Investigating Metal–Metal Bond Polarization in a Heteroleptic Tris-ylide Diiron System

Abstract

This article describes the synthesis, characterization, and S-atom transfer reactivity of a series of C_3v-symmetric diiron complexes. The iron centers in each complex are coordinated in distinct ligand environments, with one (Fe_N) bound in a pseudo-trigonal bipyramidal geometry by three phosphinimine nitrogens in the equatorial plane, a tertiary amine, and the second metal center (Fe_C). Fe_C is coordinated, in turn, by Fe_N, three ylidic carbons in a trigonal plane, and in certain cases, by an axial oxygen donor. The three alkyl donors at Fe_C form through the reduction of the appended N=PMe₃ arms of the monometallic parent complex. The complexes were studied crystallographically, spectroscopically (NMR, UV-Vis, and Mössbauer), and computationally (DFT, CASSCF) and found to be high-spin throughout, with short Fe–Fe distances that belie weak orbital overlap between the two metals. Further, the redox nature of this series allowed for the determination that oxidation is localized to the Fe_C. S-atom transfer chemistry resulted in the formal insertion of a S atom into the Fe–Fe bond of the reduced diiron complex to form a mixture of Fe₄S and Fe₄S₂ products.

Graphical Abstract

Spectroscopic and computational analyses on a neutral diiron complex supported by an N₄ binding pocket for one metal (Fe_N) and a C₃ binding pocket for the other (Fe_C) supported assignments of Fe_N(II) and Fe_C(I) physical oxidation states. The chemical oxidation of this complex occurred at Fe_C, with the Fe–Fe bond polarity governed by axial, not equatorial, interactions. Treatment of the neutral complex with Ph₃SH led to a mixture of Fe₄S_x (x = 1,2) clusters.

INTRODUCTION

From enzymatic systems to heterogeneous catalysts, the study of metal-metal interactions has far reaching implications and provides insight into the activity and properties of myriad synthetic and naturally occurring systems.^1–4 Since Cotton’s seminal reports on Re-based metal-metal bonding,⁵ bimetallic complexes have emerged as an important area of synthetic inorganic chemistry, due to their ability to perform cooperative multi-electron transformations. Bimetallic 3d transition metal complexes containing metal-metal bonds are few in comparison to their heavier congeners, partly due to the diminished orbital overlap between the smaller 3d manifolds,⁶ but the potential for performing multi-electron chemistry with two metals that readily undergo single-electron transfer, as is the case for 3d transition metals, motivates a desire to develop ligands capable of controlling the coordination environments about dinuclear cluster cores.

The vast majority of diiron complexes contain carbonyl or thiolate bridges and are not constructed with ligands that support and control the bimetallic core. Of the more systematically designed complexes, a small subset exhibit C₃-symmetry, typically containing nitrogen and/or phosphorus donors,⁷ with one metal serving as a metalloligand that has been used to modulate the electronics of a second, reactive metal. This approach has led to interesting electronic structures for the complexes as the Fe–Fe interaction is tuned by the nature of the supporting ligand framework.

The most well characterized examples of these trigonal diiron complexes are shown in Figure 1.^8–12 All share short Fe–Fe distances (2.20–2.46 Å) and high-spin electron configurations (5-7 unpaired electrons), despite the varied ligand donor types. For a symmetric complex in this trigonal field, the entire d-manifold may participate in Fe–Fe bonding, leading to the possibility of one σ-, two π-, and two δ-symmetry bonds. Even in the highest symmetry cases, however, the bond orders remain fairly low, with limited Fe–Fe multiple bonding due to occupation of the π*/δ* manifolds. The complexes of this type are primarily differentiated by the extent of metal-metal bond delocalization. With more asymmetric ligand sets, the energy gap between the d-manifolds increases, leading to more polarized σ bonding and little to no π/δ bonding.⁷ It should be noted that the ligand types do not solely determine the degree of localization.⁹ However, across a series of homologous diiron complexes, ligand effects are expected to determine the amount of Fe–Fe bond polarization.

Figure 1.

Literature examples of C₃-symmetric diiron complexes.

Herein, we describe the synthesis and characterization of axially pseudo-trigonal diiron complexes that result from the unexpected methyl group C–H activation of P₃tren, a phosphinimine-substituted tris(2-aminoethyl)amine ligand. Previously, we have used the zwitterionic character of the phosphinimines in this ligand system to create a cationic electrostatic field in the secondary coordination sphere that is capable of tuning the valence manifolds of Cu^I coordination complexes.¹³ During the course of related explorations of P₃tren iron complexes targeted at making use of the C₃-symmetric binding pocket for small molecule activation chemistry, a chemical reduction led to an unexpected diiron product with an unusual trialkyl coordination environment about one Fe center.

We have used spectroscopic and computational methods to evaluate the effect of the differing primary coordination spheres of the two Fe centers (N₄Fe vs. C₃(O)Fe) on the metal-metal bonding relative to related diiron complexes. One of these complexes was then investigated for its ability to reduce a sulfur-atom transfer reagent to form bridging sulfide products.

RESULTS AND DISCUSSION

Synthesis and characterization.

The treatment of FeCl₂ with P₃tren in acetonitrile led to the clean formation of the mononuclear, trigonal bipyramidal complex [(P₃tren)FeCl][Cl] (1, Scheme 1) in 95% yield as a white solid. Crystallographic analysis revealed the complex to be a 1:1 ion pair with an outer sphere chloride. The cation was found to exhibit local C_3v symmetry in the solid state, with unexceptional bond metrics (Table 1) considering the characterization of the metal center as high-spin (S = 2) Fe^II (solution phase, Evans’ method).

Scheme 1.

Synthesis of complexes 1-4

Table 1.

Selected and averaged bond distances for all compounds.

Compound	Fe-Nax (Å)	Fe-Fe (Å)	P-N (Å)	Fe-Cl (Å)	Fe-C (Å)	Fe-Neq (Å)	FeN-S (Å)	FeC-S (Å)	Fe-O (Å)
1	2.334(1)	-	1.596(2)	2.4761(5)	-	2.075(1)	-	-	-
2	2.279(2)	-	1.627(1)	2.2852(6)	-	1.995(2)	-	-	-
3	2.258(1)	2.2808(3)	1.612(2)	-	2.126(2)	2.028(2)	-	-	-
[4-OTf]0	2.241(8)	2.538(2)	1.611(9)	-	2.129(1)	2.012(8)	-	-	2.264(8)
[4-THF]+	2.218(2)	2.5537(4)	1.617(2)	-	2.115(3)	2.009(2)	-	-	2.377(2)
5	2.427(4)	2.6908(8) a 3.6188(8) b	1.603(4)	-	2.118(5)	2.073(4)	2.443(1)	2.301(1)	-
6	2.374(6)	3.485(1)	1.606(5)	-	2.157(7)	2.095(5)	2.446(2)	2.354(1)	-

^aRefers to the Fe-Fe distance within the Fe-S core.

^bRefers to the Fe_N-Fe_C distances.

