Food Chemistry 118 (2010) 78–83

Contents lists available at ScienceDirect

Food Chemistry
journal homepage: www.elsevier.com/locate/foodchem

Authentication of extra virgin olive oils by Fourier-transform infrared spectroscopy

M.J. Lerma-García *, G. Ramis-Ramos, J.M. Herrero-Martínez, E.F. Simó-Alfonso
Department of Analytical Chemistry, Faculty of Chemistry, University of Valencia, 46100 Burjassot, Valencia, Spain

a r t i c l e i n f o a b s t r a c t

Article history: Fourier-transform infrared spectroscopy (FTIR), followed by multivariate treatment of the spectral data,
Received 30 April 2008 was used to classify vegetable oils according to their botanical origin, and also to establish the composi-
Received in revised form 11 March 2009 tion of binary mixtures of extra virgin olive oil (EVOO) with other low cost edible oils. Oil samples cor-
Accepted 21 April 2009
responding to five different botanical origins (EVOO, sunflower, corn, soybean and hazelnut) were used.
The wavelength scale of the FTIR spectra of the oils was divided in 26 regions. The normalized absorbance
peak areas within these regions were used as predictors. Classification of the oil samples according to
Keywords:
their botanical origin was achieved by linear discriminant analysis (LDA). An excellent resolution among
Botanical origin
Extra virgin olive oil adulteration
all categories was achieved using an LDA model constructed with eight predictors. In addition, multiple
Fourier-transform infrared spectroscopy linear regression models were used to predict the composition of binary mixtures of EVOO with sun-
Linear discriminant analysis flower, corn, soybean and hazelnut oils. For all the binary mixtures, models capable of detecting a low
Multiple linear regression cost oil content in EVOO as low as 5% were obtained.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction McIntyre, & Davies, 2002; Kasemsumran & Kang, 2005; Sato, 1994;
Wesley, Pacheco, & McGill, 1996; Yang, Irudayaraj, & Paradkar,
Oil genuineness is a very important aspect of quality edible oils. 2005), FTIR (Baeten et al., 2005; Dupuy, Duponchel, Huvenne,
Extra virgin olive oil (EVOO) has unique nutritional and sensory Sombret, & Legrand, 1995; Lai, Kemsley, & Wilson, 1999; Ozen &
characteristics, being also a basic component of the Mediterranean Mauer, 2002; Tay et al., 2002; Vlachos et al., 2006; Yang et al.,
diet. The importance of EVOO is mainly attributed to its high con- 2005), FT-Raman (Baeten & Meurens, 1996; López-Díez, Bianchi,
tent of oleic acid and its richness in phenolic compounds, which & Goodacre, 2003; Yang et al., 2005), NMR (Dais & Spyros, 2007;
act as natural antioxidants (Bendini et al., 2007). On the other hand, García-González, Mannina, D’Imperio, Segre, & Aparicio, 2004; Vig-
EVOO is expensive owing to the hard and time-consuming tasks li, Philippidis, Spyros, & Dais, 2003) and MS (Catharino et al., 2005;
involved in the cultivation of olive trees, the harvesting of the fruits, Lay, Liyanage, Durham, & Brooks, 2006; Lerma-García, Ramis-Ra-
and the extraction of the oil. For these reasons, adulterations of mos, Herrero-Martínez, & Simó-Alfonso, 2007, 2008; Lerma-García,
EVOO with olive oils of lower quality, or with oils of a different Simó-Alfonso, Ramis-Ramos, & Herrero-Martínez, 2007; Marcos-
botanical origin are occasionally detected (Catharino et al., 2005; Lorenzo et al., 2002), followed by multivariate statistical analysis
Chiavaro, Vittadini, Rodriguez-Estrada, Cerretani, & Bendini, 2008; of the data, have been described. For this purpose, the contents
Marcos-Lorenzo, Pérez-Pavón, Fernández-Laespada, García-Pinto, of fatty acids (Brodnjak-Voncina et al., 2005), tocopherols
& Moreno-Cordero, 2002; Mariani, Bellan, Lestini, & Aparicio, (Lerma-García, Simó-Alfonso, et al., 2007; Sikorska et al., 2005),
2006; Poulli, Mousdis, & Georgiou, 2006; Tay, Singh, Krishnan, & volatile compounds (Marcos-Lorenzo et al., 2002), amino acids
Gore, 2002; Vlachos et al., 2006). For this reason, European Mediter- (Lerma-García, Ramis-Ramos, et al., 2007) and sterols (Lerma-
ranean countries, which are major suppliers of olive oil in the world García et al., 2008; Mariani et al., 2006), have been used.
market, have adopted common regulations to protect growers and FTIR is a rapid and non-destructive powerful analytical tool for
consumers from fraud (European Union Commission, 1991). the study of edible oils and fats, requiring minimum sample
To establish the authenticity of edible oils, a number of chro- preparation. FTIR is also an excellent tool for quantitative analysis,
matographic (Brodnjak-Voncina, Kodba, & Novic, 2005; Marcos- since the intensities of the spectral bands are proportional to con-
Lorenzo et al., 2002; Mariani et al., 2006), thermal (Chiavaro centration. For this reason, FTIR has been used to distinguish oils
et al., 2008) and spectroscopic methods, including fluorescence from different botanical origins using non-supervised classificatory
(Poulli et al., 2006; Sikorska, Górecki, Khmelinskii, Sikorski, & Koz- techniques (Dupuy et al., 1995; Lai et al., 1999; Rusak, Brown, &
iol, 2005), NIR (Christy, Kasemsumran, Du, & Ozaki, 2004; Downey, Martin, 2003). FTIR has been also used to distinguish EVOOs from
different geographical origins (Bendini, Cerretani, et al., 2007;
* Corresponding author. Tel.: +34 96 354 43 34; fax: +34 96 354 44 36. Galtier et al., 2007; Tapp, Defernez, & Kemsley, 2003) and differ-
E-mail address: [email protected] (M.J. Lerma-García). ent genetic varieties (Gurdeniz, Tokatli, & Ozen, 2007). FTIR

