Catalysis Today 326 (2019) 82–93

Contents lists available at ScienceDirect

Catalysis Today
journal homepage: www.elsevier.com/locate/cattod

Photocatalytic NOx removal: Rigorous kinetic modelling and ISO standard T

reactor simulation
Vanesa Muñoza,b, Cintia Casadoa, Silvia Suárezb, Benigno Sánchezb, Javier Marugána,
⁎

a
Department of Chemical and Environmental Technology, ESCET, Universidad Rey Juan Carlos, C/ Tulipán s/n, 28933, Móstoles, Madrid, Spain
b
FOTOAIR-Ciemat, Renewable Energy Division, CIEMAT, Madrid, Spain

ARTICLE INFO ABSTRACT

Keywords: An air purification reactor working under the standard ISO 22197-1 is studied and used for the determination of
Photocatalysis the intrinsic kinetic parameters of the NOx photocatalytic removal. A new mechanistic kinetic model considering
NOx explicitly the radiation step is proposed. The derived reaction rate expressions for NO and NO2 include the
Mechanistic kinetics dependence on the concentration of NO, NO2 and the local superficial rate of photon absorption.
Intrinsic parameters
The air purification reactor was modelled using a numerical simulation methodology to validate homogeneity
CFD
of the light distribution reaching the catalyst surface and go deep in the knowledge of fluid dynamic behaviour.
Based on these results, general guidelines of the reactor behaviour working under the ISO 22197-1 standard can
be stablished, being the first time that this reactor is modelled with a global predictive approach. Radiation
intensity over the catalyst surface was confirmed to reach the specified value of 10 ± 0.5 W m−2. Plug flow can
be applied in the studied reactor with an axial dispersion module lower than 0.001. However, the diffusion of
chemical species in the reactor was also studied, as well as the impact of mass transfer phenomena on the
accuracy of the calculated kinetic parameters. A comprehensive simulation model of the ISO reactor, including
fluid dynamics, radiation, photochemical reaction and mass transfer was developed. The simulation results
confirm that the kinetic parameters obtained without considering mass transfer limitation fail to reproduce the
experimental data, whereas a global error lower than a 17% was achieved including the diffusion of the species
in the reactor. The proposed methodology, experimentally validated, could be successfully applied to the pre-
diction of the performance of different type of photocatalytic reactors or even open systems for NOx pollutants.

1. Introduction
The air-purification performance of semiconductor photocatalytic
materials can be achieved by photocatalytic test embraced in the ISO
22197 standards. Five air-purification photocatalytic methods for the
removal of different pollutant compounds have been published from
2007 to 2016 [1]: nitric oxide (NO), ISO 22197-1; acetaldehyde
(CH3CHO), ISO 22197-2; toluene (CH3C6H5), ISO 22197-3; for-
maldehyde (CH2O), ISO 22197-4; and methyl mercaptan (CH4S), ISO
22197-5. The last two standards for the removal of formaldehyde and
methyl mercaptan are still in the review phase (necessary at least three
years after publication).
All five standards cited above use the same flat plate photoreactor
with the catalyst being immobilized on a flat surface. This type of gas
phase photocatalytic reactor is one of the generic laboratory scale de-
vice designed for the determination of kinetic models.
Nitrogen oxides (NOx, NO and NO2) have been used in this study as
a target contaminant since they are typical inorganic and major air
pollutants produced by the combustion of fuels, for instance, from ve-
hicles in urban areas. This is why there are several proposed kinetic
models for the photocatalytic oxidation (PCO) of NO and NOx [2–6],
with the aim to reproduce the degradation of pollutant compounds in
air by photoactive materials and predict the performance of air-pur-
ification photoreactors. Some kinetic models are based on Langmuir–-
Hinshelwood and do not consider explicitly the local radiation ab-
sorption rate (LVRPA) [4], limiting the extrapolation to other systems.
Other kinetic models reported in the literature do consider radiation
absorption and reaction steps based on diverse reaction mechanisms
[2,3,5].
The most rigorous kinetic model for the PCO of NOx available in
literature was reported by Yu et al [5], based on mechanistic steps
previously described [4,5,7–9]. It involves the generation of holes and
electrons by the activation of the TiO2 particles and either their re-
combination or the respective hole and electron trapping. The OH•

⁎
Corresponding author.
E-mail address: javier.marugan@urjc.es (J. Marugán).

https://doi.org/10.1016/j.cattod.2018.09.001
Received 15 February 2018; Received in revised form 10 May 2018; Accepted 7 September 2018
Available online 24 September 2018
0920-5861/ © 2018 Elsevier B.V. All rights reserved.
V. Muñoz et al. Catalysis Today 326 (2019) 82–93

Nomenclature Greek letters

a Parameter of the polynomial equation defined for mass 1 Kinetic parameter (m2 s E−1)
transport 2 Kinetic parameter (dimensionless)
APC Photocatalyst absorption coefficient (dimensionless) 3 Kinetic parameter (kmol2 m−6)
a The active surface area per unit of air volume on the 4 Kinetic parameter (m4 kmol−1 s−1)
surface (m−1) ¯ Wavelength averaged primary quantum yield
b Parameter of the polynomial equation defined for mass (kmol·Einstein−1)
transport
c Parameter of the polynomial equation defined for mass Subscripts
transport
CFD Computational fluid dynamics 0 Relative to initial conditions
h Relative to transversal height A Relative to the model A
H Distance between catalytic area and optical window B Relative to the model B
(5 mm) I Specific position along the photoreactor
Ji Diffusive flux species i (kg m−2 s−1) in Relative to the inlet stream
k Kinetic constant (m2 kmol−1 s−1) NO Relative to nitrogen oxide
K Equilibrium adsorption constant (m3 kmol−1) NO2 Relative to nitrogen dioxide
LSRPA, ea,s Local superficial rate of photon absorption PC Photocatalyst
(Einstein m−2 s-1)
NOx Nitrogen oxides (nitrogen monoxide and dioxide) Superscripts
PCO Photocatalytic oxidation
q0 Incident radiation flux (Einstein m−2 s-1) s Relative to the photocatalyst surface area
Q Volumetric flow rate (m3 h−1)
R.H. Relative humidity (%) Special symbols
RTE Radiative transfer equation
r Photooxidation reaction rate (kmol m−2 s-1) [ TiO2] Superficial concentration of sites in conduction band
rg The superficial rate of electron-hole generation (kmol (kmol m−2)
m−2 s-1) [TiO2 _ ] Concentration on the catalyst surface (kmol m−2)
¯ −1 [] Concentration (kmol m−3)
v Velocity vector (m s )
[¯] Averaged concentration (kmol m−3)
Yi Mass fraction of species i in the mixture (dimensionless)

