Chapter 1: Approaches to Controlling Homogeneous Electrochemical Reduction of Carbon Dioxide

Carbon dioxide is the end product of combustion and respiration, and its reduction back to energy-rich products has challenged chemists for well over a hundred years. The first reports of the direct electrochemical reduction of CO₂ appeared in the early 1900s.^1,2  As the field of homogeneous catalysis grew during the 1960s and 1970s, chemists recognized that molecular electrocatalysis was also feasible. The first reports of homogeneous electrochemical reduction of CO₂ by molecular catalysts appeared in the 1970s and 1980s. These included metal phthalocyanine complexes,^3,4  metal tetraaza-macrocycles,^5,6  late metal phosphine complexes,⁷  and Re (i) bipyridyl carbonyl complexes.⁸  A molecular electrocatalyst both participates in an electron transfer reaction (at an electrode) and increases the rate of a chemical reaction. In organic electrochemistry, mediators are often used to mediate the flow of charge between an electrode and a chemical substrate, thus acting as an outer sphere electron donor or acceptor toward the substrate.⁹  An electrocatalyst both supplies charge and performs inner sphere reduction or oxidation of an otherwise kinetically stable substrate. Both the electron transfer and chemical kinetics must be fast for an electrocatalyst to be efficient. Additionally, the electrocatalyst must display a good thermodynamic match between the redox potential (E⁰) for its electron transfer reaction and the chemical potential difference between the products and reactants (i.e. thermodynamic driving force) for the reaction that is being catalyzed (e.g. reduction of CO₂). A significant advantage of molecular electrocatalysis is that these factors can be optimized by chemical tuning of the metal centers via appropriate ligand design.

Over the past ten years, several comprehensive reviews of the homogeneous electrochemical reduction of CO₂ have appeared.^10–15  In addition to the effects of metal type, d-electron configuration, and ligand type, new dimensions in the performance of molecular catalysts have been identified. Many breakthroughs in our understanding of how molecular catalysts respond to their local environment in the electrochemical cell, e.g. solvent, electrolyte, and applied bias have occurred over the last decade. One example of innovative catalyst design is the use of pendent proton relays, as well as charged and hydrogen bonding groups attached to the catalysts have been reported. This chapter presents much of the new knowledge that has been gained over the past ten years in the homogeneous electrochemical reduction of CO₂, with the aim of identifying areas where new research may produce further breakthroughs in our understanding and control of the reduction of CO₂ for its conversion to high-value chemicals like CO, formate, methanol, and methane.

This section will give a brief introduction to the parameters commonly used to describe molecular electrochemical catalysis. A more in-depth discussion of these parameters can be found in the ‘Catalyst Comparisons’ section of this text.

Catalytic Tafel plots are used to visualize catalyst performance using the overpotential (η) and turnover frequency (TOF). The overpotential is the additional potential applied beyond the thermodynamically determined potential (E⁰) required for the electrochemical reaction to transpire. The reaction conditions, such as solvent and pK_a of all present proton sources (i.e. water, phenol, or acetic acid), must be considered to appropriately determine E⁰ of the redox reaction. Furthermore, the products formed (i.e. CO, MgCO₃, HCO₃⁻) are considered in the determination of E⁰_cat. Thus, the calculation of overpotential is not trivial but enables the comparison of electrocatalysts under different reaction conditions. The TOF is the number of chemical conversions of the substrate per unit of time (typically in seconds). Catalysts that have a high TOF for a selective transformation with minimal overpotential are ideal.

The catalytic potential for the electron transfer event is reported in several ways. The potential value at the peak catalytic current (E⁰_cat) is typically reported versus a standard electrode potential such as saturated calomel electrode (SCE), silver chloride electrode (Ag/AgCl), or normal hydrogen electrode (NHE). Additionally, this potential may be adjusted based on the redox couple of an internal standard like ferrocene/ferrocenium (Fc^+/0). The potential at half the catalytic current (E⁰_cat/2) is normally found at the steepest part of the catalytic wave and may also be reported. The Faradaic efficiency (FE) describes the amount of product produced per number of electrons transferred to facilitate the electrochemical reaction. The following sections will provide many of these parameters for an assortment of electrocatalysts.

Molecular catalysts can typically be divided into classes based on ligand architecture. Our research group previously described five main classes of catalysts in a 2009 tutorial review published in Chem. Soc. Rev.: metals ligated by porphyrins, cyclams, bipyridyls, polypyridyls, and phosphines.¹⁰  This section will briefly recount the molecular catalysts used for electrocatalytic CO₂ reduction by Periodic Group in the literature to date.^10–15  Improvements in catalytic activity via changes to the metal identity and/or ligand structure, including tuning of the steric and electronic properties, are discussed. The discussion will also describe new strategies for improving catalytic selectivity (i.e. pendent proton relays, Lewis acid additives, hydrogen bonding, and others).

The Kubiak and Cowan groups have examined electrocatalytic CO₂ reduction with bipyridine complexes of Group 6 metals. The metal tetracarbonyl bipyridyl, [M(R-bpy)(CO)₄] (M=Mo, W; R=H, tBu), complexes are active for CO₂ reduction with quantitative Faradaic efficiency (FE) for CO at −2.3 V vs. saturated calomel electrode (SCE) [ca. −2.7 V vs. ferrocenium/ferrocene (Fc^+/0)] in MeCN.¹⁶  The reaction rates are significantly slower than their Group 7 counterparts. Density functional theory (DFT) calculations revealed stronger π-back bonding into the CO ligands with Group 6 metals compared to Re, suggesting that CO release from [M(R-bpy)(CO)₄]²⁻ limits the CO₂ reduction rates of these complexes. The catalytic onset potential of [M(bpy)(CO)₄] (M=Mo, W, Cr) was anodically shifted at Au electrodes compared to glassy carbon (GC) electrodes.¹⁷  Cowan and co-workers used vibrational sum-frequency generation spectroscopy to show that CO loss is enabled by strong interactions with the Au surface, providing access to a lower-energy pathway for generation of the active species [M(bpy)(CO)₃]²⁻.¹⁸  This is an example of chemically relevant interactions that the electrode materials may have in “homogenous” molecular electrocatalysis.

Other Group 6 complexes with modified bidentate ligands¹⁹  or [Mo(CO)₂(η³-allyl)(α-diamine)(NCS)]²⁰  also showed modest catalytic current increases under CO₂. The former systems had low FEs for CO and reduction products were not quantified in the latter. Pyridine monoimine (PMI) analogues [(PMI)Mo(CO)₄] exhibited catalytic behavior under CO₂ on the first voltammogram sweep, but diminished substantially on the second scan.²¹  Chemical reduction studies implicated the formation of a CO₂ adduct that interacts across the metal and imine carbon. This irreversible ligand-based reactivity is responsible for the inability of the catalyst to turnover. Grice and co-workers showed that Group 6 carbonyl complexes without other ancillary ligands are capable of CO₂ electroreduction.²²  [Mo(CO)₆] was the most active in this series, producing CO at −2.8 V vs. Fc^+/0 (up to 95% FE_CO) while Group 7 analogues Re₂(CO)₁₀ and Mn₂(CO)₁₀ were not active. The reactivity of related Group 4−6 metallocenes [Cp₂MCl₂] with CO₂ were also studied by the same group.²³  The niobium, molybdenum, and tungsten complexes reduce CO₂ to CO, albeit at very negative potentials (−3 V vs. Fc^+/0) with low current efficiencies. Cleary, while bipyridine or other redox-active ligands are not required for CO₂ reduction here, the presence of such ancillary ligands enables significantly milder operating potentials. A radically different approach was demonstrated by Jayarathne et al. utilizing labile phosphine ligands and chloride ions on a tungsten complex. This complex was able to reduce CO₂ to CO via a mechanism involving a tungsten oxo intermediate.²⁴ 

Lehn and co-workers first described the activity of [Re(bpy)(CO)₃Cl] (Re-bpy, 1) for electrocatalytic CO₂ reduction in 1984.⁸  This system, sometimes referred to as the Lehn catalyst, reduced CO₂ to CO at −1.49 V vs. SCE (ca. −1.9 V vs. Fc^+/0) in 9 : 1 N,N-dimethylformamide (DMF)/H₂O with very high selectivity for CO over H₂ evolution (FE_CO=98%). However, the rate of catalysis, defined by the turnover frequency (TOF), was reported to be relatively slow (TOF=21.4 h⁻¹).

Infrared-spectroelectrochemistry (IR-SEC) studies indicate that significant Re–Re dimer formation occurs following one-electron reduction of 1, which was later confirmed by comparison to the independently prepared dimer.²⁵  Additional reduction is then required to break the Re–Re bond to generate the proposed active catalyst, [Re(bpy)(CO)₃]⁻. Introduction of t-butyl groups in the 4,4′ positions (denoted as [Re(tBu-bpy)(CO)₃Cl], 2) dramatically increased the rate of CO₂ reduction.²⁶  The steric bulk of these substituents largely prevents dimerization as shown by IR-SEC (Figure 1.1) and stopped-flow IR studies.²⁷  Furthermore, the reduced species, [Re(R-bpy)(CO)₃]⁻, reacts with CO₂ approximately ten times faster when R=tBu versus H, suggesting an additional effect of the t-butyl groups (see Section 1.9). Replacement of the halide ligand with a chelating phosphazane (PNP) ligand in [Re(bpy)(PNP)(CO)₂]OTf is another effective strategy to prevent dimerization as the ligand does not fully dissociate upon reduction.²⁸ 

DFT calculations predict that the HOMO of the two-electron reduced form of the Re(bpy)(CO)₃Cl catalyst has a mixed metal–ligand character.³⁰  It has been proposed that the delocalized electronic structure of [Re(bpy)(CO)₃]⁻ contributes to the high selectivity of CO₂ reduction over H₂ evolution in the presence of a proton source as formation of a metal hydride requires the electron density to be localized in the Re dz² orbital. This proposal was further supported by X-ray absorption studies³¹  and Raman spectroscopy³²  which indicated that the active doubly-reduced species [Re(bpy)(CO)₃]⁻ is best described as [Re(0)(bpy˙)(CO)₃]⁻ where one electron has been added to the bpy π* orbital and the other to the Re d_z² orbital. This electronic configuration was posited to enable reaction with CO₂ through σ and π interactions with a lower reorganization energy near the transition state than for the reaction with a proton, resulting in CO₂ reduction being kinetically favored over H₂ evolution.³¹  This was further corroborated via DFT calculations by the Carter group that showed that protonation of the anion to produce the Re–hydride has a higher activation energy than formation of the Re–COOH complex.³³  Manganese bipyridine catalysts have been studied as more Earth-abundant and less expensive alternatives to the Re–bpy systems.^34,35  Johnson et al. originally reported that the doubly reduced [Mn(bpy)(CO)₃]⁻ anion (analogous to the active Re catalyst) does not react with CO₂.³⁶  Bourrez et al. discovered that these Mn-bpy compounds are catalytic in the presence of water, and have lower overpotentials (around 0.40 V) when compared to Re-bpy analogues.³⁵  The overpotential (η) in this case is the potential difference between the thermodynamically determined reduction potential for aqueous CO₂ reduction and the experimentally observed reduction potential for Mn-bpy compounds under CO₂.

Fe porphyrins have been extensively studied for electrocatalytic CO₂ reduction for the past three decades, in particular by Savéant and co-workers.^37,38  Fe(iii) tetraphenyl porphyrin (FeTPP, 3) undergoes three sequential one-electron reductions in DMF under Ar. The current increases at the third reduction (ca. −1.6 V vs. SCE) under CO₂, corresponding to electrocatalytic CO₂ reduction to CO (Figure 1.2a). However, electrolysis at −1.8 V vs. SCE (ca. −2.2 V vs. Fc^+/0) revealed that the catalyst rapidly degrades under catalytic conditions, perhaps from carboxylation and/or hydrogenation of the porphyrin.³⁹ 

Ru and Os complexes of the form [M(bpy)(CO)₂Cl₂] undergo polymerization on the electrode upon reduction to form heterogeneous polymer films that are active for CO₂ reduction to CO and formate.^41,42  Following earlier reports of using bulky substituents on the bipyridine ligand to prevent metal–metal dimerization, it was demonstrated that installation of mesityl (mes) groups in 6,6′ positions prevent polymerization, and this system is an active homogeneous catalyst for CO₂ reduction to CO (95% FE) at −2.2 V vs. Fc^+/0 in the presence of phenol.⁴³  Interestingly, the yield of CO is potential-dependent: at −1.7 V vs. Fc^+/0, FE_CO decreased to 63% and formate was detected using NMR. IR-SEC studies support an electron transfer-chemical reaction-electron transfer (ECE) mechanism where one-electron reduction of Ru(mes-bpy)(CO)₂Cl₂ (4) leads to chloride loss, followed by protonation and another one-electron reduction to form a Ru(ii)–hydride complex (4b). Complex 4b can react with CO₂ to generate a formate complex [Ru-η¹-OCHO] (4c) (Scheme 1.1). Dissociation of formate from 4c may occur, closing the catalytic cycle for CO₂ reduction to formate. However, additional one-electron reduction of 4c appears to increase the rate of thermal dehydration of Ru-η¹-OCHO, generating CO and water and shifting the product ratio to favor CO. Min and co-workers made a Ru pincer compound using bipyridine with an aminophosphine substituent at the 6 position. An increase in current was observed under CO₂.⁴⁴  The current further increased under CO₂ when 3% H₂O was added. Like the M(bpy)(CO)₂Cl₂, controlled potential electrolysis (CPE) showed that the compound made both CO (FE=60.7%) and formate (FE=37.3%) with trace H₂ in 3% H₂O/MeCN. The compound also showed stability over 24 h, yet the catalyst only achieved 11.2 turnovers during that time.

