Introduction

Rapid and efficient synthesis of structurally diverse molecules is an important goal in organic synthesis, as potentially useful leading compounds were identified by high-throughput screening a library of molecules in drug development and material science. Particularly, scaffold diversity is a key feature of a compound library and crucial for the success in the screening process, because scaffolds are considered as the core structures of molecules and determine their shape, rigidity, and flexibility, as well as placing functional moieties in the correct positions to interact with the biological targets1,2,3. The conventional approaches for preparing structurally diverse scaffolds required the starting materials that are usually varied in structure and different catalysts and reagents under individually established reaction conditions, which are time-consuming and laborious. In transition metal-catalyzed reactions, the chemo-, regio- and stereoselectivity can be modulated by the electronic and steric properties of ligands, enabling transformations of the same set of substrates to form structurally diverse scaffolds. Therefore, ligand-controlled divergent synthesis strategy has attracted increasing attention. Specifically, reactions through divergent chemoselectivies might provide completely distinct scaffolds. However, such processes are rare and challenging due to the inherent preference of the substrates and catalysts for a specific pathway. Cobalt is an inexpensive earth-abundant transition metal of low toxicity4. Development of novel cobalt-catalyzed divergent transformations to produce useful building blocks with diverse scaffolds through the unique reactivity of organocobalt complexes fulfills the increasing demands for sustainable chemistry and provides new reaction modes for cobalt catalysis.

Oxidative cyclization of two π-bonds promoted by low-valent transition metal complexes represents a classical elementary step in organometallic chemistry. Transition metal-catalyzed transformations through oxidative cyclization constitutes one of the most important approaches for carbon–carbon bond forming reactions in organic synthesis. Alkynes and alkenes are commonly used fundamental building blocks that are easily accessible and versatile for highly diversified transformations. Therefore, catalytic coupling of alkynes and alkenes without the need of pre-functionalization initiated by oxidative cyclization is an attractive strategy for rapid construction of multifunctional complex molecules from simple starting materials due to high atom- and step-economy5,6,7,8. A central issue for such processes is the accurate control of chemoselectivity, because not only the lower-lying LUMO of alkynes led to the higher affinity to the metal center and caused self-oligomerization, but also multiple reaction pathways might occur via a common metallacycle intermediate generated from oxidative cyclization of alkynes and alkenes (Fig. 1a). For instance, oxidative cyclization of alkynes and alkenes followed by endocyclic β-hydride elimination promoted by Co-9, Ni-10, Rh-11, Ru-12,13,14 and Pd-based15 catalysts afforded 1,3-dienes (pathway a). The coupling partners are limited to monosubstituted styrenes and electron-deficient alkenes. On the other hand, Ru-16,17,18,19,20,21,22 and Co-catalyzed23,24,25 coupling of alkynes with monosubstituted alkyl alkenes delivered 1,4-dienes as the favored products through exocyclic β-hydride elimination (pathway b). Few examples of catalytic enantioselective [2 + 2] cycloaddition through oxidative cyclization followed by reductive elimination have been developed to furnish enantioenriched cyclobutenes (pathway c). Such processes promoted by Co-based catalysts are limited to monosubstituted electron-deficient alkenes or 1,3-dienes with carbonyl-activated alkynes26,27, whereas Ru-28,29, Ir-30,31, Ni-32,33 and Rh-catalyzed34 protocols are only suitable for bridged alkenes with significant limitation of substrate scopes. In the presence of a proton source, the metallacycle intermediates formed from alkynes and monosubstituted electron-deficient alkenes promoted by Co- and Ni-based catalysts were able to undergo double protonation to produce trisubstituted alkenes as the reductive coupling products (pathway d)35,36,37,38,39,40. Scarce examples of enantioselective reductive coupling with disubstituted electron-deficient alkenes have recently been developed41,42,43. However, catalytic enantioselective [2 + 2] cycloaddition as well as reductive coupling of alkynes with nonpolar alkenes remained unknown.

