Allylic C(sp3)–H arylation of olefins via ternary catalysis

Transforming C(sp3)–H bonds efficiently and selectively into C(sp3)–C(sp3) or C(sp3)–X bonds is a highly relevant task. The direct arylation of allylic C(sp3)–H bonds provides an elegant method for the formation of unconjugated aryl-substituted olefins. Although both ionic- and radical-based transition metal catalysis has been applied to achieve this transformation, numerous challenges remain. The requirement for persistent radical coupling partners, moderate selectivity and the need for tri- or tetrasubstituted olefins have limited the generality of existing methods. Now we report a ternary catalytic method that combines organic photoredox, hydrogen atom transfer and nickel catalysis, and can directly arylate allylic C(sp3)–H bonds of readily available olefins. This process operates under mild conditions and exhibits a remarkable reaction scope in both aryl halide and olefin coupling partners. Mechanistic experiments, coupled with density functional theory calculations of Ni-oxidation states and reaction energetics allowed the elucidation of a ternary catalytic cycle and the origin of regioselectivity. A ternary catalytic method combining organic photoredox, hydrogen atom transfer and nickel catalysis is reported. This combination can directly arylate the allylic C(sp3)–H bonds of a broad range of readily available olefins. Mechanistic experiments, coupled with density functional theory calculations aid the elucidation of the ternary catalytic cycle and the origin of regioselectivity.

T he attainment of more sustainable, atom-economic approaches remains among the most challenging goals in synthetic chemistry 1,2 . Transition-metal-catalysed allylic functionalization has become one such valuable synthetic tool to construct useful architectures, and has been widely applied in total synthesis, drug discovery and material science 3,4 . In general, preactivated starting materials are used, which thus requires additional steps and so increases the overall synthetic cost. In comparison, direct allylic C(sp 3 )-H bond functionalization provides a more straightforward and attractive approach for the synthesis of highly functionalized alkenes, and is established as a state-of-the-art alternative [5][6][7][8][9][10][11] . As allylarene units can be found in bioactive molecules 12 and serve as versatile intermediates in synthetic chemistry 13 , their synthesis from inexpensive precursors has been thoroughly investigated. An example is Nakamura and co-workers' seminal report of the iron-catalysed direct allylic arylation of olefins via C-H activation; Grignard reagents were required, which limited the functional group tolerance of the method (Fig. 1a) 14 . Further related developments have since been made by our group and others via an oxidative mechanism using boronic acids [15][16][17][18] or electron-poor arenes 19,20 .
To avoid the use of organometallic reagents and strong bases, and to extend the attainable scope of the transformation, transition-metal-catalysed allylic functionalization reactions that involve radicals could offer complementary advantages over traditional ionic allylations 21,22 . Building on the pioneering work by Arnold and co-workers 23 in the reaction between dicyanobenzene with olefins, Inoue 24 , MacMillan 25 , Kanai 26 and their co-workers developed elegant methods to achieve allylic C(sp 3 )-H arylation through radical-radical couplings by merging hydrogen atom transfer (HAT) catalysis with photocatalysis (Fig. 1b). Nevertheless, these methods are plagued by two main limitations: (1) only persistent radicals (for example, from cyanoarenes) can be used as the coupling partners, which thus limits the reaction scope, and (2) only a moderate selectivity is usually achieved, owing to the uncontrollable radicalradical coupling.
Recently, metallaphotoredox catalysis emerged as a tool to overcome energetically unfavourable steps in two-electron crosscouplings under mild conditions [27][28][29][30][31] . As pioneered by Molander 32 , Doyle 33 and their co-workers, photoexcited Ni(II) or Ni(III) complexes can generate halogen radicals, which then act as HAT catalysts in several C-H bond functionalizations 34 . However, to date only tri-and tetrasubstituted alkenes have been employed in allylic C-H arylation protocols (by Huang and Rueping 35 ), which therefore limits the synthetic applicability of such methods 35 . To develop a general approach towards allylic C(sp 3 )-H arylation, we hypothesized the following key steps: (1) generation of allyl radicals from simple olefins using a HAT catalyst; (2) allyl radical trapping with a metal complex; (3) cross-coupling with a suitable aryl halide via an oxidative addition/reductive elimination sequence; (4) turnover of the HAT and metal catalysts by a suitable photocatalyst (PC) (Fig. 1c) 36 . Owing to the interplay between several catalytic cycles, this ternary catalytic system has to face major obstacles regarding mutual compatibility, selectivity and efficiency 37 .

