TiO 2 -graphene/chitosan nanocomposite: preparation and its application for removal of anionic dyes

In this work, TiO 2 -graphene/chitosan nanocomposite with high photocatalytic activity was successfully synthesized and characterized by various analyses such as XRD, TEM, SEM, EDX and DRS. The photocatalytic activity was tested vs. removal of methyl red as ananionic dye under black light radiation. Based on the results, TiO 2 -graphene/chitosan nanocomposite could effectively remove methyl red, and demonstrate an excellent photocatalytic enhancement over TiO 2 and TiO 2 -graphene samples. The degradation reaction fit well to a Langmuir-Hinshelwood kinetic model implying that the reaction rate is depended on the initial adsorption step. An artificial neural network (ANN) comprising four input variables (TiO 2 -graphene/chitosan dosage, initial dye concentration, reaction time and temperature of the solution), eight neurons and an output variable (Removal efficiency %) was optimized, tested and validated for methyl red degradation by the prepared TiO 2 -graphene/chitosan nanocomposite. The results showed that the predicted data from the designed ANN model are in good agreement with the experimental data with a correlation coefficient (R 2 ) of 0.9831. Based on the results, reaction time is the most influential variable and the temperature of solution is the less influential parameter in the removal efficiency of methyl red. © 2020 by SPC (Sami Publishing Company), Asian Journal of Green Chemistry, Reproduction is permitted for noncommercial purposes. nanocomposite the anionic dye (Methyl red) from aqueous The results demonstrated that TiO 2 -graphene/chitosan nanocomposite is more effective in comparison with TiO 2 , and TiO 2 -graphene catalysts. Based on the results, removal efficiency was obviously affected by TiO 2 -graphene/chitosan dosage, initial dye concentration, reaction time and temperature of the solution. Photocatalysis of methyl red can be explained in terms of the Langmuir– Hinshelwood kinetic model (k dye = 0.85 (mg/L − 1 ) − 1 and k c = 3.21 (mg/L − 1 ) − 1 . The performance of the photocatalytic removal of methyl red was successfully predicted by applying a three-layered neural network with eight neurons in the hidden layer, and using a back-propagation algorithm. Also, a measure of the saliency of the input variables was made based on the connection weights of the neural networks, allowing the analysis of the relative importance of input variables to the value of


Graphical Abstract Introduction
Semiconductor-based heterogeneous photocatalysis is one of the most promising procedures for the removal of pollutants from wastewater. Among different semiconducting materials, titanium dioxide (TiO2) presents the most suitable properties: it is chemically and biologically inert, stable, non-toxic, cheap and easy to produce [1]. However, TiO2 has its own disadvantage that includes the fast recombination of photogenerated electron-hole pairs [2]. In order to suppress charge recombination and extend light absorption of TiO2, many strategies have been attempted to dope TiO2 and to combine TiO2 with the related materials. Especially, carbon-based materials, such as carbon nano-tube, active carbon, and graphene were widely useded to modify TiO2 [3][4][5].
TiO2-graphene composites have attracted extensive attention for the different applications, including the removal of organic pollutants, generation of H2 by water splitting, reduction of CO2 for solar fuel production, and so on [6][7][8]. These research simply that the contact of TiO2 with graphene can improve the photoactivity as compared to the bare TiO2 and due to the enhanced adsorption of pollutants, extended light absorption range and increased charge separation [9,10]. Graphene has a conjugated structure and its combination with TiO2 can be an ideal preference to achieve an enhanced charge separation in electron-transfer process [11]. Das et al., [12] deposited various nanoparticles such as ZnO, TiO2, Fe3O4 and Ni on graphene. By using first principle calculations, they showed that charge transfer occurs between graphene and the deposited nanoparticles.
Sonophototcatalytic activity of Pt doped GO-TiO2 composites for DBS degradation was carried out by Neppolian et al., [13]. However, the obligation to separate the nanoparticles from the suspension after treatment restricts the procedure development. The above problems can be prevented by immobilization of the nanoparticles over appropriate supports [14].
The usage of polymeric matrixes such as alginate, chitosanand different resins is one of the most widely applied methods for the immobilization. Chitosan is a suitable natural biopolymer for the immobilization procedure due to its hydrophilicity, biodegradability, non-toxicity and availability [15]. Also, adsorption capacity of the chitosan for sequestering anionic dyes such as methyl red because of the electrostatic attraction between the protonated amine groups on the chitosan and the sulfonic groups of the anionic dyes is beneficial to increase the adsorption of anionic dyes together with the immobilized adsorbent [16].
The purpose of this work was to prepare TiO2-graphene/chitosan nanocomposite and evaluate its photoactivity for the removal of anionic dyes such as methyl red from aqueous solutions. Also, the influence of various operational parameters such as photocatalyst dosage, initial dye concentration, reaction time, and temperature of the solution were investigated. The photocatalytic removal of methyl red in the presence of TiO2-graphene/chitosan nanocomposite was also investigated and discussed in terms of the Langmuir-Hinshelwood kinetic model. The relative importance of each factor was calculated based on the connection weights of ANN model.
Chitosan, which was of analytical grade, was purchased from Sigmae Aldrich, USA. TiO2 nanoparticles were prepared by sol-gel method according to our previous work [17].

