Homogenous Electrocatalytic Oxygen Reduction Rates Correlate with Reaction Overpotential in Acidic Organic Solutions.

Improved electrocatalysts for the oxygen reduction reaction (ORR) are critical for the advancement of fuel cell technologies. Herein, we report a series of 11 soluble iron porphyrin ORR electrocatalysts that possess turnover frequencies (TOFs) from 3 s-1 to an unprecedented value of 2.2 × 106 s-1. These TOFs correlate with the ORR overpotential, which can be modulated by changing the E1/2 of the catalyst using different ancillary ligands, by changing the solvent and solution acidity, and by changing the catalyst's protonation state. The overpotential is well-defined for these homogeneous electrocatalysts by the E1/2 of the catalyst and the proton activity of the solution. This is the first such correlation for homogeneous ORR electrocatalysis, and it demonstrates that the remarkably fast TOFs are a consequence of high overpotential. The correlation with overpotential is surprising since the turnover limiting steps involve oxygen binding and protonation, as opposed to turnover limiting electron transfer commonly found in Tafel analysis of heterogeneous ORR materials. Computational studies show that the free energies for oxygen binding to the catalyst and for protonation of the superoxide complex are in general linearly related to the catalyst E1/2, and that this is the origin of the overpotential correlations. This analysis thus provides detailed understanding of the ORR barriers. The best catalysts involve partial decoupling of the influence of the second coordination sphere from the properties of the metal center, which is suggested as new molecular design strategy to avoid the limitations of the traditional scaling relationships for these catalysts.

Introduction

Fuel cells offer clean and effective electrical power generation from chemical fuels for stationary, portable, and transport applications.[1] Widespread implementation of fuel cell technologies has been limited in large part by the inefficient oxygen reduction reaction (ORR) at the cathode, and by the typical requirement for large amounts of expensive platinum.[2,3] This motivates research in heterogeneous and homogeneous catalysts, from new platinum alloy nanocrystals supported on electrodes[4] to novel earth-abundant carbonaceous materials[5] to unique molecular structures and hybrid combinations.[6,7] Homogeneous electrocatalysts have yet to match heterogeneous materials in turnover frequencies (TOFs) per active site at modest overpotentials, but they permit facile development of important structure/activity and driving force/activation barrier relationships. Such atomic-level understanding is key to building better ORR electrocatalysts.

Iron and cobalt complexes with N4-macrocyclic ligands have been extensively studied as ORR catalysts[8,9] and are models for promising but complex solid Fe/N/C ORR electrodes.[10] Molecular ORR electrocatalysts are typically only soluble in organic solvents such as N,N′-dimethylformamide (DMF) or acetonitrile (ACN), making it difficult to directly compare and understand the effects of different design features. In addition, the standard potentials for O2/H2O in nonaqueous solvents were unknown until recently.[11,12] Therefore, previous studies have predominantly done aqueous measurements of molecular ORR electrocatalysts adsorbed on electrodes, often in complex pastes. In these situations, differing local environments complicate understanding,[13] with some elegant exceptions.[7,14]

Described here are studies on a series of 11 iron porphyrin ORR electrocatalysts ([FeIII(por)]OTf with OTf = trifluoromethanesulfonate, Figure A). Catalysts 1 and 8–11 were designed with pendant functionalities to serve as potential “proton relays”, with the aim of facilitating H+ delivery to the iron active site. This has been shown to be an effective strategy for other homogeneous electrocatalysts for H+ and CO2 reduction.[15,16] The catalysts were electrochemically evaluated for homogeneous ORR in organic solvents, with varying amounts of acid. Analysis of the ORR turnover frequencies shows strong correlations with the recently available standard potentials[11,12] under the differing conditions. This is the first such analysis for molecular ORR catalysts. The presence of the correlations is remarkable because the turnover limiting steps for these catalysts do not involve electron transfer from the electrode. Computational analysis of a subset of the catalysts shows that the correlation with overpotentials results from linear free energy relationships connecting the catalyst E1/2 with its O2 binding free energy and the proton affinity of the resulting superoxide complex, providing atomic-level understanding of this series of electrocatalysts.

