Connecting [NiFe]- and [FeFe]-Hydrogenases: Mixed-Valence Nickel-Iron Dithiolates With Rotated Structures

Abstract

A series of mixed-valence iron-nickel dithiolates is described that exhibits structures similar to those of mixed-valence diiron dithiolates. Interaction of tricarbonyl salt [(dppe)Ni(pdt)Fe(CO)₃]BF₄ ([1]BF₄, dppe = Ph₂PCH₂CH₂PPh₂, pdtH₂ = HSCH₂CH₂CH₂SH) with P-donor ligands (L) afforded the substituted derivatives [(dppe)Ni(pdt)Fe(CO)₂L]BF₄ incorporating L = PHCy₂ ([1a]BF₄), PPh(NEt₂)₂ ([1b]BF₄), P(NMe₂)₃ ([1c]BF₄), P(i-Pr)₃ ([1d]BF₄) and PCy₃ ([1e]BF₄). The related precursor [(dcpe)Ni(pdt)Fe(CO)₃]BF₄ ([2]BF₄, dcpe = Cy₂PCH₂CH₂PCy₂) gave the more electron-rich family of compounds [(dcpe)Ni(pdt)Fe(CO)₂L]BF₄ for L = PPh₂(2-pyridyl) ([2a]BF₄), PPh₃ ([2b]BF₄) and PCy₃ ([2c]BF₄). For bulky and strongly basic monophosphorus ligands, the salts feature distorted Fe coordination geometries: crystallographic analyses of [1e]BF₄ and [2c]BF₄ showed they adopt ‘rotated’ Fe(I) centers, in which PCy₃ occupies a basal site and one CO ligand partially bridges the Ni and Fe centers. Like the undistorted mixed-valence derivatives, the new class of complexes are described as Ni(II)Fe(I) (S = ½) systems according to EPR spectroscopy, although with attenuated ³¹P hyperfine interactions. DFT calculations using the BP86, B3LYP, and PBE0 exchange-correlation functionals agree with the structural and spectroscopic data, suggesting that the spin for [1e]⁺ is localized in a Fe(I)-centered d(z²) orbital, orthogonal to the Fe-P bond. The PCy₃ complexes, rare examples of species featuring ‘rotated’ Fe centers, both structurally and spectroscopically resemble mixed-valence diiron dithiolates. Also reproducing the NiS₂Fe core of the [NiFe]-H₂ase active site, the hybrid models incorporate key features of the two major classes of H₂ase. Furthermore, cyclic voltammetry experiments suggest that the highly basic phosphine ligands enable a second oxidation corresponding to the couple [(dxpe)Ni(pdt)Fe(CO)₂L]^+/2+. The resulting unsaturated 32e⁻ dications represent the closest approach to modeling the highly electrophilic Ni-SI_a state. In the case of L = PPh₂(2-pyridyl) chelation of this ligand accompanies the second oxidation.

Introduction

The hydrogenase (H₂ase) enzymes catalyze the processing of hydrogen, a reaction that is biologically significant and scientifically topical. These enzymes are prominent in anaerobic bacteria and archaea and are classified according to the metals present at their active sites.¹ In particular, the redox reaction 2H⁺ + 2e⁻ ⇌ H₂ is mediated by the [FeFe]- and [NiFe]-H₂ases, the two major classes of H₂ase. Members of a third class, labeled the [Fe]-H₂ases, have been characterized but do not perform redox on H₂ or protons.²

Although no phylogenic relationship exists between the [FeFe]- and [NiFe]-H₂ases,³ their active sites share a number of structural commonalities. Central to their function, the [FeFe]-H₂ase enzymes (45 – 130 kDa) feature an active site ensemble, the “H-cluster”, composed of a binuclear low-spin [Fe₂S₂] core linked through a cysteinate residue to a [Fe₄S₄] redox cofactor, which is itself tethered to the protein by three other cysteinate groups (Figure 1a).^4,5 Within the [Fe₂S₂] fragment, the two Fe centers are bridged by a 2-aza-1,3-propanedithiolate cofactor, and are further bound to CO and CN⁻ ligands. The latter participate in H-bonding to neighboring proline, lysine and serine residues, and thus while the [Fe₂S₂] fragment appears only loosely anchored to the backbone, the conformational freedom of the Fe coordination spheres is nevertheless restricted.

Figure 1.

Line drawings of: (a) the [FeFe]-H₂ase active site;⁶ (b) the [NiFe]-H₂ase active site;⁸ (c) a mixed-valence NiFe model complex;⁹ (d) a ‘rotated’ NiFe model complex described in this work.

In enzymes isolated in the H_ox (Fe(II)Fe(I)) state, the proximal (cysteinate-bound) Fe center interacts with a CO ligand that is shared with the other (distal) Fe atom.⁶ The latter center, believed to exist in the +I oxidation state, adopts a coordination geometry exposing a vacant site for H₂ binding. The ‘rotated’ arrangement of ligands around the distal Fe center results in it being preorganized for catalysis, which likely contributes to the high rates (and low overpotentials) associated with its function. ‘Rotation’ is also observed in [Fe]-H₂ase active sites, with Striebitz and Reiher having highlighted the structural and electronic similarities between the two classes of H₂ases.⁷

The heterodimeric [NiFe]-H₂ases each consist of a small and a large subunit (28 + 60 kDa). These more prevalent enzymes are typically located in the periplasm¹⁰ and feature heterometallic active sites not dissimilar to those of the homobimetallic [FeFe]-H₂ases. The [NiFe]-H₂ases also feature a thiolate-bound iron fragment, in this case linked to a nickel center (Figure 1b). This center exists in a conformationally rigid see-saw geometry, coordinated to four soft, strongly σ- and π-donating cysteinate ligands. These S-donors stabilize Ni in high (+III, hydride-bound), intermediate (+II), and low (+I) oxidation states. The principal states of the [NiFe]-H₂ase enzyme differ not only with respect to the oxidation state of Ni but also with the presence or absence of hydride ligands bridging the two metals (Scheme 1).^11,12

Scheme 1.

Mixed-valence states play a prominent role in the functioning of both H₂ases. In the [NiFe]-H₂ases, the EPR-active Ni-C state, an intermediate in the oxidation of H₂, features Ni(III)Fe(II) metal centers bridged by a hydride ligand.¹³ Proton loss effected by low-temperature UV-irradiation of Ni-C results in another S = ½ state, Ni-L, which features a Ni(I)Fe(II) core.⁸ In this case the oxidation states have been assigned according to EPR data (vide infra) and parallels can be drawn between this system (in which the spin resides on Ni) and the H_ox state of the [FeFe]-H₂ases (in which the spin resides on the distal Fe). Despite its low formal oxidation state, the Fe(I) center in the latter is able to activate H₂,⁷ by a mechanism which likely involves proton-coupled electron transfer.¹⁴ In contrast, while Ni-L binds CO, it is insufficiently electrophilic to activate H₂. The substrate is instead cleaved by the more oxidized Ni-SI_a state, featuring a low-spin 32e⁻ Ni(II)Fe(II) core, in which the electrophilic centers are poised to accept a bridging H⁻ ligand. CO is a potent inhibitor of Ni-SI_a, which is poisoned to afford the active site adduct (CO)(cysteinate)₂Ni(μ-cysteinate)₂Fe(CO)(CN)₂, known as Ni-SCO.¹⁵ In comparison to the [FeFe]-H₂ases, one can conclude that the NiFe enzymes operate with metal centers at higher formal oxidation states, a fact that might be attributed to the four basic cysteinate residues present at the active site.

