Optimisation of Photo-Fenton-Like Degradation of Aqueous

Polyacrylic Acid Using Box-Behnken Experimental Design

Samira Ghafoori, Mehrab Mehrvar* and Philip Chan
Department of Chemical Engineering, Ryerson University, 350 Victoria Street, Toronto, Ontario, Canada, M5B 2K3

The effectiveness of photoFentonlike process to degrade aqueous polyacrylic acid (PAA) is investigated in a batch recirculation system using two
photoreactors in series. The response surface methodology (RSM) using the BoxBehnken experimental design combined with quadratic
programming is employed for the experimental design, the statistical analysis and the optimisation. The effects of the initial concentration of PAA, the
initial concentration of H2O2:Fe3, pH and the recirculation rate on the percent removal of total organic carbon (TOC)and the pseudosecond order
rate constant as the process responses are studied. The statistical analysis of the results indicates a satisfactory prediction of the system behaviour by
the developed quadratic models. Optimum operating conditions to maximise the percent TOC removal are also determined.
Keywords: photoFentonlike, polyacrylic acid, TOC removal, statistical analysis, process optimisation, advanced oxidation technologies

INTRODUCTION

Fe3 H2 O ! Fe2 OH H
hv

H2 O2 ! 2 OH

ynthetic watersoluble polymers are gaining more attention

due to industrial developments. Generally, they are known as
specialty polymers with functional groups such as carboxyl,
hydroxyl and amino groups that have applications in diverse elds
such as packaging, agriculture, detergents, superabsorbents and
others.[1,2] Nevertheless, the knowledge of the synthetic water
soluble polymers entering the environment is still sparse and due to
their versatility, the environmental pollution by this type of
polymers has become a major concern.[3,4] Also, as a result of their
water solubility, they are able to reach sewage disposal systems
and consequently contaminate water resources.[5] On the other
hand, since they contain a low concentration of polymers, they
neither can be recycled nor incinerated. Therefore, in order to
achieve applicable and robust results, experimental and analytical
methods must be adapted to gain insight into their degradation
process.
Advanced oxidation technologies (AOTs) as an alternative
method to the conventional treatment technologies have been
applied successfully to enhance the biodegradability of recalcitrant
pollutants in wastewater.[616] AOTs through the use of shortlived
highly reactive oxygen derived species render the contaminants to
less toxic intermediates.[1721] However, only few studies are
available in the open literature on the photochemical degradation
and mineralisation of polymers[5,2224]. The photoFentonlike
process, as one of the AOTs, has received great attention in recent
decades due to its capacity in oxidising and mineralising
recalcitrant compounds that are not amenable to biodegradation.[25,26] In the photoFentonlike process, hydroxyl radical
(OH), a powerful nonselective source of oxidation, is generated
from H2O2 in the presence of Fe3 ions based on the following
simplied reactions[27,28]:
Fe

H2 O2 ! Fe

HO2

Fe2 H2 O2 ! Fe3 OH OH

VOLUME 92, JANUARY 2014

1
2

hv

The great oxidation potential of these shortlived free radicals

(especially .OH) results in the fast degradation of target organic
compounds. The hydrogen abstraction from the main chain of
polymers by generated free radicals is the main cause of the
polymer photodegradation.[29] The main advantages of the Fenton
reagent are its ability to convert a wide range of pollutants to
biodegradable products without environmental threat by its
residual reagents, the ease of operation and the relatively low
cost of the reagents.[27] The photoFentonlike process, similar to
other AOTs, is a multifactor system where its efciency is affected
by different parameters such as initial concentrations of organic
compounds, H2O2 and iron, as well as pH and other operating
conditions. Therefore, crossfactor effects must be considered to
characterise the process and determine the synergetic effects of the
process variables.
The experimental design has been applied successfully to
identify the most inuential factors in multivariable systems.
The response surface methodology (RSM) is able to optimise the
operating conditions in multifactor systems by considering the
interactions among variables.[30] Also, RSM has been proven as a
reliable statistical tool in studying chemical treatment processes
to achieve an optimal response with a minimum number of
experiments.

*Author to whom correspondence may be addressed.

Email address: mmehrvar@ryerson.ca
Can. J. Chem. Eng. 92:97108, 2014
2013 Canadian Society for Chemical Engineering
DOI 10.1002/cjce.21849
Published online 29 May 2013 in Wiley Online Library
(wileyonlinelibrary.com).

THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING

97

In this study, the experimental design for the photodegradation

of polyacrylic acid (PAA) by photoFentonlike process was
investigated in a batch recirculation system using two photoreactors in series. PAA, a carbon chain watersoluble polymer with
the pendant carboxylic acids known to be recalcitrant to
biodegradation, was considered as a model watersoluble polymer.
The effects of the initial concentrations of PAA and H2O2:Fe3, pH
and the recirculation rate on the percent TOC removal and the tted
values of pseudosecond order mineralisation rate constant as
the process responses were studied using a fourfactor threelevel
BoxBehnken experimental design combined with RSM and
quadratic programming.
MATERIALS AND METHODS

portable pH meter (230A plus, Thermo Orion). No further pH

adjustment was made during the experiments.
The following protocol was pursued in conducting each
experiment. The PAA solution was diluted to achieve the desired
PAA concentration in a 4L solution. The lamp was turned on for
30 min before the beginning of each experiment to stabilise the
light intensity. A desired dosage of oxidisers (H2O2 and Fe3) was
added to the solution at the beginning of each experiment. The
solution was fed to the system and the temperature was kept
constant at 25C during each experiment by means of the heat
exchanger. The operating pressure was at ambient. The samples
were taken from the collection tank for further analysis. The TOC
concentration of the samples was measured by a TOC analyser
(Apollo 9000, Teledyne Tekmar, Mason, OH) immediately after
taking the samples.

