Preparation of the Fe-ZSM-5 catalysts
The active sites in Fe-ZSM-5 are positively charged extra-framework Fe-oxo species bound to negatively charged framework Al sites, while the initial Fe content in the zeolite framework determines the maximum amount of extra-framework Fe produced via calcination. These two facts suggest that increasing the amount of framework Al and Fe in the initial Fe-ZSM-5 would be beneficial for generating more extra-framework Fe in the final calcined H-form catalyst. Compared to the conventional synthesis of ZSM-5 using TPA+, our proposed template-free synthesis allows more Al and Fe to be incorporated into ZSM-5 because the consequent high-density negative charges on the framework can be balanced easily by the smaller inorganic cations (Na+) in the synthetic system35–37. Moreover, Fe-ZSM-5 synthesized using the template-free method can generate a higher proportion of desired extra-framework Fe species during calcination compared to conventional Fe-ZSM-5.
Under the synthetic conditions employed in this study, the Si/Al ratio of Fe-ZSM-5 synthesized with TPA+ was > 2538,39, whereas the Si/Al ratio of Fe-ZSM-5 synthesized without TPA+ (i.e., the template-free approach) was in the range of 12−2040. For both systems, the maximum achievable amount of Fe in the zeolite framework is ~ 10 mol.% that of the Al (i.e., Al/Fe≈10). As observed using ultraviolet–visible (UV–Vis) spectroscopy, attempts to introduce more Fe to the zeolite by adding more Fe precursors resulted in the generation of iron oxide clusters or particles, (Supplementary Fig. 1).
Two representative Fe-ZSM-5 samples were selected for detailed analysis: one synthesized via the conventional method with TPA+ (Si/Al = 33.5; Al/Fe = 10.8) and the other via the template-free method (Si/Al = 17.9; Al/Fe = 12.5) (Fe-Z5-C and Fe-Z5-TF, respectively). Before catalytic testing for CH4 oxidation, the as-synthesized samples underwent ion exchange with ammonium followed by calcination to convert them to H-form zeolites (Fe-HZ5-C and Fe-HZ5-TF, respectively). Table 1 lists the chemical composition of these four samples. Powder X-ray diffraction (XRD) revealed that the four samples were pure phases with an MFI structure (Fig. 1a). No XRD peaks associated with iron oxides were observed, even for the H-forms, indicating that Fe was highly dispersed in the zeolite framework or micropores without forming bulk crystalline phases.
Table 1
┃Properties of various Fe-ZSM-5 catalysts.
Samples | Fe loading (wt%)a | Si/Ala | Al/Fea | Fe speciesb | Extra-framework Fe (wt%) | Al pairs (%)c | Al pairs/Fe2 (or Fe1)d (molar ratio) |
ISO FeFW | ESSR FeFW | FeEFW |
Fe-Z5-C | 0.24 | 34.5 | 11.1 | 6.6% | 93.4% | 0 | 0 | 9.8 | / |
Fe-HZ5-C | 0.25 | 33.5 | 10.8 | 0 | 52.1% | 47.9% | 0.12 | 9.8 | 1.25 |
Fe-Z5-TF | 0.41 | 17.1 | 12.5 | 64.6% | 35.4% | 0 | 0 | 28.8 | / |
Fe-HZ5-TF | 0.39 | 17.9 | 12.5 | 0 | 16.2% | 83.8% | 0.33 | 28.8 | 1.78 |
a Determined using ICP-OES |
b Determined using 57Fe Mössbauer spectroscopy; ISO FeFW, ESSR FeFW, and FeEFW refer to isolated framework Fe, framework Fe with the enhanced spin-spin relaxation effect, and extra-framework Fe species, respectively |
c Proportion of Al pairs in total Al species derived from the ICP-OES results for Co-exchanged zeolites |
d Fe2 and Fe1 represent binuclear Fe and mononuclear Fe, respectively |
Identification of active Fe species in the Fe-ZSM-5 catalysts
Diffuse reflectance UV–Vis spectroscopy is frequently used to analyze Fe species in zeolites because the absorption band associated with ligand-to-metal charge transfer (CT) is sensitive to the coordination geometry and environment, with framework Fe having adsorption bands in the range of 200–250 nm, compared to 250–350 nm for isolated and oligomeric extra-framework Fe confined in zeolite channels, 350–450 nm for larger Fe clusters, and > 450 nm for bulk iron oxide particles. Consistent with the literature, Fe-Z5-C exhibited two discernible CT bands in the UV–Vis spectrum (Fig. 1b), centered at 211 and 245 nm, which are characteristic of isomorphously incorporated, tetrahedrally coordinated Fe3+ ions within the zeolite framework, corresponding to t1 → t2 and t1 → e transitions, respectively33,41,42. In contrast, Fe-Z5-TF showed only one intense CT band at 248 nm (Fig. 1b), suggesting an unusual framework Fe coordination geometry or environment. In previous research on template-free synthesized Fe-ZSM-543,44, the same single-band absorption was observed and assigned to the framework Fe without questioning why it differed from the typical double-band absorption. In another study, the double-band and single-band absorption was attributed to Fe3+ in a perfect and distorted tetrahedral (Td) coordination environment, respectively33.