Preparation of graphene oxide sample
According to reference [18], graphene oxide was prepared via Staudenmaier method. First, 5 g graphite was mixed with concentrated nitric acid (45 mL) and sulfuric acid (90 mL) and stirred vigorously for 20 min to get dispersion. Next, potassium chlorate (55 g) was slowly added over 30 min, to avoid sudden increases in temperature, and the reaction mixture was stirred for 72 h at room temperature. The mixture was then added into a copious amount of distilled water and then filtered.
The obtained graphite oxide solid was rinsed repeatedly and washed with a 5% solution of HCl, and water repeatedly until the pH of the filtrate was neutral. The graphite oxide thus obtained was placed in a quartz boat and inserted into a tubular furnace preheated to 1050 °C and kept at this temperature for 30 s.

Preparation of TiO2/GO sample
In our previous work, we reported the preparation of TiO2/GO catalyst via hydrothermal process [19]. Graphene oxide (20 mg) was dissolved in a solution of distilled water (80 mL) and ethanol (40 mL) via ultrasonic treatment for 2 h. TiO2 (200 mg) was added to the synthesized graphene oxide solution and stirred for further 2 h to obtain a homogeneous suspension. The suspension was then placed in a 200 mL Teflon-sealed autoclave and maintained at 120 °C for 3 h to simultaneously achieve the reduction of graphene oxide and the deposition of TiO2 on the graphene sheets. Finally, the obtained sample was recovered via filtration, rinsed with high purity deionized water for several times, and dried at 70 °C for 12 h in the vacuum furnace.
Then, TiO2-graphene (4 g) was added to the concentrated solution and magnetically stirred for 1 h to reach homogeneity. The resulted mixture was kept for 8 h in a stable place. The weight ratio of chitosan to TiO2-graphene was 2:1. The mixture was added dropwise via a syringe to a 500 mL solution containing 15% NaOH and 95% ethanol. The volumetric ratio of NaOH/ethanol was 4. Then, they were stored in the solution for 24 h to allow the nanocomposite to be formed. The resulted sample was withdrawn from the solution and washed with deionized water several times to remove impurities. The obtained nanocomposite was dried in room temperature.

