The Reactivity Patterns of Low-Coordinate Iron Hydride Complexes

Abstract

We report a survey of the reactivity of the first isolable iron-hydride complexes with a coordination number less than five. The high-spin iron(II) complexes [(β-diketiminate)Fe(μ-H)]₂ react rapidly with representative cyanide, isocyanide, alkyne, N₂, alkene, diazene, azide, CO₂, carbodiimide and Brønsted acid containing substrates. The reaction outcomes fall into three categories: (1) addition of Fe-H across a multiple bond of the substrate, (2) reductive elimination of H₂ to form iron(I) products, and (3) protonation of the hydride to form iron(II) products. The products include imide, isocyanide, vinyl, alkyl, azide, triazenido, benzo[c]cinnoline, amidinate, formate, and hydroxo complexes. These results expand the range of known bond transformations at iron complexes. Additionally, they give insight into the elementary transformations that may be possible at the iron-molybdenum cofactor of nitrogenases, which may have hydride ligands on high-spin, low coordinate metal atoms.

Introduction

Organometallic chemists have long appreciated the many reactions of transition-metal hydride complexes, especially reductions.₁ Recently, hydride chemistry has become established in bioinorganic systems.₂ Nitrogenases are prodigious reductants that cleave double and triple bonds in N₂, CO₂, N₂O,N₃⁻, and CN⁻ at iron-sulfur clusters (FeMoco in the molybdenum-iron nitrogenases, FeVco in the vanadium-iron nitrogenases, and FeFeco in the iron-only nitrogenases).³ Because these substrates are reduced by multiples of two electrons accompanied by the addition of protons, chemists have often speculated about the potential role of hydrides.⁴

Recently, the first direct evidence for hydride intermediates in nitrogenase emerged from ENDOR (Electron-Nuclear Double Resonance) studies of nitrogenase mutants freeze-trapped during substrate turnover. For example, a Val70Ile mutant of A. vinelandii molybdenum-iron nitrogenase freeze-trapped during proton reduction shows two hydrogen nuclei with very strong coupling to the S = ½ iron-sulfur cluster, strongly suggesting the presence of Fe-H bonds.⁵ This frozen species loses two molecules of H₂ upon annealing to −20 °C, raising the possibility that these two hydride ligands can combine with nearby protons to release H₂.⁶ This species is thought to be in the redox state that reacts directly with N₂ in the wild-type MoFe nitrogenase,⁷ underscoring the importance of hydride-containing intermediates in enabling the nitrogen reduction activity characteristic of this enzyme.

From the perspective of the coordination chemist, there are a number of ways that potential FeMoco-bound hydrides differ from the majority of known synthetic transition-metal hydride complexes. First, FeMoco hydride adducts could achieve a number of different oxidation levels. One-electron redox changes undoubtedly occur in the FeMoco because electrons are supplied to the FeMoco one at a time by the Fe protein, which dissociates and reassociates before each one-electron reduction of the FeMoco.³ However, most synthetic hydride complexes have diamagnetic metal centers, and little one-electron chemistry has been reported.^9,10 Second, the iron ions in the FeMoco are expected to be high-spin, based on ligand field considerations (weak-field sulfide donor set, coordination number less than five) and the results of computational studies.¹¹ Synthetic hydride complexes, on the other hand, typically have strong-field organometallic or phosphine co-ligands, which enforce a low-spin electronic configuration.^12,13 These fundamental differences motivate synthetic research aimed at the creation of iron-hydride complexes with weak-field ligands and low coordination number, to determine their characteristic reactivity patterns and reaction mechanisms, which can in turn be correlated with nitrogenase reactions.

Often the design of functional models of an enzyme is driven by an attempt to achieve structural similarity to the enzyme’s active site.¹⁴ In the case of nitrogenase, an accurate structural mimic of the FeMoco hydride species is elusive for several reasons. First, the FeMoco of iron-molybdenum nitrogenase (Figure 1) has eight transition metals in a cluster type (M₈S₉C or M₈S₉N or M₈S₉O) that is unknown in synthetic chemistry.^15,16 Second, the crystallographically characterized form of the FeMoco (M^N) does not have hydrides, as shown by ENDOR: hydrides are only incorporated concurrent with catalytic turnover.⁵ In any case, X-ray crystal structures of proteins do not have sufficient resolution to distinguish hydrogen atoms. Third, the N₂-binding form of the FeMoco is reduced by 3-4 electrons from M^N, and the FeMoco may undergo structural rearrangements upon reduction.^7,17 Peters and Holland have independently proposed that dissociation of X is important in enabling substrate binding to iron, based on the binding of N₂ to synthetic iron complexes.¹⁸ Computational studies come to a variety of conclusions regarding the structural flexibility of the FeMoco during catalysis.¹¹ In short, there are many questions about the structure of the iron-hydride nitrogenase intermediates, and the available data are not sufficient to provide guesses about their structure(s).

Figure 1.

The iron-molybdenum cofactor (“FeMoco”) of iron-molybdenum nitrogenase in the isolated M_N form.⁸ This form is reduced (with incorporation of hydrides in an undisclosed location) to give the intermediate that reacts with N₂ and other nitrogenase substrates. X is C₄⁻, N₃⁻, or O₂⁻.

For these reasons, we have created one- and two-iron compounds that focus on two key features of the nitrogenase-hydride intermediates: the weak-field ligands and the coordination number less than five.^18a Extremely bulky bidentate β-diketiminate ligands (Figure 2) have π-donating nitrogen atoms that lead to a weak ligand field.¹⁹ The bulky diisopropylphenyl groups maintain a low coordination number that mimics the ligand-poor environment of the iron atoms in the FeMoco. Using these ligands, we have isolated the only examples of iron-hydride complexes with a coordination number less than five (Figure 2).^20,21

Figure 2.

Bulky β-diketiminate iron complexes used in this work. The ligands are abbreviated L^R, where R is the group on the diketiminate backbone. Both complexes have roughly tetrahedral geometry at each iron atom.²⁰,²¹

Our studies on low-coordinate iron are complemented by those of Peters and coworkers, who use tridentate, strong-field tris(phosphino)borate (BP₃) supporting ligands that contain “soft” phosphine donors, and more often give low-spin electronic configurations.²² In the BP₃ systems, it has not yet been possible to isolate a low-coordinate iron hydride, but the presence of a terminal hydride in four coordinate complexes is strongly implied by the isolation of products that result from activation of solvent or the supporting ligand.²³ Bridging hydrides have recently been observed in a few electronically unsaturated dinuclear complexes,^24,25 and borohydrides have been studied on iron-sulfur clusters.²⁶

Recent contributions from our group have described the reaction of [L^tBuFeH]₂ with azobenzene to cleave the N=N bond,^20,27 and the reaction of [L^MeFeH]₂ with boranes to cleave B-C bonds.²⁸ This manuscript reports a wider variety of reactions with unsaturated small molecules that are nitrogenase substrates or organic-soluble mimics of these substrates. By examining the reactions with representative compounds, one may begin to understand some of the unusual reaction patterns available to hydride complexes with a high-spin electronic configuration. In addition, a number of the products have novel structural and/or electronic features that are interesting from a fundamental perspective.

Results

Iron-Hydride Starting Materials

The synthesis and characterization of [L^MeFeH]₂ (1a) and [L^tBuFeH]₂ (1b) have been presented previously.^20,21,27,28 They are each synthesized from iron(II) chloride complexes ([L^MeFeCl]₂ or L^tBuFeCl) by reaction with potassium triethylborohydride for 0.5 h in toluene, followed by promptly removing solvent and BEt₃, then extracting with pentane and crystallizing to give brown powders or crystals. Crude reaction mixtures are contaminated with the chloride starting material and the dihydridoborate complex L^RFe(μ-H)₂BEt₂.²⁸ Multiple crystallizations are typically necessary to remove these impurities, and ¹H NMR spectroscopy is used to judge purity of the hydride complexes. The good yields of many of the products derived therefrom support the purity of 1a and 1b.

The ¹H NMR spectrum of 1a consists of seven resonances; the number of signals and their integrations are characteristic of the diketiminate ligand in a local C_2v symmetry environment. It has no unusual temperature- or concentration-dependent changes in its ¹H NMR or UV-vis spectra, suggesting that the dimeric structure in the solid state is always maintained. To learn more about the electronic structure of 1a, we examined a solid sample using Mössbauer spectroscopy. The zero-field Mössbauer spectrum at 80 K (Figure S-6) exhibits a superposition of two quadrupole doublets with 72% and 28% relative intensities. The major component has δ = 0.70(2) mm/s and ΔE_Q= 0.86(2) mm/s, whereas the minor component shows δ = 0.49(2) mm/s and ΔE_Q = 2.06(2) mm/s. The values for the minor component are relatively similar to those of a high-spin Fe(I) diketiminate complex (ΔE_Q = 2.02 mm/s, δ = 0.41 mm/s at 180 K).^19c Therefore, we assign the component to contamination from an unidentified Fe(I) impurity. In contrast, the Mössbauer parameters of the major component, particularly the isomer shift, are in the range of high-spin iron(II) diketiminate complexes (δ = 0.62 – 0.86 mm/s), and so it is assigned to 1a.^19,29 Magnetic Mössbauer spectra measured at 4.2 K with applied fields up to 7 T show that the major component is diamagnetic, suggesting antiferromagnetic coupling between the two iron(II) ions. In sum, the spectroscopic data support the formulation of 1a as an exchange-coupled pair of high-spin iron(II) ions with a ground state of S = 0.

The ¹H NMR spectrum of 1b has at least 17 overlapping peaks, because of extreme steric crowding in the dimer that presumably renders some ligand bond rotations slow on the NMR time scale.²⁷ Upon heating a solution of 1b in C₆D₆, there is growth of a simple seven-line ¹H NMR spectrum with chemical shifts much like trigonal-planar alkyl complexes L^tBuFeR.²⁰ This behavior is ascribable to equilibration between dimeric [L^tBuFeH]₂ and monomeric L^tBuFeH that is slow on the NMR time scale (ms) but rapid on the time scale of the equilibration (min). Therefore, in the discussion below it should be understood that 1b does not refer specifically to the monomer or dimer, but the equilibrating mixture of the two.³⁰

Reactions With C≡N Triple Bonds, and with N₂ and CO

Carbon monoxide (CO) is an inhibitor of N₂ reduction by nitrogenase.³ Methyl isocyanide (CH₃NC) and cyanide (CN⁻) are also inhibitors but are substrates as well: CN⁻ is reduced to CH₃NH₂, NH₃, and CH₄, while CH₃NC is reduced to CH₃NH₂, (CH₃)₂NH, and CH₄.³¹ Therefore, it is of interest to examine the reactivity of small molecules containing CO and CN triple bonds with synthetic hydride complexes that mimic potential activated forms of the FeMoco. We use alkyl cyanides as surrogates for CN⁻ and tert-butyl isocyanide in place of CH₃NC.

