Solvent and solvent density effects on the spectral shifts and the bandwidths of the absorption and the resonance Raman spectra of phenol blue

Article abstract of DOI:10.1021/jp971310f

We have measured the absorption and the resonance Raman spectra of a solvatochromic dye, phenol blue, in liquid and supercritical solvents. We have found anomalous solvent dependence of the absorption bandwidth in liquid solvents: the width has apparently no correlation with the absorption peak shift. On the other hand, we have found good linear correlation between the absorption peak shift and the peak position of the resonance Raman bands (the C=N and the C=O stretching modes). The relative intensities of the Raman bands and the bandwidth of the C=N stretching mode also show correlation with the absorption peak shift. Incorporating these Raman data, the anomalous bandwidth of the absorption spectrum is explained by the change of the intramolecular vibrational contribution to the absorption bandwidth due to the electronic structure change by the solvent, which cancels the change of the solvent contribution. We have estimated the solvent reorganization energy assuming linear dependence of the intramolecular contribution on the absorption band center and neglecting the solvent reorganization energy in alkanes such as ethane and cyclohexane. In liquid solution, the estimated solvent reorganization energy is correlated fairly well with the absorption peak shift. Solvent dependence of the Raman bandwidth of the C=N stretching mode resembles the solvent dependence of the solvent reorganization energy estimated in this way. Relatively large bandwidths of both absorption and resonance Raman spectra have been observed in supercritical solvents compared with those in liquid solvents of a similar absorption peak shift. We interpreted this as due to the small refractive indices of the supercritical solvents relative to the liquid solvents; the large refractive indices of the liquid solvents only make the absorption peak shifts without broadening the absorption spectra.

Full text of DOI:10.1021/jp971310f

9050
J. Phys. Chem. A 1997, 101, 9050-9060
Solvent and Solvent Density Effects on the Spectral Shifts and the Bandwidths of the
Absorption and the Resonance Raman Spectra of Phenol Blue
T. Yamaguchi, Y. Kimura,* and N. Hirota*
Department of Chemistry, Graduate School of Science, Kyoto UniVersity, Kyoto 606-01, Japan
ReceiVed: April 16, 1997; In Final Form: September 18, 1997X
We have measured the absorption and the resonance Raman spectra of a solvatochromic dye, phenol blue, in
liquid and supercritical solvents. We have found anomalous solvent dependence of the absorption bandwidth
in liquid solvents: the width has apparently no correlation with the absorption peak shift. On the other hand,
we have found good linear correlation between the absorption peak shift and the peak position of the resonance
Raman bands (the CdN and the CdO stretching modes). The relative intensities of the Raman bands and
the bandwidth of the CdN stretching mode also show correlation with the absorption peak shift. Incorporating
these Raman data, the anomalous bandwidth of the absorption spectrum is explained by the change of the
intramolecular vibrational contribution to the absorption bandwidth due to the electronic structure change by
the solvent, which cancels the change of the solvent contribution. We have estimated the solvent reorganization
energy assuming linear dependence of the intramolecular contribution on the absorption band center and
neglecting the solvent reorganization energy in alkanes such as ethane and cyclohexane. In liquid solution,
the estimated solvent reorganization energy is correlated fairly well with the absorption peak shift. Solvent
dependence of the Raman bandwidth of the CdN stretching mode resembles the solvent dependence of the
solvent reorganization energy estimated in this way. Relatively large bandwidths of both absorption and
resonance Raman spectra have been observed in supercritical solvents compared with those in liquid solvents
of a similar absorption peak shift. We interpreted this as due to the small refractive indices of the supercritical
solvents relative to the liquid solvents; the large refractive indices of the liquid solvents only make the absorption
peak shifts without broadening the absorption spectra.
1. Introduction
Recently intensive studies were performed on the absorption
and the fluorescence band shapes in connection with electron-
transfer reactions to extract the reaction free energies and solvent
reorganization energies, which are important parameters in
electron-transfer reactions.2-9 The theoretical treatments on the
solvent fluctuation have been in many cases based on the
dielectric continuum model, as in the case of the spectral shift.
As for the solvent reorganization, only the reorientation of the
dipole moments of solvents is considered in this model.
Recently the importance of the reorientation of multipole
moments and translational motions was both experimentally10
and theoretically11 proposed. Theoretical studies based on the
distribution function theory were applied also to the problem
on the Stokes shifts and succeeded in explaining not only the
Stokes shift12 but also the dynamic Stokes shift.13
In chemical processes in solution, solvents play such an
important role that we can hardly neglect.1 For example, the
mean potential by solvent molecules changes the chemical
potential of a solute dissolved in solvent, and sometimes even
its electronic structure. The thermal fluctuation of the solvent
mean potential is also quite important, since it reveals itself in
the electron transfer reaction, the energy relaxation, the isomer-
ization, and so on. Spectroscopic methods are powerful tools
to study such solvent effects. The spectrum in solution is
different from that in the gas phase, not only in its peak position
but also in its width. The peak shift is the transition energy
shift due to the solvent mean potential. The broadening is
induced by the solvent fluctuation, and we can extract from
them the information on the static and dynamical fluctuation
of solvents by proper treatments.
Solvent effects on the absorption peak shift (solvatochromism)
and the fluorescence peak shift have long attracted much
attention. There have been many theoretical and experimental
studies performed to reveal the relation between the solvent
properties and the solvatochromic shifts.1 Most theoretical
treatments are based on the continuum model of the solvent,
and calculate the effect of the reaction field on the energy
difference between the ground and the excited states. These
theories sometimes succeed in correlating the shift with the
solvent properties such as the dielectric constant, but sometimes
fail. In application, the solvent shift itself is used as an indicator
of the solvent property as is represented by the π*-value or the
ET value.
However the experimentally observed absorption bandwidths
cannot be directly compared with the theoretical predictions on
the solvent fluctuation. Under the structureless spectra there
are vibrational structures obscured by the large solvent fluctua-
tion. It is in general difficult to separate the bandwidth into
the intramolecular vibrational and the intermolecular solvent
contributions.14 It is often assumed that the solvent does not
drastically change the intramolecular contribution, and the
intramolecular contribution is extracted from the spectra in
nonpolar solvents such as alkanes neglecting the solvent
contribution. However, since the solvent can alter the electronic
structure of the solute, the intramolecular contribution can in
principle change with solvents.
It is important to investigate the solvent effect on the
absorption bandwidth including the change in the electronic
structure of a solute. In this work we have studied both the
absorption spectrum and the resonance Raman spectrum of a
solvatochromic dye molecule dissolved in liquids and fluids
In contrast to the extensive studies on the peak shift, the study
on the absorption bandwidth is now under development.
X Abstract published in AdVance ACS Abstracts, November 15, 1997.
S1089-5639(97)01310-8 CCC: $14.00 © 1997 American Chemical Society

Resonance Raman Spectra of Phenol Blue
J. Phys. Chem. A, Vol. 101, No. 48, 1997 9051
SCHEME 1: Phenol Blue (PB): (a) Ground State
(Neutral Form), (b) Excited State (Charge-Separated
Form)a
ground and the excited states. Absorption bandwidth reflects
the transition energy fluctuation induced by the solvent, and
the intramolecular nuclear structure change on excitation which
is included as vibrational progressions. Raman Stokes shift is
the strength of intramolecular chemical bonds, which is
determined by the electronic structure of the ground state.
Resonance Raman intensity is a measure of the intramolecular
structural change on excitation (although excitation profile of
absolute cross section is desirable, we only measured the relative
intensities at single excitation energy for experimental reasons).
