Fourier-transform infrared spectroscopy (FTIR) and attenuated total reflectance (ATR) appetizer (Nicolet 380 FTIR, Thermo Scientific, Waltham, MA, USA) with OMNIC ™ software were used to determine possible interactions between MB particles and starch. The formulations of MB adsorbed on every 50 mg of starch (SN, SM1, SM2 and SM3) were dried at 40 °C and compared with a physical mixture of MB and starch in a weight ratio of 1:10 and with the standards of the pure substances. The spectra of powders were recorded at wavelengths of 400 to 4000 cm^-1 at 32 scans per sample and a resolution of 4 cm^-1.

The analysis of starch was supplemented by X-ray diffraction measurements (PXRD). The powder PXRD data were recorded on a Bruker D2 PHASER diffractometer (Bruker AXS, Karlsruhe, Germany) with a Lynxeye detector using Cu Kα radiation (0.15418 nm). All samples were measured at 295 K with 3.0 mm slit and 1.0 mm shutter. Diffractograms were obtained between 7.5° and 40° (2) (step size of 0.02° (2) and 0.25 s per step). The X-ray generator operated at 30 kV and 10 mA. The PXRD patterns were processed using the software Diffrac.Eva V 3.2. (Bruker AXS). The percentage of crystallinity and amorphousness of the tested starch samples in relation to SN was determined.

Differential scanning calorimetry (DSC 214 Polyma, Netzsch, Selb, Germany) was performed to investigate the samples of dried adsorbed MB on 50 mg starches (SN, SM1, SM2, SM3), their physical mixtures of 1:10 w/w and pure ingredients. The grated samples of 3 to 5 mg were explored in aluminium pans with lids under a nitrogen atmosphere, with a flow rate of 50 mL min^-1. The thermograms were recorded at a constant heating rate of 5 °C min^-1 in the temperature range from 0 to 350 °C.

The absorbance study was performed using four 50 mL conical flasks (A, B, C, D) as a series. 40 mL 6 mg L^-1 MB was added for each flask. The SN, SM1, SM2 and SM3 starch probes of 2, 5, 10, 25, 50, 125, 250, 500, 750 and 1000 mg were placed in the flasks during the series, respectively. The flasks were put on the orbital shaker (at 100 rpm) at 22±0.5 °C. Four measurements of absorbance were proceeded based on the pharmacopoeial method by sampling the MB solution volumes of 3 mL in every 5, 10, 15, and 20 minutes, and than in equal periods up to 4 h. The solution was returned to the flasks. Analysis of taken samples was done by the spectrophotometer UV/VIS Jasco V-530 (Tokyo, Japan) at 663 nm, according to the available bibliography [37-39], and compared to the absorption spectrum of MB. A standard curve based on three series of measurements with five concentration points from 0.5 to 5.0 mg L^-1 was prepared. The results were examined according to pseudo-zero-order kinetics and first-order kinetics. ANOVA statistical test was performed for independent groups at the significance level α = 0.05. Freundlich, Langmuir and BET isotherm models were taken for analysis of adsorption processes (Table 2). BET isotherm equation was adapted to form BET isotherm for liquid phase adsorption [40], where k_BET(s) applies to the equilibrium constant of adsorption of the first layer and k_BET(L) refers to the equilibrium constant of adsorption of upper layers.

The values of q and q_t were calculated by the equations (1) and (2):

where C₀ is initial concentration of the solution (mg L^-1), C_t is concentration of the solution in time t (mg L^-1), V is volume of solution (L) and m is adsorbent mass (g).

Equation (3) was used for the adsoprption (%) calculation:

where m₀ - the initial mass of the adsorbate, m_t - is the adsorbate mass after time t.

The FTIR spectrum of pure MB showed a broad peak from 3219 to 3364 cm^-1, indicating an O-H bending bond. The sharp peaks at 3050 cm^-1 and in the range of 817 to 884 cm^-1 were responsible for the presence of C-H or =C-H bonds in the aromatic ring. Similarly, a sharp peak around 2700 cm^-1 may have occurred due to the C-H stretching bonds [41].

Characteristic peaks interacting at 1596 cm^-1 might have been assigned to N=C stretching bonds. The peaks in the range 1420-1490 cm^-1 most likely belonged to the stretching C=C-C bonds in the aromatic ring. Vibration from 1340 to 1360 cm^-1 revealed the presence of C-N stretching bonds in aromatic tertiary amine [41] and peaks from 1059 to 1064 cm^-1 and 661 cm^-1 are related to C-S-C stretching bonds in heterocycle [42]. The pure starches FTIR spectra indicated a broad peak of O-H stretching bond at about 3300 cm^-1, C-O stretching vibration at the range 1185-1347 cm^-1, especially the strong peak at 1240 cm^-1 for SM1 (Figure 3 b), C-O stretching bond at 930 cm^-1 [43] and O-H bending bond at 988 cm^-1 [44] (Figure 3).

