Analytica Chimica Acta 583 (2007) 138–146

Q-mode curve resolution of UV–vis spectra for structural

transformation studies of anthocyanins in acidic solutions
Paulo Henrique Março, Ieda Spacino Scarminio ∗
Laboratório de Quimiometria em Ciências Naturais, Departamento de Quı́mica, Universidade Estadual de Londrina,
Caixa Postal 6001, CEP 86051-970 Londrina, Paraná, Brazil
Received 13 May 2006; received in revised form 29 September 2006; accepted 29 September 2006
Available online 4 October 2006

Abstract
Chemometric analysis of ultraviolet–visible (UV–vis) spectra for pH values 1.0, 3.3, 5.3, and 6.9 was used to investigate the kinetics and the
structural transformations of anthocyanins in extracts of calyces of hibiscus flowers of the Hibiscus acetosella Welw. ex Finicius for the first time. Six
different species were detected: the quinoidal base (A), the flavylium cation (AH+ ), the pseudobase or carbinol pseudobase (B), cis-chalcone (CC ),
trans-chalcone (Ct ), and ionized cis-chalcone (CC − ). Four equilibrium constant values were calculated using relative concentrations, hydration,
pKh = 2.60 ± 0.01, tautomeric, KT = 0.14 ± 0.01, acid–base, pKa = 4.24 ± 0.04, and ionization of the cis-chalcone, pKCC = 8.74 ± 1.5 × 10−2 . The
calculated protonation rate of the tautomers is KH+ = 0.08 ± 7.6 × 10−3 . These constants are in excellent agreement with those measured previously
in salt form. From a kinetic viewpoint, the situation encountered is interesting since the reported investigation is limited to visible light absorption
in acid medium. These models have not been reported in the literature.
© 2006 Elsevier B.V. All rights reserved.

Keywords: Anthocyanins; Chemometric analysis; UV–vis spectrometry; Kinetic studies

1. Introduction Based on observation of a few relatively simple anthocyanins,

Anthocyanins have been the subjects of many studies due to
their importance as quality indicators in foods, important chemo-
taxonomic indicators for plants [1,2], as well as for their possible
antioxidant action [3].
It has long been known that anthocyanins bearing the same
chromophoric structure can give rise to different colors. Their
color is easily affected by a number of chemical and physical
factors, such as temperature, pH, solvent, the structure of the
pigment itself and the presence of other molecules, which can
be generally described as copigments [4,5]. A particular problem
is the pH influence on their color. Depending on their acidity or
alkalinity, anthocyanins adopt different chemical structures. In
acidic aqueous solutions [4], four species of the anthocyanin
molecule may exist in equilibrium: the quinoidal base (A), the
flavylium cation (AH+ ), the pseudobase or carbinol pseudobase
(B), and the chalcone (C), Fig. 1 [6].
∗ Corresponding author. Tel.: +55 43 3371 4366; fax: +55 43 3371 4366
E-mail address: ieda@qui.uel.br (I.S. Scarminio).
the generally accepted chemical equilibria characteristic of
structural transformation are shown in Scheme 1 [4,6,7].
According to Eqs. (1)–(3), the AH+ , B, and C species are
interconverted with their relative amounts depending on pH. At
a pH of approximately 3 or lower, anthocyanin exists mainly
as a flavilyium cation and the amount of quinoidal base is not
significant, thus the system can be described by Eqs. (1) and (2).
Kh
AH+ B+H+ hydration equilibrium (1)
KT
BC ring/chain tautomeric equilibrium (2)
Ka
AH+ A + H+ acid–base equilibrium (3)
As the pH is raised, kinetic and thermodynamic competi-
tion occurs between the hydration reaction of the flavylium
cation and the proton transfer related to acidic hydroxyl groups
of aglycone. While the first reaction gives a colorless carbinol
pseudobase, which can undergo ring opening to a chalcone pseu-
dobase, the latter reaction gives rise to a quinoidal base. The

0003-2670/$ – see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.aca.2006.09.057
 P.H. Março, I.S. Scarminio / Analytica Chimica Acta 583 (2007) 138–146 139

different equilibrium constants are expressed as follows [8]:

[B]
Kh = aH+ (4)
[AH+ ]
[C]
KT = (5)
[B]
[A]
Ka = aH+ (6)
[AH+ ]
where aH+ is 10−pH ; Ka , KT and Kh are the equilibrium constants
for, respectively, the acid–base, the hydration, and the ring chain
tautomerism equilibria. Ka and Kh have been previously found
to be 10−4 and 10−2 , respectively at 25 ◦ C [6].
Most of the studies about the measurement of the equilib-
rium constants associated with the structural transformations
of anthocyanins in aqueous solution have been conducted
by temperature-jump [9] or pH-jump [10] relaxation. The
Fig. 1. Molecular structures of flavylium cation (AH+ ), quinoidal base (A), corresponding relaxation amplitudes are recorded by means
pseudobase or carbinol (B) and chalcone (C). R is a glycosil and R are usually of ultraviolet–visible (UV–vis) absorption spectrophotometry
H or sugar. R1 and R2 are usually H, OH, or OCH3 . [11]. Soundheimer [12] was the first to measure the equi-

Scheme 1.
140 P.H. Março, I.S. Scarminio / Analytica Chimica Acta 583 (2007) 138–146
librium constant of reaction 1 in the case of pelargonidin
3-monoglucoside. Brouillard et al. [10] applied the pH-jump
method to measure the equilibrium constants of one natural
product (apigenidin) and three related 3-deoxyflavylium salts.
Little research has been made using information from the
full spectra for studying the structural transformation of
anthocyanins. The common approach has been single point
measurements at a wavelength where one component dominates
the spectral response and the contributions from the other com-
ponents are neglected. The major limitation in the identification
of pigments on the sole basis of the absorbance spectrum is
overlapping absorbance bands of individual pigments present in
the mixture. Furthermore, in some cases the molecules studied
are involved in complex equilibria such as tautomerism, pH
equilibria, dimerization, etc.
Determination of ionization constants of natural antho-
cyanins by spectrophotometric methods is complicated by the
fact that prior knowledge of the spectra of all the charged
and uncharged species is required [13]. For such systems, a
direct spectral analysis is difficult since the number of compo-
nents contributing to the spectrum and their spectral profiles are
usually unknown. Therefore, additional analysis is required to
obtain detailed information about the structural transformation
of anthocyanins. None of the spectroscopic methods is entirely
suitable for elucidation of the structures of pigments. If the rel-
ative concentrations of the species (structures) can be shifted by
changing factors that do not alter the individual spectral profiles,
chemometric methods can be used to estimate the number of
species present as well as extract the spectra of species involved
and their corresponding concentration profiles.
Curve resolution methods are a group of chemometrical
approaches suitable for the treatment of multicomponent data
matrices. Their main purpose is the correct determination of the
response profiles as spectral and concentration profiles of the
component present in unresolved mixtures [14]. All of these
methods assume Beer’s law holds, in which case the original
data matrix can be represented by the product of two smaller
matrices that contain the concentration profiles and the pure
spectra of each individual component present in the mixture [15].
Curve resolution can also be used to deconvolute spectral fea-
tures using models for equilibrium constant determination and
kinetic information [16–25]. A large number of curve resolu-
tion techniques have been published, e.g. orthogonal projection
(OPA) [26], Imbrie Q-mode factor analysis followed by vari-
max and Imbrie’s rotations [27,28], correlated spectral data by
Procrustes rotation [29], interative target transformation factor
analysis (ITTFA) [30], simple to use interactive self-modelling
mixture analysis (SIMPLISMA) [26], and heuristic evolving
latent projection (HELP) [31].
Recently we presented results of an investigation of antho-
cyanins from fresh petals of hibiscus flowers of the Hibiscus
rosa-sinensys L. var. regius maximus [32] and in the calyces
of flowers of the Hibiscus sabdariffa [33] species using the
PARAFAC method to resolve the absorption spectral profiles
as well as the kinetic concentration profiles, for complete
UV–vis spectra measured between 240 and 748 nm for pH val-
ues between 1.0 and 13.0. In this work, we extend these studies
to the kinetics of simultaneous structural transformations of
anthocyanins in extracts of fresh calyces of hibiscus flowers
