Introduction

Methane has been considered as an abundant and promising feedstock for future energy and chemical productions, especially after discovery of large reserves of shale gas and methane hydrate1,2. Direct conversion of methane to value-added chemicals has been attracting great interest, however, due to a stable C–H bond, a small polarizability, a high ionization potential and a low electron affinity stability of methane, it remains as a long-standing challenge3,4,5. Harsh reaction conditions, such as high temperatures6,7,8 and/or high pressures9,10,11,12,13,14,15, are required for traditional heterogeneous thermocatalytic selective conversion of methane. Recently, photocatalysis has been explored for selectively converting methane mainly to valuable liquid oxygenates at room temperature and ambient pressure16,17,18,19,20,21,22,23,24.

H2O2 is widely used as an oxidant for photocatalytic selective conversion of methane over oxide-based photocatalysts. Photocatalytic activation of H2O2 by photo-generated electrons into ·OH radicals (0.06 eV vs RHE)25 or ·OOH radicals (−0.38 eV vs RHE)26, depending on the conduction band edges of semiconductor photocatalysts, is generally considered as the key step. However, photocatalytic activation of H2O2 by photo-generated holes into O2 usually occurs facilely25, which strongly competes and decreases the utilization efficiency of H2O2 for the methane conversion, defined as the ratio of the H2O2 amount consumed for methane conversion against the total consumed H2O2 amount. So far, the highest utilization efficiency of H2O2, in the means of ·OH radicals, was reported as 72.3% in photocatalytic CH4 conversion over a Fenton-type Fe-based catalyst21. Adsorption of H2O2 molecules on photocatalyst surfaces is a prerequisite for occurrences of photocatalytic reactions. Here, we show O2 additive as a general strategy to enhance utilization efficiencies of H2O2 for the photocatalytic CH4 conversion over oxide-based photocatalysts up to 93.3% by suppressing the H2O2 adsorption on photocatalyst surfaces and the consequent side reaction of photocatalytic H2O2 dissociation into O2.

Results

Synthesis and structural characterizations

Anatase TiO2 nanocrystals (NCs) predominantly enclosed by the {001} facets (denoted as TiO2{001}), the {100} facets (denoted as TiO2{100}) and the {101} facets (denoted as TiO2{101}) were prepared following well-established recipes27. XRD patterns, TEM and HRETM images of as-synthesized various TiO2 NCs (Fig. 1a, Supplementary Fig. 1) agree with those reported previously27. TiO2 NCs-C3N4 composites were prepared by calcination of mixture of calculated amounts of dicyandiamide (C2H4N4) and TiO2 NCs in Ar at 550 °C and denoted as TiO2 NCs-C3N4-x, in which x was the actual TiO2:C3N4 mole ratio acquired by TGA analysis (Supplementary Fig. 2 and Table 1). TEM, HRTEM and element mapping images (Fig. 1b–d, Supplementary Fig. 3a–c) show that various TiO2 NCs preserve their original morphologies and form smooth anatase TiO2-g-C3N4 interfaces. We failed to observe clear lattice fringes of g-C3N4 in the HRTEM images (Supplementary Fig. 3d) likely due to the strong damage effect of high-energy electron beam on the structure of g-C3N4, but its presence in the TiO2 NCs-C3N4 composites is identified by XRD patterns (Supplementary Fig. 3e) and XPS spectra (Supplementary Fig. 3f).

Fig. 1: Photocatalysts and photocatalytic performance.
figure 1

a TEM image of TiO2{001}. b TEM, (c) HAADF and (d) element mapping images of TiO2{001}-C3N4−0.1. e H2O2 decomposition rate, H2O2 decomposition and O2 selectivity of photocatalytic H2O2 decomposition over TiO2{001} and TiO2{001}-C3N4−0.1 under the reaction condition of 165 μL H2O2 + 20 mL H2O in Ar or 10%O2/Ar. CH4 conversion rate (rCH4), yield(YHCOOH) and selectivity (SHCOOH) of formic acid, selectivity of oxygenates (SOxygenates), and H2O2 utilization efficiency (EH2O2) of photocatalytic CH4 conversion over (f) 20 mg TiO2{001} under the reaction condition of 8%CH4 + 92%Ar + 110 μL H2O2 + 20 mL H2O or 8%CH4 + 1.6%O2 + 90.4%Ar + 110 μL H2O2 + 20 mL H2O for 5 h and over (g) 20 mg TiO2{001}-C3N4−0.1 under the reaction condition of 8%CH4 + 92%Ar + 165 μL H2O2 + 20 mL H2O or 8%CH4 + 4%O2 + 88%Ar + 165 μL H2O2 + 20 mL H2O for 8 h at 298 K. Source data are provided as a Source Data file.

Photocatalytic performance

H2O2 barely decomposes at 300 K over various oxides (P25, ZnO, Fe2O3, WO3, CuO and V2O5) without Xe light illumination. Under Xe light illumination, H2O2 decomposition predominantly to O2 occurs slightly in an Ar atmosphere without the presence of oxides but substantially with the presence of oxides (Supplementary Table 2), demonstrating facile occurrence of photogenerated holes-mediated H2O2 decomposition to O2. Photocatalytic H2O2 decomposition over various TiO2 NCs was observed dependent on the surface structure. TiO2{001} NCs exhibit the lowest photocatalytic activity and O2 selectivity while TiO2{101} NCs exhibit the highest (Supplementary Table 3). C3N4 is poor in photocatalytic H2O2 decomposition, and comparing corresponding TiO2 NCs, TiO2 NCs-C3N4 composites exhibit much decreased photocatalytic activity and O2 selectivity (Supplementary Table 3). Interestingly, we found that photocatalytic H2O2 decomposition over oxides gets greatly suppressed in an O2/Ar atmosphere, together with slight decrease of O2 selectivity; moreover, such an O2 suppress effect varies with the structures of TiO2 NCs and TiO2 NCs-C3N4 composites (Supplementary Tables 2 and 3). As shown in Fig. 1e, the H2O2 decomposition percentage/H2O2 decomposition rate/O2 selectivity are 31.2%/610.9 μmol h−1/93.0% over TiO2{001} NCs in the Ar atmosphere and decrease to 15.4%/301.5 μmol h−1/91.8% in the 10% O2/Ar atmosphere, while they are 20.4%/399.4 μmol h−1/89.0% over TiO2{001}-C3N4−0.1 in the Ar atmosphere and decrease to 8.26%/161.7 μmol·h−1/86.4% in the 10% O2/Ar atmosphere.

