Catalytic Synthesis of 1H-2-Benzoxocins: Cobalt(III)-Carbene Radical Approach to 8-Membered Heterocyclic Enol Ethers.

The metallo-radical activation of ortho-allylcarbonyl-aryl N-arylsulfonylhydrazones with the paramagnetic cobalt(II) porphyrin catalyst [CoII(TPP)] (TPP = tetraphenylporphyrin) provides an efficient and powerful method for the synthesis of novel 8-membered heterocyclic enol ethers. The synthetic protocol is versatile and practical and enables the synthesis of a wide range of unique 1H-2-benzoxocins in high yields. The catalytic cyclization reactions proceed with excellent chemoselectivities, have a high functional group tolerance, and provide several opportunities for the synthesis of new bioactive compounds. The reactions are shown to proceed via cobalt(III)-carbene radical intermediates, which are involved in intramolecular hydrogen transfer (HAT) from the allylic position to the carbene radical, followed by a near-barrierless radical rebound step in the coordination sphere of cobalt. The proposed mechanism is supported by experimental observations, density functional theory (DFT) calculations, and spin trapping experiments.

Introduction

Medium-sized O-heterocycles are important substructures found in several natural compounds with interesting bioactivity toward the cardiovascular and neurological systems of humans (Figure ).[1,2] Screening a library of related compounds is therefore expected to uncover a range of future pharmaceutical applications. However, such compounds are difficult to synthesize, and as a direct consequence they are largely underrepresented in libraries currently used to screen for bioactivity. Having access to a broad pallet of synthetic possibilities toward these compounds would therefore be helpful in addressing challenges faced in drug discovery, and in particular, novel methods to prepare 8-membered O-heterocycles are in high demand. Developing new synthetic approaches is therefore important. This is not an easy task though, because in contrast to the fruitful synthetic approaches to 5- or 6-membered heterocyclic ring structures, developing synthetic protocols for constructing medium-sized (hetero)cycles is challenged by entropic factors and inherent transannular interactions increasing the transition state barriers. Accordingly, in the synthesis of medium-sized O-heterocyclic ring compounds, the applied flexible, linear precursors are generally more amenable to reacting in an intermolecular way, leading to low yields and formation of substantial amounts of unwanted dimerized or polymerized side products. Available protocols for constructing medium-sized ring ethers include ring-closing metathesis,[3] lactone methylenation,[4] reductive cyclization of hydroxyl ketones,[5] Barbier coupling,[6] thermal rearrangement of oxabicyclo[4.2.0] compounds[7] or acyloxybenzocyclbutenes,[8] Giese radical addition reactions,[9] and ring expansion by Claisen rearrangements.[10] Most of these methodologies require prefunctionalized precursors (such as terminal alkenes in ring-closing metathesis) or multistep synthesis, or they are restricted to specific substrate classes. Additional limitations further confine the applicability of these strategies, such as harsh reaction conditions, unavoidable byproduct generation, and the necessity of noble metal catalysts. Therefore, developing new, efficient strategies to synthesize 8-membered O-heterocycles remains an important task.

Figure 1

Bioactive 8-membered ether rings.

In recent years, catalytic radical-type transformations have proven to be attractive methods in the construction of cyclic products,[11] and one outstanding example is the application of Co(II)-based metallo-radical catalysis in ring-closing synthesis. In particular, square-planar cobalt-porphyrin complexes with well-defined open-shell doublet (S = 1/2) low-spin d7-electronic configuration display a remarkable radical-type reactivity.[12] In carbene transfer reactions mediated by these systems, the cobalt(II) complex catalyst first reacts with a carbene precursor[13] (such as a diazo compound or a N-phenyl-sulfonohydrazone, Figure B), and then transforms to a cobalt(III)-carbene radical intermediate by an intramolecular metal-to-substrate single-electron transfer from cobalt(II) to the carbene carbon atom. The carbene carbon atom in cobalt(III)-carbene radical intermediate is a carbon-centered radical involved in subsequent radical-type reaction pathways. This metallo-radical strategy has been successfully employed in the synthesis of ring structures, such as cyclopropanes,[14] furans,[15] indenes,[16] indolines,[17] ketenes,[18] dihydronaphthalenes,[19] piperidines,[20] and pyrrolidines.[21] Furthermore, an efficient synthetic protocol for the synthesis of 2H-chromenes has been developed (Figure A),[22] and we also successfully applied the intrinsic radical-type reactivity of cobalt(III)-carbene radicals to construct two types of new 8-membered carbocyclic compounds: dibenzocyclooctenes[23] and monobenzocyclooctadienes[24] (Figure A). Building on these prior results, and in view of our continued interest in cobalt(II)-based metallo-radical catalysis, we envisioned that related open-shell catalysis could also uncover new possibilities for the synthesis of 8-membered O-heterocyclic compounds (Figure C).

