Cobalt-electrocatalytic HAT for functionalization of unsaturated C–C bonds

The study and application of transition metal hydrides (TMHs) has been an active area of chemical research since the early 1960s1, for energy storage, through the reduction of protons to generate hydrogen2,3, and for organic synthesis, for the functionalization of unsaturated C–C, C–O and C–N bonds4,5. In the former instance, electrochemical means for driving such reactivity has been common place since the 1950s6 but the use of stoichiometric exogenous organic- and metal-based reductants to harness the power of TMHs in synthetic chemistry remains the norm. In particular, cobalt-based TMHs have found widespread use for the derivatization of olefins and alkynes in complex molecule construction, often by a net hydrogen atom transfer (HAT)7. Here we show how an electrocatalytic approach inspired by decades of energy storage research can be made use of in the context of modern organic synthesis. This strategy not only offers benefits in terms of sustainability and efficiency but also enables enhanced chemoselectivity and distinct, tunable reactivity. Ten different reaction manifolds across dozens of substrates are exemplified, along with detailed mechanistic insights into this scalable electrochemical entry into Co–H generation that takes place through a low-valent intermediate. A perspective is given on how an electrocatalytic approach, inspired by decades of energy storage studies, can be used in the context of efficient cobalt-hydride generation with a variety of applications in modern organic synthesis.

The study and application of transition metal hydrides (TMHs) has been an active area of chemical research since the early 1960s 1 , for energy storage, through the reduction of protons to generate hydrogen 2,3 , and for organic synthesis, for the functionalization of unsaturated C-C, C-O and C-N bonds 4,5 . In the former instance, electrochemical means for driving such reactivity has been common place since the 1950s 6 but the use of stoichiometric exogenous organic-and metal-based reductants to harness the power of TMHs in synthetic chemistry remains the norm. In particular, cobalt-based TMHs have found widespread use for the derivatization of olefins and alkynes in complex molecule construction, often by a net hydrogen atom transfer (HAT) 7 . Here we show how an electrocatalytic approach inspired by decades of energy storage research can be made use of in the context of modern organic synthesis. This strategy not only offers benefits in terms of sustainability and efficiency but also enables enhanced chemoselectivity and distinct, tunable reactivity. Ten different reaction manifolds across dozens of substrates are exemplified, along with detailed mechanistic insights into this scalable electrochemical entry into Co-H generation that takes place through a low-valent intermediate.
Transition metal hydrides (TMH) species have been a vibrant topic for exploration in organic and organometallic synthesis 1,8 . Pioneering studies in this field have led to a deep understanding of metal hydrides 9 that has allowed synthetic chemists to establish these species as selective mediators for hydrogen atom transfer (HAT) chemistry 7,10 . Such insights have led to the discovery of unique selectivity for known transformations 11 along with the development of new chemical reactivity 12 . HAT, the concerted migration of a proton and an electron from a TM-H bond to an acceptor molecule, has emerged as one of the most useful chemical processes for the hydrofunctionalization of alkenes 4 . In its common manifestation, the generation of a TMH involves the exposure of an appropriate metal complex to a stoichiometric amount of reductant, such as a silane. Its subsequent reaction with an olefin leads to the formation of a C-H bond at the less electronically stabilized position along with a carbon-centred radical on the adjacent position. This intermediate can then be trapped with various reagents to form new C-C, C-N, C-O and C-X bonds [13][14][15][16][17][18] . Although the overall process formally requires only the addition of a proton and an electron to form the active TMH catalytic species, exogenous chemical oxidants are often required to elicit this reactivity. The application of HAT chemistry on a large scale could be problematic because of the need for an excess amount of external reductants with or without oxidants, resulting in poor atom economy and scalability concerns implicit in the use of organic reductants and oxidants in the same flask ( Fig. 1a) 19 . Given the growing documented utility of such reactions in organic synthesis, it is clear that more practical and universal variants are required 20,21 .
