Divergent ynamide reactivity in the presence of azides – an experimental and computational study

An unusually divergent reactivity of ynamides in the presence of azides is reported.


Introduction
Ynamides are well established as versatile reagents in organic synthesis, as the unique polarization of their triple bond continues to enable broad synthetic applications. 1 In this context, ynamides thrive as building blocks for the synthesis of complex organic frameworks (which would be difficult to achieve otherwise) and biologically relevant heterocycles. Somewhat surprisingly, most of the contemporary efforts on ynamide chemistry appear to focus on their activation by carbophilic transition metals. Herein, we report the unusually divergent reactivity of ynamides in the presence of azides under promotion by a simple Brønsted acid (Scheme 1, bottom).
With the notable exceptions of Ye's, Liu's and Huang's recent elegant works on the gold-catalyzed syntheses of indole-and pyrrole-derivatives, 2 the activation of ynamides with azides has hitherto been comprehensively utilized for the synthesis of triazoles through (3 + 2) cycloaddition (Scheme 1, top). 3 Our work documents a novel reactivity pattern of azide/ynamide systems leading to the divergent and controlled formation of either oxazolidine-2,4-diones or b-enaminoamides (Scheme 1, bottom). In more detail, the rst part of our work describes the formation of oxazolidine-2,4-diones via a complex azide-triggered rearrangement. These scaffolds have rich biological activity 4 and can be employed as monomers for the synthesis of side-chain modied (SCM) polypeptides. 5 In the second part, a change of azide partner strikingly promoted formation of radically different, quasi-dimerization products. The dimerization of alkynes in general is rare in the absence of metal promoters, 6 and the only previously reported dimerization of ynamides, by Skrydstrup, 7 involves the use of a gold catalyst. Mechanistic studies and extensive density functional theory (DFT) calculations shed light on the details of these unique examples of divergent reaction pathways.

Results and discussion
From the outset, attempts at combining phenyl-substituted ynamide 1a and the azide 2a under acidic conditions unexpectedly led to the rearranged product 3a, the structure of which Scheme 1 Prior work on the reactivity of ynamides and azides (top) and divergent reactivity under Brønsted acid catalysis (bottom).
was elucidated by X-ray analysis (Scheme 2). Additional screening revealed dichloromethane (DCM) to be the best solvent and TfOH to be necessary to promote the reaction (details of optimization experiments are compiled in the ESI †). This enabled the preparation of 3a in 78% isolated yield 8 (X-ray structure in Scheme 2).
Given the unusual skeletal reorganization involved in the coupling of 1a and 2a into 3a (newly formed bonds highlighted in bold typeset in Scheme 2), along with the known biological relevance of oxazolidine-2,4-diones, 9 we were eager to probe the generality of this transformation and decided to investigate the use of diverse ynamide and azide partners.
The results are depicted in Scheme 3. Aryl ynamides containing halogen substituents or electron-donating/withdrawing moieties on the aromatic ring led to the skeletal rearrangement products in generally good yields. The reaction was also performed with ynamides carrying a non-aromatic substituent with moderate yields (cf. 3l, Scheme 3). A diastereoselective variant was then investigated and the use of (S)-4-isopropyl-2-oxazolidinone as a chiral auxiliary led to product 3k, formed with a d.r. of 85 : 15. Azides bearing uorine and methyl substitutions also delivered the expected products in good yield (3m and 3n). The use of an aliphatic azide led to the rearranged product in lower yield (3o).
During these investigations, we decided to look more in detail at the fate of the azide reaction partner. Considering that the efficiency of the reactions in Scheme 3 may be compromised by degradation of the azide under acidic conditions, 10 we turned to the use of diphenylphosphoryl azide (dppa) (4) as a potentially more stable species. 11 To our surprise, the use of dppa (4) under essentially the same conditions as those portrayed in Scheme 3 resulted in the exclusive formation of a new product, which was not the expected 3p. Structural elucidation revealed this to be a dimer 5a resulting from the union of two ynamide reactant molecules (Scheme 4). As mentioned previously, the direct dimerization of alkynes is almost exclusively the domain of transition-metal chemistry and the formation of a dimerization product under simple acidic conditions piqued our interest. Strikingly, the presence of dppa (4) was absolutely crucial for the formation of 5a: in its absence, only the simple hydrolysis product of ynamide 1a was detected, with no traces of any dimer. 13 Further optimization enables the generation of this ynamide dimer in 81% yield in the presence of only substoichiometric amounts of both TfOH and dppa (4) (Scheme 5). 14 With these new conditions in hand, we have been able to obtain the products of hydrative dimerization of differently substituted aryl ynamides. As shown in Scheme 5, several ynamides carrying either an aryl or alkyl side chain afforded good to excellent yields of the dimerization product. The use of an aryl-ynamide carrying a strongly electron-withdrawing CF 3 moiety led, however, to a lower yield (cf. 5e). Importantly, sulfonylynamides also enable dimerization to take place (cf. 5h).

