Controlling, understanding, and redirecting the thermal rearrangement of 3,3-dicyano-1,5-enynes

Abstract

The thermal [3,3] rearrangement of 3,3-di-cyano-1,5-enynes to γ-allenyl alkylidenemalononitriles (the “enyne Cope rearrangement”) has largely eluded synthetic value as the desired products, too, are thermally reactive and ultimately yield 6π electrocyclization products. Herein, we describe experimental and computation studies related to the thermal rearrangement of 1,5-enynes, structural features to halt the thermal rearrangement at the allene-stage, and a reductive variant for preparing bifunctional allenyl malononitriles. We also describe various ways that the bifunctional building blocks can be manipulated and converted to cyclic and acyclic architectures.

Graphical Abstract

INTRODUCTION

Thermal rearrangement of 3,3-dicyano-1,5-enynes can yield a suite of products (Scheme 1A).¹ In cascade fashion, they undergo [3,3] sigmatropic rearrangement to γ-allenyl alkylidenemalononitrile 2, “proton transfer” to the triene, 6π-electrocyclization to 3, and [1,5]CN shift to 4. This thermal sequence was originally reported by Roger Brown and co-workers with little exploration of optimization or scope.¹ Although many of the products 2 – 4 likely have synthetic value, we are interested in the allenyl intermediate 2, which bears a reactive alkylidenemalononitrile and allene. Generally speaking, the enyne Cope rearrangement, also known in the literature as the acetylenic Cope rearrangement or the propargyl Cope rearrangement, has only a few reports including physical organic studies,² a retro-variant,³ and theoretical studies.⁴ Notably, the [3,3] rearrangement of propargyl enol ethers (the “propargyl Claisen rearrangement”) has seen considerable application in modern synthesis.⁵

Scheme 1.

A: Thermal rearrangement of 3,3-dicyano-1,5-enynes. B and C: The allenyl intermediate “2” and its reduced counterpart “H₂-2” are attractive bifunctional building blocks.

We are interested in applying the enyne Cope rearrangement of 3,3-dicyano-1,5-enynes 1 to complex molecule synthesis.⁶ Enynes 1 are readily constructed from ketones, malononitrile, and propargyl bromides, which allows for significant substrate diversity. If the products 2 are rendered highly accessible by the Cope rearrangement of enynes 1, a vast array of chemical space can be easily accessed by standard chemistry (Scheme 1B). For example, alkylidenemalononitriles undergo mild reduction (e.g. to H₂-2)⁷ or can be alkylated by deconjugative alkylation (e.g. to I).⁶ Attractively, 2 are bifunctional in that they can undergo mild deconjugative alkylation then cyclization with propargyl bromides or furfuryl-derived alkyl halides to yield hydroazulenes II⁸ or decalins III⁹, respectively, in two steps. The reduced counterparts H₂-2 also have some unique and noteworthy reactivity (Scheme 2C). Mono-alkylmalononitriles undergo oxidative decyanation to yield esters or amides IV.¹⁰ It has also been found that these building blocks undergo a mild-base catalyzed cycloisomerization to cyclopentenes V.¹¹ Additionally, deconjugative alkylation/cyclization, similar to as described for 2, can yield common natural product frameworks (VI – VII). The major limitation is accessing these versatile building blocks via the enyne Cope rearrangement because, as described above, the thermal reaction results in various products (2 – 4, Scheme 1A). This is the primary contributor to the modest yields of γ-allenylalkylidenemalononitriles. Considering the potential value of the enyne Cope rearrangement of 3,3-dicyano-1,5-enynes to synthesis, we wished to better understand the transformation and improve its scope, efficiency, and applicability, which is described herein.

Scheme 2.

Thermal cascade reaction of 3,3-dicyano-1,5-enynes.