Fig. 1: Background and reaction design.
figure 1

a Possible reaction pathways for transition metal-catalyzed coupling of alkynes and alkenes. b Representative natural products and biologically active molecules containing a four-membered ring moiety. c Catalytic enantioselective carbon–carbon forming reactions of nonpolar cyclobutenes. d Strategies for divergent coupling through oxidative cyclization. e Cobalt-catalyzed pathway-divergent enantioselective coupling of alkynes and cyclobutenes (this work). EWG electron-withdrawing group, PC photocatalyst.

Full size image

Enantioenriched cyclobutenes and cyclobutanes are common motifs in natural products and pharmaceutical important molecules, as well as useful building blocks in organic synthesis (Fig. 1b)44,45,46,47,48. Establishment of a general and modular strategy for catalytic enantioselective synthesis of a broad scope of densely functionalized cyclobutenes and cyclobutanes is thus highly sought after. Catalytic enantioselective approaches for access to enantioenriched cyclobutenes involve [2 + 2], [3 + 1] cycloaddition and functionalization of four-membered carbocycles, but suffered from significant limitation of substrate scopes and the requirement of installing specific functional groups49. However, none of these protocols enabled catalytic enantioselective construction of cyclobutenes fused by another four-membered carbocycles which are core structures in a variety of natural products as well as isosteres for substituted aromatic rings and rigidified cyclohexanes50. Catalytic enantioselective methods for synthesis of cyclobutanes include [2 + 2] cycloaddition51,52,53,54,55,56,57,58,59,60,61,62, cyclization63,64,65, ring contraction and expansion reactions in multiple steps66,67,68, but specific substrate patterns are required. An alternative approach is catalytic enantioselective functionalization of preformed four-membered carbocycles, consisting of directed C–H functionalization as well as reactions of cyclobutanones and cyclobutenes69. The advantage for such approach is that highly diversified functional groups can be introduced directly into the four-membered ring scaffold through a single set of transformations and starting materials70. Although progresses have been made in this field, significant limitations remained unaddressed. A boryl, aryl, or alkenyl group could be incorporated requiring a directing group for catalytic enantioselective C–H functionalization of cyclobutanes71,72,73,74,75,76,77. Only reduction and reactions of ketone enolates have been developed for catalytic enantioselective transformations of cyclobutanones78,79,80,81,82,83. Moreover, few approaches for catalytic enantioselective conjugate addition of activated cyclobutenes, including Cu-catalyzed hydride, boryl, simple alkyl addition with dialkylzinc reagents84,85,86, organocatalyzed Diels-Alder reactions87, and Rh-catalyzed aryl addition with aryl boronic acids88. We recently disclosed the cobalt-catalyzed enantioselective alkyl conjugate addition of cyclobutenes with cobalt homoenolates generated from cyclopropanols89. A more challenging class of substrates are cyclobutenes without conjugated carbonyls. The sluggish reactivity arose from relatively lower olefinic strain (1.9 kcal/mol) in comparison with that in cyclopropenes (27.7 kcal/mol) and bicyclic olefins (norbornene: 4.8 kcal/mol)90,91,92,93. While metal-catalyzed enantioselective hydroboration94, borylallylation95, hydroamination96, and diboration of cyclobutenes have been revealed97, the enantioselective carbon–carbon bond forming reactions of cyclobutenes remained scarce, and mainly limited to processes via carbometallation (Fig. 1c). Fletcher and co-workers reported a series of Rh-catalyzed approaches for enantioselective arylation98, hydroacylation of cyclobutenes99 as well as allylic substitution of cyclobutanone ketals with aryl boronic acids100. Lu and co-workers described a Pd-catalyzed protocol for enantioselective arylation of cyclobutenes with aryl iodides101. We disclosed the only example of cobalt-catalyzed enantioselective alkynylation and cascade allyl addition of cyclobutenes89. However, methods for catalytic enantioselective carbon–carbon bond forming reactions of cyclobutenes through oxidative cyclization remained elusive. Very recently, we reported an instance of regio-, diastereo- and enantioselective reductive coupling of 1,1-disubstituted allenes and cyclobutenes through oxidative cyclization, whereas no precedence on enantioselective coupling of alkynes and cyclobutenes through oxidative cyclization has been revealed102. One of the key challenges for such processes arises from the low reactivity of the cyclobutenes without conjugated carbonyls that results in the difficulty to overcome the inherent higher reactivity of alkynes. Self-oligomerization of alkynes has to be suppressed chemoselectively. Furthermore, the presence of a four-membered carbocycles fused to the metallacycle (A, Fig. 1e) introduced additional ring strain and elevated the barrier for formation of such species. Another crucial issue is accurate control of the reaction pathway followed by oxidative cyclization. In the meanwhile, the regio- and stereoselectivity of each elementary step has to be precisely tuned. In addition, as multiple possible elementary steps might occur from the same metallacycle (Fig. 1a), it is more challenging to identify different chiral ligands to access divergent pathways. In this context, scarce examples of cobalt-catalyzed divergent enantioselective coupling of unsaturated hydrocarbons through oxidative cyclization have been revealed. The strategy in these cases was to use unsaturated hydrocarbons bearing multiple reactive sites27,103,104,105. Different ligands enabled the reactions to occur at divergent sites via different metallacycle intermediates (Fig. 1d). An alternative strategy for divergent enantioselective coupling that remained undeveloped is to modulate the following elementary step via the same metallacycle intermediate generated from oxidative cyclization of unsaturated hydrocarbons with a single reactive site (Fig. 1d)106,107. One of the most critical challenges for such processes is to overcome inherent preference for a particular pathway. We envisioned that different properties of chiral ligands might enable varied elementary steps to occur preferentially after oxidative cyclization. Herein, we disclosed a cobalt-catalyzed protocol for pathway-divergent diastereo- and enantioselective coupling of alkynes and cyclobutenes through oxidative cyclization followed by protonation or reductive elimination accurately controlled by ligands, affording a broad scope of enantioenriched cyclobutanes and cyclobutenes fused with a four-membered ring that are otherwise difficult to access in high chemo-, regio-, diastereo- and enantioselectivity.

