Mechanistic insights into cobalt(ii/iii)-catalyzed C–H oxidation: a combined theoretical and experimental study

The detailed mechanism of the cobalt(ii/iii)-catalyzed C–H oxidation has been investigated by using both theoretical and experimental methods.


Introduction
Transition-metal-catalyzed C-H bond functionalization has attracted widespread attention due to its high efficiency and atom economy. 1 However, most of these transformations have been accomplished with the aid of transition metals such as palladium, 1l,2 rhodium, 3 and ruthenium. 4 Recent improvements in sustainable catalysis have focused on the development of cheaper and naturally abundant rst-row transitionmetal alternatives with comparable catalytic efficacies. 5 Cobalt as a catalyst for C-H bond functionalization has been greatly developed with signicant progress being achieved. 6 For example, electron-rich Co I catalysts have been used in C-H activation/coupling reactions with alkynes, olens, imines, alkyl/aryl halides, and phenol derivatives. Subsequently, high-valent Cp*Co III complexes such as Cp*Co(CO)I 2 , [Cp*Co III (arene)](PF 6 ) 2 , and [Cp*Co(C 6 H 6 )][B(C 6 F 5 ) 4 ] 2 were also discovered by several groups as effective catalysts with which to activate C-H bonds for addition to unsaturated compounds (imines, enones, and aldehydes) and amidation, cyanation, allylation, and halogenations. 7 Additionally, Daugulis reported a new method for Co II -catalyzed C-H bond alkenylation, in which aminoquinoline is utilized as a directing group. 8 The formation of C-O bonds is more difficult than C-C bond generation by C-H activation, especially when alcohols are used as alkoxylation reagents 9 and alkoxyl metal intermediates are prone to undergo b-hydride elimination to form the corresponding aldehydes, ketones, or carboxylic acids. 10 So far, most of the cases that involve transition-metal-catalyzed C-H activation are focused on palladium catalysts 11 but a few reports on copper catalyzed systems have also been disclosed. 12 A Cucatalyzed alkoxylation of arenes under basic and aerobic conditions (Scheme 1a) was also developed in this laboratory. 13 Recently, we reported the rst Co II -catalyzed alkoxylation of C(sp 2 )-H bonds under basic and oxidizing conditions (Scheme 1b). 14 Due to the similarity between cobalt-and coppercatalyzed C-H functionalization, it might be expected that these reactions share a similar reaction mechanism, particularly in the early steps.
For the Cu catalyzed reaction shown in Scheme 1a, a high kinetic isotope effect (KIE ¼ 2.5) is observed, indicating that the C-H bond cleavage is the rate-determining step. It is generally accepted that the C-H bond is activated via a concerted metalation-deprotonation (CMD) mechanism. 15 The similarity between the starting materials and reaction conditions depicted in Scheme 1a and b may imply a similar reaction mechanism for the two reactions but, unexpectedly, a preliminary mechanistic study of the Co II catalytic reaction depicted in Scheme 1b revealed that KIE z 1, indicating that the C-H bond cleavage does not proceed by the same mechanism as the Cu catalyzed reaction. This phenomenon is similar to the Cu II -catalyzed example reported by Yu's group (Scheme 1c), 16 in which an intramolecular single electron transfer (SET) route, not the CMD route, was invoked. The similarity of the KIE results implies that our Co II -catalyzed system may also follow the SET mechanism. Stahl et al. observed a switch between SET-based and CMD-based pathways upon changing from acidic to basic reaction conditions in Cu II -mediated aerobic C-H oxidation, in which a KIE is observed aer the system is made alkaline. 17 Consequently, a comparative study on both the SET-based and CMD-based routes is necessary to determine the detailed reaction mechanism of the Co II -catalyzed alkoxylation.