The addition of CO or isocyanide to 1a results in the rapid release of H₂. Adding an excess of CO to 1a gives L^MeFe(CO)₃ through formal reduction of the iron. Because L^MeFe(CO)₃ has been characterized previously, the reader is referred to the earlier paper for its properties.⁴⁷

Addition of a large excess of tert-butyl isocyanide gives an intractable mixture, but treatment of a solution of 1a in pentane or toluene with 4 equiv (per dimer) of tert-butyl isocyanide results in the formation of a mononuclear iron(I) complex, L^MeFe(CN^tBu)₂ (2a). Integration of the ¹H NMR spectrum against an internal standard (L^tBuFeCl in a capillary) indicated 62% conversion to 2a, amongst other unidentified products. The production of H₂ was quantified by GC to be ca. 0.2 equiv per mole of 1a. The low yield of dihydrogen may be due to hydrogen incorporation into some of the unidentified products. Because of the low conversion, samples of 2a for further spectroscopic study were generated through an alternative method, by adding 4 equiv of ^tBuNC to L^MeFeNNFeL^Me. The ¹H NMR and X-band EPR spectra of samples prepared in this way were identical to the spectra of material generated from 1a.

Solutions of the isocyanide complex 2a decompose over several hours in solution at room temperature, making the results of bulk magnetic studies unreliable, but EPR spectra of frozen mixtures of L^MeFeNNFeL^Me and 1–4 equiv of ^tBuNC in toluene show a rhombic signal with g = 2.45, 2.24, 1.98, suggesting that 2a has an S = ½ ground state.^32,33 In the solid-state infrared spectrum of 2a, four stretching bands are observed at 2122, 2050, 1969, and 1948 cm⁻¹. Some bands almost certainly derive from decomposition products, because bands at 2050, 1969 and 1948 cm⁻¹ are observed in the IR spectrum of a sample from allowing 2a to completely decompose. Therefore, only the peak at 2122 cm⁻¹ can be assigned confidently as a C≡N stretching vibration of 2a. Computations (see below for details) predict the second stretching vibration at 2050 cm⁻¹ after correction for anharmonicity, so it is possible that the peak at 2050 cm⁻¹ derives from 2a as well as a decomposition product.

The product has also been characterized by X-ray diffraction of a crystal grown at low temperature. The solid-state structure of 2a is illustrated in Figure 2. The iron atom has a square planar geometry, with the sum of angles 360.4(5)°. The two isocyanides are not identical: one CNC angle is almost linear (170.1(2)°) while the other is bent. The bent isocyanide ligand is disordered; two positions of the ^tBu group were observed, and the ratio of components is exactly 50:50 due to the presence of a crystallographic inversion center. Both refined ligand positions show bent isocyanide ligands (CNC angles of 164.0(7)° and 148.3(7)°). The refinement model required that geometrical restraints be placed on the N^tBu portions of the two ligand disorder components, specifically that the N–C(^tBu) bond lengths and angles in both components be similar. The result is a disorder model that confirms the CNC bending, but that should be used with caution with regard to the exact values of the angles.

The ambiguous spectroscopic and crystallographic data left unanswered questions about 2a. Therefore, hybrid quantum mechanics/molecular mechanics (QM/MM) computations were performed to study complete models of 2a. The optimized geometry of the doublet state (E_relative = 0) is square planar about iron, but the higher energy quartet state (E_relative = 3.9 kcal/mol) has a tetrahedral coordination geometry upon QM/MM geometry optimization. The calculated metrical parameters for doublet L^MeFe(CN^tBu)₂ (compared to experimental ones in parentheses) are Fe-C = 1.83 Å (1.82 Å); CN = 1.18 Å (1.20, 1.17 Å). Both the lower calculated energy and the geometrical similarity to the experimental structure support the contention that 2a has an S = ½ ground state.

The computations also address the unusual difference between the CNC angles of the two isocyanide ligands (we denote the difference between these CNC angles as θ). The optimized geometry had CNC angles of 171° (close to the crystallographic angle of 170°) and 162° (between the observed 164° and 148° disorder components). A search of the Cambridge Structural Database for neutral, monometallic complexes with at least two ^tBuNC ligands indicated that the average θ is 6.9 ± 10.1° (170 examples) with a median value of 2.9°.³⁴ Hence, our computed value of θ = 9°, while somewhat large, is not outside experimental norms. Computations indicate that bending the CNC angle from 180° to 150° requires 4 kcal/mol in free ^tBuNC and only 0.4 kcal/mol in the QM/MM model of 2a.³⁵ Considering the softness of this bending distortion, it is not surprising that the isocyanide is unusually flexible, and can exist in multiple geometries in the solid state.

In the next section, we turn to reactions with nitriles, mimics of cyanide with substituents on the carbon. Scheme 1 shows the reduction products obtained from CH₃CN and ^tBuCN. Heating 1a with 2 equiv of CH₃CN in toluene at 45 °C for 18 h gives an orange-yellow precipitate in 88% yield. The X-ray crystal structure shows that the product is L^MeFe(μ-N=CHCH₃)₂FeL^Me (3a) (Figure 4, top), with a crystallographic inversion center relating the two halves of the molecule. Each nitrile molecule has been reduced to a bridging N=CHCH₃ ligand with an N=C distance of 1.258(2) Å. The iron centers in 3a have roughly tetrahedral geometries. The ¹H NMR spectrum of 3a in toluene-d₈ has peaks from 50 ppm to −40 ppm, and the number and integrations are consistent with the C_2h symmetry in the crystal structure.

Scheme 1.

Figure 4.

Molecular structures of 3a (top) and 4a (bottom), showing 50% thermal ellipsoids. The aryl isopropyl groups are omitted for clarity. Selected bond distances (Å) and angles (°) for 3a: Fe-N3 2.041(2), Fe-N3′ 2.079(2), N3-C14 1.258(2), Fe-N3-Fe′ 96.72(6), N3-C14-C24 125.4 (2). For 4a: Fe1-Fe2 2.7816(8), Fe1-N5 2.000(3), Fe2-N5 2.016(3), Fe1-H1 1.69(3), Fe2-H1 1.67(3), N5-C15 1.258(4), Fe1-N5-Fe2 87.7(1), N5-C15-C25 129.3(3).

When 1a is treated with the bulkier nitrile ^tBuCN, brown L^MeFe(μ-N=CH^tBu)( μ-H)FeL^Me (4a) is obtained in 85% yield. The diiron complex incorporates only one equivalent of the nitrile, even in the presence of excess ^tBuCN. Addition across an N≡C bond has formed a bridging N=CH^tBu ligand, while the other bridging hydride remains untouched. The crystal structure of 4a (Figure 4, bottom) indicates that the large ^tBu group pushes the two β-diketiminate ligands toward the bridging hydride, which significantly reduces the dihedral angle between the two β-diketiminate planes to around 81° in each of the two independent molecules in the crystal structure. This structural distortion makes the ¹H NMR spectrum more complicated than in 3a. At ambient temperature, approximately 16 peaks are observed. Severe overlap between peaks has made attempts to assign peaks and measure the solution magnetic moment unsuccessful. At −20 °C, these ~16 peaks split into ~25 peaks, consistent with a reduction of symmetry from C_s (in a conformation like that in the crystal structure, but with the two β-diketiminate ligand planes coplanar on average) to C₁ (the symmetry in the solid state).

Another kind of insertion product is observed when 1b is treated with ^tBuCN. In this case, the orange product is monomeric L^tBuFe(N=CH^tBu) (5b) in 67% yield. The geometry around the iron atom is trigonal planar (Figure 5). A C=N double bond is indicated by the N14-C14 bond length of 1.256(3) Å and the C=N stretching band at 1687 cm⁻¹. The Fe-N(imide) bond length is 1.857(2) Å, substantially shorter than the Fe-N distances in the bridging imides of 3a and 4a.

Figure 5.

Molecular structure of 5b, showing 50% thermal ellipsoids. Most hydrogen atoms have been omitted for clarity. Selected bond distances (Å) and angles (°): Fe-N14 1.857(2), Fe-N11 1.977(2), Fe-N21 1.966(2), N14-C14 1.256(3), N11-Fe-N21 94.69(6), N11-Fe-N14 126.53(7), N21-Fe-N14 138.70(7), Fe-N14-C14 143.0(2).

Some triple bonds do not insert into the Fe-H bond. There is no evidence for thermal reaction of hydride complexes 1a and 1b with N₂: use of high pressures of dinitrogen (up to 800 psi at 60 °C) gave L^tBuFeOFeL^tBu, the product of reaction with moisture, as the only product observable by ¹H NMR spectroscopy.^29b However, irradiation of each hydride complex under 1 atm of N₂ with a high-pressure mercury lamp yields quantitative conversion to L^RFeNNFeL^R over the course of 18 h (1a) or 3 d (1b) as shown by ¹H NMR spectroscopy. These dinitrogen complexes have a formal oxidation state of +1 at each iron atom, and therefore result from reduction.³⁶ Therefore, it is possible to induce these hydride complexes to reductively eliminate H₂ and bind N₂, but this transformation requires photolysis.

Reactions With C≡C Triple Bonds

Acetylene is the most commonly studied unnatural substrate of nitrogenases, and it is reduced to ethylene.³⁷ Mutant nitrogenase enzymes can also reduce longer-chain alkynes.^38,39 Here, we examine the reactivity of the low-coordinate iron hydride complexes with alkynes diphenylacetylene and 3-hexyne.

