Coupling Dinitrogen and Hydrocarbons through Aryl Migration

Abstract

A persistent challenge in chemistry is to activate abundant, yet inert molecules such as hydrocarbons and atmospheric N₂. In particular, forming C–N bonds from N₂ typically requires a reactive organic precursor¹, which limits the ability to design catalytic cycles. Here, we report an diketiminate-supported iron system that is able to sequentially activate benzene and N₂ to form aniline derivatives. The key to this new coupling reaction is the partial silylation of a reduced iron-N₂ complex, which is followed by migratory insertion of a benzene-derived phenyl group to the nitrogen. Further reduction releases the nitrogen products, and the resulting iron species can re-enter the cyclic pathway. Using a mixture of sodium powder, crown ether, and trimethylsilyl bromide, an easily prepared diketiminate iron bromide complex² can mediate the one-pot conversion of several petroleum-derived compounds into the corresponding silylated aniline derivatives using N₂ as the nitrogen source. Numerous compounds along the cyclic pathway have been isolated and crystallographically characterized; their reactivity outlines the mechanism including the hydrocarbon activation step and the N₂ functionalization step. This strategy incorporates nitrogen atoms from N₂ directly into abundant hydrocarbons.

Reduction or “fixation” of N₂ is accomplished by a few catalytic systems^3–8 which fully reduce it to ammonia (NH₃)⁹. It is difficult to adapt these processes to form C–N bonds from N₂, even though a number of compelling functionalizations of N₂ are thermodynamically feasible¹⁰. Efforts to form C–N bonds from N₂ generally require reduction of N₂ to make it nucleophilic enough to react with carbon electrophiles^11,12. Notable examples include synthetic cycles for C–N bond formation from Cummins, Sita, Mori, and Schneider^13–17. A frequent difficulty is the direct reaction of the reducing agents with the carbon electrophiles.

There would be a substantial advantage if N₂ could be induced to form a C–N bond with a simple hydrocarbon, without prior functionalization of the hydrocarbon to convert it into an electrophile. Such a strategy would require C–H activation of the hydrocarbon, leveraging the decades of work on oxidative addition of C–H bonds by low-valent transition metal complexes¹⁸. C–H functionalization to form cross-coupling products is well-known¹⁹, but there are no previous examples of cross-coupling of C–H bonds with the relatively unreactive N₂ molecule.

Here, we report an iron complex that couples N₂ and unactivated arenes at low temperature, taking advantage of silyl activation of the N₂ to produce silylated anilines in a one-pot procedure. The overall strategy (Fig. 1) begins with C–H bond activation of benzene to form a phenyl fragment, which could migrate to N₂ upon silylation to form the key C–N bond. To explore the feasibility of this pathway, we converted the iron(II) bromide complex [LFe(μ-Br)]₂ (1) into the iron(I) benzene complex LFe(η⁶-C₆H₆) (2)²⁰. Fig. 2 shows the structure of 2 and illustrates the β-diketiminate ligand L, which controls the coordination environment of iron²¹. Reduction of 2 with KC₈ in the presence of 18-crown-6 (18c6) at room temperature led to the isolable, purple iron(0) complex LFe(η⁴-C₆H₆)K(18c6) (3-K)². When 2 was reduced instead with Na and 15-crown-5 (15c5) in benzene, the product was the red iron(II) complex LFe(H)(Ph)Na(15c5) (4-Na), in which a C–H bond of benzene is broken to give an iron(II) complex with phenyl and hydride groups on the iron (Fig. S50). Thus, 3-K and 4-Na have isomeric anions (top of Fig. 2A). Mössbauer spectra of solid 3-K and 4-Na (Fig. S23–S25) support the difference, and density functional theory (DFT) studies indicate that the molecules have high-spin electronic configurations (S = 1 for 3-K and S = 2 for 4-Na). This is a rare case of room-temperature C–H activation using a high-spin iron complex^22,23. Reduction of 2 with Na and 15-crown-5 in toluene similarly gave LFe(H)(Tol)Na(15c5), from activation of the aryl C–H bonds of the solvent. The X-ray crystal structure refined best to a 61:39 mixture of meta and para isomers (Fig. S54).

