203
© IWA Publishing 2015 Water Science & Technology
|
71.2
|
2015
An experimental design approach for modeling As(V) adsorption from aqueous solution by activated carbon C. Bakkal Gula, E. Bilgin Simsek, D. Duranoglu and U. Beker
ABSTRACT The present paper discusses response surface methodology as an efficient approach for predictive model building and optimization of As(V) adsorption on activated carbon derived from a food industry waste: peach stones. The objectives of the study are application of a three-factor 23 full factorial and central composite design technique for maximizing As(V) removal by produced activated carbon, and examination of the interactive effects of three independent variables (i.e., solution pH, temperature, and initial concentration) on As(V) adsorption capacity. Adsorption equilibrium was investigated by using Langmuir, Freundlich, and Dubinin-Radushkevich isotherm models. First-order and second-order kinetic equations were used for modeling of adsorption kinetics. Thermodynamic
C. Bakkal Gula D. Duranoglu U. Beker (corresponding author) Chemical Engineering Department, Yildiz Technical University, Istanbul 34220, Turkey E-mail: [email protected] E. Bilgin Simsek Chemical and Process Engineering Department, Yalova University, Yalova 77100, Turkey
parameters (ΔG , ΔH , and ΔS ) were calculated and used to explain the As(V) adsorption W
W
W
mechanism. The negative value of ΔH (7.778 kJ mol1) supported the exothermic nature of the W
sorption process and the Gibbs free energy values (ΔG ) were found to be negative, which indicates that the As(V) adsorption is feasible and spontaneous. Key words
| activated carbon, adsorption, arsenic, kinetic model, response surface methodology
INTRODUCTION Arsenic, a known carcinogen in humans, is generally found in contaminated groundwater as a result of weathering of rocks, industrial waste discharges, agricultural use of arsenical herbicides, and pesticides (Bilici Baskan & Pala ). Drinking arsenic-rich water over a long period can result in adverse health effects including various types of cancers (bladder, kidneys, and lungs) and diseases of the blood vessels of the legs and feet (Dong et al. ). There are many methods for dealing with arsenic pollution and among them, adsorption is a versatile treatment technique practised widely in fine chemical and process industries for wastewater. Activated carbon (AC) based systems can remove a wide variety of toxic pollutants with very high removal efficiencies due to their well-developed pore structure (ranging from micro-pores to macro-pores), high active surface area, high degree of surface reactivity, and good mechanical characteristics (Özdemir et al. ). AC can be produced from a number of precursor materials, such as coal, peat, shell, or peach, apricot and cherry stones, or any other inexpensive material with a high carbon content. However, the usefulness of the adsorption process lies in the operational simplicity and reuse potential of adsorbents doi: 10.2166/wst.2014.491
during long-term applications (Wen & Wu ; Sahu et al. ). Wen & Wu () have stated that the adsorption process is difficult to understand in terms of the effects of independent variables and it does not depict their interactions on the dependent variable. Response surface methodology (RSM) is one of the statistical methods available to evaluate the effective factors and to design chemical and to physical processes in respect of the interactions between the input parameters. The experimental design technique gives an approximate description of an experimental region around a center of interest with validity of interpolation with a minimum number of experiments. Therefore, examining the effects of process parameters by statistical models is a more useful approach in order to understand the phenomena of As(V) adsorption onto activated carbon. The objectives of the study are: application of a three-factor 23 full factorial and central composite design technique for maximizing As(V) removal by activated carbon produced, and investigation of the interactive effects of three independent variables (i.e., solution pH, temperature, and initial concentration) on As(V) adsorption capacity.
C. Bakkal Gula et al.
204
|
An experimental design approach for modeling As(V)
Water Science & Technology
|
71.2
|
2015
MATERIALS AND METHODS
Statistical analysis
Materials
RSM is essentially a particular set of mathematical and statistical methods for designing experiments, building models, evaluating the effects of variables, and searching optimum conditions of variables to predict targeted responses (Song et al. ). RSM also offers a relationship between the controllable input parameters and response function. The three-factor, five-level central composite (CCD), and three-level 23 full factorial design (FFD) techniques were used in order to examine the effects of parameters. The quadratic model was defined by three selected parameters namely, pH (x1), initial As(V) concentration (x2), and temperature (x3). As(V) adsorption capacities (mg g1) of PSAC sample were designated as dependent variables (Yi). The levels of each factor chosen for CCD and 23 FFD were shown in Tables 1 and 2, respectively. For the three variables, the FFD for each categorical variable that consists of eight factorial points and six replicates at the center points (pH: 6.0, Ci: 4.5 mg L1, T: 45 C) were employed. The center points are used to estimate the experimental error and the duplicability of the data. STATISTICA (Ver. 8.0, StatSoft Inc., USA) package was used to improve the mathematical model and estimate the regression and graphical analyses.
