The photophysics of 2-cyanoindole probed by femtosecond spectroscopy

The photophysics of 2-cyanoindole (2-CI) in solution (water, 2,2,2-trifluoroethanol, acetonitrile‚ and tetrahydrofuran) was investigated by steady-state as well as time resolved fluorescence and absorption spectroscopy. The fluorescence quantum yield of 2-cyanoindole is strongly sensitive to the solvent. In water the quantum yield is as low as 4.4 × 10–4. In tetrahydrofuran, it amounts to a yield of 0.057. For 2-CI dissolved in water, a bi-exponential fluorescence decay with time constants of ∼1 ps and ∼8 ps is observed. For short wavelength excitation (266 nm) the initial fluorescence anisotropy is close to zero. For excitation with 310 nm it amounts to 0.2. In water, femtosecond transient absorption reveals that the fluorescence decay is solely due to internal conversion to the ground state. In aprotic solvents, the fluorescence decay takes much longer (acetonitrile: ∼900 ps, tetrahydrofuran: ∼2.6 ns) and intersystem crossing contributes.


Introduction
The side chain of the amino acid tryptophan consists of the hetero-aromatic indole moiety. This moiety contributes strongly to the intrinsic fluorescence of proteins [1]. Substitution of the indole moiety may alter its fluorescence properties which can be exploited, for instance, in biophysical studies on proteins and peptides [2][3][4][5][6]. The cyano-substituent is hereby of particular interest. It exhibits strong negative inductive and mesomeric (resonance) This publication is dedicated to Prof. Silvia E. Braslavsky, a pioneer in photobiology and photobiophysics, on the occasion of her 80th birthday. effects on the chromophore [7]. It may, thus, substantially alter the electronic structure of indole. Due to its size, a cyano-substituent causes only a small steric effect [8]. Some cyanoindoles exhibit strongly solvent dependent fluorescence lifetimes and quantum yields [9]. For the title compound 2-cyanoindole (2-CI, structure in Scheme 1) dissolved in tetrahydrofuran (THF), a fluorescence lifetime of 2.9 ns was measured [9]. Dissolving the compound in dimethylsulfoxide (DMSO) reduces the lifetime to 0.4 ns. For water as a solvent, only an upper boundary of 0.05 ns for the lifetime was reported (see also ref. [10]). Due to this dependence, the fluorescence of 2-CI and other cyanoindoles can serve as a sensor of the local environment of an amino acid residue in a protein. The cyano-group further offers the possibility to probe this environment via vibrational spectroscopy [11][12][13][14].
Here, femtosecond fluorescence and UV/Vis absorption experiments on the photophysics of 2-CI are reported. The experiments aim at the fluorescence lifetime of 2-CI in water for which heretofore only an upper boundary has been reported (see above). Via time-dependent fluorescence anisotropy measurement, information on the emissive state (L a or L b ) [15] is retrieved. With femtosecond UV/Vis absorption, transient states populated in the course of the fluorescence decay are detected. It is shown that in aprotic solvents intersystem crossing (ISC) contributes substantially to the decay. Experiments were conducted using solutions of 2-CI in water, 2,2,2-trifluoroethanol (TFE), acetonitrile (MeCN), and tetrahydrofuran (THF). TFE was selected to study the effect of the proticity on the photophysics. Like water, it is a protic solvent albeit with a lower dielectric constant. MeCN is a strongly polar but aprotic solvent and THF represents a solvent of moderate to small polarity.

Steady-state measurements
Steady-state absorption spectra were measured with a Lambda 19 spectrometer from Perkin Elmer. Fluorescence spectra were recorded on a FluoroMax-4 (Horiba Scientific). The excitation wavelength was set to 260 nm. 1 cm fused silica cuvettes (Hellma Analytics) were employed for these measurements. To determine the fluorescence quantum yields Φ fl , thymidine in water served as reference (Φ fl = 1.32 × 10 −4 [16]). Fluorescence excitation spectra were obtained by detecting the fluorescence signal around 340 nm. For both, emission and excitation spectra, the optical densities were kept below 0.05 (per cm) to avoid inner filter effect. Spectra were corrected for the spectral sensitivity of the instrument.

