Abstract
Strict environmental legislation and increasing energy demand are the foretokens to edify the need of an eco-friendly sustainable method of metal extraction from spent catalysts. The present study practices an eco-friendly chelation technology for extraction of nickel from spent catalyst. Experiments were carried out with chelating agents diethylenetriaminepentaacetic acid (DTPA) and nitrilotriacetic acid (NTA), in addition to our previous published data on (S,S)-ethylenediamine disuccinic acid [(S,S)-EDDS] and ethylenediaminetetraacetic acid (EDTA), by adopting the “one variable at a time” (OVAT) approach. However, OVAT does not take into account the interaction effect of process variables and cannot be regarded as an accomplished approach for multivariate analysis. Therefore, an attempt has been made to explore the applicability of response surface methodology (RSM) in order to minimize experimental efforts and provide elaborated quality of information to overcome the limitations of OVAT. A Box-Behnken design (BBD) was coupled with RSM to optimize the process parameters using statistical analysis. The quantitative evaluation of the data shows that the molar ratio of chelating agent to Ni, solid to liquid ratio, and time, exert a significant influence on Ni extraction. A high value of regression coefficient (R2>0.97) substantiated a well fitted quadratic empirical model. The present optimization study may prove utilitarian for hydroprocessing industries intending to extract metals from spent catalyst which may reduce processing costs also.
1 Introduction
Although, the world has been reeling from financial crisis, with repercussions being felt throughout the economy, developing countries in transition are expected to drive the world economy, growing on average by 5.6% in 2012 [1]. It has been forecasted by the Organization for Economic Co-operation and Development (OECD) center that developing countries will contribute nearly 60% of world GDP by 2030 [2]. The energy sector acts as one of the most important building blocks to ameliorate the economy of a developing nation. It is reported in the Annual Energy Outlook 2012 that world energy consumption will increase by 47% from 2010 through 2035 [3]. Total primary energy consumption is anticipated to grow by 0.3% per year from 2010 to 2035 and will reach up to 106.9 quadrillion Btu in 2035 [3]. It is worthy to mention here that catalysts serve as important contributors in most of the chemical industries. The global catalyst market is driven by factors such as growing consumption of petroleum products, environmental regulations, technology developments and emerging trends. As the catalyst consumption increases because of a steady increase in the processing of heavy feedstock, the generation of spent catalysts also increases significantly. In 2010, about 568.64 tons of spent catalyst was generated from Indian refinery industries, while it was just 203 tons in 2007 [4]. These spent catalysts contain large amounts of heavy metals and cannot be disposed off directly into the environment. Spent catalysts of the fertilizer industry contain 18–28% nickel, whereas the major metal content in spent catalysts of the refining industry contain 4–12% Mo, 15–30% Al, 1–5% Ni and 0–4% Co [5]. Spent hydrotreating catalysts and spent hydrorefining catalysts have been added to the hazardous waste list, due to the presence of toxic heavy metals in the catalyst [6]. Disposal of these spent catalysts requires compliance with stringent environmental regulations. Thus, increasing concerns of the environment and declining access of resources create an imperative need to recover the valuable metals from spent catalysts.
Metal extraction from spent catalysts can be achieved by hydrometallurgical and pyrometallurgical methods. Hydrometallurgy involves the use of aqueous chemistry, occurring at temperatures lower than the boiling point of the aqueous media. It includes leaching techniques (acid, caustic, and bio) and the chelation concept, whereas smelting, calcinations, roasting etc., high temperature techniques, are included in pyrometallurgy methods. Chelation technology is considered one of the most efficacious methods to extract heavy metals. Extraction of metals with chelating agents does not require high temperatures and the chelating agents can be reused after the extraction of metals. Due to its recoverable nature, this process is more economical than any other process. Polyphosphate, hydroxycarboxylate and aminopolycarboxylate are the predominantly used chelating agents for metal extraction. Aminopolycarboxylate chelating agents e.g., ethylenediaminetetraacetic acid (EDTA), nitrilotriacetic acid (NTA), diethylenetriaminepentaacetic acid (DTPA), (S,S)-ethylenediamine disuccinic acid [(S,S)-EDDS], are used preferably over other kinds of chelating agents because of a better binding capacity than polyphosphates and their better sequestering ability over a wide pH range than hydroxycarboxylate chelating agents [7]. Chelation technology has been a subject of a few studies as an eco-friendly approach for extraction of heavy metals from spent catalysts. Goel et al. [8] reported a recovery of 95% nickel using EDTA under optimum reaction conditions at a reaction time of 8 h, while Vuyyuru et al. [9] achieved 95% Ni extraction at a reaction time of 4 h, using EDTA under autogeneous reaction conditions at a high temperature.[S,S]-EDDS was considered an effective solvent for activity regeneration of three way automobile catalysts [10].
Experimental studies for optimization of process parameters, generally, consider the “one variable at a time” (OVAT) approach, keeping all other variables constant. The classical approach OVAT does not take into account the interaction effect of process parameters. Therefore, response surface methodology (RSM) is receiving attention these days for the multivariate analysis of response. RSM is a statistical optimization approach for designing experiments, model development and optimization of process parameters [11]. Literature suggests that RSM has been successfully employed for modeling and optimization of the laser transmission joining process [12], for determination of bio available concentrations of selenium and arsenic in estuarine sediments [13], for determination of vanadium in petroleum condensate [14], for optimization of phenolic compounds extraction from wheat [15], for optimization of lipase catalyzed esterification reactions [16] and in many other applications. The most frequently used design matrices for RSM are central composite design (CCD), Box-Behnken design (BBD) and Doehlert design. An independent quadratic BBD has been used in various studies. BBD is a rotatable second-order design which is based on three-level incomplete factorial designs. It can be viewed as a spherical, revolving cube as shown in Figure 1A, where each independent variable is placed at one of three equally spaced values. This particular layout of the three levels allows the number of design points and number of polynomial coefficients to increase at the same rate.
Box-Behnken design matrix: (A) a cubical design and (B) interlocking 22factorial design.
The other view of BBD depicts the three interlocking 22 factorial designs and a central point as shown in Figure 1B [17–19]. BBD provides the major emphasis on middle points of the variable’s range and is considered a “safer” way, since the reaction does not need to perform at the extreme points. BBD does not contain an embedded factorial design and requires fewer treatment combinations than CCD [20]. It has been frequently used for the determination of optimum conditions for extraction processes [21], for chromatography methods [22], for the preparation of nanoparticles [23] and many more processes.
The objective of the present study is to statistically determine the relationship among various process variables of the metal extraction process, to obtain a high percentage of nickel. BBD was coupled with RSM to analyze and optimize the process variables with a reduced number of experiments. The estimated regression models between response variables and process parameters were developed for different chelating agents. Experiments were also carried out with DTPA and NTA, in addition to our previous published data on EDTA and [S,S]-EDDS [24], to validate the suitability of RSM over a wide range of chelating agents and process conditions.
