Synthetic, Spectroscopic and DFT Studies of Iron Complexes with Iminobenzo(semi)quinone Ligands: Implications for o-Aminophenol Dioxygenases

Abstract

The oxidative C-C bond cleavage of o-aminophenols by nonheme Fe dioxygenases is a critical step in both human metabolism (the kynurenine pathway) and the microbial degradation of nitroaromatic pollutants. The catalytic cycle of o-aminophenol dioxygenases (APDOs) has been proposed to involve formation of an Fe(II)/O₂/iminobenzosemiquinone complex, although the presence of a substrate radical has been called into question by studies of related ring-cleaving dioxygenases. Recently, we reported the first synthesis of an iron(II) complex coordinated to an iminobenzosemiquinone (ISQ) ligand, namely, [Fe(^Ph2Tp)(ISQ^tBu)] (2a; where ^Ph2Tp = hydrotris(3,5-diphenylpyrazol-1-yl)borate and ISQ^tBu is the radical anion derived from 2-amino-4,6-di-tert-butylphenol). In the current manuscript, density functional theory (DFT) calculations and a wide variety of spectroscopic methods (electronic absorption, Mössbauer, magnetic circular dichroism, and resonance Raman) were employed to obtain detailed electronic-structure descriptions of 2a and its one-electron oxidized derivative [3a]⁺. In addition, we describe the synthesis and characterization of a parallel series of complexes featuring the neutral supporting ligand tris(4,5-diphenyl-1-methylimidazol-2-yl)phosphine (^Ph2TIP). The isomer shifts of ~0.97 mm/s obtained via Mössbauer experiments confirm that 2a (and its ^Ph2TIP-based analogue [2b]⁺) contain Fe(II) centers, and the presence of an ISQ radical was verified by analysis of the absorption spectra in light of time-dependent DFT calculations. The collective spectroscopic data indicate that one-electron oxidation of the Fe²⁺–ISQ complexes yields complexes ([3a]⁺ and [3b]²⁺) with electronic configurations between the Fe³⁺–ISQ and Fe²⁺–IBQ limits (IBQ = iminobenzoquinone), highlighting the ability of o-amidophenolates to access multiple oxidation states. The implications of these results for the mechanism of APDOs and other ring-cleaving dioxygenases are discussed.

Introduction

A critical step in the microbial degradation of numerous aromatic compounds involves oxidative ring scission by a mononuclear nonheme Fe dioxygenase.^[1] Ring-cleaving dioxygenases have been shown to oxidize an impressive array of substrates, including catechols, protocatechuates,^[2] o-aminophenols,^[3] hydroquinones,^[4] and salicylates.^[5] The general catalytic strategy employed by these dioxygenases differs substantially from the “textbook” O₂-activation mechanism exemplified by the cytochrome P450s,^[6] methane monooxygenase,^[7] and α-ketoglutarate-dependent dioxygenases.^[8] In this latter set of enzymes, O₂ is used to generate an iron(IV)-oxo (ferryl) intermediate that performs the demanding hydroxylation of an aliphatic substrate. By contrast, extensive experimental and computational studies have revealed that the ring-cleaving dioxygenase mechanism does not involve high-valent Fe intermediates, as illustrated for the extradiol catechol dioxygenases (ECDOs) and o-aminophenol dioxygenases (APDOs) in Scheme 1.^{[2c, 9]} In both cases, the bidentate substrate coordinates to the Fe(II) center as a monoanionic ligand. Displacement of the bound H₂O molecules facilitates formation of an Fe/O₂ adduct capable of reacting directly with the bound substrate. The resulting peroxy-bridged intermediate undergoes a Criegee rearrangement to generate a lactone, which is hydrolyzed by the Fe-OH unit to provide the ring-opened product. Several of these intermediates were observed crystallographically in different subunits of the enzyme homoprotocatechuate 2,3-dioxygenase (HPCD).^[10]

Scheme 1.

While there is broad agreement concerning the generic mechanism shown in Scheme 1, the electronic structure of the critical Fe/O₂/substrate intermediate remains disputed. While Fe/O₂ adducts are normally described as ferric-superoxo species (such as intermediate I in Scheme 1), early studies by Lipscomb^[11] and Bugg^[12] lead to the proposal that ring-cleaving dioxygenases proceed instead via a superoxo-Fe(II)-(imino)semiquinone species (II). The radical character of the substrate was presumed to facilitate reaction with the superoxo ligand, thereby overcoming the large kinetic barrier to formation of the peroxy intermediate. This hypothesis was partly inspired by model studies that had demonstrated the “non-innocent” nature of metal-bound catecholates^[13] and amidophenolates.^{[14] [15] [16] [17] [18] [19]} Subsequent experiments with a substrate analogue featuring a cyclopropyl moiety as a radical probe indicated formation of a semiquinonate (SQ) radical during the catalytic cycle of MhpC, a well-studied ECDO.^[20] Density functional theory (DFT) analysis by Siegbahn further supported this electronic structure description.^[21] In addition, Emerson et al. demonstrated that a pair of ECDOs, Fe-HPCD and Mn-MndD, are equally active with either Fe or Mn in their active sites, despite the intrinsically different redox potential of the two metal ions.^[22] This result lead to the conclusion that the Fe (or Mn) oxidation state does not change during the catalytic cycle; instead, the metal ion only serves to conduct an electron from the bound substrate ligand to O₂, thus yielding intermediate II directly.

Despite this large body of evidence, recent studies have challenged the prevailing notion that a substrate-based radical is generated in the catalytic cycles of ECDOs (and, by extension, other ring-cleaving dioxygenases). Firstly, Lipscomb and coworkers succeeded in isolating the Fe/O₂ adduct of a HPCD mutant (H200N) bound to the “unactivated” substrate 4-nitrocatechol. Interrogation with EPR and Mössbauer spectroscopies revealed that this species contains an Fe³⁺–O₂^•– unit with S = 2 – inconsistent with the presence of a SQ ligand.^[23] A follow-up study of H200N HPCD with the native substrate resulted in the isolation and spectroscopic characterization of novel intermediate best described as a hydroperoxo-Fe(III)-semiquinone species (III in Scheme 1).^[24] However, a recent computational study by Ye and Neese suggested that this intermediate is not catalytically viable.^[25] Their calculations of the ECDO mechanism found no evidence that a SQ-containing intermediate (either II or III) is required for catalysis. Instead, they favor a mechanism in which the ferric-superoxo adduct (I) converts directly to a hydroperoxo-bridged species (Scheme 1). The kinetic barrier for this step is lowered by concomitant transfer of a proton from the second-sphere histidine residue to the nascent peroxo group.

The development of suitable synthetic models can help resolve the ambiguous electronic structures of enzymatic intermediates. In a recent communication, we described the synthesis and X-ray structure of the five-coordinate (5C) complex 2a (Scheme 2) that represents the first synthetic example of an Fe(II) center coordinated a biologically-relevant (imino)semiquinonate ligand.^[26] The supporting ligand in 2a is hydrotris(3,5-diphenylpyrazol-1-yl)borate (^Ph2Tp), which adequately models the facial coordination geometry and monoanionic charge of the 2-His-1-carboxylate coordination motif of most ring-cleaving dioxygenases.^[27] A combination of crystallographic, spectroscopic (absorption, EPR), and computational methods were used to confirm the existence of the ^tBuISQ ligand in 2a. The overall spin of 3/2 arises from antiferromagnetic (AF) coupling between the high-spin Fe(III) center (S = 5/2) and ISQ radical. Using similar techniques, we also examined the complex [3a]SbF₆ that arises from one-electron oxidation of 2a. The X-ray structural data for [3a]⁺ are consistent with the presence of an Fe³⁺–^tBuISQ unit with S = 2, although density functional theory (DFT) calculations suggested a certain degree of Fe²⁺–^tBuIBQ character (where ^tBuIBQ is the neutral iminobenzoquinone with tert-butyl substituents at the 4- and 6-positions).^[26]

Scheme 2.

