A Unified Electro- and Photocatalytic CO2 to CO Reduction Mechanism with Aminopyridine Cobalt Complexes

Mechanistic understanding of electro- and photocatalytic CO 2 reduction is crucial to develop strategies to overcome catalytic bottlenecks. In this regard, herein it is presented a new CO 2 -to-CO reduction cobalt aminopyridine catalyst, a detailed experimental and theoretical mechanistic study toward the identification of bottlenecks and potential strategies to alleviate them. The combination of electrochemical and in-situ spectroelectrochemical (FTIR/UV-Vis SEC) studies together with spectroscopic techniques (NMR, EXAFS) lead us to identify elusive key electrocatalytic intermediates derived from complex [Co(py Me tacn)(OTf) 2 ] (1) (py Me tacn = 1-[2-pyridylmethyl]-4,7-dimethyl-1,4,7-triazacyclononane) such as a highly reactive cobalt (I) (1 (I) ) and cobalt (I) carbonyl (1 (I) -CO) species. 1 (I) was obtained by electrochemical reduction of 1 (II) , and characterized by NMR, EXAFS and FTIR/UV-Vis SEC. The combination of spectroelectrochemical studies under CO 2 , 13 CO 2 and CO with DFT disclosed that 1 (I) directly reacts with CO 2 to form the pivotal 1 (I) -CO intermediate at the 1 (II/I) ABSTRACT: Mechanistic understanding of electro- and photocatalytic CO 2 reduction is crucial to develop strategies to overcome catalytic bottlenecks. In this regard, herein it is presented a new CO 2 -to-CO reduction cobalt aminopyridine catalyst, a detailed experimental and theoretical mechanistic study toward the identification of bottlenecks and potential strategies to alleviate them. The combination of electrochemical and in-situ spectroelectrochemical (FTIR/UV-Vis SEC) studies together with spectroscopic techniques (NMR, EXAFS) lead us to identify elusive key electrocatalytic intermediates derived from complex [Co(py Me tacn)(OTf) 2 ] ( 1 ) (py Me tacn = 1-[2-pyridylmethyl]-4,7-dimethyl-1,4,7-triazacy-clononane) such as a highly reactive cobalt (I) ( 1 (I) ) and cobalt (I) carbonyl ( 1 (I) -CO ) species. 1 (I) was obtained by electrochemical reduction of 1 (II) , and characterized by NMR, EXAFS and FTIR/UV-Vis SEC. The combination of spectroelectrochemical

aminopyridine catalyst, a detailed experimental and theoretical mechanistic study toward the identification of bottlenecks and potential strategies to alleviate them. The combination of electrochemical and in-situ spectroelectrochemical (FTIR/UV-Vis SEC) studies together with spectroscopic techniques (NMR, EXAFS) lead us to identify elusive key electrocatalytic intermediates derived from complex [Co(py Me tacn)(OTf) 2 ] (1) (py Me tacn = 1-[2-pyridylmethyl]-4,7-dimethyl-1,4,7-triazacyclononane) such as a highly reactive cobalt (I) (1 (I) ) and cobalt (I) carbonyl (1 (I) -CO) species. 1 (I) was obtained by electrochemical reduction of 1 (II) , and characterized by NMR, EXAFS and FTIR/UV-Vis SEC. The combination of spectroelectrochemical studies under CO 2 , 13 CO 2 and CO with DFT disclosed that 1 (I) directly reacts with CO 2 to form the pivotal 1 (I) -CO intermediate at the 1 (II/I) redox potential. At this redox potential the theoretical energy barrier for the C-O bond cleavage was found to be as low as 12.2 kcal·mol -1 . However, the catalytic process does not proceed at the 1 (II/I) redox potential, due to the formation of 1 (I) -CO, which is a thermodynamic sink and the CO release restricts the electrocatalysis. In agreement with the experimental observed CO 2 -to-CO electrocatalysis at the ]. In contrast, under photochemical conditions, the catalytic process smoothly proceeds at the 1 (II/I) redox potential. Under the latter conditions, it is proposed that the electron transfer rate is under diffusion control and then the CO release from 1 (II) -CO is kinetically favored, facilitating the catalysis. Finally, we have found that visible light irradiation has a positive impact under electrocatalytic conditions. We envision that light irradiation can serve as an effective strategy to improve the CO 2 reduction of molecular catalysts, via alleviating bottlenecks, such as the CO poisoning.
