Available via license: CC BY 4.0
Content may be subject to copyright.
Sustainability 2019, 11, 6414; doi:10.3390/su11226414 www.mdpi.com/journal/sustainability
Article
Use of FTIR Spectroscopy and Chemometrics
with Respect to Storage Conditions of Moldavian
Dragonhead Oil
Arkadiusz Matwijczuk
1,
*, Tomasz Oniszczuk
2,
*, Alicja Matwijczuk
1
, Edyta Chruściel
1
,
Anna Kocira
3
, Agnieszka Niemczynowicz
4
, Agnieszka Wójtowicz
2
, Maciej Combrzyński
2
and Dariusz Wiącek
5
1
Department of Biophysics, University of Life Sciences in Lublin, 20-950 Lublin, Poland;
alicja.matwijczuk@up.lublin.pl (A.M.); edyta.chrusciel@up.lublin.pl (E.C.)
2
Department of Thermal Technology and Food Process Engineering, University of Life Sciences in Lublin,
20-612 Lublin, Poland; agnieszka.wojtowicz@up.lublin.pl (A.W.); maciej.combrzynski@up.lublin.pl (M.C.)
3
Institute of Agricultural Sciences, State School of Higher Education in Chelm, 22-100 Chelm, Poland;
akocira@pwsz.chelm.pl
4
Department of Analysis and Differential Equations, University of Warmia and Mazury,
10-710 Olsztyn, Poland; niemaga@matman.uwm.edu.pl
5
Department of Physical Properties of Plant Materials, Institute of Agrophysics, Polish Academy of Sciences,
20-290 Lublin, Poland; d.wiacek@ipan.lublin.pl
*
Correspondence: arkadiusz.matwijczuk@up.lublin.pl (A.M.); tomasz.oniszczuk@up.lublin.pl (T.O.);
Tel.: +48-81-445-65-64 (A.M.); +48-81-445-61-18 (T.O.)
Received: 17 August 2019; Accepted: 10 November 2019; Published: 14 November 2019
Abstract: Oils often have similar properties and can be difficult to identify based on color, smell or
taste alone. The present paper suggests the use of Fourier-transform infrared spectroscopy (FTIR)
in combination with chemometric methods to explore similarities and differentiate between
samples of Moldavian dragonhead oil subjected to different storage conditions. Dragonhead is a
plant characterized by very good honey output and ease of cultivation. Principal component
analysis (PCA) was applied to a standard, full range of FTIR spectra. Additionally, hierarchical
cluster analysis (HCA) was employed to explore the organization of the samples in groups relative
to their “proximity” (similarity), by way of Euclidean distance measurement. PC1 and PC2
accounted respectively for 85.4% and 10.1% of the total data variance. PC1 and PC2 were strongly,
negatively correlated within the entire spectral range; the only exception was the region
corresponding to νs(-C-Hvst, -CH
2
) vibrations (aliphatic groups in triglycerides), where PC2 was
positively correlated. The use of FTIR spectral analysis revealed noticeable differences in the
intensity of bands characteristic of the ageing processes (markers of oxidative processes, etc.)
taking place in oleaginous samples and related to the processes of fatty acids oxidation.
Keywords: chemometric analysis; Dracocephalum moldavica; FTIR spectroscopy; functional food
1. Introduction
One of the aspects of sustainable food production is the widest possible use of raw materials
with health-promoting properties. Currently, the bioeconomy is becoming increasingly globalized.
As a result of globalization processes, many previously new products now appear on European
markets, including plant products such as fruit or seeds, which have valuable antioxidative
properties and can potentially offer considerable health benefits to consumers. At the same time,
growing health awareness and the general problem of ageing societies dictate the direction of
research related to functional food intended for particular age groups. Maintaining a well-balanced
Sustainability 2019, 11, 6414 2 of 16
diet largely based on plant products, often those advocated by traditional medicine, is fast becoming
one of the key concerns in our everyday lives. Apart from fruit, the production of functional food
largely depends on oils cold-pressed from seeds. Such oils provide considerable food energy as well
as essential unsaturated fatty acids (UFA), phytosterols, and liposoluble vitamins. Cold-pressed oils
are considered more nutritious due to their antioxidant and provitamins content, e.g., carotenoids,
tocopherols, polyphenols. Oils obtained exclusively from pressing are characterized by a lower
content of oxyphitosterols, carcinogenic and mutagenic compounds, as well as the absence of fatty
acid trans isomers [1,2]. The sensory quality of oil depends on a number of factors such as exposure
to light and oxygen, and time and temperature of storage. Auto- and photo- oxidative processes lead
to the oxidation of unsaturated fatty acids, and consequently, to the formation of fatty acid
hydroperoxides. It is important that the oil is pressed under appropriate conditions, and stored and
packed in an appropriate way.
Moldavian dragonhead (Dracocephalum moldavica L.) is a plant endemic to the Himalayas and
southern Siberia, which has been used for medicinal purposes in Central Asia since the Middle Ages.
The species is a fragrant, annual plant producing essential oils with a strong, lemony scent,
commonly cultivated for ornamental, melliferous, and medicinal purposes [3,4]. Its florescence
usually takes place in June. It produces violet or white flowers located at the top of the shoot in the
form of an apparent ear composed of pseudowhorls. The essential oils produced by the flowers,
stem, and leaves, as well as its sugar-rich nectar, render this plant particularly attractive for
pollinating insects [5]. The plant’s essential oils contain e.g., citral, geranial, neral, geraniol, and
geranyl acetate [6]. Its above ground parts have been identified as a source of flavones, terpenes,
proteins, polypeptides, and 16 amino acid. In August, the plant produces fruit, i.e., schizocarps
containing 4 seeds each. The plant is relatively undemanding in terms of its cultivation, and does not
require particularly fertile or nutrient-rich soils; however, calcium-rich soils and well-maintained
cultures are preferred. Furthermore, cultivation in sun-filled areas facilitates higher concentrations
of the essential oil in the flowers and stalks [1]. The seed yield depends on the method of cultivation,
and the plant’s botanical form and can vary from 2500 to 2800 kg ha−1 for the white and blue cultivar,
respectively [3]. The seeds contain 18–29% of fatty oils rich in essential unsaturated fatty acids
(approx. 90%): α-linolenic (61.0%), linoleic (20%), oleic (8.5%), palmitic (6.5%), and stearic (5.0%) [7].
