You are on page 1of 13

Chemical Engineering Science 58 (2003) 2269 – 2281

www.elsevier.com/locate/ces

Modelling and design of thin-$lm slurry photocatalytic reactors


for water puri$cation
Gianluca Li Pumaa;∗ , Po Lock Yueb
a School of Chemical, Environmental and Mining Engineering, The University of Nottingham, University Park, Nottingham NG7 2RD, UK
b Department of Chemical Engineering, The Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong, People’s
Republic of China
Received 10 July 2002; received in revised form 11 December 2002; accepted 20 January 2003

Abstract

Photocatalytic oxidation processes are highly e4ective clean technologies for the degradation and mineralization of a wide variety of
priority pollutants in water and wastewater. However, the application of heterogeneous photocatalysis for wastewater treatment on an
industrial scale has been impeded by a lack of mathematical models that can be readily applied to reactor design and scale-up. As a results
current photocatalytic reactors in research and development have been designed by empirical or semi-empirical methods only.
In this paper, a simple and generic mathematical model for steady-state, continuous 9ow, thin-$lm, slurry (TFS) photocatalytic reactors
for water puri$cation using solar and UV lamps is presented. The model developed is applicable to TFS 9at plate and annular photoreactors
of (a) falling $lm design or (b) double-skin design, operating with three ideal 9ow conditions: (1) falling $lm laminar 9ow, (2) plug
9ow and (3) slit 9ow. The model is expressed in dimensionless form and scale-up of TFS photocatalytic reactors can be carried out by
dimensional analysis. In addition, the model parameters can be estimated easily from real systems and model solutions can be obtained
with little computational e4ort.
Comparison of a number of ideal 9ow systems shows that both falling $lm laminar 9ow and plug 9ow operation modes give higher
performance than the slit 9ow system. Slit 9ow operation mode results in lower conversions due to the non-correspondence of 9uid-residence
time and the transversal radiation $eld. The e4ect of optical thickness, on reactor performance and the evolution of radial pro$les of a
model pollutant with photoreactor length are presented for each of the operation modes. The falling $lm laminar 9ow system was found
to be more e?cient than the plug 9ow system when the reactor conversion is above 80%. For lower reactor conversion the plug 9ow
system was found to be marginally more e?cient than the falling $lm laminar 9ow system. A methodology for the optimal geometrical
design of a highly e?cient con$guration of TFS photocatalytic reactors is also presented. The mathematical models presented may be
used as a tool for the design, scale-up and optimization of these types of photocatalytic reactors.
? 2003 Elsevier Science Ltd. All rights reserved.

Keywords: Reaction engineering; Photoreactor; Falling $lm; Slurries; Suspension; Thin $lms

1. Introduction In heterogeneous photocatalysis, the interaction of pho-


tons of su?cient energy with a suitable semiconductor
The design and modelling of photocatalytic reactors is a e.g., titanium dioxide—produces electron–hole pairs which
relatively new area of research (Cassano, Martin, Brandi, can drive oxidative–reductive reactions (Bahnemann,
& Alfano, 1995) but is essential for the successful ex- Bockelmann, & Goslich, 1991). In the presence of water,
ploitation of heterogeneous photocatalysis as an alterna- the above process yields hydroxyl radicals that can decom-
tive method for the puri$cation of water and treatment of pose a wide variety of priority pollutants.
wastewater (Ho4mann, Martin, Chai, & Bahnemann, 1995; Thin-$lm, slurry (TFS) photocatalytic reactors provide an
Alfano, Bahnemann, Cassano, Dillert, & Goslich, 2000). excellent con$guration for e?cient excitation of the semi-
conductor photocatalyst (TiO2 ) due to the high light ab-
sorptivity of TiO2 slurry suspensions. This con$guration
∗ Corresponding author. Tel.: +44-0115-951-4170; bene$ts from a very large illuminated catalyst surface area
fax: +44-0115-951-4115. per unit volume of reactor and minimal mass transfer limi-
E-mail address: gianluca.li.puma@nottingham.ac.uk (G. Li Puma). tations (Turchi & Ollis, 1988). TFS photocatalytic reactors

0009-2509/03/$ - see front matter ? 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0009-2509(03)00086-1
2270 G. Li Puma, P. L Yue / Chemical Engineering Science 58 (2003) 2269 – 2281

