient, κ, m2 kg−1b specific scattering coefficient, σ, m2 kg−1b optical thickness, τa apparent optical thickness, τapp geometrical parameters

PR1

b

PR2 0.862 187 1163

3.24 1.87 1.197 16.385

H/L L/ηR

scattering probabilities

16.20 10.52 1.766 6.725

(1) pb = 0.71 pf = 0.11 ps = 0.045 (diffusively reflecting phase function) (2) pb = pf = ps = 1/6 (isotropical phase function)

Given values are calculated for ccat = [TiO2] = 0.4 g L−1. In some cases, TiO2 loading was varied (0.1 to 1 g L−1) for the purpose of the model validation. bTiO2 P25 optical parameters.31 a

specific scattering and absorption coefficients per unit mass of catalyst, respectively, which can be evaluated by averaging over the useful spectrum of the incident radiation. pf, pb, and ps are SFM model parameters that correspond to the probabilities of forward, backward, and side scattering, respectively. These parameters depend on the adopted scattering phase function. For isotropic scattering, they take the values pf = pb = ps = 1/6. For a diffusively reflecting phase function, they can be estimated by a Monte Carlo approach (Table 2).30 LVRPA and Fluid-Dynamic Model. PR1 was operated in the laminar flow regime. Because the concentration of OA within the reactor varied both radially and axially, the rate of oxidation of OA was computed locally (ξ*, z*) using the LVRPA calculated from eq 5. The two distinct cases of isotropic scattering and of diffusively reflecting scattering phase function were evaluated. In contrast, because PR2 was a wellmixed reactor, the LVRPA was averaged across the reactor volume and treated as a lumped parameter in the evaluation of the rate of OA oxidation. This is because the high turbulence generated by the mixer provides sufficient transversal mixing in the reactor, continuously exposing new catalyst to areas of high irradiance (such condition was not realized in PR1). Therefore, the LVRPA estimated from eq 5 was averaged over the reactor volume to calculate LumpVRPA (eq 6) and was further used for the modeling of the photocatalytic oxidation of OA in PR2. Lump VRPA(t ) =

1 VR

∫V (LVRPA)dVR R

2Q

vz =

2⎡

πR ⎣⎢(1 − η 4 ) −

(1 − η2)2 ⎤ ⎥ ln(1/η) ⎦

⎤ ⎡ ⎛ 1 − η2 ⎞ × ⎢1 − (r*)2 + ⎜ ⎟ln(r *)⎥ ⎥⎦ ⎢⎣ ⎝ ln(1/η) ⎠

(7)

where Q is the volumetric flow rate through the annulus. The hydraulic diameter for PR1 is 0.014 m, and the Reynolds number for the flow conditions used in the experiments is 1192 in the laminar flow regime. In contrast, PR2 was operated under a fully developed turbulent flow. The Reynolds number for the flow developed under magnetic stirring was estimated from eq 8:33 2

( 2a )

Ω Re =

ν

(8)

where Ω represents a rotation frequency, and a is the length of the magnetic stirring bar. For the stirring conditions used (a = 3.5 cm; 650 rpm) the Reynolds number was 3303, describing a turbulent flow regime. Reaction Kinetics Model. Oxalic acid was selected as the model contaminant because it oxidizes to carbon dioxide in one step without forming stable intermediates. During adsorption and photodegradation of OA on TiO2 P25 films,25,26,34,35 the oxalate species chemisorb on TiO2 forming two complexes. These are a more stable bidentate structure with two oxygen species binding on the TiO2 surface and with the sigma C−C bond parallel to the TiO2 surface (species A) and a more labile species binding on one oxygen with the C−C bond perpendicular to the TiO2 surface (species B). Under irradiation at initial concentrations as in this study (10 to 40 ppm) and below, the reaction follows an apparent first-order kinetics with the oxidation of species A driven by hole transfer (slow) and the oxidation of species B driven by hydroxyl radical attack (faster). These two parallel kinetic mechanisms (Table S2, Supporting Information)24 has been modeled with the following assumptions. (i) The rate controlling steps are hole trapping by adsorbed OA (slow step) and OH• radical attack (faster step). (ii) Species A and B are at equilibrium under dark

(6)

