Mono- and dinuclear Ni(I) products formed upon bromide abstraction from the Ni(I) ring-expanded NHC complex [Ni(6-Mes)(PPh3)Br]

Bromide abstraction from the three-coordinate Ni(i) ring-expanded N-heterocyclic carbene complex [Ni(6-Mes)(PPh3)Br] (1; 6-Mes = 1,3-bis(2,4,6-trimethylphenyl)-3,4,5,6-tetrahydropyrimidin-2-ylidene) with TlPF6 in THF yields the T-shaped cationic solvent complex, [Ni(6-Mes)(PPh3)(THF)][PF6] (2), whereas treatment with NaBArF4 in Et2O affords the dimeric Ni(i) product, [{Ni(6-Mes)(PPh3)}2(μ-Br)][BArF4] (3). Both 2 and 3 act as latent sources of the cation [Ni(6-Mes)(PPh3)]+, which can be trapped by CO to give [Ni(6-Mes)(PPh3)(CO)]+ (5). Addition of [(Et3Si)2(μ-H)][B(C6F5)4] to 1 followed by work up in toluene results in the elimination of phosphine as well as halide to afford a co-crystallised mixture of [Ni(6-Mes)(η2-C6H5Me)][B(C6F5)4] (4), and [6MesHC6H5Me][B(C6F5)4]. Treatment of 1 with sodium salts of more strongly coordinating anions leads to substitution products. Thus, NaBH4 yields the neutral, diamagnetic dimer [{Ni(6-Mes)}2(BH4)2] (6), whereas NaBH3(CN) gives the paramagnetic monomeric cyanotrihydroborate complex [Ni(6-Mes)(PPh3)(NCBH3)] (7). Treatment of 1 with NaOtBu/NHPh2 affords the three-coordinate Ni(i) amido species, [Ni(6-Mes)(PPh3)(NPh2)] (8). The electronic structures of 2, 5, 7 and 8 have been analysed in comparison to that of previously reported 1 using a combination of EPR spectroscopy and density functional theory.


Introduction
In a very recent review, Lin and Power referred to Ni(I) as a '... "rare" oxidation state of growing importance'. 1,2 In terms of monodentate ligands, the early dependence on tertiary phosphines to stabilise Ni(I) 3,4 has largely been superseded by the use of Nheterocyclic carbenes (NHCs) and these have facilitated the isolation of a wide range of fully characterised four-, three-and even two-coordinate Ni(I) species. 5,6 Over the last few years, we have used so-called ring expanded NHCs (RE-NHCs; carbenes with ring sizes >5) for the preparation of three-and two-coordinate Ni(I) complexes with interesting stoichiometric 7 and catalytic chemistry, 8 as well as novel magnetic properties. 9 In all cases, the starting point for our chemistry has been the threecoordinate species [Ni(RE-NHC)(PPh3)Br]. 10 The first of these to be prepared, Mes)(PPh3)Br] (1, Scheme 1), 8 has continued to be the focus of much of our attention as it tends to yield readily isolable products.
Herein, we describe the stoichiometric reactivity of 1 with a range of bromide abstracting agents to afford seven new Ni(I) complexes. Five of these are monomeric (cationic as well as neutral) and their adoption of T-or Y-shaped structures has been probed using DFT calculations.

