Arresting consecutive steps of a photochromic reaction : studies of b-thioxoketones combining laser photolysis with NMR detection †

Photochromism of monothiodibenzoylmethane has been studied in a number of environments at different temperatures. Direct laser irradiation of a sample located in the NMR magnet allowed in situ monitoring of the phototransformation products, determining their structure, and measuring the kinetics of the back reaction. These observations, along with the data obtained using electronic and vibrational spectroscopies for rare gas matrix-isolated samples, glasses, polymers, and solutions, as well as the results of quantum-chemical calculations, provide insight into the stepwise mechanism of the photochromism in b-thioxoketones. At low temperature in rigid matrices the electronic excitation leads to the formation of the –SH exorotamer of the (Z)-enethiol tautomer. In solutions, further steps are possible, producing a mixture of two other non-chelated enethiol forms. Photoconversion efficiency strongly depends on the excitation wavelength. Analysis of the mechanisms of the photochromic processes indicates a state-specific precursor: chelated thione–enol form in the excited S2(pp*) electronic state. The results show the potential of using laser photolysis coupled with NMR detection for the identification of phototransformation products.


Introduction
Huge interest in photochromism is stimulated by numerous possible applications of this phenomenon. [1][2][3][4] Photochromic transformations quite often involve proton transfer and/or conformational changes. A suitable, widely investigated system for studying the interplay between tautomeric and rotameric 25 interconversion processes is provided byβ-thioxoketones.  In principle, these compounds can assume many different forms, as depicted in Figure 1, which makes a reliable structural assignment of the forms involved in photochromism a very challenging task. After some controversies, 6,9,12,13 it has now 30 been established that β-thioxoketones exist in solution primarily as the thione-enol form (e-CCC) in equilibrium with the t-CCC form of Scheme 1, although the equilibrium constant does depend on the flanking groups (aliphatic or aromatic), temperature, and solvent. 18,23 Irradiating at 410 nm, the wavelength corresponding 35 to the S 2 (ππ*) transition of the e-CCC form of monothiodibenzoylmethane (1), carried out for rare gas, glass, polymer, and KBr matrices, leads to proton transfer and trapping of the non-hydrogen bonded t-TCC species, a form that can be kept almost indefinitely below 90 K. 24 Increasing the temperature 40 from 90 to 115 K results in over a thousand-fold decrease in the lifetime of the photochromic product. For instance, for 1 dissolved in 3-methylpentane, the lifetime drops from 1260 to 1 min. 27 It can be expected that, in a less rigid environment, other photochromic channels can be activated, e.g., those involving the 45 rotation about the C-C bond. No such structures have been reported so far. Therefore, the main goal of this work was to check for the presence of the currently unknown photochromic channels and to determine the structure of possible photochromic products. 50 In the above-mentioned cases the species were identified by means of IR and UV measurements. [24][25][26][27][28] Here, we present another, novel approach, based on laser irradiation in the magnet combined with in situ NMR detection. This methodology provides an interesting possibility for studies of photoinduced 55 reactions, although a number of technical difficulties have to be overcome. So far, the technique has been used to demonstrate the occurrence of syn-anti photoisomerization in the electron donoracceptor molecule, 2-(N-methyl-N-isopropylamino)-5cyanopyridine. 29 Other applications included studies of 60 photosensitized oxidation of imidazole and guanosine derivatives, [30][31][32] conformational changes in proteins, 33, 34 and RNA refolding. 35,36 In the present investigation, monothiodibenzoylmethane 1 (Scheme 1) is investigated, as this compound is stable and many derivatives exist. The results of 1 involves several steps. In contrast to the situation encountered in a rigid environment, where the photoconversion leads to the t-TCC form, irradiation of solutions produces a mixture of t-CTC and t-TTC. These forms revert back to the most stable e-CCC and t-CCC structures on the time scale of days at 5 183 K and minutes at 293 K.

