Chemical Engineering Journal 286 (2016) 594601

Contents lists available at ScienceDirect

Chemical Engineering Journal

journal homepage: www.elsevier.com/locate/cej

Performance of continuous stirred-tank reactors connected in series

as a photocatalytic reactor system
Atsuko Miyawaki, Shunya Taira, Fumihide Shiraishi
Section of Bio-process Design, Department of Bioscience and Biotechnology, Graduate School of Bioresource and Bioenvironmental Sciences, Kyushu University, 6-10-1,
Hakozaki, Higashi-Ku, Fukuoka 820-8581, Japan

h i g h l i g h t s

g r a p h i c a l a b s t r a c t

Performance of five CSTRs in series as

a photocatalytic reactor system is

evaluated.

Five CSTRs arranged here can
decompose DNP in the absence of
diffusional resistance.

A five-CSTR system with sufficient
mixing is close to that of a plug-flow
reactor.

Five-CSTR arrangement has practical
potential as a photocatalytic reactor
system.

a r t i c l e

i n f o

Article history:
Received 21 August 2015
Received in revised form 30 October 2015
Accepted 1 November 2015
Available online 10 November 2015
Keywords:
Photocatalytic reaction
CSTR
Film-diffusion resistance
Plug-flow reactor
Immobilized photocatalyst

a b s t r a c t
Film-diffusion resistance significantly reduces the decomposition of organic compounds over an
immobilized photocatalyst in the vicinity. To overcome this problem, the performance of a photocatalytic
system consisting of five continuous stirred-tank reactors (CSTRs) in series was investigated. Each reactor
was a square-shaped container with a glass plate immobilized with a film of TiO2 photocatalyst at the
bottom. Two 4-W blacklight blue-fluorescent lamps fixed above the container provided UV light. A
10-ppm aqueous solution of 2,4-dinitrophenol (DNP) supplied continuously to the reactor was stirred
by an impellor fixed to a shaft with a wind propeller at its upper end. The change in the DNP concentration in each reactor with increasing flow rate agreed with the result calculated by a mathematical model,
constructed for the photocatalytic system, which included a mass-transfer coefficient expressed in terms
of rotational speed. Hence, this model was used to simulate the performance of the reactor system. At
sufficiently high rotational speeds, the conversion of DNP at the exit of the fifth CSTR was substantially
higher than that at the exit of a single CSTR with the same liquid volume, and was close to the calculated
value for a plug-flow reactor without film-diffusion resistance. At low rotational speeds, however, the
DNP concentration remained remarkably high because of large film-diffusion resistance, demonstrating
that the photocatalytic reaction requires sufficient liquid mixing to enhance the reaction rate. Thus, the
system of five CSTRs in series is effective as a photocatalytic reactor system.
2015 Elsevier B.V. All rights reserved.

1. Introduction
The operational mode of a chemical reactor can be divided into
two main groups; batch-wise and continuous-flow modes [1,2].
Corresponding author. Tel./fax: +81 92 642 7603.
E-mail address: fumishira@brs.kyushu-u.ac.jp (F. Shiraishi).
http://dx.doi.org/10.1016/j.cej.2015.11.007
1385-8947/ 2015 Elsevier B.V. All rights reserved.

The typical example for the former mode is the stirred-tank reactor
and that for the latter mode is the continuous stirred-tank reactor
(CSTR) and plug-flow reactor (PFR). Fixed-bed, fluidized-bed, membrane reactors, etc. are also used according to the property of reactants and products, the specificity of chemical reactions, operating
conditions (temperatures, pressures, flow rates), etc. [3]. The CSTR
and PFR are further used to define perfectly-mixed flow and plug

