**3. Application of fluorescence in olive oil analysis**

Application of fluorescence to quality assessment of olive oils was proposed already in the beginning of the 20th century. From 1925, when mercury lamp with the Wood's filter became commercially available, visual observation of oil fluorescence induced by UV light was utilized to detect adulteration of extra virgin olive oils. It was shown that extra virgin olive oils exhibit characteristic yellow fluorescence, due to chlorophylls, while fluorescence of refined oils was blue due to the changes in chlorophyll content during the refining process. This method allowed detecting adulteration of extra virgin olive oils at the level of 5% with refined oils (Sidney & Willoughby, 1929; Glantz, 1930). The use of Wood's lamp was accepted as the U.S. official method for detection of olive oil adulteration. (Kyriakidis & Skarkalis, 2000). The authors of the first papers that reported spectral properties of fluorescence of vegetable oils also point out practical applications of fluorescence spectra as fingerprints in oil analysis (Wolfbeis & Leiner, 1984; Kyriakidis & Skarkalis, 2000).

#### **3.1 Discrimination between quality grades of olive oil**

Olive oil is an economically important product and its quality control and detection of possible fraud are of great interest. Olive oils are classified and priced according to acidity. The most expensive is the high-quality extra-virgin olive oil. This oil may be subject of both mislabeling and adulteration. Refined olive oil is obtained from virgin olive oil using

respective excitation and emission loadings, which correspond to the respective excitation and emission spectra, facilitating identification of the fluorescent constituents. This approach is called mathematical chromatography, enabling qualitative and quantitative

The analytical problem of the food quality assessment often involves assignment of a particular product to a specific category. To perform such classification, supervised pattern recognition methods are used. In these methods the information about the class membership of the samples in a certain category is used to derive classification rules, which are next applied to classify new samples into correct categories on the basis of patterns present in their measurements (Berrueta et al., 2007). A number of classification techniques were used for analysis of food fluorescence data in the supervised mode: linear discriminate analysis, factorial discriminate analysis, k - nearest neighbors, discriminate partial least squares regression (DPLS), soft independent modeling of class analogy, and artificial neural

Multivariate calibration is the most important and successful combination of chemometrics with spectral data used in analytical chemistry. The calibration consists of building a relationship between a desired chemical, biological or physical property of a sample, and its spectrum. The advantage of such approach is the replacement of the wet chemical measurements of a concentration, which are usually slow and expensive, by the spectral measurements, which are nondestructive and fast, requiring little or no sample preparation and producing no waste chemicals. The multivariate regression methods most frequently used in fluorescence analysis are partial least-squares regression (PLS) and principal component regression (PCR). N-way partial least-squares regression (N-PLS) is used for

Application of fluorescence to quality assessment of olive oils was proposed already in the beginning of the 20th century. From 1925, when mercury lamp with the Wood's filter became commercially available, visual observation of oil fluorescence induced by UV light was utilized to detect adulteration of extra virgin olive oils. It was shown that extra virgin olive oils exhibit characteristic yellow fluorescence, due to chlorophylls, while fluorescence of refined oils was blue due to the changes in chlorophyll content during the refining process. This method allowed detecting adulteration of extra virgin olive oils at the level of 5% with refined oils (Sidney & Willoughby, 1929; Glantz, 1930). The use of Wood's lamp was accepted as the U.S. official method for detection of olive oil adulteration. (Kyriakidis & Skarkalis, 2000). The authors of the first papers that reported spectral properties of fluorescence of vegetable oils also point out practical applications of fluorescence spectra as

calibration analysis of fluorescence excitation-emission matrices (Geladi, 2003).

fingerprints in oil analysis (Wolfbeis & Leiner, 1984; Kyriakidis & Skarkalis, 2000).

Olive oil is an economically important product and its quality control and detection of possible fraud are of great interest. Olive oils are classified and priced according to acidity. The most expensive is the high-quality extra-virgin olive oil. This oil may be subject of both mislabeling and adulteration. Refined olive oil is obtained from virgin olive oil using

**3. Application of fluorescence in olive oil analysis** 

**3.1 Discrimination between quality grades of olive oil** 

analysis of the individual mixture components (Bro, 2003; Christensen et al., 2006).

networks.

refining methods that do not alter the initial glyceridic structure; pure olive oil (or simply olive oil) consists of a blend of virgin and refined olive oil. The potential of fluorescence to discriminate olive oils of different quality was the subject of several studies. Both total luminescence and synchronous fluorescence spectra combined with various chemometric approaches were successfully used for this purpose.

Scott et al. (2003) used total luminescence spectra of four different types of edible oils: extra virgin olive, non-virgin olive, sunflower and rapeseed oils. The spectra of undiluted oil samples were measured in the excitation range from 350 to 450 nm with 10 nm intervals and in the emission range from 400 to 720 nm with 5 nm interval. Three supervised neural network algorithms were used for sample classification: simplified fuzzy adaptive resonance theory mapping, traditional back propagation and radial basis function. The 100% correct classification was obtained using all methods.

Guimet et al. (2004a) in a series of studies investigated possibility of application of total fluorescence spectra for discrimination between various quality grades of olive oils. The excitation-emission matrices of undiluted oils were measured using right angle geometry. The hierarchical agglomerative clustering method with the Euclidean distance as a similarity measure and the average linkage method were applied to discriminate between three classes of commercial Spanish olive oils (virgin olive oils, pure olive oils, and olivepomace oils). To optimize the sample grouping into clusters, different preprocessing methods and two spectral ranges were tested, which either included or not the fluorescence peak of chlorophylls. The oils were distinguished using the unfolded excitation–emission fluorescence matrices in the 300-400 nm excitation range and 400-600 nm emission range, thus excluding the chlorophyll band, Fig. 5. The large variations in the chlorophyll band intensity, even between samples of the same type, tend to deteriorate oil discrimination. The optimal preprocessing included normalization of the unfolded spectral excitation–emission fluorescence matrices, followed by column autoscaling. The comparison of the results obtained from the excitation–emission fluorescence matrices to those from a single emission (ex=345, 360, 390 nm) and excitation (em=410 nm) fluorescence spectrum analysis showed the advantage of the total fluorescence data, which result in a significantly better discrimination.

