Excited-state intramolecular proton transfer of 2-acetylindan-1,3-dione studied by ultrafast absorption and fluorescence spectroscopy

We employ transient absorption from the deep-UV to the visible region and fluorescence upconversion to investigate the photoinduced excited-state intramolecular proton-transfer dynamics in a biologically relevant drug molecule, 2-acetylindan-1,3-dione. The molecule is a ß-diketone which in the electronic ground state exists as exocyclic enol with an intramolecular H-bond. Upon electronic excitation at 300 nm, the first excited state of the exocyclic enol is initially populated, followed by ultrafast proton transfer (≈160 fs) to form the vibrationally hot endocyclic enol. Subsequently, solvent-induced vibrational relaxation takes place (≈10 ps) followed by decay (≈390 ps) to the corresponding ground state.


I. INTRODUCTION
Excited-state intramolecular proton transfer (ESIPT) is one of the most successful model systems for the examination of complex proton transfer dynamics. [1][2][3] In a photoinduced ESIPT reaction, a proton moves from the donor to an acceptor group within the same molecule upon a change in electronic charge distribution in response to an electronic excitation. In general, ESIPT is an ultrafast process occurring on a subpicosecond time scale, but it is also reflected in spectral changes of steady-state UV/visible (VIS) absorption and fluorescence. The fluorescence emitted by the ESIPT product's excited state that is energetically much more stable than the one of the initial electronically excited reactants, not only significantly deviates from the mirror image of the absorption but is also strongly red-shifted with a Stokes shift ranging from 6000 to 10 000 cm À1 , as seen in the landmark work by Weller on methyl salicylate. 4,5 The origin of strongly red-shifted fluorescence spectra in these and related compounds is believed to originate from a profound rearrangement of the molecular structure, resulting from proton transfer along a pre-existing intramolecular H-bond after electronic excitation. To directly observe ESIPT in the electronically excited state, time-resolved measurements with femto-to picosecond time resolution are required. The real-time dynamics of such a proton transfer in various ESIPT molecular systems have been explored in the gas phase, 6-10 techniques applied to ESIPT in solution comprise time-resolved fluorescence based on streak cameras, 11 time-correlated single-photon counting, [12][13][14] and upconversion, 12,[15][16][17] as well as transient absorption (TA) with pump/probe combinations being UV/mid-IR, 2,16,[18][19][20] UV/VIS, [12][13][14]16,[19][20][21][22][23][24][25][26][27][28][29] and recently UV/UV. 27 have shown that ESIPT typically occurs on a subpicosecond  time scale, with a key role for the low-frequency Raman-active modes modulating the H-bond  distance between the proton donor and the acceptor atom. 11,25,30-38 2-acetylindan-1,3-dione (AID) is a ß-diketone with an exocyclic double bond and has been shown to exist as exocyclic enol form (see molecular structure in Figure 1) in solution by various theoretical and experimental studies. [39][40][41][42][43] Apart from being a bio-active drug molecule with antibacterial and anticoagulant activities, it is a promising chelating and sunscreen agent. [44][45][46] Its six-membered ring structure via intramolecular H-bonding is of special interest from the viewpoint of ESIPT. Steady-state measurements of AID reveal that the fluorescence spectrum is not a mirror image and largely red-shifted compared with its absorption spectrum, suggesting ESIPT. 42 Quantum-chemical calculations support the possibility of ESIPT in AID. 42 Gaining further insight into the photochemical processes taking place in the excited state of AID necessitates time-resolved spectroscopic studies.
In this work, we combine results from fluorescence upconversion in the VIS with TA covering a broad spectral range from the deep-UV to the VIS spectral region. In contrast to the previous studies on ESIPT in other molecules, this allows not only the observation of signatures from the excited state of the involved species but also of the ground-state bleach (GSB) of the reactant. The observed dynamics are discussed in relation to ESIPT, solvation, and vibrational relaxation processes.

