Introduction

High-valent iron species are highly reactive molecules involved in numerous oxidative processes of synthetic and biological relevance1,2. In particular, Fe(V) intermediates have been proposed as the oxidation agents in key organic synthesis reactions, such as C–H, C=C and arene oxidations, and in energy-related transformations, including water oxidation3. Moreover, iron-dependent enzymes such as cytochrome P450 and Rieske oxygenases presumably use formal Fe(V) intermediates to oxidize inert substrates, including alkanes or arenes. Cytochrome P450 has been shown to use an oxoiron(IV) porphyrin cation radical intermediate termed compound I (cpd I) in C–H oxidation reactions4, whereas Rieske oxygenases may use a non-detectable oxoiron(V) intermediate in the syn-dihydroxylation of arenes and in metabolic C–H oxidations, although no direct evidence has been reported thus far5,6,7. Oxoiron(V) complexes are extremely challenging targets for synthetic inorganic chemistry because of their high reactivity. Accordingly, no crystal structure is available, and spectroscopically characterized examples remain exceedingly rare8,9,10,11,12,13.

Inspired by iron oxygenases, chemists have intensively exploited iron coordination complexes as catalysts, also thanks to the availability of this metal14. Complexes with tetradentate aminopyridine ligands are particularly interesting because they can use hydrogen peroxide to catalyze enzyme-like stereoretentive C–H and C=C oxidations (Fig. 1a)15,16,17. Extensive mechanistic studies based on product analysis, isotopic labeling and computations have indirectly shown that these complexes operate via FeV(O)(X) (X = alkyl carboxylate or OH) reactive species18,19,20,21.

Fig. 1
figure 1

Mechanistic frame of iron-catalyzed oxidations with tetradentate aminopyridine ligands. a Iron catalysts based on tetradentate aminopyridine ligands that catalyze stereoretentive C–H and C=C oxidations. b Mechanistic frame for the generation of Fe(V) species in the presence (right) or absence (left) of carboxylic acids. c Schematic diagram of [FeII(CF3SO3)2(5tips3tpa)] (1) and its catalytic oxidation activity

Full size image

Carboxylic acids assist the heterolytic cleavage of the O–O bond (Fig. 1b), forming a reactive FeV(O)(O2CR) (R = alkyl) intermediate (IIIb in Fig. 1b) that epoxidizes olefins and hydroxylates alkanes20,22. Quite recently, this FeV(O)(O2CR) intermediate was accumulated using a robust ligand frame, enabling its spectroscopic and chemical characterization and, therefore, providing a solid foundation for the mechanistic proposal13,23,24.

In the absence of carboxylic acids, O–O cleavage may be assisted by a water molecule, forming an FeV(O)(OH) (IIIa in Fig. 1b) intermediate that hydroxylates alkanes and engages in the syn-dihydroxylations of olefins25. These FeV(O)(OH) species have also been proposed to oxidize the water molecule under low pH condition3. Moreover, no FeV(O)(OH) species accumulate in solution, consistently with their high reactivity.

The complex [FeII(CF3SO3)2(5tips3tpa)], 1 (Fig. 1c, 5tips3tpa = tris(5-(triisopropyl)silyl-2-methylpyridyl)amine is a remarkable example of a catalyst operating through FeV(O)(OH) intermediates. It efficiently catalyzes the syn-dihydroxylation of various olefins with H2O226,27 and is, thus, a sustainable alternative to traditional, Os- and Ru-based syn-dihydroxylating agents28,29. In addition, this catalyst shows outstanding selectivity properties; for example, olefins are highly chemoselectively syn-dihydroxylated while epoxidation is largely minimized. Moreover, electron-deficient olefins, and arenes, unreactive to Os-based reagents, are instantaneously dihydroxylated, thus indicating the involvement of extraordinarily powerful oxidizing species.

Before the present study, FeV(O)(OH) species have been detected only by mass spectrometry (MS), and their formulation was derived from experiments using isotopically labeled reagents (H218O and H218O2)30,31. Although their reactivity has been inferred from MS analysis of catalytic reaction mixtures, their spectroscopic characterization and direct assessment of their reactivity has not been performed yet.

