Crystal Structure and C−H Bond Cleaving Reactivity of a Mononuclear Co^IV-dinitrate Complex

Abstract

High-valent Fe^IV=O intermediates with a terminal metal-oxo moiety are key oxidants in many enzymatic and synthetic C−H bond oxidation reactions. While generating stable metal-oxo species for late transition metals remains synthetically challenging, notably, a number of high-valent non-oxo-metal species of late transition metals have been recently described as strong oxidants that activate C−H bonds. In this work, we obtained an unprecedented mononuclear Co^IV-dinitrate complex (2) upon one-electron oxidation of its Co(III) precursor supported by a tridentate di-anionic N3 ligand. 2 was structurally characterized by X-ray crystallography, showing a square pyramidal geometry with two coordinated nitrate anions. Furthermore, characterization of 2 using combined spectroscopic and computational methods revealed that 2 is a low-spin (S = 1/2) Co(IV) species with the unpaired electron located on the cobalt d_z2 orbital, which is well positioned for substrate oxidations. Indeed, while having a high thermal stability, 2 is able to cleave sp³ C−H bonds up to 87 kcal/mol to afford rate constants and kinetic isotope effects (KIEs) of 2–6 that are comparable to other high-valent metal oxidants. The ability to oxidize strong C−H bonds has yet to be observed for Co^IV-O and Co^III=O species previously reported. Therefore, 2 represents the first high-valent Co(IV) species that is both structurally characterized by X-ray crystallography and is capable of activating strong C−H bonds.

Graphical Abstract

Introduction

High-valent Fe^IV=O intermediates with a terminal oxo ligand have been identified as the key oxidants in many heme and nonheme metalloenzymes.^1,2 In the past 20 years, a large number of synthetic M^IV=O complexes of early and middle-transition metals including Mn and Fe have been obtained, shedding light on the structural and electronic properties and substrate C−H bond oxidation reactivities of the Fe^IV=O unit.^1,3 In contrast, high-valent terminal metal-oxo species are rarely reported for first-row late transition metals such as Co, Ni and Cu,^4–6 as predicted by classic bonding theory.^7,8 To date, only one Co^III=O complex supported by an anionic tris(carbene)borate ligand was structurally characterized by X-ray crystallography.⁴ In addition, very recently Roithova and co-workers described the generation of Co^III-oxo and Co^III-oxyl species in the gas phase. The Co^III-oxyl (equivalent to Co^IV-O) species is capable of cleaving strong C−H bonds as those in cyclohexane.⁹

Notably, a number of high-valent non-oxo-metal species of late transition metals have been recently described as strong oxidants that activate C−H bonds. Representative examples include a series of Ni(III) and Cu(III) complexes developed by the groups of Tolman,^10–12 McDonald,^13–15 Company¹⁶ and Brudvig.¹⁷ In this work, we obtained a high-valent Co^IV-dinitrate complex (Co^IV-(ONO₂)₂, 2) by one-electron oxidation of its Co(III) precursor. 2 is characterized structurally by X-ray crystallography, and its electronic structure has been probed using combined spectroscopic and computational approaches. Furthermore, 2 is able to cleave sp³ C−H bonds up to 87 kcal/mol at room temperature. Such reactivity of cleaving strong C−H bonds has yet to be observed for previously reported Co^IV-O^6,18 and Co^III=O species in the condensed phase.^4,5 Therefore, 2 represents the first high-valent Co(IV) species that is both structurally characterized by X-ray crystallography and is capable of activating strong C−H bonds.

Results and Discussion

Generation and characterization of 2

The starting complex in this work is a mononuclear Co^III-OH species (1-OH) supported by a di-anionic tridentate N3 ligand L (L = 2,6-bis((2-(2,6-dimethylphenylamino))isopropyl)pyridine), which is synthesized according to the procedure reported previously.¹⁹ The crystal structure of 1-OH (Figure 1A, Table S1) shows that the Co(III) center is in a slightly distorted square planar geometry with a Co-N_py bond length of 1.862 Å, two Co-N_amine bonds at an average distance of 1.836 Å, and one Co-OH bond of 1.792 Å. The shorter Co-N_amine bond length compared to that of the Co-N_py bond is a good indication that the cobalt center interacts more strongly with the anionic N-donors than with the neutral pyridine ligand. In acetonitrile, 1-OH exhibits a strong absorption at 570 nm (ε = 7300 M⁻¹ cm⁻¹, Figures 2 and S1). Furthermore, characterization of 1-OH using ¹H NMR (Figure S2) and electron paramagnetic resonance (EPR) spectroscopies confirms that 1-OH is a diamagnetic and EPR silent species, consistent with the assignment of the low-spin d⁶ Co(III) center.

