Origins of Endo Selectivity in Diels–Alder Reactions of Cyclic Allene Dienophiles

Abstract

Strained cyclic allenes, first discovered in 1966 by Wittig and coworkers, have recently emerged as valuable synthetic building blocks. Previous experimental investigations, and computations reported here, demonstrate that the Diels–Alder reactions of furans and pyrroles to 1,2-cyclohexadiene and oxa and aza heterocyclic analogs proceed with endo selectivity. This unprecedented endo selectivity gives the adduct with the allylic saturated carbon of the cyclic allene endo to the diene carbons. The selectivity is very general and useful in synthetic applications. Our computational study establishes the origins of this endo selectivity. We analyze the helical frontier molecular orbitals of strained cyclic allenes and show how secondary orbital and electrostatic effects influence stereoselectivity. The LUMO of carbon-3 of the allene (C-3 is not involved in primary orbital interactions) interacts in a stabilizing fashion with the HOMO of the diene in such a way that the carbon of the cyclic allene attached to C-1 favors the endo position in the transition state. The furan LUMO, allene HOMO interaction reinforces this preference. These mechanistic studies are expected to prompt the further use of long-avoided strained cyclic allenes in chemical synthesis.

Graphical Abstract

A study of the factors contributing to endo selectivity in Diels–Alder reactions with strained cyclic allenes is reported. The contributions of a new type of secondary orbital interaction resulting from the twisted nature of the strained allene, and electrostatic effects, to endo selectivity are established.

Introduction

Strained cyclic allenes are valuable building blocks for the construction of complex molecular scaffolds.^{[1, 2]} A recent application by the Schreiber group to create a DNA-encoded library synthesis exemplifies their utility.^[2a] A diverse range of stereochemically-rich products can be accessed via highly endo selective Diels–Alder cycloadditions to strained cyclic allenes, as shown by cycloadducts 1–4 (Figure 1a).^[2b-2e] Our groups^[2c,d,f-i] and others^[2j,k] have generated strained cyclic allenes from silyl triflates, 5, and have found that in situ generated heterocyclic allenes, such as unsubstituted allenes 7 and 8, participate in highly endo selective Diels–Alder reactions to provide adducts 13–22 (Figure 1b).^[2c,d]

Figure 1.

a) Adducts accessible via highly endo selective Diels–Alder reactions of strained cyclic allenes. b) Examples of highly endo selective Diels-–Alder reactions with strained heterocyclic allenes. c) The endo selective Diels–Alder reaction of cyclohexa-1,2-diene (25) with furan (24). R⁴ = carboxybenzyl.

The simplest example of this endo selectivity is the cycloaddition of cyclohexa-1,2-diene (25) with furan (24) (Figure 1c).^[2k] The reaction yields cycloadduct 26 in 80% yield with a 11:1 endo:exo ratio. Notably, there are no obvious secondary orbital interactions with unsaturated substituents attached to the reacting alkene, the usual feature that distinguishes an endo approach of a diene to a dienophile with extended unsaturation. Ordinarily, the preference for the endo Diels–Alder reaction arises from secondary interactions involving unsaturated substituents on the dienophile double bond, often polar activating groups such as an ester. Endo selectivity here refers to a CH₂ group of the cyclohexadiene (Figure 1c) oriented cis to the diene, and this lacks any unsaturation. The only analogous endo selectivity of a saturated group is the reactivity of dienes with cyclopropene, that exhibits high endo selectivity.^[3] That case is known to arise from the very large hyperconjugation of the CH₂ of cyclopropene, but we show here that the explanation for cyclic allenes is entirely different.

Previous computational studies of endo and exo selectivities in Diels–Alder reactions involving strained dienophiles such as cyclopropenes, methylenecyclopropane, and derivatives thereof implicate the role of secondary orbital, electrostatic, and steric factors in exo/endo stereoselectivity.^{[3, 4]} Acyclic allenes can participate in endo selective Diels–Alder reactions as well, but this selectivity is only achieved in cases where an ester substituent is present and a Lewis acid is used.^[5] These studies cannot be directly applied to endo selectivity in reactions with cyclohexa-1,2-diene (25), which lacks such unsaturated substituents.