0308-8146/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.foodchem.2009.04.092
 M.J. Lerma-García et al. / Food Chemistry 118 (2010) 78–83 79

applications addressed to detect olive oil adulteration with low origins, and containing 5%, 50% and 80% low cost oil were prepared.
cost edible oils (Baeten et al., 2005; Ozen and Mauer, 2002; Tay These 12 additional binary mixtures were used to validate the pre-
et al., 2002; Vlachos et al., 2006), to evaluate olive oil freshness diction performance of the regression models.
(Sinelli, Cosio, Gigliotti, & Casiraghi, 2007), to study changes
produced by frying (Valdes & Garcia, 2006) and to assess oil oxida- 2.2. FTIR spectra
tion (Guillén & Cabo, 2002; Muik, Lendl, Molina-Diaz, Valcarcel, &
Ayora-Canada, 2007; Vlachos et al., 2006) have been also FTIR spectra were obtained using a Nicolet Nexus FTIR spectro-
described. photometer (Thermo Electron Corporation, Waltham, MA, USA)
In this work, FTIR followed by linear discriminant analysis (LDA) with a resolution of 4 cm 1 at 32 scans. A small quantity of the
of the spectral data was used to classify vegetable oils according to oil samples (2 lL) was directly deposited between two well-pol-
their botanical origin. Also, data treatment by multiple linear ished KBr disks, creating a thin film. Duplicated spectra were re-
regression (MLR) was used to detect and quantify EVOO adultera- corded for all the oil samples and binary mixtures, except the 12
tion with other low cost edible oils, including sunflower, corn, mixtures used as validation set in regression studies which were
soybean and hazelnut oils. recorded three times each. Spectra were scanned in the absorbance
mode from 4000 to 500 cm 1 and the data were handled with the
2. Experimental EZ OMNIC 7.3 software (Thermo Electron Corporation).

