Ligand Substitution and Electronic Structure Studies of Bis(phosphine)Cobalt Cyclooctadiene Precatalysts for Alkene Hydrogenation

Abstract

Diene self-exchange reactions of the 17-electron, formally cobalt(0) cyclooctadienyl precatalyst, (R,R)-(^iPrDuPhos)Co(COD) (P₂CoCOD, (R,R)-^iPrDuPhos = 1,2-bis((2R,5R)-2,5-diisopropylphospholano)benzene, COD = 1,5-cyclooctadiene) were studied using natural abundance and deuterated 1,5-cyclooctadiene. Exchange of free and coordinated diene was observed at ambient temperature in benzene-d₆ solution and kinetic studies support a dissociative process. Both neutral P₂CoCOD and the 16-electron, cationic cobalt(I) complex, [(R,R)-(^iPrDuPhos)Co(COD)][BAr^F₄] (BAr^F₄ = B[(3,5-(CF₃)₂)C₆H₃]₄) underwent instantaneous displacement of the 1,5-cyclooctadiene ligand by carbon monoxide and generated the corresponding carbonyl derivatives. The solid-state parameters, DFT-computed Mulliken spin density and analysis of molecular orbitals suggest an alternative description of P₂CoCOD as low-spin cobalt(II) with the 1,5-cyclooctadiene acting as a LX₂-type ligand. This view of the electronic structure provides insight into the nature of the ligand substitution processes and the remarkable stability of the neutral cobalt complexes toward protic solvents observed during catalytic alkene hydrogenation.

Graphical Abstract

Introduction

Catalysis with Earth-abundant, first-row transition metals has witnessed a renaissance of interest recently due to potential economic, environmental advantages as well as reactivity distinct from more widely used and studied precious metal catalysts.¹ The asymmetric hydrogenation of carbon–carbon, carbon–oxygen and carbon–nitrogen multiple bonds has received considerable attention. Examples of manganese, iron, cobalt and nickel catalysts with high activity and enantioselectivities have been reported and in some cases offer performance or operating conditions superior to precious metals.^2,3,4 Bis(phosphine)cobalt complexes have emerged as a powerful and versatile class of catalysts for asymmetric hydrogenation and offer remarkable properties such as stability in alcohol solvents and broad functional group tolerance.⁵ Compounds of this type have also been reported to be versatile catalysts for other catalytic transformations such as alkene hydroformylation⁶ and hydroacylation,⁷ alkene and alkyne hydrovinylation,⁸ [4+2] and [2+2] cycloadditions⁹ and other C–C bond-forming reactions (Scheme 1).¹⁰

Scheme 1.

Examples of catalytic chemistry promoted by bis(phosphine) cobalt complexes.

High-throughput experimentation (HTE) has been an enabling technology for the rapid identification of in situ catalyst generation conditions and optimal ligands for asymmetric alkene hydrogenation. Two carbon-bridged, C₂-symmetric chiral bis(phosphine)s in combination with cobalt(II) halides and alkyl lithium activating reagent emerged as effective for the hydrogenation of functionalized alkenes such as methyl 2-acetamidoacrylate.^5a Greatly improved compatibility with essentially all of the 192 chiral bidentate ligands within the library was discovered when active cobalt catalysts were generated from zinc reduction in MeOH.^5c A pilot-scale, 200 gram asymmetric hydrogenation affording the epilepsy medication levetiracetam in 98.2% ee was performed with only 0.08 mol% loading of the cobalt catalyst. Isolation of well-defined Co(I) and Co(0) precatalysts generated from sequential one-electron reduction provided insights into catalyst activation pathways. The asymmetric hydrogenation of α,β-unsaturated carboxylic acids with diverse substitution patterns has also been achieved with the Co(0) precatalysts,^5d wherein good functional group tolerance was observed and deuterium labeling studies supported an unusual homolytic H₂ cleavage pathway with cobalt carboxylates.

