Axial Ligand Effects on Electronic Structures and H‑Atom Abstraction by Oxidoiron(IV) Complexes Featuring Triplet-Only Reactivity

Abstract

Oxidoiron­(IV) species are crucial intermediates for the functionalization of C–H bonds by biological oxidases and oxygenases, and understanding the relations between ligation, electronic structure, and reactivity of the Fe^IVO unit is key to the rational design of bioinspired catalysts. Here, a series of Fe^IVO complexes based on a macrocyclic tetracarbene ligand and with different anionic trans-axial ligands (L^ax = CF₃COO^–, ^t BuS^–, AdaS^–) has been comprehensively characterized, including X-ray crystallography in two cases, as well as zero-field and applied-field ⁵⁷Fe Mössbauer spectroscopy, magnetic susceptibility measurements, magnetic circular dichroism (MCD), and helium-tagging infrared photodissociation (IRPD). This establishes a detailed picture of the electronic structures of these ferryl complexes, featuring a well-separated S = 1 ground state, and the influence of the trans-ligand. The confinement to triplet-only pathways in C–H bond activations and comparison with the parent Fe^IVO complex (L^ax = MeCN) allow uncovering trends in hydrogen atom abstraction (HAA) ability without perturbations from different extents of two-state reactivity (TSR). Kinetic studies on HAA for C–H substrates with BDEs up to 78 kcal mol^–1 show relatively minor effects of the axial ligands, with rates following an electrophilic trend in the order L^ax = MeCN > CF₃COO^– ≈ RS^–. This contrasts the antielectrophilic trend found for a related series of tetramethylcyclam (TMC)-based Fe^IVO complexes for which TSR is dominant. For the present systems, temperature-dependent KIEs, in combination with computations at the DFT and ab initio levels of theory, suggest a semiclassical model with modest tunneling contributions that are most pronounced in the case of a strongly donating trans-thiolate. This study thus provides a combined experimental and computational basis for understanding intrinsic axial ligand effects and the effect of a trans-axial thiolate in S = 1 ferryl intermediates.

Introduction

High-valent oxidoiron species are key intermediates in many transformations in nature, most prominently in the activation of O₂ by heme and nonheme oxidases and oxygenases. These enzymes, such as Cytochrome P450, Taurine dioxygenase, halogenase CycC3, or phenylalanine hydroxylase, are known to catalyze challenging substrate transformations like the oxygenation of unactivated C–H bonds. In heme enzymes, the axial ligand coordinated to the central iron plays an important role for tuning the reactivity of the oxidoiron­(IV) (ferryl) intermediates in compounds I and II. Especially, a cysteinato ligand is proposed to enable the hydroxylation of strong C–H bonds by Cytochrome P450, because the axial thiolate ligand is assumed to increase the basicity of the Fe^IVO unit by a push effect allowing the hydrogen atom abstraction (HAA) to proceed despite relatively low redox potentials. , Therefore, understanding the influence of different axial ligands on ferryl reactivity has been an essential aim, both for the biological systems and for synthetic analogues that emulate key features of the natural blueprints and may be employed for synthetically useful C–H bond functionalization.

Most oxidoiron­(IV) nonheme model systems that have been investigated over the years feature chelating N-donor ligand scaffolds, most prominently macrocyclic ligands based on 1,4,8,11-tetraazacyclotetradecane (cyclam), pentadentate ligands with a central tertiary amine and four pyridine arms (N4Py) or tripodal tetradentate ligands like tris­(2-pyridylmethyl)­amine (TPA). These high-valent oxidoiron­(IV) complexes can engage in oxygen atom transfer (OAT), hydrogen atom abstraction (HAA), and electron transfer (ET) reactivity. In contrast to natural nonheme enzymes that have S = 2 spin states, most synthetic analogues exhibit an S = 1 ground state. However, the quintet state for nonheme enzymes leads to higher reactivity in C–H activation via the σ-pathway, where the σ*­(Fe-d _{z ²} ) is the electron-accepting orbital (EAO) interacting with the target C–H σ-bond and exchange energy is maximized. , The steric strain and the activation barrier are hereby reduced in comparison to the π-pathway, where the π*_x/y(Fe-d _xz,yz) serves as EAO. S = 1 ferryl species are usually expected to react over the higher-barrier π-trajectory, but such model systems often exhibit an unexpectedly high reactivity. This has been explained by the two-state reactivity (TSR) scenario where, during the interaction with the C–H bond, the S = 1 complex undergoes a spin crossover to the more reactive quintet (S = 2) state, enhancing the overall reaction rate (Figure ; left). ,

1.

Schematic free energy profiles for HAA processes with R–H substrates for TMC-based S = 1 ferryl complexes featuring two-state reactivity (TSR) due to a small triplet-quintet gap (left) and for S = 1 ferryl complexes based on macrocyclic tetracarbenes with a very large triplet-quintet gap (right).

In principle, HAA by high-valent oxidoiron­(IV) intermediates is an electrophilic reaction, and hence strongly electron-donating supporting ligands that decrease the electrophilicity of the Fe^IVO unit would be expected to increase the transition state barrier. Ligands trans to the oxo group are more influential than the donors of the equatorial ligand scaffold, which are cis to the oxo group, because the trans-ligand directly interacts with both σ-type (Fe-d _{z ²} ) and π-type (Fe-d _xz,yz) metal fragment orbitals. However, when Que, Shaik and Nam et al. experimentally studied the effect of different trans-ligands on the HAA reactivity for a series of S = 1 ferryl complexes bearing an equatorial tetramethylcyclam (TMC = 1,4,8,11-tetramethyl-1,4,8,11-tetraaza-cyclotetradecane), a counterintuitive antielectrophilic trend was observed. The reactivity of complexes [(TMC)­Fe^IVO­(L^ax)]ⁿ⁺ toward C–H substrates such 9,10-dihydroanthracene (DHA) increases in the order L^ax = NCCH₃ (A ^NCMe ) < CF₃COO^– (A ^TFA ) < N₃ ^– (A ^N3 ) < A ^TMCS (Figure , where TMCS represents a tethered trans-thiolate), which has been explained on the basis of the TSR model. Increasing electron donation by the axial ligand to the Fe^IVO group destabilizes the nb­(Fe-d _xy ) orbital relative to the σ*­(Fe-d _{x ²–y ²} ) orbital, which reduces the orbital energy gap and thus decreases the quintet-triplet separation ΔE _Q‑T. This leads to enhanced participation of the more reactive S = 2 state in the HAA reaction, which overshadows the intrinsic axial ligand effect. , Hence, A ^TMCS is the least electrophilic species but has the lowest ΔE _Q‑T. In line with this reasoning, the reactivity of analogous ruthenium complexes [(TMC)­Ru^IVO­(L^ax)]ⁿ⁺ that all feature well-separated triplet (S = 1) ground states, was shown to follow the reverse trend, i.e., the expected electrophilicity pattern, although overall reactivity is rather sluggish. However, the antielectrophilic trend in HAA by ferryl complexes [(TMC)­Fe^IVO­(L^ax)]ⁿ⁺ may be significantly affected by the dependence of tunneling contributions on the electron-donating ability of the trans-axial ligand L^ax. Therefore, a series of S = 1 ferryl complexes without involvement of two-state reactivity would be of particular interest to experimentally uncover the intrinsic effects of the axial ligands on Fe^IVO reactivity.

2.

Previously reported oxidoiron­(IV) complexes featuring a thiolato ligand; only complex C has been characterized by X-ray diffraction.

