Research Article  Open Access
Optimization of Bi_{2}O_{3}, TiO_{2}, and Sb_{2}O_{3} Doped ZnOBased LowVoltage Varistor Ceramic to Maximize Nonlinear Electrical Properties
Abstract
In ZnObased low voltage varistor, the two essential features of microstructure determining its nonlinear response are the formation Bienriched active grain boundaries as well as a controlled ZnO grain size by secondary spineltype phases. Besides, the microstructure and phase composition are strongly affected by the dopant concentration during sintering process. In this study, the optimal dopant levels of Bi_{2}O_{3}, TiO_{2}, and Sb_{2}O_{3} to achieve maximized nonlinear electrical property (alpha) were quantified by the response surface methodology (RSM). RSM was also used to understand the significance and interaction of the factors affecting the response. Variables were determined as the molar ratio of Bi_{2}O_{3}, TiO_{2}, and Sb_{2}O_{3}. The alpha was chosen as response in the study. The 5level3factor central composite design, with 20 runs, was used to conduct the experiments by ball milling method. A quadratic model was established as a functional relationship between three independent variables and alpha. According to the results, the optimum values of Bi_{2}O_{3}, TiO_{2}, and Sb_{2}O_{3} were obtained 0.52, 0.50, and 0.30, respectively. Under optimal conditions the predicted alpha (9.47) was calculated using optimal coded values from the model and the theoretical value is in good agreement with the value (9.43) obtained by confirmation experiment.
1. Introduction
Multicomponent semiconducting ceramics that are based on ZnO and other cationic oxides exhibit highly nonlinear currentvoltage characteristics [1]. ZnO varistor materials are mainly composed of ZnO as well as a balancing mix of other oxides (e.g., oxides from cobalt, nickel, bismuth, titanium, and antimony). These materials combined produce highly nonOhmic properties and, therefore, are widely used as surge protecting elements in the electrical transmissions and circuits against lightning or temporary overvoltages [2]. With the popularity of miniaturization and integration of electronic devices, lowvoltage varistors are in rapidly growing need [3]. The symmetric nonlinear currentvoltage (IV) response of varistor ceramics is closely related to thin insulating layers around the successive ZnO grains [4]. Microstructural studies show that the thin insulating layers around the successive ZnO grains consist of bismuthrich second phases which promote the formation of potential barriers to electrical conduction at the ZnO homojunctions [5, 6]. Besides, secondary spineltype phases are also located at these layers which can control the microstructure development during sintering process [7]. TiO_{2} is a spinelforming dopant which is commonly used as a grain growth enhancing additive in the production of lowvoltage ZnObased varistor ceramics. ZnO grains’ growth in the presence of TiO_{2} was explained by reduction insulating layer mobility due to the presence of spinel particles. As another spinelforming dopant, Sb_{2}O_{3} is typically added to produce finegrained highvoltage varistor by the presence of a spinel phase at the insulating layer [8]. Each of the dopants plays a distinctive role in the subtle tuning of the final nonlinear characteristics of the varistor ceramics and cannot be omitted. Also, the proper ratios among the dopants have to be set in order to obtain the required electrical performance of the varistor ceramics through the process of microstructure development [9]. As a multivariate case, a pervious study has demonstrated that the phase equilibrium formation and as a consequence the microstructure development are strongly influenced by the TiO_{2}/Bi_{2}O_{3} in lowvoltage varistor ceramics and the Sb_{2}O_{3}/Bi_{2}O_{3} ratio in highvoltage varistor ceramics. This report reveals the synergistic interaction of dopant in order to achieve a desired microstructure with specific electrical properties [10]. The traditional onefactoratatime approach to optimization is timeconsuming and incapable of reaching a true optimum because of taking no account of comprehensive effect of factors [11]. On the contrary, the statistical experimental design allows simultaneous investigation of the effects of several process variables, as well as their actual significance on the considered response and possible interrelationship among them, giving maximum information with the fewest number of trial experiments [12–14]. For these purposes, optimization by RSM utilizing polynomial equations has been widely used [15]. This methodology requires minimum experimentation and time, thus providing to be far more effective and also costeffective than the traditional methods of optimization [16]. In this study, RSM was used for modeling and optimizing of molar ratio of Bi_{2}O_{3}, TiO_{2}, and Sb_{2}O_{3} as additives to achieve the maximum value of the alpha for lowvoltage varistor.