The suppress effect of O2 on photocatalytic H2O2 decomposition into O2 was observed to generally enhance not only H2O2 utilization efficiency but also H2O2 conversion, and consequently CH4 conversion in aqueous-phase photocatalytic conversion of methane with H2O2 using oxide photocatalysts due to the reaction coupling between photocatalytic H2O2 and CH4 reactions (Supplementary Table 4). Under the studied condition, the H2O2 utilization efficiency and CH4 conversion with an O2 addition are 1.30–1.78 and 1.4–2.0 times of those without O2 addition, respectively. We then optimized the O2 enhancement effect and photocatalytic performance over TiO2 NCs and TiO2 NCs-C3N4 composites (Supplementary Tables 510), both of which were observed to vary with structures of TiO2 NCs. TiO2{001} NCs are more photocatalytic active than TiO2{100} and TiO2{101} NCs, and the produced liquid-phase oxygenates are CH3OH and HCOOH over TiO2{001} NCs and CH3OH over TiO2{100} and TiO2{101} NCs. Over TiO2{001} NCs (Fig. 1f), the O2 addition increases the methane conversion rate from 39.5 to 69.7 μmol·gcatalyst−1·h−1, the selectivity of liquid-phase oxygenates and HCOOH respectively from 50.8% to 70.7% and from 30.4% to 53.9%, the HCOOH yield from 12.0 to 37.6 μmol·gcatalyst−1 h−1, and the H2O2 utilization efficiency from 21.4% to 32.1%. TiO2 NCs-C3N4−0.1 composites exhibit much better photocatalytic performance and more significant O2 promotion effect than corresponding TiO2 NCs. The produced liquid-phase oxygenates are CH3OH and HCOOH over TiO2{001}-C3N4−0.1, CH3OH and CH3OOH over TiO2{100}-C3N4−0.1, and CH3OOH over TiO2{101}-C3N4−0.1. Over TiO2{001}-C3N4−0.1 (Fig. 1g), the O2 addition increases the methane conversion rate from 358.5 to 696.3 μmol gcatalyst−1 h−1, the selectivity of liquid-phase oxygenates and HCOOH respectively from 93.7% to 97.0% and from 56.4% to 69.8%, the HCOOH yield from 202.2 to 486 μmol gcatalyst−1 h−1, and the H2O2 utilization efficiency from 53.4% to 93.3%.

The above results demonstrate an interesting photocatalytic system for efficiently converting CH4 to liquid-phase oxygenates in the presence of H2O2 and O2 at room temperature and ambient pressure over oxide-based photocatalysts, which presents high H2O2 utilization efficiencies due to the suppress effect of O2 on photocatalytic H2O2 decomposition into O2. The best photocatalyst, TiO2{001}-C3N4−0.1, exhibits an unprecedented H2O2 utilization efficiency of 93.3%, leading to a liquid-phase oxygenates selectivity of 97% and formic selectivity and yield respectively of 69.8% and 486 μmolHCOOH·gcatalyst−1 h−1. Its apparent quantum efficiency at 365 nm was measured to be 0.48%.

TiO2{001}-C3N4−0.1 is stable and its performance maintains well within six cycles of photocatalytic activity evaluations (Supplementary Fig. 4). Routine structural characterization results (Supplementary Fig. 5), including XPS, VB-XPS, UV-Vis spectra and photocurrent measurements, show few difference between the as-synthesized and used TiO2{001}-C3N4−0.1 catalysts.

Reaction mechanism

The carbon balance was calculated above 96.7% for all studied photocatalytic reactions. Blank photocatalytic experiment of photocatalytic reaction in the presence of TiO2{001}-C3N4−0.1 but absence of CH4 in the reactant did not produce detectable C-contained products; meanwhile, using 13CH4, all C-contained products only contained 13C (Supplementary Fig. 6). Thus, all C-contained products exclusively form from CH4. Initial evolutions of reaction products as a function of reaction time were examined over TiO2{001}-C3N4−0.1 (Supplementary Table 11). At a reaction time of 10 min, CH3OOH, CH3OH and HCHO were detected, and CH3OOH was the major product. The CH3OOH, CH3OH and HCHO productions increased at a reaction time of 30 min, meanwhile, HCOOH and CH3CH2OH appeared. As a reaction time of 1 h, the CH3OOH production decreased and HCHO was not detected, whereas the CH3OH and HCOOH productions increased greatly and the CH3CH2OH production increased slightly, meanwhile, CH3COOH emerged. These observations suggest CH3OOH as the primary product and CH3OH, HCHO, HCOOH, CH3CH2OH and CH3COOH as the secondary products that are produced sequentially. Moreover, the reaction rate of HCHO seems to be faster than the formation rate.

18O2 and H218O were used to trace origins of oxygen atoms in the liquid-phase oxygenate products. 18O2 were observed to exert similar enhancement effects on the H2O2 utilization efficiency to 16O2 and to slightly affect the product selectivity (Supplementary Table 12). It is noteworthy that CH3OOH decomposes completely into CH3OH during the mass spectroscopy analysis (13). Over TiO2{001} NCs (Supplementary Figs. 7, 8), no 18O-labelled product was detected when H218O was used, while CH318OH and HC18O16OH were detected with CH318OH/CH316OH and HC18O16OH/HC16O16OH ratios respectively of around 0.12 and 0.11 when 18O2 was used. Over TiO2{001}-C3N4−0.1 (Fig. 2a–d and Supplementary Fig. 9), only CH3C18O16OH for CH3COOH and no other 18O-labelled oxygenate were detected when H218O was used, while CH318OH, HC18O16OH and CH3CH218OH were detected with CH318OH/CH316OH, HC18O16OH/HC16O16OH and CH3CH218OH/CH3CH216OH ratios respectively of around 0.14, 0.13 and 0.25 when 18O2 was used, and CH3C16O16OH and CH3C16O18OH were detected for CH3COOH. Therefore, the oxygen atoms in CH3OOH, CH3OH, HCOOH and CH3CH2OH are contributed majorly by H2O2 and minor by O2, but seldom by H2O. Interestingly, HCOOH is formed via CH3OH oxidation exclusively by H2O2 whereas CH3COOH is formed via CH3CH2OH oxidation exclusively by H2O, suggesting that they follow different mechanisms. This was further supported by the observations that HC16O16OH/HC18O16OH and CH3C18O16OH/CH3C18O18OH were detected when 18O2 and H218O were used (Supplementary Fig. 10). Photocatalytic CH3CH2OH oxidation with H2O to CH3COOH was reported to be mediated by ·OH radicals generated by photogenerated holes-participated activation of H2O, typically occurring in the aqueous solution28,29. Thus, photocatalytic reactions to other liquid-phase products occur on the photocatalyst surfaces. Meanwhile, only a tiny amount of C16O18O was detected in the gas phase products while no C18O and C18O2 was detected when 18O2 was used for both TiO2{001} NCs and TiO2{001}-C3N4−0.1 (Supplementary Fig. 11).