595

A. Miyawaki et al. / Chemical Engineering Journal 286 (2016) 594601

flow, respectively. These ideal flows express both extremes of the

liquid flows in chemical reactors, and the liquid flows in actual
chemical reactors are included between the ideal extremes.
The continuous-flow mode is effective in treating a large
amount of liquid solution [4,5]. Once operation of a reactor is
started in this mode, the reactor can easily be controlled compared
with that in the batch-wise mode. Thus, the continuous-flow mode
has been utilized to operate many industrial reactors [6,7]. In the
perfectly-mixed flow, the liquid solution is well mixed and the
reactant concentration is uniform throughout. Hence, the reactant
in the CSTR is diluted with the progress of the reaction, which leads
to a reduction in the rate of reaction and, in turn, lowers the performance of the reactor. In the plug flow, on the other hand, there
is no liquid mixing in the direction of flow. Hence, the reactant
concentration in this reactor steadily decreases along the flow
direction while taking a largest value at each position, which provides a theoretically highest performance for the reactor [1,2].
When many CSTRs are connected in series, the entire flow pattern
approaches the plug flow; when the number of CSTRs is infinite,
the performance of the reactor is identical to that of the PFR
[1,8]. This theoretical fact is useful when one must inevitably adopt
a CSTR system and wants to increase the performance of the system under a limited reactor volume as highly as possible. However,
little practical application is available as to the multiple CSTRs.
Titanium dioxide (TiO2) irradiated with UV light can decompose
organic compounds in its vicinity [916]. In general, such a photocatalyst is used to decompose harmful substances in an environment. When a reactor containing a suspension of TiO2 fine
particles is operated in a continuous-flow mode, the particles
easily flow out with the effluent. It is reasonable, therefore, to
use TiO2 by immobilizing it on a support [10,1719]. However,
the concentrations of harmful substances in the environment are
generally less than or on the order of 1 ppm (parts per million).
When such substances are treated using the immobilized TiO2,
the very low concentrations create a large film-diffusion resistance
in the vicinity of TiO2 in both gaseous reactions [20] and liquid
reactions [21,22], leading to a marked decrease in the rate of photocatalytic decomposition.
Since the decomposition ability of UV-irradiated TiO2 is not so
high enough to put to practical use at once, it is reasonable to
ensure a reaction field where TiO2 can substantially express its
inherent catalytic activity because of being free from the filmdiffusion resistance. However, the removal of this resistance in
water treatment is not easy compared with that in air treatment.
Although this problem may be solved by mixing a reaction mixture
at a high speed, a high energy consumption problem arises as a
result. For achievement of the practical use of photocatalytic reaction processes, we thus consider that the fundamental knowledge
of chemical engineering must be used to design a reactor that is
free from the film-diffusion resistance and has a higher
performance.
To increase the conversion of a reactant in a PFR, the reaction
mixture must be allowed to flow very slowly. Unfortunately, this
increases the film-diffusion resistance, which may significantly
decrease the rate of photocatalytic reaction. To solve this problem,
we have previously investigated the performance of a continuousflow recirculation reactor system [23]. In this arrangement, a reactant solution introduced from the reactor entrance was recirculated at a high speed between a mixed-flow container and an
annular-flow photocatalytic reactor, and a part of the reaction mixture was allowed to flow out of the exit. We found that as the rate
of photocatalytic decomposition increases with an increase in the
rate of flow of a reaction mixture through the photocatalytic reactor, the reactor performance could be enhanced. Although the ability of this reactor system to eliminate the problem of film-diffusion
resistance is beneficial for photocatalytic processes, it has the

disadvantage of large energy consumption, since the reaction mixture must constantly be recirculated at a very high speed.
An alternative to the photocatalytic reactor is the CSTR, in
which the film-diffusion resistance can be removed if the reaction
mixture is well mixed. However, the performance of the CSTR is
theoretically inferior to that of the PFR, because the rate of reaction
is lowered due to the fluid mixing [1,8]. According to the reactor
theory stated above, this problem might be solved by use of multiple CSTRs in series. Unfortunately, experimental studies to verify
whether this theoretical prediction is correct in an actual reactor
system are highly limited.
The purpose of the present study is to determine a continuousflow reactor system that is appropriate for a photocatalytic reaction using immobilized TiO2 both theoretically and experimentally.
To achieve this purpose, we first select a CSTR that is easy to
remove the film-diffusion resistance and then consider to compensate for the resulting decrease in the rate of reaction by using five
CSTRs in series. DNP (2,4-dinitrophenol) is used as a reactant. A
kinetic analysis of experimental data is carried out to construct a
mathematical model for the reactor system. The mathematical
model is used to examine an effect of mixing speed on the rate
of photocatalytic reaction by computer simulation. Finally, the usefulness of the CSTRs is discussed.
2. Theory
2.1. Kinetic expression for the photocatalytic decomposition of DNP
The DNP molecules are negatively charged [24], while the TiO2
immobilized on a glass surface is positively charged, because an
aqueous solution of DNP at 10 gm3 is acidic (pH  4.5) [25].
Therefore, the photocatalytic decomposition of DNP on the TiO2
surface is strongly affected by an electrostatic interaction. The diffusion flux, NA [gm2-catmin1], toward the photocatalyst surface
can be expressed as [2628].

NA kL C b  C i ek

1
3

2

1

where kL is the mass-transfer coefficient [m -liqm -catmin ], k

is the dimensionless electrostatic potential, and C is the DNP
concentration [gm3-liq] with the subscripts b and i denoting
the concentrations of DNP in the bulk liquid and at the TiO2
surface, respectively. In the absence of the electrostatic interaction,
the rate of the DNP decomposition reaction, r [gm3-liqmin1], is
given by

r

dC i
kK C

H i

dt
1 K H C i 1 KCipb

C b0 C b
10C b0 C b

where k is the rate constant [gm3-liqmin1], Cb0 is the initial

concentration of DNP [gm3-liq], KH is the adsorption equilibrium
constant [m3-liqg1], Kip is the product inhibition constant
[gm3-liq], and t is the time [min]. During the reaction, k changes
as follows.

k

j1 j2 C b0

Cb
C b0

j4

j4 

j4
exp j3 C b0 CC b
j2 C b0 CC b
b0

b0

In a previous study [29], we determined the following values

for the parameters; k = 0.20 gm3min1, KH = 0.146 g1m3,
Kip = 25.0 gm3, j1 = 2.44, j2 = 0.0378, j3 = 1.60, and j4 = 0.519.
The value of kL in Eq. (1) changes in proportion to the degree of liquid mixing. If a reaction mixture is well mixed, the stagnant film
near the TiO2 surface is disturbed, so that DNP molecules in the bulk
can reach the TiO2 surface rapidly. The present study expresses kL as

