Fe-Mediated Nitrogen Fixation with a Metallocene Mediator: Exploring pK_a Effects and Demonstrating Electrocatalysis

Abstract

Substrate selectivity in reductive multi-electron/proton catalysis with small molecules such as N₂, CO₂, and O₂ is a major challenge for catalyst design, especially where the competing hydrogen evolution reaction (HER) is thermodynamically and kinetically competent. One common strategy to achieve selectivity is to limit the direct reaction between acid and reductant with the intent of slowing background HER. In this study, we investigate how the selectivity of a tris(phosphine)borane iron(I) catalyst, P₃^BFe⁺, for catalyzing the nitrogen reduction reaction (N₂RR, N₂-to-NH₃ conversion) versus HER changes as a function of acid pK_a. We find that there is a strong correlation between pK_a and N₂RR efficiency. Stoichiometric studies indicate that the anilinium triflate acids employed are only compatible with the formation of early stage intermediates of N₂ reduction (e.g., Fe(NNH) or Fe(NNH₂)) in the presence of the metallocene reductant Cp*₂Co. This suggests that the interaction of acid and reductant is playing a critical role in N–H bond forming reactions. DFT studies identify a protonated metallocene species as a strong PCET donor and suggest that it should be capable of forming the early stage N–H bonds critical for N₂RR. Furthermore, DFT studies also suggest that the observed pK_a effect on N₂RR efficiency is attributable to the rate and thermodynamics, of Cp*₂Co protonation by the different anilinium acids. Experimental support for the hypothesis that Cp*₂Co plays a critical role in P₃^BFe-catalyzed N₂RR comes from electrochemical studies. Inclusion of Cp*₂Co⁺ as a co-catalyst in controlled potential electrolysis experiments leads to improved yields of NH₃. The data presented provide what is to our knowledge the first unambiguous demonstration of electrocatalytic nitrogen fixation by a molecular catalyst (up to 6.7 equiv NH₃ per Fe at −2.1 V vs Fc^+/0). While the electrocatalysis is modest in terms of turnover, the comparatively favorable Faradaic efficiencies for NH₃ (up to 31%) highlight the value of studying molecular N₂RR catalysts to define design criteria for selective N₂RR electrocatalysis. Our collective results contribute to a growing body of evidence that metallocenes may play multiple roles during reductive catalysis. While they can behave as single electron transfer (SET) reagents in the reductive protonation of small molecule substrates, ring-functionalized metallocenes, previously considered as intermediates of background HER, can also play a critical role in productive substrate bond-forming steps.

Graphical abstract

INTRODUCTION

There has been substantial recent progress in the development of soluble, well-defined molecular catalysts for N₂-to-NH₃ conversion, commonly referred to as the nitrogen reduction reaction (N₂RR).¹ Nevertheless, a significant and unmet challenge is to develop molecular catalysts, and conditions, compatible with electrocatalytic N₂RR. Progress in this area could have both fundamental and practical benefits, including access to informative in situ mechanistic studies via electrochemical techniques, and an electrochemical means to translate solar or otherwise derived chemical currency (H⁺/e⁻) into NH₃. The latter goal, which has been the subject of numerous studies using heterogeneous catalysts, is key to the long-term delivery of new ammonia synthesis technologies for fertilizer and/or fuel.²

Many soluble coordination complexes are now known that electrocatalytically mediate the hydrogen evolution reaction (HER),³ the carbon dioxide reduction reaction (CO₂RR),⁴ and the oxygen reduction reaction (O₂RR).⁵ The study of such systems has matured at a rapid pace in recent years, coinciding with expanded research efforts towards solar-derived fuel systems. In this context, it is noteworthy how little corresponding progress has been made towards the discovery of soluble molecular catalysts that mediate electrocatalytic N₂RR. To our knowledge, only two prior systems address this topic directly.^6,7,8

More than three decades ago, Pickett and coworkers reported that a Chatt-type tungsten-hydrazido(2−) complex could be electrochemically reduced to release ammonia (and trace hydrazine), along with some amount of a reduced tungsten-dinitrogen product; the latter species serves as the source of the tungsten-hydrazido(2−) complex (via its protonation by acid).^6a By cycling through such a process, an electrochemical, but not an electrocatalytic, synthesis of ammonia was demonstrated. Indeed, efforts to demonstrate electrocatalysis with this and related systems instead led to substoichiometric NH₃ yields.^6c

An obvious limitation to progress in electrochemical N₂RR by molecular systems concerns the small number of synthetic N₂RR catalysts that have been available for study; it is only in the past five years that sufficiently robust catalyst systems have been identified to motivate such studies. In addition, the conditions that have to date been employed to mediate N₂RR have typically included non-polar solvents, such as heptane, toluene, and diethyl ether (Et₂O), that are not particularly well-suited to electrochemical studies owing to the lack of compatible electrolytes.¹

Nevertheless, several recent developments, including ones from our lab, point to the likelihood that iron (and perhaps other) molecular coordination complexes may be able to mediate electrocatalytic N₂RR in organic solvent. Specifically, our lab recently reported that a tris(phosphine)borane iron complex, P₃^BFe⁺, that is competent for catalytic N₂RR with chemical reductants, can also mediate electrolytic N₂-to-NH₃ conversion,^6d with the available data (including that presented in this study) pointing to bona fide electrocatalysis in Et₂O.