Complex 1a inserts into the C≡C triple bond in PhCCPh to form L^MeFeC(Ph)=C(H)Ph (6a) in 75% yield. This vinyl product is analogous to L^tBuFeC(Et)=C(H)Et (6b), which was previously reported from the reaction of 1b with 3-hexyne.²⁰ Figure 6 presents the crystal structure of 6a. The C-C distance (1.347(2) Å) and C-C-C angles (129.0(2)°, 121.6(2)°) in the vinyl ligand confirm that both carbons of the C=C double bond have sp² hybridization. The ¹H NMR spectrum of 6a shows seven peaks for the β-diketiminate protons, indicating averaged C_2v symmetry from rapid rotation around the Fe-C14 bond on the NMR time scale.

Figure 6.

Molecular structure of L^MeFeCPh=C(H)Ph (6a), showing 50% thermal ellipsoids. Selected bond distances (Å) and angles (°): Fe-C14 2.017(2), C14-C24 1.347(2), C14-C24-C34 129.0(2), C24-C14-C94 121.6(2).

Monitoring the reaction of 1a with PhC≡CPh, or the reaction of 1b with EtC≡CEt, shows no dependence of the reaction rate on [PhC≡CPh] (Table 2). Therefore, the rate laws are rate = k[1] with first-order rate constants 1.7(2)×10⁻³ s⁻¹ (1a/PhCCPh at 31 °C) and 5.0(5)×10⁻⁴ s⁻¹ (1b/EtCCEt at 15 °C). The first-order rate law is inconsistent with the interaction of alkyne with 1a or 1b during or prior to the rate-limiting step of the reaction. Two mechanistic possibilities consistent with the rate law are shown in Scheme 2. In pathway A, the opening of a single Fe-H bond is the rate determining step. We consider this Fe-H bond opening pathway to be more likely for the reaction of alkyne with 1a because (1) there is no other evidence for any monomer form of 1a by NMR or UV-vis spectroscopy; (2) in the reaction of 1a with boranes, the rate law was first-order in [1a] and first-order in [BEt₃] but independent of the steric demands of the borane or iron complex.²⁸ These data were inconsistent with dissociation of 1a into monomers (which predicts a half-order dependence on [1a]) and most consistent with single Fe-H opening, as in the bottom reaction pathway here. In pathway B, dimer cleavage is the rate determining step, as proposed for the reaction of 1b with 3-hexyne.²⁰ This pathway is reasonable for 1b, because monomer is rapidly accessed at room temperature (see above). The rate of the reaction is consistent with the rapid equilibrium of monomer and dimer observed previously.²⁰ So, although the evidence is not definitive, the kinetic data here and elsewhere are most indicative of the pathway A in the reactions of alkynes with 1a, but pathway B for alkyne reactions with 1b. This difference is consistent with the greater steric demands of the diketiminate ligands in 1b than 1a.

Table 2.

Effect of [PhC≡CPh] on the rate of reaction with 1a to give 6a.^a,^b

[PhC≡CPh] (mM)	[PhC≡CPh]/[1a]	kobs (s−1)
184	10.5	1.7(1)×10−3
362	20.6	1.6(1)×10−3
725	41.2	1.7(2)×10−3

^a[1a] = 17.6 mM in C₆D₆ at 30.8 °C.

^bDetails of the reaction of 1b with 3-hexyne are given in ref 20.

Scheme 2.

Reactions with C=C bonds

We previously reported that L^MeFe(alkyl) complexes with β-hydrogens can act as sources of transient hydride species L^MeFe(H)(alkene). These react with alkenes to give insertion products, and complexes L^MeFeR′ (R′ = ethyl, 1-propyl, 2-propyl, 1-butyl, 2-butyl, cyclohexyl, 2-phenethyl) were fully characterized.³⁰ We have also characterized numerous alkyl complexes of iron with the larger L^tBu ligand.⁴⁰ With isolated iron hydride complexes in hand, we verified that addition of alkenes to isolated 1a or 1b gives alkyl complexes of the same type through [1,2] addition across the double bond. For example, reaction of 1a with 1-hexene quantitatively gives a 1-hexyliron complex, as judged by ¹H NMR spectroscopy. It is worth noting that the transient iron hydride L^MeFe(H)(alkene) also adds across the C=N bonds of imines (forming L^MeFeNR′CHR₂ from R₂C=NR′) and the C=O bonds of ketones (forming L^MeFeOCHR₂ from R₂C=O).³⁰ Because these reactions have already been characterized starting from the alkyl complexes,³⁰ the reactions with isolated hydride complexes were not investigated further.

Reactions With Azides

The azide ion (N₃⁻) is transformed by nitrogenase into N₂, NH₃, and N₂H₄.⁴¹ Here, substituted azides AdN₃ and Me₃SiN₃ are used as models. The reactions of organic azides with transition-metal complexes have been reviewed recently.⁴²

Reactions of complex 1a with Me₃SiN₃ give mixtures, as judged from ¹H NMR spectra. However, complex 1b, when treated with 2 equiv of trimethylsilyl azide in diethyl ether, gives a dimeric azide complex, L^tBuFe(μ-N₃)₂FeL^tBu (7b) in 79% yield (Scheme 3). Compound 7b can be prepared independently by reacting L^tBuFeCl⁴³ with sodium azide.

Scheme 3.

The X-ray crystal structure of 7b (Figure 7) shows that the compound is dinuclear in the solid state with an Fe···Fe distance of 5.0966(4) Å. Each iron atom is coordinated to two azide ions, which bridge the iron atoms in a μ_1,3-end-to-end fashion with inequivalent iron-nitrogen bond lengths of 2.061(1) and 2.025(1) Å. The eight atoms of the Fe₂N₆ core are coplanar, in contrast to the only other crystallographically characterized end-to-end bridged iron azide complex, [(PhBP₃)Fe(μ-1,3-N₃)]₂ (PhBP₃ = [PhB(CH₂PPh₂)₃]⁻), which has a chair conformation.⁴⁴ The dihedral angle between the Fe₂N₆ plane and the Fe₂N₂(diketiminate) plane is 87.41(4)°. The azide ligands are almost linear (N-N-N angle of 177.0(1)°) and the N-N bond lengths within the azide groups (1.167(2), 1.176(2) Å) show double bond character similar to N-N distances in other bridging azide complexes.³⁴ The solid-state IR spectrum of 7b shows characteristic azide bands at 2129 and 2081 cm⁻¹.⁴⁵

Figure 7.

Molecular structure of L^tBuFe(μ-N₃)₂FeL^tBu (7b), showing 50% thermal ellipsoids. Selected bond distances (Å) and angles (°): Fe-N15 2.061(1), Fe-N35 2.025(1), Fe-N11 1.988 (1), Fe-N21 1.989(1), N15-N25 1.167(2), N25-N35 1.176(2), N11-Fe-N21 97.10(4), N15-Fe-N35 95.17(4), N15-N25-N35 177.0(1).

²⁹Si{¹H} NMR analysis of the volatile products in the reaction mixture revealed a resonance at −22 ppm for Me₃SiSiMe₃, but none for Me₃SiH. The production of Me₃SiSiMe₃ implies that two hydrogen atoms were lost. However, H₂ was not detected in the headspace by gas chromatography, so the fate of the hydrogen atoms is unclear. The coupling of two trimethylsilyl units strongly implies the intermediacy of Me₃Si^· radicals. These in turn might derive from the attack of an iron radical on trimethylsilyl azide.⁴⁶ In order to test this idea, the iron(I) sources L^tBuFeNNFeL^{tBu 47} and L^tBuFeClK(solvent)_x ⁴⁷ were each treated separately with Me₃SiN₃. In each case, [L^tBuFeN₃]₂ was formed. Therefore, iron(I) species of this type are reasonable intermediates that could be formed by homolysis of Fe-H bonds. Homolysis of Fe-H bonds does not occur spontaneously in 1b at room temperature, and therefore we infer that coordination of Me₃SiN₃ weakens the Fe-H bond and brings about Fe-H homolysis.

Compound 1b undergoes an insertion reaction with AdN₃ to give 67% yield of a triazenido complex, L^tBuFe(η₂-HNNNAd) (8b), which is shown in Scheme 4 and Figure 8. The bidentate triazenide and β-diketiminate ligands are perpendicular to each other, which gives a distorted tetrahedral geometry around the metal center. The presence of an N-H bond in 8b is confirmed by the observation of ν_N-H at 3371 cm⁻¹ in the infrared spectrum. The ¹H NMR spectrum of 8b is indicative of C_2v local symmetry at temperatures from room temperature to −75 °C, despite the C_s idealized symmetry of the molecule in the solid state (each face of the diketiminate ligand should be inequivalent). This observation implies that there is rapid exchange between the two possible orientations of the triazenido ligand. Considering the stability of three-coordinate complexes L^tBuFeX and the large size of the adamantyl group, it seems most reasonable to attribute this behavior to a rapid, reversible isomerization of the triazenido ligand from η² to η¹, followed by rotation and re-coordination.

Scheme 4.

Figure 8.

Molecular structure of triazenido complex 8b, showing 50% thermal ellipsoids. Selected bond distances (Å) and angles (°): Fe-N1 2.081(2), Fe-N3 2.078(2), Fe-N11 1.878(4), Fe-N21 2.117(4), N1-N2 1.310(3), N2-N3 1.285(2), N3-C1 1.463(7), N11-Fe-N21 98.32(9), N1-Fe-N3 60.46(7), N1-N2-N3 107.6(2).

Although disubstituted triazenido complexes are well-known in the literature, there are few examples of triazenido complexes bearing H as one substituent; they come from insertion of azide into a hydride complex,^48,49 and from protonation of an azide complex.^50,51 Free H₂NNNR compounds are extremely unstable with respect to loss of N₂. Although compound 8b is thermally stable for hours in solution at room temperature, it eventually decomposes to the Fe(II) amido complex L^tBuFeNHAd with loss of N₂. (For identification, L^tBuFeNHAd was synthesized independently from L^tBuFeCl and LiNHAd by analogy to other known iron(II) amido complexes.⁵²) Quantitative transformation from triazenido to amido was observed by ¹H NMR spectroscopy after heating a sample of 8b in C₆D₆ at 80 °C for 5 h.

Reactions With N=N Double Bonds

Diazene (HN=NH) is a possible intermediate of N₂ fixation by nitrogenase. Consistent with this idea, both NH=NH and CH₃N=NH are nitrogenase substrates,⁵³ and CH₃-N=N-CH₃ is reduced by nitrogenase to give ammonia, methane and methylamine.⁵⁴ Very recently, hydrazine (N₂H₄), methyldiazene (HN=N-CH₃) and N₂ derived nitrogenase intermediates have been freeze-trapped.⁵⁵ Free diazenes bearing hydrogen substituents are extremely difficult to handle, because they decompose in seconds or minutes. As an alternative approach, we have used the stable diazenes azobenzene (PhN=NPh) and benzo[c]cinnoline.