Synthetic modeling of H₂ase active sites can provide insights into the mechanisms by which these enzymes operate.¹⁶ For example, recent work on the [FeFe]-H₂ases has highlighted the acid/base¹⁷ and redox¹⁴ functionality necessary for catalytic activity. Models for the paramagnetic states are of particular interest, owing to the novelty and synthetic challenges associated with open-shell organometallic compounds. A prominent mixed-valence model is [(IMes)(CO)₂Fe(pdt)Fe(CO)₂PMe₃]⁺ (Figure 2, left),¹⁸ in which strongly σ-donating N-heterocyclic carbene (IMes) and trialkylphosphine ligands represent surrogates for the CN⁻ ligands found in the enzyme. Notably, the (IMes)(CO)₂Fe(I) fragment features a semi-bridging CO, with the two terminal ligands oriented in such a way as to expose a vacant coordination site, thereby reproducing the rotated coordination geometry observed for the distal Fe.

Figure 2.

Mixed-valence H₂ase models: [(IMes)(CO)₂Fe(pdt)Fe(CO)₂PMe₃]⁺ (left) and [(dppe)Ni(pdt)Fe(CO)₃]⁺ (right)

The first paramagnetic models for [NiFe]-H₂ase have recently been reported, one example of which is the prototypical S = ½ species [(dppe)Ni(pdt)Fe(CO)₃]⁺ (Figure 2, right).⁹ EPR studies support the assignment of oxidation states as Ni(II)Fe(I), such that the model resembles Ni-L with oxidation states reversed. DFT calculations suggest that the spin is largely Fe-centered and that Ni exists in a square-planar coordination geometry, characteristic of a Ni(II) (d⁸) system. The electrophilicity of the Fe(I) center in these mixed-valence cations is sufficient to allow the replacement of one CO ligand for PPh₃, thereby affording a more electron-rich species (Figure 1c).⁹ This complex was also described as Ni(II)Fe(I), and evidently the (CO)₂(PPh₃) ligand set is not as donating as the (CN)₂(CO) ligands present in [NiFe]-H₂ase, given that the Ni(I)Fe(II) state is possible in the latter case.

This paper details the synthesis of new Ni(II)Fe(I) complexes of type [(dxpe)Ni(pdt)Fe(CO)₂L]⁺, characterized spectroscopically, crystallographically, electrochemically, and computationally. Through systematic modulation of the steric bulk and basicity of the mono- and diphosphine (L and dxpe, respectively), new insights are afforded into the relationship between structure and spectroscopy of mixed-valence derivatives. While our previous report only disclosed compounds in which L is poorly basic,⁹ presented here are examples incorporating highly basic ligands L, which differ structurally and spectroscopically. In particular, remarkable changes in the Fe coordination geometry are effected, resulting in rare examples of highly distorted Fe(I) species, as previewed in Figure 1d and discussed below. Studies on the oxidation of these models to 32e⁻ Ni(II)Fe(II) derivatives afford evidence for the first models for Ni-SI_a, featuring 16e⁻ Fe(II) centers.

Herein, we extend the comparison between [FeFe]- and [NiFe]-H₂ases and their model complexes. As we show, the connections between the two classes of H₂ase and their models run much deeper than simple consideration of their coordination spheres. Indeed, the relationships between the two enzymes and their models are also electronic in nature. The new complexes described below serve to exemplify this with respect to the mixed-valence active sites present in the 33e⁻ H_ox and Ni-L states, defining a new link between the [FeFe]- and [NiFe]-H₂ases.

Results and Analysis

Syntheses

In this work, a range of new derivatives of the type [(dppe)Ni(pdt)Fe(CO)₂L]BF₄ (vide supra) were prepared with an emphasis on highly basic phosphines. It was reasoned that highly basic alkylphosphines would lower the Fe(I/II) couple relative to the Ni(II/I) couple. Furthermore, it was anticipated that the new complexes could undergo oxidation to coordinatively unsaturated dicationic models for Ni-SI_a. To this end, modulating the basicity of the Ni-bound diphosphine was also of interest, given that it was also expected to have a significant effect on the electron density at the metal centers. Thus, through systematic variation of the phosphine ligands in complexes of type [(diphosphine)Ni(pdt)Fe(CO)₂(monophosphine)]⁺, new species were generated. As expected, the first electrochemical studies on complexes of this type indicate correlations between ligand basicity and redox potentials. The compounds described also differ significantly from [(dppe)Ni(pdt)Fe(CO)₂PPh₃]BF₄ in terms of structure and spectroscopy. The preparation of these new salts is now outlined.

The dppe derivative [1]BF₄ could be converted to the salts [1a–e]BF₄, while [2]BF₄ served as the precursor for [2a–c]BF₄, which feature the more strongly σ-donating dcpe ligand. Briefly, addition of in situ generated tricarbonyl salts to CH₂Cl₂ solutions containing excess monophosphine (L) afforded the substituted radicals. Addition of pentane to the mixtures precipitated yellow to green solids in yields typically exceeding 70%. The products are sensitive to O₂ and H₂O. They decompose to CO-free products at room temperature over the course of days. The salts were characterized according to analytical and ESI-MS data, the latter revealing the tendency of the compounds to ionize by loss of BF₄⁻, in some cases with dissociation of a CO ligand to afford the ions [(dxpe)Ni(pdt)Fe(CO)₂L]⁺ and [(dxpe)Ni(pdt)Fe(CO)L]⁺.

Solution IR spectra for the new compounds typically exhibited two v_CO bands (Table 1). However, the PPh(NEt₂)₂ derivative [1b]BF₄, like the previously reported PMePh₂ complex,⁹ exhibits four comparably intense bands, assigned to conformers that differ with respect to the disposition of the pdt backbone. The stretching frequencies are affected by the monodentate P-donor ligands, being lowest for complexes of the strong donor PCy₃. For the PCy₃-containing derivatives [1e]BF₄ and [2c]BF₄, values of v_CO are relatively insensitive to the identity of the Ni-bound diphosphine (dppe vs dcpe). Two v_CO bands lower in energy were observed for the ¹³CO-labeled complexes [1d′]⁺ and [1e′]⁺.

Table 1.

IR data for New Salts of Type [(dxpe)Ni(pdt)Fe(CO)₂L]BF₄ in CH₂Cl₂ Solution. Data for ¹³CO-labeled Analogs [(dppe)Ni(pdt)Fe(¹³CO)₂L]BF₄ are Included, These Compounds Being Denoted with Prime Symbols.

compound	dxpe	L	vCO /cm−1
[1a]BF4	dppe	PHCy2	1974, 1914
[1b]BF4	dppe	PPh(NEt2)2	1982, 1972, 1923, 1903
[1c]BF4	dppe	P(NMe2)3	1967, 1899
[1d]BF4	dppe	P(i-Pr)3	1966, 1899
[1d′]BF4	dppe		1921, 1857
[1e]BF4	dppe	PCy3	1966, 1899
[1e′]BF4	dppe		1921, 1856
[2a]BF4	dcpe	PPh2(2-py)	1988, 1928
[2b]BF4	dcpe	PPh3	1984, 1925
[2c]BF4	dcpe	PCy3	1964, 1898

Relation to S = ½ Hydrides

Although the new complexes contain highly basic ligands and are oxidizable (vide infra), they resist protonation. For example, treatment of the electron-rich complex [1e]⁺ with the strong acid H(OEt₂)₂BAr^F₄⁻ (BAr^F₄ = B(3,5-(CF₃)₂C₆H₃)₄⁻) (1 equiv) resulted in no initial reaction. After 1 h, IR spectroscopic analysis of the reaction solution indicated the presence of [1e]⁺, [1]⁺ (v_CO = 2057, 1986 cm⁻¹) and [1H]⁺ (v_CO = 2082, 2024 cm⁻¹). The latter cation was the only CO-containing species detected after 17 h, the solution having changed color from green to deep red. It is interesting to consider the formation of the tricarbonyl hydride [1H]⁺; the targeted paramagnetic dicarbonyl hydride [1eH]²⁺ could not be detected. Treatment of [1b]⁺ with the same acid also afforded [1H]⁺, although in this case the tricarbonyl radical [1]⁺ was the major CO-ligated product. Here it is possible that protonation of the weakly basic amine groups in PPh(NEt₂)₂ induces ligand dissociation and CO redistribution. In any case, the cannibalization that leads to the tricarbonyl products suggests that complexes of the present type are not suitable precursors to S = ½ hydrides and that other synthetic approaches are required to generate these Ni-C models.