Figure 2

Design of 8-membered ring ether synthesis by metallo-radical catalysis. A. Previous work in our group utilizing metallo-radical catalysis for the synthesis of 2H-chromenes, dibenzocyclooctenes, and monobenzocyclooctadienes. B. In situ generation of cobalt(III)-carbene radical intermediates from aldehydes. C. Present work utilizing cobalt(II) porphyrin ([CoII(TPP), TPP = tetraphenylporphyrin]) catalysis to construct 1H-2-benzoxocins via cobalt(III) carbene radicals.

Herein, we describe a highly effective metallo-radical approach to construct 8-membered oxygen heterocycles via cobalt(III)-carbene radical mediated cyclization. This protocol affords a class of unprecedented 1H-2-benzoxocins. DFT calculations suggest that the cyclization to 1H-2-benzoxocins proceeds via a unique reaction sequence involving a 1,6-HAT followed by a radical-rebound process leading to C–O bond formation directly in the coordination sphere of cobalt. This new methodology produces the novel heterocyclic products efficiently, under mild reaction conditions, and exhibits a broad substrate scope and high functional group tolerance. In addition, this reaction is easy to scale up and the O-heterocyclic products can be functionalized with various bioactive substituents.

Results and Discussion

Inspired by previous work on metallo-radical catalysis in cyclization reactions, we continued our research by seeking a new synthetic approach to 8-membered O-heterocycles. As such, we designed a substrate with an allylcarbonyl moiety, and began our study by using 1 as the model substrate to probe the feasibility of ring-closure to the desired 8-membered ring compound 1a (Table ). At first, we tested the reaction using LiOBu as the base and p-TsNHNH2 to prepare the hydrazone from the aldehyde in situ. The reaction was performed at 60 °C in one pot (Table , entry 1). However, under these conditions we obtained only a trace amount of the desired cyclic product 1a, while the aromatic 6-membered ring product 1a′ was obtained as the main product instead. This result was unexpected, and formation of naphthalene 1a′ seems to be the result of an uncatalyzed intramolecular aldol condensation reaction of aldehyde substrate 1, producing the aromatic product upon elimination of water (Scheme , red pathway).[25] Indeed, a control reaction in the absence of the [Co(TPP)] catalyst also produced 1a′ in 95% yield (see Figure ). Under the catalytic conditions, this side reaction seems to take place before 1 has the chance to react properly with the arylsulfonyl hydrazide to produce the diazo compound (Scheme , blue pathway), thus preventing the desired metallo-radical cyclization to 1a.

Table 1

Optimization of the Reaction Conditions

entrya	T [°C]	base	solvent	–Ar	yield of 1ab (%)	1a/1a′
1c	60	LiOtBu	benzene	–4-MeC6H4	trace	-
2	60	LiOtBu	benzene	–4-MeC6H4	42	1/1
3	r.t.	LiOtBu	benzene	–4-MeC6H4	55	1.3/1
4	r.t.	Cs2CO3	benzene	–4-MeC6H4	74	5/1
5	r.t.	KOtBu	benzene	–4-MeC6H4	57	4/1
6	r.t.	NaOMe	benzene	–4-MeC6H4	9	0.8/1
7	r.t.	Cs2CO3	chlorobenzene	–4-MeC6H4	67	4/1
8	r.t.	Cs2CO3	1,2-dichlorobenzene	–4-MeC6H4	71	5/1
9	r.t.	Cs2CO3	benzene	–2,4,6-iPrC6H4	83	6.5/1
10	r.t.	Cs2CO3	benzene	–4-OMeC6H4	84	6.2/1
11	r.t.	Cs2CO3	benzene	–3-NO2C6H4	65	5/1
12	r.t.	Cs2CO3	benzene	–2-ClC6H4	75	5/1
13d	r.t.	Cs2CO3	benzene	–4-OMeC6H4	36	2/1
14e	r.t.	Cs2CO3	benzene	–4-OMeC6H4	42	2/1.5

