Electrocatalytic Conversion of CO2 to Formate at Low Overpotential by Electrolyte Engineering in Model Molecular Catalysis

Abstract An electrolyte engineering strategy was developed for CO2 reduction into formate with a model molecular catalyst, [Rh(bpy)(Cp*)Cl]Cl, by modifying the solvent (organic or aqueous), the proton source (H2O or acetic acid), and the electrode/solution interface with imidazolium‐ and pyrrolidinium‐based ionic liquids (ILs). Experimental and theoretical density functional theory investigations suggested that π+‐π interactions between the imidazolium‐based IL cation and the reduced bipyridine ligand of the catalyst improved the efficiency of the CO2 reduction reaction (CO2RR) by lowering the overpotential, while granting partial suppression of the hydrogen evolution reaction. This allowed tuning the selectivity towards formate, reaching for this catalyst an unprecedented faradaic efficiency (FEHCOO−) ≥90 % and energy efficiency of 66 % in acetonitrile solution. For the first time, relevant CO2 conversion to formic acid/formate was reached at low overpotential (0.28 V) using a homogeneous catalyst in acidic aqueous solution (pH=3.8). These results open up a new strategy based on electrolyte engineering for enhancing carbon balance in CO2RR.


Introduction
Electrochemical CO 2 reduction reaction (CO 2 RR) is a promising method for CO 2 conversion into different value-added products such as carbon monoxide (CO), formic acid/formate (HCOOH/ HCOO À ), alcohols and hydrocarbons and different heterogeneous and homogeneous catalytic approaches have been already studied. [1][2][3][4][5][6] In particular, the production of formate from CO 2 RR is a promising strategy, [7] since formate is a commodity chemical. Molecular catalysis is an interesting approach for CO 2 RR since this type of catalysts offer a high degree of tunability of both the metal center and the ligand. [8][9][10][11] However, molecular catalysts are very seldom soluble in aqueous solution, [12][13][14] the solvent of choice for industrial applications, and for this reason, most studies are limited to organic solvents. In addition, since protons are required for the CO 2 RR, [15,16] hydrogen evolution reaction (HER) represents a critical competitive reaction.
The molecular catalyst for CO 2 RR is dissolved in the solvent together with the electrolyte. Thus, CO 2 is not reacting at the electrode surface. In contrast, the molecular catalyst comes into contact with the electrode for a successful electron transfer, which generates the active form of the catalyst, regardless of the chemical nature of the solid electrode used for that purpose. An alternative strategy to improve molecular catalysts' performances other than the modification of either their metal center or ligands is the modulation of the local electric field by electrolyte engineering using ionic liquids (ILs), since the local environment at the double layer is controlled by the electrolyte composition, but might evolve under operating conditions. So far, most attention has been focused on modulating catalytic electrodes such as Ag or Cu by incorporating ILs, acting as a solvent or a supporting electrolyte [17][18][19][20] to influence the catalytic performance (activity [21][22][23][24][25] and selectivity [26] ) of different electrocatalysts, [27][28][29] as well as a part of the electrolyte membrane. [30] Thus, such an electrolyte engineering strategy [31][32][33][34] aims at controlling the ions present at the electrode-electrolyte interface when the electrical double layer is built up, since it is known to impact the selectivity and the energy efficiency of CO 2 RR. Two main approaches are considered in the literature for that purpose. On the one hand, increasing the hydrophobicity of the electrode surface by addition of long-chain cationic surfactants such as hexadecyl trimethylammonium bromide (CTAB) in solution or drop casting hydrophobic polymers such as poly(vinylidene difluoride) (PVDF) on the electrode surface, which promotes in both cases HER suppression [35][36][37] by forming a nonpolar layer on the electrode. On the other hand, modulating the electric field on the electrode-solution interface, which either stabilizes or destabilizes CO 2 reaction intermediates. [38][39][40] In particular, the potential-dependent orientation of the ions in the electrical double layer implies that mainly cation adsorption happens at the interface when the electrode undergoes cathodic polarization and anion adsorption under anodic polarization. [41] In contrast, very few studies using ILs in solution have been devoted to molecular catalytic systems. [42][43][44][45] In one of those rare examples, we have already demonstrated that ILs in solution acted as catalytic promoters for CO production by decreasing the reaction overpotential, but not affecting the CO 2 RR selectivity. [43] However, the main goal of the present work is to study the effect of ILs not only on the activity, but also on the selectivity (CO 2 RR vs. HER). For that purpose, we used a model molecular catalyst with a well-established mechanism [46,47] for formate production (Scheme S1), [Rh(bpy)(Cp*)Cl]Cl (bpy = bipyridine and Cp* = pentamethylcyclopentadienyl), referred to here as complex [1] (Figure 1). This water-soluble catalyst exhibits moderate selectivity for CO 2 conversion to formate (faradaic efficiency FE HCOO À � 50 %) and also presents significant activity as a HER catalyst. Thus, complex [1] represents a suitable model for studying the IL impact on the catalyst selectivity.
