Introduction

Non-heme diiron enzymes are employed by nature to activate dioxygen for various catalytic oxidation and/or oxygenation reactions1,2. Their catalytic cycles generally employ a diferrous form that reacts with dioxygen to a peroxo-diferric intermediate (P, Fig. 1a). The active species is supposed to be either this peroxo-diferric species or a species derived from it. In soluble methane monooxygenase (sMMO)3,4,5, the peroxo intermediate P converts to a high-valent FeIVFeIV active species (Q, Fig. 1a)1,2,3,4,5,6,7,8. Kinetic studies revealed a pH-dependence indicating that this step is proton-promoted1,3,9,10. The site of protonation is still unknown. Proposals include protonation of the peroxo ligands resulting in bridging μ-1,1- or μ-1,2-hydroperoxo ligands2,3,9,10,11. In other non-heme diiron enzymes12,13,14,15,16,17,18,19,20, a peroxo activation step has been proposed by the conversion of P-type to P′-type intermediates that lack the peroxo → FeIII LMCT around 14,000–15,000 cm−1 and the higher Mössbauer isomer shift characteristic for P-type intermediates. For this peroxo activation step, also a protonation has been suggested15,16,20. For the two diiron arylamine oxygenases AurF and CmlI, μ-1,2-hydroperoxo15 and μ-1,1-peroxo intermediates21 have been proposed, respectively, or a μ-1,1-hydroperoxo intermediate for both (Fig. 1a)14.

Fig. 1: Structural formula.
figure 1

a Supposed intermediates in non-heme diiron enzymes. b The dinucleating ligands susan and susan6-Me. c The mononucleating ligand 6Me2BPP-.

Full size image

Syntheses and detailed spectroscopy and reactivity studies of μ-1,2-peroxo-diferric model complexes22,23,24,25,26,27,28,29,30,31,32,33,34,35 provided not only important structure-spectroscopy correlations to establish peroxo intermediates in the enzymes but also variations in their stabilities and reactivities by slight variations of the ligands. In most cases, the μ-1,2-peroxo-diferric species could only be identified spectroscopically as transient intermediates. Interestingly, protonation of different complexes with a {FeIII(μ-O)(μ-1,2-O2)FeIII} core afforded μ-hydroxo-bridged {FeIII(μ-OH)(μ-1,2-O2)FeIII} species22,27,36 questioning the principle accessibility of hydroperoxo-diferric species.

Here, we present the synthesis, characterization, and reactivity of the rationally stabilized μ-1,2-peroxo complex [(susan6-Me){FeIII(μ-O)(μ-1,2-O2)FeIII}](ClO4)2 using the dinucleating ligand susan6-Me (Fig. 1b)37,38,39. This μ-1,2-peroxo complex is stable even in solution at −40 °C and shows nucleophilic character of the μ-1,2-peroxo ligand attenuated for exogenous organic substrates by encapsulation of the ligand scaffold. [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}]2+ is reversibly oxidized to the high-valent μ-1,2-peroxo complex [(susan6-Me){FeIV(μ-O)(μ-1,2-O2)FeIII}]3+ and reversibly protonated to the μ-1,2-hydroperoxo complex [(susan6-Me){FeIII(μ-O)(μ-1,2-OOH)FeIII}]3+. The study of the electrophilic reactivity for oxygen-atom transfer (OAT) using PPh3 and hydrogen-atom transfer (HAT) using DHA and TEMPOH provides not only a low electrophilic character of the parent μ-1,2-peroxo-FeIIIFeIII complex but also for the oxidized μ-1,2-peroxo-FeIVFeIII and protonated μ-1,2-hydroperoxo-FeIIIFeIII species. Only the oxidized μ-1,2-peroxo-FeIVFeIII species reacts with the relatively weak substrate TEMPOH. The determination of the pKa = 9.5 ± 0.1 and the bond dissociation free energy \({{{\rm{BDFE(OH)}}}}_{{{{{\rm{CH}}}}_3}{{{\rm{CN}}}}}\) = 78 ± 2 kcal mol−1 of the protonated μ-1,2-hydroperoxo-FeIIIFeIII species quantifies the low electrophilic character even of the oxidized μ-1,2-peroxo-FeIVFeIII species. Therefore, the intramolecular C–H activation of preorganized 6-methyl pyridine groups to benzylalcoholato and carboxylato donors in the parent 1,2-peroxo-FeIIIFeIII complex should not occur via the 1,2-peroxo-ligand but via conversion to a more reactive but fluent high-valent species. The low electrophilic character and the spectroscopic signatures of this μ-1,2-hydroperoxo-diferric model are discussed in relation to assignments of reactive intermediates postulated for diiron enzymes.

Results

The complex [(susan6-Me){FeIII(μ-O)(μ-1,2-O2)FeIII}]2+

The reaction of susan6-Me and Fe(ClO4)2·6H2O provided [(susan6-Me){FeII(μ-OH)2FeII}](ClO4)2 (Fig. S1a) and subsequent reaction with O2 at −15 °C the μ-1,2-peroxo complex [(susan6-Me){FeIII(μ-O)(μ-1,2-O2)FeIII}](ClO4)2 (Fig. 2a). Single-crystal X-ray diffraction provides an asymmetric core structure: the μ-1,2-peroxo ligand is coordinated with O1 trans to a tert-amine (N2) and with O2 trans to a pyridine (N44). This results in a shorter Fe1–O1peroxo and a longer Fe2–O2peroxo bond. The resulting different charge donation is compensated by a longer Fe1–O3oxo and a shorter Fe2–O3oxo bond. The O1–O2 bond length of 1.432(2) Å is the longest yet established for a peroxo-diferric complex (1.396–1.426 Å)22,23,24,25,26.

