3.1. Design of experiment using CCD
CCD was applied to develop a polynomial regression equation to analyze the relationship between the amounts of Iodine and methylene blue adsorbed during GPa synthesis, with the results presented in Table 2. The amount of iodine adsorbed ranged from 254.51 to 704.23 mg/g, and the amount of methylene blue from 14.01 to 22.28 mg/g.
Table 2
Geopolymers synthesis using CCD matrix
Runs | X1 (g) | X2 (g) | X3 (M) | Ii (mg/g) | MBi (mg/g) |
---|
1 | -1 | -1 | -1 | 456.85 | 17.88 |
2 | 1 | -1 | -1 | 437.82 | 16.12 |
3 | -1 | 1 | -1 | 432.99 | 14.13 |
4 | 1 | 1 | -1 | 423,35 | 19.36 |
5 | -1 | -1 | 1 | 215.60 | 14.21 |
6 | 1 | -1 | 1 | 247.46 | 20.27 |
7 | -1 | 1 | 1 | 214.51 | 14.02 |
8 | 1 | 1 | 1 | 275.88 | 20.54 |
9 | -1 | 0 | 0 | 404.57 | 14.21 |
10 | 1 | 0 | 0 | 456.85 | 20.25 |
11 | 0 | -1 | 0 | 455.11 | 20.15 |
12 | 0 | 1 | 0 | 552.46 | 22.28 |
13 | 0 | 0 | -1 | 601.22 | 21.68 |
14 | 0 | 0 | 1 | 704.23 | 21.35 |
15 | 0 | 0 | 0 | 625.13 | 20.79 |
16 | 0 | 0 | 0 | 620.11 | 20.96 |
17 | 0 | 0 | 0 | 620.11 | 21.56 |
18 | 0 | 0 | 0 | 620.11 | 20.94 |
19 | 0 | 0 | 0 | 620.11 | 20.96 |
20 | 0 | 0 | 0 | 620.11 | 20.95 |
Based on the experimental data, the centered composite design develops regression models to evaluate each response as a function of the process variables [12]. Model selection is based on the maximum values of R2, Adj-R2 and Pred-R2, so the quadratic model is suggested to describe both responses. As shown in Figs. 1 and 2, the correlation coefficients (R2) between experimental and predicted data were 0.9064 (Ii) and 0.9382 (MBi); these values were close to unity. Both R2 coefficients were relatively high and in reasonable agreement with the adjusted R2 values of 0.8721 and 0.8826 for iodine and methylene blue removal respectively, indicating that the predicted values are in agreement with the actual values. The resulting model equations for Ii and MBi are given in equations 6 and 7 respectively.
$$\:\text{I}\text{i}\:(\text{m}\text{g}/\text{g})=622.22+{11.68\text{x}}_{1}{\:+\:8.64\text{x}}_{2}-{69.46\text{x}}_{3}+{4.86\text{x}}_{1}{\text{x}}_{2}+{{15.24\text{x}}_{1}{\text{x}}_{3}+8.21\text{x}}_{2}{\text{x}}_{3}-{192.88\text{x}}_{1}^{2}-{119.88\text{x}}_{2}^{2}{\:+\:29.14\text{x}}_{3}^{2}$$
6
$$\:\text{M}\text{B}\text{i}(\text{m}\text{g}/\text{g})=21.15+{2.21\text{x}}_{1}{\:+\:0.1695\text{x}}_{2}+{0.1219\text{x}}_{3}+{0.9326\text{x}}_{1}{\text{x}}_{2}+{1.14\text{x}}_{1}{\text{x}}_{3}+{0.0714\text{x}}_{2}{\text{x}}_{3}-{4.10\text{x}}_{1}^{2}-{0.1151\text{x}}_{2}^{2}{\:+\:0.1791\text{x}}_{3}^{2}$$
7
3.2. Analysis of variance
An analysis of variance (ANOVA) was applied to further substantiate the suitability of the models. The ANOVA of iodine index quadratic model listed in Table 3 shows that the model is significant with the F-value of 10.76 and Prob > F of 0,0005. There is therefore a 0.05% chance that variation may occur due to noise. Adequate precision (AP) measures the signal/noise ratio, with a ratio greater than 4 being desirable [24]. In this case, the ratio of 10.80 for the iodine index indicates a very suitable signal, which means that this model can be used to navigate the design space. Prob > F values below 0.05 indicate the significance of the model terms [25]. In this case, as shown in Table 3, only x3, x12 and x22 are the significant terms.
