|Eilers-König , Nina: Ultrafast relaxation after photoexcitation of the dyes DCM and LDS-750 in solution |
The results of section 4. are discussed and critically reflected, starting with DCM and continuing with LDS-750. An outlook on future investigations concludes this section.
The spectral dynamics of DCM after photoexcitation from the electronic ground state are characterized by strong excitation intensity dependence for all excitation wavelengths and solvents investigated. This is manifested in the pump-probe spectra, especially in the time evolution of their total spectral intensity. With increasing excitation energy the amplitude fraction of a non-instantaneous contribution to the decay of the total spectral intensity increases (Figure 4.1-22). This decay is ascribed to the rise and decay of stimulated emission in the ranges above and below approx. 575 nm in acetonitrile and methanol or approx. 525 nm in chloroform, toluene, tetrachloromethane and cyclohexane (126.96.36.199.). The spectral dynamics could be described by precursor-successor kinetics (4.1.5.), the analysis yielding a time coefficient of 0.23 ( 0.04) ps averaged over the dynamics in all solvents investigated. This initial relaxation is followed by a red shift and further rise of the emission band in acetonitrile and in methanol, whereas in chloroform and in toluene the continued rise of the emission above 525 nm is accompanied by a further decay of emission below this wavelength.
The excitation energy dependence cannot be explained by saturation; it is interpreted as due to resonant two-photon absorption of DCM. Feeding of the S1 from higher electronic states has been proposed before for DCM [Kov 96, Ruth]. Very high effective non-resonant two-photon absorption coefficients have been reported for bis-donor substituted stilbenes in
123acetone and toluene by Ehrlich et al. [Ehr 97]. From a comparison of measurements with the same excitation energy, but different pump wavelengths of 530 and 470 nm in Figures 4.1-1 c) and 4.1-10 b), the larger relative amplitude of the blue shifted emission in the data for 470 nm excitation indicates a wavelength selectivity of the two-photon absorption process. The excited state absorption, as obtained from the spectral decomposition (fig. 4.1-1 b), has its maximum at 465 nm in acetonitrile. The two-photon absorption process may therefore be pictured to be resonant, with the first singlet excited state as an intermediate level. The transition strength for the excited state absorption can be roughly estimated from the spectral decomposition to about 1.4 x the transition strength for S0S1 absorption. The maximum extinction coefficient for ground state absorption has been determined (cf. 4.2.1) as 43300 l mol-1 cm-1 in acetonitrile, so that the maximum exctinction coefficient for the excited state absorption is 61500 l mol-1 cm-1. The absorption cross section can be calculated from the extinction coefficient after to maximum values of 01 = 1.66 x 10-16 cm2 for the S0S1 transition and 1N = 2.36 x 10-16 cm2 for the S1SN transition.
The transition moment d is related to the absorption spectrum according to [Lip 68] by:
From the intensity of a lognormal approximation to the ground and excited state absorption bands, the transition moments d01 and d1N were calculated after equation 5.1 to 2.95 x 10-29 and 3.02 x 10-29 Asm, respectively.
A cross section of two-photon absorption may be defined which is normalized by the photon flux per unit area :
Schubert and Wilhelmi [Schub] deduced the following expression for :
Here n(L) 1, 0 is the vacuum dielectric constant, 0 is the (angular) transition frequency and g((0) is the absorption line shape function. For simplicity, a Lorentzian lineshape is assumed, with
where 0 is the dephasing time of the electronic transition and estimated to 0 10-13 s.
0N is the polarizability matrix element for the transition between levels 0 and N. If the intermediate level 1 between the ground state and the state N reached by two-photon absorption is far from resonance, the polarizability can be approximated by [Schub] :
so that for non-resonant two-photon absorption here 0N would be 4.21 x 10-39 Asm2 / V and 2.0 x 10-54 m4 s. A photon flux of 1033 m-2 would therefore be enough to achieve effective non-resonant two-photon absorption. The 50 fs, 470 nm-centered pulse with a diameter of 150 µm on the sample and 1 µJ pulse energy such as employed in the pump-probe experiments corresponds to an intensity of 1.1x 1015 J/m2s and a photon flux of 2.7 x 10 33 m-2s-1 . This yields 0.54 x 10 -16 cm2 for the two photon cross section and thus raises it to about one third of the cross section 01 for the one-photon transition to the first singlet excited state. In the case of resonant two-photon absorption, the interaction cross section may increase by up to 107-108 [Letok]. Unfortunately, the excitation energy could only be varied in a range from 0.2-1.4 µJ, the lower limit determined by the signal-to-noise ratio and the higher limit imposed by saturation of the S0S1 transition. Therefore, and because of the overlap of the excited state absorption band, a more thorough study of the excitation intensity dependence of the blue shifted emission could not be carried out. The shift of the early emission bands in the pump-probe spectra of DCM in chloroform when changing from a 470 nm excitation wavelength to 530 nm excitation (figures 4.1-12 b) and 4.1-13), leads to the assumption that at least in chloroform, several higher-lying electronic states may be reached by two-photon absorption. This is roughly confirmed by the semiempirical calculations yielding three transitions between 38000 and 41000 cm-1.
