Nuclear Resonance Vibrational Spectroscopic (NRVS) Definition of the Fe(IV)₂ Intermediate Q in Methane Monooxygenase and its Reactivity

Abstract

Methanotrophic bacteria utilize the non-heme diiron enzyme soluble methane monooxygenase (sMMO) to convert methane to methanol in the first step of their metabolic cycle under copper-limiting conditions. The structure of the sMMO Fe(IV)₂ intermediate Q responsible for activating the inert C-H bond of methane (BDE = 104 kcal/mol) remains controversial, with recent studies suggesting both “open” and “closed” core geometries for its active site. In this study, we employ nuclear resonance vibrational spectroscopy (NRVS) to probe the geometric and electronic structure of intermediate Q at cryogenic temperatures. These data demonstrate that Q decays rapidly during the NRVS experiment. Combining data from several years of measurements, we derive the NRVS vibrational features of intermediate Q as well as its cryoreduced decay product. A library of 90 open and closed core models of intermediate Q is generated using density functional theory (DFT) to analyze the NRVS data of Q and its cryoreduced product, as well as prior spectroscopic data on Q. Our analysis reveals that a subset of closed core models reproduce these newly acquired NRVS data as well as prior data. The reaction coordinate with methane is also evaluated using both closed and open core models of Q. These studies show that the potent reactivity of Q towards methane resides in the “spectator oxo” of its Fe(IV)₂O₂ core, in contrast to non-heme mononuclear Fe(IV)=O enzyme intermediates that H-atom abstract from weaker C-H bonds.

Graphical Abstract

1. Introduction

The superfamily of binuclear non-heme (NH2Fe) enzymes¹ plays major roles in all realms of life including metabolism,^2-3 biotechnology,^4-5 drug design,^6-7 and bioremediation.⁸ They share in common an active site composed of histidine and carboxylate residues that house the binuclear iron core. The reductive activation of dioxygen by this cofactor initiates a wide range of chemistries with substrates including N-oxygenation,⁹ electrophilic aromatic substitution (EAS), ¹⁰ and hydroxylation.^11-12 Spectroscopic studies of O₂-activated intermediates in several NH2Fe enzymes have led to an improved understanding of their reaction mechanisms.^13-14

Under low [Cu] conditions, methanotrophic bacteria utilize the NH2Fe enzyme soluble methane monooxygenase (sMMO) to convert methane to methanol in the first step of its metabolic cycle.^15,16 The quaternary structure of this enzyme complex consists of three component subunits: a hydroxylase (αβγ)₂ (MMOH), which holds the NH2Fe cofactor; an electron transfer reductase (MMOR) that supplies two electrons to the NH2Fe cofactor during catalysis; and a regulatory component (MMOB) that binds to MMOH, greatly enhancing its product yield, O₂ reactivity, and turnover frequency.^17-20 To produce methanol, this enzyme complex must activate the inert C-H bond of methane (BDE = 104 kcal mol⁻¹) at ambient conditions, a chemical feat otherwise unprecedented in iron enzymology. The catalytic cycle of sMMO has been intensely studied under single-turnover conditions, leading to the identification of several intermediates including the methane-reactive intermediate Q.^19-23 Mössbauer spectroscopy has defined intermediate Q as having a high-spin, antiferromagnetically-coupled Fe(IV)₂ electronic structure.²⁴ This intermediate stands alone in the NH2Fe superfamily, for which most other substrate-reactive intermediates to date have been characterized as biferric (hydro)peroxides¹³ (and in some cases ferric superoxide).^25,26

The structure of intermediate Q was originally assigned in the late 1990s as a closed “diamond” Fe₂(μ-O)₂ core from extended X-ray absorption fine structure (EXAFS) data²⁷ indicating a short Fe-Fe distance of 2.46 Å. This assignment was supported by a more recent (2015) continuous flow resonance Raman (rR) study,²⁸ which revealed a 690 cm⁻¹ vibration showing ¹⁸O₂ and ¹⁶O¹⁸O isotopic shifts consistent with the vibrational spectra of diamond core model complexes^29,30 and O₂ loading into the Fe₂O₂ core. An additional peak at 556 cm⁻¹ observed upon the decay of Q in this study also showed an ¹⁸O₂ (but no unique ¹⁶O¹⁸O) shift, and was assigned to an Fe(III)₂ mono-μ-O species formed after methanol production (with loss of one labeled oxygen to methanol). Kinetic studies of the reaction of intermediate Q with methane reported a low activation barrier³¹ (ΔG^‡ = 10 kcal mol⁻¹ at 4 °C) and large deuterium kinetic isotope effect³² (KIE = 50), indicating an H-atom abstraction from the strong C-H bond of CH₄. These EXAFS and reaction kinetics studies prompted a number of theoretical investigations into the reactivity of methane hydroxylation by closed core Q models.^33-39 Various closed core models have been proposed, including a Fe₂O₂ diamond core with one³³ or two μ-1,3-carboxylate bridges,³¹ and with or without an exogenous water ligand.³⁰ Irrespective of the proposed model, all candidate structures optimized with a high-spin electronic structure resulted in Fe-Fe distances of ≥ 2.65 Å, far longer than measured by EXAFS.

In 2010, a high-spin “open” core O=Fe(IV)─O─Fe(III)─OH model complex was reported that exhibited a second order rate constant for dihydroanthracene oxidation 10⁶ higher than its low-spin diamond core precursor.^40,41 This observation fueled speculation into the possibility of an open core assignment for Q. Since this report, two studies using high-energy resolution fluorescence detected (HERFD) X-ray absorption spectroscopy (XAS) on rapid freeze-quenched (RFQ) samples of Q have advocated for the possibility of an open core assignment. First, an analysis of the high intensity of the HERFD-XAS pre-edge region of the Q spectrum showed that the pre-edge was best reproduced by an open core model complex possessing a terminal oxo; this high pre-edge intensity of Q was alternatively not reproduced by a Fe₂O₂ diamond core model complex.⁴² Second, remeasurement of the EXAFS region using HERFD-XAS did not reproduce the Fe-Fe 2.46 Å scatterer, which was determined to be a background metallic iron contribution.⁴³ The new RFQ-Q HERFD-EXAFS data were fit with an Fe-Fe distance of 3.30-3.34 Å, longer than the 3.27 Å distance of the ~ 50-60% Fe(II)₂ species (MMOH_red) also present in the RFQ-Q sample. This finding was interpreted as Q possessing a longer Fe-Fe distance (3.4 Å) than MMOH_red, which required an open core structural assignment. Though these studies have cast doubt on the closed core assignment of Q, no general consensus has been reached on its geometric structure and therefore the relevance of past calculations on closed core models.

Defining the geometric structure of Q thus requires further spectroscopic insight. The continuous-flow rR feature at 690 cm⁻¹ informs on the mechanical coupling of O₂ incorporated atoms, but this feature alone provides limited structural insight. However, the ⁵⁷Fe preparations of sMMO used for RFQ-Mössbauer spectroscopy also enable the use of nuclear resonance vibrational spectroscopy (NRVS). This technique probes the vibrational sidebands in the ⁵⁷Fe nuclear Mössbauer transition and reports on all vibrations with significant ⁵⁷Fe displacement.^44-46 NRVS studies by our group have successfully characterized high-valent oxygen intermediates in binuclear⁴⁷ and mononuclear enzymes⁴⁸ and their models,^49,50 as well as iron-containing zeolites.⁵¹ In the present study we utilize NRVS to study RFQ samples of Q. This hot intermediate is found to rapidly decay in the beam. Scanning multiple samples of Q over several years of experiments has enabled the separation of the Q NRVS spectrum from its decay. Further, the decay of Q in the X-ray beam is found to involve its cryoreduction; thus the NRVS spectrum of Q decay provides a spectral perturbation to assist in the assignment of Q. DFT calculations have been used to generate 90 structural models to correlate to the Q NRVS data as well as to prior spectroscopic data. This study yields deeper insight into the structure of Q as well as its reactivity relative to mononuclear non-heme iron enzymes.

2. Results

2.1. NR VS of Intermediate Q

Intermediate Q is formed by reaction O₂ with biferrous MMOH in complex with equimolar (vs MMOH active sites) MMOB. Thus, all samples analyzed here contain a complex of MMOH and MMOB. NRVS data collected on RFQ-Q samples measured at BL09XU exhibit time-dependent changes in the beam at cryogenic temperatures. The NRVS spectrum of a high-purity preparation of Q (50% of total iron, 37% biferrous (MMOH_red), 13% biferric (MMOH_ox); see Figure S1 for quantitation) displays an intense feature at 275 cm⁻¹ in the first 10 scans (1 hour/scan) which disappears after 20 total scans (Figure S2, bottom to top). Concomitant with the decay of the 275 cm⁻¹ feature is the growth of intensity between 200-250 cm⁻¹ during 40 hours of measurement. Other RFQ samples of Q containing a lower quantity of Q (40%) and higher MMOH_red (50%) also exhibited weaker intensity at 275 cm⁻¹ that goes away with increasing scans. Therefore, in order to clearly distinguish the NRVS features of Q (early time data) and its decay product (late time points), seven samples of Q were measured up to 80 hours (ranging from 20 to 80 hours per sample). Binning sixty scans from these samples into ten-hour windows demonstrates biphasic, time-dependent NRVS changes (Figure S3, bottom to top): in the first phase, the 275 cm⁻¹ feature and a weaker (but reproducible above noise) feature at 605 cm⁻¹ decay and are no longer present after 20 hours of measurement (bottom to middle, black arrows), while features between 220-240 and at 580 cm⁻¹ grow in (middle, red arrows). In the second decay phase, the 220-240 and 580 features continue to increase in intensity and no longer change after 30 hours of measurement (top, black arrows). The decay of a 375 cm⁻¹ feature is also observed in these phases (top, black line), resulting in a local maximum in the late time point NRVS data at 365 cm⁻¹. NRVS data combined from several samples into sets of 5-hour intervals (Figure S4) corroborate the decay of the 605 and 275 cm⁻¹ features, and the growth of the 220, 240, and 580 cm⁻¹ features within the first 15 hours of measurement (first phase). Final early and late time scan Q NRVS data were generated by compiling over one-hundred scans between 1-15 and 30-80 hours of measurement, and these are shown in Figure 1A top as red and black, respectively. Additionally shown in Figure 1A top (orange) is the NRVS spectrum of a MMOH_red sample which was scaled down to reflect the 50% Fe(II)₂ present in the “Q” sample. Subtraction of 50% MMOH_red and renormalization generates the spectra at the bottom of Figure 1A for the Q (red) and Q decay (black) NRVS data. The overlay of these spectra confirms that the time-dependent changes observed between early and late Q NRVS data (red down arrows and up black arrows, respectively) are outside of noise, as are additional features at 420, 330, and 300 cm⁻¹ which do not change between the early and late Q NRVS data (lighter coincident red and black lines). Thus, NRVS data of Q reveal strong time-dependent changes in the beam at both low (200-400 cm⁻¹) and high (560-610 cm⁻¹) energies. The early time data define the NRVS features of intermediate Q, and the late time data provide the NRVS spectrum of its decayed features in the X-ray beam.

Figure 1.

NRVS spectra of Q samples indicating time-dependent changes in the beam and isotopic perturbations. (A) Top: NRVS spectra (including error) of early Q (hours 1-15; top red), and late decay scans (30+ hours; top black), and MMOH_red (multiplied by 0.5; top orange). Late decay scans include ¹⁶O₂ and ¹⁸O₂ data in order to maximize S/N. At the top of the figure, thick red lines indicate major features of early Q data. Brown lines indicate features which grow in over two phases. Black thick lines indicate major features of decayed Q data. Thin red and black lines indicate features shared by both early and decayed Q data. Bottom: MMOH_red-subtracted and renormalized spectra (Q red and decay black). In the bottom plot, based on poor S/N the intermediate Q and decay spectra past 620 cm⁻¹ are given as Early – Late scan data (no feature above noise is observed). (B) NRVS spectra of ¹⁶O₂ (71 scans) and ¹⁸O₂ (110 scans, with data accumulated at ~ 1/2 count rate as ¹⁶O₂ decay data) labeled Q after 30+ hours of measurement.

In addition, intermediate Q samples were also prepared with ¹⁸O₂. While the ¹⁶O₂ Q samples could be generated using chlorite dismutase and were thus concentrated enough to display significant NRVS intensities in the early scans, the ¹⁸O₂ samples had to be prepared with ¹⁸O₂-saturated buffer and thus were more dilute (1.00 mM total iron vs. 1.80 mM). The ¹⁸O₂ generated Q samples required more accumulation time for good S/N and isotope perturbations could only be obtained for the late time (30-80 hour) decay data. Two isotope shifts are observed in the decayed Q NRVS spectrum: a 12-15 cm⁻¹ downshift of the 420 cm⁻¹ feature, and a small downshift of the 580 cm⁻¹ feature (Figure 1B; see Figure S5 for raw data counts in this region). While limited by the S/N of ~ 2, analysis of the 580 cm⁻¹ feature (Figure S6) reveals an isotope shift of 4–7 cm⁻¹.

