Simultaneous promotion of photocatalytic CH₄ conversion and H₂O₂ production via nanopore water confinement

Abstract

Aqueous photocatalytic CH₄ oxidation offers a promising route for converting natural gas into oxygenates, a process governed by multi-electron and proton transfer at the catalyst-water interface. Here, we demonstrate that spatially confining water within Au/TiO₂@pSiO₂ core-shell catalysts—by reducing silica pore size to 1.7 nm—increases CH₄ conversion three-fold and H₂O₂ production 22-fold compared to Au/TiO₂. This strategy is generalizable to other semiconductors and cocatalysts, with Pt/TiO₂@pSiO₂-1.7 exhibiting oxygenate yields of 32.7 mmol g^-1 h^-1 and a 14.1% apparent quantum yield at 365 nm. Spectroscopic studies and molecular dynamics simulations reveal that water confined within pores, with a weakened hydrogen-bonding network, alters proton-coupled electron transfer pathways. Water oxidation transits to a concerted pathway, favoring •OH production for CH₄ conversion, while oxygen reduction shifts to a two-electron process, directly producing H₂O₂. This work highlights the potential of water confinement for designing efficient photocatalysts for CH₄ conversion.

Introduction

Aqueous photochemical processes are vital for energy conversion and environmental applications, such as water splitting¹, O₂/CO₂ reduction², and pollutants removal³. These processes typically involve multi-electron and multi-proton transfer steps that generate reactive oxygen species (ROS) as key intermediates^3,4,5. Photocatalytic methane oxidation has emerged as a green route to directly transform abundant natural gas into valuable fuels and chemicals, such as methanol^6,7,8. This approach utilizes photon energy to overcome the high activation barriers of CH₄, enabling room-temperature conversion via hydroxyl radicals (•OH) and bypassing the energy-intensive syngas route^6,9. Moreover, the aqueous environment offers unique advantages in improving product selectivity by facilitating the desorption of partially oxygenated products^9,10. Although extensive research has focused on developing photocatalysts with enhanced charge separation and surface reactions regulation through band structure engineering^11,12,13 and cocatalyst modification^{9,14,15,16,17,18}, the potential of tuning the aqueous microenvironment surrounding the semiconductor to boost CH₄ conversion remains largely unexploited.

Aqueous photocatalytic CH₄ oxidation involves a complex network of electron transfer (ET) and proton transfer (PT) reactions at the semiconductor-water interface (Fig. 1, Supplementary Fig. 1), where water serves as a reactant, proton source, and solvent^19,20,21. For hydrophilic oxide semiconductors like TiO₂, water is the primary hole acceptor, generating •OH via proton and electron transfer (Fig. 1a)^9,20. These •OH radicals activate CH₄ through hydrogen atom transfer (HAT), forming methyl radicals (•CH₃). Simultaneously, O₂ is reduced to superoxide radicals (•OOH), which further contribute to •CH₃ conversion into oxygenates¹⁵. These ET and PT events can occur stepwise or concertedly via proton-coupled electron transfer (PCET) mechanisms, which are highly sensitive to the solvent environment due to the significant energy required for solvent reorganization—especially in water^22,23. While studies over model reactions have shown examples of enhanced catalytic performance by tuning water structure^{24,25,26,27,28,29,30,31,32,33}, its application to complex photocatalytic reactions such as methane oxidation remains largely unexplored. Therefore, designing a photocatalyst that modulates water structure near the semiconductor interface presents a promising avenue for tuning proton/electron transfer pathways and optimizing methane oxidation kinetics.

Fig. 1: Design of nanoporous core-shell photocatalyst for enhanced CH₄ conversion and simultaneous H₂O₂ production.

a Schematic representation of PCET and CH₄ activation pathways in photocatalytic CH₄ oxidation over Au/TiO₂ in bulk water. b Schematic representation of PCET and CH₄ activation pathways in photocatalytic CH₄ oxidation over Au/TiO₂@pSiO₂ with confined water. The black dashed rectangle and blue dashed lines show a schematic of the disrupted water structure inside the nanopore. In (a, b), colored spheres represent specific atoms (light gray: H; dark gray: C; ice blue: O). PT: proton transfer; ET: electron transfer; PCET: proton-coupled electron transfer; CPET: concerted proton-electron transfer; HAT: hydrogen atom transfer; h⁺: holes; e⁻: electrons.

Spatial confinement via tunable nanopores offers an effective strategy to alter water structure^22,34, and consequently, the dynamics of these ET and PT processes³⁵. Confined water is known to exhibit distinct hydrogen-bonding (HB) networks, physical properties and reactivities compared to bulk water^35,36. In this work, we demonstrate that water confinement by silica nanopores in Au/TiO₂@pSiO₂ catalysts modulates PCET pathways and achieves efficient CH₄ conversion and H₂O₂ production simultaneously (Fig. 1b). We select TiO₂ as the semiconductor material for its ability to oxidize water to •OH radicals^12,20 and employ a noble metal cocatalyst to enhance CH₄ conversion^9,14,37. Water confinement is achieved by installing nanoporous and UV-transparent SiO₂ enclosures on the Au/TiO₂ core. Reducing the silica pore size from 3.1 nm to 1.7 nm leads to an approximately three-fold enhancement in CH₄ conversion and a 22-fold increase in H₂O₂ production compared to Au/TiO₂. This confinement strategy is applicable to other metal cocatalysts and semiconductors, as demonstrated by Pt/TiO₂@pSiO₂-1.7, achieving maximum oxygenate yields of 32.7 mmol g⁻¹ h⁻¹ with 95.5% selectivity and an apparent quantum yield (AQY) of 14.1% at 365 nm. Mechanistic investigations and molecular dynamics (MD) simulations with machine learning potentials (MLPs) reveal that the weakened HB network of confined water modulates PCET pathways for both water oxidation reaction (WOR) and oxygen reduction reaction (ORR). Specifically, WOR shifts from stepwise to concerted PCET, enhancing •OH radical production for CH₄ conversion, while ORR shifts from a one-electron (1e⁻) to a 2e⁻ process, promoting H₂O₂ production. These findings underscore the potential of engineering water structures via spatial confinement to promote photocatalytic activity and selectivity.

Results

Synthesis and characterization of Au/TiO₂@pSiO₂ catalysts

TiO₂ is a well-established semiconductor for photocatalytic methane oxidation, with a valence band position sufficient for oxidizing water to •OH radicals^12,20. Noble metal cocatalyst modification is known to promote CH₄ conversion^9,14,37. Therefore, we selected Au/TiO₂ as the model catalyst to study the effect of water confinement. We chose the SiO₂ enclosure due to its facile growth, controllable pore size, optical transparency in the UV region, and good stability. Furthermore, our previous studies have shown that a thin, non-porous silica shell can prevent oxygenates overoxidation while still allowing •OH radicals transport and convert CH₄²⁰, indicating good compatibility of silica with the TiO₂-based photocatalysts.

Gold nanoparticles (NPs) were loaded onto TiO₂ (P25) via NaBH₄ reduction^9,38, and a nanoporous silica shell was constructed on Au/TiO₂ using the Stöber method with alkylammonium surfactants as templates (Fig. 2a)³⁹. By varying surfactant carbon chain length or adding auxiliary organics^40,41, we obtained Au/TiO₂@pSiO₂-X with different pore sizes (X = pore size). The surface area and porosity of Au/TiO₂@pSiO₂-X were characterized by nitrogen physisorption (Fig. 2e, Supplementary Table 1). Using cetyltrimethylammonium bromide (C₁₆TAB) resulted in mesopores ~2.6 nm, while a shorter surfactant, dodecyltrimethylammonium bromide (C₁₂TAB), yielded smaller pores around 1.7 nm (Fig. 2f). Adding hexane as a pore-expanding agent yielded larger pores around 3.1 nm. For comparison, Au/TiO₂@SiO₂ without nanopores were synthesized similarly but without alkylammonium surfactants (Supplementary Fig. 2).