Figure 1.

Crystal structures of (A) 1-OH, selected bond lengths: Co-N1, 1.862 Å; Co-N2, 1.833 Å; Co-N3, 1.840 Å; Co-O, 1.792 Å, and (B) 2, selected bond lengths and angle: Co-N1, 1.863 Å; Co-N2, 1.831 Å; Co-N3, 1.834 Å; Co-O1, 2.100 Å; Co-O2, 1.927 Å; O1-Co-O2, 81.96°. Thermal ellipsoids are drawn at the 50% probability level. Hydrogen atoms (except the one on the hydroxyl group) are ignored for clarity.

Figure 2.

(A) UV-vis spectra for the conversion of 1-OH (black) to 2 (red) by 1 equivalent of CAN in acetonitrile at room temperature. Inset: titration of CAN monitored at 570 nm (black) and 420 nm (red). (B) Perpendicular-mode X-band EPR spectrum of a 5 mM frozen-solution of 2 in acetonitrile at 30 K (black) and simulation (red). Experimental conditions: microwave frequency 9.6403 GHz, microwave power 7.962 mW, modulation amplitude 0.4 mT, modulation frequency 100 kHz, and time constant 163.84 ms.

The introduction of one equivalent of cerium(IV) ammonium nitrate (CAN, (NH₄)₂Ce^IV(NO₃)₆), a one-electron oxidant, into the blue solution of 1-OH at room temperature causes rapid decrease of the 570 nm absorption with the concomitant appearance of a new intense band at 420 nm (ε = 5900 M⁻¹ cm⁻¹) and two weaker features at 565 nm (ε = 2400 M⁻¹ cm⁻¹) and 750 nm (ε = 2000 M⁻¹ cm⁻¹), indicating the formation of a new species 2 (Figures 2A and S1). Two isosbestic points at 486 nm and 682 nm were observed, suggesting that no other intermediate was involved in the conversion of 1-OH to 2. Titration experiments further show that the stoichiometry between 1-OH and CAN is 1:1 (Figure 2A inset). The addition of excess CAN does not cause 2 to be formed in a higher yield, suggesting that 2 is likely a Co(IV) species. 2 has a lifetime of several hours at room temperature and is stable at cryogenic temperatures, which allows us to obtain its single crystals suitable for X-ray crystallographic studies. As shown in Figure 1B and Table S2, 2 is a neutral mononuclear five-coordinated Co^IV-(ONO₂)₂ complex. The source of nitrate ligands appears to be CAN. The Co(IV) center is in a pseudo-square pyramidal geometry with a N3O2 donor set. The Co-N_py/amine distances compare similarly with those in 1-OH. The Co^IV-ONO₂ bond length for the axial nitrate ligand is 2.10 Å, which is ~0.17 Å longer than that for the equatorial nitrate O-donor, indicating that the axial nitrate interacts less strongly with the cobalt center than the equatorial counterpart. The other two square pyramidal Co(IV) complexes previously characterized by X-ray crystallography also exhibited a long bond length between cobalt and the axial ligand. For example, an organometallic corrole-Co^IV-C₆H₅ species has an axial Co-C bond of 1.937 Å.²⁰ Another Co(IV) complex [Co^IV{S₂C₂(CF₃)₂}₂(PPh₃)] showed a long axial Co-P bond length of 2.22 Å.²¹ Furthermore, comparison of selected carbon-nitrogen and carbon-carbon bond lengths of the N3 ligand (Table S4) in the crystal structures of 1-OH and 2 (and also 1-NO₃, see below) clearly shows that the ligand scaffold is not impacted during the conversion of 1-OH to 2, indicating that the oxidation is a metal-centered instead of ligand-based process.

