Chemical Physics Letters 410 (2005) 182–187

www.elsevier.com/locate/cplett

A DFT investigation of anthocyanidins

Jeffrey N. Woodford *

Department of Chemistry and Biochemistry, Eastern Oregon University, One University Blvd.,
303J Badgley Hall, La Grande, OR 97850-2899, United States

Received 19 November 2004; in final form 12 May 2005

Available online 14 June 2005

Abstract

A theoretical analysis of anthocyanidins, a class of natural plant pigments, has been conducted using density functional theory at
the B3LYP/6-31G(d) level. It is found that these compounds are non-planar, with the 2-phenyl ring twisted relative to the benzo-
pyrylium rings. TD-DFT calculations reveal a first excited state dominated by two orbital transitions, including the HOMO–LUMO
transition. From a comparison of these molecular orbitals with those of related compounds, and from the known chemistry of these
compounds, it is suggested that anthocyanidins should be regarded as natural, stable carbocations.
Ó 2005 Elsevier B.V. All rights reserved.

1. Introduction and the colorless quinonoidal base at mildly acidic to

neutral pHs [4–6]. Other factors, such as copigmenta-
Anthocyanidins are a class of natural, polyphenol- tion, serve to modulate the color of anthocyanidins in
based colored compounds that are largely responsible their natural settings.
for the vibrant colors in most fruits and flowers. They Surprisingly, this biologically important class of com-
are commonly found in red wine and fruit juices, and pounds has been the subject of only a small number of the-
have been shown to have properties beneficial to human oretical studies. Furthermore, those that have been
health, including antioxidant [1] and antitumor [2] activ- conducted have been almost exclusively at the semiempir-
ity. Synthetic anthocyanidins have also been shown to ical [7–10] or non-correlated Hartree-Fock (HF) [11]
exhibit photochromic behavior [3]. Hence, they may also levels. One study [12] performed post-Hartree-Fock
be used in advanced devices, serving as a basis for calculations, employing both MP2 and DFT, but was lim-
organic materials in optical data storage. ited only to flavylium and its monohydroxylated deriva-
The structures of some naturally occurring anthocy- tives. In this study, we present results obtained from a
anidins studied in the present work are given in Fig. 1 theoretical investigation of commonly occurring anthocy-
along with that of flavylium, the parent compound. Also anidin compounds, as well as flavylium. Density functional
shown in the figure is the labeling scheme for the three theory has been chosen, since it is capable of recovering a
rings (A, B, C) and the numbering scheme for the atoms. substantial amount of the correlation energy for common
All the anthocyanidins discovered to date have largely organic molecules. In addition, time-dependent density
the same chemical structure, yet they are responsible functional theory (TD-DFT) is applied to obtain, as a first
for wide variations in color. This is due in part to the approximation, the properties of the first excited states of
anthocyanidinsÕ equilibrium in aqueous solution be- these molecules. From the DFT results, it is predicted that
tween the pigmented benzopyrylium form at low pH anthocyanidins should be regarded as non-planar, thereby
lacking extended conjugation throughout the molecular
*
Fax: +1 541 962 3873. system; furthermore, anthocyanidins are best represented
E-mail address: jeff[email protected]. as carbocationic species.

0009-2614/$ - see front matter Ó 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.cplett.2005.05.067
 J.N. Woodford / Chemical Physics Letters 410 (2005) 182–187 183

imum on the potential energy surface by a normal mode

analysis; no imaginary frequencies were obtained. This
analysis was performed numerically by displacing each
atom by ±0.01 a.u. along each Cartesian coordinate, cal-
culating the analytic gradient for each displacement, and
then calculating the second derivative numerically. In
constructing geometries for the molecule, care was taken
to orient the C3–O3–H torsion angle, when present, in a
trans configuration. This configuration was shown to be
a lower energy configuration, by a few millihartrees,
than the anti-configuration in which the hydroxyl is ori-
ented towards the 2 0 -phenyl ring. To estimate the excita-
tion energy to the first excited state, the TD-DFT
method was employed, using the B3LYP/6-31G(d) equi-
librium geometry. Molecular orbital plots were made
with the MOLDEN software package [17].