Experimental and computational details
Synthesis and preparation of samples 15 1 was synthesized as described in Ref. 19 The concentration of the NMR sample was usually 2 millimolar. The NMR solvents used, i.e. CD 2 Cl 2 , CDCl 3 , CD 3 CN, CD 3 OD, THF-d 8 , toluene-d 8 , acetone-d 6 , were obtained from Cambridge Isotopes Laboratories, Inc. or from Armar Chemicals. Each sample was prepared from 20 fresh solvent from an ampoule without further purification. Solutions were bubbled with dry nitrogen for at least 10 min. inside NMR tubes directly before measurements.

NMR-laser setup
The setup with the laser is depicted in Figure 4. The laser beam 25 is transmitted via three aluminium mirrors and focused with a quartz lens on the end of the quartz rod (length 1 m and diameter 3.8 mm) inserted into the NMR magnet. The other end of the quartz rod is immersed in solution in NMR tube hanging on the rod (for details see Supporting Materials). It is placed 1.5 mm 30 above the upper end of the 1 H receiving coil as a compromise between the best irradiation of a measured volume and good magnetic field homogeneity. 35 NMR measurements were performed on Bruker AVANCE II 300 spectrometer operating at 300. unit. Temperature was set with use of the methanol thermometer as a probe before each experiment. The temperature during irradiation was increased by at most 2 o C. Irradiation time was controlled by the NMR spectrometer by means of a shutter. After a period of laser irradiation a 1 H 50 spectrum was recorded (16 scans, 2 min.) and the procedure was repeated until a steady-state was achieved. The reverse reaction was monitored by measuring 1 H spectra every 10-30 minutes. The product was examined using the following NMR techniques: 1 H-1 H NOESY -(noesygpph, mixing time 1 s), 1 H-1 H DQF 55 COSY (cosygpmfph), 1 H-13 C correlations, i.e. HSQC -(hsqcgpph, optimized for 170 Hz), HMBC -(hmbcgplpndqf, zero and double quantum coherence, optimized for 8 Hz with low-pass J-filter for 170 Hz). 60 The UV/VIS spectra were recorded on a Shimadzu UV2700 spectrophotometer, equipped with a home-made temperature control system. To monitor changes in the electronic spectra upon irradiation, the sample was cooled down to 205 K using a mixture of solid carbon dioxide and ethanol. DHOM-M-445-1000 and laser.

UV-VIS experiments
For the measurements of Raman spectra, a Renishaw InVia microscopic system was used, with two excitation sources: an HPNIR785 diode laser emitting at 785 nm and Stellar Pro Modu-Laser (LLC) emitting the Ar + 514.5 nm line. The laser light was 5 focused on a sample with a 30 mm focusing distance objective. The laser power at the sample was about 100 mW. The Ramanscattered light was collected by the same objective through a cutoff filter to eliminate Rayleigh scattering. Gratings of 1800 and 1200 grooves/mm were used for 514.5 and 785 nm laser 10 lines, respectively. The resolution was 5 cm −1 and the wavenumber accuracy 2 cm −1 , both calibrated with the Rayleigh line and the 520.6 cm −1 line of silicon. The Raman spectra were recorded by a 1024 × 256 pixel Peltier-cooled RenCam CCD detector. A Linkam THMS600 climatic chamber coupled to the 15 system translation stage was used for measurements at low temperatures, down to 183 K.

X-ray studies
Single crystal X-ray diffraction measurements were carried out on an Agilent Supernova diffractometer, at 100K with graphite 20 monochromated Cu Ka radiation (1.54184 Å). The data reduction was made by using CrysAlisPRO 37 software. The structures were solved by direct methods and refined on F 2 by full-matrix leastsquares by using SHELXS97 and SHELXL97. 38 All nonhydrogen atoms were refined as anisotropic while hydrogen 25 atoms were placed in calculated positions, and refined in riding mode.  30 CCDC 974809 contains the supplementary crystallographic data for this paper. These data can be obtained free of charge from The Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/data_request/cif. 35

