NH₃ Binding to Fe^II in MeCN

Catalyst degradation reported in previous studies of ammonia oxidation was proposed to occur from either competitive binding of NH₃ to the metal center^{29, 38} or protonation of an anionic ligand by [NH₄⁺].³⁵ Considering these limitations, we explored the use of an iron complex, trans-[LFe^II(MeCN)₂][PF₆]₂, bearing a macrocyclic ligand (abbreviated hereafter as L) with four N-heterocyclic carbene (NHC) donors. The strong-field tetradentate macrocyclic ligand is expected to confer enhanced stability. In addition, the Fe center in the cyclic tetra-NHC environment is electron-rich because of the strongly σ-donating NHC ligand, facilitating the first oxidation step in catalytic ammonia oxidation. Trans-[LFe^II(MeCN)₂][PF₆]₂ was first reported by Kühn and co-workers,⁵⁶ who showed that the axial MeCN ligands of trans configurations can be replaced by DMSO, NO, CO, ^tBuNC or Ph₃P.^56-58 The reaction of [LFe^II(MeCN)₂]²⁺ with O₂ and catalytic epoxidation of alkenes have also been reported.^{57, 59-61} Notably, NH₃ binding to an acyclic tetra-NHC Fe^II complex with cis-labile binding sites was recently reported by our group.⁶²

Addition of ¹⁵NH₃ gas (1.2 atm) to a light yellow solution of [LFe^II(MeCN)₂]²⁺ in CD₃CN at 22 °C produced a deepened yellow solution that contained approximately 2.8 M of ¹⁵NH₃. The ¹H NMR spectra revealed a mixture of [LFe^II(MeCN)₂]²⁺, [LFe^II(NH₃)(MeCN)]²⁺ and [LFe^II(NH₃)₂]²⁺ (Scheme 1) in a molar ratio of 5 : 63 : 32. All the ¹H NMR spectroscopic signals were observed in the normal range of chemical shifts, indicating diamagnetic low-spin Fe^II species. Free ¹⁵NH₃ was observed at 0.52 ppm in the ¹H NMR spectrum, compared to 0.45 ppm for ¹⁵NH₃ in CD₃CN at 1.0 atm with no Fe^II complex added, suggesting that the bound ¹⁵NH₃ ligands exchange with free ¹⁵NH₃. In addition to the ¹H NMR spectroscopic signals of [LFe^II(MeCN)₂]²⁺, two sets of new resonances assigned to ¹⁵NH₃-bound Fe^II complexes were observed (Figure S11). Doublets for bound ¹⁵NH₃ at −0.93 ppm (¹J_15N−1H=67.5 Hz) and −1.49 ppm (¹J_15N−1H=67.2 Hz) are assigned to [LFe^II(¹⁵NH₃)(MeCN)]²⁺ and [LFe^II(¹⁵NH₃)₂]²⁺, respectively. In the ¹⁵N NMR spectrum, two quartets for bound ¹⁵NH₃ ligands were observed at −442.5 (¹J_15N−1H=67.3 Hz) and −450.5 (¹J_15N−1H=66.8 Hz) ppm, both of which are upfield from free ¹⁵NH₃ (−370.0 ppm; Figure S19). The observed doublets in the ¹H NMR spectrum and quartets in the ¹⁵N NMR spectrum for bound ¹⁵NH₃ ligands all collapsed into singlets in the ¹H{¹⁵N} and ¹⁵N{¹H} NMR spectra (Figures S15 and S20).

The assignment was further supported by ¹H NMR experiments on [LFe^II(MeCN)₂]²⁺ with incrementally increasing concentrations of NH₃, in which the ¹H signals of [LFe^II(NH₃)(MeCN)]²⁺ emerged first, followed by those of [LFe^II(NH₃)₂]²⁺(Figure S25). Upon increasing [NH₃] from 0.012 to 0.27 M, the signals of bound NH₃ for both [LFe^II(NH₃)(MeCN)]²⁺ and [LFe^II(NH₃)₂]²⁺ shifted slightly towards that of free NH₃, moving downfield from −1.13 to −1.11 ppm for the former and from −1.66 to −1.64 ppm for the latter, again indicative of dynamic binding equilibria of NH₃ to the Fe^II complexes.

At [NH₃] <0.051 M, only the equilibrium between [LFe^II(MeCN)₂]²⁺ and [LFe^II(NH₃)(MeCN)]²⁺ was observed, with no [LFe^II(NH₃)₂]²⁺ detected. At higher NH₃ concentrations (0.10–2.8 M), [LFe^II(NH₃)₂]²⁺ was observed, leading to two interlinked equilibria. The first equilibrium constant (K_eq,1 in Scheme 1) was determined by eq. S2 in Scheme S3 using the equilibrium concentrations of NH₃, [LFe^II(MeCN)₂]²⁺, and [LFe^II(NH₃)(MeCN)]²⁺. The second equilibrium constant (K_eq,2) was determined through two distinct methods. The first method paralleled the approach employed for K_eq,1, relying on the equilibrium concentrations of NH₃, [LFe^II(NH₃)(MeCN)]²⁺, and [LFe^II(NH₃)₂]²⁺ (eq. S4 in Scheme S4). The second method was derived based on the concept of a defined fractional saturation, which was originally developed for binding of O₂ to hemoglobin (eq. S5 in Scheme S5, details in the Supporting Information).^{63, 64} The equilibrium constants for binding of NH₃ at 298 K were determined as K_eq,1≈3.6 M⁻¹ and K_eq,2≈0.17 M⁻¹. As estimated using the obtained K_eq,1 and K_eq,2, the population of [LFe^II(NH₃)(MeCN)]²⁺ is approximately 69 %, whereas [LFe^II(MeCN)₂]²⁺and [LFe^II(NH₃)₂]²⁺ constitute about 19 % and 12 %, respectively, at 298 K and 1.0 M NH₃ in MeCN.

