Introduction

Medium-sized ring-containing organic molecules, especially the seven-membered rings, are important structural motifs that are found in drug candidates as well as in bioactive molecules1,2,3,4,5,6,7. However, due to entropic effects and transannular interactions, such frameworks are regarded as difficult structures to access8,9. Compared to the synthesis of five- and six-membered rings, the construction of seven-membered rings can be more challenging through traditional cyclization pathways. Thus, the development of an efficient method to build these seven-membered rings has attracted a significant amount of research attention. Among the numerous methods developed so far, the Büchner-type ring-expansion reaction which has become an effective method for the preparation of seven-membered rings has attracted much attention during the last decade10,11,12,13,14. Traditionally, the Büchner reaction is triggered by the cyclopropanation of the benzene ring to give a norcaradiene, and then the electrocyclic ring opening provides a cycloheptatriene. However, this strategy is hindered by the nature of these α-diazo ketone precursors, which are hazardous, not easily accessible, and potentially explosive. Consequently, the development of catalytic approaches is highly desirable, especially from readily and generally available precursors.

Recently, gold-catalyzed intermolecular alkyne oxidation presumably involving α-oxo gold carbenes has burgeoned, as it avoids the use of difficult and hazardous diazo compounds15,16,17,18,19,20,21,22,23,24,25,26.

In 2010, Zhang et al.27 disclosed an elegant protocol for the gold-catalyzed intermolecular oxidation of alkynes via a reactive α-oxo gold carbene intermediate (Fig. 1). In addition, the Tang group described that rhodium could also catalyze such an intermolecular alkyne oxidation28. Following this notion, numerous efficient synthetic methods have also been disclosed by Hashmi29,30, Liu31,32,33,34, Ye35,36 and others based on this strategy, affording functionalized heterocycles37,38,39,40,41,42,43,44,45,46,47,48,49. Despite these great successes, this intermolecular pathway is evidently not atom-economical due to the fact that the procedure generates a stoichiometric amount of pyridines or quinolines, as waste, which may coordinate and poison metal catalysts. Furthermore, a noble-metal catalyst usually is required, and may severely hamper the practical application of this strategy owing to the high cost and toxicity of the catalyst. In our ongoing program of expanding copper catalysis into alkyne transformation50,51,52,53,54,55,56, we develop a copper-catalyzed alkyne oxidation/Büchner-type ring-expansion sequence, leading to the benzo[6,7]azepino[2,3-b]quinolines and pyridine-based diones. In particular, the pyridine or quinoline partner could be further utilized to construct N-heterocycles in such an oxidative copper catalysis. Most importantly, mechanistic studies and theoretical calculations demonstrate that the reaction presumably proceeds by a Büchner-type pathway, which is distinctively different from the related gold-catalyzed oxidative cyclization. General access to highly valuable seven-membered rings via Büchner-type reaction remains a formidable challenge. Herein, we describe a copper-catalyzed alkyne oxidation/Büchner-type ring-expansion sequence, thus providing practical access to synthetically useful fused seven-membered ring cyclic compounds. Cyclopropanations of heteroarenes are shown in an intermolecular Büchner-type reaction, while circumventing the use of hazardous diazo carbonyl substrates.

Fig. 1: Typical ways for the generation of α-oxo metal carbenes.
figure 1

a Previous work. b Our initial design. M metal.

Results and discussion

Optimization of the reaction conditions

At the outset, alkynone 1a and 8-methylquinoline N-oxide 2a were chosen as model substrates and a range of experiments were executed in order to authenticate our opinion. As documented in Table 1, our initial examination focused on the reaction of the alkynone 1a with 8-methylquinoline N-oxide 2a in DCE at 80 °C in the presence of a copper catalyst. To our pleasure, the expected benzo[6,7]azepino[2,3-b]quinoline 3a was certainly formed in 23% yield, albeit with a lower yield (Table 1, entry 1). The molecular structure of 3a was further confirmed by X-ray diffraction57. Subsequent other copper catalyst screenings indicated that the Cu(hfacac)2 performed obviously better (entries 2–6). Other Lewis acids, including Zn(OTf)2, Y(OTf)3 and Sc(OTf)3, failed to further improve the reaction efficiency (entries 7–9). In addition, the desired 3a was detected in 43% yield when the solvent was changed from DCE to toluene (entry 10). Raising the reaction temperature to 120 °C improved the product yield considerably to 58% (entries 11–12). Doubling of catalyst loading led to an even better yield, affording 3a in 69% yield (entry 13). It should be mentioned that the reaction failed to give even a trace of 3a in the absence of the catalyst (entry 14). Finally, the addition of 4 Å MS led to a slight increase in the yield, forming 3a in 76% yield (entry 15).