3. Results and discussion

In Table 1 selected structural data are presented for

Fig. 1. Chemical structures of the anthocyanidin compounds studied the anthocyanidins studied in this work, along with rel-
in this work, along with their common abbreviations. evant crystal structure data [18,19]. The agreement be-
tween the calculated bond lengths and experimental
ones is satisfactory, indicating that this DFT method
2. Theoretical methodology is well suited for study of these compounds. It is interest-
ing to note that the longest bond in all of the com-
Geometry optimization and frequency calculations pounds is the C2 –C10 bond. This result implies
were performed with GAMESS [13], while TD-DFT calcu- substantial single-bond character, and thus a lack of res-
lations were performed with GAUSSIAN 03 [14]. The equi- onant conjugation between the benzopyrylium ring sys-
librium geometry was obtained by optimization with the tem and the 2-phenyl ring. This lack of conjugation is
B3LYP functional [15,16] and the 6-31G(d) basis set. also manifested in the O1 –C2 –C10 –C20 torsion angle,
The geometries thus obtained were verified to be a min- which for all of the molecules studied deviates several

Table 1
Selected structural data for anthocyanidins and comparison to experimental values
Propertya F Ap Apb Cy Cyc Dp Lt Mv Pg Pn Pt
d(O1–C2) 1.337 1.349 1.34 1.351 1.349 1.352 1.349 1.353 1.349 1.353 1.352
d(C2–C3) 1.409 1.404 1.39 1.424 1.396 1.424 1.404 1.424 1.423 1.424 1.410
d(C3–C4) 1.378 1.381 1.37 1.384 1.380 1.383 1.383 1.385 1.385 1.383 1.393
d(C4–C4a) 1.418 1.406 1.39 1.407 1.382 1.407 1.408 1.407 1.406 1.407 1.399
d(C4a–C5) 1.416 1.427 1.42 1.427 1.432 1.426 1.428 1.429 1.427 1.429 1.434
d(C5–C6) 1.381 1.383 1.37 1.384 1.366 1.384 1.385 1.380 1.383 1.379 1.378
d(C6–C7) 1.413 1.414 1.41 1.415 1.413 1.415 1.413 1.415 1.416 1.415 1.416
d(C7–C8) 1.389 1.403 1.40 1.399 1.387 1.399 1.402 1.399 1.399 1.400 1.399
d(C8–C8a) 1.394 1.382 1.36 1.384 1.376 1.384 1.381 1.390 1.384 1.390 1.388
d(C8a–O1) 1.367 1.370 1.37 1.362 1.361 1.362 1.368 1.360 1.361 1.362 1.359
d(C4a–C8a) 1.415 1.419 1.41 1.413 1.408 1.413 1.419 1.410 1.414 1.411 1.417
dðC2 –C10 Þ 1.448 1.440 1.45 1.437 1.453 1.436 1.439 1.436 1.438 1.435 1.446
dðC10 –C20 Þ 1.415 1.419 1.40 1.425 1.409 1.422 1.419 1.425 1.425 1.428 1.417
dðC20 –C30 Þ 1.387 1.377 1.38 1.380 1.371 1.384 1.382 1.378 1.376 1.378 1.387
dðC30 –C40 Þ 1.400 1.408 1.39 1.425 1.400 1.416 1.424 1.423 1.410 1.426 1.411
dðC40 –C50 Þ 1.400 1.409 1.39 1.400 1.378 1.411 1.401 1.410 1.406 1.397 1.404
dðC50 –C60 Þ 1.388 1.384 1.38 1.385 1.383 1.381 1.385 1.391 1.382 1.384 1.392
dðC60 –C10 Þ 1.415 1.417 1.40 1.417 1.404 1.421 1.412 1.415 1.421 1.420 1.415
sðO1 –C2 –C10 –C20 Þ 7.52
4.98 4.14 2.19 5.9 3.75 2.21
7.91 1.82
10.70
28.14
a
Bond lengths in Angstroms; torsion angles in degrees.
b
From [18].
c
From [19].
184 J.N. Woodford / Chemical Physics Letters 410 (2005) 182–187

