Central composite design optimization of Bismarck Dye oxidation from textile effluent with Fenton’s reagent

Coloured wastewater is discharged in a huge amount from textile industry as a result of using synthetic dyes that causes threat water pollution. A simple sol–gel technique was used to synthesis haematite nanocrystals as a source of photo-Fenton’s reagent. The current work is dealing with the mineralization of Bismarck Brown dye (BBD) using such reagent, and the process parameters were optimized using response surface methodology. Central composite design of the two most important variables of the Fenton’s reagent process (Fe3+ and H2O2 doses) was employed for the experimental design, and BBD removal was maximized. The suggested model investigated that the dye removal increases with the increasing of both reagents up to a certain limit then decreases afterwards. The reduction percentages of colour and chemical oxygen demand under the optimum operating parameters (37.21 mg/L of the haematite nanocrystal and 412.31 mg/L of H2O2) are found to be 76.5% and 73%, respectively. Analysis of variance and the high regression coefficient (98%) confirm a good estimation of the experimental values by polynomial regression model. Therefore, the exploration of using haematite nanocrystals as a Fenton’s reagent source for treating textile effluents is recommended from this study.


Introduction
Due to the rapid development in the textile dying industries and the extensive use of synthetic dyes, wastewater discharges from such industry represent a major environmental pollution issue. Azo dyes are considered the major classes of synthetic dyes since up to 70% of the dyes used for commercial purposes are azo dyes (Hussein 2012). According to the presence of the azo bonds (-N=N-) such dyes are classified to mono-, dia-, tri-azo, etc. In addition, some of those dyes or their precursors are from toxic aromatic amines complexes. Thus, both human and the marine life are seriously affected due to their toxic and highly carcinogenic effect. To develop the environmental performances, different treatment technologies have been applied in order to mange those aquatic effluents (Abu Amr et al. 2014;Kumaravel et al. 2015;Tony et al. 2018). Over the last decades, to overcome the increasing demand of human being for water, several techniques have been developed for industrial wastewater treatment. According to the literature, for instance, some researchers applied non-reagent physical techniques for treatment such as filtration, coagulation and precipitation (Zhao et al. 2009;Kotut et al. 2011;Ashour and Tony 2017). Additionally, some studies reported the application of adsorption methods for wastewater treatment (Parker et al. 2012;Ashour et al. 2014;Torrades and García-Montano 2014;Tony 2019). However, such techniques are not efficient or adequate for treatment since they transfer the pollutant from one phase to another without mineralizing them besides the high cost of maintenance and treatment. Thus, there is an urgent need to improve feasible alternative techniques for treating the industrial wastewater in an economic and viable approach that environmentally friendly and does not produce a second pollution. Recent studies suggested the significant use of the oxidation techniques such as Fenton's reagent as an 1 3 108 Page 2 of 9 attractive alternative method for treating dyed wastewater (Chen et al. 2012;Torrades and García-Montano 2014).
Recent advances in the wastewater treatment oxidation are the innovative nanomaterials application. Nanomaterials revealed a promising improvement in the hazards and toxic contaminant remediation and removal (Babuponnusami 2014;Tony et al. 2016). Nanomaterial is applied previously as an alternative way to enhance the Fenton's reagent oxidation technique. Nanomaterials as a source of Fenton's reagent have many applications in wastewater treatment with various pollutants, for instance, humic acids (Nie et al. 2010), chlorophenol (Lu et al. 2002) and phenol (Prucek et al. 2002). However, according to the literature, nanocrystals as a source of Fenton's reagent have not been applied so far for treating BBD effluents.
Response surface methodology (RSM), which is a combination of mathematical and statistical procedures, is appropriate for modelling and evaluating the several variables influencing the responses even in the existence of involved interactions. To date, the traditional optimization researches of wastewater processes have focused on one-variable-at a time methodology. However, this study does not consider the cross-effects of the considered variables. Thus, this time consuming method with a lack of accurate optimum values (Torrades and García-Montano 2014;Tony et al. 2011;Mourabet et al. 2017). The main objective of RSM application in the treatment can result in improving the Fenton's reagent process by determining the optimum variables values and maximizing the response with a decrease in the experiments numbers. Also, the extensive literature examination demonstrates none of researchers examined the Fenton's parameters optimization using CCD for BBD removal. In this respect, the present investigation introduces the application of two-variable central composite design (CCD) for BBD removal using photo Fenton reagent to evaluate the colour (C) removal and COD reduction in the BBD wastewater as dependent variables. The two selected variables applied in this study are nanocrystals dose and hydrogen peroxide reagent as independent variables. Subsequently, the mathematical models correlating C and COD removals of the two variables were established.

