Synthesis and thermal analysis of the ternary complexes

The complexity of the starting materials, especially that of hazelnut oil, as well as the differences among their characteristics (hydrophobicity and water solubility) suggest the kneading method as the most appropriate one for obtaining β-CD/hazelnut (Corylus avellana L.) oil/flavonoid glycoside or flavonolignan ternary complexes. Kneading allows for higher recovery yields of complexes in comparison with the co-crystallization method because only small amounts of solvent are needed for preparation. On the other hand, similar methods such as spray-drying do not provide intimate contact between the three types of components for a sufficient period of time to reach the association–dissociation equilibrium [1,27,67]. In this study, the recovery yields were in the range of 51.5–85.3%, and significantly higher for the 3:1:1 complexes. Equimolar X1H, X1N, X1R and X1S ternary complexes were obtained with yields of 57.7 (± 8.8), 54.6 (± 1.9), 74.3 (± 1.8), and 64.7 (± 2.6)%, respectively. For the 3:1:1 ternary complexes (single samples) these yields were in the range of 74.5–85.3%. The difference in the yield can be explained by the level of hydration, as was determined by TG (see below). For the 1:1:1 complexes, the mass loss is half in comparison with the water content of β-CD (6.4–7.4% for complexes and 14% for β-CD hydrate). On the other hand, the mass loss of the 3:1:1 complexes is much higher (e.g., 11.8% for X3N complex). As a consequence, the 1:1:1 complexes lose relatively more hydration water than the corresponding 3:1:1 complexes. This can be explained by the high level of complexation for the 1:1:1 complexes. This aspect could be confirmed by thermal analysis, especially by DSC.

Both TG–DTG and DSC thermal analyses provide information about the molecular inclusion of guest molecules into the β-CD cavity. Unfortunately, these methods cannot differentiate between the encapsulated components and their entrapment efficiency. However, the goal of the study was the discrimination of such ternary complexes on the basis of FTIR. The evaluation of the encapsulation competitiveness of such multicomponent mixtures is very challenging (highly hydrophobic FA triglycerides, mono- and diglycerides, free FAs, as well as more hydrophilic flavonoid glycoside, namely hesperidin, naringin and rutin, or flavonolignan – silibinins). According to TG-DTG and DSC analyses, the ternary complexes are highly stable up to 200 °C. The TG and DTG plots were similar for ternary complexes at a 1:1:1 molar ratio, in comparison with the β-CD hydrate at temperatures up to ≈200 °C. The only significant difference was observed for the mass loss corresponding to water/moisture release up to ≈110 °C, with values of 6.37–7.38% and 9.45% for β-CD hydrate, respectively. A lower mass loss was observed for β-CD hydrate in comparison with the water content provided by the manufacturer (maximum 14% by oven drying). This could be due to the TG protocol, which assumes the pre-equilibration of the microbalance prior to analysis. Consequently, loss of surface water could have taken place before the start of the analysis. However, the difference of 2–3% for the ternary complexes at 1:1:1 molar ratios can be explained by a partial replacement of water molecules during the molecular encapsulation of the FA triglyceride and flavonoid guest molecules. On the other hand, the mass loss for the 3:1:1 ternary complexes was similar to the one observed of β-CD hydrate or even higher (see Supporting Information File 1, Figures S1–S4 and Tables S1 and S2). This means that a significant amount of β-CD is not involved in the formation of complexes and remains as β-CD hydrate in the mixture. These observations are in agreement with other studies on the complexation of vegetable (common bean lipids) and fish oil (common barbel, Pontic shad, European wels catfish, common bleak) by CDs [11,17]. Moreover, this TG behavior does not depend on the method of synthesis (kneading or co-crystallization) or the method of water determination (TG as mass loss or Karl Fischer water titration, KFT) [6,68]. It was observed that the difference between the water content or TG mass loss up to ≈110 °C is lower for binary complexes of CD/flavonoids in comparison with CD/fish oil (Atlantic salmon or European anchovy) [12,14,43]. The TG results are in agreement with the DSC data, where the calorimetric effect corresponding to water/moisture release is lower for the ternary complexes (378 J/g for X1N and 432 J/g for β-CD hydrate, Supporting Information File 1, Figure S5 and Table S3). There are two aspects that can be observed in the DSC but not in the TG–DTG analyses. The first aspect is the presence of two types of water molecules in the ternary complexes. They appear at two specific DSC peak temperatures of 44.5 °C for surface water and 82.0 °C for the stronger retained water molecules. While the surface water-related temperature is quite similar to β-CD hydrate, the stronger retained water has a higher DSC peak temperature value for β-CD (94.7 °C). This observation confirms the partial replacement of strongly retained water molecules during the complexation process. The second observation on DSC results is related to the absence of an endothermal–exothermal calorimetric peak in the case of the X1N ternary complex. This peak appears at 218.9 °C for β-CD hydrate and means that the complex obtained by kneading has an amorphous structure, in comparison with crystalline β-CD hydrate. The calorimetric peak observed for β-CD hydrate at this temperature is due to the transition of anhydrous β-CD (after water release) from the crystalline to the amorphous state [6]. Finally, TG analysis indicates a mass loss of 1.4–4.0% in the temperature range of 110–275 °C for the 1:1:1 ternary complexes and only 1.25% for the 3:1:1 complexes, whereas almost no mass loss was observed for β-CD hydrate (0.05%). The degradation of β-CD appears above 275 °C, with a maximum degradation rate at 299.4–326.0 °C as determined by DTG (the highest for β-CD) and at ≈322 °C by DSC. The degradation of the encapsulated hazelnut oil – of the triglyceride components – appears at a higher temperature of 394–407 °C (DTG and DSC).

