Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Science. Author manuscript; available in PMC 2012 November 11.
Published in final edited form as:
PMCID: PMC3218428

N2 Reduction and Hydrogenation to Ammonia by a Molecular Iron-Potassium Complex


The most common catalyst in the Haber-Bosch process for the hydrogenation of dinitrogen (N2) to ammonia is an iron surface promoted with K+, but soluble iron complexes have neither reduced the N-N bond of N2 to nitride nor produced large amounts of NH3 from N2. We report a molecular iron complex that reacts with N2 and a potassium reductant to give a complex with two nitrides, which are bound to iron and potassium cations. The product has a Fe3N2 core, implying that three iron atoms cooperate to break the N-N triple bond through a six-electron reduction. The nitride complex reacts with acid and with H2 to give substantial yields of N2-derived ammonia. These reactions, though not yet catalytic, give structural and spectroscopic insight into N2 cleavage and N-H bond-forming reactions of iron.

The Haber-Bosch process is a large-scale method for catalytic reduction of dinitrogen (N2) with dihydrogen (H2) to give ammonia (NH3). This process is vital for production of synthetic fertilizer, which is needed to produce food for the world's expanding population (1). Several metals catalyze the Haber-Bosch process (iron, ruthenium, osmium, uranium and cobalt-molybdenum), but iron has received the most industrial and theoretical attention due to its hardiness and low cost (24). The most common iron catalyst is promoted by K+ additives, and was developed by Mittasch in 1909–1910 (2). In the century since then, chemists have studied the nature of active iron and iron/potassium surfaces, as well as the kinetics of N2 reduction (3). These studies have shown that iron is predominantly in the zero oxidation state in the catalyst. Importantly, the rate-limiting step of the catalytic process is N2 chemisorption and N-N bond cleavage to give surface-bound nitrides (N3−), which react with H2 to form the N-H bonds in NH3 (Fig. 1). However, many details of the bond-breaking and bond-forming steps are difficult to discern using surface science techniques: for example, how many iron atoms are involved in the N-N dissociation step, and whether the K+ promoter interacts with N2 during the N-N dissociation step (4). Similar uncertainties surround postulated mechanisms for N2 reduction by nitrogenase enzymes, which have iron (in the +2 and +3 oxidation states) at their active sites (5, 6).

Fig. 1
Simplified scheme of the ammonia formation pathway in the Haber-Bosch process.

Reactions of iron coordination compounds with N2 are of interest because the products can shed light on iron-N2 interactions in atomic detail. Though the behavior of iron atoms in a coordination compound (where they have a positive formal charge) is inevitably different than the zerovalent iron atoms on a metallic surface, coordination compounds are more amenable to detailed structural characterization. Unfortunately, iron coordination complexes are poor at reducing N2 (7, 8). Known iron-N2 complexes have not been observed to give iron-nitride products, and produce little (if any) ammonia upon exposure to acids or H2 (913). Here, we report an iron/potassium system that completely cleaves the N-N triple bond of N2 to give a product with two nitride (N3−) ligands derived from N2. The N2-derived nitride complex reacts with H2 to give a substantial yield of ammonia. This work displays the structural characteristics of an iron-based N2-cleaving system, and suggests that cooperation between metal atoms helps to achieve N-N bond cleavage and N-H bond formation on iron.

Consistent with their low reactivity toward N-N bond cleavage, the N-N bonds in most mononuclear and dinuclear iron-N2 complexes are within 0.04 Å of the N-N distance in free N2 (1.10 Å) (7, 8). A few research groups have reported significant N-N bond weakening in iron-N2 complexes by using bulky, electron-rich ligands with N or P donors (1517). Our research has revealed that iron-N2 complexes with low-coordinate iron weaken the N-N bond to between 1.17 and 1.23 Å (14, 15). The greatest weakening (N-N = 1.23 Å; νN-N = 1589 cm−1) is in β-diketiminate complex K2[L1FeNNFeL1] (L1 = HC[C(tBu)N(2,6-iPr2C6H3)]2; Fig. 2A), which comes from treating L1FeCl with potassium graphite (KC8) under N2 (15). Similar N-N weakening is observed in analogues containing the slightly less hindering L2 (L2 = HC[C(Me)N(2,6-iPr2C6H3)]2; Fig. 2B). The research reported here uses L3 (L3 = MeC[C(Me)N(2,6-Me2C6H3)]2; Fig. 2C), which is less sterically hindering by virtue of having methyl rather than isopropyl groups on the aryl rings. The new iron(II) chloride complex [L3Fe(μ-Cl)]2 (1) can be prepared like earlier β-diketiminate-supported iron(II) chloride complexes (18, 19).

