Food Chemistry xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

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

Evaluation of olive oils from the Mediterranean region by UV–Vis

spectroscopy and Independent Component Analysis
⁎
Francieli C.G.B.S. Alves, Aline Coqueiro, Paulo H. Março, Patrícia Valderrama
Universidade Tecnológica Federal do Paraná (UTFPR), P.O. Box 271, 87301-899 Campo Mourão, Paraná, Brazil

A R T I C L E I N F O A B S T R A C T

Keywords: Extra-virgin olive oil (EVOO) from Mediterranean were analyzed by Ultraviolet–Visible (UV–Vis) spectroscopy
Dienes and Independent Component Analysis (ICA). The use of ICA resolution provided information over dienes (pri-
Tocopherol mary oxidation compound), polyphenolic compounds, tocopherol, carotenoids and chlorophylls when EVOO
Polyphenols was evaluated by UV–Vis spectroscopy. Based on these data, ICA could be used to determine the contribution of
Carotenoids
chemical compounds to the composition of EVOO produced in different regions from Mediterranean. The results
Chlorophylls
suggest that the combination of UV–Vis measurements and ICA makes possible the EVOO evaluation, and can
contribute to suggesting that a foodstuff comes from an alleged origin. The proposed methodology is a low cost,
fast and sample preparation free methodology to highlights the EVOO characteristics produced in the
Mediterranean region.

1. Introduction characteristics of the EVOO produced in the Mediterranean region and

Extra-virgin olive oil (EVOO) has nutritional and sensory char-
acteristics that make it unique and a basic component of the
Mediterranean diet. EVOO is obtained exclusively by mechanical and
physical processes including collecting, washing, and crushing olives,
malaxation of olive paste, centrifugation, storage, and filtration
(Calabriso, Scoditti, Pellegrino, & Carluccio, 2015).
It is known that EVOO present high monounsaturated fatty acids
composition (55–85%), low polyunsaturated fatty acids (2–21%) and,
the presence of highly antioxidative phenolic compounds (Tena, García-
González, & Aparicio, 2009). EVOO is a rich source of α-tocopherol, a
compound belonging to the group generic named as Vitamin E (α-, β-,
γ-, δ-tocopherol and tocotrienols) (Guinazi, Milagres, Pinheiro-
Sant’Ana, & Chaves, 2009). Furthermore, α-tocopherol is suggested as
presenting highest antioxidant activity among the tocopherols
(Yoshida, Niki, & Noguchi, 2003).
The EVOO production is a very important industry for the
Mediterranean countries, both in terms of wealth, health, and tradition
(Souilem et al., 2017). Foods with certificates of quality are gaining
more and more relevance in people's diets. For example, in Andalusia
(southern Spain) there are companies on the olive-oil gastronomic route
and their designations of origin involved in the development of the
tourism product of “olive oil tourism” (de la Torre, Arjona-Fuentes, &
Amador-Hidalgo, 2017). In this sense, is mandatory the development of
analytical methodologies that contribute to highlights the
that can contribute to the evaluation, certification, and designation.
EVOO samples from Mediterranean region have been analyzed
using spectroscopy techniques such as photoluminescence (Sena,
Isnaeni, & Juliastuti, 2017), fluorescence (Tena et al., 2009), and in-
frared (Tena, Aparicio-Ruiz, & García-González, 2013). The coupling of
the spectroscopy techniques with resolution methods in oil analysis
made it possible to resolve the signals corresponding to the various oil
constituents without resorting to physical separation methods
(Gonçalves, Março, & Valderrama, 2014). In this sense, the Multivariate
Curve Resolution with Alternating Least Squares (MCR-ALS) was cou-
pled with ultraviolet–visible (UV–Vis) spectroscopy in order to verify
the modifications that occur in oil samples when heated from ambient
to high temperatures such as happens during frying (Gonçalves et al.,
2014). Infrared spectroscopy and MCR-ALS was employed in the eva-
luation of the aging of edible oils (Le Dréau, Dupuy, Artaud, Ollivier, &
Kister, 2009).
MCR-ALS is based on a bilinear model which assumes that the ob-
served spectra are a linear combination of the spectra of the pure
components in the system (Jaumot, de Juan, & Tauler, 2015). The al-
gorithm steps include the determination of the number of components
by rank analysis methods and an initial estimation of concentration
and/or spectra with as many profiles as the number of components
estimated from the rank analysis. Once the initial estimate is generated,
the iterative optimization step is started (Valderrama, Gonçalves,
Março, Rutledge, & Valderrama, 2016). However, sometimes, suitable

⁎
Corresponding author.
E-mail address: patriciav@utfpr.edu.br (P. Valderrama).

https://doi.org/10.1016/j.foodchem.2018.01.126
Received 28 July 2017; Received in revised form 4 January 2018; Accepted 19 January 2018
0308-8146/ Published by Elsevier Ltd.

