Introduction

The rapid transition towards net-zero carbon (CO2) emissions is an imperative undertaking by science and society to stave off potentially catastrophic climate change1. Hence, urgency arises: (1) to develop low-carbon transition pathways to turn traditional fossil resources into high-value-added chemicals, and (2) to displace traditional carbon-intensive manufacturing processes. Formaldehyde is an important high-volume industrial chemical with a market value of over 8 billion USD, expanding at a compound annual growth rate (CAGR) of 5.7%. It is widely used for household, commercial, aviation, medical and automotive products, and is a valuable precursor for melamine, urea-formaldehyde and phenolic resins, etc., due to its high reactivity and versatility2. It is also safely in use for the manufacture of vaccines, anti-infective drugs and hard-gel capsules. Currently, formaldehyde is produced by methanol oxidation-dehydrogenation using silver or metal-oxide catalysts at a high reactor temperature of over 500–600 °C, incurring both high CO2 emission and energy penalties.

On the other hand, with the continuous discovery of abundant methane (CH4) resources, especially shale/natural gas, the direct CH4 conversion into value-added chemicals such as methanol, formaldehyde and formic acid offers considerable economic and environmental benefits3,4,5,6,7. However, due to the high C–H bond dissociation energy (439 kJ·mol−1), CH4 serves as the most stable and inert industrial feedstock among alkanes8,9,10,11,12, and its industrial utilisation through indirect steam-reforming and subsequently Fischer-Tropsch synthesis is usually energy-intensive due to the high operating temperature (700–1100 °C)13,14,15,16. Therefore, sustainable CH4 utilisation under mild conditions is highly desirable.

As a renewable technology, photocatalysis has shown unprecedented opportunities for overcoming thermodynamic barriers and achieving direct CH4 conversion to various chemicals at ambient temperature. Recently, several efforts on direct photocatalytic CH4 oxidation to methanol (CH3OH) and formaldehyde (HCHO) have been reported. High selectivity to CH3OH (>90%) and HCHO (100%) but with very moderate yields (350 μmol·g−1·h−1 of CH3OH and 300 μmol·g−1·h−1 of HCHO), have been achieved over optimised FeOx/TiO2 and Au/WO3 photocatalysts, respectively17,18. Noble-metal (Au, Ag, Pd, Pt) modified ZnO and TiO2 performed acceptable yields of various oxygenates but with relatively low selectivity and mainly under UV/near UV irradiation (<80%)17,18,19,20,21,22,23. These advances encourage further investigations especially in the search for low-cost noble-metal-free photocatalysts with wide-spectrum response and simultaneous optimisation of activity and selectivity. Extending photoabsorption and enhancing charge separation generally improve photoactivity on various solar-driven reactions such as H2O splitting24,25, CO2 reduction26,27, organic synthesis28,29 and contaminant elimination30,31. However, such an approach is insufficient for CH4 conversion due to its rather low electron and proton affinity. Moreover, another challenge for the CH4 conversion into desirable oxygenates (i.e., CH3OH and HCHO) is that the targeted oxygenates are usually more reactive than CH4, which tend to be over-oxidised, leading to poor selectivity23,32. Hence, rationally regulating CH4 reaction dynamics and promoting charge separation are equally significant to tune the overall photocatalytic performance.

In industry, HCHO is solely produced from CH3OH oxidation and accounted for 30% consumption of CH3OH33. Multi-step production from CH4 via steam reforming, Fischer-Tropsch synthesis and methanol conversion is energy-intensive and requires harsh conditions33,34. Nevertheless, one-step conversion from CH4 to HCHO with high selectivity and yield is still lacking and no commercial catalyst is available even for high temperature. Hence, it is highly desirable to develop a suitable catalyst and/or co-catalyst as a promising alternative to promote photoabsorption, enhance charge separation and optimise selectivity under mild conditions35. The key to realise such direct conversion and selectivity depends on regulating the reactive oxygen species (ROS), enhancing the conversion of methane and promoting timely desorption of the desired products, whilst suppressing its mineralisation to CO218. Suitable redox cocatalysts can be beneficial to charge transfer and promote the activation of the adsorbed O2 and H2O to form reactive oxygen species such as superoxide (·O2-), hydroperoxyl (·OOH) and hydroxyl (·OH) radicals. Introducing point defects in the photocatalyst may also modulate local electronic environment, facilitating reactant polarisation, chemical adsorption and hence electron/hole trapping to promote charge separation36,37. The integration of such dual reaction sites can simultaneously address the concerns on both charge separation and surface reaction dynamics38,39. Besides, a highly dispersed co-catalyst, especially single atoms or atomic clusters with a low co-ordination number, provides unique reaction sites for ready identification of the active local environment for CH4 conversion40,41,42. Accordingly, it is of considerable benefit to develop highly dispersed noble-metal-free dual-site coordinated catalysts.

Herein, tungsten oxide (WO3) nanocrystals were used as the substrate due to its visible-light responsive property largely extending natural light utilisation. Atomic copper co-catalyst and Wδ+ sites associated with oxygen vacancies (Ov) were hybridised to regulate synergistically charge transfer and surface reaction dynamics. Under 420 nm light irradiation, CH4 was converted to HCHO with 96.5% selectivity and a maximum time yield of 12.4 μmol·h−1. Computational simulations indicated that Ov are essential to stabilise the single Cu atoms and induce the formation of adjacent Wδ+ sites. Further mechanistic investigations proved that Cu species acted as the electron acceptors, while Wδ+ species facilitated hole transfer and the preferred adsorption and activation of H2O to generate reactive hydroxyl radicals and then activated CH4. The ensemble co-ordination of the adjacent dual sites of single Cu atoms and Wδ+ species thus resulted in the superior activity and selectivity of CH4 conversion into HCHO at ambient temperature.