In the limit of zero excitation energy, the time coefficient for the decay of the integrated spectral intensity reduces to approx. 0.09 ps in acetonitrile (4.1.4.), close to the timescale of inertial motion for acetonitrile molecules of 100 fs [Cho 92, Horng 95, Ruth 98]. A
125similar tendency towards lower values for in chloroform and tetrachloromethane cannot be relied on, because the large scattering of values from different days of experiment would require a larger data set. The trend in acetonitrile can be interpreted in two ways: either a very fast reaction takes place or is observed only for low excitation energy, or solvent relaxation gradually alters the transition strength for S1S0 fluorescence. The lack of excitation energy dependence of the time coefficients from the precursor-successor fits and the fact that there is no fast decay of the emission in the blue and green spectral region support the second assumption. It may also explain the absence of a decrease of with excitation energy for methanol, where most of the energy relaxation from solvent reorientation is slower than 0.2 ps (see discussion of LDS-750 results). The solvent-dependent continued rise of the emission band in acetonitrile and methanol on a picosecond timescale would also be in line with a transition strength dependence on solvent orientation. It could not be examined whether these amplitude changes are reflected in the total intensity on a picosecond timescale, because the noise level is too high to detect a small decay component of the integrated spectral intensity. The amplitude rise could also be due to emission band narrowing as observed by Glasbeek et al. and Gustavsson et al. [vdMeul 98, Gust 95].
On the other hand, the appearance of a second band around 460 nm in addition to that at 540 nm (fig 4.1-10) with higher excitation energy indicates a three-state scheme such as proposed by Ruthmann [Ruth], who found evidence for a charge transfer reaction of DCM in S1 in methanol with = 246 fs and feeding of the reactant state from an emissive higher-lying electronic state with = 120 fs.
Figure 5.1-1: Three-state scheme for DCM relaxation in methanol as proposed by Ruthmann.
126This cannot be corroborated here, as no indication of a decay as fast as 120 fs resulted from the precursor-successor modelling at high excitation energies. If there are three states involved in the relaxation, the reaction rate appears not to be influenced by the reaction path. It is also nearly independent of the solvent (table 4.1). The absence of dielectric relaxation effects on electron transfer have been reported for porphyrine-dichloroquinone cyclophanes [Pöll 92], so that a charge transfer reaction from two precursor states, the population of which depends on excitation energy, might be thought of.
In the first 400-600 fs the evolution of the dump-probe spectra is determined by a decay of a structured absorption band at wavelengths larger than 519, 537 or 558 nm in acetonitrile, methanol and propylene carbonate, respectively, and the rise of another absorption to the blue of this temporary isosbestic point. The dynamics can also be modelled by precursor-successor kinetics with a time coefficient of 0.28 ( 0.07) ps, averaged over similar dynamics in acetonitrile, methanol and propylene carbonate. After 400-600 fs, the absorption band shifts continously to the blue, accompanied by a decrease of the integrated spectral intensity on a solvent-dependent, picosecond time scale. From the empirical comparison of the dynamical Stokes-shift of DCM with that of coumarin 153 (4.1.3.), the relaxation of DCM after 0.6 ps in the ground and excited electronic states in dipolar solvents is assigned mainly to solvation dynamics. The decrease of the integrated intensity on the same timescale is attributed to a decline in amplitude of the absorption band, corresponding to a modification of the transition strength to the S1 state by solvent reorientation. The dynamics before 0.6 ps are governed by a state-to-state transition, which could be either intramolecular vibrational redistribution or a conformational relaxation of DCM, possibly connected with a charge redistribution.