Bromide abstraction from 1 by [Tl] + , NaBAr F 4 and [(Et3Si)2(-H)][B(C6F5)4].
We have previously shown that the addition of free 6-Mes to 1 results in transfer of the bromide ligand to the outer-sphere to give the two-coordinate, cationic product [Ni(6-Mes)2]Br. 9 Initial efforts to abstract bromide from 1 with more typical halide abstractors 4 such as AgX reagents (X = BF4, NO3, OTf) yielded only mixtures of products containing the pyrimidinium salt X and the plating out of what appeared to be metallic nickel. However, when 1 was treated with TlPF6 in THF, the three-coordinate cationic THF complex [Ni(6-Mes)(PPh3)(THF)][PF6] (2) was isolated as a pale yellow solid in 85% yield (Scheme 1).
The Ni-O distance of 2.0956(17) Å was intermediate between those reported for the neutral -diketiminato species [L R Ni(THF)] (L R = [HC(C( t Bu)NC6H3( i Pr)2)2] -, 2.000 (1) Å)) [11][12][13]  The 1 H NMR spectrum of 2 displayed a series of broad resonances between ca.  17-0 which could not be integrated. As the signals for the bound THF could not be assigned, we were unable to establish spectroscopically the lability of the THF ligand. 6 However, X-ray crystallography repeatedly revealed the presence of THF following recrystallization of 2 from a number of solvents (CH2Cl2, C6H5F, C6H6) suggesting that the THF cannot be easily dissociated from the nickel.
The formation of TlBr as a side-product in the synthesis of 2 proved problematic, as even following multiple recrystallisations, complete removal was not always acheivable. This manifested itself in EPR spectra of 2 (ESI), but more obviously in reactions with CO (vide infra). Fig. 2 shows the EPR spectrum of a 'clean' sample of complex 2 (Fig 2d). The spin Hamiltonian parameters of the EPR spectra of all of the species shown in Fig. 2 are listed in Table 1, and are discussed further below.
The B2 based borohydride has a similar coordination mode once experimental errors are taken into consideration. However, H2E may be closer to Ni1 (1.92(4) Å) than to Ni2 (2.12(4) Å), which would indicate a tendency towards an even more unusual  2 ,  2 :  1 coordination mode. 25 In an effort to further probe the bonding of the borohydrides, a neutron dataset was collected, but a phase transition hampered acquisition of any additional insights (see experimental). IR spectroscopy provided little in the way of diagnostic characterisation of any particular coordination mode, as only a single, broad (B-H) absorption band was measured at 2378 cm -1 in KBr. 125.14 (7), C(1)-Ni(1)-N(3) 132.38 (9), P(1)-Ni(1)-N(3) 102.47 (7).
We have previously reported that 1 reacts with NaO t Bu to provide a low yielding route to the Ni (0)   The overall geometric changes in fully relaxing the crystal structures are small (see ESI). Most importantly, the striking consistency of the Ni-P and Ni-C bond lengths in the crystal structures across the series (variation < 0.03 Å and < 0.02 Å, respectively) is preserved upon geometry optimisation (variation < 0.03 Å and < 0.04 Å, respectively: ESI). The tendency of three-coordinate transition metal d 9 complexes to form either Tor Y-shaped geometries is due to the Jahn-Teller effect, thus lifting orbital degeneracy (dxy, dx2-y2) at the ideal D3h symmetry (Fig. 10a). MO theory predicts that the SOMO in a T-shaped d 9 complex will be of dx2-y2 character, whereas in a Y-shaped d 9 complex, it will be of dxy character (Fig. 10a), in agreement with the dominant character of the DFT-20 calculated orbitals (Fig. 10b). For Ni(I) complexes, this was most recently discussed by the groups of Holland and Lee, 13,21 and prior to that, by Pietrzyk. 38 Holland and coworkers rationalised the formation of T-vs Y-shaped complexes with a charge donation analysis (natural bond order analysis, NBO). 39 Their findings indicated that a T-shape is inherently favoured by d 9 complexes, but a Y-shape can result when there is increased donation of charge from the ligands to the metal centre, thus effectively partially reducing the metal centre. In the present case, the analysis of Mulliken 40  The CW X-band EPR spectra of complexes 1, 2, 5 and 8 were shown in Fig. 2.
The resulting spin Hamiltonian parameters, notably the g-tensor and A( 31 P)-tensor components were extracted by simulation, and are listed in Table 1. All spectra display a rhombic g profile, with one component (g1) Table 1 and are in reasonable agreement with the experimentally determined The relative orientations of the g-and A-tensors for the cationic complex 5 are shown in Fig. 11, alongside the spin density. As a comparison, g-and A-tensor orientations and spin densities derived from the DFT calculations of EPR parameters for starting complex 1 are given in Fig. 11b (corresponding figures for complexes 2, 7 and 8 are given in the ESI).
As already mentioned, the 31 P superhyperfine interaction is almost entirely isotropic, therefore an explanation for the much smaller HFC in the case of complex 5 compared to starting complex 1 (see aiso( 31 P) in Table 1) can be found by simply looking at the overall spin density on the 31 P nuclei, neglecting the relative orientations of the A( 31 P) frames in each of the complexes (isotropic interaction is orientation independent).
As the insets clearly show, there is a significantly less spin density on the 31 P nucleus of 5 when compared to 1, which readily explains the much lower hyperfine interaction found experimentally and computationally. In fact, the spin density on the 31 P nucleus of 5 is so small that two of the principal values of the A( 31 P) tensor for this complex are smaller than the overall broadening caused by g-strain effects and are completely unresolved at X-band. Only the A3( 31 P) component of the tensor is visible at X-band. In the spectrum in  character (simultaneously to dxy contributions approaching zero) when moving away from ideal D3h symmetry towards T-shape symmetry. Both these observations seem to be in good agreement with what was described previously and represented in Fig. 9. Very interesting is the case of complex 8, which as we noted above may be regarded as a Tshape complex with C-Ni-L as the largest angle. However, a fully geometry optimised version of the same complex where the amido phenyl substituents were replaced by methyl substituents (8-Me, vide supra) showed angles that are similar to the Y-shape complexes 1 and 7. Our interpretation is that this compound is electronically inclined to be a Y-shape (similar to the other neutral compounds of the present series), however large steric strain pushes the amido group towards the carbene ligand, thus geometrically distorting it towards a T-shape. Orbital distribution and coordination geometry should reflect the shape and magnitude of the g tensor associated with the paramagnetic centre.
In Fig. 12b, experimental grel, a parameter used to evaluate the shape of the diagonalised g tensor and calculated according to Equation 1, is also reported as a function of the (bond angle).
It can be seen that an increase in the dx2-y2 contribution to the SOMO corresponds to a shift of the g2 value away from g3 towards g1, and indeed for T-shape complexes g2 is closer to g1 than to g3, highlighting a geometry induced shape shifting of the g tensor.