Results and discussion

Reaction optimization

Our studies commenced with reaction of alkyne 1a with cyclobutene 2a promoted by Co complexes derived from various phosphine ligands (Fig. 2a). The utility of photoredox catalysis for single-electron transfer provided cleaner reactions (see Supplementary Information for more details). Although various Co complexes formed from bisphosphines bearing axial stereogenicity (4a) and chiral alkyl linkers (4b–d) were able to induce the reaction, hydroalkenylation product 3a through reductive coupling pathway was afforded in 68–85% yield with 60:40–72:28 er as a single diastereomer (entries 1–4). Improved enantioselectivity was obtained with bisphosphines containing P-stereogenic centers (4e) and a ferrocene skeleton (4f) (entries 5–6). No reaction occurred in the presence of chiral bisphosphines 4g and 4i (entries 7 and 9). We found that Co complex generated from phosphine bearing chiral phospholane fragments (4h) enabled production of 3a in 56% yield with 99:1 er (entry 8). Switching the Hantzsch ester to water led to less side products albeit in lower yield (entry 10). Increasing the amount of water significantly enhanced the efficiency with slight erosion of enantioselectivity (entry 11). It is worth mentioning that minor amount of [2 + 2] cycloaddition product 5a was formed with 4a (20% yield, 50:50 er) and 4e (14% yield, 90:10 er). Only a single regioisomer was observed. We next surveyed another reaction pathway to access [2 + 2] cycloaddition products (Fig. 2b). Such reaction mode is more challenging, as highly strained 4,4-fused ring scaffold was generated through reductive elimination. Although a number of chiral ligands were able promote the oxidative cyclization/protonation sequence, [2 + 2] cycloaddition product 5a furnished through oxidative cyclization/reductive elimination was more difficult to be produced with the same ligand (entries 3–5 and 8, Fig. 2a vs. entries 3–5 and 8, Fig. 2b), indicating that inherently protonation was a more favorable elementary step for subsequent reaction of the metallacycle formed from oxidative cyclization compared with reductive elimination. Although transformations of alkyne 1a and cyclobutene 2a in the presence of bisphosphines 4a, b and 4f afforded [2 + 2] cycloaddition product 5a exclusively, low enantioselectivity was obtained (entries 1–2 and 6). We discovered that introduction of electron-deficient aryls to the chiral phosphine containing a ferrocene skeleton resulted in higher enantioselectivity (entry 9). Further investigations revealed that reaction in DME provided improved enantioselectivity albeit with lower efficiency (entry 10). Alteration of CoI2 to CoBr2 enhanced the yield without erosion of enantioselectivity (entry 11). Reduction of the amount of Zn powder resulted in production of 5a in 66% yield with 93:7 er as a single diastereomer (entry 12).