Possible evidence for a charge transfer reaction from the analysis of the stationary spectra and from quantum chemical calculations is discussed next. The solvatochromic analysis (4.2.4.) does not give evidence for a sizeable change in dipole moment after optical transition to the electronic excited or ground state, since the polarity dependence of the
127peaks of the absorption and emission spectra could be described with nearly the same set of parameters, only differing by 0.3 Å in the Onsager cavity radius. Usually the molecular radius is not treated as a parameter but estimated from the van-der-Waals radius of the molecule. The value obtained here for r0 ( 10 Å) is large compared to 5.7 Å for DCS from a similar analysis of [Il'ich 96]. It is closer to the 8 Å assumed for DCM by Marguet et al. [Marg 92] from extrapolated literature data of the subunits forming the DCM molecule. A crystallographic structure analysis would be necessary to clarify these differences. The deviations from the fit in Figure 4.2-7 are not due to measurement errors, but signify the shortcomings of the continuum model which does not account for microscopic solvent structure.
The excited state dipole moment of 23 Debye used in the solvatochromic analysis could not be reproduced in the quantum chemical calculations (9.9 D and 14.9 D for S0 and S1), which limits their reliability or at least their transferability to the description of the excited state in solution. The calculated dipole moment changes after excitation into S1 or S2 were very small (-0.44 D for S1 and + 1.7 D for S2) so that a large charge alteration after photoexcitation seems not very probable for DCM. Also, the small changes in dipole moment with torsion of the anilino, dimethylamino or dicyanomethylene-methyl-pyran subunits in S0, S1 or S2 are not in favour of a TICT mechanism. This differs from the result of Marguet et al. concerning the dimethylamino group, which at a 90° twist in S2 led to marked charge transfer character of the configuration and a dipole moment of 22.45 D. For the other single bond rotations, the results are in line with the findings in [Marg 92].
The simulation of the stationary absorption spectra (4.2.5.) would be more exact if carried out using the frequencies from the resonance Raman spectrum (4.2.3.). Since the transition strenghts could not be estimated from the spectrum in methanol, but only of the crystallic form, and the amplitudes as well as the frequencies of the non-resonant Raman spectrum were found to be influenced by the environment, this could not be done. Especially the Raman peaks corresponding to vibrations of the ethylene bonds involving electrons were subject to solvent polarizabilitiy and polarity depending variations. Also, the analysis would not have permitted a comparison of the absorption and fluorescence bands. In the frame of the simplified picture of one effective harmonic mode coupled to the optical transition, the simulation of fluorescence and absorption spectra gave evidence for gradual changes in the equilibrium position and frequencies of the S1 and S0 vibrations with polarity of the solvent.
128The same tendency towards a smaller equilibrium displacement with increasing polarity was found for the fluorescence and absorption bands. This can be attributed rather to a static modification of the energy hypersurface of a single excited or ground state than to a polarity-dependent barrier height for the transition between two discrete conformer states.
In summary of this subsection, configurational changes after Franck-Condon projection into the electronic ground and excited states cannot be ruled out and would be even expected from the different effective harmonic frequencies and mode displacements found in the analysis of the absorption and fluorescence bands. There is no evidence, though, of the existence of several stable conformers, or of a sizeable charge separation or large alteration of electronic structure connected with those changes.
After two-photon absorption and internal conversion to S1, the molecules should be highly vibrationally excited. A lower limit for the vibrational energy relaxation time in the electronic ground state was derived from the width of the resonance Raman line at 1552 cm-1 to 310 fs (4.2.3.). A fit of the lineshape of the other dominant mode at 1213 cm-1 to a Gaussian gives a fwhm of 11 cm-1 , corresponding to T1 > 480 fs. Vibrational relaxation in the excited state may be faster than in the ground state, but it is unlikely that its rate should increase by a factor of 10, as would be necessary for it to remain unobserved in the pump-probe experiment. For SN, a much larger rate for intramolecular vibrational redistribution can be assumed from the higher density of vibronic states. Vibrational coherences due to impulsive excitation of low-frequency modes for DCM observed at probe wavelengths around 470 nm ( Figure 5.1-2 ) persist for longer than a picosecond, implying small rates for vibrational dephasing and energy transfer also for low-frequency modes. This is important, because vibrational relaxation by energy transfer to low-frequency solvent modes via low-frequency "exchange" solute modes has been proposed [Shelby 79]. Vibrational dephasing time coefficients as large as 440-1330 fs have been found from the analysis of vibrational coherences in the time-resolved pump-probe spectra of a cyanine dye, HDITC, in ethylene glycol [Yang 99]. It is concluded that for high excitation energies, the dynamics observed in the pump-probe spectra of DCM with an average time coefficient 0.23 ( 0.04) ps can be attributed to intramolecular vibrational energy redistribution in S1 after fast internal
129conversion from a higher-lying electronic state SN. The approximate independence of the reaction rate from solvent polarity supports this assumption, as IVR for trans-stilbene has been found to take place in about 2 ps in methanol and in hexane [Schultz 97]. This can be explained by the presence of a large number of intramolecular low-frequency modes in molecules of the size of stilbene, which permit the redistribution of energy independently of the solvent low frequency modes and therefore render IVR insensitive to the solvent spectral density for such modes.