Conclusions
Treatment of the three-coordinate Ni(I) complex [Ni ( either T-shaped or Y-shaped geometries. These structural differences manifest in different electronic structure characteristics, namely that the SOMO for a T-shape complex is expected to be of dx2-y2 character, whereas for a Y-shape complex, it will be of dxy character. Electron paramagnetic resonance spectroscopy was used to derive spin Hamiltonian parameters for this series of three-coordinate Ni(I) complexes, which showed that all complexes have a rhombic g-tensor profile and that the 31 P superhyperfine couplings are predominantly isotropic. The much lower magnitude of 31 P superhyperfine coupling constants observed for the CO-containing complex 5 was explained with a smaller spin density found at the phosphorus ligand as predicted by density functional theory calculations. The overall computed spin densities in this series are polarised differently for the Y-and T-shaped complexes, namely with a larger 29 lobe trans to the phosphine ligand in the former case as opposed to a larger lobe trans to the ligand in the latter case. This directly affects the shape and magnitude of the g-tensor: while all complexes have a rhombic g-tensor with g1<<g2<g3, a larger dx2-y2 contribution to the SOMO shifts g2 closer to g1.

General considerations
All manipulations were carried out using standard Schlenk, high vacuum and glovebox techniques. Solvents were purified using an MBraun SPS solvent system (hexane, Et2O) or under a nitrogen atmosphere from sodium benzophenone ketyl (benzene, THF

[Ni(6-Mes)(PPh3)(CO)][PF6] (5)
To a degassed THF solution (0.5 mL) of 2 (20 mg, 0.02 mmol), 1 atm of CO was added to the stirring solution. An immediate colour change to dark yellow/green occurred, and after 1 minute the solution was reduced to dryness. The residue was extracted into THF (0.5 mL), filtered and layered with hexane (2 mL
The solution was reduced to dryness and the residue extracted into benzene (2 x 10 mL).
Upon removal of the benzene, the orange residue was washed with hexane (2 x 10 mL) to using Cu(K) radiation. All experiments were conducted at 150 K, with the exception of that for 6, which was achieved at 100 K. Details of the data collections and refinements are given in Table 2. The structures were uniformly solved using SHELXS, 42 and refined 33 using full-matrix least squares in SHELXL 43 via the Olex-2 44 software suite. Only noteworthy refinement details follow.
A small amount of racemic twinning was accounted for in the refinement of 2 (ESI only). This structure represents a P21 polymorph of compound 2, the latter solving in space group P21/n. In 3, the asymmetric was seen to contain one anion, one cation and trifluoromethyl groups) C-F and FF distances within each disordered region were restrained to being similar in the final least squares. In addition, the ADPs for fractional occupancy atoms were also restrained, to assist convergence.
The cation in the asymmetric unit is of 4 was also fell prey to disorder. In particular, there is a 50:50 ratio of the tolyl-Ni-carbene moiety present versus the tolylpyrimidinium pair, the latter being stabilised by a C-H interaction. In 5, the asymmetric unit was seen to comprise one cationic nickel containing species, one [PF6]anion and one THF molecule. The crystal was small, which contributed to weak diffraction at higher Bragg angles. Hence, data were truncated to a  value of 24.7.
The borohydride hydrogen atoms in the structure of compound 6 were readily located and refined with a common Uiso in each [BH4]moiety. No distance restraints employed. C25 was modelled for 87:13 disorder, and the minor component of this atom was refined isotropically. A data collection was also performed on this compound, at room temperature (designated 6a, ESI), in which the asymmetric unit was seen to consist of one half of a dimer molecule, wherein the metal centres and carbene carbon atoms were noted to coincide with a crystallographic 2-fold rotation axis. This necessarily means that the apical NHC carbons (C3 and C15) are each disordered in a 50:50 ratio.
This disorder precluded addition of the hydrogen atoms bound to C2 and C14 using the riding model; hence, they were omitted from the refinement. The borohydride hydrogens were located, and refined without restraints, but their credibility is somewhat questionable given their associated Uiso values and the overall atomic displacement parameters. The reason for implementing a room temperature data collection for 6a was to resolve a phase transition that arose in the course of a neutron experiment conducted on 6, using VIVALDI, at the ILL. The rationale for doing a neutron experiment arose because, at 100 K, the borohydride moieties appeared to coordinate unsymmetrically to the nickel centres. Unfortunately, during cooling at the neutron source, the large crystals cracked. This ultimately resulted in collection of a neutron data set at room temperature, which suggested a different space group (C2/c) to that for the structure determined at 100 K using X-rays (P21/c).
This phase transition, from a diffraction perspective, results in averaging the electron density that arises from the borohydrides across the sample and, overall, the ambient temperature neutron data did not afford any additional insight into the bonding subtleties which the experiment aimed to probe.
The asymmetric unit in 7 was seen host to one molecule of THF in addition to one molecule of the nickel complex. C3 in the latter was equally disordered over two sites, and the four chemically equivalent C-C distances involving C3/C3A were restrained to being similar in the final least squares. Three of the five atoms in the solvent were also refined to take account of 75:25 disorder. Once again, the chemically equivalent distances involving fractional occupancy atoms in this moiety were restrained to being similar, and ADP restraints were also incorporated to assist convergence.
In addition to one molecule of the complex, the asymmetric unit in 8 was noted to contain one molecule of guest diethyl ether.
Analysing the crystal structure of [Ni ( L of dry THF (in all cases, a small quantity of dry toluene was also added to improve the quality of the polycrystalline glass formed in frozen solution, and thereby enhance the quality of the EPR spectra). The solutions were transferred to an EPR tube, sealed in the glove box and then cooled to 77 K before rapid transfer to the pre-cooled EPR cavity.
The X-band CW EPR measurements were performed on a Bruker EMX spectrometer utilizing an ER4119HS resonator, 100 kHz field modulation at 140 K.