Please cite this article as: Alves, F.C.G.B.S., Food Chemistry (2018), https://doi.org/10.1016/j.foodchem.2018.01.126
F.C.G.B.S. Alves et al. Food Chemistry xxx (xxxx) xxx–xxx

results may not be achieved when MCR-ALS is applied. The main reason
for this is the initial estimates step that reaches a local minimum re-
sulting in insufficient curve resolution due to a problem of local
minimum instead of the global minimum. There is a large possibility of
occurrence of this by using the UV–Vis spectroscopy technique due to
the high degree of band overlapping and the lack of selectivity in this
technique (Valderrama et al., 2016). In this sense, other curve resolu-
tion tools can be investigated in order to evaluate UV–Vis spectra data
set, since it is a fast and low-cost technique.
In this paper, EVOO samples from Mediterranean region were
evaluated by UV–Vis spectroscopy coupled with the resolution method
of Independent Component Analysis (ICA). ICA searches for the de-
composition of signals from the mixture into a more statistically in-
dependent component. There are several different algorithms to per-
former ICA resolution (Bell & Sejnowski, 1995; Hyvärinen & Oja, 1997)
however the Joint Approximate Diagonalization of Eigenmatrices
(JADE) (Cardoso & Souloumiac, 1993) present an advantage over the
others. The JADE algorithm is based on matrix computation, involving
matrix diagonalization to calculate Independent Components (ICs).
Other algorithms rely on an optimization procedure and hence may
yield variable results depending on the starting point and on the opti-
mization path followed by the search algorithm (Rutledge &
Bouveresse, 2013). Moreover, some ICA algorithms allow applications
of constraints such as non-negativity (for concentration and spectra)
Fig. 1. UV–Vis spectra to extra-virgin olive oil from different Mediterranean regions.
which drastically reduces the independence between ICs (Parastar,
Jalali-Heravi, & Tauler, 2012).
Here it is employed ICA with JADE algorithm in order to avoid the
problem of local minimum instead of the global minimum of the MCR-
ALS and, break away from the problem reported with other algorithms
in the computation of the ICs. So, the main objective is to verify simi-
larities between EVOO samples from Mediterranean region due to its
constituents from UV–Vis spectra and contribute to the proposal of a
methodology to assist in the highlights the EVOO characteristics pro-
duced in the Mediterranean region.
ICA is a blind source separation method that extracts the pure un-
derlying signals from a signals data set where they are mixed in un-
known proportions. The ICA general model is (Stone, 2004):
X = A. S
where X is the matrix of observed signals (UV–Vis spectra in our case). S
is the matrix of unknown pure source signals. A is the mixing matrix of
unknown coefficients, related to the corresponding proportions. ICA
aims to maximize the non-gaussianity of the extracted signals. Then,
ICA assumes that statistically independent source signals have intensity
distributions that are less Gaussian than are their mixtures (Ammari,
Jouan-Rimbaud-Bouveresse, Boughanmi, & Rutledge, 2012). Recent
researches have shown the ICA with JADE algorithm employed to un-
supervised pattern recognition applications (Ammari, Redjdal, &
Rutledge, 2015; Kassouf, Maalouly, Rutledge, Chebib, & Ducruet, 2014;
Mishra et al., 2016).
Fig. 2. Percentage of variance captured by PCA.

