Combining scaling relationships overcomes rate versus overpotential trade-offs in O2 molecular electrocatalysis

See allHide authors and affiliations

Science Advances  13 Mar 2020:
Vol. 6, no. 11, eaaz3318
DOI: 10.1126/sciadv.aaz3318


The development of advanced chemical-to-electrical energy conversions requires fast and efficient electrocatalysis of multielectron/multiproton reactions, such as the oxygen reduction reaction (ORR). Using molecular catalysts, correlations between the reaction rate and energy efficiency have recently been identified. Improved catalysis requires circumventing the rate versus overpotential trade-offs implied by such “scaling relationships.” Described here is an ORR system—using a soluble iron porphyrin and weak acids—with the best reported combination of rate and efficiency for a soluble ORR catalyst. This advance is achieved not by “breaking” scaling relationships but rather by combining two of them. Key to this behavior is a polycationic ligand, which enhances anionic ligand binding and changes the catalyst E1/2. These results show how combining scaling relationships is a powerful way toward improved electrocatalysis.


Improving the rates and efficiencies of electrocatalytic reactions is critical to the development of chemical-to-electrical energy conversion technologies. The oxygen reduction reaction (ORR)—the combination of dioxygen (O2), protons, and electrons to give water—is an important example common to fuel cells (1, 2). Because current ORR technologies use platinum catalysts, current research seeks replacements sourced from earth-abundant materials (3). One approach to developing catalytic systems with high rates [turnover frequencies (TOFs)] and high efficiencies (low overpotentials, η) has been the development of scaling relationships. The properties of heterogeneous electrocatalysts can often be understood and even predicted using a single scaling descriptor, typically the energy of substrate binding to the catalyst surface (4). In contrast, soluble molecular electrocatalysts, which can be studied in more atomistic and mechanistic detail, follow multiple kinetic/thermodynamic scaling relationships (5). These describe how the maximum TOFs [log(TOFmax)] scale with the various terms contributing to the effective overpotential (ηeff) (58) (see the Supplementary Materials). To date, the reported log(TOFmax)/ηeff relationships always describe a trade-off: Faster catalysis is only achieved at lower efficiencies (higher ηeff).

We report here an inverse scaling relationship, one that allows for faster rates at higher efficiencies, using a polycationic iron porphyrin catalyst with buffered weak acids in acetonitrile (MeCN) (Fig. 1A). With acetic acid/acetate buffer (AcOH/AcO), this system achieves a TOFmax of 170 s−1 at 0.54 V ηeff, which is ~104 faster than any previously reported molecular ORR catalyst at this ηeff (2, 9). As described below, this unprecedented result is predictable by combining two kinetic/thermodynamic scaling relationships. The coupling of the two scaling relationships is a result of the electrostatic ligand design of 1 that enables cooperativity between the catalyst and buffer.

Fig. 1 Catalytic system efficiencies, reaction mechanism, and structure of catalyst 1.

(A) Plot of log(TOFmax)/ηeff values and fits (dashed lines) for 1 (diamonds; data and conditions in Table 1) and for previously reported iron porphyrin [Fe(por)] catalysts (circles; 0.1 M [DMF-H+] in DMF or MeCN) (5). The uncertainties are smaller than the data points. The yellow shading indicates an aspirational region. (B) Fe(por) catalyzed O2 reduction mechanism, as described in the main text (11). (C) Drawings of 1 and of the cation in the solid-state x-ray crystal structure, [1•2H2O]OTf5 (Fe, orange; N, blue; C, white; O, red; H atoms omitted for clarity; thermal ellipsoids at 50% probability).


Iron αβαβ-tetra(o-N,N′,N″-trimethylanilinium)porphyrin (1) was prepared as reported (10) and was isolated and characterized as the di-aquo, penta-triflate salt of the αβαβ isomer by single-crystal x-ray diffraction (Fig. 1C). Cyclic voltammograms (CVs) of 1 in MeCN containing 0.1 M tetra-n-butylammonium tetrafluoroborate electrolyte ([n-Bu4N][BF4]) showed three reversible redox features (FeIII/II, FeII/I, and FeI/0 (10). Solutions containing 1, O2, and buffered acid (1:1 [HA]/[A]) showed a large, irreversible current at the FeIII/FeII reduction potential, indicative of catalysis (5, 7, 11). All CV experiments were buffered (1:1 acid-to-conjugate base) to define the ORR equilibrium potential and ηeff, which was calculated using Eq. 1 (5, 7, 11).

