Tuning Cobalt(II) Phosphine Complexes to be Axially Ambivalent.

We report the isolation and characterization of a series of three cobalt(II) bis(phosphine) complexes with varying numbers of coordinated solvent ligands in the axial position. X-ray quality crystals of [Co(dppv)2][BF4]2 (1), [Co(dppv)2(NCCH3)][BPh4]2 (2), and [Co(dppv)2(NCCH3)2][BF4]2 (3) (dppv = cis-1,2-bis(diphenylphosphino)ethylene) were grown under slightly different conditions, and their structures were compared. This analysis revealed multiple crystallization motifs for divalent cobalt(II) complexes with the same set of phosphine ligands. Notably, the 4-coordinate complex 1 is a rare example of a square-planar cobalt(II) complex, the first crystallographically characterized square-planar Co(II) complex containing only neutral, bidentate ligands. Characterization of the different axial geometries via EPR and UV-visible spectroscopies showed that there is a very shallow energy landscape for axial ligation. Ligand field angular overlap model calculations support this conclusion, and we provide a strategy for tuning other ligands to be axially labile on a phosphine scaffold. This methodology is proposed to be used for designing cobalt phosphine catalysts for a variety of oxidation and reduction reactions.

Introduction

Cobalt complexes have been studied extensively as relatively earth-abundant (∼25,000 times more abundant than Rh or Ir)[1] redox catalysts for electrochemical applications.[2−5] In the case of multielectron activation of small molecules, tuning substrate and product binding to the catalytic center is critical: substrate binding that is too weak can result in large electrocatalytic overpotentials, while product binding that is too strong negatively affects the overall reaction kinetics.[6] Because of this, complexes which can reversibly bind ancillary ligands are of particular interest. The square-planar Vaska’s complex, IrCl(CO)[P(C6H5)3]2, for example, is well-known to reversibly bind substrates such as dioxygen via oxidative addition.[7] However, the scarcity of third-row transition metals complicates the use of iridium for most large-scale applications.

Cobalt complexes featuring phosphine ligands (e.g., dppv, cis-1,2-bis(diphenylphosphino)ethylene) have also been reported to form 1:1 dioxygen adducts under a variety of conditions, albeit irreversibly.[8] These adducts are best described as side-bound cobalt(III)-peroxo species. For example, the brown, 6-coordinate complex [Co(dppv)2O2][BF4] can be prepared by dissolving [Co(dppv)2][BF4] in CH2Cl2 in the presence of air, by adding dppv to an oxygenated solution of [Co(H2O)6][BF4]2 in acetone, or by bubbling air through a refluxing methanolic solution of [Co(dppv)2][BF4]2. Interestingly, similar reactivity is not observed in complexes of the analogous bidentate ligand dppe (1,2-bis(diphenylphosphino)ethane).[8] It has been proposed that the more rigid dppv backbone allows for less steric hindrance of the axial position, though there is less structural information about cobalt dppv complexes in the literature. Although there are nearly 300 structures of M(dppv) complexes in the Cambridge Structural Database, only 59 are M(dppv)2 structures and only 27 are of M = groups 9, 10, and 11.[9] None of the cobalt structures are 4-coordinate, even though the d8 Ni(II) complex is square-planar.[10]

In our recent work on d6–d10 cobalt complexes of dppv, we found that the d7 CoII(dppv)22+ core crystallizes with one axial ligand (acetonitrile), generating a pseudo-square pyramidal geometry which we previously reported (2).[11] Herein, we further report that the core also crystallizes with 0 or 2 axial ligands (1 and 3), and we provide the structural characterization of each of these complexes. The former is a structure of 4-coordinate divalent Co(dppv)22+ in a rigidly planar geometry. This is a rare (if not unprecedented) example of 4-coordinate Co(II) with bidentate redox-innocent ligands. The ability of Co(dppv)22+ (1) to crystallize with or without axial ligands suggests that this cobalt cation is energetically precarious. This behavior is reminiscent of the work done in 1964 by Langford and Gray, in which anionic square-planar cobalt complexes with redox-active dithiolate ligands were interconvertible between 4-, 5-, and 6-coordinate environments upon addition of certain ligand(s).[12] We have analyzed the electronic structure of the multiple coordination environments using electron paramagnetic resonance (EPR) spectroscopy, UV–visible spectroscopy, and electrochemical measurements. To explore the underlying structural elements that give rise to the unusual axial-ligation ambivalence of this complex, we have developed a simple ligand-field model based on angular overlap calculations. This model maps out energy landscapes for 6- versus 5- versus 4-coordinate ML4 cores as a function of readily tunable parameters, such as σ and π ligand donor strength and L–M–L “bite angles,” providing a qualitative predictive tool for designing specific metal–ligand combinations engineered for the reversible binding of specific substrates.