Fourier-transform infrared spectroscopy (FTIR) of ternary complexes

FTIR is a fast method that allows the evaluation of the presence of a compound in a complex through specific absorption bands. β-CD consists of seven 1→4-linked α-ᴅ-glucopyranose units forming a macrocycle. As a consequence, the FTIR specific bands especially appear for OH, CC and CH/CH2 bonds and groups. However, CD specific bands also appear for CH groups in the CD ring and α-type glycosidic bonds. Thus, a broad FTIR band corresponding to the stretching vibration of the O–H bonds in β-CD and hydration water molecules appears at ≈3301 cm−1. A weak band for the asymmetric stretching vibrations of the C–H groups appears at 2924.8 (± 1.4) cm−1, while the bending vibrations (in-plane, asymmetric, and symmetric) of the OH and CH groups appear as weak bands in the range of 1205–1643 cm−1. The stretching vibrations of the C–O and C–C groups in the glucoside moieties appear as medium-strong bands in the range of 998–1152 cm−1. A specific band for CD appears at 939.2 (± 1.8) cm−1 and is assigned to the stretching vibrations of the C–H groups from the β-CD ring. Also, the band at 852.9 (± 0.8) cm−1 is attributable to the bending vibrations of the C–C–H groups related to the α-type glycosidic bonds in the CD. Other bands appear at wavenumbers lower than 800 cm−1 and were tentatively assigned to the bending vibrations of the CH and OCC groups (574–754 cm−1), as well as to the stretching vibrations of the CC bonds at 526.3 (± 1.3) cm−1[69,70]. Relevant data from the FTIR analysis of β-CD is presented in Figure 2 and Figure 3 and in Supporting Information File 1 (Figures S6–S11 and Table S4).

Figure 2: Superposition of the FTIR spectra for the β-cyclodextrin/Corylus avellana oil/hesperidin ternary complex at a 1:1:1 molar ratio (blue), β-cyclodextrin hydrate (red), C. avellana oil (pink), and hesperidin (green).

Figure 3: Superposition of the FTIR spectra for the β-cyclodextrin/Corylus avellana oil/hesperidin ternary complex at a 3:1:1 molar ratio (blue), β-cyclodextrin hydrate (red), C. avellana oil (pink), and hesperidin (green).

