Exploring photochemistry of p-bromophenylsulfonyl , p-tolylsulfonyl and methylsulfonyl azides by ultrafast UV-pump – IR-probe spectroscopy and computations †

The photochemistry of p-bromophenylsulfonyl azide (BsN3), p-tolylsulfonyl azide (TsN3) and methylsulfonyl azide (MsN3) was studied by femtosecond time-resolved infrared spectroscopy with CH2Cl2 and CCl4 as solvents along with quantum chemical calculations. The photolysis of these azides after 267 nm light excitation leads to the population of each respective azide S1 excited state. Decay of the S1 excited state gives rise to singlet nitrene formation. In the case of BsN3, the decay was found to correlate with the formation of a pseudo-Curtius photoproduct (PCP) BrC6H4NSO2. Transient electronic ground states of the three azides on their way to singlet nitrenes and PCPs were shown by locating the corresponding transition states on the potential energy surfaces. The lifetime of singlet (BsN) and (TsN) nitrenes is tS = B20 ps in CH2Cl2 and B700 ps in CCl4. Singlet (MsN) was not detected. Due to fast intersystem crossing (ISC), singlet nitrenes are converted into the triplet spin isomers lying lower in energy, the formation time constants being equal to the corresponding singlet nitrene lifetime. The formation of (MsN) was shown and the formation time constant in CH2Cl2 was found to be tISC = 34 3 ps. Internal conversion of the S1 excited state to the ground state of the azide was low (F E 0.15) for BsN3 and TsN3 and was not found in the case of MsN3.


Introduction
Sulfonyl azides, RSO 2 N 3 , are important reagents in synthetic organic chemistry. [1][2][3][4] Similar to the reactions of the structurally related carbonyl azides, 5 both thermal and photochemical decomposition reactions of sulfonyl azides have been extensively studied in solution [6][7][8][9][10][11][12][13][14][15] and demonstrated to exhibit rich and complex chemistry. 5,6,[16][17][18] The photochemistry of 2-naphthylsulfonyl azide (2-C 10 H 7 SO 2 N 3 ) was recently studied by femtosecond time-resolved infrared (fs-TRIR) spectroscopy and the azide S 1 excited state has been observed. 6 This S 1 state decays to produce the singlet nitrene 1 (2-C 10 H 7 SO 2 N) as a short-lived species (t S E 0.70 AE 0.30 ns in CCl 4 ) that decays to the lower-energy and longer-lived triplet nitrene 3 (2-C 10 H 7 SO 2 N). The triplet spin state is the ground state for sulfonyl nitrenes which has been proved both theoretically 6,19,21 and experimentally by ESR and IR spectroscopy in matrices at low temperature. 16,17,20 However, neither singlet nor triplet sulfonylnitrenes are global minima, because the most stable species are N-sulfonylimines RNSO 2 , which are formed with a large energy gain. 22 So far, evidence for the formation of the pseudo-Curtius rearrangement product (PCP) after azide photolysis is still inconclusive. 6 The possibility of concerted pseudo-Curtius Rearrangement In the Excited State (RIES) of sulfonyl azides as well as other sulfonylnitrene precursors has also been reported. 23 N-Mesyl and N-tosyldibenzothiophene sulfimides were studied by ns-TRIR spectroscopy as the predecessors of MsN and TsN, respectively. 23 The time resolution (50 ns) of this experiment did not allow the authors to detect nitrenes, but sulfonoazepine was detected as a result of singlet nitrene attack on dibenzothiophene. Analysis of stable products formed upon photolysis of N-mesyldibenzothiophene sulfilimines suggests that triplet nitrene was produced upon irradiation. No evidence of a Institute of Biophysics, Johann Wolfgang Goethe-University, Max-von-Laue-Str. 1, pseudo-Curtius rearrangement of the sulfilimines precursor was found. This result suggests that neither singlet nor triplet nitrenes are the precursors of PCPs in solution at room temperature, as was concluded in ref. 6. However, the very recently studied photochemistry of trifluoromethylsulfonyl azide CF 3 SO 2 N 3 at low temperatures in matrices showed that the triplet nitrene 3 (CF 3 SO 2 N) generated using the technique of post-pulse irradiation at 193 nm is converted to CF 3 NSO 2 and CF 3 S(O)NO by a Curtiustype rearrangement. 16 Moreover, another new species CF 2 NQSO 2 F and FSNO were identified along with CF 2 NF, SO 2 , F 2 CO, CF 3 NO, and SO as side products. This experiment indicates that the possibility of the formation of the PCP as well as other products of rearrangement of nitrenes may depend on the substituent R (in RSO 2 N).
Such a rich and diverse photochemistry of sulfonyl azides prompted us to investigate three sulfonyl azides that are structurally related to the aforementioned compounds, i.e. p-bromophenylsulfonyl azide (BsN 3 ), p-tolylsulfonyl azide (TsN 3 ) and the simplest member of the family, methylsulfonyl azide (MsN 3 ), by femtosecond time-resolved UV-pump-IR-probe spectroscopy in conjunction with computational studies.

Experimental and computational details
Synthesis Sulfonylazides were synthesized by the reaction of the corresponding sulfonyl chlorides with sodium azide by known procedures. [24][25][26] Sodium azide was purified as in ref. 27. The structure of the azides was confirmed by 1 H, 13 C NMR and IR spectroscopy (see ESI †).

