Investigating N₂ Fixation Using a Bulky Fe(bisphosphine)₂ Framework

ABSTRACT

An iron(0) dinitrogen complex incorporating highly sterically encumbered bisphosphine ligands, Fe(N₂)(dibpe)₂ (dibpe = iBu₂PCH₂CH₂PiBu₂), ^{i Bu}1·N₂ , has been prepared and thoroughly characterized, and its N₂ fixation reactivity assessed. ^{i Bu}1·N₂ is a more hindered analogue of Fe(N₂)(depe)₂ (depe = Et₂PCH₂CH₂PEt₂), ^Et1·N₂ , which has previously been shown to be an efficient N₂ reduction catalyst with unusual selectivity for N₂H₄, and it was anticipated that greater bulk might make ^{i Bu}1·N₂ less prone to deleterious side reactivity, improving performance. The N₂ ligand in ^{i Bu}1·N₂ displays a similar degree of activation to ^Et1·N₂ , and the two complexes can stoichiometrically fix N₂ with similarly high efficiency upon treatment with suitable acids, giving mixtures of NH₃ and N₂H₄. However, attempts to catalytically fix N₂ via treatment of ^{i Bu}1·N₂ with mixtures of excess acids and reductants led to poor results. Mechanistic investigations implicate a combination of more sluggish reaction kinetics and weaker binding of N₂ to the intermediate Fe(I) cation [Fe(dibpe)₂]⁺, [ ^{i Bu}1]⁺, whose study was aided by isolation as its BAr^F ₄ ⁻ salt (Ar^F = 3,5‐bis(trifluoromethyl)phenyl), and which has also been fully characterized. This weak interaction hinders regeneration of the initial ^{i Bu}1·N₂ via reduction of [ ^{i Bu}1·N₂ ]⁺, which appears to be necessary to close the catalytic cycle.

The bulky Fe(0) N₂ complex Fe(N₂)(dibpe)₂ has been prepared and efficiently produces NH₃/N₂H₄ upon treatment with acid, similarly to previously investigated Fe(N₂)(depe)₂. However, it performs much less effectively during attempted catalytic N₂ fixation, allowing investigation of the crucial role of steric hindrance in steps such as N₂ binding and protonation.

1. Introduction

In recent years there has been great interest in the development of transition metal (TM) complexes capable of mediating the fixation of inert molecular N₂ into chemically useful N‐containing compounds [1, 2, 3]. Stoichiometrically, this is typically achieved through protonation of low‐valent TM·N₂ complexes using suitable Brønsted acids to yield the azanes NH₃ and N₂H₄ (Scheme 1a, i) [4]. These reactions may sometimes be rendered catalytic by the addition of excess H⁺ and e⁻ sources under an atmosphere of N₂ (Scheme 1a, ii) [5, 6, 7]. Analogous use of the Lewis acid Me₃SiCl in place of Brønsted acids had also led to numerous protocols for the catalytic transformation of N₂ into N(SiMe₃)₃ (Scheme 1a, iii), again in the presence of additional electron sources [8, 9, 10, 11, 12].

SCHEME 1.

(a) General stoichiometric and catalytic fixation of N₂ mediated by molecular TM complexes, including examples of Fe‐based catalysis, and (b) reduction of N₂ to N₂H₄ catalyzed by Fe(N₂)(PP)₂ (PP = R₂PCH₂CH₂PR₂; [Me₃Si⁺]. = Me₃SiX, X = Cl/Br/I/OTf).

Much of the motivation for the study of these reactions lies in the related industrial and biological processes for N₂ fixation, which are among the most important of all global chemical reactions. In both cases, study of the active Haber‐Bosch (industrial) [13]. or nitrogenase (biological) [14, 15] catalysts is highly challenging, and synthetic molecular catalysts can provide useful analogues that are much more amenable to mechanistic investigation. This is amplified by the ease with which molecular complexes can be modified, allowing the importance of specific structural features (e.g., steric bulk, coordination geometry) to be assessed in a controlled manner via direct comparisons between different, rationally tuned analogues.

Given the presence of Fe in both the Mittasch catalyst (used by the Haber‐Bosch process) and all known nitrogenases (FeMo/FeFe/FeV) [13, 14, 15], it is unsurprising that the use of molecular Fe complexes has attracted particular interest for the fixation of N₂ [16]. Notable work by the groups of Peters [17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28], Nishibayashi [29, 30, 31, 32, 33], and Mezailles [34], among others [35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49] has exposed a variety of Fe‐based homogeneous molecular complexes capable of catalytically transforming N₂ to NH₃. Among our own contributions to this field, [50, 51, 52, 53, 54] we have shown that a surprisingly simple bisphosphine complex, Fe(N₂)(depe)₂ (depe = Et₂PCH₂CH₂PEt₂; ^Et1·N₂ ), is capable of efficient stoichiometric N₂ fixation to give NH₃ and N₂H₄ (up to 55% electron yield) [54]. ^Et1·N₂ was subsequently shown to be effective in the catalytic fixation of N₂ (up to 25 N₂ reduced per Fe) [53]. Notably, this was found to proceed with unprecedented selectivity for N₂H₄, which was formed almost exclusively, instead of the more commonly observed NH₃ (Scheme 1b) [18, 19].

