Isomerization Reactions in Anionic Mesoionic Carbene‐Borates and Control of Properties and Reactivities in the Resulting CoII Complexes through Agostic Interactions

Abstract We present herein anionic borate‐based bi‐mesoionic carbene compounds of the 1,2,3‐triazol‐4‐ylidene type that undergo C−N isomerization reactions. The isomerized compounds are excellent ligands for CoII centers. Strong agostic interactions with the “C−H”‐groups of the cyclohexyl substituents result in an unusual low‐spin square planar CoII complex, which is unreactive towards external substrates. Such agostic interactions are absent in the complex with phenyl substituents on the borate backbone. This complex displays a high‐spin tetrahedral CoII center, which is reactive towards external substrates including dioxygen. To the best of our knowledge, this is also the first investigation of agostic interactions through single‐crystal EPR spectroscopy. We conclusively show here that the structure and properties of these CoII complexes can be strongly influenced through interactions in the secondary coordination sphere. Additionally, we unravel a unique ligand rearrangement for these classes of anionic mesoionic carbene‐based ligands.


Synthetic procedures and instrumentation
Unless otherwise noted, all reactions were performed using standard Schlenk-line techniques under an inter atmosphere of argon (Linde, Argon 4.8, purity ≥ 99.998) or an MBraun glove box fitted with a gas purification and recirculation unit. Commercially available chemicals were used without further purification. THF, diethyl ether, toluene and n-hexane were dried and distilled from sodium, dichloromethane and acetonitrile from phosphorus pentoxide. Other solvents were available from MBRAUN MB-SPS-800 solvent system and additionally degassed using standard techniques. 1 H NMR and 11 B NMR were recorded on Joel ECS 400 or JOEL 400R spectrometer. Proton decoupled 13 C NMR were recorded on AVANCE700 proton resonance spectrometer.
NMR spectra based on rearrangement experiments were recorded on a Bruker AV3 or AV 400 MHz spectrometer operating at 400. 13 7 Li. All 13 C spectra were proton decoupled. 1 H and 13 C NMR spectra were referenced against the appropriate solvent signal. 7 Li NMR spectra were referenced against LiCl in D 2 O at 0.00 ppm and 11 B spectra were referenced against BF 3 •OEt 2 in CDCl 3 at 0.00 ppm Chemical shifts are reported in ppm (relative to the TMS signal) with reference to the residual solvent peaks. [1] Multiplets are reported as follows: singlet (s), duplet (d), triplet (t), quartet (q), quintet (quint), and combinations thereof. NMR experiments were conducted in J. Young's NMR tubes oven dried and flushed with argon prior to use.
Mass spectrometry was performed on an Agilent 6210 ESI-TOF.
Cyclic voltammograms were recorded with a PAR VersaStat 4 potentiostat (Ametek) by working in anhydrous and degassed tetrahydrofuran with 0.1 м NBu 4 PF 6 (dried, > 99.0%, electrochemical grade, Fluka) as supporting electrolyte. Concentrations of the compounds were about 1·10 -4 м. A threeelectrode setup was used with a glassy carbon or gold working electrode, a coiled platinum wire as counter electrode, and a coiled silver wire as a pseudo-reference electrode. The ferrocene/ferrocenium couple was used as internal reference.
FTIR-ATR spectra were recorded with a Thermo Scientific TM Nicolet TM iS TM 10 FT-IR spectrometer equipped with a smart orbit unit.
X-ray data were collected on a Bruker Smart AXS or Bruker D8 Venture systems at 140(2) K or 100 (2) K, respectively, using graphite-monochromated Mo Κα radiation (λ α = 0.71073 Å). The strategy for the data collection was evaluated by using the Smart software. The data were collected by the standard omega scan or omega + phi scan techniques, and were scaled and reduced using Saint+ and SADABS software. The structures were solved by direct methods using SHELXS-97 or intrinsic phasing using SHELXL-2014/7 and refined by full matrix least-squares, refining on F 2 . Non-hydrogen atoms were refined anisotropically. [2][3][4][5][6][7][8] Crystallographic data of [Li(L1 N,N )] 2 were collected on an Oxford Diffraction instrument with Cu Kα radiation (λ = 1.54184 Å). Structures were solved using SHELXS-97 [4] and the software OLEX2, [9] while refinement was carried out on F2 against all independent reflections by the S3 full matrix least-squares method using the SHELXL-97 program. All non-hydrogen atoms were refined using anisotropic thermal parameters.
Room temperature X-band CW-EPR spectra of an oriented single crystal [Co(L1 N,N ) 2 ] were acquired on a Magnettech MS5000 EPR spectrometer equipped with a rectangular TE 102 cavity. A single crystal was adhered with vacuum grease on one of the faces of a cubic KCl crystal (approximately 2x2x2 mm), thus defining an orthogonal coordinate system xyz. The KCl crystal was placed on a flat surface inside a cavity performed on a 4 mm diameter quartz rod. The rod was inserted in the cavity of the EPR spectrometer, solidary to a home-made goniometer. This set up allowed us to acquire EPR spectra from 0º to 180º, in 5º intervals, on three orthogonal planes (xy, zx, zy), by subsequently placing down a different face of the KCl cube. Phenyl azide and 1,5-diphenyl-1H-1,2,3-triazole (diPhtz) were synthesized according to a literature procedure. [10] Under an inert atmosphere of argon 1,5-diphenyl-1H-1,2,3-triazole (770 mg, 3.48 mmol) and dicyclohexyl-trifluoromethanesulfonate-borane (378 mg, 1.16 mmol) were dissolved in dry diethyl ether (60 mL) and refluxed for 24 h under an inert atmosphere. During the reaction, the product precipitated as white solid. The white solids were collected by filtration, washed with diethyl ether and dried under high vacuum to give the desired product in good yields of 87%.

