Nitrogen fixation and reduction at boron

See allHide authors and affiliations

Science  23 Feb 2018:
Vol. 359, Issue 6378, pp. 896-900
DOI: 10.1126/science.aaq1684

Boron learns to give back to nitrogen

Although diatomic nitrogen is famously inert, a variety of transition metals can bind to it through a process termed backbonding. As the nitrogen weakly shares its own electrons, some electrons from the metal reach back out to it. Nonmetals would not seem to have the capacity for this type of bonding, but now Légaré et al. show that conventionally electron-deficient boron can be coaxed into it (see the Perspective by Broere and Holland). The authors treated boron-based precursors with potassium under a nitrogen atmosphere to produce several compounds with sandwiched dinitrogen between two boron centers in reduced motifs reminiscent of metal complexes.

Science, this issue p. 896; see also p. 871


Currently, the only compounds known to support fixation and functionalization of dinitrogen (N2) under nonmatrix conditions are based on metals. Here we present the observation of N2 binding and reduction by a nonmetal, specifically a dicoordinate borylene. Depending on the reaction conditions under which potassium graphite is introduced as a reductant, N2 binding to two borylene units results in either neutral (B2N2) or dianionic ([B2N2]2–) products that can be interconverted by respective exposure to further reductant or to air. The 15N isotopologues of the neutral and dianionic molecules were prepared with 15N-labeled dinitrogen, allowing observation of the nitrogen nuclei by 15N nuclear magnetic resonance spectroscopy. Protonation of the dianionic compound with distilled water furnishes a diradical product with a central hydrazido B2N2H2 unit. All three products were characterized spectroscopically and crystallographically.

The element nitrogen is essential for life on Earth and makes up over three-quarters of the atmosphere by volume, yet its common elemental form (dinitrogen, N2) is extremely stable and thus difficult to utilize. Nature overcomes this through the nitrogen-binding enzymes called nitrogenases (1, 2), whereas industry relies on the energy-intensive, transition-metal–catalyzed Haber-Bosch process (3, 4) to convert dinitrogen to ammonia for the production of fertilizer. In the century since the discovery of the Haber-Bosch process, a number of transition-metal (TM) species have been found to bind (and even functionalize) N2 at low temperatures (514). The rare ability of certain transition-metal complexes to bind N2 is attributed to their advantageous combination of unoccupied and occupied d orbitals, which are of appropriate energy and symmetry to synergically accept electron density from and backdonate to N2 (Fig. 1A). The latter process, termed π backdonation, weakens the N-N bond while simultaneously strengthening the metal-nitrogen bond and is thus crucial in effective N2 binding and activation. In contrast to transition metals, main-group compounds generally lack the combination of empty and filled orbitals required to form bonds of σ and π symmetry, respectively, and thus very few are able to provide π backbonding. Accordingly, N2 binding by p-block compounds is currently restricted to a handful of highly reactive, most often strongly Lewis acidic, species generated in the gas phase or under matrix isolation conditions (1518). Of the main-group elements, only the strongly reductive element lithium reacts with N2 at room temperature (albeit slowly) to give an isolable product, the binary nitride Li3N (19).

Fig. 1 Nitrogen binding motifs to transition metals and borylenes.

(A) Simplified schematic of the bonding in well-known end-on-bound transition metal N2 complexes (left) and prospective application to monovalent boron species (BRL) (right, this work). (B) Known photodecarbonylation and ligand binding chemistry of tri- (i and iii) and dicoordinate (ii) borylene species. (C) Outcomes of various reduction reactions of dihaloborane adduct 1, including generation of a transient dicoordinate borylene species (2) and its reaction with dinitrogen. Dip, 2,6-diisopropylphenyl; Dur, 2,3,5,6-tetramethylphenyl.

