Dinitrogen as ligand for coordination complexes

Beyond the heterogeneous nature of the N2-activating compounds in the HB process, lies N2 -fixation on molecular defined complexes. A first characterised example of dinitrogen as ligand was found by Allen and Senoff in 1965, in form of [Ru(NH3)5(N2)]²⁺, see Figure 2.¹¹ As highlighted in interesting background narratives by Leigh and Jones,^12,13 the original purpose of the work was to synthesize the hexa-ammine ruthenium complex via reaction of [RuCl3(H2O)3] and hydrazine as reduction agent and ammonia source. Yet without the addition of an ammonium salt, the penta-ammine dinitrogen complex was formed, where the N2 originates from hydrazine.

Especially the appearance of a stretching frequency at 2100 cm^-1 by IR spectroscopy was indicative for coordinated N2, even though it was first interpreted as a metal-hydride. Later, it was shown that the N2 ligand can also originate from atmospheric uptake via reduction of a Ru(III) precursor under a dinitrogen atmosphere. The original work was welcomed with a lot of criticism, that was first tempered by the publication of the crystal structure of [Ru(NH3)5(N2)]Cl2. This showed a N-N bond length of 1.12 Å, which is only moderately activated compared to free N2 (1.10 Å, see Table 1).¹⁴

Figure 2. Left: first characterised N2-bound complex by Allen.¹¹ Right: qualitative σ- and π-interactions in an end-on M-N2 fragment.

Thereafter, and with the realisation how to identify a possible N2-complex, many complexes were developed. The metal-dinitrogen interaction relies on a σ-donation from N2 to the metal, and a π-backdonation from the metal into the π*-orbital of N2 (Figure 2). Both interactions will decrease the bond-order in dinitrogen. The reminiscence with a carbonyl ligand is evident and from the early complexes the same basic rules for N-N bond elongation were found, i.e. a shorter N-N bond is found upon oxidation of the metal centre or the coordination of π-accepting ancillary ligands, and vice versa. The degree of activation can be expressed by N-N bond lengths or stretching frequencies, as listed for various oxidation states of N2 in Table 1. Weak orbital overlap between N2 and the metal only results in weak bond activation compared to CO, as reflected in typical N-N bond lengths in M-N2 complexes usually in the range of 1.10-1.12 Å.¹⁵

Table 1. Bond lengths and stretching frequencies of coordinated N2 and free N2, H2N2 and H4N2.¹⁶

A higher degree of activation can be reached when a multi-metallic approach is pursued. Within this strategy, the coordination modes of N2 are various, and an overview of common binding motives is given in Figure 3. Evolving from mononuclear, end-on coordination, the binding to a second metal centre leads to an end-on N2-bridge (= μ-η¹:η¹), which is a well-known motif.

Additionally, dinitrogen can be coordinated in a side-on bridging μ-η²:η² mode, or a mixed side-on end-side-on μ-η¹:η² mode. Considering the relevance for the research as performed within this work, the focus in the next sections will be on end-on bridging complexes.

Figure 3. Common binding motives for dinitrogen as ligand.

Continuing from [Ru(NH3)5(N2)]²⁺, Taube and co-workers isolated in 1969 the first example of a bimolecular end-on N2-complex, in form of [{Ru(NH3)5}2(μ-η¹:η¹-N2)]⁴⁺,¹⁷ which was right thereafter crystallised by Gray and co-workers.¹⁸ Many N2-bridging complexes followed and less