radicals produced at the surface of the photocatalyst (TiO2 _OH•) pro-
gressively oxides the adsorbed species -NO, -HNO2 and -NO2 to finally
be led to nitrates. The kinetic model was successfully applied in the
operation conditions of the reported study. However, as it will be
shown in the present work, the extrapolation of this model into some
limiting cases led to meaningless predictions. For these reason, a new
generalized kinetic model is proposed in this study.
Some efforts have been done previously by different research groups
in the modelling of photoreactors for indoor air purification. Boyjoo
et al. [10] summarized in a recent review the reactors simulated with
the traditionally called Computational Fluid Dynamics (CFD) methods.
That includes photoreactors with a wide range of geometries: corru-
gated plate [11], slit-shaped flat bed [12], annular [13], or monolith
system [14] among others. However, to the best of the author’s
knowledge, Pelzers et al. [15] reported the only approach found in
literature to model the radiation field of a photo-reactor following the
ISO 22197-1:2007 standard. They use the RADIANCE lighting simula-
tion tool, based on the backward ray-tracing modelling method, in a
mirrored system to calculate irradiation fluxes. However, no global
modelling approach including radiation field, fluid dynamics and a
mechanistic kinetic model description of the PCO reaction has been
reported yet.
In the present work, modelling of the reactor radiation field and
fluid dynamics were integrated with the developed intrinsic kinetic
model to predict the NOx removal reaction under a wide range of op-
erational conditions of irradiation, NO concentration and volumetric
flow rate.
Experimental validation of the developed kinetic model was carried
out using an air purification reactor in accordance with the ISO 22197-1
standard. Experiments under controlled conditions of the operational
variables were performed to obtain the intrinsic kinetic parameters that
defines PCO reaction rate with and without considering mass transport
limitations along the cross section of the reactor. In addition to that, a
numerical simulation model of the reactor was developed using the
finite volume method based ANSYS Fluent computational tool. Fluid
dynamics and radiation transport was modelled, simulated and vali-
dated, verifying the fulfillment of the ISO standard requirements. The
photochemical reaction and mass balances of NOx coupled with the
obtained velocity field, were also modelled. Results provided some
general guidelines of the simplifications that can be considered when
using this photoreactor under standardized conditions.
Predicted results were successfully validated with experimental
data, confirming that the proposed kinetic model, parameters and
methodology are a useful contribution to air purification apparatus
design, scaling up and NOx conversion prediction, allowing the future
design of more efficient apparatus.
2. Experimental
2.1. Photoreactor details
The photocatalytic experiments were carried out in a continuous flat
plate photoreactor designed according the ISO 22197-1:2007 (Fig. 1A).
The complete experimental set-up is represented schematically in
Fig. 1B. The photoreactor consists of a stainless steel rectangular cavity
with internal dimensions of 200 mm × 50 mm × 20 mm (length ×
width × depth) which embeds a ceramic planar sample inside with a
size of 100 mm × 50 mm × 15 mm (l × w × d). The ceramic sample
surface is coated with a commercial TiO2 synthesized through a sol-gel
method. According to standard ISO the sample surface is 5 mm beneath
of a borosilicate glass cover plate with high transmittance allowing the
irradiation of the photocatalytic materials. A completely developed
flow inside the reactor is assured by a reactor length double that the
sample length.

83
V. Muñoz et al.
Fig. 1. A) 3D Schematic view of the air purification photoreactor. B) Scheme of the photocatalytic experimental set-up.
Two UVA Phillips TL-8 W/05 fluorescent lamps (UV-A emission
17.5 W m−2 measured experimentally) were used as a radiation source
with a maximum emission at 365 nm wavelength. To avoid the end-
effects both lamps length 28.8 cm, being the extremes outside the re-
actor and placed in the longitudinal axis of the reactor situated on a
mobile support to set the irradiance on the photocatalyst surface at
10 W m-2 according to the ISO (distance needed 3.6 cm).
As it can be seen from Fig. 1B, the gas mixture is composed of pure
air (humidified to achieve 50% RH) and the NO pollutant stabilized in
N2 from a gas cylinder (104 ppm). The total gas flow is fed to the re-
actor, being fixed by mass-flow controllers to reach the desired NO
concentration in the inlet. Two three-way valves direct the flow either
to the photoreactor or to a by-pass mode. Then, only 700 cm3 min−1
flow pass through a NOX Analyzer (Signal NOx analyzer, model 447)
operating under the chemiluminescence detection principle according
to ISO standard 7996 and the rest is discarded through a duct. The
evolution of the NOx concentration with reaction time is measured and
the signal send to the datalogger (Combilog 1022).
2.2. Experimental reaction procedure
According to the ISO specification, the sample must be irradiated at
least 5 h, to ensure the removal of organic matter present on the pho-
tocatalytic sample. Thus, a preconditioning step of the test sample was
carried out to consider the photocatalyst surface exposed to the gas
stream as pristine. The set temperature of the whole system and relative
humidity (R.H.) during the experiments are 25 ± 2.5 ᵒC and 50%,
respectively. Flow rate, temperature, R.H. and pressure are controlled
by automated plant, using electronic mass flow controllers, specific
sensors and automated valves.
After cleaning the system with air at by-pass mode during 30 min,
the desired concentration of NO/N2 was incorporated to the gas feed.
Then, the air flow containing NO pass through the photoreactor con-
taining the photocatalytic sample. After stabilization of the possible
adsorption-desorption phenomena, the UV-A lamps were turned on and
84
Catalysis Today 326 (2019) 82–93
the photocatalytic oxidation process (PCO) starts. The distance between
the lamp centre and the catalyst was carefully adjusted to achieve the
10.0 ± 0.5 W m−2 irradiance defined in the ISO standard. The varia-
tion of the concentration of NO, NO2 and the total NOx was monitored
during the whole experiment. The results of the PCO reaction are
gathered once the steady state was achieved and it ended turning off the
lamps.
Three different experimental conditions have been selected to in-
vestigate their influence on the degradation of NOx and to study the
kinetic of the physicochemical process to obtain the intrinsic para-
meters:
(i) The inlet NO concentration (CNO, in) in the range 500 ppb to
2000 ppb fixing the mass flow of the NO/N2 and air stream to
reach the required concentration.
(ii) The incident radiation flux (q0) in the range 7.62 10−6 to 3.05 10-5
Einstein m-2 s-1 using different lamps and varying the distance to
the photocatalyst surface.
(iii) The inlet volumetric flow rate (Q0) in the range 1.5 dm3 min−1 to
4.0 dm3 min−1 varying the value set by the mass-flow controllers.
3. Kinetic modelling
3.1. Derivation of the kinetic model
As mentioned in the introduction section, Yu et al. [5] reported the
development of a mechanistic kinetic model for the PCO of NO, taking
into account the influence of several indoor air conditions such as
pollutant concentration, volumetric flow rate, relative humidity, irra-
diance, dosage of the photocatalyst, reactor size and the influence of the
nitrogen dioxide (NO2) on the PCO process. The reaction rates for NO
and NO2 proposed in this study were:
V. Muñoz et al. Catalysis Today 326 (2019) 82–93

