Hydrogen Atom Transfer Thermodynamics of Homologous Co(III)- and Mn(III)-Superoxo Complexes: The Effect of the Metal Spin State.

Systematic investigations on H atom transfer (HAT) thermodynamics of metal O2 adducts is of fundamental importance for the design of transition metal catalysts for substrate oxidation and/or oxygenation directly using O2. Such work should help elucidate underlying electronic-structure features that govern the OO-H bond dissociation free energies (BDFEs) of metal-hydroperoxo species, which can be used to quantitatively appraise the HAT activity of the corresponding metal-superoxo complexes. Herein, the BDFEs of two homologous CoIII- and MnIII-hydroperoxo complexes, 3-Co and 3-Mn, were calculated to be 79.3 and 81.5 kcal/mol, respectively, employing the Bordwell relationship based on experimentally determined pK a values and redox potentials of the one-electron-oxidized forms, 4-Co and 4-Mn. To further verify these values, we tested the HAT capability of their superoxo congeners, 2-Co and 2-Mn, toward three different substrates possessing varying O-H BDFEs. Specifically, both metal-superoxo species are capable of activating the O-H bond of 4-oxo-TEMPOH with an O-H BDFE of 68.9 kcal/mol, only 2-Mn is able to abstract a H atom from 2,4-di-tert-butylphenol with an O-H BDFE of 80.9 kcal/mol, and neither of them can react with 3,5-dimethylphenol with an O-H BDFE of 85.6 kcal/mol. Further computational investigations suggested that it is the high spin state of the MnIII center in 3-Mn that renders its OO-H BDFE higher than that of 3-Co, which features a low-spin CoIII center. The present work underscores the role of the metal spin state being as crucial as the oxidation state in modulating BDFEs.

Introduction

Efficient and selective functionalization of substrates directly utilizing dioxygen, O2, represents a desired goal of synthetic chemistry.[1] To the best of our knowledge, only a few large-scale industrial processes can realize this type of reactions, whereas such transformations have been frequently identified in the catalytic cycle of a diverse array of metalloenzymes.[2−4] In enzymatic processes, in situ generated metal-superoxo, -peroxo, and -oxo intermediates derived from O2 activation by low-valent metal cofactors often employ hydrogen atom transfer (HAT) to oxidize their substrates. For instance, treatment of isopenicillin N synthase (IPNS) and myo-inositol oxygenase (MIOX) with O2 was found to initially afford an FeIII-superoxo intermediate that is capable of performing HAT reactions to produce an FeIII-hydroperoxo species.[5,6] A range of metal-superoxo model compounds, such as LCu(O2•) (L, a bis(arylcarboxamido)pyridine ligand),[7] (PIm)Fe(O2•) (PIm, a porphyrinate ligand with an appended axial imidazolyl group),[8] (PAr)Fe(O2•) (PAr, a porphyrinate ligand in four varied derivatives),[9,10] [Cu2(XYLO)(O2•)]2+ (XYLO, a bis(2-{2-pyridyl}ethyl)amine chelating ligand with a bridging phenolate moiety),[11] L′Cu2(μ-O2•) (L′, a tacn/pyrazolate hybrid ligand),[12] and Co(O2•)(Me3TACN)(S2SiMe2),[13] have been demonstrated to carry out HAT reactions and furnish metal-hydroperoxo complexes. Because the Gibbs free energy change of a HAT reaction can be estimated to be the difference of the X–H (X = C, N, O) bond dissociation free energy (BDFE) of the substrate relative to the OO–H BDFE of the hydroperoxo product, thermodynamically the occurrence of such a transformation requires that the latter value, as observed for all systems shown in Scheme , be greater than the former. Therefore, the HAT capability of a given metal-superoxo intermediate to a large extent, if not completely, depends on the OO–H BDFE of its hydroperoxo product. Consequently, it is of fundamental significance to elucidate the underlying electronic-structure features that govern the OO–H BDFE, because the insights thus obtained can guide the design of transition metal catalysts that directly utilize O2 for functionalization of substrates. Though the OO–H BDFEs of a plethora of metal-hydroperoxo species have been determined thus far (Scheme ), systematic investigations aimed at rationalizing their varying values remain lacking.

Scheme 1

OO–H BDFEs of Biomimetic Metal-Hydroperoxo Complexes: (A) LCu(OOH),[7] (B) (PIm)FeIII(OOH),[8] (C) (PAr)FeIII(OOH),[9,10] (D) [Cu2(XYLO)(OOH)]2+,[11] (E) L′Cu2(μ-OOH),[12] (F) Co(OOH)(Me3TACN)(S2SiMe2),[13] (G) 3-Co, and (H) 3-Mn

Converted from BDE (83–87 kcal/mol, CG,sol = 66.0 kcal/mol for BDE and 60.4 kcal/mol for BDFE in THF).