In 2013, some of us reported the isolation of an organometallic oxidoiron­(IV) complex that features in the equatorial plane a macrocyclic ligand with four N-heterocyclic carbene donors, [L^NHCFe^IVO­(MeCN)]­(OTf)₂ (3; L^NHC = 3,9,14,20-tetraaza-1,6,12,17-tetraazoniapenta-cyclohex-acosane-1 (23),4,6(26),10,12(25),15,17(24),21-octaene; Figure ). Complex 3 is topologically reminiscent of TMC-based A ^NCMe , and it was shown that the bonding within the Fe^IVO unit is not significantly affected by the different equatorial scaffolds. Furthermore, a series of {Fe-NO} ⁿ complexes of L^NHC demonstrated that redox changes are largely confined to the metal ion. However, the very strong σ-donation of the macrocyclic tetracarbene in oxidoiron­(IV) complex 3 raises the energy of its d _{x ²–y ²} orbital above the d _{z ²} , which gives rise to a very large triplet-quintet energy gap, preventing accessibility of the quintet state (Figure ; note that the lowest-energy quintet excited states of A ^NCMe and 3 have different electron configurations, nb­(Fe-d _xy )¹π*­(Fe-d _xz,yz)²σ*­(Fe-d _{x ²–y ²} )¹σ*­(Fe-d _{z ²} )⁰ versus a unique nb­(Fe-d _xy )¹π*­(Fe-d _xz,yz)²σ*­(Fe-d _{z ²} )¹σ*­(Fe-d _{x ²–y ²} )⁰ in the case of 3). As a consequence, the HAA reactivity of 3 is confined to the triplet surface, and any two-state reactivity is quenched (Figure , right). While representing the intrinsic reactivity of a ferryl species in its triplet state, 3 was shown to exhibit HAA reactivity toward C–H substrates DHA or 1,4-cyclohexadiene (CHD) that are 2–3 orders of magnitude higher than for A ^NCMe under the same experimental conditions. L^NHC-based oxidoiron­(IV) complexes therefore represent an ideal platform for the systematic investigation of intrinsic axial ligand effects on S = 1 ferryl reactivity without the ambiguities arising from different extents of two-state reactivity. A recent computational study indeed predicted that different axial ligands barely alter the triplet state orbital ordering of [(L^NHC)­Fe^IVO­(L^ax)]ⁿ⁺ complexes and do not significantly influence the reactivity.

3.

Structures and electronic ground states of 3, A^NCMe , and C.

In the present work, we now synthesize several new oxidoiron­(IV) complexes based on the tetracarbene scaffold L^NHC, namely [L^NHCFe^IVO­(L^ax)]­(OTf) with varying anionic trans-ligands L^ax = CF₃COO^–, ^t BuS^– and AdaS^– (Ada is 1-adamantyl); these L^ax were chosen to enable a direct comparison with the related TMC-based systems. Since oxidoiron­(IV) intermediates with S-bound cysteine are the key oxidants in the catalytic cycles of many heme and nonheme iron enzymes, a focus of this study lies on species [L^NHCFe^IVO­(SR)]⁺ with a trans-thiolate. During the last two decades, extensive efforts have been devoted to the generation and isolation of synthetic nonheme oxidoiron­(IV) complexes with thiolato ligation, aiming at understanding how thiolates provide favorable conditions for O₂ binding and reactivity enhancement. However, the number of well-characterized nonheme oxidoiron­(IV) complexes with a thiolato ligand has remained very limited, mainly because the stabilization of a thiolato group in a highly oxidizing environment in the absence of a protecting protein pocket is challenging; previously reported examples are shown in Figure .

A crystallographically characterized cis-thiolato-ligated high-valent oxidoiron­(IV) complex C with an extremely short Fe–O bond (1.603(7) Å) was reported by Kovacs et al. and was shown to be capable of cleaving strong C–H bonds (96 kcal mol^–1). Experimental spectroscopic data for C are lacking, but its properties were rationalized by DFT calculations, which indicated that C is an S = 1 Fe^IVO species with the σ*­(Fe–O) and one of the π*­(Fe–O) orbitals being empty, while the lower-lying π*­(Fe–S) orbital is doubly occupied and the σ*­(Fe–S) orbital is singly occupied (Figure ), leading to a long Fe–S bond and a very short Fe–O bond as well as an unusually high DFT-calculated ν_Fe–O stretching frequency of 918 cm^–1. This situation clearly differs from the situation in the only nonheme Fe^IVO model complex featuring a trans-thiolato ligand, [(TMCS)­FeO]⁺ (A ^TMCS ) reported by Que and coworkers. The structure of complex A ^TMCS , which is based on a cyclam derivative with a pendant thiolato arm (TMCS), has been determined by extended X-ray absorption fine structure (EXAFS), showing d(Fe–O) = 1.70(2) Å and d(Fe–S) = 2.33(2) Å. Unfortunately, no experimental vibrational data for ν_Fe–O of A ^TMCS are available. Complex A ^TMCS is the current benchmark system that has been used to investigate the effect of axial thiolato ligation on Fe^IVO reactivity. Oxidoiron­(IV) complex B with a silanedithiolato ligand cis to the FeO unit has also been characterized by EXAFS and shows d(Fe–O) = 1.687(6) Å. The Fe–O stretch was identified by resonance Raman (rR) spectroscopy at ν_Fe–O = 735 cm^–1. However, the authors indicated that some Fe^III(O)­(thiyl radical) character for B cannot be ruled out. B was found to be unreactive toward weak C–H bonds in substrates such as xanthene or acridine. In view of the very limited number of reported thiolato-ligated oxidoiron­(IV) species, the isolation of further Fe^IVO complexes with a trans-thiolate and their structural characterization by single-crystal X-ray diffraction (scXRD) as well as studies on their HAA reactivity would be highly desirable.

Here, the new complexes [L^NHCFe^IVO­(L^ax)]­(OTf) are comprehensively characterized by different spectroscopic methods including applied-field ⁵⁷Fe Mössbauer as well as MCD spectroscopy, and some of them are also structurally characterized by single-crystal X-ray diffraction (scXRD). Furthermore, the HAA reactivity toward a set of C–H substrates has been studied, and temperature-dependent kinetic isotope effects (KIEs) are examined, and the results are evaluated and rationalized computationally. Together with 3, this series of new [L^NHCFe^IVO­(L^ax)]­(OTf) complexes thus allows us to experimentally assess the intrinsic axial ligand effect for pure “triplet-only” ferryl model systems.

Results and Discussion

Synthesis and Structural Characterization of Fe^IVO Complexes

[L^NHCFe^IVO­(NCMe)]­(OTf)₂ (3) was synthesized by the addition of 2.8 eq of the soluble iodosobenzene derivative 2-( ^t BuSO₂)­C₆H₄IO (sPhIO) to the Fe^II complex [L^NHCFe­(NCMe)₂]­(OTf)₂ (1) at −35 °C in MeCN, as reported previously. , The new complex [L^NHCFe^IVO­(OOCCF₃)]­OTf (4) with an exogeneous trifluoroacetate (TFA) was now obtained by adding 1 eq of NaOOCCF₃ to the reaction mixture of freshly prepared 3 (Scheme ). For installing a trans-axial thiolate, the thiolato ligands have to be bound prior to the formation of the oxidizing ferryl species to avoid the oxidation of free thiolate. Sterically demanding ^t BuS^– and AdaS^– were used in this work in order to disfavor intermolecular degradation pathways involving disulfide bond formation. Hence, Fe^II complexes [L^NHCFe­(SR)­(NCMe)]­OTf (2a: R = ^t Bu; 2b: R = Ada) were prepared according to recently reported procedures by adding the corresponding NaSR to a solution of 1 in MeCN, and the resulting thiolato complexes 2a/2b were subsequently oxygenated by the addition of 2.8 eq of sPhIO at −35 °C to give complexes [L^NHCFe^IVO­(SR)]­OTf (5a, 5b; Scheme ). Alternatively, NBu₄[IO₄] can also be used as an oxidant.

