Intramolecular Hydrogen-Bond Interactions Tune Reactivity in Biomimetic Bis(μ-hydroxo)dicobalt Complexes

Abstract

Active site hydrogen-bond (H-bond) networks represent a key component by which metalloenzymes control the formation and deployment of high-valent transition metal-oxo intermediates. We report a series of dinuclear cobalt complexes that serve as structural models for the nonheme diiron enzyme family and feature a Co₂(μ–OH)₂ diamond core stabilized by intramolecular H-bond interactions. We define the conditions required for the kinetically controlled synthesis of these complexes: [Co₂(μ–OH)₂(μ-OAc)(κ¹-OAc)₂(py^R)₄][PF₆] (1^R), where OAc = acetate and py^R = pyridine with para-substituent R, and we describe a homologous series of 1^R in which the para-R substituent on pyridine is modulated. The solid state X-ray diffraction (XRD) structures of 1^R are similar across the series, but in solution, their ¹H NMR spectra reveal a linear free energy relationship (LFER) where, as R becomes increasingly electron-withdrawing, the intramolecular H-bond interaction between bridging μ–OH and κ¹-acetate ligands results in increasingly “oxo-like” μ–OH bridges. Deprotonation of the bridging μ–OH results in the quantitative conversion to corresponding cubane complexes: [Co₄(μ-O)₄)(μ₃-OAc)₄(py^R)₄] (2^R), which represent the thermodynamic sink of self-assembly. These reactions are unusually slow for rate-limiting deprotonation events, but rapid-mixing experiments reveal a 6000-fold rate acceleration on going from R = OMe to R = CN. These results suggest that we can tune reactivity by modulating the μ–OH pK_a in the presence of intramolecular H-bond interactions to maintain stability as the octahedral d⁶ centers become increasingly acidic. Nature may similarly employ dynamic carboxylate-mediated H-bond interactions to control the reactivity of acidic transition metal-oxo intermediates.

Graphical Abstract

INTRODUCTION

Over the last 2 billion years, life has evolved to wield the highly potent chemical reactivity of O₂, harnessing the energy gained from the O–O bond homolysis to drive the functionalization of some of nature’s strongest bonds (e.g., the C–H bond of methane, which has a BDE of 104 kcal/mol).¹ Among the enzymes capable of managing the reactivity of high-valent metal-oxo intermediates are the diverse and highly distributed family of nonheme diiron proteins.² Present in organisms across all kingdoms of life, these metallocofactors perform activities as “simple” as O₂ transport (hemerythrin) to those as thermodynamically challenging as methane hydroxylation (soluble methane monooxygenase, sMMO). In all cases, these dinuclear metalloenzymes feature histidine and carboxylate-rich active sites and bridging oxo, hydroxo, and carboxylato ligands. They also contain distinct H-bond networks that are critical to catalysis.^3–6 Such networks facilitate the gated and prompt transfer of protons with electrons that enable the enzymes to overcome kinetic and thermodynamic barriers to the multi-e⁻/multi-H⁺ processes that they mediate.⁷ These proton-coupled electron transfer (PCET) reaction steps are indispensable in maintaining control during O₂ activation and represent a mechanistic paradigm we need to understand if we are to successfully employ O₂ as the terminal electron donor in synthetic and artificial catalytic processes.

Carboxylate residues have been shown to play a key role in the PCET processes facilitated by nonheme diiron enzymes. Their ability to adopt multiple binding modes and serve as H-bond acceptors allow them to act as molecular switches in directing the synchronicity of individual mechanistic steps.^8,9 Specifically, they provide a means with which macroscopic protein conformational changes can be translated into individual PCET events. These types of reactions are referred to as “carboxylate shift reactions” and have been implicated in many dinuclear oxidase enzymes including sMMO,^10,11 β-hydroxylase Cm1A,¹² and ribonucleotide reductase.^13,14 Since carboxylate residues that are buried in a peptide fold exhibit minimal unique spectroscopic features, examining these types of mechanisms in biological systems remains a challenge. Accordingly, insights from synthetic model systems can provide opportunities to explore the consequences of carboxylate shift reactivity and define the specific ways in which the dynamicity of these ligands can be used in catalysis.^15–18

Here, we employ a complex first reported by Sumner, [Co₂(μ–OH)₂(μ-OAc)(κ¹-OAc)₂(py)₄][PF₆] (1^H),¹⁹ to add to these modeling efforts, extending them to explore the effects of carboxylate-mediated H-bond interactions. Numerous other examples of bis(μ-hydroxo)dicobalt(III) complexes have been reported.^19–28 The most similar example to the present work and to that of Sumner comes from Christou et al., who prepared an analogue of 1^H in which two 2,2′-bipyridine (bpy) replace four py ligands.²⁰ In both of these complexes, the Co₂(μ–OH)₂ diamond core is stabilized through intramolecular H-bond interactions with the κ¹-acetate ligands. This feature makes them attractive models for dinuclear M₂(μ–OH)₂ metallocofactor active sites.

We have now built on these works to prepare the homologous series 1^R in which R is the para-substituent of py and R = NMe₂, OMe, ^tBu, Me, H, C(O)Me, CF₃, or CN. These efforts allowed us to tune the pK_a and H-bond interactions present in the complexes and to observe LFERs by ¹H NMR and stopped-flow (SF) UV–vis absorption spectroscopies. We also report the conditions for kinetic control in the synthesis of these complexes, where the presence of acidic water is critical in preventing the formation of higher-order Co₄O₄ clusters, which represent the thermodynamic sink of self-assembly.^29,30 Together, these findings highlight the role that carboxylate-mediated H-bond interactions play in maintaining control over the formation and reactivity of transiently formed metal-oxo species.

EXPERIMENTAL SECTION

Materials.

All chemicals were used as received from the following suppliers: peracetic acid (32% in dilute acetic acid), high-purity grade (9385) silica gel, and 4-cyanopyridine (py_CN) were from Sigma-Aldrich; preparative TLC plates (silica gel GF, UV254, 1000 μm, 20 × 20 cm) from AnalTech; Co(OAc)₂·4H₂O from Mallinckrodt Inc.; NaOAc·3H₂O, MgSO₄, pyridine (py), ethyl ether, 4-N,N-dimethylaminopyridine (py^NMe2), and dichloromethane (DCM) from Fisher Scientific; 30% H₂O₂ from Macron Fine Chemicals; 7-methyl-1,5,7-triazabicyclo[4.4.0]dec-5-ene (MTBD) from TCI; ammonium hexafluorophosphate, 4-methoxypyridine (py^OMe), 4-^tbutylpyridine (py^tBu), 4-methylpyridine (py^Me), 4-acetylpyridine (py^C(O)Me), and 4-trifluoromethylpyridine (py^CF3) from Oakwood Chemical; deuterated solvents and pyridine-d₅ (py-d₅) 99.5% from Cambridge Isotope Laboratory; dry solvents were obtained from a Jorg C. Meyer Solvent Purification System; laboratory grade water (18.2 MΩ.cm) was obtained from a Milli-Q IQ 700 system. Final concentrations of solutions were confirmed by UV–vis absorption spectroscopy with the following experimentally obtained extinction coefficients: 1^H (ε₅₅₇ = 175 ± 2 M⁻¹ cm⁻¹); 1^OMe (ε₅₆₀ = 199 ± 3 M⁻¹ cm⁻¹); and 1^CN (ε₅₅₆ = 221 ± 11 M⁻¹ cm⁻¹).