s kmol kNO CNO CH 2O 4 E photocatalyst among adsorbed species, approximating the photo-
rNO = · 1+ 1 catalyst thickness to zero,
m2s1 (4kNO CNO + 2kNO2 CNO2 ) CH 2O (1)
(iii) the pseudo steady state approximation (MSSA) for the concentra-
kmol (kNO CNO kNO2 CNO2) CH 2O 4 E tion of electron, holes, radicals and intermediate compounds that
s
rNO 2 = · 1+ 1 does not accumulate,
m2s1 (4kNO CNO + 2kNO2 CNO2) CH 2O (2)
(iv) equilibrium conditions for adsorption steps between the con-
Where CNO , CNO2 and CH 2O are the concentration of NO, NO2, and H2O in centration of species at the surface and the bulk,
the bulk, respectively. E is the irradiance on the surface of photocatalyst (v) no competitive adsorption mechanisms for the species above the
and , , kNO and kNO2 are the intrinsic kinetic parameters. surface photocatalyst, and
This model was able to successfully reproduce the experimental data (vi) oxygen and water concentration in the bulk are assumed constant
analysed in their work. However, the model fails to predict the ex- during the reaction.
perimental behaviour of the system under the limiting of NOx con-
centration. Specifically: The details of the derivation are shown in the Appendix A and the
kinetic rate expressions for the NO and NO2 are as follows:
(i) when the concentration of NO2 in the bulk is zero, the predicted kmol
s 2
reaction rate of superficial NO would be independent of NO con- rNO
m2s1
= 4 [[NO]s + 2 [NO ]s [NO 2 ]s ]·

centration in the bulk, and only depends on the radiation absorp- 3

1+ 1+ ( 1+ 1 + 1 e a, s )
tion. Furthermore, the predicted reaction rate of superficial NO2 is [NO]2s + 22 [NO 2 ]2s + 2 2 [NO]s [NO2 ]s
not zero, in contrast to what it should be expected. This experi- (3)
mental condition is usually employed in the measurements under
kmol
the ISO 22197 standard, where the inlet to the reactor only contains s
rNO 2 = 2
4 [[NO ]s
2 2
2 [NO ]s [NO 2 ]s 2 2 [NO2 ]s ] ·
m2s1
NO and the concentration of NO2 is negligible.
3
(ii) When the concentration of NO in the bulk is zero, the predicted 1+ 1+
[NO ]2s + 22 [NO2 ]2s + 2 2 [NO]s [NO2 ]s
( 1+ 1 + 1 e a, s )

reaction rate of superficial NO2 would be independent of NO2

concentration in the bulk.
The reason of this dysfunction is that, in the considered reaction
mechanism, the OH• radicals can only react either with NO or NO2 .
Therefore, when the concentration of NO2 in the bulk is zero, the OH •
radicals only have the possibility to oxidise the NO molecules. As the
total amount of hydroxyl radicals can only follow this reaction
pathway, the reaction rate becomes independent of the concentration of
NO and only dependent on the radiation absorption. Something similar
occurs when the concentration of NO is zero.
To solve this problem, it is necessary to consider an additional re-
action pathway for OH • radicals to provide an alternative reaction route
in which they can be consumed. Table 1 shows the proposed extended
reaction mechanism. It includes the potential dimerization of hydroxyl
radical to form H2 O2 [16] that can be desorbed from the photocatalyst
surface and decomposes into O2 and H2O [17].
Based on the proposed reaction mechanism, the kinetic model can
be derived applying the following assumptions:
(i) there are no mass transfer limitations,
(ii) the photocatalytic reaction takes places at the surface of the
(4)
where s
and
rNO are the reaction rates for the removal of NO and the
s
rNO 2
net generation of NO2, respectively, [NO]s and [NO2]s are the NO and
NO2 concentration in the catalytic surface, ea,s is the local superficial
rate of photon absorption (LSRPA), and α1, α2, α3 and α4 the lumped
intrinsic kinetic parameters. This kinetic model solves the limitations of
the model developed by Yu et al. [5], being its prediction capability
analyzed in next sections. Anyway, it is important to remark that the
kinetic model has being derived under the specific conditions of con-
stant temperature and relative humidity specified in the ISO 22197-1
standard (25 ± 2.5 ᵒC and 50%, respectively).
3.2. Estimation of kinetic parameters
The estimation of the kinetic parameters was carried out by em-
ploying the method of least squares residues between the concentra-
tions predicted by the model and the experimental data obtained
varying different operational conditions following the reaction proce-
dure detailed in Section 2.2.
The local superficial rate of photon absorption (LSRPA, ea,s) is
considered as an uniform parameter, q0 (Eq. (5)), an assumption whose
validity will be proven in a later section. The photocatalyst absorption

Table 1
Proposed reaction mechanism for the photocatalytic oxidation of nitrogen oxides.
Step Reaction Rate