Vegetable oils and animal fat especially contain FA triglycerides, but mono-, diglycerides and free FAs also exist. As a consequence, the broad band corresponding to the stretching vibrations of the O–H groups is attributable to free fatty acids, monoglycerides, diglycerides and water. In the hazelnut samples, this band was observed at 3287.8 (± 10) cm−1. In this study, very useful was the weak band at 3005 (± 0.2) cm−1, which corresponds to the symmetric stretching vibrations of the =CH groups from the mono- and polyunsaturated FA moieties (especially oleic acid, but also palmitoleic and linoleic acids). The asymmetric and symmetric stretching vibrations of the CH groups provide strong bands at 2952.5 (± 0.3), 2922.5 (± 0), and 2853.2 (± 0) cm−1 due to the high number of CH2 and CH3 groups in the triglyceride structures. Another important and characteristic FTIR band for glycerides is that corresponding to the stretching vibrations of the ester C=O groups that appears as very strong band at 1744 (± 0) cm−1 for hazelnut oil. The stretching vibration of the cis-RHC=CHR’ group is observed as a weak band at 1652.7 (± 0.3) cm−1. Medium and strong bands are those related to the bending vibrations of the CH2 and CH3 groups at 1458.7 (± 0.2) cm−1, bending vibrations of the CH2 groups at 1236.8 (± 1.3) and 1158.1 (± 2.3) cm−1, the stretching vibrations of the C–O groups at 1027.9 (± 5.7) cm−1, as well as the out-of-plane bending vibrations in the C–H groups at 722 (± 0.1) cm−1. Degradation/isomerization of oil components (low level) can be observed at 956.7 (± 8.7) cm−1, where the band corresponding to the bending vibrations of the C=C groups in trans-RHC=CHR’ groups appears (sometimes at slightly higher values). Details of the FTIR analysis of hazelnut oil samples can be seen in Figure 2 and Figure 3 and in Supporting Information File 1 (Figures S6–S11 and Table S5) [71].

Hesperidin, naringin, and rutin are flavonoid glycosides derived from the corresponding flavanones hesperetin and naringenin and the flavonol quercetin, respectively. These compounds have a disaccharide moiety connected to the aglycones through an ether linkage with the hydroxy groups in the 7 and 3 positions (Figure 1a). On the other hand, silibinins (the main components of silymarin) are flavanonol derivatives, having a coniferyl alcohol moiety connected through the hydroxy groups in the 3’ and 4’ positions of the aglycone (Figure 1b). FTIR analysis of these flavonoids revealed stretching and bending vibrations corresponding to OH bonds (phenolic or alcoholic, glycosidic and OH groups from water molecules), CH bonds (especially from the CH2 and CH3 groups), bands corresponding to the aromatic CC bonds, and the carbonyl C=O bond. The most relevant FTIR band for these compounds is the asymmetric stretching vibration of the C=O bonds, νasC=O, which appears around 1633–1651 cm−1. The lowest value for this band was observed for silymarin at 1634.1 (± 0.4) cm−1 and the highest one for rutin at 1651 (± 0.1) cm−1. For hesperidin and naringin this band appears at approximately the same value (≈1645 cm−1). The stretching vibrations of phenolic, glycosidic or water O–H bonds appear as broad bands in the range of 3263–3541 cm−1. Asymmetric and symmetric stretching vibrations of the C–H bonds in CH3 and CH2 groups appear at 2931–2941 cm−1. Similar FTIR bands also appear at 2982, 2907–2914, and 2876–2897 cm−1 in flavonoid glycosides. In the spectra of these compounds the bending vibrations of the aromatic CC groups appear at 1583–1604 cm−1 and ≈1518 cm−1, some of them being superimposed by the stretching vibrations of the C–C group in the ring C of aglycones. The stretching of a C–C group also appears in silymarin/silibinins at 1509.9 (± 0.6) cm−1, while this value is significantly lower for flavonoid glycosides (1502–1504 cm−1). Other bending vibrations were observed for CH bonds in the range of 1393–1468 cm−1, while the stretching vibrations for CC and CO bonds and the bending vibrations for HOC, OCH, an HCC groups were superimposed in the range of 1011–1364 cm−1. The stretching vibration of the O–C groups in all flavonoids appears at 968–995 cm−1. Finally, out-of-plane bending vibrations of CH groups and twisting bending vibrations of COH and HCCC groups appear in the range of 742–921 cm−1[72-77]. All wavenumber values corresponding to the specific FTIR bands as well as the superimposed FTIR spectra of flavonoids with the other components of the ternary complexes are presented in Figure 2 and Figure 3 and in Supporting Information File 1 (Figures S6–S11 and Tables S6–S9).