Fig. 2
Iron complexes with varying steric bulk give different N2 products upon reduction by potassium (introduced as KC8, potassium graphite). A) Bulky ligand L1 (14); B) Bulky ligand L2 (15); C) Less bulky ligand L3 (this work). tBu = tert-butyl.

Mixing 1 with two molar equivalents of KC8 in tetrahydrofuran (THF) under an atmosphere of N2 gives a red-brown tetrairon bis-nitride complex (2) (Fig. 2C). This product can be isolated as analytically pure crystals in 43% yield. The use of THF as reaction solvent is critical: formation of 2 is not observed by proton nuclear magnetic resonance (NMR) spectroscopy when the reaction is performed in other solvents such as pentane and toluene. No formation of 2 is observed when 1 is treated with KC8 under an atmosphere of argon rather than N2; instead, a mixture of unidentified products is evident. Use of 15N2 gives an isotopically labeled product (see below), demonstrating that the nitrides originate from N2. Solutions of purified 2 decompose to unidentified products with a half-life of 8 hours at 25 °C in benzene-d6, though crystalline samples of 2 suffer no apparent decomposition when stored for up to one month at −40 °C.

X-ray crystallography shows the molecular structure of 2 (Fig. 3), which has a core of three iron centers (Fe1, Fe2, Fe3) surrounding two bridging nitrogens. The N-N distance of 2.799(2) Å indicates that there is no N-N bond; thus six-electron reduction of N2 has given two bridging nitrides (N3−). One of the nitrides (N2) is attached to three irons, and the other (N1) bonds to two irons as well as two potassiums. Each potassium interacts with two chlorides and with a diketiminate aryl group through a π-cation interaction. A fourth iron center (Fe4) is bound to the chlorides, and displays a pseudotetrahedral geometry. The Fe-N distances to the potassium-bound N1 are shorter than the Fe-N distances to the triply-bridging N2, most likely from steric clashes between the three β–diketiminate ligands bound to Fe1, Fe2, and Fe3.

Fig. 3
Molecular structure of 2 using 50% probability ellipsoids. Hydrogen atoms and cocrystallized solvent molecules have been omitted for clarity. Selected bond lengths (Å) and angles (°): Fe1-N1, 1.812(2); Fe1-N2, 1.906(2); Fe2-N1, 1.809(2); ...

NMR spectroscopy suggests that the solution structure of 2 is similar to the solid-state structure. Fifteen paramagnetically shifted resonances are observed in the 1H NMR spectrum of 2, and the number and integrations of the resonances are consistent with C2v symmetry in solution, with equivalent β-diketiminate ligands on Fe1 and Fe2. The solubility of 2 in nonpolar solvents like hexane and diethyl ether also suggests that the potassium cations are held tightly within the structure in solution.

The zero-field Mössbauer spectrum of a crystalline sample of 2 at 80 K (Fig. 4A) indicates three different iron environments, and the isomer shifts (δ) can be used to gauge their oxidation states by comparison to literature values (Table S2). The isomer shift for subspectrum (I) (δ = 0.29 mm/s) is similar to those for high-spin Fe3+ sites in nitride and sulfide clusters (2022). Subspectrum (I) has twice the intensity of the other subspectra, indicating that the equivalent Fe1 and Fe2 sites are high-spin Fe3+. Subspectra (II) and (III) of 2 exhibit higher isomer shifts that are typical of high-spin Fe2+ sites. The Mössbauer parameters of subspectrum (II) (δ = 0.68 mm/s, ΔEQ = 1.54 mm/s) resemble those for the planar three-coordinate Fe2+ complexes L1FeNHR (R = p-tolyl and tert-butyl), which have a similar environment as Fe3 (23). Subspectrum (III) shows parameters (δ = 0.96 mm/s, ΔEQ = 1.80 mm/s) that are typical of quasi-tetrahedral high-spin Fe2+ compounds such as L2Fe(μ-Cl)2Li(ether)2 (24). Comparison of the Fe-N(diketiminate) bond lengths in Fe4 to Fe2+ and Fe3+ compounds of L3 also suggests a Fe2+ assignment for Fe4 (Table S1). Thus signal (II) can be confidently assigned to Fe3, and signal (III) to Fe4.

Fig. 4
A) Zero-field Mössbauer spectrum of solid 2 at 80 K. The red line is a superposition of the three Lorentzian doublets (I), (II), and (III) with intensity ratio 2:1:1. B) Temperature dependence of the effective magnetic moment of 2 measured at ...

