Mimicking the Constrained Geometry of a Nitrogen-Fixation Intermediate

Abstract

Both biological and industrial nitrogen reduction catalysts activate N₂ at multinuclear binding sites with constrained Fe–Fe distances. This contrasts with molecular diiron systems, which routinely form linear N₂ bridges to minimize steric interactions. Model compounds that capture the salient geometric features of N₂ binding by the nitrogenase enzymes and Mittasch catalysts would contribute to understanding their high N₂-reduction activity. It is shown in the present study that use of a geometrically flexible, dinucleating macrocycle allows for the formation of a bridging N₂ ligand with an unusual Fe–Ct_N2–Fe angle of 150° (Ct_N2 = centroid of N₂), a geometry that approximates the α-N₂ binding mode on Fe(111) surfaces that precedes N₂ bond cleavage. The cavity size the macrocycle prevents the formation of a linear Fe–N₂–Fe unit and leads to orbital interactions that are distinct from those available to the linear configuration.

Ammonia (NH₃) is produced annually on a ~180 megaton scale, largely to provide fixed nitrogen for plants.¹ The generally accepted mechanism by which industrial NH₃-synthesis catalysts bind and activate N₂ invokes multinuclear Fe sites, and recent advances in the understanding of biological N₂ fixation points to a similar need for multiple Fe centers at the site of N₂ binding. As such, the coordination chemistry of N₂ continues to be an area of interest as chemists attempt to adapt the superior performance of these ammonia synthesis catalysts into molecular systems.

Both biological and industrial catalysts constrain the geometries of their multinuclear, N₂-binding sites. Biological nitrogen fixation occurs at Fe₇MS₉C cofactors (M = Mo, V, Fe), found within the active sites of the nitrogenase enzymes.^2–4 While many potential sites of N₂ binding in these cofactors have been proposed,^5–7 recent work has focused on a “belt” position that spans two, low-coordinate Fe ions (d_Fe–Fe = ca. 2.6 Å, Figure 1).^{8–9,10–11} Industrial ammonia synthesis is also understood to use specific multi-iron centers to activate N₂.^12–13 The catalysts used in the Haber-Bosch process are typically based on metallic iron with main group promotors, the most active face of which is Fe(111).^14–15 Following initial physisorption through end-on N₂-binding (γ-N₂, Figure 1), the adsorbate moves into a bridging, side-on binding mode (α-N₂) prior to N–N bond cleavage.^16–21 This binding mode (and the proposed α’-N₂ structure that follows^15,18) is thought to be critical to N–N bond activation.

Figure 1.

Top: Proposed location of N₂ activation at the FeV cofactors. Bottom: Side- and top-views of the binding modes that precede rate-limiting formation of μ-nitrides on Fe(111) surfaces. The top-, second-, and third-layers of Fe are depicted with orange, dark grey, and light grey spheres, respectively.

The geometry of the α-N₂ intermediate appears to be impacted by the constrained Fe–Fe distance of ca. 4.1 Å for top-layer Fe atoms, which leads to an estimate Fe–Ct_N2–Fe angle to ca. 130°. This contrasts with the vast majority of known molecular systems,²² which form linear Fe–Ct_N2–Fe bridges to minimize steric repulsion. The development of molecular systems that can offer constrained, multiiron sites for N₂ binding is thus of interest for its potential to inform our understanding of industrial (and biological) N₂ fixation.

In this Communication, we describe the synthesis, characterization, and electronic structure of unique diiron-μ-N₂ complexes. The use of a dinucleating macrocycle constrains the geometry of the Fe–N₂–Fe unit, leading to orbital interactions that are unavailable to traditional, linear Fe₂(μ-N₂) complexes but are of interest for their potential to contribute to multinuclear N₂ bond cleavage. We use the ³PDI₂ macrocycle, which contains two pyridyldiimine groups connected by propylene linkers (Figure 2). This ligand was recently shown to stabilize a range of bimetallic systems with different core oxidation states and geometries.^23–25 This work involved the Fe–Fe bonded complexes illustrated at the top-left of Figure 2, [(³PDI₂)Fe₂(μ-Cl)(PR₃)₂][OTf] (^R[Fe₂Cl]⁺, R = Me, Ph), which provided a starting point for the present study.

Figure 2.

Top-left: Synthetic scheme for the formation of ^R[Fe₂N₂]⁰. Right: Crystal structure of ^Ph[Fe₂N₂]⁰ and illustration of the acute Fe–Ct_N2–Fe angle exhibited by this series of complexes. Bottom-left: View along the Fe–Fe vectors for ^Me[Fe₂Cl]⁺ and ^Me[Fe₂N₂]⁰, highlighting the change in the angle between i) a plane created by the imino nitrogens and the linker-carbons α to the imino nitrogens (N– C_α–C_α–N) and ii) a plane created by the three linker carbons (C_α–C_β–C_α).