EPR Spectra for [(dppe)Ni(pdt)Fe(CO)₂L]⁺

All mixed-valence derivatives were characterized by X-band EPR spectroscopy at 110 K; selected spectra were simulated in order to extract g and A(³¹P) values (Table 2). Spectra not presented here can be found in the Supporting Information.

Table 2.

EPR simulation parameters. Each compound has an entry for each of the two isomers, the relative abundances of which are given in the last column. Parameters for [1e]BF₄ were derived from both X- and Q-band data. Note: edtH₂ = HSCH₂CH₂SH.

compound	g-factor	A(31P) /MHz	linewidth /G	weight
[1a]BF4	2.076, 2.046, 2.000	54, 60, 59	9, 13, 7	75
	2.079, 2.039, 2.005	27, 8, 39	8, 6, 9	25
[1b]BF4	2.079, 2.052, 1.997	101, 161, 132	14, 9, 11	52
	2.075, 2.041, 2.002	66, 31, 52	11, 11, 10	48
[1e]BF4	2.089, 2.036, 2.008		6, 4, 10	54
	2.087, 2.036, 2.010		11, 7, 9	46
[1e′]BF4	2.087, 2.036, 2.011		14, 17, 17	54
	2.089, 2.036, 2.007		9, 7, 11	46
Ni-L8	2.298, 2.116, 2.043
Hox20	2.097, 2.039, 1.999
[(IMes)(CO)2Fe(pdt) Fe(CO)2PMe3]+18	2.180, 2.096, 2.052
[(PCy3)(CO)2Fe(edt) Fe(CO)dppv]+21	2.096, 2.042, 2.001	69, 67, 73

The EPR spectra of all derivatives prepared feature two overlapping rhombic signals assigned to conformers related by the flipping of the pdt²⁻ chelate ring. The new complexes could be classified into two families, one of which is similar to previously reported species, the other being very distinct. For the triarylphosphine complexes [2a]⁺ and [2b]⁺, each of the resonances is split by a single ³¹P nucleus (Figure S29, S32), indicating a Ni(II)Fe(I) description for these species. Indeed, they are spectroscopically almost identical to their dppe analogs, which were shown by ¹³CO-labeling studies to adopt a structure in which the monophosphine occupies the apical Fe coordination site.⁹

Whereas substitution of the diphosphine only subtly influences the Fe center, variation of the Fe-bound monophosphine ligand has a marked effect. Motivated by the synthesis of high fidelity Ni(I)Fe(II) (and Ni(II)Fe(II)) models, strongly donating alkylphosphines were employed, as these more basic σ-donors might stabilize the Fe(II) centers in the targeted complexes. The basic ligands PHCy₂, PPh(NEt₂)₂, P(NMe₂)₃ and P(i-Pr)₃ were be incorporated into the complexes [1a]⁺, [1b]⁺, [1c]⁺ and [1d]⁺, respectively, using the general procedure outlined above. Notably, their EPR spectra indicate the ³¹P hyperfine interactions for these complexes to be greatly attenuated (A ≈ 50 MHz) relative to triarylphosphine complexes (A ≈ 200 MHz). Similar values were observed for the ‘rotated’ H_ox model [(PCy₃)(CO)₂Fe(edt)Fe(CO)dppv]⁺,²¹ suggesting a connection between Fe(I) coordination geometry and hyperfine interactions. The couplings found here are related to phosphine basicity – they are strongest in the case of PHCy₂ and weakest in the complex of P(i-Pr)₃, the most basic of the four P-donor ligands.²² The influence of the phosphine is clearest for the PCy₃-containing complexes [1e]⁺ and [2c]⁺. In contrast to the complexes of PPh₃ and PPh₂(2-py), which are yellow in color, the PCy₃-containing complexes are deep green (for representative UV-vis spectrum, see Figure S17), suggesting a significant difference in electronic structure.

Unique among the series of complexes, [1e]BF₄ and [2c]BF₄ give EPR spectra with no ³¹P hyperfine splitting. The complex [1e]⁺ gives rise to two overlapping rhombic signals at 110 K, the z components of which are at lower field (Figure 3); such a pattern is consistent with spin localization on Fe(I). The individual simulated signals, as well as their sum, are presented for X- and Q-band spectra of [1e]⁺ (Figure S21, S22). Lastly, the EPR spectroscopy of [1e]⁺ is also distinctive in that a single line was observed at room temperature (see Figure S20), whereas all other complexes reported here give spectra reflecting the presence of multiple isomers.

Figure 3.

X-band EPR spectra (CH₂Cl₂/PhMe, 110 K) of [1e]BF₄ (exp.) and [1e′]BF₄ (exp.′). Simulated spectra are also presented.

EPR analysis was also conducted on the ¹³CO-labeled derivative [1e′]⁺. In frozen solution, the spectrum features the expected rhombic signals but no ³¹P coupling. Furthermore, coupling to the two ¹³CO ligands is observed, this being most evident at g = 2.036 where the triplet pattern indicates that the CO ligands are equivalent. If the ¹³CO ligands were indeed basal, the monophosphine ligand would occupy the apical position. However, such a stereochemistry appeared inconsistent with the large A(³¹P) values that would be expected for apically-bound P-donor ligands. The spectral properties were rationalized in terms of the unusual structures adopted by these cations, which are discussed in the following section.

Structure of [(dppe)Ni(pdt)Fe(CO)₂PCy₃]BF₄

Structural confirmation of a substituted derivative of [1]⁺ was obtained in the case of the PCy₃ complex salt [1e]BF₄ (Figure 4).

Figure 4.

ORTEP of [1e]BF₄ (left) with ellipsoids drawn at the 50% probability level. The H atoms and disordered BF₄⁻ anion are omitted for clarity. A view down the Fe-Ni vector (right) is also provided, in which the carbocyclic rings are omitted. Selected distances (Å, values calculated using BP/TZVP given in parentheses): Ni1-Fe1, 2.637 (2.77); Ni1-P1, 2.168 (2.23); Ni1-P2, 2.171 (2.23); Ni1-S1, 2.213 (2.28); Ni1-S2, 2.224 (2.25); Fe1-S1, 2.305 (2.36); Fe1-S2, 2.352 (2.33); Fe1-C30, 1.808 (1.78); Fe1-C31, 1.754 (1.76); Fe1-P3, 2.283 (2.36). Selected angles (°, calculated values in parentheses): Ni1-Fe1-C30, 75.0 (78.4); Fe1-C30-O1, 175.6 (178.3).