Room temperature monitoring study:
In a dry oxygen-free glovebox ligand (H 2 L1)OTf (0.063 mmol) and LiTMP (0.125 mmol) were added to a J. Young NMR tube and dissolved in THF-d 8 (0.5 mL). The solution was then monitored by 1 H NMR spectroscopy at room temperature. The monitoring study is presented in Figure S1.

Low temperature experiment:
The reaction was repeated with the following alterations: As soon as the sample was prepared, it was frozen in a Dewar of liquid nitrogen to prevent reaction. The reaction was warmed to -30 °C and monitored by 1 H NMR spectroscopy. The first point to note from the initial spectrum reveals that the deprotonation is instantaneous, as illustrated by the appearance of resonances corresponding to TMPH, and importantly no remaining LiTMP. After 1 hour at this temperature, there is only minimal rearrangement. The sample was next warmed to -20 °C and, once more conversion to the rearranged complex is very slow over a further hour. The sample was then warmed in the spectrometer to 0 °C. At this point the conversion occurs more rapidly. The spectrum after 2 hours at this temperature is shown in Figure S2. Warming to room temperature results in more rapid conversion to the rearranged isomer in ca. 15 minutes.   Under an inert atmosphere of argon (H 2 L2)OTf (220.9 mg, 0.292 mmol) was dissolved in dry THF At -50 ⁰C (H 2 L1)OTf (0.5 mmol) was added to LiTMP (1.0 mmol) in n-hexane (5 mL). THF (5 mL) was added, causing the suspension to pass into solution. This mixture was stirred for a further 2 hours while gradually warming. All volatiles were removed in vacuo, and toluene was added. The white precipitate was separated and the filtrate was collected and all volatiles were removed. The pale-yellow solid was then recrystalised by suspending in n-hexane (2 mL) and adding toluene (1.5 mL). At this stage gentle heating was required to completely dissolve the remaining solid. Colourless white crystals then grew at room temperature in 68.0% yield.   Figure S4: 1 H NMR spectra of (H2L1)OTf in CDCl3.