1. Complexes Studied in This Work and Their Syntheses.

Oxidation of the Fe^II complexes and generation of the oxoferryl species can be conveniently monitored by UV–vis spectroscopy (Figure and Figure S1). The parent complex [L^NHCFe^IVO­(NCMe)]­(OTf)₂ (3) shows characteristic absorption bands at 400 nm and 600 nm. ,, In the case of [L^NHCFe^IVO­(OOCCF₃)]­OTf (4), these bands are slightly red-shifted to 408 and 629 nm, respectively (the assignment of these absorptions is discussed below). Both thiolato complexes [L^NHCFe^IVO­(S ^t Bu)]­OTf (5a) and [L^NHCFe^IVO­(SAda)]­OTf (5b) show almost identical UV–vis spectra with absorption bands at 442 and 577 nm (5a) or 447 and 586 nm (5b). In contrast to complexes 3 and 4, however, 5a/5b feature an additional weaker absorption around 710 nm, which will be discussed in more detail below. In case of thiolato-ligated Fe^II complexes 2a/2b and using NBu₄[IO₄] as an oxidant for UV–vis monitoring (see SI for details), a transient intermediate with an absorption around 800 nm can be observed en route to 5a/5b (Figure , bottom). This intermediate could not be isolated or characterized so far, but one may speculate that it is an initial IO₄ ^– adduct of 2a/2b.

4.

In situ formation of complexes 3 (1.0 mM; top), 4 (1.0 mM; middle), and 5a (0.25 mM; bottom) at −40 °C upon the addition of excess oxidant (sPhIO or NBu₄[IO₄]).

Single crystals of 4 could be obtained by the slow diffusion of diethyl ether into a MeCN solution of the complex at −35 °C and were analyzed by X-ray diffraction. The molecular structure of the cation is shown in Figure (left) and confirms the presence of the O-bound TFA trans to the oxido ligand. The FeO bond length is 1.671(1) Å, which is only slightly longer than in the parent complex 3* (1.661(3) Å; 3* is the derivative of 3 with trans-axial EtCN, which was characterized by scXRD). The tetra­(NHC) macrocycle L^NHC adopts a saddle-shaped nonplanar structure, and each −CH₂CH₂– linker forms one short C–H···O contact with the ferryl oxygen atom (dashed lines in Figure ), indicative of a weak interaction that was also observed for the parent complex 3* and was computationally suggested to contribute to the relative stability of 3/3*, , similar to the stabilization of a tripodal oxidoiron­(IV) complex with three N–H···O hydrogen-bonding interactions.

5.

Plot (30% probability thermal ellipsoids) of the molecular structure of the cationic parts of 4 (left) and 5b (right); hydrogen atoms omitted for clarity.

Dark green single crystals of 5b could be grown by layering diethyl ether onto a solution of the complex in PrCN at −35 °C. X-ray diffraction analysis confirms the molecular structure with equatorial ligation of the central iron atom by the L^NHC macrocycle, a terminal oxido ligand and a trans-axial thiolate (Figure , right). However, the quality of the crystallographic data is rather poor, so we refrain from a detailed discussion of metric parameters.

The new complex 5b is only the second isolated nonheme oxidoiron­(IV) species with a trans-thiolate, and the first to be structurally authenticated by scXRD, with A ^TMCS (Figure ) reported by Que, Nam, and coworkers being the benchmark system so far. While structural parameters of A ^TMCS such as d(Fe–O) = 1.70(2) Å and d(Fe–S) = 2.33(2) Å were determined by EXAFS, vibrational data for ν_Fe–O of A ^TMCS are unknown. On the other hand, the electronic structure of C (Figure ) featuring a cis-thiolato-ligated S = 1 Fe^IVO unit is clearly very different from that of Fe^IVO complexes with a trans-thiolate (vide supra). To identify the effect of the trans axial thiolate on the tetra­(NHC)-ligated Fe^IVO core and to provide a foundation for evaluating reactivity trends, the spectroscopic properties and electronic structures of the new complexes 4 and 5a/5b have been studied in detail in this work and are compared with those of parent 3.

Spectroscopic Characterization and Electronic Structures of Fe^IVO Complexes

⁵⁷Fe Mössbauer spectroscopy of the newly synthesized oxidoiron complexes 4, 5a, and 5b confirmed the iron­(IV) character. 4 features a very low isomer shift of δ = −0.14 mm s^–1 and a large quadrupole splitting of ΔE _Q = 3.13 mm s^–1 (Figure , left). These values are very similar to those of the parent oxidoiron­(IV) compound [L^NHCFeO­(NCMe)]­(OTf)₂ (3: δ = −0.13 mm s^–1, ΔE _Q = 3.08 mm s^–1), which have previously been attributed to the extremely strong equatorial σ-donor strength of the macrocyclic tetra­(NHC) ligand with pronounced charge donation into the iron Fe­(4s) orbital as well as to the oblate charge distribution around the Fe nucleus arising from the strongly asymmetric covalent bonds and the large electronic charge donated by the equatorial NHCs into the d _{x ²‑y ²} orbital. In contrast, oxidoiron­(IV) complexes based on the TMC ligand scaffold show isomer shifts within the narrow range δ = 0.15–0.20 mm s^–1 and much smaller quadrupole splittings in the range ΔE _Q = 0.16–1.39 mm s^–1. Complexes 5a and 5b show less negative isomer shifts δ = −0.09 mm s^–1 and significantly smaller ΔE _Q = 2.21 mm s^–1 or ΔE _Q = 2.23 mm s^–1, respectively (Figure , right, and Figure S3). This can be attributed to the electronic nature of the axial thiolato ligand, which acts both as σ- and strong π-donor. Due to the latter interaction, the electron density in the d-orbitals of Fe increases, causing a higher shielding of the s-electron density and a lowering of the anisotropy of charge distribution around the iron core; this increases the isomer shift and lowers the quadrupole splitting. A similar effect on ΔE _Q was found for TMC-based oxidoiron­(IV) complexes, where A ^NCMe and A ^TFA (the latter having a trans-axial CF₃COO^–) show similar quadrupole splittings of 1.23 mm s^–1 and 1.39 mm s^–1, respectively, while A ^TMCS has a much smaller ΔE _Q = 0.22 mm s^–1 that reflects the larger basicity of the axial thiolato ligand. ,

6.

Zero-field ⁵⁷Fe Mössbauer spectra of solid samples of [L^NHCFe^IVO­(OOCCF₃)]­OTf (4) and [L^NHCFe^IVO­(S ^t Bu)]­OTf (5a) at 80 K.

Magnetic susceptibility and variable temperature–variable field (VTVH) data for solid samples of 4, 5a, and 5b, obtained by SQUID measurements, are shown in Figure and Figure S4. They confirm an S = 1 ground state for all three compounds and rather small zero-field splitting parameters of D = 19.0 cm^–1 (for 4), 14.5 cm^–1 (for 5a) and 15.7 cm^–1 (for 5b). These D values are similar to that of parent 3 (D = 16.4 cm^–1) but much smaller than for TMC-based systems, which exhibit values around D = 30 cm^–1 (A ^NCMe : D = 29 cm^–1, A ^TFA : D = 31 cm^–1 and A ^TMCS : D = 35 cm^–1). ,, This reflects a large triplet-quintet gap in the tetra­(NHC)-ligated oxidoiron­(IV) species, where the mixing of the S = 1 and S = 2 states is minimal. Theoretical results are consistent with a large triplet-quintet gap for the new complexes 4, 5a, and 5b, ∼15–20 kcal mol⁻¹ at the DFT-B3LYP level and ∼25–30 kcal mol⁻¹ at the CASSCF­(12,12)/NEVPT2 level (Table S11).

7.

Magnetic susceptibility and variable temperature–variable field (VTVH) measurements of solid samples of [L^NHCFe^IVO­(OOCCF₃)]­OTf (4) (left) and [L^NHCFe^IV(O)­(S ^t Bu)]­OTf (5a) (right).

The above zero-field splitting parameter values D were confirmed by applied magnetic field Mössbauer measurements on 4 and 5b, as shown in Figure . Simulation of the spectra (see SI for details) gave D = 17.6 cm^–1 for 4 and D = 18.4 cm^–1 for 5b, in good agreement with the values derived from the SQUID data. The magnetic hyperfine coupling constants A/g _Nμ_N [xx, yy, zz] for 4 [−20.6, −24.6, −2.8 T] and for 5b [−24.9, −22.2, 3.2 T] are quite similar. The weak rhombicity is evident from the small difference between A _xx and A _yy , whereas the tensor remains predominantly axial. The values obtained for 4 and 5b are in good agreement with those reported for 3 [−17.1, −18.8, −2.0 T] or A ^TMCS [−23, −25, −5 T], except for the sign of A _zz . , The sign change may reflect the pronounced sensitivity of the z component of the A tensor to the axial ligand field, where even subtle changes in axial bonding can modify the balance between the isotropic Fermi-contact term and anisotropic spin-dipolar and spin–orbit contributions along the L^ax–FeO axis, thereby affecting both the magnitude and sign of A _zz .

