π-Frontier molecular orbitals in S = 2 ferryl species and elucidation of their contributions to reactivity

Abstract

S = 2 Fe^IV═O species are key intermediates in the catalysis of most nonheme iron enzymes. This article presents detailed spectroscopic and high-level computational studies on a structurally-defined S = 2 Fe^IV═O species that define its frontier molecular orbitals, which allow its high reactivity. Importantly, there are both π- and σ-channels for reaction, and both are highly reactive because they develop dominant oxyl character at the transition state. These π- and σ-channels have different orientation dependences defining how the same substrate can undergo different reactions (H-atom abstraction vs. electrophilic aromatic attack) with Fe^IV═O sites in different enzymes, and how different substrates can undergo different reactions (hydroxylation vs. halogenation) with an Fe^IV═O species in the same enzyme.

The key reactive intermediates in the catalytic cycles of many mononuclear nonheme iron (NHFe) enzymes are known to be S = 2 Fe^IV═O species capable of abstracting an H-atom from an inert C─H bond as strong as 106 kcal/mol (1–5). These transient intermediates have been difficult to trap in appreciable concentrations amenable to spectroscopic studies, but recent developments hold promise for future geometric and electronic structural characterization (4, 6).

The synthesis and structural characterization of stable biomimetic model complexes have greatly aided in understanding the geometric and electronic structures and reactivity of the Fe^IV═O unit. A series of intermediate-spin S = 1 Fe^IV═O model complexes, supported by amine, pyridine, and deprotonated amido ligands, has become available for structural, reactivity, and spectroscopic investigations (7–13). Recently, high-spin S = 2 Fe^IV═O complexes have also become available, with two being structurally defined (14–20), but notably their reactivity does not surpass that of the most reactive S = 1 complexes (15, 18, 21).

S = 1 Fe^IV═O complexes have been spectroscopically shown to possess one channel for reactivity—the β-d_xz/yz π^∗-frontier molecular orbital (FMO) (22)—and the S = 2 species had been computationally predicted to have an additional σ^∗-FMO for reactivity by stabilization of its α-d_z² orbital through spin polarization (i.e., exchange stabilization) (22–24). In the S = 2 complex (TMG₃tren)Fe^IV═O (1), which is the subject of this study, it was shown that the accessibility of this α-d_z² σ^∗-FMO is restricted by an axial steric wall of methyl groups, giving a large steric contribution to its reaction barrier, rendering it only as reactive as the S = 1 complex (N4Py)Fe^IV═O (25). Importantly, (N4Py)Fe^IV═O also has a large steric barrier because the π-channel requires an approach perpendicular to the Fe─O bond for orbital overlap, and this leads to the steric clash of the substrate with the chelate ligand. In one case where the ligand completely excludes π-approach, an S = 1 complex can utilize a σ-channel for reactivity through a low-lying S = 2 excited state (26).

In this study, detailed variable-temperature (VT) magnetic circular dichroism (MCD) spectroscopic studies and analyses are performed on the structurally defined S = 2 model complex 1 (17) (Fig. 1A, Inset) and further correlated to density functional theory (DFT) and multiconfigurational ab initio calculations to elucidate the electronic structure and contribution of electronic structure to reactivity. MCD spectroscopy provides a direct probe of the reactive FMOs and, when applied to 1, reveals new S = 2 π-channels at the transition state (TS). In the absence of a steric barrier (as would be the case for NHFe enzymes), these π-channels would lead to new reactivity with interesting Fe spin-state contributions. Furthermore, the MCD data reveal the presence of extensive excited-state spin-orbit coupling (SOC) interactions in the absorption spectrum—hitherto unobserved in other Fe^IV═O species—that lead to interesting spectral consequences, including a dip in the absorption (abs) band and different vibronic progressions within one excited state that reflect distorted potential energy surfaces (PES).

Fig. 1.

(A) UV-visible absorption spectrum and (Inset) structure of 1. (B) VT MCD spectrum of 1 and (Inset) VT MCD data of a higher-concentration sample in the 16,000–22,000 cm^-1 region allowing resolution of the vibronic progression. VT MCD spectra taken at 2, 10, 20, 40, and 80 K. Arrows show intensity behavior of bands with respect to increasing temperature.