Perhaps the most remarkable finding is the distorted geometry of the Fe(I) center, which adopts a ‘rotated’ structure. One CO ligand occupies a semi-bridging position between the Fe and Ni centers although the Ni-CO distance is long (2.784 Å) and the Fe-C-O atoms are almost collinear (176°).²³ The central methylene of the pdt²⁻ bridge is poised over the vacant Fe coordination site, in contrast to the ‘unrotated’ derivative [(dppe)Ni(pdt)Fe(CO)₂PPh₃]⁺⁹ and the hydride [(dppe)Ni(pdt)HFe(CO)₂PPh₃]⁺.²⁴ Unlike the dcpe-containing tricarbonyl [2]⁺, which features a square pyramidal Fe center,⁹ the Fe coordination geometry in [1e]⁺ is intermediate between square pyramidal and trigonal bipyramidal, its Addison τ parameter²⁵ (here the difference between the angles S1-Fe1-C31 and S2-Fe1-P3 divided by 60) being 0.42 (τ = 0.02 for [2]⁺). The mean Fe-C distance in [2]⁺ (1.807 Å) is somewhat greater than that in [1e]⁺ (1.781 Å), owing to the stronger π-backbonding in the substituted derivative. The coordination environment of Ni is distorted from planarity, the twist angle between the NiP₂ and NiS₂ planes being 11.6°. The Ni-S (2.213, 2.224 Å) and Ni-P distances (2.171, 2.168 Å) are similar to those in [2]⁺. These observations support a +II oxidation state assignment for Ni.

The Ni-Fe distance in [1e]⁺ (2.637 Å) is shorter than that in [2]⁺ (2.818 Å). The greater intermetallic separation in [1e]⁺ translates to a larger dihedral (“butterfly”) angle between the NiS₂ and FeS₂ planes for this species (105.2°) relative to that for [2]⁺ (116.4°). Covalent radii for Ni (1.24 Å) and low-spin Fe (1.32 Å) have been determined from CSD entries,²⁶ their sum (2.56 Å) representing an upper limit for the length of a formal Ni-Fe bond. Accordingly, such a bond may be considered absent in [1e]⁺, although the possibility of weak interactions between these centers is not ruled out. The distance is very similar to that observed crystallographically for [NiFe]-H₂ase (2.60 Å for Ni-C/Ni-R state in Desulfovibrio vulgaris Miyazaki F)²⁷ and [FeFe]-H₂ase (2.62 Å for H_ox state in Clostridium pateurianum),²⁸ although it is noted that in the latter case the Fe centers are within bonding distance.

Structure of [(dcpe)Ni(pdt)Fe(CO)₂PCy₃]BF₄

According to the X-ray crystallographic analysis, the dcpe-containing salt [2c]BF₄ is virtually isostructural with [1e]BF₄ (Figure 5).

Figure 5.

ORTEP of [2c]BF₄·CH₂Cl₂·0.5 pentane (left) with ellipsoids drawn at the 50% probability level. The H atoms, disordered CH₂Cl₂/pentane solvate molecules, and BF₄⁻ anion are omitted for clarity. A view down the Fe-Ni vector (right) is also provided, in which the carbocyclic rings are omitted. Selected distances (Å): Ni2-Fe1, 2.990; Ni2-P2, 2.192; Ni2-P3, 2.197; Ni2-S1, 2.226; Ni2-S2, 2.231; Fe1-S1, 2.345; Fe1-S2, 2.321; Fe1-C19, 1.781; Fe1-C20, 1.744; Fe1-P3, 2.306. Selected angles (°): Ni2-Fe1-C19, 84.4; Fe1-C30-O1, 173.4.

Once more, the Fe center exists in a highly distorted square-pyramidal coordination environment (τ = 0.39), although the CO closest to Ni is much further from a bridging position (Ni2-Fe1-C19 = 84°) than is the case with [1e]⁺, perhaps suggesting that the present system is more of a ‘pure’ Fe(I) system. Moreover, compared to [1e]⁺, the Ni coordination geometry in [2c]⁺ is closer to square planar (twist angle between NiP₂ and NiS₂ planes = 2.8°), consistent with a Ni(II) site. Perhaps owing to the increased basicity of dcpe over dppe, the divalent state for Ni is particularly stabilized for [2c]⁺. The large “butterfly” angle (122.5°) reflects the ‘open’ nature of [2c]⁺, with the intermetallic separation being significantly greater in this complex (2.990 Å) relative to [1e]⁺. It appears the interactions between the metal centers in [2c]⁺ are of a negligible nature, indicative of the strongly donating ligands at Ni satisfying its Lewis acidity such that no further contacts with Fe1 or C19 are necessary.

Cyclic Voltammetry

The effects of the phosphine substituents on electronic structure of the mixed-valence complexes were probed by cyclic voltammetry. Analysis of selected derivatives in CH₂Cl₂/NBu₄PF₆ electrolyte solution was performed under an N₂ atmosphere. Well-defined reduction and oxidation waves were observed, these being assigned to metal-centered events, which on the basis of the peak separations (ΔE_p), involve the transfer of a single electron. For example, in the case of [2c]BF₄, separations of 0.071 and 0.072 V were measured for the anodic and cathodic waves, respectively; under these conditions the value for the ferrocene/ferrocenium (Fc/Fc⁺) couple was found to be 0.069 V. Reduction of the Ni(II)Fe(I) cations afforded neutral Ni(I)Fe(I) complexes, with i_pc/i_pa values being close to unity (Table 3). This reversibility suggests that the neutral species are stable; indeed, in addition to the robust tricarbonyl complexes 1 and 2 used as precursors in the present study, the substituted Ni(I)Fe(I) derivative (dppe)Ni(pdt)Fe(CO)₂(PPh₃) has also been isolated as a stable solid.²⁴ Consistent with the reduction being Ni-centered, use of the more basic dcpe ligand in place of dppe caused, in most cases, a cathodic shift on the order of 0.3 V for the couple [(dxpe)Ni(pdt)Fe(CO)₂L]^0/+. Similar shifts have been reported for the couples [(dxpe)Ni(edt)Fe(CO)₃]^0/+ and [(dxpe)Ni(pdt)Fe(CO)₃]^0/+ when the diphosphine is changed.²⁹

Table 3.

Redox Potentials (V vs Fc/Fc⁺) and Associated Electrochemical Data for [(dxpe)Ni(pdt)Fe(CO)₂L]BF₄ (1 mM) in NBu₄PF₆ Electrolyte (100 mM CH₂Cl₂ solution) Recorded at 0.1 Vs⁻¹. Oxidations are Given as Half-Wave Potentials and Anodic Peak Potentials for Partially Reversible and Irreversible Couples, Respectively. Measurements for (dxpe)Ni(pdt)Fe(CO)₃ Were Recorded on PhCN Solutions.

compound	dxpe	L	E(NiIIFeI/NiIIFeII) (|ipc/ipa|,|ΔEp|)	E½(NiIFeI/NiIIFeI) (|ipc/ipa|, |ΔEp|)
129	dppe	CO	−0.04 (irrev.)	−0.58 (1.08)
-	dppe	PPh2(2-py)	−0.19 (irrev.)	−0.83 (1.00, 0.072)
-	dppe	PPh3	−0.04 (0.51, 0.068)	−0.80 (0.99, 0.062)
[1e]BF4	dppe	PCy3	−0.14 (0.68, 0.076)	−0.92 (1.00, 0.072)
229	dcpe	CO	0.22 (irrev.)	−0.82 (1.05)
[2a]BF4	dcpe	PPh2(2-py)	−0.29 (irrev.)	−1.11 (1.10, 0.067)
[2bBF4	dcpe	PPh3	−0.01 (0.52, 0.063)	−1.11 (1.09, 0.073)
[2c]BF4	dcpe	PCy3	−0.12 (0.97, 0.071)	−1.21 (1.09, 0.072)

