Subnanometric MoO_x clusters limit overoxidation during photocatalytic CH₄ conversion to oxygenates over TiO₂

Abstract

Direct photocatalytic oxidation of methane to high-value-added oxygenated products remains a great challenge due to the unavoidable overoxidation of target products. Here, we report an efficient and highly selective TiO₂ photocatalyst anchored with subnanometric MoO_x clusters for photocatalytic methane oxidation to organic oxygenates by oxygen. A high organic oxygenates yield of 3.8 mmol/g with nearly 100% selectivity was achieved after 2 h of light irradiation, resulting in a 13.3% apparent quantum yield at 365 nm. Mechanistic studies reveal a photocatalytic cycle for methane oxidation on the MoO_x anchored TiO₂, which not only largely inhibits the formation of hydroxyl and superoxide radicals and the overoxidation of oxygenate products but also facilitates the activation of the first carbon-hydrogen bond of methane. This work would promote the rational design of efficient non-noble metal catalysts for direct conversion of methane to high-value-added oxygenates.

Introduction

Direct conversion of methane (CH₄) to high-value-added oxygenates (such as CH₃OOH, CH₃OH, HCHO, HCOOH, and CH₃COOH) with molecular oxygen (O₂) is one of the most ideal approaches to realize the optimization and utilization of methane and reduce the dependence on crude oil^1,2,3,4,5. However, due to the intrinsic inertness of CH₄, high temperatures and pressures are normally required to activate C–H bonds, which not only greatly decrease selectivity of the organic products but also give rise to operational risks and environmental problems^6,7. Photocatalysis is a potential way to drive CH₄ oxidation by utilizing solar energy instead of thermal energy to overcome thermodynamic barriers, which is drawing keen attentions due to its safe, green and economic advantages^8,9,10,11. Upon excitation by photons, a series of active oxygen-containing radicals (such as •OH and O₂^•−) formed during photocatalytic CH₄ conversion with O₂ can activate the C–H bond under mild conditions^12,13,14. However, the highly active radicals are much easier to oxidize oxygenates than CH₄^15,16,17, and thus it is a great challenge to simultaneously optimize the activity and selectivity for the photocatalytic reaction, unless expensive oxidants (such as H₂O₂ and N₂O) instead of O₂ are utilized^{18,19,20,21,22}.

Choosing appropriate cocatalysts plays a key role in photocatalytic CH₄ oxidation on semiconductor photocatalysts, as they not only achieve efficient separation of photogenerated charge carriers to promote methane activation, but also regulate surface catalytic reactions to minimize the overoxidation of desired products. Among numerous cocatalysts, owing to the surface plasmon resonance effect and the electron trapping effect, noble metals (Pt, Pd, Au, Ag etc.) generally exhibit the excellent performance for photocatalytic CH₄ oxidation^15,23,24,25. It has been found that dual metal cocatalysts involving noble metals and less-expensive metal species would integrate both electron acceptor and donor cocatalysts with the photocatalyst, which could boost the separation of photogenerated carriers and weaken the oxidative potentials of photocatalysts to properly suppress overoxidation^16,26. For example, a promising result obtained by Tang and coworkers²⁶ showed that ZnO with state-of-the-art dual Au–Cu cocatalysts achieved nearly 100% selectivity and high activity for 4 h with a 14.1% apparent quantum yield at 365 nm for CH₄ conversion to oxygenates. Recently, highly efficient alternative cocatalysts (including transition metals and their derivative species) to noble metals have been widely studied^{27,28,29,30,31}. However, their performance remains far from satisfactory due to the sluggish separation of photogenerated carriers. Most recently, Ye and coworkers²⁸ reported that atomically dispersed Ni on nitrogen doped carbon/TiO₂ composite (Ni−NC/TiO₂) achieved a high yield of oxygenates of 198 μmol with a selectivity of 93%, despite of the low apparent quantum yield (1.9% at 365 nm) for oxygenates. It is worth noting that most of the reported photocatalytic CH₄ oxidation reactions mainly follow the radical reaction mechanism, leading to unavoidable overoxidation of the products, and therefore a long-time (>360 min) accumulation of the oxygenate products cannot be achieved.

To address these problems and further improve the catalytic performance of non-noble metal-modified photocatalysts, rational and applicable designs of suitable cocatalyst are of great urgency. Highly dispersed cocatalysts of transition metal oxide, including single sites and clusters, can sufficiently expose active sites, and maximize metal oxide-support interaction to facilitate the separation of photogenerated carriers^32,33,34,35. More importantly, a rich tunability of the chemical state of transition metals and oxygen coordination may enable non-radical catalytic mechanism proceed, modulating the activity and selectivity by creating fitting electronic structure and active sites.

Molybdenum oxide (MoO_x), in particular, tends to exhibit unique physical and chemical properties due to the variable Mo valence range of +2 to +6 and the presence of a variety of Mo−O bonds (including Mo=O bonds, short Mo−O bonds, and long Mo−O bonds), whose atomic and electronic structures can be sensitively tuned by heteroatoms (Bi, Ce, Co, Fe, etc.) and oxide supports (Al₂O₃, TiO₂, ZrO₂, SiO₂)^36,37. As such, MoO_x based materials have been utilized to selective thermocatalytic oxidation of hydrocarbons, such as ethane, isobutane and isobutene^38,39. Most recently, Shen et al. dispersed MoO_x monolayer with Mo⁴⁺, Mo⁵⁺, and Mo⁶⁺ species on TiO₂, which exhibited high activity and selectivity in thermocatalytic oxidation of isobutene to produce methacrolein⁴⁰. In addition, some MoO_x and their derivatives (such as MoO_x/SiO₂ and MoO_x/SBA-15) were used to improve the efficiency of selective CH₄ oxidation^41,42,43,44. The volatility of active Mo species in the water-containing thermocatalytic reactions (≥500 °C) usually hinders the development of Mo-containing materials from an academic to a commercial catalyst^45,46,47. The bottleneck can be resolved by photocatalysis under ambient temperature.