In summary, Q shows NRVS peaks at 605, 420, 375, 330, 300, and 275 cm⁻¹. The decay of Q results in the loss of the 605, 375, and 275 cm⁻¹ features and produces new peaks at 580, 365, 240, and 220 cm⁻¹. The features at 580 and 420 cm⁻¹ in the decayed Q spectrum show a ¹⁶O/¹⁸O effect of 4-7 cm⁻¹ and 12-15 cm⁻¹, respectively.

2.2. Mössbauer of Q NRVS Samples and Parallel Data On Cryoreduced Q

It was hypothesized that the decay of Q in the X-ray beam during NRVS data collection was due to its photoreduction. Mössbauer spectra were obtained on RFQ Q samples before and after NRVS experiments to quantify the amount of Q decay and identify any new species formed due to photoreduction. To track the decay of Q during NRVS measurements, parallel experiments were performed on frozen-solution Q samples at BL19LXU at SPring-8. Samples measured at BL19LXU were prepared in 11 x 4 x 1.2 mm rectangular holders; illumination of these samples results in an estimated 25% total volume exposure to the NRVS X-ray beam. The NRVS spectrum of a BL19LXU Q sample after 14 hours overlays well with the NRVS spectrum of BL09XU Q decay (Figure S7; due to the 3 times higher flux at BL19LXU). Importantly, The Mössbauer spectrum of the BL19LXU sample (Figure 2A red, speciation prior to NRVS experiment given in Figure S8) after NRVS measurement at BL19 for 14 hours shows two changes: 1) 20% reduction of the original Q in the full sample volume (Figure 2A, black, arrows pointing up at 0.50 mm/s and −0.14 mm/s where the second component of the Q doublet is located); and 2) concomitant growth of a new species with δ ≈ 0.6 and ΔE_Q ≈ 1.60 mm/s (Figure 2A, arrows pointing down at 1.45 mm/s and −0.14 mm/s; note the downward arrow at −0.14 mm/s reflects a compensating increase in signal for the loss of Q at this velocity) indicative of the formation of a ferric species and possibly photoreduction of the NRVS sample.

Figure 2.

Mössbauer spectra collected at 6 K of intermediate Q exposed to the NRVS X-ray beam and to cryoreduction. (A) Mössbauer spectra before (red) and after (black) NRVS measurement of Q. Arrows indicate major changes pre- and post-NRVS measurement. (B) Mössbauer spectra of a Q sample irradiated to 6 Mrad (top to middle). Arrow in the middle spectrum indicates a new feature present after cryoreduction. Bottom spectrum post-cryoreduction in an applied 73 mT parallel field showing the appearance of paramagnetic features (arrows).

In order to evaluate whether the new ferric species in Figure 2A (downward arrows) is a product of Q photoreduction, samples of Q were independently cryoreduced using 6 Mrad of γ-ray emission from a radioactive ⁶⁰Co source. Similar experiments were reported by Valentine et al. on MMOH Q isolated from Methylococcus capsulatus (Mc Bath);⁵² however, Mössbauer data have not yet been reported for cryoreduced Q from Methylosinus trichosporium OB3b (Mt OB3b) MMOH. Mössbauer spectra of Q before (Figure 2B top) and after cryoreduction revealed formation of a ferric species (Figure 2B, middle, arrow) with a feature at 1.45 mm/s, similar to the signal observed in the post-NRVS Mössbauer spectrum (Figure 2A, black), confirming cryoreduction of the NRVS sample. Application of a 73 mT parallel magnetic field resulted in no change of the 1.45 mm/s signal indicating it is a coupled biferric species, and revealed additional features at δ = −3.7, −1.7, and 4.8 mm/s (Figure 2B, bottom, arrows), indicating the presence of a paramagnetic Fe(III)Fe(IV) species, similar to intermediate X^52,53 (here labeled Qx) in ribonucleotide reductase. Subtracting the Fe(II)₂, an initial Fe(III)₂ component present in Q before cryoreduction, and the Fe(III)Fe(IV) species from the bottom spectrum in Figure 2B yields the diamagnetic biferric species (Q3; Figure S9, bottom), consisting of two inequivalent high-spin Fe(III) sites (δ₁ = 0.50 mm/s, ΔE_Q1 = 1.99 mm/s; δ₂ = 0.63, mm/s ΔE_Q2 = 1.33 mm/s) of equal intensity. Thus, the loss of the Q Mössbauer signal at 0.50 mm/s (Figure 2A) and the formation of a new ferric signal in the post-NRVS Mössbauer spectrum at −0.14 and 1.45 mm/s can be attributed to photoreduction of Q and formation of a high-spin antiferromagnetically coupled biferric decay product (through the mixed valent Qx). From simulating the BL19 post-NRVS Mössbauer spectrum (Figure S10), all of Q in the BL19 NRVS sample has decayed, leaving a mixture of Qx and Q3 (roughly 40%:60%). While the ratio of these two reduced species would vary with beam exposure, Qx and Q3 are calculated to have (vide infra) and do exhibit similar NRVS features (scan dependence in Figure S3). Thus, this variable is not an issue in the analysis of the NRVS decay spectra in Section 3.4.

3. Analysis

3.1. Structural Models for Intermediate Q

To determine possible structures for Q which best reproduce the NRVS and prior spectroscopic data, a library of 90 DFT models (Table S1) that contain two antiferromagnetically-coupled high-spin Fe(IV) centers was constructed by modifying the X-ray crystal structure of MMOB-associated MMOH from Methylococcus capsulatus (Bath. PDB 4GAM).⁵⁴ Though these crystals appear to have cryoreduced in the X-ray beam, a crystal structure of oxidized MMOH (without MMOB bound, PDB 1MHY) possesses three oxygen ligands, two O(H)_x bridges and one terminal O(H)_y moiety.⁵⁵ Therefore, a model with two (μ-O) bridges and a terminal H₂O (bis(μO)-H₂O-1) was chosen to preserve charge neutrality of the Q Fe(IV)₂ active site and served as a starting structural scaffold for optimization of Q models. This model choice has been justified by the recent report of the higher resolution crystal structure of fully oxidized MMOB-associated MMOH from Methylosinus trichosporium OB3b (see Section 3.4),⁵⁶ and has been further calibrated with DFT (Figure S11) for determination of the optimal truncation scheme.

The full library of Q models possesses either “closed core” structures with two O(H)_x bridges (denoted bis(μO)-L-# in Table S1, where L is a terminal ligand to one Fe(IV)) or “open core” structures with one oxo bridge (denoted OC## in Table S1), with further variations in the protonation level of the carboxylate ligands and oxygenic terminal ligands including oxide, hydroxide, and water, and hydrogen bonding interactions. Closed core Q models with a terminal H₂O ligand (bis(μO)-H²O subclass, 16 models) were initially constructed through ligand rearrangements of bis(μO)-H₂O-1. These models were further modified by deprotonating or removing the terminal H₂O ligand, resulting in bis(μO)-OH and bis(μO) subclasses (17 and 6 models, respectively) containing a variety of Glu coordination, conformation, and protonation states, as well as H-bonding interactions with the bis(μO) core or a non-coordinating H₂O molecule. Two other subclasses, (μOμOH) (8 models) and bis(μO)=O (4 models), are included in the SI and will be briefly presented. Representative bis(μO), bis(μO)-OH, and bis(μO)-H₂O structures are given at the bottom of Figure 3.

Figure 3.

Structural model subtypes for Q. Includes structural representatives for the OC -OH/-OH (OC10), OC -OH/=O (OC26) subclasses at the top, and bis(μO) (bis(μO-1)), bis(μO)-OH (bis(μO)-OH-16), and bis(μO)-H₂O (bis(μO)-H₂O-6) subclasses at the bottom.

Open core models for Q were prepared by translocating a proton from the terminal water in bis(μO)-H₂O-1 to a μ-O bridge and reoptimizing on the Fe(IV)₂ level. Opening of the core in this manner yielded a 5C square pyramidal and 6C site with a terminal OH⁻ trans-axial to a His. Models generated in this way with further variations in the conformations and H-bonds of the 1st-sphere Glu residues form the OH-Fe-O-Fe-OH (OC -OH/-OH) subclass. By either protonating or deprotonating the terminal OH⁻ ligand, 37 total open-core models were generated, including the OH-Fe-O-Fe=O (OC -OH/=O) subclass and three other open core subclasses, O=Fe-O-Fe=O (OC=O/=O), O=Fe-O-Fe-H₂O (OC -H₂O/=O), and OH-Fe-O-Fe-H₂O (OC -OH/-H₂O), to be presented in less detail (Table S1). Representative OC -OH/-OH and OC -OH/=O structures are given at the top of Figure 3.

3.2. Vibrational Analysis of Closed and Open Core Models of Q

The 90 possible models for intermediate Q generated in Section 3.1 were initially evaluated by computing their vibrational spectra using DFT methods (with the BP86 functional plus 10% HF) and comparing these results to the prior rR data on Q. This rR study²⁸ revealed that Q exhibits a vibration at 690 cm⁻¹, with a ¹⁸O₂ isotopic downshift of 36 cm⁻¹ and a ¹⁶O¹⁸O mixed isotope downshift of 12-15 cm⁻¹ (with no observable ²H perturbation, for an experimental resolution of ~ 3 cm⁻¹). Furthermore, rR data of the methane-reacted product complex, T show a 556 cm⁻¹ vibration with a ¹⁸O₂ downshift to 533 cm⁻¹, and two ¹⁶O¹⁸O mixed isotope signals at 556 and 533 cm⁻¹ reflecting the presence a single labeled oxygen atom in T. These rR data on T define it as having a μ-O bridge derived from O₂. Therefore, these data together show that Q contains an O₂ derived core, with one of these oxygens loaded in a μ-O bridge. To screen candidates for Q from the library of 90 models, vibrational frequencies for ¹⁶O₂, ¹⁸O₂, ¹⁶O¹⁸O, and ²H isotopomers for each of the models were obtained from the DFT calculations and correlated to the experimental rR vibrational data (¹⁶O₂ spectra for all models given in Table S1, along with tabulated ¹⁶O¹⁸O, ¹⁸O₂ and ²H perturbations).

Closed core Q models of the bis(μO), bis(μO)-OH, and bis(μO)-H₂O subclasses exhibit three modes above 600 cm⁻¹: a symmetric (A_g) breathing mode at 700 cm⁻¹, as well as two non-symmetric (B_3u and B_2u) modes (using effective D_2h symmetry for a Fe₂O₂ core) between 600 and 650 cm⁻¹ (ν₁, ν₂, ν₃ in Figure 4 top, respectively). Incorporation of ¹⁸O₂ and ¹⁶O¹⁸O into the (μ-O)₂ core results in downshifts of the breathing mode by 30-40 cm⁻¹ and 15-20 cm⁻¹, respectively. The absence of proton involvement in the breathing mode results in no predicted ²H effect. The breathing mode of all bis(μO) models, as well as 12 out of 16 bis(μO)-H₂O models, reproduces the Q rR 690 cm⁻¹ feature well (highlighted in red in Figure 4 top), and is similar to rR data collected on bis(μO) model complexes.^29,30 However, the symmetry of the Fe₂O₂ core mediates the degree of coupling between the (μ-O)₂ units. For example, in bis(μO)-H₂O-16, hydrogen bonding (heavy atom distance = 2.5 Å) between E144 and one (μ-O) introduces a C_2ν distortion into the core (Fe₁-O_a = 1.83, Fe₂-O_a = 1.95 Å, Fe₁-O_b = 1.74, Fe₂-O_b = 1.85). As a result, bis(μO)-H₂O-16 possesses two uncoupled ν(Fe₁-O_a) and ν(Fe₂-O_b) stretches above 600 cm⁻¹ which localize onto the short Fe-O bonds, and prevents the appropriate mixed isotope shift in the ¹⁶O¹⁸O isotopomer (similar distortions are found in bis(μO)-H₂O-5, 7, and 13; see Table S1). In the bis(μO)-OH subclass, the strong ligand field of the terminal OH⁻ produces a large C_2h distortion and uncoupling of the core vibrations (models bis(μO)-OH 1-5, 7, 9, 17-18). In contrast to the bis(μO)-H₂O subclass, hydrogen bonding to the (μ-O)₂ unit in the bis(μO)-OH structures decreases the distortion. This increases the (μ-O)₂ mechanical coupling and results in a breathing mode vibration near 690 cm⁻¹ with the appropriate mixed isotope shift. The influence of the OH⁻ ligand is such that, of the remaining 9 bis(μO)-OH models (bis(μO)-OH 6, 8, 10-16) possessing a coupled A_g breathing mode, 7 of them contain a hydrogen bond to the (μ-O)₂ unit (models bis(μO)-OH 6, 10-14, 16). For all three bis(μO), bis(μO)-OH, and bis(μO)-H₂O subclasses, the 690 cm⁻¹ vibration is well reproduced by a breathing mode in models without strong C_2ν or C_2h distortions (i.e., in models for which all Fe-O bond lengths in the Fe₂O₂ core are similar)

Figure 4.