Ar = 2,6-difluorophenyl.

Ar = pentafluorophenyl.

Ar = phenyl.

Ar = 2,4,6-trimethylphenyl.

In our long-term research work on O2 activation mediated by the first-row transition metals, we have successfully prepared a series of FeIII-, CoIII-, and MnIII-superoxo complexes through adding O2 to their divalent precursors at low temperatures[14−16] and explored their reactivities toward various substrates, which, in particular, demonstrated ambiphilicity of metal-superoxo species.[17,18] Specifically, CoIII- and MnIII-superoxo complexes, Co(BDPP)(O2•) (2′-Co, H2BDPP = 2,6-bis((2-(S)-diphenylhydroxylmethyl-1-pyrrolidinyl)methyl)pyridine), Mn(BDPP)(O2•) (2′-Mn), and Mn(BDPBrP)(O2•) (2-Mn, H2BDPBrP = 2,6-bis((2-(S)-di(4-bromo)phenylhydroxylmethyl-1-pyrrolidinyl)methyl)pyridine), were found to react with TEMPOH to furnish the corresponding hydroperoxo complexes, Co(BDPP)(OOH) (3′-Co),[15] MnIII(BDPP)(OOH) (3′-Mn),[16] and MnIII(BDPBrP)(OOH) (3-Mn).[16] Furthermore, treatment of 2-Mn with trifluoroacetic acid (TFA) led to formation of a MnIV-hydroperoxo complex, [Mn(BDPBrP)(OOH)]+ (4-Mn), which could be reversibly reduced to 3-Mn.[17] In contrast, the reaction of TFA with 2′-Co does not result in one-electron oxidation of the metal center to yield a CoIV-hydroperoxo complex, but induces protonation of the BDPP2– ligand and generation of [CoIII(HBDPP)(O2•)]+ (4′-Co). In analogy to a HAT process, only in the presence of both electron and proton donors does transformation of 2′-Co to 3′-Co occur.[18]

The purpose of the present work is to investigate the HAT thermodynamics of the aforementioned CoIII- and MnIII-superoxo complexes and clarify the electronic-structure origin for their difference. To circumvent the complexity arising from the slightly different ligand environment, we first synthesized Co(BDPBrP)(O2•) (2-Co) with the same ligand as 2-Mn and then determined the OO–H BDFEs of Co(BDPBrP)(OOH) (3-Co) and 3-Mn. The differential HAT capability of 2-Co and 2-Mn was ascertained by examining their reactions with assorted substrates having varied O–H BDFEs. Finally, density functional theory (DFT) calculations were employed to pinpoint the electronic-structure characteristics that dictate the OO–H BDFEs of 3-Co and 3-Mn. Our combined experimental and theoretical studies enable us, for the first time, to demonstrate that the distinct metal spin state is responsible for the varied HAT capability of the homologous CoIII- and MnIII-superoxo complexes. Of note, the effects of the metal spin state on the HAT reaction rates of FeIV=O species have been well documented in the literature.[19−22] Furthermore, it has been reported that the different metal spin state also affects the strength of metal–NO interactions.[23] Equally important is that the metal oxidation state has been shown to play a significant role in controlling O–H BDFEs of a range of Fe-, Mn-,[24] and Cu-hydroxo[25] complexes.

Results and Discussion

Synthesis of Co(BDPBrP)(O2•) (2-Co) and Determination of the OO–H BDFEs of Co(BDPBrP)(OOH) (3-Co)

To prepare the homologous CoIII-superoxo complex Co(BDPBrP)(O2•) (2-Co), a green CoII starting material, Co(BDPBrP) (1-Co), was synthesized by reacting H2BDPBrP with 1 equiv of CoCl2 in the presence of NaH (2.5 equiv) in 1:1 tetrahydrofuran (THF)/CH3CN mixed solvent (Figure S1). The molecular structure of 1-Co determined by crystallographic analysis revealed that the CoII center possesses a distorted square-pyramidal geometry with τ = 0.47 (Figure ). The crystallographic data and the selected bond lengths and angles are listed in Tables S1 and S2. The effective magnetic moment of 1-Co measured by the Evans method is 4.4 μB (Figure S2), indicative of an ST = 3/2 ground state, similar to that found for 1′-Co.

Figure 1

ORTEP of 1-Co with ellipsoids set at 50% probability. Hydrogen atoms are omitted for clarity.