Experimental Section

Materials and Methods

All reagents were purchased from Sigma-Aldrich unless otherwise stated.

All compounds were synthesized air-free using Schlenk techniques under argon. All solvents were dried using molecular sieves and thoroughly degassed prior to use.

Synthesis of 4-Coordinate [Co(dppv)2][BF4]2 (1)

1 was prepared by dissolving cobalt(II) tetrafluoroborate hexahydrate (0.100 g, 0.294 mmol) in a minimal amount of dry, degassed acetone in a Schlenk flask. Two equivalents of the ligand, cis-1,2-bis(diphenylphosphino)ethylene, were dissolved in a minimal amount of acetone in a separate Schlenk flask. The ligand solution was then cannulated into the cobalt solution, and the overall solution began turning yellow with a precipitate of the same color. The flask was cooled in a refrigerator for at least 1 h to maximize the amount of precipitation. The solid was filtered-off using a Schlenk frit and washed/dried several times with dry, degassed diethyl ether via cannulation, resulting in a 58% yield of a yellow solid. Complex 1 was crystallized by vapor diffusion of ether into a concentrated solution of [Co(dppv)2][BF4]2 in acetone. The solid-state UV–vis reflectance spectrum of 1 has peaks at 900 and 1200 nm (ε900 = 2ε1200, Supporting Information).

Synthesis of 5-Coordinate [Co(dppv)2(NCCH3)2][BPh4]2 (2)

2 was prepared by dissolving compound 1 (0.100 g, 97.5 μmol) in a minimal amount of dry, degassed acetonitrile in a Schlenk flask (orange-colored solution). In a separate Schlenk flask, an excess (5–10 fold) of sodium tetraphenylborate was dissolved in a minimal amount of dry, degassed acetonitrile. With positive argon pressure, the cobalt solution was then cannulated into the sodium tetraphenylborate solution, and immediately a glittery orange precipitate was observed. The flask was stored in a refrigerator for at least an hour to maximize precipitation. Subsequently, the orange solid was filtered off using a Schlenk frit and washed/dried several times with dry, degassed diethyl ether via cannulation. A yield of 62% was obtained. The solid-state UV–vis reflectance spectrum of 2 has peaks at 700, 1200, and 1450 nm (ε1200 = 2ε700 = 0.9ε1450, Supporting Information).

Synthesis of 6-Coordinate [Co(dppv)2(NCCH3)2][BF4]2 (3)

3 was prepared by dissolving compound (1) (0.100 g, 97.5 μmol) in a minimal amount of dry, degassed acetonitrile in a Schlenk flask (orange-colored solution). The flask was capped and was stirred at room temperature for 5 min, during which the solution color changed from pale pink to dark orange. Over the next minute, 100 mL of ether was added to the orange solution while stirring vigorously, during which an orange precipitate formed. This solid was filtered in air and rinsed with minimal ether, dried under high vac, and was isolated in 65% yield. The solid-state UV–vis reflectance spectrum of 3 has peaks at 700, 1150, and 1400 nm (ε1150 = 0.8ε1400 = 0.9ε700, Supporting Information).

Recrystallization of 5- and 6-Coordinate, [CoII(dppv)2(MeCN)y][BX4]2, Where X is Ph or F and y is 1 or 2, Respectively

Compounds 2 and 3 were both recrystallized via vapor diffusion of diethyl ether into saturated solutions of 2 or 3 in acetonitrile.

UV/Visible/NIR Spectroscopy

Solution spectra were collected on a Cary 14 spectrometer in the dual-beam mode. Specular reflectance spectra were collected on a Cary 5000 spectrometer with an integrating sphere.