2. Experimental Procedures
2.1. Sample Preparation
The samples were fabricated using a conventional ceramic process with nominal composition of ZnO (99.99%), Bi_{2}O_{3} (99.975%), Sb_{2}O_{3} (99.6%), and TiO_{2} (99.9%) as raw powders. The raw chemicals of analytical grade produced by Alfa Aesar were weighed according to the experimental designs (Table 2) based on molar ratio [mol%]. The molar ratio was converted into weight, and electronic balance was used for precise measurement up to 10^{−4 }g. The weighed powders were mixed and milled with highresistance zirconium balls and acetone in a polypropylene jar for 24 h. The resultant slurries were dried in oven at 100°C for 8 h. The dried pastes were crushed into powders, sieved, and pressed into pellet forms of 5.0 mm radius and 0.7 thickness at 20 MPa by Specac Hydraulic Presser machine. The green compacts were sintered at 1260°C for 2 h, with heating and cooling rates of 5°C min^{−1} [17, 18] by a box furnace (CMTS Model HTS 1400) to prepare the ceramic. To determine the electrical properties, both faces of the ceramic pellets were painted by conductive silver electrode which is called varistor.
2.2. Materials Characterization
The electric fieldcurrent density (EJ) characteristics of the varistors were recorded with Keithley 236 source meter at room temperature. The varistors were scanned by DC voltage from 0 to 100 V in step size of 2.5 V. The IV data was used to calculate the alpha of the varistor by where and are the electrical fields corresponding to the current densities and at 0.1 and 1 mA/cm^{2}, respectively. The breakdown field () was determined at 1 mA/cm^{2} while the leakage current density () was determined at 80% of breakdown field.
The crystalline phase compositions of the sintered sample were identified by Xray diffractometer (XRD, PANalytical (Philips) X’Pert Pro PW3040/60, CuKα, and ) and the data were analyzed by using X’Pert High Score software. In order to make microstructural investigations, one of the surfaces of samples was ground with silicon carbide paper and polished with 1 μm diamond suspension, which led to a mirrorlike surface. Then, they were thermally etched at 1100°C for 10 min to reveal the microstructural details. To reduce charging effects and to improve the resolution of the image, the etched samples were metalized with a thin coating of gold mounted on Al stub. The surface microstructure and compositional analyses of the sintered samples were examined by VPSEM (LEO 1455) which is attached to EDX.
2.3. Experimental Design
The most popular RSM design is the central composite design (CCD) [19, 20]. A CCD has three groups of design points: (a) twolevel factorial or fractional factorial design points, (b) axial points (sometimes called “star” points), and (c) center points. CCDs are designed to estimate the coefficients of a quadratic model. The factorial points, which are one unit distance away from the center of the design space, are used to fit the linear and interaction terms. The star points can provide additional levels of the factor for prediction of the quadratic terms, and their distances from the center are α unit [19]. The α value is equal to , where n is the number of factors. In this study, is equal to three effective variables, namely, the molar ratio of Bi_{2}O_{3} (), TiO_{2} (), and Sb_{2}O_{3} (), so that the α value is 1.682. Table 1 summarizes the ranges and levels of the effective variables in five different levels (−1.682, −1, 0, +1, and +1.682) involved in the design strategy.


Accordingly, 20 experiments which determined by consisting 8() full factorial points, 6 (2n) axial points and 6 center points designed as replications to get a good estimate of experimental error (pure error). Then the design (Table 2) was performed according to the procedure in Section 2.1 and the calculated alpha (Section 2.2) is presented in column of actual value (Table 2). The completed design matrix was used for regression process by Design Expert software version 8.0.7.1 (StatEase Inc., USA).
