An amide-based second coordination sphere promotes the dimer pathway of Mn-catalyzed CO2-to-CO reduction at low overpotential

The [fac-Mn(bpy)(CO)3Br] complex is capable of catalyzing the electrochemical reduction of CO2 to CO with high selectivity, moderate activity and large overpotential. Several attempts have been made to lower the overpotential and to enhance the catalytic activity of this complex by manipulating the second-coordination sphere of manganese and using relatively stronger acids to promote the protonation-first pathway. We report herein that the complex [fac-Mn(bpy-CONHMe)(CO)3(MeCN)]+ ([1-MeCN]+; bpy-CONHMe = N-methyl-(2,2′-bipyridine)-6-carboxamide) as a pre-catalyst could catalyze the electrochemical reduction of CO2 to CO with low overpotential and high activity and selectivity. Combined experimental and computational studies reveal that the amide NH group not only decreases the overpotential of the Mn catalyst by promoting the dimer and protonation-first pathways in the presence of H2O but also enhances the CO2 electroreduction activity by facilitating C–OH bond cleavage, making [1-MeCN]+ an efficient CO2 reduction pre-catalyst at low overpotential.

In 2012, Savéant, Costentin and their co-workers reported an Fe porphyrin catalyst with phenolic groups which as a local proton source can enhance the performance of CO 2 electroreduction, 21 demonstrating the strong effect of the second coordination sphere on CO 2 reduction. Aer that, several research groups have used this strategy to improve CO 2 reduction activity and designed a few family of Mn/Re based CO 2 reduction catalysts with functional groups, such as thiourea, 18 imidazolium, 11,17 phenol, [38][39][40] and ether 41 moieties as either a hydrogen bond donor or acceptor, resulting in enhanced CO 2 reduction performances. On the other hand, for earth-abundant Mn-based CO 2 reduction catalysts, it is difficult to achieve high catalytic activity at low onset potential towards CO 2 reduction without using strong acids. 6,9,41,42 Very recently, Nippe, Panetier and their co-workers reported a family of imidazoliumfunctionalized fac-Mn(CO) 3 bipyridine catalysts, which showed moderate catalytic activity at mild potentials (ca. 1.1 mA cm À2 at À1.55 V vs. Fc +/0 ; all reduction potentials are reported vs. Fc +/0 ) using H 2 O as the proton source. 11 For Mn-based CO 2 reduction catalysts, there are three proposed catalytic pathways for CO production: (i) the reduction-rst pathway, (ii) the protonation-rst pathway and (iii) the dimer pathway. The reduction-rst pathway oen occurs when weak acids are used as the proton source, while the protonation-rst pathway generally requires the introduction of stronger acids. 12 The dimer pathway has been studied by Chardon-Noblat 43 and Cowan 44 groups. Their studies indicate that the fac-Mn(CO) 3 bipyridine catalyst could catalyse the CO 2 -to-CO conversion via the dimer pathway at a low onset potential but with low catalytic activity. Among these three reaction pathways, the dimer pathway so far has showed the lowest onset potential but on the other hand the catalyst design that could promote this pathway is mostly unexplored. Herein, we tailored the bpy ligand by installing amide groups and prepared [fac-Mn(bpy-CONHMe)(CO) 3 Br] ([1-Br]; bpy-CONHMe ¼ N-methyl-(2,2 0 -bipyridine)-6-carboxamide; Scheme 1) with an amide -NHMe group with hydrogen bonding and proton donor capability in the second coordination sphere and [fac-Mn(bpy-CONMe 2 )(CO) 3 Br] ( ; bpy-CONMe 2 ¼ N,N 0dimethyl-(2,2 0 -bipyridine)-6-carboxamide; Scheme 1), which lacks the amide proton. Under catalytic conditions, complexes  and  are respectively converted to   +

and [2-
MeCN] + . The amide -NHMe group of [1-MeCN] + not only decreases the overpotential requirement of the Mn catalyst by promoting the dimer and protonation-rst pathways in the presence of H 2 O but also enhances the CO 2 electroreduction activity by facilitating the rate-limiting C-OH bond cleavage. To our knowledge, this is the most active Mn-bpy catalyst utilizing the dimer pathway at mild potentials (as positive as À1.50 V) with H 2 O as the proton source.