Calculations
All the theoretical calculations were carried out using the Gaussian 09 code. 39 Molecular geometries were fully optimized using the B3LYP variant of the density functional theory (DFT) 40,41 with the 6-311++(d,p) basis set. Time-dependent DFT (TD- 40 DFT) method has been used for geometry optimization of the excited singlet states. For the triplet state geometry optimizations, both TD-DFT/B3LYP/6-311++(d,p) and DFT/UB3LYP/6-311++(d,p) methods were used, yielding similar structures and relative energies. The T 1 structures presented in Figures 6S-12S 45 correspond to the UB3LYP calculations. The NMR nuclear shieldings were calculated with B3LYP/6-311++(d,p) using the GIAO method. 42, 43 The electronic transition energies were calculated for both ground and excited singlet and triplet state optimized geometries using TD-DFT with the same functional 50 and basis set (B3LYP/6-311++(d,p)).

Irradiation experiments
Monitoring photochromism by electronic absorption spectroscopy 55 Figure 2 shows the results of photoirradiation of 1 in dichloromethane solution at 205 K using a 445 nm laser line. The evolution of the absorption spectra demonstrates that complete photoconversion can be readily achieved in less than 10 minutes. The quantum yield of the photoprocess was estimated as 60 1.5±0.6%. This value is practically the same as the quantum yield of the photochromic transformation of 1 in 3-methylpentane glass at 90 K (1.3±0.6%).
Contrary to the behavior observed for 445 nm excitation, when using the 532 laser line, the photochromic process is extremely 65 slow. Correcting for the amount of photons absorbed at 445 and 532 nm, the yield of photoconversion was found to be at least 100 times lower when irradiating into the lower energy band. The lower and higher energy regions correspond to S 1 (nπ*) and S 2 (ππ*) transitions of the enol e-CCC form, respectively. The 70 observed difference in photoconversion efficiencies raises an important question about the precursor state of the photochromic product: it can be either the enol tautomer in its S 2 (ππ*)electronic state or the t-CCC thiol form, for which, as discussed below, the electronic absorption is predicted to be blue-shifted with respect 75 to the enol. The electronic spectrum of the photoproduct after irradiation at 445 nm ( Figure 2), with the highest intensity peak at 318 nm resembles that obtained for the product of irradiation of 1 in high-85 viscosity media, such as rare gas matrices, polymers and glasses. However, the spectra are not identical, suggesting that the photochromic species produced under the two regimes are different. This is proven by the decay kinetics of the photoproduct. It disappears in less than a minute in a rigid IR/Raman measurements 445 nm irradiation of a sample of 1 in dichloromethane at 183 K resulted in the disappearance of the Raman signal of the substrate. However, no signal from a photoproduct could be 5 detected either. We interpret this as due to the fact that the electronic absorption of the most stable e-CCC form extends well to the red, thus fulfilling a pre-resonance condition for Raman excitation at 785 nm. This is not the case for the photoproduct, of which the absorption is significantly blue-shifted. As described 10 below, this hypothesis was confirmed by calculations.
On the other hand, the IR spectrum of the photoproduct could be registered even at room temperature. This spectrum was obtained while irradiating the solution of 1 in CH 2 Cl 2 contained in the sample chamber of the FTIR spectrometer with 445 nm 15 laser light. This led to the establishment of a photostationary regime, with about 20% depletion of the enol e-CCC form. The difference IR spectrum, presented in Figure 3 was compared with the patterns theoretically predicted for various forms. Nice agreement between experiment and simulation is obtained for 20 three thiol forms: t-TTC, t-CTC, and t-TCC. We note, however, that the IR spectrum of the latter, obtained in rare gas matrices is definitely different from the spectrum obtained for room temperature solution ( Figure 3). This suggests that the photoproduct generated at higher temperatures/lower viscosities 25 may correspond to either t-TTC or t-CTC, or the mixture of both. Based on the IR spectra alone, this problem cannot be solved unambiguously. K. The asterisks denote the bands corresponding to the photoproduct. For Ar matrix, it has been identified as t-TCC. 24 The hatched regions 35 correspond to solvent absorption. Bottom, the IR spectra simulated for e-CCC (black, inverted), t-TTC (red, left) and t-TCC (blue, right). The spectrum simulated for t-CTC is practically identical to that of t-TTC .

