Reaction conditions development

To test our hypothesis, we initiated the study by examining the cross-coupling between benzoyl fluoride (1a) and 4-ethyl anisole (2a) with the triazolium salt A as the catalyst, which has been shown to be efficient in multi component couplings³⁴. [Ir(dF(CF₃)ppy)₂(dtbbpy)]PF₆ was selected as a photocatalyst because its excited state is a strong oxidant (E_1/2[Ir^*III/II] = +1.21 V)⁴⁷. Although 4-ethyl anisole shows a higher oxidation potential (E_1/2 = +1.52 V)⁴⁸, we speculated that the endothermic SET can be productive if the subsequent deprotonation is fast⁶. Indeed, after irradiation for 24 h in dichloromethane, a 15% yield of the targeted 3a was obtained. However, conversion was low (Table 1, entry 1). Solvent optimization revealed that improved yields can be achieved in dimethylformamide and acetonitrile with the latter showing the best result (entries 2–8). The NHC-screening identified the triazolium salt A as the ideal precatalyst, and all other triazoliums, imidazoliums, and thiazolium salts B-H tested, provided lower yields. In two cases (E and G), no conversion was noted (entries 9–15, Fig. 2). Yield and conversion could be further improved upon running the cascade at higher concentration (entries 16–18). It is notable that the donor–acceptor dye 2,4,5,6-tetra(carbazol-9-yl)isophthalonitrile (4CzIPN)⁴⁹ showed the same catalytic efficiency as the Ir-based photocatalyst (entry 19). Decreasing the amount of base, NHC precatalyst, or photocatalyst led to lower yields (SI, Supplementary Table 1). Finally, control experiments validated the necessity of carbene and the photocatalyst for successful C–H benzoylation of 2a. No product was observed in the absence of light, photocatalyst, or carbene (entries 20–22). Importantly, the same two substrates that qualify for classical Friedel-Crafts acylation using a Lewis acid as an activator will provide a different product upon switching to the NHC/photocatalyst activation mode in a chemodivergent manner. For example, the reaction of benzoyl fluoride (1a) with 4-ethyl anisole (2a) in the presence of a stoichiometric amount of AlCl₃ afforded the Friedel-Crafts product in 72% yield. Acylation occurred with excellent regioselectivity at the ortho-position of the activating methoxy group (see SI for details).

Table 1

Optimization of the reaction conditions^a.

Entry	Solvent	NHC precatalyst	2a, (M)	Conversion (%)b	Yield (%)c
1	CH2Cl2	A	0.05	17	15
2	CHCl3	A	0.05	24	22
3	DMSO	A	0.05	20	20
4	DMF	A	0.05	43	42
5	1,4-Dioxane	A	0.05	5	5
6	Toluene	A	0.05	5	4
7	CH3CN	A	0.05	60	59
8	Ethyl acetate	A	0.05	4	4
9	CH3CN	B	0.05	43	42
10	CH3CN	C	0.05	3	3
11	CH3CN	D	0.05	7	6
12	CH3CN	E	0.05	0	0
13	CH3CN	F	0.05	3	3
14	CH3CN	G	0.05	0	0
15	CH3CN	H	0.05	2	0
16	CH3CN	A	0.07	77	77
17	CH3CN	A	0.10	86	85
18	CH3CN	A	0.17	92	88 (83)
19d	CH3CN	A	0.17	–	83
20e	CH3CN	A	0.17	0	0
21f	CH3CN	A	0.17	0	0
22g	CH3CN		0.17	0	0

Fig. 2

Different NHC catalysts tested.

A–D are precursors leading to triazol-5-ylidene-type carbenes, E and F provide imidazole-2-ylidines, and G and H deliver thiazol-2-ylidenes. NHC N-heterocyclic carbene.