Computational details
All calculations were performed with full geometry optimization using the Becke three-parameter hybrid exchange functional and the Lee-Yang-Parr correlation functional (B3LYP) 28,29 by employing 6-311++G(3df,3pd) and M06-2X 30 density functional theory with 6-311++G(d,p) basis sets. However, first singlet S 1 and triplet T 1 excited state energies and frequencies of azides were calculated using the single point methodology on the S 0 ground state geometry, all trials to optimize the excited states led to fragmentation of the species. Since the cost of S 1 state calculations by B3LYP/6-311++G(3df,3pd) was extremely high even for the smallest molecule MsN 3 , we first calculated it using the 6-311++G(d,p) basis set. However, it led to principal errors in predicting the IR band shift: S 1 showed a blue shift with respect to the S 0 , whereas a red-shift is observed experimentally. 6 Changing the B3LYP by M06-2X functional gave the results which are in compliance with the experiment. Because the correspondence of calculation and experiment differs for each species, the results of both the B3LYP/6-311++G(3df,3pd) and M06-2X/6-311++G(d,p) computations are given where possible.
For each stationary point, second derivatives of the energy were calculated to confirm whether these structures were local minima or transition states. All transition state calculations were accompanied by intrinsic reaction coordinate (IRC) calculations 31 to verify that each transition state connected the corresponding reactants and products. The calculated vibrational frequencies were not scaled. The calculated and experimental frequencies along with their assignment are summarized in Table S1 (ESI †).
Vertical excitation energies of azides were computed at the time-dependent TD-B3LYP and TD-M06-2X 32,33 levels of theory using the corresponding S 0 ground state geometries. To characterize the vertically excited states, electron density difference plots were computed (between S 0 and the S 1 -S 3 states) as described previously. 6,34 All calculations were performed using the Gaussian09 suite of programs. 35

Ultrafast experiments
Ultrafast time-resolved UV-pump-IR-probe experiments were performed on an amplified Ti:sapphire laser system (100 fs pulse length, 800 nm laser wavelength and 1 kHz repetition rate). The amplified output is split into two beams and used to pump two optic parametric amplifiers (OPAs). One OPA was used to produce broad-band IR probe and reference pulses, and another OPA to generate UV pump pulses at 267 nm via third harmonic generation of the fundamental. The pump energy was set at 2 mJ, and the exposure dose was 40 J mol À1 or 2 J ml À1 solution (but 10 mJ for MsN 3 , the exposure dose was 100 J mol À1 or 10 J ml À1 solution). The instrument response function was typically about 350 fs in the semiconductor GaAs (FWHM). To avoid contributions of rotational diffusion to the pump-probe signal, the polarization angle between pump and probe beams was set at the magic angle (54.71). The pump and probe beam diameters were about 200 mm and 160 mm, respectively. Kinetic traces were analyzed utilizing single-trace fitting using a sumof-exponentials instead of a global routine in view of the fact that the spectra of some species contained continuously shifting spectral bands which are related to vibrational cooling processes and difficult to model. The general equation for the used exponential terms is: where DA is the experimental difference absorption signal (pump on -minus-pump off , in mOD); A n -pre-exponential factor (in mOD); t -delay time between pump and probe beams (in ps); t n -formation/decay time constant (lifetime, in ps); offset -an 'infinite' spectrum at 'infinite' time (in mOD); n -integer number denoting the number of exponentials used. The featureless signal at À20 ps was subtracted as the background signal. In the ESI † the resulting paramaters required to fit the time evolution of discussed single wavenumbers can be found. All experiments were performed at room temperature. The final sample concentration was 50 mM in the case of BsN 3 and TsN 3 , and 100 mM for MsN 3 . CCl 4 and CH 2 Cl 2 (Sigma Aldrich) were used as nonpolar and polar solvents in order to estimate the solvent polarity effect on the reaction mechanism and on the time constants. The solution (about 6 ml) was constantly circulated between two 2 mm CaF 2 windows separated by a 50 mm spacer in a closed loop flow cell, 36 using a peristaltic pump at a speed of 4.4 ml min À1 . A comparison of the intensity of the N 3 stretch vibrations before and after the experiment via steady-state FTIR measurements reveals that about 8% of BsN 3 and TsN 3 , and B20% of MsN 3 is photoconverted after 10 h and 14 h of laser experiments, respectively.