Table 1 Optimization of reaction conditionsa.

Substrate scope with different alkynones

With the optimized reaction conditions in hand, we then set out to assess the scope of the reaction by varying alkynones 1. The results are presented in Fig. 2. Alkynones with varied aryl groups (Ar = 4-XC6H4, X = F, Cl, Br, CF3, CN, Ph, Me, Et, tBu and OMe) at the C-4 position were first examined, delivering the corresponding benzo[6,7]azepino[2,3-b]quinolines 3ak in 62–94% yields (entries 1–11). In addition, aryl-substituted alkynones bearing both electron-withdrawing and electron-donating substituents, such as F, Cl and Me, on the phenyl ring were also compatible with this tandem reaction, thus leading to the resulting benzo[6,7]azepino[2,3-b]quinolines 3lr in 50–91% yields (entries 12–18). In particular, the reaction proceeded smoothly with alkynones bearing sterically hindered ortho substituents. The molecular structure of 3r was further confirmed by X-ray diffraction57. To our satisfaction, thiophene, styryl, nBu and cyclopropyl-substituted alkynones were also suitable substrates for this transformation, affording the expected benzo[6,7]azepino[2,3-b]quinolines 3sv in 67–86% yields (entries 19–22). For alkynones bearing different R2 substituents, their desired products 3wz were obtained in 75–82% yields (entries 23–26).

Fig. 2: Reaction scope with different alkynones 1.
figure 2

Reaction conditions: [1] = 0.05 M; yields are those for the isolated products.

Substrate scope with different quinoline N-oxides

We next extended the reaction to a variety of quinoline N-oxides 2 (Fig. 3). Unsubstituted quinoline N-oxide was first performed, giving the corresponding benzo[6,7]azepino[2,3-b]quinoline 3aa in 67% yield (entry 1). The products 3ab and 3ac were also formed in 84 and 62% yields, respectively, when 8-methylquinoline N-oxide was replaced by 8-ethylquinoline N-oxide or 8-isopropylquinoline N-oxide (entries 2–3). Compounds 2eg (N-oxide = 7-chloroquinoline N-oxide, 7-trifluoromethylquinoline N-oxide and 7-methylquinoline N-oxide) were converted smoothly into benzo[6,7]azepino[2,3-b]quinolines 3adaf in 87–95% yields (entries 4–6). This tandem reaction also proceeded for 6-substituted quinoline N-oxides, including substrates with fluoro, chloro, bromo, methyl formate, nitro, methyl, n-butyl, t-butyl and methoxy substituents, and the resulting 3agao were obtained in 58–98% yields (entries 7–15). The related reactions of quinoline N-oxides with additional substitutions at the 5-position and 4-position were either equally or more efficient, affording the expected benzo[6,7]azepino[2,3-b]quinolines 3apar in 67–95% yields (entries 16–18). Accordingly, this approach provided a general and highly efficient strategy for the construction of polycyclic N-heterocycles in organic synthesis. Notable is that the reaction substrates were not only limited to 8-alkylquinoline N-oxides as the oxidants58.

Fig. 3: Reaction scope with different quinoline N-oxides 2.
figure 3

Reaction conditions: [1] = 0.05 M; yields are those for the isolated products.

Reaction scope for the formation of pyridine-based diones

Besides quinoline N-oxides, the reaction also proceeded well with pyridine N-oxides to furnish unexpected pyridine-based diones. Thus, the treatment of alkynones 1 with pyridine N-oxide 4a under copper catalysis furnished the resulting pyridine-based diones 5aq in 60–75% yield (Fig. 4). The molecular structure of 5p was further confirmed by X-ray diffraction57. The reaction presumably involved a copper-catalyzed oxidation-initiated tandem alkyne oxidation/Büchner-type/[1,2]-H shift, and the formation of pyridine-based diones instead of the previous benzo[6,7]azepino[2,3-b]quinolines was attributed to the relatively lower activity (mechanism for the formation of 5a is depicted in Supplementary Information).

Fig. 4: Reaction scope for the formation of pyridine-based diones 5.
figure 4

Reaction conditions: [1] = 0.05 M; yields are those for the isolated products.