Meyer and co-workers performed detailed studies of [Ru(tpy)(bpy)X]²⁺ (X=MeCN, Cl⁻) to elucidate the pathway for CO₂ reduction in MeCN.⁴⁵  [Ru(tpy)(bpy)(MeCN)]²⁺ (5) and [Ru(tpy)(MebImpy)(MeCN)]²⁺ (6) exhibit two reversible reductive features in cyclic voltammetry under Ar, corresponding to the reduction of terpyridine (tpy) and bipyridine (bpy) or 3-methyl-1-pyridylbenzimidazol-2-ylidene (MebImpy), respectively. Current enhancement is observed at the second reduction in the presence of CO₂, as well as at more negative potentials (Figure 1.3a). The catalytic peak current is scan rate dependent and the wave does not show the typical plateau shape for a rate-limiting catalytic reaction. The rate of catalysis also exhibits saturation behavior with respect to CO₂ concentration. When considered together, these observations are consistent with the rate-limiting step being substitution of the MeCN ligand by CO₂ in the doubly-reduced species [Ru(tpy⁻)(bpy⁻)(MeCN)]⁰. This substitution generates [Ru(tpy⁻)(bpy⁻)(CO₂²⁻)]⁰ which undergoes further reduction leading to overall CO₂ reduction and disproportionation to CO and CO₃²⁻ (Scheme 1.2).

Substitution of MeCN by CO₂ in 5 proceeds via a dissociative mechanism; therefore, modifications that increase the rate of MeCN dissociation should increase the rate of catalysis. Meyer and co-workers⁴⁵  developed the benzimidazolylidene analogue 6, where the carbene ligand is more donating than bipyridine (Figure 1.4). The increased electron density at the metal increases the lability of MeCN and increases the rate of catalysis from 5.5 to 19 s⁻¹ (Figure 1.3b). Notably, 5 is also active for the hydrogen evolution reaction (HER) and the oxygen evolution reaction (OER). These reactions have been coupled with CO₂ reduction to develop single-catalyst systems capable of producing pure CO or controlled syngas ratios (CO/H₂) at the cathode and O₂ at the anode.^46–48  Additional studies by Ott and co-workers⁴⁹  found a correlation between the rate of CO₂ reduction and the rate of monodentate ligand dissociation, which was varied by systematically changing the electron-donating character of the bidentate ligand in 5. Increasing the ligand donating ability promotes MeCN loss and increases the catalytic rate. Other variations to the original catalysts, [Ru(tpy)(bpy)Cl]⁺ and [Ru(bpy)₂(CO)Cl]⁺, have been explored.^50–52  However, in these reports, the new complexes exhibit similar catalytic behavior as the parent systems and are proposed to proceed via an analogous mechanism for CO production involving a key [Ru(tpy)(bpy)(CO₂²⁻)]⁰ intermediate.

A related Ru complex [CpRu(bpy)(MeCN)]⁺ (8), (Cp=cyclopentadiene), was recently reported to facilitate rapid electrochemical CO₂ reduction and disproportionation to CO and CO₃²⁻ triggered by initial one-electron reduction at −1.9 V vs. Fc^+/0.⁵³  An ECE pathway for initial CO₂ activation analogous to that shown in Scheme 1.3 for 5 was proposed. Substitution of the MeCN ligand by CO₂ occurs with one-electron reduction, after which addition of the second electron occurs at a more positive potential than the first reduction to generate a [CpRu(bpy)(CO₂)]⁻ complex. However, this system is not catalytic at this potential due to the high stability of the resulting [CpRu(bpy)(CO)]⁺ species.

Co porphyrins have also been shown to exhibit excellent activity for electrocatalytic CO₂ reduction.^54–56  Carbon monoxide is the main CO₂ reduction product using Co porphyrins but the selectivity over H₂ evolution varies depending on the solution pH. Hydrogen dominates at pH 1 while up to 60% CO is produced at pH 3.⁵⁴  DFT calculations by the Koper group⁵⁵  and by Leung and co-workers⁵⁴  supported a mechanism in which a reduced Co(i) complex reacts with CO₂ to generate a [Co(i)Por–CO₂]⁻ adduct (Por=porphyrin).

The electrochemical conversion of CO₂ to CO by Co(ii) catalysts bearing N₄-tetradentate ligands at a glassy carbon electrode (GCE) was investigated by Wang and co-workers.⁵⁷  The most effective catalyst cis-[Co(PDP)Cl₂] (PDP=1,1′-bis(2-pyridinylmethyl)-2,2′-bipyrrolidine) exhibited FE_CO=96% at −1.7 V vs. SCE (ca. −2.1 V vs. Fc^+/0) with no concomitant H₂ evolution.

While phosphine ligands are commonly used for other catalytic processes, they are rarely utilized in electrocatalytic CO₂ reduction catalysts. Early examples for CO₂ reduction with group 9 metal phosphine complexes date back to the 1980s. Slater et al. reported electrochemical CO₂ reduction with Rh(diphos)₂, (diphos=1,2-Bis(diphenylphosphino)ethane) and trinuclear Ni clusters with phosphine ligands have also been reported as electrocatalysts of modest activity.^7,58  CpCo(P₂N₂) was found to catalyze the reduction of CO₂ to formate and is discussed in detail in Section 1.5.⁵⁹ 

The first report demonstrating electrocatalytic CO₂ reduction with molecular Ni and Co catalysts was published by Eisenberg and co-workers in 1980.⁵  Reasonable FE_CO (up to 65%) using Ni(ii) cyclam derivatives were achieved in water/MeCN (2 : 1) between −1.5 V and −1.6 V vs. SCE (ca. −1.9 to −2.0 V vs. Fc^+/0) at a Hg electrode. Hydrogen made up the remainder of the charge balance. Furthermore, Sauvage and co-workers showed that [Ni(cyclam)]²⁺ (9) is exceptionally selective for CO (96%) in pure water at −1.05 V vs. NHE at a Hg electrode.^6,60  As shown in Figure 1.5, a large catalytic wave is observed by cyclic voltammetry in water under CO₂, corresponding to TOF=32 h⁻¹ and making 9 one of the more active molecular catalysts for electrocatalytic CO₂ reduction to CO. Ni(ii) cyclams substituted with C-alkyl groups were shown to have higher TOF than 9 in 20% MeCN/H₂O.⁶¹  Schiff bases have been widely explored as planar, tetradentate ligands for catalysis, but have only recently been investigated for CO₂ reduction. A series of Ni(ii) and Cu(ii) Schiff base complexes exhibited current enhancement under CO₂, but further studies are needed to quantify their activity.^62,63  Mukherjee and co-workers⁶⁴  prepared cathode materials of related Ni(ii) and Cu(ii) salen complexes, which reduced CO₂ to CO and C1+ hydrocarbons. The activity of these materials differed from that of pure metals but since production of hydrocarbons with molecular catalysts is exceedingly rare, further studies to identify the true active species are warranted.

Moving beyond cyclam and related macrocycles, several research groups have made efforts to develop new CO₂ reduction catalysts based on other tetradentate nitrogen ligands (Figure 1.6). A series of Ni(ii) complexes (10a-c) with an N-heterocyclic carbene–pyridine ligand yielded CO as the major product of CO₂ reduction at −1.5 V vs. SCE (ca. −1.9 V vs. Fc^+/0) at GC in MeCN.^65,66  Unfortunately, Faradaic efficiencies were not reported and the catalysts decomposed over extended times. Further optimization of this ligand architecture led to improved stability and up to 0.2 V decrease in operating potential for CO₂ reduction, but the catalytic rates remained far slower than for 9.⁶⁶ 

Group 10 metal phosphines were studied primarily by the Dubois group in the 1990s and 2000s.^11,58,67  A library of [Pd(triphosphine)(S)](BF₄)₂ complexes were screened as CO₂ reduction catalysts.⁶⁷  It was found that triphosphine complexes with an additional labile ligand produce CO. In contrast, later studies with tetraphosphine ligands show that the reactions proceed through a hydride intermediate to produce formate. Substitution of one or more phosphines with heteroatoms resulted in complete loss of activity. One of the main degradation pathways is the formation of dimers by the reduced state. Several new derivatives were synthesized, including dimeric catalysts that showed extraordinary enhancement in catalytic performance.⁶⁸  The bimetallic bridged [Pd₂(CH₃CN)₂(eHTP)](BF₄)₄ (eHTP=bis(bis((diethylphosphino)ethyl)phosphino)methane) complex has a very similar structure to the [NiFe] CO dehydrogenase enzyme and di-metallic cooperativity is proposed to play a crucial role in catalysis.¹¹  The flexible modular synthetic pathways to those ligands gave rise to a large variety of similar ligands with different functional groups attached to enhance cooperative catalysis; those catalysts are discussed in further detail in Section 1.5.⁶⁹ 

In recent years, carbene pincer complexes of group 10 metals were shown to reduce CO₂ selectively in the presence of acids. In 2014 Wolf et al. reported the use of tridentate N-heterocyclic carbene (NHC) ligands with Pd to produce CO with moderate efficiency and selectivity.⁷⁰  The NHC ligands bind to the metal in a CNC configuration (two NHC carbon atoms and 1 pyridine nitrogen atom). Sheng et al. showed that Pd can be substituted by the more abundant Ni.⁷¹  Independently, Cope et al. reported a CCC-NHC pincer Ni complex that exhibited similar behavior.⁶⁹  “CCC-NHC” denotes the binding of the metal by 3 C atoms of the ligand. In both cases, the main product was CO with water present as a proton source. Notably, systematic studies comparing Ni, Pd, and Pt by Wolf et al. found that a very similar catalyst mainly produces hydrogen in the presence of CO₂ and added TFA, and only the Pd catalyst is active for CO production.⁷²  These findings emphasize the influence of the proton source on the performance of CO₂ reduction catalysts. The performance of these catalysts was only slightly improved, FE_CO increased to 28% from 23%, by tuning the electronic properties of the ligand.^11,70 

While the appropriate combination of metal and ligand is crucial for the design of an active catalyst for CO₂ reduction, tuning the electronic and/or steric properties of the ligands often only leads to an incremental increase in catalytic performance. Furthermore, simultaneously improving the catalytic rate and decreasing the overpotential is very challenging with simple ligand optimization. In evaluating the comprehensive body of work in this field, we conclude that other factors can have just as much of an effect on improving catalysis, if not more, than basic changes to the metal and/or ligand. These other factors include: selection of the appropriate Brønsted acid source (although water is preferred), addition of stoichiometric or co-catalytic additives, introduction of pendent functional groups, and multinuclear metal ensembles. Here, we highlight the recent literature where these parameters are considered to control the performance of many of the catalysts presented in Section 1.3.

Savéant and co-workers demonstrated that the rate of CO₂ reduction and FeTPP catalyst lifetime are increased by the addition of weak Brønsted acids (Figure 1.7).^73,74  Notably, the product distribution depends on the strength of the Brønsted acid. Specifically, the formic acid (HCOOH) yield is inversely correlated with the Brønsted acid strength. For example, the addition of 1-propanol resulted in both CO and formate (FE≈60% and 35%, respectively), while the stronger acid trifluoroethanol (TFE) yielded only CO (FE>96%). This counterintuitive behavior arises from the greater hydrogen bonding stabilization of catalytic intermediates with stronger acids. In the presence of Brønsted acid, water is formed as a co-product as shown in Scheme 1.4, thereby eliminating the issue of precipitation of carbonate salts on the electrode.

A more robust ligand, bis-hydroxyphenyl bipyridine, was recently explored for CO₂ reduction by Machan and co-workers (Figure 1.8).⁷⁵  In the absence of Brønsted acid, the Fe complex, Fe(tBudhbpy)Cl (12) (tBudhbpy=6,6′-di(3,5-di-tert-butyl-2-hydroxybenzene)-2,2′-bipyridine) facilitated CO₂ reduction and disproportionation to CO and CO₃²⁻ at −2.5 V vs. Fc^+/0 in DMF at GC; however, only a FE_CO of 1.1% is observed due to strong binding of CO to the Fe center. This behavior is similar to that of FeTPP (3) at GC electrodes: CO dissociation from the metal is rate limiting and the high CO binding affinity eventually leads to catalyst decomposition. However, upon addition of phenol to 12, an increase in current is seen, and formate is observed as the main CO₂ reduction product (FE=68%). IR-SEC supports a mechanism involving a Fe(iii)-hydride species, which upon further reduction is capable of hydride transfer to CO₂ to generate formate.