Some authors entitled the latter as a volumetric rate of photon absorption, VRPA.32 However, VRPA can be found in the literature also as an integrated value of LVRPA across the reaction volume, meaning the sum of absorbed photons per unit of time in the entire reaction volume. To avoid confusion, the average amount of photons absorbed per unit of time and unit of volume was denoted as LumpVRPA hereafter. The radial velocity profile in PR1 was modeled with 13705

dx.doi.org/10.1021/es403472e | Environ. Sci. Technol. 2013, 47, 13702−13711

Environmental Science & Technology

Article

The concentration of OA at the inlet of the well-mixed reservoir tank is a function of time (a permanent nonstationary state), which means that a time-dependent homogeneous concentration of OA exits the reservoir tank and re-enter the reactor. PR1, which was operated under the laminar flow regime, was modeled considering streamline flow with negligible mixing in the lateral direction. The radial diffusive flux of OA concentration was neglected because the reaction kinetics were never too fast to generate sharp radial concentration gradients during each pass. However, the rate of OA oxidation varies along the reactor length (z-direction) and radius (r-direction), respectively. The mass balance is represented by eq 14. The average concentration of OA at the reactor outlet equals

and under irradiation conditions and follow a Langmuirian adsorption equilibrium model. (iii) Species A from solution occupies the sites in which species B has reacted maintaining an approximately constant surface concentration of oxalate onto the catalyst. (iv) Oxygen adsorbs on sites other than those occupied by the oxalate ions. (v) The unstable reaction intermediates such as hydroxyl and other radicals, electrons, and semiconductor holes are at steady-state. The rate-limiting step for oxalate oxidation [Ox] becomes eq 924 rOx = k OHθ(t )[Ox]B [•OH]B + k hole[1 − θ(t )][Ox]A [h+]A (9)

where kOH and khole are the rate constants for the hydroxyl radical and hole transfer attack on species B and A, respectively, and θ(t) is the fraction of sites occupied by B. The rate of electron−hole pair generation is the product of the primary spectral quantum yield Φ and the LVRPA rg( x) = Φ × LVRPA

R

[Ox]out Reactor =

(10)

Finally, applying the steady-state approximation on OH•, h , and e− species, a simple expression for the intrinsic local rate of oxalate oxidation can be derived, eq 1124 −rOx( x , t ) = {kBobsθ(t )[Ox]aq + kAobs[1 − θ(t )] [Ox] aq } LVRPA

(11)

with the fraction of sites occupied by B modeled as a decreasing linear function with the reaction time t proportional to LVRPA (eq 12), and x is the vector that indicates the position in the (z, r, φ) coordinate system. θ(t ) = β′[Ox]aq,0 [1 − α′t × LVRPA]

VTank

dt

out = Q ([Ox]in Tank − [Ox]Tank )

d[Ox] = rOx(z ,r ,t ) dz

(15)

in [Ox]out Reactor = [Ox]Tank

(16)

out [Ox]in Reactor = [Ox]Tank

(17)

rOx = d[Ox]aq /dt

(18)

with the LVRPA treated as a lump parameter (eq 6).



(13)

RESULTS AND DISCUSSION TiO2 Photon Extinction Coefficient. The photon extinction coefficient (σ + κ) of TiO2 P25 in the aqueous suspension was estimated by measuring (radiometrically in PR1) the photon flux emerging from the outer wall (eq 19) at different photocatalyst concentrations.31 This photon flux estimated from the SFM

for the perfectly mixed reservoir tank and vz(r)

Q

Prior to the illumination, the solution was recirculated in the dark. Therefore at (t = 0), the concentration of OA in the reservoir tank [Ox]Tank and reactor [Ox] were equal to the concentration of OA introduced into the system, [Ox]aq,0. The numerical simulation was performed dividing the (z, r) domain in sufficiently small intervals in each direction. The local velocity and the LVRPA for each point in the reaction space were initially computed solving eqs 3, 4, 5, and 7. A small time increment (Δt) equal to the reactor space time (τ = VReactor/Q) was then introduced. The mass balance in the tank (eq 13) was solved for a time period between t − Δt and t considering eq 16. Because the reservoir tank was perfectly mixed, eq 17 was applied, and the material balance in the reactor was solved at time t. The time step counter was increased, and the procedure repeated. In contrast, PR2 was modeled by the mass balance for a wellmixed constant volume and constant temperature batch reactor given by

(12)