Results

MCD and Absorption Spectroscopy.

Fig. 1 shows the UV-visible abs and VT MCD spectra of 1. The abs spectrum shows a weaker feature at 12,000 cm^-1 and an intense feature at 25,000 cm^-1 with a shoulder at 19,000 cm^-1. The abs band at 12,000 cm^-1, in fact, exhibits a dip in its intensity that corresponds to a sharp positive feature in the MCD spectrum at the same energy (Fig. 1, vertical dashed line). These otherwise featureless abs bands become much more well-resolved and feature-rich in the MCD spectrum, which thus allows definitive band assignments for electronic-structure elucidation.

In the near-IR (NIR) 6,000–15,000 cm^-1 region of the MCD spectra, three distinct bands centered around 11,500 cm^-1 can be identified: a positively signed Franck–Condon vibronic progression, a negatively signed vibronic progression, and a sharp positive feature (indicated by the dashed vertical line in Fig. 1) overlapping the positive progression. The temperature-dependence behaviors of the two progressions indicate that they are both (x, y)-polarized (analyzed in Scheme S1 and Fig. S1). This pair of vibronic progressions forms a derivative-shaped pseudo-A term (with its negative component at higher energy) that necessarily results from in-state spin-orbit splitting of a ⁵E excited state of the S = 2 Fe^IV═O species in C_3v symmetry. Importantly, as indicated by brackets in Fig. 1B, the vibronic spacing of the negative progression is considerably larger than that of the positive progression (ν = 880 and 710 cm^-1, respectively).

In the 15,000–30,000 cm^-1 region of the MCD spectrum, there are four bands under the envelope of the intense 25,000 cm^-1 absorption feature and its weaker low-energy shoulder. At 19,500 cm^-1 and 23,500 cm^-1, there are two z-polarized bands (from the temperature dependence of the MCD data; Scheme S1 and Fig. S1) that can be assigned to charge-transfer (CT) transitions on the basis of their small MCD-to-absorption intensity (C₀/D₀) ratios of 0.040 and 0.002, respectively. In contrast, the C₀/D₀ ratio of the pseudo-A term at 11,500 cm^-1 (vide supra) is 0.250, which indicates that this ⁵E state corresponds to the lowest-energy ligand-field (LF) transition. From its z-polarization, the negative band at 19,500 cm^-1 can be assigned as the lowest-energy oxo π → Fe CT transition. This has a vibronic progression with a spacing of 490 cm^-1 and an intensity distribution indicating an excited-state distortion of the Fe─O bond of approximately 0.21 Å (Table S1). Finally, the two bands at 25,500 cm^-1 and 27,500 cm^-1 have C₀/D₀ ratios of 0.002 and 0.003, respectively, and these can also be assigned as CT transitions. Their overlap precludes analysis of the VT MCD data in terms of polarization.

DFT/Time-Dependent (TD)–DFT/Complete Active Space Second-Order Perturbation Theory (CASPT2) Calculations.

DFT calculations give a geometry-optimized structure of 1 in accord with the crystal structure (Fig. S2) and predict an S = 2 ground state, in agreement with Mössbauer spectroscopy (15). This gives the (spin-unrestricted) molecular orbital (MO) energy-level diagram shown in Fig. 2. As would be expected for a trigonal bipyramidal C_3v ligand field with a strong oxo axial ligand, the d manifold is split into two degenerate e levels, the d_xy/x²-y² at lowest energy followed by the d_xz/yz pair, and finally a nondegenerate a₁ level (d_z²) that is highest in energy. Thus, for a high-spin d⁴ Fe^IV species, the four electrons are distributed unpaired into the two e sets (i.e., 4 α-d orbitals are occupied as in Fig. 2, Left), resulting in a ground state. Consequent to this electronic structure, in a spin-unrestricted formalism the α-d_z² unoccupied orbital is spin polarized to lower energy, comparable to that of the β-d_xz/yz π^∗-FMOs, a phenomenon that was identified in previous studies (22–24). Compared to Fe^IV═O S = 1 species, which possess only a π-pathway for reactivity, this shows the availability of an additional α-d_z² σ^∗-FMO σ-pathway for reactivity in an S = 2 Fe^IV═O species.

