A Highly Reactive Co^III,IV₂(μ-O)₂ Diamond Core Complex that Cleaves C-H Bonds

Abstract

The selective activation of strong sp³ C–H bonds at mild conditions is a key step in many biological and synthetic transformations, and an unsolved challenge for synthetic chemists. In nature, soluble methane monooxygenase (sMMO) is one representative example of nonheme dinuclear iron-dependent enzymes that activate strong sp³ C–H bonds by a high-valent diiron(IV) intermediate Q. To date, synthetic model complexes of sMMO-Q have shown limited abilities to oxidize strong C–H bonds. In this work, we generated a high-valent Co^III,IV₂(μ-O)₂ complex 3 supported by a tetradentate TPA ligand via one-electron oxidation of its Co^III₂(μ-O)₂ precursor 2. Characterization of 2 and 3 using X-ray absorption spectroscopy and DFT calculations showed that both species possess a diamond core structure with a short Co•••Co distance of 2.78 Å. Furthermore, 3 is an EPR active species showing an S = 1/2 signal with clear hyperfine splittings originated from the coupling of the ⁵⁹Co nuclear spin with the electronic spin. Importantly, 3 is a highly reactive oxidant for sp³ C–H bonds, and an oxygenation reagent. 3 has the highest rate constant (1.5 M⁻¹ s⁻¹ at −60 °C) for oxidizing 9,10-dihydroanthracene (DHA) compared to diamond core complexes of other first-row transition metals including Mn, Fe and Cu reported previously. Specifically, 3 is about 4–5 orders of magnitude more reactive than the diiron analogs Fe^III,IV₂(μ-O)₂ and Fe^IV₂(μ-O)₂ supported by TPA and related ligands. These findings shed light on future development of more reactive approaches for C–H bond activation by bio-inspired dicobalt complexes.

Graphical Abstract

Introduction.

The activation of carbon-hydrogen (C–H) bonds is the first step of functionalizing inert hydrocarbons. This transformation is a key step in many biological and synthetic processes.^1–7 Nature offers highly efficient and selective solutions for carrying out aliphatic C–H bond functionalization using metalloenzymes that employ earth-abundant transition metals such as iron and copper in the active sites.⁸ One representative example is soluble methane monooxygenase (sMMO), a nonheme dinuclear iron-dependent enzyme that catalyzes the hydroxylation of the strong C–H bond of methane (bond dissociation energy BDE = 105 kcal/mol) using O₂ as the oxidant.^9–10 The catalytic cycle of sMMO has been extensively studied over decades, and features a high-valent bis-μ-oxo Fe^IV₂(μ-O)₂ “diamond core” intermediate called Q as the active oxidant for C–H bond activation. Spectroscopic characterization of sMMO-Q has revealed that the Fe^IV₂(μ-O)₂ diamond core has a short Fe•••Fe distance of 2.46 Å,¹¹ with both Fe centers in a high spin (S = 2) state.^12–13 Lipscomb and co-workers have further provided confirmative evidence using resonance Raman spectroscopy for the structural assignment of sMMO-Q.¹⁴ However, recent work by DeBeer et al. employing high-energy-resolution fluorescence detected X-ray absorption spectroscopy (HERFD XAS) argued that sMMO-Q possesses an open core structure with an Fe•••Fe distance of ~3.4 Å based on HERFD XAS data.^15–16 On the other hand, synthetic high-valent model complexes of sMMO-Q have been reported. A number of dinuclear μ-oxo and bis-μ-oxo manganese,^17–20 iron^9,21–22 and copper^23–24 complexes exhibited C–H bond cleavage and O–O bond formation activities. However, only one of them, a (μ-oxo)diiron(IV) complex supported by an anionic pentadentate N4O ligand, is capable of activating strong sp³ C–H bonds such as those in cyclohexane (BDE = 99.3 kcal/mol).²¹

In contrast, the high-valent diamond core chemistry of cobalt has been much less investigated, likely due to the difficulty of generating high-valent oxocobalt(IV) species for characterization.^25–26 There is evidence showing that high-valent oxocobalt(IV) species are involved in a number of water oxidation^27–28 and C–H bond cleavage reactions.²⁹ To date, six Co^III₂(μ-O)₂ complexes supported by tridentate or bidentate ligands have been structurally characterized.^30–35 However, no higher-valent derivative is available. In the present study, we report an unprecedented high-valent Co^III,IV₂(μ-O)₂ complex (3) supported by a tetradentate tris(2-pyridylmethyl)amine (TPA) ligand (Scheme 1). This species can be generated by one-electron oxidation of its Co^III₂(μ-O)₂ precursor (2). Characterization of 3 using combined spectroscopic and computational approaches confirmed that it has a diamond core structure with a short Co•••Co distance of 2.78 Å. More importantly, 3 is highly reactive and activates C–H bonds at a rate constant that is 3–5 orders of magnitude higher than its diiron and dimanganese analogs measured at higher temperatures.