Interestingly, replacing ^Et1·N₂ with the slightly smaller but otherwise very similar complex Fe(N₂)(dmpe)₂ (dmpe = Me₂PCH₂CH₂PMe₂; ^Me1·N₂ ) led to inferior results in analogous reactions (see refs. 53 and 54 and Table S2, entries 16 and 17). This was proposed to be due to rapid, thermodynamically favored protonation of ^Me1·N₂ at the metal, leading to oxidized Fe(II) hydrides that are inactive toward N₂ fixation [54]. In comparison, the bulkier depe ligands in ^Et1·N₂ should confer greater steric protection, thereby channeling reactivity away from this pathway and toward productive protonation at the more accessible terminal N_β, which is kinetically favored (Scheme 2).

SCHEME 2.

Competing protonation of complexes 1·N₂ at Fe (left) or N_β (right), leading to complex deactivation or N₂ functionalization, respectively.

Extrapolation from this conclusion suggests that further increases in the steric bulk of the bisphosphine ligand might continue to lead to improved catalytic performance. To test this hypothesis and to provide further insight into the mechanistic factors that underpin the catalytic N₂ reduction activity of ^Et1·N₂ , we report herein the preparation of the analogous complex Fe(N₂)(dibpe)₂ (dibpe = iBu₂PCH₂CH₂PiBu₂; ^{i Bu}1·N₂ ) and assess its performance in stoichiometric and catalytic N₂ fixation. Consistent with expectations, ^{i Bu}1·N₂ is an effective mediator of the former, but in contrast with expectations, it fails to achieve the latter. Accompanying mechanistic studies help to account for this surprising apparent contradiction.

2. Results and Discussion

2.1. Synthesis and Characterization of ^{i Bu}1·N₂

The previously studied complexes ^Me1·N₂ and ^Et1·N₂ could both be prepared through preparation of the corresponding Fe(II) chloride precursor, FeCl₂(PP)₂ (PP = dmpe, ^Me1·Cl₂ ; depe, ^Et1·Cl₂ ), and subsequent reduction (Scheme 3a) [4, 54, 55, 56, 57]. However, attempts to isolate the dibpe‐substituted analogue ^{i Bu}1·Cl₂ through analogous reactions were unsuccessful. Instead, the combination of FeCl₂ and dibpe led to broad resonances at chemical shifts corresponding to the free ligand in the ³¹P{¹H} and ¹H NMR spectra, and solvent peaks were shifted relative to an internal standard, indicating the formation of paramagnetic species [58]. Nor was any color change observed (cf. vivid green colors for ^Me1·Cl₂ and ^Et1·Cl₂ ) [56]. These observations are tentatively attributed to reversible formation of the high‐spin, four‐coordinate complex FeCl₂(dibpe), ^{i Bu}2, with coordination of a second dibpe ligand being prevented on steric grounds (Scheme 3b). Nevertheless, it was anticipated that, while the intended Fe(dibpe)₂ core may be too bulky to accommodate the two chloride ligands in ^{i Bu}1·Cl₂ , the formation of ^{i Bu}1·N₂ should still be possible due to both the reduced steric profile of the N₂ ligand and the greater size of Fe(0) relative to Fe(II). Thus, a ‘one‐pot’ synthesis of ^{i Bu}1·N₂ from FeCl₂, dibpe, and reductant under an N₂ atmosphere was pursued, in the hope that in situ reduction of the postulated ^{i Bu}2 would allow a second dibpe ligand to coordinate after loss of chloride. Gratifyingly, the reaction of FeCl₂ and dibpe with a combination of Mg and MgBr₂·(OEt₂)₂ in THF under N₂ led to the formation of an orange/red solution similar in color to ^Me/Et1·N₂ (and clearly distinct from colorless ^{i Bu}2 or green ^Me1·Cl₂/ ^Et1·Cl₂ ), and extraction into pentane and crystallization at low temperature conveniently afforded the target complex ^{i Bu}1·N₂ in reasonable yield (47%; Scheme 3c). Notably, the use of an elevated N₂ pressure (4 bar) was crucial to achieve a useful yield in this reaction. In its absence, while ^{i Bu}1·N₂ was still formed, numerous side products were also evident by ³¹P NMR spectroscopy (Figure 1). Associated small peaks in the hydride region of the ¹H NMR spectrum were also observed, and it is proposed that these may arise from cyclometallation of the iBu groups of the dibpe ligands, as has been reported for related reduced Fe(PP)₂ species [59]. This side‐reactivity could be facilitated by slower binding of N₂ to the bulky Fe center (vide infra).

SCHEME 3.

(a) Previous synthesis of ^Me1·N₂ and ^Et1·N₂ ; [4, 54, 55, 56, 57] (b) attempted synthesis of ^{i Bu}1·Cl₂ ; and (c) synthesis of ^{i Bu}1·N₂ .

FIGURE 1.

³¹P{¹H} NMR spectra for the synthesis of ^{i Bu}1·N₂ under 1 bar (top) or 4 bar (bottom) N₂ pressure, at equal concentrations (* = PPh₃ in capillary insert). Spectra not to equal scale (equivalent capillary insert used for both spectra).