Fig. 2.

Spin-unrestricted MO energy-level diagram of 1 as determined by DFT calculations. Note the proximity in energy of the reactive α-d_z² and β-d_xz/yz FMOs and the high-lying oxo π orbitals poised for oxo → Fe d CT.

TD-DFT calculations predict that the lowest-energy LF excited state at 18,100 cm^-1, originating from the transition, involves excitation from the degenerate α-d_xz/yz to the α-d_z² level (Fig. S3). This is overpredicted in energy by around 6,500 cm^-1 as compared with the MCD experimental value of 11,500 cm^-1 for the lowest-energy ⁵E excited state (Fig. 1B). In the MCD spectrum two low-energy z-polarized CT transitions are observed at 19,500 and 23,500 cm^-1—the former is assigned in MCD and Absorption Spectroscopy based on polarization and vibronic structure as the lowest-energy oxo π → Fe CT transition. The TD-DFT calculations predict that the two lowest oxo π → Fe CT transitions are z-polarized at 23,000 and 27,000 cm^-1 (Fig. S3); thus, the two low-energy CT transitions in MCD are assigned as oxo π → Fe CT transitions, the lowest in energy being the oxo π → d_xz/yz CT. In addition, the TD-DFT calculations predict that there are several (x, y)-polarized CT transitions originating from the oxo group and equatorial nitrogens; two of these could correspond to the CT bands in MCD at 25,500 and 27,500 cm^-1, but because of the uncertainty in the polarizations of these two MCD bands, no definitive assignments can be made.

These TD-DFT assignments of the transitions for 1 are consistent with multireference CASPT2 calculations (SI Materials and Methods), showing the α(d_xz/yz → d_z²) transition to be lowest in energy and the oxo π(p_x,y) → d_xz/yz CT to be the lowest-energy CT (Table S2 and Fig. S4). For the 15,2000–30,2000 cm^-1 energy region, CASPT2 states are approximately 10,000 cm^-1 higher in energy than their corresponding bands in the MCD data; this is ascribed to the lack of inclusion of double-shell effects of the Fe 4d and oxo 3p orbitals (27).

In summary, the band at 12,000 cm^-1 in the abs data exhibits an (x, y)-polarized pseudo-A term in the MCD spectrum, allowing its assignment as the transition involving the α(d_xz/yz → d_z²) excitation. This transition is overlapped by a sharp peak in MCD leading to a dip in abs intensity. The pseudo-A term in MCD further shows different vibronic progression spacings for its two components. The lowest-energy CT band in MCD at 19,500 cm^-1 is z-polarized and assigned as the oxo π(p_x,y) → d_xz/yz CT transition from the TD-DFT and CASPT2 calculations.

Analysis

Electronic Structure.

Fano interference.

The dip in the abs spectrum and the sharp positive peak in MCD, both observed at 11,820 cm^-1 (Fig. 3), originate from a Fano-type interference/antiresonance (28–30) that occurs when a sharp transition from a spin-forbidden state overlaps and weakly interacts with a broad vibronic continuum associated with an allowed state (i.e., an intense band that only changes slowly in the vicinity of the sharp peak). This Fano interference arises from SOC between the spin-forbidden and spin-allowed states, and the fact that it is a peak in MCD but a dip in abs is caused by the different contributions of the M_s sublevels to the Fano intensity for MCD versus abs (SI Analysis).

Fig. 3.

(A) NIR abs spectrum showing a dip in the continuum background profile; (B) NIR MCD spectrum at 2 K (bold solid line) fit with three separate vibronic Franck–Condon progressions and one sharp peak, where S is the Huang–Rhys parameter (see MCD and Absorption Spectroscopy), ΔQ is the Fe─O bond distortion, and E₀₀ is the zero-phonon energy; Fano analyses of the abs dip and sharp positive MCD feature indicated by the arrow provided in Fig. S5.

MCD ⁵E pseudo-A term.