596

A. Miyawaki et al. / Chemical Engineering Journal 286 (2016) 594601

a function of the rotational speed of mixing, R [revmin1], and

determines their relationship experimentally. If there is no diffusional resistance and the reaction is affected by an electrostatic
interaction alone, the rate of decomposition is expressed as

rb

kK H C b


ek K H C b 1 CKb0ip

C b0 C b
10C b0 C b

f kL r exp
 Ef
0

kK H C b0
0
ek K H C b0

10

To obtain a value of kL satisfying a relation of f(kL) = 0, we apply

the NewtonRaphson method to Eq. (10). A derivative value of f(kL)
with respect to kL is calculated using
0

f kL

df kL
dEf kK H C b0

dkL
dkL ek K H C b0

11

At a steady state, the rate of decomposition is equal to the diffusion

flux, given by

and a new estimate of kL is obtained from

r aNA

kL kL

where a is the specific surface [m2-catm3-liq] given as the ratio of

the surface area of catalyst A [m2-cat] and the liquid volume of a
reaction mixture V [m3-liq]. Since A = 9.14  103 m2-cat and
V = 9.0  105 m3-liq, the value of a is equal to 101.5 m2-catm3liq. Since Eq. (5) contains Ci, which is difficult to measure because
it is the DNP concentration at the glass surface, this is solved with
respect to Ci and the resulting equation is substituted into Eq. (2)
in order to delete Ci and then express the rate of decomposition
as a function of Cb, which can be measured easily. To perform this
derivation simply, we introduce the following dimensionless
numbers.

P 1

C b0
C b0  C b
b
bb0  bb
1 b0
;
K ip 10 C b0  C b
x c bb0  bb

bb K H C b ;

/ K H k=akL ;

bi K H C i ;

x K H K ip ; c 10K H

bi
Pbb  bi ek
1 bi

Solving Eq. (6) with respect to the dimensionless DNP concentration

at the TiO2 surface, bi, leads to

bi

/  Pbb Pek

q
2
/ Pbb Pek  4P/bb
2Pek

The catalytic effectiveness factor, Ef, is given as the ratio of the

actual rate of reaction, r, and the ideal rate of reaction, rb, as follows.
bi

Ef

r
1bi

b
r b ekbb


q
2
ek bb / Pbb Pek  / Pbb Pek  4P/bb
b

2/bb

Thus, the rate of photocatalytic reaction actually taking place on the

TiO2 surface is expressed in terms of only the measurable concentration, Cb, as follows.

r Ef

kK H C b


ek K H C b 1 CKb0ip

C b0 C b
10C b0 C b

i1

i1

 f kL

i1

=f kL

12

The convergence criterion is given as



ki1  ki 
 L
L 
 < e i 1; 2; . . .

 ki1 

13

where e was set at 1014. Although dEf/dkL in Eq. (11) can be calculated using analytically derived equations [23], we carried out this
calculation by a highly-accurate differentiation method [31]. For
this series of calculations of kL, the rates of decomposition were
measured at various rotational speeds.
In general, kL is expressed in terms of a power-law function of
the flow rate of a reaction mixture in a reactor [30,32,33]. However,
we express kL in terms of R instead of the flow rate, given by

kL p Rq

Consequently, Eqs. (1) and (2) are rewritten as

2.2. Determination of the mass-transfer coefficient

The value of kL greatly depends on the geometry of a reactor,
and the flow pattern of fluid within it [23,30]. Hence, kL should
be determined fundamentally for each reactor. Since in the present
study, mixing in each CSTR is performed by a wind-driven impellor, the value of kL changes in proportion to the rotational speed
of mixing. When the initial rate of decomposition, rexp
0 , is experimentally measured at a given rotational speed, R, and the initial
DNP concentration, Cb0, the following algebraic equation with an
unknown value of kL can be formulated.

14

Thus, kL is plotted against R in a double-logarithmic chart, and the

experimental constants, p and q, are determined from the slope
and intercept of a straight line obtained by a least squares method.
2.3. CSTR model
When expressing the liquid volume in each reactor by V [m3liq] and the flow rates of an influent and effluent at each reactor
by t [m3-liqmin1], we can formulate the following design equations for five CSTRs in series.

dC b;i
t C b;i1  C b;i  r i V
dt

i 1; . . . ; 5

15

Assuming a steady state leads to

f i C b;i t C b;i1  C b;i  ri V 0 i 1; . . . ; 5

16

where the subscript i represents the reactor number. Since Cb,0 is

known, applying the NewtonRaphson method to Eq. (16) allows
us to calculate the DNP concentration in each reactor, Cb,i
(i = 1, . . ., 5), from the first reactor in order.
2.4. Design equation for a plug-flow reactor
The reaction efficiency of a plug-flow reactor, where a reaction
mixture is not mixed in an axial direction, is higher than that of a
mixed-flow reactor, where a reaction mixture is uniformly distributed in a liquid solution [1,2]. In a plug-flow reactor with immobilized TiO2, a very low flow rate must be maintained in order to
attain a high conversion of a reactant. Unfortunately, since the
film-diffusion resistance is likely to become significant under such
a condition, the reaction efficiency of the plug-flow reactor would
be lowered substantially. However, this type of reactor is theoretically predicted to provide the highest performance in the absence of
diffusional resistance. Consequently, we set the conversion in this
reactor as a target value for the multiple CSTRs in series.
In the absence of film-diffusion resistance, the design equation
for a plug-flow reactor with immobilized TiO2 is given as

t

dC b
kK H C b

dV
ek K H C b 1 CKb0ip

C b0 C b
10C b0 C b

17

A. Miyawaki et al. / Chemical Engineering Journal 286 (2016) 594601

with the initial condition

C b C b;in at V 0

18

where t is the flow rate [m -liqmin ], V is the liquid volume in the

reactor [m3] and the origin of the coordinate is set at the reactor
entrance. Integrating Eq. (17) from V = 0 to the total liquid volume,
Vtotal, provides the DNP concentration at the reactor exit, Cb,out
[gm3]. Numerical integration was performed by the RungeKutta
method.
3