Compound ^{i Bu}1·N₂ is stable for at least several months when stored at −30°C under Ar or N₂ in the solid state and is also stable for extended periods in aliphatic hydrocarbon solutions. However, rapid decomposition was observed in aromatic solvents such as benzene, and slower decomposition was also observed in donor solvents, which is believed to be due to displacement by solvent of the N₂ ligand and subsequent cyclometallation of the dibpe ligand (as seen for N₂‐free Fe⁰(depe)₂) [60, 61] over the course of several hours. As a result, where THF or Et₂O have been employed as solvents for subsequent studies, these solutions were prepared immediately prior to use.

Complex ^{i Bu}1·N₂ has been fully characterized and shows a singlet ³¹P NMR resonance at 79.7 ppm in per‐deuterated methylcyclohexane. Single crystals suitable for analysis by X‐ray diffraction analysis (XRD) were grown from pentane at −30°C and the solid‐state structure confirms the proposed formulation, showing a geometry similar to both ^Et1·N₂ and (^Me1)₂·N₂ (the latter being a dimeric structure with a bridging N₂ ligand that exists in equilibrium with monomeric ^Me1·N₂ ) [54]. Nevertheless, ^{i Bu}1·N₂ shows a significant distortion away from the expected trigonal bipyramidal geometry and toward a square‐based pyramidal structure (Figure 2). This can be attributed to the increased steric demands of the dibpe ligand and is quantified by a τ parameter of 0.6 (τ = 0 for idealized square‐based pyramidal, 1 for ideal trigonal bipyramidal geometries; [62] cf. τ = 0.89 for ^Et1·N₂ [63], τ = 0.99, 1.05 for (^Me1)₂·N₂ [54]).

FIGURE 2.

X‐ray crystallographic structure of ^{i Bu}1·N₂ . Selected bond lengths: Fe1‐P1, 2.1997(5); Fe1‐P2, 2.2077(5); Fe1‐N1, 1.791(3); N1‐N2, 1.125(4) Å. Selected bond angles: P1‐Fe1‐P1′, 140.72(3)°; P2‐Fe1‐P2′, 176.61(4)°; Fe1‐N1‐N2, 180.0° (required by symmetry). Ellipsoids drawn at 50% probability. Minor occupancy disorder positions and H atoms omitted for clarity. Fe atoms are shown in rust, P in magenta, C in grey, and N in blue. The two ligands are related by C₂ symmetry about the Fe‐N₂ axis, and one of the iBu groups is disordered over two independent orientations, only one of which is shown.

The crystallographic N—N distance of 1.125(4) Å in ^{i Bu}1·N₂ is identical (within error) to that of ^Et1·N₂ (1.142(7) Å) [63]. However, this is known to be a poor measure of the degree of N₂ activation in such complexes [16]. To obtain a more reliable indication of this key parameter, solution‐phase IR spectroscopy was carried out in Et₂O, which revealed an absorption at 1984 cm⁻¹ that was assigned to the N—N stretching frequency. This assignment was confirmed through preparation of the isotopically labelled complex ^{i Bu}1·¹⁵N₂ , for which this feature was shifted to a higher frequency of 1899 cm⁻¹ (ν(¹⁴N₂)/ν(¹⁵N₂) = 1.045 (experimental), 1.035 (theoretical; simple harmonic oscillator model)). The observed frequency is slightly higher than those for related Fe(N₂)(PP)₂ complexes (e.g. 1955 cm⁻¹ for ^Et1·N₂ , 1975 cm⁻¹ for ^Me1·N₂ ) [56, 57] and is consistent with a slightly less activated N≡N bond [64]. This conclusion is further supported by ¹⁵N NMR spectroscopic studies of ^{i Bu}1·¹⁵N₂ , with the ¹⁵N{¹H} spectrum showing two doublets at very close chemical shifts (δ = 331.1, 330.8 ppm, Δδ  = 0.3 ppm), indicating that the N atoms are in very similar chemical environments. For comparison, ^Me1·N₂ (for example) shows a larger separation of Δδ = 1.8 ppm, although the coupling constants ¹ J _NN for the two compounds are similar (6.5 Hz and 5.9 Hz, respectively) [54].

2.2. N₂ Reduction Using ^{i Bu}1·N₂

With compound ^{i Bu}1·N₂ in hand, investigations were carried out into its capacity for N₂ fixation, beginning with stoichiometric N₂ reduction (i.e. acidification with no external reductant, cf. Scheme 1a, i). Previously, excellent results had been obtained by treating ^Et1·N₂ with the strong acid TfOH in pentane at ambient temperature (20°C) [54] Gratifyingly, very similar results could be obtained using ^{i Bu}1·N₂ , with mixtures of NH₃ and N₂H₄ being obtainable with electron yields up to 51% (cf. up to 55% using ^Et1·N₂ ; for full details, including discussion of minor differences between the two systems, see SI Section 2.1 and Table S2). The fate of the final Fe/dibpe fragments in these reactions was not investigated in detail but is likely to include protonated bisphosphine ([dipbe(H)₂]²⁺) and Fe²⁺ salts, in line with previous studies of ^Et1·N₂ [53].