Focusing on the P₃^BFe⁺ catalytic N₂RR system, a development germane to the present study was its recently discovered compatibility with reagents milder than those that had been originally employed.^1c Thus, decamethylcobaltocene (Cp*₂Co) and diphenylammonium acid are effective for N₂RR catalysis; these reagents give rise to fast, and also quite selective (> 70% vs HER), N₂RR catalysis at low temperature and pressure in ethereal solvent. In addition, based on preliminary spectroscopic evidence and density functional theory (DFT) predictions, it appears that a protonated metallocene species, Cp*(η⁴-C₅Me₅H)Co⁺, may be an important intermediate of N₂RR catalysis under such conditions. Indeed, we have suggested that Cp*(η⁴-C₅Me₅H)Co⁺ may serve as a proton-coupled-electron-transfer (PCET) donor (BDE_C–H(calc) = 31 kcal mol⁻¹), thereby mediating net H-atom transfers to generate N–H bonds during N₂RR.⁹ The presence of a metallocene mediator might, we therefore reasoned, enhance N₂RR during electrocatalysis.

We present here a study of the effect of pK_a on the selectivity of P₃^BFe⁺ for N₂RR vs HER. By using substituted anilinium acids, we are able to vary the acid pK_a over 9 orders of magnitude and find that the selectivity is highly correlated with the pK_a. In our efforts to investigate the origin of the observed pK_a effect, we found, to our surprise, that in stoichiometric reactions, the catalytically competent anilinium triflate acids are unable to faciltate productive N–H bond formation with early-stage N₂-fixation intermediates. We therefore hypothesize that the formation of a protonated metallocene species, Cp*(η⁴-C₅Me₅H)Co⁺, plays a critical role in N–H bond-forming reactions, either via PCET, PT, or a combination of both during N₂RR catalysis. DFT studies support this hypothesis and also establish that the the observed pK_a correlation with N₂RR selectivity can be explained by the varying ability of the acids to protonate Cp*₂Co. The suggested role of this protonated metallocene intermediate in N–H bond forming reactions led us to test the effect of Cp*₂Co⁺ as an additive in the electrolytic synthesis of NH₃ mediated by P₃^BFe⁺. We find that the addition of co-catalytic Cp*₂Co⁺ enhances the yield of NH₃ without decreasing the Faradaic efficiency (FE), and furnishes what is to our knowledge the first unambiguous demonstration of electrocatalytic N₂RR mediated by a soluble, molecular coordination complex.

RESULTS AND DISCUSSION

pK_a studies

In our recent study on the ability of P₃^BFe⁺ to perform N₂RR with Cp*₂Co as the chemical reductant,⁹ we found that there was a marked difference in efficiency for NH₃ generation with diphenylammonium triflate ([Ph₂NH₂][OTf]) versus anilinium triflate ([PhNH₃][OTf]). In that study, we posited that this difference could arise from several possibilities, including the differential solubility, sterics, or pK_a’s of these acids.⁹

To investigate the last possibility, we have studied the efficiency of the catalysis by quantifying the NH₃ and H₂ produced when using substituted anilinium acids with different pK_a values (Table 1). The table is organized in increasing acid strength, from [^4-OMePhNH₃][OTf] as the weakest acid to the perchlorinated derivative ([^per-ClPhNH₃][OTf]) as the strongest. Importantly, good total electron yields (85.8 ± 3.3) were obtained in all cases. As can be seen from the table, the NH₃ efficiencies are found to be strongly correlated with pK_a.¹⁰

Table 1.

Literature and calculated pK_a values and efficiencies observed in catalytic N₂-to-NH₃ conversion

	pKaexp (THF)10	pKacalc (298 K)a	pKdcalc (195 K)b	Equiv of NH3/Fe	% yield of NH3/e−	% yield of H2/e−c	Total Yield/e−
[4-OMePhNH3][OTf]	8.8	9.6	15.7	0.04 ± 0.01	0.2 ± 0.1	89.1 ± 0.2	89.3
[PhNH3][OTf]	7.8	7.7	13.8	7.3 ± 0.1	40.4 ± 0.5	48.6 ± 0.7	87.5
[2,6-MePhNH3][OTf]	6.8	7.3	13.2	8.6 ± 0.7	47.5 ± 4.0	37.8 ± 0.2	85.6
[Cp*(exo-η4-C5Me5H)Co][OTf]	N/A	9.2	11.8	―	―	―	―
[2-ClPhNH3][OTf]	5.6	5.6	6.0	10.7 ± 0.1	53.9 ± 0.4	26.1 ± 1.9	80.0
[2,5-ClPhNH3][OTf]	4.3	4.0	5.0	13.9 ± 0.7	77.5 ± 3.8	10.5 ± 1.1	87.7
[2,6-ClPhNH3][OTf]	3.4	3.4	3.4	13.8 ± 0.9	76.7 ± 4.9	12.6 ± 2.5	89.3
[2,4,6-ClPhNH3][OTf]	2.1	2.7	1.8	12.8 ± 0.4	70.9 ± 2.2	12.0 ± 0.8	83.1
[per-ClPhNH3][OTf]	1.3	0.8	0.4	3.6 ± 0.1	19.9 ± 0.5	63.5 ± 1.1	83.5

[^4-OMePhNH₃][OTf] = 4-methoxyanilinium triflate, [PhNH₃][OTf] = anilinium triflate, [^2,6-MePhNH₃][OTf] = 2,6-dimethylanilinium triflate, [^2-ClPhNH₃][OTf] = 2-chloroanilinium triflate, [^2,5-ClPhNH₃][OTf] = 2,5-dichloroanilinium triflate, [^2,6-ClPhNH₃][OTf] = 2,6-dichloroanilinium triflate, [^2,4,6-ClPhNH₃][OTf] = 2,4,6-trichloroanilinium triflate, [^per-ClPhNH₃][OTf] = 2,3,4,5,6-pentachloroanilinium triflate.