Reaction conditions: Substrate 1 (0.1 mmol, 1.0 equiv) and the arylsulfonyl hydrazide (0.105 mmol, 1.05 equiv) were mixed in methanol (1.0 mL), and stirred for 15 min at room temperature. The thus-obtained crude hydrazone (= carbene precursor) (0.1 mmol) was mixed with [Co(TPP)] (0.005 mmol, 0.05 equiv) and base (0.11 mmol, 1.1 equiv) in benzene (1.0 mL), and stirred at room temperature for 20 h.

Yields were determined by integration of the 1H NMR signals in the presence of dimethyl sulfone as internal standard.

Substrate 1 (0.1 mmol, 1.0 equiv), the arylsulfonyl hydrazide (0.105 mmol, 1.05 equiv), [Co(TPP)] (0.005 mmol, 0.05 equiv) and base (0.11 mmol, 1.1 equiv) were directly mixed in benzene (1.0 mL) in a one-pot reaction, and the mixture was stirred at 60 °C for 20 h.

[Co(ppIX)] (0.003 mmol, 0.03 equiv) and Aliquat 336 (0.015 mmol, 0.15 equiv) were used instead of Co(TPP).

[Co(ppIX-OMe)] (0.003 mmol, 0.03 equiv) and 4-DMAP (0.005 mmol, 0.05 equiv) were used instead of [Co(TPP)].

Scheme 1

Desired Formation of 8-Membered O-Heterocycle 1a via Hydrazone 1′ and Diazo Compound 1″ (blue) versus Undesired Formation of Naphthalene 1a′ via an Aldol Condensation Pathway (Red)

Figure 5

Mechanistic studies. A. Control experiment. B. Spin trapping of the cobalt(III)-carbene radical intermediate by thiol, BHT, and BQ. C. Spin trapping of the cobalt(III)-carbene radical by PBN, detected with EPR spectroscopy.

To constrain this side reaction, we therefore adjusted the experimental procedure by preparing the hydrazone 1′ first, in the absence of added base, and then using the crude hydrazone as a carbene precursor for the catalytic reaction (Scheme ). To our delight, under these conditions the desired 8-membered enol ether ring-product could be obtained in a much higher yield (Table , entry 2).

Lowering the reaction temperature increased the selectivity for the desired product further (Table , entry 3). After screening various bases and solvents, Cs2CO3 was found to be the most suitable base to trigger the reaction (Table , entries 4–6). Changing to solvents with a somewhat higher polarity proved to have only a small influence on the yield of 1a (Table , entries 4, 7, and 8). Further evaluation using differently substituted arylsulfonyl hydrazides revealed that p-methoxy-benzenesulfonyl hydrazide and 2,4,6-triisopropylbenzenesulfonyl hydrazide work best (Table , entries 9–12). To explore the importance of the meso-substituents on the porphyrin ligand we also tested the bioderived [CoII(ppIX)] and [CoII(ppIX-OMe)] catalysts (ppIX = protoporphyrin IX; ppIX-OMe = methyl ester of ppIX) lacking meso-substituents (Table , entries 13 and 14). These catalysts also produce the desired 8-membered product 1a, albeit in a moderate yield. We ascribe the lower yields to self-aggregation of these catalysts in solution. Nonetheless, meso-substituents on the porphyrin ring are clearly not essential for the reaction. Interestingly, the 6-membered dihydronaphthalene ring product 2a was not observed in the whole series of optimization reactions, while its formation by carbene insertion into the allylic C–H bond was expected to be feasible and competitive, based on both thermodynamic considerations and an expected easier cyclization to 6-membered rings than to 8-membered rings (for entropic reasons and due to transannular interactions). Formation of 1a in high yields is in fact not trivial at all, as besides formation of 1a′ instead of the expected 6-membered ring product 2a, the instability of the substrate under both acidic and basic conditions (caused by the reactive α-H position of the allylcarbonyl moiety) could have led to a variety of other unwanted side reactions. Ring-closure reactions involving (formal) carbene addition to a ketone group are rare in general,[26] and ring-closure of a carbene precursor onto a ketone group to produce an 8-membered O-heterocycle is unprecedented.