Fig. 2: Structural characterizations of Au/TiO2@pSiO2 with varying silica pore sizes.

a Schematic illustration of the preparation of Au/TiO₂@pSiO₂-X catalysts. The black dashed box highlights the template agents used in material synthesis: C₁₂TAB, dodecyltrimethylammonium bromide; C₁₆TAB, cetyltrimethylammonium bromide. b HR-TEM image of Au/TiO₂@pSiO₂-1.7 catalysts. The white dashed line outlines the porous SiO₂ shell, and the white arrow points to the SiO₂ shell. c A High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image of Au/TiO₂@pSiO₂-1.7 catalysts. White arrows point to the highly dispersed Au NP and TiO₂ core, respectively. d EDS mapping images of Au/TiO₂@pSiO₂-1.7. Different colors correspond to different elements (red: Ti; green: Si; yellow: Au; cyan: O). e N₂ adsorption-desorption isotherms of Au/TiO₂ and Au/TiO₂@pSiO₂-X. f Pore size distribution of Au/TiO₂ and Au/TiO₂@pSiO₂-X. g UV-Vis DRS of Au/TiO₂ and Au/TiO₂@pSiO₂-X.

Transmission electron microscopy (TEM), energy-dispersive X-ray spectroscopy (EDS), and X-ray diffraction (XRD) confirmed full encapsulation of Au/TiO₂ by a uniform, amorphous SiO₂ shell across all Au/TiO₂@pSiO₂-X catalysts (Fig. 2b–d, Supplementary Figs. 3–5). The silica layer had an average thickness of 8 nm (Fig. 2b–d), which is tunable by varying the amount of tetraethyl orthosilicate (TEOS) (Supplementary Fig. 6). Au NPs ranging 2–6 nm were randomly dispersed on TiO₂ (Fig. 2b–d, Supplementary Fig. 3), with a loading of ~0.1 wt% (Supplementary Table 2). Contact angle measurements demonstrate that catalysts with different pore sizes exhibit similar hydrophilic properties (Supplementary Fig. 7). UV-visible diffusive reflective spectra (UV-Vis DRS) of Au/TiO₂ and Au/TiO₂@pSiO₂-X were nearly identical (Fig. 2g, Supplementary Fig. 8), suggesting minimal effect of the SiO₂ coating on light absorption. X-ray photoelectron spectra (XPS) showed similar Ti oxidation states among all samples (Supplementary Fig. 9). These results indicate that the nanoporous coating does not significantly modify the intrinsic properties of the semiconductors.

Pore-size dependent CH₄ photocatalytic oxidation by Au/TiO₂@pSiO₂-X

Photocatalytic reactions were conducted in a batch reactor with 100 mL of distilled water under 2.5 MPa CH₄ and 1 MPa O₂ at 25 °C (Supplementary Fig. 10). Liquid products (CH₃OOH, CH₃OH, and HCOOH) were analyzed by ¹H-NMR, while HCHO was quantified by colorimetry methods (Supplementary Fig. 11). Control experiments, including dark and catalyst-free controls (Supplementary Table 3), and a ¹³CH₄ isotope-labeling study (Supplementary Fig. 12), validated that oxygenates were generated from photocatalytic methane oxidation over Au/TiO₂@pSiO₂. Compared to bare Au/TiO₂, Au/TiO₂@pSiO₂ significantly enhanced oxygenate yields, which increased as pore size decreased (Fig. 3a, b). Au/TiO₂@pSiO₂-3.1 exhibited a total oxygenates yield of 18.0 ± 0.4 mmol g⁻¹ h⁻¹, with 92.5% selectivity. Further reducing the pore size to 1.7 nm led to higher yields, achieving 27.4 ± 0.6 mmol g⁻¹ h⁻¹—approximately triple that of Au/TiO₂. The AQY for Au/TiO₂@pSiO₂-1.7 was 10.9 ± 0.3% at 365 nm (Supplementary Fig. 13, Supplementary Table 4), substantially exceeding Au/TiO₂ (3.3 ± 0.2%). Furthermore, this performance trend remained consistent under a broad-spectrum Xenon lamp, demonstrating that the catalytic improvement is independent of the specific light source (Supplementary Fig. 14). Notably, Au/TiO₂@pSiO₂ also produced a significant amount of H₂O₂ as a valuable byproduct (Fig. 3a, c, Supplementary Figs. 15–17 and Supplementary Note 1). The H₂O₂ production also increased with decreasing pore size, reaching 44.6 ± 1.3 mmol g⁻¹ h⁻¹ for Au/TiO₂@pSiO₂-1.7 (Fig. 3c), representing a 22-fold increase over Au/TiO₂ and competitive with state-of-the-art catalysts (Supplementary Table 5).

Fig. 3: Photocatalytic performance for methane oxidation.

a Schematics of products formation from CH₄ and O₂ over Au/TiO₂@pSiO₂-X. b Yields of oxygenates from CH₄ oxidation over different photocatalysts. c H₂O₂ production over different photocatalysts. Reaction conditions for (b, c): 10 mg catalysts, 100 mL H₂O, 2.5 MPa CH₄, 1 MPa O₂, 1 h reaction time, reaction temperature of 25 ± 3 °C, light source of 365 nm UV LED, 80 mW/cm². d O₂ partial pressure dependence on Au/TiO₂@pSiO₂-2.6. Reaction conditions: 10 mg catalysts, 100 mL H₂O, 2.5 MPa CH₄, 1 h reaction time, reaction temperature of 25 ± 3 °C, light source of 365 nm UV LED, 80 mW/cm². e Productivity assessment of Au/TiO₂@pSiO₂-1.7 obtained at different light intensities. Reaction conditions: 10 mg catalysts, 100 mL H₂O, 2.5 MPa CH₄, 1 MPa O₂, 1 h reaction time, reaction temperature of 25 ± 3 °C, light source of 365 nm UV LED. In (b–e), bars represent product yields (left y-axis), while the pink diamonds and corresponding pink arrow indicate oxygenate selectivity (right y-axis). The error bars in (b–d) represent the standard deviations derived from three independent experiments.

Further optimization of parameters, including light intensity, water volume, and Au loading, was performed (Fig. 3e, Supplementary Figs. 18, 19). By increasing the light flux to 500 mW/cm², we achieved a yield of 62.9 mmol g⁻¹ h⁻¹ with 80% selectivity for oxygenates (Fig. 3e). Au/TiO₂@pSiO₂-1.7 remained stable over ten cycles, with consistent production and selectivity (Supplementary Fig. 20). XRD and TEM showed no obvious structure changes in the spent catalyst (Supplementary Fig. 21), and N₂ adsorption isotherms showed pore size distribution and surface area similar to those of fresh catalysts (Supplementary Fig. 22). ICP-OES analysis verified that the gold loading remained unchanged after cycles, with negligible gold leaching observed (Supplementary Table 6).