In this manuscript, we seek to develop detailed electronic-structure descriptions of 2a and [3a]⁺ using an assortment of spectroscopic methods, including UV-vis-NIR absorption, Mössbauer, magnetic circular dichroism (MCD), and resonance Raman (rR) spectroscopies. In addition, we have prepared a parallel series of complexes (1b, 2b, and [3b]⁺) containing the facial N₃-donor ligand tris(4,5-diphenyl-1-methylimidazol-2-yl)phosphine^[28] (^Ph2TIP; Scheme 2). X-ray structures of complexes [1b]BPh₄ and [3b](OTf)₂ are presented to complement those already published for 1a, 2a, and [3a]SbF₆. By employing this neutral supporting ligand, we are able to evaluate the role of coordination environment in modulating the oxidation states of the Fe center and ISQ ligand. In all cases, the spectroscopic data were analyzed with the aid of DFT calculations. By elucidating the spectroscopic signatures of these synthetic complexes, we anticipate that our results will assist in the interpretation of comparable data from the biological systems, thereby allowing researchers to determine whether Fe–(I)SQ species are viable intermediates in the catalytic cycles of ring-cleaving dioxygenases.

Results and Analysis

1. Preparation, Structural Characterization, and Electrochemical Properties

1.A. Fe(II) Complexes

The synthesis and X-ray structure of [Fe²⁺(^Ph2Tp)(^tBuAPH)] (1a) were reported in our previous paper.^[26] The analogous complex [1b]⁺ based on the neutral ^Ph2TIP ligand was generated by mixing [Fe²⁺(^Ph2TIP)(OTf)₂]^[28] and 2-amino-4,6-di-tert-butylphenol (^tBuAPH₂) with one equivalent of triethylamine in CH₂Cl₂; recrystallization from toluene/pentane provided [1b]OTf as a yellow solid. Following counteranion metathesis with NaBPh₄, X-ray quality crystals of [1b]BPh₄ were obtained by layering a concentrated 1,2-dichloroethane (DCE) solution with MeOH. The resulting structure (Figure 1) contains two symmetrically independent [1b]⁺ units with nearly identical geometries. Relevant structural parameters for 1a, [1b]BPh₄, and 4a are provided in Table 1. Similar to 1a, complex [1b]⁺ features a 5C monoiron(II) center in a distorted trigonal-bipyramidal coordination environment (τ = 0.60^[29]). The amino and phenolate donors of the bidentate ^tBuAPH ligand are found in axial and equatorial positions, respectively (Figure 1). The average Fe-N_TIP bond distance of 2.19 Å in [1b]⁺ is ~0.04 Å longer than the average Fe-N_Tp distance in 1a, reflecting the weaker donating ability of neutral ^Ph2TIP relative to anionic ^Ph2Tp. In both 1a and [1b]⁺, the Fe-N/O bond lengths are indicative of high-spin Fe(II) centers, consistent with the presence of paramagnetically-shifted peaks in the corresponding ¹H NMR spectra (Figure S1).

Figure 1.

Thermal ellipsoid plots (50% probability) derived from the X-ray structures of [1b]BPh₄•1.5DCE (top) and 4a•4CH₂Cl₂ (bottom). Non-coordinating solvent molecules, counteranions, and most hydrogen atoms have been omitted for clarity. Key metric parameters are provided in Table 1.

Table 1.

Selected Bond Distances (Å) and Bond Angles (deg) from the X-ray Structures of 1a•MeCN•0.5DCE, 1b[BPh₄] • 1.5DCE, and 4a•4CH₂Cl₂.

	1a•MeCN•0.5DCE	1b[BPh4] • 1.5DCE	4a•4CH2Cl2
Fe1-N1	2.101(1)	2.175(2)	2.121(2)
Fe1-N3	2.127(1)	2.175(2)	2.158(2)
Fe1-N5	2.223(1)	2.222(2)	2.236(2)
Fe1-O1	1.931(1)	1.929(2)	1.920(2)
Fe1-N7	2.214(1)	2.229(2)	2.335(2)
N1-Fe1-N3	92.82(5)	92.50(7)	100.11(9)
N1-Fe1-N5	83.30(5)	85.88(7)	80.90(9)
N3-Fe1-N5	86.11(5)	84.43(6)	81.79(9)
O1-Fe1-N1	136.61(5)	134.21(7)	131.71(9)
O1-Fe1-N3	129.47(5)	130.73(7)	127.63(9)
O1-Fe1-N5	105.94(4)	109.45(6)	109.69(9)
O1-Fe1-N7	80.76(4)	80.24(6)	78.07(8)
N1 -Fe-N7	92.60(5)	88.33(7)	100.11(9)
N3 -Fe-N7	88.68(5)	87.90(7)	98.07(9)
N5 -Fe-N7	173.20(5)	170.17(7)	170.40(9)
τ-valuea	0.61	0.60	0.65

The ^Ph2Tp-based complex 4a contains the DMAP(1-) ligand – the N,N-dimethylated version of ^tBuAPH (Scheme 2). This complex is intended serve as a “control”, since methylation of the – NH₂ group hinders conversion to the ISQ state. The overall structure of 4a, as revealed by X-ray crystallography (Figure 1), is quite similar to 1a. The most prominent difference is the 0.12 Å elongation of the Fe1-N7 bond in 4a relative to 1a (Table 1), presumably due to steric repulsion between the −NMe₂ group and 3-Ph substituents of the ^Ph2Tp ligand. Additionally, the plane of the DMAPH ligand is bent away from the O1-Fe-N7 chelate ring by 26°, whereas the two planes are nearly colinear in 1a and [1b]⁺.

1.B. Fe(II)-Iminobenzosemiquinonate Complexes

Previously, we demonstrated that reaction of 1a with 2,4,6-tri-tert-butylphenoxy radical (TTBP^•) provides the complex [Fe(^Ph2Tp)(^tBuISQ)] (2a) via H-atom transfer from the ^tBuAPH –NH₂ donor to the organic radical.^[26] In the X-ray structure of 2a, the ^tBuISQ ligand exhibits short O-C and N-C bond distances and a “four long/two short” distortion of the C-C bonds (Table S1) – well-known traits of the iminobenzosemiquinone oxidation state as elucidated by Wieghardt^[15-17], Kaim,^[14] and others.^[19] Further spectroscopic evidence in support of this electronic-structure description is provided later in the manuscript.

Treatment of [1b]OTf with TTBP^• in CH₂Cl₂ produces a light green chromophore [2b]⁺ with an absorption spectrum similar to the one collected for 2a (Figure 2). In addition, the EPR spectra of [2b]⁺ and 2a are nearly identical; both exhibit a sharp peak at g ~ 6.5 and a broad derivative feature centered at g < 1.8 (Figure S2). Such spectra are characteristic of S = 3/2 systems with large, negative D-values and moderate rhombicities (E/D = 0.24 and 0.18 for 2a and [2b]⁺, respectively). Unfortunately, despite repeated attempts, it was not possible to obtain X-ray quality crystals of [2b]⁺. However, the strong resemblance between the spectroscopic (UV-vis/EPR) features of 2a and [2b]⁺ suggests similar Fe²⁺–ISQ electronic configurations – an assumption verified by analysis with Mössbauer spectroscopy (vide infra).