2.1. Oil samples and mixtures 2.3. Data treatment and construction of matrices

The vegetable oils employed in this study (Table 1) were either FTIR spectra were divided in the 26 wavelength regions de-
purchased at the local market or kindly donated by the manufac- scribed in Table 2. Each selected spectral region corresponds to a
turers. The botanical origin and quality grade of all the samples peak or a shoulder, representing structural or functional group
were guaranteed by the suppliers. As indicated in Table 1, four information, either about the lipids or minor components of the
samples of each botanical origin were used for training purposes oil samples (see Table 2). For each region, the peak/shoulder area
in classification studies, being the other two samples of each cate- was measured. In order to reduce the variability associated to
gory used to evaluate the prediction capability of the classification the total amount of oil sample used, and to minimize other sources
models. To estimate the adulteration of EVOO with low cost oils by of variance also affecting the intensity of all the peaks, such as the
using regression models, binary mixtures containing EVOO and thickness of the sample and radiation source intensity, normalized
increasing percentages of low cost oil (sunflower, corn, soybean rather than absolute areas were used. Two normalization proce-
or hazelnut) were prepared. To improve robustness of MLR models, dures were tried. In procedure A, for each spectrum, the area of
the objects of the calibration matrix were prepared using EVOOs each region was divided by the sum of the areas of the 26 regions.
and low cost oils from different geographical origins. For instance, In procedure B, the area of each region was divided by each one of
for the sunflower-EVOO pair, oils from different geographical ori- the areas of the other 25 regions; in this way, and since any pair of
gins were selected to prepare a total of seven mixtures containing areas should be considered only once, (26  25)/2 = 325 normal-
0%, 5%, 10%, 30%, 50%, 75% and 100% sunflower oil. Sets of mixtures ized variables were obtained.
containing the same percentages of low cost oil were also prepared For classification studies, two matrices containing 20 objects
for the corn-, soybean- and hazelnut-EVOO pairs. The resulting 28 each, which corresponded to the averages of the duplicated spectra
mixtures were used as calibration set to construct regression mod- of the training samples of Table 1, were constructed. These matri-
els. Additional mixtures of the sunflower-, corn-, soybean- and ces had either 26 or 325 predictors, according to normalization
hazelnut-EVOO pairs, also using oils from different geographical procedures A and B, respectively. A response column, containing
the categories corresponding to the five botanical origins of the oils
Table 1 (corn and corn germ were considered as a single category), was
Botanical origin, number of samples, brand and use during LDA model construction of added to the training matrices. In order to reduce the internal dis-
the oil samples. persion of the categories, which was important to reduce the num-
Origin No. of samples Brand LDA set ber of variables selected during model construction, the means of
the two spectra of each sample, instead of the individual spectra,
Hazelnut 2 Guinama Training
2 Percheron Training were used. For evaluation purposes, two more matrices containing
2 Flumen Evaluation 10 objects each, which corresponded to the averages of the dupli-
Sunflower 2 Koipesol Training cated spectra of each evaluation sample of Table 1, were con-
2 Hacendado Training structed. These matrices also had either 26 or 325 predictors,
1 Capicua Evaluation according to normalization procedures A and B, respectively.
1 Coosol Evaluation
Concerning to the regression studies, and for the sunflower-
Corn 1 Guinama Training EVOO pair, two calibration matrices containing seven objects each,
1 Asua Training
which corresponded to the averages of the duplicated spectra of
1 Artua Evaluation
1 Mazola Evaluation the calibration mixtures, were constructed. According to normali-
zation procedures A and B, the number of predictors was either
Corn germ 1 Guinama Training
1 Hacendado Training 26 or 325, respectively. A response column, containing the low cost
oil percentages of the mixtures, was added to these matrices. For
Extra virgin olive 1 Carbonell Training
1 Grupo Hojiblanca Training validation of the prediction performance, two more matrices con-
1 Borges Training taining three objects each, which corresponded to the averages of
1 Torrereal Training triplicate spectra of validation mixtures, were constructed. The
1 Coosur Evaluation number of predictors was either 26 or 325 predictors, as indicated.
1 Hacendado Evaluation
Analogously, matrices for the corn-, soybean- and hazelnut-EVOO
Soybean 2 Guinama Training pairs were also constructed. Statistical treatment of the data was
2 Biolasi Training
performed using SPSS (v. 12.0.1, Statistical Package for the Social
2 Sojola Evaluation
Sciences, Chicago, IL, USA).
80 M.J. Lerma-García et al. / Food Chemistry 118 (2010) 78–83