Herein, we report that subnanometric MoO_x clusters anchored on TiO₂ can fully expose the active sites of Mo species for photocatalytic CH₄ oxidation with O₂. A high C1 oxygenates yield of 3.8 mmol/g with nearly 100% selectivity was achieved after 2 h of light irradiation, resulting in a 13.3% apparent quantum yield at 365 nm, which outperforms some recent reports (Table S1). A long-time (1800 min) accumulation of the products has been realized during reaction process with almost constant productivity and high selectivity (> 95%). Mechanistic studies reveal a photocatalytic cycle for CH₄ oxidation on MoO_x−TiO₂ surface. The MoO_x-cocatalysts not only trap photogenerated electrons to boost the carrier separation, but also largely inhibit the formation of •OH and O₂^•− radicals to jam the overoxidation of desired oxygenates. This work would promote the rational design of efficient and economical non-noble metal catalysts for direct conversion of methane (CH₄) to high-value-added oxygenates.

Results

Synthesis and characterization of the MoO_x−TiO₂ catalyst

After the synthesis of anatase TiO₂ nanosheet with predominantly exposed (001) facet, a series of MoO_x−TiO₂ photocatalysts with different Mo loading (0.05–0.7 wt.%) were prepared via a facile impregnation method. X-ray diffraction (XRD) patterns (Fig. 1a) show that all diffraction peaks correspond to pure anatase and no signal of Mo species is detectable on the MoO_x−TiO₂ samples. Scanning electron microscopy (SEM) image (Fig. 1b) shows that the 0.5%MoO_x−TiO₂ sample exhibits a typical nanosheet morphology with an average size of ca. 120 nm and a thickness of ca. 10 nm. The high-resolution transmission electron microscopy (HRTEM) (Fig. S1 in supporting information) shows clear lattice fringe and the lattice spacing parallel to the lateral facets is 0.356 nm, corresponding to the (101) facet of anatase TiO₂. Although the Mo species is not observed by HRTEM on the surface of 0.5%MoO_x−TiO₂, aberration-corrected high-angle annular dark-field scanning TEM (AC HAADF-STEM) images (Figs. 1c, d and S2) clearly exhibits subnanometric MoO_x clusters of ca. 0.6 nm on 0.5%MoO_x−TiO₂. The energy-dispersive X-ray spectroscopy (EDX) elemental mapping images (Fig. 1e) indicate that elemental Mo is uniformly dispersed throughout the entire surface of 0.5%MoO_x−TiO₂. Moreover, the MoO_x structures on 0.1%MoO_x−TiO₂ and 0.7%MoO_x−TiO₂ have also been characterized by the AC HAADF-STEM experiments (Figs. S3 and S4). It can be found that the MoO_x species mainly exist as single sites and small amount of subnanometric cluster on 0.1%MoO_x−TiO₂, while aggregated sub-nanometer MoO_x clusters with slightly increased size (ca. 0.77 nm) are present on 0.7%MoO_x−TiO₂.

Fig. 1: Synthesis and characterization of MoO_x−TiO₂ catalysts.

a XRD patterns of bare TiO₂ and MoO_x−TiO₂ with 0.05–0.7% Mo loading. b SEM, (c, d) AC HAADF-STEM images with different scale bar and (e) EDX elemental mapping images of 0.5%MoO_x−TiO₂. Source data are provided as a Source Data file.

The surface compositions and chemical states of MoO_x−TiO₂ were investigated by X-ray photoelectron spectroscopy (XPS). As shown in the high-resolution Mo 3d XPS spectra (Fig. 2a), two peaks at 232.6 and 235.8 eV can be ascribed to Mo 3d_5/2 and Mo 3d_3/2 of Mo⁶⁺ species, respectively. As the Mo loading increases, more MoO_x species are bond to TiO₂ surface, and more MoO_x clusters are formed, leading to the shift of Mo 3d, O 1s, and Ti 2p XPS peaks towards higher binding energy as shown in Figs. 2a and S5. X-ray absorption fine structure spectra (XAFS) were further acquired to investigate the coordination environment of Mo in MoO_x−TiO₂ using Na₂MoO₄, MoO₃, MoO₂, and Mo foil as references. The Mo K-edge X-ray absorption near-edge structure (XANES) spectra (Fig. 2b) show that the absorption edge position of MoO_x−TiO₂ is very similar to Na₂MoO₄, and the pre-edge transition at 20006.1 eV is due to Mo 1s to 4d transition with low symmetry of coordination atoms. Thus, the Mo species anchored to titanium dioxide should have tetrahedral [MoO₄] characteristics similar to Na₂MoO₄.

Fig. 2: The structure of Mo sites on MoO_x−TiO₂.

a Mo 3d XPS spectra of MoO_x−TiO₂ with different Mo loading. b Mo K-edge XANES spectra of 0.5%MoO_x−TiO₂, Na₂MoO₄, MoO₃, MoO₂, and Mo foil. c FT Mo K-edge EXAFS spectra of 0.5%MoO_x−TiO₂, Na₂MoO₄, MoO₃, and Mo foil. d EXAFS fitting curve of 0.5%MoO_x−TiO₂. The inset is the model MoO_x structure on TiO₂. The O, Ti, and Mo atoms are in red, grayish and cyan, respectively. e UV–Vis absorption spectra and (f) the corresponding plots of transformed Kubelka–Munk function versus photon energy of TiO₂, 0.5%MoO_x−TiO₂ and MoO₃. Source data are provided as a Source Data file.