All the chemicals/reagents used in this work were of analytical reagent grade. As(V) stock solutions were prepared by dissolving Na2HAsO4·7H2O (Sigma–Aldrich) in deionized water. Peach stones were obtained from a fruit juice factory in Bursa, Turkey.
Preparation of adsorbent The peach-stone-based activated carbon was prepared by the steam activation method as described in a previous study (Duranoğlu & Beker ). The carbonization experiments were conducted in a horizontal quartz tube reactor with a diameter of 6 cm and a length of 55 cm. The reactor was heated using a Protherm ASF tube furnace (Alserteknik, Turkey). The nitrogen flow rate was kept constant at 500 mL/min by using a Dwyer MMA flow meter (Dwyer MMA, MI, USA). Briefly, the pretreated peach stones were placed into the quartz tube reactor and steam activation was performed at 800 C with the heating rate of 5 C/min in nitrogen atmosphere. Steam/ nitrogen flow was passed through the reactor for 2 h at 800 C. Steam activated carbon was rinsed with distilled water until neutral pH. The sample was denoted as peach stone-based activated carbon (PSAC). W
W
W
W
Table 1
|
Levels of each factor for central composite design
Independent variables
Adsorption experiments Arsenate (As(V)) adsorption experiments were conducted by varying amount of adsorbent of PSAC (2–500 mg) with different concentrations (0.5–8.5 mg L1) at various pHs (2.0–10.0) and temperatures (25, 45, and 65 C). Thermodynamic studies were conducted at different temperatures (25, 45, and 65 C) at pH 4.0 with 4.5 mg L1 initial As(V) concentration. Samples were shaken at 130 rpm using water bath with orbital shaker (Memmert WB14-SV1422). The pH values of solutions were adjusted with HCl and NaOH twice a day. Equilibrium As(V) concentration was determined at 193.696 nm by atomic absorption spectrophotometer (Analytik Jena ContrAA 700 TR). Analysis were conducted at a wavelength of 193.7 nm by the graphite furnace system using Pd/Mg(NO3)2 as a matrix modifier. W
Level
x1, pH
x2, Ci (mg L1)
x3, T ( C)
2
2
0.5
25
1
4
2.5
35
0
6
4.5
45
1
8
6.5
55
2
10
8.5
65
W
W
Table 2
|
Levels of each factor for 23 full factorial design
Independent variables Level
x1, pH
x2, Ci (mg L1)
x3, T ( C)
1
4
2.5
35
0
6
4.5
45
1
8
6.5
55
W
C. Bakkal Gula et al.
205
|
An experimental design approach for modeling As(V)
RESULTS AND DISCUSSION
Table 3 represents range of three variables employed and their levels of independent reaction conditions (2, 0, þ2) according to the CCD. As(V) adsorption capacities were in the range of 34–229 μg g1. Moreover, the matrix of independent variables and the observed response values for FFD are given in Table 4. The comparison of observed and predicted As(V) sorption capacities were found in good agreement for both statistical methods. To determine the main and interaction effects of the independent variables on As(V) adsorption onto adsorbent, an analysis of variance (ANOVA) was performed. ANOVA results showed that interactions of pH, temperature, and concentration were highly significant according to the p-values. The coefficient of determination (R 2) values of the CCD and FFD models were found to be 0.931 and 0.992, respectively, indicating the accuracy and general availability of the proposed models. The ANOVA results of the response surface models for As(V) adsorption on PSAC showed that the effects and the |
|
Table 4
Response surface methodology and statistical analysis
Table 3
Water Science & Technology
The observed and predicted values for the factors (central composite design)
|
2015
The observed and predicted values for the factors (full factorial design)
Run
x1 (pH)
x2 (Ci, mg L1)
x3 (T, C)
qe,exp (μg g1)
qe,model (μg g1)
1
0
0
0
140.11
144.05
2
1
1
1
80.71
84.76
3
1
1
1
218.84
214.79
4
0
0
0
145.32
144.05
5
0
0
0
145.65
144.05
6
0
0
0
140.59
144.05
7
1
1
1
97.18
93.13
8
1
1
1
229.48
233.53
9
1
1
1
49.55
45.50
10
1
1
1
61.76
65.81
11
0
0
0
137.14
144.05
12
1
1
1
94.98
90.93
13
0
0
0
155.51
144.05
14
1
1
1
56.22
60.27
W
interactions of pH, temperature, and concentration were highly significant according to the p-values (P < 0.05) (Tables 5 and 6). Moreover, the applied polynomial model was found to be significant according to the adjusted determination coefficient and lack of fit test. After neglecting the insignificant terms, the response data were fitted to a quadratic polynomial regression equation with a significance level of p 0.05 and the model equations of the CCD (Equation (1)) and FFD (Equation (2)) models were established as
Run
x1 (pH)
x2 (Ci, mg L1)
x3 (T, C)
qe,exp (μg g1)
qe,model (μg g1)
1
0
2
0
180.71
181.01
2
0
0
0
140.11
140.83
3
1
1
1
80.71
69.10
4
2
0
0
98.61
126.18
5
1
1
1
218.84
196.35
6
0
0
0
145.32
140.83
28:82x1 x2 3:53x1 x3 3:14x2 x3 18:60x21
7
0
0
0
145.65
140.83
8
0
0
2
8:93x22 þ 0:29x23 ,
182.02
170.51
9
0
0
0
140.59
140.83
10
1
1
1
97.18
113.77
11
0
2
0
34.48
29.21
12
1
1
1
229.48
224.12
13
0
0
2
106.90
113.43
14
1
1
1
49.55
59.91
15
2
0
0
39.19
6.64
16