Single-Crystal X-ray Diffraction (1, 3)

A suitable crystal was mounted on a polyimide MiTeGen loop with STP Oil Treatment and placed in the 100 K nitrogen stream of an Oxford Cryosystems 700 Series cryostream cooler. X-ray data for both compounds 1 and 3 were collected with a Bruker AXS D8 KAPPA APEX II diffractometer (Siemens KFF Mo 2 K-90 fine-focus sealed tube with Mo Ka = 0.71073 Å operated at 50 kV and 30 mA; TRIUMPH graphite monochromator; CCD area detector with a resolution of 8.33 pixels/mm) using a combination of ω- and φ-scans. All diffractometer manipulations, including data collection, integration, and scaling, were carried out using Bruker APEX3 software.[13] An absorption correction was applied using SADABS-2016/2.[14] The space group was determined, and the structure was solved by intrinsic phasing using XT-2014/5[15] (1 and 3 are both in the monoclinic space group P21/n (#14) with the cation on a center of symmetry). Refinement was full-matrix least-squares on F2 using XL-2018/3.[16] All non-hydrogen atoms were refined using anisotropic displacement parameters. Hydrogen atoms were placed in idealized positions and the coordinates refined. The isotropic displacement parameters of all hydrogen atoms were fixed at either 1.2 or 1.5 (for methyl) times the Ueq value of the bonded atom.

EPR Spectroscopy

X-band EPR spectra were measured using a Bruker EMX spectrometer equipped with a cavity operating in the standard perpendicular mode, TE102. Spectra at 77 K were collected using a finger Dewar filled with liquid nitrogen. Spectra at other temperatures were measured using an Oxford helium cryostat (ESR900). Due to the air-sensitive nature of these compounds in solution, the samples were prepared air-free under argon on a Schlenk line and immediately transferred into a custom air-free quartz Schlenk EPR tube. Simulations of the EPR spectra were performed using the EasySpin 5.2.28 toolbox[17] within the Matlab R2018a software suite (The Mathworks Inc., Natick, MA).

Certain ligand-field excited state energies can be estimated based on the shift of the observed g-value from the free electron g-value (ge = 2.0023) using first-order perturbation theory formulas developed by Pryce.[18] Specifically, for a 3dz2-based ground statewhere ζ is the single electron spin–orbit coupling constant for cobalt (515 cm–1) and Δ is the energy difference between the z2-based ground state and a state described by the unpaired electron in the xz/yz orbitals.

For a 3d-based ground statewhere Δ is the energy difference between the x2–y2-based ground state and a state described by the unpaired electron in an xy-based orbital; Δ is the energy difference between the x2–y2-based ground state and a state described by the unpaired electron in the xz/yz orbitals.

Results

Synthesis and Structural Characterization

Analysis of X-ray quality crystals of 1 reveals a rigorously planar, 4-coordinate structure. When 1 is dissolved in CH3CN, the slow diffusion of diethyl ether into the solution affords orange crystals, the X-ray diffraction analysis of which show a 6-coordinate complex with two CH3CN ligands coordinated in the axial positions ([Co(dppv)2(NCCH3)2][BF4]2, compound 3). When the same recrystallization method is used (CH3CN/ether) but in the presence of excess NaBPh4, however, crystals of either a 5-coordinate complex with only a single CH3CN ligand coordinated ([Co(dppv)2(NCCH3)][BPh4]2, compound 2) or a similar 6-coordinate complex ([Co(dppv)2(NCCH3)2][BPh4]2, compound 3b) are isolated (Figure ).

Figure 1

Structural representation of 1, 2, and 3 in the solid state with displacement ellipsoids shown at the 50% probability level. Hydrogen atoms and counter anions are omitted for clarity. Dark blue = Co, light blue = N, orange = P, and gray = C.

The crystal structure data from 1, 2, and 3 reveal subtle but distinct differences in bonding of the phosphine ligands among the series in the solid state (Table ). For the two that are symmetric above and below the plane of the phosphine ligands, compounds 1 and 3, the Co(II) sits directly in the center of the plane of the phosphorus atoms. This is apparent from the trans-phosphine bond angles being 180° for those complexes. In the 5-coordinate complex 2, Co(II) is distorted out of the plane of the phosphorus ligands in the direction of the bound CH3CN ligand. There is a slight elongation of the average Co–P bond as the coordination number increases. Additionally, for complex 2, the two chelating phosphine ligands are more asymmetric with respect to each other than in complexes 1 and 3. This is apparent in the differences in the P1/P2 versus P3/P4 bite angles and the differences in the dppv ligand dihedral angles on the same complex.