Materials

Experimental Design and Optimisation Studies by RSM

PAA (35 wt%) with an average molecular weight (Mw) of

100,000 g mol1, H2O2 (30 wt%) and FeCl36H2O (SigmaAldrich)
were used as received. NaOH (99%) and H2SO4 (99%; EMD
Chemicals, Gibbstown, NJ) were used as received to adjust pH.

The optimum conditions for maximising the percent TOC removal

were determined by means of a fourfactor threelevel Box
Behnken experimental design combined with the response surface
methodology to correlate experimentally obtained criteria and
experimental conditions given by BoxBehnken experimental
design. The BoxBehnken design, a modied central composite
design, is known as an independent and rotatable quadratic design
having no fractional factorial points.[32] In this type of design, the
variable combinations are at the centre and the midpoints of the
edges of the variable space.[33] Also, compared to other types of
experimental designs such as full factorial design, the Box
Behnken experimental design needs fewer experimental trials.
Therefore, the effects of four independent variables on the response
functions were investigated. The independent variables were
initial concentrations of PAA (X1) and H2O2:Fe3(X2), pH (X3), and
recirculation rate (X4) that were coded as 1, 0 and 1 as shown in
Table 1. The total number of experimental trials was 27 based on
three levels and a four factor experimental design, with three
replicates at the centre of the design, to estimate a pure error sum of
squares. The percent TOC removal and the pseudosecond order
rate constant were considered as the dependent factors (process
responses). The independent variables and their critical experimental levels as shown in Table 1 were chosen based on the
preliminary experimental results and the values reported in the
open literature.[3436] In the RSM, as the initial step, an appropriate

Experimental Setup and Procedure

As shown in Figure 1, a laboratoryscale batch recirculation
photoreactor set up including two photoreactors in series providing
uniform light distribution was used in this study. A small annular
photoreactor (SLLAB; Siemens Inc.) with the annular space of
1.33 cm was used as a part of the recycle system including a
centrifugal magnetic pump (Model RK7201210; ColeParmer,
Tokyo, Japan), an all glass heat exchanger for controlling the
temperature, and a large volume tank with provisions for sampling
and temperature measurements. The system was also equipped
with a bypass valve to control the ow rate and to provide a relief
to the pump pressure. The lowpressure Hg lamp with input power
of 14 W with UV emission peaks at 254 nm (LP4130, Siemens Inc.)
sealed with the quartz sleeve was positioned at the centerline of
each photoreactor with stainless steel housing. This particular
geometry (very small annular space) and the method of irradiation
lead to have a good approximation of an isoactinic condition
(uniform light distribution) in the photoreactor.[31] The pH was
adjusted at the beginning of each experiment by adding a few drops
of 1 N NaOH or H2SO4 as needed and it was measured by a

Figure 1. Schematic diagram of the experimental setup in batch recirculation mode. (1) Reservoir tank, (2) Pump, (3) Flow meter, (4) UV photoreactor,
(5) Heat exchanger, (6) Cooling water inlet, (7) Water outlet, (8) Bypass and (9) UV photoreactor.

98

THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING

VOLUME 92, JANUARY 2014

Table 1. Independent variables and their coded levels based on BoxBehnken design
Coded Levels
Independent variable
1

[PAA]o (mg L )
[H2O2:Fe3]o (mg L1:(mg L1))
pH
Recirculation rate (L min1)

Symbol

X1
X2
X3
X4

10
400:4
3
1

30
900:6
5
6

50
1600:8
7
10

approximation which is generally a rstorder model is applied to

nd the true functional relationship between the response function
and the set of factors. However, in the rstorder model, there is a
lack of t due to the existence of the surface curvature. Therefore,
the rstorder model was upgraded by adding higher order
terms.[37,38] Consequently, in the next step, the experimental
data from the BoxBehnken experimental design was tted to the
following quadratic equation:
Y bo

k
X
i1

bi X i

k
X

bii X 2i

i1

k1 X
k
X

bij X i X j e

i1 j2

where Y, bo, bi, bii and bij are the predicted response, the constant
coefcient (intercept term), the linear coefcients, the quadratic
coefcients and the interaction coefcients, respectively. The
parameters Xi and Xj are independent variables, where k (in this
case, k 4) and e are the number of factors and the residual term
allowing uncertainties between observed and predicted values,
respectively. The statistical softwares, STATISTICA (trial version
10.0) and DesignExpert (trial version 8.0) were used for the
regression analysis and the parameter estimation of the response
functions, respectively. The statistical signicance of the model
equations was completely analysed using analysis of variance
(ANOVA) at 95% condence intervals. Threedimensional response surface plots and twodimensional contour plots were
developed while holding a variable constant in the quadratic
model. The experimental and predicted values were compared to
validate the developed models. The optimal operating conditions
to maximise the percent TOC removal were also determined
using a numerical technique built in the software DesignExpert
8.0. The pseudosecond order rate constant was calculated at the
determined optimal conditions. Also, another experimental trial
was carried out to verify the obtained optimal conditions by the
developed models for both response functions.