The seeds also contain 21% protein, with a desirable amino acid composition, mucilage with a
soluble dietary fiber fraction, and essential oil [8]. Moldavian dragonhead oil is considered very
valuable due to its chemical composition; indeed, with its high content of UFAs, the oil obtained
from Dracocephalum moldavica seeds may be classified as one of the most sought after biooils in
phytomedicine and cosmetology. For this reason, the extracts and oil obtained from this plant are
commonly used in the pharmaceutic, cosmetic, and food industries [3].
The widespread use of dragonhead oil has drawn the attention of researchers to the problems
related to its storage and the influence of various factors affecting its quality.
A multivariate data analysis allows us to model the chemical and physical properties of simple
and complex compounds on the basis of spectroscopic data. The scope and applicability of
qualitative and quantitative analyses employing infrared spectroscopy can be enhanced by
embracing a statistical methodology in approaching certain research problems. Researchers working
in a variety of fields are increasingly encouraged to take advantage of this analytical tool, which
further confirms its great scientific potential.
Spectroscopic analyses in the infrared range entail the measurement of the vibration
frequencies in the chemical bonds of functional groups such as e.g., C-C, C-H, O-H, C-O, or N-H,
following the absorption of radiation [9,10]. The measured values are processed by applying a
number of mathematical procedures (including Fourier transform) to the registered absorption
spectrum, which is, in turn, correlated to the actual concentration of respective ingredients in the
sample in a process of calibration. The data were compressed and further processed statistically
using multivariate chemometric techniques, including Principal Component Analysis (PCA) and
Hierarchical Cluster Analysis (HCA).
Sustainability 2019, 11, 6414 3 of 16
In the last few years, numerous scientific publications have demonstrated the usefulness of
spectroscopic methods in the study of the properties of vegetable oils and, above all, the assessment
of the quality of edible oils [11,12]. It has been recognized that these methods are valuable for
monitoring the quality of edible oils. The study of oils can be carried out with the use of absorption
spectroscopy (UV-VIS) [13–15] and infrared spectroscopy (IR) [16–21]. The latter method can be
applied for qualitative and quantitative measurements of parameters of edible oils such as free fatty
acid [22], peroxide value [23], iodine and anisidine value [24,25], lipid classes, and fatty acid
composition [26,27].
The application of chemometric tools for the description, classification, organization,
determination, and exploration of geographic origin and quality control of food products has
recently become a very active research area (e.g., [28,29] and references therein). Many authors have
attempted to use these tools to classify plant foods or other objects. For example, in [30] the authors
applied chemometrics to classify pomegranate juices on the basis of their antioxidant activity. They
reported the main determinant of this parameter to be cultivar. In [31], Wang and coauthors carried
out PCA to gain an overview of the similarities and differences among 10 algal species, and also
investigated the relationships between total phenolic content and different antioxidant activity
assays. There are many examples of such research directions at present.
The present study involved the use of multivariate PCA and HCA analyses to identify the main
sources of variance between the Moldavian dragonhead oil samples stored under different
conditions. The PCA and HCA results were used for the purposes of early classification and
interpretation of differing oil samples in the analyzed set.
The main goal of the study presented in the present paper was to analyze the usability of FTIR
spectroscopy combined with chemometrics for the purposes of controlling the quality of oil obtained
from Moldavian dragonhead seeds, relative to the time and conditions of its storage. Moreover,
relevant spectra were analyzed in detail with the use of the aforementioned analytical methods to
attempt to identify the spectroscopic (infrared) markers (relevant bands) which reflect the varying
rate of ageing processes relative to the conditions under which a given product is stored and the
external stimuli to which it is exposed.
2. Materials and Methods
The research material consisted of oil pressed from the seeds of Moldavian dragonhead
(Dracocephalum moldavica L.). Before pressing, the seeds were stored in bags at room temperature.
Both unheated and thermally-processed seeds were used. The heat treatment entailed heating seeds
to 70 °C, 100 °C, and 130 °C, on a metal tray placed in a laboratory drier for a period of 1 h. The oil
pressing process was conducted using a DUO screw press by Farmet (Czech Republic) with an
efficiency of 18 to 25 kg·h−1 and an engine speed of 1500 rpm. A 10 mm nozzle was used. After
pressing, the oil was left for 2 days to allow natural sedimentation to occur, after which it was placed
in 10 cm3 dark-glass bottles which were impenetrable by sunlight. Some samples were placed in an
argon atmosphere, while others were exposed to oxygen. Directly prior to pressing, control samples
were collected for the respective temperatures of seed drying and pressing atmospheres. The
remaining 80 samples were stored at two different temperatures: 7 °C (refrigerator) and 20–22°C
(air-conditioned laboratory room) for a period of 1, 2, 3, or 6 months, with and without exposure to
light.
The four samples selected for the study were characterized by constant pressing temperature,
i.e., 130 °C, and storage temperature, i.e., refrigerated at 7 °C. Therefore, the variables were: the
storage atmosphere (argon or oxygen), the color of the bottle (dark or clear), and the storage time.
The oil obtained from pressing was analyzed in terms of its fatty acids profile, acid value (AV),
peroxide value (PV), anisidine value (AnV), and iodine value (IV). The general color (GC) was
determined, along with the content of carotenoid and chlorophyll pigments, β-carotene, tocopherols,
and PC-8.
The fatty acids profile was determined with the use of gas chromatography combined with
mass spectrometry. The oil samples were used to obtain methyl esters in accordance with PN-EN
Sustainability 2019, 11, 6414 4 of 16
ISO 12966-2, and their division was conducted using a Trace GC Ultra chromatograph with an ITQ
1100 spectrometer (Thermo Scientific, USA) with the use of a Rtx-2330 column (105 × 0.25 × 0.25 μm)
by Restek. The carrier gas was helium, applied at a constant flow-through rate of 1mL/min.; the
temperature range was from 60 to 250 °C (5 °C/min.), and the injection temperature was 250 °C.
FTIR Measurements: Measurements of ATR-FTIR background corrected spectra (25 scans for
each sample) were carried out with a HATR Ge trough (45° cut, yielding 10 internal reflections)
crystal plate at 20 °C, and were recorded with a 670-IR spectrometer (Agilent Technologies, Santa
Clara, CA 95051, USA). The Ge crystal was cleaned with ultra-pure organic solvents (Sigma-Aldrich,
Darmstadt, Germany). The instrument was continuously purged with argon for 40 min before and
during measurements. Absorption spectra at a resolution of one data point per 1 cm−1 were obtained
in the region between 4000 and 400 cm−1. Scans were Fourier-transformed and averaged with
Grams/AI 8.0 software (Thermo Electron Corporation; Waltham, MA, United States, USA).