Upon bubbling O2 into a THF solution of 1-Co at −90 °C, the appearance of two characteristic absorption bands at 485 and 570 nm in the UV–vis spectrum signaled the generation of 2-Co, as the same features were found for the formation of 2′-Co (Figure S1).[15] Furthermore, 2-Co registers an almost identical EPR spectrum (Figure a) to 2′-Co, reflecting that 2-Co is composed of a low-spin (SCo = 0) CoIII center interacting with a superoxo ligand as unequivocally identified for 2′-Co.[15]2-Co also performs a HAT reaction toward TEMPOH in THF at −90 °C to form a low-spin (SCo = 0) CoIII-hydroperoxo complex, Co(BDPBrP)(OOH) (3-Co), and a TEMPO radical in 93% yield (Figures S4 and S5).

Figure 2

X-band EPR spectra of (a) 2-Co and (b) 4-Co. Measurement conditions: T = 100 K, fmw = 9.393 and 9.410 GHz, respectively. The simulations were obtained by using the following parameters: g = 2.011, 2.097, 1.994, ACo = 24, 57, 29 MHz, a = 27° for 2-Co, and g = 2.012, 2.062, 2.003, ACo = 17, 75, 20 MHz, a = 30° for 4-Co. The definition of a is described in the Supporting Information.[26]

Treating 2-Co with 1 equiv of HOTf led to protonation of the alkoxide group of the BDPBrP2– ligand and generated [CoIII(HBDPBrP)(O2•)]+ (4-Co). In analogy to the situation found for 2′-Co and 4′-Co, 2-Co and 4-Co exhibit similar EPR spectra (Figure b), and two characteristic features at 480 and 640 nm of 4-Co emerge as those found for 4′-Co (Figure a). As seen in Figure a, 2-Co can be retrieved by adding 1 equiv of 1,8-bis(dimethylamino)naphthalene (proton sponge) to 4-Co as indicated by the disappearance of the signature absorption band of 4-Co at 640 nm and the reemergence of the absorption band of 2-Co at 485 nm. A pKa value of 10.3 in MeTHF at −120 °C was obtained by proton sponge titration. Actually, only around 70% of 2-Co was regenerated, as estimated by the UV–vis spectral changes of 4-Co to 2-Co (Figure S6). The electrochemical property of 3-Co was also examined by performing cyclic voltammetry (CV) measurements in a n-PrCN solution vs an Ag wire reference electrode with 0.1 M NBu4BF4 as electrolyte. Owing to the limited solubility of the electrolyte, CV measurements of 3-Co had to be carried out at −80 °C instead of −120 °C. A quasi-reversible 1 e– redox event at 0.210 V (E1/2 vs Fc/Fc+) was found, attributed to the 3-Co/4-Co couple as displayed in Figure a, primarily due to the partial decay of 4-Co. With the obtained redox potential E1/2 and pKa, we were able to establish a thermodynamic scheme, shown in Scheme a, and determine the OO–H BDFE of 3-Co to be 79.3 kcal/mol from the Bordwell relationship (eq ,[27]CG,sol = 60.4 kcal/mol for BDFE in THF[28]).

Figure 3

UV–vis spectral changes of the reactions of (a) 4-Co and (b) 4-Mn (1.0 mM, blue trace) with 1 equiv of base (proton sponge for 4-Co and NEt3 for 4-Mn) in MeTHF at −120 °C. Inset: (a) Titration of 4-Co provides a slope offering a Keq of 5.81, which gives a pKa value of 10.3 for 4-Co, and (b) titration of 4-Mn provides a slope offering a Keq of 2.23, which gives a pKa value of 12.1 for 4-Mn.

Figure 4

CV of (a) 3-Co (1.0 mM) in n-PrCN with ipc/ipa = 0.63, ΔE = 215 mV (scan rate = 150 mV/s), and (b) 3-Mn (1.0 mM) in DCM with ipc/ipa = 0.90, ΔE = 140 mV (scan rate = 75 mV/s) at −80 °C.

Scheme 2

Thermodynamic Scheme of the Discussed Co/O2 and Mn/O2 Species (in THF/MeTHF)

Determination of the OO–H BDFE of 3-Mn

Our earlier combined spectroscopic and computational studies unequivocally established that 2-Mn possesses a high-spin (SMn = 2) MnIII center antiferromagnetically coupled with the superoxo ligand to give an overall ST = 3/2 spin ground state.[16] Reactions of 2-Mn with TEMPOH and trifluoroacetic acid gave hydroperoxo complexes 3-Mn and 4-Mn composed of a high-spin (HS, SMn = 2) MnIII and a HS (SMn = 3/2) MnIV center,[17] respectively. Protonation of 2-Mn to yield 4-Mn was found to be reversible,[17] and the pKa value of 4-Mn in MeTHF at −120 °C was determined to be 12.1 by NEt3 titration, as seen in Figure b. Different from the Co system, regeneration of 2-Mn reached a yield of 91%,[17] indicating that 4-Mn is more stable compared to 4-Co; therefore, CV measurements of 3-Mn in a CH2Cl2 solution at −80 °C (Figure b) revealed a nearly reversible 1e– redox wave at 0.196 V (E1/2 vs Fc+/Fc for 3-Mn/4-Mn). On the basis of the measured E1/2 and pKa, the OO–H BDFE of 3-Mn was estimated to be 81.5 kcal/mol (Scheme b).