In parallel, the same types of species have been efficiently and sustainably generated in the hydrogen production field with a proton as the hydride source (Fig. 1b) 2 . For example, hydrogen evolution by electrochemically generated Co-H species was known as early as 1985 (ref. 22 ). The field was dormant for over two decades until recently, with interest in cobalt-catalysed electrochemical hydrogen evolution for green energy storage being the subject of a large body of studies spanning hundreds of publications 23 . These robust Co-H based processes feature high turnover numbers and have been optimized to high levels with more than 90% efficiency of H 2 production from simple protic systems, indicating that their commercial implementation is imminent. From a mechanistic standpoint, Co-H is formed in situ by the protonation of low valent Co(I)/Co(0) intermediates after direct cathodic reduction 24,25 . Subsequently, it can react by two different pathways to form hydrogen and regenerate the catalyst. In the first suggested mechanism, the generated Co(II)-H species decomposes by proton attack and evolves hydrogen by an intermediate dihydrogen metal complex. Alternatively, Co(III)-H can be reduced to Co(II)-H, which is Article followed by a similar protonation step. Interestingly, the described process operates with high faradaic efficiency in aqueous or non-aqueous mediums and various types of proton sources, such as water, acids and alcohols. Amongst the many cobalt complexes enlisted, many do not require complex ligand architectures 26,27 .
Inspired by the well-established cobalt-electrocatalytic hydrogen evolution chemistry precedent outlined above, presented here (Fig. 1b) is a set of chemoselective, tunable electrochemical HAT (e-HAT) protocols free of either chemical reductants and oxidants (for example, silanes and peroxides) or rigorous experimental protocols (for example, moisture tolerant and glove-box free). Thus, a versatile range of tunable reactivities with alkenes and alkynes-such as isomerization, selective reduction and hydrofunctionalization (Fig. 1c)-can be realized with unmatched efficiency and chemoselectivity beyond that observed under purely chemical conditions. In support of these claims, this electrochemically enabled reactivity is benchmarked with several of the most popular and recently disclosed chemical methods. In addition, in-depth mechanistic analysis using cyclic voltammetry, ultraviolet (UV)-visible (vis) spectroelectrochemistry, computation and kinetics provides insight into e-HAT and rationalizes the observed selectivity. Finally, the scalability of this process is demonstrated in both batch and recycle flow (on a gram-centigram (0.05-0.8 mol) scale).
As a proof of concept, alkene isomerization was selected as a model transformation, as only a substoichiometric amount of Co-H is required for the reaction to proceed with complete conversion 28 . Alkene 1 was selected for the initial optimization of the e-HAT reaction. Trial runs using the literature precedents for classical HAT isomerization with alkene 1 provided poor conversion to the desired product. For example, the use of 50 mol% of silane and 10 mol% of cobalt catalyst in benzene gives a 29% yield of the desired product and a 7% yield of other chain-walking isomers (entry 1, Shenvi's protocol) 29 . The method of Norton, which relies on a high pressure of the hydrogen gas, delivered only traces of product (entry 2) 30 . First forays into e-HAT isomerization followed the guiding principles 31 from previous findings in electrochemistry 32,33 and HAT chemistry 34 to aid in the selection of proper ligands, cathodic materials and proton sources. An abbreviated summary of more than 200 experiments is depicted in Fig. 1d (see the Supplementary Information for an extensive list). First, the cobalt catalyst screening (entries 3-6) revealed that CoBr 2 (glyme) was optimal, resulting in efficient Co-H generation with the highest yield for the alkene isomerization. However, the mass balance of these reactions consisted of a mixture of chain-walking and reduction by-products. Thus, ligand screening was performed to cleanly obtain the desired single isomerization product (entries 7-10). Although 4,4′-dimethoxy-bipyridine cleanly afforded the desired selective 1-position isomerization, the conversion was low (38% yield + 33% recovered 1, entry 10). To improve the conversion while retaining the ligand-controlled selectivity, various proton sources within a wide range of pK a were explored (entries 10-13); the use of inexpensive triethylamine hydrotetrafluoroborate (3 equiv., entry 13) as a proton source emerged as optimum, providing the desired product 2 in 72% isolated yield. This unique proton source was used owing to its ability to function as a supporting electrolyte as well; its inclusion was crucial to the reproducibility and robustness of the reaction. The final set of e-HAT conditions tolerates moisture, leads to completion with catalytic amounts of electricity (0.5 F mol -1 , entry 14) and can be set up in minutes using a simple undivided cell and a commercial potentiostat. Interestingly, similar reactivity was not observed when conventional reductants, such as zinc and manganese, were used (entries [15][16]. Of all the cathodes evaluated, tin, Ni-foam, glassy-carbon and stainless steel could be used, but a tin cathode gave the highest yield across a broad range of substrates (Supplementary Information).