To gain additional mechanistic insights, our laboratory has also been exploring the chemistry of well-defined organometallic cobalt precatalysts. Bis(phosphine)cobalt(II) dialkyls have been prepared and structurally characterized as low-spin, planar compounds and applied to the diastereoselective hydrogenation of hydroxyl alkenes.^5b In the absence of substrate, stirring (R,R)-(^iPrDuPhos)Co(CH₂SiMe₃)₂ in methanol resulted in dehydrogenation of the alcohol and isolation of [(R,R)-(^iPrDuPhos)Co]₂(μ-CO)₂.¹¹ Related, formally cobalt(0) 1,5-cyclooctadiene complexes have been prepared either from the hydrogenation of the cobalt(II) dialkyl derivatives in the presence of COD or from reduction of the corresponding cobalt(II) dihalides in the presence of the diene.^5b,c,d These compounds are one electron reduced variants of widely used bis(phosphine) rhodium(I) cations and are effective single component catalysts for the hydrogenation of a range of alkenes with high activity and enantioselectivity. Despite their anticipated reducing nature, the Co(0) precatalysts showed unusual stability to protic medium as well as broad functional group tolerance.^5c,d

Recently, our laboratory reported the synthesis, structural characterization and hydrogenation performance of [(R,R)-(^iPrDuPhos)Co(COD)][BAr^F₄] ([P₂CoCOD]⁺), a pioneering example of a long-sought-after cobalt analog of the rhodium cations.¹² An idealized square-planar geometry at cobalt was identified in the solid-state structure of [P₂CoCOD]⁺,¹² while D_2d-distorted tetrahedral geometry was observed for all neutral P₂CoCOD compounds, a result of the formally Co(0), d⁹ configuration.^5b,c,d The one-electron oxidized, cationic [P₂CoCOD]⁺exhibited much higher substitutional lability of 1,5-cyclooctadiene as compared to the neutral P₂CoCOD complex, where displacement of the diene by enamides and arenes for [P₂CoCOD]⁺ was instantaneous and substitution by tetrahydrofuran-d₈ also occurred over time.¹² “While the neutral cobalt(0) precatalysts were highly active in protic solvents such as MeOH and i-PrOH for the asymmetric hydrogenation of enamides and unsaturated carboxylic acids, optimal performance for the cobalt(I) precatalysts was observed in aprotic solvents such as tetrahydrofuran for the asymmetric hydrogenation of methyl 2-acetamidoacrylate and (Z)-methyl-2-acetamidocinnamate, highlighting the distinct reactivity/tolerance with protons for the neutral and cationic cobalt intermediates during catalysis. Isolation of otherwise identical complexes separated by one-electron highlights the unique electronic properties offered by first row transition metals. Catalytic transformations including hydrogenation in potentially two different redox cycles, initiated from Co(0) and Co(I) precatalysts, are highly desired features unique to first-row metals and likely offer new reactivity and selectivity. Here we describe a study in olefin substitution chemistry and electronic structures for both the neutral and cationic bis(phosphine) cobalt 1,5-cyclooctadiene complexes, P₂CoCOD and [P₂CoCOD]⁺, to better understand the basis of their distinct stability and reactivity profiles.

Results and Discussion

Kinetic Studies of Cyclooctadiene Substitution Reactions.

The rate and mechanism of substitution of the cyclooctadiene ligands in both cobalt(0) and cobalt(I) was of fundamental interest and may provide insights into precatalyst activation and substrate coordination events. Our studies commenced with a self-exchange experiment between free and coordinated 1,5-cyclooctadiene with neutral P₂CoCOD. Because of the paramagnetically shifted ¹H NMR resonances of the formally cobalt(0) compound, 1,5-cyclooctadiene-d₁₂ would provide additional spectroscopic handles for these experiments. The iron-catalyzed [4+4] cycloaddition of dienes, pioneered by tom Dieck^13a and later optimized by Ritter^13b and expanded by our laboratory^13c was well suited for the preparation of cyclooctadiene-d₁₂ from commercially available 1,3-butadiene-d₆. The targeted cobalt(0) diene complex, (R,R)-(^iPrDuPhos)Co(COD-d₁₂) (P₂CoCOD-d₁₂), was successfully synthesized following the previously reported procedure^5c in 92% isolated yield. The natural abundance compound, P₂CoCOD exhibits paramagnetically shifted resonances in the benzene-d₆ ¹H NMR spectrum at ambient temperature that range between −80.23 and 34.35 ppm. Comparing these data to the analogous spectrum of P₂CoCOD-d₁₂ identified the six resonances corresponding to the coordinated cyclooctadiene (see Supporting Information, Figure S2).

