Experimental and computational studies on the solvatochromism and thermochromism of 4-pyridiniophenolates

Article abstract of DOI:10.1039/b204654g

The observed solvatochromism of the betaine 4-(2,4,6-triphenyl-1-pyridinio)phenolate (2a) in solvents of low and very high relative permittivity has been assessed both experimentally and theoretically using molecular orbital methods. The PM3/COSMO method

Full text of DOI:10.1039/b204654g

View Article Online / Journal Homepage / Table of Contents for this issue
Experimental and computational studies on the solvatochromism
and thermochromism of 4-pyridiniophenolates
John O. Morley* and James Padfield
2
Chemistry Department, University of Wales Swansea, Singleton Park, Swansea, UK SA2 8PP
Received (in Cambridge, UK) 14th May 2002, Accepted 5th August 2002
First published as an Advance Article on the web 13th September 2002
The observed solvatochromism of the betaine 4-(2,4,6-triphenyl-1-pyridinio)phenolate (2a) in solvents of low and
very high relative permittivity has been assessed both experimentally and theoretically using molecular orbital
methods. The PM3/COSMO method suggests that there are a number of possible conformations for the betaine
involving clockwise or anti-clockwise rotation of the four pendant phenyl groups relative to the central heterocyclic
ring. The large thermochromic effect observed for the dye in acetone or tetrahydrofuran on moving from room
temperature to Ϫ78 ЊC has been attributed to a combination of conformational changes coupled with an increase in
the relative permittivity of the respective solvent. The calculated spectroscopic shifts for the betaine using a multi-
electron configuration interaction treatment show similar trends to those found experimentally in aprotic solvents. In
solvents with acidic hydrogens, the large hypsochromic shift observed for the betaine in the visible region arises from
both a dielectric effect and a hydrogen bonding effect. Stable hydrogen bonded structures are predicted to be formed
between either water, chloroform, or acetonitrile and the exocyclic oxygen atom of the betaine. The overall shifts
observed in these solvents show a good correlation with those calculated for the postulated solvates using a version of
the CNDO/S method.
Introduction
The visible absorption spectra of polar conjugated organic
molecules are strongly influenced by the solvent used for
measurement.1 This is especially true for merocyanines such
as (E)-1-(4-methyl-4-azaphenyl)-2-(4-oxyphenyl)ethene (1)2,3
which changes colour from blue to orange in moving from a
solvent of chloroform to water, and also for betaines such as 4-
(2,4,6-triphenyl-1-pyridinio)phenolate (2a),4,5 and the more well
known 2,6-diphenyl-4-(2,4,6-triphenyl-1-pyridinio)phenolate
(2b),5–7 called Reichardt’s dye, where the long wavelength
absorption moves from 810 nm in diphenyl ether to 453 nm
in water. The very large negative solvatochromism or hypso-
chromic shift observed for 2b in moving from non-polar to
polar solvents provides the basis of the so-called ET(30) scale of
solvent polarity developed by Reichardt.1,5–7
direction in the change of solvent from hexane to water with the
former moving from 552 to 684 nm and the latter from 484 to
593 nm.
Recently we have reported on the structure and solvato-
chromism of both the merocyanine (1)17,18 and phenol blue
(3),19 and showed that the former exists as a zwitterion in all
solvents17 while the latter exists as a quinoneimine19 as shown
in the structural representations (above). Our calculations
strongly suggested that the large solvatochromic shifts observed
for each dye could be satisfactorily rationalised in terms of the
dielectric effect of the solvent and its ability to hydrogen bond
at the exocyclic oxygen atoms of the respective molecule.17,19
However, despite a very large number of experimental studies
on the spectroscopic properties of Reichardt’s betaine (2b),
there have been relatively few theoretical studies partly because
of the molecular size which precludes the use of accurate
molecular orbital methods, and partly because of the large
number of possible conformations which can be obtained by
rotating the phenoxide ring B relative to the pyridinium ring A
(Scheme 1) coupled with the independent rotation of the five
peripheral phenyl rings relative to both of these. However, there
have been a number of quantitative studies,20–26 including four
molecular orbital studies23–26 at the semi-empirical AM1 level27
on the structure and electronic properties of the dye. In the first
of these studies,23 spectroscopic calculations were carried
out on the AM1 structure of 2b, which was optimised in the
gas phase, using a limited CI expansion coupled with a self
consistent reaction field (SCRF) procedure.28 An approximate
correlation was found between the calculated results and
experimental data though the predicted absorptions of 400–
450 nm in solvents with relative permittivitives ranging from
There are other polar molecules, such as Phenol Blue [N-(4-
dimethyaminophenyl)-1,4-benzoquinoneimine (3), also called
Indoaniline]8–14 and Nile Red [7-diethylamino-3H-[1,2]-benzo-
phenoxazine-3-one (4)],8–15,16 which exhibit positive solvato-
chromism with the absorption maximum moving in the opposite
1698
J. Chem. Soc., Perkin Trans. 2, 2002, 1698–1707
This journal is © The Royal Society of Chemistry 2002
DOI: 10.1039/b204654g

View Article Online
structure of the simpler pyridiniophenolate (2a) has been
determined by X-ray crystallography40 (see later).
Some of the theoretical studies discussed above can be
criticized, however, either because the optimised gas phase
structure used for the subsequent spectroscopic calculations
does not correlate particularly well with experimental data
(see later), or because the conformational effects involving the
rotation of ring A relative to ring B around the central N1–C7
bond appear to have been neglected, or because specific
hydrogen bonding interactions between the solvent and the exo-
cyclic oxygen, O13, of the dye have been ignored in some cases.
Furthermore, there is a paucity of structural information dis-
closed in the previous theoretical papers, and few comparisons
are made between the calculated structures and relevant crystal-
lographic data present in the Cambridge Structural Database41
especially in relation to the key N1–C7 and C10–O13 bond
lengths of the dye (Scheme 1), which would be expected to have
a marked effect on the transition energies. There has been little
attempt also to compare the calculated dipole moments with
experimental data42 and no reference has been made at all
Scheme
1
Numbering convention adopted for the 4-pyridinio-
phenolates (2).
1
around 5 to 50 occur at much shorter wavelength than those
found experimentally at 630–700 nm. In a related study,25 the
electronic and spectroscopic properties of the AM1 optimised
structure were evaluated in the gas phase and in water only
using the GRINDOL package,29 which models the effect of
solvent using a Langevin dipole/Monte Carlo approach, to give
good correlations between the calculated and experimental
spectra. This work was extended26 to show that the minimum
energy conformation of the dye was obtained with the phen-
oxide ring B rotated relative to the pyridinium ring A by around
60Њ in the gas phase and around 90Њ in water (Scheme 1). These
studies showed also that the molecular hyperpolarisability
of the dye is at a maximum when rings A and B are forced to be
co-planar. The electronic contributions to the two-photon
absorption cross section of the pyridiniophenolates have also
been modelled using the GRINDOL method.30 In another
study,24 the AM1 method was used to optimise the structure of
the dye using three separate conformations involving rotation
of all the phenyl rings, both in the gas phase and in solution,
with hydrogen bonding effects evaluated using a supermolecule
approach. A good correlation was claimed between the cal-
culated results evaluated using the INDO/S method31 and
experimental data though the predicted absorptions of 832 and
779 nm in chloroform and acetonitrile occur at longer wave-
length than those found experimentally at 730 and 622 nm
respectively.
Recently, density functional theory and a SCI treatment has
been used to probe the structure and visible absorption spectra
of pyridiniophenolates and pyridinium betaines32 but the struc-
tures obtained do not correlate particularly well with experi-
mental data particularly for the key aryl–oxygen bond lengths
which are much shorter than expected (see later). Ab initio
methods have also been used to model the ground state struc-
ture and spectra of pyridiniophenolates in the gas phase and in
solution.33–35 In water and other hydroxylic solvents, moderately
strong hydrogen bonds are predicted33 though the predicted
spectra are less satisfactory.
to detailed H and 13C NMR studies43 on the dye in various
solvents.
The present work has been carried out to probe the structure
of betaine dyes in solution and to attempt to relate the calc-
ulated structure and spectra to experimental results. We have
synthesised the simpler betaine (2a) and carefully measured its
spectrum in a much wider variety of solvents of differing rela-
tive permittivity and proton donating ability than previously
reported.4,5 While the omission of two phenyl rings relative to
the betaine (2b) moves the absorption band of 2a to shorter
wavelength, the overall shifts observed in moving from solvents
of low to high relative permittivity are very similar for 2a and
2b, but the NMR spectrum is simplified. The omission of two
phenyl rings also simplifies the computational problem in terms
of the number of possible conformations and makes it easier to
understand the large effect of solvation on the structure both in
terms of the relative dielectric effect and hydrogen bonding.