Results and discussion
Azide N 3 stretch region, 2000-2200 cm À1 Irradiation at 267 nm of BsN 3 in CCl 4 produces a negative absorption signal at 2128 cm À1 which, according to calculations, corresponds to the (bleached) N 3 stretch vibration of the azide S 0 ground state ( Fig. 1 and 5 in the Computational results section for the complete FTIR spectrum). As was observed for 2-C 10 H 7 SO 2 N 3 , 6 the bleach recovers after about 10 ps due to the repopulation of the azide ground state from its S 1 excited state. The internal conversion process (IC, i.e. the S 1 -S 0 transition) was found to have a time constant t IC (CCl 4 ) = 45 AE 2 ps for BsN 3 in CCl 4 . Based on the fitting results we conclude that almost 15% of the excited molecules return to the azide ground state (corresponding to a quantum yield F IC = 0.15 at 3 ns). The reaction rate R IC = F IC t IC À1 is estimated to be (3.3 AE 0.3) Â 10 9 s À1 in CCl 4 . Percentages of recovery were the same for both BsN 3 and TsN 3 ground state nN 3 modes in both CCl 4 and CH 2 Cl 2 . However, the fitted recovery time constant at 2133 cm À1 for BsN 3 in CH 2 Cl 2 was different from that obtained in CCl 4 and was found to be t IC (CH 2 Cl 2 ) = 31 AE 1 ps (see Fig. S1, ESI †). In the case of TsN 3 t IC (CCl 4 ) = 33 AE 2 ps at 2127 cm À1 in CCl 4 and 28 AE 3 ps at 2129 cm À1 in CH 2 Cl 2 were obtained (Fig. S1, ESI †). Temporal evolution of the signals did not show any dynamics after 200 ps up to 3 ns. Similarly, no recovery of the ground state was observed after photolysis of MsN 3 (Fig. S3, ESI †). As was shown previously, 330 nm excitation of 2-C 10 H 7 SO 2 N 3 in the same solvents resulted in higher values of the ground state F IC (up to 0.6 in CH 2 Cl 2 ) and two times higher R IC . 6 In the case of BsN 3 and TsN 3 the internal conversion is low after irradiation at 267 nm. Its yield is not affected by the nature of the solvent and the R IC is in the range of a few 10 9 s À1 . In terms of the potential energy surface (PES) that means that S 1 and S 0 states of these azides have a rather large energy gap and obviously sufficiently separated PESs, so there is no conical intersection between them.
TD-B3LYP calculations predict that 267 nm light directly populates the S 1 excited state of BsN 3 , TsN 3 and MsN 3 (vide infra). Unfortunately, we failed to experimentally detect any noticeable signal within 30 ps which could be assigned to the azide excited state S 1 in this spectral window for all studied azides. Interestingly, according to TD-M06-2X/6-311++G(d,p) results, the BsN 3 S 1 excited state is 56 cm À1 red-shifted with respect to 2337 cm À1 in S 0 , and has eight times lower intensity than that for the S 0 state of azide, 83 vs. 691 km mol À1 (Fig. S7 and Table S8, ESI †). That means that maximum DA of the S 1 signal is predicted to be B1.1 mOD since the S 0 state has a negative intensity signal of BÀ9 mOD. However, the S 1 state signal was not spectroscopically observed in the azide spectral window. The obtained experimental results allowed us to conclude that the S 1 states of BsN 3 , TsN 3 and MsN 3 are probably weak IR-active species, showing signals which are even weaker than predicted, because otherwise they would have been detected. A similar behavior was recently observed in studying photochemistry of 5-azido-2-(N,N-diethylamino)pyridine. 37 The authors named the nonobservable S 1 state a 'spectroscopically dark' excited state. It can be assumed that such a spectroscopic behavior of the S 1 state is due to identical or close symmetry of the ground and excited states rather than the np* nature of S 0 -S 1 transition of the compounds. However, a low-intense signal appearing within 30 ps in the 1500-1600 cm À1 spectral window and its decay within 50 ps were observed for BsN 3 in CH 2 Cl 2 (vide infra). We have found no evidence of triplet azide formation while the theory predicts strong signals (e = B1000 km mol À1 ) for such species.
Although the azide spectral region did not contain signals from azide's excited state S 1 , it was clearly possible to observe Fig. 1 Transient IR spectra produced upon BsN 3 photolysis in CCl 4 (l ex = 267 nm) at selected time delays (left). The (scaled) light-minus-dark FTIR difference spectrum is shown for comparison (spheres). Transient kinetics of BsN 3 ground state recovery at 2128 cm À1 (right), also depicting the parameters resulting from the fit (continuous line) to the data curve (spheres). intramolecular vibrational redistribution or 'vibrational cooling' (VC) of the hot azide ground state formed from the S 1 state for BsN 3 and TsN 3 in both CCl 4 and CH 2 Cl 2 (Fig. S2, ESI † for BsN 3 ). VC typically exhibits itself by showing a time-dependent upshift in wavenumber of a vibrational mode. For instance, a vibrationally hot ground state of BsN 3 dissolved in CCl 4 was observed in the spectral range of 2110-2121 cm À1 (2102-2119 cm À1 in CH 2 Cl 2 , see Fig. S2, ESI †). The onset and decay time constants are 13 AE 5 ps at 2112 cm À1 and 20 AE 19 ps at 2119 cm À1 in CCl 4 for BsN 3 (and 15 AE 1 ps at 2102 cm À1 and 16 AE 3 ps at 2119 cm À1 in CH 2 Cl 2 ). Similar VC values were obtained for TsN 3 while for MsN 3 no VC was found since no ground state recovery was observed.