Compared to Fe-Z5-C, Fe-HZ5-C exhibited a significantly broader spectrum, in which the two bands at 211 and 245 nm remained discernable but with reduced intensity, and the long tail covered a broad range up to > 500 nm (Fig. 1b). These observations are consistent with results reported elsewhere22,27−29, indicating that the partial migration of Fe atoms yielded various extra-framework Fe species, mainly oligomeric species and trace amounts of clusters and particles. Interestingly, the sample prepared using the template-free method in the present study exhibited distinctive behavior. The UV–Vis spectrum of Fe-HZ5-TF was similar in shape to Fe-Z5-TF, but shifted towards higher wavelengths by ~ 30 nm, with no absorption at > 400 nm (Fig. 1b), which suggests that almost all Fe migrates out of the zeolite framework during calcination to form homogeneous extra-framework Fe species confined in the micropores.
In the X-ray absorption near-edge structure (XANES) spectra for Fe-Z5-TF and Fe-Z5-C (Fig. 1c), there was a pre-edge peak at 7114 eV, which is indicative of Fe3+ in the tetrahedral zeolite framework, corresponding to the transition from 1s to 3d-like levels (i.e., the strong mixing of 3d and 4p metal orbitals due to electric-quadrupole coupling)45,46. This peak was almost absent for Fe3+ with centrosymmetric octahedral coordination (e.g., in Fe2O3, which was used as a reference). Compared to Fe-Z5-C, the pre-edge peak of Fe-HZ5-C was less intense but still clearly visible, which corresponds to the partial transformation of framework Fe to octahedrally coordinated extra-framework Fe. In the XANES spectrum of Fe-HZ5-TF, the pre-edge peak intensity of Fe-HZ5-TF was very low, indicating the complete migration of framework Fe. Therefore, the conclusions drawn from XANES were consistent with those from the UV–Vis analysis.
57Fe Mössbauer spectroscopy was conducted to quantitatively analyze Fe species in different states. The spectra for Fe-Z5-C and Fe-Z5-TF were deconvoluted into a singlet component with an isomer shift (IS) of ∼0.25 mm/s and a broad magnetic relaxation component with an IS of ~ 0.36 mm/s (Figs. 1d and e), both of which were related to tetrahedrally coordinated high-spin Fe3+ in the zeolite framework47. The singlet component corresponded to isolated framework Fe atoms that were far apart from each other (denoted as ISO FeFW), while the broad magnetic relaxation component corresponded to framework Fe atoms that were close enough to exhibit the enhanced spin-spin relaxation effect (denoted as ESSR FeFW)48,49. Quantitative analysis revealed that most of the Fe (93.4%) in Fe-Z5-C was ESSR FeFW, whereas most of the Fe (64.6%) in Fe-Z5-TF was ISO FeFW (Table 1), meaning that Fe-Z5-TF had a higher dispersion degree of framework Fe than Fe-Z5-C. The high proportion of ESSR FeFW in Fe-Z5-C can be attributed to the preferential localization of TPA+ at the channel intersections in ZSM-5, resulting in the enrichment of trivalent Fe therein50. In contrast, the distribution of Fe in Fe-Z5-TF is not restrained by the organic template and thus is more dispersed and homogeneous.