2 Materials and methods
2.1 Methodology of chelant assisted extraction process
The spent catalyst, obtained from a primary reforming unit of a fertilizer industry, contained 25% Ni, a small quantity of MgO as the promoter and the support material α-Al2O3. The catalyst was calcined in the presence of atmospheric oxygen flow in a temperature controlled muffle furnace for 5 h at 550°C, for the removal of carbon material. The extraction experiments were carried out at atmospheric pressure under reflux conditions in a similar manner as described in the literature [24]. All of the experiments were carried out at a stirring speed of 500 rpm, with a particle size of 100 µm, to eliminate the effect of mass transfer. The individual effect of various process parameters was studied to optimize the extraction efficiency. Spectrophotometric analysis was performed to determine the concentration of Ni in the aqueous solution of NiSO4. The process flow chart of chelant assisted metal extraction process is shown in Figure 2.
Process flow chart of chelant assisted Ni extraction process.
The amount of extracted Ni is calculated as the ratio of Ni present in aqueous solution to Ni present in the spent catalyst. The reactions associated with the chelation and dechelation process are shown in Eqs. (1)–(3).
2.2 Multivariate experimental design
The multivariate technique RSM was employed to investigate the influence of various process variables on a metal extraction process. RSM is used to investigate an optimum region, response surface behavior in the optimum region, estimation of the optimal condition and verification of the model [25]. A BBD for five independent variables at each of the three levels was employed to fit the model. The total number of experiments for the BBD matrix was calculated using Eq. (4):
where N=number of experiments; k=number of process variables; and Cp=center points. Eight axial and 32 factorial experimental runs were carried out with a sixfold repetition of center points, to evaluate the pure error. The five independent variables for the extraction process were molar concentration of the chelating agent to Ni (MR) (X1), solid to liquid ratio (S/L) (X2), reaction time (X3), reaction temperature (X4) and reaction pH (X5). Three levels for each ith independent variable (-1, 0, +1) were coded as xi according to the following transformation equation (5):
where Xi=the real value of the ith independent variable; Xi0=its value in the central point of the interval; and ΔXi=the step change. Center points were assigned at zero level for all of the variables. Table 1 lists the coded and experimental values of the independent variables at three levels.
Experimental range and levels of independent process variables.
| Independent variables | Units | Notation | Parameter values | ||
|---|---|---|---|---|---|
| Low level (-1) | Middle level (0) | High level (+1) | |||
| MR | X1 | 1.2 | 3.6 | 6 | |
| S/L | g/ml | X2 | 1/10 | 1/20 | 1/30 |
| Reaction temperature | °C | X3 | 50 | 70 | 90 |
| Reaction time | h | X4 | 2 | 6 | 10 |
| Reaction pH | X5 | 7 | 9 | 11 | |
It is necessary to fit a mathematical model according to the levels of variables studied to describe the behavior of the response. Extraction of Ni (%) using [S,S]-EDDS (Sigma Aldrich, New Delhi, India) (Y1), EDTA (Fisher Scientific, New Delhi, India) (Y2), DTPA (Merck, New Delhi, India) (Y3) and NTA (Merck, New Delhi, India) (Y4) were chosen as the observed response and were statistically analyzed using Design Expert 8.0.7.1 software. The correlation between the observed response(s) and the set of independent variables in RSM was incurred by second order polynomial, as shown in Eq. (6) [26, 27].
where Y=the predicted response, i.e., (%) Ni extraction using a chelating agent; β0=a constant; βi=the linear coefficient; βii=the squared coefficient; βij=the cross-product coefficient; and k=number of factors (k=5). The results were analyzed by the least-square method and response surfaces were generated to find the optimum reaction conditions for the extraction process. It is assumed in the least-square method that errors present a random distribution profile with a zero mean and a common unknown variance and that these errors are independent of each other [28].
2.3 Statistical analysis
The quality of the model fit was evaluated by the application of analysis of variance (ANOVA) and coefficients of determination (R2). The statistical significance of the model was determined by the application of Fisher’s F-test. The three-dimensional response surfaces and the two-dimensional contour plots were also developed to describe the individual and the cumulative effects of the variables, as well as the mutual interactions between the independent variables and the dependent variables. The optimal values of studied process parameters predicted by RSM were calculated using Design Expert 8.0.7.1.software.
3 Results and discussion
3.1 Regression models of response
The BBD with the observed response(s) for each set of reaction parameters is summarized in Appendix A.1. The data of the design matrix were used for the regression analysis and for the estimation of the regression coefficients. The observed responses were correlated with independent variables employing multiple regressions through the least-square method, to fit the second order polynomial equation. Non-significant factors were eliminated using the stepwise elimination method. The mathematical models for [S,S]-EDDS, EDTA, DTPA and NTA are shown in Eqs. (7)–(10), respectively.
Ni Extraction ([S,S]-EDDS)
Ni Extraction (EDTA)
Ni Extraction (DTPA)
Ni Extraction (NTA)
It was observed that a quadratic polynomial equation could best fit the experimental data. It may also be interpreted from derived empirical models that the MR, S/L and reaction time exert a significant influence on metal extraction with a positive linear effect. A negative quadratic effect of S/L and MR is also evident as shown in Eq. (7), which depicts that excess of the chelating agent in the solution may cause a negative impact on the extraction efficiency. The empirical model for EDTA, DTPA and NTA showed the substantial mutual interaction between process variables with positive coefficients as shown in Eqs. (8)–(10). The mathematical model for NTA [Eq.(10)] with a negative coefficient of quadratic values of S/L depicts that nickel extraction efficiency may decrease at higher S/L. Experimental runs are in concordance with the mathematical observation. It was concluded from the model equations that a significant amount of nickel extraction results from a higher MR, S/L and time. Negative quadratic effects and a negative interaction between reaction parameters must also be considered.
3.2 Analysis of variance (ANOVA) of RSM model
The mathematical quadratic models cannot be considered as the only way to describe the experimental results; therefore, it is necessary to evaluate the quality of the deduced model. Adequacy of the developed mathematical models was investigated using ANOVA methods and comparing variation sources with Fisher distribution (F-test). The significance test of regression model and analysis of observed responses are shown in Table 2 for [S,S]-EDDS. The significance of each of the coefficients of the empirical model, as given in Eq. (7), were investigated on the basis of the probability of error value (p-values), which is considered as an indicator of the interaction strength of each parameter [29]. The p-values <10-4 shown in Table 2, depict confidence of the proposed model in predicting the response values. The level of significance (α) was chosen to be 0.05 for the statistical analysis. A high F-value (152.12) for the model implies that the upper bound for p-values is less than the level of significance. Thus, the mathematical model is significant at 95% confidence interval. The p-value (0.16) for the lack-of-fit test was found to be nonsignificant, which is considered favorable for the data fitting model. A lower value of 4.67% of coefficient of variation indicated a higher degree of precision and reliability of the experiments. It was also observed that the p-value is quite high for the parameter pH when compared to other independent variables, which shows the trivial effect of pH on Ni extraction using the chelating agent [S,S]-EDDS. It may be concluded from the data matrix and statistical observations that even at neutral pH, [S,S]-EDDS can give a better Ni extraction as compared to other chelating agents.