Herein, the impact of two different molecular solvents (acetonitrile and water), two types of ILs (pyrrolidinium-and imidazolium-based ILs, represented in Figure 1), and two proton sources (water and acetic acid), on the selectivity and energy efficiency of CO 2 RR displayed by the selected model molecular catalyst (complex [1]) has been evaluated. We show that, thanks to an IL-based electrolyte, the catalyst allows greater selectivity for formate and higher energy efficiency not only in acetonitrile but, remarkably, also in purely aqueous acidic conditions, which is a rarely reported performance in the case of a molecular complex. This results in an enhanced carbon balance during CO 2 RR regarding the input CO 2 thanks to the acidic electrolyte, which significantly reduces the amount of CO 2 captured as bicarbonate (HCO 3 À ) and carbonate (CO 3 2À ) in the bulk solution [48,49] by conventional strong alkaline or neutral aqueous solutions. However, the main drawback associated with an acidic aqueous electrolyte for CO 2 RR is the more favorable environment for the competitive HER, which highlights the present need of developing new strategies for suppressing HER, such as the one presented here based on an IL electrolyte. Density functional theory (DFT) calculations provide insights into the influence of the ILs on the electronic structure of the catalyst and on the reaction mechanisms at play in both CO 2 RR and HER.

Results and Discussion
The electrochemical characterization of complex [1] Figure 2. Those experimental conditions represent the benchmark conditions previously reported [46] to study electrocatalytic CO 2 conversion to formate using this model molecular catalyst in organic solvents. According to the literature, [46,47,50] the first quasi-reversible reduction wave observed in the black and red plots represented in Figure 2 and centered at À 1.21 V vs. Fc + /Fc has been attributed to the metal center reduction from Rh III into Rh I . This metal-centered redox wave is strongly affected by the simultaneous addition of a proton source and CO 2 in solution ( Figure 2, blue plot), whereby it becomes irreversible due to the chemical reoxidation of the catalyst triggered by the catalytic reduction of CO 2 (see below). In contrast, the second reduction wave, which has been assigned to a one-electron reduction of the bipyridine ligand shifts from À 2.60 V in the absence of CO 2 and proton source in solution to À 2.14 V under those conditions. This is accompanied by a significant increase of current density (j cat /j p = 17.5, where j cat corresponds to the maximum catalytic current   density) confirming a catalytic process. Figure S1 shows the effect on the electrochemical response of complex [1] Figure 1). Moreover, their influence is also evaluated under catalytic conditions, as shown in Figure 3, which compares the benchmark supporting electrolyte and the different types of ILs studied here in a concentration of 0.5 m, since we have already demonstrated in a previous article [43] that no additional effect is provided upon increasing the IL concentration beyond 0.5 m. Moreover, it must be noted that the catalytic current displayed in Figure 3 is independent of the scan rate in all different electrolytes. Table 1 reports the values, determined from Figure 3 and Figure S1, for the following parameters: catalytic potential (potential at the maximal peak current, E cat ), half wave catalytic potential (E cat/2 ), catalytic peak current under CO 2 (j cat ), the ratio of currents under CO 2 and under argon (j p ) compared at the peak potential value (j cat /j p ). However, j p cannot be observed in some cases because imidazolium cations undergo a reduction process at about À 2.5 V, hindering any process occurring at more negative potentials (see Figure S2). [22,51,52] The data shown in Figure 3 and  6 ], which seems to point out that no