As presented above, in addition to the sharp positive feature at 11,820 cm^-1, the NIR MCD spectrum exhibits a pseudo-A term with vibronically resolved progressions, resulting from in-state SOC of the LF ⁵E(xz,yz) state. Considering the C_3v double group and the SOC operator λL_zS_z [appropriate for an (x, y)-polarized transition] acting on a set of wave functions (where λ is the many-electron state-specific SOC constant) the ⁵E state is split into five twofold degenerate sublevels separated in energy by λ (Fig. 4A). For an (x, y)-polarized transition, MCD intensity requires an applied magnetic field H parallel to the z axis of the molecule, which lowers the symmetry to C₃ and splits the doublets. For an applied field of 7 T, the positive axial zero-field splitting of the ground state (D = +5 cm^-1) (15) results in the M_s = -1 component being the lowest-energy MCD-active sublevel for the ground state (Fig. 4A). Because transitions require ΔM_s = 0, the two sublevels from the ⁵E excited state that are MCD-active are and , resulting in left circularly polarized (LCP, positive) and right circularly polarized (RCP, negative) MCD C-term intensities, respectively, with lying lower in energy than by 2λ (+200–300 cm^-1, from a fit of the pseudo-A term with two overlapping bands; SI Analysis).

Fig. 4.

(A) The ⁵E excited-state in-state spin-orbit splitting and ground-state zero-field splitting without and with magnetic field along z [for an (x, y)-polarized transition]. Sublevels labeled in terms of (, M_S). LCP and RCP transitions in MCD (numbered according to Fig. 3) indicated by vertical arrows. (B) Energy diagram connecting nonrelativistic states [(B, a) ³E and ⁵E; (B, a^′) and ⁵E)] with corresponding spin-orbit states (B, d and Fig. S6A) through in-state spin-orbit perturbed states (B, b and b^′) and two-state spin-orbit perturbed states [(B, c) ³E/⁵E; (B, c^′) ). MCD-active states indicated in red (LCP) and green (RCP).

Interactions of the ⁵E state with spin-forbidden ³E and states.

The fact that (in Fig. 1B) both the RCP and LCP progressions in the MCD spectrum of the ⁵E state have different vibronic spacings, with the RCP negative progression having a larger vibronic splitting than the ground-state value (820 cm^-1) (25), implies strong SOC interaction between one component of the ⁵E state and other nearby spin-forbidden excited states (31). Also, the presence of a Fano dip in the abs spectrum (Fig. 1A) indicates a weak SOC to the second component of the ⁵E. As presented in Fig. 4B, starting at a and a^′, two candidates are identified to be close in energy to the ⁵E state—a ³E state (Fig. 4B, a) at -30 cm^-1 and a state (Fig. 4B, a^′) at +1,210 cm^-1, relative to the ⁵E—from nonrelativistic CASPT2 calculations at the ground-state equilibrium geometry. In-state SOC is introduced (Fig. 4B, a → b and a^′ → b^′) with CASPT2/spin-orbit complete active space state interaction (SO–CASSI) calculations, splitting the ⁵E state into five doublets, each pair separated by 49 cm^-1. Next, out-of-state SOC is turned on (Fig. 4B, b → c and b^′ → c^′), showing that the ³E state interacts strongly with one sublevel (indicated in green) and the state interacts weakly with another sublevel (indicated in red). Finally, out-of-state SOC among the lowest-energy states (Fig. S6A) is included (Fig. 4B, c, c^′ → d) to assess the overall SOC perturbations to the ⁵E state.

Evaluation of the PES interactions of the ³E and states with the ⁵E state along the Fe─O coordinate (Fig. 5 C and D) shows that the strongest interactions occur within ± 0.05 Å of the ground-state equilibrium geometry. This reveals two significant insights into the electronic structure, which can be correlated with the NIR MCD spectra: (i) among the 10 sublevels (from five doublets) originating from ⁵E in-state SOC, only the sublevel, which results in LCP, positive MCD intensity (in Fig. 4B), is coupled to the spin-forbidden state, with a SOC matrix element of approximately 160 cm^-1 leading to ≤ 1% mutual admixture of the two states; and (ii) the sublevel of the ⁵E state, which results in RCP, negative MCD intensity, is coupled to the sublevel of the ³E state, with a SOC matrix element of approximately 220 cm^-1 giving rise to considerable (approximately 50%) mixing of both states caused by their energy proximity.