ANOVA analysis of regression model developed for (S,S)-ethylenediamine disuccinic acid {[S,S]-EDDS}.
| Source | Sum of squares | df | Mean square | F value | p-value prob>F | |
|---|---|---|---|---|---|---|
| Model | 15062.50 | 10 | 1506.2 | 152.12 | <0.0001 | Significant |
| X1 | 3717.95 | 1 | 3717.95 | 375.50 | <0.0001 | |
| X2 | 4195.80 | 1 | 4195.80 | 423.76 | <0.0001 | |
| X3 | 1753.51 | 1 | 1753.51 | 177.10 | <0.0001 | |
| X4 | 3410.56 | 1 | 3410.56 | 344.45 | <0.0001 | |
| X5 | 87.89 | 1 | 87.89 | 8.876 | 0.0052 | |
| X1X2 | 70.56 | 1 | 70.56 | 7.12 | 0.0114 | |
| (X1)2 | 660.72 | 1 | 660.72 | 66.73 | <0.0001 | |
| (X2)2 | 753.20 | 1 | 753.20 | 76.07 | <0.0001 | |
| (X3)2 | 123.63 | 1 | 123.63 | 12.48 | 0.0012 | |
| (X4)2 | 1157.41 | 1 | 1157.41 | 116.89 | <0.0001 | |
| Residual | 345.54 | 35 | 9.90 | |||
| Lack of fit | 324.40 | 30 | 10.8 | 2.44 | 0.1616 | Not significant |
| Pure error | 22.13 | 5 | 4.42 | |||
| Cor. total | 15409.04 | 45 | ||||
| Std. dev. | 3.14 | R-squared 0.98 | ||||
| Mean | 67.37 | Adj R-squared 0.97 | ||||
| C.V. % | 4.67 | Pred R-squared 0.96 | ||||
| PRESSa | 761.45 | Adeq precision 43.57 | ||||
aPredicted residual of sum of squares.
Regression coefficients were also calculated to corroborate the adequacy of the models. A high regression coefficient (R2=0.98), along with the high value of the adjusted regression coefficient (Radj2=0.97) for Eq. (7) indicates the potential of the empirical model to satisfactorily describe the system behavior within the investigated range of studied parameters [26]. The high value of R2(0.98) indicates 98% compatibility of experimental data with the data predicted by the model [30]. The value of the predicted R2 (0.96) is comparable with the adjusted R2 (0.97) and validates the precision of the deduced model. Adequacy of the mathematical models was also investigated for EDTA, DTPA and NTA. The ANOVA results of these three chelating agents are given in Appendix A.2, A.3 and A.4. It was observed from ANOVA that the F-values for EDTA, DTPA and NTA are high, with the probability values (pmodel >F)=10-4; therefore, the models could be considered highly significant. The predicted R2 value is in agreement with the adjusted R2 value. The value of adequate precision was 40.6 for EDTA, 40.4 for DTPA, and 37.9 for NTA, which indicates an adequate signal. It may be concluded from ANOVA that the linear and quadratic effects are significant, whereas the interaction effects could not show a considerable effect on the extraction efficiency.
ANOVA results were presented graphically in Figures 3A and 3B. A plot between the observed and predicted response for all four chelating agents is shown in Figure 3A. All of the points were equally scattered around the diagonal line, which demonstrate their low discrepancies. The predicted values of the extraction efficiency lie within ±10% of the experimentally observed response. Good correlations between observed and predicted data with R2 of 0.98, 0.97 and 0.97 were obtained for Ni extraction using EDTA, DTPA and NTA, respectively.
(A) Graphical representation of comparative analysis of observed and predicted (%) Ni extraction efficiency of different chelating agents and (B) plot of internally studentized residuals vs. predicted values.
Residuals provide a clear picture for any discrepancies in fitting to the model and are important to test the assumption of constant variance [26]. Therefore, residuals vs. fitted plots were drawn for all four chelating models, to determine the possibility of non-linearity and unequal error variances. The plot of internally studentized residuals vs. predicted values is presented in Figure 3B for [S,S]-EDDS, EDTA, DTPA and NTA. The pattern shown in Figure 3B indicates that residuals are randomly scattered around zero deviation, which validates the assumption of linearity. All of the points lie in the acceptable range between -3 to 3, thus the existence of systematic errors may be cast away.
The layouts of the plots presented in Figures 3A and 3B lead to the conclusion that the predicted model given in Eqs. (7)–(10) can successfully establish the correlation between the process parameters of the studied systems.
3.3 Localization of optimum conditions
The interaction effect of process variables on Ni extraction was investigated using 3D response surface plots and corresponding contours. 3D response surface plots give a visualization of the influence of process parameters on the response and help to get the optimal values of independent variables for maximization of extraction efficiency [31]. Figures 4A–4D show the 3D response surfaces and their corresponding contours for Ni extraction using the chelating agent [S,S]-EDDS. These plots demonstrate the effect of two process factors on the response at a time, while other factors are kept at level zero.
![Figure 4 (A) Mutual effect of MR (X1) and S/L (X2) on (%) Ni extraction using (S,S)-ethylenediamine disuccinic acid {[S,S]-EDDS}, (B) mutual effect of MR (X1) and temperature (X3) on (%) Ni extraction using [S,S]-EDDS, (C) mutual effect of MR (X1) and reaction time (X4) on (%) Ni extraction using [S,S]-EDDS and (D) mutual effect of MR (X1) and reaction pH (X5) on (%) Ni extraction using [S,S]-EDDS.](/document/doi/10.1515/gps-2013-0016/asset/graphic/gps-2013-0016_fig4.jpg)
(A) Mutual effect of MR (X1) and S/L (X2) on (%) Ni extraction using (S,S)-ethylenediamine disuccinic acid {[S,S]-EDDS}, (B) mutual effect of MR (X1) and temperature (X3) on (%) Ni extraction using [S,S]-EDDS, (C) mutual effect of MR (X1) and reaction time (X4) on (%) Ni extraction using [S,S]-EDDS and (D) mutual effect of MR (X1) and reaction pH (X5) on (%) Ni extraction using [S,S]-EDDS.