Methods of calculation
Molecular orbital calculations were carried out on empirical
structures for the betaine (2a) and (2b) using the AM127 and
PM344 methods of the MOPAC 93 Program45 with full
optimisation of all bond lengths, angles, and torsion angles
except where stated otherwise. The atom numbering convention
used is shown in Scheme 1. The effect of solvents of varying
relative permittivity (ε) on the structures and energies of the
betaine (2a) were assessed at the AM1 and PM3 levels using the
COSMO method47 incorporated in the MOPAC 93 program.
In this solvation model, the solute molecule is embedded in a
cavity constructed from the intersecting van der Waals spheres
of the component atoms surrounded by a dielectric continuum
of permittivity, ε. The surface between the continuum and the
solute is then partitioned into a large number of segments and
the interaction between the charge density at each segment
polarises the surrounding medium and produces a reaction
field which in turn acts on the solute. Typical solvents used in
the calculations included tetrahydrofuran (ε = 7.52), pyridine
(ε = 13.4), acetone (ε = 21.0), and dimethyl sulfoxide (ε = 47.2)
(keywords for the calculation in tetrahydrofuran: prec pm3 ef
xyz geo-ok eps = 7.52). The most stable conformation for the
betaine was assessed using mainly the PM3 method by selecting
a number of different starting geometries for the optimisation
by rotating rings B, C, D, and E, either clockwise or anticlock-
wise relative to the plane of ring (A). While there are many
conformations possible for each of the pendant phenyl rings B,
C, D, and E where the ring edges C8–C9, C18–C19, C24–C25,
and C27–C28, lie clockwise or anti-clockwise to the plane of
the central ring A, on calculation only a limited number were
found to be stationary points on the potential energy surface.
The excited state geometry of Reichardt’s dye (2b) has been
calculated and the results predict that the central phenoxide and
pyridinium rings move from a twisted conformation in the
ground state into a perpendicular position in the first excited
state.36 This intramolecular rearrangement is supported by
experimental femtosecond transient absorption studies in polar
solvents which shows that the excited and ground state con-
formations are different.37 This change, which is related to rapid
intramolecular electron transfer,37 has been probed also by
measuring the electromagnetic waveform broadcast by the
charge transfer process itself.38 Very recently, the temperature
dependent absorption spectrum or thermochromism of
Reichardt’s dye (2b) has been analysed in acetonitrile39 and the
J. Chem. Soc., Perkin Trans. 2, 2002, 1698–1707
1699

View Article Online
These were identified by using force calculations on each
separate structure (keywords for the calculations in water:
pm3 xyz eps = 80.1 force large let isotope) and only those
showing positive vibrational frequencies were considered
further. Spectroscopic calculations were carried out on the fixed
optimised ground state geometry using the multi-electron con-
figuration interaction (MECI) treatment in MOPAC 93 which
considers 4900 configurations generated between all com-
binations of up to eight electrons in the four highest occupied
and four lowest unoccupied molecular orbitals (keywords for
the ground state calculation in water: pm3 prec 1scf xyz eps =
80.1 ef geo-ok meci singlet c.i. = 8 root = 1 vectors), and the
transition energy evaluated from the difference between the
heats of formation of the modified ground state energy and the
first excited singlet state energy (same keywords except: root =
2 and open(2,2) added). The spectra were calculated also using
the CNDOVS method which has been specifically developed
for dyes and pigments,46 using a solvent dependent value for the
spectroscopic constant (see later), a core coefficient of 0.33,
and 50 singly excited configurations between seven occupied
and seven unoccupied molecular orbitals. Molecules and crys-
tal structures were displayed and analysed using the SYBYL
Molecular Modelling package.48
N–C bond length and average C–C bond lengths between
the pendant phenyl groups and the central pyridinium ring to
2b with values of 1.470 and 1.481 Å, but the C–O bond length
is shorter at 1.279 Å. The 2-, 4-, and 6-phenyl groups in this
structure are inclined at torsion angles of 51.0, 44.3, and 57.4Њ
to the plane of the central heterocyclic ring. A further highly
relevant crystal structure present in the Cambridge Database
is the simpler 1-(4-hydroxyphenyl)-2,4,6-triphenylpyridinium
cyanate (7),52 which shows N–C and average Ph–C bond lengths
of 1.459 and 1.478 Å respectively, with the B and C rings twisted
in opposite directions by 68.5 and 22.9Њ relative to the plane
of the pyridinium ring. The remaining phenyl rings, D and
E, are also twisted in opposite directions by 54.2 and 66.4Њ
relative to the same plane.
Results and discussion
1
Structural data
An examination of the Cambridge Structural Database41
reveals a number of structures which are highly relevant to the
present studies on the betaine (2a). For example, the solvated
crystal structure49 of the closely related 2,6-diphenyl-4-[(4-
bromophenyl-2,6-diphenyl)-1-pyridinio]phenolate (2c) shows
N1–C7 and C10–O13 bond lengths of 1.479 and 1.290 Å and
average C–C bond lengths between the pendant phenyl groups
and the central rings A (Ph–A) and B (Ph–B) of 1.494 Å. In the
crystal structure, which contains heavy atoms only, the exo-
cyclic oxygen, O13, is close to the oxygen of an ethanol
molecule at an O13–OC2H5 distance of 2.174 Å. The addition
of a hydrogen atom to the sp3 hybridized oxygen of the solvent
at an O–H bond length of 0.95 Å and a torsion angle H–O–
C(H2)–C(H3) of Ϫ105Њ using the SYBYL molecular modelling
program,48 generates a strong intermolecular hydrogen bond
between O13 and the acidic hydrogen of ethanol at a O13–H
distance of 1.766 Å. Both ring B and ring C are rotated anti-
clockwise relative to ring A with torsion angles C2–N1–C7–C8
of 67.8Њ and C3–C4–C14–C19 of 16.9Њ so that the right hand
edges of both attached rings as shown in the Scheme lie above
the plane of ring A. Rings D and E are also rotated but in
different directions relative to ring A with torsion angles C5–
C6–C20–C25 of Ϫ51.9Њ and C3–C2–C26–C27 of 68.0Њ so that
the C24–C25 edge of D lies below the plane of ring A, and the
C27–C28 edge of E lies above. In contrast, the two phenyl
groups attached to ring B are rotated in the same direction with
torsion angles of 62.1 and 64.8Њ.
In the better resolved and related crystal structure of
4-(2,4,6-triphenyl-1-pyridinio)phenyltoluene-4-sulfonaminide
(5),50 and using the same numbering convention for the pyr-
idinium part of the molecule as adopted for 2a, similar N1–C7
and average Ph–A bond lengths are found with values of 1.463
and 1.478 Å respectively. Rings B and C are rotated anticlock-
wise to ring A with torsion angles C2–N1–C7–C8 of 105.9Њ and
C3–C4–C14–C19 of 18.9Њ both above the plane. Rings D and E
are also rotated relative to ring A with torsion angles C5–
C6–C20–C25 of 50.6Њ above the plane, and C3–C2–C26–C27
of Ϫ48.1Њ below the plane. The pendant rings, B, D and E, in
this structure therefore adopt different conformations to those
found in (2c).
Earlier this year, and after this paper had been submitted
for publication, details of the crystal structure of 4-(2,4,6-
triphenyl-1-pyridinio)phenolate (2a) have appeared.40 The
results reported are somewhat different to those discussed for
the pyridiniophenolate above, with a shorter N1–C7 bond
length of 1.446 Å and a longer C10–O13 bond length of 1.311
Å than expected, possibly because of the presence of six water
molecules in the unit cell.40
Most of the previous studies on Reichardt’s betaine (2b) have
used the AM1 method to generate the ground state structure
though there is little information reported on bond lengths.
We have repeated this work and find that our fully optimised
structure is slightly lower in energy (0.61 kcal molϪ1) than that
reported previously.24 This structure shows N1–C7 and C10–
O13 bond lengths of 1.409 and 1.248 Å respectively which are
clearly far too short by comparison with the experimental
values of 1.459 to 1.479 Å and 1.279 to 1.290 Å respectively
found for the same bonds in 2c, 5 and 6 and may lead to serious
errors in the calculation of spectroscopic properties (see later).