Raman bandwidth reflects the magnitude and modulation
frequency of vibrational frequency induced by solvent fluctua-
tion. These five quantities and the relationships among them
will be discussed in detail.
a The marks “a” and “b” are used to distinguish the position of the
isotope substitution.
from near the critical to the liquidlike densities using so-called
supercritical fluids. Resonance Raman study is considered to
be suitable for this purpose, since the vibrational modes strongly
coupled to the optical transition make large contributions to both
absorption and resonance Raman spectra.14 We have chosen
phenol blue (PB, Scheme 1) as a probe molecule. PB is a well-
known solvatochromic dye and its absorption band maximum
is often used as a solvent parameter.15 It has a well-separated
absorption band in the visible region (around 600 nm), whose
position moves from blue to red as the polarity or the hydrogen
donating ability of the solvents increases. PB is considered to
have a “neutral” form (Scheme 1a) in its ground state and a
“charge-separated” form (Scheme 1b) in its excited state. These
two forms constitute resonance structures, and the contribution
of the “charge-separated” form is expected to increase as the
solvent polarity or the hydrogen donating ability increases,
which was assured by semiempirical molecular orbital (MO)
calculations by Warshel et al.16 and by Zerner et al.17 As is
shown later, such a change in the electronic structure is also
observed in the resonance Raman spectra, which indicates that
the solvent dependence of intramolecular vibrational modes
should be taken into account.
We have measured the spectra in a variety of solvents, ranging
from protic solvents such as ethylene glycol to nonpolar solvents
such as ethane at the density near to its critical density. We
consider that the so-called supercritical fluids are suitable for
the study of the solvent effects, since we can change the density
continuously to gaseous one where the solvent effects are small.
In addition, the fluids near the gas-liquid critical points have
unique characters both in static and dynamic properties as
represented by the local density enhancement. Such characters
of the supercritical fluids currently attract interests in a wide
field of chemistry, both academic and industrial.18 Vibrational
spectroscopy has also been applied to the study of the solute-
solvent interaction in such a density region, but mainly by means
of IR spectroscopy.19 Most studies on fluids in the medium-
density region by the Raman spectroscopy have been restricted
to the solvent properties,20-23 or small-size solute molecules
such as hydrogen,24 iodomethane,25,26 or acetone,27 due to the
low solubilities of ordinary organic molecules and the low
sensitivity of the Raman spectroscopy, although the study on
naphthalene in CO2 has been done in the higher density region.28
In this work, we have developed a new high-pressure optical
cell which is suitable for the resonance Raman spectroscopy,
and studied how the solvent density affects the vibrational
structure of a large organic molecule.
This paper consists of six sections. In section 2 we briefly
review the theoretical basis to discuss the band shape of the
absorption spectrum and the intensity of the resonance Raman
spectrum, the Raman bandwidth, and the connection with
solvent properties. Section 3 is the experimental section.
Experimental results are given in section 4. A tentative
assignment based on the 15N isotope shift and vibrational spectra
of related species is proposed there. Section 5 is the discussion,
and concluding remarks are given in section 6.
2. Theoretical Basis of Analysis
2.1. Electronic Absorption Spectrum. We consider the
model where both intramolecular vibrational modes and the
solvent fluctuation are expressed as harmonic oscillators coupled
to the electronic transition from the ground to the excited state.29
The harmonic oscillator is characterized by three quantities: the
vibrational frequency ω, the dimensionless displacement ∆, and
the magnitude of the frictional force on it. Intramolecular modes
are usually high frequency (ω . kBT) and underdamped (low
friction), whereas solvent modes are low frequency (ω , kBT)
and overdamped (high friction). The displacement is also
expressed in terms of reorganization energy λ ) ω∆2/2. In
our model, the intramolecular vibrational mode is represented
by a single oscillator with the frequency of ωv, and the
reorganization energy of λv. In the Brownian oscillator model
the solvent fluctuation is described by a single (or several)
classical harmonic oscillator, whose properties are characterized
by two parameters: the solvent reorganization energy λs and
the correlation time Λ-1. We assume here that the relaxation
of the solvent is much slower than the optical dephasing time
(2kBTλs)-1/2 (static limit), since a large solvent reorganization
energy is expected for broad and structureless spectra30 (Figure
4). In such a model, the absorption spectrum is expressed as31
ꢀ(ω) ωIA(ω)
IA(ω) )
exp(-λv/ωv) (λv/ωv)n
{ω-(λs + ∆G + nωv)}2
4λskBT
∞
exp -
∑
[
]
n!
n)0
4πk Tλ
x
B
s
(1)
where ꢀ(ω) denotes the molar extinction coefficient, and IA(ω)
is the line shape function of the absorption spectrum. ∆G is
the energy gap between the bottoms of the ground and the
excited states. In the static limit, the band shape due to the
solvent contribution corresponds to the static distribution
function of the transition energy itself (which is described as a
Gaussian), and the whole spectrum is a vibrational progression
convoluted with the solvent contribution. The first-order and
the second-order cumulants, ω0 and σ2, respectively, are given
by
We have measured the absorption and the resonance Raman
spectra of PB, and analyzed comprehensively the solvent
dependence of the following five quantities: absorption peak
shift, absorption bandwidth, resonance Raman peak shift,
resonance Raman intensity and resonance Raman bandwidth.
Absorption peak is the averaged energy difference between the

9052 J. Phys. Chem. A, Vol. 101, No. 48, 1997
ω0 ) ∆G + λs + λv
Yamaguchi et al.
where σR is the scattering cross section, ωL and ωS are the
frequencies of the excitation laser and the scattered photon,
respectively. We have assumed that ωv of the excited state is
the same as that of the ground state. Equation 7, together with
eq 3, means that the intramolecular vibrational modes of strong
Raman bands (vibrational modes with large reorganization
energy) give large contributions to the bandwidth of the
absorption spectrum. We can also infer the structure of the
excited state from the resonance Raman spectrum, since the
nuclear displacement on the transition is related to the reorgan-
ization energy.
The band shape of the Raman spectrum is the Fourier
transform of the time correlation function of the Raman
scattering tensor, and its width is determined by the vibrational
and the rotational relaxation.36 We neglect here the rotational
relaxation, since our probe molecule PB is such a large molecule
that the rotational relaxation is expected to be much slower than
the vibrational one. The vibrational relaxation is usually
dominated by the pure dephasing. In the stochastic formalism
dephasing mechanism is divided into two, homogeneous and
inhomogeneous broadenings.
(2)
(3)
σ2 ) 2kBTλs + ωvλv
which stand for the center and the squared width of the
spectrum, respectively. The lifetime effect on the absorption
bandwidth is considered to be negligible in the present case.32
In our model there are only two parameters which characterize
the solvent effect, that is, the mean transition energy ω0 and
the solvent reorganization energy λs. Our principal interest is
how the solute-solvent and the solvent-solvent interactions
affect the observed spectra. In principle it is desirable to know
the relation between the solvent reorganization energy and the
intermolecular specific interactions at the molecular level.
However, the dielectric continuum model of solvents is often
useful for an order estimation, especially when the electronic
transition is accompanied with a large change in charge
distribution. Under such an approximation the absorption peak
shift and the solvent reorganization energy are, for example,
expressed as follows33
2
µ2g - µe n - 1
In the homogeneous case the rate of the frequency modulation
Λ is larger than the inverse of the root-mean-squared fluctuation
(fast modulation). In such a case the Raman band
shape is a Lorentzian, and fwhm (full width at half-maximum)
2
ꢀ - 1 n2 - 1
ꢀ + 2
2µg(µg - µe)
δω0 )
+
-
(
)
4πꢀ0a3 2n2 + 1
4πꢀ0a3
n2 + 2
2 1/2
|δω|
(4)
of the Raman band ∆ν1/2 is36
(µe - µg)2
4πꢀ0a3
ꢀ - 1 n2 - 1
ꢀ + 2
λs )
-
(5)
2
(
)
n2 + 2
∆ν1/2 ) 2 |δω| /Λ
(8)
The Raman band becomes narrower as the modulation becomes
faster, because the frequency fluctuation is averaged, which is
called “motional narrowing”.