Equation 12 derives from the consideration that the rate of reaction of species B and the fractional catalyst coverage by B is a function of time and of the photonic energy absorbed (LVRPA). As a first approximation, such dependence is taken to follow a linear dependence. The catalytic sites released of species B are then reoccupied by species A. The kinetic rate law combines an apparent first-order dependence on oxalate concentration [Ox]aq in solution for the adsorbed fraction reacting with OH radicals and a square root dependence on [Ox]aq for the adsorbed fraction of oxalate reacting by hole obs transfer. The intrinsic reaction rate constants kobs A and kB and the parameters α′, β′, were determined from matching the experimental and modeling results. Reactors Material Balances. PR1 was operated in total recirculation flow-through mode with the reactor outlet flow entering a well-mixed reservoir tank that fed back to the reactor. The coupled differential equations for the material balance on OA are d[Ox]out Tank

ηR

where the radial distribution of the OA concentration at the reactor outlet section, [Ox]out x , is obtained as an integrated value by solving the eqs 5, 11, 12, and 14 along the (z, r) domain inside PR1, with the boundary condition: [Ox] = [Ox]inTank at z = 0 at all t. Since the system was operated in a total recirculation, it follows:

+

×

2π ∫ rvz[Ox]out r dr

(14)

for the laminar flow annular photoreactor. rOx(z, r, t) is the local reaction rate calculated from eq 11 independent of φ for symmetry reasons. Q is the volumetric recirculation flow rate. 13706

dx.doi.org/10.1021/es403472e | Environ. Sci. Technol. 2013, 47, 13702−13711

Environmental Science & Technology I(r * ,z*),forward

⎛ 2 η I(ηR , z*) ⎜ 2 1 − ωcorr = r * 1 − γ ⎜⎝ 1 + 1 − ω 2 corr

Article

⎞ ⎟exp( −τ ) app ⎟ ⎠

suspension containing 0.4 g/L of TiO2. In PR1 the absorption of radiation appears to approach an optimum because the entire reactor volume can be irradiated with photons with minimal photon losses at the reactor boundary. In contrast, the volume of PR2 was poorly irradiated at this catalyst concentration, except for the region close to the lamp (less than 1 cm). The poor irradiation occurs as a result of the large value of the optical thickness (τ = 16.2, Table 2). The spatial distribution of the LVRPA in PR2 can be improved by decreasing the photocatalyst concentration. The general conclusion of this analysis is that catalyst dosage must be tailored to the specific reactor geometry and reactor dimensions. There is no single catalyst concentration that can produce an optimum rate of photon absorption in photocatalytic reactors that differ in size or geometry. The methodology to determine the optimum catalyst concentration, which maximizes the rate of photon absorption, has been described in simple terms in recent studies.18 This analysis requires the knowledge of the optical thickness and the apparent optical thickness that depend on the reactor geometry (optical path length), photocatalyst (type, particle size in suspension, bad-gap, and optical properties), and irradiation source (emission spectrum). The prospect of catalyst particle size and optical properties varying during the photocatalytic reaction process, for example, as a result of pH changes, may add another level of complexity. Photocatalytic Oxidation of Oxalic Acid in Multiple Photoreactors and Model Validation. The photocatalytic oxidation of OA over irradiated aqueous suspensions of TiO2 P25 was studied in PR1 and PR2, which differ in geometry, irradiation conditions, and hydrodynamics. Figure S5 of the Supporting Information shows the results obtained in PR1 for a set of experimental conditions and the model fitting when the LVRPA in the rate (eq 11) was replaced with a lumped volume averaged LVRPA value (eq 6) and when the LVRPA was treated as a local value and the reaction rate determined locally (eq 5 directly replaced in eq 11). In the latter case, the solutions with either the isotropic or the diffusely reflecting phase functions of scattering were compared. Lumping the LVRPA in one volume averaged value results in an unsatisfactory fitting of the experimental results in PR1. In contrast, the SFM with either the isotropic scattering or the diffusely reflecting phase functions could model the experimental results of the photocatalytic oxidation of OA. The inherent differences in the representation of the scattering phenomena by these two different phase functions appear to have a minor effect, although isotropic scattering yielded slight lower rates of oxidation. The scattering phase function might have a more significant effect on the LVRPA and the predicted contaminant degradation;36 however, such differences were attenuated in the present results because of the half-order dependence of the rate from LVRPA and by the dependence of the oxidation rate from other terms. In further model simulations, we exclusively adopted the parameters for the diffusively reflecting phase function (Table 2). That lumping the rate could produce unsatisfactory results in flow-through photocatalytic reactors had to be expected. In fact, in the absence of strong radial mixing, all tubular or annular flow-through photoreactors operated under laminar or turbulent flow, including compound parabolic collectors (CPC), have to be analyzed by estimating local contaminant reaction rates and by integrating these over the entire reactor volume. The analysis of such reactors must be carried out through an accurate modeling of the radiation field (e.g., by the