2.4. Statistical Analysis
The multiple regression equation was used to fit the secondorder polynomial equation based on the experimental data (Table 2) as follows: where represents the predicted response, is the model intercept, , , and are linear coefficients, , , and are quadratic coefficients, , , and are cross product coefficients, and , , and are the independent variables influencing the response. Fitting (2) to the experimental data by the method of least squares (MLS) allowed estimation of all the coefficients [21]. The MLS is a multiple regression technique used to enquire the relationship between the independent and dependent variables and the estimator can be written as follows [22, 23]: where is a vector of regression coefficients; is an extended designed matrix of the coded levels of the input variables; and is a column vector of response determined according to the arrangements points into the experimental design.
Once has been determined, it is then possible to predict and so calculate a few numbers of statistical lines of evidence whose results appear in analysis of variance (ANOVA): value, probability value ( value), lack of fit, coefficient of determination squared (), adjusted squared (), and predicted squared (). These lines of evidence confirm the quality of the fitted model by conducting Fisher’s test [24, 25].
3. Results and Discussion
3.1. Model Fitting and Statistical Analysis
In the regression process, the software fitted the actual values with the polynomial equation (2) to obtain the predicted values by using RSM (Table 2). The residuals of the actual and predicted values were based on statistical analysis and model suggestion. The suggested model for the initial additives is a function of the molar ratio of Bi_{2}O_{3} (), TiO_{2} (), and Sb_{2}O_{3} () (4). Consider
The results of ANOVA were used to evaluate the statistical significance of the quadratic model in general and its terms in detail (Table 3) [26] were listed in Table 3. According to the results, the model value for the model was 28.79, implying the significance of the suggested model. The regression model can explain most of the variation in the response with the large value of . In addition, the associated value was less than 0.05, which confirmed that this model was very significant. On the base of the results of analysis of error, the lack of fit can be achieved simply by differences between residual errors and replicated error which can provide an estimate of the pure error at the center point. A value (0.1846) greater than 0.05 implies that the lack of fit is much less than the pure error which is not significant. Nonsignificant lack of fit is good and displays that the model is suitable to accurately predict the response (alpha).

The is calculated on the basis of the change in the response relative to the total variation of the response over the range of the independent factors. As shown in Figure 1, the good correlation between the observed and the predicted values () for this model indicated that this model could well explain the 96.28% of the variability in the responses. The and values should be within 0.2 of each other. The of 0.779 was in good agreement with the of 0.929 for the model. Adequate precision is a signaltonoise ratio. Ratios greater than 4 are desirable. In the present study, a ratio of 13.279 shows that the studied model is an efficient application in the design space according to standard error.
According to the terms of statistical analysis, Table 3 also shows the significance effect of each term in the model equation with respective values. A significant value (<0.05) for each term indicates an active term and a reasonable estimate of its effects. In this case, the independent variables, including , , , , and , are significant model terms. In particular, the linear effect of TiO_{2} and the quadric effect of TiO_{2} and Sb_{2}O_{3} are highly significant terms with values <0.0001. The linear and quadric effects of Bi_{2}O_{3} are significant with . In addition, none of the interaction terms had a significant effect on the response () (). In order to determine the influence rank of each term in the model, Pareto analysis was used in the form of Pareto chart according the following equation [27]: where represents the percentage effect of each factor and represents the coefficient of each term in the polynomial model. As illustrated in Figure 2, the most statistically significant variable in the model was quadratic effect of Sb_{2}O_{3} (72%), followed by linear effect of Sb_{2}O_{3} (23.5%). However, the antagonistic effect and synergistic effect of each variable on the response (alpha) can be determined by the negative and positive signs of regression coefficients in the model equation.
3.2. RSM Analysis
In order to better understand the relationship between the response (alpha) and the independent variables , , and , threedimensional surfaces plots were formed based on the model polynomial function (Figure 3). Meanwhile, the maximum achievable alpha point was determined by the partial derivative of model (4) with respect to its variables , , and as presented in following equations: By solving the system of (6), the maximum achievable alpha point was found to be 9.47. The corresponding parameters that yielded this maximum value are Bi_{2}O_{3} (0.44 mol%), TiO_{2} (0.4 mol%), and Sb_{2}O_{3} (0.29 mol%).