The perpendicular-mode, X-band EPR spectrum of 2 shows one broad resonance at g ≈ 2.2 (320 mT) at 30 K (Figure 2B). The negatively signed component of this signal shows seven resolved hyperfine lines attributed to the ⁵⁹Co (I = 7/2) nucleus (Figure S3). The EPR signal is unusually broad, spanning ca. 270–400 mT. The breadth of this signal, as well as the observation of ⁵⁹Co hyperfine coupling, are consistent with predominant localization of the unpaired spin on the Co center, as expected for Co(IV). While the EPR signal of 2 does not closely resemble any EPR signal previously reported for a Co(IV) complex, there are a limited number of examples of Co(IV) centers in square pyramidal environments for comparison (see Table S5).^18,20–23 Two Co(IV)-corrole complexes with an axial PPh₃ or C₆H₅ ligand displayed rhombic signals,^20,22 the later was described as having a delocalized valence, with the unpaired electron distributed between both the cobalt and the corrole macrocycle.²³ In another case, a Co(IV) bis(dithiolene) complex with an axial PPh₃ ligand showed a nearly axial EPR signal, with resolved hyperfine-splitting on both g_┴ and g_∥.²¹ Finally, a Co(IV)-oxo-Lewis acid complex with a tetraamido macrocyclic ligand showed a rhombic EPR signal with g-values from 2.57–2.03.¹⁸ The EPR spectrum of 2 can be well-simulated using g₁ = 2.23, g₂ = 2.08, g₃ = 2.00 (Figures 2B and S4), which fall within the range of g values reported for square-pyramidal, low-spin d⁵ Co(IV) systems (Table S5). The hyperfine coupling constant A(⁵⁹Co) is simulated with A₁ = 135, A₂ = 165, A₃ = 320 MHz. The large A₃ value is required to reproduce the hyperfine features readily observed in the experimental spectrum (the simulated values for A₁ and A₂ should be viewed with less certainty, although values in the range of 100 – 200 MHz were necessary to reproduce the breadth of the EPR signal; see Supporting Information). The A₃ value is higher than previously reported A values for Co(IV) centers (207, 156 and 72 MHz; see Table S5), which could be due to a higher spin density on Co(IV) center in 2.

2 can be reduced by ferrocene to re-generate a Co(III) species with an absorption band at 585 nm (ε = 9200 M⁻¹ cm⁻¹, Figures S1 and S6). Its absorption maximum red-shifted by ~15 nm compared to that of 1-OH, and we assign this new species to a nitrate bound Co^III-ONO₂ complex 1-NO₃. Titration of ferrocene to the solution of 2 (Figure S6 inset) reveals that 2 is generated in about 90% yield. 1-NO₃ can also be formed by reacting 1-OH with excess tetrabutylammonium nitrate (Figure S7). Furthermore, 1-NO₃ can be oxidized to 2 using stoichiometric amount of CAN (Figure S8). Therefore, the conversion between 1-NO₃ and 2 is chemically reversible. The crystal structure of 1-NO₃ (Figure S9, Table S3) shows that 1-NO₃ has a similar square planar structure as 1-OH with only one bound nitrate ligand. Therefore, the axial nitrate ligand of 2 must dissociate from the cobalt center during the reduction reaction.

Computational studies of 2

The electronic structure of 2 was examined using both DFT and CASSCF/NEVPT2 computations. The DFT-optimized structure of 2 (Figure 3A) shows metric parameters in excellent agreement with those observed in the X-ray crystal structure of 2 (Figure 1B). Notably, the long Co-ONO₂ axial distance is reproduced in the DFT structure, which suggests that this bond elongation is not an artifact of the solid-state sample but should persist in solution. This axial bond length plays an important role in defining the orbital ground state for an S = 1/2 Co(IV) system, as discussed below.

Figure 3.

(A) DFT-calculated structure of 2 with the corresponding bond distances (Å). (B) Co(IV) d orbital splitting pattern for 2 from DFT computations. Surface contour plots of quasi-restricted orbitals are shown for each d orbital.