Fig. 2: Characterization of [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}](ClO4)2.
figure 2

a Molecular structure of [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}]2+ in single-crystals of [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}](ClO4)2•0.85CH3CN•0.7H2O. Hydrogen atoms have been omitted for clarity. b 57Fe Mössbauer spectrum of [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}](ClO4)2 at 80 K. The solid line is a simulation with δ = 0.53 mm s−1, |ΔEQ| = 1.69 mm s−1, and Γ = 0.26 mm s−1. c Temperature-dependence of the effective magnetic moment, μeff, of [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}](ClO4)2. The solid line is a simulation to the spin-Hamiltonian (1) in the Supplementary Information with J = −155 cm−1, gi = 2.05, 0.2% p.i. (S = 5/2) of same molecular mass and Θw,pi = −8 K. d UV-Vis-NIR spectrum of [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}](ClO4)2 dissolved in CH3CN at −10 °C and the ligand susan6-Me for comparison.

Full size image

Despite this structural asymmetry, the Mössbauer spectrum exhibits one quadrupole doublet with isomer shift δ = 0.53 mm s−1 and quadrupole splitting |ΔEQ| = 1.69 mm s−1 (Fig. 2b and Fig. S7). This isomer shift is higher than 0.47 mm s−1 of [(susan6-Me){FeIIIF(μ−O)FeIIIF}]2+ (Table 1)40, which is in line with higher values frequently observed for peroxo-diferric complexes1,41. Magnetic measurements (Fig. 2c) revealed an exchange coupling of J = −155 cm−1 (in the convention H = −2 J S1S2) providing an exact value for a μ-1,2-peroxo,μ-oxo-diferric complex. This antiferromagnetic coupling is significantly stronger than −100 ± 20 cm−1 usually found for μ-oxo-diferric complexes42,43 including those with our dinucleating ligands38. This applies also to μ-oxo,μ-carboxylato-diferric complexes which exhibit decreased ∢(FeIII-(μ-O)-FeIII) angles as the μ-1,2-peroxo,μ-oxo-diferric complex. This comparison indicates a significant contribution of the FeIII-μ-1,2-peroxo-FeIII exchange pathway due to shorter (1.88 and 1.93 Å) and hence more covalent FeIII-μ-Operoxo bonds44 compared to the longer FeIII-μ-carboxylato bonds (e.g., 1.97 and 2.07 in [(susan){FeIII(μ-O)(μ-OAc)FeIII}]3+)45 that are considered not to contribute significantly to the exchange coupling.

Table 1 Compilation of experimental (on solids or on frozen CH3CN solutions) and DFT-calculated (shown in italics) 57Fe Mössbauer parameters of the complexes reported here and some complexes with the ligands susan and susan6-Me for comparison.
Full size table

The UV-Vis-NIR spectrum (Fig. 2d) exhibits three prominent bands at 19,300 cm−1 (ε = 1180 M−1 cm−1), 15,400 cm−1 (ε = 1000 M−1 cm−1), and 11,800 cm−1 (ε = 190 M−1 cm−1) that are characteristic for {FeIII(μ-O)(μ-1,2-O2)FeIII} complexes29 with the 19,300 cm−1 band assigned to the μ-oxo→FeIII LMCT and the 15,400 cm−1 band to the μ-1,2-peroxo→FeIII LMCT involving π*π → t2g transitions11. This low LMCT energy demonstrates strong covalency of the FeIII–Operoxo bond representing the origin of the strong antiferromagnetic coupling.

We synthesized all four possible 18O/18O2-isotopomers as microcrystalline solids. The resonance Raman (rR) and FTIR spectra (Fig. 3) both reveal several 18O-sensitive vibrations, which allow their assignments to the {FeIII(μ-O)(μ-1,2-O2)FeIII} core (Table 2)22,29. Only slight changes are observed upon dissolution in CH3CN (Fig. S9). The 831 cm−1 band for the two 16O2-isotopomers is assigned to the ν(O–O) stretch by isotopic labeling (Δ(16O2−18O2) = 46 cm−1 consistent with a Hooke’s law calculation for a harmonic O–O vibration of 48 cm−1).

Fig. 3: Vibrational characterization of [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}](ClO4)2.
figure 3

Resonance Raman (bottom, 633 nm (15,800 cm−1) excitation) and FTIR spectra (top) measured on solids at room temperature for the four different iostopomers of [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}](ClO4)2 indicated by the representation of their central cores and the color scheme used.

Full size image
Table 2 Vibrational (resonance Raman and FTIR) frequencies (cm−1) and isotopic shifts together with some metrical parameters for [(susan6-Me){FeIII(μ-O)(μ-1,2-O2)FeIII}]2+ and comparisons to those for A.
Full size table

The ν(O–O) stretch of the crystallographically characterized [(6Me2BPP){FeIII(μ-O)(μ-1,2-O2)FeIII}(6Me2BPP)] (A) appears at higher energy at 847 cm−1 (Table 2)22. Higher ν(O–O) stretches were attributed22 to increasing ∢(Fe–O–O) angles11. However, ∢(Fe–O–O) is slightly smaller in A than in our complex (115° vs 117°). Interestingly, the lower ν(O–O) stretch correlates with a longer O–O distance (1.432(2) Å vs 1.411 Å) indicating a stronger dependence of ν(O–O) on d(O–O) than on ∢(Fe–O–O).

[(susan6-Me){FeIII(μ-O)(μ-1,2-O2)FeIII}]2+ shows no indication of decay for hours in CH3CN at −40 °C (Fig. S10). This stability provides the opportunity for the electro- and spectroelectrochemical investigation of a peroxo-diferric complex. [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}]2+ can be reversibly oxidized at E1/2ox = 0.55 V and irreversibly reduced at Epred = −1.28 V vs Fc+/Fc (Fig. 4a).