A crystal of [LFe^II(NH₃)(MeCN)][PF₆]₂ was grown by vapor diffusion of Et₂O to a MeCN solution in the presence of concentrated NH₃ at −30 °C, and the structure was determined by single crystal X-ray diffraction (Figure 1).⁶⁵ The flexibility of four CH₂ linkers allows the ligand to adopt a distorted saddle conformation, while the Fe^II maintains a slightly distorted octahedral geometry, similar to that found for [LFe^II(MeCN)₂]²⁺.⁵⁶ The Fe^II−NH₃ bond length of 2.0603(16) Å is in the typical range of low-spin NH₃-bound Fe^II complexes (2.03–2.08 Å).^{30, 62, 66-68}

NH₃ binding to Fe^III in MeCN.

The Fe^III complex, [LFe^III(MeCN)₂]³⁺, was prepared by one-electron oxidation of [LFe^II(MeCN)₂]²⁺ (E_1/2=0.15 V in MeCN, all reduction potentials are referenced to [Cp₂Fe]^+/0) by thianthrene radical cation (E_1/2=0.86 V in MeCN).⁵⁹ A deep purple solution of [LFe^III(MeCN)₂]³⁺ was treated with 2.5 equiv. of NH₃ in MeCN at −30 °C, resulting in an immediate color change to yellow (Scheme 2). After recrystallization by layering Et₂O onto the MeCN solution, [LFe^III(NH₃)₂]³⁺ was isolated as a yellow solid in ∼61 % yield. Note that excess NH₃ would reduce Fe^III to Fe^II (see details below).

In the ¹H NMR spectrum of the paramagnetic Fe^III complex in CD₃CN, two distinct singlets with equal integrals (8H) were observed at −16.8 and 39.4 ppm, assigned to CH and CH₂, respectively, along with a broad resonance for bound NH₃ at 257.1 ppm (Figure S27). The assignment of bound NH₃ was confirmed by its almost complete absence in the ¹H NMR spectrum of [LFe^III(ND₃)₂]²⁺ (99 atom % D) (Figure S31), and was further verified by a sharp signal at 256.6 ppm in the ²H NMR spectrum in CD₃CN (Figure S32).

Slow dissociation of the axial NH₃ ligands of [LFe^III(NH₃)₂]³⁺ was observed, as indicated by the conversion of 25 % of [LFe^III(NH₃)₂]³⁺ after 3 days of vacuum drying. This drying process followed by dissolution in CD₃CN generated [LFe^III(NH₃)(CD₃CN)]³⁺, which exhibited ¹H NMR signals at −16.0 ppm for CH, 39.1 and 45.5 ppm for its diastereotopic CH₂, and 252.7 ppm for the bound NH₃ (Figures S33 and S34). Similarly, the replacement of NH₃ by the solvent CD₃CN becomes more pronounced as the concentration of [LFe^III(NH₃)₂]³⁺ decreases, as indicated by a concomitant increase in the amount of [LFe^III(NH₃)(CD₃CN)]³⁺ from 5 % in an 8 mM solution to 21 % in a 0.8 mM solution (Figure S35). These observations imply stronger binding of Fe^III−NH₃ compared to Fe^II−NH₃ (see below), as expected, since Fe^III is more electron deficient than Fe^II and behaves as a harder Lewis acid towards σ-donor ligands like NH₃.

A solution of [LFe^III(NH₃)₂]³⁺ in MeCN (∼55 μM Fe), which includes both [LFe^III(NH₃)₂]³⁺ and [LFe^III(NH₃)(MeCN)]³⁺ upon dissolution, displays a broad adsorption band at 450 nm (ϵ=1944 M⁻¹ cm⁻¹) in the UV/Vis spectrum, which shifts from 500 nm (ϵ=2202 M⁻¹ cm⁻¹) and 570 nm (ϵ=2197 M⁻¹ cm⁻¹) for [LFe^III(MeCN)₂]³⁺, consistent with the striking color change from deep purple to bright yellow (Figure S29). [LFe^III(NH₃)₂]³⁺ is likely a low-spin Fe^III complex (S=$$ ), as determined by the Evans method (see details in the Supporting Information).⁶⁹ [LFe^III(NH₃)₂]³⁺ displays N−H stretching frequencies at 3322, 3297 and 3231 cm⁻¹ in the solid-state infrared spectrum (3337, 3450 cm⁻¹ for gaseous NH₃),⁷⁰ compared to the N−D bands at 2481, 2386 and 2345 cm⁻¹ for [LFe^III(ND₃)₂]³⁺. The ratios of N−D vs N−H stretching frequencies (0.75, 0.72, and 0.73) are consistent with the value of 0.73 calculated from Hooke's law.