Similar results are obtained for the N1–C7 and C10–O13 bond
lengths in the simpler betaine (2a) with calculated values of
1.399 and 1.248 Å versus reported values of 1.446 and 1.311 Å
respectively40 (Table 1). We found the same trends for the C–O
bond length in the merocyanine (1)17 where the AM1 gas phase
calculated value of 1.242 Å compares unfavourably with the
experimental value53 of 1.304 Å. The crystallographic data
suggest that the C–O bond length of 1.290 Å for the betaine
(2c) lies closer to C–O single bond length of around 1.33 Å
found in hydrated sodium phenoxide,54 than the C᎐O
᎐
double bond length of around 1.22 Å found in benzoquinone.55
Furthermore, 13C NMR studies show that the structure of
the betaine (2b) does not appear to be greatly affected by the
solvent used for measurement,43 with the resonance of C10
which is adjacent to the exocyclic oxygen, O13, in ring B
(Scheme 1), occurring at 161.3 ppm in CDCl3, 161.0 ppm in
[2H6]dimethyl sulfoxide, 162.9 ppm in [2H6]acetone, and 164.4
ppm in CD3OD.43 Our results on betaine (2a) are similar, with
the same carbon, C10, resonating at 158.2 ppm in [2H6]acetone,
and 158.5 ppm in [2H6]dimethyl sulfoxide. These values are
comparable to those found for model compounds of the
The corresponding crystal structure of 1-(4-hydroxyphenyl)-
2,4,6-triphenylpyridinium-3-olate (6)51 also shows a similar
1700
J. Chem. Soc., Perkin Trans. 2, 2002, 1698–1707

View Article Online
Table 1 Calculated geometries of 4-(2,4,6-triphenyl-1-pyridinio)phenolate (2a) versus crystallographic data
Gas phase
AM1
Waterb
AM1
X-Ray data
Parametera
PM3
1.400
PM3
1.383
2cc
2ad
1.380
1.369
1.390
N1–C2
C2–C3
C3–C4
C4–C5
C5–C6
C6–N1
N1–C7
C7–C8
C8–C9
C9–C10
C10–C11
C11–C12
C12–C7
C10–O13
C2–C26
1.390
1.404
1.399
1.399
1.404
1.390
1.399
1.436
1.355
1.460
1.460
1.355
1.436
1.248
1.473
1.473
1.458
44.2
1.380
1.404
1.403
1.403
1.404
1.380
1.438
1.412
1.379
1.429
1.429
1.379
1.412
1.289
1.474
1.474
1.458
82.3
1.356
1.390
1.418
1.399
1.386
1.368
1.479
1.393
1.391
1.448
1.445
1.410
1.379
1.290
1.508
1.493
1.485
67.8
1.393
1.396
1.396
1.393
1.400
1.395
1.437
1.349
1.468
1.468
1.349
1.437
1.228
1.476
1.476
1.465
1.395
1.399
1.399
1.395
1.383
1.452
1.405
1.375
1.430
1.430
1.375
1.405
1.279
1.475
1.475
1.465
81.5
1.2
67.6
67.8
Ϫ0.766
ϩ0.685
90.12
1.446
1.391
1.375
1.404
1.311
1.487
C6–C20
C4–C14
1.487
60.0
26.1
C2–N1–C7–C8
C3–C4–C14–C19
C5–C6–C20–C25
C3–C2–C26–C27
Charge on O13
Charge on N1
∆Hf
43.6
0.5
67.4
37.8
54.4
54.4
Ϫ0.368
ϩ0.141
148.62
30.9
58.6
58.6
Ϫ0.715
ϩ0.044
109.91
16.9
51.9
68.0
67.4
47.7
Ϫ0.383
ϩ0.827
125.51
a Bond lengths in Angstroms and torsion angles in degrees; ∆Hf is the heat of formation in kcal molϪ1 b Optimised structures using the COSMO
.
method with a relative permittivity of 80.1. c From the Cambridge Structural Database (ref. 49). d Ref. 40.
betaine (2), such as the phenoxide anion, where the carbon
the di-tert-butyl derivative (2d) of 14.8
solvent.42
1.2 D in the same
attached to oxygen resonates at 168.1 ppm,56a but are different
from alternative structures such as benzoquinone, where the
carbon attached to oxygen resonates at 187.1 ppm.56b There
seems little doubt, therefore, that the betaine (2) exists mainly in
the zwitterionic form.
There appears to be little to choose between the semi-
empirical methods in terms of the predicted structure, but
because the PM3 method gives a more reasonable charge distri-
bution and also provides a much better account of inter-
molecular hydrogen bonding between the exocyclic oxygen and
several water molecules for the merocyanine (1),17 the PM3/
COSMO method was adopted for all subsequent calculations.
The AM1 results obtained in the gas phase for 2a and 2b,
both in this work and in previous studies, are more consistent
with the presence of a C᎐O double bond than the C–O single
᎐
bond length. Furthermore, the error between the calculated
N1–C7 bond length of 1.409 Å and the expected value of
1.446 Å for the betaine (2a), based on crystallographic data,40
is even larger and suggests that the gas phase structure used
for previous calculations of the visible spectra is inappropriate.
In our previous studies on the merocyanine (1), we found that
the C–O bond length increased significantly in length17 when a
dielectric field was imposed during the structure optimisation
using the COSMO routine present in MOPAC 93. Accordingly,
we have re-examined the AM1 structure optimisation of the
betaines (2a) and (2b) using the relative permittivity of water.
The results show a significant improvement to the N1–C7 and
C10–O13 bond lengths with values of 1.438 and 1.289 Å for 2a
and 1.434 and 1.271 Å for 2b respectively (Table 1). These
results compare favourably with those obtained for 2b using the
3–21G basis set in the dielectric field of water which gives values
of 1.468 and 1.260 Å respectively for the same bonds.35 The
corresponding optimisations for the same betaine (2a) and (2b)
in water using the PM3/COSMO method gives N1–C7 and
C10–O13 bond lengths of 1.452 and 1.279 Å for 2a and 1.446
and 1.261 Å for 2b (Table 1). However, the calculated atomic
charges at the aryl nitrogen, N1, and exocyclic oxygen atom,
O13, differ considerably between the methods with the AM1
method giving values of 0.044 and Ϫ0.715 for 2a and 0.069 and
Ϫ0.584 for 2b respectively, while the PM3 method is more
consistent with the formal written structure of the betaines with
values of 0.685 and Ϫ0.766 for 2a and 0.735 and Ϫ0.671 for 2b.
The choice of the PM3 method also gives a good account of the
dipole moment of the betaine (2a) with a calculated value
of 15.6 D in dioxane versus an experimental value reported for
2
Conformational studies
In terms of the molecular conformation of the betaine (2a),
there are a very large number of conformers possible, which
involve both clockwise and anticlockwise rotations of rings B,
C, D, and E relative to ring A, so that ring edges C8–C9, C18–
C19, C24–C25, and C27–C28 can be rotated from 0 to 180Њ
relative to the plane of ring A though symmetry will remove a
number of these possibilities. In this study, the initial point of
the optimisations were taken with the edges C8–C9 and C18–
C19 of rings B and C either both rotated clockwise (or anti-
clockwise) relative to the plane of ring A, or the former rotated
clockwise and the latter rotated anticlockwise resulting in an
approximately eclipsed or staggered conformation for the two
pendant rings when viewed from O13 along the axis C7–N1–
C4–C14. Similarly, the edges C24–C25, and C27–C28 of rings
D and E can also adopt an approximately eclipsed or staggered
conformation for the two pendant rings relative to the plane of
ring A. Thus in one set of conformers, rings B and C can be
rotated in the same directions relative to the plane of ring A,
while in another set they can be rotated in opposite directions,
with rings D and E rotated in the same or opposite directions
relative to the plane of ring A.
Around 90 representative conformers were selected initially
from the large number possible for 2a in different solvents,
where ring B was rotated by fixed 5 to 10Њ increments relative to
ring A, starting with the rings co-planar (C2–N1–C7–C8 = 0Њ),
and ending with them orthogonal to one another (C2–N1–C7–
C8 = 90Њ), and the structures fully optimised at the PM3/
J. Chem. Soc., Perkin Trans. 2, 2002, 1698–1707
1701

View Article Online
COSMO level. A substantial number of valid conformations
were found to lie within 1 kcal molϪ1 of the lowest energy con-
former with all showing the pendant rings B, C, D, and E to be
rotated relative to ring A. In the optimisation procedure, there
is clearly a compromise between the steric interactions of ring
hydrogens on the pendant rings B, D, and E, and the degree
of overlap between each of these rings and the central ring A.
For example, while a planar arrangement of the pendant rings
relative to the central ring A would maximise the overlap
between them, there is considerable clash between the hydro-
gens at carbons C8 and C31, and those at C12 and C21.
Conformational studies on Reichardt’s betaine (2b) and the
simpler analogue 2e at the AM1 level, report that the minimum
of the potential energy is reached in the gas phase when ring B
is twisted relative to ring A, by 60Њ in the former and 25Њ in the
latter, and by 90Њ for both molecules in water.26 Our results at
the PM3 level for 2a are similar in the gas phase to 2b and 2e,
with the minimum energy conformation showing ring B twisted
by 43.6Њ relative to ring A, but differ in solution where two
minimum energy conformers are predicted by the stepwise
rotation of ring B relative to ring A (Scheme 1). In tetrahydro-
furan, one of these corresponds to a torsion angle of around
65Њ between the rings, while the other corresponds to a torsion
angle of 90Њ, with both conformers showing the same calculated
heats of formation of 99.9 kcal molϪ1. In ethylene carbonate,
the lowest energy conformer is calculated at a torsion angle of
around 70Њ with another minima again predicted at 90Њ, with
calculated heats of formation here of 89.9 and 90.0 kcal molϪ1
respectively.