1

3. Experimental method
3.1. Materials and reagents
DNP (special grade) was purchased from Wako Pure Chemical
Industry, Ltd. (Osaka, Japan). A TiO2 (anatase)-coating solution
was donated from the Kon Corporation (Saga, Japan) and consisted
of an aqueous 35% hydrogen peroxide solution containing 2.4 wt%
TiO2 prepared from aqueous titanium tetrachloride (Toho Titanium
Co., Ltd.; 16 wt% Ti). The detail of this solution is described elsewhere [29]. A square Pyrex glass (9.6 cm long, 9.6 cm wide, and
0.3 cm thick) with a circular hole of diameter 1 cm in the center
was used to immobilize TiO2. Two 4-W blacklight bluefluorescent lamps (Panasonic, Tokyo, Japan) with a UV wavelength
of 300400 nm were used as a light source in each reactor.
3.2. Immobilization of TiO2 [29]
A TiO2 coating solution was applied by a brush onto one side of
the square glass plate, which was then inclined diagonally for
about 10 min to remove an excess coating solution and then

597

heated in a muffle furnace at 500 C for 60 min. This procedure

was repeated four times in order to load a sufficient amount of
TiO2 on the plate. The resulting TiO2 film on the glass plate was
stable and no reduction in the photocatalytic activity was observed
at least during repeated experiments. The immobilized TiO2 glass
plate was always stored in distilled water to protect the film
surface from drying.

3.3. Experimental apparatus and procedure

A stainless-steel square container (10 cm long and 4 cm deep)
was used as the reactor. The TiO2-immobilizing glass plate was
placed at the bottom of the reactor. A stainless steel rod (3 mm
in diameter and 10.5 cm long), with four impeller blades at the
lower end and four propeller blades at the upper end, was set in
the center of the reactor (Fig. 1). First, 9.0  105 m3 of an aqueous
DNP solution at 10 gm3 was poured into the reactor and mixed
with the impeller, rotated by blowing air from a pump to the propeller. The photocatalytic reaction was initiated by irradiating the
TiO2 immobilized on the glass plate (0.9 cm in liquid depth) with
UV light from the two blacklight blue-fluorescent lamps. About
4  106 m3 of the reaction mixture was sampled from the reactor
at appropriate time intervals and the DNP concentration was
determined by measuring the absorbance of the sample solution
at 357 nm with a spectrophotometer (Amersham Biosciences Corporation; Ultrospec 1100 pro). The sample solution was immediately returned to the reactor. UV irradiation was halted during
the measurement.
The multiple CSTR experiment was conducted by connecting
five such reactors in series. The reactors were placed one-by-one
on each step of five stairs, as shown in Fig. 1. The whole reactor

Fig. 1. Schematic diagram and photograph of the five CSTRs in series. Each square-shaped reactor has a glass plate covered with TiO2 on the bottom of the vessel, two
blacklight blue fluorescent lamps for UV irradiation, and a wind propeller for liquid mixing.

598

A. Miyawaki et al. / Chemical Engineering Journal 286 (2016) 594601

system was quickly filled with an aqueous solution of 10 gm3

DNP. The flow rate was set up at 6.0  106 1.8  105 m3min1
and the reaction was initiated by turning on the blacklight bluefluorescent lamps. The reaction mixture overflowed from an upper
reactor and fed into a lower reactor. The depth of the liquid in each
reactor was kept at roughly 9 mm during reactor operation. In this
depth, the UV light could sufficiently excite the TiO2 film and provide its catalytic activity [11]. The UV intensity at the liquid surface
was 0.992 mWcm2, which was measured with an instrument
(CUSTOM Co., Tokyo; UVA-365). For reference, the UV intensity
of the sunlight was 1.76 mWcm2 on a clear day in the summer
(noon, July 15 in 2015) in Fukuoka city. The liquid volume in each
reactor fluctuated around 9.0  105 m3 because of a slight
differences in the heights of the effluent exits of each reactor,
which changed the residence time of the reaction mixture in each
reactor. When the volume of the effluent from the fifth reactor
amounted to three times of the total volume of the reaction mixture in the five reactors (4.5  104 m3), the DNP concentration
had become almost stable. The reaction mixture in each reactor
was sampled to determine the DNP concentration. The expression
for the calibration curve is as follows; Absorbance []
= 0.0591  DNP concentration [gm3] with R2 = 0.988.
The photocatalytic reaction in a liquid solution may cause a lack
of oxygen [3437]. In the present reaction system, however, oxygen was supplied from the air to the liquid solution without any
delay and the photocatalytic reaction proceeded smoothly. This is
because the rate of photocatalytic decomposition is generally not
high, the DNP concentration was very low, the upper part of the
reactor was opened, and the liquid solution was well mixed.