Results and discussion
Reaction development. The simple feedstocks 1-hexene 1a and aryl bromide 2a were chosen for the initial optimization of the direct allylic C(sp 3 )-H arylation (DAA; Fig. 2a). Recently, the concept of ternary catalysis was developed by MacMillan and co-workers 38 and Kanai and co-workers [39][40][41][42] to accomplish several elegant C-H transformations by employing a suitable HAT catalyst, metal catalyst and photocatalyst (PC). Inspired by these exciting examples, we were pleased to find that the DAA reaction could be performed in the presence of 5 mol% nickel catalyst, 4 mol% thiophosphoric imide (HAT-1) and 7.5 mol% organophotocatalyst (PC-1); additionally, product 3 was the only detectable regioisomer formed (68% isolated yield, 80:20 E/Z, >20:1 regioselectivity ratio (r.r.); Fig. 2b, entry 1). Different ligands, solvents and catalyst loadings were investigated and it was shown that the above reaction conditions are optimal (71%) to achieve the DAA reaction (Fig. 2b, entries 2-4 and 12-14).

Allylic C(sp 3 )-H arylation of olefins via ternary catalysis
Huan-Ming Huang 1,3 , Peter Bellotti 1,3 , Pan-Pan Chen 2,3 , Kendall N. Houk 2 ✉ and Frank Glorius 1 ✉ Transforming C(sp 3 )-H bonds efficiently and selectively into C(sp 3 )-C(sp 3 ) or C(sp 3 )-X bonds is a highly relevant task. The direct arylation of allylic C(sp 3 )-H bonds provides an elegant method for the formation of unconjugated aryl-substituted olefins. Although both ionic-and radical-based transition metal catalysis has been applied to achieve this transformation, numerous challenges remain. The requirement for persistent radical coupling partners, moderate selectivity and the need for tri-or tetrasubstituted olefins have limited the generality of existing methods. Now we report a ternary catalytic method that combines organic photoredox, hydrogen atom transfer and nickel catalysis, and can directly arylate allylic C(sp 3 )-H bonds of readily available olefins. This process operates under mild conditions and exhibits a remarkable reaction scope in both aryl halide and olefin coupling partners. Mechanistic experiments, coupled with density functional theory calculations of Ni-oxidation states and reaction energetics allowed the elucidation of a ternary catalytic cycle and the origin of regioselectivity.
Interestingly, only the presence of HAT-1, which was developed by Kanai and co-workers 39 , gave high conversion (entries 5-8). Several common photocatalysts were also screened ( Fig. 2b, entries 9-11) and, interestingly, an easily prepared organic photoredox catalyst (E 1/2 (PC*/PC ·-) = 2.0 V versus the saturated calomel electrode in MeCN)-previously designed and applied in photocatalytic carbonyl-olefin cross-metathesis by our group 43 -showed the best result in this ternary catalytic system. This organophotocatalyst 44 is similar to that applied in site-selective arene C-H amination by Nicewicz and co-workers 45 . Control experiments were also performed to prove the necessity of light, catalyst and base in this DAA reaction (Fig. 2b, entries [15][16][17][18][19][20]. A condition-based sensitivity screening 46 was applied ( Fig. 2f): the ternary catalytic system is not sensitive towards concentration fluctuations, irradiation under high intensity and reaction scale (5 mmol). In addition, simple conduction of the reaction under argon is sufficient to ensure optimal activity. A moderate yield reduction was observed under reduced light intensity, and temperature variations. The use of wet solvents, or the performance of the reaction under an oxygen-containing atmosphere, caused the complete inhibition of the catalytic manifold.
The mildness of the protocol was further confirmed by the successful employment of alkenes that bare difluorophenol (37)

Fig. 1 | Direct allylic C(sp 3 )-H arylation, state-of-the-art and design of this work. a, Selected example of well-known allylic arylation via an ionic mechanism.
The key steps involve C-H bond activation and C-C bond formation via reductive elimination. b, Early example of allylic arylation via a radical mechanism. The key steps are the allyl hydrogen atom abstraction, followed by a radical-radical coupling. c, This work: general and redox-neutral approach for allylic C(sp 3 )-H arylation via a radical strategy (right). The key steps are allyl hydrogen abstraction, radical trapping by nickel and C-C bond formation via reductive elimination. The challenge of tertiary catalysis is to achieve compatibility of the HAT system with nickel and the photocatalyst (PC) system (right). A, nitrogen or carbon; acac, acetylacetonate; Ar, aryl; ppy, 2-phenylpyridinato-C 2 ,N; Mes, mesityl; Tf, triflate; xantphos, (9,9-dimethyl-9Hxanthene-4,5-diyl)bis(diphenylphosphane).