Time resolved fluorescence measurements
A femtosecond fluorescence Kerr gating was used to trace the decay of 2-CI in water for 266 and 310 nm excitation pulses. A detailed description for this setup was given before [17,18]. A Ti:Sa laser amplifier system (Libra, Coherent) served as pulse source with a repetition rate of 1 kHz, 100 fs pulse duration and a wavelength of 800 nm. To generate 266 nm pump pulses, a part of the output was first converted to a wavelength of 400 nm by frequency doubling (in a BBO crystal type I, 29°, 1 mm). Subsequently, from the doubled and the fundamental, the sum frequency was generated to yield a wavelength of 266 nm (in another BBO crystal, type II, 55.5°, 0.5 mm). The beam diameter was 80 μm (full width half maximum, FWHM) and the energy per pulse amounted to 1.1 μJ at the sample location. For the generation of 310 nm excitation pulses, a part of the output was fed in to a TOPAS-White non-collinear optical parametric amplifier system. The TOPAS was set to generate pulses peaking at 620 nm. By frequency doubling (BBO crystal, type I, 37°, 0.15 mm), they were converted to the pump wavelength of 310 nm with an energy per pulse of 0.9 μJ. The instrumental response function (IRF) was about 430 and 440 fs (FWHM) for 266 and 310 nm excitation light, respectively. The rather long IRF times can (partially) be attributed to the group velocity mismatch of fluorescence and gate light in the Kerr medium [19]. The optical gate operated as described before [18]. For the experiment with 266 nm excitation pulse, the integration time for each spectrum was set to 2.5 s. To record each scan, there were 50 steps from − 2 to 5 ps on a linear scale, then 35 steps on a logarithmic one up to 35 ps. For Scheme 1 Chemical structure of the molecule 2-CI. Purple and blue dashed arrows indicate the orientation of the transition dipole moments (TDM) of the L a and L b states [10] this measurement, a total of 12 scans were averaged. For the measurement with 310 nm excitation pump, the integration time was set to 1 s. Each scan covered a time range of 40 ps which consisted of 80 equidistant steps. For this measurement, a total of 40 scans were averaged. The solutions were circulated by a peristaltic pump (REGLO Analog MS-2/8 from ISMATEC ® ) through a flow cell (custom made QX, Hellma Analytics) with 1 mm path length. Measurements on solutions of 2-CI (concentrations of ~ 24 μM and ~ 32 μM for the excitation of 266 and 310 nm light, respectively) as well as the neat solvent were performed. Solvent signals were subtracted from the solution ones after proper scaling.

Time resolved fluorescence anisotropy measurements
Measurements of the fluorescence anisotropy for both 266 and 310 nm excitation pulses were performed on the Kerr gate set up as well. To this end, a fluorescence signal S || ( , t) with the pump pulse parallel to the first polarizer of the Kerr gate was recorded. Then the polarization of the pump light was rotated by 90° with a half-wave plate. Care was taken to ensure the same excitation energy for both polarizations were employed. The anisotropy was then calculated via [1] Thymidine in water was chosen for the calibrations of the set up. To record the parallel signal S || ( , t) the solution was excited at 266 nm and the integration time amounted to 1 s. The time range covered was 7 ps. The range was divided in 70 equidistant steps. It was averaged over 13 delay line scans. The perpendicular signal S ⟂ ( , t) was recorded with the same settings. Thymidine solution had a concentration of ~ 16 μM. The resulting anisotropy value from Eq. (1) was calculated. The initial anisotropy for thymidine in water was determined to be 0.31 ± 0.01. It compares favorably with the value of 0.35 reported by Gustavsson et al. [20]. For the anisotropy experiment on 2-CI solutions with 266 nm excitation, the same settings were applied to the parallel and perpendicular measurements. The integration time for each spectrum was set to 2.5 s. For recording each scan, there were 50 steps from − 2 to 5 ps linearly, then 35 steps on a logarithmic scale up to 35 ps. In total, 12 scans were averaged. To record the anisotropy measurement for the solution of 2-CI with 310 nm excitation light, the integration time amounted to 1 s. Each scan covered a time range of 40 ps which consisted of 80 linearly steps. For this measurement, a total of 7 scans were averaged. Time resolved anisotropy experiments were recorded as a function of detection wavelength and delay time, however, only one selected wavelength was shown (see result section). For dissolved 2-CI (concentrations of ~ 20 μM and ~ 30 μM for the excitation of 266 and 310 nm light, respectively) and the neat solvent data sets were recorded. After proper scaling the solvent signals were subtracted.