With the optimal reaction conditions being established, we started to evaluate the generality of the new cyclization protocol. First, different substituents on the R1 = aryl moiety were evaluated. Pleasingly, these substrates were compatible with this reaction as well and afforded a broad range of new 8-membered O-heterocyclic products (Figure A). It seems that substrates with electron-donating substituents on the R1 = aryl ring result in a somewhat higher yield of the desired 8-membered products (Figure A: 2a) than substrates substituted with electron-withdrawing groups (Figure A: 3a–9a).[27]

Figure 3

Substrate scope for formation of 1H-2-benzoxocins. Standard reaction conditions: the substrates (0.1 mmol, 1.0 equiv), p-methoxybenzenesulfonyl hydrazide (0.105 mmol, 1.05 equiv) were mixed in methanol (1.0 mL) and stirred for 15 min at room temperature; The thus obtained crude hydrazones (= carbene precursors) (0.1 mmol) were mixed with [Co(TPP)] (0.005 mmol, 0.05 equiv) and Cs2CO3 (0.11 mmol, 1.1 equiv) in benzene (1.0 mL), and stirred at room temperature for 20 h. Isolated yields are shown. For 23a, 24a, 26a, and 27a, the isolated yields corrected for the E/Z ratio of the substrates are shown between parentheses.

No obvious influence of the position of the substituents on the yields was observed. Substrates with bulkier groups at the aromatic R1 position seem to reduce the chemoselectivity slightly (Figure A: 1a, 10a, and 11a, yield drops from 84% to around 60%), and in those cases a bit more of the aromatic aldol side products were formed, thus lowering the yield of the desired 8-membered ring products. We also noticed that some of the 8-membered enol ether ring products are rather unstable. This is the case for 12a, which decomposes rapidly, and as a result this compound was always obtained as a mixture containing a small amount of unidentifiable decomposition products, which is possibly due to the increased electron density of the enol moiety. However, in general the protocol works well for various substrates with different R1 = aryl moieties and affords the 8-membered ring products in yields between 50% and 88%.

We continued our investigation of the substrate scope by changing the aryl groups adjacent to the carbonyl moiety to aliphatic groups (Figure B). As the carbonyl moiety is directly involved in the cyclization process, changing substituents on the carbonyl from aryl to alkyl could affect the geometry and electron distribution of the intermediates and transition states, and hence could in principle have a substantial influence on the outcome of the reaction. A change in mechanism could even lead to a switch of the preferred reaction pathway, for example, leading to preferred formation of 6-membered ring products such as dihydronaphthalenes. However, formation of 8-membered cyclic products was still the preferred pathway for these substrates, and the selectivity over 6-membered naphthalene ring formation proved to be even better for R1 = alkyl than for R1 = aryl substrates. Almost no aldol cyclization products could be detected in these reactions, and the desired 8-membered heterocycles were obtained in near-quantitative yield. Even bulky alkyl substituents do not hinder the reaction (Figure B: 15a–19a). Notably, the substrates containing cyclopropyl and cyclohexyl moieties also smoothly reacted to produce the desired 8-membered enol ethers in high yield, and no cyclopropyl or cyclohexyl ring-opened products were observed (Figure B: 17a and 18a).

To further expand the scope, substrates with different ester groups were synthesized and tested. Gratifyingly, changing the ester groups has little effect on the reactivity, and all desired 8-membered cyclic products were obtained in good yields (Figure C). We continued to screen the substrate scope of the reaction by checking the feasibility of different R3 substituents at the phenyl ring of the benzaldehyde moiety. Various R3 substituents also proved to be well-tolerated, and products 23a–28a were obtained in yields between 80% and 92% (Figure D).

