31 October 1997

Chemical Physics Letters 10222 Ž1997. 380–390

Substructures and different energy relaxation time within the first

electronic transition of pinacyanol
B. Voigt, F. Nowak, J. Ehlert, W.J.D. Beenken, D. Leupold, W. Sandner
¨ Nichtlineare Optik und Kurzzeitspektroskopie, D-12474 Berlin, Germany
Max-Born-Institut fur
Received 31 May 1997; in final form 11 September 1997

Abstract

The S 0 –S1 absorption band substructure of pinacyanol dissolved in ethanol has been investigated with nonlinear
polarisation spectroscopy in the frequency domain ŽNLPF.. The evaluation of the experimental results has been done in the
frame of global analysis for a set of four NLPF spectra at two test wavelengths. Four bands with maxima located at 520,
556, 565 and 607 nm have been identified. The energy relaxation after excitation in the 565 nm band is slower ŽT1 s 13 ps.
than in the red-most band ŽT1 s 7 ps.. This behavior, inverse to usual excess-energy relaxation of organic molecules,
suggests that these subbands belong to different isomers of pinacyanol. q 1997 Published by Elsevier Science B.V.

1. Introduction
Pinacyanol Ž1,1X-diethyl-2,2X-carbocyanine chloride. ŽPI. belongs to the class of symmetric polymethine dyes.
PI has been widely applied as an optical sensitizer in the photographic process and has been used successfully as
a mode locking dye for rhodamine 6G and rhodamine B lasers. In the last ten years PI has been investigated as
an example for studying the photoisomerization of polymethine dyes. It forms short-lived photoisomers after the
S 0 –S 1 excitation w1–8x. Fig. 1 presents the red-most absorption spectrum with three maxima located at 607, 564
and ; 520 nm. The first one is usually interpreted as the vibrationless electronic S 0 –S 1 transition, the other two
as the same electronic transition under vibration cooperation. Nevertheless there are some hints from other
cyanine dyes, that the red-most absorbtion band is the result of superposition of several ground-state isomers
formed by rotation around a bond in the conjugated chain w2x.
The aim of this work was to test this hypothesis about the absorption band substructure.
Isomerization reactions have been extensively used to test theories on reaction dynamics in solutions w9–14x.
The most general theory for barrierless isomerization is proposed by Bagchi, Fleming and Oxtoby ŽBFO. w14x in
which they describe the influence of viscosity on electronic relaxation processes. In this theory the dependence
of relaxation on the excitation wavelength has also been predicted. The decay of the excited state depends on the
wavelength of excitation due to the fact that light of different wavelengths will excite the molecule to different
positions on the excited-state potential surface. The shorter the wavelength the further away from the sink is the
initial configuration and it should therefore need more time to reach the sink position.
˚
Akesson et al. w24x have investigated the dynamics of the photoisomerization of PI, t-stilbene and
1,1X-diethyl-4,4-cyanine in n-alcohol solutions by means of picosecond ground-state recovery ŽGSR.. They have
found isomerization without any potential barrier in the excited state by measuring the viscosity and temperature

0009-2614r97r$17.00 q 1997 Published by Elsevier Science B.V. All rights reserved.

PII S 0 0 0 9 - 2 6 1 4 Ž 9 7 . 0 1 0 4 3 - 9
 B. Voigt et al.r Chemical Physics Letters 278 (1997) 380–390 381

Fig. 1. The long-wavelength part of the absorption spectrum of pinacyanol dissolved in ethanol. The NLPF test wavelengths 552.052 and
570.040 nm are indicated by the arrows A and B.