File list (1) download file view on ChemRxiv Manuscript_ACS_20-06-2019.pdf (3.44 MiB) INTRODUCTION CO2 reduction is one of the most promising approaches for sustainable production of renewable fuels and chemicals. 1 The design of efficient catalysts for CO2 reduction entails a fundamental understanding of the parameters that control the catalytic activity and selectivity. 2 However, to obtain insights into the CO2 reduction mechanism is highly challenging and the mechanism still remains poorly understood. 3 In this regard, coordination complexes serve as platforms to implement different strategies to interrogate the operative mechanisms. Among the different families of active catalysts for CO2 reduction, cobalt complexes containing polypyridyl or aminopyridyl ligands are easily tunable and therefore of interest for intermediate characterization and mechanistic studies. These cobalt complexes, in combination with specific photosensitizers (e.g. Ru 4,5 or Ir 6 ) show high activity and selectivity for the light-driven CO2to-CO reduction. 7 However, except for some specific examples, 6a,b,8 their electrocatalytic performance is still limited by their low stability, resulting in low turnover numbers and faradaic yields for CO formation. 8b,9 Despite many efforts, further understanding of the reaction mechanism is required to identify the bottlenecks of the electrocatalytic reaction and to explain the differences in performance between the photo-and electrocatalytic conditions. 7b, 8a,10,11 Scheme 1. Summary of the hypothesized mechanisms for electrocatalytic CO2 reduction to CO by molecular cobalt complexes bearing neutral nitrogen chelating ligands. A formal oxidation state is given for the different Co species. i) CO 2 binding to Co I ii) CO 2 binding to Co 0 L N4 = tetradentate neutral nitrogen donor ligand From the main attempts to elucidate the CO2-to-CO mechanism, the most commonly accepted hypotheses for Co catalysts based on nitrogen donor ligands are summarized in Scheme 1. For several reported Co II complexes containing highly basic ligand frameworks, the 1e --reduced Co I species are nucleophilic enough to coordinate CO2. The CO2 coordination to Co I complex can be indirectly detected and measured by the anodic shift of the half-wave potential (E1/2) of the Co II/I redox couple under CO2. 12 A number of computational studies performed on these systems lead to a proposed general scheme (Scheme 1, route i) for either the electro-or photocatalytic CO2-to-CO reaction based on two critical steps 5b,c,6b : 1) CO2 binding at the in-situ generated Co I species to form a Co III CO2 adduct; 2) the cleavage of the C-O bond, which can be promoted by a second molecule of substrate to give free CO3 2in aprotic media (route i.a) or by protons (route i.b). In the latter case, the reaction could take place through a PCET pathway with the formation of OH -(or water after further protonation). On the other hand, for systems bearing less basic ligands, the CO2 binding and catalytic conversion to CO might require the formation of a formal Co 0 intermediate, without any prior interaction between the Co I species and CO2 (Scheme 1, route ii). 8,9b,13 Chart 1. Selected cobalt catalysts for electrochemical CO2to-CO reduction.
In comparison to the C-O bond cleavage, the CO release step from the final Co II -CO intermediate is generally considered facile and received much less attention (Scheme 1). Nevertheless, recent spectroelectrochemical studies under CO2, suggested that the formation of stable low oxidation state carbonyl complexes under CO2 may result in the deactivation of the molecular transition metal catalysts. In particular, it was proposed that a Fe 0 -CO species is formed in the course of the electrochemical CO2 reduction by a quaterpyridine Fe complex. 8a Analogous Ni I -CO intermediates were detected for cyclam-type and aminopyridyl Ni complexes. 14 However, a direct observation of carbonyl intermediates, formed during electrocatalytic CO2 reduction mediated by cobalt complexes, is rare and mostly unexplored. 15 Herein we present a compelling mechanistic study of the electrochemical CO2-to-CO reduction mediated by a [Co(py Me tacn)(OTf)2] (1) (py Me tacn = 1-[2-pyridylmethyl]-4,7-dimethyl-1,4,7-triazacyclononane) complex, recently studied for light-driven H2 evolution 16 and organic substrates reduction 16a,17 (Chart 1). Insitu spectroelectrochemistry (SEC) studies reveal that a crucial and rarely reported cobalt(I) carbonyl species (1 (I) -CO) is formed under catalytic conditions at redox potential values of the Co II/I redox couple. Electrochemical and spectroscopic techniques (UV-Vis, FTIR, EXAFS and NMR) were employed to characterize the reduced species observed under Ar and CO2 atmosphere. These results together with DFT studies have served to present, for the first time, a full catalytic cycle integrating the pH and redox potential effects along the mechanism. Additionally, we have explored the use of visible light irradiation as an effective strategy to induce the CO release from the Co-CO species improving the performance of CO2 reduction catalysis. Finally, we propose a unified view of the CO2-to-CO reduction mechanism under both electro-and photocatalytic conditions.

RESULTS AND DISCUSSION
Compound 1 has labile triflate ligands that exchange fast with coordinating solvent molecules to form the doubly charged CH3CN complex (1 (II) ), as characterized by X-ray diffraction (Figure 1). The two main features of the cyclic voltammogram of 1 (II) , in anhydrous acetonitrile under argon atmosphere, are an irreversible wave at -1.74 V assigned to a Co II/I process (potential values referenced vs. Fc +/0 unless indicated) 16a and a quasi-reversible peak at -2.36 V, assigned to a formal Co I/0 process (Figure 1, Figure S1). Inset: magnification range between -2.0 to -0.5 V. Scans window from -0.5 to -1.9 V are shown with dotted lines. Bottom) under Ar (black), and with added H2O (0.55 M) under Ar (blue) and CO2 (green). Inset: X-ray crystal structure of complex 1 (II) , triflate counterions and hydrogen atoms have been omitted for clarity.