The anodic waves observed for the mixed-valence salts are assigned to Fe-centered oxidations, generating dicationic Ni(II)Fe(II) complexes. In contrast to the reductions, oxidations are virtually unaffected by the identity of the Ni-bound diphosphine, dppe vs dcpe, despite strong differences in their basicities.³⁰ Similarly, the diphosphine also has little influence on the v_CO values (vide supra), further confirming the assignment of the anodic waves. The [(dxpe)Ni(pdt)Fe(CO)₂L]^+/2+ couple is sensitive to the Fe coordination sphere. For example, substitution of PPh₃ in [2b]BF₄ for the stronger σ-donor PCy₃ ([2c]BF₄) results in a cathodic shift of 0.11 V. The latter compound, decorated with three alkylphosphines, exhibits the mildest (and most reversible) oxidations. Replacement of the N₂ atmosphere with H₂ did not affect any of the results described.

The anodic waves observed for the PPh₂(2-py) species [(dppe)Ni(pdt)Fe(CO)₂PPh₂(2-py)]⁺ and [(dcpe)Ni(pdt)Fe(CO)₂PPh₂(2-py)]⁺ ([2a]⁺) were considerably different to those for other substituted derivatives. While voltammograms of the dppe complex are complicated, [2a]⁺ exhibits more well-defined behavior. In such cases, the electrophilic Fe(II) centers generated upon oxidation are thought to bind the pyridyl groups such that complexes of type [(diphosphine)Ni(pdt)Fe(CO)₂(κ₂-PPh₂(2-py))]²⁺ form. Thus, chelation occurs to ‘quench’ the Lewis acidic Fe site by completing its octahedral coordination sphere. The process is very rapid and the irreversibility of the oxidation waves even at high scan rates (1 Vs⁻¹) precluded extraction of kinetic data. Representative data are given for [2a]BF₄ and [2c]BF₄ (Figure 6) and further electrochemical data can be found in the Supporting Information.

Figure 6.

Cyclic voltammograms for the PPh₂(2-py) derivative [2a]BF₄ (dotted trace) and PCy₃ derivative [2c]BF₄ (solid trace).

Reactivity

The irreversibility of the [1]^+/2+ couple suggested that the electrophilicity of the tricarbonyl cations might result in enhanced reactivity. Indeed, treatment of 1 with FcBF₄ (2 equiv) in CH₂Cl₂, afforded a complex mixture including the fluoride complex [(dppe)Ni(pdt)FFe(CO)₃]⁺ ([1F]⁺), identified according to IR and MS data. The same fluoride complex was generated by treating 1 with 1-fluoropyridinium tetrafluoroborate (1 equiv) in CH₂Cl₂/MeCN (v_CO = 2114, 2067, 2028 cm⁻¹; m/z 733.7 [1F – CO + MeCN]⁺, 720.7 [1F]⁺; Figures S38, S39).

The observation of F⁻ abstraction suggested the use of the non-coordinating anion BAr^F₄⁻, whose Fc⁺ salt allows for clean conversion of 1 to [1]BAr^F₄.⁹ A mixture containing [1]BAr^F₄ and excess FcBAr^F₄ was used in preliminary hydride abstraction experiments. No reactivity was observed when the solution was exposed to H₂, even in the presence of the (non-coordinating) external base P(o-tolyl)₃. However, the targeted hydride [1H]⁺ was observed when the stronger hydride source Ph₂SiH₂ was employed. Thus, treatment of a CH₂Cl₂ solution of [1]BAr^F₄ and FcBAr^F₄ (1 equiv) with Ph₂SiH₂ (10 equiv) resulted in complete conversion to [1H]⁺ after 4 h, as evidenced by IR (Figure S40, S41) and ESI-MS measurements. An analogous experiment, in which [1]BAr^F₄ alone was allowed to interact with Ph₂SiH₂ afforded some [1H]⁺, although in this case the reaction was slower and several unidentified coproducts were observed by IR spectroscopy (Figure S42). A possible mechanism for the reaction involves the binding of Ph₂SiH₂ to [1]⁺, followed by oxidation and heterolytic Si-H bond cleavage. Nevertheless, other pathways are also possible, especially given the propensity of Ph₂SiH₂ to participate in H atom transfer reactions.³¹

DFT Calculations

Density functional theoretical (DFT) calculations were performed for the ‘rotated’ complex [1e]⁺ in an effort to rationalize EPR data and assign the oxidation states. The calculations support a Ni(II)Fe(I) description, with the predicted structural parameters agreeing with those determined experimentally to within 0.1 Ǻ for bond lengths and 3° for bond angles (Figure 4 caption, vide supra). As was described above, the flipping of the pdt²⁻ chelate ring gives rise to two conformers, which differ in the relative orientation of the central −CH₂− group. These ‘flipamers’ are referred to as ‘a’ or ‘b’ depending on whether this group is closest to the Ni or Fe centers, respectively. Optimized structures for the flipamers [1e]^+a and [1e]^+b (the conformer found in the solid state) were determined. The transition state (TS) for their interconversion was also characterized; chemical structures and isocontour plots of unpaired spin densities are presented in Figure 7.

Figure 7.

Schematic drawings of the [1e]⁺ conformations and the transition state between the two. Below these are isocontour plots of the singly-occupied molecular orbital (SOMO) at 0.03 h for [1e]^+a, TS and [1e]^+b.

The Ni-Fe-P angle in [1e]^+a and [1e]^+b is 152° and 158°, respectively, with the PCy₃ ligand being displaced from an ideal ‘apical’ position. In each case one CO ligand is semi-bridging, with the CO-Ni distance in [1e]^+a and [1e]^+b being 3.58 Å and 3.39 Å, respectively. In the TS, which is characterized by a single imaginary frequency corresponding to chelate ring inversion, the pdt²⁻ ligand is planar and perpendicular to the Ni-Fe vector. The ring flip has a significant effect on the Ni-Fe distance, which increases from 2.87 Å in [1e]^+a to 3.00 Å in the TS, before decreasing to 2.77 Å in [1e]^+b. A further conformer ([1e]^+a′) was analyzed, it being similar in structure to [1e]^+a but with PCy₃ occupying a basal site (Figure S24). This stereochemistry is of particular relevance to the enzyme, in which the strongly σ-donating CN⁻ ligands are basal. Selected metrics for the optimized structures are given in Table S1.

The two flipamers of [1e]⁺ are closely matched in energy, yet have a large barrier to interconversion (Table 4). Given the steric demand of the apical PCy₃ ligand, it was expected that flipamer ‘a’ is more stable than ‘b’. Indeed, all DFT calculations suggest that the Gibbs free energy of [1e]^+a is lower by 2–4 kcal mol⁻¹. The activation barrier for conversion of [1e]^+a to [1e]^+b is 6–9 kcal mol⁻¹. The calculated barrier is only slightly affected when van der Waals interactions are considered, with BP/TZVP and BP86⁺VDW/TZVP calculations giving 7.6 and 5.8 kcal mol⁻¹, respectively. The hybrid functionals B3LYP and B3LYP⁺VDW/TZVP predicted barriers of 8.6 and 9.0 kcal mol⁻¹, respectively.