The proposed mechanism for CO₂ reduction with the Re-bpy catalysts is shown in Scheme 1.5. Following CO₂ binding at the active catalyst [Re(bpy)(CO)₃]⁻, protonation yields a hydroxycarbonyl species that undergoes further reduction and protonation to generate [Re(bpy)(CO)₄] via rate-limiting C-O cleavage.⁷⁶  The Re(COOH)(bpy)(CO)₃Cl intermediate for 1 and the t-butyl analogue 2 have been observed by stopped-flow IR spectroscopy.⁷⁷  This proposed mechanism was also supported by DFT calculations from Carter and co-workers.⁷⁸  In the absence of Brønsted acid, ligand-bound carboxylate is produced.⁷⁹  DFT calculations on the Re centered radical [·Re(bpy)(CO)₃] purported the insertion of CO₂ into a Re carboxylate dimer [Re(Me-bpy)(CO)₃]₂(µ-CO₂).⁷⁹ 

To better understand the mechanism of catalyst regeneration, the tetracarbonyl species, [Re(bpy)(CO)₄]⁺ and [Re(tBu-bpy)(CO)₄]⁺, were synthesized.⁷⁹  IR-SEC and chemical reduction studies revealed that these complexes undergo substitution at reducing potentials, releasing CO and generating [Re(R-bpy)(CO)₃(CH₃CN)]⁺ in an electron-transfer catalyzed process. This reaction occurs at potentials almost 0.5 V more positive than the catalytic operating potential for 1, which implies that CO elimination and regeneration of the active catalyst is facile and not rate-limiting in electrocatalytic CO₂ reduction.³⁶ 

Mn-bpy compounds have important mechanistic differences from their Re-bpy counterparts. Detailed mechanistic studies of Mn(tBu-bpy)(CO)₃Br revealed that one-electron metal-based reduction is followed by immediate loss of bromide to give a five-coordinate Mn-based radical.⁸⁰  DFT calculations indicate that the Mn-based radical is more stable than the six-coordinate MeCN-coordinated complex, contrary to the Re-bpy analogue.³³  The dimer [Mn(tBu-bpy)(CO)₃]₂ is rapidly formed, at a rate 10⁹ times faster than the analogous Re–Re dimer.^33,81,82  The doubly reduced anionic species [Mn(R-bpy)(CO)₃]⁻ is formed after reductive cleavage of the dimer, and this is the chemical process that ensues the second reduction observed by cyclic voltammetry.^35,81,83  A CO₂ molecule is then added to the nucleophilic metal center of the anionic species, and subsequent protonation of the bound CO₂ ligand leads to a hydroxycarbonyl [Mn(R-bpy)(CO)₃(C(O)OH]⁰ complex. The mechanism can proceed via two different pathways: reduction-first or protonation-first (Scheme 1.6). In the reduction-first pathway, the hydroxycarbonyl complex is reduced to give [Mn(bpy-R)(CO)₃(C(O)OH]⁻, which undergoes C–OH bond cleavage initiated by protonation of the OH to form water and CO. In the protonation-first pathway, the hydroxycarbonyl complex is protonated, leading to C–OH bond cleavage and formation of water and the tetracarbonyl cationic complex [Mn(bpy-R)(CO)₄]⁺. The tetracarbonyl complex generates CO upon further reduction. There was a considerable discourse about whether the Mn–Mn dimer or the anionic species is the active species for CO₂ reduction. While the Mn⁰−Mn⁰ dimer has been shown to be catalytically active, photochemically and electrochemically reducing CO₂ to formic acid or CO,^33,81,82  the rates for these catalytic reactions are much slower than the electrocatalytic reduction of CO₂ to CO by the anionic [Mn(R-bpy)(CO)₃]⁻ species.^33,82  Furthermore, the experimental catalytic reaction order is first-order in catalyst, suggesting the catalytic intermediate should be mononuclear, i.e. not dimers.^33,84,85 

Electrocatalytic reduction of CO₂ to CO and water necessarily means a proton source must be present in the reaction medium, according to eqn (1.1). While select catalysts are capable of the reduction and disproportionation of CO₂ to CO and CO₃²⁻ under aprotic conditions (eqn (1.2)), the vast majority of CO₂ electrocatalysts require a Brønsted acid such as water, phenol, or TFE in order to drive the catalytic cycle to generate CO and water. In the absence of added acid, some catalysts are still active for CO₂ reduction according to eqn (1.1) and must therefore scavenge protons from the bulk solution.^8,86  Tetraalkylammonium electrolyte salts, MeCN (solvent), and/or trace water have been proposed as possible proton sources in these cases, however, the actual proton source when none is deliberately added remains poorly investigated.

The selectivity and activity of CO₂ reduction typically depend on the strength of the acid source. For example, while Re tricarbonyl catalysts such as [Re(tBu-bpy)(CO)₃Cl] (2) are capable of CO₂ reduction to CO in the absence of added proton source, it has been well established that weak Brønsted acids such as water increase the rate of catalysis due to more facile formation of the Re-COOH intermediate, the most stable species in the catalytic cycle.^76,86,87  CO₂ reduction is observed with quantitative FE_CO with no H₂ production upon addition of phenol or TFE (pK_a=29.1 and 35.4 in MeCN, respectively). The rate of CO₂ reduction is comparable to that in the absence of added proton source, but the catalytic potential is shifted positive by nearly 0.2 V. In the presence of acetic acid (pK_a=23.5 in MeCN^88,89 ), the selectivity decreased to FE_CO=35% with increased H₂ evolution (35%). As shown in Figure 1.9, a larger maximal current enhancement is observed for 2 with acetic acid compared to the weaker Brønsted acids; likely due to the additional contributions of competitive H₂ evolution to the current. Thus, it is critical to strike a balance between a Brønsted acid that is sufficiently acidic to rapidly generate stable hydroxycarbonyl species, but not too strong as to increase the favorability of H₂ evolution.

The acid pK_a dependence on selectivity can also vary with the electronic structure of the ligands: increasing the donating ability of the ligands leads to a more basic metal center, which will be more susceptible to protonation by weaker Brønsted acids. This trend is exemplified by comparing the reactivity of 2 to that of a related Re complex bearing a bidentate pyridyl-N-heterocyclic carbene ligand.⁹⁰  This carbene complex is active for electrocatalytic CO₂ reduction to CO at high FE_CO>90% with water present, similar to the behavior of 2. However, FE_CO with the carbene catalysts drops to 43% and 21% for TFE and phenol, respectively, and a significant amount of H₂ was also generated with phenol. CO selectivity remained high for [Re(tBu-bpy)(CO)₃Cl] under the same conditions with TFE and phenol.⁸⁶  The difference in CO selectivity is attributed to the increased donating ability of the carbene ligand compared to bipyridine and highlights the importance of optimizing the relative pK_as of the chosen acid and the reduced form of the electrocatalyst.

Decreased selectivity for CO₂ reduction has been observed for the Fe porphyrin catalyst FeTPP (3) with increasingly stronger acids. Increased H₂ evolution is observed for FeTPP when using stronger Brønsted acids such as triethylammonium which is well established.⁹¹  The Mn-bpy catalysts required the addition of weak Brønsted acids in order to achieve CO₂ reduction to CO, and also shows an increase in current density with increasing acid strength, but no selectivity change.⁸²  Additionally, it was found that [Mn(tBu-bpy)(CO)₃(MeCN)] (13) is more active than the Re version when using water as the proton source, but is less active when using methanol or TFE.

One aspect of proton source selection that is rarely considered is the resulting conjugate base generated upon deprotonation of the Brønsted acid. For the weak O-H acids typically used for electrocatalytic CO₂ reduction, the conjugate bases are alkoxide species, themselves capable of reacting with CO₂ to form alkyl carbonates. The trifluoroethoxide (generated from deprotonation of TFE) was proposed by Gray and co-workers to assist [Mn(tBu-bpy)(CO)₃]⁻ in electrocatalytic CO₂ reduction by providing an additional driving force for protonation of the [Mn-CO₂]⁻ intermediate and facilitating the dehydroxylation of the hydroxycarbonyl intermediate. DFT calculations suggest that this activity is due to the homoconjugation (trifluoroethoxide H-bonding dimer) and/or carboxylation of the conjugate base of this dimer.⁹²  While the formation of these alkoxide species from weak Brønsted acids may be unavoidable, the consequences of their presence in solution and their effects on the reaction thermodynamics must be considered. This is especially important in discussions of overpotential, as the thermodynamics of CO₂ reduction coupled to alkyl carbonate formation is clearly different than for an aqueous system.

Many of the literature reports state that water was chosen as a very weak Brønsted acid source for CO₂ reduction, having a pK_a(H₂O)≈38-40 in MeCN. However, upon saturation of the reaction solution with CO₂, carbonic acid will be formed to some degree according to eqn (1.3). Carbonic acid is relatively acidic in organic solvent (pK_a=7.37 in DMF), and therefore will very likely be the strongest Brønsted acid present in solution. This should be considered when evaluating the thermodynamics of CO₂ reduction in mixed organic/aqueous systems, as described by Savéant and co-workers.⁹³ 

Fe and Ru catalysts containing the redox- and proton-active cyclopentadienone ligand have been widely utilized for hydrogenation of various organic substrates, including CO₂. The tricarbonyl Fe complex, tricarbonyl(η⁴-2,5-bistrimethylsilylcyclopentadienone)iron (14), originally reported by Knölker,⁹⁴  exhibited an irreversible reduction at −1.4 V vs. NHE (ca. −2.0 V vs. Fc^+/0) followed by a smaller reversible feature at −1.5 V vs. NHE in MeCN under Ar.⁹⁵  The introduction of CO₂ resulted in a dramatic current enhancement at lower potentials in the absence of added Brønsted or Lewis acid sources (Figure 1.10). By CPE, CO is selectively produced (FE_CO=96%) with no H₂ evolution. Under these conditions, the overpotential for CO₂ reduction to CO is 0.85 V.

Water was produced during CPE in roughly equimolar quantities to CO. The formation of water suggests that catalysis does not proceed via CO₂ reduction and disproportionation, as is often observed in the absence of added acid. In order to generate water with no added Brønsted acid, the catalyst must scavenge protons from the solvent, residual water, or supporting electrolyte. However, as the catalyst turns over, more water will be generated which will be a better proton source for CO₂ reduction. The reaction will thus become catalytic in water and a continual increase in water production is not expected. Further studies into the proton source during catalysis may be warranted. An additional avenue of exploration is the influence of weak Brønsted acids on the rate and selectivity of CO₂ reduction which were not reported. The important role of the added proton source on CO₂ reduction with [Re(R-bpy)(CO)₃X] catalysts is discussed elsewhere in this section.⁸⁶ 

The hydride complex of tricarbonyl(η⁴-2,5-bistrimethylsilylcyclopentadienone)iron (15) is a key intermediate in the catalytic hydrogenation of other substrates. Surprisingly, exposure of 15 to CO₂ does not result in any reaction (Scheme 1.7a), indicating that this species is not involved in electrocatalytic CO₂ reduction. DFT calculations predict that CO₂ reduction proceeds via two-electron-one-proton reduction of 14 to the hydroxycyclopentadienyl complex (14b), from which loss of a CO ligand occurs readily followed by association of CO₂ (Scheme 1.7b). The Fe-CO₂ adduct of tricarbonyl(η⁴-2,5-bistrimethylsilylcyclopentadienone)iron (14c) is stabilized by hydrogen bonding between the carboxylate and hydroxyl group. Similar behavior was noted for FeTPP catalysts bearing O–H groups on the ligand.⁹³  This hydrogen bond also facilitates cleavage of the carboxylate C–O bond and lowers the barrier for water loss. Thus, the ability to reduce the cyclopentadienone ligand and the optimal positioning of the resulting hydroxyl group are both critical for the electrocatalytic activity of 14.

Electrocatalytic reduction of CO₂ has been extensively explored over the past three decades with Ru(ii) complexes bearing polypyridyl ligands. Since reduction of Ru(ii) is generally inaccessible at potentials relevant to CO₂ reduction, the polypyridyl ligands play a critical role in catalysis by storing reducing equivalents. In a seminal report, Tanaka and co-workers⁹⁶  showed that [Ru(bpy)₂(CO)₂]²⁺ (16) is an active electrocatalyst for CO₂ reduction. In 1 : 1 DMF : H₂O, CO₂ reduction is observed at −1.40 V vs. SCE (ca. −1.78 V vs. Fc^+/0), producing CO and formate with only trace H₂. Notably, a Brønsted acid such as water is required for catalysis, and the solution pH dictates the major product. Following two-electron reduction of 16 and loss of one CO ligand, a [Ru(bpy)₂(CO₂⁻)]⁺ adduct is generated under CO₂ that either favors CO or formate production under acidic or basic conditions, respectively.

A closely related catalyst [Ru(tpy)(bpy)(CO)]²⁺ (17) was later reported by Tanaka and co-workers to also perform electrocatalytic CO₂ reduction.⁹⁷  In 20% DMF/H₂O, CO and formate are both produced via CPE at −1.6 V vs. Ag/Ag⁺ (ca. −1.73 V vs. Fc^+/0), along with trace methanol. Varying the solvent and temperature has a significant effect on the products, demonstrating the importance of solvent effects. In 80% EtOH/H₂O at −20 °C, a range of products were detected including formaldehyde, glyoxylic acid, glycolic acid, and methanol.⁹⁷ 

Considering the importance of the proton source for electrocatalytic CO₂ reduction, it is not surprising that significant efforts have been undertaken to incorporate pendent acid or base groups as proton shuttles into known molecular catalysts. Enzyme active sites are highly optimized to shuttle multiple protons and electrons to and from the substrate as well as shuttle the product away from the active site; hence, many bio-inspired molecular catalysts have been developed with pendent proton shuttles to mimic enzymatic behavior and enhance catalytic performance.⁹⁸  This has been especially well utilized for H₂ evolution catalysts but is also becoming more commonly integrated into CO₂ reduction catalysts.