Herein, we spectroscopically characterized the proposed [FeV(O)(OH)(5tips3tpa)]2+ reactive intermediate in the gas phase by helium tagging infrared photodissociation (IRPD) spectroscopy32. We conclusively identify the terminal FeV=O and FeV–OH stretching vibrations of the Fe(O)(OH) unit. Furthermore, we confirm that [FeV(O)(OH)(5tips3tpa)]2+ hydroxylates C–H bonds in a rebound mechanism and performs the syn-dihydroxylation of alkenes and arenes. These reactions, previously described in enzymes and bioinspired oxidation catalysts, have only been previously understood based on product analysis and computational methods5,6,19,30,33,34,35.

Thus, the present study reports the experimental characterization of the FeV(O)(OH) species and demonstrates its chemical competence in bioinspired reactions, particularly in reactions relevant to Rieske oxygenases.

Results

Generation and ion-spectroscopy characterization of intermediates

The reaction of 1 (0.4 mM) with H2O2 (10 equiv.) in acetonitrile at −40 °C, monitored by ultraviolet-visible (UV–vis) spectroscopy, produces a metastable purple species 2 (λmax = 544 nm, ε = 1300 M−1 cm−1) (Fig. 2 and Supplementary Fig. 2). After 2 was formed in acetonitrile solution, the reaction mixture was analyzed by electrospray ionization mass spectrometry (ESI-MS). Two peaks at m/z = 444 and 424 stand out in the ESI-MS spectrum (Supplementary Fig. 1). The former corresponds to the expected dicationic species [FeIII(OOH)(CH3CN)(5tips3tpa)]2+ (22+)25, whereas the latter can be tentatively formulated as either [FeIII(OOH)(5tips3tpa)]2+ (2a2+) or [Fe(O)(OH)(5tips3tpa)]2+ (32+), wherein the O–O bond has been broken. Using helium-tagging IRPD spectroscopy32 we were able to measure IR spectra of the mass-selected ions with m/z 424 generated by electrospray ionization from the solution of 2 and ascertain that these indeed correspond to [FeV(O)(OH)(5tips3tpa)]2+ (32+).

Fig. 2
figure 2

Generation of the iron(V) intermediate 3. Schematic diagram of the formation of ferric hydroperoxide species 2 in solution and subsequent transfer of this species to the gas phase where the FeV species 3 is generated

Full size image

We measured the IRPD spectrum of the ions (corresponding to the iron(V) intermediate 32+) generated from the reaction mixture of 1 and H216O2. The spectrum was assigned by comparison with the spectra of isotopically labeled ions resulting from the oxidation of 1 with H218O2, H216O18O, and D216O2 (the mass-selected complexes contained the 56Fe isotope if not mentioned otherwise, Fig. 3). We also analyzed the spectrum of naturally occurring 54Fe-labeled ions (32+(54Fe), Supplementary Fig. 3). The IRPD spectrum of [FeV(16O,16OH)(5tips3tpa)]2+ (32+, m/z 424) (Fig. 3a, black) shows bands at 827 cm−1 and 638 cm−1 that shift to 797 cm−1 and 616 cm−1, respectively, upon double 18O labeling [FeV(18O,18OH)(5tips3tpa)]2+, (32+(18O18O), m/z 426, Fig. 3a). These bands can be interpreted as either Fe=O and Fe–OH stretching vibrations9,13,36,37 or as O–O stretching and O–O–H bending vibrations (see the comparison with the theoretically predicted IR spectra in Fig. 3c)38. The frequencies of the characteristic n(Fe–O) and n(O–O) bands of [FeIII(OOH)(tpa)(S)]2+ (S = solvent) have been determined by resonance Raman to be 626 cm−1 and 789 cm−1,38, respectively, which may be considered in reasonable agreement with those observed for 3. To differentiate {Fe(O)(OH)} and {Fe(OOH)} binding motifs, we measured the IRPD spectrum of singly 18O labeled ions [FeV(16/18O,18/16OH)(5tips3tpa)]2+ (32+(16O18O), m/z 425). If the 827 cm−1 band would correspond to the O–O stretching mode, the band should redshift (the bond would be always labeled by 18O). Conversely, if the band would correspond to the Fe=O stretch, only half the band should disappear (the Fe=O bond is 18O labeled in only 50% ions). Indeed, the second variant is observed in the experiment (Fig. 3b). This result allows us to assign the IRPD spectra to the [FeV(O)(OH)(5tips3tpa)]2+ intermediates. Fully consistent with this interpretation, the vibrational spectrum of 32+(54Fe) (m/z = 423) shows that both bands are blueshifted by 3 cm−1, as expected for spectral shifts of Fe–O bonds (Supplementary Fig. 3). Interestingly, Hooke’s Law analysis of a Fe–O oscillator predicts shifts of 3 cm−1 for both vibrations