Synthesis and characterization
Complexes ,  and [fac-Mn(bpy)(CO) 3 Br] ([3-Br]) (Scheme 1) were prepared by the reaction between the corresponding ligands and [Mn(CO) 5 Br] according to previously reported procedures (see Fig. S1 and S2 in the ESI †) 7 and were fully characterized by 1  As shown in Fig. 1 (le), the X-ray crystal structure of the complex  reveals that it crystallized in the space group P2(1)/c and the geometry is facial octahedral. The second coordination sphere of the amide -NHMe group is out of the bpy plane, while being close to the active site (the Br position), with a dihedral angle N3-C12-C11-N2 of 119.90 . The doubly reduced species [Mn 0 (bpy-CONHMe)c À (CO) 3 ] ([1] À ) was chemically generated by the reaction of  and KC 8 . The single crystals of [K(18-crown-6)] + [1] À were successfully grown through diffusing pentane into the THF solution of the complex and its X-ray crystal structure is depicted in Fig. 1(right). Compared with , [1] À lost the axial bromide, forming a ve-coordinate species that crystallized in the space group P2(1)/c and its geometry is intermediate between square pyramidal and trigonal bipyramidal with s 5 ¼ 0.24. For perfect square pyramidal and trigonal bipyramidal geometries, the s 5 is 0 and 1, respectively. 46 The dihedral angle N3-C12-C11-N2 of [K(18crown-6)] + [1] À is 52.15 , 67.75 smaller than that of , indicating that the amide -NHMe group is closer to the metal center in the reduced state of the Mn complex in the solid state. The C6-C7 bond in the bpy ring shortens from 1.480 in the crystal structure of [1-Br] to 1.401 in [1] À (Fig. 1), in good agreement with Kubiak's observations and reecting the noninnocent character of the bpy ligand upon reduction. 7 The crystals of complex  were prepared by employing the same method of  and crystallized in the space group P2(1)/n ( Fig. S19 †). The selected bond lengths and crystallographic data of ,  and [K(18-crown-6)] + [1] À are given in Tables S1-S6. †