of Hibiscus acetosella Welw. ex Finicius. Imbrie Q-mode fac-
tor analysis followed by varimax and Imbrie’s oblique rotations
and K matrix treatment were applied to the UV–vis spectral
data. Since individual pH curves were treated using this tech-
nique, higher curve resolution was obtained with more accurate
determinations of equilibrium and rate constants than with the
PARAFAC model. This method offers several advantages. The
chemometric methods used here permit one to deduce the rela-
tive concentrations from observed spectral changes, resolve the
absorption spectra of the species, as well as to obtain the equi-
librium constants. Also these studies, in contrast to other works
measuring these constants in which the pigments are generally
isolated and purified as flavylium cations, determined equilib-
rium and kinetic rate constants of anthocyanin systems without
isolation and purification of the individual components.
2. Theory
2.1. Principal component analysis (PCA)
Principal component analysis [34] is a multivariate technique
used for reducing the dimensionality of a data set. It is based
upon eigen-analysis of the correlation or covariance matrix.
Considerable amounts of literature exist on the PCA method;
therefore this section will provide only a summary of PCA. The
data matrix A consists of n objects (spectra) represented by p
variables (absorbencies at different wavelengths). PCA decom-
poses the matrix A into a product of two matrices,
A = Tn×q PTq×p + En×p , (7)
where T is a score matrix, P a loading matrix, and E a matrix of
residuals. The number of principal components describing the
major part of the data variance is represented by q, and PT is the
transpose of P.
2.2. Q-mode principal component analysis
Paatero and Tapper have compared theoretical aspects of
Q-mode and R-mode principal component analyses as general
purpose methods [35]. For equilibrium constant determination,
Q-mode analysis is especially convenient since it evaluates inter-
relationships between samples and concentration proportions
are determined directly [36]. R-mode analysis is primarily used
to evaluate correlations between variables although its scores
can be used to determine concentrations. The original matrix
can be transformed to describe an index of proportional similar-
ity. Imbrie Q-mode factor analysis defines the similarity between
samples based on the proportions of their constituents. The con-
tribution of each object is obtained by an eigen-analysis of a
real symmetric matrix obtained from the data matrix. Imbrie
and Purdy [37] define an index of proportional similarity, called
cos θ. For two objects, k and , cos θ is calculated by
p
j=1 akj aj
cos ␪k =  p 2 (8)
p
j=1 akj j=1 aj
2
 P.H. Março, I.S. Scarminio / Analytica Chimica Acta 583 (2007) 138–146
For positive a elements, this index varies from zero, no sim-
ilarity, to one, identity. This normalization does not affect the
proportionality relationship between variables but removes the
effects of size differences among samples. The cos θ are calcu-
lated for all pairs of objects in the data set and then arranged in an
n × n matrix of associations, H. The loadings are used to describe
the relative proportions of the constituents in the sample.
According to [36], the mathematical procedure starts by data
normalization. The raw data matrix A is normalized by
W = D−1/2 A
where D = diag(AAT ). The association matrix is defined as the
major product moment of the row normalized data matrix,
H = WW . T
The row normalized data matrix can be expressed as the prod-
uct of a loading matrix and a score matrix,
W ≈ TPT .
Here T is an n × q matrix and P is p × q, where q is the
approximate rank of W. The relationship between W, H, P and
T is given by
H = WWT = TPT PTT .
The constraint that the PT P = I, leads to the following equa-
tion from Eq. (12),
H = TTT .
The symmetric matrix H can be factored according to
H = UΛ UT ,
where U is the matrix of eigenvectors and  the diagonal matrix
of associated eigenvalues. Thus,
T = Uq 1/2
q .
The matrix of loadings is then the matrix of eigenvectors,
scaled by the squares roots of the eigenvalues.
The matrix of scores may be defined by
W ≈ TPT
Premultiplication by TT gives
TT W ≈ TT TPT ,
but TT T = Λ so that PT = Λ−1 TT W, or
P = WT TΛ−1 .
It must be emphasized that the Tn×q and PTq×p matrices so
derived are not the same as those derived in R-mode analysis.
Once obtained, T may be rotated to B by varimax rotation. Each
row of B corresponds to an object or a linear combination of
objects and each column represents a factor.