A particularly successful example of using pendent acid groups to improve CO₂ reduction was reported by Savéant and co-workers.^93,99  As discussed in Section 1.4, the addition of weak Brønsted acids such as phenol to the reaction solution enhances the rate of CO₂ reduction by Fe porphyrin catalysts including FeTPP (3).^40,73  The installation of phenol substituents at each ortho position on the phenyl rings in iron 5,10,15,20-tetrakis(2′,6′-dihydroxylphenyl)porphyrin (FeTDHPP) (18) resulted in an extremely fast rate of CO₂ reduction in DMF with 2 M water, TOF=1.6×10⁶ s⁻¹ (Figure 1.11).⁹³  This rate is equivalent to the activity expected for the binary catalyst system (i.e. 3 and phenol) at a phenol concentration of 150 M.⁹³  Notably, it is not physically possible to run this reaction at this phenol concentration. Later studies revealed that this rate enhancement is not only due to the high local proton concentration near the reacting center but also due to stabilization of the [Fe-CO₂]²⁻ intermediate by intramolecular hydrogen bonding between the carboxylate and the pendent O–H groups (Figure 1.11).^99,100 

Since this report, the activities of other CO₂ reduction catalysts have been enhanced in a similar fashion. Marinescu and co-workers investigated Co aminopyridine macrocycles for electrocatalytic CO₂ reduction.^101,102  Complexes containing alkyl substitutions on the pendent amine groups were compared to cobalt(ii) 2,4,6,8-tetraaza-1,3,5,7(2,6)-tetrapyridinacyclooctaphane 19 (Figure 1.12). The TON for 19 was 300 times higher than that for the methyl or allyl substituted analogous Co macrocycles. Furthermore, 19 was highly selective for CO (FE_CO=98%), and the presence of the N–H moieties positively shifted the reduction potential. In a computational study of Co porphyrin, Koper and co-workers¹⁰³  found that the mechanism of the formation of the neutral hydroxycarbonyl adduct [Co–CO₂H] changes from concerted proton-coupled electron transfer (CPET) to mixed CPET and sequential proton-electron transfer (SPET) around pH 3.5, the pK_a of this intermediate. This transition mirrors the observed increase in CO production upon increasing the pH from 1−3, indicating the importance of accessing the [Co-CO₂]⁻ adduct via an SPET pathway in order to favor CO₂ reduction over proton reduction. For [Mn(R-bpy)(CO)₃Br]-type complexes, the introduction of a phenol substituent in close proximity to the metal center provided a local Brønsted acid source for enhanced proton transfer to the [Mn-CO₂]⁻ adduct intermediate, which enabled electrocatalytic CO₂ reduction to CO to proceed even in anhydrous MeCN.^104,105  Analogous complexes without the pendent acid group showed no catalytic response by cyclic voltammetry under identical aprotic conditions. Furthermore, a seven-fold increase in the rate of CO₂ reduction was observed upon addition of water as a proton source, as compared to the activity of the parent complex.¹⁰⁶  A series of detailed investigations into hydrogen bonding interactions for [Re(R-bpy)(CO)₃X] catalysts containing 4,4′-substituted bipyridine ligands with methylacetamidomethyl (dacbpy)¹⁰⁷  or tyrosine groups (Tyrdac-tBu) have also been reported.^108,109  It was found that dacbpy promotes the formation of a hydrogen-bonded dimer, enabling access to an alternate bimolecular mechanism for CO₂ reduction (see Section 1.8).¹⁰⁷  The phenol group of Tyrdac-tBu similarly participates in the structural assembly of a bimetallic active species, and can additionally function as a local proton source for catalysis.^108,109 

While the rather weakly acidic phenol group seems to be the pendent acid of choice for early transition metal catalysts that primarily produce CO, a stronger protic substituent is required for later transition metal complexes that produce formate. The prototypical example is the bidentate phosphine P₂N₂ ligands containing pendent amine groups, which have already been widely exploited as ligands for H₂ evolution catalysts.¹¹  Hydride complexes of [M(P₂N₂)₂]²⁺ systems, where M=Group 10 transition metals, are generally not sufficiently hydridic to react with CO₂.^110,111  Artero and co-workers recently demonstrated that a series of complexes [CpCo(P^R₂N^R′₂)I]⁺ (20) are active electrocatalysts for CO₂ reduction to formate.⁵⁹  These complexes undergo two sequential one-electron reductions to Co(i) near −0.9 and −1.2 V vs. Fc^+/0 in DMF under Ar. A large current enhancement is observed around −2 V vs. Fc^+/0, upon addition of CO₂ and various concentrations of water, which is significantly more negative than the Co(ii)/Co(i) couple. While all three complexes were active catalysts, an impressive maximum TOF was calculated for the fastest system [CpCo(P^Cy₂N^Bn₂)I]⁺, where TOF>1000 s⁻¹. CPE studies confirmed that formate is produced with excellent FE (up to 98%) at −2.05 to −2.25 V vs. Fc^+/0, corresponding to an overpotential of approximately 0.4−0.6 V. Notably, the related complex [CpCo(dppp)I]I (dppp=diphenylphosphinopropane) exhibited a very small current increase under identical conditions. The proposed catalytic mechanism involves a net three-electron-one-proton reduction of the Co(iii) starting complex to generate a Co(ii)-hydride intermediate, which is a strong hydride donor and is thermodynamically capable of hydride transfer to CO₂ (Scheme 1.8). Although some questions regarding the exact details of the mechanism remain unanswered at this stage, it is evident that the pendent amine plays a crucial role, not only as a proton shuttle but also to stabilize the hydride transfer intermediates. However, we note that the pendent amine group as well as the identity of the amine substituent both have a significant influence on the electronic properties of the catalyst, which makes it difficult to independently parse out the influence of the protic functionality of the amine group on catalysis.

A tetradentate Co complex with an N-H group directly bound to the Co center was examined for CO₂ reduction by the Peters’ group.¹¹²  In MeCN with water as a proton source, CO₂ reduction to CO was observed at −1.88 V vs. Fc^+/0 with an FE of 45%. Simultaneously, H₂ evolution (30%) was observed under these conditions. The CO₂ adduct is stabilized through hydrogen bonding to one of the protons of a coordinated amine, significantly lowering the energy of the transition state.¹¹³  The identity of the metal was shown to affect selectivity with a similar pentadentate macrocylic ligand. When Co(ii) was used in electrochemical CO₂ reduction, CO production dominated, but when Fe(ii) was used, formic acid selectivity was observed at low overpotentials. It is unclear if the N-H group's interaction with the metal plays a role in this selectivity difference and further mechanistic studies are required.¹¹⁴ 

Inspired by Savéant, the Nocera and Chang groups have recently examined so-called Fe hangman porphyrins.^115–117  Fe hangman porphyrins containing phenol, guanidine, and sulfonic acid groups electrocatalytically reduced CO₂ to CO between −2.1 and −2.2 V vs. Fc^+/0 with FE_CO>93%. DFT calculations suggest that the CO₂ adduct is stabilized by the intramolecular binding with the pendent group. In this series of results, the complex with the strongest CO₂ interaction (i.e. phenol hangman) exhibited the highest apparent rate constant for CO₂ reduction. Notably, phenol was the only Brønsted acid source used in this study. The effect of varying acid strength with the different pendent groups may be worth exploring. The proximity of the proton shuttle to the metal center was examined by the Chang group. The activity of FeTPP catalysts with an amide at different positions was compared.¹¹⁶  The ortho-amide pendent groups both significantly enhanced the rate of catalysis, with the distally placed amide exhibiting greater enhancement (Figure 1.13).

The nature of the pendent acid must be carefully selected for the particular catalyst system of interest. If the pK_a of the pendent group and the metal center are not appropriately matched, the activity and/or selectivity for CO₂ reduction over H₂ evolution can decrease substantially.¹¹⁰  This consideration is especially important for catalysts that produce formate since a metal hydride species is typically the branching point between CO₂ reduction to formate versus H⁺ reduction to H₂.¹¹¹  The stability of pendent groups should also be considered. Fujita and co-workers attempted to use redox-active phenol ligands and observed a tendency to deprotonate upon reduction and react with CO₂.^118,119  Das and co-workers used the redox activity of a pendent carboxylate to their advantage. The labile pendent carboxylate acts to stabilize the metal center, provides a local proton source, and provides a binding site for CO₂.¹²⁰ 

There are other consequential factors to consider when pendent proton shuttles are installed close to the catalytic site. These substituents can significantly alter the electronic properties of the ligand and can exert significant electronic and Coulombic influence on the metal center. Thus, the presence of these groups can dramatically change the catalyst activity due to these other second coordination sphere effects, either in addition to or instead of their protic functionality. Therefore, control experiments with complexes bearing inactive substituents having similar electronic and steric properties are highly advised in order to confirm that no other competing effects from the second coordination sphere functional groups are operative. For example, addition of a carboxylic acid group to the cyclam ligand in [Ni(cyclam)]²⁺ (9) resulted in an increased reaction rate and improved selectivity for aqueous CO₂ reduction at low pH (2–5).¹²¹  The authors propose that these improvements may be due to the availability of a local proton source; however, electronic effects from the electron-withdrawing carboxylic acid were not specifically ruled out with control experiments.

The pendent group may also have other functions, such as facilitation of CO dissociation.⁸⁰  Compain and co-workers synthesized dicarbonyl and tricarbonyl Mn-terpyridine (tpy) complexes and examined their CO-releasing properties. The tpy ligand exhibits bidentate binding in the tricarbonyl Mn complex and switches to tridentate upon externally triggered CO-release to generate the dicarbonyl Mn complex. It has also been shown that the tridentate tpy coordinated dicarbonyl Mn complex was active for CO₂ reduction but degrades faster than the Mn-bpy catalysts.¹²²  A bipyridine with a pendent NAD-like structure (Nicotinamide adenine dinucleotide) was proposed to facilitate hydride transfer to metal-bound CO₂ to explain the observed enhanced selectivity for formate.¹²³  The pendent groups function as more than proton shuttles in each of these cases and further studies should critically analyze the chemical behavior of these systems. A pendent thiourea group in the second coordination sphere of the Re-bpy catalyst acts as a hydrogen bond promoter and proton donor.¹²⁴  DFT calculations indicated an N–H bond of the thiourea group stabilizes the Re–carboxylate and acts as a proton source to form the hydroxycarbonyl species. Pendent groups act as more than local proton sources and their influence on the CO₂ reduction reaction mechanism should be evaluated holistically.

A very different trend is observed using charged substituents to effect through-space Coulombic interactions, as demonstrated by Savéant and co-workers.¹²⁵  The incorporation of trimethylammonium groups at the para positions in iron tetraphenylporphyrin (21) shifted the reduction potentials in a more positive direction, decreasing the overpotential from 0.6 V for FeTPP (3) to 0.4 V (Figure 1.14). Most notably, however, the catalytic rate was not diminished compared to 3. The reduced basicity of the Fe center in 21 is now counterbalanced by increased stabilization of the [Fe-CO₂]²⁻ adduct through coulombic interactions between the positively-charged trimethylammonium groups and the negatively-charged oxygens of the carboxylate. This effect was even more pronounced in the ortho derivative (22): closer proximity of the charged substituents led to stronger coulombic attraction and greater stabilization of [Fe-CO₂]²⁻ resulting in a nearly two-fold increase in rate and decrease in overpotential to 0.2 V (Figure 1.14). As expected, the opposite behavior was observed for the sulfonate-substituted version (23): the anionic substituents shifted the catalytic operating potential to a more negative value and decreased the rate of CO₂ reduction. This approach of using through-space coulombic effects to move beyond typical through-bond scaling relationships could be applied to other catalyst systems involving intermediates with a localized charge similar to the [Fe-CO₂]²⁻ adduct.

The para-substituted catalyst 21 was originally developed as a water-soluble catalyst. Electrocatalytic CO₂ reduction to CO in neutral water (pH 6.7) was achieved using 21, where the cationic trimethylammonium groups enable aqueous solubility.¹²⁷  Under these conditions, an extremely large maximum TOF of 10⁷ s⁻¹ was estimated, corresponding to a second-order rate constant of k=2.5×10⁸ M⁻¹ s⁻¹. Only trace amounts of H₂ were detected by CPE at −0.97 V vs. NHE, indicating a significant preference of CO₂ reduction over H₂ evolution. One explanation is that the [Fe-CO₂]²⁻ intermediate is stabilized via hydrogen bonding with water and via Coulombic attractions with the cationic trimethylammonium groups. Further mechanistic studies are required to confirm the origins of the impressive rate and selectivity of 21 for CO₂ reduction in water.