Results and discussion

Highly selective methane oxidation to formaldehyde by dioxygen

CH4 conversion reaction was conducted under 420 nm light irradiation with 120 mL distilled H2O, 1 bar O2 and 19 bar CH4 for 2 h at 25 °C. With no photocatalyst or light irradiation, no HCHO or other oxygenate products was detected, suggesting the critical role of photocatalyst and light irradiation. Over pristine WO3, relatively low yields of HCHO (579.5 μmol·g−1) and CO2 (82.1 μmol·g−1) were produced (Fig. 1a), which was mainly ascribed to the severe charge recombination over single-component nanocrystals compared with multicomponent photocatalysts that could facilitate charge separation and transfer24. The selectivity of HCHO was calculated to be 87.6%. To enhance photocatalytic CH4 conversion efficiency, Ov linked with Wδ+ sites, as point defects, were introduced, due to the potential advantages of (i) improving light absorption through injection of sub-band gap or traps43, (ii) enhancing charge separation and transfer36, (iii) promoting chemical adsorption and activation of the symmetric molecule37. Defective WO3–x (denoted def-WO3) was prepared through temperature-programmed thermal reduction from pristine WO3 in a hydrogen atmosphere (5 vol.% H2/Ar). Compared with the pristine counterpart, def-WO3 showed 1.6 times higher HCHO production, being 933.2 μmol·g−1. Moreover, only a trace amount of CO2 by-products (72.2 μmol·g−1) was detected, indicating high selectivity of HCHO (92.8%) and significantly suppressed over-oxidation of CH4 to CO2 or CO.

Fig. 1: Selective CH4 oxidation.
figure 1

a HCHO production over WO3, def-WO3 and Cux-def-WO3 photocatalysts in 2-h reaction. Optimisation of b molar ratio of CH4 to O2, c H2O dosage and d total pressure of CH4 and O2 mixture on HCHO production over Cu0.029-def-WO3. e Reusability of Cu0.029-def-WO3 under five-cycle run. f Comparison of CH4 conversion to HCHO over Cu0.029-def-WO3 and the typical reported photocatalysts10,18,19,20,77. Reaction conditions in a: 5 mg photocatalyst, 120 mL distilled H2O, 420 nm light irradiation and operated at 25 oC. Reaction conditions in b, c, d are identical to a except varied O2 pressure, H2O dosage and total pressure, respectively.

Full size image

To further promote HCHO production, all period 4 transition metal elements, including Sc, Ti to Cu and Zn, were hybridised with the defective WO3–x photocatalyst via. a highly reproducible impregnation method, along with subsequent thermal reduction. The as-prepared photocatalysts with different metal contents were denoted Mx-def-WO3, where x% represented the actual mass percentage of the specific metal (M). The actual metal content was measured by inductively coupled plasma optical emission spectrometry (ICP-OES). Reaction evaluation under identical conditions (Fig. S1) suggested that Cu was the most suitable co-catalyst for HCHO production among these 3d metals. The intrinsic reason why Cu is the most effective among the selected 3d candidates lie not only in its ability to catalyse oxygen reduction reaction, but also its synergistic cooperation with the substrate to promote H2O oxidation so as to activate CH4 as discussed below.

The effect of copper content on HCHO production was then optimised, which exhibited a volcanic trend with increasing percentage of Cu, as shown in Fig. 1a. The highest HCHO yield was over Cu0.029-def-WO3 (4979.0 μmol·g−1). Excessive Cu results in clustering, which may provide recombination centres for the charge carriers44. The dramatically enhanced activity is attributed to the increased availability of photo-generated charge carriers and efficient reactant activation as discussed below. Apart from the excellent HCHO production, a superior HCHO selectivity of 96.5% was also achieved. Moreover, a higher production rate means a higher HCHO concentration around the catalytic site, which may cause over-oxidation. However, no discernible level of CO2 was detected, suggesting that the produced HCHO molecules could be desorbed from the surface of the catalyst in time to avoid deep oxidation.

Effect of oxygen vacancy density on photocatalytic performance was also investigated by subtle control of the thermal reduction temperature with the same impregnated precursors. It was expected that a higher reduction temperature would lead to more Ov via. oxygen extraction by H2. As shown in Fig. S2, HCHO production reached the highest when it was prepared at 250 °C. Too high a temperature may result in the generation of deep trapping sites for charge recombination, while too low a temperature may not generate sufficient regulation of the structural defects.

As shown in Fig. 1b–d, the reaction conditions, including the molar ratio of CH4 to O2, H2O dosage and total pressure, were studied to optimise the HCHO production over Cu0.029-def-WO3. Initially, the effect of the molar ratio of CH4 to O2 was analysed, at a fixed pressure of 20 bar with varied CH4 and O2 dosages (Fig. 1b). Without O2 dosage, the HCHO conversion was 823.9 μmol·g−1 over Cu0.029-def-WO3, suggesting H2O could be the alternative oxygen source for HCHO. In the presence of a very small amount of molecular oxygen, HCHO production gradually increased with raising O2 pressure. Further increasing O2 while decreasing CH4 pressure led to a reduction of HCHO production to 3210.2 μmol·g−1 at CH4/O2 = 15/5. Such suppressed photoactivity would be primarily attributed to the reduced concentration of dissolved CH4. The optimal molar ratio of CH4 to O2 is 19:1, which is clearly higher than the stoichiometric ratio for the production of HCHO (CH4 + O2 → HCHO + H2O) and suggests a relatively large chemical gradient is needed to achieve an effective collision rate of the reactants at the catalytic local environment.

