The pentadehydro-Diels–Alder reaction

Abstract

In the classic Diels–Alder (DA) [4+2] cycloaddition reaction¹, the overall degree of unsaturation of the 4π (diene) and 2π (dienophile) pairs of reactants dictates the oxidation state of the newly formed six-membered carbocycle. For example, in the classic DA reaction, butadiene and ethylene combine to produce cyclohexene. More recent developments include variants in which the hydrogen atom count in the reactant pair and in the resulting product is reduced by², for example, four in the tetradehydro-DA (TDDA) and by six in the hexadehydro-DA (HDDA^3,4,5,6,7) reactions. Any oxidation state higher than tetradehydro leads to the production of a reactive intermediate that is more highly oxidized than benzene. This significantly increases the power of the overall process because trapping of the benzyne intermediate^8,9 can be used to increase the structural complexity of the final product in a controllable and versatile manner. In this manuscript, we report an unprecedented net 4π+2π cycloaddition reaction that generates a different, highly reactive intermediate known as an α,3-dehydrotoluene. This species is at the same oxidation state as a benzyne. Like benzynes, α,3-dehydrotoluenes can be captured by various trapping agents to produce structurally diverse products that are complementary to those arising from the HDDA process. We call this new cycloisomerization reaction a pentadehydro-Diels–Alder (PDDA) reaction—a nomenclature chosen for chemical taxonomic rather than mechanistic reasons. In addition to alkynes, nitriles (RC≡N), although non-participants in aza-HDDA reactions, readily function as the 2π-component in PDDA cyclizations to produce, via trapping of the α,3-(5-aza)dehydrotoluene intermediates, pyridine-containing products.

The overall oxidation states of the π-bond-containing pair of reactants in Diels-Alder processes can be viewed as the total amount of 'dehydroness' (cf. refs^10,11) of those species (Fig. 1). This can be identified either from the overall hydrogen atom count or by the number of sp-hybridized carbon atoms that engage to create the newly formed six-membered ring (cf. □ and □ in reactants 4 + 5 and 4 + 8). It occurred to us that a 6π-electron net [4+2] cycloaddition of reactants containing a total of five sp-hybridized carbon atoms—namely, an allenyne + alkyne pair like 11 + 4—might produce an α,3-dehydrotoluene (cf. 12, Fig. 1d). The parent species 12 itself has been characterized by photoelectron spectroscopy in the gas-phase¹², and derivatives of 12, which comprise a little-explored class of reactive intermediate, are generated by the cycloisomerization reaction of allenyl enynes like 14 (Fig. 1e) by the Myers-Saito reaction^{13,14,15,16,17} (Fig. 1e). However, we can find no previous examples of the generation of α,3-dehydrotoluene derivatives by a formal cycloaddition event, the process we term here as the pentadehydro-Diels–Alder (PDDA) reaction. We should emphasize that in choosing this nomenclature, we do not mean to imply anything about the mechanism of this net [4+2] cycloaddition. Specifically, it should not be inferred that transformations like 4 + 11 to 12 are required to proceed by a concerted reaction pathway. Finally, it is interesting to note that α,3-dehydrotoluene 12 is merely a tautomer of 3-methyl-1,2-dehydrobenzene (9). Accordingly, trapping of 12 leads to a toluene derivative that has been newly functionalized at its benzylic position, whereas capture of the tautomeric benzyne 9 gives rise to a newly substituted benzenoid product.

Figure 1. Terminology associated with various cyclizations in the Diels–Alder family of 4π+2π reactions.

a, The classic example of a diene and dienophile to give a cyclohexene. b, The absence of four hydrogen atoms amounts to a tetradehydro (TDDA) variant; the product is at the benzene oxidation state. c, The absence of six hydrogen atoms amounts to the hexadehydro Diels–Alder (HDDA) variant. d, The unprecedented pentadehydro-Diels–Alder (PDDA) reaction proceeds via an α,3-dehydrotoluene (cf. 12); importantly, both the HDDA and PDDA reactions result in formation of trappable reactive intermediates. e, α,3-Dehydrotoluenes have previously been generated principally by cyclization of allenyl enynes like 14.