Under CO2, the Co II/I peak potential shifts to positive values depending on the CO2 concentration; higher the substrate concentration larger the positive shift of the Ep(Co II/I ) value, reaching a maximum value of ΔEp = 52 mV under CO2 saturation ([CO2] = 0.28  . This is indicative of a fast reaction between the electrochemically generated Co I species and CO2 in the timescale of the CV experiment. In addition, CO2 saturation induces a small but significant current increase (10%) at the Co II/I wave in comparison with the CV under inert atmosphere (Figure 1), suggesting that the process cannot be described as mere coordination of CO2 to the reduced Co I species (Scheme 2). By reversing the potential just after the Co II/I couple, a small new oxidation peak at -0.82 V appears in the back scan, indicating the formation of a new species under CO2 ( Figure  1). The same anodic peak also appears under CO atmosphere, which suggests the formation of a common intermediate, most likely a cobalt carbonyl species ( Figure S2, S3). Under CO, the Co II/I reduction wave shifts to even higher redox potentials (about twice than under CO2, ΔEp= 118 mV), which is a sign of a strong interaction between Co I and CO. Unfortunately, the irreversibility of the reduction wave prevents a precise calculation of the equilibrium constant (KCO) by cyclic voltammetry. However, it can be broadly estimated as >> 10 4 M -1 based on the peak potential shift observed under CO (see section 2.1 in the SI). Nevertheless, we were able to obtain the kinetic binding constants (kCO2 = 2·10 3 M -1 s -1 and kCO = 6·10 6 M -1 s -1 ) by applying equation 1, which relates the Ep shift with the rate constant (k), the scan rate (v) and the substrate concentration (C). 8a Further reduction under CO2 produces a catalytic wave at redox values close to the Co I/0 process, reaching more than 4-fold current increase (icat/ip = 4. The electrochemistry of 1 (II) in the presence of water (0.55-9.26 M) also gives some insight into the mechanism. Under Argon or CO2, the presence of H2O (0.55 M) did not produce any modification to the Co II/I reduction process (Figures 1 and S8), which implies that the proposed chemical reaction between the electroreduced cobalt complex and CO2 is not significantly influenced by water (equation 3). 16a However, at lower redox potentials the presence of water (0.55 M) induces a clear shift and current increase in both cases Ar and CO2. Interestingly, controlled-potential electrolysis at this new catalytic wave under CO2 and in presence of water shows an excellent CO/H2 selectivity (no H2 detected, 3.6 TON CO after 3 h at Eappl = -2.37 V, 0.5 M H2O, Figure S9). This selectivity is remarkable considering that under Argon there is also an induced catalytic current by the presence of water at the same redox potential.
On the other hand, preparative-scale electrolysis of 1 (II) (1 mM, Eappl = -2.46 V) under a constant flow of CO2 (30 mL min -1 ) in anhydrous CH3CN yields 5.5 TONs of CO after 6 h. This result provides evidence for catalytic CO2 reduction even in the absence of an added proton source ( Figure S10). Gas-chromatographic analysis (see experimental section in the SI) indicates that CO is the major product formed, along with the formation of carbonate. H2 was not detected and formate was only detected in traces (TON HCO2 -~ 0.1).
Spectroscopic and theoretical evidence for the formation of 1 (I) -CO and 1 (I) . The already mentioned current increase at the Co II/I reduction peak under CO2 (c.a. 10 %) suggests further reactivity. To confirm this hypothesis, we employed in-situ spectroelectrochemical techniques 18 (UV-Vis-SEC, FT-IR-SEC) and spectroscopic characterization ( 1 H-NMR and EXAFS) of electrochemically generated intermediates by bulk electrolysis at the Co II/I reduction peak. FTIR-SEC experiments in an OTTLE cell 19 revealed the formation of a new species at the first reduction event in a CO2 saturated electrolyte. A stepwise scan to negative potentials showed the progressive formation of a new band (νCO = 1910 cm -1 ) when the applied potential reaches the first reduction peak (ca. -1.7 V, Figure S11), which was not observed in absence of CO2. The same IR feature is also formed under CO, but about 3-fold more intense than under CO2 (Figure 2, S12). Labeling experiments with 13 CO2 indicated that the detected intermediate derives from CO2 reduction (ν 13 CO = 1866 cm -1 , stretching band shifts 43 cm -1 towards lower energy). The observed vibration and isotopic shift are comparable to an uncommonly reported Co I carbonyl complex. The direct reaction of the chemically synthesized [Co I (L C1 )] + (L C1 = 5,7,7,12,14.14-hexamethyl-l, 4,8,11-tetraazacyclotetradeca-4,11diene)  To further confirm the nature of the putative cobalt carbonyl intermediate, we computationally modeled the theoretical IR spectra of possible cobalt carbonyl species bearing the py Me tacn ligand, as well as known homoleptic cobalt carbonyl complexes (section 3.5 of the SI). 21 This together with previously reported values, 22 allowed us to discard homoleptic cobalt carbonyl complexes and other cobalt carbonyl complexes in oxidation state II and 0 bearing the py Me tacn ligand. The calculated 1912 cm -1 feature of [Co I (py Me tacn)(CO)] + (1 (I) -CO) matches with the experimental νCO value of 1910 cm -1 complex as well as the theoretical 13 C shift ( Figure 2).
Upon an oxidative back scan after the formation of 1 (I) -CO in SEC under either CO2 and CO atmosphere, the 1910 cm -1 feature of 1 (I) -CO is preserved until about -0.8 V. Further oxidation leads to the disappearance of the 1 (I) -CO signal recovering the original spectrum (Figures S12 B and S11 B, respectively), suggesting that the anodic peak at -0.8 V observed in the CVs under CO2 and CO corresponds to the reoxidation of 1 (I) -CO. Similar results were obtained under CO2 in the presence water (0.5 M), showing a mixture of 1 (I) -CO (νCO = 1910 cm -1 ) and carbonate species (1676-1631 cm -1 ) when the applied potential matches the Co II/I process ( Figure  S13).
Altogether offers a compelling evidence for the formation of [Co I (py Me tacn)(CO)] + (1 (I) -CO) at the Co II/I redox potential through CO2 reduction to CO. We propose that the formation of Co(I) carbonyl species may be more general since we have also detected by IR spectroelectrochemistry the formation of [Co I (tpa)(CO)] + (see section 2.5 of the SI). In addition, the formation of this carbonyl species is necessarily fast because it is detected in the CV time scale (Figure 1 inset).