To elucidate the role of the nanoporous silica shell, a series of control experiments was performed. A physical mixture of Au/TiO₂ and porous SiO₂ (pSiO₂-1.7) showed similar catalytic performance to Au/TiO₂ alone, highlighting the importance of the core-shell architecture (Supplementary Figs. 23, 24). Notably, Au/TiO₂@SiO₂ without pores yielded more than Au/TiO₂ but less than the nanoporous catalysts (Fig. 3b), probably due to the greater diffusion barrier in non-porous silica. This trend signifies the importance of controlling silica pore size for efficient methane conversion.

Silica thickness also significantly impacts performance, showing a non-monotonic effect. For Au/TiO₂@pSiO₂ catalysts with an identical pore size and surface hydrophilicity, an 8 nm shell was optimal (Supplementary Figs. 6, 25–27). A 30 nm shell reduced yields (Supplementary Fig. 27), likely due to an increased diffusion barrier. However, thinner shells (3.6 nm and 6 nm) also resulted in lower yields; the catalyst with a 3.6 nm shell, in particular, was substantially less active. This is likely attributed to an ultra-thin shell failing to provide complete and uniform encapsulation over TiO₂, thus being unable to establish an effective water confinement environment. These findings reveal a critical trade-off: the shell must be thick enough to create an effective confined microenvironment but thin enough to minimize mass transport resistance. Thus, a shell architecture with both optimal porosity and thickness is essential.

Given the critical role of the nanoporous shell, we investigated its influence on gaseous reactant transport. Temperature-programmed desorption of CH₄ (CH₄-TPD) showed weak adsorption on both Au/TiO₂ and Au/TiO₂@pSiO₂-1.7 (Supplementary Fig. 28), with the latter showing an even lower adsorption temperature. These results rule out CH₄ enrichment as the source of improved catalytic performance. We then examined the effect of O₂, whose direct participation in oxygenates formation was confirmed by ¹⁸O isotopic labeling (Supplementary Fig. 29). Both Au/TiO₂ and Au/TiO₂@pSiO₂-X exhibited low yield under anaerobic conditions, with yields increasing with O₂ pressure (Fig. 3d, Supplementary Fig. 30). Au/TiO₂ achieved maximum yield at 0.042 MPa O₂, whereas Au/TiO₂@pSiO₂-2.6 and Au/TiO₂@pSiO₂-1.7 required higher pressures (up to 1 MPa), suggesting hindered O₂ transport by the silica shell (Fig. 3d, Supplementary Fig. 30a, b and Supplementary Note 2). Furthermore, TiO₂ and TiO₂@pSiO₂ required higher O₂ pressures to attain optimal yields compared to their Au-incorporated counterparts (Supplementary Fig. 30c, d), and relocating Au NPs to the SiO₂ surface decreased product yields (Supplementary Fig. 31). These results indicate that the enhanced performance of Au/TiO₂@pSiO₂ is not due to O₂ enrichment from its high surface area. Instead, O₂ transport within silica shell is hindered in aqueous solution; however, the adverse effect of limited O₂ transport on overall product yield is mitigated by a combination of increasing O₂ pressure and effective O₂ utilization enabled by the supported Au cocatalysts. The function of Au as an electron acceptor for O₂ reduction was confirmed using in situ XPS, where the Au 4 f signal shifted to lower binding energies upon illumination (Supplementary Fig. 32).

The counter-intuitive observation that the SiO₂ shell promotes catalytic performance despite impeding O₂ transport implies the enhancement originates not from gaseous reactant enrichment. We therefore hypothesized that the origin is the confinement of water layers within the hydrophilic nanopores. To further decouple these factors, we systematically modified the optimal Au/TiO₂@pSiO₂-1.7 catalyst with methyltrimethoxysilane (MTMS) to increase its surface hydrophobicity while preserving the pore structure (Supplementary Figs. 33–36). Fourier transform infrared (FTIR) spectra confirmed the successful surface grafting (Supplementary Fig. 37). At low modification levels (e.g., 0.1–0.5 wt%), the shell thickness, pore size, and surface area of the catalyst remained unchanged (Supplementary Figs. 33, 34), while a higher 5% MTMS loading caused a decrease in surface area, likely attributed to pore blockage. In principle, a more hydrophobic surface is expected to enrich nonpolar reactants like CH₄ and O₂ and potentially accelerate the reaction^42,43,44. However, the catalytic results show the opposite trend: oxygenate yields decreased drastically compared to the unmodified, hydrophilic catalyst (Supplementary Fig. 36). Even a tiny 0.1 wt% MTMS modification, corresponding to an estimated Si–CH₃ surface coverage of <0.5% (Supplementary Table 7), was sufficient to significantly suppress the reaction. This extreme sensitivity to a miniscule fraction of hydrophobic groups provides strong evidence that the interfacial water environment, rather than the mass transport of gaseous reactants, plays the dominant role in the catalytic enhancement.

Generalization of the nanoporous confinement strategy

We explored this confinement strategy with different metal cocatalysts and semiconductors. Substituting Au with Pd, Pt, or Rh exhibited a similar trend: confined pores around metal-TiO₂ surface significantly promoted oxygenates and H₂O₂ yields compared to bare metal-TiO₂ catalysts, with smaller pores yielding higher productions (Fig. 4a, Supplementary Figs. 38, 39 and Supplementary Note 3). Among these, Pt/TiO₂@pSiO₂-1.7 achieved oxygenate yields of 32.7 ± 0.8 mmol g⁻¹ h⁻¹ and an AQY of 14.1 ± 0.3% for CH₄ oxidation at 365 nm, representing competitive performance relative to previously reported catalysts (Fig. 4, Supplementary Tables 4, 8). We further extended our study to other semiconductors, including BiVO₄, ZnO and anatase TiO₂ (A-TiO₂) (Fig. 4a, Supplementary Figs. 40–42 and Supplementary Note 3). Applying the porous silica coating improved oxygenate yields and H₂O₂ production for these catalysts. For example, Au/A-TiO₂@pSiO₂-1.7 achieved oxygenate and H₂O₂ formation rates approximately 5 and 21 times higher than those of Au/A-TiO₂, respectively (Fig. 4a). These results underscore the versatility of the nanoporous confinement design, demonstrating its broad applicability across different photocatalyst systems.

Fig. 4: Generalization of the photocatalyst confinement strategy.

Yields of oxygenated product and H₂O₂ from photocatalytic methane oxidation over core-shell photocatalysts with different metal cocatalysts and semiconductors. Reaction conditions: 10 mg catalysts, 100 mL H₂O, 2.5 MPa CH₄, 1 MPa O₂, 1 h reaction time, reaction temperature of 25 ± 3 °C, light source: 365 nm UV LED, 80 mW/cm². The error bars represent the standard deviations derived from three independent experiments. The vertical dashed lines categorize the different cocatalysts and semiconductor supports.

Key ROS for photocatalytic CH₄ oxidation

We also investigated the role of SiO₂ shell in photocarrier separation using photoluminescence (PL) and photocurrent measurements, and electrochemical impedance spectroscopy (EIS). Steady-state and time-resolved PL (TRPL) spectroscopy revealed that the silica shell has a negligible impact on photocarrier separation and lifetimes (Supplementary Figs. 43, 44). Furthermore, Au/TiO₂ showed a higher photocurrent and lower impedance compared to Au/TiO₂@pSiO₂ (Supplementary Figs. 45, 46). These results, together with findings on gas transport and surface modifications (Supplementary Note 2), demonstrate that improved performance in Au/TiO₂@pSiO₂ is not due to improved photocarrier separation or gas transport, but rather to effects associated with the structure of the confined water layers in contact with the semiconductors.