Fig. 4: Electrochemical/spectroelectrochemical characterization of [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}](ClO4)2 and oxidation to [(susan6-Me){FeIV(μ-O)(μ-1,2-O2)FeIII}]3+.
figure 4

a Cyclic voltammograms of [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}](ClO4)2 at −20 °C in CH3CN solution (0.2 M (NBu4)PF6) recorded at a GC working electrode. Scan rate 200 mV s−1 unless noted otherwise. Spectroelectrochemical measurements at −40 °C in CH3CN (0.2 mM with 0.1 M (NBu4)PF6) during b oxidation at 0.68 V vs Fc+/Fc and c re-reduction at 0.16 V vs Fc+/Fc. d Chemical oxidation with one equivalent (thia)ClO4 at −60 °C in CH3CN/CH2Cl2 (1:1) and re-reduction with excess NEt3. The chemically oxidized species (green) almost superimpose with the electrochemically oxidized species (red).

Full size image

Oxidation to [(susan6-Me){Fe(μ-O)(μ-1,2-O2)Fe}]3+

Coulometric oxidation of [(susan6-Me){FeIII(μ-O)(μ-1,2-O2)FeIII}]2+ at 0.68 V vs Fc+/Fc (1.18 C, 98% of one-electron) in CH3CN at −40 °C resulted in slight changes of the absorption features (Fig. 4b). Re-reduction at 0.16 V vs Fc+/Fc (0.92 C, 76% of one-electron) restored the initial UV-Vis spectrum to ~95% (Fig. 4c and Fig. S11). The lower charge necessary for re-reduction implies some chemical reduction during the coulometric experiments (1 h, Fig. S12).

Chemical oxidation with thianthrenium perchlorate ((thia)ClO4) generated the UV-Vis spectrum of [(susan6-Me){Fe(μ-O)2(μ-O2)Fe}]3+ (Fig. 4d) within 10 s. Addition of excess NEt3 as reductant regenerated the starting spectrum. This emphasizes again the chemical reversibility of the oxidation and indicates the conservation of the μ-oxo,μ-1,2-peroxo motive in the oxidized species.

Characterization of [(susan6-Me){Fe(μ-O)(μ-1,2-O2)Fe}]3+

The EPR spectrum of [(susan6-Me){Fe(μ-O)(μ-1,2-O2)Fe}]3+ (Fig. 5a, S13) shows a broad anisotropic St = 1/2 signal with g = (2.272, 2.152, 2.021). Mössbauer spectroscopy (vide infra) demonstrates that the iron ions remain high-spin ruling out an interpretation as FeIII l.s. species. The severe deviation of gav = 2.15 from 2.0023 and the large g-anisotropy demonstrate a strong contribution of orbital angular momentum and rule out a ligand-centered oxidation to a μ-oxo,μ-1,2-superoxo-FeIIIFeIII complex, disclosing a metal-centered oxidation to a μ-oxo,μ-1,2-peroxo-FeIVFeIII complex with antiferromagnetically coupled FeIV(S1 = 2) and FeIII(S2 = 5/2) ions. The FeIV h.s. S1 = 2 configuration results in g1 < 2.0 by spin-orbit coupling, while the FeIII h.s. is close to isotropic g2 ≈ 2.0046. Projection of the local spins onto the antiferromagnetic St = 1/2 ground state47 results in gav > 2.00 for the St = 1/2 as observed experimentally39. Despite significant efforts, we were not able to obtain resonance-enhanced Raman features of [(susan6-Me){Fe(μ-O)(μ-1,2-O2)Fe}]3+ by excitation at 647 (15,500), 568 (17,600), and 514 nm (19,500 cm−1) corresponding to the three absorption features of this oxidized complex.

Fig. 5: Spectroscopic characterization of the oxidized complex [(susan6-Me){FeIV(μ-O)(μ-O2)FeIII}]3+.
figure 5

a X-band EPR spectrum of chemically oxidized [(susan6-Me){FeIV(μ-O)(μ-O2)FeIII}]3+ in ≈0.4 mM CH3CN solution at 10 K (top, 9.63312 GHz, 80 μW power, 0.75 mT modulation) and its simulation (bottom; simulation red trace, green and blue traces subspectra) with the parameters provided in the text. b 57Fe Mössbauer spectra of 57Fe-enriched, chemically oxidized [(susan6-Me){FeIV(μ-O)(μ-O2)FeIII}]3+ in frozen CH3CN solution at 180 K. The solid lines are simulations with parameters provided in Table 1.

Full size image

The 180 K Mössbauer spectrum of 57Fe-labeled [(susan6-Me){Fe(μ-O)(μ-1,2-O2)Fe}]3+ (Fig. 5b) exhibits a 4-line spectrum suggesting the presence of two quadrupole doublets. Two different fit models are possible, but considerations explained in the Supplementary Information (Fig. S14 and S15) strongly favor the model with δ1 = 0.39 mm s−1/ΔEQ1 = −1.29 mm s−1 and δ2 = 0.27 mm s−1 / ΔEQ2 = +0.57 mm s−1 (Table 1). Interestingly, adding excess NEt3 as reductant to the oxidized [(susan6-Me){Fe(μ-O)(μ-1,2-O2)Fe}]3+ restores the Mössbauer spectrum of the starting complex [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}]2+ (Fig. S16) confirming the chemical reversibility.

The decrease of the isomer shift from 0.53 mm s−1 of the starting complex to 0.27 and 0.39 mm s−1 confirms a mainly metal-centered oxidation to a high-valent μ-1,2-peroxo complex with both FeIII ions involved resulting in a mixed-valence FeIVFeIII species. In the Robin-and-Day classification for mixed-valence systems48, class I implies no interaction (ruled out by the coupled St = 1/2 spin ground state) while class III stands for quantum-mechanically delocalized states. In class II systems, two different states exist that correspond roughly to the excess electron localized on the one or the other metal ion. Between these two states is an energy barrier and there can be temperature-dependent and light-induced mechanisms to transfer the excess electron from the reduced to the oxidized metal ion. In symmetric cases, the two states are energetically degenerate while the asymmetry observed here results in an energy difference between these two states (Fig. S17). The two quadrupole doublets can arise from one of these states populated exclusively up to 180 K or from an electron hopping between these two states at a rate faster than the Mössbauer timescale (10−7 s). To differentiate between these two possibilities, we recorded Mössbauer spectra at lower temperatures (Fig. S18). Unfortunately, the spectra broadened with decreasing temperature due to a relaxation process that is fast relative to the Mössbauer timescale at only 180 K. The origin cannot only arise from a decrease of the electron hopping rate but also from paramagnetic effects. DFT calculations (Supplementary Information) provided two localized configurations FeIV1FeIII2 and FeIII1FeIV2 that both reproduce the isomer shift decrease of both iron ions (Table 1) and hence confirm the assignment to class II. However, although these DFT calculations provided FeIV1FeIII2 being lower in energy by ≈820 cm−1, more advanced MO calculations are required to obtain further insight that are beyond this study.