The plot of the spectrum of adsorbed MB on the starch (red line) mostly followed the plot of the spectrum of the respective starch in its pure state (green line). However, there were locations where individual peaks were in excess. The peaks assigned to N=C stretching bonds at about 1600 cm^-1 appeared in the case of MB adsorbed on the SN and SM2, however, to a very slight extent (Figure 3 a,c). The peak indicating the presence of the C=O bond of the acetyl group at 1730 cm^-1 [45] was spotted on the SM1 spectra (Figure 3b).

Due to the appearance of peaks at 1599 cm^-1 assigned to the C=N bond on the FTIR spectrum of samples of adsorbed MB on SN and SM2, we postulate the existence of an interaction between the N⁺ ion derived from MB and the OH^- group belonging to starch, which probably leads to enhanced adsorption of the dye on the polymer. This observation coincides with the mechanism of MB adsorption on black olive stone carried out by Al-Ghouti et al. [48].

According to Table 3, the SN was the most crystalline, and SM2 the least. The crystallinity of SM increased according to the following pattern, SM2<SM3<SM1<SN, which suggests that the implementation of ionic groups resulted in the highest amorphicity, whereas the non-ionic acetyl groups in the lowest extent influenced the increase of amorphicity. This remark agrees with the results of Bhargav et al. [49], who assessed the combination of PVA with NaI. As the salt content of the polymer increased, the DSC diffraction peaks became less intense. They suggested high susceptibility of the amorphous structure of the polymer to ionic diffusivity of ionic particles. Similarly, Hodge et al. [50] found a correlation between peak intensity and the degree of crystallinity. They observed that the addition of water resulted in a decrease in XRD peak intensity and an increase in PVA amorphousness.

The DSC thermograms of pure MB in the first heating cycle displayed an endothermic peak at about 80 °C responsible for the dehydration (Figure 5 a-d). The thermogram of the second heating cycle no longer showed this peak, while an endothermic peak in the 100 °C range appeared (Figure 5 e-h). MB forms several hydrates. In addition, it reacts with an aluminum container for DSC [51]. This endothermic peak may indicate a transition from one form of MB hydrate to another. In the thermogram of the first heating cycle, it was most likely obscured by the broad peak of dehydration. These speculations are confirmed by the shifted exothermic peaks occurring at 171 and 233 °C in the first heating cycle (Figure 5 a-d) and 201 °C in the second cycle (Figure 5 e-h). They testify to the decomposition of MB [52]. Omer et al. [53] indicated the occurrence of oxidation of MB at 260 °C. A difference was observed between the spectrum of adsorbed MB and physical mixed MB. An exothermic peak appeared, indicating the presence of MB in the physical mixture with SN (177.6 °C), SM1 (175.9 °C), SM2 (174.0 °C) and SM3 (177.2 °C) (Figure 5 a,b,c,d respectively). No MB peak was observed in MB adsorbed at SN and SM1 (Figure 5 a,b), which may indicate the absence or complete decomposition of MB. For MB adsorbed on SM2 and SM3, the course of these peaks was flattened at 174.4 and 176.7 °C, respectively, which may indicate the interaction of MB with these starches (Figure 5 c,d).

Glass transition temperature T_g value readings showed the lowest T_g for SM2 (47.2 °C) and the highest for SM3 (55.9 °C) (Table S1). Hydrogen bonds are among the factors affecting T_g [54]. Taylor et al. [55] observed significantly lower T_g of sucrose compared to other disaccharides. This was due to the less extensive hydrogen bonding network in the amorphous state. Similarly, Zhou et al. [56] demonstrated the dependence of T_g on hydrogen bonds occurring between polymer and copolymer particles (maleimide-isobutene alternating copolymers). A network of hydrogen bonds of carboxyl groups resulted in a higher T_g value than hydrogen bonds between hydroxyl groups. The lowest T_g occurred in the absence of hydrogen bonds between polymer molecules. In our case, the holders of the hydroxyl groups were SN and SM1 (T_g 52.2 and 51.15 °C respectively) (Table S1) and the carboxyl groups were present in the SM2 molecule. In addition, phosphate groups appeared in the SM3 molecule. High T_g values for SM3 may suggest the self-association of these starch molecules, whose T_g decreased to 53.0 °C after bond formation with MB [55]. In the case of SN, T_g values increased after MB adsorption (54.7 °C), what may indicate low self-association of SN particles in the pure state.