When an ester is attached to a reactive cyclic allene double bond, the ester is exo in products, while the saturated ring CH₂ is endo (see 19–22 above). A recent report by West provides many examples of endo cycloadditions of this type, in which unsaturated groups, CN and COR, are forced into an exo position.^[2l] Previous computational investigations of Diels–Alder reactions of cyclic allenes^[8] have not discussed endo selectivity, while West describes electrostatic interactions between the oxygen of furan (24) and a cyano group on the allene that stabilize the exo-CN transition, endo-cyclohexene transition state. This kind of interaction is not known to operate in normal Diels-Alder reactions of acrylonitrile. While we discuss such cases, our major goal was to understand why the CH₂ group invariably is endo in the favored transition state (Figure 1b).

We previously studied the reaction mechanism for the Diels–Alder reaction of cyclohexa-1,2-diene (25) with furan (24) at the (U)B3LYP/6-31G(d) level of theory.^[2n] At that computational level, we proposed that the reaction proceeds via a stepwise, diradical mechanism with favored formation of the endo adduct, but did not explore the origins of the endo selectivity. Our more recent investigation on the Diels–Alder reaction of substituted azacyclohexa-3,4-dienes with the more accurate functional, ωB97X-D,^[6] shows that these reactions occur instead by a highly asynchronous, but concerted, Diels–Alder mechanism.^[2c] These calculations were consistent with experimentally observed endo selectivities, but no attempts were made at that time to analyze the factors leading to this endo selectivity.

Results and Discussion

Molecular Orbital Analysis of Cyclohexa-1,2-diene (25)

Previous computational studies have provided an examination of the electronic structure of cyclic allenes and, in some cases, a comparison to that of linear allenes.⁷ Bickelhaupt, Hamlin, et al. have extensively analyzed the regioselectivity of 1,3-dipolar cycloadditions of linear and bent substituted allenes and concluded that orbital interactions control reactivity and regioselectivity.^7f We provide an orbital interaction analysis of a related, but entirely different phenomenon here, the stereoselectivity of Diels-Alder cycloadditions of cyclic allenes. We begin by describing the π MOs of linear and cyclic allenes and the interactions of diene orbitals with these allene π orbitals to explain the origins of stereoselectivity in their Diels–Alder reactions.

We first analyzed the electronic structure of the linear allene 2,3-pentadiene (27) and compared it to that of cyclohexa-1,2-diene (25). Computations were performed using Gaussian 16. Geometry optimizations were carried out using the ωB97X-D functional^[6] and the 6-311+G(d,p) basis set. Energy minima were verified through vibrational analysis. Molecular orbitals and orbital energies were calculated using HF/6-31G(d). Similar conclusions can be made based upon Kohn-Sham orbitals at the ωB97X-D/6-31G(d) level. Molecular orbitals are illustrated using PyMol 2.4.0. Figure 2a shows the computed frontier molecular orbitals (FMOs) of 2,3-pentadiene (27). The degenerate HOMOs and LUMOs of 27 have helical topologies,^[8] and can be represented by the combination of localized orthogonal π and π* orbitals, respectively. The HOMO–a and LUMO–a of 27 are left-handed helices, whereas the HOMO–b and LUMO–b are right-handed helices. It has been proposed that a slight split in degeneracy occurs in 2,3-pentadiene (27) due to the reduction of symmetry from the D_2d symmetry of the allene.^[8a]

Figure 2.

a) Localized and b) Delocalized (canonical) FMOs of 2,3-pentadiene (27). MO structures and energies obtained with HF/6-31G(d). MO energies at the ωB97X-D/6-31G(d) level are provided in parentheses.

A significant degree of distortion is required to constrain the allene into a six-membered ring such that the aforementioned representation of the MOs as involving only orthogonal p orbitals is no longer suitable for understanding the MOs of cyclic allene 25 and their involvement in [4+2] cycloadditions. As a result of twisting and bending the allene (i.e., altering the bond angle formed by the three allenic carbon atoms from 180° to 133°), the FMOs of cyclohexa-1,2-diene (25) also become significantly non-degenerate (Figure 3). The energies of the LUMO (2.9 eV) and LUMO+1 (5.9 eV) differ by 2.9 eV whereas the HOMO (−8.5 eV) and HOMO–1 (−10.0 eV) are non-degenerate by 1.5 eV. In addition, allene bending results in increased s-character at C2 of the LUMO of 25.