Physical Methods and Instrumentation.

¹H NMR data were collected on a Varian Unity Inova 400 or Varian VXR 500 spectrometer and ¹³C NMR data was collected on a Carver B500 Bruker Advance III HD NMR spectrometer at ambient temperatures unless specified. Chemical shifts are referenced to residual NMR solvent peaks: CDCl₃ (δ 7.26 ppm), CD₃CN (δ 1.94 ppm for ¹H and δ 188.26 for ¹³C), MeOD (δ 3.31 ppm).³¹ UV–vis absorption spectra were obtained on an Agilent Technologies 8454 spectrophotometer, and SF UV–vis absorption spectroscopy was performed on an Applied Photophysics SX20 system equipped with a direct mount photodiode array. Both optical systems were held at 23 °C via Fisher Scientific ISOTEMP 1006S water circulators. Fourier transform infrared (FT-IR) spectra were recorded on a PerkinElmer Spectrum 100 FT-IR spectrometer using a CaF₂ cell for detection in solution phase and a PerkinElmer Universal ATR Sampling Accessory for solid phase measurements. High-resolution mass spectra (HRMS) were obtained on a Waters Synapt G2-Si ESI instrument at the UIUC School of Chemical Sciences Mass Spectrometry Laboratory. CHN and CHDN elemental analyses were performed on an Exeter Analytical CE 440 instrument at the UIUC School of Chemical Sciences Microanalysis Laboratory.

X-ray diffraction data were collected on a Bruker D8 Venture kappa diffractometer equipped with a Photon II CPAD detector. An Iμs microfocus source provided the Mo Kα radiation (λ = 0.71073 Å) that was monochromated with multilayer mirrors. The collection, cell refinement, and integration of intensity data were carried out with the APEX3 software.³² Multiscan absorption corrections were performed numerically with SADABS.³³ The initial structure solution was solved with either intrinsic phasing methods SHELXT³⁴ or direct methods and refined with the full-matrix least-squares SHELXL³⁵ program within the OLEX2³⁶ GUI. Detailed data collection parameters and refinement methods are included in Tables S1–S5 and the Supporting Information.

General Synthetic Procedures.

[Co₂(μ–OH)₂(μ-OAc)(κ¹-OAc)₂(py)₄][PF₆] (1^H) was synthesized by two different procedures: Protocol I produces 1^H in higher yield (34%) but cannot be used for the synthesis of derivatives in which R is more electron-withdrawing than H; in these cases, 2^R predominates. Protocol II produces 1^H in lower yield but is capable of producing 1^CN and 1^C(O)Me. Ligand exchange reactions from either 1^CN or 1^H provided all additional 1^R complexes as described below. For maximal yields and purity of products, 4-methoxypyridine, 4-methylpyridine, 4-acetylpyridine, and 4-trifluoromethylpyridine were used brand new from the manufacturer.

[Co₂(μ–OH)₂(μ-OAc)(κ¹-OAc)₂(py)₄][PF₆] (1^H).

Protocol I is explained in this paragraph. NaOAc·3H₂O (10.805 g, 79.4 mmol, 4 eq) in water was adjusted to pH 5.4 through addition of conc. HCl and then diluted to 28 mL with water. MeOH (28 mL) was added to this solution followed by Co(OAc)₂·4H₂O (5.00 g, 20.1 mmol, 1 eq) and py (5.14 mL, 54.4 mol, 2.7 eq). The solution was heated to reflux in a preheated oil bath (104 °C). H₂O₂ (30% solution, 2.98 mL, 29.2 mmol, 1.45 eq) was added dropwise over 7 min, producing bubbles and a black solution that was refluxed for an additional 45 min. After cooling to room temperature, MeOH was removed in vacuo. The remaining aqueous layer was extracted with DCM (3 × 70 mL), washed with brine (2 × 50 mL), and dried over Na₂SO₄. Removing the DCM in vacuo provided a burgundy solid which was dissolved in minimal amount of H₂O and layered with a solution of NH₄PF₆ (4.127 g, 25.3 mmol, 1.26 eq, in 34 mL H₂O) before storage overnight at 4 °C. The resulting precipitate was filtered and washed with cold H₂O and wet-loaded onto a 600 mL silica column. The product was eluted with 2.5% MeOH in DCM (R_f = 0.12). 1^H was the pink band that eluted first and was characterized by TLC (R_f = 0.32 in 5% MeOH in DCM). Pooled fractions were dried over MgSO₄, filtered through Celite, and concentrated. Overnight crystallization from layered DCM/pentane yielded 1^H as purple crystals (2.678 g, 33.8% yield). X-ray quality crystals (purple rectangular prisms) were obtained by vapor diffusion of petroleum ether in DCM at −20 °C. ¹H NMR (400 MHz, CDCl₃) δ 8.35–8.31 (m, 8H), 7.93 (tt, J = 7.5, 1.5 Hz, 4H), 7.53–7.48 (m, 8H), 7.40 (s, 2H), 2.33 (s, 3H), 2.10 (s, 6H). ¹H NMR data is in agreement with previous reports.^{19 1}H NMR (400 MHz, CD₃CN) δ 8.30–8.25 (m, 8H), 8.02 (tt, J = 7.6, 1.5 Hz, 4H), 7.54–7.49 (m, 8H), 7.26 (s, 2H), 2.25 (s, 3H), 2.03 (s, 6H).¹³C NMR (126 MHz, CD₃CN) δ 190.0, 186.5, 153.6, 140.8, 126.2, 55.3, 25.9, 25.8. Exact mass calcd for C₂₆H₃₁Co₂N₄O₈: 645.0806, found: m/z 645.0791. Anal. Calcd for C₂₆H₃₁Co₂F₆N₄O₈P•H₂O: C, 38.63; H, 4.11; N, 6.93, found: C, 38.96; H, 3.96; N, 6.96. Isolation of 2^H by this protocol (as the second fraction in chromatographic separations) was in agreement with previously reported characterization.²⁹

[Co₂(μ–OH)₂(μ-OAc)(κ¹-OAc)₂(py)₄][PF₆] (1^H).