Mechanistic studies
In order to understand the factors determining the different product outcomes observed experimentally and to gain insight Scheme 2 Unexpected coupling of 1a with 2a under acidic conditions to generate the oxazolidine-2,4-dione 3a (X-ray structure shown; newly formed bonds highlighted in bold).
into the mechanisms and energetics of these two divergent pathways, we turned to DFT calculations. Oxazolidinone 1a served as the model substrate for all investigated processes.

Computational methods
All DFT calculations were performed using the B3LYP hybrid functional 15 with the D3 dispersion correction of Grimme 16 and the polarized double-z basis set 6-31G*, 17 as implemented in the Gaussian09 program package. 18 The effect of solvation was taken into account implicitly by adding a polarizable continuum model, with dichloromethane as solvent (dielectric constant 8.93). 19 Geometry optimizations were carried out without any constraints. Ground state minima and transition states were conrmed by frequency calculations, yielding none and one imaginary frequency, respectively. The connectivity of the transition state structures was veried by intrinsic reaction coordinate calculations. 20 Electronic energies obtained were converted to relative enthalpies DH 0 and free energies DG 0 at 298.15 K and 1 atm by using zero-point energy and thermal energy unscaled corrections obtained in the frequency calculation. All energies are reported in kcal mol À1 and all distances are reported inÅngstrom (Å).

Preliminary DFT studies
It is well recognized that the treatment of ynamides with TfOH leads to the formation of transient keteniminium triates A (Scheme 6). 21 These reactive species rapidly collapsein the absence of any nucleophilic reaction partnerto the energetically more stable N,O-ketene aminals 6. 22 Although any attempt to characterize these reactive intermediates using spectroscopic methods was met with failure, 22,23 we and others have disclosed chemical transformations involving the discrete intermediacy of keteniminium triates. 22,23a,24 This, however, raises the question as to the true nature of the stability and reactivity of these transient keteniminium species Ain particular in the presence of additional reagents.
Intrigued by the fundamental question on the stability of keteniminiums A, and in order to better understand the exact nature of the interaction between reagents/additives and that species, we decided to begin our computational studies by simply modelling the generation of aminal 6a (R 1 ¼ Ph, N(R 2 ) EWG ¼ oxazolidinone) from ynamide 1a in the absence of any additional reaction partners, such as (2-azidoethyl)benzene (2a) or dppa (4). As shown in Fig. 1, aer initial association of TfOH and ynamide 1a, which is highly exothermic in energy (DDH 0 ¼ À17.4 kcal mol À1 ), protonation of the ynamide in complex B occurs at the b-carbon providing keteniminium triate A 0 (TS B-A 0 ). Iminium A 0 was found to collapse without any signicant activation barrier . This result is in accordance with the spectroscopic studies carried out by Thibaudeau and Evano. 22 Moreover, the low barrier of activation of the triateaddition to keteniminium A 0 explains why the experimental characterization of any intermediate keteniminium is highly challenging. Overall, the conversion of ynamide 1a to aminal 6a is both highly exothermic (DDH 0 ¼ À21.7 kcal mol À1 ) and exergonic (DDG 0 ¼ À37.5 kcal mol À1 ) in energy.