Fig. 3. Relationship between the mass-transfer coefficient and the rotational speed
of liquid mixing with a wind propeller. The experiments were performed using a
batch reactor system. The solid line shows the values calculated by Eq. (19).

4. Results and discussion

4.1. Effect of rotational speed on the rate of DNP decomposition
DNP at 10 gm3 in a 9.0  105-m3 aqueous solution was photocatalytically decomposed in a batch-wise manner, and r0 determined at different rotational speeds. Fig. 2 shows a plot of the
measured r0 against R. The value of r0 increases with an increase
in R, and then approaches a maximum of 0.189 gm3min1, which
is the allowable value at R = infinity, C0 = 10 gm3, and the resulting electrostatic interaction. This is because the stagnant solution
in proximity to the immobilized TiO2 is well mixed, with an
increase in R, thereby reducing the film-diffusion resistance.
The values of kL were determined by fitting the calculated
values of r0 to the corresponding experimental values of r0 in

Fig. 4. Simulation of the relationship between the catalytic effectiveness factor and
the rotational speed of liquid mixing with a wind propeller in a batch reactor
system.

Fig. 2 (see the procedure described in Section 2.2). Only the values
of r0 shown by the symbol s were used in this determination,
because the values of r0 shown by the symbol D in the region of
high DNP concentration are almost the same as the rate constant
k and a slight measurement error of r0 caused an abnormal
increase in the estimated value of kL. Consequently, the values of
kL in this region have been excluded. The values calculated for kL
are plotted against R in Fig. 3. Applying the least-square method
to the relationship between these variables provides

kL 4:23  106 R1:16

19

that was applied to Eq. (8) to calculate Ef and finally obtain r0 at a

given value of R using Eq. (9). It is clear that the calculated line successfully expresses a variation in the measured values. Fig. 4 shows
the relationship between Ef and R in the photocatalytic treatment of
10 gm3 DNP in a 9.0  105 m3 aqueous solution in a batch reactor
with a wind propeller. The value of Ef is close to unity at R > 130
rev.min1, indicating that the film-diffusion resistance is negligible
in this region, and suggesting that it is possible to operate the five
CSTRs in series under the condition of negligible film-diffusion resistance if the reaction mixture is well mixed by the wind propeller.

Fig. 2. Relationship between the initial rate of decomposition of DNP and the
rotational speed of liquid mixing with a wind propeller. The experiments were
performed using a batch reactor system. The solid line shows the values calculated
by the mathematical model.

4.2. Steady-state DNP concentrations at different flow rates

Fig. 5 shows the DNP concentrations at a steady state in
each reactor under different flow rates. It is clear that the DNP

A. Miyawaki et al. / Chemical Engineering Journal 286 (2016) 594601

Fig. 5. Relationship at a steady state between the DNP concentration in each

reactor and the flow rate, for the decomposition of DNP using five CSTRs in series.
The calculated values are shown by solid lines for five CSTRs, by a broken line for
the plug flow reactor, and by a chain line for a single CSTR with the same liquid
volume as the five CSTRs. The data symbols show the experimental for each reactor:
(s) 1st; (h) 2nd; (4) 3rd; (e) 4th; (r) 5th.

concentration in the same reactor increases with an increase in the

flow rate, and that in the lower-step reactor certainly decreases at
the same flow rate. The solid lines show the calculated results by
the mathematical model using the parameter values that were
determined from the experimental data, successfully expressing
the experimental relationship between the DNP concentration in
each reactor and the flow rate.
The DNP concentration calculated at the fifth reactor is close to
that in the plug-flow reactor without diffusional resistance,
indicating that when five reactors are used in series, it is possible
to operate the CSTR system with roughly the same performance
as that of the plug-flow reactor. This result is supported by a
theoretical prediction [8]. Furthermore, when the same treatment
is carried out using a single CSTR with a liquid volume equivalent
to that of the five CSTRs in series combined (4.5  104 m3), the
DNP concentration in the region of low flow rate is higher; that
is, the DNP conversion is lower in the single CSTR than the five
CSTRs in series. For example, at 6  106 m3min1, the conversion
of DNP at the fifth reactor is 94.9%, while that in the single CSTR is
86.4%. This difference is due to the decrease in the DNP concentration caused by liquid mixing, which results in a decrease in the rate
of decomposition of DNP. In the treatment of a wastewater containing harmful chemical compounds, it is favorable to increase
the conversion to a value that is as high as possible. The use of five
CSTRs in series enables us to achieve a higher conversion at a
shorter spacetime. It is thus obvious that the present reactor system is useful.
The agreement of the calculated results with the experimental
data indicates that the mathematical model is useful to characterize the five CSTRs connected in series. In the following, therefore,
computer simulation is carried out using the mathematical model
in order to elucidate the effect of film-diffusion resistance on the
photocatalytic reaction and judge if the present reactor system is
appropriate for practical application.
4.3. Effect of rotational speed on the conversion of DNP
In general, the rate of decomposition is significantly reduced
over the immobilized TiO2 due to the large film-diffusion resistance. To evaluate the seriousness of this problem, the relationship
between the DNP concentration in the fifth reactor and the flow
rate as a function of the rotational speed is shown in Fig. 6. It is

