Electrocatalytic Reduction of CO2 to CO by Molecular Cobalt–Polypyridine Diamine Complexes

Cobalt complexes have previously been reported to exhibit high faradaic efficiency in reducing CO2 to CO. Herein, we synthesized capsule-like cobalt–polypyridine diamine complexes [Co(L1)](BF4)2 (1) and [Co(L2) (CH3CN)](BF4)2 (2) as catalysts for the electrocatalytic reduction of CO2. Under catalytic conditions, complexes 1 and 2 demonstrated the electrocatalytic reduction of CO2 to CO in the presence or absence of CH3OH as a proton source. Experimental and computational studies revealed that complexes 1 and 2 undergo two consecutive reversible one-electron reductions on the cobalt core, followed by the addition of CO2 to form a metallocarboxylate intermediate [CoII(L)–CO22−]0. This crucial reaction intermediate, which governs the catalytic cycle, was successfully detected using high resolution mass spectrometry (HRMS). In situ Fourier-transform infrared spectrometer (FTIR) analysis showed that methanol can enhance the rate of carbon–oxygen bond cleavage of the metallocarboxylate intermediate. DFT studies on [CoII(L)–CO22−]0 have suggested that the doubly reduced species attacks CO2 on the C atom through the dz2 orbital, while the interaction with CO2 is further stabilized by the π interaction between the metal dxz or dxz orbital with p orbitals on the O atoms. Further reductions generate a metal carbonyl intermediate [CoI(L)–CO]+, which ultimately releases CO.

Scheme 1.The structures of the target cobalt complexes 1 and 2.

Synthesis, Characterization, and X-ray Structures
The mononuclear cobalt complex [Co(L 1 )](BF 4 ) 2 (1) was conveniently prepared by reacting Co(BF 4 ) 2 •6H 2 O with L 1 in acetonitrile at room temperature under ambient air conditions (see the experiment section).A mauve crystalline solid of complex 1 was obtained with a yield of 85%.Complex 2 was synthesized and purified using the same method, yielding 82%.The high-resolution mass spectra (HRMS) and elemental analysis data of complexes 1 and 2 are consistent with the proposed composition of [Co(L 1 )](BF 4 ) 2 and [Co(L 2 )(CH 3 CN)](BF 4 ) 2 .
The molecular structures of complexes 1 and 2 were determined by single crystal X-ray crystallography (Figure 1).Complex 1 adopts a distorted octahedral geometry.The Co-N bond distances for two axial pyridyl-nitrogen atoms (Co-N5 (N6) = 2.292 (2.326) Å) are significantly longer than that of the other four nitrogen atoms on the equatorial plane (Co-N1 (N2, N3, N4) = 2.193 (2.164, 2.147, 2.186) Å).Two pyridine rings on the equatorial plane in complex 1 form an angle of extrusion of 34 • .By replacing one of the 6-methylpyridinyls in 1 with phenyl, complex 2 features one CH 3 CN solvent molecule coordinated in the equatorial site.Complex 2 shows a normal octahedral geometry, however, its bond lengths and angles are very similar to those in complex 1. Detailed crystallographic data and processing parameters are collected in Tables S1 and S2.