degrees from planarity. This result emerges even for

flavylium, for which we calculate a torsion angle of
7.6°. This lack of planarity even for flavylium is notable
because flavylium is unsubstituted at the C3 and C60
positions, where steric contact is closest between the
benzopyrylium system and the phenyl ring. Thus, in
flavylium, there is minimal steric interaction between
the ring substituents which would force the two ring sys-
tems to deviate from planarity.
Previous theoretical studies have offered contradic-
tory results on the question of whether anthocyanidins
are planar. Early semi-empirical studies predicted a flat
structure, but the HF study of reference [11] predicts
nonzero inter-ring torsion angles. However, because
HF consistently underestimates bond lengths, these re-
sults are called into question. A recent B3LYP/6-
31G(d) analysis of flavylium [12] yielded a torsion angle
of 179°, or a deviation of only one degree from planar-
ity; this is in contradiction to our current results. This
discrepancy may be due to differences in the integration
grid for the hybrid HF-B3LYP method, or even differ-
ences in the implementation of the B3LYP functional
itself. Nevertheless, our theoretical results are in agree- Fig. 2. Calculated HOMO
1 (a), HOMO (b), and LUMO (c) of
ment with the available crystal structure data, although flavylium.
this agreement may be fortuitous owing to the neglect of
crystal packing forces in the ab initio calculations.
It is also interesting to note that the C–C bond length tion, recent results have shown that in the aging pro-
alternation in the C-ring of the anthocyanidins is signif- cess of red wine, condensation reactions occur between
icant, with C–C bond lengths deviating by as much as anthocyanidins and other compounds present in wine
0.04 Å from one another. The shortest carbon–carbon [24–27]. The condensation occurs most readily at carbon
bond length in the C-ring is C3–C4, implying that this C4, perhaps because carbon C2 is sterically protected
bond has the largest double-bond character. This bond from reaction with these relatively large molecules, but
alternation trend is also manifested in the other two also involves reaction at C5. These separate reactions
rings, but is not as significant. This finding suggests that all share a common mechanism, recently elucidated
the canonical Lewis structure presented in Fig. 1 for the [28], with the initial step being nucleophilic attack at
benzopyrylium rings, in apparent analogy with naphtha- C4. In 4-methyl derivatives of flavylium salts, it is known
lene, may not accurately represent the entire ground that the 4-methyl group exhibits unusually high acidity
state electronic picture of these compounds. Indeed, this [29]; this too may be explained as an inductive effect if
same suggestion has already been made on the basis of the C4 carbon is viewed as an electropositive center.
results from a semiempirical study [9]; we validate this Finally, to explore the optical properties of these
suggestion with the present DFT results. interesting molecules, time-dependent density functional
To explore the ground-state reactivity of anthocyani- theory was applied; the TD-DFT results for the various
dins, their HOMOs were investigated. The HOMO for anthocyanidins are summarized in Table 2, along with a
flavylium is displayed in Fig. 2b; the HOMOs for the comparison with available experimental data for the
other anthocyanidins are qualitatively similar to flavy- electronic excitations [30]. Among the multitude of ap-
liumÕs. The HOMO is of p-type, with an excess of elec- proaches in common use to calculate excited-state prop-
tron density at all of the ring atoms except carbons C2, erties, TD-DFT was chosen in the present study due to
C4, C5, and C8. This implies that these four carbon its computational efficiency and relatively good agree-
atoms will be electropositive, and hence likely candi- ment with experimental results [31]. Qualitatively, agree-
dates for nucleophilic attack. This interpretation is con- ment between the calculated values and experiment is
sistent with the known chemistry of anthocyanidins. For reasonable, as the TD-DFT method is able to reproduce
example, it is well known that one pathway for decom- the correct relative ordering of the electronic transitions
position of anthocyanidins is nucleophilic attack by among the anthocyanidins (except the order for Mv and
water at carbon C2. This reaction leads to the eventual Dp is reversed). Quantitatively, however, agreement is
formation of a colorless chalcone [4–6]. In flavylium, only somewhat satisfactory, due to the limitations of
nucleophilic attack at C4 is also observed [23]. In addi- the TD-DFT method and comparison of calculated
 J.N. Woodford / Chemical Physics Letters 410 (2005) 182–187 185