dependence of the relaxation rate. For PI they observed a distinct dependence of the isomerization rate on the
position of excitation wavelength in relation to the absorption profile. Further investigations on 1,1X-diethyl-4,4-
cyanine have been done on a time scale up to 100 fs by means of GSR, fluorescence lifetime and excited-state
absorption ŽESA. measurements w25x.
The properties of PI are investigated by application of a multitude of spectroscopic methods. The
fluorescence lifetime has been investigated not only by fluorescence quantum yield measurements but also by
picosecond time-resolved fluorescence spectroscopy. The reported fluorescence lifetimes for PI dissolved in
ethanol range from 9 to 50 ps, depending on the type of excitation w15–21x. The first indication of a wavelength
dependence of T1 in the absorption band of the S 0 –S 1 transition was given in Refs. w20,21x. For PI dissolved in
methanol a decrease of T1 from 9.8 to F 5 ps between 595 and 619 nm and for PI dissolved in ethanol from 15
to 11 ps between 585 and 633 nm has been observed. In particular, Sundstrom ¨ et al. and Rentsch et al.
investigated the energy relaxation time T1 of the first excited singlet state as well as position and shape of the
transient absorption of the created photoisomer by means of GSR and transient absorption measurements
w22,23x. The absorption band of the created photoisomer is red-shifted with a maximum approximately at 630
nm for methanol and 640 nm for ethanol. Moreover, kinetic measurements have indicated that the photoisomer
of PI dissolved in methanol does not absorb light shorter than 605 nm w22x.
However, there is a critical inspection of all these experimental results w26,27x. Bagchi et al. have pointed out
that in case of a barrierless reaction the absence of a clear separation of time scales between the reactive motion
and the motion on the rest of the potential surface leads to the problem that different spectroscopic techniques
ŽGSR, ESA or fluorescence lifetime. are sensitive to different aspects of the motion on the excited potential
surface w26x. The exact result of a fluorescence decay measurement will depend on the selected wavelength and
spectral bandwidth as well as the results of GSR measurements only under special conditions will reflect the
overall decay of the excited states. A direct observation of the excitation-wavelength dependence predicted by
BFO theory has not been verified.
Therefore we investigated the red-most transition of PI to verify a possible wavelength dependence of the
relaxation time T1 by means of nonlinear polarization spectroscopy in the frequency domain ŽNLPF.. We have
chosen NLPF, because it connects the determination of energy and phase relaxation times on a femto- and
picosecond time-scale with a high resolution of spectral substructures 1.

1
It is noteworthy, that in contrast to saturated absorption spectroscopy techniques such as hole burning NLPF provides a large
temperature independent signal-to-noise ratio wich is important in particular for biological objects. Consequently, investigations at
physiologically relevant ambient temperatures are possible w28x.
382 B. Voigt et al.r Chemical Physics Letters 278 (1997) 380–390

2. Experimental

The first experimental realization of NLPF has been described by Song et al. w29x. In this Letter we introduce
a new advantageous pump–test beam configuration based on the fact that NLPF requires no phase-matching.
The phase-matching conditions necessary in three- and four-wave-mixing techniques result in more or less
parallel directions of the pump and test beam. For a finite optical density ŽOD., this leads to an imperfect
illumination of the sample which may result in a falsification of the measured spectra. To avoid these difficulties
we have chosen an orthogonal configuration of pump and test fields. By adequate rotation of the polarization
planes of the two linearly polarized lasers it is possible to get the same physical situation for the electric field
vectors as in the basic arrangement used by Song et al. w29x. The advantage of this configuration is a
homogeneous excitation of the sample in the direction of the test field. Fig. 2 shows the principle of NLPF
under these conditions. The sample is located between two crossed polarizers. The pump and test fields are
linearly polarized at an angle of 458. The idea is to measure polarization changes of a weak test field of
frequency v t s 2 p crlt , passing an absorbing medium in the presence of a strong pump field of variable
frequency v p s 2 p crl p . The NLPF signal behind the analyzer is observed in the direction and at the frequency
of the test field. The origin of the NLPF signal can be traced back to the non-linear optical modification of the
absorbing medium by the strong pump field. The polarization change is measured as a function of detuning
D s v p y v t . The measurements are performed under stationary conditions, i.e. the duration of the laser pulses
is much longer than the longest relaxation time of the system under investigation.
The basic item of our experimental arrangement is a selected set of two Glan-Thompson polarizers ŽHalle
Nachf., Berlin. with an cut-off ratio better than 10y8 . Pump and test laser fields are provided by dye lasers,
simultaneously pumped by an excimer laser ŽLPX 105 Lambda Physik.. The pump laser field is created by a
near grazing incidence system ŽLDL 205 Laser Analytical System., continuously tunable over the entire gain
range of the used dye solution. The test laser beam is fixed in wavelength and provided by the laser system LPD
3000 ŽLambda Physik; Littrow configuration with intracavity etalon.. The spectral band width of both lasers is
- 0.05 cmy1 , the pulse-width ŽFWHM. is ; 15 ns. For our measurements we have used the dye Coumarin 153
dissolved in propylene carbonate, tunable from 520 to 585 nm.
Pump and test laser energies are continuously monitored. Spectral bandwidth and wavelength calibrations are
controlled by a wave meter WA-4500 ŽBurleigh Instruments; spectral resolution 0.05 cmy1 ; accuracy "0.02
cmy1 ..
By means of lenses, the pump and test beams can be focused into the sample up to a diameter of some tens
of mm. In all cases the geometry of the pump and test laser spots has been controlled by a laser beam analyser
system ŽSpiricon LBA-100A.. Care has been taken that the intensity distribution of test laser spots was Gaussian
shaped and the illumination of the sample by the pump beam has been homogeneous. By means of neutral glass

