Computational Study for CO₂-to-CO Conversion over Proton Reduction Using [Re[bpyMe(Im-R)](CO)₃Cl]⁺ (R = Me, Me₂, and Me₄) Electrocatalysts and Comparison with Manganese Analogues

Abstract

The Nippe group has previously reported a series of imidazolium-functionalized rhenium bipyridyl tricarbonyl electrocatalysts, [Re[bpyMe(Im-R)]-(CO)₃Cl]⁺ (R = Me and Me₂), for CO₂-to-CO conversion using H₂O as the proton source [Sung, S.; Kumar, D., et al. Electrocatalytic CO₂ Reduction by Imidazolium-Functionalized Molecular Catalysts. J. Am. Chem. Soc. 2017, 139, 40, 13993–13996. 10.1021/jacs.7b07709]. These compounds feature charged imidazolium ligands in the secondary coordination sphere and exhibit higher catalytic activities as compared to the Lehn catalyst [Re(bpy)(CO)₃Cl] (where bpy = 2,2′-bipyridine). However, the reaction mechanism for the CO₂ reduction reaction (CO₂RR) over the competing hydrogen evolution reaction (HER) is unclear. Here, we employ density functional theory (DFT) and restricted active space self-consistent field (RASSCF) methods to study the selectivity for CO₂ fixation using [Re[bpyMe(ImMe)](CO)₃Cl]⁺ (1⁺) in water and compare its reactivity to [Re[bpyMe(ImMe₂)](CO)₃Cl]⁺ (2⁺) and [Re[bpyMe(ImMe₄)](CO)₃Cl]⁺ (3⁺). Our results reveal that the turnover frequency (TOF) for CO₂RR using 1⁺ is 4 orders of magnitude higher than for proton reduction, consistent with controlled potential electrolysis (CPE) experiments in which CO was the only detectable reduction product. The imidazolium moiety in the secondary coordination sphere stabilizes the metallocarboxylate species and assists the C–O cleavage through intermolecular hydrogen-bonding stabilizations. Furthermore, our calculations imply that the strongest hydrogen-bonding interactions at the C2 position in 1⁺ contribute to the faster reaction rate observed experimentally with respect to 2⁺. More significantly, the use of the energy span model demonstrates that the turnover frequency-determining transition state (TDTS) corresponds to the formation of the Re–CO₂ adduct, contrasting with manganese analogues in which the C–O bond cleavage step is the TDTS. We attribute this distinction based on the electronic structures of doubly reduced active catalysts. Indeed, RASSCF calculations indicate that rhenium compounds are best described as a rhenium(I) coupled with a doubly reduced bipyridine ligand, [Re^I[bpyMe(ImMe)²⁻](CO)₃]⁰. In contrast, manganese analogues feature a metal center in a formal zero oxidation state antiferromagnetically coupled with an unpaired electron on the bpy, [Mn⁰[bpyMe(ImMe)^•−](CO)₃]⁰.

Graphical Abstract

INTRODUCTION

The fixation of CO₂ to reduced carbon compounds of higher energy will significantly impact the economy and the environment.¹ However, current CO₂ reduction catalysts lack fundamental mechanistic insights and exhibit high over-potentials for the selective CO₂ reduction reaction (CO₂RR) over the competing hydrogen evolution reaction (HER).^2–17 There is, therefore, a critical requirement to study and develop catalysts that will selectively reduce CO₂ to fuels and chemicals using H₂O as the proton source while operating at low overpotential.¹⁷ Not meeting this need will limit our efforts toward finding a solution to our environmental and energy-related challenges.

Recently, Grills and co-workers revealed that the over-potential for CO₂-to-CO could be decreased by 450 mV when using [Re(bpy)(CO)₃Cl] in neat 1-ethyl-3-methylimidazolium tetracyanoborate (EMIM-TCB) as both the solvent and electrolyte related to performing catalysis in CH₃CN containing 0.1 M [NBu₄][PF₆] (Figure 1).¹⁸ Based on this discovery, the Nippe group went a step further by synthesizing a series of imidazolium-functionalized rhenium and manganese bipyridyl tricarbonyl electrocatalysts that incorporate positively charged imidazolium moieties in the secondary coordination sphere (Figure 1).^19,20

Figure 1.

Schematic representation of Grills’ study for CO₂-to-CO conversion in neat 1-ethyl-3-methylimidazolium tetracyanoborate (EMIM-TCB)¹⁸ and homogenous electrocatalysts reported by Nippe and co-workers.^19,20

Experimentally, [Re[bpyMe(Im-R)](CO)₃Cl]⁺ (1⁺, R = Me; 2⁺, R = Me₂) improved the selectivity and catalytic properties compared to the unfunctionalized Lehn catalyst.^19,21–23 However, the reaction mechanism is unclear. Notably, catalytic activities for CO₂-to-CO conversion using 1⁺ and 2⁺ were measured in different water concentrations. In this case, the catalytic current reached a peak upon the addition of H₂O up to 2.8 M.¹⁹ The continuing addition of water inhibits catalysis and leads to gradually decreased catalytic currents at concentrations higher than 3.0 M, differing from the Lehn catalyst. Indeed, the catalytic current of the parent compound increased significantly with the addition of water.¹⁹ This proton source concentration-dependent behavior agrees with results reported by the Lau group, where the catalytic current of the Lehn catalyst reaches its limit value at high concentrations of water (10.4 M H₂O).²⁴ These experimental data imply a distinct mechanism between these imidazolium-functionalized rhenium complexes and Lehn-based catalysts.