Briefly, for all measurements, a 1 kHz Ti:Sa femtosecond laser amplifier system (Libra, Coherent) was used to obtain the frequency-tripled output (266 nm) as pump pulses. The beam had a diameter of 160 μm on the sample position and the pump pulse energy at the sample was adjusted to ~ 1 μJ. Absorption changes were probed by a single-filament white light continuum generated in CaF 2 with a diameter of 100 μm on the sample. The relative orientation of pump and probe polarization time was set at the magic angle. The instrumental response time was ∼180 fs (FWHM). Each scan consisted of 139 steps, from -1 to 1 ps being equidistant on a linear time scale and between 1 ps and 3.4 ns on a logarithmic time scale. 2000 spectra were recorded and a total of 4 scans were averaged for the measurement. Using a flow cell (Hellma, Suprasil) with 1 mm optical path length, sample solutions were circulated over the course of a measurement. The absorption of samples was set to ~ 0.7 per mm at the excitation wavelength. The solvent signal was recorded as a separate measurement and subtracted with proper scaling [24]. The instrumental time zero shift was determined as a function of wavelength via the optical Kerr effect and corrected for.

Nanosecond transient absorption spectroscopy
A detailed description of the setup is given elsewhere [18]. Nanosecond transient absorption spectra were acquired using an Edinburgh Instruments LP980 spectrometer in a right-angle geometry. The fourth harmonic (266 nm) pulses from a Nd:YAG laser (Spitlight 600, Innolas, 5 Hz repetition rate, pulse duration of 12 ns (FWHM)) served as the excitation. The beam had a diameter of ∼8 mm. The average pulse energy was set to 2 mJ. As a probe beam a pulsed xenon lamp (Osram XBO 150 W/CR OFR) was utilized. A photomultiplier (Hamamatsu, PMT-900) was used to detect the transmitted probe light after being dispersed by a grating monochromator. Every 5 nm, kinetic traces were acquired and averaged over 64 acquisitions to obtain transient spectra.
The sample solution was circulated by a peristaltic pump (REGLO Analog MS-2/8 from ISMATEC ® ) through a rectangular flow cell. In the direction of the pump its length amounted to 5 mm, in the probe direction to 10 mm. The samples absorption at the excitation wavelength amount to 1.6 per cm.

Data analysis
The measured data sets S(λ, t) were analyzed globally with multi-exponential fit function convoluted with the instrumental response function (IRF): The procedure yields time constants τ i , decay associated spectra S i (λ) (DAS, in the case of fluorescence measurements) and decay associated difference spectra (DADS, in the case of absorption measurements) [25,26]. A spectrum S i ( ) parametrizes the spectral changes in the course of a process with time constant τ i . For S i ( ) > 0 , it describes a signal decay and for S i ( ) < 0 a rise. The sum of the spectra

Steady state spectroscopy of 2-CI in different solvents
The absorption band of 2-CI lowest in energy is located at wavenumbers larger than 30,000 cm −1 (see Fig. 1, Table 1 and ref. [9]). The band consists of a shoulder around 32,500 cm −1 and a maximum at 35,500 cm −1 . The absorption band features a weak positive solvatochromism, i.e., it shifts to smaller wavenumbers with increasing solvent polarity. Peak absorption coefficients max determined here are in the range 17,000-18,000 M −1 cm −1 and depend only slightly on the solvent (see also Table 1). The present value for water of 17,800 M −1 cm −1 is somewhat higher than the value of 16,000 M −1 cm −1 reported earlier [9].