In the inhomogeneous case where the solvent modulation is
slower than the inverse of the root-mean-squared fluctuation,
the Raman band shape is a Gaussian and its fwhm is36
Here the dipole moments of the ground and the excited states
are denoted as µg and µe, respectively. δω0 ) ω0(gas) -
ω0(solution) is the mean peak shift from the absorption peak
without solvent to that in solution. The radius of the dielectric
cavity (which corresponds to the distance of the center of solute
and solvent molecules) is a. ꢀ and n2 are the zero frequency
and optical frequency dielectric constants, respectively. The
first term of eq 4 corresponds to the contribution from the
solvent electronic polarizability, and the second term arises from
the nuclear reorientational motion. The former immediately
responds to the electronic transition, and does not show the
additional reorganization, whereas the reorganization energy of
the latter is given by eq 5. If we neglect the first term in eq 4,
the following linear relationship between the peak shift and the
reorganization energy holds,
2
∆ν1/2 ) 2 2 ln 2 |δω|
(9)
x
The solvent fluctuation is frozen here, and the observed band
shapes represent the distribution of the vibrational frequency,
as in the case of the absorption spectrum. In both cases, of
course, the Raman bandwidth is larger when the magnitude of
the frequency fluctuation is larger, although its dependence on
the frequency fluctuation is different.
δω0:λs ) -2µg:µe - µg
(6)
3. Experiment
2.2. Resonance Raman Spectrum. The resonance Raman
scattering is a two-photon process; a molecule is first excited
by the excitation photon, then the nuclear wave packet
propagates on the excited state potential surface, and finally
the molecule returns to the ground state by emitting the scattered
photon. The intensity of each vibrational band is determined
by two factors, that is, Albrecht A-term and B-term.34 The
former is the nuclear wave function overlap of the final state
with the wave packet of the initial state after the time
development on the excited state surface. The latter is the
nuclear coordinate dependence of the transition dipole moment.
Usually A-term dominates the spectrum under the resonance
condition. When the optical dephasing is fast, the time
development of the nuclear wave packet is determined by the
slope of the Franck-Condon region of the excited-state potential
surface. In such a case, the relative intensity is given by14,35
3.1. Materials. PB-14N (Aldrich 97%) is purified twice by
recrystallization from ethyl acetate-ligroine 1:1 mixture. Further
purification by alumina column37 did not alter the observed
spectra. We prepared two types of isotope substituted PB; PB-
15Na and PB-15Nb (for the position of the substituted nitrogen,
see Scheme 1). The preparation method is as follows (Scheme
2); Aniline N-methylation was performed by the reaction with
formaldehyde and sodium borohydride in sulfuric acid.38 Then
nitroso group was introduced at para position by the usual
method. p-Nitrosodimethylaniline was reduced by sodium
borohydride to p-Dimethylphenylenediamine,39 from which PB
was prepared by the method proposed by Vittum et al.40 Crude
dyes were purified by alumina column.37 Isotope sources,
aniline-15N and sodium nitrite-15N, were purchased from Isotec
Inc. (15N 99%). Sample production was assured from the
absorption, the resonance Raman and the IR spectra.
All liquid solvents were purchased from Nacalai Tesque and
used as received. CO2 (Sumitomo Seika, >99.99%), CF3H
σR(ωL,ωS) ωLω3Sωvλv
(7)

Resonance Raman Spectra of Phenol Blue
J. Phys. Chem. A, Vol. 101, No. 48, 1997 9053
SCHEME 2: Reaction Scheme of the Synthesis of PB
(Asahi Glass, >99.999%), N2O (Seitetsu Kagaku, >99.99%)
and C2H6 (Sumitomo Seika, >99%) were used without further
purification.
3.2. Apparatus. The 514.5 nm line of an argon ion laser
(Cyonics, 10mW) was used as the excitation light in resonance
Raman experiments. The scattering at 90° was collected and
focused onto the slit of a 60 cm single spectrograph (Engis,
Model 600). A holographic filter (Kaiser, Super Notch Plus)
was placed in front of the slit. Signal was detected by an
intensified silicon photodiode array (EG&G, Model 1421) and
stored in a personal computer. Spectral resolution was about 4
cm-1. Wavenumber calibration was performed by measuring
the Raman spectra of solvents with known frequencies (toluene
and acetone). Detector sensitivity was calibrated by measuring
the fluorescence of quinine.41 Sample concentrations were
controlled so as to obtain the best signal-to-noise ratios.
In measuring the resonance Raman band with broad band
absorption, it is quite important to reduce both the absorption
of the laser beam before the scattering position and the
reabsorption of the scattering light by the sample. They are
crucial especially in the measurement of supercritical fluid
solutions where the sample cell is required to stand for high
pressure. To accomplish this aim, we have designed a new high-
pressure optical cell of a simple design with which a number
of spectroscopic measurements such as absorption, fluorescence,
transient absorption, and Raman spectroscopies can be made
by changing the optical windows. In designing we referred to
the optical cell designed by Jodl and Holzapfel for the high-
pressure Raman measurement at low temperatures.42 Figure 1
shows the cross section of the high-pressure Raman cell used
for the measurement. The cell is made of titanium and can be
used at least up to 50 MPa. The optical cell is equipped with
four windows, three of them are used for the Raman measure-
ment: one for the laser input, another for the laser output, and
the last one for the scattering light. The laser beam was
introduced into the cell through a pinhole of 2 mm diameter
and reflected at 90° by a prism attached to the optical window
(BK-7). The distance between the edge of the prism and the
optical window for the scattering was set to be as small as
possible, in order for the laser beam to pass through the cell
nearest to the optical window. This reduces both the absorption
of the laser beam before the scattering position and the
reabsorption of the scattering light by the sample solution. The
aperture of the optical window for the scattering light was taken
to be large (1.9 of f-value). Special care was taken to get rid
of stray light and wall scattering as possible as we could.
In the measurement a proper amount of solid sample was
placed in the cell, and the cell was purged by a fluid in use.
The optical cell is connected to the reservoir of the high-pressure
fluid by a 1/16-in. capillary tube, and the fluid is introduced
from the reservoir. The sample was dissolved within the cell
by a magnetic stirrer. Temperature was controlled within (0.5K
by flowing thermostated water through the cell. The pressure
Figure 1. The cross section of the high-pressure optical cell. See details
in the text.
was measured by a stainless gage (Kyowa, PGM 500KH), and
the densities of the fluids were calculated by empirical equations
of states.43 All experiments were performed at the reduced
density Fr (density divided by the critical density Fc) larger than
1.0 due to the low solubility of PB in fluids.
All experiments on liquid solution spectra were performed
at room temperature. Spectra in supercritical fluids were
measured at 50 °C. We have also measured the temperature
effect on the CF3H solution at 10 °C. Absorption spectra were
measured by a multichannel photodetector (Otsuka Electronics,
MCPD-2000). Absorption spectra at pressures higher than 30
MPa were measured by using a different optical cell44 and a
spectrometer (Shimadzu, UV 2500PC). No concentration effect
was found in the observed spectra. Long time laser exposure
did not alter the observed Raman spectra.
4. Results
4.1. Absorption spectra. Absorption spectra of PB are
shown in Figure 2. The absorption peak wavelength λmax agrees
with the previously reported value.15,45,46 It can be easily
noticed that λmax shifts from blue to red with increasing the
solvent polarity or the hydrogen donating ability. On the other
hand, the absorption bandwidth is almost the same in liquid
solvents, which contradicts to the theoretical prediction (eq 6)
that λs increases in proportion to δω0. Interesting is the behavior
in fluids that the widths are much larger than those in liquids
although the peak shifts are smaller, as easily noticed by
comparing the spectra in fluids with that in cyclohexane (Figure
2, lower panel).
We have fitted these spectra to eq 1 to estimate the first and
the second-order cumulants, ω0 and σ,2 respectively. In the
fitting procedure, the intramolecular vibrational frequency ωv
was fixed at 1500 cm-1. Examples of the fittings are shown in
Figure 3. Although the deviation of the simulation is found on
the blue edge, we consider it is mainly due to the overlap with
the S0-S2 absorption, and that the bandwidth calculated from
this fitting does not deviate largely from the real value. In the
case of such structureless spectra, division of the bandwidth to
the intramolecular part and the solvent part is in general difficult.
We indeed found some uncertainty in the division. How-
ever, the summation of the two contributions stayed within (20
cm-1
.
The parameters ω0 and σ2 are correlated in Figure 4. In liquid
solvents virtually no correlation is found, although a slight

9054 J. Phys. Chem. A, Vol. 101, No. 48, 1997
Yamaguchi et al.