The molecular structure of [LFe^III(NH₃)₂]³⁺ was determined by single-crystal X-ray diffraction on a crystal grown by vapor diffusion of Et₂O into a MeCN solution at −30 °C. (Figure 2). The conformation of the supporting ligand and the coordination geometry of Fe in [LFe^III(NH₃)₂]³⁺ are similar to those of [LFe^III(MeCN)₂]^3+[59] and [LFe^II(NH₃)(NCMe)]²⁺; all the three complexes exhibit two axial ligands in a trans configuration. Notably, the Fe^III−NH₃ bond (1.964(8) Å) in [LFe^III(NH₃)₂]³⁺ is shorter than those of previously reported Fe^III complexes with NH₃ ligands (1.994–2.080 Å), which all bear more electron-rich di- or tri-anionic supporting ligands.^{20, 34, 71-73}

Electrochemical Studies

The slow replacement of axial NH₃ by MeCN upon dissolution allows an investigation of the electrochemical properties of [LFe^III(NH₃)₂]³⁺ (bottom left in Scheme 3). Three redox events were observed using differential pulse voltammetry (DPV, also used for differential pulse voltammogram) and cyclic voltammetry (CV, also used for cyclic voltammogram). In Figure 3, E₁, E₂ and E₃ represent the E_1/2 values, based on DPV, while E_z,pc and E_z,pa (z=1, 2, 3) represent the peak potential of cathodic and anodic events in the CV, respectively. In Scheme 3, electrochemical reductions and oxidations are denoted as E_z,red and E_z,ox (z=1, 2, 3). As shown in the DPV, a redox wave at 0.15 V (E_1,red) is assigned to the reduction of [LFe^III(MeCN)₂]³⁺, in agreement with the previously reported redox potential of Fe^III/II at 0.15 V for [LFe^II(MeCN)₂]²⁺.⁵⁶ The reduction wave at −0.15 V (E_2,red) likely results from [LFe^III(NH₃)(MeCN)]³⁺. The observed peak current corresponds to 25 % [LFe^III(NH₃)(MeCN)]³⁺ in a 0.8 mM solution, which agrees with the result obtained from the ¹H NMR experiment above where 21 % [LFe^III(NH₃)(MeCN)]³⁺ was detected in a solution of the same concentration. The major redox event at −0.44 V (E_3,red) is assigned to the reduction of [LFe^III(NH₃)₂]³⁺.

In the CV, the reduction of [LFe^III(NH₃)₂]³⁺ (E_3,pc=−0.50 V) on the cathodic scan leads to replacement of the axial NH₃ ligands by the solvent MeCN at a scan rate of 0.1 V/s, resulting in a very small anodic peak at E_3,pa=−0.42 V for [LFe^II(NH₃)₂]²⁺ along with increased anodic peaks at E_2,pa=−0.12 V for [LFe^II(NH₃)(MeCN)]²⁺ and E_1,pa=0.20 V for [LFe^II(MeCN)₂]²⁺ on the return scan. The observation agrees with the favorable formation of [LFe^II(NH₃)(MeCN)]²⁺ and [LFe^II(MeCN)₂]²⁺ over [LFe^II(NH₃)₂]²⁺ in the Fe(II) oxidation state. When the scan rate was increased to 10 V/s, a quasi-reversible redox event of E₃ with i_pa/i_pc=0.74 and ΔE_p=187 mV (compared to ΔE_p=82 mV for [Cp₂Fe]^+/0) was observed, allowing determination of the Fe^III/II redox potential of [LFe^III(NH₃)₂]³⁺ as approximately −0.47 V (Figure S52).

As described above, upon replacement of each MeCN in [LFe^II(MeCN)₂]²⁺ by NH₃, the redox potential of Fe^III/II shifts cathodically by approximately 300 mV, and substitution of two axial MeCN ligands by NH₃ ligands results in a cumulative cathodic shift of about 600 mV. Similar trends have been previously reported. Upon replacing two MeCN ligands by NH₃, a cathodic shift of ∼300 mV for [(TPA)Fe^II(MeCN)₂]²⁺, 460 mV for [^NMe2(NCCN)Fe(MeCN)₂]²⁺, and 680 mV for [acyclic-(NHC)₄Fe(MeCN)₂]²⁺ was observed, respectively.^{29, 62}

in which K_eq=K_eq,1K_eq,2=0.62 M⁻² and ΔE°=E₃–E₁ (more details are shown in Scheme S13).

Based on this approach, K_eq' was estimated to be 2.4×10¹⁰ M⁻², which is ten orders of magnitude greater than K_eq, providing quantitative evidence on the superior binding affinity of NH₃ to Fe^III compared to Fe^II.

Chemical Catalysis of Ammonia Oxidation

We previously discovered that oxidation of NH₃ to N₂ is catalyzed by Ru^II(TMP)(NH₃)₂ (TMP=tetramesitylporphyrin) when treated with an aryloxyl radical that abstracts hydrogen atoms.⁴⁰ In contrast, no N₂ was detected by ¹⁵N NMR spectroscopy when a solution of [LFe^II(MeCN)₂]²⁺ and ¹⁵NH₃ in CD₃CN was treated with either TEMPO (bond dissociation free energy (BDFE) of TEMPO−H=66 kcal/mol)⁷⁵ or tri-tert-butyl aryloxyl radical (BDFE of ArO−H=74.8 kcal/mol).⁷⁵