Fig. 2. Principle of NLPF in the new orthogonal configuration Ž k p and k t represent the wave vector of pump and test radiation fields,
respectively; k S describes wave vector of the NLPF signal..
 B. Voigt et al.r Chemical Physics Letters 278 (1997) 380–390 383

filters different pump energies can be obtained. After passing a monochromator Žsuppression of scattered light.
the NLPF signal is detected by a photomultiplier tube with preamplifier ŽR636, Hamamatsu. and a boxcar
system ŽV1, Laser Analytical System.. The entire experiment Žtuning range of pump laser frequency, spectral
resolution, signal accumulation, etc.. is controlled by the program ‘Wavescan’, being part of the dye laser
system LDL 205.
The spectral bandwidth and tuning range of the laser systems determine the time resolution of the
measurement. The upper limit for determination of T1 is restricted by the finite spectral bandwidth Žwithout
deconvolution in our case T1 s 50 ps.. The lower limit for the determination of T2 is restricted by the tuning
range of the pump beam and the signal-to-noise ratio of the NLPF signal, which is essentially specified by the
cut-off ratio of the two crossed polarizers and the electronic equipment.
The data analysis requires the pump intensity to be constant over the used wavelength range. But the
intensity of the pump laser in our experiment depends on the wavelength. Therefore the original NLPF spectra
are weighted. The theoretically predicted, quadratical dependence of the NLPF signal on pump intensity is
experimentally fulfilled for a wide intensity range Ž4 = 10 5 –4 = 10 6 W cmy2 .. Consequently, the above-men-
tioned weight factor could be eliminated mathematically. Additional measurements of nonlinear absorption gave
no hints on a significant depopulation of the singlet ground state under our experimental conditions.

3. Theory

3.1. NLPF signal and model functions

Over the past two decades several non-linear spectroscopic techniques have been applied to condensed matter
to elucidate ultrafast relaxation times as well as mechanisms of spectral broadening. NLPF has been used by
Song et al. with the aim of determining ultrafast energy relaxation times ŽT1 . in dye solutions w29x. Further
developments have shown that this method can reveal more information than the determination of T1 and phase
relaxation times T2 w30,31x. Neef and Mory w32x used NLPF to elucidate the kind of spectral broadening
observed for large organic molecules in solution. The fundamental difference in the spectral NLPF line-shape
between an extreme-inhomogeneously broadened and a homogeneously broadened electronic transition band has
been pointed out.
The line-shape function describes the dependence of the third-order susceptibility x 3 on the relaxation
processes in a dissolved molecule. For a two-level system with damping it is calculated by the off-diagonal
density matrix element r 01 to in third order as
i 1 1 1 1 1
f Ž D, v 0 t . s
G q i v0t g ž G q iŽ v 0 t y D.
q cc. q
/ gqi D ž G q i v0t
q
G y iŽ v 0 t y D. / , Ž 1.