Reactions of 2-Co and 2-Mn toward Substrates with Different O–H BDFEs

Owing to the thermal instability of 4-Co and 4-Mn, MeTHF was employed as the reaction solvent to obtain the pKa values of 4-Co and 4-Mn at −120 °C; however, different solvents (n-PrCN and dichloromethane (DCM)) have to be applied for their CV measurements. In addition to these experimental limitations, CG,sol in THF was used to compute BDFEs; thus, we surmised that all of these factors may cause some errors in the OO–H BDFEs of 3-Co and 3-Mn. Therefore, three substrates with varying O–H BDFEs were employed to validate the determined OO–H BDFEs of 3-Co and 3-Mn. Specifically, 1-hydroxy-2,2,6,6-tetramethyl-4-oxo-piperidine (4-oxo-TEMPOH), which has an O–H BDFE (68.9 kcal/mol in THF)[28] lower than the OO–H BDFEs of 3-Co and 3-Mn, was selected as the first substrate for the comparison. As expected, both 2-Co and 2-Mn reacted with 4-oxo-TEMPOH (5 equiv) in a THF solution at −80 °C to cleanly form 3-Co and 3-Mn (Figure a and b). The organic product of 4-oxo-2,2,6,6-tetramethylpiperidine-1-oxyl radical (4-oxo-TEMPO) was detected by EPR spectroscopy in a yield of 99% for 2-Co and 90% for 2-Mn (Figure S9a and b). When 3,5-dimethylphenol (3,5-DMP, 10 equiv) was employed, not surprisingly, no reaction with 2-Co or 2-Mn occurred, because the O–H BDFE (85.6 kcal/mol in THF)[29−32] of 3,5-DMP exceeds the OO–H BDFE of 3-Co and 3-Mn (Figure S10a and b). When THF solutions of 2-Co and 2-Mn were treated by 2,4-di-tert-butylphenol (2,4-DTBP) at −80 °C, whose O–H BDFE (80.9 kcal/mol in THF)[29−32] is intermediate between the OO–H BDFEs of 3-Co and 3-Mn, no reaction was found in the mixture of 2-Co and 2,4-DTBP (Figure a), whereas the HAT reaction with 2-Mn proceeded to furnish 3-Mn and 2,4-di-tert-butylphenoxyl radical as monitored by the UV–vis measurements of the reaction mixture (Figure b). After workup by mixing the resulting solution with HCl, the organic product of dimeric 3,3′,5,5′-tetra-tert-butyl-2,2′-biphenol in the resulting mixture was estimated to be in 84% yield relative to 2-Mn by GC-MS detection (Figure S11). These experimental results clearly demonstrated that 2-Mn is a more potent H atom abstractor than 2-Co, in line with the varying OO–H BDFEs of 3-Co and 3-Mn (Scheme ).

Figure 5

UV–vis spectral changes of the reaction of (A) 2-Co and (B) 2-Mn (1.0 mM, blue trace) with 5 equiv of 4-oxo-TEMPOH in THF at −80 °C.

Figure 6

UV–vis spectral changes of the reactions of (a) 2-Co and (b) 2-Mn (1.0 mM, blue trace) with 2,4-DTBP (100 equiv) in THF at −80 °C.

Scheme 3

Reactivity of 2-Co and 2-Mn toward Three Different Substrates Possessing Varying O–H BDFEs

Calculations for BDFEs in THF

The BDFEMH of M–H was defined as follows:

Following our earlier protocol to compute hydricity,[33,34] a property closely related to BDFE, we first chose a range of species (Table S4) with known X–H BDFEs in THF and then calculated their ΔGM• and ΔGMH in THF by using the DFT approach (for details, see the Supporting Information). Finally, a linear equation between ΔGM• – ΔGMH and the experimental BDFE value was set up. Because ΔGH• in THF is a constant, it can be neglected in the linear regression analyses. The equation has a high degree of predicting power as suggested by R2 = 0.923, the mean absolute error, MAE = 2.62 kcal/mol, and the root-mean-square deviation, RMSD = 3.14 kcal/mol. As a consequence, the predicted OO–H BDFEs of 3-Co and 3-Mn are 74.5 and 79.3 kcal/mol, respectively, which are in good agreement with the experimental values within the uncertainty range of the calculations.