Secondly, H2O dosage was studied under the optimised CH4/O2 molar ratio (19/1). According to Raoult’s law, the molar ratio of the dissolved CH4/O2 is fixed in water. By variation of the H2O dosage, the concentration of the photocatalyst varies, further influencing the light transmittance. As shown in Fig. 1c, the higher the H2O dosage, the higher the HCHO production. The trend is almost linear: With the gradual increase of H2O content from 25 to 120 mL, the HCHO production is enhanced from 890.5 μmol·g−1 to 4979.0 μmol·g−1, corresponding to a concentration of HCHO from 178.1 μmol·L−1 to 207.5 μmol·L−1, respectively (Fig. 1c). The amount of oxidation of HCHO to CO2 would be higher at a higher concentration of HCHO, as indeed observed here. The CO2 production increases from 125.5 to 166.2 μmol·g−1 when water dosage is changed from 25 to 120 mL, or the concentration of HCHO varies from 178.1 to 207.5 μmol·L−1. Interestingly, an improved selectivity to HCHO from 87.6% to 96.5% is also observed, suggesting that the change in CO2 production does not affect the selectivity. Moreover, the HCHO production increases by ~4.6 folds from 890.5 to 4979.0 μmol·g−1 for the H2O dosage increase from 25 to 120 mL, respectively. Therefore, the dilution (or more water content) should preferentially promote the production and timely desorption of HCHO, rather than its oxidation on the surface of the photocatalyst. In other words, the increased HCHO production and selectivity can be attributed to the boosted light absorption and mass transfer of the products as a result of the higher amount of water used, at least under the range of H2O dosages considered here. More importantly, without H2O, a larger amount of CO2 (1038.8 μmol·g−1) is detected, suggesting that H2O can suppress HCHO over-oxidation into CO2. This may be attributed to the solvation effect of H2O, promoting the desorption of oxygenate products23,45.

Thirdly, the HCHO production was assessed over varying feeding pressure of CH4 and O2 under a constant molar ratio of 19 (Fig. 1d). It reveals that HCHO production over Cu0.029-def-WO3 responded almost linearly to the total feeding pressure. As mentioned above, this indicates that no side reaction occurred when increasing the dissolved CH4/O2 at the constant ratio. The intrinsic HCHO product reached 1505.7 μmol·g−1 at ambient pressure. Finally, as shown in Fig. 1e, Cu0.029-def-WO3 performed a relatively stable CH4 conversion to HCHO, with no evident deactivation after five cycles, which confirms the stability of Cu0.029-def-WO3 as a desirable photocatalyst. Meanwhile, the XRD (Fig. S3) and XPS spectra (Fig. S4) comparison of the freshly prepared Cu0.029-def-WO3 and the one after 5 cycles exhibited no discernible difference. Meanwhile, copper was not detected in the filtered reactant by ICP-OES, which further suggests the stable topology of Cu0.029-def-WO3.

The normalised mass activity was calculated using the formula (\({{{{{\rm{Normalized\; activity}}}}}}=\,\frac{{{{{{\rm{n}}}}}}}{{{{{{\rm{m}}}}}}*{{{{{\rm{t}}}}}}}\)), where n, m and t represent the molar production of HCHO (μmol), the mass of Cu co-catalyst (g) and reaction time (h), respectively. Comparison between the Cu0.029-def-WO3 photocatalyst and typically reported photocatalysts for CH4 conversion to HCHO is shown in Fig. 1f. Most of the photocatalysts show relatively low normalized activity for HCHO production. Some noble-metal-modified ZnO and TiO2 photocatalysts under full-arc irradiation exhibit a higher normalized activity between 1.8 × 106 to 5.5 × 106 μmol(HCHO)·g−1(cocatalyst)·h−1, but the selectivity for HCHO is lower than 80%19. Regarding the selectivity, the best-performed Au-WO3 catalyst presents a 100% HCHO selectivity but more than serval order of magnitudes lower normalized activity than our work18. Comparatively, the present noble-metal-free Cu0.029-def-WO3 photocatalyst shows much superior photocatalytic activity with nearly 100% HCHO selectivity and a normalized activity as high as 8.5 × 106 μmol(HCHO)·g−1(cocatalyst)·h−1 (with an apparent quantum yield of 1.14%) under 420 nm irradiation.

Observation of active sites

In order to clarify the mechanism of the high performance, we compared three samples (WO3, def-WO3 and Cu0.029-def-WO3), all of which show a typical and well-crystallised monoclinic WO3 crystal structure (PDF#43-1035) as noted in the X-ray diffraction patterns (XRD) (Fig. S5). The incorporation of copper species through high-temperature reduction did not change the XRD patterns, indicating the stable topology of the WO3 nanocrystal. In addition, there are no extra copper-associated diffraction peaks, due to its high dispersion and/or low loading. Raman spectra (Fig. 2a) further support the WO3 crystalline structure according to the typical WO3 peaks observed at 130.8, 270.4, 713.2 and 804.2 cm−1, which are contributed by lattice vibration, δ(O-W-O) deformation vibration and stretching ν(O-W-O) mode of the bridging oxygen of WO3, respectively46.

Fig. 2: Structural characterisation.
figure 2

a Raman spectra and b high-resolution W4f XPS spectra of WO3, def-WO3 and Cu0.029-def-WO3. c HRTEM and d aberration corrected HAADF-STEM images of Cu0.029-def-WO3. e x–y line scan curve measured along the yellow rectangle region marked in d. f The optimised configuration of the atomic Cu at A site of def-WO3 over the (002) surface in the top view.