We recently reported a detailed study of the the reaction of 1b with PhN=NPh, which leads first to the [1,2]-addition product L^tBuFeNPhNHPh, and subsequently to the amido complex L^tBuFeNHPh.²⁷ Mechanistic studies were most consistent with a radical chain mechanism, mediated by an iron(I) carrier. ²⁷ Reaction of the smaller 1a with PhN=NPh at ambient temperature gives a mixture of products as judged by the ¹H NMR spectrum. It has not been possible to purify and isolate these compounds, but they may be tentatively assigned as L^MeFeNPhNHPh and L^MeFeNHPh based on the similarity of their ¹H NMR spectra with the L^tBu analogues.²⁷ Longer reaction time or heating does not drive the reaction mixture to L^MeFeNHPh. Instead decomposition occurs, probably due to the instability of L^MeFeNHPh.⁵²

We also investigated the reaction of 1a with 2 equiv of benzo[c]cinnoline, which gave 90% conversion to deep green 9a. The detection of 1.07(3) equivalents of H₂ by GC is consistent with the reaction stoichiometry shown in Scheme 5. In this case, there is no addition across the N=N double bond: rather, H₂ is lost and the benzo[c]cinnoline coordinates to iron. Alternatively, 9a could be isolated in 91% yield from the reaction of L^MeFeNNFeL^Me and benzo[c]cinnoline.

Scheme 5.

The crystal structure of 9a (Figure 9) shows that benzo[c]cinnoline binds face-on to iron(I) through the N=N π-bond, and the N-N distance is 1.138(5) Å. The dihedral angle between the benzo[c]cinnoline and the β-diketiminate ligand is 55.35(5)°. A side-on interaction between iron and an N=N double bond has been crystallographically confirmed in only two other compounds. In [Fe(NO)₂-{PPh₂CH₂CH₂PPh₂NNAr}][PF₆], the diazene (N=N 1.403(5)) is constrained to bind using a bidentate phosphine;⁵⁶ and our recently reported L^tBuFe(PhNNPh) has an N=N distance of 1.398(2) Å.²⁷

Figure 9.

Two views of the molecular structure of 9a, showing 50% thermal ellipsoids. In the picture on the right, aryl groups are removed for clarity. Selected bond distances (Å) and angles (°): Fe-N2 1.967(1), N2-N2′ 1.385(2), N2-Fe-N2′ 41.22(7). The dihedral angle between the β-diketiminate FeNCCNFe plane and the benzo[c]cinnoline plane is 55.35(5)°.

The peaks in the ¹H NMR spectra of C₆D₆ solutions of 9a are unusually broad. Mixing different ratios of 1a and benzo[c]cinnoline indicates that the broadness of the peaks increases with a greater concentration of benzo[c]cinnoline, and becomes substantially sharper at a 1:1 ratio of 1a to benzo[c]cinnoline (see Supporting Information for spectra). This behavior suggests that free and bound benzo[c]cinnoline exchange at a rate near the NMR time scale (milliseconds). The spectra are substantially sharper in C₆D₁₂, indicating that exchange of the aromatic with C₆D₆ is part of this process. We have previously characterized L^MeFe(C₆H₆), in which the benzene ligand is bound relatively weakly (it is displaced rapidly by phosphines, alkenes, and alkynes).⁵⁷ Since benzene competes effectively with benzo[c]cinnoline as a ligand, it shows that the heteroaromatic π-ligand is bound weakly as well.

Reactions With Carbon Dioxide and Carbodiimide

After azide, CO₂ is another heterocumulene that is a substrate of nitrogenase. Seefeldt and co-workers have shown that CO₂ is slowly reduced to CO by the nitrogenase enzyme.⁵⁸

Complexes 1a and 1b each react with two equivalents of carbon dioxide at room temperature, generating yellow solids that are formulated as formate-bridging diiron complexes (10a, 10b) based on X-ray crystallography (Figure 10). Only a few crystallographically characterized η²-formato bridging iron complexes have been reported in the literature. These complexes were synthesized from iron formate, Fe(O₂CH)₂·2H₂O,⁵⁹ reaction of iron metal and formic acid,⁶⁰ or Fe(ClO₄)·10H₂O with NaO₂CH⁶¹. Reactions of CO₂ with iron hydride complexes are uncommon. ⁶²

Figure 10.

Molecular structure of L^MeFe(μ-OCHO)₂FeL^Me (10a) showing 50% thermal ellipsoids. Isopropyl groups on the aryl rings are omitted for clarity. The structure of 10b is similar, and is shown in the Supporting Information. Selected bond distances (Å) and angles (°) for 10a: Fe-O 1.978(2), 2.006(2), C-O 1.245(3), 1.246(3), O-Fe-O 115.57(7), O-C-O 127.9(2). For 10b: Fe-O 1.967(1), 1.994(1), C-O 1.248(2), 1.245(2), O-Fe-O 106.98(6), O-C-O 128.2(2).

The formate-bridged diiron complexes are insoluble in pentane, and only somewhat soluble in aromatic solvents, a problem that was especially severe for 10b. The solution magnetic moment of 10a is 8.8(2) μ_B, consistent with two nearly uncoupled high-spin iron(II) ions, and 10b was too insoluble to derive a reliable value. The ¹H NMR spectrum of 10a at room temperature has six peaks with chemical shifts that range from 18 ppm to −60 ppm, suggesting idealized D_2h symmetry. Although this observation is consistent with either monomer or dimer in solution, no decoalescence is seen in the ¹H NMR spectrum of 10a in toluene from −35 °C to 120 °C, indicating that the crystallographically observed dimer is the most likely solution species. In 10b, the ¹H NMR spectrum at 85 °C has only eight peaks that integrate as expected for a fully symmetric diketiminate ligand, and the chemical shifts are similar to those in 10a. Interestingly, new peaks appear at lower temperatures. These peaks are consistent with splitting of a isopropyl methine and one isopropyl methyl resonance, which correspond to the protons closest to one another in the dimer. Unfortunately, the poor solubility of 10b, especially at low temperature, prevents us from fully characterizing this phenomenon. The available data are consistent with either a monomer-dimer equilibrium, or a change from slow to rapid flipping of the diketiminate planes to each side of a dimer (change from C_2h to effective D_2h point group symmetry).

To overcome the issues caused by the low solubility of these products, we investigated the analogous reaction of diisopropyl carbodiimide (ⁱPrN=C=NⁱPr) with 1b. This reaction proceeded rapidly and quantitatively at room temperature to give the formamidinate complex L^tBuFe(ⁱPrNCHNⁱPr) (11b) in 89% yield. This compound is monomeric, in contrast to the dimeric formate complexes derived from CO₂, a result that is most reasonably attributed to the steric hindrance of the isopropyl groups on the carbodiimide. In the solid state structure of 11b (Figure 11) both the diketiminate ligand and the formamidinate ligand are coordinated in an η² fashion to the Fe atom, giving a distorted tetrahedral geometry around the metal center. The N-Fe-N bite angle of each ligand is typical.³⁴

Figure 11.

Molecular structure of 11b, showing 50% thermal ellipsoids. Selected bond distances (Å) and angles (°): Fe-N1 2.1358(9), Fe-N2 2.0675(8), Fe-N11 2.0663(8), Fe-N21 1.9908(8), N1-C1 1.318 (1), N2-C1 1.323(1), N1-Fe-N2 64.50(3), N11-Fe-N21 97.44(3), N1-C1-N2 116.27(8).

Reactions With Brønsted Acids

As described above, EPR investigations suggest that an FeMoco species with two hydrides can lose two equivalents of H₂ to return to the hydride-free resting state.⁶ This suggests that some low-coordinate iron hydride species could be protonated by nearby Brønsted acids to give H₂. To investigate the susceptibility of synthetic Fe-H compounds to protonation of the hydride, we explored the reactions of 1a and 1b with weak acids.

In the presence of excess water, the low-coordinate iron hydride complexes decompose to intractable mixtures that contain free β-diketimine, suggesting that the acid protonates the α position of the β-diketiminate ligand. Reaction with smaller amounts of water often gives mixtures as well. In a few cases, using solutions of H₂O in tetrahydrofuran gives mixtures in which one compound predominates, enabling isolation. For example, we reported that addition of 1 equiv of H₂O to 1b (0.5 equiv per iron) gives the unique oxodiiron(II) complex L^tBuFeOFeL^tBu in 71% yield.^29b In another reaction, the addition of 2 equiv of water to 1a gives a green iron complex, L^MeFe(μ-OH)₂FeL^Me (12a) in 67% yield (Scheme 6)

Scheme 6.

The solid state structure of 12a (Figure 12) has pseudo-D_2h symmetry, and half of the molecule is related to the other half by a crystallographic inversion center. The metrical parameters compare well with other complexes containing a Fe₂(μ-OH)₂ core.^34,63 Seven paramagnetically shifted peaks are observed in the ¹H NMR spectrum of 12a, consistent with D_2h symmetry in solution. No peak clearly corresponding to the bridging OH is observed in the ¹H NMR spectrum, but O-H stretching vibrations are seen at 3709 and 3668 cm⁻¹ in the IR spectrum. This binuclear complex has a solution magnetic moment of 7.5(1) μ_B per dimer, suggesting two uncoupled high-spin Fe(II) centers. Dihydrogen is observed by GC as another product of the reaction. Unfortunately, the THF and H₂O peaks overlap with the peak from the internal standard CH₄, which prevented quantitative measurement of the H₂ produced.

Figure 12.

Molecular structure of 12a, showing 50% thermal ellipsoids. Isopropyl groups removed for clarity. Selected bond distances (Å) and angles (°): Fe-O1 2.082(3), 2.059(6), Fe-O1-Fe′ 98.7(2), O1-Fe-O1′ 81.3(2).

