Catalytic Dinitrogen Reduction to Ammonia at a Triamidoamine-Titanium Complex.

Catalytic reduction of N2 to NH3 by a Ti complex has been achieved, thus now adding an early d-block metal to the small group of mid- and late-d-block metals (Mo, Fe, Ru, Os, Co) that catalytically produce NH3 by N2 reduction and protonolysis under homogeneous, abiological conditions. Reduction of [TiIV (TrenTMS )X] (X=Cl, 1A; I, 1B; TrenTMS =N(CH2 CH2 NSiMe3 )3 ) with KC8 affords [TiIII (TrenTMS )] (2). Addition of N2 affords [{(TrenTMS )TiIII }2 (μ-η1 :η1 -N2 )] (3); further reduction with KC8 gives [{(TrenTMS )TiIV }2 (μ-η1 :η1 :η2 :η2 -N2 K2 )] (4). Addition of benzo-15-crown-5 ether (B15C5) to 4 affords [{(TrenTMS )TiIV }2 (μ-η1 :η1 -N2 )][K(B15C5)2 ]2 (5). Complexes 3-5 treated under N2 with KC8 and [R3 PH][I], (the weakest H+ source yet used in N2 reduction) produce up to 18 equiv of NH3 with only trace N2 H4 . When only acid is present, N2 H4 is the dominant product, suggesting successive protonation produces [{(TrenTMS )TiIV }2 (μ-η1 :η1 -N2 H4 )][I]2 , and that extruded N2 H4 reacts further with [R3 PH][I]/KC8 to form NH3 .

The conversion of dinitrogen, N2, into ammonia, NH3, is essential for supplying N2 in a fixed form into the Earth's biosphere,1 and key to providing NH3 to chemical industry on a vast scale.2 However, the N≡N triple bond, with a bond strength of 944 kJ mol−1, is one of the strongest chemical bonds known, and with a high ionisation potential, negative electron affinity, poor nucleophilicity and electrophilicity, large HOMO–LUMO gap, and no permanent dipole, there are major kinetic and thermodynamic barriers to activating N2, and thus to fixing it as NH3.3, 4

Nature uses homogeneous nitrogenases based on V, Mo, and Fe to execute multiple single‐electron transfer and protonation steps to convert N2 into NH3.5 In contrast, chemical industry uses N2 and H2 to produce NH3 over heterogeneous catalysts in the Haber–Bosch process.6 However, each process is energy‐intensive, reflecting the challenge of N2 activation, and so there is interest in studying reactivity at molecular complexes to improve our understanding of these elementary transformations.7

After the report that a molybdenum–triamidoamine complex catalytically reduces N2 to NH3 in the presence of H+/e−,8 a few Mo, Fe, Ru, Os, and Co complexes have been shown to be catalytically competent in H+/e− mediated N2 reduction cycles,9 and stoichiometric reduction and protonation/hydrogenation of N2 and nitrides by a variety of metals have been reported.10 However, no abiological, early metal complex preceding Group 6 has ever been shown to catalytically convert N2 into NH3, and even V‐, Cr‐, and W‐triamidoamine complexes11 do not facilitate N2 reduction/protonolysis to NH3 like Mo analogues.8 Where Ti is concerned, cleavage of N2 by molecular polyhydrides and low valent species and/or stoichiometric protonolysis has been reported.12 Recently, a heterogeneous Ti hydride was found to catalytically convert N2 and H2 into NH3, whilst TiO2 is known to photolytically convert N2 and H2O into NH3 and O2.13 These reports hint that Ti could hold significant promise in this arena. This is appealing because Ti is the ninth most abundant element in the Earth's crust, and second only to Fe for metals that can fix N2.

Herein, we report the first abiological, homogeneous Ti complex that is competent for catalytic reduction of N2 to NH3. We find that N2 binding and partial activation occurs at TiIII supported by one of the simplest triamidoamine ligands, priming the N2 for cooperative reduction by KC8. Protonation to give NH3 in a catalytic cycle is facilitated by phosphonium salts that are the weakest proton source used in any catalytic system to date.