Table 2
TD-DFT results for the first several electronic excited states of anthocyanidins
Compound DE (eV) f kcalc (nm) Composition kexpta (nm)
F 3.11 0.007 399 HOMO ! LUMO (0.102) n/a
(HOMO
1) ! LUMO (0.675)
3.30 0.626 375 HOMO ! LUMO (0.617) n/a
(HOMO
1) ! LUMO (0.101)
3.34 0.007 371 (HOMO
2) ! LUMO (0.672) n/a
Ap 2.88 0.445 431 HOMO ! LUMO (0.536) 476
(HOMO
1) ! LUMO (0.359)
3.21 0.328 386 HOMO ! LUMO (0.316)
(HOMO
1) ! LUMO (0.559)
3.63 0.004 341 (HOMO
2) ! LUMO (0.684)
Cy 2.53 0.470 489 HOMO ! LUMO (0.574) 535
(HOMO
2) ! LUMO (0.244)
2.76 0.024 449 (HOMO
1) ! LUMO (0.660)
(HOMO
3) ! LUMO (0.104)
3.17 0.299 391 HOMO ! LUMO (0.177)
(HOMO
2) ! LUMO (0.628)
Dp 2.53 0.181 490 HOMO ! LUMO (0.318) 546
(HOMO
1) ! LUMO (0.555)
2.55 0.446 486 HOMO ! LUMO (0.454)
(HOMO
1) ! LUMO (0.389)
2.89 0.098 428 (HOMO
2) ! LUMO (0.608)
(HOMO
3) ! LUMO (0.122)
Lt 2.65 0.424 468 HOMO ! LUMO (0.596) 493
(HOMO
2) ! LUMO (0.220)
2.99 0.049 415 (HOMO
1) ! LUMO (0.665)
(HOMO
3) ! LUMO (0.102)
3.29 0.296 377 HOMO ! LUMO (0.638)
(HOMO
2) ! LUMO (0.162)
Mv 2.42 0.111 513 HOMO ! LUMO (0.356) 542
(HOMO
1) ! LUMO (0.557)
2.60 0.533 477 HOMO ! LUMO (0.459)
(HOMO
1) ! LUMO (0.384)
2.94 0.045 422 (HOMO
2) ! LUMO (0.631)
Pg 2.64 0.417 470 HOMO ! LUMO (0.556) 520
(HOMO
1) ! LUMO (0.295)
3.13 0.367 397 HOMO ! LUMO (0.243)
(HOMO
1) ! LUMO (0.583)
3.46 0.012 359 (HOMO
2) ! LUMO (0.681)
Pn 2.60 0.560 476 HOMO ! LUMO (0.574) 532
(HOMO
2) ! LUMO (0.208)
2.88 0.001 430 (HOMO
1) ! LUMO (0.654)
(HOMO
3) ! LUMO (0.127)
3.15 0.210 393 HOMO ! LUMO (0.160)
(HOMO
2) ! LUMO (0.642)
Pt 2.47 0.394 502 HOMO ! LUMO (0.572) 543
(HOMO
1) ! LUMO (0.255)
2.61 0.142 475 HOMO ! LUMO (0.213)
(HOMO
1) ! LUMO (0.632)
2.91 0.003 426 (HOMO
2) ! LUMO (0.642)
(HOMO
3) ! LUMO (0.157)
a
In methanol/HCl solution; from [31].

gas-phase values to values obtained experimentally in first three electronic excitations predicted by the TD-
solution. While it is believed that the transition to the DFT method, except for flavylium, only two have
first excited state is responsible for the vivid coloration non-negligible oscillator strengths (f > 0.1). Further-
of anthocyanidins, the present results suggest that high- more, both of these are qualitatively similar for all com-
er excited states may also play a prominent role. Of the pounds studied, involving a pair of orbital transitions:
186 J.N. Woodford / Chemical Physics Letters 410 (2005) 182–187