Despite the great progress that has been made in understanding cobalt catalysis, the detailed reaction mechanisms for these reactions remain poorly understood. In particular, Co II in the reaction system can be oxidized to Co III by the excess oxidant additive (Scheme 1b), and it is unclear whether the exact oxidation state of the catalytic Co species is Co II or Co III . In recent months, Co III -catalyzed C(sp 2 )-H functionalization reactions have been reported (Scheme 2a and b), 7e,h but the mechanisms of these reactions continue to be hotly debated. On the one hand, recent literature studies report that Co III compounds are capable of catalyzing many chelate-directed C(sp 2 )-H functionalization reactions. 7c,e,f,h,18 None of these works indicate the existence of a radical intermediate, all of them supporting organometallic mechanisms, even when the C-H activation is shown by KIE assessment to be not always rate-limiting. On the other hand, the oxidation of a p system by cobalt complexes is a well-known process which involves the intermediacy of carbon-centered radicals. 19 For example, Kochi employed the intermolecular SET concept to elucidate the Co III (TFA) 3 -mediated oxidation of aryl C-H bonds (Scheme 2c) 20 and concluded that an aromatic cation-radical intermediate is formed aer an arene electron is transferred to Co III . Likewise, it is reasonable to speculate that the intermolecular SET mechanism may be appropriate in the Co III -catalyzed C(sp 2 )-H functionalization described above.
No computational study on the mechanism of Co II /Co IIIcatalyzed C(sp 2 )-H alkoxylation appears to have been performed, and experiments on the Co III -catalyzed C(sp 2 )-H alkoxylation of arenes and olens have not been reported. Several questions on the possible mechanisms of this kind of reaction need however to be resolved: (a) which mechanism (CMD or SET) is favored when Co II is the pre-catalyst? (b) In which state (Co II or Co III ) is the actual catalyst in high-valency Co-catalyzed C(sp 2 )-H alkoxylation? (c) Can the Co III compound catalyze the C(sp 2 )-H alkoxylation? If so, which mechanism (CMD or SET) is favorable?
All of these questions led us to study the detailed mechanisms of Co-catalyzed C(sp 2 )-H alkoxylation not only in theory but also experimentally. In the present study, we rst conducted a theoretical investigation of the detailed mechanism of Co IIand Co III -catalyzed C(sp 2 )-H alkoxylation. This predicted that the Co III compound might also catalyze the C-H alkoxylation. Subsequently, experimental results of the Cp*Co III (CO)I 2 -catalyzed C(sp 2 )-H alkoxylation were found to conrm and support the theoretical prediction.
For practical assessment of the energetic viability of the possible routes, the Co II (OAc) 2 $4H 2 O catalyzed C(sp 2 )-H alkoxylation reaction between methanol and (E)-2-(2-methylbut-2enamido)pyridine-1-oxide (denoted as R) was chosen for the theoretical investigation. Ag 2 O acts as the oxidant and NaOPiv as the base. It should be noted that Ag 2 O can react with methanol to generate AgOMe by the equation: and we believe AgOMe may be the actual oxidant in this reaction. This is supported by our observation that AgOTf can also work as the oxidant. 14 The calculations were performed at the M06-L level of density functional theory (DFT), 21 incorporating solvation effects via the appropriate SMD continuum solvation model. 22

Results and discussion
Several possible reaction pathways of Co II /Co III -catalyzed C(sp 2 )-H alkoxylation have been studied by DFT, and further experimental studies have been performed to support the results predicted by the theoretical study.

Possible pathways of Co II -catalyzed C(sp 2 )-H alkoxylation
In order to coordinate with Co, the nitrogen in the reactant R should initially be deprotonated. According to our calculated results shown in Scheme 3, the base OPiv À in the reaction system can abstract the proton attached to the nitrogen easily via transition state TS1, whose formation requires only 14.0 kcal mol À1 . The carboxamide anion intermediate INT1 can coordinate with Co II forming INT2, as shown in Scheme 4. Two kinds of mechanisms, including SET and CMD, to activate the C-H bond are possible. We set the energy of reactant R at 0.0 kcal mol À1 in the energy proles, and all the energies of the other minima discussed below are relative to that unless otherwise specied.