8.

Field-dependent ⁵⁷Fe Mössbauer measurements of solid samples of [L^NHCFe^IVO­(OOCCF₃)]­OTf (4; top) and [L^NHCFe^IVO­(SAda)]­OTf (5b; bottom). For the fitting, g-values were fixed to 2.0 to avoid overparameterization. Note that insertion of the sample into the spectrometer at room temperature led to a small impurity of degradation product in the case of complex 5b, which, however, could be easily distinguished during the fitting of the spectrum; it was not possible to record spectra of 5a due to its higher instability at rt.

The FeO stretching vibration of the new complexes 4 and 5a/5b was identified by helium-tagging infrared photodissociation (IRPD) measurements in combination with ¹⁸O labeling, , where the Fe¹⁸O species were synthesized by using ¹⁸O-labeled sPhIO (for 5a/5b) or by the addition of H₂ ¹⁸O into the cold MeCN solution of the complexes (for 3 and 4). This method gives gas-phase ν̃_Fe–O values for the cations [L^NHCFeO­(X)]⁺ (X^– = CF₃COO^–, ^t BuS^–, AdaS^–) unperturbed by solvation. In solution, the oxygen atom of the FeO bond interacts with solvent molecules, leading to slightly more polarized FeO bonds. On average, in acetonitrile, the gas-phase FeO stretching frequency blue-shifts by +9 cm^–1. Hence, the effect is small and consistent across similar complexes in a given solvent. The same method was previously used to determine ν̃_Fe–O = 832 cm^–1 (Δ­(¹⁸O–¹⁶O) = −33 cm^–1) for the parent complex 3 with a trans-axial MeCN (Figure A), which is comparable to the FeO stretching vibration of TMC-based A ^NCMe (ν̃_Fe–O = 839 cm^–1, Δ­(¹⁸O–¹⁶O) = −35 cm^–1).

9.

Infrared photodissociation (IRPD) spectra for complexes A) 3, B) 4, and C) 5a (black) and their respective Fe¹⁸O isotopologues (red).

In case of 4, the FeO stretch appears at slightly lower energy at ν̃_Fe–O = 816 cm^–1 (Δ­(¹⁸O–¹⁶O) = −39 cm^–1; Figure B), which is in line with the slightly elongated FeO bond length determined by scXRD (1.671(1) Å for 4 vs 1.661(3) Å for 3*); the difference in ν̃_Fe–O indeed translates into Δ­[d(Fe–O)] of ∼0.01 Å according to a Badger’s rule analysis. , The slight weakening of the FeO bond in 4 can be explained by the weak π-donor CF₃COO^– in the trans-axial position competing with the oxido moiety, in contrast to the weak π-acceptor MeCN in 3 (or EtCN in 3*). However, the FeO stretch for A ^TFA (854 cm^–1, Δ­(¹⁸O–¹⁶O) = −37 cm^–1) occurs at higher energy than for A ^NCMe (839 cm^–1). This difference is likely due to solvation of the CF₃COO^– anion, which reduces its π-donor ability, an effect that is absent in the IRPD measurements (note that IRPD affords ν̃_Fe–O for the bare cation in the gas phase, while the Raman spectra for A ^NCMe and A ^TFA were recorded in MeCN solution).

In the present work it was possible for the first time to determine the FeO stretching frequency for a synthetic nonheme oxidoiron­(IV) complex with trans-axial thiolate: a very low ν̃_Fe–O = 761 cm^–1 is observed for 5a, shifting to 730 cm^–1 upon ¹⁸O-labeling (Δ­(¹⁸O–¹⁶O) = −31 cm^–1; Figure C). A similar value ν̃_Fe–O = 753 cm^–1 was obtained for complex [L^NHCFeO­(SAda)]⁺ (5b; see Figure S5). According to a normalized Badger’s rule analysis, , this would correspond to a long FeO bond of d(Fe–O) ≈ 1.69/1.70 Å in 5a/5b, very similar to d(Fe–O) = 1.70(2) determined by EXAFS for A ^TMCS . Also, DFT calculations give a long FeO bond for 5a (1.672 Å) compared to 3 and 4 (1.645 and 1.644 Å, respectively; see Table S12). These values indicate substantial weakening of the FeO bond in 5a/5b, which has been assumed to translate into higher reactivity of oxidoiron­(IV) intermediates with trans-axial thiolate toward hydrogen atom abstraction (HAA). The results also underline the very different electronic structures for 5a/5b and oxidoiron­(IV) complex C with a cis-thiolato ligand, the latter showing an extremely short Fe–O bond (vide supra). All the important experimental parameters are summarized in Table .

1. Experimental Structural and Spectroscopic Parameters for All Complexes [L^NHCFeO­(L)]­(OTf)₂ (L = MeCN/EtCN) and [L^NHCFeO­(X)]­OTf (X = TFA, S ^t Bu, SAda).

complex	λmax [nm]	d(Fe–O) [Å]	δ [mm s–1]	ΔE Q [mm s–1]	D [cm–1] (SQUID)	D [cm–1] (MB)	ν̃Fe−O [cm–1]
3/3* ,	400, 610	1.661	–0.13	3.08	16.4	15	832
4	408, 629	1.671	–0.14	3.13	19.0	17.6	816
5a	442, 577	-	–0.09	2.21	14.5	-	761
5b	447, 586	-	–0.09	2.23	15.7	18.4	753

Magnetic Circular Dichroism (MCD) and UV−vis Studies

Magnetic Circular Dichroism Spectroscopy (MCD) is a powerful technique to investigate the electronic structures of the ground state as well as the excited states, especially for transition metal complexes. Earlier detailed investigations showed that the MCD spectra of complex 3 are dominated by three ligand field transitions that were attributed to promoting an electron from Fe-3d _xy to π*_x/y (Fe-d _xz,yz ), from π*_x/y to Fe-3d _{x ²‑y ²} , and from π*_x/y to Fe-3d_z² , respectively. The first excitation is situated at ∼12000 cm^–1 in the near-infrared region and is distinguished by a characteristic vibronic progression along the Fe–O stretch of the associated excited state. The MCD signals of the transitions π*_x/y→Fe-3d_x²‑y² and π*_x/y→Fe-3d_z² at 20 K are opposite in sign; the former one gives a positive pseudo-A feature at ∼24000 cm^–1 and the latter a negative one at ∼17000 cm^–1. On the basis of these findings, the MCD spectra of the new complexes [L^NHCFeO­(X)]­OTf with X^– = CF₃COO^– (4) or ^t BuS^– (5a) can be unequivocally assigned. As depicted in Figures S7–S10, the relative energies of the three transitions all follow the same ordering. Therefore, all L^NHC-ligated oxidoiron­(IV) complexes 3, 4, and 5a feature comparable electronic structures in the sense that the Fe-3d_x²‑y² orbital lies energetically above Fe-3d_z² due to the very strong σ-donating ability of the equatorial L^NHC ligand.

MCD spectra typically exhibit higher resolution compared to UV–vis spectra and allow for a more reliable determination of the energy-level variation due to the ligand variance (shown in Figure ). Compared with parent complex 3 (MeCN), the excitation energy for π*_x/y→Fe-3d_x²‑y² in complex 4 (CF₃COO^–) is slightly lower, while this transition is significantly lower in energy by about 2000 cm^–1 for complex 5a ( ^t BuS^–). This is reflected in the UV–vis absorptions at 400, 408, and 442 nm for 3, 4 and 5a, respectively (cf. Figure ). Since Fe-3d_x²–y² is a σ-antibonding orbital in the equatorial plane, its energy is hardly affected by the variation of the axial ligand. Therefore, this ordering discloses the varying destabilization degree of π*_x/y resulting from the differential π-donating ability of the axial ligand, congruent with the well-accepted sequence MeCN < CF₃COO^– < ^t BuS^–. However, the transition energy for π*_x/y→Fe-3d_z² follows a different order, namely ^t BuS^– (5a) > MeCN (3) > CF₃COO^– (4) (cf. UV–vis absorptions at 577, 610, and 629 nm). This reflects the combined effect of σ- and π-donating capability of the axial ligand, which ought to destabilize Fe-3d_z2 and π*_x/y, respectively, yet to a varying extent. CASSCF/NEVPT2 calculations nicely reproduced the observed trend of the transition energy variations (Table S10), although the energy differences are not significant, presumably due to the weak binding of the axial ligand.