With these results in hand, the scope of the e-HAT isomerization of monosubstituted olefins was investigated (Fig. 2a). A wide range of functionalities was tolerated, including free and protected amines (3, 6), anilines (5), amides (4), lactams (7), alcohols (13) and aliphatic nitriles (14) with over 80% yield on average. In addition, the e-HAT  isomerization exhibited a broad scope across a range of different arenes. Under optimized conditions, alkenes were isomerized in the presence of pyridines (8), thiophenes (9), electron-deficient indoles (10), redox-active aryl bromides (11) and aryl-pinacolato boronyl (2). Notably, this method can deliver the isomerization of an allylic ether to the corresponding enol ether adduct (16). Terminal disubstituted olefins were, however, untouched by the bipyridine complex (conditions A). As a result, another round of optimization was conducted revealing that commercially available Co II (t-Bu,t-Bu-cyclohexylsalen) (Co(salen)-1, see the Supplementary Information for the exact structure) could be used to exclusively isomerize such olefins to the thermodynamic trisubstituted alkenes by using hexafluoroisopropanol (HFIP) as the proton source and a Ni-foam cathode in acetone (conditions B, see Supplementary Information for optimization). As e-HAT relies on the in situ formation of a cobalt hydride, chemoselective reactions are thus possible simply by tuning the conditions. A similarly broad scope was observed for this isomerization as well (17)(18)(19)(20)(21)(22)(23)(24)(25). The ability to achieve olefin isomerizations in the presence of free phenols, pyridines, anilines, nitriles and epoxides has, to the best of our knowledge, not been seen before. Selected examples of the scope have been directly compared to existing conventional Co-H isomerization methods to show the advantages of e-HAT chemoselecticity (see the Supplementary Information for the comparisons, specifically the comparison section). In addition, the generation of β,γ-unsaturated amides has not, to our knowledge, been previously reported by isomerization methods, presumably owing to a tendency to isomerize into conjugation. Given the radical nature of intermediates in HAT-based reactions, conditions B not surprisingly initiated intramolecular radical cyclizations of dienes to form new C-C bonds by cycloisomerization. Accordingly, methallyl prenyl malonate can undergo intramolecular cycloisomerization to yield the corresponding trisubstituted cyclopentane (26) in a high yield with no isomerization side products. The malonate can be exchanged by an ether (27) or an amine (28) without compromising the high efficiency of the transformation. Endocyclic alkenes can be similarly used as effective cyclization partners to form the cis five-six bicyclic systems with high diastereoselectivity (29). Even a cyclic enol ether can be used as the radical acceptor (31). As shown in Fig. 2, the reaction shows high efficiency for five-membered  Article ring formation but is less suitable for the formation of six-membered rings (30), which gave only a 50% yield along with a linear isomerization side product. The previously discussed isomerization reactions are net-redox-neutral transformations. Therefore, a substoichiometric amount of cobalt hydride is needed to proceed efficiently as the active catalyst is regenerated during the reaction pathway (see below). Alternatively, cobalt-hydride chemistry can be used to reduce unsaturated systems. Such reactions, by definition, will require 'stoichiometric' electrons to be added with the right tuning of the proton source and cobalt complex to achieve the desired transformation. As a proof of concept, using e-HAT logic, a new e-HAT set of conditions for Z-selective (cis) alkyne semi-reduction using HFIP as the hydride surrogate was developed. To place this into context, the most frequently used reagents (that is, outside a glove-box) to achieve such a reaction on unactivated (non-conjugated) alkynes involve the use of Pd (Lindlar) catalysis and diimide. For such selectivity, the 6,6′-dimethyl-bipyridine ligand combined with CoBr 2 revealed the best reactivity, delivering high Z-selectivity and minimal over-reduction (conditions C, see the Supplementary Information for optimization). With this new set of conditions in hand, a range of substituted alkynes could be reduced with e-HAT to provide Z-alkenes in good yield (Fig. 3) rapidly. tert-Butyloxycarbonyl-protected amines (32), pyridines (33), ethers (34), free and silyl-protected alcohols (36,41), aryl chlorides (37), aryl-pinacolato boronyls (38), carbonyls (39,42), alkenes (39) and alkyl phosphates (40) were all tolerated. The highest Z/E selectivity (cis/trans) was observed with primary, secondary and tertiary carbons adjacent to the alkyne moiety. However, the process was less selective with quaternary carbons adjacent to the alkyne (see 43 and 44), presumably owing to the increased steric hindrance surrounding the putative cobalt-alkene intermediate.