Our first evidence for a PDDA process came from the reaction of tetrayne 15-Ms in an ambient temperature solution of piperidine. The benzylic amine 18a (Fig. 2a) was the only characterizable product formed in this experiment and was isolated in 81% yield following chromatographic purification. Tetrayne 15-Ms was consumed with a half-life of approximately 15 hours at room temperature. Based on several lines of evidence we have gathered and present below, we believe that the generation of 18a is best described by (i) initial, rate-limiting piperidine-catalyzed isomerization of 15-Ms to produce the allenyne 16, (ii) rapid PDDA cyclization to the dehydrotoluene 17, and (iii) even more rapid trapping by protic piperidine of that reactive intermediate. Base-promoted isomerization of propargyl to allenyl sulfonamides is known.¹⁸

Figure 2. PDDA cascades of tetraynes.

a, The first example: substrate 15-Ms undergoes base-promoted PDDA reaction and in situ trapping by piperidine to provide the adduct 18a; the HDDA cyclization of 15-Ms is slower. b, Examples indicating some of the scope of the PDDA cascades of substrates 15.

This reaction process is quite distinct from the course followed during an HDDA cascade (i.e., sequential benzyne formation and trapping). Formation of a product in which one of the tethering atoms separating the diyne and diynophile has become functionalized has not been observed in any HDDA cascade⁷. We established that, in the absence of base, the tetrayne 15-Ms is a well-behaved HDDA substrate, but only at elevated temperature (half-life of ca. 4 h at 80 °C in C₆D₆). We have trapped the resulting benzyne 19 with methanol^19,20 or acetic acid^5,19 to produce 20a or 20b, respectively (Fig. 2a). Together these observations indicate that formation of the piperidine-trapped product 18a, occurring at a substantially lower temperature, is not derived from the benzyne 19. Thus, isomerization of the tetrayne 15-Ms to the allene tautomer 16 occurs faster than does the thermal HDDA cyclization of 15-Ms.

The non-nucleophilic base DBU mildly accelerates the initial, rate-limiting isomerization of 15-Ms to 16. The rate of formation of adduct 18a was approximately doubled (t_1/2 of ca. 7 vs. 15 h) when five equivalents of DBU were added to the initial piperidine solution of 15-Ms. Other secondary and primary amines participate in this transformation (Fig. 2b, entries 2–5). Other amides in the tether are also compatible (entries 6 and 7), although the reaction is slower with the benzamide 15-Bz.

Oxygen nucleophiles will also trap the intermediate dehydrotoluene derivative 17 (Fig. 2b, entries 8 and 9). When carried out in methanol or aqueous acetonitrile, DBU-promoted PDDA reaction of the methanesulfonamide 15-Ms gave the methoxylated or hydroxylated adducts 18h or 18i, respectively.

We have also achieved PDDA cyclizations with substrates in which the triply bonded acceptor moiety is a cyano rather than an alkynyl group (Fig. 3a). When dissolved in neat piperidine, diynyl nitriles 21 gave rise to the pyridine derivatives 22, thereby establishing the viability of an aza-PDDA reaction. The ability of a cyano group to enter into the PDDA cycloisomerization is particularly noteworthy and decidedly distinct from its inertness to HDDA cyclization—we, nor others²¹, have ever observed nitriles to engage in that process. This is the case for the nitriles 21 as well. In the absence of base, none gave evidence of cyclizing to a pyridyne such as 23 (a 3,4-dehydropyridine), even upon heating to 150 °C in the presence of an excellent HDDA-aryne trap like acetic acid; only extensive decomposition was observed. Thus, the base-promoted tautomerization to 24 and subsequent PDDA cycloisomerization of that allene to the α,3-dehydroazatoluene (or α,3-dehydropicoline) intermediate 25 is considerably more facile than the HDDA cycloisomerization of its precursor tautomer 21. This is entirely in parallel with the observation presented earlier—namely, that the PDDA cyclization of 16 is much faster than the HDDA reaction of 15-Ms (Fig. 2a).