The Co(IV) d-orbital splitting pattern for 2 reflects highly covalent interactions between the Co(IV) center and the N3 ligand. Using a conventional coordinate system, where the z axis is along the axial Co-ONO₂ bond and the x and y axes lie along the equatorial Co-ligand bonds (Figures 3B and S10, Table S6), the d_xy, d_xz, and d_yz orbitals are of π-type, and the d_z2 and d_x2-y2 orbitals are of σ-type. The Co(IV) d_xy and d_xz orbitals have the weakest interactions with the ligands and are therefore at lowest energy and doubly-occupied. Each of these orbitals shows some mixing with the π-system of the pincer ligand but has only minor interactions with the nitrate ligands (Figure 3B). In contrast, the Co(IV) d_yz orbital is ideally suited for strong π-interactions with the pincer ligand (Figure 3B). The strength of these π-interactions, coupled with the weak Co-ONO₂ axial interaction, places the Co(IV) d_yz MO at higher energy than the Co(IV) d_z2 MO (Figure 3B). Consequently, the Co(IV) d_z2 MO is the singly-occupied MO (SOMO) for this complex. At highest energy is the Co(IV) d_x2-y2 MO that is σ-antibonding with the pincer ligand and the equatorial nitrate ligand.

CASSCF calculations with NEVPT2 corrections offer a complementary means of examining the electronic structure of 2 (Figure S11, Table S7). Using a moderately-sized active space of 13 electrons in 9 orbitals (so-called CAS(13,9), see Supporting Information for details), the CASSCF/NEVPT2 calculations converge to a doublet ground state, consistent with the EPR results described above. This ground state is well separated from other states; the lowest-lying doublet and quartet excited states are near 5000 cm⁻¹. The ground state is dominated by a (d_xy)²(d_xz)²(d_z2)¹(d_yz)⁰(d_x2-y2)⁰ configuration (61%), with 12% admixture of a (d_xy)²(d_xz)⁰(d_z2)¹(d_yz)²(d_x2-y2)⁰ configuration. Thus, these multi-reference calculations identify the Co(IV) d_z2 MO as the SOMO, in accordance with the DFT results. The compositions of the CASSCF natural orbitals are similar to the DFT MOs, and reveal the strong π-bonding interaction between L and the Co(IV)d_yz orbital that places this orbital above the Co(IV) d_z2 MO (Figure S11). An isosurface plot of the spin density for the CASSCF wavefunction shows the localization of the spin on the Co(IV) d_z2 MO (Figure S12). The CASSCF/NEVPT2 calculations predict g values of 2.39, 2.07, and 2.01, in reasonable agreement with the values obtained by spectral simulations (2.23, 2.08, and 2.00; see Figure S4). The A values from the CASSCF/NEVPT2 computations are in more modest agreement with experiment (|A_NEVPT2| = 61, 326, 451 MHz; A_simulation = 135, 165, 320 MHz), although the uncertainty in two of the experimental components of the A matrix renders the comparison less meaningful.

The electronic structure computations offer some important insights into the potential reactivity of 2 in hydrogen atom transfer reactions (vide infra). In these reactions, the Co(IV) d_z2 SOMO is expected to accept an electron from substrate. The axial NO₃⁻ ligand could serve as the proton acceptor. The covalency between the Co(IV) d_z2 orbital and the axial NO₃⁻ ligand in the SOMO (Figure 3B) appears well-suited for facilitating proton and electron transfer in a concerted fashion.

C−H bond cleavage reactivity of 2

With the spectroscopic and computational characterization of the Co(IV) species 2 in hand, we sought to investigate its ability to activate sp³ C−H bonds. We first selected 9,10-dihydroanthracene (DHA, BDE = 78 kcal/mol) as a diagnostic substrate. As shown in Figure S13, the introduction of 250 mM of DHA into a solution of 0.1 mM 2 at room temperature causes its characteristic 420 nm band to disappear within a few minutes. Meanwhile, the growth of the absorption at 585 nm was observed, indicating that 1-NO₃ is the cobalt reaction product. The typical recovery yield of 1-NO₃ is ~70–80% and is independent of the amount of O₂ present in the reaction medium. These results suggest that 2 is a one-electron oxidant. Furthermore, the sharp optical features observed below 400 nm are indicative of the formation of anthracene as the oxidation product of DHA. Quantification of anthracene by GC-MS for the reaction shows that its formation yield is ~35%, accounting for ~70% of the oxidizing equivalent used. The formation yield of anthracene is independent of the amount of O₂ present in the solution as well. Other products include anthrone and anthraquinone in trace amounts.