The selective reduction of monosubstituted alkenes was similarly achieved by relying on e-HAT (conditions C). Canonical hydrogenation with H 2 over Pd/C can accomplish this type of reduction; however, the chemoselectivity of that method can be poor when competing reductively labile functionalities are present 35 . Accordingly, e-HAT was tested with substrates that can be challenging with such well-established reduction protocols. Thus, olefins containing amine-diol functionalities (45), tri-substituted alkenes (46,47), benzyloxycarbonyl-protected amines (48), benzyl-protected carboxylic acids (48) and thioanisole moieties (49), which can all be problematic under typical hydrogenation conditions, were smoothly reduced.
Although classical HAT chemistry has been studied for over 30 years, precise control of chemoselectivity when multiple olefins are present in a substrate has remained underdeveloped, owing to the complexity of tuning the hydride donor, oxidant and catalyst 36,37 . The modularity of e-HAT can potentially address such a challenge to achieve unique and useful selectivity with polyunsaturated systems, and was thus pursued using the three distinct e-HAT conditions (conditions A-C) presented above. Compound 50, which contains two different mono-and disubstituted terminal alkenes, was chosen as a case study (Fig. 4a(i)). By applying conditions B (Co(salen)-1/HFIP), the exocyclic alkene was selectively isomerized to form the thermodynamically favoured trisubstituted alkene product 51 in a 59% yield. By contrast, exposing compound 50 to conditions A (CoBr 2 /4,4′-MeO-bipyridine/Et 3 NHBF 4 ) led to the formation of disubstituted alkene 52 in a 92% yield with an E:Z isomeric ratio of 3:1. This result of e-HAT could be contextualized through a direct comparison with known isomerization methods . Various canonical Co, Fe, Pd and Ru isomerization methods resulted in an inseparable mixture of double isomerized and reduced products (see the Supplementary Information for a detailed product distribution, specifically the comparison section). Similarly, evaluation of the e-HAT alkene reduction displayed exquisite selectivity towards terminal monosubstituted alkenes. Triene 53 was subjected to conditions C that selectively reduced the desired alkene-over the 1,1-disubstituted and endocyclic olefins-in a 95% isolated yield (Fig. 4a(ii)). Likewise, diene 55 was also examined under the same conditions and exhibited similar chemoselectivity towards the monosubstituted olefin in the presence of an internal disubstituted alkene to furnish 56 in an 81% yield. Using H 2 with Pd/C, diimide, photoinduced electron transfer (Co-salen with Ru catalyst and light) 38 , Co/H + -based 39 and cobalt-hydride-based 40 systems as a direct comparison resulted in an inseparable mixture of reduced products (see the Supplementary Information for a detailed product distribution, specifically the comparison section), highlighting the singular efficacy of the e-HAT method.
A different and useful case of e-HAT chemoselectivity is exemplified with the isomerization of 1,1-disubstituted olefins to produce the less hindered isomerization products (Fig. 4a(iii)). In the case of compound 57, the double bond migration process has essentially two unpredictable directions to form different trisubstituted alkenes. By applying the conditions B (Co(salen)-1/HFIP) conditions with compound 57, the regioselectivity control exclusively furnished product 58 (in an 89% yield). Such regioselectivity could be explained through steric discrimination as guided by the catalyst. To strengthen this hypothesis, compound 59 was also tested under the same conditions; although there is a driving force to form the tetrasubstituted thermodynamic alkene, the kinetically favoured trisubstituted alkene was identified as the major product with a 7:1 isomeric ratio.