Rings C, D, and E of 2a are also predicted to be rotated by
varying degrees both in the gas phase and in solution. For
example, in one favoured conformation for the betaine (2a) in
acetone, which has a calculated heat of formation of 93.3 kcal
molϪ1, rings B and C are both rotated anticlockwise 75.6 and
4.7Њ relative to ring A (viewed along the axis O13–N1–C17
shown in Scheme 1), while rings D and E rotated in opposite
directions to one another by 71.4 and 71.5Њ respectively relative
to ring A. However, there is a slightly more stable conformation
predicted in acetone, which has a heat of formation of 92.9 kcal
molϪ1, with rings B and C rotated by 90.0 and 0.2Њ in the same
direction, while rings D and E rotated in opposite directions by
87.1 and 79.1Њ respectively all relative to ring A. Force calcula-
tions showed that all these conformers gave positive vibrational
frequencies thus ruling out the possibility that these are transi-
tion states rather than genuine minima on the potential energy
surface. Similar results were obtained when the optimisations
were carried out in other solvents.
While the small energy differences observed between the
conformers obtained by rotating ring B relative to ring A, are
consistent with the broad absorption bands observed for the
dye in almost all solvents, some of these especially those where
ring B is orthogonal to ring A, may not contribute to the visible
absorption spectrum because the exocyclic oxygen, O13, and
the pyridinium nitrogen, N1, are no longer conjugated. In these
studies, we have explored structures where ring B is at least
partially conjugated with ring A, mainly because the crystal
structures of the related betaines (2c) and (7) show the rings
twisted by 67.8 and 68.5Њ to each other respectively. As the
energy differences involved in rotating ring B relative to ring A
from around 55Њ to 75Њ is generally less than 1 kcal molϪ1 in
almost all the solvents studied, this range of rotation was sub-
sequently used to calculate the spectra of the dye in different
solvents (see later).
Fig.
1
(a) Absorption spectra of 4-(2,4,6-triphenyl-1-pyridinio)-
phenolate (2a) at room temperature and at Ϫ78 ЊC: (a) in tetra-
hydrofuran; and (b) in acetone.
conformers in solution. The change in colour of the dye with
varying temperature is supportive suggesting that the more
stable conformers predominate at low temperature. In the
original work on 2a in acetone, it is reported4 that the cherry-
red solution at 20 ЊC, changes to orange on cooling to Ϫ80 ЊC,
and then darkens to violet–red on warming to 68 ЊC. A similar
phenomenon has been reported also for Reichardt’s dye (2b)57
in chloroform where the absorption maximum changes from
663 nm at Ϫ75 ЊC to 730 nm at 25 ЊC and then to 744 nm
at 55 ЊC, with an associated colour change from green–
blue to yellow–green. Recently, the temperature dependent
absorption of the same dye 2b has been analysed in aceto-
nitrile using resonance Raman spectra, where the solvent
re-organisation after excitation, is found to decrease with
increasing temperature.39
Our results show that the absorption maximum of 2a in
acetone moves from 621 nm at room temperature to 512 nm
when the solution is cooled at Ϫ78 ЊC (CO2–acetone bath) with
a concomitant change in colour from blue to red. The original
position of the absorption maximum and blue colour are
restored on quickly warming to room temperature. The colour
changes reported in the original work4 suggest that the acetone
used contained water which would account for the shorter
wavelength absorptions observed. We observe a similar but
larger shift for 2a in tetrahydrofuran where the absorption
maximum moves from 716 nm at room temperature to 552 nm
when the solution is cooled at Ϫ78 ЊC (Fig. 1). Again the
change is reversed on warming. The broad shape of the absorp-
tion bands in acetone and tetrahydrofuran is similar at either
of these temperatures (Fig. 1) suggesting that the energy dif-
ferences among the conformers are small though the distri-
butions may be significantly changed by cooling. These con-
3
Thermochromic effects
The experimental spectrum of the dye 2a is consistent with the
conformational analysis carried out here with the very broad
absorption bands in different solvents (Fig. 1) reflecting the
small energy differences between the large number of possible
1702
J. Chem. Soc., Perkin Trans. 2, 2002, 1698–1707

View Article Online
formers are clearly much closer in energy to one another than
those we have reported for the merocyanine (1),18 which shows
a series of discrete well defined absorption bands in solvents of
low relative permittivity such as tetrahydrofuran.18
The explanation for the hypsochromic shift of the dye 2a
observed on cooling, or thermochromic behaviour, is attribut-
able at least in part, to the increase in the relative permittivity
of the solvent with decreasing temperature. For example, the
relative permittivity (ε) has been related to the temperature
using the polynomial expression:58
orbital method combined with continuum solvation model,
which does not invoke a range of empirical solvent parameters.
The shift observed in protic solvents was related to two main
effects only.17,19 The first effect arises from interactions between
the molecule and the dielectric field of the solvent, while the
second arises from specific hydrogen bonding interactions
between the solvent and the lone pair electrons at the exocyclic
oxygen of the respective molecule. Theoretically, good correl-
ations were found between the calculated transition energies
and the experimental data for Brooker’s merocyanine (1)17 and
Phenol Blue (3)19 in aprotic solvents, but the calculated results
for protic solvents which show strong hydrogen bonding inter-
actions were less satisfactory.
ε(T) = a ϩ bT ϩ cT2
We have used the same approach in these studies also to
model the observed solvatochromic shifts of the betaine (2a) on
the grounds that (i) the experimental shift in the absorption
maximum of polar dyes, such as 1, appears to be directly related
to relative permittivity of aprotic solvents,17 and (ii) the crystal
structures of 1 and 2c show discrete solvent molecules attached
to the exocyclic oxygens, with two distinct water molecules
hydrogen bonded in the first case,42 and an ethanol molecule
in the second.49 While there have been previous spectroscopic
studies on 4-(2,4,6-triphenyl-1-pyridinio)phenolate (2a), the
recorded data are limited to solvents such as water, a range of
alcohols, chloroform, aqueous acetone and aqueous pyridine
only.5 For this reason we have synthesised the dye (using a
modified procedure), and measured the long wavelength
absorption in a wide range of solvents (after this paper had
been submitted for publication, however, details of the spectra
of the anhydrous and hydrated dye in other solvents have been
reported40).
where T is given in K. Using the appropriate coefficients for
acetone obtained in the temperature range of 0 to 50 ЊC (where
a = 88.157, b = Ϫ0.3430, c = 38.925 × 10Ϫ5),48 the relative
permittivity is calculated to increase from 20.8 at 22 ЊC to 36.1
at Ϫ78 ЊC. Although the result at Ϫ78 ЊC, must be treated with
caution, a similar increase is found in tetrahydrofuran (a =
30.739, b = Ϫ0.12946, c = 17.195 × 10Ϫ5),58 where the relative
permittivity is calculated to increase from 7.51 at 22 ЊC to 12.0
at Ϫ78 ЊC. In this case, the temperature used for the calibration
of the coefficients ranged from Ϫ49 to 22 ЊC, suggesting that
the results at Ϫ78 ЊC are probably valid for both tetrahydro-
furan and acetone.
The absorption maxima estimated by interpolation for
the betaine (2a) using the calculated relative permittivitives at
Ϫ78 ЊC fall in the range 570–590 nm for acetone and 650–670
nm for tetrahydrofuran and represent a maximum of 33% of
the observed thermochromic shift in each case. In addition
to the substantial increase in the relative permittivity of each
solvent on cooling to Ϫ78 ЊC, the viscosity of the respective
solution would be expected to increase on cooling and promote
aggregation of the dye particularly in tetrahydrofuran. In the
case of the merocyanine (1), we found that solvents of high
viscosity and low relative permittivity promoted the formation
of a dimer, which is predicted to absorb at shorter wavelength
than the monomer.17,18 However, the viscosity of both acetone
and tetrahydrofuran (given in units of 10Ϫ3 kg mϪ1 sϪ1 through-
out) only increases from 0.306 and 0.456 respectively at 25 ЊC to
0.540 and 0.849 at Ϫ25 ЊC;59 extrapolation gives approximate
values of 1.05 and 1.40 at Ϫ78 ЊC. This increase is insufficient to
promote aggregation as other solvents used in this study such
as dioxane have a similar viscosity at 25 ЊC (1.18)59 to acetone
and tetrahydrofuran at Ϫ78 ЊC. Furthermore, the high relative
permittivity of acetone at Ϫ78 ЊC would almost certainly pre-
vent the formation of dimers or aggregates.