The kinetic trace for DHA oxidation monitored at 420 nm can be well fitted using the first-order mode to obtain the pseudo-first-order rate constant (k_obs). Furthermore, k_obs correlates linearly with the concentration of DHA (Figure S14), which allows us to determine the second-order rate constant (k₂) of 0.079(3) M⁻¹ s⁻¹. A slower reaction was observed when DHA-d₄ was used as the substrate, affording a k₂ of 0.034(2) M⁻¹ s⁻¹. The H/D kinetic isotope effect (KIE) of 2.3 indicates that the cleavage of a C−H bond by 2 is the rate-determining step for DHA oxidation.

We then extended our investigation to the reaction of 2 with other substrates having C−H bond strength in a range of 76–87 kcal/mol, including xanthene, cyclohexene, tetralin and ethylbenzene (Figure S15, Table S8). As predicted, the rate constant is slower when the C−H bond being cleaved becomes stronger. Specifically, 2 reacts with xanthene to afford a k₂ of 0.15(5) M⁻¹ s⁻¹ and a KIE of 3.3 (Figure S16). The benzylic C−H bond of ethylbenzene represents the strongest bond that 2 can cleave, affording a k₂ of 0.0033(1) M⁻¹ s⁻¹ and a KIE of 5.4 (Figure S17). The oxidation products of ethylbenzene include styrene and acetophenone. For this group of substrates, the log k₂’ (k₂’ is normalized k₂ on a per hydrogen basis) correlates linearly with the strength of the C−H bond being cleaved with a slope of −0.13 (Figure 4).

Figure 4.

Plots of log k₂’ as a function of the C−H bond strength for substrate oxidation by 2 in acetonitrile at 25 °C (black) and (13-TMC)Co^IV-O at −40 °C (red, data from ref. 6). The lines represent the best linear fittings.

We further carried out electrochemical studies of these cobalt complexes to gain more insights about the C−H bond cleavage reactivity of 2. As shown in Figure S18, the cyclic voltammogram of 1-NO₃ exhibits an anodic peak at 0.67 V vs. ferrocene, with the corresponding cathodic feature at 0.43 V upon returning the scan at 1.10 V. The peak currents of these two processes are comparable; however, the peak separation of 0.24 V is much larger than 0.06 V expected for a reversible electron transfer. This phenomenon typically indicates that the redox process is coupled to a chemical reaction, likely the coordination/dissociation of the second nitrate ligand in the axial position during the conversion between 1-NO₃ and 2, as clearly observed in the crystal structures of these two complexes. We assign the mid-point potential (0.55 V vs. ferrocene) to the Co^IV/III couple. On the other hand, attempts to measure the pK_as of the nitrate ligands in complex 2 were not successful. Instead, we use the pK_a of free nitrate anion in acetonitrile (8.8) as an estimate.^24,25 This value should be considered as the upper bound of the pK_a, as the coordination of the nitrate to the Co(IV) center should make these anions less basic. Following the protocol of Mayer in applying the Bordwell-Polanyi relationship to C−H bond oxidations by a set of metal-oxo complexes (eq. 1),^26,27 we estimated the Co^III(O−H)NO₂ bond strength of the bound nitrate ligand that is protonated after hydrogen atom abstraction by 2 to be <84 kcal/mol. This value appears to be consistent with our observation that 2 is able to cleave C−H bonds up to 87 kcal/mol.

$${\text{D}}_{\text{O}-\text{H}}=23.06E+1.37\text{p}{K}_{\text{a}}+\text{C}$$ (1)