Chemometric analysis: All the registered spectra were subjected to multivariate analyses,
specifically, hierarchical cluster analysis (HCA) and principal component analysis (PCA), conducted
with the use of the OriginPro software (OriginLab, Northampton, MA, USA) and PCA for spectra
application. The numbers associated with the names of every sample in the chemometrics analysis
correspond to the conditions of FTIR spectra measurements, that is, number 1 corresponds to the
conditions described on Figures 1 and 2, etc.
3. Results and Discussion
The cold-pressed Moldavian dragonhead oil was characterized by a unique content of fatty
acids [1,2]. Over 90% of the fatty acids present in this oil are unsaturated, of with over 80% are
polyunsaturated fatty acids. The content of palmitic acid (C16:0) was 3.84%, palmitoleic acid (C16:1)
–0.19%, stearic acid (C18:0)–1.71%, oleic acid (C18:1)–6.81%, linoleic acid (C18:2 (9,12), n-6 omega-6
fatty acid)–19.01%, α-linolenic acid ALA (C18:3 (9,12,15), n-3 omega-3 fatty acid)–67.91 %, and other
acids approx. 0.54%. The results showed that the three predominant fatty acids in the Moldavian
dragonhead oil were linolenic (67.9%), oleic (6.8%) and linoleic (19.0%) acids. The content of
saturated fatty acids was very low (less then 6%), whereas the oil was rich in unsaturated ones. The
contents of mono and polyunsaturated fatty acids were 7.0% and 86.9%, respectively. Compared to
other important high-linolenic oils, such as flax and chia oils, the linolenic acid content in Moldavian
dragonhead seed oil was higher than that of flax (50%) and chia (62%) oils. The n-3 to n-6 ratio (3.5)
was higher than that of flax (3.3) and chia (3.2) oils. The human body cannot synthesize linolenic
acid, and therefore, it is known, along with linoleic acid, as an essential fatty acid*. Due to the high
content of this fatty acid and high ratio of n-3/n-6, Moldavian dragonhead seed and the extracted oil
can be used as a food supplement, where enrichment with omega-3 fatty acids is needed.
Figures 1–4 present the ATR-FTIR spectra for the analyzed samples of the oil obtained from
Moldavian dragonhead seeds stored at 7 °C (refrigerated) in an Ar or O2 atmosphere, in a dark or
clear bottle, respectively: a—immediately after pressing, b—two weeks after pressing, c—four weeks
after pressing, d—10 weeks after pressing. The oil was pressed at a temperature of 130 °C. The
experimental constants were the oil pressing temperature (130 °C) and the particular storage
conditions. The samples were spread on a Zn–Se crystal and analyzed under a N2 atmosphere.
Sustainability 2019, 11, 6414 5 of 16
1800 16 50 150 0 1350 1200 10 503000 2850
Wavenumber [cm
-1
]
0
0.02
0.04
0.06
0.08
A
bsorbance [a. u.]
Oils:
a
b
c
d
1745
3008
2957
2925
2852
1701
1652
1558
1513
1459
1429
1395
1369
1304
1268
1236
1101
1160
1064
967
1029
Figure 1. ATR-FTIR spectra for selected Moldavian dragonhead oil samples stored at 7 °C
(refrigerated) in an argon atmosphere in a clear bottle, respectively: a—immediately after pressing,
b—two weeks after pressing, c—four weeks after pressing, d—8 weeks after pressing. The spectra
are presented with the spectral range of 900–3150 cm−1.
1800 16 50 150 0 1350 12 00 10 503000 2850
Wavenumber [cm
-1
]
0
0.02
0.04
0.06
0.08
Absorbance [a. u.]
Oils:
a
b
c
d
1743
3012
2852
2956
2925
1698
1650
1557
1619
1515
1459
1370
1313
1270
1237
1163
1097
1030
975
Figure 2. ATR-FTIR spectra for selected Moldavian dragonhead oil samples stored at 7 °C
(refrigerated) in an O2 atmosphere in a clear bottle, respectively: a—immediately after pressing,
b—two weeks after pressing, c—four weeks after pressing, d—8 weeks after pressing. The spectra
are presented with the spectral range of 900–3150 cm−1.
Sustainability 2019, 11, 6414 6 of 16
Figure 3. ATR-FTIR spectra for selected Moldavian dragonhead oil samples stored at 7 °C
(refrigerated) in an argon atmosphere in a dark bottle, respectively: a—immediately after pressing,
b—two weeks after pressing, c—four weeks after pressing, d—8 weeks after pressing. The spectra
are presented with the spectral range of 900–3150 cm−1.
Figure 4. ATR-FTIR spectra for selected Moldavian dragonhead oil samples stored at 7 °C
(refrigerated) in an O2 atmosphere in a dark bottle, respectively: a—immediately after pressing,
b—two weeks after pressing, c—four weeks after pressing, d—8 weeks after pressing. The spectra
are presented with the spectral range of 900–3150 cm−1.
Table 1 and Tables S1–S3 (in Supplementary Materials) provide a description of all the
characteristic bands present in the oil samples selected for the study, along with the related
vibrations of particular functional groups.
Sustainability 2019, 11, 6414 7 of 16
Table 1. Positions of the maxima of absorption spectra and assignment to the relevant vibrations as
recorded for Moldavian dragonhead oil samples stored at 7 °C (refrigerated), in an argon atmosphere
and a clear bottle, respectively: a—immediately after pressing, b—two weeks after pressing, c—four
weeks after pressing, d—8 weeks after pressing.
