Iron Complexes of a Proton-Responsive SCS Pincer Ligand with a Sensitive Electronic Structure

Abstract

SCS pincer ligands have an interesting combination of strong-field and weak-field donors, a coordination environment that is also present in the nitrogenase active site. Here, we explore the electronic structures of iron(II) and iron(III) complexes with such a pincer ligand, bearing a monodentate phosphine, thiolate S donor, amide N donor, ammonia, or CO. The ligand scaffold features a proton-responsive thioamide site, and the protonation state of the ligand greatly influences the reduction potential of iron in the phosphine complex. The N–H bond dissociation free energy, derived from a square scheme, is 56 ± 2 kcal/mol. EPR spectroscopy and SQUID magnetometry measurements show that the iron(III) complexes with S and N as the fourth donors have an intermediate spin (S = 3/2) ground state with large zero field splitting, and X-ray absorption spectra show high Fe–S covalency. The Mössbauer spectrum changes drastically with the position of a nearby alkali metal cation in the iron(III) amido complex, and DFT calculations explain this phenomenon through a change between having the doubly-occupied orbital as d_z2 or d_yz, as the former is more influenced by the nearby positive charge.

Graphical Abstract

INTRODUCTION

The organometallic chemistry of iron has been dominated by strong-field supporting ligands such as CO, CN, and phosphines, and by macrocyclic N ligands like porphyrins.^1–6 The active sites of hydrogenase enzymes incorporate S-based ligands, and these are low-spin due to the influence of carbonyl and cyanide donors.⁷ However, the interesting reactions of nitrogenase enzymes, and a new generation of low-valent iron catalysts, have predominantly weak-field ligands and have led to increasing attention to organometallic iron complexes with higher spin states.^8–11 We focus here on iron coordination environments that result from a mixture of C and S donors – choices that are particularly compelling since the six “belt” iron atoms in the iron-molybdenum cofactor (FeMoco) of nitrogenase have a mixed C/S coordination sphere.^12–14 This unusual combination of potentially strong-field C donors and weak-field S donors could lead to changes in spin states during catalysis, which has been linked to changes in barriers and selectivity.^15–18 Thus, the study of C- and S-ligated iron has relevance for both fundamental coordination chemistry and bioinorganic mechanisms.^19–21 However, CO-free iron complexes with supporting ligands that coordinate through only carbon and sulfur donors are rare.^22–28 Of these examples, only one multidentate C/S ligand is known to support N₂ binding.²⁶ In addition, Qu has provided important studies on Cp*-supported iron dimers bridged by dithiolates, and this C/S ligand sphere can bind and facilitate the reduction of nitrogenase-relevant N_xH_y substrates.^{23–25, 29}

Our group’s recent work on SCS systems began with a dithiolate ligand having a central arene that can interact with iron through backbonding into its π-system in low oxidation states.^{26, 30} While this hemilabile interaction stabilizes the reduced complexes, complete dissociation of the SCS ligand occurred in the presence of Brønsted acids, and backbonding into N₂ competed with backbonding into the arene. In order to address this instability, we then moved to tridentate pincer scaffolds in which the carbon forms a σ-bond to the iron center. Namely, we synthesized a new SCS pincer comprised of an N-heterocyclic carbene (NHC) ligand with two bulky thiolate arms, and isolated a number of iron complexes with this scaffold.²⁸ The geometries at iron sites with this supporting ligand were sometimes tetrahedral and sometimes square planar, indicating significant flexibility in the core. However, this NHC ligand was also sensitive to acids, and the ligand synthesis involved time-consuming protection and deprotection steps.

Here, we study the first iron complexes of pincer ligands containing a central aryl C donor and two S-donor arms. SCS pincers of this type have been primarily studied in complexes of group 10 metals (Chart 1, top), but only have precedent in group 8 metal complexes of ruthenium.^31–35 Thioamide arms are compelling, because they are easily prepared from readily available precursors and offer the potential for proton responsiveness.^36–38 There is no precedent for iron complexes with thioamide-aryl ligands, although there is a related iron SNS pincer complex with a central pyridine (Chart 1, lower left).^39–40

Chart 1.

Typical palladium SCS pincer complexes and closely-related iron complexes with SNS and SCS pincers. In the bottom pictures, the square represents a coordination site with various ligands.

Thioamides have two tautomers, which form the same anion upon deprotonation (Figure 1, right). Most often, the thioamide coordinates as an iminothiolate through the S donor rather than the potential N donor as a result of the relative weakness of the C=S π bond and the better overlap of the S atom with metal orbitals.⁴¹ After coordination, protonation at the N atom gives a formally neutral thione donor that maintains the M–S interaction (Figure 1, brackets). Even as a formally neutral ligand, the thione C–S bond is significantly polarized toward a C^δ+–S^δ– form, which affords some anionic character at S.⁴² The ability to tune the donor properties of iron-coordinated sulfur via protonation/deprotonation of the supporting ligand could be used to probe how protonation state affects the redox potential, spectroscopic characteristics, and reactivity of the iron center.

Figure 1.

Illustration of thioamide tautomerization and deprotonation to form an anionic donor.

The potential for proton responsiveness is also relevant to nitrogenase mechanisms. The Thorneley-Lowe kinetic scheme for nitrogenases proposes that one proton is transferred to the FeMoco with each reduction step,⁴³ but the location of these protons has been controversial. Computational studies have showed that there are many potential sites of protonation on the FeMoco, including the sulfides,^44–47 the Mo-coordinated homocitrate,^{45, 48} and even the carbide.^{46–47, 49} Studies on multi-site proton-coupled electron transfer (PCET) have demonstrated that various protonation sites can influence the redox potentials and bond dissociation free energies (BDFE).^50–53 These inspirations motivated us to pursue thermochemical studies on well-characterized (SCS)Fe complexes with proton-responsive ligands to examine the effects of distant protonation events on potentially biologically-relevant iron sites.

In the present study, we describe a new ligand scaffold that incorporates thioamides and an anionic aryl group to mimic the sulfur and carbon coordination environment found in the FeMoco. The ligand is easy to prepare on a useful scale, and the stability of the iron complexes is highlighted by the ability of an iron(III) complex to be handled in air and water. Although no iron complex of the new SCS ligand was observed to bind N₂, we use EPR and Mössbauer spectroscopy alongside magnetometry measurements to elucidate the unusual electronic structures of complexes with different exogenous donors, including biologically-relevant amide, thiolate, ammonia, and CO. In the amide complex, the electronic structure is influenced by the presence of nearby cations, as demonstrated by computations and spectroscopy. We also demonstrate the ability of the thioamide arm to be proton responsive, and quantify the connection between protonation state and redox state of iron.

RESULTS

Ligand Synthesis and Metalation.

Using a procedure modified from the literature,⁵⁴ isophthalic acid was treated with thionyl chloride followed by 2,6-diisopropylaniline, which provided the diamide (Scheme 1). Reaction of this compound with P₂S₅ in toluene at 100 °C yielded the bis(thioamide) 1, in an overall yield of 69% in two steps from commercial starting materials. Treating 1 with a slight excess of Fe(PMe₃)₄ in Et₂O at room temperature led to effervescence and a color change from yellow-brown to dark green, and the iron(II) complex 2 was isolated in 96% yield (Scheme 1).^55–56

Scheme 1.

Synthesis and metalation of SCS pincer ligand 1. Dipp = 2,6-diisopropylphenyl.

