Pendent Relay Enhances H₂O₂ Selectivity during Dioxygen Reduction Mediated by Bipyridine-Based Co–N₂O₂ Complexes

Abstract

Generally, cobalt-N₂O₂ complexes show selectivity for hydrogen peroxide during electrochemical dioxygen (O₂) reduction. We recently reported a Co(III)-N₂O₂ complex with a 2,2′-bipyridine-based ligand backbone which showed alternative selectivity: H₂O was observed as the primary reduction product from O₂ (71±5%) using decamethylferrocene as a chemical reductant and acetic acid as a proton donor in methanol solution. We hypothesized that the key selectivity difference in this case arises in part from increased favorability of protonation at the distal O position of the key intermediate Co(III)-hydroperoxide species. To interrogate this hypothesis, we have prepared a new Co(III) compound with which contains pendent –OMe groups poised to direct protonation towards the proximal O atom of this hydroperoxo intermediate. Mechanistic studies in acetonitrile (MeCN) solution reveal two regimes are possible in the catalytic response, dependent on added acid strength and the presence of the pendent proton donor relay. In the presence of stronger acids, the activity of the complex containing pendent relays becomes O₂ dependent, implying a shift to Co(III)-superoxide protonation as the rate-determining step. Interestingly, the inclusion of the relay results in primarily H₂O₂ production in MeCN, despite minimal difference between the standard reduction potentials of the three complexes tested. EPR spectroscopic studies indicate the formation of Co(III)-superoxide species in the presence of exogenous base, with greater O₂ reactivity observed in the presence of the pendent –OMe groups.

Graphical Abstract

Introduction

The reduction of dioxygen (ORR) to water (H₂O, 4e⁻, 4H⁺) or hydrogen peroxide (H₂O₂, 2e⁻, 2H⁺) is important for the development of new technologies for energy storage, understanding dioxygen-dependent biological systems, and the development of new oxidative chemical transformations via reactive oxygen species.^1,2 Understanding reaction mechanisms and controlling selectivity between the two pathways is vital: for both energy storage and the biological systems like cytochrome c oxidase (CcO), H₂O is the ideal product.^3–5 In energy storage technologies, the generation of H₂O₂ or other reactive oxygen species can be damaging to the cell-dividing membranes.^3,4 While H₂O is the desired product for the study of energy conversion and biomimetic use of ORR to simultaneously drive chemical oxidation reactions, the selective generation of reactive oxygen species or hydrogen peroxide is an attractive route to the discovery of new direct oxidation reactions using O₂ as the terminal oxidant.²

The study of molecular Co complexes for the reduction of dioxygen (O₂) has been examined extensively with N₄ macrocyclic ligand frameworks, including derivatives of porphyrins,^6–16 corroles,^17,18 phthalocyanines^19–23, chlorins,^24,25 and cyclam,^13,26–30 with the majority demonstrating selectivity for H₂O₂. For certain catalysts, Brønsted acid-scaling relationships have been used to alter product selectivity through thermodynamic bracketing; by thermodynamically excluding the H₂O₂ pathway through the tuning of proton activity, a selectivity switch to H₂O is observed.^1,12

Cobalt complexes containing non-macrocyclic N₂O₂ salen, salophen, and acen derivatives were recently examined by Stahl and coworkers and found to be competent catalysts for the reduction of O₂ to H₂O₂ in methanol (MeOH) solution with acetic acid (AcOH) as a proton donor and decamethylferrocene (Cp*₂Fe) as a chemical reductant.^31–33 However, our own report employing a bipyridine-based Co(III)(N₂O₂) complex, initially studied in the context of O₂ binding to the Co(II) state in the presence of pyridine by Thomas and coworkers,³⁴ [Co(^tbudhbpy)(py)₂][PF₆] (1) showed selectivity for the production of H₂O (71±5%) under identical conditions (Scheme 1).³⁵ Here, [^tbudhbpy]²⁻ = 6,6′-di(3,5-di-tert-butyl-2-phenolate)-2,2′-bipyridine and py = pyridine. We hypothesized that the Co(III) hydroperoxo resting state was resistant to net protonation and H₂O₂ release due to stronger π-backbonding of the Co center with the bpy backbone relative to other N₂O₂ ligands. The relative decrease in electron density at Co from π-backbonding into the bpy fragment allows the hydroperoxo moiety to donate more π electron density to Co, favoring protonation at the distal O atom relative to the Co center and directing product selectivity towards two equivalents of H₂O.