reduction using UV alone. In the UV/H2O2 process, a drastic

improvement in the TOC removal efciency was observed in both
single photoreactor and two photoreactors in series. The TOC
reduction by the UV/H2O2 process was due to the attack of the free
radicals to the polymer molecules. The free radicals (mainly .OH)
are generated by Reaction (4) and the following summarised
reactions:
H2 O2 OH ! HO2 H2 O

H2 O2 HO2 ! OH H2 O O2

RESULTS AND DISCUSSION

Preliminary Experiments
In order to obtain a preliminary insight into the TOC removal
efciency, different processes including UV alone, UV/H2O2, UV/
Fe3 and UV/H2O2/Fe3 for one photoreactor and two photoreactors in series were carried out. In each case, the TOC of the PAA
solution was determined during the total reaction time of 120 min.
Dark experiments did not show any TOC removal after 120 min.
The results of TOC removal for these processes, operated in
recirculating batch mode, are presented in Figure 2. As illustrated
in Figure 2, the TOC removal efciency in one single reactor using
UV alone was negligible (Figure 2a). Also, the TOC removal
efciency in two photoreactors in series showed only 33%

VOLUME 92, JANUARY 2014

Figure 2. Proles of TOC removal in: (a) one photoreactor; and (b) two
photoreactors in series for UV, UV/H2O2, UV/Fe3 and UV/H2O2/Fe3
processes ([PAA]o 30 mg L1, [H2O2]:[Fe3] 900:30 mg L1:mg L1,
pH 3 and recirculation rate 4 L min1).

THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING

99


HO2 ! O
2 H

second order polynomial equations:


H2 O2 O
2 ! OH O2 OH

TOC OH !    ! CO2 H2 O

10

Y 1 89:63  3:66X 1 2:16X 2  6:26X 3 0:21X 4 1:26X 1 X 2

 1:08X 1 X 3 0:26X 1 X 4 0:38X 2 X 3 0:33X 2 X 4
 0:075X 3 X 4  0:91X 21  0:46X 22  3:10X 23 0:29X 24
12

However, the TOC reduction by the UV/H2O2 process in

two photoreactors in series was higher because of the higher light
intensity resulting in the generation of more free radicals (Reaction
4). As illustrated in Figure 2a and b, the photoFenton process
(UV/Fe3) in the single photoreactor and two photoreactors in
series did not show high TOC removal efciencies (less than 45%).
This is due to the lower quantum yield of the hydroxyl radical
production by Reaction (3) (0.2 mol Einstein1) than the quantum
yield of Reaction (4) (0.5 mol Einstein1).[39] The highest TOC
removal efciency was correspondent to the photoFentonlike
process (UV/H2O2/Fe3) using two photoreactors in series as
shown in Figure 2b. Therefore, the effectiveness of the photo
Fentonlike process with two photoreactors in series as the
most efcient process was studied using response surface
methodology.
RSM Model Development
The efciency of the photoFentonlike process (UV/H2O2/Fe3)
for the treatment of aqueous PAA in a batch recirculation
system using two photoreactors in series was investigated by
BoxBehnken experimental design including four factors at
three levels. The process parameters taken into account were the
initial concentrations of PAA and H2O2:Fe3, pH and the
recirculation rate. The effectiveness of the process was evaluated
through the percent TOC removal for the model wastewater. Also,
the performance of the RSM on the pseudosecond order rate
constant was studied in order to indirectly introduce another
parameter (treatment time) in the experimental design.[40] In order
to compare the reaction rates, the reaction order has to be
determined and then the rate constant could be easily calculated
and compared through the linear regression if all of them could be
tted to the same reaction order.[41] Therefore, several functional
dependences of the TOC concentration on time representing
different integral order of reactions (zero, rst and second)
were examined. It was determined that the reaction rates of
the TOC removal for all 27 experimental trials were tted well to
the pseudosecond order kinetics ((1/TOCt)  (1/TOCo) kobs  t)
as shown in Figure 3. The expression to achieve the pseudo
second order rate constant is found by integrating the following
equation:


dTOC
kobs TOC2
dt

11

The range of linear regression coefcients (R2) was from 0.91 to

0.99. The fourfactor BoxBehnken experimental design and the
observed and predicted values for the percent TOC removal and
the pseudosecond order reaction constant by the developed
quadratic models are shown in Table 2. As mentioned earlier, the
RSM was used to estimate the parameters indicating an empirical
relationship between the input variables and the responses as
shown in Equation (5). Therefore, the quadratic model equations
for predicting the percent TOC removal (Y1) and the pseudo
second order rate constant (Y2) are presented in the following