Figure 3.

a) Localized, and b) Delocalized (canonical) FMOs of cyclohexa-1,2-diene (25). The MO coefficients of the LUMO of 25 are shown in red. MO structures and energies obtained with HF/6-31G(d). MO energies at the ωB97X-D/6-31G(d) level are provided in parentheses.

Figure 3 shows how the FMOs can be constructed from two distorted and twisted localized π and π* orbitals of two ethylenes in an allene. The two bonding MOs, prior to mixing, are depicted as π₁ and π₂. Bending at the allene central carbon gives s orbital admixture with the p orbital (towards sp²). Combination of the p orbitals at the central carbon provides tilted p orbitals that overlap with the terminal hybrid orbitals. Additive and subtractive combinations of the bonding MOs π₁ and π₂ provide the HOMO and HOMO–1 of cyclohexa-1,2-diene (25), respectively. Because they interact, they are stabilized and destabilized and no longer remain degenerate. The antibonding MOs of cyclohexa-1,2-diene (25) can be described similarly. Prior to mixing, the antibonding MOs can be represented as π₁* and π₂*. Additive and subtractive combinations of π₁* and π₂* provides the LUMO and LUMO+1 of cyclohexa-1,2-diene (25), respectively. The LUMO of 25 has a larger MO coefficient at C2 and, as such, Diels–Alder reactions with electron-rich dienes (e.g. furan) are expected to occur with preferential bonding at C2 of cyclohexa-1,2-diene (25).

Transition States of the [4+2] Cycloaddition Reaction of Cyclohexa-1,2-diene (25) with Furan (24), and Analysis of Exo/Endo Differences

The Gibbs energy barriers for the reaction of 25 with furan (24) to provide endo and exo adducts 26-endo and 26-exo, respectively, were calculated and are shown in Figure 4a. Calculations were performed as above. Solvation effects of tetrahydrofuran (THF) or acetonitrile (MeCN) were included using the solvation model SMD^[9] with a standard state of 1 M. Energy minima and transition states were verified through vibrational analysis. Truhlar’s quasiharmonic correction was applied by setting all positive frequencies below 100 cm⁻¹ to 100 cm⁻¹.^[10] Hirshfeld charges and electrostatic potential maps were obtained using ωB97X-D/6-311+G(d,p). Optimized structures and are illustrated using CYLview.^[11] Formation of endo adduct 26 is predicted to occur with an energy barrier of 19.4 kcal mol⁻¹. The exo transition state is 1.7 kcal mol⁻¹ higher in energy than the endo transition state, which correlates well with experimental results (11:1 endo:exo). The exo and endo transition structures TS-1-exo and TS-2-endo, respectively, reveal the highly asynchronous nature of the cycloaddition (Figure 4b). TS-1-exo and TS-2-endo also show that orbital overlap of the HOMO of furan (24) with the p orbital on C2 of allene 25 is maximized in the endo approach of 24 to 25. The endo reaction is slightly thermodynamically favored (ΔG = −33.9 kcal mol⁻¹) over the exo reaction (ΔG = −32.7 kcal mol⁻¹).

Figure 4.

a) Calculated energy barriers for exo and endo Diels–Alder reaction of cyclohexa-1,2-diene (25) and furan (24). b) Exo and endo transition state geometries. c) D/IAS analysis of endo and exo TSs. Energies in kcal mol⁻¹ are for the ωB97X-D/6-311+G(d,p)/SMD(THF) level.