II. MATERIAL AND METHODS
AID (CAS No. 1133-72-8) was obtained from Sigma-Aldrich and used as received. Spectroscopic-grade solvents were also obtained from Sigma-Aldrich. The steady-state absorption and fluorescence spectra were recorded on a Jasco V670 UV-VIS and an Edinburgh Instruments FLS900 spectrometer, respectively. The wavelength-dependent sensitivity of the fluorometer was checked by comparing a measurement of Coumarin 153 with a reference spectrum from Ref. 52. For the time-resolved measurements, a fresh sample of 5 Â 10 À4 M concentration was prepared each time.
A detailed description of the utilized TA setup can be found elsewhere. 27 Briefly, a fraction of the fundamental beam from a commercial Ti:sapphire regenerative-amplifier laser system (Spitfire Pro, Spectra-Physics, center wavelength k c ¼ 800 nm, pulse duration Ds ¼ 120 fs, repetition rate rep ¼ 1 kHz) was used to seed a commercial non-collinear optical parametric amplifier (TOPAS White, Light Conversion) which was frequency-doubled to generate the UV pump beam. Another part of the fundamental beam was used to generate the third harmonic (267 nm) which was then focused into a linearly moving 5 mm thick CaF 2 window to generate a broadband UV probe continuum ranging from 220 to 330 nm. A custom-designed spectral filter (Laser Components) was used to suppress the high spectral power around 267 nm contained within the broadband UV probe beam, which also caused unwanted reduction in the absorption intensity around 267 6 8 nm. For the VIS probe continuum, the fundamental beam, rather than the third harmonic, of the amplifier system was focused into the moving CaF 2 . Pump and probe were focused in a nearly collinear geometry at the sample position with magic-angle configuration for the linear polarization directions. 47 After passing through the sample, the probe beam was spectrally resolved via a spectrograph (Acton SP2500i) equipped with a CCD camera (Princeton Instruments Pixis 2 k). The detection was performed on a shot-to-shot basis, i.e., a mechanical chopper was utilized to block every other pump pulse. Thus, consecutive probe signals were recorded to acquire pump-induced changes in the absorbance of the sample. The analysis of the TA maps, consisting of a three-dimensional dataset (wavelength, pump-probe delay, and absorbance change), was performed with the help of the software GloTarAn, which also accounts for instrument response function (IRF), dispersion, and coherent artifact during the fitting process. [48][49][50] TA measurements were also performed with the neat solvents only. Pumpinduced signals arising purely from the solvents were found to be negligible. 27,29 For the fluorescence upconversion measurement, a commercial setup (Halcyone, Ultrafast Systems) was used with the details provided elsewhere. 51 Briefly, part of the fundamental beam of another amplified Ti:Sapphire laser (Solstice, Spectra-Physics, k c ¼ 800 nm, Ds ¼ 100 fs, rep ¼ 1 kHz) was used to pump an optical parametric amplifier (TOPAS, Light Conversion) to generate the UV pump beam centered at 300 nm. The fluorescence collected from the sample was focused on a 0.5-mm b-barium borate (BBO) crystal for type-I frequency upconversion with another part of the fundamental beam acting as gate pulse. The gate pulse was timedelayed with respect to the fluorescence using a motorized optical delay line. The upconverted light was focused onto the entrance slit of a double monochromator and detected by a photomultiplier tube. The instrument response was estimated to be 350 fs by measuring stray light or Raman signals of the neat solvents. 51 The femtosecond fluorescence decays were deconvoluted globally (with OriginPro 2015 software) using a Gaussian-shaped IRF (fixed at 350 fs) and a multi-exponential function, Iðk; tÞ ¼ P i a i ðkÞ exp ½Àt=s i ðkÞ, where a i (k) are pre-exponential factors, s i (k) are the decay times, and P a i ðkÞ ¼ 1. Next, we computed a new set of intensity decays, normalized so that the time-integrated intensity at each wavelength was equal to the steady-state intensity at that wavelength. 52 The equation used was where I ss (k) is the steady-state emission spectrum and a 0 i ðkÞ ¼ HðkÞa i ðkÞ. 52 The values of I 0 ðk; tÞ were used to calculate the intensity at any wavelength and time, and thus the timeresolved emission spectra (TRES). The use of Eq. (1) is well-suited when the fluorescence signal decays to zero signal within the measured time window. The TRES were fitted to a lognormal line-shape function 53 as available in OriginPro 2015 software.