1.1. Concerted metalation-deprotonation mechanism for Co II -catalyzed C(sp 2 )-H alkoxylation. The CMD mechanism of Co II -catalyzed C(sp 2 )-H alkoxylation (pathway 1) consists of two key reaction steps (Scheme 4). The rst step proceeds through a six-membered ring transition state TS2 leading to the formation of intermediate INT3, in which the C(sp 2 )-H bond is activated and the proton is transferred to ligand OAc À . The second step is OMe group transfer from AgOMe to the carbon atom via transition state TS3 affording intermediate INT4. Aer the dissociation of intermediate INT4 to INT5 and Co I OAc, the Co I is oxidized to Co II by AgOMe, and the protonation of INT5 produces the product P. The energy barriers for transition states TS2 and TS3 in the CMD pathway are 28.0 and 20.4 kcal mol À1 (Scheme 4), respectively. Apparently, the C-H activation step associated with TS2 is rate-determining, which is not consistent with the previous KIE (z1.0) results and experimental electron spin resonance (ESR) results. Thus, we can conclude that the CMD mechanism will not be the favorable pathway. 14 In addition, the energy difference between the lowest and the highest stationary points in the energy prole is 34.4 kcal mol À1 , indicating that this pathway is quite unlikely and it is unnecessary to explore the details aer transition state TS3.
1.2. Possible single electron transfer pathways for Co IIcatalyzed C(sp 2 )-H alkoxylation 1.2.1 Intramolecular SET pathways. As described above, C-H bond activation may follow the SET mechanisms, and here two possible intramolecular SET pathways (including pathways 2 and 3) were explored. As shown in Scheme 5, pathway 2 consists of four steps: the ligand exchange between Co II and AgOMe via transition state TS4, the intramolecular single electron transfer of intermediate INT7, the OMe group transfer from Co I to the electron-decient carbon via transition state TS5, and the solvent MeOH-assisted proton transfer from the olenic carbon to the ligand OAc À via transition state TS6. The structure of intermediate INT7 still has only one unpaired electron, a doublet, aer the intramolecular SET, and the two structures of intermediate INT7, before and aer the SET, as in Scheme 5, resemble resonance structures of one another. We assumed that the SET may occur in this reaction pathway, and the details of this SET process are discussed in the next section. With the exception of the SET step, the Gibbs free energy barriers of the rst and third steps in pathway 1 are only 3.6 and 19.8 kcal mol À1 (Scheme 5), respectively. However, the energy barrier of transition state TS6 for C-H bond breaking is extremely high (68.1 kcal mol À1 ), suggesting that pathway 2 is not energetically favorable.
As shown in Scheme 6, pathway 3 consists of three reaction steps. The rst is the intramolecular SET step for INT2. Here, we Scheme 6 Possible intramolecular SET pathway (pathway 3) of Co IIcatalyzed C(sp 2 )-H alkoxylation (energy: kcal mol À1 , distance:Å).  also assumed that the SET can occur in this reaction pathway. The OMe group can also be transferred directly from AgOMe to INT2 via transition state TS7. The last step is a direct proton transfer process via transition state TS8. The energy barrier of the second step is only 14.9 kcal mol À1 , but the energy barrier for the proton transfer is as high as 46.8 kcal mol À1 , indicating that pathway 3 is unlikely to occur under the experimental conditions.
According to the free energy proles shown in Schemes 5 and 6, the two intramolecular SET pathways will not occur even if it is assumed that the SET process in INT2 and INT7 can take place. In an effort to nd out whether the intramolecular SET can really happen, we performed time-dependent density functional theory (TD-DFT) calculations to evaluate the transition energy from the ground-state (GS) to the lowest excited state (ES). As shown in Fig. 1 and 2, the lowest-energy internal SET product arises from transfer of a a-spin electron from the olen p electrons, the singly occupied orbital in INT2/INT7, to the lowest unoccupied orbital of the Co. The excitation energies for the transitions are 37.2 and 44.3 kcal mol À1 for INT2 and INT7 respectively, indicating that the intramolecular SET processes of Co II -catalyzed C(sp 2 )-H alkoxylation cannot occur.
