|Home | About | Journals | Submit | Contact Us | Français|
Reported N2 complexes of cobalt do not have substantial weakening of the N-N bond. Using diketiminate ligands to enforce three-coordinate geometries, we have synthesized several novel CoNNCo complexes. In formally univalent complexes, cobalt is poorer than iron at weakening the N-N bond, but in formally zerovalent complexes, cobalt and iron give similar N-N weakening. The weakening is due to cobalt-to-N2 π-backbonding, and cations pull more electron density into N2. These results show that the low coordination number of a trigonal-planar geometry is impetus enough to make even the electronegative cobalt weaken the N-N bond of N2.
The binding of N2 to transition metal complexes is a topic of great interest in the chemistry community.1 While the reactivity of N2 complexes does not always correlate with their ground-state properties, it has proven useful to estimate “activation” of the N2 using the N-N bond length and N-N stretching frequency, which are readily measurable. One trend that has emerged is that the “early” transition metals in groups 4−6 often give N2 complexes with weaker, longer N-N bonds, whereas N2 complexes of the “late” transition metals in groups 8−10 have shorter, stronger N-N bonds (dNN < 1.15 Å, νNN > 1900 cm−1).2 For example, all N-N distances in satisfactory (R < 10%) structures of metal-N2 compounds of group 9 metals are less than 1.18 Å, suggesting relatively little activation.3 This trend is consistent with the importance of π-backbonding, where the low-energy d electrons of the electronegative “late” metals are not transferred effectively to the π* orbitals of N2. Paradoxically, biological (nitrogenase) and industrial (Haber-Bosch) catalysts for N2 reduction use iron, a late metal whose isolated N2 complexes typically have little N-N weakening. The α form of N2 on elemental Fe surfaces has an N-N stretching frequency of 1415 cm−1, indicating significant weakening that has never been achieved in a synthetic complex.4 It is not clear which properties of the catalysts enable them to achieve N-N weakening and cleavage.
Recently, we described a series of low-coordinate Fe-N2-Fe complexes that have unprecedented weakening of the N-N bond (1.18−1.23 Å; 1583−1810 cm−1), and explored their electronic structure in detail.5 Experimental and theoretical studies showed that the N2 weakening in these complexes arises from the unusual low-coordinate geometry at iron, which gives high-energy, singly-occupied d orbitals that transfer electron density into the π* orbitals of N2.6 Tetrahedral Fe-N2-Fe complexes can also give long N-N bonds.3g,7 Therefore, it is of interest to learn whether low-coordinate geometries would also enable cobalt, which has lower-energy d orbitals than Fe, to weaken the N-N bond of N2 as well. Here, we report that a dicobalt(I) complex has little N-N weakening ability, but that a formally dicobalt(0) complex has significant weakening of the N-N bond from backbonding.
Treatment of LCoCl (L = 2,2,6,6-tetramethyl-3,5-bis(2,4,6-triisopropylphenylimido)hept-4-yl)8 with one equiv of KC8 gives the bimetallic dinitrogen complex LCoNNCoL,9 and with 2 equiv of reductant gives K2LCoNNCoL (Scheme 1). Reduction with metallic Na gives the analogous sodium salt Na2LCoNNCoL. Each deep purple dinitrogen complex is obtained in a yield of 60−70%, and each has a 7-line 1H NMR spectrum in C6D6 solution at room temperature that is consistent with averaged D2h or D2d symmetry on the NMR time scale. Solid-state magnetic susceptibility measurements show that LCoNNCoL has a quintet (S = 2) ground state, and K2LCoNNCoL is a triplet (S = 1). Interestingly, in each case the CoNNCo group behaves as a single spin system (see Supporting Information for details).10
X-ray crystal structures were determined for each of the new cobalt-dinitrogen complexes. Table 1 compares the CoNNCo cores to the FeNNFe5 and NiNNNi11 analogues. The Co-N bonds (1.8401(8) Å) are longer than the Fe-N bonds (1.771(5) Å), and the N-N bond is significantly shorter in the Co complex than in the Fe complex (1.1390(15) Å for Co vs. 1.189(4) Å for Fe). In the nickel analogue LNiNNNiL, the N-N bond is even shorter (1.120(4) Å). Because of the exact correspondence of ligands, it is possible to conclude unambiguously that LCo and LNi are less effective at weakening the N-N bond than LFe at this oxidation level. We attribute this difference to the lower d orbital energies of Co and Ni, which give weaker backbonding into the π* orbital of N2.