Fig. 2: Reaction mechanism.
figure 2

Mass spectra of (a) methanol, (b) formic acid, (c) ethanol and (d) acetic acid formed during photocatalytic CH4 conversion over TiO2{001}-C3N4−0.1 under the reaction condition of 8%CH4 + 4%O2 + 88%Ar + 165 μL H2O2 + 20 mL H2O, 8%CH4 + 4%18O2 + 88%Ar + 165 μL H2O2 + 20 mL H2O, or 8%CH4 + 4%O2 + 88%Ar + 165 μL H2O2 + 20 mL H218O. Photocatalyst amount: 20 mg; reaction temperature: 25 °C; reaction time: 8 h. e In situ ESR spectra of H2O, H2O + O2, H2O + H2O2 and H2O + O2 + H2O2 solutions under UV light illumination at 298 K in the presence of DMPO over TiO2{001} NCs and TiO2{001}-C3N4−0.1. f In situ ESR spectra of CH4 + H2O mixture under UV light illumination at 298 K in the presence of DMPO over TiO2{101} (olive), TiO2{100} (dark yellow) and TiO2{001} (magenta) NCs, TiO2{101}-C3N4−0.1 (blue), TiO2{100}-C3N4−0.1 (red) and TiO2{001}-C3N4−0.1 (black) composites. Schematic diagrams of proposed dominant photocatalytic aqueous-phase CH4 reaction paths to liquid-phase oxygenates in the presence of H2O2 and O2 over (g) TiO2 NCs and (h) TiO2 NCs-C3N4. The 0.06 eV and −0.38 eV refer to the redox potential of H2O2 activated to ·OH radicals and ·OOH radicals at pH = 7. Source data are provided as a Source Data file.

In order to further clarify the role of O2, the O2 concentration in the reactant was increased from 4% (8%CH4 + 4%O2 + 88%Ar + 165 μL H2O2 + 20 mL H2O) to 12% (8%CH4 + 12%O2 + 80%Ar+165 μL H2O2 + 20 mL H2O), and the photocatalytic reaction was studied over TiO2{001}-C3N4−0.1 comparatively with 16O2 or 18O2. Using 16O2 or 18O2 gave similar H2O2 utilization efficiencies of around 94% and slightly different CH4 conversion rates and product selectivity (Supplementary Table 13). Using 18O2, the CH318OH/CH316OH, HC18O16OH/HC16O16OH and CH3CH218OH/CH3CH216OH ratios in the liquid-phase products were measured respectively as around 0.19, 0.17 and 0.22 (Supplementary Figs. 1214), similar to the case of the reactant with 4% O2; however, C18O and C18O2 were detected and the fraction of C16O18O in CO2 is much larger than that of C16O16O, different from the case of the reactant with 4% O2. Therefore, during photocatalytic aqueous-phase CH4 conversion in the presence of H2O2 and O2, CH4 preferentially reacts with H2O2 to produce liquid-phase oxygenates, while O2 acts mainly as a promoter to enhance H2O2 utilization efficiency and consequently CH4 conversion, and minorly as a reactant.

Using 5, 5-dimethyl-1-pyrroline N-oxide (DMPO) as the radical trapping agent, in situ EPR was used to probe radicals generated by photo-induced activation of various reactants. As shown in Fig. 2e and Supplementary Fig. 15, under UV light illumination, H2O is activated to ·OH radicals30 by photogenerated holes (h+) over various TiO2 NCs and TiO2 NCs-C3N4−0.1 composites, which barely changes upon the addition of O2. ·O2 radicals formed by O2 activation with photogenerated electrons (e) can not be observed in ESR spectra due to the instability in the aqueous solution, but their formation is evidenced by in situ ESR spectra in the methanol solution31 (Supplementary Fig. 16). Over TiO2 NCs, the ·OH radical signal grows slightly upon the addition of H2O2 and greatly upon the co-addition of H2O2 and O2. Over TiO2 NCs-C3N4−0.1 composites, the ·OH radical signal does not vary upon the addition of H2O2, while the ·OOH radical signal25,26 appears, and its intensity increases greatly upon the co-addition of H2O2 and O2. Thus, under UV light illumination, in addition to the h+-mediated decomposition into O2, H2O2 undergoes the e-mediated activation into ·OH radicals over TiO2 NCs and ·OOH radicals over TiO2 NCs-C3N4−0.1 composites. The e-mediated formation of dominant ·OOH radicals but few ·OH radicals over TiO2 NCs-C3N4−0.1 composites indicates that e for H2O2 activation is located on the conduction band mainly of C3N4 but seldom of TiO2, pointing to efficient interfacial transfer of e from the conduction band of TiO2 to the conduction band of C3N4. When CH4 was introduced to the aqueous solutions containing TiO2 NCs or TiO2 NCs-C3N4−0.1 composites under UV light illumination (Fig. 2f), ·CH3 radicals22,26, in addition to ·OH radicals, were detected. They greatly grew when isopropanol was added to quench ·OH radicals (Supplementary Fig. 17), but could not be detected in the presence of H2O2 and O2 when h+ was quenched using methanol (Supplementary Fig. 18). Thus, photocatalytic CH4 activation to ·CH3 radicals is mainly mediated by h+, instead of by·OOH, ·OH and ·O2 radicals.

Based on the above isotope-labelled and ESR results, photocatalytic CH4 conversion with H2O2 is initiated by the reaction of h+-generated ·CH3 with e-generated ·OH to CH3OH over TiO2 NCs (Fig. 2g) and with e-generated ·OOH to CH3OOH over TiO2 NCs-C3N4−0.1 composites (Fig. 2h). The addition of O2 opens up minor reaction pathways, including the reaction of h+-generated ·CH3 with O2 to CH3OO· radicals that facilely transform to CH3OOH22,32 and the reaction of CH4 with Ti–O· formed by e-mediated ·O2 reactions on TiO2 surfaces directly to CH3OH33. We consider that the ·CH3 + O2 reaction occurs mainly for TiO2 NCs-C3N4−0.1 composites due to the lack of enough e on the TiO2 components while the CH4 + Ti–O· reaction occurs mainly for TiO2 NCs due to the absence of CH3OOH in the products. Moreover, the addition of O2 greatly enhances the intensities of ·OOH radicals over TiO2 NCs-C3N4−0.1 composites and ·OH radicals over TiO2 NCs formed by the e-mediated H2O2 activation, and consequently the photocatalytic CH4 conversions. Since the presence of O2 efficiently suppresses the h+-mediated H2O2 decomposition to O2 under UV light illumination, the enhancement effect of O2 on ·OH and ·OOH generations from photocatalytic H2O2 activation is probably due to O2-suppressed h+-mediated H2O2 decomposition to O2 rather than O2-promoted e-mediated H2O2 decomposition to ·OH and ·OOH radicals. O2 does not compete with H2O2 for h+ that is localized on the TiO2 surface, thus O2 likely suppresses H2O2 adsorption on TiO2, instead of reaction of adsorbed H2O2 with h+, to suppress the h+-mediated H2O2 decomposition to O2. Both TiO2 NCs and TiO2 NCs-C3N4−0.1 composites exhibit TiO2 facet-dependent intensities of various radicals. The ·OH radicals are strongest over TiO2{001} NCs among all TiO2 NCs and the ·OOH radicals are strongest over TiO2{001}-C3N4−0.1 composite among all TiO2 NCs-C3N4 composites (Supplementary Fig. 19). The ·CH3 radicals are strongest over TiO2{100} NCs among various TiO2 NCs and over TiO2{001}-C3N4−0.1 composite among various TiO2 NCs-C3N4 composites (Supplementary Fig. 20). Meanwhile, TiO2 NCs-C3N4 composites exhibit more reactive radicals than corresponding TiO2 NCs. These results are consistent with the results of photocatalytic activity.