Electrochemistry
To demonstrate how the local amide -NHMe group in the secondary coordination sphere inuences the electrochemical properties of [fac-Mn(bpy-R)(CO) 3 Br] systems, cyclic voltammetry (CV) curves of complexes ,  and  were recorded ( Fig. S20-S25 †). It is well known that the axial bromo ligand of [fac-Mn(N^N)(CO) 3 Br] can be partially replaced by CH 3 CN in the acetonitrile solution. 5 The solvolysis of  to Br occurs much faster in mixed acetonitrile/water than in dry CH 3 CN (see the section of ligand exchange in the ESI †). In addition, FTIR measurements in the mixed acetonitrile/water solutions also showed that the solvolysis of the complex  occurs faster than that of complexes [ S22a †), and the rst and second reduction potentials are more negative than those of [1-MeCN] + by À90 and À120 mV, respectively.
In CO 2 -saturated dry acetonitrile, a minor change in the reduction peaks (Fig. S20b †) induced by redox-silent solvolysis of Br À to MeCN-bound species is observed, while the addition of H 2 O induced a strong enhancement of the cathodic current ( Fig. 2 and S29 †). The current increase corresponds to the electrocatalytic reduction of CO 2 to CO, as veried by controlled potential electrolysis (CPE) experiments (vide infra). As shown in Fig. S29, † when the concentration of H 2 O was increased to 5.51 M, the complex [1-MeCN] + exhibited the optimal electrocatalytic performance, and three catalytic waves were observed at ca. À1.55, À1.85 and À2.05 V. This phenomenon is different from the catalytic properties of published [fac-Mn(bpy-R)(CO) 3 Br] complexes, which have one or two catalytic waves.
For those that displayed only one catalytic wave, Kubiak, Carter and their co-workers proposed a reduction-rst pathway, 7,47 where the metallocarboxylic acid intermediate fac-Mn(bpy-R)(CO) 3 (COOH) generated through the reaction between the two-electron reduced anionic species [fac-Mn(bpy-R)(CO) 3 ] À and CO 2 was reduced rstly (reduction-rst step) and the protonation of the carboxylic acid group in the second step promotes the C-OH bond cleavage. Once the second coordination sphere effects are introduced into the bpy ligand, some of the [fac-Mn(bpy-R)(CO) 3 Br] complexes start to exhibit two catalytic waves. The waves at low and high overpotential are assigned to the protonation-rst and reduction-rst pathways respectively as theoretically predicted by Carter and experimentally corroborated by Rochford, Grills and Ertem. 41,47 In the protonation-rst pathway, the protonation of the metallocarboxylic acid intermediate fac-Mn(bpy-R)(CO) 3 (COOH) occurs rst followed by H 2 O evolution via the second coordination sphere through weak hydrogen bonding; the reduction of the resulting tetracarbonyl intermediate fac-Mn(bpy-R)(CO) 4 , as the second step, leads to fast dissociation of CO and the regeneration of the catalyst. Interestingly, the complex [1-MeCN] + with a local amide -NHMe group in the secondary coordination sphere displayed three catalytic waves at ca. À1.55, À1.85 and À2.05 V (Fig. 2). On the basis of theoretical calculations, the two catalytic waves at À1.85 and À2.05 V are proposed to be the protonation-rst pathway and the reduction-rst pathway, respectively (see the Computational section below for more details). The rst catalytic wave was proposed to follow the dimer pathway. As shown in Fig. S30, † before the rst catalytic wave there is a small reduction wave, which overlapped fully with the rst reduction wave of the complex In order to validate the electrochemical CO 2 reduction ability of [1-MeCN] + (the real form of the pre-catalyst  in solution) as observed in the CV measurements, controlled potential electrolysis (CPE) was carried out under different applied potentials. Firstly, a 2 hour CPE experiment (Fig. S31 †) was performed at an applied potential of E app ¼ À1.55 V in CO 2 saturated CH 3 CN solution (5.51 M H 2 O). Gas chromatography (GC) analysis shows that CO was the main product during the 2 h electrolysis with a high faradaic efficiency FE CO z 90%, and no hydrogen gas was detected. A charge of 7.24 C passed over 2 h, corresponding to a TON CO value of 7.5. Our results clearly show that [1-MeCN] + could catalyze electrochemical CO 2 reduction at low overpotential. The main catalytic product in the second catalytic wave (E app ¼ À1.85 V) is also CO with FE CO z 90% (Fig. S32 †). A charge of 19.8 C passed over 2 h and the associated TON co is 19. The CPE experiment at E app ¼ À2.05 V also produces CO as the sole gas product with FE CO z 93%   It is also worth noting that the catalytic current of [1-MeCN] + starts to increase from À1.5 V (Fig. 2). To the best of our knowledge, [1-MeCN] + is the most effective Mn-based CO 2 reduction pre-catalyst with low over-potential using water as the proton source (Table S8 †). Overpotential is dened as the difference between the catalytic potential and the equilibrium potential. Appel and Helm suggested the use of the potential at half of the catalytic current as the catalytic potential, denoted as E cat/2 . 48 Then the catalytic potentials for the three catalytic waves of [1-MeCN] + are À1.51, À1.73 and À1.91 V. Matsubara experimentally estimated the standard electrode potential (E 0 ) for the reduction of CO 2 to CO in an acetonitrile-water mixture and showed that the equilibrium potential (E eq ) for the electrochemical reduction of CO 2 to CO can be formulated as: 49 where F is the Faraday constant, R is the gas constant, T is the temperature, c eq CO is the concentration of CO in the solution with CO at 1 bar in the gas phase, c * 0 is the concentration of the catalyst, and D CO (2.2 Â 10 À5 cm 2 s À1 ), 50 D HCO3 À (1.0 Â 10 À5 cm 2 s À1 ), 51 D CO2 (2.0 Â 10 À5 cm 2 s À1 ) 51 and D O (0.5 Â 10 À5 cm 2 s À1 ) 52 are the diffusion coefficients of CO, HCO 3 À , CO 2 and the cata-  (Fig. S37 †). In addition, the current density of all three catalytic waves showed scan rate dependence in the range of v ¼ 0.1 to 1.4 V s À1 (Fig. S38 †); as the scan rate further increases, the catalytic current plateaus of the three waves are relatively scan rate independent in the range of v ¼ 1.4-1.8 V s À1 . The plot of i cat /i p versus inverse square root of the scan rate highlights that steady-state conditions are accomplished at high scan rates over 1.4 V s À1 . These results implied that the pure kinetic regime was approached by increasing the scan rate to 1.4 V s À1 . 41 As shown in Fig. S38b-d, † the TOF max of the third (246 AE 1 s À1 ), second (170 AE 4 s À1 ) and rst (79 AE 2 s À1 ) catalytic processes is calculated using eqn (2) with the data obtained at 1.4, 1.6 and 1.8 V s À1 : 53,54 where F is the Faraday constant, R is the gas constant, T is the temperature, v is the scan rate, n p is the number of electrons involved in the non-catalytic faradaic process (1 electron for [1-MeCN] + ), n cat is the number of electrons required for a single catalytic cycle (2 electrons for CO 2 to CO), and i cat /i p is the ratio of the catalytic current and the non-catalytic faradaic peak current.

