**3. Results**

## *3.1. Photoluminescence*

The photoluminescence (PL) spectra acquired from pure and adulterated EVOO with different amounts of olive-pomace oil adulterant is depicted in Figure 1a. PL spectra for all of the different adulterated EVOO samples that were adulterated with different oils are depicted in Figure S3a–e, Supplementary Materials. Pure adulterant oils (HO sunflower, sunflower, corn, and soy–nut) do not present any PL activity under this 532 nm excitation. In the case of pure olive-pomace oil, which is a common adulterant oil, the PL is very weak with a clear blue shift in its PL peak relative to that of pure EVOO. Despite the fact that both EVOO and olive-pomace oil are derived from olives, their PL spectra present large differences from one another due to the low concentration of compounds with luminescent properties, such as pigments (e.g., chlorophyll, carotenes, and derivatives), phenols, and tocopherols in the adulterant oils [37]. For EVOO, the strong luminescence around 670 nm and 720 nm is mainly associated to the photosystem of chlorophyll [38]. The first peak is attributed to photosystem I (PSI) and the second peak is due to the combination of photosystems I and II [38]. The strong photoluminescence can be seen even by naked eye (Figure S1b,c, Supplementary Materials).

**Figure 1.** (**a**) Photoluminescence spectra of EVOO adulterated with different concentrations of olivepomace oil. (**b**) Integrated photoluminescence spectra of the different oil mixtures as a function of the adulterant-oil concentration. The light-grey shaded region represents the 95% range of confidence region around the best-fit line. (Inset) peak position of the PL spectra as a function of adulterantoil concentration.

The numerically integrated PL intensity of all of the spectra as a function of the adulterant-oil concentration of different adulterant oils is shown in Figure 1b. The lightgrey shaded region represents the 95% range of confidence region around the best-fit line. We note that the best-fit line passes through pure EVOO, but a clear linear decrease and variation in the integrated intensity is observed due to the negligible luminescent activity of the adulterant oils. Additionally, we observe a small blue shift of the chlorophyll/pheophytin peak as the concentration of adulterant oil increases (see inset Figure 1b). The origin of this

blue shift is even more pronounced when comparing the PL spectra of pure EVOO with pure adulterant oil, since the PL peak in pure adulterant oil sits at lower wavelengths (see Figure S3, Supplementary Materials).

#### *3.2. Raman Spectroscopy*

The Raman spectra of pure EVOO, olive-pomace, HO sunflower, corn, and soy–nu<sup>t</sup> blend oils are depicted in Figure 2. Four common bands can be observed in all of the oils, located at ~1265, 1305, 1440 and 1656 cm<sup>−</sup>1, which correspond to the common Raman modes of unsaturated fatty acids such as: oleic (OA, C18:1), linoleic (LA, C18:2), and linolenic (ALA, C18:3) acid [39]. These molecules are all 18-carbon carboxylic acids with one, two, and three *cis*-double bonds, respectively. Each of the oils under study has a comparable fatty-acid composition (see Table S2 in the supplementary information), which leads to these common carboxylic acid peaks in their Raman spectra. These characteristic Raman bands have already been previously studied by El-Abassy et al. and Lv et al. [39,40]. The attribution of each of the observed peaks to their associated vibrational mode for all spectra in Figure 2 is summarized in Table 1. The remaining two Raman bands located at ~1155 and 1523 cm<sup>−</sup>1, which are unique to EVOO, have previously been associated with C–C and C=C stretching vibrations of the main polyene chain of carotenoids [41,42]. These additional two bands are not detected in any of the adulterant oils, including the olive-pomace oil. As was the case with the photoluminescence, Raman spectroscopy clearly distinguishes a spectroscopic difference between EVOO and all of the other edible oils, including olive-pomace oil, which shares a common derivation from olives. The absence of carotenoids in refined oils results from the degradation that they suffer during food processing, storage, and thermal treatment. Thermal treatment during the refinement process leads to the isomerization of the carotenoids and a consequent change in their molecular structure [43].