The study of secondary-sphere effects in O₂ reduction based on hydrogen-bonding motifs is motivated by similar proposed effects in bioinorganic processes, such as the CcO O₂ reduction mechanism. In CcO, selectivity towards H₂O production is proposed to derive from hydrogen atom transfer from a tyrosine residue to O₂ bound at the heterobimetallic active site, yielding a tyrosyl radical, Fe(IV)(O), and a Cu(II)(OH) species as the result of net O–O bond cleavage.³⁶ This proposed mechanism has led to the development of a variety of “hangman”-type designs where hydrogen-bonding residues are placed above the active site of catalysis and facilitate O₂ binding, often leading to enhanced rates of O–O bond scission through interaction with the distal O atom of intermediate O₂ adducts.^37–40

Here we take an alternative approach: hydrogen-bond acceptor –OMe moieties are placed such that they can direct protonation towards the proximal O atom of O₂ fragments bound to Co in a similar fashion to a strategy we have previously used to enhance electrocatalytic CO₂ reduction to formate.⁴¹ We describe the synthesis, characterization, and ORR reactivity of [Co(^nPrbpy)(py)₂][PF₆], 2 (Figure 1) in acetonitrile (MeCN) employing benzoic acid derivatives (AH) with varying proton activity, where [^nPrbpy]²⁻ = 6,6′-di(3-methoxy-5-n-propyl-2-phenolate)-2,2′-bipyridine. Interestingly, the activity of 1 in MeCN is negligible, and instead a comparison with the activity of 2 in MeOH was assessed in parallel. In MeCN and MeOH solution, 2 is selective for H₂O₂ production. Complex 2 also undergoes a change in mechanism in MeCN at high proton activities, with the observed rate showing a dependence on [O₂]. To the best of our knowledge, comparable kinetic control for H₂O₂ with high efficiency has not previously been demonstrated for homogeneous ORR catalysts. Consistent with the proposed kinetic relay effect, although 1 and 2 have identical standard reduction potentials, only 2 exhibits quantifiable activity under otherwise identical conditions.

Figure 1.

Summary of complexes studied in this work and their ORR selectivity.

Results

Synthesis and Characterization

The synthesis of ^nPrbpy[H]₂ was completed through modification of our previously reported procedure (Supporting Information).⁴¹ Isolation of 2 was achieved by adding equivalent amounts of ^nPrbpy[H]₂ and Co(OAc)₂•4H₂O to a refluxing MeOH solution, followed by the addition of excess pyridine and ammonium hexafluorophosphate. After a 16h reflux, the solution was cooled to room temperature, filtered, and the volatile components were removed under reduced pressure. The isolated solid was recrystallized from a saturated MeOH solution to yield a red-brown diamagnetic crystalline solid. UV-vis spectroscopy in an MeCN solution (Figure S1) revealed two absorbance bands with λ_max of 253 nm (ε = 2.2 × 10⁴ M^–1 cm^–1) and 345 nm (ε = 5.9×10³ M^–1 cm¹). Complex 1 was synthesized according to our previously reported procedure.³⁵

Initial Reactivity Screening in MeCN

Initial testing for ORR was carried out in MeCN with decamethylferrocene (Cp*₂Fe) and acids of varying strengths (pK_a) under O₂-saturating conditions. Suitable conditions for the ORR were initially found with benzoic acid (pK_a(MeCN) = 20.4).⁴² A structurally homologous series of benzoic acid derivatives (AH) of increasing strength was chosen for subsequent mechanistic study: 2-chlorobenzoic acid (pK_a(MeCN) = 19), 2,4-dichlorobenzoic acid (pK_a(MeCN) = 18.4), and 2,6-dichlorobenzoic acid (pK_a(MeCN) = 17.6) (Figure 2).⁴² Interestingly, while complex 2 exhibited significant rates of catalysis, complex 1 did not show appreciable activity in MeCN in comparison to the background reaction (Figures S4–S7), although this complex is active in MeOH solution with AcOH present (Figure S8).³⁵

Figure 2.