We directed our focus to the removal of electrons by a one-electron chemical oxidant and deprotonation by an organic base. This approach inherently presents the challenge arising from the commonly observed poor compatibility between oxidizing agents and electron-rich bases.⁷⁶ Pioneering work by Nishibayashi, Sakata, and co-workers addressed the chemical compatibility issue by conducting the oxidations at low temperatures.^{28, 32} They carried out catalytic oxidations of NH₃ using [(p-BrC₆H₄)₃N]⋅⁺[SbCl₆]⁻ (E_1/2=0.67 V in MeCN)⁷⁷ as the oxidant and 2,4,6-trimethylpyridine (collidine) as the base. Note that the basicity of 2,4,6-trimethylpyridine is weaker than that of NH₃ in MeCN (pK_a=15.0 for 2,4,6-trimethylpyridinium; pK_a=16.46 for [NH₄⁺]).^{78, 79} Consequently, only a small fraction of free NH₃ would be generated from 1 : 1 [NH₄][OTf]/collidine (Figure S42 and Table S11, ∼9 % at room temperature, though it may vary at lower temperatures). While NH₃ is being consumed during the catalysis, additional NH₃ would be continually produced in the presence of [NH₄][OTf] and collidine, thereby possibly maintaining a constant, low concentration of NH₃. Additionally, we envision that using bases stronger than NH₃ in the catalytic reaction can lead to other complications, since strong bases are more easily oxidized. Our experiments utilized free NH₃ in solution for the chemical catalysis, in which NH₃ serves both as the substrate to be oxidized and as the base for deprotonation.^27-37

The speciation of [LFe^II(MeCN)_2-x(NH₃)_x]²⁺ (x=0, 1, 2) displayed redox potentials of Fe^III/II lower than about 0.15 V vs [Cp₂Fe]^+/0 (see details in Figures 3 and 4). The oxidizing agent [(p-MeOC₆H₄)₃N]⋅⁺[PF₆]⁻ (E_1/2=0.16 V in MeCN)⁸⁰ was selected for chemical catalysis. Treatment of [LFe^II(MeCN)₂]²⁺ with 24 equiv. of deep blue [(p-MeOC₆H₄)₃N]⋅⁺ in an NMR tube afforded clean oxidation to [LFe^III(MeCN)₂]³⁺, with no apparent decomposition,⁸¹ after sitting at room temperature overnight, as indicated by ¹H NMR spectroscopy (Figure S48). The deep blue solution was then refilled with ∼60 equiv. ¹⁵NH₃ at −30 °C after being degassed by five cycles of freeze-pump-thaw. The solution was maintained at −30 °C for 2 hours, then warmed to room temperature overnight, followed by continuous mixing at room temperature for 36 hours. The reaction was complete, as indicated by the disappearance of the paramagnetic signals in the ¹H NMR spectrum, along with a color change from intense blue to dark brown. The observation of ¹H NMR spectroscopic signals of [LFe^II(MeCN)₂]²⁺ and [LFe^II(NH₃)(MeCN)]²⁺ indicate that the Fe complex survived under the reaction conditions (Figure S49). ¹⁵N₂ was detected at −71.2 ppm vs Me¹⁵NO₂ by ¹⁵N NMR spectroscopy (Figure S50), confirming NH₃ as the source of the N₂ product (Scheme 4).

After verifying the formation of N₂ by ¹⁵N NMR spectroscopy in NMR tube experiments, we conducted larger-scale catalytic reactions to determine the yield of N₂ by gas chromatography (GC) using a thermal conductivity detector (TCD) and helium as a standard for integration. Using the oxidant [(p-MeOC₆H₄)₃N]⋅⁺ and NH₃ in a ratio of 2.5, we achieved 91 turnovers of N₂ per Fe (Table 1, Entry 1), indicating a 50 % increase in the yield of N₂ compared to the results reported in Entry 5 of Table S5 (see detailed results of initial explorations with use of [(p-tolyl)₃N]⋅⁺ (E_1/2=0.38 V in MeCN)⁸⁰ as the oxidant in the Supporting Information). The weak oxidant [(p-MeOC₆H₄)₃N]⋅⁺ provides more flexibility and significant advantages in determining the optimal reaction temperature. Catalysis performed at −30 °C and −20 °C produced improved yields of N₂, generating 104 and 98 turnovers per Fe, respectively (Table 1, Entries 2 and 3). However, further warming the reaction temperature afforded lower yields of N₂, with 77 turnovers at −10 °C (Table 1, Entry 4), 59 turnovers at 0 °C (Table 1, Entry 5), and 25 turnovers at 22 °C (Table 1, Entry 6). These results underscore the value of low temperatures in this catalytic reaction, which may stabilize intermediates and thereby enhance the requisite N−N bond formation, as first reported in the Ru-catalyzed ammonia oxidation by Nishibayashi and co-workers,²⁸ or improve the compatibility between oxidizing agents and free NH₃. Moreover, the lower redox potential of [(p-MeOC₆H₄)₃N]⋅⁺ compared to [(p-tolyl)₃N]⋅⁺ reduces the driving force required for the catalysis, implying a lower “chemical overpotential”.

Increasing the concentration of the NH₃/oxidant pair to a ratio of 5.0 further improved the turnovers of N₂ per Fe to 148 (Table 1, Entry 7). By reducing the catalyst loading, a substantial increase in the yield of N₂ was achieved, affording 182 turnovers of N₂ per Fe (Table 1, Entry 8), which represents the highest reported thus far for the chemical catalysis of ammonia oxidation by a soluble molecular complex. Recharging the catalytic system with fresh (p-MeOC₆H₄)₃N⋅⁺ and NH₃ produced additional 2.4 turnovers of N₂ per Fe (see details in the Supporting Information), implying nearly complete deactivation of the catalysts in the last round of catalysis. Note that in Table 1, Entry 8, an extremely low loading of catalyst (0.043 mM, 0.004 mol %) worked well, indicating that the Fe catalyst is highly efficient in chemical catalysis. Low loading of a Ru catalyst (0.5 mM, 0.0035 mol %) for electrocatalytic ammonia oxidation in aqueous solutions was recently reported by Ménard and co-workers.³⁷ We emphasize that catalysis with low loadings of first-row transition metal catalysts⁴⁴ is typically more challenging, as the d orbital energy splitting in first-row transition metal complexes is generally smaller than that in second- and third-row transition metals, and the difference is expected to contribute to the potentially greater difficulty in achieving high stability of first-row transition metal catalysts in the presence of ammonia. Control experiments in the absence of Fe catalyst show no N₂ formation (Table 1, Entry 9).