Normal modes for the Q closed core bis(μO), bis(μO)-OH, and bis(μO)-H₂O subclasses (top), and Q open core OC -OH/-OH (middle) and OC -OH/=O subclasses (bottom). Closed core modes defined in D_2h symmetry. The calculated vibration in each of these models that corresponds to the reported rR 690 cm⁻¹ vibration is highlighted in red. Arrows denote in-plane motion, and “x” and “o” characters denote out-of-plane motions.

The two other closed core Q subclasses, bis(μO)=O and (μOμOH), do not possess a feature above 600 cm⁻¹ with an appropriate mixed isotope downshift. In bis(μO)=O models, the terminal oxo facilitates an opening of the core to a O=Fe-O-Fe=O structure; therefore, only one antisymmetric μ-O stretch (ν_as(Fe-O-Fe)) is observed between 600-740cm⁻¹. This is true as well for (μOμOH) models which lack a coupling of the μ-OH stretches to the ν_as(Fe-O-Fe) mode. Thus, these models do not predict the observed average mixed isotope downshift in the ¹⁶O¹⁸O isotopomer and are excluded from further consideration.

Though previous literature had identified closed core models as structural candidates for Q based on the 690 cm⁻¹ vibration and ¹⁶O¹⁸O mixed isotope shift, an open question remains as to how an open core model as suggested by recent EXAFS data could reproduce these rR data. This question is now addressed. In Q open core models, the Fe-OH stretch (ν(Fe-OH)) and ν_as(Fe-O-Fe) appear between 600 and 750 cm⁻¹. If these vibrations occur close in energy, they mix to form symmetric (ν(Fe-OH) + ν_as(Fe-O-Fe)) and antisymmetric (ν(Fe-OH) − ν_as(Fe-O-Fe)) combinations (Figure 4, ν₁ and ν₂ middle row, and ν₂ and ν₃ bottom row).

For OC -OH/=O models possessing an OH-Fe₂-(μ-O)-Fe₁=O structure, The Fe₁=O bond weakens the Fe₁-(μ-O) bond resulting in a shorter Fe₂-(μ-O) bond, localizing ν_as(Fe-O-Fe) on the Fe₂ side. This facilitates ν(Fe-OH) + ν_as(Fe-O-Fe) coupling, resulting in a symmetric mode ν₂ which would agree with the 690 cm⁻¹ experimental data (Figure 4, bottom row, red box), as anticipated in a recent computational study of Q.⁵⁷ Isotope labeling with ¹⁸O₂ must be incorporated into the μ-O bridge to correlate to the decay product T, as well as either into the -OH or =O terminal ligands. Incorporation of ¹⁸O into the =O position does not lead to an observed mixed isotope signal, since the oxo vibration is too high in energy (800 cm⁻¹) to influence ν₂ (Figure S12). Therefore, the ¹⁸O₂ would have to be incorporated into the -OH and μ-O ligands. OC -OH/=O ¹⁸O₂ and ¹⁶O¹⁸O isotopomers display a downshift of ν₂ by 20-40 and 10-20 cm⁻¹ (respectively), which reproduce the Q rR experimental data. Importantly, the mixed isotope signal strongly depends on the similarity between Fe₂-OH and Fe₂-(μ-O) bond lengths. Fe₂-OH and Fe₂-(μ-O) bond lengths differing by more than 0.05 Å results in the uncoupling of the ν(Fe-OH) and ν_as(Fe-O-Fe) vibrations and the absence of the appropriate mixed isotope shift (OC2, OC22, OC24, OC25). This ν(Fe-OH) + ν_as(Fe-O-Fe) mode coupling is also facilitated by hydrogen-bond donation to the Fe₁=O oxo, which weakens the Fe₁=O bond and strengthens the Fe₁-(μ-O) bond, resulting in elongation of the Fe₂-(μ-O) bond. Of the remaining OC -OH/=O models, OC17 and OC26 exhibit ν₂ matching the experimental 690 cm⁻¹ vibration and its isotope shifts, facilitated by hydrogen-bond donation (from a glutamic acid or OH⁻) to the terminal oxo. Note, however, that two of these models (OC1 and OC3) exhibit ν₂ ²H downshifts of 8 cm⁻¹ that is too large based on the rR data; an explanation for this effect is provided below.

In OC -OH/-OH models, three vibrations appear between 600 and 750 cm⁻¹: two ν(Fe-OH) (denoted ν(Fe₁-OH) and ν(Fe₂-OH)) and one ν_as(Fe-O-Fe) vibration which may mix to give symmetric and antisymmetric combinations (Figure 4, middle). In these OH-Fe₁-(μ-O)-Fe₂-OH models, the Fe₁-OH vector is oriented out-of-plane with respect to the Fe-O-Fe plane and ν(Fe₁-OH) couples weakly with the ν(Fe₂-OH) and ν_as(Fe-O-Fe) vibrations. Thus two coupled modes are observed near 700 cm⁻¹: ν(Fe₂-OH) + ν_as(Fe-O-Fe) and ν(Fe₂-OH) − ν_as(Fe-O-Fe) (Figure 4, middle row, ν₁ and ν₂), as well as one uncoupled ν(Fe₁-OH) mode (ν₃). Hence, to reproduce the experimental isotope shifts the two atoms of ¹⁸O₂ must incorporate into the in-plane hydroxide (Fe₂-OH) and the μ-O ligands. For OC -OH/-OH models mode ν₁ exhibits ¹⁸O₂ and ¹⁶O¹⁸O isotopic shifts of 20-40 and 10-20 cm⁻¹ (respectively), agreeing well with the experimental Q rR data. Due to the similarity of the Fe₂-OH and Fe₂-(μ-O) bond lengths (Δ = 0.02-0.03 Å), 13 out of 14 OC -OH/-OH models display ν₁ with a mixed isotope downshift of 10-20 cm⁻¹. However, 9 of these 13 models exhibit a large ν₁ downshift upon deuterium labeling (∣Δ²H∣ ≥ 5 cm⁻¹ for OC9, OC11, OC13, OC18, OC19, OC20, OC23, OC33, OC34), as observed in the OC -OH/=O OC1 and OC3 models (vide supra). This is due to significant mixing of the OH torsional twist into the in-plane ν(Fe₂-OH) stretch; this torsion mode downshifts by more than 100 cm⁻¹ upon ²H labeling and hence downshifts ν(Fe₂-OH) (Figure S13). The remaining models (OC10, OC15, OC16, OC32) possess a ν₁ with a mixed isotope signal that reproduces the 690 cm⁻¹ rR vibration well with no significant ²H effect.

The remaining OC =O/=O, -H₂O/=O and -OH/-H2O subclasses do not exhibit a 690 cm⁻¹ vibration with an appropriate mixed isotope downshift for the ¹⁶O¹⁸O isotopomer and are excluded from further consideration. Both OC =O/=O and OC -H₂O/=O subclasses display a single ν_as(Fe-O-Fe) vibration between 600-750 cm⁻¹ with no coupling partner. In the OC -OH/-H₂O subclass, the Fe-(μ-O) bond on the OH- side is too long to permit ν(Fe-OH) and ν_as(Fe-O-Fe) coupling.

Thus, both closed core and open core Q models predict vibrations near 690 cm⁻¹ with a ¹⁶O¹⁸O labeled mixed isotope shift that would correspond to the experimental rR data. The bis(μO), bis(μO)-OH, and bis(μO)-H₂O models display the canonical A_g breathing mode, which upon ¹⁶O¹⁸O labeling into the (μ-O)₂ core exhibits an appropriate mixed isotope shift. For OC -OH/=O models, the ν_as(Fe-O-Fe) mode localizes on the Fe₂-OH side and mixes with the ν(Fe₂-OH) vibration to produce a symmetric ν(Fe₂-OH) + ν_as(Fe-O-Fe) mode. Labeling of the OH- and μ-O ligands with ¹⁶O¹⁸O and ¹⁸O₂ would result in its downshift of 10-20 and 20-40 cm⁻¹ respectively, which would agree well with experiment. The OC -OH/-OH subclass also exhibits a ν(Fe₂-OH) + ν_as(Fe-O-Fe) mode which agrees well with experiment, provided minimal rotation of the Fe₂-OH unit (that would result in a large deuterium isotope shift) and selective ¹⁸O₂ labeling of the in-plane Fe₂-OH and μ-O ligands. All calculated isotope perturbations are included in Table S1 for the 90 Q model library.

3.3. TD-DFT Correlations to Absorption and Excited State Distortions

Normal mode vibrations in both Q closed core (ν₁, Figure 4 top) and open core models (ν₁ and ν₂, Figure 4 middle and bottom) have been identified as candidates for the 690 cm⁻¹ feature in its rR spectrum. The UV-vis absorption spectrum for Q exhibits an intense (ε = 10 mM⁻¹cm⁻¹) band at 350 nm (28,600 cm⁻¹; Figure 5A) which, when probed with a 351 nm (28,500 cm⁻¹) laser line, results in the resonance-enhancement of the characteristic 690 cm⁻¹ vibration. The absorption spectrum additionally shows a weak feature at 800 nm (12,500 cm⁻¹; ε = 1.2 mM⁻¹cm⁻¹. Insert Figure 5A), as well as rising intensity between 15,000 and 28,000 cm⁻¹ with a local maximum at 23,000 cm⁻¹ (ε = 8 mM⁻¹cm⁻¹). To correlate the Q models identified in Section 3.2 to the experimental absorption spectrum and its associated rR data, there are two points of focus: 1) prediction of the absorption spectra for structural subtypes of Q and in particular the nature of the transitions in the 350 nm region; and 2) determination of the excited state distortions associated with the 350 nm transitions that would lead to resonance enhancement. Therefore, TD-DFT (using the CAM-B3LYP functional to obtain reasonable carboxylate charge transfer energies)⁵⁸ was employed to calculate and assign the absorption spectra of the Q models (identified in Section 3.2) and predict the excited state distortions associated with 351 nm excitation.

Figure 5.

Q UV-vis absorption spectrum and TD-DFT calculated spectra and predicted rR enhancement of Q models. (A) Experimental absorption spectrum for intermediate Q. Inset: expanded scale for low energy region. Adapted from reference 16. (B) TD-DFT spectrum (calculated using CAM-B3LYP, downshifted by 3,000 cm⁻¹ based on calibration in Figure S14), left, of bis(μO)-H₂O-6 as representative of the bis(μO)-L (L = OH⁻, H₂O, or vacant site) models. A qualitative description of transition 2 is shown on the right, along with the LCT for transition 2 calculated for modes ν₁, ν₂, and ν₃ (Figure 4 top). (C) TD-DFT calculated spectrum (left) of OC10 as a representative spectrum of the OC -OH/-OH models. A qualitative depiction of transition 1 is given on the right, along with the LCTs for transition 1 along modes ν₁, ν₂, and ν₄ (Figure 4, middle). (D) TD-DFT spectrum (left) of OC26 as representative of the OC -OH/=O models. Transition 1 is schematically given on the right, along with LCTs for transition 1 for modes ν₂, ν₃, and ν₄ (Figure 4, bottom).

The TD-DFT calculated spectra for closed core and OC -OH/-OH and OC -OH/=O models are shown in Figure 5. The predicted electronic transitions for the three classes are similar and may be described together (see SI, Figures S14-24 and Tables S2-12 for a full description of the TD-DFT calculations and analysis). The weak feature at 12,500 cm⁻¹ in the experimental data (Figure 5A, inset) is predicted to consist of Fe(IV) d-d transitions (Figure 5B-D, insets). The rising intensity above 15,000 cm⁻¹ in the experimental data (Figure 5A) may be described as LMCT transitions involving μ-O, oxo, and OH⁻ donation into the Fe(IV) dσ* LUMO orbitals (Figure 5B-D; the dσ* LUMOs are given in Figure S17, S20, and S23). Importantly, the TD-DFT calculations predict the third critical spectral region incorporating the 28,600 cm⁻¹ experimental peak (Figure 5A) to contain LMCT transitions involving μ-O and OH⁻ donation into the Fe(IV) dπ* orbitals (Figure 5B-D, numbered transitions; the dπ* orbitals are given in Figure S17, S20, and S23) that are highly localized in nature (Note that for the open core models, the acceptor orbitals reside on Fe₂ containing the in-plane OH⁻ ligand in the OC -OH/-OH subclass and the OH⁻ ligand in the OC -OH/=O subclass; for examples, see Tables S8 and S10). In the bis(μO), bis(μO)-H₂O, and bis(μO)-OH subclasses, the (μ-O)₂ → Fe(IV)₂ dπ* LMCT would result in an excited state with an elongation of all four Fe₂-(μ-O)₂ bonds and resonance enhancement of the A_g breathing mode (Figure 4, top), as observed in M₂O₂ (M = metal) systems.^29,30 This prediction based on past model studies was evaluated by using TD-DFT to calculate the linear coupling terms (LCT) for the different possible normal modes in the 650 – 750 cm⁻¹ region, where a large slope in the LCT (i.e., a large change in excited state transition energy with distortion along a normal mode) for a given vibration indicates a large exited state distortion, and thus significant resonance enhancement in its Raman spectrum. As observed in Figure 5B (right), the LCT for the bis(μO)-L (L = OH⁻, H₂O, or vacant axial position) in the symmetric breathing mode has a larger slope (relative to other modes) and is thus calculated to be the dominant resonance enhanced vibration. From extension to the OC -OH/-OH and OC -OH/=O subclasses, we find that the ν₁ mode in the former and the ν₂ mode in the latter (both at ~ 690 cm⁻¹) do have large LCT slopes (Figure 5C-D on right; relative to other modes in this region with small slopes) and are also predicted to have significant resonance enhancement. This derives from the assignments of the transitions in this region of the absorption spectrum as μ-O and OH⁻ → Fe₂(IV) dπ* LMCT (see Tables S8-9). This would result in significant elongations of both the Fe₂-OH and Fe₂-(μ-O) bonds that corresponds to a positive distortion along the ν₁ (OC -OH/-OH) and ν₂ (OC -OH/=O) modes. All three modes identified as candidates for the 690 cm⁻¹ rR feature in the possible Q structural types are thus predicted to be resonance enhanced by excitation at 351 nm (28,500 cm⁻¹). Hence, the TD-DFT calculations herein reproduce findings from previous studies that demonstrate resonance enhancement of the A_g mode (at 690 cm⁻¹) in Fe₂-(μ-O)₂ closed core systems. Importantly, they also predict resonance enhancement of the symmetric ν₁ (OC -OH/OH) and ν₂ (OC -OH/=O) modes at 690 cm⁻¹ in Q open core models upon excitation at 351 nm, which results in μ-O and OH⁻ donation into localized Fe₂(IV) dπ* orbitals.