Fig. 1. Strategy for converting benzene and N₂ into silylated aniline without the use of carbon electrophiles.

Reduction steps are not shown here for simplicity, but are elaborated in Fig. 4 below. Inset: the iron β-diketiminate fragment used in this implementation of the strategy.

Fig. 2. Activation of benzene.

(a) Reduction of 2 with KC₈/18c6 formed 3-K (left) while reduction with Na/15c5 formed 4-Na (right). In THF solvent the cations dissociate, and isomeric anions 3 and 4 are in equilibrium. (b) Conversion of 4-Na to 6-Na occurred spontaneously in 18% yield (top pathway), but gave a higher yield of 75% with the hydride acceptor BPh₃ (bottom pathway), suggesting that hydride loss is likely to be the main reaction pathway.

The solution ¹H NMR (nuclear magnetic resonance) spectra of 3-K and 4-Na in C₆D₆ showed a few peaks that disappeared over several hours, suggesting that these benzene or phenyl groups exchange with C₆D₆. Additionally, a minor component was observed in the ¹H NMR spectrum of the C₆D₆ solutions of 3-K that resembled those of 4-Na, and vice versa. These observations suggested that arene-bound iron(0) complex 3 and C-H activated iron(II) complex 4 are in equilibrium with one another, and the iron(0) arene isomer (3) predominates with K(18c6)⁺, while the oxidatively added iron(II) phenyl hydride isomer (4) predominates with Na(15c5)⁺ (see Tables S7–S8). This hypothesis was strongly supported through solution studies in THF (tetrahydrofuran), a solvent that can disrupt the interactions with the alkali metal. Dissolving either purified 3-K or purified 4-Na in THF gave mixtures with characteristic peaks of both isomers 3 and 4 in ¹H NMR and Mössbauer spectra (Fig. S9, S26). The ability of each isomer to produce the other one demonstrates that there is reversible oxidative addition of the aryl C–H bond of 3 to form 4.

Purple THF solutions of LFe(H)(Ph)Na(15c5) (4-Na) turned green upon standing at room temperature for several hours, and ¹H NMR spectra showed formation of the iron(I) phenyl complex [LFePh][Na(15c5)] (6-Na) (Fig. 2B). Though the formation of 6-Na from 4-Na corresponds to formal loss of H•, the spectroscopic yield of 6-Na from 4-Na in THF was only 18% (Fig. S17), and the best yield of 6-Na (44%) came from addition of excess 15c5 and Na to 2 in diethyl ether (Fig. S18). These results suggested that the reaction could have a different pathway. In a mechanistically revealing experiment, 4-Na transferred the hydride to BPh₃ (triphenylborane) to give 5 and [Na(15c5)][HBPh₃] (Fig. S19–S20) in a higher yield of 75%. Therefore, it is possible that in the presence of excess reducing agent, 4-Na can similarly lose hydride to give 5, which is subsequently reduced by 3 or by Na to give the observed 6-Na.

Next, we explored N₂ binding. Cooling a solution of 6-Na in THF under an atmosphere of N₂ led to changes in the electronic absorption (UV-vis) spectrum (Fig. 3A) and Mössbauer spectrum (Fig. 3B) that did not occur in control experiments under argon. Van’t Hoff analysis of the UV-vis data for 6-Na gave ΔH = −17 ± 2 kJ mol⁻¹ and ΔS = −55 ± 10 J K⁻¹ mol⁻¹ (Fig. S40), which are consistent with the binding of N₂. In order to gain crystallographic verification of N₂ binding, we cooled a concentrated sample of the potassium analogue 6-K at −78 °C for 3 h, which led to crystals of the N₂ complex 7-K. The X-ray crystal structure of 7-K (within Fig. 4 below) demonstrates end-on binding of the N₂ unit and a pseudotetrahedral geometry at the iron(I) site, and the spectroscopic similarity to the Na system (Figs. S36–S39) indicates that 7-Na has a similar structure.

