Ring Enlargement of Three-Membered Heterocycles by Treatment with In Situ Formed Tricyanomethane.

Although the chemistry of elusive tricyanomethane (cyanoform) has been studied during a period of more than 150 years, this compound has very rarely been utilized in the synthesis or modification of heterocycles. Three-membered heterocycles, such as epoxides, thiirane, aziridines, or 2H-azirines, are now treated with tricyanomethane, which is generated in situ by heating azidomethylidene-malonodinitrile in tetrahydrofuran at 45 °C or by adding sulfuric acid to potassium tricyanomethanide. This leads to ring expansion with formation of 2-(dicyanomethylidene)oxazolidine derivatives or creation of the corresponding thiazolidine, imidazolidine, or imidazoline compounds and opens up a new access to these push-pull-substituted olefinic products. The regio- and stereochemistry of the ring-enlargement processes are discussed, and the proposed reaction mechanisms were confirmed by using 15 N-labeled substrates. It turns out that different mechanisms are operating; however, tricyanomethanide is always acting as a nitrogen-centered nucleophile, which is quite unusual if compared to other reactions of this species.

Introduction

The history of tricyanomethane (5) dates back to 1864 when the first isolation of so‐called cyanoform was reported.1 But this and a second synthesis2 could not be reproduced by other chemists,2, 3 and later it was found that 5, even if generated, would not have been able to survive the described drastic reaction conditions. In 1896, Schmidtmann treated sodium tricyanomethanide (1 a) with dilute sulfuric acid and then with diethyl ether (Scheme 1).4 He obtained a three‐phase system, which comprised a greenish yellow middle layer 2 that was claimed to include cyanoform (5), although first experiments to remove the solvents did not lead to characterizable substances. Three years later, Hantzsch and Osswald confirmed these phenomena and stated that tautomerism of 5 is likely to form ketenimine 3.5 In 1963, Trofimenko used the salt 1 b to repeat the synthesis of 2 and called this compound “aquoethereal cyanoform”.6 He successfully liberated 2 from the solvents by rapid evaporation and vacuum sublimation to receive unstable white crystals, to which he erroneously assigned the structure of 3. When Dunitz et al. performed the rapid evaporation of 2 without sublimation, they were able to isolate single crystals of hydronium tricyanomethanide (4), and that facilitated the structure verification by X‐ray diffraction studies.7 In 2017, it was shown that Trofimenko's experiment did not produce 3 because rapid evaporation and sublimation of 2 led to the isolation of a mixture of 4 and 5.8 Under special conditions of vacuum sublimation, single crystals of tricyanomethane (5) were accessible, which allowed structure confirmation by X‐ray diffraction. If moisture was excluded, such single crystals could be handled at ambient temperature for a short time. But cyanoform (5) could not be analyzed in solution by NMR spectroscopy at room temperature due to rapid decomposition. Moreover, 5 was easily converted into 4 even if only trace amounts of water were present. Especially, mixtures of 4 and 5 tended to dynamic processes, which led to extremely broad 1H and 13C NMR signals at temperatures between −50 and 0 °C. Thus, even lower temperatures, for example, −70 °C, were necessary to detect sharp 13C NMR signals for pure 5 in [D8]THF.8

Scheme 1

Previous syntheses of tricyanomethane (5).

Cyanoform (5) was prepared not only from 1 a or 1 b via 2, but also by reacting 1 c with hydrogen sulfide,9 or alternatively by treatment of 1 d with an excess of anhydrous hydrogen fluoride,10 and finally by short‐time thermolysis or photolysis of vinyl azide 7.8 The products 5 and 6 could not be separated in the case of precursor 1 d; however, convincing spectroscopic identification of 5 was nevertheless possible.10 Whereas thermolysis of 7 produced only 5 and small amounts of 4, irradiation of 7 in solution led to 5 and the 2H‐azirine 8.8 The photolysis of 7 isolated in an argon matrix did not generate 5 since ketenimine 3 and the heterocycle 8 were formed instead.

During a period of more than 150 years, the chemistry of cyanoform (5) has been studied thoroughly.11 Its gas‐phase structure was investigated by photoelectron spectroscopy12 and microwave spectroscopy.9 The relative stabilities, spectroscopic data, and isomerization reactions of 3, 5, and other C4HN3 species were analyzed by utilizing quantum‐chemical methods.13 In particular, the acidic properties of 5 as well as its tautomer 3 were discussed intensively.5, 6, 14 Hence, 5 is presented in textbooks of organic chemistry as one of the strongest carbon acids.

