A total of 15 oil samples were purchased from local supermarkets, samples 1 to 6 were labeled as EVOO (3 samples imported from Spain, 3 samples imported from Italy). Samples 7 to 10 were labeled as ROO. Samples 11 to 15 were labeled as different seed oils (SO), such as peanut oil, soybean oil, rapeseed oil, corn oil and blended oil. All of the samples were kept in dark glass bottles and stored at room temperature.

For the investigation of olive oil adulteration, fresh EVOO samples were mixed with ROO and soybean oil samples. Two sets of mixtures of 1, 3, 5, 10, 15, 20, 25, 30, 35, 40, 50, 80% (w/w) for ROO or soybean oil adulterant in EVOO were prepared. These mixtures were analyzed immediately after preparation.

NMR. All solvents were of reagent or analytical grade: Hexafluorobenzene (99%) was purchased from Alfa Aesar (Tianjin, China). The deriving reagent (4-fluorobenzoyl chloride, purity: 98%) and 4-tert-butylphenol were purchased from Sigma-Aldrich (Shanghai, China). Pyridine and Chloroform-d were purchased from Macklin (Shanghai, China).

Titration. Acetic anhydride, potassium hydrogen phthalate, n-butyl alcohol, ethanol, potassium hydroxide, phenolphthalein, sodium hydrogen sulfate, diethyl ether were bought from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China).

GC-MS. Analytical grade trimethylpentane, potassium hydroxide, methyl alcohol, and sodium hydrogen sulfate monohydrate were purchased from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China).

A stock solution (100 mL) composed of CDCl₃ and pyridine in 1.5:1.0 (v/v), with 0.1 mL hexafluorobenzene and 150 mg 4-tert-butylphenol was prepared. 4-Tert-butylphenol was used as an internal standard for quantification purposes and hexafluorobenzene was used as an internal standard for ¹⁹F NMR chemical shift at δ-164.90 ppm. 400 mg olive oil sample were mixed with the stock solution (2 mL) in a 4 mL centrifuge tube. The required volume of the mixed solution (0.5 mL) and the reagent (30 μL) were added into a NMR tube with 5 mm diameter. The reaction mixture was left in the NMR tube to react for 30 minutes at room temperature. Upon completion of the reaction, the ¹⁹F NMR spectra of the solution were determined immediately.

120 mg of olive oil sample were dissolved in 0.5 mL chloroform-d, which contained 0.03 % trimethylsilane (TMS). The resulting solution was placed in a 5 mm NMR tube and then the ¹H NMR spectra were recorded.

All NMR spectra were obtained on a Bruker AMX500 spectrometer, operating at 470 and 500.1 MHz for fluorine-19 and proton nuclei, respectively, at 26 ±1 ^oC. Typical ¹⁹F spectral parameters for this study were as follows: 90^o pulse width, 19.3μs; sweep width, 100 kHz; relaxation delay, 1s; memory size, 64K. 64Ttransients were accumulated for each spectrum. For all FIDs, line broadening of 0.3Hz was applied and drift correction was performed prior to Fourier transformation. A polynomial fifth-order base-line correction was performed before integration. All ¹⁹F chemical shifts were reported relative to hexafluorobenzene, which gave a sharp signal in pyridine/CDCl₃ at δ-164.90. High-resolution ¹H NMR spectra were acquired with the following acquisition parameters: time domain, 32K; 90^o pulse width, 9.3 μs; spectral width, 20.7ppm; relaxation delay, 1 s; acquisition time, 3.2 s. 32 scans were accumulated. A long delay time ensures the return of the excited nuclei to their thermal equilibrium prior to the next pulse, which is an imperative condition for a quantitative measurement. Base-line correction was performed carefully by applying a polynomial fourth-order function in order to achieve a quantitative evaluation of all signals of interest. The ¹H NMR chemical shift signals in CDCl₃ were referenced to TMS at δ 0.00.

These parameters were determined by employing the official methods of titration (AOCS method Cd 13-60, ISO 660:1996).

A solution of about 60 mg of oil dissolved in 4 mL trimethylpentane was made up in a test tube with glass stopper. After the addition of 200 µL of 0.2 M solution of potassium hydroxide in methanol, the tube was shaken for 30 s, and left until the phases separated. One gram of sodium hydrogen sulfate monohydrate was added into the solution in order to neutralize potassium hydroxide. It was then stored in a refrigerator until subjected to GC-MS (ISO 5509:2000, ISO 5508:1990).

All GC-MS analyses were performed on Shimadzu GC2010A (Kyoto, Japan) gas chromatography instrument coupled with a GCMS-QP2010 quadrupole mass spectrometer (Shimadzu). In the gas chromatographic system, an Rtx^®-Wax capillary column, 30 m length, 0.25 mm i.d., and 0.25 μm film, consisting of cross bond polyethylene glycol (Restek) was used. The column temperature was programmed from 140 to 250 ^oC at the rate of 4 ^oC/min, held for 1 min at 140 ^oC and then held for 6 min at 250 ^oC. The injection temperature was kept at 220 ^oC, the carrier gas was nitrogen, and the column flow (nitrogen flow rate) was 1.36 mL/min. A sample of 1 μL was injected with a split ratio of 30:1.