normally operate at higher catalyst concentrations than con- catalyst. As the major operative costs of a hypothetical pho-
ventional photoreactors. This increases the number of avail- tocatalytic treatment plant derive from the electricity and
able sites for hydroxyl radical generation per unit volume of lamp replacement cost, (Li Puma & Yue, 1999), the selec-
reactor and maximizes the e?ciency of photon utilization, tion of the radiation source is a critical step for optimal re-
oxidation rates and reactor throughput. TFS photocatalytic actor design.
reactors are therefore suitable for large-scale industrial ap- UV radiation sources for TFS photocatalytic reactors are
plications of photocatalysis for water treatment and puri$- either solar or arti$cial. Solar-driven TFS photoreactors are
cation. normally 9at slabs of simple design (TFSFW con$guration
The most common con$gurations of TFS photocatalytic in Table 1). The con$guration can be either of a falling $lm
reactors are that of falling $lm or thin-$lm annular/9at reac- (Turchi, Mehos, & Pacheco, 1993) or 9ow between paral-
tors (Li Puma & Yue, 1999; Alfano et al., 2000). Other less lel plates (Alfano et al., 2000). Solar radiation can hit the
conventional designs include the fountain photocatalytic re- slab directly or can be concentrated by the use of heliostats
actor (Li Puma & Yue, 2001a–c). Falling $lm con$gurations (Malato, Blanco, Vidal, & Richter, 2002). In the former
have the advantage over thin-$lm annular/9at reactors of a case both the di4use and the direct components of solar
very high mass transfer rate of the reducing agent oxygen UV radiation will illuminate the 9at slab while in the latter
and do not su4er from the $lming problem (Braun, Maurette, case only the direct component will be used. However, solar
& Oliveros, 1991) which is typical of reactor con$guration irradiation is available only during daylight hours and its’
in which photons enter the reactor through a transparent intensity is dependent upon the terrestrial latitude and the
window such as in thin-$lm annular/9at reactors. However, local weather conditions. In addition, land cost and avail-
the latter reactor con$gurations are preferred when the solu- ability can also be a major drawback, as these reactors have
tions to be detoxi$ed contain high concentrations of volatile a planar geometry.
organic carbon (VOC) which will simply be lost from the TFS photocatalytic reactors illuminated by arti$cial
system in falling $lm reactors. sources of UV radiation could represent a more versa-
The modelling of photocatalytic reactors requires a com- tile system. In this case, under controlled conditions, 24 h
plex analysis of the radiation $eld in the photoreactor irradiation can be achieved, although this would be ob-
(Cassano et al., 1995; Alfano et al., 2000). This analy- tained at the expense of electricity costs. In addition, the
sis, linked to the modelling of the 9uid-dynamics and the height-to-volume ratio of such reactors can be greatly in-
reaction kinetics, produces integro-di4erential equations creased thus reducing land costs. The geometry of TFS
which in the majority of cases require demanding numerical photocatalytic reactors illuminated by arti$cial sources of
solutions. Further development of the heterogeneous pho- UV radiation depends on the shape and size of the ra-
tocatalysis process on an industrial scale will be facilitated diation source. Arti$cial sources of UV radiation for the
by the availability of simpler mathematical models that can design of photocatalytic reactors are almost invariably long
be easily used for scale-up and design. Recently, we have cylindrical lamps which can be either mercury discharge or
presented a di4erent simpler approach to the analysis of 9uorescent UV lamps. Medium and high pressure mercury
TFS photocatalytic reactors which brings in complete and discharge lamps are high wattage UV lamps that can pro-
experimentally validated dimensionless models for falling duce high intensities of UV radiation. Although these lamps
$lm and fountain photocatalytic reactors (Li Puma & are currently used in the industrial environment, they would
Yue, 1998a, b, 2001a–c). not be appropriate for use in TiO2 photocatalytic systems
In the present paper, we expand this approach to include because of the diminished quantum yields of photocatalytic
steady-state continuous 9ow TFS photocatalytic reactors reactions at high intensity values of the incident radiation
for water puri$cation using solar and arti$cial sources (Ollis, Pelizzetti, & Serpone, 1989). In addition, the unit
of UV radiation and we present modelling results for cost of the lamp and of the ballast system is high.
a number of ideal 9ow operation modes. In addition, a In a photocatalytic reactor, a higher e?ciency of photon
methodology for the optimal geometrical design of a highly utilization would be obtained using dilute arti$cial sources
e?cient con$guration of TFS photocatalytic reactors is also of UV radiation such as 9uorescent or low pressure mercury
presented. arc lamps. Comparison of the e4ectiveness of photon-based
oxidation processes in a falling $lm photoreactor has shown
that low pressure mercury arc lamps are more e?cient than
2. Geometry of TFS photocatalytic reactors and the 9uorescent UV sources for wastewater treatment by photo-
radiation source catalysis (Li Puma & Yue, 1999).
Two feasible con$gurations of TFS photocatalytic reac-
The initial step in the design of a TFS photocatalytic re- tors are shown in Table 1:
actor is the selection of an appropriate radiation source(s). TFSFW—the liquid is 9owing along a 9at wall (FW),
The remaining design features are a trade-o4 between reac- as a falling $lm or between parallel plates and the reactor
tor geometry, spatial arrangement of the photon source(s), is irradiated by lamps equally distributed above and with a
photon entry system and selection of an appropriate photo- light re9ector or directly by solar radiation.
G. Li Puma, P. L Yue / Chemical Engineering Science 58 (2003) 2269 – 2281 2271

Table 1
TFS photocatalytic reactors and idealized 9ow operation modes

TFSIW—the liquid $lm is 9owing along the internal wall 3. The geometrical design of TFSIW photocatalytic
(IW), of a cylindrical reactor as a falling $lm or in an annulus reactors
with the lamp mounted along the central axis.
Of these two options, the TFSIW is the preferred con$g- A schematic representation of the TFSIW geometry is
uration for commercial design when using UV lamps. The shown in Fig. 1. The cylindrical symmetry of a TFSIW con-
advantages of this reactor geometry are: $guration allows a two-dimensional analysis along the (r; z)
plane to be made. The geometrical parameters of interest for
(1) The cylindrical symmetry of the system allows all parts the design of a photoreactor of con$guration TFSIW are:
of the photoreactor to be uniformly illuminated with (i) the radius and length of the lamp, and (ii) the length and
little loss of e?ciency providing that the lengths of the radius of the reactor. From a qualitative point of view, the
reactor and lamp are comparable; following points should be considered:
(2) There is improved photon utilization as back-scattered
photons have a low probability of escaping from the (1) The larger the ratios of lamp length-to-lamp radius
reactor and can be recaptured by the liquid $lm; and reactor radius-to-lamp radius, the less important
(3) There is no need for a light re9ector which inevitably is the consideration of the lamp radius as a design
introduces a loss in photon utilization and adds to the parameter.
overall cost of the unit; (2) The wider the reactor, the larger the number of pho-
(4) Improved safety aspects such as the containment of the tons that will escape from the top and bottom of the
lamp inside the reactor in case of $re or explosions; reactor.
(5) Simple design that require the use of a single lamp and (3) The longer the reactor, the greater the proportion of
occupies minimum 9oor space. photons that will be absorbed within the reactor.
2272 G. Li Puma, P. L Yue / Chemical Engineering Science 58 (2003) 2269 – 2281

This assumption is good when the scattering coe?cient of


the suspension is much smaller than the absorption coe?-
cient. A more rigorous method to account for the e4ects of
scattering will be published in a future paper.
Under these assumptions, the integration of Eq. (1) yields
the incident radiation $eld at the inner wall of the reactor,
(r = R):
    