In-ring C-C stretch region, 1500-1600 cm À1 As was mentioned before, in the 1500-1600 cm À1 spectral region it was possible to observe the S 1 excited state signal for the BsN 3 sample dissolved in CH 2 Cl 2 (see Fig. 2). It was not possible to collect data in CCl 4 because the signal reproducibility was low. A negative absorption at 1575 cm À1 was observed, which, based on the B3LYP results (calculated at 1607 cm À1 , e = 65 km mol À1 ), corresponds to the C-C ring stretch mode of the BsN 3 ground state. This band has a recovery time constant of t IC (CH 2 Cl 2 ) = 28 AE 1 ps, which is similar to the time constant observed for the ground state N 3 stretch mode (2133 cm À1 , t IC (CH 2 Cl 2 ) = 31 AE 1 ps, Fig. S1, ESI †). Calculations at the M06-2X level of theory predict the BsN 3 S 1 state vibration at 1625 cm À1 (e = 306 km mol À1 , Fig. S7 and Table S8, ESI †) which is 4.5 times more intense than that for the ground state and is 24 cm À1 red-shifted. Indeed, a broad positive signal was detected within 30 ps after the laser pulse with maximum at 1559 cm À1 corresponding to the azide S 1 state. The lifetime of the BsN 3 S 1 state in CH 2 Cl 2 was found to be t S 1 (CH 2 Cl 2 ) = 22 AE 2 ps. The decay constant nicely correlates with the t IC (CH 2 Cl 2 ) = 28 AE 1 ps recovery time constant of the S 0 . It also deserves to be mentioned that the BsN 3 S 1 lifetime is five times longer than that of 2-NpSO 2 N 3 in the same solvent. 6 This fact can be considered as a confirmation of the aforementioned assumption about weak intersection between the S 1 and S 0 PESs. A persistent signal with its maximum near 1564 cm À1 is observed from 50 ps up to 3 ns and is assigned to triplet nitrene 3 (BsN) based on vibrational calculations at the B3LYP level (1602 cm À1 , 177 km mol À1 , see Table S3, ESI †) and its absence in the FTIR spectrum (Fig. 2). SO 2 asymmetric stretch region, 1300-1400 cm À1 Calculations suggest the presence of another interesting region where sulfonylnitrenes could be detected, which is the 1300-1400 cm À1 spectral window (experiment in Fig. 3; see also calculated frequencies in Fig. S4-S9, ESI †). In this section we present a detailed discussion of the results obtained upon irradiation of BsN 3 in both CCl 4 and CH 2 Cl 2 followed by a concise description of the results of TsN 3 and MsN 3 .
BsN 3 . Similar to the bands at 2128 and 1576 cm À1 , which are bleached in CH 2 Cl 2 , the n as (SO 2 ) band near 1377 cm À1 vanishes upon photolysis due to depletion of the azide ground state ( Fig. 3b and Fig. S10, ESI † for the spectral slices). The second higher wavenumber bleach corresponding to the S 0 state is the second nN 3 stretch band (1393 cm À1 ). When CCl 4 was used as the solvent, the position of the two bands was reversed ( Fig. 3b-d).
As in the case of the N 3 stretch spectral window, the band at 1393 cm À1 (in CH 2 Cl 2 ) does not have a corresponding positive excited state signal at lower wavenumbers that appears within 30 ps and has a lifetime of around 20 ps. Interestingly, the M06-2X calculations predict the signal of the azide S 1 state of nearly but slightly less intensity and wavenumber to overlap with that of the S 0 state (around 1380 cm À1 , see Fig. S7, ESI †), explaining the absence of a measured S 1 signal.
The 1377 cm À1 feature in Fig. 3b shows little recovery (about 7%, but B20% in the case of CCl 4 as the solvent) and its time constant is 38 AE 9 ps (CH 2 Cl 2 ). This is about 50% slower than t IC = 28 AE 1 ps (CH 2 Cl 2 ) obtained from the trace at 1576 cm À1 . We believe that the observed behavior of this band is due to three different processes. One of them is the S 1 azide excited state formation absorbing at 1377 cm À1 , and the second one is its decay. 6 Assuming that the S 1 decay and S 0 recovery time Fig. 2 Transient IR spectra in the range of 1500-1600 cm À1 upon BsN 3 photolysis in CH 2 Cl 2 at selected time delays (left). Transient kinetics at 1576 cm À1 (BsN 3 ground state) and 1559 cm À1 (BsN 3 S 1 state) in CH 2 Cl 2 (right), also depicting the parameters resulting from the fits (solid line) to the data curves (dots).
constants and extinction coefficients are equal they cancel each other, as predicted by the M06-2X calculations (see above). The third process, according to M06-2X calculations, is the formation of p-bromo-N-sulfonylaniline (BrC 6 H 4 NSO 2 ) as the pseudo-Curtius photoproduct (B-PCP). The evolution of the n as (SO 2 ) band at 1391 cm À1 in CCl 4 shows B20% recovery of S 0 , close to that for the azide ground state in the N 3 stretch region. This trace is well separated from the band assigned to the PCP (1365 cm À1 , see Fig. 3b and d). One point that remains unclear is why in CH 2 Cl 2 the ground state n as (SO 2 ) band has two times less recovery than that in the N 3 stretch region.
Since the S 1 excited state of sulfonyl azides is a dissociative species capable of elimination of molecular nitrogen, singlet nitrene 1 (BsN) is observed in both CH 2 Cl 2 and CCl 4 . The singlet appears as a vibrationally hot species (see Fig. 3a-c, marked as 'hot singlet nitrene', and Fig. S10, ESI †), evident by a band shift from 1325 to 1350 cm À1 in CH 2 Cl 2 and from 1340 to 1357 cm À1 in CCl 4 . Fig. 3a-c show the results in CCl 4 because the kinetics of the vibrationally hot singlet nitrene cooling process is slower than that observed in CH 2 Cl 2 , and because the cooling process in the latter suffers from significant spectral overlap between hot singlet and triplet nitrenes (evident by the shifting isosbestic point in Fig. S10, ESI †). The time constant of hot 1 (BsN) formation is observed on a 2 ps timescale on the low wavenumber side of the signal, however, due to the spectral overlap of the cooling and formation processes, the actual formation timescale might actually be a bit slower; we conservatively estimate this value to be less than 20 ps. When CCl 4 is used as the solvent, the lifetime of singlet 1 (BsN) nitrene is substantially longer than that of VC. Thus, in CCl 4 the cooled singlet nitrene 1 (BsN) observed at 1360 cm À1 has a formation time constant of t SN (CCl 4 ) = 20 AE 1 ps and its relaxation lifetime is t ISC (CCl 4 ) = 0.75 AE 0.10 ns.