Protocol II is explained in this paragraph. An aqueous 200 mL solution containing Co(OAc)2·4H₂O (10.050 g, 40.35 mmol, 1 eq) and NaOAc·3H₂O (54.913 g, 403.5 mmol, 10 eq) was adjusted to pH 3.90 by the addition of conc. HCl. The solution was transferred to a 500 mL RB flask equipped with a stir-bar, reflux condenser, and blast shield. The reaction flask was charged with py (6.50 mL, 80.7 mmol, 2 eq) followed by the slow addition of peracetic acid (32% wt. in dilute acetic acid, 12.70 mL, 1.5 eq) dropwise over 20 min. The resulting deep purple solution was slowly heated to reflux and held there for 13 min. A saturated solution of NH₄PF₆ in H₂O (50 mL) was then added, and the flask was removed from heat to afford a dark brown precipitate. The reaction flask was stored at 4 °C overnight, and the precipitate was filtered and washed with H₂O. The crude product was wet-loaded onto an 800 mL silica gel column and purified using a gradient of 2.5 to 10% MeOH in DCM. [Co₃(μ₃-O)(μ–OH)-(OAc)₅(py)₃][PF₆] (3^H) eluted first, and its ¹H NMR spectrum matched the previous reports.¹⁹ 1^H eluted second (R_f = 0.12 in 2.5% MeOH in DCM) and was dried over MgSO₄, filtered over Celite, and solvent removed in vacuo. 1^H was crystallized with pentane layered on DCM to provide the crystalline solid (0.467 g, 2.9% yield). Warning: peracetic acid has a flash point below 100 °C; therefore, a blast shield was used and the reaction flask heated slowly.

[Co₂(μ–OH)₂(μ-OAc)(κ¹-OAc)₂(py^NMe2)₄][PF₆] (1^NMe2).

4-N,N-Dimethylaminopyridine (0.622 g, 5.09 mmol, 12 eq) was added to a stirring solution of 1^H (0.336 g, 0.425 mmol, 1 eq) in CH₃CN (21.6 mL) and heated to 80 °C for 25 h. The solvent was removed in vacuo, and the resultant solid was triturated in diethyl ether for 30 min. The burgundy powder was filtered, washed with diethyl ether, and then purified by preparative TLC (R_f = 0.46 in 5% MeOH in DCM). The product was further purified by layering hexane on DCM to obtain crystalline material (0.127 g, 19.5%). X-ray quality crystals were obtained by vapor diffusion of petroleum ether into chloroform at −20 °C. ¹H NMR (400 MHz, CDCl₃) δ 7.71 (d, J = 6.6 Hz, 8H), 6.70 (s, 2H), 6.54 (d, J = 6.7 Hz, 8H), 3.08 (s, 24H), 2.13 (s, 3H) 2.01 (s, 6H). ¹H NMR (400 MHz, CD₃CN) δ 7.65–7.61 (m, 8H), 6.67–6.63 (m, 10H, overlapping with –OH), 3.05 (s, 24H), 2.12 (s, 3H), 1.93 (m, overlapping with CH₃CN). ¹³C NMR (126 MHz, CD₃CN) δ 188.7, 185.3, 156.4, 151.3, 118.3, 108.4, 39.5, 25.8, 25.7. Exact mass calcd. for C₃₄H₅₁Co₂N₈O₈: 817. 2494, found: m/z 817.2499. Anal. Calcd for C₃₄H₅₁Co₂F₆N₈O₈P: C, 42.42; H, 5.34; N, 11.64, found: C, 42.34; H, 5.37; N, 11.74.

[CO₂(μ–OH)₂(μ-OAc)(κ¹-OAc)2(py^OMe)₄][PF₆ (1^OMe).

4-Methoxypyridine (9.64 mL, 94.94 mmol, 150 eq) was added to a stirring solution of 1^H (0.500 g, 0.633 mmol) in CH₃CN (30 mL) and heated at 60 °C for 14 days. The reaction was cooled to room temperature, and the solvent removed in vacuo. The resulting purple oil was triturated in diethyl ether (40 mL) for 24 h. The purple powder was filtered, washed with diethyl ether, and wet-loaded onto a 200 mL silica column. The product was eluted with 2.5% MeOH in DCM (R_f = 0.12). The first pink band was collected, and the solvent removed in vacuo to yield the purple solid product characterized by TLC (R_f = 0.29 in 5% MeOH in DCM). X-ray quality crystalline material was obtained by crystallization in layered pentane/DCM over 2 days (0.2741 g, 47.6%). ¹H NMR (400 MHz, CDCl₃) δ 8.06 (d, J = 6.5 Hz, 8H), 7.08 (s, 2H), 7.05–6.95 (m, 8H), 3.94 (s, 12H), 2.25 (d, J = 0.9 Hz, 3H), 2.05 (s, 6H). ¹H NMR (400 MHz, CD₃CN) δ 8.03–8.00 (m, 8H), 7.06–7.03 (m, 8H), 7.02 (s, 2H), 3.94 (s, 12H), 2.21 (s, 3H), 2.00 (s, 6H). ¹³C NMR (126 MHz, CD₃CN) δ 189.7, 186.3, 169.4, 154.1, 112.4, 57.1, 25.8, 25.8. Exact mass calcd. for C₃₀H₃₉Co₂N₄O₁₂: 765.1228, found: m/z 765.1232. Anal. Calcd for C₃₀H₃₉Co₂F₆N₄O₁₂P: C, 39.57; H, 4.32; N, 6.15, found: C, 39.42; H, 4.26; N, 6.15.

[Co₂(μ–OH)₂(μ-OAc)(κ¹-OAc)₂(py^tBu)₄][PF₆] (1^tBu).

4-^tButylpyridine (0.93 mL, 6.321 mmol, 12 eq) was added to a stirring solution of 1^H (0.416 g, 0.527 mmol, 1 eq) in CH₃CN (28 mL) and heated at 80 °C for 45 h. The solvent was removed in vacuo and the remaining oily solid was triturated in hexanes for 24 h before filtering to afford a burgundy solid. The solid was purified by preparative TLC (R_f = 0.34 in 10% hexanes in diethyl ether) to afford 1^tBu as a burgundy solid (0.188 g, 35%). ¹H NMR (400 MHz, CDCl₃) δ 8.20 (d, J = 6.3 Hz, 8H), 7.46 (d, J = 6.4 Hz, 8H), 7.23 (s, 2H), 2.31 (s, 3H), 2.09 (s, 6H) 1.36 (s, 36H). ¹H NMR (400 MHz, CD₃CN) δ 8.13 (d, J = 6.1 Hz, 8H), 7.55–7.51 (m, 8H), 7.25 (s, 2H), 2.23 (s, 3H), 2.02 (s, 6H), 1.35 (s, 36H). Exact mass calcd. for C₄₂H₆₃Co₂N₄O₈: 869.3310, found: m/z 869.3306. Anal. Calcd for C₄₂H₆₃Co₂F₆N₄O₈P: C, 49.71; H, 6.26; N, 5.52, found: C, 50.14; H, 6.4; N, 5.74.

[CO₂(μ–OH)₂(μ-OAc)(κ¹-OAc)₂(py^Me)₄][PF₆] (1^Me).