Herein, we perform density functional theory (DFT) and restricted active space self-consistent field (RASSCF) calculations to elucidate the electronic structure and reactivity of [Re[bpyMe(ImMe)](CO)₃Cl]⁺ (1⁺), [Re[bpyMe(ImMe₂)](CO)₃Cl]⁺ (2⁺), and [Re[bpyMe(ImMe₄)](CO)₃Cl]⁺ (3⁺, Figure 2) for CO₂RR. However, for clarity, in the main text, we focus on understanding the mechanism for CO₂ fixation using $1_{C2}^{+}$, which features a hydrogen atom at the C2 position and relates its reactivity to $2_{C5}^{+}$ (hydrogen atom at the C4/C5 position) and $3_{C2}^{+}$ (methyl substituent at the C2 position, Figure 2) for CO₂RR. The computed Gibbs free energy reaction profiles for the other conformers are presented in the Supporting Information (Figures S20–S25).

Figure 2.

Schematic representation of the three imidazolium-functionalized rhenium electrocatalysts that were investigated for CO₂-to-CO conversion. The subscript of a given species indicates the position of the substituent that interacts with the substrate (i.e., hydrogen atom at the C2 position in $1_{C2}^{+}$), while the superscript corresponds to the overall charge of the complex (in this case, cationic for all starting materials).

COMPUTATIONAL DETAILS

Density functional theory (DFT) calculations were run using Gaussian 09 (Revision E.01)²⁵ and Gaussian 16 (Revision A.03) packages.²⁶ Geometry optimizations were carried out at the unrestricted ωB97X-D level of theory in acetonitrile (ε = 35.688) using the SMD approach.²⁷ All species were optimized using the Def2-TZVP basis set on Re and Fe (i.e., ferrocene and ferrocenium), while the Def2-SVP basis set was employed for all other atoms except Cl, in which the Def2-SVPD basis set was used (denoted BS1).²⁸ Additional single-point energy calculations were performed with the Def2-TZVPP basis set on all atoms except Cl, in which the Def2-TZVPPD basis set was employed (denoted BS2). Benchmarking calculations using different functionals were also performed to calculate the reduction potentials of species 1⁺ and 2⁺ (Tables S1–S5). The redox potentials using the ωB97X-D(SMD)/BS2//ωB97X-D(SMD)/BS1 level of theory are underestimated by approximately 300 mV with respect to the experimental values (Tables S1 and S2). However, the MN15 functional gave better agreements for one- and two-electron reduced complexes.²⁹ Therefore, instead of shifting our redox potentials by an arbitrary value, all calculated reduction potentials are corrected using the MN15 functional. Exchange correlation integrals were evaluated with a quadrature grid of 99 radial shells and 590 angular points per shell. Stability analyses were performed in addition to analytical frequency calculations on all stationary points to ensure that geometries correspond to local minima (all positive eigenvalues) or transition state (one negative eigenvalue). IRC calculations and subsequent geometry optimizations were used to confirm the minima linked by each transition state.^30,31

All reported redox potentials were calculated using the direct approach rather than a thermodynamic cycle involving gas-phase energies. All redox potentials reported herein (in V) are versus the Fc^+/0 redox couple. Additional information can be found in the Supporting Information.

All energies are modified for zero-point vibrational energy, and free energies (quoted at 298.15 K and 1 atm) are corrected using the modified harmonic oscillator approximation proposed by Grimme, where low-lying vibrational modes are treated by a free-rotor approximation.³² Unless otherwise stated, only lowest energy structures are discussed in the main manuscript. Finally, an applied potential (Φ = −2.00 V) was employed to model each electrochemical step, corresponding to the redox potential of the second reduction in 3⁺.

Natural bond orbital (NBO) analyses were performed with NBO version 3.1,³³ as implemented in Gaussian 09.²⁵

Restricted active space self-consistent field (RASSCF) calculations were done to probe the electronic configurations of the nonreduced, one- and two-electron reduced imidazolium-functionalized rhenium complexes and compared those with manganese derivatives. All calculations were done using MOLCAS 8.4.³⁴ The double-ζ basis set ANO-RCC-VDZP was used for all atoms except H, in which the minimal basis set ANO-RCC-MB was employed.^35,36 Scalar relativistic effects were included using the Douglas–Kroll–Hess (DKH) Hamiltonian,^37,38 and the Cholesky decomposition for two-electron integrals was employed.³⁹ The active spaces are denoted as RAS(n,mh,me;N1,N2,N3),⁴⁰ where n is the total number of active electrons and mh and me are the maximum number of holes and electrons in RAS1 and RAS3, respectively. Here, we allowed a maximum of two holes (mh = 2) in RAS1 and a maximum of two electrons in RAS3 (me = 2). Finally, N1, N2, and N3 define the number of orbitals in RAS1, RAS2, and RAS3, respectively. In all RASSCF calculations, we started with an active space composed of the π and π* orbitals of bipyridine and imidazolium ligands, as well as 5d and 6d orbitals for Re (or 3d and 4d for Mn). Finally, active orbitals were kept consistent to relate the electronic structures of rhenium and manganese complexes. All isosurface plots of active orbitals can be found in the Supporting Information (Figures S33–S38).