Characterization of TiO2-graphene/chitosan nanocomposite
The major phase of the samples was determined using X-ray diffraction (XRD) (Siemens/D 5000) with Cu Kα radiation (0.15478 nm) in the 2θ scan range of 20-60°. Size of the prepared nanocomposite was obtained by TEM instrument (Philips CM-10 HT-100 kV). The texture and morphology of the prepared nanocomposite was measured via scanning electron microscope (SEM) (Philips XL-30ESM). The chemical composition of the synthesized nanocomposite was analyzed by an energy dispersive X-ray spectroscopy (EDX) system. Ultraviolet/visible diffuse reflectance spectra (DRS) of samples were performed by Avaspec-2048 TEC spectrometer in order to detect the band gap energy (Eg) of products which estimated via the following formula:

Studies and analysis
The photocatalytic activity of the prepared samples was evaluated in the removal of model compound methyl red. The photocatalytic removal of methyl red was measured at ambient pressure and room temperature in a batch quartz reactor. Artificial illumination was provided by 36 W (UV-A) mercury lamp (Philips, Holland) with a wavelength peak at 365 nm positioned above the photoreactor. In each run, desired amounts of prepared sample and methyl red were fed into the batch quartz reactor and placed in the dark condition for 30 min with continuous stirring for adsorption-desorption equilibrium and then exposed to black-light irradiation.
where C0 is the initial concentration of dye (mg/L -1 ) and Ct isthe final concentration of dye (mg/L -1 ).

Results and Discussion
Characterization of prepared samples X-ray diffraction XRD pattern of TiO2-graphene/chitosan nanocomposite is shown in Figure 1. As shown in Figure   1, the diffraction peaks of prepared sample were observed at 2Ө=25. 43 be assigned to pure anatase phase [21]. However, no typical diffraction peaks belonging to the separate graphene are observed in the composite. The fact is that the main characteristic peak of graphene at 24.5° might be shielded by the main peak of anatase TiO2 at 25.4°. Figure 1c shows XRD patterns of TiO2-graphene/chitosan nanocomposite. It can be seen that there is no difference between XRD pattern of TiO2-graphene/chitosan nanocomposite and XRD pattern of bare TiO2, which reveals TiO2-graphene/chitosan nanocomposite synthesis did not destroy the characteristic structure of TiO2. This result is similar to what was reported by Alagumuthu and Anantha Kumar [22].
TEM analysis of TiO2 -graphene/chitosan nanocomposite TEM image of TiO2-graphene/chitosan nanocomposite has been illusterated in Figure 2. The average size of the primary particles estimated from TEM image is about 8-10 nm. It can be seen that the particles demonsterate a relatively uniform particle size distribution. The degree of agglomeration can be prevented through the growth of TiO2 nanoparticles from carboxyl groups on the surface of graphene nanosheets. Therefore, graphene nanosheets can facilitate uniform dispersion of the nanoparticles on its surfaceas concluded by Loh et al., [23]. As shown in TEM image, chitosan layers on the TiO2 surface attached together and generated the porous structure. The porous structure of chitosanon the TiO2 surface and elsewhere is an important characteristic that allows specific interactions of TiO2-graphene/chitosan nanocomposite with methyl red molecules, making it an important feature for the photocatalytic and adsorption performances [24].

SEM analysis of TiO2-graphene/chitosan nanocomposite
SEM image of TiO2-graphene/chitosan nanocomposite is shown in Figure 3. Figure 3 implies that TiO2-graphene/chitosan nanocomposite has a highly porous structure which suggests the appropriate of the nanocomposite as adsorbent for removal of dye molecules from aqueous solution.
Nawi et al. reported the fabrication of bilayer system consisting of TiO2 and chitosan (CS) biopolymer.
They found that the highly porous structure of immobilized TiO2 layer allows better diffusion of pollutants, increases the light penetration and improves the optical property [25].

Elemental analysis with EDX spectroscopy
The elemental composition of the synthesized nanocomposite was determined using energy dispersive X-ray spectroscopy analysis. C, O, N, and Ti peaks can be clearly observed from Figure 4.
EDX analysis demonstrated no significant levels of impurities which could have originated from the procedure. This result is similar to what was reported by Hasmath Farzana and Meenakshi [26].