In the 57Fe Mössbauer spectra for Fe-HZ5-C and Fe-HZ5-TF, a doublet component (IS = 0.32 mm/s and quadrupole splitting = 0.81 mm/s) was observed (Figs. 1f and g), which corresponded to extra-framework octahedral Fe3+ species (denoted as FeEFW). The spectra could be deconvoluted with satisfactory accuracy based on the coexistence of FeEFW and ESSR FeFW (Figs. 1f and g). The results indicated that during the transformation of Fe-Z5-C to Fe-HZ5-C, 47.9% of the framework Fe was converted to extra-framework Fe; in comparison, this value was as high as 83.8% for the transformation of Fe-Z5-TF to Fe-HZ5-TF (Table l).
Collectively, these spectroscopy characterization results identified two significant differences between the Fe-ZSM-5 samples synthesized with and without the organic template. First, the framework Fe in Fe-Z5-C and Fe-Z5-TF exhibited different coordination geometries (i.e., Td vs. distorted Td) and distribution (i.e., ESSR FeFW vs. ISO FeFW as the primary species). Second, compared to Fe-Z5-C, a significantly higher proportion of Fe in Fe-Z5-TF migrates out of the zeolite framework to form extra-framework Fe species when converted to the H-form. It is reasonable to speculate that the first observation was the cause of the second. Compared to Fe-Z5-C, Fe-Z5-TF had an inherently higher Fe content and higher Fe migration, and consequently, it yielded approximately three times more extra-framework Fe in the H-form catalyst (Table 1).
Figure 2a presents the k2-weighted Fourier-transformed extended X-ray absorption fine structure (EXAFS) spectra for Fe-HZ5-C and Fe-HZ5-TF using Fe foil and Fe2O3 as references. Unlike the reference materials, which have multiple peaks associated with Fe−Fe interactions, Fe-HZ5-C and Fe-HZ5-TF exhibited only one intense peak at 1.4 Å (uncorrected for the phase shift) corresponding to Fe−O scattering, suggesting that Fe was dispersed atomically without the formation of bulk oxides. The EXAFS spectra were analyzed further using wavelet transforms (WTs) based on Morlet wavelets with an optimal resolution (Fig. 2b). A comparison of the WT contour plots for the Fe foil and Fe2O3 revealed that the intensity maxima at ~ 4.0 and ~ 8.0 Å−1 correspond to Fe−O and Fe−Fe paths, respectively (Fig. 2b). The WT contour plot for Fe-HZ5-TF exhibited no sign of the Fe−Fe path but only a concentrated intensity distribution with a maximum at 4.0 Å−1, which corresponds to the Fe−O path according to the reference materials. The absence of a Fe−Fe path suggests that the extra-framework Fe species in Fe-HZ5-TF are predominantly mononuclear Fe. In contrast, Fe-HZ5-C had a diffuse intensity distribution along the k-vector direction, centered at 7.0 Å−1. This diffuse intensity can be attributed to a Fe−O path perturbed by a coexisting Fe−Fe path because its R + α value (~ 1.4 Å) is in the range of the Fe−O bond distance, suggesting that Fe-HZ5-C contains multinuclear extra-framework Fe species.
Least-squares EXAFS fitting was employed to obtain the average fine structure parameters, including the interatomic distance (ID), coordination number (CN), and Debye-Waller factor (σ2) (Figs. 2c,d and Table 2). For Fe-HZ5-C, given the presence of framework Fe and extra-framework multinuclear Fe, three paths (Fe−O, Fe−Al, and Fe−Fe) were included simultaneously for the fitting. The results revealed a dominant Fe−O path (ID: 1.89 Å) with a CN of 4.7, a Fe−Al path (ID: 2.49 Å) with a CN of 1.0, and a Fe−Fe path (ID: 3.77 Å) with a CN of 0.7 (Fig. 2c and Table 2). Given that the amount of framework tetrahedral Fe and extra-framework Fe in Fe-HZ5-C was very similar (52:48), the average Fe−O CN (≈ 5) confirmed that the extra-framework Fe species were octahedrally coordinated, while the average Fe−Fe CN (≈ 0.5) indicated that the extra-framework Fe species were predominantly binuclear Fe (because binuclear Fe has a theoretical Fe−Fe CN of 1).
