Introduction

Dyes are used as coloring agents in various industries such as the textile, leather, paper, plastic, cosmetics, printing, pharmaceuticals, and food (Anithaa et al. 2016; Dashamiri et al. 2017; Jothirani et al. 2016). They generate colored wastewaters, depending on the fixation level of dye stuffs on the substrates, which change with the nature of the substance, the intensity of coloration, and the applied method (Amin 2009). Wastewaters coming from dyeing industries can be toxic to aquatic life by weakening the light flux which strongly inhibits the photosynthetic process (Kumar et al. 2018; Giannakoudakis et al. 2015) and can cause severe effects to the human health (Suganya et al. 2017; Kumar et al. 2014a, b), due to their mutagenic and carcinogenic aspects (Jamshidi et al. 2016). Furthermore, some recalcitrant dyes are resistant to bio-degradation [Ghaedi et al. 2016; Bardajee et al. 2017; Khan et al. 2011). Hence, many treatments have been conducted for removing the dyes from the textile effluents (Kumar and Subramaniam 2013; Setiabudi et al. 2016). The adsorption is frequently used to diminish the color from effluents (Kumar et al. 2011) due to its simplicity of design and operation (Senthamarai et al. 2013; Kumar et al. 2015) and insensitivity to toxic substances (Kumar et al. 2014a, 2014b; Chatterjee et al. 2012; Kaçan and Kütahyah 2012). The commercial activated carbons with very large specific surface areas and pore structure possess high adsorption activities (Fernandez et al. 2014), but their applications remain limited because of their high cost. So, the use of inexpensive agriculture by-products to produce activated carbon provides an economical solution to these environmental problems (Deniz and Karaman 2011); one can cite activated carbon developed from the orange peels (Köseoğlu and Başar 2015), peanut shells (Georgin et al. 2016), and tomato (Lycopersicon esculentum Mill) (Sayğılı and Güzel 2016). The main objective of this research is devoted to the removal of the dye Acid Blue 25 (AB 25) from aqueous solution using activated carbon prepared from wild date stones (WDS).

Previous studies of dyes adsorption investigated the effect of individual parameters while maintaining the other constants at unspecified levels. However, this approach does not depict the combined effect of all parameters. This process is time-consuming and requires an increased number of experiments and more products to reach the optimum levels, which may be unreliable. These restrictions can be eliminated by optimizing the process parameters collectively by statistical experimental design such as the response surface methodology (RSM). In this respect, various parameters including the biomass dose, initial AB25 concentration (C0), contact time, and temperature were optimized by central composite design (CCD) combined with (RSM).

Materials and methods

Materials

Wild date stones (WDS) were used as precursor material for the preparation of activated carbon. They were collected from the University campus (Algiers), thoroughly washed with distilled water to remove dust, and dried at 105 °C for 24 h.

The carbonization temperature of the natural material was investigated by Thermogravimetric Analyzer model (SDT Q600, TA instruments); the biomass was heated up to 1000 °C at a heating rate of 10 °C min−1 under N2 atmosphere (100 mL min−1). The carbonization temperature can be selected by plotting the weight loss percentage versus temperature. It is observed from Fig. 1 that the weight loss percentage was negligible up to 850 °C, indicating that the main structure of char was formed. So, this temperature was selected for the carbonization of WDS biomass.

Fig. 1
figure 1

TGA curve of natural material

Full size image

The pyrolysis of the raw material was performed in a carbonization device under N2 atmosphere. The sample was heated at 900 °C (10 °C min−1) under N2 flow (150 mL min−1, 2 h) in a tubular furnace. Following the carbonization, the chars were heated under N2 flow up to 850 °C. Once the final temperature was reached, N2 flow was switched to CO2 and activated for 2 h. Finally, the CO2 flow was replaced with N2 during cooling to room temperature. The samples were kept in desiccators for further use.

The Acid Blue 25 dye (AB25: C20H13N2NaO5S; MW = 416.38 g mol−1; chemical index (CI) of 62055) was purchased from the textile Algerian Company and is characterized by a weak biodegradability and high toxicity. The adsorption tests were performed on synthetic solutions prepared with non-purified commercial product.