Other studies used unfold PCA and PARAFAC to explore the excitation–emission fluorescence matrices of virgin and pure olive oils (Guimet, 2004b). The spectral ranges studied were ex=300–400 nm, em=400–695 nm and ex=300–400 nm, em=400–600 nm. The first range included chlorophylls, whose peak was much more intense than those of the other components. The second range did not include the chlorophyll peak, being limited to the fluorescence spectra of the oxidation products and vitamin E. The three-component PARAFAC model on the second range (chlorophylls excluded) was found to produce the most useful results. With this model, it was possible to distinguish well between the two groups of oils and to calculate the underlying fluorescent spectra of the three families of compounds. Both unfold PCA and PARAFAC applied to the excitation–emission matrices showed clear differences between fluorescence of the two main groups of olive oils (virgin and pure). Chlorophylls had a strong influence on the models because of their high fluorescence intensity and high variability. Differentiation between the two types of oils was better when the chlorophyll fluorescence region was excluded from the models. The oxidation products are the species that most contribute to the separation between the two

Analysis of Olive Oils by Fluorescence Spectroscopy: Methods and Applications 77

including peroxide value, K232, and K270 (Guimet et al., 2005c). Multiway methods were applied to the data analysis: PARAFAC with multi-linear regression and N-PLS regression. Better regression fits and lower prediction errors were obtained using N-PLS. The best results were obtained for prediction of K270. The detection of extra virgin olive oils was highly degraded at early stages (with high peroxide value) and little oxidized pure olive oils

Synchronous fluorescence spectroscopy combined with multi-dimensional chemometric techniques was applied to the classification of virgin olive oils according to their quality by Poulli et al. (2005). They studied the fluorescence of virgin olive oils and lampante oils, using total fluorescence, synchronous and total synchronous fluorescence spectra. Total luminescence spectra recorded in the 350–720 nm range while exciting in the 320 to 535 nm range showed different shapes and intensities for the two classes of oils. Lampante olive oil had a broad emission with its maximum at 450-500 nm in addition to the 685-690 nm peak. Total synchronous fluorescence spectra measured at =20-180 nm had emission peaks between 500-700 nm, depending on , for both classes of oils. However, lampante oil had additional fluorescence in the of 360–500 nm range, which is not observed for edible virgin olive oil. Classification of virgin olive oils based on their synchronous fluorescence spectra (=80 nm) was performed by hierarchical cluster analysis and PCA using the 429–545 nm spectral range. The authors conclude that the fluorescence in the 429–545 nm range, which they used for data analysis, originates from oleic acid. PCA provided 100% correct discrimination between the two classes, while hierarchical cluster analysis allowed 97.3%

A few papers were published in recent years on the use of fluorescence to assess adulteration detection of virgin olive oils. Adulteration of virgin olive oils has been a common fraud practice that involves addition of cheaper oils, including olive oils of lower quality or other plant oils. The most common adulterants found in virgin olive oil are refined olive oil, pomace oil, residue oil, synthetic olive oil–glycerol products, seed oils, and nut oils. The current analytical standards for olive oil enable detection of the presence of almost all of the possible adulterants; however, they require the measurement of several parameters established by the EU Regulations: (EEC) No 2568/91 and (EC) No 796/2002.. Thus, rapid methods to detect olive oil adulteration are important for quality control

Hazelnut oil is chemically similar to virgin olive oil; its presence is difficult to detect at low concentration levels using standard methods. A different approach was tested to detect this type of adulteration using fluorescence (Sayago et al., 2004; Sayago et al., 2007). The emission spectra of undiluted olive oil mixtures with virgin and refined hazelnut oils with excitation at 350 nm were measured (Sayago et al., 2004). The spectra were subjected to mathematical treatment by calculation of the first derivative. One-way analysis of variance was used for the selection of suitable wavelengths to differentiate oil samples. The response to the addition of adulterant, as evaluated by multiple linear regression, was linear for virgin olive and refined hazelnut oil mixtures (R2=0.99), and for virgin olive and virgin hazelnut oil mixtures (R2=0.98). Stepwise linear discriminant analysis used to discriminate genuine from adulterated olive oil samples allowed 100% correct classifications for each

(with low K270).

correct classification.

**3.2 Adulteration detection of olive oils** 

purposes (Karoui & Blecker, 2011).

groups. PCA was calculated from the emission spectra of oils between em=400 and 695 nm measured at ex=365 nm (Guimet, 2004b).

Fig. 5. Dendrogram of the 56 x 861 unfolded matrix (*λ*ex =300-400 nm, *λ*em = 400-600 nm) using the Euclidean distance as similarity measure and the average linkage method. The distance is expressed as a percentage of dissimilarity (normalized and autoscaled spectra): virgin V, pure, P; and olive-pomace, OP oils. Reprinted with permission from Guimet et al., 2004a. Copyright 2004, American Chemical Society.

Non-negative matrix factorization with Fisher's linear discriminant analysis were applied for discriminating between different types of olive oils: 1) discrimination between commercial Spanish olive oils of different quality (virgin, pure, and olive-pomace oil); 2) discrimination between virgin oils from two ''Siurana'' (Protected Denomination of Origin) regions; 3) discrimination between the original ''Siurana'' virgin olive oils and oils adulterated with olive-pomace oil at 5% (w/w) level (Guimet et al., 2006). In all cases, classifications at above 90% confidence were achieved. The proposed method was also compared to PARAFAC and discriminant N-PLS regression. The classification results were better with non-negative matrix factorization than PARAFAC for two data sets out of three. Non-negative matrix factorization combined with Fisher's linear discriminant analysis was also comparable with discriminant N-PLS regression, giving better classifications for the second data set, but slightly worse results for the other two. The main advantage of non-negative matrix factorization with respect to discriminant N-PLS regression is that its basis functions are more interpretable than the regression loadings, because they are positive and correspond to parts of the spectra than can be directly related to the fluorescent components of oils.

The fluorescence excitation-emission matrices (ex=300-390 nm and em=415-600 nm) were used in studies of the Spanish extra virgin, virgin, pure, and olive pomace oils, to investigate the relationship between oil fluorescence and the conventional quality parameters, including peroxide value, K232, and K270 (Guimet et al., 2005c). Multiway methods were applied to the data analysis: PARAFAC with multi-linear regression and N-PLS regression. Better regression fits and lower prediction errors were obtained using N-PLS. The best results were obtained for prediction of K270. The detection of extra virgin olive oils was highly degraded at early stages (with high peroxide value) and little oxidized pure olive oils (with low K270).

Synchronous fluorescence spectroscopy combined with multi-dimensional chemometric techniques was applied to the classification of virgin olive oils according to their quality by Poulli et al. (2005). They studied the fluorescence of virgin olive oils and lampante oils, using total fluorescence, synchronous and total synchronous fluorescence spectra. Total luminescence spectra recorded in the 350–720 nm range while exciting in the 320 to 535 nm range showed different shapes and intensities for the two classes of oils. Lampante olive oil had a broad emission with its maximum at 450-500 nm in addition to the 685-690 nm peak. Total synchronous fluorescence spectra measured at =20-180 nm had emission peaks between 500-700 nm, depending on , for both classes of oils. However, lampante oil had additional fluorescence in the of 360–500 nm range, which is not observed for edible virgin olive oil. Classification of virgin olive oils based on their synchronous fluorescence spectra (=80 nm) was performed by hierarchical cluster analysis and PCA using the 429–545 nm spectral range. The authors conclude that the fluorescence in the 429–545 nm range, which they used for data analysis, originates from oleic acid. PCA provided 100% correct discrimination between the two classes, while hierarchical cluster analysis allowed 97.3% correct classification.