To understand the origin of this stereoselectivity, a distortion/interaction activation-strain (D/IAS) analysis was performed.^[12,13] In a D/IAS analysis, activation potential energies (ΔE^‡) are analyzed. For each transition state, ΔE^‡ is further broken down into the distortion energy (ΔE_dist^‡) and the interaction energy (ΔE_int^‡). The ΔE_dist^‡ is the energetic cost of deforming the ground states of the reactants into their transition state geometries. The ΔE_int^‡ is an energetic benefit resulting from stabilizing electronic interactions between fragments in the transition state. As shown in Figure 4c, ΔE_dist^‡ is only slightly higher in the TS-1-exo (ΔΔE_dist^‡ = −0.3 kcal mol⁻¹). ΔE_int^‡ was found to be significantly more stabilizing for TS-2-endo (ΔΔE_int^‡ = −1.5 kcal mol⁻¹), indicating that the reaction is selective due to more stabilizing electronic interactions in TS-2-endo than in TS-1-exo.¹⁴

To better understand the favorable electronic interactions that lead to stabilization of ΔE_int along the endo reaction pathway, an energy decomposition analysis (EDA) was performed along both the endo and exo reaction coordinates. EDA was carried out using the ADF.2018.106 program^[15,16] at the ωB97X-D/TZ2P level of theory on the geometries optimized at ωB97X-D/6-311+G(d,p)/SMD(THF) Gaussian 16. An EDA involves the decomposition of ΔE_int into electrostatic (ΔV_elstat), Pauli repulsive (ΔE_Pauli), and orbital interactions (ΔE_orb) as shown in Equation 1.

$$\Delta {\mathrm{E}}_{\text{int}}=\Delta {\mathrm{V}}_{\text{elstat}}+\Delta {\mathrm{E}}_{\text{Pauli}}+\Delta {\mathrm{E}}_{\text{orb}}$$ (Eq. 1)

ΔV_elstat accounts for the electrostatic interaction between the deformed reactants, ΔE_Pauli accounts for the destabilizing interactions between occupied orbitals (i.e., steric repulsion), and ΔE_orb represents the degree of charge transfer between the two reactants (e.g., HOMO–LUMO interactions or polarization). In Figure 5, the EDA terms for endo and exo pathways are plotted along the reaction coordinate. The x-axis corresponds to the length of the forming bond to the central carbon of the allene unit in cyclohexa-1,2-diene (25). ΔE_Pauli is slightly less destabilizing in the exo reaction pathway, which is likely due to the longer lengths of the forming bonds and thus, a smaller degree of steric repulsion. In contrast, ΔE_orb and ΔV_elstat are more favorable in the endo reaction pathway and account for the more stabilizing ΔE_int associated with formation of endo product 26-endo. Thus, the EDA analysis provides evidence that orbital and electrostatic interactions enable more stabilizing ΔE_int in the formation of endo product 26-endo.

Figure 5.

EDA of the reaction of cyclohexa-1,2-diene (25) with furan (24) leading to endo and exo adducts. Energies were calculated at the ωB97X-D/TZ2P level.

To explain why ΔE_orb favors the endo transition state, we analyzed the FMO interactions in TS-2-endo and TS-1-exo (Figure 5). The MOs of the allene LUMO and diene HOMO at TS-2-endo and TS-1-exo were calculated at the TS geometry with HF/6-311+G(d,p). There is a secondary orbital interaction that stabilizes TS-2-endo and has little effect on TS-1-exo. Figure 6 shows two perspectives of TS-2-endo and TS-1-exo that highlight this secondary orbital interaction.

Figure 6.

HOMO furan (24) – LUMO allene 25 interactions in the endo and exo transition states. a) The stabilizing interaction in TS-2-endo is shown as the dashed red line and involves orbital overlap of the HOMO at C3' of furan (24) with the LUMO at C3 of 25. This secondary orbital interaction is not present in TS-1-exo because the HOMO of C3' of furan (24) and the LUMO of C3 of 25 are out-of-phase and overlap is negligible. b) A second perspective of the orbital interactions in TS-2-endo and TS-1-exo.