The 3D response surface plot and the contour plot in Figure 4A gives the (%) extraction of Ni as function of MR and S/L. The spherical shape of the 3D surface and concentric contour plots demonstrate the equable effect of the Ni extraction efficiency. The process factor MR (X1) was varied from 1.2 to 6 and S/L (X2) was varied from 1/10 to 1/30. It may be seen from Figure 4A that much less extraction efficiency (27.8%) was predicted when both factors are at low levels. Ni extraction efficiency increased with the increase in MR values and reached up to 63% at MR=6 (high level) and S/L=1/10 (low level). Increase in S/L from 1/10 to 1/30 with a constant value of MR=6, also demonstrates a nearly 35% increase in (%) Ni extraction efficiency. This behavior is acceptable, because an excess amount of chelating agent moves the reaction in a forward direction, due to the reversible nature of the chelation reaction and extraction efficiency increases. Maximum 90% Ni extraction was achieved at MR=6 and S/L=1/30, while other parameters were set at zero level.
The mutual effect of MR (X1) and reaction temperature (X3) is shown in Figure 4B with the help of response surfaces and contours. As the temperature increases from 50°C to 90°C, Ni extraction increases from 71.4% to 92.2% at MR=6. It may be related primarily with the effect of enhanced kinetics at higher temperatures. The 3D response surface demonstrates that about 92% Ni extraction is predicted at MR=6 and 90°C, whereas only 41% extraction of Ni was predicted at MR=1.2 and 50°C. The BBD matrix (Appendix A.1) indicates <4% deviation between experimental observations and predicted responses at extreme values of MR and S/L (when both factors are at a low level or a high level).
A synergistic effect was observed between the MR and reaction time by plotting the contours and response surface as shown in Figure 4C. Beyond MR=3.6, Ni extraction becomes nearly constant, therefore, MR=3.6 was found to be favorable for a reaction time in the range of 2–10 h. Maximum 94.6% Ni extraction was predicted at MR=6 within 10 h of reaction time, when all other parameters are kept constant at level zero. An experimental investigation was also carried out at the same operating conditions and 91.5% Ni extraction efficiency was observed. Thus, the experimental observations were found to be comparable with the predicted extraction efficiency. The optimum extraction of Ni was 79% at center points, i.e., at level zero.
The response surface of Figure 4D illustrates the effect of pH and MR together on Ni extraction. Reaction pH was varied from neutral (pH=7) to alkaline (pH=11) conditions. It may be seen that with the increase in reaction pH, the extraction efficiency only increased from 53% to 57.3% at a low level of MR value and from 83% to 87.7% at a high level of MR value. 3D response surface, therefore, leads to the conclusion that the effect of solution pH did not have a significant effect on Ni extraction from spent catalysts using [S,S]-EDDS, however, at a higher value of MR, a little mutual interaction effect of pH and MR was observed, which helps to increase the extraction efficiency. Straight lines in the contours also explain the nonsignificant behavior of reaction pH.
It was observed from 3D response surfaces and its contours that a mutual interaction effect exists between reaction parameters, although it is not as significant as the individual effect of reaction parameters. Literature [11] suggests that the contours with concentric elliptical ridges within the design boundary are the essential requirement to show a significant mutual interaction effect. However, in the present study, the semi-elliptical (Figures 4A–4C) or non-elliptical (Figure 4D) contour plots confirm the less significant mutual interaction effect of process parameters. ANOVA studies also suggest that the p-values are not significant (p>10-4) for an interaction effect of reaction parameters.
It is evident from response surfaces that the most predominant factors for Ni extraction process are MR, S/L and reaction time. MR values in the range of 3.6–4.8, S/L in the range of 1/15–1/20 and reaction time of 6–7 h, reaction temperature 90°C and reaction pH between 8 and 9 were considered as the most favorable reaction conditions, on the basis of response surfaces. Plots of 3D response surface and corresponding contours for Ni extraction using EDTA, DTPA and NTA were also obtained in similar patterns. The layouts signify the similar optimal regions for EDTA, DTPA and NTA to extract Ni from the spent catalyst.
3.4 Experimental validation of the RSM model
The feasibility of the regression models was also investigated experimentally for all four chelating agents in a stirred batch reactor. The effect of various process parameters on Ni extraction from spent catalysts was studied to achieve the optimum reaction conditions. The effect of the molar ratio of the chelating agent to Ni (MR) was studied by varying the MR from 1.2 to 6 with a temperature of 90oC, S/L 1/20 (g/ml) and reaction time 8 h. It was observed from Figure 5A, that with an increase in MR from 1.2 to 6, there is a gradual increase in the (%) extraction of Ni. The experimental results are in concordance with the RSM model, where a positive effect was observed with the increase in MR. [S,S]-EDDS showed the highest (%) extraction of Ni, whereas the least (%) Ni extraction was obtained using NTA. These behavioral differences can be attributed to the complexation characteristics of the ligand. [S,S]-EDDS is a hexadentate ligand. Structurally, all six ligand groups of the [S,S]-EDDS fill the available coordination sites of the metal ion. Since all major functional groups of the [Ni-EDDS] are involved in the complexation, fewer sites are available for bridging the catalyst surface [32]. It may also be seen from Figure 5A, that extraction of Ni was not appreciable at low concentrations (MR=1.2) of the chelating agents. A minimum chelant concentration of 0.16 M is required to form 1:1 chelant/Ni complexes. A chelant concentration above the stoichiometric amount is desired for higher extraction of Ni from the catalyst. The amount of extracted Ni was seen to be less affected beyond MR=3.6, therefore, all experiments were performed further at MR=3.6.
(A) Effect of molar ratio of chelating agent to Ni on (%) Ni extraction.
(Reaction conditions: S/L=1/20 g/ml, temperature=90°C, pH=11, reaction time=8 h), (B) effect of solid to liquid ratio on (%) Ni extraction. (Reaction conditions: MR=3.6, temperature=90°C, pH=11, reaction time=8 h), (C) effect of reaction temperature on (%) Ni extraction. (Reaction conditions: MR=3.6, S/L=1/20, pH=11, reaction time=8 h) and (D) effect of pH on (%) Ni extraction. (Reaction conditions: MR=3.6, temperature=90°C, S/L=1/20, reaction time=8 h).