### **3.2 Adulteration detection of olive oils**

76 Olive Oil – Constituents, Quality, Health Properties and Bioconversions

groups. PCA was calculated from the emission spectra of oils between em=400 and 695 nm

Fig. 5. Dendrogram of the 56 x 861 unfolded matrix (*λ*ex =300-400 nm, *λ*em = 400-600 nm) using the Euclidean distance as similarity measure and the average linkage method. The distance is expressed as a percentage of dissimilarity (normalized and autoscaled spectra): virgin V, pure, P; and olive-pomace, OP oils. Reprinted with permission from Guimet et al.,

Non-negative matrix factorization with Fisher's linear discriminant analysis were applied for discriminating between different types of olive oils: 1) discrimination between commercial Spanish olive oils of different quality (virgin, pure, and olive-pomace oil); 2) discrimination between virgin oils from two ''Siurana'' (Protected Denomination of Origin) regions; 3) discrimination between the original ''Siurana'' virgin olive oils and oils adulterated with olive-pomace oil at 5% (w/w) level (Guimet et al., 2006). In all cases, classifications at above 90% confidence were achieved. The proposed method was also compared to PARAFAC and discriminant N-PLS regression. The classification results were better with non-negative matrix factorization than PARAFAC for two data sets out of three. Non-negative matrix factorization combined with Fisher's linear discriminant analysis was also comparable with discriminant N-PLS regression, giving better classifications for the second data set, but slightly worse results for the other two. The main advantage of non-negative matrix factorization with respect to discriminant N-PLS regression is that its basis functions are more interpretable than the regression loadings, because they are positive and correspond to parts of the spectra than can

The fluorescence excitation-emission matrices (ex=300-390 nm and em=415-600 nm) were used in studies of the Spanish extra virgin, virgin, pure, and olive pomace oils, to investigate the relationship between oil fluorescence and the conventional quality parameters,

measured at ex=365 nm (Guimet, 2004b).

2004a. Copyright 2004, American Chemical Society.

be directly related to the fluorescent components of oils.

A few papers were published in recent years on the use of fluorescence to assess adulteration detection of virgin olive oils. Adulteration of virgin olive oils has been a common fraud practice that involves addition of cheaper oils, including olive oils of lower quality or other plant oils. The most common adulterants found in virgin olive oil are refined olive oil, pomace oil, residue oil, synthetic olive oil–glycerol products, seed oils, and nut oils. The current analytical standards for olive oil enable detection of the presence of almost all of the possible adulterants; however, they require the measurement of several parameters established by the EU Regulations: (EEC) No 2568/91 and (EC) No 796/2002.. Thus, rapid methods to detect olive oil adulteration are important for quality control purposes (Karoui & Blecker, 2011).

Hazelnut oil is chemically similar to virgin olive oil; its presence is difficult to detect at low concentration levels using standard methods. A different approach was tested to detect this type of adulteration using fluorescence (Sayago et al., 2004; Sayago et al., 2007). The emission spectra of undiluted olive oil mixtures with virgin and refined hazelnut oils with excitation at 350 nm were measured (Sayago et al., 2004). The spectra were subjected to mathematical treatment by calculation of the first derivative. One-way analysis of variance was used for the selection of suitable wavelengths to differentiate oil samples. The response to the addition of adulterant, as evaluated by multiple linear regression, was linear for virgin olive and refined hazelnut oil mixtures (R2=0.99), and for virgin olive and virgin hazelnut oil mixtures (R2=0.98). Stepwise linear discriminant analysis used to discriminate genuine from adulterated olive oil samples allowed 100% correct classifications for each

Analysis of Olive Oils by Fluorescence Spectroscopy: Methods and Applications 79

fluorescence spectra for olive oils show a spectral band at around 660 nm, attributable to pigments of chlorophyll group. Moreover, all studied oils save the virgin olive oil show a band at above 315 nm when using =20 nm. This band could be attributed to linoleic acid. It has been suggested that differentiation of virgin olive oil from low quality oils can be achieved using this wavelength region. The PLS regression model was used to quantify adulteration using 20 nm synchronous fluorescence spectra. This technique enabled detection of olivepomace, corn, sunflower, soybean, rapeseed and walnut oil in virgin olive oil at levels of 2.6,

Fig. 6. Predicted versus actual concentrations of sunflower oil in virgin olive oil at a

**3.3 Authentication of virgin olive oils geographical origin** 

wavelength interval of 80 nm: open circles - calibration samples, filled triangles - validation samples. Reprinted with permission from Poulli et al., 2006. Copyright 2006, Springer.

For adulteration detection of extra virgin olive oil with rapeseed oil a neural network method – a simplified fuzzy adaptive resonance theory mapping - was found to be very efficient, resulting in a total of 99.375% correctly classified oil samples at the 5% v/v adulteration level (Scott et al., 2003). It was shown for extra virgin olive oil adulterated by rapeseed oil that the percentage of adulteration may be described by either a radial basis functional network (2.435% RMSE) or a simple Euclidean distance relationship of the PCA scores (2.977% RMSE).

Classification of virgin olive oils according to variety and/or geographical origin is of great importance for the producers, importers, and consumers. Dupuy et al. (2005) developed a method to discriminate olive oils according to their geographic origin. Samples of virgin olive oil from five French registered designations of origin (RDOs), namely, Nyons, Vallé des Baux, Aix-en-Provence, Haute-Provence, and Nice, were discriminated by applying multivariate

3.8, 4.3, 4.2, 3.6, and 13.8% (w/w), respectively (Poulli et al., 2007).

mixture separately, and also for the entire set of samples. Another work explored application of the fluorescence spectroscopy to differentiate between refined hazelnut and refined olive oils (Sayago et al., 2007). Classification of these oils based on their excitation (in 300-500 nm spectral range, using em=655 nm) and emission spectra (in the 650-900 nm range, using ex=350 nm) was performed, using PCA and artificial neural networks. Both methods provided good discrimination between the refined hazelnut and olive oils. Using the artificial neural networks model, the presence of refined hazelnut oils in refined olive oils was robustly detected at levels exceeding 9%.

Several studies devoted to the detection of of adulteration of virgin olive oil with sunflower oil. Poulli et al. (2006) applied total synchronous fluorescence to differentiate virgin olive from sunflower oil and synchronous fluorescence combined with PLS regression for quantitative determination of olive oil adulteration. Total synchronous fluorescence spectra were acquired in the 270–720 nm range, using the wavelength interval variable from Δ=20 to 120 nm. The emission band at around 660 nm was only observed in virgin olive oil, attributed to pigments of the chlorophyll group. For sunflower, in contrast to virgin olive oil, a fluorescence band in the 325–385 nm excitation range is observed. This band was attributed to linoleic acid, however, there are no published data on fluorescence of this compound. In contrast, virgin olive oil has only small signals in this range if scanned at 30 to 50 nm wavelength interval. Synchronous fluorescence spectra of virgin olive oil recorded at Δ=20 nm show a double band at 275 and 297 nm and a single band at 660 nm, in stark contrast to sunflower oil that has an intense band at around 300 nm and a weak one at 325 nm. For quantification of the adulteration, the PLS regression model was used for analysis of synchronous fluorescence spectra of mixtures of virgin olive oil and sunflower oil at Δ=20 and 80 nm, Fig. 6. The detection limits were 3.6% and 3.4% (w/v) when using the 20 and 80 nm wavelength intervals, respectively.