where G s 1rT2 and g s 1rT1 being phase relaxation rate and energy relaxation rate, respectively. v 0 is the
resonance frequency for the two-level system and v 0 j s v 0 y v j , the detuning between the resonance
frequency and the pump or test frequency. The quantity directly measured in our experiment ŽNLPF signal. is
proportional to the square modulus of the line-shape function < f Ž D, v 0t .< 2 of the third-order polarization Žsee
Fig. 5: homogeneous case..
For a two-level system with an additional third reservoir state, filled by relaxation from the excited state with
rate g 12 and depleted to the ground state with rate g 01 , we deduced from Song et al. w29x the line-shape function
i 1 g 12 1
fr Ž D , v 0 t . s
Gqi v0t ž g 02 q g 12 ž 1q
2g 01 /ž G q iŽ v 0 t y D.
q cc.
/
1 g 12 1 1
q
g 02 q g 12 q i D ž 1q
2 Ž g 01 q i D . /ž Gqi v0t
q
G y iŽ v 0 t y D. // . Ž 2.
384 B. Voigt et al.r Chemical Physics Letters 278 (1997) 380–390

The form of its square modulus differs only slightly from that of a pure two-level system in the vicinity of
D s 0 Žnarrow peak. and is identical for a broad-band spectrum. In any case there is no additional asymmetry in
line-shape compared to the pure two-level system. On the contrary if g 12 )) g 01 , g 02 we can approximate
f r Ž D, v 0 . by a line-shape function like Eq. Ž1., where g is substituted by g 01.

3.2. NLPF signal and kind of broadening

In general, the absorption spectra of dissolved organic molecules can be characterized by three types of
broadening.

3.2.1. Homogeneous broadening

There is only one resonant frequency v 0 . The NLPF signal is given by < f Ž D, v 0t .< 2 , which corresponds to a
homogeneously broadened absorption spectrum.

3.2.2. Inhomogeneous broadening

An inhomogeneously broadened absorption band is a superposition of a sufficiently large number of
homogeneously broadened subbands with only slightly different resonant frequencies v 0 . The total line-shape
function is given by
q`
f tot Ž D ; v L 2 . f Hy` g Ž v 0 . f Ž D, v 0 t . d v 0 , Ž 3.
with g Ž v 0 . being the distribution function of the molecular resonant frequencies v 0 . For the sake of simplicity
we assume g Ž v 0 . is a Lorentzian with width d and maximum at v L w32x. Thus the NLPF signal is given by
2 1 1 1
ftot Ž D ; v Lt . s
G q d q i v Lt ž g G q d q i Ž D q v Lt .
q
Ž g q i D . Ž G q d q i v Lt .
2
1 1 2 Gq2dqi D
q
ž g
q.
gqi D / G q d y i Ž D q v Lt . Ž 2 G q i D .
,

Ž 4.
where v Lt , is the distance of test frequency from v L .

3.2.3. Heterogeneous broadening

The superposition of k homogeneously or inhomogeneously broadened subbands leads to a NLPF signal
which is given by
2
k
2
ftot Ž D , vct ; j . s Ý a j f j Ž D , vct ; j . , vct ; j s v 0 t ; j or v Lt ; j , Ž 5.
js1

where the a j are weight factors for the contribution of the corresponding molecular group to the total non-linear
polarization dispersion shape including the dipole moments and the number of molecular dipoles in a unit
volume. The functions f j in Eq. Ž5. are given by Eqs. Ž1., Ž2. or Ž4. modified by the index j taking into account
different detuning frequencies v 0t; j or v L2; j , respectively, as well as different relaxation rates Gj and g j . Using
this model of broadening an asymmetry in the narrow peak caused by different T1 times of the subbands
appears. This can be verified by model calculations based on Eq. Ž5. and is demonstrated in Fig. 3 in case of
two pure two-level systems with different energy relaxation rates g 1 and g 2 , respectively.
Summing up, NLPF spectra consist of a narrow peak centred at v p s v t , revealing the information on T1
timeŽs. and of one or several broad maxima corresponding to resonant transitionŽs. of the different molecular
 B. Voigt et al.r Chemical Physics Letters 278 (1997) 380–390 385