Electrocatalytic Ammonia Oxidation

We examined [LFe^II(MeCN)₂]²⁺ as a potential electrocatalyst to compare to the chemical catalysis of ammonia oxidation described above. Electrochemical properties of [LFe^II(MeCN)₂]²⁺ in the presence of NH₃ were evaluated using a solution of 1.0 mM Fe^II and 0.8 M NH₃ in MeCN. In the presence of NH₃, three redox events were observed on the anodic scans of CV and DPV (Figure 4), corresponding to redox potentials of [LFe^II(NH₃)₂]²⁺, [LFe^II(NH₃)(MeCN)]²⁺, and [LFe^II(MeCN)₂]²⁺, respectively, in agreement with the electrochemical results obtained with the isolated [LFe^III(NH₃)₂]³⁺ above. On the return cathodic scan, the predominant reduction wave corresponds to [LFe^III(NH₃)₂]³⁺. The irreversibility observed in all three redox waves implies an electron transfer-chemical (EC) process involving the oxidation-induced replacement of axial MeCN ligands by NH₃, driven by the more favorable binding of NH₃ vs. MeCN to the Fe^III relative to the Fe^II center, as demonstrated above. No clear evidence was obtained for PCET events during the anodic scan (see details in the Supporting Information).

Based on the apparent lack of electrocatalysis when replacing the chemical oxidant with an electrode, we next explored the feasibility of mediated electrocatalytic ammonia oxidation using [(p-MeOC₆H₄)₃N], which can be electrochemically oxidized to generate its radical cation. CVs in MeCN at 295 K and 1 atm N₂ in the absence and presence of NH₃ are shown in Figure 5. In the absence of NH₃, Fe^II complex (light orange) and (p-MeOC₆H₄)₃N (light blue) showed quasi-reversible redox events at nearly identical potentials (0.15 V). A solution of 1.0 mM Fe^II complex and 2.0 equiv. (p-MeOC₆H₄)₃N reveals an increase in peak current (light red), which was merely the sum of the peak current observed independently from 1.0 mM Fe^II complex and from 2.0 mM (p-MeOC₆H₄)₃N (Table S8).

In the presence of NH₃, Fe^II complex displayed three redox events (orange, see above for details), while (p-MeOC₆H₄)₃N still exhibited a quasi-reversible redox event at 0.15 V (blue). A solution containing 1.0 mM Fe^II complex, 2.0 equiv. (p-MeOC₆H₄)₃N and 0.8 M NH₃ in MeCN revealed a significant current enhancement, with a ratio of 3.7 for i_cat/i$$ at room temperature (red trace, i_cat represents the maximum current of catalytic waves), implying electrocatalytic oxidation of NH₃ synergistically enhanced by both the Fe catalyst and (p-MeOC₆H₄)₃N (see CVs in Figure S64 for electrocatalysis at −20 °C). E_cat/2, defined as the potential at which the catalytic wave reaches half of its maximum current (i_cat/2),⁸² was observed at 0.18 V in MeCN (red trace in Figure 5). In this CV experiment, [LFe^II(MeCN)₂]²⁺, [LFe^II(NH₃)(MeCN)]²⁺ and [LFe^II(NH₃)₂]²⁺ were present in the bulk solution. While E_1/2 of [LFe^II(MeCN)₂]²⁺ (0.15 V) is in close proximity to E_cat/2, it is unlikely that the Fe^III/II redox potential directly accounts for the catalytic current response, as neither [LFe^II(MeCN)₂]²⁺ nor [LFe^III(MeCN)₂]³⁺ have coordinating NH₃ ligands. Instead, the E_1/2 values of [LFe^II(MeCN)₂]²⁺ (0.15 V), [LFe^II(NH₃)(MeCN)]²⁺ (−0.15 V), and [LFe^II(NH₃)₂]²⁺ (−0.47 V) collectively govern the formation of their respective Fe^III species, all of which are likely rapidly transformed to [LFe^III(NH₃)₂]³⁺ in the presence of NH₃. We speculate that the NH₃-bound Fe complexes cooperate with (p-MeOC₆H₄)₃N⋅⁺, which has an E_1/2 of 0.15 V that also closely aligns with E_cat/2, thereby electrochemically catalyzing the oxidation of NH₃ (see details in the proposed mechanism below).

The overpotential was estimated using the equation η=E_cat/2–E°_N2/NH3.^{82, 83} Our observed E_cat/2 value was +0.18 V. E°_N2/NH3 is the thermodynamic equilibrium potential of the half reaction shown in eq. 6 (Scheme 4) under standard-state conditions at 298 K in MeCN, with a value of −0.94 V.⁸⁴ The resulting value of the overpotential is approximately 1.1 V. It should be noted that the conditions are not identical for E_cat/2 (0.8 M NH₃) compared to E°_N2/NH3 (1 : 1 buffered condition of 1 M NH₃ and 1 M [NH₄⁺]), so the overpotential should be viewed as an estimate, not an absolutely accurate value. However, the impact of this discrepancy is expected to be small, relative to the magnitude of the overpotential.⁸²