Laser photolysis-NMR studies
The measurements were performed generally at 183 K in several 40 different solvents: CD 2 Cl 2 , CDCl 3 , CD 3 OD, CD 3 CN, acetone-d 6 , and toluene-d 8 . The light source for most experiments was an argon laser, but some experiments were repeated with DPSS lasers emitting the light at 445 or 532 nm. For a description of the light path see Figure 4. The output was measured immediately 45 before entering the light guide (see Figure 1S for light guide and NMR tube arrangement). The dominant frequencies of the argon laser are the 514.5 and 488 nm lines together with a weaker line at 457.9 nm. In most of the present experiments the weak S 1 (nπ*) transition of e-CCC is therefore primarily irradiated together with 50 the S 2 (ππ  *) transition. Experiments showed early on that oxygen in solution had a dramatic effect ( Figure 2S). However, this type of photochemistry is not the subject of the present paper. Therefore, the solutions were freed of oxygen by thorough bubbling with dry nitrogen gas for at least 10 min. 55 Figure 4. Block diagram of the hyphenated NMR setup. For a full description see Experimental.
Using CD 3 OD, CD 3 CN or acetone-d 6 as solvent no product was observed. Using CD 2 Cl 2 as solvent one well defined species 60 (1id) was produced (see Figure 5). Typical irradiation time for maximal conversion of the starting material using the argon laser (0.3 W at the sample) was 2-3 hours in CD 2 Cl 2 . Using a laser at 445 nm (0.1W at the sample) the time for full conversion was reduced to less than 20 min. Using a laser at 532 nm no product 65 could be observed. Discontinuing irradiation led to conversion to the original compound, 1, as seen from the spectrum in Figure 5. A very similar situation was found in CDCl 3 . In this case the formation of the product (1ic) upon irradiation was observed with similar speed as for CD 2 Cl 2 . Also the reverse reaction in the 75 darkness took place.
Irradiation of 1 in a thoroughly degassed toluene-d 8 solution at 183 K for hours resulted in a product 1it as seen in Figure 6. Discontinuation of the irradiation led to 1: at 183K only slowly (lifetime of a day), but quicker at higher temperatures (lifetime of an hour at 273 K and of minutes at 298 K). 5 Back reaction rates 1/k [s] are as follows: CD 2 Cl 2 at 183 K: 18000; CDCl 3 at 213 K 40000 and toluene-d 8 at 183 K: 30000. For CD 2 Cl 2 at 183 K a much slower back reaction 1/k = 200000) is found if the solvent is purified by passing through a column of basic aluminum oxide. For CD 3 CN, methanol-d 4 and acetone-d 6 10 no products could be observed, so the back reaction rates could not be estimated.

NMR assignments
Based on the set of the following NMR spectra: 1 H, 1 H NOESY, 1 H-1 H correlation (COSY), 1 H-13 C correlations (HSQC, HMBC) 20 we were able to assign most of the 1 H and 13 C signals. A complete analysis was done for the spectra recorded in CD 2 Cl 2 , a partial one, for CDCl 3 and toluene-d 8  A NOESY spectrum of the 1it product (irradiation in toluened 8 ) is depicted in Figure 3S. At 183 K, it only shows a few cross peaks and these of positive phase due to the high viscosity of 50 toluene at low temperature. Therefore we have repeated this experiment at 213 K to get negative cross peaks. Furthermore, the spectrum is dominated by very strong signals from toluene. A cross peak is seen from the doublet at 7.67 ppm to a triplet at 6.92 ppm and one from a doublet at 7.26 ppm to a signal at 6.95 ppm. 55 In addition, a cross peak is seen from the SH signal at 2.8 ppm to the singlet at 6.5 ppm. The latter is the signal from the H-2 proton. Based on these and supported by the theoretical calculations the following assignment can be made: 7.67 ppm (H-2´, d), 7.26 ppm (H-2", d), 6.92 ppm (H-3´, t), 6.95 ppm (H-3"), 60 5.85 ppm (H-2, s) and 2.8 ppm (SH,s). 13 C NMR. Based on HSQC and HMBC spectra the 13 C spectrum of 1id at 183 K can be assigned as follows:

DFT calculations of the NMR parameters
As described above, many different enol and thioenol structures can be constructed together with several diketo forms. Ground state structures, energies, and 1 H and 13 C chemical shifts have 75 been calculated. The NMR parameters were not computed for the diketo forms, as the NMR spectra clearly show their absence. It is interesting to notice that all but two lowest thiol forms have very similar energies, especially when zero point energies are taken into account (Figure1). Moreover, except for the lowest energy e-CCC structure, all other enol forms have energies higher than the thiol species. 5 NMR data. The calculated 1 H nuclear shielding of the thiol forms for the H-2 and SH protons are discussed in terms of the structures given in Scheme 2. The structures in Scheme 2 can be described in the following fashion: A: H-2 is "trans" to CO; B: H-2 is "cis" to CO; C: SH is cis to Ph and S is close to O; D: SH is 10 cis to Ph and S is not close to O, and finally E: SH is trans to Ph and S is not close to O. For H-2 of A the following shieldings are found: 24

Structure of 1id.
It is obvious from the 1 H NMR data that 1id is not a diketo form, as no resonances are found in the 5-6 ppm range. In order to 40 distinguish whether the enol or thiol tautomer is formed, we have to analyse the 13 C chemical shifts of C=X. The finding of the 13 C resonance of C=X at 187.2 ppm is a clear proof that this chemical shift corresponds to C=O. The signals of C=S are typically shifted more downfield 44 and the 1d form, dominated by e-CCC, 45 exhibits a signal at 211.5 ppm, corresponding to C=S. The calculated chemical shifts are around 230-250 ppm for C=S (enol forms) and around 190-200 ppm for C=O (thiol forms). Naturally, hydrogen bonding to the solvent can also take place influencing the chemical shifts to some extent. 23 50 The finding that the SH form is dominant is also supported by the 1 H chemical shifts of the XH proton at 3.86 ppm, because such a value is characteristic rather for SH, whereas OH signals are usually shifted more downfield, as shown from the NMR calculations (Table 1S). The hydrogen-bonded thiol form t-CCC 55 is likewise excluded. The t-TCC form found in the experiment irradiating at 415 nm at 90 K is likewise not found, in good agreement with the low barrier to return found for this intermediate. The structural assignment of 1id is based on a combination of chemical shifts and NOE data. The finding of a 60 cross peak between the SH proton and H-2 clearly shows that those two protons are on the same side of the double bond. The chemical shifts of the H-2 proton (7.13 ppm) assign this proton and the oxygen to be on opposite sides of the double bond (structure A in Scheme 2). These two criteria only leave two 65 possible structures for 1id, t-CTC and t-TTC (see Figure 1). The finding of a NOE cross peak between the SH proton and the protons of the double prime ring (Scheme 1) points towards the t-TTC form. On the other hand, the SH chemical shift (3.86 ppm) points towards E of Scheme 2, corresponding to t-CTC and, most 70 probably, both species exist in equilibrium. The relevant protonproton distances for all calculated enol and thiol tautomers are presented in Table 2S.
The product obtained in CDCl 3, 1ic, is identical to 1id, as judged from the 1 H NMR spectrum and from the value of the 75 reverse reaction rate.

Structure of 1it
As seen in the results section, the 1 H chemical shifts of 1it and 1id are rather different, which could suggest that both structures are not the same. On the other hand, the solvent effect on the 80 chemical shifts in toluene is specific. Ring currents of toluene lead to the low-frequency shifts of 1 H signals, which makes the comparison to the spectra in the other solvents complicated. However, the 2D-NOESY experiment exhibits cross-peaks between the same protons as in the case of 1id (see Figures 7 and 85 3S). The identical structures of 1id and 1it are also supported by the experiment in mixed solvents (CD 2 Cl 2 : toluene-d 8 5:1 and 1:1), as seen in Figure 8. The changes of the SH and CH chemical shifts with growing concentration of toluene have a smooth character and are similar to the behaviour of solvent peaks. 90 Moreover, the reverse reaction in the dark is as slow as for 1id. Therefore, we can conclude that 1it has the same structure as 1id.

