Photoinduced reactivity in a dispiro-1,2,4-trioxolane: adamantane ring 1 expansion and first direct observation of the long-lived triplet diradical 2 intermediates 3

14 The dispiro-1,2,4-trioxolane 1 , an ozonide with efficient and broad antiparasitic activity, was 15 synthesized and investigated using matrix isolation FTIR and EPR spectroscopies together 16 with both and theoretical 17 methods. Irradiations ( λ ≥ 290 nm) of the matrix isolated 1 (Ar or N 2 ) afforded exclusively 18 4-oxahomoadamantan-5-one 4 and 1,4-cyclohexanedione 5 . These results suggested that the 19 reaction proceeded via a dioxygen-centered diradical intermediate, formed upon homolytic 20 cleavage of the labile peroxide bond, which regioselectively isomerized to form the more 21 stable (secondary carbon-centered)/oxygen-centered diradical. In situ EPR measurements 22 during the photolysis of 1 deposited in a MeTHF-matrix led to the detection of signals 23 corresponding to two triplet species, one of which was short-lived while the


INTRODUCTION 33
Organic dispiro-endoperoxides, in particular dispiro 1,2,4-trioxanes and 1,2,4-34 trioxolanes (ozonides), have been attracting much attention since the discovery of artemisinin 35 more than 40 years ago 1-3 , which provided a completely new antimalarial structural prototype 36 of pharmacophore. The mechanism of bioactivation and action of the artemisinins and related 37 endoperoxide based drugs has been discussed thoroughly 4-14 and bioactivation is known to 38 require iron(II)-induced reductive cleavage of the peroxide bond to form oxygen-centered 39 radicals, followed by rearrangement to generate carbon-centered radical species that act In all the above chemical processes, radical or diradical species have been hypothesized 45 to be the key reactive intermediates, following the initial homolytic cleavage step. For 46 asymmetrically substituted dispiro-1,2,4-trioxolanes, represented by compound 1, and also 47 for their dispiro-1,2,4-trioxane analogues, it has been postulated, based on the final reaction 48 products obtained, that after the cleavage of the peroxide bond generating the dioxygen-49 centered diradical 2 a regioselective β-scission occurs concomitantly with the rearrangement, 50 leading to the carbon-centered/oxygen-centered diradical species 3 (see Scheme 1) 21-23 .   4-Oxahomoadamantan-5-one 4, was synthesized using the procedure described by 138 Renoud-Grappin et al. 35 To a suspension of 2-adamantanone (0.5 g, 3.33 mmol) and NaHCO 3 139

Matrix isolation infrared spectroscopy measurements 155
To prepare the low temperature matrices a sample of the solid compound to be studied 156 was placed in an especially designed thermoelectrically heatable mini-oven attached to the 157 vacuum chamber of a helium cryostat (APD Cryogenics closed-cycle helium refrigerator 158 system with a DE-202A expander). Before the measurements, the samples were subjected to 159 additional purification by continued high-vacuum (~10 -7 mbar) pumping, during 160 approximately 1 hour, at room temperature. The samples were then sublimed and the vapors 161 of the compound to be studied were deposited, together with a large excess of argon or N 2 , 162 onto a cesium iodide (CsI) substrate mounted at the cryostat cold tip (15 ± 0.1 K, as measured 163 by a silicon diode sensor connected to a Scientific Instruments digital temperature controller). 164 This temperature was kept during the overall experiments. 165 The IR spectra of the matrix-isolated compounds were recorded in the 400−4000 cm −1 166 range and with 0.5 cm −1 spectral resolution, using a Nicolet 6700 Fourier transform infrared 167 (FTIR) spectrometer, equipped with a deuterated triglycine sulfate (DTGS) detector and a 168 Ge/KBr beam splitter. The optical bench was continuously purged with a flux of dry and CO 2 169 filtered N 2 , to avoid interference from atmospheric H 2 O and CO 2 . 170

Matrix isolation EPR spectroscopy measurements 171
A 100 mM solution of compound 1 in MeTHF (100 μL) was degassed under high vacuum 172 (~3.0 x 10 -2 Pa) in a quartz EPR tube, which was sealed after three freeze-pump-thaw cycles 173 under the vacuum conditions. The MeTHF solution of compound 1 was irradiated in the EPR 174 cavity at 10 K. The X-band EPR signals were obtained at a resonance frequency of 9.40 GHz 175 using a Bruker E500 spectrometer at 10-50 K. 176 177

In situ UV irradiation experiments 178
In the steady state infrared spectroscopy experiments carried out in Ar and N 2 matrices, 179 in situ UV irradiation of the samples was undertaken using different approaches. In the case 180 of the experiments carried out in the Ar matrices, broadband UV radiation was used, as 181  Table 1), which are represented in 213 are the values resulting from the relative energies obtained in the B3LYP and M06-2X 218 calculations, respectively), i.e., both forms are predicted to be significantly populated in these 219 experimental conditions, and can then be expected to be present in the cryogenic matrices 220 investigated in this work (see below). The Cartesian coordinates of the two conformers, as 221 predicted by the two methods used in this study, are provided as Supporting Information 222 (Tables S1 and S2).     is very similar to that obtained in argon and is provided in the Supporting Information ( Figure S2).

