Box–Behnken Design for Optimizing the Synthesis Condition and Adsorption Capability of Covalently Crosslinked Chitosan/coal Fly Ash Composite

: The growing concern regarding the environmental protection has encouraged researchers to focus their efforts on developing better and more effective possibilities of removing pollutants. In this research, a biocomposite adsorbent of covalently crosslinked chitosan-epichlorohydrin/coal fly ash (CHT-ECH/CFA) was synthesized and applied for reactive red 120 dye (RR120) removal. The CHT-ECH/CFA was characterized by BET, pH potentiometric, pH pzc , XRD, FTIR and SEM-EDX. Box – Behnken Design (BBD) was employed to assess the effects of the adsorption key parameters such as CFA loading into CTH-ECH matrix, adsorbent dose, solution pH, working temperature and contact time. The optimized CFA loading, adsorbent dose, temperature, time, and pH were observed to be 50 %, 0.07 g, 45 °C, 60 min, and pH 4 respectively. From BBD, the highest removal of RR120 removal at optimum conditions was found to be 90.2%. The results showed that adsorption performance can be modelled perfectly by both Langmuir and Freundlich isotherm models with maximum adsorption capacity of 237.7 mg/g at 45 °C. Moreover, the adsorption kinetics were well fitted to the pseudo-second order model. Based on the findings from the experiments conducted, the hybrid biocomposite adsorbent offers adequate potential for the treatment of anionic dye-polluted water.


Introduction
Industrialization and urbanisation have benefited society in different ways. However, releasing a large amount of polluted water into natural water system is like a two-edged sword [1]. The presence of colour remnants, industrial dyes, catalytic chemicals and inorganic matters in the wastewater originating from textile, printing inks, fluorescent lamp, plastic and tanning industries are regarded as serious and formidable threats [2]. The disastrous emergence in these industries has left a wide environmental footprint, with chemicals employed in their manufacturing lines potentially escaping into the adjacent waterways [3]. Therefore, as a result of the growing awareness concerning the adverse effects of dyes in water, the industries are compelled to meet the requirements of increasingly stringent regulations promulgated by governments and regulatory bodies [4].
Chitosan (CHT) is a polymer containing hydroxyl (-OH) and amino (−NH2) functional groups, these functional groups make CHT as a benchmark adsorbent due to its potential ability for removing a wide range of pollutants such as heavy metal ions [5], pharmaceutical compounds [6] and dyes [7]. The availability of the abundance of amino and hydroxyl groups in CHT backbone are not only essential for a strong electrostatic attraction for capturing various water contaminants. Consequently, these functional groups can serve as excellent binding sides to modify and upgrade CHT for any desirable purpose [8]. However, direct application of pristine CHT in wastewater treatment technologies is not preferable due to high swelling capacity and poor chemical stability in acidic environment. Therefore, the physical and chemical modification of CHT is not a dispensable option for enhancing its chemical stability and adsorptive property.
Thus, selection of the right modification method and choosing the right modifying agent is a very important issue to determine the quality and functionality of the synthesised CHT product. In this regard, there are plenty of nano scale organic/inorganic fillers that had been utilized to improve the surface area property, chemical stability, thermal stability, functionality, and adsorptive performance such as nano titania [9], Mg-doped hydroxyapatite [10], Silicon dioxide [11], Cerium oxide [12], Graphene oxide [13], Zinc oxide (ZnO) [14], nano silicon carbide [15]. On the other hand, crosslinking process is a very demanded approach to improve the chemical stability, functionality, and hydrophobicity of CHT by forming irreversible network with either (-OH) or (-NH2) functional group of CHT, and it depends on the nature and chemical structure of the crosslinked agent.
For instance, glutaraldehyde (GLA) is the benchmark dialdehyde crosslinked agent that binds covalently with CHT's amino group (-NH2) at C-3 position to form Schiff's-base system and as a result, improving mechanical performance and chemical stability of CHT [1]. In fact, amino (-NH2) group is the prime active adsorption site in the CHT backbone, especially when the CHT is very much needed for removal of anionic species of water containment. Therefore, it is hypothesized that finding an alternative crosslinker agent that interacts covalently with −OH group at C-6 in the CHT backbone may present the same inserting physicochemical properties as CHT-GLA, while avoid the shortcomings of making amino (−NH2) sites occupied to result in preferable adsorption performance [16]. To avoid any reduction in the adsorption capacity of crosslinked CHT's derivative, choosing epichlorohydrin (ECH) as an epoxy chloropropane (C3H5ClO) crosslinker agent that interacts only with the −OH group in CHT backbone will overcome this operation issue.
Coal fly ash (CFA) which is a fine ash collected from the flue gas after coal combustion is the primary solid waste emitted from thermal power stations [17]. By following the principles of the circular economy, recycling the CFA will reduce the landfill pressure and accomplish the idea of waste-free production, paving the way for sustainable reuse of an industrial byproduct [18]. As a typical aluminosilicates material with particle size varying from 1-100 µm, CFA is mainly composed of SiO2, Al2O3, Fe2O3 and CaO [19]. Thus, CFA is predominantly utilized to manufacture bricks, cement additives, concrete blocks and other construction materials [20]. In terms of high-value added material, efforts have been devoted on its possible utilizations as an adsorbent to remove various pollutants from aqueous system [21,22]. Therefore, the main objective of this study is to modify crosslinked chitosanepichlorohydrin with coal fly ash and to produce a hybrid crosslinked chitosan-epichlorohydrin/ coal fly ash (CHT-ECH/CFA) biocomposite as a promising hybrid adsorbent for removal of reactive red (RR120) dye from aqueous environment. Box-Behnken design in response surface methodology was applied to optimize the synthesis and adsorption characteristics of CHT-ECH/CFA as a function of process parameters namely CFA loading, adsorbent dose, solution pH, working temperature, and contact time. The regression analysis was performed to investigate the fitness of the experimental data to full-quadratic model with a coefficient of determination (R 2 ). Moreover, the adsorption kinetic, isotherm, thermodynamic, and mechanism of RR120 by CHT-ECH/CFA were discussed.