Still in Table 3, the analysis results for the methylene blue index response during GPa synthesis show that the signal/noise ratio (adequate precision) of 12.21 indicates an adequate signal. The low predicted R2 value implies that several non-significant effects have been included in the model. Thus, the main effect of volcanic ash mass (x1) and its quadratic effect (x12) are significant for the model response. As for the interactions, x1x2 and x1x3 are significant. An adequate precision of 12.21 also indicated an adequate signal, allowing us to deduce that this model can also be used to navigate in the design space.
Table 3
ANOVA for the response surface model for both responses.
| Ii | | MBi |
---|
Source | DF | Sum of squares | F-value | Prob > F | Source | DF | Sum of squares | F-value | Prob > F |
Model | 9 | 44327.08 | 10,76 | 0,0005 | Model | 9 | 16.58 | 16.88 | < 0.0001 |
x1 | 1 | 1364.96 | 0,3313 | 0,5776 | x1 | 1 | 48.74 | 49.62 | < 0.0001 |
x2 | 1 | 745.67 | 0,1810 | 0,6796 | x2 | 1 | 0.2872 | 0.2924 | 0.6005 |
x3 | 1 | 48240.30 | 11,71 | 0,0065 | x3 | 1 | 0.1486 | 0.1513 | 0.7055 |
x1x2 | 1 | 189.12 | 0,0459 | 0,8347 | x1x2 | 1 | 6.96 | 7.08 | 0.0238 |
x1x3 | 1 | 1857.65 | 0,4508 | 0,5171 | x1x3 | 1 | 10.37 | 10.56 | 0.0087 |
x2x3 | 1 | 538.78 | 0,1308 | 0,7252 | x2x3 | 1 | 0.0408 | 0.0416 | 0.8425 |
x12 | 1 | 1.023E + 05 | 24,83 | 0,0006 | x12 | 1 | 46.27 | 47.11 | < 0.0001 |
x22 | 1 | 39472.47 | 9,58 | 0,0113 | x22 | 1 | 0.0364 | 0.0371 | 0.8511 |
x32 | 1 | 2334.71 | 0,5666 | 0,4690 | x32 | 1 | 0.0882 | 0.0898 | 0.7706 |
R2 = 0.9064 Adj-R2 = 0.8721 Pred-R2 = 0.2720 Adequate precision = 10.8064 | R2 = 0.9382 Adj-R2 = 0.8826 Pred-R2 = 0.0200 Adequate precision = 12.2123 |
3.3. 3D response surface plots
Table 3 clearly shows that phosphoric acid concentration has the strongest effect on the iodine index due to the highest F-value of 11.71, while volcanic ash and rice husk ash masses and had weak effects on this response with F-values of 0.33 and 0.18 respectively. The effects of the interactions are all inconsistent. Nevertheless, the effect of the interaction between volcanic ash mass and phosphoric acid concentration is higher than the other two, with an F-value of 0.45.
The 3D response surface plot of the interaction between volcanic ash mass and phosphoric acid concentration shown in Fig. 3 shows that the iodine index is highest when the volcanic ash mass is at the ash end of the range and the phosphoric acid concentration decreases.
For the methylene blue index response, we can see that the effect of volcanic ash mass is more significant with an F value of 49.62, while the other two effects, namely rice husk ash mass and phosphoric acid concentration, are similar and inconsistent (Table 3). The effect of the volcanic ash mass/phosphoric acid concentration interaction is greater than that of the volcanic ash/rice husk ash mass interaction, with F values of 10.56 and 7.08 respectively. On the other hand, the effect of the interaction rice husk ash mass/phosphoric acid concentration is weak and inconsistent with an F-value of 0.04.
The 3D plot representing the variation of the methylene blue index response with the interaction of volcanic ash mass/rice husk ash mass (Fig. 4) shows that this response is maximal in the center of the domain. However, we note that an increase in the mass of volcanic ash and the mass of rice husk also has a positive effect on the methylene blue index response. For the volcanic ash mass/phosphoric acid concentration interaction at a fixed rice husk ash mass (x2 = 1.5 g) shown in Fig. 5, we observe that an increase in methylene blue index is due to a simultaneous increase in volcanic ash mass and phosphoric acid concentration. Indeed, the formation of geopolymers from volcanic ash and phosphoric acid can increase the specific surface area and adsorption capacity of the resulting materials.