Table 1

Structural Parameters for Complexes 1, 2, and 3a

	Co–P bond distance (Å)				P–Co–P bite angle (deg)		dppv dihedral angle (deg)		trans-phosphine bond angle (deg)
complex	P1	P2	P3	P4	P1/P2	P3/P4	P1/P2	P3/P4	P1/P3	P2/P4
1	2.282	2.281	2.281	2.282	83.45	83.45	1.31	1.31	180	180
2	2.277	2.291	2.314	2.272	82.65	84.21	4.96	3.96	174	171
3	2.332	2.318	2.318	2.332	82.04	82.04	2.65	2.65	180	180

For 1, P1 and P3 lie on symmetry elements which make the parameters equivalent.

Strikingly, the 4-coordinate [Co(dppv)2]2+ is shown to be rigorously planar by X-ray crystallography. Several square-planar cobalt(II) complexes have been reported, but those contain higher-order hapticity ligands such as pincers (tridentate, neutral, or monoanionic),[19] porphyrins (tetradentate and dianionic),[20] or salens (tetradentate and dianionic).[21,22] There are rare examples of square-planar Co(II) complexes featuring bidentate ligands but only when those ligands are negatively charged (e.g. aminophenol).[23] To the best of our knowledge, this is the first crystallographically characterized square-planar Co(II) complex containing only neutral, bidentate ligands.

UV–Visible Spectroscopy

It was unclear from the solid-state structural data why the addition of Na[BPh4] results in the isolation of either 5- or 6- coordinate cobalt complexes from acetonitrile solutions. While there are no obvious structural interactions between BPh4– and the cobalt complex in the solid state, UV–visible absorbance spectroscopy was used to probe a possible equilibrium in solution.

Indeed, when [TBA][BPh4] (TBA = tetrabutylammonium; BPh4 = tetraphenylborate) is titrated into a solution of [Co(dppv)2(NCCH3)2][BF4]2 in CH3CN, minor changes in the spectra are observed (Figure ), and changes in the spectra versus the concentration of added [TBA][BPh4] can be fit with a single equilibrium binding isotherm (see the Supporting Information). This indicates that there is some equilibrium in solution mediated by [TBA][BPh4], presumably the ligation and dissociation of a second CH3CN ligand. This equilibrium explains why both 5- and 6-coordinate complexes were isolated when recrystallization occurred in the presence of BPh4– (Scheme ). One plausible model is that the Co2+/BPh4– ion pair is not as stable in acetonitrile as the Co2+/BF4– ion pair. At high concentrations of BPh4–, the equilibrium is shifted to the 5-coordinate/BPh4– complex, presumably with binding of the phenyl group.[24] Consequently, one BF4– anion does mildly encroach on the pocket of the cobalt phosphine core in the 4-coordinate complex, indicating stronger ion pairing. This pairing is non-negligible in acetonitrile.[25]

Figure 2

UV–visible absorbance spectra of the titration of [TBA][BPh4] into a solution of [Co(dppv)2(NCCH3)2][BF4]2 in CH3CN (TBA = tetrabutylammonium; BPh4 = tetraphenylborate).

Scheme 1

Ligation Chemistry of [Co(dppv)2]2+ under Various Conditions to form 4-, 5-, and 6-Coordinate Complexes

EPR Characterization

All EPR spectra described below are consistent with mononuclear low-spin (S = 1/2) d7 Co(II)-containing complexes. Depending on the ligand field, the ground electronic state single unpaired electron is found in either the 3dz2 or 3d-based molecular orbital (vide infra).

The continuous-wave (CW) EPR spectrum of solid 1 (Figure A) is characterized by a rhombic g-tensor (g = [2.75, 2.47, 1.978]) and exhibits splitting attributed to a large magnitude 59Co (I = 7/2) hyperfine interaction (HFI) (A59Co = [490, 350, 350] MHz). These magnetic properties are consistent with the unquenched orbital angular momentum in the ground state owing to the high local symmetry of the cobalt spin-center in the square-planar ligand sphere. Further hyperfine splittings from the four ligating 31P nuclei (I = 1/2) are not apparent, though it was convenient to model the spectral lineshape using four 31P hyperfine tensors, A31P = [125, 115, 55] MHz. These magnetic parameters are almost identical to those observed for [Co(dppe)2][ClO4]2,[26] which was surmised to possess a (d)1 ground-state electron configuration.