Crystallization of 2 gave green blocks suitable for X-ray diffraction (Figure 2, top). The diffraction data showed an octahedral iron site with a meridional pincer ligand as expected. The Fe−C bond length is 1.9502(18) Å, and the average Fe−S length is 2.268(9) Å. The three Fe−P bonds are only slightly different, with a length of 2.2484(3) Å for the phosphine trans to C and lengths of 2.2553(4) Å and 2.2608(4) Å for the phosphines cis to C. There was disorder at the thioamide N−H site, which was satisfactorily modeled with each thioamide having 50% hydrogen atom occupancy, consistent with single protonation of the supporting ligand in 2.

Figure 2.

Molecular structures of monoprotonated iron(II) complex 2 and fully deprotonated iron(III) complex 4, plotted as 50% probability thermal ellipsoids. All hydrogen atoms except those bonded to N are hidden for clarity, and each N−H site is modeled as having 50% hydrogen atom occupancy.

The Mössbauer spectrum of 2 has a doublet with δ = 0.21 mm/s and |ΔE_Q| = 1.25 mm/s (Figure S46). Proton NMR spectra in either C₆D₆ or THF-d₈ show signals in the range expected for a diamagnetic complex of low-spin iron(II). The most downfield resonance integrates to one proton and disappears upon addition of D₂O, and therefore it is assigned to the thioamide proton (Figure S5). Based on the NMR spectra, 2 is unsymmetric in solution, and it is best described has having one neutral thioamide S donor and one anionic iminothiolate donor. The solid-state IR spectrum of 2 shows a broad peak centered at 3336 cm^–1 (Figure 3, red trace), which is distinct from the analogous signal in free ligand 1 at 3136 cm^–1 (Figure 3, gray trace), consistent with an N–H stretching vibration.

Figure 3.

Solid state IR spectra of neutral ligand 1 and monoprotonated ferrous complex 2, ferric iminothiolate complex 4, and ferrous iminothiolate complex 3-Na in the range of 4000 to 2000 cm⁻¹. The bands assigned to N−H stretches (3136 and 3336 cm⁻¹) in the spectra of 1 and 2 are labeled.

Electron and Proton Transfer in Phosphine Complexes.

We serendipitously discovered that neutral, fully-deprotonated iron(III) compound 4 could be prepared in one pot from 1 and Fe(PMe₃)₄ by simply exposing the crude product 2 to air after metalation was complete. Proton NMR and UV-visible spectroscopy of air-exposed 2 in situ demonstrates rapid and complete conversion to 4, with no monoprotonated iron(III) transient species observed. This one-pot procedure afforded 4 in 88% yield from 2, and this method was used to prepare 4 for subsequent experiments.

Additional proton and electron transfer reactions were carried out on 2 and 4 (Scheme 2). Treating the neutral compound 2 with stoichiometric sodium bis(trimethylsilyl)amide provided anionic 3-Na. Oxidation of 3-Na using AgPF₆ regenerated 4. Alternatively, 4 could be formed in one step by hydrogen atom abstraction from 2 using TEMPO. It was also possible to generate the potassium analogue of 3-Na (3-K) by treatment of neutral iron(III) compound 4 with stoichiometric KC₈. Both 3-Na and 3-K have sharp peaks in their ¹H NMR spectra suggesting low-spin iron(II), and their structures were inferred based on these spectra (Figures S7 and S9).

Scheme 2.

Conversion of 2 to 4 using stepwise and concerted H atom removal. Yields were determined by ¹H NMR using an internal standard.

The X-ray crystal structure of the fully-deprotonated iron(III) complex 4 (Figure 2, bottom) shows an Fe−C bond length of 1.9413(12) Å, which is indistinguishable from the analogous distance in the monoprotonated iron(II) complex 2. The average Fe−S length is 2.23(2) Å in 4, which is slightly shorter than in 2. Again, all three Fe−P bond distances are different, with the phosphine trans to C having the shortest bond of 2.2692(6) Å while the cis phosphines have lengths of 2.3007(6) Å and 2.3050(5) Å. The average Fe−P bond in 4 is shorter than in 2 (2.23(1) Å vs. 2.255(4) Å). Thus, there is a general contraction of bonds in ferric 4 when compared to ferrous 2, consistent with 4 being more oxidized.

As expected for complexes lacking protonated thioamides, 3-Na and 4 do not have any IR bands above 3050 cm^–1 that correspond to N–H stretching vibrations (Figure 3, blue and green traces). Proton NMR spectra of ferric 4 contain very broad peaks between −20 and +12 ppm (Figure S11), and the ³¹P NMR spectrum is featureless, as expected for nuclei directly bound to a paramagnetic metal. The relative integrations of the evident peaks in the ¹H NMR spectra of 3-M in THF-d₈ are consistent with C_2v symmetry. Taken together, the IR and NMR data support double deprotonation of the thioamide groups of the supporting ligand in 3-Na and 4.

Cyclic voltammetry (CV) studies of 4 showed a reversible Fe^3+/2+ couple at E_1/2 = −1.42 V vs. Fc⁺/Fc in THF (Figures S60–61). Next, we tested the ability of different organic bases to deprotonate 2, using changes in the ³¹P NMR chemical shifts to determine the ratio of 2 and deprotonated 2 since the conjugate base and acid are in rapid equilibrium. Addition of neither Et₃N (pK_aH = 12.5 in THF) nor DBU (pK_aH = 16.9 in THF) gave changes in the NMR spectra of 2.⁵⁷ Addition of 1.10 equiv of triazabicyclodecene (TBD, pK_aH = 21.0 in THF)⁵⁷ to 2, on the other hand, gave a ³¹P NMR spectrum after equilibration with peaks at 12.1 ppm and 10.8 ppm. These shifts are 4.8 ppm upfield from the ³¹P resonances of 3-K but 1.4 ppm downfield of ³¹P resonances of 2 (Figure S77). The presence of both the protonated and deprotonated species in this equilibrium indicates that the pK_a of 2 is 22 ± 1 in neat THF-d₈. An equivalent NMR experiment in 0.3 M [NⁿBu₄][PF₆] in THF-d₈ (the electrolytic solution used for CV) indicated a pK_a of 21 ± 1 (Figure S78). These data enable us to define the bond dissociation free energy (BDFE) of the N–H bond using the Bordwell equation, with C_G(THF) = 60.4 ± 2 kcal/mol.^58–59 This analysis gave a BDFE(N–H) of 56 ± 2 kcal/mol (Scheme 3).

Scheme 3.

Thermodynamic values used to calculate the BDFE.

Other redox and protonation events were tested as well. Electrochemical oxidation of 2 is irreversible under the same CV conditions used for 4 (Figure S59). To test whether 4 could be protonated, it was treated with one equiv of [H(OEt₂)₂][BF₄] in THF. A dark brown solid precipitated, and its IR spectrum showed a band at 3250 cm⁻¹, which is much lower than the N–H stretching frequency of 2 at 3336 cm⁻¹ (Figure S79). Though further studies on putative protonated 4 was prevented by its insolubility, the IR spectrum of acid-treated 4 may indicate the feasibility of thioamide protonation.

Spectroscopy of Phosphine Complexes.