Figure 5.1-2: Quantum beats in the kinetic trace at 577 nm of DCM in methanol from pump-probe measurement with excitation at 470 nm and an excitation pulse energy of 0.4\µJ.
The simulations of vibrational relaxation and solvent reorientation (4.4.) reflect the principal dynamic features of the stimulated emission spectra at high and intermediate pump energies, but can neither account for the exact frequency spacing of the emission bands, nor for the correct timescale of the spectral dynamics in chloroform. There are several reasons to consider: the simplest, that several optically bright excited (electronic) states are involved (see 5.1.1.). The dielectric relaxation times L were set to the average time coefficient of the coumarin 153 Stokes shift dynamics, which may have led to an overestimation of the relevant L in chloroform. For example, fast translational motion of the chloroform molecules might be the relevant movement here for energetic stabilization of the excited state. Furthermore, the value of 3620 cm-1 for the excess energy was rather arbitrarily selected from the energetic spacing of the emission bands in the early pump-probe spectra of DCM in methanol and might be too low, if following the quantum chemical calculations. On the other hand, vibrational relaxation could be faster for higher vibrational states than
130determined by the quadratic dependency on excess energy in the model of Montroll and Shuler (2.4.). A bottleneck for vibrational relaxation at low excess energies has been proposed for trans-stilbene by Nakabayashi et al. [Nak 98]. Nevertheless, the simulations demonstrate why spectral changes appear more pronounced in chloroform (thanks to the relatively long reorientation time of the large chloroform molecules) not only for DCM, but also for LDS-750. Another interesting feature is the observation of isosbestic points in the cyclohexane and chloroform simulations, which is usually interpreted as a sign of direct inter(electronic)-state relaxation and here is due to intramolecular vibrational energy redistribution.
Cyclohexane has been treated analogously to the dipolar solvents, which at first view is justified by the observation of nonpolar solvation dynamics for other molecules [Loch 99, Gard 95, Yam 97], see also 2.3., and by the large Stokes shift of DCM in liquids with small 0. At a closer look, this process has been explained by [Gard 95] as being due to quadrupol and higer order electrostatic moments of non-dipolar solvent molecules which are not present for 2-methyl butane. The Stokes-shift of DCM in 2-methyl butane is as large as 870 cm-1, though (4.2.5.). Dispersive forces, which should then be responsible for the Stokes shift, are characterized by a 1/r6 potential, whereas Coulombic interaction energy decreases with distance as 1/r. Therefore, a treatment of non-dipolar solvation dynamics by a Debye model may seem inappropriate, and mechanical relaxation theory would be more suitable [Berg 94, Ma 95]. Ladanyi and Klein [Lad 96b] have shown in molecular dynamics simulations that in spite of the more translational character of non-dipolar solvation dynamics most of the solvation process arises from the innermost region of the solvent shell, as with dipolar solvation, and may therefore be as fast as the latter, only missing the Gaussian inertial component. The continuum model in the form used here only roughly approximates the dipolar solvation dynamics, since multiple timescales and Gaussian inertial behaviour are ignored, so that it serves as an approximation to both dipolar and non-dipolar solvation. A model describing both solvation processes in more detail and on a physical foundation would be more appropriate and might give results closer to the data.
The optical properties and the reaction dynamics of DCM as could be elucidated in this work are summarized and discussed in conclusion.
Franck-Condon excited state
The larger instantaneous relative amplitude of the excited state absorption in the pump-probe spectra of DCM in non-dipolar solvents implies a solvent-dependent transition strength to higher electronic states or to the ground state. It exceeds by far the solvent-dependence of the transition strength for the S0 S1 transition as reflected in the maximum extinction coefficients for the stationary absorption in 5.2.1. Therefore, it is mainly ascribed to a variation of the transition strength to higher electronic states. This is explained by different instantaneous electron density distributions of the S1 state in non-dipolar and dipolar solvents. It should be interesting to check on this conclusion by performing semiempirical calculations under consideration of the solvent environment. The altered charge distribution implies a variation of the equilibrium displacement of the vibrational modes coupled to the electronic transition. Indeed, a gradual variation of the displacement of an effective harmonic vibrational mode with solvent polarity has been noted (4.2.5).