Activation TiO2 + h e + h+ rgs

Recombination e + h+ heat k2 [e ][h+]
Hole trapping h+ + TiO2 _H2 O TiO2 _OH • + H + k3 [h+][TiO2 _H2 O]
Electron trapping e + TiO2 _O2 TiO2 _O2 k 4 [e ][TiO2 _O2]
Hydroxyl attack TiO2 _NO + TiO2 _OH • TiO2 _HNO2 k5 [TiO2 _NO][TiO2 _OH •]
TiO2 _HNO2 + TiO2 _OH • TiO2 _NO2 + H2 O k 6 [TiO2 _HNO2][TiO2 _OH •]
TiO2 _NO2 + TiO2 _OH • TiO2 _NO3 + H + k7 [TiO2 _NO2][TiO2 _OH •]
Dimerization of hydroxyl radical TiO2 _OH • + TiO2 _OH • TiO2 _H2 O2 k8 [TiO2 _OH •]2
Decomposition of hydrogen peroxide H2 O2 H2 O +
1
O2 k 9 [H2 O2]
2
Adsorption TiO2 + H2 O TiO2 _H2 O k H2 O [ TiO2 ][H2 O] k H2 O [TiO2 _H2 O]
TiO2 + O2 TiO2 _O2 kO2 [ TiO2 ][O2] kO2 [TiO2 _O2]
TiO2 + NO TiO2 _NO kNO [ TiO2 ][NO] kNO [TiO2 _NO]
TiO2 + NO2 TiO2 _NO2 k NO2 [ TiO2 ][NO2] k NO2 [TiO2 _NO2]
TiO2 + H2 O2 TiO2 _H2 O2 k H2 O2 [ TiO2 ][H2 O2] k H2 O2 [TiO2 _H2 O2]

85
V. Muñoz et al. Catalysis Today 326 (2019) 82–93

coefficient, APC , can be considered constant for the selected material, ¯ i] + (x i + 1 x i)·a · rNO
[NO¯ i + 1] = [NO S
,i
because all the experiments were carried out with the optimum catalyst vair (9)
thickness above which absorption remains constant.
¯ 2, i] + (x i + 1 x i)· a · rNO
[NO¯2, i + 1] = [NO S
e a, s qo · APC (5) vair 2, i
(10)
The value of APC for the studied materials was experimentally de- where rNO
S
,i and rNO
were calculated with the kinetic model expressions
S
2, i
termined to be 0.071. (Eqs. (3) and (4)).
The concentration predicted at the outlet of the reactor was calcu- These two equations (Eqs. (9)–(10)) are used in an iterative opti-
lated through the mass balance of the NOx species in the reactor fol- mization process to calculate the values of the kinetic parameters that
lowing two different mass transfer scenarios: minimize the sum of the square of the errors with respect to the ex-
perimental data.
3.2.1. Case A: the reactor is operating without mass transfer limitations
Considering that the cross section of the reactor is very thin, several 3.2.2. Case B: the reactor is operating under mass transfer limitations
authors have assumed that the reaction is not controlled by the inter- In this case, the concentration at the outlet of the reactor was cal-
facial mass transport, and a one dimensional convection-controlled culated considering plug flow in the flow direction but assuming dif-
model can be assumed [2,4,18]. fusion in the cross section of the reactor to account for mass con-
Since the reaction only occurs on the surface of the photocatalyst, centration profiles between the bulk and the catalytic surface. This
only the air volume located over the plate was used in the kinetic cal- approach allows dealing with the mass transfer problems identified by
culations, as shown in Fig. 2A. some authors during the standard operation of this reactor [18,19].
Assuming a plug flow (validation of this hypothesis will be analysed In this case, the concentration of NOx along the cross section, is
later) the mass balance was calculated as follows [5]: calculated assuming a quadratic expression (see Supplementary
Information S1). In this equation, h is the distance between the catalytic
vair
d [NOx ] S
= a rNOx surface and cross section position, that changes from 0 and the max-
dx (6) imum value of hmax = H = 5 mm (distance among the catalytic surface
where, a is a coefficient with the purpose of change from superficial to and the optical window):
volumetric units by Eq. (7), and [NOx ¯ ], is the averaged NOx con- [NOx (h)] = a h2 + b h + c (11)
centration in the cross section. The linear air velocity is defined as υair,
being calculated by Eq. (8). In the Supplementary Information S1 is also included a schematic
representation of the boundary conditions used for the mass balance
Active surface area calculation including mass diffusion. At the catalyst surface (h = 0), it
a (m 1) =
air volume (7) is assumed that the result of surface reaction must be equal to diffusive
flux, in the transversal direction:
volumetric flow
vair (m s 1) =
cross section area (8) S
rNOx
d [NOx (h)]
, i = JNOx , i = DNOx
dh (12)
Assuming that there are no mass transport limitations, the con-
centration in the catalytic surface can be assumed to be equal to the where JNOx , i is the diffusive flux given by first Fick law of diffusion, and
concentration in the bulk. Therefore, the theoretical concentration of DNOx is the mass diffusivity.
NO and NO2 at each point along the reactor can be calculated as fol- Combining Eq. (12) and derivative of Eq. (11) at h = 0, parameter b
S
rNOx
lows: is obtained: = DNOx
.

Fig. 2. A) Air domain used in kinetic calculations. B) Geometry used in the radiation computational model. C) Geometry used for fluid-dynamics calculations and
convective-diffusive transport.

86
V. Muñoz et al. Catalysis Today 326 (2019) 82–93

Additionally, at h = 0, at the reactive surface, parameter 3.3. Computational model

c = [NOx ]s
At h = 5 mm (reactor window), it is assumed zero gradient 3.3.1. Radiation modelling
boundary condition:
d [NOx (h)]
=0 Simulation of the reactor behaviour based on the computation of the
dh
Consequently, parameter a can be calculated as: reaction rate according to the developed kinetic expressions required
the calculation of the local superficial rate of photon absorption
d [NOx (h)] rS S
rNOx (LSRPA, ea,s). The modelling of the radiation transport in the reactor
= 0 = 2· a·H + NOx a=
dh DNOx 2·DNOx ·H will also allow confirmation of the homogeneity in the irradiation
conditions required by the ISO 22197-1 standard.
By substituting in Eq. (11): The modelling of photocatalytic reactors usually requires the re-
solution of the radiative transfer equation (RTE) for every spatial di-
rection, which includes the energy loss due to absorption and out-
S
rNOx rS
[NOx (h)] = h2 + NOx h + [NOx ]s
2·DNOx · H DNOx (13) scattering and energy gain due to in-scattering from other directions.
The solution of this equation allows the evaluation of the radiation field
And, integrating through h, the volumetric average concentration is at any point over the reactor and catalyst surface. In the case of the
obtained: studied reactor, as the catalyst is immobilized, the calculation of the
radiant flux is relatively easy to accomplish, as the gas filling the re-
H
[NOx (h)]·dh S
rNOx ·H actor can be considered as non-absorbing media, concentrating radia-
¯ ]=
[NOx 0
= + [NOx ]s
H
dh 3· DNOx tion absorption on the thin catalyst layer.
(14)
0
The computational details are described below. For a better un-
derstanding, Fig. 2B represents the geometrical models used identifying
The solution of the system of equations given by:
the regions of interest.
In this work, the RTE was solved using the Discrete Ordinate
(i) surface concentration as function of averaged concentration (Eq.
Method with an angular discretization of 15 × 15, values previously
14)),
optimized to achieve accurate results at a reasonable computational
(ii) averaged concentration as function of surface reaction rate (Eqs.
costs [20,21]. The simulations were carried out using ANSYS Fluent v.
(9) and (10), and
14.5.
(iii) surface reaction rate as function of surface concentration (Eqs. (3)
The lamps (UVA Phillips TL-8 W/05) were modelled as diffuse
and (4)),
(isotropic) emitting surfaces, defining only the outer wall of the lamp.
The lamp emission values were obtained experimentally taking mea-
allows the estimation on the intrinsic kinetic parameters by em-
sures along its surface, with an averaged radiant flux value of 16.9 W
ploying an iterative optimization process that minimize the sum of the
m−2 and minor spatial changes. An air domain was defined between
square of the errors between model predictions and experimental data.
the lamps and the catalyst support to allow the calculation of the ra-
diation transport. The air was considered a non-participating media