Imbrie’s oblique rotation [36] is used to rotate the factor axes
of the B matrix so that they coincide with the most divergent
samples and then express all the other samples as proportions of
141
these end member samples. This is accomplished by construct-
ing a Vq×q matrix that contains the highest absolute values of the
varimax weights in each object column. The oblique projection
matrix is given by
C = BV−1 (19)
where V−1 is the inverse of V. The C matrix row vectors furnish
the proportional contributions of all the objects in the terms of
the end member samples.
(9)
2.3. K matrix method
The method (sometimes referred to as classical least squares,
CLS) is based on the well-known Beer–Lambert law [38]. If a
(10)
mixture set of known composition is used as a calibration set,
a series of simultaneous equations may be utilized to describe
mixture spectra. The corresponding matrix equation is
(11) ATp×n = Kp×q Cq×n + Ep×n (20)
where A contains the sample spectra C, the concentrations of the
q species in the N samples (n > q) and E is the error matrix. The K
matrix is the spectra of the pure species for unitary concentration
and path length. In this work K was estimated from C, which
(12) was obtained by Imbrie’s oblique rotation, Eq. (19).
3. Experimental
(13) Plants were cultivated in the garden of the Chemistry Depart-
ment of the Universidade Estadual de Londrina (UEL), in Lon-
drina, PR, Brazil. The plants were identified by Dr. Manuel
(14) RC Paiva, Departamento de Biologia, Universidade Estadual de
Londrina. A voucher was deposited in the herbarium of the Uni-
versidade Estadual de Londrina. The access numbers is FUEL
33816.
All reagents were of analytical grade. Buffer solutions with
(15)
four different pH (1.0; 3.3; 5.3, and 6.9) values were prepared
with phosphate following the detailed procedure described by
Levi et al. [32]. The pH of each solution was measured with a
HANNA HI model HI9321 pH Meter.
Pigments were extracted by maceration of fresh calyces
(16) (11.80 g) of the hibiscus flowers with 0.1% HCl in ethanol.
The ethanolic extracts were passed through filter paper. Twenty
milliliter of buffer solutions were added to the aliquots (3.00 mL)
(17) of the ethanolic extracts of anthocyanin, and immediately the
spectra were recorded every 20 s during a 30 min interval.
The analyses were carried out using an Ocean Optics model
(18) CHEM2000 spectrophotometer with a quartz cuvette (d = 1 cm).
The temperature was fixed at 25 (±0.1) ◦ C. Ultraviolet–visible
absorption spectra at different pH values were recorded from 240
to 748 nm. The 240–245 nm region was very noisy and elimi-
nated from subsequent data treatment.
For the chemometric study, each pH set (time versus wave-
length) was characterized by 1450 absorbance values obtained
in an interval from 245 to 748 nm. Chemometric methods were
applied to the data matrix, composed of 90 rows (time) and 1450
columns (wavelength), resulting in a 90 × 1450 matrix. Since
142 P.H. Março, I.S. Scarminio / Analytica Chimica Acta 583 (2007) 138–146
Q-mode analysis is performed, a 90 × 90 matrix is diagonalized
instead of a 1450 × 1450 matrix necessary for R-mode analysis.
3.1. Computer programs
The FORTRAN programs used for the K matrix and Q-mode
principal component analysis, varimax and Imbrie’s oblique
rotations calculations were developed in our laboratory. These
programs can be compiled on any digital computer and are avail-
able from the corresponding author.
4. Results and discussion
Representative spectra of the ethanolic extract of anthocyanin
at pH 1.0 are given in Fig. 2a. Absorption in the 522 nm region
is attributed to the presence of flavylium cation. The absorptions
at 281 and 330 nm correspond to a carbinol pseudobase (B) and
chalcone (C), respectively. After the addition of buffer solution,
increases in absorbencies at the 281, 330 and 522 nm wave-
lengths occur until attaining equilibrium at this pH value, Fig. 2a.
The absorbance increase can be due to a decrease in water activ-
ity or to the copigmentation reaction. This result is in agreement
with the literature and suggests that the copigment effect par-
allels the hydration reaction [39]. Copigmentation reduces the
extent of the hydratation reaction and therefore increases the
stability of the colored species [40]. The copigmentation reac-
tion consists of an increase in the visible band intensity for the
pigment (hyperchromic effect) and frequently in a shift of this
Fig. 2. Spectra vs. time matrices corresponding to the following pH values: (a) 1.0; (b) 3.3; (c) 5.3, and (d) 6.9.