Protonation to [(susan6-Me){FeIII(μ-O)(μ-1,2-OOH)FeIII}]3+

Treatment of a solution of [(susan6-Me){FeIII(μ-O)(μ-1,2-O2)FeIII}]2+ with HClO4 at −60 °C resulted in the loss of the 15,400 cm−1 band while the 11,800 and 19,300 cm−1 bands are only slightly affected (Fig. 6a, b). A Job plot analysis49,50 (Fig. 6c) provided a 1:1 stoichiometry for the reaction between the μ-1,2-peroxo complex and H+. Adding NEt3 as a base restores the initial spectrum (Fig. 6a) showing the reversibility of this protonation.

Fig. 6: Protonation of [(susan6-Me){FeIII(μ-O)(μ-1,2-O2)FeIII}]2+ and spectroscopic characterization of the protonated complex [(susan6-Me){FeIII(μ-O)(μ-1,2-OOH)FeIII}]3+.
figure 6

a UV-Vis spectroscopic characterization of protonation and deprotonation of [(susan6-Me){FeIII(μ-O)(μ-1,2-O2)FeIII}]2+ in CH3CN/CH2Cl2 (1:2) at −60 °C (c = 0.77 mM). b Selected time traces for the protonation/deprotonation in a. c Job-plot analysis at 15,400 cm−1 of the protonation in a. d 57Fe Mössbauer spectrum of a frozen solution of [(susan6-Me){FeIII(μ-O)(μ-1,2-OOH)FeIII}]3+ at 80 K generated by treating [(susan6-Me){FeIII(μ-O)(μ-1,2-O2)FeIII}]2+ with 1.5 equivalents HClO4 in CH3CN at −40 °C. The solid lines are simulations with parameters provided in Table 1.

Full size image

Characterization of [(susan6-Me){FeIII(μ-O)(μ-1,2-OOH)FeIII}]3+

The disappearance of the μ-1,2-peroxo → FeIII LMCT at 15400 cm−1, while the μ-oxo → FeIII LMCT around 19,000 cm−1 persists, is consistent with protonation of the μ-1,2-peroxo ligand. In contrast, protonation of complex A to the μ-hydroxo,μ-1,2-peroxo complex is accompanied by a shift of the μ-1,2-peroxo → FeIII LMCT from 17300 cm−1 to 15,500 cm−122. A similar shift was observed using a linear N4 ligand36. Moreover, the persistence of the weaker feature at 11800 cm−1 assigns this to a μ-oxo → FeIII LMCT. However, the absence of a μ-1,2-hydroperoxo→FeIII LMCT prohibits excitation for rR spectra at 647 nm (15,500 cm−1). Neither were resonance-enhanced vibrations observed by excitation at 568 (17,600) and 514 nm (19,500 cm−1) close to the μ-oxo→FeIII LMCT (Fig. 6a).

The Mössbauer spectrum provided two quadrupole doublets with δ1 = 0.49 / |ΔEQ | 1 = 2.48 mm s−1 and δ2 = 0.45 / |ΔEQ | 2 = 1.37 mm s−1 (Fig. 6d). Since the parent [(susan6-Me){FeIII(μ-O)(μ-1,2-O2)FeIII}]2+ exhibits only one quadrupole doublet despite the significant differences in its two coordination sites (vide supra) requires a different source of asymmetry for the protonated species to explain its two strongly differing quadrupole doublets. This rules out a μ-1,1-hydroperoxo-bridge, while a μ-1,2-hydroperoxo-bridge provides the different source of asymmetry as the protonated site becomes much less charge-donating. Although a terminally bound hydroperoxo-ligand would be in-line with the absence of a μ-1,2-peroxo→FeIII LMCT and two quadrupole doublets, the persistence of the μ-oxo → FeIII LMCTs around 19,000 and 11,800 cm−1 strongly favors a doubly-bridged structure of almost the same ∢(FeIII-(μ-O)-FeIII) angle hence ruling out an almost linear {FeIII(OOH)(μ-O)FeIIIX} core38,40 that is also inaccessible with the ligand susan6-Me (vide infra).

The formation of a μ-1,2-hydroperoxo-bridged complex is supported by DFT calculations (Supplementary Information). Geometry optimizations were achieved for three different tautomers with protonation of the μ-peroxo-O1 being 1200 cm−1 higher in energy. Although the μ-hydroxo species (protonation of the μ-oxo-O3) seems energetically feasible, its isomer shifts are even higher than for the starting peroxo complex (Table 1). In contrast, protonation at the μ-peroxo-O2 decreases the isomer shifts as experimentally observed. Thus, the μ-1,2-hydroperoxo ligand is most likely protonated at O2.

Further reactivity studies

The straightforward fast protonation of [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}]2+ with HClO4 demonstrates the nucleophilic character of the μ-1,2-peroxo ligand. On the other hand, the reaction of [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}]2+ with 2-phenylproponal as a typical substrate to evaluate the nucleophilic character of peroxo ligands51,52 is slow in CH3CN at −5 °C (Fig. S19). However, the formation of roughly one equivalent of acetophenone by performing this reaction on a preparative scale for 5 days supports the slow nucleophilic reactivity of the μ-1,2-peroxo ligand.