We propose a mechanism (Scheme 1) that summarizes our experimental and computational findings. The oxidation of 1-OH to generate 2 is a one-electron process coupled with the substitution of hydroxide by nitrate and the coordination of the second nitrate ligand. 2 cleaves the C−H bond of the substrate by hydrogen atom transfer to generate a carbon radical and 1-NO₃. While Co(IV) is reduced to Co(III), one nitrate ligand must be protonated and dissociate from the cobalt center. According to the crystal structure and calculations of 2, we believe that the weakly bound axial nitrate ligand plays such role in the course of C−H bond cleavage. The classic radical rebound does not occur after C−H bond cleavage because 2 is a one-electron oxidant. The substrate radical is likely oxidized by another molecule of 2 to generate the desaturated product or is trapped by O₂ in the solvent to form the oxygenated product(s).

Scheme 1.

Schematic illustration of the reactions described in this work.

2 thus represents the first Co(IV) complex that is both characterized by X-ray crystallography and is able to cleave strong sp³ C−H bonds in the condensed phase. Compared to 2,6-pyridinedicarboxamidate, another di-anionic tridentate ligand that is used to generate Ni(III) and Cu(III) complexes, the N3 ligand employed in this work lacks the electron-withdrawing carbonyl groups, making the anionic N-donors more electron-donating to stabilize the high-valent Co(IV) state and preventing possible coordination isomerism.²⁸ Our ongoing effort aims to use this ligand to generate high-valent species of other late-transition metals and to investigate their oxidation chemistry.

Notably, while having a high thermal stability, 2 is capable of activating sp³ C−H bonds up to 87 kcal/mol, which has yet to be observed by previously reported Co^IV-O^6,18 and Co^III=O species in the condensed phase.^4,5 The plot of the logarithm of the normalized k₂ values as a function of the strength of the C−H bond being cleaved yielded a linear correlation with a slope of −0.13 (Figure 4). This slope is less negative than that (−0.31) of (13-TMC)Co^IV-O, an oxocobalt(IV) species recently characterized by Nam and co-workers (Figure 4).⁶ Interestingly, further comparisons of slopes of such BDE plot for a variety of oxo-metal and non-oxo-metal complexes (see Table S9 and references therein) revealed that, in terms of the C−H bond cleavage reactivity, 2 is less sensitive to the C−H bond strength of the substrate than most of other hydrogen atom abstraction reagents (typical slopes = −0.2 to −0.5). The basis for shallow slopes in C−H bond oxidation reactions has been discussed, but a consensus opinion regarding the origin for these shallow slopes is lacking.^29,30 Intuitively, it would indicate that the reaction proceeds through an early transition state, at which the C−H bond is not fully broken.

Furthermore, the rate constant for ethylbenzene oxidation (0.0033 M⁻¹ s⁻¹) is comparable to that of a Cu^III-acetate species (0.0016 M⁻¹ s⁻¹),¹³ but is about 5 to 50-fold slower than those of reported Ni(III) oxidants.^15,16 On the other hand, 2 is unable to cleave C−H bonds stronger than 87 kcal/mol. This thermodynamic limit is consistent with the estimated Co^III(O−H)NO₂ bond strength of <84 kcal/mol. Our current effort is focusing on modifying the N3 ligand in order to generate more reactive Co(IV) complexes and study their structure-reactivity relationship.

Conclusion

In summary, we have obtained a novel mononuclear Co^IV-dinitrate complex 2 supported by a di-anionic tridentate ligand upon one-electron oxidation of its Co(III) precursor. 2 is fully characterized using spectroscopic and computational approaches, including X-ray crystallography, EPR, UV-vis and DFT and CASSCF calculations, revealing that it is a low-spin (S = 1/2), square pyramidal Co(IV) complex with the unpaired spin located on the cobalt center. Importantly, 2 is reactive with sp³ C−H bonds up to 87 kcal/mol to afford decent rate constants at room temperature. These interesting findings clearly demonstrated that stable but reactive high-valent Co(IV) species can be generated using the rigid, electron-donating N3 ligand scaffold. This seed contribution inspires future development of more reactive high-valent late-transition metal complexes for C−H bond activation.