4-Methylpyri-dine (9.24 mL, 94.94 mmol, 150 eq) was added to a stirring solution of 1^H (0.500 g, 0.633 mmol) in CH₃CN (30 mL) and heated at 60 °C for 14 days. The reaction was cooled to room temperature and the solvent removed in vacuo. The resulting brown-purple oil was triturated in diethyl ether (40 mL) for 24 h. The purple powder was filtered, washed with diethyl ether, and wet-loaded onto a 200 mL silica column. The product was eluted with 2.5% MeOH in DCM. The first pink band was collected and the solvent was removed in vacuo to yield 1^Me as a purple solid (0.187 g, 34.9%) characterized by TLC (R_f = 0.33 in 5% MeOH in DCM). ¹H NMR (400 MHz, CDCl3) δ 8.09 (d, J = 6.0 Hz, 8H), 7.27 (d, J = 6.2 Hz, 9H, overlaps with CDCl₃), 7.18 (s, 2H), 2.49 (s, 12H), 2.24 (s, 3H), 2.05 (s, 6H). ¹H NMR (400 MHz, CD₃CN) δ 8.09–8.05 (m, 8H), 7.36–7.33 (m, 8H), 7.11 (s, 2H), 2.49 (s, 12H), 2.21 (s, 3H), 2.00 (d, J = 1.5 Hz, 6H).¹³C NMR (126 MHz, CD₃CN) δ 189.7, 186.3, 153.4, 152.7, 127.0, 25.9, 25.8, 21.0. Exact mass calcd. for C₃₀H₃₉Co₂N₄O₈: 701.1432, found: m/z at 701.1429. Anal. Calcd for C₃₀H₃₉Co₂F₆N₄O₈P: C, 42.57; H, 4.64; N, 6.62, found: C, 42.49; H, 4.6; N, 6.7.

[CO₂(μ–OH)₂(μ-OAc)(κ¹-OAc)₂(py^C(O)Me)₄][PF₆] (1^C(O)Me).

Protocol II was followed under the following conditions: Co(OAc)₂·4H₂O (2.592 g, 10.40 mmol, 1 eq), Na(OAc)·3H₂O (17.000 g, 124.90 mmol, 12 eq), 4-acetylpyridine (2.30 mL, 20.80 mmol, 2 eq), and peracetic acid (3.28 mL, 1.5 eq) in total volume of 50 mL of H₂O with a final pH of 3.99. The solution was refluxed for 13 min, and saturated NH₄PF₆ (30 mL) was added. An extra step was included in this work up: the precipitate was dissolved in a minimal amount of H₂O and extracted into DCM (3 × 65 mL), dried with MgSO₄, and filtered over Celite. The solvent was removed in vacuo prior to loading the crude product onto a silica column (R_f = 0.13 in 2% MeOH in DCM,). Fractions were further purified by preparative TLC if needed to yield a pink powder characterized by TLC (R_f = 0.45 in 7.5% MeOH in DCM) (0.362 g, 7.26%). ¹H NMR (400 MHz, CDCl₃) δ 8.50 (d, J = 6.2 Hz, 8H), 7.96–7.89 (m, 8H), 7.69 (s, 2H), 2.70 (s, 12H), 2.35 (s, 3H), 2.11 (s, 6H). ¹H NMR (400 MHz, CD₃CN) δ 8.49–8.45 (m, 8H), 7.91–7.87 (m, 8H), 7.51 (s, 2H), 2.65 (s, 12H), 2.30 (s, 3H), 2.07 (s, 6H). ¹³C NMR (126 MHz, CD₃CN) δ 197.6, 190.4, 186.9, 154.8, 146.5, 123.7, 27.4, 26.0, 25.9. Exact mass calcd. for C₃₄H₃₉Co₂N₄O₁₂: 813.1228, found: m/z 813.1238. Despite repeated attempts, 1^C(O)Me could not be recrystallized and the EA failed. The closest to target was as follows: Anal. Calcd for C₃₄H₃₉Co₂F₆N₄O₁₂P: C, 42.6; H, 4.1; N, 5.85, found: C, 46.46; H, 4.37; N, 6.01.

[Co₂(μ–OH)₂(μ-OAc)(κ¹-OAc)₂(py^CF3)₄] (1^CF3).

4-Trifluoromethylpyridine (0.48 mL, 4.212 mmol, 75 eq) was added to a stirring solution of 1^CN (0.050 g, 0.0564 mmol, 1 eq) in CH₃CN (5 mL), and the mixture was heated at 60 °C for 22 h. The solvent was removed in vacuo, and the resultant solid was purified via preparative TLC (3% MeOH in DCM). The first pink band was collected and characterized by TLC (R_f = 0.29 in 5% MeOH in DCM). To purify further, the solid was dissolved in CH₃CN and extracted with pentane; the CH₃CN layer was collected and the solvent was removed to afford a pink solid. X-ray quality crystalline material was obtained by crystallization in layered toluene/chloroform over 1 week at room temperature (0.021 g, 34.9%). ¹H NMR (400 MHz, CDCl₃) δ 8.59 (d, J = 6.1 Hz, 8H), 7.77 (d, J = 5.9 Hz, 8H), 7.69 (s, 2H), 2.36 (s, 3H), 2.14 (s, 6H). ¹H NMR (400 MHz, CD₃CN) δ 8.54 (d, J = 6.0 Hz, 8H), 7.83 (d, J = 6.1 Hz, 8H), 7.63 (s, 2H), 2.30 (s, 3H), 2.08 (s, 6H). Exact mass calcd. for C₃₀H₂₇Co₂F₁₂N₄O₈: 917.0301, found: m/z 917.0301. Anal. Calcd for C₃₀H₂₇Co₂F₁₈N₄O₈P: C, 33.92; H, 2.56; N, 5.27, found: C, 33.74; H, 2.64; N, 5.35.

[Co₂(μ–OH)₂(μ-OAc)(κ¹-OAc)₂(py^CN)₄][PF₆] (1^CN).

Protocol II was followed with the following amounts and conditions: an 100 mL aqueous solution of Co(OAc)₂·4H₂O (5.000 g, 20.07 mmol, 1 eq) and NaOAc·3H₂O (27.320 g, 200.70 mmol, 10 eq) adjusted to pH 4.0, 4-cyanopyridine (4.180 g, 40.10 mmol, 2 eq), peracetic acid (32% wt. in dilute acetic acid, 6.33 mL, 1.5 eq), and 25 mL of NH₄PF₆ saturated solution. The crude product was wet-loaded onto a silica-column (R_f = 0.51 in 25% acetone in DCM). The first pink band was isolated and the solvent was removed in vacuo to yield a pink powder (0.587 g, 6.57%). X-ray quality crystals were obtained by layering toluene onto acetonitrile at room temperature over the course of few days. ¹H NMR (400 MHz, CDCl₃) δ 8.58 (d, J = 6.0 Hz, 8H), 7.79 (d, J = 6.4 Hz, 10H, overlapping with –OH), 2.42 (s, 3H), 2.13 (s, 6H). ¹H NMR (400 MHz, CD₃CN) δ 8.52–8.45 (m, 8H), 7.89–7.82 (m, 8H), 7.62 (s, 2H), 2.28 (s, 3H), 2.06 (s, 6H).¹³C NMR (126 MHz, CD₃CN) δ 190.8, 187.3, 154.7, 128.2, 124.7, 118.3, 116.2, 55.2, 25.8. Exact mass calcd. for C₃₀H₂₇Co₂N₈O₈: 745.0616, found: m/z 745.0624. Anal. Calcd for C₃₀H₂₇Co₂F₆N₈O₈P: C, 40.47; H, 3.06; N, 12.58, found: C, 40.17; H, 3.06; N, 12.39.