To visualize the surface morphology of WDS and WDS-AC, scanning electron microscopy (SEM, JEOL JSM-6360) was used. The specific surface areas of adsorbents were measured from the N2 adsorption–desorption isotherm using a Micrometrics ASAP 2010 V5.00 H Surface Analyzer in the relative pressure (P/P0 = 0.308). The micropore volumes were evaluated with the accumulative pore volume using the t-plot method. The total pore volume was calculated from the volumes of N2 absorbed at relative pressure (P/P0 = 0.349). In order to identify the functional groups of WDS and WDS-AC, the Fourier transform infrared spectroscopy (FTIR, PerkinElmer spectrometer) was carried out in the range (400–4000 cm−1) by mixing 1 mg of adsorbents with 200 mg of KBr (spectroscopic quality) to get translucent disk.

Procedure and analysis

All adsorption tests were done in batch mode according to the same experimental protocol. A stock solution of 1000 mg L−1 was prepared by dissolving AB25 in distilled water and stored in glass bottles. The solutions of various AB25 concentrations were made up by dilution; the pH was natural without any adjustment. In the batch method, fixed quantities of adsorbent were added in the specific volume of AB25 solutions and agitated at 400 rpm at controlled temperature until equilibrium. The aqueous samples were immediately centrifuged for 5 min at 4000 rpm for 15 min. The final AB25 concentrations were analyzed with a double-beam UV–visible spectrophotometer (Specord 210 Plus Analytik Jena) at the maximum wavelength (602 nm). The percentage of AB25 removed by the adsorbent (Y) was calculated from the following equation (Boumaza et al. 2012; Prola et al. 2013):

$$ Y\left( \% \right) = \frac{{C_{0} - C_{e} }}{{C_{0} }}*100 $$
(1)

where C0 and Ce (mg L−1) are the liquid-phase concentrations of AB25 at t = 0 and equilibrium, respectively. The amounts qe (mg g−1) adsorbed per unit mass of adsorbent are evaluated from the following formula (Kaouah et al. 2013):

$$ q_{e} = \frac{{\left( {C_{0} - C_{e} } \right)*V}}{m} $$
(2)

where V is the volume of the dye solution (mL) and m the adsorbent dose (g).

RSM is a group of statistical and mathematical techniques for analyzing and evaluating the interactive effect of several variables taken from the fit of empirical models to the experimental data [Arulkumar et al. 2011; Vecino et al. 2012; Fakhri 2015). In this work, four independent parameters were investigated by using RSM under CCD method, including the AB25 concentration (X1), adsorbent dose (X2) temperature (X3), and time (X4). These variables were set in five levels to evaluate the influence of the quantities of removal yield for the optimization process. The independent variables are coded in the range (−1, +1) with a low (−1) and high levels (+1). The axial points are located at a distance α from the center, making the design rotatable (Roosta et al. 2014) and α fixed at 2 (rotatable). Replicates of the test at the center are very important since they provide an independent estimation of the experimental errors and a reproducibility of the data runs (Cronje et al. 2011). For four variables, the recommended number of tests at the center is 6. Therefore, the number of experimental runs is calculated as follows (Chatterjee et al. 2012; Azargohar and Dalai 2005):

$$ N = 2^{n} + 2n + n_{c} = 30 \left( {n = 4} \right) $$
(3)

The range and levels of the independent numerical parameters in terms of actual and coded values are given in Table 1.

Table 1 Experimental range and levels of the independent variables for AB25 adsorption
Full size table

The mathematical relationship between the four independent factors can be approximated by the second-order polynomial model (Eq. 4) (Cronje et al. 2011):

$$ Y = \beta_{0} + \mathop \sum \limits_{i = 1}^{n} \beta_{i} X_{i} + \mathop \sum \limits_{i = 1}^{n} \beta_{ii} X_{i}^{2} + \mathop \sum \limits_{i = 1}^{n - 1} \mathop \sum \limits_{j = 1}^{n} \beta_{ij} X_{i} X_{j} $$
(4)

where Y is the predicted response (removal percentage) and Xi’s the independent variables (AB25 concentration, adsorbent dose, temperature, and time) known for each experimental run. The parameter β0 is the model constant, βi the linear coefficient, βii the quadratic coefficients, and βij the cross-product coefficients. The analysis of variance (ANOVA) method is applied to estimate the significance level of factors (Cheng et al. 2015; Hou et al. 2012).

