Water Research 37 (2003) 1125–1135

Solar photocatalytic thin film cascade reactor for treatment of

benzoic acid containing wastewater
Alex H.C. Chan, Chak K. Chan, John P. Barford, John F. Porter*
Department of Chemical Engineering, The Hong Kong University of Science and Technology (HKUST), Clear Water Bay, Kowloon,
Hong Kong SAR, People’s Republic of China
Received 26 February 2002; received in revised form 27 July 2002; accepted 25 September 2002

Abstract

A solar photocatalytic cascade reactor was constructed to study the photocatalytic oxidation of benzoic acid in water
under various experimental and weather conditions at HKUST. Nine stainless steel plates coated with TiO2 catalyst
were arranged in a cascade configuration in the reactor. Photolytic degradation and adsorption were confirmed to be
insignificant total organic carbon (TOC) removal mechanisms. A turbulent flow pattern and, hence, improved mixing in
the liquid film were achieved due to the unique cascade design of the reactor. The photoinduced consumption of oxygen
during reactions was demonstrated in a sample experiment. The proposed rate equations provided good fits to 90 data
points from 17 experiments. The regression results showed that the TOC removal rates averaged over 30 min intervals
did not illustrate significant dependence on TOC0 and that Imean was more important in affecting the photocatalytic
process within the ranges of the data examined. The percentage removal of TOC in 7 l of 100 mg/l (or 100 ppm) benzoic
acid solutions increased from 30% to 83% by adding 10 ml of hydrogen peroxide solution (30 wt%). Hydrogen
peroxide was also shown to enhance the efficiency of the degradation process at elevated temperatures. Ortho-, meta-
and para-hydroxybenzoic acids were identified by HPLC analysis as the intermediates of benzoic acid during reactions
without the addition of hydrogen peroxide solutions.
r 2002 Elsevier Science Ltd. All rights reserved.

Keywords: Photocatalysis; Cascade; Solar; Titanium dioxide (TiO2); Peroxidation

1. Introduction inexpensive, non-toxic and commercially available. The

Photocatalysis by titanium dioxide (TiO2) has been
demonstrated to be an inexpensive and effective method
for treating a variety of organic pollutants in water.
Matthews [1–3], Ollis et al. [4,5] and Mills [6] summar-
ized different types of organic pollutants treated by
TiO2. Examples include alkanes, alkenes, aromatics,
aliphatic alcohols, aliphatic and aromatic carboxylic
acids, phenol, polymers, surfactants, pesticides, herbi-
cides, dyes, etc. Apart from being indiscriminate and
capable of degrading many different types of organic
pollutants, the TiO2 catalyst employed is relatively
*Corresponding author. Tel.: +852-2358-7132; fax: +852-
2358-0054.
E-mail address: kejep@ust.hk (J.F. Porter).
band gap value of anatase TiO2 is approximately 3.2 eV
enabling UV light of a wavelength less than 400 nm to
activate the catalyst, resulting in the promotion of an
electron to the conduction band and the generation of a
hole in the valence band. The redox potential of the
valence band hole of anatase TiO2 is sufficiently positive
to generate hydroxyl radicals by reacting with adsorbed
water molecules or hydroxyl groups, while the redox
potential of the conduction band electron is sufficiently
negative to reduce adsorbed O2 to superoxide [6].
Hydroxyl radical oxidation is considered to be the
primary mechanism for the destruction of organic
pollutants [7].
The UV radiation required for photocatalytic pro-
cesses may come from an artificial source or the sun. The
artificial generation of UV radiation contributes to a