Wastewater
The textile dye Bismarck Brown (BBD), with a molecular formula of (C 21 H 24 N 8 ·2HCl), its molecular weight is 461.39 g/mol and water solubility is 11 g/L at 25 °C is the target pollutant in this study. Real textile dying wastewater samples were collected from a service of dying facility of Jeans garments using Bismarck Dye in Menoufia Governorate, Shebin El-Kom City, north of Egypt between the period of February and April 2018. Wastewater was collected subsequent the washing facilities conducted after the re-dying process to remove the extra dyes. The main characteristics of wastewater are BBD load is 13.28 mg/L, and COD 1240 and the Suspended Solids (SS) are 23 mg/L with an effluent pH of 8.0.

Synthesis and Characterization of Fe 2 O 3 nanocrystals
Nano-crystalline Fe 2 O 3 from aqueous solutions of iron chloride precursor was synthesized using a simple sol-gel technique, and the final product is attained after calcination. The synthesized nanocrystals by this method were confirmed using X-ray diffraction (XRD) patterns. IR spectroscopy using FTIR was used for identification of functional groups and confirmed such type of nanocrystals. Also, the surface morphology was performed using transmission electron microscope (TEM), which found the particle size is varied from 6.1 to 18.3 nm as previously found in our preliminary experiments.

Photochemical reactor and procedures
A lab-scale batch mode test of the experimental set-up is shown in Fig. 1. A 12 W UV lamp inside a glass sleeve is immersed in a glass container. Firstly, nanocrystals are added to 100 mL of wastewater solution with pH 3.0 (the optimum pH according to our preliminary investigation). Thereafter, the reaction is initiated by adding H 2 O 2 reagent. The reagents are well mixed in the BBD aqueous solution, and then, the solution is poured to the photoreactor container. All experiments were conducted at room temperature at stirring speed of 800 rpm and pH 3.0. After 10 min of reaction time treatment, the substrate of the treated wastewater samples is subjected to COD and colour removal analysis.

Central composite experimental design (CCD)
Based on response surface methodology (RSM), central composite experimental design (CCD) technique was chosen to conduct the optimization of the experimental conditions for BBD removal from aqueous solution using photo-Fenton reagent. RSM is a valuable statistical tool that maximizes the response surface affected by operating variables based on a few sets of experiments within a chosen range. RSM is applied widely in the optimization of operating variables (Bhatia et al. 2007), experimental conditions of advanced oxidation process, adsorption processes and biological techniques (Raquel et al. 2015), etc. In this study, CCD is applied for optimization of operating parameters, namely nanocrystals Fe 2 O 3 dose and H 2 O 2 reagent concentration for maximizing BBD removal. The leading objectives of BBD reduction are based on colour removal and COD reduction efficiencies of the dyed wastewater. In order to evaluate the process, five levels of the two main factors: nanocrystals (E 1 ) and H 2 O 2 (E 2 ) doses, were chosen as shown in Table 1.
After conducting the experiments, the actual experimental data aimed at developing the model (shown in Table 2) were analysed through least-square method; thus, the correlation of the independent variables and the responses was estimated by the following second-order polynomial model (Khuri and Cornell 1996): where f i is the predicted response for C and COD removal, respectively; β o , β i , β 2 and β ii , are the model coefficient and e 1 and e 2 are the independent variables.
RSM was applied to the experimental data using statistical analysis software (SAS) to predict the models through regression analysis and analysis of variance (ANOVA) (SAS 1990). Three-dimensional response surface and two-dimensional contour plots were created by applying Mathematica software (V 5.2) to visualize the interactive effects of the independent factors. Optimum region was also located based on the main parameters in the overlay plot.