THERMAL TRANSFORMATION OF SIMPLE, MODEL 3,3-DICYANO-1,5-ENYNES

As noted in previous work from Brown,¹ our own studies,⁶ and further explored herein, the thermal rearrangement of 1a begins to occur at temperatures above ~110 °C. We first evaluated the effect of time on the product distribution (Scheme 2, entries 1 and 2). The allene is favored at shorter reaction times (entry 1); however, as the heating time is increased, the reaction proceeds to favor the electrocyclization products (entry 2). We saw similar results with the five-membered ring variant (1b, entries 3 –5). In trace amounts, we also observed products of auto-oxidation to the benzodinitriles 5 in these studies. Brown and co-workers noted significantly more benzenoid formation under their conditions (200 °C). Furthermore, we found that the acidic additive trifluoroethanol (TFE) helped promote the proton transfer step as well as the [1,5]CN shift. In a similar fashion, we also examined the 1,5-enyne 1c and 1d bearing an internal alkyne. Under analogous conditions (150 °C, toluene), no reaction occurred, indicating a high barrier to [3,3] rearrangement. When increasing the temperature, conversion began to occur, but in no cases were allenes observed. This is likely because the additional energy necessary for the [3,3] also facilities the additional isomerizations.

COMPUTATIONAL ANALYSIS OF THE [3,3] REARRANGEMENT OF MODEL 3,3-DICYANO-1,5-ENYNES

We next turned to computation and performed state-of-the-art density functional theory computations to better understand the sigmatropic rearrangement step (Figure 1). Our main finding was that the enyne Cope rearrangement is sensitive to whether the starting alkyne is terminal or internal. In particular, we computed a free energy barrier of 31.5 kcal/mol for the parent terminal alkyne 1a, but larger barriers of 33.6 kcal/mol and 33.5 kcal/mol respectively for the related internal alkynes 1c and 1d. The breaking and forming bonds in the internal alkyne transition states were approximately 0.1 Å longer than in the terminal alkyne transition state, suggesting that the higher barrier comes from steric hindrance that prevents the enyne termini from approaching each other as closely. This computational result is consistent with the experimental finding that internal alkynes require higher temperatures for the Cope rearrangement, making them more prone to rearrangement through the pathway shown in Scheme 1A.

Figure 1.

Free energy barriers for the enyne Cope rearrangement of terminal and internal alkynes.

We found a similar trend with bicyclic scaffolds (Figure 1), with internal alkynes 1f and 1g having free energy barriers that were 1–2 kcal/mol higher and bond lengths that were ~0.1 Å longer than terminal alkyne 1e. However, as compared to the monocyclic scaffolds, the bicyclic scaffolds exhibit a lower free energy barrier, and computations reveal that this is due to a raised ground state that possesses significant torsional strain. As shown in Figure 2, the bicyclic scaffold 1e requires the starting material to adopt a near-eclipsed conformation with an illustrated HCCH dihedral angle of 19.6 degrees. This high-energy dihedral angle is subsequently relaxed to 39.7 degrees in the transition state (and 61.8 degrees in the product). In contrast, monocyclic starting material 1a has a corresponding dihedral angle of 44.2 degrees, so it is not significantly destabilized and consequently has a higher free energy barrier for the enyne Cope rearrangement.

Figure 2.

In bicyclic scaffold 1e, the illustrated HCCH dihedral angle increases from 19.6 degrees in the starting material to 39.7 degrees in the transition state and 61.8 degrees in the product.

Because some allene products undergo further reaction (Scheme 2), we performed a conformational study of product allene 2a formed from terminal alkyne 1a. As shown in Figure 3, the axial conformation of the allene formed directly in the Cope rearrangement is more stable than the equatorial conformation by 2.8 kcal/mol. The equatorial conformation suffers from allylic strain between the allene and alkylidenemalononitrile moieties, which is alleviated by having the allene eclipse the axial γ-C–H bond. Nevertheless, the equatorial conformation of the allene is readily accessible via a chair-chair inter-conversion under the reaction conditions. We hypothesize that this conformation is critical to decomposition because the axial γ-C–H bond is aligned with the π system of the alkylidenemalononitrile for deprotonation and, therefore, further reaction. In contrast, in the bicyclic systems 1e, 1f, and 1g, the allene is locked in the axial conformation and the γ-C–H bond is orthogonal to the π system of the alkylidenemalononitrile, so no further reaction can occur.

Figure 3.

The axial conformation of allene 2a is morestable than the equatorial conformation by 2.8kcal/mol.

Based on the above experimental and computational data, we have come to the following conclusions:

1. The transformation is sensitive to steric effects at the enyne termini. Experimentally, internal alkynes require higher temperature and therefore decompose unless structural features are present which prevent access to the reactive conformer. This is supported by the computational studies showing a ~2 kcal/mol difference in transition state energy between terminal alkyne 1a and internal alkynes 1c and 1d.