Scheme 1.

Schematic illustration for the synthesis of 2 and 3.

Results and discussion.

Generation and characterization of 2.

Instead of first making the dihydroxo-bridged dicobalt(II) precursor Co^II₂(μ-OH)₂ described by Hikichi et al.,³⁶ we hypothesized that the direct reaction of a mononuclear cobalt(II)-Cl species [(TPA)Co^II-Cl]⁺ (1)³⁷ with H₂O₂ in the presence of base could afford the diamond core complex [(TPA)Co^III(μ-O) Co^III(TPA)]²⁺2 (2) according to Eq.1.

$$2{\left[\left(\text{TPA}\right){\text{Co}}^{\text{II}}-\text{Cl}\right]}^{+}+{\text{H}}_2{\text{O}}_2+2{\text{OH}}^{-}\to {\left[\left(\text{TPA}\right){\text{Co}}^{\text{III}}{\left(\mathrm{\mu }-\text{O}\right)}_2{\text{Co}}^{\text{III}}\left(\text{TPA}\right)\right]}^{2+}+2{\text{H}}_2\text{O}+2{\text{Cl}}^{-}$$ (1)

Indeed, the reaction of 1 directly with 0.5 eq. H₂O₂ at −40 °C in the presence of 1 eq. tetrabutylammonium hydroxide as the base produced a new brown species 2 (Figure 1 red). 2 exhibits an intense absorption at 460 nm (ε = 4700 M⁻¹ cm⁻¹) and a weak shoulder at ~600 nm. No reaction was observed between 1 and H₂O₂ alone. The stoichiometry between 1, H₂O₂, and OH⁻ was determined to be 2:1:2 by a titration experiment (Figure S1). The addition of excess H₂O₂ or OH⁻ did not cause 2 to be formed in a higher yield. The formation of 2 could be carried out in a variety of solvents including acetonitrile (MeCN), methanol (MeOH) and dichloromethane (DCM). 2 is relatively stable at −40 °C and below (t_1/2 = ~4 hours at −40 °C), and undergoes fast thermal decay at higher temperatures. Above −20 °C, 2 is formed at much reduced yields. 2 has a much lower thermal stability compared to the μ-hydroxo, μ −1,2-peroxo-bridged dicobalt(III) species of the same TPA ligand, whose single crystals can be obtained at room temperature.^38–39 The optical features of 2 are distinct from those reported for Co^III₂(μ-OH)₂, Co^III (μ-OH)(μ-OO)Co^III and mononuclear cobalt(III)-peroxo species Co^III(O₂).^38,40–45 In contrast, the intensity of the 460 nm absorption resembles those of the Co^III₂(μ-O)₂^30–31,35 and Ni^III₂(μ-O)₂^46–47 complexes reported previously, which was assigned to an oxo-to-metal charge transfer transition.

Figure 1.

Optical spectra of 1 (black), 2 (red) and 3 (blue) obtained in methanol at −60 °C. Inset: time trace of the absorption at 480 nm during the formation of 3.

The characterization of 2 using high-resolution electrospray ionization mass spectrometry (ESI-MS) shows a primary signal at m/z = 884.1755 (Figure S2). The mass and the isotope distribution pattern correspond to [Co^III₂(TPA)₂ (O)₂ (CF₃COO)(CH₃CN)]⁺. When the formation of 2 was carried out using ¹⁸O labeled H₂O₂, the ESI-MS signal was observed at m/z 888.2312. This upshift of four mass units clearly indicates that the oxygen atoms from H₂O₂ are incorporated into the molecular formula of 2. Complex 2 is EPR silent, consistent with the assignment of two d⁶ Co(III) centers (Figure S3). In contrast, the mononuclear precursor 1 exhibited an EPR signal typical of d⁷ S = 3/2 Co(II).^48–49 The lack of any residual Co(II) signal in the spectrum of 2 indicates that the conversion of 1 to 2 is complete. Furthermore, characterization of 2 using ¹H NMR spectroscopy at −55 °C in methanol confirmed that 2 is a diamagnetic species (Figure S4). Interestingly, we observed two sets of signals assignable to pyridine aromatic protons with the integration ratio of 2:1. This is clear evidence that the three pyridines of each TPA ligand are in two different environments---likely two are trans- to each other and the third one is trans to a bridging oxo ligand, a pattern that resembles that of the Co^III₂(μ-OH)₂ species supported by the same TPA ligand.⁴⁰ Quantification by integrating the aromatic proton signals using 4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS) as the internal standard showed that the formation of 2 is quantitative (> 90%).