Fig. 5.

(A) PES of the ground state and LF ⁵E and CT excited states along the Fe─O bond coordinate, evaluating (B) the evolution of wave function characters. Note that the CT state is composed of two dominant configurations. Energies of CASPT2 PESs shifted to match experimental values (for unshifted CASPT2 PES, see Fig. S7). (C) PES of the , LF ⁵E, and CT excited states, and one triplet state (³E) interacting with ⁵E. Only strong SOC between ³E and ⁵E in the vicinity of the PES minimum is considered; otherwise, PES are calculated with nonrelativistic CASPT2 approach. PES shifted to match experimental values (for unshifted PES, see Fig. S7). (D) SOC-driven strong CI mixing between ³E and ⁵E PES.

These theoretical findings directly correlate to the experimental MCD data. First, the state can be assigned as the sharp Fano interference. Second, the MCD pseudo-A term with different progressions within the formally-same transition can be attributed to strong SOC between the ³E state and the RCP component of the ⁵E (leading to a close-to-equal mixing of both states). This strong SOC leads to two RCP (negative) vibronic progressions, whose PES are shaped by strong SOC-driven configuration interaction (CI) in the vicinity of the ground-state geometry (see Fig. 5 C and D). Alternatively, the LCP (positive) component of the ⁵E is unperturbed by the ³E and only weakly interacts with the (Fig. 4B). Reasonable agreement between this model and experimental data is obtained from a fit of the pseudo-A term with one positive and two negative vibronic progressions (labeled “2,” “1,” and “3,” respectively, in Fig. 3B; see also Table S1): (i) The relative energy positions of the three vibronic progressions from the MCD Franck–Condon fits match well with the corresponding calculated CASPT2 vertical-transition energies for the three states responsible for the progressions; (ii) excited-state distortions relative to that of the ground state (ΔQ_Fe─Os), as estimated from the CASPT2 PES in Fig. 5C, reproduce the change in the Fe─O bond length in the excited state, relative to the ground state obtained from the MCD Franck–Condon analyses (Fig. 3B and Table S1); and (iii) the frequencies of the excited-state Fe─O stretch, derived from the curvatures of the CASPT2 PES, are in accord with the experimental values (Table S1).

PES of oxo π → Fe d_xz/yz CT state.

In analogy to the vibronically resolved NIR MCD features, the excited-state parameters obtained from Franck–Condon analysis of the CT progression (Fig. 1B, Inset), along with theoretical values derived from the corresponding CASPT2 PES (Fig. 5C), are provided in Table S1. The calculated oxo π → d_xz/yz (π^∗) CT transition has a lower excited-state frequency and a larger excited-state distortion than any of the three LF ⁵E progressions as observed experimentally, reflecting a weak Fe─O bond in this excited state.

Evolution of Fe^III-oxyl character along Fe─O coordinate.

The spectroscopically calibrated electronic-structure model with the two lowest excited states, ⁵E (d_xz/yz → d_z²) and CT (oxo π → d_xz/yz), can be extended to examine the evolution of the wavefunction along the Fe─O stretching coordinate that is relevant to reaching the TS in reactivity (see Reactivity). Using the multireference CASPT2 approach, the wavefunction character is inspected in terms of the dominant electronic configurations.

As presented in Fig. 5 A and B, elongation along the Fe─O coordinate to bond lengths close to the TS at approximately 1.75 Å (Reactivity) leads to a large change in the ground- and excited-state wavefunctions. The ground-state wavefunction develops Fe^III(S = 5/2)—oxyl(p_z,σ) character (at the Fe─O distance of 1.88 Å there is approximately 45% oxyl character in ; see Fig. 5 A and B and Fig. S8), whereas the ⁵E and oxo π CT states develop Fe^III(S = 5/2)—oxyl(p_x,π) and Fe^III(S = 3/2)—oxyl(p_x,π) character, respectively. For the first excited state, the oxyl character dominates starting from an Fe─O distance of 1.72 Å (Fig. 5 A and B and Fig. S8). Moreover, both CASPT2 excited-state PES cross the ground-state PES (at 1.9 and 2.0 Å, respectively), which indicates accessibility of the two π-oxyl channels in addition to the σ-oxyl channel for the S = 2 Fe^IV═O species. Thus, the π- as well as the σ-channels can play important roles in reactivity, and, importantly, all three states have dominant oxyl character at these crossing points, indicating their effectiveness in electrophilic attack.