DFT results for oxazolidine-2,4-dione formation and experimental support Next, we embarked on DFT calculations for the reaction of ynamide 1a, TfOH and (2-azidoethyl)benzene (2a). At the outset of the mechanistic theoretical investigations on the formation of oxazolidine-2,4-dione 3a, we evaluated the competitive Scheme 5 Scope of the dppa-promoted ynamide dimerization. Reagents and conditions: 1 (0.6 mmol), 4 (0.3 mmol) and TfOH (0.3 mmol) in DCM (1.5 mL). Yields refer to isolated products. The related enthalpy prole for the proposed underlying mechanism is presented in Fig. 2. In accordance with the preliminary studies, pre-association of all three reagents involved (ynamide 1a + TfOH + azide 2a ¼ complex E) is accompanied by a large gain in energy (DDH 0 ¼ À28.4 kcal mol À1 ). Following migration of TfOH to the electron-rich alkynyl moiety, protonation at the b-carbon ensues, leading to keteniminium triate F (TS E-F ) in a slightly endothermic process (DDH 0 ¼ +1.5 kcal mol À1 ). The obtained iminium F can then follow two competitive pathways: (1) triate addition generating the thermodynamically more stable N,O-ketene acetal D (TS F-D ) or (2) azide addition forming N,N-ketene acetal C (TS F-C ), the kinetic product. Three points are worth mentioning: rst, in light of the low barriers of activation for both processes (DDH 0 # 2.7 kcal mol À1 ), keteniminium F can be considered to be a rather short-lived intermediate. Second, in comparison with the isolated system (Scheme 6, Fig. 1), the activation barrier for triate addition is energetically more demanding, albeit the difference is small DDH 0 ¼ +2.3 kcal mol À1 . 25 Third, the low difference in activation energies (TS F-D vs. TS F-C ) would suggest that both ketene acetals C and D are present as discrete intermediates towards the generation of oxazolidine-2,4-dione 3a.
The nal steps en route to 3a are depicted in Fig. 3. Although calculations have indicated that N,O-ketene acetal D is favored over N,N-ketene acetal C by DDG 0 ¼ 6.8 kcal mol À1 , DFT-studies have attributed further viable reaction pathways only to acetal product C. 26 In light of the energetically accessible transition state TS F-D in the reverse direction (D / F), it appears reasonable to assume that upon consumption of available C, successive displacement of the equilibrium D # F # C would ensure full conversion. As shown in Fig. 3, once N,N-ketene acetal C is formed, it can undergo spontaneous loss of dinitrogen. Interestingly, computations have revealed that the initial orientation of the oxazolidinone moiety concurrent to the dinitrogen expulsion dictates the presence (TS C-G ) or the absence (TS C-H ) of a stable cationic intermediate G en route to oxazolium triate H (Fig. 4). Thus, if the carbonyl oxygen, O(6), points towards the emerging benzylic cationic system (TS C-H ) the resulting transient cation G 0 (not shown) collapses in a second, yet barrierless step into bicycle H. 27 Conversely, if the carbonyl oxygen and the benzylic cation adopt a quasi-antiperiplanar conformation (F(C]C(1)-N(2)-O(6)) ¼ +145, Fig. 4) during the nitrogen-extrusion, imine G is formed as a stable intermediate. The identied torsional transition state TS G-H connecting imine G and oxazolium H is characterized by the counter-clockwise rotation of the oxazolidinone moiety with respect to the axis as dened by the N(2)-C(1) bond. 28 Notably, since the energetics of these two described nitrogen-extruding transition states are very close, we anticipate that both pathways are operative for aryl-bearing ynamides, such as 1a. Further characteristic features of these key transition states are (Fig. 4): (1) the rather long incipient N(4)-N(5) and C(1)-N(3) bond lengths, as well as the rather short incipient N(3)-N(4) bond length, indicative of early transition states. (2) The assistance of the double bond (p(C]C(1))) in the rupture of the N(3)-N(4) bond, as judged by both NBO and MO analyses. Key to this observation is the existence of a distinct orbital in the highest occupied molecular orbital (HOMO) of both transition states (TS C-H and TS C-G ) connecting all three bonding atoms of the enamine-like system (C]C(1)-N(3)). 29 Associated to this nding is (3) the stereodened nitrogen extrusion, servicing the stereoselective formation of (E)-imines G and G 0 , a necessary prerequisite for the further pathway (vide infra). In accordance with the above statements made, systems hypothetically leading  to (Z)-imines lack the crucial orbital overlap between p(C]C(1)) and s*(N(3)-N(4)), thus rendering these pathways highly unlikely. 30 As shown in Fig. 3, the proposed onward reaction of oxazolium triate H to 3a involves two successive S N 2 reactions (H / I and I / J). The rst of these involves nucleophilic opening of the oxazolium moiety (TS H-I ) giving rise to primary triate I. A structurally different bicycle is then reformed by intramolecular ring closure via substitution of the primary triate by the amidine functionality in I (TS I-J ) generating the nal intermediate, amidinium triate J. Both transition-state geometries (TS H-I and TS I-J ) nicely reect the second-order character of the nucleophilic substitutions typical for primary alkyl reaction centers.