Computational Details
All density functional theory (DFT) calculations were carried out with ORCA (version 4.0.0.2). 45 The geometries were taken from crystallographic refinements, either optimising only the positions of the hydrogen atoms or fully relaxing the geometry. The geometry optimisations used the BP86 density functional, 46 making use of the zeroth order relativistic correction ZORA retaining onecenter terms. 47 The scalar-relativistically recontracted versions of Ahlrich's triple-zeta quality basis sets (ZORA-def2-TZVP) were used on all atoms except carbon and hydrogen for which ZORA-def2-SVP basis sets were used. 48 The resolution of the identity (RI) approximation and the auxiliary basis SARC/J were used. 49 The integration accuracy was increased to 7.0, the grid was set to 7 in ORCA nomenclature, and 'tight' SCF criteria were used. The optimisations considered solvent effects through the conductor-like polarisable continuum model, with the solvents as indicated in the experimental part. 50 Dispersion effects were taken into account with Grimme's D3BJ model including Becke-Johnson damping. 51 Mulliken spin populations were inspected to confirm convergence to the targeted electronic structure.
Broken-symmetry DFT calculations used the functionals TPSSh, 52 B3LYP, 53 PBE0, 54 M06L, 55 additionally making use of the chain-of-spheres approximation (RIJCOSX) and using the 'flipspin' feature in ORCA to generate the initial guess for the brokensymmetry solution, with otherwise unchanged calculation setups. 56 The exchange 37 coupling constants were taken directly from the ORCA output, using the definition by Yamaguchi. 57 For the calculation of EPR parameters, it was found that calculations with a different family of basis sets gave superior results. Generally, the IGLO-II basis set was used on all atoms, with CP for Ni and aug-pc-3 for Br, 58 in conjunction with the PBE0 density functional and the RIJCOSX approximation as for the BS-DFT calculations, making use of the AutoAux feature in ORCA. The grid sizes were set to Grid6 and GridX9 in ORCA nomenclature, with increased grids (7) on the Ni ion and all directly bound atoms as well as the nitrogen atoms in the carbene ligand. The spin-orbit mean field operator (SOMF(1X)) was used, and the origin for the g-tensor was taken at the centre of the electronic charge. 59 All tensor orientations, spin densities and molecular orbitals depicted and discussed in the main text and the ESI are derived from calculations at this level of theory.  (12) 14.2120 (4) 18.99247 (14) b/Å 14.8045 (1) 17.3278 (5) 10.32800 (10) 16.4910 (3) 20.24524 (15) 16.5650 (4) 11.96944(9) c/Å 19.0391 (1) 25.2732 (6) 28.8100 (3) 18.8500 (4) 14.83697 (15) 18.0386 (6) 20.99341 (16)