Fig. 3
figure 3

IRPD spectrum of the ions generated from the reaction mixture of 1 with H2O2 and their prediction. a IRPD spectra of 32+ (black trace) and 32+(18O18O) (orange trace). b IRPD spectra of 32+ and 32+(18O16O). c B3LYP-D3/def2TZVP predictions of the IR spectra for the 32+ and 2a2+ complexes. d IRPD spectra of 32+ and 32+(2H). e B3LYP-D3/def2TZVP predictions of the IR spectra of 32+ and 2a2+ complexes

Full size image

The vibrational features of 32+ agree well with the DFT spectra of the [FeV(O)(OH)(5tips3tpa)]2+ complex with the S = 3/2 ground state as predicted at the B3LYP-D3/def2TZVP level (432+, Fig. 3c, e). In addition to reproducing the experimentally determined energy and isotopic shifts of Fe=O and Fe–OH stretching vibrations, the computations predict a distinctive δ-OH vibration, sensitive to deuteration, and a blueshift of the Fe=O stretch in 32+(2H). The blueshift of the Fe=O stretch in 32+(2H) (Fig. 3d) is caused by coupling between the ν(Fe=O) and δ(OH) vibrations; this coupling is also evident when we calculate the IR spectrum of 32+ labeled only at the Fe=O oxygen, wherein the δ(OH) vibration redshifts (Supplementary Fig. 4e). We also considered the doublet state, but the calculations predict that this state is 12.3 kcal mol−1 higher in energy than the quartet state. Its predicted IR spectrum is quite similar to that of the quartet state complex, except for a higher frequency of the Fe=O stretching vibration (Supplementary Fig. 4b).

The computed spectroscopic features of the [FeIII(OO(H/D)(5tips3tpa)]2+ species were also considered. Interestingly, the computed O–O stretching frequency is basically insensitive to deuteration of the hydroperoxide ligand, in line with the resonance Raman analysis of [FeIII(OO(H/D)(N4Py)]2+ 39, (N4Py = (1,1-di(pyridin-2-yl)-N,N-bis(pyridin-2-ylmethyl)methanamine)) and in sharp contrast with the shift of [FeV(O)(OD)(5tips3tpa)]2+. In conclusion, the vibrational spectra provide compelling evidence that 32+ must be formulated as [FeV(O)(OH)(5tips3tpa)]2+, wherein the iron center is in the quartet state. This formulation actually reproduces the structure and spin ground state predicted by Siegbahn and Que for the parent [FeV(O)(OH)(tpa)]2+, based on DFT calculations40.

The isotopic composition of 32+ shows that both oxygen atoms originate from a single H2O2 molecule, thus indicating that its formation is not assisted by a water molecule (Fig. 1b). In contrast, isotopic analyzes of diol products formed in catalytic olefin oxidation reactions conducted in acetonitrile show that 32+ is formed in solution with the assistance of a water molecule, that is, 32+ contains one oxygen atom from H2O2 and another from water26. Therefore, 32+ must be formed by different mechanisms in solution and in gas phase. DFT calculations suggest that 32+ is more than 6 kcal mol−1 lower in energy than 2a2+ in the gas phase (see Supplementary Table 1). Therefore, elimination of acetonitrile from 22+ in the gas phase may likely lead directly to the rearranged product 32+.