FTIR
Type and Origin of Vibrations
Position of Bands (cm−1)
a b c d
3010 3008 3007 3013
ν
(=C-Hm
,
cis-)
2956 2963 2965 2961 νas(-C-Hvst, -CHa) and νs(-C-Hvst, -CHa) (aliphatic groups in
triglycerides)
2926 2926 2925 2928
2853 2853 2852 2854
1743 1742 1743 1743
ν
(-C=Ovst) in esters
1708 1701 1709 1701
ν
(-C=Ovw) in acids
1656 1647 1656 1651
νvw(-C=C-, cis-)
1588 591 1600 1613
- - 1589 1557
- - - 1540
- - - 1515
1459 1460 1460 1457 δvw(-C-H) w CH2 and in CH3, groups, deformation
(scissoring) νvw(-C-H, cis-) deformation (ring)
- - 1429 1428
1373 1375 1375 1369/1393 νw, m, vw (-C-H, -CH3) and deformation
1302 1318 1304/1348 - δm(-C-H, -CH3)
1271 1262 1268 1267 νm(-C-O) or δm(-CH2-)
1236 1240 1237 - νm(-C-O) or δm(-CH2-)
1161 1159 1160 1161 νm(-C-O)
1097 1099 1100 1098
νm,vw(-C-O) 1067 1025 1065 1028
1027 - 1030 -
966 968 964 964 δw(-HC=CH-, trans-) out-of-plain deformation
ν—stretching vibrations, δ—deformation vibrations, s—symmetric, as—asymmetric, st—strong, w—weak.
3.1. FTIR Spectroscopic Analysis of Moldavian Dragonhead Oil Samples
All the infrared (FTIR) spectra for the selected Moldavian dragonhead oil samples revealed
very intensive bands which correspond to specific vibrations of the respective functional groups
contained in ingredients typically found in this type of food. Plant fats and potential oleaginous
materials are substances composed primarily of various fractions of triglyceride groups, mainly
differing in terms of the degree and form of the acyl groups’ unsaturation, as well as the length of
their chains [9]. Numerous publications provide the appropriate associations of the particular
spectral bands in oils, both animal and vegetable, and other fats [10,32–34] with specific vibrations in
particles or groups thereof, although many bands are not easily assigned to respective functional
groups. Table 1 and Tables S2–S3 (in Supplementary Materials) present in detail the frequencies of
characteristic spectra, including the most significant broadenings/enhancements of the respective
spectral bands for the four analyzed time-frames of oil sample storage, as well as their association
with the respective functional groups (with a detailed review and comparison with data available
from the literature [9,35–37]. The subscript text indicates the intensity of the observed bands within
typical IR spectra for this type of biological sample. It should be pointed out that, in this case, the
association of the maxima corresponding to the mode of stretching vibrations in the IR spectra of the
analyzed samples is considerably easier than assigning the bands corresponding to deformation
vibrations. This is due to the fact that bands corresponding to the vibrations of the latter type often
tend to overlap. The presented FTIR spectra reveal vibrations of the methylene group, located in the
Sustainability 2019, 11, 6414 8 of 16
spectral range from 1350 to 1150 cm−1 [9]. They are stretching vibrations of the -C-H group bonded
with CH3 (approx. 1350–1360 cm−1, in out samples 1370 cm−1), as well as deformation vibrations in
this group (~1160 cm−1, in our samples 1159–1165 cm−1). In this case, the stretching vibrations of the
ester bond ν(C-O) are a combination of two asymmetric vibrations, namely those of C-C(=O)-O and
O-C-C [10,38]. The former vibrations are typically considerably more intensive [12]. The bands are
located at approx. 1300 (as C-C(=O)-O, in our case, approx. 1270 cm−1, as visible enhancement of the
band with the maximum at 1235–9 cm−1) and at approx. 1000 cm−1 (in our case 1028 to 1036 cm−1 for
these groups).
In turn, bands related to the vibrations of saturated esters C-C(=O)-O occur between 1240 and
1160 cm−1 (in our case approx. 1234–40 cm−1) [39], whereas for unsaturated esters, the vibrations are
more often generated at lower frequencies [9]. On the other hand, the O-C-O band originating from
primary alcohols appears in the region from 1100 to 1020 cm−1 (in our case approx. 1029–31 cm−1, as
mentioned above), whereas in the case of secondary alcohols, the band usually appears with the
maximum at approx. 1100 cm−1 (in our case 1090–1093 cm−1). Both types of esters descried above are
present in triglyceride particles. In the literature, the mentioned band (at approx. 1239–4 cm−1) has
been associated exclusively with out-of-plane bending vibrations of the methylene group [40].
Another two bands presented in Table 1 and Tables S1–S3 (in Supplementary Materials) (as
well as in Figures 1 and 2) are somewhat more difficult to identify: the maximum of the first band is
at approx. 1416–18 cm−1, and that of the second at approx. 1320 cm−1 (most likely a band broadening,
see Figures 1–4). The first group of vibrations with the maximum at approx. 1416–18 cm−1
(depending on the duration of the experiment) is often assigned to the vibrations of the methyl
groups in the aliphatic chains of the analyzed oils [36,40]. The second group of bands (most likely
band broadening or enhancement) with the maximum at approx. 1320 cm−1 (in all the samples—not
shown so as not to obscure the presentation) is observed simultaneously with the bands with the
maximum at approx. 980 cm−1 and lower wave numbers. It should be noted that the band at approx.
920 cm−1 (depending on duration of the experiment, i.e., more or less intensive), which appears in all
oil samples, is related to the stretching vibrations of cis-substituted olefin groups [35], or can be
connected with the vibrations of the vinyl group.
The oil samples examined at the initial stages of the experiment produced largely similar
spectra in the infrared range. Depending on the duration of the experiment (storage time,
irrespective of analogous storage conditions), the particular spectra started to reveal significant
differences in terms of the intensity and position of the respective bands (the shifts were not large
but very important; discussed further in the text). In each case, we observed the maximum
absorbance, which was clearly correlated to the particular storage conditions (i.e.,
duration/atmosphere and bottle color). All spectra in Figures 1–4 are shown analogically, and
indicate very evident ageing effects in the case of Moldavian dragonhead oil samples.
Other very important vibration regions were also observed with respect to the bands with
maxima at approx. 1745-1 cm−1, which were typical of the stretching vibrations of the carbonyl C=O
group [9] in ester groups. Next to the band (characteristic of the vibrations of the carbonyl group in
esters), we observed, on the lower wavenumber side, a clearly-visible enhancement with the
maximum at approx. 1700-15 cm−1 (whose intensity also increased together with the ageing effect),
which corresponded to the vibrations of a carbonyl group, but in this case, found in the acidic
groups of the analyzed samples [9].