very band towards longer wavelengths (bathocromic effect). It
is affected by several factors of which pH is an important one. It
has been demonstrated that the copigment effect occurs from pH
values close to 1 to neutrality. At a pH of 1, no statistical method
was applied considering that there was no change in λmax but
only in the intensity of the absorbance.
When the ethanolic extract of anthocyanin is added to the
buffer at pH 3.3, the red solution fades and a decrease in inten-
sity of absorption at 522 nm occurs. Spectral modifications occur
which predominantly reflect the partial transformation of the
flavilium cation into carbinol pseudobase (B) and chalcone (C)
during the period required to establish equilibrium, shown in
Fig. 2b. In comparison with Fig. 2a the spectral profiles appear
similar but not identical. A shoulder found at 422 nm is charac-
teristic of trans-chalcone. Since there are at least three species,
it is not possible to calculate the true or relative concentrations
from the absorption data by conventional chemical analysis. In
order to investigate the effect of pH on the structural transforma-
tion more fully and thereby gain an insight into the species that
may be present in solution, as well as to calculate their relative
concentrations, Imbrie Q-mode factor analysis followed by vari-
max and Imbrie’s oblique rotations and K matrix calculations
were carried out.
Imbrie’s Q-mode factor analysis showed that two factors
explain 99% of the total spectral data variance. Since the fac-
tors are linearly independent, it was concluded that three main
species are present in the structural transformation of the investi-
gated system: the flavylium cation, the carbinol pseudobase, and
 P.H. Março, I.S. Scarminio / Analytica Chimica Acta 583 (2007) 138–146
Fig. 3. Concentration profiles obtained by the model at pH 3.3 based on the
deconvolution of UV–vis absorption spectra. AH+ is the flavylium cation, B is
the pseudobase or carbinol pseudobase, and C is the chalcone.
chalcone. At this pH it was not possible to obtain the relative
concentration of the trans-chalcone, because it only appears in
negligible quantities. Besides that, the absorption spectrum of
the trans-chalcone isomer is practically identical to that of the
cis-chalcone isomer due to band overlap of the latter with the
carbinol pseudobase form [41]. Therefore, three factors were
subjected to varimax rotation and oblique rotation to obtain
the final species proportions. Since this a critical step in the
analysis [42], alternative calculations were carried out with addi-
tional principal components. However, increases in the number
of rotated principal components resulted in increased noise lev-
Fig. 4. Time-dependence of the absorbance at (a) 281 nm; (b) 330 nm; (c) 420 nm, and (d) 522 nm.
143
els for the factor loadings. The relative concentrations for the
three species were calculated by Eq. (19), and their profiles are
shown in Fig. 3. Relative concentrations after equilibrium have
been established allowing determination of the apparent acid-
ity constants Kc = ([C]/[AH+ ])aH+ = Kh KT , for the chalcone
formation reaction starting from flavylium cation AH+ , as well
as for the hydratation and tautomeric reactions associated with
the structural transformations of anthocyanins.
The experimental values determined for pKh , KT , and pKc
obtained in this study were 2.60 ± 0.01, 0.14 ± 0.01, and
3.54 ± 0.02, respectively. These values are in good agreement
with our results obtained by the PARAFAC method for Hibiscus
sabdariffa species [33], pKh = 2.70, KT = 0.14, and pKc = 3.54,
and the literature values of 2.60 ± 0.02, 0.12, and 3.54 ± 0.04
obtained by the pH-jump method for pure malvidin-3-glucoside
chloride [10]. Kinetic analysis was performed by utilizing plots
of the dependence of the variation of absorption against wave-
length monitored at 281 nm (carbinol pseudobase), 330 nm
(cis-chalcone), 422 nm (trans-chalcone) and 522 nm (flavylium
cation). The variation of the absorption with time is shown in
Figs. 4a–d. Generally, anthocyanin degradation is explained by
one first-order rate constant. Here, this model is not sufficient
to describe and explain the whole transformation process. At
281, 330, and 422 nm the models for the kinetic behavior were
fitted to a sum of exponentials, Eq. (21), with the goodness of fit
determined by chi-square. The results for the kinetic parameters
and their errors are shown in Table 1.
At − A0 = α1 (1 − e−k1 t ) + α2 (1 − e−k2 t ) (21)
144 P.H. Março, I.S. Scarminio / Analytica Chimica Acta 583 (2007) 138–146