Observed changes in UV-Vis absorbance at 780 nm resulting from catalytic ORR in MeCN. Conditions: 40 μM 2, 25 mM AH, 1.5 mM Cp*₂Fe, and 8.1 mM O₂. Dashed lines are extrapolated fits of the initial rates regions.

Electrochemistry in MeCN

Differential pulse voltammetry (DPV) was employed to examine the relevant electrochemical responses of 1 and 2 under Ar and O₂ saturation. For 1, a single reduction feature is observed with E_1/2 = −0.76 V vs Fc⁺/Fc. Under aprotic conditions, O₂ saturation induces a cathodic shift in the reduction potential of 1 to −0.85 V vs Fc⁺/Fc, consistent with O₂ binding (Figure S9). Upon the addition of the 25 mM of each AH under Ar saturation, shifts in the reduction potential of 1 of 30–50 mV towards more negative potentials are observed (Figures S10–S13, Table S1). Addition of 25 mM 1:1 buffers of each AH:A⁻ pair led to a shift towards negative potentials of ~0.4 V in all cases, consistent with displacement of py from Co(III) and coordination of A⁻.

With 2, a reduction is observed at the same potential as 1, E_1/2 = −0.76 V vs Fc⁺/Fc. Under aprotic conditions, O₂ saturation causes a positive shift in the reduction potential of 2 to −0.65 V vs Fc⁺/Fc, and a second reduction event is observed at −0.94 V vs Fc⁺/Fc consistent with O₂ binding and further reduction (Figure S14). Upon the addition of each of the four acids under Ar saturation, shifts in the first reduction potential of 2 of 10–20 mV towards more positive potentials are observed (Figures S15–18, Table S1). Addition of 25 mM 1:1 buffers of each AH:A⁻ pair again led to a shift in towards negative potentials, although for 2 these amounted to only ~0.1 V in all cases. Interference from the working electrode precludes electrochemical analysis of both 1 and 2 in the presence of O₂ and all of the chosen AH. Although O₂ binding is apparent by CV, competing py loss inhibits the determination of the corresponding second-order rate constant (Figure S19).

Selectivity Determination in MeCN

Selectivity for H₂O₂ was determined by a modified spectrophotometric titration method using an acidified Ti(O)SO₄ solution (Figure S20).^43,44 Briefly, air-saturated solutions of 2 and AH were added to an air-saturated solution of Cp*₂Fe to give final concentrations of 40 μM catalyst, 25 mM AH, and 0.9–1.5 mM Cp*₂Fe. The reaction was allowed to run to completion based on the kinetic studies discussed below. Since complex 1 did not exhibit activity in MeCN solution, no reaction selectivity measurements were run (Figures S4–S7).

For complex 2, a range in H₂O₂ selectivity from 76–87% was observed (Table 1, Figures S21–S28). We postulate that the shift of selectivity is consistent with the enhancement of the rate of protonation at the proximal O atom in the bound Co(III)-hydroperoxo species relative to the distal O atom. We emphasize that there is no difference in the standard potentials of 1 and 2, suggesting that the pendent –OMe relay greatly enhances activity and places the system under kinetic control during the ORR. Control experiments indicate that no H₂O₂ disproportionation occurs under catalytic conditions with 2 for any AH (Figure S29–S32). Effective overpotentials of 0.33–0.52 V were calculated for the production of H₂O₂ eq(1) and 0.81–0.99 V for H₂O eq(2) for complexes 1 and 2, dependent upon AH. We emphasize again that these values are consistent between the two complexes under otherwise identical conditions and that 1 exhibits no activity in MeCN. Note that effective overpotential for the electrochemical ORR is defined by the difference between the standard reduction potential that initiates catalysis and the thermodynamic potential of the reaction.^1,32,33,45

$$O_2+2C{p}_2^{*}Fe+2AH\stackrel{\text{Co}}{\to }{H}_2{O}_2+2C{p}_2^{*}F{e}^{+}+2A^{-}$$ eq(1)

$$O_2+4C{p}_2^{*}Fe+4AH\stackrel{\text{Co}}{\to }2H_{2}O+4C{p}_2^{*}F{e}^{+}+4A^{-}$$ eq(2)

Table 1.