The electrophilic character was not only evaluated for [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}]2+ but also for the oxidized μ-1,2-peroxo-FeIVFeIII and the protonated μ-1,2-hydroperoxo-diferric species as both oxidation and protonation should increase the electrophilic character. This is already reflected in the different stabilities of these species (Fig. S10). As [(susan6-Me){FeIII(μ-O)(μ-1,2-O2)FeIII}]2+ shows no indication of decay for hours in CH3CN at −40 °C, oxidized [(susan6-Me){FeIV(μ-O)(μ-O2)FeIII}]3+ and protonated [(susan6-Me){FeIII(μ-O)(μ-1,2-OOH)FeIII}]3+ decay with half-lives of τ1/2 ≈ 90 min τ1/2 ≈ 11 min, respectively. Therefore, the clean characterization of the latter two species required lower temperatures of −60 °C and hence addition of a certain amount of CH2Cl2.

The electrophilic character of the three complexes were initially investigated using 9,10-dihydroanthracene (DHA) and PPh3 as typical substrates for HAT and OAT, respectively. The reactions with [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}]2+ were performed in CH3CN at −40 °C, while that with [(susan6-Me){FeIV(μ-O)(μ-O2)FeIII}]3+ and [(susan6-Me){FeIII(μ-O)(μ-1,2-OOH)FeIII}]3+at −60 °C in CH3CN/CH2Cl2 mixtures (vide supra). The parent [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}]2+ showed no reactivity towards DHA and PPh3 (Figs. S20, S21). The oxidized [(susan6-Me){FeIV(μ-O)(μ-O2)FeIII}]3+ also showed no reactivity towards DHA (Fig. S22), while the reaction with PPh3 resulted in the reoccurence of the UV-Vis signature of the parent [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}]2+ (Fig. S23). The reformation of the μ-1,2-peroxo→FeIII LMCT excludes an OAT reactivity between [(susan6-Me){FeIV(μ-O)(μ-O2)FeIII}]3+ and PPh3 but suggests an oxidation of PPh353 by [(susan6-Me){FeIV(μ-O)(μ-O2)FeIII}]3+. Analogous observations were made with the protonated [(susan6-Me){FeIII(μ-O)(μ-1,2-OOH)FeIII}]3+ that showed no reactivity with DHA (Fig. S24) and with PPh3 the partial recovery of the UV-Vis signature of the parent [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}]2+ (Fig. S25). Again, the reformation of the μ-1,2-peroxo→FeIII LMCT excludes an OAT reactivity between [(susan6-Me){FeIII(μ-O)(μ-1,2-OOH)FeIII}]3+ and PPh3. The partial recovery of [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}]2+ indicates a protonation equilibrium between PPh354 and [(susan6-Me){FeIII(μ-O)(μ-1,2-OOH)FeIII}]3+.

As it is not surprising35,55 that [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}]2+ exhibits no electrophilic reactivity against DHA, the non-reactivity of both oxidized [(susan6-Me){FeIV(μ-O)2(μ-O2)FeIII}]3+ and protonated [(susan6-Me){FeIII(μ-O)(μ-1,2-OOH)FeIII}]3+ is quite surprising. To further understand this non-reactivity, we determined the \({{{\rm{BDFE(OH)}}}}_{{{{{\rm{CH}}}}_3}{{{\rm{CN}}}}}\) of [(susan6-Me){FeIII(μ-O)(μ-1,2-OOH)FeIII}]3+. In this respect, we determined the pKa of [(susan6-Me){FeIII(μ-O)(μ-1,2-OOH)FeIII}]3+ in CH3CN that provided 9.5 ± 0.1 (Fig. S26). Using the typical square scheme (Fig. 7a) and the Bordwell relation Eq. (1)56,57,58

$${{{{{{\rm{BDFE}}}}}}({{{{{\rm{O}}}}}}-{{{{{\rm{H}}}}}})}_{{{{{{\rm{CH}}}}}}_3{{{{{\rm{CN}}}}}}}=1.37\,{pK}{{{{{\rm{a}}}}}}+{23.06\,{E}}^{0}{^{\prime}}+{52.6\,{{{{{\rm{kcal}}}}}}\,{{{{{\rm{mol}}}}}}}^{-1}$$
(1)

provided BDFE(O–H)CH3CN = 78 ± 2 kcal mol−1. This means that [(susan6-Me){FeIV(μ-O)(μ-O2)FeIII}]3+ should be capable as oxidant for HAT for substrates with a lower BDFE(X-H)\(_{{{{{\rm{CH}}}}_3}{{{\rm{CN}}}}}\). The BDE(C–H) of DHA is 76.3 kcal mol−159. However, the intrinsic difference between BDFE and BDE57, the temperature-dependence of BDFE especially for transition metal complexes, and the experimental error explain the non-reactivity of [(susan6-Me){FeIV(μ-O)(μ-O2)FeIII}]3+ with DHA. In this respect, TEMPOH should be a suitable HAT substrate with BDFE(O–H)\(_{{{{{\rm{CH}}}}_3}{{{\rm{CN}}}}}\) = 66.5 kcal mol−1 and BDE(O–H) = 70.6 kcal mol−157 especially for [(susan6-Me){FeIV(μ-O)(μ-O2)FeIII}]3+.

Fig. 7: Reactivity studies.
figure 7

a Square scheme showing the PCET thermochemistry of [(susan6-Me){FeIII(μ-O)(μ-1,2-OOH)FeIII}]3+ in CH3CN. b Molecular structure of the decay product of [(susan6-Me){FeIII(μ-O)(μ-1,2-O2)FeIII}]2+ including the ligand disorders (please see also Supplementary Fig. 1c): at Fe1 80% carboxylate and 20% hydroxide (shown with dotted lines), at Fe2 35% benzylalcoholato and 65% hydroxide (shown with dotted lines). c Space-filling model of [(susan6-Me){FeIII(μ-O)(μ-1,2-O2)FeIII}]2+ to illustrate the encapsulation of the peroxo ligand by a CH2 group (right Fe) and a 6-methyl group (left Fe).