The influence of the S/L on Ni extraction was investigated at a wide range of S/L ranging from 1/10 to 1/30 and the results are illustrated in Figure 5B. It may be observed from Figure 5B that the removal efficiency of the chelating agents was increased with an increase in S/L. The extraction efficiency achieved an asymptotic value at 1/20 (g/ml) S/L, and therefore, 1/20 (g/ml) was considered as the optimum reaction condition for Ni extraction. The optimum value was found to be in agreement with RSM results. A slight decrease in the extraction efficiency of nickel was observed with a stoichiometric excess of NTA after S/L=1/25. This may be due to the possibility of readsorption. It could be explained by the fact that NTA is quadridentate and Ni complexation leaves two metal coordination sites available for interaction with the catalyst surface, thus readsorption of the [Ni-NTA]- could occur through the coordination sphere of the metal. Elliott and Brown [32] performed experiments for Pb extraction from soil, and found that the stability constant for the formation of [Pb-NTA]- was nearly seven orders of magnitude less than that of [Pb-EDTA]2-, due to structural differences between these two chelating agents. The other possible reason for the decrease in extraction efficiency is that excess of NTA added to the system leads to the formation of the 1:2 complex [Ni-(NTA)2]-4. The stereochemical arrangement of M(NTA)2-4 species is such that two uncoordinated NTA functional groups are exposed to the solution [32].
The temperature of the reaction was also varied from 50oC to 100oC, while keeping all other reaction parameters constant in Figure 5C. It may be depicted from Figure 5C that with the increase in temperature, Ni extraction efficiency also increases. This may be related primarily with the effect of enhanced kinetics due to higher operating temperatures. It was observed that at a lower temperature, classical aminopolycarboxylate chelating agents (EDTA, DTPA) could not extract a significant amount of metal, whereas [S,S]-EDDS, the new biodegradable chelating agent, showed about 68% Ni extraction at a reaction temperature of 50°C [24]. The RSM results justified the observation where the response surface did not show a significant increase in Ni extraction with respect to an increase in temperature, as shown in Figure 4B. The extraction efficiency of different chelating agents was investigated for variations in reaction time, for a period of 2 to 10 h. It was observed that the equilibrium time was 8 h for EDTA to achieve a higher extraction, while for [S,S]-EDDS and DTPA, the equilibrium stage was achieved at nearly 6 h reaction time. NTA recovered the equilibrium condition at 10 h reaction time. The differences in reaction time to achieve equilibrium may be correlated with the structural differences.
Experiments were conducted to investigate the effect of pH on metal extraction for different values of pH varying from 7 to 11, and the results are presented in Figure 5D. At pH=11, a higher metal extraction was achieved over EDTA and DTPA than at pH=7. The biodegradable chelating agent [S,S]-EDDS did not show any significant increase in extraction efficiency with increase in pH from pH=7 to pH=11. This deviation in pH range may be explained on the basis of the stability constant of metal chelate complexes [24]. An anion induced adsorption takes place when any ligand (L) forms a bridge between the surface and the metal species, as demonstrated by Anson [33].
ML+C-L↔C-L-ML
where M represents the metal (Ni) and C is the catalyst surface. Under these conditions, the metal should exhibit a pH-dependent adsorption behavior similar to that of the unbound ligand [34]. The alumina surface carries a net positive charge below pH=9, so electrostatic attraction for the anionic ligand formed plays a role in its adsorption. Catalyst components become negatively charged above pH=9 and electrostatic repulsion exists, therefore Ni recovery increases at a higher pH. The optimized process conditions obtained from the stirred batch experiments were MR=3.6, S/L=1/20, reaction time 6–8 h for different chelating agents, reaction temperature 90°C and pH=7 for [S,S]-EDDS, and pH=9 for the other three chelating agents.
The chelation process is economically favorable, due to complete recycling of chelating agents. Dechelation was performed to recover the chelating agents by lowering the pH of the solution. It may be mentioned here, that a large number of binding sites of DTPA provides a stronger binding capacity than EDTA, therefore DTPA showed a higher extraction efficiency compared to EDTA, but cannot be recovered easily due to a complicated process to break the coordination bond. Recovery of [S,S]-EDDS, EDTA and NTA was nearly 96%, while recovery of DTPA was not more than 40%, due to its higher complexing abilities. Dechelation of Ni-DTPA was performed at 4°C for 3 days and DTPA recovery was improved up to 80–82%. Therefore, a stronger binding capacity but poor recovery of DTPA makes it a less preferable chelating agent for metal extraction.
3.5 Validation of RSM model using numerical optimization
Numerical optimization was performed to obtain the best possible solutions for maximization of Ni extraction. Ramp function graphs were drawn to test the desirability of each factor and response, as well as the combined desirability. A number of different possible combinations of reaction parameters were obtained using Design Expert 8.0.7.1 software to optimize the extraction efficiency. Figure 6 shows the ramp function for one of those best possible combinations of reaction parameters, to investigate the optimized solution for chelating agent [S,S]-EDDS. It was seen from Figure 6 that optimized solution has good desirability under the given set of constraints.
Ramps of the numerical optimization.
The dot point on the ramp shows both the horizontal movement of the point and goal satisfaction. The desirability is a method to find the optimum region and completely dependent on the relative difference of lower and upper limit with respect to the observed optimum. The desirability of the process optimization was found to be 0.907. Different treatment combinations were chosen randomly among the series of best possible combinations of reaction parameters and confirmation experiments were performed. Experiments were carried out at the nearest possible practical values and results were found according to the constraint set for the desired extraction efficiency. Reproducibility was investigated by performing each experiment three times and the actual value of the response was calculated as the average of these three trials. Table 3 suggests that a maximum of 5% deviation was observed between the experimental results and the predicted extraction efficiency. All of the validation checks for the empirical model substantiated a well fitted regression model for the extraction process.
Validation test results.