Over the past decade, advances in the chemistry of low-valent, low-coordinate main-group elements have indicated that these compounds, often bearing reactive lone pairs of electrons as well as vacant orbitals at the same atom, effectively mimic transition metals in many reactions (20). Undoubtedly the most well-developed examples of these are the singlet carbenes (:CR2), particularly the σ-donating and π-accepting carbenes developed by Bertrand and co-workers (21). The combination of filled and empty orbitals proximal in space and energy in these compounds has facilitated binding and activation of a number of challenging small molecules such as H2, NH3, and P4 under mild conditions. Base-stabilized borylenes (22, 23), compounds containing a monovalent boron atom and one or two neutral donor groups (:BRLn; n = 1, 2; R = anionic substituent; L = neutral donor; Fig. 1B), are a promising recent addition to the family of main-group metallomimetics. In 2014, Stephan and Bertrand reported the fixation of CO at the boron atom of a borylene-like allenic dicoordinate (RBL) species (Fig. 1B, ii and iii) (24), and we have subsequently detailed photolytic decarbonylation and donor exchange reactions (Fig. 1B, i to iii) at monovalent boron centers (25, 26)—all characteristic reactions of transition metals. Our observation of strong π-backbonding from boron to CO in a number of borylene complexes (Fig. 1B, i), which is also a bonding motif critical to end-on dinitrogen binding to metal centers (Fig. 1A), inspired us to test our borylene fragment as a scaffold to bind N2. The suitability of borylenes as candidates for N2 binding is further underlined by a recent report of N2 binding by the unstabilized borylene “PhB:” under matrix conditions (18).

In previous work, we showed that borylene species [DurB(CO)(CAAC)] [3, Fig. 1C; Dur = 2,3,5,6-tetramethylphenyl; CAAC = 1-(2,6-diisopropylphenyl)-3,3,5,5-tetramethylpyrrolidin-2-ylidene] undergoes photolytic decarbonylation to generate the transient dicoordinate borylene [DurB(CAAC)] (2), which could be trapped with Lewis bases (pyridine, isonitrile, N-heterocyclic carbene) or would undergo C-H activation in the absence of a base (26). However, similar attempts to bind N2 to the boron atom by photodecarbonylation under a N2 atmosphere provided only the previously reported C-H activation product 4.

During a search for a more convenient synthetic route to 3 {previously accessible only by addition of CAAC to transition-metal borylene complex [Fe(BDur)(CO)3(PMe3)] (26)}, we found that this compound could be prepared by reducing [(CAAC)BBr2Dur] (1) with potassium graphite (KC8) under an atmosphere of CO (Fig. 1C). Similarly, reduction of 1 under an argon atmosphere produced the aforementioned C-H activation product 4. These reactions again suggested the intermediacy of dicoordinate borylene species [DurB(CAAC)] (2, Fig. 1C)—but notably without photolytic conditions that could be detrimental to the formation of N2 adducts—prompting us to attempt this reaction under an atmosphere of N2. To this end, toluene was added to a solid mixture of [(CAAC)BBr2Dur] (1) and five equivalents of KC8 at –80°C under 1 atm of dinitrogen, which led to an immediate change from colorless to fuchsia, a color not encountered in analogous reactions under either argon or CO. The resulting mixture was allowed to warm up to room temperature and after overnight reaction showed the presence of C-H activation product 4 by 11B nuclear magnetic resonance (NMR) [δB ~70 and 86 parts per million (ppm), corresponding to two diastereoisomers (26)] and provided a very low yield of dark purple crystals of new compound 5 (Fig. 1C) with a substantially upfield-shifted 11B NMR signal (δB 10.7 ppm).