Calculation methods
BBD dye removal from wastewater was calculated according to the following equations: where f 1 and f 2 were the colour (C) removal and chemical oxygen demand (COD) removal; C o , COD o and C t , COD t were the initial and finial values of C and COD.

Mathematical model build-up
In the present study, photo-Fenton reagent based on the nanocrystalline iron oxide was investigated to treat BBD wastewater under different operational conditions: Fe 3+ (11.72-68.28 mg/L) and H 2 O 2 (117.2-682.8 mg/L) at the operating pH 3.0. A central composite design (CCD) was applied as it is very commonly used in the RSM technique and the simple linear or quadratic models can be related to the response values (Colour and COD removals in this study) (Tak et al. 2015). The thirteen experiments are necessary for the design matrix of the two factors (Fe 3+ and H 2 O 2 doses) with five levels of each factor, which were considered to cover the experimental domain. Values of the experimental responses variable (C and COD reduction) are given in Table 2 along with their predicated responses values. The predicted responses of C and COD removal values were obtained from the second-order polynomial equation fitting using SAS software [Eqs. (4), (5)]. The response values (f 1 ) for %C removal and (f 2 ) for %COD removal are presented by the following equations to attain the interaction between dependent and independent variables: where f 1 and f 2 are %C and %COD removal of BBD from wastewater, E 1 and E 2 are the codified values of the operating parameters nanocrystals iron oxide and hydrogen peroxide concentrations, respectively. At this point, the experimental (2) (4) responses for %C and %COD removals are presented in Table 2. The corresponding predicted responses obtained from the models [Eqs. (4) and (5)] are very close as shown in Fig. 2, demonstrating that both values were accurate and reliable.
In order to test the suitability and the significance of the model, analysis of variance (ANOVA) was conducted using SAS software. The obtained regression equations of the polynomial models from ANOVA indicated that the correlation coefficient (R 2 ) for f 1 and f 2 are 98.47 and 98.80, respectively. This indicates a good correlation between experimental and predicted models. In fact, the quality of the model is well fitted when R 2 is suggested to be at least 0.80 (Arslan-Alaton et al. 2009;Tony and Bedri 2014). In the ANOVA test of the two models presented in Table 3, the Fisher variance ratio, F value is used as a measure to estimate the way that the factors explains the variation in the mean of data (Montgomery 1991). F value is 89.98 for C removal and 115.73 for COD removal. Simultaneously, the given very low probability (P-value) (0.0001) is considered significant when it is (P model > F) < 0.05 in surface response analysis (Montgomery 1991;Tony et al. 2015). Moreover, the coefficient of variation, CV, recorded low values of 3.39% C removal model and 4.08% for COD removal model that confirms the reliability and good precision of the experimental data as recommended by Kuehl (2000). Therefore, it is concluded that the model fitted well and highly significant and the estimated factors effects are real.