Generation and characterization of 3.

To obtain a high-valent derivative of 2, we employed a one-electron chemical oxidant Ce^IV (in the form of (NH₄)₂Ce(NO₃)₆, CAN).⁵⁰ The introduction of 2-to 3-fold excess of CAN into the methanol solution of 2 at −60 °C and below afforded a new, transient species 3 with an intense absorption maximum at 480 nm (Figure 1 blue). 3 was formed from 2 within seconds (Figure 1, inset), much faster than the formation of 2 from 1 at −40 °C. 3 is less stable than 2 and decomposes even at −60 °C (t_1/2 = ~90 s at −60 °C), indicating that 3 is likely a higher-valent species. Furthermore, 3 is much less stable than the superoxo-bridged dicobalt(III) species of the TPA ligand, which can be obtained by one-electron oxidation of the peroxo-dicobalt(III) precursor and characterized at room temperature.^38–39 3 can be reduced back to 2 using ferrocene (Figure S5) and acetylferrocene (E = 0.27 V), but not diacetylferrocene (E = 0.49 V). The one-electron redox potential of 3 was thus estimated to fall in the range of 0.27–0.49 V vs. ferrocene.

3 is an EPR active species, showing a broad X-band (9.65 GHz) EPR signal at g = 2.13, 2.03 and 1.99 measured at 16 K, which is indicative of an S = 1/2 system (Figure 2 blue). The quantification of this S = 1/2 signal revealed that 3 was formed in a typical yield of ~70–80% from 2. Based on the EPR spin quantification, the molar extinction coefficient of the 480 nm band is estimated to be ~15,000 M⁻¹ cm⁻¹. We performed careful control experiments to exclude possible interference from CAN and its one-electron reduced species (both are EPR silent). Due to the broadness of this EPR signal, the hyperfine splitting originated from the coupling of the ⁵⁹Co nuclear spin (I = 7/2, natural abundance 100%) with the electronic spin cannot be established by the X-band data. We further measured S-band (3.5 GHz) spectra on the sample containing 3, which clearly demonstrated the existence of ⁵⁹Co hyperfine coupling (Figure 2 purple). The hyperfine splitting centered at g = 2.13 is consistent with a pattern involving two ⁵⁹Co nuclei with almost identical A values having a magnitude ~70 MHz (Figure S6). A spectral simulation including two identical ⁵⁹Co nuclei is the best solution to account for both X-band and S-band data (see Figure 3 for the simulation parameters, and Figures S6 and S7 for more discussions on the spectral simulations). The relatively small g anisotropy (g_max – g_min = 0.14) and the small A(⁵⁹Co) tensor suggest that the spin density may not be predominantly localized on a single Co center. Similar observations have been reported for a [Co(III)₃Co(IV)O₄] species with a complete spin delocalization on the cluster.⁵¹ In addition, several reported μ-superoxo-dicobalt(III) species have also exhibited an broad S = 1/2 EPR signal at cryogenic temperatures similar to the one observed for 3.^{38–39,45,52–54} Interestingly, these μ-superoxo-dicobalt(III) species have much higher thermal stability than 3 that allows them to be characterized even at room temperature, under which their EPR spectra exhibit much better resolved hyperfine structures.^39,55 Although EPR alone is unable to distinguish between bis-μ-oxo dicobalt(III,IV) and μ-superoxo-dicobalt(III) species, combined evidence obtained from X-ray absorption spectroscopy (XAS), Density Functional Theory (DFT) calculations, and oxygen exchange and C–H bond cleavage reactivities supports that 3 is a mixed-valent Co^III,IV₂(μ-O)₂ diamond core species having a spin-delocalized electronic ground state (vide infra).

Figure 2.

X-band (blue) and S-band (purple) EPR spectra of 3 recorded at 16 K. The spectra are aligned against g = 2 marker. The spectral simulations for the corresponding EPR signal are shown in black underneath the experimental spectra. The simulation parameters for 3 are g = [2.13, 2.02, 1.99], σ_g = [0.01, 0.008, 0.008], A(⁵⁹Co) = [75, 35, 18] MHz, with Euler angle [α, β, γ] = [0, 20°, 0], line width = 0.7 mT. The measurement conditions: microwave frequency, 9.64 GHz (X-band) and 3.5 GHz (S-band); microwave power, 20 μW (X-band) and 300 μW (S-band); modulation frequency, 100 kHz; modulation amplitude, 1 mT (X-band) and 0.5 mT (S-band).

Figure 3.