1.2.2 Intermolecular SET pathways. Inspired by the intermolecular SET mechanism of Cu II -catalyzed C(sp 2 )-H functionalization reported by Stahl and Ertem, 17 we considered the possible intermolecular SET transition processes for this reaction. As shown in Scheme 7, the Gibbs free energy differences for intermolecular SET processes range from 31.2 to 50.3 kcal mol À1 , demonstrating that it would be difficult for the intermolecular SET to occur.

The possible pathways of Co III -catalyzed C(sp 2 )-H alkoxylation
We have explored almost all the possible mechanisms for Co IIcatalyzed C-H alkoxylation under basic conditions, but the above calculated results fail to explain the experimental results reasonably. As mentioned above, Co II might be oxidized to Co III by AgOMe, and thus the Co III -catalyzed C(sp 2 )-H functionalization reaction pathway would be feasible. As shown in Scheme 8, the energy barrier for oxidation of Co II to Co III by AgOMe is only 2.5 kcal mol À1 , showing that it would be very facile. Then, starting from Co III (OAc) 2 (OMe), there are also two kinds of reaction mechanisms including SET and CMD starting with the Co III catalyst and INT1.
2.1 Concerted metalation-deprotonation pathway for Co IIIcatalyzed C(sp 2 )-H alkoxylation. To facilitate comparison, the ligands around the cobalt atom are unchanged from INT2 to INT16. Two key transition states, TS10 associated with concerted metalation-deprotonation process and TS11 associated with the OMe transfer, are assigned to the CMD mechanism of Co III -catalyzed C(sp 2 )-H alkoxylation, shown as pathway 4 depicted in Scheme 9. The structure of transition state TS11 indicates that the OMe is transferred by a three membered (C-O-Co) ring, which is remarkably different from the direct OMe transfer from AgOMe in transition state TS3 of pathway 1. Obviously, the rst energy barrier via the CMD pathway, at 37.0 kcal mol À1 (Scheme 9), is extremely high, indicating that C-H activation via this pathway is very energy-demanding.
Scheme 11 Possible intermolecular SET pathway 7 of Co III -catalyzed C(sp 2 )-H alkoxylation (energy: kcal mol À1 , distance:Å). If the pre-catalyst Co II (OAc) 2 were to be oxidized to a Co III intermediate, then the intermolecular SET pathway 5 becomes the most energetically favorable among all the eight possible pathways being considered, and we believe that the Co III should be the actual catalyst for this kind of C(sp 2 )-H alkoxylation irrespective of the Co II or Co III compound that was added to the reaction system. The INT11 has the highest energy (19.7 kcal mol À1 ) in the energy prole of pathway 5, so we can conclude that the intermolecular SET step rather than the C-H activation should be rate-determining in this reaction.

Combined experimental and theoretical exploration of Co III -catalyzed C(sp 2 )-H alkoxylation
The above computational results prompt us to further explore whether the SET mechanism for the direct Co III -catalyzed alkoxylation of C(sp 2 )-H bond is experimentally feasible. To Scheme 15 Controlled experiments for Co III -catalyzed C(sp 2 )-H alkoxylation.
conrm the plausibility of this prediction, experimental research was initiated with the reaction between ethanol and 2-benzamidopyridine 1-oxide (1a) catalyzed by Cp*Co(CO)I 2 . When a solution of benzamide (1a) in ethanol was treated at 70 C with Cp*Co(CO)I 2 (20 mol%), Ag 2 O (2 equiv.) and NaOAc (1 equiv.) under an air atmosphere (Scheme 13), the desired ethoxylated product 3aa was obtained in a good yield.
Diversely substituted amides were tolerated under the Cp*Co(CO)I 2 -catalyzed alkoxylation to furnish the desired products (3) in moderate to good yields (Scheme 14). The aryl substrates possessing electron-rich and electron-decient functional groups underwent the transformation successfully (3aa-3ka). For meta-substituted amides (1g, 1h), the reaction tended to take place at the less hindered position and an iodo group on the aryl ring was also compatible. The heterocyclic substrate (1k) afforded the corresponding ethoxylated product in 63% yield. Several alcohols were tested and demonstrated to give the products in yields ranging from 48-85% (3ab-3ae). Olenic carboxamides were also found to follow the protocol under the reaction conditions (3la-3oa).