A cyclic voltammogram of 1 revealed one reversible oxidation at −0.6 V vs. Fc^0/+ (see Supporting Information). Upon chemical oxidation with excess AgOTf, the high-spin Fe^III complex [(P₃tren)FeCl][OTf]₂ (2, Scheme 1) was formed as a bright orange solid. Crystallographic characterization of 2 revealed a slight elongation of the P–N bond lengths compared to those in 1 and a ~0.2 Å decrease in the Fe–Cl bond distance. Similarly, the Fe–N_ax distance decreased from 2.334(1) Å for 1 to 2.279(2) Å; both values are within the range of previously reported Fe(II)/Fe(III) TBP complexes.¹⁴ Paramagnetically shifted resonances between 150 and 0 ppm were observed in the ¹H NMR spectra for both complexes (see Supporting Information), though unequivocal assignments could not be made from these data.

While the only electrochemical feature observed by cyclic voltammetry within the solvent window of MeCN for 1 was a reversible one-electron oxidation, the addition of 2 equiv of KC₈ to 1 in THF resulted in a color change to deep green along with gas evolution. Optimization of the synthetic procedure found that the use of toluene as a solvent with 4 equiv of KC₈ led to reproducible crystallization of large green blocks of a new product in 30-40% yield following crystallization from concentrated hexane solutions at −20 °C. In an attempt to account for the Fe-based stoichiometry of the reaction, an additional equivalent of FeCl₂ was introduced prior to the reductant. Doing so had a marginal effect on the outcome of the reaction, increasing the yield to a maximum of 42%. The use of either 3 or 5+ equiv of KC₈ was not found to lead to alternative, isolable products but instead affected the yield of 3. A dark-brown, insoluble material forms in these reactions, and the only species observed in solution were 3, P₃tren, and PMe₃.

A crystal structure of the reduction product 3 (Figure 2) revealed an unexpected C–H activation on each of the PMe₃ groups to give a diiron complex with three alkyl-like CH₂-coordinated ylide ligands on the “top” iron (Fe_C) and the N₄ coordination environment of the tren backbone on the “bottom” Fe (Fe_N, Scheme 1). Repeating the reaction in a sealed tube allowed us to identify by ¹H NMR spectroscopy that H₂ was a byproduct of this reaction. Despite the relatively low yields, no other organic or organometallic byproducts were observed during the reduction, and analytically pure material was isolated following crystallization.

Figure 2.

Thermal ellipsoid plots of complexes 1, 3, [4-OTf]⁰, and [4-THF]⁺. Thermal ellipsoids are shown at the 50% probability level. Hydrogen atoms, counter anions, solvates and disordered atoms are omitted for clarity.

A close analysis of the crystallographic data revealed important features of the coordination environments about the metal centers. Fe_N has a distorted trigonal bipyramidal geometry enforced by the tetradentate ligand and the second iron atom. The Fe–Fe distance of 2.2808(3) Å is short but not unprecedented, and it is consistent with some degree of metal-metal bonding. The formal shortness ratio (FSR) for this interaction is 0.97, which is indicative of a M–M single bond.¹⁵ Fe_N lies slightly above the plane created by the phosphinimine nitrogens, with the N_P–Fe–N_P angles summing to 354°. Fe_C exhibits a similarly idealized trigonal planar environment, with the sum of C–Fe–C angles totaling 358°. If the Fe–Fe interaction is taken into account, Fe_C has a trigonal pyramidal geometry with a τ₄ value of 0.857 (ideal trigonal pyramid τ₄ is 0.851).¹⁶

Electrochemical investigation of the redox properties of 3 using cyclic voltammetry revealed two ostensibly reversible oxidation features at −0.70 V and −2.06 V. An additional feature was observed at ca. −1.0 V, but its intensity was inconsistent over multiple trials (see Supporting Information). Chemical oxidation of 3 with [Cp₂Co][OTf] led to the formation of a one-electron oxidation product: [4-THF]⁺ or [4-OTf]⁰, depending on the crystallization solvent (THF and toluene, respectively; Scheme 1). Both are largely isostructural to 3, except for an added axial oxygen-based donor (THF or OTf) to Fe_C. [4-THF]⁺ and [4-OTf]⁰ display identical spectroscopic features when dissolved in non-coordinating solvents, signifying the facile formation of [4-OTf]⁰ from [4-THF]⁺ through a weakly bound THF ligand. When dissolved in THF, the triflate appeared to be coordinated or in equilibrium with the solvent, based on the broadened and shifted triflate signal in the ¹⁹F NMR spectrum (see Supporting Information). The Fe–Fe distance in [4-THF]⁺ elongated to 2.5537(4) Å (FSR 1.10), indicating a weakening of the metal-metal interaction. This expansion occurred at the expense of the Fe_N-N_ax distance, which decreased from 2.258(1) Å in 3 to 2.218(2) Å in [4-THF]⁺. The rest of the analysis will be based on [4-OTf]⁰, due to the similar NMR and UV-Vis spectroscopic data and similar computational results for it and [4-THF]⁺.

[4-OTf]⁰ was found to be paramagnetic in solution, owing to a ¹H NMR spectrum that shows three broad signals between 80 and 350 ppm. The solution-phase effective magnetic moment was determined to be 9.1 ± 0.1 μ_B (Evans’ method), corresponding to an S = 4 spin state with modest contributions from spin-orbit coupling. These data may be compared to those for complex 3, which was also found to be paramagnetic in solution. Two broad signals were visible in its ¹H NMR spectrum, and the solution-phase effective magnetic moment was determined to be 8.2 ± 0.1 μ_B (Evans’ method), which is most readily interpreted as resulting from an S = 7/2 system, again with a modest amount of spin-orbit coupling. These spin state values track closely with those observed for complexes C and D (isoelectronic to 3) and B (isoelectronic to [4-OTf]⁰).

DFT Modelling.

Density functional theory (DFT) calculations were used to model the structures of complexes 1-4, using the structural and magnetochemical data for guidance. The geometry optimized structures for the cationic portions of complexes 1 and 2 closely matched the experimental data, and their electronic structures were consistent with high-spin Fe(II) and Fe(III) centers, respectively, in trigonal bipyramidal ligand fields. As expected, the Fe–N_eq/Fe–N_ax distances contract on oxidation. The experimental data show a contraction of ca. 0.06 Å for each type of interaction, consistent with the decrease in ionic radius following an oxidation event that involves depopulation of a primarily non-bonding orbital. The most notable deviation between the experimental and computed bond lengths was found in the Fe–N_ax distance for 2, for which the predicted value of 2.102 Å was 0.177 Å shorter than the experimental distance of 2.279(2) Å.