Figure 1

(A) Iron(III) 5,10,15,20-tetraarylporphyrin electrocatalysts 1–11. (B) CV of 1 in ACN under 1 atm of N2 (blue) or O2 (red), 20 mM [DMF-H]OTf, 100 mV/s scan rate; the inset shows the electrocatalytic onset (foot-of-the-wave) region.

Results

The electrochemistry of complexes 1–11 has been examined in DMF or ACN, with excess [DMF-H]OTf as the acid. Most catalysts could only be evaluated in one of these solvents, due to limited solubility. Cyclic voltammograms (CVs) of the complexes all showed a reversible FeIII/II redox couple in the absence of O2. Changing the substituents on the porphyrin scaffold allowed tuning of the FeIII/II reduction that initiates electrocatalysis (EFe(III/II)) over a significant range in ACN (∼100 mV) and DMF (∼270 mV). In the presence of excess acid and 1 atm of O2, CVs of 1–11 showed a substantial increase in current, indicative of ORR electrocatalysis (Figure B).

Turnover frequencies (TOFs, s–1) were quantified electrochemically using foot-of-the-wave analysis (FOWA; see Supporting Information). This TOF is defined by the rate law –d[O2]/dt = TOF[FeII], as is common;[16,17] it corresponds to the more electrochemically precise TOFmax recently introduced by Savéant et al.[18,19] FOWA enables determination of TOFs for soluble electrocatalysts despite limitations such as substrate consumption and catalyst decomposition.[18] The TOFs in the presence of 1 atm of O2 and 20 mM [DMF-H]OTf are given in Table . The TOFs in general increase linearly with increasing partial pressures of O2 and [DMF-H]OTf (Figures S1 and S20–S32). The TOF values reported in Table are in regions where the TOF scales linearly with the [DMF-H+], however some catalysts become independent of the acid strength at higher concentrations (see below). In general, the TOF (or TOFmax[18,19]) values are well-defined properties of the molecular catalysts. For the catalysts studied here, the rate of O2 consumption near the electrode surface (and thus current, Figure B) reaches a plateau when E < E1/2, representing the maximum chemical rate under conditions where electron transfer from the electrode to the soluble catalyst is rapid and complete.[18] This is in contrast to heterogeneous electrocatalysts where the current typically continues to increase exponentially with the applied potential. The effective overpotential for different molecular electrocatalysts can then be defined by the potential of the catalysis-initiating redox couple (in this case the E1/2 of FeIII/II, EFe(III/II)) and the ORR equilibrium potential under the relevant catalytic conditions (O2 pressure, proton and water activity).[12]

Table 1

Catalyst Reduction Potentials (vs Ferrocene/Ferrocenium, Fc+/0), ORR Overpotentials (See Text), and TOF from FOWAa

catalyst	EFe(lll/ll) (V)	η Fe/ORR (V)	TOF (s–1)
1ACN	–0.375	1.21	2.2 × 106
2ACN	–0.390	1.24	2.2 × 106
9ACN	–0.326	1.15	6.5 × 104
10ACN	–0.296	1.13	2.2 × 104
11ACN	–0.280	1.11	2.2 × 102
1DMF	–0.630	1.18	2.0 × 103
2DMF	–0.611	1.16	2.5 × 103
3DMF	–0.547	1.10	1.6 × 102
4DMF	–0.536	1.09	1.8 × 102
5DMF	–0.530	1.08	2.7 × 101
6DMF	–0.491	1.04	5.0 × 100
7DMF	–0.486	1.04	1.5 × 101
8DMF	–0.362	0.91	3.0 × 100

In ACN (superscript “ACN”) or DMF (superscript “DMF”), 0.1 M [Bu4N]PF6, 1 atm of O2, 20 mM [DMF-H]OTf, 100 mV/s scan rate. Experimental uncertainties are ±5 mV for EFe(III/II), ±20 mV for overpotentials, and ±15% for TOFs.