The mechanism of CO₂ reduction with [Ni(cyclam)]²⁺ (9) was initially proposed by Sauvage and co-workers (Scheme 1.9).⁶⁰  DFT calculations on the homogeneous activity of 9 by Ye and co-workers¹²⁸  further supported this proposal: following reduction to Ni(i), generation of a Ni-η¹-CO₂ carboxylate adduct that leads to CO formation is energetically favored by 14 kcal mol⁻¹ over the Ni-η¹-OCO complex, which leads to formate production. Thus, this initial binding event determines the observed product selectivity. Further reduction of [Ni(iii)-CO₂]⁺ to the Ni-hydroxycarbonyl likely occurs via a PCET pathway, followed by C–O cleavage and loss of CO.

The rate of CO₂ reduction with [Ni(cyclam)]²⁺ does not exhibit a linear increase with catalyst concentration and catalysis occurs approximately 0.3 V more positive than the reversible Ni(ii/i) couple, both suggesting the involvement of adsorbed species.⁶⁰  Anson and co-workers later showed that 9 is only weakly adsorbed at potentials relevant to catalysis, but the one-electron reduced species [Ni(cyclam)]⁺ is strongly adsorbed and plays a crucial role in CO₂ reduction.¹²⁹  Computational studies support the role of the Hg as a promoter by maintaining noncovalent dispersive interactions with the cyclam N-H groups.¹³⁰  These interactions favor flattening of the cyclam ring and destabilize the Ni(i)–CO form of the catalyst, enabling CO release and suppressing catalyst degradation to Ni(0) carbonyl species. Further investigations by Fujita and co-workers highlighted the important role of the cyclam conformation. For example, C-RRSS-[Ni(HTIM)]²⁺ (2,3,9,10-tetramethyl-1,4,8,11-tetraazacyclotetradecane) (24) is a better catalyst than 9 due to the nearly flat geometry of the ligand enforced by the methyl substituents, which enables closer approach of the catalyst to the Hg surface (Figure 1.15).¹³¹  The C-RSSR-isomer 25, on the other hand, exhibited significantly diminished catalytic activity due to the steric bulk of the axial methyl groups that prohibited close surface interactions.

While 9 is one of the best catalysts for CO₂ reduction at Hg electrodes, it is also active at GC and other electrode materials, albeit at much lower rates.^130,132,133  The Kubiak group¹³²  demonstrated that 9 is highly selective for CO (FE_CO=90%) with no H₂ evolution at GC at −1.30 V vs. NHE, but the TON (=4) and TOF (=90 s⁻¹) were both significantly lower than at Hg. DFT calculations predict that CO₂ binding to the metal is more favorable in the SSSS-isomer of the freely-diffusing catalyst, perhaps due to hydrogen bonding with the N–H groups. In the absence of surface interactions with Hg to stimulate the loss of CO, [Ni(cyclam)(CO)]⁺ accumulates in solution and undergoes decomposition to [Ni(CO)₄], limiting the catalyst lifetime at the GCE. The addition of [Ni(TMC)]²⁺ (TMC=1,4,8,11-tetramethyl-1,4,8,11-tetraazacyclotetradecane), which itself is not a CO₂ reduction catalyst at GC but has a strong affinity for CO after reduction to [Ni(TMC)]⁺, dramatically increases the rate of CO₂ reduction with 9 by scavenging CO and inhibiting catalyst deactivation.¹³³  As shown in Figure 1.16, the catalytic current density is increased by nearly 40 times in the presence of 20 equivalents of [Ni(TMC)]²⁺.

In the Mn-bpy class of electrocatalysts, Mn–Mn dimer formation negatively shifts the reduction potential of the second reduction, and different strategies have been adopted to avoid dimerization. Steric bulkiness is a proven way to avoid dimerization.^84,134,135  A bulky 6,6′-dimesityl-2,2′-bipyridine (mesbpy) ligand was used to make Mn(mesbpy)(CO)₃Br (28) and [Mn(mesbpy)(CO)₃(MeCN)](OTf) (29).³³  These Mn complexes exhibit a single, two-electron reduction wave under a nitrogen atmosphere with no dimerization (Figure 1.17). IR-SEC and chemical reductions with KC₈ further confirmed the formation of both the singly reduced and doubly reduced Mn complexes at the same potential, suggesting that elimination of dimerization lowers the second reduction potential. The active species [Mn(mesbpy)(CO)₃]⁻ binds CO₂ at −1.6 V vs Fc^+/0, but catalysis does not occur until more negative potentials are applied. IR-SEC experiments under CO₂/H⁺ indicate that reduction of a Mn(i)−C(O)OH catalytic intermediate may determine the unusual “over-reduction” required to initiate catalysis. Moreover, replacing the axial bromide for a pseudohalogen (CN) in Mn(bpy)(CO)₃(CN) can also avoid dimerization.¹³⁶  IR-SEC and cyclic voltammetry indicate a disproportionation mechanism of two one-electron-reduced species, generating the catalytically active species.

Similar to the modifications to the Re catalyst, many nitrogen-containing aromatic ligands can be used to replace bipyridines. N-heterocyclic carbene (NHC) ligands are popular choices since they are versatile for different kinds of modifications.^136–138  However, their enhanced σ-donor character compared to bipyridines results in an increased HOMO–LUMO gap and hence a cathodic shift (more negative) of the one- and two-electron reduction potentials.¹³⁷  Replacing the Br⁻ ligand with NCS⁻ and CN⁻ in Mn-NHC compounds shows similar negative impacts on reduction potentials.¹³⁶  On the other hand, the increased π acidity of pyrimidine shifts the two-electron reduction to −1.77 V vs Fc^+/0, 70 mV more positive than that for the Mn-bpy catalyst.¹³⁸  Similar positive shifts can be achieved by extending the π network of imidazole ligands.¹³⁶  Extension of the bipyridine ligand to a phenylazopyridine positively shifts the reduction potential to −0.93 V vs. Fc^+/0.¹³⁹  Nonetheless, when bipyridines are replaced with other nitrogen-containing ligands, the catalytic performance is usually lower than the bulky Mn(mesbpy)(CO)₃Br catalysts, because these catalysts do not solve the problem of dimerization and may exhibit stability issues during catalysis.^{80,122,135,140,141} 

The lability of the monodentate ligand in [Ru(tpy)(bpy)X]ⁿ⁺ (5) can also be increased by altering the steric properties of bipyridine. Ott and co-workers¹⁴²  found that introducing an ortho-methyl on bipyridine significantly improves the rate of MeCN dissociation in (30) compared to the 4,4′-dimethyl version (31) (Scheme 1.10). The steric bulk of the methyl group causes bipyridine to tilt towards the MeCN ligand and promotes dissociation.⁴⁰  Both 30 and 31 are active for CO₂ reduction to CO below −2 V vs. Fc^+/0 by the typical ECE pathway. However, for 24, catalysis is also observed at the first reduction potential (−1.82 V vs. Fc^+/0). The increased lability of MeCN enables substitution by CO₂ to occur after only one-electron reduction, forming [Ru(tpy)(mbpy)(CO₂˙⁻)]⁺ to which the addition of the second electron is facile and generates the key [Ru(ii)-CO₂²⁻] species (Scheme 1.10). While the TOF at the first potential is low (1.14 s⁻¹), the ability to access this ECE mechanism with 30 lowers the overpotential by 0.4 V without any loss in FE_CO.

The stability and reaction rates of Fe(iii) tetraphenyl porphyrin FeTPP (3) are improved by the addition of Lewis acidic cations such as Mg²⁺ (Figure 1.18), with CO and formate produced in FE´s of 60% and 30%, respectively.^39,40  Although the addition of Lewis acid cations dramatically improved the catalysis, carbonate is also formed, which precipitates as MgCO₃, and quickly passivates the electrode.

The mechanism of CO₂ reduction by Fe porphyrins has been studied in detail by Savéant and others and is shown in Scheme 1.11.⁴⁰  Following reductions of FeTPP (3), nucleophilic attack of CO₂ by [FeTPP]²⁻ generates an [Fe−CO₂]²⁻ adduct. This adduct can be stabilized by ion pairing between the carboxylate group and Mg²⁺, which weakens the C–O bonds by “pulling” an electron pair out of CO₂ and facilitates formation of an [Fe(ii)–CO] species. Alternatively, weak Brønsted acids can stabilize the carboxylate adduct via hydrogen bonding and similarly enable C–O cleavage. This behavior is termed as a two-electron “push-pull” mechanism, where an electron pair is pushed from the catalyst ([FeTPP]²⁻) into the substrate (CO₂) and then pulled out of [Fe−CO₂]⁻ with the help of an electrophile (Brønsted or Lewis acid). The electronic structure of [FeTPP]²⁻ remains under debate. A recent study by Neese and co-workers¹⁴³  suggested that [FeTPP]²⁻ is an intermediate-spin Fe(ii) center antiferromagnetically coupled to a porphyrin diradical dianion. Savéant and co-workers proposed an Fe(0) description due to its reactivity toward alkyl halides.¹⁴⁴  This question of non-innocent ligand contributions often arises in electrocatalysis. Highly reduced catalytic intermediates containing redox-active ligands such as [FeTPP]²⁻ can have the charge distributed over both metal-based and ligand-based orbitals.

Despite this important advance being initially reported in 1991, it would be more than two decades later before the use of Lewis acids to promote CO₂ electroreduction with a different catalyst system was published. Inspired by this early publication, the authors’ group examined the CO₂ reduction activity of [Mn(mesbpy)(CO)₃(MeCN)] (29) in the presence of Lewis acids.¹³⁴  Generation of the [Mn(i)-COOH] hydroxycarbonyl species occurs at the two-electron reduction of 29, under CO₂ with weak Brønsted acids such as TFE. The [Mn(i)-hydroxycarbonyl] species must be further reduced in order to access the “fast catalysis” regime (Figure 1.19).¹³⁴  However, an alternate mechanism for CO₂ reduction is likely operative in the “slow catalysis” regime. DFT calculations suggest that additional protonation of [Mn(i)-hydroxycarbonyl] leads to C–O bond cleavage, analogous to the Brønsted acid-assisted catalysis with FeTPP (3) (Scheme 1.11). Clearly this route with 29 is comparatively slow, and using stronger acids to increase the rate resulted in H₂ evolution.¹⁴⁵  It was shown that the Lewis acidic Mg²⁺ could be used in place of a Brønsted acid, triggering CO₂ reduction at a significantly lower overpotential (roughly 0.3−0.45 V) with high selectivity for CO generation (FE_CO=96%).¹³⁴  However, similar to FeTPP (3) in the presence of Mg²⁺,⁵  the stoichiometric use of Mg²⁺ and the formation of insoluble MgCO₃ that quickly blocks the electrode are major drawbacks of this approach. To overcome this issue, the Kubiak group later developed an improved catalytic system using [Zn(cyclam)]²⁺ as a soluble Lewis acid cation in conjunction with 29.¹⁴⁶  Here, the equatorial coordination of the ligand did not block the Lewis acid functionality of the Zn²⁺ center but the cyclam ligand prevents precipitation of Zn carbonate during catalysis. Thus, [Zn(cyclam)]²⁺ could be used in co-catalytic amounts as opposed to a sacrificial additive like Mg²⁺. The TOF for 29 in the presence of [Zn(cyclam)]²⁺ (30 mM) was estimated to be 105 s⁻¹, five times greater than that observed in the presence of Mg²⁺ (120 mM, TOF=20 s⁻¹) suggesting that [Zn(cyclam)]²⁺ is indeed functioning as a co-catalyst. Importantly, high selectivity for CO was maintained with [Zn(cyclam)]²⁺ (FE_CO=82% at −1.6 V vs. Fc^+/0).