With the P₂CoCOD-d₁₂ in hand, exchange with natural abundance 1,5-cyclooctadiene was explored (Figure 2A). Addition of 20 equivalents of free 1,5-cyclooctadiene to a benzene-d₆ solution of P₂CoCOD-d₁₂ and monitoring the progress of the reaction by ¹H NMR spectroscopy established appearance of coordinated COD resonances over the course of 8 hours at ambient temperature (Figure S7). Notably, the isotopic composition of the cyclooctadiene ligand also impacted the ¹H NMR chemical shifts of the ^iPrDuPhos protons (Figure 2B). The resolution of the ¹H NMR resonances of the ^iPrDuPhos ligand in P₂CoCOD and P₂CoCOD-d₁₂ provided a convenient measure for determining the degree of conversion for diene exchange. The reaction reached 10 and 50% conversions in approximately 4 and 27 hours, respectively and equilibrium was established after 96 hours.

Figure 2.

A) Diene self-exchange experiment between P₂CoCOD-d₁₂ and natural abundance 1,5-cyclooctadiene. B) Stack plot of ¹H NMR spectra of (R,R)-^iPrDuPhos ligand resonances (partial) showing disappearance of P₂CoCOD-d₁₂ signals (square) and appearance of P₂CoCOD signals (triangle) as an indicator of conversion.

To determine the reaction order in both cobalt and 1,5-cyclooctadiene, initial rate (<10% conv.) measurements were performed. Three separate reactions: (A) P₂CoCOD-d₁₂ (32 mM) and 1,5-cyclooctadiene (640 mM); (B) P₂CoCOD-d₁₂ (32 mM) and 1,5-cyclooctadiene (1280 mM) and (C) P₂CoCOD-d₁₂ (64 mM) and 1,5-cyclooctadiene (640 mM) in benzene-d₆ solutions were monitored by ¹H NMR spectroscopy and the formation of [P₂CoCOD] product as a function of time is shown in Figure 3. Doubling the concentration of 1,5-cycloctadiene did not result in a discernable acceleration in initial rate (A vs B) while an approximate two-fold enhancement in rate was observed when the initial P₂CoCOD-d₁₂ concentration was doubled (A vs C), indicating a 0^th order in [COD-d₁₂] and 1^st order in [P₂CoCOD-d₁₂], respectively (Figure 3B). These findings support a dissociative substitution pathway where ligand coordination is fast compared to rate-determining cobalt-carbon bond breaking. Although 17-electron complexes have been reported to undergo associative substitutions through 19-electron intermediates,¹⁴ the steric properties of the ancillary ligands are known to dramatically influence the preferred substitution mechanism.¹⁵ The solid-state structure of P₂CoCOD demonstrates that the iso-propyl groups on the rigid DuPhos ligand backbone impose a large steric hindrance above and below the P–Co–P plane, likely inhibiting formation a five-coordinate, 19-electron intermediate with two COD ligands coordinated in (η², η²) and η² arrangement. Substitution of P₂CoCOD with 1,5-cyclooctadiene-d₁₂ was also monitored and exhibited the same kinetic profile, suggesting there was no secondary kinetic isotope effect. An overall first order rate constant k = 6.6(±0.5) ×10⁻⁶ s⁻¹ was derived from the linear fits of the initial rates, supporting slow diene substitution. The entire reaction time courses for trials A-C were followed over 103 hours until [P₂CoCOD] plateaued and the equilibria were established (Figure 3C). The increase in [P₂CoCOD] slowed significantly at higher conversions as the reverse reaction rate of the equilibrium became significant. Plotting conversions versus time for trials A-C showed similar traces but slightly different conversions (positions of final equilibrium) relevant to the initial [P₂CoCOD-d₁₂]:[COD] ratio. (Figure 3D).

Figure 3.

A) Conditions and initial rates of ligand substitution. B) Initial rate experiments, P₂CoCOD concentrations versus time followed until 10% conversion. C) Product concentrations versus time followed over entire reaction. D) Conversion to product versus time followed over entire reaction. Trials A, B and C reached equilibrium at 91% conv. (initial concentration [P₂CoCOD-d₁₂]:[COD] = 1:20), 94% conv. (initial concentration [P₂CoCOD-d₁₂]:[COD] = 1:40) and 88% conv. (initial concentration [P₂CoCOD-d₁₂]:[COD] = 1:10) respectively.