The electronic structures of the dinuclear complexes 3 and [4-OTf]⁰ were more complex than those of the mononuclear species and invited comparison to related C₃-symmetric diiron systems that have been described in the literature (Figure 1). All known C₃-symmetric diiron systems have N- or P-atom donors coordinated to the iron centers, making the carbon-based ligands in compounds 3 and [4-OTf]⁰ unique and worth further investigation. Not only are C-based donors rare in diiron complexes, but even mononuclear iron complexes coordinated by three alkyl groups are scarce, with only a handful of known examples.^17–21 Notably, Smith and co-workers recently described several iron complexes coordinated by bis(ylide)diphenylborate ligands, which feature chelation of two ylides per chelate,²² and Neidig and coworkers have provided recent examples of homoleptic 3- and 4-coordinate iron alkyl complexes.^{17, 21} Both the ylide and phosphinimine ligand sets in 3 and [4-OTf]⁰ are expected to serve as strong σ-donors.^{23, 24} Related ylides have actually been shown to be even stronger donors than both NHCs and phosphinimines. This was demonstrated by César and co-workers, who synthesized a series of rhodium carbonyl complexes bound by various neutral donors that included a phosphinimine, an NHC, and an ylide. The latter showed the most activated CO stretching frequency from the series.²⁵ We thus sought to investigate the effect of these ligand environments around each iron on the complexes’ spectroscopic properties and extents of M–M bonding, with a particular interest in the ability of ligand inequivalence to create spectroscopic inequivalence and polarization of the metal-metal interaction.

The DFT-optimized geometry of 3 closely matched the experimental data for this neutral diiron species. The optimized structure retained the pseudo-C_3v geometry observed crystallographically, and the Fe–C and Fe–N distances all lay within 2% of the experimental values. DFT calculations also reproduced the Fe–Fe distance, yielding a calculated value of 2.265 Å, within 0.02 Å of the experimental distance. DFT was similarly robust in predicting the Fe–Fe (−0.039 Å), Fe–C, and Fe–N_eq distances in [4-OTf]⁰ but lagged for the Fe–O (−0.124 Å) and Fe–N_ax (+0.120 Å) distances. For [4-THF]⁺, the Fe–O distance (+0.015 Å), Fe–N_ax (+0.09 Å), and Fe–Fe (−0.066 Å) distances were all predicted with excellent agreement to the experimental data.

UV-vis-NIR and Mössbauer Spectroscopic Characterization.

The UV-Vis-NIR absorption data for complexes 1-[4-OTf]⁰ provided useful insight into the electronic structures of each (see Figure 3 and the Supporting Information). Both 1 and [4-OTf]⁰ are nearly featureless in the visible region, while 2 and 3 have several intense visible transitions, as expected based on their colors. In the NIR region, where we would expect to see features that result from both inter- and intra-metal transitions, the mononuclear HS Fe^II complex 1 shows low-intensity d-d transitions (ε < 30 M⁻¹cm⁻¹), while the HS Fe^III complex 2 is featureless. Compounds 3 and [4-OTf]⁰ exhibited more intense features in this region, which may similarly reflect intrametal d-d transition or transitions within a weakly interacting Fe₂ bonding manifold. 3 has a weak absorption at 1044 nm (ε = 157 M⁻¹cm⁻¹), and [4-OTf]⁰ exhibits a comparable feature at 1410 nm (ε = 164 M⁻¹cm⁻¹) (Figure 3). These absorption bands are similar to the features observed by Lu for a series of complexes bearing the ligand in C (Figure 1). Dunbar and coworkers also observed features in the same range for a series of thiolate bridged dirhodium complexes.²⁶ They characterized those transitions as intervalence charge transfer (IVCT) bands; however, their observed extinction coefficients are nearly an order of magnitude larger than those for 3 and [4-OTf]⁰. TD-DFT computational analyses were not able to accurately reproduce the experimental results, likely due to the significant ground state static correlation in the Fe–Fe interaction (see below), limiting our ability to model the origin of these features.(Figure 4)

Figure 3.

NIR spectra of 1, 2, 3, and [4-OTf]⁰.

Figure 4.

⁵⁷Fe Mössbauer spectra of 1 (left) and [4-OTf]⁰ (right) collected at 80 K with a parallel 50 mT applied field. Experimental data are shown with open circles; simulations are shown in solid red and dashed lines. The difference between the experimental data and the simulations are given with grey lines.

Mössbauer spectroscopic data were collected at 80 K. The experimental Mössbauer spectrum of 1 showed a doublet at 1.00±0.03 mm/s (|ΔE_Q| =2.57±0.06 mm/s). This isomer shift is typical of high-spin Fe(II) and was used to aid in validating our computationally predicted isomer shift data (Table 2). The DFT calculated isomer shift for 1 was determined to be 0.98 mm/s (|ΔE_Q| = 2.70 mm/s), in good agreement with the experimental data.

Table 2.

Experimental and calculated Mössbauer parameters, electron count, and oxidation state assignment information.

Complex		δ (mm/s)a	|ΔEQ| (mm/s)b	Total d-Electron Count	EANc
1	Exp Calc	1.00 0.98	2.57 2.70	6	16
2	Exp Calc	~0 0.45	– 0.78	5	15
3 (FeC, FeN)	Exp Calc	~0.5, ~0.8 0.47, 0.70	~1.4, ~0.8 0.27, 0.73	13	27
[4-OTf]0 (FeC, FeN)	Exp Calc	0.42, 0.62 0.52, 0.73	1.46, 0.99 1.00, 0.67	12	28
[4]+ (FeC, FeN)	Calc	0.35, 0.66	0.99, 2.44	12	26
A	Exp	0.65	0.32	13	25
B (FeN, Fepy)	Exp	0.48, 0.58	1.31, 0.38	12	28
C (FeN,FeP)	Exp	0.42, 0.55	0.13, 0.12	13	27
D (FeN, FeP)	Exp	0.56, 0.44	1.11, 1.41	13	27

^aEstimated uncertainty of ±0.03 mm/s.

^bEstimated uncertainty of ±0.06 mm/s.

^cEffective atomic number.

Compared to the well-defined doublet for 1, the spectra for 2 and 3 were broadened and showed small quadupole splitting values that complicated the fits to the experimental data (see Supporting Information). This quality is common for half-integer-spin iron complexes under no/low field conditions.¹⁰ Despite our inability to extract quantitative data from the spectra for 2 and 3, qualitative isomer shift and quadrupole splitting information can be determined to a degree that allows for useful comparisons both across the series and to the computational data.

The experimental spectrum of 2 shows a single, broad feature at ~0 mm/s. The observation of only a single feature in a no/low field Mössbauer spectrum indicates the presence of minimal quadrupole splitting. The DFT calculated isomer shift for 2 was determined to be 0.46 mm/s (|ΔE_Q| = 0.78 mm/s), which is consistent with oxidation of the metal center to Fe(III). The poor quantitative agreement with the experimental data may be explained in part by the use of the dicationic unit of 2 in the computational experiments. Holland and co-workers have recently demonstrated the sensitivity of ⁵⁷Fe Mössbauer features to the electrostatic environment created by the specific locations of counterions.²⁷ Furthermore, the low quadrupole splitting that is apparent from the experimental data is consistent with the presence of a minimal electric field gradient about the Fe nucleus that commonly results from a high-spin S = 5/2 electron configuration.