Free N≡N 1.10 Å 2331 cm^-1

than ten years later, Chatt reviewed the over 30 (more or less well characterised) examples reported within that short time frame.¹⁵ More recent reviews give an overview of the circa 50 years of research.^1,19,20 To rationalise the bonding situation of these dinuclear end-on complexes, the orbital interactions between the metal and the N2-ligand where considered, as initiated by Gray and Richards.^18,21 As frontier orbitals, a π-orbital dominated manifold was found, based on the interaction between the metal dxz/dzy and the N2 π/π*-orbitals, see Scheme 1. This constructs four MO’s ranging from fully bonding (1eu, πππ), to fully anti-bonding (2eg, π*π*π*), with respect to the {MNNM}-manifold. A second degenerate set is oriented orthogonally. Within the same energy range, the combination between the metal dz2- and the N2 σ*-orbital to a σσ*σ MO is found, which has a strong NN anti-bonding character and will be relevant for N2-splitting (vide infra). If the ancillary ligand arrangement on the metal centre is threefold, the ligand-metal interactions raise the energy of the remaining metal orbitals (dxy and dx2-y2). In a fourfold coordination, the orthogonal ancillary ligands interact different with the d-orbitals, and the dxy

orbitals appear within the relevant energy level for the {MNNM} manifold. Without finding interaction with N2, they remain non-bonding, metal centred orbitals, as highlighted in green in Scheme 1. It was found by calculations that the fully π-bonding set (1eu) is mainly N2-centred and the ππ*π set (2eu) is mainly metal-centred in nature. Depending on the nature of the ancillary ligands or the metal centres, the energy levels or degeneracy of the orbitals as sketched in Scheme 1 can be different.

Scheme 1. Schematic qualitative orbital interaction scheme for end-on dinuclear {MNNM} fragments, as originating between N2 and two ‘ML3’ fragments in a threefold coordination (left) or N2 and two ‘ML4/5’ fragments in a fourfold coordination (right), where L represents an ancillary ligand. The orbitals are drawn with the following colour code: π orbitals (black, only the set in the y,z-direction is visualised), δ non-bonding (n.b.) orbitals (green) and σ orbital (blue).

Based on the metal valence d-electrons and the population of the orbitals, this model can be used to rationalise the N-N bond orders as found by XRD or vibrational spectroscopy. For pioneering Ru^II [{Ru(NH3)5}2(μ-η¹:η¹-N2)]⁴⁺, the N-N bond length (1.124 Å) is barely activated, compared to free N2 (Table 1). This is in line with occupation of 16 available electrons (4 from the dinitrogen, and two times 6 valence electrons from Ru^II) up to and including the 2eu set (in a so-called ‘δ⁴π¹²’ configuration). Since this latter π-orbital is N-N bonding in nature, its population will strengthen the N-N bond. A contrary example is the paramagnetic heteronuclear Re^I/Mo^V [{ReCl(PMe2Ph)4}{MoCl4(OMe)}(μ-η¹:η¹-N2)] as found by the group of Richards, with an elongated N-N bond of 1.18(3) Å. This elongation is in line with its δ³π⁸ configuration, where the N2 π-bonding orbital π*ππ* remains unpopulated.^22,23

In addition to these isolated examples that illustrate the use of this MO description to rationalise the N-N bond metrics, various redox-series of end-on N2-bridged complexes were prepared, of which some will be discussed herein. Dinuclear [{Mo(NArtBu)3}2(μ-η¹:η¹-N2)] (Ar = 3,5-dimethylphenyl) is the pioneering example for N2-bond splitting into terminal nitrides (vide infra), as characterised by Cummins. Cyclic voltammetry (CV) of this compound suggested synthetic access to both mono- and di-cationic congeners at mild potentials and both complexes were synthesised (Figure 4, left). Upon oxidation, the N-N bond order decreases as shown by XRD and vibrational spectroscopy. This initially seems a paradox: in a classical mononuclear binding situation electron removal decreases the metal to ligand backdonation, which strengthens the N-N bond. Yet, by considering the π¹⁰ configuration within the {MNNM}

manifold of the neutral complex, it gives a coherent picture: subsequent removal of electrons from the N-N bonding π*ππ*-orbital to a π⁹ and π⁸ configuration gradually elongated the NN bond. In addition, both the absence of a N2-stretching band in the IR spectrum or a typical mixed-valent electronic absorption band for [{Mo(NArtBu)3}2(μ-η¹:η¹-N2)]⁺, highlights the high covalency within the {MNNM} core for this S = ½ compound.²⁴