from the HOMO to the LUMO, and from another near-

by occupied MO (either HOMO
1 or HOMO
2) to
the LUMO. As a representative example, the calculated
HOMO
1, HOMO and LUMO of flavylium are dis-
played in Fig. 2; for purposes of comparison, the same
orbitals for benzopyrylium (BP) and 2-phenylpyrylium
(2PP) are also presented in Figs. 3 and 4, respectively.
These two compounds may be considered possible Ôpar-
ent compoundsÕ, in an electronic sense, of flavylium,
with benzopyrylium isoelectronic with naphthalene
and 2-phenylpyrylium isoelectronic with biphenyl. For
flavylium and 2PP, the HOMO
1 is of p-type and is
clearly localized on the 2-phenyl ring; for BP, the
HOMO
1 is of p-type and is delocalized. The HOMO
for each molecule is also of p-type and, for all three, is
delocalized over the entire molecule. The HOMOs of
flavylium and 2PP bear a striking resemblance to one
another; while the HOMOs of flavylium and BP are sim-
ilar in the C-ring only. The LUMO for all three mole-
cules is of p*-type and is also delocalized. This time,
while the comparison of LUMOs between flavylium
and 2PP remains striking, the comparison of LUMOs
between flavylium and BP is substantially more satisfac-
tory as well.

Fig. 4. Calculated HOMO
1 (a), HOMO (b), and LUMO (c) of

benzopyrylium (BP).

The nearly identical nature of the frontier orbitals of

flavylium and 2PP suggest that flavylium, and its antho-
cyanidin derivatives, should be considered derivatives of
2PP, rather than derivatives of BP. Electronically, anth-
ocyanidins bear more resemblance to biphenyl than to
naphthalene. It is well known that biphenyl does not ex-
hibit resonant delocalization of its p electrons, and is
twisted in its ground state [32,33]; much the same may
be said of anthocyanidins, albeit the twist angle in
biphenyl is much larger. Consequently, the canonical
Lewis structure representation of anthocyanidins, pre-
sented in Fig. 1, should be abandoned in favor of a car-
bocation-based one shown in Fig. 5. In the structures
shown in this figure, the positive charge is located alter-
natively on C2 and C4. They are consistent with the
structural data presented above, since the non-zero in-
ter-ring torsion angles clearly demonstrate that anthocy-
anidins are not planar, and hence are unable to
experience resonant conjugation throughout the entire
Fig. 3. Calculated HOMO
1 (a), HOMO (b), and LUMO (c) of 2- three-ring system. Indeed, the potential energy barrier
phenylpyrylium (2PP). to phenyl ring rotation has been calculated to be
 J.N. Woodford / Chemical Physics Letters 410 (2005) 182–187 187

O O

Fig. 5. Resonance pair of Lewis structures for flavylium that more accurately correspond to the calculated results.