hydrophobicity modification takes place at the electrodeelectrolyte interface. Note that the effect of imidazolium-based ILs on both the overpotential and the catalytic current density cannot be ascribed to an increase of the local concentration of CO 2 at the electrode surface, since CO 2 is less soluble in imidazolium-based ILs than in acetonitrile [24] and the short-chain ILs studied do not form any nonpolar layer at the electrode, as surfactants do. Next, we aim to study the CO 2 RR in purely aqueous solution, which remains unaddressed so far for complex [1] despite being a water-soluble catalyst. In the following, we explore the increase of the amount of water in acetonitrile and finally, the use of purely aqueous solutions at different pH values. For this purpose, and due to the low solubility in aqueous solution provided by electrolytes containing the PF 6 (Table 1 and Figure S4, blue plots) and furthermore, HER is significantly shifted towards more cathodic potentials, which significantly improves the j cat /j p ratio (Table 1 and Figure S4, black plots). Figure 4 shows the role of adding a weak Brønsted acid (acetic acid) as a more acidic proton donor than H 2 O in solution, together with [TBA] + or [EMIM] + . Addition of acetic acid in the presence of CO 2 (green plots) greatly enhances the catalytic activity of complex [1] with, in both cases, 5 times larger j cat value (by comparing Figure 4 and Figure S4). This can be ascribed to a higher concentration of protonated catalyst when reaching E cat , granted by acetic acid molecules acting as proton donors for direct protonation of the Rh I intermediate (see Scheme S1 for the catalytic cycle). However, acetic acid has almost no effect on E cat/2 under CO 2 and [EMIM] + (1.39 V in Figure 4 vs. 1.38 V in Figure S4). Figure 4 (black plots) also shows control experiments in the presence of acetic acid and complex [1], but in the absence of CO 2 , with either [TBA] + or [EMIM] + in solution, which demonstrates a minor contribution from HER catalyzed by complex [1] within the potential range  studied herein by electrolysis under acidic aqueous conditions (see Table 3). The effect on activity, products selectivity, and energy efficiency of the CO 2 RR due to the presence of ILs in acetonitrile and aqueous solution was studied by controlled-potential (CPE) and controlled-current (CCE) electrolysis. In all experiments, formic acid/formate was detected as the sole product in the liquid phase, and only H 2 was observed in the gas phase. Table 2 shows the overpotential, faradaic efficiencies obtained for both products, and energy efficiency for CO 2 conversion to formate as a function of the electrolyte composition, H 2 O content and either applied potential or current during electrolysis in acetonitrile solution (see Figure S5). As formate can partially migrate from the catholyte to the anolyte, [7] a systematic analysis of both catholyte and anolyte solutions was performed in all electrolysis reported here, proving that between 15 and 20 % of the total formate generated during the electrolysis was detected within the anolyte solution. Thus, analyzing the presence of reaction products in both compartments allowed closing quite efficiently the mass balance of the electrolysis reaching in most cases a total FE (FE HCOO À + FE H 2 ) between 78 and 100 %. Table 2 Table 2 allows to rule out any contribution in CO 2 RR from Rh 0 nanoparticles deposited on the glassy carbon (GC) electrode as a result of complex [1] decomposition/electrodeposition during the electrolysis. For that control experiment, the electrode used during a first electrolysis (entry 4) was recovered, smoothly rinsed and used for a second electrolysis under the same conditions, but in the absence of complex [1], H 2 being almost the sole product formed in that case (entry 7). This result is very   [1] among the top selective molecular