DFT Calculations of HAT Thermodynamics

To gain in-depth insight into the effect of the different metal centers on the OO–H BDFEs, we chose the homolytic cleavage of the O–H bond of a hydroperoxyl radical (HOO•) to form a triplet dioxygen (O2) and a hydrogen atom (H•), i.e., HOO• → O2 + H•, as the reference reaction and evaluated how the interaction of the divalent metal precursor, MII(BDPBrP) (M = Co (1-Co) and Mn (1-Mn)), with HOO• and O2 influences the OO–H BDFE. Specifically, for a given metal, a thermodynamic cycle was set up based on the Gibbs free energy charges in THF computed for the coordination of 1-M with HOO• and O2 to form 3-M and 2-M, respectively, and the difference between them (ΔΔG) corresponds to the change in the OO–H BDFE from its original value of HOO• to those found of 3-Co and 3-Mn.

To our knowledge, the experimental OO–H BDFE of HOO• in THF has not been reported yet, but those in H2O and DMSO are 60.6 and 58.0 kcal/mol,[35,36] respectively. As shown in Figure , those values are comparable to our computed ones in THF. Moreover, the OO–H BDFEs of 3-Co and 3-Mn (shown in maroon in Figure ) estimated by using the thermodynamic cycle agree reasonably well with the experimental values, which suggests the reliability of this approach to dissect OO–H BDFEs.

Figure 8

Thermodynamic cycles relating the BDFE of HOO• to those of (a) 3-Co, (b) 3-Mn, (c) 3-Co-HS, and (d) 3-Mn-HS. Experimental OO–H BDFEs are shown in blue, computed values based on the thermodynamic cycles are shown in maroon, and those obtained by the linear regression (Figure ) are shown in black.

Figure 7

Correlation plot between the calculated ΔGM• – ΔGMH and the experimental BDFEs of a range of H atom donors in THF (for details, please refer to the Supporting Information).

The coordination of HOO• and O2 to both divalent metal precursors to generate the hydroperoxo and superoxo complexes was computed to be all downhill. As elaborated for the bonding of a copper fragment to HO•,[25] such exergonicity primarily results from the overwhelming stabilizing effect of the charge transfer from the metal center to HOO• and O2 that dominates over the e–e repulsion. Remarkably, the exergonicity of the association of O2 was found to substantially fall behind that of the addition of HOO•. Relative to HOO• only having one π bond, O2 possesses a pair of two-center, three-electron π bonds with two electrons occupying the π bonding orbital and one electron populating the π* antibonding orbital. Consequently, O2 is expected to feature a significantly higher degree of resonance stabilization than HOO•. In line with this notion is that Borden, Hoffmann, and co-workers estimated the resonance energy of O2 to be 37 kcal/mol above that of two HOO•.[37] As such, the charge transfer from the metal center to O2 alleviates the resonance stabilization much more pronouncedly than that to HOO•, which likely accounts for the observation that the interaction of HOO• with the divalent metal precursors is more favored than that of O2. Consequently, the OO–H BDFEs of 3-Co and 3-Mn appreciably exceed that found for HOO•. Specifically, interaction of HOO• and O2 with 1-Co to generate 3-Co and 2-Co is exergonic by −20.4 and −3.4 kcal/mol, respectively; thus ΔΔG = 17.0 kcal/mol. The corresponding values for 3-Mn and 2-Mn are −32.6 and −10.6 kcal/mol; thus, ΔΔG = 21.9 kcal/mol. The more positive ΔΔG computed for the Mn system ultimately leads to a greater OO–H BDFE of 3-Mn.

To probe the effect of the hydrogen bond between the hydroperoxo ligand and the alkoxide group of the BDPBrP2– ligand in 3-Co and 3-Mn, we calculated the corresponding hypothetical hydroperoxo models in which the H atom of the OOH ligand points upward and hence has vanishing interaction with the supporting ligand. The thus-obtained OO–H BDFE of the Mn complex (72.0 kcal/mol) still surpasses that of the Co complex (67.7 kcal/mol).