Figure 4. The relation between the first-order (ω0) and the second-
order (σ2) cumulants of absorption spectra. O, in liquid solvents; 0, in
CF3H (1.00∼2.57) at 50 °C; ×, in CF3H (2.20∼2.80) at 10 °C; +, in
CO2 (1.00∼1.90); 4, in N2O (1.00∼1.98); ], in ethane (1.00∼2.00).
In each supercritical fluid, ω0 is smaller at the higher density. Data
points in supercritical fluids at Fr ) 1.60 are united by the solid line.
The dotted line is explained in section 5.
Figure 2. Absorption spectra of PB. Upper: spectra in liquid solvents
at room temperature; from left to right, in ethylene glycol, methanol,
dimethyl sulfoxide (DMSO), chloroform, acetonitrile, benzene, ethyl
acetate, carbontetrachloride, cyclohexane. Lower: spectra in super-
critical fluids at 323.2 K. From left to right, in CF3H (2.00), CO2 (1.90),
N2O (1.98), ethane (2.00). Their reduced densities are written in
parentheses. The spectrum in cyclohexane is drawn by the dotted curve
for reference. Excitation energy for the resonance Raman measurement
is indicated by the vertical lines.
Figure 5. Resonance Raman spectra in liquid solvents; from upper to
lower, in ethylene glycol, methanol, DMSO, chloroform, acetonitrile,
benzene, ethyl acetate, carbontetrachloride, cyclohexane. Solvent bands
are marked with *.
means that σ2 is really larger in fluids than in liquids with similar
values of ω0 at the same temperature.
The absorption bandwidth contains both the intramolecular
and the solvent contributions, as mentioned before. When the
electronic structure depends on solvents, the intramolecular
contribution is also expected to vary with solvents. Such a
change should be reflected in the resonance Raman spectra, so
we will return to the problem of the absorption bandwidth in
section 5 after the examination of the resonance Raman spectra.
4.2. Resonance Raman Spectra. Resonance Raman spectra
of PB in liquid solvents and in supercritical fluids are shown in
Figures 5 and 6, respectively. Solvent bands in the spectrum
of the ethylene glycol solution were subtracted since a strong
solvent band overlaps with the CdN stretching band of PB
(described below). The spectra drastically change with solvents
and solvent densities. No fluorescence due to PB was observed
at least around 560 nm, where the Raman bands were ob-
served.47
We first try to assign each vibrational mode to discuss the
origin of these solvent dependent peak shifts and bandwidths,
since this is the first study on the vibrational spectra of PB. For
assignment we prepared two 15N substituted species, PB-15Na
and PB-15Nb (Scheme 1). Their resonance Raman spectra in
carbontetrachloride, acetonitrile and methanol are shown in
Figure 7 together with those of PB-14N. It is easily noticed
that the Raman band around 1500 cm-1 (whose position strongly
depends on solvents) shows a large isotope shift in PB-15Na,
which clearly indicates that this band is the CdN stretching
mode. The strong band around 1640 cm-1, whose frequency
Figure 3. Examples of the fitting of the absorption spectra; left: in
methanol, right: in CF3H (2.00). The observed and the fitted spectra
are drawn by the solid and the broken lines, respectively.
increase of σ2 with a decrease of ω0 is found. In supercritical
fluids, however, σ2 is significantly larger than those in liquid
solvents. Furthermore, σ2 decreases almost linearly as the
solvent density increases. All of these phenomena violate the
linear relationship between the peak shift and the bandwidth
(eq 6). However, comparing the spectra in various supercritical
fluids at the same reduced density, σ2 increases with decreasing
ω0 as is expected from the solvent polarity. It is possible that
the different thermal energy (50 °C) in supercritical fluids may
bring about apparent broadening of the absorption spectrum,
since the solvent fluctuation contribute to the bandwidth as
λskBT(eq 3). To check this effect, we have measured the
spectrum in CF3H at 10 °C. Lowering the temperature of course
sharpened the spectra, but the bandwidths of the CF3H solutions
at 10 °C were still larger than those of liquid solutions. This

Resonance Raman Spectra of Phenol Blue
J. Phys. Chem. A, Vol. 101, No. 48, 1997 9055
TABLE 1
frequency in CCl4 (cm-1
)
PB-14N
PB-15Na
PB-15Nb
intensitya
assignment
1641
1596
1568
1550
1505
∼1480
1445
1412
1371
∼1355
1315
1247
1235
1641
1593
1566
1546
1488
c
1443
1411
1365
∼1355
1313
1245
1229
1641
1593
1566
1549
1505
∼1480
1542
1409
1367
1339
1311
1245
1232
s
CdO str
m
m
w
vs
sh
m
w
m
m
w
w
m
DMA 8ab
CdC sym str
CdNa str
δs(CH3)
?d δa(CH3)
ring-Nb str
?d ring-Na str
a vs, very strong; s, strong; m, medium; w, weak; sh, shoulder. b The
mode that corresponds to the 8a mode of DMA. c Not observed.
d Tentative assignments.
polarity, as is expected by the electronic structure changes from
the neutral to the charge separated forms (Scheme 1).
Between the CdO and the CdC stretching modes a band of
medium strength is found (near 1600 cm-1) which shows a slight
red shift with decreasing the electronic transition energy. We
assigned this band to the 8a vibration of the dimethylaniline
(DMA) part of the PB (hereafter called 8a), because of the
similarity of this frequency to the 8a mode of DMA (1605
cm-1).48 Since the DMA part of the PB in its charge separated
form (Scheme 1) resembles the DMA cation radical, the solvent
shift also follows the change of the vibrational shift from DMA
to its radical cation. In fact, the frequency of the 8a mode in
Figure 6. Resonance Raman spectra of PB in various solvent fluids
and at various densities. (a) Spectra at Fr ) 1.60 (except for CF3H
solution at Fr ) 1.61) and 323.2 K. From upper to lower, in C2H6,
N2O, CO2, CF3H. (b) Spectra in CF3H at various densities. From top
to bottom: Fr ) 1.03, 1.22, 1.41, 1.61, 1.80, and 2.00 at 50 °C.
DMA cation radical is reported as 1568 cm-1 49
whose shift
,
relative to the band of DMA is consistent with our experimen-
tally observed shift of the corresponding mode.
Only a small change is found on the substitution of amino-
nitrogen (PB-15Nb, Figure 7). Most significant change was
found in the band near 1370 cm-1 in PB-14N, which consists
of sharp and broad components. In PB-15Nb, the broad
component moves to the lower frequency and the sharp one
remains. This broad component is considered to be the Nb-
ring stretch. The corresponding band of DMA lies near, 1348
cm-1 48
which supports this assignment. Our assignments,
,
including tentative ones for bands not mentioned in the text,
are summarized in Table 1.
In the following, we will mainly discuss the peak position
and bandwidth of the CdO and the CdN stretching modes
which show large solvent dependence. The peaks, the widths
and the intensities of Raman bands were determined by fitting
the bands to a Gaussian after the baseline (determined from
polynomial fit) was subtracted. Examples of the fitting are
shown in Figure 8. The low-frequency side of the CdN
stretching mode was omitted in the fitting, since a shoulder was
found here. In methanol, DMSO and ethylene glycol we could
not fit the whole band shapes of the CdO stretching mode by
a single Gaussian function, and we used the summation of two
Gaussian functions with the same peak frequency.
Figure 9 shows the correlation between the absorption band
center and the vibrational frequency shifts of the CdN stretching
and the CdO stretching modes. Both bands show large shifts
with increasing the solvent polarity or the hydrogen donating
ability in a variety of solvents, and with increasing solvent
density in supercritical fluids. Good correlation is found in both
bands, which indicates that the ground-state electronic structure
has some correlation with the transition energy. The density
dependence in supercritical fluids is also in correlation with the
absorption band center as in the solvent dependence within liquid
Figure 7. 15N isotope shifts of Raman bands. Each spectrum is obtained
in (a) carbontetrachloride, (b) acetonitrile, and (c) methanol, respec-
tively. From upper to lower, PB-14N, PB-15Na, PB-15Nb.
shifts largely, is assigned to the carbonyl CdO stretching mode,
since we have observed a strong band at nearly the same position
in the FT-IR spectra measured in KBr. A small band around
1560 cm-1, which also shows a low-frequency shift with
increasing the solvent polarity, is assigned to the CdC sym-
metric stretch of the quinone ring. The frequencies of the above-
mentioned three modes decrease with increasing the solvent

9056 J. Phys. Chem. A, Vol. 101, No. 48, 1997
Yamaguchi et al.
