Synthesis and characterization of a formal 21-electron cobaltocene derivative

Metallocenes are highly versatile organometallic compounds. The versatility of the metallocenes stems from their ability to stabilize a wide range of formal electron counts. To date, d-block metallocenes with an electron count of up to 20 have been synthesized and utilized in catalysis, sensing, and other fields. However, d-block metallocenes with more than formal 20-electron counts have remained elusive. The synthesis and isolation of such complexes are challenging because the metal–carbon bonds in d-block metallocenes become weaker with increasing deviation from the stable 18-electron configuration. Here, we report the synthesis, isolation, and characterization of a 21-electron cobaltocene derivative. This discovery is based on the ligand design that allows the coordination of an electron pair donor to a 19-electron cobaltocene derivative while maintaining the cobalt–carbon bonds, a previously unexplored synthetic approach. Furthermore, we elucidate the origin of the stability, redox chemistry, and spin state of the 21-electron complex. This study reveals a synthetic method, structure, chemical bonding, and properties of the 21-electron metallocene derivative that expands our conceptual understanding of d-block metallocene chemistry. We expect that this report will open up previously unexplored synthetic possibilities in d-block transition metal chemistry, including the fields of catalysis and materials chemistry.


Preparation of [Mn(CpNCp)] (4)
In a nitrogen glovebox, a 20 mL vial equipped with a Teflon coated stirring bar was charged with MnBr2 (22.0 mg, 0.102 mmol) and 4 mL THF. The vial was sealed with a plastic cap, and the solution was stirred, and heated intermittently using a heat gun until all pink chunks of MnBr2 dissolved. The resulting slightly cloudy colorless solution was cooled to 25 °C. To the solution was then added a solid of Na2CpNCp (28.5 mg, 0.102 mmol) portionwise for ca. 2 min. The vial containing Na2CpNCp was washed with 1 mL THF, and the THF solution was added to the MnBr2 solution. A yellow solution and white precipitate were formed as soon as Na2CpNCp was added. The solution was stirred for 20 min and concentrated to dryness. The product was extracted using ether and the yellow ether solution was filtered using plug of Celite, and the Celite was

Examination of possible coordination of pyridine to cobaltocene
In a nitrogen glovebox, a 4 mL vial was charged with 10 μL Me3SiOSiMe3, 0.50 mL pyridine, and 0.10 mL C6D6. Cobaltocene (5.7 mg, 0.030 mmol) was completely dissolved in 0.50 mL of this solvent mixture, and the solution was transferred to a J. Young NMR tube. To the tube was then added a glass capillary containing the same solvent mixture. Effective magnetic moment (μeff) of the complex was calculated using 1 H NMR spectra recorded at 80, 25, and -40 °C, and was 1.8, 1.9, and 1.9 μB, respectively. The slight change of μeff is likely due to change of solvent density upon temperature changes, which causes slight change of solution concentration. 1 H NMR signals with chemical shift similar to 1 were not detected. The NMR sample was transferred to an EPR tube in a nitrogen glove box, and EPR spectra was recorded at 4.2 K. Signals expected for the formation of pyridine-coordinated, S = 3/2 cobaltocene species, [CoCp2(pyridine)], were not detected.

EPR studies
EPR samples were prepared in a nitrogen glovebox using 5 mm outer diameter quartz EPR tubes sealed with rubber septums and parafilm. Continuous wave (cw) X-band EPR spectra of 1 were recorded using a JEOL JES-X330 X-band EPR spectrometer equipped with a liquid helium cooling system. The instrumental parameters employed were as follows: power: 1 mW; time constant: 100 ms; experimental frequency: 9.079 GHz; modulation amplitude: 40 and modulation width: ±0.06 mT.
Pulse EPR measurements were performed using a Bruker ESP380E spectrometer equipped with an Oxford liquid He-flow temperature controller. The spectrometer is customized with a TTL/ECL conversion unit and controlled by the SpecMan4EPR software 2 . Cw-and pulse-EPR spectra were simulated by using the EasySpin software 3,4 .

Supplementary Fig. 28.
Hahn-echo-detected EPR spectra of 1 observed at 4 K. Hahn-echo signal intensities were sensitive to the applied pulse length and pulse-duration time. The middle EPR spectrum dominantly includes the Z components of the spin system in the angular selective manner. Below 300mT, the intensity gradually decreases due to the fast phase-memory time T2, indicating the difficulties of the detection pulse-EPR signals in the lower magnetic field range.