RESULTS AND DISCUSSION

First and Second Reduction Steps.

Experimental reduction potentials for first and second catalysts were used to benchmark different DFT functionals (Tables S1–S5).¹⁹ The calculated redox potentials at the MN15(SMD)/BS2//ωB97X-D(SMD)/BS1 level of theory best match the experimental values. Notably, the computed reduction potentials for $1_{C2}^{+}$ and $2_{C5}^{+}$ are within 40 mV of the observed values (Table 1).

Table 1.

Experimental and Calculated Redox Potentials (in V) versus the Fc^0/+ Couple for First and Second Reductions in $1_{C2}^{+}$, $2_{C5}^{+}$, and $3_{C2}^{+}$^a

catalysts	E11/2 (exp.)	E11/2 (calc.)	E21/2 (exp.)	E21/2 (calc.)
$1_{C2}^{+}$	−1.65	−1.66	−1.88	−1.90
$2_{C5}^{+}$	−1.65	−1.69	−2.01	−1.97
$3_{C2}^{+}$	N/A	−1.70	N/A	−2.00

^aAll reduction potentials (in V) are given for standard states at room temperature. Experimental condition: 1 mM of Re, 0.1 M [NBu₄]PF₆ in acetonitrile solution under Ar condition, and 0.1 V/s. Reference electrode: Ag/AgCl; internal standard: Fc^0/+ couple.¹⁹ E₁^1/2 (calc.) is the redox potential for the first reduction of six-to-six-coordinate complexes. E₂^1/2 (calc.) represents the second reduction potential for six-to-five coordinate species (after chloride dissociation).

As previously reported by Kubiak et al., chloride dissociation could yield a rhenium(0) metal center after the first reduction.^41,42 Similarly, the Nippe group described an irreversible initial redox behavior at a slow scan rate for the first two catalysts, indicating that chloride dissociation is feasible at the one-electron reduced complex.¹⁹ At the ωB97X-D(SMD)/BS2//ωB97X-D(SMD)/BS1 level of theory, chloride dissociation after the first reduction is competitive in $1_{C2}^{+}$ and $2_{C5}^{+}$ (i.e., ΔG = 0.0 kcal/mol from 1_C2 (red1-Cl), Figure 3 and Table S3). However, this process is nonspontaneous in 3_C2 (red1-Cl) (ΔG = 1.9 kcal/mol). While the computed Gibbs free energies are within the range of density functional theory errors, these results indicate that chloride dissociation is accessible after the initial reduction in $1_{C2}^{+}$ and $2_{C5}^{+}$. However, these results are functional-dependent (Table S3). For instance, at the MN15(SMD)/BS2//ωB97X-D(SMD)/BS1 level of theory, chloride dissociation is uphill by 6.5 and 6.8 kcal/mol from 1_C2 (red1-Cl) and 2_C5 (red1-Cl), respectively (Table S3). In this case, the calculated redox potentials for the second reduction for five-to-five coordinate species are −1.62 and −1.68 V, which are not in good agreement as to the calculated reduction potentials for six-to-five coordinate complexes (Table 1 and Figure 3). Therefore, we propose that chloride dissociation happens after the second reduction, as previously reported by Kubiak et al.⁴³

Figure 3.

Computed Gibbs free energies of key intermediates (kcal/mol) obtained after the first and second reductions for $1_{C2}^{+}$. The plain values were calculated at the ωB97X-D(SMD)/BS2//ωB97X-D(SMD)/BS1 level of theory, while the values in italics were obtained at the MN15(SMD)/BS2//ωB97X-D(SMD)/BS1 level of theory. The redox potentials (in V) are reported versus the Fc^0/+ couple. The right subscript of a species indicates the position of the substituent that interacts with the substrate (i.e., hydrogen atom at the C2 position in $1_{C2}^{+}$), while the right superscript corresponds to the overall charge of the complex. The left superscript represents the spin multiplicity of the species.

One interesting aspect of our study is that the first reduction is ligand-centered for both five- (after chloride dissociation) and six-coordinate complexes, consistent with the π* orbital of the bipyridine ligand being lower in energy than the 5d_z² orbital on rhenium.⁴² In this case, the Mulliken spin on Re for the six-coordinate species 1_C2 (red1-Cl) is close to zero. We note that Kubiak et al. reported a shift of 21 cm⁻¹ in carbonyl IR stretching frequencies in going from the nonreduced species [Re(bpy)(CO)₃Cl] to the one-electron complex [Re(bpy)(CO)₃Cl]⁻.⁴² This shift was assigned to the formation of a bpy radical, consistent with our DFT results. In addition, previous studies on related [Re(bpy)(CO)₃Cl] complexes have shown that a ligand-to-metal charge transfer (LMCT) upon chloride dissociation could occur to form a rhenium(0) metal center.^42,44,45 However, DFT and RASSCF calculations do not support such LMCT for the imidazolium-functionalized rhenium species (vide infra). For example, the computed Mulliken spin population on rhenium for the five-coordinate species $1_{C2}^{+}$ (red1) is 0.07, contrasting with imidazolium-functionalized manganese analogues, where the Mn(I) center is reduced to Mn(0) upon bromide dissociation after the initial reduction.^20,46 We have also considered acetonitrile binding after the first and second reductions; however, these steps are endergonic for all catalysts, implying that such a process does not occur experimentally.