The Mössbauer signal for 2 (δ = 0.21 mm/s and |ΔE_Q| = 1.25 mm/s, Figure S46) is similar to those for 3-Na (δ = 0.24 mm/s and |ΔE_Q| = 0.94 mm/s, Figure S47) and 3-K (δ = 0.25 mm/s and |ΔE_Q| = 1.05 mm/s, Figure S48), consistent with low-spin iron(II) complexes. Solid 4 has a lower isomer shift of 0.16 mm/s and much larger quadrupole splitting of 3.45 mm/s, and its doublets are highly asymmetric, with Γ_L = 0.57 mm/s, Γ_R = 0.32 mm/s. To determine the spin state of 4, solid state and C₆D₆ solution magnetic moments were obtained. Both measurements gave μ_eff = 1.6 μ_B at 298 K, which indicates a ground-state spin of S = 1/2.

Spectroscopy and DFT calculations using B3LYP/ZORA-def2-TZVP were further employed to reveal the electronic structure of 4. The EPR spectrum of 4 in 2-methyltetrahydrofuran (MeTHF) at 77 K was modeled as an S = 1/2 spin system with g = [2.200, 2.120, 1.998] and coupling to three ³¹P nuclei: two with |A| = [71, 83, 71] MHz and one with |A| = [0, 37, 37] MHz (Figure 4). Mulliken spin population analysis showed spin populations of −0.036 on the axial phosphorus atoms and −0.019 on the equatorial phosphorus, in agreement with the fit to two equivalent ³¹P nuclei having larger A than a third ³¹P nucleus. The spin population analysis also showed an averaged spin density of 0.031 on the two ligand nitrogen atoms, though the smaller nuclear magnetic moment of ¹⁴N relative to ³¹P likely contributed to smaller, unresolvable ¹⁴N hyperfine splitting, which was modeled by anisotropic H-strain of [66, 27, 0] MHz in the simulation.

Figure 4.

X-Band EPR spectrum of 4 at 77 K (1.0 mM in 2-MeTHF) measured with a microwave frequency of 9.384 GHz and power of 1.779 mW. Parameters for simulation: g₁ = 2.200, |A₁(³¹P)| = 71, 71, 0 MHz; g₂ = 2.120, |A₂(³¹P)| = 83, 83, 37 MHz; g₃ = 1.998, |A₃(³¹P)| = 71, 71, 37 MHz; HStrain = [66, 27, 0] MHz. Inset: Mulliken spin densities are shown next to the respective phosphorus atoms (shown in magenta). Hydrogen atoms and dipp groups are omitted for clarity.

Synthesis of Phosphine-Free SCS Iron Complexes.

When iron(III) species 4 was treated with a slight excess of triphenylborane in THF, the color of the reaction mixture changed from green to orange, and precipitation of Me₃P–BPh₃ was observed (Scheme 4). Crystallization from Et₂O gave 5-Et₂O in 58% yield (Figure S70). Dissolving 5-Et₂O in THF and removing solvent under vacuum several times led to the formation of 5-THF (Figure S69). We refer to these compounds collectively as 5-Solv because of their similarity and ability to interconvert as a function of solvent.

Scheme 4.

Phosphine abstraction with triphenylborane. Dipp = 2,6-diisopropylphenyl.

Both compounds 5-Solv are dimeric in the solid state (Figure 5), with each iron center in a square pyramidal geometry with a coordinated solvent molecule occupying the axial site. The dimers adopt a butterfly shape in which the two iron centers are bridged by one sulfur from each ligand. In 5-Et₂O, the planes containing each pincer intersect at 75°. Comparing the two arms of the pincer ligand in 5-Et₂O, the bridging Fe–S bond is longer than the non-bridging Fe–S by 0.08(1) Å. Other bond metrics of 5-Et₂O are comparable to those in phosphine precursor 4, and the long C–S (1.78(3) Å average) and short C–N bonds (1.26(1) Å average) indicate two anionic iminothiolate donors. The metrical parameters of 5-THF are nearly identical to those of 5-Et₂O, with Fe−C bond lengths of 1.961(2) Å (5-THF) and 1.966(3) Å (5-Et₂O), Fe−S_bridging bond lengths of 2.2557(7) Å (5-THF) and 2.278(1) Å (5-Et₂O), and Fe−S_non-bridging bond lengths of 2.1797(7) Å (5-THF) and 2.187(1) Å (5-Et₂O).

Figure 5.

Molecular structure of 5-Et₂O plotted at 50% probability thermal ellipsoids, showing its dimeric butterfly shape from side and top views. Dipp groups and Et₂O are shown in wireframe, and hydrogen atoms are omitted.

Mössbauer spectra of solid 5-Et₂O and 5-THF are also comparable, with parameters of δ = 0.34 mm/s and |ΔE_Q| = 3.91 mm/s for 5-Et₂O and δ = 0.34 mm/s, |ΔE_Q| = 3.73 mm/s for 5-THF (Figures S50 and S51). Mössbauer spectra of 5-THF differ, however, between the solid state and a frozen THF solution. Specifically, there is a higher isomer shift of 0.45 mm/s for the frozen-solution spectrum, which may suggest that the dimer is broken up with THF replacing the bridging sulfur (Figure S80). The frozen solution also shows an increased quadrupole splitting and marked doublet asymmetry (|ΔE_Q| = 4.26 mm/s, Γ_L = 0.73 mm/s, Γ_R = 0.49 mm/s). The ¹H NMR spectrum of 5-Et₂O in THF-d₈ shows six broad signals between 12 and −81 ppm (Figure S12). The number and integration of the peaks is consistent with C_2v symmetry; a seventh expected peak integrating to 4H was not resolved, likely due to broadening. This is also consistent with the dimer breaking up, although it is also possible that there is some other dynamic phenomenon. The ¹H NMR spectra of 5-Solv in C₆D₆ have at least 12 peaks with considerable broadness, which prevented further analysis but points to a lowered symmetry in the dimer.

Cyclic voltammetry of 5-Solv showed three irreversible reduction events (Figure S62). An attempt to reference the reduction potentials to a ferrocene internal standard was unsuccessful due to reactivity with Fc and Fc⁺. The lack of electrochemical reversibility may be due to large structural changes occurring at different oxidation states. Undeterred, we proceeded to stoichiometrically reduce 5-Solv with 1 equiv of KC₈ per iron at −78 °C in THF, which resulted in a deep purple solution. Crystallization from diethyl ether at −40 °C gave iron(II) complex 6 in 46% yield. Its X-ray crystal structure, shown in Figure 6, shows a tetrameric assembly in the solid state, with K⁺ bridging S atoms and Dipp groups.

Figure 6.

Molecular structure of 6. Thermal ellipsoids are displayed at the 50% probability level. Dipp groups are shown in wireframe and hydrogen atoms are omitted. Two-thirds of the asymmetric unit is shown.

The ¹H NMR spectrum of 6 in THF-d₈ shows at least 19 resonances between −50 and +15 ppm, reflecting C₁ symmetry in solution. We were unable to assign resonances to specific proton environments due to peak broadness and overlapping in the 0 to 8 ppm region (Figure S13). Despite the complexity of the ¹H NMR spectrum and crystal structure, the solid-state Mössbauer spectrum of 6 (Figure S52) has a single quadrupole doublet with an isomer shift similar to that of the starting material (0.34 mm/s for 6 vs. 0.35 mm/s for 5-Et₂O), but a smaller quadrupole splitting (1.87 mm/s for 6 vs. 3.91 mm/s for 5-Et₂O). The Mössbauer spectrum suggests that all iron sites are equivalent.