3.4. Responses optimization
When optimizing GPa synthesis, the aim was to maximize iodine and methylene blue indices. By exploiting the response surfaces, optimal conditions during GPa synthesis were obtained for volcanic ash mass = 3.72 g, rice husk ash mass = 1.97 g and phosphoric acid concentration = 5 M with predicted responses of 704.23 mg/g and 21.33 mg/g respectively for iodine and methylene blue indices with a desirability of 0.84. However, the experimental values obtained under the same conditions for iodine index and methylene blue index were 703.88 and 21.82 mg/g respectively (see Table 4).
Table 4
GPa synthesis parameters optimization
| Variables | Iodine index (mg/g) | Methylene blue index (mg/g) | Desirability |
---|
| X1(g) | X2 (g) | X3 (M) | Exp | Pred | residual | Exp | Pred | residual | |
GPa | 3.72 | 1.97 | 5.00 | 703.88 | 704.23 | 0.35 | 21.82 | 21.33 | 0.49 | 0.94 |
3.6. Feedstocks and geopolymer characterization
The chemical composition of volcanic ash and rice husk ash determined by XRF analysis is given in Table 5. For volcanic ash, SiO2, Al2O3, Fe2O3 and CaO are the main oxides with 41.52, 15.90, 14.74 and 9.67 wt.%, respectively. SiO2/Al2O3 ratio (2.61) and SiO2 + Al2O3 sum (57.42) in VA present this material as those used for geopolymer synthesis with their basic ingredient values contained within the ranges reported in the literature [26–29]. It was observed that after heat treatment, RHAs contain mainly SiO2 (80.20 wt.%) with a significant amount of K2O (4.95 wt.%).
Table 5
Chemical composition of VA and RHA.
Oxides | SiO2 | Al2O3 | Fe2O3 | CaO | MgO | TiO2 | Na2O | K2O | P2O5 | LOI |
---|
VA | 41,52 | 15,90 | 14,74 | 9,67 | 8,29 | 3,45 | 2,30 | 0,63 | 0,75 | 2,44 |
RHA | 80,20 | 1,54 | 0,48 | 0,57 | 3,22 | 0,09 | 0,15 | 4,95 | 3,03 | 5,63 |
The FTIR spectra of rice husk ash, volcanic ash and geopolymer are shown in Fig. 6. The spectrum of RHA (Fig. 6a) shows bands around 873, 793 and 466 cm− 1 which are respectively attributed to the symmetrical stretching vibration of Si-OH and/or Al-OH, Si-O or Al-O and to the bending vibration of Si-O-Si. The band at 1091 cm− 1 is attributed to asymmetric stretching of the SiO4 unit. For VA and GPa, the absorption bands at 1007 and 1091 cm− 1 are attributed to the asymmetric stretching vibrations of SiO4 units of various connections (Si-O-Si, Si-O-Al and Si-O-Fe) [30, 31]. The characteristic bonds of the O-H group and water molecules are located at 3428 cm− 1, 1646 and 1612 cm− 1 [32, 33]. The peaks appearing at 572 cm− 1 correspond to the vibration of the Si-O bond [33, 34] and the peak at 466 cm− 1 represents the vibration of the Si-O-Fe bond [3, 35].
The XRD spectra of rice husk ash, volcanic ash and geopolymer are shown in Fig. 7. The mineralogical composition of RHA (Fig. 7a.) includes calcite, CaCO3 (PDF# 86-2343), cristobalite, SiO2 (PDF#87-2096), diopside sodian, (Ca0.52Na0.29Fe0.10Mg0.09) (Mg0.057 Fe0.14Al0.27Mn0.01Ti0.01) (Si2O6), (PDF#85–1692), albite Na(AlSi3O8), (PDF#71-1156) and anorthoclase, Na0.75K0.25(AlSi3O8) (PDF#89–1459). In VA's DRX spectrum (Fig. 7b.), we find : albite disordered Na(AlSi3O8), (PDF#20–0572), Forsterite ferrous Mg1.641Fe0,359SiO4, (PDF# 88-1993) dolomite CaMg(CO3)2, (PDF#75-1759), anorthoclase, Na0.75K0.25(AlSi3O8) (PDF#89–1459), cristobalite SiO2, (PDF#76–0939), anorthite sodian disordered, (Ca, Na) (Si, Al)4O8, (PDF# 10–0360), Diopside manganian, Ca0.87Mn0.19Mg0.94Si2O6), (PDF#83-1834) and Labradorite Ca0.65Na0.32(Al1.62Si2.38O8), (PDF# 83-1367). Comparing the diffractograms of volcanic ash to that of optimized geopolymer (GPa), we can observe a decrease in the intensity of certain peaks, evidence of the dissolution of minerals during the geopolymerization process [36]. We also note the presence of a dome between 20° and 40°, which suggest that the rice husk ash contributed amorphous silica during geopolymerization.