To unravel the underlying mechanism, we next sought to identify the key ROS responsible for methane activation and H₂O₂ formation within the confined environment. This can be achieved by introducing sacrificial agents that selectively quench different ROS and monitor the change in methane conversion^15,45. Both holes and •OH radicals are capable of activating methane^9,14,15. We used salicylic acid and (NH₄)₂C₂O₄ to scavenge •OH radicals and holes (h⁺), respectively¹⁵. The addition of salicylic acid nearly completely inhibited CH₄ conversion, showing a stronger suppression than (NH₄)₂C₂O₄ (Supplementary Fig. 47). This suggests that •OH radicals, rather than h⁺, are the main reactive species for CH₄ activation.

The Au/TiO₂@pSiO₂ catalyst with smaller pores produced more •OH radicals, as monitored by in situ electron paramagnetic resonance (EPR), which correlates with their increased performance in CH₄ conversion (Fig. 5a, b). Coumarin was also employed as a fluorescent probe to assess •OH production¹⁵, which further supports the findings from EPR (Supplementary Figs. 48, 49). Typically, there are two distinct pathways by which •OH radicals are formed: one involves H₂O oxidation by holes (1e⁻ WOR), while the other involves O₂ reduction by electrons (ORR) (Supplementary Fig. 1). To identify its origin, we performed ¹⁸O isotopic labeling with benzoic acid as a probe to capture •OH (Supplementary Fig. 50). Under H₂¹⁶O/¹⁸O₂ conditions, over 95% of products corresponded to H¹⁶OC₆H₄CO₂H (m/z = 137), while with H₂¹⁸O/¹⁶O₂, approximately 80% of products corresponded to H¹⁸OC₆H₄CO₂H (m/z = 139). This revealed that •OH radicals primarily originate from 1e⁻ WOR rather than from O₂ reduction over Au/TiO₂@pSiO₂-1.7 (Fig. 5c, Supplementary Fig. 51). In contrast, for Au/TiO₂ and TiO₂, approximately half of the •OH radicals derive from H₂O oxidation and the other half from O₂ reduction (Supplementary Figs. 52, 53). Thus, the enhanced 1e⁻ WOR to •OH radicals within confined SiO₂ pores are responsible for the improved CH₄ conversion over Au/TiO₂@pSiO₂.

Fig. 5: Key ROS detection for photocatalytic CH4 oxidation.

a In situ EPR spectra of DMPO-OH for monitoring the generation of •OH radicals over Au/TiO₂ and Au/TiO₂@pSiO₂-X. b Correlation between CH₄ conversion rate with •OH radical production rate over different photocatalysts. Black and red arrows denote the trends in methane conversion and •OH generation versus pore size, respectively. c ¹⁸O isotopic studies on the origins of •OH radicals over Au/TiO₂@pSiO₂-1.7, with benzoic acid used as a probe to capture •OH in situ. d In situ EPR spectra of DMPO-OOH for monitoring the generation of •OOH active species over Au/TiO₂ and Au/TiO₂@pSiO₂-X. e Correlation between H₂O₂ generation and •OOH radical production rate over different photocatalysts. Blue and green arrows denote the trends in •OOH and H₂O₂ generation versus pore size. f ¹⁸O isotopic studies on the origins of H₂O₂ over Au/TiO₂@pSiO₂-1.7.

EPR spectra of DMPO-OOH show that Au/TiO₂@pSiO₂ with smaller pores produced less •OOH radicals (Fig. 5d), consistent with their enhanced H₂O₂ production as the main product in ORR (Fig. 5e). Electron scavenger experiments using AgNO₃ or NaIO₃ and ¹⁸O₂/H₂¹⁸O isotope labeling confirmed that H₂O₂ is generated exclusively through ORR (Fig. 5f, Supplementary Figs. 54, 55). The ORR selectivity to H₂O₂ is 75 % over Au/TiO₂@pSiO₂-1.7, compared to 11% over Au/TiO₂ (Supplementary Fig. 13, Supplementary Note 4). The high H₂O₂ production and negligible •OOH production over Au/TiO₂@pSiO₂ suggest that the silica nanopores confinement alters the ORR pathway from a 1e⁻ process to a 2e⁻ process.

Thus, facilitated 1e⁻ WOR producing •OH and an altered ORR pathway account for the improved catalytic performance of Au/TiO₂@pSiO₂. Similar enhancements in •OH production from 1e⁻ WOR and altered ORR pathways were also observed with Pt/TiO₂@pSiO₂ (Supplementary Figs. 56–58, Supplementary Note 3). These results indicate that this alteration is a general phenomenon when water is spatially confined within nanopores. We hypothesize that both the accelerated 1e⁻ WOR and the modified ORR pathway result from the distinct structure of water confined within these nanopores compared to bulk water.

KIE investigations for CH₄ oxidation process

H/D kinetic isotope effect (KIE) studies were used to analyze whether PT is involved in the rate-determining step (RDS)^46,47,48, which would further shed light on the mechanism of altered reactivity and selectivity. For the 1e⁻ WOR that produces •OH radicals, Au/TiO₂@pSiO₂–1.7 shows a pronounced KIE value of 2.0–2.3, indicating a strong involvement of PT in the RDS. In contrast, Au/TiO₂ exhibits a low KIE value of 1.1–1.2 (Fig. 6a, Supplementary Figs. 59, 60), suggesting a distinctly different 1e⁻ WOR pathway. The absence of KIE for Au/TiO₂ is consistent with the widely accepted stepwise proton-electron transfer (SPET) mechanism^32,49,50, where water readily deprotonates on TiO₂ before being oxidized by the photogenerated hole, making the electron transfer step (OH⁻ + h⁺ → •OH) the RDS. Conversely, the large KIE for Au/TiO₂@pSiO₂–1.7 suggests that molecular water confined in the nanopores acts directly as the hole acceptor, generating •OH radicals through a concerted proton-electron transfer (CPET) pathway (H₂O + h⁺ → •OH + H⁺). As a result, the slower dissociation and deuteron transfer in D₂O compared to H₂O greatly reduce the •OH radical production.

Fig. 6: H/D KIE and in situ DRIFTS studies over Au/TiO₂@pSiO₂ catalysts.

a Comparison in •OH radical production over Au/TiO₂ and Au/TiO₂@pSiO₂-1.7 measured in H₂O and D₂O. b Comparison in CH₄ conversion rate over Au/TiO₂ and Au/TiO₂@pSiO₂-1.7 measured in H₂O and D₂O. c Comparison in H₂O₂ generation rate over Au/TiO₂ and Au/TiO₂@pSiO₂-1.7 measured in H₂O and D₂O. d DRIFTS analysis of the OH stretching band of the water in different photocatalysts. e DRIFTS analysis of the HOH bending band of the water in different photocatalysts. In (d, e), colored curves represent the deconvoluted components of specific HB environments (blue: DA; orange: DDAA; green: DDA; purple: DAA); The black lines denote the raw data, while the hollow circles (yellow in (d), gray in (e)) represent the cumulative fitted curves. f Spectroscopically observable hydrogen-bonding configurations of H₂O. g Mean HB state of water in Au/TiO₂ and Au/TiO₂@pSiO₂-X samples derived from DRIFTS. Dashed lines represent hydrogen bonds between O (red) and H (gray). h Free energy profiles for water dissociation and •OH generation/deprotonation on the anatase TiO₂(101) surface in bulk water. i Corresponding free energy profiles in confined water. Both energy profiles (h, i) were calculated assuming a hole concentration of C_h+ = 10⁻⁹ ML.