^aAcidities calculated at 298 K in THF and referenced to the known literature value for [^2,6-ClPhNH₃][OTf].

^bAll species calculated as the ion-paired OTf⁻ species in Et₂O at 195 K and referenced to the known literature value for [^2,6-ClPhNH₃][OTf] in THF.

In particular, a comparison of the efficiency for NH₃ with the pK_a of the anilinium acid used gives rise to four distinct activity regimes (Table 1, Figure 1). One regime that is completely inactive for N₂RR, but active for HER, is defined by the weakest acid, [^4-OMePhNH₃][OTf] (pK_a = 8.8).¹¹ A gradual increase in observed NH₃ yields, coupled with a decrease in H₂ yield, comprises a second regime, in which the acid is strengthened from [PhNH₃][OTf] (pK_a = 7.8), to [^2,6-MePhNH₃][OTf] (pK_a = 6.8), to [^2-ClPhNH₃][OTf] (pK_a = 5.6). Yet stronger acids, [^2,5-ClPhNH₃][OTf] (pK_a = 4.3), [^2,6-ClPhNH₃][OTf] (pK_a = 3.4), and [^2,4,6-ClPhNH₃][OTf] (pK_a = 2.1), constitute another, most active N₂RR regime, one in which the H₂ yields are nearly invariant.¹² The highest selectivity for N₂RR (~ 78%) was observed using [^2,5-ClPhNH₃][OTf] as the acid. A final regime of very low N₂RR activity is encountered with [^per-ClPhNH₃][OTf] (pK_a = 1.3) as the acid. We suspect this last acid undergoes unproductive reduction via ET, thereby short-circuiting N₂RR. The only other N₂RR system for which this type of acid-dependent correlation has been systematically studied is the enzyme MoFe-nitrogenase.^13,14 As shown in Figure 1, the N₂RR vs HER activity of P₃^BFe⁺ as a function of acid strength, is, in broad terms, similar to the behavior of the enzyme¹³ across a wide pH range.

Figure 1.

(top) Percentage of electrons being used to form NH₃ or H₂ at different pH values by the FeMo-nitrogenase in A. vinelandii. Reprinted with permission from Pham, D. N.; Burgess, B. K. Biochemistry 1993, 32, 13725. Copyright 1993 American Chemical Society. (bottom) Percentage of electrons being used to form NH₃ or H₂ at different pK_a values by P₃^BFe⁺.

In a previous study of Cp*₂Co-mediated N₂RR by P₃^BFe⁺,⁹ we identified that P₃^BFeN₂⁻ forms under the catalytic conditions. Earlier studies on the reactivity of P₃^BFeN₂⁻ with an excess of soluble acids, including HOTf and [H(OEt₂)₂][BAr^F₄] (HBAr^F₄, BAr^F₄ = tetrakis(3,5-bis(trifluoromethyl)phenyl)borate)), at −78 °C in Et₂O, established rapid formation of the doubly protonated species, P₃^BFeNNH₂⁺.¹⁵ Recent computational work from our group suggests that, under catalytic conditions with a soluble acid, different efficiencies for N₂RR (versus HER) by P₃^EFe catalysts (E = B, C, Si) are likely correlated to the rate of formation and consumption of early N₂RR intermediates (i.e., P₃^EFeNNH and P₃^EFeNNH₂^+/0).¹⁶ Thus, we were interested in the reactivity of these anilinium triflate acids with P₃^BFeN₂⁻, reasoning they may show differential efficiency in the formation of P₃^BFeNNH₂⁺.

To our surprise, a freeze-quench EPR spectrum of the reaction of excess [^2,6-ClPhNH₃][OTf] (high N₂RR efficiency regime) at −78 °C in Et₂O with P₃^BFeN₂⁻ does not show any P₃^BFeNNH₂⁺. Also, freeze-quench Mössbauer analysis shows the formation of the oxidized products P₃^BFeN₂ and P₃^BFe⁺, but nothing assignable to P₃^BFeNNH₂⁺ (see SI for relevant spectra). Analysis of such a reaction for NH₃ or N₂H₄ after warming shows no fixed-N products. The observation of exclusive oxidation, rather than productive N–H bond formation, is analogous to what is observed upon addition of 1 equiv of a soluble acid (HBAr^F₄ or HOTf) to P₃^BFeN₂⁻. We have previously suggested that if unstable P₃^BFeNNH is formed (eq 1) without excess acid to trap it (to form more stable P₃^BFeNNH₂⁺, eq 2), then it can decay bimolecularly with the loss of 1/2 H₂ to form P₃^BFeN₂ (eq 3).

$${{\mathrm{P}}_3}^{\mathrm{B}}{{\text{FeN}}_2}^{-}+{\mathrm{H}}^{+}\to {{\mathrm{P}}_3}^{\mathrm{B}}\text{FeNNH}$$ (1)

$${{\mathrm{P}}_3}^{\mathrm{B}}\text{FeNNH}+{\mathrm{H}}^{+}\to {{\mathrm{P}}_3}^{\mathrm{B}}{{\text{FeNNH}}_2}^{+}$$ (2)

$${{\mathrm{P}}_3}^{\mathrm{B}}\text{FeNNH}\to {{\mathrm{P}}_3}^{\mathrm{B}}{\text{FeN}}_2+1/2{\mathrm{H}}_2$$ (3)

The low solubility of the anilinium triflate acids studied herein, in excess (25 equiv) and under the catalytically relevant conditions (Et₂O, −78 °C), likely leads to a similar scenario; consequently, P₃^BFeNNH that is generated is not efficiently captured by excess acid, leading instead to bimolecular H₂ loss. In accord with this idea, a freeze-quench EPR spectrum of the addition of 25 equiv of [^2,6-ClPhNH₃][BAr^F₄], a far more ether soluble derivative of the same anilinium, to P₃^BFeN₂⁻ shows P₃^BFeNNH₂⁺ formation, and the detection of fixed-N products upon warming (0.20 ± 0.04 eq. NH₃ per Fe).

