Selectivity of tungsten mediated dinitrogen splitting vs. proton reduction

An N2-bridged ditungsten complex is presented that undergoes N2-splitting or hydrogen evolution upon protonation depending on the acid and reaction conditions. Spectroscopic, kinetic and computational results emphasize the impact of hydrogen bonding on the reaction selectivity.


Introduction
Homogeneous N 2 xation under ambient conditions has made remarkable progress over the past 15 years. 1 Nishibayashi and co-workers recently obtained over 4000 equiv. of NH 3 with the proton coupled electron transfer (PCET) reagent H 2 O/SmI 2 as H + /e À sources and a molecular Mo pincer catalyst. 2 Accordingly, nitrogen reduction (NR) via electrochemically or light-driven PCET with related systems has attracted a lot of attention. 3,4 Lindley et al. estimated a suitable overpotential window of 1-1.5 V for selective NR (in MeCN) prior to competing hydrogen reduction (HR) at a glassy carbon cathode. 5 However, besides the thermochemical framework, mechanistic models that account for NR vs. HR selectivities of molecular catalysts are generally poorly developed.
Several M(N x H y ) intermediates relevant to N 2 xation (Scheme 1) exhibit low N-H bond dissociation free energies (BDFEs) below that of free H 2 (BDFE(H 2 , gas) ¼ 97.2 kcal mol À1 ) as possible branching points into HR. 6,7 Computational evaluation of NR vs. HR selectivities for a series of Fe catalysts pointed at bimolecular H 2 loss from species with low N-H BDFEs. 8,9 Attempts to stabilize such Fe(N x H y ) species by hydrogen bonding with pendant bases so far resulted in shutdown of catalysis. 10 But, in fact, such secondary interactions might also be relevant for Nishibayashi's catalyst as indicated by selectivities obtained with 2,6-lutidinium acids ([LutH] + [X] À ) as the proton source. These strongly depend on the X À counter anion: NH 3 /H 2 (X À ) ¼ 7.0 (Cl À ), 0.9 (OTf À ), and 0.14 (BAr 4 À ). 11 In this contribution, we address the role of hydrogen bonding for the selectivity of proton induced N 2 splitting into molecular nitrides vs. proton reduction. N 2 splitting has evolved as an entry into N 2 functionalization for a wide variety of metals 12 and was proposed as the key step in N 2 xation with Mo pincer catalysts. 2,13 Cleavage of Cummins' seminal complex 1 (Fig. 1) was attributed to the {p 1 2 p 2 2 p 3 2 p 4 2 p 5 1 p 6 1 } conguration of the Mo 2 N 2 -core, which enables population of Scheme 1 Transition metal species relevant to N 2 fixation with a low BDFE N-H . 7 a destabilizing s-antibonding molecular orbital (MO) in the transition state. 14,15 In contrast, the {p 8 } oxidation product 2 2+ exhibits strong N 2 activation (Table 1) but lacks two electrons to form stable Mo VI (N 3À ) nitrides. 16 Similar electronic structure considerations can be applied to metal pincer platforms developed by our group. For example, the square-pyramidally coordinated dinuclear Re complex [(N 2 ) {ReCl(PNP)} 2 ] (3, PNP ¼ N(CH 2 CH 2 PtBu 2 ) 2 ) also splits into nitrides at r.t. and exhibits a {p 10 d 4 } conguration that is isolobal with 1 (Fig. 1) (Table 1), but is thermally stable. 19 Unexpectedly, splitting of 4 at r.t. was obtained upon protonation of the pincer backbone (Scheme 2), which was rationalized by a protonation induced low-spin {p 8 d 4 } to high-spin {p 10 d 2 } transition that facilitates electron transfer to N 2 .
Given the impressive N 2 xation rates with Mo pincer catalysts which possibly proceed via N 2 splitting; 2,13 W nitride formation from N 2 is surprisingly rare. 20 We here report tungsten mediated N 2 splitting that competes with proton reduction upon protonation of a {W 2 N 2 } pincer complex. Our results provide evidence for the signicance of hydrogen bonding for the reaction selectivity.