The coordination structure of MoO_x species is resolved by combining Fourier transformed (FT) Mo K-edge extended X-ray absorption fine structure (EXAFS) analysis and atomistic modeling based on DFT calculations. Figure 2c compares the FT Mo K-edge EXAFS spectra for MoO_x−TiO₂, Na₂MoO₄, MoO₃, and Mo foil. The MoO_x−TiO₂ photocatalyst exhibits first-shell scattering at 1.19 Å in R space (without phase correction), which is similar to the values, 1.13 and 1.10 Å, found for Na₂MoO₄ and MoO₃, respectively. This is distinct from the case of Mo foil, in which the first-shell scattering locates at 2.38 Å. As such, we tentatively assign the primary scattering pair at 1.19 Å in the R-space spectrum of MoO_x−TiO₂ to two types of Mo−O bonding. To reveal the configuration of Mo species on TiO₂, the curve fitting for EXAFS spectrum was performed (Fig. 2d and Table S2). The fitting results of the first-shell scattering (1–2.2 Å) show that one Mo−O bond length is 1.68 Å with an average coordination number of 2.5, and the other is 2.26 Å with an average coordination number of 1.0 on MoO_x−TiO₂. Meanwhile, the wavelet transformation analysis of MoO_x−TiO₂ catalyst and MoO₃ has been conducted (Fig. S6), which is used to provide resolutions in both the R- and k-spaces and therefore has the potential to distinguish paths with similar coordination distances⁴⁸. The peak centered at (k, R) = (5, 1.2) is associated with the Mo−O single scattering path, which is similar to that in MoO₃. While the peaks centered at (k, R) = (8.5, 2.4) and (8.8, 3.1) probably correspond to the Mo−Ti and Mo−Mo scattering paths, respectively. The assignment is based on the fact that the scattering atoms are heavier with higher centering k values, and the peak at (k, R) = (8.8, 3.1) matches well with that of the Mo−Mo path in MoO₃ centered at (k, R) = (10.0, 3.1) (Fig. S6b). Combined with the AC HAADF-STEM result, we deduce that the Mo species exist as subnanometric hexavalent Mo oxide (MoO_x) clusters anchored on TiO₂ surface as shown in Fig. 2d, insert. According to the UV–Vis absorption spectra (Fig. 2e), the light absorption of bare TiO₂ occurs at wavelengths shorter than 400 nm (the ultraviolet region) while that of 0.5%MoO_x−TiO₂ extends to visible light region (400–800 nm). Then, the UV–Vis absorption data are converted into Tauc plots to determine the bandgap (Fig. 2f). Compared to the bare TiO₂, the bandgap of 0.5%MoO_x−TiO₂ exhibits a slight reduction (0.06 eV), which can be attributed to the formation of Ti−O−Mo bonds on TiO₂ anchored with subnanometric MoO_x clusters. Similar to the MoO₃, the visible light absorption of 0.5%MoO_x−TiO₂ should be mainly associated with the subnanometric MoO_x clusters.

The photocatalytic performance of CH₄ oxidation with O₂

Photocatalytic CH₄ oxidation reactions were performed on bare TiO₂ and MoO_x−TiO₂ photocatalysts with variable Mo loading. As shown in Fig. 3a, the oxygenate products of CH₄ photooxidation include CH₃OOH, CH₃OH, HCHO, HCOOH CO, and CO₂. C₂₊ products and H₂ could not be detected (Fig. S7). The produced CH₃OOH, CH₃OH, and HCOOH were detected by ¹H NMR spectra (Fig. S8), while HCHO was quantified using the calibration curves established by the acetylacetone colorimetric method (Fig. S9). CO₂ and CO as the overoxidized products, were determined by gas chromatography (Fig. S10). Xenon lamp was used as light source (Fig. S11). The bare TiO₂ exhibits a relatively low liquid oxygenate yield of 1.58 mmol/g for the 2 h reaction (0.79 mmol/g/h) with an oxygenate products selectivity of 87.2%. The yield of liquid oxygenates increases slightly when the Mo loading is increased to 0.1% and increases significantly to maximum (1.90 mmol/g/h) when the Mo loading is increased to 0.5%, while it declines slightly with the further increase of Mo loading to 0.7%. According to the AC HAADF-STEM results, the MoO_x species are fully transformed into subnanometric MoO_x clusters with the increase of Mo loading to 0.5%, and aggregated sub-nanometer MoO_x clusters with slightly increased size (ca. 0.77 nm) are present when the further increase of Mo loading to 0.7%. These experimental results indicate that the subnanometric MoO_x clusters can enhance the activity of photocatalytic CH₄ oxidation over MoO_x−TiO₂ more effectively than the single MoO_x sites. Moreover, the larger size and aggregated MoO_x clusters are slightly less favorable to the increase in the catalytic activity probably due to the reduced exposure of MoO_x sites. Compared to single-site catalysts, subnanometric-cluster catalysts not only have higher tunability in chemical composition and atomic arrangements of active sites, but also provide multiple adsorption sites for multi-reactant catalytic systems^49,50. Recently, Lu et al. prepared rGO catalysts anchored with Ru₃O₂ clusters, which were efficient and selective for the oxidative dehydrogenation reaction of 1,2,3,4-tetrahydroquinoline (THQ). They found that Ru₃O₂ clusters promoted the adsorption of reactants and desorption of products, and the stronger interaction of Ru₃O₂ clusters with the rGO support facilitated the electron transfer compared with that of single atoms and nanoparticles⁵¹. Dong et al. also demonstrated that the unique reactive interfaces between subnanometric BaO clusters and TiO₂ facilitated the activation and dissociation of nitrate, leading to the efficient and selective photocatalytic conversion of nitrate to ammonia⁵².

Fig. 3: Photocatalytic CH₄ oxidation by O₂.

a Product yield and selectivity of liquid organic oxygenates for a series of MoO_x−TiO₂ with different Mo loading. b Product yield and selectivity of liquid organic oxygenates for 0.5%MoO_x−TiO₂ with variable H₂O amount. Time course for product yield and selectivity of liquid organic oxygenates for (c) bare TiO₂ and (d) 0.5%MoO_x−TiO₂. Reaction condition in (a–d): 10 mg catalyst, 100 mL water, 2 MPa CH₄, 0.1 MPa O₂, 25 °C, light source: 300 W Xe lamp, 13 mW/cm² (UV light), 420 mW/cm² (full spectrum). Three repeated experiments were performed under the same conditions, and the total yield (the sum of all products) was obtained for each of the three experiments. Error bars represent the standard deviations of the three total yield. Source data are provided as a Source Data file.

The photocatalytic performance of MoO_x−TiO₂ exhibits a volcanic trend with the increase of Mo loading, and the highest yield of liquid oxygenates reaches 3.80 mmol/g for the 2 h reaction (1.90 mmol/g/h) over 0.5%MoO_x−TiO₂. On the other hand, the product selectivity also exhibits a volcanic trend along with the increase of Mo loading, and the optimal selectivity of liquid oxygenates is nearly 100% over 0.5%MoO_x−TiO₂. According to the AC HAADF-STEM results, the MoO_x species are transformed into subnanometric MoO_x clusters with the increase of Mo loading to 0.5%, and the product selectivity also increases. With the further increase of the Mo loading, aggregated sub-nanometer MoO_x clusters with slightly increased size are predominant, resulting in a slightly decrease of the product selectivity. Thus, the formation of subnanometric MoO_x clusters should be also the key to improve the selectivity of liquid oxygenates. In addition, the catalyst shows high stability, and there is no obvious decrease in activity after 6-run cycling tests (12 h in total) (Fig. S12). Noteworthily, due to the difference of light sources in different studies, including light intensity and composition, it is difficult to precisely compare the catalyst performance by reaction rate. The apparent quantum efficiency (AQE) provides a reliable assessment of photocatalytic efficiency. The AQE for the liquid oxygenates over 0.5%MoO_x−TiO₂ is determined to be 13.3% (Table S1) under the irradiation of monochromatic light of 365 nm.