Full size image

High-resolution X-ray Photoelectron Spectroscopy (XPS) was conducted to examine the chemical structure of the defective surface of the photocatalyst. As shown in Fig. 2b, the two deconvolution peaks from the W4f XPS spectra at 35.65 and 34.56 eV are observed for the pristine WO3 and attributed to W6+ and W5+ species, respectively. The minor amount of W5+ species (3.4%) could be ascribed to the low concentration of the intrinsic defects in the nanocrystal47. For the def-WO3, the level of the W5+ species increased to 14.4%, over 4 times higher than that of the counterpart, suggesting the successful introduction of Ov linked with the W5+ species. High-resolution O1s XPS spectra (Fig. S6) provides further evidence of the successful introduction of Ov, where the peak at 531.77 eV can be deconvoluted and ascribed to the Ov48. The other two peaks centred at 530.32 and 532.88 eV can be assigned to the crystalline oxygen (W-O) and adsorbed moisture, respectively49. Accordingly, the results indicate that the hydrogen reduction is effective to introduce surface Ov species, which regulate the local electronic environment and generate Wδ+ sites therein50,51,52. Similar to those in the def-WO3, the percentages of W5+ and Ov species of the Cu0.029-def-WO3 are 15.0% and 8.83%, respectively. This indicates that the Ov are preserved after the Cu loading. Besides XPS spectra, Ov are also confirmed through TEM images, EPR and UV-DRS as discusses later.

The transmission electron microscopy (TEM) images of the pristine WO3 and the def-WO3 are shown in Fig. S7a, b and S8, respectively. The former is characterised by a well-crystallised nanocrystal structure with smooth surfaces. High-resolution transmission electron microscopy (HRTEM) is accordance with the XPS analysis. Such defects in the subsurface are reported to result in electron delocalisation and benefit the stabilisation of reaction intermediates53,54,55. The crystalline d-spacings of 0.395 and 0.312 nm in Fig. 2c can be ascribed to (002) and (−112) facets of the WO3 nanocrystal, respectively, indicating the primary structure of WO3 is retained, consistent with the XRD observation. The unsaturated sites of the amorphous/defective layer should promote chemical adsorption and activation of the reactants43. Aberration corrected high-angle annular dark field scanning transmission electron microscopy (HAADF-STEM) images (Fig. 2d) were captured to further investigate the dispersion of the copper species. Irregular tungsten atoms that present on the edge of the nanocrystal further support the existence of Ov-induced Wδ+ sites on the Cu0.029-def-WO3. Besides, it is also shown that W atoms distributed in the regular array, while some additional and smaller dots are clearly observed among the regular array, which may be assigned to Cu atoms. Cu K-edge X-ray absorption near edge structure (XANES) spectra of Cu0.029-def-WO3 with Cu-foil and CuO references were then measured to further study the structural microenvironment of the Cu atoms (Fig. S10). Fourier transforms of the Cu K-edge (Fig. 3a) exhibit only a predominant peak at ca. 1.51 Å for Cu0.029-def-WO3, which can be ascribed to the first shell of the Cu–O bond with reference to the CuO sample (Fig. S11 and Table S1). In parallel, no peaks corresponding to Cu-O-Cu and Cu-Cu at 2.47 Å and 2.23 Å were detected, confirming that the atomically dispersed Cu sites exist in the Cu0.029-def-WO3 and are coordinated with oxygen.

Fig. 3: Structural identification and photo-physical properties.
figure 3

a Fourier transforms of EXAFS of the Cu K-edge of Cu0.029-def-WO3, CuO and Cu-foil. Low-temperature in situ solid-state EPR spectra of b def-WO3 and c Cu0.029-def-WO3 under 420 nm irradiation for different time. d Steady-state PL spectra of WO3, def-WO3 and Cu0.029-def-WO3.

Full size image

The x–y elemental line scan (Fig. 2e) along with the yellow rectangle of Fig. 2d clearly shows the atomic dispersion of Cu in the oxide lattice. The distance between adjacent W and Cu atoms was measured as 0.201 nm along [001] crystalline direction of the (002) plane, which is consistent with the following density functional theory (DFT) calculation results that the Ov is energetically favourable to host a single atomic Cu dopant (Fig. 2f and Fig. S12), where the direct line distance between the nearest W and Cu atoms is 0.434 nm. When projected normal to the [001] direction along the (002) plane, it is 0.218 nm, close to the experimental observation. Thus, the pristine (002) surface model was built based on the monoclinic WO3 bulk structure, shown in Fig. S13a, b. The bottom two layers were fixed, while all other atoms were fully relaxed until the energy and force criteria were reached. The distance of the (002) surface is 3.91 Å, close to the experimentally measured value in this study (3.95 Å). The oxygen vacancy was created by the removal of a single oxygen atom on the top surface. After relaxation, the distance of the nearest W atom stretched to 4.34 Å (Fig. S14a), compared to 3.85 Å in the pristine case. This open-site gives the energetically most favourable host position for an atomic Cu. Three possible adsorption sites were comparatively investigated, as shown in Fig. S14b–d, denoted as Site A, B and C, respectively. The optimised configurations are shown in Fig. 2f. The adsorption energy was calculated based on:

$${E}_{{{{{{{\mathrm{ad}}}}}}}}={E}_{{{{{{{\mathrm{total}}}}}}}}-{E}_{{O}_{v}}-{E}_{{{{{{{\mathrm{Cu}}}}}}}}$$
(1)