Fig. 3. A) Incident radiation over the catalyst surface in the global scale and in detail. B) Velocity contour in the symmetry plane of the reactor and velocity vector
represented in the area over the catalyst frame.

87
V. Muñoz et al. Catalysis Today 326 (2019) 82–93

Fig. 4. A) Temporal NO mass fraction evolution over the catalyst surface. B) Normalized NO concentration leaving the catalyst area with time. C) E(t) function.

(absorption and dispersion coefficients equal to zero). An aluminium
reflector box cover was included in the experiments and in the com-
putational model. The aluminium surface has been considered to have
diffuse reflectance. The rest of the walls, except the emitting surface
representative of the UV-lamp, were considered transparent to radia-
tion. A value of 1 K was set in all regions and walls of the model,
avoiding the coupling of the radiation balance with the overall energy
balance of the system.
The model was discretized in approximately 460 000 structured
volume cells using ANSYS Meshing tool. This number of cells has been
found to be high enough to give mesh-independent results, corre-
sponding to a mean cell volume of 1.28 mm3.
3.3.3. Chemical reaction and mass transport modelling
To understand the importance of the kinetic model on the conver-
sion of the reaction, both set of parameters were used for the simulation
of the NOx photoreactor operation under the conditions of the ISO
22197-1 standard. Based on the previous fluid dynamics calculations,
the steady state simulation of the photochemical reaction and mass
transport was carried out introducing the developed kinetic expression
for the surface reaction rate (Eqs. (3) and (4)) in ANSYS Fluent through
an User Defined Function (UDF, external code). For each individual
chemical i in the computational domain, the mass conservation equa-
tion can be expressed as:
¯
( Yi ) + ( vYi ) = Ji + ri
t (15)

3.3.2. Fluid flow where Yi is the fraction of i in the mixture, Ji is its diffusive flux and ri its
¯
To confirm the validity of the use of a plug flow model for mass rate of production. The velocity vector, v , couples the mass balances to
balance calculation, the Residence Time Distribution curve (RTD) was the fluid dynamics calculations. Simulation of the NOx photochemical
determined with the computational model. First, the fluid dynamic degradation was carried out using a simplified system of reactions with
calculations were carried out by means of the resolution of the con- next global stoichiometry:
tinuity equation and the classical Navier-Stokes equation as described 1
elsewhere [21]. The simulation was performed considering a three-di- NO + O2 NO2
2
mensional, steady state, laminar flow (Reynolds number of 138 in the
core of the reactor). The fluid was assumed Newtonian, incompressible, NO2 +
1
O2 NO3
and isothermal, with constant physical properties (defined as air). The 2
inlet velocity was set to 1.77 m s−1 (3 L min−1, stablished in ISO Different initial NO concentrations were studied, as well as flow
22197-1:2007), and normal to the boundary. At the outlet, atmospheric rates and irradiance values. The diffusion coefficients used were
pressure was applied. A no-slip boundary condition was imposed at all DNO = 2.32 × 10−5 m2 s-1, DNO2 = 1.52 × 10−5 m2 s-1,
reactor walls. DNO3 = 1.37 × 10−5 m2 s-1 [22].
After that, to calculate the residence time distribution, the mass
transport of NO was simulated (diffusion coefficient 4. Results and discussion
DNO = 2.32 × 10−5 m2 s-1 [22] over the steady-state laminar velocity
field in the region over the catalytic surface (Fig. 2C). A non-steady 4.1. Light homogeneity over the catalyst surface
state simulation was performed for 2 s with a time step of 0.001 s. All
simulations were carried assuring a residual value lower than 10-6. The incident UV radiation over the catalyst surface is defined in the
The model used for fluid dynamic calculations was discretized in ISO 22197-1:2007, indicating that experiments must be carried out
around 470,000 structured volume cells using again ANSYS Meshing under an illumination of 10.0 ± 0.5 W m−2. Due to the nature of light
tool, corresponding to a mean cell volume of 0.33 mm3. distribution, is a difficult task to achieve a uniform light distribution
over the catalyst surface. Different catalyst to lamp heights were tested