Sl z−x z−x−L
IR; z = arctan − arctan : (3)
4 R R R
Physical considerations suggest that the maximum value of
the intensity of the incident radiation would be at the axial
position
L
z=x+ (4)
2
and its value would be
 
max Sl L
IR = arctan : (5)
2 R 2R
If the lamp is perfectly centred with respect to the reactor,
H −L
x = xc = (6)
Fig. 1. The geometry of a TFSIW photoreactor showing a section of the 2
photoreactor along the (r; z) plane. L is the length of the lamp, H is the
the reactor would give the highest performance (Esplugas,
length of the reactor, R is the distance from the lamp axis to the inner
surface of the reactor,  is the thickness of the thin $lm, x determines Ibarz, & Vincente, 1983). Under this circumstance, intro-
the position of the lamp and j is the mass 9ow rate of substrate j per ducing the geometrical design parameters of TFSIW photo-
unit width of reactor wall. catalytic reactors,
H
= ; (7)
A quantitative assessment of the e4ect of size and geome- L
try on the e?ciency of photon absorption can be performed L
= (8)
by using the Linear Source Spherical Emission (LSSE) radi- R
ation model (Jacob & Drano4, 1966). Common UV sources and the dimensionless axial coordinate,
for medium- and large-scale operations are cylindrical in z
z∗ = (9)
shape with a large ratio of length-to-radius. Consequently, H
neglecting the lamp radius in the formulation of the radi- the fraction of the intensity of the incident radiation at the
ation emission model would not introduce a large error, inner wall of the reactor can be written as follows:
especially for the purpose of obtaining simple design infor- IR;c z
mation without having to perform complex mathematical # = max
IR
analysis. The LSSE model assumes that the lamp is a line ra-
 
diation source with each one of its points emitting radiation 1 arctan[ 2 (2z ∗ −  + 1)] − arctan[ 2 (2z ∗ −  − 1)]
isotropically and in every direction. With reference to Fig. 1, =
2 arctan(  )
2
the incident radiation $eld according to the LSSE model, is
(10)
described by
 x+L  This expression gives the axial distribution of radiation
Sl exp[ − (1 − Rr ) r 2 + (z − x )2 ] 
Ir; z = dx ; in the reactor. An example calculation of the variation of
4 x r 2 + (z − x )2 # with the dimensionless axial coordinate z ∗ for  equal to
(1) 22.2 and for di4erent values of  is shown in Fig. 2. The
where sections z ∗ = 0 and 1 represent the position at top and the
bottom section of the inner wall of the reactor, respectively.
Sl = 2 rl Iw (2)
For =1, the fraction of the intensity of the incident radiation
in which  is the absorption coe?cient and Sl is the radiation in these two sections is approximately equal to 0.5. As the
emission of the lamp per unit time and unit length of the parameter  increases, i.e. as the reactor length increases
lamp. with respect to the lamp, # at these two sections decreases
For the purpose of the geometrical design of a TFSIW accordingly. Fig. 2 shows that for values of  greater than
photoreactor, the thickness of the liquid $lm can be ne- 1.3 there would be little advantage in designing a longer
glected, that is (R +  ∼
= R). It is assumed that all photons reactor. For  = 1:5 the top and the bottom sections of the
entering the liquid $lm are absorbed within the liquid $lm. reactor will be poorly illuminated.
G. Li Puma, P. L Yue / Chemical Engineering Science 58 (2003) 2269 – 2281 2273

Fig. 2. Axial pro$les of the fraction of the intensity of incident radiation Fig. 3. Fraction of radiant power transmitted to a TFSIW photoreactor
in a TFSIW photoreactor, for the parameter  = 22:2 and for di4erent from a linear radiation source as a function of the reactor geometry.
value of the parameter .


1=2
 2
The radiant power absorbed within the photoreactor can 1 
= ( + 1) + 1
be derived by integrating the di4erential radiant power  2
dW c received by a di4erential element ds of the reactor
wall,
 2 1=2 
 
dW c = IR;c z ds cos ’; (11) − ( − 1) +1 (18)
2 
where
A plot of the fraction of radiant power absorbed within
ds = 2 R d z; (12) the photoreactor, , as a function of the geometrical pa-
  rameters  and , is shown in Fig. 3. For  greater than
z − x
’ = arctan ; (13) 20, more than 95% of the energy is absorbed within the
R
reactor.
R In practical reactor design, the dimensions of the lamps,
cos ’ =  : (14)
R2 + (z − x )2 viz. radius and length, may depend on the lamp manufac-
It follows turer, therefore the only geometrical parameters which can
  be more freely changed for a given lamp are the radius
H xc +L
cS l R2 1 and the length of the reactor. The radius of the reactor
W = d z d x (15)
2 0 xc [R2 + (z − x )2 ]3=2 should be selected in relation to the design 9ow rate and
to the desired operative 9ow regime: laminar or turbulent
and the radiant power absorbed within the photoreactor
9ow. The length of the reactor required to achieve a de-
would be

1=2 sired substrate conversion will be derived from the mathe-
  2
matical modelling of the reactor. The optimal length of the
W c = Sl R ( + 1) + 1 lamp in relation to the length and radius of the reactor and
 2
for a desired transmission factor , can be estimated from

 2 1=2  Fig. 3.
 
− ( − 1) +1 : (16)
2 
4. Modelling of TFS photocatalytic reactors
The overall radiant power of the lamp is
Wl = Sl L (17) In this section a general mathematical model for the
steady-state continuous 9ow TFS reactors of Table 1 is pre-
therefore the fraction of radiant power absorbed within the sented. In accordance with the LSSE-LSPP mathematical
photoreactor (or transmitted to the photoreactor) would be model recently developed by Li Puma and Yue (1998a, b)
Wc for a laminar falling $lm photocatalytic reactor, the follow-
=
Wl ing assumptions are made:
2274 G. Li Puma, P. L Yue / Chemical Engineering Science 58 (2003) 2269 – 2281