The potential of fluorescence spectroscopy for detecting adulteration of extra virgin olive oil with olive oil has been investigated recently (Dankowska & Małecka, 2009). Synchronous fluorescence spectra were collected in the 240–700 nm range, using Δ=10, 30, 60 and 80 nm. A narrow band at around 300 nm appeared in the synchronous fluorescence spectrum at Δ=10 nm, attributed to tocopherols, and an intense band with a peak at around 665 nm, attributed to compounds of the chlorophyll group. The raw spectra were subject to calculation of the first and second derivatives to find the maximum or the intersection point. Five wavelengths at each of the wavelength intervals were chosen for further analysis. Multiple regression analysis was applied separately to the data acquired at each of the wavelength intervals. The ability to detect olive oil in extra virgin olive oil was better at the wavelength interval of 60 or 80 nm, rather than 10 or 30 nm. Using the spectra acquired at 60 and 80 nm wavelength intervals, the lowest detection limits of adulteration were 8.9% and 8.4% at 350 and 302 nm, respectively.

Fluorescence was used to detect adulteration of virgin olive with others oils (Poulli et al., 2007). Synchronous fluorescence spectra of virgin olive, olive-pomace, corn, sunflower, rapeseed, soybean and walnut oils at 20 nm wavelength interval were used for analysis. Virgin olive oil shows a double band in the 275–297 nm range and a single band at 660 nm, in contrast to other oils that show a strong band around 300 nm and a weak to moderate band near 325 nm. Total synchronous fluorescence spectra were acquired for the excitation wavelength in the 250– 720 nm range and the wavelength interval in the 20 to 120 nm range. Total synchronous

mixture separately, and also for the entire set of samples. Another work explored application of the fluorescence spectroscopy to differentiate between refined hazelnut and refined olive oils (Sayago et al., 2007). Classification of these oils based on their excitation (in 300-500 nm spectral range, using em=655 nm) and emission spectra (in the 650-900 nm range, using ex=350 nm) was performed, using PCA and artificial neural networks. Both methods provided good discrimination between the refined hazelnut and olive oils. Using the artificial neural networks model, the presence of refined hazelnut oils in refined olive

Several studies devoted to the detection of of adulteration of virgin olive oil with sunflower oil. Poulli et al. (2006) applied total synchronous fluorescence to differentiate virgin olive from sunflower oil and synchronous fluorescence combined with PLS regression for quantitative determination of olive oil adulteration. Total synchronous fluorescence spectra were acquired in the 270–720 nm range, using the wavelength interval variable from Δ=20 to 120 nm. The emission band at around 660 nm was only observed in virgin olive oil, attributed to pigments of the chlorophyll group. For sunflower, in contrast to virgin olive oil, a fluorescence band in the 325–385 nm excitation range is observed. This band was attributed to linoleic acid, however, there are no published data on fluorescence of this compound. In contrast, virgin olive oil has only small signals in this range if scanned at 30 to 50 nm wavelength interval. Synchronous fluorescence spectra of virgin olive oil recorded at Δ=20 nm show a double band at 275 and 297 nm and a single band at 660 nm, in stark contrast to sunflower oil that has an intense band at around 300 nm and a weak one at 325 nm. For quantification of the adulteration, the PLS regression model was used for analysis of synchronous fluorescence spectra of mixtures of virgin olive oil and sunflower oil at Δ=20 and 80 nm, Fig. 6. The detection limits were 3.6% and 3.4% (w/v) when using the 20

The potential of fluorescence spectroscopy for detecting adulteration of extra virgin olive oil with olive oil has been investigated recently (Dankowska & Małecka, 2009). Synchronous fluorescence spectra were collected in the 240–700 nm range, using Δ=10, 30, 60 and 80 nm. A narrow band at around 300 nm appeared in the synchronous fluorescence spectrum at Δ=10 nm, attributed to tocopherols, and an intense band with a peak at around 665 nm, attributed to compounds of the chlorophyll group. The raw spectra were subject to calculation of the first and second derivatives to find the maximum or the intersection point. Five wavelengths at each of the wavelength intervals were chosen for further analysis. Multiple regression analysis was applied separately to the data acquired at each of the wavelength intervals. The ability to detect olive oil in extra virgin olive oil was better at the wavelength interval of 60 or 80 nm, rather than 10 or 30 nm. Using the spectra acquired at 60 and 80 nm wavelength intervals, the lowest detection limits of adulteration were 8.9% and

Fluorescence was used to detect adulteration of virgin olive with others oils (Poulli et al., 2007). Synchronous fluorescence spectra of virgin olive, olive-pomace, corn, sunflower, rapeseed, soybean and walnut oils at 20 nm wavelength interval were used for analysis. Virgin olive oil shows a double band in the 275–297 nm range and a single band at 660 nm, in contrast to other oils that show a strong band around 300 nm and a weak to moderate band near 325 nm. Total synchronous fluorescence spectra were acquired for the excitation wavelength in the 250– 720 nm range and the wavelength interval in the 20 to 120 nm range. Total synchronous

oils was robustly detected at levels exceeding 9%.

and 80 nm wavelength intervals, respectively.

8.4% at 350 and 302 nm, respectively.

fluorescence spectra for olive oils show a spectral band at around 660 nm, attributable to pigments of chlorophyll group. Moreover, all studied oils save the virgin olive oil show a band at above 315 nm when using =20 nm. This band could be attributed to linoleic acid. It has been suggested that differentiation of virgin olive oil from low quality oils can be achieved using this wavelength region. The PLS regression model was used to quantify adulteration using 20 nm synchronous fluorescence spectra. This technique enabled detection of olivepomace, corn, sunflower, soybean, rapeseed and walnut oil in virgin olive oil at levels of 2.6, 3.8, 4.3, 4.2, 3.6, and 13.8% (w/w), respectively (Poulli et al., 2007).

Fig. 6. Predicted versus actual concentrations of sunflower oil in virgin olive oil at a wavelength interval of 80 nm: open circles - calibration samples, filled triangles - validation samples. Reprinted with permission from Poulli et al., 2006. Copyright 2006, Springer.

For adulteration detection of extra virgin olive oil with rapeseed oil a neural network method – a simplified fuzzy adaptive resonance theory mapping - was found to be very efficient, resulting in a total of 99.375% correctly classified oil samples at the 5% v/v adulteration level (Scott et al., 2003). It was shown for extra virgin olive oil adulterated by rapeseed oil that the percentage of adulteration may be described by either a radial basis functional network (2.435% RMSE) or a simple Euclidean distance relationship of the PCA scores (2.977% RMSE).

#### **3.3 Authentication of virgin olive oils geographical origin**

Classification of virgin olive oils according to variety and/or geographical origin is of great importance for the producers, importers, and consumers. Dupuy et al. (2005) developed a method to discriminate olive oils according to their geographic origin. Samples of virgin olive oil from five French registered designations of origin (RDOs), namely, Nyons, Vallé des Baux, Aix-en-Provence, Haute-Provence, and Nice, were discriminated by applying multivariate

Analysis of Olive Oils by Fluorescence Spectroscopy: Methods and Applications 81

Fluorescence combined with PLS regression was used to determine tocopherol homologues (α-, β-, γ-, and δ- tocopherol) in the quaternary mixture and the oils (Diaz et al., 2006). The calibration set that included mixtures of tocopherols dissolved in hexane: diethyl ether (70:30 v/v) was constructed based on the central composite plus a full factorial plus a fractionated factorial design. PLS regression was applied to analyze matrices of fluorescence excitation and emission spectra and with fluorescence excitation, emission, and synchronous spectra. For analysis of synthetic samples, recoveries around 100% were obtained. For the analysis of the oils, the samples were diluted in hexane, cleaned in silica cartridges and then tocopherols were eluted with hexane:diethyl ether (90:10 v/v). The method was applied to different edible oils giving satisfactory results for α-, β-, and γ-, but not for δ- tocopherol.

