3.3. XRD Analysis
To determine the types of IONPs, XRD analysis was necessary. As shown in
Figure 3a, the X-ray diffractograms of IONPs confirmed that the nanoparticles have peaks at 2θ positions of 18.61°, 30.57°, 35.82°, 43.47°, 53.71°, 57.35°, and 63.01°, which corresponded to the magnetite [
37], leading to the conclusion that IONPs were SPION.
Based on
Figure 3a at a reflective peak 2θ of 35.82°, combining the XRD and the Scherrer equation [
38], the SPION size was calculated to be 29.54 nm with a shape factor of 0.89, which was in accordance with the FE-SEM analysis. Moreover, the peaks correspond to their hkl indices of (220), (311), (400), (422), (511), and (440), which are similar to the literature [
39,
40]. As shown in
Figure 3b, for SPION/PVA/GR, the sharp peak at 26.63° and 54.69° indicates that GR exists in the composite, corresponding to their hkl indices of (002) and (004) planes, which is similar to the literature [
41,
42]. Additionally, as shown in
Figure 3b, the SPION peaks have shifted slightly to 18.59°, 30.36°, 35.71°, 43.51°, 57.36°, and 62.94°. Hence, the XRD confirmed that IONPs are SPION, and the adsorbent consists of GR and SPION.
3.4. FTIR Analysis
FTIR is an effective instrument to quantitate and determine the functional groups of the nanoparticles, which confirm the final structure of the material. The FTIR spectra were analyzed using the Brucker Sensor 27 in Germany. As seen in
Figure 4, IONPs have the characteristic magnetite peak due to the Fe-O-Fe band splitting into two peaks, corresponding to the first and second bands at 586 and 441 cm
−1, which is the Fe-O bond of bulk magnetite [
43]. At the wavelengths of 3429 and 1629 cm
−1, the peaks corresponded to the bending vibration of absorbed water and surface hydroxyl—O-H stretching mode, C=O stretching vibration, respectively [
1,
44,
45,
46] This FTIR analysis confirmed that the IONPs were, in fact, SPION.
As shown in
Figure 4, the PVA FTIR spectra shows peaks at 3423, 2923, 1743, 1644, 1457, 1381, 1100, and 851 cm
−1, which correspond to the O-H stretching vibration, CH
2 asymmetric stretching vibration, C=O stretching vibrational band attributed to the carbonyl function groups due to the residual acetate groups from the hydrolysis process of polyvinyl acetate, C=O carbonyl stretch, CH
2 bending, C-H dseformation vibration, C=O stretching, and C-C stretching vibration, respectively [
47,
48,
49,
50]. As shown in
Figure 4, the GR FTIR spectra shows peaks at 3422 and 1644 cm
−1, which correspond to the O-H stretching vibration, and C=C [
51]. The peaks at 2923, 2853, 1265, and 1059 correspond to the asymmetric stretching vibration of CH
2, the symmetric stretching vibration of CH
2, the C-O stretching of the epoxy group, and the C-O stretching of the alkoxy group, respectively [
52]. The FTIR spectra of SPION/PVA/GR in
Figure 4 showed the peaks at 3435, 2923, 1724, 1630, 1266, 1059, and 587 cm
−1 which matched the peaks of SPION, PVA, and GR, indicating that the adsorbent has been assembled successfully as the proposed structure in
Figure 2d.
3.7. Adsorption
The methylene blue concentration was measured using UV-Vis spectrometry (Jasco V-730, scan speed 40 nm/min, data interval 1 nm, response 0.06 sec, filter exchange step). As the absorbance for methylene blue is at 664 nm, similar to the literature [
58], the calibration curve was also obtained.
Via the UV-VIS analysis, the loading amount (
), percent loading capacity (%LC), and entrapment efficiency (%EE) for the adsorption process of methylene at 273.15, 303.15, and 333.15 K can be seen in
Table 3.
As shown in
Table 3, the loading amount and the loading capacity do not change dramatically as the temperature increases. Comparing the loading amount between temperatures, the adsorption percentage increases slightly as the temperature rises, indicating the process to be endothermic, which is in accord with the
[
59,
60]. However, as the temperature increased to 333.15 K, the adsorption capacity decreased slightly. This could be because the adsorption forces between the active sites of the adsorbents and methylene blue decreased [
61,
62] due to the increase in mobility of methylene blue ions [
63]. As the initial methylene blue concentration increased, the loading capacity increased due to the high driving force for mass transfer at a high initial concentration [
64].
After 45 h at T = 333.15 K, the UV-VIS peak wavelength of the aliquots shifted from 664 nm to 657–660 nm. This might be caused by the degradation or changes in the structure of methylene blue, which led to the unreliable concentration of the remaining methylene blue in solutions. However, for 298.15 K and 310.15 K, the peak wavelength was 664 ± 2 nm. This indicates that the change in the structure of MB was insignificant.
Comparing the experimental equilibrium adsorption capacities obtained (as shown in
Table 1), these values were much smaller compared to some [
65,
66,
67] literature and much greater compared to other literature [
64,
68].
