A Rare Low‐Spin CoIV Bis(β‐silyldiamide) with High Thermal Stability: Steric Enforcement of a Doublet Configuration†

Abstract Attempted preparation of a chelated CoII β‐silylamide resulted in the unprecedented disproportionation to Co0 and a spirocyclic cobalt(IV) bis(β‐silyldiamide): [Co[(NtBu)2SiMe2]2] (1). Compound 1 exhibited a room‐temperature magnetic moment of 1.8 B.M. and a solid‐state axial EPR spectrum diagnostic of a rare S=1/2 configuration for tetrahedral CoIV. Ab initio semicanonical coupled‐cluster calculations (DLPNO‐CCSD(T)) revealed the doublet state was clearly preferred (−27 kcal mol−1) over higher spin configurations only for the bulky tert‐butyl‐substituted analogue. Unlike other CoIV complexes, 1 had remarkable thermal stability, and was demonstrated to form a stable self‐limiting monolayer in preliminary atomic layer deposition (ALD) surface saturation experiments. The ease of synthesis and high stability make 1 an attractive starting point to investigate otherwise inaccessible CoIV intermediates and for synthesizing new materials.

Cobalt complexes are effective and cost-efficient alternatives to Ir-and Rh-based catalysts for a number of important C À H bond transformations, [1] alkynylations, [2] alkylations, and arylations. [3] Recently, high-valent Co IV species have also been shown to affect CÀH bond activation. [4] High-valence Cocontaining materials like the electrode material Li 2Àx CoO 2Àd are also essential as the cathode material of Li-ion batteries, [5] and others such as CoS x have been predicted to become key materials in future green energy photocatalysts. [6] The facile synthesis of robust Co IV complexes has remained a challenge, so their exploitation in the field of catalysis as active species and in materials science as precursors for material synthesis has remained limited. Chemical vapor phase (CVD) techniques such as atomic layer deposition (ALD) have become one of the most preferable approaches for the fabrication of exotic, emerging materials. [7] So it stands to reason to evaluate Co IV complexes not only in light of their electronic nature but also with respect to their physico-chemical properties which might qualify them as CVD/ALD precursors.
To date, most Co IV complexes have been prepared by 1 e À oxidation of Co III complexes with Br 2 , Fe(ClO 4 ) 3 , or by cyclic voltammetry. The starting Co III complexes typically comprise sterically demanding tetraamido or pyrrolato macrocycles, and the resulting Co IV species are commonly unstable above room temperature. In cases where unambiguous Co IV species could be isolated, geometries are typically limited to square planar and square pyramidal. [8] Macrocyclic Co IV oxo-congeners such as the one introduced by Wang et al. are typically more robust but still suffer from this geometrical restriction. [9] For d 5 ions, the high-spin S = 5 = 2 configuration exhibits the maximum exchange energy for transition metals; thus, examples of low-spin d 5 are rare. This is especially the case for tetrahedral geometries where the ligand field splitting is typically small in relation to octahedral fields. However, for Co IV , the increased oxidation state affords larger ligand fields and so a finer energetic balance is expected between high-and low-spin electronic configurations. Interesting examples are Dengs trigonal-planar, N-heterocyclic carbene stabilized Co IV bis-imide [10] and Groysmans tricoordinate, diarylcarbene-stabilized Co IV bis-alkoxide. [4] While computational studies showed that the assignment of the spectroscopic oxidation state for the metal centers was ambiguous, their magnetic properties supported a S = 1 = 2 ground state in each case. More recently, a squashed-tetrahedral Co IV tetrakis-(ketimide) Co(N = C t Bu 2 ) 4 was reported, but it exhibited a quartet ground state (S = 3 = 2 ). [11] Co IV tetrakis(1-norbonyl), Co(nor) 4 , has remained the only known unambiguous example of a tetracoordinate low-spin Co IV complex for over 40 years. [12,13] b-Silylamines are excellent ligands to stabilize unusual coordination numbers and oxidation states for elements across the periodic table. [14] The good thermal stability and high volatility of the well-known three-coordinate Co II complex {Co[N(SiMe 3 ) 2 ] 2 } 2 prompted us to explore other Co b-silylamides as potential precursors for vapor phase deposition of Co-containing materials.