A single-crystal x-ray diffraction study of these crystals showed them to comprise the dinitrogen bis(borylene) compound {[(CAAC)DurB]22-N2)} (Fig. 1C and Fig. 2, top). The solid-state structure of 5 showed two [(CAAC)DurB] fragments bound through their boron atoms to a central N2 unit. The two halves of the molecule are considerably different, with the B2-N2-N1 angle [146.1(2)°] significantly closer to linearity than B1-N1-N2 [131.9(2)°]. The B1-N1 [1.423(4) Å] and B2-N2 [1.403(5) Å] distances are equivalent within experimental uncertainty and are in the range of conventional B-N double bonds in aminoboranes (27). The N1-N2 bond length [1.248(4) Å] lies far closer to the range of N-N double bonds than single bonds in organic N-N compounds, as exemplified by azobenzene [d(N=N): 1.247(2) Å] (28) and 1,2-diphenylhydrazine [d(N–N): 1.394(7) Å] (29), respectively. However, the B1-N1-N2-B2 torsion angle [113.3(4)°] is not consistent with double bonding between two sp2-hybridized nitrogen atoms. In contrast to 5, dinuclear TM complexes of N2 predominantly contain linear MNNM axes (614), as the metals’ d orbitals allow overlap with the two orthogonal π orbitals of N2. Despite the nonlinear B-N-N-B unit in 5, the solid-state structure suggests that the B-N bonding in 5 resembles that of a TM-like (N2-to-M σ-donation and M-to-N2 π-backbonding) coordination complex of N2 rather than that of organic azo (i.e., C-N=N-C), hydrazido [C-N(H)-N(H)-C], or diimido (C=N-N=C) compounds. This is hinted at by the short B-Ccarbene bonds [B1-C1: 1.528(5) Å; B2-C2: 1.541(4) Å], which suggest that the boron atoms have monovalent borylene character, to which the N2 unit acts as a Lewis σ-donor and π-acceptor. Indeed, the formal boron-centered lone pairs of electrons appear to be delocalized between both the N2 and CAAC ligands in π-backbonding interactions that are plausibly the key to the stability of the complex (22, 23). Metal-to-N2 π-backdonation is also known to be an important factor in many transition-metal complexes of N2 and to contribute to the weakening of the N2 bond, critical for its functionalization (614). Accordingly, 5 is structurally similar to 3 but features slightly longer B-Ccarbene bonds [3: 1.499(2) Å], suggesting that the N2 fragment in 5 accepts a higher degree of backbonding from boron than does CO in 3, thus decreasing the B-to-CAAC backdonation. The strong B-to-N2 backbonding can also explain the long N-N bond in 5 when compared to many bridging, end-on-bound metal-dinitrogen complexes (14). The bend in the structure of 5 is not unexpected, however, as the geometries of the suitable backbonding orbitals are different in the case of boron (p) than for TMs (d). Density functional theory (DFT) calculations (fig. S1 and table S1) accurately reproduced the solid-state geometry of 5 and allowed us to study the orbitals involved in the borylene-N2 bonding. Kohn-Sham orbital analysis of this structure showed that the HOMO-1 consists of a combination of one of the N-N(π*) orbitals, the boron p and the C-Ncarbene(π*) orbitals (fig. S1). This electronic pattern is typical for the sharing of boron’s formal lone pair of electrons with both the CAAC and N2 ligands. Computed Wiberg bond indices (WBIs) on this geometry are consistent with strong B-N and N-N bonds with multiple-bond character (N1-N2: 1.51; B1-N1: 1.14; B2-N2: 1.30) (see table S1 for comparison with model compounds).

Fig. 2 Optimized syntheses and crystallographic characterization of borylene dinitrogen compounds.

(A) Reduction of radical 6 to form dipotassium complex 7. (B) Synthesis of 5 via reversible oxidation of 7 in air. (C) Product (8) of the protonation of 7 in water. All ellipsoids are shown at the 50% probability level. Hydrogen atoms (except the nitrogen-bound hydrogens of 8) and peripheral ellipsoids have been removed for clarity.

As evidenced by 1H NMR spectroscopy, 5 exists in solution as a mixture of two conformers that interconvert rapidly at room temperature (fig. S21). At –38°C, the exchange is slow enough to allow the acquisition of well-resolved spectra (fig. S22). These show that both conformers are unsymmetrical, similar to the solid-state structure, with two distinct sets of CAAC and duryl resonances for each conformer. Unfortunately, however, the dinitrogen compound 5 could not be reproducibly separated from the borylene C-H activation product 4 owing to the low yield of 5 and the very similar solubilities of the compounds.