catalysts reported in the literature, which exhibit FE HCOO À = 80-97 %. [14] Actually, previously reported results using complex [1] and the benchmark electrolyte ([TBA][PF 6 ]) in electrocatalysis [14,46] never reached FE HCOO À � 50 %. Figure 5 and Table 3 show the CCE (applied current density À 3.33 mA cm À 2 ) results obtained with complex [1] Table 3). This goes along with an evident diminution in overpotential (η = 0.65 V), which is further improved at buffered pH of 3.8 by mixing acetic acid and acetate (η = 0.38 V, entry 3 in Table 3). However, a significant decrease in selectivity towards formic acid/formate is also observed in both cases ( Figure 5 and entries 2 and 3 in Table 3 4 ] induces a rise on the production of formic acid/formate (FE HCOO À /FE H 2 ratio shifts from 0.7 to 1.2 by comparing entries 3 and 4 in Table 3, respectively). This effect cannot be ascribed to an increase of  the concentration of CO 2 in solution, although CO 2 is more soluble in imidazolium-based ILs than in aqueous solution, because the amount of IL present as electrolyte (less than 1 mol %) is not large enough to modify the CO 2 concentration in solution. However, the presence of [EMIM] [BF 4 ] at the double layer could locally increase the molecular catalyst concentration at the electrode surface. In contrast, identical formate production was obtained in additional electrolysis performed by increasing the molecular catalyst concentration from 1 to 5 mm, which demonstrates that a higher concentration of molecular catalyst is not responsible for the formate production enhancement observed in the presence of ILs. Furthermore, the minimum overpotential required for reaching À 3.33 mA cm À 2 (η = 0.28 V) is achieved by combining [EMIM][BF 4 ] and acetic acid/acetate buffer in solution (entry 4 of Table 3). Comparing those results with previously reported molecular catalysts is not easy because experimental conditions vary from one study to another. In any case, 0.28 V seems to be among the lowest overpotential values reported so far for a homogeneous molecular catalyst producing formate in aqueous solution. [14,54,55] We only found in the literature two molecular complexes that behave similarly in terms of energy efficiency, an Ir pincer complex, [54] which displays an overpotential of 0.8 V at À 0.60 mA cm À 2 , but requires small amounts ( � 1 %) of acetonitrile in solution, and an iron carbonyl cluster [Fe 4 N-(CO) 12 ] À , [55] which displays 0.35 V at À 4 mA cm À 2 . However, neutral aqueous solutions were used in both cases. Thus, as far as the authors are aware, not a single example of molecular catalyst for electrocatalytic formate production in such an acidic pH is reported in the literature so far. In addition to this, an interesting energy efficiency (32 %) ( Figure 5), together with partial HER suppression in comparison with [TBA] + are achieved (FE H 2 decreases from 58 to 40 % as comparing entries 3 and 4 in Table 3). Remarkably, this performance at low overpotential is stable in long term CCE ( Figure S8). Furthermore, buffered acidic conditions limit pH changes during CCE, as shown in entries 3-6 of Table 3. In contrast, unbuffered solutions reported in entries 1 and 2 show an undesired, progressive solution alkalization during electrolysis. Figure 5 also shows control experiments without complex [1] (entry 5 in Table 3 and Figure S7) or without CO 2 (entry 6 in Table 3 and Figure S7), which demonstrate the negligible effect of the electrode catalyzing direct CO 2 conversion and confirm CO 2 as the only source of carbon to generate formic acid/formate, respectively. In addition to this, the stability of the IL present in solution during electrolysis was demonstrated by 1 H nuclear magnetic resonance (NMR) spectroscopy, since identical spectra of the IL in solution were obtained before and after the electrochemical reaction ( Figure S9).