Measurements of the H2O/H2O2 selectivity were performed using rotating ring disk electrochemistry (RRDE). In all cases, electrocatalysis was found to give <20% H2O2 (see the Supplementary Materials). This high selectivity for H2O is similar to other iron porphyrin catalysts under comparable conditions (5, 7, 11). With AcOH/AcO buffer, catalysis occurs at potentials below the equilibrium potential for O2 to H2O2 and thus requires thermodynamic selectivity for H2O [cf., (9)].

TOFmax values were determined from the catalytic CVs by foot-of-the-wave analysis (FOWA), Eq. 2 [Table 1; (5, 11, 12)]. This widely used approach normalizes the catalytic currents (ic) to the noncatalytic peak current of the FeIII/II couple (ip) at potentials (E) near the onset of the catalytic wave, where complications such as substrate depletion are minimized. Given the high selectivity for H2O, ncat was taken to be 4 (electrons per O2 reduced), and the conservative σ = 1 value was used (see the Supplementary Materials). Measuring TOFmax values under different conditions showed that the catalytic rate law was first order in [1], PO2, and [HA], similar to other Fe(por) catalysts (5, 7, 11). [HA] is the concentration of free acid after considering homoconjugation (see the Supplementary Materials). These results implicate a catalytic mechanism of (i) initial reduction of [FeIII(por)]+ to FeII(por), (ii) pre-equilibrium O2 binding to form the superoxide complex, FeIII(por)(O2•−), and (iii) rate-limiting proton transfer to form [FeIII(por)(O2H)]+ (Fig. 1B). Experimental and computational studies of this mechanism are reported in (5, 11, 12).ηeff=EORRoE1/22.303RTncatFlog([H2O]2[A]4PO2[HA]4)(0.0592V)pKa(1)icip=2.24ncatσRTFvTOFmax1+exp[FRT(EE1/2)](2)

Table 1 Properties of catalytic systems with 1 and different buffers.

Experimental conditions: 0.1 mM 1, 0.1 M buffer (1:1 HA/A or HB+/B), 0.1 M [n-Bu4N][BF4] in MeCN (~15 mM H2O), 1 atm O2. [DMF-H+], N,N′-dimethylformamidium triflate; TFAH, trifluoroacetic acid; [Lut-H+], lutidinium tetrafluoroborate; SalOH, salicylic acid; BzOH, benzoic acid; AcOH, acetic acid.

View this table:

ORR catalysis was studied in MeCN using 1 and a series of buffers. Using N,N′-dimethylformamidium triflate ([DMF-H+]OTf; pKa =6.1, where Ka is the acid dissociation constant), TOFmax = 8.5 s−1 and ηeff = 1.16 V (Table 1). Within the log(TOFmax)/ηeff space, this [ηeff, log(TOFmax)] data point roughly fits the scaling relationships previously reported for Fe(por) electrocatalysts under similar conditions (Fig. 1A, black diamond) (5). In contrast, using acetic acid (AcOH, pKa = 23.5) gave markedly improved catalysis: a faster TOFmax (170 s−1) at less than half the ηeff (0.54 V). This result contradicts previously derived log(TOFmax)/ηeff relationships, which always predict that a lower overpotential will give a slower rate, as seen for iron tetra-arylporphyrins (see the Supplementary Materials) (7). Catalysis with 1 and trifluoroacetic acid (TFAH), salicylic acid (SalOH), or benzoic acid (BzOH) similarly show improvements in both TOFmax and ηeff versus the [DMF-H+] point (Fig. 1, blue diamonds, and Table 1). Together, the carboxylic acid [ηeff, log(TOFmax)] points define an unexpected inverse log(TOFmax)/ηeff relationship (Fig. 1A, blue dashed line), a previously unrealized goal in molecular electrocatalysis (5, 13, 14). The following sections describe a model that explains this inverse relationship as the sum of two known kinetic/thermodynamic scaling relationships (7).