The onset potential (E_onset),⁸⁵ is defined as the point at which the baseline current intersects with the linear extrapolation of the steepest portion of the rising current of the catalytic wave. E_onset of the catalytic wave was observed at 0.09 V in MeCN (Figure S59), which is ∼250 mV (estimated error: ±30 mV) more cathodic than the onset potential of uncatalyzed background ammonia oxidation on a glassy carbon electrode (Figure S61, E_onset=0.34 V). The catalytic E_onset occurs at a milder potential than that observed for [(TPA)Fe^II(MeCN)₂]²⁺ (0.70 V in MeCN),²⁹ [(bpyPy₂Me)Fe^II(MeCN)₂]²⁺ (0.45 V in MeCN),³⁰ and [(BPM^MeO)Fe(MeCN)₂]²⁺ (0.29 V in MeCN),³⁸ yet it occurs at a higher potential than that reported for [Cp*Fe^III(1,2-Ph₂PC₆H₄NH)(NH₃)]⁺ (−0.04 V in THF)³⁴ or [ⁱPr₂NN_F6]Cu^I(NH₃) (−0.24 V in MeCN).³⁵

Controlled-potential electrolysis (CPE) was performed to quantify the product of the mediated electrocatalysis. Initially, a CPE experiment was carried out at room temperature with a constant potential of 0.25 V vs [Cp₂Fe]^+/0, using a solution containing 0.12 mM Fe^II catalyst, 0.24 mM (p-MeOC₆H₄)₃N and 0.8 M NH₃. After 3 hours, 5.7 Coulombs of charge had passed (Figure S66). GC analysis revealed 1.8 turnovers of N₂ per Fe were produced in the working electrode compartment, with a Faradaic efficiency of 16 % for N₂. A CV recorded after the electrolysis showed that the redox wave of (p-MeOC₆H₄)₃N disappeared (Figure S65), indicating its decomposition during CPE at room temperature, suggesting an explanation for the low Faradaic efficiency.

Subsequently, the CPE was performed at low temperatures, as reported above for the chemical catalysis. Using the same reaction solution as for the CPE at room temperature, a potential of 0.20 V vs [Cp₂Fe]^+/0 was applied at −20±3 °C for 4 hours, at which point 6.4 Coulombs of charge had passed (Figure S68). A CV recorded after electrolysis demonstrated that the redox potential of the internal standard, [Cp₂Co][PF₆], drifted catholically by about 120 mV, indicating that the applied potential shifted from 0.20 V initially to a final potential of 0.32 V at the end of the CPE. Additionally, the previously observed current enhancement on the cyclic voltammogram prior to CPE was no longer present (Figure S67). However, a quasi-reversible oxidation peak of (p-MeOC₆H₄)₃N remained, indicating that (p-MeOC₆H₄)₃N and its radical cation at least partially survived at −20 °C during catalysis, consistent with the observation of characteristic bright blue color of the radical cation during bulk electrolysis.

The headspace of the working electrode compartment was sampled while the solution was maintained at −20 °C. GC analysis revealed that 7.2 turnovers of N₂ per Fe were produced, confirming that the catalytic oxidation of NH₃ to form N₂ occurred at low temperatures. The Faradaic efficiency for N₂ generation was determined to be 61 %. The reaction mixtures were then warmed to room temperature; headspace analysis of the working electrode compartment showed that 9.3 turnovers of N₂ per Fe were produced, giving a Faradaic efficiency of 75 % for N₂. The increased N₂ was released from the reaction mixture after warming from −20 °C to room temperature. Additionally, the headspace of the counter electrode compartment was also sampled when the reaction mixtures were left at room temperature. GC analysis revealed 40.5 turnovers of H₂ per Fe were generated, corresponding to a Faradaic efficiency of 100 % for H₂. The ratio of H₂/N₂ is approximately 4.4. Control experiments conducted in the absence of either (p-MeOC₆H₄)₃N or Fe^II catalyst showed no evidence for N₂ formation (Table S10).

In contrast to the chemical catalysis of ammonia oxidation in the reactions of [LFe^II(NH₃)₂]²⁺, (p-MeOC₆H₄)₃N⋅⁺ and NH₃ solution in MeCN at low temperatures, e.g., Entry 3 in Table 1, the electrocatalysis synergistically promoted by [LFe^II(NH₃)₂]²⁺ and triarylamines produced substantially reduced turnovers of N₂ per Fe. We attribute this observation to two factors. First, the chemical catalytic reactions occurred over 12–16 hours, while the bulk electrolysis experiments were carried out for 4 hours, as electrocatalysis seems more sensitive to temperature perturbations compared to chemical catalysis. Since the catalysis of ammonia oxidation is slow, an extended reaction time could produce an increased yield of N₂. Second, it is plausible that electrode passivation occurred as the Fe catalyst and/or triarylamine degraded during bulk electrolysis, consequently leading to a lower turnover of N₂ in electrocatalysis. This possible passivation is also consistent with difficulty we encountered in obtaining CVs after the bulk electrolysis when the voltammetry electrode remained in the solution during electrolysis.

Mechanistic Studies

We investigated the deprotonation of [LFe^III(NH₃)₂]³⁺ using ¹H NMR spectroscopy. When 10 equiv. of NH₃ were added to [LFe^III(NH₃)₂]³⁺ at room temperature, some of the paramagnetic [LFe^III(NH₃)₂]³⁺ was transformed overnight into diamagnetic [LFe^II(MeCN)₂]²⁺ (Figure S38, see experiments with sterically hindered stronger bases in Scheme S9 in the Supporting Information). The deprotonated species [LFe^III(NH₃)(NH₂)]²⁺ was not observed. This observation suggests that the first deprotonation step is either slow or thermodynamically unfavored, and is followed by a thermodynamically favored, fast conversion of the transient [LFe^III(NH₃)(NH₂)]²⁺ to [LFe^II(MeCN)₂]²⁺ (Scheme S9). Thermal instability of an Fe^III−NH₂ species was observed previously.²⁰ The slow deprotonation of [LFe^III(NH₃)₂]³⁺ may account for the absence of current enhancement on the CV timescale for a solution of [LFe^II(MeCN)₂]²⁺ and NH₃ in MeCN.

