As illustrated in Table 1, benzamides(1a) and benzyl buta-2,3-dienoate (2a) were exploited as model substrates with 10 mmol % Co(acac)2, 0.2 equiv. of NaOPiv and 5.0 mol % photoredox catalysis under visible light irradiation (15W CFL) for 24 h at room temperature under an oxygen atmosphere. The model reaction was preferentially performed using 2,2,2-trifluoroethanol as a solvent, mainly because of which is facilitated to pair with hydrogen-bond acceptor groups and interfere with the catalytic cycle . To our delight, the desired product 3a could be obtained when [Ru(bpy)3]2+Cl2 was used as a photoredox catalyst (entry 1). It was found that [Ir(dF(CF3)ppy)2(dtbbpy)]PF6 did not effectively promote this cyclization (entry 2). Instead of [Ru(bpy)3]2+Cl2, Na2-eosin Y slightly increased the annulation yields under the similar reaction conditions (entry 3). Further experiment revealed that the use of KOTf as a base could accelerate the reaction and effectively improve the yield of 3a (entry 5). Interestingly, a slight increase in the amount of cobalt made significant effect in promoting the reaction, and the excellent yield was obtained when the annulation reaction was performed with 20 mmol% Co(acac)2 under the similar reaction conditions (entry 6). In addition, control experiments indicated that in the absence of either Co(acac)2 or oxygen, the reaction was completely inhibited (entry 9 and 10). It is worthy to note that the light played a crucial role for the annulation reaction, and it stopped at 14% yield when the reaction was performed without photocatalyst in dark (entry 11). Using unprotected benzamides with weaker directing groups instead of 1a, the desired product was not obtained under standard conditions, and the corresponding starting materials were decomposed or fully recovered.
Optimization the reaction conditions
With the optimized conditions in hand, the annulation reaction scope was investigated between various substituted benzamides and allenes, and the results were presented in Scheme 1. We found that a number of arylamides bearing different substitutions at the ortho-, metal-, and para-positions were compatible with the optimized conditions (Scheme 2, 3a-3p). For para-substituents in the carboxamides, such as halogen (3f-3 h), acetyl (3i), cyano (3j), or electron-donating methyl (3 k), cyclohexyl (3 l), methoxyl (3 m) groups, the desired isoquinolinones were produced in excellent yields. It was found that the reaction of meta-substituted benzamide furnished two expected regioisomers in good yields (3n). On the other hand, the ortho-substituted substrates with incorporation of electron-withdrawing substituents, such as halogen (3b-3c) and trifluoromethyl (3d), were efficiently transformed into the corresponding products in excellent yields. Replacing with methoxyl in the ortho-position (3e), the cyclization process took a short reaction time, but slightly reduced product yield was observed. In the case of disubstituted benzamide (3o), this reaction also gave the desired product in good yield. Additionally, the naphthoamide (3p) was smoothly converted into the three rings fused heterocycle in good yield.
Scheme 2 Scope of benzamides with benzyl buta-2,3-dienoate
Next, the scope of allenes was investigated extensively. As summarized in Scheme 3, various allenes were smoothly reacted with benzamides under the optimal conditions, and the corresponding isoquinolinones were got in good to excellent yields. Notably, it was found that the cyclization reaction featured excellent chemo- and regioselectivity, and solely occurred at the allenes’ terminal position. Various electronically diverged allenes (2b-2f) were efficiently converted into the desired products (4a-4p). Heteroaromatic substrate also showed good reactivity, and thieno[3,2-c]pyridin-4(5H)-one derivative (4 h) was efficiently obtained in good yield. Gratifyingly, it is worth mentioning that the excellent yields (4i-4n) were obtained when diphenyl(propa-1,2-dien-1-yl)phosphine (2e) oxide was used. A gram-scale reaction was conducted to assess the efficiency of this protocol, and two isoquinolinones (4 l and 4n) could be obtained in 79% and 87% yield, respectively (for details, please see supporting information). The structure and regioselectivity of nitro substituted isoquinolinone (4n) were confirmed unambiguously by the X-ray crystallography. In the case of ethyl penta-3,4-dienoate (2f), the desired isoquinolinones (4o and 4p) were afforded as single regioisomers with moderate yields.