Summary of selectivity for different acids with complex 2.

AH (pKa)	[Co(nPrbpy)(py)2][PF6] H2O2 Selectivity (ncat)	Background H2O Selectivity	ηH2Oa	ηH2O2a
Benzoic Acid (20.7)	81±7% (2.38)	0±4%	0.81	0.33
2-Chlorobenzoic Acid (19)	87±2% (2.26)	0±6%	0.91	0.44
2,4-Dichlorobenzoic Acid (18.4)	81±3% (2.38)	0±2%	0.95	0.47
2,6-Dichlorobenzoic Acid (17.6)	76±3% (2.48)	0±5%	0.99	0.52

Conditions: 40 μM Co, 25 mM AH, air saturation, 0.9 or 1.5 mM Cp*₂Fe. Values for n_cat are shown in parentheses. Overpotential calculations are described in the Supporting Information.

^aValues for η are consistent between both complexes 1 and 2.

Spectrochemical Kinetic Studies in MeCN

Kinetic studies were undertaken with Cp*₂Fe as a chemical reductant using initial rates. Spectral changes in the visible region were monitored by following the rate of appearance of [Cp*₂Fe]⁺ under O₂ saturation conditions with added AH in MeCN, λ_max = 780 nm (ε = 461 M⁻¹ cm⁻¹ as determined by spectrophotometric titration). Selectivity data was used to determine number of electrons transferred (n_cat) for each set of conditions. The catalytic rate law determined for complex 2 conforms to eq(3) for benzoic acid and 2-chlorobenzoic acid (Figures 33–S40), consistent with that reported for other Co(N₂O₂) complexes in MeOH (Figure 3).^31,32,35 From these variable catalyst concentration studies, TOF values were determined with 25 mM AH (Table 2).

Figure 3.

Variable O₂ concentration studies with 40 μM 2, 1.5 mM Cp*₂Fe, and 25 mM AH. (A) Benzoic acid; (B) 2-chlorobenzoic acid; (C) 2,4-dichlorobenzoic acid; (D) 2,6-dichlorobenzoic acid.

Table 2.

Summary of turnover frequencies (TOF), and second- and third-order rate constants for different acids with complex 2 in MeCN.

AH (pKa)	TOF (s−1)	Second-Order Rate Constant (M−1s−1)a	Third-Order Rate Constant (M−2s−1)b
Benzoic Acid (20.7)	1.0 × 10−1±0.1 × 10−1 s−1	4.1	n/a
2-Chlorobenzoic Acid (19)	3.3 × 10−1±0.1 × 10−1 s−1	13	n/a
2,4-Dichlorobenzoic Acid (18.4)	7.8 × 10−1±0.1 × 10−1 s−1	n/a	3.8 × 103
2,6-Dichlorobenzoic Acid (17.6)	1.9±0.1 s−1 (O–H) 0.94±0.05 s−1 (O–D)	n/a	9.2 × 103

Conditions: variable [Co], 25 mM AH, O₂ saturation, 1.5 mM Cp*2Fe. The deuterated 2,6-dichlorobenzoic acid substrate was prepared with 92% O–D substitution.

^adetermined using Co and AH

^bdetermined using Co, AH, and O₂, where possible.

However, with 2,4-dichlorobenzoic acid (at [O₂] < 4mM) and 2,6-dichlorobenzoic acid, the rate law demonstrates a first-order dependence on [2], [HA], and [O₂], with no observed dependence on [Cp*₂Fe] (Figures S41–S48). We propose that this shift to [O₂] dependence eq(4) is consistent with a change in catalyst resting state due to participation of the pendent -OMe relay, vide infra.

$$rate=k_{cat}{[Co]}^1{[AH]}^1{\left[O_2\right]}^0{\left[C{p}_2^{*}Fe\right]}^0$$ eq(3)

$$rate=k_{cat}{[Co]}^1{[AH]}^1{\left[O_2\right]}^1{\left[C{p}_2^{*}Fe\right]}^0$$ eq(4)

Determination of Kinetic Isotope Effect in MeCN

To elucidate additional information on 2, we then evaluated the kinetic isotope effect (KIE, k_H/k_D) with the strongest acid: 2,6-dichlorobenzoic acid. Catalytic reaction rates for both naturally abundant (O–H) and deuterated (O–D) 2,6-dichlorobenzoic acid substrate under variable acid concentration are summarized in Table 2. Under these conditions, a KIE value of 2.3±0.1 was obtained for 2 (Figure 4).