Synthesis of CHT-ECH/FA
1 g of CHT was dissolved in 50 mL of 5% acetic acid solution and left to stir overnight until homogenized. The CHT beads were then prepared by injecting the CHT solution into 100 ml of 0.5M NaOH. CHT beads were then washed thoroughly with distilled water to remove NaOH residues. The step of crosslinking was performed by adding 100 mL of 2% ECH to the CHT beads and stirred for 2 h at 40°C. The CHT-ECH beads were washed with distilled water and air dried. Lastly, the CHT-ECH beads were converted into powder form with a constant particle size of ≤ 250 µm and the powder was kept in an airtight container for further use in RR120 removal applications. A series of CHT-ECH/CFA composites were synthesized using the same process described above, except that the CFA particles were mixed physically with CHT before being added to the acetic acid solution. The same steps described above were adopted for crosslinking. The composite with constant loading ratio of 25% CFA and 50% CFA were labelled as CHT-ECH/CFA25 and CHT-ECH/CFA50, respectively. The materials were further ground and sieved to constant particle size of ≤ 250 µm.

Samples characterization
The morphological characteristics of the CHT-ECH/CFA were analysed using a scanning electron microscope equipped with energy dispersive X-ray analyser (SEM-EDX, Hitachi TM3030 Plus). The surface of CHT-ECH/CFA was sputter-coated with gold before imaging. The chemical structure of the CHT-ECH/CFA was identified using FTIR Spectroscopy (Perkin-Elmer, Spectrum RX1) in the range of 4000 cm -1 -450 cm -1 applying the KBr pellet technique. The X-ray diffraction (XRD) patterns of CHT-ECH/CFA were obtained by using X'Pert PRO, PANalytical Cu Kα radiation (λ=1.54 Å). The estimation of amino (-NH2) content in CHT-ECH/CFA was made using the pH-potentiometric titration method as described by Vieira and Beppu [24]. In computing the point of zero charge (pHpzc) of the CHT-ECH/CFA, the reported method Dalvand, et al. [25] was applied.