Characterization of reaction intermediates
Fourier-transform infrared spectroelectrochemistry (FTIR-SEC) of [1-MeCN] + together with computed n CO bands of several intermediates (  (Fig. S40b †), assigned to the doubly reduced [1] À species. These conclusions are further corroborated by the computed n CO bands (Table S9 †). The doubly reduced [2] À and previously reported [3] À species also have similar CO bands ( Fig. S43b and S41b †). The singly and doubly reduced species of [1-MeCN] + can also be prepared via chemical reduction with KC 8 . As shown in Fig. S44a, † the CO bands of singly and doubly reduced species by KC 8 are very similar to the CO bands observed in the FTIR-SEC spectra, except that the band at 1817 cm À1 of [1] À generated in the FTIR-SEC experiment splits into two CO bands at 1826 and 1800 cm À1 for the chemically generated [1] À in the solution without TBAP. As depicted in Fig. S44b (Fig. 3a), and the corresponding differential IR spectra (Fig. S45 †) conrmed the formation of new species. The doubly reduced species [1] À is not observed. The strong vibrational band at 1680 cm À1 is assigned to free HCO 3 À . 10 The rest of the new bands are tentatively assigned to the metallocarboxylic acid species; specically, the broad band at 1596 cm À1 is due to the vibration of the Mn-COOH group. Kubiak and co-authors have reported the Re-COOH vibration band at 1616 cm À1 , the value of which is very similar to our observation for the Mn-COOH species. 55 Gibson and co-authors have also reported the vibration of the Re-COOH group at 1572 cm À1 . 56 On the other hand, it's not straightforward to make an assignment based on the computed n CO bands, but among them only [1-CO 2 ] À and [fac-Mn(bpy-CONMe)(CO) 2 (CO 2 H)] À exhibit low energy bands in the 1640-1650 cm À1 region together with overlapping bands with those of bicarbonates/carbonates. To further probe the assignment of this metallocarboxylic acid, [1] À was chemically generated by the reaction of [1-Br] and KC 8 , and then reacted with CO 2 . The color of the [1] À solution changed from blue to brown upon the addition of CO 2 -saturated THF solution, and the corresponding FTIR spectral change is shown in Fig. 3b. The CO bands at 1895 and 1861 cm À1 are close to the new set of CO bands at 1915 and 1867 cm À1 in the FTIR-SEC experiment (Fig. 3a) and the small difference is caused by the solvation effect (see the FTIR-SEC spectra of the complex  in THF solution as depicted in Fig. S46 †). Finally, the reaction mixture generated by the reaction of [1] À and CO 2 under dry conditions was analyzed by HR-MS, and apparently many peaks were detected (Fig. S47 †). The isotope distribution pattern at m/z À ¼ 396.0029, 397.0042 and 398.0051 ts well with the calculated isotope distribution pattern of the mass fragment [1 + CO 2 ] À (Fig. S48 †). This mass fragment could be generated from three different species: (i) the metallocarboxylate anion [1-CO 2 ] À ; (ii) the metallocarboxylic acid [1-COOH] via the deprotonation of carboxylic acid; (iii) the metallocarboxylic acid [1-COOH] via the deprotonation of the amide group (Fig. 4). Nevertheless, the formation of metallocarboxylic acid or metallocarboxylate species aer the reaction of [1] À and CO 2 was presumed to occur as a transient intermediate during catalysis.