Results and discussion

Surface characteristics

The SEM micrographs of WDS and WDS-AC are shown in Fig. 2. A significant difference between the morphology of both adsorbents is observed. Unlike smooth surface of WDS (Fig. 2a, b), the surface of WDS-AC (Fig. 2c, d) is coarse with lots of micropores and holes, where there is a good possibility for the dye to be adsorbed inside these pores. The improvement of the surface relief of the adsorbent after activation and carbonization is confirmed by the augmentation of the BET surface area and total pore volume of adsorbents. The surface area and pore volume of WDS were 0.134 m2 g−1 and 0.00048 cm3 g−1, respectively, and increase up to 610.843 m2 g−1 and 0.224 cm3 g−1 for WDS-AC.

Fig. 2
figure 2

SEM micrographs of WDS: (a, b) and WDS-AC: (c, d)

Full size image

The FTIR spectra of the precursor and WDS-AC are illustrated in Fig. 3. It can be seen that the structure of WDS is rich in functional groups than WDS-AC, thus revealing the complex nature of the precursor. The bands (O–H) at 3435 cm−1 and (C–H) at 2921 and 2856 cm−1 observed for both samples are attributed to hydroxyl, methyl and methylene groups, respectively. Additionally, the intensity of the bands between 1750 and 1450 cm−1 diminishes for the activated carbon compared with the natural precursor, due to the activating action of CO2 at high temperature. The absorption band at 1746 cm−1 indicates the existence of carboxyl or aldehydes functional group (C=O), while the peak 1641 cm−1 is assigned to alkenes bands (C=C). From the FTIR spectrum of activated carbon, many bands disappeared in the range (1400–500 cm−1) while others emerged. The vanished peaks related to the aromaticity of the material increased during activation. The peak at 1385 cm−1 is attributed to the presence of (CH3 and–CH2), whereas the broadband of C–O–C stretching vibration is observed at 1120 cm−1. Finely, the new peak at 1540 cm−1 corresponds to the C–C stretching vibration.

Fig. 3
figure 3

FTIR spectra of precursor and WDS-AC

Full size image

Experimental design results

The CCD matrix coupled with the predicted and actual results is given in Table 2; the GMP 8.0.2 software is used for the regression and graphical analysis of the obtained data. The sum of squares, mean squares, estimated coefficient, F values, and p values are also tested using ANOVA, and the results are regrouped in Table 3. In statistic, a model with large F value and small p value (< 0.05) is considered to be significant (Hou et al. 2012). So, it can be concluded that the coefficients for the linear effect of all factors: initial AB25 concentration C0 (X1), adsorbent dose (X2), temperature (X3), and time (X4) (P < 0.05 for all), are highly significant (Table 3). Moreover, the coefficients of the quadratic effect of AB25 concentration (X1*X1), adsorbent dose (X2*X2), and time (X4*X4) (P < 0.05 for all) are considered as significant, while the quadratic effect of temperature (X3*X3) does not seem significant with high P value (P = 0.2976). In the interaction effect, none of the variables was found to be significant except between the AB25 concentration (X1) and adsorbent dose (X2) (P < 0.0001) and concentration C0 (X1) and temperature (X3) (P < 0.05).

Table 2 Data statistics of model variables
Full size table
Table 3 Analysis of variance (ANOVA) of the factors on the responses
Full size table

According to the experimental data, the quadratic regression model where the insignificant terms are not included is described by the relation:

$$ \begin{aligned} & Y = 57.2123 - 5.5601x_{1} + 10.4947x_{2} + 5.4999x_{3} + 3.5178x_{4} + 6.5292x_{1}^{2} + 2.521x_{2}^{2} \\ & \quad + 1.7573x_{4}^{2} - 6.7051x_{1} x_{2} - 2.6706x_{1} x_{3} \\ \end{aligned} $$
(5)