PLS regression was utilized to develop calibration models between front face and right angle synchronous fluorescence spectroscopy for the characterization of edible oils and total tocopherol content as determined by HPLC (Sikorska et al., 2005b). The studies were performed on commercially available edible oils: olive, grapeseed, rapeseed, soybean, sunflower, peanut, and corn oils were analyzed. The regression models showed a good ability to predict tocopherol content. The best fitting results were obtained for 1% v/v diluted oils and for bulk samples using the entire spectrum, yielding the regression

The studies of thermal deterioration of oils are important because changes during oxidation involve degradation of oil constituents and formation of new products that alter quality attributes and nutritional profile, as the oxidation products are potentially toxic. Fluorophores in olive oils are compounds that can participate in oxidation, thus fluorescence spectroscopy can serve as a tool for better understanding of oil oxidation. The fluorescence was compared to other spectroscopic techniques (NIR/VIS, FT-IR and FT-Raman) and chemical and physical methods in determining the deterioration of frying oils, collected from a commercial Chinese spring roll plant, (Engelsen, 1997). Fluorescence has been measured by using five selected excitation wavelengths varying from 395 to 530 nm. Data analysis was performed using PCA and PLS regression. Overal, fluorescence provided the best models for the anisidine value, oligomers, iodine value, and vitamin E

Fluorescence spectrometry and PLS regression were used as a rapid technique for evaluating the quality of heat-treated extra virgin olive (Cheikhousman et al., 2005). Two commercial extra virgin olive oils were heated at 170C for 3 h. Changes in excitation spectra were correlated with changes in concentrations determined by other methods. The fluorescence excitation band emitting at 330 nm was attributed to vitamin E and some fluorescent polyphenols. This fluorescence decreased during the heating process, with the exponential decay constant similar to that obtained chromatographically. Fluorescence excitation spectra with the emission wavelength at 450 nm were inversely correlated with the hydroperoxide content of oil. Indeed, the degradation products generated during heating, particularly the compounds formed by reaction between amino-phospholipids and

Thermal deterioration of extra virgin olive oils was studied by Tena et al. (2009). The sample of virgin olive oil was heated at 190C for 94 h in cycles of 8 h per day. The fluorescence intensity in the spectral region between 290 and 400 nm decreased during the oxidation and

coefficient of 0.991, and the root mean square error of cross-validation of 8%.

**3.5 Monitoring thermal and photo-oxidation of olive oils** 

concentration, among the spectroscopic techniques used.

aldehydes, fluoresce in this wavelength range.

regression methods to synchronous fluorescence spectra of oils. The synchronous fluorescence spectra were collected in the 250-700 nm range at a constant wavelength difference of 30 nm between the excitation and emission wavelengths. The PLS regression analysis of synchronous fluorescence spectra allowed to determine the origin of the oils with satisfactory results, despite the similarity between two denominations of origin (Baux and Aix) that are composed by some common cultivars (Aglandau and Salonenque). The interpretation of the regression coefficients shows that RDOs are correlated to chlorophylls, pheophytins, tocopherols, and phenolic compounds, present in different amount for each origin (Dupuy et al., 2005).

Guimet et al. (2005a) developed a method based on excitation-emission matrices and threeway methods for detecting adulterations of pure olive oils in extra virgin olive oil from the protected denomination of origin (PDO) "Siurana", which is a prestigious distinction given to the extra virgin olive oils produced in a specific area in the south of Catalonia, Spain. Excitation and emission ranges were ex=300–390 nm and em=415–600 nm, respectively. Unfold PCA and PARAFAC were used for exploratory analysis. Discrimination between non-adulterated and adulterated samples was performed using Fisher's linear discriminant analysis and discriminant N-PLS regression. Using discriminant N-PLS regression, 100% correct classification was obtained. Adulteration at around 5% level was quantified, with a prediction error of 1.2% (Guimet et al., 2005a). In another study, Guimet et al. (2005b) achieved discrimination between oils from the two PDO "Siurana" regions by means of discriminant unfold PLS regression, giving correct classification for 94% of samples for "Siurana-Camp" and 100% for "Siurana-Montsant" oils.

#### **3.4 Quantification of fluorescent components in olive oils**

The olive oil autofluorescence is attributed to minor components, species such as tocopherols, phenols and chlorophylls, thus fluorescence spectroscopy has been used to analyze these compounds in olive oils. Diaz et al. (2003) used fluorescence for determination of chlorophylls *a* and *b* and pheophytins *a* and *b* in olive oil samples. The analysis was accomplished by PLS multivariate calibration using the three types of spectra (excitation, emission, and synchronous spectra of these solutions). The best results were obtained for the excitation spectra. The optimum wavelength range to record the excitation spectra (em=662 nm) was selected to minimize the contribution of pheophytin *a* and to maximize the contribution of the other pigments, which are the minor constituents in olive oil. To perform the PLS calibration, a set of samples with final concentration ranges varying from 140 to 560 ng mL-1 for pheophytin *a*, from 10 to 40 ng mL-1 for chlorophyll *a* and pheophytin *b*, and from 20 to 80 ng mL-1 for chlorophyll *b* was used as the calibration matrix. The oil samples were diluted in acetone. Recovery values from olive oil, spiked with chlorophylls *a* and *b* and pheophytins *a* and *b*, were in the ranges of 70-112, 71-111, 76-105, and 82-109%, respectively.

Fluorescence was proposed as an alternative to the Folin-Ciocalteu assay for estimation of the total phenol content in virgin olive oil, olive fruit and leaf polar extracts (Papoti & Tsimidou, 2009). Phenol content in olive oils was determined by measuring the fluorescence intensity of methanol/water extract, with the excitation/emission wavelengths set at 280/320 nm. The method was shown to be more sensitive (limit of detection and limit of quantification values 10-fold lower) and three times faster then the Folin-Ciocalteu assay. Good correlation was found with the results of colorimetric assay (r = 0.69, n = 65) for virgin olive oil extracts.

regression methods to synchronous fluorescence spectra of oils. The synchronous fluorescence spectra were collected in the 250-700 nm range at a constant wavelength difference of 30 nm between the excitation and emission wavelengths. The PLS regression analysis of synchronous fluorescence spectra allowed to determine the origin of the oils with satisfactory results, despite the similarity between two denominations of origin (Baux and Aix) that are composed by some common cultivars (Aglandau and Salonenque). The interpretation of the regression coefficients shows that RDOs are correlated to chlorophylls, pheophytins, tocopherols, and

phenolic compounds, present in different amount for each origin (Dupuy et al., 2005).

"Siurana-Camp" and 100% for "Siurana-Montsant" oils.