Design of experiment
Response surface methodology (RSM) coupled with Box-Behnken design (BBD) was applied to design the experiments and to perform the statistical analysis using Design expert software (version 13.0). A three-level, five-factor Box-Behnken design was applied in this study. The investigated independent variables were CFA loading (%) (A), adsorbent dose (B), solution pH (C), working temperature (D), and contact time (E) with their coded levels -1 (minimal), 0 (central point) and +1 (maximal) were listed in Table 1. Consequently, BBD contains Forty-six sets of experiments (40 factorial points and 6 centre points) which were conducted. Table 2 presents experimental design and test results of response (RR120 removal (%). The obtained results were statistically evaluated by quadratic response surface equation (2nd degree polynomial equation) as expressed in Eq. (1) [26].

Response determination (RR120 removal %)
The determination of response (RR120 removal (%)) was experimentally performed by transferring 100 mL RR120 dye solution of 50 mg/L into a 250 mL conical flask. A desired amount of the specific adsorbent (CHT-ECH, CHT-ECH/CFA25, CHT-ECH/CFA50) was weighed and transferred to the flasks. In thermostat water bath shaker (WNB7 -45, Memmert, Germany), these flasks were stirred at a constant stirring rate of 100 strokes/min. After completing the adsorption process, the filtration process for adsorbent/adsorbate system was carried out using a 0.45µm syringe filter to filter the residual solution. Lastly, the concentration of RR120 was monitored at λmax of 534 nm using HACH DR 3900, a UV-Vis spectrophotometer. The RR120 dye removal (DR %) was determined according to Eq. 2.
where C ο (mg/L) is the initial concentration of RR120 dye and C e (mg/L) is the equilibrium concentration of the RR120 dye.

Batch adsorption study
In this study, the batch adsorption mode was adopted to figure out the effect of initial RR120 concentration and contact time on the removal of RR120 using the optimal key adsorption parameters obtained from run 14 of BBD as listed in Table 2. These optimum adsorption conditions were implemented in the adsorption study using a conical flask (250 mL) with 100 mL of RR120 dye at different series of initial concentrations ranging from 50 mg/L -300 mg/L. The adsorbed amount of the RR120 was quantified using Eq. 3.
where, q e (mg/g) is the adsorption capacity at equilibrium, W (g) is the mass of the adsorbent and V (L) is the volume of RR120 dye solution.

Samples characterization
The specific surface area and other physical parameters of the CHT-ECH, CHT-ECH/CFA25, and CHT-ECH/CFA50 are summarized in Table 3. From the results, it is proven that the specific surface area (BET), Langmuir surface area, t-Plot external surface area, and total pore volume of CHT-ECH were improved gradually by increasing the loading of CFA in the polymeric matrix of CHT-ECH from 25% to 50%, signifying the significant role of the CFA in enhancing the surface area and pore diameter of CHT-ECH, and subsequently improving its adsorption property.
Opposite to specific surface area trend, the actual content of amino ( In other words, when the ratio of CFA increased in the CTH-ECT-CFA formulation, correspondingly the CHT content will be less according to the desired mixing ratio, consequently the amino (-NH2) group will be less. However, amino (-NH2) group content (22.05%) in the polymeric matrix of CHT-ECH/CFA50 is still preferable with high potential to serve as active sites for adsorption after protonation process [27].
The XRD patterns of CHT-ECH, CHT-ECH/CFA25, CHT-ECH/CFA50 are presented in Fig. 1. As can be seen from Fig. 1a, a broad hump around 2θ=20.5° is a signature peak for CHT biopolymer [28]. After loading 25% of CFA microparticles (CHT-ECH/CFA25), new distinguished peaks with different intensities are appeared at 2θ = 26°, 35°, 43° and 59° as shown in Fig.1b. This observation confirms the uniform distribution of the CFA microparticles in the polymeric matrix of CHT-ECH, and predominantly composed of crystalline phases such as hematite, quartz, and alumina. Besides that, the XRD profile after loading more 50% of CFA microparticles (CHT-ECH/CFA50) is shown in Fig. 1c. As can be seen, the spectrum of Fig.1c is quite similar to the spectrum of Fig.1b, except the intensity of the distinguished peaks become slightly sharper due to extra loading of CFA microparticles.
The functional groups present in CHT-ECH/CFA before and after uptake of RR120 were identified using FTIR analysis and displayed in Fig. 2 the surface of CHT-ECH/CFA50 appeared to be less porous and more compact with fewer cavities (Fig. 3c). This observation can be assigned to the loading of RR120 dye molecules on the CHT-ECH/CFA's surface. Moreover, the presence of S element on EDX analysis reconfirmed the successful capturing of the RR120 dye molecules on the CHT-ECH/CFA50 surface.