UV-Vis-SEC experiments provided complementary information to FT-IR-SEC. Under Ar atmosphere, a new d-d transition band appears at λmax 459 nm at the first reduction wave (ca. -1.9 V) (Figures 3 B, S14). The formation of the putative 1 (I) is reversible, recovering 1 (II) upon back scan oxidation. This clear indication of the formation of an elusive cobalt (I) intermediate motivated us to further explore its electronic structure and coordination environment, which is discussed below.
Conversely, the reduction of 1 (II) under CO2-saturated conditions leads to the growth of two new bands at 308 and 427 nm, indicating the formation of a new species (Figures S15). In agreement with the CV data, these features disappear at an approximated applied potential of -0.8 V during the reverse sweep ( Figure S16). In line with the above discussion, the same intense absorptions at 308 and 427 nm resulted from UV-Vis-SEC experiments under CO at the Co II/I potential, consistent again with the formation of 1 (I) -CO ( Figure  S17). UV-Vis-SEC experiments are also interesting because they provide an estimation of the concentration of the species formed in solution. By analyzing the differences in absorbance values at 427 nm under CO2 and CO, a 3-fold increase in the 1 (I) -CO concentration is observed under CO relative to CO2. This increase is comparable to that observed by FT-IR-SEC ( Figure S18).
To explore the oxidation states and coordination geometries of the possible intermediates we performed Co K-edge XAS. To calibrate our analysis, we first studied 1 (II) and 1 (III) as reference complexes relevant to catalysis (Eq. 6, Figures S20 -S22). In CH3CN, 1 (II) complex has a pre-edge arising from 1s → 3d transitions of 0.05 normalized intensity units occurring at 7710.6 eV accompanied by a rising edge at 7721.2 eV. The XANES profile is consistent with a centrosymmetric pseudo-octahedral coordination geometry ( Figure  3C). 23,24 EXAFS analysis supports two coordination shells having 2 N/O scattering atoms at 2.0 Å and 4 N/O scattering atoms at 2.16 Å, which is consistent with the optimized DFT geometry (Figure 3 D and section 2.6 of the SI for details). To study the reduced species under Ar (Eq. 1) by XAS and EXAFS, we performed bulk electrolysis at -1.8 V of a solution containing 1 (II) (5 mM, in anhydrous CD3CN at -40 ºC under Argon). After 1epassed, the solution was frozen and analyzed by Co K-edge XAS.  However, a penta/hexa-coordination equilibrium cannot be discarded since the XANES suggests a pseudo octahedral environment and DFT calculations show an almost isoenergetic penta/hexa-coordinate environments for Co I (the five-coordinate structure is 0.6 kcal·mol -1 more stable than its hexacoordinated counterpart). This is in agreement with the previously reported solid state structures of five-coordinate formal Co I complexes based on polypyridine, 25 and pyridine-bisimine 40 ligands. 13,23,13,25 The 1 H-NMR of the sample showed the formation of a new paramagnetic species is the range of 140 -5 ppm at 235 K ( Figure S23). The paramagnetic nature of the sample together with the DFT analysis of the spin density is consistent with a high-spin d 8 configuration of the metal center ( Figure  S24). To note is that the electronic structure of 1 (I) is rather unique. Previous electronic structure studies of formal cobalt (I) complexes with N-donor ligands are better described as Co(II) with a reduced ligand, resulting in a challenging characterization of this naturally elusive intermediates. On the other hand, ligands with high crystalline-field splitting favor the formation of low spin Co(I) complexes. 26 In this regard, the data presented herein are one of the few compelling evidence of the formation of a d 8 high-spin Co(I) species reported so far.
While IR-SEC is an in-situ experiment that lasts seconds, the exsitu bulk XAS experiment under CO2 lasts at least 20 min prior to sampling an aliquot for analysis, which prevents the quantitative accumulation of intermediates without decay. Nevertheless, we have performed CPE experiments under CO2 analogous to the ones under argon. In this case, XAS shows a more effectively reduced metal center than the starting Co II complex and distinct in bond metrics to the Co I obtained under Ar (Figures 4 B, S31). Although we were not able to fit a short bond distance as expected for a Co-CO bond, the pre-edge intensity (0.04) as well as the pre-edge and rising edge energies (7710.3 eV;7720.4 eV) are similar to that of a Co II carbonate reference (1 (II) -CO3) generated by mixing 1 (II) with 1 eq. of tetrabutylammonium hydrogencarbonate (TBACO3H). Furthermore, EXAFS analysis shows comparable bond metrics and coordination numbers in both the reference and electrochemically generated sample having 2N/O scattering atoms at 2.05 A and 4N/O scattering atoms at 2.16 A (Panel S1). In addition, upon deliberate exposure to ambient atmosphere, we generated a product that approaches the profile of the chemically generated Co III -carbonate (Eq. 7), both in terms of XANES and EXAFS analysis. The new Co III carbonate species exhibits a diamagnetic 1 H-NMR spectrum, as expected for a low-spin d 6 metal center ( Figure S25). X-ray diffraction of crystals obtained after electrolysis of 1 (II) under CO2 confirmed the formation of a six-coordinate Co III complex [Co III (py Me tacn)(η 2 -CO3)](PF6) (1 (III) -CO3, Figure 4 A). In addition, electrolysis of 1 (II) (1 mM, at -1.7 V in anhydrous CH3CN) under CO2 produced CO over the first 20 minutes of reaction (Figure 4 D), consistent with the reduction of CO2 through the 1 (I) -CO formation pathway. Unfortunately, low-temperature bulk electrolysis at -40 ºC did not provide further evidence of the reactivity.