Figure 3. Cyclizations of nitrile-containing diynes: the aza-PDDA.

a, Substrates 21, non-competent reactants in HDDA cyclizations, undergo smooth, base-promoted PDDA reactions to give the piperidine-trapped adducts 22a-c. b, Examples indicating some of the scope of the aza-PDDA cascade.

As with the all-alkyne series already discussed (15), the nitrile substrates 21 are also competent PDDA precursors when bearing a toluenesulfonyl, methanesulfonyl, or benzoyl electron-withdrawing group on the propargylic nitrogen atom (Fig. 3b). Again, amines (entries 1–4), water (entries 5–6), and alcohols (entries 7–11) are all effective trapping nucleophiles.

We have performed several experiments (Fig. 4) that provide support for the proposed mechanism for these transformations. When either of the substrates 15-Ms or 21-Ms was incubated with DBU in a solution of deuterochloroform (Fig. 4a), the major product (26 or 27, respectively) was a dichloroalkene in which the arene carbon corresponding to the sp-hybridized center in the putative α,3-dehydrotoluene intermediate was fully deuterated. Presumably, initially formed trichloromethylated adducts like 28 underwent facile elimination induced by the excess DBU present.

Figure 4. Mechanistic aspects of the PDDA reaction.

a,b, Indirect evidence for intermediacy of α,3-dehydrotoluenes and allenes. c, Relative free energies (G) from DFT calculations [(U)B3LYP−D3BJ/6−311+G(d,p); a solvation model (using Et₂NH) was employed (see SI)] of the directly relevant minima (32, 33, 35, and 37) and two transition state structures [34 (TS1) and 36 (TS2)] for the PDDA cyclization via the diradical intermediate 35. Values beside "a–d" for each of 32–37 are the computed energies (G) in kcal•mol⁻¹. The 2π C≡X component is either an alkyne (a,b) or a nitrile (c,d). (r.d.s. = rate-determining step.) d, An (isolable) allenyne, the phosphine oxide 40, readily cyclizes to the benzenoid product 42.

In contrast to the behavior of the gem-dimethylated substrates 21 (cf. Fig. 3), the analogous substrate 29 (Fig. 4b) lacking those methyl groups gave a different outcome when held in a piperidine solution at room temperature. No more than a trace amount of the expected PDDA cyclization product was observed. Instead, the enamine 31 was isolated as the principal product formed in this experiment. We presume that this arises by addition of the amine to the central carbon in allene 30 (to give a delocalized allylic/propargylic anion) rather than by direct hydroamination of the starting diyne 29; we have performed a number of HDDA reactions on diyne substrates, not activated for tautomerization, in the presence of secondary amine trapping agents and never observed amination of a conjugated diyne. The change in reaction course between 21 vis-à-vis 29 can be explained by the Thorpe-Ingold effect;²² the lack of the geminal substituents results in a widening of the bond angle and increase in distance (r) between the unsaturated centers in 30. This significantly slows the rate of the PDDA cyclization, permitting time for the piperidine to intercept the allene. We have observed a similar phenomenon in the rate of HDDA cyclization of an analogous pair of triyne substrates²³ and also have found computational support for this interpretation (see below, Fig. 4c).

We have carried out density functional theory (DFT) computations to gain additional mechanistic perspective about the PDDA cycloisomerization. The energy diagram in Fig. 4c shows the relative free energies for four series (a–d) of PDDA reactions of allenynes 33a–d; members of the a/b series have an alkyne as the pendant 2π-component and the c/d a nitrile. The diagram has been normalized so that each of the four PDDA reactant allenes 33 has the same free energy. The cyclizations proceed to the α,3-dehydro(aza)toluenes 37 by way of diradical intermediates 35. Houk and coworkers have recently described a DFT analysis of (bimolecular) HDDA reactions and found that the diradical pathway is more favored than the concerted.²⁴ We have observed the same for a series of intramolecular HDDA reactions and found that the (U)B3LYP-D3BJ functional did an excellent job of correlating with our experimentally observed rates.²⁵ We have not been successful locating transition state structures corresponding to concerted PDDA reactions for 33a–d at this level of theory.