100

THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING

Y 2 19  23:42X 1 4:25X 2  10:17X 3 1:83X 4 0:5X 1 X 2

8X 1 X 3  1:25X 1 X 4 0:75X 2 X 3  1:5X 2 X 4 2:25X 3 X 4
14:58X 21 1:58X 22 1:21X 23 4:21X 24
13
where X1, X2, X3 and X4 are coded terms for the four independent
variables indicating the initial concentrations of PAA and H2O2:
Fe3, pH and the recirculation rate, respectively. The negative
coefcients for the model components X1, X3, X1X3, X3X4, X 21 , X 22
and X 23 in Y1 and X1, X3, X1X4 and X2X4 in Y2 indicate the
unfavourable effects on the percent TOC removal and pseudo
second order rate constant, respectively. While, the positive
coefcients for X2, X4, X1X2, X1X4, X2X3, X2X4 and X 24 in Y1 and
X2, X4, X1X2, X1X3, X2X3, X3X4, X 21 , X 22 , X 22 and X 24 in Y2
indicate favourable effects on the percent TOC removal and
pseudosecond order rate constant, respectively. The signicance
of each model parameter will be determined by the statistical
analysis.
Statistical Analysis
The ANOVA was applied as the most important test to evaluate the
signicance of the developed quadratic model for the prediction of
the percent TOC removal and the pseudosecond order rate
constant. The results of the ANOVA are presented in Tables 3 and 4
for the percent TOC removal and the pseudosecond order rate
constant, respectively. The signicance of each coefcient in
Equations (12) and (13) was determined by the Fishers Ftest and
values of probability greater than F. The acceptance or rejection of
the model terms were based on the Pvalue with the 95%
condence level. As shown in Tables 3 and 4, small probability
values (P < 0.0001) indicate that the models were highly signicant which conrm the accuracy of the developed models to predict
the response functions. Also, high values of regression coefcient
(R2) as well as adjusted regression coefcient (R2adj ) ensure an
acceptable modication of the quadratic model to the experimental
data. In fact, R2adj modies R2 for the sample size and the number of
terms in the model. The values of R2 and R2adj were found to be
0.9363 and 0.8620 for the percent TOC removal and 0.9576 and
0.9081 for the pseudosecond order rate constant, respectively.
The closer the values of R2 and R2adj are to 1, the better the prediction
of models for the percent TOC removal and the pseudosecond
order rate constant is. The high values of regression and adjusted
regression coefcients indicate that the developed quadratic
models could adequately describe the system behaviour within
the selected range of operating parameters. Also, the adequate
precisions, the signal to the noise ratio, greater than 4 (12.80 for
percent TOC removal and 15.69 for the pseudosecond order rate
constant) show that the models could be used to navigate the
design space. The high correlations between the observed and the
predicted data are illustrated in Figures 4a and 5a. As a result of low
discrepancies, the points are very close to the diagonal line,
indicating a good agreement between the actual and the predicted
data. The normality of the data could be checked through the
normal probability plot of the residuals as shown in Figures 4b

VOLUME 92, JANUARY 2014

Figure 3. Pseudosecond order kinetic t for experimentally achieved TOC in 27 experimental trials. (See Table 2 for the conditions of each experimental
trial.)

and 5b. Since the points on the plot follow a straight line, it could
be concluded that the residuals are normally distributed. Also,
as the Sshaped curve is not formed in Figures 4b and 5b,
the response transformations are not needed.[30] Therefore, it

VOLUME 92, JANUARY 2014

could be concluded that the predictions of the experimental data

by developed quadartic models for both the percent TOC
removal and the pseudosecond order rate constant are quite
satisfactory.

THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING

101

Table 2. Fourfactor BoxBehnken design for RSM and the observed and predicted responses
Independent coded variables

kobs  104 (M1 s1)

TOC removal (%)

Run

X1

X2

X3

X4

Observed

Predicted

Observed

Predicted

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27

0
0
0
1
0
1
1
1
1
0
0
1
1
0
0
0
0
1
1
0
0
1
0
1
0
0
1

1
1
0
1
1
0
1
0
0
0
0
0
1
1
0
0
1
0
0
0
1
1
1
0
1
0
0

0
0
1
0
0
1
0
0
0
1
0
0
0
1
0
1
1
0
1
1
1

1
1
1
0
1
0
0
1
1
1
0
1
0
0
0
1
0
1
0
1
0
0
0
0
1
0
0

89.98
87.55
80.98
95.13
90.30
95.65
88.21
87.53
87.86
80.17
89.26
91.20
89.86
83.34
88.97
91.94
79.49
92.68
82.89
92.35
93.66
77.91
91.34
72.61
88.44
90.70
89.71

91.08
87.18
80.84
92.82
92.16
94.46
88.02
85.40
85.30
80.27
89.63
93.14
91.02
82.35
89.63
92.94
77.27
92.20
84.10
93.21
94.10
81.18
90.55
74.62
87.42
89.26
89.30

24
21
14
69
24
74
22
13
13
14
19
67
55
18
19
35
12
62
29
36
32
6
29
4
15
19
17

29
24
18
62
29
76
16
14
15
10
19
64
55
17
19
35
7
58
40
34
35
7
28
9
17
19
14

1
1
0
0
1

Table 3. ANOVA results for prediction of percent TOC removal

Statistics
Factors (coded)
Model
X1
X2
X3
X4
X1X2
X1X3
X1X4
X2X3
X2X4
X3X4
X21
X22
X23
X24
Residual
Pure error

SSa

dfb

MSc SS/df

Fvalue

Pvalued Prob > F

761.57
160.60
56.03
469.63
0.55
6.33
4.71
0.28
0.59
0.43
0.023
4.39
1.15
51.21
0.43
51.81
0.66

