Terminal Iron–Dinitrogen and Iron–Imide Complexes Supported by a Tris(phosphino)borane Ligand

Our currently limited understanding of the mechanism of biological dinitrogen fixation^[1] – i.e. the reduction of N₂ to two equivalents of ammonia by addition of protons and electrons – motivates a sustained effort towards the preparation of functional models of the nitrogenase MoFe-cofactor. One plausible mechanism for N₂ reduction is a so-called distal or Chatt-type cycle^[2], in which three hydrogen-atom equivalents are successively added to the distal N-atom of a metal-bound dinitrogen molecule (M–N≡N), resulting in the elimination of one equivalent of NH₃ to yield an intermediate nitrido complex (M≡N). The latter is in turn converted to ammonia by reaction with three additional protons and electrons. Indeed, such a cycle has been suggested as operative for a mononuclear, molybdenum-based catalyst.^[3] The increasing evidence for substrate coordination at iron in the MoFe cofactor^[1b] warrants the continued investigation of iron-based models in a related context.^[4]

A Chatt-type cycle at an iron center would require a single ligand scaffold to stabilize low-valent complexes with a π-acidic dinitrogen molecule as well as high valent complexes with a π-basic nitride ligand. Our group has investigated this chemistry in two different ligand-imposed geometries that model the local trigonal symmetry of the coordination environment of iron in the MoFe cofactor. The pseudotetrahedral geometry has been successfully used to stabilize terminal imide^{[5, 6]} and nitride^{[7, 8]} ligands because the π-antibonding d-orbitals of e parentage are usually empty, but most of the N₂ complexes in this geometry are dinuclear, bridged species rather than terminal ones due to the difficulty of implementing sufficient steric protection.^{[9, 10]} On the other hand, the trigonal-bipyramidal geometry enforced by tetradentate tris(phosphino)silyl ligands has been shown to stabilize terminal N₂ complexes of iron in three different oxidation states,^[11] but is less well suited for accomodating Fe–N multiple bonds because of the presence of low-lying, π-antibonding orbitals.^[12]

It has been shown that both the trigonal-bipyramidal and tetrahedral geometries can be accessed with a single NL₃-type scaffold (L = phosphine^[13], N-heterocyclic carbene^[8c]) in which the nitrogen donor is labile. We reasoned that a similar behaviour could be accessed with a Lewis-acidic borane in the apical position and therefore turned our attention to the tris(phosphino)borane (TPB) ligand 1 (Scheme 1), which has been introduced recently by the group of Bourissou.^[14] Ligand 1 has been shown to form a range of C₃-symmetrical metallaboratranes with an (M–B)¹⁰ configuration,^[15] in which the B–M distance varies from 2.168 Å in [(TPB)Ni] to 2.540 Å in [(TPB)AgCl], pointing to some flexibility in adapting the metal–boron interaction to the Lewis-basicity of the metal.^[14b,c] Herein we explore the iron chemistry of ligand 1, showing that the (TPB)Fe fragment accommodates both terminal N₂ complexes in low oxidation states and an imidoiron complex at the iron(II) state.

Scheme 1.

Synthesis and reactivity of ferraboratranes derived from tris(phosphino)borane (TPB) ligand 1.

A convenient entry into the iron chemistry of the TPB ligand was found to be the iron bromide complex 2 (Scheme 1), which was obtained as a greenish-brown powder from the comproportionation of iron(II) bromide and iron powder in the presence of TPB (THF, 90 °C, 66 h). Broad ¹H NMR resonances ranging from −23.1 to 92.5 ppm and the solution magnetic moment of 4.1 µ_B (Evans method) indicate an S = 3/2 ground state. The X-ray diffraction (XRD) crystal structure^[16] of 2 (Figure 1) exhibits a slightly distorted trigonal-bipyramidal geometry, with rather long Fe–Br (2.4138(7) Å) and Fe–P (2.3832(13) – 2.4351(11) Å) distances as expected for a high spin complex. The (M–B)⁷ electron configuration of 2 is, to our knowledge, unprecedented in metallaboratrane chemistry.

