α, β-Unsaturated ketones are common structures in functional organic molecules and are easy to synthesize. In contrast, robust synthetic methods for β,γ-unsaturated ketones are lacking, despite the fact that these moieties are found in many bioactive molecules and natural products and can be used as building blocks for complex structures (Fig. 1a)1,2,3,4,5,6. Many of the reported methods are based on disconnection of the bond between the α and β carbons which means α-alkenylation of an enolate or enolate equivalent;7,8,9,10,11,12 whereas disconnection of the bond between the carbonyl group and the α carbon—that is, allylation of an acyl donor13,14,15,16,17—has not been thoroughly explored. In addition, most methods for synthesizing β,γ-unsaturated ketones require a prefunctionalized starting material, which limits the applications of the methods to relatively simple targets. Moreover, these methods suffer from low β,γ regioselectivity18,19. To address these issues, investigators have recently developed a number of mild catalytic reactions. For example, the Helquist group reported a procedure for Ni-catalyzed cross-coupling of ketone enolates with allyl bromides (Fig. 1b, top)20, Fu and a co-worker reported Ni-catalyzed enantioselective cross-coupling of secondary α-bromoketones with vinyl zirconium reagents (Fig. 1b, bottom)21, and MacMillan and colleagues realized enantioselective α-vinylation of aldehydes with catalysis by Cu(I) and chiral amines (Fig. 1c)22,23,24. However, these elegant methods are based on disconnection of the α,β bond, which has hampered their wider applications. Only a few examples of retrosynthetic analyses based on radical reaction of homolytic disconnection of the bond between the α carbon and the carbonyl carbon have been reported25,26. Given that β,γ-unsaturated ketones are privileged scaffolds, their synthesis remains an important challenge, and there is a growing need to blossom new C–H bond activation and late-stage functionalization reactions in an unconventional manner.

Fig. 1: Bioactive compounds with β,γ-unsaturated ketone motifs and approaches for their synthesis.
figure 1

a Examples of pharmaceutically active agents possessing β,γ-unsaturated ketone motifs. b, c Classic catalytic approaches for β,γ-unsaturated ketone synthesis. d Strategy used in this study.

We reasoned that combining N-heterocyclic carbine (NHC) catalysis with other types of catalysis might be useful for β,γ-unsaturated ketone synthesis. NHCs are unique Lewis basic catalysts that use polarity reversal to mediate various organic transformations27,28,29. The revival of photocatalysis30,31,32,33,34,35,36 and electrocatalysis has accelerated the development of free radical chemistry37,38,39,40,41,42,43,44,45,46, and visible-light catalysis has been combined with NHC catalysis to achieve NHC-mediated radical reactions under mild conditions. For example, reactions of radicals generated by single-electron oxidation of Breslow intermediates have been reported by the groups of Scheidt47,48, Studer49,50, Chi51,52,53,54,55, Rovis56,57, Sun58, Ye59,60,61,62, Ohmiya63,64,65,66, and others67,68,69,70,71,72,73,74. In contrast, single-electron reduction of NHC-bound acyl azolium intermediates in radical reactions has not been as thoroughly explored75,76,77,78,79,80,81,82,83,84. In above processes, the NHC catalytic cycle is intertwined with the second catalytic cycle, and biosyntheses of complex natural products often proceed via multicatalytic pathways, combining three or more catalysts is an attractive strategy for the development of new reactions85,86. However, the use of NHCs in multicatalytic systems is in its budding stage87, and its challenge lies in compatibility issues between catalysts and intermediates, so we aim to use the unique single-electron reduction characteristics of acyl azolium intermediates with the combination of photocatalysis and hydrogen atom transfer catalysis to further bridge the existing gaps in this field. Aromatic carboxylic acids were chosen as substrates because they are inexpensive, readily available, highly reactive, can be used to quickly generate libraries of structurally complex small molecules88,89.