All complexes in Figure 1, except the symmetric complex A, have been reported to display distinct quadrupole doublets by Mössbauer spectroscopy, indicating that the metal centers in these homobimetallic systems are spectroscopically differentiable. For example, while the Mössbauer spectrum for A exhibits a single doublet at 0.65±0.03 mm/s for the symmetric Fe centers, the isoelectronic but dissymmetric species C has isomer shifts at 0.55±0.03 (Fe_P) and 0.42±0.03 (Fe_N) mm/s. The accumulated Mössbauer data for complexes A-D are listed in Table 2.

As was the case with 2, the Mössbauer data for 3 were found to be broad with poor signal-to-noise ratios (see Supporting Information, Figure S22), despite extended data collection times. Still, the spectrum for 3 did reveal two apparent doublets and could be assigned approximate values when informed by the computational data given below. This result is consistent with the presence of inequivalent iron environments, though we acknowledge that other fits to the data may be possible. These features were identified at ~0.8 and ~0.5 mm/s (|ΔE_Q| = ~0.8, ~1.4 mm/s, respectively), with the former being more positive than the typical range of 0.4–0.6 mm/s for C₃-symmetric diiron systems. We note that the LiCl adduct of C, though less structurally analogous to 3 than C itself, displays isomer shift values of 0.77±0.03 (Fe_P) and 0.50±0.03 (Fe_N) mm/s that are similar to the observed values for 3.¹⁰ The Mössbauer spectrum for [4-OTf]⁰ exhibited two inequivalent, but well resolved, quadrupole doublets, with isomer shifts of 0.62±0.03 and 0.42±0.03 mm/s (|ΔE_Q| = 1.46±0.06, 0.99±0.06 mm/s, respectively). These Mossbauer parameters are similar to those reported for complex B, which is isoelectronic to [4-OTf]⁰ (see Table 2 for d-electron count and effective atomic numbers (EAN)) and displays isomer shifts of 0.58±0.03 (ΔE_Q = 0.38±0.06 mm/s) and 0.48±0.03 (ΔE_Q = 1.31±0.06 mm/s) mm/s.⁹

The computationally informed Mössbauer data for 3 and the unambiguous data for [4-OTf]⁰ indicate both that the isomer shifts move toward more negative values on oxidation and that oxidation of 3 to [4-OTf]⁰ has a more substantive impact on the more positive isomer shift. The former point is consistent with what is typically observed for Werner-type complexes; the isomer shift would be expected to become more negative as the complex is oxidized. This trend results from diminished shielding of the nucleus by the d-manifold on oxidation, thereby allowing for greater nuclear penetration by electron density in the s-manifolds. This is worth noting because this trend can be reversed by changes in spin-state, coordination number, and M–L backbonding, where the shorter M–L bonds lead to higher s-electron density and a more negative isomer shifts.^28–30 Strong M–M interactions also tend to result in more negative isomer shifts, due to decreased d-electron density on Fe and less shielding of the s-manifolds.³¹ These two effects can be in competition, with the oxidation event causing the isomer shift to become more negative and the weakened M–M interaction causing it to become more positive.³² It does not appear, however, that the M–M interaction is a dominant factor in the present case. The typical-to-low quadrupole splitting values are much lower than those observed for some multiply bonded metal complexes (4–5 mm/s).³³ Instead, the more negative isomer shifts for [4-OTf]⁰ compared to those for 3 are consistent with an increase in the oxidation state of the diiron core being the dominant effect and a perturbation of an already weak metal-metal bonding interaction having a lesser impact.

DFT modelling of Mössbauer spectroscopic data.

Still unresolved at this stage is an assignment of which iron center, Fe_N or Fe_P, is giving rise to each of the quadrupole doublets observed in the Mössbauer spectra for 3 and [4-OTf]⁰. To aid in these assignments, we performed extensive computational modelling of our compounds and analogs from the literature. We found that the use of several known computational methods^{34, 35} and their associated fitting parameters led to unsatisfactory correlations with the experimental results. We thus formed a new test set composed of seven C_3v diiron complexes that have reported experimental Mössbauer parameters (see Supporting Information). Their ρ(0) values were calculated using a methodology similar to that developed by the Holland group,³⁴ and a regression analysis yielded a new set of fitting parameters that were then used to calculate isomer shifts for 3 and [4-OTf]⁰.

The computational results suggested that the more positive isomer shift values for 3 and [4-OTf]⁰ (δ = 0.6–0.8 mm/s) are attributable to Fe_N, with the more negative isomer shifts resulting from Fe_C (0.35–0.5 mm/s). Both sets of values lie more positive than what would be expected based on available literature precedent. In related diiron complexes containing an iron center coordinated in a TBP geometry by four nitrogens and an iron metalloligand, the N₄Fe-ligated Fe centers exhibit isomer shifts between 0.42–0.50 mm/s.¹⁰ In the available literature on trialkyl iron species, Neidig and co-workers have characterized homoleptic [FeR₃]⁻ complexes (R = Me, Et) that are structurally related to Fe_C.^{17, 21} Neidig’s Fe(II)-containing ions exhibit trigonal planar geometries with Fe–C distances of ca. 2.06 Å. The zero-field Mössbauer spectra for these salts reveal quadrupole doublets for R = Me at δ = 0.25±0.03 mm/s (ΔE_Q = 1.36±0.06 mm/s) and for R = Et at δ = 0.19±0.03 mm/s (ΔE_Q = 1.43±0.06 mm/s). Without more examples that span a larger oxidation state range, it is difficult to pinpoint a reason for the positive movement in the isomer shifts for both Fe_N and Fe_C in 3 and [4-OTf]⁰, so we next turned to in-depth electronic structure analyses in an effort to parse the origin of the trends in the Mössbauer spectra.

CASSCF modelling.