Supplementary Fig. 29.
Spin-lattice relaxation time T1 of 1 observed at 4 K. The spin-lattice relaxation time T1 was measured by the three-pulse inversion recovery method. The value of T1 gradually decreases when the magnetic field going down, indicating the spin-orbit interaction on the Co ion contribute to the acceleration of T1. Four-pulse HYSCORE spectra of 1 at 4 K. (a) Contour plots observed at 320 mT. (b) Simulated spectra using the EasySpin toolbox. Orientation selectivity at 340 mT was given on the right. Only two-protons in addition to 14 N nucelar spin were considered for the simulation because of the reduction of the size of spin-Hamiltonian.   All structures were solved by the intrinsic phasing approach using SHELXT-2018/2 and refined by the full-matrix least-squares on F 2 using SHELXL-2018/3 5,6 . Calculations were mainly performed using the WinGX-2021.3 suite of programs 7 . Non-hydrogen atoms were refined anisotropically. Hydrogen atoms were inserted at the calculated positions and refined as riding atoms. The disorder, if present, was resolved using free variables and reasonable restraints on geometry and anisotropic displacement parameters.
The multipole refinement of 1 was performed within the Hansen-Coppens multipole formalism as implemented in the MoPro software package 8 . The total electron density is considered as a superposition of pseudo-atomic electron densities, expressed as a sum of spherical core, spherical valence, and deformation valence contributions (37): where , , and are the core, monopole, and multipole populations, and and ′ are the atomic spherical and deformation valence expansion/contraction parameters, respectively. are the Slater-type radial functions. are the atom-centered real spherical harmonics. Core and spherical valence scattering factors derived from the relativistic analytical wave functions of Su and Coppens were utilized 9 . The exponential Slater-type radial functions with radial function parameters = 4,4,4,4 for Co,= 2,2,3,4 for N and C, 2 = 1 for H-atoms and the values of orbital exponents = 7.6496, = 3.8106, = 3.1303, and = 2.0000 were used. The multipole refinement was performed against with reflections that satisfy the > 1.5 ( ) condition. A reciprocal resolution sin( max ⁄ ) of the data was 1.26 Å -1 . The function minimized in the least-squares procedure was ∑ (| | − | |) 2 with weight equal to 1/(3.4293 2 ( 2 )).
The first and fourth scale factors were refined. The unit cell electroneutrality constraint was imposed during the whole refinement procedure. The C-H bond distances, -and ′ -parameters for hydrogen pseudo-atoms, and ′ -parameters for all pseudo-atoms were constrained to the theoretical values obtained from the DFT-optimized periodic structure. The same ′-parameter was used for all multipole levels of each pseudo-atom; 00 was fixed at zero. Multipole expansion was truncated at the hexadecapolar level ( max = 4) for the Co1, N1, C11-C15, and C21-C25 pseudo-atoms and octupolar level ( max = 3) for the C1-C7 pseudo-atoms. For each H-pseudoatom, only a monopole and a bond-oriented dipole 10 were refined. The anharmonic atomic motion of Co1 was modeled using the Gram-Charlier expansion of temperature factors up to the tensors of 4 th rank. A block refinement of the Gram-Charlier coefficients was applied by analogy with the procedure described earlier 10 . Anisotropic displacement parameters for the H-atoms were calculated using the SHADE3 algorithm and inserted multiple times between the refinement steps until no further change was achieved 11 . The analysis of multipole-derived electron density was carried out in WinXPRO 12,13 by analogy with the published procedures 14,15 . The experimental results were compared with the theoretical ones calculated for the optimized free molecule; for this, the Multiwfn program was applied [15][16][17] . Absolute structure parameter is -0.081 (9). Volume fraction of the minor twin component is 0.385(1).
Crystallographic data for 4.

DFT studies Test Computations.
Adding two electrons to the 19 electron configuration of cobaltocene gives either S = 1/2 or S = 3/2. Therefore, we initially optimized the molecular structure derived from SC-XRD in the doublet and quartet state. We considered a diverse set of dispersion-corrected (D4 unless otherwise noted) DFT levels such as GFN2-xTB 18,19 , B97-3c 20 , BP86 [21][22][23] , TPSS, M06-L 24,25 (using D3) 26 , TPSSh [27][28][29]and PBE0 30,31 . Optimizations in the doublet state result in scission of the Co−N bond at any level to afford a local minimum structure with formal 19-electrons (1'). In the quartet state, the 21 valence electron complex with a Co−N bond is the minimum structure. The computed Co−Cpcent bond lengths agree favorably with those obtained by single crystal X-ray diffraction at all levels of theory ( Supplementary Fig. 41A). However, for some functionals, the optimized Co−N bond differs notably from the experimentally determined one (Supplementary Fig. 41A). The influence of scalar-relativistic effects on the bond lengths is minor and BP86 and TPSS perform best. We, therefore, considered the non-relativistic TPSS level to be suitable for structure optimizations. Then, we carried out single-point energy computations to calculate the Gibbs free energy difference between the isomers 1 (S = 3/2, formal 21-electrons) and 1' (S = 1/2, formal 19electrons). At the TPSS level, 1' is 3.0/1.9 kcal/mol lower in energy at the scalar-relativistic/nonrelativistic level, respectively. However, at the TPSSh level, 1 is favored by 3.8/5.0 kcal/mol at the scalar-relativistic/non-relativistic level, respectively. Therefore, we used the non-relativistic TPSSh-D4/def2-QZVPP//TPSS-D4/def2-TZVPP level of theory for computations of the electronic structure of 1 and 1'.    (5).
[c] The net charge attributed to the entire Cp and pyridine parts is −0.52 and +0.01, respectively.
The positive charge of the cobalt is neutralized mainly by the negatively charged Cp parts, while the pyridine part is neutrally charged. The charge distribution over the Cp ligands is uniform (−0.08 to −0.13 for the individual C atoms with summed H atoms).