Table 2
┃Fe K-edge EXAFS curve fitting parameters for various Fe-ZSM-5 samples*.
Samples | Path | R (Å) | CN | δ2 (*100Å2) | ΔE0 (eV) | R-factor |
Fe-Z5-C | Fe-O | 1.85 (0.02) | 4.0 (0.8) | 0.3 (0.2) | 4 (3) | 0.015 |
Fe-HZ5-C | Fe-O | 1.89 (0.03) | 4.7 (0.4) | 0.3 (0.1) | 7 (2) | 0.013 |
Fe-Al | 2.49 (0.09) | 1.0 (1.4) | 0.9 (0.3) |
Fe-Fe | 3.77 (0.05) | 0.7 (0.4) | 0.9& |
Fe-Z5-TF | Fe-O | 1.84 (0.01) | 4.1 (0.7) | 0.4 (0.1) | 5 (2) | 0.009 |
Fe-HZ5-TF | Fe-O1 | 1.86 (0.09) | 1.9# | 0.3& | 2 (7) | 0.011 |
Fe-O2 | 2.03 (0.08) | 4.1 (1.4) | 0.3& |
Fe-HZ5-TF | Fe-O1 | 1.90 (0.03) | 2.0& | 0.5 (0.3) | 4 (3) | 0.015 |
Fe-O2 | 2.08 (0.02) | 4.0Δ | 0.3 (0.3) |
Fe-Al | 2.49 (0.10) | 1.6 (0.4) | 0.6 (0.4) |
Fe-Fe | 3.79 (0.05) | 0.3& | 0.4 (0.1) |
*R, distance between absorber and backscatter atoms; CN, coordination number; δ2, Debye-Waller factor to account for thermal and structural disorder; ΔE0, inner potential correction; R-factor indicates the goodness of the fit. # Set value; & Restrained value; Δ Define: CN(Fe-O2) = 6-CN(Fe-O1). |
Because Fe-HZ5-TF had only mononuclear extra-framework Fe and a small amount of framework Fe, the fitting of its EXAFS spectrum did not involve Fe−Al and Fe−Fe paths. Compared to Fe-HZ5-C, the raw spectrum for Fe-HZ5-TF had a markedly broader Fe−O peak (Fig. 2a), suggesting that there may be multiple Fe−O paths in the first shell of the Fe center. The best-fitting result (R-factor: 0.011) confirmed the presence of two independent Fe−O paths, one with an ID of 1.86 Å and a CN of 1.9 and the other with an ID of 2.05 Å and a CN of 4.1 (Fig. 2d and Table 2). The total CN of ∼6 was consistent with the conclusion that most of the Fe in Fe-HZ5-TF was extra-framework Fe with octahedral coordination. The attempt to include a Fe−Fe path in the EXAFS fitting for Fe-HZ5-TF resulted in a higher R-factor of 0.015 (Supplementary Fig. 2; Table 2). The resulting average CN for Fe−Fe was as low as 0.3, further confirming that Fe-HZ5-TF contained only trace amounts of multinuclear Fe species, if any.