The assignments of high-spin Fe3+ for Fe1 and Fe2, and of high-spin Fe2+ for Fe3 and Fe4, are corroborated by the magnetic properties of solid 2. The temperature dependence of the magnetic moment (Fig. 4B) can be simulated using the spin topology indicated in the inset of Fig. 4B. The triad made up of Fe1, Fe2, and Fe3 has ground state S123 = 3 because of dominant antiferromagnetic coupling of S3 to both S1 and S2, which is stronger than the antiferromagnetic coupling between S1 and S2. The effective magnetic moment of 2 fits the superposition of the moments from S123 and from the uncoupled ferrous Fe4 site, with μeff2 = μeff,1232 + μeff,42 (25).

The yield of ammonia derived from N2 upon addition of acid is a measure of N2 activation. Some carefully designed molybdenum compounds can produce ammonia catalytically with acids (26, 27), but previous Fe-N2 complexes have not given more than 10% yield of ammonia, indicating weak activation of coordinated N2 (713, 28). In contrast, treatment of a THF solution of 2 with 100 molar equivalents of ethereal HCl gives ammonia (present as its conjugate acid, NH4+) in 82 ± 4% yield. Control experiments with other complexes of L3, or N2 complexes of other diketiminate ligands, give no detectable amount of ammonia. Reaction of 2 with acid to give a high yield of ammonia demonstrates that the N-N bond is completely cleaved and that the nitrides are nucleophilic.

The reaction of 1 with KC8 may be carried out under an atmosphere of 15N2 to generate isotopically labeled samples of 2-15N. Purified 2-15N reacts with acid to form 15NH4+, which may be identified from its characteristic 1:1 doublet in the 1H NMR spectrum. An analogous reaction with unlabeled 2 and excess acid shows a 1:1:1 triplet for 14NH4Cl. These labeling studies verify that the nitrides in 2 come from N2, and that the ammonia comes from these nitrides.

Iron-nitride complexes typically have terminal and bridging nitrides that are derived from reactive N sources such as N3 (22). Recently reported iron clusters with triply bridging nitrides use N(SnMe3)3 as the nitride source (20, 21). The iron system reported here generates nitrides instead from the six-electron reduction of N2, and we suggest two possible explanations for its N2-reducing ability. First is the influence of the K+ generated from reduction of the Fe2+ ions in 1 by potassium graphite. On metallic iron-potassium catalysts, it is thought that surface K+ promotes N2 dissociation by pulling more electron density toward the surface iron atoms (3). The interaction of K+ with the nitrides in our synthetic compound suggests that direct K+-N2 interactions could also be considered as contributors to the potassium promotion effect. Second, since more hindered L1Fe and L2Fe gave bimetallic Fe2N2 species without N-N cleavage (14, 15), and the less protected L3Fe gives N-N cleavage in 2, it is likely that the ability of more metal atoms to simultaneously access the N2 molecule is important for N2 cleavage. From a thermodynamic perspective, it is reasonable that the formation of more iron-nitrogen bonds can better balance the energetic cost of breaking the N-N triple bond. On iron metal, the surfaces that are most active for N2 reduction are those with sites that expose subsurface atoms (29). N2 bound at such a site may interact in a cooperative fashion with several iron atoms to facilitate N-N cleavage, as suggested by computational studies (30).

It is most relevant to the Haber-Bosch process to show that formation of ammonia from an iron-nitride is possible through reaction with dihydrogen (H2). When electropositive early transition metal complexes cleave N2 to metal-nitride complexes, the products rarely react with H2 to give ammonia. In the best previous example, reaction of free H2 with a well-characterized N2-derived zirconium nitride complex formed ammonia in 10 to 15% yield (31). Shaking a solution of 2 in toluene with 1 atm of H2 at room temperature for 6 hours produces NH3 in 42 ± 2% yield (Eq. 1). The main iron-containing product of the reaction of 2 with H2 is [L3Fe(μ-H)]2, which is generated in 85 ± 8% yield as shown by 1H NMR spectroscopy. Thus, the soluble molecular iron system described here is capable of performing both N2 cleavage and hydrogenation.

equation image

How do these results relate to the Mittasch iron catalyst that is used for the Haber-Bosch process? It is likely that potassium reduction of the Fe2+ precursor 1 leads to a high-spin Fe1+ intermediate, by analogy to related complexes (15). The metal in a low oxidation state can donate charge to N2 that weakens the N-N bond, and this effect is heightened by a low coordination number at the metal (15). Surface irons on the Mittasch catalyst also have a low oxidation state and low coordination number (3), suggesting a potential parallel. In a second commonality, a molecule of N2 may react simultaneously with multiple unsaturated iron atoms during the formation of 2 and on an iron surface. Nitrogenase uses a cluster of iron atoms to reduce N2 (5,6), suggesting that cooperativity may be important in biological N2 reduction as well. Thirdly, potassium has a beneficial effect on N-N bond cleavage by the Mittasch catalyst, perhaps through a direct interaction between potassium and nitrogen like that observed in 2. Finally, we suggest that the presence of a three-coordinate iron atom near the nitrides may enable iron to split H2 at the unsaturated site and form N-H bonds without extensive reorganization. Further studies will test these hypotheses derived from the structure and behavior of 2.