In the case of the second reduction, our DFT calculations indicate that the additional electron is also located onto the π* orbital of the bipyridine ligand leading to 1_C2 (red2), which has a closed-shell singlet ground state (Figure 3). Similar electronic structures were observed for 2⁺ and 3⁺. Therefore, we hypothesize that the doubly reduced compound is best described as a rhenium(I) metal center coupled with a doubly reduced bipyridine ligand, [Re^I[bpyMe(Im-R)²⁻](CO)₃]. This description agrees with the C_py−C_py bond of 1.38 Å on the bipyridine ligand, indicating significant electron density onto the ligand framework.⁴⁷ This proposed electronic structure differs from the one proposed by Kubiak and Carter for the Lehn catalyst, in which, based on X-ray absorption spectroscopy and electronic structure calculations, the doubly reduced species [Re(bpy)(CO)₃]⁻ is best described as a rhenium(0) metal center with a bipyridine π* radical.⁴⁸

Since DFT calculations are based on single Slater determinants, we employed RASSCF methods to support our proposed electronic configurations for these imidazolium-functionalized rhenium complexes. Like Carter and Kubiak, we focus on the nonreduced, one- and two-electron reduced five-coordinate species since chloride dissociation occurs experimentally.⁴⁸ This approach allows us to emphasize the Mulliken charge changes on Re upon reduction. In the case of the nonreduced five-coordinate complex, $1_{C2}^{2+}$ (red0), we performed a RAS(24,2,2;10,4,9) calculation, which includes 6 e⁻ from the imidazolium π system, 12 e⁻ from the π orbitals of the bpy ligand, and 6 e⁻ for the rhenium(I) metal center (Figure S33). The main configuration state function (CSF), which has a weight of 79%, confirms that the starting material is best described as a rhenium(I) metal center (Mulliken charge = 0.71, Table S23).

For the one-electron reduced five-coordinate complex, $1_{C2}^{+}$ (red1), we have selected a RAS(25,2,2;10,5,9) formulation, which was obtained by adding an extra electron into an additional RAS2 orbital. Interestingly, the main CSF (weight of 79%) hints that the singly occupied orbital ϕ₁₃ is predominantly ligand-based (Figure S34). Indeed, the 5d character for ϕ₁₃ is only 2.8%, supporting that the initial reduction is ligand-centered (Mulliken charges on Re remain unchanged, Table S23).

Finally, the two-electron reduced compound, 1_C2 (red2), was computed by adding one electron and one RAS2 orbital to the active space, yielding to a RAS(26,2,2;10,6,9) calculation. Again, the main CSF (weight of 77%) shows that the doubly occupied orbital ϕ₁₃ (occupation = 1.91) is ligand-based (Figure S35). In addition, we note that the Mulliken charge slightly changes from 0.71 to 0.64 upon reduction due to the doubly occupied RAS1 orbital ϕ₁₀, which has a 5d character of 7.5%, while the 5d nature on ϕ₁₃ is 1.7%.

Taken together, our RASSCF calculations indicate that the doubly reduced complex is best described as a rhenium(I) coupled with a doubly reduced bipyridine ligand, [Re^I[bpyMe(ImMe)²⁻](CO)₃], contrasting with the previously proposed electronic configuration of the Lehn catalyst,^48,49 implying a distinct reaction mechanism for CO₂ reduction using 1⁺.

Formation of the Metallocarboxylic Intermediate $1_{C2}^{+}$ (I5).

Previous experimental and computational studies on Lehn-type complexes have revealed that CO₂ reduction occurs at the two-electron reduced five-coordinate intermediate.^43,50 Following the formation of this species, CO₂ binding or the formation of a Re-hydride compound can happen experimentally. Rhenium catalysts are well known to prefer engaging CO₂ kinetically, while H⁺ reduction is favored thermodynamically in the presence of weak Brønsted acids.⁵¹ The Nippe group also reported that during controlled potential electrolysis (CPE) experiments, CO was the only detectable reduction product using 1⁺ and 2⁺ in the presence of 2.8 and 9.4 M H₂O.¹⁹ To understand the selectivity for CO₂RR over HER, both pathways were considered.

Along the CO₂ fixation pathway, our proposed mechanism begins with the formation of the Re–CO₂ adduct, 1_C2 (I2) (ΔG = −4.6 kcal/mol, Figure 4A). The transition state, 1_C2 (TS1), for CO₂ binding requires an activation energy of 17.2 kcal/mol and indicates that CO₂ is preactivated with an O=C=O bond angle of 144.6° (Figure 4B). We note that this transition state has an open-shell singlet ground state (the Mulliken spin is 0.12 and 〈S²〉 = 0.13) and exhibits an increase of the C_py−C_py bond of bipyridine from 1.38 Å in 1_C2 (I1) to 1.43 Å in 1_C2 (TS1), consistent with one electron being transferred from the ligand to CO₂ through the rhenium center. It is also worth mentioning that the lowest energy structure of 1_C2 (TS1) shows that the imidazolium ligand is trans to CO₂, similar to imidazolium-functionalized manganese catalysts.⁴⁶ Another transition state where the imidazolium ligand weakly cooperates with CO₂ was also located and found to be 2.1 kcal/mol higher in energy (Figure S7).

Figure 4.