Mechanistic studies.
To shed light on the mechanism of the transformation, a series of experiments was performed. The radical trapping agent TEMPO (2,2,6,6-tetramethylpiperidinyloxyl) completely inhibited the reaction and no TEMPO · radical adduct could be detected. This might suggest-in accordance with Kattamuri and West 51 -poisoning of the catalytic system by the persistent radical species 52 . When the independently synthesized product 59 was submitted to the reaction conditions, no further allylic coupling product was detected, which indicated that a further productive HAT process did not occur, as previously reported by Cuthbertson and MacMillan (Fig. 5a) 25 . Furthermore, the predominantly Z-alkene 59 underwent Z → E isomerization to an approximately 80:20 E/Z mixture only in presence of PC-1 (radar diagram in Fig. 5a), which clearly indicates the possibility of photocatalytic stereoisomerization-probably via a single electron transfer (SET) (59, E (1/2) = 2.08 V versus the saturated calomel electrode) 53 -without concomitant isomerization, which further excludes additional HAT events on the product, which was further analysed by density functional theory (DFT) calculations (Supplementary Section 8.6). Analysis of the ultraviolet-visible spectrum of the reaction mixture and the individual components-in addition to the initial control experiments (Fig. 2b)-proved that species PC-1 acted as a photoactive component of the reaction (Supplementary Section 7.6). Furthermore, the observed primary kinetic isotope effect (KIE) when 1am-d 2 was used as the substrate hints that the HAT process to generate the required allyl radical is the likely rate-determining step (Fig. 5b) 42,54 . We then evaluated the ability of Ni(0) species within the catalytic cycle: when Ni(cod) 2 (cod, 1,5-cyclooctadiene), as the Ni(0) source, and L 1 were used, conversion towards the desired product 3 could be detected 55 , whereas the system proved to be unreactive in the absence of an ancillary ligand (Fig. 5c) 56 .  Similarly, the use of the oxidative addition complex Ni-I (Fig. 5c) as the catalyst generated 3 in 59% yield, which thus suggests that a Ni(II)-aryl species may be involved in the catalytic manifold 57,58 .
Computational studies and proposed mechanism. DFT   underlying mechanism of this transformation (Supplementary Section 8). Three key questions were addressed: (1) In the nickel catalytic cycle, do Ni(0) [59][60][61][62][63] or Ni(I) 64-67 species act as an active coupling catalyst? (2) What are the rate-determining steps, and does the radical reaction with nickel occur before or after oxidative addition?
(3) What is the origin of the observed regioselectivity? A model reaction that featured methyl 4-bromobenzoate (2a′) and 1-butene (1a′) was used for the computational study. As shown in Fig. 6a,b, starting from the Ni(I) species [64][65][66][67] , INT1 (Fig. 6c) . We performed an indepth independent analysis of each potentially productive pathway, with respect to both the organometallic elementary steps and the redox process (detailed in Supplementary Section 8). For path A, the rate-limiting step is the oxidative addition (Ni(0)-Ni(II)), and the barrier is 39.4 kcal mol −1 , which is too high to overcome under the reaction conditions, which excludes this possible reaction scenario. Analogously, the rate-limiting step for path B is the oxidative addition (Ni(I)-Ni(III)), and the corresponding barrier is 29.5 kcal mol −1 , which cannot be surmounted in the reaction system. Furthermore, the unfeasibility of both reaction paths is corroborated by the highly unfavourable single-electron reduction profile of Ni(I) by the reduced photocatalyst (PC r ), both under thermodynamic (ΔG r = 26.3 kcal mol −1 ) and kinetic (Gibbs free energy of activation from the Marcus−Hush theory 68 ΔG ≠ MH = 32.4 kcal mol −1 ) grounds. Similarly, path D features the rate-limiting oxidative addition (Ni(I)-Ni(III)), which is prohibitively high in energy (31.9 kcal mol −1 ), in addition to an overall unfavourable single electron reduction profile (Ni(II)-Ni(I)). However, path C involves an accessible rate-determining oxidative addition (Ni(I)-Ni(III)) and thermodynamically (ΔG r = −19.7 kcal mol −1 ) and kinetically (ΔG ≠ MH = 2.6 kcal mol −1 ) favoured SET. In particular, starting from INT1, aryl bromide coordination and oxidative addition occurs first via TS1, with a 23.9 kcal mol −1 energy barrier, to generate the Ni(III) intermediate INT3, which then undergoes the abovementioned SET to give INT4. The productive pathway is completed by a facile radical addition (via TS2) and reductive elimination (via TS3) to give the final product (3′) and the active nickel catalyst INT1 (Fig. 6c).
As mentioned above, of all the mechanisms discussed, C (Fig. 6a) is the most reasonable reaction path, in which oxidative addition via TS1 is the rate-limiting step and reductive elimination via TS3 is the regioselectivity-determining step. The calculated results suggest that, in this reaction system, the oxidation states of nickel should be Ni(I)-Ni(III)-Ni(II)-Ni(III).
Origins of regioselectivity. The free energy diagram (Fig. 6c) indicates that reductive elimination determines the regioselectivity, as this elementary step could occur at either terminus of the unsymmetrical allyl radical. Therefore, we compared the corresponding competitive transition states, TS3 and TS3′ (Fig. 7a), to explore the origins of regioselectivity. TS3 is lower in energy than TS3′ by 8.2 kcal mol −1 . The calculated selectivity is consistent with the experimental results, which favour the linear product formation.   2). b, KIE using the initial rate measurement. c, Studies using ni(0) and ni(II) sources. For each nickel source, a 5 mol% loading was used. yields were determined by 1 H nMr spectroscopy using dibromomethane as the internal standard.