3.4. NRVS Structural and Spectral Assignment of Q

In the previous sections, a number of models with open and closed core motifs have been evaluated and a limited number have been identified as structural candidates for Q through a correlation of their predicted absorption and rR spectra to experimental data. The remaining models have OC -OH/=O and OC -OH/-OH cores, as well as bis(μO) cores with only glutamate, or OH⁻ and H₂O axial ligands. These models are now further appraised using NRVS. The results from the NRVS experiments in Section 2.1 define unique features for Q in its PVDOS spectrum at 605, 420, 375, 330, 300 and 275 cm⁻¹ (Figure 1A, red spectrum). Decay of these features was observed in two sequential phases (Figure S3), corresponding to the cryoreductive formation of a one-electron reduced high-spin, mixed-valence, antiferromagnetically-coupled Fe(III)Fe(IV) species (Qx), followed by further reduction to an Fe(III)₂ species (Q3). In the first decay phase, the 605 cm⁻¹ and 275 cm⁻¹ features are eliminated, with concomitant intensity growing in between 560-580 cm⁻¹ and 220-240 cm⁻¹. In the second decay phase, intensity between 560-580 cm⁻¹ and 220-240 cm⁻¹ continues to grow and reaches a maximum. A feature at 375 cm⁻¹ in the early Q data also decays during the measurement. Hence, the final decay spectrum (Figure 1A, black), containing contributions from a mixture of Fe(III)Fe(IV) and Fe(III)₂ species, exhibits a feature between 560-580 cm⁻¹, and at 420, 365, 330, 300, 240, and 220 cm⁻¹. Furthermore, the NRVS spectrum for the ¹⁸O₂ Q decay sample (Figure 1B) exhibits an observable but modest (4 - 7 cm⁻¹) downshift relative to the ¹⁶O₂ 560 - 580 cm⁻¹ feature, and shows a 12-15 cm⁻¹ downshift of the ¹⁶O₂ 420 cm⁻¹ feature.

In order to screen the possible candidates for Q using NRVS, the DFT-calculated NRVS spectra for all candidate Q models (90 total) were correlated to the NRVS experimental data (Figure 1A, bottom red spectrum). Models with calculated Q NRVS spectra that reproduce the experimental spectrum were then computationally “cryoreduced” for correlation to the cryoreduced-Q NRVS and Mössbauer data as a further screen. From Section 2.2, the NRVS sample at the end of the measurement contains 0% Q, 50% MMOH_red, and approximately 20% Fe(III)Fe(IV) (Qx) and 30% Fe(III)₂ (Q3) species. Therefore, calculated cryoreduced NRVS spectra were generated using 40% Fe(III)Fe(IV) and 60% Fe(III)₂ species (matching the speciation with MMOH_red subtracted). Though this chosen component mixture is approximate, the calculations below and the experimental data in Figure S3 show that the NRVS features of Fe(III)Fe(IV) and Fe(III)₂ cryoreduced models are similar and therefore deviations from this composition would not significantly impact the calculated NRVS spectra. Two important considerations have been included in the generation of acceptable cryoreduced-Q models. First, in order to maximize the number of models possessing high-spin Fe(III)Fe(IV) and Fe(III)₂ electronic structures, cryoreduction-modeling calculations were performed with B3LYP which preferentially stabilizes high-spin states. The Fe(IV)₂ Q models obtained with 10% HF included with BP86 were reoptimized with B3LYP, which resulted in only minor changes in structures and vibrational profiles (spectra shown in Figure S25). These re-optimized Q structures were then used for cryoreduction modeling. Second, the cryoreduced-Q models were determined by fully optimizing the B3LYP Fe(IV)₂ Q models with high-spin Fe(III)Fe(IV)₂ and Fe(III)₂ electronic structures. This assumption allowed for bond length changes in the iron-ligand bonds and perturbations of hydrogen-bonding interactions involving the μ-O ligands, which we believe is reasonable at cryogenic temperatures. NRVS spectra for models “cryoreduced” in such a fashion (including spectra for both ¹⁶O₂ and ¹⁸O₂ isotopomers) were calculated and compared to the experimental data in Figure 1A (bottom, black spectrum) and Figure 1B (for the isotope perturbation) as an additional screen for the determination of the geometric and electronic structure of Q.

In the following sections, analyses of the NRVS spectra for OC -OH/=O, OC-OH/-OH, and bis(μO)-L models (where L is H₂O, OH⁻, or no water-derived ligand) are presented using the methodology described above.

3.4.1. Open Core Terminal Oxo, Terminal Hydroxide (OC -OH/=O)

Of the eight OC -OH/=O models, only OC1, OC3, OC17, and OC26 possess a ν(Fe-OH) + ν_as(Fe-O-Fe) mode (ν₂ Figure 4, bottom) near 700 cm⁻¹ which would correspond to the resonance Raman feature at 690 cm⁻¹. (Note that the ν₂ modes for OC1 and OC3 exhibit a ²H perturbation of > ∣8∣ cm⁻¹, but are included to generally explore the NRVS properties of this structural subclass.) The calculated Q NRVS spectra of these models are provided in Figure 6 (black spectra), and compared to the Early Q – MMOH_red data in red (top spectrum; note NRVS calculated spectra for other OC -OH/=O models are provided in Table S1). The experimental data correlate best to the OC17 NRVS spectrum. Below 300 cm⁻¹, the OC17 NRVS spectrum exhibits two peaks at 250 and 281 cm⁻¹ that dominate the spectrum, and may be correlated to the 275 cm⁻¹ and the 300 cm⁻¹ peaks in the experimental data. The two peaks in the calculated spectrum at 329 cm⁻¹ and 356 cm⁻¹ would correlate to the experimental data at 330 and 375 cm⁻¹, respectively, while the calculated less intense 390 and 425 cm⁻¹ peaks would correspond to the weak 420 cm⁻¹ feature in the data. Hence, below 450 cm⁻¹, the calculated OC17 NRVS spectrum reasonably correlates to the experimental Q data.

Figure 6.

Correlation of OC -OH/=O computed NRVS spectra to experimental data. The Q – MMOH_red (and renormalized) PVDOS spectrum is shown at top in red (adapted from Figure 1A, bottom). Q data (unsubtracted including error) between 610 and 720 cm⁻¹ are provided in the top right inset. OC -OH/=OH models that possess a ν₂ mode (see Figure 4, bottom) with the approximate 690 cm⁻¹ mixed isotope signal are shown below in black (OC1, OC3, OC17, OC26. Their structures are given in Table S1). The blue region indicates low energy bending modes (ν₅ & ν₆ Figure 4 bottom) and terminal ν(Fe-COO⁻) stretches. The yellow region also contains ν(Fe-COO⁻) stretches and the ν_s(Fe-O-Fe) mode (ν₄ Figure 4, bottom). The red region contains ν₂ and its antisymmetric pair (ν₃ Figure 4, bottom).

Above 450 cm⁻¹, the calculated OC17 NRVS spectrum exhibits one peak at 475 cm⁻¹, followed by a lack of intensity up until ~ 700 cm⁻¹, where two peaks are observed at 691 cm⁻¹ and 722 cm⁻¹. The experimental data above 620 cm⁻¹ show no feature above the increasing high noise within this region (Figure 6, top spectrum, inset on right). However, it is clear that the calculated spectrum does not reproduce the experimental data as there is no calculated peak at 605 cm⁻¹. We thus evaluated whether a vibration of the OC17 model (and OC -OH/=O models more generally) could reproduce the experimental 605 cm⁻¹ peak, with some limited structural perturbation. Vibrations closest to 605 cm⁻¹ in the calculated NRVS spectra are color highlighted in Figure 6 and assessed. The calculated vibrations at 691 and 722 cm⁻¹ correspond to the ν₃ (ν(Fe-OH) - ν_as(Fe-O-Fe); Figure 4 bottom) and ν₂ modes, respectively (red region, Figure 6). The ν₂ mode corresponds to the resonance Raman 690 cm⁻¹ feature, and is thus required to be near 700 cm⁻¹ based on this assignment. Since ν₃ is the antisymmetric version of ν₂ and is a result of mixing (between the ν(Fe-OH) and ν_as(Fe-O-Fe) modes), and since the splitting between ν₂ and ν₃ is ~ 30-40 cm⁻¹, the ν₃ mode cannot be assigned to the 605 cm⁻¹ peak without lowering the energy of ν₂ and sacrificing the rR assignment. Note that the 605 cm⁻¹ feature cannot be assigned as the isolated ν(Fe-OH) mode in these models, since ν(Fe-OH) + ν_as(Fe-O-Fe) mixing is required near 700 cm⁻¹ to reproduce the mixed isotope effect in rR of the ν₂ mode. This rules out models such as OC24 (see NRVS spectrum in Table S1), which possesses a particularly short Fe₁=O bond that highly localizes the μ-O ligand on the Fe₂ side, resulting in strong μ-O donation to Fe₂ and an elongated (1.82 Å) Fe₂-OH bond. This model exhibits a localized ν(Fe-OH) vibration near 600 cm⁻¹; however, only the ν_as(Fe-O-Fe) mode is observed near 700 cm⁻¹, which would not produce the mixed isotope effect in the rR spectrum (and also would not show resonance enhancement). Thus, neither the ν₂, ν₃, nor an isolated ν(Fe-OH) mode can be assigned to the 605 cm⁻¹ peak in the experimental data (indicated by the red X in OC17 Figure 6).

Vibrations below 600 cm⁻¹ were also assessed as possible assignments for the 605 cm⁻¹ feature. The region between 440 and 540 cm⁻¹ (yellow region Figure 6) includes the ν_s(Fe-O-Fe) mode (ν₄ Figure 4, bottom), as well as terminal carboxylate (ν(Fe-COO⁻)) stretches that further distribute into the 345 and 440 cm⁻¹ region (blue region Figure 6) that mostly contains low energy bending modes (ν₅ and ν₆, see Figure 4 bottom). The energy of the ν_s(Fe-O-Fe) mode (and importantly, the higher energy ν_as(Fe-O-Fe) mode) depends primarily on the Fe-O-Fe angle.⁵⁹ However, manipulation of the Fe-O-Fe angle to increase the energy of the ν_s(Fe-O-Fe) mode would decrease the energy of the ν_as(Fe-O-Fe) mode, which would unacceptably decrease the energy of the ν₂ mode that would be assigned to the 690 cm⁻¹ feature in the rR spectrum. Hence, the ν_s(Fe-O-Fe) mode cannot be assigned to the 605 cm⁻¹ NRVS feature. The ν(Fe-COO⁻) modes are too low in energy to be reasonably assigned to the 605 cm⁻¹ feature (yellow and blue regions in Figure 6). Their energies are already calculated to be slightly higher than that experimentally observed for a terminal ν(Fe-COO⁻) mode in a NRVS study⁴⁸ of a high-spin Fe(IV) oxo intermediate in SyrB2 at 370 cm⁻¹. Therefore, the 605 cm⁻¹ experimental NRVS peak cannot be reasonably correlated to either ν(Fe-COO⁻) or ν_s(Fe-O-Fe) modes that must be at energies well below 600 cm⁻¹.

In summary, while OC17 (and OC-OH/=O models in general) reproduces the Q NRVS spectrum reasonably well below 450 cm⁻¹, it fails to capture the 605 cm⁻¹ peak in the experimental data. No vibration in the OC-OH/=O models may be reasonably assigned to the 605 cm⁻¹ NRVS feature; thus, the OC -OH/=O subclass is excluded as a possible candidate for intermediate Q.