Conversion back to 1
Is the conversion of 1id and 1it back to 1 feasible? In the context it should be considered that both 1id and 1it are so-called twisted 95 push-pull ethylenes. The calculated twist (the S,C=C,C dihedral angle) is 10 o and 8 o for t-TTC and t-CTC, respectively. Sandström and co-worker have found that barriers to rotation around the double bonds in such systems are not very high.

Modeling of the photoreaction path.
In order to understand the mechanisms of the multistep photochromic transformation in 1, we considered all five experimentally observed species as well as the relevant transition states. Ground state structure optimizations were followed by 10 calculations of transition energies originating at S 0 geometries. In the next step, geometries of the lowest excited singlet and triplet states were optimized. The calculated ground state geometry of the e-CCC form is in very good agreement with the X-ray structure ( Figure 6S). The optimized S 0 , S 1 , T 1 , and transition 15 states geometries of all five forms are presented in Figure 7S. We have also repeated the calculations for thiomalonaldehyde ( Figures 8S-12S), a model system, for which all five forms are predicted to be planar in the ground state. The patterns of the relative energies obtained for both molecules turned out to be 20 very similar, which indicates that the origin of the behavior of 1 is due the electronic structure of the thiomalonaldehyde core, not to the symmetry-lowering perturbation of the substitution by phenyl rings. In this context, we also note that the X-ray study of 1-(1-methylcyclopropyl)-3-thioxobutan-l-one, a β-thioxoketone 25 with aliphatic substituents 46 revealed a very similar core geometry to that of 1 (the differences in bond lengths of about 0.01 Å, the angles between bonds agreeing within 1 degree). The relative energies calculated for different tautomers/rotamers of 1 are shown in Scheme 3. The first 30 important finding concerns the relative energies of the two most stable forms, e-CCC and t-CCC. In agreement with experiment, their energies are predicted to be similar, with the chelated thiol form higher by 4 kcal/mol (2 kcal/mol after ZPE correction). On the contrary, for the S 1 and T 1 excited states the calculations 35 predict a large destabilization, of the order of 10 kcal/mol or more, of the thiol with respect to the enol tautomer. This prediction is crucial for the understanding of photochromism, as the first photoproduct, "open" thiol t-TCC is obviously generated from t-CCC. Since the excited state e-CCC→t-CCC reaction is 40 predicted to be strongly uphill in S 1 and T 1 , the effective way to produce t-TCC is either by direct excitation of t-CCC or via the S 2 (ππ*) state of e-CCC. This conclusion is in excellent agreement with the observed huge dependence on the photoconversion efficiency on the excitation wavelength. It is well established for 45 many thiocarbonyl compounds that the ability of sulfur to form hydrogen bonds decreases upon excitation to S 1 (nπ*), but increases in S 2 (ππ*). [47][48][49] Moreover, the lifetime of S 2 (ππ*) can be as long as hundreds of picoseconds, 50, 51 but it can be effectively quenched by specific interactions with such solvents as 50 acetonitrile 52, 53 or water. 54 These findings not only provide a strong indication that the main precursor of the photochromism is the S 2 (ππ*) state of e-CCC: they also explain why there is no photoreaction in methanol or acetonitrile. Naturally, the reaction can also be initiated by direct excitation of t-CCC, but the ground 55 state population is dominated by the enol, which also has a stronger absorption than the thiol at the wavelengths used. Therefore the contribution by direct excitation of this species should be minor, especially at lower temperatures. We finally recall that the open thiol can be generated photochemically at 60 very low temperatures, when only the enol form is initially present in the ground state. 24 The calculations predict that the e-CCC to t-CCC conversion is strongly endoenergetic in both S 1 and T 1 (Scheme 3A). Attempts to optimize S 2 were not successful. For model molecules with 65 hydrogen atoms or methyl groups instead of phenyls, S 2 optimization assuming C s symmetry of t-CCC ends up in e-CCC. These results indicate the e-CCC(S 2 ) → t-CCC(S 1 ) as the most probable path.