Catalyst 1 is unique among the Fe(por) series because changing the buffer notably affects two properties of the catalytic system: the acid pKa and also the catalyst E1/2, due to carboxylate binding. Since both pKa and E1/2 each modulate the TOFmax and ηeff in different ways [see the Supplementary Materials and (5, 7) for further discussion], a single scaling relationship is not adequate for predicting the composite changes in log(TOFmax)/ηeff. Instead, the scaling relationships—which only establish directionality within the log(TOFmax)/ηeff space—must be recast as vectors. Vectors have directionality and length and are additive.

The Nernst equation for this ORR reaction (Eq. 1) shows that a one-unit increase in acid pKa decreases ηeff by 0.059 V: ∆ηeff = −(0.059 V)∆pKa (15, 16). In iron porphyrin ORR catalysis, protonation is the rate-determining step, and the TOFmax varies with pKa according to the Brønsted equation: Δlog(TOFmax) = −α(ΔpKa), where α ≈ 0.3 (5, 7, 11). For the range of carboxylate buffers used here, TFAH/TFA to AcOH/AcO, the change in pKa is 10.9. Therefore, with α = 0.3 and holding all other properties constant, the changes in ηeff and log(TOFmax) associated with the 10.9 pKa shift can be described as the vector vpKa in the [ηeff, log(TOFmax)] space (Eq. 3, dec stands for decade in TOFmax; Fig. 3).vpKa=Δηeff,Δlog(TOFmax)=0.059(ΔpKa),α(ΔpKa)=0.64 V, 3.3 dec(3)

In addition to acid pKa, the buffer identity also affects the E1/2. When buffered carboxylic acids are titrated into solutions of 1, the catalyst E1/2 shifts negatively, by as much as 350 mV (Fig. 2A). This shift results from the conjugate base of the buffer (e.g., acetate) binding more strongly to the FeIII versus the FeII form of 1. This behavior was documented and explained many years ago for chloride binding to iron tetraphenylporphyrin (17). At 0.1 M carboxylate buffer, the E1/2 values vary linearly with acid pKa (−28 ± 1 mV/pKa; Fig. 2B). In contrast, buffers with cationic acids and neutral conjugate bases ([DMF-H+]/DMF and [Lut-H+]/Lut) give only small shifts in E1/2 (<50 mV; Fig. 2B). This unique distinction is due to the highly cationic nature of 1, which enhances binding of anionic conjugate bases but not neutral ones (see the Supplementary Materials). Further supporting the importance of the cationic ligand, no notable change in catalyst E1/2 was reported when iron tetraphenylporphyrin—an analog without the cationic trimethylanilinium groups—was combined with similar buffers under similar conditions (7).

Fig. 2 Electrochemical studies of 1 under noncatalytic and catalytic conditions.

(A) CVs of 1 showing the shift of E1/2 with increasing concentrations of 1:1 AcOH/AcO buffer. (B) Plot of E1/2 versus the acid pKa at 0.1 M buffer. (C) Linear sweep voltammograms under catalytic conditions with different BzOH buffer concentrations. (D) FOWA for voltammograms in (C), with fits between ic/ip = 1 to 4.

Changes in E1/2 also affect ηeff: Δηeff = −ΔE1/2 (Eq. 2) (5, 7, 18). For the Fe(por) series, log(TOFmax) has empirically been shown to change according to log(TOFmax) = −18.5 decade/E1/2(V) (see the Supplementary Materials for further discussion) (7). Conceptually, this is because catalysts with more negative E1/2s (and thus higher ηeffs) bind O2 more strongly and form more basic superoxide complexes, both of which lead to higher TOFmax (5). As above, a vector vE1/2 can be used to represent the log(TOFmax)/E1/2 scaling relation and the 0.302 V change in E1/2 upon replacing TFAH/TFA with AcOH/AcO (Eq. 4 and Fig. 3)vE1/2=Δηeff,Δlog(TOFmax)=ΔE1/2, 18.5(ΔE1/2)=0.30 V, 5.6 dec (4)

Fig. 3 Vector analysis to predict the inverse scaling relationship for 1.

Predicted coordinates using the vectors are shown as red squares. (A) Plot of log(TOFmax) versus ηeff for catalytic systems of 1 and varying carboxylic acid buffers (data points match those in Fig. 1). Superimposed vectors (gold, black, and blue) show predicted changes caused by pKa, E1/2, and summative effects, respectively. (B) Buffers with cationic acids follow only the pKa vector (gold). Prior Fe(por) data and E1/2 scaling relationships (7) are included for reference (gray). Uncertainties are smaller than the data points.