As a Brønsted acid, tricyanomethane (5) should be able to catalyze the ring opening of three‐membered heterocycles, for example, epoxides. If no other reactive species such as competing nucleophiles are present, interesting novel types of open‐chain or ring‐expanded products may result. Herein, we report on surprising transformations which were induced by treatment of oxiranes, thiirane, aziridines, and 2H‐azirines with in situ formed cyanoform (5).

Results and Discussion

At first, we utilized the commercially available cyclohexene oxide (9 a) as a substrate that was exposed to the precursor 7 in anhydrous THF at 45 °C (Scheme 2). We expected that 5 resulting from 7 would O‐protonate 9 a, and the nucleophilic attack of the central carbon atom of the tricyanomethanide counterion would lead to the ring‐opened product 10 a, which can possibly react to 12 a by nucleophilic addition of the hydroxy group at a cyano unit. The former assumption was seemingly supported by the well‐known15 C‐alkylation of tricyanomethanide salts such as 1 b with the help of alternative electrophiles,16, 17, 18 for example, primary alkyl halides. But instead of 10 a or 12 a, we obtained the isomeric product 13 a after treatment of 9 a with 7. Obviously, protonated 9 a was attacked by a nitrogen atom, and after ring opening, the resulting ketenimine 11 a was transformed into the final product 13 a, which includes a push–pull‐substituted C=C unit. Whereas both rings are cis‐fused in the starting compound 9 a, the 1H NMR data of the ring‐expanded product 13 a indicated a large vicinal coupling of the two axial bridgehead protons with 3 J=12.0 Hz. Consequently, both rings in 13 a are trans‐fused.

Scheme 2

Ring enlargement of 9 a by treatment with 7 and in situ generation of tricyanomethane (5).

Besides 9 a, we similarly treated also the epoxides 9 b–h with vinyl azide 7 to obtain the ring‐enlargement products 13 (Table 1). While the parent oxirane (9 b) exclusively led to the single product 13 b,19 the reaction of propylene oxide (9 c) resulted in the formation of two regioisomeric oxazole derivatives, which were substituted with a methyl group in the 5‐ or in the 4‐position. Both isomers were easily separated by chromatography, and assignment by 1H and 13C NMR spectroscopy was simple since the adjacent oxygen atom always induced significantly stronger deshielding properties than the amine unit. In the case of styrene oxide (9 d), treatment with 7 regioselectively led to 13 d as the sole product, and the analogous reaction of the unsymmetrical epoxide 9 e similarly resulted in the exclusive formation of the oxazole derivative 13 e, which was isolated in 79 % yield. Thus, the well‐known20 regiochemistry of acid‐catalyzed epoxide ring opening was also observed for the conversion of 9 into 13 although the involved reagents did not include an acid on the first view. However, slight warming of vinyl azide 7 generated the Brønsted acid cyanoform (5), which initiated the desired transformation by O‐protonation of 9.

Table 1

Treatment of epoxides 9 with azide 7 to prepare the ring‐enlargement products 13.

	Substrate 9			Product 13
	R1	R2	R3	R1	R2	R3	Yield [%]]a]
a	H	(CH2)4		H	(CH2)4		56
b	H	H	H	H	H	H	72
c	H	H	Me	H	H	Me	25
				Me	H	H	22
d	Ph	H	H	Ph	H	H	58
e	Ph	Ph	H	Ph	Ph	H	79
f	Ph	H	Bz	Ph	H	Bz	27
				H	Ph	Bz	26
g	H	Ph	Ph	H	Ph	Ph	46[b]
h	Ph	H	Ph	H	Ph	Ph	47[b]

[a] Isolated yields; after separation of two products in the case of 13 c and 13 f. [b] 1,2‐Diphenylethanone and diphenylacetaldehyde were additionally formed and isolated with a total yield of 13–14 %.