In the triarylamine-mediated electrocatalysis using [LFe^III(NH₃)₂]³⁺, 2.0 equiv. of triarylamine, and excess NH₃ in MeCN, triarylamine is not expected to be capable of accelerating the direct deprotonation, as it is significantly less basic than NH₃ (for reference, pK_a=1.28 for [Ph₃NH⁺]⁷⁸ vs pK_a=16.46 for [NH₄⁺]⁷⁹ in MeCN). The oxidized triarylamine can function as an electron shuttle (Scheme 5), accelerating heterogeneous electron transfers between the electrode and the soluble molecular Fe catalyst (ΔE_p=81 mV at 0.1 V/s for (p-MeOC₆H₄)₃N in Figure S56). In addition, it is possible that the electrochemically generated (p-MeOC₆H₄)₃N⋅⁺ could facilitate a homogeneous PCET process in conjunction with NH₃.

DFT computations were used to complement experimental mechanistic studies and to assess potential pathways. Calculated reduction potentials of [(L')Fe^II(NH₃)]²⁺ (L'=MeCN or NH₃) are in relatively good agreement with experimental values at the level of theory used (see Methods in the Supporting Information).

For [(L')Fe^II(NH₃)]²⁺, both the solvent-bound species and the bis-ammine may be present, as they are close in energy (ΔG=−1.4 kcal/mol, Scheme S14). When L'_trans=MeCN, oxidation of [Fe^II(MeCN)(NH₃)]²⁺ followed by deprotonation from the oxidized species [Fe^III(MeCN)(NH₃)]³⁺ is most favorable, as it has a calculated pK_a of 25.1 (Scheme 6a, orange values). Deprotonation followed by oxidation is unfavored, as this pathway has a much higher pK_a than that of NH₃. The N−H BDFE of the Fe^II−NH₃ complex is calculated to be 82.1 kcal/mol. It is important to note that while [Ar₃NH]⁺ alone has a calculated BDFE of 57.0 kcal/mol, the Ar₃N/[NH₄]⁺ couple has an effective formal BDFE of 78.9 kcal/mol for a net hydrogen atom transfer (HAT) in MeCN.⁸⁶ The HAT from [Fe^II(MeCN)(NH₃)]²⁺ to the oxidant-base pair of Ar₃N⋅⁺ and NH₃ is uphill by only 3.2 kcal/mol. Similar trends are seen when L'_trans=NH₃ (Scheme 6a, blue values). The oxidation to form [Fe^III(NH₃)₂]³⁺ is very favorable at −0.41 V vs [Cp₂Fe]^+/0. Subsequent deprotonation of the oxidized species [Fe^III(NH₃)₂]³⁺ is feasible, as the calculated pK_a (27.4) is very close to the range identified experimentally (Scheme S9) and agrees with the experimental observation of initial deprotonation as the slowest step. Deprotonation of [Fe^II(NH₃)₂]²⁺ is not feasible, as demonstrated by its high pK_a of 45.7. The calculated BDFE of 76.5 kcal/mol is lower in energy when L'_trans=NH₃ and is less than the BDFE of the Ar₃N/[NH₄]⁺ couple, allowing for HAT from [Fe^II(NH₃)₂]²⁺ to afford [Fe^III(NH₂)(NH₃)]²⁺.

It is likely that most of the catalysis operates from [Fe^III(NH₂)(NH₃)]²⁺ where L'_trans=NH₃, since NH₃ exchange with MeCN is facile, especially at higher oxidation states. The potential for oxidation is calculated to be 0.40 V vs [Cp₂Fe]^+/0 for the formation of [Fe^IV(NH₂)(NH₃)]³⁺, which is otherwise energetically unfavorable through HAT from [Fe^III(NH₃)₂]³⁺ (BDFE of N−H bond=95.1 kcal/mol).

Further PCET events from [Fe^III(NH₂)(NH₃)]³⁺ can occur from the amidyl moiety or from the trans NH₃ ligand (Scheme 6b). A pathway through an amide species has been previously invoked,^{13, 23, 87, 88} and the formation of a bis-amide has been observed in second and third row transition metals.²⁴ For the formation of the Fe^IV imide complex in Scheme 6b-i, deprotonation of [Fe^IV(NH₂)(NH₃)]³⁺ to form [Fe^IV(NH)(NH₃)]²⁺ has a calculated pK_a of 25.0. The N−H BDFE for Fe^III−NH₂ is calculated to be 91.9 kcal/mol, which is higher than previously calculated steps. The BDFE to generate [Fe^V(NH)(NH₃)]³⁺ directly from [Fe^IV(NH₂)(NH₃)]³⁺ was calculated to be 94.6 kcal/mol, which is even more unfavorable. Alternatively, deprotonation could occur from the trans NH₃ ligand to form an Fe^IV bis-amide species, [Fe^IV(NH₂)₂]²⁺ (Scheme 6b-ii). This has a lower calculated pK_a and is lower in energy by 3.0 kcal/mol (1.37ΔpK_a), compared to deprotonation of the amidyl group. The overall N−H BDFE for trans NH₃ is still somewhat high at 88.9 kcal/mol, though it may be lowered by favorable ion pairing. Detailed evaluation of explicit solvent and counterions will be examined in future studies.