Given this complication, and the need for two borylene fragments to bind the dinitrogen in 5, we reasoned that maximizing the concentration of the borylene 2 in the reaction solution might lead to better selectivity to 5 and its isolation. We therefore sought to use the one-electron reduction product of 1, the captodatively stabilized radical [(CAAC)BBrDur] (6, Fig. 1C), as an advanced intermediate to rapidly generate borylene 2 by a simple one-electron reduction step. Radical 6, the bromide analog of previously published radical [(CAAC)BClDur] (30), was thus prepared and isolated by reduction of 1 with a 1.5-fold excess of KC8 under argon (Fig. 1C; see supplementary materials for details). Subsequently, solid 6 was treated with 10 equivalents of KC8 in toluene at –80°C under dinitrogen, and the solution turned an intense fuchsia, indicating the formation of 5. The flask was then placed under 4 atm of dinitrogen and with stirring warmed to room temperature over ~20 min, during which time the initial fuchsia color gave way to deep blue. After 4 hours of reaction, a dark blue, highly sensitive solid was isolated in good yield (64%) from this mixture and ascertained to be the dipotassium complex {[(CAAC)DurB]22-N2K2)} (7, Fig. 2), a product of two-electron reduction of dinitrogen compound 5. Conveniently, in contrast to 5, 7 can be separated from the side-product 4 and its presumed overreduction decomposition products by washing the solid with cold pentane. Complex 7 shows a 11B NMR signal at δB 31.2 ppm, substantially downfield from that of neutral dinitrogen complex 5B 10.7 ppm). Both solution-state 1H NMR spectroscopy and x-ray crystallography (Fig. 2) showed that 7 is a symmetrical species, in contrast to 5. The solid-state structure of dipotassium compound 7 shows notably shorter B-Ccarbene bonds [7: 1.470(4) Å] and longer B-N bonds [7: 1.484(4) Å] than neutral 5. This is consistent with increased B-to-CAAC backbonding resulting from a lesser degree of B-to-N backbonding, to be expected with the population of orbitals of the N2 fragment concomitant with its reduction. The N-N bond is also slightly (4.5%) longer in 7 [1.304(3) Å] than in 5. A similar, but smaller, difference was observed by Holland and co-workers between a (LFe=N=N=FeL) complex and its reduced dipotassium congener, the latter having a 2.8% longer N-N bond (31). Calculations indicate that the strength of the B-CAAC and B-N bonds mirrors the structural changes from 5 to 7 (WBI: B1-C1 = B2-C2 = 1.42; B1-N1 = B2-N2 = 0.87). The strength of the N-N bond, by contrast, does not change substantially (WBI: 1.65) and remains in the range of the strength for double bonds. The natural charges of the nitrogen atoms, however, become more negative from 5 (–0.358, –0.362) to 7 (–0.485), reflective of the reduction of the dinitrogen fragment. Similar negative charges appear on the CCAAC atom of 7, consistent with a rerouting of the boron backbonding from the N2 fragment to the CAAC ligand. Although 7 can be reliably and reproducibly isolated, it decomposes rapidly when exposed to air, moisture, or protic compounds, especially when in solution.