To determine whether the adsorption process favors higher or lower temperatures, thermodynamic studies were conducted, and the results yielded that the adsorption process was non-spontaneous, and exothermic, and the randomness decreased, as shown in
Table 4.
As shown in
Table 4,
,
, and
shows that the adsorption process was feasible, spontaneous, endothermic, with weak chemical forces between methylene blue and the adsorbents, the randomness increasing on the surface, and some changes occurring in the internal structure of the methylene blue and the adsorbent during the adsorption process [
59,
60,
69,
70,
71]. As
becomes more negative as temperature increases, this indicates that the adsorption process of MB on the adsorbents becomes favorable at higher temperatures via physical force [
60,
72]. With the
values smaller than 40 kJ/mol, the adsorption process can be considered a physisorption process [
60,
70]. In this case, since the
value is smaller than 20 kJ/mol, the physisorption interaction is dominated by Van der Waals forces [
73]. The small positive value of
also indicates that the adsorption was endothermic and physical, involving weak forces of attraction, weak electrostatic interactions, and the existence of loose bonding between methylene blue and the adsorbents [
74,
75,
76]. Moreover, the positive value of
also indicates the occurrence of monolayer adsorption [
69]. This result also shows that as the temperature increases, the degree of adsorption increases [
70]. Endothermic adsorption may be caused by the stronger interaction between the adsorbent and pre-adsorbed water than the interaction of cationic dyes with the adsorbent [
77]. In this case, the methylene blue and water molecules compete for the active sites of activated charcoal, leading to simultaneous adsorption and desoprtion of both types of molecules, resulting in positive
[
78,
79]. Different adsorption isotherm models (Langmuir, Freundlich, Dubinin-Radushkevich, Temkin and Pyzhev, and Halsey) were built, and various adsorption isotherm constants and variables were calculated as shown in
Table 5.
Based on the R
2 values of the isotherm models, which were all greater than 0.5, from
Table 5, all the models can be used to fit the experimental values. To determine whether the adsorption of the adsorbate over the adsorbent was favorable, the Langmuir adsorption isotherm (as shown in
Figure 6) can be used [
2,
16] which can be calculated using Equation (25).
where
is calculated from the Equation (7) or (8). If
, then the adsorption was favorable [
2,
16].
As shown in
Table 5, the Langmuir adsorption isotherm models (as shown in
Figure 5) can confirm that the mechanism of attaching methylene blue onto the surface of SPION/PVA/GR was adsorption linear due to the average R
L = 1 [
60,
80]. This indicates that MB was adsorbed as a monolayer onto the homogeneous surface of the adsorbent [
14].
However, the negative Langmuir isotherm constants show no physical meaning and are unacceptable [
81]. Hence, the linearized Langmuir isotherm model should not be considered the best-fitted model for all three temperatures.
After fitting the experimental values with the Freundlich isotherm model, the
value, which was obtained from the Freundlich isotherm model, as shown in
Figure 7, shows the non-favorable physical process.
The
value which is greater than one shows the adsorption process was cooperative adsorption [
82,
83]. Moreover, the Freundlich isotherm model also yields values of n
F that are smaller than 1. These values indicate that the bond energies increase with surface density [
84]. Moreover, the values of n
F also represent the poor adsorption characteristic [
60].
The Freundlich isotherm model determined that the adsorption process was a non-favorable physical process. Hence, to determine whether the adsorption process was chemical and confirm the non-physical adsorption process, the Dubinin-Radushkevich model was built [
85]. As shown in
Figure 8, since K
DR was less than unity, the adsorbent’s pore structure and the interactions between adsorbent and adsorbate caused the increase in surface heterogeneity [
86].
The magnitude of E indicates that the adsorption process was chemical ion-exchange because chemical ion-exchange occurs at a magnitude of E between 8 and 16 kJ/mol and physical adsorption occurs at a magnitude of E less than 8 kJ/mol [
87,
88,
89]. Hence, at 273.15, 310.15, and 333.15 K, the adsorption processes were chemical ion-exchange, chemical ion-exchange, and physical sorption, respectively [
90].
The Freundlich isotherm model is quite similar to the Halsey isotherm model in evaluating the multilayer adsorption system and the heterogeneous surfaces with uniform surface heat distribution [
91,
92] The Freundlich isotherm model describes the exponential distribution of the active sites and their energies [
91,
92]. However, the Halsey isotherm model not only evaluates the multilayer adsorption system but also describes its condensation at a relatively large distance from the surface [
91,
92]. Hence, overall, at all three temperatures, the Halsey adsorption isotherm was the only model that did not fit the experimental data the worst, as shown in
Figure 9.
As shown in
Figure 10, the Temkin-Pyzhev isotherm model shows that all the molecules in the layer decrease linearly with coverage due to adsorbate/adsorbate interactions, and the adsorption is characterized by a uniform distribution of binding energies up to some maximum binding energy [
93,
94].