Figure 1.

Solid state structures of 2, 4, 5b, and 6. Thermal ellipsoids set at 50%. For clarity, hydrogen atoms as well as the [Na(12-crown-4)₂]⁺ counterion and an unbound diethylether molecule in structure 5b are omitted. Only the main component of a disordered isopropyl group in 2 and only one of the two independent molecules of 6 are plotted.

Reduction of 2 with a slight excess of sodium naphthalide (THF, −60 °C to RT) afforded a brown compound that exhibits a strong IR absorption at 2011 cm⁻¹ and a solution magnetic moment of 2.8 µ_B. These data are consistent with its formulation as the terminal (Fe–B)⁸ dinitrogen complex 3 with an S = 1 ground state. Exposure of a yellowish brown solution of 3 to vacuum results in a reversible color change to dark reddish brown accompanied by major changes in the ¹H-NMR spectrum, indicating a labile N₂ ligand. Crystals of 3 invariably suffered from severe twinning that prevented an unequivoqual structure determination, but its monomeric nature was confirmed by a DOSY experiment showing that 2 and 3 have virtually equal diffusion coefficients.

The nature of complex 3 was further confirmed by reacting it with one atmosphere of CO to afford a brown compound that exhibits a single IR absorption at 1857 cm⁻¹ and is consequently formulated as [(TPB)Fe(CO)] (4). Despite the isoelectronic nature of N₂ and CO, compound 4 does not have the expected spin-triplet ground state, but is diamagnetic. The ¹H NMR spectrum of 4 resembles that of a C₃ symmetrical species, but a broad ³¹P resonance at δ 87 ppm suggests fluxionality. Upon cooling to −90 °C, this signal splits into three mutually-coupled resonances at δ 94, 87.5, and 16.7 ppm, indicating an asymmetric geometry. This was confirmed by an XRD crystal structure^[16] (Figure 1), which reveals an additional interaction of the iron center with two carbon atoms of an aromatic ring of the TPB ligand (Fe–C1 2.337(2) Å; Fe–C2 2.321(2) Å), resulting in an overall η⁴ coordination of one arm. Partial dearomatization of the iron-bound ring is evident from C–C bond length alternation (See SI). There are a few examples of related η³-coordinated phenylborane moieties,^[17] but incorporation of this motif in a metallaboratrane cage structure had been, to the best of our knowledge, unknown.

A quasi-reversible wave at –2.19 V vs Fc/Fc⁺ in the cyclic voltammogram (CV) of 3 suggested the accessibility of the anionic N₂ complex 5, which could indeed be isolated as its sodium salt 5a from the reaction of 2 with 2.5 equivalents of sodium naphthalide (THF, −60 °C to RT). The paramagnetic compound 5a has an S = 1/2 ground state, as shown by its solution magnetic moment of 1.6 µ_B and a quasi-axial EPR signal (X-Band, toluene/THF glass, 20K) with g₁ = 2.23, g₂ = 2.09, g₃ = 2.05. The IR spectrum of 5a in THF solution exhibits two intense bands at 1918 and 1877 cm⁻¹, which are attributed to the N–N stretch of the free anion Fe–N≡N⁻ and tight ion pair Fe–N≡N⁻∙∙∙Na⁺, respectively. In accord with this assignment, only the second band is observed in the less-coordinating solvent diethylether (1862 cm⁻¹) and in the solid state (1879 cm⁻¹). This was additionally confirmed by treating 5a with two equivalents of 12-crown-4 to afford the complex salt 5b, which exhibits only the free-anion band at 1918 cm⁻¹ in THF and at 1905 cm⁻¹ in solid state. XRD crystal structures^[16] were obtained for both 5a (SI) and 5b (Figure 1). Both exhibit a distorted trigonal-bipyramidal geometry with P–Fe–P angles of 107.3, 110.3, and 134.6° in 5a and 105.4, 112.3, and 135.0° in 5b, consistent with a Jahn-Teller distortion arising from having three electrons in degenerate orbitals of d_xy, and d_x²−y² character. The main difference between 5a and 5b is that the sodium counterion in 5a is terminally bound to the N₂ unit while the complex cation [Na(12-crown-4)₂]⁺ is isolated, which appears to have little effect on the N–N distance (1.149(2) in 5a vs 1.144(3) in 5b).