Figure 2.

UV-vis-NIR absorption spectra obtained in CH₂Cl₂ at room temperature.

1.C. Complexes [3a]SbF₆ and [3b](OTf)₂

Oxidation of 1a and [1b]OTf with two equivalents of AgX (X = SbF₆ or OTf) in CH₂Cl₂ gives rise to complexes [3a]SbF₆ and [3b](OTf)₂, respectively, that each display a distinctive absorption band near 780 nm (12,800 cm⁻¹; Figure 2). The magnetic moments of these complexes are near 5.0 μ_B – typical for S = 2 paramagnets – and the EPR spectrum of [3b]²⁺ exhibits a 4S signal at g = 8.7 (Figure S3). The structural data for [3a]SbF₆ reported in our earlier study suffers from large uncertainties in bond distances (3σ ~ 0.035 Å in C-C distances), which hindered evaluation of the ligand oxidation state based on geometric parameters. As noted in the introduction, DFT calculations suggested an electronic structure intermediate between Fe³⁺–^tBuISQ and Fe²⁺–^tBuIBQ.^[26]

Crystals of [3b](OTf)₂ were obtained by layering a CH₂Cl₂ solution with pentane; the resulting X-ray structure is shown in Figure 3, and relevant metric parameters for [3a]SbF₆ and [3b](OTf)₂ are compared in Table 2. The σ-values for the [3b](OTf)₂ bond distances are significantly smaller than those in the [3a]SbF₆ structure. Like [3a]⁺, complex [3b]²⁺ features a distorted trigonal-bipyramidal coordination geometry (τ = 0.57) with O1 occupying an axial position. The O1-C1, N7-C2, and C-C bond distances of the O,N-donor ligand are fully consistent with ligand-based oxidation. Recently, Brown developed a helpful procedure for assessing the oxidation state of o-amidophenolate ligands based on their metric parameters.^[30] Using this method, the ligands in [3a]⁺ and [3b]²⁺ have estimated charges of −0.54(8) and −0.48(10) – approximately half-way between the ISQ^1– and IBQ⁰ limits (Table 3). Further insights into the Fe and ligand oxidation state of [3a]⁺ and [3b]²⁺ are provided by spectroscopic and computational studies described below.

Figure 3.

Thermal ellipsoid plot (50% probability) derived from the X-ray structure of [3b](OTf)₂•CH₂Cl₂. Non-coordinating solvent molecules, counteranions, and most hydrogen atoms have been omitted for clarity. Key metric parameters are provided in Table 2.

Table 2.

Selected Bond Distances (Å) from the X-ray Structures of [3a]SbF₆•0.5DCE and [3b](OTf)₂•1.5CH₂Cl₂.

	[3a]SbF6	[3b](OTf)2
Fe1-N1	2.071(7)	2.077(3)
Fe1-N3	2.038(7)	2.089(4)
Fe1-N5	2.134(6)	2.093(4)
Fe1-O1	2.082(6)	2.037(3)
Fe1-N7	2.017(8)	2.013(4)
O1-C1	1.264(11)	1.287(5)
N7-C2	1.333(11)	1.314(6)
C1-C2	1.473(12)	1.483(6)
C2-C3	1.423(12)	1.429(6)
C3-C4	1.345(15)	1.340(7)
C4-C5	1.433(16)	1.448(7)
C5-C6	1.375(14)	1.370(7)
C1-C6	1.439(12)	1.427(7)

Table 3.

Experimental and DFT-Computed Properties of Selected Complexes.

			1a	[1b]+	2a	[2b]+	[3a]+,a	[3b]2+,a
Metric parameters	Fe-NTp/TIP ave (Å)	exp DFT	2.150 2.215	2.191 2.240	2.137 2.188	2.233	2.081 2.077	2.086 2.122
	O1-C1 (Å)	exp DFT	1.345 1.330	1.340 1.338	1.285 1.276	1.277	1.264 1.278	1.287 1.273
	N7-C2 (Å)		1.451 1.454	1.455 1.460	1.328 1.335	1.337	1.333 1.329	1.314 1.326
	Ligand charge	exp DFT	—	—	–0.69 –0.72	–0.73	–0.54 –0.62	–0.48 –0.54
Mulliken spins	Fe	DFT	+3.78	+3.76	+3.77	+3.76	+3.94	+3.93
	O,N-ligand	DFT	+0.14	+0.18	–0.86	–0.82	–0.30	–0.15
Mössbauer	δ (mm/s)	exp DFT	1.06 0.96	1.06 / 1.14 0.95	0.97 0.86	0.95 / 0.99 0.91	0.73 0.57	0.64 0.62
	ΔEQ (mm/s)	exp DFT	2.52 2.00	2.08/2.93 1.86	3.5 2.93	1.9b / 2.5 2.09	2.33 0.71	1.94 1.34

^aDFT calculation employed the BP functional

^bAverage ΔE_Q-value for the dominant species in the [2b]⁺ MB spectrum.

1.D. Electrochemical Properties

Our original intent in performing chemical oxidations of 1a and [1b]OTf was to generate the corresponding ferric complexes [Fe³⁺(^Ph2Tp)(^tBuAPH)]⁺ ([1a^ox]⁺) and [Fe³⁺(^Ph2TIP)(^tBuAPH)]²⁺ ([1b^ox]²⁺). However, treatment of 1a and [1b]OTf with a single equivalent of one-electron oxidants (such as acetylferrocenium, [N(C₆H₄Br-4)₃]⁺, or Ag⁺) instead produced 0.5 equivalent of [3a]⁺ and [3b]²⁺, respectively. Indeed, titrations of 1a and [1b]OTf with acetylferrocenium revealed a linear increase in the intensity of the [3a]⁺/[3b]²⁺ absorption features up to two equivalents of oxidant, indicating that the Fe(II) precursors exclusively undergo two-electron oxidations. This situation generally occurs when the product of the initial one-electron transfer undergoes a chemical change to yield a species that is more reducing than the starting complex.

To better understand this phenomenon, cyclic voltammetric studies of the ^Ph2Tp series were conducted in CH₂Cl₂ with 100 mM [NBu₄]PF₆ as the supporting electrolyte.^[31] All redox potentials are referenced to the ferrocenium/ferrocene couple (Fc⁺/Fc). As shown in Figure 4, 1a undergoes an irreversible oxidation at −15 mV, followed by a cathodic wave at −490 mV with half the current intensity of the anodic wave. We therefore concluded that the oxidation at −15 mV corresponds to the two-electron process 1a → [3a]⁺ + H⁺, while the subsequent reduction corresponds to one-electron reduction of [3a]⁺ to 2a. This conclusion was confirmed by the CV of 2a (Figure 4), which shows a quasi-reversible [3a]⁺/2a couple with E_1/2 = −380 mV (ΔE = 140 mV). Thus, 2a is indeed more reducing than 1a, accounting for the behavior described above.

Figure 4.

Cyclic voltammograms of 1a, 2a, and 4a. Data was collected in CH₂Cl₂ with 100 mM (NBu₄)PF₆ as the supporting electrolyte. The scan rate was either 100 mV/s (2a and 4a) or 200 mV (1a).