In this work, we disclose a strategy for coordinating triple catalysis to direct allylic acylation of alkenes with carboxylic acids (Fig. 1d). Specifically, a carboxylic acid would be activated in situ by N,N-carbonyldiimidazole (CDI), and then the addition of a NHC would afford an acyl azolium intermediate90, which would undergo single-electron reduction upon irradiation with visible light in the presence of a photocatalyst to generated an azolium radical anion. Meanwhile, single-electron oxidation of a thiolate generated in situ would provide a thiyl radical, which would undergo a HAT reaction with the olefin to produce a nucleophilic allylic radical, and coupling of this radical with the azolium radical anion would form the β,γ-unsaturated ketone. The key challenge posed by this strategy was the identification of three highly selective but independent catalysts that functioned harmoniously. Several formidable challenges comprise (1) The HAT catalyst would have to discriminate between the reactants and the products. (2) Thiyl radical would selectively undergo a HAT reaction rather than direct coupling with the azolium radical anion. (3) The formation of α,β-isomers as by-products would have to be avoided. (4) Because the redox activity of the three catalysts would be interdependent, regeneration of the ground-state photocatalyst, the NHC catalyst, and the HAT catalyst would require careful synchronization of the three catalytic cycles.


Optimization of reaction conditions

We commenced our investigation by using 1a, an acyl imidazole derivative of p-toluic acid, and cyclohexene (2a) as model substrates (Table 1). After extensive exploration of the reaction parameters, we found that upon irradiation of 1a and 2a in DCM with blue LEDs in the presence of PC-I as the photocatalyst, NHC A as the NHC catalyst, triisopropylsilanethiol (HAT-1) as the HAT catalyst, and potassium phosphate and cesium carbonate as bases (20 mol % each), we could obtain target β,γ-unsaturated ketone product 3 in 77% isolated yield (entry 1). When other photocatalysts were used, the yield of 3 decreased substantially (entries 2–4). Replacing NHC A with other catalysts led to sharply lower yields (entries 5–9). Evaluation of solvent effects revealed that CH3CN and THF were inferior to DCM (entries 10 and 11). Under the same conditions, other thiol catalysts HAT-2–HAT-5 gave considerably lower yields (entry 12). This result is mainly due to the mismatch of BDE values (e.g., methyl thiosalicylate S−H BDE = 78.7 kcal/mol; cyclohexene allylic C−H BDE = 83.2 kcal/mol and toluene benzylic C−H BDE = 89.9 kcal/mol) and other on long-standing and well-established physical properties (that is, oxidation potentials, hydrogen atom transfer exchange constants)91,92,93,94. Replacing K3PO4 or Cs2CO3 with other bases substantially decreased the yields (entries 13–17). Control experiments showed that all three catalysts, both bases, and light were crucial to the reaction (entries 18–22).

Table 1 Optimization of reaction conditionsa

Substrate scope with respect to the carboxylic acid

After optimizing the reaction conditions, we investigated the suitability of acyl imidazole derivatives of various aryl carboxylic acids (Fig. 2). In reactions with 2a, acyl imidazoles 1 with a substituent on the meta or para position of the aryl ring were tolerated, affording the corresponding β,γ-unsaturated ketone products (420) in moderate to high yields, regardless of the electronic nature of the substituent. Moreover, a product with an unprotected ortho hydroxyl group (21) could be obtained, albeit in moderate yield. Disubstituted substrates were also amenable to the reaction conditions, affording compounds 2225. Notably, halogen atoms remained intact, and thus products 1113, 18, 19, and 25 had the potential to undergo subsequent functionalization. The mild conditions were compatible with a range of functional groups, including phenyl (6), ethers (710, 16 and 24), esters (17), and alkyl-substituted phenyl rings (5, 15, 22 and 23). Moreover, several relatively sensitive but versatile functional groups—an olefin (9) and boronic esters (14 and 20)–also tolerated the reaction conditions well, indicating the potential utility of this method for pharmaceutical and synthetic chemistry. Notably, we did not detect a reaction at the allylic C − H bonds position of substrate of 9. Because of the prevalence of heteroaryl groups in pharmaceutical compounds, we were pleased to find that thiophene and pyridine substrates were amenable to the reaction, yielding 26 and 27 in 71% and 38% yields, respectively. In addition, 2-naphthoic acid underwent the desired transformation to afford 28 in 69% yield. Finally, late-stage modification of a derivative of menthol, delivered the corresponding product 29 in 41% yield. Potentially because of the instability of aliphatic azolium radicals, the scope was primarily limited to aryl substrates.

Fig. 2: Scope of the reaction with respect to the carboxylic acid.
figure 2

Reaction conditions: 1 (0.3 mmol), 2a (0.6 mmol), NHC A catalyst (0.06 mmol), PC-I (0.003 mmol), Cs2CO3 (0.06 mmol), K3PO4 (0.06 mmol), and HAT-1 (0.06 mmol) in DCM (4 mL) were irradiated with a 36 W blue LED under Ar at room temperature, isolated yields were given.