Computational studies
We performed density functional theory (DFT) calculations at the M06 level of theory 57 in conjunction with the SMD continuum solvation model for acetonitrile 58 as solvent to investigate the catalytic CO 2 reduction mechanism of the complex [1-MeCN] + . On the basis of experimental observations and theoretical calculations coupled with previous mechanistic studies of other Mn-based CO 2 reduction catalysts, 12,59 we propose the reaction mechanism depicted in Fig. 4, and further details on Computational methods, activation of the catalyst (Fig. S49 †) and computed reaction mechanism (Fig. S50-S56 †) are provided in the ESI. † The computed reaction mechanism features three main pathways, namely dimer, protonation-rst and reduction-rst pathways proposed to occur at the corresponding catalytic waves at À1. 55  expected to dimerize to generate [1 2 ] 0 with DG 1 ¼ À11.2 kcal mol À1 . Based on the observed catalytic wave at À1.55 V, we considered binding of CO 2 directly to [1] 0 in the presence of H 2 O but the energy requirement for the formation of [1-CO 2 H] + , formally a Mn II -COOH complex, was found to be prohibitively high (DG z 50.0 kcal mol À1 , Fig. S50 †). The unfavourable energetics of CO 2 binding to [1] 0 also diminishes the possibility of formation of CO 2 sandwiched dimer species [1-CO 2 -1] 0 (Fig. S50 †), as proposed for the Re(bpy)(CO) 3 class of complexes. 60 Next, we considered possible CO 2 activation pathways starting from the [1 2 ] 0 dimer and found dissociation of a CO molecule to generate a vacant site on one of the Mn centers to be energetically more accessible (DG ¼ 22.2 kcal mol À1 ) compared to [3 2 ] 0 (DG ¼ 38.1 kcal mol À1 , Fig. S52 †), in part due to the coordination of the amido group in [1 2 ] 0 to the metal center upon the dissociation of CO (the product is denoted as [1 2 ] 0 -CO). However, the optimized TS structures for CO dissociation involve signicantly high activation free energies (DG ‡ ¼ 38.3 kcal mol À1 , Fig. S51 †), indicating that this pathway is quite unlikely. We still performed an exhaustive search for possible conformers resulting from CO 2 binding to [1 2 ] 0 -CO species and found the located structures to be quite high in energy (DG z 40.0 to 60.0 kcal mol À1 , Fig. S52 †), so that the formation of mer-[1-CO 2 H] + from such intermediates is not plausible. On the other hand, the deprotonation of one of the amide groups in [1 2 ] 0 is possible (pK a ¼ 27.8) and becomes more favorable in the presence of H 2 O (pK a1 ¼ 25.7) (Fig. 4 and S53 †). Thus, we propose that deprotonation leads to stronger coordination of the amide N to the Mn center leading to the destabilization of the dimer and a consequent disproportionation to generate [Mn(bpy-CONMe)(CO) 3 ] and [1] À (DG 2 ¼ 0.0 kcal mol À1 ), the latter of which can follow the protonation rst pathway (described below) at À1.55 V for the catalytic reduction of CO 2 to CO. Although other alternative pathways could not be excluded, the proposed reaction mechanism seems to be the most plausible among the alternatives considered.
Reduction-rst pathway (E ¼ À2.05 V). In contrast, the reduction-rst pathway starts with the reduction of [1-CO 2 H] 0 (E 5 ¼ À1.91 V) followed by C-OH bond cleavage (DG ‡ 4 ¼ 19.4 and DG 6 ¼ À33.5 kcal mol À1 ) (Fig. S55 †) The theoretical calculations demonstrate the critical role of the -NHMe group in the second coordination sphere for the destabilization of the dimer [1 2 ] 0 species upon deprotonation, leading to the disproportionation reaction as well as the stabilization of the transition state of the rate-limiting chemical step of heterolytic C-OH bond cleavage. In the presence of H 2 O as a weak Brønsted acid, the -NHMe group acts as a hydrogen bond and proton donor in the second coordination sphere leading to a DG ‡ 3 of 21.7 kcal mol À1 for the C-OH bond cleavage step via the protonation-rst pathway (Fig. 5a), whereas the absence of hydrogen bonding stabilization from the -NHMe group results in an increase of approximately 7 kcal mol À1 in activation free energy, DG ‡ ¼ 28.5 kcal mol À1 (Fig. 5b), a value quite similar to that computed for the optimized TS structure of complex 3 (Fig. S56, † DG ‡ ¼ 26.3 kcal mol À1 , see the ESI † for further details). This indicates the essential role of the -NHMe group in effectively reducing the energy requirement to perform C-OH bond cleavage and promoting the dimer and protonation-rst pathways and thereby providing access to a low overpotential route for the catalytic reduction of CO 2 to CO with H 2 O as a weak Brønsted acid.