0043-1354/03/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 4 3 - 1 3 5 4 ( 0 2 ) 0 0 4 6 5 - 7
1126 A.H.C. Chan et al. / Water Research 37 (2003) 1125–1135
large portion of the operating, capital and maintenance
costs of a photocatalytic reaction system because of the
utility consumption and periodic replacement of the UV
lamps. There is, therefore, a significant economic
incentive to develop solar-powered photocatalytic reac-
tors. In addition, the environmental impact induced by
the use of solar energy is minimal and this renders the
photocatalytic process environmentally attractive. The
application of solar-powered photocatalytic reactors to
treat water contaminated with organic pollutants holds
promise for regions receiving strong sunlight throughout
the year, such as southern China and Hong Kong.
Matthews [1] reported the use of suspended TiO2 in a
solar illuminated tubular reactor equipped with a
parabolic trough concentrator to degrade a variety of
organic materials. The need for a separation system to
recover the catalysts leads to the conclusion that scaling
up such a design may be difficult. Other than the issue of
catalyst recovery, the light-concentrating reactor re-
quires reflectors, which are expensive and only able to
concentrate the direct component of the solar irradiation
for reaction. In contrast, non-light concentrating reac-
tors utilizing both the direct and diffuse components
have greater potential for process development. Wyness
et al. [8], Freudenhammer et al. [9], Goslich et al. [10]
and Sclafani et al. [11] investigated solar-powered fixed
bed reactors, in which organic-containing solutions flow
across single flat plates coated with TiO2 catalysts. The
reactions taking place in these reactors may, however, be
constrained by mass transfer limitations due to laminar
flow conditions [10,11]. Chan et al. [12] developed the
first bench-top cascade reactor and evaluated its
performance based on the removal efficiency of the
model pollutant, benzoic acid employing artificial UV
lamps as the light source. It was demonstrated that the
backflow and turbulence resulting from the novel
cascade reactor design can reduce mass transfer limita-
tions and the ‘‘waterfall’’ effect can also enhance oxygen
transfer to the solution. In this study, a pilot scale
cascade reactor was developed and tests were carried out
to specifically demonstrate the applicability of the
cascade reactor to treat benzoic acid solutions under
solar illumination. The two main differences between the
bench-top and pilot scale cascade reactors are the
number of plates (three versus nine) in the reactor and
the use of artificial UV versus solar radiation as the
illumination source.
2. Experimental method
Nine 316 stainless steel flat plates were arranged in a
cascade configuration in the solar-powered photocata-
lytic reactor. The plates were coated with TiO2 catalysts
(Degussa P-25) using an electrophoretic method pre-
viously described [12]. After drying under ambient
conditions, the plates were heated in air at 473 K for
60 min in a muffle furnace to enhance the stability of the
coatings against mechanical attrition. Representative
samples were scraped off the plates and weighed using a
digital balance (Ohaus AS 200) to estimate the catalyst
loading. The number of plates was increased from three
for the original bench-top reactor [12] to nine for this
pilot scale reactor in order to increase the total area for
reaction. The nine TiO2-coated plates were positioned
2.5 cm vertically apart from each other and inclined at
an angle of 5o to the horizontal. The area of each plate
exposed to solar light was 17.5 cm
28 cm. A solution of
a known concentration of benzoic acid (Riedel-de ha.en,
99.5% purity) in de-ionized water was introduced by an
overflow to the liquid reservoir and flowed down across
the cascade as a thin film. A centrifugal pump was used
to circulate the solution back to the liquid reservoir. A
temperature bath (Neslab Instruments RTE-211) and a
flowmeter were used to control the temperature and
monitor the flowrate of the solution, respectively. The
liquid reservoir was covered with a sheet of aluminum
foil to prevent the solution from being heated by the
sunlight directly. Seven litres of solution were treated in
all the experiments. The initial concentration of benzoic
acid solution varied from 25 to 100 mg/l. A schematic
diagram of the pilot scale cascade photoreactor is shown
in Fig. 1(a). Fig. 1(b) shows the experimental set-up as
located on the rooftop of the laboratories at HKUST.
The entire set-up was fixed in position and the TiO2-
coated plates faced west in all the experiments.
A UV sensor connected to a radiometer (Vilber
Lourmat VLX-3W) was located at the edge of the fifth
plate for measuring the cumulative solar UV energy at
the wavelength of 365 nm at 30-min intervals during the
experiments. The UV intensities at 365 nm were aver-
aged over a 30 min time period and corrected to those
between 320 and 400 nm based on calibrations obtained
from another radiometer (Spectroline DM-365XA) that
was equipped with a UV sensor measuring the UV
intensity between 320 and 400 nm. Data on the UV-A
range were employed for calculations in the following
sections because more than 98% of the solar UV
radiation is in the UV-A range. Measurements of the
UV energy at 254 nm on a sunny day were also made
and the intensity obtained was determined to be
negligible. The dissolved oxygen (DO) level and the
temperature of the solution in the liquid reservoir were
recorded by a DO meter (Cole-Parmer DO 100), while
the temperature of the solution at the photoreactor
outlet was measured by a type K thermocouple (Digi-
Sense) at 2-min intervals in the first 20 min of the
reactions and every 20 min afterwards. The pH value of
the solution in the liquid reservoir was recorded by a pH
meter (Orion 420A). Since the pH of benzoic acid
solutions only varied between 4.2 and 3.9 over the
concentration range examined, no adjustments were
 A.H.C. Chan et al. / Water Research 37 (2003) 1125–1135 1127

Fig. 1. (a) Schematic diagram of the pilot scale cascade photoreactor. A: TiO2-coated plates; B: tank with drain valve; C: control;
D: centrifugal pump; E: cooling coil in water bath; F: flowmeter; G: liquid reservoir. (b) The pilot scale cascade photoreactor.