(a) Aprotic solvents. The solvents used here for measure-
ment of the solvatochromic shifts included dioxane, chloro-
form, ethyl acetate, tetrahydrofuran, methylene chloride,
pyridine, acetone, acetonitrile, N,N-dimethylformamide, di-
methyl sulfoxide, propylene carbonate, ethylene carbonate,
N-ethylacetamide, and N-methylacetamide which have relative
permittivities ranging from 2.22 to 179. As we have pointed out
previously, although each of the N-alkylacetamides used for
measurement possesses an acidic hydrogen atom and could be
considered as a hydrogen bond donor, crystallographic data on
N-methylacetamide strongly suggests that it is unlikely to form
hydrogen bonds with the oxygen of the betaine (2a), because
in the favoured trans conformation (either alone62 or in the
presence of metal salts41), the hydrogen is partly shielded by
the adjacent methyl hydrogens on the carbonyl carbon, and the
N-alkylacetamides probably act essentially as aprotic solvents.
The effect of the relative permittivitives of aprotic solvents
on the experimental absorption maximum are considerably
more marked for this betaine (2a) than the results we obtained
for the merocyanine (1) with much larger shifts observed in
moving from solvents with low relative permittivitives to those
with high values (Fig. 2). The magnitude of the overall shifts
observed are so large that hydrogen bonding effects are clearly
discernable in the experimental data for aprotic solvents with
acidic hydrogens such as chloroform and acetonitrile (see later).
The results show that the absorption maximum of the betaine
(2a) moves from 729 nm in dioxane (ε = 2.22) to 621 nm in
acetone (ε = 21.0), through to 512 nm in ethylene carbonate (ε =
89.8). However, the dielectric effect begins to saturate in the last
solvent, and further increases in the relative permittivity have a
much smaller effect on the absorption which moves to 489 nm
in N-ethylacetamide (ε = 135) and finally to 461 nm in N-
methylacetamide (ε = 179) (Table 2). The small anomalies in the
shifts observed for some solvents are possibly attributable to
the presence of traces of water which are difficult to remove
completely.
It follows that the large hypsochromic shift observed for the
betaine (2a) on cooling in these solvents almost certainly arises
from a combination of a dielectric effect coupled with a change
in the conformer distribution.
4
Solvatochromic shifts
The large hypsochromic shifts produced in the IR–Visible
spectrum of the betaine (2b) by different solvents is well docu-
mented particularly by the single parameter ET(30) scale of
solvent polarity.1,4–7 However, there are a number of more
complex treatments of solvent polarity,1,60,61 where the solvato-
chromic shift is related to the sum of parameters such as π*, α,
and β, where the π* parameter is a measure of the solute/
solvent interactions (dipole–dipole, dipole-induced dipole), and
the α and β parameters are quantitative empirical measures
of the ability of the solvent to act either as a hydrogen bond
donor or hydrogen bond acceptor respectively. This treatment
of solvent polarity generally relies on the measured interactions
or properties of the given solvent with some other substance.
In our previous studies on Brooker’s merocyanine 117 and
Phenol Blue (3)19 we theoretically modelled the solvatochromic
shift observed in aprotic solvents using a simple molecular
The absorption band observed in all solvents is generally
broad; in acetone (see Fig. 1), the absorption maximum occurs
J. Chem. Soc., Perkin Trans. 2, 2002, 1698–1707
1703

View Article Online
Table
2
Calculated absorption energies of 4-(2,4,6-triphenyl-1-
pyridinio)phenolate (2a) using the PM3/COSMO/MECI method versus
solvents, using the PM3/COSMO/MECI approach on fixed
ground state structures, depend on the relative permittivity of
the solvent, the molecular conformation selected, and the size
of the configuration interaction treatment used, i.e., on the
number of occupied and virtual orbitals included in the calcula-
tion. The calculated potential energy surface of the betaine in
different solvents only shows small changes for rotations of ring
B relative to ring A ranging from 55 to 75Њ (< 1 kcal molϪ1) but
the change of overlap between the two rings does result in
significant changes to the calculated transition energies between
the ground and first excited singlet state. As experimental
evidence suggests that a range of different conformations are
present in solution, it was decided to calculate the structure and
transition energies of the betaine in different solvents, at fixed
torsion angles of ring B relative to ring A of 55Њ (conformer I),
65Њ (conformer II), and 75Њ (conformer III) while allowing all
other variables to be optimised. The change to the ground state
energy is very small with the calculated energy of the betaine in
acetone moving from 93.08 kcal molϪ1 for the minimum energy
structure, where ring B is twisted by 70Њ, to 93.52 kcal molϪ1 for
conformer I, to 93.13 kcal molϪ1 for conformer II, and then to
93.58 kcal molϪ1 for conformer III. After the MECI treatment
with eight molecular orbitals included in the calculation, the
ground state is always stabilised further and the energy falls, for
example, to 91.03 kcal molϪ1 for the preferred conformer and to
91.24 kcal molϪ1 for conformer I.
While the calculated absorptions of the betaine (2a) show a
reasonable correlation with the experimental data in aprotic
solvents with relative permittivitives ranging from dioxane to
dimethyl sulfoxide (Table 2), the results for solvents with larger
relative permittivitives ranging from ethylene carbonate to N-
methylacetamide are less satisfactory with the predicted absorp-
tions at longer wavelengths than those observed experimentally
(Table 2). For example, the predicted absorptions for con-
formers I to III in ethyl acetate range from 635 to 731 nm
compared with the experimental absorption maximum of
673 nm, while those in dimethyl sulfoxide range from 542 to
626 nm respectively compared with the experimental absorp-
tion maximum of 591 nm (Table 2). However, the calculated
absorptions for conformers I to III in ethylene carbonate are
less satisfactory and range from 536 to 616 nm compared with
the experimental absorption maximum of 512 nm(Table 2)
suggesting that the molecule may be less twisted in solvents
of high relative permittivity. A subsequent calculation on a
structure optimised with a fixed torsion angle of 50Њ between
rings A and B in ethylene carbonate gives a predicted absorp-
tion at 506 nm. However, the calculated results obtained for the
N-alkylacetamides are clearly unsatisfactory with the predicted
wavelengths too large by comparison with experimental data
suggesting either that these solvents may be acting as weak
hydrogen bond donors or that the PM3/COSMO/MECI
method has reached its limit of accuracy.
experimental valuesa
Solvent
ε
θ
∆Hf
λcalc.
λexptl.
1,4-Dioxane
2.22 55
115.17
115.54
116.07
103.64
101.98
102.20
100.23
99.89
100.23
95.81
95.48
95.84
93.52
93.13
93.58
91.63
91.28
91.13
91.14
90.79
91.16
90.21
90.34
90.41
90.59
90.11
90.01
89.70
89.29
89.62
677
769
971
635
655
731
613
634
710
573
601
663
558
585
642
546
569
614
542
567
626
536
574
616
544
574
608
534
555
609
729
65
75
Ethyl acetate
6.08 55
673
716
658
621
609
591
512
489
461
65
75
Tetrahydrofuran
Pyridine
7.52 55
65
75
13.36 55
65
75
Acetone
21.01 55
65
75
N,N-Dimethylformamide
Dimethyl sulfoxide
Ethylene carbonate
N-Ethylacetamide
N-Methylacetamide
38.25 55
65
75
47.24 55
65
75
89.8
55
65
75
55
65
75
55
65
75
135.0
179.0
a ε is the relative permittivity (Ref. 48); θ is the torsion angle of ring B
relative to ring A (Њ); ∆Hf is the heat of formation (kcal molϪ1); λ is the
measured or calculated long wavelength absorption maximum (nm).