The interactions are shown from two different perspectives in Figure 6a and 6b. The stabilizing secondary orbital interaction in TS-2-endo involves orbital overlap of the HOMO at C3' of furan (24) with the LUMO at C3 of cyclohexa-1,2-diene (25). In TS-1-exo, the orbital at C3' of furan (24) is out-of-phase with the LUMO of C3 of allene 25 and, therefore, cannot engage in stabilizing interactions. The stabilizing interaction in TS-2-endo is indeed related to the secondary orbital interaction that stabilizes more conventional Diels-Alder transition states, like butadiene or cyclopentadiene dimerization, but occurs here only because of the twisting of the LUMO of the distorted allene. It does not operate in acyclic linear allenes.

Orbital interactions in the endo and exo reaction pathways were quantified using extended transition state-natural orbitals for chemical valence (ETS-NOCV). ETS-NOCV analyses were carried out using the ADF.2018.106 program^[15,16] at the ωB97X-D/TZ2P level of theory on the geometries optimized at the ωB97X-D/6-311+G(d,p)/SMD(THF) level in Gaussian 16. Results from ETS-NOCV along the endo and exo reaction coordinates are shown in Figure 7a. The x-axis corresponds to reaction progress or the C─C bond forming distance to the central carbon of the allene. The strongest interaction (interaction A in Figure 7a) arises from overlap between allene (25) LUMO and furan (24) HOMO and favors the endo reaction. The second strongest ETS-NOCV interaction, interaction B, is due to overlap between the allene 25 HOMO and furan (24) LUMO and is also stronger in the endo transition state. Thus, results from ETS-NOCV reveal that in both normal and inverse electron-demand Diels–Alder reactions of 24 and 25, orbital interactions are more stabilizing along the endo pathway.

Figure 7.

a) ETS-NOCV orbital interaction energies in the Diels–Alder reaction of furan (24) and cyclohexa-1,2-diene (25) leading to endo and exo adducts (ωB97X-D/TZ2P). b) HOMO and LUMO energies of cyclohexa-1,2-diene (25) and furan (24). MO energies provided in units of eV and obtained with HF/6-31G(d). MO energies at the ωB97X-D/6-31G(d) level are provided in parentheses.

Although the allene 25 LUMO/furan (24) HOMO energy gap (11.8 eV, Figure 7b) is smaller than that of the allene 25 HOMO/furan (24) LUMO pair (12.9 eV), orbital interactions between the allene 25 HOMO and furan (24) LUMO still contribute to endo selectivity. This is supported by results from the EDA, which shows that orbital overlap between the allene 25 HOMO and furan (25) LUMO is greater along the endo pathway than along the exo pathway (Figure 8a). Thus, results from the ETS-NOCV method and EDA reveal the presence of a favorable allene 25 HOMO/furan (24) LUMO interaction in the endo TS that further contribute to stereoselectivity. This interaction is shown in Figure 8b and involves overlap of the LUMO at C3' of furan (24) with the HOMO at C3 of allene 25.

Figure 8.

a) EDA orbital overlap between the allene 25 HOMO and furan (24) LUMO in the Diels–Alder reaction (ωB97X-D/TZ2P). The x-axis corresponds to reaction progress or the C─C bond forming distance to the central carbon of allene 25. b) Orbital interactions between the allene 25 HOMO and furan (24) LUMO in endo and exo transition states. The stabilizing interaction in TS-2-endo is shown as the dashed red line. In TS-1-exo, this interaction is antibonding and small from low overlap.

As mentioned earlier, results from the EDA demonstrated that ΔV_elstat is also more favorable in the endo reaction pathway. Thus, we analyzed the ΔV_elstat term that contributes to ΔE_int and the endo selectivity by comparing electrostatic interactions in the endo and exo reaction pathways. Electrostatic charges are most comparable at analogous geometries along the reaction coordinate where the same degree of bond formation has occurred. Since the endo TS occurs earlier than the exo TS (ΔE_endo^‡ = 5.6 kcal mol⁻¹ while ΔE_exo^‡ = 7.3 kcal mol⁻¹), it was necessary to first identify analogous geometries along each reaction pathway and then calculate electrostatic charges of the corresponding structures. We chose analogous endo and exo geometries where C─C bond forming distances to the central carbon of the allene are very similar (2.25–2.26 Å).