Next, the scalability of e-HAT was evaluated in both batch and recycle flow on three different transformations: isomerization, alkyne reduction and cycloisomerization (Fig. 4b). The isomerization reaction with conditions A was tested with compound 61 on a 1 g scale using the commercial Vapourtec ion recycle flow system (Fig. 4b(i)). By keeping the same parameters with 0.8 mA cm -2 current density and 1 ml min -1 flow rate, the product was obtained in a 92% yield. In a similar fashion, the alkyne reduction was demonstrated on a 10 g scale using a batch setup to achieve a complete conversion of 63 and to obtain the desired product (64) with a 76% isolated yield (with 11:1 E:Z isomeric ratio, Fig. 4b(ii)). Finally, the cycloisomerization reaction with conditions B was conducted on a 100 g scale with compound 65 using a recycle flow apparatus containing four reaction cells (Fig. 4b(iii)). After optimization, 100 g of 65 were successfully converted to product 27 by keeping the same current value of 2 mA cm -2 (compared with the 0.2 mmol scale) with only of 0.6 F mol -1 required to achieve complete conversion (73% isolated yield).
The direct hydrofunctionalization of saturated systems represents an area in which HAT-based transformations have found extensive application in synthesis 7 . Accordingly, e-HAT could be implemented under a universal set of conditions to achieve a wide range of classical HAT reactions (Fig. 4c). By simply using a slight modification of conditions B, intramolecular HAT-Giese (67), hydroarylation (71), hydropyridination (73), retroisomerization of alkene-conjugated strained-rings (75) and deallylation as a general deprotection method (77) were achieved. For the intermolecular HAT-Giese with ethyl acrylate to form a quaternary carbon (50%), a modification of conditions B was used in which 2 mol% of Co(tetramethyl-ethanediyl salen) (Co(salen)-2) enabled the reaction to proceed, whereas no product was observed with Co(salen)-1. This phenomenon might be explained by the efficient formation of the carbon-centred radical instead of the cobalt-carbon bond with Co(salen)-2 because of geometrical differences in the cobalt catalysts (see the Supplementary Information for a mechanistic rationale, specifically the computation section) 41 .
As a testament to the unique tunability of e-HAT, a challenging trans-selective alkyne semi-reduction was achieved using conditions B by slightly modifying the amount of HFIP to 8 equiv. and passing through 5 F mol -1 of electricity. Alkyne 78 was therefore reduced to the corresponding trans-alkene in 60% isolated yield and high geometric selectivity (with 7:1 E:Z ratio). Interestingly, by performing the analogous controls with HAT chemistry using silane and the same Co(salen)-2 complex, we found that such reactivity failed to translate, once again highlighting a unique feature of the e-HAT platform (Fig. 4d). The direct, E-selective reduction of alkynes can currently only be accomplished using either forcing Birch-like conditions (Li/NH 3 ) 42 or by enlisting an expensive Ru-based catalysis system 43 .