Full size image

The parent [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}]2+ showed no reactivity with TEMPOH (Fig. S27), which is in-line that this one-electron reduced species [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}]2+ should have a driving force for HAT significantly lower than 78 kcal mol−1 of [(susan6-Me){FeIV(μ-O)2(μ-O2)FeIII}]3+. The reaction of protonated [(susan6-Me){FeIII(μ-O)(μ-1,2-OOH)FeIII}]3+ with TEMPOH resulted in the reoccurence of the UV-Vis signature of the parent [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}]2+ (Fig. S28). The reformation of the parent μ-1,2-peroxo-diferric complex excludes a HAT reactivity between [(susan6-Me){FeIII(μ-O)(μ-1,2-OOH)FeIII}]3+ and TEMPOH and suggests a protonation of TEMPOH60. The reaction of [(susan6-Me){FeIV(μ-O)(μ-O2)FeIII}]3+ with TEMPOH resulted in the reoccurence of the UV-Vis signature of the parent [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}]2+ (Fig. S29). This is in-line with HAT from TEMPOH to [(susan6-Me){FeIV(μ-O)(μ-O2)FeIII}]3+ resulting in [(susan6-Me){FeIII(μ-O)(μ-1,2-OOH)FeIII}]3+ that reacts again by protonation of excess TEMPOH to [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}]2+.

Thus, only the oxidized [(susan6-Me){FeIV(μ-O)(μ-O2)FeIII}]3+ is capable for HAT from TEMPOH corroborated by the BDFE while the parent [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}]2+ and the protonated [(susan6-Me){FeIII(μ-O)(μ-1,2-OOH)FeIII}]3+ does not exhibit enough electrophilic character. However, [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}]2+ that shows no decay at −40 °C exhibits a change in the UV-Vis spectra at room temperature with the formation of the typical signature of complexes with a {FeIIIX(μ-O)FeIIIX} core38 without the observation of intermediates accompanied by deposition of an inhomogenous solid and a few single-crystals. The crystallographic analysis provided the structure of the decay product (Fig. 7b) based on [(susan6-Me){FeIII(OH)(μ-O)FeIII(OH)}]2+61 with a disorder of the coordinated OH- ligands, which could be resolved to coordination of oxidized 6-methyl groups. At Fe1, only 20% is OH- while 80% is a carboxylate while at Fe2, 65% is OH− and 35% consists of a benzylalcoholato donor. NMR spectroscopy of the demetalated bulk decay product shows the formation of more than one product but a significant signal at 173.4 ppm for a benzoic acid group in the 13C NMR spectrum indicates that the hydroxylation of the 6-methyl group is not only a minor reaction path.

Discussion

To mimic dinuclear active sites of metalloenzymes, we have developed a dinucleating ligand system with varying terminal donors38,39. With the ligand susan, we obtained in straightforward reactions a series of μ-oxo-diferric complexes {FeIIIX(μ-O)FeIIIX} bearing anionic exogenous ligands X-37,39,45,61. The complex with hydroxides, [(susan){FeIII(OH)(μ-O)FeIII(OH)}]2+, catalyzes the oxidation of CH3OH with H2O2 to HCHO62. We could observe the μ-1,2-peroxo intermediate [(susan){FeIII(μ-O)(μ-1,2-O2)FeIII}]2+. However, the temperature-dependencies ruled out this μ-1,2-peroxo intermediate to be the active species indicating the conversion to a high-valent active species. This conversion is faster in the presence of a proton suggesting a transient μ-1,2-hydroperoxo species.

In contrast, formation of μ-oxo-diferric complexes with susan6-Me was not possible under identical aerobic conditions. Only the small ligand F- with H2O2 as oxidant allowed the isolation of [(susan6-Me){FeIIIF(μ-O)FeIIIF}]2+ 40. Its comparison to [(susan){FeIIIF(μ-O)FeIIIF}]2+ showed a steric repulsion between the 6-methyl group and the terminal F− ligand in cis-position, which explains the inaccessibility of susan6-Me complexes with larger terminal ligands as Cl− or OAc− that are easily accessible with susan. Thus, the μ-oxo-brigded core {FeIIIX(μ-O)FeIIIX} is not the thermodynamic sink for susan6-Me as it is for susan. Moreover, the steric repulsion of the 6-methyl group enforces longer Fe–N6-Me-py than Fe–Npy bonds63,64,65,66,67,68, and hence a lower electron donation. This results in an anodic shift of +250 mV making FeIV less accessible with susan6-Me than with susan.

In this respect, we thought that susan6-Me should be able to stabilize a μ-1,2-peroxo complex {FeIII(μ-O)(μ-1,2-O2)FeIII} that is with susan only a reactive intermediate decaying via a high-valent FeIV species to its thermodynamic sink {FeIIIX(μ-O)FeIIIX}. In constrast, with susan6-Me not only this thermodynamic driving force is absent but FeIV is also less accessible. Indeed, we could present here the synthesis and characterization of the stable μ-1,2-peroxo complex [(susan6-Me){FeIII(μ-O)(μ-1,2-O2)FeIII}]2+. Inspection of the space-filling model (Fig. 7c) shows that the μ-1,2-peroxo ligand is even further stabilized by a better encapsulation with the 6-methyl group of susan6-Me (left peroxo-oxygen atom in Fig. 7c) that would be absent with susan. This ligand encapsulation also explains the slower nucleophilic reactivity of the μ-1,2-peroxo ligand for the organic substrate 2-phenylpropanal than for the small H+.