The formation of 1 (III) -CO3 species can be explained by the O2 oxidation of in situ generated Co II -carbonate species during the CO2 reduction electrolysis (formally: 2CO2 + 2e -→ CO3 2-+ CO). However, we cannot fully rule out that some of the carbonate is formed via CO2 hydration since the water content in the solution is in the range of 40 -60 ppm, as analyzed by Karl-Fischer titration under our conditions. 20a The sum of these results led us to hypothesize that the solution reaches a thermodynamic equilibrium, where the 1 (I) -CO as well as the Co II -carbonate disfavor the CO2 reduction at the Co II/I redox potential (Scheme S1). Indeed, the CV of the solution after electrolysis of 1 (II) under CO2 (Figure 4 C) is consistent with the analogous CV in the presence of one equivalent of (TBACO3H). The latter experiment also showed that at the Co I/0 redox potential there is catalysis which implies that carbonate is not involved in the catalytic CO2 reduction process.
Computational modeling of the mechanism. With the aim to give additional insight into the reactivity of electrochemically generated Co I species with CO2, we studied the reaction energy profile by DFT. B3LYP-D3(SMD) / aug-cc-pVTZ(-d H ,-f C,N,O ,-g Co ) // B3LYP-D3(SMD) / 631+G* was selected as the level of theory as it has been shown to reproduce well the catalytic activity of related systems. 16a Computed Gibbs energies were corrected for the catalytic conditions, i.e. substrate (CO2) and product (CO) concentrations of 0.28 M and 50 μM, respectively. 27 For a detailed description of the computational methodology and for the optimized structure coordinates see sections 3.1 and 6 of the SI. In aprotic conditions, CO2 is known to act as an oxide acceptor assisting the reductive disproportionation reaction to CO and CO3 2-. 7b Nevertheless, residual water contained in anhydrous CH3CN may have an important role in the protonation of the cobalt-CO2 adducts. To account for available protons, we studied the pH dependency of the mechanism. At that low proton concentration, a proton assisted mechanism could be operative but competitive with an aprotic CO2 reductive disproportionation mechanism. Therefore, in the first part of this section, we will discuss possible mechanisms for the (sub)stoichiometric formation of the key 1 (I) -CO intermediate under both proton-assisted and aprotic conditions. Later, we will comment on the cobalt-catalyzed CO2 reduction mechanism at the Co I/0 redox potential focusing on the effect of the pH and the redox potential on the thermodynamics and kinetics of the catalytic reaction.
Formation of 1 (I) -CO. According to the experimental data, the (sub)stoichiometric reduction of CO2-to-CO occurs at the first Co II/I reduction wave (ca. -1.7 V), yielding 1 (I) -CO and Co II -carbonate species as the main reaction products. We have shown that, although the C-O bond cleavage can take place, the reaction does not proceed catalytically. Thus, we first focused our efforts on understanding the activation of CO2 at the Co II/I redox potential by means of computational modeling. In order to reproduce our experimental conditions, the theoretical Co II/I reduction potential (-1.91 V) was considered as the working redox potential to calculate the energy profiles ( Figure 5). As depicted in Figure 5 A, in the proton-assisted mechanism, the nucleophilic Co I species ([L N4 Co I (S)] + ) formed by 1ereduction of [L N4 Co II (S)] +2 binds CO2 to form a higher in energy carboxylate adduct ([L N4 Co III -CO2] + ), with a 8.8 kcal·mol -1 energy barrier. Then, the subsequent 1ereduction gives the slightly endergonic [L N4 Co II -CO2] at the defined redox potential. Further protonation of the highly basic [L N4 Co II -CO2] species yields the thermodynamically favored [L N4 Co II -CO2H] + (pKa = 28.4).
The subsequent C-O bond cleavage step has been proposed as the rate determining step (r.d.s.) in the light-driven CO2-to-CO reduction mechanism catalyzed by other macrocyclic Co complexes. 8b, 28 In our case, the calculated Gibbs energy barrier for the heterolytic C-O bond cleavage from [L N4 Co II -CO2H] + to give [L N4 Co II -CO(OH)] + is 16.0 kcal·mol -1 ( Figure 6A). This result is in agreement with the previously reported data for complex C6 and its variants showed in Chart 1. 5c However, we found that even at low proton concentration given by 0.4 µM of water, the C-O bond cleavage triggered via a second protonation of [L N4 Co II -CO2H] + (Figure 5 A) is kinetically more favored (G ‡ 1st CO2= 12.2 kcal·mol -1 ). The subsequent release of a water molecule to form [L N4 Co II -CO] 2+ is entropically driven due to the low concentration of water in organic solution. Likewise, the recovery of the starting [L N4 Co II (S)] 2+ is formed by the CO release from [L N4 Co II -CO] 2+ which would complete the first turnover cycle. The rate determining step of this postulated catalytic cycle is the proton-assisted C-O bond cleavage with a kinetic barrier as low as G ‡ 1st CO2 ~ 12.2 kcal·mol -1 , which is kinetically feasible at room temperature. However, it can be anticipated that, at a higher proton concentration (pH < 24.5), the kinetics will be independent of the protonation events and only governed by the CO2 binding step (G ‡ binding= 8.8 kcal·mol -1 ).
