Construction of axial chirality via asymmetric radical trapping by cobalt under visible light

The 3d metals have been identified as economic and sustainable alternatives to palladium, the frequently used metal in transition-metal-catalysed cross-couplings. However, cobalt has long stood behind its neighbouring elements, nickel and copper, in asymmetric radical couplings owing to its high catalytic activity in the absence of ligands resulting in unfavourable un-asymmetric background reactions. Here we disclose an asymmetric metallaphotoredox catalysis (AMPC) strategy for the dynamic kinetic asymmetric transformation of racemic heterobiaryls, which represents a visible-light-induced, asymmetric radical coupling for the construction of axial chirality. This success can also be extended to the reductive cross-coupling variant featuring the use of more easily available organic halide feedstocks. The keys to these achievements are the rational design of a sustainable AMPC system that merges asymmetric cobalt catalysis with organic photoredox catalysis in combination with the identification of an efficient chiral polydentate ligand. Controlling the enantioselectivity in metallaphotoredox-catalysed radical cross-couplings using cobalt has proven challenging. Now, the identification of a chiral polydentate ligand enables cobalt-catalysed enantioselective couplings with a broad scope of radicals affording chiral heterobiaryl products.

O ver the past decade, organic photochemical synthesis that utilizes sustainable, abundant visible light has witnessed a noticeable renaissance [1][2][3] . Due to the high reactivity of open-shell intermediates, asymmetric photochemical transformations under visible light 4 are believed to be challenging. Even so, stereocontrol strategies including asymmetric organocatalysis 5 , organometallic catalysis 6 and enzyme catalysis 7 have been gradually applied to this area. As highlighted in Fig. 1a, the emerging dual catalysis strategy has achieved considerable success in construction of chiral carbon stereocenters [8][9][10][11][12][13][14][15][16][17] ; however, its application in construction of axially chiral molecules is in its infancy 18 . To the best of our knowledge, there has only been one chiral bifunctional catalyst system which was recently reported by Bach, who disclosed a catalytic deracemization of allene lactams via an energy transfer mechanism under visible light (as shown in Fig. 1b) 19 and later expanded the system to primary allene amides 20 . Despite this noticeable achievement, the exploitation of efficient catalyst systems and novel asymmetric transformations are still highly desirable in this field.
Cobalt-catalysed cross-couplings have been well known as one of the most powerful approaches to forge chemical bonds 21 . However, as an important 3d metal, cobalt has been much less studied for use in asymmetric radical couplings [22][23][24] than its neighbours, nickel 25 and copper 26,27 . Recently, metallaphotoredox-catalysed radical cross-couplings that merge photoredox catalysis and 3d metal catalysis have flourished 28 and again, cobalt 29 stands behind nickel and copper in its utilization. One main reason might be that the high catalytic activity of cobalt salts themselves in the absence of additional ligand hinders chiral ligand-promoted asymmetric couplings. Because of our research interests in asymmetric photochemical transformations under visible light [30][31][32] , we recently questioned whether chiral polydentate ligands could increase the possibility of coordination with cobalt. More importantly, compared with the extensively used Cu(I) and Ni(0) which have 10 valence electrons (d10), the relatively empty 3d orbital of Co(I) (eight outermost electrons, d8) would allow polydentate binding to not affect the catalytic performance in oxidative addition; again, compared with the well-known planar coordination structures of Pd(II) and Ni(II), the octahedral coordination of Co(III) would be more favoured for asymmetric induction by the chiral polydentate ligand.

Results
Design plan. As illustrated in Fig. 1c, we assumed that oxidative addition of low-valent cobalt would afford cationic cyclic cobalt intermediates (R)-int. I and (S)-int. I. As a basis of this proposal, the coordination of cobalt with the basic nitrogen atom in the heterobiaryl scaffold would keep the two steric groups far from each other compared with their positions in starting materials 1, and the conversion of (R)-int. I into (S)-int. I via axial rotation would be more easily induced by chiral ligands. Theoretically, the smaller atomic radii of 3d metals would make the rotation process more favourable than that with 4d metals. Next, asymmetric trapping of photoredox-generated radicals by the chiral cobalt intermediate (S)-int. I would afford the desired axially chiral heterobiaryl product through reductive elimination. According to this reaction scenario, quick axial rotation and stereo-favoured radical trapping are critical for both high enantioselectivity and efficiency. If successful, this 3d-metal cobalt-involved, metallaphotoredox-catalysed radical DYKAT will add a member to the arsenal of techniques for chiral heterobiaryl construction 43 , which is difficult through precious-metal-catalysed two-electron processes. Additionally, these radical DYKAT processes will be highly valued because of their avoidance of air-and moisture-sensitive organometallic reagents, thus increasing the practicality and functional-group compatibility of the methodology. Despite these promising advantages, the underlying challenges include the satisfactory chemo-and enantioselectivity of asymmetric radical coupling, and the potential for reductive dehalogenation (triflates are a kind of halogenoid widely used in transition-metal-catalysed cross-couplings) which, according to preceding literature, would compete with the designed process through aryl-cobalt decomposition 44 . On the other hand, to date, successful cobalt-catalysed asymmetric radical cross-couplings are rare, perhaps because less efficient catalytic systems have been utilized.