Conclusions
We have reported a Mn-based CO 2 reduction pre-catalyst [1-MeCN] + that bears an amide -NHMe group in the second coordination sphere. The presence of this amide -NHMe group leads to a drastic enhancement in its catalytic activity, especially at mild potentials, using H 2 O as the proton source such that it surpasses the reference pre-catalyst [2-MeCN] + . Theoretical calculations demonstrate the critical role of the amide proton of the -NHMe group in accelerating C-OH bond cleavage and enabling a low over-potential route for the catalytic reduction of CO 2 to CO with H 2 O as a weak Brønsted acid via dimer and protonation-rst pathways. The straightforward synthesis of amide functionalized ligands and the good stability of the amide group allow the ne tuning of the catalyst properties. We believe that this work will inspire the scientic community to develop more efficient Mn catalysts and functional materials using amide groups as the second coordination sphere.
The NMR spectra were recorded on a Bruker 400 MHz spectrometer. Mass spectrometry was performed on a Q-Exactive. Fourier-transform infrared (FTIR) spectra were collected on a Bruker Alfa. Fourier-transform infrared spectroelectrochemistry (FTIR-SEC) spectra were collected on a Bruker V80 with a home-made IR spectroelectrochemical cell. Elemental analysis was performed by using a Thermoquest-Flash EA 1112 elemental analyzer for C, H, and N. The studies of single-crystal X-ray diffraction were performed on a Bruker D8 VENTURE with Mo Ka radiation. The crystals were xed on a Cryoloop to collect data under a N 2 stream at a temperature of 100 K. Bruker SAINT soware was used to integrate the collected data. The structures of the complexes were produced by using the Olex program. The CCDC numbers for complexes ,  and [K(18-crown-6)] + [1] À are 1957948, 1970506 and 1957947, respectively. †