(19)

with r* ≥ 1, was correlated to the experimental data, with the extinction coefficient (embedded in τapp) as a fitting variable. Figure S4a of the Supporting Information shows the results at z* = 0.5 for two different photon fluxes when the catalyst concentration was varied from 0 to 0.4 g/L. Since the scattering albedo of TiO2 P25 is ω = 0.8617,32 the best fit of the experimental data was obtained when the extinction coefficient was set to 1350 m2 kg−1. Figure S4b of the Supporting Information shows that with such extinction coefficient, eq 19 could also predict the photon flux emerging at different positions on the outer wall of PR1 (at 20, 14, and 3.5 cm from the upper section of the lamp) when the catalyst concentration was set at 0.4 g/L and at different photon irradiances. Photon Radiation Absorption. In a photocatalytic reaction system of suspended TiO2 particles, the volumetric distribution of photon absorption in the reaction space is required to determine local reaction rates and summing these up, the overall conversion of the contaminant. Figure 1 shows the spatial distribution of the LVRPA in PR1 and PR2 calculated by the SFM (eq 5) incorporating the data in Table 2 and Table S1 of the Supporting Information and for a

Figure 1. Spatial distribution of LVRPA in (a) PR1 and (b) PR2. [TiO2] = 0.4 g L−1. 13707

dx.doi.org/10.1021/es403472e | Environ. Sci. Technol. 2013, 47, 13702−13711

Environmental Science & Technology

Article

SFM or other radiation models) and by the evaluation of the local rate of reaction.11 Lumping rates or photon absorption in one volume-averaged parameter can produce significant errors, the magnitude of which depends on the particular reactor setup and experimental conditions. The results in Figure S5 of the Supporting Information show a 12% error. Figure 2 shows the modeling of the experimental results in PR1 when the concentration of OA (10, 20, and 40 mg/L), catalyst concentration (0.1, 0.2, 0.3, g/L; 0.4 g/L shown in Figure S5 of the Supporting Information), and the irradiance at the lamp wall (50.55, 65.53, 75.41 W/m2; 89.84 W/m2 shown in Figure S5 of the Supporting Information) were varied. The model included the estimation of local reaction rates from eqs obs 5, 7, 11−17. The kinetic parameters (α′, β′, kobs A , and kB ) of OA photocatalytic oxidation on TiO2 P25 suspensions shown in Table 3 were determined by a nonlinear least-squares optimization procedure that resulted in the best fit of the experimental data. The validation of photocatalytic degradation models in different reactors operated under significantly different experimental conditions is infrequently reported in literature. Figure 3 shows the modeling of the experimental results in PR2 when the concentration of OA, the catalyst concentration, and in consequence the levels of radiation absorption were varied in a cross-random fashion within the confidence limits of the model. In these simulations, we adopted eqs 6, 11, 12, 18 and the lumped form of the LVRPA as previously discussed. The model with the kinetic parameters determined in PR1 (Table 3) could also fit the experimental data of PR2. However, higher dosages of TiO2 in PR2 resulted in lower LumpVRPA due to poor irradiation of the reactor volume because only the region nearer to the lamp was irradiated (Figure 1). Increasing catalyst concentration beyond an optimal reduces the effective irradiated volume reducing the rate of OA oxidation in the reactor. The suspensions in PR1 were saturated with oxygen by bubbling in the recirculation tank, while oxygen was supplied by the headspace in PR2; however, dissolved oxygen was never limiting the oxidation reaction because the model could predict the degradation of OA in both reactors. The validation of the proposed model in two different photoreactors, PR1 and PR2, supports the model assumptions including the proposed mechanism of OA oxidation, the proposed rate law, radiation field represented by the SFM, and assumptions on fluid flow. The kinetic parameters reported in Table 3 can therefore be considered to be independent from reactor geometry, reactor size, operation mode, and irradiation conditions. Model Advantages, Limitations, and Suggested Modifications. The photocatalytic reactor model presented delivers a balance between the simplicity of the modeling equations and the computational effort and accuracy of the model results. The model presents analytical solutions of the incident photon flux (eq 4) and the spatial distribution of the LVRPA (eq 5) as a result the radiation field in the reactor can be easily computed. Solving the full radiative transfer equation (RTE)23 over each reactor configuration may provide a more accurate representation of the radiation field in the model, but such an approach may require specialist modeling knowledge and more significant computation efforts. Another important aspect that may affect the rate of photocatalytic degradation of water contaminant is the composition of the water matrix,37 which should be evaluated if treating real contaminated water.