Finally, we performed DFT calculations to provide insights into the effects of the [EMIM] + cation on the activity of complex [1]. Previous computational studies on the reduction of CO 2 catalyzed by complex [1] [50] showed that the [Rh III -(bpy)(Cp*)H] + intermediate tends to evolve towards the more stable [Rh I (bpy)(HCp*)] + species bearing a protonated Cp* ligand (see Figure S10a). [50] For this reason, we analyze the interactions between the [Rh I (bpy)(HCp*)] + species and [EMIM] + . Notably, our calculations reveal that the formation of a π cation···π interaction between the catalyst and [EMIM] + represented in Figure S10b stabilizes the lowest unoccupied molecular orbital (LUMO) of the [Rh I (bpy)(HCp*)] + species facilitating its reduction. Accordingly, the calculated reduction potential is lowered by around 170 mV ( Figure S10c) 6 ] as supporting electrolyte ( Table 1). As expected, the interaction between the complex and [EMIM] + is further stabilized upon reduction of the bipyridine ligand (see Figure S10d). Figure 6a compares the Gibbs free-energy profiles in acetonitrile solution for formate and H 2 production catalyzed by complex [1] in the presence or absence of an explicit [EMIM] + interacting at the bipyridine ligand. Starting from the active form of the catalyst A, that is, the [Rh III (bpy *À )(Cp*)H] species (dashed frame in Scheme S1), the reduction of CO 2 in the absence of IL takes place through TS1 overcoming a freeenergy barrier of 14.3 kcal mol À 1 . This generates a formate ion, which is spontaneously protonated, and a Rh II species B. The latter might undergo disproportionation to generate a Rh I and a Rh III species [56,57] or be easily reduced back to Rh I at the working onset potential. The HER in the absence of IL occurs through HÀ H coupling between A and a Zundel cation (H 5 O 2 + ) (TS2) overcoming a very smooth energy barrier of 0.8 kcal mol À 1 from a slightly stabilizing van der Waals adduct. The formation of the H 2 product releasing a water dimer and species B is highly exergonic (> 40 kcal mol À 1 ). Note that although the standard-state free-energy barrier for HER is significantly lower than that for CO 2 RR, the experimental concentration of protons is expected to be several orders of magnitude lower than that of CO 2 , balancing the rate of both pathways and explaining the experimentally observed product distribution (entries 1 and 4 in Table 2). As shown in red lines in Figure 6a, the incorporation of [EMIM] + at the electrode interface scarcely affects the freeenergy barrier for CO 2 RR, showing only a slight increase of 1.3 kcal mol À 1 that lies within the limits of computational uncertainty. Conversely, the HER pathway is more significantly affected, showing an increase of 4.5 kcal mol À 1 in the freeenergy barrier upon the incorporation of [EMIM] + . Notably, this reduces the free-energy difference between TS1 and TS2 from 13.8 to 10.3 kcal mol À 1 in acetonitrile when [EMIM] + is present, thus shifting the product distribution in favor of formic acid/ formate, which can qualitatively explain the experimental selectivity trend observed. Figure 6b and c display the transition states for CO 2 RR and HER in the presence of [EMIM] + . The stronger impact on the HER pathway can be ascribed to the cationic nature of the [Rh III (bpy *À )(Cp*)H]···[EMIM] + complex. The latter might prevent to some extent the approach of other positively charged species such as a free protons, disfavoring the HER process via electrostatic repulsion. In fact, this can be already appreciated in going from species A to the A···H 5 O 2 + adduct, which becomes unfavorable when [EMIM] + is attached to the catalyst structure ( Figure 6, red dashed lines). Analogous results are also obtained in aqueous solution (reducing the freeenergy difference from 13.1 to 9.8 kcal mol À 1 ). It is worth mentioning that having slightly higher free-energy barriers for the hydride transfer step in the presence of [EMIM] + together with a higher experimental current density may sound counterintuitive. However, one should note that the intensity of the catalytic curve might depend on the rate at which the catalyst is reduced at the electrode surface to generate its active species and not on the kinetics of the subsequent, thermally activated chemical step. Thus, bearing in mind the positive impact of [EMIM] + in facilitating the reduction of the catalyst (Figure S10c), the observed faster electron transfer kinetics in the presence of [EMIM] + is the expected outcome.