Fig. 3. Binding and functionalization of N₂.

(a) Electronic absorption spectra of 6-Na in THF solution at various temperatures under 1 atm N₂, demonstrating the formation of 7-Na at lower temperature. The data for 6-K are similar (Fig. S37). (b) ⁵⁷Fe Mössbauer spectra at 80 K, from samples of 6-K in THF solution frozen under 1 atm Ar (top) or 1 atm N₂ (bottom), demonstrating the formation of a new doublet attributed to the N₂ complex 7-K (green subspectrum) in addition to some remaining 6-K (red subspectrum). (c) Illustration of the equilibrium of 6 and N₂ to form 7. In a key transformation, silylation of the bound N₂ in 7 induces migration of the phenyl group to the N₂ fragment, giving 8. When a bulkier silyl group is used, it is possible to isolate the singly silylated species 9, which can subsequently form C–N bonds. Each of the iron species has been characterized through X-ray crystallography (Fig. 4 and S50–S55).

Fig. 4. Proposed cyclic reaction mechanism for the conversion of N₂ and benzene to aniline mediated by iron β-diketiminate complexes.

Thermal ellipsoid plots of the isolated intermediates are shown on the outside of the cycle (for ionic compounds, only the anion is shown). Hydrogen atoms and isopropyl groups are omitted from the thermal ellipsoid plots for clarity.

Next, we silylated the N₂ ligand that had been activated through Fe coordination. Treating a cold solution of 7-K with 2 equivalents of Me₃SiX (X = Br, I) formed the hydrazido complex LFe(N(Ph)N(SiMe₃)₂) (8). To our knowledge, the transformation of 7 to 8 is the first crystallographically-verified example of the migration of a hydrocarbyl group from a metal to the α position of N₂^25–26. This is the key C–N bond forming step during the formation of silylated anilines (Fig. 1), which differs from the previously utilized attacks on bound N₂ by carbon electrophiles.

Though we did not detect any intermediates during the conversion of 7 to 8, the treatment of a cold solution of 7-K with 0.5 molar equivalents of the bulkier triisopropylsilyl triflate (TIPSOTf; OTf = SO₃CF₃) gave the formally iron(II) diazenido complex 9, in which the N₂ is singly silylated while the phenyl group remains bound to the iron (Figure S55). The conversion of 7 to 9 is accompanied by a similar yield of the iron(II) complex 5, which results from half of 7 acting as a reducing agent. Considering this stoichiometry, the formation of 9 occurs in 67% yield. The isolation of 9, in which the phenyl has not migrated, suggests that the initial silylation of the β position of the coordinated N₂ in 7 takes place before the migration of the aryl group. It is likely that the second silylation induces the aryl migration, because Peters recently reported the migration of H from an iron center to the α position of a doubly silylated N₂ group to form an iron disilylhydrazido complex²⁷. Accordingly, addition of excess trimethylsilyl triflate (Me₃SiOTf) and excess Na to 9 gave a 14% yield of PhN(SiMe₃)₂, showing that addition of a second silyl group can initiate C–N bond formation. This could be because the second silylation leads to a formally iron(IV) complex with a Fe=N double bond, a migration that is reminiscent of alkyl migration to N in an imidoiron(IV) complex²⁸.

The conversion of 8 to the silylated aniline and amine occurred upon the addition of 1 equivalent of KC₈ and 2 equivalents of Me₃SiX (X = Br, I) to solutions of 8: this treatment led to mixtures containing PhN(SiMe₃)₂, PhN(SiMe₃)N(SiMe₃)₂ and N(SiMe₃)₃ within 30 min, either at room temperature or at −100 °C. In contrast, 8 did not react with 2 equivalents of Me₃SiBr alone within 3 days at room temperature. This result suggests that reduction precedes the electrophilic attack of silyl groups on 8, and alternatively this final step could involve the formation of Me₃Si^• radicals^28,30. Others have also studied the reductive silylation of disilylhydrazido complexes^30,31.