made to the initial pH of the solution for any of the
experiments.
Parameters, including reaction temperature, initial
concentration of benzoic acid solution, addition of
hydrogen peroxide solution and solar light intensity
were investigated. Before each experiment, the reactor
was allowed to run in the dark for 30 min by covering
the entire experimental setup. Equilibrium adsorption of
benzoic acid by the supported catalysts was achieved
within this period of time. Liquid samples of 25 ml were
withdrawn from the reactor effluent at 30-min intervals
for a period of 3 h. The coated plates were washed with
flowing de-ionized water under sunlight after each
experiment for about 10 to 15 min. The photoreactor
performance was evaluated based on the rate of removal
of total organic carbon (TOC). A Shimadzu TOC-5000
analyzer was used to analyze the TOC of the liquid
samples. Some of the intermediates from the degrada-
tion of benzoic acid were identified using an HP series II
1090 high performance liquid chromatograph (HPLC),
1128 A.H.C. Chan et al. / Water Research 37 (2003) 1125–1135
equipped with a built-in diode array detector (DAD).
The column used was a Microsorb-MV 100 A ( C-18
capillary column operating isothermally at 401C. The
eluent was a mixture of 0.05 M potassium phosphate
solution acidified to pH 3.2 and methanol (60:40) at a
flowrate of 0.7 ml/min.
3. Results and discussion
3.1. Liquid flow pattern
One disadvantage of a thin film fixed bed reactor is
the consequence of mass transfer limitations resulting
from laminar flow conditions [10,11]. The turbulent flow
pattern of the liquid film resulting from the cascade
design can be visually observed in Fig. 2. The solution
flowrate in all the experiments in this study was chosen
as 3 l/min because previous studies, employing the
bench-top three-plate reactor illuminated with artificial
UV lamps, showed that the TOC removal did not
depend on solution flowrate from 2 to 5 l/min in a 3 h
reaction [12]. The mean film thickness at the flowrate of
3 l/min was estimated to be approximately 1.1 mm. Bird
et al. [13] presented the equation for calculating the
Reynolds numbers (Re) for a falling film over a single
Fig. 2. Turbulent flow pattern of liquid film observed at the flowrate of 3 l/min.
inclined surface and stated the critical Re as 1200. For
flowrates from 2 to 5 l/min, the estimated Re would
range from 535 to 1338. However, an equation
corresponding to the cascade design is not available
theoretically and the relevant calculations, which can be
useful in demonstrating the flow pattern of the liquid
film in the cascade configuration, were thus not carried
out in this study. While only the flow pattern of the
solution at the flowrate of 3 l/min was shown in Fig. 2,
similar turbulent flow patterns could also be observed
for operations at lower flowrates, for example, 1 l/min.
Better mixing resulting from the cascade design should
reduce mass transfer limitations of organics from the
bulk of the liquid to the catalyst surface and enhance
oxygen transfer.
3.2. Photolytic degradation and adsorption
The photolytic degradation of a benzoic acid solution
of 25 mg/l was tested on a sunny day similar to other
experiments, except that the plates installed into the
reactor were uncoated. The intensity of UV at 254 nm
was measured to be negligible, while the average UV-A
intensity during 3 h was found to be 1.12 mW/cm2. Fig. 3
shows that the photolytic degradation of the 25 mg/l
solution was insignificant. Therefore, UV-A is not the
 A.H.C. Chan et al. / Water Research 37 (2003) 1125–1135
90.00
Photolysis
80.00 Adsorption, 50ppmBA
Adsorption, 100ppmBA
70.00
60.00
TOC (ppm) 50.00
40.00
30.00
20.00
10.00
0.00
0 30
Fig. 3. TOC trends in the processes of photolytic degradation and adsorption.
primary driving force for degrading benzoic acid in the
absence of TiO2 catalysts. Since the photolytic degrada-
tion of a solution of 25 mg/l was insignificant, the
degradation of benzoic acid of a higher initial concen-
tration by the photolytic process is reasonably assumed
also to be insignificant.
Adsorption studies were also conducted at initial
benzoic acid concentrations of 25, 50, 75 and 100 mg/l in
the dark by covering the entire experimental setup. The
UV-A intensity measured was negligible. The TOC
concentrations in these adsorption experiments are
also shown in Fig. 3. The adsorption process mainly
took place within the first 30 min of the 3 h reactions.
Because the mass of catalyst on each plate was small
(approximately 440 mg), the amount of benzoic acid
adsorbed was also small. As a result, photolytic
degradation and adsorption were not considered to be
significant TOC removal mechanisms in the subsequent
experiments.
3.3. Effect of light intensity and initial TOC
concentration
Two simple empirical rate equations were used to
represent the average rate of change of TOC, d[TOC]/
dt (mg/l min) as a function of the initial TOC
concentration, TOC0 (mg/l) and the mean UV-A
intensity between 320 and 400 nm, Imean (mW/cm2).
The first rate equation involving the Langmuir-type
dependence on Imean was based on one presented earlier
by Crittenden et al. [14], whereas the second rate
equation involving the power law dependence was
derived from the work of Egerton and King [15], Ollis
[16] and Chang et al. [17]. It should be noted that the
rate expressions used here are empirical—the use of a
1129
Adsorption, 25ppmBA
Adsorption, 75ppmBA
60 90 120 150 180
Time (min)
Langmuir-type dependence on the mean UV-A intensity
does not necessarily imply mechanistic significance.