[Co₂(μ–OH)₂(μ-OAc)(κ¹-OAc)₂(py-d₅)₄][PF₆] (1^D).

Pyridine-d₅ (py-d₅, 0.675 mL, 8.424 mmol, 75.6 eq) was added to a stirring solution of 1^CN (0.099 g, 0.1115 mmol, 1 eq) in CH₃CN (10 mL), and the mixture was heated at 60 °C for 13 h. The solvent was removed in vacuo, and the resulting oil was triturated in diethyl ether for 3 h. The pink powder was filtered and washed with diethyl ether. The product was further purified by layering pentane on DCM to obtain crystalline material (0.032 g, 34.9%). ¹H NMR (400 MHz, CD₃CN) δ 7.26 (s, 2H), 2.25 (s, 3H), 2.03 (s, 6H). Exact mass calcd. for C₂₆H₁₁D₂₀Co₂N₄O₈: 665.2061, found: m/z 665.2062. Anal. Calcd for C₂₆H₁₁D₂₀Co₂F₆N₈O₈P·H₂O: C, 37.69; H, 1.59; D, 4.85; N, 6.76, found: C, 37.55; H, 1.35; D, 4.90; N, 6.82.

Reactions with MTBD Observed by ¹H NMR Spectroscopy.

Solutions (0.5 mL) of 10 mM 1^CN, 1^H, or 1^OMe and 20 mM MTBD in CD₃CN were premixed, transferred to an NMR tube, and incubated at ambient temperature between data collection points. [MTBDH][BF₄] was synthesized according to published procedures.³⁷ Standard samples of 2^CN, 2^H, and 2^OMe were prepared according to previously published procedures.^29,38 Low temperature measurements were performed by placing freshly prepared samples immediately in an acetonitrile/dry ice bath before being transferred into precooled Varian VXR 500 Spectrometer at −35 °C. For later time points, samples were incubated at room temperature between spectra collected in the precooled instrument at −35 °C.

Crossover Experiments Performed with 1^H and 1^D.

Simultaneous preparations of 0.5 mL solutions containing 5 mM 1^H, 5 mM 1^D, and 20 mM MTBD were mixed and incubated at ambient temperature for 16 h prior to injection and analysis by both high- and low-resolution ESI-MS. Control reaction mixtures to assess background levels of ligand substitution contained either 5 mM 1^H and 5 mM 1^D (no MTBD), 10 mM 1^H and 20 mM py-d₅, or 5 mM 2^H with both 20 mM py, and 20 mM py-d₅.

Kinetics Measurements and Data Analysis.

The conversion of 1^R to 2^R was monitored by UV–vis absorption spectroscopy in the presteady state. Rapid mixing with five different concentrations of base (MTBD) under pseudo-first order conditions (20–100 eq of MTBD) on three independently prepared samples of 1^CN, 1^H, and 1^OMe provided 45 individual measurements. All experiments were performed at 23 °C.

MTBD solutions were individually prepared using volumetric glassware and dry CH₃CN. To ensure sample freshness, the reagent was stored under inert atmosphere at 4 °C. Fresh solutions of 3.00 mM 1^R in dry CH₃CN were prepared for each of three independent trials to provide 1.5 mM upon mixing. Solid samples of 1^R were weighed on an analytical balance and brought to volume in volumetric glassware. Final concentrations of 2^R were analyzed by UV–vis absorption spectroscopy with the following extinction coefficients: 2^H (ε₆₉₄ = 286 ± 13 M⁻¹ cm⁻¹), 2^OMe (ε₆₉₄ = 292 ± 16 M⁻¹ cm⁻¹), 2^CN (ε₆₉₄ = 300 ± 5 M⁻¹ cm⁻¹). Owing to the large differences in rates for the three complexes, both SF (1^CN and 1^H) and hand-mixing (1^OMe) experiments were performed, which had 2 ms or 5 s deadtimes, respectively. Reaction traces were fit over five lifetimes to eq 1 using OriginPro 2019

$$\mathit{y}={\mathit{y}}_0+{\mathit{A}}_1{\mathit{e}}^{-\mathit{x}/{\mathit{t}}_1}+{\mathit{A}}_2{\mathit{e}}^{-\mathit{x}/{\mathit{t}}_2}$$ (1)

The inverses of t₁ and t₂ were averaged across three independently prepared samples to provide the rate constants listed in Tables 2 and S6–S8. Error limits encompass 1σ for three independently prepared samples, compounded with error associated with fitting to eq 1.

Table 2.

Rate Constants for Base-Dependent Conversion of 1^R to 2^R

R	kdeprot (M−1 s−1)a
CN	300 ± 30
H	0.219 ± 0.004
OMe	0.048 ± 0.005

^aUncertainties are 1σ for the average of three data sets, along with error associated with fitting to eq 1.

RESULTS

First reported by Sumner in 1988, 1^H shown in Scheme 1 forms as a mixture with oxo-centered trinuclear [Co₃(μ₃-O)(μ–OH)(μ-OAc)₅(py)₃]PF₆ (3^H).^19,30,39 We hypothesized that the unique H-bond interaction between the κ¹-OAc ligands and bridging hydroxo moieties in 1^H could serve as a foray into the study of carboxylate-mediated reactions in biomimetic dinuclear systems. Upon discovering that 1^H can be formed from the same starting materials as those reported by Das et al. for the preparation of tetranuclear cubane clusters, Co₄(μ₃-O)₄(μ-OAc)₄(py)₄ (2^H),²⁹ we decided to determine which conditions lead to the formation of 1^H over 2^H. Our findings to this end are summarized in Scheme 1.

Scheme 1.

Synthesis of 1^H, 2^H, and 3^H

By modulating solvent, solution pH, oxidant strength, and length of reflux time relative to conditions previously reported by Sumner and Das, we found that 1^H can be captured as the kinetic product of the reaction of Co^II(OAc)₂, sodium acetate, and pyridine under aqueous acidic conditions. The synthesis of 2^H is typically achieved in methanol, with hydrogen peroxide as the oxidant, and by refluxing the mixture for 4 h (Scheme 1).²⁹ In contrast, the reported synthesis of 1^H and [Co₃(μ₃-O)(μ–OH)₂(μ-OAc)₃(py)₅][PF₆]₂ uses water and acetic acid in a 5:1 ratio as the solvent, peracetic acid as the oxidant, and a shorter, 15 min reflux.¹⁹ Interestingly, using water as the solvent but otherwise keeping reaction conditions consistent with those reported for the preparation of 2^H produces only trace amounts of 1^H. It is only when the length of reflux is shortened to under an hour, and the pH of the solution brought to under 6, that detectable amounts of 1^H are formed.