4.1. Radiation source model In Eq. (7), I; z is the incident radiation at a point (; z) of
the liquid $lm and  = ccat is the absorption coe?cient
In the TFSFW con$guration, the intensity of the incident averaged over the spectrum of the incident radiation. kT is
radiation is uniform and described by its value averaged over a constant that takes into account all other factors that may
the useful range of the incident radiation spectrum. a4ect the overall quantum yield, with the exception of the
In the TFSIW con$guration the emission of radiation from substrate concentration.
the lamp is modelled using the linear source spherical emis- Justi$cation for using Eq. (19) in preference to a more
sion (LSSE) model by Jacob and Drano4 (1966). The lamp popular Langmuirian rate equation has been given elsewhere
is considered to be a line source. Each point on this line is (Li Puma & Yue, 1998a). The Langmuir type of kinetic
assumed to emit radiation in every direction and isotropi- model does not o4er signi$cantly better representation of
cally. There is no light absorption, scattering or emission in the present rate data; therefore for simplicity, we choose to
the space between the lamp and the thin-$lm. use the power law model. The objective is not intended to
be a rigorous representation of the reaction mechanism but
4.2. Light absorption model to provide a satisfactory model of the rate data. The present
kinetic rate equation does not include the e4ect of reaction
In the reaction space, i.e., the thin-$lm, it is assumed that intermediates, since most of the applications of photocat-
the useful photons entering the liquid $lm travel through the alytic detoxi$cation involve lightly contaminated water in
liquid $lm only in the direction orthogonal to the main 9ow which the e4ect of the intermediates can often be neglected.
along parallel planes (LSPP model). However, if the intermediates are found to a4ect the kinetics
The LSPP model has proved to be satisfactory for TFS of destruction of the primary substrate these e4ects will be
photocatalytic reactors. Using a more complex radiation seen only when the concentration of the primary substrate
model, i.e. the LSSE model, does not bring any substantial has been reduced considerably near the end of the steeper
bene$t to the accuracy of the reactor model (Li Puma & part of the degradation curve (e.g., 70% conversion). In this
Yue, 1998a). situation, Eq. (19) will only be valid for the partial conver-
The useful photons, those photons with energy higher sion of the primary substrate.
than the band-gap of the semiconductor photocatalyst, are
absorbed only by the solid photocatalyst particles and the 4.5. Geometry
scattering albedo of the catalyst [! = =( + )] is smaller
than approximately 0.3. Under this circumstance we have In the TFSFW con$guration, the length of the 9at plate
shown that the e4ect of light scattering can be safely ne- is H and the width is T .
glected (Li Puma & Yue, 2001a). It is important to observe In the TFSIW con$guration, the lamp is perfectly cen-
that the scattering albedo given above should be evaluated tered with respect to the reactor and the reactor geometry is
using the absorption and scattering coe?cients determined described by the following parameters:
experimentally under the prevailing conditions of the slurry H
suspension (i.e., with the observed degree of agglomeration = ; (20)
L
of TiO2 in the suspension).
L
= ; (21)
4.3. Fluid-dynamics model R
In addition, the $lm thickness is much less than the reactor
Steady-state, unidirectional, continuous 9ow conditions radius:
prevail. The liquid phase is a Newtonian 9uid with constant
physical properties. The catalyst particles are considered to R; (22)
be uniformly distributed within the liquid $lm, however, the
concentration of solids is not so high as to cause substantial R
changes in the rheological properties of the 9uid. ≈ 1: (23)
R+

4.4. Reaction kinetics model 4.6. TFS dimensionless reactor model

At each point in the reaction space, the rate of the reaction A material balance for a substrate j written for an annular
of a substrate j is considered to be proportional to the local (TFSIW) or transversal (TFSFW) slice within the liquid
volumetric rate of energy absorption (LVREA) (I; z ) raised $lm reduces to the following expression (Bird, Stewart, &
to the mth power and to the nth power of the local substrate Lightfoot, 1960):
concentration Cj :
@Cj @2 Cj
vz =D + rj : (24)
− rj = kT (I; z )m Cjn : (19) @z @r 2
G. Li Puma, P. L Yue / Chemical Engineering Science 58 (2003) 2269 – 2281 2275

Table 2
Steady-state ideal 9ow systems and radiation model of TFS reactor con$gurations

If the di4usive 9ux is much smaller than the convective 9ux the following general dimensionless design equation for a
and introducing the dimensionless variables thin-$lm photoreactor can be derived:
 dCj∗
∗ = ; (25) vz∗ ∗ = −NDa (I∗∗ ; z∗ )m (Cj∗ )n ; (31)
 dz
z
z∗ = ; (26) where NDa is the DamkRohler number of the system
H
max m
I; z kT (I=0 ) (CjO )n max m
kT (I=0 ) H (CjO )n−1
I∗∗ ; z∗ = max ; (27) NDa = = ; (32)
I=0 vzmax CjO =H vzmax
Cj which is the ratio of the overall reaction rate, calculated at the
Cj∗ = O ; (28)
Cj inlet concentration and at the maximum incident radiation,
vz to the maximum input mass 9ow rate of the reactant.
vz∗ = max ; (29)
vz It should be noted that the optical thickness de$ned in
Eq. (30) is expressed in terms of  =ccat rather than Nap =
and the optical thickness of the photoreactor
( + )ccat since it has been assumed that the scattering
' = ; (30) albedo is small.
2276 G. Li Puma, P. L Yue / Chemical Engineering Science 58 (2003) 2269 – 2281

The dimensionless velocity pro$les for steady state (i) For a thin-$lm photoreactor where a fraction + of the product
falling $lm laminar 9ow, (ii) plug 9ow, and (iii) slit 9ow stream QR is recycled back to the inlet of the photoreactor
systems are given in Table 2. For all three ideal 9ow sys- the overall conversion for the system -j will be given by
tems the Reynolds number is de$ned in the same manner. *j
The thickness of the liquid $lm for laminar or turbulent 9ow -j = ; (41)
1 − +(1 − *j )
falling $lm photoreactors can be determined from correla-
tions in the literature (Perry & Green, 1984). For TFS an- where + is the recycle ratio:
nular/parallel plates photoreactors  is the dimension of the
annulus/gap. The dimensionless pro$les of the intensity of QR − QO
+= : (42)
the incident radiation (i.e., the dimensionless LVREA) for QR
the TFSFW and TFSIW reactor con$gurations are given in Analytical solution of the expression for A for the TF-
Table 2. SIW con$guration is possible only when the coe?cient
The integration of Eq. (31) with following boundary con- m is equal to one (Eq. (43)). In photocatalysis, this sit-
ditions: uation has been experimentally veri$ed at low irradi-
BC (1) z ∗ = 0; Cj∗ = 1; (33) ation intensities of less than 10−3 Ein m−2 s−1 (Ollis,
1991).
2
(( + 1) arctan[ 2 ( + 1)] − ( − 1) arctan[ 2 ( − 1)] + ln 1+[=2(−1)]
1+[=2(+1)]2
)
Am=1 = : (43)
2 arctan( 2 )
If n=0 (zero order kinetics) and m=1 an analytical solution