Because the buffer affects both the acid pKa and catalyst E1/2, both vpKa and vE1/2 are needed to describe the changes in ηeff and log(TOFmax). The effects of these changes are additive, described by the vector sum: vsum=vpKa+vE1/2 (Eq 5 and Fig. 3). While neither vpKa nor vE1/2 alone fits the observed changes in log(TOFmax)/ηeff, vsum predicts both the directionality and distance from the experimental TFAH coordinate to the AcOH coordinate—within a factor of 3 in TOFmax.vsum=vpKa+vE1/2=0.34 V, 2.3 dec(5)

Qualitatively, using a less acidic buffer improves the overall catalysis because of two factors. The change in E1/2 causes a large gain in TOFmax at relatively little cost to ηeff (a steep scaling slope). In contrast, the concomitant changes in pKa decrease ηeff substantially with only small losses in TOFmax (a shallow slope). The combination is a faster TOFmax at lower ηeff.

While vsum is required to analyze the carboxylate buffer data, only vpKa is needed to predict the change in log(TOFmax)/ηeff for the cationic buffers (e.g., [DMF-H+]/DMF and [Lut-H+]/Lut) (Fig. 3B). This is because the change in cationic buffers affects primarily the pKa, with little effect on the E1/2. The lack of change in E1/2 highlights the key role of the cationic macrocycle, which enhances anion binding but not binding of a neutral ligand. Unlike CO2 electroreduction by catalyst 1 (10), the remarkable log(TOFmax)/ηeff data reported here are not just a feature of the catalyst but rather the combination of the catalyst and buffer.

The summative or “tandem” scaling relationship approach developed here is unexpectedly accurate given that the contributing scaling relationships were obtained for a somewhat different set of catalysts, iron porphyrins with different meso-aryl substituents. Yet, these scaling relationships quantitatively hold for the different axial ligands in the 1 + buffer systems examined here. Paradoxically, the tandem approach circumvents the limitations implied by prior molecular scaling relationships—that there is always a trade-off between TOFmax and ηeff—because of the robustness and generality of the same “limiting” relationships.

Here, we show that kinetic/thermodynamic scaling relationships for molecular electrocatalysts can be additive and that this tandem scaling approach is a powerful and predictive way to improve multistep electrocatalytic processes. Specifically, summing multiple known scaling relationships predicted an unprecedented, inverse log(TOF)/ηeff relationship that allowed for simultaneous improvement in both catalysis rates and efficiencies. This approach has yielded the best reported combination of TOF and ηeff for a soluble ORR catalyst.

The ability to combine scaling relationships should be applicable beyond this case study. Optimization by this method, and perhaps even inverse scaling, only requires that the catalytic reaction have at least two properties that affect the TOFmax and ηeff in different ways (different slopes). Most molecular electrocatalysis meets this two-descriptor requirement because their mechanisms have pre-equilibrium and rate-limiting steps with different stoichiometries and/or different relationships between the k, Keq, and the ∆G° for each step. ORR catalysis by 1 is a special case of the two-descriptor requirement because a single change to the system—changing the buffer—affects both the catalyst E1/2 and the acid pKa, each of which independently affects the pre-equilibrium and rate-determining steps. Ongoing work in our lab is extending this tandem scaling approach to simultaneous but independent changes in E1/2 and pKa, by changing both the catalyst and the buffer. A similar approach may also be applicable to heterogeneous electrocatalysis, since there is increasing recognition that those mechanisms may also have pre-equilibria and kinetic steps with different properties (19). Therefore, this method of combining scaling relationships may have ramifications for the development of both homogeneous and heterogeneous electrocatalysis of energy-important processes.



High-resolution mass spectrometry was performed using a Waters Xevo G2-XS QTof mass spectrometer. Ultraviolet-visible (UV-vis) optical spectra were recorded on an Agilent 8452 diode-array spectrometer and were collected using a 1-cm path length cuvette. Infrared (IR) spectrum was recorded on a Bruker Alpha Fourier transform infrared spectrophotometer equipped with an attenuated total reflectance (ATR) attachment. The electrochemical setup is described below. Information about the x-ray diffractometer is described below.