2. The bicyclic scaffolds examined have a lower activation energy as a result of a destabilized ground state.

3. The allene prefers to be axial, which minimizes allylic strain between the alkylidenemalononitrile and the allene.

4. Restricting conformations post [3,3] rearrangement can preclude the proton-transfer step (and therefore downstream rearrangements) resulting in higher yields of the targeted allenes. This is because the γ-C–H bond is more readily deprotonated when it is aligned with the alkylidenemalononitrile π system.

This computational data supports the following mechanism for thermal cascade rearrangement (Scheme 3): [3,3] rearrangement occurs to yield the chair conformer having the allene axial (chair-ax). This conformer must undergo a chair-chair interconversion to the equatorial allene conformer (chair-eq), the acidic conformer, to allow for proton transfer to the triene, followed by electrocyclization and 1,5[CN] shift.

Scheme 3.

Detailed mechanism for thermal rearrangement

The above computational discoveries set the stage for new hypotheses aimed at improving the scope of allene synthesis by [3,3] rearrangement. Specifically, we wished to identify substrate classes that would stop thermally rearranging prior to allene isomerization as well as devise a protocol to prepare allenes by this method that is not tied to substrate structure. These are the subjects of the following two sections, respectively.

PREVENTING THE THERMAL CASCADE THROUGH SUBSTRATE STRUCTURE BIAS

We suspected that certain substructures could be incorporated onto the 1,5-enyne to prevent the proton transfer step, thus preventing thermal cascade and yielding the targeted bifunctional γ-allenyl alkylidenemalononitriles. Considering that the allene is delivered axially and that the axial allene is preferred over the equatorial allene, we reasoned that biasing one chair conformation over another could halt the thermal cascade following allene formation.

We prepared substrates 1e, 1h – 1i, which would lock the allene in a non-acidic conformation, and substrates 1j–1m, which would bias the allene in a non-acidic conformation (Scheme 4). The locked substrates underwent smooth transformation and gave high yields of the desired allenes 2e, 2h, and 2i. Notably, this transformation was tolerant to alkyne substitution, which was not possible in the simple system (e.g. 1c). A large ^tbutyl group also gave a high yield of the desired allene 2j when the starting enyne was terminal. The analogous internal enyne 1k (R = Me) also yielded the desired allene 2k. Although modest yield and conversion was observed, this is a substantial improvement, as 1c, lacking the biasing group, shows no allene product. Similarly, the menthone derived enynes successfully yielded allenes 2i and 2m. In this case, smaller alkyl groups on the cyclohexyl ring, when compounded, are effective at biasing chair conformations.

Scheme 4.

Preventing thermal cascade by substructure bias

THE REDUCTIVE ENYNE COPE REARRANGEMENT

In the above section, we described that the thermal cascade can be halted at the allene stage when utilizing sub-strates that prefer one chair conformation over the other (locked or biased substrates). Although pleased to better understand the Cope rearrangement and yield structurally unique allenes, this approach is somewhat limiting, as it requires specific biasing elements. A method not tied to a structural architecture would yield a wider scope. We hypothesized that the Cope rearrangement in the presence of a reductant that could chemoselectively react with the in situ generated alkylidenemalononitrile-moiety would divert the transformation before the γ-allenylalkyli-denemalononitrile can further isomerize (Scheme 5). If achieved, this would yield bifunctional allenyl malono-nitriles H₂-2, useful building blocks for synthesis.

Scheme 5.

The Reductive Enyne Cope Rearrangement

To test the hypothesis, we prepared the 3,3-dicyano-1,5-enyne 1b derived from cyclopentanone, malononitrile, and propargyl bromide and began examining the reductive enyne Cope rearrangement (Table 1). After some experimentation, it was found that adding Hantzsch ester (HE) to the reaction results in product 6b exclusively in good yield as a mixture of diastereomers (entry 1).⁸ Other solvents and mixtures of solvents provided the desired product with varying efficiency. We also explored a few other reductants that were unsuccessful.

Table 1.