88
V. Muñoz et al.
Fig. 5. NOx concentration in the outlet for the photocatalytic oxidation of NO at steady state conditions (ISO 22197-1) and predicted with the proposed kinetic
model. A) As a function of the initial NO concentration, B) Influence of the incident radiation, C) Varying the flow rate. D) Global adjustment between predicted and
experimental results.
to assure standardized conditions. Best results are obtained for the
configuration showed in Fig. 3A, reaching to a value of 10.1 ± 0.4 W
m−2 for a distance between lamps and catalyst surface of 36 mm.
Additionally, the uniformity index is obtained, with a value of 0.98
that represents how a specified field variable varies over a surface,
where a value of 1.00 indicates the highest uniformity [23]. Based on
the results of incident radiation on the catalytic surface, a constant
value of irradiance can be reasonably assumed over the entire surface of
the catalyst, whose value is introduced in the kinetic model expressions.
4.2. Analysis of the reactor flow
Fig. 3B shows the velocity magnitude represented over the air path
line for the air purification reactor assuming a laminar model
(Re = 138). It is clear from the velocity vectors, with a reduced velocity
scale, that the flow collides with the ceramic plate with the catalyst on
this surface leading to a backflow swirl. Similar behaviour is founded in
the area after the ceramic plate. Nevertheless, over the catalyst surface,
the velocity values and vectors are quite homogeneous.
To determine if a plug flow hypothesis can be reasonably used for
conversion prediction, the deviation of the real flow from the ideal
model is characterized in the region in contact with the catalytic
89
Catalysis Today 326 (2019) 82–93
surface. The NO convective-diffusive transport over the laminar flow is
simulated to calculate the residence time distribution. Fig. 4A shows the
evolution with time of the NO mass fraction in the area of interest. In
Fig. 4B is represented the normalized NO concentration leaving the
catalyst area against time while in the Fig. 4C, the residence time dis-
tribution curve (the E(t) function) is plotted.
The residence time (RT) in contact with the catalyst and variance
value are calculated as described in [24]. The calculate RT value was
0.55 s in contact with the catalyst. The variance value obtained from
the E(t) curve is 0.039 s2, a value low enough to assume plug flow
conditions. The axial dispersion module was additionally calculate as
D
, the non-dimensional group that characterizes dispersion in all the
uL
system [24]. The value obtained was 0.001, confirming that dispersion
is insignificant and plug flow can be accepted.
4.3. Kinetic model validation
Two different sets of kinetic parameters were calculated under the
two mass transfer considerations scenarios (see Section 3.2). The ob-
tained values of the kinetic parameters 1, 2 , 3, 4 with their respective
units are presented for both models:
V. Muñoz et al. Catalysis Today 326 (2019) 82–93

Fig. 6. NOx concentration in the outlet for the photocatalytic oxidation of NO at steady state conditions (ISO 22197-1) and predicted with CFD and the proposed
kinetic model. A) As a function of the initial NO concentration, B) Influence of the incident radiation, C) Varying the flow rate. D) Global adjustment between
predicted and experimental results.

agreement with the expected [9]. A raise in the irradiance leads into a
Case A (without mass transfer) Case B (with mass transfer) greater number of active sites and accelerate the PCO process in both
pollutant species. The oxidation rate of NO is higher than the NO2 rate,
1= 3.58 m2 s E 1 1 = 10.4 m2 s E 1 because the increase of the irradiance on the photocatalyst surface
2 = 2.29 × 10 1 (dimensionless ) 2 = 7.55 × 10 2 (dimensionless ) produces accumulation of NO2 in the bulk. Furthermore, Hüsken et al.
3 = 3.28 × 10 3 kmol2 m 6 3 = 3.61 × 10 3 kmol2 m 6 [25] overview several studies that have reported the presence of a first
4 = 69.97 m4 kmol 1 s 1 4 = 91.60 m4 kmol 1 s 1 region where the degradation has a linear dependency on irradiance
and a second one where the dependency is proportional to the square
root of irradiance, a common behaviour in photoactivated processes.
Validation of the model was carried out under different operation Different values for the change over both regimes have been set: i)
conditions such as inlet NO concentration, irradiance on the photo- 250 W m−2 and ii) 10-20 W m-2. Otherwise, the results obtained by
catalyst surface and inlet air volumetric flow rate. The prediction of Hüsken et al. are similar to ours, with a value for the change-zone of
NOx degradation by the kinetic model using both sets of kinetic para- approximately 4 W m-2. Again, the behavior is reproduced satisfactorily
meters is represented in Fig. 5. by the proposed kinetic model.
Fig. 5A shows the comparison between the NO and NO2 outlet The influence of the flow rate was also studied (Fig. 5C). A higher
concentration predicted by the model using both sets of kinetic para- volumetric air flow causes an increase in NO outlet concentration and a
meters and obtained experimentally. The linear relation between the decrease in NO2 outlet concentration, due to a lower residence time in
NO inlet concentration and the NO and NO2 outlet concentration is the photoreactor that reduces the possibility of the NO pollutant to be
clearly reflected in the figure and previously reported [3,25]. Thus, it adsorbed and oxidized on the surface. In this case, for the lowest values
can be deduced that the kinetic model behaves as a first order model. of volumetric air flow, the experimental results for the NO2 con-
Both sets of kinetic parameters reproduce satisfactory this behaviour, centration were better reproduced using the model considering mass
although NO2 concentrations slightly higher than the experimental transfer limitations. Finally, Fig. 5D shows the experimental NO and
values are predicted using the model considering mass transfer limita- NO2 outlet concentrations in all the experiments compared with those
tions. obtained by modelling. As can be derived from the proximity of the
The effect of irradiance is shown in Fig. 5B, showing a behaviour in

90
V. Muñoz et al.
Fig. 7. Molar concentrations obtained at different planes of the reactor calculated from the CFD simulations of the reactor operation under the conditions of the ISO
22197-1 standard.
obtained results with the graph diagonal, what represents the error of
the prediction, the agreement between the experiments and the pre-
dictions using both approaches is certainly good.
The value of Normalized Root Mean Square Error (NMRSE) has been
calculated to quantify the deviation of the model. NMRSE represents
the standard deviation of the differences between the predicted values
and the experimental values, normalized between the average experi-
mental values. The calculated NMRSE for the predicted results has re-
sulted in 13% without considering diffusive limitations, and a 13.6%
with parameters estimated considering diffusive transport.
4.4. Reactor simulation: kinetic model and mass transfer
The kinetic expressions (Eqs. (3) and (4)) together with the kinetic
parameters obtained, were used for the rigorous simulation of the
photoreactor operation using multiphysics CFD modelling. The CFD
model considers the diffusive and convective mass transport of the
chemical species taking into account the flow distribution in the re-
actor.
In the previous section it has been shown that a good fitting of the
91
Catalysis Today 326 (2019) 82–93
experimental results can be obtained with both versions of the plug-
flow model (with and without mass transfer). However, when the set of
kinetic parameters obtained without considering diffusion is used in the
CFD model, a huge prediction error is obtained (see Fig. 6, solid lines),
with a calculated NMRSE above 60%. In contrast, this error can be
significantly reduced to a 17% using the kinetic parameters estimated
under the consideration of diffusional control (Fig. 6, dashed lines).
This conclusion is in agreement with those reported by other research
groups [18,19], whose reports showed that mass transport limitations
could be only discarded when the conversion is below 10%.
The relatively small differences observed between the CFD model
based on the mass transport based kinetics and the experimental data of
the reaction operation can be attributed to non-ideal features of the
reactor flow that also affect mass transfer calculations. As an example,
Fig. 7 shows the concentration profiles in different reactor planes cal-
culated from the CFD simulations of the reactor operation under the
conditions of the ISO 22197-1 standard. As it can be observed, in ad-
dition to the existence of concentration profiles in the cross section,
significant variations in concentration can be observed also along the
width of the reactor, definitively rebutting the plug flow assumption.
V. Muñoz et al. Catalysis Today 326 (2019) 82–93