3.4.2. Open Core bis-terminal-hydroxide (OC -OH/-OH)

The OC -OH/-OH models (14 total) constitute the remaining open core subclass that are to this point in the analysis possible structures of Q. Of these 14 models, 13 possess a symmetric ν(Fe-OH) + ν_as(Fe-O-Fe) mode (ν₁ Figure 4) near 690 cm⁻¹ with an appropriate ¹⁶O¹⁸O mixed isotope shift, while only 4 of these models (OC10, OC15, OC16, and OC32) also exhibit a small ²H isotope effect (Δ ≤ 3 cm⁻¹) that would reproduce the rR data. (Models exhibiting a larger ²H isotope effect are OC9, OC11, OC13, OC18, OC19, OC20, OC23, OC33, OC34) The calculated NRVS spectra for OC10, OC15, OC16, and OC32 are provided in Figure 7A, and NRVS spectra for all remaining OC -OH/-OH models are included in Table S1.

Figure 7.

NRVS spectra of OC -OH/-OH models and the cryoreduction decay spectra for OC16. (A) Early Q – MMOH_red NRVS data (top), with calculated spectra given below for the four OC -OH/-OH models possessing a ν₁ mode (Figure 4 middle) with appropriate ¹⁶O¹⁸O mixed isotope and deuterium isotope effects (calculated with BP86 and 10% HF). Yellow region contains the ν₄ (Figure 4 middle) and terminal ν(Fe-COO⁻) modes. Red region contains the ν₁, ν₂, and ν₃ modes (Figure 4). (B) Top: Early Q – MMOH_red (red spectrum) and Decayed Q – MMOH_red (black spectrum) NRVS data, with arrows highlighting changes in the spectra from early to late scans. Middle: NRVS spectrum for the B3LYP DFT-optimized OC16 model (red; see Figure S25 for the close correlation of NRVS calculations between B3LYP vs. 10% HF in BP86), as well the calculated decay spectrum (OC16 Decay) consisting of a 40%/60% mixture of its Fe₂(III)Fe₁(IV) and Fe(III)₂ optimized calculated NRVS spectra, respectively (black). Arrows indicate major changes from OC16 to OC16 Decay. Bottom: Calculated NRVS spectra for the OC16 Decay from Figure 7B middle (black) and its ¹⁸O₂ labeled isotopomer (¹⁸O₂ OC16 Decay, blue).

Of the 4 models in Figure 7A, OC16 best reproduces the experimental data. The calculated NRVS spectrum of OC16 (Figure 7) below 450 cm⁻¹ exhibits an intense feature at 278 cm⁻¹, which may be associated with the 275 cm⁻¹ peak in the experimental data (Figure 7A, top red spectrum). The calculated feature at 315 cm⁻¹ would correlate reasonably to the 300 peak in the data, while the broad, intense calculated peak at 357 cm⁻¹ may be associated with the combination of both the 330 and 375 cm⁻¹ features in the data. The 406 cm⁻¹ feature may be correlated to the peaks around 420 cm⁻¹ in the data; hence, the low energy (< 450 cm⁻¹) region of the calculated OC16 spectrum reasonably reproduces the Q NRVS spectrum (with similar results for models OC10, OC15, and OC32).

Above 450 cm⁻¹, both OC16 and OC32 exhibit intensity near 605 cm⁻¹, whereas OC10 and OC15 (as well as the remaining 9 models in the SI) do not. Three vibrations contribute to the calculated OC16 NRVS spectrum in the region between 600 and 750 cm⁻¹: the rR-assigned ν₁ mode (the symmetric combination of the in-plane ν(Fe₂-OH) stretch and ν_as(Fe-O-Fe) mode, centered at ~ 712 cm⁻¹); ν₂ (Figure 4 middle), the antisymmetric version of ν₁ (625 cm⁻¹); and ν₃ (Figure 4 middle), the isolated out-of-plane ν(Fe₁-OH) mode (666 cm⁻¹). Both the ν₂ and ν₃ modes are possible assignments for the 605 cm⁻¹ feature in this structure, with the ν₂ mode closer to 605 cm⁻¹. Interestingly, the ν₁ and ν₂ splitting is larger (40-70 cm⁻¹; red region Figure 7A) in OC -OH/-OH vs. OC -OH/=O models (30-40 cm⁻¹). This is due to the higher symmetry of the OC -OH/-OH models (relative to OC -OH/=O); the two similar Fe-(μ-O) bond lengths result in a lowering of the ν_as(Fe-O-Fe) mode to be closer to ν(Fe₂-OH) (with a concomitant increase of ν_s(Fe-O-Fe)). This increases the ν_as(Fe-O-Fe) + ν(Fe₂-OH) mixing, which results in the larger split ν₁ and ν₂ modes. For models OC10 and OC15 (and the 9 related models in the SI), the ν₂ mode is at ≥ 650 cm⁻¹, which may not be lowered without sacrificing the rR-assigned ν₁ mode. However, in OC16, Glu243 is hydrogen-bonded to Fe₁-OH as well as Fe₂-OH (see structure in Table S1), which weakens its hydrogen-bonding interactions with Fe₂-OH (heavy atom distance increases by 0.1 Å, compared to OC10). This decreases the energy of the ν(Fe₂-OH) mode, which drives down the energies of both the ν₁ and ν₂ modes. Hence, up until this point in the analysis of this structural class ν₂ is a possible candidate for the 605 cm⁻¹ NRVS assignment.

As an additional screen for evaluating OC16 (and OC32; see Figure S27), this model was cryoreduced in DFT to correlate its simulated decay NRVS spectrum to the experimental data in Figure 7B (top, black spectrum). OC16 was thus optimized with high spin Fe(III)Fe(IV) and Fe(III)₂ electronic structures using the methodology outlined above Section 3.4. Optimization of OC16 to a Fe(III)Fe(IV) species afforded two possible models: Fe₂(III)Fe₁(IV) or Fe₂(IV)Fe₁(III), with the Fe₂(III)Fe₁(IV) model lower in energy by ΔE = −3.4 kcal/mol. The NRVS spectra for the cryoreduced OC16 Fe₂(III)Fe₁(IV) and Fe(III)₂ models are provided in Figure S26 (middle and bottom rows) in black (structures for the cryoreduced optimized forms are also provided).

To reproduce the decayed Q NRVS decay spectrum in Figure 7B (top in black) which contains a mixture of both Fe(III)Fe(IV) and Fe(III) species, the cryoreduced mixture spectrum (OC16 Decay) was prepared by summing 40% of the calculated OC16 Fe₂(III)Fe₁(IV) NRVS spectrum and 60% of the Fe(III)₂ spectrum from Figure S26A (black spectra). The changes from OC16 to OC16 Decay (Figure 7B middle, red to black) agree with the data (Figure 7B top, red to black), showing shifts of intensity from 288 to 260 cm⁻¹ (exp 275 to 220-240 cm⁻¹) and 612 to 590 cm⁻¹ (exp 605 to 560-580 cm⁻¹), and decay of the 357 cm⁻¹ feature (exp 375 cm⁻¹). The decay of the calculated 612 cm⁻¹ and formation of the 590 cm⁻¹ features are of particular interest, which in this model would represent the cryoreduced decay of the ν₂ mode corresponding to the 605 cm⁻¹ experimental NRVS feature in Q. In the cryoreduced calculation the 590 cm⁻¹ feature corresponds to a localized ν(Fe₂-OH) mode (Figure 7B, middle); relative to the ν₂ mode in the Fe(IV)₂ Q OC16 model (Figure 7B, middle), this vibration appears downshifted and uncoupled from ν_as(Fe-O-Fe) due elongation of the Fe₂(III)-OH bond upon cryoreduction. This lowers the energy of the ν(Fe₂-OH) mode and limits its mixing with the ν_as(Fe-O-Fe) mode. The calculated 590 cm⁻¹ feature in OC16 Decay also contains contributions from ν_s(Fe-O-Fe) (ν₄ Figure 4), which shifted up in energy (~ 40 cm⁻¹) relative to OC16 due to a decrease of the Fe-O-Fe angle upon cryoreduction. However, both of these modes would be expected to exhibit signficant ¹⁸O₂ perturbations as both oxygen atoms in the modes must be derived from O₂ based on the rR assignment of the 690 cm⁻¹ feature in the Fe(IV)₂ structure.

The ¹⁸O₂ spectrum for OC16 Decay (¹⁸O₂ Q Decay; Figure 7B bottom, blue spectrum) was prepared by summing 40% of the ¹⁸O₂ OC16 Fe₂(III)Fe₁(IV) and 60% of the ¹⁸O₂ Fe(III)₂ NRVS spectra (provided in Figure S26B, blue spectra). As anticipated, the NRVS spectrum for ¹⁸O₂ Q Decay predicts a large (22 cm⁻¹) isotope downshift for the OC16 Decay 590 cm⁻¹ feature, which does not agree with the 4-7 cm⁻¹ shift observed experimentally.

Note that from the analysis above, the ν₄ mode would also exhibit a large isotope dependence in the cryoreduced model (Figure S26B, bottom) and can be excluded. Therefore, the remaining possible assignment for the 605 cm⁻¹ feature in the experimental data is the ν₃ mode (ν(Fe₁-OH)) centered at ~ 690 cm⁻¹ (note this is slightly upshifted relative to the BP86/10% HF calculation value of 666 cm⁻¹). In the OC16 Decay spectrum this mode decreases in energy by 19 cm⁻¹. This mode is not calculated to have an ¹⁸O₂ isotope dependence in the ¹⁸O₂ OC16 Decay spectrum (Figure 7B, bottom), since the O₂ loading would occur on the Fe₂-OH and μ-O oxygen atoms. Weak mixing with the ν₄ mode could provide a small ν₃ ¹⁸O₂ isotope shift as observed experimentally. However, the energy of the ν₃ mode at > 650 cm⁻¹ (Figure 7A, red column) is too high to reproduce the 605 cm⁻¹ NRVS feature and to facilitate mixing with the ν₄ mode. In principle, a model possessing a weaker Fe₁-OH bond and thus a lower energy ν(Fe₁-OH) mode could reproduce the Q 605 cm⁻¹ NRVS feature and the 4-7 cm⁻¹ isotope shift of the Q decay 580 cm⁻¹ feature.

The approximate Fe₁-OH bond length required to produce a 605 cm⁻¹ ν(Fe₁-OH) mode was calculated using a Badger’s rule analysis (Figure S28).⁶⁰ The 605 cm⁻¹ feature could be reproduced by a ν(Fe₁-OH) mode corresponding to a 1.85 Å Fe-OH bond length. Models with hydrogen-bond donors to the Fe₁-OH oxygen (OC9, OC13, OC21, OC33) result in modest elongation of the Fe-OH bond to 1.76 – 1.78 Å, and exhibit NRVS intensity above 660 cm⁻¹ (see Table S1). Indeed, none of the 14 OC -OH/-OH models reproduce this long Fe₁-OH bond corresponding to a ν(Fe₁-OH) mode near 605 cm⁻¹.

To summarize, the OC -OH/-OH subclass reasonably reproduces the low-energy (< 450 cm⁻¹) region of the Q NRVS spectrum. However, the ¹⁸O₂ 4-7 cm⁻¹ isotope shift of the 560-580 cm⁻¹ feature in the Q decay data eliminates assignment of the Q 605 cm⁻¹ NRVS feature to vibrations strictly containing motion of the in-plane Fe₂-OH and μ-O ligands. The one plausible assignment of the 605 cm⁻¹ feature in the OC -OH/-OH structure is the out-of-plane ν₃ (ν(Fe₁-OH)) mode; however, the calculated Fe₁-OH bond length is too short in the OC -OH/-OH subclass (due to the relatively low donor strength of the ligands on Fe₁) to produce a ν₃ mode near this energy. The relatively high energy of this mode prevents mixing with the ν₄ mode that would be required for the ¹⁸O₂ 4-7 cm⁻¹ isotope shift of the Q decay 580 cm⁻¹ feature. Though the assignment of Q to the OC -OH/-OH subclass is not well-supported by our calculations, we leave it as an open possibility pending further experimental evidence.

3.4.3. Closed core

The closed core subclasses available as possible assignments for Q are the bis(μO), bis(μO)-H₂O, and bis(μO)-OH models. The subset of these models possessing an appropriate ¹⁶O¹⁸O mixed isotope shift include all 6 bis(μO) models (bis(μO)-1-6), 12 out of 16 of bis(μO)-H₂O models (bis(μO)-H₂O-1-4, 6, 8-12, 14-15), as well as 9 out of 16 of the bis(μO)-OH models (bis(μO)-OH-6, 8, 10-16). The NRVS spectra for all closed core models are included in Table S1 (along with their structures).