The fluorescence emission spectra of 2-CI in the different solvents peak around 30,000 cm −1 (see Fig. 1 and Table 1). As the absorption spectra they are subject to a weak positive solvatochromism. The impact of the solvent on the fluorescence quantum yield Φ fl is pronounced. In water, it amounts to 4.4×10 -4 . In THF, it is two orders of magnitude higher. This dependence is in agreement with previous studies Fig. 1 a Fluorescence (logarithmic y-axis, solid lines) and absorption (coefficients, dashed lines) spectra of 2-CI in different solvents. The excitation wavelength for the fluorescence spectra was tuned to 260 nm. Fluorescence spectra are scaled so that their integrals are proportional to their fluorescence quantum yields Φ fl . b Fluorescence excitation of 2-CI in comparison with its absorption spectrum. For the excitation spectrum the signal was probed at 340 nm Table 1 Photophysical properties of 2-CI derived from steady-state experiments Absorption maxima λ abs of the absorption band lowest in energy are listed. Absorption coefficients ɛ max correspond to the strongest absorption maxima. Wavelengths λ abs and λ em are retrieved from spectra obtained with constant wavelength bandpass. Experimental fluorescence quantum yields Φ fl were determined with thymidine in water as a reference (Φ fl = 1.32 × 10 −4 [16]). The radiative rate constants k rad were determined via a Strickler-Berg analysis as described in the text. The last row gives predicted fluorescence lifetimes τ fl SB , whereby τ fl SB = 1 3 [9]. The fluorescence excitation spectra overlay favorably with the properly rescaled absorption spectrum exemplified in Fig. 1b for 2-CI in water. 2-CI, thus, obeys the Kasha-Vavilov rule [27].
To arrive at a first estimate for the excited state lifetime, particularly for 2-CI in water, a Strickler-Berg analysis was undertaken [28,29]. For 2-CI, such an analysis is hampered by overlapping electronic transitions around 30,000 cm −1 . According to quantum chemical calculations [15] and gas phase spectroscopy [10] two * transitions termed L a and L b are located in this energetic region. For 2-CI, the L b transition was found to be lowest in energy. The L b transition features a smaller oscillator strength ( f = 0.092) than the L a one ( f = 0.307) [15]. To isolate the L b contribution to the band around 30,000 cm −1 , a procedure developed earlier [30] was used. The procedure rests on the assumption that a mirror image relation [29] exists between the fluorescence spectrum and that part of the absorption spectrum from which the fluorescence originates. Thus, the 2-CI fluorescence spectrum plotted on a wavenumber axis was flipped and shifted to overlay with the onset of the absorption spectrum (see Fig. 2). Hereby, the "trivial" frequency dependences of spontaneous emission (fluorescence) and absorption [29] was taken into account. The height of the flipped and shifted spectrum was scaled so that it approximately matches the absorption coefficient spectrum of 2-CI. The spectrum generated thereby entered the Strickler-Berg analysis. The radiative rate constants k rad obtained are of the order of 0.5 ×10 8 s −1 (see Table 1). With these rate constants and the fluorescence quantum yields Φ fl , fluorescence lifetimes SB fl can be predicted, according to SB fl = Φ fl k rad . The lifetimes obtained for 2-CI in THF ( SB fl = 2500 ps) and in MeCN ( SB fl = 850 ps) compare favorably with the ones measured directly (2900 and 900 ps [9]). The lifetimes for the other two solvents will be considered in conjunction with the time resolved measurements.