Results and discussion
The reaction of WCl 4 with H PNP in the presence of NEt 3 gives the pincer complex [WCl 3 (PNP)] (5) in yields of up to 60%. In the absence of a signal in the 31 P{ 1 H} NMR spectrum, the paramagnetically shied 1 H NMR signals and the solution magnetic moment derived by Evans' method (m eff ¼ 2.8 AE 0.1 m B ) are in line with a d 2 high-spin (S ¼ 1) conguration. The molecular structure obtained by X-ray diffraction closely resembles previously reported compounds [MCl 3 (PNP)] (M ¼ Re and Mo). 18a,19 Reduction of 5 with Na/Hg (2 equiv.) under N 2 (1 atm) in THF gives the green, N 2 -bridged dinuclear complex [(N 2 ) {WCl(PNP)} 2 ] (6) in up to 66% isolated yield (Scheme 3). In the solid state (Fig. 2) 6 is isostructural with the molybdenum analogue 4, regarding the N 2 binding mode, the approximate C 2 symmetry due to mutually twisted {WCl(PNP)}-fragments (Cl1-W1-W1#-Cl1#: 89.59 /92.27 ), and the distorted squarepyramidal metal coordination (s ¼ 0.35). 21 The short W-N 2 bond (1.78(2)/1.82(4)Å) indicates multiple bonding character. In turn, the N-N bond (1.33(4)/1.27(8)Å) is at the higher end for   The N 2 stretching vibration of 6 was assigned to the Raman signal at 1392 cm À1 (l exc ¼ 457 nm, THF solution; 15 N 2 isotopologue: 1347 cm À1 ) supporting strong N 2 -activation with a formal N-N bond order below the double bonding character (trans-diazene: n NN ¼ 1529 cm À1 ). 23 The closed-shell ground state and degree of N 2 activation are in line with the covalent bonding picture described in Fig. 1. The {p 8 d 4 } conguration of the W 2 N 2 core can be rationalized to arise from two low-spin W II ions. The twisted conformation enables strong back bonding of each metal ion with one p*-MO of the N 2 bridge, respectively, resulting in net transfer of approximately two electrons as judged from the Raman data. This picture is corroborated by DFT computations, which conrm the {p 8 d 4 } conguration of the W 2 N 2 core, analogous to the Mo analogue 4 and Cummins' 2 2+ . The blue-shied N 2 stretching vibration of 6 vs. 4 (Dn NN ¼ 49 cm À1 ; Table 1) indicates slightly reduced back-bonding by the 5d metal.
In the solid state, 7 + and 8 2+ resemble the twisted conformation found for 6 ( Fig. 2). Distinctly different bond lengths around the two tungsten ions of the mixed-valent complex 7 + indicate valence localization, which is further supported in solution by the large comproportionation constant (K c z 10 8 ) 24 and the X-band EPR spectrum at r.t. The isotropic signal (g av ¼ 1.93) of the low-spin (S ¼ 1/2) complex exhibits hyperne interaction (HFI) with only one tungsten (A( 183 W) ¼ 220 MHz) and two phosphorous nuclei (A( 31 P) ¼ 56 MHz), respectively. HFIs with the N 2 -bridge are not found and the 14 N 2 -and 15 N 2isotopologues give identical spectra, further supporting metal centered oxidation. In fact, the degree of N 2 activation is almost invariant within the redox series 6/7 + /8 2+ as judged from the invariance of the N-N stretching vibrations and the N-N bond lengths of the W 2 N 2 cores (Table 1). Notably, the 1 H NMR spectrum of 7 + features four signals assignable to tBu groups, in agreement with the averaged C 2 symmetry and therefore valence delocalization on the slow NMR timescale.