Several studies have appeared in recent years exploring the influence of ionic liquids as additives or as the reaction medium for CO₂ reduction with molecular catalysts. The addition of an ionic liquid, 1-butyl-3-methylimidazolium tetrafluoroborate (BMImBF₄), as an additive (0.3 M) to FeTPP (3) with trifluoroethanol (1 M) in DMF solution resulted in a lower overpotential (η) for electrocatalytic CO₂ reduction, η=0.67 V compared to 0.82 V for 3 in the absence of the ionic liquid additive.¹⁴⁷  In this case, the positively charged ionic liquid interacts with negatively charged reduced porphyrin, acting as a co-catalyst in the CO₂ reduction to CO by FeTPP. These interactions lead to the positive shift in the reduction potential of Fe^I to Fe⁰ and an overall decreased overpotential leading to increased current densities. Notably, a four-fold increase in TOF was observed under these conditions, compared to the parent catalyst without an ionic liquid. Furthermore, the selectivity for CO production over H₂ evolution remained high (FE_CO=93%). Electrocatalytic reduction of CO₂ to CO by Re-bpy in the presence of the ionic liquid 1-ethyl-3-methylimidazolium tetracyanoborate ([emim][TCB]) was described by Grills and co-workers.¹⁴⁸  The [emim][TCB] was used as a solvent and an electrolyte and decreased the overpotential for CO₂ reduction to CO by 0.45 V, with much faster catalytic rates than in MeCN. The rapid dissociation of a chloride ligand occurred at the second electron-reduction potential of fac(Re-(bpy)(CO)₃Cl in the presence of [emim]⁺ in comparison to neat MeCN. This resulted in a lower activation energy and an order of magnitude increased apparent catalytic rate constant for CO₂ reduction to CO. Notably, CO₂ is less soluble in [emim][TCB] than MeCN/H₂O (0.13 M atm⁻¹ vs. 0.26 M atm⁻¹). Thermodynamic and mechanistic studies of CO₂ reduction by fac(Re-(bpy)(CO)₃Cl in the presence of imidazolium-based ionic liquids were reported by the same group.¹⁴⁹  The proposed mechanism was supported by DFT calculations, as depicted in Scheme 1.12. The imidazolium cation interacts with doubly reduced [Re(bpy)(CO)₃]⁻, followed by protonation and CO₂ binding to the reduced metal center to form a metal–carboxylic acid that possesses more positive reduction potential of the species than in conventional electrolytes. Similarly, the effect of positively charged ammonium groups of a norbornenyl polymer with attached Re(bpy)(CO)₃Cl was demonstrated by Kubiak and Gianneschi.¹⁵⁰  The authors found that positively charged polymers with quaternary ammonium salts exhibit a 300 mV more positive shift in comparison to the neutral polymers, while the negatively charged polymers displayed a negative shift in potential with no reactivity towards CO₂. This observation is analogous to FeTPP catalysts with charged NMe₃ groups.¹²⁵ 

In a slightly different approach, Grubbs, Gray, and co-workers developed brush polymer ion gels for electrocatalytic CO₂ reduction to CO.¹⁵¹  High selectivity for CO (FE_CO=90%) and a potential shift of 450 mV was observed in this case in comparison to the conventional organic solvents, which can also be attributed to the cation effect of a charged ionic liquid as described in a previous paper. Nippe and co-workers developed and studied pendent imidazolium Re-bpy catalysts.^152,153  The synthetic design for these catalysts was inspired by nature. The protonated imidazolium and amine groups of histidine and lysine residues of Ni, Fe-carbon monoxide dehydrogenases stabilize metal carboxylate and hydroxycarbonyl intermediates through hydrogen bonding interactions. The installation of an imidazolium functional group to the Mn-bpy catalyst sought to mimic the hydrogen bonding behavior of amino acid residues in enzymes. The Mn–Me(ImMe_x)bpy catalysts reduce CO₂ at −1.5 V vs. Fc^+/0 with FE_CO=77%. This corresponds to reduction potentials that are 100 mV more positive than the Mn-mesbpy.¹⁵³  Electrochemical CO₂ reduction by similar imidazolium functionalized Re-bpy complexes also occurs at significantly more positive potentials (∼200 mV) compared to the Re-bpy catalyst.¹⁵² 

Nature uses multinuclear ensembles to enable multi-electron transfer processes. In one example, the metalloenzyme carbon monoxide dehydrogenase (CODH) contains a key dinuclear Ni-Fe complex in the active site that cooperatively facilitates electrochemical CO₂ reduction to CO at a low overpotential (η<0.1 V).¹⁵⁴  Thus, the bio-inspired incorporation of multiple reactive metal centers into a single molecular catalyst has attracted some attention in recent years in the context of electrocatalytic CO₂ reduction.

A series of co-facial Fe porphyrin dimers have been developed by Naruta and co-workers.^155,156  In their first report, two Fe triphenylporphyrin fragments were connected via an ortho- or meta-substituted benzene ring (32 and 33 in Figure 1.20, respectively).¹⁵⁵  The ortho version 32 outperformed 33, as well as the mononuclear catalyst FeTPP (3), displaying high FE_CO (=95%) and an impressive turnover frequency (TOF=4300 s⁻¹) at a moderate overpotential (η=0.66 V) in 10% H₂O/DMF. For the related Mn–Mn analogue, the two metals are held apart by a distance of approximately 3.7–6.2 Å.^157–159  The authors propose that a similar Fe–Fe distance exists in 27, which creates a molecular pocket that may be occupied by CO₂. The arrangement of the two Fe centers appears to enable an intramolecular “push-pull” mechanism to occur. In the reduced state, one Fe center may act as a Lewis base to push an electron pair onto CO₂, while the second Fe center may act as a Lewis acid to promote C–O bond cleavage. The same group later studied the effects of electron-withdrawing or donating substituents on the phenyl rings of these bis-porphyrin structures.¹⁵⁶  As observed with mononuclear Fe porphyrins, electron-withdrawing groups lowered the catalytic overpotential by positively shifting the reduction potentials of the complex; however, the decreased electron density at the Fe center resulted in a lower TOF for CO production. Electron-donating groups had the opposite effects. Further investigations should explore tuning of hetero-dinuclear complexes to overcome this conflicting scaling relationship between overpotential and rate.

Machan et al. used two hydrogen bonding acetoamido (dac) functionalized Re and Mn-bpy electrocatalysts (Re(dacbpy)(CO)₃Cl (34) and Mn(dacbpy)(CO)₃Br (35)) to make a supramolecular assembly to study the mechanism of CO₂ reduction in a heterobimetallic system.¹⁶⁰  Electrochemical studies showed that the redox features of the co-catalyst system are different from the overlay of redox features of the individual catalysts (Figure 1.21). Additional cyclic voltammetric experiments in DMF indicated that metal–metal bond formation occurs for the co-catalyst mixture under conditions where the respective homobimetallic analogues (either 34 or 35) are not generated, which is suggestive of a favorable heterobimetallic interaction at reducing potentials (Scheme 1.13). The increased catalytic current response of the co-catalyst mixture is also consistent with a cooperative effect. Control experiments with the co-catalyst mixture in DMF and with an equimolar mixture of Re(CH₃-bpy) and Mn(CH₃-bpy) in MeCN did not show an increased current response. It was concluded that Re(dacbpy)(CO)₃Cl operates via a similar bimolecular mechanism based on DFT calculations and IR-SEC observations.¹⁰⁸  A cooperative effect was also observed for a peptide modified Re-bpy catalyst which was designed to promote hydrogen bonding between complexes, and homo-dimer formation.¹⁰⁹ 

Homogeneous transition metal catalysts for electrochemical CO₂ reduction primarily produce CO and/or formate, and only in very rare cases are further reduction products observed. Bouwman and co-workers reported a Cu(i) thiolate catalyst, inspired by Ni-containing superoxide dismutase, that exists as a dinuclear dimer (36) bridging through a disulfide bond in the solid state and in solution.¹⁶¹  Upon exposure to air, 36 preferentially captures CO₂ rather than O₂, resulting in oxidation of Cu(i) to Cu(ii) and concomitant formation of oxalate (Scheme 1.14). A tetranuclear species (36b) is formed from this reaction, where the Cu(ii) centers are bridged by oxalate anions. Elimination of the product oxalate was achieved chemically by treatment with hydrochloric acid to generate oxalic acid, or electrochemically by CPE in MeCN at −0.03 V vs. NHE (ca. −0.66 V vs. Fc^+/0). Under a CO₂ atmosphere, oxalate is formed in a near quantitative FE (96%), demonstrating a successful electrocatalytic system for CO₂ reduction to oxalate. However, the addition of lithium perchlorate is critical to the activity of this system. Precipitation of lithium oxalate drives product elimination from 36b and enables catalyst turnover. At the same time, the crystallization of lithium oxalate at the electrode gradually passivates the electrode. Maverick and co-workers¹⁶²  later used a similar approach to generate oxalate from CO₂ with a dinuclear Cu(i) complex (37) (Figure 1.22), but no electrocatalytic studies with this system have been reported. The reactivity of 36 and 37 differs from that of previously reported dinuclear Cu complexes, which favor CO₂ reductive disproportionation to CO and CO₃²⁻.¹⁶³ 

It should be noted that the use of a binuclear complex does not guarantee cooperative behavior between metal centers. For example, a dinuclear version of [Ru(tpy)(bpy)(MeCN)]²⁺ (5) was developed by Oshio and co-workers, where two Ru complexes were connected via a long alkyl bridge.¹⁶⁴  Due to the flexibility of the bridging ligand, the catalytic behavior of the dinuclear species was essentially identical to that of the parent mononuclear complex (5). Indeed, careful design of the ligand structure is critical in order to promote close approach between the metal centers in the correct geometry for catalysis, either through rigid bridging structures, as in the co-facial Fe porphyrin dimer (32), or through soft hydrogen bonding interactions as in Re(dacbpy)(CO)₃Cl (34). Jurss and co-workers used rigid anthracene bridged Re(bpy)(CO)₃Cl to show that the orientation of the active Re sites dictates the reaction mechanism.¹⁶⁵  When the Re active sites adopt a cis confirmation a cooperative bimetallic reaction pathway is observed in the electrochemical studies. When the Re active sites adopt a trans confirmation a predominantly monometallic reaction pathway is observed. Constrictive ligand structures have also been utilized in the design of bimetallic Ni and Co complexes.^166,167 

The introduction of different ligand substituents in FeTPP provides a means to tune the catalyst properties in order to improve the overpotential and rate of CO₂ reduction. Aromatic substituents (biphenyl, pyrene, and phenyl–pyrene) on the porphyrin periphery had a negligible effect on the three reduction potentials of the complexes under Ar, but resulted in enhanced catalytic rates for CO production compared to (3).¹⁶⁸  The authors suggested that this rate improvement may be due to local accumulation of CO₂ in the hydrophobic space created by the π-conjugated groups; however, no evidence was presented to support this proposal, and this hypothesis seems unlikely given the known properties of CO₂. Savéant and co-workers showed that replacement of the phenyl groups in 3 with electron-withdrawing perfluorophenyl groups in FeF5TPP, FeF10TPP, and FeF20TPP (38−40, Figure 1.23) successively shifts the catalytic operating potential to a more positive value, thereby decreasing the overpotential.¹²⁷  However, these electron-deficient ligands also decrease the basicity of the reduced forms of the catalyst, which negatively affects formation of the [Fe-CO₂]²⁻ adduct and cleavage of the C–O bond and therefore decreases the catalytic rate. This is a general trend observed for many electrocatalytic systems: improving the overpotential often comes at the cost of reduced catalytic rates, and vice versa. For 38−40, the substituent inductive effects follow the linear free energy relationship shown in Figure 1.23 between the maximum turnover frequency TOF_max and the catalytic operating potential E°_cat.

The electronic influence of ligand substituents also depends on the position of the functional group, as shown by Duan, Sun, and co-workers.¹⁶⁹  In this report, a methoxy group was installed at the ortho, meta, or para position on each phenyl ring in FeTPP (3). The potential of the Fe(i/0) couple shifted positively in the order Fe-ortho-COOCH₃<FeTPP<Fe-meta-COOCH₃<Fe-para-COOCH₃, indicating that the ortho complex is the most electron-rich system, likely due to the dipole effect. Thus, Fe-ortho-COOCH₃ is expected to be the fastest CO₂ reduction catalyst with the largest overpotential in this series, which was confirmed experimentally in DMF.¹⁶⁹  Again, tuning the electronic properties of the ligand led to an improvement in one parameter (i.e. rate) at the expense of the other (i.e. overpotential). The linear correlation in Figure 1.23 clearly depicts the limitations of using through-bond inductive effects to improve catalysis.

Several substituent changes on the original Re-bpy catalyst have been made in attempts to increase the catalytic activity.⁸⁶  The introduction of various electron-donating or withdrawing substituents on bipyridine has been systematically studied by the Kubiak group.^170,171  A series of 4,4′-substituted [Re(R-bpy)(CO)₃Cl] complexes were prepared (where R=OCH₃, CH₃, tBu, CF₃, CN) and compared to the parent catalyst 2.^86,170  The first reduction potentials of these complexes range from −1.2 V to −1.9 V vs. Fc^+/0 and exhibit an almost linear correlation with the para-substituted Hammett parameter, consistent with this reduction occurring at the bpy ligand (Figure 1.24). While tBu and CH₃ have appropriate electron-donating powers for better catalytic performance than the unsubstituted catalyst, stronger donors like OCH₃ not only have higher overpotentials, but have unstable reduced states that lead to degradation of the catalyst over time. The electron-withdrawing groups CF₃ and CN were substituted onto the bipyridine to lower the overpotential. However, these groups change the electronic structure of the catalyst, shifting the site of the second reduction from the metal to the ligand, and thus making them only active at the third reduction. A similar study¹⁷¹  showed that the 5,5′- substituted catalysts have higher catalytic activity (i_cat/i_p=29.6) than the 3,3′-substituted catalysts (i_cat/i_p=17.0), where i_cat is the catalytic current and i_p is the peak current for the redox couple of the catalyst (Figure 1.25). The authors suggested that the steric hindrance from the 3,3′-substituents distorts the planarity of the bipyridines thus destabilizing the reduced intermediate.

Several π-delocalizing groups have been added to the bipyridines to improve electrocatalysis or photocatalysis. In general, the withdrawing nature of these ligands caused a positive shift in the reduction potentials of the complexes when compared to Re-bpy but they also tended to have lower Faradaic efficiencies. For example, Qiao and co-workers¹⁷²  reported that nanographene Re catalysts have up to an 800 mV positive shift of onset potential for CO₂ reduction compared to Re-bpy in a tetrahydrofuran (THF)/methanol mixture. DFT calculations by Franco and co-workers¹⁷³  suggested that this trend comes from the fact that π-conjugation lowers the SOMO-LUMO gap as compared to the Re-bpy catalyst and moves the LUMO+1 onto the ligand. When the extended π system is not co-planar, as shown in the Re(bis-pyridine anthracene)(CO)₃Cl catalyst,¹⁷⁴  the system shows a negative shift of reduction potential as well as lower FE_CO=20%.