A positive sign in the equation represents a synergistic effect of the variables, while a negative sign indicates an antagonistic effect (Garba et al. 2016). The fit of models is evaluated by the correlation coefficient R2 and adjusted coefficient R2 (\( R_{\text{adj}}^{2} \)) (Chatterjee et al. 2012). The high values of R2 (= 0.963) and \( R_{\text{adj}}^{2} \) (= 0.941) indicate that the response surface quadratic model is appropriate for predicting the performance of AB25 removal (Table 4). The predicted correlation coefficient (Q2 = 0.878) agrees with the coefficient \( R_{\text{adj}}^{2} \). Moreover, both F (= 42.6846) and p (< 0.0001) proof statistically significance of the model.

Table 4 The ANOVA for the full quadratic model for R % of AB25
Full size table

The fitted quality of Eq. 5 is also expressed through the AB25 adsorption between experimental and predicted models (Fig. 4) with a good agreement between the actual and predicted values of AB25 removal, indicating the goodness of fit and reliability of the model (R2 = 0.969).

Fig. 4
figure 4

The predicted data versus the experimental data of normalized removal of AB25 dye

Full size image

Residuals normal probability plot for the reduced model

One of the essential hypotheses for the statistical analysis of the experimental data is their normal distribution (Asfaram et al. 2016). The normal possibility for studentized residuals is illustrated in Fig. 5. The distribution of residuals is normal, if the experimental data follow a straight line (Dastkhoon et al. 2015). It is clear that the experimental points are normally distributed with no outliers and located on the normal line spread out between −4 and +4 studentized residuals (Fig. 5).

Fig. 5
figure 5

Normal probability plots of residuals for % removal of AB25 dye

Full size image

Interactive effects of different parameters on adsorption of AB25 dye

The relation between different factors and responses is elucidated by the exploration of contour plots as a function of two factors by maintaining the other factors at central level. In the present work, the effect of the AB25 concentration on its removal percentage and its interaction with the adsorbent dose and temperature are illustrated in Fig. 6a, b, respectively. The curvature natures of these plots prove the interaction of variables.

Fig. 6
figure 6

Contour plots indicating interaction effects of independent variables on variation of removal percentage: a initial concentration–adsorbent mass; b initial concentration–temperature

Full size image

Figure 6a represents the interaction between the AB25 concentration and adsorbent dose and their influence for the removal percentage. The augmentation of the initial AB25 concentration C0 from 100 to 200 mg L−1 is associated with the diminution of the removal percentage. This may be due to the fact that the ratio of adsorption sites in comparison to adsorbate AB25 dye becomes smaller and not sufficient to interact with adsorbate species, leading to decreased removal efficiency (Mazaheri et al. 2016). In contrast, the adsorbent dose has a positive effect on the percentage removal which improves from low (0.4 g L−1) to high value (0.8 g L−1). This behavior evidently indicates that the augmentation of adsorbent dose increases the number of active sites and improves the interaction adsorbate/adsorbent.

In adsorption, the temperature plays an important role; Fig. 6b shows the contour plot representing the composed influence of AB25 concentration C0 and temperature. With augmenting the temperature from 25 to 45 °C, the percentage of AB25 removal increases because of the higher active surface area and kinetic energy of AB25 molecules. In other view, the increase in the diffusion rate of AB25 molecules at higher temperature across the boundary layer and inside the pores results from the decrease in the solution viscosity (Sivarajasekar and Baskar 2014).

Process optimization

The best local maximum was determined by the reduced model: AB25 concentration C0 of 100 mg L−1, adsorbent dose (0.8 g L−1), temperature (45 °C), and adsorption time (120 min). Under these optimal conditions, the maximum AB25 elimination from aqueous solution by adsorption approaches 100%. These optimum parameters were experimentally checked to verify the adequacy of the model with a removal of 99.61%. The percentage error between the actual and predicted values is very small; this corroborates the result of the response surface optimization and indicates that the suggested model is adequate for getting the optimum values for the studied factors.