Time resolved fluorescence spectroscopy of 2-CI in water
In previous studies on the 2-CI fluorescence [9,10], its decay in water was not reported due to the limited time resolution of the instruments. Here, the decay is traced by femtosecond fluorescence Kerr gating [17,18]. In these experiments, solutions of 2-CI in water were excited at 266 and 310 nm. For 266 nm excitation, 2-CI should predominately be promoted to the L a state, the 310 nm excitation should favor L b excitation [15]. 266 nm excitation causes a fluorescence signal peaking at 341 nm ( Fig. 3) which is close to the value of the steady-state spectrum (cf. Fig. 1). The signal decays to zero on the 10 ps time scale. Closer inspection reveals a slight red-shift and broadening within ~ 1 ps. To retrieve time constants, the spectro-temporal behavior of the fluorescence was subject to a global analysis. A multi-exponential ansatz convoluted with instrumental response function served as a trial function. Two exponential terms proved necessary for a satisfactory fit of the measurement. The time constants retrieved are fl 1 =1.5 ps and fl 2 =7.9 ps (see Table 2). The value of fl 2 is close to the estimate based on the Strickler-Berg treatment (see above). The decay associated spectrum DAS 1 for the time constant fl 1 peaks at 328 nm and its height amounts to only ~ 1/3 of the one of DAS 2 which peaks at 351 nm (Fig. 5). By adjusting the polarization of the excitation light parallel and perpendicular positions, the time dependent anisotropy r(t) was recorded. The anisotropy is close to zero throughout the time range covered. Excitation of 2-CI in water with 310 nm pulses results in a time dependent fluorescence signal similar to the one for 266 nm excitation (Fig. 4). The emission again peaks around 341 nm and decays on the 10 ps time scale. The time constants retrieved by the global analysis amount to τ fl 1 = 2.5 ps and τ fl 2 = 7.6 ps and are, thus, close to the values for the 266 nm excitation. Also, the DAS exhibit similar shapes and relative heights (cf. Fig. 5). The time dependent anisotropy r(t) does differ from the 266 nm experiment. Its initial value is 0.2. Within ~ 5 ps it decays to zero. However, after a few picoseconds the r(t) curve fluctuates strongly since the fluorescence signals contributing go to zero as well. Therefore, this decay behavior of the anisotropy should not be interpreted.
Measurements with both excitation wavelengths revealed a bi-exponential decay when analyzing them with multiexponential trial function. It is well known that on the picosecond time scale fluorescence spectra red-shift due to dielectric relaxation (dynamic Stokes shift) [31]. Such a shift can "feign" a multi-exponential decay if the data set is analyzed with the trial function used here [32]. Thus, to clarify whether the bi-exponential decay is genuine, a second ansatz for the analysis was used. A normalized spectrally integrated fluorescence signal I fl (t) was computed according to Hereby, S fl (̃ , t) is the wavenumber ̃ and time t dependent fluorescence signal. S fl (̃ , 0) is this signal at time zero. The division by ̃ 3 cancels the trivial wavenumber or frequency dependence [29]. The average time dependent emission wavenumber ⟨̃ ⟩(t) was computed via As the above description has shown, the signals S fl (̃ , t) are very similar for excitation wavelengths of 266 and 310 nm. The 266 nm measurement offers a larger spectral coverage and a higher signal to noise ratio. The analysis was, thus, performed on the 266 nm measurement.
The time dependence of the spectrally integrated signal I fl (t) (Fig. 6) is very similar to wavelength resolved time traces depicted in Fig. 3. Indeed, a bi-exponential fit of I fl (t) taking the IRF into account yields time constants τ int (relative amplitude 0.4) and τ int fl 2 =8 ps (0.6) which are very close to those determined by the above global analysis. The average fluorescence wavenumber at time zero ⟨̃ ⟩(0) amounts to 29,300 cm −1 . It shifts to 27,900 cm −1 for "infinite" times. According to a single exponential fit, the time constant for this shift is 0.33 ps. This time constant is close to the average solvent relaxation time of water (0.4 ps [31]).