(a)
(b)
(c)
Figure 3(a) shows the combined effect of variables and on alpha at constant molar ratio of Sb_{2}O_{3} (0.29). As shown, the alpha increased with increasing molar ratio of Bi_{2}O_{3} between 0.3 and 0.44 and TiO_{2} between 0.3 and 0.4; however, with further increase more than the optimum (Bi_{2}O_{3}: 0.44 mol% and TiO_{2}: 0.4 mol%), the alpha decreased.
Figure 3(b) signifies the impact of changing amount of Bi_{2}O_{3} and Sb_{2}O_{3} on the alpha while the amount of TiO_{2} was fixed at 0.4. As shown in the figure, it is clear that when the molar ratio of Bi_{2}O_{3} and Sb_{2}O_{3} increases, the alpha increases firstly and then decreases after the maximum point (9.47).
The combined effect of the amount of TiO_{2} and Sb_{2}O_{3} has been presented in Figure 3(c) while the amount of Bi_{2}O_{3} was kept constant at 0.44 mol%. The results show that the maximum alpha (9.47) was recorded at 0.4 and 0.29 mol% of TiO_{2} and Sb_{2}O_{3}, respectively.
In all plots, addition of Bi_{2}O_{3} up to the optimum points increases the nonlinear coefficient (alpha) of the varistor ceramics. This shows that the bismuth content increases in the ZnO samples, someway advocating the cationic interdiffusion and the mass transport processes throughout the sintering process. It is well identified that the majority of bismuthdoped ZnO varistors have been made ready through liquid phase sintering process. On the other hand, the growth of the Bi_{2}O_{3} content leads to a growth of the potential barrier height, causing an increase in the value of alpha [28]. However, further increase beyond optimum points might cause the homogeneous segregation of the additives, causing a decline in the level of alpha [29].
As can be seen in all 3D plots with increasing amount of TiO_{2} up the optimum points, the nonlinear exponents have been enhanced. According to a previous report, TiO_{2} increases reactivity of the Bi_{2}O_{3}rich liquid phase with the solid ZnO throughout the sintering process which avoids Bi_{2}O_{3} vaporization, probably leading to increase in the nonlinear coefficient [18]. However, with more TiO_{2} doped (more than optimum points), the dopants will worsen the nonlinear electrical properties of the samples. This might be linked to the sudden reduction of the quantity of Bi_{2}O_{3} in the varistor ceramics owing to the reaction between Bi_{2}O_{3} and TiO_{2} into secondary phases, consuming out the varistorforming oxide. It was stated that, throughout sintering, Bi_{2}O_{3} provides for the development of insulating boundary layers that regulate the varistors operation [30].
As shown in all plots, the alpha increased by increasing the amount of Sb_{2}O_{3} below the optimum points. The increasing of the amount of antimony oxide favors the densification of the ceramic matrix during the sintering process of ZnObased varistors. Moreover, the higher concentration of antimony also shows a significant role in increasing the grain boundary resistivity, as it heightens the probability of segregation forming effective potential barriers and consequently increasing the level of alpha [31]. However, with more increasing beyond optimum levels the alpha values decreased. This reduction may be attributed to the decrease in the amount of Bi_{2}O_{3} since Bi is incorporated into spinel particles till Bi_{2}O_{3} ultimately fades as the amount of Sb_{2}O_{3} enhanced [32].
3.3. Confirmation Experiment
On the basis of RSM, the optimum values of the test variables were Bi_{2}O_{3}, 0.44 mol%; TiO_{2}, 0.4 mol%; and Sb_{2}O_{3}, 0.29 mol%. Under these conditions, the maximum predicted alpha was 9.47. The verification experiment was carried out by studying under optimal conditions. An actual value of 9.43 ± 0.42 was observed from real experiments, which are in close agreement with the model predicted values (9.47). The results (Table 4) indicated that verification study confirmed the predictivity of the model.