Although susan6-Me is less suited for stabilization of FeIV than susan, the principal accessibility of FeIV with susan6-Me is demonstrated by the reversible oxidation to [(susan6-Me){FeIV(μ-O)(μ-1,2-O2)FeIII}]3+, which is stabilized by the additional highly covalent μ-1,2-peroxo ligand. It is interesting to note, that this high-valent μ-1,2-peroxo species stores one oxidation-equivalent more than intermediate Q of sMMO. Comparing the FeIVFeIII/FeIIIFeIII redox potential of E1/2 = 0.55 V to 0.41 V vs Fc+/Fc for the analogous redox couple of [(tpa6-Me){FeIII(μ-O)2FeIII}(tpa6-Me)]2+ 69 shows only a slightly lower electron-donating character of the μ-1,2-peroxo-bridge than a μ-oxo-bridge.

We could further demonstrate the reversible protonation to the μ-1,2-hydroperoxo-diferric complex [(susan6-Me){FeIII(μ-O)(μ-1,2-OOH)FeIII}]3+. Generally, protonation of a Fe-coordinated peroxo ligand is regarded to enhance its reactivity, e. g. protonation of the cis-μ-1,2-peroxo intermediate P of sMMO was proposed to promote the conversion to intermediate Q11. The relatively high stability of the μ-1,2-hydroperoxo complex [(susan6-Me){FeIII(μ-O)(μ-1,2-OOH)FeIII}]3+ (no decay at −60 °C, τ1/2 ≈ 11 min at −40 °C) is thus remarkable and must owe its origin to a low stabilization of the FeIV conversion product by susan6-Me.

In contrast to [(susan6-Me){FeIII(μ-O)(μ-1,2-O2)FeIII}]2+, the μ-1,2-peroxo complex A is protonated at the μ-oxo-bridge forming a {FeIII(μ-1,2-O2)(μ-OH)FeIII} species22 indicating different nucleophilicities. The nucleophilic character of a ligand should increase with less electron donation to the FeIII ions, i.e. less covalent, longer bonds. But for A, the FeIII-μ-Ooxo bonds are shorter than for [(susan6-Me){FeIII(μ-O)(μ-1,2-O2)FeIII}]2+ (1.72/1.74 Å vs 1.82/1.89 Å), whereas the situation is reversed for the FeIII-μ-Operoxo bonds (2.07/2.10 Å vs 1.88/1.93 Å). This structural argumentation is in contrast to the experimentally determined protonation sites. However, the disorder of the μ-oxo/μ-1,2-peroxo ligands in A questions the significance of this comparison. Moreover, as protonation should occur at a pπ orbital and a Fe–O bond consists of σ- and π-bonding, a pure structural analysis does not need to provide the answer for the reactivity.

Thus, spectroscopic markers might provide a better correlation to the nucleophilic character of the peroxo group than structural parameters. Solomon and coworkers proposed to extract the donor strength of a given ligand from the integrated absorption intensities of all CT transitions associated with this ligand70. Here, the π-charge donation from the peroxo π*π donor orbital into the Fe 3dπ acceptor orbitals should be extractable from the prominent μ-1,2-peroxo → Fe LMCTs. Complex A exhibits this μ-1,2-peroxo→FeIII LMCT with ε = 1500 M−1 cm−1 and a much less intense μ-oxo→FeIII LMCT. In contrast, the μ-oxo → FeIII LMCT in [(susan6-Me){FeIII(μ-O)(μ-1,2-O2)FeIII}]2+ is more intense (ε = 1180 M−1 cm−1) than the μ-peroxo → FeIII LMCT (ε = 1000 M−1 cm−1). Note that also the integrated absorption intensity is smaller in the susan6-Me complex than in A for the μ-1,2-peroxo→FeIII LMCT indicating - without the intention of a quantitative analysis - less charge-donation and hence more nucleophilic character of the μ-1,2-peroxo ligand. This UV-Vis spectroscopic argumentation is supported by a comparison of the vibrational signature in the rR spectra. Interestingly, the strongest difference is observed for the νs(Fe–O2–Fe), which are at 465 and 448 cm−1 for A and [(susan6-Me){FeIII(μ-O)(μ-1,2-O2)FeIII}]2+, respectively, indicative for less covalent FeIII-μ-Operoxo bonds and hence a higher nucleophilicity of the μ-1,2-peroxo ligand in [(susan6-Me){FeIII(μ-O)(μ-1,2-O2)FeIII}]2+.

The study of the electrophilic reactivity demonstrated only a low electrophilic character of the parent μ-1,2-peroxo-FeIIIFeIII complex that is not unexpected for such complexes35,55. Interestingly, also protonation to the μ-1,2-hydroperoxo-FeIIIFeIII species turned out to be not sufficient to increase the electrophilic character for HAT with substrates of weak to modest BDE (TEMPOH and DHA). Only the oxidized μ-1,2-peroxo-FeIVFeIII species reacts with the relatively weak substrate TEMPOH. The determination of the pKa = 9.5 ± 0.1 and the bond dissociation free energy BDFE(O–H)\(_{{{{{\rm{CH}}}}_3}{{{\rm{CN}}}}}\) = 78 ± 2 kcal mol−1 of the protonated μ-1,2-hydroperoxo-FeIIIFeIII species quantifies this low electrophilic character even of the oxidized μ-1,2-peroxo-FeIVFeIII species. In this respect, the intramolecular C–H activation of preorganized 6-methyl pyridine groups to benzylalcoholato and carboxylato donors by the parent μ-1,2-peroxo-FeIIIFeIII complex is remarkable. Considering that this complex does not react with TEMPOH (BDE(O–H) = 70.6 kcal mol−1) and using the BDE(C–H) = 90 kcal mol−159 of the methyl group of toluene as an approximation for the BDE(C–H) of the 6-methyl groups of the coordinated pyridines, HAT should not occur via the bridging μ-1,2-peroxo-ligand. This indicates that this intramolecular reaction requires the conversion of the μ-1,2-peroxo-diferric core to a more reactive high-valent species as already postulated for the CH3OH oxidation62 of the analogous μ-1,2-peroxo-diferric complex of susan (vide supra).