To understand the mechanism of these cyclizations, several control experiments were first conducted (Fig. 5). Our attempts to extend the reaction to hydroxyl-substituted alkynone resulted in the formation of 3-(8-methylquinolin-2-yl)-2-phenyl-4H-chromen-4-one 3aba in 86% yield, and no corresponding azepine compound 3aba’ formation was detected. The molecular structure of 3aba was further confirmed by X-ray diffraction12. Furthermore, when unsubstituted alkynone 1ac and 1ad were subjected to this copper-catalyzed oxidative cyclization, 1,3-diones 3aca and 3ada were generated in 88 and 87% yields, respectively. These results showed that the amino group was crucial for the formation of benzo[6,7]azepino[2,3-b]quinolines. Alkynone without Boc at the amino group was then examined, and no corresponding azepine compound 3ae was obtained, likely due to coordinate and poison copper catalysts.

Fig. 5: Control experiments.
figure 5

a Hydroxyl-substituted alkynone was used. b Unsubstituted alkynone was used. c Alkynone without Boc at the amino group was used.

Synthetic application

The synthetic utility of the benzo[6,7]azepino[2,3-b]quinolines was examined (Fig. 6). Firstly, 3a was prepared on a gram scale in 71% yield under the optimized reaction conditions. Subsequently, a selective elimination of benzaldehyde of 3a was achieved with LiOH to furnish benzo[6,7]azepino[2,3-b]quinoline derivatives 6 in almost quantitative yield. In addition, 3a could be transformed into compound 7 bearing two contiguous quaternary carbon stereocenters in 67% yield via a 1,2-Boc shift. Furthermore, 3a could be readily converted into compound 8 in 86% yield by a 1,3-benzoyl-migration. The molecular structure of 8 was further confirmed by X-ray diffraction57.

Fig. 6: Synthetic applications.
figure 6

The synthetic utility of the 3a was examined.

Mechanistic studies

Although a detailed description of mechanistic rationale at present is not possible and deserves further detailed exploration, several control experiments were conducted to gain some further information on potential pathways (Fig. 7). Typical noble-metal catalysts were tested. The direct N-H insertion by the gold carbene in 1a-A was obtained as the main product under gold catalysis conditions. As we considered 3a’ to be possible intermediates in such a tandem sequence, we then subjected 3a’ to the optimal reaction conditions and the formation of 3a was not observed, thus ruling out 3a’ as a potential intermediate for the formation of 3a. Besides, we performed further studies using quinoline as the external nucleophiles. A 1:1 mixture of 8-methylquinoline N-oxide 2a and quinoline 2b’ under the optimized reaction conditions only led to the formation of the corresponding 3a in 72% yield, and no desired 3aa was obtained. Similarly, when quinoline N-oxide 2b and 8-methylquinoline 2a’ was treated under the optimized reaction conditions, only 3aa was obtained in 65% yield. These results indicate that α-oxo copper carbene is not presumably involved in such a copper catalysis.

Fig. 7: Control experiments.
figure 7

a Typical noble-metal catalysts were tested. b A 1:1 mixture of quinoline N-oxide and quinoline was treated.

Proposed mechanism

Based on the above experimental observations and density functional theory (DFT) computations (for details, see the Supplementary Information), a possible mechanism to clarify the formation of 3a is documented. As depicted in Fig. 8, there are two plausible mechanisms to rationalize the formation of 3a. It entails an initial copper activation of alkynone 1a in the form of complex A, followed by a nucleophilic attack by 8-methylquinoline N-oxide 2a via the transition state TS-B1 to furnish the vinyl copper intermediate B1 by overcoming a barrier of 16.5 kcal/mol. In path a, the intramolecular cyclization occurs efficiently to form the five-membered-ring intermediate C1, via transition state TS-C1 with an activation energy of 8.5 kcal/mol, which undergoes Büchner-type reaction to deliver the norcaradiene intermediate D1. It should be mentioned that the stabilization of intermediate C1 can be attributed to the coordination of the carbonyl oxygen to the copper atom according to the calculations. Subsequently, an electrocyclic step opens the cyclopropane ring to provide the seven-membered ring intermediate E1 via TS–E1 with a lower activation energy of 0.2 kcal/mol. Going a step further, intramolecular nucleophilic addition of N-Boc to the imine moiety produces the eventual polycyclic N-heterocycle 3a. The whole process is exothermal by 46.7 kcal mol−1 in free energy. In path b, upon N-O bond cleavage, B1 transforms into α-oxo copper carbene intermediate F1 via transition state TS-F1 with a higher activation energy of 11.2 kcal/mol, and thus the formation of the α-oxo copper carbene F1 is unfavorable. Meanwhile, considering the subsequent Büchner-type reaction, via TS-G1 and via TS-D1A, the activation barriers are 34.6 and 26.0 kcal/mol, respectively. Obviously, path a is much favored kinetically over path b. Besides, path a can rationalize our control experiments in Fig. 7 in which α-oxo copper carbene is not presumably involved in such a copper catalysis.