Once the excited t-CCC has been produced, it can be 70 transformed into t-TCC. The S 1 optimized t-CCC and t-TCC structures converge to identical geometries, with the SH bond almost perpendicular to the OC1C2C3S plane. This structure (tsSH1 in Scheme 3A and 3B) corresponds in S 0 to the transition state between t-CCC and t-TCC. Thus, the S 1 -S 0 relaxation 75 (favourable because of a small energy gap) can lead to both forms.
In order to understand the next step of photochromism, the transformation into t-TTC and/or t-CTC, we note that the optimized S 1 /T 1 (and probably also S 2 ) states have in these 80 structures higher energies than tsSH1. However, the structures in which the molecule is twisted about the C2C3 bond (tsR1 and tsR2 in Scheme 3) have in S 1 a lower energy than tsSH1. These structures correspond in the ground state to high energy (40-50 kcal/mol) transition states for rotation about a double bond. Full 85 geometry optimization of the twisted forms was not possible, because of close-lying S 0 and S 1 states, characteristic for this type of structure and possibly leading to a conical intersection (CI). The S 1 energies presented in Scheme 3 correspond therefore to upper limits. One can conclude that once the molecule overcomes 90 the barrier from tsSH1 to tsR1 or tsR2, it can immediately convert (via conical intersection) into t-CTC and t-CCC (from tsR1) or into t-TTC and t-CTC (from tsR2). The t-TTC and t-CTC are practically isoenergetic in S 0 (0.5 kcal/mol difference) and separated by a small barrier for the SH group rotation (6 95 kcal/mol). Most probably the equilibrium is established between the two forms.
The behaviour predicted for the triplet states using UB3LYP is somewhat similar to that described above for S 1 . Except e-CCC, no structures with planar skeleton are obtained: they all relax to 100 the twisted tsR1 and tsR2 forms (Scheme 3B). Contrary to the situation in S 1 , also the tsSH1 form is nonplanar; it now corresponds to the transition state between tsR1 and tsR2. Thus, the fate of excited molecules which underwent rapid proton transfer is similar to those in S 1 . Now, however, the twisting leading to one of the tsR forms is barrierless. Even though no CI is predicted for these forms, the T 1 -S 0 energy gap is very small. 5 On the basis of calculations we conclude that photochromism involving large amplitude motion is possible both in singlet and triplet manifolds. Because of expected large efficiency of intersystem crossing to the triplet and the barrierless character of tsSH1 → tsR forms, the triplet path seems more probable. 10 The crucial parameter for the t-TCC → t-TTC, t-CTC conversion is the C2C3 bond length (Scheme 1), as, formally, the process involves a rotation about a double bond. The values calculated for ground and excited states are very instructive in this regard. For S 0 , the calculations predict 136, 135, and 135 pm 15 for t-TCC, t-TTC, and t-CTC, respectively. For the relevant S 1 structure, tsSH1, the value of 139 pm is obtained, whereas the calculations for the triplet tsR1 and tsR2 forms yield 146 pm. These results clearly demonstrate the weakening of the double bond character in the triplet state. Together with the relative 20 energies calculated for the three forms, they provide strong arguments that the large amplitude transformation leading to the photochromic products observed in solution occurs via the triplet state.
As was already mentioned above, the rotation about the double 25 bond in the ground state involves a large energy barrier. In principle, an alternative mechanism is possible, with the participation of the diketo form. First, the SH hydrogen atom is transferred from t-CTC to C2 carbon with partial twisting, and then back to either sulphur or oxygen, with further twisting. 30 However, the calculated barriers to these processes are only slightly lower (still above 40 kcal/mol). The experimental finding that this reaction takes minutes at room temperature indicates that the real mechanism is different. The large change of the reaction rate in CD 2 Cl 2 after removal of acid traces from the solvent 35 supports the reaction pathway involving protonation of the C=O group of t-CTC and/or t-TTC. Initial calculations support this hypothesis, yielding for the protonated form a barrier of 20 kcal/mol . The barrier to rotation about the C1C2 single bond is very 40 small (ca. 2 kcal/mol). The t-TTT form is calculated to lie higher in energy than t-TTC. The experiment suggests that if it is formed, its concentration is below that required for NMR detection.