(A) Normalized XANES spectra of 1 (solid line), 2 (dotted line) and 3 (dashed line). The inset depicts an expansion of the pre-edge. (B-C) Representative best fits (bolded entries in Tables S1 and S2) to k³-weighted EXAFS data of 2 (panel B), and 3 (panel C). Experimental data is shown as a dotted line, while the best fit is shown as a solid line.

Characterization of 2 and 3 by XAS.

In light of the thermal instability of both 2 and 3 precluding the growth of suitable single crystals for X-ray crystallography, we turned to cobalt K-edge XAS on frozen solution samples of 2 and 3 for insights into their geometric and electronic structures. To provide further evidence for the progressive oxidation of 1 to 2 and 3, we first evaluated their cobalt K-edge X-ray absorption near edge spectroscopy (XANES) properties. The XANES spectrum of 2 shows a significant blue shift relative to that of 1, as evidenced by a +0.9 eV shift of pre-edge energy, and +2.8 eV shift for edge energy (Table 1, Figures 3A and S8), indicating an increase of the Co oxidation state in 2 relative to 1. Further, after correcting for the yield in the frozen solution samples of 3 (based on quantification of its EPR signal) by normalizing the spectrum after taking a scaled difference of 2 from 3, the edge energy of 3 is blue-shifted by ~+0.6 eV compared to that of 2 (7720.2 eV for 2 and 7720.8 eV for 3; see Table 1, Figures 3A and S9) while the pre-edge peak energy (7710.2 eV for both) is invariant (Figure S8). This is consistent with a partial increase of Z_eff from 2 to 3. We note that a small blue shift of the edge energy, without any change in pre-edge peak energy, was similarly observed for Co₄O₄ clusters.^56–57 The pre-edge area of the 7710.2 eV feature is 6.7(2) and 9.0(4) units for 2 and 3, respectively. There is an additional pre-edge feature at ~7713 eV that gains intensity upon oxidation of 2 (1.7 units) to 3 (5.9 units). The weak intensities of the 1s→3d pre-edge features of complexes 2 and 3 are consistent with O_h symmetry.^58–60 For octahedral low-spin Co(III) centers, the only allowed excited electron configuration is (t_2g)⁶(eg)¹, producing a E_g state.^57–58 This is responsible for the pre-edge feature at 7710 eV for complex 2. The pre-edge spectrum of mixed-valent complex 3 is expected to be an average of those of its spectroscopically resolved low-spin cobalt(III) and cobalt(IV) centers, as the XAS core-hole lifetime is orders of magnitude shorter than the intramolecular electron transfer rates for class III mixed valent complexes.^57–58,61 For octahedral low-spin Co(IV) centers, the pre-edge peak originates from ¹T₁, ³T₁, and ¹T₂ excited states of the (t_2g)⁵ (e_g)¹ electron configuration;^57–58 a transition to the ¹A₁ excited state from the (t_2g)⁶(e_g)⁰ electron configuration is expected to produce a lower-energy shoulder, which is not well resolved in the XANES spectrum of complex 3 with mixed-valent Co(III)(μ-O)₂Co(IV) core. There is a shoulder on the rising edge at ~7713 eV that gains intensity upon oxidation of 2 (1.7 units) to 3 (5.9 units). Similar features at a relative position of 3~5 eV above the pre-edge peak were observed in K₃[Fe(III)(CN)₆] (octahedral low-spin d⁵ system),⁶¹ K₃[Co(III)(CN)₆] (octahedral low-spin d⁶ system),⁶⁰ several Co(III)-peroxo complexes supported by macrocyclic N₄ ligands (distorted octahedral low-spin d⁶ system),⁶² and cobalt(III) intermediates of B₁₂ enzymes and model compounds supported by dimethylglyoxime ligands with various axial donors (distorted octahedral low-spin d⁶ system with or without Co–C bonds);⁵⁹ these features were attributed to charge transfer transitions into predominantly ligand-based orbitals and/or 1s-to-4p transitions with shakedown contributions. Such features were also reported for MCoO₂ (M = Li, Ag, Eu, and La)⁶³ and Co₄O₄ cubanes,⁵⁷ and were assigned to transitions to orbitals with significant M/M coupling interactions mediated by the oxo bridges.⁶⁴ In addition, the ³T₂ transition of the excited (t_2g)⁵(e_g)¹ electron configuration for the Co(IV) center of complex 3 could also contribute to the broad 7713 eV feature.^57–58

Table 1.

Summary of spectroscopic parameters for cobalt species 1, 2 and 3.