Generally, the electronical properties of allenes have a great influence on the reactivity of annulation. Interestingly, various isoquinolinones with exo-double bonds were obtained by coupling benzamides with propa-1,2-dien-1-ylbenzene (2 g), which demonstrated the diverse reactivity of allene. A number of ary amides (5a-5d), and heterocyclic derivative, such as thiophene (5e) showed good compatibility with the reaction condition. To our delight, when terminal di-substituted allene (2 h) was used as a coupling partner, the reaction also afforded the corresponding exo-cyclic isoquinolinones (5 g-5i) in moderate to good yields. These results might provide an interesting perspective for the mechanism information on the migratory allene insertion/isomerization manifold.
To gain further mechanistic insights, a series of control experiments were carried out. A mixture of benzamide [D5]-1a/1a (1:1) was used to react with allenylphosphonate (2e) under optimized conditions, and the KIE experiments gave a kH/kD value of 1.1. This phenomenon indicated that C − H bond activation might not be involved in the rate-determining step (Scheme 5a). Furthermore, no D/H exchange was observed when [D5]-1a was treated with phosphateallene under standard conditions. Similarly, no deuterium incorporation in the product 4n was observed when 1c was treated with isotopically labelled CD3OD as a cosolvent. These results suggested that the C − H cobaltation step should be irreversible (Scheme 5a). A control reaction was performed in the presence of stoichiometric 2,2,6,6-tetramethyl-1-piperidinyloxy (TEMPO, a common radical scavenger), and no the desired product was obtained under the standard condition (Scheme 5b). This gave some clues that a highly radical species might be involved in the catalytic cycle. Additionally, the cyclization reaction conducted under an oxygen atmosphere in the absence of Eosin Y did not generate the desired product (Scheme 5c). It excluded the possibility of oxygen acting as a single oxidant which was used to regenerate the cobalt catalyst. On the other hand, intermolecular competition studies suggested that the electron-rich amide was slightly more favorable, which could be rationalized an electrophilic-type substitution C − H metalation.
Based on the mechanistic experiments described above and relevant literature reports [81–82], a plausible mechanism for this reaction was proposed in Scheme 6. The first step began with oxidation of Co(II) to give Co(III)-species assisted simultaneously by reduction of Na2Eosin Y*. A following ligand exchange and coordination of 8-aminoquinoline derived benzamide (1a) generated Co(III)-species (intermediate A). Interestingly, this intermediate was observed by the treating stoichiometric amounts of Co(acac)2 under the standard conditions (for details, please see supporting information) [59, 77–78, 83–85]. After that, intermediate A activated the inert sp2 C − H bond to form the key metallacycle Co(III) intermediate B. Subsequently, allene coordinated with metallacycle intermediate B to give the corresponding C, and regioselective insertion of C − Co bond of intermediate C gave the seven membered cobaltacycle intermediate D or E. The selectivity of the annulation depended on the allene structure and electronic characteristics of the substituents. When R were electronic-withdrawing groups, the insertion took place in the less-substituted double bond of allene leads to the intermediate D. When R was the phenyl group, the addition was apt to occur at a more electron-rich double bond, furnishing the formation of the intermediate E with high selectivity. The reductive elimination gave the cyclic-product (I or G) and intermediate F. The newly obtained cobalt intermediate F (the in situ generated cobalt catalyst between Co(OAc)2 and benzamide substrate showed a oxidation potential at 1.19 V vs SCE)  might be oxidized by the photoexcited Na2Eosin Y* (0.83 V vs SCE) [77, 86]. Thus the Co(III) species could be regenerated, and a strong reductant, Na2Eosin Y radical anion (-1.06 V vs SCE)  was simultaneously generated. The Na2Eosin Y radical anion might be oxidized to the ground state by O2 to complete the photoredox cycle . At last, the intermediate G undergoes 1,3-hydrogen shift to furnish the corresponding final product H.