2. Experimental
2.1. Samples
EVOO samples from France (Languedoc Roussillon 0.5%, Marseille
0.5%), Spain (Catalonia 0.2% and Andalusia 0.1%), Italy (Tuscany
0.33%, Sicily 0.35%, Puglia 0.22%), Greece (Crete 0.4 and 0.5%),
Morocco (Atlasgebergte 0.15%) were purchased in the Leiden Holland
marketplace.
range 200–750 nm (steps of 1 nm) in a 1 mm quartz cuvette. Data were
analyzed using MATLAB version R2007b (The Mathworks Inc., MA,
USA) where curve resolution was performed by ICA with JADE algo-
rithm (Cardoso & Souloumiac, 1993). The mathematical steps for ICA
method are described by Rutledge and Bouveresse (2013) and Rutledge
and Bouveresse (2015), while the Matlab code for JADE algorithm can
be found on the website < http://perso.telecom-paristech.fr/
~cardoso/Algo/Jade/jadeR.m > .

2.2. Experimental parameters and software 3. Results and discussion

Samples from each EVOO were analyzed in triplicate. The spectra The UV–Vis spectra of the samples after spectral baseline adjust-
were taken at the room temperature (25 °C). UV–Vis spectra were ac- ment made by the first-order polynomial and smoothed through
quired in an Ocean Optics equipment (model USB-650-UV–VIS) in the Savitsky-Golay algorithm (Savitzky & Golay, 1964) (13 points and first