1
1
1
61.76
89.22
17
0
0
0
137.14
140.83
18
1
1
1
94.98
84.29
19
0
0
0
141.10
140.83
Residual
20
1
1
1
56.22
71.88
Total SS
W
|
71.2
Y ¼ 140:83 29:89x1 þ 37:95x2 14:27x3 (1)
Y ¼ 144:05 44:91x1 þ 39:34x2 9:76x3 28:82x1 x2 3:53x1 x3 3:14x2 x3 32:96x21 :
(2)
The linear term of the initial concentration shows a significant effect on the response compared to the Table 5
|
Analysis of variance of As(V) sorption for central composite design
Model
Sum of squares (SS)
df
Mean square
F-value
p-value
57482.05
9
6386.89
15.04
0.0001
10
424.64
4246.391 61728.44
19
C. Bakkal Gula et al.
206
Table 6
|
|
An experimental design approach for modeling As(V)
Analysis of variance of As(V) sorption for 23 full factorial design
Sum of squares
Model
df
Mean square
F-value
p-value
99.88
© IWA Publishing 2015 Water Science & Technology
|
71.2
|
2015
An experimental design approach for modeling As(V) adsorption from aqueous solution by activated carbon C. Bakkal Gula, E. Bilgin Simsek, D. Duranoglu and U. Beker
ABSTRACT The present paper discusses response surface methodology as an efficient approach for predictive model building and optimization of As(V) adsorption on activated carbon derived from a food industry waste: peach stones. The objectives of the study are application of a three-factor 23 full factorial and central composite design technique for maximizing As(V) removal by produced activated carbon, and examination of the interactive effects of three independent variables (i.e., solution pH, temperature, and initial concentration) on As(V) adsorption capacity. Adsorption equilibrium was investigated by using Langmuir, Freundlich, and Dubinin-Radushkevich isotherm models. First-order and second-order kinetic equations were used for modeling of adsorption kinetics. Thermodynamic
C. Bakkal Gula D. Duranoglu U. Beker (corresponding author) Chemical Engineering Department, Yildiz Technical University, Istanbul 34220, Turkey E-mail: [email protected] E. Bilgin Simsek Chemical and Process Engineering Department, Yalova University, Yalova 77100, Turkey
parameters (ΔG , ΔH , and ΔS ) were calculated and used to explain the As(V) adsorption W
W
W
mechanism. The negative value of ΔH (7.778 kJ mol1) supported the exothermic nature of the W
sorption process and the Gibbs free energy values (ΔG ) were found to be negative, which indicates that the As(V) adsorption is feasible and spontaneous. Key words
| activated carbon, adsorption, arsenic, kinetic model, response surface methodology
INTRODUCTION Arsenic, a known carcinogen in humans, is generally found in contaminated groundwater as a result of weathering of rocks, industrial waste discharges, agricultural use of arsenical herbicides, and pesticides (Bilici Baskan & Pala ). Drinking arsenic-rich water over a long period can result in adverse health effects including various types of cancers (bladder, kidneys, and lungs) and diseases of the blood vessels of the legs and feet (Dong et al. ). There are many methods for dealing with arsenic pollution and among them, adsorption is a versatile treatment technique practised widely in fine chemical and process industries for wastewater. Activated carbon (AC) based systems can remove a wide variety of toxic pollutants with very high removal efficiencies due to their well-developed pore structure (ranging from micro-pores to macro-pores), high active surface area, high degree of surface reactivity, and good mechanical characteristics (Özdemir et al. ). AC can be produced from a number of precursor materials, such as coal, peat, shell, or peach, apricot and cherry stones, or any other inexpensive material with a high carbon content. However, the usefulness of the adsorption process lies in the operational simplicity and reuse potential of adsorbents doi: 10.2166/wst.2014.491
during long-term applications (Wen & Wu ; Sahu et al. ). Wen & Wu () have stated that the adsorption process is difficult to understand in terms of the effects of independent variables and it does not depict their interactions on the dependent variable. Response surface methodology (RSM) is one of the statistical methods available to evaluate the effective factors and to design chemical and to physical processes in respect of the interactions between the input parameters. The experimental design technique gives an approximate description of an experimental region around a center of interest with validity of interpolation with a minimum number of experiments. Therefore, examining the effects of process parameters by statistical models is a more useful approach in order to understand the phenomena of As(V) adsorption onto activated carbon. The objectives of the study are: application of a three-factor 23 full factorial and central composite design technique for maximizing As(V) removal by activated carbon produced, and investigation of the interactive effects of three independent variables (i.e., solution pH, temperature, and initial concentration) on As(V) adsorption capacity.