Figure 4.

Kinetic isotope effect determination for complex 2. Black line is with naturally abundant 2,6-dichlorobenzoic acid (O-H). Red line is with isotopically enriched (92%) d₁-2,6-dichlorobenzoic acid (O-D); linear fit procedure with forced zero intercept.

Spectrochemical Kinetic Studies in MeOH

Given that we were unable to assess the activity of 1 directly against that of 2 in MeCN, we quantified the reaction parameters of 2 in MeOH instead, where a complete comparison with 1 is available.³⁵ With 2 in MeOH, a reduction is observed at E_1/2 = −0.74 V vs Fc⁺/Fc by DPV. Upon the addition of AcOH under Ar saturation, shifts in the first reduction potential of 2 of 50 mV towards more positive potentials are observed (Figure S49, Table S2).

The catalytic rate law determined for complex 2 in MeOH with added AcOH conforms to eq(5) (Figures S50–S53), in contrast with that reported for other Co(N₂O₂) complexes in MeOH.^31,32,35 For comparison, complex 1 was previously determined to conform to eq(3), producing H₂O with 71±5% efficiency.³⁵ Under otherwise identical conditions, complex 2 instead produces H₂O₂ with 91±5% efficiency (Figure S54). The dependence of the catalytic rate on [Cp*₂Fe] suggests that the redox-dependent O₂ binding step has become rate determining under these conditions. Consistent with the pendent relay enhancing effective proton activity, control testing without AcOH present shows that the apparent zero-order dependence of [AcOH] on the observed rate is the result of saturation kinetics at low concentrations (Figure S51).

$$rate=k_{cat}{[Co]}^1{[AH]}^0{\left[O_2\right]}^1{\left[C{p}_2^{*}Fe\right]}^1$$ eq(5)

Electron Paramagnetic Resonance Spectroscopy

In order to better understand the nature of the chemically accessible species and reaction intermediates, EPR spectroscopy was performed on frozen solutions of [Co(^tBudhbpy)]⁰ (3) and [Co(^nPrbpy)]⁰ (4) prepared aerobically in CH₂Cl₂ in the presence and absence of excess pyridine. We note that CH₂Cl₂ was used to circumvent solubility limitations of the neutral Co(II) species and provide a non-coordinating solvent environment. The X-band EPR spectrum of [Co(^tBudhbpy)]⁰ 3 at 30 K exhibits a narrow, rhombic signal spanning approximately 42 mT (Figure 5A). Hyperfine coupling to the ⁵⁹Co nucleus is evident on the low- and center-field turning points, and the spectrum is adequately simulated⁴⁶ with g_x = 2.019, g_y = 2.110, g_z = 1.990 and A_x(⁵⁹Co) = 18 MHz, A_y(⁵⁹Co) = 55 MHz, and A_z(⁵⁹Co) = 49 MHz. The signal intensity is consistent with near-quantitative population of the S = ½ species, and in the absence of exogenous base, no change is observed upon exposure to O₂ (Figure S55). We assign this species to a low-spin, pseudo-octahedral Co(II) species with weak axial ligation, likely residual MeOH or water from the sample preparation procedure, which would give rise to the low g-tensor anisotropy and hyperfine coupling observed; four-coordinate square planar and five-coordinate square-pyramidal Co(II) species, such as those observed in cobalt porphyrins, corrins, Schiff-base complexes, cobaloximes, and related compounds generally show large g-tensor anisotropy, with low-field g-values between 2.3 and 2.2, respectively.^47–51 In the presence of excess pyridine, the EPR spectrum of aerated 3 changes slightly, retaining near-quantitative intensity and exhibiting features that are consistent with the formation of a six-coordinate Co^III-O₂^•− species,^47,51–55 with g_x = 2.017, g_y = 2.095, g_z = 1.990 and A_x(⁵⁹Co) = 24 MHz, A_y(⁵⁹Co) = 45 MHz, and A_z(⁵⁹Co) = 28 MHz. Extended exposure of 3 to O₂ in the presence of pyridine leads to a >50% decrease in spin concentration. These results can be rationalized by the formation of the bridging Co(III)–O₂²⁻–Co(III) species, as noted previously for Co(salen) compounds.³² Addition of 2,6-dichlorobenzoic acid results in only EPR-silent species, which suggests formation of a Co(III) – OOH species.