d½TOC bImean
 ¼ k0 ½TOC0 n ; ð1Þ
dt 1 þ bImean
d½TOC
 ¼ k0 ½TOC0 n Imean
m
: ð2Þ
dt
In each case the apparent rate constant is denoted as
k0 ; n represents the nth power of TOC0 ; m represents the
mth power of Imean and parameter b is determined in the
data fitting process.
Values of the parameters in the proposed equations
were obtained from the regression of 90 data points
from the 17 experiments in this study. As described
above, the average UV intensities at 365 nm were
calculated from the differences in the cumulative solar
energies over 30-min intervals. Corrections of the UV
intensities at 365 nm to those between 320 and 400 nm,
Imean ; were made. Liquid samples were taken at 30-min
intervals. The average rate of change of TOC was
calculated from the difference in the TOC values in
30 min. The initial TOC concentration was taken as the
instantaneous TOC value at the onset of each 30-min
interval. If the final TOC value of the solution was zero,
the corresponding value of the TOC removal rate was
ignored because the exact time for the TOC value to
reduce to zero was not known. Sample calculations are
presented in Table 1.
The experiments concerned were carried out at a
lower temperature (181C) to minimize water evaporation
and changes in TOC concentrations were corrected for
water evaporation. The dissolved oxygen concentration,
DO was not included in the rate equations because it did
not change significantly during these experiments. The
1130 A.H.C. Chan et al. / Water Research 37 (2003) 1125–1135

Table 1
Sample calculations to study the dependence of d[TOC]/dt on TOC0 and Imean

Data Index Time (min) TOC (mg/l) TOC0 (mg/l) Imean (mW/cm2) d[TOC]/dt (mg/l min)

1 0 17.21 17.21
0.6584 0.0573
2 30 15.49 15.49
0.8571 0.0593
3 60 13.71 13.71
0.3784 0.0763
4 90 11.42 11.42
0.3421 0.0614
5 120 9.58 9.58
0.4245 0.0391
6 150 8.41 8.41
0.0762 0.0636
7 180 6.50

Table 2
Regression results for the proposed rate law equations
P
k0 ½TOC0 n ðbImean =ð1 þ bImean ÞÞ Case k0 n B |Error|