2-Nitrobenzaldehyde (Sigma-Aldrich, >95%), tin(II) chloride dihydrate (Sigma-Aldrich, >99%), aqueous hydrochloric acid [Macron, 36 to 38 weight % (wt %)], iron (II) bromide (Sigma-Aldrich, >98%), 2,6-lutidine (Sigma-Aldrich, ReagentPlus, >98%), formaldehyde (Sigma-Aldrich, 37 wt %, containing 10 to 15% methanol as stabilizer), sodium cyanoborohydride (Acros, >98%), methyltrifluoromethylsulfonate (MeOTf; Sigma-Aldrich, >98%), trifluoromethylsulfonic acid (Acros, 99%), neutral aluminum oxide (Sigma-Aldrich, >99%), silica (Sigma-Aldrich, >99%), sodium chloride (Sigma-Aldrich, >99%), sodium bicarbonate (Sigma-Aldrich, >99%), ammonium hydroxide solution (Sigma-Aldrich, 25 wt %), AcOH (Sigma-Aldrich, >99.9%), BzOH (Sigma-Aldrich, >99.5%), SalOH (Sigma-Aldrich, >99%), TFA (Sigma-Aldrich, ReagentPlus, >99%), tetra-n-butylammonium acetate (Sigma-Aldrich, 97%), tetra-n-butylammonium benzoate (Sigma-Aldrich, >99%), tetra-n-butylammonium salicylate [Tokyo Chemical Industry (TCI), 98%], and sodium trifluoroacetate (Sigma-Aldrich, 98%) were all used as received. All of the solid chemicals were stored in a N2 glovebox when not in use.

Tetrahydrofuran, N,N′-dimethylformamide (DMF), MeCN, and diethyl ether were all degassed with argon and dried using a Pure Process Technology solvent system before use. Tetra-n-butylammonium tetrafluoroborate (Acros, >98%) was stored in a desiccator containing Drierite (calcium sulfate). Bis(cyclopentadienyl)iron(II) (ferrocene or Fc; Sigma-Aldrich, 95%) was recrystallized two times from hexanes before being dried in a N2 glovebox. Pyrrole (Acros, 99%) was freshly distilled immediately before each use. The Pur-A-Lyzer Mega Dialysis Kit [Sigma-Aldrich, 1 kDa molecular weight cut-off (MWCO)] was soaked in Milli-Q water for 1 hour before dialysis of 1. Dioxygen (Airgas, Ultra High Purity) and argon (Airgas, Ultra High Purity) were used as received. N,N′-dimethylformamidium triflate ([DMF-H]OTf) and 2,6-lutidinium triflate ([Lut-H]OTf) were synthesized using previously reported methods (20, 21).

Synthesis of iron αβαβ-tetra(o-N,N′,N″-trimethylanilinium)porphyrin (1)

The synthesis of the known iron(III) α,β,α,β-5,10,15,20-tetra(ortho-N,N′,N″-trimethylanilinium)-porphyrin penta(trifluoromethanesulfonate), 1, was slightly adapted from literature preparation (10). The α,β,α,β-5,10,15,20-tetra(ortho-aminophenyl)porphyrin atropoisomer was isolated chromatographically using the conditions described in (10). Later in the synthesis, during the final methylation reaction of iron(III) α,β,α,β-5,10,15,20-tetra(ortho-N,N′-dimethylaminophenyl)porphyrin chloride [chart S1, product 4, from reference (10)] to the fully quaternized product, 300-equivalent MeOTf were used in place of the reported 100 equivalents. Full quaternization was supported by the high-resolution mass spectrum (HRMS) (which did not reveal any partially quaternized products) and by cyclic voltammetry (see the Supplementary Materials). The product, which was collected in identical yields to the initial report, was purified by dialysis against Milli-Q water (10 ml of sample inside dialyzing bag, 250 ml of surrounding water replaced every 90 min for a total of 12 hours). The product was then recrystallized via slow vapor diffusion of diethyl ether into MeCN containing the iron porphyrin. This recrystallization method yielded crystals suitable for x-ray diffraction (see the Supplementary Materials). HRMS [electron spray ionization (ESI)/quadrupole orthogonal acceleration–time-of-flight) mass/charge ratio (m/z): ([M]5+ + 4OTf) calculated (calcd) for C56H60FeN8(CF3SO3)4 1496.2373; found 1496.2493. ([M]5+ + 3OTf) calcd for C56H60FeN8(CF3SO3)3 673.6426; found 673.6395. UV-vis and IR spectra also agree with what was reported in reference (10), see the Supplementary Materials.

