Professional Documents
Culture Documents
Application of Box-Behnken Design With Response Surface Methodology For Modeling and Optimizing Ultrasonic Oxidation of Arsenite With H2O2
Application of Box-Behnken Design With Response Surface Methodology For Modeling and Optimizing Ultrasonic Oxidation of Arsenite With H2O2
Application of Box-Behnken Design With Response Surface Methodology For Modeling and Optimizing Ultrasonic Oxidation of Arsenite With H2O2
net/publication/259637687
CITATIONS READS
34 773
8 authors, including:
Some of the authors of this publication are also working on these related projects:
Full Solar Spectrum Harvesting and Storing Light Driven Materials View project
Study on guerrilla rainstorm, flood, and water pollution in megacity urban watersheds View project
All content following this page was uploaded by Kyounglim Kang on 20 June 2014.
y
Department of Environmental Engineering,
2
op
Kumoh National Institute of Technology,
Gyungbuk 730-701, Republic of Korea
Abstract: A combined ultrasound (US)/H2O2 process was used to oxidize arsenite to arsenate, yielding a synergistic effect value of 1.26. This
showed that the combined process could be an effective method of oxidizing arsenite, instead of using either ultrasonic or H2O2
oxidation processes. This combined process was successfully modeled and optimized using a Box–Behnken design with response
ho
surface methodology (RSM). The effects of the US power density, the initial concentration of arsenite, and the H2O2 concentration
on the sonochemical oxidation efficiency of arsenite were investigated. Analysis of variance indicated that the proposed quadratic
model successfully interpreted the experimental data with coefficients of determination of R2 = 0.95 and adjusted R2 = 0.91. Through
this model, we can predict and control the oxidation efficiency under different conditions. Furthermore, the optimal conditions for the
oxidation of arsenite were found to be a US power density of 233.26 W L-1, an initial arsenite concentration of 0.5 mg L-1, and an H2O2
concentration of 74.29 mg L-1. The predicted oxidation efficiency obtained from the RSM under the optimal conditions was 88.95%. A
ut
confirmation test of the optimal conditions verified the validity of the model, yielding an oxidation efficiency of 90.1%.
Keywords: Sonochemical Oxidation • Synergy • Kinetics • Mathematical modeling
© Versita Sp. z o.o.
A
* E-mail: hyeong@korea.ac.kr
164
P. Qiu et al.
Figure 1. Schematic of the experimental reactor setup: (a) cooling system, (b) transducer, (c) reactor, (d) thermometer, (e) ultrasound generator,
and (f) power meter.
y
oxidation of arsenite using potassium peroxydisulfate 2.2. Method
as an oxidizing agent. Lee et al. [12] investigated the The experimental setup is shown in Fig. 1. The
op
synergy between US and ultraviolet processes for sonoreactor consisted of a glass-cup–horn-type reactor
oxidation of As(III). However, the combination of US of capacity 2 L (Φ12.5 × 22 cm) equipped with a cup–
and an H2O2 oxidant for arsenite oxidation has not been horn-type ultrasonic transducer (Mirae Ultrasonic
studied so far. It is suggested that more ∙OH radicals MEGA-100). The frequency was 300 kHz, and the
will be generated in the US/H2O2 process because of maximum power (100 W) was applied using a single
rc
the pyrolysis of H2O2 through cavitation phenomena. In piezoelectric transducer (Tamura). The temperature in
addition, H2O2 itself has a redox potential of 1.77 eV and the solution was measured using a thermometer (Tecpel
can also oxidize arsenite. DTM-318) and maintained with a water jacket.
Response surface methodology (RSM), which The power densities were 160, 200, and 240 W L-1,
is based on factorial design, is a mathematical and respectively. The range of initial arsenite concentrations
statistical technique for designing experiments, was 0.5 mg L-1 to 1.5 mg L-1, based on the distribution of
ho
fitting models, and determining the optimal operating As in groundwaters around the world [18]. The selection
conditions for a target response [13]. It is usually used of the H2O2 concentration range was based on pretests.
to optimize and understand the performance of new and Fig. 2 shows the changes in the kinetic constants for
complex systems whose mechanisms are not clearly different H2O2 concentrations (20, 60, 120, 240, and
defined [14]. Recently, RSM has been widely used in 1000 mg L-1) at a fixed power density of 160 W L-1 and a
ut
the optimization of operating parameters in combined fixed initial arsenite concentration of 1 mg L-1. A pseudo-
systems [15-17]. first-order reaction was assumed. The chosen H2O2
In this work, the US/H2O2 process was used to concentration range was from 20 mg L-1 to 100 mg L-1.
oxidize arsenite to arsenate in a batch study. The
A
kinetics of the H2O2, US, and US/H2O2 processes were 2.3. Analysis
compared. Then, a Box–Behnken design with RSM was An anion cartridge (Supelclean LCSAX SPE 3 mL)
used to optimize and develop the US/H2O2 process. filter was used to separate the As(III) and As(V). The
The oxidation efficiency was measured as a function of mechanism for this method is clearly described in
power density, initial concentration of arsenite, and H2O2 Johnson’s study [19]. The As(III) concentration was
concentration, and a quadratic model was proposed. measured using an inductively coupled plasma atomic
emission spectrometer (Perkin Elmer 5300DV).