Note that most of the results listed in Figure were obtained using pure E-isomers of the substrates. Only a small amount of Z-isomer was found in substrates 23, 24, 26, and 27. The isolated yields of 23, 24, 26, and 27 corrected for the E/Z ratio of these substrates are shown between parentheses in Figure because only the E-isomer can convert to the final 8-membered ring.[23,24]

The broad substrate scope demonstrated in Figure reveals an exceptional functional group tolerance and high efficiency toward formation of medium-sized ring structures, which offers several possibilities to prepare 8-membered O-heterocycles containing biologically relevant substituents. Hence, to further explore the practicability of the protocol, we employed the cobalt-catalyzed cyclization reaction to synthesize the 1H-2-benzoxocins functionalized with different bioactive and druglike substituents shown in Figure A. The compounds in Figure A were obtained in good yields, using the aforementioned standard reaction conditions. Notably, some of the substrates used in these reactions contain functional groups that are known to be sensitive to intra- or intermolecular radical attack, but the desired 8-membered ring products were still obtained in high yields, which demonstrates the excellent chemoselectivity toward formation of 8-membered O-heterocyclic enol ethers.

Figure 4

Synthetic practicality and applications of 1H-2-benzoxocins. A. Modification of pharmaceutical derivatives and natural products. B. Larger-scale reaction. C. Molecular structure of 27a (ORTEP diagram with 50% probability ellipsoids), as determined by single crystal X-ray diffraction. Selected bond distances [Å] for 27a: C1–C2 1.502(4); C2–C3 1.401(4); C3–C4 1.468(3); C4–C5 1.348(4); C5–C6 1.474(5); C6–C7 1.335(4); C7–O1 1.367(3); C1–O1 1.440(4). D. Synthetic transformation of 1H-2-benzoxocins. Start from 13a or 19a. (a) NaOH (2.0 equiv); H2O/MeOH; 40 °C, 8 h. (b) Diisobutylaluminum hydride (3.0 equiv); DCM; −78 °C to room temperature; 2 h. (c) Methylmagnesium bromide (2.0 equiv); THF; 0 °C; 1 h. (d) H2 (1 bar); Rh/Al2O3; MeOH; room temperature; 20 h. (e) Toluene; 120 °C, 3 d. (f) Al2O3 (10.0 equiv); NaOCl (3.0 equiv); MeCN, 0 °C; 2 h.

To demonstrate that larger quantities can be synthesized with the newly developed protocol, we also tested the reaction on a near-gram scale (Figure B), showing that the 8-membered O-heterocycle 1a is still obtained in a high yield even when using a lower catalyst loading of 1 mol % (no other changes in reaction conditions).

The molecular structure of 1H-2-benzoxocin 27a was confirmed by single crystal X-ray diffraction (Figure C). Single crystals of 27a were harvested upon solvent evaporation of an NMR sample to dryness after chromatography. Compound 27a crystallized in the monoclinic space group P21/n. Notably, the C7–O1 bond length (1.367(3) Å) in the newly generated heterocycle is much shorter than that of C1–O1 (1.440(4) Å) in 27a, showing that C7 is sp2 hybridized, while C7–O1 may also feature some double-bond character.

Considering that the products generated in this reaction contain unusual 8-membered ring structures with a conjugated acrylate moiety and special enol ring motif, we argued that these compounds are likely to be interesting substrates for further transformations. As such, we explored the reactivity of compounds 19a and 13a as model substrates, showing a broad reactivity pattern (Figure D). 19a could be efficiently hydrolyzed under mild conditions to produce the carboxylic acid product 19aa. Reduction of the ester moiety of 13a with DIBAL-H produced the allylic alcohol 13ab in high yield, and reaction of 19a with the Grignard reagent MeMgBr efficiently produced the corresponding substituted alcohol 19ac. Note that both the DIBAL-H reduction and Grignard reaction are highly chemoselective and do not affect the alkene and enol motifs. 19a could also be hydrogenated by a heterogeneous Rh catalyst, producing the fully hydrogenation product 19ad. Interestingly, heating the 8-membered enol ring compound 19a to 120 °C for 3 days converted the 8-membered ring compound to the thermodynamically more stable 6-membered ring product, namely, dihydronaphthalene 19ae. We ascribe formation of 19ae to thermal ring opening of 19a to the corresponding ortho-quinodimethane (o-QDM) compound, followed by cyclization to 19ae (Scheme in reserve). Note that substituted dihydronaphthalenes are key motifs in many bioactive compounds, but are in fact also difficult to construct via existing organic synthetic methods.[28] Interestingly, reaction of 19a with sodium hypochlorite in the presence of neutral alumina also leads to a selective ring-contraction, but in this case to the 5-membered ring product 1,3-dihydroisobenzofuran-3-oxobutanoate 19af. The exact mechanism of this reaction is presently unclear, but it should be noted that 1,3-dihydroisobenzofuran-3-oxobutanoates are important bioactive substructures found in the vaccinol family of compounds (vaccinol H and vaccinol I),[29] and 19af is also an analogue of emefuranone and emefuran, which are bioactive secondary metabolites of a diverse genus of filamentous fungi that have potent antimicrobial activity.[30] Until now, no organic synthetic methods to prepare these type of compounds have been reported, and this is in fact the first reported efficient laboratory route to 1,3-dihydroisobenzofuran-3-oxobutanoates. These proof-of-principle transformations demonstrate that the 8-membered enol ether ring structures provide powerful platforms to synthesize a variety of potentially bioactive compounds. While screening for biological activity is beyond the scope of the present paper, it could well be worthwhile to test these compounds, as well as the 1H-2-benzoxocins, for potential pharmaceutical applications in follow-up studies. Further synthetic applications are ongoing in our lab.