Despite the high oxidation state, Fe^IV complexes have been recently invoked in catalytic ammonia oxidation. In calculations by Peters and co-workers for electrocatalytic ammonia oxidation using a dicationic catalyst [(bpyPy₂Me)Fe^II(MeCN)₂]²⁺, high-valent Fe^IV and Fe^V intermediates were proposed to account for N−N bond formation.³⁰ In recent work by Wang, Liao, Ye, and co-workers, evidence was presented suggesting that their electrocatalytic ammonia oxidations involved intermediates featuring Fe^IV and Fe^V species.³⁴

For N−N bond formation, it is possible that [Fe^III(NH₂)(NH₃)]²⁺ could couple to form a bridged bimetallic hydrazido complex; a similar mechanism was invoked in catalysis by (TMP)Ru catalysts (TMP=tetramesitylporphyrin).⁴⁰ This N−N bond formation is found to be unfavorable by 18.9 kcal/mol (Scheme 6c, upper). With each species possessing 2+ charge, however, it may be electrostatically unfavorable, even in a polar solvent like MeCN. Additionally, the formation of the partial π-bond might overstabilize the amide and render it unreactive (see below).⁸⁹ Alternatively, bimetallic coupling could occur from [Fe^IV(NH₂)₂]²⁺ (Scheme 6c, lower). The formation of the bridged complex has a much lower energy (4.1 kcal/mol). From the bridged states, amido transfer can occur by either breaking the Fe−NH₂ bond homolytically or heterolytically. In the bridged Fe^IIINH₂NH₂Fe^III complex, the homolytic route is favorable by 10.5 kcal/mol, whereas in the Fe^IVNH₂NH₂Fe^IV complex, it is favorable by 17.5 kcal/mol. The heterolytic reaction from the bridged Fe^IVNH₂NH₂Fe^IV species is unfavorable by almost 26 kcal/mol, rendering that pathway prohibitively high in energy. Following N−N formation, “NH₂−NH₂” could be easily oxidized to N₂ under the catalytic conditions, since the N−H bond of free hydrazine has a much lower first BDFE (72.6 kcal/mol in the gas phase vs 99.4 kcal/mol for NH₃)⁸⁶ and the electrochemical oxidation of hydrazine is 433 mV more favorable than oxidation of ammonia, based on E° in MeCN with NH₃ as the base.⁸⁴ Warren and co-workers observed facile oxidation of free NH₂−NH₂ to N₂ in MeCN solution with an onset potential of −0.40 V on a glassy carbon electrode.³⁵ In the case of [Fe^III(NH₂NH₂)(NH₂)]²⁺, HAT may even be able to proceed via direct Ar₃N⁺⋅ mediation, considering its weak hydrazido N−H bonds.^{86, 90} The [Fe^III(NH₂)(NH₃)]²⁺ would then enter back into the catalytic cycle.

The macrocyclic tetra-NHC ligand plays a key role as a strong donor, and may also help stabilize the [LFe^III(NH₂NH₂)(NH₂)]²⁺ complex. In previous computations with (TPP)Fe(NH₃)(NH₂) (TPP=tetraphenylporphyrin) complexes of the same oxidation state and multiplicity, the amidyl moiety is nonplanar, with considerable electron density causing the hydrogen atoms to move out of the plane.^{40, 87} As shown in Figure 6, the amidyl moiety is planar; similar geometries were also seen for computed structures of [LFe^III(NH₃)(NH₂)]²⁺ and [LFe^IV(NH₂)₂]²⁺ (Figure S77 and S78). By natural bond orbital (NBO)⁹¹ analysis, the Fe−NH₂ bond in both molecules consists of one fully occupied σ orbital and one partially occupied π orbital, suggesting the initial formation of the metal-ligand π-system. This bonding is thought to stabilize the metal-amidyl complexes⁸⁹ and is likely the source of the planar geometry. These results also illustrate the impact of macrocyclic ligand choice compared to previously studied Fe complexes.^{87, 88}

The terminal nitrogen atom of the hydrazido ligand is oriented towards the C−H bond of the macrocyclic ligand. In the NBO framework, hydrogen bonding can be treated as a charge transfer between the lone pair of the nitrogen (hydrogen bond acceptor) and the σ* orbital of the C−H bond (hydrogen bond donor).⁹² Our computations reveal a weak hydrogen bonding interaction that stabilizes the complex by 1 kcal/mol. There is a small amount of charge transfer present, though about an order of magnitude weaker than previously been observed between anions and O−H bond of water.⁹³ The lowest energy electronic state of [LFe^III(NH₂NH₂)(NH₂)]²⁺ is a doublet wherein most spin density is located on the Fe center, while some is located on the nitrogen of the trans-NH₂ ligand.

Finally, the possibility of NH₃ attack on [LFe^III(NH₂)(NH₃)]²⁺ was explored as an alternative pathway. In this reaction, a proton must be lost from the attacking NH₃. This deprotonation can be done by the tetra-NHC ligand or an additional ammonia, with energetic costs of 52.2 and 65.7 kcal/mol, respectively. Neither of these paths are feasible. It should be noted that there is an energetic cost of 10.7 kcal/mol for the approach of the additional ammonia to the [LFe^III(NH₂)(NH₃)]²⁺ species, which contributes to this large energy. NH₃ attack at later states was not considered in this work, though it will be the subject of future studies.