10.

Overlay of the 20 K MCD spectra of complexes 3 (black), 4 (red), and 5a (blue).

We performed a Franck–Condon analysis of the vibronic progression observed for the transition Fe-3d _xy →π*_x/y for 4 and 5a, following the same procedure previously employed (data for 3 are included in Figure for comparison). , The average spacing of vibronic peaks represents the excited-state Fe–O stretching frequency. Moreover, the intensity of the n ^th progression lines follows a Poisson distribution,

$$\frac{I_{0-n}}{I_{0-0}}=\frac{{S_{HR}}^n}{n!}$$

in which I _0–0 and I _0–n are the intensities of the 0–0 and 0–n vibronic transitions, while S_HR denotes the Huang–Rhys factor that describes the displacement of the excited state potential energy surface relative to that of the ground state,

$$S_{HR}=\frac{\frac{1}{2}{k}_{es}\mathrm{\Delta }{r}^2}{h{v}_{es}}$$

in which k _es is the force constant of the excited-state Fe–O stretching mode, hν _es is the corresponding vibrational energy, and Δr is the elongation fraction of the Fe–O distance.

Since the intensity of this transition first increases and then decreases upon elevating the temperature, the spectra recorded at 20 K and under a 10 T magnetic field were chosen for vibronic analyses, and the best simulations are shown in Figure and Figures S11–S13. Because hν _es, S _HR, and the line width strongly correlate with each other, especially when the spectral resolution is not sufficiently high, the error bar of the simulation is somewhat underestimated. Despite this difficulty, the excited-state Fe–O stretching frequencies obtained from the simulations (Figure ) are all substantially lower than those determined for the ground state, because this Fe-3d _xy →π*_x/y excitation weakens the Fe–O π bond. The vibrational energy of this series follows the order ^t BuS^– (5a) < MeCN (3) ≈ CF₃COO^– (4), consistent with the relative ν_Fe–O measured for their ground states.

11.

Franck–Condon progression simulation of 20 K MCD spectra of intra-t_2g transitions in [Fe^IVO­(L^ax)]ⁿ⁺ series.

In addition to those ligand field transitions discussed above, complexes 5a/5b exhibit a new band at ∼14000 cm^–1 (∼710 nm), which possesses a comparable intensity to the intra-t_2g transition in the UV−vis and MCD spectra, while no analogous transitions were detected for 3 or 4. We thus tentatively attribute it to a weak charge transfer transition from the thiolate to the Fe^IVO unit (S→Fe^IVO). In line with this reasoning, a similar band at 1040 nm was also observed in A ^TMCS , whereas complex A ^NCMe is devoid of such a low-energy NIR feature.

C–H Bond Activation Reactivity

The present series of oxidoiron­(IV) complexes with equatorial tetra­(NHC) ligation features a unique electronic structure with the Fe-3d_x²‑y² orbital high in energy and a very large triplet-quintet gap, leading to triplet-state-only reactivity. Hence, a comparison of their HAA reactivity should reveal intrinsic axial ligand effects not being overshadowed by different extents of two-state reactivity. As previously shown, complex 3 is able to activate weak C–H bonds up to around 80 kcal mol^–1. Therefore, the reactivity of 3, 4, and 5a was studied with a set of substrates containing relatively weak C–H bonds, viz. 1,4-cyclohexadiene (CHD; BDE = 78 kcal mol^–1), 9,10-dihydroanthracene (DHA; BDE = 77 kcal mol^–1), 9H-xanthene (XanH; BDE = 75.5 kcal mol^–1) and 10-methyl-9,10-dihydroacridine (AcrH; BDE = 73.7 kcal mol^–1).

For the previously studied parent complex 3, upon addition of the above substrates, the formation of a hydroxidoiron­(III) intermediate with UV–vis absorptions at 448 and 558 nm was observed. In a slower follow-up process, this intermediate reacted further to give a μ-oxidodiiron­(III) product, [(L^NHCFe)₂O]⁴⁺, with five-coordinate ferric ions. , In contrast to this, upon addition of the C–H substrates to a solution of 4 in MeCN, the UV–vis features of 4 at 408 and 629 nm decrease, and a band at 339 nm is rising (Figure , top), which can be attributed to the formation of iron­(II) complex [L^NHCFe­(NCMe)₂]²⁺ (1). Isosbestic points are present only during a first phase of the reaction (around 400 s at −40 °C for the substrate CHD), indicating that this reaction is a multistep process proceeding via a first HAA and a different follow-up reactivity compared to complex 3, likely a second HAA by the initially formed hydroxidoiron­(III) intermediate to finally give 1. Stronger binding of the anionic CF₃COO^– compared to MeCN may prevent axial ligand dissociation from the initial hydroxidoiron­(III) intermediate and subsequent formation of the dimeric μ-oxidodiiron­(III) species with five-coordinate ferric ions. Upon addition of substrate to the in situ formed complex 5a (Figure , bottom), the characteristic band at 447 nm for 5a vanishes and a prominent new absorption band at 609 nm arises. The presence of multiple isosbestic points suggests a single-step transformation during HAA, forming the hydroxidoiron­(III) intermediate [L^NHCFe^III(OH)­(S ^t Bu)]⁺. This is further supported by EPR spectroscopy of the frozen solution after HAA, which shows a rhombic spectrum characteristic for a low-spin (S = 1/2) Fe^III species (Figure S2). In contrast to the other hydroxidoiron­(III) intermediates lacking the trans-axial thiolate, [L^NHCFe^III(OH)­(S ^t Bu)]⁺ is stable over the course of 4 h at −40 °C.

12.

Spectral evolution of the optical features during the reaction of complex 4 (top) and 5a (bottom) with 200 eq CHD in MeCN at −40 °C.

Analysis of the products obtained from the oxidation of DHA and CHD by oxidoiron­(IV) complexes 3, 4, and 5a showed the formation of anthracene and benzene in yields of 82–87% and 88–100%, respectively (Table ). These yields were calculated for a nonrebound process, in which a maximum of 0.5 eq anthracene/benzene per 1 eq Fe^IVO complex can be obtained since two H atoms need to be abstracted from either substrate, and no evidence for any oxygenated products could be found. For quantifying the products in these experiments, crystalline material of the oxidoiron­(IV) complex was used in the case of 3 and 4. Due to its instability, however, complex 5a had to prepared in situ from its Fe^II precursor, which required the use of 2.8 eq of sPhIO for full conversion. Hence, some reoxidation of the hydroxidoiron­(III) intermediate (which forms after the HAA step) by excess sPhIO cannot be fully excluded, and product yields for 5a should thus be viewed with caution. For substrates AcrH and XanH, some oxygenated products were observed, likely resulting from radical rebound after the HAA step. This has been confirmed by using ¹⁸O-labeled oxidoiron­(IV) complexes (prepared from sPhI¹⁸O) and ESI mass spectrometry, showing the formation of some Xan¹⁸O (¹⁸O-labeled 9H-xanthen-9-one) and Acr¹⁸O (¹⁸O-labeled 10-methylacridin-9­(10H)-one, see SI for more information about the product quantification and labeling experiments).