To examine the feasibility of the required reactivity of our hypothesis, we initially theoretically investigated the cobalt/ photoredox-catalysed radical cross-coupling pathway using heterobiaryl substrate 1a and 1,4-dihydropyridine (DHP) substrate 2a as the model reactants. According to the preceding literature on cobalt catalysis and photoredox catalysis, the newly designed coupling reaction would include C-O oxidative addition to Co(I), ligand exchange, single-electron reduction of Co(III) to Co(II), benzyl radical addition to Co(II) and C-C reductive elimination (as shown in Fig. 2a). Our density functional theory (DFT) results suggest that formation of triplet Co(III) intermediate 3 B through C-O oxidative addition to three coordinate Co(I) complex ( 3 TS-1) is the rate-determining step and requires an activation free energy of only 20.2 kcal mol -1 . The ligand exchange between PPh 3 and TfOcan form a more stable Co(III) intermediate 3 C. Notably, single-electron transfer between 3 C and PCis exergonic by 24.3 kcal mol -1 , indicating that the single-electron reduction of Co(III) to Co(II) is thermodynamically favoured. Although the transition state of benzyl radical addition to doublet Co(II) complex 2 D is not obtained, computational results show that the generation of Co(III) intermediate 3 G after radical addition ( 2 D → 3 F) followed by PPh 3 dissociation ( 3 F → 3 G) is exergonic by 5.6 kcal mol -1 , which indicates that the formation of 3 G is a feasible process. Finally, the formation of coupling product rac-3a through the C-C reductive elimination transition state 3 TS-2 is highly exergonic  by 83.8 kcal mol -1 and has an energy barrier of only 4.8 kcal mol -1 . Therefore, the DFT calculations suggested that the designed radical coupling is theoretically feasible by merging cobalt catalysis and photoredox catalysis. To verify this theoretical study, we performed a radical coupling reaction with heterobiaryl 1a and DHP 2a in the presence of the cobalt catalyst Co(I)(PPh 3 ) 3 Cl and an organic photocatalyst under the irradiation of blue light-emitting diodes (LEDs). Indeed, the desired coupling product rac-3a was facilely obtained in 67% yield (as shown in Fig. 2b). Furthermore, computational studies also revealed that after the C-O oxidative addition to Co(I), the dihedral angle (θ) in 1a formed by the planes of the two aromatic rings significantly decreases from 64.8° to 28.4° (intermediate 3 B, as shown in Fig. 2a). This prominent geometric change can diminish the steric clash between two aromatic rings amid the C-C axial rotation, which paves the way for a rational design of the dynamic kinetic asymmetric transformation of racemic heterobiaryls.
Condition optimizations for the AMPC-enabled radical DYKAT. Encouraged by the above success, we turned our attention to experimentally studying the asymmetric process. The metallaphotoredox-catalysed DYKAT was optimized with respect to the chiral ligands, cobalt salts and other reaction parameters. The results of the ligand effect study are summarized in Fig. 3a.