Fig. 8: Plausible reaction mechanism for the formation of 3a.
figure 8

a α-oxo copper carbene intermediate was not involved. b α-oxo copper carbene intermediate was involved.

Optical properties

Our next efforts concentrated on the exploration of the optical properties of the obtained benzo [6,7]azepino[2,3-b]quinolines (Fig. 9). According to the impact of substituent on the benzene ring, the absorption and emission maxima of these compounds varied from 287 to 432 nm and from 440 to 557 nm, respectively. The absorption was red-shifted by the presence of an electron-withdrawing substituent as in 3e, and the λmax of 3e was bathochromically shifted by 83 nm compared to 3k. A similar effect was detected for the emission spectra of the 3ag, which showed a longer-wavelength emission band at 536 nm and extended the emission to 750 nm. Furthermore, the fluorescence of 3g displayed a considerable red-shift from a fluorescence maximum wavelength of 440–474 nm by conjugation with an increased benzene ring. In addition, the emission wavelengths were further red-shifted to 536 and 498 nm by the introduction of tBu substituent as in 3j and 3an. These results confirmed that the red-shifting absorption and emission band might be achieved by utilizing the strategy of combining the push–pull design, allowing the facile synthesis of near-infrared polycyclic N-heterocycles.

Fig. 9: Optical properties of 3.
figure 9

a Absorption spectra of benzo[6,7]azepino[2,3-b]quinolines in DCM (10 μM). b, c Emission spectra of benzo[6,7]azepino[2,3-b]quinolines in DCM at room temperature (10 μM). d Emission spectra of 3r in solvents of different polarities.

Besides, the effect of the solvent environment on their emission properties was also explored with 3r (Fig. 9d). Interestingly, the 3r displayed negative solvatochromism59,60,61, and the emissions of 3r were blue-shifted from 537 to 423 nm by increasing solvent polarity (the structures of the 3r were optimized and the dipole moment of 3r in the excited and ground states were calculated by Gaussian 09 at the B3LYP/6-31G(d) level). The ground state of 3r was calculated to be slightly more polar (Dipole Moment = 7.28 Debye) than the excited state (Dipole Moment = 6.45 Debye). These results supported the negative solvatochromism of 3r. The solution fluorescence quantum yield (ΦF) was estimated for polycyclic N-heterocycles, as shown in Table 2. The ΦF mainly concentrated between 0.01 and 0.05 in DCM. However, the ΦF of 3r (0.32) was much higher relative to the other compounds, presumably due to the steric effect. The two meta-position methyl groups limited the rotation of the benzene ring, which reduced the intramolecular vibration relaxation and improved the stability of the excited state, thus delivering the higher ΦF62,63.

Table 2 Photophysical data of benzo [6,7]azepino[2,3-b]quinolinesa.

To further explore the structure–property relationship of benzo [6,7]azepino[2,3-b]quinoline derivatives, DFT calculations were performed to obtain the optimized geometries and the frontier orbitals distribution of benzo [6,7]azepino[2,3-b]quinolines64,65. As shown in Fig. 10, due to the non-conjugated and non-planar heterocycle skeleton of benzo [6,7]azepino[2,3-b]quinolines, the Boc and benzoyl groups increased the intramolecular steric hindrance. The HOMO and LUMO levels of these compounds were well separated without significant overlap. The HOMOs of the benzo [6,7]azepino[2,3-b]quinolines mainly distributed on the backbone of the molecule, and the HOMO level and electron cloud distribution could be regulated by introducing different substituents into the backbone. The introduction of electron-donating substituents significantly improved the HOMO levels, such as 3ao (−5.27 eV) to 3aj (−5.85 eV). The LUMOs of benzo [6,7]azepino[2,3-b]quinolines were mainly distributed in the benzoyl groups, and the LUMO levels were regularly decreased by introducing the electron-withdrawing and conjugate groups. These results suggested that the luminescence color and band gap of the compounds could be effectively adjusted, which could have great potential applications in biological imaging and optoelectronic devices.