(A) Computed free energy (kcal/mol) profile for the formation of metallocarboxylic acid species, $1_{C2}^{+}$ (I5), in the presence of an explicit (H₂O)₅ cluster. All free energies are calculated related to separated reactants. The left superscript represents the spin multiplicity of a given species. (B) Optimized geometries of the transition state 1_C2 (TS1) for CO₂ binding. Distances are in angstroms, and the bond angle is in degree. Nonparticipating hydrogen atoms are omitted for clarity.

More significantly, as we have observed for manganese derivatives,^20,46 the metallocarboxylate intermediate 1_C2 (I2) has a closed-shell singlet ground state, similar to the carboxyl moiety in the C-cluster.⁵² This species is stabilized by the imidazolium ligand through a hydrogen-bonding interaction (C2–H⋯O = 1.63 Å, Figure S8). This stabilization is reinforced by second-order perturbation analyses (Figure S28), where the magnitude of $n_{\text{O}}\to {\sigma }_{\text{CH}}^{*}$ in 1_C2 (I2) is 17.0 and 5.9 kcal/mol. Similar interactions were obtained for manganese analogues (Table S20).⁴⁶ It is worth mentioning that in 1_C2 (I2), the imidazolium group cooperates with carbon dioxide (Figures 4A and S8). At the same time, the ligand is trans to CO₂ in the transition state (Figure 4B), indicating that an isomerization process is required. We located three transition states for the rotation of the imidazolium group, and as expected, the exchange process is facile, demonstrating that the imidazolium moiety is highly fluxional during catalysis (Figure S10), as amino acids in Ni,Fe-CODHs.⁷

As previously done for manganese derivatives, the protonation of 1_C2 (I2) was computed by adding an explicit (H₂O)₅ cluster to yield 1_C2 (I3) (ΔG = −7.5 kcal/mol, Figure 4A) from which the formation of the metallocarboxylic acid intermediate 1_C2 (I4) (ΔG = −7.8 kcal/mol) occurs via 1_C2 (TS2).^20,46 We note that the computed electronic energy for the transition state is higher than 1_C2 (I3) and 1_C2 (I4). However, 1_C2 (TS2) has a lower zero-point energy than both intermediates, offsetting its higher free energy. Second-order perturbation analyses of 1_C2 (I3) and 1_C2 (I4) indicate that the imidazolium ligand acts as a hydrogen-bond acceptor for the water cluster (Figures S29–S32). Similar hydrogen-bonding interactions were demonstrated by Rochford et al. when using [Mn([MeO]₂Ph)₂bpy-(CO)₃ (CH₃CN)](OTf).⁵³ The Marinescu group also varied pendant secondary and tertiary amines in cobalt catalysts and revealed that these functional groups in the secondary coordination sphere facilitate intermolecular proton transfer via hydrogen-bonding interactions.^54,55 Overall, the formation of the metallocarboxylic acid intermediate, $1_{C2}^{+}$ (I5), is exergonic, which is attributed to using an explicit (H₂O)₅ cluster to model the protonation steps instead of a single water molecule.⁴⁶

Reduction-First versus Protonation-First Pathways from $1_{C2}^{+}$ (I5).

After obtaining $1_{C2}^{+}$ (I5), CO and H₂O release can happen via the protonation-first or reduction-first pathways. The reduction-first mechanism begins with reducing $1_{C2}^{+}$ (I5) at −1.71 V to yield 1_C2 (I5-R) (Figure 5A).^20,46 Subsequent addition of water yields 1_C2 (I6-R) (ΔG = −7.1 kcal/mol). As earlier observed, the transition state for C–O bond cleavage, 1_C2 (TS3), displays a hydrogen-bonding network between the imidazolium ligand and the water cluster (C2–H⋯(H₂O)₅ = 1.88 Å), as well as a secondary hydrogen-bonding stabilization between the bridging methylene ligand and H₂O (Figure 5B). This step requires an activation barrier of 17.7 kcal/mol with respect to infinitely separated reactants to give 1_C2 (I7-R). More significantly, the computed Gibbs free energy for the C–O bond cleavage transition state 1_C2 (TS3) (ΔG^‡ = 2.8 kcal/mol) is kinetically more favorable than CO₂ addition (1_C2 (TS1), ΔG^‡ = 7.2 kcal/mol), indicating that CO₂ binding may be the turnover frequency (TOF) determining transition state (TDTS, vide infra). With the following hydroxide removal, species $1_{C2}^{+}$ (I8) is generated, which can then be reduced at −1.81 V to regenerate the active species 1_C2 (red2) after CO dissociation (Figure 5A). The reaction energy, ΔG_r, for CO₂-to-CO is −1.2 kcal/mol, indicating that the reaction is spontaneous.

Figure 5.

(A) Computed free energy (kcal/mol) profile for the protonation-first and reduction-first pathway from intermediate $1_{C2}^{+}$ (I5) in the presence of an explicit (H₂O)₅ cluster. All free energies are calculated related to separated reactants. The left superscript represents the spin multiplicity of a given species. (B) Optimized geometries of the transition state 1_C2 (TS3). Distances are in angstroms, and nonparticipating hydrogen atoms are omitted for clarity.