Scheme 3 summarizes the electrochemical properties of these complexes. One-electron oxidation of 1a presumably yields the ferric complex [1a^ox]⁺, although this transient species is never observed. Instead, [1a^ox]⁺ quickly sheds a proton to the surrounding medium to give 2a, since the change in Fe oxidation state dramatically lowers the pK_a of the coordinated –NH₂ donor. The fact that E_p,a([1a^ox]⁺/1a) > E_1/2([3a]⁺/2a) ensures that 2a is not stable in this environment; instead, it disproportionates to give 0.5 equivalent each of 1a and [3a]⁺ (Scheme 3B). Thus, the only way to convert 1a into 2a is to employ an H-atom abstracting agent like TTBP^• that is not an effective one-electron oxidant. Our results also indicate that deprotonation of the O,N-ligand dramatically lowers its redox potential, thus favoring ligand-based over metal-based oxidation.

Scheme 3.

The electrochemical behavior of 4a is straightforward since the DMAPH ligand cannot easily participate in proton or electron transfers. This complex displays a quasi-reversible redox event with E_1/2 = −15 mV (ΔE = 130 mV) that corresponds to the Fe³⁺/Fe²⁺ couple (Figure 4). The 4a potential can serve as an approximate value for the irreversible [1a^ox]⁺/1a couple. The free ^tBuAPH₂ and DMAPH ligands are irreversibly oxidized at more positive potentials of +280 and +450 mV, respectively, in CH₂Cl₂ (Figure S4).

2. Spectroscopic and Computational Analysis of Electronic Structures

2.A. Geometric and Electronic Structures of DFT-Optimized Models

Since the spectroscopic data presented below are frequently interpreted with the assistance of DFT calculations, it is necessary to first describe the molecular and electronic structures of our computational models. We employed truncated versions of the complexes in which the tert-butyl groups of the O,N-donor were replaced with methyl groups and the phenyl substituents at the 5-positions of the pyrazole (^Ph2Tp) and imidazole (^Ph2TIP) rings were removed. Unless otherwise noted, all calculations employed the B3LYP functional. As noted in our earlier study, the most stable geometry-optimized structure of 2a was obtained using the broken symmetry (BS) methodology^[32] and an overall spin of 3/2; this model provided metric parameters in excellent agreement with the crystallographic data (Table S1). The Mulliken spin populations, listed in Table 3, support the view that the electronic structure of 2a should be described as a high-spin Fe(II) center (3.77 α-spins) AF-coupled to an ISQ-based π-radical (0.86 β-spins). The exchange coupling constant (J) has a computed value of −223 cm⁻¹, based on the Yamaguchi approach (H = −2JS_A•S_B).^[33] The AF coupling is mediated by a nonorthogonal pair of singly-occupied molecular orbitals (SOMOs) with opposite spins, shown in the qualitative MO diagram in Figure 5. The relevant magnetic orbitals for 2a are the α-Fe(d_xy)- and β-ISQ(π*)-based MOs that display 33% spatial overlap – indicative of a ligand-based radical with rather weak interactions with the Fe(II) center. The structure and bonding scheme of [2b]⁺, also computed with the BS approach, are nearly identical to the corresponding 2a model (Tables 3 and S1), consistent with the spectral similarities between 2a and [2b]⁺ (vide supra).

Figure 5.

Qualitative molecular orbital diagrams of 2a (left) and [3a]⁺ (right) obtained from broken-symmetry DFT calculations. Isosurface plots for important MOs are provided, along with the overlap (S) between corresponding magnetic orbitals.

While the optimized [3a]⁺ geometry obtained with the B3LYP functional agrees reasonably well with the crystallographic data, the Fe-N/O and intraligand bond distances in the DFT-structure of [3b]²⁺ structure deviate significantly from the experimental values (Table S2). Better agreement was obtained with the BP functional for both [3a]⁺ and [3b]²⁺, and therefore our analysis of these complexes has employed the BP-derived structures. For each complex, DFT calculations converge to the same S = 2 state irrespective of whether the BS approach is employed. The Mulliken spin populations appear to favor an Fe²⁺–^tBuIBQ description, as the Fe center carries approximately four unpaired spins and the ligand-based spin density is reduced relative to 2a/[2b]⁺ (Table 3). The β-HOMO of [3a]⁺ contains significant amounts of both Fe (39%) and O,N-ligand (53%) character (Figure 5), indicative of intermediate Fe^2+/3+ and IBQ/ISQ oxidation states (the corresponding values for [3b]²⁺ are 53 and 37%). In addition, the spatial overlap between the pair of magnetic orbitals involved in the spin-coupling interaction is ~80%, which reflects a high degree of spin pairing (as opposed to the “diradical” situation of 2a). However, these computational results should be viewed with caution, since the experimental metric parameters for [3a]⁺/[3b]²⁺ are not consistent with a pure IBQ description, as evident in the ligand charges shown in Table 3. Spectroscopic studies are therefore required to properly evaluate the electronic structures of [3a]⁺ and [3b]²⁺.

Finally, we generated DFT models of the hypothetical Ga(III) complexes, [Ga³⁺(^Ph2Tp)(ISQ)]⁺ (Ga-ISQ) and [Ga³⁺(^Ph2Tp)(IBQ)]²⁺ (Ga-IBQ). The Ga(III) ion has been employed in previous computational studies as a closed-shell analog of Fe(III) due to similarities in charge and ionic radius.^[34] Since Ga(III) is not redox active, the Ga-ISQ and Ga-IBQ models allow us to assess the electronic and spectroscopic properties of “pure” ISQ and IBQ ligands without complications from the paramagnetic Fe center.

2.B. Mössbauer Spectroscopy

Mössbauer (MB) spectroscopy has proven capable of providing definitive assessments of Fe oxidation states in complexes with non-innocent ligands. Figure 6 displays low-temperature (6 K) MB spectra collected in an applied field of 0.04 T for polycrystalline samples of 1a, 2a, and [3a]SbF₆; the corresponding spectra of the ^Ph2TIP-based “b series” are provided in Figure S5. MB parameters summarized in Tables 3 and S3. Complex 1a displays a single doublet with an isomer shift (δ) of 1.06 mm/s and quadrupole splitting (ΔE_Q) of 2.52 mm/s, consistent with the presence of a conventional high-spin Fe(II) center.^[35] In contrast, the MB spectrum of [1b]OTf features two doublets in a ~2:1 ratio; the parameters for both species are typical of high-spin Fe(II) complexes: δ = 1.06 and 1.14 mm/s and ΔE_Q = 2.08 and 2.93 mm/s for the major and minor components, respectively. The two observed species likely correspond to conformational isomers of [1b]⁺ that lie at different points along the continuum between square-pyramidal and trigonal-bipyramidal geometries. In support of this conclusion, the X-ray structure of [1b]⁺ (vide supra) features two independent complexes in the unit cell with distinct τ-values of 0.60 and 0.55.

Figure 6.

Mössbauer spectra of complexes 1a, 2a and [3a]SbF₆ recorded at 6 K in an applied field of 0.04 T. The parameters indicated represent the principal species discussed in the text. The spectrum of sample 2a contains a contribution from 1a, amounting to 30% of the total Fe in the sample (inner doublet, indicated by the bracket). The spectrum of [3a]⁺ shown in this figure was obtained by subtracting 20 % of impurities from the raw data.