Comparing the electronic structures of 3-Co and 3-Mn, we surmised that the metal spin state is likely a determinant of their differential OO–H BDFEs. To verify this hypothesis, we initially aimed to compute a hypothetical quartet model consisting of an SCo = 2 CoIII center antiferromagnetically coupled to a superoxo radical, which features a similar electronic structure to 2-Mn. Despite repeated attempts, the calculations did not converge to the desired electronic structure, but to an SCo = 1 CoIII center that is ferromagnetically coupled to a superoxo radical. Gratifyingly, when the interaction of the two fragments was transitioned from antiferromagnetic coupling to ferromagnetic coupling, we succeeded in locating a sextet model (2-Co-HS). On top of that, we estimated the OO–H BDFE of its corresponding ST = 2 hydroperoxo species (3-Co-HS). To make a reasonable comparison, we also calculated the OO–H BDFE of 3-Mn with respect to 2-Mn-HS (3-Mn/2-Mn-HS) in which a HS MnIII center (SMn = 2) is ferromagnetically coupled to a superoxo radical ligand. All theoretical results are summarized in Figure .

The calculations predicted that the OO–H BDFE of 3-Mn/2-Mn-HS slightly surpasses that of 3-Mn/2-Mn by only 4.0 kcal/mol, whereas the difference between the OO–H BDFE of 3-Co-HS/2-Co-HS and that of 3-Co/2-Co is as large as 10.5 kcal/mol. Of note, the former energy gap is close to the uncertainty range of our computations, but the latter is far beyond that. More importantly, when both metal centers are in the HS state, the computed OO–H BDFEs of 3-Co-HS/2-Co-HS and 3-Mn/2-Mn-HS are comparable. However, upon going from 2-Co-HS and 3-Co-HS to 2-Co and 3-Co, respectively, the changes of the Co spin state impart a markedly higher degree of the stabilization to the superoxo complex (−24.6 kcal/mol) than that to the hydroperoxo complex (−13.9 kcal/mol), which lowers the OO–H BDFE of 3-Co below that of 3-Mn in the end.

Electronic–Structure Analyses

During the formation of 2-Co-HS, in order to retain the Co HS state, the CoII center in 1-Co needs to transfer a β electron to the O2 π* orbital; therefore, the electron-donating orbital has to be one of the doubly populated t2g-derived orbitals of the CoII center (Figure ). As depicted in Figure , in 2-Co-HS the O2 πip* orbital in the Co–O–O plane interacts primarily with Co d to form a pseudo σ bond, because both fragment orbitals do not have optimal symmetry for efficient overlap. More importantly, the thus-generated bonding and antibonding molecular orbitals (MOs) labeled as d + πop* and d – πop*, respectively, have almost equal parentage of O2 and Co. As such, O2 coordination is accompanied by a partial electron transfer from the CoII center of 1-Co to the incoming O2 ligand. Furthermore, together with the singly populated O2 πop* MO perpendicular to the Co–O–O plane, the electronic structure of 2-Co-HS is hence best interpreted as a resonance hybrid ranging between HS CoII–O2 and HS CoIII–O2•.

Figure 9

Schematic electronic-structure evolution for the formation of 2-Co-HS and 2-Co by adding O2 to 1-Co.

Figure 10

Comparison of the geometric- and electronic-structure changes upon addition of HOO• and O2 to 1-Co to generate 3-Co, 3-Co-HS, 2-Co, and 2-Co-HS and molecular orbitals describing Co–OOH and Co–O2 interactions thereof. Red numbers are Chelpg charges computed for the O2 and OOH ligands.

Different from that of 2-Co-HS, the electronic-structure evolution for the generation of 2-Co can be formally interpreted as follows. Upon O2 approaching the CoII center in 1-Co, the latter species shifts two α electrons in the eg set into the two O2 π* orbitals, which is concurrent with transfer of a β electron from the O2 πop* to a Co t2g-derived orbital. As a consequence, the Co center of the resulting 2-Co complex possesses a low-spin state with two completely empty Co d and d orbitals. In 2-Co the O2 πip* orbital exclusively interacts with the Co d orbital, because both fragment orbitals have appropriate symmetry and their overlap is more efficient than the Co–O2 quasi σ bond found for 2-Co-HS involving Co d and O2 πip*. For 2-Co, in addition to two electrons occupying the O2 πip* orbital, one electron resides in the O2 πop* MO. This orbital occupation pattern therefore defines an authenticated superoxo ligand in 2-Co and suggests that the aforementioned electron transfer has already occurred. Consistent with this notion, the computed Chelpg charge[38] for the O2 ligand of 2-Co-HS is appreciably lower than that of 2-Co, and the same trend was also found for Mulliken and Löwdin charges (for details, please refer to Figures S12–14 in the Supporting Information). The ease of the Co-to-O2 electron transfer in the formation of 2-Co and 2-Co-HS can be ultimately traced back to the differential energy of the employed Co donating orbitals (EDOs). For 2-Co, one of the high-energy eg-derived orbitals functions as the EDO, whereas for 2-Co-HS a low-energy t2g-derived orbital acts as the EDO. Although the Co–O2 π interactions involving the Co d and O2 πop* fragment orbitals are almost the same for both superoxo species, the flip of the Co spin state replaces an approximately half pseudo σ bond in 2-Co-HS formed by Co d and O2 πip* with a σ bond in 2-Co formed by Co d and O2 πip*. This electronic-structure change thus induces considerable shortening of the Co–O2 bond distance found for 2-Co relative to 2-Co-HS.