The band structures of various TiO2 NCs and TiO2 NCs-C3N4−0.1 photocatalysts were determined using UV–vis spectra and valence band XPS spectra (Supplementary Fig. 21). TiO2 NCs-C3N4−0.1 exhibits smaller band gaps than corresponding the TiO2 NCs, suggesting stronger capacity for light absorption and charge generation. The conduction band edges of TiO2 NCs and TiO2 NCs-C3N4 composites were measured to be −0.14−0.34 and −0.41−0.47 vs RHE, respectively, consistent with the experimental observations that H2O2 undergoes the e-mediated activation into ·OH radicals over TiO2 NCs and ·OOH radicals over TiO2 NCs-C3N4−0.1 composites (Fig. 2g, h). ESR spectra (Supplementary Fig. 22a) show that TiO2 NCs-C3N4−0.1 exhibit much lower densities of F+ and Ti3+ defects than TiO2 NCs and the defect density follows an order of TiO2{101} > TiO2{100} > TiO2{001} > TiO2{101}-C3N4−0.1 > TiO2{100}-C3N4−0.1 > TiO2{001}-C3N4−0.1. Accordingly, PL spectra (Supplementary Fig. 22b) show that the PL peak arising from the recombination of photoexcited electrons and holes displays an intensity order of TiO2{101} > TiO2{100} > TiO2{001} > TiO2{101}-C3N4−0.1 > TiO2{100}-C3N4−0.1 > TiO2{001}-C3N4−0.1. EIS spectra of various TiO2 NCs and TiO2 NCs-C3N4−0.1 photocatalysts were also measured, in which a smaller radius represents a low charge transfer resistance. All photocatalysts exhibit semicircle EIS spectra (Supplementary Fig. 22c), and the semicircle radius and consequently the charge transfer resistance follow an order of TiO2{101} > TiO2{100} > TiO2{001} > TiO2{101}-C3N4−0.1 > TiO2{100}-C3N4−0.1 > TiO2{001}-C3N4−0.1. ESR, PL and EIS are all bulk-sensitive characterization techniques, and their characterization results show that TiO2 NCs-C3N4−0.1 exhibit higher charge separation and transfer efficiencies than corresponding TiO2 NCs and that TiO2{001} is the best of various TiO2 NCs while TiO2{001}-C3N4−0.1 is the best of TiO2 NCs-C3N4−0.1 composite photocatalysts, consistent with the photocatalytic activity results.

NEXAFS acquired in a mode of total electron yield is a surface sensitive technique to probe the density of states of the orbitals involved in the electron transitions. UV light illumination excites electrons from the valence band to the conduction band, which consequently changes the density of states of the involved orbitals. We thus measured Ti L-edge, O K-edge, N K-edge and C K-edge NEXAFS spectra under dark and UV light illumination conditions of various samples (Fig. 3a–d, Supplementary Figs. 23, 24). The valence and conduction bands of TiO2 consist of the O 2p and Ti 3d orbitals, respectively, and the Ti L-edge and O K-edge NEXAFS features arise from the Ti 2p→3d and O 1 s→2p electron transitions, respectively. The valence and conduction bands of C3N4 consist of the N 2p and C 2p orbitals, respectively, and the N K-edge and C K-edge NEXAFS features arise from the N 1 s→2p and C 1 s→2p electron transitions, respectively. TiO2 NCs-C3N4 composites exhibit enhanced Ti L-edge and O K-edge features than corresponding TiO2 NCs but weakened C K-edge and N K-edge NEXAFS features than C3N4. This indicates an occurrence of TiO2→C3N4 electron transfer within TiO2 NCs-C3N4 composites, which decreases the electron density on TiO2 but increases the electron density of C3N4. Using the Ti L-edge and C K-edge NEXAFS features as examples (Fig. 3e), TiO2{001}-C3N4 composite exhibits the largest intensity variations of both Ti-L edge and C-K edge absorption features among all TiO2 NCs-C3N4 composites, demonstrating the most extensive electron transfer from TiO2{001} NCs to C3N4.

Fig. 3: Interfacial charge transfer.
figure 3

a Ti L-edge, (b) O K-edge, (c) N K-edge and (d) C K-edge NEXAFS spectra of TiO2{001} NCs (black line), TiO2{001} NCs-C3N4−0.1 composite (red line) and C3N4 (blue line) in dark (thick line) and under UV light illumination (thin line). e Ti L-edge intensity ratios of TiO2 NCs-C3N4−0.1 composites against corresponding TiO2 NCs and C K-edge intensity ratios of TiO2 NCs-C3N4−0.1 composites against C3N4. f Intensity ratios of Ti L-edge, O K-edge, C K-edge and N K-edge features of various photocatalysts under UV light illumination against in dark. Source data are provided as a Source Data file.

UV light illumination excites electrons from the valence bands of TiO2 or C3N4 to the conduction bands, and consequently results in weakened Ti L-edge NEXAFS features of TiO2 NCs, enhanced O K-edge NEXAFS features of TiO2 NCs and TiO2 NCs-C3N4−0.1 composites, and weakened C K-edge and enhanced N K-edge NEXAFS features of C3N4 and TiO2 NCs-C3N4−0.1 composites. But TiO2 NCs-C3N4−0.1 composites exhibit stronger Ti L-edge NEXAFS features under UV light illumination than under dark condition. This supports the formations of Z-scheme TiO2-C3N4 heterojunctions within TiO2 NCs-C3N4−0.1 composites34, in which the photogenerated electrons on the conduction band of TiO2 (Ti 3d orbital) efficiently transfer to the valence band of C3N4 (N 2p orbital) and recombine with photogenerated holes therein (Fig. 2h). Moreover, the total transferred electrons from the conduction band of TiO2 are more than the photogenerated electrons, likely due to a large number of photogenerated holes in the valence band of C3N4, which results in a less-occupied Ti 3d orbital and consequently a stronger Ti L-edge NEXAFS features of TiO2-C3N4−0.1 composites under UV light illumination than under dark condition. Figure 3f presents the ratios (IUV/Idark) of Ti L-edge, O K-edge, N K-edge and C K-edge NEXAFS features of various photocatalysts under UV light illumination against in dark, whose deviations from the unity reflect the photogenerated charges on the photocatalyst surfaces. Much larger concentrations of photogenerated electrons and holes are present on TiO2 surfaces than on C3N4 surface, suggesting more efficient charge separation and migration to surface within TiO2 NCs. C3N4 surface exhibits similar concentrations of photogenerated electrons and holes while TiO2 surfaces exhibit larger concentrations of photogenerated holes than of photogenerated electrons. TiO2 NCs-C3N4−0.1 composite surfaces exhibit slightly smaller concentrations of photogenerated holes than corresponding TiO2 NCs surfaces but larger concentrations of photogenerated charges than C3N4 surfaces. Thus, the Z-scheme TiO2-C3N4 heterojunctions within TiO2 NCs-C3N4−0.1 composites contribute to the charge separation and migration to surface over C3N4 component more than over TiO2 component. Among various TiO2 NCs or TiO2 NCs-C3N4−0.1 composites, the photocatalysts consisting TiO2{001} NCs exhibit the largest concentrations of photogenerated charges on the surfaces, leading to the largest concentrations of ·OH radicals over TiO2{001} NCs, ·OOH and ·CH3 radicals over TiO2{001}-C3N4 composite. However, TiO2{100} NCs, instead of TiO2{001} NCs, exhibit the largest concentration of ·CH3 radicals, which is likely relevant to the adsorption behaviors of CH4 on various photocatalysts. The adsorption heats of CH4 were measured similar for various TiO2 NCs (16.8–17.7 kJ/mol) or TiO2 NCs-C3N4−0.1 composites (11.1–14.5 kJ/mol) (Supplementary Figs. 2527), while the adsorption amounts followed orders of TiO2{100} > TiO2{101} > TiO2{001} and of TiO2{001}-C3N4 > TiO2{100}-C3N4 > TiO2{101}-C3N4.