The MB spectrum of 2a exhibits a broad doublet with a large quadrupole splitting (δ = 0.97, ΔE_Q = 3.5 mm/s), although the presence of starting material (~30% of Fe) is also evident (Figure 6). Significantly, the modest decrease of 0.09 mm/s in isomer shift upon conversion of 1a to 2a provides unequivocal evidence that the oxidation is ligand-based, in support of the Fe²⁺–ISQ formulation for 2a. Likewise, the MB parameters of [2b]⁺ are consistent with the presence of a high-spin Fe(II) center. In this case, proper fitting of the MB spectrum required three doublets with δ-values of 0.97 ± 0.02 mm/s and ΔE_Q values between 1.7–2.5 mm/s (Table S3). Given the nearly identical isomer shifts, these three species likely correspond to conformational isomers of [2b]⁺, similar to the situation discussed above for [1b]⁺. The heterogeneity observed in the [2b]⁺ spectrum is related to the inability to prepare suitable crystals of this complex, which necessitated the use of powder samples lacking the intrinsic order of crystalline material. The broadness of the doublets for both 2a and [2b]⁺ is likely due to spin-spin relaxation effects.

MB spectra of [3a]⁺ and [3b]²⁺ each display a single doublet centered at δ = 0.73 and 0.64 mm/s, respectively, with quadrupole splittings of ~2.0 mm/s (Table 3; the [3b]²⁺ spectrum also contains features arising from ferrous impurities). The lower δ-values suggest that one-electron oxidation of 2a→[3a]⁺ (or [2b]⁺→[3b]²⁺) involves significant removal of electron density from the Fe ions. However, the observed isomers shifts are larger than one would expect for high-spin Fe(III) centers, which typically display values between 0.4 and 0.6 mm/s.^[35] A survey of high-spin [Fe³⁺(ISQ)_n] complexes (n = 1, 2, or 3) prepared by Wieghardt and coworkers found δ-values ranging from 0.44 to 0.54 mm/s.^[16-17] The MB data is therefore consistent with our DFT results that suggest partial IBQ character for the O,N-ligands in [3a]⁺/[3b]²⁺. It is also important to note that the a- and b-series of complexes yield very similar MB parameters, suggesting that the supporting ligand (^Ph2Tp(1-) vs. ^Ph2TIP) has little effect on the Fe/L_O,N unit.

As shown in Table 3, isomer shifts derived from DFT calculations are quite consistent with the experimental data, although DFT generally underestimates δ-values by ~0.05-0.15 mm/s. In particular, DFT nicely reproduces the magnitude of changes (Δδ) across the 1→2→3 series. The overall agreement is less satisfactory for quadrupole splittings, but it is well-known that DFT has more difficulty computing accurate ΔE_Q values.^[36] The general agreement between calculated and experimental MB parameters indicates that our computational models faithfully represent the electronic structures of these “redox-ambiguous” complexes.

2.C. Electronic Absorption and MCD Spectroscopies

The UV-vis-NIR spectra of 2a and [2b]⁺ displayed in Figure 2 are characterized by a broad absorption manifold (ε ~ 1.0 mM⁻¹cm⁻¹) centered around 750 nm (13,300 cm⁻¹). In this region, the 2a/[2b]⁺ spectra closely resemble those reported for other ISQ-containing complexes with various metal ions (Co³⁺, Ni²⁺, and Cu²⁺),^[37] as well as the spectrum measured by Carter et al. for a “free” ISQ radical.^[38] These spectral similarities suggest that the transitions observed for 2a and [2b]⁺ are primarily ligand-based. This conclusion is also supported our time-dependent DFT (TD-DFT) calculations. The dominant contributor to the 750 nm feature is an intraligand transition in which the acceptor orbital is the SOMO of the ^tBuISQ radical. TD-DFT calculations predict this transition to appear at 16,100 and 15,200 cm⁻¹ for 2a and [2b]⁺, respectively (Table S4). In general, the TD-DFT methodology nicely replicates the energies and intensities of the experimental absorption features, as shown in Figure 7 for 2a (electron density difference maps – EDDMs – for key transitions are provided in Figure S6). These calculations indicate that another component of the absorption manifold involves excitation to the ^tBuISQ SOMO from an MO with mixed ^tBuISQ/Fe(II) character; the computed energy for this transition is 15,000 cm⁻¹ for 2a. The weak near-infrared (NIR) bands evident for both complexes likely arise from Fe(II) d-d transitions, while features at higher energies (> 20,000 cm⁻¹) are attributed by TD-DFT to ^tBuISQ→Fe(II) charge transfer (CT). These latter bands are relatively weak due to poor overlap between the ^tBuISQ SOMO and half-occupied Fe(II) orbitals (Figure 5).

Figure 7.

Experimental (dashed line) and TD-DFT computed (solid line) absorption spectra for 2a (top) and [3a]⁺ (bottom). The arrows point to features in the computed spectra arising from the indicated type of electronic transition.

Analysis of the [3a]/[3b]²⁺ spectra is more complex due to the large extent of mixing between Fe and ligand orbitals. For both complexes, the β-HOMO – containing approximately equal parts metal and ligand character (vide supra) – arises from a π-bonding interaction between the parent Fe 3d and L_O,N orbitals (Figure 5). TD-DFT calculations, which again faithfully reproduce key features of the experimental spectrum (Figure 7 and Table S4), indicate that the intense absorption feature at ~13,000 cm⁻¹ involves electron transfer from the β-HOMO to its unoccupied π-antibonding counterpart; therefore, this excitation is best described as a π→π* transition centered on the O1-Fe1-N7 unit. The broad band evident in the NIR region (ε ~ 1.0 mM⁻¹cm⁻¹) is then attributed to β-HOMO→Fe 3d transitions where the acceptor orbitals lack significant L_O,N character. Given the mixed nature of the β-HOMO, this low-energy feature possesses both Fe d-d and LMCT character, as revealed in the EDDMs in Figure S7. Interestingly, the intraligand transition that is prominent at ~13,300 cm⁻¹ in the 2a/[2b]⁺ spectra is calculated to appear at 18,900 cm⁻¹ for [3a]⁺, although this transition now contains some LMCT character. The blue-shift reflects the partial IBQ character of the O,N-coordinated ligands, since previous reports have demonstrated that the lowest-energy IBQ-based transition in metal complexes occurs near 20,800 cm⁻¹.^[37] This intraligand transition is largely obscured in the [3a]⁺ spectrum by intense N_Tp→Fe CT transitions. However, because the ^Ph2TIP-based MOs are greatly stabilized relative to their ^Ph2Tp counterparts (on account of the difference in charge), the N_TIP→Fe CT transitions are found at relatively high energies. Intraligand transitions are therefore evident in the 20,000-25000 cm⁻¹ region of the [3b]²⁺ spectrum with intensities of ~1.5 mM⁻¹cm⁻¹.

The electronic transitions of [3b](OTf)₂ were further probed through the collection of solid-state MCD spectra at low temperatures (4-30 K). All bands in the visible region exhibit temperature-dependent intensity characteristic of C-term behavior (Figure 7, top), as expected for a paramagnetic S = 2 species. Based on our analysis of the absorption spectrum, the most intense MCD feature at 12,500 cm⁻¹ corresponds to the Fe/L_O,N π→π* transition, while the bands centered around 20,000 cm⁻¹ are ascribed primarily to intraligand transitions. Since MCD C-term intensity requires spin-orbit coupling between states, features that involve metal d-orbitals, such as ligand-field (d-d) and CT transitions, are relatively more intense in MCD than absorption spectra (C/D ratio).^[39] Thus, the comparative weakness of the higher-energy features in the [3b]²⁺ MCD spectrum is further confirmation of the ligand-based nature of these transitions.