(b) Hydrogen bond donating solvents. In protic solvents, the
observed shifts are considerably larger than those expected
from the dielectric effect alone because of strong hydrogen
bonding between the acidic hydrogen of the solvent and the
electronegative oxygen of the betaine. For example, the overall
shift of Ϫ327 nm observed for betaine (2a) in moving from
dioxane to water (Table 3), arises from a hydrogen bonding
component (∆λH), and a dielectric component (∆λe) which can
be derived approximately from the shifts observed in aprotic
solvents as we have pointed out previously for the merocyanine
(1).17 Interpolation of the data for the expected shift in water
arising from the purely dielectric effect of the solvent (Fig. 2)
gives an estimated absorption at approximately 525 nm, repre-
senting a shift of Ϫ204 nm for the dielectric component relative
to dioxane. However, because the real absorption maximum
is found at 402 nm in water (Table 3), there is a further shift
of Ϫ123 nm arising from the hydrogen bonding effect of the
Fig. 2 Experimental absorption energies (λ/nm) of 4-(2,4,6-triphenyl-
1-pyridinio)phenolate (2a) versus the relative permittivity (ε) of aprotic
solvents.
at 621 nm but significant absorption is found also from around
530 nm to 680 nm with absorbance values here of 50% of
the main peak. The overall dielectric shift of Ϫ268 nm for the
absorption maximum of 2a in moving from dioxane to N-
methylacetamide (Table 2) is considerably larger than the di-
electric shift of Ϫ84 nm observed for the absorption maximum
of 1 in the same solvents,17 and larger also than the value of
Ϫ244 nm reported for Reichardt’s betaine (2b), where the
absorption moves from 794 nm in dioxane, to 634 nm in
dimethyl sulfoxide, then to 588 nm in ethylene carbonate, and
finally to 550 nm in N-methylacetamide.1b
The calculated absorption maxima for the dye 2a in different
1704
J. Chem. Soc., Perkin Trans. 2, 2002, 1698–1707

View Article Online
Table 3 Estimated effect of the relative permittivity and hydrogen
bonding ability of acidic solvents on the experimental low energy
absorption band of betaine (2a)a
molecule is clearly hydrogen bonded to the electronegative
carbonyl oxygen. In the zwitterionic betaine (2a), the corre-
sponding hydrogen bonding effect would be expected to be
even stronger and contribute to the large solvatochromic shift
of Ϫ118 nm observed in moving from dioxane to chloroform
versus an expected value of around Ϫ19 nm expected for the
dielectric effect alone (Fig. 2, Table 3).
Acetonitrile appears to exert a similar if smaller effect to
chloroform as the observed shift for the betaine in this solvent
at Ϫ169 nm is again larger than that expected from the
dielectric effect alone which is estimated at approximately
Ϫ124 nm (Fig. 2, Table 3). Again the Cambridge Structural
Database41 shows a number of polar carbonyl compounds
Solvent
ε
λlit
λ
∆λ
∆λε
∆λH
Dioxane
Chloroform
2.22
4.81 615
7.58
8.93
12.5
13.0
16.4
17.8
17.9
20.2
20.8
25.3
33.7
36.6
41.4
80.1
729
611
544
617
582
494
529
491
492
504
482
466
437
560
437
402
458
467
0
0
19
29
34
49
54
59
64
64
79
79
94
119
124
139
204
234
262
0
99
156
78
118
185
112
147
235
200
238
237
225
247
263
292
169
292
327
271
262
2-Ethylhexan-1-ol
Methylene chloride
2-Methylpropan-2-ol
n-Hexanol
Cyclohexanol
n-Butanol
2-Methylpropan-1-ol
Propan-2-ol
Propan-1-ol
Ethanol
Methanol
Acetonitrile
Ethane-1,2-diol
Water
560
98
181
141
174
173
146
168
169
173
45
484
present which contain acetonitrile as solvate with the C᎐
᎐
O ؒ ؒ ؒ H–CH2CN distance falling in the range of 2.40 Å
(bonding) to 2.80 Å (non-bonding). For example, this distance
is 2.43 Å in the crystal structure of 1-benzoyl-4-ethyl-1,3-
dihydro-5-[4-(2-methyl-1H-imidazol-1-yl)benzoyl]-2H-imida-
zol-2-one acetonitrile solvate67 suggesting that the solvent is
weakly hydrogen bonded to the electronegative oxygen as the
acidic hydrogen lies just within the combined van der Waals
radii of oxygen and hydrogen. A similar but stronger effect
would be expected for interactions between acetonitrile and the
betaine given the enhanced electronegativity of the oxygen
atom, O13.
467
442
153
123
37
412
Formamide
N-Methylformamide
111.0
189.0
0
a ε is the relative permittivity of the solvent (Ref. 48); λlit is the literature
absorption maximum (Ref. 5); λ is the measured absorption maximum
found in this work; ∆λ is the shift in absorption relative to that found in
dioxane; ∆λε is the estimated dielectric component of the shift derived
from interpolated data shown in Fig. 2; ∆λH is the hydrogen bonding
component of the shift.
In these studies we have carried out a series of calculations to
attempt to rationalise not only the shifts observed in protic
solvents such as water but also those found in aprotic solvents
such as chloroform and acetonitrile. Initial calculations were
carried out at the PM3/COSMO/MECI level on a represent-
ative low energy structure for the betaine in the dielectric field
of water, where ring B is twisted by 65Њ relative to ring A
(conformer II), to give a predicted absorption of 530 nm, which
clearly underestimates the experimental absorption of 402 nm
(Table 3) because of the absence of specific hydrogen bonding
effects. In our previous calculations at the PM3/COSMO level
in water on a hydrated merocyanine (1), which contained
hydrogen bonds between the oxygen of the dye and one or two
water molecules, we found that strongly bonded complexes were
formed.17 Similarly, the PM3/COSMO optimisation on the
betaine dihydrate in water with O13 in an sp2 conformation
gives a trigonal planar arrangement at the oxygen atom with
two O13–H2O hydrogen bond lengths of 1.762 Å, and C10–
O13–H and O13–H–O angles of 107 and 174Њ respectively, with
an overall energy of Ϫ32.4 kcal molϪ1. The corresponding
optimisation on the dihydrate where O13 adopts a pyramidal
sp3 conformation gives a slightly less stable structure with an
overall energy of Ϫ32.0 kcal molϪ1. The presence of the two
intermolecular hydrogen bonded water molecules result in
small changes only to the overall geometry of the betaine with
the central N1–C7 bond length essentially unaffected at 1.451–
1.452 Å while the C10–O13 bond is slightly stretched to 1.285 Å
relative to the value of 1.279 Å in the absence of the attached
water molecules. Furthermore, the free hydrogens and two
lone pair electrons residing on each of the oxygens of the two
complexed water molecules of the betaine dihydrate, would
be expected to form additional hydrogen bonds with further
water molecules to give polyhydrates such as a trihydrate,
hexahydrate, and heptahydrate. For example, a further PM3/
COSMO optimisation on the hexahydrate in water gives a
structure with the two O13–H2O hydrogen bond lengths now
slightly shortened to around 1.758 Å, with four additional
(O13–H2O)–H2O hydrogen bonds predicted at distances 1.775
Å. These predictions are consistent with the recently published
crystal structure of the hexahydrate,40 where two water mole-
cules are indeed hydrogen bonded to the exocyclic oxygen,
with four other water molecules also attached, though only
oxygen–oxygen distances are reported. Assuming an O–H bond
length of around 0.96 Å in H2O, and an almost linear O13–H–
OH angle, the O13–H2O hydrogen bond lengths in the crystal
solvent (Table 3). This approximate division of dielectric and
hydrogen bonding effects can be extended to all the other
solvents. Despite the approximate values estimated for the di-
electric components of protic solvents, the hydrogen bonding
components of the solvatochromic shift, evaluated for eight
primary alcohols explored in these studies are remarkably con-
sistent and fall in the range of Ϫ168 to Ϫ181 nm and tend to
dominate the overall shift observed except at high relative
permittivitives (Table 3). For example, the hydrogen bonding
component of the shift in butan-1-ol at Ϫ174 nm is much
greater than the dielectric component of Ϫ64 nm, in methanol
the margin narrows to Ϫ173 and Ϫ119 nm respectively, but
in ethane-1,2-diol the dielectric component of Ϫ139 nm is
comparable to the hydrogen bonding component of Ϫ153 nm.
For secondary alcohols such as propan-2-ol and cyclohexanol,
the hydrogen bonding shift is smaller than that found for
primary alcohols with values of Ϫ146 and Ϫ141 nm respec-
tively, while in tertiary butanol the shift is only Ϫ98 nm
(Table 3). This trend appears to fit in neatly with the pKavalues
of 16, 16.5, and 17 for the acidity of primary, secondary, and
tertiary alcohols.63
The observed shifts for acidic aprotic solvents such as chloro-
form, methylene chloride, and acetonitrile (Table 2), are also
larger than expected from the dielectric effect alone, because
they are able to act as hydrogen bond donors and complex with
the electronegative oxygen of the betaine. Crystallographic data
from the Cambridge Structural Database41 are supportive as
there are a number of polar aromatic systems present which
contain chloroform as solvate, though there are no solvated
zwitterions directly related to the betaine (2). The crystal
structures of a number of aryl ketones, which in principle can
form hydrogen bonds between the lone pair electrons on the
carbonyl oxygen and the acidic hydrogen of chloroform in
a similar way to that proposed for the betaine, show strong
intermolecular interactions with the chloroform molecule close
to the carbonyl group at a C᎐O ؒ ؒ ؒ H–CCl distance ranging
᎐
3
from 2.15 to 2.40 Å. For example, in the well resolved crystal
structure of 3,4-diphenyl-2-piperidino-5H-1-azabenz[cd]azulen-
5-one chloroform solvate,64 the distance between the carbonyl
oxygen and the hydrogen of chloroform is 2.15 Å. Given that
the van der Waals radius for hydrogen ranges from 1.01 to 1.17
Å and that for oxygen is around 1.52 Å,65,66 the chloroform
J. Chem. Soc., Perkin Trans. 2, 2002, 1698–1707
1705

View Article Online
are estimated to be around 1.82 Å, while the (O13–H2O)–H2O
bond lengths are estimated to be approximately 1.98 Å.
combination of the dielectric effect combined with the hydro-
gen bonding effect of one chloroform molecule.