Previous experimental and computational studies show that singlet sulfonyl nitrenes are unstable, highly reactive species that rapidly convert into more stable triplet isomers via intersystem crossing. 6,23 Recently, we confirmed that such transformation has an activation barrier of as low as 2 kcal mol À1 . 21 That means that the formation of triplet nitrene is, apparently, the predominant if not the only process of transformation of the singlet nitrene and the two processes should occur with the  (BsN) nitrene decay, a new persistent (lifetime 43 ns) positive signal appears in CH 2 Cl 2 at 1323 cm À1 belonging to 3 (BsN) with the time constant t ISC (CH 2 Cl 2 ) = 20 AE 1 ps (see Fig. 3b and 4). A 'rise-decayrise' profile of the 1323 cm À1 curve in Fig. 4 results from the imposed exponentials for the appearance of hot singlet nitrene, its decay, and the formation of the triplet nitrene, as can also be followed by color changes at 1323 cm À1 ( yellow-orange-yellowred) in Fig. 3b. A similar time constant of formation was also detected near 1148 cm À1 in the n s (SO 2 ) region ( Fig. 3e and 4). With CCl 4 as the solvent the band appears much later at 1347 cm À1 and has a time constant of t ISC (CCl 4 ) = 0.70 AE 0.23 ns (Fig. 4). According to the B3LYP results (Fig. S4, Table S3, ESI †) the long-lived species may be assigned to either 3 (BsN) (1342 cm À1 ) or B-PCP (1354 cm À1 ) or both. The B-PCP has also an additional predicted band at 1377 cm À1 of double intensity. Experimentally, a low intensity band was observed at 1365 cm À1 and 1360 cm À1 in CCl 4 and CH 2 Cl 2 , respectively, and assigned to B-PCP. The calculated band at 1354 cm À1 is therefore expected to exhibit an even lower intensity, leading us to assign the 1323 cm À1 feature in CH 2 Cl 2 to 3 (BsN) nitrene.
Singlet and triplet nitrenes are not the only products of photolysis derived from the azide S 1 excited state. For arylsulfonylazides, the formation of a PCP was postulated. 6,38 Both DFT functionals used here predict the presence of B-PCP, appearing as a 29 cm À1 red-shifted signal to 1377 cm À1 in the case of the B3LYP functional (Fig. S4, ESI; † it red-shifts only 8 cm À1 to 1383 cm À1 for M06-2X, see Fig. S7, ESI †) near n as (SO 2 ) of the azide ground state. For instance, for BsN 3 in CCl 4 and CH 2 Cl 2 the signal of B-PCP was detected at 1365 and 1360 cm À1 , respectively. From the trace at 1365 cm À1 in CCl 4 it is not possible to extract the parameters of B-PCP formation since it significantly overlaps with that of 1 (BsN). However, we were able to derive these time constants from the observed trace at 1360 cm À1 (see Fig. 4), and to obtain t RIES (CH 2 Cl 2 ) = 17 AE 1 ps. The B-PCP's formation time constant of 17 AE 1 ps correlates well with the time constant of the azide S 1 excited state depletion (at 1559 cm À1 , 22 AE 2 ps) as well as 1 (BsN)'s lifetime. However, previous studies have shown that singlet sulfonyl nitrene cannot be a precursor of the PCP since there is no evidence of PCP formation when a non-azide precursor is used 23 and that both B-PCP and 1 (BsN) are formed exclusively from the S 1 excited state of the azide. 6 This conclusion was also confirmed by our unsuccessful attempts to locate a transition state between 1 (BsN) and the B-PCP (vide infra).