The double oxidation product 8 2+ exhibits paramagnetically shied, yet relatively sharp 1 H NMR signals. Magnetic characterization by SQUID magnetometry reveals a c M T product of about 0.6 cm 3 mol À1 K À1 at r.t., which gradually drops to 0 at about 20 K. The data can be tted to a model with two weakly antiferromagnetically coupled (J ¼ À59 cm À1 ) low-spin (S ¼ 1/2) ions. The g-value (g av ¼ 1.90) indicates an orbital contribution in the typical range for W V complexes with multiply bound hard ligands, such as oxo or nitrido complexes. 25 Characterization of the redox series 6/7 + /8 2+ supports the electronic structure picture with {p 8 1 and Table 1). The spin orbitals of 7 + and 8 2+ are orthogonal to the W 2 N 2 core resulting in weak Scheme 3 Preparation of the N 2 -bridged ditungsten redox series 6-7 + -8 2+ . } (8 2+ ) ground states, respectively, were computed with a low lying triplet state for 8 2+ due to weak antiferromagnetic coupling of the metal centered spins (J DFT ¼ À184 cm À1 ).
Protonation induced N 2 splitting vs. proton reduction N 2 splitting of 6 into the nitride [W(N)Cl(HPNP)]OTf (9 OTf , Scheme 4) as the only detectable tungsten species (NMR/EPR spectroscopy, HR-ESI-MS) was achieved upon adding 2 equiv. of triic acid at À78 C and gradual warming to r.t. Complex 9 OTf could be isolated in over 60% yield and was fully characterized. The tungsten(V) nitride is NMR silent and features an isotropic signal (g av ¼ 1.93) in the X-band EPR spectrum (THF, r.t.) with HFIs with the tungsten and phosphorous nuclei (A( 183 W) ¼ 220 MHz; A( 31 P) ¼ 56 MHz). The W^N stretching vibration is found in the IR spectrum at 1058 cm À1 ( 15 N-9 OTf : Dn ¼ 29 cm À1 ). In the solid state (Fig. 3), 9 + is isostructural with the molybdenum analogue, 19 featuring square-pyramidally coordinated tungsten with the nitride ligand in the apical site. Hydrogen bonding of the amine proton with the triate anion is indicated by the short NH . O distance (2.03(3)Å). The W^N bond length (1.679(2)Å) is in the typical range found for the related tungsten nitrides. 1, 17a,26 In contrast to the Mo analogue 4 (Scheme 2), the selectivity of protonation induced N 2 splitting strongly varies with the reaction conditions. The addition of HOTf (2 equiv.) to 6 at r.t. results in low nitride yields and substantial amounts of the oxidation products 7 + and 8 2+  Next, the inuence of the acid counteranion on the selectivity was probed. Upon protonation with [HNEt 3 ][BAr F 24 ] (2 equiv.), 7 + was found exclusively (Scheme 5). The second oxidation is hampered by the higher pK a of this acid vs. HOTf, which prevents protonation of the monocationic product. Importantly, this selectivity changes with [HNEt 3 ]OTf (2 equiv.): in this case, 7 + is obtained in spectroscopic yields of up to only 30%. In situ HR-ESI-MS examination indicates that nitride 9 + is formed as the only other product. This observation is reminiscent of acid dependent selectivities reported by Nishibayashi for catalytic nitrogen xation (see above). 11 For this reason, [HNEt 3 ] Cl (2 equiv.) was also used. Unfortunately, sluggish mixtures of products were obtained, including substantial amounts of trichloride 5. In the next sections, experimental and computational mechanistic examinations with only [HNEt 3 ]X (X À ¼ BAr F 24 À , OTf À ) are therefore reported.