Various TiO2 NCs and TiO2 NCs-C3N4−0.1 composites show not only TiO2 facet-dependent activity but also TiO2 facet-dependent selectivity in photocatalytic CH4 conversion with H2O2 or H2O2 + O2. The photocatalysts with low photocatalytic activity exhibit low selectivity toward the liquid-phase products because more oxidizing radicals are available to eventually convert the liquid-phase intermediates to CO2. TiO2{001} NCs and TiO2{001}-C3N4−0.1 composite exhibit the highest photocatalytic activity and consequently the highest photocatalytic selectivity toward the liquid-phase products among various TiO2 NCs and various TiO2 NCs-C3N4−0.1 composites, respectively. Moreover, distributions of liquid-phase products vary with the TiO2 facets. Especially, HCOOH is the major liquid-phase product for the photocatalysts containing TiO2{001} NCs, but is barely observed for the photocatalysts containing TiO2{101} or TiO2{100} NCs. In situ DRIFTS spectra were used to explore surface reaction mechanisms of photocatalytic CH4 conversion with H2O2 + O2 over TiO2 NCs-C3N4−0.1 composites (Fig. 4). The observed vibrational bands (Supplementary Table 14) were assigned based on in situ DRIFTS spectra of CH3OH and HCOOH adsorption on various TiO2 NCs (Supplementary Fig. 28) and previous reports35,36,37,38,39. As the photocatalytic reaction prolongs over TiO2{001}-C3N4−0.1 composite (Fig. 4a), the vibrational features of adsorbed CH3 (1473 cm−1), CH2 (1445 cm−1), CH3OH (1019 and 1092 cm−1), CH3O (1042 and 1156 cm−1), CH2O (1712 cm−1), HCOO (1526, 1556 and 1564 cm−1), HCOOH (1664 cm−1) and carbonates (1504 and 1592 cm−1) species and gaseous HCOOH (1760 and 1782 cm−1) emerge and grow at the expense of gaseous CH4 (1304 cm−1). These results directly evidence the occurrences of photocatalytic oxidation of CH4 to CH3OH via the CH3 intermediate and further to HCOOH via the CH3O, CH2O and HCO (in the form of HCOOTiO2) intermediates, as schematically shown in Fig. 2g, h. Although the carbonate intermediates were observed, no signals of CO or CO2 appeared, indicating that the carbonate intermediates are very stable on TiO2{001}-C3N4 composite. Comparing TiO2{001}-C3N4−0.1 composite, TiO2{100}-C3N4−0.1 and TiO2{101}-C3N4−0.1 composites exhibit very different in situ DRIFTS spectra (Fig. 4b). The gaseous CH4 consumptions and the CH3OH(a) formation are greatly smaller over TiO2{100}-C3N4−0.1 and TiO2{101}-C3N4−0.1 composites than over TiO2{001}-C3N4−0.1 composite. Meanwhile, only very minor vibrational features of surface intermediates appear whereas obvious vibrational features of gaseous CO (2135 and 2170 cm−1) and CO2 (2340 and 2360 cm−1) emerge over TiO2{100}-C3N4−0.1 and TiO2{101}-C3N4−0.1 composites, respectively. These in situ DRIFTS results are consistent with the photocatalytic reaction data that TiO2{001}-C3N4−0.1 composite are much more photocatalytic active and selective toward the liquid-phase products in photocatalytic CH4 conversion with H2O2 + O2 than TiO2{100}-C3N4−0.1 and TiO2{101}-C3N4−0.1 composites.

Fig. 4: In situ characterization.
figure 4

a In situ DRIFTS spectra of photocatalytic CH4 conversion at 298 K under different light irradiation times over TiO2{001}-C3N4−0.1 with DRIFTS spectra prior to UV light illumination as the background spectra. b In situ DRIFTS spectra of photocatalytic CH4 conversion at 298 K under light irradiation for 80 min over TiO2{001}-C3N4−0.1, TiO2{100}-C3N4−0.1 and TiO2{101}-C3N4−0.1 with DRIFTS spectra prior to UV light illumination as the background spectra. Source data are provided as a Source Data file.

Theoretical calculations

DFT calculations were carried out to understand O2-suppressed photocatalytic H2O2 decomposition to O2 and facet-dependent photocatalytic selectivity of CH4. Since both photocatalytic H2O2 decomposition to O2 and photocatalytic CH4 conversion are mediated by photogenerated holes located predominantly on TiO2 NCs, thus we considered TiO2 facets, but not TiO2-C3N4 interfaces, during the DFT calculations. As reported previously35,40,41,42,43,44, the anatase TiO2(001) surface exposed on TiO2{001} NCs exhibits a typical reconstructed (001)-(1 × 4) surface with fourfold-coordinated Ti cations (Ti4c) at the (1 × 4) added row, fivefold-coordinated Ti cations (Ti5c) at the (1 × 1) basal surface and twofold-coordinated O anions (O2c), the anatase TiO2(100) surface exposed on TiO2{100} NCs exhibits a typical reconstructed (1 × 2) surface with the Ti5c, O2c and threefold-coordinated O (O3c) sites, and the anatase TiO2(101) surface exposed on TiO2{101} NCs exhibits a (1 × 1) unreconstructed surface with the Ti5c, O2c and O3c sites (Supplementary Fig. 29). The Ti4c sites on TiO2(001) surface show much stronger adsorption ability than the Ti5c sites on TiO2 (001), (100) and (101) surfaces. As shown in Fig. 5a and Supplementary Fig. 30, the adsorption energy of H2O2 is −1.46, −0.80 and −0.77 eV on TiO2 (001), (100) and (101) surfaces, respectively, and greatly decreases to −0.53, −0.18 and −0.10 eV on O2-covered TiO2 (001), (100) and (101) surfaces, respectively. The adsorption energy of O2 is −0.49, −0.18 and −0.14 eV on TiO2 (001), (100) and (101) surfaces, respectively (Fig. 5b and Supplementary Fig. 31). These DFT calculation results demonstrate that O2 is capable of weakening H2O2 adsorption on TiO2 to suppress the h+-mediated H2O2 decomposition to O2. The strongest adsorption of O2 on TiO2(001) surface exerts the strongest suppress effect on H2O2 decomposition to O2 on TiO2{001} NCs. CH4 adsorption on TiO2 (001), (100) and (101) surfaces are very weak with an adsorption energy of −0.17, −0.03 and −0.04 eV (Supplementary Fig. 32). Adsorption energy of CH3OOH on TiO2 (001), (100) and (101) surfaces is −0.69, −0.22 and −0.04 eV, respectively (Fig. 5c and Supplementary Fig. 33). CH3OH adsorbs both molecularly and dissociatively with adsorption energy respectively of −0.84 and −1.69 eV on TiO2 (001) surface, −0.57 and −0.16 eV on TiO2 (100) surface, −0.49 and −0.65 eV on TiO2(101) surface (Fig. 5d and Supplementary Fig. 34). The calculated adsorption energies of various liquid-phase products on TiO2 (001), (100) and (101) surfaces are consistent with the experimentally observed different selectivity toward liquid-phase products in photocatalytic CH4 conversion over TiO2 {001}, {100} and {101} NCs, suggesting that desorption of various products from TiO2 surface play a key role in determining the selectivity. Preferential dissociation of produced CH3OH on TiO2{001} NCs and TiO2{001}-C3N4−0.1 composite forms methoxy species which is further photooxidized to HCOOH (Fig. 2g, h), leading to the experimental results that HCOOH is the major liquid-phase product. The very weak adsorption of produced CH3OOH on TiO2(101) surface makes it as the sole liquid-phase product over TiO2{101}-C3N4−0.1 composite.