Synthesis of [fac-Mn(bpy-CONHMe)(CO) 3 Br] ([1-Br])
The synthetic route of the complex  is depicted in Fig. S2. † Mn(CO) 5 Br (0.165 g, 1.2 mmol) and bpy-CONHMe (0.107 g, 0.5 mmol) were added into 25 mL of Et 2 O under an Ar atmosphere. The solution was reuxed under dark. Aer 1 h, the product precipitated from the solution. The mixture was cooled to room temperature, and the orange precipitate was ltered off and washed with 10 mL Et 2 O three times. The solid was dried under vacuum to get  with a 70% yield. Single crystals of  were prepared by vapor diffusion of pentane into the THF solution of the complex at À20 C. 1

Synthesis of [fac-Mn(bpy-CONHMe)(CO) 3 (CH 3 CN)](OTf) ([1-MeCN](OTf))
[Mn(CO) 5 Br] (0.275 g, 1.0 mmol) and Ag(OTf) (0.270 g, 1.05 mmol) were added to CH 2 Cl 2 (25 mL). The mixture was stirred in the dark until all the starting material was converted to [fac-Mn(CO) 5 (OTf)] as conrmed by FTIR spectroscopy. The reaction mixture was ltered and the solution was removed by rotary evaporation, yielding a yellow solid. The solid was re-dissolved in diethyl ether (25 mL) and bpy-CONHMe (0.192 g, 0.9 mmol) was then added. The solution was reuxed in the dark overnight, and the precipitate was ltered, washed with 10 mL Et 2 O three times and dried under vacuum, yielding the target complex as a bright yellow solid (0.289 g, 60%). 1

Chemical reduction of [1-Br]
The manganese complex [1-Br] (10 mg, 0.023 mmol) was dissolved in THF in an argon-lled glovebox. In order to get singly reduced species ([1 2 ] 0 ), KC 8 (4 mg, 0.03 mmol) was slowly added into the THF solution and stirred until the IR spectrum of  completely disappeared. The solution was centrifuged to remove the black solid and a dark red solution of the neutral species was obtained. Due to its instability, an attempt to get the single crystals of this neutral species failed. To prepare the twoelectron reduced species [K(18-crown-6)] + [1] À , 18-crown-6 (15 mg, 0.058 mmol) was added into a THF solution of , and then KC 8 (7 mg, 0.053 mmol) was gradually added until the CO bands of the complex [1-Br] and [1 2 ] 0 completely disappeared, resulting in a blue solution of the doubly reduced [1] À complex. The single crystals of [K(18-crown-6)] + [1] À were successfully grown through diffusing pentane into the THF solution of the doubly reduced complex at À20 C.
In situ reaction of [1] À with CO 2 The CO 2 -saturated THF solution (0.5 mL) was mixed with the THF solution of the doubly reduced complex [1] À (1 mL, 0.023 mM under air-free conditions). Then the resulting solution was injected immediately into a Q-Exactive using CH 3 OH/ H 2 O (v/v, 1 : 1; not air-free solvents) as an eluent at a capillary temperature of 160 C.