where the \({E}_{{{{{{{\mathrm{total}}}}}}}}\), \({E}_{{O}_{v}}\) and \({E}_{{{{{{{\mathrm{Cu}}}}}}}}\) are the energy of the whole system, that of the system with a single O vacancy on the WO3 (002) surface and the chemical energy of copper, respectively. The results are listed in Table S2. Both the A and the B sites are energetically favoured, but the A site is the most stable. The Bader charge analysis was carried out to evaluate the charge changes, and the results are listed in Table S3. For the def-WO3, the Bader partial potential of the W atom decreases from −2.59e for the 2nd nearest to −2.34e for the nearest, suggesting a lowering of the valence of the nearest Wδ+ (δ < 6), induced by the oxygen vacancy. Such reduced oxidation states of W around the defect were also the case for the Cu-def-WO3, which strongly proves the existence of the Wδ+ species at the active site. By accommodating the Cu in the A site (the most stable structure, also consistent with the HAADF-STEM images), the oxidation states of W decreased by 13% from −2.52e to −2.19e. This lends further benefit to W and Cu as the dual active sites for water and oxygen adsorption, respectively, as discussed later.

Mechanistic investigation

UV–Vis diffraction spectra (UV-DRS) (Fig. S15) show similar photoabsorption characteristics among WO3, def-WO3 and Cu0.029-def-WO3, suggesting that photoabsorption is not the dominant factor influencing the photocatalysis herein. The absorption boundary at 426 nm determines the visible-light responsive properties of the as-prepared photocatalysts. Bandgap energies calculated by the Tauc plots (Fig. S16) are 2.91, 2.88 and 2.88 eV for WO3, def-WO3 and Cu0.029-def-WO3, respectively. Additionally, Mott-Schottky plots (Fig. S17) were measured to establish the flat band position. Positive slopes of the Cu0.029-def-WO3 photocatalyst indicate the n-type characteristics of WO3, for which the flat band always lies 0.1 V below the conduction band (CB)56. Thus, CB and valence band (VB) positions of the Cu0.029-def-WO3 vs. NHE (pH = 0) are established as −0.10 and 2.78 V, respectively57,58. Therefore, the potentials of VB and CB positions are sufficient to drive O2 reduction (−0.05 V vs. NHE) and H2O oxidation (2.38 V vs. NHE) to generate reactive species, such as ·OOH and ·OH radicals, respectively59,60.

Electron paramagnetic resonance (EPR) is a highly sensitive tool for the study of the paramagnetic transition of metal ions and the oxygen vacancies due to unbalanced electron spins61. As shown in Fig. S18, no evident EPR signal was detected for WO3, suggesting its pristine topology with no unpaired electrons. After the thermal reduction in hydrogen, the def-WO3 exhibited a single Lorentzian EPR signal at g = 2.002, which can be assigned to an oxygen defective structure49. In the case of the Cu0.029-def-WO3, a similar EPR signal at g = 2.002 was also observed, but weaker than that of the def-WO3, thus indicating that the introduced copper species are very likely on/around Ov so reducing its EPR response62. Meanwhile, an additional hyperfine peak at g = 2.061 is observed, which is attributed to the hybridized Cu2+ species63,64. The hyperfine structure of this peak provides further evidence of the high dispersion of the copper species.

In situ light-irradiated EPR spectra of the def-WO3 and the Cu0.029-def-WO3 were then tested to probe further details of the charge transfer mechanism. For the def-WO3, a newly added EPR signal at g = 2.007 was observed under light irradiation (Fig. 3b), which is likely due to the excited electrons at the conduction band of WO3. In parallel, the signal of Ov at g = 2.002 gradually becomes stronger under light irradiation, suggesting Ov facilitate the trapping of active electrons from the conduction band and an equilibrium is reached after 120 s. In contrast, no noticeable change at g = 2.007 and 2.002 (Fig. S19) was observed for the Cu0.029-def-WO3 under identical conditions, suggesting that the photo-induced electrons at the conduction band of WO3 and the Ov could efficiently migrate to the Cu species. In addition, the constant intensity of the Lorentzian signal at g = 2.002 would suggest that the Ov serve as the mediator for electron transfer from the conduction band to the Cu species in the Cu0.029-def-WO3. Accordingly, a hyperfine EPR signal at g = 2.056 (Fig. 3c) was observed for the Cu0.029-def-WO3, which is ascribed to the Cu2+ species as Cu+ and Cu0 are EPR-silent. Under continued light irradiation, the Cu2+ signal gradually weakens, indicating a reduced content of the Cu2+ species, implying that the Cu2+ serves as the electron acceptor. In situ Cu2p XPS spectra (Fig. S20) with and without light irradiation were obtained for further investigation of the role of the Cu species. In dark, the deconvoluted Cu2p XPS results validate the coexistence of Cu0/Cu+ (932.38 eV) and Cu2+ (933.60 eV) for the Cu0.029-def-WO3, which corresponds to the composition of 66% Cu0/Cu+ and 34% Cu2+, respectively. Under light irradiation, the concentration of Cu2+ dramatically decreased to 19% while Cu0/Cu+ increased to 81%, further confirming the active role of Cu2+ as the electron acceptor.