165
Application of Box–Behnken design with response surface
methodology for modeling and optimizing
ultrasonic oxidation of arsenite with H2O2
Table 1. Minimum and maximum levels of three factors in terms of coded and uncoded symbols.
(5)
y
1.5
op
1.0
In this model, xi (i = 1, 2, and 3) and xj (j = 1, 2, and
3) are independent variables, and β0, βi, and βii (i = 1,
0.5
2, and 3), and βij (i = 1, 2, and 3; j = 1, 2, and 3) are
the coefficients of the model obtained using polynomial
0.0 regression. Finally, ε is the random error. Design Expert
rc
1 10 100 1000 10000 V 8.0.6 (US, Stat-Ease Inc.) software was used to
H2O2 (mg L-1) evaluate this relationship.
designs can sharply reduce the number of experimental 3.1. Comparison of kinetics of each process
sets without decreasing the accuracy of the optimization Fig. 3 shows comparisons of the kinetics of the
compared with traditional factorial design methods. The H2O2, US, and US/H2O2 processes. It is clear that the
minimum and maximum ranges for the three factors are US/H2O2 process has the highest oxidation rate, when
ut
illustrated in terms of coded and uncoded symbols in each process was carried out separately. The oxidation
Table 1. The relationship between the coded variables rate in the US process is higher oxidation than that in the
and uncoded variables is described by the following H2O2 process because of the generation of ·OH radicals.
equations: The oxidation rate in the combined US/H2O2 process
A
166
P. Qiu et al.
0.7
2) Statistically designed experiments are performed.
0.6
3) The coefficients in the mathematical model are
0.5 H2O2 process estimated and the accuracy of the model is checked.
0.4
US process
US/H2O2 process
4) Response analysis is performed to predict the
optimal conditions, and these predictions are confirmed
0.3
0 10 20 30 40 50 60 70
experimentally.
Time/min
3.2.1. Model determination
Figure 3.
y
Comparison of the kinetics in each process. H2O2 Table 2 shows the statistical summary for each model
process: H2O2 concentration, 60mg L-1; US process: US
power density, 160 W L-1; US/H2O2 process: US power that was output by Design Expert V 8.0.6. A quadratic
density, 160 WL-1 and H2O2 concentration, 60 mg L-1. model was suggested, even though it has lower R2 and
op
adjusted-R2 (Adj-R2) values than a cubic model. This is
promotes the generation of ·OH radicals through because the cubic model is aliased, which means that
the decomposition of H2O2, which is caused by the the effects of each variable that cause different signals
extremely high temperatures and pressures formed in become indistinguishable. For a linear relationship,
collapsing cavitation bubbles. It has been reported that the R2 and Adj-R2 values are 0.6362 and 0.5370,
rc
the dissociation energy of the O–O bond (213 kJ mol-1) respectively. It is clear that the linear relationship is
in H2O2 is less than that of the O–H bond (418 kJ mol-1) not adequate for the experimental data. The quadratic
in water [15]. H2O2 is therefore more easily decomposed model was therefore selected to fit the experimental
than water. Furthermore, changes in the power density data.
and H2O2 concentration affect the synergistic effect [16].
ho
When the power density increases, not only the number 3.2.2. Model fitting and analysis of variance (ANOVA)
of cavitation bubbles, but also the temperature and Experiments were performed using the Box–Behnken
pressure inside the bubbles, will be increased. As the experimental design. The experimental and predicted
number of cavitation bubbles increases, the amount of oxidation efficiencies are shown along with the
H2O2 diffused into the bubbles increases, therefore more experimental conditions in Table 3.