Catalyst selectivity was investigated by electrochemical detection of hydrogen peroxide using rotating ring disk voltammetry (RRDV, see Supporting Information for details). Under identical conditions, 1–11 produced minor amounts of hydrogen peroxide (<15% H2O2). The % H2O2 converts to an average number of electrons transferred per catalytic turnover (ncat > 3.83), which has been included in the TOF calculations. Thus, these catalysts are quite selective for the direct reduction of O2 to H2O, as has been previously reported for iron porphyrin complexes under similar conditions.[20,21]

The 2-carboxyphenyl catalyst (1) and its methyl ester (2) are both remarkably fast catalysts in ACN solution, with TOFs of 2.2 × 106 s–1 for both systems under our standard conditions (Table ). Contributions from adsorbed or heterogenized catalysis on the electrode surface were found to be negligible through use of a rinse test (Figure S18). The catalytic rate of O2 reduction by 1 was also examined by a nonelectrochemical stopped-flow kinetics study (see Supporting Information). Following our earlier methodology,[22] an ACN solution of 3 μM 1, 8 mM Cp*2Fe (Cp* = pentamethylcyclopentyldienyl), and 2.2 mM [DMF-H]OTf/2.2 mM DMF was mixed with an air-saturated O2 solution. The appearance of [Cp*2Fe]+ was complete within 200 ms (Figure S46). Converting the derived third order rate constant to a TOF under the electrochemical conditions gives TOF = 1.1 × 106 s–1. Thus, spectrochemical kinetic measurements are consistent with the electrochemical results and provide confirmation of the remarkably high catalytic activities observed.

The TOFs for 1 and 2 in ACN are, to our knowledge, the fastest TOFs ever reported for homogeneous ORR electrocatalysts. Furthermore, the TOFs for 1 and 9 increase linearly with [DMF-H]+ until saturation kinetics are observed at higher acid concentrations and the TOFs become independent of [DMF-H]+. The maximum TOFs are 3 × 106 s–1 and 2.5 × 105 s–1 for 1 and 9, respectively. Based on the catalytic mechanism in Figure A,[22] when the rate becomes independent of acid concentration, the TOF is limited by O2 binding (see Supporting Information, page 5). Using TOF = kO2[O2] and [O2] = 8 mM in ACN under 1 atm of O2, the saturation TOFs correspond to second order rate constants for O2 binding, kO2 = 4 × 108 M–1 s–1 (1) and 3 × 107 M–1 s–1 (9). These are in the range of typical kO2 for ferrous porphyrins.[8] Thus, these catalysts approach the fastest possible rate, being only ∼102 below the rate constant for diffusive encounter of O2 and the catalyst.[23]

Figure 2

(A) Proposed initial steps in the ORR catalyzed by 1–11 (oval = porphyrin). (B) log(TOF) as a function of EFe(III/II) with 20 mM [DMF-H]OTf, 1 atm of O2. The four purple points for 8DMF correspond to (from left to right) 50, 20, 10, and 5 mM [DMF-H+]OTf. (C) log(TOF) as a function of the catalyst-specific overpotential ηFe/ORR (defined at catalyst E1/2, see text). The solvent and acid concentrations are indicated by the color and shape of the points, as indicated in the legend. Groups of points are labeled by the catalyst numbers in that group. 8DMF is represented in purple, and the conditions are identical to those specified in panel B.

The 2-pyridyl substituted complex 8, unique among the catalysts in the series, displays an inverse dependence on the concentration of acid. Increasing the concentration of [DMF-H]OTf causes the TOF for 8 to decrease, and also shifts the EFe(III/II) to more positive values. This is likely a consequence of increased protonation of the 2-pyridyl substituents upon addition of [DMF-H]OTf. Indeed, under 1 atm of N2 the EFe(III/II) for 8 shifts in a Nernstian fashion (90 mV/decade in [DMF-H+], Figure S19) implying that, upon reduction, the average protonation state of 8 increases by one and a half, though detailed characterization of the equilibrium mixture of proteomers and rotamers in solution has not been achieved. Qualitatively, increasing the average protonation state of the catalyst results in less active forms and decreases the TOF.