The triate-mediated rearrangement of oxazolium H into amidinium J represents a rather unusual, yet intriguing process. In light of the feasible (DDG 0 (TS H-I ) ¼ +17.6 kcal mol À1 ) and favorable (DDG 0 (H / I) ¼ À10.5 kcal mol À1 ) computed energetics of the ring-opening reaction of oxazolium H, our proposed mechanism marks one of the few cases of nucleophilic assistance of triates in chemical transformations. Straightforward aqueous basic hydrolysis of amidinium triate J during work-up gives rise to oxazolidine-2,4-dione 3a.
The overall free energy of the conversion of association complex E (¼1a + 2a + TfOH) into imidazolium triate J, at À109.9 kcal mol À1 , is evidently high, with the loss of dinitrogen (C / G/G 0 / H: DDG 0 ¼ À75.6 kcal mol À1 ) representing the main driving force of the reaction. Since all of the individual barriers towards the formation of oxazolium H are fairly small, the reaction is believed to proceed smoothly up to this point; the  Scheme 8 Interception of amidinium J by reduction (X-ray structure of 7a depicted). 31 subsequent rearrangement of the bicycle, however, contains rather high barriers of activation (DDG 0 ¼ +17.6 and +19.4 kcal mol À1 , respectively). The triate-mediated formation of amidinium J therefore represents the rate-limiting process. Since the amidinium hydrolysis only occurs upon addition of water, we anticipated that the nal intermediate J could also serve as a valuable platform for other synthetic transformations. In the event (Scheme 8), treatment of the reaction mixture obtained by combining 1a and 2a in the conditions previously described aer 24 hours with NaBH 4 , afforded a new compound. Pleasingly, X-ray crystallographic analysis showed this to be the expected N,N-acetal 7a, 31 formed as a single trans-diastereoisomer, thus lending strong support to the intermediacy of J prior to hydrolysis.

DFT results for hydrative dimerization and experimental validation
Having established plausible mechanisms for the conversion of ynamides 1 into N,O-ketene aminals 6, as well as for the reaction of alkyl azides 2 with ynamides 1, we next investigated the effect of dppa (4) in leading to the spontaneous dimerization of ynamides, such as 1a. In alignment with the studies presented above, we initially considered the formation of N,O-and N,N-ketene acetals K and L expecting to gain valuable insights into the exact nature of the interaction between dppa (4) and the transient keteniminium triate M. The computational results on the proposed mechanistic pathway, along with the schematic representation of the initial competitive pathways are delineated in Fig. 5. As shown before, pre-association of the reaction partners, ynamide 1a, dppa (4) and TfOH, proceeds with a large gain in energy (DDH 0 ¼ À33.0 kcal mol À1 ) forming complex N. Subsequent protonation at the ynamide b-carbon (TS N-M ) then affords keteniminium M. Interestingly, both the activation barrier for the protonation, as well as the energy of the transient keteniminium are markedly reduced compared to the isolated system ( Fig. 1) and the reaction of ynamide 1a with azide 2a (Fig. 2). This can be ascribed to favorable p-p stacking interactions between the pendant phenolic arm of dppa (4) and the styrene-like system in the transition state geometry TS N-M , as well as the intermediate keteniminium M. The same holds  true for the subsequent triate addition step (TS M-K ) leading to N,O-ketene acetal K. As a result, the calculated enthalpic prole for the conversion of keteniminium triate M into dppa-associated acetal K closely resembles the energetic footprint of the isolated system. While this observation might appearat rst glanceinconsistent with the experimental observations, the exceptional ability of dppa (4) to interact with the styrene-motif in an energetically favorable mode, as observed in TS N-M / K, also proves to be vital for the occurring dimerization process in the case of aryl-bearing ynamides, such as 1a (vide infra). It should be mentioned that we also investigated other processes including the addition of dppa (4) to the keteniminium species M in two ways: (1) azide-addition of dppa (TS M-L ) affording oxazolidinone L and (2) addition of the phosphoryl oxygen in dppa (4) (TS M-O ) leading to oxaphosphonium O. However, computations have indicated that both pathways are kinetically and thermodynamically disfavored with respect to the formation of N,O-ketene acetal K, and are therefore considered to be non-vital to the global picture.