On the other hand, the metallocarboxylic acid species can proceed through the protonation-first pathway. In this case, an explicit (H₂O)₅ cluster was added to yield $1_{C2}^{+}$ (I5-P). This step is endergonic by 9.1 kcal/mol. Interestingly, the activation barrier for forming the tetracarbonyl intermediate along the protonation-first mechanism (ΔG^‡ = 22.8 kcal/mol) is higher in energy than the reduction-first pathway (ΔG^‡ = 17.7 kcal/mol), indicating that $1_{C2}^{+}$ follows a reduction-first mechanism similar to what Carter and Kubiak reported for the Lehn catalyst.^43,50 Finally, removing the solvated hydroxide forms $1_{C2}^{2+}$ (I7-P), which can be reduced at −1.41 V to yield $1_{C2}^{+}$ (I8). As previously described, this last intermediate can give back the active species 1_C2 (red2) after CO dissociation.

Competitive Hydrogen Evolution Reaction (HER).

In their original work, the Nippe group demonstrated that these imidazolium-functionalized rhenium electrocatalysts could reduce CO₂ to CO with 70% Faraday efficiency while suppressing H₂ in the presence of 9.8 M H₂O.¹⁹ To understand this selectivity, the competitive HER pathway was investigated using $1_{C2}^{+}$ (Figure 6A).

Figure 6.

(A) Computed free energy (kcal/mol) profile for the hydrogen evolution reaction via the reduction-first pathway from doubly reduced species, 1_C2 (red2). All free energies are calculated with respect to separated reactants. The left superscript represents the spin multiplicity of a given species. (B) Optimized geometry of the transition state 1_C2 (TS5), leading to the formation of the metal-hydride intermediate $1_{C2}^{+}$ (I11). (C) Optimized geometry of the transition state 1_C2 (TS6) for H₂ formation. Distances are in angstroms, and nonparticipating hydrogen atoms have been omitted for clarity.

Our proposed mechanism starts with the doubly reduced species, 1_C2 (red2), from which the formation of the Re-hydride intermediate $1_{C2}^{+}$ (I11) is exergonic (Figure 6A). The optimized transition state to give $1_{C2}^{+}$ (I11) indicates that the imidazolium ligand stabilizes the water cluster via a hydrogen-bonding network (C2–H⋯(H₂O)₅ = 2.41 Å, Figure 6B) and facilitates the proton transfer from H₂O to the metal center. This step requires an activation energy of 26.8 kcal/mol related to 1_C2 (red2), which is 9.6 kcal/mol higher in energy than CO₂ binding (Figure 4). We have also considered carbonic acid as the proton source, which is 3.7 kcal/mol higher in energy than CO₂ binding (Figure S40). Thus, the experimentally observed selectivity of 1⁺ for CO₂RR over HER is due to kinetics, consistent with the literature report with the Lehn catalyst.^43,50

The second protonation of the Re-hydride species can progress through the protonation-first or reduction-first pathways to release H₂. Both paths were studied, and the reduction-first mechanism is more accessible (Figure S11). In this case, the rhenium-hydride intermediate $1_{C2}^{+}$ (I11) is reduced at −1.77 V to generate 1_C2 (I11-R) (Figure 6A), which can then be protonated by adding the water cluster into the calculations to yield 1_C2 (I12-R). The subsequent proton transfer step proceeds via 1_C2 (TS6) and gives 1_C2 (I13-R). The transition state also displays a hydrogen-bonding network between the imidazolium ligand and the water cluster (C2–H⋯(H₂O)₅ = 2.03 Å, Figure 6C). Overall, the calculated activation energy for this step is 21.7 kcal/mol related to infinitely separated reactants, which is more accessible than forming the Re–H species, suggesting that 1_C2 (TS5) is the TDTS (vide infra). Finally, after removing the solvated hydroxide from 1_C2 (I13-R), intermediate $1_{C2}^{+}$ (I14) is formed in which H₂ weakly interacts with the metal center. Complete H₂ release gives $1_{C2}^{+}$ (red1) from which 1_C2 (red2) can be generated at −1.62 V (Figure 6A). It is worth mentioning that we have also considered CO₂ insertion into the Re–H bond in $1_{C2}^{+}$ (I11) to generate formate (Figure S39) and was found to be kinetically more accessible than H₂ release.

Energy Span Model and Turnover Frequencies for CO₂RR and HER Using $1_{C2}^{+}$.

Based on the computed free energy reaction profiles, we applied the energy span model (ESM) proposed by Kozuch and co-workers to identify the TOF-determining transition state (TDTS) and TOF-determining intermediate (TDI) for CO₂RR and HER, as well as to calculate turnover frequencies (TOFs) using the AUTOF program.^56,57 The fundamental steps, related transition states, and intermediates are depicted in Figure 7. The detailed calculated TOFs and “degree of TOF control” (X_TOF) can be found in the Supporting Information (Tables S6–S9). We note that microkinetic modeling is another approach that is often used in homogeneous catalysis.^58–60 Still, the ESM allows us to compare the experimental selectivity for CO formation over H⁺ reduction by relating the computed Gibbs free energies and turnover frequency (TOF).^61–63

Figure 7.

Computed free energy (kcal/mol) profiles of one catalytic cycle for CO₂RR (in black) and HER (in maroon) using $1_{C2}^{+}$ in the presence of H₂O as the proton source. All computed Gibbs free energies are reported with respect to separated reactants under an applied potential of Φ = −2.00 V versus the Fc^0/+ couple.