At this point, our modeled 2ereduction mechanism, that catalyzed the CO2 + 2H + reduction to CO + H2O by 1 (II) , is similar to the recently proposed mechanisms for similar systems under both photo-and electrochemical conditions. 5c,29 However, none of the previously reported mechanisms gives an explanation for the general non-catalytic behavior of these systems at the Co II/I wave. Indeed, according to the Co II /Co II -CO mechanism, 1 (II) should catalyze the CO2-to-CO reduction at the thermodynamic Co II/I reduction potential with fast reaction rates due to its low kinetic barrier (G ‡ 1st CO2 ~ 12.2 kcal·mol -1 ). Nonetheless, we have shown that our cobalt complex is not catalytic within the CV timescale (100 mV/s) at the Co II/I wave, and only substoichiometric amounts of CO were accumulated during corresponding electrolysis experiments. Furthermore, we identified the formation of 1 (I) -CO, which is yet to be included as an intermediate in the CO2-to-CO reduction catalyzed by aminopyridine cobalt complexes. 7b In order to account for a model that fits our experimental observations, we considered the further reduction of the cobalt-based intermediates involved in the CO2 reduction mechanism. In this regard, it is remarkable that the 1ereduction of [L N4 Co II -CO] 2+ is highly favored at the Co II/I reduction potential (E1/2(Co II/I -CO) = -0.94 V; G(Co II/I -CO) = -22.3 kcal·mol -1 ). Then, [L N4 Co I -CO] + (1 (I) -CO) becomes the most stable intermediate of the Gibbs energy profile. Indeed, the strong Co-CO bond is responsible for this stability with respect to Co I . The nature of the CO binding and its backbonding character can be illustrated by the frontier molecular orbital analysis in the Co II , Co I and formal Co 0 oxidation states (Figure S32). In the case of Co II -CO, there is not a significant -backdonation from the Co center to the CO ligand, as it is expected for an electron poor metal center. However, regarding Co I -CO and Co 0 -CO, two of the  singly occupied d orbitals of Co I/0 contribute to the -backbonding character of the Co-CO bond as it is shown by the canonic orbitals depicted in Figure 6 B. Moreover, the enhanced stability in [L N4 Co I -CO] + , provided by the presence of a -acceptor ligand, can be explained by means of the 18 ecounting rule. While the CO release from Co II is exergonic (17 e -), the release from Co I (18 e -) is highly endergonic (GCoI-CO > 20.2 kcal·mol -1 ) which blocks the recovery of the active species and prevents catalysis at the Co II/I redox potential. Similarly, the CO release from Co 0 -CO is endergonic by 24.3 kcal·mol -1 . Indeed, the electronic structure of the formal Co 0 -CO is better described as [(L N4 ) • ‾Co I -CO] (18 e -) since the -HOMO orbital is mainly delocalized in the pyridine ring with a small contribution of the metal center.
According to the energetic span model, the overall kinetic barrier of a catalytic process (δEspan) should be calculated as where GTDI, GTDTS and ∆Gr correspond to the Gibbs energies of the TOF Determining Intermediate (TDI), the TOF Determining Transition State (TDTS) and the reaction, respectively. 30 In our case, the TDI corresponds to the [L N4 Co I -CO] + intermediate, and at a working potential of -1.91 V, the TDTS is [L N4 Co II -CO···OH2] 2+ . Then, the energy barrier of the catalytic process is given by the sum of the energy barrier of the first CO2 activation (G ‡ 1st CO2) and the thermodynamics of the CO release (Grelease) from the TDI to recover the active species (i.e. δEspan = G ‡ 1st CO2 + Grelease = 30.3 kcal·mol -1 ). Conversely to G ‡ 1st CO2, δEspan exceeds the kinetic limit for a catalytic process at room temperature. This model is in agreement with the accumulation of 1 (I) -CO at ca. -1.7 V evidenced by thin layer SEC (vide supra).  Alternatively, the reductive disproportionation mechanism has been also computed to explain the formation of 1 (I) -CO (Figure 5 B) in the absence of H + . In this case, after the first CO2 binding, another CO2 molecule can bind to [L N4 Co II -CO2] 2+ to form the thermodynamically downhill [L N4 Co II -(CO2)2] 2+ with a kinetic barrier of 10.9 kcal·mol -1 . The subsequent C-O bond cleavage to obtain a [L N4 Co II -(CO)(CO3)] 2+ is exergonic and proceeds with a barrier of 8.5 kcal·mol -1 . Then, a second Co II molecule can assist the release of carbonate to form [L N4 Co II (O2CO)] and [L N4 Co II -CO] 2+ , which reduction at working potential is strongly exergonic (equation 9). Therefore, the 1 (I) -CO formation through the disproportionation mechanism has a lower Gibbs energy barrier than in the proton assisted mechanism at pH values higher than 25.3.
On the contrary, the energy span for the reductive disproportionation mechanism (G ‡ 2nd CO2 + Grelease + GCO3 = 69.2 kcal·mol -1 ) is by far higher than in the proton assisted mechanism (G ‡ 1st CO2 + Grelease = 30.3 kcal·mol -1 ) due to the additional stability of the Co IIcarbonate species.
These results clearly show that the formation of 1 (I) -CO is both thermodynamically and kinetically favored. The high stability of 1 (I) -CO and the partial sequestration of the starting Co II in the form of cobalt carbonate kinetically prevents the catalytic CO2 reduction at the Co II/I redox potential, in agreement with the detection of cobalt carbonate species in solution after electrolysis. Both theoretical and experimental results highlight the complexity of the cobalt catalyzed CO2 reduction mechanism. As it has been shown, δEspan strongly depends on the stability of 1 (I) -CO but also on redox and protonation events which are controlled by the applied redox potential and the pH of the medium, respectively. Indeed, the variation of these two factors can switch the operative mechanism for the formation of 1 (I) -CO from a pH-independent reductive disproportionation mechanism to a proton assisted CO2 reduction mechanism.