Table 2
FTIR spectral regions selected as predictor variables for statistical data treatment.
1
Identification no. Range, cm Functional group Nominal frequency Mode of vibration
a
1 3029–2989 @C–H (trans) 3025 Stretching
@C–H (cis) 3006a Stretching
2 2989–2946 –C–H (CH3) 2953a Stretching (asym)
3 2946–2881 –C–H (CH2) 2924a Stretching (asym)
4 2881–2782 –C–H (CH2) 2853a Stretching (sym)
5 1795–1677 –C@O (ester) 1746a Stretching
–C@O (acid) 1711a Stretching
6 1486–1446 –C–H (CH2) 1465b Bending (scissoring)
7 1446–1425 –C–H (CH3) 1450b Bending (asym)
8 1425–1409 @C–H (cis) 1417a Bending (rocking)
9 1409–1396 @C–H 1400b Bending
10 1396–1382 @C–H –b Bending
11 1382–1371 –C–H (CH3) 1377a Bending (sym)
12 1371–1330 O–H 1359b Bending (in plane)
13 1330–1290 Non-assigned 1319a Bending
14 1290–1211 –C–O 1238a Stretching
–CH2– Bending
15 1211–1147 –C–O 1163a Stretching
–CH2– Bending
16 1147–1128 –C–O 1138b Stretching
17 1128–1106 –C–O 1118a Stretching
18 1106–1072 –C–O 1097a Stretching
19 1072–1043 –C–O –b Stretching
20 1043–1006 –C–O 1033a Stretching
21 1006–929 –HC@CH– (trans) 968a Bending (out of plane)
22 929–885 –HC@CH– (cis)? 914a Bending (out of plane)
23 885–802 @CH2 850b Wagging
24 802–754 –C–H –b Bending (out of plane)
25 754–701 –(CH2)n– 723a Rocking
–HC@CH– (cis) Bending (out of plane)
26 701–640 685b Bending (out of plane)
O–H 650b Bending (out of plane)
a
According to Guillén and Cabo (1998).
b
According to Silverstein, Bassler, and Morrill (1981).

3. Results and discussion

3.1. Classification of vegetable oils according to their botanical origin

using LDA

FTIR spectra of the 30 oil samples of Table 1 were collected.

Fig. 1 shows the spectra of five oils, one for each botanical origin,
tailored at two absorbance units. As it can be observed, the differ-
ences between them were small. Interpretation of the absorption
bands provides information about the molecular skeleton and
functional groups (see Table 2); however, to obtain information
about the botanical origin of the oils, band interpretation is not re-
quired. Information related to the origin was obtained by measur-
ing the peak areas at the selected wavelength ranges. These data
were conveniently handled by multivariate statistical techniques.
LDA, a supervised classificatory technique, is widely recognized
as an excellent tool to obtain vectors showing the maximal resolu-
tion between sets of objects belonging to previously defined cate-
gories. In LDA, vectors minimizing the Wilks’ lambda, kw, are Fig. 1. FTIR spectra of representative samples of EVOO, sunflower, corn, soybean
obtained (Vandeginste et al., 1998). This parameter is calculated and hazelnut oils. Data were tailored at 2 absorbance units to better show the
as the sum of squares of the distances between points belonging differences among small peaks.
to the same category divided by the total sum of squares. Values
of kw approaching zero are obtained with well resolved categories,
whereas overlapped categories made kw to approach one. Up to inclusion in the model exceeds Fin, the entrance threshold of a test
N 1 discriminant vectors are constructed by LDA, being N the of comparison of variances or F-test. However, the entrance of a
lowest value for either the number of predictors or the number new predictor modifies the significance of those predictors which
of categories. are already present in the model. For this reason, after the inclu-
Using the normalized variables, LDA models capable of classify- sion of a new predictor, a rejection threshold, Fout, is used to decide
ing the oil samples according to their botanical origin were con- if one of the other predictors should be removed from the model.
structed. To select the variables to be included in the model, the The process terminates when there are no predictors entering or
SPSS stepwise algorithm was used. According to this algorithm, a being eliminated from the model. The default probability values
predictor is selected when the reduction of kw produced after its of Fin and Fout, 0.05 and 0.10, respectively, were adopted.
 M.J. Lerma-García et al. / Food Chemistry 118 (2010) 78–83 81

Fig. 2. Score plot on an oblique plane of the 3-D space defined by the three first
discriminant functions of the LDA model constructed to classify vegetable oils
according to their botanical origin.