2. Product Quantification of the Oxidation of DHA and CHD for Complexes 3, 4, and 5a.

	3	4	5a
Yield anthracene [%]	84	87	82
Yield benzene [%]	100	90	88

Kinetic data were obtained using in situ prepared complexes 4 and 5a (0.25 mM; prepared with sPhIO at −40 °C in MeCN) and varying excess of CHD under pseudo-first-order conditions. In the case of 4, the decrease in the absorption of the oxidoiron­(IV) species at 408 nm was monitored until the break of the isosbestic points (at 0.25 mM concentration, it was not reliable to follow the absorption at 629 nm due to its low intensity). For complex 5a, the increase in the absorption band corresponding to the formed hydroxidoiron­(III) intermediate (609 nm) was monitored. Similarly, for 3, the bands corresponding to the formation of the hydroxidoiron­(III) intermediate (448 or in the range 525−550 nm) were chosen for reaction monitoring. As expected, complexes 5a and 5b show very similar reactivities (reactivity with DHA at 25 °C is depicted in Figure S20; see also Table S6), and therefore only complex 5a was studied in greater detail. Plots of the observed rates k _obs versus the concentration of CHD for complexes 4 and 5a are shown in Figure (left). The obtained second-order rate constants k ₂ were corrected by the number of abstractable hydrogen atoms of the substrate used (four abstractable hydrogen atoms for CHD) to obtain k ₂’ constants compiled in Table . The parent complex 3 and the new complexes 4 and 5a, with trans-axial CF₃COO^– and ^t BuS^–, respectively, exhibit second-order rate constants that differ by not more than a factor of 10. Remarkably, the parent complex 3 shows the fastest rates among the complexes studied, while rates for 4 and 5a are similar. This is contrary to the antielectrophilic trend observed for type A complexes based on the TMC macrocycle, where more donating axial ligands increased the reaction rate: compared to A ^NCMe , the rate for HAA from the substrate CHD was higher by 1 order of magnitude in the case of A ^TFA , and using DHA as the substrate, the reaction proceeded even 38 times faster with A ^TMCS . This antielectrophilic trend for the TMC-based series of complexes was rationalized by the reduction of the orbital energy gap between the triplet and the more reactive quintet states, ΔE _Q‑T, with increasing donor strength of the axial ligand (vide supra). , For the present tetra­(NHC)-ligated systems participation of the quintet state is prohibited, and hence the most electrophilic 3 clearly shows the highest reaction rates. i.e., the trend of HAA reactivity is inverted compared to the TMC-based systems and follows the expected electrophilicity pattern. It should also be noted that 3 (with trans-axial MeCN) reacts 2–3 orders of magnitude faster compared to the corresponding TMC-based A ^NCMe , despite 3 being confined to the triplet pathway.

13.

Plots of the observed reaction rates k _obs vs. concentration of CHD in MeCN at −40 °C (left) for complexes 4 (0.25 mM, following the band at 408 nm) and 5a (0.25 mM, following the band at 609 nm) and at 25 °C (right) for complexes 3 (1 mM, following band at 550 nm), 4 (1 mM, following the band at 629 nm), and 5a (1 mM, following the band at 609 nm).

3. Second-Order Rates k ₂ and k ₂’ for Complexes 3, 4, and 5a with CHD at −40 and 25 °C in MeCN.

	3		4		5a
T [°C]	–40	25	–40	25	–40	25
k 2(CHD) [M–1 s–1]	0.48	25.0	0.09	2.5	0.13	3.5
k 2’(CHD) [M–1 s–1]	0.12	6.3	0.02	0.6	0.03	0.9

Figure and Table also contain data from kinetic traces recorded at 25 °C and 1 mM complex concentrations; in case of 4, this allowed to monitor the disappearance of the band at 629 nm, which does not overlap with spectroscopic changes during follow-up processes. Trends in k _obs and k ₂ are similar at 25 °C, with 3 featuring around 10 times higher rates, while differences between the rates of 4 and 5a are rather minor. Computational studies suggested that a five-coordinate oxidoiron­(IV) complex [L^NHCFeO]²⁺ (3′) devoid of any trans-axial ligand should be significantly more reactive because dissociation of the trans ligand stabilizes the σ*­(Fe-3d_z² ) that could then serve as a more efficient electron-accepting orbital (EAO) in the HAA reaction step. ,, Despite the relatively small Gibbs energy for MeCN binding (2.3 kcal mol⁻¹ based on calculations, vide infra), we consider potential dissociation of the trans-MeCN from 3 and involvement of 3′ to be unlikely under the present conditions because the experiments were performed in MeCN solution; this is also supported by the much better agreement of the experimental Gibbs free activation barrier with that calculated for MeCN-ligated 3 (vide infra).

Eyring analyses of the reactions of complexes 3, 4, and 5a with CHD (Figure , left) reveal that ΔG ^‡ _20 °C is lowest for 3, but overall, the three complexes show rather similar free energy barriers (Table ), reflecting the relatively minor variations in HAA rates.

14.

Left: Eyring plot of the reactions of complexes 3, 4, and 5a with 75 eq CHD at different temperatures. Right: Plot of log­(k ₂’) against the BDE of different substrates (AcrH, XanH, DHA, CHD) for the three complexes.

4. Activation Parameters for the Reaction of Complexes 3, 4, and 5a with CHD, Derived from an Eyring Analysis.

	3	4	5a
ΔH ‡ [ kcal mol−1]	8.9 ± 0.3	5.2 ± 0.2	8.2 ± 0.3
ΔS ‡ [cal K/mol]	–22.1 ± 1.0	–38.8 ± 0.7	–28.7 ± 0.9
ΔG ‡ 20 °C [ kcal mol−1]	15.4 ± 0.6	17.6 ± 0.4	16.6 ± 0.5

Plots of log­(k ₂’) against the bond dissociation energy (BDE) of the four different substrates (Figure , right) indicate linear Bell-Evans-Polanyi type correlations with slopes of −0.31 (complex 3), −0.33 (complex 4) and −0.37 (complex 5a), which is in the range of previously studied oxidoiron­(IV) complexes ,, and is strong evidence for HAA as the rate-limiting step. It also suggests that the proton-coupled electron transfer (PCET) process is concerted and not significantly biased toward a mechanism of initial proton transfer (PT) or electron transfer (ET).

Kinetic Isotope Effects and Tunneling Contributions

To gain further mechanistic information, temperature-dependent kinetic isotope effects (KIEs) were explored. Reactions of complexes 4 and 5a with DHA and deuterated DHA (DHA-d₄) at different temperatures were compared with previously obtained data for complex 3 (three temperatures were already studied before, and the rt measurements for 3 were added in this work); the results are shown in Figure (left). KIE values obtained at rt (Table and Figure ) are higher with increasing donor strength of the axial ligand, from 4.8 for complex 3 (MeCN) via 8.7 (CF₃COO^–) for complex 4 to 17.1 for complex 5a ( ^t BuS^–). At −40 °C a rather high KIE for complex 3 of 32 is obtained (Table ). Similarly high values were observed for the natural enzyme taurine α-KG dioxygenase (∼35), whereas most TMC-coordinated iron complexes show values between 10 and 20 (at 0 °C in MeCN). , Linear extrapolation of KIE values for 4 and 5a to −40 °C gives values around 39 (±9) for complex 4 and 109 (±29) for complex 5a (see SI for details of the extrapolation; note that below 10 °C, 4 and 5a react too slowly with DHA-d₄ to determine KIEs experimentally). These enormously high KIE values in the reaction with DHA, in particular for 5a, support HAA as a rate-determining step in these reactions and suggest increased nonclassical hydrogen tunneling effects. This is in line with reports for TMC-based ferryl complexes [(TMC)­Fe^IVO­(L^ax)]ⁿ⁺ where A ^TMCS with a strongly donating trans-axial thiolate showed the largest tunneling-augmented KIE (80 at −20 °C in MeOH). It should be noted that the ΔH ^‡ extracted from Eyring plots of the H-transfer rate (Table ) may be smaller than the “true” (zero-point-corrected) barrier when substantial quantum-mechanical tunneling occurs, because tunneling allows the reaction to proceed below the classical transition-state energy; the corresponding apparent ΔS ^‡ may consequently be driven in the opposite direction.

15.

Left: Plot of k _obs against c(DHA-(d ₄)) at 25 °C; right: plot of ln­(KIE) vs. RT^–1.