Next, we tested 6 for the ability to bind N₂ at low temperatures using variable-temperature UV-visible spectroscopy. Negligible spectral changes were observed between 25 and −100 °C in THF, Et₂O, or toluene (Figure S81a–S81c). Addition of stoichiometric amounts of 18-crown-6 slightly shifted the absorption maxima, but the resulting species also did not exhibit notable temperature-dependent spectral changes (Figure S81d–S81f). Thus, it does not appear that 6 undergoes speciation changes or N₂ binding at these concentrations and temperatures. Further reduction of the iron(II) tetrameter 6 led to a silent NMR spectrum and a broad, featureless UV-Vis spectrum. The products formed in this reaction are unknown, and attempts to isolate or identify the species present were unsuccessful.

Synthesis of Monomeric Iron Complexes With S and N Donors.

To study monomeric complexes of our SCS iron framework in the absence of strong phosphine donors, we next treated 5-Et₂O with ligands containing S and N donors, which gave the products in Scheme 5 within 1 hour at room temperature. Their syntheses, NMR spectra, and IR spectra are described here, and their magnetism, EPR spectra and electronic structures are described in following sections.

Scheme 5.

Synthesis of complexes 7–10. L = triply deprotonated trianionic ligand 1; Ar = 2,6-dimesitylphenyl.

Addition of a bulky thiolate (potassium 2,6-dimesitylphenylthiolate) to 5-Et₂O in THF generated complex 7 (Scheme 5, top). Addition of 18-crown-6 followed by crystallization furnished 7-crown (Figure S72). 7-Crown displays an axially coordinated THF molecule and a nearly square pyramidal geometry with τ₅ = 0.08. The Fe−C, average Fe−S_Pincer, and Fe−S_Ar, bond lengths are 2.0006(17) Å, 2.244(6) Å, and 2.2758(5) Å, respectively. Addition of Et₂O, arenes, or alkanes to solid 7 prior to the addition of 18-crown-6 led to a color change from orange to green, indicating the formation of the THF-free analogue (8) in non-coordinating solvents in 97% yield. Compound 8 was crystallized from toluene in the presence of 18-crown-6 as 8-crown (Figure 7, top, and Figure S73). The geometry is distorted from square planar, with τ₄ = 0.20. The Fe−S_Ar bond distance is 2.2448(8) Å, which is shorter than the Fe−S_Ar bond in 7-crown (Table 1). Unlike the structure of the THF adduct 7-crown, one mesityl group of THF-free 8-crown is twisted to cover an axial iron site; however, the long 3.06 Å distance between iron and the arene centroid is inconsistent with a direct electronic interaction. We previously observed a similar orientation of this thiolate ligand in another SCS iron complex.²⁸ We were unable to crystallize four-coordinate thiolate 8 without crown ether. The ¹H NMR spectrum of 8 in THF-d₈ has ten broad peaks between 15 and −63 ppm (Figure S14). Upon dissolution in THF-d₈, the color of 8 changes from dark brown to red, indicating that the THF-adduct 7 is likely formed. The number and integration of the signals suggests C_2v symmetry, which implies that the crystallographically-observed lack of symmetry caused by the bulky aryl thiolate is not preserved in THF-d₈ solution.

Figure 7.

Molecular structures of 8-crown, 9, and 10 at 50% probability thermal ellipsoids. Hydrogen atoms are excluded for clarity, with exception of the NH₃ hydrogen atoms of 10. The outer-sphere potassium ion in 8-crown is not shown.

Table 1.

Comparison of Fe−C, Fe−S, and Fe−X (X = SAr, N(TMS)₂, NH₃) bond lengths (in Å) in 7-crown, 8-crown, 9-crown, and 10. Ar = 2,6-dimesitylphenyl. Two sets of lengths for one parameter represent the two molecules in the unit cell.

Complex	Exogenous Ligand (X)	Fe−C	Average Fe−SPincer	Fe−X
7-crown	SAr	2.0006(17)	2.244(6)	2.2758(5)
8-crown	SAr	1.999(3)	2.250(5)	2.2448(8)
9-crown	N(TMS)2	2.002(6) 2.010(5)	2.234(5) 2.241(7)	1.882(5) 1.902(4)
10	NH3	1.973(3) 1.971(3)	2.24(1) 2.25(1)	2.065(3) 2.084(3)

Amide complex 9 was synthesized in 87% yield by treating 5-solv with KN(TMS)₂, and it was crystallized with and without 18-crown-6 to give structures of 9-crown (Figure S75) and 9 (Scheme 5, middle, and Figure S74), respectively. Crown-free 9 is unstable as a solid at room temperature and in solution even at −40 °C, as evidenced by the appearance of multiple new peaks between 11 and −43 ppm in its NMR spectrum after a few days (Figure S17). Once this unknown impurity formed, we have been unable to remove it by recrystallization or washing. 9-Crown, however, is more stable. The crystallographic structures of 9 and 9-crown lack axial solvent coordination (despite 9 being crystallized from THF), and their color is the same in coordinating and non-coordinating solvents. These observations suggest that solvent coordination to the N(TMS)₂ adduct is unfavorable, possibly because the TMS groups block the axial sites.

The crystallographic structure of 9 displays a square planar molecular geometry at iron with τ₄ = 0.09–0.11. Fe–N bond lengths in 9 are Fe1–N14 = 1.915(3) and Fe2–N18 = 1.899(3), which are distinguishable. Interestingly, the structure of 9 shows a potassium cation lying in the pincer plane, with a K1−S11 distance of 3.133(1) Å and K2−S15 distance of 3.126(1) Å. The Fe−S bond lengths on the side with potassium are 2.2459(9) Å (Fe1) and 2.2463(9) Å (Fe2), which are indistinguishable. The Fe−S bond length is, however, significantly shorter on the side of the molecule without the close S−K contact, with a difference of 0.0065(12) Å at Fe1 and 0.0080(12) Å at Fe2. Metrical parameters for 9-crown are shown in Table 1. The ¹H NMR spectra of 9 and 9-crown in THF-d₈ have eight and nine, respectively, paramagnetically-shifted peaks between 19 and −68 ppm (Figures S16 and S18), consistent with C_2v symmetry. Their chemical shifts are nearly the same.

Treating 5-solv with 1 equiv of NH₃ gas in THF furnished ammine complex 10 in 86% yield (Scheme 5, bottom). In the crystallographic structure, there are two molecules of 10 in the asymmetric unit, and each has an axially-coordinated THF that gives a distorted square pyramidal geometry with τ₅ = 0.09–0.14. The Fe−N bond lengths are 2.084(3) Å (Fe1) and 2.065(3) Å (Fe2). The molecule containing Fe2 shows a hydrogen bond between one NH₃ proton and one uncoordinated THF. Hence, the shorter Fe−N length in Fe2 may be due to the hydrogen bond, which increases the basicity of the NH₃. While the Fe−S bond lengths of 10 are similar to those in 8-crown and 9-crown, the Fe−C bond length is significantly shorter in 10 (Table 1). This difference may arise because the thiolate and amido ligands exert a larger trans influence than the ammine.^60,61 The solid-state IR spectrum of 10 shows four weak bands at 3352, 3299, 3236, and 3159 cm⁻¹ that are assigned as N–H stretches (Figure S44). Their frequencies are not suggestive of any significant coordination-induced N−H bond weakening. The ¹H NMR spectrum of 10 in THF-d₈ has eight resonances between 175 ppm and −78 ppm, consistent with C_2v symmetry (Figure S19). Since the complex is C_s symmetric with THF coordination on one face, the spectrum in THF-d₈ could indicate fast THF exchange on both sides of the complex. Its spectrum in C₆D₆ has severe broadening compared to THF-d₈, and there are at least 15 resonances (Figure S20). Assuming that coordinated THF has two different proton environments and that NH₃ rotation on the NMR timescale renders its protons equivalent, a C_s symmetric geometry (with the mirror plane perpendicular to the pincer) would predict 14 peaks. Thus, the number of peaks in the C₆D₆-solvated NMR spectrum suggests 10 has no symmetry in C₆D₆, though the reason for the low symmetry is not obvious.