Treatment of suspended CoCl 2 (TMEDA) (TMEDA = tetramethylethylenediamine) with one equivalent of (LiN t Bu) 2 SiMe 2 in pentane led to a rapid change in color from ultramarine to dark brown-black. Dark gray particles were observed to precipitate from solution alongside LiCl following the disproportionation of the presumed Co II monokis(diamide) intermediate to form 1 and elemental Co 0 . Workup of the reaction solution afforded 1 as a darkbrown, almost black crystalline solid that was purified either by recrystallization from pentane at À49 8C or vacuum sublimation (85 8C/10 mTorr). The synthesis scaled well and satisfactory yields (60 % from the ligand) were obtained from 10 mmol batches with ease.
Single-crystal X-ray crystallography revealed 1 was a fourcoordinate inorganic spirocycle of D 2d symmetry ( Figure 1). With a Yang geometry index value of t 4 = 0.75 (Figure 2), [15] it lies between the two other known four-coordinate homoleptic Co IV complexes: Haytons Co(N = C t Bu 2 ) 4 (t 4 = 0.59) [11] and Co(nor) 4 (t 4 = 0.94). [13] The average Co À N bond length of 1.84 was only slightly longer than in Co(N = C t Bu 2 ) 4 (1.80 ), and notably shorter than the CoÀC bond in Co(nor) 4 (1.92 ). Coordination around each nitrogen was planar, typical of silylamides due to electrostatic repulsion of the polarized Si À N bonds. The ligand planes were arranged almost orthogonally in 1 (f = 86.98), contrasting Haytons complex which was twisted toward D 4h symmetry (f = 57.18). Both shared the same CoÀN coordination sphere, but the increased steric bulk and geometric rigidity of the b-silyldiamide locked the ligands on 1 into an orthogonal arrangement.
This being the case, we expected significant differences in the electronic structure of 1 and Co(N=C t Bu 2 ) 4 .
Room-temperature magnetic susceptibility measurements of 1 (Gouy method) revealed an effective magnetic moment of m eff = (1.8 AE 0.1) B.M. in the solid state, suggesting 1 adopts an unusual low-spin doublet (S = 1 = 2 ) ground state configuration. This was supported by solid-state EPR spectra which displayed axial spectra consistent with a doublet state (g z = 2.6, g x = g y = 2.0, Figure 3). Variable-temperature EPR studies showed no significant dependence on temperature from 125-300 K ( Figure S12). The g-tensor is similar to that of a low-spin four-coordinate d 5 Fe III complex reported by Peters which adopted C 3v symmetry (g z = 2.9, g x = 2.0, g y = 2.0). [16] The doublet ground state was also supported by density functional theory (DFT) calculations, as we calculated g z = 2.4, g x = g y = 2.0 for the doublet state and g z = g x = g y = 2.0 for the quartet and sextet states.
We turned to DFT and ab initio coupled-cluster theory to understand the electronic structure of 1. Accurate prediction of spin states with DFT is an ongoing challenge because the exact exchange-correlation functional is still unknown. While mixing a portion of exact Hartree-Fock (HF) exchange into hybrid functionals generally improves performance for closed-shell molecules, these methods are biased toward high-spin transition metal configurations. [17] Nevertheless, all of the pure functionals (BLYP, OLYP, PBE, OPBE) and one hybrid functional (PW6B95) we employed accurately reproduced the experimental structure of 1 when a doublet ground state was considered. The quartet and sextet ground state structures did not: the quartet deviated by exhibiting a smaller ligand interplane angle (f % 728), and the sextet had elongated CoÀN bonds (ca. 1.94 ). The overall geometric consistency between functionals within different spin states   suggested the structures were reasonable, and we proceeded with ab initio calculations for higher accuracy.