EPR
Single-crystal EPR spectroscopy: The EPR spectra acquired on an oriented single crystal of [Co(L1 N,N ) 2 ] show one set of eight well resolved hyperfine lines (I( 59 Co) = 7/2) arising from a single cobalt(II) centre, which reveals the triclinic P1 ̅ space group of the crystal. Selected single crystal spectra are shown in Figure S15 and the full set of spectra in the three experimental crystal planes are shown in Figure S16.  The well resolved hyperfine splittings and low linewidths indicate that intermolecular magnetic dipolar and exchange interactions are negligible, [13][14][15] which is congruent with the long Co-Co distances (12.702(1)Å) and with the presence of bulky cyclohexyl and phenyl groups which interact only weakly in the crystal. The central positions of the resonances were automatically determined using Matlab S19 scripts. From the average (B 0 ) of the eight resonance positions the g-value for each crystal orientation was obtained, and from the average separation a (mT) between adjacent resonances, the K-factor (K = agµ B /h, were µ B is the Bohr magneton and h is Planck's constant) was calculated. Analysis of the angular dependence of the g 2 -value and of the g 2 K 2 -value, (K = agµ B /h) allowed us to obtain the gmatrix and hyperfine A-matrix, following procedures described elsewhere. [16][17] The g-and A-matrices were obtained in the experimental xyz coordinate system. As the crystal structure belongs to the triclinic P1 ̅ space group, with only one molecule per asymmetric unit, there is one transformation matrix R relating the molecular and experimental coordinate frames. Determination of this matrix usually requires knowledge of the morphology of the crystal (i.e. assignment of the Miller indices of the major faces of the crystal).
However, given the high symmetry of the square planar coordination around the Co(II) centre, the directions of the eigenvectors of the g-matrix can be a priori established. Furthermore, as the CASSCF/NEVPT2 calculations which we performed reproduced the g-values remarkably well, the calculated eigenvectors could be used to establish the orientation of the g-matrix in the molecular frame.
The angular dependence of the g 2 -value in the three orthogonal planes xy, zx and zy is shown in Figure S17.  The orientation of the g-matrix in the molecular coordinate frame is shown in Figure S18. The g x and g y -values are contained in the equatorial CoN 4 plane, forming 27.8º and 24.3º angles with the Co-N bonds. The g z -value is normal to the equatorial CoN 4 plane.

Electronic Structure Calculations
In order to correlate the results from EPR experiments with the electronic structure and ligand field energetics of the different cobalt complexes we performed complete-active space self-consistent field (CASSCF) calculations, with multi-reference perturbation theory (NEVPT2) to take into account dynamical correlation effects.
All calculations were performed with the ORCA program package, version 4.0.1.2. [18] Single-crystal X-ray diffraction derived structures were used without optimization. For the truncated complex [Co(L1 N,N ) 2 ] Tr the apical CH 2 groups were removed and capping H atoms were added manually.
Multireference complete active space self-consistent field (CASSCF) calculations, [19][20] together with the N-electron valence perturbation theory of order 2 (NEVPT2) [21][22][23] were performed on the structures Ahlrichs type basis sets def2-SVP were used on all atoms except Co, for which def2-TZVP basis sets were used. [24] The resolution-of-the-identity (RI) approximation [25][26][27][28][29][30][31] with matching basis sets, [32][33][34] as well as the RIJCOSX approximation (combination of RI and chain-of-spheres algorithm for exchange integrals) were used to reduce the time of calculations. We also performed DFT calculations using the PBE functional [35] with def2-TZVP basis set on all atoms. Visualization of structures, orbitals and spin densities were done using the program Chemcraft. [36] We found that for all studied complexes the low spin multiplicity was the ground state, and that considering 20-30 doublet roots were enough to predict g-values in great agreement with experiment.   The magnetic and EPR properties of low spin cobalt(II) complexes have long been of interest to coordination chemists. One of the earliest theoretical analyses in this field were performed by Maki et al. [37] and later by McGarvey. [38] The latter author's study focused on low spin Co (II) complexes  The terms E(d xz ) and E(d z2 ) do not mean the energies of the orbitals, but the energies of the configurations where these orbitals contain the unpaired electron. Using the energies of the roots from the calculations in

Calculated IR Spectrum
We performed a single-point Density Functional Theory (DFT) calculation on the structure of [Co(L1 N,N ) 2 ], followed by a frequency calculation, in order to obtain the calculated IR spectrum. Due to the large size of the molecule and the computationally expensive nature of the frequency calculation, we used the pure-GGA PBE functional, together with a def2-TZVP basis set. Since IR spectra were acquired in the solid state, and DFT optimization would not take into account the crystal packing, the structure was not optimized. This factor, together with the relatively inaccurate theoretical level, results in only a qualitative reproduction of the experimental IR spectra (depicted in Figure S13). However it can be seen that the C-H stretching frequencies associated with the cyclohexyl substituent atoms not 12.3891 (9) 12.702 (1) 13.493 (1) 95.865 (3) 90.227 (3) 118.977 (2) 11.600 (2) 14.319 (3) 15.169 (3) 61.927 (4) 70.819 (7) 72.828 (5) 23.0013 (7) 20.3174 (6) 19.5333 (7