I 0.125 0.036 1.988 1.37286

P
k0 ½TOC0 n Imean
m
Case k0 n m |Error|

II 0.081 0.030 0.400 1.48442

III 0.079 0.034 0.500 1.53655
P
k0 ðbImean =ð1 þ bImean ÞÞ Case k0 b |Error|

IV 0.135 2.296 1.37682

P
k0 Imean
m
Case k0 m |Error|

V 0.090 0.399 1.49632

VI 0.088 0.500 1.55058

equations were derived for data for I mean ranging from
0.06 to 2.52 mW/cm2 and TOC0 ranging from 5.51 to
72.39 mg/l.
Table 2 presents the regression results for the
proposed rate equations. The data indicate that the
values of the exponent, n; in cases I, II and III were small
(m was set to 0.5 in case III). Based on the results, the
dependence of TOC removal rates on TOC0 should,
therefore, be weak. Just to confirm that the initial TOC
concentration had little effect on the TOC removal rate,
this parameter was intentionally removed from the rate
equations and the error of the data fitting was re-
checked. The values of the sum of absolute errors in
cases IV, V and VI when the exponent n was set to zero
were only slightly larger than the corresponding values
in cases I, II and III. This implies that, while the
inclusion of the extra parameter, TOC0, provided a
better fit to the experimental data, the improvement was
not significant. Thus, it is concluded that TOC0 was a
less important factor than Imean in affecting the
degradation process over the ranges of the data
examined. An earlier analysis by Chan et al. [12] used
a simplified linear form of the Langmuir–Hinshelwood
equation presented by Matthews [1–3] to model the
removal of benzoic acid in the bench-scale reactor.
While the results of this study do not directly contradict
our previous work, since the two models are based on
different concentration terms, the relatively large dataset
and non-linear model used in this analysis provide
confidence in the relative importance of TOC0 and Imean
seen here.
Ollis et al. [5,16] concluded that the photocatalytic
reaction rate by suspended TiO2 is first-order dependent
on light intensity at intensities appreciably below one
sun equivalent, but is half order dependent at signifi-
cantly higher levels of illumination. As suggested by
Egerton and King [15] and Pichat [18], a linear
dependence of the rate at lower radiant flux implies
 A.H.C. Chan et al. / Water Research 37 (2003) 1125–1135
that the majority of photoproduced charges do not
recombine, while the square-root dependence at higher
radiant flux arises from the dominance of electron–hole
recombination over other productive steps, such as
O2,ads+e CB-O2,ads and h +OHads-OHads. In con-
 + 
nection with this, it was stated that there should exist a
radiant flux, above which the reaction rate becomes
photo-independent and does not further increase. Based
on the regression results in Table 2, rate equation (1)
provides a more accurate model for the removal rate of
TOC. It should also be noted that the implication of rate
equation (1), which involves a Langmuir-type depen-
dence on Imean ; is qualitatively consistent with both Ollis’
conclusions as well as the observations of Egerton and
King, and Pichat.
The bench-top reactor previously used [12] differs
from the pilot scale reactor in this study only in terms of
the number of plates and the UV source used. Hence,
the TOC removal rates obtained for benzoic acid with
the bench-top reactor should be able to be scaled
linearly based on the number of plates for comparison
with those obtained using the pilot scale reactor under
similar illumination levels — all other conditions (the
dimensions, inclination and vertical spacing of the
plates) remaining identical. Chan [19] compared
the TOC removal rates obtained with the two reactors
as a function of illumination intensity, which has just
been illustrated to be an important factor in affecting the
degradation process. The rates obtained with the pilot
scale reactor were clearly shown to match with the rates
achieved with the bench-top reactor when the bench-top
rates were multiplied by 9/3—i.e. linearly scaled
Fig. 4. Dynamic time profiles in a sample experiment.
1131
according to the number of plates used—demonstrating
the ease of scale up.
3.4. Sample experimental results
Previous studies employing a bench-top three-plate
reactor showed that the degree of TOC removal was the
greatest at 181C in the temperature range of 18–281C
[12]. In this sample experiment, the solution temperature
was chosen as 181C. The initial benzoic acid concentra-
tion was 25 mg/l, corresponding to a TOC level of
approximately 17 mg/l. This particular reaction was
carried out from 11:30 am to 2:30 pm on a sunny day.
As illustrated in Fig. 4, the DO level dropped quickly
at the onset of the photocatalytic process, became quite
steady after that and started to increase to the original
value at approximately the second hour of the reaction,
demonstrating the photoinduced consumption of oxy-
gen. The DO level decreased at the onset of the reaction
because oxygen was consumed, while the DO level
increased at the second hour because all the organics
were degraded and oxygen consumption stopped. On
the other hand, the fact that the DO level remained
steady during the reaction instead of decreasing
continuously indicates the uptake of oxygen into the
solution. The large liquid–air interface over the plates
and air introduced by the ‘‘waterfall’’ effect of the
cascade design resulted in oxygen steadily being
replenished from the atmosphere.
The solution temperature was controlled adjusting
the temperature of cooling water in the water bath. The
slight jump in the solution temperature, +0.51C at the
1132 A.H.C. Chan et al. / Water Research 37 (2003) 1125–1135
eighth minute was actually caused by solar heating.
However, the drop in the DO level in the first few
minutes of the reaction contributed by this slight
increase in solution temperature is shown to be
insignificant. Although the solubility of oxygen in the
benzoic acid solution might decrease when other organic
compounds were formed as intermediates during the
reaction, this was unlikely the reason for the decline in
the DO level at the onset of the reaction because
their concentrations were low, as demonstrated shortly.
The increase in the solution pH during the reaction
also implies that the degradation of the organics took
place.
The average DO level in the cascade configuration of
the bench-top three-plate reactor was previously shown
to be slightly greater than that in the single flat plate
configuration [12]. Comparison between the DO profiles
in a single flat plate reactor and the pilot scale nine-plate
reactor has not been made in this study because two
technical problems have to be solved. First, it is difficult
to construct and coat a single flat plate of an area
identical to the sum of the nine plates in our laboratory.
Second, the experiments should be performed under
controlled illumination levels for comparison. The latter
can be more easily solved when artificial UV lamps
instead of the sunlight are used as the light source.
Nevertheless, a higher DO level is expected in an
extended cascade reactor with many steps than an
extended single flat plate reactor.
3.5. Reactor performance with the addition of hydrogen
peroxide
In cases where water containing high concentrations
of organic pollutants has to be treated, hydrogen
peroxide can be used to promote the degradation
process. Tests employing the bench-top three-plate
reactor showed that a dosage of 10 ml of H2O2 solution
(30 wt%) added at the onset of the reaction greatly
improved the degradation of 7 l of a 100 mg/l benzoic
acid solution [19]. The same dosage of H2O2 solution
was, therefore, chosen in this study. The initial benzoic
acid concentration and solution temperature in the two
experiments concerned were 100 mg/l and 181C accord-
ingly. The mean UV-A intensity in the 3-h experiment
with H2O2 was 1.80 mW/cm2, while that in the experi-
ment without H2O2 was 1.74 mW/cm2. Although the UV
intensity was slightly greater in the experiment with
H2O2, the percentage removal of TOC notably increased
from 30% to 83% with the addition of H2O2, as shown
in Fig. 5. This is because hydrogen peroxide can
decompose into oxygen and increase the dissolved
oxygen in the solution (Eq. (3)). The increase in the
dissolved oxygen of the solution is illustrated in Fig. 6.
In addition, hydrogen peroxide could also act as a
source of hydroxyl radicals to improve the removal of
TOC (Eq. (4)) [17].
H2 O2 -H2 O þ 1=2O2 ; ð3Þ
H2 O2 þe  -OHdþOH : ð4Þ
However, issues such as the optimum dosage and
method of addition of H2O2 (either continuously during
the reaction or in one lump sum at the reaction’s onset)
have to be further addressed. Finally, it should be noted
that the DO profile in the experiment without H2O2
illustrated in Fig. 6 did not return to the original DO
value at the end of the 3-h reaction. This is because the
initial benzoic acid concentration concerned was
100 mg/l and the organics were not completely oxidized,
hence oxygen was still being consumed at that moment.
3.6. Chemical intermediates formed during reactions
By comparing the retention time and the UV spectra
of the sample peaks with those of the standard
compounds in the HPLC analysis, three aromatic
organic compounds — ortho-, meta- and para-hydro-
xybenzoic acids — were identified as the intermediates in
the degradation of benzoic acid in reactions without
hydrogen peroxide solutions. Among them, meta-hydro-
xybenzoic acid was found to be more abundant than the
ortho- and para-isomers, as illustrated in Figs. 7 and 8.
This is in qualitative agreement with the chemistry in
that the carboxylic acid functional group is meta-
directing in electrophilic aromatic substitution reactions.
The product distribution can also be explained if meta-
hydroxybenzoic acid is more resistant to degradation by
TiO2 than the ortho- and para-isomers. Matthews
[1,2,20] had also identified these organic compounds as
the intermediates of benzoic acid during photocatalytic
oxidation by TiO2.
Figs. 7 and 8 also show that the total initial carbon
concentrations obtained by the TOC and HPLC
1.00
0.90
0.80
0.70
TOCt /TOCo
0.60
0.50
0.40
0.30
0.20 100ppmBA+10mlH2O2 (30wt-%)
100ppmBA
0.10
0.00
0 30 60 90 120 150 180
Time (min)
Fig. 5. Removal of TOC as a function of time in experiments
with and without addition of H2O2.
 A.H.C. Chan et al. / Water Research 37 (2003) 1125–1135 1133