Mechanistic examinations
Stoichiometric protonation at low temperatures was carried out to obtain spectroscopic information about intermediates. With 1 equiv. of HOTf at low T (À35 C), the NMR data are in agreement with pincer protonation to diamagnetic dinuclear Scheme 4 Protonation of 6 with 2 equiv. of different acids. C 1 -symmetric [(HPNP)ClW(m-N 2 )WCl(PNP)]OTf (10 OTf ), analogous to the respective Mo system (Scheme 6). 19 Notably, immediate formation of the oxidation product 7 + was observed with [H(OEt 2 ) 2 ][BAr F 24 ], even at temperatures down to À75 C. The enhanced stability of 10 OTf suggests an interaction of the immediate protonation product with the tri-ate anion. Contact-ion pair formation is conrmed by 19 F and 1 H DOSY NMR spectroscopy at À35 C. The diffusion coefficient of the triate anion in 10 OTf (D ¼ 2.29 Â 10 À6 cm 2 s À1 ) is in the same range as that of the cation (D ¼ 2.18-2.14 Â 10 À6 cm 2 s À1 ) and signicantly reduced compared to free triic acid (D ¼ 5.11 Â 10 À6 cm 2 s À1 ). We tentatively attribute the solution ionpairing to hydrogen bonding of the triate with the pincer N-H proton, as found in the solid state for 9 OTf (Fig. 3).
Protonation of 6 with 2 equiv. of HOTf at low temperatures in THF is associated with a color change from green to yellow. The absence of a signal in the 31 P{ 1 H} NMR spectrum and the broadened and strongly shied 1 H NMR signals indicate the formation of a paramagnetic product. The magnetic moment for the presumable product, [(N 2 ){WCl(HPNP)} 2 ] OTf2 (11 OTf2 ), was estimated with Evans' method at À60 C (m eff ¼ 4.7 m B ), i.e. close to the spin-only value for a quintet ground state (4.9 m B ). Increasing the temperature leads to fading of the color and disappearance of all 1 H NMR signals, as expected for selective N 2 -splitting into the pale, NMR silent nitride product 9 OTf .
Mechanistic information about proton reduction was obtained from kinetic studies. For this purpose, [HNEt 3 ][BAr F 24 ] was used as the acid, which selectively gives 7 + at r.t. within a convenient timescale even under pseudo rst-order conditions. Addition of the acid to 6 in THF leads to an immediate drop of absorbance without signicant change of the absorption maxima, suggesting only small changes in the electronic structure. The acid concentration dependence of the absorbance allowed for estimating the equilibrium constant and forward rate of the initial protonation of 6 (K 1 ¼ 1592 AE 578 M À1 , k 1 ¼ 163 AE 47 M À1 s À1 ; Scheme 7 and Fig. S25 and S26 in the ESI †). This step is followed by a much slower, monoexponential decay, which was monitored over 5 h (Fig. 4, le). Under pseudo rst order conditions in acid (c(HNEt 3 + ) 0 /c(6) 0 ¼ 10-25), the rate constant (k obs (2) ) linearly depends on the acid concentration (Fig. 4, right), which is in agreement with a slow, irreversible second protonation aer the initial, fast preequilibrium K 1 . However, the non-zero intercept indicates the presence of at least one competitive pathway at a low acid concentration. The rate constant k obs (2) for the formation of 7 + was therefore expressed as eqn (1) which results from the minimum kinetic model outlined in Scheme 7: The rst term accounts for the initial protonation of 6 to give 10 + , followed by irreversible H 2 release from acid and 10 + . Rapid, subsequent comproportionation of the resulting 8 2+ with 6 to 2 Â 7 + is in line with the electrochemical results (K c z 10 8 , see above). The second term in eqn (1) is ascribed to bimolecular decay of 10 + as an alternative path at low acid concentrations. The rate constant k 2 ¼ 0.018 AE 0.001 M À1 s À1 was derived from tting the experimental data to eqn (1) (with preserved K 1 ) under pseudo rst order conditions in acid (10-25 equiv.). The rate constant k 3 ¼ 0.4 M À1 s À1 for the bimolecular path at low acid concentrations was obtained from the initial rate of the reaction of 6 and an equimolar amount of [HNEt 3 ][BAr F 24 ]. ‡ Kinetic analysis suggests two pathways for H 2 formation which both go through the spectroscopically characterized common intermediate 10 + (as 10 OTf ). Path B (Scheme 7) explains the decay of 10 + even in the absence of the acid and reects a bimolecular H 2 formation route as proposed by Matson and Peters for an iron diazenide N 2 -xation intermediate. 8 However, path A is predominant with excess acid. Besides these routes for hydrogen evolution, splitting of the N 2 bridge is observed in the presence of triate as the counteranion and is even selective at Scheme 6 Oxidation of 6 with 1 equiv. of acid at different temperatures.