Figure 5.

CW X-band EPR spectra (T = 30 K; P_μw = 2 mW) of 250 μM (A) 3 and (B) 4 in the absence and presence of excess py, as indicated. Samples were prepared aerobically in CH₂Cl₂. Simulations overlaid in red dashed lines were generated with the parameters indicated in the text including isotropic A-strain of 10 MHz and 1 mT line broadening. The simulation for 3(py)O₂ used an anisotropic A-strain of 15, 15, 6.3 MHz in addition to 1 mT isotropic line broadening.

In contrast to the near-quantitative EPR signals seen for solutions of 3, only trace amounts of EPR-active species are observed for [Co(^nPrbpy)]⁰ 4 in the absence of excess base. The electronic structure of 4 differs only slightly from that of 3 but the substantial decrease in signal suggests a strong propensity for dimerization in the divalent state (Figure S56). Consistent with this hypothesis, the addition of excess pyridine restores signal intensity, with formation of a new species that closely resembles that of the 3(py)O₂ adduct is observed for 4(py)O₂, with g_x = 2.017, g_y = 2.100, g_z = 1.990 and A_x(⁵⁹Co) = 25 MHz, A_y(⁵⁹Co) = 53 MHz, and A_z(⁵⁹Co) = 32 MHz (Figure 5B). Addition of excess O₂ quenches the 4(py)O₂ signal almost fully, again indicating formation of a bridging peroxide dimer (Figure S58), and no EPR-active species are observed in the presence of 2,6-dichlorobenzoic acid. In both cases, the EPR spectral features support formation of a Co(III)-O₂^•− species, with the relative intensity differences between 3(py)O₂ and 4(py)O₂ suggesting more rapid and quantitative O₂ binding to 4.

DFT Calculations

To reconcile our mechanistic observations in MeCN, we examined the initial steps of the reaction computationally by DFT methods. Consistent with our observations³⁵ and those of Thomas and co-workers,³⁴ these predict that in the absence of reductant, the predominant Co(III) species should be a bis-py adduct, [Co(^nPrdhbpy)(py)₂]⁺ S = 0. Further, upon one-electron reduction, the dissociation of a single equivalent of py to generate a five-coordinate coordination mode is exergonic by 7.8 kcal/mol, [Co(^nPrdhbpy)(py)]⁰ S = ½ (Figure 6). The exchange of the axial py with an equivalent of MeCN to generate [Co(^nPrdhbpy)(MeCN)]⁰ S = ½ is favored slightly under reaction conditions.

Figure 6.

Reaction profile predicted by DFT methods for O₂ binding by complex 2 upon one-electron reduction; all species are in the S = ½ manifold.

Although the formation of the five-coordinate solvento species is favored slightly, the mono py adduct presents a barrier for O₂ binding of 7.7 kcal/mol, lower than that predicted for the solvento species, 14.9 kcal/mol, in a S = ½ spin configuration (Figure 6). The O₂ binding reaction is endergonic for both axial ligands, but again py (+6.8 kcal/mol) if more favorable than MeCN (+10.6 kcal/mol). Overall, these key differences place the pyassisted pathway as the lowest energy and thermodynamically favorable. Comparable quartet manifolds did not result in stable O₂ adducts and no barrier for O₂ binding could be located without an axial ligand present. EPR properties for possible resting species were also calculated (Table S3). Good agreement between the calculated and experimentally observed g- and A-tensors was obtained for the 4(py)(O₂) species, with g_calc = [1.985, 2.003, 2.047] and A_calc(⁵⁹Co) = [22, 29, 77] MHz.