2
F.C.G.B.S. Alves et al.
Fig. 3. PCA Loadings. (A) PC1 (63.63%); (B) PC2 (19.05%); (C) PC3 (12.77%); (D) PC4 (2.36%); (E) PC5 (0.69%); (F) PC6 (0.33%).
order polynomial), is shown in Fig. 1.
From this spectral (or matrix) data set the mathematical rank in
absence of experimental noise (pseudo-rank or chemical rank), is one of
the crucial points in the analysis. In this work, three methods were used
in order to determine the number of independent components that
corresponds to the matrix rank.
The first method is based on the percentage of variance captured by
Principal Component Analysis (PCA), as shown in Fig. 2. This procedure
was employed by Ribeiro et al. (2017) to NIR data and ICA with JADE
algorithm. Março, Poppi, Scarminio, and Tauler (2011) used this pro-
cedure to estimates the mathematical rank in UV–Vis data. In this way,
when a PC is referred to a small variance captured it suggest that are no
chemical information on this PC. However, sometimes is hard to decide
if the small variance captured is just noise or if this variance can bring
some information regarding on the small quantities of a chemical
compound. Then, the PCA loadings plot, shown in Fig. 3, can be done to
help in choosing the matrix rank (second method). The chemical rank in
molecular fluorescence spectroscopy was properly chosen with the PCA
3
Food Chemistry xxx (xxxx) xxx–xxx
loadings help (Valderrama, Março, Locquet, Ammari, & Rutledge,
2011). With this second method when a PC loading seems like a noise,
this PC can be considered as not informative in the rank evaluation.
Based on the two methods the rank considered was equal 4, since the
PC number 5 it presents a variance captured practically null and his
loadings seem to a noise.
The third method used to matrix rank determination was the
Durbin-Watson (DW) criterion which was applied to the residual ma-
trices after removing the contributions of the calculated signals to find
out which ones had a low signal/noise ratio and could therefore be
assumed to no longer contain informative ICs (Ammari et al., 2012).
This procedure was applied in the rank determination of front-face
fluorescence spectroscopy matrix for the study of the heat stability of
sunflower oil enriched in natural antioxidants by applying ICA (Ammari
et al., 2012). The Fig. 4 shows the results for the application of this
third method.
Based on these results, it is possible to confirm that 4 is the ap-
propriate rank for the matrix data set utilized in this work. As shown in
F.C.G.B.S. Alves et al.
Fig. 4. Durbin-Watson values of the residual matrices for increasing numbers of extracted IC signals from 2 to 6. (A) ICs equal 2; (B) ICs equal 3; (C) ICs equal 4; (D) ICs equal 5; (E) ICs
equal 6.
the Fig. 4, with rank equal to 2 or 3 there are several ranges in the
residual matrix where the DW values are close to zero, that reinforce
that are structured vectors in the residual matrix. By applying 4 ICs, the
residual matrix shows a vector with several range values close to 2.
When DW values are close to 2, the vector is noisy and can be dis-
carded. So, all information was retained on the considered 4 ICs. No
increase in the information was obtained by considering a rank equal to
5 or 6.
ICA model with 4 ICs was applied to the UV–Vis data. For each IC in
Fig. 5, it is possible to observe the recovered signals and the evolution
of its corresponding proportions (scores).
The first IC was attributed to primary oxidation compounds that
occur due to factors such as auto-oxidation and the presence of light.
One of the main constituents of the primary oxidation compounds are
the dienes, which have a maximum absorbance at 220–230 nm (Vieira,
& D’Arce, 1998). This result is in agreement with the signal observed on
the IC number 1. According to the scores for the first IC, all the samples
used in the study had an initial degree of oxidation.
IC number 2 can be attributed to polyphenolic compounds which
show absorption between 270 and 330 nm (Park, Ikegaki, Abreu, &
Alcici, 1998). These compounds have an impact on the sensorial and
quality aspects, and also have relationships with the health benefits
(García-González, Aparicio-Ruiz, & Aparicio, 2008). They have also
been used to characterize monovarietal, olive maturation and applied
extraction technologies (Dimitrios, 2006). The scores for IC number 2
show that samples from Andalusia (Spain), Languedoc Roussillon and
Marseille (France), Tuscany (Italy), and Crete (Greece) show similarity
to each other based on the polyphenolic compounds. The EVOO sam-
ples from Puglia and Sicily (Italy), and Morocco shows similarity to
each other in relation to polyphenol content. On the other hand, the
EVOO samples from Catalonia (Spain) are distinct from the others
considered in the study.
Tocopherol, which has maximum absorption at 325 nm (Gonçalves
et al., 2014), was identified as being the separate signal in IC number 3.
Tocopherols have antioxidant properties and are often considered as a
purity criterion (García-González et al., 2008). According to the scores