The synthesized ternary complexes reveal the medium and strong FTIR bands of the above-mentioned host and guest components. However, FTIR bands that appear in specific regions where no interference occurs can also be relevant for the presence of individual compounds in the complex. This is the case for the weak band corresponding to the symmetric stretching vibrations of =CH groups from unsaturated glycerides in the hazelnut oil, which appear at 3006.5 (± 1), 3006.4 (± 0.6), 3006.3 (± 1.1), and 3006.6 (± 1.6) cm−1 for the X1H, X1N, X1R, and X1S ternary complexes at 1:1:1 molar ratios, respectively. These values are slightly higher by 1.1–3.1 cm−1 for all 3:1:1 ternary complexes (see Figure 2 and Figure 3 and Supporting Information File 1, Figures S6–S11 and Tables S6–S9). The strong bands corresponding to the asymmetric and symmetric stretching vibrations of the C–H bonds in the aliphatic CH3 and CH2 groups, as well as to the stretching vibrations of the ester C=O groups in triglycerides from hazelnut oil are clearly visible in all ternary complexes at 2922–2924, 2853–2854, and 1744–1745 cm−1, respectively. These values are very close to those corresponding to the starting hazelnut oil. Among other glyceride-related bands, those at 1453–1458 cm−1 originating from bending vibrations of the CH2 and CH3 groups, and 1236–1244 and 1152–1153 cm−1 from bending vibrations of the CH2 groups are also representative in the ternary complexes. They generally appear at lower values in the first case and at significantly higher values in the latter case in comparison with the starting hazelnut oil (see Supporting Information File 1, Figures S6–S11).

The most relevant flavonoid-related FTIR bands for the ternary complexes are those corresponding to the asymmetric stretching vibrations of the C=O groups. They occur in the range of 1637–1652 cm−1 for ternary complexes. The stretching vibrations of the C–C group in the ring C of the flavonoid glycosides or the bending vibrations of the aromatic CC groups occur in the range of 1598–1608 cm−1, but without specific variations in comparison with the starting compounds. The same is true for the band correlated to the in-plane bending vibrations of CH and OCH groups that appears at 1268–1299 cm−1. Also, the stretching vibrations of the C–C groups in the flavonoid glycosides or the stretching vibrations of the C–O groups in silymarin components (lower values) are observed in the same region. Another band that is present in all ternary complexes and is assigned to flavonoids is found at 807–821 cm−1, and corresponds to the out-of-plane bending vibrations of the C–H groups. This band appears at significantly lower values in rutin and rutin-related complexes.

β-CD was selected as the host for the formation of ternary complexes with the above-mentioned biologically active compounds and its content varies in complexes at 1:1:1 and 3:1:1 molar ratios. In the FTIR spectra of β-CD as a host, besides the wavenumbers corresponding to characteristic bands of β-CD, their intensities are relevant for the discrimination of the ternary complexes. However, many β-CD-related bands are weak or have at least medium intensities in the range of 1200–4000 cm−1. The most relevant bands for ternary complexes were the medium-strong intensity bands at 1152–1154 cm−1 (stretching vibrations of the C–O–C groups in the glucoside moieties), 1077–1080 cm−1 (stretching vibrations of the C–C groups), 1022–1026 cm−1 (stretching vibrations of the C–O groups), 944–947 cm−1 (stretching vibrations of the C–H groups from the β-CD ring), and two other medium intense bands at 574–576 and 522–529 cm−1, which were tentatively assigned as bending vibrations of the O–C–C groups and stretching vibrations of the C–C groups, respectively (see Figure 2 and Figure 3 and Supporting Information File 1, Figures S6–S11 and Tables S4, and S6–S9).

Discrimination of ternary complexes by Fourier-transform infrared spectroscopy coupled with principal component analysis (FTIR–PCA)

Taking into account the differences between the wavenumbers and intensities of specific stretching and bending vibrations of β-CD hydrate, raw hazelnut oil, and flavonoids in the pure form and as ternary complexes, a multivariate statistical analysis technique was applied for the discrimination of these samples and identification of the important FTIR variables for such classifications. PCA is a widely used multivariate statistical analysis technique that can extract valuable information from a large dataset. It is the case of FTIR data (both wavenumbers and intensities), where were assigned 20, 17, 34, and 33 FTIR bands for β-CD hydrate, hazelnut oil, flavonoids, and ternary complexes, respectively (see Supporting Information File 1, Tables S4–S9). On the other hand, not all FTIR bands corresponding to the starting compounds can be observed and assigned for the ternary complexes. PCA works with a complete variable matrix. As a consequence, only the FTIR bands that were identified in both the starting materials and the ternary complexes were considered for PCA analysis (see Table 1 and Supporting Information File 1, Tables S10–S12). This matrix is transformed in order to obtain the maximum variance of the data. The new axes are denominated Factors or Principal Components (PCs). The translation coordinates will provide the scores plots that reveal the similarities/dissimilarities between cases (samples), while the representation of the rotation coordinates of the axes (direction cosines) will give information about the influence of variables to the classification of cases. Only few PCs will extract the useful information from the dataset. As a consequence, the large number of variables will be reduced to only 2–4 PCs that will explain the variance of the data.