The presence of attached water molecules at the oxygen atom
of the PM3 structure, however, has little effect on the calculated
transition energies, with the absorption of the dihydrate and
hexahydrate of conformer III moving to the slightly longer
wavelengths of 540 and 548 nm respectively compared with the
value of 530 nm obtained for the non-hydrated structure in
water. Exactly the same trends were found in our calculations
on the hydrates of both Brooker’s merocyanine (1)17 and Phenol
Blue (3).19 The failure of the MECI procedure to reproduce
the experimental trends for the hydrates arises because of the
limited number (8) of occupied and unoccupied orbitals which
can be normally included in the full configuration interaction
treatment (see Methods of calculation). In our spectroscopic
calculations on Phenol Blue (3),19 we showed that it was neces-
sary to include much higher energy molecular orbitals in the CI
so that n–π* transitions were also considered.
Acetonitrile, like chloroform, also produces a larger solvato-
chromic shift than that expected from the dielectric effect of the
solvent alone (Table 3). As in the previous case, PM3/COSMO
calculations predict that acetonitrile forms a stable solvate with
the betaine (2a) where the acidic hydrogen is hydrogen bonded
to the exocyclic oxygen, at an O13–HCH2CN bond length of
1.813 Å and an O13–H–CH2CN bond angle of 180Њ. The
hydrogen bond in this case is clearly weaker than those calcu-
lated for the corresponding betaine solvates with either water
or chloroform. Spectroscopic calculations on the same PM3/
COSMO structure optimised in the dielectric field of aceto-
nitrile, again at the CNDOVS level with the spectroscopic con-
stant set at 0.75, gives a predicted absorption of 617 nm for
conformer II of the betaine (2a) which is broadly consistent
with the expected absorption of the molecule in the dielectric
field of the solvent alone at 605 nm (Fig. 2). Following the
previous procedure, a corresponding CNDOVS calculation on
the PM3/COSMO optimised solvate in the dielectric field of
acetonitrile, gives a predicted absorption of 564 nm which is
fully consistent with the experimental absorption maximum
of 560 nm (Table 3) and confirms that the overall shift is again
satisfactorily represented by a combination of the dielectric
effect of the solvent combined with the hydrogen bonding effect
of one acetonitrile molecule.
The alternative CNDOVS method46 which we have
developed for dyes and pigments, employs singly excited con-
figurations only and gives a good account of the spectroscopic
shifts of the hydrates of both Brooker’s merocyanine (1)17 and
Phenol Blue (3).19 Spectroscopic calculations on the PM3/
COSMO structures optimised in the dielectric field of water, at
the CNDOVS level46 using a spectroscopic constant of 0.86,
gives a predicted absorption of 518 nm for conformer II of the
betaine (2a) which is broadly consistent with the expected
absorption of the molecule in the dielectric field of water alone
at 525 nm (Fig. 2). Using the same CNDOVS parameters on the
PM3/COSMO optimised structures obtained for the hydrated
betaines in the dielectric field of water, give predicted absorp-
tions of 465 nm for the dihydrate, 444 nm for the trihydrate,
425 nm for the hexahydrate and 408 nm for the heptahydrate.
These shifts are fully consistent with the experimental absorp-
tion maximum of 402 nm (Table 3) and tend to confirm that
the overall shift can be represented satisfactorily using the
PM3/COSMO/MECI approach to model the dielectric effect
of water, and the CNDOVS method to model the hydrogen
bonding effect of water molecules attached to the exocyclic
oxygen of the betaine.
In chloroform, similar calculations on the betaine (con-
former II) in the dielectric field of chloroform gives a predicted
absorption of 713 nm which underestimates the experimental
absorption of 611 nm (Table 3) probably because of the absence
of specific hydrogen bonding effects. PM3/COSMO calcula-
tions predict that chloroform forms a stable mono-solvate with
the betaine (2a) where the acidic hydrogen is strongly hydrogen
bonded to the exocyclic oxygen, at an O13–HCCl3 bond length
of 1.758 Å and an O13–H–CCl3 bond angle of 180Њ. The form-
ation of a di-solvate seems less likely as there is a potential
clash between chlorine atoms located on adjacent chloroform
molecules when the acidic hydrogens of each are attached to the
oxygen atom. These results differ markedly from those obtained
on Reichardt’s dye (2b) using the AM1 method where the
hydrogen bonding to chloroform is described as quite weak.24
Again, the presence of one attached chloroform molecule at
the oxygen atom of the PM3 structure has little effect on the
calculated MECI transition energies, with the absorption of the
solvate now at 711 nm.
Experimental
Visible spectra were measured on a Unicam UV300 UV–Vis
spectrometer using a matched pair of quartz cells of 1 cm path
length. Solutions of the betaine (2a) were recorded at room
temperature (unless indicated otherwise) in solvents of the
highest purity (spectroscopic grade and/or purified and dried by
appropriate methods prior to use) where possible.
4-(2,4,6-Triphenyl-1-pyridinio)phenolate (2a)
The preparation was carried out using a similar procedure68 to
that reported by Kessler and Wolfbeis for their synthesis of the
betaine (2b). A mixture of 2,4,6-triphenylpyrylium hydrogen
sulfate ( 0.406 g 1 mmol, prepared according to the literature
procedure69), 4-aminophenol (0.109 g, 1 mmol), and anhydrous
sodium acetate (0.41 g, 5 mmol), were added to 5 mL of
ethanol. The mixture was refluxed for 3 h, cooled, and 3.5 mL
of 5% aqueous sodium hydroxide solution was added. Most
of the ethanol was removed by distillation, and the solution
cooled in an ice bath. The resulting red crystals were collected
by filtration, recrystallised from aqueous ethanol, and dried
over P2O5 in vacuo for several days to yield deep blue crystals
(0.34 g, 85%), mp 195–199 ЊC (lit.5 198 ЊC).
References
1 (a) C. Reichardt, Solvents and Solvent Effects in Organic Chemistry,
Verlag Chemie, New York, 1988, and references therein;
(b) C. Reichardt, Chem. Rev., 1994, 94, 2319.
2 L. G. S. Brooker, C. H. Keys, R. H. Sprague, R. H. Van Dyke,
E. Van Zandt, F. L. White, H. W. J. Cressman and S. G. Dent,
J. Am. Chem. Soc., 1951, 73, 5332.
3 L. G. S. Brooker, C. H. Keys and D. W. Heseltine, J. Am. Chem. Soc.,
1951, 73, 5350.
4 W. Schneider, W. Dobling and R. Cordua, Ber. Dtsch. Chem. Ges.,
1937, 8, 1645.
5 K. Dimroth, C. Reichardt, T. Siepmann and F. Bohlmann, Liebigs
Ann. Chem., 1963, 661, 1.
6 (a) K. Dimroth and C. Reichardt, Liebigs Ann. Chem., 1969, 727, 93;
(b) C. Reichardt, Liebigs Ann. Chem., 1971, 752, 64.
7 C. Reichardt and E. Harbusch-Gornert, Liebigs Ann. Chem., 1983,
721.
8 O. W. Kolling, Anal. Chem., 1978, 50, 212.
9 L. G. S. Brooker and R. H. Sprague, J. Am. Chem. Soc., 1941, 63,
3214.
Spectroscopic calculations on the PM3/COSMO structure
optimised in the dielectric field of chloroform, at the CNDOVS
level this time with the spectroscopic constant adjusted to 0.71,
gives a predicted absorption of 710 nm for conformer II of the
betaine (2a) which coincides exactly with the expected absorp-
tion of the molecule in the dielectric field of the solvent alone
(Fig. 2). Using the same CNDOVS parameterisation to calcu-
late the PM3/COSMO optimised solvate, gives a predicted
absorption of 606 nm which is now fully consistent with the
experimental absorption maximum of 611 nm (Table 3) and
confirms that the overall shift is satisfactorily represented by a
1706
J. Chem. Soc., Perkin Trans. 2, 2002, 1698–1707

View Article Online
10 A. L. LeRosen and C. E. Reid, J. Chem. Phys., 1952, 20, 233.
11 (a) J. J. Figueras, J. Am. Chem. Soc., 1971, 93, 3255; (b) J. J. Figueras,
P. W. Scullard and A. R. Mack, J. Org. Chem., 1971, 36, 3497.
12 C. A. G. O. Varma and E. J. G. Groenen, Recl. Trav. Chim. Pays-Bas,
1972, 91, 296.