Reactivity.

Model complex 1 has been shown to perform electrophilic H-atom abstraction from exogenous substrates, including 1,4-cyclohexadiene (CHD) and THF, as well as undergo self-decay via H-atom abstraction from a methyl group on the TMG₃tren ligand (15, 17). For reactivity with exogenous substrates, because of ligand sterics a σ-attack is the only possible orientation to access the oxo moiety, whereas self-decay necessarily proceeds via a π-attack because of the approximately 90° Fe─O─C angle presented by the closest target methyl group (Fig. 6, Center). This self-decay reactivity is ideal for probing the efficacy of the β-d_xz/yz π^∗-FMOs in an S = 2 Fe^IV═O species for comparison with its α-d_z² σ^∗-FMO for exogenous-substrate reactivity.

Fig. 6.

Two possible reaction channels for 1 (Center): σ-attack for exogenous substrate (THF) and π-attack for self-decay of closest endogenous methyl group. The next-closest methyl group (circled) is a second potential target for self-decay (Fig. S10). The σ-attack (A) and π-attack (B) energy profiles, in kcal/mol. Solvated energies (Polarized Continuum Model, solvent = CH₃CN) shown in the order ΔE^‡/ΔH^‡/ΔG^‡ for TS and ΔE°/ΔH°/ΔG° for products. Experimental ΔG^‡ values calculated from second-order rate constants k₂ using the Eyring equation. Transition-state lowest unoccupied MOs involved in H-atom transfer shown. Note that DFT-calculated bond strengths of target C─H bonds are 94 (THF) and 97 (-CH₃) kcal/mol.

The σ-channel H-atom abstraction from THF and π-channel self-decay reactivities were evaluated computationally, and the DFT-calculated reaction coordinate energy profiles are shown in Fig. 5. For the THF σ-attack pathway, the Gibbs energy barrier, ΔG^‡, is calculated to be 16.8 kcal/mol, in reasonable agreement with the experimental value of 18.9 kcal/mol (17). At the TS, an electron has been partially transferred (39%) from the C─H bond of THF into the α-d_z² FMO, a description agreeing with a previous study on the reaction of 1 with CHD (25) as well as the electronic structure obtained from a CASPT2 calculation on the TS structure (Fig. S9). Further along the reaction coordinate, an S = 5/2 Fe^III─OH product is formed, consistent with the Mössbauer spectroscopic results (17). Note that the σ-attack product of 1 with CHD is also an S = 5/2 Fe^III─OH (17, 25).

For the self-decay π-pathway, the DFT-calculated and experimental ΔG^‡s (both 19.4 kcal/mol) are in agreement. In addition, ΔH^‡ was determined from temperature-dependent kinetic studies (17) to be 16.7 kcal/mol (at 0 °C), also corresponding well with the DFT-calculated value of 19.2 kcal/mol. An internal reaction coordinate scan proceeding from the TS results in a product with an S = 3/2 Fe^III─OH ferromagnetically coupled to the ligand radical. However, this is a local minimum, and relaxation of the wavefunction of this product and reoptimization of the structure leads to the energetically favored S = 5/2 Fe^III─OH + ligand radical product comparable in endergonicity to that of the THF σ-pathway reaction (Fig. 6A). Thus, DFT evaluation of the π- and σ-attack pathways of 1 shows that the reaction barriers are close in energy, in concord with experimental kinetics data. The barrier ΔE^‡ of the π-pathway (self-decay) and σ-pathway for an exogenous substrate with comparable C─H bond strength (CH₃NH₂; see Fig. S10) can be decomposed into steric and electronic components (25), showing that deformation of TMG₃tren in the π-pathway has a much larger steric contribution (13.4 vs. 4.4 kcal/mol) to its ΔE^‡, resulting in similar electronic barriers between the π- and σ-pathways (9.8 vs. 9.5 kcal/mol, respectively) for similar C─H bond strengths. Finally, the product in both cases is an S = 5/2 Fe^III─OH with similar free energies of reaction. Because both pathways have similar reaction barriers and driving forces, both the α-d_z² σ^∗-FMO and the β-d_xz/yz π^∗-FMO are equally-viable pathways in 1 for H-atom abstraction reactivity, which is in contrast to a recent computational study (32).