14
1
1
1
1
1
1
1
1
1
1
1
1
1
1
12
2

54.40
160.60
56.03
469.63
0.55
6.33
4.71
0.28
0.59
0.43
0.023
4.39
1.15
51.21
0.43
4.32
0.33

12.60
37.20
12.98
108.77
0.13
1.47
1.09
0.064
0.14
0.099
5.21E003
1.02
0.27
11.86
0.10

<0.0001
<0.0001
0.0036
<0.0001
0.7283
0.2494
0.3169
0.8048
0.7192
0.7580
0.9436
0.3331
0.6156
0.0049
0.7569

Sum of squares.
Degree of freedom.
c
Mean square.
d
P < 0.05 is considered as signicant.
b

102

THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING

VOLUME 92, JANUARY 2014

Table 4. ANOVA results for prediction of pseudosecond order rate constant

Statistics
Factors (coded)
Model
X1
X2
X3
X4
X1X2
X1X3
X1X4
X2X3
X2X4
X3X4
X21
X22
X23
X24
Residual
Pure error

SSa

dfb

MSc SS/df

Fvalue

Pvalued Prob > F

9625.94
6580.08
216.75
1240.33
40.33
1.00
256.00
6.25
2.25
9.00
20.25
1134.26
13.37
7.79
94.45
426.58
0.000

14
1
1
1
1
1
1
1
1
1
1
1
1
1
1
12
2

687.57
6580.08
216.75
1240.33
40.33
1.00
256.00
6.25
2.25
9.00
20.25
1134.26
13.37
7.79
94.45
35.55
0.000

19.34
185.10
6.10
34.89
1.13
0.028
7.20
0.18
0.063
0.25
0.57
31.91
0.38
0.22
2.66

<0.0001
<0.0001
0.0295
<0.0001
0.3078
0.8696
0.0199
0.6824
0.8056
0.6240
0.4650
0.0001
0.5511
0.6481
0.1290

Sum of squares.
Degree of freedom.
Mean square.
d
P < 0.05 is considered as signicant.
b
c

Effects of Model Parameters and Their Interactions

As the signicance of the models was conrmed (Tables 3 and 4)
and the predictivity of the experimental data by the developed
models found to be accurate (Figures 4 and 5), the next step is to
evaluate the signicance of each model parameter. Hence, the most
inuential process parameters on the percent TOC removal and the
pseudosecond order rate constant could be determined. This
evaluation is based on the F and Pvalues for each of the linear,
quadratic and interaction model paramaters as listed in Tables 3
and 4. Generally, Fischers Fvalue is used to determine the
signicance of the regression coefcients of the variables while
Pvalue determines the signicance of the variable that may
indicate the pattern of interaction among variables. The larger the
value of F and the smaller the value of P, the more siginicant is
the corresponding coefcient.[42] As shown in Tables 3 and 4, the
Pvalue less than 0.05 indicate the model parameters are
signicant. As presented in Tables 3 and 4, the initial concentrations of PAA and H2O2:Fe3 as well as pH have signicent linear
effects on the percent TOC removal and pseudosecond order rate
constant. The effect of the initial concentration of PAA on the
percent TOC removal and pseudosecond order rate constant is
mainly due to the absorption of UV radiation by PAA molecules. An
increase in the initial concentration of PAA results in inducing an
inner lter effect, that is the UV light penetration becomes less due
to less permeability of the solution.[43] This leads to a decrease in
the UV radiation absorption by H2O2 and Fe3 and subsequently
less hydroxyl radical formation which is the main cause of the PAA
degradation. Also, the high amount of PAA molecules reduces their
chances to be in contact with hydroxyl radicals and consequently
fewer molecules are attacked by the hydroxyl radicals. Based on
Reactions (1) and (2), higher concentrations of hydrogen peroxide
and ferric iron result in the formation of more ferrous iron and
subsequently more hydroxyl radicals, which are the major cause of
the polymer degradation. In more acidic pH, a higher level of

VOLUME 92, JANUARY 2014

dissolved iron will be maintained in the aqueous solution. Also, the

oxidation potential of hydroxyl radicals is higher at more acidic
conditions.[36]
The recirculation rate did not show any signicant effect on the
response functions, showing that this factor can be ignored in
comparison to the other inuential factors. In liquid phase
reactions, the recirculation rate or the rate of mixing is usually
considered as an important factor. The higher rate of mixing
increases the rate of reaction due the increase of mass transfer rate.
The insignicant effect of the rate of mixing in this study indicates
that the rate of mixing in the selected range in a small photoreactor
volume provides a good mixing condition in all studied levels.
As indicated in Table 3, the interaction model parameters (X1X2,
X1X3, X1X4, X2X3, X2X4 and X3X4) did not show any signicant
effect on the percent TOC removal while the interaction effect of the
initial concentration of polymer and the pH (X1X3) shows a
signicant effect on the pseudosecond order rate constant. The
signicant effect of X1X3 indicates that the interaction effect of the
initial concentration of polymer and pH is benecial to the percent
TOC removal or the treatment time. The quadratic model
parameters X 23 (the quadratic effect of the pH) and X 21 (the
quadratic effect of the initial concentration of the polymer) show
signicant effects on the percent TOC removal and pseudosecond
order rate constant.This nding shows that the percent TOC
removal is more pronounced by the pH and the pseudosecond
order rate constant or the treatment time is more affected by the
initial concentration of the polymer.
3D Response Surface and 2D Contour Plots
The threedimensional (3D) response surface and twodimensional
(2D) contour plots as the graphical representations of the
regression analysis are presented in Figures 6 and 7 for the percent
TOC removal and the pseudosecond order rate constant,
respectively. In such plots, the response function of two factors

THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING

103

Figure 4. Validation of models for the TOC percent removal using different
plots of: (a) observed experimental data versus predicted values and (b)
normal probability.