At a molecular level, this mechanistic divergence stems from the preferential reactivity of photogenerated holes with either molecular water or hydroxide anions (OH⁻). The SPET mechanism on Au/TiO₂ suggests extensive water dissociation, populating the surface with OH⁻ that are highly reactive toward holes. In contrast, the shift to a CPET mechanism in Au/TiO₂@pSiO₂–1.7 indicates that confined water resists dissociation, reacting instead in its molecular form. This change can be attributed to the weakened HB network and reduced solvation effect within the nanopores^26,32. The elevated •OH production over Au/TiO₂@pSiO₂–1.7 suggests that the CPET mechanism provides a more effective route for generating the •OH radicals necessary for CH₄ activation than the two-step SPET pathway. Thus, confinement-driven changes in water reactivity and PCET pathways are key to enhanced CH₄ conversion for Au/TiO₂@pSiO₂–1.7.

As the WOR-producing •OH is key to CH₄ activation, its KIE influences the KIE of overall CH₄ oxidation. Consequently, Au/TiO₂ does not show KIE effect for CH₄ conversion, and Au/TiO₂@pSiO₂-1.7 shows KIE value of 1.3–1.6 (Fig. 6b, Supplementary Figs. 61–64). The KIE decreases over time for Au/TiO₂@pSiO₂–1.7, likely due to H/D exchange between CH₄ and D₂O during reactions (Supplementary Fig. 64, Supplementary Note 5).

On the ORR side, a large KIE of 2.97 is also observed for Au/TiO₂@pSiO₂–1.7 (Fig. 6c), which is expected for the 2e⁻ process that produces H₂O₂ from oxygen and proton. In stark contrast, a reverse KIE for H₂O₂ production was observed over Au/TiO₂, indicating that the confined water significantly modulates the PT and ET dynamics. Notably, the observed shift in ORR pathway from the 1e⁻ process for Au/TiO₂ to the 2e⁻ process for Au/TiO₂@pSiO₂ can also be attributed to facilitated proton transfer due to water confinement. Compared to the 1e⁻ process, the 2e⁻ process is thermodynamically more viable but requires simultaneous transfer of two protons. Therefore, efficient proton transfer, along with increased protons generated from the enhanced WOR process, is responsible for the shift to the 2e⁻ process and enables efficient production of H₂O₂.

Confined water characterization

The weakened HB network of the confined water was analyzed by in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) measurement at various relative humidity (RH) levels (Fig. 6d–g, Supplementary Figs. 65, 66). The broad OH stretching band (2800−3600 cm⁻¹) is sensitive to the HB network configuration and can be deconvoluted into distinct Gaussian sub-bands (Fig. 6f)^31,34. From Au/TiO₂ to Au/TiO₂@pSiO₂-1.7, water in the smaller pore shows a significantly larger portion of DA to DDAA ratio along with a lower average number of HBs (<N_HB>) (Fig. 6g, Supplementary Table 9), indicating a weakened HB network. A similar trend was observed in the HOH bending (1500–1800 cm⁻¹) region (Fig. 6e), where deconvolution of the spectra can give the ratio between two peaks at 1620 cm⁻¹ and 1680 cm⁻¹, attributed to DA and DDAA water clusters³¹. These results suggest that H₂O molecules within smaller pores adopt a more linear arrangement with a weaker HB network, leading to more isolated water species that are prone to oxidation and a reduced solvation that better stabilizes the semi-hydrophobic •OH radicals.

To rationalize the role of water structure, we performed MD simulations using MLPs on the anatase TiO₂(101) surface^44,51. Two different water environments were modeled: one containing 128 water molecules to represent the bulk system and the other with an interfacial layer of 39 molecules designed to mimic the restricted content of the confined system within the 1.7 nm pores (Supplementary Fig. 67a–c, Supplementary Data 1). Simulated water structures yielded DA/DDAA ratios in reasonable agreement with the experimental DRIFTS data (Fig. 6g, Supplementary Fig. 67d), supporting the qualitative consistency of our computational model. Using this framework, we computed the energetics of the SPET and CPET pathways, corresponding to the oxidation of surface hydroxide (*OH_t⁻, reactions (1)–(3)) and adsorbed molecular water (*H₂O_ad, reaction (4)), respectively (Supplementary Fig. 68, Supplementary Data 1)⁴⁹.

Although the reaction barriers for hole-driven oxidation are low under standard conditions (barrierless for *OH⁻; ~12.5 kJ/mol for *H₂O), the reaction rate is governed by the extremely low concentration of surface-reaching holes (C_h+ = 10⁻⁹ monolayer (ML)) under realistic photocatalytic conditions^49,50,52,53. To account for this, we converted this concentration constraint into an equivalent energetic penalty, which adds an additional barrier of ~51.3 kJ/mol to the hole-involving steps (Supplementary Note 6). In bulk water, exergonic dissociation (−5.2 kJ/mol) populates the surface with *OH_t⁻, resulting in a lower effective barrier for SPET (46.1 kJ/mol) compared to CPET (63.8 kJ/mol) (Fig. 6h, Supplementary Figs. 69, 70). Under confinement, however, water dissociation becomes unfavorable (20.3 kJ/mol), which stabilizes molecular *H₂O (Fig. 6i, Supplementary Figs. 69, 70). This raises the effective barrier of SPET to 71.6 kJ/mol, thereby shifting the reaction to the CPET pathway. This mechanistic distinction is robust across a wide range of hole concentrations, even at a concentration of 10⁻⁴ ML (Supplementary Fig. 71). These computational findings align well with our experimental KIE results.

Furthermore, the nanoconfined environment enhances the stability of the generated •OH radicals. In bulk water, •OH radical is prone to deprotonation to form surface oxygen radicals (•O_t⁻) through reaction (5) (Supplementary Fig. 68)⁵³, a process with a barrier (45.2 kJ/mol) slightly lower than that of radical production, facilitating rapid consumption (Fig. 6h). However, the disrupted hydrogen-bond network in the confined system hinders this process. Quantitative analysis shows a substantial increase in both the free energy and activation barrier for this step, rising from ~45 kJ/mol in bulk to >60 kJ/mol under confinement (Fig. 6i). This suggests that the deprotonation barrier exceeds the •OH production barrier in nanopores, potentially extending the lifetime of semi-hydrophobic •OH species and facilitating their participation in CH₄ activation.

Therefore, both enhanced methane conversion and H₂O₂ production of the Au/TiO₂@pSiO₂ can be explained by the weakened HB network of the confined water in porous silica (Fig. 7): (a) for h⁺ driven WOR, confined water with disrupted HB networks suppresses dissociation and preferentially reacts with h⁺ in its molecular form, causing a shift from SPET to CPET mechanism that generates more •OH radical for CH₄ activation; (b) for e⁻ driven ORR, confined water with weakened HB transfers proton more efficiently, enabling a shift from 1e⁻ to 2e⁻ process that directly produces H₂O₂.

Fig. 7: Proposed mechanistic pathways and the role of confined water.

The schematic illustrates the distinct PCET pathways over Au/TiO₂ in bulk water (left) versus Au/TiO₂@pSiO₂-1.7 with confined water (right). Over Au/TiO₂, water oxidation to •OH radical proceeds via SPET, and O₂ reduction follows a 1e⁻ pathway; In contrast, the confined water with disrupted HB networks promotes the CPET pathway for water oxidation to •OH radicals and 2e⁻ pathway for O₂ reduction. H atom: light gray; O atom: ice blue.