Figure 10. The bandwidths of the CdN stretching mode against the
absorption band center. O, in liquid solvents; 0, PB-15Na in methanol;
acetonitrile and carbontetrachloride; ], in CF3H (1.00∼2.57); ×, in
CO2 (1.00∼1.90); 4, in N2O (1.00∼1.98); +, in ethane (1.20∼2.00).
Figure 8. Examples of the fitting of CdN Raman bands: upper, in
methanol; lower, in CF3H (1.61). The observed and the fitted spectra
are drawn by the solid and the dotted lines, respectively. The residue
of the fit is drawn by the broken lines.
Figure 11. The relative intensity ratio of the CdO stretching and the
8a modes against the band center. Longitudinal axis is drawn in
logarithmic scale. O, in liquid solvents; 0, in CF3H (1.03∼2.00); ],
in CO2 (1.00∼1.90); 4, in N2O (1.00∼1.98); 3, in ethane (1.40∼2.00).
15Na. However, the width is still large in the case of PB-15Na,
where the effect of the overlap is small. Furthermore, the width
is small in nonpolar solvents (such as cyclohexane and CCl4),
and the band does not seem to suffer from the overlap with
other bands. Therefore, we conclude that the width of the CdN
stretching band is indeed large, although we may slightly
overestimate the bandwidth of PB-14N. As is shown in Figure
10, a fairly good linear relationship between the absorption band
center and the CdN bandwidth is found, if we compare the
bandwidths in liquid solutions only. This suggests that the
origin of the broadening of the bandwidth is the same as that
of the peak shift of the absorption band. This point is discussed
in detail in section 5. It is to be noted that the bandwidths are
larger in fluids than in liquid solvents if compared at the same
transition energy, as found in the absorption bandwidth.
The intensity of the resonance Raman band is an important
quantity in considering the bandwidth of the absorption spec-
trum. It can be easily noticed from Figures 5 and 6 that the
relative intensity of each band also depends on solvents. In
particular the CdO stretching band grows as the absorption peak
shifts to red. Figure 11 shows the relative intensity of the CdO
stretch and the 8a against the absorption band center. We have
chosen these bands, because they are strong in all solvents,
separated from other resonance and solvent bands, and so close
to each other that reabsorption correction is not required. In
the plot, we show the relative intensity by a log scale. The
relative intensity in all solvents used here has good correlation
with the absorption band center, although the results of the CF3H
solution have a somewhat large deviation.
Figure 9. Raman band shifts of the CdN stretching and the CdO
stretching modes. O, CdN in liquid solvents; ], CdN in CF3H
(1.03∼2.00); 0, CdN in CO2 (1.00∼1.90); 3, CdN in N2O (1.00∼1.98);
4, CdN in ethane (1.00∼2.00); b,CdO in liquid solvents; [, CdO
in CF3H; 9, CdO in CO2; 1, CdO in N2O; 2, CdO in ethane.
solvents. This means that the solvent density effect on the
frequency of the CdN stretching and the CdO stretching modes
can be explained in terms of the effect on the transition energy,
as in the case of the solvent effect in liquid solvents.
As is shown in Figures 5 and 6, the bandwidth of the CdN
stretching band is not only relatively large in comparison with
other bands but also shows large solvent dependence. Figure
10 shows the plot of the bandwidth of the CdN stretching
modes obtained by the spectral fitting against the absorption
band center. The bandwidth of the CdO stretching mode was
almost constant irrespective of the solvent, and are smaller than
that of the CdN stretching mode. We should at first examine
the possibility that the broad band we assigned to the CdN
stretching mode is actually composed of several bands due to
other vibrational modes whose widths are not broad. In fact
the obtained bandwidths for PB-14N are larger than those for
PB-15Na in acetonitrile and methanol. We consider this is due
to the shoulder peak observed in the case of PB-14N, which is
totally hidden by the CdN stretching band in the case of PB-
The Raman intensity consists of two contributions, one is
the nuclear coordinate displacement on excitation (Albrecht

Resonance Raman Spectra of Phenol Blue
J. Phys. Chem. A, Vol. 101, No. 48, 1997 9057
A-term), and the other is the vibronic interaction (B-term). If
the former dominates the band intensity, which is usual in the
resonance condition, the change in the relative intensity means
that the intramolecular vibrational reorganization energy changes
with the solvent. To check the assumption, we have measured
the spectra of several liquid solutions excited by a krypton ion
laser (647 nm).55 The relative intensity of the CdO band was
slightly smaller in the Kr+ excited spectra. Since the excitation
energy is fixed in our experiments, the relative position in the
absorption spectrum moves to red as the absorption peak moves
blue. Even if we admit that the change in the relative intensity
shown in Figure 11 is partly explained from the B-term
contribution, the solvent dependence was much larger than the
excitation energy dependence. We also measured the depolar-
ization ratio in CCl4 solution. The values were close to 1/3,
which is the characteristic value of the A-term enhancement.
Therefore the B-term contribution is considered to be minor in
this case.
an ab initio calculation to NMA(H2O)3 cluster and obtained the
spectrum close to the experimentally observed one in water.53
We consider that vibrational frequency shifts due to the
electronic structure change are not so rare, and that our
observation is another example.
The problem is the way the solute-solvent interaction affects
the spectra. It is well-known that the absorption peak shift of
PB is not correlated with the dielectric constants, and depends
on both the solvent polarity and the hydrogen donating ability.15
Although both effects lower the transition energy, their mech-
anisms are different from each other; the former works on the
solute as a whole such as a reaction field, whereas the latter
works almost exclusively on the oxygen atom. If the hydrogen
donating ability is reflected on the electronic structure, we should
be able to observe it in the resonance Raman spectra. For
example, we can expect the peak shift of the CdO stretching
band to be larger in protic solvents. However, no meaningful
difference can be found between the effects of polar and protic
solvents on the peak shift and the relative intensity of the Raman
spectra (Figures 5, 6, and 9). This means that whatever the
interaction brings about the shift of the absorption spectrum,
the structural changes for both the ground and the excited states
are characterized by only one parameter, ω0. We consider that
this behavior is originated in the electronic structures of PB,
which is easily changed by its environment as expressed by
the resonance structure, although further theoretical studies are
needed to ascertain this point.
The next question is how the structural changes discussed
above manifests itself in the absorption bandwidth. The
resonance Raman intensity is an important quantity in this
regard, since the stronger Raman bands contributes more to the
absorption bandwidth as is represented by eqs 3 and 7. As
shown in Figure 11, the relative intensity ratio of the 8a and
the CdO stretching modes strongly depends on solvents, which
indicates the possible change of the intramolecular contribution
to the absorption spectrum. Unfortunately we could not
experimentally determine the relative values of λv among
different solvents, since we could not obtain the excitation
profiles of the absolute cross section with our equipment. A
simple resonance consideration, however, predicts that λv
decreases with decreasing ω0, because the ground and the
excited states become more similar to each other as they mix.
Figueras found that the transition dipole moment increases as
the transition energy decreases.15 It means that the overlap of
wave functions of both states increases, which also supports
our consideration.
5. Discussion
We summarize the experimental results briefly as follows.
(a) The center of the absorption band ω0 does not correlate with
the bandwidth σ2 (Figure 4), in contradiction to the linear
correlation between ω0 and λs predicted by the dielectric
continuum model (eq 6). (b) The peak shift and the relative
intensity of the resonance Raman band correlate well with ω0
(Figures 9 and 11). (c) The widths of the absorption spectrum
and the CdN stretching band in supercritical fluids do not
correlate with those in liquid solvents.