12.0 40.0

35.0
10.0

Solution temperature ( C)
30.0

o
8.0

DO level (ppm)
25.0

6.0 20.0

15.0
4.0
DO level: 100ppmBA + 10ml H2O2(30wt-%) 10.0
DO level: 100ppmBA
2.0 Temperature: 100ppmBA + 10ml H2O2(30wt-%)
Temperature: 100ppmBA 5.0

0.0 0.0
0 20 40 60 80 100 120 140 160 180
Time (min)

Fig. 6. Solution DO and temperature time profiles in experiments with and without addition of H2O2.

20.0 2.0 70.0 7.0

18.0 BA-C 1.8
Sum-C 60.0 6.0
16.0 TOC 1.6
Carbon concentration (ppm)

Carbon concentration (ppm)

Carbon concentration (ppm)

Carbon concentration (ppm)

2-OHBA-C
14.0 3-OHBA-C 1.4 50.0 5.0
4-OHBA-C BA-C
12.0 1.2
40.0 Sum-C 4.0
10.0 1.0 TOC
2-OHBA-C
8.0 0.8 30.0 3-OHBA-C 3.0
4-OHBA-C
6.0 0.6 20.0 2.0
4.0 0.4
10.0 1.0
2.0 0.2

0.0 0.0 0.0 0.0

0 30 60 90 120 150 180 0 30 60 90 120 150 180
Time (min) Time (min)

Fig. 7. Carbon concentration profiles associated with benzoic Fig. 8. Carbon concentration profiles associated with benzoic
acid (BA), intermediates and TOC. Temperature=181C, initial acid (BA), intermediates and TOC. Temperature=181C, initial
BA conc.=25 mg/l. BA conc.=100 mg/l.

analyses were approximately the same in each case. the aromatic intermediates. Benzoic acid was thus
However, differences between them increased with time, deemed to be the major component contributing to the
with the TOC becoming increasingly larger than the sum TOC of the liquid samples, although the contribution of
of carbon associated with benzoic acid and the identified the unidentified aliphatic organic species to TOC
intermediates (from the HPLC analysis). This implies (difference between the trends of TOC and sum of
that more carbon existed as organic species other than carbon) varied in the two cases. Further analysis in the
benzoic acid and the aromatic intermediates as the degradation rates of the aliphatic organic species, once
reaction progressed. Matthews [3] proposed that an identified, should prove useful in explaining the trends.
aliphatic compound, mucondialdehyde (C6H6O2) is one Nevertheless, caution should be taken if one attempts to
of these unidentified organic species, accompanied by interpret the TOC removal rates presented in this study
those breakdown products resulting from it. Further as the removal rates of benzoic acid. Since the
investigation has to be carried out to isolate and identify unidentified aliphatic organic species were not expected
subsequent organic intermediates in the degradation to be the principal component of TOC and are not
reaction pathway. necessarily volatile, air contamination resulting from the
In addition, Figs. 7 and 8 show that the carbon escape of these species would not be significant if it
concentration associated with benzoic acid was one might happen. Indeed, both open reaction systems that
order of magnitude higher than those associated with have a large liquid–air interface as well as systems that
1134 A.H.C. Chan et al. / Water Research 37 (2003) 1125–1135