Another difference between the LFeNNFeL and LCoNNCoL is the geometry at the metal. The iron atoms in LFeNNFeL have local C2v symmetry (Y-shaped geometry), with the N2 ligand on the C2 axis of the diketiminate-iron unit. However, in the Co complex, the Co atoms have a significant distortion toward a T shape, with NL-Co-NN2 angles of 101.01(3)° and 162.69(3)°. An independent crystal structure in a different space group, though poorly refining, showed a similar T distortion with N-Co-N angles of 108° and 156°, suggesting that the T distortion is not from crystal packing effects. There is a low barrier for interconversion of the T shapes in solution, because the two aryl groups of the diketiminate ligand are equivalent on the 1H NMR time scale (apparent D2h/D2d symmetry) at temperatures from 205 to 355 K.
In contrast to these differences between formally cobalt(I) and iron(I) species, the formally cobalt(0) compounds with a [LCoNNCoL]2− core have metrical parameters that are similar to their Fe analogues, with N-N distances of 1.21−1.22 Å (Table 1). These N-N distances are similar to the N-N double bond in azobenzene of 1.24(2) Å.12 These are the longest N-N distances in any cobalt dinitrogen complex, including tetrahedral examples.3 They are also significantly longer than the N-N bond in the nickel analogue K2LNiNNNiL (1.185(8) Å).11 Resonance Raman spectra (λex = 406.7 nm) of the [LCoNNCoL]2− complexes in toluene solution show bands below 1600 cm−1 that shift 50−60 cm−1 in the 15N2 isotopomer. The very low N-N stretching frequency confirms the crystallographically observed weakening of the N-N bond. Thus, even though Co is less activating than Fe in LMNNML species, it is as effective as Fe at weakening N2 in [LMNNML]2− compounds. The short Co-N bonds of 1.73−1.75 Å suggest some multiple bond character from π-backbonding.13
Is the N-N weakening in the Co2 dianions due to the additional electron density at cobalt, or from the influence of the alkali metal cations? Because the alkali-metal-free dianions have not been isolable, DFT calculations are used to elucidate the reasons for the exceptional N-N bond weakening. Calculations use a simplified β-diketiminate ligand (L’ = C3N2H5−) with a pure functional and an extended all-electron basis set, BPW91/6−311+G(d). The triplet dianion [L'CoNNCoL']2− optimizes to a Y-shape minimum with Co-N = 1.76 Å, N-N = 1.19 Å, and νNN = 1742 cm−1. This shows that reduction alone gives significant N-N stretching, but not as much as the observed compounds. Interestingly, geometry optimization of triplet K2L'CoNNCoL’ gives a model that has excellent agreement with the experimental metrical data, with Co-N = 1.74 Å (expt. 1.75), N-N = 1.22 Å (expt. 1.22), and νNN = 1603 cm−1 (expt. 1599). The presence of the potassium ions heightens the electron transfer from Co to N2; Natural Bond Order (NBO) analyses show that the N-N bond order changes from 2.05 to 1.82 upon incorporation of the K+ ions. Clearly, both reduction and alkali metal binding work in concert to weaken N2, whereby the positively charged alkali metal pulls electrons into N2.
Thus, iron and cobalt have different abilities to weaken N2 in formally monovalent compounds, and similar abilities in formally zerovalent compounds. In neutral LMNNML, Fe is significantly better than Co at back-bonding into the π* orbitals of N2. Reduction of LCoNNCoL by two more electrons gives the first Co complex with a significantly weakened N2 ligand, and this NN bond weakening is similar between the formally zero-valent complexes (alkali)2[LMNNML] with Fe and Co. These results demonstrate that the coordination geometry and presence of alkali metals can have a major impact on the reduction level of bound N2, even with a poorly backbonding metal like cobalt.
The authors acknowledge a gift of 15N2 from Cambridge Isotopes, and financial support from the NIH (GM065313 to P.L.H., AI072719 to K.R.R.) and the NSF (CHE-0112658 to P.L.H., CHE-0701247 to T.R.C.).