These observations must next be reconciled with the seemingly contradictory observation that comparatively efficient N₂RR catalysis is observed when [^2,6-ClPhNH₃][OTf], and other anilinium triflate acids, are employed under catalytic conditions. For example, [Ph₂NH₂][OTf] leads to better efficiency for NH₃ formation versus [Ph₂NH₂][BAr^F₄] (72 ± 3% and 42 ± 6%, respectively). A key difference between the stoichiometric reactions described above, and the catalytic reaction, is the presence of Cp*₂Co in the latter.

We have suggested that Cp*₂Co can be protonated under the catalytic reaction conditions, to form Cp*(η⁴-C₅Me₅H)Co⁺,⁹ which may then play a role in N–H bond forming steps.¹⁷ The results presented here (and below) suggest that such a mechanism is not only plausible, but is likely necessary, to explain the observed catalytic results with anilinium triflate acids. Given the effect of pK_a on the efficiency for N₂RR, we now hypothesize that this effect can arise from the relative energetics and rates of Cp*₂Co protonation by the different anilinium triflate acids.

Computational Studies

To investigate the kinetics and thermodynamics of Cp*₂Co protonation by anilinium triflate acids we turned to a computational study. DFT-D₃¹⁸ calculations were undertaken at the TPSS/def2-TZVP(Fe); def2-SVP¹⁹ level of theory, as used previously for studies of this P₃^BFe⁺ system.²⁰ The free energy of H⁺ exchange (ΔG_a) was calculated for all of the anilinium acids used (representative example shown in eq 4), and also for Cp*(exo-η⁴-C₅Me₅H)Co⁺, in Et₂O at 298 K. These free energies were then used to determine the acid pK_a’s, with inclusion of a term to reference them to the literature pK_a value for [^2,6-ClPhNH₃][OTf] at 298 K in THF (eq 5).

$${\text{PhNH}}_2+{}^{2,6-\text{Cl}}{{\text{PhNH}}_3}^{+}\to {{\text{PhNH}}_3}^{+}+{}^{2,6-\text{Cl}}{\text{PhNH}}_2$$ (4)

$$\mathrm{p}{K}_{\mathrm{a}}({{\text{PhNH}}_3}^{+})=-\mathrm{\Delta }{\mathrm{G}}_{\mathrm{a}}/(2.303\times \text{RT})+\mathrm{p}{K}_{\mathrm{a}}({}^{2,6-\text{Cl}}{{\text{PhNH}}_3}^{+})$$ (5)

Because we presume that variable triflate hydrogen bonding effects (0.5–10 kcal mol⁻¹) are likely to be important under the catalytic conditions (low temperature and low polarity solvent), we additionally calculated the free energy for net HOTf exchange reactions (ΔG_d) at 195 K in Et₂O (representative example shown in eq 6). The free energies of these reactions can then be used to determine a pK_d, referenced to the pK_a value for [^2,6-ClPhNH₃][OTf] at 298 K in THF, for ease of comparison (eq 7). Hereafter, we use these pK_d values for discussion, but note that use of the pK_a values instead does not substantively alter the conclusions drawn.

$${\text{PhNH}}_2+[{}^{2,6-\text{Cl}}{\text{PhNH}}_3][\text{OTf}]\to [{\text{PhNH}}_3][\text{OTf}]+{}^{2,6-\text{Cl}}{\text{PhNH}}_2$$ (6)

$$\mathrm{p}{K}_{\mathrm{d}}([{\text{PhNH}}_3][\text{OTf}])=-\mathrm{\Delta }{\mathrm{G}}_{\mathrm{d}}/(2.303\times \text{RT})+\mathrm{p}{K}_{\mathrm{a}}({}^{2,6-\text{Cl}}{{\text{PhNH}}_3}^{+})$$ (7)

Calculations of the pK_d of all of the relevant species (Table 1) shows that the pK_d of [Cp*(exo-η⁴-C₅Me₅H)Co][OTf] (pK_d^calc = 11.8; Table 1) falls within the range of the anilinium acids studied (0.4 ≤ pK_d^calc ≤ 15.7), suggesting there should be a significant acid dependence on the kinetics and thermodynamics of Cp*₂Co protonation. To better elucidate the differences in Cp*₂Co protonation between the acids studied, we investigated the kinetics of protonation for three acids, [^2,6-ClPhNH₃][OTf] (high selectivity; pK_d^calc = 3.4), [^2,6-MePhNH₃][OTf] (modest selectivity; pK_d^calc = 13.2), and [^4-OMePhNH₃][OTf] (poor selectivity; pK_d^calc = 15.8).

Transition states for Cp*₂Co protonation were located for all three acids. To confirm that these transition states accurately reflect proton transfer, internal reaction coordinates (IRC) were followed to determine the reactant (IRC-A) and product (IRC-B) minima (Figure 2). These minima represent hydrogen bonded arrangements of the reactants and products. Protonation is found to have only a moderate barrier (ΔG^‡ in kcal mol⁻¹) in all three cases: ([^4-OMePhNH₃][OTf], +4.5; [^2,6-MePhNH₃][OTf], +3.8; [^2,6-ClPhNH₃][OTf], +1.3). This suggests that Cp*₂Co protonation is kinetically accessible in all cases, in agreement with the experimental observation of background HER with each of these acids (see SI).