**Figure 2.** Raman spectrum of pure EVOO, olive-pomace, sunflower, corn, and soy–nu<sup>t</sup> oils.


**Table 1.** Assignment of the Raman bands of the edible oils.

Figure 3a shows the normalized Raman spectrum of pure EVOO adulterated with different amounts of HO sunflower. As the HO sunflower concentration increases, the intensity at 1523 cm<sup>−</sup><sup>1</sup> (carotenoid peak) decreases. Similar results were also found for the rest of the adulterant oils (see Figure S4, Supplementary Materials). A summary of these results are shown in Figure 3b, which presents the ratio of the numerical integration of the areas of the Raman peaks at 1523 cm<sup>−</sup><sup>1</sup> and 1656 cm<sup>−</sup>1. A clear decrease in I1523/I1656 ratio can be observed as the adulterant oil content increases. This effect comes from the zero Raman activity for carotenoids peaks shown by all adulterant oils studied here. Notably, the Raman spectra can be directly measured from as-packaged oil without opening and manipulating the oil, allowing for non-invasive verification even from an unopened oil bottle (see Figure S5, Supplementary Materials). Similar to the PL, the addition of adulterant oils leads to a decrease in the I1523/I1656 integral ratio. Qui et al. recently observed a similar result using the I1523/I1656 ratio to determine the free-fatty-acid (FFA) content of olive oils and found that this intensity ratio decreases linearly with FFA content, although the FFA content was obtained from the nutrition label of each of the oils [21]. Thus, the I1523/I1656 integral ratio is an additional useful figure of merit to quantify EVOO purity.

**Figure 3.** (**a**) Raman spectra of EVOO adulterated with different concentrations of high oleic sunflower oil. (**b**) Intensity ratio of the carotenoid peak (1523 cm<sup>−</sup>1, I1523) to the C=C stretching peak (1656 cm<sup>−</sup>1, I1656) as a function of the adulterant oil concentration.

#### *3.3. Fourier-Transform Infrared Spectroscopy*

The IR spectra of pure EVOO, corn, soy, and olive-pomace oils are depicted in Figure 4. For ease of visualization, the spectra were separated in two wavenumber ranges: (3150–2800) cm<sup>−</sup><sup>1</sup> and (1500–1000) cm<sup>−</sup>1. The first window shows the characteristic IR peaks resulting from hydrogen stretching functional groups, while the second window shows other bond deformations and bending that are primarily associated with vibrations

of CH*i* groups (with *i* = 1, 2, 3) and C–O bonds [44]. Unlike the PL and Raman results, the FTIR spectra shows remarkable similarities between the spectra of the studied samples, making them difficult to differentiate. Therefore, a deep analysis using principal component analysis (PCA), a technique that allows for patterns and variations within a dataset to be readily visualized, was performed to allow for facile differentiation of each of the spectrum from one another. PCA analysis is relatively common in food chemistry, as optical spectra tend to be very similar within particular foods and their associated derivatives. The results of this analysis are displayed in the inset of Figure 4a,b. Our results showed that EVOO and olive-pomace oils could not be differentiated from one another in FTIR spectroscopy, as the PCA scores were almost identical. However, the PCA scores of corn, soy–nu<sup>t</sup> oil, and sunflower oils showed clear differences when compared with EVOO.

**Figure 4.** FTIR spectra of four oils: EVOO (pale green), olive-pomace (blue), corn (orange) and soy (dark green) over two different wavenumber ranges focusing on the (**a**) hydrogen stretching functional groups and (**b**) CH*i* (i = 1, 2, 3) functional groups present in each oil.

At the most superficial level, a quick differentiation of the IR spectra of the oils was established via PCA, though a deeper analysis of the PCA scores of the adulterated EVOO is possible. Figure 5 shows the subsequent PCA analysis of IR spectra of adulterated EVOO. This rapid and simple PCA analysis highlights the impurities added to EVOO by showing a shift in the scores of adulterated samples as the adulterant oil increases. The shift is observed even with less than 5% of added adulterant (Figure 5a,b,d,e). Similar results have been observed by Vanstone et al. [45], who demonstrated the potential of a combination of near-infrared spectroscopy with PCA to detect EVOO adulteration at levels as low as 2.7%, given an unadulterated reference sample (i.e., pure EVOO). However, we demonstrate similar conclusions with FTIR in the mid-IR spectral range, which is advantageous as molecular fundamental vibrational modes lie in the mid-IR, while spectral measurements in the near-IR are measurements of molecular vibrational overtones. While the PCA alone exhibits potential in its ability to discriminate similar spectra, the addition of a multivariable regression model will be necessary to obtain true quantification of EVOO adulteration.

**Figure 5.** PCA score plots of oil mixtures at the 3000 cm<sup>−</sup><sup>1</sup> and 1300 cm<sup>−</sup><sup>1</sup> window: (**<sup>a</sup>**,**d**) corn–EVOO, (**b**,**<sup>e</sup>**) soy–nut–EVOO, and (**<sup>c</sup>**,**f**) sunflower–EVOO. The inset in (**f**) shows a zoom around the PCA scores of the sunflower-based sample. The color gradient in each figure indicates the evolution of the PCs from pure adulterant (darker colors) to smallest amount of adulterant (lighter colors).