DRS analysis
In order to investigate the influence of graphene oxide and chitosan on the optical absorption properties of titanium dioxide, DRS analysis has been performed ( Figure 5). The reflectance spectra  [27]. These phenomenon result not only from the high conductivity of graphene that can facilitate separation of photogenerated e -/h + pairs, but also from the formation of Ti-O-C bonding between TiO2 and graphene sheets [28]. Also, this can be attributed to the charge transfer between conduction band of TiO2 and graphene sheets. However, the electron transition from the valence band to the conduction band can occur easier with the band-gap narrowing in TiO2 nanoparticles [29].  So, the recombination of photo-generated electron-hole pairs can be prevented [30]. Therefore, the improvement in photocatalytic activity of TiO2/GO is imaginable.
II) The electrons on the graphene can be trapped by oxygen and water on the surface of TiO2/GO catalystand produce the hydroxyl and superoxide radicals. As a result, electron-hole recombination is largely inhibited and this further facilitates the formation of more . because of the valence band of TiO2 and the superoxide radicals anion ( 2 .− ) at the surface of catalyst, which in turn result in a rapid decolorization of methyl red [31].
From Figure 6, photocatalytic activity of TiO2-graphene/chitosan nanocompositeis considerably higher than that of TiO2-graphene. TiO2-graphene/chitosan nanocomposite became all dark red in color, that is, methyl red was strongly attracted. These results, in accordance with what reported by Hasmath Farzana and Meenakshi, showed that chitosan/TiO2 composite was preferential to TiO2 in photodegradation of different types of dyes such as reactive red, methylene blue, and rhodamine [26]. The incorporation of TiO2 with Ti-OH groups into chitosan with positively charged amino groups led to the grafting of these functional groups which is desirable for the adsorption of methyl red containing negatively charged carboxilic groups.

Effect of TiO2-graphene/chitosan dosage onphotocatalytic removal of methyl red
The examined the dependence of the photooxidation kinetics of methylene blue on TiO2 loading and found that the rates increase with TiO2 loading up to a limit and then decrease to a constant value.

Effect of initial dye concentration
The effect of change in initial dye concentration on the removal percentage in range (10-65 mg/L −1 ) was studied via keeping other experimental conditions constant at room temperature, reaction time equal to 17 min, and 0.45 g/L −1 TiO2-graphene/chitosan concentration. As can be observed from Figure 8, the removal efficiency decclines as the initial dye concentration enhances.
This behavior can be expessed via following parameters: By enhancing of methyl red concentration, more and more organic substances and intermediates may be adsorbed on the surface of TiO2-graphene/chitosan, and so the reaction of pollutant molecules with photogenerated holes or hydroxyl radicals is prevented because of the lack of any direct contact between them [35].
At high concentration of methyl red, TiO2-graphene/chitosan surface was covered mainly by dye molecules and so production of • OH and • 2 − superoxide radicals was declined [36].
These results, in accordance with what reported by Sayilkan and Emre, showed that the initial concentration of dye could affect the photocatalytic activity of TiO2/chitosan nanocomposite and the photodegradation rate decreased with increasing initial concentration of dye [37].

Kinetic analysis as function of initial dye concentration
Finding a simple rate equation, which fits the experimental rate data, is beneficial for engineering purposes. The removal of methyl red by TiO2-graphene/chitosan nanocomposite obeys apparently pseudo-first-order kinetics at low initial dye concentration and the rate expression is given by  where kap is the pseudo-first-order rate constant, Ct and Co are the concentration at time "t" and "t = 0", respectively. The values of kap can be estimated by applying a least square regression analysis. It is clear that the removal rate is dye concentration-dependent. As the concentration of methyl red was enhanced, the rate of dye removal declined. It is related to the adsorption of more dye molecules on the surface of catalyst. If more pollutant molecules are adsorbed on the surface of catalyst, the reaction of dye molecules with holes or hydroxyl radicals is prevented because of the lack of direct contact between them [38]. Furthermore, solubility of oxygen in water is low. So, during the photoreaction, oxygen concentration in the solution may be lost and photocatalytic removal efficiency decreased [39].  [42].

