¹⁷O-NMR spectra of compound 1 dissolved in chloroform-d₁ (CDCl₃) were recorded at 14.1 T and various temperatures ranging between −30 and 110 °C (Figure 4). Two signals of similar integrated intensity are observed at low temperature (see Supporting Information). On the basis of chemical shift data reported for similar compounds [8,9,10,11,12], the signal at about 350 ppm is assigned to the carbonylic O₁ oxygen atoms and, conversely, the signal at about 250 ppm is assigned to the ester-like O₂ oxygen atoms. At −30 °C, the chemical shift (δ) and the full linewidth at half height (LW), as determined by fitting Lorentzian lines, are respectively 349 ppm and 3.28 kHz for the O₁ signal and 246 ppm and 1.49 kHz for the O₂ signal. Both these signals broaden for increasing temperature and coalescence occurs at about 5 °C. At 10 °C, a single signal is observed at δ ≈ 300 ppm, i.e., at the average of the O₁ and O₂ chemical shift values measured at low temperature. The LW of this signal decreases for increasing temperatures while the chemical shift is essentially constant.

These observations are consistent with the evolution expected for an equally populated two-site exchange. At the coalescence temperature (T_c), the first order rate constant characterizing such an exchange process is directly related to the difference in resonance frequency in each site (Δν = ν₁ ‑ ν₂); the activation free energy at T_c can then be determined via the Eyring equation (see Supporting Information). For a chemical shift difference Δδ = (δ₁ − δ₂) = 103 ppm (Δν = 8.30 kHz at 14.1 T) and T_c = 5 °C, ΔG^# is estimated to be (45.2 ± 0.5) kJ/mol. The error on ΔG^# was estimated considering an absolute error on T_c of ± 3 °C. It is noteworthy that an error of ±5 ppm on Δδ affects ΔG^# by only ±0.1 kJ mol⁻¹.

Further information can be gained from the temperature dependence of the LW, which is shown in Figure 5 and was analyzed according to Equations (1–4), as in our previous ¹⁷O-DNMR investigation [10].

In Equation (1), LW_q is the natural linewidth due to quadrupole relaxation and LW_ex is the contribution due to chemical exchange. The temperature dependence of LW_q is accounted for by an Arrhenius-type expression (Equation (2)) where the pre-exponential factor is the natural linewidth at 298 K and E_q is the activation energy. Identical activation energy was assumed for the quadrupole relaxation of both the O₁ and O₂ oxygen nuclei. Above the coalescence temperature, LW_q is given by the arithmetic average of the natural linewidth values predicted for O₁ and O₂. The broadening due to chemical exchange between two equally populated sites is given by Equations (3a) or (3b), depending on whether the kinetics is slow or fast on the NMR spectral time scale. In Equation (3b), ν₁− ν₂ is the frequency difference between the O₁ and O₂ signals, k is the first-order rate constant, the temperature dependence of which is given by the Eyring equation (Equation (4)). ΔS^# and ΔH^# are the activation entropy and the activation enthalpy, respectively, which are assumed to be temperature independent, and the other symbols have their usual meaning. The best-fit of Equations (1)–(4) to the experimental LW data, using a least-squares technique based on the relative residuals [13], yields ΔG^# = (45.5 ± 0.1) kJ/mol at 25 °C, which agrees with the value estimated at the coalescence (vide supra), ΔH^# = (45 ± 1) kJ/mol and ΔS^# = (−1 ± 5) J mol⁻¹ K⁻¹ [14]. Interestingly, the LW analysis reveals that the entropic contribution is not significant.

In order to investigate possible solvent effects, ¹⁷O-NMR spectra were recorded at 14.1 T and several temperatures for compound 1 dissolved in dichloromethane-d₂ (CD₂Cl₂) and in acetonitrile-d₃ (CD₃CN). The spectra recorded in these more polar solvents are highly similar to the spectra recorded at the same temperature in CDCl₃ (Figure 6). T_c is found to be about 10 °C in CD₂Cl₂ and is somewhat higher, ~15 °C, in CD₃CN. Similar chemical shift values are observed, leading to similar ΔG^# values (Table 1).

¹⁷O-NMR spectra of compound 2 dissolved in CDCl₃ were recorded at 14.1 T and various temperatures ranging between −20 and 90 °C (Figure 7A). At low temperature, the signal of trifluoroacetic acid, which is present as an impurity in the sample, is the most intense signal because 2 is not entirely soluble and also because the ¹⁷O-NMR signals of 2, which could be detected at about 350 ppm (O₁) and 220 ppm (O₂), are highly broadened. At 0 °C in CD₃CN (Figure 7B), the O₁ and O₂ signals are observed at 341 ppm and 218 ppm, respectively. Similarly to the spectra of compound 1, the spectra of 2 reveal the occurrence of a dynamic process which interchanges O₁ and O₂. The coalescence is observed at about 20 °C in CDCl₃ and ΔG^# is estimated to be (47.3 ± 0.5) kJ/mol.