When the solubility of CH₄ and O₂ in water is increased by elevating the pressure, the selectivity of liquid oxygenates reaches the maximum at CH₄ of 2 MPa and O₂ of 0.1 MPa (Figs. S13 and S14). The excessively dissolved CH₄ and O₂ cause a decrease in selectivity of liquid oxygenates from nearly 100% to 98% due to the overoxidation of products. The effect of water on the photocatalytic CH₄ oxidation over bare TiO₂ and 0.5%MoO_x−TiO₂ is also investigated (Figs. 3b and S15). With the increase of H₂O amount, both product yield and selectivity of liquid oxygenates are slightly boosted on bare TiO₂ due to the production of hydroxyl (•OH) and superoxide (O₂^•−) radicals and the enhanced desorption of the liquid oxygenates, in consistent with the previous reports^15,53,54. Unlike these radical reactions, with increasing H₂O amount from 25 to 100 mL, the yield of liquid oxygenates decreases from 5.40 to 3.80 mmol/g for 2 h reaction, while their selectivity increases from 97% to nearly 100% in the aqueous photocatalytic CH₄ oxidation over 0.5%MoO_x−TiO₂ (Fig. 3b), which may mainly occur on the MoO_x active sites of the catalyst surface. The increase of H₂O amount can reduce the concentration of the catalyst, and the probability of reactants to contact with the active sites on catalyst, leading to the decline in the product yield. Moreover, the increase of H₂O amount can reduce the concentration of the oxygenate products, and inhibit the further oxidation of these oxygenates to CO₂. Thus, the selectivity of the liquid oxygenates increases with the increase of H₂O amount. We also performed the photocatalytic CH₄ oxidation reaction on TiO₂ with variable Mo loadings under the test condition: 25 mL water, 2 MPa CH₄, 0.1 MPa O₂ (Fig. S16). The product yield and selectivity of liquid organic oxygenates are optimal for the 0.5%MoO_x−TiO₂ sample.

Figure 3c shows time course for product yield and selectivity of liquid organic oxygenates on bare TiO₂. The product yield slightly grows up after 360 min of irradiation, but the selectivity of liquid organic oxygenates decreases considerably from 90% to 77% along with irradiation time (up to 1800 min), which can be ascribed to the overoxidation of oxygenate products. With the increase of reaction time, more oxygenate products accumulate in solution, which boosts their further oxidation to CO₂. Thus, as the reaction time increases, the selectivity of liquid oxygenates on bare TiO₂ declines. In contrast, on 0.5%MoO_x−TiO₂, when the reaction time is gradually prolonged up to 1800 min, the product yield successively increases to 25.1 mmol/g with a high and stable selectivity of liquid organic oxygenates (95–100%) (Fig. 3d), indicating that the MoO_x clusters can not only largely improve the activity of photocatalytic CH₄ oxidation, but also hinder the overoxidation of these oxygenates to reach a high selectivity. Thus, a long-time accumulation of products can be achieved during the reaction process over the MoO_x−TiO₂ photocatalyst.

Reactive intermediates characterized by NMR and EPR

There is no product detectable without photocatalyst or light, or replacing CH₄ with N₂ (Table S3), indicating that all of the products originate from photocatalytic conversion of CH₄. Isotope labeling NMR experiments using ¹³CH₄ and ¹⁷O₂ were performed to trace the source of carbon and oxygen atoms of these products. The ¹³C NMR spectrum (Fig. 4a, upper) shows four obvious peaks assigned to CH₃OOH (65.0 ppm), CH₃OH (49.0 ppm), HOCH₂OH (82.3 ppm, derived from the hydration of HCHO in aqueous solution), and HCOOH (171.6 ppm), indicating that the formed oxygenates are derived from reactant ¹³CH₄. Similar result is also found from the ¹H NMR spectra (Fig. 4a, lower). For photocatalytic CH₄ oxidation in a ¹²CH₄ and ¹⁶O₂ atmosphere, two ¹H peaks at 3.26 and 8.42 ppm correspond to ¹²CH₃OH and H¹²COOH, respectively. Using ¹³CH₄ instead of ¹²CH₄, both peaks split into two peaks due to ¹H–¹³C J coupling (140 Hz) in the formed ¹³CH₃OH and H¹³COOH. Furthermore, for the reaction in a ¹³CH₄ and ¹⁷O₂ atmosphere, the full width at half maximum of the signals for ¹³CH₃OH and H¹³COOH obviously increases due to the weak ¹H–¹⁷O J coupling (1.96 Hz), indicating that the oxygen atoms in CH₃OH and HCOOH are originated from reactant ¹⁷O₂. Additionally, when the N₂ gas was used instead of O₂ in the photocatalytic reaction (Table S3), only trace yield of products (HCHO, CO, and CO₂) was generated, suggesting that the oxygen in HCHO is mainly derived from O₂. When photocatalytic CH₄ oxidation was carried out under direct heating conditions without a light source, only trace amounts of CO₂ were produced (Table S3), indicating that the reaction was light-initiated rather than thermally driven.

Fig. 4: Reactive intermediates on catalyst surface.

(a, upper) ¹³C and (a, bottom) ¹H NMR spectra for the product at a 2 h reaction time using 30 ml water, 80 kPa ¹³CH₄/¹²CH₄, 20 kPa O₂/¹⁷O₂, and 10 mg 0.5%MoO_x−TiO₂. b EPR spectra for detecting free radicals in the aqueous solution over bare TiO₂ and 0.5%MoO_x−TiO₂. DMPO was added as the radical trapping agent. c In-situ EPR spectra of 0.5%MoO_x−TiO₂ in N₂, CH₄, O₂ and (CH₄ + O₂) atmospheres before and during light irradiation. d In-situ ¹³C NMR spectra for bare TiO₂ (bottom) and 0.5%MoO_x−TiO₂ (upper) in a (¹³CH₄ + O₂) atmosphere with increasing irradiation time. Source data are provided as a Source Data file.