Some notable points from these calculations are: (i) the free energy differences between the 1,3-diynes 32 and tautomeric allenynes 33 are small, which serves as a reminder that the potential energies of the participating functional groups in 33 are also high, comparable to those in 32, (ii) as with triyne to benzyne conversions, the overall energies of reaction from 33 to the reactive α,3-dehydrotoluenes 37a/b are exergonic (by ΔG° of ca. 35 kcal•mol⁻¹), although not to as large an extent as those computed for HDDA cyclizations to benzynes (ca. −50 kcal•mol⁻¹)^26,5, (iii) the corresponding energy differences between the nitrile-containing allenynes 33c/d vis-à-vis the product α,3-dehydroazatoluenes 37c/d is even smaller, reflecting the inherently lower potential energy of a nitrile triple bond vs. that of an alkyne, (iv) the magnitude of the computed ΔG^‡s for the first (and slower) step in the PDDA cyclization [cf. 34 (TS1), Fig. 4c] are not inconsistent with the fact that our PDDA cyclizations are proceeding rapidly at less than near-ambient temperatures, and (v) the difference in the computed ΔG^‡ for the first step in the PDDA reaction of the nitrile 33c vs. that of 33d (23.5 vs. 18.0) is consistent with the differing behavior of allenyne 24 (Fig. 3a) vs. the aza-analog 30 (Fig. 4b). The former, recall, underwent smooth PDDA cyclization to 25 en route to the piperidine-trapped adduct 22b (Fig. 3a), whereas the latter cyclized so slowly that interception by piperidine interceded to produce the enamine 31. The optimized geometries computed for the reactive conformers of 33c vs. 33d (lacking vs. having gem-dimethyl groups²²) differ substantially (0.3 Å) in the distance (r) between the nitrile and central allene carbons, indicating that the former is better poised for surmounting the TS1 activation barrier.

Finally, we devised an experiment to unambiguously demonstrate the intermediacy of an allenyne (cf. 40, Fig. 4d). The hydroxytetrayne 38 reacted readily with chlorodiphenylphosphine to produce the transient phosphinite 39, which smoothly rearranged^27,14 at sub-ambient temperature to the allenyldiphenylphosphine oxide 40. This labile compound was observed to degrade upon handling at room temperature, but could be rapidly purified and characterized. Dissolving 40 in methanol or water/acetonitrile spontaneously gave rise to the corresponding trapped product 42a or 42b, respectively. We view this as strong support for the intermediacy of 41 and the PDDA mechanism posited here throughout. That these reactions proceed smoothly in the absence of base also argues against a mechanism initiated by nucleophilic attack on the allene terminal carbon of the transient intermediates 16 (Fig. 2a) or 24 (Fig. 3a). The known reaction of methanol with Myers-Saito reaction-derived α,3-dehydrotoluenes in similar fashion is also relevant^28,29,30.

In summary, we have described a new class of reaction—the formal [4+2] cycloaddition of an allenyne with a pendant alkyne (or nitrile) to produce an α,3-dehydro(aza)toluene derivative. We call this a pentadehydro-Diels–Alder (or PDDA) reaction. In the majority of examples reported, the base-promoted isomerization of a precursor N-1,3,-diynyl sulfonamide to the reactive allenyl tautomer is overall rate-limiting. The PDDA then proceeds rapidly, far more so than the HDDA cyclization of the precursor 1,3-diyne. We also have shown that the PDDA-derived dehydrotoluene, itself a reactive intermediate, can be trapped by a variety of N-, O-, and C-centered nucleophiles (Figs. 2b and 3b). In one instance the conjugated allenyne has been isolated (40, Fig. 4d) and its facile cyclization and in situ trapping observed. Finally, nitriles, which do not participate in HDDA reactions, now enter the realm of reactivity, resulting in pyridine-containing products (Fig. 3).