The DFT computational results used for the modelling the Mössbauer spectroscopic data indicated that the spin state energetics for 3 and [4-OTf]⁰ were consistent with the solution-phase experimental data, which had identified high-spin ground states for both complexes: S_tot = 7/2 for 3 and S_tot = 4 for [4-OTf]⁰. However, the single-determinant nature of DFT prevented evaluation of the extent to which multiple configurations contribute to the ground state electronic structures of these M–M bonded systems. Given the extensive static correlation that is possible for systems with weak-overlap covalent bonds, we employed complete active space self-consistent field (CASSCF) calculations on 3 and [4-OTf]⁰, with the natural orbital output from DFT as a starting point. The active spaces for each complex were chosen to include the valence d-manifolds for each species. Doing so resulted in CAS(13,10) and CAS(12,10) active spaces for 3 and [4-OTf]⁰, respectively; CASSCF natural orbital representations are shown in Figure 5. The ground state of 3 has a leading configuration (61% of the ground state) that is best described as (σ)²(π)²(π)²(π*)¹(π*)¹(σ*)¹(Fe_C-d_x2_-y2)¹(Fe_C-d_xy)¹(Fe_N-d_x2_-y2)¹(Fe_N-d_xy)¹. This is similar to the ground state of C.¹⁰ Here, the δ-symmetry components of the d-manifold are expressed as atomic orbitals, since their half-occupation negates any net bonding interaction. The Fe–Fe formal bond order (FBO) for this leading configuration is 1.5, with the bonding combinations exhibiting polarization toward Fe_C. This polarization is reflected in the antibonding components, which show majority Fe_N character. The secondary configuration accounts for 16% of the ground state wavefunction and formally represents an excitation from π to π*, while the remaining configurations each contribute < 5% to the ground state and result from formal excitations within the β-spin manifold. When the entire ground state is considered, which includes partial occupations of the σ (1.8127), π (3.4409), σ* (1.1170), and π* (2.5541) systems, an effective bond order (EBO) of 0.79 is determined. Applying this same analysis to [4-OTf]⁰ (and [4-THF]⁺, see Supporting Information) revealed that oxidation results in depopulation of the π-bonding manifold. This and inclusion of the axial donor leads to an EBO of ~0, consistent with the substantial 0.26 Å increase in Fe–Fe distance observed experimentally. From here, we used the orthogonal valence bond (OVB) method to gain quantitative predictions of the relative electron densities at each metal center in the dinuclear complexes 3 and [4-OTf]⁰. This method uses an active space that is transformed through an orbital localization procedure into a new set of orbitals that are more easily interpreted by way of a valence bond analysis. This orbital localization yields a new wavefunction that is energetically indistinguishable from the original, making it an equally valid interpretation of the CASSCF results. Upon orbital localization of the active space of 3, an isolated d-manifold on each iron is observed. Examination of the electron distributions and configuration weights reveals that the oxidation states of Fe_N and Fe_C in 3 are best described as Fe(II) (6.04 e⁻) and Fe(I) (6.96 e⁻), respectively. A similar analysis on [4-OTf]⁰ suggests that the oxidation occurs primarily at Fe_C to give an Fe(II)/Fe(II) (6.00 e⁻ for each) oxidation state distribution within the diiron core. Nearly identical results were obtained for [4-THF]⁺ and [4]⁺, the latter of which is a theoretical complex that has no axial ligand coordinated to Fe_C (see Supporting Information).

Figure 5.

Natural orbital representations of the CASSCF active spaces of 3 (top, CAS(13,10)) and [4-OTf]⁰ (bottom, CAS(12,10)), along with orbital occupation numbers and orbital assignments.

Physical Oxidation State Assignments.

The picture that emerges from the combined experimental and computational data may be summarized as follows. The crystallographic data for 1 indicate that high-spin Fe(II) supported by the N₄ coordination environment of P₃tren will exhibit Fe–N_eq bond distances of ca. 2.07 Å and a Mössbauer isomer shift of δ = +1.0 mm/s. The shortening of the Fe–N_eq distances to ca. 2.03 Å and, though the spectral data are somewhat ambiguous, the apparent modest movement of δ to ~0.8 mm/s on formation of 3 indicate that Fe_N within the complex’s [Fe₂]³⁺ core is in a comparable oxidation state to the metal center within 1, albeit with some added covalency through binding to Fe_C in 3. The computational data similarly indicated that Fe_N within 3 is best described as an Fe(II) center, which is found to take part in weak-overlap covalent bonding to Fe_C. This latter attribution results from the observed low-energy electronic transitions for 3 (λ_max = 1044 nm), which are made possible by the presence of a compact valence manifold for this species. The CASSCF data for this compound support the notion that a wide number of electron configurations lie within a small energy range and contribute to the complex’s ground-state electronic structure.

Since the [Fe₂]³⁺ core of 3 contains a ferrous Fe_N, one is left to assume that Fe_C is best described as Fe(I); this description is supported by the data. The Fe–C distances (ca. 2.13 Å) are lengthened from known Fe(II) alkyl complexes in a trigonal geometry (ca. 2.08 Å),^{19, 21} and the computationally-informed isomer shift assignment ascribed to Fe_C in 3 (~0.5 mm/s) is more positive than the values for these same [FeR₃]⁻ Fe(II) complexes (δ = ca. 0.23 mm/s). Further, the CV data for 3 reveal an electrochemical event at −0.70 V that is close to the Fe(II/III) couple in 1 (−0.60 V), which suggests that the −0.70 V feature for 3 may be attributable to an Fe_N(II/III) electrochemical event. The feature at −2.06 V for 3 could then be assigned to the Fe(I/II) couple at Fe_C, which is consistent with the data accumulated for the 1 electron oxidation products [4-OTf]⁰ and [4-THF]⁺. Finally, the assignment of an Fe(I) physical oxidation state to Fe_C in 3 is further supported by the computational data, in that an OVB analysis of the CASSCF data found that Fe_C was best described as Fe(I).

The computational data further suggest that the polarization of the Fe–Fe bond in 3 may be accounted for by the inequivalent axial coordination environments between Fe_C and Fe_N. The axial nitrogen bound to Fe_N serves to raise the energy of the d-manifold on Fe_N with respect to Fe_C, allowing it to attain a higher oxidation state despite its formally neutral coordination environment. This difference is represented in the polarization of the calculated σ-bonding interaction toward Fe_C, suggesting either that the equatorial ligands are only able to play a minor role in the polarization of the Fe–Fe interaction in these compounds or, more likely, that the phosphinimine-ylide ligands show strongly anionic character at both the N- and C-based donors, resulting in little effective polarization of the Fe–Fe interaction as a result of the equatorial ligand fields. The shorter P–C_Fe distances (ca. 1.74 Å) compared to those for the P–C_Me bonds (ca. 1.81 Å) support the notion that resonance pictures like those depicted in Figure 6 provide a non-negligible contribution to the ground state of this ligand.

Figure 6.

Resonance structures of 3 showing the zwitterionic character of the phosphinimines and the ylidic character of the P–CH₂–Fe linkages.

The impact of the axial coordination was investigated further through a comparison of 3 to the oxidized species [4-OTf]⁰. The combination of the lowered electron count and the presence of an axial donor in [4-OTf]⁰ leads to an increase in the Fe–Fe distance by 0.26 Å. CASSCF calculations were consistent with disruption of the Fe–Fe bond; this outcome was supported by the NIR electronic absorption data for [4-OTf]⁰, which were red-shifted compared to those in 3 (Figure 3). The changing coordination number upon oxidizing 3 to [4-OTf]⁰ was also found to play a role in the Mössbauer parameters. The calculated isomer shift assigned to Fe_N in 3 (+0.70 mm/s) is close to the calculated values for Fe_N in [4-OTf]⁰ (+0.73 mm/s) and [4]⁺ (+0.66 mm/s). With the locus of oxidation at Fe_C, we anticipated that the isomer shift for Fe_C would move significantly toward a less positive number from the calculated value for 3 of +0.47 mm/s, but the value for [4-OTf]⁺ was found to be more positive (+0.53 mm/s). This appears to be a coordination number effect, in that the isomer shift of Fe_C decreased to +0.35 mm/s in [4]⁺, a species that is isomorphic to 3.