Steady-state and time-decay photoluminescence (PL) spectra (Fig. 3d and S21) were conducted to study the charge transfer dynamics. The strong PL peak of the pristine WO3 is correlated with its severe charge recombination, as expected for the single component24. Comparatively, the def-WO3 shows relatively weaker PL intensity than WO3, indicating an enhanced charge separation efficiency65, which should originate from the role of Ov in the def-WO3 as the electron trapping sites. In the case of the Cu0.029-def-WO3, PL intensity decreased further, being the lowest, among the different Cu-based photocatalysts, implying the most suppressed charge recombination case after the incorporation of the atomic Cu. Notably, different from the def-WO3, the Ov in the Cu0.029-def-WO3 contribute little to charge separation, as evidenced by the unchanged EPR intensity under light irradiation (Fig. S19), suggesting the highly dispersed Cu co-catalyst is more efficient than the Ov in acting as the photo-induced electron acceptors. DFT simulations were then employed to show the charge density distribution of the conduction band minimum (Fig. 4a) and the valence band maximum (Fig. 4b) of the Cu-def-WO3. The charges are found to significantly accumulate around the Cu and the Ov region in the conduction band minimum (Fig. 4a). From Bader charge analysis in Table S3, the partial charge of the related O atom changes from 0.58e for the pristine WO3, to 1.06e for the def-WO3, and 0.98e for Cu-def-WO3. It confirms our speculations in EPR results. From the in situ EPR and XPS analyses, with further support from DFT simulations, it may be inferred that photo-excited electrons in the conduction band are firstly trapped by the Ov, and then transferred to the Cu species to facilitate the formation of Cu+ (Cu2+ + e- → Cu+), where Cu+ would next act as the electron donor to activate the adsorbed O2 on the surface (Cu+ + O2 → Cu2+ + ·OO-). Under reaction conditions in the presence of H2O, superoxide radicals (·OO-) tend to react with a proton and form ·OOH radicals. In parallel, photo-induced holes locate around the hybridized O and Wδ+ (Fig. 4b).

Fig. 4: Photo-chemical properties.
figure 4

a Conduction band minimum and b Valence band maximum charge distributions of Cu-def-WO3. The grey, red and blue balls represent W, O and Cu atoms, respectively. In situ EPR monitor of c ·OOH and d ·OH radicals trapped by DMPO over WO3, def-WO3 and Cu0.029-def-WO3 under light. e Time-dependent PL intensity of 7-hydroxycoumain from the reaction of coumarin and ·OH over WO3, def-WO3 and Cu0.029-def-WO3. f GC-MS of the produced HCHO with isotopic labelled H218O or 18O2 in the presence of 19 bar CH4 and 1 bar O2 and 3 mL H2O over Cu0.029-def-WO3.

Full size image

Time-decay PL spectra (Fig. S22) were fitted by the two-exponential decay. All samples exhibit similar exponential emission properties. Compared with the WO3 and the def-WO3, a much slower decayed emission of the Cu0.029-def-WO3 demonstrates slower kinetics of the fluorescent decay and the suppressed charge recombination. The average PL lifetime of the Cu0.029-def-WO3 was calculated to be 4.61 ns (Table S4), longer than that of the WO3 (3.14 ns) and the def-WO3 (3.27 ns), indicating that the Cu0.029-def-WO3 most facilitates the charge carrier separation. Photocurrent responses were measured to evaluate further the charge separation dynamics. As shown in Fig. S23, the def-WO3 shows a photocurrent intensity of 12.2 μA·cm−2, 1.9 times stronger than that of the pristine WO3 (6.3 μA·cm−2), demonstrating the efficient transfer of photo-induced electrons to the defects. For Cu0.029-def-WO3, photocurrent density is further increased to 14.8 μA·cm−2, about 1.2 times higher than that of the def-WO3, suggesting the role of the Cu species in promoting charge separation. Electrochemical impedance spectroscopy (EIS) plots (Fig. S24) show a much smaller radius of Cu0.029-def-WO3 than those of def-WO3 and WO3, indicating its smallest resistance for interfacial charge transfer. From the above analysis, it is clear that the charge recombination is greatly suppressed by suitable ensemble corporation of the Ov-Wδ+ sites and highly dispersed Cu-co-catalyst atoms, which result in efficient charge separation and transfer.

Reactive oxygen species, including hydroperoxyl (·OOH) and hydroxyl radicals (·OH), were monitored with 5, 5-dimethyl-1-pyrroline N-oxide (DMPO) as the trapping agent by the detection of DMPO-OOH and DMPO-OH adducts, respectively. The intermediate of O2 reduction was measured in methanol solution under light irradiation (Fig. 4c). There were clearly six prominent characteristic signals of the DMPO-OOH adduct at hyperfine splitting constants of AN = 15.4 G and AH = 10.6 G over WO3, def-WO3 and Cu0.029-def-WO3 photocatalysts66. Cu0.029-def-WO3 and def-WO3 show 2.0 and 1.8 times stronger ·OOH production than the pristine WO3, due to the highly suppressed charge recombination induced by the cooperation of Cu single atoms and Ov. Figure 4d shows the four signal peaks (1: 2: 2: 1) of ·OH over the three photocatalysts. The trend of ·OH production from H2O oxidation with photo-induced holes (H2O + h+ → ·OH + H+) is similar to ·OOH production among WO3, def-WO3 and Cu0.029-def-WO3, due to the enhanced charge separation. Quantification experiments further support that the production of ·OH radicals is directly associated with charge separation. The ·OH test using coumarin as the probe molecule (Fig. 4e) supports that the Cu0.029-def-WO3 is more efficient than the WO3 and the def-WO3. Compared with the pristine WO3, the generation of the trapped species (coumarin + ·OH → 7-hydroxycoumain) is improved by 2.2 and 2.6 times for def-WO3 and Cu0.029-def-WO3 over a 30 min reaction, confirming improved ·OH radicals due to the synergy between Cu single atoms and Ov. Such trend is consistent with the in situ EPR under light. As ·OH could activate CH4 to produce methyl radical (·CH3) (·OH + CH4 → ·CH3 + H2O), a higher amount of ·OH produced would be more beneficial to CH4 activation. Accordingly, both ·OOH and ·OH radicals are the reactive oxygen species during the photocatalytic CH4 conversion.