ut
radicals will be generated. However, the increases in the Based on the model analysis in the first part,
temperature and pressure inside the bubbles will intensify a quadratic model was chosen to fit the data. The
the cavitation effect, leading to a higher decomposition relationship between the oxidation efficiency and the
rate of H2O2. Finally, the H2O2 concentration also affects three chosen factors is shown in Eq. 7:
A
167
Application of Box–Behnken design with response surface
methodology for modeling and optimizing
ultrasonic oxidation of arsenite with H2O2
y
1 -1 -1 0 160 0.5 60 71.6 69.5
op
3 -1 1 0 160 1.5 60 64.6 61.6
Table 4 shows the ANOVA results for the acquired the effect of cavitation increases, more H2O2 tends to
model. ANOVA is an analytical technique that is used to be decomposed, which significantly affects the optimal
identify the importance of the model and its parameters, concentration of H2O2. This result corresponds to the
A
using Fisher’s F-test and Student’s t-test [23]. Student’s discussion in the chapter 3.1.
t-test was used to determine the significance of the
regression coefficients using a p-value standard. In 3.2.3. Model modification
general, larger F-values and smaller p-values indicate After the significance of the parameters has been
more significant coefficient terms [24]. The coefficients evaluated, the model can be improved by eliminating
for the quadratic terms for the US power density and the terms that are not very significant. The final model
H2O2 concentration are shown to be very significant, for describing the relationship between the oxidation
which means that these factors have very large effects efficiency and the US power density, initial concentration
on the removal efficiency. However, the initial arsenite of arsenite, and H2O2 concentration is shown in Eq. 8:
concentration is only very significant in terms of its linear
effect, whereas its quadratic effect is very small. For the (8)
interaction effects, only the interaction between power
density and H2O2 concentration is significant. This is
because the size of the reaction zone increases with Although the R2 value of 0.9505 has decreased
increasing power density, thereby increasing the amount slightly, the Adj-R2, which is a more important factor for
of cavitation occurring per volume [25]. However, when determining the fit of a regression model, has changed
168
P. Qiu et al.
y
Residuals 53.28 5 10.66
op
2 2
90
from 0.8826 to 0.9134 [16]. Furthermore, the p-value
of less than 0.0001 indicates that the model is more
Predicted oxidation efficiency
85
significant.
80
rc
75 3.2.4. Model accuracy check
To obtain an adequate model, an accuracy check
70
is necessary. The model accuracy was checked by
65
comparing the predicted and experimental oxidation
60 efficiencies. Fig. 4 shows the linear relationship between
ho
55
the predicted and experimental oxidation efficiencies. In
addition, a normal plot of residuals between the normal
50
50 55 60 65 70 75 80 85 90 probability (%) and the internally studentized residuals
Experimental oxidation efficiency was also obtained. In this way, the residuals can be
checked to determine how well the model satisfies the
Figure 4. Comparison of predicted and experimental oxidation
ut
169
Application of Box–Behnken design with response surface
methodology for modeling and optimizing
ultrasonic oxidation of arsenite with H2O2
(a) (a)
y
op
rc
(b) (b)
ho
Figure 6. Effect of power density and initial concentration of Figure 7. Effect of power density and H2O2 concentration on the
arsenite on the arsenite oxidation efficiency with the H2O2 arsenite oxidation efficiency with the initial concentration
concentration at the zero level: (a) response surface and of arsenite at the zero level: (a) response surface and
(b) contour plot of the predicted oxidation efficiency. (b) contour plot of the predicted oxidation efficiency.
As can be seen from the plots, the oxidation from H2O2, thereby increasing the oxidation efficiency.
ut
efficiency increased with increasing power density. However, an overdose of H2O2 may inhibit radical-
However, increases in the power density beyond a induced oxidation, because H2O2 itself can serve as an
certain value did not lead to any further increase in ·OH radical scavenger. The ·OH-radical-scavenging
oxidation efficiency, and even caused a small decrease. reaction can be explained by Reactions 9 and 10
A
170
P. Qiu et al.
4. Conclusions
In this study, a combined US/H2O2 process was used to
oxidize arsenite to arsenate, yielding a synergistic effect
value of 1.26. This indicates that the combined process
is an effective process for oxidizing arsenite, instead of
using either ultrasonic or H2O2 oxidation processes. A
(a) Box–Behnken design with the RSM was successfully
applied to the sonochemical oxidation of arsenite in the
presence of H2O2. A quadratic model was proposed to
y
describe the relationship between the sonochemical
arsenite oxidation efficiency and the US power density,
op
initial concentration of arsenite, and H2O2 concentration.
The model fitted the experimental data well, with a
coefficient of determination, R2, of 0.9501 and an Adj-R2
of 0.9134. In addition, the p-value of this model was
less than 0.0001, which indicates that the model is very
rc
significant. The optimal conditions for the sonochemical
oxidation of arsenite were a US power density of
233.26 W L-1, an initial arsenite concentration of
0.5 mg L-1, and an H2O2 concentration of 74.29 mg L-1.