S-atom Transfer to 3.

Finally, we performed a preliminary investigation into the reactivity of 3 toward the sulfur atom transfer reagent HSCPh₃ as a way of both demonstrating the reducing capacity of Fe_c and probing the reactivity of the ylide-based ligands. The Lu and Thomas compounds shown in Figure 1 generally retain their architectures through a range of chemical transformations. While many reactivity studies remain to be done with 3 and 4, the results presented below reveal the ready ability of this system to dissociate on group addition chemistry.

Upon slow addition of the thiol at low temperature and subsequent warming to room temperature, a vibrant red solution was formed. Two products, 5 and 6 (Scheme 2), were consistently co-crystallized from this solution in low yield. The low conversion and inability to separate the two precluded extensive characterization but nonetheless provided an example of how this weak M-M bond and unique ligand set may be used to support group transfer reactivity.

Scheme 2.

Synthesis of Fe-S insertion products

Intriguingly, both complexes involve both the formal insertion of a sulfur atom into the Fe–Fe bond and the formation of a dimer (Scheme 2 and Figure 7). Complex 5 is an Fe₂S₂ complex in which one ylidic arm has gained an H-atom, presumably from the thiol. The Fe_N–Fe_C distances average 3.6188 Å, demonstrating cleavage of the Fe–Fe bond, and the Fe_N–S bond distances (2.443(1) Å) are also elongated. The Fe₂S₂ core is planar with an Fe–Fe distance of 2.6964(8) Å. The core contains a mirror plane and has Fe–S distances of 2.301(1) and 2.443(1) Å, which are on the upper end of the expected range.³⁶ Complex 6 is the product of the formal addition of 0.5 equivalents of an S-atom into 3 and has a μ₄-S moiety bound to the four iron atoms. Again, disruption of the Fe–Fe bond is required to form this complex, and the product appears to contain four ferrous centers. The sulfur is in a distorted tetrahedral configuration with Fe-S-Fe angles ranging from 80.9° to 106.5°. The Fe–S bond lengths are again long, with average Fe_N–S and Fe_C–S bond lengths of 2.446(2) Å and 2.354(1) Å, respectively. Both complexes are paramagnetic, and like the parent complex, NMR spectroscopic peak assignments cannot definitively be made, especially as they are only observed as a mixture. The weak M–M bond observed in 3 along with its apparent propensity to dissociate an ylidic arm can thus lead to interesting S atom inserted products. This reactivity contrasts with that of compounds B, C, and D (Figure 1), for which the dinuclear core is maintained and chemistry primarily occurs on the distal metal. Future work will focus on ways to expand and exploit this dinuclear participation.

Figure 7.

Thermal ellipsoid plots of complexes 5 and 6. Thermal ellipsoids are shown at the 50% probability level. Hydrogen atoms, solvates, and disordered atoms are omitted for clarity.

CONCLUSION

In conclusion, the reduction of a phosphinimine-supported mononuclear Fe(II) complex led to the serendipitous formation of an unusual trigonal diiron complex following C–H activation and H₂ extrusion. This result is notable given the unusual set of three ylidic carbon donors coordinated the “top” metal (Fe_C) and the strong phosphinimine and amine nitrogens supporting the “bottom” iron (Fe_N). This complex and its one electron oxidation product, [4-OTf]⁰/[4-THF]⁺, exhibit high-spin ground states that allow for a modest amount of Fe–Fe bonding. However, the varying coordination environments of the metal centers lead to polar metal-metal interactions that reflect distinct oxidation states for the two metal centers in 3. Importantly, it appears that the difference in axial donor, instead of the C- vs. N-equatorial donor identity, bears significant responsibility for the polarization of the Fe–Fe interaction. Upon addition of a sulfur atom donor to 3, cleavage of the weak M–M bond occurred, and two bridging sulfide complexes were formed. Both 3 and [4-OTf]⁰ were characterized using spectroscopic and computational methods and add a unique ligand architecture to the repertoire of trigonal, high-spin diiron complexes that have applications ranging from small molecule catalysis to single-molecule magnets.

EXPERIMENTAL

General Considerations

All reactions containing transition metals were performed under an inert atmosphere of N₂, using standard Schlenk line or glovebox techniques. Glassware, stir bars, filter aid (Celite), and 3 Å molecular sieves were dried in an oven at 150 °C for at least 12 h prior to use. All solvents (THF, acetonitrile, toluene, n-pentane, and diethyl ether) were dried by passage through a column of activated alumina, deoxygenated by passage through a copper Q5 column where applicable, sparged with N₂, and stored over activated 3 Å molecular sieves under an inert atmosphere. Deuterated solvents were purchased from Cambridge Isotope Laboratories, Inc., and dried over either Na/benzophenone (THF-d₈, C₆D₆) or CaH₂ (MeCN-d₃), isolated by vacuum transfer (THF-d₈, MeCN-d₃) or distillation (C₆D₆), and stored under an inert atmosphere over 3 Å sieves. P₃tren was synthesized following previously published procedures.¹³ KC₈ was synthesized according to a literature procedure³⁷ and stored under nitrogen at −35 °C in a glovebox prior to use. PMe₃ (98%) was either purchased from Strem Chemicals or synthesized according to a literature procedure.¹³ AgOTf was dried over P₂O₅ and stored in the dark. [ⁿBu₄N][PF₆] (98%) was purchased from MilliporeSigma and recrystallized twice from hot ethanol, followed by drying at 60 °C under 30 mbar for 10 h and stored in a glovebox under dry nitrogen prior to use.³⁸ Ferrocene (98%) was purchased from Acros Organics, purified by sublimation three times,³⁹ and stored in a glovebox under dry nitrogen prior to use. Cobaltocenium triflate was prepared using a modified literature procedure, replacing indium triflate with silver triflate.⁴⁰ All other chemicals were used as received. ¹H and ¹⁹F{¹H} NMR spectra were recorded on a UNI 400 spectrometer. All chemical shifts (δ) are reported in units of ppm, with references to the residual protio-solvent resonance for proton chemical shifts. Internal PhF was used for referencing ¹⁹F NMR spectra.⁵ Elemental analysis was performed by Midwest Microlab, LLC or at the CENTC Elemental Analysis Facility, Department of Chemistry, University of Rochester. IR spectra (KBr pellet) were collected on a JASCO FT/IR-480 Plus spectrometer FTIR, and the IR spectra are given in the Supporting Information. Solution phase effective magnetic moment data were determined using Evans’ method.⁴¹ P₃tren’ refers to the activated methyl group variant of P₃tren.

Cyclic voltammetry experiments were performed in a VAC OMNI-LAB glovebox with an Epsilon E2 Potentiostat. The data were processed with BASi Epsilon-EC software version 2.13.77. All experiments were performed under an N₂ atmosphere in a glovebox using an electrochemical cell that consists of a glassy carbon (3 mm outer diameter) working electrode, a platinum wire counter electrode, and a Ag/AgNO₃ (0.01 M in acetonitrile) reference electrode. All experiments were conducted in acetonitrile, with 1 mM analyte and 100 mM [ⁿBu₄N][PF₆] as the supporting electrolyte. Potentials were reported versus Cp₂Fe^+/0.