Discussion

These data demonstrate that complex 1 is inactive for ORR under all examined conditions in MeCN, while 2 is 76–87% selective for H₂O₂. This is noteworthy, given that there is no quantifiable difference in their standard reduction potentials. To assess if the minimal activity of complex 1 in MeCN was the result of steric effects, an additional control complex was prepared. For this new catalyst, [Co(^p-tBudhbpy)(py)₂][PF₆] 5 (where ^p-tBudhbpy = 2,2’-([2,2’-bipyridine]-6,6’-diyl)bis(4-(tert-butyl)phenol)), an H atom was substituted adjacent to the Co-coordinated O atom of the ligand, where complexes 1 and 2 have tert-butyl and methoxy functional groups, respectively. Under Ar saturation in MeCN, complex 5 has a catalytically relevant E_1/2 = −0.78 V vs Fc⁺/Fc, which is only 20 mV more negative than both 1 and 2. Under catalytic conditions in MeCN with 2,6-dichlorobenzoic acid as the proton donor, complex 5 has a selectivity for H₂O₂ of only 21±5% (79±5% selectivity for H₂O). In addition to suggesting a steric inhibition for complex 1, the observed selectivity for complex 5 is consistent with a secondary-sphere effect altering mechanism,^41,56–59 since it is expected that for catalysts within the same family, activity and selectivity will scale with overpotential.

We propose the overall enhanced kinetic selectivity for H₂O₂ observed for 2 under all conditions is the result of the alkyl ether pendent relay enhancing proton transfer to the proximal O atom of a Co(III)–(OOH) intermediate, accelerating H₂O₂ dissociation (Figure 7). With benzoic acid and 2-chlorobenzoic acid, this protonation step (Figure 8, Step B) remains rate-determining and the catalytic rate law shows no [O₂] dependence but is dependent on [Co] and [AH].^31–33

Figure 7.

Proposed pendent relay interaction relevant to the kinetic selectivity of ORR mediated by complex 2

Figure 8.

General proposed ORR mechanism for complex 2 in MeCN.

Interestingly, additional mechanistic consequences of the pendent relay become apparent with the stronger acids, 2,4-dichlorobenzoic acid and 2,6-dichlorobenzoic acid. Not only is the kinetic selectivity for H₂O₂ retained, but with the stronger acids 2,4-dichlorobenzoic acid and 2,6-dichlorobenzoic acid as the proton source, the rate becomes dependent on [O₂], as well as [Co] and [AH]. These data imply that the relay also alters earlier steps of the cycle and shifts the rate-determining step. Based on these observations, we propose that the resting state of catalysis for 2 (Figure 8) with the stronger acids shifts to [Co(II)(^nPrdhbpy)(py)(O₂)]⁰ (ii). Initially, an equilibrium reaction between i and O₂ occurs, which we propose is followed by rate-determining proton-transfer reaction (Step B) to produce a [Co(III)–OOH]⁺ species. Under reaction conditions, it is likely that subsequent reduction occurs rapidly. From species (iii), the pendent –OMe relay again directs a proton equivalent towards the proximal O atom of the hydroperoxide ligand, which is again the selectivity-determining step of the reaction. The protonation of Co(III)-OOH to generate a bound H₂O₂ adduct precedes net dissociation of H₂O₂ and binding of py to regenerate i (Figures 7 and 8).

Evidence of proton transfer involving the relay being the rate-determining step with stronger acids is directly obtained from the KIE data: for complex 2, a KIE of 2.3±0.1 is observed (Figure 8).^60–62 The absence of catalytic rate dependence on the concentration of the reductant for stronger acids suggests that this proton transfer is decoupled from electron transfer, eliminating the possibility of a PCET step under these conditions. We note that the formation of bridging Co(III)–O₂²⁻–Co(III) species cannot be excluded, based on the observed shift to diamagnetic species in the EPR under O₂ saturation. However, the observation of EPR signals characteristic of Co(III)-O₂^•− species under low [O₂] exposures supports the intermediacy of monomeric species^63–66 and we postulate that a dynamic off-cycle equilibrium exists between monomer and dimer. Stahl and co-workers have previously demonstrated that at the low [Co] used under catalytic conditions, monomer speciation is generally >95%.³² Consistent with this, DFT calculations on a probable dimer structure (Figure S59) suggest that its formation from the O₂ adduct (ii) and [Co(II)(^nPrdhbpy)(py)]⁰ is endergonic by 6.4 kcal/mol under catalytic conditions.