5. Second-Order Rate Constants for the Reaction of Complexes 3, 4, and 5a with DHA and Deuterated DHA-d₄ at 25 °C and the Corresponding KIE Values.

	3	4	5a
k 2’(DHA) [M–1 s–1]	10.0	1.0	3.2
k 2’(DHA-d 4) [M–1 s–1]	2.1	0.1	0.2
KIE	4.8	8.7	17.1

6. Experimental Temperature-Dependent KIE Values for Complexes 3, 4, and 5a and Their Analysis.

	3	4	5a
KIE (−40 °C)	32	-	-
KIE (−20 °C)	18	-	-
KIE (0 °C)	11	-	-
KIE (10 °C)	-	11.4	23.8
KIE (15 °C)	-	10.6	22.5
KIE (20 °C)	-	9.1	18.8
KIE (25 °C)	4.8	8.7	17.1
E a(D)–E a(H) [kcal mol−1]	4.0 ± 0.4	3.2 ± 0.5	4.0 ± 0.6
ln(A H/A D)	–5.1 ± 0.7	–3.3 ± 0.9	–3.8 ± 1.0

The importance of tunneling can be evaluated by temperature-dependent KIE analysis, which allows to assess the influence of the tunneling effects on the activation energy and on the Arrhenius prefactor. , Therefore, ln­(KIE) was plotted versus (RT)⁻¹, where the slope of this plot gives the differences in activation energy (E _a(D)–E _a(H)) and the intercept represents the difference between the exponential prefactors (ln­(A _H/A _D)). With these parameters, different regimes with increasing importance of tunneling can be defined. In semiclassical treatments, KIEs are often seen as the loss of zero-point energy in the transition state. This scenario mainly affects the isotope effect on the activation energy, whereas there should be just a minimal effect on A. In this case, the intercept is expected to be zero, while the slope should be smaller than or equal to 1.4 kcal mol^–1. In this context, proton coupling would increase the magnitude of k _H and decrease the temperature dependence, yielding values for the slope greater than 1.4 kcal mol^–1 and extrapolated intercepts below zero. If quantum mechanical treatments are applied, both H and D can tunnel intensively. This would lead to a small effect on the slope and a positive intercept. All complexes 3, 4, and 5a studied in this work show an extrapolated intercept lower than zero and a temperature dependence steeper than 1.4 kcal mol^–1 (Figure right and Table ). This is in agreement with a semiclassical transition state theory with modest tunneling contributions. Therefore, experimental findings indicate that all complexes show similar tunneling behavior, and that differences are rather subtle, in accordance with the overall rather similar free energy barriers derived from Eyring analyses. A more detailed comparison was then pursued computationally.

Theoretical Investigations on HAA Reactivity

To obtain detailed insights into the reactivity differences between complexes 3, 4 and 5a, we performed DFT-based investigations of the reaction energetics and electronic structures. Along with complexes 3, 4, and 5a, the five-coordinated version of complex 3 devoid of the trans ligand (3′) was also investigated for the HAA reactivity with the substrate CHD. The reaction energetics (ΔG, kcal mol⁻¹) are presented with respect to the free species (complex + CHD) to approximately account for the entropic contribution to the reaction barriers (Figure ). For complexes 3 and 3′, both HAA transition states for the σ- (³TS_σ) and π-pathway (³TS_π) were successfully located. In line with previous calculations, , for the five-coordinated complex 3′ the σ-pathway involves a lower barrier (11.0 kcal mol⁻¹) than the π-pathway (15.7 kcal mol⁻¹), whereas the scenario is the opposite in the case of the six-coordinated species 3 with trans-bound MeCN, where the π-pathway features a slightly lower barrier (15.1 kcal mol⁻¹) than the corresponding σ-pathway (15.3 kcal mol⁻¹). Most importantly, this direct comparison of the HAA barriers reflects the destabilization of the σ*­(Fe-3d_z² ) by the trans-axial ligand, which disfavors the σ-pathway where the Fe-3d_z² serves as EAO. While the formation of the five-coordinate species 3′ through the dissociation of MeCN appears feasible in light of the almost thermoneutral MeCN binding free energy of +2.3 kcal mol⁻¹, the experimentally determined value of ΔG ^‡ _20 °C (15.4 ± 0.6 kcal mol⁻¹) matches closely with the DFT calculated value for six-coordinate 3 (15.1 kcal mol⁻¹), while the calculated barrier for the five-coordinate complex 3′ is much lower (11.0 kcal mol⁻¹). This strongly supports the assumption that 3′ is not relevant under the present conditions when the reactions are performed in MeCN solution (vide supra).

16.

Computed Gibbs free energy (ΔG, kcal mol⁻¹) profiles of the HAA from CHD by complexes 3′ (five-coordinated), 3, 4, and 5a at the DFT-B3LYP level of theory. Optimized geometries of the lowest energy transition states, their key geometrical parameters (in red), and imaginary frequency are presented along with the corresponding reaction profile. The lower energy pathway is presented with black lines, and the higher energy one is presented with gray lines.

For complex 4 with a trans-CF₃COO^– both the σ and π HAA transition states could be successfully located, with the π-pathway lying marginally lower in energy (Figure ). The calculated Gibbs free energy barrier for the lower-energy π-pathway is 17.3 kcal mol⁻¹, which is in excellent agreement with the experimentally determined value of 17.6 ± 0.4 kcal mol⁻¹ (Table ). Unlike the five-coordinate complex, 3′, both six-coordinate complexes 3 and 4 have the π-pathway (³TS_π) lower in energy as compared to the σ-pathway, because any trans-ligand raises the energy of the EAO, Fe-d_z² , involved in the σ-pathway.

The CASSCF­(12,12)-derived natural orbitals do not show any appreciable differences in the electronic structures of complexes 3′, 3, 4, and 5a. However, the calculated d-d transition energy d _xz,yz → d_z² is appreciably higher for complex 5a (Table S10). This indicates an energetically high-lying Fe-d_z² orbital for complex 5a as compared to the other complexes, in agreement with the UV–vis and MCD data (vide supra; Figures and ). Therefore, the Fe-d_z² EAO for the σ-pathway becomes inaccessible for 5a. In line with this reasoning, the σ transition state for HAA could not be located in case of 5a despite the best of our efforts. Only the π transition state (³TS_π) involving a Gibbs free energy of 18.4 kcal mol⁻¹ could be located, which is about ∼1 kcal mol⁻¹ higher as compared to that of complex 4. While this DFT-calculated HAA barrier for 5a agrees reasonably well with the experimentally determined value (16.6 ± 0.5 kcal mol⁻¹), the agreement is worse than for 3 and 4, and the relative order of ΔG ^‡ _20 °C for 4 and 5a in the calculations and experiments differs. We attribute this to a more significant tunneling contribution for the HAA pathway of 5a, which effectively lowers the barrier compared to complex 4, in agreement with the significantly higher KIE value determined for complex 5a (KIE­(25 °C) = 17.1 for 5a vs 8.7 for 4; Table ). Indeed, if we take a closer look at the calculated intrinsic reaction coordinates (IRCs) of complexes 3, 4, and 5a (Figure ), the transition state (³TS_π) associated with complex 5a involves an appreciably smaller width as compared to complex 4, suggesting a lower effective barrier due to increased H-tunneling. This is in accordance with reports on the TMC-based ferryl complexes, where A ^TMCS with a strongly electron-donating trans-thiolate favors tunneling in comparison to other [(TMC)­Fe^IVO­(L^ax)]ⁿ⁺ complexes. ,

17.

Intrinsic reaction coordinates (IRCs) calculated for the lowest energy transition states ³TS_π for complexes 3, 4, and 5a. The estimated widths (in bohr·amu^1/2) and full width at half maxima (fwhm), are shown along with the IRCs.