Table 4.

Change in Gibbs free Energy (kcal mol⁻¹) for the: (i) Ring-Flipping Process Between [1e]^+a and [1e]^+b, and (ii) Turnstile Rotation Interconverting [1e]^+a and [1e]^+a′.

functional/basis set	ΔGa→b	ΔGa→TS	ΔGTS→b	ΔGa→a′
BP/TZVP	+2.5	+7.6	−5.1	−0.2
BP+D/TZVP	+1.4	+5.8	−4.4	+1.1
B3LYP/TZVP	+2.4	+8.6	−6.2	+0.5
B3LYP+D/TZVP	+4.2	+9.0	−4.8	+0.9

Despite the steric bulk of PCy₃, the calculations indicate that this ligand, not CO, is slightly favored to occupy the pseudo-apical Fe site. Indeed, the turnstile rotation process converting [1e]^+a (pseudo-apical PCy₃) to [1e]^+a′ (basal PCy₃) is calculated to be endergonic by about 1 kcal mol⁻¹ (Table 4) when dispersion corrections are considered. Similar results were obtained when a COSMO model was employed to simulate interactions with CH₂Cl₂ solvent (data not shown).

Theory (BP/TZVP) indicates that the unpaired spin for [1e]⁺ resides almost exclusively on the Fe center, irrespective of which conformation the complex adopts (Table 5). As is evident in the isocontour plots for the [1e]⁺ conformers (Figure 7), pdt²⁻ chelate ring flipping or Fe turnstile rotation have little influence on the spin density distribution. Similar results were obtained from spin-unrestricted calculations (values obtained using B3LYP/TZVP can be found in Table S2).

Table 5.

Atomic Spin Populations for the Conformers of [1e]⁺ Calculated Using BP/TZVP.

conformer	ρ(Fe)	ρ(Ni)	ρ(Pdppe)	ρ(PCy3)
[1e]+a	0.94	0.03	0.01, 0.01	0.007
TS	0.94	0.02	0.01, 0.01	−0.01
[1e]+b	0.87	0.07	0.01, 0.01	0.03
[1e]+a′	0.79	0.05	0.01, 0.01	−0.02

Despite the spin distributions in [1e]^+a and [1e]^+b being almost identical, their isotropic ³¹P hyperfine interactions (+10 and +72 MHz, respectively) differ somewhat, which might be expected given the increase in unpaired spin density ρ(PCy₃) from 0.007 for [1e]^+a to 0.03 for [1e]^+b (BP86/TZVP). Nevertheless, the magnitudes of these A(³¹P) values are low and they could not be resolved in the experimental EPR spectra. The extremely weak nature of the hyperfine interactions is proposed to result from the unusual geometry of [1e]⁺, as similar DFT calculations for the ‘unrotated’ analog [(dppe)Ni(pdt)Fe(CO)₂PPh₃]⁺ (in which the triphenylphosphine occupies an axial position) predicted much larger isotropic hyperfine interactions (+178 and +204 MHz).⁹

The g-tensor principal values were calculated from spin-unrestricted CP-SCF and spin-restricted ZORA calculations (Table S3). The BP86 functional was found to underestimate g-shifts, while PBE0 overestimated g-shifts; the best agreement was obtained using the B3LYP hybrid functional in spin-unrestricted calculations. The self-consistent consideration of spin-orbit coupling in spin-restricted calculations with Slater basis functions gave larger g-shifts compared to the effective potential approach through the coupled-perturbed SCF equation. All methods afforded rhombic g-tensors for the various isomers of [1e]⁺ (Table S2), with the smallest g-component g₃ being close to g_e. This is suggestive of a d(z²)-centered SOMO; pure Ni(I) species can be ruled out as larger g-shifts would be expected in such a case, owing to the significant d(x²−y²) character of the SOMO in a low-spin d⁹ square-planar system. In contrast, the g-values reported for Ni-L (2.298, 2.116 and 2.043),¹⁹ in particular the deviation of the g_z component from g_e (2.0023), are indicative of the presence of Ni(I) in this state of the enzyme.

Discussion

The [FeFe]- and [NiFe]-H₂ases, in their H_ox and Ni-L states, respectively, share several key features. Aside from their similar coordination spheres, the binuclear active sites are both EPR-active, with the 33_e⁻ clusters existing in low-spin (S = ½) configurations. However, while the spin in H_ox is located on the rotated Fe(I) site, in Ni-L it resides on the Ni center. In the latter case Fe is in the +II oxidation state and DFT calculations suggest that it does not adopt a rotated structure.¹⁹ This is unsurprising given that a vacant Fe coordination site is necessary for H_ox to bind H₂,⁷ whereas the [NiFe]-H₂ases (in the Ni-SI_a state) probably activate H₂ by a different mechanism.³¹

The triarylphosphine complexes [2a]⁺ and [2b]⁺ are spectroscopically similar to the Ni(II)Fe(I) complexes reported earlier⁹ in that the ³¹P nucleus of the weakly basic monophosphine couples strongly to the Fe(I) center. Given that Ni-L has a Ni(I)Fe(II) core with strongly donating CN⁻ ligands at the basal Fe coordination sites, improved models were expected to result from the coordination of highly basic phosphines to the Fe center. It was necessary to employ bulky alkylphosphines, as we have found that phosphines that are both basic and small (e.g. PMe₃, P(n-Bu)₃) trigger disproportionation-type reactions when interacted with [1]⁺.⁹ These new Ni(II)Fe(I) complexes, also of formula [(dxpe)Ni(pdt)Fe(CO)₂L]⁺, differ from those previously described in terms of their structure and spectroscopy. It is proposed that the magnitude of the ³¹P hyperfine interaction is highly sensitive to the position of the monophosphine and is greatest in the case when L is apically bound, where overlap with the Fe-centered SOMO is maximized. In turn, the position of this ligand is governed by its basicity. While triarylphosphines are bound apically, complexes of more basic ligands such as PHCy₂ are thought to be more structurally distorted. This effect, perhaps caused by repulsion between the P lone pair and Fe-centered SOMO, serves to decrease overlap between the electron in the latter and the ³¹P nucleus. In the extreme case of PCy₃, complexes featuring a highly distorted Fe(CO)₂(PR₃) site are afforded, for which no hyperfine interactions were observed. Indeed, the smallest Ni-Fe-C angle in [1e]⁺ (75°) and, to a lesser extent, that in [2c]⁺ (84°), suggests that Fe exists in a ‘rotated’ coordination geometry, a theme prevalent in mixed-valence Fe(II)(pdt)Fe(I) complexes.^17,21 For example, in [(IMes)(CO)₂Fe(pdt)Fe(CO)₂PMe₃]⁺, the smallest C-Fe-Fe angle is even more acute (57°). It is likely that the ‘rotated’ Fe stereochemistry found for [1e]⁺ and [2c]⁺ results from the presence of strongly σ-donating ligands, although the large steric profile of the trialkylphosphines employed (cone angles: P(i-Pr)₃ 160°, PCy₃ 170° vs PPh₃ 145°)³³ may also play a role.