Finally, we tested 2,6-diisopropylaniline (dippNH₂), which has a similar acidity as water in organic solvents (the pK_a values in DMSO are 32 for H₂O and 30.6 for PhNH₂).⁶⁴ A mixture of 1a and dippNH₂ in C₆D₆ quantitatively gives the iron amido complex L^MeFeNHdipp within a few hours (Scheme 6), as judged by ¹H NMR spectroscopy. Dihydrogen (1.6 equiv) is detected from the reaction, close to the 2 equiv expected. This iron amido complex was previously synthesized from [L^MeFeCl]₂.⁵²

Discussion

Addition of Fe-H Across a Multiple Bond

The low coordination number at the iron atoms in these hydride complexes enables reactions with many unsaturated organic compounds. Rapid [1,2] additions are seen with alkenes, alkynes, imines, ketones, nitriles, carbodiimides, and carbon dioxide. These reactions generally place the iron fragment on the more electronegative atom of the double bond, and the hydrogen on the less electronegative atom. These are not uncommon reactions of transition-metal hydride complexes,¹ but they appear to be quite rapid in this system relative to literature iron complexes of higher coordination number. For example, a mixture of trans-[FeHCl(dppe)₂] and TlBF₄ reacts only with alkynes in which there is an ester functionality that can chelate to the metal.⁶⁵ The reactions of phosphine-supported octahedral iron hydrides with alkynes often give a variety of products, including acetylide and vinylidene complexes.⁶⁶ Known saturated iron hydride complexes do not react with alkenes⁶⁶ except cyclopropene.⁶⁷ Although there are not many literature examples of heterocumulenes (CO₂, CS₂, carbodiimides) reacting with iron hydrides, these reactions appear to be facile in one case.⁶⁸

Cyanide compounds (nitriles) are not usually reactive toward insertion into the Fe-H bonds of iron-hydride complexes, aside from one example of insertion from a dinuclear iron-carbonyl complex.⁶⁹ Acetonitrile more typically coordinates to iron.⁷⁰ In the β-diketiminate complexes described here, the iron hydrides reduce RCN to a RC(H)=N⁻ ligand under mild conditions. Interestingly, the outcome of the insertion reactions is dependent on the steric demands of the β-diketiminate ligand and the nitrile. With the bulkiest β-diketiminate (L^tBu) and nitrile (^tBuCN), the product is a monomeric complex. With the smallest β-diketiminate (L^Me) and nitrile (CH₃CN), the product is a dimer with two bridging imidoyl ligands. Finally, with ^tBuCN and the less bulky hydride complex, a single insertion occurs and the steric bulk prevents access of the second nitrile to the remaining bridging hydride.

The reaction of adamantyl azide with 1b gives a triazenido complex that results from formal [1,1]-addition. A few late-metal alkyl complexes have been observed to give [1,1]-addition to azides.⁷¹ The lone example of a triazenido ligand on iron is in a dinuclear iron-carbonyl complex, where the triazenide derives from protonation of an anionic azide complex.⁷² To our knowledge, the only other reaction of an iron hydride complex with an azide involves the addition of sodium azide to L₂FeH₂ (L = bidentate phosphine ligand) in methanol, which replaces the hydride ligand with azide.⁷³ A search of the CSD shows that L^tBuFe(HNNNAd) is the first structurally characterized complex of a triazenido ligand bearing a hydrogen atom.³⁴ Mechanistic and spectroscopic studies indicate the formation of hafnium, rhenium and tungsten triazenido complexes from the reaction of azides with hydride species.^51,74 In the hafnium complex, mechanistic studies indicate that the HNNN(aryl)⁻ ligand decomposes to NH(aryl)⁻ through a shift of the hydrogen atom to the carbon-bound nitrogen atom, followed by extrusion of N₂. In the iron system studied here, a similar mechanism can explain the conversion of the triazenido complex to L^tBuFeNHAd.

Recent reports of other low-coordinate iron systems are also indicative of high reactivity, although hydride species were not isolable in these systems. For example, Chirik has used pyridinediimine-iron and α-diimine-iron complexes as precatalysts for catalytic alkene hydrogenations, though hydride complexes were not isolated.⁷⁵ Peters has proposed a transient four-coordinate iron hydride complex that adds across an aromatic C=C bond of benzene,⁷⁶ and Ohki and Tatsumi have proposed the intermediacy of Cp*FeH in the cleavage of N=N bonds.⁷⁷ These precedents imply that the high activity of the isolable low-coordinate iron-hydride complexes described here is likely to have broader applicability with other supporting ligands.

Reductive Elimination of Dihydrogen

Numerous polyhydride complexes, both monometallic and multimetallic, are capable of formal reduction through loss of H₂ when a ligand is added.⁷⁸ Some synthetic hydride complexes have been observed to lose H₂ with addition of N₂.^79,80 Specifically with iron, a few polyhydride complexes reductively eliminate H₂ with ligand binding.^25,81 Iron-dihydrogen complexes have been studied in octahedral systems.⁸² The reactions typically involve reduction of iron(II) to iron(0), not to iron(I) as observed here. In a recent paper, Peters has presented evidence supporting reversible metallation of a ligand alkyl group in a low-coordinate iron(I) complex.^23b

We see no evidence of stable H₂ complexes in thermal or photochemical elimination reactions of the diketiminate-supported iron hydride complexes, because any putative transient “Lfe” or “LFe(H₂)” is trapped by N₂ (in the photochemical reaction) or the added ligand (thermal reactions with CO or CNR). The thermal reactions must proceed through a mechanism in which substrate binding precedes H₂ loss, because 1a and 1b are stable at room temperature in solution and under vacuum. We infer that interaction with ligands activates the iron hydride species toward H₂ loss. As described above in the context of alkene/alkyne insertion, the evidence described above suggests that [L^MeFeH]₂ reacts through a low-population isomer in which one iron-hydrogen bond has opened in the Fe₂H₂ core. The subsequent L^MeFe(substrate)(μ-H)Fe(H)L^Me complex could be unstable with respect to breaking the remaining Fe-H bond to give an iron(I) product and an unstable iron(III) dihydride complex L^MeFeH₂, which could lose H₂ leading to the remaining iron(I) product. Though other mechanisms are consistent with the data, this is a working model for further studies of these reactions; it also rationalizes why the stronger ligands and those more resistant to insertion reactions (isocyanides, benzo[c]cinnoline, some azides) lead to iron(I) products. In the case of Me₃SiN₃, the product is not the iron(I) compound itself, but instead the result of reductive coupling of trimethylsilyl groups in Me₃SiN₃. Because the reductive coupling product also results from treatment of Me₃SiN₃ with the iron(I) source LFeNNFeL, iron(I) species are likely intermediates in this reaction.

The product of the reaction of 1a with benzo[c]cinnoline is especially interesting. Despite the presence of lone pairs on the two nitrogen atoms of this heterocycle, the iron coordinates instead to the π-system. There are only four previous crystal structures in which benzo[c]cinnoline is a ligand to a transition metal: Ti and Yb complexes with the metal in the plane of the aromatic rings,⁸³ a complex with the diazene bridging the iron atoms in a Fe₂(CO)₆ fragment,⁸⁴ and the cobalt(0) complex (PMe₃)₃Co(benzo[c]cinnoline).⁸⁵ The latter complex is the most similar to the one shown here, from its formal d⁷ electronic configuration to the similar angle between the MN₂ and aromatic planes (68° vs. 55° here).⁸⁵ We have previously shown that benzene coordinates in a η⁶ binding mode in L^MeFe(C₆H₆),⁴⁷ but the η² binding of benzo[c]cinnoline is clearly stronger, since the benzocinnoline complex is stable in benzene solution. Why, then, is the binding stronger for the more hindered benzo[c]cinnoline? Previous studies on the binding constants of alkenes, alkynes, and other ligands to the iron(I) L^MeFe fragment showed that backbonding dominates the selectivity of binding.⁵⁷ Because of the greater electronegativity of nitrogen than carbon, the stabilizing shift of electron density from the iron to the unsaturated ligand is more effective in the iron(I) complex of benzo[c]cinnoline.

We also note that strongly hydridic M-H bonds like those in 1a and 1b are characteristic of early transition metal hydrides,⁸⁶ consistent with the Fe^δ+–X^δ− character of iron alkyl, amido, and alkoxo complexes demonstrated in our earlier studies,^30,52,57 Further research will show whether the reactivity of other low-coordinate late-metal systems will also tend to mimic the behavior of transition metals with a lower d electron count.

Protonation to Give Dihydrogen

Many iron-hydride complexes are protonated by acids to release H₂. For example, in iron complexes of a PNP ligand, the necessary pK_a of the acid was determined to be less than 10.5 (a coordinated nitrogen was protonated in preference to the hydride ligand).^87,88 On the other hand, the diketiminate-bound iron hydrides are protonated by weak acids such as diisopropylaniline and diphenylhydrazine.²⁷ Using literature values in DMSO,⁶⁴ one can conclude that acids with pK_a value up to 32 (and perhaps higher) can protonate 1a and 1b.

Protonation of a hydride ligand may be a key step in nitrogen fixation by nitrogenases. In this hypothesis, one of the metal-bound hydride ligands is protonated and released as dihydrogen to open a coordination site for dinitrogen binding.⁸⁹ This hypothesis is consistent with Hoffman’s ENDOR-based conclusion that the N₂ binding form of the iron-molybdenum cofactor is at the E₄H₂ state.⁶ There are a number of acidic Arg residues near the binding face of the iron-molybdenum cofactor.³ This hypothesis finds support here, in that weak acids are capable of protonating the hydride ligand to form H₂.

Insight into Potential Nitrogenase Mechanisms

Some nitrogenase researchers have explained the production of H₂ by nitrogenase during N₂ reduction as the result of ligand-assisted reductive elimination.³ In this model, creation of hydrides provides a way for the FeMoco to store electrons for N₂ reduction without reaching an unrealistically low charge and reduction potential. Release of H₂ from the FeMoco is thought to be concurrent with N₂ binding, explaining why H₂ is a competitive inhibitor of N₂ reduction.⁷ Other substrates are not inhibited by H₂ because the extra “boost” in reducing power from H₂ loss is not necessary. The reactivity of complexes 1a and 1b can be viewed using this model, in that they have sufficient reducing power to react thermally with all nitrogenase substrates except N₂. Thus, the reactivity of compounds 1 mimics not the E₄ state, but a less reduced FeMoco intermediate (E₂ or E₃) that reduces alternative substrates like alkynes, CO₂, cyanide, and azide. Photolysis of the hydride complexes 1a and 1b provides the “boost” required to bind the weak ligand N₂.