Complex	λmax, nm (ε, M−1 cm−1)	Spin State	g1, g2, g3	Pre-edge peak		Rising-edge shoulder		Edge energy (eV)	Co•••Co (Å)
				E (eV)	Area	E (eV)	Area
1	490 (250) 615 (150)	3/2	7.13, 3.03, 2.06a	7709.3(1)	10.7(4)	7713.9(2)	10.1(15)	7717.4	NA
2	460 (4,700)	0	NA	7710.2(1)	6.7(2)	7713.1(1)	1.7(5)	7720.2	2.78d 2.784e
3	480 (15,000)	1/2	2.14, 2.02, 1.98	7710.2(1)	8.4(3)b 9.0(4)c	7713.4(1)	4.9(16)b 5.9(19)c	7720.5b 7720.8c	2.78d 2.769e

^aeffect g values from m_s = +/− 1/2 levels of an S = 3/2 spin state.

^bfrom fittings of the raw data. Pre-edge areas are multiplied by100 for convenience.

^cfrom fittings after correction for the formation yield determined by EPR of the corresponding species.

^dfrom EXAFS analysis.

^efrom geometry optimized calculation.

Extended X-ray absorption fine structure (EXAFS) characterization of 2 and 3 provided further structural insights. The k³χ(k) EXAFS of 2 and 3 have highly analogous oscillation patterns, which is mirrored by the very similar appearance of the EXAFS Fourier transforms for these two species (Figures 3B, C). Both observations suggest that the coordination environment around the cobalt centers does not alter appreciably upon the oxidation of 2 to 3. Indeed, the first coordination shell of 2 and 3 are both best fit by 5 N/O scatterers surrounding the cobalt center at an average distance of 1.91–1.92 Å, accounting for nitrogen donors from the TPA ligand as well as oxo donors. Attempts to split this shell into two subshells of N_TPA and O_oxo donors yielded poorer quality fits, as evidenced by higher reduced χ² and R-factor values. As shown in the DFT geometry-optimized structures (vide infra), the N_TPA and O_oxo subshells are separated by less than the EXAFS resolution of 0.13 Å for our k range of 2.0–14.0 Å⁻¹. Therefore, the shorter O_oxo sub-shell is not resolved from the N_TPA scatterers.

The second coordination sphere for both complexes is best modeled with a combination of long range carbon scatterers and a Co•••Co vector, with the best structural solution for both complexes (i.e. Fit 17 for 2 in Table S1 and Fit 18 for 3 in Table S2) including a cobalt scatterer at a short distance of 2.78 Å away from the absorbing cobalt. Fits excluding this cobalt scatterer and comprising only first-shell N/O scatterers and one or two shells of second sphere C atoms yielded sizably worsened fit quality. The improvement of fit quality by introducing the cobalt scatterer is also observed in the Fourier transform spectrum (Figures S10 and S11). Such a short Co•••Co distance has previously been reported for the crystal structures of Co^III₂(μ-O)₂ complexes (2.67–2.74 Å)^30–35 and other high-valent M₂(μ-O)₂ complexes with a diamond-core structure (Table S3), and is also confirmed in our DFT geometry optimization (vide infra). Notably for both complexes, we did not observe significant features at path-lengths longer than 3 Å in the Fourier transform characteristic of peroxo-bridged core structures such as Co^III(μ-OH)(μ-OO)Co^III and Cu^II(μ-η²:η²(O₂))Cu^II where the M•••M separation is significantly longer (Table S3). Taken together, the characterizations of 2 and 3 by EXAFS support their structural assignment as the Co₂(μ-O)₂ diamond core complexes.

DFT calculations of 2 and 3.

We then turned to DFT optimization of the structures of 2 and 3 (Figures S12 and 4, respectively; also see Table S4) to substantiate our spectroscopic observations. Geometry optimization at the BP86/6–31G(D) level provided low-spin ground states for 2 and 3, i.e. singlet and doublet, respectively. For each species calculated, an optimized structure with a symmetric Co₂(μ-O)₂ diamond core and coordinatively saturated cobalt centers was successfully obtained. Two structural isomers differing in the relative position of two tertiary amine nitrogen atoms (i.e. cis- vs. trans-, see Figures 4, S12 and Scheme S1) were optimized. For both 2 and 3, the trans model was found to be 31.1 kJ/mol and 27.4 kJ/mol lower in energy than the cis-model, respectively. In addition to symmetric Co₂(μ-O)₂ structures, we also optimized open-core and asymmetric Co₂(μ-O)₂ structures (Scheme S1). However, all of these alternative structures of 2 and 3 are energetically less stable than the symmetric trans-Co₂(μ-O)₂ counterparts. Notably, spin density on the two Co atoms was distributed equally in complex 3 (Figures 5 and S13, Table S5), which implies that 3 is a valence-delocalized system in agreement with the simulation of the EPR data. The Co•••Co distances in the DFT-optimized trans structure of 2 and 3 are 2.78 Å and 2.77 Å, respectively, consistent with those determined by EXAFS analysis (Table 1). Furthermore, the average distances of Co–N_TPA/O_oxo bonds for the trans structures of complexes 2 and 3 are 1.92 Ǻ and 1.91 Ǻ (Table S6), respectively, in excellent agreement with EXAFS analysis for the first coordination shell. The average distances of Co–O_oxo bonds of DFT geometry optimized trans structures 2 and 3 are 1.86 Ǻ and 1.83 Ǻ, respectively. These Co–O_oxo distances are consistent with those of several Co₄O₄ cubane complexes with six-coordinated cobalt sites (values are from X-ray crystallography and DFT calculations),^57,65 and are slightly longer than those of diamond core Co₂O₂ crystal structures with cobalt centers of lower coordination numbers (1.78–1.83 Ǻ).^30–35 Furthermore, these geometry-optimized structures clearly showed that the six pyridines are in two distinct local environments with a ratio of 2:1, which is experimentally observed in our ¹H NMR results obtained for 2 (vide supra, Figure S4). Similar results were obtained using other commonly used DFT functionals M06-L, B3LYP* (15% HF), and B3LYP (Table S7).

