Azirinium ylides from α-diazoketones and 2H-azirines on the route to 2H-1,4-oxazines: three-membered ring opening vs 1,5-cyclization

Strained azirinium ylides derived from 2H-azirines and α-diazoketones under Rh(II)-catalysis can undergo either irreversible ring opening across the N–C2 bond to 2-azabuta-1,3-dienes that further cyclize to 2H-1,4-oxazines or reversibly undergo a 1,5-cyclization to dihydroazireno[2,1-b]oxazoles. Dihydroazireno[2,1-b]oxazoles derived from 3-aryl-2H-azirines and 3-diazoacetylacetone or ethyl diazoacetoacetate are able to cycloadd to acetyl(methyl)ketene generated from 3-diazoacetylacetone under Rh(II) catalysis to give 4,6-dioxa-1-azabicyclo[3.2.1]oct-2-ene and/or 5,7-dioxa-1-azabicyclo[4.3.1]deca-3,8-diene-2-one derivatives. According to DFT calculations (B3LYP/6-31+G(d,p)), the cycloaddition can involve two modes of nucleophilic attack of the dihydroazireno[2,1-b]oxazole intermediate on acetyl(methyl)ketene followed by aziridine ring opening into atropoisomeric oxazolium betaines and cyclization. Azirinium ylides generated from 2,3-di- and 2,2,3-triaryl-substituted azirines give rise to only 2-azabuta-1,3-dienes and/or 2H-1,4-oxazines.