relatively small, from 4.4 kcal/mol [11] to 8.96 kcal/mol [3] See, e.g., M.C. Moncada, A.J. Parola, C. Loderio, F. Pina, M.
[12] in flavylium. This implies, then, that rotation be- Maestri, V. Balzani, Chem. Eur. J. 10 (2004) 1519, and references
cited therein.
tween the B-ring and C-ring is free. From the structural [4] R. Brouillard, J.-E. Dubois, J. Am. Chem. Soc. 99 (1977)
data, we conclude that the resonance structure with the 1359.
positive charge located at C2 is dominant over the one [5] R. Brouillard, B. Delaporte, J. Am. Chem. Soc. 99 (1977)
with the positive charge located at C4, since the C3–C4 8461.
bond is shorter than the C2–C3 bond. While similar con- [6] R. Brouillard, B. Delaporte, J.-E. Dubois, J. Am. Chem. Soc. 100
(1978) 6202.
clusions have been drawn from semiempirical and HF [7] W.E. Kurtin, P.-S. Song, Tetrahedron 24 (1968) 2255.
studies [9,11], this is the first time these conclusions have [8] P.-S. Song, T.A. Moore, M. Sun, Adv. Food Res. 3 (1972) 33.
been validated using density functional methods which [9] G.K. Pereira, P.M. Donate, S.E. Galembeck, J. Mol. Struct. 392
include electron correlation. (1997) 169.
[10] G.K. Pereira, S.E. Galembeck, Spectrochim. Acta A 54 (1998)
339.
[11] G.K. Pereira, P.M. Donate, S.E. Galembeck, J. Mol. Struct. 363
4. Conclusion (1996) 87.
[12] M. Meyer, Int. J. Quant. Chem. 76 (2000) 724.
DFT calculations on commonly occurring anthocy- [13] M.W. Schmidt, K.K. Baldridge, J.A. Boatz, S.T. Elbert, M.S.
anidins have predicted structural properties in reasonable Gordon, J.H. Jensen, S. Koseki, N. Matsunaga, K.A. Nguyen, S.
Su, T.L. Windus, M. Dupuis, J.A. Montgomery, J. Comput.
agreement with experiment; the TD-DFT method applied Chem. 14 (1993) 1347.
to the DFT equilibrium geometry resulted in a first ex- [14] M.J. Frisch et al., GAUSSIAN 03, Revision B.02, Gaussian Inc.,
cited state dominated by two orbital transitions: the Wallingford, CT, 2004.
HOMO to LUMO transition, and a HOMO
1/ [15] A.D. Becke, J. Chem. Phys. 98 (1993) 5648.
HOMO
2 to LUMO transition. On the basis of these [16] C. Lee, G. Yang, R.G. Parr, Phys. Rev. B 37 (1988) 785.
[17] G. Schaftenaar, J.H. Noordik, J Comput.-Aided Mol. Design 14
calculations, a pair of canonical Lewis structures for the (2000) 123.
ground state of anthocyanidins is suggested in Fig. 5, to [18] P.B. Buseta, J.C. Colleter, M. Gadret, Acta Cryst. B 30 (1974)
replace the one commonly in use (cf. Fig. 1). This pair 1448.
consists of a positive charge located either on C2 or on [19] K. Ueno, N. Saito, Acta Cryst. B 33 (1977) 114.
C4, instead of O1, with the C2-positive resonance structure [23] R.A. McClelland, S. Gedge, J. Am. Chem. Soc. 102 (1980)
5838.
being the dominant member of the pair. These Lewis [24] J. Bakker, P. Bridle, T. Honda, H. Kuwano, N. Saito, N.
structures are consistent with much of the chemistry of Terahara, C.F. Timberlake, Phytochemistry 44 (1997) 1375.
anthocyanidins, and would be one of only a small handful [25] H. Fulcrand, C. Benabdeljalil, J. Rigaud, V. Cheynier, M.
of examples of a stable carbocation existing in nature. Moutonet, Phytochemistry 47 (1998) 1401.
[26] H. Fulcrand, P.-J.C. dos Santos, P. Sarni-Manchado, V.
Cheynier, J. Favre-Bonvin, J Chem. Soc. Perkin Trans. 1 (735)
(1996).
Acknowledgement [27] Y. Lu, Y. Foo, Y. Sun, Tetrahedron Lett. 43 (2002) 7341.
[28] M. Schwarz, T.C. Wabintz, P. Winterhalter, J. Agric. Food
The author gratefully acknowledges the financial sup- Chem. 51 (2003) 3682.
port of the EOU Faculty Scholars fund. [29] C. Roehri-Stoeckel, E. Gonzalez, A. Fougerousse, R. Brouillard,
Can. J. Chem. 79 (2001) 1173.
[30] J.B. Harbone, Biochem. J. 70 (1958) 22.
[31] R. Bauernschmitt, R. Ahlrichs, Chem. Phys. Lett. 256 (1996)
References 454.
[32] J. Cioslowski, P. Piskorz, G. Liu, D. Moncrieff, J. Phys. Chem.
[1] A.D. Sarma, R. Sharma, Phytochemistry 52 (1999) 1313. 100 (1996) 19333.
[2] S. Meiers, M. Kemény, U. Weyand, R. Gastpar, E. von Angerer, [33] A. Almenningen, O. Bastiansen, L. Fernholt, B.N. Cyvin, S.J.
D. Marko, J. Agric. Food Chem. 49 (2001) 958. Cyvin, S. Samdal, J. Mol. Struct. 128 (1985) 59.