Although we could not isolate pure samples of neutral dinitrogen compound 5 from the reduction of radical 6 under N2 as it rapidly converts to 7, we were gratified to observe that the product dipotassium complex 7 can be readily oxidized by ambient air to provide pure 5 in good yield (79%) as deep purple crystals (Fig. 2B). Compound 5 can also be quantitatively reduced back to its dipotassium complex 7 by treatment with KC8 under argon. Alternatively, treating the dipotassium complex 7 with distilled water led to a turquoise solution showing no 11B NMR signal, from which a turquoise solid was isolated in 74% yield. This compound was shown by single-crystal x-ray diffraction, electron paramagnetic resonance (EPR) spectroscopy, high-resolution mass spectrometry, and infrared spectroscopy (N-H band at ν = 3403 cm–1; see Fig. 2C and figs. S35 to S37) to be the paramagnetic diradical diborahydrazine compound {[(CAAC)DurB]22-N2H2)} (8, Fig. 2C, bottom). The EPR spectrum of a powder sample of 8 at 290 K (fig. S35) shows the typical signature of a triplet state, with a half-field signal at about g = 4 and zero-field splitting parameters of |D/hc| = 0.021 cm−1 and |E/hc| = 0.00083 cm−1. The solid-state structure of 8 also supports its description as a hydrazine complex. With a N-N bond length of 1.402(2) Å, 8 is the only compound of our series to bear a distinct N-N single bond. The B-N distances [1.435(4) and 1.417(4) Å] are in the range of double bonds, which explains the planar geometry of the N atoms and indicates that their nonbonding electron pairs are donated to boron. Mirroring the conversion of the dinitrogen moiety to a hydrazido fragment, the computed WBI for the N-N is also the smallest of the series (1.13), which is consistent with the assignment of a single bond and similar to the parent hydrazine and diphenylhydrazine calculated as model compounds (1.04 and 1.02, respectively). The N atomic charges are also the most negative of the series (–0.581, –0.572). That hydrazino compound 8 is a diradical is unsurprising, given that stable radicals of the form [(CAAC)BR2] are known, of which the aforementioned radical 6 is an example (30).

As a final confirmation that the products reported herein arise from the fixation of dinitrogen, we prepared the 15N isotopologues of the diamagnetic species 5 and 7 and characterized them by 15N NMR spectroscopy. Dipotassium complex 7-15N was synthesized in a similar manner to 7 by the reduction of 6 with ~10 equivalents of KC8 under an atmosphere of 15N2 (98% isotopic purity) and was isolated in 40% yield as a blue solid showing analogous 11B and 1H NMR spectra to 7, and a broad singlet at 235 ppm in its 15N NMR spectrum (fig. S42). High-resolution mass spectroscopy (HRMS) allowed determination of the exact empirical formula of 7-15N, showing a molecular ion signal and isotopic distribution corresponding to replacement of two K+ ions with H+ ions (as also observed in the HRMS of its 14N isotopologue 7) and full 15N enrichment (fig. S43). A C6D6 solution of 7-15N was subsequently exposed to ambient air directly in the NMR tube to give quantitative conversion to 5-15N as confirmed by 11B and 1H NMR and HRMS (fig. S47). Analogous to 5, 5-15N appears to be present in solution as a pair of unsymmetrical conformers. Consequently, its 15N NMR spectrum of 5-15N displayed four broad signals at δ = 58.8, 52.7, 46.7, and 42.6 (fig. S46).

These results provide an important bridgehead for dinitrogen activation with elements of the p block, having the potential to lead to a range of nitrogen-containing molecules with exciting applications.

Supplementary Materials

Materials and Methods

Figs. S1 to S54

Tables S1 and S2

References (3250)

References and Notes

Acknowledgments: This work was financially supported by the Deutsche Forschungsgemeinschaft under its Research Grants program (H.B.) and Research Training Group GRK2112 - Molecular Biradicals: Structure, Properties and Reactivity (B.E.). M.-A.L. thanks the National Sciences and Engineering Research Council of Canada for a postdoctoral fellowship. G.B.-C. thanks the Alexander von Humboldt Foundation for a postdoctoral fellowship. We thank R. Bertermann for NMR experiments, K. Radacki for crystallographic insight, and C. Mahler for mass spectrometry. X-ray data are available free of charge from the Cambridge Crystallographic Data Centre under reference CCDC-1578524 (5), -1578525 (1), -1578526 (8), -1578527 (6), and -1578528 (7). Further spectroscopic, crystallographic, and computational data are included in the supplementary materials.

Stay Connected to Science

Navigate This Article