238
The dispiro-1,2,4-trioxolane 1 was sublimed under high-vacuum at room temperature and 239 co-deposited with large excess of argon or N 2 onto the cold (15 K) CsI substrate of the used 240 cryostat. The infrared spectra of the prepared matrices were subsequently recorded. The 241 spectrum obtained for the compound in the argon matrix is shown in Figure 2, while that 242 obtained in the N 2 matrix is provided in the Supporting Information ( Figure S2). The two 243 spectra were found to be very similar, testifying the fact that the molecules of the compound 244 were well-isolated in both matrices and also that the conformational population of 1 existing 245 in the gas phase prior to deposition was efficiently trapped in the matrices. Also, the peroxide 246 The comparison of the spectrum obtained in the argon matrix with those theoretically 253 predicted (see Figure 2), shows that the B3LYP functional reproduces better the experimental 254 data throughout the spectrum than the M06-2X functional. Thus, spectra calculated using the 255 B3LYP functional will be used in this study for the analysis of the experimentally obtained 256 spectroscopic data of the remaining compounds under study. 257 The UV-induced reactivity of 1 was investigated in the two types of matrices used (Ar, 258 N 2 ). Argon matrix was used as standard work medium, while N 2 matrix, known to stabilize 259  (Tables S6-S14 and Figures S4 and S5). The B3LYP computed vibrational frequencies of 4 and 295 5 were scaled by 0.983 and 0.982, respectively.

296
According to the obtained results , the diradical rearrangement is regioselective , with no 297 evidence of formation of the alternative products, 2-adamantanone and oxocane-2,7-dione. 298 This can be clearly seen when one compares the infrared spectra of the photoproduced 299 species, generated after UV irradiation of the matrix-isolated 1, with those of 2-300 adamantanone and oxocane-2,7-dione (see Figure 3; additional calculated structural and 301 spectroscopic data on these two compounds are presented in the Supporting Information, 302 Figure S6 and Tables S15 and S16). The observed regioselectivity can be explained by the 303 expected greater stability of the (secondary carbon-centered)/oxygen-centered diradical 304 species 3, in comparison with the putative alternative (primary carbon-centered)/oxygen-305 centered diradical that would lead to the non-observed final products. In addition, spin 306 density calculations (performed at the B3LYP/6-311++G(3df,3dp) level) on the triplet state 307 dioxygen-centered diradical result in a considerable larger electron spin density in the tertiary 12 β-carbon of the spiroadamantyl substituent as compared to that of the secondary β-carbon of 309 the spiro-p-cyclohexanonyl substituent, which can also be considered an indication of a most 310 favorable rearrangement of the initially generated dioxygen-centered diradical into the 311 (secondary carbon-centered)/oxygen-centered diradical species 3, leading to the observed 312 final products. The in situ EPR measurements in the photolysis of 1 in MeTHF matrix were conducted 337 using 266 nm laser light (5 mJ) at 10-50 K. The X-band EPR signals were measured at a 338 resonance frequency of 9.40 GHz ( Figure 6). During the photolysis at 10 K (Figure 6 Ib), a 339 half-field signal at ~1550 G (triplet species A) was observed after 120 seconds, which is a 340 typical triplet species of diradicals, and another half-field signal at 1670 G (triplet species B) 341 was detected after 420 seconds, together with the 1550 G signal (Figure 6  To gather further insight regarding the reactivity of the triplet species B, the photolystate 357 at 10 K was warmed to 50 K under dark conditions (Figure 6 IIb). The typical triplet signal 358 disappeared and did not recover to the original one in Figure 6 Ic after re-cooling the sample 359 to 10 K, indicating that the triplet species B was also thermally labile at ~50 K. The triplet 360 signal of B was obtained by the difference spectrum of Figure 6 IIa-b (Figure 6 IIc). The 361 zero-field splitting parameters D/hc and E/hc of the triplet species B were determined to be 362 0.160 cm -1 and 0.001 cm -1 after the simulation (Figure 6 IId). From the D/hc value, the 363 distance between two spins was estimated to be ~ 5 Å. Adamantane-2-spiro-3'-8'-oxo-1',2',4'-trioxaspiro[4,5]decane 1, a reported potent 375 antiparasitic dispiro-1,2,4-trioxolane, has been synthesized and studied from the view point 376 of its molecular structure and monomeric photochemistry, using matrix isolation techniques 377 coupled to FTIR and EPR spectroscopy. It is noteworthy that the ozonide 1 remained 378 thermally stable upon sublimation required for matrix deposition. All experimental results 379 have been supported by quantum chemical calculations using two levels of theory [B3LYP/6-380 311++G(3df,3dp) and M06-2X/6-311++G(3df,3dp)], the comparison of the spectrum 381 obtained in the argon matrix with those theoretically predicted showing that the B3LYP 382 functional reproduces better the experimental data throughout the spectrum than the M06-383 2X functional. Both methods predict two conformers for the dispiro-1,2,4-trioxolane 1 with 384 very close energies, differing in the orientation of the spiro-p-cyclohenanonyl moiety, which 385 in the most stable form is directed to the peroxide group, while in the less stable one is 386 oriented towards the ring ether fragment. In the experimental spectra, several bands were 387