Figure 3

CW EPR spectra (black traces) of Co(II) complexes with various axial ligands. (A) Crushed powder of [Co(dppv)2][BF4]2 (1), temperature = 77 K; (B) [Co(dppv)2][BF4]2 (1) + dppe, temperature = 70 K; (C) frozen solution of [Co(dppv)2][BF4]2 (1) in acetone, temperature = 10 K; (C′) computed numerical derivative showcasing the 31P hyperfine coupling that is evident along g||. Spectrometer settings: (A) microwave frequency = 9.434 GHz, microwave power = 2 mW; (B) microwave frequency = 9.377 GHz, microwave power = 6.4 mW; (C) microwave frequency = 9.377 GHz, microwave power = 2 mW. Simulations (red traces) achieved with magnetic parameters given in Table and these inhomogeneity parameters: (A) linewidth = 2 mT, g-strain = [0.08, 0.05, 0.06]; (B) linewidth = 31.4 mT, g-strain = [0.30, 0.05, 0.11]; (C) linewidth = 2 mT, g-strain = [0.036, 0.011, 0.00].

Table 2

Magnetic Parameters for Cobalt-Phosphine Complexesa

complex	g	59Co (MHz)	31P (MHz)	refs
1 solid	[2.75, 2.47, 1.978]	[490, 350, 350]	[125, 115, 55]	this study
1 solid + dppe	[2.673, 2.508, 2.005]	[275, 47, 80]	n.d.	this study
1 in acetone	[2.400, 2.278, 1.999]	[99, 52, 220]	[55, 55, 55]	this study
2 solid	[2.290, 2.262, 2.009]	[12, 35, 212]	n.d.	this study
3 solid	[2.163, 2.159, 2.009]	[248, 278, 48]	n.d.	this study
3 solid + DPPE	[2.305, 2.230, 2.007]	[22, 31, 202]	n.d.	this study
3 in acetone	[2.298, 2.222, 2.005]	[31, 20, 206]	n.d.	this study
3 in CH3CN	[2.154, 2.134, 1.999]	[275, 276, 29]	n.d.	this study
[Co(dppe)2][ClO4]2	[2.797, 2.442, 1.968]	[482, 396, 402]	n.a.	(26)
[Co(dppv)2Br]+	[2.309, 2.239, 2.010]	[210, 42, 66]	[50, 50, 50]	(30)
CoSaltBu NO2	[2.56, 2.30, 1.98]	[410, 196, 346]	n.a.	(31)

n.d. Not detected. n.a. Not applicable.

No 59Co HFI was resolved in the spectrum of the solid 4-coordinate complex 1 with added dppe (Figure B). The reported A59Co values in Table were chosen to model the non-Gaussian lineshape of the resonance; these values are likely upper bounds for the metal HFI. Spectra 3A and 3B are of crushed powders and are therefore likely distorted from ideality due to partial ordering and/or dipolar broadening.[27]

Dissolution of the 4-coordinate complex 1 (Figure C) into acetone results in a change in the EPR spectrum indicating coordination of a fifth ligand, likely a molecule of acetone, and analogous to earlier observations of solvent coordination in cobalt-corrinoids.[28] The axial g-tensor (with g⊥ > g|| ≈ ge) and 59Co hyperfine (A|| ≫ A⊥) are indicative of a cobalt-based 3d ground state. Using Pryce’s perturbative formulas,[18] the prediction of g-shift from ligand-field splittings and a spin–orbit coupling constant for Co(II) of ζ = 515 cm–1, the xz/yz singly occupied excited state lies 9200 cm–1 above the ground state. We note that ζ is ∼90% of the free-ion value for Co(II), which is high for very covalent complexes. Values of 80–100% are routinely used for complexes of less covalency (such as square planar complexes), and often ζ is only accurate to within 10%.[29] Hyperfine coupling from the four equatorial 31P (I = 1/2) ligand atoms is evident at the mI −1/2 ↔ +1/2 resonances along g|| (c.a. 330 and 340 mT, Figure C). This was modeled as four identical isotropic couplings of 55 MHz. Note that only along g|| are these splittings resolved. Along the other g-axes, the tabulated 31P HFI merely contributes to the line broadening and should not be considered accurate measures of the coupling along these dimensions.

The CW EPR spectrum of the solid, singly–CH3CN–ligated 5-coordinate complex 2 (Figure A) is axial and reminiscent of many pseudo-C4-symmetric cobalamin and cobalt(II) porphyrin spectra with g⊥ > g|| ≈ ge. The magnitude of the g-shift for g⊥ corresponds to a Δ = 11,300 cm–1. As in the spectrum of the 4-coordinate complex dissolved in acetone, the g|| resonance is split into eight peaks owing to a strong (A|| = 212 MHz) HFI with the 59Co nucleus along this dimension.