The observation of a dissociative substitution mechanism for the 17-electron, neutral cobalt(0) complex prompted related studies on cyclooctadiene exchange with the 16-electron, cationic cobalt(I) complex, [P₂CoCOD]⁺. The high propensity of [P₂CoCOD]⁺ to form the more stable 18-electron, η⁶-arene complexes [P₂Co(arene)]⁺ precluded the use of benzene-d₆ as solvent.¹² While [P₂CoCOD]⁺ was also unstable upon prolonged standing at ambient temperature in THF-d₈ due to solvent substitution and formation of high-spin, paramagnetic cobalt species,¹² adding 20 equivalents of 1,5-cyclooctadiene-d₁₂ to a THF-d₈ solution of [P₂CoCOD]⁺ and monitoring the reaction by ¹H NMR spectroscopy after 10 minutes demonstrated that the reaction had reached equilibrium (See Supporting Information, Figure S8). No additional spectroscopic changes were observed until competing THF-d₈ substitution occurred. Although no quantitative kinetic data could be obtained due to the fast rate of substitution, these observations clearly establish the substantial increase in the substitutional lability of [P₂CoCOD]⁺ as compared to its one-electron reduced analogue, P₂CoCOD.

Cyclooctadiene Substitution Reactions with Carbon Monoxide.

Substitution of the diene with stronger field ligands such as CO was also explored. Substitution of the 1,5-cyclooctadiene ligand by carbon monoxide with P₂CoCOD was previously studied and afforded the coordinatively saturated, dimeric cobalt(0) dicarbonyl complex with bridging, terminal CO ligands and a cobalt–cobalt bond, [(R,R)-^iPrDuPhosCo(η¹-CO)(μ²-CO)]₂ (Scheme 2A).¹¹ The carbonylation of [P₂CoCOD]⁺ was studied for comparison. Treatment of a diethyl ether solution of [P₂CoCOD]⁺ with 1 atm CO resulted in an instant color change from blue to orange. The product was isolated in 86% yield and X-ray diffraction established formation of the 18-electron cobalt carbonyl complex, [(R,R)-^iPrDuPhosCo(η¹-CO)₃][BAr^F₄]. (Scheme 2B, Figure 4) An idealized trigonal bipyramidal geometry at cobalt was observed. The ¹H, ³¹P{¹H} and ¹³C{¹H} NMR spectra (THF-d₈, 298K) of [(R,R)-(^iPrDuPhos)Co(η¹-CO)₃][BAr^F₄] suggest chemically equivalent phosphine atoms (singlet, 98.15 ppm) and carbon atoms of CO ligands (199.5 ppm) at ambient temperature (Supporting Information, Figures S3, S4, S5). The ¹H and ¹³C NMR signals are consistent with an overall C₂-symmetric environment for the ligand, despite the overall C₁-symmetry of the molecule. This observation is consistent with previously reported for η⁶-Ph(BPh₃) and η², κ¹-methyl 2-acetamidoacrylate coordination.¹² Solution infrared spectra (Et₂O, 298 K) of [(R,R)-^iPrDuPhosCo(η¹-CO)₃][BAr^F₄] showed terminal CO stretching frequencies at 2079, 2035 and 2011 cm⁻¹ which are higher than those of the neutral [(R,R)-^iPrDuPhosCo(η¹-CO)(μ²-CO)]₂ (Scheme 2), consistent with weaker back-bonding to CO from Co(I)⁺ than the neutral, formal Co(0) center. The observation that both the neutral P₂CoCOD and cationic [P₂CoCOD]⁺ underwent fast and complete displacement of COD ligand by CO to afford coordinatively saturated carbonyl complexes highlight their substitutional lability to the stronger field ligand.

Scheme 2.

A) Rapid substitution of P₂CoCOD by CO and formation of bis(phosphine)cobalt dicarbonyl dimer (ref. 11). B) Fast substitution of [P₂CoCOD]⁺ by CO and formation of cationic bis(phosphine)cobalt tricarbonyl.