ANOVA and empirical model fit equations
The statistical evaluation for input effective variables (A, B, C, D, E) and their response (Y) for this study was performed using ANOVA test and shown in Table 4. It may be observed from  [23]. The low value (12.68 %) of coefficient of variation (C.V.) for RR 120 removal (%) reconfirmed the precision, higher reliability, and excellent reproducibility of the given quadratic model equation [32]. The full quadratic regression model for the RR120 removal (%) using BBD based on Eq. 1 is given in Eq. 4 along with the statistical variables obtained from the ANAOVA.

Diagnostic of the responses
The normality of the data was assessed using a normal probability plot and all the data points are relatively close to the diagonal line as shown in Fig. 4a. As the data points were distributed along the diagonal line indicating that the errors were normally distributed with mean zero and constant [34]. The relationship between predicted values and actual values of the response is expressed in Fig. 4b. The actual values are depicted by square dots in this plot, whereas the predicted values are represented by a straight line, indicating that there was good agreement between the predicted and actual values for the response. This observation shows that the BBD is well presented into the model and reconfirmed its suitability to carry out the optimization operation of the process [35]. The plots of residuals versus the predicted value of the responses are presented in Fig. 4c, while the plots of the studentized residuals versus run are presented in Fig. 4d. From Figs. 4c and 4d the adequacy can be predicted and the random scattering of the residuals around zero reveal that the model was well presented, and the constant variance assumption was appropriate [36]. Consequently, the plot of the Cook's distance is presented in Fig. 4e. Generally, the acceptable Cook's distance should be less than 1 [37]. As presented in Fig. 4e, all the observed values are less than 1. Moreover, all 46 runs are even below 0.4 and mainly equal to zero which indicates the significant effect on the predictive power of the model. Consequently, the Box-Cox plot for power transforms plot is presented in Fig. 4f. This graph indicates the most appropriate power transformation to apply for the response values. Hence, the best transformation should be closest to the ideal lambda value (the minimum point of the curve) contains 1 and lies inside the confidence interval. Thus, there is no specific transformation recommended for this BBD-RSM model [38].

Model graphs for factors plot
The model graphs for each independent factor are presented in Fig. 5. It was observed that the loading of CFA (%) in the polymeric matrix of CHT-ECH has a very desirable effect on the output response (RR120 removal (%)), and RR120 removal increased remarkably by increasing the CFA lading as shown in Fig. 5a. Moreover, it seems the most dominant input factor over other input factors that presented in Fig. 5a. Moreover, similar trend can be observed with the adsorbent (CHT-ECH/CFA50) dose towards the output response (RR120 removal (%)) as shown in Fig. 5b, and comparatively less steep than the loading of CFA (%) curve of Fig.   5a. In contrast, Fig. 5c indicates that the solution pH has a very insignificant impact on the output response, and the RR120 removal (%) reduced significantly by rising the solution pH towards basic environment (the reason why will be given in the following section).
Subsequently, the effect of the working temperature and contact time factors on the output response is illustrated in the Fig. 5d and Fig. 5e, respectively. As can be seen, both input factors behaved similarly, and RR120 removal (%) increased by increasing the working temperature and extending the contact time.