To test the ability of the (TPB)Fe platform to support a two electron redox process, compound 3 was treated with p-methoxyphenyl azide (benzene, RT) to yield the stable diamagnetic imido complex 6 (Scheme 1), which possesses the unusual (Fe–B)⁶ configuration. Compound 6 was thoroughly characterized by multinuclear NMR, UV-Vis spectroscopy, and XRD.^[16] Two independent molecules were found in the asymmetric unit of 6, with an average Fe–N distance (1.668 Å) only marginally longer than that of 1.651 Å found in the pseudotetrahedral iron(II) imide [(BP₃)Fe≡N(1-Ad)][ⁿBu₄N] ((BP₃) = [PhB(CH₂PPh₂)₃]⁻).^[5c] The large Fe–N–C angle (170.2°) additionally supports the description of the Fe–N linkage as a triple bond. The very long Fe–B distance (2.608 Å) causes us to describe the coordination geometry as derived from pseudotetrahedral, as confirmed by a sum of P–Fe–P angles (333.0°) close to the value of 328.4° for a regular tetrahedron.

More insight into the electronic structure of 6 is gained by considering the frontier Kohn-Sham orbitals (Figure 2) obtained from density functional theory (DFT) calculations^[18] on the closely related compound [(TPB)Fe≡NPh]. The highest occupied molecular orbital (HOMO) is of d_z² parentage but still possesses some Fe–B σ-bonding character arising from a stabilizing interaction with the empty boron-centered orbital. Two orbitals of d_xy, and d_x²−y² parentage, respectively, lie within a 0.4 eV interval of the HOMO. Accordingly, the two orbitals of π-antibonding character derived from d_xz and d_yz are unoccupied, in accord with the presence of a Fe≡NAr triple bond.

Figure 2.

Frontier Kohn-Sham orbitals of [(TPB)Fe≡NPh] calculated at the B3LYP/6-31G(d) level. Energies in eV.

While metallaboratranes have now been isolated for all transition metals from groups 8 to 11,^[20] examples of metal-based reactivity in such compounds are scarce. These include oxidative cleavage of Fe–B^[21] and Ni–B^[22] bonds as well as reversible B–H bond formation.^[23] To our knowledge, the only previously known redox reactions that preserve the metal–boron bond are deprotonation of the (Pt–B)⁸ cation {[B(mt)₃]Pt(H)(PPh₃)}⁺ (mt = methimazolyl) and oxidative additions to the resulting (Pt–B)¹⁰ complex {[B(mt)₃]Pt(PPh₃)}.^[24] The series of compounds described herein is unique in that they span four different electronic configurations ((Fe–B)⁶ in 6 to (Fe–B)⁹ in 5a,b) that can be interconverted via formal one or two-electrons processes.