CjH ∗
for the reactor conversion can be obtained:
BC (2) z = 1; Cj∗ = = CjH ; (34)
CjO vzmax Am=1 NDa
*j = a + average (1 − exp '); (44)
yields the radial distribution of the dimensionless concen- vz '
tration of substrate j at the reactor outlet section, CjH . where a = 0 for falling $lm laminar 9ow and plug 9ow
For kinetics other than $rst-order the solution of Eqs. systems, and a = 0:5 for the slit 9ow system.
(31)–(34) is The methodology for the evaluation of the model param-
 1=(1−n) eters including the kinetic parameters m and n is reported
H∗ NDa (1 − n) exp(−m'∗ )
Cj = 1 − A ; elsewhere, (Li Puma & Yue, 1998a, b). The only adjustable
vz∗
parameter of the model is kT .
n = 1 (35) The formulation of the mathematical model in dimen-
sionless form allows photoreactor scale-up by dimensional
and for $rst-order kinetics is analysis. If a small-scale or pilot plant TFS photocatalytic
 
∗ NDa exp(−m'∗ ) reactor is optimized with respect to the dimensionless pa-
CjH = exp 1 − A ; n = 1: (36) rameters of the TFS model, scale-up to a larger unit could
vz∗
be obtained using similarity concepts.
In Eqs. (35) and (36) A is a dimensionless coe?cient that is
dependent only on the geometry of the reactor and m (Table
4.7. Absorption of radiation power
2).
The dimensionless mass 9ow rate of j per unit width of
The e4ective radiant power absorbed within the reaction
the reactor wall at the reactor outlet is c
space Wabs can be obtained by integrating the LVREA with
∗ jH respect to the volume occupied by the liquid $lm. The e4ec-
jH = ; (37) tive fraction of radiant power absorbed within the reaction
jO
space (e4ective transmission factor) is e4 . The expressions
where jO is the inlet mass 9ow rate of j per unit width of c
of Wabs and e4 are shown in Table 2 for both the TFSFW
reactor wall, and the TFSIW con$gurations. Once the absorption of radi-
c
ation Wabs is known the apparent quantum yield of a reaction
jO = CjO : (38)
could be determined.
It follows that:
 1  1
H∗ H∗ ∗ vzmax ∗
5. Model simulations
j = j;  d = average vz∗ CjH d∗ (39)
0 v z 0

and the reactor conversion can be computed from The results of the model simulations for the idealized 9ow

conditions in Table 2 are presented for the TFSIW reactor
*j = 1 − jH : (40) con$guration. The photocatalytic reaction is the oxidation of
G. Li Puma, P. L Yue / Chemical Engineering Science 58 (2003) 2269 – 2281 2277

Fig. 4. Model simulations for photocatalytic oxidation of salicylic Fig. 5. Model simulations for photocatalytic oxidation of salicylic
acid in a TFSIW photocatalytic reactor. E4ect of the optical thick- acid in a TFSIW photocatalytic reactor. E4ect of the optical thick-
ness on conversion of salicylic acid with di4erent 9ow conditions. ness on conversion of total organic carbon with di4erent 9ow condi-
Model parameters:  = 1:008,  = 918:5, H = 50 m, NRe = 1454, tions. Model parameters:  = 1:008,  = 918:5, H = 50 m, NRe = 1454,
=1296 m2 kg−1 , Iw =76:6 W m−2 , Csalicylic
O
acid =8:5×10
−3 kg m −3 ,
=1296 m2 kg−1 , Iw =76:6 W m−2 , Csalicylic
O −3 kg m −3 ,
acid =8:5×10
 = 3:64 × 10−4 m2 s−1 , + = 0. Kinetic parameters: m = 0:5, n = −0:44,  = 3:64 × 10 −4 2 −1
m s , + = 0. Kinetic parameters: m = 0:5; n = −0:16,
kT = 5:55 × 10−8 kg(1−n) s−1 W−m m(3m+3n−3) . kT = 6:92 × 10−8 kg(1−n) s−1 W−m m(3m+3n−3) .

salicylic acid in aqueous slurry suspensions of TiO2 Degussa ial distance from the reactor inlet with the three ideal 9ow
P25 which has been reported elsewhere with the estimation systems is shown in Fig 7. In Fig. 6a it can be seen that

of the model parameters (Li Puma & Yue, 1998b). The the pro$le for j;z ∗ for the falling $lm laminar 9ow system
simulation results of the conversion of salicylic acid and is depleted from both sides of the liquid $lm, the side of
total organic carbon as a function of the optical thickness for the liquid $lm close to the internal wall of the reactor and
di4erent idealized 9ow systems are shown in Figs. 4 and 5, the side close to the external wall of the reactor. In conse-

respectively. For these simulations we have chosen a rather quence, the remaining pro$le of j;z ∗ , at the reactor exit,
unrealistic reactor—50 m in height and 0:108 m in diameter, extends from $lm sections from ∗ = 0:25 to 0.55 only. By
operating in single pass—in order to be able to describe the making the reactor slightly longer, the remaining pro$le of

full e4ect of the 9ow conditions on reactor performance. j;z ∗ , at the reactor exit, can be easily depleted as the light
Since the model has been developed in dimensionless form intensity in this region of the transversal section of the $lm
and uses dimensionless parameters, the conclusions of this is reasonably high. Conversely, with the plug 9ow system

analysis can be easily extended to shorter and more practical (Fig. 6b) the pro$le of j;z ∗ is depleted from the side of
reactors. In practice, multiple reactors may be connected in the liquid $lm close to the internal wall of the reactor only.