Under an applied potential of Φ = −2.00 V versus the Fc^0/+ couple, the reaction energy, ΔG_r, for CO₂RR and HER is negative (ΔG_r = −1.2 and −19.2 kcal/mol, respectively, Figure 7). Consistent with the literature reports,^43,50 HER is thermodynamically more favorable but kinetically slower than CO₂RR due to a higher activation energy for yielding the critical rhenium-hydride intermediate, the TDTS for H₂ formation (Tables S8 and S9). In this case, the calculated TOF is 1.4 × 10⁻⁷ s⁻¹. This is about 4 orders of magnitude slower than CO₂RR (TOF = 2.6 × 10⁻³ s⁻¹, Tables S6 and S7), supporting CPE experiments where CO was the only detectable reduction product.¹⁹ More interesting, the TDTS for CO₂RR corresponds to the CO₂ addition step (X_TOF = 1.00), in contrast with previous studies suggesting that the C–O bond cleavage is the TDTS for rhenium- and manganese-based CO₂ electrocatalysts.^43,46,50 We attribute this distinction in reactivity to the proposed electronic structure of the doubly reduced imidazolium-functionalized rhenium(I) catalyst, which is less nucleophilic than the initially proposed rhenium and manganese metal centers (i.e., zero oxidation state).

Substituent Effects on the Imidazolium Moiety for CO₂ Reduction.

To this end, for the first time, we have proposed a reaction mechanism using $1_{C2}^{+}$ for CO₂RR in H₂O as the proton source. Additionally, as previously described, the imidazolium moiety in the secondary coordination sphere can introduce intra- and intermolecular interactions to facilitate the CO₂ binding step (i.e., the TDTS) and assist the C–O bond cleavage step. To further understand the effect of the substituents on the imidazolium moiety, reaction profiles for [Re[bpyMe(ImMe₂)](CO)₃Cl]⁺ (2⁺) and [Re[bpyMe(ImMe₄)](CO)₃ Cl]⁺ (3⁺) were computed and compared to [Re[bpyMe(ImMe)](CO)₃Cl]⁺ (1⁺).

All three catalysts show similar reaction mechanisms and potential energy surfaces for $2_{C5}^{+}$ and $3_{C2}^{+}$ are shown in Figures S12–S19. As discussed in detail in the Supporting Information, the formation of the Re–CO₂ adducts 2_C5 (I2) and 3_C2 (I2) is endergonic related to doubly reduced catalysts. More significantly, the computed free energies of metallocarboxylate intermediates for the second and third compounds are higher in energy than 1⁺ (1_C2 (I2), ΔG = −4.6 kcal/mol; 2_C5 (I2), ΔG = −1.2 kcal/mol; and 3_C2 (I2), ΔG = −1.0 kcal/mol). This difference in energies is attributed to weaker intramolecular stabilizations between CO₂ and substituents on the imidazolium ligand, going from 1⁺ to 3⁺ (Figures S8 and S15 and Table S20). Additionally, the computed free energies for the CO₂ binding transition state increase across the series (1_C2 (TS1), ΔG^‡ = 7.2 kcal/mol; 2_C5 (TS1), ΔG^‡ = 8.8 kcal/mol; 3_C2 (TS1) ΔG^‡ = 10.1 kcal/mol), suggesting that 1⁺ is more reactive than 2⁺ and 3⁺ (vide infra).

In H₂O as the proton source, metallocarboxylic acid species are exergonic with respect to starting materials (Figures 4A, S12, and S13). Optimized geometries for transition states demonstrate that imidazolium ligands assist the proton transfer from H₂O to the carboxyl group. However, such facilitation directly depends on substituents at the C2 and C4/C5 positions of the imidazolium moiety (Figure S16 and Table S21).

Subsequent protonation steps of $2_{C5}^{+}$ (I5) and $3_{C2}^{+}$ (I5) were considered to cleave the C–O bond and release H₂O. Like the first catalyst, the reduction-first pathway is preferred due to the lower computed free energies of critical intermediates and transition states (Figures S17 and S18). Again, as observed for manganese analogues, the optimized geometries of transition states display that the imidazolium moiety assists the C–O bond cleavage step by stabilizing the proton source near the carboxylic group (Figure S19). The strength of such stabilization decreases from 1⁺ to 3⁺ due to substituents on imidazolium ligands, as revealed by second-order perturbation interaction energies (Table S22). We note that the selectivity for CO₂RR over HER using $2_{C5}^{+}$ and $3_{C2}^{+}$ was also studied, and our calculations imply that forming metal-hydride species required an activation energy of at least 25.0 kcal/mol, meaning that these imidazolium-functionalized catalysts are also selective for CO₂ fixation over H⁺ reduction, as seen experimentally for 2⁺.¹⁹

Following computed reaction profiles for all three catalysts, the ESM was used to calculate TOFs, as well as to identify the TDTS and TDI. The TDI corresponds to the metallocarboxylic acid complex for all catalysts, while the transition state for CO₂ binding is the TDTS (Figure 8 and Tables S12–S19). This contrasts with earlier studies for Lehn-based catalysts, where the C–O bond cleavage step is the TDTS. Moreover, the calculated TOFs for $1_{C2}^{+}$, $2_{C5}^{+}$, and $3_{C2}^{+}$ are, respectively, 2.6 × 10⁻³ s⁻¹ (Table S6), 1.4 × 10⁻³ s⁻¹ (Table S12), and 4.6 × 10⁻⁴ s⁻¹ (Table S16). While the calculated TOFs are underestimated compared to the experimental values, our predicted catalytic performance is consistent with the experimental data in which 1⁺ has a higher catalytic activity than 2⁺. Our study also predicts that 3⁺ will have the slowest reactivity for CO₂RR, similar to manganese analogues.⁴⁶ Taken together, we propose that functional groups at the C2 position of the imidazolium moiety in the secondary coordination sphere directly impact the catalytic performance of all three catalysts for CO₂-to-CO in H₂O, which follow the trend 1⁺ > 2⁺ > 3⁺.