Development of the reductive Cope rearrangement

entry	Reductant(2 equiv)	Solvent	(h)	% conv.	6b	2b	3b	4b
1	Hantzsch Ester	tol	2	40	1	0	0	0
2	Hantzsch Ester	tol	7	>95	1 (70%)	0	0	0
3	Hantzsch Ester	propanol	2	35	1	0	0	0
4	Hantzsch Ester	tol:PrOH (5:1)	2	52	1	0	0	0
5	Hantzsch Ester	tol:TFE (5:1)	2	64	3	0	0	1
6	Hantzch Amide	tol	2	39	0	0	1	0
7	NaCNBH3	tol	2	40	0	0	1	1

With optimized conditions in hand, we found that many allenyl malononitriles could be prepared (Scheme 6). The scope was broad, in part, due to the ready availability of the 1,5-enynes examined. For example, products 6a – 6h were prepared from malononitrile, propargyl bromide and a variable cycloalkanone: cyclohexanone (6a), cyclopentanone (6b), protected 1,4-cycloalkandione (6c), N-Boc piperidinone (6d), α-tetralone (6e), β-tetralone (6f), 4-methylcyclohexanone (6g), and 2-methylcyclohexanone (6h). We also examined acyclic ketones (6i – 6m) and aldehydes (6n – 6p). The reaction is generally tolerant to functional groups such as ketals (6c), acetals (6l), Boc-protected amines (6d), esters (6j), and cyclopropanes (6m). Regarding diastereoselectivity, the reduction step is poorly selective, generally speaking. However, the Cope rearrangement-step is diastereoselective (6g and 6h). In both these cases, the reduction step still yields a diastereomeric mixture. Products 6a, 6e, 6l, 6m, and 6n were prepared on the > gram scale. Generally speaking, the standard scale (50 – 100 mgs) and the scale up procedures (2 – 5 grams) gave similar results.

Scheme 6.

Scope of the reductive Cope rearrangement.

^a standard conditions: 30 – 100 mgs of enyne, Hantzsch ester (HE) (2 equiv) tol (0.1 M), 130 – 180 °C ^b the minor product is epimeric at the starred position ^c 180 °C in lieu of 140 – 150 °C ^d 4 equiv. HE

In final examples related to the scope of the reductive enyne Cope rearrangement, methyl and allyl cyanoacetate derived substrates were examined (Scheme 6). Although they required a higher temperature to react (180 °C in lieu of 140 – 150 °C), they still yielded the desired allenes 6q and 6r in good yield with the expected modest diastere-oselectivity.

In a few cases, we found that the reductive Cope product was inseparable from the Hantzsch ester. We found that “telescoping” the reaction by malononitrile alkylation with 7a would allow for isolation of target building blocks, in this case the furan-allene building block 8a (Scheme 6).

We next examined the utility of the allene/malononitrile building blocks (Scheme 7). Functionally dense methyl-cyclopentenes 9 can be prepared by a base-catalyzed cycloisomerization.¹¹ Decalin ring systems 10 are accessible by an alkylation (with 7a)/thermal [4+2] strategy.⁹ Hydroazulenes 11 are prepared by allenic-Pauson-Khand reaction (PKR) under the conditions reported by Brum-mond.⁸

Scheme 7.

Utility of allenyl malononitrile building blocks

^a see Scheme 7 for the synthesis of the starting material 8a. ^b + 25% recovered starting material ^c major diatereomer shown. minor diastereomer is epimeric at labled carbon ^d major diastereomer separable by chromatography, dr determined from crude NMR analysis

As final experiments, we explored how the malono-nitrile moiety can be manipulated (Scheme 8). Monoal-kylmalononitriles can be thought of as carboxylate equivalents: Using magnesium monoperoxyphthalate (MMPP) as an oxidant in methanol, allene/carboxylic esters 12 are prepared.¹⁰ In the cases of disubstituted malononitriles, formal decyanation can mildly be achieved by Pinner reaction with allyl alcohol to the allyl cyanoacetates 13 followed by Pd-catalyzed decarboxylative protonation yielding mono-nitriles 14.¹²

Scheme 8.

Malononitrile functional group interconversions.

In conclusion, we have explored various aspects of the rearrangement of 3,3-dicyano-1,5-enynes. Through experimental and computational studies, we have a better understanding of how to control and halt the rearrangement at the prior to allene isomerization, as well as a method to divert from the thermal pathway by chemoselective reduction. The bifunctional allenyl alkyli-denemalononitriles and allenyl malononitriles are also well suited for manipulation into a variety of cyclic or acyclic building blocks. Future directions include complex molecule synthesis, scope expansion, improving the diastereoselectivity, and identifying enantioselective variants.