Even though 1a and 1b do not reduce N₂, their reducing power is superior to octahedral Fe-H complexes, a difference that might arise from the low coordination number, the high-spin electronic configuration, the weak Fe-H bond strength, or (most likely) a combination of these effects. Considering that iron atoms in the FeMoco have a low-coordinate geometry and high-spin electronic configuration, these factors are clearly identified as reasonable targets for evaluation in the enzyme.

It is notable that nitrogenase inhibitors (or analogues thereof) like CO, RNC, and RN₃ typically cause the synthetic hydride complexes 1a and 1b to lose H₂ and bind the small molecule that was added. Therefore, one may hypothesize that these small molecules might inhibit enzymatic N₂ reduction by binding strongly to the FeMoco, causing premature loss of H₂. Addition of 1a and 1b to small molecules that are substrates of nitrogenase (or analogues thereof), like RCN, alkynes, RN₃, and RN=NR, typically forms a new substrate-H bond, and no H₂ is lost. An exception is benzo[c]cinnoline: although it mimics the nitrogenase substrate diazene, it displaces H₂ like an inhibitor. This may be attributed to the aromatic nature of benzo[c]cinnoline,⁹⁰ which perhaps makes it more resistant to reduction than other diazenes. Azides and cyanides are inhibitors and substrates in the enzyme, and in the synthetic compounds they sometimes give H₂ and sometimes are reduced. Therefore, the reactions of the synthetic compounds 1a and 1b recall the bifurcated roles of these molecules toward the enzyme.

It is also worth comparing the specific products of substrate reduction by nitrogenase and the products from the synthetic iron complexes. The synthetic compounds typically do not complete the reduction and release the appropriate product, but maintain a bond between the iron atom and the partially reduced substrate. For example, 1a and 1b reduce alkynes to iron-vinyl complexes, but do not release alkene. They reduce one C=O bond of CO₂, but do not eliminate water to give CO.⁹¹ They reduce C≡N bonds to C=N bonds, but do not continue the reduction to methylamine or methane/ammonia. These outcomes may be attributable to the absence of protons in the synthetic system that would release the anionic ligand, either as a product or as an intermediate that is subsequently reduced to the final product. Future studies will examine the further transformations of these small molecules at iron.

Conclusions

In summary, we have shown that the first isolable low-coordinate iron hydride complexes are highly reactive towards various unsaturated small molecules. The reactions can be categorized into three types: (1) insertions of a multiple bond of the substrate into the Fe-H bond with two-electron reduction of the substrate; (2) reductive elimination of H₂ and coordination (or one-electron reduction) of the substrate; (3) loss of H⁻ to an acidic proton of the substrate, releasing H₂ and giving an iron complex of the conjugate base. Some of the products isolated have novel structural features, such as a square-planar iron(I) complex, a face-bound complex of benzo[c]cinnoline, and a monoalkylated triazenido complex. The reaction patterns are reminiscent of elementary steps in the proposed nitrogenase catalytic mechanism, and support the idea that hydrides on the low-coordinate, high-spin iron atoms of the FeMoco would be capable of some of the characteristic reactions of the FeMoco. In addition, these studies identify some possible reasons why certain small molecules are substrates of nitrogenases, and others are inhibitors.

Experimental Section

General Procedures

All manipulations were performed under a nitrogen atmosphere by Schlenk techniques or in an M. Braun glove box maintained at or below 1 ppm of O₂ and H₂O. Glassware was dried at 150 °C overnight. NMR data were recorded on a Bruker Avance 500 spectrometer (500 MHz). All peaks in the NMR spectra are referenced to residual protiated solvents (benzene δ 7.16 ppm; toluene δ 2.08 ppm; cyclohexane δ 1.38 ppm). In parentheses are listed integrations and assignments. Resonances were singlets unless otherwise specified. Infrared spectra (450–4000 cm⁻¹) were recorded on KBr pellet samples in a Shimadzu FTIR spectrophotometer (FTIR-8400S) using 32 scans at 2 cm⁻¹ resolution. UV-vis spectra were measured on a Cary 50 spectrophotometer using screw-cap cuvettes; descriptions of the spectra include extinction coefficients in parentheses. Solution magnetic susceptibilities were determined by the Evans method.⁹² Elemental analyses were determined by Desert Analytics, Tucson, AZ.

Pentane, tetrahydrofuran (THF), diethyl ether and toluene were purified by passage through activated alumina and “deoxygenizer” columns from Glass Contour Co. (Laguna Beach, CA). Deuterated solvents were first dried over CaH₂, then over Na/benzophenone, and then vacuum transferred into a storage container. Before use, an aliquot of each solvent was tested with a drop of sodium benzophenone ketyl in THF solution. Celite was dried overnight at 200 °C under vacuum. CO₂ was purchased from Air Products, dried by passing through 4Å molecular sieves, and freed from O₂ with three freeze-pump-thaw cycles. H₂O, ^tBuNC, ^tBuCN, MeCN and 2,6-diisopropylphenylamine were degassed before using. Diphenylacetylene, benno[c]cinnoline, and 1-azidoadamantane (AdN₃) were crystallized from pentane in the glovebox. Compounds 1a and 1b were prepared by published procedures.^20,21

L^MeFe(CN^tBu)₂ (2a)

Reaction of 1a and ^tBuNC in C₆D₆ gives complex 2a, in a crude yield of 62% based on ¹H NMR integration with an internal integration standard. However, it can be more conveniently synthesized from L^MeFeNNFeL^Me and ^tBuNC as follows. A solution of L^MeFeNNFeL^Me (205 mg, 0.21 mmol) in pentane (15 mL) was treated with ^tBuNC (105 μL, 0.93 mmol) with stirring. Immediate bubbling was observed and the solution color changed from deep brown to bright orange. The reaction solution was stirred at ambient temperature for 5 h. Volatile materials were removed under vacuum and the residue was extracted with pentane (20 mL) and filtered through Celite. The solution was concentrated to 3 mL and cooled to −35 °C to give bright orange crystals of 2a (145 mg, 54%). ¹H NMR (500 MHz, C₆D₆): δ9.1 (2H, aryl p-H), 5.8 (4H), 3.5 (12H, ⁱ Pr-CH₃), 2.5 (18H, ^tBu), 1.3 (4H), −2.0 (6H, backbone CH₃), −5.8 (12H, ⁱPr CH₃) ppm. (Peaks integrated as 4H could be aryl m-H or ⁱPr methine CH; the backbone CH was not observed). UV-vis (pentane): 317 (25.3 mM⁻¹cm⁻¹), 447 (3.9 mM⁻¹cm⁻¹), 808 (0.2 mM⁻¹cm⁻¹) nm. IR (KBr pellet): 2122 (vs), 2050 (m), 1969 (m), 1948 (m) cm⁻¹. The X-band EPR spectrum is shown in the Supporting Information. Attempts to measure the solution magnetic moment of 2a were not successful because the internal standard peak was obscured. Complex 2a is thermally unstable (decomposition of a solid sample occurred overnight inside the glove box), and all characterizations were done with fresh samples. The instability prevented elemental analysis measurements.

L^MeFe(μ-N=CHMe)₂FeL^Me (3a)

A stirring solution of [L^MeFeH]₂ (1a) (146 mg, 0.15 mmol) in toluene (30 mL) was treated with MeCN (18.0 μL, 0.34 mmol). The reaction solution was heated at 45 °C for 18 h and the color changed from deep brown to yellow. All volatile materials were removed under vacuum and the residue was extracted with toluene (30 mL), filtered and concentrated to 5 mL. Cooling to −35 ºC gave orange-yellow crystals of 3a (155 mg, 88%). ¹H NMR (500 MHz, C₇D₈): δ50 (6H, N=CHCH₃), 20 (12H), 13 (4H), 12(4H), −2 (12H), −7 (12H), −12 (4H), −16 (4H), −21 (12H), −22 (4H), −31(2H, diketiminate backbone), −40 (12H) ppm. (No peak corresponding to the imine proton was identified. Peaks integrated as 4H could be ⁱPr methine, aryl p-H or aryl o-H; peaks integrated as 12H could be ⁱPr CH₃ or backbone CH₃.) UV-vis (toluene): 337 (34.2 mM⁻¹cm⁻¹), 392 (4.8 mM⁻¹cm⁻¹) nm. IR (KBr pellet): 1637 cm⁻¹ (C=N). μ_eff (1% Me₃SiOSiMe₃ in THF-d₈, 25 °C): 5.9(1) μ_B per dimer. Elem. Anal. Calcd. for C₆₂H₉₀N₆Fe₂: C, 72.22; H, 8.80; N, 8.15. Found: C, 72.04; H, 8.98; N, 7.72.

L^MeFe(η¹-N=CH^tBu)(μ-H)FeL^Me (4a)

A solution of 1a (100 mg, 0.11 mmol) in pentane (10 mL) was treated with ^tBuCN (12.0 μL, 0.11 mmol) and stirred at room temperature for 2 h. Volatile materials were removed under vacuum and the residue was extracted with toluene (30 mL), filtered and concentrated to 5 mL. Crystallization at −35 ºC gave 4a as a brown powder (97 mg, 85%). ¹H NMR (500 MHz, C₆D₆): The narrow range of the NMR spectrum, the broad peaks, and the low symmetry of the molecule lead to severe overlap between peaks, which made assignment unsuccessful. The spectrum is shown in the Supporting Information. UV-vis (pentane): 332 (30 mM⁻¹cm⁻¹). IR (KBr pellet): 1629 cm⁻¹ (C=N). Elem. Anal. Calcd for C₆₃H₉₃N₅Fe₂: C, 73.31; H, 9.08; N, 6.79. Found: C, 73.50; H, 9.40; N, 6.46.