[(susan6-Me){FeIII(μ-O)(μ-1,2-OOH)FeIII}]3+ is an μ-1,2-hydroperoxo model complex and provides spectroscopic signatures for the assignment of postulated hydroperoxo intermediates in diiron enzymes: Upon protonation, the prominent μ-1,2-peroxo→FeIII LMCT around 14,000–16,000 cm−1 disappears and the isomer shift decreases. The presence of an μ-oxo-bridge is indicated by the typical μ-oxo→FeIII LMCTs around 12,000 and 19,000 cm−1 and large values of | ΔEQ | ≥ 1.3 mm s−1. The best signature to differentiate between a μ-1,1-hydroperoxo and a μ-1,2-hydroperoxo is the appearance of two strongly differing quadrupole doublets for the latter due to the strongly differing donation abilities of the two μ-1,2-hydroperoxo-oxygen atoms.

The protonation of [(susan6-Me){FeIII(μ-O)(μ-1,2-O2)FeIII}]2+ to [(susan6-Me){FeIII(μ-O)(μ-1,2-OOH)FeIII}]3+ reflect the UV-Vis-NIR and Mössbauer spectroscopic differences between P and the more reactive P′ intermediates in non-heme diiron enzymes2,15,16,20. For the latter, a μ-1,2-hydroperoxo structure has been suggested, which is thus strongly supported by the results of this study.

The diferrous form of AurF exhibits one quadrupole doublet demonstrating structurally rather similar iron places12. Reaction with O2 provides a diferric species that lacks the typical μ-1,2-peroxo → FeIII LMCT around 14,000 cm−1. The Mössbauer spectrum contains two quite different quadrupole doublets with δ1 = 0.54, ΔEQ1 = 0.66 mm s−1 and δ2 = 0.61, ΔEQ2 = 0.35 mm s−112,13. This coupled with the low values of | ΔEQ | and a lack of the μ-1,2-peroxo → FeIII LMCT leads us to suggest the formulation of the P′-type intermediate in AurF as a {FeIII(μ-1,2-hydroperoxo)FeIII} core without a μ-oxo-bridge, which also supports a recent proposal15.

Methods

Synthesis of [(susan6-Me){FeII(μ-OH)2FeII}](ClO4)2·H2O

A solution of susan6-Me (595 mg, 1.00 mmol) in MeOH (25 mL) was added to a solution of Fe(ClO4)2·6H2O (726 mg, 2.00 mmol, 2.0 equiv) in MeOH (20 mL). This yellow solution was stirred at room temperature for 10 min followed by an addition of a 25% aqueous solution of ammonia (0.16 mL, 2.1 mmol, 2.1 equiv) resulting in a slight intensity increase of the yellow color. Yellow crystals of [(susan6-Me){FeII(μ-OH)2FeII}](ClO4)2·MeOH precipitated at 0 °C, which were filtered off, washed three times with water, a small amount of cold MeOH, three times with Et2O, and dried under reduced pressure. Yield: 704 mg (7.50 × 10−4 mol, 75%). IR (KBr): \(\widetilde{v}\)/cm−1 = 3642 w, 3083 w, 2976 w, 2856 m, 1603 s, 1576 m, 1462 s, 1429 m, 1093 vs, 1009 m, 956 m, 847 m, 780 s, 723 w, 623 s, 490 m. ESI-MS (+) (CH2Cl2): m/z = 361.2 [(susan6-Me){Fe(μ-O)Fe}]2+. Anal. Found: C 44.90, H 5.76, N 11.51. Calcd. for [(susan6-Me){Fe(μ-OH)2Fe}](ClO4)2·H2O C36H54N8Cl2Fe2O11: C 45.16, H 5.68, N 11.70.

Synthesis of [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}](ClO4)2·2H2O

A solution of [(susan6-Me){FeII(μ-OH)2FeII}](ClO4)2 (63 mg, 6.7 × 10−5 mol) in CH2Cl2 (2 mL) was added to a solution of 25% aqueous ammonia (20 μL, 2.6 × 10−4 mol, 3.9 equiv) in MeOH (6 mL) at −15 °C. Pouring an O2 stream of approximate 0.1 L min−1 through this yellow solution for 1.5 h at −15 °C results in immediately color change to dark green followed by precipitation of a black microcrystalline solid, which was filtered off, washed three times with cold water, a small amount of cold MeOH, three times with Et2O, and dried under reduced pressure. Yield: 52 mg (5.3 × 10−5 mol, 78%). Anal. Found: C 43.34, H 5.20, N 11.02. Calcd. for [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}](ClO4)2·2H2O C36H54N8Cl2Fe2O13: C 43.30, H 5.55, N 11.22. For single-crystal X-ray diffraction and magnetic measurements, the sample was recrystallized by slow evaporation of a filtered (0.2 μm pores size PTFE filter) solution of [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}](ClO4)2·2H2O (200 mg, 2.02 × 10−4 mol) in MeCN (20 mL) at −30 °C. The resulting black crystals of [(susan6-Me){FeIII(μ-O)(μ-O2)FeIII}](ClO4)2·0.85 MeCN·0.7 H2O were filtered off, washed three times with cold water, three times with cold MeOH, three times with Et2O, and dried under reduced pressure. Yield: 52 mg (5.2 × 10−5 mol, 26%). IR (KBr): \(\widetilde{v}\)/cm−1 = 3079 w, 2999 w, 2918 w, 2881 w, 2811w, 1605s, 1573 w, 1466 s, 1454 s, 1090 vs, 1002 m, 955 m, 933 w, 842 m, 832 w, 788 s, 686 m, 624 s, 500 w, 447 w. ESI-MS (+) (CH2Cl2/MeCN): m/z = 377.1 [(susan6-Me){Fe(μ-O)(μ-O2)Fe}]2+. Anal. Found: C 44.66, H 5.21, N 12.27. Calcd. for [(susan6-Me){Fe(μ-O)(μ-O2)Fe}](ClO4)2·0.8MeCN·H2O C37.6H54.4N8.8Cl2Fe2O12: C 44.97, H 5.46, N 12.27.