Electrochemistry
Cyclic voltammograms (CVs) of complex 1 under Ar in CH 3 CN solution feature two reversible couples at −1.66 and −1.89 V (all potentials given in this paper are versus ferrocene (Fc +/0 )) (Figure 2, black line).The free ligand L 1 is a silent electrochemical in the potential range (Figure S1).Additionally, the absence of any redox processes for a Cu analogue [Cu(L 1 )][BF 4 ] 2 between the potentials of −1.66 to −1.89 V (Figure S2) confirms that the ligand is not responsible for the redox events observed for complex 1.The results reveal that the two couples can be assigned to Co II/I and Co I/0 , respectively.The CV of complex 2 (Figure S3) shows almost the same redox couples as 1.Under the same test conditions, linear sweep voltammograms (LSV) of both complexes 1 and 2 also performed the same redox couples as the reduction waves in the CVs (Figures S4 and S5).Both the first and second couples of complexes 1 and 2 show a good linear relationship between the peak current and squared root of the scan rate, but a difference in slope, indicating that complexes 1 and 2 are functioning in a diffusion-controlled regime and are molecular in nature (Figures S6 and S7).As the scan rate decreases, the ratio of the second peak current to the first increases.Differential pulse voltammograms (DPVs) of complexes 1 and 2 both exhibit two reduction peaks around −1.64 and −1.85 V (Figure S8), which are in accordance with the reduction waves in CVs.Differences in amplitude between Co II/I and Co I/0 for both complexes 1 and 2 in the CVs and DPVs may have resulted from minor changes in configuration following the first electron transfer process of Co II/I .Under CO 2 , complex 1 exhibits a slight current enhancement around −1.85 V, which is near the potential for Co I/0 couple (Figure 2, red line; Figure S4).A controlled potential electrolysis (CPE) experiment of complex 1 was performed at −1.9 V under CO 2 .Thereafter, an analysis of the gas mixture, in the headspace of the working compartment of the electrolysis cell, by gas chromatography (GC) identified CO as the exclusive gaseous product under the test condition (Figure S9).The catalytic peak current (i cat ) varies linearly with the concentration of complex 1, consistent with a mechanism for catalytic CO 2 reduction that is a first-order reaction in complex 1 (Figure S10).Under CO 2 in the presence of 4 M methanol, the current exhibits a 4.8-fold increase under CO 2 conditions in the absence of methanol at −2.1 V (Figure 2 and Figure S11).As documented in the literature, this phenomenon is ascribed to methanol's role as a proton source, leading to the modification of the catalytic reaction pathway, thereby enhancing the rate of carbon-oxygen bond cleavage of the metallocarboxylate intermediate [4,6,41] (See details in the "The Analysis of Process and Intermediates").In addition, the catalytic peak current has linear dependence on methanol concentration, ranging from 0.15 to 0.98 M, and reaching saturation at around 4 M (Figure S11).Changing the sequence of bubbling with CO 2 and adding methanol results in negligible differences in CVs.In contrast, the addition of methanol results in a 260 mV negative shift of the second couple under Ar (Figure 2 and Figure S12).This observation indicates that CO 2 is much more competitive than methanol to coordinate with the reduced catalyst.Under the same test conditions, complex 2 performs an almost identical phenomenon with complex 1 (Figures S5 and S13-S17).

The Catalytic Rate and Stability of Cobalt Complexes
Under slower scan rates (0.01 to 0.075 V s −1 ) with 4 M MeOH, the CVs of complexes 1 and 2 show a steady "S-shaped" wave (Figures S18 and S19).Higher scan rates lead to the combination of catalytic current plateaus, followed by subsequent increases in current.Using methods described by Savéant [42,43], the turnover frequency (TOF) values for complexes 1 and 2 are estimated to be 2.03 and 1.57 s −1 , respectively, under the scan rate at 0.05 V s −1 , with an overpotential around 550 mV (Figures S18 and S19).
A CPE experiment of complex 1 at −2.1 V under CO 2 shows the CO concentration increases as time goes on (Figure S20 inset).The gas chromatography (GC) analysis confirms that the measured CO production matches well with the theoretical yield (Figure S20), indicating that the faradaic efficiency is close to 100%.However, decomposition of the catalyst under highly reductive conditions results in the CO production deviating from its theoretical yield after 3 h (Figures S20-S22).An average turn-over number (TON) of 9.5 was achieved by complex 1. Between the changes in complex 1 and 2, an insignificant difference in catalytic activity and stability was found (Figures S23-S25 and Table S4).
In the presence of methanol under CO 2 atmosphere, an 80 min CPE experiment of complex 1 performed at −2.1 V. GC analysis shows that CO and H 2 were the main products.The faradaic efficiency of CO is approximately 30% within 40 min and shows a gradually decreasing trend, dropping to 7% at 80 min.The faradaic efficiency of H 2 gradually increased, reaching 85% at 80 min (Figure S26).Additionally, obvious deposits were observed on the surface of glass carbon electrode after electrolysis for 80 min (Figure S27).This phenomenon is commonly observed in published papers for molecular catalysts, where the predominant component of deposition is a zero-valent metal, accompanied by a minor presence of metal oxide [31, [44][45][46].The chemical state of the deposits on the surface of glass carbon electrode were investigated by XPS.As shown in Figure S28, the peak with binding energies at 778.5 eV are associated with the zero-valent cobalt, and the peak of 779.5 and 781.1 eV binding energies stem from the cobalt in the oxidation state.The area of the three peaks indicates that the zero-valent cobalt is the predominant component in our experimental deposition.The obtained results align with the findings reported in published papers on cobalt molecular catalysts [31].The deposited glass carbon electrode, acting as the working electrode, was used at −2.1 V in the presence of methanol under CO 2 atmosphere without complex 1 or 2. Thereafter, GC analysis shows that H 2 was the sole product.The results reveal that deposits are incapable of catalyzing the reduction of CO 2 ; their catalytic activity is limited to the reduction of protons for hydrogen gas production.The actual catalyst responsible for the conversion of CO 2 into CO is complexes 1 or 2 in CH 3 CN solution.