We found that chiral monodentate phosphine ligand L1 failed to induce any enantioselectivity, although the product was obtained in a high yield (90% yield, 50:50 enantiomeric ratio (e.r.)). Chiral bidentate bisoxazoline ligands, which showed high efficiency and enantioselectivity in cobalt-catalysed asymmetric radical couplings of α-bromocarbonyls 22   L5-L8, which were exploited by Zhang's group usually as chiral bifunctional organocatalysts 45 . To our delight, chiral bisphosphine ligand L8 was determined to be the best ligand, resulting in 90% yield and 85:15 e.r. To understand the function of the sulfinyl amine unit, two additional ligands L9 and L10 were tested. The results show that the chiral sulfinyl amine unit plays an important role in both reaction efficiency and good stereocontrol (L8 and L9), and we speculate that this amide might not function as a chiral anionic ligand like Liu's catalyst systems with ligand L4 because the difference in enantioselectivity was not obvious at 30 °C (L8 versus L9) 23 . Later, we found that increasing the reaction temperature and using 90 W blue light can further improve the enantioselectivity, while retaining high yields (Fig. 3b, entries 1-4). Next, a series of cobalt catalyst precursors was evaluated (Fig. 3b, entries 4-9), and CoCl 2 gave a slightly better result (Fig.  3b, entry 6, 93% yield and 96:4 e.r.). Finally, after further optimizing the amount of each component of the reaction (see details in Supplementary Tables 1-9), we confirmed that the combination of CoCl 2 (5 mol%), chiral ligand L8 (10 mol%), i-Pr 2 NEt (1.0 equiv.) and ZnCl 2 (20 mol%) provided optimal conditions for this visible-light-induced radical DYKAT process, affording the desired 3a in 92% isolated yield with 96:4 e.r. (Fig. 3b, entry 10). Omission of ZnCl 2 resulted in a slightly eroded yield and enantioselectivity, but the reason is not clear at the current stage (Fig. 3b, entry 11; see Supplementary Table 3 for more results on the study of ZnCl 2 effect). To exclude the potential effect of the impurities, a higher-purity CoCl 2 (99.998%) was used and the same results as the CoCl 2 with 98% purity were observed (entry 12 versus 10). Moreover, as a comparison, other 3d metals including iron, copper and nickel, 4d metals including ruthenium, rhodium and palladium, and the 5d metal iridium were examined under the optimized conditions (Fig. 3b, entries [13][14][15][16]. We found that all of them failed to generate the desired coupling product 3a except nickel salts, which resulted in a high yield but in a racemic form.      Fig. 3b). All the yields are isolated yields, and all e.r. values were determined by chiral HPLC analysis. Ar, 3,4-2(MeO)-C 6 H 3 .

Substrate scope for AMPC-enabled radical DYKAT.
Under the optimal reaction conditions, we examined the generality of this radical DYKAT process. As highlighted in Fig. 4, a variety of DHP reagents were suitable for this transformation. When benzyl-substituted DHPs were used, variations in the electronic character and substitution pattern on the benzene ring were tolerated, affording the target products in good yields and high enantioselectivities (3b-3h, 70-98% yield, up to 97.5:2.5 e.r.). Among them, the structure of product 3g was unambiguously confirmed through single-crystal X-ray diffraction analysis. Moreover, DHP reagents containing heteroaryls, such as thiophene, indole and carbazole, can also effectively participate in this transformation, delivering the corresponding chiral heterobiaryl products 3i-3o in 65-99% yields and up to 95:5 e.r. In addition to benzyl radicals, asymmetric radical coupling reactions with DHP reagents generating heteroatom-stabilized radicals (O or N) can also proceed well, affording the desired products 3p-3t in 68-99% yields and up to 96.5:3.5 e.r. To prove the practicality of this methodology, two gram-scale reactions were performed with heterobiaryl 1a and DHPs 2k and 2p under the standard AMPC conditions, affording chiral product 3k and 3p in the similar efficiency and enantioselectivity (Fig. 6). Unfortunately, secondary and tertiary carbon radicals are not applicable at the current stage; presumably steric repulsion slowed the reaction such that competitive reductive detriflation of heterobiaryl triflates was observed.
Next, we probed the scope of racemic heterobiaryl components. First, we found that substitution on the isoquinoline ring, that is, fluoro and chloro at the 4-position and various substituents at the 5-7-positions, are compatible, generating chiral products 3u-3af in 71-97% yields and up to 97.5:2.5 e.r. In addition to isoquinoline, heterobiaryl substrates with other nitrogen-containing heterocycles, such as pyridine and pyrazine, were found to be suitable for this radical DYKAT, generating chiral products 3ag-3al in 53-99% yields and up to 98.5:1.5 e.r. Moreover, when substituents were introduced to the 4-, 6-and 7-positions of the naphthalene ring (3am-3ap) or when the naphthalene ring was replaced with a 6-methyl phenyl group (3aq), the reactions also proceeded in high yields and enantioselectivities (92-99% yields and up to 97.5:2.5 e.r.). It is worth noting that, owing to the mild reaction conditions and the avoidance of organometallic reagents that are usually utilized in cobalt-catalysed cross-coupling reactions, impressive tolerance of functional groups on both reaction components (ester groups, cyano groups, aldehydes, siloxy groups, olefins, halides, and so on) was observed.