These characterization results raise the question of why extra-framework Fe takes a binuclear form in Fe-HZ5-C but a mononuclear form in Fe-HZ5-TF. This difference can be explained by the higher Al content in Fe-HZ5-TF than in Fe-HZ5-C, because more Al means more Al pairs and thus more anchoring sites to promote the dispersion of extra-framework Fe in a more isolated form. This speculation was verified using an established method to measure the number of Al pairs. Two Fe-free ZSM-5 samples were prepared using TPA and the template-free method (denoted as Z5-C and Z5-TF, respectively) for this to avoid interference from Fe. Briefly, Co2+ was introduced into the calcined zeolites via ion exchange. Due to the one-to-one correspondence between Co2+ and Al pairs, the number of Al pairs can be estimated from the Co uptake measured using inductively coupled plasma mass spectrometry51,52, while the distribution of Al pairs in the zeolite framework can be inferred from the UV–Vis spectra for dehydrated Co-exchanged zeolites (Supplementary Fig. 3). The results showed that 28.8% of the framework Al in Z5-TF forms Al pairs, compared to only 9.8% for Z5-C, which is consistent with expectations. In both samples, Al pairs are predominantly located at the intersections of the straight and sinusoidal channels (i.e., the β sites). The data summarized in Table 1 and Supplementary Fig. 4 revealed two interesting facts. First, the absolute number of Al pairs in Z5-TF was approximately five times higher than in Z5-C, which perfectly matches the number ratio of mononuclear Fe in Fe-HZ5-TF to binuclear Fe in Fe-HZ5-C. Second, the numbers of Al pairs are sufficient to anchor the corresponding active Fe centers in both systems.
Based on the CN and primary locations determined from the above experimental results, three models for the octahedral mononuclear [(H2O)n=1,2,3-Fe(III)-OH]2+ complex anchored by an Al pair at the β site were constructed and optimized by the DFT method (Supplementary Figs. 5a-c). Of the three optimized models, [(H2O)2-Fe(III)-OH]2+ bound to three zeolite framework oxygen (Of) atoms (Fig. 2e; Supplementary Table 1) exhibited the closest agreement with the EXAFS fitting results in terms of the Fe−O bond lengths (Table 2). The possibility of stabilizing mononuclear Fe with a single Al site rather than an Al pair was also investigated using [(H2O)2-Fe(III)-(OH)2]+ complexes with different ligand configurations at the octahedral apical sites (Supplementary Fig. 5). It was found that the Fe−O bond lengths obtained from the most energetically stable configurations (Supplementary Fig. 5f and Supplementary Table 1) did not conform to the EXAFS results (Table 2). Therefore, it was concluded that the mononuclear Fe species in Fe-HZ5-TF was in the form of [(H2O)2-Fe(III)-OH]2+ bound to three Of atoms (Fig. 2f). Figure 2g displays the binuclear Fe model previously described in the literature for conventional Fe-ZSM-5 as a comparison26,28.
Catalytic performance
The prepared Fe-HZ5-C and Fe-HZ5-TF catalysts were evaluated for the selective oxidation of CH4. The reactions were conducted in a 100-mL autoclave reactor containing 10 mL of an aqueous H2O2 solution (0.5 M) and 27 mg of catalyst at 75°C and 30.5 bar CH4 for 25 min. All tested catalysts exhibited a similar selectivity, with HCOOH as the predominant product (> 85%) and other minor products including CH3OH, CO2, CH3OOH, and HOCH2OOH. Therefore, we focus the following discussion on catalyst activity and use the overall C1 oxygenate productivity (µmol) as the main criterion for comparison.
We first tested a series of Fe-HZ5-C samples with a fixed Fe content (0.25 wt%) but different Si/Al ratios (approximately 200, 100, and 35). The sample with the highest Al content (i.e., Si/Al ≈ 35) exhibited the highest activity (Fig. 3a), which can be attributed to the favorable migration of framework Fe and the dispersion of extra-framework Fe associated with high Al levels27. Therefore, the Si/Al ratio was fixed at 35 (i.e., close to the highest Al content available for Fe-Z5-C) in the subsequent tests to examine the effect of Fe loading on the activity of Fe-HZ5-C. To this end, two additional Fe-HZ5-C samples (Si/Al ≈ 35) were prepared, which contained less and more Fe (0.15 wt% and 0.40 wt%, respectively) compared with the standard Fe-HZ5-C sample (0.25 wt%). As shown in Fig. 3a, both samples produced lower C1 oxygenate levels than the standard Fe-HZ5-C. In particular, the C1 productivity of Fe-HZ5-C with 0.15 wt%, 0.25 wt%, and 0.40 wt% Fe was 277.5, 427.8, and 403.7 µmol, corresponding to turnover frequency (TOF) of 918.4, 849.5, and 501.0 h− 1, respectively. These results indicated that at low Fe loads, the productivity increased almost in proportion with the Fe content, whereas there was an optimal Fe loading beyond which further increases in Fe led to a decrease in productivity. More systematic analysis revealed that the optimal Fe loading was dependent primarily on the Al content in the zeolite framework (i.e., Al/Fe ≈ 10). This can be attributed to the fact that active (i.e., binuclear) Fe species require Al pairs to stabilize them, while excess Fe relative to framework Al leads to the formation of clusters and particles that adversely affect the reaction through the nonselective decomposition of H2O227,28. UV–Vis spectra confirmed that the Fe-HZ5-C sample with 0.4 wt% Fe had more pronounced absorption at > 350 nm compared to the samples with lower Fe content (Supplementary Fig. 6).