Under argon, reduction of yellow [TiIV(TrenTMS)X] (X=Cl, 1 A;14 I, 1 B; TrenTMS=N(CH2CH2NSiMe3)3) with KC8 yields green [TiIII(TrenTMS)] (2).15 Three broad 1H NMR resonances are observed (C6D6, 3.6, 1.0, and −22.3 ppm), whilst no 13C or 29Si NMR resonances could be detected, and an EPR spectrum of 2 in frozen toluene‐pentane exhibits g=1.984, 1.943, and 1.891. This is consistent with the presence of TiIII, unfortunately 2 has resisted exhaustive attempts to isolate it. However, slow cooling of an N2‐saturated pentane solution of 2 yields red crystals of [{(TrenTMS)TiIII}2(μ‐η1:η1‐N2)] (3), Scheme 1.15 In the solid state 3 has an effective magnetic moment of 2.50 μB at 298 K, which is close to the spin‐only value for two s= species (μ eff=2.5 βe for g=2.0). The near temperature‐independence of μ eff indicates that any Ti⋅⋅⋅Ti interaction is very weak.15 EPR spectra of solid 3, give apparently axial g ⊥=1.989 and g ∥=1.940 at room temperature; however on cooling these diverge to rhombic g=1.993, 1.951, and 1.902 and are temperature‐independent below 100 K, suggesting that molecular motion at higher temperatures gives higher effective symmetry. The magnetism and EPR spectra of 3 thus suggest the presence of TiIII ions, and the structural and spectroscopic data (Table 1) are consistent with modest activation of the N≡N triple bond.16

Scheme 1

Synthesis of 2–5 from 1A/B. Reagents and conditions: i) Ar, 2 KC8, −2 KX, −2 C8; ii) N2, cool; iii) −N2, warm; iv) 2 KC8; v) I2, −2 KI; vi) 4 B15C5; vii) N2, 4 KC8, −2 KX, −2 C8.

Table 1

Key crystallographic bond lengths and Raman spectroscopic data for 3–5 and N2H benchmarks (x=0, 2, 4).15, 24

Cmpd	ν(N2) [cm−1]exp(14N/15N); calc	d(N−N)[Å]	d(Ti−NN)[Å]	∡(Ti−N−N)[°]
N2	2330	1.098(1)	–	–
N2H2	1583,1529	1.25	–	–
N2H4	1076	1.45	–	–
3	1701/1644;1724	1.121(6)	2.022(3)	180
4	1201/1164;1247	1.315(3)	1.814(2)1.810(2)	166.6(2)172.9(2)
5	1246/1203;1307	1.461(7)	1.712(4)	178.8(5)

The formulation of 3 is corroborated by DFT calculations;15 all attempts to model a TiIV/TiIV/N2 2− combination were intractable or produced the TiIII/TiIII/N2 formulation. Computed Ti−NN and N−N Mayer bond orders of 0.56 and 2.38 are consistent with a weakly bound and activated N2. This is also supported by computed Ti spin densities/charges of −0.55/0.99 and a total net spin density/charge of −0.9/‐0.56 on the N2‐unit. The two SOMOs are orthogonal Ti→N2 π* back‐bonding interactions.

The 1H NMR spectrum of 3 (C6D6) reveals identical resonances to those of 2, consistent with N2 dissociation in solution. Complex 2 is similar to [Ti(TrenDMBS)] [TrenDMBS=N(CH2CH2NSiMe2But)3],17 and their optical spectra exhibit broad absorptions at about 620 and about 640 nm, respectively,15 which is responsible for their green colours. In line with this, [Ti(TrenTMS)] exhibits computed HOMO to LUMO+3/+4 (d–d) energy separations of 655/657 nm.