According to the Bell–Evans–Polanyi (BEP) principle, the reaction barrier is correlated to the driving force. Therefore, the bond dissociation free energies (BDFE) of the Fe^III–OH intermediates that result after the HAA step would be a reflection of the relative HAA reactivity. The calculated BDFEs of the intermediates associated with the lowest energy ³TS_π pathways for complexes 3, 4, and 5a (³Int_π) are 80.0, 75.6, and 77.8 kcal mol⁻¹, respectively (and 88.2 kcal mol⁻¹ for 3′, ³Int_σ). This nicely mirrors the trends in reactivity observed experimentally (3 > 5a > 4) and the relative ΔG ^‡ _20 °C values derived from the Eyring analysis. Importantly, a higher driving force is found for the π-pathway of 5a as compared to that of 4.

Summary and Conclusions

In this work, a set of Fe^IVO complexes supported by a macrocyclic tetracarbene ligand and bearing different anionic trans-axial ligands CF₃COO^– (4), ^t BuS^– (5a) and AdaS^– (b) has been isolated and thoroughly investigated. Together with our previously reported tetracarbene-ligated Fe^IVO complex with trans-axial MeCN (3) (or EtCN in 3′) this constitutes a valuable series [(L^NHC)­Fe^IVO­(L^ax)]ⁿ⁺ to experimentally probe the effect of axial ligands on the HAA reactivity of biorelevant Fe^IVO intermediates toward C–H substrates, complementing and contrasting the findings for the only other comparable series of Fe^IVO species in a tetragonal ligand field, viz. the cyclam-based [(TMC)­Fe^IVO­(L^ax)]ⁿ⁺. The molecular structures of 4 and 5b have been established crystallographically, which in combination with the previously reported structure of 3 provides a unique set of structurally authenticated organometallic Fe^IVO species that differ only by the trans-axial ligand, including a trans-thiolate in 5b. Their electronic structures have been comprehensively investigated using a bouquet of methods including magnetic measurements, field-dependent ⁵⁷Fe Mössbauer and MCD spectroscopies, and all [(L^NHC)­Fe^IVO­(L^ax)]ⁿ⁺ exhibit a well-isolated S = 1 ground state with nb­(Fe-d _xy )²π*­(Fe-d _xz,yz)²σ*­(Fe-d _{z ²} )⁰σ*­(Fe-d _{x ²–y ²} )⁰ configuration where the strong σ-donor character of the equatorial tetracarbene ligand lifts the Fe-d _{x ²–y ²} energetically above the Fe-d _{z ²} . In agreement with DFT and CASSCF calculations in this and prior work, ,, this unusual orbital ordering leads to a very large triplet-quintet gap that makes the quintet state inaccessible. This prevents significant mixing of the S = 1 and S = 2 states, which translates into rather small zero-field splitting parameters D < 20 cm^–1 determined by SQUID magnetometry and applied field Mössbauer spectroscopy. Both experiment and theory show Fe^IVO stretching frequencies in the order 3 ≥ 4 > 5a, with a very low ν̃_Fe–O = 761 cm^–1 for 5a (the first experimentally determined ν̃_Fe–O for a synthetic nonheme oxidoiron­(IV) complex with trans-axial thiolate). This correlates with the order of FeO bond lengths, and the elongation of the Fe^IVO bond upon trans-thiolato coordination likely contributes to the increase of the ⁵⁷Fe Mössbauer isomer shift δ in 5a/5b (−0.09 mm s^–1) compared to 3 and 4 (−0.13 and −0.14 mm s^–1, respectively).

Because of their very large triplet-quintet gap, the HAA reactivity of all L^NHC-based Fe^IVO complexes is restricted to triplet pathways, which allows intrinsic trends in hydrogen atom abstraction (HAA) to be identified experimentally without interference from two-state reactivity (TSR). DFT calculations suggest that for all these six-coordinate complexes the π-pathway, where the π*_x/y(Fe-d _xz,yz) are the EAOs, is favored. For 3 and 4 with weak σ-donor trans-axial MeCN and CF₃COO^–, ³TS_σ is only slightly above ³TS_π, but in the case of 5a/5b the destabilization of the σ*­(Fe-3d _{z ²} ) by the trans-axial thiolato ligand is so pronounced that the Fe-d _{z ²} EAO for the σ-pathway becomes inaccessible. Kinetic measurements with substrates featuring C–H bond dissociation energies up to 78 kcal mol^–1 reveal that the different axial ligands have only a modest impact on reactivity, following roughly an electrophilic order of L^ax = MeCN > CF₃COO^– ≈ RS^–. This unusual minor effect of axial ligands on HAA corroborates that TSR is not involved for the present series of L^NHC-ligated complexes, and it contrasts the findings for the related series of TMC-based systems, where TSR dominates and leads to opposite HAA reactivity trends (L^ax = MeCN < CF₃COO^– < RS^–). ,

The L^NHC-based Fe^IVO complexes show high KIE values in the HAA reaction with DHA, the KIE being highest for 5a with a trans-thiolate. Such large KIEs for HAA processes are usually indicative of reactive S = 1 spin states of the TS. A similarly large KIE has also been observed for the TMC-based Fe^IVO complex with a trans-thiolate, A ^TMCS , which proceeds via an S = 2 TS. This was reported to reflect increased tunneling with stronger donicity of the trans-axial ligand L^ax, which was proposed to be responsible for the antielectrophilic HAA reactivity trend found for the TMC-based systems. For the present series of L^NHC-based Fe^IVO complexes, experimentally determined temperature-dependent KIEs, supported by density functional theory and ab initio calculations, are consistent with a semiclassical reaction model involving moderate tunneling contributions. Indeed, calculated intrinsic reaction coordinates show a barrier for the relevant (³TS_π) transition state that is only slightly narrower in the case of 5a with a trans-thiolate, which we surmise is related to the FeO bond elongation in 5a/5b inducing a slightly more symmetric O···H···C transition state.

For TMC-based systems with various axial ligands, ⁵TS_σ on the S = 2 surface generally is the lowest energy TS, with a close-to-linear Fe–O···H angle and σ*­(Fe-d _{z ²} ) being the EAO. In that case, a stronger donicity of the trans-axial ligand, L^ax, was found to have a major effect on the slimness of the potential energy barrier along the reaction coordinate by increasing the alternating charges in the O···H···R arrangement, thus leading to a prevailing tunneling effect on HAA (with A ^TMCS being much more reactive than A ^NCMe ). ,, In contrast, the present L^NHC-based triplet-only systems follow a π-trajectory with a bent Fe–O···H angle and the π*_x/y(Fe-d _xz,yz) serving as EAO. Here, the donicity of L^ax appears to have a much smaller effect on the barrier width along the reaction coordinate, and hence tunneling only attenuates the electrophilic trend, but it does not reverse it. Hence, 3 with L^ax = MeCN remains more reactive in HAA than thiolato-ligated 5a/5b. We conclude that TSR remains a dominant factor leading to the antielectrophilic trend in the case of the TMC-based Fe^IVO complexes, while the present series of L^NHC-based Fe^IVO systems reveals the intrinsic and rather small effect of the trans-axial ligands for a pure triplet state scenario. Overall, this combined experimental and computational study clarifies how axial ligandsincluding thiolatesaffect the spectroscopic signatures, properties, and reactivity of S = 1 ferryl intermediates.

Glossary

Abbreviations

AcrH

9-methylacridine

BDE

bond dissociation energy

CHD

cyclohexadiene

DFT

density functional theory

DHA

dihydroanthracene

ET

electron transfer

HAA

hydrogen atom abstraction

HAT

hydrogen atom transfer

IRPD

infrared photodissociation

KIE

kinetic isotope effect

MCD

magnetic circular dichroism

OAT

oxygen atom transfer

TFA

trifluoroacetate

Xan

xanthene

TPA

tris­(2-pyridylmethyl)­amine

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacsau.6c00781.

• Synthetic procedures and materials; further spectroscopic (Mössbauer, UV–vis, EPR, MCD) and magnetic data; details of the kinetic experiments; detailed crystallographic information; details about DFT and CASSCF calculations; Cartesian coordinates of all optimized structures (PDF)

∇.

I.B. and M.M. contributed equally to this work.

The authors declare no competing financial interest.

All other data for this paper, including data from experimental methods and computational studies, are available at the Göttingen Research Online server at https://doi.org/10.25625/SS9KST.