Fig. 10: Density functional theory calculations.
figure 10

Molecular orbital and energy levels of benzo [6,7]azepino[2,3-b]quinolines, calculated at the B3LYP/6-31G(d) level.

Conclusions

In conclusion, we have described a copper-catalyzed dearomative annulation between alkynones and quinoline N-oxides, delivering the practical and efficient synthesis of fused polycyclic N-heterocycles. This strategy provides the first example of non-noble-metal catalyzed intermolecular alkyne oxidation via an atom-economical and environmentally friendly pathway where the quinoline partner could be further utilized to construct N-heterocycles instead of the previously reported the departure of a stoichiometric amount of quinolines as waste. Cyclopropanations of heteroarenes are shown in an intermolecular Büchner-type reaction, while circumventing the use of hazardous diazo carbonyl substrates. Of note, mechanistic studies revealed that this copper-catalyzed alkyne oxidation presumably proceeds by a Büchner-type ring-expansion pathway, which is distinctively different from the related gold catalysis. Moreover, such oxidation of alkynes can afford the divergent synthesis of pyridine-based diones with pyridine N-oxides as oxidants. Work to exploit enantioselective variants and acquire a deeper mechanistic understanding is underway.

Methods

Materials

Unless otherwise noted, materials were obtained commercially and used without further purification. All the solvents were treated according to general methods. Flash column chromatography was performed over silica gel (300–400 mesh). See Supplementary methods for experimental details.

General methods

1H NMR spectra were recorded on a Bruker AV-400 spectrometer in chloroform-d3. Chemical shifts are reported in ppm with the internal TMS signal at 0.0 ppm as a standard. The data is being reported as (s = singlet, d = doublet, t = triplet, m = multiplet or unresolved, brs = broad singlet, coupling constant(s) in Hz, integration). 13C NMR spectra were recorded on a Bruker AV-400 spectrometer in chloroform-d3. Chemical shifts are reported in ppm with the internal chloroform signal at 77.0 ppm as a standard. Infrared spectra were recorded on a Nicolet iS 10 spectrometer as thin film and are reported in reciprocal centimeter (cm−1). Mass spectra were recorded with Micromass Q-Exactive Focus mass spectrometer using electron spray ionization. 1H NMR, and 13C NMR are supplied for all compounds: see Supplementary Figs. 168. More mechanism studies are supplied: see Supplementary Figs. 6974. Representative synthetic procedures for the preparation of alkynones are supplied: see Supplementary Fig. 75. General procedure for the synthesis of benzo[6,7]azepino[2,3-b]quinolines 3 is supplied: see Supplementary Fig. 76. General procedure for the synthesis of pyridine-based diones 5 are supplied: see Supplementary Fig. 77. Synthetic applications are supplied: see Supplementary Figs. 7880. Crystal data are supplied: see Supplementary Tables 15. TD-DFT computational data are supplied: see Supplementary Tables 615. See Supplementary methods for the characterization data of compounds not listed in this part.

General procedure for the synthesis of benzo[6,7]azepino[2,3-b]quinolines 3

Quinoline N-oxide 2 (0.4 mmol), 4 Å molecular sieves (40 mg) and Cu(hfacac)2 (0.04 mmol, 19.1 mg) were added in this order to the alkynones 1 (0.2 mmol) in toluene (4.0 mL) at room temperature. The reaction mixture was stirred at 120 °C (120 °C, heating mantle temperature) and the progress of the reaction was monitored by TLC. Upon completion, the mixture was then concentrated and the residue was purified by chromatography on silica gel (eluent: petroleum ether/ethyl acetate) to afford the desired product 3.

General procedure for the synthesis of pyridine-based diones 5

2-Chloropyridine N-oxide 4a (0.4 mmol, 51.8 mg) and Cu(hfacac)2 (0.04 mmol, 19.1 mg) were added in this order to the alkynones 1 (0.2 mmol) in toluene (4.0 mL) at room temperature. The reaction mixture was stirred at 120 °C (120 °C, heating mantle temperature) and the progress of the reaction was monitored by TLC. Upon completion, the mixture was then concentrated and the residue was purified by chromatography on silica gel (eluent: petroleum ether/ethyl acetate) to afford the desired product 5.