Figure 5. Validation of the prediction of pseudosecond order rate

constant using different plots of: (a) observed experimental data versus
predicted values; and (b) normal probability.

are presented while all others are at xed levels. The 3D response
surface plots are formed based on the quadratic model Equations (12) and (13) while the relationship between the model
factors and the response functions are illustrated by corresponding
contour plots. The nonlinear nature of all 3D response surfaces in
Figures 6 and 7 show considerable interactions between independent variables and their corresponding response functions.
Moreover, the nonlinear contour plots in Figures 6 and 7 demontrate that there is no direct linear relationship among the selected
independent variables. Figures 6a and 7a present the inuence of
the initial concentration of PAA (X1) and the initial concentration
of H2O2:Fe3 (X2) on the percent TOC removal and pseudosecond
order rate constant in photoFentonlike process, respectively. As is
evident in these gures, higher initial concentrations of PAA have
negative effects on the response functions while higher H2O2:Fe3
dosage results in higher TOC removal and higher rate constant. The
high negative coefcient of X1 in the response functions(Equations 12 and 13) also conrms the antagonistic effect of the initial
concentrations of PAA on the response functions. Furthermore, as
shown in these gures, the effect of the initial concentration of PAA
is more inuential than the other factor as it is in agreement
with the signicance of these two process parameters on the basis
of Pvalues (Tables 3 and 4). However, as shown by the ANOVA
results (Tables 3 and 4), the interaction between the initial
concentrations of PAA and the oxidant ratio (H2O2:Fe3) does not
show a signicant effect on the response functions.

Figures 6b and 7b present the interaction effects between the

initial concentration of the polymer and the pH. As is evident in
Figure 6b, the higher percent of TOC removal at different initial
concentrations of the polymer is achieved at lower pH (35). As
illustrated in Figure 7b, the variation of the initial concentration of
the polymer on the pseudosecond orderrate constant is more
inuential at lower pH. The Pvalue <0.05 in Table 4 also shows
the signicant effect of the interaction between the initial
concentration of PAA and the pH on the pseudosecond order
rate constant. At low concentration of PAA and low pH, a higher
rate constant is achieved which could be conrmed by the high
negative values of coefcients of X1 and X3 in Equation (13).
The pH was selected in the acidic range since it has been proven
that in the photo Fentonlike process, a high level of dissolved ferric
iron species is achieved at pH buffered in acidic range.[44] Also, in
the acidic condition, the prolonged reactivity of H2O2 is higher and
the lifetime of H2O2 is decreased at buffer pH as it can be seen in
Figures 6b and 7b.
As illustrated in Figures 6c and 7c and as conrmed previously in
Tables 3 and 4, the recirculation rate does not show a signicant
effect on the percent TOC removal and the pseudosecond order
rate constant. Also, the interaction effects of the initial concentration of the polymer and the recirculation rate on the response
functions are insignicant.
Figure 6d illustrates the signicant effects of the initial dosage of
H2O2:Fe3 and the pH on the percent TOC removal. As shown in

104

THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING

VOLUME 92, JANUARY 2014

Figure 6. Interaction effects of different parameters on the percent TOC removal using 3D response surface and 2D contours. (a) [PAA]o and [H2O2:Fe3]o;
(b) [PAA]o and pH; (c) [PAA]o and recirculation rate; (d) [H2O2:Fe3]o and pH; (e) [H2O2:Fe3]o and recirculation rate; (f) pH and recirculation rate.
X1 [PAA]o, X2 [H2O2:Fe3]o, X3 pH and X4 recirculation rate.

this gure, the pH changes are more important than the other factor
while in Figure 7d, both factors show small changes on the
pseudosecond order rate constant (Figure 7d). Figure 6e also
conrms the signicant effect of the oxidants concentration (H2O2:
Fe3) on the percent TOC removal while Figure 7e shows that
the changes of the oxidant concentration are less important in the

VOLUME 92, JANUARY 2014

pseudosecond order rate constant. Figure 6f also shows the

signicant effect of the pH on the percent TOC removal at different
recirculation rates while as is evident in Figure 7f, the pH is less
inuencial factor on the pseudosecond order rate constant. Also,
as this gure shows, at higher recirculation rates,decreasing pH has
more positive effect on the pseudosecond order rate constant.

THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING

105

Figure 7. Interaction effects of different parameters on the pseudosecond order rate constant using 3D response surface and 2D contours. (a) [PAA]o and
[H2O2:Fe3]o; (b) [PAA]o and pH; (c) [PAA]o and recirculation rate; (d) [H2O2:Fe3]o and pH; (e) [H2O2:Fe3]o and recirculation rate; (f) pH and recirculation
rate. X1 [PAA]o, X2 [H2O2:Fe3]o, X3 pH and X4 recirculation rate.