Time resolved UV/Vis absorption spectroscopy of 2-CI in different solvents
Obviously, time resolved fluorescence spectroscopy cannot reveal which states are populated in the course of the depletion of the primarily excited singlet state. Therefore, also time resolved UV/Vis absorption experiments were  conducted. According to the above fluorescence experiments, the excitation wavelength has little impact on the decay kinetics. Therefore, for all absorption experiments, the excitation was tuned to 266 nm. For 2-CI in water, the excitation results in positive absorption changes throughout the spectral range covered (Fig. 7). No negative signals due to ground state bleach (GSB) or stimulated emission (SE) are recorded. The absence of GSB is due to small vanishing absorption coefficients of 2-CI in this range, cf. Fig. 1. The SE signal is expected in the spectral region of the fluorescence. Its absence indicates that it is overcompensated by excited state absorption (ESA). The ESA at time zero is characterized by peaks at 440 nm and ~ 750 nm. Within ~ 1 ps a slight signal decrease around 435 nm and a weak increase around 570 nm is observed. This results in a less structured difference spectrum. The overall signal decays to essentially zero on the 10 ps time scale. The behavior in another protic solvent, namely TFE, is similar (Fig. 7). Around time zero the ESA peaks mentioned before at ~ 430 and 720 nm are observed. In addition, a positive difference absorption feature increasing toward the UV is discernible. Again, minor signal changes within a few picoseconds are recorded. The overall signal decay is slower than in water occurring with 20-30 ps. After that decay a weak absorption band peaking at 405 nm persists until the end of time range covered (3 ns).  In the aprotic solvents MeCN and THF, similar spectral signatures as in the protic ones are observed around time zero (Fig. 8). Again, small spectral changes occur within a few picoseconds. The decay of the initial signature now takes much longer and occurs on the nanosecond time range. This is in line with the solvent dependence of the fluorescence lifetime reported earlier [9]. Importantly, on this time scale also a signal increase around 410 nm is observed. This rise is due to the population of a longer lived transient species-presumably the triplet state of 2-CI. According to the transient spectra at 3 ns, the species features an absorption peak at 420 nm and a shoulder at 380 nm. The same signature was observed in a nanosecond transient absorption experiment (Fig. 8). According to these experiments, the species exhibits a lifetime in the microsecond range in deoxygenated solutions (1.56 μs for 2-CI in MeCN as well as THF) which is in line with the tentative triplet assignment [33].
For a quantitative comparison of the 2-CI transient absorption in different solvents, the same analysis approach as for time resolved fluorescence was adopted. Measurements in all solvents could be fitted with a tri-exponential trial function. Hereby, one time constant was set to infinity to account for the signal persisting longer than 3 ns ("offset" in the following). The first time constant 1 ranges between 0.9 ps (water) and 5.3 ps (THF), see Table 2. The value for water is close to one determined by time resolved fluorescence spectroscopy ( fl 1 =1.5 ps). The respective decay associated difference spectra DADS 1 are similar to each other with maxima around 460 nm and minima at 560 nm (Fig. 9). The minima are more pronounced in protic solvent. Relative to the other DADS the amplitudes are smaller and, thus, DADS 1 parametrizes small spectral changes. The second time constant τ 2 spreads between 8.3 ps (water) to 2600 ps (THF). The value for water is again very close to fluorescence time constant fl 2 =7.9 ps. The values for MeCN ( 2 =640 ps) and THF ( 2 =2600 ps) are close to the reported fluorescence lifetimes (900 and 2900 ps [9]). In the DADS 2 maxima are located at 420, 560, and ~ 720 nm for the protic solvents. In the aprotic solvents, minima are found at 410 nm. These negative features parametrize the rise due to the population of a species with "infinite" lifetime. Accordingly, in the DADS 3 positive bands around 410 nm are seen for the aprotic solvents. For the protic ones, a weak if any signature is seen here.

Discussion
The essential experimental findings of this study are the first report of the fluorescence decay of 2-CI in water, the anisotropy of this fluorescence, the bi-exponential decay of the singlet state in all solvents considered as well as the population of the triplet state in aprotic solvents. Steady state (Strickler-Berg analysis) as well as time resolved fluorescence and absorption spectroscopy reveal a fluorescence lifetime of ~ 8 ps for 2-CI in water. In the Strickler-Berg analysis, one only arrives at this value when isolating the L b contribution to the absorption band lowest in energy (cf. Fig. 2). This gives strong evidence that the emission originates from the L b state. As all lifetimes derived via Strickler-Berg match the ones measured directly (cf. Tables 1 and 2), emission  Fig. 3 using spectral integrals. a The spectrally integrated fluorescence signal I fl (t) (cf. Equation (3)) is plotted as a function of time (dotted black line). The red line represents the result of bi-exponential fit which takes the IRF into account. b The evolution of the average emission wavenumber ⟨̃ ⟩(t) (cf. Equation (4), dotted black line) and the result of a single-exponential fit (solid red line) seems to occur from the L b state in all solvents. The L b lifetime decreases with increasing solvent polarity and proticity. The impact of the latter is particularly pronounced as a comparison of the behavior in MeCN (aprotic) and TFE (protic) shows. In MeCN (dielectric constant of 35.9 [33]) the lifetime is around 900 ps. In TFE with its smaller dielectric constant (26.7 [33]) the lifetime is around 30 ps.