In addition, control experiments revealed that the addition of 2,2,6,6-tetramethylpiperidine-N-oxyl (TEMPO, 1.5 equiv.) as a radical quencher completely inhibits the reaction (Scheme 15a). This result indicates that a radical pathway (SET mechanism) is involved. A 1 : 1 mixture of 1a and [D5]-1a was then treated with ethanol. No kinetic isotope effect (KIE ¼ 1.0) was obtained (Scheme 15b), suggesting that C-H bond cleavage of arenes is not the rate-limiting step, in agreement with our calculated results. Additionally, the electron paramagnetic resonance (EPR) spectrum of the reaction system demonstrated the existence of the single electron (g ¼ 2.23003, see the ESI †). The above experiments can be explained by our computational study, which they support.
We have performed a theoretical study on the mechanism of Cp*Co III complex catalyzed C-H activation. As shown in Scheme 16, it is very similar to the favorable pathway 5 catalyzed by the Co III complex, and the SET process is still the rate-determining step based on the energy prole, which is also in agreement with the KIE and EPR experiments. Based on the above computational and experimental results, we can propose the general mechanism for not only the Co II -but also the Co IIIcatalyzed C(sp 2 )-H alkoxylation reaction. As shown in Scheme 17, AgOR (OR ¼ alkoxy group) is generated and is the actual oxidant, and the reactant I is transformed to intermediate II by proton transfer to the base. First, it is an intermolecular SET process between intermediate II and Co III L 3 (L ¼ ligand) generating intermediate III, L À , and Co II L 2 , which can be oxidized to Co III L 2 (OR) by AgOR. Second, intermediate IV is formed by the coordination of intermediate III with Co III L 2 (OR). Subsequently, the OR group is transferred from the Co to the substrate leading to the formation of intermediate V. Intermediate VI is formed by the coordination of base B À . As shown in Scheme 16, it should be noted that the coordination would not occur when the catalyst is Cp*Co III . The next step is the proton transfer to base ligand and, nally, intermediate VII dissociates to VIII and Co II . Co II can be oxidized again to Co III which catalyzes the next cycle, and intermediate VIII continues to be converted to the product IX. Although the tridentate chelate VII should be very stable, we still think it can be dissociated directly, because the structural transformation to VIII (i.e. INT18 in Scheme 10 or INT18 0 in Scheme 16) is a highly exothermic process, which can compensate for the energy for the direct dissociation.

Conclusions
We have explored multiple possible SET and CMD mechanisms of Co II /Co III -catalyzed alkoxylation of C(sp 2 )-H bond in theory, and found that the intermolecular SET pathway in which the Co III works as the actual catalyst (pathway 5) should be the most favorable pathway, even when the Co II (OAc) 2 was added as a catalyst. Generally, there are three key steps in the favorable pathway of the high-valent Co-catalyzed C(sp 2 )-H alkoxylation, i.e. the intermolecular SET between the reactant and Co III , the alkoxylation and, nally, the breaking of C(sp 2 )-H bond. The calculated results indicate that the SET step, rather than the C(sp 2 )-H activation, is rate-determining. Guided by the computed mechanism, the rst Co III -catalyzed C(sp 2 )-H alkoxylation is reported. A variety of amides with electron-rich and electron-poor functional groups were experimentally suitable for the Cp*Co(CO)I 2 -catalyzed alkoxylation, thus extending the application for cobalt-catalyzed functionalization of C-H bonds. All observations in the experiment can be explained by our computational results.
This work provides another special SET-based example in the context of chelate-directed C(sp 2 )-H functionalization. The detailed Co II /Co III -catalyzed SET and CMD models and pathways should be helpful for chemists to understand the general mechanism and the roles of the additives and catalysts in the Co II /Co III -catalyzed C-H functionalization, and thus provide valuable insights into rational prediction and design of the more efficient catalysts in this kind of reaction.