Electrochemical methods

CV was performed on a CH Instruments model 650D potentiostat. RRDE was conducted using a BASi Epsilon potentiostat and a Pine Instruments rotator (see the Supplementary Materials for further details). CVs used a 3-mm glassy carbon working electrode, a platinum wire counter electrode, and a Ag wire pseudoreference. The Ag wire pseudoreference was prepared using the method reported by Dempsey et al. (22), where a silver wire was sanded and fit snuggly into a capillary containing MeCN and 0.1 M [n-Bu4N][BF4]. The solution inside the capillary was separated from bulk solutions using a Vycor tip that had been mechanically sealed to the capillary using heat shrink tubing. The capillary was stored in a solution containing 0.1 M [n-Bu4N][BF4] in-between experiments. For analysis, each voltammogram was internally referenced to ferrocene. The glassy carbon working electrodes were polished after every voltammogram by vigorously polishing on a Buehler felt pad to a mirror-like finish using an alumina slurry (wetted 0.05-μm alumina powder). After polishing, the electrode was rinsed with water and MeCN. Internal resistance (iR) compensation was performed before each voltammogram using the CH Instruments integrated software. The typical resistance value was <60 ohm. Considering the typical amount of current passed during catalysis (~200 μA), the iR compensation shifted the potential response of <10 mV for most experiments.

All of the buffered electrochemical experiments were performed with a one-to-one buffer of acid and conjugate base. Throughout, we identify the initial buffer concentrations rather than expressly list the acid and conjugate base concentrations. For example, a “10 mM AcOH buffer” means that the solution was prepared to initially contain 10 mM AcOH and 10 mM AcO. These values do not necessarily reflect the concentrations of nonhomoconjugated acid and base (see the Supplementary Materials).

Electrocatalytic, aerobic measurements were performed after sparging the buffered solution and headspace with O2. The sparging was performed with pure O2 at 1 atm unless otherwise noted. To minimize solvent evaporation and accompanying temperature changes, the O2 was first sparged through an initial bubbler containing pure MeCN. Between each catalytic voltammogram, the working electrode was vigorously polished (see above). To ensure reproducibility, each catalytic voltammogram was also performed in duplicate, and the rates were averaged (see the Supplementary Materials). Unless otherwise specified, every voltammogram reported was collected at 0.1 V/s in a MeCN solution containing 0.1 mM 1 and 0.1 M [n-Bu4N][BF4].

Single-crystal x-ray diffraction

Crystals of [Fe-o-TMA]OTf5•2H2O (identification code 007b-17062) were grown by slow vapor diffusion of diethyl ether into MeCN on the benchtop. Two water molecules were identified as axial ligands in the solid-state structure. Low-temperature diffraction data (ω-scans) were collected on a Rigaku MicroMax-007HF diffractometer coupled to a Saturn994+ charge-coupled device detector with Cu Kα (λ = 1.54178 Å) for the structure of [Fe-o-TMA]OTf5•2H2O. The diffraction images were processed and scaled using Rigaku Oxford Diffraction software. The structure was solved with SHELXT and was refined against F2 on all data by full-matrix least squares with SHELXL (23). The data were refined as an inversion twin. The fractional volume contributions of the second twin component were refined to a value of 0.432(18). All nonhydrogen atoms were refined anisotropically. Hydrogen atoms were included in the model at geometrically calculated positions and refined using a riding model. The isotropic displacement parameters of all hydrogen atoms were fixed 1.2 times the U value of the atoms to which they are linked (1.5 times for methyl groups). The Fe-O distance was measured to be 2.101(8) Å, which suggested the assignment as water.

The asymmetric unit of the crystallographic model contains 0.25 of the formula unit C56H64FeN8O2•5(CF3O3S). The iron resides on the crystallographic 4¯ rotation axis at (0, ½, ¼). The asymmetric unit also contains a quarter of the porphyrin and 1.25 triflates. One triflate site is on a general position (x,y,z), and the other site is near the crystallographic 4¯ rotation axis (½, ½, ½).