The fluorescence anisotropy experiment confirms the emission out of the L b state. With 266 nm excitation the (initial) anisotropy of the transient fluorescence is close to zero. As rotational depolarization occurs on a much longer time scale [34,35] than the IRF time (~ 430 fs) of the experiment, this anisotropy value has to be related to non-parallel transition dipoles for excitation and emission. An initial anisotropy of zero implies that the two transition dipoles span an angle of 54.7° [34]. According to quantum chemical computations, the angle between the transition dipoles of the L a and L b states are very close to this value, namely 57° [15]. According to these computations, the 266 nm excitation addresses the L a state. Within our IRF time (~ 430 fs), an internal conversion (IC) transition to the L b state seems to occur from which all fluorescence detected here originates. Such a short transition time is in agreement with recent measurement on tryptophan, for which the L a → L b transition was shown to occur within less than 50 fs [36]. A similar transition time was derived from fluorescence anisotropy measurements on 5-methoxyindole [37]. The anisotropy of zero implies that also in terms of adiabatic energies the L b state is the lowest. Thus, no L a -L b state switching occurs (situation as described by Fig. 8a in ref. [15]). Hebestreit et al. [10] have calculated the excited state dipole moments of the L a and L b state, respectively, and compared them to the experimental values from electronic Stark spectroscopy. They found a considerably smaller dipole moment for the L a (3.44D) compared to the L b state (4.95D) in 2-CI. This behavior is strikingly different from that of the other substituted indoles and also from the parent indole, in which the dipole moment of the L a state is always larger than of the L b , and 5-CI [40] the L a dipole moment is the larger one, leading to state inversion upon solvation in strongly polar solvents. It can thus be stated safely that it is the small dipole moment of the L a state in 2-CI, which impedes state switching here. One has, however, to keep in mind that the notation of L a and L b for the excited ππ* states is somehow misleading, since the states tend to mix considerably and the original notation, introduced by Platt for cata-condensed hydrocarbons [41] and later extended to indole by Weber [42] is difficult to employ, the closer both ππ* states are located.
For 310 nm excitation, the (initial) anisotropy amounts to ~ 0.2. For this excitation wavelength, the L b state is predominately excited [15] and from this state emission ought to occur. Thus, only one transition dipole is involved and the initial anisotropy should amount to 0.4 [34]. The observation that the experimental value is smaller than this prediction could be due to two reasons. (i) Even in chromophores with no overlapping electronic transitions the initial anisotropies are commonly somewhat smaller than 0.4 [1]. This matches our observation on the anisotropy behavior of thymidine in water which was investigated for calibration purposes (see Experimental Section). For thymidine, an initial anisotropy of 0.4 is expected. Our measurement yielded a value of ~ 0.3 in good agreement with an earlier study [20]. (ii) The L a and L b transition partially overlap and the L a transition exhibit a much larger oscillator strength. Therefore, it is very likely that with the 310 nm excitation the L b state is not excited exclusively. As discussed above a partial L a excitation ought to shift the anisotropy toward zero. Presumably both contributions are responsible for the value of 0.2 for the 310 nm excitation.
These anisotropy experiments give clear evidence that for 2-CI in water the L b state is populated within less than ~ 100 fs irrespective of the excitation wavelength. The similarity of the transient absorption signatures around time zero and the above Strickler-Berg argument strongly suggest that this also implies for all other solvents. The depopulation of the L b state (time constant 2 ) occurs within ~ 8 ps to ~ 2500 ps, depending on the solvent. This depopulation goes along with the fluorescence decay. The solvent dependence of the time constant 2 can be rationalized by a crossing of the L b state with a 1 πσ* one as put forth by Domcke and Sobolewski [43,44]. The 1 πσ* state gets stabilized in polar solvents, thereby reducing the barrier for the IC transition. The present results indicate that in addition to the polarity also the proticity of the solvents has a huge impact on the lifetime 2 . Gas phase data have shown that 2-CI forms binary cluster with water. Water seems to act as a hydrogen bond acceptor in this cluster [45]. For the isolated 3-CI-(H 2 O) 1 cluster a direct lifetime measurement shows a distinct shortening of the fluorescence lifetime from 9.8 ns (monomer) to 3.6 ns (1:1 water cluster) [38]. Determination of fluorescence lifetimes for different cluster sizes and different solvent clusters are on the way to resolve the problem of how the solvent influences the fate of the excited state.