Initial understanding of the TOFs for the various iron catalysts came from plotting log(TOF) versus EFe(III/II), which revealed linear free energy relationships (LFERs) within each solvent system (Figure B). The LFERs fit very well for each unique system analyzed (R2 > 0.98, 8DMF excluded due to its unusual acid dependence and 11ACN excluded for reasons described below). The ORR TOFs are larger for catalysts with more negative EFe(III/II). When the catalyst EFe(III/II) is more negative, its reduced form is more electron rich (and the TOF reports on the rate under conditions when the catalyst is fully reduced, see above). The log(TOF)/EFe(III/II) correlation is limited, however, because it focuses only on the electrochemical potential while the ORR is an electron/proton process. Thus, this LFER does not account for the differences in proton activity (such as in the case of 8, where each data point represents the same initial catalyst with varying amounts of [DMF-H+]. In addition, electrochemical potentials are not easily compared between solvents.

It is more appropriate to correlate TOF with the overall driving force of the catalytic reaction. For each of the reaction conditions used here, there is a characteristic equilibrium potential for O2 reduction (EO2/H2O), depending on the solvent, proton activity, etc.[11] Similarly, each of the molecular catalysts has a characteristic redox potential under these conditions (EFe(III/II)), which we approximate as the measured E1/2 of the catalyst (see Supporting Information, page 5). We therefore define a characteristic overpotential under these conditions, ηFe/ORR = EO2/H2O – EFe(III/II). This ηFe/ORR is the characteristic value within the general definition of overpotential as the difference between EO2/H2O and the applied potential E. We emphasize that ηFe/ORR is a well-defined single value for a molecular electrocatalyst, because the current density (and thus reaction rate) are constant at overpotentials beyond EFe(III/II).[24] This contrasts with the situation for heterogeneous electrocatalysts, where the current density increases exponentially with η.

Correlating log(TOF) with ηFe/ORR for the 11 catalysts analyzed in different conditions shows that all of the results fall on two parallel lines (Figure C). The simple relationship between TOF and ηFe/ORR has been independently observed in three different ways: (i) changing the catalyst, which changes EFe(III/II); (ii) changing the solvent, which alters EO2/H2O; and (iii) changing the [DMF-H]+, which shifts EO2/H2O in a Nernstian fashion. Catalyst 8 is unique since both i and iii occur upon changing the [DMF-H]+. Considering both changes for 8 in Figure C places all of its values on a LFER of similar slope to catalysts 1–7 in DMF. This analysis allows, for the first time, the direct comparison of homogeneous ORR electrocatalysts in different solvents and with different acid concentrations. Thus, catalysts 1, 2, 9, and 10 in ACN and 8 in DMF are clearly better catalysts than the other, because they lie on a line that is above and to the left. At a given ηFe/ORR, the catalyst systems on the left line have higher TOFs, or stated another way, these catalysts require a lower ηFe/ORR to reach the same TOF.

Dissecting the TOF/ηFe/ORR correlations provides additional understanding. The log(TOF) vs ηFe/ORR lines in Figure C have inverse slopes of 57 and 54 mV per decade in TOF for the left and right correlations, respectively. This is (by coincidence we believe) remarkably close to the Nernstian slope of 59 mV per decade in concentration. We emphasize that this plot is distinct from the traditional Tafel plot analysis of a single electrocatalyst at different applied potentials (molecular or heterogeneous). Figure C relates the properties of different catalysts under different conditions. Fundamentally, these correlations show how the barrier height of turnover limiting step(s) changes as a function of overall driving force, varied via EO2/H2O, EFe(III/II), or both for 8.

The presence of these correlations is remarkable. For these catalysts, the measured TOF is limited by the O2 binding equilibrium and the rate of protonation of the superoxide intermediate, as shown in the mechanism in Figure A. The TOFs are not determined by any electrochemical step. The correlations indicate that electronic modifications of the catalysts linearly tune the free energies of the relevant intermediates and reaction steps of the catalytic cycle.