The triflate on the general position is disordered over two sites. The thermal ellipsoids for the triflate atoms were restrained to behave as rigid bodies. The site occupancies were fixed at 0.50, and all chemically equivalent 1,2 and 1,3 distances were restrained to be similar. The other triflate was disordered with respect to the 4¯ rotation axis. This modeled triflate was placed as a constrained, rigid group based on the difference map (see the Supplementary Materials). The special position constraints were suppressed, and the model occupancy was fixed at 0.25. The program SQUEEZE (24) was used to compensate for the contribution of disordered solvents contained in voids within the crystal lattice from the diffraction intensities. This procedure was applied to the data file, and the submitted model is based on the solvent removed data. On the basis of the total electron density found in the voids (181 e/Å3), it is possible that some combination of crystallization solvents is present in the unit cell. See “_platon_squeeze_details” in this .cif for more information.

The full details of the x-ray structure determination (CIF) can be found in the Cambridge Crystallographic Data Center (CCDC), number 1947226 ([Fe-o-TMA]OTf5•2H2O). These data can be obtained free of charge from the CCDC via


Supplementary material for this article is available at

Section S1. Spectroscopic characterization of 1

Section S2. Voltammetry of 1

Section S3. UV-vis spectroscopy of 1 + varying buffers

Section S4. Homoconjugation

Section S5. Kinetic analysis

Section S6. Effective overpotential determination

Section S7. Selectivity for H2O versus H2O2

Section S8. Conceptual background for E1/2 and pKa scaling relationships

Section S9. Single crystal x-ray structure

Fig. S1. Full high-resolution ESI mass spectrum of 1 with identified peaks, as labeled.

Fig. S2. High-resolution ESI mass spectrum and isotopic fits for the [M5+ + 3OTf] ion.

Fig. S3. UV-vis spectrum of 1 in N,N′-dimethylformamide.

Fig. S4. IR spectrum of 1.

Fig. S5. A CV of an Ar-sparged solution of 1.

Fig. S6. Scan rate investigation of the FeIII/FeII redox couple of 1 in unbuffered solution.

Fig. S7. Voltammograms of an Ar-sparged solution of 1 before and after addition of buffered AcOH.

Fig. S8. Investigation of the FeIII/FeII redox couple of 1 with titrations of AcOH buffer.

Fig. S9. Scan rate investigation of the FeIII/FeII redox couple of 1 in AcOH-buffered solution.

Fig. S10. Investigation of the FeIII/FeII redox couple of 1 with titrations of BzOH buffer.

Fig. S11. Scan rate investigation of the FeIII/FeII redox couple of 1 in BzOH-buffered solution.

Fig. S12. Investigation of the FeIII/FeII redox couple of 1 with titrations of SalOH buffer.

Fig. S13. Scan rate investigation of the FeIII/FeII redox couple of 1 in SalOH-buffered solution.

Fig. S14. Investigation of the FeIII/FeII redox couple of 1 with titrations of TFA buffer.

Fig. S15. Scan rate investigation of the FeIII/FeII redox couple of 1 in TFA-buffered solution.

Fig. S16. Investigation of the FeIII/FeII redox couple of 1 with titrations of [DMF-H]OTf-buffer.

Fig. S17. Investigation of the FeIII/FeII redox couple of 1 with titrations of [Lut-H]BF4 buffer.

Fig. S18. Changes in E1/2(FeIII/FeII) with varying buffers (and concentrations).

Fig. S19. CVs of an O2-sparged MeCN solution containing 100 mM [AcOH] buffer before (black) and after (blue) adding 30 μM 1.

Fig. S20. Voltammograms of 1 with AcOH buffer under various solution conditions.

Fig. S21. Voltammograms of 1 with BzOH buffer under various solution conditions.

Fig. S22. Voltammograms of 1 with SalOH buffer under various solution conditions.

Fig. S23. Voltammograms of 1 with TFA buffer under various solution conditions.

Fig. S24. Voltammograms of 1 with [Lut-H+] buffer under various solution conditions.

Fig. S25. Voltammograms of 1 with [DMF-H]OTf buffer under various solution conditions.

Fig. S26. Rinse tests for all of the buffers used in this study.

Fig. S27. UV-vis spectra of MeCN solutions containing 1 (~0.05 mM), [n-Bu4N][BF4] (~0.05 M), and varying 1:1 buffers (~0.05 M, as identified).