Computational and experimental details
All of the calculations were performed using Gaussian 09. 23 Computed structures are illustrated using CYLView. 24 The density functional theory (DFT) method was applied since it has been successfully used in many studies of organocatalysis, 25 organometallic catalysis, 26 and biological reaction mechanisms. 27 The density functional theory calculations were carried out with the M06-L functional in the presence of the SMD continuum solvation model with methanol (or ethanol) as the solvent. All the structures were completely optimized using a combined basis set: the LanL2DZ basis set 28 was used for Co, I, and Ag along with the 6-31G(d, p) basis set for C, N, H, and O. The frequency calculations were performed at the same level at 298 K and 1 atm, and vibrational analysis was performed to conrm the optimized stationary points as true minima with no imaginary frequency, or transition states with one and only one imaginary frequency, on the potential energy surface and to obtain the thermodynamic data. On the basis of the optimized structures at the M06-L/6-31G(d, p)//LanL2DZ level, the energies were then rened by M06-L/6-311++G(2df, 2pd)//SDD 29 singlepoint calculations with the same solvent effects. We checked the S 2 values of all the stationary points obtained above, and found that all the differences between the obtained S 2 value and the normal S 2 value are less than 7.5%, so spin contamination can safely be ignored.
We chose to conduct discussions based on Gibbs free energies rather than Born-Oppenheimer energies, which are the electronic (including nuclear-repulsion) energies plus ZPEs. Free energy contributions were added to single-point M06-L electronic energies computed with the SDD basis set on Co, I, and Ag and the 6-311++G(2df, 2pd) basis set on all other atoms to arrive at nal, composite free energies.
We have calculated and compared the free energy DG tot and the DG 50% (the entropy is cut by 50%) proles of the favorable reaction pathway (pathway 5). As shown in Schemes 10 and 19, although a computational error does exist in the entropy calculation for a multimolecular reaction step (such as a SET process), the calculated results can still predict exactly that the SET process is the rate-determining step, which is in agreement with the KIE experiment. Therefore, we think that the calculated results can reliably explain the experimental results.
Recently, Singleton et al. investigated the mechanism of alcohol-mediated Morita-Baylis-Hillman (MBH) reactions using the most popular DFT methods, i.e. B3LYP and M06-2X. 34 They found that the calculated results by the two methods are remarkably different, and the rate-determining step cannot be predicted correctly by using the M06-2X method, so they concluded that the computations aid in interpreting observations but fail utterly as a replacement for experiment. In this work, similar computational tests to Singleton have been performed for the key C-H activation in the favorable reaction pathway (pathway 5) by using B3LYP and M06-L methods. As summarized in Table 2, the calculated results indicate that the free energy barriers DDG tot , DDG 50% , and DDG explicit (calculated in the explicit solvents without implicit model) are close and the free energy barriers obtained by the different DFT methods are not signicantly different, which is remarkably different from Singleton's work. As mentioned above, we believe that the computational errors in this system are not very signicant and the calculated results using the M06-L method are consistent with experiment. In addition, multiple pathways involving explicit methanol for this kind of reaction 35 are considered in both implicit and explicit models, but the H atom of HOMe returns automatically when we try to locate the possible structures of the transition states TS19, TS20, TS21, and intermediate INT27 (depicted in Scheme S1 of the ESI †), which is mainly because the basicity of OMe À is stronger than that of OAc À , so we did not further explore details of these pathways.
The M06-L method has been successfully used in the theoretical report on the mechanism of transition-metal catalyzed C-H oxidation, 17 and the above computational tests also indicate that it should be suitable for investigating the systems including Co complexes.
Alcohol (1.2 mL) was added to a mixture of amide (0.15 mmol), Ag 2 O (69 mg, 0.3 mmol, 2 equiv.), NaOAc (12 mg, 0.15 mmol, 1 equiv.), Cp*Co(CO)I 2 (14 mg, 20 mol%). Then, the reaction system was stirred at 70 C for 12 h under an air atmosphere. Aer cooling to RT, the alcohol was evaporated under reduced pressure, and 2 N HCl (5 mL) was added to the residue. The mixture was extracted with CH 2 Cl 2 and the organic layer was dried over anhydrous Na 2 SO 4 . Purication of the residue gave the product 3.