Catalytic CO2 reduction. According to cyclic voltammetry, further reduction to formal Co 0 intermediates is needed in order to activate the catalytic process. Moreover, we have evidence to support that catalysis is assisted by the presence of H + in solution since the catalytic wave increases in current when H2O is added to the solution and it shifts to more positive potentials. As above shown, the catalytic wave it is not affected by the presence of added carbonate, and then it can be excluded from the mechanism. These experimental evidence, together with the previous DFT study, led us to hypothesize a reaction mechanism in which a second CO2 binding occurs after the 1ereduction of [L N4 Co I -CO] +  Since the mechanistic scenario is complex and the reaction mechanism is redox-and pH-dependent, we have evaluated how the thermodynamics (Gr) and kinetics (δEspan) of the catalytic process are modified in terms of both the redox potential and pH. Although this type of analysis has its precedents in heterogeneous catalysis, it is uncommon in the study of molecular systems. 31 To this end, we have developed a software tool that identifies the TDI and TDTS and calculates δEspan and Gr of a given reaction mechanism as a function of different parameters such as the redox potential and concentration of the reactant species including [H + ]. 32 The summary of the resulting analysis is shown in Figure 7. . A) Proposed reaction mechanism for the CO2 reduction to CO catalyzed by 1 (I) -CO. B) Contour-plots of the kinetic energy span δEspan (color scale, kcal·mol -1 ) of the calculated reaction mechanism versus the applied reduction potential (horizontal axis, V vs. Fc/Fc + ) and the pH (vertical axis). Regions A-F are delimited by dashed red lines and the dashed black line spares the thermodynamic (Gr < 0 kcal·mol -1 ) regime from the non-thermodynamic one (Gr > 0 kcal·mol -1 ). C) Gibbs energy profiles associated with Regions A and B.
In order to illustrate the variation of the reaction kinetics, the δEspan value is represented by a color scale in a contour plot, where the vertical and horizontal axes correspond to the pH and redox potential (E vs. Fc +/0 ), respectively. In the resulting 2D map, we can identify regions where the kinetics either depend on the pH (A), on the redox potential (B), or on both pH and redox potential (C-F). The black dashed line in the 2D map represents the pH and redox potential values with Gr = 0 for the catalytic cycle. For simplicity, we will focus on regions A, B and C, as they are the ones where the reaction mechanism is thermodynamically favored (Gr < 0 kcal·mol -1 , Figure S37). The Gibbs energy profiles of regions E-F are given in Figure S38.
Region A corresponds to a regime where the redox potential is more negative than -2.3 V and the concentration of protons is very low. According to the corresponding energy profile (Figure 7 C), the 1ereduction of [L N4 Co I -CO] + to form [L N4 Co 0 -CO] presents a non-negligible kinetic barrier. However, the formal Co 0 species is reactive enough to bind CO2 forming [L N4 Co II -CO2(CO)] via a barrierless reaction with CO2, unlike in the case of Co I where the CO2 binding is endergonic and with a barrier of 8.8 kcal·mol -1 (Figure 5). The following protonation of [L N4 Co II -CO2(CO)] to form L N4 Co II -CO2H(CO)] + becomes the rate-determining step of the reaction. For instance, at pH 35 and E < -2.35 V the δEspan is 20.0 kcal·mol -1 . As anticipated above, further reduction and protonation yields [L N4 Co I -(CO)2] + together with the extrusion of a water molecule. Finally, CO is thermodynamically favorable released to recover the key [L N4 Co I -CO] + intermediate.
As it can be inferred from the 2D plot, the increase in the proton concentration will drive the reaction to region B. Once in region B, the reaction rate is given by the Co I/0 electron transfer process, which becomes the rate-determining step of the reaction instead of the protonation steps. In fact, we have included the Marcus electron transfer barrier to better describe the reaction kinetics of this region. Then, as δEspan solely involves an electron transfer, the reaction rate only depends on the reduction potential. The subsequent CO2 binding, protonation and reduction steps are thermodynamically favored, and the overall energy profile becomes downhill in Gibbs energy. For instance, at -2.35 V and pH < 33 the δEspan is 18.0 kcal·mol -1 . Finally, region C is in between A and B because the kinetic barrier depends on both the Co I/0 electron transfer kinetics and also on the protonation of the carboxylate adduct [L N4 Co II -CO2(CO)] ( Figure S38). In summary, our model allows for the rationalization of the experimental observations. First, it describes a regime where the catalytic reaction is kinetically unfavorable at low overpotentials and high pH values. This data is also in agreement with the lack of catalytic current at the Co II/I redox potential, even upon addition of water to the reaction media, and with the detection of Co I carbonyl species. In addition, our model gives an explanation of the peak shift and current increase measured by CV at the Co I/0 redox potential in the presence of water (vide supra).
The mechanistic proposal for the CO2 reduction at the Co II/I redox wave suggests that catalysis could be activated by avoiding the 1 (II/I) -CO reduction. However, we noticed that this 1ereduction is much more favored than the Co II/I process which it makes difficult to avoid the formation of 1 (I) -CO under electrochemical conditions. Another beneficial strategy to promote catalysis could be based on the metal carbonyl labilization. In this regard, photocatalysis can facilitate both. On the one hand, it is well-known that light induces the M-CO bond cleavage in organometallic carbonyl species. On the other hand, photocatalysis can operate at very low concentrations. For bimolecular catalysis/photosensitizer reactions it implies that the electron transfer rate is under diffusion control. Therefore, at low enough catalyst concentration, the 1 (II/I) -CO reduction rate could be lower than the CO release allowing the Co II /Co II -CO mechanism.
Catalysis and the effect of light irradiation. With the aim of testing our hypothesis, we designed the following experiments to promote catalysis at the 1 (II/I) redox couple via the Co II /Co II -CO mechanism.