Figure 4.

Solid-state structure of [(R,R)-^iPrDuPhosCo(η¹-CO)₃][BAr^F₄] at 30% probability ellipsoids with H atoms and BAr^F₄₋ anion omitted for clarity.

Electronic Structure Studies of [P₂CoCOD]⁺ and P₂CoCOD.

The electronic structures of the neutral and cationic cobalt complexes were investigated by full molecule density functional theory (DFT) using the B3LYP functional. The qualitative d-orbital splitting diagrams for [P₂CoCOD]⁺ are shown in Figure 5A. The HOMO is principally d_z² in character as expected for a square-planar d⁸ metal complex with σ-donating phosphines and π-accepting alkene (diene) ligands.¹⁶ The qualitative d-orbital splitting diagram and the Mulliken spin density plot for P₂CoCOD are likewise presented in Figures 5B and 6. The spin density distribution on cobalt approximates the shape of the d_z² orbital, suggesting d_z² as the singly occupied molecular orbital (SOMO) of the S = 1/2 complex. This contrasts expectations for a cobalt(0), d⁹ complex where Jahn-Teller distortion resulting in a D_2d geometry would be expected and the d_xy, d_xz and d_yz orbitals would have higher energies than d_x²-_y², d_z² orbitals and d_z² would be filled.

Figure 5.

Qualitative molecular orbital diagrams representing the corresponding orbitals possessing significant d-orbital character of A) [P₂CoCOD]⁺; B) P₂CoCOD computed by DFT at B3LYP/def2-SVP/def2-TZVP level of theory. Calculations were initiated from optimized solid-state structures.

Figure 6.

Spin-density plot for P₂CoCOD obtained from the Mulliken population analysis (red, positive spin density; yellow, negative spin density). DFT calculation was performed at the B3LYP/def2-SVP/def2-TZVP level of theory. Calculations were initiated from optimized solid-state structure.

A closer examination of the metrical parameters from the solid-state structure of P₂CoCOD^5c revealed that one alkene double bond of COD was longer than the other (1.417(3) Å vs. 1.396(3) Å), while the average Co–C bond length was shorter than the other (2.044(3) Å vs. 2.128(2) Å), indicating stronger back-bonding interaction and metallacyclopropane character¹⁷ of one “Co–C=C” bonds (Figure 7A). The elongated C=C bond length is comparable to that (1.423 Å) of a structurally characterized cobalt ethylene complex, (PMe)₃Co(Ph)(η²-C₂H₄).¹⁸ By comparison, both alkene double bond distances of COD of the cationic [P₂CoCOD]⁺ are 1.39(1) Å (Figure 7A), suggesting weaker back-bonding from Co(I)⁺ and a lack of metallacyclopropane character. The solid-state structure of P₂CoCOD showed distortion from an idealized D_2d geometry towards a more square-pyramidal configuration, where the phosphines, cobalt and one C=C bond of the cyclooctadiene define the basal plane while the other alkene is apical. A similar distortion towards square-pyramidal geometry and metallacyclopropane character have also been identified in the solid-state structures of (dppe)Co(COD)^5b (dppe = 1,2-Bis(diphenylphosphino)ethane) and (R,R)-(BenzP*)Co(COD)^5d ((R,R)-BenzP* = (R,R)-1,2-Bis(t-butylmethylphosphino)benzene) (Figure 7A). As such, these formally P₂Co(0)COD complexes are best described as five-coordinate, d⁷, cobalt(II) compounds with the cyclooctadiene being viewed as an LX₂-type ligand. The four strong-field donors consisting of phosphines and alkyls occupy the xy-plane and largely determine the relative d-orbital energies similar to those in a square-planar field with perturbation from the weaker field alkene donor coordinating through the z-axis and slightly raising the energies of the d_z², d_xz and d_yz parentage orbitals. Accordingly, the spin density analysis and qualitative d-orbital splitting diagram suggest that the SOMO of the low spin, d⁷ cobalt complex has primarily d_z² character with contributions from the cyclooctadiene carbon p orbitals, while the d_xz, d_yz and d_xy orbitals are filled and d_x²-_y² orbital is empty (Figures 6, 5B).

Figure 7.