Table 1
Kinetic parameters and their errors for the structural transformation of anthocyanins at pH 3.3
281 nm 330 nm 422 nm 522 nm

χ2a 1.29 × 10−6 1.29 × 10−6 1.34 × 10−6 2.24 × 10−6

A0 359(± 1.69) × 10−3 327(± 1.44)× 10−3 8.72(± 1.19) × 10−3 1171(± 3.10) × 10−4
␣1 48.4(± 1.69) × 10−3 26.9(± 2.28) × 10−3 24.1(± 1.64) × 10−3 230(± 6.80) × 10−4
␣2 26.9(± 2.28) × 10−3 48.4(± 1.69) × 10−3 23.0(± 3.44) × 10−3 –
k1 80(± 7.6) × 10−3 57(± 8.5) × 10−2 5.8(± 2.0) × 10−2 14.9(± 1.1) × 10−2
k2 57(± 8.5) × 10−2 80(± 7.6) × 10−3 40(± 0.8) × 10−1 –
a χ2 —Chi-square distribution with (n − 1) degrees of freedom.

This equation is characteristic of consecutive first-order kinetic parameters at this pH, as well as their errors are pre-
reactions. Since the data in Table 1 confirms the relations sented in Table 2. As observed at pH 3.3, the kinetic behavior
k1␭=281 nm /k2␭=330 nm ∼ 1, and KT « 1, the mechanism of tau- fitted to a sum of exponentials, Eq. (21), is not seen at 330 nm
tomeric interconversion between the B and C species proceeds and 422 nm but only at 281 nm. Inspection of Table 2 clearly
through the formation of an intermediate by acid catalysis by shows that the tautomeric interconversion between B and C can-
the proton [43]. The calculated protonation rate of the tautomers not be seen as acid catalysis by the proton at this pH. On the
(KH+ ) is 0.08 ± 7.6 × 10−3 . When the jump relaxation method other hand, the species monitored at 422 nm showed the pres-
←
is used, the amplitude associated with the B→C equilibrium ence of two separate linear behaviors, i.e. the presence of two
can be affected by the presence of other species [44]. However, pH-independent reactions, with different rates. The first can be
our results, using chemometric methods, show that it is pos- observed between 0 and 10 min and the second after 10 min. The
sible to obtain the KT values without interference from other kinetic behaviors for 330 and 522 nm are described by Eq. (23).
species. Although the above values cannot be compared directly This equation shows that two separate kinetic processes occur at
with results in the literature, we obtained the same behaviour this wavelength. The first obeys first-order kinetics and the sec-
for anthocyanin extracted from fresh petals of hibiscus flowers ond indicates that the kinetics does not depend on the species
of the Hibiscus rosa-sinensys L. var. regius maximus, using the concentrations anymore.
PARAFAC model [32].
The behaviour at 522 nm, Fig. 4d, obeys first-order kinetics, At − A0 = α1 e−k1 t + α2 t (23)
Eq. (22). At this pH, a significant copigmentation effect also
exists, with an increase in intensity, Fig. 4d. In the case of pH 6.9, the spectra obtained, Fig. 2d, shows
a bathochromic effect at 600 nm due to the formation of the
At − A0 = α1 e−k1 t (22) quinoidal base. It is possible to observe that at 600 nm, very
small absorbance differences from the initial values occur until
At pH 5.3, Fig. 2c, the analysis of the spectral changes showed equilibrium has been established. On the other hand, an increase
that the disappearance of AH+ causes a bathochromic shift of in absorption at 240–350 nm was observed.
522–560 nm due to the formation of the quinoidal base. Chemo-
metric analysis shows that three factors explain 100% of the total
spectral data variance; therefore, four factors were used in the
rotations. The relative concentrations for the four species were
calculated by Eq. (19), and their profiles are shown in Fig. 5.
The results based on the calculated spectral profiles using K
matrix methods, suggests that the pseudo-equilibrated mixture
of products is constituted essentially by the cis-chalcone (CC ),
the carbinol pseudobase (B) and small amounts of the quinoidal
base (A). At this pH value, the transformation is very fast and
the trans-chalcone (Ct ) concentration cannot be calculated.
Relative concentrations after equilibrium has been estab-
lished allowing determination of the values for pKh and KT ,
2.61 ± 0.04 and 0.12 ± 5.6 × 10−4 , respectively, with good
agreement with the previous results at pH 3.3. The quinoidal base
formation allowed measuring Ka , by Eq. (6). The estimated value
for pKa obtained in this study was 4.24 ± 0.04. This value agrees
with that measured by the pH-jump [10] method of 4.25 ± 0.1.
Fig. 5. Concentration profiles obtained by the model at pH 5.3 based on the
The relative concentrations of the flavylium cation (AH+ ), deconvolution of UV–vis absorption spectra. AH+ is the flavylium cation, B2
carbinol pseudobase (B2 ), cis-chalcone (Cc ) and quinoidal base is the pseudobase or carbinol pseudobase, CC is the cis-chalcone, and A is the
(A) at a pH of 5.3 are shown in Fig. 5. The results for the quinoidal base.
 P.H. Março, I.S. Scarminio / Analytica Chimica Acta 583 (2007) 138–146 145

Table 2
Kinetic parameters and their errors for the structural transformation of anthocyanins at pH 5.3
281 nm 330 nm 422 nm 522 nm

χ2a 7.99 × 10−6 1.56 × 10−6 – 1.12 × 10−6

A0 315(± 3.51) × 10−3 48.4(± 1.69) × 10−3 – 575(± 3.1) × 10−4
␣1 51.7(± 8.19) × 10−3 − 40.6(± 1.12) × 10−3 – 52.8(± 1.63) × 10−3
␣2 54.4(± 7.81) × 10−3 92(± 5.0) × 10−5 – 37(± 2.0) × 10−5
k1 1.85 ± 0.423 2160(± 9.0) × 10−4 1.34 × 10−2 0.91 ± 0.048
k2 9.69 ± 1.99 – 6.49 –
a χ2 —Chi-square distribution with (n − 1) degrees of freedom.