Upon going from 2-Co to 3-Co, due to the interaction with the proton, the originally doubly populated O2 πop and πop* orbitals split into a lone pair that interacts with the metal center and an OO–H σ bonding orbital. Comparison of the estimated charges of the HOO moiety of 3-Co-HS and 3-Co showed that lowering the Co spin state also induces more favorable electron transfer from the metal center to HOO•. This is primarily due to the covalent Co–OOH π interaction in 3-Co-HS; therefore, three electrons are delocalized into the Co d and HOO πop* fragment orbitals, which thus defines a half Co–OOH π bond. Because 3-Co-HS features a HS state, its Co d orbital is singly occupied, which results in a half σ Co–OOH bond formed by Co d and HOO πip*. The Co–OOH bonding strength of 3-Co-HS is hence more or less the same as that of 3-Co, having mere one Co–OOH σ interaction, consistent with their nearly identical Co–OOH bond lengths. As such, one can conclude that the change of the Co spin state results in more pronounced electronic-structure changes in the superoxo complexes than those in the hydroperoxo complexes; consequently, the exergonicity computed for 2-Co-HS → 2-Co exceeds that for 3-Co-HS → 3-Co, which reduces the OO–H BDFE from 85.0 kcal/mol for 3-Co-HS to 74.5 kcal/mol for 3-Co. Taken together, the distinct metal spin state of homologous 3-Co and 3-Mn complexes is the main reason for their varying OO–H BDFEs.

Conclusion

We report here the HAT thermodynamics of the homologous CoIII- and MnIII-superoxo species 2-Co and 2-Mn in THF. Both complexes can be reversibly protonated to yield 4-Co and 4-Mn, which undergo reversible one-electron reduction to generate the corresponding CoIII- and MnIII-hydroperoxo complexes 3-Co and 3-Mn. The redox potentials of 3-Co and 3-Mn and the pKa values of 4-Co and 4-Mn have been directly determined by CV measurements and titration experiments. On the basis of them, the OO–H BDFEs of 3-Co and 3-Mn were estimated to be 79.3 and 81.5 kcal/mol, respectively, by using the thermodynamic scheme and Bordwell relationship. It has been experimentally shown that 2-Mn is capable of abstracting H atoms from 4-oxo-TEMPOH with an O–H BDFE of 68.9 kcal/mol and 2,4-DTBP with an O–H BDFE of 80.9 kcal/mol, but failed to activate the O–H bond of 3,5-DMP with an O–H BDFE of 85.6 kcal/mol, while 2-Co can react only with 4-oxo-TEMPOH. These experimental findings thus affirmed the calculated OO–H BDFEs of 3-Co and 3-Mn. To probe the effect of the metal spin state on OO–H BDFEs, we computationally investigated the corresponding hypothetical high-spin CoIII-superoxo and -hydroperoxo complexes 2-Co-HS and 3-Co-HS. Detailed theoretical analyses showed that 2-Co-HS features weaker CoIII–O2–• interactions relative to 2-Co, because in formation of 2-Co-HS a low-energy t2g-derived orbital serves as the EDO that reluctantly donates the electron into O2, whereas a high-energy eg-derived orbital functions as the EDO in the generation of 2-Co. Hence, 3-Co-HS possesses a much stronger OO–H bond than 3-Co and even surpasses that of 3-Mn. A different metal spin state is therefore responsible for the varying OO–H BDFEs of 3-Co and 3-Mn. The present work for the first time pinpoints, in addition to the metal oxidation state, the metal spin state, which is another crucial factor governing BDFEs. Therefore, our findings would add a new dimension to the design of transition metal systems that catalyze functionalization of substrates directly using O2 as the oxidant.

Experimental Section

Materials and Methods

All manipulations were operated under a nitrogen or argon atmosphere using standard Schlenk techniques or in a glovebox. Acetonitrile and dichloromethane were distilled under nitrogen over CaH2 prior to use; THF, MeTHF, pentane, and diethyl ether were purified by Na/benzophenone and distilled prior to use. n-PrCN was purified using Na2CO3 and KMnO4 and then distilled prior to use. Mn(BDPBrP) (1-Mn), Mn(BDPBrP)(O2•) (2-Mn), Mn(BDPBrP)(OOH) (3-Mn), [Mn(BDPBrP)(OOH)]+ (4-Mn), 1-hydroxy-2,2,6,6-tetramethylpiperidine, and 1-hydroxy-2,2,6,6-tetramethyl-4-oxo-piperidine were prepared by utilizing the method reported previously.[16,17,39−41] Other chemical reagents were obtained from commercial sources and used as received unless stated otherwise. UV–vis spectra were recorded with an Agilent 8454 spectrophotometer equipped with a cryostat from Unisoku Scientific Instruments, Osaka, Japan. Product analyses were performed with an Agilent GC6890 gas chromatograph (GC-MS). Elemental analyses for C, H, and N were performed on an Elementar Vario EL cube analyzer at the Instrumentation Center at National Taiwan University.