Of these remaining models, four models (bis(μO)-OH-6, bis(μO)-OH-8, bis(μO)-OH-10, bis(μO)-H₂O-14) lack in their calculated NRVS spectra the characteristic 275 cm⁻¹ signal present in the early Q data (Figure 8A, red spectrum) and are therefore excluded from further analysis. The calculated NRVS spectrum of bis(μO)-H₂O-6 corresponds particularly well to the early Q NRVS data, and is representative of the NRVS spectral behavior of the possible bis(μO) and bis(μO)-H₂O closed core models. Below 450 cm⁻¹, the bis(μO)-H₂O-6 NRVS spectrum (Figure 8B) exhibits one intense feature at 267 cm⁻¹, which may be associated with the intense 275 cm⁻¹ peak in the early Q data (Figure 8A, red spectrum). At higher energy, four peaks of moderate intensity are calculated at 306, 349, 378 and 413 cm⁻¹, which would correspond to the features in the experimental data at 300, 330, 375, and 420 cm⁻¹. Thus, below 450 cm⁻¹, the calculated NRVS spectrum for bis(μO)-H₂O-6 reproduces the early Q NRVS data reasonably well.

Figure 8.

Calculated NRVS spectra of bis(μO)-H₂O-6 as representative of bis(μO) and bis(μO)-H₂O subclasses. All spectra simulated with B3LYP. (A) Early Q – MMOH_red (red) and Decayed Q – MMOH_red (black) NRVS data. Arrows indicate major spectral differences. (B) Calculated NRVS spectrum of bis(μO)-H₂O-6. (C) Calculated NRVS spectra bis(μO)-H₂O-6 (red) and its simulated decay product bis(μO)-H₂O-6 Decay (black). Arrows indicate major spectral differences. (D) Calculated NRVS spectra for bis(μO)-H₂O-6 Decay (black) and its ¹⁸O₂ isotopomer (blue).

Above 450 cm⁻¹, two features of similar intensity are calculated at 614 and 631 cm⁻¹. These two features correlate well to the early Q NRVS feature at 605 cm⁻¹; indeed, all 23 possible closed core models possess features between 600 and 640 cm⁻¹ which could correspond to the 605 cm⁻¹ feature in the early Q NRVS data. In bis(μO)-H₂O-6, the two features calculated at 614 and 631 cm⁻¹ correspond to the B_2u (ν₃ Figure 4) and B_3u (ν₂ Figure 4) non-symmetric modes of the Fe₂O₂ core. The remaining high-energy Fe₂O₂ stretching mode, the A_g breathing mode of the core (ν₁ Figure 4) that would be assigned to the 690 cm⁻¹ rR vibration, is calculated at 693 cm⁻¹ and is much weaker in NRVS than the non-symmetric modes. Therefore, at this point in the analysis, the high-energy (> 450 cm⁻¹) region of the early Q NRVS data may reasonably be assigned to the non-symmetric Fe₂O₂ stretches of bis(μO)-H₂O-6 (as well as in the other closed core models).

The bis(μO)-H₂O-6 and related acceptable models were then further screened through NRVS calculations based on their cryoreducted DFT structures for correlation to the decayed Q data (Figure 8A, black spectrum). For the H₂O-Fe₂-(μ-O)₂-Fe₁ structure of bis(μO)-H₂O-6, optimization to an Fe₂(III)Fe₁(IV) electronic structure was lower in energy by 2.8 kcal/mol. The decay spectrum for bis(μO)-H₂O-6 (bis(μO)-H₂O-6 Decay) is shown in Figure 8C (40% Fe₂(III)Fe₁(IV) and 60% Fe(III)₂, their individual NRVS spectra are shown in Figure S29A (ii) and (iii)). The spectral differences between bis(μO)-H₂O-6 and bis(μO)-H₂O-6 Decay (Figure 8C red to black) correlate well to the decay of Q features in the beam (Figure 8A, red to black), which include decay of the intense 267 cm⁻¹ feature (exp 275 cm⁻¹) with growth of features between 190 and 215 cm⁻¹ (exp 220-240 cm⁻¹), decay of the 378 cm⁻¹ feature (exp 375 cm⁻¹), and a downshift in energy of the 631 cm⁻¹ feature (exp 605 cm⁻¹) to 560 cm⁻¹ (exp 560 to 580 cm⁻¹). Though the downshift of the calculated 631 cm⁻¹ feature agrees with experiment, this mode corresponds to the ν₂ non-symmetric stretch of the Fe₂O₂ core and is calculated to have an ¹⁸O₂ isotopic shift (> 20 cm⁻¹; Figure 8D) larger than accommodated by the experimental data (4-7 cm⁻¹ from Figure 1B). However, the related closed core models with the axial H₂O replace by an OH⁻ are consistent with experiment.

Unlike the bis(μO) and bis(μO)-H₂O subclasses, the bis(μO)-OH models possess an additional mode between 600-700 cm⁻¹ corresponding to the ν(Fe-OH) stretching mode, and are therefore further evaluated. The calculated NRVS spectrum for bis(μO)-OH-12 is given in Figure 9B (and its structure is given in Figure 10). Below 500 cm⁻¹, the calculated NRVS spectrum correlates well to the experimental data (Figure 9A, red spectrum), exhibiting one intense peak at 270 cm⁻¹ (exp 275 cm⁻¹), three additional features at 300 cm⁻¹ (exp 300 cm⁻¹), 336 cm⁻¹ (exp 330 cm⁻¹) and 381 cm⁻¹ (exp 375 cm⁻¹), and intensity at 428 and 467 cm⁻¹ (exp 420 cm⁻¹). Above 450 cm⁻¹, the calculated spectrum shows a shoulder at 589 cm⁻¹, followed by strong intensity between 604 - 621 cm⁻¹ (with a shoulder at ~ 620 cm⁻¹ and a weak feature at 692 cm⁻¹. The intense peak at ~ 604 cm⁻¹ corresponds to the ν(Fe-OH) mode, while the remaining intensity corresponds to the breathing mode (692 cm⁻¹; assigned to the rR feature at 690 cm⁻¹) and two of the non-symmetric core modes (589 and ~ 620 cm⁻¹). Interestingly, this ν(Fe-OH) mode is downshifted by more than 50 cm⁻¹ compared to the out-of-plane ν(Fe-OH) mode in the OC - OH/-OH subclass (Section 3.4.2). This reflects the longer Fe-OH bond length in the bis(μO)-OH (~ 1.83 A) vs. OC -OH/-OH (~ 1.77 A) models, due to the bis(μO) bridges providing a stronger donating equatorial core. From correlation to the experimental Q NRVS data, any one of the calculated three modes in the bis(μO)-OH-12 NRVS spectrum between 580 cm⁻¹ and 630 cm⁻¹ could correspond to the Q NRVS feature at 605 cm⁻¹.

Figure 9.

Calculated NRVS spectra of bis(μO)-OH-12 as representative of the bis(μO)-OH subclass. All spectra simulated with B3LYP. (A) Early Q – MMOH_red (red) and Decayed Q – MMOH_red (black) NRVS data. Arrows indicate major spectral differences. (B) Calculated NRVS spectrum of bis(μO)-OH-12. (C) Calculated NRVS spectra for bis(μO)-OH-12 (red) and its simulated cryoreduced product (bis(μO)-OH-12 Decay in black; see text for details). Arrows indicate major spectral differences. (D) Calculated NRVS spectra for bis(μO)-OH-12 Decay (black) and its ¹⁸O₂ isotopomer (blue).

Figure 10.

Structure for bis(μO)-OH-12.

The bis(μO)-OH-12 model was further evaluated by DFT calculated cryoreduction to Fe(III)Fe(IV) and Fe(III)₂ electronic structures. Optimization on the Fe(III)Fe(IV) level yielded both Fe₂(III)Fe₁(IV) and Fe₂(IV)Fe₁(III) models (hydroxide bonded to Fe₂) differing in energy by only 1.5 kcal/mol. Therefore, the NRVS spectra of both models are included in the analysis of the cryoreduction data. Additionally, optimization of bis(μO)-OH-12 to an Fe(III)₂ structure resulted in proton transfer from Glu144 to the μ-O bridge it hydrogen-bonds to in the Q structure (Figure 10); structures and bond lengths of these cryoreduced models are found in Figure S30 and Table S13, respectively.

The cryoreduced NRVS spectrum of bis(μO)-OH-12 (bis(μO)-OH-12 Decay) was generated by combining 20%, 20%, and 60% of the bis(μO)-OH-12 Fe²(III)Fe₁(IV), Fe₂(IV)Fe₁(III), and Fe(III)₂ NRVS spectra (Figure S30 (ii), (iii), and (iv)). The spectral changes from bis(μO)-OH-12 to bis(μO)-OH-12 Decay (Figure 9C, red to black) correlate reasonably well to the changes in the Q NRVS spectrum observed in the beam (Figure 9A, red to black). Decay of the calculated feature at 270 cm⁻¹ (exp 275 cm⁻¹) is observed with concomitant growth of intensity between 180-215 cm⁻¹ (exp 220-240 cm⁻¹), as well as decay of a feature at 381 cm⁻¹ (exp 375 cm⁻¹). At high energy, a downshift of calculated intensity is observed in the 604-621 cm⁻¹ region (exp 605 cm⁻¹) to 560-583 cm⁻¹ (exp 560-580 cm⁻¹), corresponding to a decrease in the energies of the non-symmetric stretches and ν(Fe-OH) mode. Note that, in the distorted Fe(III)Fe(IV) and (μ-OH)(μ-O) Fe(III)₂ structures of the cryoreduced models, the Fe-O core stretches localize (therefore, Fe-(μ-O) stretching modes are indicated as ν(Fe-O) in black in Figure 9C).

The calculated ¹⁸O₂ NRVS spectrum of bis(μO)-OH-12 Decay (¹⁸O₂ bis(μO)-OH-12 Decay) reveals a striking difference in the isotope sensitivity of the high energy region ~ 600 cm⁻¹) relative to the other bis(μO) subclasses (Figure 9D black to blue). Vibrational modes with Fe-(μ-O) stretching motion above 500 cm⁻¹ exhibit ¹⁸O₂ downshifts larger than 20 cm⁻¹, which are not observed in the experimental data (Δ = 4-7 cm⁻¹) and is similar to the findings for the bis(μO) and bis(μO)-H₂O subclasses. However, a significant peak at 583 cm⁻¹ in the bis(μO)-OH-12 Decay NRVS spectrum does not exhibit an isotope shift in the calculated ¹⁸O₂ spectrum (Figure 9D). This corresponds to the ν(Fe₂-OH) mode, for which the hydroxide oxygen atom remains unlabeled upon ¹⁸O₂ loading. Note that the limited 4-7 cm⁻¹ isotope shift observed would reflect some mixing with the non-symmetric Fe₂O₂ core modes in this energy region (as observed in the calculated bis(μO)-OH-16 NRVS spectra: see Figure S31). Indeed, in comparison to all other subclasses of models analyzed, the bis(μO)-OH subclass best reproduces the two defining characteristics of the Q 605 cm⁻¹ NRVS feature: the small ¹⁶O₂ → ¹⁸O₂ 4-7 cm⁻¹ isotope shift in the Q decay NRVS spectrum, which reflects the unlabeled Fe-OH bond; and the low energy of this ν(Fe-OH) stretch, reflecting the increased Fe₂-OH bond length due to the strong equatorial (μ-O)₂ donation in the bis(μO)-OH core. Hence, of the 90 models evaluated in this study, the closed core bis(μO)-OH subclass (Figure 10) best reproduces the Q and Q decay NRVS data in the higher energy region.

With the assignment of this subclass as the best candidate for the structure of Q, assignment of the Q and Q decay NRVS spectra is now provided (Figure 9B) using bis(μO)-OH-12. The most intense mode (Figure 9B) calculated at 270 cm⁻¹ (exp 275 cm⁻¹) corresponds to two ν(Fe-N) stretching modes. The decay of these features with concomitant growth of intensity between 180 - 215 cm⁻¹ (exp 220-240 cm⁻¹) in bis(μO)-OH-12 decay reflects elongation of the Fe-N bonds upon cryroeduction. The two features at 300 cm⁻¹ (exp 300 cm⁻¹) and 336 cm⁻¹ (exp 330 cm⁻¹) are assigned to the A_g accordion (ν₅ Figure 4) and A_u butterfly (ν₆ Figure 4) modes of the closed core, respectively. These vibrations primarily reflect bending motions of the Fe₂O₂ core, with force constants less dependent on Fe-ligand bond lengths; thus, upon cryoreduction, the energy of these vibrations remains relatively unchanged. The feature at 381 cm⁻¹ (exp 375 cm⁻¹) is assigned to a ν(Fe-COO⁻) terminal carboxylate stretch, and decay of this feature reflects an elongation of the Fe-COO⁻ bond upon cryoreduction. The two calculated peaks between 428 and 467 cm⁻¹ (exp 420 cm⁻¹) correspond to the Fe₂O₂ core B_1g mode (ν₄ Figure 4), mixed into symmetric and antisymmetric combinations with a nearby terminal carboxylate stretching mode. In the cryoreduced models, the structure distorts and the core modes localize (indicated in Figure 9C by ν(Fe-O) in black). The cryoreduced Fe(III)₂ (μ-O)(μ-OH) model now has μ-OH antisymmetric ν_as(Fe-OH-Fe) and symmetric ν_s(Fe-OH-Fe) modes. The lowest energy ν(Fe-O) and ν_s(Fe-OH-Fe) modes are calculated to be between 400 and 450 cm⁻¹, in a similar energy regime as the B_1g mode in bis(μO)-OH-12, and exhibit an ¹⁸O₂ downshift in ¹⁸O₂ bis(μO)-OH-12 Decay (16 cm⁻¹) that corresponds well to the isotope shift of the 420 cm⁻¹ NRVS feature in Q decay (12-15 cm⁻¹). Finally, as presented above, non-symmetric Fe₂O₂ core stretches as well as the ν(Fe₂-OH) mode are calculated to be between 589 – 621 cm⁻¹ in the Q model, which lower in energy to 560-583 cm⁻¹ in bis(μO)-OH-12 Decay. However, from the data the ν(Fe₂-OH) mode (calc ~ 604 cm⁻¹) is mostly observed in the NRVS spectrum, based on the limited ¹⁶O₂ and ¹⁸O₂ sensitivity of the 560 – 580 cm⁻¹ region in the Q decay data.