We set out to explore the binding of N₂ to reduced forms of the ^R[Fe₂Cl]⁺ complexes. Chemical reduction with 2.0 equiv of KC₈ under an atmosphere of N₂ resulted in color changes from dark brown to purplish red. After workup, the products (³PDI₂)Fe₂(μ-N₂)(PR₃)₂ (^R[Fe₂N₂]⁰, Figure 2) were isolated in moderate spectroscopic yields (ca. 50–60 %) and low yields on crystallization (11 % for R = Me and 35 % for R = Ph). Crystallographic analysis of both ^R[Fe₂N₂]⁰ complexes revealed μ-N₂-κ¹(N),κ¹(N’) units that bridge between the metal centers. The Fe–Ct_N2–Fe angles of 150.1° (R = Me) and 151.7° (R = Ph) are similar to the cyclic M₂Fe₃(N₂)₃ complexes described by Holland (see below) but significantly more acute than known dinuclear Fe₂(μ-N₂) complexes (Fe–Ct_N2–Fe > 165°). Three μ-N₂ complexes are known that contain (PDI)Fe units; all display linear Fe–Ct_N2–Fe angles (174.1–177.8°).^26–27 The N–N distance of 1.135(3) Å in ^Me[Fe₂N₂]⁰ [1.139(4) Å in ^Ph[Fe₂N₂]⁰] shows mild activation compared to free dinitrogen (1.098 Å) and is similar to those observed in related (PDI)Fe–N₂ complexes (ca. 1.10–1.13 Å; see Supporting Information).^26–34

Recent reports in the literature have highlighted the use of ligand design to support multinuclear iron systems that can activate N₂. These include Murray’s use of a rigid, trinucleating cyclophane ligand that supports the iron-mediated reduction of N₂ into μ-NH_x groups (x = 1, 2; Figure 3).³⁵ Use of the same ligand scaffold with Cu led to a Cu₃(μ³-N₂) species,³⁶ but a corresponding Fe₃(N₂) species has yet to be detected. Holland and co-workers reported the first example of Fe-mediated N₂ cleavage into nitrides (Figure 3).³⁷ Extensive computational investigations identified a candidate for N₂ activation that was proposed to bind N₂ in a geometry that involves both end-on and side-on coordination modes.³⁸ In related chemistry, Holland described the cyclic M₂Fe₃(N₂)₃-containing complexes (Figure 3; M = K, Rb, Cs),³⁹ which form bridging N₂ geometries that had only previously been found with a handful of early transition metals.^40–43 N₂ is weakly bound in these examples, and before the present case, it was unclear if the acute Fe–Ct_N2–Fe angles of ca. 150° can exist in the absence of the stabilizing alkali metal ions. Further, the extent to which this change in geometry impacts the Fe–N₂ bonding interaction is currently unknown, prompting us to examine the electronic structures of the ^R[Fe₂N₂]⁰ complexes and the factors that determine their unusual N₂-binding geometries.

Figure 3.

Top-left, Middle: Multinuclear iron systems used for N₂ bond cleavage. Top-right: N₂-binding by a tricopper cyclophane complex. Bottom: Use of an M₂Fe₂ unit to bind N₂ with an acute Fe–Ct_N2–Fe angle.

Both compounds were found to be diamagnetic in solution, displaying C_2v symmetry, as determined by a characteristic 4:4:2:2 ratio of integrals for the features that result from the diastereotopic methylene groups. The retention of phosphine ligands was apparent from the ³¹P{¹H} NMR spectra, which revealed singlets at 22.33 ppm for R = Me and 60.97 ppm for R = Ph (THF-d₈, 298 K). The N₂ unit was found to be kinetically inert with respect to dissociation. Either application of a dynamic vacuum or storage of the material under an atmosphere of Ar did not result in detectable decomposition. Both compounds display high N–N stretching frequencies: 2003 cm⁻¹ for ^Me[Fe₂N₂]⁰ and 1959 cm⁻¹ for ^Ph[Fe₂N₂]⁰. The latter shifts to 1896 cm⁻¹ (Δν_NN = 63 cm⁻¹; cf. Δν_NN would be 66 cm⁻¹ assuming a simple oscillator) on use of ¹⁵N₂ during the synthesis, consistent with our assignment of this feature as resulting from an isolated, diatomic N–N stretching mode. Bridging N₂ complexes are typically IR inactive due to the inversion symmetry of linear M–N₂–M units. In this case, the out-of-plane positioning of the N₂ unit results in significant intensity in the IR spectrum for the N–N stretch. This positioning appears to result from constraints imposed by the macrocyclic ligands in ^R[Fe₂N₂]⁰, which display an expanded, arch-shaped geometry compared to the more contracted ligand geometries in ^R[Fe₂Cl]⁺. The angle between the planes of the pyridyl rings illustrates the difference between the structures, with ∠py-py = 89.7° for ^Me[Fe₂N₂]⁰ (82.2° for ^Ph[Fe₂N₂]⁰) compared to 21.7° in ^Me[Fe₂Cl]⁺. The expansion of the macrocycle appears to be limited, however, by torsion within the aliphatic linker (Figure 2). Further unfurling of the macrocycle would force the central methylene into the pocket between the iron centers.