599

Fig. 6. Simulation of the relationship between the DNP concentration in the fifth
reactor and the flow rate, for the decomposition of DNP using five CSTRs in series.
The calculated values for different rotational speeds of liquid mixing using a wind
propeller are shown by solid lines. Those for infinite rotational speed are shown by
a broken line.

clear that decreasing the rotational speed lowers the conversion

of DNP because of increased film-diffusion resistance. At a flow
rate of 5  106 m3min1, the conversion of DNP calculated at
R = infinite is 97.0%, but decreases remarkably when the mixing
is insufficient; for example, 77.2% at R = 30 rev.min1 and 21.5%
at R = 5 rev.min1. This result indicates the importance of liquid
mixing with an immobilized TiO2 catalyst. On the other hand, the
result at R = 180 rev.min1 is almost equal to that at R = infinity,
indicating that if the reaction mixture is well mixed, it is possible
to operate the present reactor system under the condition of negligible film-diffusion resistance.
The photocatalytic reactor system shown in Fig. 1 has been
designed for outside installation. The UV light from the blacklight
blue fluorescent lamps and the flow of air supplied from the air
pumps are substitutes for the sunlight and wind, respectively.
When the sunlight is used as a UV light source, the UV intensity
changes frequently, so that the rate constant must be expressed
as a function of the UV intensity. The UV intensity of the sunlight
in the summer in Fukuoka is about two times larger than that of
two 4 W blacklight blue fluorescent lamps placed on each reactor.
Consequently, it is expected that the natural light source completes
the DNP decomposition in a shorter time. The performance of the
reactor system in this environment will be discussed in a subsequent study.
4.4. Effect of the number of reactors on the conversion of DNP
The decomposition of DNP in reactor systems consisting of
between one and five CSTRs in series, with a total reactor volume
of 4.5  104 m3, was simulated to investigate how the number
of reactors affects the DNP concentration in the end reactor. The
rotational speed was set at 180 rev.min1. As is evident from
Fig. 7, when the number of reactors is increased from one to two,
the conversion of DNP increases remarkably. A further increase
in the number of reactors gradually increases the conversion and
the reactor performance approaches that of the plug-flow reactor
without diffusional resistance. When the number of reactors is five,
the conversions in both the reactor systems are roughly the same.
The best number of reactors must be determined according to
the easiness of reactor control, which depends on the reactor configuration, initial concentration of reactant, final conversion, etc.
Motegh et al. [38] theoretically investigated the performance of
rectangular slurry bubble column reactors connected in series for

600

A. Miyawaki et al. / Chemical Engineering Journal 286 (2016) 594601

percentage change in a parameter /j (C0, k, KH, Kip, v, V, R, and a)

[39], given as

SC i ; /j

Fig. 7. Simulation of the relationship between the DNP concentration in the fifth
reactor and the flow rate, for the decomposition of DNP using one to five CSTRs with
a total liquid volume of 4.5  104 m3.

degradation of cyanide in water and estimated that the use of eight

column reactors is the best when one aims at a 99% conversion.
Although additional reactors give further slight increases in the
conversion in our reactor system, this would not be effective
because of making it difficult to control the reactor system.
4.5. Sensitivity analysis
Fig. 8 shows normalized sensitivities, i.e., the ratio of the
percentage change in a DNP concentration in each reactor to the

@C i
@/j

!

/j
C i

i 1; . . . ; 5

20

where the symbol * indicates that result is evaluated at a steady

state. A positive values for S(Ci, /j) indicates that when one of the
parameter values is increased, the DNP concentration becomes larger at a new steady state. On the other hand, a negative value indicates a response in the opposite direction, meaning that when the
sensitivity has a negative value, the unconverted fraction of DNP
decreases and the reactor performance increases.
When Ef = 1, i.e., there is no film-diffusion resistance, an
increase in C0 or t increases the unconverted fraction of DNP, while
an increase in k, KH, Kip, and V decreases it. The change in the
unconverted fraction is large, especially when C0, k, t, and V
change.
At R = 180 rev.min1, i.e., in the presence of slight diffusional
resistance, an increase in C0 or t increases the unconverted fraction, whereas an increase in k, KH, Kip, V, R, or a decreases it. The
unconverted fraction is affected, especially by changes in C0, k, t,
and V. The result is similar to that for Ef = 1. A change in R or a provides a small sensitivity, meaning that such a change does not
affect the unconverted fraction greatly, even when the filmdiffusion resistance is removed by liquid mixing. This is because
the diffusional resistance is small.
At R = 30 and 60 rev.min1, i.e., in the presence of larger
diffusional resistance, a change in C0, k, t, or V does not affect the
unconverted fraction greatly. This is because increases in these

Fig. 8. Normalized sensitivities of a DNP concentration in each CSTR with respective to parameters at t = 6  106 m3min1; (a) without film-diffusion resistance, (b) at
R = 180 rev.min1, (c) at R = 60 rev.min1, and (d) at R = 30 rev.min1.