employ gas sparging for air or oxygen supply are prone 10.0 50.0
to vaporization of volatile species present in the liquid 9.0 45.0
phase. The proposed cascade reactor will likely suffer 8.0 40.0
from this problem and thus should not be used to deal

Solution temperature(°C)
7.0 35.0
with water polluted with volatile organic contaminants

DO level (ppm)
6.0 30.0
(e.g. trichloroethylene) nor other compounds which
5.0 25.0
form volatile, toxic intermediates on photocatalytic
4.0 20.0
degradation.
3.0 15.0
DO level: with temperature control
2.0 DO level: without temperature control 10.0
3.7. Reaction at higher ambient temperatures 1.0 Temperature: with temperature control 5.0
Temperature: without temperature control
0.0 0.0
Two sample experiments, in which the mean UV-A 0 30 60 90 120 150 180
intensities were very close to each other, were chosen to Time (min)
compare the reactor performances with and without Fig. 9. DO level and solution temperature profiles in two
temperature control. The temperature of the solutions in sample experiments.
the liquid reservoir was kept at 181C or allowed to
fluctuate with the ambient conditions. The initial
benzoic acid concentration in the two experiments was 2.5
100 mg/l. A total of 175 ml solution out of 7 l was taken
as liquid samples for TOC analysis in each case. Table 3
presents experimental data from the two sample experi- 2.0
UV-A intensity (mW/cm2)

ments averaged over a 3 h period. The details are shown

in Figs. 9 and 10. The experiment performed at the 1.5
higher temperature was found to exhibit a slightly lower
removal of TOC than the experiment performed at the
lower temperature. An initial inspection suggests that 1.0
this may be as a result of the decrease in the DO level of
the solution in the reaction lasting for 3 h. To account 0.5 With temperature control
for the loss of water by evaporation at an elevated Without temperature control
temperature, the results of TOC removal rates are
reported in the total amount of carbon removed per 0.0
minute (mg-C/min). 0 30 60 90 120 150 180
Cooling of the solution, however, requires extra Time (min)
utility. Chan [19] demonstrated that the addition of Fig. 10. UV-A intensity profiles in two sample experiments.
hydrogen peroxide is not only effective at treating highly
polluted solutions but also capable of improving the
reactor performance at elevated temperatures. In parti- should favor its choice over cooling of the solution to
cular, 27.3 percent of TOC of 7 l of a 100 mg/l benzoic maintain reactor performance at elevated temperatures.
acid solution was removed at 306 K with H2O2 versus More detailed cost information for a number of
8.2 percent at 291 K without H2O2, other conditions advanced oxidation processes was published by Ollis
remaining constant [19]. Chan [19] also suggested that [21] and Li Puma and Yue [22].
the relative enhancement in the TOC degradation rate
and the lower operating costs incurred using hydrogen
peroxide solution as a supplementary oxidizing agent 4. Conclusions