Having established that stoichiometric N₂ fixation can be achieved with high efficiency using ^{i Bu}1·N₂ , investigations proceeded to the use of ^{i Bu}1·N₂ for the catalytic reduction of N₂. For these studies, CoCp*₂ (Cp* = C₅Me₅) was used as an additional chemical reductant, having been shown to give optimal performance in analogous reactions using ^Et1·N₂ [53]. Surprisingly, however, with ^{i Bu}1·N₂ these reactions were found to give significantly inferior results. Indeed, even after surveying a range of different reaction conditions (see Table S3), the best result using ^{i Bu}1·N₂ yielded only 1.6 eq. of fixed nitrogen atoms (Scheme 4), which is an order of magnitude lower than for ^Et1·N₂ (up to ca. 30 eq.) and falls below the threshold of formal catalysis. Moreover, only trace N₂H₄ was observed (vide infra).

SCHEME 4.

Attempted catalytic fixation of N₂ by ^{i Bu}1·N₂ in the presence of CoCp*₂. All other reaction conditions tested gave inferior results (see Table S3).

2.3. Synthesis, Characterization, and Mechanistic Role of [ ^{i Bu}1]⁺

Though unexpected, the poor performance of ^{i Bu}1·N₂ under the catalytic reaction conditions may partially be explained by the low reaction temperatures, which are needed to minimize the rate of unproductive H₂ evolution (via direct side reaction between the acid and reductant). This was found to lead to significantly diminished results for the stoichiometric acidification procedure (see SI Section 2.1 and Table S2), likely due to kinetic factors. This suggests that the poor catalytic performance of ^{i Bu}1·N₂ is at least partially due to slow N₂ protonation and reduction kinetics (due in turn to the bulky dibpe ligands), rendering these steps uncompetitive with the competing, uncatalyzed H₂ evolution. Nevertheless, the magnitude of the apparent discrepancy in performance between the two regimes (and between ^{i Bu}1·N₂ and ^Et1·N₂ ) suggests that other factors could also be relevant, which prompted further mechanistic investigations. In particular, it was speculated that the contrast between the excellent performance of the optimized stoichiometric reaction and the poor performance of the optimized catalytic reaction may imply difficulties in regenerating the initial ^{i Bu}1·N₂ , as required to close the catalytic cycle, rather than just the transformation of ^{i Bu}1·N₂ into NH₃/N₂H₄.

Our previous studies of ^Et1·N₂ have highlighted the importance of the Fe(0/I) redox couple during N₂ fixation [53, 54]. During stoichiometric acidification this strongly reducing couple is believed to be crucial in supplying electrons for the reduction of N₂ to N _x H _y . Indeed, it is notable that the electron yields achieved for stoichiometric acidification of 1·N₂ seem to be limited to a maximum of approx. 50%. Since calculation of these yields assumes oxidation of Fe(0) to Fe(II) [4, 49] this may imply that, in practice, only one electron per Fe is readily available for N₂ reduction, presumably originating from the most reducing couple present, that is, Fe(0/I) only. Catalytic turnover can then be achieved by re‐reduction of the resulting Fe(I) species by an external reductant, to regenerate the Fe(0) starting material (Scheme 5) [65, 66].

SCHEME 5.

A simplified, outline mechanism for N₂ fixation mediated by complexes 1·N₂ .

To assess the reducing power of the Fe(0/I) couple in ^{i Bu}1·N₂ , electrochemical investigations were carried out. Cyclic voltammetry (CV) measurements on an Et₂O solution of ^{i Bu}1·N₂ under an N₂ atmosphere ([nBu₄N][BAr^F ₄] supporting electrolyte, Ar^F  = 3,5‐bis(trifluoromethyl)phenyl) revealed that initial oxidation of ^{i Bu}1·N₂ occurs irreversibly at very low potential (E _p,ox ∼−1.56 V vs Fc/Fc⁺ at 200 mV/s, Fc = ferrocene; Figure 3). While this irreversibility prevents a simple evaluation of the precise Fe(0/I) redox potential, it is clear that ^{i Bu}1·N₂ is a powerful reductant in line with other, similar Fe(N₂)(PP)₂ species, albeit not quite as powerful as ^Et1·N₂ (for which the equivalent redox potential is measured as −2.0 V vs. Fc/Fc⁺, with E_p,ox ∼ −1.88 V) [53, 54]. Notably, however, ^Et1·N₂ oxidation was fully reversible over scan rates between 100 and 2000 mV/s. The contrasting irreversibility for ^{i Bu}1·N₂ is consistent with rapid loss of the N₂ ligand upon oxidation and hence with a weaker interaction of N₂ with the oxidized [Fe(dibpe)₂]⁺ core vs. [Fe(depe)₂]⁺. This helps to explain the need for elevated N₂ pressures during the synthesis of ^{i Bu}1·N₂ (vide supra), since the Fe(0) product of this reaction is presumably formed via Mg reduction of the Fe(I) intermediate.

FIGURE 3.

Cyclic voltammogram of ^{i Bu}1·N₂ in Et₂O under N₂ using [nBu₄N][BAr^F ₄] electrolyte, referenced vs. Fc/Fc⁺ (200 mV/s).