Heating azadiene 3e in DCE under reflux (84 °C) for 3.5 h gave a 6:1 equilibrium mixture of 1,4-oxazine 4e and azadiene 3e (according to 1 Н NMR spectroscopy). Oxazine 4e is stable enough at room temperature to be isolated by column chromatography (    adduct 7h, which were isolated in 58 and 13% yields, respectively (Table 1, entry 9). Compounds 6-8 were characterized by standard spectral methods and the structures of adducts 6f and 7h were additionally confirmed by X-ray diffraction analysis ( Figure 2). The reaction of diazoacetylacetone 2c with 2,3diphenyl-2H-azirine (1a) provides oxazine 4i as a sole product (  [12] or 5-arylfuran-2,3-diones [13]. Therefore, the presence of compounds 8f,g among the reaction products provides evidence for the formation of some amounts of acetyl(methyl)ketene (12) under the reaction conditions, which, in turn, gives us insight to the mechanism of formation of adducts 6 and 7 (Scheme 3). A separate experiment was performed to show that these compounds are formed via independent pathways, as they do not interconvert under the reaction conditions (Rh 2 (Oct) 4 , 84 °C, DCE).
Several examples of the 1,5-cyclization of azomethine ylides bearing an α-keto group into oxazole derivatives were reported [27][28][29]. As also known, the azomethine ylide derived from N-benzylideneanisidine and diazoacetylacetone under Rh 2 (OAc) 4 -catalysis undergoes 1,3-cyclization to an aziridine derivative in high yield, rather than 1,5-cyclization [22]. However, no cyclizations of azirinium ylides, cyclic analogs of azomethine ylides, are known. This is not surprising in view of the high strain of the azirinium system, and until now ring opening in these systems seemed much more preferable than annelation of a new cycle. Nevertheless, we decided to study two competing pathways for isomerization of the model azirinium ylide 9j: ring opening into azadiene 3j and 1,5cyclization into azirenooxazole 10j (Scheme 4), by means of DFT calculations (B3LYP/6-31+G(d,p)). In addition, two reasonable pathways for the formation of adducts 6j and 7j formed from azirenooxazole 10j and ketene 12 were studied at the same level of theory (Scheme 4).
According to the calculations, the barrier to the 1,5-cyclization of ylide 9j to compound 10j was found to be even slightly lower (7.9 kcal/mol) than the barrier to the ring opening across the N-С2 bond to azadiene 3j (10.1 kcal/mol) (Figure 3). Azirenooxazole 10j is thermodynamically more stable than ylide 9j, by ca. 15 kcal/mol, but the barrier to the reverse reaction 10j→9j is not too high (22.6 kcal/mol). The ring opening in azirinium ylide 9j into azadiene 3j, too, has a low activation barrier but occurs irreversibly. Therefore, azirenooxazole 10j might form in this reaction, and, moreover, its formation from  ylide 9j is kinetically preferred over the formation of azadiene 3j. However, in view of the reversibility of the 1,5-cyclization 9j →10j and in the absence of an active trap for azirenooxazole 10j in the reaction mixture, it isomerizes via ylide 9j to a much more thermodynamically stable open-chain form 3j.
Curiously, the same energy profile (see Supporting Information File 1, Scheme S1) was obtained for isomerization of azirinium ylide 5 (Scheme 1, R 1 = Me, R 3 = R 4 = H, R 5 = Ph, Alk = Me) which contains one acyl group and one ester group at the carbanion center. From these results it follows that 1,5-cyclization of azirinium ylides is a general stabilization route for ylides derived from diazo compounds containing an α-keto group. However, azirenooxazoles 10 reversibly formed in these reactions could not be detected in the absence of an active electrophilic trap in the reaction mixture. 3-Diazoacetylacetone under Rh(II)-catalysis gives, along with a Rh(II) carbenoid, a highly electrophilic acetyl(methyl)ketene (12) via Wolff rearrangement (Scheme 3).
The nucleophilic attack of the azirenooxazole 10j nitrogen on the ketene sp-carbon followed by cyclization provides two isomeric adducts 6j and 7j. There are two reasonable pathways to these unexpected products. The first pathway involves the addition of bicycle 10j to ketene 12 to form atropoisomeric azirenooxazolium betains 13j,13′j (Scheme 4) which can further undergo aziridine ring opening to give atropoisomeric oxazinium betaines 14j,14′j and their cyclization into adducts 6j and 7j, respectively. Besides, betaines 14j,14′j can result from aziridine ring opening in azirenooxazole 10j into carbonyl ylide 15j and its addition to ketene 12 (Scheme 4).
The energy profiles for both the cycloaddition pathways of 10j to 12 are represented in Figure 3 with the use of relative scale for total energy ΔΔE. According to the calculations, the two attack modes of azirenooxazole 10j to ketene 12 (red and blue lines on the plot) give rise to atropoisomeric azirenooxazolium betains 13j,13′j (not shown in the plot) which undergo a virtually barrierless aziridine ring opening to give oxazinium betaines 14j,14′j. The activation barriers to both the transformation pathways of 10j to 13 are close to each other (10.5 and 11.6 kcal/mol, respectively) but much lower than that to the ring opening of azirenooxazole 10j to azirinium ylide 9j. 1] adduct 6j seems to be more reasonable than 10j→15j→14j due to the much higher concentrations of the reacting species. Actually, the activation barrier to the bicyclic C-N bond cleavage in 10j is very low, but the equilibrium between 10j and 15j is strongly shifted toward bicyclic isomer 10j, and, therefore, carbonyl ylide 15j should be formed in an extremely low concentration.
It is worthy to notice that the distribution of products 6,7 strongly depends on the electronic effects of the para substituents in the aryl group of 3-aryl-2H-azirine 1. It is known that the Rh(II)-catalyzed reaction of 3-aryl-2Hazirines with ethyl diazoacetoacetate (2d) gives rise to 2H-1,4oxazines as a single product [15]. The absence of products like 6 or 7 may be caused by a decreased propensity of the carbenoid derived from 2d to undergo a Wolff rearrangement into a ketene derivative. To obtain experimental evidence for the formation of azirenooxazole intermediates 10 in the reactions of azirines 1 with α-diazo-β-ketoesters, we reacted azirine 1g with a mixture of two diazo compounds, ethyl diazoacetoacetate (2d) and diazoacetylacetone (2c), in the presence of Rh 2 (Oct) 4 (Scheme 5). We suggested that the azirenooxazole formed from the azirine and diazo compound 2d will be trapped by ketene 12 generated from diazoacetylacetone 2c via Wolff rearrangement. The 1 H NMR and TLC analysis of the reaction mixture revealed, along with oxazines 4k, 4h and adduct 7h, one compound. The latter was not detected among the products of the reactions of azirine 1g separately with each of the diazo compounds. This product was isolated by column chromatography on silica gel, and its structure corresponding to the [4.3.1] adduct 7k formed by cycloaddition of ketene 12 to azirenoxazole 10k was assigned on the basis of the 1 H and 13 C NMR and mass spectra. According to the 1 H NMR spectrum of the reaction mixture, the 4k:4h:7k:7h ratio was 16:7:1:2. In the analogous reaction of 3-(p-tolyl)azirine (1e) with a mixture of diazo compounds 2c,d, no other products than [3.2.1] adduct 6l (see Supporting Information File 1, Scheme S1) formed from the ethoxycarbonyl-substituted azirenooxazole derivative and ketene 12 were detected.
Obviously, the bridgehead nitrogen in azirenooxazoles 10 must be sterically accessible for smooth addition of 10 to ketene 12. For example, methyl or phenyl substitution in the one-atom bridge in the related 5-oxa-1-azabicyclo[4.1.0]hept-3-ene system completely suppresses the reactivity of the latter toward acylketenes [12]. This is a possible reason for the formation of neither 6 nor 7 adduct in the reaction of diazoacetylacetone (2c) with 2,3-diphenyl-2H-azirine (1a, Table 1, entry 10). The moderate yield of compound 4i is explained by the formation of an unstable by-product of the transformation of the acetyl group in target oxazine 4i under the reaction conditions. We succeeded in isolating a by-product of this type in the analogous reaction of spiroazirine 1h with diazo compound 2c (Scheme 6). Along with oxazine 16 (57%), small amount of 1,4-oxazine 17 with a modified acetyl group at С5 was isolated by column chromatography. Compound 17 obviously resulted from cycloaddition of acyl ketene 12 derived from 2c under the reaction conditions to the carbonyl group of oxazine 16. Actually, compound 17 formed in the reaction of a pure oxazine 16 with diazo compound 2c in the presence of Rh 2 (OAc) 4 . Its structure was assigned by standard spectral methods and confirmed by X-ray diffraction analysis (Figure 4). It was found that, when kept for a week in a CDCl 3 solution in the dark at room temperature, oxazine 17 undergoes reversible ring opening to form a 1:2.3 equilibrium mixture of 17 and 18. Azadiene 18 was isolated by chromatography and characterized by standard spectral methods.

Supporting Information
Supporting Information File 1 Experimental part, computational details and copies of 1 H and 13 C NMR spectra.