Fig. 5: DFT calculations.
figure 5

Calculated adsorption energies of (a) H2O2 on clean and O2-covered TiO2 (001), (100) and (101) surfaces, (b) O2, (c) CH3COOH and (d) molecular and dissociative CH3OH adsorption on TiO2 (001), (100) and (101) surfaces. Source data are provided as a Source Data file.

Discussion

Therefore, O2 is a general and efficient molecular additive to suppress H2O2 adsorption on oxide photocatalysts and consequently photogenerated holes-mediated H2O2 decomposition to O2 during photocatalytic reactions. Such a suppress effect, together with efficient charge separation within TiO2{001}-C3N4 heterojunctions, photogenerated holes-mediated activation of CH4 into ·CH3 radicals on TiO2{001} and photogenerated electrons-mediated activation of H2O2 into ·OOH radicals on C3N4, and preferential dissociative adsorption of methanol on TiO2{001}, leads to an unprecedented high H2O2 utilization efficiency of 93.3% and highly active and selective to liquid-phase oxygenates with formic acid as the major product during photocatalytic conversion of methane with H2O2 and O2. H2O2 production is known as an environment-unfriendly and economic-costly process45, therefore, our findings point to co-use of H2O2 and O2 in photocatalytic oxidation reactions over oxide-based photocatalysts as a promising strategy to achieve high H2O2 utilization efficiency and excellent photocatalytic performance.

Methods

Materials

H2O2 aqueous solution (20 wt.%), HF aqueous solution (40 wt.%), acetate, acetic acid, methanol, isopropanol, Ti(OBu)4, K2TiO(C2O4)2, P25, ZnO, Fe2O3, WO3 and V2O5 were all with the analytical grade and purchased from Sinopharm Chemical Reagent Co. CuO (≥99%), dicyandiamide (≥98%), pentane-2,4-dione (≥98%), 5, 5-dimethyl-1-pyrroline N-oxide (DMPO) (≥97%) and 3-(trimethylsilyl)−1-propanesulfonic acid sodium salt (DSS) (≥97%) were purchased from Sinopharm Chemical Reagent Co. Reactants of CH4 (8%) + O2(4%) + Ar (88%) and CH4 (8%) + O2 (4%) + O2 (10%) + Ar (78%) were purchased from Nanjing Shang Yuan Industry Factory. 13CH4 (13C enrichment > 99%atom), 18O2 (18O enrichment ≥ 98%atom) and H218O (18O enrichment ≥ 98%atom) were purchased from Wuhan Newradar Gas Co. All chemicals and gases were used as received.

Catalyst synthesis

TiO2 NCs predominantly exposing different types of facets were prepared following previous procedures27.

Synthesis of anatase TiO2{001} NCs: typically, 25 mL Ti(OBu)4 and 3 mL HF aqueous solution (40 wt%) were mixed under stirring at RT (Caution: Hydrofluoric acid (HF) is extremely corrosive and a contact poison, and it should be handled with extreme care! Hydrofluoric acid solution is stored in Teflon containers in use.). The solution was then transferred into a 50 mL Teflon lined stainless steel autoclave and kept at 180 °C for 24 h. The resulted white precipitate was collected by centrifugation, washed repeatedly with ethanol and water, and dried at 70 °C for 12 h. The acquired powder was dispersed in 700 mL NaOH aqueous solution (0.1 mol/L), stirred for 24 h at RT, centrifuged, and washed repeatedly with water until the pH value of aqueous solution was of 7–8.

Synthesis of anatase TiO2{100} and TiO2{101} NCs: typically, 6.6 mL TiCl4 was added dropwise into 20 mL HCl aqueous solution (0.43 mol/L) at 0 °C. After stirring for an additional 0.5 h, the solution was added dropwise into 50 mL NH3 aqueous solution (5.5 wt%) under stirring at RT. Then the pH value of the solution was adjusted to between 6 and 7 using 4 wt% NH3 aqueous solution, after which the system was stirred at RT for 2 h. The resulted precipitate was filtered, washed repeatedly with water until no residual Cl could be detected, and then dried at 70 °C for 12 h to acquire Ti(OH)4 precursor. To prepare anatase TiO2-{100} nanocrystals, 2.0 g Ti(OH)4 and 0.5 g (NH4)2SO4 were dispersed in a mixture of 15 mL H2O and 15 mL isopropanol under stirring at RT, then the mixture was transferred into a 50 mL Teflon-lined stainless steel autoclave and kept at 180 °C for 24 h. The obtained white precipitate was collected and washed repeatedly with water. To prepare anatase TiO2-{101} nanocrystals, 2.0 g Ti(OH)4 and 0.2 g NH4Cl were dispersed in a mixture of 15 mL H2O and 15 mL isopropanol under stirring at RT, then the mixture was transferred into a 50 mL Teflon-lined stainless steel autoclave and kept at 180 °C for 24 h. The obtained white precipitate was collected and washed repeatedly with water.

Synthesis of anatase TiO2 NCs-C3N4 composites: calculated amounts of dicyandiamide (C2H4N4) and TiO2 NCs were mixed in a crucible. The crucible was placed into a tube furnace, purged in Ar 1 h, and heated to 550 °C at a rate of 2.5 °C/min and kept for 4 h, then cooled to room temperature. The acquired powders were taken out and grind to obtain TiO2 NCs-C3N4 composites.