Competing reaction channels
The large dipole moment of DCM in S1 implies a notable stabilization of the first singlet excited state by dipolar solvation. It is manifested in the spectral shift of the stimulated emission spectra, and also in the "Anti-Stokes" shift of the time-dependent ground state absorption spectra after 400-600 fs in dipolar solvents. Thus, the spectral evolution after 400-600 fs in dipolar solvents in the electronic ground and in the excited state is found to be determined mainly by diffusive solvation dynamics. Stabilization of the excited state by dipolar solvation would be coincident with an increased barrier height for isomerization in dipolar solvents, if the latter proceeded via an intermediate state of a moderate dipole moment. A lack of evidence for photoisomerization in cyclohexane together with the low fluorescence quantum yield in that solvent hint at the importance of another desactivation channel apart from isomerization, though, e.g. inter-system crossing to a close-lying triplet state.
The spectral evolution after 400-600 fs in dipolar solvents is attributed primarily to
132solvation dynamics in the electronic ground and in the excited state. But what is the nature of the relaxation observed in the first half of a picosecond, and how to explain the changes in bandshape or (for dipolar solvents) in transition strength observed on a picosecond timescale?
For high excitation intensities (>0.5x1015 J/m2s) , vibrational energy redistribution after fast internal conversion from one or more higher-lying electronic state(s) SN is likely to be the dominant process observed in the pump-probe measurements (see 5.1.1., 5.1.4.). In contrast to the findings of Ruthmann [Ruth], no differences were observed between the relaxed pump-probe spectra at 20-40 ps measured with high and low excitation intensities in methanol, acetonitrile and toluene. It is concluded that from any populated state SN, internal conversion to S1 is the only accessible reaction channel.
Low excitation intensities are difficult to realize, and even if present, vibrational excess energy may be conferred upon the molecules depending on the choice of the excitation wavelength. In the stimulated emission pumping experiments, the dump pulse was centered to the blue of the fluorescence maximum of DCM in the solvents investigated. If the total Stokes shift of DCM is attributed to dipolar, multipolar or nonpolar solvent reorganization, as presumed in 4.2.5., no excess energy is expected for the molecules after stimulated emission pumping. The same argumentation applies to optical excitation at 530 nm of DCM in methanol and acetonitrile with low excitation intensity. Resonant two-photon excitation can be more easily evaded here, as this wavelength is far from the excited state absorption maximum. The excitation wavelength 530 nm is also situated on the red edge of the absorption spectrum in methanol and acetonitrile. Unfortunately, the larger excited state absorption in non-dipolar solvents and its proximity to the ground state absorption band prevent similar measurement conditions for these solvents. But for dipolar solvents, in the frame of the stated assumption, vibrational relaxation can be excluded from the causes of the ground state dynamics and excited state dynamics for low-intensity 530 nm excitation.
Under these experimental conditions, the decaying and rising bands are very close, the maxima spaced by approximately 900-1100 cm-1 (fig. 4.1-1 c) and 4.1-8 a). In the isolated ground state absorption spectra, the frequency difference between the maxima of the
133precursor and successor species is 2100-2300 cm-1. This is reflected in the precursor / successor spectral amplitudes as presented in figures 4.24 b) and 4.25 b). An assignment of the structured precursor spectra to a locally excited (LE) state as in [Kov 96] is not possible, because the ground state precursor spectra should then belong to the charge transfer conformation and be broad and without structure. A conformational relaxation is tentatively assigned to the observed relaxation in the electronic ground state and for low-intensity excitation conditions, also in the excited state.
The maximum dipole moment differences corresponding to the energetical spacing of the precursor and successor species can be estimated as follows. The solvent reorganization energy was estimated to 4320 cm-1 in acetonitrile (see 4.4.). The dipole moments of the relaxed ground state and excited state conformations were assumed as 9 and 23 Debye, and their absorption and fluorescence spectrum were assumed to be coincident with the stationary absorption and emission spectrum. The dipole moments of the precursor conformations in the ground and excited states were then calculated via equation 2.31, substituting - 0 in equation 2.31 by 4320-2300 cm-1 and 4320 -1100 cm-1, respectively. The dipole moment of the precursor conformation was thus estimated to <13.4 D in the ground state and to > 21.1 D in the excited state, so that the maximum dipole moment variation between reactant and product state would be 2 D in the electronic excited state and 4.4 D in the electronic ground state. The observed reaction therefore cannot be characterized as a charge transfer reaction, as it could be expected from 5.1.3. Since the reaction rate is nearly the same in the highly viscous propylene carbonate and in methanol ( (0.3 ps)-1), large amplitude motion such as a 90° single bond twist does not seem probable. Such a conformational relaxation was attributed to the reaction of DCS after photoexcitation by comparison with the bridged compound DCS-B24 [Abr 97]. Bicimer formation should also be ruled out for the DCM concentrations utilized.