Supplementary Material



This work was supported by a grant from the National Institutes of Health (GM-065313). We thank Christopher Scarborough and Ryan Cowley for valuable discussions. Materials and methods are available as supporting material on Science Online. Crystallographic data for 2 (two independent structures with different solvents of crystallization), L3FeCl2, L3FeCl2K(18-crown-6), and [L3FeH]2 have been deposited with the Cambridge Crystallographic Data Centre with deposition numbers CCDC 836870-836874.


Publisher's Disclaimer: "This manuscript has been accepted for publication in Science. This version has not undergone final editing. Please refer to the complete version of record at The manuscript may not be reproduced or used in any manner that does not fall within the fair use provisions of the Copyright Act without the prior, written permission of AAAS."

Notes and References

1. Smil V. Sci. Am. 1997;277:76.
2. Mittasch A. Geschichte der Ammoniaksynthese. Weinheim: Verlag Chemie; 1951.
3. Schlögl R. Handbook of Heterogeneous Catalysis. 2nd Ed. vol. 5. Weinheim: Wiley-VCH; 2008. pp. 2501–2575.
4. Hellman A, et al. J. Phys. Chem. B. 2006;110:17719. [PubMed]
5. Burgess BK, Lowe DJ. Chem. Rev. 1996;96:2983. [PubMed]
6. Hoffman BM, Dean DR, Seefeldt LC. Acc. Chem. Res. 2009;42:609. [PMC free article] [PubMed]
7. Crossland JL, Tyler DR. Coord. Chem. Rev. 2010;254:1883.
8. Hazari N. Chem. Soc. Rev. 2010;39:4044. [PubMed]
9. Schrauzer GN, Guth TD. J. Am. Chem. Soc. 1976;98:3508.
10. Hills A, Hughes DL, Jimenez-Tenorio M, Leigh GJ, Rowley AT. J. Chem. Soc., Dalton Trans. 1993:3041.
11. George TA, Rose DJ, Chang Y, Chen Q, Zubieta J. Inorg. Chem. 1995;34:1295.
12. Hall DA, Leigh GJ. J. Chem. Soc., Dalton Trans. 1996:3539.
13. Gilbertson JD, Szymczak NK, Tyler DR. J. Am. Chem. Soc. 2005;127:10184. [PubMed]
14. Smith JM, et al. J. Am. Chem. Soc. 2001;123:9222. [PubMed]
15. Smith JM, et al. J. Am. Chem. Soc. 2006;128:756. [PubMed]
16. Betley TA, Peters JC. J. Am. Chem. Soc. 2004;126:6252. [PubMed]
17. Chomitz WA, Arnold J. Chem. Commun. 2007:4797. [PubMed]
18. Eckert NA, Smith JM, Lachicotte RJ, Holland PL. Inorg. Chem. 2004;43:3306. [PubMed]
19. Materials and methods are available as Supporting Online Material (SOM) on Science Online.
20. Bennett MV, Stoian S, Bominaar EL, Münck E, Holm RH. J. Am. Chem. Soc. 2005;127:12378. [PubMed]
21. Bennett MV, Holm RH. Angew. Chem. Int. Ed. 2006;45:5613. [PubMed]
22. Additional references may be found in the SOM.
23. Andres H, et al. J. Am. Chem. Soc. 2002;124:3012. [PubMed]
24. Stoian SA, Smith JM, Holland PL, Münck E, Bominaar EL. Inorg. Chem. 2008;47:8687. [PubMed]
25. The g values were fixed to g = 2 for the sake of unambiguity except for g4, which was allowed to vary to 2.49 in order to match exactly the experimental values. See SOM for details and other fit values.
26. Yandulov DV, Schrock RR. Science. 2003;301:76. [PubMed]
27. Arashiba K, Miyake Y, Nishibayashi Y. Nat. Chem. 2011;3:120. [PubMed]
28. The yields given here are per nitrogen atom of N2, in order to accurately reflect the fate of the two nitrogen atoms. In previous papers with mononuclear iron-N2 complexes, the yield of NH3 is typically calculated per iron atom.
29. Strongin DR, Carrazza J, Bare SR, Somorjai GA. J. Catal. 1987;103:213.
30. Mortensen JJ, Hansen LB, Hammer B, Nørskov JK. J. Catal. 1999;182:479.
31. Pool JA, Lobkovsky E, Chirik PJ. Nature. 2004;427:527. [PubMed]