Structure characterizations

Powder X-ray diffraction (XRD) patterns were recorded on a Philips X’Pert Pro Super diffractometer with Cu Kα radiation (λ = 0.15406 nm) operated at 40 kV and 50 mA. Transmission infrared spectra were recorded on a Nicolet 8700 spectrometer at room temperature. Electron paramagnetic resonance (ESR) spectra with and without Xenon lamp irradiation were recorded on a JEOL JES-FA200 ESR spectrometer (9.063 GHz, X-band) at 130 K with employed microwave power, modulation frequency, and modulation amplitude of 0.998 mW, 100 kHz, and 0.35 mT, respectively. Steady-state photoluminescence spectra were measured on a HORIBA LabRAM HR spectrograph with a continuous wave 325 nm laser as the exciting source and the signal was collected by passing through a filter with cut-off wavelengths below 380 nm. UV–vis diffuse reflectance spectra (UV–vis DRS) were obtained on a Shimadzu DUV-3700 spectrophotometer equipped with an integrating sphere attachment. X-ray photoelectron spectroscopy (XPS) measurements were performed on an ESCALAB 250 high-performance electron spectrometer using monochromatized Al Kα (hv = 1486.7 eV) as the excitation source, and the likely charging of samples was corrected by setting the C 1 s binding energy of the adventitious carbon to 284.8 eV. Near-edge X-ray absorption fine structure (NEXAFS) spectra were measured at photoelectron spectroscopy end-station of National Synchrotron Radiation Laboratory. Transmission electron microscopy (TEM), high-resolution transmission electron microscopy (HRTEM) and element mapping images were performed with a JEOL JEM-2100F instrument at an acceleration voltage of 120 kV.

Adsorption microcalorimetry measurements were carried out on a home-setup equipment consisting of a Setaram Sensys EVO 600 Tian-Calvet microcalorimeter and an Micromeritics Autochem II 2920 automated chemisorption apparatus46. Typically, 50 mg sample was placed in the sample quartz tube and degassed in He (flow rate: 50 mL/min) at 200 °C for 60 min, then the sample was cooled to −100 °C, and the gas stream was switched to 2% CH4/He (flow rate: 50 mL/min) for adsorption. After CH4 adsorption reached saturation, the gas stream was switched back to He (flow rate: 50 mL/min) for desorption. The adsorption/desorption amounts and accompanying heat flows were quantified by the chemisorption apparatus and microcalorimeter, respectively, from which the adsorption/desorption heats were calculated.

In situ DRIFTS experiments were performed at 298 K on a Thermo Scientific Nicolet iS50 FTIR Spectrometer with a mercury cadmium telluride detector cooled with liquid nitrogen. The spectrometer was equipped with a Harrik Praying Mantis diffuse reflection accessory and a Harrick high-temperature reaction cell with ZnSe windows. The reaction cell was connected to a SH-110 dry scroll vacuum pump (Agilent Technologies), H2O2 aqueous solution stored in a quartz tube welded with Kovar, and 8%CH4 + 4%O2 + 88%Ar gas via three closed valves. The 30% H2O2 aqueous solution was purified by repeated cycles of freeze−pump−thaw treatments. The UV light irradiation on the sample was accomplished through the front window of the high-temperature reaction chamber using a 100 W high-pressure Hg arc lamp (Oriel 6281), which provides a pressure-broadened emission spectrum from gaseous Hg in the UV-light region. A water filter was used to remove the IR portion of the emission spectrum. Typically, the sample was loaded in the sample holder of the reaction cell, then the reaction cell was evacuated by opening the valve connecting the vacuum pump. After the pressure decreased to 10 Torr, the valve connecting the vacuum pump was closed, and the valve connecting the H2O2 aqueous solution was opened to reach a stable pressure, and then the valve connecting to 8%CH4 + 4%O2 + 88%Ar gas was open to allow the pressure of the reaction cell to 1 atm, and finally both valves were closed. The DRIFTS spectrum of the sample prior to UV light illumination was firstly taken as the background spectrum, then the UV light was turned on to irradiate the sample and the DRIFTS spectra were taken in a sequential mode. The DRIFTS spectrum of the sample was also taken after the turn off of the UV light. All DRIFTS spectra were measured with 128 scans at a resolution of 4 cm−1.

Photocatalytic activity measurements

Photocatalytic activity of various samples in aqueous-phase methane conversion was evaluated in a quartz reactor with a cooling-water jacket to maintain the reaction temperature at 25 °C under atmospheric pressure using a 300 W Xe lamp as the light source whose spectrum is shown in the Supplementary Fig. 35. Typically, 20 mg photocatalyst, 20 mL deionized water and a certain amount of H2O2 aqueous (1 mol/L) solution were mixed in the reactor. The reaction system was adequately deaerated by reaction gas for 1 h, and then was irradiated by the Xe lamp. then the photocatalytic reaction was carried out. After a desirable reaction time, 0.5 mL gas was sampled from the reaction system and its composition products was analyzed by a Fuli GC9720 gas chromatography equipped with FID and TCD detectors.

Liquid-phase oxygenate products were analyzed and quantified by 1H nuclear magnetic resonance (NMR) spectra acquired on a JEOL ECS 400 MHz NMR spectrometer. A DSS solution in D2O (0.020 wt.%) with the 1H chemical shift at δ = 0.0 ppm was prepared to calibrate the chemical shift. Typically, 0.70 mL clear aqueous solution was sampled from the reaction system and mixed with 0.10 mL DSS solution in a NMR tube and the 1H NMR spectrum was taken. The intensity of measured 1H NMR peak of various products were compared to the corresponding 1H NMR working curve acquired using pure product of different concentrations (Supplementary Fig. 36). Since pure CH3OOH could not be purchased while both CH3OOH and CH3OH have the methyl group, the amount of CH3OOH in the liquid-phase products was quantified using the working curve of CH3OH22.

The concentration of HCHO was quantified by the colorimetric method22. Typically, 100 mL of the reagent aqueous solution was prepared by dissolving 15 g ammonium acetate, 0.3 mL acetic acid, and 0.2 mL pentane-2,4-dione in water. Then, 0.5 mL liquid product was mixed with 2.0 mL water and 0.5 mL reagent solution. The mixed solution was maintained at 35 °C and measured by UV − vis absorption spectrum until the absorption intensity at 412 nm did not further increase. The concentration of HCHO in the liquid product was determined by the standard curve (Supplementary Fig. 37).

The concentration of H2O2 in the aqueous solution was quantified by the colorimetric method13,21. Typically, a reagent aqueous solution was prepared by dissolving 0.636 g K2TiO(C2O4)2 and 20 μL concentrated H2SO4 (98%) in 100 mL deionized water. 0.2 mL aqueous solution was exacted from the reaction system and mixed with 4.0 mL reagent solution. Then the UV − vis absorption spectrum of the mixed solution was measured, and the intensity of the absorption peak at 398 nm arising from the complex formed by K2TiO(C2O4)2 and H2O2 was compared to the working curve acquired using pure H2O2 aqueous solution of different concentrations (Supplementary Fig. 38) to quantify the H2O2 concentration.