In line with this interpretation of the CV results, treatment of an ethereal solution of ^{i Bu}1·N₂ with [Fc][BAr^F ₄] led to an immediate color change to deep purple. Notably, a similar transient purple color had also been observed during many of the earlier ^{i Bu}1·N₂ acidification reactions. Removal of solvent in vacuo, washing with pentane, and recrystallization from Et₂O yielded the four‐coordinate, 15 valence‐electron Fe(I) product [Fe(dibpe)₂][BAr^F ₄] ([ ^{i Bu}1][BAr^F ₄]; Scheme 6). XRD analysis of single crystals of [ ^{i Bu}1][BAr^F ₄] grown from Et₂O at −30°C revealed a square‐planar coordination geometry with only a slight tetrahedral distortion (Figure 4), very similar to the related depe‐substituted salt [ ^Et1][BAr^F ₄] [51]. Notably, the solid‐state structure showed no evidence of an N₂ ligand bound to the Fe center, even under N₂ atmosphere (vide infra).

SCHEME 6.

Synthesis of [ ^{i Bu}1][BAr^F ₄]. Fc = Fe(C₅H₅)₂.

FIGURE 4.

X‐ray crystallographic structure of [ ^{i Bu}1][BAr^F ₄]. Selected bond lengths: Fe1‐P1, 2.2449(7); Fe1‐P2, 2.2390(8); Fe1‐P3, 2.2433(7); Fe1‐P4, 2.281(7); Fe1‐P4', 2.245(9) Å. Selected bond angles: P1‐Fe1‐P3, 171.81(3)°; P2‐Fe1‐P4, 166.86(15)°; P2‐Fe1‐P4′, 173.9(3)° (P4′ not pictured). Ellipsoids drawn at 50% probability. Minor occupancy disorder positions in dibpe ligands, H atoms, and [BAr^F ₄]⁻ anion omitted for clarity. Fe atoms are shown in rust, P in magenta, and C in grey. P9 is disordered over two possible positions, resulting in two orientations of the dibpe ligand. Only the major occupancy orientation of that ligand is shown.

Analysis by the Evans NMR method [67] provided a solution‐phase magnetic moment of 1.98 µ_B, consistent with a low‐spin S  =  ½ electronic configuration (no observable change in magnetic moment is evident between 25 and −80°C, indicating that deviation from the spin‐only value of 1.73 µ_B is due to an orbital contribution to µ_eff rather than thermal population of a higher spin state). As such, solutions of [ ^{i Bu}1][BAr^F ₄] are silent to ³¹P NMR spectroscopy, although a pair of characteristic broad resonances were visible in the room temperature ¹H NMR spectrum (approx. −1.3 and −12.6 ppm in Et₂O; similar signals were observed for [ ^Et1][BAr^F ₄]) [51]. UV‐vis spectroscopy, meanwhile, revealed that the purple color of [ ^{i Bu}1][BAr^F ₄] is due to a number of electronic absorptions in the visible region, most notably a strong absorption at 520 nm (2960 M⁻¹ cm⁻¹).

Upon cooling below approx. 40°C under an atmosphere of N₂, solutions of [ ^{i Bu}1][BAr^F ₄] were observed to lose their purple coloration. Correspondingly, the UV‐vis features associated with [ ^{i Bu}1][BAr^F ₄] were no longer observed; instead, a new feature was observed at 365 nm. Extensive studies of the depe analogue [ ^Et1][BAr^F ₄] have shown an analogous color change to be due to reversible binding of N₂ to the [Fe(PP)₂]⁺ core, although in that case the transition was observed at significantly higher temperature [51]. Furthermore, while EPR spectroscopic analysis (CW, X‐band) of [ ^{i Bu}1][BAr^F ₄] under Ar at 40 K showed a clearly rhombic signal (consistent with an S  =  ½ species, g = [2.2697, 2.1798, 1.9973], g_av = 2.15), under N₂ a less anisotropic signal was observed (g = [2.0038, 2.088, 2.123], g_av  =  2.072; Figure 5 and Figure S8). These spectra are again very similar to those observed for uncoordinated [ ^Et1][BAr^F ₄] and N₂‐bound [ ^Et1·N₂ ][BAr^F ₄], respectively [51]. Collectively, these data strongly indicate reversible binding of N₂ to the [ ^{i Bu}1]⁺ cation and that this binding is favored only at low temperatures.

FIGURE 5.

CW X‐band EPR spectra of [ ^{i Bu}1][BAr^F ₄] recorded at 40 K as toluene glasses. The signal under Ar is shown in purple; degassing the tube and backfilling with N₂ results in the yellow signal.