To further understand the e-HAT processes and support the critical steps of the proposed mechanism, cyclic voltammetry (CV), UV−vis spectroelectrochemistry, density functional theory (DFT) computations, and kinetic analysis were used. First, we investigated the CV profiles of the Co(salen)-1 isomerization system. As shown in Fig. 5b(i), the CV profile of Co(Salen)-1 at 50 mV s -1 showed a reversible redox peak at -1.98 V (versus Fc/Fc + ) corresponding to the Co(II/I) redox couple (step A). The addition of HFIP to the Co(Salen)-1 solution resulted in the observation of heightened current response. This is expected to arise from an ECEC cat (catalytic electron transfer, chemical reaction, electron transfer, chemical reaction) process corresponding to the generation of H 2 (see the Supplementary Information for a detailed discussion and Fig. 44) 24,25 . Such an observation is indicative of in situ protonation of the low-valent Co(I) to generate the Co(III)-H (step B). The addition of both HFIP and alkene to the Co(salen)-1 solution (conditions B) results in a slight dampening of the current response. This could indicate that the alkene is reacting with the Co-H species to catalyse the isomerization reaction, diverting the catalyst from the H 2 generation cycle (step C). To further support the proposed mechanism of the Co(salen)-1 system, we sought to observe the respective catalyst resting states with in situ UV−vis spectroelectrochemistry (Fig. 5b(ii)). Accordingly, we examined the spectroelectrochemical properties of the Co(salen)-1 under the olefin isomerization reaction mixture. Application of a stepwise reducing potential to a Co(II)-salen solution found modest decreases in absorbance for the electronic transitions at 416 nm and 492 nm, which could result from either the direct reduction of the metal centre to a Co(I)-salen (step A) or redox non-innocence of the salen backbone (Fig. 5b(ii), and see the Supplementary Information). Reduction of Co(salen)-1 at −2.03 V in the presence of HFIP led to the complete disappearance of the Co(II)-salen electronic spectral features and the appearance of a single UV-vis absorbance at 375 nm. On the basis of our previous observations, on application of reducing potentials, we speculate that this new spectral feature at 375 nm could arise from the presence of a cobalt-hydride catalyst resting state (Step B), a commonly invoked intermediate within the field of cobalt-promoted HER catalysis 24,44 . Intriguingly, spectroelectrochemical studies of the Co(salen)-1 in the presence of HFIP and alkene lead to the same absorbance feature at 375 nm during active electrocatalysis. Presumably this putative cobalt hydride could be competent for both cobalt-promoted HER as well as cobalt-catalysed olefin isomerization, thus appearing in both spectroelectrochemical experiments as a plausible catalyst resting state.
To investigate the CoBr 2 /4,4′-dimethoxybipyridine isomerization system (conditions A), the ligation state of the active cobalt catalyst was first identified (Fig. 5b(iii)). A 1:1 mixture of CoBr 2 and 4,4′-dimethoxybipyridine resulted in a CV profile with two distinct reduction peaks for Co(II) to Co(I) (Fig. 5a, step A). These peaks were preliminarily assigned as the bisligated L 2 CoBr 2 (which overlaps with unligated CoBr 2 ) for the less negative reduction peak and the monoligated L 1 CoBr 2 for the more negative reduction peak. This assignment is consistent with previous studies showing that the reduction potential of bisligated cobalt complexes is, in general, less negative than its corresponding monoligated species 45 . Next, CV studies were performed by adding a concentration of 4,4′-dimethoxybipyridine. At 1:1.5 Co/ligand ratio, both reduction peaks increased. The addition of more ligand caused the peak assigned as the monoligated complex to decrease. Owing to the overlapping peak on the reduction wave associated with Article the bisligated system, the oxidation wave was examined (as the reduction of unligated CoBr 2 is irreversible, see Supplementary Fig. 50), for which an increased oxidation peak was observed on increasing the ligand concentration beyond 1.5 equiv. These studies provide evidence for the preliminary assignment outlined above. Next, the effect of the proton source and the alkene substrate on the CV behaviour of the Cobalt catalyst was studied. Using a 1:2 mM ratio of CoBr 2 /ligand at which both the mono-and bisligated species were present, the effect of adding 1 mM Et 3 NHBF 4 was investigated (Fig. 5b(iv)). A significant increase in peak current was observed for the monoligated cobalt complex and a negligible change was observed for the bisligated species. These results indicate that the monoligated cobalt complex is active towards reaction with Et 3 NHBF 4 to generate a cobalt-hydride intermediate (Fig. 5a, step B). The addition of monosubstituted alkene to this solution resulted in a heightened current response indicative of a chemical reaction with alkene that regenerates the active cobalt catalyst. We therefore propose this as a series of chemical steps for the isomerization of alkenes (Fig. 5a, steps C-E).