We studied 1 (II) as a homogeneous catalyst for the light-driven CO2 reduction in combination with two different cyclometalated Ir photosensitizers. The typically used [Ir III (ppy)3] (PSIr1) with an E1/2(PSIr1 0/-) redox potential of -2.67 V, low enough to promote the reduction of 1 (I/0) -CO and [Ir III (ppy)2(bpy)](PF6) (PSIr2), with a E1/2(PSIr2 +/0 ) of -1.78 V at which the formation of 1 (0) -CO is not accessible (Figure 8). Experiments were performed with 1 (II) (50 µM) and the photosensitizer (200 µM) in CO2 saturated CH3CN:Et3N mixed (4:1 v/v) irradiated at 447±20 nm for 24 h at 25 °C. Gases evolved were quantified by GC, with CO and H2 as the only detected products (Figures 9, S39). Remarkably, although PSIr2 provides a redox potential 820 mV less negative than PSIr1, both photosensitizers result in a similar reaction rate and catalytic activity (TON of CO 69±2 and 68±3 for PSIr1 and PSIr2, respectively). These data confirmed that the in-situ generated Co I species is able to promote a selective conversion of CO2-to-CO as anticipated from the electrochemical and computational studies.
On the other hand, in an attempt to avoid the CO-poisoning process under electrochemical conditions, we also performed electrolysis experiments under blue light irradiation. Previous studies by T. C. Lau, M. Robert and co. suggested that light irradiation could indeed facilitate the CO release in the case of the [Fe I (qpy)CO] + adduct over the reduction to Fe 0 carbonyl species. 8a For these set of experiments, we carefully controlled the reaction temperature (25 ºC) with a jacketed electrochemical cell connected to a cryostat. CV of 1 (II) under blue LED light (447±20 nm) in CO2-saturated solution showed the disappearance of the reoxidation peak at -0.8 V ( Figure  S40). 10 This feature is reproducible upon successive switch on/off cycles.   When a constant Eappl potential of -2.46 V is held for 6 h under irradiation a substantial improvement of the catalytic activity of 1 (II) is observed (TONCO = 13, FYCO = 38%) with respect to the performance in dark (TONCO = 5.5, FYCO = 26%), in terms of both catalytic turnovers and faradaic yield for CO production ( Figure S41). Prolonged electrolysis highlights a sustained electrocatalytic current, leading to almost 20 turnovers of CO after more than 10 h and maintaining the same average efficiency. This is consistent with a beneficial effect of blue-light photoirradiation on catalysis, consisting of a light-induced cleavage of the accumulated stable Co-CO species in solution, thus favoring a partial regeneration of the catalyst. On the other hand, the effect of irradiation is barely observed during light-assisted electrolysis at -1.70 V under CO2 atmosphere, suggesting a smaller effect of light absorption on the 1 (I) -CO species ( Figures S43, S44 and Table S16).

CONCLUSIONS
We have presented a detailed mechanistic investigation of electrochemical CO2-to-CO reduction catalyzed by a new cobalt catalyst for CO2-to-CO reduction (1 (II) ) based on a highly basic tetradentate aminopyridyl ligand. To the best of our knowledge, FTIR-SEC provides the first in-situ spectroscopic evidence for the formation of a Co I -CO (1 (I) -CO, νCO = 1910 cm -1 ) resulting from the electrochemical CO2-to-CO reduction at the non-catalytic Co II/I redox wave under CO2. This observation has relevant mechanistic implications since it shows that: 1) the electrochemically generated Co I species (1 (I) ) is nucleophilic enough to bind the CO2 molecule and 2) the C-O bond cleavage can occur at room temperature, at mild applied potentials and with no added protons in acetonitrile. DFT modeling of the reaction mechanism has corroborated that both the CO2 binding and the C-O bond cleavage steps are kinetically feasible at the Co II/I redox potential. However, the CO release from 1 (I) -CO is a key limiting step which prevents the recovery of the catalytically active species 1 (I) . Computational modeling of the different catalytic mechanisms in a broad potential and pH windows allowed for the rationalization of our experimental observations. The catalytic mechanism is triggered by the one-electron reduction of 1 (I) -CO to the corresponding formal Co 0 which can only be afforded close to the Co I/0 redox potential. Photocatalytic experiments under blue-light irradiation confirm the ability of 1 (I) towards catalytic CO2 reduction, even when the E1/2 of the PSIr is not suitable for the 1 (I/0) -CO reduction. It is proposed that under photocatalytic contitions the CO release from 1 (II) -CO is kinetically favored over the 1 (I) -CO reduction due to the low concentration of catalyst and photosensitizers.
Finally, light-assisted electrocatalysis was successfully employed to improve the catalytic performance of 1 (II) , favoring the activation of inactive carbonyl species and reaching higher stability and efficiency for CO production. In view of these findings, light-induced metal carbonyl dissociation was revealed as a promising strategy to mitigate CO catalyst poisoning. Finally, we have proposed a unified mechanistic view of the existing differences between photo-and electrochemical CO2-to-CO reduction catalysis ( Figure 4). The results presented here will help to rationalize the behavior of other reported cobalt-based molecular electrocatalysts and to find out new approaches for the optimization of earth-abundant molecular catalysts.
CERCA Programme and DIUE 2014SGR931 (Generalitat de Catalunya) for financial support and MINECO project CTQ2016-80038-R and PGC2018-098212-B-C22. We acknowledge SOLEIL and DIAMOND for provision of synchrotron radiation facilities and we would like to thank Dr. Gautier Landrot for assistance in using beamline SAMBA.