Fig. 2: Optimization of reaction conditions.
figure 2

a Optimization for enantioselective reductive coupling pathway. b Optimization for enantioselective [2 + 2] cycloaddition pathway. aYield of isolated product. bRatios of regioisomers (rr) and diastereomers (dr) were determined by analysis of 1H NMR spectra of unpurified mixtures. cEnantiomeric ratios (er) were determined by analysis of HPLC spectra. dNot available. eThe reaction was performed with 5.0 mol % CoI2, 5.0 mol % 4h, 2.5 equiv of H2O for 24 h. fThe reaction was conducted with 5.0 mol % CoI2, 5.0 mol % 4h, 25 equiv of H2O for 24 h. gThe reaction was performed in DME. hThe reaction was conducted in the presence of 10 mol % CoBr2. iThe reaction was performed with 10 mol % Zn. 4CzIPN, 1,2,3,5-tetrakis(carbazole-9-yl)−4,6-dicyanobenzene; THF tetrahydrofuran.

Full size image

Scope for cobalt-catalyzed pathway-divergent enantioselective reactions of alkynes and cyclobutenes

With the optimal conditions, we investigated the scope of cobalt-catalyzed enantioselective reductive coupling of alkynes and cyclobutenes (Fig. 3). We found that 100 equiv of H2O as well as 10 mol % catalyst loading was required for high efficiency in a lot of cases. Alkynes bearing electron-rich (3b–f, 3o–r, 3t), electron-deficient (3h–n), halogenated (3k, l, 3n, 3s) and sterically congested aryl group (3s–u) underwent the reductive coupling, enabling incorporation of a broad scope of stereochemically defined trisubstituted alkenyl groups in high efficiency and stereoselectivity (30–95% yield, 92:8– > 99.5:0.5 er, >98:2 rr, >98:2 dr). Functional groups such as silyl (3g), cyano (3i), and carbonyl (3j) groups are well tolerated. Alkynes containing pharmaceutically important heterocycles including benzofuran (3w), thiophene (3v, 3x), pyridine (3y) as well as indole (3z) served as suitable substrates that delivered enantioenriched cyclobutanes in 54–82% yield with 98:2– > 99.5:0.5 er as a single diastereomer. Alkynes substituted with primary alkyls (3aa–ae) and secondary alkyls (3af) were transformed into the hydroalkenylation products in 45–55% yield with 91:9–99:1 er as a single regioisomer and diastereomer. Unexpectedly, reaction of trimethylsilyl substituted alkyne provided reversed regioselectivity (3ag). Transformation of 1,3-enyne delivered cyclobutane 3ah bearing a dienyl group in 47% yield with 97:3 er. Cyclobutenes bearing functionalized alkyls (6a), allyl (6b) as well as aryl (6c–e) and easily removable silyl (6f) groups are compatible with the reaction. It is worth mentioning that in some cases, the transformations proceeded with higher efficiency in the presence of Hantzsch ester (3d, 3h–j, 3m, 3p, 3y).

Fig. 3: Scope for cobalt-catalyzed enantioselective reductive coupling of alkynes and cyclobutenes.
figure 3

aThe reactions were conducted with 10 mol % CoI2 and 10 mol % 4h in the presence of 2.5 equiv of Hantzsch ester instead of H2O. bThe reactions were performed with 10 mol % CoI2 and 10 mol % 4h in the presence of 100 equiv of H2O for 48 h. 4CzIPN, 1,2,3,5-tetrakis(carbazole-9-yl)−4,6-dicyanobenzene.