This property largely stems from the ability of the Fe–B interaction to respond to changes in the electronic properties of the metal, as is evident from the structural properties summarized in Table 1 and the schematic orbital correlation diagram depicted in Figure 3. In the highly reduced compound 5a, the short Fe–B distance (2.311(2) Å) and strong pyramidalization of the boron atom (Σ(C–B–C) = 329.8°) indicate a strong metal–boron interaction that pulls the iron center into the P₃ plane (Σ(P–Fe–P) = 352.2°). As a result, the σ(Fe–B) bonding orbital is expected to be lower in energy than the d orbitals and the electronic structure of 5a,b parallels that of a d⁷ metal in a trigonal-bipyramidal geometry. This is in accord with the covalent bond classification (CBC) system^[19], in which all σ-bonding electrons are substracted from the d-electron count. In contrast, the Fe–B distance of 2.608 Å in the imidoiron(II) complex 6 points to a weak interaction resulting in a pseudotetrahedral geometry (Σ(P–Fe–P) = 333.0°) in which the two sets of d orbitals of e parentage are inverted as compared to 5a,b. The similarity between this orbital diagram and that previously obtained for the related pseudotetrahedral iron(II) imide [(BP₃)Fe≡N(1-Ad)][ⁿBu₄N]^[5c] suggests that 6 closely resembles a low-spin d⁶ compound, even though the d_z² orbital has some σ-bonding character. The geometry of 2 – and presumably 3 – is intermediate between the trigonal-bipyramidal and pseudotetrahedral extremes, and the consequent energetic proximity of the two sets of degenerate orbitals is consistent with the S = 3/2 ground state of 2. Finally, the neutral carbonyl complex 4 adds another twist with the η⁴-BCCP coordination that results in a Fe–B distance (2.227(2) Å) even shorter than that in 5a,b. It is worth noting that all the observed Fe–B bond length are significantly longer than that of 2.108 Å in the octahedral complex {[B(mt^tBu)₃]Fe(CO)₂}.^[21] This can be attributed to the lower angular flexibility of the tertiary phosphine donors of TPB (<C(Ar)–P–Fe> = 107.5° in 5a) as compared to the terminal thione (<C–S–Fe> = 99.2° in {[B(mt^tBu)₃]Fe(CO)₂)} of the [B(mt^tBu)₃] ligand.

Table 1.

Geometrical Features of Compounds 2, 4, 5a,b, and 6 Relevant to the Fe–B Interaction.

Compound	Fe–B [Å]	Σ(C–B–C)	Σ(P–Fe–P)
2	2.459	341.9°	341.9°
4	2.227	352.0°	357.4°
5a	2.311	329.8°	352.2°
5b	2.293	331.0°	352.7°
6	2.608[a]	338.3°[a]	333.0°[a]

^[a]Averaged over two independent molecules.

Figure 3.

Schematic orbital correlation diagram for compounds 2, 3, 5a,b, and 6.

It is of interest to compare the chemistry of the neutral TPB ligand with that of the closely related anionic (SiP^iPr₃) ((SiP^iPr₃) = [2-(iPr₂P)C₆H₄]₃Si⁻).^[11] Both ligands are able to stabilize low-valent iron–N₂ compounds in a trigonal-bipyramidal geometry, and the degree of activation of the N₂ is similar for compounds having the same electrical charge: ν_NN = 2011 and 1905 cm⁻¹ for 3 and 5b versus 2008 and 1920 cm⁻¹ for [(SiP^iPr₃)Fe(N₂)] and [(SiP^iPr₃)Fe(N₂)][Na(12-crown-4)₂], respectively. This observation may at first appear surprising because the (SiP^iPr₃) complexes have one more valence electron than their TPB counterparts, but it can be easily understood by considering that the additional electron lies in an orbital of d_xy or d_x²−y² parentage that has no overlap with the σ and π orbitals of N₂. A major difference between TPB and (SiP^iPr₃) resides in the greater flexibility of the Fe–B bond length as compared to Fe–Si,^[11] which explains why the triply-bonded imido complex 6 is stable while imidoiron complexes of (SiP^iPr₃) are short-lived intermediates in the formation of azo compounds from aryl azides.^{[12, 25]}

In summary, the TPB ligand has been shown to stabilize both low-valent iron–dinitrogen complexes and a mid-valent imido species with a Fe≡NAr triple bond, thanks to its ability to shuttle between trigonal-bipyramidal and pseudotetrahedral geometries by elongation of the apical iron–boron bond. The adaptability of the coordination environment in ferraboratranes derived from triphosphinoborane ligands makes them promising candidates as functional models for biological N₂ fixation, which will be the subject of further investigation in our laboratories.