3.4. Structural and Morphological Properties of Validated Varistor
Figure 4 represents the XRD spectrum of the validated polycrystalline ceramic used as core in ZnObased lowvoltage varistor. The patterns confirmed the presence of dominant ZnO phase (ICSD code: 000050664) with hexagonal wurtzite structure and secondary phases. Many secondary phases with small peaks were detected in the ceramics at all sintering temperatures, namely, Bi_{4}Ti_{3}O_{12} (ICSD code: 000080258), Bi_{12}TiO_{20} (ICSD code: 000340097), Zn_{2}TiO_{4} (ICSD code: 000251164), Zn_{2}Ti_{3}O_{8} (ICSD code: 000130471), ZnTiO_{3} (ICSD code: 000261500)), Zn_{7}Sb_{2}O_{12} (ICSD code: 000361445), and ZnSb_{2}O_{4} (ICSD code: 000040563).
The morphology and microstructure of the sample fabricated under the optimum conditions were studied by VPSEM (Figure 5). As shown, the grains were uniformly distributed throughout the microstructure of sample with nearly no pore which resulted the high relative density and consequently the reliable nonlinear electrical property. From EDAX analysis, the Bi, Ti, and Sb were found at the grains boundaries (Figure 6).
The electrical properties of the varistor were basis of IV characteristic measurement that shows breakdown voltage was 120 V/mm with alpha 9.43. The leakage current was 0.013 mA/cm^{2}.
4. Conclusion
In this study, lowvoltage Bi_{2}O_{3}TiO_{2}Sb_{2}O_{3} doped ZnO varistor ceramics have been successfully prepared by conventional fabrication technique. The RSM was used to optimize the molar ratio of Bi_{2}O_{3}, TiO_{2}, and Sb_{2}O_{3} as additives in starting powder and a secondorder polynomial equation was developed for describing the influence of key variables on response (alpha). The results of model fitting and statistical analysis demonstrated that variables , , , , and played a key role in the response, while the linear effect of and interactions of , , and demonstrated a negligible effect on the response. The optimal conditions for the response (alpha) were found as Bi_{2}O_{3} of 0.44 mol%, TiO_{2} of 0.40 mol%, and Sb_{2}O_{3} of 0.29 mol%. The predicted alpha value determined as 9.47 under the optimal conditions and the verification study (9.43) confirmed the suitability of the predicted model. As a conclusion, RSM through the CCD model is appropriate for determining the optimal conditions for alpha, understanding the relationships among the independent and response variables, and maximizing the alpha.
Conflict of Interests
The authors declare that there is no conflict of interests regarding the publication of this paper.
Acknowledgment
The authors would like to express gratitude and acknowledgement to the Ministry of Higher Education, Malaysia, for funding this project under Research University Grant Scheme (RUGS) of Project no. 0502121878.
References
 M. Matsuoka, “Non ohmic properties of zinc oxide ceramics,” Japanese Journal of Applied Physics, vol. 10, no. 6, pp. 736–746, 1971. View at: Google Scholar
 L. M. Levinson and H. R. Philipp, “Zinc oxide varistor—A review,” The American Ceramic Society Bulletin, vol. 65, no. 4, pp. 639–646, 1986. View at: Google Scholar
 W. R. W. Abdullah, A. Zakaria, and M. S. M. Ghazali, “Synthesis mechanism of lowvoltage praseodymium oxide doped zinc oxide varistor ceramics prepared through modified citrate gel coating,” International Journal of Molecular Sciences, vol. 13, no. 4, pp. 5278–5289, 2012. View at: Publisher Site  Google Scholar
 M. A. Alim, M. A. Seitz, and R. W. Hirthe, “Complex plane analysis of trapping phenomena in zinc oxide based varistor grain boundaries,” Journal of Applied Physics, vol. 63, no. 7, pp. 2337–2345, 1988. View at: Publisher Site  Google Scholar
 S. Anas, R. V. Mangalaraja, M. Poothayal, S. K. Shukla, and S. Ananthakumar, “Direct synthesis of varistorgrade doped nanocrystalline ZnO and its densification through a stepsintering technique,” Acta Materialia, vol. 55, no. 17, pp. 5792–5801, 2007. View at: Publisher Site  Google Scholar
 M. Peiteado, J. F. Fernández, and A. C. Caballero, “Processing strategies to control grain growth in ZnO based varistors,” Journal of the European Ceramic Society, vol. 25, no. 12, pp. 2999–3003, 2005. View at: Publisher Site  Google Scholar