3.7. Study of operating parameters for CV dyes adsorption by GPa
3.7.1. Effect of initial pH solution
This experiment was carried out at a pH between 2 and 12, a contact time of 60 min, an initial CV concentration of 60 mg/l, an adsorbent dose of 0.2 g and a temperature of 25°C. The results shown in Fig. 8. indicate that the adsorption rate of CV increases with increasing solution pH. When the solution pH is below the pHpzc (5.4) of the geopolymer material (GPa), protonation of the active binding sites was enhanced by the sharp increase in proton (H+), so the number of active sites became lower for sorption of CV molecules. On the other hand, at pH above pHpzc, the surface acquires an increasingly negative charge from 6 onwards, increasing the adsorption of cationic CV dye molecules due to electrostatic interaction [37–40].
3.7.2 Effect of adsorbent dose
To assess the influence of adsorbent dose on the percentage adsorption of CV by GPa geopolymer, given quantities of GPa in the range (0.1–0.6 g) were dispersed in a series of experiments containing 30 ml for an initial CV concentration of 60 mg/l. The solution was then stirred at room temperature at pH 6 for 60 minutes. The effect of adsorbent dose on crystal violet removal is shown in Fig. 9. From this figure, we can see that increasing the adsorbent mass leads to an increase in the adsorption rate. This can be attributed to the increase in surface area available and the increase in the number of active sites on the surface of the adsorbent used [12, 37]. On the other hand, the decrease in adsorption rate observed at 0.5 g may be due to agglomeration and overlapping of adsorption sites, resulting in a reduction in the total number of sites accessible to the pollutant [41–43].
3.7.3. Effect of contact time and initial concentration
Thermodynamic equilibrium between the adsorbate in the liquid phase and the adsorbate bound to the solid is achieved at a rate that depends not only on the rate at which the constituents in the adsorbent and in the fluid, but also on the interaction between the adsorbent/adsorbate interaction. Studying the adsorption of a compound on an adsorbent enables us to examine the influence of contact time and initial concentration on retention. Figure 10. shows the effect of contact time and initial concentration on the adsorption of crystal violet by GPa. From this figure, we can see that the crystal violet has a good affinity for GPa. The shape of the curves shown is typical of saturation curves, with a slight qualitative and quantitative difference. We can see that CV adsorption took place very rapidly from the start of the experiment, and equilibrium was reached very quickly after 60 minutes. Beyond that, there was almost no further increase in adsorption. These results confirm that CV adsorption on the GPa surface is a spontaneous speed phenomenon in the initial moments, slowing down after 60 minutes in a state of adsorption saturation. This phenomenon can therefore be divided into two stages: a rapid initial phase due to the availability of the more frequently active sites that were vacant and spontaneously accessible to the CV particles, followed by a slower phase in which the remaining unoccupied surface sites diminish due to the repulsive forces between the solute and the solid phases [44, 45].
With regard to the influence of initial concentration, we note that the adsorption capacity of crystal violet increases with initial concentration. This may be due to the increasing mass gradient, which acts as a driving force to overcome the mass transfer resistance of pollutants. Nevertheless, the increase in initial concentration can lead to the appearance of a plateau indicating saturation of the support, which can be explained by the exhaustion of all the active sites existing on the surface of the support [46, 47].
3.8. Adsorption kinetics
The design of adsorption treatment systems requires knowledge of kinetic processes due to the diversity of chemical systems, the nature of the different adsorbents and the different designs of contact systems. For this reason, two liquid-phase adsorption kinetic models, the pseudo-first-order model [48] and the pseudo-second-order model [49], were used in this study to analyze experimental adsorption kinetic data.