Variable‐temperature (VT) UV‐vis spectroscopy allowed the strength of this binding to be quantified (Figure 6). Van't Hoff analysis between 188 and 233 K yielded values of ΔH = −6.9(5) kcal mol⁻¹ and ΔS = −25(2) cal K⁻¹ mol⁻¹ for N₂ binding to [ ^{i Bu}1][BAr^F ₄] (assessed based on the intense absorption at 520 nm observed for [ ^{i Bu}1][BAr^F ₄] and literature data for the solubility of N₂; for full details see SI Section 3). While the entropy value is similar to that obtained previously for [ ^Et1][BAr^F ₄] (ΔS = −27.6(1) cal K⁻¹ mol⁻¹), the enthalpy is significantly lower (cf. ΔH = −13.1(1) kcal mol⁻¹ for [ ^Et1][BAr^F ₄]) [51]. Notably, the VT UV‐vis data indicate that N₂ binding is incomplete even at the low temperatures investigated for catalysis (ca. 195 K). This confirmation that N₂ interacts much more weakly with the bulkier [ ^{i Bu}1]⁺ core suggests a further explanation for the poor catalytic performance of ^{i Bu}1·N₂ . In the absence of strong (and rapid) binding to N₂, it seems likely that reduction of [ ^{i Bu}1]⁺ does not lead cleanly to reformation of ^{i Bu}1·N₂ , but instead also produces significant quantities of four‐coordinate Fe⁰(dibpe)₂ ( ^{i Bu}1), which can then rapidly decompose via ligand cyclometallation reactions [59, 60, 61]. Indeed, attempted reduction of [ ^{i Bu}1]⁺ with a single equivalent of CoCp*₂ under one atmosphere of N₂ was found to regenerate ^{i Bu}1·N₂ alongside significant quantities of side products spectroscopically identical to those previously observed during attempts to synthesize ^{i Bu}1·N₂ under only 1 bar of N₂ (cf. Figure 2).

FIGURE 6.

VT UV‐vis spectra of [ ^{i Bu}1][BAr^F ₄] recorded in THF (0.15 mM) under N₂ atmosphere.

3. Conclusions

We have synthesized and thoroughly characterized the low‐valent iron dinitrogen complex Fe(N₂)(dibpe)₂, ^{i Bu}1·N₂ , which is a bulkier analogue of the related complex Fe(N₂)(depe)₂, ^Et1·N₂ . This has revealed that, despite showing comparable efficiency in the stoichiometric fixation of N₂, ^{i Bu}1·N₂ performs much more poorly than ^Et1·N₂ in catalytic protocols, which mechanistic investigations suggest may be due to a combination of slower N₂ functionalization kinetics and weaker N₂ binding to the Fe(dibpe)₂ moiety, particularly in the Fe(I) oxidation state. These results shed further light on the catalytic efficiency of Fe(N₂)(depe)₂ and suggest that it exists in a steric ‘sweet spot’: bulky enough to minimize the rate of unproductive protonation at Fe, but not bulky enough to slow N₂ binding and subsequent protonation at N_β to the point that they can be outcompeted by other unproductive side‐reactions (such as H₂ evolution and cyclometallation).

Also conspicuous is the very different distribution of azane products observed during (attempted) catalytic runs with ^{i Bu}1·N₂ when compared with ^Et1·N₂ or ^Me1·N₂ . This is again likely due to the greater steric profile of the dibpe ligands and may specifically be attributable to the greater steric protection afforded to the proximal N_α site. This should hinder functionalization at this less accessible position and so disfavor the “alternating” N₂ functionalization pathways that could lead to N₂H₄ formation and instead encourage “distal” pathways that form exclusively NH₃. This would be particularly true if N₂ reduction is achieved via H atom transfer from bulky donors such as [Cp*Co(C₅Me₅H)⁺], whose in situ formation by protonation of CoCp₂* has been described in detail by the Peters group [23, 68].

These observations highlight the subtle and complex steric balances that underpin N₂ reduction mediated by Fe(N₂)(PP)₂ complexes and the importance of comprehensive mechanistic understanding for the design of efficient catalytic systems. Similar conclusions are likely to apply to many other examples of homogeneous N₂ reduction catalysis.

4. Experimental Section

4.1. General Methods

All chemical manipulations were performed under an N₂ atmosphere using standard Schlenk line techniques or an MBraun Labmaster DP glovebox, unless stated otherwise. Solvents were purchased from VWR: MeOH was dried over several batches of 3 Å molecular sieves; benzene was dried over several batches of 5 Å molecular sieves; pentane and toluene were dried using an Innovative Technology Pure Solv SPS‐400; THF and Et₂O were distilled from dark green Na/fluorenone indicator. Solvents were degassed by thorough sparging with N₂ gas and stored in gas‐tight ampoules; pentane, benzene, toluene, and Et₂O were stored over a K mirror. Solvents for use under Ar were thoroughly sparged with Ar (BOC, Zero grade), followed by freeze‐pump‐thaw degassing. Deuterated solvents were freeze‐pump‐thaw degassed, dried, and stored in gas‐tight ampoules over a K mirror (C₆D₆, methylcyclohexane‐d₁₄) or 4 Å molecular sieves (THF‐d₈, CDCl₃). ¹⁵N₂ (Cambridge Isotope Laboratories, 98% ¹⁵N) was transferred from a breakseal flask using a Toepler pump. All glassware and stainless steel cannulae were dried in an oven at 160°C overnight before use.