Fig. 3. Predicted (MLR) versus nominal oil percentages for binary mixtures of EVOO
with sunflower (d), corn (j), soybean (+) and hazelnut () oils. The straight-line is
Table 3 y = x.
Predictors selected and corresponding standardized coefficients of the LDA model
constructed to predict the botanical origin of vegetable oils.
3.2. Determination of EVOO adulteration by MLR
Predictora f1 f2 f3 f4
8/9 19.00 8.20 5.14 3.63 In order to quantify EVOO adulteration, the binary mixtures of
10/12 0.75 9.37 4.03 0.92 EVOO with sunflower, corn, soybean and hazelnut oils (Section
11/13 28.35 5.40 3.19 2.13
14/17 16.01 3.49 1.50 0.65
2.1), were measured. To select the predictors used in the construc-
15/17 6.81 5.74 1.01 1.68 tion of MLR models, the SPSS stepwise algorithm, with the default
19/25 13.73 4.33 1.25 0.91 probability values of Fin and Fout, 0.05 and 0.10, respectively, was
21/26 12.51 3.76 0.64 0.95 adopted. MLR models, one for each binary combination of oils,
22/24 7.84 5.13 2.26 1.36
and in each case according to normalization procedures A and B,
a
Pairs of wavelength regions identified according to Table 2. were constructed. Normalization procedure B, which gave models
with higher linear regression coefficients than normalization pro-
cedure A, was selected. A plot showing the predicted versus the
To classify oils according to the five botanical origins, two LDA nominal oil percentages for all the binary mixtures is shown in
models, one for each normalization procedure, were constructed. Fig. 3. The predictors selected and their corresponding non-stan-
Normalization procedure B, which gave a model with a lower kw dardized model coefficients are given in Table 4. It is interesting
value and a smaller number of predictors than normalization pro- to observe that, except in the case of the corn-EVOO mixtures, a
cedure A, was selected for further studies. As shown in Fig. 2 for single predictor was selected to predict the composition of the
normalization procedure B, an excellent resolution between all cat- other three binary mixtures. According to Table 2, the main wave-
egory pairs was achieved (kw = 0.362). Taking into account that a length transitions selected to construct the MLR model for the sun-
large number of categories were simultaneously distinguished, this flower-EVOO pair were @CH2 (wagging), O–H (bending out of
kw value was quite low. The predictors selected by the SPSS step- plane) and (bending out of plane). Analogously, for the hazel-
wise algorithm, and the corresponding standardized coefficients nut-EVOO pair, the selected transitions were @C–H (trans and
of the model, showing the predictors with large discriminant capa- cis, stretching), –C–O (stretching) and –CH2– (bending). Those se-
bilities, are given in Table 3. According to Table 2, the main wave- lected for the soybean-EVOO pair were –(CH2)n – (rocking) and –
length regions selected by the algorithm to construct the LDA HC@CH– (cis, bending out of plane), and those selected for the
model corresponded to @C–H (bending), O–H (bending in plane), corn-EVOO pair were –C–O (stretching) and @C–H (trans and cis,
–C–H (CH3, bending sym), –C–O (stretching), –CH2– (bending),
–HC@CH– (trans, bending out of plane), O–H (bending out of
plane) and (bending out of plane). When leave-one-out valida- Table 4
tion was applied, all the objects of the training set were correctly Predictors selected and their corresponding non-standardized model coefficients
classified. Concerning to the prediction capability of this model, (coef.), linear regression coefficients (R2), average prediction errors (av. pred. error)
and limits of detection (LODs) obtained for the MLR models constructed to predict
and using a 95% probability, all the objects of the evaluation set
olive oil adulteration.
were correctly assigned; thus, the prediction capability of the mod-
el was 100%. On the other hand, the variables obtained by measur- Binary mixture Predictor Coef. R2 Av. pred. LOD,%
error,%
ing the peak areas along the 26 spectral regions of Table 2 were
forced to be included in the model. An improvement of the LDA Hazelnut-EVOO 1/15 1080 0.912 2.0 4.8
Constant 348
model was achieved but at the cost of introducing a large number
of predictors with low significances. For this reason, variable selec- Sunflower-EVOO 23/26 325 0.989 1.7 4.0
Constant 270
tion by the stepwise algorithm was maintained.
Finally, an LDA model with all the available objects, including Corn-EVOO 1/17 353 0.994 1.5 1.3
1/16 221
those used above for either training or evaluation, was constructed.
Constant 309
The predictors selected by the SPSS stepwise algorithm were basi-
Soybean-EVOO 13/25 324 0.980 1.9 1.7
cally the same that those selected above, but the kw value was
Constant 228
smaller (0.202), indicating a satisfactory stability of the model.
82 M.J. Lerma-García et al. / Food Chemistry 118 (2010) 78–83