Based on the above discussion, it is clear that a careful attention
must be paid when using kinetic parameters obtained with simplified 1-
D or 2-D models reported in the literature [2,4,18]. Even in the studied
case, in which most of the assumptions were satisfactory validated
(plug-flow in the axial direction, homogeneous radiation), not con-
sidering mass transfer in the cross section makes useless model pre-
dictions of the reactor operation.
In contrast, the model here proposed demonstrated that considering
the diffusion in the cross section during the estimation of the kinetic
parameters leads to meaningful predictions of the performance of the
ISO 22197-1 photoreactors for NOx photocatalytic oxidation.
5. Conclusions
experimental data, being able to predict the influence of different op-
erating conditions such as pollutant concentration, irradiance and vo-
lumetric flow in the removal of NOx. However, whereas the kinetic
parameters estimated without mass transport limitations fails to predict
the performance of reactor, the use of the kinetic parameters calculated
considering diffusional control and the application of a rigorous CFD
model of the reactor coupling fluid dynamics, radiation, photoreaction
and mass transfer is able to predict satisfactorily the experimental re-
sults in the reactor with an error below 17%. This confirms the validity
of the proposed kinetic modelling approach, parameters and metho-
dology and represent a significant advance towards the commercial
application of this technology at large scale.

The photocatalytic oxidation process of nitrogen oxides (NOX) has Acknowledgements

been rigorously modelled according to the ISO standard photoreactor.
The theoretical assumptions considered in the modelling of the PCO
process have been made, were substantiated and validated by the
photoreactor CFD simulation.
The kinetic model has been successfully applied to reproduce the
The authors gratefully acknowledge the financial support of the
Spanish Ministry of Economy and Competitiveness (MINECO) through
the project WATER4FOOD (CTQ2014-54563-C3-1-R) and Comunidad
de Madrid through the program REMTAVARES (S2013/MAE-2716).

Appendix A. Derivation of the kinetic model

The kinetic model for the PCO of NOx is derived based on the reaction scheme summarized in Table 1. It is assumed that the adsorption of species
at the reaction surface take place in equilibrium conditions and through a no competitive mechanism, consequently, the adsorbed species con-
centration can be expressed as:
[TiO2 _NO] = KNO [ TiO2 ][NO] (A.1)

[TiO2 _NO2] = K NO2 [ TiO2 ][NO2] (A.2)

[TiO2 _H2 O] = K H2 O [ TiO2][H2 O] (A.3)

[TiO2 _O2] = KO2 [ TiO2 ][O2] (A.4)

[TiO2 _H2 O2] = K H2 O2 [ TiO2 ][H2 O2] (A.5)

Applying the pseudo steady state approximation (MSSA) for the concentration of electrons, holes, the intermediate compound TiO2 _HNO2 that is
consider not to accumulate and the hydroxyl radical, the following expressions are derived:
rgs
[e ] =
k2 [h+] + k4 [TiO2 _O2] (A.6)

rgs
[h+] =
k2 [e ] + k3 [TiO2 _H2 O] (A.7)

k5 [TiO2 _NO][TiO2 _OH •]

[TiO2 _HNO2] =
k6 [TiO2 _OH •] (A.8)

k5 [TiO2 _NO] + k6 [TiO2 _HNO2] + k 7 [TiO2 _NO2 ] k8·4·k3 [h+][TiO2 _H2 O]

[TiO2 _OH •] = · 1+ 1+
2k 8 ( k5 [TiO2 _NO] + k6 [TiO2 _HNO2] + k 7 [TiO2 _NO2 ]) 2 (A.9)
The superficial rate of electron-hole generation at the photocatalyst surface is defined by:
rgs = ¯ e a, s (A.10)
where is the local superficial rate of photon absorption (LSRPA) and ¯ wavelength averaged primary quantum yield.
e a, s
Introducing Eq. (B.6) and (B.10) in Eq. (B.7) the concentration of holes is obtained:
k 4 [TiO2 _O2]
[h+] = ( 1+ 1+ 1 e a, s )
2k2 (A.11)
where
4 ¯ k2
1, B =
k3 k 4 [TiO2 _H2 O][TiO2 _O2] (A.12)
From Eqs. (A.9) and (A.11), and introducing the concentration of the adsorbed species Eqs. (A.1)–(A.5), the concentration of the hydroxyl radical
can be expressed as:

[ TiO2 ] k5 KNO 3
[TiO2 _OH •] = [[NO] + 2 [NO2]]· 1+ 1+ ( 1+ 1+ 1 e a, s )
k8 [NO]2 + 2
2 [NO
2]
2 +2 2 [NO][NO2 ] (A.13)

92
V. Muñoz et al. Catalysis Today 326 (2019) 82–93

where
k 7 K NO2
2 =
2 k5 KNO (A.14)

k8 k3 K H2 O [H2 O] k 4 K O2 [O2 ]
3 =
2 k2 k52KNO 2 (A.15)
The superficial reaction rate for the removal of NO and the net generation of the intermediate compound NO2 are expressed as follows:
s
rNO = k5 [TiO2 _NO][TiO2 _OH •] (A.16)
s
rNO • •
2 = k 6 [TiO2 _HNO2 ][TiO2 _OH ] k 7 [TiO2 _NO2 ][TiO2 _OH ] (A.17)
Introducing Eq. (A.13) and the concentration of the adsorbed species Eq. (A.1) – (A.5) in Eq. (A.16) and (A.17), the final expressions for the
superficial reaction rate of NO and NO2 are:

s 3, B
rNO (kmol· m 2 ·s 1) = 4 [[NO]2 + 2 [NO2 ][NO]]· 1+ 1+ ( 1+ 1+ 1 e a, s )
[NO]2 + 2
2 [NO 2
2] + 2 2 [NO][NO2 ] (A.18)

s 2 1 2 2 [NO ]2 ] 3
rNO 2 (kmol ·m · s ) = 4 [[NO] 2 [NO][NO2] 2 2 2 · 1+ 1+ ( 1+ 1+ 1 e a, s )
[NO]2 + 2
2 [NO
2]
2 +2 2 [NO][NO2 ] (A.19)
where
[ TiO2 ]2 k52KNO 2
4 =
k8 (A.20)
A simplification of the dependence of reaction rates with the absorbed energy can be done by taking the first term of the square root Taylor
expansion:
( 1+ 1+ 1 e a, s ) (0.5· 1· e a, s )

Appendix B. Supplementary data

Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.cattod.2018.09.001.