Addition of two or four equivalents of KC8 to 3 or 1A, respectively, under N2 affords red–brown [{(TrenTMS)TiIV}2(μ‐η1:η1:η2:η2‐N2K2)] (4; Scheme 1).15 The structure of 4 (Table 1) shows the N2 ligand is bound in a near‐linear manner. The N−N bond distance is extended compared to 3 and free N2 and the Ti−NN bond distances are short, inferring some Ti–imido character. These data along with Raman spectroscopy suggest strong activation of N2. Multinuclear NMR spectra of C6D6 solutions of 4 and 4‐ N are consistent with a C 3‐symmetric TiIV/TiIV/N2 4− formulation.

Only a closed‐shell formulation for 4 gave a converged DFT calculation,15 where the Ti ions and N2 unit carry computed charges of 0.52 (av.) and −1.1 (total), respectively; this implies a covalent bonding picture for the Ti−(μ‐N2)−Ti unit that is supported by Ti−NN and N−N Mayer bond orders of 1.25 and 1.44, respectively. For comparison, the Ti−Namide and Ti−Namine Mayer bond orders are about 0.7 and 0.25, respectively. The HOMO and HOMO−1 are two orthogonal, doubly occupied Ti→N2 π* back‐bonding interactions. The KI ions clearly play a stabilising role, but this is electrostatic, with K−N Mayer bond orders of <0.05.

Addition of 4 equivalents of benzo‐15‐crown‐5 (B15C5; Scheme 1) gives red [{(TrenTMS)TiIV}2(μ‐η1:η1‐N2)][K(B15C5)2]2 (5).15 The solid‐state molecular structure of 5 (Table 1) reveals that the Ti−N−N−Ti axis is closer to linearity than in 4. The N−N and Ti−NN bond distances in 5 are longer and shorter, respectively, suggesting strong N2 activation and significant Ti–imido character. However, for 5 the ν(N2) Raman stretch, a better indicator of N2 activation than bond distances, suggests reduced N2 activation compared to 4, Table 1. UV/Vis spectra of 4 and 5 are essentially identical in THF, but different in benzene, suggesting that the KI ions in 4 are labile in polar donor solvent, but remain coordinated in non‐polar solvents.

Only a closed‐shell formulation for the dianion part of 5 gave a converged calculation. The Ti−NN and N−N Mayer bond orders in 5 are 1.33 and 1.62, which are larger than the corresponding data for 4; the latter is in‐line with the Raman data, whilst the former suggests that more Ti–imido character results from removal of KI ions. The fact that 5 contains a charge‐rich dianion is reflected by Ti−Namide and Ti−Namine Mayer bond orders that are lower than in 4 at about 0.6 and 0.16, respectively. This is consistent with computed Ti charges of 0.99 in 5 that are higher than those in 4, but the N2 unit is less charged at −0.86 overall.

Having established N2 reduction, attention turned to fixation (Table 2). Treatment of 4 (10 mm, pentane) with ethereal 1 m HCl (10 equiv) yielded 0.88 N2H4 and 0.13 NH3 equiv (entry 1). The near‐stoichiometric yield of N2H4 confirms 4 as a hydrazido complex. In a control, when 1A is identically quenched only 0.04 NH3 equiv were detected. This suggests that the small amount of NH3 produced is the result of minor degradation of TrenTMS under the action of very strong acid.

Table 2

Catalytic acidification experiments for the reaction of 4 with acid and reductant under N2 to produce NH3 and N2H4.[a]