Figure 4.

BP86/6–31G(D) optimized symmetric diamond core structures for 3 in the trans- (panel A) and cis- (panel B) configuration. Cobalt: pink, oxygen: red, nitrogen: blue, carbon: gray, hydrogen atoms are omitted for clarity.

Figure 5.

Spin density plot of 3 [BP86/6–31G(D)+PCM(CH3OH)].

C–H bond cleavage reactivity of 3.

With the spectroscopic characterization of the mixed-valent Co^III,IV₂(μ-O)₂ diamond core 3 in hand, we sought to investigate its ability to cleave aliphatic C–H bonds. Similar reactivity has yet to be described for superoxo-bridged dicobalt(III) species. We first selected 9,10-dihydroanthracene (DHA, BDE = 78 kcal/mol) as a diagnostic substrate. As shown in Figure S14, the introduction of 8 mM DHA into a freshly prepared solution of 3 at −60 °C in 2:1 methanol/DCM mixed-solvent caused the complete disappearance of its absorption maximum at 480 nm within less than 2 minutes. The kinetic trace could be well fitted using a first-order model to obtain the pseudo-first-order rate constant (k_obs) of 0.019 s⁻¹. Furthermore, the second-order rate constant k₂ of 1.5(1) M⁻¹ s⁻¹ for DHA oxidation could be extracted from the slope of a linear correlation between k_obs and the substrate concentration. When deuterated DHA-d₄ was used as the substrate, a slower reaction with k₂ = 0.25(1) M⁻¹ s⁻¹ was observed, yielding a H/D kinetic isotope effect (KIE) of 6.0 (Figure S15). The large KIE value indicates that the cleavage of an aliphatic C–H bond of DHA by 3 through hydrogen atom transfer (HAT) is the rate-determining step. We further measured the DHA oxidation rate constants in a temperature range of −55 °C to −70 °C, and obtained the activation parameters (ΔH^≠ = 9.0(1) kcal/mol and ΔS^≠ = −14.8(2) cal/(mol ● K)) by Eyring analysis (Figure S16). The small activation energy is consistent with the fast rate of DHA oxidation even at cryogenic temperatures.

Quantification of the oxidation product(s) of DHA by GC-MS indicated the formation of primarily anthraquinone (19% yield) and trace amount of anthrone and anthracene, accounting for ~80% of the oxidizing equivalents used (Table S8). We carried out careful control experiments to exclude possible interference from the excess amount of CAN present in the reaction solution. This observation shows that 3 is a two-electron oxidant. Furthermore, the yield of anthraquinone is independent of the amount of O₂ present in the reaction solution, suggesting that anthraquinone is not the product of a radical auto oxidation by O₂. No ligand oxidation product was found, presumably due to geometric constraints of the ligand backbone and the unlikehood of the intermolecular process between two oxidant molecules. After the reaction, a new cobalt species that is distinct from 2 was formed (Figure S12). Given that 3 is a two-electron oxidant, we hypothesized that this new species is a dicobalt(II,III) complex. The characterization of this species will be described elsewhere.