A. Miyawaki et al. / Chemical Engineering Journal 286 (2016) 594601

parameters are not reflected directly on the unconverted fraction

in the presence of film-diffusion resistance. In contract, R and a
have larger effects on the unconverted fraction because liquid
mixing decreases the film-diffusion resistance, which is directly
associated with a change in the unconverted fraction.
Thus, it is clear that the liquid mixing is effective in the immobilized TiO2 system when the system has a large film-diffusion
resistance.
5. Conclusions
The rate of the photocatalytic decomposition of organic compounds by immobilized TiO2 is markedly reduced by large filmdiffusion resistance. To address this problem, the present study
investigated the performance of a system of CSTRs in series. A CSTR
that is easy to remove this resistance was selected as a reactor unit
and the number of CSTRs was increased to compensate for the
resulting decrease in the rate of reaction. Consequently, the
following results are obtained. First, sufficient liquid mixing
reduces the film-diffusion resistance, clearly increasing the rate
of the photocatalytic decomposition. Second, with a constant total
reactor volume of a system divided equally between a varying
number of reactors in series, the reactor performance is enhanced
remarkably for an increase from one to two reactors, and by smaller increments with additional reactors. Third, sufficient mixing of
a reaction solution in a system of five CSTRs in series clearly
increases the rate of photocatalytic decomposition. Fourth, the performance of this system approaches that of a plug-flow reactor
without diffusional resistance. Thus, it is concluded that the multiple CSTR system is appropriate for photocatalytic reactions. This
reactor system is one of the very few examples indicating the
validity of a theoretical prediction that an increase in the number
of reactors is effective to enhance the performance of the multiple
CSTR system.
References
[1] O. Levenspiel, Chemical Reaction Engineering, John Wiley & Sons, New York,
1972.
[2] F. Froment, K.B. Bischotf, Chemical Reactor Analysis and Design, Wiley, New
York, 1990.
[3] J.M. Smith, Chemical Engineering Kinetics, McGraw-Hill, New York, 1981.
[4] M.A. Behnajady, N. Modirshahla, N. Daneshvar, M. Rabbani, Photocatalytic
degradation of an azo dye in a tubular continuous-flow photoreactor with
immobilized TiO2 on glass plates, Chem. Eng. J. 127 (2007) 167176.
[5] Z. Cheng, X. Quan, J. Xiang, Y. Huang, Y. Xu, Photocatalytic degradation of
bisphenol A using an integrated system of a new gas-liquid-solid circulating
fluidized bed reactor and micrometer Gd-doped TiO2 particles, J. Environ. Sci.
24 (2012) 13171326.
[6] G.D. Maria, 5th Symposium on continuous flow reactor technology for
industrial applications, Chim. Oggi-Chem. Today 31 (2013) 3341.
[7] M.R. Nanda, Z. Yuan, W. Qin, H.S. Ghaziaskar, M.-A. Poirier, C. Xu, Catalytic
conversion of glycerol to oxygenated fuel additive in a continuous flow
reactor: Process optimization, Fuel 128 (2014) 113119.
[8] K. Kusunoki, Mixing characteristics in cotinuous-flow reactors and a
mathematical model for CSTR, Kyushu University, 1965. pp. 1114.
[9] S. Wang, F. Shiraishi, K. Nakano, A synergistic effect of photocatalysis and
ozonation on decomposition of formic acid in an aqueous solution, Chem. Eng.
J. 3998 (2002) 111.
[10] S. Wang, F. Shiraishi, K. Nakano, Decomposition of formic acid in a
photocatalytic reactor with a parallel array of four light sources, J. Chem.
Tech. Biotechnol. 77 (2002) 805810.
[11] F. Shiraishi, M. Ueno, R. Chand, Y. Shibata, H.N. Luitel, Effect of silanization of
titanium dioxide on photocatalytic decomposition of 2,4-dinitrophenol under
irradiation with artificial UV light and sunlight, J. Chem. Tech. Biotechnol. 89
(2014) 8187.
[12] H. Wang, H.L. Wang, W.T. Jiang, Z.Q. Li, Photocatayltic degradation of 2,4dinitrophenol (DNP) by multi-walled carbon nanotubes (MWCNTs)/TiO2
composite in aqueous solution under solar irradiation, Water Res. 43 (2009)
204210.