Synthesis and Characterization of an Iron(II) Carbonyl Complex.

Treating a THF solution of iron(III) complex 5-Solv with 1 atm of CO gave little reaction, with 85% unreacted 5-Solv remaining after 2 hours at room temperature, as judged by NMR spectroscopy. In contrast, stirring iron(II) complex 6 under 1 atm of CO afforded 11 in 79% yield after 1 hour (Scheme 6). Analytically pure material was isolated by removing THF under vacuum and washing the pale-yellow solids with Et₂O. Material generated in this way gave a Mössbauer spectrum with a quadrupole doublet at δ = 0.01 mm/s and |ΔE_Q| = 0.99 mm/s (Figure S58). Low isomer shifts such as that of 11 are frequently observed in iron carbonyl complexes, reflecting the withdrawal of d-electron density from the iron nucleus by backbonding.⁶² The ¹H and ¹³C NMR spectra of 11 show 7 and 14 narrow peaks, respectively (Figures S21 and S22). These spectra are consistent with a C_2v-symmetric diamagnetic species. The two most-downfield ¹³C signals have a relative integration of 2:1 and are assigned to the CO ligands. They have comparable chemical shifts to the CO signals in an SNS-supported iron(II) tricarbonyl complex.³⁹

Scheme 6.

Synthesis of tricarbonyl complex 11. L = triply deprotonated trianionic ligand 1.

The solid-state IR spectrum of 11 shows three bands in the region expected for CO stretching at 2078, 2013, and 1988 cm^–1 (Figure S45). To compare our experimental spectrum with theoretical IR spectra of SCS-supported complexes with either two or three CO ligands, we used DFT calculations (BP86/ZORA-def2-TZVP) to optimize and predict the CO stretching frequencies of [LFe(CO)₂]⁻ and [LFe(CO)₃]⁻. The model with two CO ligands predicted two IR-active bands of equal intensity, while the model with three CO ligands predicted stretching frequencies whose number and relative intensities matched the experimental spectrum (Figures S84 and S85). Thus, we assign the bands at 2078, 2013, and 1988 cm^–1 to CO stretching vibrations in an octahedral tricarbonyl complex.

X-Ray Absorption Spectra of Thiolate Complex 8.

To gain insight into the Fe–S bonding in 8-crown, which has iron(III) in the most biomimetic S₃C coordination sphere, Fe K-edge X-ray absorption spectra were measured. The spectrum showed two pre-edge features at ca. 7112 and 7114 eV. TD-DFT calculations reproduced the pre-edge features of the experimental spectrum (Figure 8, bottom). Frontier quasi-restricted orbital (QRO) analysis showed three singly-occupied orbitals of primarily Fe 3d character with < 10% S 3p mixing and a LUMO composed of 39.3% Fe 3d character and 21.3% S 3p character (Figure 8, top). The greater covalency between the Fe and S ligands in the LUMO is responsible for its increased energy. Based on the QRO analysis, the first pre-edge feature at 7112 eV is assigned to excitations of Fe 1s electrons into the three nearly-degenerate singly-occupied orbitals, while the second feature at 7114 eV is assigned to an Fe 1s to LUMO excitation. As expected for iron in the +3 oxidation state, the large contributions of ligand S orbitals to the frontier QROs indicate a large degree of Fe–S bond covalency.

Figure 8.

Top: Frontier MO diagram of 8. Energies were obtained from the QROs (calculated using the B3LYP level of theory with the CP(PPP) basis set on Pd and ZORA-def2-TZVP(-f) on all other atoms). Ligand orbitals are labeled in red. Orbitals are plotted at an isovalue of 0.03 au with hydrogen atoms removed for clarity. Bottom: Overlay of the experimental (red) and TD-DFT calculated (black) Fe K-edge spectra for 8-crown. Black sticks represent individual transitions. The calculation employed the CP(PPP) basis set for Fe and the ZORA-def2-TZVP(-f) basis set for all other atoms.

Electronic Structures of S and N Donor Complexes from Magnetometry and EPR Spectroscopy.

The solution magnetic moments of the iron(III) complexes were measured at 298 K in THF-d₈, giving μ_eff(7/8) = 4.0 ± 0.1 μ_B, μ_eff(9) = 4.1 ± 0.1 μ_B, and μ_eff(10) = 4.1 ± 0.1 μ_B. These moments are consistent with a well-isolated S = 3/2 ground state. Note that in THF solution, thiolate complex 8 is in equilibrium with its THF adduct 7. The solid-state magnetic susceptibilities of 8, 9, and 10 are each consistent with an S = 3/2 ground state and little mixing of excited states (Figures S63–S65). To further evaluate the zero-field splitting (ZFS) parameters, we measured the low-temperature solution X-band EPR spectra of each in MeTHF and magnetization curves at ≤ 10 K under field strengths of 1 T to 7 T. The data were fit within the constraints of the usual spin Hamiltonian for an S = 3/2 spin system,

$$\widehat{H}=\left(g_x+g_y+g_z\right){\mu }_{\text{B}}\mathit{\text{B}}·\mathit{\text{S}}+D\left({\widehat{S}}_z^2-\frac{1}{3}S\left(S+1\right)\right)+E\left({\widehat{S}}_x^2-{\widehat{S}}_y^2\right).$$

Variable-temperature (3 K to 10 K) magnetization data of thiolate complex 8 showed nesting of curves obtained at different fields, with saturation around 1.6 μ_B (Figure S66, top). These data were fit to g = [2.30, 2.30, 1.89] (g_iso = 2.16) and D = −28(1) cm⁻¹; an adequate fit could not be obtained with positive D. At 10 K, complex 8 has an extremely broad EPR spectrum with poorly-resolved signals at g ≈ 5.7, 5.1, 4.5, 4.3, 2.2, 2.1, and 1.2 (Figure 9, top). The sharp set of signals near g = 2.00 is assigned to a S = 1/2 impurity (1%) with g = [2.08, 2.03, 2.00]. An acceptable simulation of the remainder of the EPR spectrum was obtained using S = 3/2, g and D values derived from the magnetization data, and E/D of 0.20. D strain of 28 cm⁻¹ (100% of D) was required to account for the broad spectrum, indicating that a wide distribution of ZFS parameters were sampled in the spectrum. EPR spectra with similarly strain-dominated line shapes have been observed for other S = 3/2 iron complexes.^63,64 In our system, the crystallographically-observed flexibility of the thiolate ligand may result in a mixture of conformers with varying ZFS (E/D between 0.1 and 0.2).

Figure 9.

X-band (frequency = 9.37 GHz) EPR spectra from 50 to 500 mT of thiolate complex 8 (top, green) at ~10 K, amide complex 9 (middle, purple) at ~5 K, and ammonia complex 10 (bottom, red) at ~10 K as 1 mM solutions in MeTHF. Each spectrum was collected using a 19 G modulation amplitude and a microwave power of either 0.064 mW (8) or 0.020 mW (9 and 10). Fits are plotted as black traces beneath their respective experimental spectrum. Signals marked by asterisks are assigned to impurities, which were modeled as S = 1/2 species for 8 and 10 in 1% relative concentration to the S = 3/2 species. The simulations with +D and −D for 10 are shown; the latter is preferred because the lowest-field component has an intensity that fits better.