NMR spectra were recorded using Bruker AV‐400 (400.4 MHz) spectrometers. Chemical shifts, δ, are reported in parts per million (ppm). ¹H chemical shifts are given relative to Me₄Si and referenced internally to the residual proton shift of the deuterated solvent employed. ¹⁹F, ³¹P, and ¹⁵N chemical shifts were referenced (δ = 0) externally to CFCl₃, 85% H₃PO₄ (aq), and neat CH₃NO₂, respectively. ¹H and ³¹P NMR spectra of solutions prepared in nondeuterated solvents incorporate an internal reference capillary containing a solution of approx. 0.1 M PPh₃ in C₆D₆ and are referenced to residual C₆D₅H and PPh₃ (δ = −5.3) resonances, respectively. Samples for low‐temperature measurements incorporated an internal reference capillary containing a solution of approx. 0.1 M PPh₃ in THF‐d₈ and are referenced to residual THF‐d₇. The paramagnetic susceptibility of samples was determined using the Evans NMR method [67]. An internal reference capillary containing C₆H₆ in THF‐d₈ was used for variable temperature Evans measurements.

Infrared (IR) spectra were recorded using a PerkinElmer FT‐IR Spectrum GX spectrometer. Solutions were recorded in an airtight Specac Omni CellTM filled with 0.01‐0.015 M solutions via syringe in a glovebox. The optics, detector area, and sample chamber of the spectrometer were purged with dry N₂ gas before each measurement to remove residual CO₂ and H₂O from the spectra. Samples measured as KBr pellets were combined with KBr (Sigma, FT‐IR grade) and ground to a powder, which was transferred into an airtight Specac die inside the glovebox. A Specac manual hydraulic press was then used outside the glovebox to press the pellet, which was then immediately measured to minimize sample decomposition.

Electronic spectra were recorded using a Perkin Elmer Lambda 20 UV‐visible spectrophotometer. Samples were prepared inside the glovebox using a quartz cuvette with an optical path length of 1 cm and fitted with a J. Young valve.

Electrochemical experiments were carried out using an AutoLab potentiostat controlled by Nova. Measurements were performed inside an Ar or N₂ glovebox on room temperature solutions containing the sample (2 mM) and [nBu₄N][OTf] electrolyte (50 mM). A three‐electrode configuration was employed: a Pt working electrode (PWE) (BASi, Indiana, USA); a Pt wire counter electrode (99.99%; GoodFellow, Cambridge, UK); and an Ag wire pseudo‐reference electrode (99.99%; GoodFellow, Cambridge, UK). All electrodes were polished using alumina/H₂O, and all electrodes were rinsed with Et₂O and dried in a 100°C oven prior to each measurement. Measurements were calibrated to the ferrocene/ferrocenium couple (10 mM) at the end of each run, and iR was compensated to within 80 ± 5% of the solution's uncompensated resistance.

CW‐EPR X‐band measurements were performed with a Bruker‐Biospin Micro EMXplus spectrometer equipped with a PremiumX microwave bridge, a cylindrical TE011 resonator (SHQE‐W), an ESR‐900 liquid helium cryostat, and an Oxford Instruments ITC‐503s temperature controller.

Single crystal X‐ray diffraction measurements were performed with an Oxford Diffraction Xcalibur unit; crystals were mounted on a nylon MicroLoop using perfluoropolyether oil and measured in a stream of N₂ at 173 K. Structures were solved in Olex2 [69] either by direct methods using the SHELXS solution program [70], charge flipping using Superflip [71], or charge flipping using the Olex2.solve structure solution program [72]. All data were subsequently refined with the ShelXL refinement package [73]. Deposition Number(s) 2475900 (for ^{i Bu}1·N₂) and 2475899 (for [ ^{i Bu}1][BAr^F ₄]) contain(s) the supplementary crystallographic data for this paper. These data are provided free of charge by the joint Cambridge Crystallographic Data Centre and Fachinformationszentrum Karlsruhe Access Structures service.

Elemental analyses were performed by Mr. S. Boyer of the London Metropolitan University.

MgBr₂·2Et₂O (Sigma‐Aldrich, 99%), FeCl₂ (Sigma‐Aldrich, 98%), NaBH₄ (Sigma‐Aldrich, 99%), 1, 2‐bis(dichlorophosphino)ethane (Sigma‐Aldrich, 97%), NH₄Cl (VWR, 99.5%), [CoCp*₂][PF₆] (Alfa Aesar, 98%), and TfOH (Sigma‐Aldrich, 99%) were purchased and used without further purification. Cp₂Fe was sublimed and recrystallized from pentane. Mg was heated at 140°C under high vacuum (1×10⁻² mbar) for 12 h. Para‐dimethylaminobenzaldehyde (pdmab) was purified by recrystallization from cold EtOH. KC₈ was prepared according to a literature procedure and ground into a fine powder using a mortar and pestle inside an Ar glovebox [74]. [Cp₂Fe][BAr^F ₄], [75] [Ph₂NH₂][BAr^F ₄], [53] CoCp*₂ [53, 76], 1, 2‐bis(diisobutylphosphino)ethane (dibpe), [77] and [nBu₄N][BAr^F ₄] [78] were prepared according to literature procedures.