The next band, with a maximum at 1655-3 cm−1, corresponded to the stretching vibrations of the
-C = C- group (particularly the cis-transformation) [33]. It is noteworthy that the intensity of those
bands increased with longer storage times of the respective samples, which clearly evidences
ongoing ageing processes (discussed further in the text). A very characteristic region was also
observed for the vibrations with the maximum at 1461–3 cm−1 and originating from the -C-H
deformation vibrations in CH2 and CH3 groups (bending vibrations). One should also mention the
vibrations in the region from 900 to 650 cm−1 (partially not presented due to low intensity)
corresponding, in the analyzed case, to the characteristic deformation vibrations of the -HC=CH-
Sustainability 2019, 11, 6414 9 of 16
groups (out-of-plain cis-conformation) and ring vibrations of the aforementioned groups (δ(-(CH2)n-
and -HC=CH- (cis-)) [9].
The next very important band corresponded to the =C-H stretching vibrations
(trans-transformation) with a maximum at approx. 3063–4 cm−1 (not shown), which originated from
the vibrations of triglyceride fractions [34]. With respect to the =C-H stretching vibrations in the
cis-configuration, very characteristic and intensive vibrations were observed with the maximum at
approx. 3007/12 cm−1 (Figures 1 and 4, Table 1 and Tables S1–S3 (in Supplementary Materials)).
Vibrations with the maxima at approx. 2952/8, 2922/8, and 2852/7 cm−1 originated, respectively, from
the -C-H stretching vibrations in -CH3, CH2 groups belonging to the aliphatic groups in triglycerides
[34,41].
It should be emphasized that the spectra of the analyzed oil samples revealed clear
discrepancies in the shape of the bands, particularly in the region from 1780 to 1670 cm−1 [36]. Most
of the analyzed samples showed a clearly-defined, slight enhancement of the band at 1743/6 cm−1
(corresponding to the vibrations of the C=O group, as discussed above) on the lower wavenumber
side, with a clear maximum at approx. 1700–16 cm−1 [42], which can be associated with the formation
of a hydrogen bond between C=O…H-O-H groups. Simultaneously to the emergence of the band at
1700–16 cm−1, we observed an increase in intensity at approx. 1350–70 cm−1 [22,42], which can also be
associated with the stretching vibrations of C-O and C-C groups (as described above). Furthermore,
the area between 1100 and 1300 cm−1 also corresponded to stretching vibrations of the C-O group,
but the same indicated minor discrepancies between the analyzed oil samples, regardless of the
storage time. The bands may display a slight increase in intensity with the decreasing affinity of the
particles that generate them toward for the formation of the hydrogen bond between C=O…H-O-H,
and, as such, constitute a perfect marker of the preliminary ageing processes taking place in the
analyzed samples.
In summary, the results of the spectroscopic studies revealed significant differences with
respect to certain bands which constitute important spectroscopic markers of the ageing processes
taking place in the analyzed oil samples. In particular, the observation of spectra within the range
from 1715 to 1500 cm−1 in the samples stored for 8 weeks revealed significant changes in terms of the
position and intensity of the band characteristic of the carbonyl group, with the maximum at approx.
1744 cm−1. The region from 1715 to 1500 cm−1 is related mainly to various vibrations originating from
the C-C and C=C groups and evidencing the progress of ageing processes (with the oxygenation of
fatty acids contained therein). One should also mention the band with a maximum at approx. 1426
cm−1, related to the vibrations of C-H groups in acids. Very significant changes, particularly in the
8th month of the experiment, were observed in the shape of the band with the maximum at approx.
1369 cm−1, as well as with regard to the shift of the band at 1237 cm−1, which also reflected the
aforementioned changes. The impact of the manner of storage was also noticeable: significant
spectral changes were correlated with the varying storage conditions. This confirms the significant
impact of storage conditions on the quality and durability of the product.
3.2. Chemometrics Studies
The HCA analysis allows for the visualization of the group and sub-group arrangement of the
spectra. The HCA (Figure 5) revealed the intragroup similarity within the considered samples and
generated clusters in each group. The difference in the spectral range was established by considering
similar areas of groups in all the samples. While the HCA dendrogram indicates differences in the
groups of investigated samples of oils, there are important questions that remain unanswered. For
example, which variations in the functional group between the samples bring about the difference in
the HCA analysis? How do vibrations of different functional groups in samples vary in the terms of
their intensity and shift? These questions need to be considered to specify the measurement of the
FTIR data. Therefore, PCA was used further in order to get answers to the aforementioned
questions.
Sustainability 2019, 11, 6414 10 of 16
Figure 5. Hierarchical Cluster Analysis (HCA) for FTIR Spectra of all oil samples.
A principal component analysis (PCA) [43] allowed us to visualize a given dataset with respect
to several main components, while accounting for possibly the highest possible percentage of the
set’s variance. After applying the PCA, the initial set of variables is reduced to a number of hidden
variables of principal components (PC) [44–46]. The scree plot (Figure 6) reveals that the greatest
impact on the variance of the analyzed spectra registered for our oil samples was related to the first
three principal components. Figures 7 and 8 present the score plot for the principal components PC1
vs. PC2 in the PCA model corresponding to the Moldavian dragonhead oil samples stored under
various conditions. The results for all samples and the first two principal components PC1 and PC2,
which jointly accounted for 95.5% of the data matrix variance, are presented in Figures 7 and 8. Oil
samples were clearly classified into three groups (Figures 6–8). The first, Group A, includes oil
samples a1, i.e., immediately after pressing (argon, clear bottle), and d1, i.e., 8 weeks after pressing
(argon, clear bottle). The above were samples where no enhancement was observed of the band at
1743 cm−1, characteristic of the vibrations of the C=O group in esters. However, in the remaining
bands (Table 1), clear differences in terms of their intensity and position could be identified.
A sample from Group C, d2, i.e., oil sample stored in an argon atmosphere, in a clear bottle,
analyzed 8 weeks after pressing, clearly stood out from the other samples. This sample formed its
own, one-element cluster. Such an organization into groups of the samples stems from the
measurements of the sample spectra [47]. When analyzing the spectra of all oil samples stored in an
oxygen atmosphere and in clear bottles (a2, b2, c2, d2), we concluded that the highest intensity was
observed for oil sample d2. On the other hand, the remaining oil samples analyzed after 8 weeks in
storage (d3, d4) revealed significant similarities, as evidenced by their relative proximity on the
dendrogram (Figure 5). It can be observed that oils d3 and d4, constituting a subgroup of Group C,
were located in the vicinity of oil d2 (Group C), which indicated significant similarity. At the same
time, sample d1, included in Group A, revealed the highest spectral intensity of all oil samples
analyzed after 8 weeks of storage.