Interestingly, a Re I complex catalyzing CO 2 conversion to CO [43,44] also exhibits π + -π interactions with [EMIM] + and shares bypiridine with complex [1] as a common ligand. Then, it is highly probable that other active molecular catalysts for CO 2 RR containing bipyridine ligands will exhibit a significant promoting effect by incorporating imidazolium-based ILs at the electrode/solution interface.

Conclusion
Using a Rh-based model molecular catalyst in solution for CO 2 conversion to formic acid/formate (complex [1]), we demonstrate the significant impact of tuning the electrical double layer by electrolyte engineering with ionic liquids (ILs) on both the catalytic activity and selectivity. Firstly, the presence of imidazolium-based ILs was found to decrease the overpotential both in acetonitrile and acidic aqueous solution. Density functional theory (DFT) calculations suggested the formation of π +π interactions between the catalyst and [EMIM] + . The latter facilitate the reduction of the catalyst to generate its active form, explaining thus the decrease in overpotential. Secondly, [EMIM] + cations were found to play a key role partially inhibiting the hydrogen evolution side reaction via electrostatic repulsion between [EMIM] + and free protons, which significantly improves the selectivity of the CO 2 reduction reaction (CO 2 RR) to formic acid/formate production. Any potential hydrophobic effect provided by the presence of imidazolium-ILs at the electrode-electrolyte interface was ruled out, since the enhancement observed in the faradaic efficiency of HCOO À (FE HCOO À ) was not accompanied by any drop in the current density, which actually increased (Figures 3 and 4). This is indeed in contrast with the effect of adding long-chain cationic surfactants in solution, which provokes a significant drop in current density. [35,37] Therefore, complex [1] in the presence of [EMIM] + in acetonitrile exhibits a FE HCOO À � 90 % and a maximal energy efficiency for CO 2 conversion to formate of 66 %, thus placing complex [1] among the top-performing molecular catalysts reported in the literature. [14,54,55] The IL-dependent partial inhibition effect on the hydrogen evolution reaction also allowed, for the first time, efficient CO 2 RR catalyzed by a molecular catalyst under acidic aqueous conditions. A remarkable energy efficiency of 32 % was achieved with complex [1] in acetate buffered solution, thanks to a FE HCOO À of around 45 % coupled with an overpotential of 0.28 V for achieving 3.3 mA cm À 2 , one of the lowest overpotential values reported thus far. [14] Overall, these results in acidic aqueous solution are very promising in order to improve the carbon balance in CO 2 RR by limiting CO 2 losses due to carbonate and bicarbonate generation, which commonly happens in alkaline and neutral aqueous solutions.

Synthesis of complex [1]
The following synthesis was adapted from existing protocols in the literature. [58,59] A methanol solution (30 mL) of 1 equiv. [Rh(Cp*)Cl 2 ] 2 (200 mg, 0.32 mmol) and 2 equiv. 2,2'-bipyridine (120 mg, 0.76 mmol) was stirred at RT for 2 h in the dark. The resulting clear orange-yellow solution was evaporated until dry. The yellow solid was dissolved in a minimal quantity of acetonitrile and precipitated upon the addition of ethyl acetate, then collected on a Buchner funnel and dried under vacuum. The purity of the final precipitate was verified by 1 H NMR spectroscopy according to the literature. [58] Figure S11 shows the 1  and CO 2 (> 99.99 %) gases used to saturate solutions were purchased from Air Liquide. The cyclic voltammetry (CV) experiments were carried out in a three-electrodes setup, with a 3 mm diameter GC disc electrode (0.07 cm 2 ) as a working electrode (BioLogic), which was polished on a polishing cloth on a 1 μm diamond suspension (Struers), sonicated for 10 s in water, and dried prior to experiments. A platinum wire was used as a counter electrode (diameter = 0.5 mm, Alfa Aesar, 99.5 % purity) and was previously flame annealed. The reference electrode used in all cases was a conventional Ag/AgCl/KCl sat reference electrode (BioLogic) separated from the solution by a salt bridge. In acetonitrile however, all potentials were calibrated using the ferrocenium/ ferrocene (Fc + /Fc) redox couple as an internal standard, which was added in the solution at the end of each experiment. CVs were run at 0.01 V s À 1 scan rate and only the third steady state cycle of all CVs is shown, unless otherwise stated in the text.