A pilot scale solar cascade reactor was set up to

Table 3 evaluate the feasibility of utilizing the UV portion of
Experimental data in two sample experiments averaged in 3 h solar light to deal with wastewaters polluted by benzoic
Average Average DO Average Average acid. Photolytic degradation and adsorption were found
solution level (mg/l) UV-A TOC to be insignificant in the experiments. The ‘‘waterfall’’
temperature intensity removal rate effect introduced by the cascade design can promote
(1C) (mW/cm2) (mgC/min) mass transport and aeration in the solution film.
The rate equation involving Langmuir-type depen-
18.2 8.5 1.74 0.9736
dence on Imean was found to provide comparatively good
25.9 7.3 1.73 0.8692
fits to 90 data points from 17 experiments performed at
 A.H.C. Chan et al. / Water Research 37 (2003) 1125–1135
181C. The average TOC removal rates did not demon-
strate significant dependence on TOC0 and the intensity
of the light was found to be the dominant factor
affecting the degradation process within the ranges of
the data, for the reaction system employed. The
proposed cascade reactor was capable of treating the
organic-contaminated wastewater. As demonstrated in a
sample experiment, the reactor degraded the starting
reactant and the intermediates identified in the 3-h
reaction. The same experiment also demonstrated that
the photoinduced consumption of oxygen took place
during the degradation process.
Ortho-, para- and meta-hydroxybenzoic acids were
identified as intermediates from the degradation of
benzoic acid. The major intermediate was meta-hydro-
xybenzoic acid. Addition of 10 ml of hydrogen peroxide
solution (30 wt%) increased the percentage removal of
TOC from 7 l of a 100 mg/l benzoic acid solution from
30% to 83%. Hydrogen peroxide solutions could also be
used to improve the efficiency of the degradation process
at elevated temperatures, which in turn reduced the
utility cost for temperature control.
Acknowledgements
The support of the Hong Kong RGC through its
Competitive Earmarked Research Grant HKUST 6026/
98P is gratefully acknowledged.
References
[1] Matthews RW. Photo-oxidation of organic material in
aqueous suspensions of titanium dioxide. Water Res 1986;
20(5):569–78.
[2] Matthews RW. Photooxidation of organic impurities in
water using thin films of titanium dioxide. J Phys Chem
1987;91:3328–33.
[3] Matthews RW. Kinetics of photocatalytic oxidation of
organic solutes over titanium dioxide. J Catal 1988;111:
264–72.
[4] Ollis DF, Pelizzetti E, Serpone N. Heterogeneous photo-
catalysis in the environment: application to water purifica-
tion. In: Serpone N, Pelizzetti E, editors. Photocatalysis:
Fundamentals and Application. New York: Wiley, 1989.
p. 603–37.
[5] Ollis DF, Pelizzetti E, Serpone N. Destruction of water
contaminants. Environ Sci Technol 1991;25(9):1523–9.
[6] Mills A, Davies RH, Worsley D. Water purification by
semiconductor photocatalysis. Chem Soc Rev 1993;22:
417–25.
1135
[7] Turchi CS, Ollis DF. Photocatalytic degradation of
organic water contaminants: mechanisms involving hydro-
xyl radical attack. J Catal 1990;122:178–92.
[8] Wyness P, Klausner JF, Goswami DY, Schanze KS.
Performance of nonconcentrating solar photocatalytic
oxidation reactors, part 1: flat-plate configuration. J Sol
Energy Eng 1994;116:2–7.
[9] Freudenhammer H, Bahnemann D, Bousselmi L, Geissen
SU, Ghrabi A, Saleh F, Si-Salah A, Siemon U, Vogelpohl
A. Detoxification and recycling of wastewater by solar-
catalytic treatment. Water Sci Technol 1997;35(4):157–64.
[10] Goslich R, Dillert R, Bahnemann D. Solar water treat-
ment: principles and reactors. Water Sci Technol 1997;
35(4):137–48.
[11] Sclafani A, Sciascia A, Rizzuti L. Pollutants photocatalytic
degradation in a falling film photoreactor. J Adv Oxid
Technol 1999;4(1):91–6.
[12] Chan AHC, Porter JF, Barford JP, Chan CK. Photo-
catalytic thin film cascade reactor for treatment of organic
compounds in wastewater. Water Sci Technol 2000;44(5):
187–95.
[13] Bird RB, Stewart WE, Lightfoot EN. Transport Phenom-
ena. New York: Wiley, 1960, p. 34–70.
[14] Crittenden JC, Zhang Y, Hand DW, Perram DL, Machard
EG. Solar detoxification of fuel-contaminated ground-
water using fixed-bed photocatalysts. Water Environ Res
1996;68(3):270–8.
[15] Egerton TA, King CJ. The influence of light intensity on
photoactivity in titanium dioxide pigmented systems. J Oil
Colour Chem Assoc 1979;62:386–91.
[16] Ollis DF. Solar-assisted photocatalysis for water purifica-
tion: issues, data, questions. In: Pelizzetti E, Schiavello M,
editors. Photochemical Conversion and Storage of Solar
Energy. Dordrecht: Kluwer Academic Publishers, 1990.
p. 593–622.
[17] Chang HT, Wu NM, Zhu F. A kinetic model for
photocatalytic degradation of organic contaminants in a
thin film TiO2 catalyst. Water Res 2000;34(2):407–16.
[18] Pichat P. Fundamentals of heterogeneous photocatalysis.
In: Yue PL, editor. Proceedings of the Croucher Advanced
Study Institute. PTC, HKUST, Hong Kong, 2000, p. 3–14.
[19] Chan AHC. A cascade solar-powered photocatalytic
reactor for wastewater treatment. MPhil. Thesis, Hong
Kong University of Science and Technology 3-1–4-24,
2001.
[20] Matthews RW. Photocatalytic oxidation of organic con-
taminant in water: an aid to environmental preservation.
Pure Appl Chem 1992;64(9):1285–90.
[21] Ollis DF. Process economics for water purification: a
comparative assessment. In: Schiavello M, editor. Photo-
catalysis and environment trends and applications. Dor-
drecht: Kluwer Academic Publishers, 1988. p. 663–77.
[22] Li Puma G, Yue PL. Comparison of the effectiveness of
photon-based oxidation processes in a pilot falling film
photoreactor. Environ Sci Technol 1999;33:3210–6.