Unambiguous assignment of the doublet ground state to 1 was achieved with ab initio semicanonical coupled-cluster calculations of our best DFT-optimized structures (DLPNO-CCSD(T)/def2-TZVPP//PW6B95/def2-SVP). This recent method provides comparable performance to canonical CCSD(T), the "gold standard" of quantum chemistry, but with computational cost comparable to DFT. [18] The doublet was calculated to be 28.1 kcal mol À1 and 27.1 kcal mol À1 more stable than the quartet and sextet, respectively. Our prior DFT calculations had substantial errors of + 22 kcal mol À1 and + 24 kcal mol À1 for the pure and hybrid functionals, respectively. We hope that these benchmarked results of a difficult case may act as reference values in the future.
Only three Co IV complexes with unambiguous doublet ground states have been reported before: Co(nor) 4 , Dengs Co(=Ndipp) 2 (NHC) (dipp = 2,6-diisopropylphenyl, NHC = N-heterocyclic carbene), and Groysmans Co(OR) 2 (=CPh 2 ). But their apparent oxidation states are their only similarity to 1. Each belongs to a different symmetry point group (T d and C 2v , respectively); their bonds are distinct from the Co À N bonds in 1; Co( = Ndipp) 2 (NHC) and Co(OR) 2 ( = CPh 2 ) are heteroleptic and three coordinate; and the oxidation states in Co(=Ndipp) 2 (NHC) and Co(OR) 2 (=CPh 2 ) were tentatively assigned alongside possible Co III resonant forms for which no precedence was found in 1. Collectively, these dissimilarities made them less useful in explaining the doublet state of 1. Despite their different ground states, consideration of Haytons Co(N = C t Bu 2 ) 4 was instructive as it shared the same Co-N coordination sphere as 1.
We constructed a qualitative molecular orbital (MO) diagram of the doublet and quartet configurations of 1 using DFT (Figure 4). Although both states had molecular D 2d symmetry, the molecular orbitals were mostly of lower symmetry in the D 2 point subgroup. Exceptions to this were observed for unpaired electrons: the doublet SOMO (SO = singly occupied) retained full D 2d symmetry, but the quartet SOMOs were of even lower C 1 symmetry. The doublets MOs were predictable from its symmetry, with the orthogonal ligand orientation stabilizing s-and p-bonds involving the b 2 (d xz ) and b 3 (d yz ), and b 1 (d x 2 Ày 2 ) DOMOs (DO = doubly occupied), respectively. The corresponding p*-antibonding b 1 (d x 2 Ày 2 ) SOMO showed p-backbonding in doublet 1, which was also observed by Mulliken spin-population analysis (0.52 e on Co and 0.52 e across four nitrogens). Despite their differences, we found an important parallel between the quartet state of 1 and Haytons Co(N = C t Bu 2 ) 4 . The twisted geometry and quartet ground state of Co(N = C t Bu 2 ) 4 was due to p-backbonding of the unpaired electrons to the N=C pbonds on the ligands, which was maximized with D 2d symmetry and S = 3 = 2 . [11] Clear p-backbonding to the N porbital was observed in the three highly distorted SOMOs of quartet 1, with more than one unpaired electron residing on the nitrogens (1.79 e on Co and 1.21 e across four nitrogens). Unfortunately, we could not reliably deconvolute the energetic contributions of these p-systems by energy decomposition analysis (EDA) because of the poor performance of DFT for 1.
Steric bulk clearly played an important role in stabilizing the doublet ground state of 1, so we optimized the geometries and computed the doublet-quartet energy gaps for less bulky N-substituents in the same way as for 1 to quantify its effect ( Figure 5). All the successfully optimized doublet and quartet geometries had features consistent with those in 1, and their DFT energy gaps also did not agree with the ab initio results. From the reliable ab initio calculations, we found that reducing the bulk from just tert-butyl to isopropyl made the quartet the most stable electronic configuration by a large margin (DE =+ 25 kcal mol À1 ). The SOMOs and spin populations of the quartet states were essentially independent of steric bulk, suggesting that the ligand plane angle f was dependent on the N-substituent more than on the extent of pbackbonding. Additionally, the geometric effect of the pbackbonding was clearly visible in the spin density plots of 1. These results together demonstrate how the tert-butyl sub-  stituents bulk was necessary to destabilize the otherwise preferred quartet ground state and it enforced the doublet ground state by orthogonal ligand arrangement. Analogues of 1 with less steric bulk are expected to have electronic structures similar to Co(N=C t Bu 2 ) 4 .