Discussion

In summary, we have demonstrated a core-shell photocatalyst design that leverages spatial confinement to modulate water structure, leading to a significant enhancement in both CH₄ conversion and H₂O₂ production. The confined water with weaker HB networks significantly modulates the PT and ET dynamics in the reaction involving photogenerated electrons and holes. Specifically, the confined water suppresses dissociation and preferentially reacts with h⁺ in WOR, causing a shift from SPET to CPET mechanism and generating more •OH radicals to activate CH₄. Concurrently, the weakened HB network facilitates more efficient proton transfer in ORR, enabling a shift from the 1e⁻ to the 2e⁻ process and directly producing H₂O₂. This study provides a fundamental understanding of how interfacial water structure can be engineered to optimize photocatalytic performance, opening new avenues for photocatalyst design.

Methods

Materials

All reagents were commercially obtained without purification. Titanium dioxide (P25, ≥99.5%), gold (III) chloride trihydrate (HAuCl₃•3H₂O, 99.99%), chloroplatinic acid (H₂PtCl₆, ≥99.5%), palladium chloride (PdCl₂, 99.9%), and Sodium sulfate (Na₂SO₄, ≥99%) were purchased from Sigma Aldrich. Rhodium(III) chloride trihydrate (RhCl₃•3H₂O, 98%), methyltrimethoxysilane (98%), dodecyltrimethylammonium bromide (C₁₂TAB, ≥99%), cetyltrimethylammonium bromide (C₁₆TAB, ≥99%), n-hexane (95%), tetraethyl silicate (TEOS, ≥99%), ammonium acetate (≥99%), coumarin (≥98%), 7-hydroxycoumarin (≥99%), acetylacetone (≥99.5 %), benzoic acid (≥99%), deuterium water (≥99.9atom%), H₂¹⁸O (≥97%), hydroxybenzoic acid (≥99%), 5,5-dimethyl-1-pyrrolidinium N-oxide (DMPO, ≥97%), Tetra-methylpiperidine N-oxide (TEMPO, ≥99%) were purchased from Aladdin. Ammonium hydroxide (NH₃•H₂O, 25–28%), dimethyl sulfoxide (≥99.9%), ethanol (99%), sodium borohydride (NaBH₄, 98%), sodium iodate (NaIO₃, 99.8%), ammonium oxalate (NH₄)₂C₂O₄, 99.5%) and salicylic acid (HOC₆H₄COOH, 99%) were purchased from Sinopharm Chemical Reagent Co., Ltd., China. The Ag/AgCl electrode was purchased from Shanghai Chenhua.

Preparation of M/TiO₂ photocatalyst

Metal cocatalyst (Au, Pt, Pd, Rh) was loaded onto TiO₂ (P25) using a simple NaBH₄ reduction method^9,38. Typically, 50 mg of TiO₂ was ultrasonically dispersed in 30 mL of deionized water, after which a specific amount of metal precursor solution (HAuCl₃•3H₂O, H₂PtCl₆, PdCl₂, RhCl₃•3H₂O) was added. Following 10 min of stirring, 2 mL of freshly prepared 0.1 M NaBH₄ was added. After an additional 30 min of stirring, the product was collected by centrifugation, washed with deionized water, and dried in an oven at 60 °C. The metal loadings for the catalyst are 0.1 wt%, unless specifically noted.

Preparation of M/TiO₂@pSiO₂-X

To synthesize M/TiO₂@pSiO₂-1.7 (M denoted to Au, Pd, Pt, Rh), 100 mg of C₁₂TAB was used as the templating agent, which was dissolved in 10 mL of deionized water and 40 mL of anhydrous ethanol. Subsequently, 50 mg of M/TiO₂ was added, sonicated for 30 mins, followed by the addition of TEOS. After 15 min of stirring, 0.5 mL ammonium hydroxide solution (25–28%) was added, and the mixture was stirred overnight. The product was collected via centrifugation, washed twice with DI H₂O and ethanol, dried at 60 °C, and calcined at 550 °C for 2 h in static air to remove the C₁₂TAB templates. TEOS volume controlled the SiO₂ thickness; 300 μL produced an 8 nm SiO₂ shell. For the M/TiO₂@pSiO₂-2.6 sample, the synthesis process was identical, except C₁₆TAB was used as the template. The preparation of M/TiO₂@pSiO₂-3.1 followed the same procedure as M/TiO₂@pSiO₂-2.6, with the exception that 10 mL of n-hexane was added to the mixture prior to the addition of TEOS. The metal loadings for the catalysts are 0.1 wt%, unless specifically noted.

Preparation of TiO₂@pSiO₂ and TiO₂@pSiO₂-Au

The synthesis procedure for TiO₂@pSiO₂ was the same as the synthesis of M/TiO₂@pSiO₂ except using TiO₂ as the starting material. The synthesis procedure for TiO₂@pSiO₂-Au was the same as the synthesis of Au/TiO₂, except using TiO₂@pSiO₂ as the supporting material.

Hydrophobic modification of Au/TiO₂@pSiO₂-1.7

The Au/TiO₂@pSiO₂-1.7 catalyst (50 mg) was ultrasonically dispersed in 30 mL of n-hexane, followed by the addition of methyltrimethoxysilane (MTMS) at varying mass ratios (y wt% relative to the catalyst’s mass). The resulting suspension was ultrasonicated for 2 h and subsequently stirred at room temperature for 24 h. The solid product was then collected by centrifuge, washed sequentially with n-hexane (three times) and anhydrous ethanol (three times), and dried at 80 °C. The obtained catalyst was denoted as Au/TiO₂@pSiO₂-y wt% MTMS. The y wt% is the added MTMS amount; the actual MTMS loading should be lower.

Characterization

The crystal structure of the catalysts was characterized using a powder X-ray diffractometer D2 (Bruker, Switzerland) with Cu Kα radiation (λ = 1.542 Å). Morphologies of the photocatalysts were studied using TEM (JEM-1400plus) and scanning electron microscopy (SEM, JEOL JEM-7800F). High-resolution TEM (HRTEM) images, HAADF-STEM images, and corresponding EDS maps were collected using a JEOL JEM-F200 field-emission microscope operated at 200 kV. Nitrogen sorption isotherms were collected on a BELSORP-Max 2 apparatus. The metal loading in the photocatalysts was determined by an inductively coupled plasma optical emission spectrometer (ICP-OES, Icap7400, Thermo). UV-Vis DRS were recorded on a Cary 5000 spectrometer (Agilent) equipped with an integrating sphere. The absorption spectra were obtained using the Kubelka–Munk transformation. XPS were collected using a Thermo Fisher ESCALAB 250Xi spectrometer with monochromatic Al-Kα X-rays as the excitation source. All spectra were calibrated to the C 1 s peak at 284.8 eV. In-situ XPS spectra were measured at BL02B01 in the Shanghai Synchrotron Radiation Facility (SSRF), where the data were collected under dark or during irradiation. All spectra were calibrated to an Au foil. The hydrophobicity of the as-prepared catalyst was measured using a contact angle goniometer (SL200KS). Steady-state PL spectra were recorded on the Horiba Fluorolog-3 Spectrometer with an excitation wavelength of 350 nm. Time-resolved PL spectra were measured on the same instrument with an excitation wavelength of 367 nm. All electrochemical measurements (photocurrent, EIS, and Mott-Schottky tests) were performed on a Gamry Instrument Interface 1010E workstation using a three-electrode system. A 1 M Na₂SO₄ solution (pH = 6.8 ± 0.1) served as electrolyte, with Ag/AgCl (3 M KCl) as the reference electrode, platinum plate as the counter electrode and catalysts loaded on indium-tin oxide (ITO) as the working electrode. Automatic iR compensation was applied for Mott-Schottky measurements.