Fig. 3. T1 peak line-shapes for different values of g 1 rg 2 . Inset: broad-band line-shape with g 2 s g 1 .

species. Only in case of extreme inhomogeneous broadening, where the spectral dependence of the distribution
function can be neglected, does an unequivocal relation between NLPF signal and type of broadening exist.
Then one only observes the narrow peak at v p s v t . Examples of NLPF spectra for the different kinds of
broadening are given in Fig. 4. A: extreme inhomogeneous Žonly narrow peak.; B: homogeneous Žnarrow as
well as broad band peak.; C: moderate-inhomogeneous Žnarrow peak and the initial appearance of a broad band
peak, here: G s d .; D: heterogeneous Žnarrow peak and one or more broad band peaks depending on distances
between the k resonance frequencies, here: two homogeneously broadened subbands, test frequency centered
between the two resonance frequencies..

4. Data analysis
The data are fitted using the method of least squares so that the unknown parameters Že.g., v 0; j , Gj , g j . in a
given model can be obtained. This has been realized in the scope of global analysis which implies the
simultaneous analysis of several data sets obeying the same model with the same model parameters, in our case
measurements at different test wavelengths l t . In this way, the identifiability of model parameters can be
improved w33x. Thus, to analyse m data sets the global goodness function
Nj 2
1 m Si j y a j Si j Ž lt , j , p k .
g k Ž pk . s Ý Ý Ž 8.
N y nk js1 is1 si 2j
has to be minimized with respect to the set of n k unknown parameters p k , where Si j represents the measured
value at the ith measuring point in the jth data set, si 2j is the associated variance estimated by repeated
measurements and Si j Ž l t, j ; p k . is the model value in the case of heterogeneous broadening with k subbands.
N s N1 q N2 q . . . qNm is the total number of observations and the factors a j are determined by means of a
simple linear least-squares fit.
The result of the minimization procedure is a minimum value g k) and an estimate p k) for the model
parameters. A good fit can be qualified by the x 2-probability of g k) . If the number k of subbands is unknown,
386 B. Voigt et al.r Chemical Physics Letters 278 (1997) 380–390

Fig. 4. Examples of theoretical NLPF signal line-shapes expected for different kinds of spectral broadening.

the relative improvement in the fits for the k and k q 1 subbands can be examined by an F-test because the
expression
Ž N y n k . g k) y Ž N y n kq1 . g kq1
)
Fk , kq1 s ) Ž 9.
Ž n kq 1 y n k . g kq1
has approximately an F-distribution with n kq 1 y n k and N y n kq1 degrees of freedom w34x.
By means of an error analysis, bounds for the estimated model parameters p k) can be computed. This error
analysis is based on a confidence region B which is defined by all parameter sets p k with g k Ž p k . F g, where
g ) g k) is chosen in such a manner that the p k do not differ significantly from p k) with respect to x 2- or
F-statistic and a given confidence level w34x. The computation of the upperrlower bound for one parameter of
interest means to maximizingrminimizing this parameter in B, allowing all other parameters to vary in B.