**3.4 Quantification of fluorescent components in olive oils** 

were in the ranges of 70-112, 71-111, 76-105, and 82-109%, respectively.

olive oil extracts.

Guimet et al. (2005a) developed a method based on excitation-emission matrices and threeway methods for detecting adulterations of pure olive oils in extra virgin olive oil from the protected denomination of origin (PDO) "Siurana", which is a prestigious distinction given to the extra virgin olive oils produced in a specific area in the south of Catalonia, Spain. Excitation and emission ranges were ex=300–390 nm and em=415–600 nm, respectively. Unfold PCA and PARAFAC were used for exploratory analysis. Discrimination between non-adulterated and adulterated samples was performed using Fisher's linear discriminant analysis and discriminant N-PLS regression. Using discriminant N-PLS regression, 100% correct classification was obtained. Adulteration at around 5% level was quantified, with a prediction error of 1.2% (Guimet et al., 2005a). In another study, Guimet et al. (2005b) achieved discrimination between oils from the two PDO "Siurana" regions by means of discriminant unfold PLS regression, giving correct classification for 94% of samples for

The olive oil autofluorescence is attributed to minor components, species such as tocopherols, phenols and chlorophylls, thus fluorescence spectroscopy has been used to analyze these compounds in olive oils. Diaz et al. (2003) used fluorescence for determination of chlorophylls *a* and *b* and pheophytins *a* and *b* in olive oil samples. The analysis was accomplished by PLS multivariate calibration using the three types of spectra (excitation, emission, and synchronous spectra of these solutions). The best results were obtained for the excitation spectra. The optimum wavelength range to record the excitation spectra (em=662 nm) was selected to minimize the contribution of pheophytin *a* and to maximize the contribution of the other pigments, which are the minor constituents in olive oil. To perform the PLS calibration, a set of samples with final concentration ranges varying from 140 to 560 ng mL-1 for pheophytin *a*, from 10 to 40 ng mL-1 for chlorophyll *a* and pheophytin *b*, and from 20 to 80 ng mL-1 for chlorophyll *b* was used as the calibration matrix. The oil samples were diluted in acetone. Recovery values from olive oil, spiked with chlorophylls *a* and *b* and pheophytins *a* and *b*,

Fluorescence was proposed as an alternative to the Folin-Ciocalteu assay for estimation of the total phenol content in virgin olive oil, olive fruit and leaf polar extracts (Papoti & Tsimidou, 2009). Phenol content in olive oils was determined by measuring the fluorescence intensity of methanol/water extract, with the excitation/emission wavelengths set at 280/320 nm. The method was shown to be more sensitive (limit of detection and limit of quantification values 10-fold lower) and three times faster then the Folin-Ciocalteu assay. Good correlation was found with the results of colorimetric assay (r = 0.69, n = 65) for virgin Fluorescence combined with PLS regression was used to determine tocopherol homologues (α-, β-, γ-, and δ- tocopherol) in the quaternary mixture and the oils (Diaz et al., 2006). The calibration set that included mixtures of tocopherols dissolved in hexane: diethyl ether (70:30 v/v) was constructed based on the central composite plus a full factorial plus a fractionated factorial design. PLS regression was applied to analyze matrices of fluorescence excitation and emission spectra and with fluorescence excitation, emission, and synchronous spectra. For analysis of synthetic samples, recoveries around 100% were obtained. For the analysis of the oils, the samples were diluted in hexane, cleaned in silica cartridges and then tocopherols were eluted with hexane:diethyl ether (90:10 v/v). The method was applied to different edible oils giving satisfactory results for α-, β-, and γ-, but not for δ- tocopherol.

PLS regression was utilized to develop calibration models between front face and right angle synchronous fluorescence spectroscopy for the characterization of edible oils and total tocopherol content as determined by HPLC (Sikorska et al., 2005b). The studies were performed on commercially available edible oils: olive, grapeseed, rapeseed, soybean, sunflower, peanut, and corn oils were analyzed. The regression models showed a good ability to predict tocopherol content. The best fitting results were obtained for 1% v/v diluted oils and for bulk samples using the entire spectrum, yielding the regression coefficient of 0.991, and the root mean square error of cross-validation of 8%.

#### **3.5 Monitoring thermal and photo-oxidation of olive oils**

The studies of thermal deterioration of oils are important because changes during oxidation involve degradation of oil constituents and formation of new products that alter quality attributes and nutritional profile, as the oxidation products are potentially toxic. Fluorophores in olive oils are compounds that can participate in oxidation, thus fluorescence spectroscopy can serve as a tool for better understanding of oil oxidation. The fluorescence was compared to other spectroscopic techniques (NIR/VIS, FT-IR and FT-Raman) and chemical and physical methods in determining the deterioration of frying oils, collected from a commercial Chinese spring roll plant, (Engelsen, 1997). Fluorescence has been measured by using five selected excitation wavelengths varying from 395 to 530 nm. Data analysis was performed using PCA and PLS regression. Overal, fluorescence provided the best models for the anisidine value, oligomers, iodine value, and vitamin E concentration, among the spectroscopic techniques used.

Fluorescence spectrometry and PLS regression were used as a rapid technique for evaluating the quality of heat-treated extra virgin olive (Cheikhousman et al., 2005). Two commercial extra virgin olive oils were heated at 170C for 3 h. Changes in excitation spectra were correlated with changes in concentrations determined by other methods. The fluorescence excitation band emitting at 330 nm was attributed to vitamin E and some fluorescent polyphenols. This fluorescence decreased during the heating process, with the exponential decay constant similar to that obtained chromatographically. Fluorescence excitation spectra with the emission wavelength at 450 nm were inversely correlated with the hydroperoxide content of oil. Indeed, the degradation products generated during heating, particularly the compounds formed by reaction between amino-phospholipids and aldehydes, fluoresce in this wavelength range.

Thermal deterioration of extra virgin olive oils was studied by Tena et al. (2009). The sample of virgin olive oil was heated at 190C for 94 h in cycles of 8 h per day. The fluorescence intensity in the spectral region between 290 and 400 nm decreased during the oxidation and

Analysis of Olive Oils by Fluorescence Spectroscopy: Methods and Applications 83

of the chlorophyll pigments involved in photo-oxidation. Total synchronous fluorescence spectra were obtained by scanning the excitation wavelength in the same spectral range and changing the wavelength interval from 20 to 120 nm at 20 nm steps. These spectra showed considerable changes during oxidation for all of the oils studied, Fig. 7. Fluorescence intensity in the 600–720 nm range, attributed to chlorophylls, decreased significantly. In contrast, the fluorescence bands in the low-wavelength range expanded up to 590 nm. PCA applied to the synchronous fluorescence spectra recorded at =80 nm in the 300–500 nm

range reveals five different classes of oils depending on their oxidation degree.