The fact that the rate coefficient for relaxation in the excited state in most solvents ( (0.23 ps)-1) is found independent of the excitation intensity indicates a similar reaction mechanism for the conformational and the vibrational relaxation. Electron-transfer has been reported to proceed via higher vibrational levels of the product state. With only approximately 1000 cm-1 energy difference between the reactant and product states and no excess energy provided by photoexcitation, this reaction mechanism can be once more excluded. If some part of the Stokes shift of DCM was due to intramolecular vibrational
134reorganization, though, the observed conformational relaxation in the excited and in the ground state could simply reflect the redistribution of vibrational excess energy.
The changes of the pump-probe spectra on a picosecond timescale may be explained by vibrational cooling, if the short-time dynamics are assigned to IVR. The manifestation of vibrational cooling should be more prominent for higher excitation intensity, when a large fraction of molecules is highly vibrationally excited. The fact that the observed spectral changes on a picosecond timescale are independent of excitation energy (figures 4.1-9, 4.1-11 and 4.1-12) renders vibrational cooling unlikely as their cause, and favours (continued) conformational changes on a picosecond timescale as an explanation. This is similar to the findings of [Mart 95, 97] and also accounts for the observation of an isosbestic region in the pump-probe experiments in chloroform and toluene on a picosecond timescale. In dipolar solvents, the conformational relaxation is held responsible for the decrease in transition strength with solvent reorientation in the electronic ground state, and for the amplitude growth of the stimulated emission spectra on the same timescale. An increase of transition strength in the excited state and a decrease in the ground state correspond to forward and backward reaction, respectively. The slightly different equilibrium displacements of an effective harmonic mode in the absorption and fluorescence spectra are attributed to this conformational relaxation. No charge transfer reaction as in [Mart 95, 97] is concluded. In contrast to the short-time dynamics, the characteristic timescale of this relaxation depends strongly on the solvent. For dipolar solvents, the nature of this reaction can be characterized as highly adiabatic, since no change in spectral bandshape occurs. This leads to the impression of a gradual variation of transition strength with solvent reorientation. It should be kept in mind, though, that for the excited state the small picosecond rise of the emission band in dipolar solvents might also be due to emission band narrowing by nonlinear solvation [Mat 97, vdMeul 98].
In contrast to DCM, LDS-750 shows a solvent-dependence of the dynamic changes in its stimulated emission band on a very short (sub-80 fs until 500 fs) timescale as well as in the picosecond range. The initial fast decay of an emission band centered around 635-655 nm is most pronounced in chloroform and propylene carbonate, whereas in methanol and acetonitrile it is harder to detect, taking place nearly within the instrumental time resolution. It is followed by a further decay of this band within 0.6 - 1.8 ps and the simultaneous rise of a red shifted emission band centered around 700-720 nm in all solvents except acetonitrile. Here the lower-frequency emission band rises with a time coefficient of approximately 0.5 ps with only small changes in differential optical density in the region of the higher-frequency emission. This is paralleled by the spectral evolution in propylene carbonate for long delay times, when the red shifted emission also grows with a time coefficient of 4.4 ps with only a very small corresponding decay of the higher-frequency emission band.
The time coefficients have been deduced from an analysis of the kinetic traces at prominent wavelengths and are subject to errors especially in methanol, where the observed frequency shift of the slowly rising emission band is large. The multiple timescales of the rise and decay of emission cannot be extracted from the integrated spectral intensity, as here the simultaneous rise and decay components partly compensate and therefore the other components are overweighted. In addition, the time coefficients for fits with sums of more than two exponential functions are often correlated if they do not differ by at least a factor of five, which naturally also limits the reliability of the analysis of the kinetic traces. The estimated error for the given time coefficients from different fits and measurements is about 20%.