plot of IC number 3, the EVOO samples from Andalusia (Spain), Lan-
guedoc-Roussillon (France) and Morocco are similar in relation to to-
copherol contents. Samples from Marseille (France), Puglia and Sicily
4
Food Chemistry xxx (xxxx) xxx–xxx
(Italy) are similar in tocopherol composition. Likewise, samples from
Tuscany (Italy) and Crete (Greece) show similarity in the amounts of
tocopherol. Moreover, the samples from Catalonia (Spain) are more
different from the other samples considered in the study in relation to
the tocopherol quantities.
IC number 4 presents signals characteristic for carotenoids and
chlorophylls, that are responsible for the color of EVOO. Chlorophyll
pigments, such as pheophytin, are responsible for the greenish color;
compounds like lutein and β-carotene from yellow pigments. The
amount and proportions of these pigments depend on the cultivar,
maturation and the olive processing system, besides the storage con-
ditions (García-González et al., 2008). Carotenoids present spectral
absorption in the visible region: α-carotene – 447 nm, β-carotene –
451 nm, and γ-carotene – 462 nm (Koplík, 2015). Chlorophyll has a
strong absorption band overlapping with carotenoid uptake at 420 nm,
and present absorption at 670 nm (Domenici et al., 2014; Giuffrida,
Salvo, Salvo, Pera, & Dugo, 2007; Psomiadou & Tsimiodou, 2002).
The scores plot for IC number 4 shows that samples from Catalonia
(Spain) are similar to samples from Languedoc-Roussillon and Marseille
(France), and Morocco regarding carotenoid and chlorophyll contents.
The samples from Puglia are similar to samples from Sicily (Italy). The
samples from Crete (Greece) and Andalusia (Spain) are similar to each
other. While the samples from Tuscany (Italy) are the most different in
relation to these constituents.
The results suggest that the combination of UV–Vis measurements
and ICA makes possible the EVOO evaluation, and can contribute to
highlights the EVOO characteristics produced in the Mediterranean
region.
4. Conclusion
The proposition of a low cost, fast and sample preparation free
methodology to evaluation, and highlights the EVOO characteristics
produced in the Mediterranean region, that can contribute to certifi-
cation, and designation of the EVOO produced in this region is feasible
by using UV–Vis spectroscopy coupled with ICA.
The use of ICA resolution provided information over dienes (pri-
mary oxidation compound), polyphenolic compounds, tocopherol,
carotenoids and chlorophylls when EVOO was evaluated by UV–Vis
F.C.G.B.S. Alves et al.
Fig. 5. ICA results. (A) IC1; (B) IC2; (C) IC3; (D) IC4. ○ Andalusia-Spain; Catalonia-Spain;
Crete-Greece; Crete-Greece; Marocco.
spectroscopy. Based on these data, ICA could be used to determine the
contribution of chemical compounds to the composition of EVOO pro-
duced in different regions of Mediterranean. The comparison of EVOO
according to their quantities of the polyphenolic compounds, toco-
pherol, carotenoids, and chlorophylls was possible since different score
behavior regarding these compounds could be achieved.
References
Ammari, F., Jouan-Rimbaud-Bouveresse, D., Boughanmi, N., & Rutledge, D. N. (2012).
Study of the heat stability of sunflower oil enriched in natural antioxidants by dif-
ferent analytical techniques and front-face fluorescence spectroscopy combined with
independent component analysis. Talanta, 99, 323–329.
Ammari, F., Redjdal, L., & Rutledge, D. N. (2015). Detection of orange juice frauds using
front-face fluorescence spectroscopy and independent component analysis. Food
Chemistry, 168, 211–217.
Bell, A. J., & Sejnowski, T. J. (1995). An information-maximization approach to blind
separation and blind deconvolution. Neural Computation, 7, 1129–1159.
Calabriso, N., Scoditti, E., Pellegrino, M., & Carluccio, M. A. (2015). Olive oil. In V. R.
Preedy, & R. R. Watson (Eds.). The Mediterranean diet – An evidence-based approach
5
Food Chemistry xxx (xxxx) xxx–xxx
Languedoc-France; Marseille-France; * Puglia-Italy; Sicily-Italy; Tuscany-Italy;
(pp. 135–142). Elsevier: Academic Press.
Cardoso, J. F., & Souloumiac, A. (1993). Blind beamforming for non-Gaussian signals. IEE
Proceedings F: Radar Signal Processing, 140, 362–370.
de la Torre, M. G. M.-V., Arjona-Fuentes, J. M., & Amador-Hidalgo, L. (2017). Olive oil
tourism: Promoting rural development in Andalusia (Spain). Tourism Management
Perspectives, 21, 100–108.
Dimitrios, B. (2006). Sources of natural phenolic antioxidants. Trends in Food Science &
Technology, 17, 505–512.
Domenici, V., Ancora, D., Cifelli, M., Serani, A., Veracini, C. A., & Zandomeneghi, M.
(2014). Extraction of pigment information from NIR-UV-Vis absorption spectra of
extra virgin olive oils. Journal Agricultural Food Chemistry, 62, 9317–9325.
García-González, D. L., Aparicio-Ruiz, R., & Aparicio, R. (2008). Virgin olive oil –
Chemical implications on quality and health. European Journal Lipid Science
Technology, 110, 602–607.
Giuffrida, D., Salvo, F., Salvo, A., Pera, L. L., & Dugo, G. (2007). Pigments composition in
monovarietal virgin olive oils from various sicilian olive varieties. Food Chemistry,