C. Bakkal Gula et al.
204
|
An experimental design approach for modeling As(V)
Water Science & Technology
|
71.2
|
2015
MATERIALS AND METHODS
Statistical analysis
Materials
RSM is essentially a particular set of mathematical and statistical methods for designing experiments, building models, evaluating the effects of variables, and searching optimum conditions of variables to predict targeted responses (Song et al. ). RSM also offers a relationship between the controllable input parameters and response function. The three-factor, five-level central composite (CCD), and three-level 23 full factorial design (FFD) techniques were used in order to examine the effects of parameters. The quadratic model was defined by three selected parameters namely, pH (x1), initial As(V) concentration (x2), and temperature (x3). As(V) adsorption capacities (mg g1) of PSAC sample were designated as dependent variables (Yi). The levels of each factor chosen for CCD and 23 FFD were shown in Tables 1 and 2, respectively. For the three variables, the FFD for each categorical variable that consists of eight factorial points and six replicates at the center points (pH: 6.0, Ci: 4.5 mg L1, T: 45 C) were employed. The center points are used to estimate the experimental error and the duplicability of the data. STATISTICA (Ver. 8.0, StatSoft Inc., USA) package was used to improve the mathematical model and estimate the regression and graphical analyses.
All the chemicals/reagents used in this work were of analytical reagent grade. As(V) stock solutions were prepared by dissolving Na2HAsO4·7H2O (Sigma–Aldrich) in deionized water. Peach stones were obtained from a fruit juice factory in Bursa, Turkey.
Preparation of adsorbent The peach-stone-based activated carbon was prepared by the steam activation method as described in a previous study (Duranoğlu & Beker ). The carbonization experiments were conducted in a horizontal quartz tube reactor with a diameter of 6 cm and a length of 55 cm. The reactor was heated using a Protherm ASF tube furnace (Alserteknik, Turkey). The nitrogen flow rate was kept constant at 500 mL/min by using a Dwyer MMA flow meter (Dwyer MMA, MI, USA). Briefly, the pretreated peach stones were placed into the quartz tube reactor and steam activation was performed at 800 C with the heating rate of 5 C/min in nitrogen atmosphere. Steam/ nitrogen flow was passed through the reactor for 2 h at 800 C. Steam activated carbon was rinsed with distilled water until neutral pH. The sample was denoted as peach stone-based activated carbon (PSAC). W
W
W
W
Table 1
|
Levels of each factor for central composite design
Independent variables
Adsorption experiments Arsenate (As(V)) adsorption experiments were conducted by varying amount of adsorbent of PSAC (2–500 mg) with different concentrations (0.5–8.5 mg L1) at various pHs (2.0–10.0) and temperatures (25, 45, and 65 C). Thermodynamic studies were conducted at different temperatures (25, 45, and 65 C) at pH 4.0 with 4.5 mg L1 initial As(V) concentration. Samples were shaken at 130 rpm using water bath with orbital shaker (Memmert WB14-SV1422). The pH values of solutions were adjusted with HCl and NaOH twice a day. Equilibrium As(V) concentration was determined at 193.696 nm by atomic absorption spectrophotometer (Analytik Jena ContrAA 700 TR). Analysis were conducted at a wavelength of 193.7 nm by the graphite furnace system using Pd/Mg(NO3)2 as a matrix modifier. W
Level
x1, pH
x2, Ci (mg L1)
x3, T ( C)
2
2
0.5
25
1
4
2.5
35
0
6
4.5
45
1
8
6.5
55
2
10
8.5
65
W
W
Table 2
|
Levels of each factor for 23 full factorial design
Independent variables Level
x1, pH
x2, Ci (mg L1)
x3, T ( C)