Sustainability 2019, 11, 6414 11 of 16
The remaining oil samples were classified under Group B. This distribution could be the result
of their particular physicochemical properties. Moreover, an analysis of the loading plot (Figures 9
and 10) reveals that PC1 was negatively correlated with all the characteristic spectra of the oil
samples, whereas PC2 was positively correlated only with the νs(-C-Hvst, -CHa) vibrations (aliphatic
groups in triglycerides) (approx. 2854 cm−1).
Figure 6. Plot of eigenvalues for PCA of FTIR spectra.
Figure 7. 3D Score plot of PCA.
Sustainability 2019, 11, 6414 12 of 16
Figure 8. 2D score plot of PCA. PC1 vs. PC2.
Sustainability 2019, 11, 6414 13 of 16
Figure 9. 3D loading plot of PCA.
Figure 10. Loading plot of PC1 and PC2 for reference spectrum a1, a2, a3, a4.
4. Conclusions
1. Principal component analysis (PCA) was used to identify the main sources of variance in the
Fourier-transforms infrared (FTIR) spectra of oil samples obtained from Moldavian dragonhead
seeds and stored under different conditions. PCA combined with HCA allowed the samples to
be explored in terms of their similarities, relative to the storage method with respect to their
FTIR spectra. Due to its inherent simplicity, quick and non-invasive character, this method may
prove useful in monitoring the physicochemical changes in oils or e.g., the oxidative state in oils
relative to the time and conditions under which they are stored.
2. The analyzed oil samples were characterized by a very good fatty acids profile, which
confirmed their value as food products with significant health benefits. Spectral analysis
revealed significant changes with respect to bands associated in the literature to various fat
fractions contained in the oil. The noticeable changes occurring after 8 weeks in storage in
infrared spectra located within the ranges of 1720–1500 cm-1 and ~1426 cm−1, 1369 and 1237 cm−1,
constituted markers which are evidence of the advancement of the ageing processes in the
analyzed samples. Changes related to aging of the sample were related to the intensification of
bands reflecting the vibrations of C-C, C=C, and C=O groups; as such, they constitute perfect
marker bands which can be easily correlated with the given oil’s shelf life and the oxidative
processes that affect it. However, only a detailed chemometric analysis allowed us to
complement and fully follow differences between the respective samples which reflected the
particular storage conditions.
3. The advent of FTIR with multivariate analysis has revolutionized many research fields. FTIR
offers unique advantages, as it reflects the overall vibrations of the components and their
Sustainability 2019, 11, 6414 14 of 16
interactions within the samples as spectra, in addition to being non-invasive and label-free,
unlike conventional methods of this kind.
In this study, we utilized technique FTIR to analyze oil samples from Moldavian dragonhead
seeds. This technique, combined with chemometric analysis, was capable of differentiating the
sample response in relation their similarity and value as food products with significant health
benefits. It is expected that the presented results may prove useful in defining the spectroscopic
markers of the ageing processes that take place in oil samples, which significantly affect the quality
and shelf-life of oil products. Moreover, the study illustrated a reliable, quantitative method of
detecting preliminary differences between oil samples without the need to resort to costly, standard
chemical methods.
Supplementary Materials: The following are available online at www.mdpi.com/xxx/s1.
Author Contributions: Conceptualization, A.M. (Arkadiusz Matwijczuk), T.O. and A.K.; methodology, A.M.
(Arkadiusz Matwijczuk), T.O. and A.M. (Alicja Matwijczuk); validation, A.W., E.C. and M.C.; formal analysis,
A.M. (Arkadiusz Matwijczuk), A.W., D.W. and A.N; data curation. A.M. (Arkadiusz Matwijczuk), T.O., A.K.,
A.M. (Alicja Matwijczuk), A.N., A.W. and D.W. writing—original draft, A.M. (Arkadiusz Matwijczuk), T.O.,
A.K., A.M. (Alicja Matwijczuk), A.N., A.W. and D.W. All authors read and approved the final manuscript.
Funding: This research received no external funding.
Acknowledgments: The part of research of Agnieszka Niemczynowicz in publication was written as a result
internship in Valencia, Spain, co-financed by the European Union under the European Social Fund (Operational
Program Knowledge Education Development), carried out in the project Development Program at the
University of Warmia and Mazury in Olsztyn (POWR.03.05. 00-00-Z310/17). The authors Agnieszka
Niemczynowicz and Arkadiusz Matwijczuk acknowledge the Cost project CA 15126 and CA15216.
Conflicts of Interest: The authors declare no conflict of interest.
References
1. Oniszczuk, A.; Olech, M.; Oniszczuk, T.; Wojtunik-Kulesza, K.; Wójtowicz, A. Extraction methods,
LC-ESI-MS/MS analysis of phenolic compounds and antiradical properties of functional food enriched
with elderberry flowers or fruits. Arab. J. Chem. 2016, doi:10.1016/j.arabjc.2016.09.003, in press.
2. Wójtowicz, A.; Oniszczuk, A.; Oniszczuk, T.; Kocira, S.; Wojtunik, K.; Mitrus, M.; Kocira, A.; Widelski, J.;
Skalicka-Wożniak, K. Application of Moldavian dragonhead (Dracocephalum moldavica L.) leaves addition
as a functional component of nutritionally valuable corn snacks. J. Food Sci. Technol. 2017, 54, 3218–3229,
doi:10.1007/s13197-017-2765-7.
3. Wolski, T.; Kwiatkowski, S. Biology of growth and development of Moldavian dragonhead
(Dracocephalum moldavica L.)–aromatic and medicinal plant. Post. Fitoter. 2006, 1, 2–10.
4. Kocira, S.; Sujak, A.; Kocira, A.; Wójtowicz, A.; Oniszczuk, A. Effect of Fylloton application on
photosynthetic activity of Moldavian dragonhead (Dracocephalum moldavica L.). Agric. Agric. Sci. Procedia
2015, 7, 108–112, doi:10.1016/j.aaspro.2015.12.002.
5. Dmitruk, M.; Weryszko-Chmielewska, E.; Sulborska, A. Flowering and nectar secretion in two forms of
the Moldavian dragonhead (Dracocephalum moldavica L.)—A plant with extraordinary apicultural
potential. J. Apic. Sci. 2018, 62, 97–110, doi:10.2478/jas-2018-0010.