Electrochemistry
All electrochemical experiments were carried out with a CHI 760 electrochemical workstation. CV measurements were performed in anhydrous CH 3 CN (5 mL) with 0.1 M TBAP as the supporting electrolyte. Glassy carbon (3 mm) was used as the working electrode, Pt wire as the counter electrode and Ag/ AgNO 3 was used as a reference electrode. Ferrocene is used as an internal standard and all potentials reported herein were converted to the Fc +/0 reference scale using E(Fc +/0 ) ¼ E(Ag/ AgNO 3 ) À 0.08 V.
Controlled potential electrolysis (CPE) experiments were carried out in a home-made H-type electrochemical cell in which the cathodic and anodic compartments were separated by glass frit. Typical working conditions are as follows: a glass carbon plate as the working electrode, a Pt plate as the counter electrode, Ag/AgNO 3 as the reference electrode, and 0.1 M TBAP/ CH 3 CN with 5.51 M (v/v) H 2 O as the electrolyte. Solutions were saturated with carbon dioxide before electrolysis. Gas analysis for CPE experiments was carried out using an on-line FuLi Instruments GC9790 Plus gas chromatograph with a ame ionization detector under a constant gas ow (20 mL min À1 ; carry gas ¼ Ar).

Fourier-transform infrared spectroelectrochemistry (FTIR-SEC)
A home-made IR spectroelectrochemical cell was used for this study. The cell consists of a glassy carbon working electrode, a Pt counter electrode, and an Ag pseudo-reference electrode. Ferrocene was used as an internal standard to calibrate the Ag pseudo-reference electrode.

Computational methods
Density functional theory. All geometries were fully optimized at the M06 level of density functional theory 57 with the SMD continuum solvation model 58 for acetonitrile as the solvent using the Stuttgart [8s7p6d2fj6s5p3d1f] ECP10MDF contracted pseudopotential basis set 63 for Mn and the 6-31G(d) basis set 64 for all other atoms. Non-analytical integrals were evaluated using the integral¼grid¼ultrane option as implemented in the Gaussian 16 soware package. 65 The nature of all stationary points was veried by analytic computation of vibrational frequencies, which were also used for the computation of zero-point vibrational energies and molecular partition functions, and for determining the reactants and products associated with each transition-state structure (by following the normal modes associated with imaginary frequencies). Partition functions were used in the computation of 298 K thermal contributions to the free energy employing the usual ideal-gas, rigid-rotator, harmonic oscillator approximation. Free-energy contributions were added to single-point, SMD-solvated M06 electronic energies computed with the optimized geometries obtained with the initial basis with the SDD basis set for Mn and the 6-311+G(2df,p) basis set for all other atoms to arrive at nal, composite free energies.
Solvation and standard reduction potentials. As mentioned above, solvation effects for acetonitrile were accounted for by using the SMD continuum solvation model. A 1 M standard state was used for all species in solution (except for acetonitrile as solvent for which the standard state was assigned as 19.14 M). Thus, the free energy in solution is computed as the 1 atm gas-phase free energy, plus an adjustment for the 1 atm to the 1 M standard-state concentration change of RT ln(24.5), or 1.9 kcal mol À1 , plus the 1 M to 1 M transfer (solvation) free energy computed from the SMD model. The free energy of solvation of protons in acetonitrile is taken as À260.2 kcal mol À1 . 66 Standard reduction potentials were calculated for various possible redox couples to assess the energetic accessibility of different intermediates at various oxidation states. For a redox reaction of the form where O and R denote the oxidized and reduced states of the redox couple, respectively, and n is the number of electrons involved in the redox reaction, the reduction potential E 0 OjR relative to the SCE was computed as where DG 0 OjR is the free energy change associated with eqn (3) (using Boltzmann statistics for the electron) and DE 0 ref is taken as 0.141 V, 67 which is required for the conversion of calculated E 0 OjR versus the normal hydrogen electrode (NHE) in aqueous solution (E NHE ¼ À4.281 V) 68 to E 0 OjR versus the saturated calomel electrode (SCE) in acetonitrile (E SCE ¼ À4.422 V). 69 We obtained reduction potentials referenced to the ferricenium/ferrocene couple by using a shi of À0.384 V from E 0 OjR vs. SCE.

Conflicts of interest
There are no conicts to declare.