The Process and Intermediates Analysis
The initial step in the reduction of CO 2 molecules involves its binding.As shown in Figure 2, in the presence of CO 2 , [Co 0 (L 1 )] 0 is sensitive to CO 2 such that slightly more negative scanning following Co II/I would result in the irreversibility of the Co I/0 and Co II/I couple (Figure S29).In the DPV test of complex 1 under CO 2 , the first peak shows no difference with that under Ar, but the second peak expands to a plateau (Figure S30), indicating that CO 2 binding and subsequent reduction coincide with the Co I/0 couple.In the presence of 4 M methanol, the plateau splits into two stronger peaks.The more positive one could be ascribed to the formation of a CO 2 adduct [Co II (L 1 )-COOH − ] + .The other corresponds to the following reduction.Additionally, the reduction of complex 1 is restricted to the [Co I (L 1 )] + state by mixing complex 1 with 1 equiv of bis-pentamethylcyclopentadienylcobalt(II) (Cp* 2 Co, the oxidation potential of Cp* 2 Co is −1.82V vs. Fc +/0 in CH 3 CN).Then, the addition of dry CO 2 brings no new absorption band attributed to carbonyl.Herein, it can be concluded that [Co 0 (L 1 )] 0 is the active state for CO 2 rather than [Co I (L 1 )] + , and two-electron reduction are necessary to activate complex 1.
[Co 0 (L 1 )] 0 as a dark blue solution in a mixture of CH 3 CN/THF (1:1) can be easily prepared by chemical reduction of complex 1 with 2 equiv of NaHg.When the solution is bubbled with dry CO 2 gas, the color changes back to mauve in 30 s (Figure S31).As we know, the blue color is attributed to the electron transfer between d orbitals of cobalt.The significant change in color indicates two electrons are transferring from cobalt to CO 2 .Then, we monitor the process of CO 2 coordinating with [Co 0 (L 1 )] 0 by in situ FTIR (Figure 3 and Figure S32).In the range of 1200 to 2500 cm −1 , [Co 0 (L 1 )] 0 shows a slight IR absorption.When dry CO 2 is added, a weak broad absorption appears at 1669 cm −1 , indicating that CO 2 coordinates with [Co 0 (L 1 )] 0 to form the intermediate [Co II (L 1 )-CO 2 2− ] (Figure 3).The frequency is close to that reported for Co(salen)CO 2 Na (1680 cm −1 ) [47].Then, upon addition of methanol, three new bands appeared at 1634, 1481, and 1390 cm −1 (Figure 3 and Figure S33).The assignment was confirmed by isotope labelling experiment.When dry 13 CO 2 is introduced, followed by the addition of methanol, these new bands shift toward lower wavenumbers by 30 to 60 cm −1 (Figure 3).We conclude that methanol disrupts the balance of electron distribution on the two oxygen atoms of CO 2 , resulting ultimately in coordinated transformation into the metallocarboxylate intermediate (Figure 3).Therefore, in this catalysis system, methanol could lead to the CO 2 binding equilibrium moving toward the forward reaction.The fresh reaction solution is then analyzed by high resolution mass spectrometry (HRMS) [48].The isotope distribution pattern at m/z − = 584.2305,585.2333 and 586.2363 fit well with the calculated isotope distribution pattern of the mass fragment [Co II (L 1 )-COOH − ] + .The results reveal that we successfully detected the protonated product of [Co II (L 1 )-CO 22− ] 0 (Figure 4 and Figure S34).In the same way, the protonated product of [Co II (L 2 )-CO 22− ] 0 (Figures S35 and S36), prepared from the analogue complex 2, was also detected by HRMS and it was found that the acetonitrile ligand had dissociated from cobalt.These results provide clear evidence for the incorporation of CO 2 into the [Co 0 (L 1 )] 0 intermediate.As referred to in the DPV test of complex 1 under CO 2 , CO 2 binding and the following reduction coincide with the Co I/0 couple (Figure S30).To investigate the subsequent reduction of the CO 2 -coordinated cobalt metallocarboxylate, in situ infrared spectroelectrochemistry (IR-SEC) experiments are performed at −1.85 V.The differential IR-SEC spectrum of 1 reveals a decrease in CO 2 intensity around 2340 cm −1 and the emergence of a band at 1917 cm −1 (Figure 5).Complex 1 itself does not possess IR absorption modes with sufficient intensity that can be distinguished within the range of 1500 to 2500 cm −1 , under varied electrolysis potential from −1.0 to −2.4 V, and in the absence of CO 2 .The absorption peak observed at 1917 cm −1 can be ascribed to the twoelectron reduction species [Co I (L 1 )-CO] + [26,49], which is obtained from through further reduction of [Co II (L 1 )-CO 22− ] 0 .Fujita and colleagues have reported that [Co I (macrocycle)-CO] + (macrocycle = 5,7,7,12,14,14-hexamethyl-1,4,8,11-tetraazacyclotetradeca-4,11-diene) exhibits an absorption peak at 1916 cm −1 [49,50].Upon reduction of complex 1 with 1 equiv of Cp* 2 Co under CO, a weak IR absorption band resembling [Co I (macrocycle)-CO] + was observed at 1917 cm −1 (Figure S37).In the presence of 1 M methanol under the same experimental condition, two strong peaks at 1916 and 1664 cm −1 were ascribed to [Co I (L 1 )-CO] + and the combination absorption of [Co II (L 1 )-COOH − ] + , respectively (Figure S38).The differential IR-SEC spectrum of complex 2 was also recorded at −1.85 V. Similar to complex 1, the peak of the reduction species [Co I (L 2 )-CO] + was also observed at 1917 cm −1 .The findings suggest that complex 2 undergoes the same catalytic mechanism as complex 1.
(Figure S39).Consequently, there will be further reductions of CO 2 -coordinated cobalt intermediates to generate CO.