Effect of the investigated variables on colour and COD removal
Subsequently, the verification of significance of the model, the suitability of the variables and interactions is identified following the Student t test (Montgomery 1991).
Based on the F values (Table 3), nanocrystals dose (E 1 ) has significant effect (P < 0.001) and (P < 0.0002) on the responses of C and COD removal (f 1 , f 2 ), respectively as it had a larger coefficient than hydrogen peroxide concentration (E 2 ). The interaction effects between E 1 and E 2 , were moderate. Positive coefficient of H 2 O 2 (E 2 ) indicated the linear effect of increase in f 1 and f 2 . However, iron nanocrystals dose (E 2 2 ) had a negative effect along with and quadratic terms (E 1 2 and E 2 2 ) that decreases f 1 and f 2 responses. The positive effect of the interaction term (E 1 E 2 ) increases f 1 response. Furthermore, in order to attain a better understanding of the results, 3D surface graph and 2D contour plots were used to attain a better understanding of the results, the dependent (f 1 and f 2 ) and independent variables (E 1 and E 2 ) using Matlab (7.11.0.584) software. The interaction effects of Fe 3+ nanocrystals and H 2 O 2 concentrations on the C removal efficiency presented in the 2D contour plots in Fig. 3b that simulated from Eq. 4. As seen from Fig. 3a, b, the surface and contour plots showed that the %C removal was increased with increasing both the nanocrystals and H 2 O 2 reagent doses. However, further increase in both reagents doses after a certain point, the reaction is retarded. This phenomenon indicates that the optimum operating conditions of Fe 3+ and H 2 O 2 according to colour removal are 37.21 and 412.31 mg/L, respectively. Based on the maximum COD reduction, the optimum values are 36.24 and 421.88 mg/L, respectively (as seen in Fig. 4a, b). Notably, this is obviously near to the optimal points with that from colour removal test.
From Figs. 3 and 4, it can be estimated that the key factor for the colour and COD removals is the nanocrystals Fe 3+ concentration. However, even the hydrogen peroxide reagent effect is also important, but its effect is at a lower level. A negative effect in both responses, C and COD reduction, is observed. This effect is related to the scavenging effect of radicals. However, optimal doses are needed for photo-Fenton reaction to occur. This could be illustrated by the increase in H 2 O 2 in the presence of iron crystals resulted in the increase in·OH radicals generation. These radicals are the main responsible of the Fenton's reagent as those radicals are attaching the organic pollutant and mineralizing

Parameters optimization within designated constraints and verification
The optimal values for two independent variables, naoncrsyals of iron oxide and hydrogen peroxide reagents for colour and COD removals were attained using the numerical optimization future of Mathematica software (V 5.2).  Additionally, a graphical optimization presented in Fig. 5 displays the area (yellow portion zone) of feasible response values in the factors space. All the factors and responses with the particular high limit and low limit experimental region are significant and fit the standards that described the optimum values as stated in Table 4. The optimum values for Fe 3+ and H 2 O 2 are 37.21 and 412.31 mg/L with maximum response of 75.9 for C removal. By comparing the optimum values in Fig. 3 with Fig. 4, it can be observed that the location of optimal operating conditions from C test was rational from COD test.
Ultimately, to validate the model adequacy, an additional experiment was performed using those amounts. The experimental colour and COD removals are 76.5% and 73%, compared to 75.9% and 72.6%, respectively, for the predicted values (Table 5). This confirms that the CCD based on RSM was a powerful tool for locating the accurate optimal values of the independent variables.

Conclusion
The performance of photo-Fenton technique for colour removal and COD reduction in BBD wastewater was investigated. Among this study, by observing the effects of interactions among the variables on colour, C, and COD removal efficiencies, the RSM experimental conditions were optimized. The results demonstrated significant effects of the two operating variables (iron oxide nanocrystals and hydrogen peroxide concentrations) as well as their interactive effects on the two responses. CCD was applied to locate the optimum removal rate for both responses. According to ANOVA, the proposed second-order polynomial model   is verified and thus the model is accepted. Notably, the obtained high R 2 values for the two models are attained (98.47 and 98.80 for C and COD removals, respectively) which confirms the high accuracy of the models. The predicted colour and COD removals are 75.9 and 72.6% at the optimum operating conditions at pH 3.0. Furthermore, the economical experimental values of the two responses maximized to 76.5% and 73% for C and COD models, respectively.