DFT computational analysis provided additional evidence for the proposed mechanism of olefin isomerization. On the basis of the divergent reactivity of the two complexes throughout the synthetic studies, it stands to reason that the salen-derived catalysts operate with a different general mechanism than that of the bipyridine system. To explore this, the bond dissociation energies of various possible intermediates in such isomerization processes were calculated for both systems. According to the previous analysis, the reaction is initiated by Co-H generation, followed by HAT onto the olefin to form an alkyl radical that may undergo a radical pair collapse to form an alkyl-cobalt species. The alkyl-cobalt intermediates of salen ligands contain weak Co-C bonds (27 kcal mol -1 for secondary carbon, 20 kcal mol -1 for tertiary carbon, Fig. 5b(v)) that resemble diradicals 46 . Accordingly, we propose that steps C and E (Fig. 5a) proceed by HAT with the Co(salen)-1 system. Similarly, Fig. 5b(vi) shows that the strength of the Co-H bond in the Co(salen)-hydride complex indeed matches the C-H bonds of the substrate, which further supports the thermodynamic feasibility of the HAT pathway 38 . In the case of the cobalt-bipyridine system (conditions A), the strong Co-C bonds of the putative intermediates ( Fig. 5b(vi)) are perhaps more consistent with an organometallic process of inner-sphere migratory insertion and β-hydride elimination. Consistent with this conclusion, the bipyridine cobalt-hydride species potentially involved in an HAT version of step C contain Co-H bonds much stronger than the C-H bonds of the substrate (Co I and Co II , Fig. 5b(v)), whereas those possibly involved in step E contain Co-H bonds much weaker than the C-H bonds of the substrate (Co III , Fig. 5b(v)). Both assertions point to a thermodynamically unfavourable HAT pathway in the presence of bipyridine ligands.
Kinetic studies carried out on the isomerization of 5-phenyl-1-pentene (15a) using the cobalt-bipyridine system (conditions A) and the cycloisomerization of 30 using the Co(salen)-1 system (conditions B) revealed intriguing differences between the two cases ( Fig. 5b(vii)). For conditions B, the reaction exhibits first-order kinetics in the [substrate] (that is, concentration of the substrate), whereas, for conditions A, the reaction is zero order in the [substrate]. Neither conditions A nor conditions B are influenced by the concentration of the proton source. These observations indicate that the rate-determining step for conditions A is steps D and/or E and for conditions B is step C. The reaction is first order in [Co] under conditions B, whereas, under conditions A, the reaction exhibits an unusual zero-order dependence on [Co] at low current (2.5 mA). No evidence of the deposition of Co on the electrode was found, precluding the zero-order dependence being due to an active surface bound Co species. Another possible explanation could be that at low current the electrons released are insufficient to fully engage all the solution Co species, in which case increasing the catalyst concentration would not influence the rate. This hypothesis was confirmed by carrying out reactions at higher currents, at which the reaction under conditions A becomes first order in [Co].
The reaction rate under conditions A is not influenced by the current from 5.0 to 7.5 mA, whereas, under conditions B, the rate increases proportionally with the increasing current from 2.5 to 5.0 mA and is not influenced by the current from 5.0 to 10 mA. Both catalyst systems exhibit an induction period, the length of which decreases with increasing current (middle plot in Fig. 5b(vii)). For conditions B, the induction period could be removed by pre-activating the catalyst with electrolysis approximately proportional to the catalyst concentration. Neither system requires 1 F mol -1 to achieve full conversion of the substrate, and thus both are substoichiometric in electrons (left and middle plots, Fig. 5b(vii)).
These kinetic results indicate subtle differences between the two catalyst systems in the isomerization mechanism presented in Fig. 5a. The mechanism consists of two coupled cycles; the product turnover is shown on the left, and the electrochemical catalyst activation cycle is shown in purple. The isomerization product cycle can theoretically be sustained in the absence of electrochemistry, consistent with the fact that the reaction with either catalyst requires less than 1 F mol -1 substrate. Turnover in the cycloisomerization reaction using Co(salen)-1 continues, albeit more slowly, even after the current is stopped (right plot in Fig. 5b(vii)). However, reactions using the cobalt-bipyridine system do not proceed further when the current is stopped. Differential electrochemical mass spectrometry (DEMS) enables operando measurement with gaseous or volatile products and was used to further investigate such phenomena in the cobalt-bipyridine system 47 . The transient response behaviours of H 2 mass spectrometric signals were studied after removing the applied current. With only a proton source, the hydrogen formation was terminated and hydrogen signals displayed an obvious exponential decay with similar relaxation time (8-10 s), independent of the applied currents ( Supplementary Fig. 64). Therefore, the larger relaxation time with addition of the Co catalyst and alkene (conditions A) compared with only a proton source indicated extra hydrogen release from the Co-H intermediate after the current was stopped (Fig. 5b(viii)). This DEMS observation served as compelling evidence for the existence of a Co-hydride intermediate and its conversion back to Co(II) after termination of the electrolysis. In the absence of current, LCo(III)-H cannot go through steps A and B, and the cycle becomes stalled after of Co(III)-H conversion to Co(II) by step I.