Catalytic response (j cat /j p ) from CV was calculated as the ratio between the highest value of reduction peak current density exhibited under catalytic conditions (CO 2 ) (j cat ) and the highest value of reduction peak current density exhibited under inert conditions (Ar) (j p ). Catalytic potential (E cat ) corresponds to the value at the maximum of the catalytic current density and (E cat/2 ) corresponds to the half wave catalytic potential.
A gastight two-compartment electrochemical H-type glass cell with a glass frit separating anolyte (5 mL) and catholyte (10 mL) solutions was used in all electrolyses reported here. Controlled potential or current electrolysis (CPE and CCE, respectively) were performed in acetonitrile solution containing 5 % v/v H 2 O and 0.1-0.5 m of supporting electrolyte previously saturated with CO 2 by gas bubbling in both catholyte and anolyte, but no continuous CO 2 gas was purged during the electrolysis. 1 mm of complex [1] was only added in the catholyte. The working electrode was a 1 cm 2 GC plate (1 mm thick, type 2, from Alfa Aesar) the counter electrode was a 5 cm 2 GC rod (Alfa Aesar) and a conventional Ag/AgCl/KCl sat electrode separated from the solution by a salt bridge, which was calibrated with ferrocene as an internal redox reference, was used as a reference electrode. Ohmic losses in the cell were minimized by achieving the minimal distance between electrodes and keeping magnetic stirring during the electrolysis.

Analytical quantification of products
Gas products were quantified by gas chromatography (Model 8610 C SRI Instruments) equipped with thermal conductivity detector (TCD) and flame ionization detector (FID) from 50 μL aliquots of the headspace of both compartments. Only hydrogen (H 2 ) was detected as a gas product. Liquid products were evaluated using an ionic exchange chromatograph (IC) (Metrohm 883 Basic IC) equipped with a Metrosep A Supp 5 column and a conductivity detector. Only formate was detected. A typical quantification of formate by IC required the sampling of 50 μL of solution from catholyte and/or anolyte, followed by a (200-400) dilution in ultrapure water and a final injection of 20 μL into the IC chromatograph. FE of each reaction product is calculated from the ratio between the charge consumed to form each product and the total circulated charge. [60] However, the total circulated charge is corrected to discount the initial three electrons consumed by complex [1] (1 mm in solution) necessary to generate its active form. Catalyst activation charge = [number of electrons × Faraday constant × mol of catalyst] = [3 × 96485 × 6.06 × 10 À 6 ] = 1.82 C. In order to compare all CPE and CCE results, a constant total charge (15 C in acetonitrile solutions and 10 C in aqueous solutions) has been used in all electrolysis. The overpotential (η) was calculated from the difference between the electrolysis applied potential and E 0 CO 2 =HCOO À (CH 3 CN, H 2 O) = À 1.32 V vs. Fc + /Fc or E 0 CO 2 =HCOOH (H 2 O) = À 0.199 V vs. standard hydrogen electrode (SHE) in acetonitrile [53] and aqueous solutions, [61] respectively. Additionally, E 0 CO 2 =HCOOH in aqueous solution was transferred from SHE to the Ag/AgCl/KCl sat reference electrode taking into account the solution pH and using the following Equation (1) The cathodic half reaction energy efficiency (EE) was calculated for CO 2 conversion to formate reaction according to the following Equation (2): [19] EE %