The optimal response obtained from the RSM was
ho
confirm the validity of the predicted optimal response, This work was supported by the Basic Science Research
three additional experiments were carried out under Program through a National Research Foundation of
the optimal conditions. The average removal efficiency Korea (NRF) grant funded by the Ministry of Education,
among the three repeated experiments was 90.1%. Science and Technology (KRF-2009-0092799) and
A
The good agreement between the experimental and supported by a NRF (2013R1A1A2006586).
References
[1] M. Bissen, F.H. Frimmel, Acta Hydrochim. [7] S.J. Hug, O. Leupin, Environ. Sci. Technol. 37,
Hydrobiol. 31, 97 (2003) 2734 (2003)
[2] Technical Application Bulletin for Arsenic (Water [8] H. Yang, W.Y. Lin, K. Rajeshwar, J. Photochem.
Quality Association, Lisle, IL, USA, 2004) Photobiol. 123, 137 (1999)
[3] M. Pettine, L. Campanella, F.J. Millero, Geochim. [9] T. Xu, Y. Cai, S.P. Mezyk, K.E. O’Shea, In:
Cosmochim. Acta 63, 2727 (1999) P.A. O’Day, D. Vlassopoulos, X. Meng, L.G. Benning
[4] W. Driehaus, R. Seith, M. Jekel, Water Res. 29, (Eds.), Advances in Arsenic Research, ACS
297 (1995) Symposium Series, Vol. 915 (ACS Publications,
[5] M.J. Kim, J. Nriagu, Sci. Total Environ. 247, 71 Washington DC, 2005) 333
(2000) [10] M.C. Cui, S.B. Lee, M. Jang, B. Kweon, H.Y. Jo,
[6] X.Y. Wei, C.A. Brockhoff-Schwegel, J.T. Creed, J.Y. Khim, Jpn. J. Appl. Phys. 50, 07HE13 (2011)
Analyst 125, 1215 (2000) [11] B. Neppolian, A. Doronila, M. Ashokkumar, Water
171
Application of Box–Behnken design with response surface
methodology for modeling and optimizing
ultrasonic oxidation of arsenite with H2O2
Res. 44, 3687 (2010) [21] M. Burton, K.C. Kurien, J. Phys. Chem. 63, 899
[12] S.B. Lee, M.C. Cui, S.M. Na, J.Y. Khim, Jpn. (1959)
J. Appl. Phys. 5, 07GD15 (2012) [22] A.M. Joglekar, A.T. May, Cereal Foods World, 32,
[13] G.E.P. Box, N.R. Draper, Empirical Model-Building 857 (1987)
and Response Surface (Wiley, New York, 1987) [23] J. Segurola, N.S. Allen, M. Edge, A.M. Mahon,
669 Prog. Org. Coat. 37, 23 (1999)
[14] N. Draper, J.A. John, Technometrics 30, 423 (1988) [24] R. Arunachalam, G. Annadurai, J. Environ. Sci.
[15] H. Zhang, H. Fu, D.B. Zhang, J. Hazard. Mater. Technol. 4, 65 (2011)
172, 654 (2009) [25] B.J. Lifka, B. Ondruschka, J. Hofman, Eng. Life Sci.
[16] Z.M. Zhang, H.L. Zheng, J. Hazard. Mater. 172, 3, 6 (2003)
1388 (2009) [26] H.L. Liu, Y.R. Chiou, Chem. Eng. J. 112, 173 (2005)
[17] M.S. Secula, G.D. Suditu, I. Poulios, C. Cojocaru, [27] J.F. Fu, Y.Q. Zhao, Q.L. Wu, J. Hazard. Mater. 144,
I. Cretescu, Chem. Eng. J. 141, 18 (2008). 499 (2007)
[18] P.L. Smedley, Arsenic in Ground Water (Kluwer, [28] M. Sivakumar, A.B. Pandit, Ultrason. Sonochem. 8,
Norwell, MA, USA, 2003) 179 233 (2001)
y
[19] R.L. Johnson, J.H. Aldstadt III, Analyst. 127, 1305 [29] G. Andaluri, E.V. Rokhina, P.S. Suri, Ultrason.
(2002) Sonochem. 19, 953 (2012)
op
[20] M. Khajeh, F.M. Zadeh, Bull. Environ. Contam. [30] K.J. Zhang, N.Y. Gao, Y. Deng, T.F. Lin, Y. Ma, L. Li,
Toxicol. 89, 38 (2012) rc M.H. Sui, J. Environ. Sci. 23, 31 (2011)
ho
ut
A
172