45
Scheme 3. Calculated relative ground (black) and T1 (blue), S1 (green) and S2 (red) excited state energies of the experimentally observed forms of 1 and of the predicted transition states. Solid lines indicate optimized geometries. The vertical energies computed for the optimized geometries are indicated by dashed (S1 and S2,at S0, S0 at S1) or dotted (T1 at S0 and S0 at T1) lines. The vertical T1 energies were computed by both TD-DFT (short dash lines) and UB3LYP (dots). Transition states are indicated by asterisks. CI indicates non fully optimized structure, due to proximity of states.

Conclusions
The use of "hyphenated techniques" such as laser irradiation combined with NMR detection at low temperature enables the observation and identification of new unstable species. By 55 varying the temperature and viscosity, it was possible to observe the consecutive steps of photochromic transformation in monothiodibenzoylmethane. Irradiation of 1 in a rigid environment at temperatures of about 100 K or lower leads to the formation of t-TCC, a structure characterized by opening of the 60 intramolecular hydrogen bond. Such a rotamer is stable below 100 K, but returns to the main e-CCC form upon heating up. The situation in a non-rigid environment, i.e. liquid solution, is different. Now, the end products of the irradiation are t-CTC and t-TTC. The reaction path we propose includes photoinduced 65 transformation from e-CCC in the S 2 (ππ*) state to t-CCC, followed by the hydrogen bond opening which leads to t-TCC. This step is possible even at very low temperatures and in rigid media, contrary to the second stage of the photochromic transformation, rotation about the C2C3 bond of the thiol t-TCC form. This reaction can only proceed in less viscous environments. The opening of hydrogen bond may occur in both the excited singlet and the triplet state of t-CCC, but the next stage most probably goes only through T 1 . The dark reaction, i.e., 5 the recovery of the most stable form of 1 is much longer than the photoinduced reaction. It is because the rotation along double C=C bound in the ground state has a relatively high barrier. The proposed scheme of the photochromism of 1 is shown in Scheme 4.
10 Scheme 4. The proposed photochromic pathways. Red and blue arrows denote singlet and triplet routes, respectively. Dashed lines indicate hot ground state molecules with geometry corresponding to S1 or T1 excited state.

15
An observation relevant for the understanding of the mechanism of photochromism is the extremely small yield for excitation into the low energy region where only e-CCC absorbs, into its S 1 (nπ*) state. Combined with the theoretical prediction of a large shift of the enol-thiol equilibrium towards the enol form, 20 this finding strongly suggests lack of proton transfer from e-CCC to t-CCC in the S 1 and T 1 excited states. In consequence, the precursor of the photochromism is state-specific: e-CCC enol in S 2 (ππ*). Another argument for this postulate is the lack of photochromism in methanol and acetonitrile, as for these solvents 25 the S 2 lifetime can be drastically reduced. The directly excited t-CCC can only contribute at high temperatures, and even then the absorption will be dominated by e-CCC.
The structural assignments of t-CTC and t-TTC provided by NMR for CD 2 Cl 2 and CDCl 3 solutions seem very reliable. The 30 product in toluene-d 8 is most likely the same. Quantitative description of the kinetics of photochromism in β-thioxoketones requires time-resolved experiments spanning many orders of magnitude, from femtoseconds to minutes or even days. Such studies are now underway. We note the application potential of 35 compounds for which the photochromic properties depend on solvent, temperature, viscosity, and excitation wavelength. Finally, one should stress that using the NMR detection for direct, in situ monitoring of unstable products of photochemical transformations opens up many new avenues in all areas 40 requiring unequivocal structural assignments.
We hope that the complicated scheme of processes responsible for the photochromism in β-thioxoketones can provide a challenge for the rapidly growing area of theoretical description of photoinduced proton/hydrogen transfer phenomena. 55-65 45