When the epoxide 9 f, which includes a phenyl group and an electron‐withdrawing benzoyl substituent, was subjected to 7, the product 13 f was formed with perfect regioselectivity. However, 13 f was composed of two stereoisomers isolated with nearly equal yields; the structural assignment of cis‐13 f and trans‐13 f was based on 1H NMR spectroscopy utilizing vicinal coupling constants and NOE experiments. In the case of trans‐4,5‐diphenyloxazole derivative 13 g/h, the synthesis was successful starting with oxirane 9 g or alternatively with the stereoisomer 9 h. Thus, both reactions with 7 are stereoselective, but not stereospecific since the same product (13 g=13 h) was formed. The transformations of 9 g and 9 h were accompanied by the generation of 1,2‐diphenylethanone and diphenylacetaldehyde (total isolated yield: 13–14 %). These side products obviously resulted from ring opening of the O‐protonated epoxides followed by hydride shift or phenyl migration of the substituted benzyl cation. Similar carbocations were involved in the reactions of 9 d or 9 e with 7 which explains the regioselectivity in the formation of 13 d and 13 e, respectively. The structure of 13 g was confirmed not only by the usual spectroscopic characterization, but also by single crystal X‐ray diffraction analysis (Figure 1).

Figure 1

ORTEP (30 % ellipsoid probability) of the molecular structure of 13 g. A second molecule in the asymmetric unit has been omitted for clarity. Selected bond properties ([Å]/[°]): C1−O1 1.45(3), C1−N1 1.31(3), C1−C4 1.37(3), O1−C2 1.45(3), C2−C3 1.60(3), C3−N1 1.43(2), C4−C5 1.40(3), C5−N3 1.15(3), C4−C6 1.43(3), C6−N2 1.21(3); N1‐C1‐O1 111(2), C1‐N1‐C3 114(2), C5‐C4‐C6 120(2), C7‐C2‐C3‐C13 128(2).

The ring‐enlargement reaction with the help of vinyl azide 7 was also transferred from oxiranes 9 to thiirane (14) and aziridines 16 and 18 to obtain the five‐membered heterocycles 15,21 17, and 19, respectively (Scheme 3). The 13C NMR spectrum of imidazole derivative 17, measured in [D6]DMSO at room temperature, showed only a broad signal for both cyano groups, whereas the corresponding spectrum of a solution in [D6]acetone exhibited two signals at ambient temperature. When the temperature was raised to 45 °C in the latter case, coalescence of the two signals was observed. Rapid exchange of both cyano groups was obviously initiated by rotation about the push–pull‐substituted C=C bond. The π system of this bond is weakened owing to a dipolar resonance structure, which includes a positive charge, stabilized by the ring nitrogen atoms, and a negative charge delocalized by the cyano groups. The molecular structures of the heterocyclic products 15, 17, and 19 were confirmed by the usual spectroscopic characterization and also by single crystal X‐ray diffraction analysis of 17 (Figure 2).

Scheme 3

Ring enlargement of 14, 16, and 18.

Figure 2

ORTEP (50 % ellipsoid probability) of the molecular structure of 17. A second molecule in the asymmetric unit has been omitted for clarity. Selected bond properties ([Å]/[°]): S1−N1 1.6929(13), N1−C8 1.4979(18), C8−C9 1.525(2), C9−N2 1.463(2), N2−C10 1.3239(18), C10−N1 1.4121(18), C10−C11 1.388(2), C11−C12 1.418(2), C12−N3 1.147(2), C11−C13 1.421(2), C13−N4 1.1508(19); C8‐N1‐C10 105.08(11), C9‐N2‐C10 112.78(13), N1‐C10‐N2 110.05(13), C12‐C11‐C13 117.06(13), N1‐C8‐C9‐N2 26.89(14).

Finally, we treated the 2H‐azirines 21 with the azide 7 to induce the reaction of in situ generated cyanoform (5) with these highly strained three‐membered heterocycles (Scheme 4). Since the addition of 5 at the C=N bond of azirine 8, which was performed by irradiating of 7 at low temperatures and warming the resulting mixture of the photoproducts 5 and 8 to ≥−30 °C, was reported to lead to the unstable aziridine 20,8 we expected analogous final products from substrates 7 and 21. But the latter starting compounds afforded novel ring‐extended imidazole derivatives 22 instead of aziridines. The products 22 a–d were formed with 56–62 % yield; and in the case of the very unstable22 2H‐azirine 21 b, which does not possess a substituent in the 3‐position, the yield was even slightly higher than that for the transformation of the robust23 substrate 21 a, although the corresponding products 22 a and 22 b are identical (Table 2). It turned out that acetonitrile was a better solvent for this ring‐enlargement reaction than tetrahydrofuran since more convenient workup was possible and pure products were obtained easily. The new heterocycles 22 were characterized not only by the usual spectroscopic methods, but also by 15N NMR spectra of 22 a and 22 c. After assignment of the two NH proton signals of 22 a by homonuclear NOE experiments, 2D‐15N,1H shift correlation (HSQC) also facilitated the allocation of the 15N NMR signals. In the 13C and 15N NMR spectra of the unsymmetrical heterocyclic compound 22 a, both cyano groups led to a single signal because of a rapid rotation about the exocyclic C=C bond. Even at low temperature (−60 °C, 100 MHz, [D7]DMF), the 13C NMR signal of the two cyano groups was not split into two lines. Apparently, the dipolar resonance structure of 22 a,24 in which the positive charge is delocalized in an aromatic imidazolium ring, is more pronounced than in the case of non‐aromatic heterocycle 19. This assumption might be supported by the fact that 13C NMR chemical shifts of the exocyclic C=C moiety in 22 a (δ=24.5 and 149.3, Δδ=124.8) show a greater difference than those of 19 (Δδ=109.6).