Scheme 3

(A) DFT Computed Mechanism for the Formation of Thermodynamically Favored 1,2-Dihydronaphthalene (2a) versus the Kinetically Favored 1H-2-Benzoxocin (1a) from Allyl Radical Intermediate D (BP86, def2-TZVP, m4 grid, disp3). Gibbs Free Energies (ΔG°298 K in kcal mol–1) Relative to the Energy of Intermediate D. (B) DFT Computed Pathway for Formation of Allyl Radical Intermediate D via a Keto-Enol Mechanism Producing Intermediate C_enol, Followed by HAT via TS8 (BP86, def2-TZVP, m4 grid, disp3). Gibbs Free Energies (ΔG°298 K in kcal mol–1) Relative to the Energy of Intermediate C. The Ellipse Represents the Porphyrin Ligand (Tetraphenylporphyrin). Full Atom Models.

Mechanistic Investigations

Several control experiments were performed to obtain a better understanding of the reaction mechanism (Figure ). The base is indispensable to trigger the desired catalytic reaction (Figure A, a), but direct contact between the base and the aldehyde in the reaction mixture is also the principal reason for the formation of the aldol side product (Figure A, b and c). Formation of the desired 8-membered enol ether ring product was not observed in the absence of hydrazide (Figure A, c), and using the preformed hydrazone as the carbene precursor (instead of mixing aldehyde + hydrazide) strongly inhibits the unwanted aldol side reaction (Figure A, b and d). Due to the reversibility of hydrazone–aldehyde equilibrium, a small amount of intramolecular aldol cyclization product still occurs, depending on the structure of the substrates (Figure A, b and d).

The cobalt catalyst is not involved in the aldol side reaction (Figure A, b, c, and d), and it also does not activate the aldehyde precursor directly (Figure A, c). Only after deprotonation of the hydrazone by the base, leading to formation of the necessary diazo compound, the cobalt catalyst starts to unveil its unique metallo-radical catalytic trait resulting in formation of the desired medium-sized O-heterocyclic products.

We note here that despite [FeII(ppIX)] (FeII heme) being known as an active carbene transfer catalyst using diazo compounds,[31] it is ineffective in these reactions (Figure A, f). Replacing [CoII(TPP)] by FeII heme in these reactions (in situ generated by reduction of hemin) leads to the substrate being consumed, but no desired 8-membered ether ring is formed (see Supporting Information for more details), showing that the unique carbene radical species generated at the cobalt(II) catalyst are essential.

In order to confirm the generation of radical species in the pathway, we set a series of radical-trapping experiments with thiol (t-butyl thiol), BHT (butylated hydroxytoluene), and BQ (1,4-benzoquinone), and all these experiments showed obvious reaction inhibition (Figure B). Notably, the thiol-trapped intermediate was observed by high-resolution mass spectrometry (HRMS) analysis (see SI for more information).