As the PCA of FTIR spectra did not show significant differences between EVOO and olive-pomace, we applied two-dimensional correlation spectroscopy (2DCOS) to gain greater insight into the FTIR spectra. The 2DCOS technique is a mathematical method for analyzing changes in a signal produced by an external perturbation (e.g., a change in temperature, pressure, pH, concentration of mixtures, etc.). To calculate the 2DCOS map we used the concentration of olive-pomace oil as an external perturbation and the spectra dataset was ordered from pure EVOO (0% of oil adulterant) to pure pomace (100% of oil adulterant), i.e., 0, 5, 10, 20, 50, 100%. The raw spectra were baselined and normalized using the most-intense band for each frequency window in Figure 4. The average spectrum was used as a reference spectrum following the same procedure as reference [46]. The 2DCOS analysis was performed with the Mat2dcorr Matlab toolbox [47]. Figure 6 shows the synchronous 2DCOS map at the 3000 cm<sup>−</sup><sup>1</sup> (Figure 6a) and 1300 cm<sup>−</sup><sup>1</sup> (Figure 6b) FTIR frequency windows. The respective autocorrelation and FTIR-averaged spectra are shown above each frequency window. Autocorrelation spectra are defined by a diagonal line along the 2DCOS map and their bands are known as autopeaks. The autopeaks represent real changes between the FTIR spectra that are produced by the external perturbation (such as the addition of olive-pomace oil, in this case). The autocorrelation spectra show three autopeaks located at 2844, 2900, and 2974 cm<sup>−</sup><sup>1</sup> in the 3000 cm<sup>−</sup><sup>1</sup> window and only one autopeak around 1330 cm<sup>−</sup><sup>1</sup> for the 1300 cm<sup>−</sup><sup>1</sup> window. The comparison of the auto correlation and FTIR spectra show that the larger changes among the spectra occur at wavelengths where the FTIR spectra is very weak, which indicates why the PCA analysis was not able to find significant differences between the olive-pomace oil and EVOO. Furthermore, a 2DCOS analysis of pure EVOO oils was also performed using the same data treatment (Figure S6, Supplementary Materials) to verify that the observed variations are not artificial variations resulting from the data treatment such as background subtraction and/or normalization. In this analysis, the same large variation in autopeaks are

in fact observed around 2844 and 2900 cm<sup>−</sup>1, indicating that such variations are dependent not only on real significant differences in oil concentration, but also on experimental fluctuations. Notably, autopeaks around 2974 cm<sup>−</sup><sup>1</sup> are only present in the EVOO/olivepomace 2DCOS map. Consequently, while this type of data processing enables even such small fluctuations to be used as identifiers for authentication between oils of similar origin, additional processing and identification of 2DCOS peaks may first be required.

**Figure 6.** Contour map of the synchronous 2D FTIR correlation spectra of the EVOO–pomace mixtures at the (**a**) 3000 cm<sup>−</sup><sup>1</sup> and (**b**) 1300 cm<sup>−</sup><sup>1</sup> frequency windows. The spectra above the 2D plots provide the auto correlation spectrum (black solid lines) of each 2DCOS map. The average of the FTIR spectra in each window is also included for comparison (grey dashed lines).

#### *3.4. Thermal Conductivities*

The temperature dependence of the thermal conductivity (*k*) of pure oils and of three adulterated mixtures are shown in Figure 7a,b, respectively. A monotonic decrease in the thermal conductivity as the temperature increases can be observed in all of the studied samples. A similar temperature dependence was also reported by Turgut et al. [31]. Interestingly, the temperature dependence of the k of the EVOO–soy–nut oil mixture follows the same temperature dependence as the pure EVOO, even at 50% of soy–nu<sup>t</sup> oil concentration. A comparative analysis of the k values at room temperature (Figure 7c) and at 400 K (Figure 7d) demonstrate noticeable differences between each pure oil and between EVOO and its adulterated mixtures. This highlights the ability of the k—which, to date, has tended to be overlooked as a useful figure of merit in food authentication—to provide information that enables the distinction of pure and adulterated oils, or, more generally, other food products as well.

**Figure 7.** (**a**) Thermal conductivity of pure oils as a function of temperature. (**b**) Thermal conductivity of adulterated oils as a function of temperature. (**c**) Room-temperature thermal conductivity as a function of adulterant oil concentration. (**d**) Comparison of the thermal conductivity at 400 K for different pure oils or adulterated EVOO.