To determine the effect of pH on the formation of 1^H, we incorporated a buffer system of AcOH/NaOAc and adjusted the pH with HCl. The yield of 1^H increased as the pH decreased from 8 to 4. Under acidic conditions, 2^H was not detected. Finally, we found that using the stronger oxidant, peracetic acid, rather than hydrogen peroxide, also increases the yield of 1^H relative to that of 2^H. Despite these optimizations, we note that 1^H is never the main product formed. Rather, it is always isolated in smaller amounts relative to either 2^H or 3^H. However, under optimized conditions, we were able to isolate 1^H as a purple crystalline solid in 34% yield.

Examination of the X-ray diffraction (XRD) structure of 1^H (Figure 1c) reveals O…O distances of 2.646 and 2.629 Å between bridging μ–OH ligands and κ¹-OAc H-bond acceptors (Table 1). With an aim to explore the impact of this intriguing H-bond interaction, we set out to functionalize 1^H with pyridine derivatives containing various substituents in the para-position (py^R) to provide 1^R. In general, ligand exchange reactions could be performed for the exchange of stronger for weaker σ-donating ligands. For R = NMe₂, OMe, ^tBu, and Me, ligand exchange reactions were performed on 1^H; for R = CF₃ and py-d₅ ligand exchange was performed on 1^CN. Complexes with R = C(O)Me and CN were synthesized de novo as outlined in the Experimental section.

Figure 1.

Thermal ellipsoid (50%) plots of the solid-state structures of 1^CN (a), 1^CF3 (b), 1^H (c), 1^OMe (d), and 1^NMe2 (e). Green, blue, red, orange, black, and white ellipsoids represent Co, N, O, F, C, and H, respectively. H Atoms not involved in H-bonding interactions, counterions, and cocrystallized solvent molecules are omitted for clarity.

Table 1.

Selected Bond Lengths (Å), Distances (Å), and Angles (deg) in 1^CN, 1^CF3, 1^H, 1^OMe, and 1^NMe2 Structures^a

	1CN	1CF3	1H	1OMe	1NMe2
O1⋯O3	2.639(3)	2.680(2)	2.646(2)	2.653(2)	2.728(3)
O2⋯O4	2.695(3)	2.638(2)	2.629(2)	2.643(2)	2.653(3)
O1’⋯O3′b	2.621(6)	2.59(4)
O2’⋯O4’b	2.664(5)	2.58(4)
Co⋯Co	2.813(6)	2.7992(3)	2.816(2)	2.802(3)	2.827(6)
Co⋯Navg	1.96 ± 0.02	1.96 ± 0.01	1.952 ± 0.005	1.949 ± 0.007	1.949 ± 0.009
Co⋯(μ-O)avg	1.892 ± 0.006	1.901 ± 0.011	1.893 ± 0.005	1.893 ± 0.004	1.903 ± 0.006
Co–O–Coavg	96.1 ± 0.4	95.5 ± 0.9	96.1 ± 0.3	95.6 ± 0.3	96.0 ± 0.2

^aParentheses represent uncertainty from modeling XRD data, while standard deviations from averaging are represented as ±1σ.

^bDisorder in these structures required two models. Both models are provided in Figure S1.

Single crystal XRD structures of 1^CN, 1^CF3, 1^H, 1^OMe, and 1^NMe2 are presented in Figure 1. There is minimal structural variation across the series. Each Co^III center resides in a nearly perfect octahedral coordination geometry, with distortion factors of ~10⁻⁴.⁴⁰ Notably, all reported Co^III₂(μ–OH)₂ complexes are also nearly perfectly octahedral.^19–28 The Co…Co distances for 1^CN, 1^CF3, 1^H, 1^OMe, and 1^NMe2 span 2.7992(3) – 2.827(6) Å, (Table 1), placing them in the center of the range reported for other Co^III₂(μ–OH)₂ complexes (2.6577(4) – 2.944(1) Å).^{19,20,24–28} Among these complexes, the Co…Co distance increases as the number of bridging ligands decreases. The largest structural variations across the 1^R series occur in the H-bond distances between μ–OH and κ¹-OAc oxygen atoms (Table 1). Fitting the diffraction data for 1^CN and 1^CF3 required two structural models in which the κ¹-OAc ligands adopt both possible orientations. These models are presented in Figure S1. It is interesting to note that only derivatives with electron-withdrawing pyridines ligands featured this disorder.

Similarly to the XRD data, UV–vis absorption and IR spectroscopies also did not reveal significant variances or patterns across the series of complexes. The λ_max of d → d transitions measured by UV–vis absorption spectroscopy are provided in Table S9, and the solid- and solution-state IR spectra are provided in Tables S10–S11 and Figures S2–S3, respectively. Here, the most notable changes across the series correspond to alkene and imine stretching modes associated with py ligands, while frequencies that correlate with symmetric and antisymmetric OAc stretching modes¹⁷ remain essentially unchanged (Figures S2–S3). However, we note that the doublet centered at approximately 1380 cm⁻¹ in both solid- and solution-phase measurements demonstrates a trend in which the lower energy shoulder increases in intensity as R becomes more electron-withdrawing, while the higher energy shoulder increases in intensity as R becomes more electron-donating. Work is underway to substitute carboxylate ligands and perform isotopic labeling studies to investigate these observations further.

To directly assess how H-bond interactions might change across the series, we turned to a technique that is more sensitive to the detection of protons: ¹H NMR spectroscopy. This technique revealed a direct correlation between py^R substitution and intramolecular H-bond interactions (Figure 2). The μ–OH resonance in 1^H appears at δ7.40 ppm in CDCl₃ (Figure S4), supporting that the H-bonded proton is deshielded relative to a proton that is not involved in a strong, directed, H-bond interaction. Figure 2 provides a plot of the chemical shift of the μ–OH resonances of 1^R relative to that of 1^H as a function of Hammett parameter (σ_p) for the para-substituent on pyridine.⁴¹ These data reveal a LFER that has a slope of 0.73 in CDCl₃ (Figure 2) and 0.64 in CD₃CN (Figures S5 and Table S12). These findings are similar to the LFER reported for 2^R, where substitution of py for py^R modulates reduction potentials by 0.17 V.³⁸ Our findings support that increasingly electron-poor pyridine ligands deshield the μ–OH proton via H-bond interaction with the κ¹-acetate ligands.

Figure 2.

¹H NMR chemical shifts of μ–OH protons in 1^R derivatives in CDCl₃ (R-groups as indicated) relative to 1^H, whose resonance occurs at δ7.40 ppm plotted as a function of Hammett parameter (σ_p) as defined in ref 41. The solid line is the linear least-squares fit with slope ρ = 0.73 and R² = 0.987.