Variable-temperature variable-field MCD (VTVH-MCD) data were collected at 790 nm (12660 cm⁻¹) for [3b]²⁺. In these experiments, MCD intensity was monitored at five temperatures (2, 4, 8, 15, and 30 K) as the magnetic field (H) was varied from 0 to 7 T; by convention, the resulting magnetization curves are plotted against βH/2kT (Figure 8, bottom). As demonstrated by Solomon and Neese, quantitative analysis of the VTVH-MCD curves provides valuable information regarding spin-Hamiltonian parameters and transition polarizations.^[40] The VTVH-MCD method is particularly powerful for non-Kramers systems, such as [3b]²⁺, that are often inaccessible by EPR. The magnetization curves obtained at 790 nm saturate rapidly with field and exhibit very little “nesting” (i.e., spread between curves obtained at different temperatures). Such behavior is characteristic of S = 2 systems with negative zero-field-splitting (ZFS).^[41] Indeed, the best fit of the 790 nm data was obtained with D = –5.0 cm⁻¹ and E/D = 0.20 (Figure 8), although a broad range of negative D-values (< 4 cm⁻¹) with moderate rhombicities (0.1 < E/D < 0.25) provided acceptable fits. Our analysis indicates that the corresponding transition is polarized in the xz-direction, which requires the O,N-donor ligand to lie in the xz-plane of the D-tensor (Figure 5).

Figure 8.

Top: Variable-temperature solid-state MCD spectra of [3b](OTf)₂. Spectra were measured at a magnetic field of 7 T and at temperatures of 4, 8, 15, and 30 K. Bottom: VTVH-MCD data collected at 790 nm for [3b](OTf)₂. Data were obtained at the indicated temperatures with magnetic fields (H) ranging from 0 to 7 T. The best fit (⨀) was obtained with the following spin Hamiltonian parameters: S = 2, D = −5 cm⁻¹, E/D = 0.20, g_iso = 2.0.

2.D. Resonance Raman Spectroscopies

Vibrational spectroscopy has proven valuable in the characterization of metal-bond phenoxyl and semiquinone radicals.^[42] In the case of dioxolene complexes, the C–O stretching frequency is a sensitive indicator of ligand oxidation state, ranging from 1400-1500 cm⁻¹ for semiquinones and 1620-1640 cm⁻¹ for benzoquinones.^[43] By contrast, the vibrational features of iminobenzo(semi)quinones have not been examined in detail. We therefore collected resonance Raman spectra on frozen samples of [3a]SbF₆ in CD₂Cl₂ (Figure 9). The experiments utilized 488 nm laser excitation, in resonance with the intraligand transitions that appear in this region (vide supra). To aid in peak assignments, we prepared a sample of [3a]⁺ in which the O,N-ligand was labeled with the ¹⁵N isotope; the difference spectrum revealed several peaks that are sensitive to ¹⁵N substitution (Figure 9). Interpretation of the rRaman data was further aided by DFT frequency calculations employing the BP functional.

Figure 9.

Resonance Raman spectra obtained with 488.0 nm excitation (45 mW) of frozen CD₂Cl₂ solutions of (top) natural abundance [3a]SbF₆, and (middle) [3a]SbF₆ with incorporation of ¹⁵N isotope in the L_O,N ligand. The difference spectrum is shown on the bottom. Frequencies (in cm⁻¹) are provided for selected peaks, with the corresponding ¹⁴N→¹⁵N shifts shown in parenthesis. Peaks marked with an asterisk (*) arise from the frozen solvent, while the peak marked with a triangle arises from condensed liquid O₂.

Based on literature precedents involving metal-semiquinone complexes,^{[42c, 42d]} the two isotopically-sensitive peaks at 520 and 577 cm⁻¹ in the [3a]⁺ spectrum are attributed to motions of the five-membered chelate ring formed by the Fe center and O,N-ligand. In support of this assignment, DFT predicts two modes at 532 and 588 cm⁻¹ (with ¹⁵N isotope shifts of 7 and 4 cm⁻¹, respectively) with large degrees of chelate stretching character. The mode at 520 cm⁻¹ (calculated at 532 cm⁻¹) is best described as a chelate breathing mode, whereas the chelate vibrations in the higher-energy mode are strongly mixed with intraligand C-C bond movements. Graphical depictions of these normal modes, along with further information concerning the computed vibrational spectrum, are provided in Figure S8.

The [3a]⁺ spectrum exhibits numerous peaks in the 1100-1650 cm⁻¹ region. These features are not present in the rR spectrum of 1a (Figure S9); we therefore conclude that they arise from the oxidized O,N-ligand (an exception is the 1609 cm⁻¹ peak, which appears in both spectra). Studies of analogous semiquinone complexes^{[42c, 42d]} found that the most intense peaks in this region correspond to modes that couple O-C stretching and C-C ring motions. Indeed, DFT predicts two ν(O-C)-based modes for [3a]⁺ at 1342 and 1404 cm⁻¹ that are slightly sensitive to ¹⁵N-substitution (calculated shifts of ~1 cm⁻¹). These modes likely corresponding to the prominent peaks at 1308 and 1383 cm⁻¹ in the experimental spectrum. Modes with significant ν(N-C) character have computed frequencies of 1319, 1384, and 1495 cm⁻¹, with ¹⁵N isotope shifts of ~1-3 cm⁻¹ in each case. The two lower-energy modes nicely match the experimental peaks at 1279 and 1373 cm⁻¹, respectively, while the higher-energy mode is not observed.

Vibrational frequency calculations were also performed for the Ga-ISQ and Ga-IBQ models, which possess O,N-ligands with unambiguous oxidation states (vide supra). DFT predicts two ν(O-C) modes at 1319 and 1424 cm⁻¹ for Ga-ISQ, while the corresponding modes for Ga-IBQ possess much higher frequencies of 1540 and 1584 cm⁻¹. Since the putative ν(O-C) modes of [3a]⁺ have experimental energies of 1308 and 1383 cm⁻¹, the rRaman data appear to support an ISQ assignment for the ligand oxidation state.

Discussion

A variety of spectroscopic and computational methods have been employed to evaluate the electronic structures of mononuclear Fe complexes bound to redox-active ligands derived from 2-amino-4,6-di-tert-butylphenol (^tBuAPH₂). Our studies included the ^Ph2Tp-supported complexes 2a and [3a]SbF₆ (previously reported), as well as a parallel “b-series” of complexes prepared with the neutral ^Ph2TIP supporting ligand (Scheme 2). The results described here strongly support our earlier conclusion – made on the basis of XRD and EPR data – that 2a contains a high-spin Fe(II) center antiferromagnetically coupled to an ISQ^tBu ligand radical (overall spin of 3/2). Specifically, the presence of a ferrous ion in 2a was verified by MB spectroscopy (δ = 0.97), and the unique absorption features were shown to arise from ^tBuISQ ligand-based transitions. Although we were not able to obtain a crystal structure of [2b]⁺, similarities between the spectral properties of 2a and [2b]⁺ indicate that the complexes share identical electronic configurations. Analysis of the more-oxidized [3a]⁺ and [3b]²⁺ species is less straightforward, since a preponderance of the data indicate that actual electronic structures lie between the Fe³⁺–ISQ and Fe²⁺–IBQ limits. For instance, the O,N-ligand in the [3b]²⁺ crystal structure exhibits a more pronounced quinoid distortion than expected for a typical ISQ ligand, and the isomer shifts measured for [3a]⁺ and [3b]²⁺ (δ_ave ~ 0.68) are considerably outside the range expected for high-spin ferric ions. However, the ν(C-O) frequencies observed by rRaman spectroscopy are more consistent with an ISQ oxidation state.