Entry[b]	Solvent	Acid	Reductant	Acid[eq.]	Reductant[eq.]	NH3 [eq.]	N2H4 [eq.]	Fixed‐N[h] [eq.]	Efficiency[i] [%]
1	pentane	1 m HCl	–	10	–	0.13	0.88	1.89	–
2	Et2O	[Cy3PH][I]	–	10	–	0.05	0.52	1.09	–
3	Et2O	[Cy3PH][I]	KC8	120	120	6.41	0.15	6.71	17
4	Et2O	[Cy3PH][I]	KC8	300	300	11.91	0.06	12.03	12
5	Et2O	[Cy3PH][I]	KC8	400	300	10.81	0.10	11.01	11
6	Et2O	[Cy3PH][I]	KC8	600	600	17.77/17.4[g]	0.03	17.83	9
7[c,d]	Et2O	[Cy3PH][I]	KC8	600	600	17.70/17.6[g]	0.08	17.86	9
8[c,e]	Et2O	[Cy3PH][I]	KC8	300	300	1.53	0	1.53	–
9	pentane	[Cy3PH][I]	KC8	300	300	5.82	0.29	6.40	6
10	toluene	[Cy3PH][I]	KC8	300	300	3.89	0.21	4.31	4
11	THF	[Cy3PH][I]	KC8	300	300	8.95	0	8.95	9
12	Et2O	[Cy3PH][I]	K(Nap)(THF)	300	300	0.36	0.07	0.50	0
13	Et2O	[Cy3PH][Cl]	KC8	300	300	2.45	0.05	2.55	3
14	Et2O	[Cy3PH][BArF 4]	KC8	300	300	4.77	0	4.77	5
15	Et2O	[nBu3PH][I]	KC8	300	300	11.73	0.09	11.91	12
16	Et2O	[tBu3PH][I]	KC8	300	300	7.37	0.30	7.97	8
17[f]	Et2O	[Cy3PH][I]	KC8	25	25	1.32	0.34	2.00	–

[a] Diazene (N2H2) was not analysed owing to its expected instability under these reaction conditions; however, complete disproportionation of N2H2 to N2H4 and N2 can only be expected to produce a maximum N2H4 yield of 50 %. [b] All experiments were performed under N2 (unless otherwise noted) at −78 °C (2 h), followed by gradual warming to 25 °C and additional stirring for 15 h. [c] 4‐ N. [d] Under 15N2. [e] Under Ar. [f] [N2H5][I]. [g] Yield calculated from 1H NMR. [h] Fixed‐N (eq.)=[NH3 (eq.)]+2 [N2H4 (eq.)]. [i] Efficiency=100 %[Red. (eq.)]/{3[NH3 (eq.)]+4 [N2H4 (eq.)]}.

To achieve catalytic turnover of N2, it was concluded that a milder acid than ethereal HCl (pK a (Et2O⋅H)+=−3.59) would be required. Also, the synthesis of 4 requires strong K‐based reductants, which react rapidly with strong, soluble acids. Confirming this, using HCl/KC8 (30:30:4) resulted in a sub‐stoichiometric yield of N2H4 and NH3 (0.13 and 0.36 equiv, respectively). It was anticipated that a weaker acid could be effective for the protonation of activated N2 whilst minimising deleterious side‐reactions. However, commonly used N‐based acids, such as lutidinium [2,6‐(CH3)2C5H3NH]+ and arylammonium/alkylammonium salts were found to be susceptible to reduction by KC8, and in some controls led to NH3 formation; as such, we anticipated false‐positive results. Thus, more stable, non‐N‐based acids were sought.

Trialkylphosphonium salts, [R3PH][X], were examined because of their mild acidity (pK a≈8–12) and tuneable nature. Initial experiments were conducted with [Cy3PH][I] (pK a=9.7). Addition to 4 (10:1 ratio) in Et2O produced 0.5 N2H4 and 0.05 NH3 equiv (entry 2). This is a lower yield of N2H4 compared to using HCl, but [Cy3PH][I] is a weaker, less soluble acid so it was expected to be less reactive. However, with excess [Cy3PH][I]/KC8, 4 catalyses the production of up to 18 equiv of NH3 per 4 (entries 3–11). Using 4‐ N under 15N2 confirmed the incorporation of 15N2 into the 15NH3/15NH4Cl product.

A sub‐stoichiometric yield was obtained using [K2(C10H8)2(THF)] as the reductant (entry 12), which is most likely due to its solubility in Et2O and thus greater propensity to react with H+ in situ, consuming acid and reductant for H2 production. Comparative runs in ethers (Et2O and THF), in which intermediates may be solvated/stabilised, gave higher yields than those in toluene or pentane (entries 4 and 9–11). The acid anion was also varied: [Cy3PH][Cl] (Cl−, higher coordinating ability) or [Cy3PH][BArF 4] (BArF 4 −=[{3,5‐(CF3)2C6H3}4B]−, non‐coordinating; entries 4, 13, and 14). For both, this resulted in lower yields of NH3, but supposed variation of a single parameter will simultaneously affect several properties that all influence catalytic turnover, preventing meaningful correlation. Similar catalytic turnovers were obtained using [Bun 3PH][I] and [But 3PH][I] (pK a=8.4 and 11.4, respectively; entries 15 and 16), or using a 4:3 ratio of [Cy3PH][I] to KC8 (entry 5), which accounts for an additional equiv of acid being consumed through protonation of NH3.