Figure 2.

The kinetics and thermodynamics of protonation of Cp*₂Co for three acids from different catalytic efficiency regimes ([^4-OMePhNH₃][OTf] = poor selectivity]; [^2,6-MePhNH₃][OTf] = modest selectivity; [^2,6-ClPhNH₃][OTf] = high selectivity).

The small differences in rate, and the large variance in the equilibrium constant K_eq defined in eq 8, points to a significant difference in the population of protonated metallocene, [Cp*(exo-η⁴-C₅Me₅H)Co][OTf], for these anilinium acids during catalysis.

$$K_{\text{eq}}=\frac{[{}^{\mathrm{R}}{\text{PhNH}}_2-{\text{Cp}}^{\ast }(\mathit{\text{exo}}-{\mathrm{\eta }}^4-{\mathrm{C}}_5{\text{Me}}_5\mathrm{H}){\text{Co}}^{+}]}{[{}^{\mathrm{R}}{\text{PhNH}}_3^{+}-{\text{Cp}}_2^{\ast }\text{Co}]}$$ (8)

We reason that the low solubility of the anilinium triflate acids, and the low catalyst concentration, leads to a scenario in which the interaction between the acid and the Cp*₂Co, the latter being present in large excess relative to the iron catalyst, significantly affects the overall kinetics of productive N–H bond formation. As such, the difference in [Cp*(exo-η⁴-C₅Me₅H)Co][OTf] concentration and formation rate should relate to, and likely dominate, the origin of the observed pK_a effect. This explanation, rather than one that involves differences in rates for the direct interaction of a given P₃^BFeN_xH_y species with an anilinium acid, better captures the collected data available.²¹

[Cp*(exo-η⁴-C₅Me₅H)Co][OTf], characterized by a very weak C–H bond, should be a strong PCET donor, and we presume it serves such a role under the catalytic conditions being discussed herein.²² Its reactions with P₃^BFeN_xH_y intermediates may occur in a synchronous fashion, akin to HAT, or in an asynchronous fashion more akin to a PT-ET reaction.²³ While many P₃^BFeN_xH_y intermediates may, at least in part, be generated via PCET with [Cp*(exo-η⁴-C₅Me₅H)Co][OTf],²⁴ available experimental data point to a critical role for such a reaction via trapping of the highly reactive first fixed intermediate, P₃^BFeNNH (Figure 3), before it can bimolecularly release H₂ (eq 3). We hence investigated this reaction in more detail.

Figure 3.

The calculated thermodynamics and kinetics of synchronous PCET and asynchronous PCET (PT-ET), between P₃^BFeNNH and [Cp*(exo-η⁴-C₅Me₅H)Co][OTf] to generate P₃^BFeNNH₂. Note: k_rel for ET is defined as 1 M⁻¹ s⁻¹.

Both a synchronous PCET (ΔG_PCET = −17.3 kcal mol⁻¹; eq 9) and an asynchronous PCET path (ΔG_PT = −5.7 kcal mol⁻¹, ΔG_ET = −11.6 kcal mol⁻¹; eq 10 and 11), are predicted to be thermodynamically favorable.

$${{\mathrm{P}}_3}^{\mathrm{B}}\text{FeNNH}+[\text{Cp}\ast (\mathit{\text{exo}}-{\mathrm{\eta }}^4-{\mathrm{C}}_5{\text{Me}}_5\mathrm{H})\text{Co}][\text{OTf}]\to {{\mathrm{P}}_3}^{\mathrm{B}}{\text{FeNNH}}_2+[{\text{Cp}\ast }_2\text{Co}][\text{OTf}]$$ (9)

$${{\mathrm{P}}_3}^{\mathrm{B}}\text{FeNNH}+[\text{Cp}\ast (\mathit{\text{exo}}-{\mathrm{\eta }}^4-{\mathrm{C}}_5{\text{Me}}_5\mathrm{H})\text{Co}][\text{OTf}]\to [{{\mathrm{P}}_3}^{\mathrm{B}}{\text{FeNNH}}_2][\text{OTf}]+{\text{Cp}\ast }_2\text{Co}$$ (10)

$$[{{\mathrm{P}}_3}^{\mathrm{B}}{\text{FeNNH}}_2][\text{OTf}]+{\text{Cp}\ast }_2\text{Co}\to {{\mathrm{P}}_3}^{\mathrm{B}}{\text{FeNNH}}_2+[{\text{Cp}\ast }_2\text{Co}][\text{OTf}]$$ (11)

To evaluate the kinetics of these reactions the Marcus theory expressions²⁵ and the Hammes-Schiffer method²⁶ were used to approximate relative rates of bimolecular ET and PCET. We find that there is a slight kinetic preference for the fully synchronous PCET reaction (k_rel^PCET ~ 3 × 10³ M⁻¹s⁻¹) compared to the fully asynchronous PT-ET reaction (k_rel^PT-ET ≈ k_rel^ET ≡ 1 M⁻¹ s⁻¹; Figure 3).²⁷

The above discussion leads to the conclusion that the efficiency for NH₃ formation in this system is coupled to the kinetics and/or thermodynamics of the reaction between the anilinium triflate acid and the Cp*₂Co reductant. This conclusion is counterintuitive, as the protonation of Cp*₂Co is also the requisite first step for background HER.²⁸ The fact that a key HER intermediate can be intercepted and used for productive N₂RR steps is an important design principle for such catalysis. Similar design strategies are currently being used to repurpose molecular cobalt HER catalysts for the reduction of unsaturated substrates.²⁹

Efforts are often undertaken to suppress background reactivity between acid and reductant in catalytic N₂RR systems.^1a-b We were hence particularly interested to explore whether the inclusion of a metallocene co-catalyst, in this case Cp*₂Co, might improve the yield, and/or the Faradaic efficiency (FE), for N₂RR versus HER, in controlled potential electrolysis (CPE) experiments with P₃^BFe⁺ under N₂.