| Run no. | X1 | X2 | X3 | X4 | X5 | Y1 | Y2 | Y3 | Y4 | |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Predicted | 3.6 | 26.1 | 87.9 | 9.8 | 7.2 | 92.2 | 64.9 | 72.9 | 51.4 |
| Actual | 3.6 | 26.0 | 90 | 10 | 7 | 88.6 | 61.6 | 74.3 | 52.6 | |
| Deviation | 4.0% | 3.6% | 1.9% | 2.2% | ||||||
| 2 | Predicted | 2.9 | 29.5 | 88.6 | 9.3 | 10.6 | 90.5 | 73.3 | 81.9 | 54.8 |
| Actual | 2.9 | 30 | 90 | 9 | 10 | 88.8 | 76.8 | 83.2 | 56.7 | |
| Deviation | 1.9% | 4.5 % | 1.5% | 3.3% | ||||||
| 3 | Predicted | 3.1 | 25.3 | 90 | 7.6 | 7 | 84.5 | 60.7 | 68.7 | 46.8 |
| Actual | 3.1 | 25 | 90 | 7 | 7 | 86.7 | 62.4 | 65.9 | 49.1 | |
| Deviation | 2.5% | 2.7% | 4.2% | 4.7% | ||||||
| 4 | Predicted | 4.0 | 18.9 | 82.4 | 8.7 | 9 | 83.2 | 58.9 | 67.8 | 45.1 |
| Actual | 4.0 | 20 | 80 | 9 | 9 | 85.1 | 61.2 | 70 | 47.3 | |
| Deviation | 2.2% | 3.7% | 3.1% | 4.6% | ||||||
| 5 | Predicted | 2.7 | 29.6 | 90 | 10 | 8 | 91.2 | 74.1 | 82.9 | 56 |
| Actual | 2.7 | 30 | 90 | 10 | 8 | 90 | 76.4 | 84 | 57.1 | |
| Deviation | 1.3% | 3.0% | 1.3% | 1.9% | ||||||
| 6 | Predicted | 4.7 | 21.2 | 73.8 | 7.2 | 11 | 81.6 | 55.9 | 62.3 | 41.2 |
| Actual | 4.7 | 20 | 70 | 7 | 11 | 78.1 | 53.4 | 60 | 39.7 | |
| Deviation | 4.4% | 4.6% | 3.7% | 3.8% |
4 Conclusion
It has been demonstrated that a BBD coupled with the RSM allows the optimization of the process variables for the extraction of Ni metal from the spent catalyst, using different chelating agents. A quadratic model was found to be best fitted to the data, with a regression coefficient of >0.96 for all four chelating agents. The predicted models, as well as experimental validation, suggest that pH does not play a significant role in Ni extraction using [S,S]-EDDS, whereas, a high S/L may cause a negative effect on the extraction efficiency in the case of NTA. A mutual interaction effect exists between reaction parameters, but it is not as significant as the individual effect of reaction parameters. Response surface plots of the model showed that the Ni extraction from spent catalysts depends predominantly on the molar ratio of chelating agent to Ni, S/L and reaction time. MR=3.6, S/L=1/22.3, reaction temperature of 85.2°C, reaction time of 8.2 h and pH=7.3 were obtained as the optimum reaction conditions for all four chelating agents, with a desirability factor of 0.907. The proposed design correlations may prove to be a useful tool in designing the pilot plant/commercial plant for the extraction of heavy metals using environmental friendly chelation technology.
About the authors
Ms. Garima Chauhan graduated from University of Rajasthan (India) with a bachelor’s degree (BTech) in Biotechnology in 2006. She obtained her master’s degree (MTech) from Malaviya National Institute of Technology (India) in 2009. Currently, she is pursuing her PhD at the Indian Institute of Technology Delhi since 2009, under the supervision of Professor K.D.P. Nigam and Professor K.K. Pant. Her research interests are focused on the development of eco-friendly technology for extraction of heavy metals from waste material and reusability of the extracted metals in other useful applications.
Dr. K.K. Pant is currently Professor in the chemical engineering department at the Indian Institute of Technology (IIT) Delhi. He has published more than 70 papers in various international research journals, and is currently actively engaged in research areas of the application of heterogeneous catalysis for cleaner and greener environment, catalytic hydrocarbon conversion process, hydrogen fuel generation from renewable energy sources, and biomass utilization. He is a reviewer of several international journals and also a member of several chemical engineering professional societies.
Professor Krishna Nigam has been on the faculty of Chemical Engineering, Indian Institute of Technology Delhi since 1976 and served in various capacities. He has authored 127 research publications which have received more than 1700 citations in peer reviewed international journals. He served on the Editorial Board of many Elsevier journals, e.g., Chemical Engineering Research and Design, Chemical Engineering and Processing-Process Intensification, Education for Chemical Engineer and Guest Editor for special issues of Chemical Engineering Science. Currently, he is on the International Advisory Panel of Chemical Engineering Science. He has the distinction of being the 1st Indian Chemical Engineer recipient of the prestigious Humboldt Research Award of Germany since its inception in 1982.
Appendix
Box-Behnken Design matrix with observed and predicted (%) nickel extraction efficiency of different chelating agents (in coded values).