As shown in
Table 6, the intraparticle diffusion model showed that at 333.15 K, the adsorption was both film diffusion and intra-particle diffusion since
[
60].
From the data in
Table 6, at 298.15, 310.15, and 333.15 K, the I values showed that the film diffusion and intra-particle diffusion occurred at the same time [
95,
96,
97].
In addition to film diffusion, intra-particle diffusion, and chemisorption, the methlylene blue adsorption mechanism of SPION/PVA/GR could result from electrostatic interactions between the negatively charged surface and the positively charged methylene blue, hydrogen bonds, and the π−π* stacking with the methylene blue aromatic ring [
60,
70,
73,
95,
96,
97]. Additionally, intraparticle diffusion, boundary layer diffusion, and external diffusion can regulate the adsorption processes [
60,
75,
98].
3.8. Desorption
The percentage of release average is given as in
Table 7.
As shown in
Table 7, despite releasing MB at a constant temperature and pH of 3.85, the release percentage varies and depends on the initial loading conditions, such as temperature and initial MB concentration. As the initial loaded MB concentration increases, the release percentage decreases when MB is loaded at 310.15 and 333.15 K. As the loaded temperature increases, the release percentage decreases at the initial loaded MB concentration of 0.02 mg/mL. However, the standard deviation of the release percentage when loaded MB at 298.15 K was much larger than T = 310.15 and 333.15 K. Hence, the highest release percentage occurred when loaded MB was at 310.15 K and 0.017 mg MB/mL.
From
Table 7, the percentage of release average can be predicted, as shown in
Figure 11 and
Figure 12, based on the loading conditions after 7 days. The predicted models can be represented in 3D or 2D. In each model, the predicted equations were generated.
As shown in
Figure 11, the calculated/predicted values of the percentage of release were calculated using the following equation:
where x, y, p00, p10, p01, p20, p11, p02, p30, p21, and p21 is T
loaded in Kelvin, [MB]
loaded in mg/mL, 16.45, −7.527, −2.278, −0.01926, −1.112, −0.1102, 3.233, 0.05589, 1.59, respectively. The above equation yields R
2 and root-mean square values of 0.83 and 2.71, respectively.
On the other hand, in
Figure 12, the calculated/predicted values of the percentage of release were calculated using the following equation:
where T
loaded and [MB]
loaded in Kelvin and mg/mL, respectively. The above equation yields
of 3.16.
To determine whether the release temperature or the initial loaded MB concentration affect the percentage of average release, the Taguchi and Factorial methods in Mintab-17 software were used to generate the signal-to-noise (SN) SN ratios (larger is better, as shown in Equation (19)), interaction plots for SN ratios (larger is better), and a Pareto chart of the standardized effects, as shown in
Figure 13 [
99].
By using the Factorial method, in
Figure 13, the calculated/predicted values of the percentage of release were calculated using the following equation:
where T
loaded and [MB]
loaded in Kelvin and mg/mL, respectively. The above equation yields
of 3.05. As shown in
Figure 13a,b, as the temperature decreases, the average release percentage (%R) decreases. On the other hand, no trend was determined for the effects of the initial MB loading concentration on the %R. However, the optimum release conditions were 298.15 K and 0.018 mg/mL MB. As shown in
Figure 13b,c, the factor that affects %R the most is temperature since the horizontal bar passed the vertical line on the Pareto charts, which is statistically significant at a 95% confidence level [
100], despite some interactions (no parallel lines observed in
Figure 13b) between initial loading MB concentration and release temperature [
99].
As shown in
Table 8 and
Figure 14, the zeroth order, Higuchi, and Korsmeyer-Peppas released models of MB loaded at different temperatures and different initial MB concentrations released after 7 days at T = 298.15 K and pH = 3.85 were calculated.
As shown in
Table 8 and
Figure 14, the Korsmeyer-Peppas models yield the release exponent with values of 0.45 < n
KP < 0.89, which indicates that the releasing mechanism is anomalous diffusion or non-Fickian diffusion, which is the combination of diffusion and case-II relaxation [
101,
102,
103,
104]. Moreover, with the values of n
KP, the release kinetics are dependent on time. However, with an initial loaded MB concentration of 0.017 mg/mL at a loaded temperature of 298.15 K and an initial loaded MB concentration of 0.018, 0.019, and 0.02 mg/mL at a loaded temperature of 333.15 K, the n
KP was smaller than 0.5, indicating that the releasing kinetic was Fickian diffusion [
101]. However, with the Fickian diffusion scenarios, the
values were all greater than 1 (much greater than 0), indicating that the model did not fit the experimental values well. Overall, the best-fitted kinetic model for releasing MB at 298.15, 310.15, and 333.15 K with a pH of 3.85 is the Higuchi model due to the smallest
values. The Higuchi models indicate that the diffusion process was based on Fick’s law, which is root-time dependent [
105]. If diffusion via water-filled pores in the matrix under constant diffusivity primarily controls the release of the MB, the Higuchi model is applicable [
102,
106]. In other words, the release of MB was controlled by diffusion through the pores and cracks of the adsorbents [
102].