After release of the silylated nitrogen products from 8, the large excess of bromide is expected to give iron(II) bromide species that are poised to be reduced with further arene binding. This suggests the feasibility of a cyclic process (Fig. 4) in a single pot, which forms silylated anilines from arenes and N₂. However, the C–H activation and hydride loss to reform complex 6 requires room temperature treatment with Na, and at this temperature Na degrades Me₃SiBr. Further, N₂ binds to 6 at low temperatures. Thus, we treated 8 with Na (25 equivalents), benzene (20 equivalents), Me₃SiBr (6 equivalents), and 15c5 (5 equivalents) at −100 °C in diethyl ether, then warmed to room temperature for 1 h, then cooled again and treated with additional Me₃SiBr (6 equivalents). This led to a 92% yield of PhN(SiMe₃)₂ and a 135% yield of N(SiMe₃)₃ (yields vs. Fe). In order to verify that the overall process is indeed cyclic, 8 was treated with the same conditions with added toluene in place of benzene. Analysis of the organic reaction products after two cycles of Me₃SiBr addition showed a 163% yield of N(SiMe₃)₃, a 62% yield of PhN(SiMe₃)₂, and a 16% yield of (tolyl)N(SiMe₃)₂ (both meta and para isomers). The ability of the phenylhydrazido complex to give products from toluene amination demonstrates that the iron containing products of the hydrazido reduction can activate toluene and continue to another reaction cycle, though the yields are low.

To test the overall cycle, we reacted the easily-prepared iron(II) complex 1 with 30 molar equivalents (equivalents relative to [Fe]) of Na, 20 equivalents of C₆H₆, and 5 equivalents of 15c5 at ambient temperature in diethyl ether, followed by addition of 6 equivalents of trimethylsilyl bromide (Me₃SiBr) at −108 °C, afforded PhN(SiMe₃)₂ in a yield of 24% per iron atom as determined by gas chromatography (GC). The yield of the reaction could be increased by adding the Me₃SiBr in several portions with temperature cycling. First, a mixture of 1, 35 equivalents of Na, 20 equivalents of benzene, and 5 equivalents of 15c5 in diethyl ether was stirred vigorously for 1.5 hours at room temperature until it became green, corresponding to the color of 6. Cooling the mixture to −108 °C under 1 atm N₂ resulted in a color change to dark red, corresponding to 7. Then, addition of 2 equivalents of Me₃SiBr (per iron) to this cold solution and warming to room temperature for 1 hour resulted in another green reaction mixture, suggesting that 6 was regenerated. Cooling again caused the same color change to red (7), and more Me₃SiBr was then added in a second cycle. Repeating 10 cooling/silylation/warming cycles with 2 equivalents of Me₃SiBr per cycle gave a cumulative yield of 85 ± 14% of PhN(SiMe₃)₂ (vs. Fe; average and standard deviation of 6 trials; Table S1). The ability to produce more product with repeated Me₃SiBr additions suggests a cyclic process, albeit one where a significant amount of the active species decomposes in each cycle. Though aniline formation was attenuated with repeated cycles, the yield of N(SiMe₃)₃ continued to increase with the number of additions of Me₃SiBr, reaching 380 ± 41% (vs. Fe) (Fig. S1–S2). Under the same conditions but in the absence of benzene, no PhN(SiMe₃)₂ was produced yet a similar catalytic yield of N(SiMe₃)₃ was observed. Neither product was detected in the absence of 1. These results suggest that the iron decomposition products lose the ability to aminate benzene, but remain competent for the silylation of N₂ to N(SiMe₃)₃, a more common reaction that has been reported with other homogeneous catalysts and decomposition products^32–40.