Based on these analyses, the optimal composition of Fe-HZ5-C was determined to be Si/Al = 33.5 and Al/Fe = 10.8 (i.e., a Fe loading of 0.25 wt%). Under typical reaction conditions, the optimized Fe-HZ5-C exhibited the highest achievable yield at 38.0 mmol of C1 products per gram of the catalyst per hour (mmolgcat−1h− 1) (Fig. 3d). When the reaction temperature was reduced to 50°C, the optimized Fe-HZ5-C had a C1 oxygenate yield of 20.2 mmolgcat−1h− 1 (Fig. 3d), which was similar to the previously reported highest value for conventional Fe-ZSM-5 under similar conditions (Supplementary Table 2).
Because the Si/Al ratio of Fe-HZ5-TF cannot vary greatly, we only investigated the effect of the Fe loading on its activity at a fixed Si/Al ratio of ∼18. When the Fe loading was low (0.15−0.40 wt%), the oxygenate productivity was approximately proportional to the Fe content, while the TOF was almost constant (Fig. 3b). Increasing the Fe loading to 0.80 wt% resulted in lower oxygenate productivity and a significant drop in the TOF (Fig. 3b). These observations were consistent with the Fe-HZ5-C system and were explained and verified in a similar manner (Supplementary Fig. 7). On the other hand, the observed catalytic activity for the optimal Fe-HZ5-TF (0.40 wt% Fe) was approximately three times higher than that for the optimal Fe-HZ5-C (0.25 wt% Fe). Specifically, Fe-HZ5-TF (0.40 wt% Fe) produced 1195.9 µmol of oxygenates (Fig. 3b), corresponding to 106.3 mmolgcat−1h− 1 (Fig. 3d). When the reaction temperature was lowered to 50°C, the oxygenate productivity was 667.0 µmol (59.3 mmolgcat−1h− 1), which was approximately three times higher than that of Fe-HZ5-C under the same conditions (Fig. 3d). Given that Fe-HZ5-TF (0.40 wt% Fe) had approximately three times more active Fe species than Fe-HZ5-C (0.25 wt% Fe) (Table 1), the TOF based on active Fe species was similar for both catalysts (Supplementary Fig. 8).
Figure 3d compares the oxygenate yields obtained from various catalysts under different reaction conditions, showing that Fe-HZ5-TF has the highest activity of the reported catalysts (see Supplementary Table 2 for more detail). The high conversion capacity per unit mass of Fe-HZ5-TF can be attributed to the high number of active Fe sites, which were maximized using the template-free synthesis method. Although the Fe loading on ZSM-5 can be easily increased to higher levels using post-synthesis methods, such as wet impregnation, ion exchange, and solid-state ion exchange, the Fe-ZSM-5 catalysts prepared in this way contained many undesirable clusters and particles and exhibited lower activity than Fe-HZ5-TF (Supplementary Fig. 9). In addition to the high activity, Fe-HZ5-TF demonstrated good recyclability without a noticeable drop in productivity over five consecutive reaction runs (Fig. 3c), indicating that the extra-framework mononuclear Fe species were firmly confined to the zeolite channels.
Methane conversion at mononuclear Fe sites
Conventional Fe-ZSM-5 is generally considered to activate CH4 molecules through a synergistic effect at binuclear Fe sites. Because mononuclear Fe can also act as an active site for CH4 oxidation in the presence of H2O2, as demonstrated in this study, it is crucial to understand how CH4 is activated and converted by mononuclear Fe.