The initial, fast decay of the higher-frequency emission is attributed to a fast depopulation of the Franck-Condon excited state. Since there is no simultaneous growth of the lower-frequency emission, the optical transition from the product state to the ground state should be forbidden. The reaction rate is close to the inverse time for inertial solvent motion (see 4.5.) and explains, why these motions were first claimed to have been observed in the spectroscopy of LDS-750. In chloroform, where the fastest solvent reorientation time is in the range of hundreds of femtoseconds and the spectral broadening and the decay is
136therefore slower than for the other solvents (see 5.1.4.), the initial higher-frequency emission still exhibits some substructure with bands spaced by approximately 1000 cm-1.
The Franck-Condon excited state is depopulated nearly completely by this fast reaction in acetonitrile, as the emission around 650 nm is found drastically reduced after 240 fs. This is not the case for the other solvents; especially in chloroform the amplitude of the lower-frequency emission band is still larger than that of the lower-frequency emission band after 1.3 ps. In methanol, propylene carbonate and chloroform population is transferred from the Franck-Condon excited state to the final emissive state on a 0.6 - 5 ps scale depending on solvent. Finally, in acetonitrile and propylene carbonate, the population of the final emissive state grows by feeding from an optically quasi-dark state, as now the higher-frequency emission rises accompanied only by small changes of differential optical density in the spectrum below 680 nm. The reaction rates for the population of the final emissive state vary with the relaxation times for diffusive solvent reorientation and agree with these apart from methanol (see 4.5.).
The reaction scheme thus involves three electronic states or conformers in the first electronic excited state ( Figure 5.2-1 ), namely the Franck-Condon excited state A, a second state with a quasi-forbidden optical transition to the electronic ground state B, and the state from which the emission around 720 nm occurs (C). Since the transition S0 S2 was found at 9460 cm-1 higher energy than the S0 S1 transition [Ruth], the states A, B and C must correspond to conformers of LDS-750 in the first excited singlet state. The larger amplitude of the excited state absorption and bleach bands relative to the emission band in the relaxed spectrum of LDS-750 in chloroform compared to the more strongly dipolar solvents indicates a solvent-dependent population distribution between these conformers. The equilibrium population of the quasi-dark state appears to be significant in chloroform. It could not be checked if in consequence the fluorescence quantum yield in chloroform is lower than in the other solvents, since the fluorescence lifetime in chloroform was not known. For several dipolar solvents, [Cast 87] found an inverse proportionality between the fluorescence lifetime (160 ps in acetonitrile and 230 ps in methanol) and solvent visosity. It was concluded that the fluorescence lifetime of LDS-750 is shortened by some non-radiative process, most likely an excited state isomerization around one of the butadiene bonds.
The small red shift of the stationary emission spectra of LDS-750 with solvent polarity
137signifies a small dipole moment difference between state C and the (Franck-Condon) ground state. On the other hand, the blue shift of the stationary absorption band indicates a stabilization of the electronic ground state relative to the Franck-Condon excited state A, so that state A is expected to posess a smaller dipole moment than the ground state. This argumentation ignores possible differences in the dipole moment of the Franck-Condon and relaxed ground state configuration. An absolute dipole moment for ionic molecules can be specified only relative to a reference position, but the relative difference in dipole moment of different configurations could be estimated from semiempirical calculations as in [Ruth].
An analysis of the stationary spectra similar to that in 4.2.5 was performed by [Kov 97] yielding a dimensionless displacement of 1.4 for a harmonic 1500 cm-1 mode from the absorption and around 0.6 from the emission spectrum in acetonitrile. The observed dynamics after photoexcitation at 530 nm were explained by an ultrafast isomerization followed by a solvent-dependent process with a time coefficient of 0.6 ps in acetonitrile. Later, the latter mechanism was assigned to feeding from higher excited states after two-photon absorption [Ruth]. Due to the missing excitation intensity dependence observed in the experiments here, the quasi-dark state B cannot be populated by two-photon absorption, but instead is thought to be fed directly from the Franck-Condon excited state A ( Figure 5.2-1 a). The further reaction mechanism is pictured as follows: The diffusive reorientation of the solvent molecules leads to an energetic stabilization of state C relative to state A ( Figure 5.2-1 b). On the timescale of slow solvent motion, therefore population is transferred from A to C in all solvents except acetonitrile, where the inertial component of solvent reorientation is dominant and leads to depletion of state A. The stabilization of state C in strongly dipolar solvents finally leads to a lowering of state C beneath state B and causes additional feeding of state C from state B ( Figure 5.2-1 c). Why this feeding is not observed in methanol, which is of comparable polarity to acetonitrile and propylene carbonate, can only be explained by the smaller relative influence of inertial motion in the solvent reorganization and consequently smaller population of the quasi-dark state B. The reaction mechanism AB is not clear, nor is the meaning of the connection between the reaction rate and inertial solvent reorientation understood. The isomerization of the retinal chromophore bacteriorhodopsin was reported by [Du 93] with nonexponential dynamics to proceed on various timescales, the fastest components lying in the range of 90-240 fs. The conformer B might be an intermediate in an isomerization reaction of LDS-750, in
138conformity with the interpretation of [Kov 97]. Since the viscosity of propylene carbonate (2.53 cP) is large compared to methanol (0.55 cP) and acetonitrile (0.34 cP), the next step in isomerization might be hindered by solvent friction. This would acount for the elevated amplitude of the slow feeding of state C from B in propylene carbonate. The inversion of state C relative to A and B by dipolar solvation implies a markedly increase in dipole moment for the transitions AC and BC, so that in contrast to DCM LDS-750 is considered to be subject to a charge transfer reaction. Until a confirmation of the existence of such conformers from quantum chemical calculations can be given, the proposed scheme of Figure 5.2-1 should be regarded as preliminary.