III. RESULTS Figure 1 shows the linear electronic absorption (solid) and emission spectra (dashed) of AID in hexane and acetonitrile. The electronic absorption spectrum consists of two bands in the region of 260-330 nm. The long-wavelength side exhibits a well-defined vibrational structure with two maxima of almost equal intensity peaking at 299 and 309 nm in hexane. The peaks undergo %2 nm red shifts on changing the solvent from non-polar (hexane) to polar (acetonitrile). The position of the other absorption maximum is negligibly dependent on the solvent polarity. On excitation with 300 nm light, fluorescence emission starting from 440 nm is observed peaking at 520 and 535 nm in hexane and acetonitrile, respectively. The fluorescence spectra are broad, structureless, and lack mirror symmetry with the longest-wavelength FIG. 1. The absorption (solid) and emission spectra (dashed, excitation wavelength 300 nm) of AID in hexane (black) and acetonitrile (orange). The absorption originates from AID molecules in the displayed tautomeric form. absorption band. In addition, there is a large Stokes shift of %13 000 cm À1 (calculated from the longest-wavelength absorption band and the fluorescence maximum). These changes in the linear spectra suggest a significant geometry change of the initially electronically excited molecule preceding emission caused by ESIPT. Like the longest-wavelength absorption band, the fluorescence maximum also shifts to the red upon increase of the solvent polarity.
To get a conclusive picture of the excited-state processes that set in after absorption of UV light (300 nm) in AID, we recorded TA of AID in acetonitrile covering from the deep-UV to the VIS spectral region (Figure 2). The time evolution can be divided into three different regions. The negative signal (blue) covering the spectral region of 240-285 nm corresponds to the GSB population. The other two spectral regions give rise to positive signals (yellow/red) around 285-300 nm and 430-610 nm and are due to the excited-state absorption (ESA) from the excited species, as will be discussed below.  Table I.
To determine whether the transient absorption signal observed in the VIS domain contains contributions from stimulated emission (SE) and to determine the excited-state lifetime, timeresolved fluorescence measurements were carried out. Fluorescence upconversion signals were measured in acetonitrile for the wavelength range from 470 to 580 nm with the excitation wavelength set to 300 nm. Four exponential components were found to be sufficient for the global fitting [Table II and Figure 4(a)] of the fluorescence intensity decays at all wavelengths for AID in acetonitrile while in hexane, only three exponential components were required (Table II and Figure 5). The fluorescence decay for each wavelength has an ultrafast rise component TA map from Fig. 2(a)  (s 1 ). Though the determined value for s 1 is close to the time resolution of the instrument (IRF: 350 fs), it is needed in the global fitting, otherwise the fitting is poor for early delay times (up to %500 fs). The decay times s 2 and s 3 are connected to a spectral shift, which can be described satisfactorily by a sum of exponentials. 17,19,53 The fourth component (s 4 ) represents the excited-state lifetime. The best fitted curves for three kinetic traces at wavelengths from the blue edge (470 nm), center region (500 nm), and red edge (570 nm) of the AID fluorescence spectrum are shown in Figure 4(a) along with the result of the global fit. Although the fluorescence appears instantaneously after photoexcitation at all the observed wavelengths, the subsequent dynamics vary significantly. At all wavelengths, a rising component of 0.25 (s 1 ) ps is observed, which represents ESIPT. At lower wavelengths, the fluorescence exhibits a fast decay (s 2 < 1 ps) whose amplitude first increases slightly followed by subsequent decrease in amplitude with increasing wavelength. The relatively slower time component (s 3 % 10 ps), which first appears as decay at the blue edge, becomes a rising component at the red edge. At a time scale beyond 10 ps, the fluorescence at all wavelengths exhibits a mono-exponential decay with a time constant of s 3 % 390 ps, reflecting the lifetime of the emitting excited-state of AID.
The reconstructed TRES are shown in Figure 4(b). Both the tautomer A and B exist right after excitation because the reconstructed TRES can be fitted well with a log-normal function having two peaks one centered at 453 nm [dark grey color in Figure 4 Figure 4(b)]. Interestingly, while going from 0 ps to 0.35 ps, not only there is an increase in the intensity (area) of the tautomer B but also the TRES fitting just requires a single peak log-normal function from then on. The emission maximum undergoes a continuous red-shift with increasing delay time up to 10 ps at which the emission maximum is close to the steady-state emission value.