The molecular structure of [1e]⁺ calculated by DFT closely matches that obtained by X-ray crystallography. Indeed, both indicate a ‘rotated’ structure for the Fe(I) site, with the PCy₃ ligand occupying a pseudo-apical position, despite its steric bulk. The calculations also allowed for characterization of the electronic structure, which features a Fe-centered SOMO with substantial d(z²)-character. Importantly, the orbital is roughly orthogonal to the Fe-P bond such that it is unaffected by the ³¹PCy₃ nucleus, which accounts for the lack of hyperfine splitting in the experimental EPR spectra. While spectra of the ¹³CO-labeled complex [1e′]⁺ suggested the ¹³CO ligands could be equivalent, upon consideration of the X-ray structure, it is likely that the they are fact inequivalent, and give rise to a pseudo-triplet owing to the A(¹³C) values being similar in magnitude. For the diiron model [(IMes)(CO)₂Fe(pdt)Fe(CO)₂PMe₃]⁺, single-point spin-restricted (ROB3LYP) calculations predicted significant spin density on the rotated Fe center, consistent with its assignment as Fe(I).³⁴ As is the case with [1e]⁺, the SOMO appears to have significant d(z²)-character, as might be expected for a low-spin, square-pyramidal d⁷ system.

Despite their heterobimetallic nature, [1e]⁺ and [2c]⁺ can be considered models for H_ox. This state of [FeFe]-H₂ase is thought to adopt a rotated structure to reduce electronic asymmetry by virtue of the semi-bridging CO, onto which spin may be delocalized. The rotated Fe(CO)₂(PCy₃) centers in these models structurally mimic the distal Fe(CN)(CO)₂ unit present in the [FeFe] enzyme, although the semi-bridging CO is not within bonding distance of Ni.

The relevance of complexes [1e]⁺ and [2c]⁺ to H_ox is confirmed not only crystallographically, but also spectroscopically. As with the triarylphosphine complexes, the g-shifts obtained are not dissimilar to those of H_ox (Table 2), although the signals are considerably more rhombic. Indeed, the values obtained for the PCy₃ complex [1e]⁺ (2.089, 2.036, 2.008), when compared to those for [(dppe)Ni(pdt)Fe(CO)₂PPh₃]⁺ (2.066, 2.036, 2.006),⁹ indicate significant g-strain and movement away from idealized C_s symmetry (i.e. by ‘rotation’) in the former case. The lack of ³¹P hyperfine coupling in these ‘rotated’ models is consistent with the DFT calculations in that of the SOMO is orientated towards a vacant coordination site, as it is in H_ox.

The description of new derivatives as Ni(II)Fe(I) complexes is further corroborated upon consideration of the effects of phosphine substitution on their electrochemical properties. Variation of the diphosphine greatly alters the [(dxpe)Ni(pdt)Fe(CO)₂L]^0/+ couple, whereas the [(dxpe)Ni(pdt)Fe(CO)₂L]^+/2+ couple is more strongly affected by the monophosphine L. Thus, the data are consistent with the oxidations Ni(I)Fe(I) → Ni(II)Fe(I) → Ni(II)Fe(II), a key finding being that oxidation to afford Ni(II)Fe(II) species can be reversible and occurs at relatively mild potentials. In contrast to the substituted derivatives, the tricarbonyl dications are unstable, as evidenced by the irreversibility of the couples [1]^+/2+ and [2]^+/2+.²⁹ These results emphasize the role of terminal ligands on the Fe electronic structure and are significant in that the electrochemically generated substituted dications represent the closest approaches to the 32e⁻ Ni-SI_a state of [NiFe]-H₂ase. Indeed, these models have highly electrophilic Ni(II)Fe(II) cores, a key feature that allows for the heterolytic activation of H₂ in Nature. Moreover, dicationic models bearing 5-coordinate Fe centers would differ from previously reported Ni(II)Fe(II) dithiolates, in which this center is coordinatively saturated.^35,36

Summary

New examples of mixed-valence (S = ½) nickel-iron dithiolates of formula [(diphosphine)Ni(dithiolate)Fe(CO)₂L]⁺ have been prepared and fully characterized. Their unambiguous description as Ni(II)Fe(I) species, achieved experimentally and by DFT, contrasts the Ni(I)Fe(II) core of Ni-L. However, the new Fe-centered radicals (‘inverse’ Ni-L models) bear remarkable similarity to H_ox. While weakly basic ligands (L = triarylphosphine) in the mixed-valence complexes occupy the apical Fe site and participate in strong ³¹P-Fe(I) hyperfine interactions, more basic monophosphines (L = trialkylphosphine) do not give rise to such coupling. Significantly, structural and DFT studies on complexes in the latter class allow for this to be rationalized in terms of the ‘rotation’ of the Fe fragments. Thus, in these compounds bearing strong σ-donors, overlap between the Fe-centered SOMO and the monophosphine is avoided by adoption of the rotated structure. The distortion in the Fe coordination sphere is a key feature of the distal Fe site in the H_ox state of [FeFe]-H₂ase. Reproducing the NiS₂Fe core present in the [NiFe]-H₂ases, while also mimicking the spectroscopy and ‘rotated’ structure of the [FeFe]-H₂ases, complexes of the present type represent unprecedented hybrid models, from which new parallels between the different H₂ases can be drawn.

Experimental Section

Unless otherwise stated, chemicals were purchased from commercial sources and used as received. The compounds 1²⁴ and 2²⁹ and FcBAr^F₄³⁷ were prepared according to the literature methods. All reactions were conducted in an MBraun glovebox equipped with a solvent purification system; the concentrations of O₂ and H₂O in the N₂ atmosphere were less than 1 ppm. The mixed-valence salts were stored at −28°C. IR spectra of complexes (in CH₂Cl₂) were recorded on a Perkin-Elmer Spectrum 100 FTIR spectrometer. EPR spectra of complexes (~1 mM in CH₂Cl₂/PhMe, 1:1) were recorded on either a Varian E-line 12″ Century Series X-band or a 15″ Q-band CW spectrometer. ESI-MS data of compounds in CH₂Cl₂ were acquired using a Waters Micromass Quattro II spectrometer. Cyclic voltammetry experiments were carried out in a one compartment glass cell using a CH Instruments CHI600D electrochemical analyzer. The working, counter and pseudo-reference electrodes were glassy carbon, platinum and silver, respectively. The analyte (1 mM) and NBu₄PF₆ (100 mM) were dissolved in CH₂Cl₂ and potentials (reported here relative to internal Fc/Fc⁺) were swept at 0.1 Vs⁻¹. Analytical data were acquired using an Exeter Analytical CE-440 elemental analyzer. UV-vis data were acquired on a Varian Cary 50 Bio spectrophotometer. Crystallographic data were collected using either a Bruker X8 ([1e]BF₄) or a Siemens SMART diffractometer ([2c]BF₄), each of which was equipped with a Mo K_α source (λ = 0.71073 Å) and an Apex II detector.

Phosphine-substituted derivatives ([1a-e]BF₄ and [2a-c]BF₄)

[(dxpe)Ni(pdt)Fe(CO)₃] (20 μmol) and FcBF₄ (20 μmol) were dissolved in CH₂Cl₂ (2 mL) with rapid stirring. After 1 min the solution was added dropwise to the appropriate phosphine (100 μmol) in CH₂Cl₂ (0.5 mL) with stirring. After 0.5 min, pentane (−28°C, 15 mL) was added and the mixture allowed to stand at −28°C for 1 h. The solids were isolated by filtration, washed with pentane (−28°C, 2 × 2 mL) and dried briefly to afford the respective phosphine complexes. The ¹³CO derivatives [1d′]BF₄ and [1e′]BF₄ were prepared analogously using [(dppe)Ni(pdt)Fe(¹³CO)₃]⁹ as the precursor.