Figure 5.2-1: Proposed reaction scheme for LDS-750 relaxation after excitation into the first excited electronic singlet state. In the course from a) to c), an inversion of state C relative to states A and B occurs, due to larger energetic stabilization of state C by solvent reorientation.
The interpretation of the pump-probe spectra is based on assumptions about or ignorance of possible changes of the spectral bands overlapping with the stimulated emission. A global analysis taking into account the excited state absorption kinetics would involve the unknown properties of at least one more higher excited electronic state. To reduce the information gained by the experiment to the S1S0 transition, the fluorescence upconversion technique could be used instead. The spectral and temporal resolution of fluorescence upconversion experiments has up to now been limited by the nonlinear gating process of the spontaneously emitted photons, requiring a bandwidth-limiting phase-matching condition of the fluorescence and gating fields in the nonlinear crystal. An extension of this technique to the broadband, sub-100 fs regime is currently under way in our laboratory. The dump-probe technique combines high spectral and temporal resolution with the investigation of only the lowest optical transition and is therefore an extremely valuable tool to study reactions in the electronic ground state. It would be instructive to carry out these experiments on LDS-750, too, to see how the conformational back-reaction is reflected in the ground state absorption spectra. Unfortunately, the limited tunability of the CPM system did not allow for enough energy in amplified pulses above 700 nm, and below that wavelength ground state absorption was found to impede effective stimulated emission pumping of LDS-750.
The high two-photon absorbance of DCM and, as has been shown in [Kov 97, Eil 98], also of LDS-750 suggests the necessity to reduce the pump pulse intensity to values well below 1015 J/m2s in the investigations of large chromophores in solution. An absolute noise of < 10-3 OD is required to maintain the signal-to-noise ratio of 20-100 of the measurements demonstrated here. The further reduction of noise requires a still deeper understanding of the continuum generation process, since the spatial fluctuations of the probe continuum were found here to be the main source of noise.
Slow intramolecular vibrational energy redistribution has been proposed rarely up to now for complex dyes (see 2.4.) and should be considered in the interpretation also of other donor-acceptor substituted stilbenes' dynamics. It would be interesting to perform
140stimulated emission pumping experiments with various excess energies or IR pump-probe experiments to study the vibrational relaxation process in the ground state. Time-resolved Anti-Stokes resonance Raman scattering investigations could also provide information on the excited state vibrational energy relaxation, but they are limited by the necessary frequency resolution of 10-15 cm-1 to about one picosecond in time resolution.
Finally, reliable semiempirical calculations including solvent polarization or ab-initio calculations (now also available for large chromophores [Schol 98, Krueg 98]) would be extremely useful to find out about possible pathways of photoexcited dyes on the multidimensional intra- and intermolecular energy hypersurface and design tailor-made experiments for the chromophore and problem under investigation.
© Die inhaltliche Zusammenstellung und Aufmachung dieser Publikation sowie die elektronische Verarbeitung sind urheberrechtlich geschützt. Jede Verwertung, die nicht ausdrücklich vom Urheberrechtsgesetz zugelassen ist, bedarf der vorherigen Zustimmung. Das gilt insbesondere für die Vervielfältigung, die Bearbeitung und Einspeicherung und Verarbeitung in elektronische Systeme.
DiML DTD Version 2.0||
der Humboldt-Universität zu Berlin
|HTML - Version erstellt am:|
Tue Apr 4 13:42:16 2000