Interestingly, there are two possible electronic descriptions of the TS of the π-pathway, raising implications for π-reactivity. The DFT calculations describe the TS as transferring 57% of a β-spin electron from the C─H bond into the Fe─O π^∗-FMO, giving Fe^III(S = 3/2) character (Scheme S2, Left), whereas in CASPT2 calculations the dominant (35–40%) configuration contributing to the TS involves an α(oxo π → d_z²) excitation, resulting in an Fe^III(S = 5/2) center and an oxyl radical ready to accept an α electron from the C─H bond (Fig. S9B and Scheme S2, Right).

Importantly, these two electronic structure descriptions involved in π reactivity directly correlate to the first two excited states observed in MCD spectroscopy and corresponding CASPT2 PESs (Fig. 5 A and C). The ⁵E LF excited state evolves into an Fe^III(S = 5/2) + π-oxyl configuration (at long Fe─O distances), whereas the first oxo π → d_xz/yz CT excited state corresponds to an Fe^III(S = 3/2) + π-oxyl configuration (Fig. 5 A and B), both highly activated to accept an electron from C─H into the oxyl radical. Thus, in addition to the σ-channel considered earlier (Fig. 6A, Center), MCD spectroscopy has defined two possible π-channels for the reactivity of S = 2 Fe^IV═O intermediates.

Discussion

The S = 2 Fe^IV═O biomimetic model complex 1 is capable of H-atom abstraction through a σ-attack by the Fe─O moiety on an exogenous substrate. Underpinning this reactivity is the spin polarization of the unoccupied α-d_z² MO, which lowers it in energy and makes it accessible as an FMO for electrophilic attack. For 1, however, the methyl groups of the ligand present a steric barrier hindering axial access to the Fe^IV═O unit, as discussed above and emphasized in previous studies on 1 (15, 25). In the present study, this σ-attack has been shown to proceed via an Fe^III(S = 5/2)–oxyl state, where elongation of the Fe─O bond transfers an α electron from the oxo p_z σ orbital to the Fe d_z², leaving a low-lying oxo p_z σ unoccupied orbital available to accept an α electron from the C─H bond of the substrate.

Additionally, 1 self-decays by attacking one of its own ligand methyl groups, and this takes place necessarily via a side-on π-attack because of the bent Fe─O─C_methyl angle (approximately 90°) imposed by the covalent linkage of the methyl substrate to the ligand. Spectroscopic data on 1 show a near featureless absorption profile, providing minimal insight, but become rich in information content upon going to VT MCD spectroscopy. From VT MCD data, two excited states have been identified, both capable of providing additional π-channels for reactivity in an S = 2 Fe^IV═O species (Scheme 1). In the case of the first ⁵E LF excited state where a d_xz/yz electron is excited into the d_z² orbital, elongation of the Fe─O bond along this ⁵E PES allows it to mix with the ground-state PES (because of the C_s distortion associated with the substrate interaction) while transferring an α electron from the oxo p_x π orbital into the d_xz orbital, creating an Fe^III(S = 5/2)─O^•- species with an oxo π-hole (Scheme 1, Right). Alternatively, the first oxo π → d_xz/yz (π^∗) CT transition results in an Fe^III(S = 3/2)─O^•- species that decreases in energy with elongation of the Fe─O bond to the TS and can similarly accept an electron into its oxo π-hole. These provide two viable channels for π-reactivity, supported by experimental spectroscopic and kinetic data.

Scheme 1.