Substrate scope with respect to the alkene

The reaction conditions could also be applied to alkenes 2 with a diverse array of substitution patterns and electronic properties (Fig. 3). Both cyclic and acyclic alkenes were suitable substrates for reactions with 1a. For example, a series of simple cyclic olefins with various ring sizes afforded the corresponding β,γ-unsaturated ketones (3033).

Fig. 3: Reactions of allylic/benzylic C(sp3)–H substrates.
figure 3

Reaction conditions: 1a (0.3 mmol), 2 (0.6 mmol), NHC A catalyst (0.06 mmol), PC-I (0.003 mmol), Cs2CO3 (0.06 mmol), K3PO4 (0.06 mmol), and HAT-1 (0.06 mmol) in DCM (4 mL) were irradiated with a 36 W blue LED under Ar at room temperature, isolated yields were given.

Evaluation of various acyclic alkenes revealed that compounds with long or short alkyl chains, terminal olefins, or internal olefins with two or three substituents could be used as substrates and afforded the corresponding products (3440) in moderate to good yields. Notably, the isomeric products (37 and 37′) of an asymmetrically substituted olefin could be separated. The exclusive formation of the benzyl acylation product (40) though bearing a terminal doble bond can be readily rationalized by consideration of the ortho oxygen reducing the BDE value of this site and its stabilizing effect on free radicals. Moreover, a series of benzylic C(sp3)–H substrates were successfully employed in this coupling reaction; specifically, primary, secondary, and tertiary benzyl substrates gave products 4143, respectively, in 66–77% yields. Finally, the mild reaction conditions were suitable for the modification of natural products with various biological activities. For example, acylation of flavor molecules isopulegol, terpinyl acetate, limonene, 3-carene, and dicyclopentadiene afforded 4448, respectively. Unique site selectivity mainly benefits from steric hindrance or more stable secondary radicals. Naturally occurring γ-terpinene, which have multiple allylic C(sp3)–H sites, selectively reacted at the cyclic carbon to deliver products 49 in 63% yield. Butadiene dimer, which is used as an indigo dye, underwent direct allylic acylation at two sites, affording a mixture of isomers (50). In addition, modification of derivative of (S)-β-citronellol, a molecule with pesticidal activity, could be achieved by means of this method, which gave product 51 in 49% yield.

Transformations of β,γ-unsaturated ketone products

The β,γ-unsaturated ketones generated by means of our method could be scaled up in a slightly lower yield (71%) after 3 days of irradiation and easily be transformed to other useful molecules (Fig. 4). For example, treatment of 3 with hydroxylamine hydrochloride afforded β,γ-unsaturated ketoxime 52, which was a convenient raw material for the synthesis of isoxazoline 5395,96,97,98,99,100. The C=C bond of 3 could also be selectively reduced to afford 54. Moreover, terminal olefin 55, β,γ-unsaturated alcohol 56, and secondary amine 57 could be obtained via Wittig reaction, sodium borohydride reduction, and reductive amination, respectively.

Fig. 4: Transformations of a representative β,γ-unsaturated ketone.
figure 4

Except for the gram-scale synthesis, the other reactions were performed on a 0.2 mmol scale. PMPNH2: 4-methoxyaniline.

Development of one-pot protocol

In addition, to demonstrate the operational simplicity and utility of the method, we synthesized β,γ-unsaturated ketone 3 in one pot directly from p-toluic acid (Fig. 5). The acyl imidazole was generated in situ by reaction of the carboxylic acid with CDI, and then reaction with cyclohexene (2a) under the standard conditions produced 3 in 47% yield (compared with the 77% yield obtained when the acylimidazole was used as the substrate). Furthermore, a one-pot reaction of p-toluic acid, CDI, and cyclohexene under the standard conditions afforded 3 in 32% yield.

Fig. 5: One-pot synthesis of β,γ-unsaturated ketone 3 from p-toluic acid.
figure 5

Reactions were performed on a 0.2 mmol scale. CDI: N,N-carbonyldiimidazole.