Lagergren's first-order equation is one of the most widely used for the sorption of a solute from a liquid solution [50] and is represented as follows :
$$\:\text{ln}\left({q}_{e}-{q}_{t}\right)=ln{q}_{e}-{k}_{1}t$$
8
Where qe is the quantity of dye adsorbed at equilibrium (mg/g), qt is the quantity of dye adsorbed at time t (mg/g), K1 is the first-order reaction rate constant (L/min).
The expression of the pseudo-second-order model is in the form quoted by Ho and Mckay [49] :
$$\:\frac{t}{{q}_{t}}=\frac{1}{{k}_{2}{q}_{e}^{2}}+\frac{1}{{q}_{e}}t$$
9
Where k2 (g/mg/min) are the constants of pseudo-second-order model.
The curves for the two models are shown in Fig. 11. and the constants obtained from the different models are summarized in Table 6.
Table 6
Pseudo-first-order and pseudo-second-order adsorption rate constants for the different initial CV concentrations
| | Concentration (mg/L) |
---|
Kinetic models | Parameters | 20 | 40 | 60 | 80 | 100 |
| qe (mg/g) | 3.93 | 6.37 | 8.64 | 12.68 | 17.14 |
Pseudo-first order | qemax (mg/g) | 0.319 | 0.461 | 0.705 | 0.759 | 0.487 |
K1 (min− 1) | 0.039 | 0.042 | 0.046 | 0.032 | 0.032 |
R2 | 0,960 | 0.981 | 0.990 | 0.991 | 0.966 |
Pseudo-second order | qecal (mg/g) | 3.949 | 6.399 | 8.684 | 12.729 | 17.170 |
K2 (g/mg.min) | 0.379 | 0.263 | 0.187 | 0.131 | 0.214 |
h (mg/g.min) | 5.912 | 10.785 | 14.106 | 21.173 | 63.091 |
R2 | 0.999 | 0.999 | 0.999 | 0.999 | 0.999 |
The R2 values shown in Table 6 suggest that the pseudo-second-order model best describes the CV adsorption process on GPa. We also note that the adsorbed quantities calculated by this model are closer to those determined experimentally. This suggesting that chemisorption might be rate-determining step for controlling the adsorption process [51–53].
3.9. Adsorption isotherms
The adsorption isotherm is a simple tool, but it plays a very important role in understanding adsorption mechanisms and finding the best adsorbent for large-scale applications. These isotherms provide information on adsorbent/adsorbate affinity and an idea of the binding energy between adsorbate and adsorbent. Crystal violet adsorption per GPa was studied as a function of initial dye concentration. The results obtained were modeled using the two empirical models.
The Freundlich model uses the exponential distribution of adsorption sites and energies in an adsorption process, assuming that the sites on the adsorbent surface are heterogeneously distributed, and is represented by the following equation [54]:
$$\:{q}_{e}={k}_{F}{C}_{e}^{\frac{1}{n}}$$
10
Where, kF (L/g) and n are Freundlich constants for adsorption capacity and adsorption intensity respectively.
Langmuir's isothermal model (Eq. (11)) assumes a homogeneous adsorbent surface and monolayer adsorption.
$$\:{q}_{e}=\frac{{q}_{m}{K}_{L}{C}_{e}}{1+{K}_{L}{C}_{e}}$$
11
Where: qm is monolayer adsorption capacity (mg/g), KL is the Langmuir isotherm constant related to the affinity of the binding sites and energy of adsorption (L/mg).
The estimated model parameters with the correlation coefficient (R2) for the different models are presented in Table 7. The R2 values are taken as a measure of the goodness of fit of the experimental data to the isotherm models. The applicability of the two isotherm models to current data follows the following order: Langmuir > Freundlich. The essential characteristics of the Langmuir isotherm can be expressed in terms of the dimensionless separation parameter RL, which indicates the shape of the isotherm that predicts whether an adsorption system is favorable or unfavorable [55]. RL is defined as follows [56] :
$$\:{R}_{L}=\frac{1}{1+{R}_{L}{C}_{0}}$$
11
The RL value for the current experimental data lies between zero and one, indicating favorable adsorption of crystal violet on GPa.
Table 7
The Langmuir and Freundlich isotherms model constants.
| | Langmuir isotherm | | Freundlich isotherm |
---|
Adsorbants | KL | qemax | RL | RMSD | R2 | KF | n | RMSD | R2 |
GPa | 0.235 | 14.668 | 0.040 | 0.498 | 0.996 | 4.187 | 2.117 | 0.166 | 0.987 |