6. Ehsani, A.; Alizadeh, O.; Hashemi, M.; Afshari, A.; Aminzare, M. Phytochemical, antioxidant and
antibacterial properties of Melissa officinalis and Dracocephalum moldavica essential oils. Vet. Res. Forum 2017,
8, 223–229.
7. Domokos, J.; Peredi, J.; Halasz-Zelnik, K. Characterization of seed oils of dragonhead (Dracocephalum
moldavica L.) and catnip (Nepeta cataria var. citriodora Balb.). Ind. Crop. Prod. 1994, 3, 91–94,
doi:10.1016/0926-6690(94)90081-7.
8. Dziki, D.; Miś, A.; Gładyszewska, B.; Laskowski, J.; Kwiatkowski, S.; Gawlik-Dziki, U. Physicochemical
and grinding characteristics of dragonhead seeds. Int. Agrophys. 2013, 27, 403–408,
doi:10.2478/intag-2013-0010.
Sustainability 2019, 11, 6414 15 of 16
9. Guillén, M.D.; Cabo, N. Relationships between the composition of edible 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. J. Agric. Food Chem. 1998, 46, 1788–1793, doi:10.1021/jf9705274.
10. Rohman, A.; Man, Y.C. Fourier transform infrared (FTIR) spectroscopy for analysis of extra virgin olive oil
adulterated with palm oil. Food Res. Int. 2010, 43, 886–892, doi:10.1016/j.foodres.2009.12.006.
11. Nawrocka, A.; Lamorska, J. Determination of food quality by using spectroscopic methods. In Advances in
Agrophysical Research; Intech Open: Rijeka, Croatia, 2013; 347-368, doi:10.5772/3341.
12. Yasushi. E. Analytical methods to evaluate the quality of edible fats and oils: The JOCS standard methods
for analysis of fats, oils and related materials (2013) and advanced methods. J. Oleo Sci. 2018, 67, 1–10, doi:
10.5650/jos.ess17130.
13. Gonçalves, R.P.; Março, P.H.; Valderrama, P. Thermal edible oil evaluation by UV–Vis spectroscopy and
chemometrics. Food Chem. 2014, 163, 83–86, doi:10.1016/j.foodchem.2014.04.109.
14. Leder, P.J.S.; Porcu, O.M. The importance of UV-Vis spectroscopy: Application in food products
characterization. Scholar. J. Food Nutr. 2018, 1, doi:10.32474/SJFN.2018.01.000111.
15. Kruzlcova, D.; Mocak, J.; Katsoyannos, E.; Lankmayr, E. Classification and characterization of olive oils
byUV-Vis absorption spectrometry and sensorial analysis. J. Food Nutr. Res. 2008, 47, 181–188.
16. Alexa, E.; Dragomirescu, A.; Pop, G.; Jianu, C.; Drago, D. The use of FT-IR spectroscopy in the
identification of vegetable oils adulteration. J. Food Agric. Environ. 2009, 7, 20–24, doi: 10.1234/4.2009.1525.
17. Nigri, S.; Oumeddour, R. Fourier transform infrared and fluorescence spectroscopy for analysis of
vegetable oils. MATEC Web Conf. 2013, 5, 04028, doi: 10.1051/matecconf/20130504028.
18. Kluczyk, D.; Matwijczuk, A.; Górecki, A.; Karpińska, M. M.; Szymanek, M.; Niewiadomy, A.; Gagoś, M.
Molecular organization of dipalmitoylphosphatidylcholine bilayers containing bioactive compounds
4-(5-heptyl-1, 3, 4-thiadiazol-2-yl) benzene-1, 3-diol and 4-(5-methyl-1, 3, 4-thiadiazol-2-yl) benzene-1,
3-diols. J. Phys. Chem. B, 2016, 120, 12047-12063, https://doi.org/10.1021/acs.jpcb.6b09371.
19. Rohman, A. Infrared spectroscopy for quantitative analysis and oil parameters of olive oil and virgin
coconut oil: A review. Int. J. Food Prop. 2017, 20, 1447–1456, doi: 10.1080/10942912.2016.1213742.
20. Siddiqui, N.; Ahmad, A. Infrared spectroscopic studies on edible and medicinal oils. Int. J. Sci. Environ.
Tech. 2013, 2, 1297–1306.
21. Karim, M.M.; Karim, S.F.E.; Rana, A.A.; Masum, S.M.; Mondol, A.; Israt, S.S. ATR-FTIR spectroscopy and
chemometric techniques for the identification of edible vegetable oils. Bangladesh J. Sci. Ind. Res. 2015, 50,
233–240, doi: 10.3329/bjsir.v50i4.25830.
22. Al-Alawi, A.; Van De Voort, F.R.; Sedman, J.; Ghetler, A.; Automated, FTIR analysis of free fatty acids or
moisture in edible oils. J. Assoc. Lab. Autom. 2006, 11, 23–29, doi: 10.1016/j.jala.2005.11.002.
23. Hayati, I.N.; Che Man, Y.B.; Tan, C.P.; Aini, I.N. Monitoring peroxide value in oxidized emulsions by
Fourier transform infrared spectroscopy. Eur. J. Lipid Sci.Technol. 2005, 107, 886–895, doi:
10.1002/ejlt.200500241.
24. Hendl, O.; Howell, J.A.; Lowery, J.; Jones, W. A rapid andsimple method for the determination of iodine
values usingderivative Fourier transform infrared measurements. Anal. Chim. Acta 2001, 427, 75–81, doi:
10.1016/S0003-2670(00)01193-4.
25. Dubois, J.; Van De Voort, F.R.; Sedman, J.; Ismail, A.A.; Ramaswamy, H.R. Quantitative Fourier transform
infrared analysis for anisidine value and aldehydes in thermally stressed oils. J. Am. Oil Chem. Soc. 1996,
73, 787–794.
26. Sherazi, S.T.H.; Talpur, M.Y.; Mahesar, S.A.; Kandhro, A.A.; Arain, S. Main fatty acid classes in vegetable
oils bySB-ATR-Fourier transform infrared (FTIR) spectroscopy. Talanta 2009, 80, 600–606, doi:
10.1016/j.talanta.2009.07.030.