Full size image

We next explored the scope for cobalt-catalyzed enantioselective [2 + 2] cycloaddition of alkynes and cyclobutenes (Fig. 4). A variety of alkynes bearing electron-rich (5b–d, 5m, 5n–p), electron-deficient (5f–l), halogenated (5f, g, 5l), and sterically congested aryl (5n–p) groups could be converted into the desired products containing a 4,4-fused ring system in 38–67% yield with 84:16–98:2 er as a single diastereomer. Functional groups such as silyl (5e), cyano (5i) and carbonyl (5j) groups are compatible with the reaction conditions. Alkynes containing heterocycles (5q) are suitable substrates. Transformations of alkynes substituted with primary alkyl (5r–v) and secondary alkyl (5w) groups furnished the densely functionalized 4,4-fused carbocycles in 45–66% yield with 91:9–92:8 er as a single diastereomer. Symmetric alkynes were competent substrates (5x). Reactions of 1,3-enynes provided 4,4-fused carbocycles (5y, z) in high efficiency albeit with lower enantioselectivity. Cyclobutenes bearing functionalized alkyl (7a), allyl (7b), aryl (7c–e) as well as easily removable silyl (7f) groups were able to participate in the reaction, delivering [2 + 2] cycloaddition products in 43–64% yield with 87:13–93:7 er. It should be noted that the succinimide moiety was crucial for the reactivity in both reductive coupling and [2 + 2] cycloaddition pathways. No reaction occurred with cyclobutenes in the absence of such moiety.

Fig. 4: Scope for cobalt-catalyzed enantioselective [2 + 2] cycloaddition of alkyne and cyclobutenes.
figure 4

aThe reactions were performed in the presence of 10 mol % dppf instead of 4i. DME dimethoxyethane, SEM trimethylsilylethoxymethyl.

Full size image

Scalable reactions and functionalization

The [2 + 2] cycloaddition can be performed on gram scale (Fig. 5a). Reaction of alkyne 1a (0.82 g) with cyclobutene 2a (1.00 g) in the presence of Co complex derived from 4i afforded 4,4-fused ring product 5a (0.82 g) in 53% yield with 93:7 er. The enantioenriched products can undergo functionalization in various ways to provide valuable and otherwise difficult-to-access building blocks (Fig. 5b, c). As the succinimide scaffold was indispensable for the coupling reactions, it is crucial to convert the succinimide to other functional groups. Although catalytic approaches for direct hydrogenation of the succinimide have been disclosed108,109, the intolerance of the alkene group prompted us to conduct a multistep sequence110. Diols 8 and 16 were produced in 52% and 55% overall yield with 99:1 and 92:8 er. The diols are versatile to be converted to a variety of other functional groups. Reduction of the succinimide moiety with LiAlH4 generated tetrahydropyrroles 9 and 17 without diminishment of stereoselectivity. For functionalization of the trisubstituted alkene in the reductive coupling products, chemo- and diastereoselective hydrogenation promoted by Rh complexes 10 and 11 provided 12 and 13 in high efficiency without touching the benzyl group, enabling incorporation of a chiral alkyl group into the cyclobutane ring (Fig. 5b)111. Both diastereomers can be accessed, as the stereoselectivity was controlled by the chiral Rh catalysts. Catalytic diastereoselective epoxidation of the trisubstituted alkene delivered epoxide 15 in 58% yield with 90:10 dr112. For the enantioenriched cyclobutenes containing a 4,4-fused ring system, epoxidation of the tetrasubstituted alkene supplied a highly strained 3,4,4-fused ring scaffold 18 in 95% yield as a single diastereomer, whereas hydrogenation in the presence of Pd/C delivered 19 in 70% yield as a single diastereomer (Fig. 5c). Desilylation of 3ag enabled introduction of a 1,1-disubstituted alkenyl group to the four-membered scaffold (Fig. 5d).

Fig. 5: Gram-scale reaction and functionalization.
figure 5

a Gram-scale reaction. b Functionalization of the reductive coupling product. c Functionalization of the [2 + 2] cycloaddition product. d Desilylation. DME dimethoxyethane, DMM dimethoxymethane.