Figure 9 shows the analogous geometries Endo Structure A (ΔE = 4.9 kcal mol⁻¹) and Exo Structure B (ΔE = 6.2 kcal mol⁻¹). It also summarizes the Hirshfeld charges of the oxygen atom of furan (24) and the allenic protons H1 or H3 (see the Supporting Information Part II for Hirshfeld charges of additional atoms) and shows the corresponding electrostatic potential maps. In Endo Structure A, the oxygen atom of 24 and H1 of allene 25 are in closer proximity (2.8 Å) than are the oxygen atom of 24 and H3 in Exo Structure B (3.6 Å). There is a stabilizing electrostatic interaction and stabilization in the endo reaction pathway, which is further supported by the electrostatic potential map of Endo Structure A. The reduced endo selectivity calculated for the reaction of cyclohexa-1,2-diene (25) with cyclopentadiene, which lacks the partially negative oxygen, also serves as evidence for this.^[17] The electrostatic potential maps also show that the Exo Structure B is destabilized by a repulsive H2’-H3 interaction that is much smaller in Endo Structure A. Overall, our analysis reveals the electrostatic and orbital interactions that allow for endo selectivity in the Diels–Alder reaction of cyclohexa-1,2-diene (25) and furan (24). Secondary orbital interactions stabilize the endo approach of diene 24 to allene 25 and involve orbital overlap of the HOMO at C3' of diene 24 with the LUMO at C3 of allene 25.

Figure 9.

Analogous geometries on the endo and exo reaction pathways of furan (24) and allene 25. Hirshfeld charges and electrostatic potentials maps were calculated with ωB97X-D/6-311+G(d,p).

Diels–Alder Reactions of Strained Heterocyclic Allenes

We have also computed the endo/exo selectivity in Diels–Alder cycloadditions with other strained heterocyclic allenes and dienes (Figure 10). The Gibbs energy barriers (ΔG^‡) for Diels–Alder reactions of allenes 7, 29, and 31 with dienes 24, 28, or 34 were computed (Figure 10).^[18] Previous experimental studies demonstrated that Diels–Alder reactions of allenes 7, 29, and 31 proceed with good to excellent endo selectivity (7.4:1 to >20:1 endo:exo, Figure 10).^{[2c, 2d]} Interestingly, methyl-substituted allene 31 undergoes the [4+2] cycloaddition with higher stereoselectivity (>20:1 endo:exo) than unsubstituted allenes 7 and 29. Computational analysis of the reaction with 31 in comparison to the reactions with allenes 7 and 29 were undertaken to provide insight into allene substituent effects on endo selectivity in Diels-–Alder reactions with strained cyclic allenes.

Figure 10.

Energy barriers for Diels–Alder reactions of strained heterocyclic allenes and various dienes leading to endo and exo adducts. Energies in kcal mol⁻¹ computed at the ωB97X-D/6-311+G(d,p)/SMD(MeCN) level. ^aThe corresponding allene derivative containing a carboxybenzyl group in place of the carbomethoxy group was used in the trapping experiment.

The calculations on cycloadditions of oxacyclic allene 7 with 2,5-dimethylfuran (28) and azacyclic allene 29 and furan (24) indicate that formation of endo adducts 13-endo and 30-endo proceed with barriers of ΔG^‡ = 17.4 kcal mol⁻¹ and ΔG^‡ = 16.8 kcal mol⁻¹, respectively. Formation of exo adducts 13-exo (ΔG^‡ = 19.5 kcal mol⁻¹) and 30-exo (ΔG^‡ = 18.3 kcal mol⁻¹) is disfavored by 1.5–2.1 kcal mol⁻¹. Our calculated ΔΔG^‡ of 2.1 kcal mol⁻¹ (35:1 endo:exo ratio) for reaction of allene 7 and 2,5-dimethylfuran (28) and 1.5 kcal mol⁻¹ (13:1 endo:exo ratio) for reaction of allene 29 and furan (24) are in reasonable agreement with the experimental endo:exo ratios of 9.2:1 and 7.4:1, respectively.