To gain insights into this process, we characterized two samples using powder XRD, UV-Vis, and X-ray absorption spectra: Fe-HZ5-TF treated with 0.5 M H2O2 (H2O2-treated Fe-HZ5-TF) and Fe-HZ5-TF recovered after the CH4 oxidation reaction (spent Fe-HZ5-TF). The results indicated that, following the addition of H2O2 or during the catalytic reaction, the mononuclear Fe species remained isolated and did not agglomerate to form FexOy clusters or particles (Supplementary Figs. 10). However, compared to pristine Fe-HZ5-TF, the H2O2-treated and spent Fe-HZ5-TF both exhibited higher white line intensities and lower pre-edge peak intensities in the XANES spectra (Fig. 4a). These spectral changes suggested that both the valence state and CN of the Fe center increased during contact with H2O2. Furthermore, H2O2-treated and spent Fe-HZ5-TF both demonstrated enhanced peak amplitudes for the Fe−O shell compared to pristine Fe-HZ5-TF (Fig. 4b), also suggesting that the CN of Fe increased (see the fitting results in Supplementary Fig. 11 and Supplementary Table 3).
Electron paramagnetic resonance (EPR) spectroscopy showed that adding Fe-HZ5-TF could promote the generation of •OH radicals in the H2O2 solution (Fig. 4c). EPR spectroscopy also detected •OH and •CH3 radicals in the solution after the CH4 selective oxidation reaction (Fig. 4c). In situ diffuse reflectance-infrared Fourier-transform spectroscopy (DRIFTS) was also conducted to identify the CH4 oxidation reaction intermediates for Fe-HZ5-TF at 75°C. As the reaction proceeded, three peaks of increasing intensity appeared at 3365, 3227, and 1624 cm− 1, which could be assigned to •OH, •OOH, and •OH2 (likely the result of •OH extracting •H from CH4 or −Fe-OH), respectively (Fig. 4d)53,54. Ex situ EPR spectroscopic analysis did not detect •OOH and •OH2 radicals, possibly because of their low concentrations and short lifetimes. Moreover, characteristic peaks for HCOH at 2720 and 1741 cm− 1 were observed in DRIFTS (Fig. 4d), indicating the presence of HCOH as an intermediate product during the reaction.
Overall, CH4 activation requires high-valence Fe, while the subsequent conversion of CH4 is based on radical reactions. This is consistent with a previous finding that Fe(IV) = O, which is surrounded by weak field ligands to have a high spin state for Fe, is reactive for the abstraction of H from CH426,55. Accordingly, the following CH4 activation/conversion process at mononuclear Fe sites is proposed. First, the initial mononuclear Fe(III) sites are oxidized by H2O2, generating Fe(IV) = O groups and •OH radicals:
[(H2O)2-Fe(III)-OH]2+ + H2O2 \(\to\) [(H2O)3-Fe(IV) = O]2+ + •OH
CH4 is then adsorbed onto the resulting Fe(IV) = O, followed by the homolytic cleavage of one of the C–H bonds to form Fe(III)-OH and •CH3 radicals. These •CH3 radicals subsequently react with •OOH and •OH radicals in solution to form CH3OOH and CH3OH. The resulting Fe(III)-OH can be re-oxidized to Fe(IV) = O by H2O2 to restart this process (Fig. 4e). The formation of other oxygenate products can be explained by the same C–H bond activation combined with radical reactions and subsequent deep oxidation (Fig. 4e). This process does not require the simultaneous participation of two Fe centers in close proximity.
The evolution of various products was monitored during CH4 selective oxidation (Supplementary Fig. 12). The yields of CH3OOH and HOCH2OOH were the highest at the beginning of the reaction and then decreased monotonically over time. In contrast, the yields of HCOOH and CO2 increased continuously during the reaction, while the yield of CH3OH increased for the first 25 min of the reaction, but extending the reaction time further caused it to decrease. These results indicate the presence of peroxygenate-to-oxygenate conversion and sequential oxidation during the reaction.