The electron paramagnetic resonance (EPR) with DMPO as the radical trapping agent was used to detect free radicals in the aqueous photocatalytic CH₄ oxidation with O₂ over bare TiO₂ and 0.5%MoO_x−TiO₂ (Fig. 4b). For bare TiO₂, the typical and strong signals of •OH and O₂^•− radicals trapped by DMPO (DMPO−OH and DMPO−OOH) were observed under light irradiation, indicating the presence of •OH and O₂^•− radicals. Thus, the CH₄ oxidation on the bare TiO₂ should undergo a radical process, which can lead to unavoidable overoxidation^55,56 and difficult accumulation of the liquid oxygenates for a long-time (Fig. 3c). For 0.5%MoO_x−TiO₂, both •OH and O₂^•− radicals decline remarkably, while its photocatalytic performance is much better than that of bare TiO₂. It can be concluded that the CH₄ oxidation mainly occurs on the metal active sites (MoO_x) on the surface of the photocatalyst, rather than via a radical reaction initiated by the •OH and/or O₂^•− radicals. When the same amount of Mo was loaded on silica, there was no product generated (Table S3), suggesting that TiO₂ plays a key role in photocatalytic CH₄ oxidation, which should be the source of photogenerated carrier formation.

We used in-situ EPR spectroscopy to follow the formation and reaction of paramagnetic intermediates on the surface of bare TiO₂ and 0.5%MoO_x−TiO₂. For 0.5%MoO_x−TiO₂, the EPR signal at g_┴ = 1.929, g_‖ = 1.874 appeared upon light irradiation (Fig. 4c), which is associated with tetra-coordinated Mo⁵⁺ sites (Mo_4C⁵⁺)⁵⁷ arising from trapped photogenerated electrons at the hexavalent MoO_x sites. The in-situ Mo 3d XPS experiments were conducted to confirm the transfer of photogenerated carriers on surface MoO_x species (Fig. S17a). In addition to the Mo_4C⁵⁺ species, a small amount of Mo⁴⁺ species were formed on MoO_x−TiO₂ under irradiation, which could not be detected by EPR. When O₂ was introduced into the 0.5%MoO_x−TiO₂ and the CH₄/0.5%MoO_x−TiO₂ system, both of the EPR signals (Fig. 4c) decreased significantly upon light irradiation. Thus, it can be deduced that the reduced Mo sites (Mo_4C⁵⁺ and Mo_4C⁴⁺) transfer the trapped electrons to activate adsorbed O₂, and are re-oxidized to hexavalent MoO_x sites. Meanwhile, according to the in-situ O 1s XPS experiments (Fig. S17b), the photogenerated hole can oxidize surface oxygen to generate oxygen vacancies. However, unlike the presence of surface superoxide (Ti−O₂^•) on the bare TiO₂ in the O₂/(O₂ + CH₄) atmosphere during light irradiation (Fig. S18), no such active surface radicals are present on 0.5%MoO_x−TiO₂. According to the in-situ ATR-FTIR spectra (Fig. S19), the reduced Mo sites can react with adsorbed O₂ to form surface peroxide sites Mo−OO and Mo−OOH, corresponding to the O−O stretching bands at 982 and 800 cm⁻¹, respectively^58,59. In the presence of CH₄ and O₂, the surface peroxide sites (especially Mo−OO) are significantly reduced upon light irradiation.

In-situ ¹³C MAS NMR experiments were used to follow the evolution of CH₄ photooxidation by O₂ on bare TiO₂ and 0.5%MoO_x−TiO₂ (Fig. 4d). The formation rate of products (including gas and liquid) remains almost constant within 1 h (Fig. S20). For bare TiO₂ in a (¹³CH₄ + O₂) atmosphere, four NMR signals evolve along with irradiation time. CH₄ was oxidized by active oxygen-contained radicals (•OH and O₂^•−) to physically adsorbed CH₃OH (52.0 ppm) and chemically adsorbed CH₃OH (65.0 ppm), which can be subsequently oxidized to HOCH₂OH (106.0 ppm) and adsorbed HCOOH (170.0 ppm). For 0.5%MoO_x−TiO₂ in a (¹³CH₄ + O₂) atmosphere, besides these oxygenate products, two types of surface intermediates are detected at 10–30 and 135.2 ppm, which can be assigned to physisorbed CH₄ and chemisorbed HCHO on MoO_x sites as confirmed by the following theoretical calculations. Compared with bare TiO₂, the amount of physisorbed CH₄ is greatly boosted on 0.5%MoO_x−TiO₂ surface upon the irradiation, which facilitates the photocatalytic oxidation of CH₄. The unique reaction intermediates and reaction process suggest that the photocatalytic reaction mechanism is entirely different from the previous radical reaction mechanism^60,61,62.

Photocatalytic mechanism and theoretical calculation of CH₄ oxidation by O₂

It can be found that HCHO is the main product of photocatalytic CH₄ oxidation on MoO_x−TiO₂ (60% of the total oxygenate products, Fig. 3a). More interestingly, CH₃OH tends to be further oxidized on bare TiO₂, while it is hardly further oxidized on MoO_x−TiO₂ (Fig. S21). As such, the HCHO should be mainly produced by direct CH₄ oxidation on MoO_x−TiO₂ rather than by the oxidation of product CH₃OH by radicals. Furthermore, due to the low content of •OH and O₂^•− radicals formed on MoO_x−TiO₂ (Fig. 4b), the HCOOH can be effectively accumulated with little overoxidation to CO₂ in the photocatalytic HCHO oxidation (Fig. S22). In contrast, owing to the relatively high content of •OH and O₂^•− radicals formed on bare TiO₂, the generated HCOOH tends to be overoxidized to CO₂. These results explain the difference in HCOOH formation between TiO₂ and MoO_x−TiO₂ (Fig. 3c, d).