UV–Vis absorption spectra were collected from 200 to 1000 nm using an Agilent Cary 60 UV–vis spectrophotometer at room temperature. NIR spectra were recorded on a Perkin Elmer 950 UV-Vis/NIR spectrophotometer at room temperature. The samples were prepared under an N₂ atmosphere in a glovebox. Stock solutions were prepared by dissolving a known mass (ca. 10 mg) of sample in 10.00 mL of solvent (3, [4-OTf]⁰ (THF); 1 and 2 (CH₃CN)). Varying amounts of the stock solutions (50.0–400.0 μL) were then diluted to 5.00 mL and transferred to a 10 mm path-length quartz cuvette with a screw cap for data collection. For each species, at least four spectra at different concentrations were collected. The absorption intensities of various peak maxima were plotted vs. concentration to ensure that absorption data were being collected in the linear response range of the spectrometer. Linear regression fits to the plotted data resulted in R² values ≥ 0.994.

⁵⁷Fe Mossbauer spectra were collected on a spectrometer from SEE Co. (Edina, MN) operating in the constant-acceleration mode in a transmission geometry. Samples were cooled in an SVT-400 cryostat from Janis Research Co. (Wilmington, MA), using liquid N₂ as a cryogen for 80 K measurements. A parallel applied magnetic field of ca. 50 mT was applied. Samples were prepared under nitrogen in custom Delrin cups equipped with tight-fitting Delrin inserts. The zero velocity of each Mossbauer spectrum refers to the centroid of a 25 micron metallic iron (α-Fe) foil collected at room temperature. The WMOSS program (SEE Co., formerly WEB Research Co., Edina, MN) was used to analyze the data, which were fit to simple quadrupole doublets with Lorentzian line shapes.

Crystallographic X-ray intensity data were collected on a Bruker D8QUEST CMOS (2, [4-THF]⁺, 6) or a Bruker APEXII CCD (1, 3, [4-OTf]⁰) area-detector diffractometer with graphite-monochromated Mo K_α radiation (λ = 0.71073 Å) at 100 K or a Rigaku XtaLAB Synergy-S HyPix-6000HE HPC (5) area-detector diffractometer with confocal multilayer optic-monochromated Cu-Kα radiation (λ=1.54184Å) at a temperature of 100K. Rotation frames were integrated using CrysalisPro⁴² for 5 and SAINT⁴³ producing a listing of unaveraged F² and σ(F²) values. The intensity data were corrected for Lorentz and polarization effects and for absorption using SCALE3 ABSPACK⁴⁴ or SADABS.⁴⁵ The structures were solved by direct methods by using SHELXT⁴⁶ and refined by full-matrix least-squares, based on F² using SHELXL-2017.⁴⁷ For [4-OTf]⁰ the asymmetric unit consists of four crystallographically-independent molecules. Non-hydrogen atoms were refined anisotropically and hydrogen atoms were refined using a riding model. CCDC entries 2154284-2154290 contain the supplementary crystallographic data for this paper. These data can be obtained free of charge from the Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/data_request/cif.

DFT geometry optimizations were performed with unrestricted Kohn-Sham DFT using the B97-D3 pure GGA functional^48–50 with Orca v4.2.1.⁵¹ The initial geometries for all optimizations were obtained from crystallographic coordinates. This functional was chosen due to its reliable ability to predict bond lengths and angles within this system. The def2-TZVP basis sets and def2-TZVP/J auxiliary basis sets (used to expand the electron density in the resolution-of-identity (RI) approach) were used for all Fe and Fe-bound atoms.^52–54 The def2-SV(P) basis sets were used for all other atoms. The RIJCOSX approximation was used for computational feasibility. The SCF calculations were tightly converged (1x10⁻⁸ E_h in energy, 1x10⁻⁷ E_h in the density change, and 5x10⁻⁷ in the maximum element of the DIIS error vector). In all cases the geometries were considered converged after the energy change was less than 1x10⁻⁶ E_h, the gradient norm and maximum gradient element were smaller than 3x10⁻⁴ E_h-Bohr⁻¹ and 1x10⁻⁴ E_h-Bohr⁻¹, respectively, and the root-mean square and maximum displacements of all atoms were smaller than 6x10⁻⁴ Bohr and 1x10⁻³ Bohr, respectively.

CASSCF calculations were carried out in Orca v4.2.1 to characterize the ground state across the diiron series. The same basis sets were used as above. RIJCOSX approximation was once again used to reduce the computational cost. All orbitals in the 3d manifold were included for both iron centers, consisting of 13 electrons for 3, and 12 electrons for both [4-THF]⁺ and [4-OTf]⁰.

The process of modelling Mössbauer isomer shifts with DFT requires computation of the nuclear electron density, ρ(0), followed by the use of fitting parameters (α, β, and C) that allow ρ(0) to be translated into a calculated isomer shift. The values for α and β are obtained by performing a linear regression analysis between calculated ρ(0) values and experimental isomer shifts for a test set of complexes, while C is an unrefined constant that is used to prevent β from becoming unreasonably large. These values are dependent on the level of theory but often translate well to novel compounds, provided the new species are structurally and electronically related to those in the test set. Isomer shifts for each iron center were estimated using single point calculation data carried out in Orca v4.2.1. EPR/NMR analysis was carried out to get accurate values for ρ(0) for each center. Quadrupole splitting parameters and isomer shifts were then estimated based on the ρ(0) values for each Fe using a published method with parameters from test set 1.⁵⁵

Synthetic Procedures

[(P₃tren)FeCl]Cl (1):

FeCl₂ (353.5 mg, 2.79 mmol) was dissolved in ca. 15 mL of acetonitrile with stirring. An acetonitrile solution (10 mL) of P₃tren (1.0310g, 2.80 mmol) was added to the FeCl₂ solution via cannula at room temperature. The solution turned light yellow and was allowed to stir for 20 min before being filtered through Celite. Volatile materials were removed in vacuo, and the resulting white solid was washed with THF before being dried again in vacuo for 10 min to yield a white powder (1.2433g, 90%). Large, colorless blocks suitable for single-crystal X-ray diffraction were grown by vapor diffusion of diethyl ether into a saturated acetonitrile solution of 1. ¹H NMR (400 MHz, CD₃CN, 300 K): δ = 119.21 (br s, 6H), 106.42 (br s, 6H), −1.77 (br s, 27H) ppm. Anal. Calcd. for C₁₅H₃₉Cl₂Fe₁N₄P₃ (495.17 g/mol): C, 36.38; H, 7.94; N, 11.31%. Found: C, 35.81; H, 7.89; N, 11.16%.