1
4
2.5
35
0
6
4.5
45
1
8
6.5
55
W
C. Bakkal Gula et al.
205
|
An experimental design approach for modeling As(V)
RESULTS AND DISCUSSION
Table 3 represents range of three variables employed and their levels of independent reaction conditions (2, 0, þ2) according to the CCD. As(V) adsorption capacities were in the range of 34–229 μg g1. Moreover, the matrix of independent variables and the observed response values for FFD are given in Table 4. The comparison of observed and predicted As(V) sorption capacities were found in good agreement for both statistical methods. To determine the main and interaction effects of the independent variables on As(V) adsorption onto adsorbent, an analysis of variance (ANOVA) was performed. ANOVA results showed that interactions of pH, temperature, and concentration were highly significant according to the p-values. The coefficient of determination (R 2) values of the CCD and FFD models were found to be 0.931 and 0.992, respectively, indicating the accuracy and general availability of the proposed models. The ANOVA results of the response surface models for As(V) adsorption on PSAC showed that the effects and the |
|
Table 4
Response surface methodology and statistical analysis
Table 3
Water Science & Technology
The observed and predicted values for the factors (central composite design)
|
2015
The observed and predicted values for the factors (full factorial design)
Run
x1 (pH)
x2 (Ci, mg L1)
x3 (T, C)
qe,exp (μg g1)
qe,model (μg g1)
1
0
0
0
140.11
144.05
2
1
1
1
80.71
84.76
3
1
1
1
218.84
214.79
4
0
0
0
145.32
144.05
5
0
0
0
145.65
144.05
6
0
0
0
140.59
144.05
7
1
1
1
97.18
93.13
8
1
1
1
229.48
233.53
9
1
1
1
49.55
45.50
10
1
1
1
61.76
65.81
11
0
0
0
137.14
144.05
12
1
1
1
94.98
90.93
13
0
0
0
155.51
144.05
14
1
1
1
56.22
60.27
W
interactions of pH, temperature, and concentration were highly significant according to the p-values (P < 0.05) (Tables 5 and 6). Moreover, the applied polynomial model was found to be significant according to the adjusted determination coefficient and lack of fit test. After neglecting the insignificant terms, the response data were fitted to a quadratic polynomial regression equation with a significance level of p 0.05 and the model equations of the CCD (Equation (1)) and FFD (Equation (2)) models were established as
Run
x1 (pH)
x2 (Ci, mg L1)
x3 (T, C)
qe,exp (μg g1)
qe,model (μg g1)
1
0
2
0
180.71
181.01
2
0
0
0
140.11
140.83
3
1
1
1
80.71
69.10
4
2
0
0
98.61
126.18
5
1
1
1
218.84
196.35
6
0
0
0
145.32
140.83
28:82x1 x2 3:53x1 x3 3:14x2 x3 18:60x21
7
0
0
0
145.65
140.83
8
0
0
2
8:93x22 þ 0:29x23 ,
182.02
170.51
9
0
0
0
140.59
140.83
10
1
1
1
97.18
113.77
11
0
2
0
34.48
29.21
12
1
1
1
229.48
224.12
13
0
0
2
106.90
113.43
14
1
1
1
49.55
59.91
15
2
0
0
39.19
6.64
16
1
1
1
61.76
89.22
17
0
0
0
137.14
140.83
18
1
1
1
94.98
84.29
19
0
0
0
141.10
140.83
Residual
20
1
1
1
56.22
71.88
Total SS
W
|
71.2
Y ¼ 140:83 29:89x1 þ 37:95x2 14:27x3 (1)
Y ¼ 144:05 44:91x1 þ 39:34x2 9:76x3 28:82x1 x2 3:53x1 x3 3:14x2 x3 32:96x21 :
(2)
The linear term of the initial concentration shows a significant effect on the response compared to the Table 5
|
Analysis of variance of As(V) sorption for central composite design
Model
Sum of squares (SS)
df
Mean square
F-value
p-value
57482.05
9
6386.89
15.04
0.0001
10
424.64
4246.391 61728.44
19
C. Bakkal Gula et al.
206
Table 6
|
|
An experimental design approach for modeling As(V)
Analysis of variance of As(V) sorption for 23 full factorial design
Sum of squares
Model
df
Mean square
F-value
p-value
99.88