Full size image

Preliminary mechanistic studies

To gain some mechanistic insight, we conducted the experiment with D2O (Fig. 6a). The incorporation of deuterium at the trisubstituted olefin and α position of the carbonyl suggested that the reductive coupling product 3a was afforded through protonation of the alkenyl–Co bond in the metallacycle generated from oxidative cyclization of the alkynes and the cyclobutenes, and subsequent β-H elimination followed by insertion and protonation. A single diastereomer of the deuterated stereogenic center at the cyclobutene was furnished, indicating that the possible isomerization of the alkenyl–Co(III) species was not observed. In addition, re-insertion of the Co(III)–H species to the cyclobutene formed from β-H elimination of the metallacycle proceeded regio- and stereoselectively. Parallel kinetic experiments for the reductive coupling pathway were conducted, implying that cleavage of the H–O bond might be the rate-determining step (Fig. 6b). Only product 3a was obtained in the competitive experiments in the presence of both H2O and D2O, supporting that the rate of reaction with H2O was faster than that of reaction with D2O and protonation was irreversible (Fig. 6c). Transformations of deuterated cyclobutene 2a-D with H2O or D2O enabled selective access enantioenriched cyclobutanes (3a-D2, 3a-D3) bearing stereogenic C–D bonds at different sites (Fig. 6d, e). These results also suggested that the reaction proceeded through β-H elimination followed by insertion and protonation. Parallel kinetic experiments with cyclobutene 2a and deuterated cyclobutene 2a-D suggested that β-H elimination and re-insertion might not be the rate-determining steps (Fig. 6f). Transformation of a mixture of 2b and 2a-D provided 6b and 3a-D2 (Fig. 6g). No deuterium was incorporated into 6b, implying that no dissociation of the cyclobutene generated from β-H elimination occurred. Reaction of deuterated cyclobutene 2a-D through [2 + 2] pathway resulted in no migration of deuterium, illustrating that reductive elimination of the metallacycle intermediate was faster than possible β-H elimination (Fig. 6h).

Fig. 6: Preliminary mechanistic studies.
figure 6

a, e Reactions with D2O. b, f Parallel kinetic experiments. c Competitive kinetic experiments. g Cross-over experiment. d, h Reactions of deuterated cyclobutene. 4CzIPN, 1,2,3,5-tetrakis(carbazole-9-yl)−4,6-dicyanobenzene; DME dimethoxyethane.