Kinetic study

The kinetics is fundamental for the adsorption of AB25 (Aljeboree et al. 2017), and it characterizes the rate-limiting step and elucidates the adsorption process (Sivarajasekar and Baskar 2014; Bouhamidi et al. 2018). In this respect, three models, namely the pseudo-first order, pseudo-second order, and intraparticle diffusion, were tested to evaluate the experimental data.

The pseudo-first-order model proposed by Lagergren and Ho and pseudo-first-order model developed by McKay are defined by Eqs. 6 and 7 (Deze et al. 2012; Puchana-Rosero et al. 2016), and the model parameters and regression coefficient (R2) are evaluated by the nonlinear regression using the ORIGIN 8.6 software.

$$ q_{t} = q_{e} \times \left[ {1 - \left. {{ \exp }\left( { - k_{1} \times \left. t \right)} \right.} \right]} \right. $$
(6)
$$ q_{t} = \frac{{k_{2} q_{e}^{2} t}}{{1 + k_{2} q_{e} t}} $$
(7)

where qe (mg g−1) is the amount of AB25 adsorbed at equilibrium, qt (mg g−1) the amount of AB25 adsorbed at time t (min), k1 (min−1) the pseudo-first-order rate constant, and k2 (g mg−1 min−1) the rate constant of pseudo-second order.

In addition to the coefficient (R2), the Chi-square statistical test (χ2) is also calculated to select the best model to the experimental data (Unuabonah et al. 2009)

$$ \chi^{2} = \mathop \sum \limits_{i = 1}^{n} \frac{{(q_{{e,{ \exp }}} - q_{{e,{\text{cal}}}} )^{2} }}{{q_{{e,{\text{cal}}}} }} $$
(8)

where n is the number of data points, \( q_{{e,{ \exp }}} \) the observation from the experiments, and \( q_{{e,{\text{cal}}}} \) the calculation from the models.

The data are fitted for both models in the concentrations range (90–150 mg/L), maintaining the other parameters at their optimum values (adsorbent dose 0.8 g L−1, temperature 45 °C, and adsorption time 120 min) which are illustrated in Fig. 7, and the corresponding kinetic parameters (R2 and χ2) are gathered in Table 5.

Fig. 7
figure 7

Pseudo-first-order and pseudo-second-order models of AB25 adsorption onto WDS-AC at different concentrations a 90 mg L−1; b 130 mg L−1; c 150 mg L−1)

Full size image
Table 5 Pseudo-first-order and pseudo-second-order models constants, correlation coefficients, and Chi-square statistical test for BA25 dye adsorption onto WDS-AC
Full size table

On the basis of R2 and χ2 for both models (Table 5), the experimental data are well fitted by the pseudo-second-order model with the initial concentrations. Additionally, the k2 values for the AB25 system diminishes with augmenting the AB25 concentration C0; the reason for this phenomenon can be attributed to the small competition for the surface active sites at lower dye concentration. On the contrary, at higher concentrations, the competition for the adsorption sites increases leading to lower k2 values (Sayğılı and Güzel 2016).

The intraparticle diffusion model proposed by Weber and Morris (Weber and Morris 1963; Kumar et al. 2010a, b) is represented mathematically by the following relation:

$$ q_{t} = k_{\text{dif}} t^{0.5} + C $$
(9)

where qt (mg g−1) is the amount of adsorbed AB25 at time t, kdif (mg g−1min−1/5) the rate of intraparticle diffusion model and C the intercept, determined from the linear plots of qt versus t0.5 (Fig. 8).

Fig. 8
figure 8

Evolution of qe versus t0.5 at different concentrations a 90 mg L−1; b 130 mg L−1; c 150 mg L−1)

Full size image

The intraparticle diffusion is the sole controlling step, if the plot of qt versus t½ satisfies the linear relationship and the C value is fitted to zero (Islama et al. 2015). If not, the external diffusion and the boundary layer diffusion along with intraparticle diffusion are also involved in the adsorption process (Cheung et al. 2007; Zhou et al. 2019)

The evolution of qt versus t0.5 (Fig. 8) is nonlinear, over the adsorption time; nevertheless, the process can be divided into three linear portions which do not pass by the origin whenever the AB25 concentration. This observation indicates that the intraparticle diffusion is not the rate-controlling step and other mechanisms are involved of the adsorption of AB25 onto WDS-AC (Deniz and Karaman 2011). The first linear form corresponds to the rapid external surface adsorption, i.e., the AB25 molecules are transferred to the easily accessible actives sites of the adsorbent which takes place at the film diffusion. The second line denotes intraparticle diffusion where the gradual adsorption stage is controlled. The third straight line indicates the last equilibrium step when the diffusion rate decreases and becomes stable by the diminution of the concentrations gradient between the adsorbate and adsorbent surface.