In the present work, the values of ΔT were assessed to determine the thermal stability of the tested samples. The higher the value of ΔT, the better the thermal stability. The value for SN was the highest (221.07 °C), demonstrating the highest thermal stability of this starch. At the same time, its highest percentage of crystallinity was confirmed (Table 3). On the contrary, SM2 showed the lowest value of ΔT (179.8 °C) - the lowest thermal stability. Keshk et al. [43] demonstrated the dependence of thermal stability on ΔT-values, which correlated with the degree of crystallinity of corn starch-cellulose mixtures, depending on the mixture ratio. They found an inversely proportional relationship between ΔT values and ΔC_p values. Our ΔC_p studies confirmed such a relationship outside of the SM2 samples having the lowest ΔC_p value (0.465 J g^-1·K^-1) according to the lowest ΔT (Table S1). This phenomenon may be supported by the observation of Liu et al. [57]. They studied various starches whose structure was more or less amorphous. According to them, amorphousness gives the compounds a liquid-like structure. More amorphous compounds have a less ordered structure than those with a crystalline structure. They showed that the process of lowering the temperature below T_g leads to the reorganization of the structure. In our study, SM1, SM2 and SM3 starches were hardened at 2 degrees below T_g. Moreover, the research is confirmed by PXRD studies, where SM2 had the highest degree of amorphousness (Table 3).

The thermograms of the second heating cycle can show the structure of the compound when structural reorganization occurs after the cooling cycle. In the case of SM2 there was an endothermic peak at 215 °C, from which an increase in heat flow is observed. The thermograms of adsorbed MB on starches are characterized by different temperatures of the start of the exothermic process. The increase in heat flow started fastest for SM2_MB (adsorbed) at 157 °C and slowest for SM3_MB (adsorbed) at 163 °C (Figures 5g and 5h, respectively). This endothermic event might be attributed to reduced hydrogen bonding as well as the interference of molecular organization due to the interaction between SM2, SM3 and MB. For SN_MB (adsorbed) and SM1_MB (adsorbed), no increase in heat flow was observed (Figures 5e and 5f).

MB adsorption on starch predominantly matched a pseudo-second-order model. A study by Lin et al. [58] of MB adsorption on activated carbon indicated a better pseudo-second-order fit to both linear and nonlinear models. Similarly, Ma et al. [45] matched the adsorption of heavy metal ions on modified potato starch to the pseudo-second-order model. In our study, the highest k₂ values obtained for SN, SM3, SM2 and SM1 were 1.9623, 1.5367, 0.9240, 0.1848 g mg^-1·min^-1, respectively, and occurred at a weight of 1000 mg for SN and SM1 and 2.0 mg for SM2 and SM3 (Table S2). This suggests that MB adsorbed fastest on SN, but this was dependent on the surface area of the adsorbent: the larger the surface area, the faster the saturation. In the case of SM2 and SM3, MB adsorbed fastest already on the smallest starch particle, suggesting a high affinity of the adsorbate to the adsorbent. The slowest adsorption was observed in SM1.

The q values in Table S2 differ from q_m obtained according to Langmuir and BET isotherms. Moreover, some of the Langmuir adsorption constants (q_m and K_L) showed negative values (Table 4). Langmuir isotherm is known to get different parameters according to the calculation method [59-61], therefore, we used the course of the isotherm of MB on the starches to establish q_m values (Figure 9).

Methylene blue is a substance that has been thoroughly studied by scientists because of its harmfulness to the environment. As a dye used in industry, it leaks into water waste in large quantities. Their purification is a major challenge for researchers. Many study substances that are adsorbents for MB and can serve as filters [32,61-65]. Some of these include starches and other polysaccharides. Table 5 shows the q_m values of example adsorbents for MB.

The sigmoidal course of the SM2 and SM3 starch isotherms falls into type IV according to the IUPAC classification. SN and SM1 isotherms are more aligned with type VI [76]. Type IV represents single and multilayer adsorption plus capillary condensation. Type VI illustrates that the adsorption isotherm can have one or more steps. In our study, the adsorptions on SN and SM1 starches best fitted to the Freundlich isotherm model. On the other hand, for SM2 and SM3 starches, the adsorption courses more closely matched the BET isotherm (Table 4).

The Freundlich model describes multilayer adsorption on a surface, with an inhomogeneous distribution of binding sites and, therefore, an inhomogeneous energy distribution. The constant k_F expresses the maximum adsorption on the surface of the adsorbent. The BET isotherm describes the phenomenon of physisorption on a non-microporous surface. The constant k_BET is related to the adsorption energy. Bhattacharyya et al. [36] found graphene oxide-potato starch almost perfectly fitted to the Freundlich isotherm model. Dipa et al. [37] adapted MB adsorption on kaolinite to the Langmuir model. Bestani et al. [62] studied MB adsorption on carbon-activated leaves, which they fitted to the Langmuir and Freundlich models. El Qada et al. [31] matched MB adsorption on activated carbon to Redlich-Peterson and Langmuir isotherm models. The Redlich-Peterson model is a compilation of the elements in the equations of the Langmuir and the Freundlich isotherms. Langmuir isotherm mostly applies to chemisorption on a monolayer. Activated carbon is highly porous and is a well-known adsorbent used in industry as a filter [77]. The dependence of the isotherm model on the structure of the adsorbent can also be supported by the study of Ma et al. [45], who used potato starch modified with high temperature to increase its porosity. Adsorption of heavy metal ions on such modified starch showed good adherence to the Langmuir model. In the case of our experiment, only SM2 was subjected to high-temperature modification at 100 °C. This could most likely increase its adsorption properties.