Mechanistic studies

Next, we turned our attention to the mechanism of this multicatalytic process. First, to determine whether radicals were involved, we performed radical-trapping experiment using TEMPO (2,2,6,6-tetramethylpiperidine oxide) as a radical scavenger. We found that reaction of 1a and 2a was greatly suppressed by the scavenger, and the addition product of 2a and TEMPO was detected by mass spectrometry. In addition, radical capture product 58 was isolated from the reaction, supporting the formation of an acyl radical intermediate (Fig. 6a). Notably, reaction of 2a with acyl azolium ion 59 or 59’ under photoredox catalysis conditions provided ketone 3 in 41% and 25% yields; this result indicates that the acyl azolium species, generated in situ from the acylimidazole and NHC catalyst, might be the intermediate (Fig. 6b). Next, we performed Stern–Volmer quenching experiments using intermediate 59, 59’, iPr3SiSH (HAT−1) and iPr3SiS (int. I) as quenching reagents (Fig. 6c–e). The results of these experiments showed that reduction of the excited photocatalyst by the thiol in the presence of nBu4NOAc resulted in strong fluorescence quenching that was linearly correlated with thiolate concentration (Fig. 6e, red line). In contrast, thiol (HAT−1), 59 and 59’ did not obviously quench the excited catalyst (Fig. 6e, black, blue and green line). Finally, a light on/off experiment showed that the reaction stopped completely in the absence of light and then resumed when the light was turned back on, indicating that light was essential (data not shown). This result indicates that any radical chain processes were short-lived.

Fig. 6: Mechanistic experiments.
figure 6

a Reaction of 1a and 2a in the presence of TEMPO (2,2,6,6-tetramethylpiperidinooxy). b Reaction of 2a with intermediate 59 or 59’. c Stern–Volmer quenching experiments involving PC-I and thiol HAT−1. d Stern–Volmer quenching experiments involving PC-I and thiolate int. I. e Stern–Volmer analysis.

Based on the above mechanism experiments and literature reports, we propose that the mechanism of photoredox/NHC/HAT catalyzed formation of β,γ-unsaturated ketones from carboxylic acids and olefins is carried out through the mechanism in Fig. 7. Blue light irradiation converts the IrIII photocatalyst into a long-lived triplet excited state IrIII* complex, which is reduced to an IrII species by thiolate I generated in situ. The resulting electrophilic thiyl radical (II) can be used as a powerful HAT catalyst to abstract allylic hydrogen from cyclohexene to generate transient radical III. At the same time, the carboxylic acid is activated in situ by reaction with CDI, and then NHC catalyst is added to the activated acid (1a) to generate azolium intermediate IV, which can be reduced by IrII species to provide azolium radical V and regenerate the ground state IrIII photocatalyst. Subsequently the coupling of V with radical III ultimately yields the target β, γ-unsaturated ketone 3 and releases the NHC catalyst.

Fig. 7: Proposed mechanism.
figure 7

NHC, HAT and photocatalyst co-catalyzed mechanism.


In summary, we have developed a method for preparing β,γ-unsaturated ketones from carboxylic acids and alkenes by means of a process involving single-electron reduction of acyl azolium intermediates. The co-catalytic mode of triple catalysis involving photoredox, NHC, and HAT catalyst, enable the single-electron reduction of the acyl azolium intermediates; and subsequent radical–radical coupling readily forms a C–C bond under mild conditions to generate β,γ-unsaturated ketones. The method uses easy-to-prepare or commercially available starting materials and has a wide substrate scope and excellent functional group tolerance; our successful modification of various natural product and drug molecules demonstrates the potential utility of the method. The products could also undergo subsequent transformations. Continued research on such multicatalytic processes can be expected to facilitate the development of additional carbene-mediated reactions.


General procedure for the radical reaction

An 8 mL glass vial was charged with PC-I (3.4 mg, 0.003 mmol, 1 mol%), 1 (0.3 mmol, 1.0 equiv), 2 (0.6 mmol, 2.0 equiv), NHC A (13.5 mg, 0.06 mmol, 20 mol%), Cs2CO3 (21.2 mg, 0.06 mmol, 20 mol%), HAT-1 (11.4 mg, 0.06 mmol, 20 mol%), K3PO4 (12.7 mg, 0.06 mmol, 20 mol%), and 4 mL of anhydrous DCM. Argon was bubbled through the reaction mixture for 15 s via an outlet needle, and the vial was sealed with a PTFE cap. The mixture was then stirred rapidly while being irradiated with a 36 W blue LED (placed approximately 2 cm away from the vial) at room temperature for 24 h. The mixture was concentrated in vacuo, and the residue was subjected to flash chromatography on silica gel to afford the desired product in pure form.