[(P₃tren)FeCl][OTf]₂ (2):

A solution of AgOTf (217.4 mg, 0.85 mmol) in 2-3 mL of acetonitrile was added at room temperature to a stirred solution of 1 (204.5 mg, 0.41 mmol) in 5 mL of acetonitrile, resulting in a rapid color change from light yellow to bright orange-red with the formation of a dark precipitate. The mixture was stirred for 5 min before all volatile materials were removed in vacuo. The reddish solid was washed with THF (2 x 2 mL) and then extracted into a minimal amount of acetonitrile (ca. 2 mL). Vapor diffusion of ether into the saturated acetonitrile solution at room temperature yielded large red blocks after 2 days (244.4 mg, 78%). ¹H NMR (400 MHz, CD₃CN, 300 K): δ = 14.05 (br s) ppm. ¹⁹F{¹H} NMR (376 MHz, CD₃CN, 300 K): δ = −78.53 ppm (s). Anal. Calcd. For C₁₇H₃₉Cl₁Fe₁N₄O₆P₃S₂ (757.85 g/mol): C, 26.94; H, 5.19; N, 7.39%. Found: C, 26.91; H, 4.88; N, 7.28%.

(P₃tren’)FeFe (3):

Safety note: The filtrand from this reaction is pyrophoric. 1 (1.2437g, 2.51 mmol) and FeCl₂ (322.0 mg, 2.54 mmol) were suspended in 15 mL of toluene with stirring. A slurry of KC₈ (1.3697g, 10.1 mmol) in 5 mL of toluene was added to the iron suspension at room temperature. After ~30 min the mixture became a deep green color and was stirred overnight with venting to prevent build-up of H₂. The mixture was filtered through Celite and the volatiles were removed in vacuo. The resulting green solid was extracted into 20 mL of n-hexane, filtered through Celite, and concentrated to 10 mL. Storage of the resulting solution overnight at −35 °C yielded large, dark-green blocks (387.1 mg, 32%). The crystals were washed with cold HMDSO (3 x 1 mL). The yield may be marginally increased upon further concentration and cooling of the mother liquor to yield additional crops of 3. Typical total yields ranged from 30-40%. ¹H NMR (400 MHz, C₆D₆, 300 K): δ = 217.07 (br s), 83.55 (br s) ppm. Anal. Calcd. for C₁₅H₃₆Fe₂N₄P₃ (477.09 g/mol): C, 37.76; H, 7.61; N, 11.74%. Found: C, 38.13; H, 7.45; N, 11.57%.

[(P₃tren’)FeFe(thf)][OTf] ([4-THF]⁺):

3 (113.4 mg, 0.24 mmol) was dissolved in 5 mL of THF, and a suspension of [Cp₂Co][OTf] (75.0 mg, 0.22 mmol) in 3 mL THF was added dropwise over 1 min. The solution turned dark brownish-green and was stirred for 5 min. The volatile materials were removed in vacuo, giving a red-brown solid. The solid was washed with pentane (4 x 2 mL) to remove cobaltocene, and the remaining yellow-green solid was extracted into THF (3 mL), and the solution was filtered through Celite. The THF solution was then layered with pentane (10 mL) and stored overnight at −25 °C. The yellow-green needles that grew during this time were washed with pentane and dried in vacuo to yield a crystalline product (81.7 mg, 49%). ¹H NMR (400 MHz, THF-d₈, 300 K): δ = 349.75 (br s), 152.62 (br s), 84.13 (br s) ppm. ¹⁹F{¹H} NMR (376 MHz, THF-d₈, 300 K): δ = −34.53 ppm (br s). Anal. Calcd. For C₂₀H₄₄F₃Fe₂N₄O₄P₃S₁ (698.26 g/mol): C, 34.40; H, 6.35; N, 8.02%. Found: C, 33.63; H, 5.98; N, 7.74%. We could not obtain passing elemental analysis due to the thermal instability of [4-THF]⁺.

(P₃tren’)FeFeOTf ([4-OTf]⁰):

3 (79.4 mg, 0.17 mmol) was dissolved in 2 mL of THF and a suspension of [Cp₂Co][OTf] (56.3 mg, 0.17 mmol) in 3 mL THF was added dropwise over 1 min. The solution turned dark brownish green and was stirred for 5 min. The volatile materials were removed in vacuo, giving a red-brown solid. The solid was washed with pentane (4x2 mL) to remove cobaltocene, and the remaining yellow-green solid was extracted into toluene (5x1 mL). The resulting solution was filtered through Celite, then layered with pentane (10 mL). Yellow-green needles grew after cooling to −25ºC overnight. The crystals were washed with pentane and dried in vacuo (12.6 mg, 12%). ¹H NMR (400 MHz, C₆D₆, 300 K): δ = 351.86 (br s), 147.58 (br s), 86.51 (br s) ppm. ¹⁹F{¹H} NMR (376 MHz, C₆D₆, 300 K): δ = −18.54 ppm (br s). ¹H NMR (400 MHz, THF-d₈, 300 K): δ = 349.84 (br s), 152.76 (br s), 84.19 (br s) ppm. ¹⁹F{¹H} NMR (376 MHz, THF-d₈, 300 K): δ = −36.90 ppm (br s). Anal. Calcd. for C₁₆H₃₆F₃Fe₂N₄O₄P₃S₁ (626.16 g/mol): C, 30.69; H, 5.80; N, 8.95%. Found: C, 26.81; H, 5.26; N, 7.41%. We could not obtain passing elemental analysis due to the thermal instability of [4-OTf]⁰.

[(P3tren’)FeFe]₂(S₂) (5) and [(P3tren’)FeFe]₂S (6):

3 (23.0 mg, 0.05 mmol) was dissolved in 2 mL of Et₂O and frozen in a cold well. HSCPh₃ (13.2 mg, 0.05mmol) was dissolved in 3 mL of Et₂O and partially frozen in a cold well. The solution of 3 was removed from the cold well and the chilled solution of HSCPh₃ was added to the solution of 3 with stirring in 3 aliquots over 1–2 minutes. A precipitate formed and the solution turned brown-green. Upon further warming to room temperature the solution turned deep red with a red precipitate and was stirred for 2 minutes. The volatile materials were removed in vacuo, giving a red solid. The solid was sequentially extracted with hexane (1 mL), Et₂O (1 mL), and THF (1 mL), with each extract filtered through Celite. X-ray quality yellow plates of 6 were grown from the concentrated light-pink Et₂O or hexane solutions at −25ºC after several days. Vapor diffusion of pentane into the saturated red THF solution at −25ºC for at least 1 week yielded a mixture of dark red blocks of 5 and small amount of 6 (6.4 mg combined, ~26%). Despite varying the rate of addition and equivalents of HSCPh₃ (0.5–1.0 equiv), the separability issues and product distribution precluded isolation of each product. ¹H NMR (400 MHz, C₆D₆) δ 137.18 (s, 1H), 121.49 (s, 1H), 96.67 (s, 1H), 90.65 (s, 1H), 70.19 (s, 1H), 61.31 (s, 1H), 58.64 (s, 1H), 41.25 (s, 1H), 21.12 (s, 3H), −2.63 (s, 4H).