stretching). For all binary combinations of oils, a transition imply- Dais, P., & Spyros, A. (2007). 31P NMR spectroscopy in the quality control and
authentication of extra-virgin olive oil: A review of recent progress. Magnetic
ing a double bond was selected, which suggest that unsaturated
Resonance in Chemistry, 45, 367–377.
fatty acids were important in the quantification of the mixtures. Downey, G., McIntyre, P., & Davies, A. (2002). Detecting and quantifying sunflower
The regression coefficients, average prediction errors (calculated oil adulteration in extra virgin olive oils from the Eastern Mediterranean by
as the sum of the absolute differences between expected and pre- visible and near-infrared spectroscopy. Journal of Agricultural and Food
Chemistry, 50, 5520–5525.
dicted oil percentages divided by the number of predictions) and Dupuy, N., Duponchel, L., Huvenne, J. P., Sombret, B., & Legrand, P. (1995).
the limits of detection (LODs, calculated as three times the stan- Classification of edible fats and oils by principal component analysis of Fourier
dard deviation at 5% low cost oil level) are also given in Table 4. transform infrared spectra. Food Chemistry, 57, 245–251.
European Union Commission (1991). Regulation 2568/91, Official Journal of the
For all the MLR models and using leave-one-out validation, the European Communities.
average prediction errors were lower than 2%. The four sets of val- Galtier, O., Dupuy, N., Le Dréau, Y., Ollivier, D., Pinatel, C., Kister, J., et al. (2007).
idation mixtures, one for each binary combination of oils Geographic origins and compositions of virgin olive oils determinated
by chemometric analysis of NIR data. Analytica Chimica Acta, 595, 136–
(4  3 = 12 mixtures), gave average prediction errors below 5%. 144.
LODs were in all cases below 5%, being lower than those previously García-González, D. L., Mannina, L., D’Imperio, M., Segre, A. L., & Aparicio, R. (2004).
reported in literature for binary mixtures of the same oils, namely Using 1H and 13C NMR techniques and artificial neural networks to detect the
adulteration of olive oil with hazelnut oil. European Food Research and
4% versus 6% (Vlachos et al., 2006) for sunflower-EVOO, 1.3% versus Technology, 219, 545–548.
9% (Vlachos et al., 2006) for corn-EVOO, 1.7% versus 6% (Vlachos Guillén, M. D., & Cabo, N. (1998). Relationships between the composition of edible
et al., 2006) for soybean-EVOO and 4.8% versus 25% (Ozen and oils and lard and the ratio of the absorbance of specific bands of their Fourier
transform infrared spectra. Role of some bands of the fingerprint region. Journal
Mauer, 2002) and versus 8% (Baeten et al., 2005) for hazelnut-
of Agricultural and Food Chemistry, 46, 1788–1793.
EVOO mixtures. Guillén, M. D., & Cabo, N. (2002). Fourier transform infrared spectra data versus
peroxide and anisidine values to determine oxidative stability of edible oils.
Food Chemistry, 77, 503–510.
4. Conclusions Gurdeniz, G., Tokatli, F., & Ozen, B. (2007). Differentiation of mixtures of
monovarietal olive oils by mid-infrared spectroscopy and chemometrics.
European Journal of Lipid Science and Technology, 109, 1194–1202.
The possibility of classifying vegetable oils according to their Kasemsumran, S., & Kang, N. (2005). Partial least squares processing of near-
botanical origin by using FTIR data has been demonstrated. Using infrared spectra for discrimination and quantification of adulterated olive oils.
Spectroscopy Letters, 38, 839–851.
eight wavelength regions, all the oil samples belonging to five dif- Lai, Y. W., Kemsley, E. K., & Wilson, R. H. (1999). Potential of Fourier transform
ferent botanical origins were correctly classified with an excellent infrared spectroscopy for the authentication of vegetable oils. Journal of
resolution among the categories. In addition, four MLR models con- Agricultural and Food Chemistry, 42, 1154–1159.
Lay, J. O., Liyanage, R., Durham, B., & Brooks, J. (2006). Rapid characterization of
structed with one predictor for sunflower-, soybean- and hazelnut-
edible oils by direct matrix-assisted laser desorption/ionization time-of-flight
EVOO mixtures, and with two predictors for corn-EVOO mixtures, mass spectrometry analysis using triacylglycerols. Rapid Communications in
were capable of detecting a low cost oil content in EVOO as low as Mass Spectrometry, 20, 952–958.
5%. Thus, the capability of the proposed method to identify the Lerma-García, M. J., Ramis-Ramos, G., Herrero-Martínez, J. M., & Simó-Alfonso, E. F.
(2007). Classification of vegetable oils according to their botanical origin using
botanical origin of the oils, and to quantify low cost oils in EVOO amino acid profiles established by direct infusion mass spectrometry. Rapid
has been demonstrated. Communications in Mass Spectrometry, 21, 3751–3755.
Lerma-García, M. J., Ramis-Ramos, G., Herrero-Martínez, J. M., & Simó-Alfonso, E. F.