To discuss the origin of the above-mentioned phenomena,
we first investigate the electronic structures of the ground and
the excited states, since they determine the vibrational reorgan-
ization energy λv, which is another source of absorption
bandwidth (eq 3). Resonance Raman spectra can be used for
this purpose, and we will discuss it in section 5.1. The Raman
bandwidth is discussed in connection with the absorption
bandwidths in 5.2. At first sight, the result (c) seems to reflect
the character of the density fluctuation in fluids in the medium-
density region, which has been extensively discussed in the
absorption and fluorescence peak shifts.18 However, our result
can be explained without invoking such an enhanced fluctuation,
based on the discussion in section 5.1 and 5.2. Finally, the
solvent density dependence of the solvent fluctuation will be
discussed in 5.3 in connection with its molecular origins.
5.1. The Origin of the Anomalous Absorption Bandwidths
in Liquids. The large solvent dependence of the resonance
Raman spectra of PB can be considered to reflect the electronic
structure change of PB induced by solvents. Provided that the
resonance between the neutral and the charge separated forms
(Scheme 1) is enhanced in polar solvent as mentioned in section
1, it will weaken the CdN, CdO, and CdC bonds, which can
explain their solvent shifts. In principle, the vibrational
frequency of polar bonds can shift in polar solvents, even if
the electronic structure does not change.50 However, considering
the large magnitude of the shift, especially that of the CdN
stretching mode, and the change of the relative Raman intensi-
ties, the electronic structure change of the solute is likely to be
the main cause for the shift.
On the basis of the above argument, we have tried to divide
σ2 into the contributions of λv and λs by simply assuming that
ωvλv is a linearly increasing function of ω0, and that λs is nearly
equal to zero in alkanes such as cyclohexane and ethane. The
assumption is not necessarily strict, since the solvent density
may also contribute to the reorganization as is discussed later.
However the contribution of the density is expected to be smaller
than that of the electric multipoles. The dotted line in Figure
4 shows the relation of ωvλv with ω0 obtained under the above
assumptions, explicitly given by
ωvλv ) 400ω0 - 5 200 000
(10)
The vibrational frequency shifts due to the electronic structure
change were in fact found previously in other species.51,52 For
example, Spiro et al. showed that in the case of N-methyl-
acetamide (NMA) the frequency and strength of the resonance
Raman bands assigned to the C-N and the CdO stretching
modes change with solvents.51 They explained the behavior
using the Hu¨ckel MO method. Recently Markham et al. applied
In drawing this line, the coefficients are chosen to give a parallel
line to the line which connects σ2 in ethane at Fr ) 1.0 and
cyclohexane, and to give a small portion of λs (ca. 100 cm-1
)
in these solvents. The value of λs in various solvents can be
calculated by subtracting ωvλv estimated by eq 10 from σ2 in
eq 3. The estimated λs is shown against ω0 in Figure 12. We
obtain a fairly good correlation between ω0 and λs estimated

9058 J. Phys. Chem. A, Vol. 101, No. 48, 1997
Yamaguchi et al.
broadened by the slow relaxation (eq 8). However, the
correlation with the solvent viscosity is not so good in the
present case. We consider that the main reason is δω, although
the solvent dependence may be enhanced by the motional
narrowing.
In general the solvent fluctuation which appears in vibrational
spectra is different from the fluctuation which determines the
absorption bandwidth. The vibrational bandwidth comes mainly
from the collisional interaction, whereas the absorption band-
width stems from the long-range electrostatic interaction.
However, we cannot explain the anomalous Raman bandwidth
of the CdN stretching band from the collisional interaction.
The hydrogen bonding can broaden the bandwidth, for Na is
another possible hydrogen accepting site. But the effect should
be larger on the CdO stretching band. Considering the
similarity between Figures 10 and 12, the main part of the CdN
bandwidth is likely to arise from the fluctuation of the electronic
transition energy. We in fact found that the Raman Stokes shift
of the CdN stretching mode depends on the excitation energy
in polar or protic solvents, which can be the evidence of the
correlation of the both fluctuation.54,55
Figure 12. An estimate of the solvent reorganization energy. All marks
are the same as Figure 4.
in this way, and the order of λs in liquid solvents is comparable
to the values reported for other solvatochromic dye systems.10
The tendency of the larger λs with the larger absorption peak
shift is also reasonable, since generally a large-energy difference
between two states accompanies large rearrangements or re-
orientational dynamics of the solvents. The values of λs for
supercritical fluids still deviates from this correlation, which is
the same situation as is observed in ∆ν1/2 for the CdN stretching
mode. We will discuss this point in section 5.3. It is interesting
that almost a linear correlation between ω0 and λs is obtained
as is expected from the dielectric continuum model given by
eq 6. If we simply apply this relation, µe/µg is estimated as
3.4. However, it is to be noted that the estimated λs in liquid
solvents are not necessarily correlated well with the dielectric
constants as predicted by eq 5, as is the case of the spectral
shift.
5.2. The Width of the CdN Raman Band. The bandwidth
of the Raman spectrum is determined by the fluctuation of the
vibrational frequency (eqs 8 and 9), which stems from the
thermal motions of solvents. We found that the bandwidth of
the CdN stretching mode is extraordinarily large and shows
peculiar solvent dependence. It will be interesting to investigate
the origin of the width in connection with the absorption
bandwidth, which also originates from the solvent fluctuation.
The bandwidth is determined by two quantities, the magnitude
and the speed of the vibrational frequency modulation (eqs 8
and 9). The large bandwidth of the CdN stretching mode
means that either δω is larger or Λ is smaller for the CdN
bond. We think the former more probable, since the very large
solvent dependence of the frequency of the CdN stretching
mode suggests that its frequency easily changes with environ-
ment. As is shown in Figure 9, the transition energy shift and
the CdN peak shift are strongly correlated. This means that
the fluctuation of the transition energy should lead to the
fluctuation of the CdN frequency. The former appears in the
absorption bandwidth as λs, and the latter contributes to the
Raman bandwidth of the CdN stretching band as discussed
above. It is quite natural that there is a good correlation between
λs and ∆ν1/2 within this argument, and this is confirmed by the
similarity between Figures 10 and 12. From the same consid-
eration, the width of the CdO stretching band is also expected
to be large. However, the width of the CdO stretching band
is not so large as that of the CdN stretching mode, although it
is slightly larger than those of other bands. We consider that it
is due to the smaller solvent dependence of the vibrational
frequency of the CdO stretching mode.54
5.3. The Origins of the Solvent Induced Absorption Peak
Shift and the Bandwidth. The remaining problem is the
relatively large solvent fluctuation in fluids compared with that
in liquid solvents, as is represented in Figures 10 and 12. To
begin with, we consider the reason the nonpolar solvent
cyclohexane gives the absorption peak shift as large as CF3H.
The shift of ω0 in cyclohexane from that of isolated PB vapor
is about 2000 cm-1, if we estimate the ω0 of isolated PB from
the zero-density extrapolation of the density dependence in
ethane. Such a large shift is not limited to the case of PB.
Solvatochromism of 2-nitroanisole (π*-value) in supercritical
fluids has been well-studied,56 and cyclohexane is known to
give a large shift as supercritical solvents whose dielectric
constants are much larger than that of cyclohexane. This
indicates that the solute specific interaction is not the reason
for the large absorption peak shift of PB in cyclohexane.
In recent years, it has been demonstrated that a substantial
amount of the solvation energy arises from multipole interactions
even in nondipolar solvents. Maroncelli et al. studied the
fluorescence Stokes shift of Coumarin 153, and they showed
that the solvent with a large quadrupole or an octapole moment
gives rise to a large fluorescence Stokes shift even if it has no
dipole moment.10 Cyclohexane, however, does not have any
large higher order multipole moment, and Maroncelli et al.
classified it to a genuine nonpolar solvent. The large solvent
shift of PB in cyclohexane does not arise from the reorientation
of the solvent.
