Introduction

Pioneered by Drago and Mukaiyama in the late 1980s, Markovnikov-selective hydrofunctionalizations of alkenes via hydrogen atom transfer (HAT) have evolved into versatile tools in the synthesis of complex molecules1,2. Among the selected recent contributions by Carreira3,4,5,6,7, Boger8,9, Baran10,11,12,13,14, Shenvi15,16,17,18,19,20, Herzon21,22,23, and others24,25,26,27,28, the HAT hydrogenation of alkenes was introduced independently by Shenvi and Herzon and has since been established as a complementary method for commonly used noble-metal-catalyzed hydrogenation reactions. Shenvi has reported Mn- and Co-catalyzed protocols that use phenylsilane or isopropoxy(phenyl)silane reductants in the presence of tert-butyl hydroperoxide, which tolerates benzyl group and C(sp2)–halogen bonds29,30. In addition, thermodynamically controlled stereoselectivity is characteristic of these transformations and supports the proposed carbon-based radical intermediates (Fig. 1a). Herzon has reported Co-catalyzed hydrogenation protocols using triethylsilane, tert-butyl hydroperoxide, and 1,4-cyclohexadiene31,32,33. This method possesses high functional-group tolerance and was studied in detail for the hydrogenation of haloalkenes and the selective hydrogenation of substituted alkenes (Fig. 1b). While these systems have been increasingly appreciated in natural-product synthesis34,35, stoichiometric amounts of organosilanes and peroxides are required in order to achieve a wide substrate scope, which remains problematic in terms of sustainability and safety concerns. In this context, Norton’s cobaloxime-catalyzed protocol allowed hydrogenation of acrylamides using only molecular hydrogen as the sole stoichiometric reactant, and should thus be considered as an insightful example of the silane- and peroxide-free HAT hydrogenation of alkenes (Fig. 1c)36. Norton’s studies revealed fundamental physical parameters associated with HAT hydrogenations, albeit that the substrates hydrogenated in high yield were limited to alkenes activated with electron-withdrawing groups. During the review of this manuscript, Kattamuri and West reported iron and thiol-cocatalyzed oxidant-free HAT hydrogenation of alkenes37. Nonetheless, stoichiometric silane is required as a reductant in their catalytic system.

Fig. 1: Catalytic HAT hydrogenation of alkenes.
figure 1

a Shenvi’s method. b Herzon’s method. c Norton’s method. d This work; dpm 2,2,6,6-tetramethyl-3,5-heptanedionato, FG functional group, acac 2,4-pentanedionato, Cy cyclohexyl, DTBMP 2,6-di-tert-butyl-4-methylpyridine, dmg dimethylglyoximato, EWG electron-withdrawing group.

We were motivated to devise silane- and peroxide-free catalytic HAT hydrogenation reactions in order to expand the scope of HAT hydrogenations beyond the canonical synthesis of complex molecules. We hypothesized that this might be possible by using ascorbic acid38,39,40,41 and a combined cobalt/photoredox catalysis42,43,44,45,46,47,48,49,50,51,52 (Fig. 1d). Inspired by König’s proposed mechanism for the cobalt/photoredox-catalyzed alkene isomerization48, we designed our HAT hydrogenation by metal-photoredox cooperative catalysis53,54,55,56,57 as shown in Fig. 2a. The photoredox catalyst58,59,60,61,62 (PC) is excited by visible light and oxidizes ascorbate38, which exists in equilibrium with ascorbic acid, to afford the reduced PC and the corresponding ascorbic acid radical. The reduction of cobalt(II) with the reduced PC generates cobalt(I), which is converted to cobalt(III) hydride upon protonation. HAT from the cobalt(III) hydride to an unactivated alkene should regenerate the cobalt(II) catalyst and afford an alkyl radical. The second HAT to the resulting alkyl radical from ascorbic acid should realize the hydrogenation of the alkene. The resulting ascorbic acid radicals disproportionate to ascorbic acid and dehydroascorbic acid (Fig. 2b)63.

Fig. 2: Proposed mechanism for the HAT hydrogenation using ascorbic acid and a combined cobalt/photoredox catalytic system.
figure 2

a Hypothetical catalytic cycle of the silane- and peroxide-free HAT hydrogenation. b Disproportionation of ascorbic acid radicals; PC photoredox catalyst.