Substrate scope

With the optimized conditions established, we examined the reaction scope with respect to the acyl fluoride first (Fig. 3a). A wide range of benzoyl fluorides bearing electron-donating or electron-withdrawing substituents could be used for the C–H aroylation of 4-ethyl anisole, affording the desired products in moderate to excellent yields (3b–3m, 48–95%). As is evident from these results, aryl halides, most of which would be incompatible with Ni catalysis^29–33, engage in this cascade, albeit a moderate yield was obtained for the iodo-congener (3f, 48%). α-Naphthoyl and β-Naphthoyl fluoride were successfully used in the direct aroylation of 2a (3n, 47%; 3o, 76%). The latter was more reactive due to lower steric hindrance during formation of the β-naphthoyl azolium intermediate. Furthermore, heteroaroyl fluorides containing the furan and thiophene moieties were also amenable to the coupling with 2a to afford 3p and 3q in 47% and 82% yields, respectively.

Fig. 3

The reaction scope.

a Benzoyl scope. b Arene scope. c Late-stage functionalization. Reaction conditions: all the reactions were carried out with benzoyl fluoride (0.4 mmol), arene (0.1 mmol), NHC precatalyst A (0.02 mmol), Cs₂CO₃ (0.2 mmol), and 4CzIPN (0.002 mmol) in anhydrous CH₃CN (0.6 mL), irradiation with blue LEDs at room temperature for 24 h. Isolated yields were given. ^*[Ir(dF(CF₃)ppy)₂(dtbbpy)]PF₆ (0.002 mmol) was used instead of 4CzIPN. See SI for more details. 4CzIPN, 2,4,5,6-tetra(carbazol-9-yl)isophthalonitrile. LEDs light-emitting diodes.

The scope of the reaction with respect to the benzylic component was explored next using benzoyl fluoride 1a as the acyl donor (Fig. 3b). A variety of functional groups are tolerated, giving rise to products bearing ester (3r, 82%; 3w, 60%), amide (3s, 84%), azide (3t, 47%), ketone (3u, 66%), and ether (3v, 31%) functionalities. It is important to note that the 1,2-dihydrobenzfuran and chroman substructures can be found in pharmaceutical drugs such as Darifenacin and Nibivolol^50,51. With this in mind, we tested them as radical coupling partners and both systems worked well to afford the ketones 3x and 3y in 82% and 87% yields, respectively. Benzoylation in α-position to the O-atom was not observed, clearly showing that C-radical formation occurs with complete regioselectivity at the benzylic position. Compared with 4-ethyl anisole, 1-ethyl-4-phenoxybenzene showed a lower efficiency (3z, 47%). Primary benzylic C–H bonds could also be acylated to give the desired products (3aa, 48%). However, reaction was sluggish and 48% of starting material was recovered. Owing to the steric hindrance of tertiary C–H bonds, lower yields were obtained for 3ab (28%). Importantly, an activating O-substituent decreasing the arene oxidation potential is not required and (4-bromobutyl)benzene could be benzoylated with 1a, albeit with moderate yield (3ac, 41%).

We next turned our attention to examine the site selectivity of the direct C-H benzoylation. Even in the presence of weaker benzylic C–H bonds, exclusive functionalization of the ethyl moiety in para-position to the alkoxy group was noted (3ad, 65% yield). Along these lines, methoxy-substituted dihydroindene and tetrahydronaphthalenes reacted with excellent site selectivity at the benzylic position that is located ortho or para to the activating methoxy group (3ae, 44%; 3af, 74%; 3ag, 70%). Likely, intermolecular C–H abstraction with an electrophilic radical would be not regioselective in these cases, showing that the deprotonation of an arene radical cation under our conditions is a highly regioselective process. It is noteworthy that monobenzoylation of 5,6-dimethoxy-2,3-dihydro-1H-indene was achieved (3ah, 62% yield). 1,2-Dihydrobenzfuran and chroman bearing a methyl substituent adjacent to the reactive benzylic site delivered the targeted ketones with high diastereoselectivities in good yields (3ai, 69%, trans/cis = 98:2; 3aj, 57%, trans/cis = 91:9). Finally, to further demonstrate the potential of our method, we applied the radical/radical cross-coupling reaction to the late-stage benzylic benzoylation of biologically interesting compounds (Fig. 3c). Highly regioselective benzoylation of the methylene moiety of a δ-tocopherol analog was achieved and benzoylation of the methyl group did not occur (3ak, 34%, dr = 1:1). Additionally, both epiandrosterone and dopamine analogs could be regioselectively functionalized at the benzylic positions, with no byproduct observed derived from reactions adjacent to the ketone or amide moieties (3al, 73%; 3am, 52%).