Fig. 7. Total synchronous fluorescence spectra: contour plots of olive oils before (left) and after 12 h (right) exposure to UV light at 80C. Grayscale indicates fluorescence intensities. Reprinted from *Food Chemistry*, (2009) Vol.117, No.3, Poulli K. I.; Mousdis G. A. & Georgiou

C. A., "Monitoring Olive Oil Oxidation Under Thermal and UV Stress Through Synchronous Fluorescence Spectroscopy and Classical Assays", pp. 499-503, Copyright

(2009), with permission from Elsevier.

a bathochromic shift of the maximum from 350-360 to around 420-440 nm was observed. The fluorescence observed in the 300-390 nm range was assigned to tocopherols together with polyphenols; the information collected from the spectra was compared to the results of the HPLC analysis of these compounds. The observed changes in the spectral profile were explained by the decrease of the tocopherols and phenols and the increase of the oxidation products of vitamin E homologues correlated to K232 and K270, and hydrolysis products. The intensity of the band between 630 and 750 nm, associated with chlorophylls and pheophytins, decreased exponentially with the thermal oxidation time.

The fluorescence intensity recorded at 350 nm and at the wavelength of the spectral maximum occurring in the range of 390-630 nm allowed to explain the increase of the percentage of polar compounds during the experiment. It was stated that the spectra of the undiluted heated oils with maxima at 490 nm or higher correspond to polar compounds exceeding 25%, which is the maximum percentage acceptable for edible oils used in frying.

Poulli et al. (2009b) studied the effect of heating to 100, 150 and 190 C on extra virgin olive, olive pomace, sesame, corn, sunflower, soybean, and a commercial blend of oils. The changes in fluorescence were assessed by measuring total synchronous fluorescence spectra, in the 250- 720 nm excitation range, with the wavelength interval, λ, from 20 to 120 nm at 20 nm step. The synchronous fluorescence intensities below 315 nm recorded at λ=80 nm decreased during heating, presumably due to the consumption of phenolic antioxidants by the lipid radicals generated. The decrease of the fluorescence bands in the 250-350 and 350-400 nm ranges for extra virgin olive and olive pomace oil, respectively, was in accordance with the percentage of trolox equivalent antioxidant capacity reduction. The bands in the total synchronous fluorescence spectra at below 350 nm disappeared during heating, with those at 600-700 nm also decreasing, probably due to the decay of antioxidant compounds and chlorophyll, respectively. The bands in the 400-450 nm range increased, probably due to the formation of secondary oxidation products. PCA of synchronous fluorescence spectra obtained at λ=80 nm allowed oil discrimination according to the degree of oxidation. For extra virgin olive, olive pomace, and sesame oil the spectral range of 300-500 nm was used for classification, while the 320-520 nm range was more appropriate for corn, soybean, and sunflower oil, and a commercial blend of oils. Spectroscopic changes are indicative of oxidative deterioration as measured through wet chemistry methods: peroxide value, panisidine value, totox value, and radical-scavenging capacity (Poulli, 2009a).

Extra virgin olive oil is very stable in the dark; it is susceptible to oxidation under UV light. An accelerated thermal and photooxidation under UV light was studied by Poulli et al. (2009a, 2009b) on samples of extra-virgin, regular-quality and pomace olive oils. Synchronous fluorescence spectra were collected using the 250–720 nm excitation range at =80 nm. Extra virgin olive oil bands in the 300–330 nm range decreased during oxidation, while the fluorescence in the 350–550 nm range increased during the initial 8 h and then remained almost constant for up to 12 h. Regular quality olive oil exhibited fluorescence in the 300–550 nm range. The bands in the 300–370 nm range decreased during oxidation, whereas fluorescence bands in the 370–550 nm range increased during the initial period and remained almost constant afterwards. Also, the fluorescence bands of pomace oil in the 350– 550 nm range decreased during the initial period of the experiment and then a small additional increase was observed. All olive oils show fluorescence bands in the 550–700 nm range, attributed to chlorophyll pigments, intensive in extra virgin olive oil and with very low intensity in olive-pomace oil. These bands decreased significantly due to deterioration

a bathochromic shift of the maximum from 350-360 to around 420-440 nm was observed. The fluorescence observed in the 300-390 nm range was assigned to tocopherols together with polyphenols; the information collected from the spectra was compared to the results of the HPLC analysis of these compounds. The observed changes in the spectral profile were explained by the decrease of the tocopherols and phenols and the increase of the oxidation products of vitamin E homologues correlated to K232 and K270, and hydrolysis products. The intensity of the band between 630 and 750 nm, associated with chlorophylls and

The fluorescence intensity recorded at 350 nm and at the wavelength of the spectral maximum occurring in the range of 390-630 nm allowed to explain the increase of the percentage of polar compounds during the experiment. It was stated that the spectra of the undiluted heated oils with maxima at 490 nm or higher correspond to polar compounds exceeding 25%, which is the maximum percentage acceptable for edible oils used in frying. Poulli et al. (2009b) studied the effect of heating to 100, 150 and 190 C on extra virgin olive, olive pomace, sesame, corn, sunflower, soybean, and a commercial blend of oils. The changes in fluorescence were assessed by measuring total synchronous fluorescence spectra, in the 250- 720 nm excitation range, with the wavelength interval, λ, from 20 to 120 nm at 20 nm step. The synchronous fluorescence intensities below 315 nm recorded at λ=80 nm decreased during heating, presumably due to the consumption of phenolic antioxidants by the lipid radicals generated. The decrease of the fluorescence bands in the 250-350 and 350-400 nm ranges for extra virgin olive and olive pomace oil, respectively, was in accordance with the percentage of trolox equivalent antioxidant capacity reduction. The bands in the total synchronous fluorescence spectra at below 350 nm disappeared during heating, with those at 600-700 nm also decreasing, probably due to the decay of antioxidant compounds and chlorophyll, respectively. The bands in the 400-450 nm range increased, probably due to the formation of secondary oxidation products. PCA of synchronous fluorescence spectra obtained at λ=80 nm allowed oil discrimination according to the degree of oxidation. For extra virgin olive, olive pomace, and sesame oil the spectral range of 300-500 nm was used for classification, while the 320-520 nm range was more appropriate for corn, soybean, and sunflower oil, and a commercial blend of oils. Spectroscopic changes are indicative of oxidative deterioration as measured through wet chemistry methods: peroxide value, p-

pheophytins, decreased exponentially with the thermal oxidation time.

anisidine value, totox value, and radical-scavenging capacity (Poulli, 2009a).

Extra virgin olive oil is very stable in the dark; it is susceptible to oxidation under UV light. An accelerated thermal and photooxidation under UV light was studied by Poulli et al. (2009a, 2009b) on samples of extra-virgin, regular-quality and pomace olive oils. Synchronous fluorescence spectra were collected using the 250–720 nm excitation range at =80 nm. Extra virgin olive oil bands in the 300–330 nm range decreased during oxidation, while the fluorescence in the 350–550 nm range increased during the initial 8 h and then remained almost constant for up to 12 h. Regular quality olive oil exhibited fluorescence in the 300–550 nm range. The bands in the 300–370 nm range decreased during oxidation, whereas fluorescence bands in the 370–550 nm range increased during the initial period and remained almost constant afterwards. Also, the fluorescence bands of pomace oil in the 350– 550 nm range decreased during the initial period of the experiment and then a small additional increase was observed. All olive oils show fluorescence bands in the 550–700 nm range, attributed to chlorophyll pigments, intensive in extra virgin olive oil and with very low intensity in olive-pomace oil. These bands decreased significantly due to deterioration of the chlorophyll pigments involved in photo-oxidation. Total synchronous fluorescence spectra were obtained by scanning the excitation wavelength in the same spectral range and changing the wavelength interval from 20 to 120 nm at 20 nm steps. These spectra showed considerable changes during oxidation for all of the oils studied, Fig. 7. Fluorescence intensity in the 600–720 nm range, attributed to chlorophylls, decreased significantly. In contrast, the fluorescence bands in the low-wavelength range expanded up to 590 nm. PCA applied to the synchronous fluorescence spectra recorded at =80 nm in the 300–500 nm range reveals five different classes of oils depending on their oxidation degree.