Table 3
Kinetic parameters and their errors for the structural transformation of anthocyanins at pH 6.9
281 nm 330 nm 422 nm 522 nm

χ2a 7.42 × 10−6 1.81 × 10−6 1.44 × 10−6

A0 320.7(±4.82) × 10−3 331.96(±7.27) × 10−3 96.63(±2.54) × 10−3
␣1 −25.94(±4.18) × 10−3 −24.56(±6.67) × 10−3 4.09(±3.59) × 10−3
NO FIT
␣2 6.1(±1.70) × 10−3 6.7(±2.0) × 10−3 8.4(±4.09) × 10−3
k1 0.16 ± 0.046 0.097 ± 2.96 × 10−2 0.97 ± 1.5
k2 4.5 ± 1.7
a χ2 —Chi-square distribution with (n − 1) degrees of freedom.

The estimated kinetic parameters are shown in Table 3. From
this table it is obvious that slightly larger structural transforma-
tions occur compared to those observed at pH 5.3.
In order to obtain the relative concentrations, three factors
were used in the rotations. The relative concentration pro-
files are presented in Fig. 6. Here a decrease in cis-chalcone
concentration and the formation of ionized cis-chalcone and
trans-chalcone can be observed. Evidence for this asser-
tion comes from the estimated acidity constant obtained in
Fig. 6. Concentration profiles obtained by the model at pH 6.9 based on the
deconvolution of UV–Vis absorption spectra. CC is the cis-chalcone, Ct is the
trans-chalcone, and CC − is the ionized cis-chalcone.
this study, [C− −2
C ][H ]/[CC ] = 8.74 ± 1.5 × 10 , which closely
+
agrees with the literature value of 8.93 [45].
5. Conclusion
Imbrie Q-mode factor analysis followed by varimax and
Imbrie’s oblique rotations and the K matrix method were used
to investigate the kinetics and the structural transformations of
anthocyanins. It was possible to identify the species involved at
equilibrium and obtain their relative concentrations as well as
the equilibrium constants.
Six different species were detected: the quinoidal base (A),
the flavylium cation (AH+ ), the pseudobase or carbinol pseu-
dobase (B), the cis-chalcone (CC ), trans-chalcone (Ct ), and
ionized cis-chalcone (CC − ). Four equilibrium constant values
were calculated by using the relative concentrations: for hydra-
tion, pKh = 2.60 ± 0.01, for tautomerism, KT = 0.14 ± 0.01, for
acid–base equilibrium, pKa = 4.24 ± 0.04, and for the ioniza-
tion of the cis-chalcone, pKCC = 8.74 ± 1.5 × 10−2 . These
values are in good agreement with literature values. When the
jump relaxation method is used, the amplitude associated with
←
the B→C equilibrium can be affected by the presence of the
other species [43]. However, our results using chemometric tech-
niques show that it is possible to obtain the pK values without
interference from other species. From a kinetic viewpoint, the
situation encountered is interesting since the reported investi-
gations are limited to visible light absorption. At pH 3.3, the
mechanism of tautomeric interconversion between B and C
species proceeds through the formation of an intermediate by
acid catalysis by the proton. The calculated protonation rate of
146 P.H. Março, I.S. Scarminio / Analytica Chimica Acta 583 (2007) 138–146
the tautomers encountered was KH+ = 0.08 ± 7.6 × 10−3 . This
model has not been reported in the literature.
Acknowledgement
P.H. Março and I.S. Scarminio thank Fundação Araucária for
financial support of this project.
References
[1] H.S. Lee, V. Honh, J. Chromatogr. 624 (1992) 221.
[2] J.P. Goiffon, P.P. Mouly, E.M. Gaydou, Anal. Chim. Acta 382 (1999) 39.
[3] Pi-Jen Tsai, J. McIntosh, P. Pearce, B. Camden, B.R. Jordan, Food Res.
Int. 35 (2002) 351.
[4] C.E. Lewis, J.R.L. Walker, J.E. Lancaster, Food Chem. 54 (1995) 315.