To investigate variance in the pK_as of bridging hydroxo ligands across the series, we sought to assess the deprotonation reactions of 1^R. However, with six d-electrons localized on each Co^III ion, molecular orbital theory suggests that deprotonation would result in an unstable Co^III₂(μ-O)₂ intermediate.⁴² In agreement with this prediction, we found that addition of two eqalents of the strong, non-nucleophilic base 7-Methyl-1,5,7-triazabicyclo[4.4.0]dec-5-ene (MTBD), which has a pK_a 25.5 in CH₃CN,⁴³ results in the quantitative conversion of 1^R to the corresponding 2^R derivatives. As shown in Scheme 2, we observe the clean conversion of 1^R to 2^R along with the byproducts expected, in the quantities expected, based on the reaction stoichiometry. A similar observation was made by Christou et al. upon the reaction of [Co₂(μ–OH)₂(μ-OAc)(κ¹-OAc)₂(bpy)₂][ClO₄] with Li₂O₂, which resulted in the formation of the analogous tetrameric cubane complexes.²⁰

Scheme 2.

Deprotonation Converts 1^R to 2^R

Figure 3 displays truncated ¹H NMR spectra for the reactions of 1^CN, 1^H, and 1^OMe with 2 eq MTBD after 1 h and when completed, along with the spectra of pure 1^CN, 1^H, or 1^OMe and their corresponding 2^R derivatives. Figures S6–S11 provide full NMR spectra for these reactions at additional time points, along with the spectra of MTBD and protonated [MTBDH][BF₄]. We can account for every peak in the reaction spectra shown in Figures 3 and S6–S11 based on the stoichiometry of the reaction shown in Scheme 2. The 4 eq of free py^R released during the reaction appear as broad features in the aromatic region of the spectra. We reason that this broadening occurs as the result of H-bonding interactions with other byproducts released during the reaction. The aliphatic peaks associated with MTBD shift downfield as the reaction progresses toward MTBDH⁺ (Figures S6, S8, and S10). Reactions do not proceed to completion in the presence of just 1 eq of MTBD and only proceed to 2^R in aprotic solvents.

Figure 3.

Truncated ¹H NMR spectra of 1^CN (a), 1^H (b), and 1^OMe (c) before (black), 1 h after (dark gray), and 24 h after addition of 2 eq MTBD (light gray), and spectra of purified 2^R (blue). Reactions were conducted in CD₃CN at room temperature.

The kinetics of 1^R to 2^R conversion were examined through a series of rapid-mixing experiments monitored by UV–vis absorption spectroscopy. The UV–vis absorption spectra of 1^H and 2^H are displayed in Figure 4, and those of 1^CN/2^CN and 1^OMe/2^OMe are shown in Figures S12 and S13, respectively. These spectra highlight the fact that 2^R absorbs ~4× stronger than 1^R in the UV region, and ~2× stronger in the visible region. Therefore, it was not possible to track the reaction kinetics by monitoring the decrease in 1^R. Accordingly, we elected to track the kinetics of product formation by monitoring the reaction at 694 nm, where only 2^R absorbs and 1^R are optically silent.

Figure 4.

UV–vis absorption spectra of 0.30 mM 1^H (solid black line), 0.075 mM 2^H (solid grey line), and inset 3.00 mM 1^H (dotted black line), and 1.50 mM 2^H (dotted grey line). Spectra were collected in CH₃CN and plotted as a function of extinction coefficient.

Using a combination of SF (dead time = 2 ms) and hand-mixing (dead time = 5 s) methodologies, the kinetics of 2^R formation were examined under pseudo-first order conditions. Equal volumes of 3 mM (1.5 mM final conc.) 1^CN, 1^H, or 1^OMe were mixed with a large excess (≥20 eq) of MTBD, and the resulting optical spectral changes were monitored at 694 nm. Exemplary data sets are shown in Figures S14 and S15 for 1^CN and 1^OMe, respectively, and in Figures 5a and S16 for 1^H. In all cases, five freshly prepared MTBD samples were mixed with an independently prepared sample of 1^R. This process was repeated three times for each derivative. For each sample type, the choice of MTBD concentration range was dictated by the limitations of the experimental setup, solubilities of derivatives, and the rates of the reactions. SF mixing experiments with 1^CN required lower concentrations of MTBD to ensure that a significant portion of the absorbance change could be observed in the early time points. In contrast, hand-mixing experiments with 1^OMe required higher concentration ranges to ensure that experiments finished in a timely manner. We note that under these high [MTBD] conditions, a small amount of product precipitation was observed and results in diminished absorbance changes in the kinetics traces (Figure S15). These effects may introduce some additional uncertainty in the values presented in Table 2, however we do not anticipate that they will change the overall trend observed, namely that rate constants vary over 4 orders of magnitude depending on the identity of R in 1^R.

Figure 5.

Absorbance changes at 694 nm upon mixing equal volumes of 3.0 mM 1^H with 25 (red), 30 (green), 40 (blue), 60 (gray), and 70 eq (black) of MTBD, fit to eq 1 (dashed lines) (a); plots of k₁ vs [MTBD] fit to a linear least-squares regression (dashed line) for 1^H (black), 1^CN (blue), and 1^OMe (red) (b). Error bars represent 1σ from averaging across three independently prepared samples, as well as error associated with the exponential fits to eq 1.

All reactions exhibited biphasic kinetics (eq 1) where the two phases each contribute substantially to the total absorbance change observed (Tables S7–S9). Exemplary data sets fit to monoexponential and biexponential rises are provided in Figures S14–S16. As shown in Figure 5b, the slow phase obtained from these data (k₁) shows a linear dependence on [MTBD]. In contrast, the fast phase (k₂) showed no [MTBD] dependence (Figure S17). Along with the plotting of k₁ vs [MTBD] and fitting to a linear least-squares regression (Figure 5b), the slope provided the bimolecular rate constants, k_deprot, listed in Table 2. Rate constants obtained in this way follow a LFER for para-substituent effects⁴¹ with an R² of 0.996, and a slope of 4.1 ± 0.3 (Figures S18).

Intermediates in the conversion of 1^R to 2^R were not observed optically or by low temperature (−35 °C) ¹H NMR spectroscopy across a range of acquisition times (0.1–4 s, Figures S19–S20). Along with evidence from rapid-mixing experiments, these observations provide the basis for the mechanistic model presented in eq 2 of the Discussion section, where rate-limiting deprotonation is followed by fast ‘oligomerization steps.’ A possible mechanism for this oligomerization process could be the dimerization of 2 eq 1^R to give 1 eq 2^R. We investigated this possibility by performing the crossover experiments described in Table 3. Mixing 0.5 eq of 1^H and 0.5 eq 1^D (where 1^D is the derivative of 1 that contains per-deuterated py-d₅ ligands) with 2 eq of MTBD, we examined the resulting product distribution by mass spectrometry. We postulated that if 2^H forms through dimerization of 1^H, then we would observe the formation of the molecular ion peak (M⁺), M⁺ + 10 amu, and M⁺ + 20 amu in a 1:2:1 ratio. If instead, 2^H is formed via a more complicated mechanism, then we would observe M⁺, M⁺+5, M⁺+10, M⁺+15, and M⁺+20 species in the relative ratios 1:4:6:4:1 based on the amounts of free py and py-d₅ generated in solution during the reaction (Scheme 2).