Ligands with similar electronic structures are used as a substitute for bipyridines to lower the overpotential of catalysis. There are detailed studies of Re catalysts with imines which show a similar mechanism to the Re(bpy) compound but generally have lower FE_CO.^36,175–177  For example, Re with diazabutadiene (DAB) ligands show a similar mechanism to bpy-based compounds under CO₂ in electron paramagnetic resonance (EPR) and IR-SEC studies. Under CO₂ they undergo a two-electron disproportionation reaction during catalysis to form CO (FE_CO<10%) and CO₃²⁻.¹⁷⁶  The authors used IR-SEC data to propose that product inhibition was due to the weaker electron donation by DAB ligands to the metal center in the reduced state. Additionally, the CO₃²⁻ formed during disproportionation requires a more negative potential for dissociation.¹⁷⁷  When more donating diamine compounds are used, like Re-pytacompounds (pyta=2-pyridyl-1,2,3-triazole),¹⁷⁸  the selectivity for CO production increases. Despite the high selectivity for CO production, they have a higher overpotential and lower TON than the Re-bpy catalyst under the same conditions. Nganga and coworkers found that Re(pyridine-oxazoline)(CO)₃Cl complexes have faster catalysis than the Re-bpy catalyst. The authors argue that this is due to the superior σ donation of the oxazoline ligand compared with bipyridine.¹⁷⁹  Dinuclear Re α-diimine complexes,¹⁸⁰  and Re imidazole complexes¹⁸¹  with extended π networks have positive shifts of reductions but lower FE or turnovers for CO than Re-bpy catalysts. Overall, diimine and other di-nitrogen ligands are less selective or slower than the Re-bpy catalyst.

NHC containing ligands provide an accommodating framework for electronic and steric tuning. The Agarwal group¹⁸²  focused on tuning Re-NHC complexes. They substituted the atoms next to the carbene with sulfur or oxygen for steric and electronic tuning. While tuning changed the product distribution slightly, the trend is that these catalysts have significant H₂ production and minor formate formation in addition to CO formation (FE_CO=60%). Liyanage and co-workers studied the electronic factors with four electron-deficient Re-pyNHC complexes. The best catalyst Re(PyNHC-PhCF₃)(CO)₃Br shows 4.5 times higher TON_CO than Re(bpy)(CO)₃Br while maintaining a high FE_CO>90% over 1 h in a 2 M H₂O/MeCN mixture. Importantly, this catalyst also showed a lower FE_CO when weaker acids, such as TFE and PhOH, were used.⁹⁰  Their stability was not tested for longer than an hour. Re-Pyridyl-NHCs catalysts undergo the reductive disproportionation reaction with two equivalents of CO₂ to give CO and CO₃²⁻. Overall, NHC ligands tend to shift the product distribution compared to the Re-bpy system and have yet to show efficient, long term stability.

To summarize, catalytic rates and selectivity of Re-bpy complexes can be improved by adding moderately electron-donating groups, but at the cost of increasing the overpotential for catalysis. Modifications to lower the overpotential by using a π-conjugated network to bipyridines, replacing bipyridines with imines, NHCs, diazine and pyridine-oxazoline usually lead to a lower FE_CO.

The reaction of CO₂ with Ir(iii)-hydrides such as iridium(iii) 2,6-bis((diisopropylphosphaneyl)methyl)pyridine trihydride[Ir(PNP)(H)₃] (41) was initially explored for CO₂ hydrogenation. Nozaki and co-workers¹⁸³  demonstrated that CO₂ insertion into one of the hydride bonds in 41 occurs readily under ambient conditions in THF, yielding an equilibrium mixture of the hydride and formate complexes (Figure 1.26a). Similar reactivity was later shown by Meyer, Brookhart and co-workers¹⁸⁴  with the five-coordinate [Ir(PCP)(H)₂] (42), where the favorability of CO₂ insertion is increased by κ²-formate coordination.¹⁸⁵  This result was applied to develop the first example of selective electrocatalytic CO₂ reduction to formate.

The cyclic voltammogram of [Ir(PCP)(H)₂] (42) in 5% H₂O/MeCN under Ar does not exhibit any reduction features within the solvent window, but current enhancement is observed under CO₂ at −1.4 V vs. NHE (ca. −2.0 V vs. Fc^+/0), as shown in Figure 1.26b.¹⁸⁴  The origin of this current increase was confirmed by CPE where formate was generated with 85% FE. The methylene version is also an active catalyst for formate production but requires more negative potentials (−1.8 V vs. NHE, ca. −2.4 V vs. Fc^+/0). Additionally, Bernskoetter, Hazari, Palmore and co-workers¹⁸⁶  demonstrated that a related complex [Ir(PN^HP)(H)₂] (43) can catalyze CO₂ reduction to formate. While the operating potential is less negative than that of 42, lower current densities are achieved due to a slow loss of formate which interacts with the ligand N–H group through hydrogen bonding (Figure 1.26a). Addition of NaPF₆ increases the catalytic current perhaps by facilitating formate release.¹⁸⁷  However, precipitation of a solid on the electrode rapidly shuts down catalysis.

The oxidized form of [Ir(PCP)(H)₂], [Ir(PCP)(MeCN)₂H]⁺ (44), displays a similar current enhancement to that of 42 under CO₂ in 5% H₂O/MeCN with a catalytic TOF of ∼20 s⁻¹. The rate of catalysis increases upon addition of water up to ca. 4%, indicating the important role of water in the mechanism. Furthermore, NMR studies show that MeCN coordinates [Ir(PCP)(H)₂] to give [Ir(PCP)(MeCN)(H)₂] (45) in anhydrous MeCN and does not react with CO₂, but the addition of 5% water triggers CO₂ insertion and release of formate. Water is proposed to stabilize the formate anion via hydrogen bonding, increasing the favorability of the CO₂ insertion equilibrium and facilitating CO₂ reduction. The reactivity of 45 is consistent with the proposed mechanism (Scheme 1.15) which is supported by DFT studies.¹⁸⁸  The rate-limiting CO₂ insertion step involves direct hydride transfer via an [Ir−H⋯CO₂] intermediate, and this step has a lower calculated transition state barrier in water than MeCN. DFT calculations also support that formate release is promoted by hydrogen bonding between formate and water.

While the original catalyst [Ir(PCP)(H)₂] (42) is insoluble in aqueous solution, introduction of quaternary amine groups on the ligand renders these complexes water soluble.¹⁸⁹  The electrochemical behavior of 46 in water is directly analogous to that of 44 in 5% H₂O/MeCN. The TOF under these conditions is 7.3 s⁻¹, corresponding to a second-order rate constant of k_CO2=220 M⁻¹ s⁻¹ for the reaction of 46 with CO₂. The relationship between TOF and k_CO2 is discussed in the ‘Catalyst Comparisons’ section. This rate constant is ∼3.5 times faster compared to [Ir(PCP)(H)₂] in 5% H₂O/MeCN. However, the observed TOF, also denoted k_cat by the authors, is nearly three-fold slower due to the poor solubility of CO₂ in water (0.033 M). CPE confirms formate production with FE_HCOO- of up to 93% in 0.1 M NaHCO₃ with 1% MeCN (pH 7). Notably, 1% MeCN is critical for sustaining catalysis by facilitating loss of a κ²-formate ligand from Ir and regenerating the starting complex (46) to close the catalytic cycle (Scheme 1.15, Figure 1.27).

Further DFT calculations by Ahlquist and co-workers¹⁹⁰  confirmed that the mechanism shown in Scheme 1.15 is feasible in water, with an activation barrier of 16.6 kcal mol⁻¹. However, an alternate pathway was found involving a square-planar Ir(i)-hydride, which likely forms via reduction of [Ir(PCP)(MeCN)₂H]⁺ (44). The calculated barrier for CO₂ insertion into [Ir(PCP)H]⁻ (only 12.3 kcal mol⁻¹) is less than that for 45, suggesting that an Ir(i) mechanism is more energetically accessible than an Ir(iii) pathway. Computational studies from Nielsen and co-workers corroborate that both pathways may be operative.¹⁹¹ 

Nielsen and co-workers calculated the energetics of H₂ evolution via protonation of an Ir(iii) or Ir(i) hydride.¹⁹¹  In all cases, the activation barrier for HER is prohibitively large, consistent with the high selectivities for CO₂ reduction. Notably, both H₂ evolution from water and CO₂ insertion to formate are thermodynamically favorable reactions based on the calculated hydricities of the Ir hydrides. While not a kinetic parameter, hydricity (ΔG_H⁻) does provide some useful guidance in catalyst design. Hydricity is the hydride donor ability of a molecule. From their computational results, the authors conclude that using metal hydride catalysts just sufficiently hydridic to reduce CO₂ coupled with weak Brønsted acids is critical for achieving selective formate production. This conclusion is in line with a recent report from the Kubiak lab,¹¹¹  where similar limitations are outlined for the reaction conditions as a function of catalyst hydricity that are necessary for selective CO₂ reduction to formate.

A series of Fe carbonyl clusters have been explored by Berben and co-workers¹⁹²  for electrocatalytic CO₂ reduction in the presence of weak acids. The butterfly-shaped monoanionic cluster (Bu₄N)-[Fe₄N(CO)₁₂] (47), exhibits a reversible one-electron reduction at −1.23 V vs. SCE (ca. −1.61 V vs. Fc^+/0) in MeCN (Figure 1.28). The addition of acids such as tosylic acid leads to H₂ evolution at this potential. Further electrochemical studies revealed that the rate of H₂ evolution depends on the acid strength, and the use of weaker organic acids enables the electrochemical generation of the hydride cluster [HFe₄N(CO)₁₂]⁻ (47b) by one-electron-one-proton reduction of 47 without subsequent protonation to H₂. These observations were used to select conditions where the reaction of [HFe₄N(CO)₁₂]⁻ with CO₂ was favored over H₂ evolution. Indeed, CO₂ reduction to formate is observed at the [HFe₄N(CO)₁₂]^−/0 couple in the presence of benzoic acid (Figure 1.24), albeit only in trace quantities.

The [(diglyme)₂Na]⁺ salt of [Fe₄N(CO)₁₂]⁻ is water soluble and allowed for electrochemical studies in aqueous solution. In the absence of CO₂, 47 is an active catalyst for H₂ evolution at −1.25 V vs. SCE (−1.01 V vs. NHE, pH 5).¹⁹³  Introduction of CO₂ results in a complete change in selectivity at this potential, with formate generated at high FE (up to 96%) between pH 5–13.¹⁹³  The best results were obtained at pH 7, which corresponds to an overpotential of 0.44 V for CO₂ reduction to formate. This work represents a remarkable advance for homogeneous CO₂ reduction catalysis: selective generation of formate is achieved with a first-row transition metal catalyst in water near-neutral pH at mild overpotentials.

Further studies by cyclic voltammetry and IR-SEC were performed to gain insight into the operative mechanism in MeCN or water. In both solvents, the reaction is first order in [Fe₄N(CO)₁₂]⁻ and CO₂. One-electron reduction of Fe₄N(CO)₁₂]⁻ in 5% H₂O/MeCN under N₂ leads to quantitative generation of the hydride (42b) as seen using IR-SEC. A crystal structure of [HFe₄N(CO)₁₂]⁻ was also obtained, indicating that the hydride is bridged across an edge of the butterfly cluster. Exposure of [HFe₄N(CO)₁₂]⁻, prepared independently or generated in situ by IR-SEC, to CO₂ results in formate production, suggesting that (42b) is the key species that reacts with CO₂. These results are consistent with the mechanism shown in Scheme 1.16.

The high selectivity of [Fe₄N(CO)₁₂]⁻ for CO₂ reduction over H₂ evolution in aqueous solution is notable. Additionally, the reaction rate and formate selectivity in MeCN is greatly improved by addition of 5% water.¹⁹⁴  Here, it is important to recognize how the solvent affects the properties of (47b). The hydricity of this complex in MeCN is 49 kcal mol⁻¹ and 15.5 kcal mol⁻¹ in aqueous solution.¹⁹⁵  The hydricity of formate decreases significantly less, going from 44 kcal mol⁻¹ in MeCN to 24.1 kcal mol⁻¹ in water. Therefore, hydride transfer from 47b to CO₂ is unfavorable by 5 kcal mol⁻¹ in MeCN, but favorable in water by 8.6 kcal mol⁻¹. These energies are consistent with the improvement in catalytic rate and formate selectivity observed with increasing water concentration. Furthermore, it is a general trend that the hydricities of metal hydrides decrease much more in aqueous solution than that of formate due to the greater stabilization of the formate anion from hydrogen bonding with water.¹⁹⁶  This suggests that the development of new metal hydride catalysts capable of hydride transfer to CO₂ should be easier in water than in organic solvents.