The electronic spectrum of 32+ could be also determined by photodissociation spectroscopy (Supplementary Fig. 5). The spectrum is characterized by two absorption bands at 440 nm and 530 nm, corresponding to a charge transfer transition, which is also well reproduced by TD-DFT calculations of 432+.

The complex 432+ is one of the few spectroscopically characterized FeV=O complexes thus far (Table 1) and the first example of an FeV(O)(OH) species. These species are frequently postulated as catalytic intermediates in iron-catalyzed biomimetic oxidations16 and in the catalytic cycle of Rieske oxygenases6. In addition, 32+ is also the single experimentally characterized example of an Fe(V) complex with a postulated S = 3/2 spin state. The energy of the Fe=O stretch in the doublet state complexes range from 798 cm−1 to 862 cm−1 8,13; thus, the energy of the Fe–O bond of 432+ falls within the same range. Moreover, FeIV=O complexes show Fe–O vibrations from 798 cm−1 to 850 cm−1 2. This suggests that the strength of the Fe=O bond is unaffected by electron removal from FeIV=O presumably because the electron is removed from non-bonding orbitals, with respect to the Fe=O bond (electronic configuration of S = 1 FeIV=O: dxy2 dxz1 dyz1; of S = 3/2 FeV=O: dxy1 dxz1 dyz1; Fe=O bond is along z-axis). Interestingly, the related iron(V) complex with oxo and acyloxo ligands had the doublet ground state (Table 1)13. According to our calculations, replacing of OH by CH3COO in 32+ would result in the spin change of the ground state to S = ½. This change is also associated with an energetic preference for the closed {LFeIII(OOCOCH3)} form over the open, high valent {LFeV(O)(OCOCH3)} in the gas phase (see in the Supplementary Discussion and Supplementary Fig. 8)42.

Table 1 Structural and spectroscopic features of previously described FeV(O) complexes
Full size table

Reactivity studies

After establishing the structure of the iron(V) intermediate, we probed its reactivity with a series of substrates in collisional experiments in the gas phase43,44. We studied reactions of mass-selected ions where each ion interacted with only one molecule of a given reactant R. The detected ionic products are thus formed from a well-defined reactant complex [432+∙R] without involvement of any additional molecules such as water or another reactant molecule. The reactions of 432+ proceed efficiently, attesting its high reactivity. Remarkably, when the ion corresponding to [FeIII(OOH)(CH3CN)(5tips3tpa)]2+ (22+, m/z = 444.3) was tested in similar experiments, no reactivity was observed (Supplementary Fig. 7). This lack of reactivity of the hydroperoxide species against organic molecules reproduces well the rather sluggish oxidant character of these species in solution25,45.

The reaction of 432+(2H) with cyclohexene (Fig. 4a) yields two products. The dominant product (m/z 465.3, 84%), an adduct between cyclohexene and 432+(2H), corresponds to a dihydroxylation reaction (Supplementary Table 2). The second ion product (m/z = 416.3, 16%) results from the oxygen transfer from 432+(2H) to olefin, most likely in an epoxidation reaction. Thus, these findings are in line with experimental results under catalytic oxidation conditions, thus showing that 1 catalyzes olefin oxidation, largely favoring syn-dihydroxylation over epoxidation reactions26.

Fig. 4
figure 4

Ion-molecule reactivity of 32+ and 32+(2H) in the gas phase. a 0.1 mTorr of cyclohexene, b 0.1 mTorr of 1,3-cyclohexadiene, c < 0.1 mTorr of naphthalene, and d 0.2 mTorr benzene (asterisks indicate impurities from previous measurements). All reactions were measured at nominally zero-collision energy determined from the retarding potential analysis

Full size image

The reaction of 432+ with 1,3-cyclohexadiene also dominantly leads to the adduct resulting from the dihydroxylation of an olefinic site alongside with the oxygen transfer reaction (Supplementary Table 2). The addition reaction is accompanied in approximately 10% by subsequent water elimination probably driven by restoring the conjugated double bond system.