The Z-selective semi-reduction of alkynes using CoBr 2 /6,6′dimethylbipyridine (conditions C) was also investigated. Using a mixture of a 1:1.5 ratio of the CoBr 2 /ligand (predominantly monoligated, and similar to the ratio in the reaction conditions), the effect of HFIP and alkyne addition (Fig. 5b(ix)) was studied. The addition of 1 equiv. of HFIP showed a shift in the cathodic peak, denoting a chemical step that generates a new electroactive cobalt intermediate, proposed as Co(III)-H (Fig. 5a, step B). The addition of alkyne to this mixture resulted in an increase in cathodic peak current with reduction potential similar to the active catalyst. This denotes a chemical reaction of the cobalt-hydride intermediate with the alkyne, accompanied by regeneration of the monoligated Co(II) catalyst (Fig. 5a, steps F-H). Computational analysis corroborated this proposed mechanism of proton migration from HFIP to the coordinated alkyne (Fig. 5b(x), step G) as it demonstrated Fig. 4 | Selectivity, scalability and HAT reactions of e-HAT. a, Selectivity of alkene isomerization and reduction. a Pdt (product) mixture, inseparable mixture of isomers. b Pdt mixture, inseparable mixture of starting material, desired product and over-reduced product. See Supplementary Information for further information. r.r., regioisomeric ratio; THF, tetrahydrofuran. b, Flow scale-up of cycloisomerization, isomerization and reduction e-HAT reactions. c Flow rate, 1 ml min -1 ; current density, 0.4 mA cm -2 ; reaction time, 11 h, 0.5 F mol -1 . d Concentration, 0.05 M; current density, 6 mA cm -2 ; reaction time, 20 h, 5.2 F mol -1 . e Flow rate, 25 ml min -1 ; current density, 2 mA cm -2 ; reaction time, 14.5 h, 0.6 F mol -1 . c, Universal conditions for HAT reactivities. f HFIP (7 equiv.), TBABF 4 (0.08 M). g HFIP (2 equiv.), TBABF 4 (0.08 M). d, New HAT reaction: E-selective alkyne semi-reduction. The images show the reaction setups: B1, 1 g Vapourtec ion recycle flow system; B2, 10 g electrochemical batch reactor; B3, 100 g recycle flow apparatus. rt, room temperature .     feasible energetics. The barrier for such a process is notably low (∆G ‡ = 9.7 kcal mol -1 ; where ∆G ‡ is the Gibbs free energy) as it proceeds through a concerted mechanism (for discussion on multiple spin and oxidation states for cobalt complexes, see the Supplementary Information, computation section). These values provide support for step G of the alkyne Z-reduction proposed catalytic cycle. A mechanistic comparison of the E-selective (condition D) and Z-selective (condition C) alkyne reductions was demonstrated by bond dissociation energy analysis (Fig. 5b(xi)). Consistent with our argument made for Fig. 5b(v), (vi), the relatively weak Co−H (42.4 kcal mol -1 ) and Co−C (37.8 kcal mol -1 ) bond strengths in the Co-salen complexes (condition D) indicate that a radical type of HAT pathway is more probable than the organometallic pathway, which favours the E reduction of the alkyne.

Article
The contribution reported provides a new perspective on how an electrocatalytic approach, inspired by decades of energy storage research, can be used in the context of efficient cobalt-hydride genration with a variety of applications in modern organic synthesis. This electroreductive protocol can be performed in an undivided cell, on multiple scales, without strict removal of air or water, and in the absence of expensive silanes and/or boranes or stoichiometric oxidants. Ten different reactions spanning isomerization, reduction and hydrofunctionalization manifolds across dozens of substrates demonstrate the broad scope of this electrochemical entry into Co-H chemistry.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-022-04595-3.