Three-dimensional (3-D) surface plots
The 3-dimensional responses plots of the independent factors such as A (CFA loading (%)), B (dose), C (pH), D (temperature) and E (time) on the output response (RR120 removal %) are presented in Fig. 6. Hence, Fig. 6a depicts the interactive effect of CFA loading (%) (A) and dose (B) with p-value of 0.0041 which is statistically significant as proven by ANOVA test ( Table 4). The efficiency of CHT-ECH/CFA50 to remove RR120 (%) elevated from 12.7 to 90.2% as the dose increases from 0.04g to 0.10g and the increase of CFA loading from 0 to 50 %. This observation can be assigned to the uniform distribution of the CFA micro particles in the polymeric matrix of CHT-ECH, and as a result the specific surface area and pore diameter of CHT-ECH/CFA50 were developed and enhanced as proven in Table 3. This enhanced surface area property will lead to efficient diffusion of the RR120 dye molecules into inter pores and/or adsorption active sites. Subsequently, adding more doses of the adsorbent (CHT-ECH/CFA50) to the dye solution will offer more active adsorption sites for capturing anionic dye molecules such as RR120 [39]. Besides that, the interaction between dose (B) and temperature (D) was observed to be statistically significance with p-value of 0.0460. It can be confirmed from Fig.   6b, the removal of RR120 (%) significantly improves as the temperature increases from 30 ºC to 60 °C. This result shows that the process of RR120 removal (%) by CHT-ECH/CFA50 is an endothermic in nature. A similar trend can be observed from Fig. 6c and Fig. 6d where the RR120 removal (%) increases as the temperature increases. Furthermore, the relationship between input variables of solution pH (B) and working temperature (D) on the output response was investigated and presented in Fig. 6c. The interaction is statistically significant as indicated by p-value (0.0353) from Table 4. As can be seen from solo input variable (Fig. 5c) and reconfirmed in the significant interaction of input variables (Fig. 6c), the RR120 removal (%) is very sensitive to the solution pH change. In this respect, it can be observed clearly that the RR120 removal (%) was greater in acidic pH range and decreased with an increase in solution pH towards basic environment. The acidity of the medium is a significant factor affecting the RR120 adsorption onto CHT-ECH/CFA50 surface. This could be due to the change in surface change of the adsorbent explained by pHpzc as displayed in Fig. 5d, and the value of pHpzc was found to be 7.1. This observation demonstrates that the CHT-ECH/CFA50 is a positively charged adsorbent at acidic environment due to the protonation of -NH2 into -NH3 + group, which is very much preferable in capturing anionic dye molecules from treated solution [27][28][29].

The effect of contact time and kinetic study
The influence of contact time on RR120 adsorption at various initial concentrations (50-300 mgL) was investigated by adopting the optimal input variables which were previously determined from experiment 14 in Table 2 as follow: CHT-ECH/CFA50 = 0.07 g, solution pH=4, and working temperature = 45 ºC. Fig. 7a shows the RR120 uptake by CHT-ECH/CFA50 versus contact time at different initial RR120 concentrations. As can be seen, the adsorption occurred quickly in the initial steps due to availability of unoccupied adsorption active sites, and then gradually filled over time, resulting in no more change in adsorption capacity [7,29].
The adsorption kinetics were determined by applying Lagergren pseudo-first order (PFO) model [40] and pseudo-second order (PSO) model [41]. The non-linear forms of these PFO and PSO models can be expressed in the Eqs. (4) and (5) respectively.
q t = q e 2 k 2 t 1 + q e k 2 t where, qt (mg/g) is the amount of RR120 adsorbed by CHT-ECH/CFA50 at time and qe (mg/g) is the amount of RR120 uptake by CHT-ECH/CFA50 at equilibrium. k1 (1/min) and k2 (g/mg min) are the corresponding rate constants of the two models, respectively. Table 4 presents the correlation coefficients (R 2 ) of the kinetic parameters of these models, which depicts that the adsorption kinetic is well described by PSO rather than PFO kinetic. Moreover, the calculated qe values of PSO model are very close to the experimental qe values. The dominancy of the PSO over PFO indicates that the overall rate of the adsorption process was controlled by chemisorption which involves valency forces through electrons sharing between the adsorbent and adsorbate [42].