nm [light grey color in
In non-polar solvents like hexane where solvent-shell organization is less important (but may contribute weakly 54 ), we also carried out fluorescence upconversion experiments; the corresponding data are displayed in Figure 5. There is a marked difference (Table II) between the fluorescence decay of AID in hexane compared to that in acetonitrile at 470 nm, but the decay at the red side is similar for both solvents, with only marginal differences above 500 nm.

A. Accessible tautomers
In the electronic ground state (S 0 ), the most stable conformer is the tautomer A (S 0 A ; reactant state) with exocyclic double bond (Scheme 1), in agreement with experimental results and calculations. [39][40][41][42][43] Therefore, other rotamers are not considered as reactants, as they have not been spectroscopically observed in solution at room temperature. However, in the excited state (S 1 ), this is no longer the case, with tautomer B (S 1 B ; ESIPT product state) being lower than S 1 A by 85.39 kJ/mol (7138 cm À1 ). 42 This inversion of the energy order of the two tautomers in the S 1 state compared with S 0 initiates ESIPT. Enchev and co-workers 42 further showed that the S 1 state exhibits a much lower potential barrier (16.86 kJ/mol) for A ! B conversion compared with the barrier (72.80 kJ/mol) in S 0 state for B ! A conversion and vibrational analysis confirmed that tautomers A and B are the true minima in the S 0 and S 1 state and that the transition state in both ground and first excited electronic states has only one imaginary frequency 55 corresponding to the proton transfer coordinate. Their AM1 calculations 42 also identified an increase of 3.31 D in the dipole moment during the ESIPT reaction (S 1 A to S 1 B ) which is consistent with the small shift of the emission maxima of AID from non-polar (hexane) to polar (acetonitrile) solvent (confer Figure 1). Note that there are also molecules exhibiting ESIPT where the emission blue-shifts with increasing polarity, indicating a reduction of the dipole moment when going to the excited product state. 12 Based on spectroscopic data and the available quantum-chemical calculations, the excitedstate processes in AID can be summarized as follows. The electronic excitation of AID promotes SCHEME 1. An illustrative energy diagram connecting the key steps of the photoinduced processes of AID upon electronic excitation. the tautomer A to S 1 A via a vertical transition, subsequent rapid charge distribution results in an electronic potential surface possessing a slope, albeit with a small barrier, 42 and an energy minimum shifted towards the proton position in the tautomer B. This redistribution displaces some of the normal modes of the molecule from their equilibrium positions, which are determined by the new potential surface. As a result, the excited system moves on S 1 toward the new equilibrium position. Once the redistribution is over (expectedly occurring on a time scale below 100 fs), the low-frequency, large-amplitude modes (e.g., either in-or out-of-plane motions of the skeletal modes) change the relative position of atoms associated with the H-bond and hence channel into the proton transfer process. 25,56 Note that the excitation energy (300 nm; 33 333 cm À1 ) used in the present study is close to the onset of the AID (S 1 ! S 0 ) transition and hence it is likely that the lowest energy level of the tautomer A is directly excited.

B. Signatures of S 1 A and of ESIPT
Depending on the solvent, fluorescence spectra comprise contributions from both S 1 A and S 1 B , giving rise to "dual" fluorescence. 19 Since ESIPT is ultrafast, emission yields from the excited reactant state however may be very low (see, e.g., in Ref. 12 where the emission yield is in the order of 10 À5 -10 À6 for 10-hydroxybenzo[h]quinolone). Nevertheless, the observation of a corresponding emission lasting only until ESIPT is complete has been reported for different molecules. 17,23 The fitting of the TRES (Figure 4(b), dark grey area) recognizes such a weak emission of tautomer A (only a tail of the complete spectrum). Such a dual fluorescence right after excitation in molecules undergoing either intra-or intermolecular proton transfer has been reported earlier. 17,57 Due to the continuum probing both in the deep-UV and the VIS, we have access to two other signals evidencing how fast S 1 A is depopulated. The ultrafast ESA decay in the deep-UV region [ Figure 3 , respectively. Thus, the proton transfer time based on our TA measurement is close to %160 fs (process "a" in Scheme 1) for AID which is similar to the proton transfer rates of related ESIPT molecular systems. 12,14,16,23 Our fluorescence upconversion measurement also recognizes such an ultrafast ESIPT signature, however, the determined time constant s 1 (ESIPT) is smaller than our time resolution.

C. Solvent reorganization
The time-dependent spectral shift can arise from solvation and vibronic relaxation. 53,58 In low-viscosity polar solvents, the solvent-shell reorganization has been observed on a subpicosecond time scale, e.g., for Coumarin 153, which is complete within 1 ps in acetonitrile. 53 The time-dependent shift in the TRES clearly indicates that solvation contributes to the early excited-state processes. The time constant s 2 determined for the fluorescence in acetonitrile, whose amplitude first increases slightly followed by subsequent decrease with increasing wavelength, might therefore be assigned to solvent reorganization. The absence of such a shift in the non-polar solvent hexane ( Figure 5 and Table II) corroborates this assignment.