On the basis of the aforementioned experimental results, the photogenerated electrons are trapped by hexavalent MoO_x sites to form Mo_4C⁵⁺ and Mo_4C⁴⁺ sites in the initial step of CH₄ activation on MoO_x−TiO₂, instead of trapping by surface OH/H₂O and O₂ to generate •OH and O₂^•− radicals, which has been detected by in-situ EPR and Mo 3d XPS techniques (Figs. 4c and S17a). On the other hand, as shown in in-situ O 1s XPS spectra (Fig. S17b), the photogenerated holes can oxidize surface oxygens to form oxygen vacancies. Subsequently, the reduced Mo sites (Mo_4C⁵⁺ and Mo_4C⁴⁺) react with adsorbed O₂ to generate surface peroxide sites, including Mo−OO and Mo−OOH, and the former can react with CH₄ as revealed by in-situ ATR-FTIR spectra (Figure S19), suggesting that these peroxide sites are the reactive sites.

To understand the enhanced reactivity for selective photocatalytic CH₄ oxidation to high-value-added oxygenate products, we have performed DFT calculation to simulate the reaction pathways on the proposed Mo−OO active site (Figs. 5 and  S23). The reaction energy for the formation of Mo−OO (intermediate A) through interaction of Mo−O center and O₂ was 1.76 eV on MoO_x−TiO₂, which is much lower than that (2.72 eV) for the formation of the peroxide intermediate (Ti−OO) on the bare TiO₂. According to our theoretical calculations, the Mo−OO active sites can dissociate into the original structure (Mo−O) and O₂ with an exothermic energy of −1.76 eV, indicating that the dissociation of Mo−OO site is thermodynamically favorable. This should be the reason that the peroxide species (including Mo−OO and Mo−OOH) formed upon irradiation decline obviously after the irradiation as shown in the in-situ ATR-FTIR spectra (Fig. S19). Thus, the formation of the Mo−OO active sites could be maintained by the photon energy available from light irradiation. The adsorption of CH₄ at the Mo−OO site (intermediate B) releases an energy of -0.08 eV, accounting for the increase in CH₄ physisorption upon irradiation (Fig. 4d). The activation of the first C−H bond of CH₄ produces chemisorbed CH₃OH (CH₃O−Mo₂, intermediate C), which is the rate-determining step of the catalytic cycle with an endothermic energy of 0.77 eV. According to the experimental results, CH₄ activation on bare TiO₂ can proceed via a radical mechanism. Our theoretical calculations indicate that the energy required for the first C−H bond activation decreases from 1.65 eV on bare TiO₂ (Fig. S24) to 0.77 eV on MoO_x−TiO₂ (intermediate B → C, Fig. 5), highlighting the effective promotion of CH₄ oxidation by MoO_x−TiO₂. The hydrolysis of CH₃O−Mo₂ would lead to the formation of product CH₃OH. According to the previous reports²⁵, the bare TiO₂ with predominantly exposed (001) facets can inhibit the formation of •CH₃ and •OH radicals. In addition to the radical reactions in solution, the C−H activation of CH₄ also occurs partly on surface peroxide active sites (Ti−OO) of the bare TiO₂. The MoO_x loading on the TiO₂ further reduces the formation of the radicals (Fig. 4b), so that the C−H activation of CH₄ occurs mainly on the surface peroxide active sites (Mo−OO) of MoO_x−TiO₂. Therefore, the similar type of active sites and surface photocatalytic mechanism lead to the less change in product distribution before and after MoO_x loading on TiO₂ as shown in Fig. 3a.

Fig. 5: Theoretical calculations.

The calculated reaction energy profile of photocatalytic CH₄ oxidation on the subnanometric MoO_x cluster. (Bottom) The corresponding structural models (structure A−G) in the figures. The H, C, O, Ti, and Mo atoms of MoO_x−TiO₂ are in white, dark gray, red, grayish, and cyan, respectively, while the O atoms from O₂ are in purple. Source data are provided as a Source Data file. The atomic coordinates of corresponding structural models data are provided as Supplementary Data 1.

In the following step, CH₃O−Mo₂ is transformed to chemisorbed HCHO (CH₂O−Mo₂), and one H₂O molecule (intermediate D) is generated by Mo−OOH abstracting one hydrogen atom from CH₃O−Mo₂ with an activation energy −1.54 eV. Therefore, the formation of chemisorbed HCHO through the second C−H bond activation is a thermodynamically favorable process. Two photocatalytic cycles of CH₄ conversion on MoO_x−TiO₂ are proposed (Fig. 6). In the H₂O-assisted mechanism of CH₄ oxidation, the medium H₂O molecule could facilitate desorption of H₂O molecule and interact with the C atom of chemsorbed HCHO (H₂O···CH₂O, intermediate E₁) with an exothermic reaction of −2.12 eV. Then the formed H₂O···CH₂O could be converted into chemsorbed dihydroxymethane (HOCH₂OH−Mo, intermediate F₁) through a H transfer step with an endothermic energy of 0.80 eV. After that, the attack of O₂ on the Mo−O site results in the formation of terminal Mo−OO site, companying with the cleavage of HOCH₂OH−Mo bond with an exothermic reaction of −0.94 eV, indicating that the generation of Mo−OO site and physisorbed CH₂(OH)₂ (intermediate G₁) is thermodynamically favorable. Finally, the CH₂(OH)₂ is released with a low endothermic energy of 0.15 eV, suggesting that the formed CH₂(OH)₂ can be readily separated from the MoO_x−TiO₂ surface. The CH₂(OH)₂ has been detected by ¹³C NMR spectra (Fig. 4a), which is a derivative from the hydration of HCHO. The formation of CH₂(OH)₂ would prohibit the further oxidation of HCHO, in agreement with our experimental results. As a long-time accumulation of HCHO proceeds in the reaction, the amount of HCHO adsorbed on the active site increases, leading to a tendency for HCHO to be further oxidized to HCOOH (Fig. 3d).

Fig. 6: Photocatalytic mechanism.

Proposed photocatalytic mechanism for CH₄ oxidation by O₂ on MoO_x − TiO₂.

In the anhydrous mechanism of CH₄ oxidation, the direct desorption of H₂O molecule with an endothermic energy of 0.45 eV results in the formation of intermediate E₂, and then one O₂ molecule adsorbs on its Mo−O site to form intermediate F₂ with an adsorption energy of −0.32 eV. The adsorbed O₂ molecule also facilitates the conversion from chemsorbed HCHO to physisorbed HCHO (intermediate G₂) with an exothermic reaction of −1.92 eV. The desorption energy of HCHO is as low as 0.16 eV. As shown in Fig. 5, although the anhydrous mechanism of CH₄ oxidation is thermodynamically feasible, the H₂O-assisted reaction mechanism is more energetically favorable.