Synthesis of Co(BDPBrP) (1-Co)

1-Co was synthesized by reacting CoCl2 (26.0 mg, 0.2 mmol) with H2BDPBrP (184.6 mg, 0.2 mmol) and NaH (12.0 mg, 0.5 mmol) in 1:1 THF/CH3CN mixed solvent (20 mL) in a 50 mL Schlenk flask. The solution was stirred for 12 h at ambient temperature, and the solvent was removed under vacuum. The green residue was dissolved in CH2Cl2 (15 mL) and filtered. The filtrate was concentrated under vacuum and recrystallized by slow diffusion of Et2O into the concentrated THF filtrate at ambient temperature. Green crystals of 1-Co were obtained over 1 day in 62% yield. The molecular structure of 1-Co features a distorted square pyramidal geometry (τ5 = 0.47). UV–vis (THF): 375 nm (306 M–1 cm–1). Anal. Calcd for C41H41Br4CoN3O2·C4H8O (FW = 1054.41): C, 51.26; H, 4.30; N, 3.99. Found: C, 51.649; H, 4.432; N, 4.048.

Formation of Co(BDPBrP)(O2•) (2-Co)

Complex 2-Co was generated by bubbling O2 into a THF solution of 1-Co at −90 °C from an oxygen balloon for 2 min. Formation of 2-Co was monitored by UV–vis spectroscopy on characteristic absorption bands of 2-Co at 485 and 570 nm (Figure S1).

Formation of Co(BDPBrP)(OOH) (3-Co)

Reaction of 2-Co with excess TEMPOH in THF at −90 °C gave a deep blue solution of 3-Co, which displayed three absorption bands at 375, 580, and 640 nm (Figure S4).

X-ray Data Collection and Structure Determination

X-ray diffraction data of 1-Co was collected on a Bruker D8 Venture diffractometer employing Mo Kα radiation (λ = 0.7107 Å) at 200 K and with a θ–2θ scan mode. The space group for 1-Co was determined on the basis of systematic absences and intensity statistics. Their structures were solved by direct methods using SIR92 or SIR97 and refined using SHELXL-97 with anisotropic displacement factors for all non-hydrogen atoms. The detailed crystallographic data of 1-Co were provided in its CIF file.

EPR Measurements

X-Band CW EPR measurements were performed at the temperature of 100 K using a Bruker EMXmicro-6/1/S/L spectrometer equipped with a Bruker E4119001 HS-W1 resonator and Bruker E7003500 temperature controller. Microwave power was in the range of 17 to 20 mW. The magnetic field modulation amplitude was 10 G. Simulations with an S = 1/2 spin Hamiltonian including an electronic Zeeman term and a 59Co hyperfine interaction were carried out with EasySpin.[42]

Computational Setup

All computations were performed with the ORCA-4.2.0 program package.[43] Geometry optimizations were performed by using the PBE[44] functional in combination with the D3BJ noncovalent corrections.[45,46] The triple-ζ def2-TZVP basis set was used for the first coordination sphere, and the def2-SVP basis set for the remaining atoms.[47−49] Solvation effects were accounted for by the CPCM model[50] with THF being the solvent. To reduce computational costs, the resolution of identity (RI) approximation in tandem with the def2/J auxiliary basis set was employed.[51] Due to difficulties of SCF convergence in some cases, damping parameters were altered using the SlowConv function in ORCA. Frequency calculations confirmed that all optimized structures had converged to local minima on the potential energy surface (i.e., no imaginary frequencies). The PBE density functional is employed for geometry optimizations and frequency analyses, because, in addition to the high computational efficiency, calculations with GGA density functionals often deliver reliable structures and frequencies.[52,53] Final single-point calculations of electronic energies were carried out by using the TPSSh,[54] B3LYP,[55,56] and BP86 functionals[57,58] with the def2-TZVPP basis set for all atoms[59] and then benchmarked by DLPNO-CCSD(T) computations.[60,61] As elaborated in the Supporting Information, TPSSh gives the most satisfactory results of O–H BDFEs, as found in an earlier study.[25]