Optimisation of Process Parameters

The developed quadratic model for the percent TOC removal
(Equation 12) was used as an objective function to nd the
optimum operating conditions (the initial concentration of PAA,
the initial dosage of H2O2:Fe3 the pH and the recirculation rate)
that maximise the percent TOC removal. Therefore, the percent

106

THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING

TOC removal (Y1 in Equation 12) was maximised at dened

optimisation criteria for the factors (10 < X1 < 50; 4:400 < X2
< 1600:8; 3 < X3 < 7; and 1 < X4 < 10). Consequently, the optimum
process parameters were found by using the numerical technique
built in the Design Expert Software 8.0.5 based on the predicted
model and the factors in their critical range as the constraints. The
numerical optimisation searches the design space using the

VOLUME 92, JANUARY 2014

developed model in the analysis to nd factor settings that meet

the goal of maximising the percent TOC removal. The optimum
values to achieve the maximum 95.7% TOC removal after 120 min
were 10 mg L1 PAA, 1600 m L1 H2O2, 8 mg L1 Fe3, pH 3.42 and
10 L min1 recirculation rate. The pseudosecond order rate
constant at the optimal condition based on the developed quadratic
model was found to be 82 M1 s1. The TOC removal efciency was
tested experimentally at the optimum operating conditions to
validate the model prediction. 92.5% TOC removal and 86 M1 s1
pseudosecond order rate constant achieved experimentally at the
optimal operating conditions conrm the results of the developed
models.
CONCLUSIONS
The photoassisted Fentonlike degradation of PAA as a model
watersoluble compound was investigated using two photoreactors
in series. The initial concentrations of the polymer, H2O2:Fe3,
the pH and the recirculation rate were considered as the process
parameters. The percent TOC removal and the pseudosecond
order rate constant were the response functions. The threelevel
BoxBehnken statistical experiment design applied in this study
was found to be an appropriate response surface methodology to
determine the effects of process parameters on the response
functions in the studied range. Two quadratic models predicting
the percent TOC removal and the pseudosecond order rate
constant were developed. The statistical analysis using ANOVA
indicated that the developed quadratic models were highly
accurate and predictive. Also, the efciency of the photoFenton
like process was strongly affected by the initial concentration of the
polymer and the pH. The oxidant ratio (H2O2:Fe3) showed less
signicant effect on the response functions while the recirculation
rate had no signicant effect on the response functions. The
optimal operating conditions to achieve the maximum percent
of TOC removal (95.7%) in the selected range was determined to
be 10 mg L1 PAA, 1600 mg L1 H2O2, 8 mg L1 Fe3, pH 3.42 and
10 L min1 recirculation rate. The pseudosecond order rate
constant at the optimum conditions was found to be 82 M1 s1.
The obtained optimal operating conditions were also validated
experimentally. The good agreement between the model predictions and the experimental results conrmed the accuracy of the
developed models. Obtaining the signicant operating conditions
and their optimal values could be considered in the photoreactor
design and modelling for the purpose of photoreactor scaleup.
NOMENCLATURE
e
Fvalue
k
kobs
Mw
Pvalue
R2
R2adj
t
Xi, Xj
X1
X2
X3
X4
Y
Y1
Y2

residual term
Fisher test value
number of factors
pseudosecond order rate constant (M1 s1)
molecular weight (g mol1)
probability value
regression coefficient
adjusted regression coefficient
time (min)
independent variables
initial concentration of PAA (mg L1)
initial concentration of H2O2:Fe3 (mg L1:mg L1)
pH
recirculation rate (L min1)
response function
percent TOC removal (%)
pseudosecond order rate constant (M1 s1)

VOLUME 92, JANUARY 2014

Greek Symbols
bo
bi
bii
bij

intercept term
linear coefficient
quadratic coefficient
interaction coefficient

Acronyms
2D
3D
ANOVA
AOT
df
MS
PAA
RSM
SS
TOC

twodimensional
threedimensional
analysis of variance
advanced oxidation technology
degree of freedom
mean square
polyacrylic acid
response surface methodology
sum of square
total organic carbon