Isofield lines from 2 to 10 K for amide complex 9 also show a field dependence, with saturation near 2.0 μ_B. These magnetization curves were fit to g = [2.15, 2.54, 2.33] (g_iso = 2.34) and D = −30(2) cm⁻¹ (Figure S66, middle). No reasonable fit was found using D > 0. The EPR spectrum at 5 K shows a sharp, intense resonance at g_eff = 6.35 with broader signals at g_eff = 2.55 and 1.57 (Figure 9, middle). These features were simulated as an S = 3/2 spin system with the g and D values from the magnetization, also introducing E/D = 0.333. Similar to the spectrum of 8, D strain of 5 cm⁻¹ (17% of D) was necessary to account for line shapes. Unresolved hyperfine coupling to the amido ¹⁴N nucleus may also contribute to broadening. There are small signals at g = 4.4 and 2 that may come from the unknown decomposition product discussed in the synthesis section.

The magnetization curves for ammonia complex 10 show a pronounced field dependence with saturation occurring around 2.7 μ_B. They were fit to g = [2.01, 2.01, 2.07] (g_iso = 2.03) and D = ±3.5(1) cm⁻¹ (Figure S66, bottom); in this case the magnetization data did not distinguish the sign of D. At 10 K, the EPR spectrum of 10 shows features at g_eff = 6.15, 4.75, and 3.40, while signals at 2.36, 2.07, 2.05, and 1.99 are assigned to impurities (Figure 9, bottom). The spectrum was simulated as S = 3/2 with E/D = 0.106 using g from the magnetization fit, and the use of the negative D value gave a better fit to the intensity of the g_eff = 6.15 component (see both simulations in Figure 9). An S = 1/2 impurity with g = [2.07, 2.05, 1.99] was included in the simulation in 1% relative abundance to the major S = 3/2 signal. A spectrum similar to that of 10 was reported for a square pyramidal ferric intermediate spin complex with E/D = 0.107.⁶⁵

The magnetization and EPR fit parameters are summarized in Table 2. Our magnetization data suggest that D < 0 for 8 and 9, and the intensity of the component at g = 6.15 in the EPR spectrum of 10 suggests that the sign of D is negative in this compound as well.

Table 2.

Key parameters for the simulations of the EPR spectra and magnetization data of complexes 8-10.

	g x , g y , g z	g iso	D	E/D
8	2.30, 2.30, 1.89	2.16	− 28(1)	0.200
9	2.15, 2.54, 2.33	2.34	− 30(2)	0.333
10	2.01, 2.01, 2.07	2.03	− 3.5(1)	0.106

Mössbauer Spectroscopy of S and N Donor Complexes.

Zero-field Mössbauer spectra of solid thiolate complexes 8 and 8-crown, N(TMS)₂ amide complexes 9 and 9-crown, and ammonia complex 10 were recorded at 80 K. The spectra of 8, 9, and 10 are shown in Figure 10, and fit parameters in Table 3. The isomer shifts and large quadrupole splitting values of 8, 8-crown, and 10 are within the expected ranges for intermediate spin iron(III).⁶² The spectrum of 9, however, shows an isomer shift of only 0.18 mm/s, a markedly low quadrupole splitting of 0.93 mm/s, and narrower lines with Γ = 0.3 mm/s. The Mössbauer parameters of 8-crown (δ = 0.30 mm/s, |ΔE_Q| = 4.03 mm/s) and 9-crown (δ = 0.17 mm/s, |ΔE_Q| = 0.88 mm/s) are nearly the same as their crown-free versions 8 and 9.

Figure 10.

Zero field Mössbauer spectra of solid 8, 9, and 10 at 80 K. Data are plotted as black circles and residual as gray trace. 8: δ = 0.31 mm/s, |ΔE_Q| = 3.88 mm/s, Γ_L = 0.61 mm/s, Γ_R = 0.54 mm/s. 9: δ = 0.18 mm/s, |ΔE_Q| = 0.93 mm/s, Γ_L = 0.31 mm/s, Γ_R = 0.33 mm/s; there is a 4% impurity with parameters matching 5-Et₂O. 10: δ = 0.35 mm/s, |ΔE_Q| = 4.28 mm/s, Γ_L = 0.55 mm/s, Γ_R = 0.43 mm/s.

Table 3.

Comparison between the experimental and calculated Mössbauer parameters for anionic 8, anionic 9, and neutral 10. The sign of the experimental quadrupole splitting is undetermined.

Complex	δ (mm/s)		ΔEQ (mm/s)
	Exp.	Calc’d	Exp.	Calc’d
8	0.31	0.28	|3.88|	+ 3.97
9	0.18	0.29	|0.93|	− 4.24
10	0.35	0.33	|4.28|	+ 4.34

In order to understand this difference, we calculated the Mössbauer parameters of 8, 9, and 10 (omitting the countercations) using DFT at the B3LYP/ZORA-def2-TZVP level of theory.^66–68 These potassium-free calculated parameters were in agreement with the experimental Mössbauer parameters of thiolate complex 8 and ammonia complex 10, but the calculated and experimental values for δ and |ΔE_Q| of amide complex 9 diverged by 0.11 mm/s and 3.18 mm/s, respectively (Table 3).

Since the crystal structure of 9 shows a potassium cation only 3.13 Å from one S atom in the SCS pincer, we hypothesized that including K⁺ may be necessary to properly describe its electronic structure. Thus, we started from the crystallographic structure of 9 and performed Mössbauer calculations on an optimized structure that includes the close K⁺ (Figure 11A). In the crystal structure, the potassium is also coordinated to a Dipp aryl group of a neighboring molecule of 9, and this second arene was modeled as benzene in the calculations. Calculations on this model gave predicted Mössbauer parameters in agreement with experiment for 9 despite a significant distortion of the pincer plane (Figure S86). Thus, replication of the Mössbauer parameters requires the inclusion of the potassium ion (Figure 11, comparing parts A and B).

Figure 11.

Structures viewed along the N−Fe−C bonds of three models of 9 after BP86/ZORA-def2-TZVP optimization, with their predicted Mössbauer parameters stated below. A: optimized structure with K⁺. B: optimized structure of the anion without K⁺. C: optimized structure with [K⁺(18-crown-6)(THF)₂]. Potassium is represented as a purple sphere.

Analogously, we performed Mössbauer calculations on an optimized model of 9-crown, which has a formally outer-sphere potassium with an Fe···K distance of 8.026 Å (Figure 11C). This structure gave δ = 0.16 mm/s and ΔE_Q = −0.91 mm/s, in agreement with experimental values for 9-crown (and very close to those of 9). The Löwdin charge on potassium was found to be −0.06 in this structure, suggesting significant delocalization of positive charge onto the coordinated THF and crown ether components.