TsN 3 . In general, due to the structural resemblance, the photochemistry of TsN 3 is similar to that of BsN 3 . However, some differences deserve to be mentioned. In the SO 2 spectral region ( Fig. S11-S13, ESI †) a bleach of the azide S 0 state was observed at 1379 cm À1 in CH 2 Cl 2 (1373 cm À1 in CCl 4 ). The evolution of the azide S 0 state at 1379 cm À1 reveals t IC (CH 2 Cl 2 ) = 20 AE 3 ps (1373 cm À1 , 30 AE 7 ps in CCl 4 ) which correlates with that observed in the azide spectral region. The calculations also predict an intense signal of TolNSO 2 (T-PCP) which overlaps with the S 0 state of the azide. The recovery of the 1373 cm À1 band was estimated from the curve fitting and found to be 25% (in CH 2 Cl 2 ), which is 10% more than that for the 2129 cm À1 trace (assigned to nN 3 ). This allows us to conclude that PCP formation also occurs for TsN 3 . A high value of 1373 cm À1 band recovery in CH 2 Cl 2 is due to its overlap with the n as (SO 2 ) band of T-PCP (1353 cm À1 ), and it has a formation time constant of t RIES (CH 2 Cl 2 ) = 14 AE 3 ps. This value is similar to the recovery time constant of n as (SO 2 ) of the azide ground state (t IC (CH 2 Cl 2 ) = 20 AE 3 ps at 1373 cm À1 ). The formation time constants of 3 (TsN) are somewhat different from those of 3 (BsN) but equal to the singlet nitrene lifetime t ISC (CH 2 Cl 2 ) = 25 AE 2 ps (1305 cm À1 ) and t ISC (CCl 4 ) 0.44 AE 0.26 ns (1337 cm À1 ), while lifetimes t S of 1 (TsN) were estimated to be B20 ps (CH 2 Cl 2 ) and 0.67 AE 0.10 ns (1349 cm À1 , CCl 4 ), respectively. MsN 3 . Unfortunately, we found no signals in the SO 2 spectral window except for a bleach of the MsN 3 ground state at 1366 cm À1 (CH 2 Cl 2 , see Fig. S14, ESI †) and 1378 cm À1 (CCl 4 ) showing no IC process. Nevertheless, one persistent (43 ns) signal at 1134 cm À1 that is assigned to n s (SO 2 ) of triplet nitrene was observed in CH 2 Cl 2 (Fig. S3, ESI †), where sulfonylazides showed fast transient dynamics. The signal has a Fig. 4 Transient kinetics at 1323 cm À1 (triplet 3 (BsN) and hot singlet nitrene peak), at 1148 cm À1 (n s (SO 2 ), triplet 3 (BsN)) and at 1360 cm À1 (B-PCP) in CH 2 Cl 2 (left). Transient kinetics at 1360 cm À1 (assigned to vibrationally cooled singlet 1 (BsN)) and 1347 cm À1 (with overlapping contributions from the hot singlet nitrene peak and triplet 3 (BsN)) in CCl 4 (right). Both panels contain the parameters resulting from the fit (solid line) to the data curves (dots).
formation time constant of 34 AE 3 ps. Based on the B3LYP results (Fig. S6, ESI †) this signal corresponds to a mixture of 3

(MsN) and MeNSO 2 (M-PCP). Predicted wavenumber values
are 1155 and 1157 cm À1 , respectively. The intensity ratio of the two species is about 5 : 1. However, the steady state FTIR spectrum does not show a permanent signal near 1134 cm À1 . Therefore, the observed signal at 1134 cm À1 corresponds to the triplet nitrene 3 (MsN).

Computational results
In order to assign the vibration bands of all species which could be formed by irradiation of azides we employed B3LYP/6-311++G(3df,3pd) and M06-2X/6-311++G(d,p) functionals/basis sets. FTIR spectra of azides along with the B3LYP predicted bands are shown in Fig. 5. The ground state N 3 vibrational frequencies are well reproduced by calculations; the error does not exceed 6%, while the M06-2X error is B10%. Previous studies showed that a better approximation of SO 2 frequencies may be obtained when large basis sets including d and f functions are used. 6,39 However, we have found that both DFT functionals provide a very good approximation of SO 2 frequencies, while 6-311++G(d,p) basis sets in combination with the M06-2X functional show an excellent agreement with experimental frequencies of the SO 2 group (Table 1).
Since B3LYP/6-311++G(3df,3pd) calculations of the azide S 1 state are extremely time-consuming, we present the vibrational frequencies of S 1 states calculated at the M06-2X level for all azides and at the B3LYP level only for MsN 3 . Full lists of the vibrational frequencies for the S 0 , S 1 and T 1 states of azides (RSO 2 N 3 ), singlet nitrenes 1 (RSO 2 N), triplet nitrenes 3 (RSO 2 N) and pseudo-Curtius photoproducts (RNQSO 2 ) are given in Tables S2-S13 (ESI †) and plotted in Fig. S4-S9 (ESI †). It should be noted that M06-2X/6-311++G(d,p) adequately predicts PCP and azide ground state vibrational frequencies, but shows some inaccuracy in the prediction of singlet and triplet nitrenes while B3LYP/6-311++G(3df,3pd) adequately predicts the vibrational frequencies for the azide ground state, singlet and triplet nitrenes, but not for the PCP.
The B3LYP/6-31G(3df,3pd) optimized geometries of the azides were used for calculations of vertical excitations by the TD-B3LYP method. The lowest electronic transition occurs at 278, 277 and 258 nm in the case of BsN 3 , TsN 3 and MsN 3 , respectively, which corresponds to the HOMO-LUMO electronic transition. In MsN 3 the transition has a very low predicted oscillator strength f = 0.0001 at 258 nm. The predicted transitions are in good agreement with the ground-state electronic absorption spectra of the azides (Fig. 6). Note that the laser excitation wavelength was centered at 267 nm (having 1.2 nm FWHM), and because the difference in wavelength between the S 0 -S 1  Table 1 Calculated and experimental IR frequencies (in CCl 4 , cm À1 ) for BsN 3 , TsN 3 and MsN 3 ground states using B3LYP/6-311++G(3df,3pd) and M06-2X/6-311++G(d,p) functionals and S 0 -S 2 electronic transitions is calculated to be 420 nm, the collected experimental data refer to the S 0 -S 1 transition.
In addition, electron density difference plots of the singlet excited states were calculated at the TD-B3LYP/6-31G(3df,3pd) level of theory for the azides in order to identify the character of the excited states, according to ref. 6 (Fig. S15, ESI †). This approach shows that the S 1 excited state corresponds to the promotion of an electron from the sulfonyl oxygen lone pair to the p*-orbital of the azide group, and in the case of the aromatic compounds also to the in-plane p*-orbital of the aromatic system. Accumulation of electron density on the p*-orbital of the terminal N b QN g moiety suggests that the S 1 excited state is a dissociative state and that the corresponding nitrene can be formed from the initially excited sulfonyl azide. This prediction is consistent with our experimental observations (vide supra).