With KC8, the major product is NH3, with little N2H4 (<0.15 per 4) detected. That 4 does not react further with KC8 in Et2O implicates an initial protonation step rather than further reduction of 4 to form a TiIV–nitride species, which could be responsible for NH3 production. A further control demonstrated that [N2H5][I] is converted into NH3 in the absence of 4/4‐ N (entry 17), and with entry 1, this suggests that the N2H4 to NH3 reduction/protonation may occur after dissociation from the active species. We cannot rule out the presence of transient TiIV–nitrides,18 but these control experiments suggest that successive protonation may produce “[{(TrenTMS)TiIV}2(μ‐η1:η1‐N2H4)][I]2”, with subsequent extrusion of N2H4, that reacts with [Cy3PH][I]/KC8 to form NH3; concomitant formation of 1B closes the catalytic cycle.

Under conditions given in entry 4, 5 is catalytically competent producing 10 equiv of NH3. Complex 3 produces 6 equiv of NH3 under such conditions, showing catalytic competence, but on dissolution a significant proportion of 3 converts into 2 (Scheme 1), retarding reactivity. In isolation, 3 reacts with acid to produce 0.03 and 0.1 equiv of NH3 and N2H4, respectively, underscoring the importance of KC8 activation of the coordinated N2 in 3.

The reactivity of 5 is significant, because it is a clear‐cut end‐on:end‐on bridging N2 complex. To date, almost all N2‐fixing catalysts contain end‐on terminal N2,7 whereas those with end‐on:end‐on bridging N2 have been proposed to dissociate to form end‐on terminal complexes, or undergo N2 cleavage to generate nitrides.7, 9b,9d, 18, 19, 20 Side‐on:side‐on bridging, which usually results in strong N2 activation, has been reported for stoichiometric N2 activation only. Conversion of end‐on:end‐on bridging into side‐on:side‐on bridging in 5 seems unlikely on steric grounds, so the reactivity of 5 suggests that consideration might be given to recognising end‐on:end‐on bridging as a catalytically competent N2 coordination mode. If this is the case, it would provide an unusual symmetrical functionalisation of N2 to N2H4 (in the absence of reductant to cleave the N−N bond) compared to the traditional Chatt‐type cycle,21 since only one molecular Fe complex is known to be selective for catalytic reduction of N2 to N2H4 rather than NH3.22

To conclude, we have reported the first abiological early base metal complex that is competent for the homogeneous catalytic reduction of N2 to NH3 under ambient conditions, and this chemistry is supported by the simple TrenTMS ligand. The proton sources in this catalysis, [R3PH][I], are the weakest yet used in N2 reduction to NH3, underscoring the importance of balancing acid sources to strong reducing agents. We propose a plausible mechanism, where TiIV is reduced to TiIII, which binds and weakly activates N2, priming it for cooperative K‐mediated reduction10e, 23 to a strongly activated state, which can then be protonated. Transient nitrides cannot be ruled out at this point, but control experiments suggest that N2H4 is formed and then converted into NH3 in the presence of acid and reductant. The catalytic activity of one of the Ti complexes suggests that the end‐on:end‐on bridging mode of N2 should possibly no longer be discounted as a catalytically active coordination mode. These results add Ti to the small number of previously exclusively mid‐ and late‐d‐block metal ions (Mo, Fe, Ru, Os, Co) that can execute catalytic reduction of N2 to NH3.24

Conflict of interest

The authors declare no conflict of interest.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supplementary

Click here for additional data file.