The remaining steps in forming the dimer 5a mandate the introduction of a second molecule of ynamide 1a. As shown in Fig. 6, the combination of keteniminium M and an additional ynamide 1a results in the exothermic formation (DDH 0 ¼ À11.9 kcal mol À1 ) of association complex P. Computations point out that the energetically most favorable spatial arrangement of all four components (dppa, keteniminium, triate and ynamide) prefers the pseudo-substitution of the triate counterion in M by ynamide 1a. As a result, association complex P is decisively limited in its reactivity, as once an unreacted ynamide coordinates to a generated keteniminium species, any reaction pathway building up on the initial triate-addition (as shown in Fig. 5) is shut down.
The generation of experimentally observed product 5a hinges on the synthetic combination of an unreacted ynamide with a keteniminium species. Computations reveal that, in the presence of dppa (4), the barrier of activation for such a decisive transformative step (TS P-Q ) is remarkably low (DDH 0 ¼ +0.8 kcal mol À1 ; DDG 0 ¼ +2.8 kcal mol À1 ). In detail, following migration and rotation of ynamide 1a in complex P towards alignment of the b-carbon of 1a and the positively charged keteniminium carbon (NBO charge ¼ +0.52), a new C-C bond can arise forming keteniminium triate Q. The addition P / Q is highly exothermic (DDH 0 ¼ À30.8 kcal mol À1 ) and exergonic (DDG 0 ¼ À26.9 kcal mol À1 ) in energy. As stated earlier, we consider the favorable p-p stacking interaction between the pendant phenolarm of dppa (4) and the styrene-like motif in TS P-Q to be decisive for the occurrence of the hydrative dimerization of aryl-bearing ynamides, such as 1a. 32 This hypothesis is also corroborated by the experimental observation that both triphenylphosphate (P(O)(OPh) 3 ) and ethoxyacetylene, both of which entail a Lewisbasic site as well as electron-rich p-systems, can also promote formation of the dimeric product 5a. 16 Triate-addition (TS Q-R ) generates the nal intermediate, N,O-ketene acetal R. Once more, straightforward aqueous workup furnishes the hydrative dimer 5a. Importantly, we once more sought experimental evidence of the presence of R prior to hydrolysis. In the event, we found that adding acetonitrile to the reaction mixture of dimerization leads to the interesting formation of the congested, fully-substituted pyridine 8 in modest unoptimized yield (Scheme 9). This remarkable process likely involves capture of R by the nitrile nitrogen followed by cyclisation and aromatization, forming two new C-C bonds and one C-N bond overall.
In summary, the hydrative dimerization of ynamides is represented by the combination of the energy proles in Fig. 5 and 6. The formation of N,O-ketene acetal R requires the overcoming of a maximum free energy barrier of only +4.2 kcal mol À1 (corresponding to the nal transition state TS Q-R ). Along with a free energy of reaction (N + 1a / R) of À53.1 kcal mol À1 the process should occur rapidly under the reaction conditions employed.

Conclusions
In conclusion, we have documented an unusually divergent reactivity mode of ynamides in the presence of azides to provide either b-enaminoamides or biologically important oxazolidine-2,4-diones in a concise and highly selective manner. The formation of either product is fully controlled by the azide partner under the inuence of a simple Brønsted acid promoter. The presented work therewith displays a rare example of a Brønsted acid-mediated, divergent reaction that leads to radically different products. Extensive DFT calculations combined with designed mechanistic experiments provided detailed insights into the underlying mechanistic pathways. The mechanistic investigations also highlighted the crucial factors key to each pathway and presented important information about the general stability and reactivity of transient keteniminium species in the presence or absence of additional reagents. Given the fundamental importance of ynamides in organic synthesis, our mechanistic insights may well evoke the development of novel, conceptually intriguing methods involving keteniminium species. 34