5. Results

PI ŽHoechst. was dissolved in ethanol. A dye concentration of 7.4 = 10y5 M and a test path length of 5 mm
were used. Our measurements took place around the 564 nm maximum. NLPF spectra were investigated at test
wavelengths of 552.052 and 570.040 nm with corresponding optical densities ŽOD. of 1.20 and 1.46,
 B. Voigt et al.r Chemical Physics Letters 278 (1997) 380–390 387

Table 1
Statistical parameters for subband analysis including results of the F-test
Number of subbands k Degrees of freedom N y n k g k) Ž x 2-probability. Fk, kq1 Ž F-probability.
3 901 1.329 Ž < 0.1%. F3,4 s 44.2 Ž < 0.1%.
4 896 1.072 Ž6.3%.

respectively. Test wavelengths and the tuning range of pump laser are indicated by arrows in Fig. 1. The applied
pump power density was 7 = 10 5 W cmy2 at 560 nm.
For the two different test wavelengths we get four NLPF spectra in all. The NLPF spectra in the case of low
and high resolution are depicted in Fig. 5a and b. Each point in the data representation is the mean value of ten
measurements. For high resolution Ž1 pm. the entire tuning range was selected in that way that the complete
narrow peak was recorded. For low resolution Ž0.5 nm. the investigated part of the absorption band was limited
by the tuning range of our pump laser.
At first sight the low-resolution NLPF spectra suggest a homogeneously broadened system with the narrow
peak at l p s l t and a second broad peak looking like a resonance peak next to the absorption maximum at 564
nm. However, the strong asymmetry of the narrow peaks ŽFig. 5b. indicates the existence of more than one
subband with different T1 Žsee above..
The investigation of the NLPF spectra assumed that the absorption band of PI is represented by a
heterogeneous inhomogeneously broadened system. That means the appropriate line shape formula is the
superposition of several inhomogeneous broadened subbands as given in Eq. Ž4.. To extract the physical
parameters of interest, we have tried to fit different model curves to the experimental NLPF spectra. Data points
in the nearest vicinity of l p s lt were not taken into account for the fitting procedure, because the NLPF signal
here may be superimposed by additional grating effects w30x.
The wavelengths l 0 at 520 and 607 nm have been fixed according to the absorption spectrum Žcf. Fig. 1..
This seems to be adequate because both subbands are outside the spectral range of our measurement. Moreover,
T1 for the subband at 520 nm has been fixed at 13 ps w16x. The other values are free parameters in the scope of
global analysis. As we were interested in a likely substructure and in a verification of a possible wavelength
dependence of T1 , models with three or four subbands have been fitted to the NLPF spectra. A comparison of
the relevant statistical parameters including the result of the F-test for the three- and four-subband analyses is
given in Table 1. We obtain a value F3,4 s 44.2 representing a probability < 0.1% that the fits are of same
content. This implies a significantly better fit for the four-subband model. The parameters of the best fit are
summarized in Table 2. The simulated NLPF spectra for the common data set depicted in Table 1 are
represented by solid lines in Fig. 5.
By means of an error analysis, limits for the T1 times of the subbands at 556, 565, and 607 nm have been
calculated. This error analysis is based on a confidence region which is defined by a probability of x 2 G 5%
that all parameter sets p k are of the same content as p k) . The lower and upper limits for T1 of the different
subbands according to global analysis are given in Table 3.

Table 2
Parameters for the best fit in frame of global analysis of NLPF spectra shown in Fig. 5 according to the heterogeneous—inhomogeneous
broadened model system discussed in the text
Subband l0 Žnm. T1 Žps. D lhom. Žnm.
520 Žfix. 13 Žfix. 19 Žvar.
556 Žvar. 4456 Žvar. 25 Žvar.
565 Žvar. 13.2 Žvar. 33 Žvar.
607 Žfix. 7.1 Žvar. 50 Žvar.
388 B. Voigt et al.r Chemical Physics Letters 278 (1997) 380–390

Fig. 5. NLPF signal in low Ža. and high Žb. resolution Žtest wavelength l2 s 552.052 nm and l2 s 570.040 nm.. Solid line: model curves
with parameters of the best fit according to Table 2.
 B. Voigt et al.r Chemical Physics Letters 278 (1997) 380–390 389

Table 3
Upper and lower bounds for the T1 given in Table 2
Subband l0 Žnm. Lower bound Žps. T1 Žbest fit. Žps. Upper bound Žps.
556 250 4400 unbounded
565 11.0 13.2 16.2
607 6.4 7.1 7.9

The subband analysis of our NLPF spectra results in the identification of four subbands located at 607, 565,
556 and 520 nm.