13 (a) O. W. Kolling and J. L. Goodnight, Anal. Chem., 1973, 45, 160;
(b) O. W. Kolling, Anal. Chem., 1981, 53, 54.
43 J. G. Dawber and A. Williams, J. Chem. Soc., Faraday Trans. 1, 1986,
82, 3097.
44 J. J. P. Stewart, J. Comp. Chem., 1989, 10, 209.
45 MOPAC 93 (J. J. P. Stewart and Fujitsu Limited, Tokyo,
Japan: copyright (Fujitsu Limited, 1993) obtained from QCPE,
Department of Chemistry, Indiania University, Bloomington,
Indiana 47405, USA.
14 (a) S. Kim and K. P. Johnston, ACS Symp. Ser., 1987, 329, 42;
(b) S. Kim and K. P. Johnston, Ind. Eng. Chem. Res., 1987, 26, 1206;
(c) K. P. Johnston and S. Kim, ACS Symp. Ser., 1989, 406, 52.
15 M. M. Davies and H. B. Hetzer, Anal. Chem., 1966, 38, 451.
16 J. F. Deye, T. A. Berger and A. G. Anderson, Anal. Chem., 1990, 62,
615.
17 J. O. Morley, R. M. Morley, R. Docherty and M. H. Charlton,
J. Am. Chem. Soc., 1997, 119, 10192.
18 J. O. Morley, R. M. Morley and A. L. Fitton, J. Am. Chem. Soc.,
1998, 120, 11479.
19 J. O. Morley and A. L. Fitton, J. Phys. Chem. A, 1999, 103, 11442.
20 I. Jano, J. Chim. Phys., 1992, 89, 1951.
21 C. Streck and R. Richert, Ber. Bunsen-Ges. Phys. Chem., 1994, 98,
619.
46 M. H. Charlton, R. Docherty, D. J. McGeein and J. O. Morley,
J. Chem. Soc., Faraday Trans., 1993, 89, 1671.
47 (a) A. Klamt and G. Schuurmann, J. Chem. Soc., Perkin Trans. 2,
1993, 799; (b) See also A. Klamt, J. Phys. Chem., 1996, 100, 3349.
48 Sybyl, Version 6.3; Tripos, Inc.: 1699 S. Hanley Road, St. Louis, MS
63144–2913.
49 R. Allman, Z. Kristallogr., 1969, 128, 115.
50 P. Milart and K. Stadnicka, Liebigs Ann. Chem., 1997, 2607.
51 (a) A. R. Katritsky, C. A. Ramsden, Z. Zakaria, S. H. Harlow
and S. H. Simonsen, J. Chem. Soc., Chem. Commun., 1979, 363;
(b) A. R. Katritsky, C. A. Ramsden, Z. Zakaria, S. H. Harlow and
S. H. Simonsen, J. Chem. Soc., Perkin Trans. 1, 1980, 1870.
52 P. Milart, D. Mucha and K. Stadnika, Liebigs Ann. Chem., 1995,
2049.
22 B. C. Perng, M. D. Newton, F. O. Raineri and H. L. Friedman,
J. Chem. Phys., 1996, 104, 7177.
23 G. Rahaut, T. Clark and T. Steinke, J. Am. Chem. Soc., 1993, 115,
9174.
24 R. B. de Alencastro, J. D. Da Motta Neto and M. C. Zerner, Int.
J. Quantum Chem., Quantum Chem. Symp., 1994, 28, 361.
25 J. Lipinski and W. Bartkowiak, J. Phys. Chem. A, 1997, 101, 2159.
26 W. Bartkowiak and J. Lipinski, J. Phys. Chem. A, 1998, 102, 5236.
27 M. J. S. Dewar, E. G. Zoebisch, E. F. Healy and J. J. P. Stewart,
J. Am. Chem. Soc., 1985, 107, 3902.
53 D. J. A. De Ridder, D. Heijdenrijk, H. Schenk, R. A. Dommisse,
G. L. Lemiere, J. A. Lepoivre and F. A. Alderweireldt, Acta
Crystallogr., Sect. C Cryst. Struct. Commun., 1990, 46, 2197.
54 J. Sielar, M. Pink and G. Zahn, Z. Anorg. Allg. Chem., 1994, 620,
743.
55 F. van Bolhuis and C. T. Kiers, Acta Crystallogr., Sect. B: Struct.
Crystallogr. Cryst. Chem., 1978, 34, 1015.
56 H.-O. Kalinowski, S. Berger, S. Braun, Carbon-13 NMR
Spectroscopy, John Wiley & Sons, New York, 1988; (a) p. 315;
(b) p. 310.
28 S. Miertus, E. Scrocco and Tomasi, J. Chem. Phys., 1981, 55, 117.
29 J. Lipinski, Int. J. Quantum Chem., 1988, 34, 423.
30 R. Zalesny, W. Bartkowiak, S. Styrcz and J. Leszczynski, J. Phys.
Chem. A, 2002, 106, 4032.
31 M. M. Karlesson and M. C. Zerner, J. Phys. Chem., 1992, 96, 6949.
32 J. Fabian, G. A. Rosquete and L. A. Montero-Cabrera, J. Mol.
Struct. (THEOCHEM), 1999, 469, 163.
33 S. R. Mente and M. Maroncelli, J. Phys. Chem. B, 1999, 103, 7704.
34 (a) T. Ishida and P. J. Rossky, J. Phys. Chem. A, 2001, 105, 558;
(b) J. Lobaugh and P. J. Rossky, J. Phys. Chem. A, 2000, 104, 899.
35 P. G. Jasien and L. L. Weber, J. Mol. Struct. (THEOCHEM), 2001,
572, 202.
36 S. Hogiu, J. Dreyer, M. Pfeiffer, K. W. Brzezinka and W. Werncke,
J. Raman Spectrosc., 2000, 31, 797.
57 K. Dimroth, C. Reichardt and A. Schweig, Liebigs Ann. Chem.,
1963, 669, 95.
58 See C. Wohlfarth, CRC Handbook of Chemistry and Physics,
eds. D. R. Lide and H. P. R. Frederiske, CRC Press, Boca Raton,
Florida, 1998, 78th edn., p. 6–139.
59 See C. Wohlfarth, CRC Handbook of Chemistry and Physics,
eds. D. R. Lide and H. P. R. Frederiske, CRC Press, Boca Raton,
Florida, 1998, 78th edn., p. 6–191.
60 (a) T. M. Krygowski and W. R. Fawcett, J. Am.Chem. Soc., 1975, 97,
2143; (b) T. M. Krygowski and W. R. Fawcett, Aust. J. Chem., 1975,
28, 2115; (c) T. M. Krygowski and W. R. Fawcett, Can. J. Chem.,
1976, 54, 3283; (d ) P. K. Wrona, T. M. Krygowski and Z. Galus,
J. Phys. Org. Chem., 1991, 4, 439.
61 M. J. Kamlet and R. W. Taft, J. Am. Chem. Soc., 1976, 98, 377–2886.
62 F. Hamzaoui and F. Baert, Acta Crystallogr., Sect. C Cryst. Struct.
Commun., 1994, 50, 757.
63 J. March, Advanced Organic Chemistry, Third Edition, John Wiley &
Sons, New York, 1988, p. 221.
37 S. A. Kovalenko, N. Eilers-Konig, T. A. Senyushkina and
N. P. Ernsting, J. Phys. Chem. A, 2001, 105, 4834.
38 M. C. Beard, G. M. Turner and C. A. Schmuttenmaer, J. Am. Chem.
Soc., 2000, 122, 11541.
39 X. H. Zhao, J. A. Burt, F. J. Knorr and J. L. McHale, J. Phys. Chem.
A, 2001, 105, 11110.
40 K. Stadnicka, P. Milart, A. Olech and P. K. Olszewski, J. Mol.
Struct., 2002, 604, 9.
64 N. Abe, T. Murafuji, Y. Sugihara and A. Kakehi, Heterocycles, 1995,
41, 2289.
65 A. Gavezotti, J. Am. Chem. Soc., 1983, 105, 5220.
66 A. Bondi, J. Phys. Chem., 1968, 68, 441.
41 Cambridge Structural Database, Cambridge Crystallographic Data
Centre, 12 Union Road, Cambridge, CB2 1EZ, UK.
42 See W. Liptay in Excited States, ed. E. C. Lim, Academic Press,
London, 1974, vol. 1, p. 203.
67 K. J. Shaw, P. W. Erhardt, A. A. Hagedorn III, C. A. Pease,
W. R. Ingebretson and J. R. Wiggins, J. Med. Chem., 1992, 35, 1267.
68 M. A. Kessler and O. S. Wolfbeis, Synthesis, 1988, 635.
69 T. C. Chadwick, Anal. Chem., 1974, 64, 1326.
J. Chem. Soc., Perkin Trans. 2, 2002, 1698–1707
1707