Results

Investigation of reaction conditions

Based on the proposed reaction design, we started to examine suitable cobalt-based co-catalysts for the photoredox-mediated HAT hydrogenation. Initially, we attempted the hydrogenation of 1a using Ru(bpy)3Cl2 as the PC and ascorbic acid as a hydrogen source. Hydrogenated 2a was not observed in the presence of cobalt chloride (Table 1, entry 1) or Co(acac)2·2H2O (acac: 2,4-pentanedionato), which is the optimal catalyst for the cobalt/photoredox-catalyzed isomerization of alkenes48 (entry 2). However, 2a was obtained when salcomine 3a (entry 3) or cobalt porphyrin 3b (entry 4) was used as a cocatalyst. A further evaluation of cobalt salen complexes revealed that a complex with cyclohexanediamine linker (3c) showed a very low catalytic performance (entry 5), while a complex with phenylenediamine linker (3d) generated 2a in higher yield (entry 6). The hydrogenation was further facilitated by increasing the amount of the cobalt catalyst and the photocatalyst (entry 7). Subsequently, we investigated the effects of the substituents, and discovered that complexes that contain tert-butyl groups (3e) show high catalytic performance (entry 8). Eventually, 3f, which was previously studied as a catalyst for electrocatalytic hydrogen evolution64, was identified as an optimal cobalt cocatalyst (entry 9). Using 3f, hydrogenated product 2a was obtained in 87% yield when the reaction was carried out with 0.2 M of 1a in the presence of catalytic amounts of tricyclohexylphosphine (entry 10). It is expected that the sterically demanding phosphine additive facilitated the reaction by preventing the accumulation of catalytically inactive cobalt-alkyl species31,65. Finally, 2a was obtained in 92% isolated yield on a larger reaction scale by increasing the amount of photon source (entry 11).

Table 1 Evaluation of the reaction conditions for the combined cobalt/photoredox-catalyzed hydrogenation of 1a.

The involvement of each reaction component was confirmed by control experiments (entries 12–15). In the absence of Co complex (entry 12), photocatalyst (entry 13), light (entry 14) or ascorbic acid (entry 15), no hydrogenation proceeded, supporting the proposed reaction design described in Fig. 2.

Substrate scope

The scope of the HAT hydrogenation of alkenes using the combined cobalt/photoredox catalysis is summarized in Table 2. Substrates with electron-rich arenes substituted with methyl (2b) and methoxy group (2c) were hydrogenated in high yield. Electron-deficient arenes containing trifluoromethyl and cyano groups were not affected during the hydrogenation and the hydrogenated products (2d, 2e) were obtained in 86% and 87% yield, respectively. Despite the putative intermediacy of a low-valent cobalt catalyst, C(sp2)–Cl, C(sp2)–Br and C(sp2)–I bonds of haloarenes were compatible with the applied reaction conditions and the chemoselective hydrogenation of alkenes proceeded in good yield (2f, 2g, 2h). These results might be due to the fast protonation of the low-valent cobalt by ascorbic acid compared to the abstraction of halogen atoms. A substrate with an aromatic ketone was converted into the corresponding hydrogenated product in 67% yield (2i). Benzyl groups attached to oxygen or nitrogen remained intact and the desired hydrogenated products were obtained in 78% (2j) and in 87% yield (2k), again demonstrating a distinct chemoselectivity between the conditions used in this study and conventional heterogeneous palladium-catalyzed hydrogenation conditions. An aliphatic ketone in the substrate remained intact during the hydrogenation to afford 2l in 88% yield. Substrates containing sulfone amide or benzyl carbonate moieties were also hydrogenated in good yield (2m, 2n). 2o was obtained in 87% yield while its C(sp3)–Cl bond remained untouched, again suggesting that HAT hydrogenation outcompeted halide abstraction. In addition to trisubstituted alkenes, 1,1-disubstituted alkene derived from γ,δ-unsaturated alcohol was efficiently hydrogenated in 86% yield (2p). The hydrogenation of α,β-unsaturated esters also proceeded in 64% yield under the standard conditions (2q). The hydrogenation of an exocyclic alkene (1r) afforded the trans diastereomer of 2r as the major product, which is consistent with the thermodynamically controlled stereoselectivity of the HAT hydrogenation. The aqueous solvent system required for this catalytic hydrogenation should be partly responsible for low yield of 2r. Nevertheless, hydrogenation of alkenes without polar functional groups proceeded in synthetically useful yield (2s, 2t, 2u). γ,δ-Unsaturated amide was hydrogenated in 84% yield (2v). High functional-group tolerance of the present HAT hydrogenation was further demonstrated by the hydrogenation of amino-acid derivatives. A glutamine derivative was hydrogenated in 84% yield, indicating that the hydrogenation was not affected in the presence of a primary amide moiety (2w). The free hydroxyl group of serine and the thioether group of methionine were also compatible with this hydrogenation and afforded the desired products in 89% (2x) and 85% yield (2y). The substrate derived from glycylglycine was hydrogenated in 79% yield (2z), indicating applicability of the present hydrogenation to dipeptides. Given that trisubstituted alkenes are frequently found in terpenes, we also studied the hydrogenation of these natural products. We found that (−)-isopregol, dihydrolinalool, and (−)-β-citronellol were hydrogenated to afford 2aa-2ac in 78–93% yield using the standard protocol. Due to the high compatibility of the present HAT hydrogenation with a variety of functional groups, we next examined the potential late-stage hydrogenation of bioactive small molecules. An immunosuppressant mycophenolic acid with free carboxyl and phenol groups was hydrogenated to furnish 2ad in 86% yield. It is noteworthy that the 1,2-disubstituted alkene of an analgesic agent capsaicin was hydrogenated to afford 2ae in 95% yield. The catalytic hydrogenation was readily applicable to C-glycosides derived from ribofuranose (2af) and galactose (2ag). Importantly, the present protocol is even suitable for the hydrogenation of an unprotected galactose derivative and afforded the gene inducer isobutyl-β-C-galactoside (IBCG, 2ah)66 in 89% yield.