Energetic barriers were also evaluated for the cycloadditions of alkyl-substituted allene 31 with furan (24), 2,5-dimethylfuran (28), and N-phenylpyrrole (34). Previous computational studies performed by our lab demonstrated that the alkyl substituent in allene 31 allows for regioselective addition onto the more electron poor π bond of allene 31 via an electronic effect.^[2c,19] Therefore, reaction pathways corresponding to regioselective addition onto the unsubstituted π bond of allene 31 were considered for the stereoselectivity studies shown in Figure 10. Our calculated ΔΔG^‡ values of 0.7–3.9 kcal mol⁻¹ for the cycloadditions of allene 31 with dienes 24, 28, and 34 demonstrate that endo adducts 32-endo, 33-endo, and 35-endo are generated preferentially over exo adducts 32-exo, 33-exo, and 35-exo.²⁰ Computed ΔΔG^‡ values qualitatively agree with the experimental endo:exo ratios shown in Figure 10.

To gain insight regarding substituent effects on stereoselectivity, D/IAS analysis was performed for reactions of allene 29 with furan (24) and allene 31 with dienes 24, 28, and 34. We hypothesized that hyperconjugative interactions of the allenic methyl substituent with the non-reactive π bond of 31 allows for stronger secondary orbital interaction between the diene and dienophile in the endo reaction. To provide support for our hypothesis, we performed a D/IAS analysis of the reaction of allene 31 with diene 24. Based on the analysis, the comparatively higher endo selectivity in Diels–Alder reactions of allene 31 with 24 (ΔΔG^‡ = 3.2 kcal mol⁻¹) is attributed to both more stabilizing ΔE_int and lower ΔE_dist along the endo reaction pathway.²¹ The more stabilizing ΔE_int supports the idea that hyperconjugative interactions allowing for stronger secondary orbital interactions in the endo reaction. Additionally, the lower ΔE_dist emerges from the presence of unfavorable steric interactions between the H at C2' of furan (24) and the methyl substituent on allene 31 along the exo reaction pathway.

D/IAS analysis for the reaction of allene 31 with 2,5-dimethylfuran (28) (ΔΔG^‡ = 3.9 kcal mol⁻¹) was performed next, revealing that the enhanced endo selectivity arises from steric, rather than electronic, factors.²¹ In this case, the methyl group at C2' of 2,5-dimethylfuran (28) has unfavorable steric interactions with the methyl substituent in allene 31 in the exo approach of the diene to the dienophile. Interestingly, selectivity in the reaction of allene 31 with N-phenylpyrrole (34) decreases (ΔΔG^‡ = 0.7 kcal mol⁻¹) due to competing ΔE_int and ΔE_dist along endo and exo reaction coordinates. While ΔE_dist is lower in the endo reaction, N-phenylpyrrole (34) does not have as stabilizing secondary orbital interactions with allene 31 as does furan (24).²² This may be due to differences in orbital localization in furans and pyrroles.

Conclusion

This study elucidates the factors contributing to endo selectivity in Diels–Alder reactions with strained cyclic allenes. In the endo reaction pathways of dienes with cyclohexa-1,2-diene and heterocyclic and substituted analogs, there is a stabilizing interaction between C3' (attached to the furan atom that is bonding to the central carbon of allene) and the LUMO orbital at C3 of the allene. The latter is a new type of secondary orbital interaction that results from the near perpendicular approach of the diene to the dienophile and the twisted nature of the strained allene. In the exo TS, this interaction is small but destabilizing. The attractive electrostatic interaction of the oxygen atom of furan (24) with the proximal allenic proton H1 of allene 25 provides additional stabilization in the endo reaction pathway. West proposed an electrostatic interaction of the furan O with an exo CH carbon.²¹ We also studied the selectivities of Diels–Alder reactions of various dienes with heterocyclic allenes, including oxacyclohexa-3,4-diene (7), azacyclohexa-3,4-diene (29), and 3-methyl-1-azacyclohexa-3,4-diene (31). Computed endo selectivities agreed with experimental results. These selectivities are generally 1–2 kcal mol⁻¹, larger than the well-known endo selectivities for monosubstituted dienophiles with cyclopentadiene (0.5–1 kcal mol⁻¹).²³ These results are expected to prove useful in future synthesis design involving Diels–Alder reactions with strained cyclic allenes.