In an attempt to get more information about the carbene radical intermediate, spin trapping reagent PBN (N-t-butyl-α-phenylnitrone) was used to trap the radical intermediate, and the trapped species was detected with X-band electron paramagnetic resonance (EPR) spectroscopy (Figure C). The EPR spectrum displayed an isotropic signal with a triplet of doublets hyperfine pattern (g = 2.00604, A = 40.46 MHz, A = 7.01 MHz), which is characteristic for a PBN-trapped carbon-centered radical, in agreement with formation of Co(III) radical species in the reaction pathway. Note that the PBN fragment in Figure C is drawn attached to the carbene carbon (species C trapped by PBN), but it could also be attached to the allylic position (species D trapped by PBN), see Supporting Information.

DFT calculations were carried out to gain additional information about the mechanism. Calculations were performed at the BP86/def2-TZVP level of theory, using Grimme’s D3 dispersion corrections (“zero” damping), at the doublet spin surface (see Supporting Information for details). This method has been properly benchmarked against experimental data for cobalt(II)-porphyrin systems.[16−20,22−24,32] The proposed mechanism of 1H-2-benzoxocin generation based on our computational studies and experimental observations is shown in Scheme .

Scheme 2

Proposed Mechanism for the [Co(porphyrin)]-Catalyzed Formation of 1H-2-Benzoxocins, Based on DFT Calculations (BP86, def2-TZVP, m4 grid, disp3)

All Gibbs free energies (ΔG°298 K in kcal mol–1), including those of TS1–TS3, are reported relative to the energy of intermediate A. Full atom models.

Since the experimental reaction mixture contains several donor atoms that could potentially bind to cobalt under the catalytic reaction conditions, we explored the DFT mechanism without (green lines in Scheme ) and with an axial donor bound to cobalt at the sixth coordination site (purple lines). We took NH3 as a model axial ligand, which is a reasonably strong donor ligand and should be representative for axial ligand donor effects of several potential donor ligands (e.g., tosyl hydrazine, MeOH). As is clear from Scheme , axial ligand coordination has only a small influence on the reaction profile, and differences in the relative energy barriers with and without NH3 binding are almost negligible (Scheme ). The same is true for the effect of Ph-substituents at the meso-position of the porphyrin ligand (for the DFT calculated energy profile of the reactions using a catalyst without meso-Ph-substituents, see Supporting Information). This is confirmed by the experimental results, which also showed that phenyl substituents on the porphyrin of the catalyst are not essential (Table , entries 13 and 14, using [Co(ppIX)] and [Co(ppIX-OMe)] as the catalysts).

First, the hydrazone precursor converts to the corresponding diazo compound in situ, which is triggered by base in a noncatalyzed pathway. The catalytic cycle starts by trapping of the diazo compound by the cobalt(II) catalyst to produce intermediate B. Subsequently, intermediate B converts to cobalt(III) carbene radical intermediate C by irreversible loss of N2, and simultaneous intramolecular single electron transfer from cobalt(II) to carbon, which is an exergonic pathway with a low energy barrier (without L: +10.9 kcal mol–1; +12.9 kcal mol–1 for L = NH3). In the next step, the carbene radical carbon atom abstracts a hydrogen atom from the allylic position of C via an intramolecular 1,6-HAT process in TS2 to produce the delocalized allyl radical intermediate D. This is again a process with a low barrier (without L: +9.2 kcal mol–1; +10.7 kcal mol–1 for L = NH3). After generation of intermediate D, the carbonyl oxygen atom bearing part of the delocalized radical attacks the carbon bound to cobalt in TS3, leading to formation of a new C–O bond with simultaneous homolysis of the Co–C bond to produce the desired final product 1a. This step is again exergonic, has a low energy barrier (without L: +7.7 kcal mol–1; +6.4 kcal mol–1 for L = NH3) and regenerates the cobalt(II) catalyst. The ring-closure step in TS3 could perhaps also be described as a concerted process involving simultaneous homolysis of the Co–C bond and immediate 8π-cyclization of the thus generated o-QDM intermediate within the coordination sphere of cobalt (see below).