Figure 2. Oxalic acid photocatalytic oxidation kinetics in PR1 (pH 4) at (a) different initial concentrations of oxalic acid with [TiO2] = 0.4 g L−1 and I0 = 89.84 W m−2, (b) different TiO2 loadings and I0 = 89.84 W m−2, and (c) different irradiances with [TiO2] = 0.4 g L−1.

In such cases, the reactor model presented here should be modified to include (1) the influence of the water matrix species on the adsorption kinetics of the target contaminants onto the catalyst, (2) the potential scavenging effects of the water matrix components for light absorption and for the radical reactive species, and (3) the effect of the water matrix 13708

dx.doi.org/10.1021/es403472e | Environ. Sci. Technol. 2013, 47, 13702−13711

Environmental Science & Technology

Article

Table 3. Intrinsic Kinetic Parameters of the Photocatalytic Oxidation of Oxalic Acid

a

kinetic constant

PR1

PR2

m LumpVRPA, W m−3 kA, kmol1/2 s−1 W−1/2 kB, m3/2 s−1 W−1/2 α′ m3 s−1 W1− β′ m3 kmol−1

0.5 LVRPA (r,z)a

0.5 25.43 ± 1.39b

(2.21 ± 0.25) × 10−6 0.56 ± 0.06 45 (1.3 ± 0.3) × 10−4

In PR1, LVRPA was integrated along the space coordinates with respect to the reaction time. bGiven value is calculated for [TiO2] = 0.4 g L−1.

which are independent of reactor type, catalyst concentration, irradiance levels, and hydrodynamic conditions. These aspects are fundamental for further development of the general field of photocatalysis. In this study, we focused on the photocatalytic degradation of oxalic acid; however, this modeling approach can be easily extended to the analysis of other water contaminants. The model requires a suitable replacement of the contaminant rate law, which ideally should be derived from an elementary reaction kinetics mechanism, although the model can also accommodate semiempirical contaminants rate laws. A major gap in literature is the lack of scientific rigor in the evaluation of apparent rate constants of photocatalytic reactions. This is relevant for evaluating catalyst activity, quantum yields, comparing catalysts, or for predicting reaction conversion in multiple photoreactors. Contaminants reaction rates are often taken to follow apparent Langmuir−Hinshelwood or first-order rate laws without the development of a specific reaction mechanism or without the explicit inclusion of LVRPA in the rate. Since the absorption of photons is essential to promote photocatalysis, its effect cannot be lumped within an apparent kinetic constant, as often reported. Another important implication of the results presented in this study is that rate equations and reaction rate constants evaluated using lumped radiation absorption terms (e.g., evaluated by actinometry) results in an incorrect modeling of the degradation rate of contaminants because actinometry alone cannot distinguish between the volumetric variations of the LVRPA. Radiation absorption evaluated by actinometry is equivalent to estimating LumpVRPA and multiplying this by the reactor volume, which results in an underestimation of the OA degradation in PR1. The only circumstance in which it may be appropriate to use LumpVRPA is when the photoreactor is perfectly mixed, coupled to an expression of the contaminant rate law, which presents separable terms (i.e., the terms including the effect of radiation absorption can be separated from the concentration dependent terms). In such cases, the vigorous mixing makes the concentration of reactants uniform in the entire reactor volume, and the concentration dependent term can be taken out of the integral in the material balance leaving an integration term of the LVRPA over the reactor volume. The photocatalytic activity (or rate constants) of different materials cannot simply be evaluated at a fixed catalyst concentration because each catalyst, when dispersed in an aqueous suspension, will absorb and/or scatter photons differently. In consequence, the spatial distribution of the LVRPA would change for each material with consequent effects on the observed degradation rate of contaminants. An alternative method would be to evaluate catalysts activity at an equal rate of photon absorption by varying the concentration of each catalyst as a function of its optical properties, maintaining constant the reactor optical thickness.