Electrolysis studies

To set the context for this section of the present study, we had shown previously that ~ 2.2 equiv NH₃ (per Fe) could be generated via controlled potential electrolysis (CPE; −2.3 V vs Fc^+/0) at a reticulated vitreous carbon working electrode, using P₃^BFe⁺ as the (pre)catalyst in the presence of HBAr^F₄ (50 equiv) at −45 °C under an atmosphere of N₂. This yield of NH₃ corresponded to a ~ 25% FE which, while modest in terms of overall chemoselectivity, compares very favorably to FE’s most typically reported for heterogeneous electrocatalysts for N₂RR that operate below 100 °C (< 2%).^2,30

To further explore the possibility of using P₃^BFe⁺ as an electrocatalyst for N₂RR, various conditions were surveyed to determine whether enhanced yields of NH₃ could be obtained from CPE experiments. For example, various applied potentials were studied (ranging from −2.1 to −3.0 V vs Fc^+/0), the concentrations of P₃^BFe⁺ and HBAr^F₄ were varied, the ratio of acid to catalyst was varied, and the rate at which acid was delivered to the system was varied (e.g., initial full loadings, batch-wise additions, reloadings, or continuous slow additions). None of these studies led to substantial improvement in N₂RR; in all cases, < 2.5 equiv of NH₃ was obtained per P₃^BFe⁺. Attempts to vary the ratio of the electrode surface area to the working compartment solution volume, either by employing smaller cell geometries or by using different morphologies of glassy carbon as the working electrode (e.g., reticulated porous materials of different pore density or plates of different dimensions), also failed to provide substantial improvement in NH₃ yield. The replacement of HBAr^F₄, the original acid used in our electrolysis studies,^6d by 50 equiv of [Ph₂NH₂][OTf] led to similar yields of NH₃ (Table 2, entry 1).

Table 2.

Yields and Faradaic Efficiencies of NH₃ from CPE Experiments with P₃^BFe⁺

Entry	Equiv Cp*2Co+	Equiv NH3 (per Fe)	Equiv NH3 (per Co)	NH3 FE (%)
1	0	2.6 ± 0.3d	―	24 ± 5
2a	0	2.6 ± 0.3	―	18 ± 1
3	1	4.0 ± 0.6	4.0 ± 0.6	28 ± 5
4	5	4.0 ± 0.6d	0.8 ± 0.1	25 ± 3
5a	5	5.5 ± 0.9e	1.1 ± 0.2	19 ± 1
6	10	4 ± 1	0.4 ± 0.1	24 ± 7
7b	5	1.9 ± 0.2	0.4 ± 0.1	10 ± 1
8c	5	0.9 ± 0.4	0.2 ± 0.1	6 ± 3

All CPE experiments conducted at −2.1 V vs Fc^+/0 with 0.1 M NaBAr^F₄ in Et₂O as solvent at −35 °C under an N₂ atmosphere, featuring a glassy carbon plate working electrode, Ag^+/0 reference couple isolated by a CoralPor™ frit referenced externally to Fc^+/0, and a solid sodium auxiliary electrode. Working and auxiliary chambers separated by a sintered glass frit. See SI for further experimental details, controls, and additional data. Averages represent two runs unless noted.

^aAfter initial electrolysis with 50 equiv [Ph₂NH₂][OTf], an additional 50 equiv [Ph₂NH₂][OTf] in 0.1 M NaBAr^F₄ Et₂O solution was added to the working chamber, via syringe through a rubber septum, followed by additional CPE at −2.1 V vs Fc^+/0.

^b[^2,6-ClPhNH₃][OTf] employed as the acid.

^c[PhNH₃][OTf] employed as the acid.

^dAverage of three runs.

^eAverage of five runs.

We next investigated the effect of Cp*₂Co⁺ as an additive on the electrolysis/electrocatalysis. Traces of relevant cyclic voltammograms (Figure 4A and 4B) collected with glassy carbon as the working electrode in Et₂O under glovebox atmosphere N₂ at −35 °C are provided. Background traces including only [Ph₂NH₂][OTf] are present in both panels (gray traces). Cp*₂Co⁺ (panel A, yellow trace), Cp*₂Co⁺ with the addition of ten equiv of [Ph₂NH₂][OTf] (panel A, green trace), P₃^BFe⁺ (panel B, dark blue trace), P₃^BFe⁺ with the addition of ten equiv of [Ph₂NH₂][OTf] (panel B, light blue trace), and P₃^BFe⁺ with the addition of one equiv of Cp*₂Co⁺ and ten equiv of [Ph₂NH₂][OTf] (both panels, red trace).

Figure 4.