6. Discussion

Our results show a wavelength dependence of the T1 for the subbands at 607 and 565 nm. The outcome of
the narrow peak analysis of the NLPF spectra are different T1 with the T1 for the long-wave subband at 607 nm
shorter than for the short-wave subband at 565 nm.
The significant difference between them Ž13 and 7 ps. leads to the conclusion that this part of the band
structure cannot be interpreted in terms of a vibrational progression of the first excited singlet state. For a
vibrational progression one expects relaxation rate g 12 in the time range of sub-ps. This leads in the model of a
reservoir system to similar T1 times measured by NLPF as shown above.
One possible explanation of the difference in the T1 times is an isomerization in the excited-state with a
barrier, which produces the splitting into the two subbands and hinders a fast vibrational relaxation into the S 10
state from the state excited at 565 nm. This model is in contradiction with the results of other authors Žsee
above.. Another explanation is two barrierless excited states with different electronic characteristics. This would
not allow for a vibrational relaxation from the one to the other.
Our results support the original model of two co-existing ground-state species with slightly different
absorption spectra and relaxation times suggested as a possible source for the measured wavelength dependence
of the GSR and fluorescence lifetimes, e.g. Refs. w5,22x.
An interpretation of the subband at 520 nm as part of a vibrational progression to the electronic transition is
consistent with the theory.
The completely different T1 of the subband at 556 nm indicates that it probably does not belong to the
electronic system of PI. This is supported by additional fluorescence and chromatographic investigations. The
fluorescence excitation spectrum, detected at lem s 620 nm differs from the absorption spectrum in the region
around 550 nm, indicating the existence of another species which absorbs in this part of the spectrum.
¨
Klein-Bolting w35x has measured the time-resolved fluorescence of PI dissolved in ethanol Ž lex s 579.5 nm,
lem s 625 nm. and observed a long-living component Ž960 ps. contributing less than 1% to the overall signal.
By means of thin layer chromatography Žadsorbance: Al 2 O 3 60 neutral type E, Merck. a separation of two
components was possible. Although we must necessarily introduce a fourth subband at 556 nm, this subband
gives only small contribution to the entire absorption as a comparison of the resulting weight factors a j shows
Ž a565 nm s 1; a556 nm s 0.002..
Three different experimental methods have been used previously to probe the dynamics of barrierless
isomerization reactions ŽGSR, ESA, and fluorescence lifetime measurements.. Unfortunately, all these tech-
niques necessarily measure a superposition of several dynamic events Žradiation- and radiationless transitions,
transient absorption, etc.. and therefore they cannot give clear experimental evidence for a wavelength
dependence of the relaxation rate. By way of contrast, NLPF offers the possibility of a selective excitation of
higher states. Therefore, T1 times in NLPF are directly connected with the excited states selected by the
test-field lt .
390 B. Voigt et al.r Chemical Physics Letters 278 (1997) 380–390

Acknowledgements

The authors thank K. Teuchner and W. Freyer for the fluorescence and thin layer chromatography
measurements. This work has been supported by a grant from the Deutsche Forschungsgemeinschaft, Bonn ŽLe
729r2-2..

References

w1x G. Scheibe, J. Heiss, K. Feldmann, Ber. Bunsenges. 70 Ž1966. 51.