Scheme 7 Proposed mechanistic pathways for proton reduction at high (Path A) and low (Path B) acid concentrations. lower temperatures. These effects are rationalized computationally in the next section. F 24 ] was rst examined computationally with trimethylammonium as the model acid (Scheme 8). Two different sites, a metal ion and a pincer nitrogen atom, respectively, were considered for the rst protonation step. A hydride product [(PNP)W(H)Cl(m-N 2 )WCl(PNP)] + (12 + ) adopts an electronic singlet (S ¼ 0) ground state and was found to be the global protonation minimum at DG 298 K ¼ À4:7 kcal mol À1 below 6 and [NMe 3 H] + . Hence, the model computation is in excellent agreement with the experimental equilibrium constant K 1 . The computed structure of 12 + features a bridging hydride between the metal ion and a pincer phosphorous atom. A similar structure was previously found experimentally by Schrock and co-workers for the protonation of a PCP molybdenum(IV) nitride by [NEt 3 H][BAr F 24 ]. 27 All efforts to experimentally verify hydride intermediates like 12 + were unfortunately unsuccessful. However, pincer protonation to 10 + is only slightly less exergonic ðDG 298 K ¼ À2:9 kcal mol À1 Þ. Importantly, this state is further stabilized upon use of [NEt 3 H] OTf as the acid due to hydrogen bonding of the pincer amine moiety with the triate anion by ðDDG 298 K ¼ À2:1 kcal mol À1 Þ. In contrast, the hydride ligand is not involved in hydrogen bonding, rendering 10 OTf ðDG 298 K ¼ À5:0 kcal mol À1 Þ the global minimum of the rst protonation in the presence of triate. Overall, the metal and pincer protonation products 12 + and 10 + (and 10 OTf in the presence of triate) should be in rapid equilibrium under these conditions, which is slightly shied towards pincer protonation by hydrogen bonding with the counteranion. Notably, hydrogen bonding with the conjugate base NMe 3 was not observed, presumably for steric reasons.

Protonation with [NEt 3 H][BAr
Starting from the amine/hydride equilibrium, the second protonation with [NMe 3 H] + can ultimately lead to hydrogen evolution or N 2 splitting, respectively. The formation of H 2 and dicationic 8 2+ , which represents Path A (Scheme 7), was computed to be exergonic by ðDG 298 K ¼ À12:8 kcal mol À1 Þ with respect to 6. The most reasonable pathway (Scheme 8, right branch) proceeds via hydride protonation of 12 + leading to the dihydrogen intermediate [(PNP)W(H 2 )Cl(m-N 2 )WCl(PNP)] 2+ (13 2+ ), which is unstable and readily releases H 2 without barriers. While the transition state that leads from 12 + to 13 2+ could not be reliably located due to the at potential energy prole of protonation, the free energy of 13 2+ (DDG 298 K ¼ þ19:7 kcal mol À1 with respect to 12 + ) was used as an estimate for the kinetic barrier of hydride protonolysis. § Notably, this value is in excellent agreement with the experimentally derived barrier for Path A ( Alternatively, splitting of the dinitrogen bridge (Scheme 8, le branch) was computed to proceed via protonation of the second pincer nitrogen. In the absence of triate, [(N 2 ){WCl(HPNP)} 2 ] 2+ (11 2+ ) was located at DG 298 K ¼ þ5:0 kcal mol À1 vs. 6 (DG 298 K ¼ þ9:7 kcal mol À1 vs. the global rst protonation minimum 12 + ) adopting an electronic quintet (S ¼ 2) ground state in accordance with the experimental ndings for 11 OTf2 . From there, N 2 cleavage into the nitrides 9 + was computed to be strongly exergonic ðDG rise to an overall effective barrier for protonation induced N 2 splitting from the most stable monoprotonation intermediate, hydride 12 + , of DG ‡ eff ¼ 31.4 kcal mol À1 . This value is considerably higher than the estimate for the hydrogen evolution pathway (DDG ‡ eff ¼ +11.5 kcal mol À1 ), which is in line with selective proton reduction with [NEt 3 H][BAr F 24 ] as the acid. Importantly, the relative energetics of these two reaction channels are perturbed in the presence of triate as the counteranion. As was found for the rst pincer protonation (see above), triate hydrogen bonding stabilizes the pincer diprotonation product 11 OTf2 by À8.8 kcal mol À1 . Consequently, the estimated effective barrier for hydrogen evolution (DG ‡ 298 K ¼ 20.2 kcal mol À1 vs. the global rst protonation minimum in the presence of triate 10 OTf ) is slightly raised. On the other hand, the N 2 splitting pathway (DG ‡ 298 K ¼ 21.5 kcal mol À1 vs. 10 OTf ) is almost isoenergetic, in full agreement with the experimental ndings. The triate induced effect on selectivity is therefore attributed to Curtin-Hammett controlled reactivity wherein N-H hydrogen bonding to the counteranion modies the energetics of the protonation pre-equilibria.