series to obtain high conversion. Therefore, the remaining pro$le of j;z ∗ at the reactor exit
In Figs. 4 and 5 it can be observed that the plug 9ow and extends from ∗ = 0:6 to 1. This pro$le will be di?cult to
laminar falling $lm 9ow systems give higher reactor per- deplete, even in a longer reactor, as the illumination in this
formance when compared to the slit 9ow system. In Fig. 4, region is poor. As a consequence, the laminar falling $lm
it can be seen that the plug 9ow system is superior to the system has higher performance than the plug 9ow system
laminar falling $lm 9ow system for values of the optical when reactor conversions are higher than 80% (Fig. 7).
thickness up to 1.4 at which the performance of the plug At intermediate values of the conversion and providing

9ow system reaches a maximum. At higher values of the that there is a non-zero pro$le of j;z ∗ in the $lm sections
optical thickness the laminar falling $lm system is superior from ∗ = 0 to 0.4, the plug 9ow system is superior to the
to the plug 9ow system and it can reach a higher maximum laminar falling $lm system as, on average, the 9uid resi-
in conversions. Conversely, in Fig. 5 the plug 9ow system dence time of the $lm sections exposed to higher intensity
is marginally superior to the falling 9ow system at any opti- of radiation is longer than for the corresponding $lm sec-
cal thickness. The reason for these observation can be more tions in the laminar falling $lm system.
easily explained by plotting the radial pro$les of the dimen- The slit 9ow system (Fig. 6c), has reduced performance
sionless mass 9ow rate of salicylic acid per unit width of when compared to falling $lm laminar 9ow and plug 9ow
reactor wall at di4erent axial distances from the reactor in- systems as there is almost complete non-correspondence of

let j;z ∗ , with the three ideal 9ow systems (Figs. 6a–c). The radiation $eld and 9uid residence time for sections from
partial conversion of salicylic acid as a function of the ax- ∗ = 0 to 0.5 of the liquid $lm. In the slit 9ow system the
2278 G. Li Puma, P. L Yue / Chemical Engineering Science 58 (2003) 2269 – 2281

Fig. 6. Model simulations for photocatalytic oxidation of salicylic acid in a TFSIW photocatalytic reactor. Radial pro$les of the mass 9ow rate of
salicylic acid per unit width of reactor wall at di4erent axial distances from the reactor inlet. (a) falling $lm laminar 9ow; (b) plug 9ow; (c) slit 9ow.
The partial conversion is shown in parenthesis. Model parameters: ' = 2, other model parameters from Fig. 4.

dence of the radiation $eld and residence time distribution in


the whole sections ∗ =0 to 1 of the liquid $lm, therefore the
optimal operating conditions would be least a4ected by dif-
ferences in the kinetic coe?cient m. Conversely, in the plug
and slit 9ow systems, this correspondence is lost and there-
fore the di4erences in kinetics of degradation and mineral-
ization are more evident. This suggests that the optimization
of a laminar 9ow system would be easier than for the plug
and the slit 9ow systems.

6. Non-ideal TFS photocatalytic reactors

The inclusion of non-ideal e4ects such as lateral and axial


mixing in the model, would introduce a second adjustable
parameter. In this case, the non-ideal photoreactor can be
Fig. 7. Model simulations for photocatalytic oxidation of salicylic acid simulated to be made of alternating ideal photoreactor sec-
in a TFSIW photocatalytic reactor. Partial conversion of salicylic acid as tions and mixer in series (Fig. 8). The number of sections,
a function of the axial distance from reactor inlet with di4erent velocity N , is the model parameters that takes into account the de-
pro$les. Model parameters: ' = 2, other model parameters from Fig. 4. gree of non-ideality. The range is N = 1 for an ideal reactor,
to N = ∞ for complete lateral and axial mixing (CSTR).
9uid with the least residence time is at the centre of the liquid The algorithm for performing this simulation with the
$lm where the intensity of the incident radiation is relatively TFSIW con$gurations is given in the appendix. The partial

low. Therefore although the pro$le of j;z ∗ is depleted from conversion *j; i and the substrate concentration Cj;Hi at the

both sections ∗ = 0 and 1, the remaining pro$le of j;z ∗ exit of the $rst ideal photoreactor section are calculated. Cj;Hi
at the reactor exit is signi$cant and the reactor conversion then becomes the feed concentration for the following ideal
(Fig. 7) is always less than that of falling $lm laminar 9ow photoreaction section, and the calculations are repeated un-
and plug 9ow systems. til the total length of the non-ideal photoreactor is covered.
The values of the optical thicknesses for optimal conver- Finally, the reactor conversion for the entire non-ideal pho-
sion of salicylic acid and TOC are identical for the falling toreactor is calculated. It should be observed that the length
$lm laminar 9ow system, but are di4erent for the plug and of the last ideal photoreactor section will be shorter then the
the slit 9ow systems (Figs. 4 and 5). In the falling $lm previous one(s) when the model parameter N is a fractional
laminar 9ow system, there is almost complete correspon- number.
G. Li Puma, P. L Yue / Chemical Engineering Science 58 (2003) 2269 – 2281 2279

in the model by a negative value for the kinetic parameter


n. If n was a positive number a reversed result would have
been observed.

7. E'ect of light scattering

The model presented assumes that the e4ect of light scat-


tering can be neglected. This assumption may be acceptable
with TFS photoreactors operating at optimal values of the
optical thickness (i.e. high catalyst concentration) and with
catalyst with scattering albedo less than approximately 0.3.
However, at low catalyst concentration and with catalysts
with scattering albedo greater than 0.5, the e4ect of light
scattering may become important and cannot be neglected
in the model. To take into account the e4ect of light scat-
tering without compromising the simplicity of the model,
the rigorous approach advocated by Cassano et al. (1995)
can be simpli$ed to include a two-9ux absorption-scattering
model by Brucato and Rizzuti (1997) with simpler radiation
attenuation terms. This approach was shown previously in
Li Puma and Yue (2001a–c) for a fountain photocatalytic
reactor.
The extension of the present model of TFS photocatalytic
reactors to include the e4ect of light scattering and the new
model results will be presented in a future publication.