L^tBuFe(NCH^tBu) (5b)

A solution of 1b (91 mg, 81 μmol) in diethyl ether (8 mL) was treated with ^tBuCN (18 μL, 163 μmol) and stirred at room temperature for 30 min. The volatile materials were removed under vacuum. The orange residue was extracted with toluene, filtered through Celite, concentrated, and stored at −35 °C to afford 5b as orange-red crystals (70 mg, 67%). ¹H NMR (500 MHz, C₆D₆): δ 95 (9H, ^tBu of imide), 92 (1H, backbone C-H), 36 (18H, ^tBu of diketiminate), 3 (4H, mAr), −24 (12H, ⁱPr-CH₃) −99 (4H, ⁱPr-CH), −103 (12H, ⁱPr-CH₃), −108 (2H, p-Ar) ppm. There was no clear peak for NCH^tBu. UV-vis (toluene): 335 (25 mM⁻¹cm⁻¹), 397 (6.7 mM⁻¹cm⁻¹), 454 (sh), 515 (sh) nm. μ_eff (C₆D₆, 298 K): 4.6 μ_B. IR (KBr pellet): 3058(w), 2960(vs), 2906(w), 2867(m), 2717(w), 2631(w), 1687(s), 1506(s), 1462(w), 1433(w), 1383(vs), 1362(vs), 1317(s), 1255(w), 1217(w), 1182(w), 1151(w), 1095(m), 1024(w) cm⁻¹. Anal. Cald. for C₄₀H₆₃FeN₃: C, 74.86; H, 9.89; N, 6.55. Found: C, 74.52; H, 9.80; N, 6.80.

L^MeFe-hexyl

In a J. Young NMR tube, 1a (5.6 mg, 5.9 μmol) and 1-hexene (1.5 μL, 12.1 μmol) were dissolved in 0.5 mL of C₆D₆. After 1 h, a single paramagnetic species was observed by ¹H NMR spectroscopy: δ116 (2H), 77 (1H, backbone CH), 51 (2H), 33 (6H, diketiminate CH₃), 32 (3H, hexyl CH₃), 15 (4H), −19 (12H, ⁱPr-CH₃), −21 (4H), −30 (2H), −44 (2H), −75 (12H, ⁱPr-CH₃), −91 (2H) ppm. Peaks integrated as 4H could be aryl m-H or ⁱPr methine CH; peaks integrated as 2H could be aryl p-H, hexyl β-CH_2, γ-CH_2, δCH_2, or ε-CH2. The hexyl α–CH₂ protons were not observed in the ¹H NMR spectrum. This ¹H NMR spectrum is very similar to the spectra of other L^MeFe(alkyl) species we have previously reported.³⁰

L^MeFe(CPh=CHPh) (6a)

A mixture of 1a (88 mg, 93 μmol) and PhCCPh (33 mg, 187 μmol) in ppentane (15 mL) was stirred at room temperature for 17 h. Volatile materials were removed under vacuum. The product was extracted with pentane (20 mL), filtered through Celite, and concentrated to 2 mL. Cooling to −35 °C gave 6a as a yellow powder (87 mg, 75%). ¹H NMR (500 MHz, C₆D₆): δ123 (1H), 60 (1H), 59 (2H), 57 (6H, diketiminate CH₃), 51 (1H), 48 (2H), 33 (2H), −9 (4H), −19 (12H, ⁱPr CH₃), −78 (2H), −118 (12H + 4H, ⁱPr CH₃ + another peak, possible assignments are listed below) ppm. Peaks integrated as 4H could be aryl m-H or ⁱPr methine CH; peaks integrated as 2H could be aryl p-H, vinyl phenyl m-H (2 peaks for the two phenyl groups), or phenyl o-H; peaks integrated as 1H could be backbone CH, vinyl phenyl p-H (2 peaks for the two phenyl groups). The vinyl C=CH and o-H from the vinyl phenyl closer to the iron were not observed. μ_eff (C₆D₆, 298 K) = 5.15(2) μ_B. UV-vis (pentane): 320 (37.9 mM⁻¹cm⁻¹), 495 (1.5 mM⁻¹cm⁻¹) nm.

L^tBuFe(μ-N₃)₂FeL^tBu (7b)

A solution of 1b (71 mg, 64 μmol) in diethyl ether (10 mL) was treated with Me₃SiN₃ (18 μL, 128 μmol) and stirred at room temperature for 2 h. An orange precipitate formed within 30 minutes. The volatile materials were removed under vacuum. The orange residue was extracted with hot toluene and filtered through Celite while hot to give an orange solution. This solution was concentrated and cooled to −35 °C to afford 7b as orange crystals (70 mg, 79%). 7b is insoluble in benzene, and slightly soluble in toluene, at room temperature. ¹H NMR (500 MHz, C₇D₈, 75 °C): δ15, 1, −3, −24, −46 ppm. The broadness of the peaks prevented accurate integration and assignment of the peaks. μ_eff (C₇D₈, 348 K): 4.1 μ_B. IR (KBr pellet): 2962 (s), 2929 (w), 2869 (m), 2129 (vs), 2081 (w), 1529 (w), 1490 (m), 1463 (w), 1436 (w), 1380 (m), 1357 (s), 1317 (s), 1259 (m), 1215(w), 1190 (w), 1097 (m), 1054 (w), 1024 (m) cm⁻¹. UV-vis (toluene): 337 (15.5 mM⁻¹cm⁻¹) nm. Anal. Cald. for C₇₀H₁₀₆N₁₀Fe₂ C 70.10, H 8.91, N 11.68. Found C, 70.04, H 8.71, N 11.02.

L^tBuFe(HNNNAd) (8b)

A solution of 1b (88 mg, 79 μmol) in diethyl ether (8 mL) was treated with a solution of AdN₃ (28 mg, 157 μmol) in diethyl ether (5 mL) and stirred at room temperature for 30 min to give an orange-red solution. The volatile materials were removed under vacuum. The residue was extracted with toluene and filtered through Celite. This solution was concentrated and cooled to −35 °C to afford L^tBuFe(HNNNAd) as brown crystals (77 mg, 67%). ¹H NMR (500 MHz, C₆D₆): δ 39 (3H, Ad-α, β, or γ), 21 (18H, ^tBu), 15 (4H, m-Ar), 10 (3H, Ad- α, β, or γ), 8 (3H, Ad-α, β, or γ), 3.7 (3H, Ad-α, β, or γ), −7 (12H, ⁱPr-CH₃), −27 (1H, NH), −40 (12H, ⁱPr-CH₃), −51 (3H, Ad-α, β, or γ), −61 (2H, p-H) ppm. UV-vis (toluene): 340 (18 mM⁻¹cm⁻¹), 415 (sh), 530 (sh) nm. μ_eff (C₆D₆, 298 K) 4.6 μ_B. IR (KBr pellet): 3371(w), 3059(w), 2960(vs), 2926(s), 2905(vs), 2850(m), 1622(w), 1585(w), 1526(w), 1491(s), 1464(m), 1429(w), 1380(vs), 1356(vs), 1311(m), 1263(w), 1215(w), 1182(w), 1153(w), 1099(m), 1022(m) cm⁻¹. Elemental analysis could not be obtained for this compound as 8b is thermally unstable.

Thermal conversion of 8b to L^tBuFeNHAd

A sample of 8b (~5 mg) in C₆D₆ was heated to 80 °C for 5 h. The ¹H NMR spectrum was identical to a sample prepared independently through the following method. A mixture of L^tBuFeCl (257 mg, 0.433 mmol), LiNHAd (68 mg, 0.43 mmol), and diethyl ether (10 mL) was stirred at room temperature for 2 h. The yellow-orange solution was filtered through Celite, and the volatile materials were removed under vacuum. The orange residue was dissolved in n-hexane (3 mL), and cooling the solution to −45 ºC afforded L^tBuFeNHAd as bright orange crystals in 2 crops (230 mg, 75%). ¹H NMR (500 MHz, C₆D₆): d 95 (6H, Ad-α), 67 (3H, Ad-β or γ), 42 (3H, Ad-β or γ), 37 (18H, ^tBu), 31 (3H, Ad-β or γ), −1 (4H, m-Ar), −24 (12H, ⁱPr-CH₃), −93 (4H, ⁱPr-CH), −102 (2H, p-Ar), −118 (12H, ⁱPr-CH₃) ppm. The N-H and backbone C-H protons were not located. μ_eff (C₆D₆, 298 K): 4.6(3) μ_B. IR (KBr pellet): 3277 (vw, ν_N–H), 3052 (w), 3013 (w), 2959 (vs), 2925 (s), 2903 (vs), 2845 (m), 1534 (m), 1502 (s), 1459 (m), 1443 (m), 1430 (m), 1385 (vs), 1364 (vs), 1319 (s), 1303 (m), 1253 (w), 1218 (m), 1197 (m), 1131 (m), 1095 (s), 1055 (m), 1029 (m), 948 (w), 933 (w), 890 (w) cm⁻¹. UV-Vis (pentane): 340 (14 mM⁻¹cm⁻¹), 510 (sh, ~0.4 mM⁻¹cm⁻¹) nm.

L^MeFe(benzo[c]cinnoline) (9a)

Although reaction of 1a and benzo[c]cinnoline gives complex 9a (yield 90% by ¹H NMR spectroscopy), it is more easily synthesized from L^MeFeNNFeL^Me and benzo[c]cinnoline. A vial was loaded with L^MeFeNNFeL^Me (94 mg, 96 μmol) and benzo[c]cinnoline (35 mg, 190 μmol). Pentane (15 mL) was added, causing an immediate color change from brown to green. The reaction solution was stirred at room temperature for 5 h, filtered through Celite, concentrated to 3 mL, and cooled to −35 °C to give dark green plates of 9a (113 mg, 91%). ¹H NMR (500 MHz, C₆D₁₂): δ140 (4H), 111 (4H), 17 (12H, ⁱPr-CH₃), −17 (6H, backbone CH₃), −21 (4H), −52 (12H, ⁱPr-CH₃), −78 (2H, p-Ar) ppm (peaks integrated as 4H could be m-Ar, ⁱPr CH, or overlapped peaks for benzo[c]cinnoline ligand). UV-vis: 305 (16.3 mM⁻¹cm⁻¹), 325 (13.1 mM⁻¹cm⁻¹), 388 (9.7 mM⁻¹cm⁻¹), 419 (7.0 mM⁻¹cm⁻¹), 585 (2.3 mM⁻¹cm⁻¹) nm. μ_eff (C₆D₆, 25 °C): 3.1(1) μ_B. Elem. Anal. Calcd for C₄₁H₄₉N₄Fe: C, 75.33; H, 7.55; N, 8.57. Found: C, 73.31; H, 7.68; N, 7.98. Repeated attempts at elemental analysis did not give better agreement, indicating a possible small impurity not visible by NMR spectroscopy.