Another example of a redox-series comes from our group: [{WCl(PNP)}2(μ-η¹:η¹-N2)] was characterised by Dr. B. Schluschass within our group, assigned to a δ⁴π⁸ configuration. Electron removal towards either the one- or two-fold oxidised products has no influence on the dinitrogen bond order as the bond length and stretching frequency are basically invariant within this series (see Figure 4, middle). The HOMO orbital of [{WCl(PNP)}2(μ-η¹:η¹-N2)] consist of the metal-based δ-orbitals, and oxidation to the δ³π⁸ andδ²π⁸ configuration does indeed not interfere with the {MNNM} core. For the di-cationic product, two weakly anti-ferromagnetically coupled S=½ ions indicate a δ¹δ¹π⁸ configuration, as substantiated by SQUID magnetometry and DFT calculations.²⁵ Recently, the analogue Mo series was isolated and characterised, showing the

same invariant behaviour. When going from W to Mo, the potentials are circa 0.25 V milder to access these oxidised analogues highlighting the decreased reducing properties of Mo vs. W.²⁶

Figure 4. Selected end-on bridging N2-complexes of Mo, W and V and their N-N bond order characteristics (bond length by XRD and stretching frequency by Raman spectroscopy).^24–27

Floriani decided to apply pure σ-donating carbon-based ancillary ligands; the lack of metal to ligand back-bonding was envisioned to access electron-rich metal centres favoured for N2 -fixation. Their vanadium based end-on dinuclear complex [{VMes3}2(μ-η¹:η¹-N2)] could be isolated as neutral, mono-anionic and di-anionic series (Figure 4, right). Unfortunately, no crystal structure of the neutral compound was obtained, but the one or two-fold reduction has virtually no effect on the N-N bond. DFT corroborated that the LUMO orbitals proved low-lying δ-symmetric metal-based orbitals, and their population upon reduction does not affect the N2 -bond metrices. The neutral complex has interesting magnetic behaviour: the low-lying vacant δ-orbitals mix into the singlet (π⁸) ground state causing temperature-independent paramagnetism (TIP).^27–29

Figure 5. An end-on bridging N2 complex from Chirik in different redox states and its N-N bond metrics upon oxidation (left), the ligand-based mono-reduced compound (top right), and the corresponding scheme of the frontier orbitals (bottom right).³⁰

The work by Chirik represents a remarkable example of an end-on bridging N2 complex with a redox-active ligand, which is stable over five redox states (Figure 5). The complex [{Mo(PPh2Me)2(^PhTpy)}2(μ-η¹:η¹-N2)]²⁺ (^PhTpy = 4’-Ph-2,2’:6’,2’’-terpyridine) shows a moderately activated N2-bridge (1.203(2)Å), in line with the population of the π*ππ* orbital in a δ⁴π¹⁰ configuration. Due to interactions with the π-manifold of the terpyridine ligand, which is in the similar energy range, the degeneracy of the π*ππ* -orbitals is lifted. Oxidation to the tris-cationic complex (δ⁴π⁹) leads to a more activated N-N bond. Yet, the subsequent oxidation does not induce further bond elongation, as one of the δ-orbitals is raised in energy and a S=1 ground state is found for [{Mo(PPh2Me)2(^PhTpy)}2(μ-N2)]⁴⁺ with a δ³π⁹ configuration (see Figure 5).

Reduction of the two-fold oxidised compound populates the LUMO, which has mainly terpyridine π* character with a small contribution of the ππ*π set. Consequently, the first reduction is accompanied by a slight decrease in N-N bond order (f(N2) = 1530 cm^-1), yet the electron density is mainly located on the ligand. The two-fold reduced neutral species was synthesised, but unfortunately its N-N bond metrics could not be characterised. It is a diamagnetic species, in analogy to the MO-scheme in Figure 5.³⁰