 M. Peiteado, J. F. Fernández, and A. C. Caballero, “Varistors based in the ZnOBi_{2}O_{3} system: microstructure control and properties,” Journal of the European Ceramic Society, vol. 27, no. 13–15, pp. 3867–3872, 2007. View at: Publisher Site  Google Scholar
 N. Daneu, N. Novak Gramc, A. Rečnik, M. Maček Kržmanc, and S. Bernik, “Shocksintering of lowvoltage ZnObased varistor ceramics with Bi_{4}Ti_{3}O_{12} additions,” Journal of the European Ceramic Society, vol. 33, no. 2, pp. 335–344, 2013. View at: Publisher Site  Google Scholar
 S. Bernik and N. Daneu, “Characteristics of ZnObased varistor ceramics doped with Al_{2}O_{3},” Journal of the European Ceramic Society, vol. 27, no. 10, pp. 3161–3170, 2007. View at: Publisher Site  Google Scholar
 S. Bernik, N. Daneu, and A. Rečnik, “Inversion boundary induced grain growth in TiO_{2} or Sb_{2}O_{3} doped ZnObased varistor ceramics,” Journal of the European Ceramic Society, vol. 24, no. 1516, pp. 3703–3708, 2004. View at: Publisher Site  Google Scholar
 J. Li, J. Peng, S. Guo, and L. Zhang, “Application of response surface methodology (RSM) for optimization of the sintering process of preparation calcia partially stabilized zirconia (CaOPSZ) using natural baddeleyite,” Journal of Alloys and Compounds, vol. 574, pp. 504–511, 2013. View at: Publisher Site  Google Scholar
 M. Shah and K. Pathak, “Development and statistical optimization of solid lipid nanoparticles of simvastatin by using 2^{3} fullfactorial design,” AAPS PharmSciTech, vol. 11, no. 2, pp. 489–496, 2010. View at: Publisher Site  Google Scholar
 J. Malakar and A. K. Nayak, “Formulation and statistical optimization of multipleunit ibuprofenloaded buoyant system using 2^{3}factorial design,” Chemical Engineering Research and Design, vol. 90, no. 11, pp. 1834–1846, 2012. View at: Publisher Site  Google Scholar
 P. R. Guru, A. K. Nayak, and R. K. Sahu, “Oilentrapped sterculia gumalginate buoyant systems of aceclofenac: development and in vitro evaluation,” Colloids and Surfaces B: Biointerfaces, vol. 104, pp. 268–275, 2013. View at: Publisher Site  Google Scholar
 A. K. Nayak and D. Pal, “Development of pHsensitive tamarind seed polysaccharidealginate composite beads for controlled diclofenac sodium delivery using response surface methodology,” International Journal of Biological Macromolecules, vol. 49, no. 4, pp. 784–793, 2011. View at: Publisher Site  Google Scholar
 J. Malakar, A. K. Nayak, and D. Pal, “Development of cloxacillin loaded multipleunit alginatebased floating system by emulsiongelation method,” International Journal of Biological Macromolecules, vol. 50, no. 1, pp. 138–147, 2012. View at: Publisher Site  Google Scholar
 S. Bernik, “Microstructural and electrical characteristics of ZnO based varistor ceramics with varying TiO_{2}/Bi_{2}O_{3} ratio,” Advances in Science and Technology, pp. 151–158, 1999. View at: Google Scholar
 Y. Abdollahi, A. Zakaria, R. S. Aziz et al., “Optimizing Bi_{2}O_{3} and TiO_{2} to achieve the maximum nonlinear electrical property of ZnO low voltage varistor,” Chemistry Central Journal, vol. 7, no. 1, p. 137, 2013. View at: Publisher Site  Google Scholar
 L. Sun and C. Zhang, “Evaluation of elliptical finnedtube heat exchanger performance using CFD and response surface methodology,” International Journal of Thermal Sciences, vol. 75, pp. 45–53, 2014. View at: Publisher Site  Google Scholar
 A. Zuorro, M. Fidaleo, and R. Lavecchia, “Response surface methodology (RSM) analysis of photodegradation of sulfonated diazo dye Reactive Green 19 by UV/H_{2}O_{2} process,” Journal of Environmental Management, vol. 127, pp. 28–35, 2013. View at: Publisher Site  Google Scholar
 M. Yolmeh, M. B. Habibi Najafi, and R. Farhoosh, “Optimisation of ultrasoundassisted extraction of natural pigment from annatto seeds by response surface methodology (RSM),” Food Chemistry, vol. 155, pp. 319–324, 2014. View at: Publisher Site  Google Scholar
 R. H. Myers and C. M. AndersonCook, Response Surface Methodology: Process and Product Optimization Using Designed Experiments, John Wiley & Sons, New York, NY, USA, 2009.