Comparing the data obtained in MeOH for complex 2 with the effective overpotentials, selectivity, and rates of our previous report also leads to a few observations. As mentioned above, complex 1 produces H₂O with 71±5% selectivity in MeOH solution with AcOH, whereas 91±5% selectivity is achieved for H₂O₂ during ORR mediated by 2.³⁵ However, the near quantitative kinetic selectivity for complex 2 in protic solvent cannot be exclusively attributed to the inclusion of the pendent - OMe relay. The steric profile at Co also appears to play an important for ORR selectivity in MeOH; control experiments with complex 5, under otherwise identical conditions show 71±7% selectivity for H₂O₂. Overall, for complex 2 the shift to a rate dependence on [Co], [O₂], and [Cp*₂Fe] and the saturation of rate at low [AcOH] suggests that in MeOH the rate-determining step has shifted to O₂ binding following one-electron reduction (Figure 8, Step A).

We emphasize that DPV experiments described above with 1:1 buffered conditions of HA:A⁻ indicate coordination by benzoate anions in the Co(III) oxidation state, which has previously been shown to alter reaction mechanism in other systems for ORR.⁶⁷ This precludes kinetic measurements under buffered conditions and suggests that over time, as the reaction proceeds to completion, the resting state of the catalyst and the observed rate are both likely to change as benzoate anion concentrations increase and change the catalyst standard potential. Since the homoconjugation values of the benzoic acid derivatives are known in MeCN, it is possible to account for their thermodynamic contribution⁶⁸ to the estimated standard potential of the reaction (See SI). The resultant effective overpotentials should be regarded as a lower-limit estimate for this value, given that we were unable to measure buffered conditions.

Conclusions

We have reported a new Co(III)(N₂O₂) complex bearing pendent –OMe relays for the ORR. Mechanistic studies reveal a change in mechanism at high proton activities for this complex, with the reaction shifting to [O₂] dependence. This change is consistent with a shift in the rate determining step to protonation of a Co(III)(O₂^•− ) (ii) intermediate (Figure 8), decoupling it from electron transfer. This assignment is supported by KIE studies and the absence of a concentration dependence on the reductant.

For ORR mediated by 2, selectivity for H₂O₂ (76–81% H₂O₂) is observed in MeCN with TOF = 1.0 × 10⁻¹ to 1.5 s⁻¹ at estimated effective overpotentials of 330–520 mV. In contrast to this, complex 1 is inactive in MeCN under otherwise identical conditions, which control experiments with complex 5 suggest can be attributed to steric effects. Since all three complexes have near identical standard reduction potentials (which defines the effective reaction overpotential as almost identical for all three systems^1,32,33,45), these results are consistent with a breaking of the expected reaction scaling relationship through secondary-sphere effects involving the alkyl ether groups.^41,56–59 For complex 2, these relays place the reaction under kinetic control and change the underlying mechanism. Control complex 5, where the pendant relay has been replaced with an H atom, instead exhibits a selectivity for H₂O of 79±5% under the same conditions in MeCN. Interestingly, in MeOH solution with AcOH, 2 shows near quantitative selectivity toward H₂O₂ (91±5%) and 5 shifts to a H₂O₂ selectivity of 71±7%, suggesting that in the protic solvent the steric profile of the ligand alters reaction selectivity as well. This demonstrates a suppression of the pendent relay in protic solvent and the role of steric effects in the protonation of the hydroperoxo intermediate during the ORR.

This work helps broaden the understanding of the ORR utilizing non-macrocyclic Co complexes and has the potential to lead to active catalysts for the ORR to either H₂O₂ or H₂O with high selectivity and activity. Further studies on the underlying non-covalent solvent effects, mechanistic dependence on proton activity, and the tunability of axial ligand effects on O₂ binding are currently underway.