Mass Spectroscopy Conditions: The ion source temperature was 230 ^oC, and the interface temperature was 280 ^oC. Ionization voltage was 0.2 kv.

The relative content was calculated by using the peak area normalization method.

The content in DGs of the various EVOO, ROO and SO were detected by ¹⁹F NMR. This method is based on the derivatization of the labile hydrogens of the hydroxyl groups of the DGs with 4-fluorobenzoyl chloride according to the reaction shown in Figure 1 and the integration of the appropriate peaks in the ¹⁹F NMR spectrum. However, the acidity cannot be detected by ¹⁹F NMR properly, because the fatty acids were not shown to react with the deriving regent quantitatively (Zhou et al., 2015). Therefore, the conventional method was applied to the acidity determination.

Table 1 contains the percentage contents of the 1,2-DGs, 1,3-DGs, TDGs, the ratio D (1,2-DGs/TDGs), acidity and D_T/A [(TDGs/620)/(acidity/282)]. These parameters appear to differentiate the oil samples. The origin of EVOO is divided into two countries, Spain and Italy. The EVOO samples from Italy show the higher content of TDGs, with acidity and D_T/A, present in a lower D ratio. EVOO is characterized by low values of TDGs and high values of ratio D compared to ROO and SO. The TDGs in EVOO range from 1.6 to 2.2%. The level of TDGs is higher in ROO, which range from 5.0 to 8.0 %. The ratio D in all EVOO samples is equal to or greater than 0.4. Moreover, the ratio D of EVOO is much higher than ROO and SO. It has been suggested that the ratio D is a useful index of the quality of olive oils because of the fact that the isomerization of 1,2-DGs to 1,3-DGs usually occurs during olive oil storage and the refinement. Previous research on the ratio D of virgin olive oils freshly extracted from olives of normal ripeness should be close to 1 according to Vigli et al. (2003). However, due to the fact that all the samples used in our work were stored for half a year, then the ratio D is closer to 0.5, although it is still much higher than ROO and SO. As for the acidity, EVOO ranges from 0.25 to 0.30, and a good linear concentration (r = 0.81) was obtained between acidity and TDGs. According to Fronimaki et al. (2002), a much better correlation (r = 0.89) is observed between the acidity and the amount of 1,3-DG. Therefore, there are some differences. The acidity of ROO and SO was much lower than EVOO because ROO and SO had been refined.

Theoretically, every triglyceride molecule is hydrolyzed to form one DG molecule and one free fatty acid (FFA) molecule. However, it was shown that the ratio (D_T/A) of molecular number of TDGs to that of FFAs was in the range of 2.81~3.31 in EVOO in Table 2. This means that the mole content of DGs is much higher than that of FFA. The question is: where are the FFAs going? The answer is that more FFAs are removed than DGs from EVOO during the water washing process because FFAs are more polar than DGs. Low quality olive oils are mainly pomace olive oil and lampant olive oil. Both of them have high acid values and TDGs and are not edible in their crude form. They become editable and are called ROO after refining, together or alone. The FFAs are removed almost completely, but little DGs are lost. So ROO samples have very high TDGs and low acidity, hence their D_T/As are large (21.47~48.71) and much larger than those of EVOOs (Table 2). Definitely, D_T/A could be used as a very important parameter to detect whether EVOO was adulterated with ROO.

ROO, as it is much cheaper than EVOO, is usually added into EVOO by some unscrupulous traders because both of them have the same components, especially the same fatty acid composition if they are from same olive trees. The addition of ROO to EVOO is expected to deteriorate the antioxidant properties and organoleptic characteristics of EVOO. Also, most of the nutritional minor components in ROO are lost due to de-acidification, decoloration and deodorization under high temperatures, and high vacuum with absorbents. Several studies have suggested that the radio D (1,2-DG / TDGs) can distinguish different grades of olive oil (Sacchi et al., 1997; Zhou et al., 2015). Therefore, using ¹⁹F NMR as the method to detect the adulteration of EVOO with ROO is acceptable.

In figure 2, it can be seen how, with the increase in adulteration level, the content of 1,2-DG, 1,3-DG TDGs increased. However, the radio D decreased. A good correlation (r = 0.96) was observed between the radio D and the adulteration level.

The hydroxyl value (OHV) is defined as the number of milligrams KOH equivalent to the hydroxyl groups found in one gram of the sample and is expressed in mg of KOH/g (Srk et al., 2013). In olive oil, DGs contain the most hydroxyl groups. In Figure 2, good linear correlations are obtained (r = 0.99) both in (a) and (b). The slope of TDGs to adulteration level is 6.7, which is close to the slope of the hydroxyl value (6.5). The result demonstrates that ¹⁹F NMR is useful in detecting TDGs.

OHV cannot discriminate the OH group in hydroxyl fatty acids (ricinoleic acid) or the OH group in glycerol bases of MGs and DGs, but our _¹⁹ F NMR method can. It can even discriminate 1,2-diglycerides and 1,3-diglycerides clearly (Zhou et al., 2015).