We rationalized the reaction pathways based on experiments with deuterated complex ([FeV(O)(OD)(5tips3tpa)]2+ (432+ (2H), Fig. 4b). The deuterium atom allows us to follow the subsequent fragmentation pathways. The initially formed adduct complex (the dihydroxylation product) is long-lived and therefore allows for hydrogen scrambling (complex is isolated in the gas phase and does not interact with any other molecules/ions)46,47. Note that this complex is isolated in the gas phase, therefore contains the energy released by the exothermic interaction between the reactants and cannot dissipate this energy by interaction with other molecules. The subsequent dehydration of the adduct thus features as elimination of HDO or H2O in the 3:2 ratio.

Most interestingly, in the formal oxygen transfer reaction, 432+(2H) yielded not only the expected product ([FeIII(OD)(5tips3tpa)]2+, m/z = 416.3), but also a product in which the OD group was replaced by OH (i.e. [FeIII(OH)(5tips3tpa)]2+, m/z = 415.8). This observation can be explained by a two-step rebound mechanism. In the first step, hydrogen atom abstraction generates [FeIV(OD)(OH)(5tips3tpa)]2+ and a short-lived carbon-centered radical. The radical can be then rebound with either OH or OD from the Fe(OH)(OD) unit, to finally form the corresponding alcohol. The observation of rebound mechanism contrasts with the previously-reported reactivity of iron(IV)-oxo complexes in the gas phase, where the observed oxygen transfer is exclusively due to the epoxidation of C=C double bonds36. The opening of the C–H activation pathway in the reaction with 1,3-cyclohexadiene increases the overall abundance of the formal oxygen atom transfer channel over cyclohexene (see Supplementary Table 2). This path occurs in the reaction with 1,3-cyclohexadiene but not with cyclohexene because the latter has a stronger C–H bond (BDEC–H = 74.3 vs 87.0 kcal mol−1)48. Exactly the same product pattern, identified in all previous reaction channels, was also observed in the reaction of 432+(2H) with 1,4-cyclohexadiene (Supplementary Fig. 6).

Lastly, we investigated reactions of 432+ with aromatic compounds. Reactions with benzene and naphthalene yield addition products followed by water elimination. Furthermore, only in the case of naphthalene, we also observed a single electron transfer reaction, yielding the naphthalene radical cation and a product of single-electron reduced 432+ 49,50. Because gas phase reactions only occur when they are exothermic, the electron affinity of 432+ must be higher than ionization energy of naphthalene (8.14 eV.)51. In turn, this value is higher than the electron affinities of oxoiron(IV) porphyrin cation radicals (cpdI models), which are always lower than 7.5 eV, thus indicating that 432+ is a stronger one-electron oxidant than oxoiron(IV) porphyrin cation radicals50,52. The addition/water elimination reaction is similar to reactions with cyclohexadiene reactants, but the reaction fully shifts towards final water elimination. The final product regains aromaticity, thereby likely driving the dehydration step kinetically and thermodynamically. This is particularly relevant in the gas phase because the initially formed syn-dihydroxylated product cannot be stabilized by interaction with solvent molecules. On the contrary, we observed the catalytic syn-dihydroxylation of naphthalene by complex 1 and H2O2 in solution (see supporting information) as also previously observed in reactions with the [Fe(CH3CN)2(tpa)]2+ complex53. We also probed the reaction of 32+ with D6-benzene, and we observed addition followed by HDO elimination with almost 100% selectivity (Fig. 4d). This reaction is highly interesting because these substrates are inert against high-valent Ru and Os oxides and, therefore, show the uniquely powerful oxidation ability of 432+.