101, 833–837.
Gonçalves, R. P., Março, P. H., & Valderrama, P. (2014). Thermal edible oil evaluation by
UV–Vis spectroscopy and chemometrics. Food Chemistry, 163, 83–86.
Guinazi, M., Milagres, R. C. R. M., Pinheiro-Sant’Ana, H. M., & Chaves, J. B. P. (2009).
Tocopherols and tocotrienols in vegetable oils and eggs. Química Nova, 32(8),
2098–2103.
F.C.G.B.S. Alves et al.
Hyvärinen, A., & Oja, E. (1997). A fast fixed-point algorithm for Independent Component
Analysis. Neural Computation, 9, 1483–1492.
Jaumot, J., de Juan, A., & Tauler, R. (2015). MCR-ALS GUI 2.0: New features and ap-
plications. Chemometrics and Intelligent Laboratory Systems, 140, 1–12.
Kassouf, A., Maalouly, J., Rutledge, D. N., Chebib, H., & Ducruet, V. (2014). Rapid dis-
crimination of plastic packaging materials using MIR spectroscopy coupled with in-
dependent components analysis. Waste Management, 34, 2131–2138.
Koplík, R.. Advanced strategies in food analysis. (2015). http://web.vscht.cz/~poustkaj/EN
%20ASFA%20AU%20Koplík%20UV_VIS_spectrometry.pdf/ Accessed 21.11.15.
Le Dréau, Y., Dupuy, N., Artaud, J., Ollivier, D., & Kister, J. (2009). Infrared study of
aging of edible oils by oxidative spectroscopy index and MCR-ALS chemometric
method. Talanta, 77, 1748–1756.
Março, P. H., Poppi, R. J., Scarminio, I. S., & Tauler, R. (2011). Investigation of the pH
effect and UV radiation on kinetic degradation of anthocyanin mixtures extracted
from Hibiscus acetosella. Food Chemistry, 125, 1020–1027.
Mishra, P., Cordella, C. B. Y., Rutledge, D. N., Barreiro, P., Roger, J. M., & Diezma, B.
(2016). Application of independent components analysis with the JADE algorithm
and NIR hyperspectral imaging for revealing food adulteration. Journal of Food
Engineering, 168, 7–15.
Parastar, H., Jalali-Heravi, M., & Tauler, R. (2012). Is independent component analysis
appropriate for multivariate resolution in analytical chemistry? TrAc Trends in
Analytical Chemistry, 31, 134–143.
Park, Y. K., Ikegaki, M., Abreu, J. A. S., & Alcici, N. M. F. (1998). Study of the preparation
of the propolis extracts and your applications. Food Science and Technology, 18,
313–318.
Psomiadou, E., & Tsimiodou, M. (2002). Stability of virgin olive oil. 1. Autoxidation
studies. Journal Agricultural Food Chemistry, 50, 716–721.
Ribeiro, G. M., Madivadua, D. A., Curti, S. M. M., Pantean, L. P., Março, P. H., &
Valderrama, P. (2017). Pseudo-univariate calibration based on independent compo-
nents analysis for determination of the carbendazim concentration in orange juice.
Microchemical Journal, 134, 114–118.
Rutledge, D. N., & Bouveresse, J.-R. D. (2013). Independent components analysis with the
JADE algorithm. TrAC Trends in Analytical Chemistry, 50, 22–32.
Food Chemistry xxx (xxxx) xxx–xxx
Rutledge, D. N., & Bouveresse, J.-R. D. (2015). Corrigendum to ‘Independent Components
Analysis with the JADE algorithm’. TrAC Trends in Analytical Chemistry, 67, 220.
Savitzky, A., & Golay, M. J. E. (1964). Smoothing and differentiation of data by simplified
least squares procedures. Analytical Chemistry, 36, 1627–1639.
Sena, A. W., Isnaeni, & Juliastuti, E. (2017). Optical characterization of major compounds
in different types of commercial olive oil using photoluminescence method. Procedia
Engineering, 170, 357–362.
Souilem, S., El-Abbassi, A., Kiai, H., Hafidi, A., Sayadi, S., & Galanakis, C. M. (2017).
Olive oil production sector: Environmental effects and sustainability challenges. In C.
M. Galanakis (Ed.). Olive mill waste – Recent advantages for sustainable management
(pp. 1–28). Elsevier: Academic Press.
Stone, J. V. (2004). Independent component analysis: A tutorial introduction. Cambridge:
Massachusetts.
Tena, N., Aparicio-Ruiz, R., & García-González, D. L. (2013). Time course analysis of
fractionated thermoxidized virgin olive oil by FTIR spectroscopy. Journal of
Agricultural and Food Chemistry, 61, 3212–3218.
Tena, N., García-González, D. L., & Aparicio, R. (2009). Evaluation of virgin olive oil
thermal deterioration by fluorescence spectroscopy. Journal of Agricultural and Food
Chemistry, 57, 10505–10511.
Valderrama, L., Gonçalves, R. P., Março, P. H., Rutledge, D. N., & Valderrama, P. (2016).
Independent component analysis as a means to have initial estimates for multivariate
curve resolution-alternating least squares. Journal of Advanced Research, 7, 795–802.
Valderrama, P., Março, P. H., Locquet, N., Ammari, F., & Rutledge, D. N. (2011). A
procedure to facilitate the choice of the number of factors in multi-way data analysis
applied to the natural samples: Application to monitoring the thermal degradation of
oils using front-face fluorescence spectroscopy. Chemometrics and Intelligent
Laboratory Systems, 106, 166–172.
Vieira, T. M. F. S., & D’Arce, M. A. B. R. (1998). Stability of oils heated by microwave: UV-
Spectrophotometric evaluation. Ciência & Tecnologia de Alimentos, 18, 433–437.
Yoshida, Y., Niki, E., & Noguchi, N. (2003). Comparative study on the action of toco-
pherols and tocotrienols as antioxidant: Chemical and physical effects. Chemistry and
Physics of Lipids, 123, 63–75.