(Left) MCD spectroscopy describes the electronic structure of the Fe^IV═O unit at its equilibrium geometry, and (Middle) computational extension to the TS Fe─O bond length leads to (Right) two π-Fe^III-oxyl and one σ-Fe^III-oxyl descriptions activated for electrophilic attack.

Similarly, computational evaluation of the self-decay π reaction coordinate of 1 revealed that the TS can be described by two possible electronic configurations: (i) DFT calculations show a β electron being partially transferred from the substrate to the π^∗-FMO to give Fe^III(S = 3/2)─O^•- character, whereas (ii) CASPT2 calculations show a dominant contribution from an Fe^III(S = 5/2)─O^•-(p_x,π) configuration. These two states are observed directly in the MCD data and their evolution into oxyl-dominant character is observed in the PES calculated by CASPT2 in Scheme 1. The importance of these two π-channels in the reactivity of S = 2 Fe^IV═O NHFe enzyme intermediates and the associated spin-state differences on Fe along the reaction coordinate are the subjects of current study.

The availability and viability of both σ- and π-pathways to NHFe enzyme Fe^IV═O intermediates, which both have similar electronic reaction barriers but different substrate orientation requirements for electrophilic reaction, can dictate both the specificity of their initial reaction and the path of their subsequent reactivity. In (4-hydroxyphenyl)pyruvate dioxygenase (HPPD) and (4-hydroxy)mandelate synthase (HmaS), the same substrate, depending on its orientation, undergoes electrophilic aromatic attack via the σ-pathway, or H-atom abstraction via the π-pathway (24). The presence of both pathways allows substrate orientation by the protein to control reactivity; in the halogenase SyrB2, either Cl^• transfer (native) or OH^• rebound (nonnative) reactivity can take place after the initial rate-determining H-atom abstraction step (33), depending on the carbon-chain length of the substrate and thus the orientation of the C─H bond relative to the Fe═O bond. This also implies that the enzyme can modulate its σ vs. π reactivity through the active site ligand environment; for example, differences in axial ligation could modulate the α-d_z² σ^∗-FMO for electrophilic attack. Thus, it is now important to apply this combined spectroscopic and computational approach to Fe^IV═O enzyme intermediates to evaluate their selectivity in reaction coordinates.

Both the abs and MCD spectra exhibit a sharp feature at 11,820 cm^-1; in abs it is a dip, whereas in MCD it is a peak. Fano analysis of the spectroscopic data was correlated to advanced quantum-chemical methods to reveal the shared origin of these features, which is a weak spin-orbit interaction between the broad spin-allowed ⁵E(xz,yz) and the nearby spin-forbidden state (Scheme 2A). In addition, the Franck–Condon analyses correlated with the same level of calculations demonstrated that the MCD pseudo-A term with variable vibronic spacing is a result of the in-state spin-orbit–perturbed ⁵E(xz,yz) state having strong SOC with another nearby spin-forbidden ³E state; this strong CI distorts the interacting PES (Scheme 2B). Together, these give rise to one positive and two negative vibronic progressions and one sharp positive peak in the NIR MCD spectrum. It is essential to note that these important effects of excited-state CI (a dip and a distortion of the PES) could only be identified and quantified through their differential effects on the MCD relative to the abs spectra because of their different selection rules, which result in different coupling interactions.

Scheme 2.

(A) Weak CI between one component of the ⁵E state with the results in a sharp Fano antiresonance feature, whereas (B) strong CI mixing of the ⁵E with the ³E (represented by dotted lines) results in one narrow and one broad PES (solid lines) with different vibrational-level spacings.

These experimental observations and theoretical findings provide fundamental insight into the electronic structure of S = 2 Fe^IV═O species, the interactions among its excited states, and the evolution of these states along the Fe─O reaction coordinate for catalysis.

Materials and Methods

Samples of 1 were prepared as described previously (15, 17), but in butyronitrile. UV-visible abs data were recorded on an HP8453A and a Cary 500 spectrophotometer. MCD was performed on Jasco J-730 and J-810 spectropolarimeters equipped with Oxford Instruments SM-4000 superconducting magnets. DFT calculations were performed with Gaussian 03/09 and multiconfigurational/multireference calculations with MOLCAS 7.4.