Figure 3. Oxalic acid photocatalytic oxidation kinetics in PR2 (pH ≈ 4) at two different initial concentrations of oxalic acid at different TiO2 loadings and different levels of radiation absorption.

species on the optical behavior of the photocatalyst and therefore the spatial distribution of the LVRPA, which for example may be caused by particles agglomeration or deagglomeration. Each of these effects requires a comprehensive investigation. The first aspect can be handled in the present model by a modified contaminant rate law including the effects of ionic strength,38 pH, and competitive adsorption. The second aspect requires the modification of the reaction kinetics mechanism considering the relevant radicals scavenging reactions. Such modification would also result in a modified contaminant rate law. The effect of all types of light scavengers species should also be considered because this may influence the spatial distribution of LVRPA. Their net effect can be grossly quantified in the model by an increase in the absolute value of the optical thickness of the suspension; however, the effect of each light absorbing species should be explicitly included in the calculation of the LVRPA. Finally, the water matrix composition and pH may affect the optical properties of the catalyst and, in turn, the distribution of LVRPA in the reactors. In such cases, it is recommended that the optical properties of the photocatalyst (σ and κ) are determined in the actual water matrix suspension and under the prevalent conditions of mixing in the reactor. The absorption and scattering effects of the water matrix alone can also be measured in the absence of the catalyst and subtracted from the previous results. Wider Implications and Significance in Heterogeneous Photocatalysis. This study has shown a general and relatively simple methodology for determining “intrinsic” photocatalytic reaction kinetic constants of water contaminants, 13709

dx.doi.org/10.1021/es403472e | Environ. Sci. Technol. 2013, 47, 13702−13711

Environmental Science & Technology

Article

the photocatalytic oxidation of cyanide with TiO2 and silica-supported TiO2 suspensions. Appl. Catal., B 2008, 85, 48−60. (11) Colina-Márquez, J.; Machuca-Martínez, F.; Li Puma, G. Photocatalytic mineralization of commercial herbicides in a pilotscale solar CPC reactor: Photoreactor modeling and reaction kinetics constants independent of radiation field. Environ. Sci. Technol. 2009, 43, 8953−8960. (12) Richardson, S. D.; Plewa, M. J.; Wagner, E. D.; Schoeny, R.; DeMarini, D. M. Occurrence, genotoxicity, and carcinogenicity of regulated and emerging disinfection by-products in drinking water: A review and roadmap for research. Mutat. Res. 2007, 636, 178−242. (13) Krasner, S. W.; Weinberg, H. S.; Richardson, S. D.; Pastor, S. J.; Chinn, R.; Sclimenti, M. J.; Onstad, G. D.; Thruston, A. D., Jr. Occurrence of a new generation of disinfection byproducts. Environ. Sci. Technol. 2006, 40, 7175−7185. (14) Hernando, M. D.; Rodríguez, A.; Vaquero, J. J.; Fernández-Alba, A. R.; García, E. Environmental risk assessment of emerging pollutants in water: Approaches under horizontal and vertical EU legislation. Crit. Rev. Env. Sci. Technol. 2011, 41, 699−731. (15) Rizzo, L.; Manaia, C.; Merlin, C.; Schwartz, T.; Dagot, C.; Ploy, M. C.; Michael, I.; Fatta-Kassinos, D. Urban wastewater treatment plants as hotspots for antibiotic resistant bacteria and genes spread into the environment: A review. Sci. Total Environ. 2013, 447, 345− 360. (16) Prieto-Rodriguez, L.; Miralles-Cuevas, S.; Oller, I.; Agüera, A.; Puma, G. L.; Malato, S. Treatment of emerging contaminants in wastewater treatment plants (WWTP) effluents by solar photocatalysis using low TiO2 concentrations. J. Hazard. Mater. 2012, 211−212, 131−137. (17) Mueses, M. A.; Machuca-Martinez, F.; Li Puma, G. Effective quantum yield and reaction rate model for evaluation of photocatalytic degradation of water contaminants in heterogeneous pilot-scale solar photoreactors. Chem. Eng. J. 2013, 215−216, 937−947. (18) Colina-Márquez, J.; Machuca-Martínez, F.; Puma, G. L. Radiation absorption and optimization of solar photocatalytic reactors for environmental applications. Environ. Sci. Technol. 2010, 44, 5112− 5120. (19) Zhang, L.; Anderson, W. A.; Zhang, Z. J. Development and modeling of a rotating disc photocatalytic reactor for wastewater treatment. Chem. Eng. J. 2006, 121, 125−134. (20) Wang, D.; Li, Y.; Li, G.; Wang, C.; Zhang, W.; Wang, Q. Modeling of quantitative effects of water components on the photocatalytic degradation of 17α-ethynylestradiol in a modified flat plate serpentine reactor. J. Hazard. Mater. 2013, 254−255, 64−71. (21) Li Puma, G. Modeling of thin-film slurry photocatalytic reactors affected by radiation scattering. Environ. Sci. Technol. 2013, 37, 5783− 5791. (22) Li Puma, G.; Brucato, A. Dimensionless analysis of slurry photocatalytic reactors using two-flux and six-flux radiation absorptionscattering models. Catal. Today 2007, 122, 78−90. (23) Brandi, R. J.; Alfano, O. M.; Cassano, A. E. Evaluation of radiation absorption in slurry photocatalytic reactors. 1. Assessment of methods in use and new proposal. Environ. Sci. Technol. 2000, 34, 2623−2630. (24) Pozzo, R. L.; Brandi, R. J.; Cassano, A. E.; Baltanás, M. A. Photocatalytic oxidation of oxalic acid in dilute aqueous solution, in a fully illuminated fluidized bed reactor. Chem. Eng. Sci. 2010, 65, 1345− 1353. (25) Mendive, C. B.; Bredow, T.; Blesa, M. A.; Bahnemann, D. W. ATR-FTIR measurements and quantum chemical calculations concerning the adsorption and photoreaction of oxalic acid on TiO2. Phys. Chem. Chem. Phys. 2006, 8, 3232−3247. (26) Mendive, C. B.; Blesa, M. A.; Bahnemann, D. The adsorption and photodegradation of oxalic acid at the TiO2 surface. Water Sci. Technol. 2007, 55, 139−145. (27) Li Puma, G.; Khor, J. N.; Brucato, A. Modeling of an annular photocatalytic reactor for water purification: Oxidation of pesticides. Environ. Sci. Technol. 2004, 38, 3737−3745.