Control experiments

To gain insights into the reaction mechanism, control experiments were performed. C-H-benzoylation of chroman with 1a was fully suppressed in the presence of 2,2,6,6-tetramethyl-piperidin-1-oxyl (TEMPO) (Fig. 4a). When CD₃CN was used in place of CH₃CN, there was no deuterium incorporation into the product, as well as in the recycled starting chroman, which indicates that H-atoms or protons of the solvent do not participate in the cascade (Fig. 4b). Moreover, a kinetic isotope effect (KIE) was observed in the intermolecular competition experiment, demonstrating that the deprotonation might be involved in the rate-determining step (Fig. 4c) ⁵².

Fig. 4

Control experiments.

a Radical inhibition experiment. b Deuterium incorporation experiment. c KIE experiment. TEMPO 2,2,6,6-tetramethyl-piperidin-1-oxyl, KIE kinetic isotope effect.

Reaction mechanism

According to the above results and previous reports, a possible mechanism is suggested in Fig. 5. Under blue LEDs irradiation, [Ir(dF(CF₃)ppy)₂(dtbbpy)]PF₆ will be excited and the excited state will be reductively quenched by the electron-rich arene 2, leading to an arene radical cation III with concomitant formation of a radical anion of [Ir(dF(CF₃)ppy)₂(dtbbpy)]PF₆ (E_1/2(P/P^•-) = − 1.37 V vs SCE) or 4CzIPN (E_1/2(P/P^•-) = − 1.21 V vs SCE)^47,49. The radical cation III will be deprotonated (Cs₂CO₃) at the benzylic position to generate a transient benzylic radical IV. Based on the KIE studies, the initial electron transfer from 2 to PC* is reversible. The reduced photocatalyst will then transfer an electron to the acyl azolium salt I (E_1/2 = −1.29 V vs SCE)⁴⁰ itself formed in situ from the acyl fluoride 1 and the NHC catalyst to generate a persistent ketyl type radical II, closing the photoredox catalysis cycle. Radical/radical cross coupling of the transient radical IV with the persistent ketyl radical II steered by the persistent radical effect⁵³ will lead to intermediate V. Fragmentation of the NHC will eventually provide the product ketone 3 thereby liberating the NHC catalyst.

Fig. 5

Plausible reaction mechanism.

Cooperative NHC/photoredox catalysis where the acylazolium ion I is involved in both catalytic cycles. PC photocatalyst, NHC N-heterocyclic carbene, SET single-electron-transfer.

In this work, we have developed a strategy to accomplish acylation of benzylic C–H bonds via cooperative NHC and photocatalysis. The key step of the cascade is a radical/radical cross-coupling. The acylation occurs with excellent site selectivity and broad functional group compatibility. The protocol is amenable to functionalize important structural motifs with good to excellent diastereoselectivities, as well as to the late-stage benzoylation of more complex natural product derived compounds. The method will open avenues in the area of direct C–H bond acylation and also complements existing transition metal catalyzed C–H bond functionalization methodologies.

General procedure for the cross-coupling between an acyl fluoride and a benzylic component

To a Schlenk tube the carbene precatalyst A (6.3 mg, 0.02 mmol), 4CzIPN (1.6 mg, 0.002 mmol) or [Ir(dF(CF₃)ppy)₂(dtbbpy)]PF₆ (2.2 mg, 0.002), and Cs₂CO₃ (65.2 mg, 0.2 mmol) were added. Then the reaction tube was evacuated and backfilled with argon two times. Subsequently, a benzylic component (0.10 mmol) and an acyl fluoride (0.40 mmol) (if solid, they should be added at the beginning) and CH₃CN (0.6 mL) were added. The resulting mixture was degassed under vacuum two times and then the mixture was irradiated with blue LEDs at room temperature for 24 h. After that, the residue was purified by silica gel chromatography using a mixture of n-pentane and ethyl acetate or pentane and diethylether as an eluent to get the desired product. Each reaction was carried out twice and the average value was used as the final yield.