(2008). Classification of vegetable oils according to their botanical origin using
Acknowledgements sterol profiles established by direct infusion mass spectrometry. Rapid
Communications in Mass Spectrometry, 22, 973–978.
Lerma-García, M. J., Simó-Alfonso, E. F., Ramis-Ramos, G., & Herrero-Martínez, J. M.
Project CTQ2007–61445 (MEC and FEDER funds). M.J.L.-G. (2007). Determination of tocopherols in vegetable oils by CEC using
also thanks the Generalitat Valenciana for an FPI grant for Ph.D. methacrylate ester-based monolithic columns. Electrophoresis, 28, 4128–
4135.
studies. López-Díez, E. C., Bianchi, G., & Goodacre, R. (2003). Rapid quantitative assessment
of the adulteration of virgin olive oils with hazelnut oils using Raman
spectroscopy and chemometrics. Journal of Agricultural and Food Chemistry, 51,
References 6145–6150.
Marcos-Lorenzo, I., Pérez-Pavón, J. L., Fernández-Laespada, M. E., García-Pinto, C., &
Baeten, V., Fernández Pierna, J. A., Dardenne, P., Meurens, M., García-González, D. L., Moreno-Cordero, B. (2002). Detection of adulterants in olive oil by headspace-
& Aparicio-Ruiz, R. (2005). Detection of the presence of hazelnut oil in olive oil mass spectrometry. Journal of Chromatography A, 945, 221–230.
by FT-Raman and FT-MIR spectroscopy. Journal of Agricultural and Food Mariani, C., Bellan, G., Lestini, E., & Aparicio, R. (2006). The detection of the presence
Chemistry, 53, 6201–6206. of hazelnut oil in olive oil by free and esterified sterols. European Food Research
Baeten, V., & Meurens, M. (1996). Detection of virgin olive oil adulteration by and Technology, 223, 655–661.
Fourier transform Raman spectroscopy. Journal of Agricultural and Food Muik, B., Lendl, B., Molina-Diaz, A., Valcarcel, M., & Ayora-Canada, M. J. (2007). Two-
Chemistry, 44, 2225–2230. dimensional correlation spectroscopy and multivariate curve resolution for the
Bendini, A., Cerretani, L., Carrasco-Pancorbo, A., Gómez-Caravaca, A. M., Segura- study of lipid oxidation in edible oils monitored by FTIR and FT-Raman.
Carretero, A., Fernández-Gutiérrez, A., et al. (2007). Phenolic molecules in virgin Analytica Chimica Acta, 593, 54–67.
olive oils: a survey of their sensory properties, health effects, antioxidant Ozen, B. F., & Mauer, L. J. (2002). Detection of hazelnut oil adulteration using
activity and analytical methods. An overview of the last decade. Molecules, 12, FT-IR spectroscopy. Journal of Agricultural and Food Chemistry, 50, 3898–
1679–1719. 3901.
Bendini, A., Cerretani, L., Di Virgilio, F., Belloni, P., Bonoli-Carbognin, M., & Lercker, G. Poulli, K. I., Mousdis, G. A., & Georgiou, C. A. (2006). Synchronous fluorescence
(2007). Preliminary evaluation of the application of the FTIR spectroscopy to spectroscopy for quantitative determination of virgin olive oil adulteration with
control the geographic origin and quality of virgin olive oils. Journal of Food sunflower oil. Analytical and Bioanalytical Chemistry, 386, 1571–1575.
Quality, 30, 424–437. Rusak, D. A., Brown, L. M., & Martin, S. D. (2003). Classification of vegetable oils by
Brodnjak-Voncina, D., Kodba, Z. C., & Novic, M. (2005). Multivariate data analysis in principal component analysis of FTIR spectra. Journal of Chemical Education, 80,
classification of vegetable oils characterized by the content of fatty acids. 541–543.
Chemometrics and Intelligent Laboratory Systems, 75, 31–43. Sato, T. (1994). Application of principal-component analysis of near-infrared
Catharino, R. R., Haddad, R., Cabrini, L. G., Cunha, I. B. S., Sawaya, A. C. H. F., & Eberlin, spectroscopic data of vegetable oils for their classification. Journal of the
M. N. (2005). Characterization of vegetable oils by electrospray ionization mass American Oil Chemists Society, 71, 293–298.
spectrometry fingerprinting: Classification, quality, adulteration, and aging. Sikorska, E., Górecki, T., Khmelinskii, I. V., Sikorski, M., & Koziol, J. (2005).
Analytical Chemistry, 77, 7429–7433. Classification of edible oils using synchronous scanning fluorescence
Chiavaro, E., Vittadini, E., Rodriguez-Estrada, M. T., Cerretani, L., & Bendini, A. spectroscopy. Food Chemistry, 89, 217–225.
(2008). Differential scanning calorimeter application to the detection of refined Silverstein, R. M., Bassler, G. C., & Morrill, T. C. (1981). Spectrometric identification of
hazelnut oil in extra virgin olive oil. Food Chemistry, 110, 248–256. organic compounds. Chichester: John Wiley & Sons.
Christy, A. A., Kasemsumran, S., Du, Y., & Ozaki, Y. (2004). The detection and Sinelli, N., Cosio, M. S., Gigliotti, C., & Casiraghi, E. (2007). Preliminary study on
quantification of adulteration in olive oil by near-infrared spectroscopy and application of mid infrared spectroscopy for the evaluation of the virgin olive oil
chemometrics. Analytical Sciences, 20, 935–940. freshness. Analytica Chimica Acta, 598, 128–134.
 M.J. Lerma-García et al. / Food Chemistry 118 (2010) 78–83 83