The discrimination of EVOO with respect to the other oils can be seen in Figure. 3, where the ratio D is plotted against the TDGs. It is seen clearly that the EVOO samples are clustered in the upper part of the graph, whereas ROO and SO are dispersed in the lower part of the graph. What is most interesting in this graph is the observation that adulterated EVOO samples with ROO lie between the group of EVOO and ROO. Along with the increase in the adulteration of low price oil, the plots move to the lower right.

In summary, TDGs and D_T/A could be used as two key parameters to determine whether EVOO is adulterated with ROO or not. If the TDGs of EVOO is less than 2.5 and sometimes its D_T/A is less than 4, it could be concluded that EVOO is not adulterated with ROO. D is only a parameter to indicate the freshness of EVOO. The larger D is, the fresher the EVOO is.

The compositions of the unsaturated fatty acids (oleic acid, linoleic acid, linolenic acid) and saturated fatty acids (SFAs) in different oils were calculated according to the various signal intensities in the ¹H NMR spectra (Vigli et al., 2003; Sacchi et al.,1996; Sacchi et al., 1997).

There is one intense peak at δ = 1.68 in the ¹H NMR spectrum of virgin olive oils. This peak was identified as the methyl protons of the CH₃-17 and CH₃-29 of squalene when compared with the ¹H NMR spectrum of squalene standard substance, which has been reported previously (Mannina et al., 2009). Because the two methyl groups in squalene are equivalent, the weight percentage of squalene can be calculated by a similar method to the determination of fatty acid composition according to Formula 1, but their different molar weights are also considered. In Formula 1, A_S is the area of integration of the signal of the two methyl groups of squalene at 1.68 ppm. The chemical shift of all the methyl protons of fatty acids is at 0.88 ppm (signal J) except linolenic (signal I) whose chemical shift is 0.97 ppm. A_I and A_J are the areas of integration of signals I and J, respectively. 410 is the molar weight of squalene and 296 is the molar weight of methyl oleate which is about 1/3 the molar weight of a triglyceride in olive oil.

The percentages of the unsaturated fatty acids (oleic acid, linoleic acid, linolenic acid) saturated fatty acids (SFAs) and squalene in different oils obtained by ¹H NMR spectroscopy are listed in Table 2. Careful analysis of these data reveals some interesting trends as follows:

The percentages of linolenic acid in all olive oil samples ranged from 0.8 to 1.6 %. The linoleic content exit in EVOO samples ranged from 3.0 to 5.5 %, which is lower than ROO (about 7 %) and SO (16~51 %). The oleic acid percentages in all olive oils are relatively stable and range from 75 to 82%, which is much higher than SO. The percentages of SFA (total saturated fatty acids) in all the oil samples were relatively stable. As for squalene, which exits in olive oils in higher amounts, it can be a very important index to distinguish between olive oils and seed oils.

The content of fatty acid of the oil samples was also detected by the GC-MS, which is a traditional method to detect fatty acids (Capote et al., 2007). 6 different kinds of fatty acids (palmitic acid, palmitoleic acid, stearic acid, linolenic acid, linoleic acid, oleic acid) were identified by GC-MS.

In Table 2, it is clearly shown that the sum of the palmitic acid content and the stearic acid content detected by GC-MS was very close to the content of SFAs detected by ¹H NMR. However, GC-MS can differentiate palmitic acid from stearic acid easily, while ¹H NMR cannot. The contents of oleic, linoleic and linolenic acids can be determined by ¹H NMR as well as GC-MS and both methods get close results. Comparing the three parameters (oleic acid, linoleic acid, linolenic acid) detected by ¹H NMR and GC-MS showed that the D-value of linolenic acid ranged from 0.2 to 1. As for the linoleic acid, it was 1.2 to 4.6. However, the D-value of oleic acid ranged from 3 to 7. This indicates that the detection of linolenic acid by these two methods is relatively accurate. It is also clearly shown that EVOO and ROO have very similar fatty compositions but ROO contains lower squalene contents than EVOO. Compared to other vegetable oils, the content of squalene is richer in olive oil. So squalene content may be a very important parameter for the determination of EVOO if adulterated with other seed oils.

Soybean oil is a typical additive for the adulteration of olive oils due to its low price. Five parameters (linolenic acid, linoleic acid, oleic acid, SFAs, squalene) of the adulteration of EVOO with soybean oils were determined by ¹H NMR. In Figure 4, except for SFAs, the other four parameters, oleic acid, linolenic acid, linoleic acid, squalene are all in good relationship with the adulteration level. With the increase in the adulteration level, the contents of linoleic acid and linolenic acid increased; although the opposite occurred with the contents of oleic acid and squalene. According to Table 2, the content of linolenic acid in EVOO cannot exceed 1.3%. On the basis of the linear equation y = 0.0745x + 0.6636, the limit of detection of the adulteration level was 4.5% by ¹H NMR method.

Squalene content and fatty acid composition can indicate whether EVOO is adulterated with SO or not. Squalene content in EVOO is larger 0.6% (w/w) and oleic acid content is larger than 65%.