Density Functional Theory (DFT) Calculation and Mechanistic Consideration
To get a better understanding of the electronic distribution on the reduced catalyst [Co 0 (L 1 )] 0 , we employed DFT calculations using Gaussian 09, Revision D.01.Firstly, the geometry-optimized initial structure of [Co I (L 1 )] + shows that both pyridine groups in the axial direction dissociate from cobalt (Figure S40a).The dissociation is consistent with the longer C-N distance in the axial direction.The dissociation results in a minor change in configuration, which leads to the difference in current amplitude for the two reduction peaks in complex 1 (Figure 2).As proven above, CO 2 coordination is coupled with the second reduction.After which, the CO 2 molecule is joined to the optimized structure [Co I (L 1 )] + at the axial open site.Meanwhile, the second electron is added.The geometryoptimized electronic structure of [Co II (L 1 )-CO 2 2− ] 0 shows the electron density distribution in the frontiers (Figure S40b).The optimized structure provides evidence supporting the hypothesis that CO 2 can undergo an attack by the d z 2 orbital of the doubly reduced cobalt from the axial direction.The formation of a σ bond between the d z 2 orbital and the carbon atom of CO 2 (HOMO) is observed, and this interaction with CO 2 can be further stabilized through π interactions between the metal's d xz or d yz orbitals and p orbitals on the oxygen atoms of CO 2 (HOMO-5, Figure S41).The results of the Kubiak and Fujita groups are also similar in that the stabilizing interaction is with oxygen atoms of CO 2 rather than the carbon atom [51].This tridentate interaction makes the coordination of CO 2 to [Co 0 (L 1 )] 0 irreversible.Furthermore, the attack on the oxygen atom by Brönsted acid becomes more favorable.The optimized [Co II (L 1 )-CO 2 2− ] 0 shows that one dissociated pyridine in [Co I (L 1 )] + state stabilizes the CO 2 adduct from the axial direction (Figure S38).Compared with 1, 2 features one labile CH 3 CN ligand.The reductive dissociation of the labile ligand acetonitrile from the polypyridine-cobalt complexes at the [Co I ] state has been previously verified by other groups.By replacing acetonitrile with CO 2 , while two electrons are also added, the geometry-optimized electronic structure of [Co II (L 2 )-CO 2 2− ] 0 gives rise to almost the same electronic structure as [Co II (L 1 )-CO 2 2− ] 0 (Figures S42 and S43).Thus, we conclude complex 2 to be a simpler model for CO 2 reduction.
Taking complex 1 as an example, Scheme 2 summarizes a proposed mechanism for the catalytic reduction of CO 2 to CO by complexes 1 and 2. Initially, a two-electron reduction is necessary to generate the active complex [Co 0 (L 1 )] 0 (I).As [Co 0 (L 1 )] 0 is so sensitive, CO 2 is coordinated to generate [Co II (L 1 )-CO 2 2− ] 0 (IIa).Subsequently, the carbonoxygen bond cleavage have two paths:

Materials and Instruments
All manipulations for the preparation and handling of the organometallic complexes were carried out under air.All the solvents were used as received.Other commercially available chemicals, such as Co(BF 4 ) 2 •6H 2 O, 1,2-ethanediamine,6-methylpicolinaldehyde, N-benzylethane-1,2-diamine, were purchased from local suppliers and used as received.Water was deionized with the Millipore Milli-Q UF Plus system.Glass carbon discs (3 mm), AgCl/Ag electrodes and platinum wire were purchased from CHI for electrochemical studies.Arrows are employed in the CV diagram to denote the initial scanning direction.
The NMR spectrum was acquired by employing the Varian INOVA 600 NMR spectrometer, an instrument manufactured by the renowned American company Varian.Mass spectra were recorded with HP 1100 HPL/ESI-DAD-MS and Waters/Micromass LC/Q-TOF-MS instruments.Elemental analyses were performed with a Thermoquest-Flash EA 1112 elemental analyzer.FTIR in situ was monitored by React IR 15 instrument equipped with a MCT detector and a Dsub AgXSiComp in situ probe.

Synthesis
Synthesis of ligands L 1 and L 2 .L 1 and L 2 were prepared according to the procedures in the literature [52,53] Preparation of complexes 1 and 2. Compound Co(BF 4 ) 2 •6H 2 O (1.1 equiv) and L 1 (1.0 equiv) or L 2 (1.0 equiv) were added to the CH 3 CN solution (5 mL).The mixture was stirred at room temperature for 12 h.The mauve solution was then transferred to tubes, which were placed in the flask containing ether, and needle-shaped crystals were obtained.
Complex In situ preparation of [Co 0 (L 1 )] 0 .Complex 1 was mixed with 2.0 equiv of Na(Hg) in a highly pure THF and CH 3 CN (1:1) mixture under Ar.The sample was stirred under Ar overnight.A dark blue solvent appeared after stirring stopped.The freshly prepared [Co 0 (L 1 )] 0 was used in the next step without further handling.
In situ preparation of [Co II (L 1 )-CO 2 2− ] 0 .The reaction was monitored by a React IR 15 instrument equipped with a MCT detector and a Dsub AgXSiComp in situ probe.The experiment was performed at room temperature.[Co II (L 1 )-CO 2 2− ] 0 was prepared by bubbling dry CO 2 into the solution of complex [Co 0 (L 1 )] 0 .In about 30 s, the dark blue color changed back to pink.After which, 2 M of CH 3 OH was added into the solvent.The fresh reaction solution was tested directly by a mass spectrum.

IR-SEC Test Details
The spectroelectrochemical experiments were conducted using an IFS 66v/S FTIR spectrometer (Bruker) equipped with an MCT-A detector.Electrode potential was controlled using a model CHI potentiostat.The SEC experiments were performed in a homemade three-neck cell with a glassy carbon electrode serving as the working electrode, while a platinum foil acted as the counter electrode, and Ag/AgCl (3 M KCl) as the reference electrode.IR-SEC spectra were collected in a single beam mode at either 2 cm −1 or 4 cm −1 , and differential absorption spectra were presented relative to the reference spectrum and were recorded immediately prior to applying the potential.

Computational Methods
All DFT computations were calculated at the B3LYP (the Becke three-parameter Lee-Yang-Parr hybrid functional) level using the 6-31G(d) basis set for all atoms via the Gaussian 09 program package [54,55].All structures were fully optimized and confirmed as minima by analytical vibrational frequency calculations with B3LYP/6-31G(d).The solvent acetonitrile was modeled as a dielectric continuum using a polarizable continuum model (PCM).

Figure 1 .
Figure 1.Molecular structures of complexes 1 and 2. Counter ions and hydrogen atoms are omitted for clarity.

Figure 2 .
Figure 2. Cyclic voltammograms of complex 1 in CH 3 CN with 0.1 M nBu 4 NPF 6 as the supporting electrolyte (I) under Ar; under CO 2 (II) in the absence of and (III) in the presence of 4 M methanol.The arrow in the CV diagram represents the initial scan direction.

Figure 3 .
Figure 3. IR spectrum of freshly prepared [Co 0 (L 1 )] 0 (black line), which is bubbled with dry CO 2 (above) and 13 CO 2 (down) (red line), then 2 M methanol is added (blue line).These lines were recorded by in situ FTIR during the reaction process.The tests are performed in a CH 3 CN/THF (1:1) mixture under Ar or CO 2 with 10.0 mM [Co 0 (L 1 )] 0 .
Scheme 2. A proposed mechanism for electroreduction of CO 2 to CO catalyzed by complex 1.