A) Summary of bond distances of P₂CoCOD complexes supporting the metallacyclopropane assignments. Bond distances of [P₂CoCOD]⁺ are reported for comparison. B) Truncated solid-state structure of P₂CoCOD at 30% probability ellipsoids with H atoms omitted for clarity.

In addition, the X-band EPR spectra of P₂CoCOD, (dppe)CoCOD, (R,R)-(BenzP*)CoCOD and (R,R)-(^PhBPE)CoCOD ((R,R)-^PhBPE = 1,2-bis[(2R,5R)-2,5-diphenylphospholano]-ethane) all exhibit characteristic hyperfine coupling of one g tensor to the ⁵⁹Co nucleus (I = 7/2) but no coupling to the ³¹P nuclei,^5b–d indicating a lack of d_x²-_y² character(phosphines are on the xy plane) of the SOMO and further supporting a d⁷ instead of d⁹ electronic configuration. Nevertheless, the relatively small g anisotropy for all P₂CoCOD complexes^5b–d compared to the large g anisotropy characteristic of square-planar, low spin, d⁷ cobalt(II) complexes, including (dppe)Co(CH₂SiMe₃)₂ ^5b and (^iPrmPNP)Co(X) (X = CH₃, Cl and H, ^iPrmPNP is an L₂X-type ligand ),¹⁹ also suggest that the P₂CoCOD compounds are electronically differentiated from low spin, planar L₂CoX₂ complexes. For a related structurally characterized P₂Co(0)(η⁶-arene) precatalyst, (R,R)-(^PhBPE)Co(η⁶-C₆H₆),^5c DFT-computed Mulliken spin density distribution on cobalt approximate the shape of the d_x²-_y² orbital (Supporting Information, Figure S10), supporting a Co(0), d⁹ assignment for the η⁶-arene complex.

Taken together, the slow substitution rate of the 17-electron P₂CoCOD may be attributed to the strong binding of one cyclooctadiene C=C bond to the cobalt through strong π-back donation, resulting in a rate-determining alkene dissociation step. In addition, the shift of electron density from cobalt to the alkene fragment likely makes the metal less reducing and consequently less reactive towards protic solvents such as methanol. A stepwise substitution mechanism accounting for the overall dissociative process established from kinetic measurements is proposed in Scheme 3, where the transition states have primarily cobalt–carbon bond-breaking character. Dissociation of the apical alkene arm provided space for an incoming cyclooctadiene molecule affording a 17-electron bis(η²-cyclooctadiene) intermediate, while formation of a 15-electron mono(η²-cyclooctadiene) intermediate is less likely. Subsequent tautomerization between the apical and equatorial COD on the metallacyclopropane component and alkene dissociation generated the product. In comparison, oxidation by one electron to furnish the corresponding cation, [P₂CoCOD]⁺, resulted in markedly higher substitutional lability and faster ligand self-exchange rate likely originating from the electrophilicity of the cobalt(I) center and a stabilization effect from coordination of another alkene. An associative substitution for [P₂CoCOD]⁺ is likely operative and occurs through an 18-electron intermediate. The planar geometry observed with [P₂CoCOD]⁺ may also provide greater access of an incoming ligand from the apical direction, resulting in an increased rate of substitution.

Scheme 3.

Proposed mechanism of ligand self-exchange reaction of P₂CoCOD-d₁₂ and COD.

Conclusions

A dissociative substitution mechanism for the 17-electron, neutral P₂CoCOD complex was established by kinetic studies on the self-exchange reaction between 1,5-cyclooctadiene and 1,5-cyclooctadiene-d₁₂. Rapid displacement of 1,5-cyclooctadiene by carbon monoxide and formation of coordinatively saturated bis(phosphine)cobalt carbonyl complexes was observed for both P₂CoCOD and the 16-electron cobalt cation, [P₂CoCOD]⁺. Electronic structure studies and metrical parameters from the solid-state structures of P₂CoCOD complexes support a cobalt(II)–metallacyclopropane assignment. The strengthened interaction within the metallacyclopropane and a more electropositive cobalt(II) center provide a rationale for the unusual protic stability of these formally cobalt(0) precatalysts.

Figure 1.

Examples of previously reported well-defined bis(phosphine)cobalt diene compounds.