The solution resistance (R_s) was determined via EIS at open-circuit potential (100 kHz –3 Hz, 10 mV amplitude). The high-frequency intercept of the Nyquist plot yielded an average R_s of 19.2 ± 0.2 Ω. All potentials in this work are converted to the RHE scale by E (vs RHE) = E (vs Ag/AgCl) + 0.210 V + 0.0591 × pH.

Photocatalytic activity measurements

Evaluation of photocatalytic activity for CH₄ oxidation was performed using a high-pressure autoclave (250 mL) with an optical sapphire window. After ultrasonically dispersing the catalyst (10 mg) in deionized water (100 mL) for 30 min, the reactor was purged with argon (99.999%). The gaseous reactants, CH₄ (99.999%) and O₂ (99.999%), were subsequently introduced into the vessel at partial pressures of 2.5 MPa and 1.0 MPa, respectively. The reactor was irradiated from the top using an LED lamp (365 nm, PLS-LED100C, Beijing Perfectlight Technology Co., Ltd.). The photocatalytic reaction was conducted for 1 h at 25 °C. A thermocouple was inserted into the solution to directly detect the temperature of the solution. The photoreactor was directly connected to a gas chromatograph (Agilent GC 8890) equipped with a flame ionization detector (FID) and a methanizer for the gas product analysis of CO₂ and ethane. The liquid products (CH₃OOH, CH₃OH, and HCOOH) were analyzed by ¹H-NMR (Bruker 400 MHz), with DMSO added as an internal standard.

HCHO quantification

The amount of HCHO was quantified using the acetylacetone color-development method^37,54. Typically, 0.5 mL of the sample liquor, 2 mL of deionized water and 0.5 mL of reagent solution (consisting of 15 g of ammonium acetate, 0.3 mL of acetic acid, and 0.2 mL of acetylacetone dissolved in 100 mL of H₂O) were mixed and placed in a water bath at 35 °C for 40 min. The absorption value of the mixture at 412 nm was then measured using UV-visible absorption spectroscopy (PerkinElmer, Lambda 265).

H₂O₂ quantification

The concentration of produced H₂O₂ was determined using the titanium oxalate spectrophotometric method⁵⁵. To prepare the test solution, 0.636 g of potassium titanium oxalate and 0.02 mL of concentrated sulfuric acid were dissolved in 99.98 mL of DI H₂O. Then, 1 mL of the above test solution was mixed with 1 mL of the reaction sample. The absorption of the mixture was measured at 390 nm using UV-visible absorption spectroscopy (PerkinElmer, Lambda 265).

CH₄ conversion rates and oxygenates selectivity calculation

The methane conversion and oxygenates selectivity in this process are calculated according to the following equations:

$${Conversion}\left(C{H}_{4}\right)=\frac{{{mol}}_{{total}\,{formed}\,{products}}}{{{mol}}_{{initial}\,{methane}}}\,\times 100\%$$

$${Oxygenates}\,{selectivity}=\frac{{{mol}}_{{produced}\,C{H}_{3}{OH},\,C{H}_{3}{OOH},\,{HCHO}\,{and}\,{HCOOH}}}{{{mol}}_{\,{total}\,{formed}\,{products}}}\times 100\%$$

The initial methane amount in the system is calculated according to the following equation (V is the volume of the headspace of the reactor):

$${Initial}\,{methane}\,\left({{\rm{mol}}}\right)=\,\frac{{P}_{{methane}}\times V\,}{{{\rm{R}}}\times T\,(298{{\rm{K}}})}$$

AQY of CH₄ conversion was calculated according to the following equation:

$${AQY}=\frac{N\left({electrons}\right)}{N({photons})}\times 100\%$$

where N(electrons) and N(photons) represent the number of reacted electrons and the number of incident photons, respectively. N(photons) = IAt/E, where I, A, t and E represent incident light intensity (W/cm²), irradiation area (cm²), light incident time (s) and photo energy (J), respectively. N(electrons) was calculated based on the following reported work^9,20: N(electrons) = n(CH₃OOH) + 3n(CH₃OH) + 5n(HCHO) + 7n(HCOOH) + 9n(CO₂), where n(CH₃OOH), n(CH₃OH), n(HCHO), n(HCOOH) and n(CO₂) represent the number of moles of produced CH₃OOH, CH₃OH, HCHO, HCOOH and CO₂, respectively. The AQY for H₂O₂ production is calculated similarly, with N(electrons) = 2n(H₂O₂), where n(H₂O₂) represents the number of moles of H₂O₂ produced. The light intensity was calibrated using a PL-MW2000 optical power meter (Beijing Perfectlight Technology Co., Ltd.). Error bars represent the standard deviation (SD) from three independent experiments. Data points without error bars represent single measurements.

Catalyst cycling test

To study the reusability of the catalyst, the solid catalyst was separated by centrifugation after each reaction run. The catalysts were re-used in the next run after drying at 90 °C overnight under vacuum and annealing at 300 °C in air to remove any adsorbed organic species. Aq

CH₄-TPD experiments

CH₄-TPD were performed using a Chemisorption analyzer (ChemStar). Catalysts were pretreated at 150 °C for 30 min under a He flow of 30 sccm. After cooling to 50 °C, adsorption was conducted under a pure CH₄ atmosphere for 120 min. The sample was then purged with He for 30 min, and the TPD profile was recorded at a heating rate of 5 °C/min under a He flow.

Analysis of photogenerated •OH radicals

The •OH production was measured using PL with coumarin, which reacts with •OH to form fluorescent 7-hydroxycoumarin (7-HC) (Supplementary Fig. 48). 2 mg catalyst was dispersed in 100 mL of 1.0 mM coumarin solution in the dark, irradiated for 10 min, then centrifuged and analyzed with a Horiba FluoroLog-3 spectrofluorometer at 332 nm excitation.

Isotope experiments for identifying the origin of products

For carbon source investigation with isotopically labeled ¹³CH₄: 10 mg Au/TiO₂@pSiO₂ was dispersed in 5 mL H₂O, and the reactor was purged with Ar to remove air. The photoreactor was then filled with 0.55 MPa ¹³CH₄ (99%), 0.45 MPa ¹²CH₄ (99.999%) and 0.2 MPa O₂ (99.999%). After 2 h of light irradiation, gas products were analyzed by gas chromatograph-mass spectroscopy (GC-MS, Shimadzu), and liquid products by ¹H-NMR.

For oxygen source investigation with ¹⁸O isotope labeling with ¹⁸O₂/H₂¹⁶O and ¹⁶O₂/H₂¹⁸O: 10 mg Au/TiO₂@pSiO₂ was dispersed in 10 mL H₂¹⁶O or (H₂¹⁸O), and the reactor was purged with Ar. The photoreactor was charged with 2.5 MPa CH₄ (99.999%) and 0.5 MPa ¹⁸O₂ (98%) (or ¹⁶O₂). Products were analyzed by GC-MS after 2 h of light irradiation.