Figure 4

CW EPR spectra (black traces) of Co(II) complexes with various axial ligands. (A) Crushed powder of [Co(dppv)2(NCCH3)][BPh4]2, temperature = 4 K; (B) crushed powder of [Co(dppv)2(NCCH3)2][BPh4]2, temperature = 77 K; (C) crushed powder of [Co(dppv)2(NCCH3)2][BPh4]2 + dppe, temperature = 50 K. (D) Frozen solution of [Co(dppv)2(NCCH3)2][BPh4]2 dissolved in acetone, temperature = 77 K; (E) frozen solution of [Co(dppv)2(NCCH3)2][BPh4]2 dissolved in CH3CN, temperature = 4 K. Spectrometer settings: (A) microwave frequency = 9.377 GHz, microwave power = 0.02 mW; (B) microwave frequency = 9.861 GHz, microwave power = 2 mW; (C) microwave frequency = 9.377 GHz, microwave power = 0.08 mW; (D) microwave frequency = 9.436 GHz, microwave power = 2 mW; (E) microwave frequency = 9.376 GHz, microwave power = 0.004 mW. Simulations (red traces) achieved with magnetic parameters given in Table and these inhomogeneity parameters: (A) linewidth = 6.7 mT, g-strain = [0.026, 0.050, 0.007]; (B) linewidth = 6.4 mT, g-strain = [0.017, 0.023, 0.02]; (C) linewidth = 8.3 mT, g-strain = [0.013, 0.041, 0.027]; (D) linewidth = 8.0 mT, g-strain = [0.024, 0.046, 0.007]; (E) linewidth = 7.4 mT, g-strain = [0.00, 0.011, 0.0003]. Insets show spectra (black) and simulations (red) of g|| region for spectra B, D, and E, multiplied by a factor of 5 to better show the 59Co hyperfine splittings.

The 6-coordinate Co(II) complex 3 possesses two CH3CN axial ligands leading to a significant alteration of the corresponding EPR spectrum (Figure B) compared to the 5-coordinate species. The g-tensor is still axial with g⊥ > g|| ≈ ge, but the strong 59Co hyperfine splitting is now observable along g⊥, not along g||. This spectral change suggests a change in the ground-state orbital description for the unpaired electron from being housed in a cobalt-based 3d molecular orbital to one that is more 3d in character. Analysis of the g-values suggests that the xz/yz excited state lies ≈6500 cm–1 above the x2 – y2 ground state.

Addition of dppe (Figure C) reverts the spin system back to the 3dz ground state, as does dissolution of the complex in acetone (Figure D). However, dissolution of 3 in CH3CN (Figure E) seems to maintain the 3d ground state, and the corresponding g-values and 59Co HFI are very similar to those found for the solid 6-coordinate complex. This behavior suggests that the second CH3CN axial ligand is maintained in CH3CN solvent but is lost when the solvent is acetone. This loss of an axial ligand stabilizes the 3d orbital to an energy below that of 3d, leading to its occupation by the unpaired electron.

Angular Overlap Model Calculations

The one electron orbital energy levels of 1, 2, and 3 were calculated by diagonalizing matrices of overlap integrals of the angular components of the d functions along with the parameterized energies of sigma (eσ) and pi (eπ) interactions with the ligands.[32] For the equatorial dppv ligands, we used 4000 cm–1 (eσ) and −2500 cm–1 (eπ) for each of the four P–Co interactions.[33] These are the reported[33] angular overlap model (AOM) parameters for triphenylphosphine, which are likely close to the (unmeasured) parameters for dppv. Ultimately, changing the absolute values of eσ and eπ results only in a scaling relationship for our calculated orbital energies and does not affect the relative orderings. A P–Co–P “bite angle” of 83° was used for the equatorial ligand geometry since this is the average of the bite angles for 1 through 3 (σ = 0.9°).

One benefit of the AOM methodology is that the overlap integrals can be parameterized precisely.[34] Factoring energies in terms of the well-known Racah parameters A, B, and C requires some fundamental symmetry components to make the calculation feasible and meaningful;[35] unfortunately, the AOM model does not explicitly deal with the multielectron problem and configuration interaction. In the present case of a Co(II) center with an S = 1/2 ground state, configuration interaction does not significantly affect the energies of the lowest-energy doublet transitions. It should be noted, however, that more sophisticated perturbation treatments show non-negligible interactions between the doublet ground state and low-lying quartet states.[36] Additionally, a ligand field theory/AOM hybrid approach has been used to solve cases where interelectronic repulsion dominates [e.g., S = 1 Cr(II)].[37]

The full Mathematica[38] code to generate the enclosed plots is included in the Supporting Information.