Fig. 8. Schematic representation of model for non-ideal photoreactors.


8. Conclusions

This study has provided a systematic and generic method-


ology for the design and modelling of steady-state continu-
ous 9ow TFS photocatalytic reactors for solar and arti$cial
light sources. The model developed is applicable to the pho-
toreactor con$gurations: TFSFW, (a) falling $lm design or
(b) parallel plates design, and TFSIW (a) falling $lm de-
sign or (b) annular design, operating with three ideal 9ow
conditions: (1) falling $lm laminar 9ow, (2) plug 9ow and
(3) slit 9ow. The model developed, in the present form is
applicable to photocatalysts with scattering albedo less than
approximately 0.3 or to cases in which light scattering can
be neglected. The mathematical model is expressed in di-
mensionless form and scale-up of TFS can be carried out by
dimensional analysis. In addition, the model parameters can
be estimated easily from real systems and model solutions
can be obtained with little computational e4orts. Non-ideal
Fig. 9. Model simulations for photocatalytic oxidation of salicylic acid
in a TFSIW photocatalytic reactor. Non-ideal falling $lm laminar 9ow.
9ow in TFS photocatalytic reactors was modelled using the
E4ect of modelling parameter N on conversion of salicylic acid as a photoreactors and mixers in series model.
function of the optical thickness. Model parameters from Fig. 4 except The comparison of a number of ideal 9ow systems has
O −2 kg m −3 .
Csalicylic acid = 1 × 10 shown that falling $lm laminar 9ow and plug 9ow systems
give high performance depending on the degree of conver-
An example calculation for the oxidation of salicylic acid sion obtained in the photocatalytic reactors. The slit 9ow
in a non-ideal falling $lm laminar 9ow photocatalytic reactor operational conditions should be avoided due to penalties
is shown in Fig. 9. In this example, reactor conversions are in reactor e?ciency. The extension of the present model
higher as N is increased due to inhibition e4ects that have of TFS photocatalytic reactors to include the e4ect of light
been observed in the photocatalytic oxidation of salicylic scattering and the new model results will be presented in a
acid (Cunningham & Al-Sayyed, 1990) that are represented future publication.
2280 G. Li Puma, P. L Yue / Chemical Engineering Science 58 (2003) 2269 – 2281

Notation  mass 9ow rate per unit width of reactor wall,


kg m−1 s−1
∗ dimensionless mass 9ow rate per unit width of re-
a coe?cient in Eq. (44): a=0 for falling $lm laminar
actor wall
9ow and plug 9ow systems, and
 $lm thickness, m
a = 0:5 for the slit 9ow system, dimensionless
 absorption coe?cient averaged over the spectrum
A geometrical coe?cient, dimensionless
of the incident radiation (=ccat × ), m−1
ccat absorbing substrate concentration, kg m−3
Nap mass Napierian extinctance averaged over the spec-
C substrate concentration, kg m−3
trum of the incident radiation (=ccat × 3Nap ), m−1
C∗ dimensionless substrate concentration (=C=C O )
3Nap speci$c mass Napierian coe?cient averaged over
D di4usion coe?cient, m2 s−1
the spectrum of the incident radiation, (= + )
H length of the reactor, m
m2 kg−1
I intensity of incident radiation, W m−2  speci$c mass scattering coe?cient averaged over
IRmax maximum value of intensity of incident radiation the spectrum of the incident radiation, m2 kg−1
at the $rst illuminated surface of the  speci$c mass absorption coe?cient averaged over
$lm in the TFSIW con$guration, W m−2 the spectrum of the incident radiation, m2 kg−1
I∗∗ ; z∗ dimensionless intensity of incident radiation or di-  $lm thickness coordinate, m−1
max
mensionless LVREA (=I; z =I=0 ) ∗ dimensionless $lm thickness coordinate (==)
(1−n) −1
kT kinetic constant, kg s W−m m(3m+3n−3) ratio of circumference to diameter of a circle (∼ =
l thickness of solution traversed by the light, m 3:1428)
L lamp length, m 2 9uid density, kg m−3
m order of the reaction with respect to the LVREA, # fraction of intensity of incident radiation
dimensionless (=IR;c z =IRmax ) dimensionless
n order of the reaction with respect to substrate con- * conversion per pass, dimensionless
centration, dimensionless - total conversion for the system, dimensionless
N number of reactors in non-ideal reactor model, di- + recycle ratio, dimensionless
mensionless ' optical thickness, dimensionless
max m
NDa DamkRohler number (=kT (I=0 ) H (CjO )n−1 =vzmax ) 1 kinematic viscosity, m2 s−1
dimensionless transmission factor (=W c =WL ) dimensionless
NRe Reynolds number (=4vzaverage =1) or (=4=21) di- e4
e4ective transmission factor (=Wabs c
=(THIO )) for
mensionless TFSFW reactor con$guration (=Wabs c
=WL ) for TF-
QO volumetric 9ow rate through system inlet, m3 s−1 SIW reactor con$guration using the LSSE-LSPP
QR volumetric 9ow rate through the reactor, m3 s−1 model, dimensionless
r radial coordinate, m ! scattering albedo (==( + )) dimensionless
rj rate of the reaction with respect to substrate j,
kg s−1 m−3 Subscripts
rl lamp radius, m
R distance between lamp axis and the $rst illuminated abs absorbed
surface of the $lm in the TFSIW j substrate
con$guration, m l lamp
s surface area, m2 r direction along the radial coordinate
S radiation emission per unit time and unit length, R position at r = R
W m−1 w lamp wall
T width of the reactor in the TFSFW con$guration, m z; z ∗ direction along the axial coordinate
x axial position of the lamp, m ; ∗ direction along the $lm transversal coordinate
x supplementary axial coordinate, m
v 9uid velocity, m s−1 Superscripts
v∗ dimensionless 9uid velocity (=vz =vzmax )
W radiant power, W average average value
z axial coordinate, m ∗ dimensionless variable
z∗ dimensionless axial coordinate (=z=H ) c lamp centred
H reactor outlet
Greek letters max maximum value
O system inlet
 geometrical parameter (=H=L) dimensionless R through reactor
 geometrical parameter (=L=R) dimensionless z∗ direction along the axial coordinate
G. Li Puma, P. L Yue / Chemical Engineering Science 58 (2003) 2269 – 2281 2281