27. Maggio, R.M.; Kaufman, T.S.; Del Carlo, M.; Cerretani, L.; Bendini, A.; Cichelli, A.; Compagnone, D.
Monitoring of fatty acid composition in virgin oliveoil by Fourier transformed infrared spectroscopy
coupled with partial least squares. Food Chem. 2009, 114, 1549–1554, doi: 10.1016/j.foodchem.2008.11.029.
28. Alonso-Salces, R.M.; Herrero, C.; Barranco, A.; López-Márquez, D.M.; Berrueta, L.A.; Gallo, B.; Vicente, F.
Polyphenolic compositions of Basque natural ciders: A chemometric study. Food Chem. 2006, 97, 438–446,
doi: 10.1016/j.foodchem.2005.05.022.
29. Woodcock, T.; Downey, G.; Kelly, J.D.; O’Donnell, C. Geographical classification of honey samples by
near-infrared spectroscopy: A feasibility study. J. Agric. Food Chem. 2007, 55, 9128–9134, doi:
10.1021/jf072010q.
Sustainability 2019, 11, 6414 16 of 16
30. Çam, M.; Hışıl, Y.; Durmaz, G. Classification of eight pomegranate juices based on antioxidant capacity
measured by four methods. Food Chem. 2009, 112, 721–726, doi: 10.1016/j.foodchem.2008.06.009.
31. Wang, T.; Jonsdottir, R.; Ólafsdóttir, G. Total phenolic compounds, radical scavenging and metal chelation
of extracts from Icelandic seaweeds. Food Chem. 2009, 116, 240–248, doi: 10.1016/j.foodchem.2009.02.041.
32. Silverstein, R.M.; Webster, F.X.; Kiemle, D.J. Spectrophotometric Identification of Organic Compounds, 7th ed.;
Wiley: New York, NY, USA, 2005.
33. Yang, H.; Irudayaraj, J. Comparison of near-infrared, Fourier transform-infrared, and Fourier
transform-Raman methods for determining olive pomace oil adulteration in extra virgin olive oil. J. Amer.
Oil Chem. Soc. 2001, 78, 889, doi:10.1007/s11746-001-0360-6.
34. Vlachos, N.; Skopelitis, Y.; Psaroudaki, M.; Konstantinidou, V.; Chatzilazarou, A.; Tegou, E. Applications
of Fourier transform-infrared spectroscopy to edible oils. Anal. Chim. Acta 2006, 573, 459–465,
doi:10.1016/j.aca.2006.05.034.
35. Dupuy, N.; Duponchel, L.; Huvenne, J.P.; Sombret, B.; Legrand, P. Classification of edible fats and oils by
principal component analysis of Fourier transform infrared spectra. Food Chem. 1996, 57, 245–251,
doi:10.1016/0308-8146(95)00213-8.
36. Guillén, M.D.; Cabo, N. Infrared spectroscopy in the study of edible oils and fats. J. Sci. Food Agric. 1997, 75,
1–11, doi:10.1002/(SICI)1097-0010(199709)75:1<1::AID-JSFA842>3.0.CO;2-R.
37. Elzey, B.; Pollard, D.; Sayo, O.F. Determination of adulterated neem and flaxseed oil compositions by FTIR
spectroscopy and multivariate regression analysis. Food Control 2016, 68, 303–309,
doi:10.1016/j.foodcont.2016.04.008.
38. Rohman, A.; Che Man, Y.B. Analysis of cod-liver oil adulteration using Fourier transform infrared (FTIR)
spectroscopy. J. Am. Oil Chem. Soc. 2009, 86, 1149, doi:10.1007/s11746-009-1453-9.
39. Hirri, A.; Bassbasi, M.; Platikanov, S.; Tauler, R.; Oussama, A. FTIR spectroscopy and PLS-DA
classification and prediction of four commercial grade virgin olive oils from Morocco. Food Anal. Methods
2016, 9, 974–981, doi:10.1007/s12161-015-0255-y.
40. Gurdeniz, G.; Ozen, B. Detection of adulteration of extra-virgin olive oil by chemometric analysis of
mid-infrared spectral data. Food Chem. 2009, 116, 519–525, doi:10.1016/j.foodchem.2009.02.068.
41. Lai, Y.W.; Kemsley, E.K.; Wilson, R.H. Potential of Fourier transform infrared spectroscopy for the
authentication of vegetable oils. J. Agric. Food Chem. 1994, 42, 1154–1159, doi:10.1021/jf00041a020.
42. Bendini, A.; Cerretani, L.; Di Virgilio, F.; Belloni, P.; Bonoli-Carbognin, M.; Lercker, G. Preliminary
evaluation of the application of the FTIR spectroscopy to control the geographic origin and quality of
virgin olive oils. J. Food Qual. 2007, 30, 424–437, doi:10.1111/j.1745-4557.2007.00132.x.
43. De Luca, M.; Restuccia, D.; Clodoveo, M.L.; Puoci, F.; Ragno, G. Chemometric analysis for discrimination
of extra virgin olive oils from whole and stoned olive pastes. Food Chem. 2016, 202, 432–437,
doi:10.1016/j.foodchem.2016.02.018.
44. Luca, M.; Terouzi, W.; Ioele, G.; Kzaiber, F.; Oussama, A.; Oliverio, F.; Tauler, R.; Ragno, G. Derivative
FTIR spectroscopy for cluster analysis and classification of morocco olive oils. Food Chem. 2011, 124,
1113–1118, doi:10.1016/j.foodchem.2010.07.010.
45. Cebia, N.; Yilmaz, M.T.; Sagdic, O.; Yuce, H.; Yelboga, E. Prediction of peroxide value in omega-3 rich
microalgae oil by ATR-FTIR spectroscopy combined with chemometrics. Food Chem. 2017, 225, 188–196,
doi:10.1016/j.foodchem.2017.01.013.
46. Mahboubifar, M.; Hemmateenejad, B.; Javidnia, K.; Yousefinejad, S. Evaluation of long-heating kinetic
process of edible oils using ATR–FTIR and chemometrics tools. J. Food Sci. Technol. 2017, 54, 659–668,
doi:10.1007/s13197-017-2502-2.
47. Granato, D.; Santos, J.S.; Escher, G.B.; Ferreira, B.L.; Maggio, R.M. Use of principal component analysis
(PCA) and hierarchical cluster analysis (HCA) for multivariate association between bioactive compounds
and functional properties in foods: A critical perspective. Trends Food Sci. Technol. 2018, 72, 83–90, doi:
10.1016/j.tifs.2017.12.006.
© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).