Ni-I
As shown in Fig. 7a, the major difference in geometry between TS3 and TS3′ involves the bound allylic fragment. In TS3, the primary carbon of the allylic ligand reacts in the bond formation/cleavage process, whereas in TS3′, the secondary carbon is involved in the same process. Therefore, from the perspective of steric interaction, TS3′ has a greater steric hindrance compared with that of TS3.
As shown in Fig. 7a In addition to steric hindrance, hyperconjugation could also positively impact the stability of the transition state (Fig. 7b). In TS3, the allylic ligand retains hyperconjugation between the olefin and the terminal methyl, and thus stabilizes the corresponding transition   state. In TS3′, hyperconjugation between the olefin and methyl is absent because the secondary carbon is involved in the bond formation/cleavage process, and thus the energy of the corresponding transition state (TS3′) is higher. To further support the important influence of hyperconjugation, we compared the energy of the linear (3′) and branched product (3′′). The linear product (3′), which retained hyperconjugation (σ C−H → π* C=C ), is 4.1 kcal mol −1 more favourable than the branched one (3′′), consistent with our hypothesis. The steric hindrance and the hyperconjugation effects also apply to other transition states, as detailed in Supplementary Fig.21.

Conclusion
In summary, we have shown that a ternary catalytic system that combines organic photoredox, HAT and nickel catalysis can be successfully applied to a direct allylic C(sp 3 )-H arylation to yield unconjugated alkenes, which are highly challenging to obtain under mild and redox neutral conditions. The protocol exhibits a high tolerance towards various functional groups with regards to both coupling partners, and offers a complementary method to transition-metalcatalysed allylic functionalization. Additionally, the severe substrate limitations inherent to dual HAT-photoredox allylation manifolds (a persistent radical) can be overcome, which thus offers an increased generality and synthetic usefulness. In-depth mechanistic and theoretical studies clearly demonstrate that a radical mechanism is involved, and elucidate that the origin of regioselectivity can be attributed to steric hindrance and hyperconjugation effects. Based on DFT calculations, we found that in the most favourable reaction pathway, the oxidation states of nickel should be Ni(I)-Ni(III)-Ni(II)-Ni(III) along the reaction coordinate. Ni(I) is the active catalyst in the studied reaction system, because further reduction of Ni(I) to Ni(0) is both kinetically and thermodynamically unfavoured. After the oxidative addition of Ni(I) with aryl bromide, the reaction undergoes radical addition and reductive elimination to give the product. Of these, radical addition is reversible, and the reductive elimination determines the regioselectivity, which is the result of steric hindrance and hyperconjugation. We envision that this newly developed ternary catalysis could potentially become a general and versatile platform for other C-H functionalization and metallaphotoredox-catalysed reactions.

Methods
In the general procedure for the allylic sp 3 -arylation of olefins, NiBr 2 ·L1 (5-15 mol%), HAT-1 catalyst (4 mol%), PC-1 catalyst (7.5 mol%) and the coupling partners bromide (0.2 mmol, 1.0 equiv., if solid or thick gum) and alkene (1.00 mmol, 5.0 equiv., if solid or thick gum), were charged under air into an oven-dried Schlenk tube equipped with a PTFE-coated stirring bar. The vessel was evacuated and back-filled with argon three times, then dried and degassed DCE (3.0 ml, 0.067 M) was added under an argon counterflow, followed by 2,4,6-collidine (53 μl, 0.40 mmol, 2.0 equiv.), the appropriate bromide (0.2 mmol, 1.0 equiv., if liquid) and/or alkene (1.0 mmol, 5.0 equiv., if liquid). The reaction vessel was sealed, then irradiated for the indicated amount of time at 450 nm using the standard set-up. After irradiation, the insoluble precipitates were removed by filtration over a short pad of silica, rinsing thoroughly with CH 2 Cl 2 . The volatiles were removed in vacuo and the residue was purified by flash column chromatography on silica.

Data availability
Materials and methods, detailed optimization studies, experimental procedures, mechanistic studies and copies of the NMR spectra are available in the Supplementary Information. NMR data in a mnova file format and gas chromatography-mass spectroscopy data for KIE analysis are available at Zenodo at https://zenodo.org/record/5614753#.Yaq8DN8kGUk, under the Creative Commons Attribution 4.0 International license.