Reductive couplings for the AMPC-enabled radical DYKAT. Inspired by the above success, we further successfully developed   a reductive cross-coupling process, which demonstrated the generality of the synergistic cobalt-photoredox catalysis strategy on asymmetric radical DYKATs. Through a simple optimization of conditions (see the details in Supplementary Tables 11 and 12), replacement of DHP 2a with more easily available benzyl chloride as the alkyl source and Hantzsch ester (HE) as the reducing reagent facilely gives the same coupling product 3a in good yield and high enantioselectivity. Next, we probed the generality of this radical DYKAT via reductive coupling. As highlighted in Fig. 5, similar reaction efficiency and selectivity to the above transformation were   Demonstration of the synthetic utility of the methodology. As privileged scaffolds, axially chiral heterobiaryls are widely used in asymmetric catalysis 43 . We supposed that the new chiral heterobiaryl molecules achieved through this radical DYKAT could be used as platforms for developing new axially chiral catalysts or ligands with isoquinoline as the hydrogen bond acceptor or the metal coordination site. Then, we treated chiral heterobiaryl 3p with hydrazine hydrate in ethanol giving chiral primary amine 5 in 95% yield with conserved optical purity. Subsequently, condensation of this amine with isothiocyanate 10, aldehyde 11 and acids 12 and 13 produced potential bifunctional thiourea catalyst 6 and polydentate ligands 7-9 in good yields, respectively (Fig. 6a, i-iv, 70-90% yields). Studies of applications of these compounds are ongoing in our laboratory. The aggregation-induced emission (AIE) caused by hindered intramolecular movement is one of the most popular design strategies for luminescence materials. In this technique, axially chiral compounds are often used, mostly by linking the inherent axially chiral skeleton with the luminescent group 46 . Here, we find a significant π-π stacking effect (approximately 3.5 Å) between the aryl ring of the side chain and isoquinoline by analysing the crystal structure of chiral heterobiaryl product 3g. According to this finding, we speculate that a donor-acceptor-type structure would make this type of axially chiral molecules a new class of AIE molecules with a different mechanism of action. To verify this conjecture, 3k, which contains a good electron-donating carbazole group, was chosen as the platform molecule for the AIE experiments. We found that 3k exhibited almost no fluorescence when dissolved in an acetonitrile solution, but as the proportion of water in the system increased, the emitted light intensity continued to increase, showing an obvious AIE effect (Fig. 6b, v). Subsequently, to reduce the band gap of the D-A structure, 3k was converted into the more electron-deficient quinolinium salt Me-3k by a routine methylation procedure involving isoquinoline (Fig. 6b, vi). The   emission wavelength (λ em ) of this salt was obviously redshifted to the yellow region, and the powder exhibited strong fluorescence under 365 nm light irradiation. More significantly, a Me-3k film also showed a strong circularly polarized luminescence signal (|g lum | = 3.02 × 10 −2 ; Fig. 6b, vii), which is difficult to achieve with axially chiral light-emitting organic small molecules 47 . Through the crystal analysis and the non-covalent interactions (NCI) analysis of Me-3k, we also found an obvious intra-and intermolecular π-π stacking effect between the aromatic rings (Fig. 6b, viii and Supplementary Fig. 24 for the NCI plots). Therefore, a layered structure was formed, and the naphthyl group on the axially chiral framework was spaced from the hexafluorophosphonate group to form a side wall. This highly ordered structure may be the cause of the highly circularly polarized signal.