Methane conversion, product selectivity, H2O2 conversion and H2O2 utilization efficiency were calculated as the following:

$$ {{{{{\rm{Methane}}}}}}\; {{{{{\rm{conversion}}}}}} \, (\%)=({{{{{\rm{n}}}}}}\left({{{{{\rm{CH}}}}}}_{4}\right)_{{{{{{\rm{before}}}}}}\; {{{{{\rm{reaction}}}}}}}\\ \quad - {{{{{\rm{n}}}}}}({{{{{\rm{CH}}}}}}_{4})_{{{{{{\rm{after}}}}}}\; {{{{{\rm{reaction}}}}}}})/{{{{{\rm{n}}}}}}({{{{{\rm{CH}}}}}}_{4})_{{{{{{\rm{before}}}}}}\; {{{{{\rm{reaction}}}}}}}\times 100\%$$
$${{{{{\rm{Product}}}}}}\; {{{{{\rm{selectivity}}}}}} \, (\%)={{{{{{\rm{n}}}}}}}_{{{{{{\rm{Product}}}}}}}/({{{{{{\rm{n}}}}}}({{{{{{\rm{CH}}}}}}}_{4})}_{{{{{{\rm{before}}}}}}\; {{{{{\rm{reaction}}}}}}}-{{{{{{\rm{n}}}}}}({{{{{{\rm{CH}}}}}}}_{4})}_{{{{{{\rm{after}}}}}}\; {{{{{\rm{reaction}}}}}}})\times 100\%$$
$$ {{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2}{{{{{\rm{conversion}}}}}} \, (\%)=\left({{{{{{\rm{n}}}}}}({{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2})}_{{{{{{\rm{before}}}}}}\; {{{{{\rm{reaction}}}}}}}\right .\\ \quad \left .- {{{{{{\rm{n}}}}}}({{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2})}_{{{{{{\rm{after}}}}}}\; {{{{{\rm{reaction}}}}}}}\right)/{{{{{{\rm{n}}}}}}({{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2})}_{{{{{{\rm{before}}}}}}\; {{{{{\rm{reaction}}}}}}}\times 100\%$$
$$ {{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2}{{{{{\rm{utilization}}}}}}\; {{{{{\rm{efficiency}}}}}} \, (\%)=\big({{{{{\rm{n}}}}}}({{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2})_{{{{{{\rm{before}}}}}}\; {{{{{\rm{reaction}}}}}}}\\ \quad - {{{{{\rm{n}}}}}}({{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2})_{{{{{{\rm{after}}}}}}\; {{{{{\rm{reaction}}}}}}}-{{{{{\rm{n}}}}}}({{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2})_{{{{{{\rm{decomposition}}}}}} \; {{{{{\rm{to}}}}}} \; {{{{{\rm{O}}}}}}2}\big )/{{{{{\rm{n}}}}}}({{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2})_{{{{{{\rm{before}}}}}} \; {{{{{\rm{reaction}}}}}}}\times 100\%$$
$${{{{{{\rm{n}}}}}}({{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2})}_{{{{{{\rm{decomposition}}}}}}\; {{{{{\rm{to}}}}}} \; {{{{{\rm{O}}}}}}2}={{{{{{\rm{n}}}}}}({{{{{{\rm{O}}}}}}}_{2})}_{{{{{{\rm{after}}}}}} \; {{{{{\rm{reaction}}}}}}} - {{{{{{\rm{n}}}}}}({{{{{{\rm{O}}}}}}}_{2})}_{{{{{{\rm{before}}}}}} \; {{{{{\rm{reaction}}}}}}} - {{{{{{\rm{n}}}}}}}_{{{{{{\rm{O}}}}}}2{{{{{\rm{reacted}}}}}}}$$
$$ {{{{{\rm{Carbon}}}}}}\; {{{{{\rm{balance}}}}}} \, (\%)={{{{{{\rm{n}}}}}}}_{{{{{{\rm{carbon}}}}}}\; {{{{{\rm{in}}}}}}\; {{{{{\rm{all}}}}}}\; {{{{{\rm{products}}}}}}}/\\ \quad ({{{{{{\rm{n}}}}}}({{{{{{\rm{CH}}}}}}}_{4})}_{{{{{{\rm{before}}}}}}\; {{{{{\rm{reaction}}}}}}}-{{{{{{\rm{n}}}}}}({{{{{{\rm{CH}}}}}}}_{4})}_{{{{{{\rm{after}}}}}}\; {{{{{\rm{reaction}}}}}}})\times 100\%$$

Where n was the quantified amount of reactants or products, while nO2 reacted was calculated from the amount of products and the ratio of the products formed by O2 based on the isotope-labelling results. For photocatalytic reactions using 18O2, n18O2 reacted was calculated by (n(18O2)before reaction-n(18O2)after reaction), in which n(18O2) was quantified using GC-MS. The carbon balance was calculated not less than 96.7% for all studied photocatalytic reactions.

Product analysis of photocatalytic reactions using isotope-labelled reactants

Liquid-phase oxygenates produced by aqueous-phase photocatalytic methane conversion using 13CH4 were analyzed by 1H NMR and 13C with decoupling NMR spectrometer as described above. Products of aqueous-phase photocatalytic methane conversion using 18O2 and or H218O were analyzed by mass spectrometer as the following: 0.5 mL gas was sampled from the reaction system and its composition was analyzed on a Trace GC/ISQ MS; and 3 mL clear aqueous solution was sampled from the reaction system and transferred into a quartz tube welded with Kovar and then connected to a QIC20 mass spectrometer (Heiden Analytical Ltd.) and a Hicube 80 Eco pump station by two closed valves. The aqueous solution was purified by repeated cycles of freeze−pump−thaw treatments and its composition was analyzed by the QIC20 mass spectrometer.

Theoretical calculations

All theoretical calculations were carried out using the Vienna ab initio simulation package (VASP)47,48, and the exchange-correlation term was described by the Perdew, Burke and Ernzerhof version within the generalized gradient approximation (PBE-GGA)49. The project-augmented wave (PAW)50,51 method was used to represent the core-valence electron interaction. The titanium 3 s, 3p, 3d, 4 s, and the carbon and oxygen 2 s, 2p electrons were treated as valence electrons and an energy cutoff of 400 eV for the basis-set expansion was used. The anatase TiO2(001)-(1 × 4), TiO2(101) and TiO2(100) surface was modeled as a periodic slab with six O-Ti-O trilayers of oxide. A vacuum between slabs >15 Å and corresponding 1 × 1 × 1 k-point mesh were used during the calculations. Adsorption was modeled on one side of the slab, and during structural optimizations, all of the atoms except those in the bottom TiO2 trilayer of the slab, were allowed to relax until atomic forces reached below 0.05 eV/Å. The adsorption energy (Eads) was expressed using the average adsorption energy calculated by Eads = Ead∕sub – (Esub + Ead) in which Ead∕sub is the total energy of the interacting system containing adsorbed molecules and TiO2 support in a surface cell, Esub is the total energy of the anatase TiO2 slab and Ead is the total energy of the molecule in gas phase.

Details on structural characterizations, activity evaluations, and DFT calculations can be found in the supplementary information.