Fig. S28. Foot-of-the-wave analysis for the buffer concentrations used in this study (all at 1 atm O2).

Fig. S29. Foot-of-the-wave analysis for all the partial pressure O2 measurements performed in this study (all at 20 mM buffer).

Fig. S30. TOFmax versus [substrate] plots for the buffers used in this study.

Fig. S31. Plot of kobs versus partial pressure of O2 for each of the buffers used in this study.

Fig. S32. RRDE analysis using ferrocene to estimate collection efficiencies.

Fig. S33. RRDE analysis for the ORR catalyzed by 1 using [DMF-H]OTf buffer.

Fig. S34. RRDE analysis for the ORR catalyzed by 1 using TFAH buffer.

Fig. S35. RRDE analysis for the ORR catalyzed by 1 using SalOH buffer.

Fig. S36. RRDE analysis for the ORR catalyzed by 1 and BzOH buffer.

Fig. S37. RRDE analysis for the ORR catalyzed by 1 and AcOH buffer.

Fig. S38. The complete x-ray model of [Fe-o-TMA]OTf5•2H2O represented with balls and sticks.

Fig. S39. The complete numbering scheme of the cation-only portion of [Fe-o-TMA]OTf5•2H2O with 50% thermal ellipsoid probability levels.

Fig. S40. The x-ray model of the cation-only portion of [Fe-o-TMA]OTf5•2H2O from a secondary perspective showing the αβαβ isomer of porphyrin structure.

Fig. S41. The complete numbering of the disordered triflate at a general position in the model of [Fe-o-TMA]OTf5•2H2O with 50% thermal ellipsoid probability levels.

Fig. S42. The unit cell of [Fe-o-TMA]OTf5•2H2O, with a surface that represents a level of 1.5 e/Å3.

Fig. S43. All orientations of the disordered triflate are shown in relation to the 4¯ rotation axis, represented with red lines.

Table S1. Summary of E1/2(FeIII/FeII) (V) versus Fc+/Fc values measured under the conditions reported in this study; errors are ±0.005 V.

Table S2. Q-band region λmax values for ~0.05 mM 1 in MeCN containing ~0.05 M [n-Bu4N][BF4] and ~0.05 M buffer.

Table S3. Homoconjugation formation constants for the buffers used in this study.

Table S4. Calculated values for nonhomoconjugated acid and conjugate base ([HA]free = [A]free and [HB+]free = [B]free) for varying buffer identities and concentrations.

Table S5. Average percent H2O2 formed for 1-catalyzed ORR in MeCN containing various buffers.

Table S6. Crystal data and structure refinement for [Fe-o-TMA]OTf5•2H2O.

Table S7. Atomic coordinates (×104) and equivalent isotropic displacement parameters (Å2 × 103) for [Fe-o-TMA]OTf5•2H2O.

Table S8. Bond lengths (Å) and angles (°) for [Fe-o-TMA]OTf5•2H2O.

Table S9. Anisotropic displacement parameters (Å2 × 103) for [Fe-o-TMA]OTf5•2H2O.

Table S10. Hydrogen coordinates (×104) and isotropic displacement parameters (Å2 × 103) for [Fe-o-TMA]OTf5•2H2O.

Table S11. Torsion angles (°) for [Fe-o-TMA]OTf5•2H2O.

References (2541)

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We acknowledge M. L. Pegis for very helpful discussions and F. Menges of the Yale Chemical and Biophysical Instrumentation Center for collecting the HRMS data. Funding: We acknowledge support from the NSF grant 1828190 for support of mass spectrometry facilities. This research was supported as part of the Center for Molecular Electrocatalysis, an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences. D.J.M. gratefully recognizes support from an NSF Graduate Research Fellowship. Author contributions: D.J.M. and J.M.M. conceived the project, constructed the scientific arguments, and wrote the paper. B.Q.M. performed x-ray crystallography and solved the structures. D.J.M. performed all other experiments and analyzed and interpreted the data. Competing interests: The authors declare that they have no competing interests. Data and materials availability: Crystallographic data are available free of charge from the Cambridge Crystallographic Data Centre under reference number 1947226 (1). All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.
View Abstract

Stay Connected to Science Advances

Navigate This Article