| Std. order | Experimental details | Observed responses | Predicted response | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| X1 | X2 | X3 | X4 | X5 | Y1 | Y2 | Y3 | Y4 | Y1P | Y2P | Y3P | Y4P | |
| 1 | -1 | -1 | 0 | 0 | 0 | 29.3 | 18.9 | 28.7 | 13.3 | 27.8 | 18.1 | 27.1 | 11.4 |
| 2 | 1 | -1 | 0 | 0 | 0 | 61.5 | 51.8 | 54.5 | 39.1 | 62.6 | 48.5 | 55.5 | 37.7 |
| 3 | -1 | 1 | 0 | 0 | 0 | 67.6 | 45.2 | 49.1 | 34.1 | 64.4 | 43.5 | 50.3 | 33.5 |
| 4 | 1 | 1 | 0 | 0 | 0 | 89.2 | 72.8 | 79.1 | 60.9 | 90.4 | 73.9 | 78.7 | 62.9 |
| 5 | 0 | 0 | -1 | -1 | 0 | 41.5 | 18.6 | 29.7 | 11.2 | 44.5 | 18.4 | 33.6 | 13.2 |
| 6 | 0 | 0 | 1 | -1 | 0 | 66.8 | 39.5 | 49.7 | 33.1 | 65.3 | 37.6 | 53.8 | 36.2 |
| 7 | 0 | 0 | -1 | 1 | 0 | 70.9 | 46.8 | 59.1 | 39.8 | 73.7 | 47.8 | 62.2 | 42.8 |
| 8 | 0 | 0 | 1 | 1 | 0 | 90 | 70.7 | 81 | 58.8 | 94.5 | 69.8 | 86.4 | 55.8 |
| 9 | 0 | -1 | 0 | 0 | -1 | 53.1 | 31.3 | 37.4 | 18.8 | 51.2 | 36.4 | 40.7 | 20.3 |
| 10 | 0 | 1 | 0 | 0 | -1 | 82.1 | 60.8 | 65.1 | 43.6 | 83.4 | 61.8 | 63.9 | 45.5 |
| 11 | 0 | -1 | 0 | 0 | 1 | 56.8 | 39.8 | 45.9 | 23.9 | 55.8 | 41.8 | 48.7 | 26.7 |
| 12 | 0 | 1 | 0 | 0 | 1 | 86.7 | 68.9 | 73.9 | 50.9 | 88 | 67.2 | 71.9 | 51.9 |
| 13 | -1 | 0 | -1 | 0 | 0 | 39.7 | 21.4 | 31.2 | 13.3 | 41 | 23.4 | 36.1 | 17.6 |
| 14 | 1 | 0 | -1 | 0 | 0 | 69.2 | 55.9 | 58.7 | 44.8 | 71.4 | 53.8 | 64.5 | 47 |
| 15 | -1 | 0 | 1 | 0 | 0 | 64.9 | 41.7 | 44.8 | 28.8 | 61.8 | 39 | 53.3 | 30.6 |
| 16 | 1 | 0 | 1 | 0 | 0 | 90.3 | 68.8 | 76.2 | 59.8 | 92.2 | 69.4 | 81.7 | 60 |
| 17 | 0 | 0 | 0 | -1 | -1 | 53.1 | 23.4 | 28.2 | 15.1 | 56.2 | 24.9 | 31.1 | 17.1 |
| 18 | 0 | 0 | 0 | 1 | -1 | 80.2 | 63.4 | 68.9 | 46.9 | 85.4 | 60.7 | 66.7 | 46.7 |
| 19 | 0 | 0 | 0 | -1 | 1 | 56.9 | 32.3 | 41 | 24.9 | 60.8 | 30.3 | 39.1 | 23.5 |
| 20 | 0 | 0 | 0 | 1 | 1 | 87.1 | 64.7 | 75.1 | 55.7 | 90 | 66.1 | 74.7 | 53.1 |
| 21 | 0 | -1 | -1 | 0 | 0 | 43.5 | 26.9 | 31.1 | 19.8 | 36.5 | 22.3 | 33.2 | 16.4 |
| 22 | 0 | 1 | -1 | 0 | 0 | 68.9 | 48.3 | 56.1 | 43.1 | 71.7 | 46.5 | 59.4 | 41.6 |
| 23 | 0 | -1 | 1 | 0 | 0 | 62.7 | 40.5 | 49.8 | 33.4 | 60.3 | 36.7 | 53.4 | 29.4 |
| 24 | 0 | 1 | 1 | 0 | 0 | 89.9 | 71.3 | 79.1 | 56.4 | 92.5 | 73.3 | 83.6 | 54.6 |
| 25 | -1 | 0 | 0 | -1 | 0 | 33.8 | 9.9 | 20.8 | 9.2 | 35 | 6.6 | 14.9 | 8.7 |
| 26 | 1 | 0 | 0 | -1 | 0 | 67.2 | 38.8 | 50 | 29.8 | 65.4 | 37 | 48.5 | 30.9 |
| 27 | -1 | 0 | 0 | 1 | 0 | 63.7 | 34.9 | 51.1 | 34.3 | 64.2 | 32.4 | 45.7 | 31.1 |
| 28 | 1 | 0 | 0 | 1 | 0 | 91.5 | 73.4 | 76.1 | 67.6 | 94.6 | 72.8 | 78.9 | 67.7 |
| 29 | 0 | 0 | -1 | 0 | -1 | 60.2 | 37.8 | 45.5 | 28.7 | 62.2 | 41.7 | 49.7 | 29.6 |
| 30 | 0 | 0 | 1 | 0 | -1 | 82.6 | 54.6 | 65.4 | 43.2 | 83 | 57.3 | 66.9 | 42.6 |
| 31 | 0 | 0 | -1 | 0 | 1 | 64.7 | 43.8 | 52.6 | 37.8 | 66.8 | 47.1 | 57.7 | 36 |
| 32 | 0 | 0 | 1 | 0 | 1 | 86.4 | 62.4 | 72.1 | 49.2 | 87.6 | 62.7 | 74.9 | 49 |
| 33 | -1 | 0 | 0 | 0 | -1 | 50.1 | 30.6 | 33.4 | 21.7 | 52.7 | 33.1 | 37.1 | 21.5 |
| 34 | 1 | 0 | 0 | 0 | -1 | 81.9 | 61.4 | 67.8 | 51 | 83.1 | 63.5 | 65.5 | 50.9 |
| 35 | -1 | 0 | 0 | 0 | 1 | 55 | 38.1 | 40.7 | 24.8 | 57.3 | 38.5 | 45.1 | 27.9 |
| 36 | 1 | 0 | 0 | 0 | 1 | 88.5 | 68.6 | 73.1 | 57.8 | 87.7 | 68.9 | 73.5 | 57.3 |
| 37 | 0 | -1 | 0 | -1 | 0 | 29.8 | 9.8 | 24 | 8.9 | 33.5 | 13.3 | 21.1 | 7.2 |
| 38 | 0 | 1 | 0 | -1 | 0 | 63.8 | 31.9 | 41.6 | 27.8 | 65.7 | 31.9 | 44.3 | 25.8 |
| 39 | 0 | -1 | 0 | 1 | 0 | 59.7 | 36.6 | 58.2 | 26.7 | 62.7 | 42.3 | 56.7 | 30.2 |
| 40 | 0 | 1 | 0 | 1 | 0 | 89.7 | 72.3 | 76.6 | 60 | 94.9 | 74.5 | 79.9 | 62 |
| 41 | 0 | 0 | 0 | 0 | 0 | 81 | 56.6 | 65.8 | 49.8 | 78.5 | 56.8 | 65.9 | 49.5 |
| 42 | 0 | 0 | 0 | 0 | 0 | 76.9 | 56 | 67.2 | 51.3 | 78.5 | 56.8 | 65.9 | 49.5 |
| 43 | 0 | 0 | 0 | 0 | 0 | 75.6 | 59 | 64.1 | 47.8 | 78.5 | 56.8 | 65.9 | 49.5 |
| 44 | 0 | 0 | 0 | 0 | 0 | 77 | 54.7 | 68.4 | 49.4 | 78.5 | 56.8 | 65.9 | 49.5 |
| 45 | 0 | 0 | 0 | 0 | 0 | 79.4 | 58.7 | 64.9 | 50.2 | 78.5 | 56.8 | 65.9 | 49.5 |
| 46 | 0 | 0 | 0 | 0 | 0 | 81.8 | 56.2 | 65.2 | 48.7 | 78.5 | 56.8 | 65.9 | 49.5 |
ANOVA analysis of regression model developed for ethylenediaminetetraacetic acid (EDTA).
| Std. dev.a | 3.03 | R-squared | 0.98 |
| Mean | 46.83 | Adj R-squaredd | 0.97 |
| C.V.b % | 6.48 | Pred R-squarede | 0.96 |
| PRESSc | 821.22 | Adeq precisionf | 40.61 |
aStandard deviation, bCoefficient of variation, cPredicted residual sum of squares, dAdjusted R-squared, ePredicted R-squared, fAdequate precision.
ANOVA analysis of regression model developed for diethylenetriaminepentaacetic acid (DTPA).
| Std. dev.a | 2.97 | R-squared | 0.97 |
| Mean | 54.8 | Adj R-squaredd | 0.97 |
| C.V.b % | 5.42 | Pred R-squarede | 0.95 |
| PRESSc | 764.52 | Adeq precisionf | 40.49 |
aStandard deviation, bCoefficient of variation, cPredicted residual sum of squares, dAdjusted R-squared, ePredicted R-squared, fAdequate precision.