Discrimination of flavonoid glycosides and flavonolignans

Twenty-two variables were considered for the discrimination of flavonoids (flavonoid glycosides – hesperidin, “H”, naringin, “N”, rutin, “R”, and flavonolignans – silymarin, “S”). They correspond to wavenumbers and intensities of the FTIR bands identified for all flavonoids (Supporting Information File 1, Table S10). The flavonoid samples were clearly grouped, according to the PC2 vs PC1 or PC3 vs PC1 scores plot (Supporting Information File 1, Figures S12 and S13). Better results were obtained when only wavenumbers were used as PCA variables (Figure 4). All flavonoid glycosides are classified in the positive region of the PC1, in comparison with flavonolignans (silymarin components). According to FTIR–PCA analysis, hesperidin, naringin, and rutin are more similar and all of them are dissimilar to silymarin. This classification is especially due to the bands corresponding to stretching vibrations of the C=O groups and bending vibrations for the CH groups for the positive region of PC1, as well as to the stretching vibrations of the CO and CC bonds for the negative part (Table 1 and Supporting Information File 1, Figures S14–S18 and Table S10). In this latter case, only the first three PCs explain 97.41% of the variance of the FTIR data, with the highest value for PC1 (47.29%; see the eigenvalues greater than 1 in Figure S19, Supporting Information File 1).

Figure 4: PC2 versus PC1 scores plot from the FTIR–PCA analysis of the flavonoid glycoside and flavonolignan antioxidants (codes: “H” – hesperidin, “N” – naringin, “R” – rutin and “S” – silymarin); only wavenumbers of the FTIR bands were used as input variables.

Table 1: Factor coordinates (principal components, PCs) of the variables, based on correlations from the FTIR–PCA analysis of the flavonoid glycoside and flavonolignan antioxidants; only wavenumbers (“v” – for stretching vibrations, “d” – for bending vibrations) of the FTIR bands were used as input variables.

  PC1 PC2 PC3 v(OH) 0.763 −0.616 −0.182 vas(CH) −0.090 0.565 −0.780 vs(CH) 0.233 −0.781 −0.563 d(OH)/vas(C=O/C=C) 0.930 0.323 0.165 d(arC#C) 0.595 0.714 0.353 d1(CH2/3) −0.350 0.026 −0.931 v1(CO)/d1(CO) −0.416 0.797 −0.435 d1(CH) 0.986 −0.142 −0.061 v(CO)/v(CC) 0.937 0.128 −0.321 v(CO)/v(CC/CO) −0.940 0.077 0.302 d4(CH) −0.557 −0.739 0.049 Discrimination of ternary complexes and β-CD hydrate samples

In the same way, ternary complexes and native β-CD hydrate samples were classified according to specific FTIR wavenumbers and intensities of the bands identified in all samples. β-CD hydrate samples were classified in the top-right region of the PC2 vs PC1 scores plot (codes “Y”), in comparison with the ternary complexes in the center-left and bottom of the plot. Moreover, such grouping can also be observed for some ternary complexes types (e.g., “X1H” in the left and “X3R” in the top-left of the plot, Figure 5). Few FTIR variables are responsible for the discrimination of ternary complexes and β-CD samples, especially those related to band intensities corresponding to bending vibrations of CH2 groups and stretching vibrations of various bonds including those from CCO, CCC, CO and COC systems (PCA results are not presented).

Figure 5: PC2 versus PC1 scores plot from the FTIR–PCA analysis of the β-CD/hazelnut oil/flavonoid ternary complexes (codes: “X1H/N/R/S” and “X3H/N/R/S” for the 1:1:1 and 3:1:1 ternary complexes with hesperidin/naringin/rutin/silymarin, respectively) and β-CD hydrate (code: “Y”); all wavenumbers and intensities of the FTIR bands were used as input variables.