¹⁷O-NMR spectra of compound 3 dissolved in CDCl₃ were recorded at 14.1 T and various temperatures ranging between −25 and 12.5 °C (Figure 8A). The ¹⁷O-NMR signal of the methoxy group was detected at 62 ppm, i.e., in the expected range of chemical shift [15] (see Supporting Information). At −25 °C, the O₁ and O₂ signals are observed at 349 ppm and 254 ppm, respectively. The coalescence occurs somewhat below 0 °C; using T_c = −5 °C, ΔG^# is estimated to be 43.7 kJ/mol.

For solubility reasons, compound 4 was dissolved in a 98/2 (v/v) mixture of CDCl₃ and DMSO-d₆ rather than in pure CDCl₃. The ¹⁷O-NMR spectra were recorded at 14.1 T and various temperatures ranging between −25 and 37.5 °C (Figure 8B). The signal of the NO₂ group was observed at 573 ppm, i.e., in the expected range of chemical shift [15] (see Supporting Information). At −25 °C, the O₁ signal is observed at 359 ppm. The solubility of 4 is rather low and, consequently, acetic acid gives rise to an intense signal which masks the O₂ signal at about 255 ppm. Coalescence is observed between 0 and 12.5 °C; using T_c = 5 °C, ΔG^# is estimated to be 45.2 kJ/mol.

Table 1 shows that the chemical shift of O₁ is barely affected by the trifluoromethyl group but the chemical shift of O₂ is reduced by about 30 ppm with respect to the value measured for bis(acetoxy)iodobenzene 1. In contrast, the ¹⁷O chemical shift data are quite insensitive to the para-substitution of the aromatic ring. This confirms the previous observations made for the average signal detected in the fast exchange regime [7]. In the present work, the O₁ and O₂ signals are, for the first time, distinctively observed and it can now be stressed that the previous observations are not the consequence of opposite chemical shift variations.

The ¹⁷O-DNMR results obtained for compound 1 indicate that, in CDCl₃, the activation entropy characterizing the oxygen atom exchange is negligible and suggest that the activation free energy, about 46 kJ/mol, is rather insensitive to the dielectric constant of the solvent. The ¹⁷O-DNMR results also indicate that ΔG^# is hardly affected by the chemical alteration present in compounds 2–4. Indeed, the ΔG^# data measured for 4 and 1 are not significantly different. The replacement of the acetoxy groups of 1 by trifluoroacetoxy groups increases ΔG^# by less than 2 kJ/mol while a methoxy group in para-position has the opposite effect.

As mentioned in the Introduction, the oxygen atom exchange in 1 was ascribed to a [1,3]-sigmatropic shift of iodine [8]. The activation energy barrier (ΔE) of this process can be calculated as the total energy difference between the transition state (TS) and the most stable rotamer. Our previous DFT study of compound 1 yielded ΔE = 45.2 kJ/mol in vacuo [8], a value which is in excellent agreement with the experimental results (Table 1). In our previous study on 1 and related compounds, we observed that the calculated energy profiles do not vary substantially when including the solvent, which was described in the framework of the polarizable continuum model (PCM) [16,17,18,19]. The DFT calculations yielded ΔE = 48.3 kJ/mol for 1 in chloroform. No important solvent effect on the activation energy barrier was observed and this also agrees with the experimental results of this work. On the base of these considerations, we believe that in vacuo calculations provide reliable results for the systems under investigations. In principle, it is the calculated activation free energy, and not ΔE, which should be compared to the experimental ΔG^# values. However, various comparisons of DFT and experimental results have shown that the best agreement is obtained when no thermodynamical correction is applied to the calculated total energy; this point is discussed in the recent review of Casarini et al. [20]. Reasonably, the comparison between calculated energy barriers and experimental activation free energies is meaningful if the activation entropy is negligible or small, as it is the case for compound 1. In the following, it is assumed that there is no significant entropic contribution to the activation free energy measured for compounds 2–4.

DFT calculations were completed for compounds 2–4 with the main objective of verifying if the interchange of O₁ and O₂ observed by NMR for these λ³ iodanes can also be explained by the [1,3]-sigmatropic shift of iodine. The minimum energy structures of 2–4 show strong resemblance with those of 1 (see Figure 9 for 2), although some differences in the relative energies can be noted (Table 2). The stable rotamers of 2 exhibit the smallest variations in total energy with essentially the same value for the conformations A and B.

Compound 2 is the one showing the largest experimental differences with compound 1. To verify whether the differences might be related to differences in the exchange mechanism, the latter was studied as previously done for compound 1 in reference [8]. The calculated energy profile is reported in Table 3 together with relevant structural data. It is similar to the profile found for 1, in having the maximum for a I-C_ac-O₁ angle comprised between 55° and 60° and in involving a rotation of 90° of the carboxylate plane during the shift (in this way going from conformation A to B).