Appendix of re9ectance greater than zero. Industrial and Engineering Chemistry


Research, 36, 4748–4755.
Algorithm for the calculation of reactor conversion in a Cassano, A. E., Martin, C. A., Brandi, R. J., & Alfano, O. M. (1995).
Photoreactor analysis and design: Fundamentals and applications.
non-ideal TFSIW photoreactor. Industrial and Engineering Chemistry Research., 34, 2155–2201.
10 N = number of ideal photoreactor sections; Cunningham, J., & Al-Sayyed, G. (1990). Factors in9uencing e?ciencies
of TiO2 -sensitized photodegradation. Part 1–Substituted benzene
acids: Discrepancies with dark-adsorption parameters. Journal of the
20 i = 1; Chemical Society Faraday Transactions, 86, 3935–3941.
Esplugas, S., Ibarz, A., & Vincente, M. (1983). In9uence of lamp position
30 Cj;Oi = CjO ; on the performance of the annular photoreactor. Chemical Engineering
Journal, 27, 107–111.
max m
kT (I=0 ) H (Cj;Oi )n−1 Ho4mann, M. R., Martin, S. T., Choi, W., & Bahnemann, D. W. (1995).
40 NDa; i = ; Environmental applications of semiconductor photocatalysis. Chemical
vzmax Reviews, 95, 69–96.
 i=N   
1  ∗
Jacob, S. M., & Drano4, J. S. (1966). Radial scale-up of perfectly mixed
50 Ai = arctan (2z −  + 1) photochemical reactors. Chemical Engineering Progress Symposium
[2arctan( 2 )]m (i−1)=N 2 Series, 62, 47–55.
 m Li Puma, G., & Yue, P. L. (1998a). A laminar falling $lm
− arctan 2 (2z ∗ −  − 1) d z∗ ; slurry photocatalytic reactor. Part I—model development. Chemical
Engineering Sciences, 53, 2993–3006.
 1=(1−n) Li Puma, G., & Yue, P. L. (1998b). A laminar falling $lm slurry
Cj; i NDa; i (1 − n) exp (−m'∗ )
60 Cj;∗i = = 1 − A i ; photocatalytic reactor. Part II—experimental validation of the model.
Cj;Oi vz∗ Chemical Engineering Science, 53, 3007–3021.

3 1 ∗ ∗ ∗ Li Puma, G., & Yue, P. L. (1999). E4ectiveness of photon-based oxidation
70 *j; i = 1 − v C d ; processes in a pilot falling $lm photoreactor. Environmental Science
2 0 z j and Technology, 33, 3210–3216.
80 Cj;Hi = Cj;Oi (1 − *j; i ); Li Puma, G., & Yue, P. L. (2001a). A novel fountain photocatalytic
reactor for water treatment and puri$cation: Modelling and design.
Industrial Engineering Chemical Research, 40, 5162–5169.
90 IF i ¡ int(N ) THEN i = i + 1 : Cj;0 i = Cj;Hi : GO TO 40; Li Puma, G., & Yue, P. L. (2001b). A novel fountain photocatalytic
reactor: Model development and experimental validation. Chemical
100 IF i ¡ N THEN i = N : Cj;Oi = Cj;Hi : GO TO 40; Engineering Science, 56, 2733–2744.
Li Puma, G., & Yue, P. L. (2001c). The modelling of a fountain
H
Cj;i photocatalytic reactor with a parabolic pro$le. Chemical Engineering
100 *j = 1 − ; Science, 56, 721–726.
CjO Malato, S., Blanco, J., Vidal, A., & Richter, C. (2002). Photocatalysis
110 END: with solar energy at pilot-plant scale: An overview. Applied Catalysis
B: Environment, 37, 1–15.
Ollis, D. F. (1991). Solar-assisted photocatalysis for water puri$cation:
Issue, data, questions. In E. Pelizzetti & M. Schiavello (Ed.),
References Photochemical conversion and storage of solar energy (pp. 593–622).
Ollis, D.F., Pelizzetti, E., & Serpone, N. (1989). Heterogeneous
Alfano, O. M., Bahnemann, D., Cassano, A. E., Dillert, R., & Goslich, photocatalysis in the environment: Application to water puri$cation.
R. (2000). Photocatalysis in water environments using arti$cial and In N. Serpone & E. Pelizzetti (Eds.), Photocatalysis fundamentals
solar light. Catalysis Today, 58, 199–230. and applications (pp. 603– 637). New York: Wiley.
Bahnemann, D., Bockelmann, D., & Goslich, R. (1991). Mechanistic Perry, R.H., & Green, D.W. (1984). Perry’s chemical engineering
studies of water detoxi$cation in illuminated TiO2 suspensions. Solar handbook (6th ed.) (pp. 5 –59). New York: McGraw-Hill Inc.
Energy Material, 24, 564–583. Turchi, C.S., Mehos, M., & Pacheco, J. (1993). Design issues for
Bird, R. B., Stewart, W. E., & Lightfoot, E. N. (1960). Transport solar-driven photocatalytic systems. In D.F. Ollis & H. Al-Ekabi,
phenomena. New York: Wiley. (Eds.), Photocatalytic puri?cation and treatment of water and air.
Braun, A. M., Maurette, M. T., & Oliveros, E. (1991). Photochemical Amsterdam, Elsevier, pp. 789 –794.
technology. New York: Wiley. Turchi, C. S., & Ollis, D. F. (1988). Photocatalytic reactor design: An
Brucato, A., & Rizzuti, L. (1997). Simpli$ed modelling of radiant $elds in example of mass-transfer limitations with an immobilized catalyst.
heterogeneous photoreactors. 2 Limiting “two 9ux” model for the case Journal of Physical Chemistry, 92, 6852–6853.

You might also like