Tapp, H. S., Defernez, M., & Kemsley, E. K. (2003). FTIR spectroscopy and Vigli, G., Philippidis, A., Spyros, A., & Dais, P. (2003). Classification of edible oils by
multivariate analysis can distinguish the geographic origin of extra virgin employing 31P and 1H NMR spectroscopy in combination with multivariate
olive oils. Journal of Agricultural and Food Chemistry, 51, 6110–6115. statistical analysis. A proposal for the detection of seed oil adulteration in virgin
Tay, A., Singh, R. K., Krishnan, S. S., & Gore, J. P. (2002). Authentication of olive oil olive oils. Journal of Agricultural and Food Chemistry, 51, 5715–5722.
adulterated with vegetable oils using Fourier transform infrared spectroscopy. Vlachos, N., Skopelitis, Y., Psaroudaki, M., Konstantinidou, V., Chatzilazarou, A., &
Lebensmittel-Wissenschaft und-Technologie, 35, 99–103. Tegou, E. (2006). Applications of Fourier transform-infrared spectroscopy to
Valdes, A. F., & Garcia, A. B. (2006). A study of the evolution of the physicochemical edible oils. Analytica Chimica Acta, 573–574, 459–465.
and structural characteristics of olive and sunflower oils after heating at frying Wesley, I. J., Pacheco, F., & McGill, A. E. J. (1996). Identification of adulterants in olive
temperatures. Food Chemistry, 98, 214–219. oils. Journal of the American Oil Chemists Society, 73, 515–518.
Vandeginste, B. G. M., Massart, D. L., Buydens, L. M. C., De Jong, S., Lewi, P. J., & Yang, H., Irudayaraj, J., & Paradkar, M. M. (2005). Discriminant analysis of edible
Smeyers-Verbeke, J. (1998). Data handling in science and technology part B (p. oils and fats by FTIR, FT-NIR and FT-Raman spectroscopy. Food Chemistry, 93,
237). Amsterdam: Elsevier Science B.V. 25–32.