It is reasonable to assume that the same exchange mechanism occurs for compound 3 and 4. In all cases, the calculated TS (see Supporting Information) are similar to that of 1 and the normal mode corresponding to the imaginary frequency clearly involves the oxygen atoms exchange. The energy barriers calculated in vacuo for compounds 2–4 are given in Table 1. The agreement with the corresponding experimental activation free energy is excellent, indicating that all of the studied bis(acyloxy)iodobenzenes are fluxional as a consequence of the [1,3]-sigmatropic shift of iodine and corroborating that this process is indeed quite insensitive to chemical alterations such as para- substitution of the arene moiety or modification of the acyl groups.

As mentioned in the Introduction, para-substitution of bis(acetoxy)iodobenzenes does not significantly affect the ¹³C nor ¹⁷O chemical shift of the acetoxy groups [7]. This contrasts with the observations on relatively similar systems, such as para-substituted benzyl alcohols [21] and phenylacetic acids [22]. We are not able, for the moment, to give a clear-cut explanation of this difference but the lack of electronic effect transmission between iodine and oxygen might be related to the bond order. X-ray measurements [23,24] on compound 1 showed that the I-O bond length is 216 pm, a figure slightly greater than the sum (201.9 pm) of the iodine (134.5 pm) and oxygen (67.4 pm) covalent radii [25], and smaller than the sum (348 pm) of the iodine (196 pm) and oxygen (152 pm) van der Waals radii [26]. Similar results are observed for the I-S distance, as reported by Varvoglis on some carbamoyl(diaryl)iodanes [27,28]. X-ray studies showed this distance to be 305.3 pm, a value comprised between the sum of the iodine and sulphur covalent radii (238.7 pm) and the sum of their van der Waals radii (400 pm). From these data, as well as other significant arguments, the I-S bond order was estimated to be 0.5 [27]. Good estimate of bond orders can be obtained by computation. In fact, Kiprof [29] has shown for a series of hypervalent iodine derivatives that there is a quite good correlation between the experimental I-O bond length and the I-O bond order as estimated by the Wiberg bond index. Kiprof's value of the I-O bond order of 1 is 0.41, in full agreement with the values computed in this work for the minimum energy conformation (Table 4). We note that the I-O bond order, which has been shown [29] to be affected by variation of the substituent at the iodine atom, is actually insensitive to change of the acyl groups or to para-substitution of the aromatic ring. It thus seems reasonable that the rather low energy barrier of the [1,3]-sigmatropic shift of iodine in compounds 1–4 is related to the low I(III)-O bond order in the stable conformations. Finally, it is noteworthy that the I-O bond order is somehow conserved in the transition state.

The NMR spectra were recorded on a Varian VNMRS spectrometer operating at 14.1 T (599.9 MHz for ¹H and 81.33 MHz for ¹⁷O) equipped with a 5 mm autoX dual broadband probe and temperature regulation. The sample was left to reach equilibrium at the desired temperature within the magnet for at least 15 min before the NMR measurements. The ¹⁷O-NMR spectra were recorded with the improved RIDE (ring down elimination) pulse sequence of Kozminsky et al. [32] using a 100 μs cawurst adiabatic inversion pulse and the following acquisition parameters: spectral width of about 850 ppm (~69 kHz) centered at 290 ppm, 5 ms relaxation delay, 10 μs (90°) excitation pulse, 10 μs (rof2) preacquisition delay (alfa = 0), 5 ms acquisition time and a number of transients ranging between 5 × 10⁴ and 5 × 10⁶. The receiver ddrtc parameter was optimized to obtain spectra without first-order phase error. The spectra were recorded lock-on without sample spinning. Unless otherwise stated, the processing comprised correction of the first three points of the Free Induction Decay (FID) by backward linear prediction, zero filling, Fourier transform, zero-order phase correction and baseline correction. The chemical shift scale was calibrated at 25 °C with respect to a sample of pure H₂O used as an external chemical shift reference (0 ppm).

Structure optimizations of all compounds were performed using the same theory level used in our previous investigations of hypervalent iodine compounds [8,9], which was shown to perform well on these classes of compounds. More specifically the calculations have been carried out at the DFT level employing the PBE0 functional [33], a parameter-free hybrid variant of the Perdew, Burke, and Ernzerhof (PBE) generalized gradient functional [34], as implemented in the commercially available Gaussian 03 suite of program [35]. The effective core-potential valence basis set LANL2DZ (i.e., D95V [36] basis set for the first row elements and the Los Alamos ECP plus DZ on iodine) [37] extended with polarization (d) and diffuse (p) functions [38,39] was employed for all atoms. Numerical integration was performed using a pruned grid having 99 radial shells and 509 angular points per shell. Vibrational analysis was carried out at the same level of theory to check the character of all stationary points. The starting geometries of the minima and TS states of compounds 2–4 were constructed by modifying (with appropriate substitutions) the minima and the TS states of compound 1, as obtained in a previous investigation [8]. The freely available program Molden was used for this purpose [40]. Cartesian coordinates and energies of the optimized geometries of minima and TS are reported in the Supporting Information. NBO version 3 [41] as implemented in Gaussian03 was employed for the analysis of the Wiberg bond index [42]. Graphics of molecular models were generated using the freely available VMD software [43].