[(dppe)Ni(pdt)Fe(CO)₂PHCy₂]BF₄ ([1a]BF₄). Yield: 74%, brown powder. ESI-MS: m/z 872.3 [M – BF₄⁻]⁺. Anal. Calcd for C₄₃H₅₃BF₄FeNiO₂P₃S₂·0.25CH₂Cl₂: C, 52.93; H, 5.49; N, 0.00. Found: C, 52.77; H, 5.52; N, 0.00.

[(dppe)Ni(pdt)Fe(CO)₂PPh(NEt₂)₂]BF₄ ([1b]BF₄). Yield: 69%, olive powder. ESI-MS: m/z 926.1 [M – BF₄⁻]⁺. Anal. Calcd for C₄₅H₅₅BF₄FeNiN₂O₂P₃S₂·0.5CH₂Cl₂: C, 51.71; H, 5.34; N, 2.65. Found: C, 51.79; H, 5.39; N, 2.68.

[(dppe)Ni(pdt)Fe(CO)₂P(NMe₂)₃]BF₄ ([1c]BF₄). Yield: 66%, olive powder. ESI-MS: m/z 836.8 [M – BF₄⁻]⁺. Anal. Calcd for C₃₇H₄₈BF₄FeN₃NiO₂P₃S₂·0.67CH₂Cl₂: C, 46.07; H, 5.06; N, 4.28. Found: C, 46.17; H, 5.01; N, 4.49.

[(dppe)Ni(pdt)Fe(CO)₂P(ⁱPr)₃]BF₄ ([1d]BF₄). Yield: 75%, olive powder. ESI-MS: m/z 834.1 [M – BF₄⁻]⁺, 806.2 [M – CO – BF₄⁻]⁺. Anal. Calcd for C₄₀H₅₁BF₄FeNiO₂P₃S₂: C, 52.09; H, 5.57; N, 0.00. Found: C, 51.65; H, 5.72; N, 0.10.

[(dppe)Ni(pdt)Fe(¹³CO)₂P(ⁱPr)₃]BF₄ ([1d′]BF₄). Yield: 70%, olive powder. ESI-MS: m/z 836.3 [M – BF₄⁻]⁺. Anal. Calcd for C₃₈¹³C₂H₅₁BF₄FeNiO₂P₃S₂·0.25CH₂Cl₂: C, 51.13; H, 5.49; N, 0.00. Found: C, 51.16; H, 5.59; N, 0.00.

[(dppe)Ni(pdt)Fe(CO)₂PCy₃]BF₄ ([1e]BF₄). Yield: 94%, green powder. ESI-MS: m/z 954.3 [M – BF₄⁻]⁺, 926.4 [M – CO – BF₄⁻]⁺. Anal. Calcd for C₄₉H₆₃BF₄FeNiO₂P₃S₂·0.67CH₂Cl₂: C, 54.27; H, 5.90; N, 0.00. Found: C, 54.49; H, 6.14; N, 0.26.

Green hexagonal single crystals were grown by layering a concentrated CH₂Cl₂ solution with pentane and allowing the mixture to stand at –28 °C. One crystal (0.322 × 0.197 × 0.054 mm) was subjected to X-ray diffraction at 193 K. Its space group was determined to be trigonal P-3 with cell parameters: a 20.768 Å, b 20.768 Å, c 27.300 Å, α 90°, β 90°, γ 120°. Integration of 5773 reflections and solution by direct methods using SHELXTL V6.12^38,39 afforded a model with R₁ = 0.0631 and wR₂ = 0.1684.

[(dppe)Ni(pdt)Fe(¹³CO)₂PCy₃]BF₄ ([1e′]BF₄). Yield: 73%, green powder. ESI-MS: m/z 956.2 [M – BF₄⁻]⁺, 673.3 [M – ¹³CO – PCy₃ – BF₄⁻]⁺. Anal. Calcd for C₄₉ ¹³C₂H₆₃BF₄FeNiO₂P₃S₂·0.75CH₂Cl₂: C, 53.92; H, 5.87; N, 0.00. Found: C, 53.93; H, 5.88; N, 0.00.

[(dcpe)Ni(pdt)Fe(CO)₂PPh₂(2-py)]BF₄ ([2a]BF₄). Yield: 80%, yellow powder. ESI-MS: m/z 961.3 [M – BF₄⁻]⁺, 933.3 [M – CO – BF₄⁻]⁺. Anal. Calcd for C₄₈H₆₈BF₄FeNNiO₂P₃S₂·0.25CH₂Cl₂: C, 47.38; H, 6.45; N, 1.31. Found: C, 54.18; H, 6.12; N, 1.36.

[(dcpe)Ni(pdt)Fe(CO)₂PPh₃]BF₄ ([2b]BF₄). Yield: 79%, yellow powder. ESI-MS: m/z 960.8 [M – BF₄⁻]⁺. Anal. Calcd for C₄₉H₆₉BF₄FeNiO₂P₃S₂·3.25CH₂Cl₂: C, 47.38; H, 5.75; N, 0.00. Found: C, 47.43; H, 5.87; N, 0.00.

[(dcpe)Ni(pdt)Fe(CO)₂PCy₃]BF₄ ([2c]BF₄). Yield: 62%, green powder. ESI-MS: m/z 978.2 [M – BF₄⁻]⁺, 950.2 [M – CO – BF₄⁻]⁺, 670.2 [M – PCy₃ – CO – BF₄⁻]⁺. Anal. calcd for C₄₉H₈₇BF₄FeNiO₂P₃S₂·2.5CH₂Cl₂: C, 48.36; H, 7.25; N, 0.00. Found: C, 48.19; H, 7.30; N, 0.00.

Green prismatic single crystals of [2c]BF₄·CH₂Cl₂·0.5 pentane were grown by layering a concentrated CH₂Cl₂ solution with pentane and allowing the mixture to stand at −28 °C. One crystal (0.514 × 0.133 × 0.042 mm) was subjected to X-ray diffraction at 193 K. Its space group was determined to be monoclinic P2₁/n with cell parameters: a 21.873 Å, b 10.784 Å, c 26.362 Å, α 90°, β 107.63°, γ 90°. Integration of 3186 reflections and solution by direct methods using SHELXTL V6.12^38,39 afforded a model with R₁ = 0.1217 and wR₂ = 0.2317.

Calculations

Calculations of structural parameters and the electronic structure were performed using ORCA.⁴⁰ Full geometry optimizations were performed using the B3LYP⁴¹ and BP86^41,42 exchange-correlation functionals and a triple-ζ basis set with polarization functions that were obtained from the TURBOMOLE library.⁴³ This combination of exchange correlation functional and basis set was shown to give accurate structural parameters. In addition, single-point calculations using the hybrid B3LYP⁴¹ and PBE0^45,46 functionals were carried out on the BP86/TZVP geometry optimized structures. IR spectra were generated by numerically calculating second derivatives; calculations of g-tensors were performed using an effective mean-field spin-orbit coupling operator, with the center-of-mass as the origin of the g-tensor.⁴⁷ Additional g- and A-tensor calculations were performed with ADF^48,49 using the zero order regular approximation (ZORA)⁵⁰ for relativistic effects and a self-consistent inclusion of spin-orbit coupling. A Slater-orbital DZ basis set was used for spin-restricted g-tensor calculations⁵¹ and a TZP basis set for spin-unrestricted scalar relativistic hyperfine coupling tensor calculations.^52,53

Scheme 2.