Table 2 Scope of the ascorbic-acid-mediated hydrogenation of alkenes.

Comparison with previous HAT hydrogenation protocols

Despite the abundance of sugar derivatives in biologically active substances, HAT hydrogenations that were performed in the presence of unprotected sugar derivatives are, to the best of our knowledge, not known. Such protecting-group-free transformation67 has deemed a challenge partly due to the chelating nature of the sugar derivatives, which can poison metal-based HAT catalysts. We speculated that 3f, which does not possess vacant cis-coordination sites because of its planar tetradentate salophen ligand, might serve as a good metal-based catalyst for the HAT transformation of unprotected sugar derivatives. Indeed, the present system exhibits superior performance for the HAT hydrogenation of 1ah relative to previously reported catalytic methods (Table 3, entries 2 and 3). Both the manganese catalyst with bidentate dipivaloylmethane ligands29 and the cobalt catalyst with bidentate acetylacetone ligands31 afford 2ah in lower yield compared to the HAT hydrogenation catalyzed by 3f. For a fair comparison, catalytic hydrogenation of 1ah can be performed in 96% yield by heterogeneous palladium-catalyzed hydrogenation (entry 4). Nevertheless, the dual cobalt and photoredox-catalyzed HAT hydrogenation potentially serve as a chemoselective alternative for hydrogenation of alkenes in carbohydrate chemistry.

Table 3 Comparison of the HAT hydrogenation performance for the preparation of IBCG 2ah.

Electrochemical study of the cobalt catalyst

Electrochemical analysis of 3f was conducted in order to assess our mechanistic proposal. Cyclic voltammetry of 3f in DMF afforded quasi-reversible voltammogram (Fig. 3a) with E1/2 = −1.36 V vs. SCE, which reasonably corresponds to the CoII/CoI redox couple64. The E1/2 value suggests that the electron transfer to 3f from the reduced photocatalyst (Ered(RuII/RuI) = −1.33 V vs. SCE)68 is within an accessible range.

Fig. 3: Cyclic voltammograms of 3f in the presence of various amounts of ascorbic acid.
figure 3

Recorded at 100 mV/s in a DMF solution of 3f (0.50 mM) and Bu4NClO4 (0.1 M). a In the absence of ascorbic acid. b 0.63 mM of ascorbic acid. c 1.5 mM of ascorbic acid. d 3.0 mM of ascorbic acid. The potential was corrected using Fc/Fc+ as an internal standard; SCE saturated calomel electrode, Fc ferrocene.

An increase in the voltammetric current in the voltammogram of 3f was observed when increasing amounts of ascorbic acid were added (Fig. 3b–d), while cathodic peak potentials remained similar as the ratio of ascorbic acid/3f increased. These changes in the voltammograms were in accord with reported electrochemical behavior of 3f in the presence of increasing amounts of acetic acid64 and indicate the generation of cobalt hydride intermediate from 3f via cathodic reduction followed by protonation by ascorbic acid.

Thus, the electrochemical studies support the two processes described in Fig. 4a: (i) One electron reduction of 3f by an electron transfer from the electrode, and (ii) Successive protonation of the reduced cobalt catalyst by ascorbic acid. In the absence of an alkene, the resulting cobalt hydride should be consumed by proton reduction64. In turn, in the photocatalytic reaction process, it is likely that the cobalt hydride, generated via one-electron transfer from the reduced ruthenium photocatalyst followed by protonation by ascorbic acid, reacts with an alkene to afford the carbon radical intermediate (Fig. 4b).