Initially, individual reactant adsorption was also simulated on the optimised photocatalysts. For oxygen adsorption on the Cu-def-WO3, simulation results reveal that the O2 bond length is largely stretched to 1.41 Å in the molecular form on the Cu (Fig. S25). Due to the low oxidation states of the Cuδ+, this stretched O2 is further readily protonated to ·OOH. Further, water adsorption on the Wδ+ site near an oxygen vacancy was compared on the optimised Cu-def-WO3 and the pristine WO3 surface (Fig. S26). The calculated free reaction energy is −0.78 eV and −0.65 eV, respectively, which indicates that both Cu and Wδ+ enhance the interaction with water. The Wδ+ with lower oxidation states (δ < 6) near the Ov defect is ready to host the lone-pair electrons from the water. Thus, both water and Wδ+ present near the Ov is the key to providing hole attraction and reaction sites for CH4 activation. Since CH4 could competitively react on the electron trapping site (Cu) or the hole trapping site (Wδ+), the calculated free energy of the reaction is −0.04 and −0.43 eV on these two sites, respectively, as shown in the optimised geometries (Fig. S27), which indicates Wδ+ site is the active site for methane activation. Nevertheless, it is still more difficult for CH4 to be activated on Wδ+ compared with H2O due to the much smaller free reaction energy of water (−0.78 eV), which then indicates CH4 would be more likely to be directly activated by the ·OH radicals, rather than the photo-holes, through H2O oxidation with the photo-holes (Figs. S28 and S29).

Oxygen sources of HCHO were further investigated through the isotopic experiment with 18O2 and H218O, separately. To confirm O2 as the oxygen source, 1 bar 18O2 and 19 bar CH4 over 20 mg Cu0.029-def-WO3 with 3 mL H2O were utilised for a 6-h reaction. GC-MS results (Fig. 4f) clearly show that the detected HCHO is composed of 81% HCH16O and 19% HCH18O, suggesting both O2 and H2O are the oxygen sources for HCHO production. Meanwhile, a much higher content of HCH16O suggests H2O serves as the major oxygen source. Photocatalytic CH4 conversion in the presence of isotopic labelled H218O further supports the primary role of H2O on HCHO production due to the reversed composition of 20% HCH16O and 80% HCH18O. The participation of H2O in HCHO formation suggests that the ·OH radicals formed from the oxidation of H2O with photo-induced holes (H2O + h+ → ·OH + H+) could directly react with CH4 or ·CH3 to form *CH3OH (“*” indicates it’s still adsorbed), and then to HCHO. Isotopic experiments were also conducted with 5 bar 13CH4 and 1 bar O2 over 20 mg Cu0.029-def-WO3 in 3 mL H2O for a 6-h reaction. 13C nuclear magnetic resonance (NMR) spectroscopy in Fig. S30 shows only one signal at 82.2 ppm, assigned to HO13CH2OH, which was the diol structure of H13CHO in an aqueous solution (H13CHO + H2O → HO13CH2OH)18,20. This result confirms that the carbon source of the produced HCHO comes from CH4. Moreover, it also confirms the high selectivity with no CH3OH and HCOOH production, consistent with the 1H NMR results (Fig. S31).

Based on the above results, particularly based on the strong evidence of the characteristics of the ·OH and ·OOH radicals by in situ EPR, a radical mechanism (Fig. 5a, b) of aerobic photocatalytic CH4 conversion to HCHO is proposed here over the Cu0.029-def-WO3 catalyst, which consists of nanocrystalline WO3 substrate with single atoms Cuδ+ and Ov /(Wδ+). In general, under visible-light irradiation, electrons are excited from the valence band (VB) to the conduction band (CB) of WO3. The active electrons then migrate to Cu2+ to form Cu+ (Cu2+-e-). The adsorbed O2 on the Cu site next reacts with the electrons trapped by Cu+ to form the ·OOH radical, which requires a potential of −0.046 eV [63]. In parallel, photo-induced holes are trapped by the Wδ+ with low oxidation (Wδ+ + h+ → W(δ+1)+), and then react with the chemical adsorbed H2O at the oxygen vacancy induced Wδ+ site to form ·OH radicals. Meanwhile, the ·OH radical could directly activate CH4 and then react with ·CH3 to form *CH3OH intermediate, then being transformed to the desired HCHO. The absence of CH3OOH and CH3OH in 1H NMR spectra (Fig. S32) may be attributed to the immediate conversion of such intermediates to HCHO. On the basis of efficient activation of the different reactants (H2O, O2 and CH4) with the assistance of the specific synergy of Ov induced Wδ+ site and highly dispersed Cu atomic co-catalyst, the selectivity of HCHO over Cu0.029-def-WO3 was successfully maintained at nearly 100% over a long period of time to 10 h.

Fig. 5: Proposed mechanism.
figure 5

a, b Schematic of charge transfer steps during selective CH4 oxidation over Cu0.029-def-WO3.

Full size image

In summary, an effective strategy to overcome the dilemma of enhancing the efficient activation while suppressing the over-oxidation of CH4 has been developed for the selective production of value-added HCHO over a noble-metal-free Cu0.029-def-WO3 photocatalyst. The WO3 substrate provides the visible-light responsive activity for CH4 conversion, while the atomically dispersed Cu acts as an effective electron acceptor, as indicated from the analysis of the in situ XPS and EPR spectra under light irradiation. Ov induce the formation of reactive Wδ+ species, which further enhances the selective chemical adsorption and activation of H2O. The coordinated ensemble of the dual active sites synergistically leads to efficient charge separation and transfer. As a result, under visible-light irradiation at room temperature, a superior photocatalytic CH4 conversion efficiency has been achieved with a high TOF of 8.5 × 106 μmol(HCHO)·g−1(cocatalyst)·h−1, out-performing previously reported photocatalysts, even much better than the noble-metal photocatalytic processes19. Nearly 100% selectivity and a high HCHO evolution rate of 4979.0 μmol·g−1 have been achieved under 420 nm light irradiation at room temperature. Isotopic experiments provide strong evidence that both O2 and H2O are the oxygen sources for HCHO production, with H2O being the major one. This work provides a new avenue for simultaneous regulation of CH4 activation under visible-light irradiation and suppression of over-oxidation by incorporation of adjacent dual active sites, which is of great interest in net-zero green conversion and upgrading of natural resources to high-value chemicals.