601

[13] M.N. Chong, B. Jin, Photocatalytic treatment of high concentration

carbamazepine in synthetic hospital wastewater, J. Hazard. Mater. 199200
(2012) 135142.
[14] T.E. Agustina, D.M. Ang, V.K. Pareek, Treatment of winery wastewater using a
photocatalytic/photolytic reactor, Chem. Eng. J. 135 (2008) 151156.
[15] M. Pelaez, P. Falaras, A.G. Kontos, A.A. de la Cruz, K. Oshea, P.A.M. Dunlop, J.A.
Byrne, D.D. Dionysiou, A comparative study on the removal of
cylondrospermopsin and mycrocystins from water with NF-TiO2-P25
composite films with visible and UV-vis light photocatalytic activity, Appl.
Catal. B: Environ. 121122 (2012) 3039.
[16] M.Y. Ghaly, T.S. Jamil, I.E. El-Seesy, E.R. Souaya, R.A. Nasr, Treatment of highly
polluted paper mill wastewater by solar photocatalytic oxidation with
synthesized nano TiO2, Chem. Eng. J. 168 (2011) 446454.
[17] A. Verma, N.T. Prakash, A.P. Toor, An efficient TiO2 coated immobilized system
for the degradation studies of herbicide isoproturon: durability studies,
Chemosphere 109 (2014) 713.
[18] N.M. Garca, S. Surez, B. Snchez, J.M. Coronado, S. Malato, M.I. Maldonado,
Photocatalytic degradation of emerging contaminants in municipal
wastewater treatment plant effluents using immobilized TiO2 in a solar pilot
plant, Appl. Catal. B: Environ. 103 (2011) 294301.
[19] M.F.J. Dijkstra, H.J. Panneman, J.G.M. Winkelman, J.J. Kelly, A.A.C.M.
Beenackers, Modeling the photocatalytic degradation of formic acid
in a reactor with immobilized catalyst, Chem. Eng. Sci. 57 (2002) 4895
4907.
[20] F. Shiraishi, K. Toyoda, H. Miyakawa, Decomposition of gaseous formaldehyde
in a photocatalytic reactor with a parallel array of light sources: 2. reactor
performance, Chem. Eng. J. 114 (2005) 145151.
[21] F. Shiraishi, T. Nakasako, Z. Hua, Formation of hydrogen peroxide in
photocatalytic reactions, J. Phys. Chem. A 107 (2003) 1107211081.
[22] F. Shiraishi, C. Kawanishi, An effect of diffusional film on formation of
hydrogen peroxide in photocatalytic reactions, J. Phys. Chem. A 108 (2004)
1049110496.
[23] F. Shiraishi, M. Nagano, S. Wang, Characterization of a photocatalytic reaction
in a continuous-flow recirculation reactor system, J. Chem. Tech. Biotechnol.
81 (2006) 10391048.
[24] G.A. Parks, The isoelectric points of solid oxides, solid hydroxides, and aqueous
hydroxo complex systems, Chem. Rev. 65 (1965) 177198.
[25] R. Chand, F. Shiraishi, Reaction mechanism of photocatalytic decomposition of
2,4-dinitrophenol in aqueous suspension of TiO2 fine particles, Chem. Eng. J.
233 (2013) 369376.
[26] M.L. Shuler, R. Aris, H.M. Tsuchiya, Diffusive and electrostatic effects with
insolubilized enzymes, J. Theor. Biol. 35 (1972) 6776.
[27] X. Cao, F. Shiraishi, A mechanism of photocatalytic and adsorptive treatment of
2,4-dinitrophenol on a porous thin film of TiO2 covering granular activated
carbon particles, Chem. Eng. J. 160 (2010) 651659.
[28] F. Shiraishi, Computational methods for analysis of immobilized enzyme
reactions: from reaction kinetics to reactor design methods, Sangyo Tosho,
Tokyo, 1997.
[29] F. Shiraishi, A. Miyawaki, R. Chand, A mechanism of the photocatalytic
decomposition of 2,4-dinitrophenol on TiO2 immobilized on a glass surface,
Chem. Eng. J. 262 (2015) 831838.
[30] R. Lortie, D. Thomas, Heterogeneous one-dimensional model for fixed bed
enzyme reactors, Biotechnol. Bioeng. 28 (1986) 12561260.
[31] F. Shiraishi, S. Furuta, T. Ishimatsu, J. Akhter, A simple and highly accurate
numerical differentiation method for sensitivity analysis of large-scale
metabolic reaction systems, Math. Biosci. 208 (2007) 590606.
[32] H. Miyakawa, F. Shiraishi, A film diffusional effect on the apparent kinetic
parameters in packed-bed immobilized enzyme reactors, J. Chem. Tech.
Biotechnol. 69 (1997) 456462.
[33] F. Shiraishi, H. Miyakawa, T. Hasegawa, S. Kasai, A computational method for
determination of the mass-transfer coefficient in packed-bed immobilized
enzyme reactors, J. Chem. Tech. Biotechnol. 66 (1996) 405413.
[34] C.-Y. Hsiao, C.-L. Lee, D.F. Ollis, Heterogeneous photocatalysis: Degradation of
dilute solutions of dichloromethane (CH2Cl2), chloroform (CHCl3), and carbon
tetrachloride (CCl4) with illuminated TiO2 photocatalyst, J. Catal. 82 (1983)
418423.
[35] H. Gerischer, A. Heller, A comparative study on the removal of
cylondrospermopsin and mycrocystins from water with NF-TiO2-P25
composite films with visible and UV-vis light photocatalytic activity, J. Phys.
Chem. 95 (1991) 52615267.
[36] M.N. Chong, Y.J. Cho, P.E. Poh, B. Jin, Evaluation of titanium dioxide
photocatalytic technology for the treatment of reactive Black 5 dye in
synthetic and real grey water effluents, J. Clean. Prod. 89 (2015) 196202.
[37] S. Zhou, A.K. Ray, Kinetic studies for photocatalytic degradation of Eosin B on a
thin film of titanium dioxide, Ind. Eng. Chem. Res. 42 (2003) 60206033.
[38] M. Motegh, J.R. Ommen, P.W. Appe, M.T. Kreutzer, Scale-up study of a
multiphase photocatalytic reactor: degradation of cyanide in water over TiO2,
Environ. Sci. Technol. 48 (2014) 15741581.
[39] A. Varma, M. Morbidelli, H. Wu, Parametric Sensitivity in Chemical Systems,
Cambridge Univ. Press, Cambridge, 1999.