w2x W. West, S. Pearce, F. Grum, J. Phys. Chem. 71 Ž1967. 1316.
w3x V.A. Kuzmin, A.P. Darmanyan, Chem. Phys. Lett. 54 Ž1978. 159.
w4x ¨ T. Gillbro, Chem. Phys. Lett. 95 Ž1983. 580.
V. Sundstrom,
w5x ˚
E. Akesson, ¨ V. Sundstrom,
H. Bergstrom, ¨ T. Gillbro, Chem. Phys. Lett. 126 Ž1986. 385.
w6x ¨ T. Gillbro, J. Phys. Chem. 86 Ž1982. 1788.
V. Sundstrom,
w7x S.K. Rentsch, Chem. Phys. 69 Ž1982. 81.
w8x M. Kaschke, J. Kleinschmidt, B. Wilhelmi, Chem. Phys. Lett. 106 Ž1984. 428.
w9x ¨ T. Gillbro, H. Bergstrom,
V. Sundstrom, ¨ Chem. Phys. 73 Ž1982. 439.
w10x ¨ T. Gillbro, J. Chem. Phys. 81 Ž1978. 159.
V. Sundstrom,
w11x F. Heisel, J.A. Miehe, J. Rachidi, Chem. Phys. Lett. 134 Ž1987. 379.
w12x ˚
E. Akesson, ¨ J. Chem. Phys. 95 Ž1991. 6508.
A. Hakkarainen, E. Laitinen, V. Helnius, T. Gillbro, J. Karppi-Tomol, V. Sundstrom,
w13x ˚
E. Akesson, ¨ T. Gillbro, Chem. Phys. 106 Ž1986. 269.
V. Sundstrom,
w14x B. Bagchi, G.R. Fleming, D.W. Oxtoby, J. Chem. Phys. 78 Ž1983. 7375.
w15x A.T. Eske, K. Razi Naqvi, Chem. Phys. Lett. 63 Ž1979. 128.
w16x C.J. Tredwell, C.M. Keary, Chem. Phys. 43 Ž1979. 307.
w17x J.C. Mialoqc, J. Jaraudias, P. Gougon, Chem. Phys. Lett. 47 Ž1977. 123.
w18x ¨ Chem. Phys. Lett. 74 Ž1980. 188.
T. Gillbro, V. Sundstrom,
w19x P.J. Thistletwaite, H.J. Grieser, Chem. Phys. Lett. 91 Ž1982. 5862.
w20x ¨ T. Gillbro, Chem. Phys. 61 Ž1981. 257.
V. Sundstrom,
w21x ¨ T. Gillbro, Appl. Phys. B 31 Ž1983. 235.
V. Sundstrom,
w22x ¨ T. Gillbro, Chem. Phys. Lett. 94 Ž1983. 580.
V. Sundstrom,
w23x S.K. Rentsch, T.V. Danielius, R.A. Gadonas, Chem. Phys. 59 Ž1981. 119.
w24x ˚
E. Akesson, ¨ V. Sundstrom,
H. Bergstrom, ¨ T. Gillbro, Chem. Phys. Lett. 126 Ž1986. 385.
w25x ˚
U. Aberg, ˚
E. Akesson, ¨ Chem. Phys. 183 Ž1994. 269.
J.-L. Alvarez, I. Fedchenia, V. Sundstrom,
w26x ˚
B. Bagchi, U. Aberg, ¨ Chem. Phys. Lett. 162 Ž1989. 227.
V. Sundstrom,
w27x ˚
U. Aberg, ˚
E. Akesson, ¨ Chem. Phys Lett. 215 Ž1993. 388.
V. Sundstrom,
w28x D. Leupold, B. Voigt, M. Pfeiffer, M. Bandilla, H. Scheer, Photochem. Photobiol. 57 Ž1993. 24.
w29x J.J. Song, J.H. Lee, M.D. Levenson, Phys. Rev. A 17 Ž1978. 1439.
w30x S. Saikan, J. Sei, J. Chem. Phys. 79 Ž1983. 4146.
w31x ´
L. Marques, I. Reif, H.J. Franco, A. Marcano, O. Paz, J.L. Paz, Phys. Rev. A 47 Ž1993. 4185.
w32x E. Neef, S. Mory, Exp. Tech. Phys. 39 Ž1991. 385.
w33x J.R. Lakowicz ŽEd.., Topics in Fluorescence Spectroscopy, vol. 2, Plenum, New York, 1991.
w34x R.I. Jennrich, Computational Statistics, Prentice Hall, Englewood Cliffs, NJ, 1995.
w35x ¨
Klein-Bolting, ¨ Bochum 1988, p.41
Ph.D. Thesis, Ruhr-Universitat,