Discussion

Figure presents the ligand field splitting (one-electron orbital energies) for the 6-coordinate complex 3. The diagram was determined by considering (1) an electronic absorption at 1200 nm (8300 cm–1) in the diffuse reflectance spectrum of solid 3, (2) the existence of an xz/yz excited state 6500 cm–1 above the x2 – y2 ground state, and (3) a second, asymmetric ligand-field transition at 700 nm (14,300 cm–1). Notably, the d orbital is close in energy to the d orbital. This can be predicted by EPR spectroscopy, which shows that the ground state changes going from 5-coordinate to 6-coordinate. We believe that this is the primary reason for axial ambivalence: a small energy gap allows for the crossing of the two frontier orbitals with increasing axial contribution. Occupation of the d orbital would therefore encourage axial binding, while occupation of the d orbital would favor axial ligand loss (d is σ* with respect to ligands along the primary rotation axis, z). Furthermore, we know that the ground state is (d)1 for 1 and 2 (owing to an axially symmetric EPR spectrum, with the largest element of the 59Co HFI appearing along g||), which places the crossover point somewhere between the geometries of complexes 2 and 3. From this perspective, the significant geometric distortion of complex 2 is likely explained by a second-order Jahn–Teller effect in which vibrations are coupled to a low-lying excited state.[39] Using the AOM model, we can generate the one-electron orbital energies as we progress from 4-coordinate to 5-coordinate to 6-coordinate (Figure ).

Figure 5

Ligand-field splitting diagram illustrating the electronic structure of the pseudo-octahedral 3 (6-coordinate) complex. The diagram is consistent with both EPR and electronic absorption spectroscopies, including the presence of a weak band at 1200 nm (8300 cm–1), which is seen in the NIR diffuse reflectance spectrum of solid 3. Other magnitudes are derived from the g-values in the EPR analysis (above).

Figure 6

One-electron orbital energies of the d orbitals as a function of the axial electronic interaction eσ. Note that the eπ component of the axial ligands does not affect the energies of the two highest energy levels because of orthogonality. The lower levels are slightly dependent on the axial eπ, but this interaction is removed for clarity. Here, eσ is treated as the sum of the two axial acetonitrile interactions.

Notably, these calculations confirm the prediction of a crossover point in the energy diagram. The energies of the d and dz orbitals are independent of the axial eπ component because of the orthogonality of those orbitals with the π symmetry orbitals of the axial ligand. The crossover is calculated at eσ(axial) ≈ 8000 cm–1. In terms of only radial functions, eσ is treated as the sum of two axial acetonitrile interactions. The energy gap between d and d for 3 is ∼1800 cm–1, so the gap between d and d for 2 must be <1800 cm–1. By plotting the transition or “cross-over” energy versus eσ(axial), we can calculate that eσ(CH3CN) is ∼5900 cm–1 (Figure ). This is a reasonable value of eσ(CH3CN) but indicates that acetonitrile is a better σ-donor than typically thought.[40] It is important to emphasize that eσ is bond length-dependent (radial integrals are not considered in AOM). Therefore, while eσ(CH3CN, @ 2 Å) is calculated to be 5900 cm–1 (5-coordinate), eσ(CH3CN, @ 2.1 Å) will be much smaller (6-coordinate). The splitting determined by EPR (∼2000 cm–1) validates the notion that binding two axial acetonitrile molecules does not double the σ interaction. A general rule-of-thumb is that the lengthening of a bond by ∼0.1 Å indicates a change in the antibonding level by ∼10 Dq/Δo; this is the equivalent of adding one antibonding electron. Thus, we conclude that eσ(CH3CN, @ 2 Å) ≈ 2eσ(CH3CN, @ 2.1 Å).

Figure 7

Locations of the 5- and 6-coordinate complexes (vertical lines) while visualizing the d/d energy gap in terms of eσ axial. The 5-coordinate complex 2 gives an exact value for eσ(MeCN), while the 6-coordinate complex 3 can be approximated as < 2eσ(MeCN) due to elongation of the Co–N bonds in 3.