Discussion

The current study describes the vibrational and electronic spectroscopic characterization of FeV(O)(OH) species with a key role in biomimetic oxidations. These FeV(O)(OH) species have long been proposed to be ultimately responsible for a wide array of oxidations, including enzymatic reactions. However, the inability to accumulate them in solution has thus far prevented their spectroscopic and chemical characterization. Herein, we used gas phase ion spectroscopy methods to address this problem. The electronic and vibrational spectra of these species were finally determined, providing experimental data to unambiguously determine its atomic and electronic structure. Gas-phase reactivity analysis of these well-defined species showed their competence in C–H hydroxylation and syn-dihydroxylation of olefins and arenes. Overall, the data highlights that the particular architecture of the FeV(O)(OH) species, featuring two reactive ligands in cis-relative positions, translates into singular reactivity properties, unattainable with hemes. For example, high-valent heme iron-oxo complexes consistently epoxidize olefins4. However, the current study shows that FeV(O)(OH) species readily engage in syn-dihydroxylation rather than in epoxidation reactions and, most remarkably, react with arenes. Furthermore, gas phase studies on the hydroxylation of C–H bonds provide direct experimental evidence of a stepwise rebound mechanism, wherein rebound can occur with two different ligands at the Fe center. This behavior differs from that observed in reactions with previously described synthetic FeIV=O complexes, which engage in HAT followed by diffusion of the carbon-centered radical36,54,55. Furthermore, this study provides experimental evidence of the rebound of the carbon-centered radical with the two cis-labile ligands at the iron center, which is not possible for hemes because HAT and rebound can only occur at the same oxygen atom4. Conversely, in non-heme iron-dependent enzymes and model complexes56, the presence of labile sites adjacent to the Fe=O moiety enables the transfer of the incipient hydroxyl ligand, or of ligands adjacent to the ferryl. For example, halides, azides and nitrates are transferred in non-heme halogenases57,58.

Finally, the report shows that gas-phase ion characterization can address questions relevant to catalysis and enzymology, related to highly reactive intermediates, currently unanswerable by other methods.

Methods

Generation of intermediate 2

Initially, 15 µL of a 0.22 M H2O2 solution in acetonitrile (diluted from 30% in aqueous solution) were directly added over a 2 mL acetonitrile solution of catalyst 1 (0.2 mM). The resulting mixture was cooled to −40 °C with a CH3CN/N2 (l) bath. At this point, the characteristic purple color of 2 was observed. The solution was kept at −33 °C during the measurements in a two-stage Peltier cooler device.

Gas phase reactivity

Mass-spectrometric measurements were performed in a TSQ 7000 quadrupole–octopole–quadrupole spectrometer43,44. The ions were transferred to the gas phase using an ESI ion source. Ionization conditions were typically: 6 kV spray voltage, 0 V capillary voltage, 90 V tube lens voltage, 150 °C capillary temperature, 30 psi sheath gas pressure, 300 l h−1 auxiliary gas flow. The spray voltage was connected directly to the solution in the vial with a stainless steel wire. The solution was kept at −33 °C and was pumped into the ESI source through a 30-cm-long fused-silica capillary with a 100-µm internal diameter by ~1 psi overpressure of nitrogen gas in the vial with the solution. The ions of interest were mass-selected by the first quadrupole and transferred to the octopole equipped with a collision cell; collision gas pressure was determined using a baratron. The collision energy was set to nominally zero, as determined by retarding potential analysis. The products were extracted from the octopole to the second quadrupole, mass-analyzed and detected with a Daly-type detector.

Helium-tagging infrared/visible photodissociation (IRPD/visPD) spectroscopy

IRPD/visPD spectra were measured with the ISORI instrument based on the TSQ 7000 platform32,59,60. The ions were generated and mass selected exactly as above. The mass-selected ions were transferred via a quadrupole bender and octopole to a cryogenic ion trap operating at 3 K. The ions were trapped with 250 µs helium pulse and formed weakly bound complexes with helium. The trapped ions were irradiated by IR light from an OPO/OPA system or by visible light from a continuum laser wavelength-filtered by acusto-optic tunable filter. After irradiation, all ions were ejected from the trap, mass-analyzed in a second quadrupole and counted by a Daly-type detector. IRPD spectra are plotted as wavenumber-dependent attenuation of the number of helium complexes (1−Ni (ν)/Ni0). The total number of the helium complexes (Ni0) was obtained in alternative cycles with blocked photon beam. The visPD spectra were corrected by dividing the attenuation by the laser power.

DFT calculations

DFT calculations were performed with Gaussian 0961 at B3LYP-D3/def2TZVP level. All structures were fully optimized and characterized by frequency calculations. The frequencies in IR spectra were scaled by 0.99. Reported energies include zero-point vibrational energy corrections; the molecular coordinates are provided in the Supplementary Note 1 and the relative energies in Supplementary Table 3.