Isotope labeling experiments were used to verify that the aniline product arises from benzene and N₂. When performing the reaction under an atmosphere of ¹⁵N₂, GC-MS indicated the formation of Ph¹⁵N(SiMe₃)₂ and ¹⁵N(SiMe₃)₃ (Fig. 5A), demonstrating that N₂ is the source of the N atoms. Performing the reaction with C₆D₆ as the arene substrate gave (C₆D₅)N(SiMe₃)₂, showing that benzene is the source of the phenyl group. The reaction with an equimolar mixture of C₆H₆ and C₆D₆ gave a 1:1 mixture of (C₆H₅)N(SiMe₃)₂ and (C₆D₅)N(SiMe₃)₂ (Fig. 5B), but partially deuterated 1,3,5-d₃-benzene gave (C₆H₂D₃)N(SiMe₃)₂ and (C₆H₃D₂)N(SiMe₃)₂ in a ratio of 2.01 ± 0.05 indicating a normal primary kinetic isotope effect for the C–H cleavage step. The difference between the intramolecular vs. intermolecular isotope effects⁴¹ indicates that arene binding/exchange is not much more rapid than the irreversible step after C–H cleavage, which is qualitatively consistent with the timescales of the experiments described above with the 3/4 equilibrium (which is established over 1–2 hours in diethyl ether) and the formation of 6 (which takes a few hours).

Fig. 5. Aniline products from amination of arenes with N₂.

(a) Mass spectra of PhN(SiMe₃)₂ products from reaction under 1 atm of natural abundance N₂ (top) and from reaction under 1 atm of ¹⁵N₂ (bottom), showing that the nitrogen atom in the product derives from N₂. (b) Yields (relative to Fe) of N-containing products during N₂-based amination of different arenes, using 5 cycles of Me₃SiBr addition (2 equivalents per cycle).

Other arene substrates were also tested using 5 cycles of Me₃SiBr addition (Fig. 5C), conditions under which the yield of benzene to PhN(SiMe₃)₂ was 68 ± 4% per iron. Toluene gave a mixture of (m-tolyl)N(SiMe₃)₂ and (p-tolyl)N(SiMe₃)₂ in a 3:1 ratio with a total yield of 61 ± 7% per iron. The overall yield is similar to that for benzene, and the ratio of isomers is comparable to that observed in the crystal of LFe(H)(Tol)Na(15c5) (Fig. S54). When o-xylene was used, a 12 ± 2% yield of N,N-bis(trimethylsilyl-3,4-xylidine) was observed. These silylated aniline products could be hydrolyzed to the deprotected anilines with weak aqueous acid when desired. Arenes with easily reducible functionalities such as aryl halides, aryl ethers, and polycyclic aromatics did not give aminated products. Formation of the silylated aniline and N(SiMe₃)₃ also occurred with other Me₃SiX reagents as well (X = Cl, I, and OTf), though the reaction gave the highest yields with Me₃SiBr.

The strategy outlined here differs fundamentally from previously described strategies for formation of C–N bonds from N₂. Typically, carbon electrophiles have been used to create C–N bonds, either from N₂^13,42–46 or nitrides that result from cleavage of N₂^15–17,47. In the new reaction, silylation plays a key role by making the coordinated N₂ sufficiently reactive to accept the migrating aryl group from the metal center. The silylated amines that are formed can be used in further synthetic steps or can be deprotected to the parent anilines using mild aqueous acid. In this method, the C–N bond comes from the migration of a hydrocarbyl from a metal to an N₂-derived group, a strategy that has been used in few stoichiometric C–N bond formation reactions^18,48. The ability of these iron complexes to generate a hydrocarbyl group on the iron through C–H activation and then transfer it to an activated N₂ provides a new tactic for coupling hydrocarbons to N atoms from atmospheric N₂, combining the powers of these bond-cleaving reactions.