Full size image

Computational studies

Density functional theory (DFT) calculations were carried out to shed light on the reaction mechanism and probe the origin for ligand-controlled chemo- and stereoselectivity (Fig. 7). We first explored the pathway using relatively electron-rich bisphosphine ligand 4h. The reaction commenced with oxidative cyclization to generate metallacycle INT-IV-4h. Both high spin and low spin states were taken into consideration due to the presence of close-lying spin states in first-row transition metal complexes113,114,115,116. The oxidative cyclization in singlet via TS1-4h was calculated to be significantly more favored than that in triplet. This preference in spin state is reserved in the product intermediate INT-IV4h which is favored over 3INT-IV-4h by 7.0 kcal/mol. The oxidative cyclization transition state (TS) lies higher than the subsequent steps (vide infra), establishing it as the stereo-determining step. Our calculations showed that TS1-4h is more favored than TS1-ent-4h, being consistent with the experimental observed configuration. A careful comparison of the geometries of these two TSs (depicted in the upper right of Fig. 7A) revealed that in the more favored TS1-4h, the bulky cyclobutane fragment is oriented to a less congested quadrant of ligand pocket (Q2), thereby reducing steric repulsions. Conversely, in TS1-ent-4h, the cyclobutane fragment experiences greater steric hindrance with the phenyl group of the ligand in Q1, destabilizing the transition state. Given that the reaction yields distinct products with different ligands, we explored both protonation and reductive elimination processes that share the same metallacycle intermediate. Our calculations demonstrated that while the reductive elimination in triplet is significantly more favored than the singlet, the protonation process via singlet TS3-4h, which involves the breaking of Co–C bond and the formation of C–H bond, exhibit lowest energy barrier. This finding is consistent with experimental observations that ligand 4h eventually leads to the hydroalkenylation product 3a instead of cyclization product 5a. Natural population analysis (shown in bottom right of Fig. 7A) revealed that the natural atomic charge (NPA charge) of Co is −0.454 in singlet protonation TS and 0.179 in triplet reductive elimination TS. The strong field ligand 4h donates more electron to metal center, enhancing the nucleophilicity of Co-alkenyl and destabilizing high spin state, resulting in a preference to protonation process occurring in low spin state. In contrast, when the relatively electron-poor ligand 4i was used in calculations, the high spin state generally exhibits greater stability in energy profiles due to a weakened ligand field (see Fig. 7B). The oxidative cyclization favored triplet, with the energy difference of metallacycle with different spin states decreasing by 4 kcal/mol. The electron-poor ligand 4i diminishes the electron density on Co center and imposes larger steric hindrance, resulting in an increase in energy barrier for the singlet oxidative cyclization and thus reverse order of spin state preference (see Supplementary Fig. 2). The ligand spiral direction is reversed with 4i, resulting in a preferred configuration for oxidative cyclization that is opposite to that observed with 4h. As illustrated in the upper right in Fig. 7B, in 3TS1-ent-4i, the ligand distorts significantly to mitigate steric repulsions from the cyclobutane fragment in Q2, leading to an increase in energy. After metallacycle formation, the subsequent reductive elimination in triplet was calculated to be more favored than the singlet protonation process by 4.1 kcal/mol, being consistent with experimentally reported cyclization product using the weak field ligand 4i. The calculated NPA charge of Co in TS3-4i decreased to −0.300, indicating a weaker nucleophilicity. In the meanwhile, the more positive NPA charge of Co in 3TS2-4i is beneficial to reductive elimination where electron densities accumulated in metal center. In protonation transition state TS3-4i, the biting angle of ligand deviates from its optimum 95.81° to 92.60° to realize a square-pyramidal geometry, contributing to destabilizations. By comparing the geometries of both singlet protonation and triplet reductive elimination transition states with different ligands, a common observation is that with weak field ligand 4i, metal-ligand interaction is weakened with elongated Co–P bond lengths, further articulating the aforementioned preference to high spin state. In summary, the calculated energy profiles with different ligands are in good agreement with the experimentally observed ligand-controlled chemo- and stereoselectivity. A detailed analysis demonstrates that both electronic effects and geometric constraints of ligands significantly influence the relative stability of various spin states, leading to distinct pathways.

Fig. 7: Computational studies.
figure 7

Energy profiles with ligands 4h and 4i. The relative free energies and electronic energies (in parentheses) are given in kcal/mol. A, Energy profiles with ligand 4h. B, Energy profiles with ligand 4i.

Full size image

Proposed catalytic cycles

Based on all the observations above, we proposed the catalytic cycles for the cobalt-catalyzed pathway-divergent enantioselective coupling (Fig. 8). Visible light irradiation excited 4CzIPN, which underwent single electron reduction by i-Pr2NEt to afford 4CzIPN• –. The 4CzIPN• – was able to reduce the Co(II) complex I to Co(I) species II, which was evidenced by DFT calculations to be a facile process with a barrier of 1.8 kcal/mol. Coordination of the alkyne 1a and cyclobutene 2a to the Co center followed by oxidative cyclization generated the metallacycle IV. In the presence of more electron-rich chiral ligand 4h, low spin organocobalt species were energetically favored and the pathway of protonation of the alkenyl–Co bond with subsequent β-H elimination and re-insertion followed by protonation delivered the reductive coupling product 3a occurred preferentially. The possible single-electron reduction of the metallacycle VI with subsequent protonation of C–Co(II) bonds cannot be ruled out. With electron-deficient chiral ligand 4i, high spin organocobalt intermediates were favored and chemoselective reductive elimination of the metallacycle IV to release the [2 + 2] cycloaddition product 5a proceeded exclusively.

Fig. 8: Proposed catalytic cycles.
figure 8

Catalytic cycles for pathway-divergent cobalt-catalyzed enantioselective reductive coupling and [2 + 2] cycloaddition of alkynes and cyclobutenes were proposed based on experimental observations and DFT calculations.