We investigated the reaction of 3 with other hydrocarbons whose C–H bond strength varies in a range of 78–87 kcal/mol. As anticipated, the k₂ value decreases as the C–H bond being cleaved becomes stronger (Figure S17). The benzylic C–H bond of ethylbenzene (BDE = 87 kcal/mol) is the strongest one that 3 can cleave, with a rate of 0.025(2) M⁻¹ s⁻¹ at −60 °C. Product analysis showed the formation of acetophenone as the only product at the yield of 15%, accounting for ~43% (after correcting for the formation yield of 3) of the oxidizing equivalents available for 3. The oxidizing equivalents unaccounted for may be due to the self-decay of 3, which has first-order rate constant of ~0.008 s⁻¹. Furthermore, a H/D KIE of 8.8 and 7.8 was determined for the oxidation of fluorene and ethylbenzene (Figures S18 and S19), respectively, clearly indicating that 3 cleaves the C–H bond of these substrates via a HAT process. For this group of substrates studied, the logarithm of k₂’ (k₂’ = k₂/number of equivalent hydrogen), when normalized on a per hydrogen basis, correlated linearly with the strength of the C–H bond being cleaved with a slope of −0.17 (Figure 6). The analysis of product formation for these substrates showed that ketone was the primary product (Table S7). In some cases alcohol and desaturated products were also formed but in much smaller yields.

Figure 6.

Plot of log k₂’ as a function of the C–H bond strength for substrate oxidation by 3 in methanol at −60 °C (red), Fe^III,IV₂(μ-O)₂ with 1 M H₂O at −30 °C (black) and Cu^III₂(μ-O)₂ at −70 °C (blue). The lines represent the best linear fittings.

3 is thus an oxygenation reagent for C–H bonds. The rate-determining HAT process from the substrate likely generates a Co^III₂(μ-O)(μ-OH) species and a carbon centered radical. The rebound of the hydroxyl group back to the carbon radical then yields the hydroxylated C–OH product (Scheme 2). The over-oxidation of the alcohol product must occur by another molecule of 3 to generate the ketone product observed in product analysis. Our observation that the product distribution and yield are independent of the presence of O₂ in the reaction media indicates that the rebound of the hydroxyl group to the carbon radical must be fast so that the carbon radical does not escape from the radical cage to be trapped by O₂.

Scheme 2.

Proposed mechanism for C–H bond oxygenation by complex 3.

We further studied the exchange kinetics of the oxo ligands in complexes 2 and 3 with added H₂¹⁸O in methanol at −60 °C in order to understand this fundamental reactivity of bis-μ-oxo complexes.^66–67 In these experiments, H₂¹⁸O was either added to the solution of 2, allowing oxygen exchange to occur for a period of incubation time followed by rapid conversion of 2 to 3, or introduced directly to the solution of 3 for oxygen exchange. The addition of water does not affect the formation yield of 3. We then employed ethylbenzene as a substrate probe to react with 3, and monitored the incorporation of ¹⁸O isotope in the oxidation product acetophenone by GC-MS. Complex 2 has sufficient lifetime at −60 °C that allows us to vary the incubation time up to 4 hours after the addition of H₂¹⁸O. In contrast, complex 3 is too short-lived for a similar set of incubation time dependence studies.

As shown in Figure 7A, the incorporation of ¹⁸O isotope in acetophenone was found to be 17% when 0.3 M H₂¹⁸O was added to the solution of 2 without any incubation time. The fraction of ¹⁸O acetophenone increases when increasing the incubation time of the solution of 2 after the same amount of H₂¹⁸O was added (Figure 7A), up to 35% at the maximum tested incubation time of 4 hours. This slight dependence of the ¹⁸O incorporation as a function of the incubation time indicates that oxygen exchange on complex 2 is in fact slow (within hours), consistent with the kinetic inertness of low-spin Co(III) centers.⁶⁸ Furthermore, the fraction of ¹⁸O acetophenone increases as a function of the concentration of H₂¹⁸O added to the solution of 2 with a fixed incubation time (2 hours, Figure 7B). At the maximum H₂¹⁸O concentration tested (3 M, almost the saturation concentration in methanol at −60 °C), we observed ~50% ¹⁸O incorporation in acetophenone. In addition, we performed a set of control experiments by adding H₂¹⁸O directly to the solution of 3 followed by immediate ethylbenzene oxidation (without any incubation time). For the two H₂¹⁸O concentrations tested (red data points in Figure 7B), comparable ¹⁸O acetophenone contents were obtained compared to the set of studies in which H₂¹⁸O was added to complex 2 with a 2-hour incubation time (black data points in Figure 7B). This observation suggests that oxygen exchange proceeds much faster on 3 than on 2.

Figure 7.

Plots of the ¹⁸O incorporation percentage in acetophenone as a function of (A) incubation time after 0.3 M H ₂¹⁸O was added to 3 mM 2, and (B) the concentration of H ₂¹⁸O introduced to either 3 mM 2 with a 2-hour incubation time (black squares) or 3 mM 3 with no incubation time (red squares). Other conditions: in methanol at −60 °C.