Dolat et al., have applied this method to evaluate the photoactivity of N-C-codoped TiO2 photocatalysts.39 Further efforts in these directions are needed to endorse heterogeneous photocatalysis as a powerful method for the treatment of contaminated water and wastewater.



ASSOCIATED CONTENT

S Supporting Information *

Details on photoreactor dimensions and lamp specifications, forward photon flux emerging from the outer wall of PR1 at different TiO2 powder loadings and levels of irradiance, reaction scheme for the heterogeneous photodecomposition of oxalic acid on TiO2, and further modeling results on PR1. This material is available free of charge via the Internet at http://pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Tel.: 0044(0)1509222510. Fax: 0044(0)1509223923. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS Professor Natalija Koprivanac and Dr. Natalia Quici are gratefully acknowledged for useful discussions. Dr. Grčić is thankful for the financial support from the British Scholarship Trust (visiting scholar grant).



REFERENCES

(1) Han, F.; Kambala, V. S. R.; Srinivasan, M.; Rajarathnam, D.; Naidu, R. Tailored titanium dioxide photocatalysts for the degradation of organic dyes in wastewater treatment: A review. Appl. Catal., A 2009, 359, 25−40. (2) Rehman, S.; Ullah, R.; Butt, A. M.; Gohar, N. D. Strategies of making TiO2 and ZnO visible light active. J. Hazard. Mater. 2009, 170, 560−569. (3) Shen, X. -Z.; Liu, Z. -C.; Xie, S. -M.; Guo, J. Degradation of nitrobenzene using titania photocatalyst co-doped with nitrogen and cerium under visible light illumination. J. Hazard. Mater. 2009, 162, 1193−1198. (4) Pelaez, M.; de la Cruz, A. A.; Stathatos, E.; Falaras, P.; Dionysiou, D. D. Visible light-activated N-F-codoped TiO2 nanoparticles for the photocatalytic degradation of microcystin-LR in water. Catal. Today 2009, 144, 19−25. (5) Li, Q.; Shang, J. K. Self-organized nitrogen and fluorine co-doped titanium oxide nanotube arrays with enhanced visible light photocatalytic performance. Environ. Sci. Technol. 2009, 43, 8923−8929. (6) Wang, D.; Li, Y.; Li Puma, G.; Wang, C.; Wang, P.; Zhang, W.; Wang, Q. Visible-light driven plasmon photocatalyst Ag/AgCl @ helical chiral TiO2 nanofibers. Chem. Commun. 2013, 49, 10367− 10369. (7) Salaices, M.; Serrano, B.; De Lasa, H. I. Experimental evaluation of photon absorption in an aqueous TiO2 slurry reactor. Chem. Eng. J. 2002, 90, 219−229. (8) Braham, R. J.; Harris, A. T. A complete multi-scale simulation of light absorption within a fluidized bed photoreactor using integrated particle, fluid and photon behaviour models. Phys. Chem. Chem. Phys. 2013, 15, 12373−12385. (9) Imoberdorf, G. E.; Taghipour, F.; Keshmiri, M.; Mohseni, M. Predictive radiation field modeling for fluidized bed photocatalytic reactors. Chem. Eng. Sci. 2008, 63, 4228−4238. (10) Marugán, J.; van Grieken, R.; Cassano, A. E.; Alfano, O. M. Intrinsic kinetic modeling with explicit radiation absorption effects of 13710