According to eq 4, the transition energy shift is divided into
the contributions from the electronic polarizability and the
solvent reorientation. The former is often neglected in the
discussion on the liquid solution since the solvent dependence
of the polarizability is much smaller than the multipole
interaction. On the other hand, supercritical solvents as a whole
have small polarizabilities compared with liquid solvents57 (their
low critical temperature means the weak solvent-solvent
attractive interaction, substantial amount of which arises from
the dispersion force). We consider the absorption peak shift in
CF3H is for the most part due to the solvent reorientation,
whereas that in cyclohexane arises from the solvent polariz-
ability.
Motional narrowing can also modify the solvent dependence
of the Raman bandwidth. Since the solvent with large band
shift, such as ethylene glycol, has larger viscosity than nonpolar
solvents such as cyclohexane, the Raman spectra may be
This difference in the origin of the absorption peak shift
between cyclohexane and CF3H can also explain their difference
in the bandwidth of both the absorption and the resonance
Raman spectra. According to the dielectric continuum theory,

Resonance Raman Spectra of Phenol Blue
J. Phys. Chem. A, Vol. 101, No. 48, 1997 9059
the fluctuation due to the dipolar reorientation accompanies the
reorganization energy, while the polarization contribution does
not. In our assumption, the absorption peak shift in CF3H is
for the most part due to the solvent reorientation, whereas the
absorption peak shift in cyclohexane arises from the solvent
polarizability. The former has the corresponding reorganization
energy whereas the latter not, which can cause the difference
in the solvent reorganization energy, i.e., the absorption
bandwidth.
which brings the transition energy shift without fluctuation.
Solvent density dependence of the fluctuation of the number of
solvents around a solute should be clarified in order to realize
the density dependence of bandwidths in fluids.
Acknowledgment. We thank Mr. F. Amita (Kyoto Univer-
sity) for constructing the optical cell for the resonance Raman
measurement in high-pressure fluids. We are grateful to Dr. Y.
Yoshimura (Kyoto University) for his help in the measurement
of the absorption spectra at the pressure higher than 30 MPa.
Although we consider that the difference in solvent polariz-
ability is the principal reason for the difference between liquid
and supercritical solvents, there still remains a question if the
electronic polarizability does not contribute to the reorganization
energy in supercritical fluids. At the molecular level, the
absorption band shift is proportional to the solvent polarization
multiplied by the solvent density within the range of the
electrostatic interaction (we neglected here the effect of the local
field). If the solvent density around the solute fluctuates, a
genuine nonpolar solvent can also cause the corresponding
transition energy fluctuation.58,59 Since the fluctuation of the
bulk density is large in fluids near the critical point, the density
fluctuation around the solute also seems large in the medium-
density region. On the other hand, the density fluctuation seems
small in the high-density region since molecules are closely
packed here. Although the solvent density fluctuation around
a solute is sometimes discussed in terms of compressibility (the
fluctuation of the bulk solvent density), we know only a few
studies in which it was actually examined in supercritical fluids.
These results indicate that the density fluctuation around a solute
is not always proportional to the solvent compressibility.60-62
Solvent density fluctuation around a solute is a remaining
problem to be studied in the future.
References and Notes
(1) See, e.g., Reichardt, C. SolVents and SolVent effects in Organic
Chemistry; VCH: Weinheim, 1988.
(2) (a) Gould, I. R.; Noukakis, D.; Goodman, J. L.; Young, R. H.;
Farid, S. J. Am. Chem. Soc. 1993, 115, 3830. (b) Gould, I. R.; Noukakis,
D.; Gomez-Jahn, L.; Young, R. H.; Goodman, J. L.; Farid, S. Chem. Phys.
1993, 176, 439.
(3) Kjaer, A. M.; Ulstrup, J. J. Am. Chem. Soc. 1987, 109, 1934.
(4) Corte´s, J.; Heitele, H.; Jortner, J. J. Phys. Chem. 1994, 98, 2527.
(5) (a) Reid, P. J.; Barbara, P. F. J. Phys. Chem. 1995, 99, 3554. (b)
Johnson, A. E.; Levinger, N. E.; Jarzeba, W.; Schlief, R. E.; Kliner, D. A.
V.; Barbara, P. F. Chem. Phys. 1993, 176, 555. (c) Walker, G. C.; Akesson,
E.; Johnson, A. E.; Jevinger, N. E.; Barbara, P. F. J. Phys. Chem. 1992,
96, 3728. (d) Akesson, E.; Walker, G. C.; Barbara, P. F. J. Chem. Phys.
1991, 95, 4188.
(6) (a) Tominaga, K.; Kliner, D. A. V.; Johnson, A. E.; Levinger, N.
E.; Barbara, P. F.; J. Chem. Phys. 1993, 98, 1228. (b) Walker, G. C.;
Barbara, P. F.; Doorn, S. K.; Dong, Y.; Hupp, J. T. J. Phys. Chem. 1991,
95, 5712.
(7) (a) Wynne, K.; Gailli, C.; Hochstrasser, R. M. J. Chem. Phys. 1994,
100, 4797. (b) Wynne, K; Reid, G. D.; Hochstrasser, R. M. J. Chem. Phys.
1996, 105, 2287.
(8) (a) Markel, F.; Ferris, N. S.; Gould, I. R.; Myers, A. B. J. Am.
Chem. Soc. 1992, 114, 6208. (b) Kulinowski, K.; Gould, I. R.; Myers, A.
B. J. Phys. Chem. 1995, 99, 9017. (c) Kulinowski, K.; Gould, I. R.; Ferris,
N. S.; Myers, A. B. J. Phys. Chem. 1995, 99, 17715.
(9) Zerg. Y.; Zimmt, M. B. J. Phys. Chem. 1992, 96, 8395.
(10) Raynords, L.; Gerdecki, J. A.; Frankland, S. J. V.; Horng, M. L.;
Maroncelli, M. J. Phys. Chem. 1996, 100, 10337.
6. Conclusion
(11) Perng, B.-C.; Newton, M. D.; Raineri, F. O.; Friedman, H. L. J.
Chem. Phys. 1996, 104, 7153.
We have for the first time measured the resonance Raman
spectra of PB in various solvents, ranging from protic solvents
such as ethylene glycol to nonpolar supercritical solvents such
as ethane near their critical densities. We have found that the
CdN and the CdO stretching bands show large solvent shifts,
reflecting the solvent dependence of the electronic structures
of PB. Their shifts are correlated well with the absorption peak
shift, irrespective of the solute-solvent interaction specificity.
The relative intensities of the Raman bands also change with
solvents and are correlated with the absorption peak shift. Both
relations stand for the correlation of the ground and the excited
electronic structures with the transition energy shift. We further
found that the bandwidth of the CdN stretching mode broadened
with the red shift of the absorption. The width is, however,
larger in supercritical solvents compared with that in liquid
solvents at the same absorption peak.
We have also measured the bandwidth of the absorption
spectrum, the solvent dependence of which is at first sight very
confusing. In liquid solvents it apparently has no correlation
with the absorption peak shift. In fluids the width is larger than
in liquid solvents, and it increases with a decrease of the solvent
density. This anomaly is ascribed to the change of the
intramolecular vibrational reorganization energy due to solvents,
since the electronic structure changes with solvents. We
subtracted the intramolecular contribution from the absorption
bandwidth by simply assuming linear dependence of the
intramolecular reorganization energy on the electronic transition
energy. The estimated solvent reorganization energies are
reasonable within liquid solvents and also correlated with the
Raman bandwidth of the CdN stretch mode. The narrow
bandwidth in liquids is originated in the large polarizabilities
(12) Perng, B.-C.; Newton, M. D.; Raineri, F. O.; Friedman, H. L. J.
Chem. Phys. 1996, 104, 7177.
(13) Hirata, F.; Munakata, T.; Raineri, F.; Friedman, H. L. J. Mol. Liq.
1995, 65/66, 15.
(14) (a) Myers, A. B. Chem. Phys. 1994, 180, 215; (b) Myers, A. B.
Chem. ReV. (Washington, D.C.) 1996, 96, 911.
(15) Figueras, J. J. Am. Chem. Soc. 1971, 93, 3255.
(16) Luzhkov, V.; Warshel, A. J. Am. Chem. Soc. 1991, 113, 4491.