A) Cyclic voltammograms of 10 equiv [Ph₂NH₂][OTf] (gray trace), 1 equiv [Cp*₂Co][BAr^F₄] (Cp*₂Co⁺) (yellow trace), 1 equiv Cp*₂Co⁺ with 10 equiv [Ph₂NH₂][OTf] (green trace), and P₃^BFe⁺ with 1 equiv of Cp*₂Co⁺ and 10 equiv [Ph₂NH₂][OTf] (red trace). B) Cyclic voltammograms of 10 equiv [Ph₂NH₂][OTf] (gray trace), P₃^BFe⁺ (dark blue trace), P₃^BFe⁺ with 10 equiv [Ph₂NH₂][OTf] (light blue trace), and P₃^BFe⁺ with 1 equiv of Cp*₂Co⁺ and 10 equiv [Ph₂NH₂][OTf] (red trace). All spectra are collected in 0.1 M NaBAr^F₄ solution in Et₂O at −35 °C using a glassy carbon working electrode, and externally referenced to the Fc^+/0 couple. Scan rate is 100 mV/s. catalyzed by Cp*₂Co⁺ (panel A, green trace)

The cyclic voltammogram of Cp*₂Co⁺ is shown in panel A (yellow trace), displaying the reversible Cp*₂Co^+/0 couple at −2.0 V. The addition of [Ph₂NH₂][OTf] to Cp*₂Co⁺ causes an increase in current at this potential, consistent with HER catalyzed by Cp*₂Co⁺ (panel A, green trace).

Panel B provides the cyclic voltammogram of P₃^BFe⁺ in the absence (dark blue trace, showing previously assigned and (pseudo)reversible P₃^BFeN₂^0/− couple at ~ −2.1 V) and in the presence (light blue trace) of [Ph₂NH₂][OTf].³¹ The latter is indicative of modest HER and N₂RR. Also evident upon the addition of acid is the disappearance of a wave corresponding to the P₃^BFe^+/0 couple at ~ −1.6 V. This wave, in the absence of acid, is broad and shows a large peak-to-peak separation, likely due to the presence of both P₃^BFe⁺ and P₃^BFeN₂⁺ in solution at −35 °C. The addition of a large excess of [Ph₂NH₂][OTf] presumably results in triflate binding (to generate P₃^BFeOTf, thereby attenuating the waves associated with the reduction of P₃^BFe⁺ and P₃^BFeN₂⁺). The red trace of Panel A reproduced in Panel B to illustrate the marked increase incurrent observed when Cp*₂Co is added.

CPE studies were undertaken to characterize the reduction products associated with the red trace at ~ −2.1 V vs Fc^+/0. These studies employed a glassy carbon plate electrode, a Ag^+/0 reference electrode that was isolated by a CoralPor™ frit and referenced externally Fc^+/0 redox couple, and a solid sodium auxiliary electrode. The latter was used to avoid excessive, non-productive redox cycling between the working and auxiliary chambers.³² Unless otherwise noted, CPE experiments were performed at −2.1 V versus Fc^+/0, with 0.1 M NaBAr^F₄ as the ether-soluble electrolyte, under a glovebox N₂ atmosphere at −35 °C. The electrolysis was continued until the current had dropped to 1% of the initial current measured, or until 21.5 hours had passed. The Supporting Information provides additional details.

CPE experiments were conducted with the inclusion of 0, 1, 5, and 10 equiv of Cp*₂Co⁺ with respect to P₃^BFe⁺, using excess [Ph₂NH₂][OTf] as the acid. In the absence of added Cp*₂Co⁺, a significant amount of NH₃ was generated (2.6 ± 0.3 equiv per Fe, entry 1), consistent with the previous finding that, in the presence of a strong acid, P₃^BFe⁺ can electrolytically mediate N₂-to-NH₃ conversion.^6d Notably, when a CPE experiment that did not include Cp*₂Co⁺ was reloaded with additional acid after electrolysis and electrolyzed again, the total yield of NH₃ (2.6 ± 0.6 equiv NH₃ per Fe, entry 2) did not improve.

We found that inclusion of 1.0 equiv of Cp*₂Co⁺ enhanced the NH₃ yield, by a factor of ~ 1.5 (Table 2, entry 3) without decreasing the FE. The data provide a total yield, with respect to both Fe and Co, that confirm modest, but still unequivocal, N₂RR electrocatalysis. In single run experiments, the highest NH₃ yield in the absence of Cp*₂Co⁺ was 2.8 equiv, compared with 4.4 equiv in the presence of 1 equiv of Cp*₂Co⁺. Conversely, the lowest single run NH₃ yield in the absence of Cp*₂Co⁺ was 2.3 equiv, compared with 3.5 equiv in the presence of 1 equiv of Cp*₂Co⁺.

Increasing the amount of added Cp*₂Co⁺ did not affect the NH₃ yield (entry 4). However, the addition of a second loading of [Ph₂NH₂][OTf] following the first electrolysis (entry 5), followed by additional electrolysis, led to an improved yield of NH₃, suggesting that some active catalyst is still present after the first run.^6d,9 Even higher Cp*₂Co⁺ loading did not lead to higher NH₃ yields (entry 6).

CPE of P₃^BFe⁺ in the presence of Cp*₂Co⁺ was also explored with other acids. Replacing [Ph₂NH₂][OTf] in these experiments with [^2,6-ClPhNH₃][OTf] led to lower yields of NH₃, and with [PhNH₃][OTf] even lower yields of NH₃ were observed (entries 7 and 8 respectively). The lower, but nonzero, yield of NH₃ provided by [PhNH₃][OTf] in these CPE experiments is consistent with chemical trials employing various acids (vide supra) and can be rationalized by the relative pK_a of the acids (Table 1). The intermediate yield of NH₃ provided by [^2,6-ClPhNH₃][OTf] in these CPE experiments is less consistent with simple pK_a considerations, suggesting that additional factors are at play, perhaps including the relative stability of the acid or conjugate base to electrolysis.