Scheme 4

Reaction of different 2H‐azirines with cyanoform (5); for the key of R1 and R2, see Table 2.

Table 2

Treatment of 21 a–d with 7 to produce 22 a–d.

21/22	R1	R2	Yield of 22 [%][a]
a	Ph	H	58
b	H	Ph	61
c	Ph	Ph	56
d	Ph	Me	62

[a] Isolated yields.

Whereas the formation of 20 is connected with an addition at the C=N bond of N‐protonated 8 and obviously induced by tricyanomethanide, acting as a carbon nucleophile, the genesis of 22 can only be explained by interaction of the protonated 2H‐azirines 21 with tricyanomethanide, which reacts as a nitrogen nucleophile. Since at least two mechanisms may rationalize the 21→22 ring enlargement, we utilized the 15N‐labeled substrate 15N‐21 a,25 which was prepared by thermolysis of the corresponding α‐azidostyrene, and included 49 % of the nitrogen label (Scheme 5). After treatment of 15N‐21 a with 7, we obtained, besides unlabeled 22 a, the products 15N‐22 a and 15N‐22 a′ in a ratio of 11:1. The quantitative analysis of 15N‐22 a was facilitated by the 1H NMR signal of 15NH‐1 at δ=12.50, which indicated 1 J(15N,1H)=97.6 Hz with 1H,1H triplet splitting (J=2.2 Hz), whereas 15N‐22 a′ showed a 1H NMR signal of 15NH‐3 at δ=12.68 with 1 J(15N,1H)=96.0 Hz and additional 1H,1H triplet splitting (J=2.2 Hz). These results were supported by the 15N NMR spectrum, which included the 15NH‐1 signal of 15N‐22 a at δ=−233.6 with 1 J(15N,1H)=97.4 Hz and additional long‐range coupling (dd, J=4.6 and 3.7 Hz) as well as the 15NH‐3 signal of 15N‐22 a′ at δ=−238.9 with 1 J(15N,1H)=96.4 Hz and triplet splitting with 3 J=3.4 Hz. The assignment of all 1H and 15N NMR signals were based not only on the previous allocation of the corresponding NMR signals of 22 a (see above), but also on additional NOE experiments with 15N‐22 a/15N‐22 a′.

Scheme 5

Ring enlargement of 15N‐21 a.

We assume that formation of tricyanomethane (5) from 7 and subsequent N‐protonation of 15N‐21 a with generation of the ion pair 23 are always the first steps in the ring enlargement reaction of 15N‐21 a (Scheme 6). Thereafter, attack of the nitrogen nucleophile tricyanomethanide at the highly activated C=N unit of 23 leads to the addition product 24, which undergoes a ring expansion by a 1,3‐migration process. The resulting imidazole derivative 25 tautomerizes to yield the main product 15N‐22 a. However, a minor pathway with nitrogen‐centered nucleophilic attack of tricyanomethanide at the sp3‐hybridized carbon atom achieves ring opening of 23 with formation of 26; and successional cyclization of this ketenimine intermediate, followed by tautomerism of 25′, generates 15N‐22 a′. The sequence 15N‐21 a→23→26→25′ is similar to the mechanisms, which we suggest for the corresponding ring‐enlargement reactions of epoxides 9, thiirane 14, and aziridines 16 and 18.

Scheme 6

Reaction mechanisms which explain the formation of 15N‐22 a and 15N‐22 a′ from 7 and 15N‐21 a.