In the process of completing this study, a higher resolution XFEL structure was published (Srinivas et al., 2020)⁵⁶ of MMOB-associated MMOH. Compared to the crystal structure (PDB: 4GAM) of the MMOB/MMOH complex used for the analysis herein,⁵⁴ the XFEL-derived structure exhibits two oxygenic bridging ligands as well as monodentate ligation of Glu114, Glu243, and Glu209 (in contrast to their bidentate ligation in 4GAM, likely due to cryoreduction).⁵⁶ Small differences in the C_α constraints are also observed in Glu114 (Δ = 0.80 Å) and Glu243 (Δ = 0.46 Å). The influence of these structural changes on our computational models was determined by generating bis(μO)-H₂O-1 and subsequent models of Q from the XFEL-derived structure and resimulating their NRVS spectra (Figure S32). These XFEL-derived structures are very similar and their calculated NRVS spectra show negligible deviations from those used in this analysis.

3.5. Correlation to Reactivity

Intermediate Q in sMMO (Mt OB3b) abstracts an H atom from methane with a kinetic barrier ΔG^‡ of 10.3 kcal/mol.^31,61 For comparison, the mononuclear non-heme iron enzyme Fe(IV)=O intermediate in SyrB2 has a ΔG^‡ ~ 17.7 kcal/mol for H-atom transfer (HAT) from threonine at the same temperature.⁶² This can be propagated to the stronger C-H bond of CH₄ (cf. ΔH_BDE(threonine) ~ 94 kcal/mol vs. ΔH_BDE(CH₄) ~ 104.5 kcal/mol). This 10 kcal/mol worse thermodynamics would raise the HAT reaction barrier between the Fe(IV)=O site in SyrB2 and CH₄ to ~ 24 kcal/mol. Thus, the binuclear Fe(IV) intermediate Q is far more reactive than the mononuclear NH Fe(IV)=O intermediate. This substrate-adjusted kinetic barrier for SyrB2 is well reproduced by reaction coordinate calculations, which predict ΔG^‡ ~ 22.1 kcal/mol for the CH₄ reaction in Figure 11(a) and Table 1.

Figure 11.

Reaction coordinates calculated for hydrogen atom abstraction from methane by (a) the mononuclear Fe(IV)=O site in SyrB2, (b) the open core HO-Fe(IV)-μO-Fe(IV)=O site in OC26, (c) the open-core HO-Fe(IV)-μO-Fe(IV)-OH site in OC15, and (d) the closed core bis(μO)-bridged Fe(IV)₂ site in bis(μO)-OH12. The Fe, C, N, O, Cl, and H atoms are displayed in green, gray, blue, red, yellow, and white, respectively.

Table 1.

Potential energy, enthalpy, and free energy changes and intrinsic barriers (in kcal/mol) involved in hydrogen atom abstraction from methane by models in Figure 11

Systems	ΔE; ΔH; ΔG	ΔE‡; ΔH‡; ΔG‡	$\Delta E_{int}^{‡}$; $\Delta H_{int}^{‡}$; $\Delta G_{int}^{‡}$
Fe(IV)=O	17.1; 14.7; 14.3	25.1; 20.1; 22.1	15.3; 11.6; 14.0
HO-Fe(IV)-μO-Fe(IV)=O	10.8; 7.2; 10.4	24.8; 19.7; 25.0	19.0; 15.9; 19.4
HO-Fe(IV)-μO-Fe(IV)-OH	0.7; −0.3; −2.1	19.8; 14.8; 15.4	19.4; 14.9; 16.4
Fe(IV)-(μO)2-Fe(IV)-OH**	3.6; 2.1; −1.0	15.7; 11.2; 12.4	13.9; 10.1; 12.9

^*ΔG values are calculated at 5 °C.

^**ΔG and

^‡ΔG values for Fe(IV)-(μO)₂-Fe(IV)-OH₂ are 3.1 and 11.1 kcal/mol.

Three binuclear models, OC26 (-OH/=O), OC15 (-OH/-OH), and bis(μO)-OH12, have been evaluated for their H-atom abstraction ability relative to this mononuclear site. The Fe(IV)=O site of the OC -OH/=O model is predicted to react with CH₄ at a similar rate as SyrB2, with the calculated ΔG^‡ ~ 25.0 kcal/mol. While this reaction appears more driven by the thermodynamics (ΔΔE ~ 6 kcal/mol and ΔΔG ~ 4 kcal/mol), this difference is due to an H bond that forms to the OH group of OC26 generated upon H-atom abstraction (HAA). Without this H-bond, the HAA product of the OC -OH/=O model is destabilized by 5 and 8 kcal/mol in potential energy and free energy, respectively (Figure S33). These energy profiles indicate that the activity of the Fe(IV)=O site is similar, regardless of whether it is present in a mononuclear or OC μ-O bridged binuclear site.

The H atom abstraction from CH₄ by the Fe(IV)-OH site in the open-core -OH/-OH model has a kinetic barrier of ΔG^‡ ~ 15.4 kcal/mol and the calculated thermodynamics of ΔG ~ −2.1 kcal/mol. These values are 9.6 and 12.5 kcal/mol lower than those of the Fe(IV)=O site in the OC -OH/=O model, respectively, showing that the Fe(IV)-OH system can be more reactive than the Fe(IV)=O system. When Marcus theory is used to remove the thermodynamic contribution to the barrier, the intrinsic barriers obtained for the Fe(IV)=O and Fe(IV)-OH systems are similar (last column in Table 1; Figure S34). Thus, the lower HAA barrier of the Fe(IV)-OH system is attributed to its greater thermodynamic driving force relative to the Fe(IV)=O system. Indeed, a hypothetical mononuclear Fe(IV)-OH site in SyrB2 was evaluated and gave similar thermodynamics to the -OH/-OH model (Figure S34). Upon HAA, the Mayer bond order between the Fe and the O decreases by 0.883 for the Fe(IV)=O to Fe(III)-OH conversion while it is 0.718 for the Fe(IV)-OH to Fe(III)-OH₂ conversion, implying that the latter costs less energy and is thus more favorable. The calculated barrier for the open-core -OH/-OH model of ΔG^‡ ~ 15.4 kcal/mol is comparable to the experimental barrier (ΔG^‡ ~ 10.3 kcal/mol) measured for intermediate Q. Interestingly, the strong HAA reactivity of the M(IV)-OH unit (M = metal) agrees with findings from a recent study on corrole model complexes, in which the Mn(IV)-OH complex was found to be more reactive than the Mn(V)=O analog.⁶³

The preferred closed core model based on the NRVS data contains the bis(μO) bridges and the terminal OH ligand, which differs from previously suggested models that instead contain a terminal water ligand or two μ-1,3-carboxylato bridges with no water-derived ligands.^64-68 This closed core model is calculated to be much more reactive than the mononuclear Fe(IV)=O and OC -OH/=O models, with ΔG^‡ ~ 12.4 kcal/mol and ΔG ~ −1.0 kcal/mol (Table 1). This low barrier agrees well with the experimental barrier of intermediate Q. The origin of this high HAA activity of the closed-core relative to the open-core systems reflects their key structural difference. As presented above, the open-core models are very similar to the corresponding mononuclear systems in terms of HAA activity. In contrast, the 2^nd μ–O bridge of the closed core systems cooperates with the H atom-accepting μ–O bridge, signifying a key role of the 2^nd Fe center in the HAA process. More details will be discussed below.

4. Discussion

In this study, we have systematically evaluated a significant range of open and closed core candidate structures of intermediate Q and assessed their viability in reproducing prior experimental as well as newly acquired NRVS data. Though the original 2.46 Å Fe-Fe distance favoring a closed core structure for Q has been determined to be an artifact,⁴³ still reasonable evidence for this assignment was provided by resonance Raman spectroscopy²⁸ that showed the presence of a 690 cm⁻¹ vibration exhibiting an averaged ¹⁶O¹⁸O mixed isotope shift. Herein, our analysis of closed core models of Q has shown that a subset of these models also possessing carboxylate (bis(μO)), water (bis(μO)-H₂O), and hydroxide (bis(μO)-OH) axial ligands would show a resonance-enhanced Raman vibration near 690 cm⁻¹ with the appropriate isotope perturbations. From TD-DFT calculations of absorption spectra, the transitions associated with the 351 nm line excitation for the closed core structure correspond to charge transfers from both μ-O bridges to both Fe(IV) ions. Such an excited state would result in elongations of all four Fe-(μ-O) bonds, leading to resonance enhancement of the A_g breathing mode characteristic of M₂(μ-O)₂ (M = metal) cores^29,30,69 (ν₁ at 690 cm⁻¹ in Fe(IV)₂(μ-O)₂, Figure 4, top). Incorporation of ¹⁶O¹⁸O into the (μ-O)₂ core would provide the observed mixed isotope shift.

However, the analysis presented here shows that the experimental 690 cm⁻¹ rR feature and its ¹⁶O¹⁸O shift are not necessarily indicative of a closed core geometry. Two subclasses of open core models have been identified that are predicted to reproduce the rR data for Q: HO-Fe₂(IV)-O-Fe₁(IV)=O (OC -OH/=O) and HO-Fe₂(IV)-O-Fe₁(IV)-OH (OC -OH/-OH; Fe₁-OH vector oriented out of the Fe-O-Fe plane). Interestingly, the calculated TD-DFT absorption spectra for both subclasses (as well as the closed core subclasses) are similar, predicting a low-energy region (8,000 – 15,000 cm⁻¹) comprised of Fe(IV) d-d transitions, followed by a region (15,000 – 23,000 cm⁻¹) of μ-O and OH⁻ LMCTs to the exchange-stabilized dx²-y² LUMO. Above 23,000 cm⁻¹ (and relevant to the 351 nm/28,500 cm⁻¹ rR excitation) TD-DFT calculations predict μ-O and OH⁻ charge transfer to localized dπ* orbitals of the Fe(IV)₂ centers. Such an excited state would result in a symmetric elongation of the in-plane Fe₂-OH and Fe₂-(μ-O) bonds. In the OC -OH/=O and -OH/-OH subclasses, this excited state distortion would result in the resonance enhancement of the ν(Fe₂-OH) + ν_as(Fe-O-Fe) symmetric stretching mode (ν₂ Figure 4 bottom, and ν₁ Figure 4 middle, respectively) near 700 cm⁻¹ that would reproduce the 690 cm⁻¹ experimental feature. Hence, the resonance Raman data for Q do not rule out an open core structural assignment.

Unlike resonance Raman, NRVS observes all vibrations of a ⁵⁷Fe active site with iron motion and has been used in this study to probe the vibrational spectrum of intermediate Q up to 720 cm⁻¹. The NRVS data of Q (Figure S3) exhibit time-dependent changes in the beam, most dramatically the rapid decay of an intense feature at 275 cm⁻¹ in the early time data with concomitant growth of intensity between 220-240 cm⁻¹ in later scans. Tracking this feature during the NRVS measurement allowed for quantitative chronological binning of the data and elucidation of intermediate Q vs. Q decay NRVS features (Figure 1A), which allowed identification of a key additional feature at 605 cm⁻¹ associated with the Q NRVS spectrum that downshifted to 560 - 580 cm⁻¹ in Q decay. The DFT simulations of the NRVS spectra for the Q models predict an intense low energy (< 300 cm⁻¹) feature for both open and closed core structural subtypes due to Fe(IV)-N(His) vibrations; hence, the 275 cm⁻¹ NRVS feature is a distinguishing feature of Q but is not characteristic of a particular structure. In contrast, these simulations have shown that the energy of the 605 cm⁻¹ NRVS feature, as well as the ¹⁶O₂ → ¹⁸O₂ 4 – 7 cm⁻¹ isotope shift of the 560 - 580 cm⁻¹ NRVS feature of Q decay (Figure 1B), does enable evaluation of possible structural subtypes for intermediate Q.