Figure 8.

Computed free energy (kcal/mol) profiles of one catalytic cycle for CO₂-to-CO conversion using $1_{C2}^{+}$, $2_{C5}^{+}$, and $3_{C2}^{+}$ in the presence of H₂O as the proton source. All relative free energies are reported with respect to separated reactants under an applied potential of Φ = −2.00 V versus the Fc^0/+ couple.

Comparison between the Imidazolium-Functionalized Rhenium and Manganese Electrocatalysts.

In collaboration with the Nippe group, we have reported a series of imidazolium-functionalized manganese catalysts for CO₂-to-CO conversion.²⁰ The experimental and computational data suggested that [Mn[bpyMe(ImMe)](CO)₃Br]⁺ can reduce CO₂ to CO over HER at mild electrochemical potentials. Additionally, a synergistic relationship between functionalized catalysts and H₂O was proposed. More recently, a systematic approach based on computational methods using [Mn[bpyMe(Im-R)](CO)₃Cl]⁺ (R = Me, Me₂, Me₄) for CO₂ fixation was studied by our group,⁴⁶ demonstrating that the C–O bond cleavage step was the TDTS, consistent with the proton concentration-dependent activity observed experimentally.²⁰ Additionally, the predicted TOFs and catalytic performance for three catalysts followed the trend of 1⁺ > 2⁺ > 3⁺ due to the strongest hydrogen interaction induced by C2–H groups on the imidazolium ligand in 1⁺. In the case of rhenium derivatives, we have demonstrated that the TDTS corresponds to the formation of the Re–CO₂ adduct. We attribute this change in the TDTS based on the electronic configurations of doubly reduced active catalysts.

To further support our hypothesis, we performed RASSCF calculations on manganese catalysts using the same active spaces as for rhenium (Figures S36–S38). For the one-electron reduced species, the main CSF (weight of 76%) indicates that the singly occupied orbital ϕ₁₃ is largely metal-centered (3d character is 80.9%, Figure S37), which agrees with our proposed electronic configuration,^20,46 and the change of the Mulliken charge from 0.59 in $1_{C2}^{2+}$ (red0) to 0.41 in $1_{C2}^{+}$ (red1) (Table S23).

In the case of the two-electron reduced compound, the main CSF (weight of 77%) indicates that the electron density on the active orbital ϕ₁₃ (occupation = 1.77 e⁻) is shared between the ligand and metal center (3d character is 36.9%, Figure S38). Interestingly, a second occupied orbital, ϕ₁₄, which has 0.23 electron, features a Mn 3d character of 53.8% (total of 90.7% for both ϕ₁₃ and ϕ₁₄). Thus, taken together, our RASSCF results support that doubly reduced manganese analogues are best described as manganese(0) coupled with an unpaired electron on the bpy ligand, [Mn⁰[bpyMe(ImMe)^•−](CO)₃]⁰, and, therefore, differ from the electronic configurations of the rhenium species. This last observation is consistent with CO₂ binding being the TDTS for rhenium with an activation energy of 17.2 kcal/mol. In comparison, the same step using a more nucleophilic manganese(0) is 14.4 kcal/mol.⁴⁶

CONCLUSIONS

DFT and RASSCF calculations were used to study the electronic configuration and reactivity of three imidazolium-functionalized rhenium electrocatalysts for CO₂-to-CO conversion using H₂O as the proton source. Our work implies that the doubly reduced catalyst is best described as a rhenium(I) coupled with a doubly reduced bipyridine ligand, [Re^I[bpyMe(ImMe)²⁻](CO)₃]⁰, contrasting with the proposed electronic structure of Lehn-type catalysts.^48,49 A direct implication of this electronic configuration is that CO₂ binding is the TDTS rather than the C–O bond cleavage.⁴³ This is consistent with the experimental data where the catalytic currents of these imidazolium-functionalized catalysts gradually decreased at concentrations higher than 3.0 M, indicating that protonation steps are not turnover-determining in contrast to the parent compound.

We have also considered different configurations and substituents at the C2 and C4/C5 positions of the imidazolium moiety. Our calculations reveal that the crucial hydrogen-bonding interaction of C2–H in 1⁺ stabilizes the metallocarboxylate intermediate and assists the protonation step, like amino acid side chains in the C-cluster,^52,64,65 as well as contributes to the fastest reaction rate for CO₂ reduction in comparison to 2⁺ and 3⁺.

Furthermore, the calculated TOF of 1⁺ for CO₂RR was shown to be 4 orders of magnitude higher than for H₂ formation, consistent with CPE experiments in which CO was the only detectable reduction product. This selectivity is attributed to the redox-active bipyridine ligand, which serves as an electron reservoir, like the [3Fe-4S] cluster in Ni–Fe carbon monoxide dehydrogenases,^52,64 providing a way to avoid the formation of the critical metal-hydride intermediate for H₂ formation.^{17,48,66–69}

Taken together, our computational approach provides a quantitative and qualitative way to understand molecular catalysts that mimic key structural features and functions of energy-relevant metalloenzymes.