This tuning is revealed by density functional theory (DFT) calculations of the reduction potentials EFe(III/II), O2 binding equilibrium constants pKO2, and the pKas of the protonated superoxide intermediate ([FeIII(por)(O2•H)]+). All of the catalysts were calculated except for 1, 2, and 8, for which many rotamers exist in solution. The pKa of the protonated superoxide intermediate ([FeIII(por)(O2•H)]+) is directly related to the free energy of protonation of the superoxide, the rate-determining step. The calculations show clear correlations among the computed catalyst EFe(III/II), the pKO2s, and the pKas (Figure ). Electron donating substituents that shift EFe(III/II) to more negative values increase the O2 affinity and the superoxide basicity, increasing the TOF. Quantitatively, using simple transition state theory and Bell–Polanyi–Evans (BPE) arguments, it is possible to show that log(TOF) = −pKO2 + α·pKa + C. Taking α = 1/2 (see Supporting Information), the computations predict a LFER of 43 mV/decade in TOF, in good agreement with the experimental value.[25] For α = 0.23, computations and experiments match exactly.

Figure 3

(A) Computed correlations between O2 binding (pKO2, left) and pKa[FeIII(por)(O2•H)]+ – pKa[DMF-H]+ (right) vs EFe(III/II). (B) Reaction coordinate qualitatively depicting the barrier height change as a function of driving force for proton transfer. (C) Optimized structures of [FeIII(por)(O2•H)]+ for 9–11ACN, depicting the changes in NH···O hydrogen bond length.

The correlations and calculated energetics of elementary steps demonstrate that installing pendant protic or hydrogen bond donor/acceptor functionalities in the second coordination sphere (i.e., 1, 8–11) in most cases has a negligible effect on the TOFs for the ORR. The 2-carboxylic acid catalyst 1DMF has a higher TOF than the isomeric 4-carboxylic acid isomer 7DMF (2040 s–1 vs 15 s–1 in DMF) simply because 1 has the more negative reduction potential. Catalyst 1 and its methyl ester 2 have EFe(III/II) values within 19 mV and very close TOFs (within 20%), in both ACN and DMF, even though only 1 would appear to have proton relays. The presence of potential proton relays does not displace the pKO2 and the pKa of the catalyst from the correlation lines. This conclusion contrasts with the great value of such “proton relays” in electrocatalysts for H+ and CO2 reduction[15,16] and shows that the observation of improved rates for catalysts containing potential relays does not demonstrate a relay effect.

The lack of a “relay effect” arises from a thermodynamic mismatch; the pendant proton donors (PyH+, PhCO2H) are not strong enough to protonate [FeIII(por)(O2•)]. Using extrapolated values from Figure A for 1DMF and 1ACN, this mismatch may be >12 pKa units. The proton must be delivered from the much stronger [DMF-H]+ acid in solution, which is the proton source in both DMF and ACN solvents. While there is no definitive evidence for intramolecular proton transfer in these systems, the second coordination sphere may play a noninnocent role in defining the TOF/η correlations. Such is the case for catalysts 9–11. Despite all having identical ortho-amide NH groups[26] that are highly pKa-mismatched with the external acid, catalysts 9 and 10 fall upon a more efficient TOF/η correlation than 11, discussed below.

Figure C shows two LFER lines of nearly identical slope for this set of catalysts and conditions. The lower η line (to the left) primarily has the systems in ACN, suggesting that this is the better solvent for ORR in general, but 8DMF falls on that line and 11ACN falls on the higher η line with most of the systems in DMF. As emphasized above, the TOFs are limited by O2 binding and the subsequent superoxide ligand protonation. Computationally, there is a single correlation between pKO2 and EFe(III/II) that includes all of the catalysts in both solvents (slope ∼100 mV/unit pKO2, Figure A, left). Therefore, this is not the origin of the better catalysis by 9 and 10. Their advantage originates solely from their better proton transfer thermodynamics, as their protonated superoxide complexes have higher pKa values than would be expected from their EFe(III/II) (Figure A, right). Indeed, 9 and 10 have the most favorable Δ for proton transfer (−3.5 and −1.8 kcal/mol, respectively) of all catalysts analyzed. They are easier to protonate and have higher TOFs. We suspect that similar issues explain why 1, 2, and 8 fall on the better correlation line, but their complicated mixture of rotamers precluded computational analysis, as noted above. The related CF3-substituted amide catalyst 11 does not have this elevated pKa and does not have the elevated TOF (it falls on the LFER line to the right).