Prior to that a process with the time constant 1 of a few picoseconds occurs in all solvents. The analysis based on band integrals (cf. Fig. 6) excludes a dynamic Stokes shift as the underlying mechanism. Also, for the other solvents, the dynamic Stokes shift is an unlikely explanation. In acetonitrile, for instance, the time constant 1 amounts to 2.8 ps as compared to 0.9 ps for water, despite the fact that dielectric relaxation in acetonitrile is even faster than in water [31]. The underlying process might be vibrational cooling which is known to occur on the picosecond time scale [46,47]. Yet, also vibrational cooling ought to retain the band integral [47][48][49]. Furthermore, for 2-CI in water, the 1 process is observed for 266 and 310 nm excitation. The 266 nm excitation deposits ~ 5000 cm −1 more vibrational excess energy in the molecule than the 310 nm one. If the 1 process were related to vibrational cooling, it should be less pronounced for 310 nm excitation. As this is not the case (cf. Fig. 5), vibrational cooling as the underlying process seems unlikely. Considering the rigid structure of 2-CI, different conformers and thereby a kinetic heterogeneity do not seem to be a plausible explanation. Aggregation can also cause such a heterogeneity [50,51]. Yet, aggregation is expected to be solvent dependent [52][53][54][55]. As a bi-exponential behavior with similar relative amplitudes is observed in all solvents considered, also aggregation is not a probable explanation. Thus, at present, no consistent assignment can be given.
The assignment of the time constant 3 is less involved. The respective transient peaks around 420 nm (cf. Fig. 8) which are close to the absorption peak of the indole triplet in water [56][57][58][59][60]. In the respective experiments on indole [57], the triplet signature was accompanied by the ones of the indole cation and the solvated electron. In the femto-and nanosecond experiments on 2-CI these signatures are absent. This might be related with the strong electron withdrawing character of the cyano substituent. The measured time constant 3 for 2-CI in MeCN and THF is well in line with triplet lifetimes in general [33]. For indole and derivatives lifetimes in the range 12-50 µs were reported [56-58, 60, 61]. The value of ~ 2 µs is, thus, on the shorter side. We assign the species decaying with the time constant 3 to the triplet state of 2-CI. The signal of this state is nearly negligible in water as well as TFE and much larger in MeCN and THF (see DADS 3 in Fig. 9). As at present, no difference absorption coefficients of the 2-CI triplet are available, the triplet quantum yield Φ T cannot be quantified. Assuming that the difference absorption coefficients are similar in all solvents and normalizing to identical excitation conditions, one can derive the following relative yields (with respect to MeCN): water (0.008), TFE (0.034), MeCN (1), and THF (0.7). Thus, triplet yields scale approximately as the lifetimes 2 . This suggests that the rate constant (k ISC ) for intersystem crossing is similar in all solvents. In protic solvents, ISC is outcompeted by IC.

Conclusion
Femtosecond transient fluorescence experiments on 2-CI in water revealed a fluorescence decay of ∼8 ps for both excitation wavelengths of 266 and 310 nm (see Fig. 10). This is a result of the ultrafast depletion of the L a state in less than 100 fs. The respective anisotropy experiments provided strong evidence of the origin of this emission from the L b state, regardless of the excitation wavelength. In time resolved transient absorption measurements, for all solvents, a bi-exponential decay was observed. The longer of the two ( 2 ) proved to be strongly solvent dependent. In aprotic solvents (MeCN and THF), intersystem crossing to the triplet state contributes to the singlet decay. The triplet assignment was confirmed by nanosecond transient absorption experiments. The relative triplet yields are proportional to the lifetime 2 . Quantum chemical computations aiming at the nature of the 1 process including channels through the 1 πσ* state are under way.
Funding Open Access funding enabled and organized by Projekt DEAL.

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations
Conflict of interest On behalf of all authors, the corresponding author states that there is no conflict of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.