Fig. 4: Evaluation of the proposed mechanism by electrochemical analysis.
figure 4

a Interpretation of the observations in cyclic voltammetry. b Plausible mechanisms of the cobalt-photoredox catalyzed hydrogenation in analogy with the electrochemical reaction; SCE saturated calomel electrode, AscH2 ascorbic acid.

Theoretical and experimental evaluation of the HAT pathway

In order to gain insight into the HAT process, quantum chemical calculations were performed using the cobalt(salophen) hydride complex derived from 3f and isobutene as a model substrate. The calculated pathway using UωB97X-D density functional [PCM (2-Propanol), LANL2DZ(f) for Co, 6-31G(d,p)] (see Section 5 of Supplementary Information for details) revealed that the barrier for the exergonic HAT between the two starting materials (SM) is only 7.3 kcal/mol higher in free energy relative to the SM at 25 °C (Fig. 5a), supporting that the HAT should proceed rapidly under the current reaction conditions. In the computed transition state, the cobalt center possesses square pyramidal geometry without cis-vacant sites (Fig. 5b). This molecular geometry closely resembles the reported transition states of the HAT between alkenes and cobalt(porphyrin) hydride complexes69,70. The observed similarity suggests that, like the [H-CoIII(porphyrin)] complex, the [H-CoIII(salophen)] complex favors HAT compared to concerted migratory insertion when reacting with alkenes.

Fig. 5: HAT pathway elucidated by quantum chemical calculations.
figure 5

a Computed energy diagram of the HAT between cobalt(salophen) hydride complex and isobutene. b Computed structure of the transition state of the HAT between cobalt(salophen) hydride complex and isobutene; ΔGrel relative free energy, SM starting materials, TS transition state, PD products. See the Supplementary Information for computational details.

As for an experimental support for HAT, the intermediacy of an alkyl radical in the hydrogenation reaction was confirmed by a radical-trapping experiment. When the hydrogenation of 1ac was attempted in the presence of TEMPO, no hydrogenated product 2ac was obtained. Instead, TEMPO adduct 4 was unambiguously observed by LCMS analysis of the reaction mixture (Fig. 6, see Sections 2–5 of Supplementary Information for details). This observation is in accord with the intermediacy of tertiary alkyl radical presumably generated by HAT from cobalt hydride to 1ac.

Fig. 6: Detection of the radical intermediate by TEMPO trapping.
figure 6

The formation of TEMPO adduct supports the presence of alkyl radical intermediate which is likely to be formed by HAT to 1ac; Cy cyclohexyl, bpy 2,2’-bipyridyl, TEMPO (2,2,6,6-Tetramethylpiperidin-1-yl)oxyl.

Discussion

In summary, we have developed a combined cobalt/PC system that enables the silane- and peroxide-free, ascorbic-acid-mediated HAT hydrogenation of alkenes in aqueous media. While the aqueous conditions were generally beneficial for the hydrogenation of polar substrates, substrates without polar functional groups were also hydrogenated in synthetically useful yield. The present reaction offers not only a more sustainable and safer alternative to previously reported HAT hydrogenation methods, but also a highly functional-group tolerant hydrogenation protocol suitable for the late-stage hydrogenation of amino-acid derivatives, terpenes, and drug molecules. The direct hydrogenation of an unprotected galactose derivative should be noted with regard to the potential future prospects for the HAT functionalization of unprotected sugar derivatives. The feasibility of the proposed HAT process was supported by electrochemical analysis, theoretical investigation and the radical-trapping experiment. Applications of this catalytic system to other Markovnikov-selective hydrofunctionalization reactions of alkenes are currently in progress in our group.

Methods

General procedure for ascorbic-acid-mediated HAT hydrogenation of alkenes

In an argon-filled glove box, a flame dried reaction vial was charged with an alkene 1 (0.20 mmol), ascorbic acid (106 mg, 0.60 mmol), 3f (12.0 mg, 20 µmol), tricyclohexylphosphine (11.2 mg, 40 µmol) and Ru(bpy)3Cl2•6H2O (3.0 mg, 4.0 µmol). The vial was capped and removed from the glove box. A mixed solvent (2-propanol/H2O = 3:1, 1 mL) was added to the vial via syringe, and the syringe hole was carefully sealed with a vinyl tape. The reaction vial was placed in front of the light source (ca. 3 cm from two blue LED panels) in a cold room (4 °C) so that the temperature of the reaction mixture was kept approximately at 25 °C. After stirring for the indicated time, the reaction mixture was cooled in an ice bath and sat. aq. NaHCO3 was added. Organic material was extracted with EtOAc (x 3) and the combined organic layer was washed with brine. The organic layer was concentrated under reduced pressure and purified by column chromatography (silica gel) to afford the hydrogenated product 2.