L^MeFe(μ-OCHO)₂FeL^Me (10a)

A resealable flask was loaded with 1a (254 mg, 0.27 mmol) and pentane (15 mL). CO₂ (256 mbar, 61.7 mL, 0.64 mmol) was condensed in the reaction flask at 77 K over 30 min. The flask was warmed to room temperature and stirred for 22 h. Volatile materials were removed from the yellow mixture under vacuum and the residue was extracted with toluene (30 mL), filtered and concentrated to 5 mL. Cooling to −35 ºC gave yellow blocks of 10a (138 mg, 50%). ¹H NMR (500 MHz, C₆D₆): δ18 (4H), 5 (12H, ⁱPr CH₃), −11 (12H + 4H, ⁱPr CH₃ + another 4H; possible assignments for this 4H peak are listed below), −19 (1H, backbone CH), −42 (2H, aryl p-H), −60 (6H, backbone CH₃) ppm. (Peaks integrated as 4H could be aryl m-H or ⁱPr methine CH; the two bridging formate H were not observed in the NMR spectrum). UV-vis (pentane): 332 (e = 29.9 mM⁻¹cm⁻¹) nm. IR: 1628 cm⁻¹(C-O). μ_eff (C₆D₆, 25 °C): 8.8(1) μ_B per dimer. Elem. Anal. Calcd for C₆₀H₈₄N₄O₄Fe₂: C, 69.49; H, 8.16; N, 5.40. Found: C, 73.52; H, 9.52; N, 5.62. Repeated attempts at elemental analysis did not give better agreement, indicating a possible small impurity not visible by NMR spectroscopy.

L^tBuFe(μ-OCHO)₂FeL^tBu (10b)

This was identical to the synthesis of 10a, but using 1b (17 mg, 15 μmol), diethyl ether (15 mL), and CO₂ (99 mbar; 7.73 mL, 30 μmol). The yield of 10b was 15 mg (82%). ¹H NMR (500 MHz, C₇D₈, 85 °C) δ 30 (1, backbone C-H), 19 (18H, ^tBu), 16 (4H, ⁱPr-CH), −8 (12H, ⁱPr-CH₃), −42 (12H, ⁱPr-CH₃), −50 (2H, p-H or O₂CH), −70 (2H, p-H or O₂CH) ppm. A 4H peak for the m-Ar protons was not observed, and is possibly obscured by the solvent. UV-vis (toluene): 339 (43 mM⁻¹cm⁻¹), 393 (6.0 mM⁻¹cm⁻¹), 516 (sh) nm. IR (KBr pellet): 3059(w), 2962(vs), 2929(s), 2869(m), 1626(vs), 1529(w), 1493(s), 1462(m), 1437(m), 1381(vs), 1360(vs), 1317(s), 1259(m), 1215(w), 1190(w), 1157(w), 1099(s), 1024(s) cm⁻¹. Elem. Anal. Calcd. for C₇₂H₁₀₈Fe₂N₄O₄: C, 71.75; H, 9.03; N, 4.65. Found: 72.79; H, 8.63; N, 4.68.

L^tBuFe(ⁱPrNCHNⁱPr) (11b)

A solution of 1b (57 mg, 51 μmol) in diethyl ether (8 mL) was treated with N,N′-diisopropylcarbodiimide (15.8 μL, 102 μmol) and stirred for 1 h. The reaction mixture turned orange within a couple of minutes. The volatile materials were removed under vacuum. The resultant orange solid was extracted with toluene, filtered through Celite, concentrated, and stored at −35 °C to afford orange-red crystals of 11b (62 mg, 89%). ¹H NMR (500 MHz, C₆D₆): δ 142 (2H, p-Ar), 120 (1H, backbone C-H of diketiminate), 19 (4H, m-Ar or ⁱPr-CH), 10 (18H, ^tBu), 5 (12H, ⁱPr-CH₃), 2.9 (12H, ⁱPr-CH₃), 2.0 (4H, m-H or ⁱPr-CH), −9 (12H, ⁱPr-CH₃), -36 (2H, ⁱPr-CH of amidinate), −85 (1H, backbone C-H of amidinate) ppm. UV-vis (toluene): 349 (14 mM⁻¹cm⁻¹), 405 (10.2 mM⁻¹cm⁻¹), 475 (sh) nm. μ_eff (C₆D₆, 298 K) 4.2(3) μ_B. IR (KBr pellet): 3057(w), 2962(vs), 2929(w), 2867(m), 1657(w), 1547(s), 1479(m), 1460(m), 1431(w), 1382(vs), 1360(s), 1317(m), 1257(vs), 1191(w), 1097(m), 1020(s) cm⁻¹. Elem. Anal. Cald. for C₄₂H₆₈FeN₄: C, 73.66; H, 10.01; N, 8.18. Found: C, 73.80; H, 10.42; N, 8.10.

[L^MeFe(μ-OH)]₂ (12a)

To a solution of 1a (189 mg, 0.20 mmol) in THF (10 mL), a H₂O solution of THF (0.28 mM, 1.45 mL, 0.41 mmol) was added dropwise with stirring. Bubbling was observed and the solution turned to green. The solution was stirred at room temperature for 20 h. Volatile materials were removed under vacuum with heating at 120 °C. The residue was extracted with pentane (20 mL), filtered, and concentrated to 5 mL. Cooling to −35 °C gave a green powder, which was further dried under vacuum with heating at 120 °C for 10 h to give dry 12a (131 mg, 67%). ¹H NMR (500 MHz, C₆D₆): δ9 (4H), 4 (12 H, ⁱPr-CH₃), −5 (1H, backbone CH), −6 (6H, diketiminate CH₃), −8 (4H), −33 (2H, aryl p-H), −38 (12H, ⁱPr-CH₃) ppm (peaks integrated as 4H could be ⁱPr CH or aryl o-H; the bridging OH protons are not observed.). UV-vis (pentane): 326 (27.1 mM⁻¹cm⁻¹) nm. IR (KBr pellet): 3709, 3668 cm⁻¹ (O-H). μ_eff (C₆D₆, 25 °C): 7.5(1) μ_B per dimer. Elem. Anal. Calcd for C₅₈H₈₄N₄O₂Fe₂: C, 71.01; H, 8.63; N, 5.71. Found: C, 71.11; H, 8.72; N, 5.53.

Detection of H₂ Using Gas Chromatography

²⁸ The reaction solution from a given reaction (3 mL) and a magnetic stir bar were placed in a 25 mL round bottom flask. The flask was capped by an adaptor with a stopcock leading to a rubber septum. Using a syringe, 8 mL of the gas inside was removed and 8 mL CH₄ (1043 mbar) was injected into the flask as an internal standard. An aliquot (20 μL) of the gas was withdrawn and injected into a GC (Shimadzu GC-17A) with a 5 Å molecular sieve column (30 m × 0.25 mm) at 26 °C, carrier gas N₂, 600 kPa. The ratio of integrated H₂:CH₄ responses were compared to a calibration plot previously determined²⁸ by injecting known amounts of H₂ into the same flask with 3 mL of toluene.

Mössbauer Spectroscopy

Mössbauer data were recorded on a spectrometer with alternating constant acceleration. The minimum experimental line width was 0.24 mm/s (full width at half-height). The sample temperature was maintained constant either in an Oxford Instruments Variox or an Oxford Instruments Mössbauer-Spectromag cryostat. The latter is a split-pair super-conducting magnet system for applied fields up to 8 T where the temperature of the sample can be varied in the range 1.5 K to 250 K. The field at the sample is perpendicular to the -beam. The ⁵⁷Co/Rh source (1.8 GBq) was positioned at room temperature inside the gap of the magnet system at a zero-field position. Isomer shifts are quoted relative to iron metal at 298 K. Magnetic Mössbauer spectra for the paramagnetic contamination of 1a were simulated by using a spin-Hamiltonian description of the electronic ground state:

$$H_{\text{e}}=D[{S_{\text{t},\text{z}}}^2-S_{\text{t}}(S_{\text{t}}+1)/3+(E/D)({S_{\text{t},\text{x}}}^2-{S_{\text{t},\text{y}}}^2)]+{\mu }_{\text{B}}B·S_{\text{t}}$$ (1)

where S_t is the total spin of the system, and D and E/D are the axial and rhombic zero-field parameters. The hyperfine interactions for ⁵⁷Fe were calculated by using the usual nuclear Hamiltonian.⁹³ For 1a only the nuclear Hamiltonian was used.

Computational Methods

The Gaussian 03 package⁹⁴ was used for all calculations described herein. Hybrid quantum mechanics/molecular mechanics (QM/MM) calculations were used to study full experimental models of L^MeFe(CN^tBu)_2. The QM/MM calculations utilized the ONIOM⁹⁵ methodology. The MM region was modeled with the Universal Force Field (UFF)⁹⁶ and included the Ar and Me substituents of L^Me,Ar and the methyl groups of ^tBuNC. The remainder of L^MeFe(CN^tBu)₂ was modeled at the B3LYP/6-311+G(d) level of theory. Geometries were fully optimized using gradient methods unless otherwise noted. The unrestricted Kohn-Sham formalism was used for the description of all open-shell species. The energy Hessian was calculated for all stationary points and thus confirmed the calculated stationary points as minima (no imaginary frequencies). All reported enthalpies are calculated at 1 atm and 298.15 K.

Figure 3.

Molecular structure of L^MeFe(CN^tBu)₂ (2a), showing 50% thermal ellipsoids. There is 50:50 disorder in the position of one isocyanide ligand. Selected bond distances (Å) and angles (°): Fe-C14 1.817(1), Fe-C15 1.821(1), C14-N14 1.222(9), C14-N14′ 1.153(9), C15-N15 1.173(2), C14-N14-C24 148.3(7), C14-N14′-C24′ 164.0(7), C15-N15-C25 170.1(2).

Scheme 7.

Table 1.

Comparisons between C≡N insertion products.

Property	3a	4a	5b
Symmetry from 1H NMR at RT	C2h	Cs	C2v
Fe···Fe distance (Å)	3.0790(5)	2.7816(8)	N/A
N=C distance (Å)	1.258(2)	1.258(4)	1.256(3)
Dihedral angle between β-diketiminate ligand planes (°)	0a	81.09(9), 80.6 (1)	N/A
N=C stretching frequency (cm−1)	1637	1629	1687
N-C-C angle in imide (°)	125.4(2)	129.3(3)	127.2(3)
Fe–N(imide) distance (Å)	2.041(2), 2.079(2)	2.000(3), 2.016(3)	1.857(2)

^aThe two β-diketiminate ligand planes are related by a crystallographic inversion center.