Discrimination of ternary complexes and flavonoids

More interesting were the results obtained for the FTIR–PCA analysis of ternary complexes and flavonoids. A total of 18 FTIR variables (both wavenumbers and intensities, Supporting Information File 1, Tables S11 and S12) were identified in all ternary complexes and flavonoids. They were used as input variables for the discrimination of complexes and guest compounds. Also, the wavenumbers and intensities sets were used separately for the discrimination. Flavonoids were clearly classified in the left side of the PC2 vs PC1 scores plot (Figure 6). Wavenumbers of the bands corresponding to the stretching vibrations of the CO and CC bonds for the positive side, as well as the intensity of the band corresponding to the asymmetric stretching vibration of the CH bond for the negative side of the PC1 were the most important for this classification (see also Supporting Information File 1, Figure S20 for the PC3 vs PC1 scores plot, Figures S21and S22 for the corresponding loadings plots, and Table S11 for the influence of variables on the classification). Better results were obtained if only wavenumbers were used as input variables for the FTIR–PCA analysis of ternary complexes and the starting flavonoids. All flavonoids were grouped in the right side of the PC2 vs PC1 scores plot, with higher similarity for hesperidin, naringin, and rutin (Figure 7). On the other hand, all ternary complexes were located in the left side of this plot, also sub-classified according to the presence of specific flavonoids. In a similar manner, ternary complexes based on silymarin are dissimilar with the other complexes, which have a high level of similarity. These observations are also sustained by the other scores plots, all with very good classifications of the samples (Figure 8 and Figure 9). Responsible for these classifications are the variables corresponding to the FTIR bands related to symmetric and asymmetric stretching vibrations of the CH bonds (positive PC1), stretching vibrations of the CC and CO bonds (negative PC1), stretching and bending of C=O and OH/CH, respectively (negative PC2) (Figure 10 and Figure 11, Supporting Information File 1, Table S12). Only the first three PCs were used for obtaining these valuable discrimination results. They explain almost all variances of the FTIR data, as is presented in Figure 12 (85.69%).

Figure 6: PC2 versus PC1 scores plot from the FTIR–PCA analysis of the β-CD/hazelnut oil/flavonoid ternary complexes (codes: “X1H/N/R/S” and “X3H/N/R/S” for the 1:1:1 and 3:1:1 ternary complexes with hesperidin/naringin/rutin/silymarin, respectively) and flavonoids (codes: “H” – hesperidin, “N” – naringin, “R” – rutin and “S” – silymarin); all wavenumbers and intensities of the FTIR bands were used as input variables.

Figure 7: PC2 versus PC1 scores plot from the FTIR–PCA analysis of the β-CD/hazelnut oil/flavonoid ternary complexes (codes: “X1H/N/R/S” and “X3H/N/R/S” for the 1:1:1 and 3:1:1 ternary complexes with hesperidin/naringin/rutin/silymarin, respectively) and flavonoids (codes: “H” – hesperidin, “N” – naringin, “R” – rutin and “S” – silymarin); only wavenumbers of the FTIR bands were used as input variables.

Figure 8: PC3 versus PC1 scores plot from the FTIR-PCA analysis of the β-CD/hazelnut oil/flavonoid ternary complexes (codes: “X1H/N/R/S” and “X3H/N/R/S” for the 1:1:1 and 3:1:1 ternary complexes with hesperidin/naringin/rutin/silymarin, respectively) and flavonoids (codes: “H” – hesperidin, “N” – naringin, “R” – rutin and “S” – silymarin); only wavenumbers of the FTIR bands were used as input variables.

Figure 9: PC3 versus PC2 scores plot from the FTIR–PCA analysis of the β-CD/hazelnut oil/flavonoid ternary complexes (codes: “X1H/N/R/S” and “X3H/N/R/S” for the 1:1:1 and 3:1:1 ternary complexes with hesperidin/naringin/rutin/silymarin, respectively) and flavonoids (codes: “H” – hesperidin, “N” – naringin, “R” – rutin and “S” – silymarin); only wavenumbers of the FTIR bands were used as input variables.

Figure 10: PC2 versus PC1 loadings plot from the FTIR–PCA analysis of the β-CD/hazelnut oil/flavonoid ternary complexes and flavonoids; only wavenumbers of the FTIR bands were used as input variables (see Table S12 in Supporting Information File 1 for codes).

Figure 11: PC3 versus PC1 loadings plot from the FTIR–PCA analysis of the β-CD/hazelnut oil/flavonoid ternary complexes and flavonoids; only wavenumbers of the FTIR bands were used as input variables (see Table S12 in Supporting Information File 1 for codes).

Figure 12: Eigenvalues of the correlation matrix from the FTIR–PCA analysis of the β-CD/hazelnut oil/flavonoid ternary complexes and flavonoids; only wavenumbers of the FTIR bands were used as input variables (see Table S12 in Supporting Information File 1 for codes); the first three PCs can be retained, which explain 85.69% from the variance of the data.