Full size image

In conclusion, a cobalt-catalyzed protocol for pathway-divergent regio-, diastereo- and enantioselective coupling of alkynes and cyclobutenes to furnish densely functionalized cyclobutanes as well as cyclobutenes containing a 4,4-fused ring scaffold in high efficiency and selectivity has been developed. The transformations proceeded through regio-, diastereo- and enantioselective oxidative cyclization followed by protonation or reductive elimination accurately controlled by chiral ligands. The starting materials are easily accessible. Inexpensive cobalt salts and commercially available bisphosphines are required for generation of the catalysts. The synthetic utility was demonstrated by diverse transformations of the products, delivering a variety of enantioenriched building blocks containing four-membered rings that are otherwise difficult to access. Mechanistic studies and DFT calculations were conducted to reveal that the properties of the chiral ligands induced different spin states of the cobalt center, resulting in varied reaction pathways. The origin of stereoselectivity was elucidated as well. Such discoveries unveiled a concept for developing divergent reactions, unknown reaction pathways for cobalt catalysis, opening up chances for designing novel transformations promoted by Co-based catalysts, providing new strategies for complex molecule synthesis and pushing forward the development of organocobalt chemistry. In-depth analysis highlighted the potential to manipulate the reactivities of different spin states by adjusting the electronic and geometric factors of ligand. Further investigations on other Co-catalyzed pathway-divergent enantioselective reactions through oxidative cyclization are underway.

Methods

General procedure for cobalt-catalyzed enantioselective reductive coupling of alkynes and cyclobutenes

In a N2-filled glove box, an oven-dried 8-mL vial equipped with a stirring bar was charged with CoI2 (3.1 mg, 0.01 mmol, 5.0 mol %), 4h (5.1 mg, 0.01 mmol, 5.0 mol %), and MeCN (2.0 mL). The solution was allowed to stir at 22 oC for 30 min. 4CzIPN (3.1 mg, 0.004 mmol, 2.0 mol %), i-Pr2NEt (51.7 mg, 0.4 mmol, 2.0 equiv), 1a (34.8 mg, 0.3 mmol, 1.5 equiv), 2a (42.6 mg, 0.2 mmol, 1.0 equiv) and Hantzsch ester (126.6 mg, 0.5 mmol, 2.5 equiv) or H2O (90.0 mg, 5.0 mmol, 25.0 equiv) were added to the solution. Then the vial was sealed with a cap (phenolic open top cap with red PTFE/white silicone septum), removed from the glove box. The mixture was irradiated by 40 W blue LEDs (450–455 nm, λmax = 454 nm) and allowed to stir at 60 oC for 24 h. Upon cooling to room temperature, the reaction mixture was filtered through a short plug of silica gel (100–200 mesh) eluted with diethyl ether (30 mL × 3). The filtrate was concentrated in vacuo. The residue was purified by silica gel column chromatography (petroleum ether:acetone = 15:1) to afford the 3a as white solid (51.0 mg, 0.154 mmol, 77% yield, >98:2 rr and dr).

General procedure for cobalt-catalyzed enantioselective [2 + 2] cycloaddition of alkynes and cyclobutenes

In a N2-filled glove box, an oven-dried 8-mL vial equipped with a stirring bar was charged with CoBr2 (4.4 mg, 0.02 mmol, 10 mol%), 4i (17.0 mg, 0.02 mmol, 10 mol%), Zn powder (1.3 mg, 0.02 mmol, 10 mol%) and DME (2.0 mL). The solution was allowed to stir at 22 oC for 30 min. 1a (34.8 mg, 0.3 mmol, 1.5 equiv) and 2a (42.6 mg, 0.2 mmol, 1.0 equiv) were added to the solution. Then the vial was sealed with a cap (phenolic open top cap with red PTFE/white silicone septum), removed from the glove box. The mixture was immediately moved to a thermostatic bath and allowed to stir at 70 oC for 24 h. Upon cooling to room temperature, the reaction mixture was filtered through a short plug of silica gel (100–200 mesh) eluted with diethyl ether (30 mL × 3). The filtrate was concentrated in vacuo. The residue was purified by silica gel column chromatography (petroleum ether:ethyl acetate = 15:1) to afford the 5a as yellow solid (42.8 mg, 0.130 mmol, 65% yield, >98:2 dr).