Adsorption isotherm
Adsorption isotherm is an important function to investigate the most appropriate adsorption isotherms for predicting the pathways of the adsorption mechanism of RR120 onto CHT-ECH/CFA50. The isotherm parameters obtained will aid in understanding the adsorbent/adsorbate interactions [43]. Thus, many isotherm models such as Langmuir model [44], Freundlich model [45] and Temkin model [46] were used in this study. The Langmuir model is a typically adopted adsorption isotherm that portrays the mechanism of monolayer adsorption on a defined homogeneous surface. The nonlinear Langmuir isotherm equation is expressed in Eq. 6.
Where and represent the amount of RR120 adsorbed at equilibrium (mg/g) and the equilibrium RR120 concentration (mg/L). is the maximum adsorbed quantity of RR120 (mg/g) and Ka is the Langmuir constant related to the active adsorption sites affinity (L/mg).
Conversely, the Freundlich isotherm is an empirical equation which is applicable for adsorption processes occurring on heterogeneous sites with different adsorption energies [45]. The empirical description of the Freundlich isotherm model is given in Eq. (7).
q e =K f C e 1 n ⁄ Where the Freundlich constant K f is related to the adsorption capacity (mg/g) and n provides the information regarding the adsorption process's favourability. The Temkin model is based on the presumption that as the coverage increases, the adsorption energy reduces linearly [46].
The Temkin isotherm can be described in Eq. 8.
Where R (J/mol K) is the universal gas constant (8.314), T (K) is temperature, the Temkin constants bT is associated to the heat of adsorption and KT (L/mg) is the constant related to the equilibrium binding. The plots of the adsorption isotherm models are presented in Fig. 7b and their corresponding parameters computed based on the models are summarized in Table 6 computed from the Langmuir model is 237.7 mg/g. In addition, higher Kf value (12.1) and the value of n which is greater than 1 (n = 1.29), suggest that the adsorption process is favourable [47]. Moreover, the adsorption capacity of CHT-ECH/CFA50 of various adsorbents towards RR210 adsorption was assessed, and the summary of the adsorption capacities are presented in Table 7. Thus, the reported data indicates that CHT-ECH/FA50 can be considered as a potential low-cost adsorbent for treatment of acidic (anionic) dye in the aqueous environment.
where T is the absolute temperature and R is the universal gas constant (8.314 J/mol K). kd is the distribution coefficient, qe is the amount of dye adsorbed by CHT-ECH/CFA50 at equilibrium (mg/g) and Ce is the equilibrium dye concentration in the liquid phase (mg/L). The calculated thermodynamic parameters are presented in Table 8 and from the gradient and intercept of Van't Hoff plots of lnkd against 1/T (Fig. 7), the values of ΔH° and ΔS° were computed respectively. Based on the thermodynamic calculations, the uptake of RR120 at solution pH 4 onto CHT-ECH/CFA50 surface was indicated as a spontaneous process if the value of Gibbs energy change (∆° ) at all working temperatures were negative. In addition, as the temperature increases, ∆° decreases, implying a stronger adsorption force and thus promoting the adsorption spontaneity to occur at high temperature [55]. The entropy change (ΔS°) is positive (0.24 kJ/mol K) which signifies an enhancement of the system randomness at the solid/solution interface during the adsorption process. The positive enthalpy change (ΔH°= 71.37 kJ/mol) advocates an endothermic nature and suggesting that energy is required in transferring RR120 from the aqueous phase to the solid phase [56].

Adsorption mechanism
The proposed mechanism for the adsorption of RR120 dye onto CHT-ECH/CFA50 is illustrated in Fig. 9. CHT contains large amounts of amino (-NH2) and hydroxyl (-OH) functional groups, making it an ideal adsorbent for removing anionic pollutants from the aqueous systems. These amino (-NH2) groups can easily be protonated (-NH3 + ) in acidic medium, enabling them to strongly adsorb anionic dyes through strong electrostatic attractions [1,4]. Other possible forces are dipole-dipole hydrogen bonding and Yoshida H-bonding which can be formed by the interaction of existing H and -OH on the CHT-ECH/CFA50 surface with nitrogen (N) atom in the aromatic ring of RR120 dye structure [7,23]. Furthermore, possible n-π stacking interactions can be formed between the delocalization of lone pair electrons on an FTIR spectra of (a) CHT-ECH/CFA50, (b) CHT-ECH/CFA50 after adsorption of RR120 dye Figure 3 SEM images of (a) CHT-ECH, (b) CHT-ECH/CFA50 (c) CHT-ECH/CFA50 after adsorption of RR120 dye at magni cation power =3000x.   (a) Effect of the contact time on RR120 adsorption at different initial concentrations, and (b) adsorption isotherms of RR120 by CHT-ECH/CFA50 at the following experimental condition: (dosage 0.07 g, pH of solution 4, temperature 45 °C, agitation speed=100 strokes/min, and volume of solution =100 mL).

Figure 9
Illustration of the possible interaction between CHT-ECH/CFA50 surface and RR120 dye including electrostatic attraction, hydrogen bonding interactions, Yoshida H-bonding, and n-π interactions