We were surprised that the effect of potassium was maintained despite the crown ether. To test the isolated effect of a distant and free potassium ion, we changed the computational model. We removed the two THF molecules and the 18-crown-6 on the potassium in optimized 9-crown and repeated the Mössbauer calculations without changing the geometry (Figure S86). This model gave δ = 0.33 mm/s and ΔE_Q = −4.08 mm/s, which is much closer to that calculated for potassium-free, anionic 9. In contrast to the structure containing THF and 18-crown-6, the Löwdin charge on potassium was found to be +0.99 in this structure. These results indicate that in 9-crown, the THF and 18-crown-6 may be accepting most of the positive charge of the potassium cation, which then places enough positive charge close to the iron to alter its electronic structure. Charge reorganization, albeit in varying degrees, has been implicated as a stabilizing factor in the formation of alkali-crown complexes in previous theoretical studies.^69–70

To compare the electronic structures of the computational models whose Mössbauer parameters match experiment to those that do not, we plotted the frontier quasi-restricted orbitals (QROs) for models A (reproduces experimental Mössbauer parameters; optimized with K⁺) and B (does not reproduce experimental Mössbauer parameters; optimized without K⁺) in Figure 12. The QROs show that the shape of the doubly-occupied d orbital (DOMO) for model A resembles d_z² while the doubly-occupied d orbital for model B resembles d_yz. The d_z² DOMO has 3.7% Fe s character, while the d_yz DOMO has 0.0% Fe s character, explaining the significant change in the isomer shift. The ordering of the d_xz and d_xy orbitals is also switched, but the d_x²_−y² orbital is always higher in energy than the remaining four d orbitals due to its σ antibonding interactions with the SCS pincer. QRO analysis of the 18-crown-6 species (9-crown, Figure 11C) shows the same d orbital ordering as model A (optimized with K⁺), though d_z² is 0.90 eV more stabilized than d_xy in 9-crown (Figure S87).

Figure 12.

QRO plots at an isovalue of 0.03 au of the d-assigned orbitals for amide complex 9 from model A (left, optimized with K⁺) and model B (right, optimized without K⁺). Energies relative to the doubly-occupied orbital are listed below the d orbital assignment. The structures of the complex are in gray wireframe, and potassium is represented as a purple sphere.

Additional experiments were performed to probe the influence of the cation on the electronic structure of 9. First, we computationally sampled a wide range of potential locations for a positive charge to learn about the orientational dependence, but the results were not conclusive (details in Supporting Information, Figure S88). In addition, we sought to experimentally “remove” the potassium completely by reacting starting material 5-Et₂O with N(TMS)₂ in the presence of an excess (2 equiv per N(TMS)₂ group) of cryptand-2.2.2 (Scheme S1). The reaction conditions were otherwise identical to the syntheses of 9 and 9-crown. The Mössbauer spectrum of the dried crude reaction mixtures showed a mixture of two components: one with parameters close to 9 (δ = 0.17 mm/s, |ΔE_Q| = 0.82 mm/s) and one with δ = 0.37 mm/s and |ΔE_Q| = 3.27 mm/s (FigureS83). The parameters of the component with a larger isomer shift and quadrupole splitting than 9 are more similar to those predicted for anionic 9, which represents the extreme in which iron and K⁺ are too far apart to interact. However, we were unable to isolate this species for further study.

DISCUSSION

Relationship between Protonation State and Reduction Potential.

To our knowledge, bond dissociation free energies of thioamide SCS pincer compounds have not been reported. The N–H BDFE of 2 (56 kcal/mol) is substantially lower than the computed BDEs for free thioamides (90–100 kcal/mol), and its pK_a of ~21 indicates that the N−H proton is much less acidic than free thioamides, which lie in the range 11–15 (Figure 13).^71–73 It is seen from the Bordwell equation that the high pK_a of 2 raises the BDFE relative to organic thioamides, but the very negative reduction potential (E_1/2 = −1.42 V vs. Fc⁺/Fc) of the iron(II) center lowers the BDFE to a much larger extent. This potential is 1.0 V more negative than E_1/2 for the Fe^3+/2+ couple in a four-coordinate SNS pincer complex with an NHC as the fourth ligand, and is 1.4 V more negative than the analogous five-coordinate SNS scaffold with two THF molecules.⁴⁰ The more negative reduction potential in our system may result from the negatively-charged aryl pincer in place of the pyridine pincer, and also from the coordination of three strong phosphine donors.

Figure 13.

Summary of the effect of iron coordination on the bond energy and pK_a of the thioamide N−H site.

Although the N–H BDFE of 2 is significantly lower than those of free thioamides, it is comparable to other iron systems with N–H bonds in the supporting ligand. In a recent example, the N–H BDFEs of aminobenzenethiolate iron(II) complexes supported by a TACN ligand were determined by measurement of pK_a and E_1/2.⁷⁴ In the parent system, the Fe^3+/2+ redox couple was found at E_1/2 = − 0.84 V, and the pK_a was determined to be ≥ 19.5, giving a BDFE of ≥ 62 kcal/mol; bracketing showed an exact BDFE near 68 kcal/mol. Studies of nickel, platinum, and ruthenium SCS compounds also corroborate decreased N–H bond acidity and more negative reduction potential upon metalation, though these examples do not report bond energies.^{38, 75–77}

It is interesting that deprotonating a distant thioamide site gives a highly-reducing iron(II) species, because of the potential for creating electron-rich active sites for small molecule activation using mild methods. This supports a growing body of evidence that reduction potentials may be modulated by the protonation state of sites outside the vicinity of metals in metallocofactors like the nitrogenase FeMoco.^{50, 78}

Electronic Structure.

The EPR spectrum of 9 is notable for its high rhombicity. While the spectrum resembles those for some other S = 3/2 examples of well-defined iron complexes with large E/D,^79–80 the most salient comparison is to a square planar S = 3/2 cobalt(II) ONO pincer complex with a chloride as the fourth donor.⁸¹ Structurally, this dianionic complex is reminiscent of 9, with the ONO pincer coordinating through a central amido N donor and lithium counterions flanking the alkoxide arms of the pincer ligand. The EPR spectrum of this complex was also highly rhombic, with E/D ≥ 0.25, maximum g_eff of 7.35, and g_y > g_x, g_z. DFT calculated d orbital energies revealed an unusual doubly-occupied d_z² orbital, and the EPR spectrum was rationalized on the basis of strong spin-orbit coupling of the ground state with the d_z²(β) → d_yz and d_xy(β) → d_xz excited states. In our case, mixing of the ground state with an excited state created by promotion of d_z²(β) into the half-filled d_yz orbital could also account for the large deviation in g from g_e in 9.

Sulfur, nitrogen, or mixed sulfur/nitrogen donors like in 8-11 are found in several other non-porphyrin ferric intermediate spin complexes.^{79, 82–87} These examples all also have near square planar or square pyramidal geometries. Mössbauer spectra of ferric S = 3/2 complexes generally have isomer shifts around 0.3 mm/s and large quadrupole splitting.⁶² Though Mössbauer spectra of complexes 8 (thiolate) and 10 (ammine) show normal δ and |ΔE_Q| for this class of compounds, the spectrum of 9 (amide) is an outlier in line width, isomer shift, and quadrupole splitting. In regard to line widths, the 80 K spectra of both 8 and 10 consist of broadened asymmetric doublets. Broad, asymmetric doublets at 80 K in intermediate-spin iron(III) complexes are a marker of slow magnetic relaxation in the range of 10⁻⁸ s to 10⁻¹¹ s; i.e., faster than the nuclear excited state lifetime of ⁵⁷Fe (~10⁻⁸ s) but slower than the ~10⁻¹² s that would give negligible paramagnetic broadening.^{62, 86, 88} The smaller line widths of 9 relative to those in 8 and 10 indicate that paramagnetic relaxation is faster, which is attributable to its larger E/D that mixes the Kramers doublets.^89–90 To our knowledge, the quadrupole splitting values of 9 and 9-crown are the lowest reported for any intermediate-spin iron(III) complex, though anomalously small quadrupole splitting has also been reported in a few S = 2 iron(II) complexes with planar ligand environments.^91–92 Planar ligand scaffolds can induce large positive contributions to the electric field gradient (EFG) on the axis perpendicular to the ligand plane (usually assigned as the z axis), and the Mössbauer quadrupole splitting is proportional to the z-component of the EFG.^{62, 91} Interestingly, computational analysis of the cited systems with anomalously small ΔE_Q also revealed a doubly-occupied d_z² orbital, as we found for 9 (Figure 12). Thus, the small ΔE_Q values imply that the large positive contribution to the z-component of the electric field gradient by the planar ligand is counteracted by a large negative contribution from the d_z² ground state. Double occupation of the d_z² orbital also accounts for the decreased isomer shift (which decreases with increasing iron s-electron density) of 9 by greater mixing of the doubly-occupied 3d_z² with the 4s orbital,⁹³ which is corroborated by our DFT calculations.