Table 3.

Crossover Experiments for 1^H to 2^H Conversion Analyzed by ESI-MS (a.m.u.)^a

	M+	+5	+10	+15	+20
1H + 2py-d5	1	0	0	0	0
2H + 4py +4py-d5	1	0.11	0	0	0
1H + 1D + 4MTBD	0.21	0.75	1	0.56	0.13

^aRelative abundances of heavy derivatives relative to the molecular ion peak (M⁺) observed by positive-mode ESI-MS 16 h after mixing at room temperature.

Control experiments assessing the extent of ligand exchange in the absence of MTBD were performed and in accordance with previous reports,^38,44 and ligand field arguments,⁴⁵ we found that the py ligands of 2^H are somewhat labile, resulting in 11% exchange to give the monosubstituted M⁺+5 complex during the course of the 16 h incubation period with py-d₅ (Table 3, entry 2, Figure S21). In contrast, 1^H did not undergo ligand exchange either in the presence of free py^D (Table 3, entry 1, Figure S22) or in the presence of 1^D (Figure S23). In the presence of MTBD, a statistical distribution of mono-, di-, tri-, and tetra-substituted complexes were observed (Table 3, entry 3, Figure S24), ruling out the possibility of a dimerization mechanism.

DISCUSSION

In contrast to the kinetically controlled biological assembly of metalloprotein active sites, laboratory syntheses of multi-nuclear metal complexes often rely on thermodynamic control to dictate the number and configuration of metal ions incorporated.^45,46 The tetranuclear cobalt complex 2^H first described by Beattie et al. is no exception to this trend and the ease of preparation of this complex has facilitated a wide range of experimental investigations.^{29,38,54,44,47–53} Here, we define alternate conditions for the combination of these reagents (Co(OAc)₂, NaOAc, py, and H₂O₂) with a focus on kinetic, rather than thermodynamic control. We also define conditions for the quantitative conversion of the resultant dinuclear Co₂(μ–OH)₂ kinetic complexes to their tetranuclear Co₄(μ₃-O)₄ thermodynamic analogues. Low pH and presence of water were critical to these efforts.

The instability of 1^R following deprotonation prompted us to examine reports of bis(μ-oxo)dicobalt(III) systems. As opposed to the abundance of bis(μ-hydroxo)dicobalt(III) examples,^19–28 relatively few bis(μ-oxo)dicobalt(III) systems have been reported. With only one exception,^27,55 none of these complexes contain Co^III in octahedral or tetragonal coordination geometries, and none are stable over long periods of time at room temperature.^56–63 Two groups have separately reported the preparation of [(TPA)₂Co^III₂(μ-O)₂][ClO₄]₂ (4), which is the only example of a Co₂(μ-O)₂ core within an octahedral coordination geometry.^27,55 Kojima et al. provide only UV–vis absorption spectra in support of the formation of 4, suggesting that it is stable at room temperature.²⁷ In contrast, Wang et al. provide a mass spectrum, and UV–vis, ¹H NMR, and X-ray absorption spectroscopic evidence for the preparation of 4, reporting that it is only stable below −40 °C (decaying with t_1/2 ~ 4 h).⁵⁵ However, the ¹H NMR spectrum reveals significant impurities and is truncated. Additionally, the mass spectrum contains two unexpected inner sphere ligands (trifluoroacetate and acetonitrile) for which no explanation is provided. The only other report of an octahedral Co^III-oxo complex (in this case, a mononuclear Co^III with a terminal oxo) was prepared through cryogenic gas-phase irradiation of a molecular precursor.⁶⁴ The rarity of these types of complexes is consistent with their anticipated instability.⁴²

The intramolecular H-bond interactions present in 1^R allow us to stabilize “oxo-like” hydroxo dicobalt complexes. Examining the reactions of 1^R with base, we show that increasing the acidity of the μ–OH group through trans-electronic effects increases the rate constant for deprotonation by 6000-fold. Equation 2 describes our model for the kinetics of 1^R to 2^R conversion. The linear dependence of k₁ on [MTBD] along with the fact that intermediates are not observed by NMR or UV–vis absorption spectroscopies, support that deprotonation is rate-determining in these reactions. However, as shown in Figures S14–S16, the data do not fit to a monoexponetial rise. Fitting to a biexponential equation (eq 1) was required to capture the early time points and to accurately model the shape of the curves. Analyses of both the residuals plots and the fits themselves reveal how poorly a monoexponential equation serves in modeling the data; the fit does not overlap with the data at the critical turning point from the rise to the plateau.

$$2{\mathbf{1}}^{\mathbf{R}}\stackrel{\begin{array}{c}{k}_{\mathrm{deprot}}\\ \mathit{excess\; base}\end{array}}{\to }2\left[{\mathbf{X}}^{\mathbf{R}}\right]\stackrel{k_2}{\to }{\mathbf{2}}^{\mathbf{R}}$$

Studies aimed at assigning k₂ to a mechanistic step are ongoing. We explored the possibility of a dimerization process as was postulated for the related [Co₂ (μ–OH)₂(OAc)₃(bpy)₂][ClO₄] system published by Christou et al.²⁰ However, our crossover experiments suggest that a more complex mechanism is at play. Another possibility is that k₂ represents a second deprotonation event. The observation that addition of <2 eq of base produce only mixtures of 1^R and 2^R supports that an inverted pK_a may be operative such that pK_a2 < pK_a1. This situation would explain the lack of MTBD concentration dependence of k₂. Ultimately, there are many possibilities for the mechanistic assignment of k₂ among the multiple structural rearrangements that could precede formation of 2^R.

In contrast to the ambiguity of k₂, our assignment of k₁ to rate-limiting deprotonation of 1^R is supported by the linear dependence on [MTBD] which provides k_deprot. The LFERs shown in Figures 2 and S18 illustrate the striking impact that trans-electronic effects can have on metal-hydroxo moieties stabilized by H-bond interactions. Here, the magnitude of the slope demonstrates the extent to which the measured parameter is affected by the electronic structure of the derivatives. Figures 2 and S18 have slopes of 0.73 and 4.1, respectively. While there are no systems reported to adequately compare our chemical shift LFER (examples are primarily from organic molecules), a system has been reported that modulates H-bond interactions through substituent effects para- to the H-bond donor. Observing the rate of reaction of C–H bond cleavage, the authors report a slope of 1.57.⁶⁵

In our system, the H-bond interaction is modulated by trans-electronic effects. This type of interaction is similar to that present in the Cytochrome P450, where thiolate ligation at the heme affects the pK_a of the trans-oxo ligand.⁶⁶ By leveraging H-bond interactions to stabilize increasingly acidic μ–OH ligands, we have been able to demonstrate a 6000-fold enhancement on the conversion of kinetically trapped dinuclear species to their tetranuclear thermodynamic counterparts. Molecular orbital theory suggests that our ability to capture a Co₂μ-O)₂ intermediate via deprotonation of the Co^III₂(μ–OH)₂ core will not be possible without concomitant oxidation. Our work is set for future studies.