In order to confirm the proposed reaction mechanisms of the transformation 7+21 a→22 a, we utilized also the 15N‐labeled azide 15N‐7, which was prepared from (chloromethylidene)malonodinitrile and sodium azide that included 98 % of the isotope label at one of the terminal positions (Scheme 7). Since we used a smaller amount of 15N‐7 and because of the label distribution on the three cyano positions of the resulting cyanoform, the experiment with 15N‐7 was significantly less sensitive than that with 15N‐21 a. As expected, we detected just a small NH signal of 15N‐22 a besides strong signals of 15N‐22′ and NH‐unlabeled 22 a and 15N‐22 a′′ in the 1H NMR spectrum, which was measured after treating 15N‐7 with 21 a. The corresponding 15N NMR spectrum indicated only 15NH‐3 (δ=−238.6, dt, 1 J=96.4 Hz, 3 J=3.4 Hz) of 15N‐22 a′ and C≡15N (δ=−113.3, s) of 15N‐22 a′′. Thus, the transformation of 15N‐7 proved to be a complementary confirmation of the experiment with 15N‐21 a.

Scheme 7

Ring enlargement of 21 a with the help of 15N‐7.

Ring‐expansion reactions of 9, 14, 16, 18, and 21 were performed with the aid of cyanoform precursor 7, which was used under aprotic starting conditions and without competing nucleophiles. Hence, the question arose whether the corresponding ring‐enlargement products can also be prepared if aquoethereal cyanoform (2) or similar reagents are utilized. In order to find an answer to this question, we treated epoxide 9 a with compound 2 and obtained the desired product 13 a in 75 % yield (Scheme 8). However, 13 a was accompanied by a small amount of trans‐cyclohexane‐1,2‐diol, which was simultaneously generated from 9 a owing to the presence of water. Thus, the separation and purification of 13 a was quite tedious if compared to the workup after the transformation 7+9 a→13 a and similar reactions. Consequently, we tried anhydrous acidification of 1 b and added concentrated sulfuric acid to a solution of 1 b and 9 a in 1,2‐dimethoxyethane (DME). After optimization of the reactions conditions, we isolated 13 a by simple washing of the crude product with a limited amount of dichloromethane in 79 % yield. When we exposed the azirine 21 a to the reagent 2, we obtained the wanted product 22 a, but the yield was disappointing low (10 %). By using the procedure with 1 b and concentrated sulfuric acid in DME, we did not get the desired compound 22 a from 21 a at all. Hence, ring enlargement of three‐membered heterocycles with the help of the alternative cyanoform precursor 1 b is possible, but there are limitations in some cases.

Scheme 8

Ring enlargement of 9 a and 21 a with cyanoform generated from tricyanomethanide salts.

Conclusions

In summary, we have demonstrated that tricyanomethane (5), in situ generated by thermal decay of vinyl azide 7 or by acidification of tricyanomethanide salt 1 b, can successfully be used to perform ring‐expansion reactions with three‐membered heterocycles. The resulting products are formed via varying reaction mechanisms; however, tricyanomethanide always acts as a nitrogen nucleophile. This is quite different to the known simple alkylation of this nucleophilic species in the presence of alkyl halides, which leads to 1,1,1‐tricyanoalkanes by attack of a carbon nucleophile.15 It can be argued that different reaction conditions of ring enlargement and alkylation are the reason for distinct nucleophilic properties of tricyanomethanide. But we will show in the near future that tricyanomethane (5), also generated in situ by warming of 7 or by acidification of 1 b, operates as a carbon nucleophile in Michael addition reactions.

Products of ring expansion, such as 13, 15, 17, 19, and 22, include an exocyclic push–pull‐substituted C=C unit, and similar compounds were previously investigated intensively because of their structures, electronic properties, and rotational barriers.24, 26 Moreover, such substances were utilized for organic synthesis, in particular, as precursors of (other) heterocyclic skeletal structures,27 and in some cases, the biological properties were tested, for example, as plant growth regulators28 or for histamine H3 receptor‐binding affinities.29 Several methods are known to prepare these push–pull‐substituted olefinic compounds. In most cases, however, the access required a multi‐step synthesis.19, 21, 30 Thus, the simple cyanoform‐induced transformation of three‐membered heterocycles into ring‐enlargement products, such as 13, 15, 17, 19, and 22, offers a new synthetic approach for the desired push–pull‐substituted alkenes.

Experimental Section

Experimental details, 1H, 13C, 15N NMR spectra, crystal structure data and refinement details are given in the Supporting Information.

Conflict of interest

The authors declare no conflict of interest.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supplementary

Click here for additional data file.