dx.doi.org/10.1021/es403472e | Environ. Sci. Technol. 2013, 47, 13702−13711

Environmental Science & Technology

Article

(28) Grčić, I.; Koprivanac, N.; Vujević, D.; Papić, S. Removal of atrazine from simulated groundwater by AOTs. J. Adv. Oxid. Technol. 2008, 11, 91−96. (29) Jacob, S. M.; Dranoff, J. S. Radial scale-up of perfectly mixed photochemical reactors. Chem. Eng. Prog. Symp. Ser. 1966, 62, 47−55. (30) Brucato, A.; Cassano, A. E.; Grisafi, F.; Montante, G.; Rizzuti, L.; Vella, G. Estimating radiant fields in flat heterogeneous photoreactors by the six-flux model. AIChE J. 2006, 52, 3882−3890. (31) Toepfer, B.; Gora, A.; Li Puma, G. Photocatalytic oxidation of multicomponent solutions of herbicides: Reaction kinetics analysis with explicit photon absorption effects. Appl. Catal., B 2006, 68, 171− 180. (32) Orozco, S. L.; Arancibia-Bulnes, C. A.; Suárez-Parra, R. Radiation absorption and degradation of an azo dye in a hybrid photocatalytic reactor. Chem. Eng. Sci. 2009, 64, 2173−2185. (33) Halász, G.; Gyüre, B.; Jánosi, I. M.; Szabó, K. G.; Tél, T. Vortex flow generated by a magnetic stirrer. Am. J. Phys. 2007, 75, 1092− 1098. (34) Araujo, P. Z.; Mendive, C. B.; Garcia Rodenas, L. A.; Morando, P. J.; Regazzoni, A. E.; Blesa, M. A.; Bahnemann, D. T-IR-ATR as a tool to probe photocatalytic interfaces. Colloids Surf. A 2005, 265, 73− 80. (35) Mendive, C. B.; Bahnemann, D. W.; Blesa, M. A. Microscopic characterization of the photocatalytic oxidation of oxalic acid adsorbed onto TiO2 by FTIR-ATR. Catal. Today 2005, 101, 237−244. (36) Brandi, R. J.; Citroni, M. A.; Alfano, O. M.; Cassano, A. E. Absolute quantum yields in photocatalytic slurry reactors. Chem. Eng. Sci. 2003, 58, 979−985. (37) Frontistis, Z.; Daskalaki, V. M.; Hapeshi, E.; Drosou, C.; FattaKassinos, D.; Xekoukoulotakis, N. P.; Mantzavinos, D. Photocatalytic (UV-A/TiO2) degradation of 17α- ethynylestradiol in environmental matrices: Experimental studies and artificial neural network modeling. J. Photochem. Photobiol., A 2012, 240, 33−41. (38) Dionysiou, D. D.; Suidan, M. T.; Bekou, E.; Baudin, I.; Laîné, J. -M. Effect of ionic strength and hydrogen peroxide on the photocatalytic degradation of 4-chlorobenzoic acid in water. Appl. Catal., B 2000, 26, 153−171. (39) Dolat, D.; Quici, N.; Kusiak-Nejman, E.; Morawski, A. W.; Li Puma, G. One-step, hydrothermal synthesis of nitrogen, carbon codoped titanium dioxide (N,C-TiO2) photocatalysts. Effect of alcohol degree and chain length as carbon dopant precursors on photocatalytic activity and catalyst deactivation. Appl. Catal., B 2012, 115−116, 81− 89.

13711

dx.doi.org/10.1021/es403472e | Environ. Sci. Technol. 2013, 47, 13702−13711