The question that arises is then: why do amide complexes 9 and 9-crown show evidence for a d_z² ground state electronic structure while thiolate complexes 8 and 8-crown as well as ammonia complex 10 do not? First, we propose that the π-donating orbitals of the N(TMS)₂ ligand in 9 exert a stronger destabilizing influence on the π-symmetry iron d_xy and d_yz orbitals than the thiolate donor in 8 and certainly more than the ammine donor in 10. Our proposition of a stronger interaction in 9 than 8 is tied to its structure, which shows a short Fe−N bond and a nearly 90° angle between the N(TMS)₂ plane and the pincer plane, permitting excellent overlap between d_xy and the N p_x orbital and between the d_yz and sp²-like orbitals aligned roughly along the N−Si bonds. This highly π-donating ligand field closes the energy gap between d_z² and d_xz/d_yz in the typical square planar splitting pattern. This explanation alone does not account, however, for the extreme variability in the calculated Mössbauer parameters for 9 depending on whether there is a positive charge near iron. On the basis of our DFT results, we therefore also suggest that the d_z² orbital can be stabilized by the presence of a nearby positive charge. Though the complete description of the cation interaction with each of the d orbitals is not elucidated by our work, selective stabilization of the d_z² orbital may be a result of its shape. A positive charge in the plane of the pincer could interact with the torus-shaped portion of the d_z² orbital, and one along the z axis can interact with the lobes pointing above and below the pincer plane. The stabilizing interaction of charge with the lobes of d_z² is illustrated by the comparison between the QRO energies of model A in Figure 12 with optimized 9-crown (Figure S87). Model A has positive charge with non-zero projections both onto the z axis and xy-plane while 9-crown has positive charge that is along the z axis. Accordingly, the d_z² orbital is 0.35 eV lower in energy than the DOMO+1 in 9-crown versus model A.

The influence of the alkali metal cation is a distinctive aspect of the work reported here, and it emphasizes the potential tuning role of cations near a transition-metal center. In a relevant example, Yang has synthesized systems in which appended cations cause substantial electric fields,⁹⁴ which can have beneficial influences on catalysis.^95–96 Electronic structure calculations in these systems did not show that the orbital energies were differentially affected by the cation, leading to the conclusion that the cation mainly exerts an electrostatic effect. Tolman has shown the influences of nearby cations on O–H bond dissociation energies.⁹⁷ In more recent work, Tomson observed shifts in redox potentials ascribed to selective stabilization of a d_z² orbital, and supported this idea with DFT calculations.⁹⁸ Cation effects in these examples were typically assessed using redox potentials, but in our complexes we utilize Mössbauer spectroscopy, which can report the electronic structure of iron. Through the combination of spectroscopy and DFT, we are able to substantiate the presence of cation-dependent changes in orbital energetics. While the overall spin state is conserved throughout, a β electron shifts to a cation-stabilized d_z² orbital, manifested in Mössbauer spectra by low δ and |ΔE_Q|. In the future, this effect could be utilized to control reactivity that depends on the ordering of orbital energies.

The conclusions from our electronic structure studies of complexes 8 (thiolate), 9 (amide), and 10 (ammine) utilizing EPR spectroscopy, SQUID magnetometry, Mössbauer spectroscopy, and DFT calculations are summarized as follows. EPR spectroscopy in combination with variable-temperature magnetometry measurements demonstrated large negative D with E ≠ 0 in each case. Mössbauer spectra of 8 and 10 showed isomer shifts and quadrupole splitting values consistent with other known intermediate spin iron(III) complexes, and DFT could accurately predict these parameters without including a counterion in 8. The quadrupole splitting of 9, however, was anomalously low, and DFT predicted a large |ΔE_Q| when no counterion was included. Inclusion of potassium in the model of 9 gave calculated Mössbauer parameters in agreement with experiment, and the energetic ordering of d_z² relative to the other d orbitals is likely to determine whether a typical large |ΔE_Q| versus an atypical small |ΔE_Q| is observed. Complex 9 is responsive because it is poised near a transition between two electronic states, which arises from the relatively low distortion from planarity, strong π donation from the amide ligand, and short Fe−N bond. Clearly, the SCS scaffold gives access to a range of electronic structures depending on the identity of the fourth donor ligand and on the cation location.

Comparisons to the Nitrogenase FeMoco.

The thioamide supporting ligand has the same donor atoms as the iron sites in the FeMoco, and we briefly explore the comparison of molecular and electronic structures. The most thoroughly-characterized state of the FeMoco is the enzyme resting state (E₀), and in this structure, the high-spin belt irons are bridged by sulfide ligands and share a central carbide, leading to a 3S/1C environment.¹² Our complex 8 likewise has an iron center ligated exclusively by three S donors and one C donor. Its Fe−C bond length of 2.00 Å is very close to the 2.01 Å Fe−C distance in the FeMoco.^{13–14, 99} The Fe–S lengths, which range from 2.24 to 2.25 Å, are also highly similar to Fe−S bond lengths in the FeMoco (2.25 to 2.27 Å). Despite being in an intermediate spin state, these bond distances are closer to those in the resting state than the two previously-reported (SCS)iron complexes, which were in lower oxidation states and thus had longer Fe−C/S bonds.^{26, 28}

The final state of the enzyme during catalysis is an iron-ammonia adduct (E₈).¹⁰⁰ Because turnover from E₈ to E₀ is redox neutral, ammonia would be expected to bind to either the iron(II) or iron(III) oxidation states.¹⁰⁰ We were able to isolate the stable ammonia adduct 10 in high yield in an intermediate-spin iron(III) oxidation state, thus demonstrating the feasibility of an iron(III) ammonia adduct in a C and S-ligand field.

CONCLUSIONS

A new pincer ligand bearing only sulfur and carbon donor atoms can support new iron complexes. The ligand scaffold has a proton-responsive site, gives access to unusual electronic structures, and mimics the sulfide and carbide donors found in the cofactor of nitrogenase enzymes. Adducts of thiolate, amide, ammonia, and CO demonstrate the versatility of this SCS pincer system to support a variety of nitrogenase-relevant donors, and induce an electronic structure that is highly sensitive to the iron surroundings.

Synopsis:

Iron(III) complexes of a proton-responsive SCS pincer ligand have unusual properties, which include finely balanced d-orbital energies in the electronic structure.