Effect of reaction time
The effect of blacklight illumination time on the photocatalytic removal of methyl red was studied from 1 to 21 min, at 20 mg/L −1 methyl red concentration, 0.45 g/L −1 catalyst concentration, and at room temperature (Figure 9). The results showed that the percentage of methyl red removal increased with an increase of the illumination time and reached up to 98.5% after 15 min of illumination time. Such data indicates the relatively higher activity of TiO2-graphene/chitosan enables the higher percentage of discoloration of methyl red in such short illumination time and has active sites for carrying out the reaction [43]. This result is similar to what was reported by Wang et al. [44].

Effect of temperature
The  [45]. It is well known that electron-hole pair production in the presence of light source is responsible for initiation of photoreaction [46]. So, the photo decolorization systems are usually operated at room temperature.

ANN modeling
ANN is an appropriate method for its ability in learning, simulation and prediction of experimental data. The topology of an ANN can be detected via the number of layers, the number of nodes in each layer and the nature of the transfer functions. Optimization of ANN topology is probably the most important step in the development of a model [47]. We applied a three-layered, feed for wardbackpropagation neural network in this work. The network examined in our study contained four in puts, representing TiO2-graphene/chitosan dosage, initial methyl red concentration, reaction time, and temperature. The ranges of operational parameters are given in Table 1  where Xmin and Xmax refer to the lowest and the highest value of the input variable Xi, respectively. It is the hidden layer structures that essentially explain the topology of a feed-forward network. It is well known that the selection of neurons in the hidden layer may have an important effect on network performance. In our research work, we tested various numbers of neurons, from 1 to 16, in the hidden layer. Figure 10 shows the relation between the network error and the number of neurons in the hidden layer. Each topology was repeated 3 times at least to prevent random correlation due to the random initialization of the weights. Optimization was based on minimizing the mean square error, MSE, which is defined as follows: where ti and ai are the predicted and experimental data of the dependent variable, respectively and N is the number of data. The relationship between MSE and the number of neurons in the hidden layer is presented in Figure 11. As can be seen that the performance of the network stabilized after inclusion of an adequate number of hidden units just about eight; therefore, 8 neurons were selected for the best performance of neural network model. The resulting ANN is schematically illustrated in Figure 12. Figure 12 shows the optimized ANN structure characterized by one hidden layer containing eight neurons.
To test the precision of the ANN model, a comparison was made between experimental and predicted removal efficiency (%) values. Figure 13 shows    Table 2 represents the weights generated via the ANN models applied in this work for removal of methyl red. The relative importance of the effect of each input variable on output variable can be obtained through the neural weight matrix [49].
For every input variable, the percentage change in the output, as a result of the change in the input variable, was estimated by the following equation 8 [50]: respectively, refer to input, hidden and output neurons. Figure 15 represents a comparison between the relative importances of input variables as calculated by equation 2 on the removal of methyl red.
As can be seen, reaction time, with a relative importance of 48 %, appeared to be the most influential

Conclusion
The present study was performed to synthesize and investigate the efficiency of TiO2graphene/chitosan nanocomposite to remove the anionic dye (Methyl red) from aqueous solutions.
The results demonstrated that TiO2-graphene/chitosan nanocomposite is more effective in comparison with TiO2, and TiO2-graphene catalysts. Based on the results, removal efficiency was obviously affected by TiO2-graphene/chitosan dosage, initial dye concentration, reaction time and temperature of the solution. Photocatalysis of methyl red can be explained in terms of the Langmuir-Hinshelwood kinetic model (kdye = 0.85 (mg/L −1 ) −1 and kc = 3.21 (mg/L −1 ) −1 . The performance of the photocatalytic removal of methyl red was successfully predicted by applying a three-layered neural network with eight neurons in the hidden layer, and using a back-propagation algorithm. Also, a measure of the saliency of the input variables was made based on the connection weights of the neural networks, allowing the analysis of the relative importance of input variables to the value of the dye removal efficiency. The results confirmed that ANN modelling could effectively reproduce experimental data and predict the behavior of the process.