These data thus provide strong support for the structural assignment of 2 and 3 as having the bis-μ-oxo diamond cores, since the oxygen exchange of H₂¹⁸O with peroxo- and superoxo-bridged dinuclear species is highly unlikely. Furthermore, the incorporation of ¹⁸O isotope into the reaction product of ethylbenzene is a clear demonstration that the oxygen in acetophenone is from the diamond core complex 3. This observation is consistent with the rebound mechanism (Scheme 2) proposed for C–H bond oxygenation reactions.

Our results clearly revealed that the Co^III,IV₂(μ-O)₂ diamond core 3 is a highly reactive oxidant towards C–H bonds. The BDE plots shown in Figure 6 comparing C–H bond cleavage reactivities of 3 with Fe^III,IV₂(μ-O)₂ (in the presence of 1 M H₂O)²² and Cu^III₂(μ-O)₂ (in an equilibrium with a Cu^II₂(μ-η²:η²-O₂) species)²³ showed that 3 is up to two orders of magnitude more reactive than the other two diamond core species. The slope of the linear fitting for 3 (−0.17) is less steep than those of Fe^III,IV₂(μ-O)₂ and Cu^III₂(μ-O)₂ species (~−0.3)_, indicating that the rate constants of 3 are less sensitive to the C–H bond strength of the substrates than the other two species.

Specifically, 3 has the highest reaction rate for DHA oxidation (1.5 M⁻¹ s⁻¹ at −60 °C) compared to diamond core complexes of other first-row transition metals including Mn,¹⁷ Fe⁶⁹ and Cu²³ reported previously at a variety of temperatures (see also Table S9 and references therein). 3 is about 3 orders of magnitude more reactive than the di-manganese species Mn^III,IV₂(μ-O)₂ (1.2 × 10⁻³ M⁻¹ s⁻¹ at 32 °C),¹⁷ and is about 4–5 orders of magnitude more reactive than the diiron analogs Fe^III,IV₂(μ-O)₂ and Fe^IV₂ (μ-O)₂ supported by TPA and related ligands (10⁻⁴–10⁻⁵ M⁻¹ s⁻¹ at −30 °C).⁶⁹ Furthermore, 3 oxidizes ethylbenzene at a rate of 0.025(2) M⁻¹ s⁻¹ at −60 °C, which is about two orders of magnitude faster than that of Fe^III,IV₂(μ-O)₂ at −30 °C in the presence of 1 M H₂O (2 × 10⁻⁴ M⁻¹ s⁻¹).²² Surprisingly, the redox potential of 3 (0.27–0.49 V) is comparable or even lower than those of these two high-valent diiron counterparts.⁶⁹ Therefore, other factors such as the pK_a of the bridging oxo ligands and the activation energy of cleaving a C–H bond likely contribute to the high reactivity of 3. On the other hand, 3 is unable to cleave C–H bonds stronger than 87 kcal/mol. This thermodynamic limit likely reflects the strength of the Co^IIIO–H bond formed after the HAT step. Our ongoing investigations aim to measure these thermodynamic values in order to better understand the reactivity of 3.

In addition, the DHA oxidation rate of 3 is at least 7-fold higher than those measured at higher temperatures of two mononuclear oxocobalt(IV) species recently described.^25–26 In fact, the DHA oxidation rate is only about one order of magnitude slower than the one measured for the open core species HO-Fe^III-O-Fe^IV=O (28 M⁻¹ s⁻¹).⁷⁰ It is thus conceivable that further development of this high-valent dicobalt-oxo system could lead to the generation of even more reactive species for sp³ C–H bond activation.

Conclusion.

In conclusion, we have generated an unprecedented high-valent dicobalt(III,IV) complex Co^III,IV₂(μ-O)₂ 3 by one-electron oxidation of its Co^III₂(μ-O)₂ precursor 2. Characterization of 3 by spectroscopic and computational methods including EPR, XAS and DFT revealed that 3 has an S =1/2 ground state and a bis-μ-oxo diamond core structure (preferably in a trans configuration) with a short Co•••Co distance. The bridging oxo ligands of 3 are exchangeable with H₂¹⁸O. 3 is able to cleave C–H bonds having bond strength up to 87 kcal/mol, affording fast kinetic rate constants and the oxygenated products. These features (spectroscopy, stability and reactivity) are distinct from those of the peroxo- and superoxo-bridged dicobalt(III) complexes supported by the same TPA ligand recently reported by other groups.^38–39 Interestingly, the rate constant for DHA oxidation by 3 is 4–5 orders of magnitude higher than those determined for the diiron analogs, despite the fact that 3 has a moderate redox potential that is comparable to those of the diiron species. These interesting findings shed light on better understanding of the high-valent diamond core chemistry for cobalt and inspire future development of more effective approaches for C–H bond activation by bio-inspired dicobalt complexes.