In the same way, the C value presents a concept of the boundary layer thickness (De Luna et al. 2013). It can be noticed that for the three initial AB25 concentrations, the kdif values decrease with raising the C value. Such behavior is due to the increased boundary layer (Ӧztürk and Malkoc 2014), while the decrease in kdif is attributed to the driving force with the initial concentration C0 and adsorption into both mesopores and micropores (Beltrame et al. 2018).

Isotherm study

In this work, the two-parameter models, namely the Langmuir, Freundlich, and Dubinin–Radushkevich (Table 6), were chosen to give an idea about the relationship between the distribution of the dye molecules in the liquid phase and in the solid surface at equilibrium (Jothirani et al. 2016; Suganya et al. 2017). The AB25 concentration ranges from 50 to 150 mg L−1, while other parameters were preserved at optimum conditions. The fitted data for different isotherm models are illustrated in Fig. 9 by nonlinear regression form using the ORIGIN 8.6 software, and their constants parameters along with R2 and χ2 values are summarized, respectively, in Table 6.

Table 6 Summary of constants parameters for isotherm models and correlation coefficients estimated for the adsorption of AB25 onto WDS-AC
Full size table
Fig. 9
figure 9

Adsorption isotherm models fitted to experimental adsorption of AB25

Full size image

On the basis of R2 and χ2 values (Table 6), the isotherm data were suitably fitted by Langmuir model with the highest R2 and lowest χ2 values. This result suggests that the AB25 adsorption occurs specific homogeneous sites and as a monolayer adsorption of AB25 onto WDS-AC (Asfaram et al. 2015; Ghaedi and Kokhdan 2015). Additionally, the mean free energy of transfer (E) was determined to verify the nature of the adsorption process using the Dubinin–Radushkevich model. It is concluded from the E value (E = 3.269 kJ mol−1) that the adsorption proceeds through physical sorption (1–8 kJ mol−1) (Khan et al. 2016).

In order to situate our novel adsorbent among those used to treat acid dyes from synthetic water, a comparison based on the maximum monolayer adsorption capacities was performed. The results (Table 7) showed that the WDS-AC possesses an excellent adsorption capacity when compared to other adsorbents and could be recommended for the removal of dyes from the textile effluents.

Table 7 Comparison of maximum monolayer adsorption capacities of acid dyes onto various adsorbents reported in the literature
Full size table

Conclusion

In the present study, wild date stones were used as novel material for an activated carbon preparation to be applied for the removal of Acid Blue 25 dye from aqueous solution. The carbonization temperature of the raw material was selected at 850 °C using thermo-gravimetric analysis. Several structural characterization analyses such as BET, SEM and FTIR confirmed the successful activation and carbonization of precursor. The BET surface area of activated carbon was found to be 611 m2 g−1. The central composite design (CCD) was used to optimize significant adsorption variables including the initial dye concentration, adsorbent dose, temperature, and time, and the optimum conditions for 99.61% removal of AB25 dye from synthetic water were achieved to be initial AB25 concentration of 100 mg L−1, adsorbent dose (0.8 g L−1), temperature (45 °C), and adsorption time (120 min). Analysis of variance (ANOVA) was suggested that the experimental data were fitted to the quadratic model. The analysis of kinetics data revealed the pseudo-second-order model was the best fitted model, while the intraparticle diffusion model is not the only controlling step; other mechanisms might be involved in the adsorption process. The Langmuir isotherm provided the best fit with the high correlation coefficient (R2) of 0.993 and the maximum monolayer adsorption capacity of 181.59 mg g−1. Finally, the results showed that the wild date stones as a low-cost material may be a suitable alternative for the elimination of AB25 from colored synthetic water.