ACKNOWLEDGEMENTS
The nancial support of the Natural Sciences and Engineering
Research Council of Canada (NSERC) and Ryerson University are
gratefully appreciated.
REFERENCES
[1] A. C. Bnard, V. Darcos, C. Drakides, C. Casellas, J. Coudane,
M. Vert, J. Polym. Environ. 2011, 19, 40.
[2] M. H. AlRashed, M. Lemanowicz, A. T. Gierczycki, Int. J.
Miner. Process 2012, 104105, 1.
[3] F. Hassouna, G. Mailhot, S. MorlatThrias, J. L. Gardette, J.
Photochem. Photobiol. A 2011, 218, 239.
[4] G. E. Luckachan, C. K. S. Pillai, J. Polym. Environ. 2011, 19,
637.
[5] J. A. Giroto, A. C. S. C. Teixeira, C. A. O. Nascimento, R.
Guardani, Ind. Eng. Chem. Res. 2010, 49, 3200.
[6] M. Mehrvar, W. A. Anderson, M. MooYoung, Int. J.
Photoenergy 2000, 2, 67.
[7] M. Mehrvar, W. A. Anderson, M. MooYoung, Int. J.
Photoenergy 2001, 3, 187.
[8] S. H. Venhuis, M. Mehrvar, J. Environ. Sci. Heal 2005, A40,
1731.
[9] M. B. Johnson, M. Mehrvar, Ind. Eng. Chem. Res. 2008, 47,
6525.
[10] J. M. Poyatos, M. M. Munio, M. C. Almecija, J. C. Torres, E.
Hontoria, F. Osorio, Water Air Soil Pollut. 2010, 205, 187.
[11] W. Cao, M. Mehrvar, Chem. Eng. Res. Des. 2011, 89, 1136.
[12] M. Barrera, M. Mehrvar, K. A. Gilbride, L. H. McCarthy, A. E.
Laursen, V. Bostan, R. Pushchak, Chem. Eng. Res. Des. 2012,
90, 1335.
[13] S. Ghafoori, M. Mehrvarand, P. K. Chan, Iran. Polym. J. 2012,
21, 869.
[14] S. Ghafoori, M. Mehrvar, P. K. Chan, Ind. Eng. Chem. Res.
2012, 51, 14980.
[15] M. Mohajerani, M. Mehrvar, F. EinMozaffari, Can. J. Chem.
Eng. 2012, 90, 719.
[16] M. Mohajerani, M. Mehrvar, F. EinMozaffari, Sep. Purif.
Technol. 2012, 90, 173.

THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING

107

[17] M. Mehrvar, W. A. Anderson, M. MooYoung, Int. J.

Photoenergy 2002, 4, 141.
[18] M. Mehrvar, W. A. Anderson, M. MooYoung, Adv. Environ.
Res. 2002, 6, 411.
[19] G. B. Tabrizi, M. Mehrvar, J. Environ. Sci. Heal 2004, A39,
3029.
[20] A. Asadi, M. Mehrvar, Int. J. Photoenergy 2006, 2006, 1,
Article ID 19790.
[21] M. Mohajerani, M. Mehrvar, F. EinMozaffari, Ind. Eng.
Chem. Res. 2010, 49, 5367.
[22] T. Aarthi, M. S. Shaama, G. Madras, Ind. Eng. Chem. Res.
2007, 46, 6204.
[23] L.J. Hsu, L.T. Lee, C. Lin, Chem. Eng. J. 2011, 173, 698.
[24] R. Vinu, G. Madras, Ultrason. Sonochem. 2011, 18, 608.
[25] J. Farias, E. D. Albizatti, O. M. Alfano, Catal. Today 2009,
144, 117.
[26] I. Grcic, M. Maljkovic, S. Papic, N. Koprivanac, J. Hazard.
Mater. 2011, 197, 272.
[27] M. A. Tarr, Chemical Degradation Methods for Wastes and
Pollutants. Marcel Dekker Inc., New York 2003, pp. 225228.
[28] M. Tokumura, R. Morito, R. Hatayama, Y. Kawase, Appl.
Catal. B 2011, 106, 565.
[29] N. B. Shukla, N. Daraboina, G. Madras, Polym. Degrad. Stab.
2009, 94, 1238.
[30] M. S. Tanyildizi, Chem. Eng. J. 2011, 168, 1234.
[31] J. Marugan, R. V. Grieken, C. Pablos, M. L. Satuf, A. E.
Cassano, O. M. Alfano, Appl. Catal. B 2011, 102, 404.
[32] H. Zhang, X. Ran, X. Wu, D. Zhang, J. Hazard. Mater. 2011,
188, 261.
[33] F. Ay, E. C. Catalkaya, F. Kargi, J. Hazard. Mater. 2009, 162,
230.
[34] M. R. Rojas, F. Prez, D. Whitley, R. G. Arnold, A. E. Sez,
Ind. Eng. Chem. Res. 2010, 49, 11331.
[35] T. C. F. Silva, C. M. Silva, C. Reis, C. R. Bellato, L. A. Lucia,
Ind. Eng. Chem. Res. 2010, 49, 11214.
[36] S. P. Sun, A. T. Lemley, J. Mol. Catal. A Chem. 2011, 349, 71.
[37] R. H. Myers, D. C. Montgomery, Response Surface Methodology: Process and Product Optimisation Using Designed
Experiment, 2nd edition. John Wiley and Sons, New York
2002, pp. 1355.
[38] K. Yetilmezsoy, S. Demirel, R. J. Vanderbei, J. Hazard. Mater.
2009, 171, 551.
[39] M. Simunovic, H. Kusic, N. Koprivanac, A. L. Bozic, Chem.
Eng. J. 2011, 173, 280.
[40] H. Kusic, M. Jovic, N. Kos, N. Koprivanac, V. Marin, J.
Hazard. Mater. 2010, 183, 189.
[41] M. Dopar, H. Kusic, N. Koprivanac, Chem. Eng. J. 2010, 173,
267.
[42] E. Bayraktar, Process Biochem. 2001, 37, 169.
[43] B. Xu, N.Y. Gao, H. Cheng, S.J. Xia, M. Rui, D.D. Zhao, J.
Hazard. Mater. 2009, 162, 954.
[44] Y. S. Jung, W. T. Lim, J.Y. Park, Y.H. Kim, Environ. Technol.
2009, 30, 183.

Manuscript received October 5, 2012; revised manuscript received

November 21, 2012; accepted for publication December 15, 2012.

108

THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING

VOLUME 92, JANUARY 2014