Fig. 7. Total synchronous fluorescence spectra: contour plots of olive oils before (left) and after 12 h (right) exposure to UV light at 80C. Grayscale indicates fluorescence intensities. Reprinted from *Food Chemistry*, (2009) Vol.117, No.3, Poulli K. I.; Mousdis G. A. & Georgiou C. A., "Monitoring Olive Oil Oxidation Under Thermal and UV Stress Through Synchronous Fluorescence Spectroscopy and Classical Assays", pp. 499-503, Copyright (2009), with permission from Elsevier.

Analysis of Olive Oils by Fluorescence Spectroscopy: Methods and Applications 85

Fluorescence spectra of olive oils contain information about fluorophores (tocopherols, phenolic compounds, and chlorophylls) that are important for oil quality. The spectra may be used to monitor either selected constituents or to determine overall sample characteristics, which may serve as the spectral fingerprint. The analytic potential of fluorescence is enhanced

The review of literature data demonstrates that fluorescence measurements conducted directly on olive oil samples with subsequent multivariate data analysis can be efficiently used for qualitative and quantitative analysis as a valid alternative to conventional, chemical methods of quality assessment. These methods can be used for oil discrimination and for quantitative determination of fluorescent components after an appropriate calibration. Further studies are needed to resolve various issues that are important for practical application of the fluorescence techniques, among which are the method verification for specific kinds of oil and identification

Grant NN312428239, 2010-2013, from the Polish Ministry of Science and Higher Education is

Berrueta L. A.; Alonso-Salces R. M. & Heberger K. (2007). Supervised Pattern Recognition in

Bro R. (2003). Multivariate Calibration. What is in Chemometrics for the Analytical Chemist?. *Analytica Chimica Acta*, Vol.500, No.1-2, pp. 185-194, ISSN 0003-2670 Cert A.; Moreda W. & Perez-Camino M. C. (2000). Chromatographic Analysis of Minor

Cheikhousman R.; Zude M.; Bouveresse D. J. R.; Leger C. L.; Rutledge D. N. & Birlouez-

Christensen J.; Norgaard L.; Bro R. & Engelsen S. B. (2006). Multivariate Autofluorescence of

Dankowska A. & Małecka M. (2009). Application of Synchronous Fluorescence Spectroscopy

Díaz, T. G.; Durán-Merás, I.; Cáceres, M. I. R.; Murillo, B. R. (2006). Comparison of Different

*Science and Technology*, Vol.111, No.12, pp. 1233-1239, ISSN 1438-9312 Díaz T. G.; Merás I. D.; Correa C. A.; Roldan B. & Cáceres M. I. R. (2003). Simultaneous

*Chemistry*, Vol.51, No.24, pp. 6934-6940, ISSN 1618-2650

Food Analysis. *Journal of Chromatography* A, Vol.1158, No.1-2, pp. 196-214, ISSN

Constituents in Vegetable Oils. *Journal of Chromatography* A, Vol.881, No.1-2, pp.

Aragon I. (2005). Fluorescence Spectroscopy for Monitoring Deterioration of Extra Virgin Olive Oil During Heating. *Analytical and Bioanalytical Chemistry*, Vol.382,

Intact Food Systems. *Chemical Reviews*, Vol.106, No.6, pp. 1979-1994, ISSN 1520-6890

for Determination of Extra Virgin Olive Oil Adulteration. *European Journal of Lipid* 

Fluorometric Determination of Chlorophylls A and B and Pheophytins A and B in Olive Oil by Partial Least-Squares Calibration. *Journal of Agricultural and Food* 

Fluorimetric Signals for the Simultaneous Multivariate Determination of Tocopherols in Vegetable Oils. *Applied Spectroscopy*, Vol.60, No.2, pp. 194-202, ISSN 1943-3530

by application of multivariate data analysis methods for the analysis of spectra.

and quantification of other fluorescent oil constituents.

**4. Conclusions** 

**5. Acknowledgment** 

gratefully acknowledged.

0021-9673

131-148, ISSN 0021-9673

No.6, pp. 1438-1443, ISSN 1618-2650

**6. References** 

An interesting study on thermal oxidation of extra virgin olive oils has been published recently (Navarra et al., 2011). Among different experimental techniques (including FTIR and rheology) time-resolved luminescence was used to investigate early steps of the thermally induced oxidative process. The oxidation process was followed at three different heating temperatures (30, 60 and 90C) as a function of time for up to 35 days. The chlorophyll fluorescence lifetime increased from 6.0 ± 0.1 ns, measured before, to 6.3 ± 0.1 ns, measured after 35 days of experiment. These changes were in agreement with the observed viscosity rise, resulting from formation of polar molecules with propensity to form hydrogen bonds. The viscosity increase reduced the frequency of collisions between the chromophore and its environment, consequently lowering the non-radiative contribution to the luminescence decay.

#### **3.6 Assessing quality changes of olive oil during storage**

Fluorescence spectroscopy was applied to monitoring changes in virgin olive oil during storage (Sikorska et al. 2008b). The extra virgin olive oil samples were stored for the period of 12 month in different conditions: in clear and green glass bottles exposed to light, and in darkness. Changes occurring in olive oil during storage were assessed by total fluorescence and synchronous scanning fluorescence spectroscopy techniques. In the total fluorescence spectra the intensity of emissions ascribed to tocopherols and chlorophyll pigments decreased during storage, depending on the storage conditions. Additional bands appeared in oils exposed to light in the intermediate range of excitation and emission wavelengths. Bands attributed to tocopherols, chlorophylls and those tentatively ascribed to phenolic compounds were observed in the synchronous scanning fluorescence spectra, allowing monitoring of the storage effects on these constituents. PCA of the synchronous fluorescence spectra revealed systematic changes in the overall emission characteristics dependent on the storage conditions, such as exposure to light, and packaging, Fig. 8.

Fig. 8. (a) Scores plot for the two most significant principal components, PC1 vs. PC2, of a PCA of the synchronous scanning fluorescence (SSF) spectra (λ = 10 nm) of virgin olive oil samples. A nonstored sample (Start), and samples stored in different conditions: in darkness (D), in green glass bottles (DG) and in clear glass bottles (CG). The samples are numbered according to the months of storage. The values in brackets describe the fraction of the total variation explained by each of the PC. Each point represents an average of the spectra obtained from three replicates . (b) One-vector loading plots for the PC1 and PC2 principal components. Reprinted with permission from Sikorska et al., 2008b. Copyright 2008, John Wiley and Sons.