D. Vibrational relaxation
The occurrence of ESIPT along a highly unsymmetrical, exergonic potential energy surface should result in a highly vibrationally excited tautomer B. Since the tautomer B is higher in energy than the tautomer A by 33.68 kJ/mol (2815 cm À1 ) in the S 0 state 42 and knowing the S 0 -S 1 energy gap of tautomers A and B from the absorption (311 nm; 32 154 cm À1 ) and fluorescence maxima (535 nm; 18 691 cm À1 ), the zero-point energy of the tautomer A in the S 1 state should be %10 650 cm À1 above that of tautomer B in the S 1 state. Hence, ESIPT should provide %10 000 cm À1 of excess energy in S 1 B . The intermediate time component of s 2 % 10 ps (process "b" in Scheme 1) observed in TA can be ascribed to the vibrational relaxation process of the vibrationally hot tautomer B which is also manifested in the fluorescence measurement. Vibrational relaxation comprises intramolecular vibrational relaxation (IVR; vibrational redistribution) and intermolecular vibrational relaxation (vibrational cooling), with the former being faster than the latter. 58,59 Characteristic time scales for vibrational cooling are consistent with our measured %10 ps vibrational relaxation time. 12,17,58,60,61 Moreover, this smallamplitude rise component of %10 ps was present in both non-polar (hexane) and polar (acetonitrile) solvents, so that vibrational cooling rather than solvation gives rise to these dynamics. E. Lifetime of S 1 B TA data can contain various overlapping signals, both with negative sign as GSB and SE, and positive sign as ESA and absorption from product molecules. Our fluorescence upconversion measurements confirm that the longest lifetime found in the TA corresponds to an excitedstate and not to a ground-state product, because the lifetimes found in TA and fluorescence upconversion are basically identical. Hence, the S 1 B !S 0 B transition (process "c" in Scheme 1) proceeds with a time constant of %390 ps.
Since fluorescence from S 1 B is observable, a corresponding SE signal should be present in the TA data. The TA data in the VIS region ( Figure 2, right panel) resemble two positive peaks with identical temporal behavior, but this double-peak appearance actually is a consequence of a broad ESA signal (positive) and a weak SE (negative) right in the center, consistent with the fluorescence spectrum. Hence, the minimum in intensity near 520 nm in Figure 2 is due to overlap of weak SE (quantum yield of AID is very low %0.01) 42 with comparatively strong ESA, so that in sum only a positive signal is observed. Noteworthy, this behavior of a broad ESA modulated by a weak SE has been observed in a variety of ESIPT systems. 14,24 F. Lifetime of S 0 B When the emission has occurred, the product ground state S 0 B is reached, and hence the full cycle displayed in Scheme 1 is not yet complete. According to calculations, 42,46 there is also a ground-state barrier that has to be overcome to get from S 0 B to S 0 A . The TA data with deep-UV probing provide insight into the ground-state processes, because we can directly monitor the GSB of tautomer A [ Figure 3(a)], which is refilled on the same time scale observed for the emission decay. Excluding the rather unlikely case that tautomers A and B have identical absorption spectra, this shows that the rate-limiting step for the full cycle is the S 1 B !S 0 B transition rather than the S 0 B !S 0 A relaxation. Hence, within less than 1 ns, the excited molecules have returned to their initial structure, i.e., tautomer A.

V. CONCLUSION
This work quantifies the steps involved in the ultrafast ESIPT reaction of the molecule AID in the condensed phase. UV excitation of AID initiates ESIPT from the tautomer A to tautomer B which occurs with a time constant of %160 fs. The tautomer B thus prepared undergoes vibrational relaxation in the order of %10 ps along the exergonic potential energy surface in the excited state. Subsequently, the thermalized tautomer B relaxes to its corresponding ground state with a time constant of %390 ps. The ground-state tautomer B then quickly turns into tautomer A. The data are well described with such a simple four-state model as shown in Scheme 1, with the characteristic time scales for the involved processes derived from TA and fluorescence upconversion measurements. The broadband TA measurements, covering also the deep-UV region, directly reveal when the full ESIPT cycle is complete.