To probe whether electrode-immobilized iron might contribute to the N₂RR electrocatalysis, X-ray photoelectron spectroscopy (XPS) was used to study the electrode. After a standard CPE experiment with P₃^BFe⁺, 5 equiv of Cp*₂Co⁺, and 50 equiv [Ph₂NH₂][OTf], the electrode was removed, washed with fresh 0.1 M NaBAr^F₄ Et₂O solution, then fresh Et₂O, and probed by XPS. A very low coverage of Fe (< 0.3 atom % Fe) was detected in the post-electrolysis material; no Fe was detected on a segment of the electrode which was not exposed to the electrolytic solution. This observation implies a detectable but likely small degree of degradation of P₃^BFe⁺ over the course of a 15 hour CPE experiment. Worth noting is that no Co was detected on the post-electrolysis electrode.

To test whether the small amount of deposited Fe material might be catalytically active for N₂RR, following a standard CPE experiment the electrode was removed from the cold electrolysis solution, washed with fresh 0.1 M NaBAr^F₄ Et₂O at −35 °C (the electrode itself was maintained at −35 °C at all times), and then used for an additional CPE experiment, under identical conditions except that P₃^BFe⁺ was excluded. This CPE experiment yielded no detectable NH₃. The charge passed, and H₂ yield, were very similar to a “no P₃^BFe⁺” control experiment conducted with a freshly cleaned electrode (See SI for further details). Accordingly, a CPE experiment in the absence of P₃^BFe⁺ demonstrated that Cp*₂Co⁺ serves as an effective electrocatalyst for HER with [Ph₂NH₂][OTf] as the acid source, but does not catalyze the N₂RR reaction (0% FE for NH₃, 75% FE for H₂; see SI). This background HER, and the observed catalytic response to the addition of [Ph₂NH₂][OTf] at the Cp*₂Co^+/0 couple, provides circumstantial evidence for the formation of a protonated decamethylcobaltocene intermediate, Cp*(η⁴-C₅Me₅H)Co⁺, on a timescale similar to that of the N₂RR mediated by P₃^BFe⁺.

To probe whether the sodium auxiliary electrode used in the CPE experiments might play a non-innocent role as a chemical reductant, a standard CPE experiment with P₃^BFe⁺, 5 equiv Cp*₂Co⁺, and 50 equiv [Ph₂NH₂][OTf] was assembled, but was left to stir at −35 °C for 43 hours without an applied potential bias. This experiment yielded 0.3 equiv NH₃ (relative to Fe), suggesting that background N₂RR due to the sodium auxiliary electrode is very minor.

To ensure the NH₃ produced was derived from the N₂ atmosphere during these electrolysis experiments, as opposed to degradation of the anilinium acid used, a standard CPE experiment using P₃^BFe⁺, 5 equiv Cp*₂Co⁺, and 50 equiv of [Ph₂¹⁵NH₂][OTf] was performed. Only ¹⁴NH₃ product was detected.

We also sought to compare the chemical N₂RR catalysis efficiency of the P₃^BFe⁺ catalyst under conditions similar to those used for electrocatalysis. Hence, chemical catalysis with P₃^BFe⁺, employing Cp*₂Co as a reductant and [Ph₂NH₂][OTf] as the acid at −35 °C instead of the more typical temperature of −78 °C, in a 0.1 M NaBAr^F₄ Et₂O solution, afforded lower yields of NH₃ (1.8 ± 0.7 equiv of NH₃ per Fe) than the yields observed via electrolysis with Cp*₂Co⁺ as an additive. The lower yields of NH₃ in these chemical trials, compared with our previously reported conditions (12.8 ± 0.5 equiv of NH₃ per Fe at −78 °C),⁹ may be attributable to increased competitive HER resulting from a more solubilizing medium (0.1 M NaBAr^F₄ Et₂O vs pure Et₂O) and a higher temperature (−35 °C vs −78 °C).⁹ These results suggest that an electrochemical approach to NH₃ formation can improve performance, based on selectivity for N₂RR, of a molecular catalyst under comparable conditions.

CONCLUSION

Herein we described the first systematic pK_a studies on a synthetic nitrogen fixation catalyst and find a strong correlation between pK_a and N₂RR vs HER efficiency. Chemical studies reveal that, on their own, the anilinium triflate acids employed in the catalysis are unable to generate the N–H bonds of early-stage N₂RR intermediates such as P₃^BFeNNH₂⁺. We propose that the insolubility of these triflate acids prevents the sufficiently rapid proton transfer necessary to capture the critical but unstable first fixed intermediate, P₃^BFeNNH. Under catalytic conditions, we believe that the presence of the metallocene reductant (Cp*₂Co) is essential, as this species can be protonated in situ to form Cp*(η⁴-C₅Me₅H)Co⁺, which in turn is effective in N–H bond formation with early intermediates. This leads to the intriguing conclusion that an intermediate of the background HER pathway is redirected for productive N₂RR chemistry during catalysis.

DFT studies illustrate that the pK_a effect on the N₂RR efficiency may be explained by the variation in the kinetics and thermodynamics of Cp*₂Co protonation by the different acids. Investigation of the reactivity of Cp*(exo-η⁴-C₅Me₅H)Co⁺ with the P₃^BFeNNH intermediate revealed that PCET reactivity, either synchronous or asynchronous, is favorable and may proceed with only a small barrier, suggesting that P₃^BFeNNH can be rapidly trapped by Cp*(exo-η⁴-C₅Me₅H)Co⁺. We suspect Cp*(η⁴-C₅Me₅H)Co⁺ may be involved in a variety of N–H bond forming reactions during the overall catalysis.

Despite the fact that Cp*₂Co⁺ itself catalyzes HER under the conditions employed for electrocatalytic N₂RR, we found that its inclusion in CPE experiments containing P₃^BFe⁺ and acid under an N₂ atmosphere led to modest improvements in the overall catalytic yield of NH₃. This system represents what is to our knowledge the first unambiguous example of electrocatalytic N₂RR mediated by a soluble, molecular coordination complex.