Sita prepared an extensive dataset of dinuclear N2-bridging complexes with various early transition metals all with virtually the same ligands, from which the metal influence on the N2

bond can be examined. For M = Ti, V, Nb, Ta, Mo, and W, end-on N2-bridging complexes were found of the general formula [{M(Cp*)(NiPrC(Me)NiPr)}2(μ-η¹:η¹-N2)].^31–33 As shown in Figure 6, all N-N bonds are elongated and are within the diazenido (N2)^2- range. Within a group (e.g. V – Nb – Ta of group 5) a gradual elongation of the N-N bond is observed, as assigned to the more

electron donating properties for the heavier metals. Notably, for M = Zr and Hf, a side-on coordination mode was found [{M(Cp*)(NiPrC(NMe2) NiPr)}2(μ-η²:η²-N2)],³⁴ attributed to the larger covalent radius for these metals (≈ 1.75 Å (Zr/Hf)). The N-N bond for these compounds is extraordinarily elongated, and upon decreasing the steric demand on the amidinate substituents, N-N bonds up to 1.635(5) Å are found. Therefore, these complexes are best described as two Zr^IV/Hf^IV with a bridging (N2)^4- moiety. The contrasting description of two Ti^III bridging a diazenido ligand for the first-row analogue is attributed to its low oxidation potential inhibiting the Ti^IV state.

Figure 6. Dinuclear N2-bound series of early transition metals by Sita and co-workers. For M = Mo, the C of the amidinate bears a dimethylamide substituent. For M = Nb, the C of the amidinate bears a phenyl substituent.^31–34

Controlling the coordination mode of N2 upon ligand-exchange was found by Fryzuk, for the side-on N2-bridging [{ZrCl(N(SiMe2CH2PiPr2)}2(μ-η²:η²-N2)]. When replacing the ancillary chloride ligand for a cyclopentadienyl ring, a rearrangement takes place to the end-on N2 -bridging analogue (Scheme 2). The latter is associated with a π⁸ configuration, in agreement with a significant elongated N-N bond (1.301(3)Å). These different bridging modes were rationalised by considering the different MO-schemes for side-on versus end-on bridging. For the side-on bridging mode, interaction between the metal d- and the N2 π*-orbitals result in a π- and δ-symmetric MO (in contrast to two π δ-symmetric MOs for end-on bridging, see Scheme 2). When a Cp-ligand is coordinated, the required metal d-orbital for this δ-symmetric orbital is involved in the bond with the Cp ring instead, which enforces an end-on N2-bridging mode where this d-orbital remains metal-centred. In order to rule out that the altered steric shielding play a role, a similar compound with a smaller metal (Ta) and a smaller ancillary ligand (neopentylidene) is synthesised. This compound is assumed to also adopt the end-on mode, yet without the indicative prove of XRD analysis.^35,36

Scheme 2. Side-on to end-on bridging N2 isomerisation upon ligand exchange by Fryzuk.^35,36

Bercaw and Chirik systematically addressed increased steric demand on the same ligand. It was found that different N2-bridging modes were adapted depending on the number of Me-substituents on the ancillary Cp-ligand. [{Zr(Cp*)2(N2)2}2(μ-η¹:η¹-N2)] and [{Zr(Cp*)(Cp’)(N2)2}2(μ-η¹:η¹-N2)] (Cp’ = C5Me4H⁻) are in an end-on N2-bridging mode, whereas [{Zr(Cp’)2(N2)2}2(μ-η²:η²-N2)] bridges N2 in a side-on coordination (Scheme 3). Computations suggested that the side-on coordination is in principle preferred. Yet, when more than 4 methyl groups are present at the Cp ring, the end-on mode is enforced because of steric repulsion (in the end-on mode the ancillary ligands are substantially further apart). All complexes adapt a twisted structure with regard to the four Cp’^/* rings, that allow both appropriate (orthogonal) metal d-orbitals to overlap with the orthogonal π*-orbitals of N2, which induces a stronger bond activation of the latter.^37–41

Scheme 3. Dinuclear zirconium N2-complexes by Bercaw and Chirik with Cp-ligands bearing from left to right reduced steric demand; the successive -Me for -H replacement is highlighted in orange.^38–41