If the pseudo-Curtius rearrangement products cannot be formed from the corresponding nitrenes (which we have experimentally confirmed, and as reported earlier 6,23 ), there must exist an alternative route to them from the original azides. Indeed, two independent transition states (TSs), each with only one imaginary mode, were located on the PESs of the studied azides. One of them leads to nitrene, while another one, lying B2 kcal mol À1 (R = p-BrC 6 H 4 and p-Tol) or B10 kcal mol À1 (R = Me) higher, leads to the PCP. Although the structures of the two TSs depicted in Fig. 7 look similar, they are principally different as proved by IRC calculations and by the analysis of the imaginary modes. The first TS (top row in Fig. 7) leading to nitrenes is located mainly on the N a Á Á ÁN b bond and reflects the process of elimination of the N 2 molecule. The second TS (bottom row) leading to PCPs has a B150 cm À1 higher imaginary wavenumber, suggesting a steeper reaction valley and being in line with a larger barrier. 40 Comparable contributions of the N a Á Á ÁN b bond elongation and simultaneous decrease of the CÁ Á ÁN a distance due to contraction of the NSC angle, together with notable C-S bond elongation, are clearly indicative of changes finally leading to the PCP formation. These results can be considered as an independent theoretical confirmation of the aforementioned conclusion that singlet sulfonyl nitrenes cannot be precursors for the PCP formation.
We also tried to locate a transition state on the way from singlet or triplet nitrenes to PCPs. For singlet nitrenes no TS could be found. In contrast, we did locate a TS (with an energy of 42 kcal mol À1 ; not shown in Table 2) for triplet 3 (MsN) leading to triplet PCP. This is in agreement with the results of matrix isolation experiments of the triplet nitrene 3 (CF 3 SO 2 N) and its further UV-promoted transformation into the final product CF 3 NSO 2 via the elusive 3 (CF 3 NSO 2 ). 16 Photochemical reaction paths Irradiation of BsN 3 , TsN 3 and MsN 3 with 267 nm light directly populates the S 1 excited state of the azides (vide supra). The observed photochemical reaction paths are summarized in Fig. 8. The S 1 excited state of the azides is a short-lived dissociative species (k S 1 (CH 2 Cl 2 ) B 45 Â 10 9 s À1 for BsN 3 ) and expels the molecule of nitrogen to give singlet nitrene ( 1 (RSO 2 N); with k SN (CH 2 Cl 2 ) of B50 Â 10 9 s À1 for BsN 3 ), and the Curtius-like photoproduct RNSO 2 , where R = p-BrC 6 H 4 , p-Tol or Me; with k RIES (CH 2 Cl 2 ) B 59 Â 10 9 s À1 for BsN 3 /TsN 3 . Singlet nitrenes were detected as vibrationally hot species since the experimental 267 nm excitation wavelength corresponds to 107.5 kcal mol À1 of energy, whereas the energy required for the azide degradation is ca. 40 kcal mol À1 (see Table 2 below). The excess energy absorbed by the azide molecule is therefore about 70 kcal mol À1 .
As shown in Table 2, transformation of the azide S 0 state to 1 (RSO 2 N) involves an energy barrier of B39 kcal mol À1 . Calculations show that the formation of a PCP from the azide S 0 state is influenced by the substituent at the SO 2 group and exhibits an energy barrier of B40 kcal mol À1 for arylsulfonyl azides and 50 kcal mol À1 for MsN 3 . Thus, since the barriers of transformation of arylsulfonyl azides to singlet nitrenes and PCPs are similar, one could expect the formation of both species for all compounds. However, singlet nitrene formation from MsN 3 is B10 kcal mol À1 more preferable than the formation of a PCP. The rest of the excitation energy of 107.5 kcal mol À1 absorbed by the azide molecules is available for product formation. It also    3 and TsN 3 in both solvents but not for MsN 3 . Subsequent relaxation of hot nitrene gives rise to the formation of a cooled singlet species 1 (RSO 2 N). For 1 (RSO 2 N) the B3LYP calculations predict n as = 1375, 1370 and 1382 cm À1 for R = p-BrC 6 H 4 , p-Tol and Me, respectively. These wavenumbers are in good agreement with the experimentally observed absorption signals of relaxed singlet nitrenes 1 (BsN) and 1 (TsN) detected at 1360 and 1349 cm À1 in CCl 4 . Their decay constants were found to be k ISC (CCl 4 ) = (1.3 AE 0.4) Â 10 9 s À1 and (1.5 AE 0.4) Â 10 9 s À1 , respectively. Determination of singlet nitrene decay constants in CH 2 Cl 2 is complicated by VC and we tentatively estimate it to be approximately 50 Â 10 9 s À1 .