Me-3k
Mechanistic considerations of the AMPC-enabled DYKAT reaction. Next, we performed mechanistic studies to understand the radical DYKAT process. First, a set of control experiments was carried out (Supplementary Table 10), and the results confirmed that visible light and the cobalt catalyst are indispensable for this reaction, omission of the photocatalyst results in a much lower yield (27% yield), albeit with a similar enantioselectivity, and the addition of i-Pr 2 NEt increases the yield by efficiently trapping the TfOH by-product rather than by serving as an electron donor. The last conclusion was supported by the two control experiments shown in Fig. 7a, where the replacement of DHP reagent 2a with N-Me-2a yielded nearly the same results in the absence of i-Pr 2 NEt. Next, stoichiometric experiments were performed by treating DHP reagent N-Me-2a, CoCl 2 and chiral polydentate ligand L8 under photocatalytic reduction conditions. As indicated in Fig. 7b, the chiral Co(I)/L8 complex, a proposed initial catalytic species in this asymmetric radical coupling, was detected by high resolution mass spectrometry (HRMS). By analysing the redox potential of the Co(II)/L8 complex (Fig. 7e, P1, standard reduction potential (E 1/2 red ([Co II /L8]/[Co I /L8])) = −1.07 V versus saturated calomel electrode (SCE) in CH 3 CN/THF, the initial Co(I)/L8 species was generated through reduction by the reduced state of the 2,4,5,6-tetra -9H-carbazol-9-yl-1,3-benzenedicarbonitrile (4CzIPN) photocatalyst (E 1/2 red (4CzIPN/4CzIPN ·− ) = −1.24 V versus SCE in CH 3 CN) 48 or the excited state of the DHP reagent (E(2a + /2a * ) ≈ −1.6 V versus Ag/Ag + in CH 3 CN) 49 . The stoichiometric oxidative addition process with Co(I)(PPh 3 ) 3 Cl, ligand L8 and heterobiaryl substrate 1a was implemented, and the proposed Co(III) intermediate was also detected by HRMS (Fig. 7c). Although we failed to obtain single crystals of chiral cobalt complexes through long-term effort, we confirmed that this radical DYKAT process was catalysed by the cobalt/L8 complex in a 1:1 ratio according to the results of nonlinear experiments (Fig. 7d) and HRMS analysis (Fig. 7b,c).

Conclusions
We have developed a metallaphotoredox-catalysed DYKAT process of racemic heterobiaryls under visible light irradiation. The identification of a chiral polydentate ligand for cobalt-catalysed enantioselective radical couplings was critical to the successful implementation of this research. As a result, a variety of chiral heterobiaryl products were afforded with generally high efficiency and selectivity. In addition, this free radical DYKAT process has been proved to have a certain universality and can be applied to reductive coupling reactions. The value of this methodology was further proven by converting these axially chiral heterobiaryl molecules to promising chiral catalysts and ligands, as well as by finding chiral fluorescent materials. We believe that this research not only opens a new window for asymmetric photochemical synthesis under visible light and cobalt-catalysed asymmetric radical couplings but also creates an avenue to significant chiral heterobiaryl scaffolds and related functional materials.

Methods
General procedure for the AMPC-enabled DYKAT reaction. In an argon-filled glove box, a 10 ml vial equipped with a magnetic stirrer bar was charged sequentially with L8 (0.02 mmol, 10 mol%) and CoCl 2 (0.01 mmol, 5 mol%), followed by addition of THF (2 ml) and CH 3 CN (2 ml). After stirring at room temperature for 30 min, substrates 1a (0.2 mmol) and 2a (0.3 mmol), photocatalyst 4CzIPN (2 mol%, 0.004 mmol), i-Pr 2 NEt (1.0 equiv., 0.2 mmol) and ZnCl 2 (20 mol%, 0.04 mmol) were added to the resulting mixture. Then, the mixture was stirred at a distance of ~10 cm from 90 W blue LEDs at 60 °C for approximately 10 h until the reaction was completed, as monitored by thin-layer chromatography analysis. The concentrated reaction residue was purified by flash column chromatography on silica gel to afford the desired product 3a.
General procedure for the reductive couplings. In an argon-filled glove box, a 10 ml vial equipped with a magnetic stirrer bar was charged sequentially with L8 (0.02 mmol, 10 mol%) and CoCl 2 (0.01 mmol, 5 mol%), followed by addition of THF (2 ml) and CH 3 CN (2 ml). After stirring at room temperature for 30 min, substrates 1 (0.2 mmol) and 4 (0.5 mmol), photocatalyst 4CzIPN (2 mol%, 0.004 mmol), i-Pr 2 NEt (2.0 equiv., 0.4 mmol), HE (1.0 equiv., 0.2 mmol) and ZnCl 2 (1.0 equiv., 0.2 mmol) were added to the resulting mixture. Then, the mixture was stirred at a distance of ~10 cm from 90 W blue LEDs at 60 °C for approximately 2-6 h until the reaction was completed, as monitored by thin-layer chromatography analysis. The concentrated reaction residue was purified by flash column chromatography on silica gel to afford the desired product 3.

Data availability
Crystallographic data for the structures reported in this article have been deposited at the Cambridge Crystallographic Data Centre, under deposition numbers CCDC 2124717 (3g) and 2124714 (Me-3k). Copies of the data can be obtained free of charge via https://www.ccdc.cam.ac.uk/structures/. Data related to materials and methods, optimization of conditions, experimental procedures, mechanistic experiments, DFT calculations, HPLC spectra and spectra are provided in the Supplementary information. All data are available from the corresponding authors upon reasonable request.