4.2. Synthesis of Fe(N₂)(dibpe)₂ ( ^{i Bu}1·N₂)

FeCl₂ (178 mg, 1.40 mmol) was added to THF (40 mL) in a Rotaflo ampoule and refluxed at 75°C until the solid dissolved. dibpe (892 mg, 2.80 mmol), MgBr₂·2Et₂O (181 mg, 0.70 mmol), and Mg (269 mg, 11.2 mmol) were then added successively; the reaction was stirred for 2 min between each addition. The ampoule was then frozen in a liquid N₂ bath, the headspace evacuated, and backfilled with dry N₂. The ampoule was thawed at 0°C behind a blast shield, such that a pressure of ∼4 bar N₂ was achieved. The stirred reaction mixture was then allowed to warm to room temperature over the course of 2 h, and subsequently left stirring overnight. The pressure was released, and 1, 4‐dioxane (2 mL) was added. After stirring for a further 90 min, formation of a fine white precipitate was observed. The reaction mixture was filtered through a Celite pad on a sintered glass frit, and the solvent was removed in vacuo. The dark residue was extracted into pentane; this gave an orange solution, which was filtered 2x times through a Celite pad before being concentrated in a stream of N₂ and left at ‐30°C overnight. Red crystalline ^{i Bu}1·N₂ formed and was separated from the mother liquor, washed 2x with cold pentane, and dried under high vacuum (1×10⁻² mbar). Crystals obtained by this method were suitable for x‐ray diffraction measurements. 470 mg of ^{i Bu}1·N₂ was obtained (47%). (Note: the mother liquor from crystallization typically contains significant quantities of dibpe, which can be recycled by following the distillation procedure described in the synthesis of the ligand [ 77 ]).

Anal. Calcd. for C₃₆H₈₀FeN₂P₄: C, 59.99; H, 11.19; N, 3.89. Found: C, 59.84; H, 11.09; N, 3.80.

¹H NMR (400.4 MHz, methylcyclohexane‐d₁₄) δ / ppm: approx. 1.1 – 2.7 (br, m).

³¹P{¹H} NMR (162 MHz, methylcyclohexane‐d₁₄) δ / ppm: 79.7 (s).

IR (0.01 M Et₂O solution): ν(¹⁴N‐¹⁴N) = 1984 cm⁻¹.

4.3. Synthesis of Fe(¹⁵N₂)(dibpe)₂ ( ^{i Bu}1·¹⁵N₂)

FeCl₂ (36 mg, 0.28 mmol) was added to THF (40 mL) in a J. Young ampoule with side arm attachment and refluxed at 75°C until the solid dissolved. dibpe (178 mg, 0.56 mmol), MgBr₂·2Et₂O (36 mg, 0.14 mmol), and Mg (54 mg, 2.25 mmol) were added successively; the reaction was stirred for 2 min between each addition. The ampoule was then frozen in a liquid N₂ bath, and a single freeze‐pump‐thaw cycle was performed. Under static vacuum, the ampoule was connected to a Toepler pump, and ¹⁵N₂ was delivered. A pressure of ∼1 bar was obtained. The reaction and work‐up were then performed as detailed above, with all further manipulations carried out under Ar. Red crystalline ^{i Bu}1·N₂ was obtained on recrystallization from pentane (42 mg, 21%).

¹⁵N NMR (40.55 MHz, pentane) δ / ppm: 331.1 (d, ¹J_NN = 6.5 Hz), 330.8 (d, ¹J_NN  =  6.5 Hz).

IR (0.01 M Et₂O solution): ν(¹⁵N‐¹⁵N) = 1899 cm⁻¹.

4.4. Synthesis of [Fe(dibpe)₂][BAr^F ₄] ([ ^{i Bu}1][BAr^F ₄])

[Cp₂Fe][BAr^F ₄] (69 mg, 0.07 mmol) was dissolved in Et₂O (5 mL) and added dropwise to a stirred solution of ^{i Bu}1·N₂ (50 mg, 0.07 mmol) in Et₂O (5 mL) under Ar. An immediate color change to deep purple was observed, and the solution was stirred for 10 min. Pentane (10 mL) was added, and the solution slowly cooled to ‐30°C and maintained at that temperature overnight. The supernatant was decanted, and the purple solid was washed with pentane. The solid was then recrystallized from minimum Et₂O at ‐30°C and again washed with pentane, then dried in vacuo. Purple crystalline [ ^{i Bu}1][BAr^F ₄] was obtained (80 mg, 79%).

Anal. Calcd. for C₆₈H₉₂BFeF₂₄P₄: C, 52.48; H, 5.96. Found: C, 52.42; H, 5.93.

¹H NMR (400.4 MHz, Et₂O) δ / ppm: 7.70 (s, meta‐Ar^F), 7.50 (s, para‐Ar^F), −1.32 (br, s), −12.64 (br, s).

¹¹B NMR (128.4 MHz, Et₂O) δ: −6.6 (s).

¹⁹F NMR (376.4 MHz, CDCl₃) δ: −62.4 (s).

UV‐VIS (THF, nm {m² mol–¹}): 269 {698}; 391 {130}; 520 {296}.

4.5. Acidification of ^{i Bu}1·N₂

Acidification reactions (with or without additional reductant) were performed, and their outcomes evaluated, in line with our previously published procedures [53, 54]. Reactions were performed using 0.008 mmol of ^{i Bu}1·N₂ in 1.75 mL of solvent.

Conflicts of Interest

The authors declare no conflict of interests.

Supporting information

Supporting File 1: chem70489‐sup‐0001‐SuppMat.pdf.