The related carbide cluster [Fe₄C(CO)₁₂]⁻ (48) is a better hydride donor than 47 (Figure 1.29) with hydricities of 44 and <15 kcal mol⁻¹ in MeCN and water, respectively.¹⁹⁵  These stronger hydricities increase the favorability of hydride transfer to CO₂ but also increase the driving force for HER, and the rate of H₂ evolution is greater than C–H bond formation. Berben concludes that catalysts with modest hydricities within a narrow range (the “formate window”) are critical to favoring CO₂ reduction over competitive hydrogen evolution, which is in agreement with the findings of Nielsen and co-workers¹⁹¹  and Kubiak and co-workers.^111,195 

Inspired by previous reports on the beneficial role of proton relays (Section 1.5), one carbonyl in [Fe₄N(CO)₁₂]⁻ was replaced with a phosphine bearing a pendent hydroxyl group (Figure 1.29).¹⁹⁷  Reduction of 50 occurred at more negative potentials compared to 49 due to the greater donating nature of phosphine. Notably, no current increase is observed at the [HFe₄N(CO)₁₂]^−/0 couple under CO₂, and CPE at −1.4 V vs. SCE (ca. −1.78 V vs. Fc^+/0) confirmed that only H₂ is produced (FE=97%). On the other hand, when L=PPh₃, electrocatalytic CO₂ reduction occurs under otherwise identical conditions, generating formate with 61% FE (background reduction of water produces 36% H₂). This sharp change in selectivity arises from the close proximity of the hydroxyl group to the metal hydride, which facilitates rapid H₂ evolution. This report highlights the potential pitfalls of installing pendent acidic groups for CO₂ reduction, and emphasizes the need to balance the kinetics and thermodynamics of CO₂ reduction versus HER to obtain selective formate production from metal hydride catalysts.

This section offers a brief description of the best practices when comparing electrocatalysts. Savéant and co-workers pioneered the development of these techniques and we direct the reader to several references which describe them more comprehensively.^198–200  The text, Elements of Molecular and Biomolecular Electrochemistry: An Electrochemical Approach to Electron Transfer Chemistry, serves as a thorough guide to electrochemical catalysis.²⁰¹ 

Catalytic Tafel plots are an effective way to visualize the difference between catalysts in the context of turnover frequency (TOF) and overpotential (η). An optimal electrocatalyst undergoes many selective molecular reactions per second, i.e. has a high TOF. An ideal electrocatalytic system also operates with minimal additional potential exceeding the thermodynamically determined reduction potential (E⁰_cat) for the redox event. The calculated E⁰_cat for proton-coupled electron transfer CO₂ reduction to CO in MeCN (eqn (1.1)) is −0.54 V vs. Fc^+/0 considering standard states for CO₂ and CO as 1 M. The calculated E⁰_cat for CO₂ disproportionation to CO and CO₃²⁻ (eqn (1.2)) was estimated to be −1.15 V to −1.3 V vs. Fc^+/0 based on the standard reduction potential for aqueous CO₂ disproportionation and the free energy of formation for MgCO₃ by Sampson et al.¹³⁴  It is important to bear in mind that these calculated standard potentials are sensitive to the proton source used (i.e. water vs. phenol) and the products made (i.e. MgCO₃ vs. HCO₃⁻). Determination of the overpotential is not trivial and reaction conditions should be considered when calculating E⁰_cat. Nonetheless, Tafel plots can enable catalyst benchmarking for different experimental conditions. The most optimal catalyst will have an “elbow” in the top left corner of the plot (low overpotential and high TOF).

Kinetically limited conditions are steady-state conditions in catalyst consumption. A cyclic voltammogram of a kinetically limited process, where the substrate consumption in the diffusion layer is negligible, yields an S-shaped wave that is independent of scan rate where the forward scan (to more negative potential) is overlaid with the backward scan (to less negative potential). The catalytic rate constant (k_cat) is calculated using the plateau current obtained from the S-shaped wave (i_cat) and the peak current for the reversible redox potential of the wave (i_p). The equations shown are for purposes of illustration and refer specifically for the catalytic reduction of CO₂ by Re-bpy. The relationship between the peak catalytic current observed and the catalytic rate constant for a reversible electron transfer-fast catalytic reaction (EC’) mechanism is shown in eqn (1.4). All electron transfers are assumed to occur at the electrode surface.

In this equation, the number of electrons transferred (n_cat) in the reaction (2 e⁻ for CO₂ reduction of CO), the amount of charge carried by one mole of electrons (F, Faraday's constant), the electrode surface area (A), the catalyst concentration ([cat]), the diffusion coefficient for the catalytically-active species (D), the rate constant of the catalytic reaction (k_cat), and the substrate concentration ([CO₂]) is related to the peak catalytic current (i_cat). This relationship assumes the reaction is first order in the catalyst and substrate (y=1) and the substrate is in large excess of the catalyst (pseudo-first-order conditions).

The Randles–Sevcik equation (eqn (1.5)) describes the relationship between the peak current for a reversible electron transfer event and the potential scan rate. In this equation, the universal gas constant (R), temperature (T), Faraday's constant (F), the diffusion coefficient for the catalytically-active species (D), the electrode surface area (A), and the scan rate (υ) are related to peak current (i_p).

Normalization of the observed catalytic current to the peak current for the reversible wave (i_cat/i_p) circumvents the need to accurately determine the electrode surface area and catalyst diffusion coefficient (eqn (1.6)). This derivation of the equation is for a 2 e⁻ CO₂ reduction and a 1 e⁻ reversible redox couple. The catalyst diffusion coefficient can change between resting and activated states. For example, the Re-bpy catalyst is anionic after reduction and could be better solvated than its neutral resting state. This difference is assumed to be negligible under ideal conditions. The k_cat also represents the theoretical maximum number of catalyst turnovers per unit time (TOF_max).

The catalytic Tafel plot is created by determining TOF at each η value using eqn (1.7). For this relationship the standard potential for the reduction of CO₂ (E⁰_CO2) under the specific catalytic conditions must be known. A plot of log (TOF) vs. overpotential generates the elbow plots shown below and enables catalyst benchmarking independent of reaction conditions.

While catalytic Tafel plots are critical quantitative benchmarks, accurate determination of catalytic values (TOF and η) requires catalysts that are well behaved. Many catalyst systems are not ideal, and side phenomena such as product inhibition of the catalyst or catalyst degradation occur. These side-phenomena greatly influence the quality of the catalytic kinetic data as they can affect the shape of the wave and the observed peak potential of catalysis. When the scan rate is increased, the amount of charge passed is decreased and the influence of side-phenomena on wave shape are diminished. Typical S-shaped curves are not obtained, and the catalytic wave may exhibit scan-rate dependence. A scan-rate dependent catalytic current cannot be reliably used to calculate a catalytic rate constant. Savéant and co-workers developed the foot-of-the-wave analysis (FOWA) to derive catalytic rates from non-ideal catalytic waves where a plateau current is not achieved.^198–200  The onset potential of a catalytic wave is an approximation for the least negative electrode potential required for a reductive catalytic process. FOWA is an examination of the current response observed for the voltammetric wave at this onset potential (halfway up the wave), where side-phenomena have only minor contributions. Eqn (1.8) is applied to determine the observed rate constant (k) which is the slope of the linear region of the plot of i/i_p vs. 1/(1+exp[(F/RT)(E−E⁰_cat)]. This is the relationship between the current (i) that corresponds to the potential (E) and the reaction rate constant (k_cat).

Figure 1.30 shows a comparison of the log TOF vs. η for Re(tBu-bpy)(CO)₃Cl, FeTPP with four paraphenyl trimetylammonium groups (WSCAT; 21), [Ni(cyclam)]²⁺; 9, Mn(mes-bpy)(CO)₃Br; 28, and 28 with Mg²⁺ added. The addition of Mg²⁺, as Mg(OTf)₂, significantly increases the reaction rate for CO₂ reduction by the Mn(mes-bpy)(CO)₃Br catalyst at low overpotentials. The Mg²⁺ enhances catalysis at the CO₂ binding potential (−1.5 V vs. Fc^+/0) for Mn(mes-bpy)(CO)₃Br by enabling a reduction and the reductive disproportionation mechanism. The beneficial effects of Lewis acids added to FeTPP were previously reported by Savéant and co-workers (as described in Section 1.7). The addition of Lewis acids to catalytic systems which show CO₂ binding and “slow catalysis” at lower overpotentials is currently an underutilized strategy for improving catalysis.

Rhenium is less abundant in the Earth's crust than platinum, thus catalysts which use manganese are more attractive from a sustainability perspective. This has stimulated research in the development of Mn-based catalysts with catalytic performance on par with their 2nd row counterparts. Figure 1.31 shows a comparison of the log TOF vs. η for Group 7 metal catalysts. Notably, Mn-tBubpy operates at a lower overpotential compared to Re-tBubpy. The Re-bpy can operate without an added acid since the CO₂ adduct is so strongly basic it can deprotonate MeCN.⁷⁶ 

Fontecave and co-workers recently prepared a series of bis-terpyridine complexes of first-row transition metals.²⁰²  Of the metals examined, Co, Ni, and Zn exhibit CO₂ reduction behavior by cyclic voltammetry in 5% H₂O/DMF. However, low Faradaic efficiencies are observed for [Co(tpy)₂]²⁺ and [Ni(tpy)₂]²⁺ (12% CO and 5% H₂ for Co at −2.03 V vs. Fc^+/0, and 20% CO for Ni at −1.72 V) and no productive catalysis is seen with [Zn(tpy)₂]²⁺, which is likely due to reductive degradation of the terpyridine ligands. Using mixtures of terpyridine and CoCl₂ led to improved activity, with a 76% Faradaic efficiency for CO obtained using a 1 : 1 Co : ligand ratio, suggesting that only one terpyridine is present in the active species. Further studies with [Co(R-tpy)₂]²⁺ showed that by tuning the ligand substituents the rate of competitive H₂ evolution could be significantly decreased, effectively “turning off” H₂ production while still enabling CO₂ reduction.²⁰³  This approach of tuning the catalyst to disfavor competing reactions in order to indirectly improve CO₂ selectivity may be a useful approach for other catalysts.

To provide a useful guide for further improvement of different catalysts, we compared catalytic performances of different catalysts with their turnover numbers and frequencies, Faradaic efficiencies, i_cat/i_p, k_cat and other properties. Tafel plots, which are less commonly reported, are recommended as they give direct visual measurement of catalytic performances, while eliminating the unfairness brought by different experimental conditions. Bearing in mind the significance of direct and fair comparisons among different catalysts, we have several recommendations as protocols for optimization of catalytic conditions.

Different acids with a range of pK_a should be tested in addition to anhydrous conditions under catalytic conditions. Many catalysts, such as porphyrins and Mn bipyridine compounds, are highly catalytic only in the presence of weak Brønsted acids or Lewis acids. In the case of Re bipyridine catalysts, while much effort has been devoted to new ligand designs to enhance catalysis, the use of appropriate weak Brønsted acids or Lewis acid for the best performance has been underestimated, and often provides more leverage. The acidity also affects the selectivity towards CO₂ reduction over hydrogen evolution. Generally, TFE, PhOH, MeOH, and water are common choices of weak acids in MeCN and DMF, and stronger acids such as acetic acid often result in significant hydrogen production.

Careful CPE studies are the most important evaluation of long-term catalytic performance. Although i_cat/i_p and k_cat obtained in cyclic voltammetry are convenient measurements of catalytic performance, the calculated efficiencies are usually higher than those obtained in CPE due to the imperfect curve shape in cyclic voltammetry, and catalyst degradation over time. Since the long term stability of a catalyst is vital for applications on an industrial scale, CPE studies with different weak acids should be done with at least 3−4 turnovers for catalysis. These results can reveal the selectivity for CO₂ reduction, which is not measurable by cyclic voltammetry. In addition, gas leakage is one of the biggest concerns in CPE studies with gaseous products, therefore, standard conditions, e.g. Re(tBu-bpy) with 1 M TFE in MeCN, should be tested and reported for fair comparison of catalytic performance of new catalytic systems. Turnover numbers or turnover frequencies should be reported together with Faradaic efficiencies for clearer understanding of catalysts.

Mechanistic investigations are the key to understanding catalytic cycles and effects of catalyst modifications. Apart from electrochemistry, common tools like DFT calculations, IR and UV–vis measurements, NMR and EPR studies, and laser investigations have been used frequently for many catalysts. For example, IR-SEC has been used to great advantage the in Re-bpy and Mn-bpy systems as their three carbonyl groups show significant shifts in the IR over the range of ∼2100 and 1800 cm⁻¹, reflecting electron density on metal centers by back-bonding in different redox states. One point in DFT computations that must be included in the model is the role of solvent molecules.

Design strategies for obtaining higher catalytic rates at a lower overpotential need to address more than the electronic structural changes that attend various ligand modifications. Through space substituent effects may play an equally important role in the continued development of homogeneous electrochemical CO₂ reduction catalysts. The design strategies discussed in this chapter were selected to provide some examples of real solutions to the challenges at hand, but also to stimulate further thinking on how to move off the trade-offs of linear scaling relationships. By now, it is quite clear that the secondary coordination sphere can be tuned in much the same way that the primary coordination sphere has been tuned over the last forty years.

JAB gratefully acknowledges funding from the UCSD Chancellor's Postdoctoral Fellowship Program (CPFP/PPFP). CPK gratefully acknowledges funding from the Air Force Office of Scientific Research (AFOSR), U. S. Department of Energy (DOE) through the Joint Center for Artificial Photosynthesis (JCAP), and DARPA and the many graduate students and postdocs who contributed to our group's efforts on the electrochemical reduction of CO₂.