Similar to the features observed for singlet nitrenes, weak persistent (43 ns) bleaches of triplet 3 (BsN) and 3 (TsN) nitrenes were observed at 1347 and 1337 cm À1 in CCl 4 . The formation constants are k ISC (CCl 4 ) = (1.4 AE 1.0) Â 10 9 s À1 and (2.3 AE 1.7) Â 10 9 s À1 , respectively, which are comparable to the singlet nitrene decay constants, thus indicating that both transitions correspond to the same process. 6 The decay constants of singlet nitrenes in CH 2 Cl 2 are deduced from the formation constants of triplet nitrenes which were found to be k ISC (CH 2 Cl 2 , 3 (BsN)) = (50 AE 5) Â 10 9 s À1 and k ISC (CH 2 Cl 2 , 3 (TsN)) = (40 AE 6) Â 10 9 s À1 , and therefore are significantly larger than in CCl 4 . Note also that a persistent band (43 ns) of 3 (MsN) was detected at 1134 cm À1 in CH 2 Cl 2 , which was assigned to the triplet nitrene formation with the formation constant k ISC (CH 2 Cl 2 ) = (29 AE 5) Â 10 9 s À1 . The latter value is slightly lower than that for 3 (BsN) and 3 (TsN), apparently due to the fact that two single time traces are compared, which do not necessarily need to represent the absorption of a single species (i.e. their transient absorptions could overlap). It is an indicator of a transient kinetic difference between aliphatic and aromatic azide photochemistry along with the absence of the azide ground state recovery.
Another permanent product of acyl-and sulfonylazide photochemistry is the PCP. 5,6,16 Its formation was clearly detected for BsN 3 and TsN 3 in CH 2 Cl 2 , while in CCl 4 the PCP's SO 2 band overlaps with that of the corresponding singlet nitrene. The trace near 1360 cm À1 in CH 2 Cl 2 , corresponding to cooled B-PCP, was found to have the formation constant k RIES (CH 2 Cl 2 , B-PCP) = (59 AE 7) Â 10 9 s À1 . The value correlates well with the azide S 1 state decay (k ISC (CH 2 Cl 2 , BsN 3 ) = (46 AE 8) Â 10 9 s À1 at 1559 cm À1 ) and falls in the range of the singlet nitrene 1 (BsN) decay constants. Previous experiments showed that only azide, not singlet nitrene, is a predecessor of the PCP. 6,23 Furthermore, the transition state has been located to connect the ground state sulfonyl azide and PCP structures ( Table 2). This transformation requires B41 kcal mol À1 for the aromatic azides, a similar value as calculated here for TS1 leading to singlet nitrene formation. However, the energy of transition state TS2 leading to MeNSO 2 was calculated to be 10 kcal mol À1 higher than that leading to 1 (MsN). According to calculations, the formation of 1 (MsN) is B3.3 Â 10 4 times faster than that of MeNSO 2 and one could expect exclusive formation of 1 (MsN) rather than MeNSO 2 .

Conclusions
The photochemistry of p-bromophenylsulfonyl-, p-tolyl-and methylsulfonyl azides has been investigated experimentally by ultrafast time-resolved UV-pump-IR-probe spectroscopy and computationally at the B3LYP/6-311++G(3df,3pd) and M06-2X/6-311++G(d,p) levels of theory. All compounds showed the absence of spectral signatures corresponding to the azide S 1 and T 1 excited states in the regions of nN 3 or nSO 2 . Only for BsN 3 the S 1 excited state with a lifetime of t S 1 (CH 2 Cl 2 ) = 21 AE 3 ps was observed by a low intensity feature in the in-ring C-C stretch region. The internal conversion of the undetected S 1 excited state to the ground state of azide is low, representing an S 0 state recovery of B15%. Still, it represents an efficient way of the azide ground state recovery in BsN 3 and TsN 3 , but was not observed for MsN 3 . The used laser excitation wavelength provides more energy than that required for N 2 release, and the 1 (BsN) and 1 (TsN) singlet nitrenes were detected as vibrationally hot species in both CH 2 Cl 2 and CCl 4 . In the former solvent the lifetime t VC (CH 2 Cl 2 ) is B20 ps. The lifetimes of relaxed singlet 1 (BsN) and 1 (TsN) nitrenes are t VC (CCl 4 ) = 0.75 AE 0.10 and 0.66 AE 0.10 ns, respectively. Corresponding triplet nitrenes were detected as persistent (43 ns) species and their formation time constants correlate with the lifetime of singlet nitrenes. Singlet 1 (MsN) was not observed but its triplet spin isomer was detected. The formation of relaxed B-PCP correlates with the azide S 1 state decay and represents an example of rearrangement in the excited state. The assignment of the experimentally detected vibrational bands to species was supported by quantum chemical calculations.
In conclusion, for many years chemists believed that the formation of N-sulfonylamines RNSO 2 and their better known carbonyl analogues, isocyanates RNCO proceeds in two steps via generation of the corresponding nitrenes followed by their Curtiustype rearrangement. 42 However, recent investigations 6,43 as well as the results of the present work employing ultrafast laser techniques call into question these notions and suggest an alternative, concerted mechanism of the Curtius-type rearrangement, at least as far as the formation of N-sulfonylamines from sulfonyl azides is concerned. Indeed, both our and other's experiments 6,43 prove the formation of PCPs and singlet nitrenes from the S 1 excited state of sulfonyl azides by comparing the formation/degradation time constants. Also, our calculations predict the existence of two different transition states, one connecting the azide and the singlet nitrene and another one connecting the azide and the pseudo-Curtius rearrangement product. This may encourage chemists to reinvestigate and reconsider the mechanism of other similar sextet rearrangements, like Lossen, Beckmann, Schmidt, Hofmann and Wolff rearrangements.