(17) Karelson, M. M.; Zerner, M. C. J. Phys. Chem. 1992, 96, 6949.
(18) See, e.g., Savage, P. E.; Gopalan, S.; Mizan, T. I.; Martino, C. J.;
Brock, E. E. AIChE J. 1995, 41, 1723.
(19) See, e.g., Poliakoff, M.; Howdle, S. M.; Karaian, S. G. Angew.
Chem., Int. Ed. Engl. 1995, 34, 1275.
(20) Garrabos, Y.; Chandrasekharan, V.; Echargui, M. A.; Marsault-
Herail, F. Chem. Phys. Lett. 1988, 160, 697.
(21) Ben-Amtoz, D.; LaPlant, F.; Shea, D.; Gardecki, J.; List, D. ACS
Syp. Ser. 1992, 488, 18.
(22) Salmoun, F.; Dubessy, J.; Garrabos, Y.; Marsault-Herail, F. J.
Raman. Spectrosc. 1994, 25, 281.
(23) Okazaki, S.; Matsumoto, M.; Okada, I.; Maeda, K.; Kataoka, Y. J.
Chem. Phys. 1995, 103, 8594.
(24) Howdle, S. M.; Bagratashvili, V. N. Chem. Phys. Lett. 1993, 214,
215.
(25) Fan, R.; Kalbfleisch, T.; Ziegler, L. D. J. Chem. Phys. 1996, 104,
3886.
(26) Ziegler, L. D.; Fan, R. J. Chem. Phys. 1996, 105, 3984.
(27) Akimoto, S; Kajimoto, O. Chem. Phys. Lett. 1993, 209, 263.
(28) Zerda, T. W.; Song, X.; Jonas, J. Appl. Spectrosc. 1986, 40, 1194.
(29) See, e.g. Mukamel, S. Principles of Nonlinear Optical Spectroscopy;
Oxford University press: New York, 1995.
(30) Li, B.; Johnson, A. E.; Mukamel, S.; Myers, A. B. J. Am. Chem.
Soc. 1994, 116, 11039.
(31) Marcus, R. A. J. Phys. Chem. 1989, 93, 3078.
(32) We have totally neglected the lifetime (T1) effect on the absorption
bandwidth in our model. We have performed transient absorption measure-
ments on PB in several solvent fluids. Since we could not detect the transient
absorption spectrum due to the excited state clearly, we could not obtain
the lifetime of the excited state directly. The ground-state recovery shows
decay with two time constants; near the center of the absorption peak, the

9060 J. Phys. Chem. A, Vol. 101, No. 48, 1997
Yamaguchi et al.
first component is about 0.3 ps, and the slow one is nearly 10 ps in methanol.
The fast component depends on the solvent and has a tendency to become
larger as the absorption peak shift blue. According to these results, it is
probably safe to say that the lifetime of the excited state (T1) is no shorter
than 0.3 ps. The T1 effect on the bandwidth is of the order (2πcT1)-1, which
stronger than the Raman emission. However, we consider it reasonable that
we did not detect the fluorescence emission for the following two reasons.
First, the fluorescence spectrum is expected to be very broad as the
absorption spectrum. Therefore, the fluorescence will be very weak
compared with the Raman emission even if it is 30 times greater when
integrated over the whole spectrum region. Second, the large solvent
reorganization energy naturally leads to the large fluorescence Stokes shift.
Considering the mirror image relationship between the absorption and the
fluorescence spectra, the fluorescence spectrum will appear in the near-
infrared region, far from the region where we observed the Raman spectrum.
(48) Guichard, V.; Buorkba, A.; Lautie, M. F.; Poizat, O. Spectrochim.
Acta 1989, 45A, 187.
is smaller than 20 cm-1. Since the total width σ is larger than 1000 cm-1
we can safely neglect the T1 effect.
,
(33) McRae, E. G. J. Phys. Chem. 1957, 61, 562.
(34) Albrecht, C. A. J. Chem. Phys. 1961, 34, 1476.
(35) Heller, E. J.; Sundberg, R. L.; Tannor, D. J. Phys. Chem. 1982,
86, 1822.
(36) See, e.g. Rothchild, W. G. Dynamics of Molecular Liquids;
Wiley: New York, 1984.
(37) Kolling, O. W.; Goodnight, J. L. Anal. Chem. 1973, 45, 160.
(38) Guimanini, A. G.; Chiavari, G.; Musiani, M. M.; Rossi, P. Synthesis
1980, 743.
(39) Nelson, T.; Wood, H. C. S.; Wylie, A. G. J. Chem. Soc. 1962,
371.
(40) Vittum, P. W.; Brown, G. H. J. Am. Chem. Soc. 1946, 68, 2235.
(41) Hamaguchi, H.; Appl. Spec. ReV. 1988, 24, 137.
(42) Jodl, H. J.; Holzapfel, W. B. ReV. Sci. Instrum. 1979, 50, 340.
(43) (a) Huang, F. H.; Li, M. H.; Lee, L. L.; Starling, K. E.; Chung, F.
T. H. J. Chem. Eng. Jpn. 1985, 18, 490. (b) Rubio. R. G.; Zollweg, J. A.;
Streett, W. B. Ber. Bunsen-Ges. Phys. Chem. 1989, 93, 791. (c) Couch, E.
J.; Hirth, L. J.; Kobe, K. A. J. Chem. Eng. Data 1961, 6, 229. (d) Younglove,
B. A.; Ely, J. F. J. Phys. Chem. Ref. Data 1987, 16, 577.
(44) Kimura, Y.; Yoshimura, Y.; J. Chem. Phys. 1992, 96, 3824.
(45) Kim, S.; Johnson, K. P. AIChE J. 1987, 33, 1603.
(46) Kim, S.; Johnson, K. P. Ind. Eng. Chem. Res. 1987, 26, 1206.
(47) Very weak fluorescence emission usually means very short lifetime
at the excited state. However, we do not consider that it contradicts with
our estimate of the excited-state lifetime T1 no shorter than 0.3 ps.32
According to our discussions in section 5.1, solvent reorganization energies
are smaller than 3000 cm-1, which leads to the pure dephasing time T2
longer than 20 fs. The ratio of the intensities of fluorescence and Raman is
estimated by (2T1/T2).29 Therefore the fluorescence should be about 30 times
(49) Poizat, O.; Guichard, V.; Buntiinx, G.; J. Chem. Phys. 1989, 90,
4697.
(50) For example, the frequency of the CtN stretching mode of CH3CN
varies about 13 cm-1 from gas phase to neat liquid (Ben-Amotz, D.; Meng-
Rong, L.; Cho, S. Y.; List, D. J. J. Chem. Phys. 1992, 96, 8781).
(51) Wang, Y.; Purrello, R.; Georgiou, S.; Spiro, T. G. J. Am. Chem.
Soc. 1991, 113, 6368.
(52) Musilli, M. M.; Loppnow, G. R. Chem. Phys. Lett. 1996, 261, 691.
(53) Markham, L. M.; Hudson, B. S. J. Phys. Chem. 1996, 100, 2731.
(54) Yamaguchi, T.; Kimura, Y.; Hirota, N. In preparation.
(55) Yamaguchi, T.; Kimura, Y.; Hirota, N. J. Chem. Phys. 1997, 107,
4436.
(56) Yonker, C. R.; Frye, S. L.; Kalkwalf, D. R.; Smith, R. D. J. Phys.
Chem. 1986, 90, 3022.
(57) Besserer, G. J.; Robinson, D. B. J. Chem. Eng. Data, 1973, 18,
137
(58) Matyushov, D. V.; Schmid, R. J. Chem. Phys. 1995, 103, 2034.
(59) Matyushov, D. V.; Schmid, R. Mol. Phys. 1995, 84, 533.
(60) Kalbfleisch, T. S.; Ziegler, L. D.; Keyes, T. J. Chem. Phys. 1996,
105, 7034.
(61) Yoshimura, Y. Mol. Phys. 1995, 86, 1419.
(62) Stephens, M. D.; Saven, J. G.; Skinner, J. L. J. Chem. Phys. 1997,
106, 2129.