[5] P. Markakis, Anthocyanins as Food Colors, Academic Press, New York,
1982, p. 17.
[6] R. Brouillard, J.E. Dubois, J. Am. Chem. Soc. 99 (1997) 1359.
[7] P. Figueiredo, M. Elhabiri, K. Toki, N. Saito, O. Dangles, R. Brouillard,
Phytochemistry 41 (1996) 301.
[8] O. Dangles, R. Brouillard, Can. J. Chem. Soc., Perkin Trans. 2 (1992) 247.
[9] O. Dangles, R. Brouillard, Can. J. Chem. Soc. 70 (1992) 2174.
[10] R. Brouillard, G.A. Iacobucci, J.G. Sweeny, J. Am. Chem. Soc. 104 (1982)
7585.
[11] R. Brouillard, B. Delaport, J.E. Dubois, J. Am. Chem. Soc. 100 (1978)
6202.
[12] E. Soundheimer, J. Am. Chem. Soc. 75 (1953) 1507.
[13] R.E. Asenstorfer, P.G. Iland, M.E. Tate, G.P. Jones, Anal. Biochem. 318
(2003) 291.
[14] A.G. Frenich, M.M. Galera, J.L. Martı́nez Vidal, D.L. Massart, J.R. Torres-
Lapasió, K. De Braekeleer, J.-H. Wang, P.K. Hopke, Anal. Chim. Acta 411
(2000) 145.
[15] S. Bijlsma, D.J. Louwerse, A.K. Smilde, J. Chemometr. 13 (1999) 311.
[16] H. Gampp, M. Maeder, C.J. Meyer, A.D. Zuberbühler, Talanta 32 (1985)
1133.
[17] H. Gampp, M. Maeder, C.J. Meyer, A.D. Zuberbühler, Talanta 33 (1986)
943.
[18] A. de Juan, M. Maeder, M. Martinez, R. Tauler, Chemometr. Intell. Lab.
Syst. 54 (2000) 123.
[19] J. Saurina, S. Hernandez-Cassou, R. Tauler, A. Izquierdo-Ridorsa, J.
Chemometr. 12 (1998) 183.
[20] S. Bijlsma, A. Smilde, K. Smilde, Anal. Chim. Acta 396 (1999) 231.
[21] P.J. Gemperline, Anal. Chem. 71 (1999) 5398.
[22] S. Bijlsma, H.F.M. Boelens, H.C.J. Hoefsloot, A.K. Smilde, Anal. Chim.
Acta 419 (2000) 197.
[23] S. Bijlsma, H.F.M. Boelens, A.K. Smilde, Appl. Spectr. 55 (2001) 77.
[24] Z.L. Zhu, W.Z. Cheng, Y. Zhao, Chemometr. Intell. Lab. Syst. 64 (2002)
157.
[25] C. Chapados, M. Trudel, Biophys. Chem. 47 (1993) 267.
[26] F.C. Sánchez, B. van den Bogaert, S.C. Rutan, D.L. Massart, Chemometr.
Intell. Lab. Syst. 34 (1996) 139.
[27] M.M. Sena, I.S. Scarminio, K.E. Collins, C.H. Collins, Talanta 53 (2000)
453.
[28] I.S. Scarminio, W.J. Barreto, D.N. Ishikawa, E.L. Paczkowski, I.C. Arruda,
Anal. Sci. 16 (2000) 1337.
[29] I.S. Scarminio, M. Kubista, Anal. Chem. 65 (1993) 409.
[30] B.G.M. Vandeginste, W. Derks, G. Kateman, Anal. Chim. Acta 173 (1985)
253.
[31] O.M. Kvalheim, Y.-Z. Liang, Anal. Chem. 64 (1992) 936.
[32] M.A.B. Levi, I.S. Scarminio, R.J. Poppi, M.G. Trevisan, Talanta 62 (2004)
299.
[33] P.H. Março, M.A.B. Levi, I.S. Scarminio, R.J. Poppi, M.G. Trevisan, Anal.
Sci. 21 (2005) 1523.
[34] K.R. Beebe, R.J. Pell, M.B. Seasholtz, Chemometrics: A Practical Guide,
John Wiley & Sons, New York, 1998, pp. 1–180.
[35] P. Paatero, U. Tapper, Chemometr. Intell. Lab. Syst. 18 (1993) 183.
[36] R. Reyment, K.G. Jöreskog, Applied Factor Analysis in Natural Science,
Cambridge University Press, New York, 1996, pp. 89–119.
[37] J. Imbrie, E. Purdy, Mem. Am. Assoc. Petrol. Geol. 1 (1962) 253.
[38] D.M. Haaland, G. Easterling, D.A. Vopicka, Appl. Spectr. 39 (1985) 73.
[39] G. Mazza, R. Broulliard, Food Chem. 25 (1987) 207.
[40] R. Broulliard, Phytochemistry 22 (1983) 1311.
[41] H. Santos, D.L. Turner, J.C. Lima, P. Figueiredo, F.S. Pina, A. Maçanita,
Phytochemistry 33 (1993) 1227.
[42] M. Amrhein, B. Srinivasan, D. Bonvin, M.M. Schumacher, Chemometr.
Intell. Lab. Syst. 33 (1996) 17.
[43] O. Bensaude, M. Dreyfus, G. Dodin, J.E. Dubois, J. Am. Chem. Soc. 99
(1987) 4438.
[44] R. Brouillard, J. Lang, Can. J. Chem. 68 (1990) 755.
[45] R.A. McClelland, S. Gedge, J. Am. Chem. Soc. 102 (1980) 5838.