We tested the robustness of 1 to examine its suitability for vapor phase deposition. As prior introduced, ALD is especially good for depositing metastable phases and exotic materials because of its self-limiting chemical mechanisms and mild operating conditions. [19] To our knowledge, there have been no reports of either ALD or CVD of Co IVcontaining materials because of the lack of Co IV precursors. None of the few previously reported Co IV complexes displayed sufficient thermal stability or volatility. Co(nor) 4 decomposes at 100 8C and cannot be sublimed, Haytons Co(N = C t Bu 2 ) 4 decomposes in solution at ambient temperature, Dengs Co( = Ndipp) 2 (NHC) undergoes reductive C À H bond activation at 50 8C, while Groysmans Co(OR) 2 (=CPh 2 ) is likely nonvolatile due to its high molecular weight and numerous phenyl groups.
We quantitatively evaluated the suitability of 1 as a precursor with respect to its thermal stability and volatility by thermogravimetric analysis (TGA) and differential thermal calorimetry (DSC). When a 10.1 mg sample was heated with a linearly increasing temperature ramp, a clean single-step mass loss was observed between 150-230 8C leaving behind a small residual mass of 3.2 %, indicating that most of the sample evaporated during the experiment (Figure 6 a). TGA can also be used to efficiently estimate vapor pressure by employing the Langmuir equation, and the temperature at which 1 displays 1 torr of vapor pressure was estimated with a Clausius-Clapeyron model to be (150.4 AE 0.1) 8C (Figure 6 b). [20] DSC revealed the melting point to be 143 8C and the onset of decomposition (defined as 5 % of the maximum of the first exothermic event) [21] to be 197 8C. The "thermal range" between the 1 torr vapor pressure temperature and the decomposition temperature is often a practical and reliable benchmark of a compounds practicality as an ALD precursor. With a thermal range spanning roughly 47 8C, compound 1 represents a promising precursor candidate for ALD.
To contextualize the performance of 1, we subjected two recently reported Co II precursors to thermal analysis: namely CoCl 2 (TMEDA) [22] and Co(DAD) 2 [23] (DAD = tert-butyldiazadienyl) (see the Supporting Inforamtion). We applied our recently developed "Figure of merit" s that takes key parameters such as thermal range, vapor pressure, and extent of decomposition during TGA into account to quantitatively compare precursors (Table S2). [21,24] A positive Figure of merit indicates the suitability of a precursor while negative indicates a poor candidate. With s = 35, complex 1 ranks between [Co(DAD) 2 ] (s = 64) and [CoCl 2 (TMEDA)] (s = À2) where the weak performance of the latter originates from its low volatility. Therefore, 1 was not only competitive with successfully applied Co II precursors but, to the best of our knowledge, represents the only known potential Co IV precursor candidate.
The most important characteristic of an ALD precursor is the ability to form a stable self-limiting monolayer on a surface. [25] After all reactive surface sites have undergone a reaction with precursor molecules, no further adsorption should occur, allowing conformal and uniform atomic layer deposition to take place during the next self-limiting reaction. [26] Employing a quartz crystal microbalance (QCM) in our home-built ALD tool, we investigated the potential for 1 to fulfill this basic yet critical requirement. The precursor was heated to 135 8C for delivery and the QCM crystal was heated to 150 8C for saturation experiments. Compound 1 demonstrated self-limiting adsorption on alumina with a mass gain of (49.2 AE 0.3) ng cm À2 that was stable over roughly 35 minutes (Figure 6 c). This encouraging initial result motivated us to begin developing ALD processes with 1 in ongoing research.
In summary, a thermally stable, volatile, homoleptic, spirocyclic Co IV bis(b-silyldiamide) was synthesized in a facile one-step salt-metathesis exploiting Co II disproportionation. Experiment and theory confirmed that it exhibits a rare lowspin d 5 doublet configuration enforced by the steric bulk and geometric rigidity of its ligands. The remarkable stability of 1 makes it a promising precursor for vapor deposition, and its ability to form a stable self-limiting monolayer makes it a promising ALD precursor. Beyond this, 1 represents a simple, inexpensive, and accessible starting material for further high-valent Co IV chemistry to be explored.