To elucidate the effect of different structure of MoO_x species on catalytic activity, in-situ EPR experiments were also conducted on 0.1%MoO_x−TiO₂ and 3.0%MoO_x−TiO₂ (Fig. S25). It has been found that the MoO_x species mainly exist as single sites on 0.1%MoO_x−TiO₂ (Fig. S3). Similar to the case of 0.5%MoO_x−TiO₂, the reduced Mo sites (Mo_4C⁵⁺) were also formed on the 0.1%MoO_x−TiO₂ under illumination as shown in Fig. S25a. Subsequently, the reduced Mo sites can react with adsorbed O₂ to generate surface active peroxide sites (Mo−OO) that can activate CH₄. As such, the single MoO_x sites should also exhibit excellent photocatalytic activity of CH₄ oxidation. However, the content of the active Mo sites exposed on the single MoO_x sites of 0.1%MoO_x−TiO₂ is much lower than that exposed on the subnanometric MoO_x clusters of 0.5%MoO_x−TiO₂. This should be the reason that properties of part of the bare TiO₂ surface are preserved, while a small amount of surface superoxide species (Ti−O₂•) are formed on the 0.1%MoO_x−TiO₂. We have performed DFT calculations to predict the activation of the first C−H bond of CH₄ on TiO₂ loaded with single MoO_x site (Fig. S26), which is the rate-determining step of the photocatalytic CH₄ oxidation. It can be found that the first C−H bond activation in CH₄ exhibits thermodynamic favorability with a calculated reaction energy −0.80 eV, which is much lower than that (0.77 eV) on TiO₂ loaded with subnanometric MoO_x cluster. While the formation of oxygen vacancy (Mo−O_v) requires a high energy of 4.35 eV on TiO₂ loaded with single MoO_x site, which is the key to generate the Mo−OO active sites, much higher than that of 3.30 eV on TiO₂ loaded with subnanometric MoO_x cluster (which is similar to the band gap energy of anatase TiO₂). Therefore, the photocatalytic activity of CH₄ oxidation on 0.5%MoO_x−TiO₂ is much higher than that on 0.1%MoO_x−TiO₂ (Fig. 3a). According to the XRD experiment of 3.0%MoO_x−TiO₂ (Fig. S27), the MoO_x species loaded on TiO₂ should be MoO₃ nanoparticles. As shown in the in-situ EPR spectra (Fig. S25b), reduced Mo sites (hexa-coordinated Mo⁵⁺ sites, Mo_6C⁵⁺)⁶³ were formed on the 3.0%MoO_x−TiO₂ under illumination. Unlike 0.1%MoO_x−TiO₂ and 0.5% MoO_x−TiO₂, under O₂ and (O₂ + CH₄) atmospheres, the signal intensity of the Mo_6C⁵⁺ site remains almost unchanged on 3.0%MoO_x−TiO₂, indicating that the formed Mo_6C⁵⁺ sites cannot transfer the photogenerated electron and activate CH₄ effectively. Therefore, the activity of photocatalytic CH₄ oxidation on 3.0%MoO_x−TiO₂ (0.91 mmol/g/h) is much lower than that on 0.5%MoO_x−TiO₂ (1.90 mmol/g/h).

Discussion

In summary, we report a highly efficient and selective photocatalytic oxidation process of CH₄ into high-value-added oxygenates by molecular O₂ on Mo oxide anchored TiO₂ catalysts (MoO_x−TiO₂). The AC HAADF-STEM and X-ray absorption spectra show that the MoO_x species with tetrahedral [MoO₄] characteristics are present as subnanometric clusters of hexavalent MoO_x sites, uniformly dispersed throughout TiO₂. Based on in-situ EPR, XPS, and NMR results, a full photocatalytic cycle for CH₄ oxidation by MoO_x−TiO₂ is proposed, which is completely different from the conventional radical reaction mechanism. The initial step in the activation of CH₄ on MoO_x−TiO₂ involves the capture of photogenerated electron-hole pairs by the hexavalent MoO_x sites to form Mo_4C⁵⁺ and Mo_4C⁴⁺ sites, followed by reduction of adsorbed O₂ to generate a surface peroxides site (Mo−OO), which largely inhibit the formation of •OH and O₂^•− radicals and the overoxidation of oxygenate products. Combined with the DFT calculations, it can be found that the Mo−OO site is the reactive site which can promote the physisorption of CH₄ and the activation of the first C−H bond of CH₄. As a result, a high C1 oxygenates yield of 3.8 mmol/g with nearly 100% selectivity was achieved on the MoO_x−TiO₂ after 2 h of irradiation, resulting in a 13.3% apparent quantum yield at 365 nm. Furthermore, a long-time (1800 min) accumulation of oxygenate products has been realized during reaction process with almost constant productivity and high selectivity (> 95%). The results presented herein would be helpful for the rational design of efficient non-noble metal photocatalysts for direct conversion of CH₄ to high-value-added oxygenates.

Methods

Materials

The following chemicals are used in the experiment, including Tetra-n-butyl titanate (Ti(OBu)₄, Macklin, 98%), Hexaammonium heptamolybdate tetrahydrate ((NH₄)₆Mo₇O₂₄•4H₂O, Sinopharm Chemical Reagent Co., Ltd., ≥ 99.0%), Anhydrous ethanol (CH₃CH₂OH, Sinopharm Chemical Reagent Co., Ltd., ≥ 99.7%), Sodium hydroxide (NaOH, Sinopharm Chemical Reagent Co., Ltd., ≥ 96.0%), Hydrofluoric acid (HF, Sinopharm Chemical Reagent Co., Ltd., ≥ 40%), 5,5-dimethyl-1-pyrroline-N-oxide (DMPO, Dojindo Laboratories, ≥ 99%), Deuterium oxide (D₂O, Energy Chemical, 99.9%). The water in all the experiments is de-ionized (DI) water in resistivity of 18.2 MΩ·cm.