 M. Khayet, C. Cojocaru, and M. Essalhi, “Artificial neural network modeling and response surface methodology of desalination by reverse osmosis,” Journal of Membrane Science, vol. 368, no. 12, pp. 202–214, 2011. View at: Publisher Site  Google Scholar
 D. R. Cox and C. A. Donnelly, Principles of Applied Statistics, Cambridge University Press, New York, NY, USA, 2011. View at: Publisher Site  MathSciNet
 D. Freedman, R. Pisani, and R. Purves, Statistics, WW Norton, New York, NY, USA, 2007.
 Y. Yuan, Y. Gao, L. Mao, and J. Zhao, “Optimisation of conditions for the preparation of βcarotene nanoemulsions using response surface methodology,” Food Chemistry, vol. 107, no. 3, pp. 1300–1306, 2008. View at: Publisher Site  Google Scholar
 M. Zarei, A. Niaei, D. Salari, and A. Khataee, “Application of response surface methodology for optimization of peroxicoagulation of textile dye solution using carbon nanotubePTFE cathode,” Journal of Hazardous Materials, vol. 173, no. 1–3, pp. 544–551, 2010. View at: Publisher Site  Google Scholar
 Y. Romaguera, Y. Leyet, F. Guerrero, L. Aguilera, J. Pérez, and J. D. L. S. Guerra, “Influence of Bi^{3+} cation on microstructure and electrical properties of the ZnO,” Revista Cubana de Química, vol. 21, no. 3, 2010. View at: Google Scholar
 L. Cheng, G. Li, K. Yuan, L. Meng, and L. Zheng, “Improvement in nonlinear properties and electrical stability of ZnO varistors with B_{2}O_{3} additives by nanocoating method,” Journal of the American Ceramic Society, vol. 95, no. 3, pp. 1004–1010, 2012. View at: Publisher Site  Google Scholar
 H. Feng, Z. Peng, X. Fu et al., “Effect of TiO2 doping on microstructural and electrical properties of ZnOPr_{6}O_{11}based varistor ceramics,” Journal of Alloys and Compounds, vol. 497, no. 12, pp. 304–307, 2010. View at: Publisher Site  Google Scholar
 F. L. Souza, J. W. Gomes, P. R. Bueno et al., “Effect of the addition of ZnO seeds on the electrical proprieties of ZnObased varistors,” Materials Chemistry and Physics, vol. 80, no. 2, pp. 512–516, 2003. View at: Publisher Site  Google Scholar
 T. Watanabe, Y. Tokoro, Y. Sato, and S. Yoshikado, “Effects of Sb, Zr, and Y addition on the electrical characteristics of Bibased ZnO varistors,” Journal of Physics, vol. 339, no. 1, Article ID 012007, 2012. View at: Google Scholar
Copyright
Copyright © 2014 Masoumeh Dorraj et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.