To further understand the origin of the distinct pKa values for the amide catalysts 9–11, we compared the optimized structures of the relevant intermediates. There is a hydrogen bond from the amide NH groups in the second coordination sphere to the distal O (Od) of the O2-derived ligand (Figure C). In principle, the hydrogen bonds should be strongest—greatest stabilization by the second coordination sphere—for the more electron rich intermediates. Indeed, for the protonated superoxide complexes [FeIII(por)(O2•H)]+, the calculations show that the intermediate from 9 has the shortest NH–O distance, nearly 0.1 Å shorter than that in 11: d(NH–Od) = 2.864 Å, 2.907 Å, and 2.966 Å for 9–11, respectively. However, the trend is the opposite when comparing the NH–O distances of the [FeIII(por)(O2•)] intermediates, where the NH–Od distances are 2.379 Å, 2.361 Å, and 2.296 Å for 9–11, respectively (see Figure S72). Therefore, proton transfer to form [FeIII(por)(O2H•)]+ requires the largest changes in the NH–O distance for 11 (0.670 Å) when compared to 10 and 9 (0.546 and 0.485 Å). Given that the majority of the other bond lengths and angles are quite similar for 9–11, the NH–O hydrogen bond seems to play an important role in modulating the E1/2 to pKa (and thus TOF) relations. The stronger H-bonding interaction for the protonated superoxide complexes 9 and 10 allow for a more exoergic proton transfer, resulting in TOFs roughly 2 orders of magnitude larger than 11 under identical conditions (after correcting for the difference in overpotentials). This H-bonding effect of the “relay” was unexpected since the proton is delivered from the exogenous donor, [DMF-H+]. Differences in proton transfer energetics may also be the reason for 1 and 2 behaving so much better in ACN than DMF, and also for 8 compared to 1–7. Understanding the better catalysis by these complexes is challenging because of the presence of multiple conformations and/or protonation states. The selective stabilization of one intermediate over another for 9 and 10 is an effect of the second coordination sphere groups, due to hydrogen bonding rather than proton delivery. The occurrence of this effect for complexes 9 and 10 but not for 11 shows the subtle interplay of factors contributing to such second coordination sphere effects.

The combined experimental and theoretical studies reported herein provide a detailed look at the factors that underlie molecular ORR electrocatalysis. Changing substituents on the aryl rings of these metalloporphyrin catalysts changes their EFe(III/II), and all of the key parameters change in parallel for most of the catalysts. This approach of finding linear free energy relationships is the same principle that has become a primary way to analyze sets of heterogeneous catalysts, including electrocatalysts.[27] The widely used “volcano plots” and “scaling relationships” correlate catalyst activity with a single descriptor such as a surface–substrate binding energy or the d-band center, and the relationships hold when the free energies of the different catalytic steps scale linearly with the descriptor. A recent computational study of iron porphyrin ORR catalysts showed the presence of scaling relations for the binding of OH, O, and OOH, but did not consider separate electron or proton transfer steps.[28] The combined experimental and computational study reported here provides a more detailed analysis and shows one way to overcome the scaling relationships. Further studies are in progress to design improved catalysts by further decoupling the properties of the second coordination sphere from those of the first. This is a new design principle for catalysis of multielectron/multiproton reactions.

The results and analysis developed here allow, for the first time, comparisons of molecular O2 reduction electrocatalysts under diverse conditions. The correlations observed in this study show that the ORR activities for catalysts 1–11 correlate with the well-defined overpotentials (η) for these molecular catalysts under a range of conditions, on two parallel linear free energy relationships. The correlations hold for a variety of electrocatalysts and media across 6 orders of magnitude in TOF, and show that the extraordinarily high TOFs for 1 and 2, close to the diffusion limit for the 4e–/4H+ ORR reaction, are mostly due to their high η. These correlations are shown computationally to stem from the chemical steps of O2 binding and proton transfer. The origin of the improved behavior of some of the electrocatalysts is shown to derive from deviations in the correlation of catalyst properties (, KO2) with the basicity of the superoxide intermediate (pKa[Fe(por)(O2•H)+]). This concept of decoupling the thermochemical properties of the first and second coordination spheres is a new approach to catalyst design.