EPR measurements

In situ electron paramagnetic resonance (EPR) spectra were acquired using a Bruker ESR5000 spectrometer equipped with a 300 W xenon lamp (PLS-SXE300 + /UV, Perfectlight) for illumination. 5,5-Dimethyl-1-pyrroline-N-oxide (DMPO) served as the radical trap. For •OH radical detection, 5 mg of catalyst was dispersed in 5 mL of distilled H₂O in the dark and purged with ultrapure argon (99.999 vol.%) for 2 min. For •OOH radical detection, 5 mg of catalyst was dispersed into a 5 mL of DMSO with O₂ bubbling for 2 min.

Investigation of the photochemical pathway of H₂O₂ production

To determine if H₂O₂ production originates from WOR or ORR, liquid samples from ¹⁸O₂ or H₂¹⁸O isotope CH₄ oxidation experiments were analyzed. As shown in Supplementary Fig. 55, 2 mL of the liquid samples was injected into a sealed vial (Vial A) and purged with He for 30 min. Another vial (Vial B), containing 200 mg of MnO₂ and 5 mL H₂O, was also purged with He for 30 min. Then, 2 mL from Vial A was injected into Vial B. After complete decomposition of H₂O₂ over MnO₂, 1 mL of gas from the headspace of Vial B was extracted for GC-MS analysis.

Investigation of the photochemical pathway for •OH generation

To analyze photochemical pathways of •OH production,¹⁸O₂ isotope experiments and H₂¹⁸O isotope experiments were conducted. In the ¹⁸O₂ experiments, 2 mg of catalyst was dispersed in 1 mL of 1 mM benzoic acid solution (H₂¹⁶O as solvent). The reactor was purged with Ar to remove air and filled with 0.1 MPa ¹⁸O₂ (98%). The reaction was carried out under UV irradiation for 600 s. The products were analyzed by liquid chromatography–mass spectrometry (LC–MS, Thermofisher). Similarly, for the H₂¹⁸O isotope experiments, the setup was identical except using benzoic acid dissolved in H₂¹⁸O and 0.1 MPa ¹⁶O₂.

KIE measurements

For •OH radical generation, 5 mg of catalyst were dispersed in 2 mL of either D₂O or H₂O, and in situ EPR measurements were performed to track the formation of •OH radicals. For CH₄ conversion, 10 mg of catalysts were dispersed in 10 mL of D₂O or H₂O. The reactor vessel was then pressured to 2.5 MPa CH₄ and 1 MPa of O₂, followed by light irradiation, and liquid products were collected at specific time intervals for analysis. For H₂O₂ generation, 10 mg of catalysts were dispersed in 10 mL of D₂O or H₂O. The reactor vessel was then pressured to 1 MPa of O₂, followed by light irradiation. The resulting liquid product was collected for analysis.

Active species scavenging experiments

The experiments were conducted by adding different types of radical scavengers was used to understand the role of radical species involved in the photocatalytic CH₄ oxidation. Tetra-methylpiperidine N-oxide (TEMPO) as a scavenger for all radicals, and ammonium oxalate ((NH₄)₂C₂O₄), sodium iodate (NaIO₃), salicylic acid (HOC₆H₄COOH) as scavengers for photogenerated holes, photogenerated electrons, hydroxyl radicals, respectively¹⁵. In a typical experiment, 10 mg photocatalyst was dispersed in 100 mL distilled water containing 1 mmol radical scavengers. After being finely sealed, the suspension was purged with ultrapure Ar (99.999%) for 20 min to completely remove air. Then O₂ (99.999%) and CH₄ (99.999%) were injected to acquire the desired pressure. The reaction was conducted for 10 mins. Products including CH₃OH, CH₃OOH, HCHO, HCOOH and CO₂ were measured according to the method shown above.

In situ DRIFTS measurements

In situ DRIFTS measurements were performed using a Thermo Scientific iS50 FTIR spectrometer equipped with a liquid-nitrogen-cooled MCT/A detector. An infrared cell (Harrick) with ZnSe windows was used for measurements. Each spectrum was recorded by averaging 64 scans at a resolution of 4 cm^–1. Samples were first pretreated at 673 K (5 K min⁻¹) for 30 min in flowing O₂ (20 mL min⁻¹) to remove impurities adsorbed on the surface. Background spectra were then taken at 303 K in dry Ar. Wet Ar (Ar/H₂O with controlled humidity) was subsequently introduced (20 mL min⁻¹), and spectra were acquired at the 303 K. By using the following equation proposed by previous work^31,34, we determined the average number of hydrogen bonds of each water molecule:

$$ < {N}_{{HB}} > =4\times {A}_{{DDAA}}+3\times {A}_{{DDA}}+3\times {A}_{{DAA}}+2\times {A}_{{DA}}$$

MLP-MD simulations

All MD simulations were conducted using LAMMPS⁵⁶ with an implementation of an MLP trained on the optB88-vdW density functional, which accurately captures the energetics and dynamics of water dissociation on TiO₂ surfaces. This approach, developed by Zeng et al⁵¹, enables large-scale simulations of water-TiO₂ interfaces with near DFT accuracy. As MLPs are trained on DFT data, they suffer from the same limitations as DFT⁵⁷. A timestep of 1 fs was used for all simulations. Free-energy landscapes at 300 K were evaluated using metadynamics, as implemented in LAMMPS patched with PLUMED⁴⁹. Metadynamics simulations were performed to explore the free-energy surfaces associated with •OH generation and subsequent transformations. The PLUMED input file specifying the collective variables (CVs) is provided in Supplementary Table 10. All simulations were carried out in the NVT ensemble using a Nosé-Hoover thermostat. Each independent metadynamics trajectory was run for 20 ns. Following G. Bussi et al⁵⁸, we obtained the average 1D free energy profile after t_fill that all the free energy minima are filled with Gaussian hills, and then computed the values of the free energy barrier and reaction free energy. According to the free energy barrier and reaction free energy curves shown in Fig. 6h, i, Supplementary Fig. 71, we determined the t_fill as 10 ns for these simulations.

Structure of Water/Anatase TiO₂ (101) interface in MD simulations

The anatase TiO₂(101) surface exhibits a terraced morphology characterized by step-edge twofold-coordinated oxygen atoms (O_2c), which is critical for proton transfer, and fivefold-coordinated titanium sites (Ti_5c) for water adsorption. The slab size of p(1 × 3) TiO₂(101) surface is 10.5 Å × 11.5 Å × 48.6 Å (Supplementary Fig. 67). This surface was represented using a ten-trilayer O-Ti-O slab comprising 180 atoms, a thickness validated to eliminate slab-size artifacts in surface energetics and water adsorption behavior. To mimic the bulk water environment surrounding Au/TiO₂, a cell containing 128 water molecules was constructed, corresponding to a density of 1 g cm⁻³. To model the confined water for Au/TiO₂@pSiO₂-1.7, a system with 39 water molecules, corresponding to a three-layer hydration film with reduced water content, was constrained at the interface between the TiO₂(101) surface and a virtual silica pore wall. The HB network of water molecules was analyzed using a workflow developed on top of the MD Analysis toolkit⁵⁹. For each trajectory snapshot, the HB topology was determined by first assigning to each oxygen atom its two nearest hydrogen atoms. The remaining hydrogen atom in each water molecule was subsequently associated with the nearest oxygen atom based on the shortest O–H distance. HBs were identified using standard geometric criteria, namely an O₁–O₂ distance of <3.0 Å and an O₁–H₂ distance of <2.5 Å, where subscripts 1 and 2 refer to atoms belonging to the hydrogen-bond acceptor and donor, respectively. Interfacial hydrogen bonds were quantified using a donor-acceptor distance cutoff of 3.0 Å together with an angular cutoff of 30°, following established protocols⁶⁰.