The unprecedented example of an Fe(II) center bound to an ISQ radical in complexes 2a and [2b]⁺ is made possible by the combination of high-spin states and trigonal-bipyramidal 5C geometries, which result in Fe centers with enhanced electron affinities. Concerning the role of spin state, our DFT calculations indicate that isomers of 2a with intermediate- or low-spin centers converge to the alternate Fe³⁺–amidophenolate(2-) configuration. And the importance of coordination number is highlighted by comparison to the six-coordinate (6C) [(L_N4)Fe³⁺(ISQ)]²⁺ complexes prepared by Wieghardt (where L_N4 = cis-cyclam or tren).^[16] These complexes undergo only ligand-based reductions to give the corresponding Fe(III) species with closed-shell amidophenolate(2-) ligands, even when the Fe center is high-spin.^[44] These insights have implications for the Fe/O₂/substrate intermediate in the catalytic cycle of APDOs. Like 2a/[2b]⁺, this species contains a high-spin center, yet it also has a 6C geometry due to the presence of the O₂ ligand. Thus, extrapolation from the synthetic models would seem to favor the closed-shell electronic structure of intermediate I in Scheme 1. However, unlike Wieghardt's [(L_N4)Fe³⁺(ISQ)]²⁺ models, the sixth ligand in the enzymatic intermediate (i.e., superoxide) is redox-active and electron-withdrawing, and therefore capable of facilitating electron-transfer from the substrate ligand to the Fe center. We are currently examining the reactivity of our Fe²⁺–APH and Fe²⁺–ISQ complexes with O₂ and NO to assess the impact of small molecule binding on the electronic configuration of the Fe/L_O,N unit.

Finally, it is worthwhile to consider why complexes like 2a/[2b]⁺ are viable whereas the corresponding Fe(II)–SQ complexes have not been reported despite extensive efforts in modeling catechol dioxygenases. Indeed, a catecholate analogue of 2a has already been reported and crystallographically characterized by Moro-oka, namely, [Fe²⁺(^tBu,iPrTp)(DBC)] (5; where DBC = dianion of 3,5-di-tert-butylcatechol).^[45] Like 2a, complex 5 contains a trigonal bipyramidal Fe/Tp unit bound to a bidentate “substrate” ligand; however, the catecholate lacks quinoid distortion (C-O bond distances of 1.35 and 1.38 Å), and the collective metric and spectroscopic data demand an Fe(III)-catecholate description. The divergent electronic structures of 2a and 5 point to the intrinsic difference in redox potentials between catecholate and amidophenolate dianions. This difference may have mechanistic implications for the ring-cleaving dioxygenases. Normally, it is assumed that ECDOs and APDOs share a common catalytic mechanism; however, our synthetic experience indicates that a species like intermediate II in Scheme 1 is more feasible for APDOs than ECDOs. Thus, the two catalytic cycles may differ at this point, with the ECDOs adopting an intermediate I structure (as proposed by Ye and Neese^[25]) and the APDOs adopting an intermediate II structure with considerable radical character on the substrate.

Experimental Section

Materials and physical methods

All reagents and solvents were purchased from commercial sources and used as received, unless otherwise noted. Acetonitrile, dichloromethane, and tetrahydrofuran were purified and dried using a Vacuum Atmospheres solvent purification system. The synthesis and handling of air-sensitive materials were performed under inert atmosphere using a Vacuum Atmospheres Omni-Lab glovebox. The ligands K(^Ph2Tp),^{[46] Ph2}TIP,^[28] and ^tBuAPH₂^[47] were prepared according to literature procedures. ¹⁵N-labeled ^tBuAPH₂ was prepared using ¹⁵NH₄OH purchased from Cambridge Isotopes. Elemental analyses were performed at Midwest Microlab, LLC in Indianapolis, IN.

Samples of [2b]OTf for spectroscopic studies were prepared by treating [1b]OTf with one equivalent of TTBP^•[48] in CH₂Cl₂. After stirring for two hours, the green solution was filtered and the solvent removed under vacuum. The resulting powder exhibited the following peaks in the IR spectrum: 3342 [ν(N-H)], 3054, 1439, 1259, 1222 cm^-1.

Spectroscopic Methods

UV-vis absorption spectra were obtained with an Agilent 8453 diode array spectrometer; NIR absorption spectra were measured using an Agilent Cary 5000 spectrophotometer. Fourier-transform infrared (FTIR) spectra of solid samples were measured with a Thermo Scientific Nicolet iS5 FTIR spectrometer equipped with the iD3 attenuated total reflectance accessory. EPR experiments were performed using a Bruker ELEXSYS E600 equipped with an ER4415DM cavity resonating at 9.63 GHz, an Oxford Instruments ITC503 temperature controller, and an ESR-900 He flow cryostat. MCD spectra were obtained using a Jasco J-715 spectropolarimeter in conjunction with an Oxford Instruments SM-4000 8T magnetocryostat. Solid-state samples of [3a]SbF₆ were prepared as uniform mulls in fluorolube. All MCD spectra were obtained by subtracting the -7 T spectrum from the +7 T spectrum to eliminate potential artifacts. Resonance Raman spectra were measured using 488.0 nm excitation from a Spectra Physics Ar⁺ laser (model 2025-05), with 45 mW at the sample point in a 180° backscattering geometry. The sample was placed in a transparent quartz Dewar cell filled with liquid nitrogen and spun at 800 rpm. Spectra were collected with the Spec-10 system (Princeton Instruments) installed on the 1269 spectrograph (SPEX Industries) equipped with a standard 1200 grove/inch grating at 80 mm slit width. Spectra were calibrated with fenchone and indene standards. Low-field (0.04 T) variable temperature (5 K – 200 K) Mössbauer spectra were recorded on a closed-cycle refrigerator spectrometer, model CCR4K, equipped with a 0.04 T permanent magnet, maintaining temperatures between 5 and 300 K. Mössbauer spectra were analyzed using the software WMOSS (Thomas Kent, SeeCo.us, Edina, Minnesota). The samples were polycrystalline powders, suspended in nujol, placed in Delrin 1.00 mL cups and frozen in liquid nitrogen. The isomer shifts are quoted at 6 K with respect to iron metal spectra recorded at 298 K.

Electrochemical measurements

Cyclic voltammetric (CV) measurements were conducted in the glovebox with an epsilon EC potentiostat (iBAS) at a scan rate of 100 mV/s with 100 mM (NBu₄)PF₆. A three-electrode cell containing a Ag/AgCl reference electrode, a platinum auxiliary electrode, and a glassy carbon working electrode was employed. Under these conditions, the ferrocene/ferrocenium (Fc^+/0) couple has an E_1/2 value of +0.52 V in CH₂Cl₂.