ANOVA analysis of regression model developed for nitrilotriacetic acid (NTA).
| Std. dev.a | 3.00 | R-squared | 0.97 |
| Mean | 37.75 | Adj R-squaredd | 0.96 |
| C.V.b % | 7.96 | Pred R-squarede | 0.95 |
| PRESSc | 636.02 | Adeq precisionf | 37.99 |
aStandard deviation, bCoefficient of variation, cPredicted residual sum of squares, dAdjusted R-squared, ePredicted R-squared, fAdequate precision.
References
[1] Global economic outlook: world economic situation and prospects, 2012. Available at: http://www.un.org/en/development/desa/policy/wesp/wesp_current/2012wesp_prerel.pdf. [Accessed on 26 July, 2012].Search in Google Scholar
[2] Perspectives on global development, June 2010. Available at: http://www.oecd.org/document/14/0,3746,en_2649_33959_45467980_1_1_1_1,00.html. [Accessed on 26 July, 2012].Search in Google Scholar
[3] Annual Energy Outlook, U.S. energy information administration. DOE/EIA-0383. 2012. Available at: http://www.eia.gov/forecasts/aeo/pdf/0383%282012%29.pdf. [Accessed on 8 December, 2012].Search in Google Scholar
[4] Sustainable development report, Bharat Petroleum Corporation Limited, 2009–2010. Available at: http://www.bharatpetroleum.com/pdf/BPCL_SDR_2009_10.pdf. [Accessed on 23 March, 2012].Search in Google Scholar
[5] Siemens RE, Jong BW, Russell JH. Conserv. Recycling 1986, 9, 189–196.10.1016/0361-3658(86)90107-4Search in Google Scholar
[6] Federal Register, 1998. Hazardous waste management system – Identification and history of hazardous waste; petroleum refining process waste. Federal Register 63, 151, 42032–42048.Search in Google Scholar
[7] Howard JL, Shu J. Environ. Pollut. 1996, 91, 89–96.Search in Google Scholar
[8] Goel S, Pant KK, Nigam KDP. J. Hazard. Mater. 2009,171, 253–261.Search in Google Scholar
[9] Vuyyuru KR, Pant KK, Krishnan VV, Nigam KDP. Ind. Eng. Chem. Res. 2010, 49, 2014–2024.Search in Google Scholar
[10] Subramanian B, Christou SY, Efstathiou AM, Namboodiri V, Dionysioua DD. J. Hazard. Mater. 2011, 186, 999–1006.Search in Google Scholar
[11] Li Y, Cui F, Liu Z, Xu Y, Zhao H. Enzyme. Microb. Tech. 2007, 40, 1381–1388.Search in Google Scholar
[12] Wang X, Song X, Jiang M, Li P, Hu Y, Wang K, Liu H. Opt. Laser Technol. 2012, 44, 656–663.Search in Google Scholar
[13] Lopes Wda L, Santelli RE, Oliveira EP, de Carvalho MdFB, Bezerra MA. Talanta 2009, 79, 1276–1282.10.1016/j.talanta.2009.05.039Search in Google Scholar PubMed
[14] Santelli RE, Bezerra MA, Freire AS, Oliveira EP, de Carvalho MdFB. Fuel 2008, 87, 1617–1622.10.1016/j.fuel.2007.08.008Search in Google Scholar
[15] Liyana-Pathirana C, Shahidi F. Food Chem. 2005, 93, 47–56.Search in Google Scholar
[16] Manohar B, Divakar S. Process Biochem. 2004, 39, 847–853.Search in Google Scholar
[17] Souza Anderson S, dos Santos Walter NL, Ferreira Sergio LC. Spectrochimica Acta Part B. 2005, 609, 737–742.Search in Google Scholar
[18] Massart DL, Vandeginste BGM, Buydens LMC, Jong SD, Lewi PJ, Smeyers JV. Handbook of Chemometrics and Qualimetrics, Part A. Elsevier: Amsterdam, 2003.Search in Google Scholar
[19] Aslan N, Cebeci Y. Fuel 2007, 86, 90–97.10.1016/j.fuel.2006.06.010Search in Google Scholar
[20] NIST/SEMATECH e-Handbook of Statistical Methods, Available at: http://www.itl.nist.gov/div898/handbook/. [Accessed on 15 June, 2012]Search in Google Scholar
[21] Turner C, Whitehand LC, Nguyen T, McKeon T. J. Agric. Food. Chem. 2004, 52, 26–32.Search in Google Scholar
[22] Stafiej A, Pyrzynska K, Ranz A, Lankmayr E. J. Biochem. Bioph. Meth. 2006, 69, 15–24.Search in Google Scholar
[23] Mohamed R, Mkhalid I, Azaam E. Mater. Sci. App. 2011, 2, 981–987.Search in Google Scholar
[24] Chauhan G, Pant KK, Nigam KDP. Ind. Eng. Chem. Res. 2012, 51, 10354–10363.Search in Google Scholar
[25] Tanyildizi MS, Ozer D, Elibol M. Process Biochem. 2005, 40, 2291–2296.Search in Google Scholar
[26] Juretic D, Kusic H, Koprivanac N, Bozic AL. Water Res. 2012, 46, 3074–3084.Search in Google Scholar
[27] Myer RH, Montgomery DC. Response Surface Methodology: Process and Product Optimization using Designed Experiment, 2nd ed., John Wiley and Sons: New York, USA, 2002.Search in Google Scholar
[28] Bezerra, MA, Santelli RE, Oliveira EP, Villar LS, Escaleira LA. Talanta 2008, 76, 965–977.10.1016/j.talanta.2008.05.019Search in Google Scholar PubMed
[29] Pradhan S, Madankar CS, Mohanty P, Naik SN. Fuel 2012, 97, 848–855.10.1016/j.fuel.2012.02.052Search in Google Scholar
[30] Mohanty P, Majhi S, Sahu JN, Pant KK. Ind. Eng. Chem. Res. 2012, 51, 4843–4853.Search in Google Scholar
[31] Almeida LC, Garcia-Segura S, Bocchi N, Brillas E. Appl. Catal. B – Environ. 2011, 103, 21–30.Search in Google Scholar
[32] Elliott HA, Brown GA. Water, Air, Soil Pollut. 1989, 45, 361–369.Search in Google Scholar
[33] Anson FC. Acc. Chem. Res. 1975, 8, 400–407.Search in Google Scholar
[34] Benjamin MM, Leckie JO. Environ. Sci. Tech. 1981, 15, 1050–1057.Search in Google Scholar
©2013 by Walter de Gruyter Berlin Boston
This article is distributed under the terms of the Creative Commons Attribution Non-Commercial License, which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