Catalyst preparation

TiO₂ nanosheets with dominant (001) facet were synthesized by hydrothermal method. Briefly, 1.0 ml of hydrofluoric acid (HF) was added to 25.0 ml of Ti(OBu)₄ in a Teflon-line autoclave and then kept at 473 K for 24 h (Caution: HF is corrosive and a contact poison, and it must be handled with care!). After the autoclave was cooled to room temperature, the resulting product was separated by centrifugation and washed three times with anhydrous ethanol and NaOH solution (0.1 M), respectively. The catalyst was then stirred in the NaOH solution for 24 h and washed with deionized water until the supernatant was neutral. Finally, it was dried at 80 °C for 8 h. Catalyst yields are around 50%.

The MoO_x-loaded TiO₂ catalyst was prepared by simple impregnation method. Dissolve 0.5 g ammonium molybdate in 10.0 ml water to form ammonium molybdate solution. The Mo loading was achieved by mixing a certain amount of ammonium molybdate solution (0.56 ml) and water with TiO₂ nanosheets (0.2 g) and stirring until the liquid is completely evaporated. The mixture was calcined at 673 K for 4 h in an air atmosphere. Catalyst yields are around 85%.

Catalyst characterization

XRD measurements of the prepared samples were conducted on a Panalytic X’ Pert³ diffractometer with Cu Kα radiation (λ = 1.54184 Å) operating at 40 kV and 40 mA. SEM was measured on a S4800 apparatus with an acceleration voltage of 3 kV. HRTEM images were acquired on Tecnai G2 F30 S-TWIN instrument at 300 kV. AC HAADF-STEM images and energy dispersive X-ray spectroscopy (EDS) mapping were measured on JEM ARM 200 F apparatus with an acceleration voltage of 200 kV. X-ray photoelectron spectra (XPS) were collected on a Thermo Scientific Escalab 250Xi instrument using Al Kα (1486.6 eV) irradiation with the C 1 s characteristic peak of 284.8 eV as the reference. X-ray Absorption Spectroscopy spectra were acquired at the Mo K-edge on beamline 1W1B at the Beijing Synchrotron Radiation Facility, which operates at 2.5 GeV with a current of 250 mA. The UV–Vis absorption spectra were obtained using a Cary 4000 UV–Vis spectrometer with BaSO₄ as a reference. Fourier transform infrared spectroscopy (FTIR) experiments were conducted on Nicolet iS50 spectrometer with a homemade spectral cell for attenuated total reflectance (ATR) FTIR experiments.

EPR experiments were performed using a JEOL JES-FA200 spectrometer to detect photogenerated reactive oxygen species⁶⁴. Hydroxyl radicals (·OH) and superoxide radicals (${{{{\rm{o}}}}}_{2}^{\_}$) were captured via the spin-trapping technique with 5, 5-dimethyl-1-pyrroline N-oxide (DMPO). All these EPR experiments were conducted with 0.1 mT of the modulation amplitude and 3 mW of the microwave power at room temperature. The microwave frequency was 9.1 GHz.

Solid-state ¹³C CP/MAS NMR spectra were acquired on a Bruker Advance III 400 spectrometer (9.4 T) with resonance frequencies of 399.33 MHz (¹H) and 100.42 MHz (¹³C). A 4 mm MAS probe at a spinning rate of 12 kHz was utilized. For ¹H −¹³C cross-polarization, Hartmann-Hahn matching conditions were optimized using hexamethylbenzene (HMB), with a contact time of 6 ms and a recycle delay of 1.5 s.

Photocatalytic measurements

Photocatalytic CH₄ oxidation reactions were conducted in a 230 mL batch photoreactor equipped with a quartz window on the top. In a typica procedure, 10 mg of catalyst was uniformly dispersed in 100 mL deionized water via ultrasonication for 5 min. The photoreactor was then sealed and purged using O₂ (purity, 99.999%) for 15 min to exhaust air. Subsequently, the reactor was pressurized with 0.1 MPa O₂ and 2.0 MPa CH₄ (purity, 99.999%). The mixture was continuously stirred at 1000 rpm in a water bath under illumination. After the reaction, the reactor was cooled in an ice bath to below 10 °C. Gaseous products were collected in gas bags and quantified by gas chromatography equipped with a methanizer and flame ionization detector. Liquid-phase products were filtered and analyzed via ¹H NMR spectroscopy. Formaldehyde was quantified by the colorimetric method²⁵. The selectivity of products were calculated according to the following equations:

$${Oxygenates\; selectivity} \, \left(\%\right)=\frac{n\left(C{H}_{3}{OOH}\right)+n\left(C{H}_{3}{OH}\right)+n\left({HCHO}\right)+n\left({HCOOH}\right)}{n\left({total \, products}\right)}\times 100\%$$

Isotopic tracer experiments

¹³C-labeled CH₄ and/or ¹⁷O-labeled O₂ were used to trace the fate of carbon atoms from methane and oxygen atoms from oxygen. Typically, 10 mg catalyst was dispersed in 30 ml of water and transferred to a glass unit for sealing. The catalyst solution was degassed on a vacuum system and labeled gas was added for the reaction. A similar approach was adopted to do ¹³C-labeled CH₃OH photocatalytic conversion experiment. Bruker Avance-600 liquid NMR spectrometer was employed to analyze the liquid products obtained from the isotopic tracing experiments. The ¹H NMR spectra were acquired using a water suppression pulse sequence.

Calculational method

Spin-polarized DFT calculations with Perdew-Burke-Ernzerhof exchange-correlation functional were performed by applying the Vienna Ab initio Simulation Package (VASP) code^65,66. The plane-wave basis set in conjunction with the projected augmented wave method was utilized to describe the valence electrons and the valence-core interactions.

The kinetic energy cut-off of the plane wave basis set and energy convergence threshold for each iteration was set to 400 eV and 10⁻⁵ eV, respectively. The geometries were considered to have converged when the forces acting on each atom was less than 0.05 eV/Å. The Brillouin zone-sampling was restricted to the gamma point. The van der Waals (vdW) interactions were incorporated by using Grimme’s DFT−D3 method as implemented in VASP. In this study, Mo₄O₁₂ and Mo₃O₁₁−Mo−OO clusters were constructed on the (001) facet of anatase TiO₂ (Fig. S23). Based on the structure of anatase TiO₂, we simulated the (001)−2 × 2 surface using 6 atomic layers with a 15 Å of vacuum between periodic slabs. The bottom three atomic layers were held fixed in their bulk positions while the remaining atoms were allowed to relax.