Synthesis of Fe complexes

Synthetic procedures for complexes 1a, 2a, and [3a]SbF₆ were previously published in Bittner et al.^[26]

[(^Ph2TIP)Fe²⁺(^tBuAPH)]X ([1b]X; X = OTf, BPh₄): [(^Ph2TIP)Fe²⁺(CH₃CN)₃](OTf)₂^[22] (1.12 g, 0.927 mmol) and ^tBuAPH₂ (205.1 mg, 0.927 mmol) were mixed in CH₂Cl₂ (10 mL), followed by addition of triethylamine (142 μL, 1.02 mmol). After stirring overnight, the reaction was filtered and the solvent removed under vacuum. The brown solid was taken up in toluene and the product precipitated with excess pentane. The yellow solid was collected, washed again with pentane, and then dried to yield [1b]OTf as a bright yellow powder (yield = 921 mg, 86%). IR (solid): ṽ = 3348, 3059, 2949, 2901, 2861, 1064, 1462, 1442, 1273, 1255, 1152, 1030 cm⁻¹. Counteranion metathesis was performed by mixing [1b]OTf (125 mg, 0.108 mmol) with NaBPh₄ (37.6 mg, 0.110 mmol) in MeOH (3 mL). After stirring for 15 minutes, a yellow precipitate formed that was collected and dried under vacuum. X-ray quality crystals of [1b]BPh₄ were obtained by layering a concentrated 1,2-dichloroethane (DCE) solution with MeOH (yield = 83.1 mg, 48%). IR (solid): 3340, 3050, 2948, 2863, 1578, 1478, 1443, 1379, 1305, 1264, 1142, 1074, 1023 cm⁻¹. UV/Vis (CH₂Cl₂): λ_max (ε) = 815 (10), 399 (1160 M⁻¹cm⁻¹). The X-ray structure revealed three uncoordinated DCE molecules in the asymmetric unit, and elemental analysis suggests that a small amount of solvent (~0.3 equiv./Fe) remains after drying. Elemental analysis calcd (%) for C₈₆H₈₁BFeN₇OP•0.3DCE: C 76.70, H 6.11, N 7.23; found: C 76.70, H 6.13, N 7.13.

[(^Ph2TIP)Fe³⁺(^tBuISQ)](OTf)₂ ([3b]OTf₂): Complex [1b]OTf (184 mg, 0.159 mmol) and two equiv. of AgOTf (83 mg, 0.32 mmol) were stirred in THF (6 mL) for one hour, yielding a dark green solution. The solution was filtered through celite and the solvent removed under vacuum. The resulting solid was taken up in CH₂Cl₂ (or 1,2-dichloroethane) and layered with n-hexane, providing dark green needles suitable for XRD experiments (yield = 208 mg, 94%). IR (solid): 3270, 3058, 2959, 1603, 1578, 1469, 1445, 1364, 1257, 1221, 1147, 1073, 1026 cm⁻¹. UV/Vis (CH₂Cl₂): λ_max (ε) = 798 nm (2800 M⁻¹cm⁻¹). Elemental analysis calcd (%) for C₆₄H₆₀F₆FeN₇O₇PS₂: C 58.94, H 4.64, N 7.52; found: C 54.88, H 4.84, N 6.79. The disagreement indicates that small amounts of impurities and/or solvent are present.

[(^Ph2Tp)Fe²⁺(DMAPH)] (4a): 2-(dimethylamino)-4,6-di-tert-butylphenol (DMAPH) was prepared according to the previously published procedure.^[49] Fe(OTf)₂ (248 mg, 0.701 mmol), K(^Ph2Tp) (497 mg, 0.701 mmol), and DMAPH (175 mg, 0.701 mmol) were mixed in a 2:1 solution of CH₂Cl₂:MeCN (15 mL), followed by addition of triethylamine (108 μL, 0.774 mmol). After stirring for several hours, the solvent was removed. The resulting solid was taken up in CH₂Cl₂ (6 mL) and filtered. Storage of this solution in a freezer (−30 °C) for several days provided bright yellow crystals suitable for X-ray crystallographic analysis (yield = 160 mg, 24%). IR (solid): 3056, 2948, 2855, 2637, 1545, 1462, 1412, 1356, 1302, 1264, 1246, 1167, 1060, 1028, 1007 cm⁻¹. UV/Vis (CH₂Cl₂): λ_max (ε) = 365 nm (1950 M⁻¹cm⁻¹). Elemental analysis calcd (%) for C₆₁H₆₀BFeN₇O: C, 75.23; H, 6.21; N, 10.07; found: C, 74.97; H, 6.18; N, 10.07.

Crystallographic studies

X-ray diffraction (XRD) data were collected with an Oxford Diffraction SuperNova kappa-diffractometer (Agilent Technologies) equipped with dual microfocus Cu/Mo X-ray sources, X-ray mirror optics, Atlas CCD detector, and low-temperature Cryojet device. The data were processed with CrysAlis Pro program package (Agilent Technologies, 2011) typically using a numerical Gaussian absorption correction (based on the real shape of the crystal), followed by an empirical multi-scan correction using SCALE3 ABSPACK routine. The structures were solved using SHELXS program and refined with SHELXL program^[50] within Olex2 crystallographic package.^[51] B- and C-bonded hydrogen atoms were positioned geometrically and refined using appropriate geometric restrictions on the corresponding bond lengths and bond angles within a riding/rotating model (torsion angles of methyl hydrogens were optimized to better fit the residual electron density).

Crystal data for [1b](BPh₄)·1.5 DCE: Formula= C₈₉H₈₇BCl₃FeN₇OP; M_w= 1474.68; triclinic; space group P−1; a= 17.3220(4), b= 20.6267(4), c= 22.0768(4) Å; α= 81.955(2), β= 78.343(2), γ= 87.958(2)°; V= 7649.0(3) ų; T = 100 K; Z= 4; 101066 reflections measured; 37578 unique reflections (R_int = 0.0357) ; final R indices [I>2σ(I)] R1= 0.0552 and wR2= 0.1469; R indices (all data) R1=0.0708 and wR2=0.1608.

Crystal data for [3b](OTf)₂·CH₂Cl₂: Formula= C₆₅H₆₂Cl₂F₆FeN₇O₇PS₂; M_w = 1389.08; monoclinic; space group P2₁/c; a= 20.0160(13), b= 15.3008(11), c= 22.2057(18) Å; α= 90, β= 94.320(8), γ= 90°; V= 6781.4(9) ų; T= 100 K; Z= 4; 35453 reflections measured; 13347 unique reflections (R_int = 0.0787); final R indices [I>2σ(I)] R1=0.0723 and wR2=0.1786; R indices (all data) R1=0.1153 and wR2=0.2105.

Crystal data for [4a]·4CH₂Cl₂: Formula= C₆₅H₆₈BCl₈FeN₇; M_w= 1313.56; monoclinic; space group P2₁/c; a= 17.1045(5), b= 14.4938(5), c= 26.5165(8) Å; α= 90, β= 100.482(3), γ= 90°; V=6464.0(3) ų; T = 100 K; Z= 4; 41153 reflections measured; 12682 unique reflections (R_int= 0.0356); final R indices [I>2σ(I)] R1=0.0606 and wR2=0.1488; R indices (all data) R1=0.0764 and wR2=0.1588.

DFT calculations

DFT calculations were performed using the ORCA 2.8 software package developed by Dr. F. Neese.^[52] Geometry optimizations employed either the Becke-Perdew (BP86) functional^[53] or Becke's three-parameter hybrid functional for exchange along with the Lee-Yang-Parr correlation functional (B3LYP).^[54] Ahlrichs’ valence triple-ζ basis set (TZV), in conjunction with the TZV/J auxiliary basis set,^[55] were used for all calculations. Time-dependent DFT (TD-DFT) calculations^[56] computed absorption energies and intensities within the Tamm-Dancoff approximation.^[57] In each case, at least 60 excited states were calculated. Vibrational frequency calculations were performed with a truncated [3a]⁺ model with hydrogen atoms at the 3- and 5-positions of the Tp ligand. Calculation of the harmonic force fields proved that the optimized structure is a local minima on the potential energy surface. MB parameters were calculated using the approach described in Römelt et al.^[58]; in these calculations, the size of the integration grid for Fe, O, and N atoms was increased. The gOpenMol program^[59] developed by Laaksonen was used to generate isosurface plots of molecular orbitals.