A similar picture evolves for the reaction with triic acid (see the ESI, Scheme S1 †). However, the potential energy of protonation is augmented by the higher driving force with the stronger acid (pK THF a (Et 3 NH + ) À pK THF a (HOTf) ¼ 4.7). 28 This affects the selectivity as the effective barrier for the N 2 splitting branch versus hydrogen evolution is close in energy. Furthermore, all rate determining states are below the starting point 6. In consequence, under these conditions (HOTf as the acid at r.t.), Curtin-Hammett conditions do not apply resulting in the experimentally observed low selectivity.
Reduction of the temperature to À80 C further perturbs the relative energetics of the two reaction pathways with HOTf. The computed amine ðDG 193 K ð10 OTf Þ ¼ 15:1 kcal mol À1 Þ vs. hydride ðDG 193 K ð12 þ Þ ¼ À10:3 kcal mol À1 Þ equilibrium is even more shied towards the amine due to the lower entropic penalty for hydrogen bonding at low T. The negligible population of the hydride tautomer is in agreement with the exclusive experimental observation of 10 OTf and 11 OTf2 upon single and double protonation with HOTf at À80 C. From 11 OTf2 , the dihydrogen complex 13 2+ ðDG 193 K ð13 2þ Þ À DG 193 K ð11 OTf2 Þ ¼ þ27:7 kcal mol À1 Þ is much higher in free energy than the barrier for N 2 splitting (DG ‡ 193 K ¼ 19.9 kcal mol À1 ), in line with selective N 2 splitting upon double protonation with HOTf at À80 C and slow warming.

Concluding remarks
In summary, an anion effect on the selectivity of proton induced dinitrogen splitting (NR) vs. hydrogen evolution (HR) at the N 2 bridged ditungsten complex 6 was demonstrated and rationalized. Our spectroscopic, kinetic and computational studies suggest some guidelines to improve NR over HR yields: (a) Nitrogen vs. metal protonation offers separate reaction channels with a proposed hydride isomer leading to hydrogen evolution analogous to the highly active Mo-oxo polypyridyl HR catalysts. 29 The tautomerisation equilibrium can be offset by hydrogen bonding with protic N-H hydrogen atoms favoring the use of an acid [BH] + X À where the anion X À is prone to form H-bonds for high NR selectivity. (b) Protonation under Curtin-Hammett control with weak acids can become irreversible with strong acids. Hence, the pK a of the acid can have a decisive kinetic effect on the selectivity.
(c) Lower temperatures favour hydrogen bonding interactions due to the reduced entropic penalty as a strategy for increased NR yields.
Besides the immediate application to the current system, these ndings might be considered as a model reaction for nitrogen xation schemes. The studies of Peters and of Nishibayashi have emphasized the importance of proton coupled electron transfer for N 2 xation under ambient conditions. Our kinetic model might therefore offer some general strategies regarding the choice of acid to improve NR selectivities with respect to unproductive proton reduction.

Conflicts of interest
There are no conicts to declare.