We recently adapted the Δ parameter, an empirical metric used to describe the physical oxidation states of PDI ligands based on characteristic bond lengths of the ligand backbone,⁴⁴ into units suitable for ³PDI₂.²⁵ The Δ values for ^R[Fe₂N₂]⁰ (0.067 for ^Ph[Fe₂N₂]⁰, 0.060 for ^Me[Fe₂N₂]⁰) lie in a range most closely associated with a (³PDI₂)⁴⁻ electron distribution. Cyclic voltammetry on ^Ph[Fe₂N₂]⁰ revealed oxidations at −1.52 and −1.26 V, which likely correspond to PDI-based oxidations. No reduction features of comparable intensity were observed out to the limits of the electrochemical window for the solvent (THF, see Supporting Information). Together with the diamagnetism of the complex and the apparent equivalence of the coordination spheres about the two metal centers, the Δ values suggest the presence of either two low-spin Fe^II ions or two intermediate-/high-spin ions that are strongly antiferromagnetically coupled. We can immediately discount the high-spin assignment due to the short Fe–N_PDI and Fe–P distances (see Supporting Information). DFT calculations were used for further analysis of the electronic structure.

Optimization of a truncated form of ^Me[Fe₂N₂]⁰ provided excellent agreement with the experimental data (see Supporting Information), including a Fe–Ct_N2–Fe angle of 149.7°, an N–N bond length of 1.148 Å, and ν_NN = 2010 cm⁻¹ (Δν_NN = 67 cm⁻¹). The experimental Δ value that led to the (³PDI₂)⁴⁻ oxidation state assignment was also reproduced computationally (Δ_calc = 0.064). The ^R[Fe₂N₂]⁰ complexes formally contain d⁸, Fe⁰ metal centers, but the low-energy π* manifold of the PDI fragment is well known to accept electron density in lieu of Fe^<II oxidation states.^{28, 31, 45} Consistent with the oxidation state assignments inferred from the crystallographic bond lengths, the metal centers are best described as Fe^II. The Fe^II ions further appear to be in low-spin electron configurations. The d_π manifold (xz±xz, yz±yz, xy±xy) was found to be fully occupied, with typical π-backbonding interactions between out-of-phase d_π fragment orbitals combinations (xy–xy, yz–yz) and the N₂ π* manifold. Doing so yields the HOMO-6 [(yz–yz) + π*_z)] and HOMO-7 [(xy–xy) + π*_x)] (Figure 4; for a quantitative MO diagram, see Figure S23 in the Supporting Information).

Figure 4.

Qualitative molecular orbital interaction diagram describing the admixture of the π* system of N₂ to the d-manifold within ^Me[Fe₂N₂]⁰.

For mononuclear (PDI)Fe complexes, the presence of two electron’s worth of electron density on the ligand typically results from single occupation of both redox-active orbitals (b₁ and a₂).^{25, 45} In the case of ^R[Fe₂N₂]⁰, however, the filled, in-phase (HOMO) and out-of-phase (HOMO-1, Figure 4) combinations [(b₁+z²)±(b₁+z²)] are primarily responsible for the four electron reduction of ³PDI₂.

Interestingly, the (b₁+z²) fragment would normally be non-bonding with respect to the N₂ π* manifold, but HOMO-1 reveals a bonding interaction between π*_z and [(b₁+z²)–(b₁+z²)]. Thus, it appears that the constrained geometry imposed by the macrocycle allows for an additional interaction between the Fe-d and N₂-π* manifolds compared to linear Fe–N₂–Fe motifs. The extent of mixing is too low in the present case to result in a significant increase in the degree of N₂ activation, but it is noteworthy that this type of interaction repeats in LUMO+4 (Figure 4), which results from mixing of π*_z with an out-of-phase combination of x²-y². Complexes with weaker ligand field donor sets, in conjunction with constrained Fe–Ct_N2–Fe angles, may be able to take advantage of these orbital interactions for N₂ bond cleavage.

The importance of constrained-geometry multinuclear iron sites for N₂ binding and activation repeat throughout biological and heterogeneous catalysts, but few molecular systems are available that can mimic these interactions. This report described the dinuclear binding of N₂ in a coordination environment reminiscent of the α-N₂ binding mode to Fe(111). The use of a macrocycle with flexible alkyl linkers allowed the metal centers to separate in space following reduction of an Fe–Fe bonded starting material, providing a location for N₂ binding. However, the limited size of the macrocycle cavity prevented the Fe₂N₂ unit from adopting a linear linkage, as is common in diiron μ-N₂ chemistry. The orbital interactions made possible by this geometry provide an avenue for facilitating N₂ reduction, thereby highlighting the manner in which geometric tuning of the active site constitutes an important factor in catalytic ammonia synthesis.