To gain more information about the mechanism, in particular, to find alternative pathways to generate 1H-2-benzoxocins, and to explain why 1H-2-benzoxocins (8-membered rings) are formed instead of the thermodynamically more stable 1,2-dihydronaphthalenes (6-membered rings), we performed additional DFT calculations. In Scheme A, we evaluated alternative pathways to form 1H-2-benzoxocins in which the substrate first fully dissociates from the cobalt catalyst, generating a free o-QDM intermediate E_2 before the ring-closure step (Scheme A, dashed blue arrows). Subsequently, the free o-QDM intermediate converts to the final product 1a via a separate 8π-cyclization step leading rearomatization. Dissociation of o-QDM intermediate E from D is endergonic (without L: +12.1 kcal mol–1; +9.9 kcal mol–1 for L = NH3; i.e., more endergonic than the barrier of TS3), while the 8π-cyclization step is essentially barrierless after bond rotation in E (the corresponding rotamer of E simply collapses to 1a).[33] The computed mechanisms and energies for 1,2-dihydronaphthalene (2a) and 1H-2-benzoxocin (1a) formation are also compared. Hypothetical formation of 1,2-dihydronaphthalene 2a could occur via different pathways. A direct radical rebound over TS5 from intermediate D would produce 2a in one step (without L: +10.6 kcal mol–1; +10.9 kcal mol–1 for L = NH3), but 2a could also be formed via a 6π-cyclization process in which o-QDM intermediates E_1 (without L: +14.8 kcal mol–1; +12.7 kcal mol–1 for L = NH3) or E_2 (without L: +12.1 kcal mol–1; +9.9 kcal mol–1 for L = NH3) are generated in endergonic processes, respectively, which then convert to 2a over TS6 (+27.3 kcal mol–1) or TS7 (+6.5 kcal mol–1).

As is clear from Scheme , all these processes (red pathways) have higher barriers than the pathways leading to 1H-2-benzoxocins 1a (blue pathways). Formation of the 8-membered enol ether ring products must therefore be kinetically controlled, and the DFT calculations explain the high chemoselectivity for formation of 1H-2-benzoxocins without detectable amounts of the more stable 6-membered dihydronaphthalenes in the experimental reactions (DFT predicted selectivity for 8- versus 6-membered ring formation based on the Eyring equation: kTS3/kTS5 = 134 without L; kTS3/kTS5 = 1680 for L = NH3). We believe that the kinetic preference for 8-membered ring formation over 6-membered ring formation in these reactions is at least partially caused by the fact that the aromatic ring and adjacent double bond prevent adaptation of stabilizing boat or chair conformations in the 6-membered transition states TS5 and TS6.

A reaction similar to the computed thermal endergonic ring opening of 1a to form o-QDM compound E_2 (+21.4 kcal mol–1) via TS4, followed by thermal ring closure of E_2 to 2a via TS7, readily explains the experimentally observed thermal ring contraction of 8-membered ring compound 19a to form the 6-membered dihydronapthalene ring 19ae upon prolonged heating (Figure D, reaction e).

Hypothetically, formation of 1a could have also proceeded via a keto–enol tautomerization pathway, generating enol compound C_enol from C followed by radical-type carbene insertion into the O–H bond of C_enol (i.e., HAT via TS8 to form D, followed by radical rebound via TS3). However, such pathways can be safely excluded based on the high computed reaction barriers of the HAT step (TS8) in these pathways (∼16 kcal mol–1, so much larger than TS2; see Scheme B).

Conclusions

We developed a new catalytic metalloradical protocol for the construction of unprecedented 1H-2-benzoxocins. The reaction provides an efficient protocol for the synthesis of medium-sized O-heterocyclic rings via cobalt carbene radical intermediates, with a broad substrate scope and an excellent functional group tolerance, and enables the synthesis of a variety of 8-membered cyclic enol ethers in good to excellent yields under mild reaction conditions. The adaptable functionalization of the products showcases the potential for a concise strategy to synthesize a broad range of potentially bioactive structures, containing novel medium-sized heterocycles. Thus, formed 1H-2-benzoxocins also proved to be useful and versatile platforms to prepare a variety of other potentially bioactive substructures. DFT calculations and spin trapping experiments reveal that cobalt carbene radicals are produced, which are involved in a subsequent intramolecular hydrogen atom transfer step followed by product formation via a radical rebound step. The proposed mechanism and the structure of the reaction products are confirmed by control experiments, 2D-NMR spectroscopy, radical trapping experiments, and X-ray diffraction studies. In contrast to most of the existing strategies for the synthesis of medium-sized ether rings, the newly developed metallo-radical catalyzed protocol enables cyclization of easily accessible linear precursors, is high-yielding, and circumvents the use of scarce precious metal catalysts.