For the sake of completeness, we note that complexes cannot exist within ± λ of the crossover, where λ is the reorganization energy. This so-called “forbidden region,” which arises from the fact that bond lengths are quantized in spin- and orbital–state transitions, is also applicable in Tanabe–Sugano diagrams for multielectron states.[41] This exclusion region primarily affects our predicted location of 2 relative to the crossover point since λ may be larger than 1800 cm–1, as assumed above. If the value of λ is greater than 1800 cm–1, eσ(MeCN) would necessarily be smaller as 2 moves away from the crossover point. Therefore, 5900 cm–1 is an upper-bound for eσ(MeCN).

The above calculations depend only on the values of eσ and eπ for the phosphine ligands. (At 90° there is no eπ contribution, using a similar argument as for the axial π interaction.) What happens as the values of eσ and eπ for the four equatorial ligands change? We can perform these calculations analytically. The most informative calculations are those that minimize the d/d energy gap because those would describe complexes at (or near) the crossover point between 5- and 6-coordination numbers. Thus, by knowing the appropriate equatorial ligand parameters of the square-planar core—phosphines or otherwise—we can predict whether potential substrates would be axially labile.

To visualize the results of these calculations, we generated contour plots that relate any combination of equatorial eσ and eπ values of the square-planar core to the eσ (axial) values of potential axial ligands that would be predicted to yield “axial ambivalent” complexes. Thus, by following a contour of a specific axial ligand sigma-donor strength, the corresponding values of the equatorial ligands that would be required for labile axial binding become apparent (Figure ). Plots are shown for three different equatorial bidentate ligand “bite angles”: 75, 83, and 90°. Interestingly, smaller bite angles cause significant distortions of the contours due to mixing of orbitals of π symmetry. Thus, in addition to modifying the σ-only strength of the equatorial ligands, one might also consider building scaffolds with varying “bites.” The large range of σ-donation and π-acceptor abilities of various phosphines has previously been explored.[42]

Figure 8

Contour plots that minimize the d/d energy gap as a function of eσ and eπ of the equatorial phosphines. Color indicates the magnitude of the axial eσ (total) at the minimum. White areas indicate regions where other crossings have occurred and the d/d transition is no longer relevant. Different “bite angles” of the phosphine ligands are calculated; (top) 75°, (middle) 83° (the average “bite angle” of dppv), and (bottom) 90°.

These contour plots provide a methodology for developing cobalt complexes which will reversibly bind a variety of substrates. We might be very interested in engineering cobalt cores which will preferentially bind some substrates over others. For example, how can we use these plots to explore complexes suitable for dinitrogen reduction in aqueous solutions? The axial sigma interaction [eσ (axial)] of dinitrogen bound end-on has been estimated to be 6900 cm–1.[43] With an 83° bite angle, eσ(equatorial) for the square-planar core should be ∼4750 cm–1, a slightly weaker sigma-donor than dppv. With a tighter “bite angle,” eσ(equatorial) could be as large as 5000 cm–1 if there were a small-to-moderate eπ(equatorial) contribution.

A major problem in dinitrogen reduction catalysis, however, is the simultaneous reduction of protons in aqueous solution. In the case of preferentially binding dinitrogen (rather than protons/hydrides), one quickly realizes that due to the lack of eπ variation in the plots, one cannot preferentially bind a weaker σ-donor. This is made apparent by the monotonic nature of the plots in Figure . One can show that this is true for any tetragonal geometry. As such, trigonal or otherwise significantly distorted complexes are of particular interest due to the admixture of states. Extending the model to include low-symmetry complexes is an ongoing work.

Conclusions

We report here the structural and spectroscopic characterization of a series of cobalt complexes that are 4-, 5-, and 6-coordinate with the same Co(dppv)22+ core. The 4-coordinate complex was isolated from acetone, while the 5- and 6-coordinate complexes were isolated from acetonitrile in the presence of and in the absence of BPh4–, respectively. Titration of BPh4– salt into a solution of the isolated 6-coordinated complex in acetonitrile shows subtle changes to the UV–vis spectrum, indicating an equilibrium between 6- and 5- coordinate species in solution. The crystal structures of each complex are reported, which includes a very rare example of a rigorously planar 4-coordinate cobalt(II) complex with neutral ligands. EPR spectroscopic results reveal that there is a change in the ground-state electronic configuration between the 5- and 6-coordinate structures; AOM calculations suggest readily tunable ligand parameters for accessing this “axial ambivalent” region for any potential axial-ligand substrate. A series of diagrams are provided that can assist in the intelligent design of equatorial phosphines (or any other scaffold) that will make catalytically relevant small molecules labile.