References photocatalytic reactor, Chem. Eng. Res. Des. 90 (2012) 1297–1306.

[1] A. Mills, C. Hill, P.K.J. Robertson, Overview of the current ISO tests for photo-
catalytic materials, J. Photochem. Photobiol. A: Chem. 237 (2012) 7–23.
[2] R. Dillert, A. Engel, J. Große, P. Lindner, D.W. Bahnemann, Light intensity de-
pendence of the kinetics of the photocatalytic oxidation of nitrogen(ii) oxide at the
surface of TiO2, Phys. Chem. Chem. Phys. 15 (2013) 20876–20886.
[3] R. Dillert, J. Stötzner, A. Engel, D.W. Bahnemann, Influence of inlet concentration
and light intensity on the photocatalytic oxidation of nitrogen(II) oxide at the
surface of Aeroxide® TiO2 P25, J. Hazard. Mater. 211–212 (2012) 240–246.
[4] M.M. Ballari, M. Hunger, G. Hüsken, H.J.H. Brouwers, NOx photocatalytic de-
gradation employing concrete pavement containing titanium dioxide, Appl. Catal. B
Environ. 95 (2010) 245–254.
[5] Q.L. Yu, M.M. Ballari, H.J.H. Brouwers, Indoor air purification using heterogeneous
photocatalytic oxidation. Part II: kinetic study, Appl. Catal. B Environ. 99 (2010)
58–65.
[6] M. Hunger, G. Hüsken, H.J.H. Brouwers, Photocatalytic degradation of air pollu-
tants - from modeling to large scale application, Cem. Concr. Res. 40 (2010)
313–320.
[7] J.V.S. de Melo, G. Trichês, Evaluation of the influence of environmental conditions
on the efficiency of photocatalytic coatings in the degradation of nitrogen oxides
(NOx), Build. Environ. 49 (2012) 117–123.
[8] J. Ângelo, L. Andrade, L.M. Madeira, A. Mendes, An overview of photocatalysis
phenomena applied to NOx abatement, J. Environ. Manage. 129 (2013) 522–539.
[9] M.M. Ballari, Q.L. Yu, H.J.H. Brouwers, Experimental study of the NO and NO2
degradation by photocatalytically active concrete, Catal. Today 161 (2011)
175–180.
[10] Y. Boyjoo, H. Sun, J. Liu, V.K. Pareek, S. Wang, A review on photocatalysis for air
treatment: from catalyst development to reactor design, Chem. Eng. J. 310 (2017)
537–559.
[11] C. Passalía, O.M. Alfano, R.J. Brandi, A methodology for modeling photocatalytic
reactors for indoor pollution control using previously estimated kinetic parameters,
J. Hazard. Mater. 211–212 (2012) 357–365.
[12] S.W. Verbruggen, S. Lenaerts, S. Denys, Analytic versus CFD approach for kinetic
modeling of gas phase photocatalysis, Chem. Eng. J. 262 (2015) 1–8.
[13] F. Jović, V. Kosar, V. Tomašić, Z. Gomzi, Non-ideal flow in an annular
[14] X. Wang, X. Tan, T. Yu, Modeling of formaldehyde photocatalytic degradation in a
honeycomb monolith reactor using computational fluid dynamics, Ind. Eng. Chem.
Res. 53 (2014) 18402–18410.
[15] R.S. Pelzers, Q.L. Yu, R.A. Mangkuto, Radiation modeling of a photo-reactor using a
backward ray-tracing method: an insight into indoor photocatalytic oxidation,
Environ. Sci. Pollut. Res. 21 (2014) 11142–11154.
[16] H. Czili, A. Horváth, Applicability of coumarin for detecting and measuring hy-
droxyl radicals generated by photoexcitation of TiO2 nanoparticles, Appl. Catal. B
Environ. 81 (2008) 295–302.
[17] V. Augugliaro, M. Litter, L. Palmisano, J. Soria, The combination of heterogeneous
photocatalysis with chemical and physical operations: a tool for improving the
photoprocess performance, J. Photochem. Photobiol. C Photochem. Rev. 7 (2006)
127–144.
[18] C. Minero, A. Bedini, M. Minella, On the standardization of the photocatalytic gas/
solid tests, Int. J. Chem. React. Eng. 11 (2013) 717–732.
[19] A. Mills, L. Burns, C. O’Rourke, S. Elouali, Kinetics of the photocatalysed oxidation
of NO in the ISO 22197 reactor, J. Photochem. Photobiol. A: Chem. 321 (2016)
137–142.
[20] C. Casado, R. Timmers, A. Sergejevs, C.T. Clarke, D.W.E. Allsopp, C.R. Bowen,
R. van Grieken, J. Marugán, Design and validation of a LED-based high intensity
photocatalytic reactor for quantifying activity measurements, Chem. Eng. J. 327
(2017) 1043–1055.
[21] C. Casado, J. Marugán, R. Timmers, M. Muñoz, R. van Grieken, Comprehensive
multiphysics modeling of photocatalytic processes by computational fluid dynamics
based on intrinsic kinetic parameters determined in a differential photoreactor,
Chem. Eng. J. 310 (2017) 368–380.
[22] M.J. Tang, R.A. Cox, M. Kalberer, Compilation and evaluation of gas phase diffusion
coefficients of reactive trace gases in the atmosphere: volume 1. Inorganic com-
pounds, Atmos. Chem. Phys. 14 (2014) 9233–9247.
[23] Ansys Fluent Theory Guide, Version 14.5, Fluent Inc., (n.d.).
[24] O. Levenspiel, The Chemical Reactor Omnibook, OSU book stores Corvallis, Oregon,
1996.
[25] G. Hüsken, M. Hunger, H.J.H. Brouwers, Experimental study of photocatalytic
concrete products for air purification, Build. Environ. 44 (2009) 2463–2474.

93