Indeed, the presence of this high-energy NRVS feature eliminates the assignment of an OC -OH/=O structure to Q. This subclass had been postulated as a structural candidate for Q based on the high C-H bond reactivity of an OC -OH/=O model complex relative to its closed core precursor,⁴⁰ as well as the intense pre-edge region of Q that derives from 3d-4p mixing requiring a non-centrosymmetric environment, thought to reflect the terminal Fe(IV)=O unit in a μ-O bridged open core.⁴² However, the OC -OH/=O model calculations cannot reproduce both the experimental 605 cm⁻¹ NRVS and 690 cm⁻¹ rR features. A strong Fe₁=O oxo bond would result in a long Fe₁-(μ-O) and short Fe₂-(μ-O) bond, resulting in a long Fe₂-OH bond that would exhibit a stretching mode near 605 cm⁻¹ (Figure 12); however, this would result in an isolated ν_as(Fe-O-Fe) mode above 700 cm⁻¹ that would not reproduce the resonance-enhanced mixed isotope shift of the 690 cm⁻¹ rR feature. A weak Fe₁=O oxo bond (facilitated through hydrogen-bonding interactions) would reverse this effect, raising the energy of the ν(Fe₂-OH) mode to near 690 cm⁻¹ that allows its mixing with ν_as(Fe-O-Fe) thus reproducing the rR feature and its isotope shifts. However, in this case, no mode near 605 cm⁻¹ would be observed that could be assigned to the NRVS feature in Q. Therefore, due to the donor strength of the terminal oxo and its distortion of the OC -OH/=O core, this subclass cannot accommodate both a 605 cm⁻¹ NRVS and a 690 cm⁻¹ rR feature. A strong oxo can distort the core to provide the NRVS feature, and a weak oxo can distort the core to reproduce the rR feature – but the presence of the oxo prevents both.

Figure 12.

Representation of the ligand donations required to produce a 605 cm⁻¹ NRVS feature in the OC -OH/=O (left), OC -OH/-OH (middle), and bis(μO)-OH (right) subclasses. The direction of donation is symbolized with arrows. In the OC subclasses, the donation required to create a long Fe-OH bond to reproduce the low ν(Fe-OH) 605 cm⁻¹ NRVS feature of Q results in the absence of a ν(Fe-OH) + ν_as(Fe-O-Fe) symmetric mode that would be assigned to the rR feature at 690 cm⁻¹. In the bis(μO)-OH subclass, the (μ-O)₂ core provides strong donation to facilitate a long Fe₂-OH bond and thus a ν(Fe₂-OH) mode near 605 cm⁻¹, and the A_g breathing mode (ν₁ Figure 4) satisfies the 690 cm⁻¹ rR assignment.

The possibility of a new open core structural type, OC -OH/-OH, was further raised in this study and evaluated against the Q and Q decay NRVS data. With two Fe-OH bonds, this model could in principle account for both the 605 cm⁻¹ NRVS feature and the small isotope shift of the Q decay 560 - 580 cm⁻¹ feature (via the out-of-plane Fe₁-OH, which has not incorporated O₂ but could mix weakly with the μ-¹⁸O bridge), and the 690 cm⁻¹ resonance Raman feature and its mixed isotope shifts (due to mixing of the O₂-labeled in-plane Fe₂-OH and μ-O modes). However, all OC -OH/-OH models evaluated in this study possess ν(Fe₁-OH) modes above 650 cm⁻¹ reflecting their relatively short Fe₁-OH bonds (1.76 Å). Hydrogen bonding to the Fe₁-OH oxygen resulted in a modest elongation of the bond, but did not result in a calculated ν(Fe₁-OH) mode near 605 cm⁻¹ that would require a ~ 1.85 Å Fe₁-OH bond length. However, even if a long Fe₁-OH bond was obtained, this would shorten the Fe₁-(μ-O) bond and elongate the Fe₂-(μ-O) bond, thus strengthening the Fe₂-OH bond (Figure 12). A structure with short Fe₂-OH and long Fe₂-(μ-O) bonds would likely eliminate the ν(Fe₂-OH) + ν_as(Fe-O-Fe) coupling required to reproduce the mixed isotope shift of the rR 690 cm⁻¹ feature. While our study has not identified a viable OC -OH/-OH model that satisfies the isotope sensitivity of the Q 690 and Q decay 560 - 580 cm⁻¹ features, we leave this model as a possible assignment for Q contingent on further experimental data.

Analysis of the predicted NRVS spectra for closed core models has yielded an assignment for Q that satisfies both the rR and NRVS data. Of all 90 DFT models considered in this report, the bis(μO)-OH subclass best reproduces the low energy of the 605 cm⁻¹ feature in the Q NRVS spectrum and the 4-7 cm^{−1 18}O₂ isotope shift in the cryoreduced data (between 560 - 580 cm⁻¹) and is our favored assignment for Q. The closed core bis(μO)-OH subclass possesses an additional mode near 605 cm⁻¹ that corresponds to the ν(Fe₂-OH) stretching mode. The low energy of this mode reflects the strongly donating (μ-O)₂ equatorial ligand field, which results in an elongated Fe₂-OH bond (~ 1.83 Å) compared to the OC -OH/-OH models (Figure 12). The ν(Fe₂-OH) mode would show only limited ¹⁸O₂ isotope dependence in the Q decay NRVS spectrum, since O₂ would be loaded into the (μ-O)₂ core and thus the isotope shift of this mode would be contingent on mixing with non-symmetric core modes. It should be qualified that this assignment of Q using NRVS assumes the DFT-calculated cryoreduced structures are reasonable allowing use of the isotope perturbations. However, in the DFT analysis of the cryoreduced data the NRVS spectra clearly exclude the OC -OH/=O subclass.

This assignment of Q is also consistent with past spectroscopic results. From Mössbauer spectroscopy, the two iron ions are fairly equivalent but show a measurable difference in the Mc Bath variant. Mössbauer calculations herein (Table S14) on a variety of models demonstrate that the isomer shifts are fairly similar for all subclasses, while the quadrupole splitting shows some variability in the difference (ΔE_Q1-ΔE_Q2). In the bis(μO)-OH subclass, this difference in quadrupole splitting between the two iron ions is found to primarily depend on the axial carboxylate Glu144-Fe₂ interaction (Figure 10), in which a stronger interaction leads to a smaller ΔE_Q1-ΔE_Q2. From TD-DFT XAS calculations in the SI (Figure S35), the bis(μO)-OH subclass would be consistent with the increased pre-edge intensity of Q as long as at least one Fe(IV) is 5C. Additionally, while an active site with two 6C Fe(IV) would have a K-pre-edge spectrum with a well split dπ-dσpattern with only the dσ being intense, the presence of at least one 5C Fe(IV) site reduces the energy of its dz² peak and produces additional dipole character in the dπ peak, consistent with the experimental spectrum (Figure S35). The Fe-Fe distance of 3.4 Å from HERFD-EXAFS⁴³ is more of an issue. However, the published data contain 50-60% biferrous with a distance of 3.27 Å, and had been extrapolated from a narrow k range of 2 to 11 Å⁻¹ which affects⁷⁰ the Fourier transform that had been fit in the analysis. These fits show two Fe-Fe contributions at 3.05 and 3.30 Å. While the longer distance would rule out a closed core species, elongation of the Fe-Fe distance to 3.0 Å in bis(μO)-OH models costs 2-5 kcal/mol and shifts the resonance Raman active A_g breathing mode down by at least 40 cm⁻¹ (with the correct isotope shifts; see Table S15). The NRVS spectra of these closed core Fe-Fe elongated models exhibit only modest changes near 600 cm⁻¹, due to elongation of the Fe-μ-O bonds that shifts core modes to lower energy while shortening the ν(Fe₂-OH) bond and increasing its vibrational energy by ~ 10 cm⁻¹ (Figure S36).

The DFT-calculated HAA activity of the closed-core bis(μO)-OH models reproduces the experimental value for intermediate Q. Compared to the mononuclear non-heme Fe(IV)=O site in SyrB2, this closed core model has a lower kinetic barrier by ~10 kcal/mol for HAA from CH₄ (Table 1). Given the similarity in intrinsic barriers, the lower barrier of the closed core model can be attributed to the greater thermodynamic driving force for HAA relative to the mononuclear Fe(IV)=O system (ΔΔG ~ −15 kcal/mol). Figure 13 illustrates the origin of this favorable thermodynamics of the binuclear closed-core system based on the concerted motion of the bis(μO) bridges. For the case of the mononuclear system, upon HAA, only the Fe(IV)=O bond significantly changes, while the other ligand-Fe bonds only show minor changes (Figure S37). Alternatively, for the closed-core models, all four Fe(IV)-μO bonds undergo significant variations upon HAA; while Fe₂ and μO_a in Figure 13 accept the electron and proton of the H atom, respectively, the Fe₂-μO_b bond is comparably weakened with a MBO decrease of 0.381. Notably, the weakening of the Fe₂-μO_b bond is compensated by forming a stronger μO_b bond to the Fe₁ center with an increase in MBO of 0.463 (Figure S38). If the Fe₁-μO bonds are kept frozen during HAA, the reaction thermodynamics is predicted to become more endergonic by 12 kcal/mol (Figure 13(b), bottom), equivalent to that of the mononuclear system. Thus, the concerted motion of the bis(μO) bridges coupled to the second Fe center drives the HAA process. After HAA, the rebound of the CH₃· radical can occur with a barrier of ΔG^‡ ~ 8.4 kcal/mol, lower than the HAA barrier of ΔG‡ ~ 12.4 kcal/mol, via opening of the core with the dissociation of the Fe₂(III)-μO_a bond (Figure S39).

Figure 13.

H atom abstraction from CH₄ by (a) the mononuclear Fe(IV)=O system and (b) the binuclear closed-core Fe₂(μO)₂ system with (top) and without (core frozen, bottom) the bis(μO) concerted motions. The electron-accepting Fe centers are displayed in green, while the redox-neutral Fe center is colored in orange.

For the case of the OC -OH/-OH model, this compensation effect from the second Fe center is not significant, as the OC -OH/-OH model is calculated to show similar thermodynamics to that of a hypothetical mononuclear Fe(IV)-OH system (Figure S34). Based on its similar calculated low-energy ΔG^‡ (15.4 vs. 12.4 kcal/mol for the closed core), the -OH/-OH model could also be a candidate for HAT from CH₄. Importantly, the reactivity of the two terminal OH ligands would be constrained by the O₂ labeling results. From the rR data, one oxygen atom from O₂ remains in the μO bridge after HAT, while the other oxygen atom from O₂ is incorporated into the CH₃OH product.³² Since the HAT and the hydroxyl rebound cannot occur from the same OH group, the OH group that abstracts the H atom from CH₄ must be water-derived. However, unlike the bridging oxygen, the terminal OH ligands can undergo isotope scrambling, which was not observed in the rR data.²⁸ Further, this HAT-active OH group would also have to be out of the Fe-μO-Fe plane, because the in-plane OH group must be derived from O₂ to reproduce the ¹⁶O¹⁸O mixed isotope effect for the rR feature²⁸ at 690 cm⁻¹. Therefore, to match existing data, the out-of-plane OH group of the -OH/-OH model would have to abstract the H atom, while the in-plane OH group would need to react with the CH₃ radical to form CH₃OH. However, if CH4 can access the in-plane OH group, it is also able to perform HAT with an even lower barrier of ΔG^‡ ~ 10.2 kcal/mol (Figure S40). Thus, while the binuclear nature of the site might enable the formation of the Fe(IV)-OH group in the -OH/-OH model, which is capable of HAT and rebound to form CH₃OH with a low barrier (Figure S41), there would have to be limited terminal OH⁻ ligand exchange and major positional control of non-polar molecules (CH₄ and CH₃·) by the active-site pocket.

While finalizing this manuscript, another study was published which computationally evaluates previous data on intermediate Q.⁷¹ This study raised the idea that the resonance Raman data of Q can be consistent with an open core structure, which is consistent with our results. However, the authors of this work favor an OC -OH/=O active site geometry, inconsistent with our NRVS data and in particular the 605 cm⁻¹ NRVS feature that excludes this assignment of Q. Additionally, we calculate that this open core structural subclass possesses similar reactivity towards methane as a mononuclear Fe(IV)=O system, inconsistent with the higher reactivity of Q relative to mononuclear non-heme iron enzymes. Alternatively, we do find that closed core models predict larger reactivity due to the compensating effect of the spectator μ-oxo bridge, and additionally find that the OC -OH/-OH subclass may also reproduce the high reactivity of intermediate Q.

5. Conclusion

In this study, NRVS has been used to probe all vibrational modes of intermediate Q and its decay product having Fe motion, thus providing important complimentary information to findings from other spectroscopic techniques. Analyzing these data with DFT calculations provides an experimentally supported geometric/electronic structure for Q that has been combined with kinetic data to obtain insight into its unique reactivity. The analysis herein supports a closed bis(μO) Fe(IV)₂ core for Q, where its high reactivity reflects the compensating effect of the μ–O bridge not involved in the H-atom abstraction. This application of NRVS to intermediate Q highlights the power of this technique in characterizing transient intermediates in iron metalloenzymes.