Methods

Preparation of Cux-def-WO3

Cux-def-WO3 photocatalysts were prepared by a highly reproducible incipient impregnation method and subsequent hydrogen reduction. In a typical experiment, 5.0 g WO3 was first mixed with 2.0 mL CuCl2 aqueous solution containing a certain amount of CuCl2. After uniformly stirring and ageing at ambient temperature overnight, the obtained solid was dried at 60 oC for another 12 h. The dried samples were then grounded and calcined at 250 oC for 2 h at a ramping rate of 5 oC/min in 5 vol.% H2/Ar flow (80 mL/min). The as-prepared samples were denoted Cux-def-WO3, where x% represented the mass percentage of the copper atom. For comparison, the defective WO3–x (denoted def-WO3) was prepared under identical conditions but without CuCl2 addition.

Characterisations

XRD patterns were acquired on the D8 ADVANCE diffractometer (Bruker Co., Ltd). Raman spectra were acquired on the DXR 2DXR2 instrument (Thermo Fisher Scientific, Co., Ltd). XPS spectra were measured with the PHI 5000 Versa Probe III instrument (ULVAC-PHI Co., Ltd). In situ XPS spectra were collected in dark or under visible-light irradiation for 20 min on the Thermo ESCALAB 250Xi instrument with an Al Kα radiation source. In situ solid-state EPR spectra were measured on the ELEXSYS II EPR instrument with 20 mg photocatalyst in dark or under 420 nm light irradiation. HRTEM images were captured on the Talos F200X instrument (FEI Co., Ltd). Where photocatalyst was pre-dispersed under sonication and dipped on the molybdenum mesh as the copper-free supporting substrate. UV–vis DRS spectra were recorded on the UV-3600 Plus spectrometer (Shimadzu Co., Ltd) with spectroscopic pure BaSO4 as the references. Steady-state and time-decay PL spectra were recorded at room temperature on the QM400 and FLSP920 spectrofluorometers, respectively.

Photocatalytic methane conversion

Photocatalytic CH4 conversion reaction was conducted in the 200 mL stainless-steel high-pressure reactor equipped with Teflon lining. Typically, 5 mg photocatalyst was first dispersed in 120 mL distilled H2O under stirring. The reactor was then sealed and purged with ultrapure oxygen (99.99 vol.%) under moderate stirring for about 20 min to remove the air completely. After acquiring atmospheric oxygen or desired partial pressure, ultrapure CH4 (99.99 vol.%) was injected into the reactor to achieve total pressure of 20 bar. The reactor with top-irradiation was then irradiated with the LED lamp source (420 nm, PLS-LED100C, Beijing Perfectlight Technology Co., Ltd). Photocatalytic reaction was conducted for 2 h at 25 oC. The gaseous products were measured by gas chromatograph (GC2014, Shimadzu Co., Ltd) equipped with a thermal conductivity detector and flame ionisation detector. The dissolved CO2 was also analysed through the precipitation method with excessive Ba(OH)2, where excess Ba(OH)2 was added into 10 mL supernate after reaction67. No discernible precipitation was observed, indicating the dissolved CO2 was too low to be measured. The possible existence of other products, including CH3OOH, CH3OH and HCOOH, was analysed by 1H nuclear magnetic resonance spectroscopy (AVANCE III, JEOL Ltd). The results (Fig. S31) confirmed that no CH3OOH, CH3OH and HCOOH were produced during the CH4 conversion reaction by the photocatalyst. HCHO was measured by the colorimetric method based on the reaction between acetylacetone and HCHO in the presence of acetic acid and ammonium acetate68. A trace amount of CO2 was detected during all photocatalytic reactions, demonstrating the ultrahigh HCHO selectivity.

Reusability of Cu0.029-def-WO3 was tested by photocatalytic CH4 conversion. Parallel experiments was conducted under identical conditions to replenish the losses of photocatalyst. After each experiment, the reactant was centrifuged and dried at 60 oC in vacuum. Then 5 mg collected photocatalyst was re-used for the next cycling experiment.

Isotopic experiments

For carbon source investigation with isotopic labelled 13CH4: 20 mg Cu0.029-def-WO3 photocatalyst was dispersed in 3 mL H2O, then the reactor was vacuumed for 30 min and refilled with 1 bar O2 and 5 bar 13CH4. The reaction was conducted for 6 h to gain more concentrated products for detection facilitation. As H13CHO tended to exist as HO13CH2OH in aqueous solution, the products were identified by 13C NMR (AVANCE III, JEOL Ltd).

For oxygen source investigation with isotopically labelled 18O2: 20 mg Cu0.029-def-WO3 photocatalyst was dispersed in 3 mL H2O, then the reactor was vacuumed for 30 min and refilled with 1 bar 18O2 and 19 bar CH4. The reaction was conducted for 6 h to gain more concentrated products for detection facilitation. The as-produced HCH18O was measured with GC-MS (QP2020, Shimadzu Co., Ltd) equipped with the Cap WAX column to identify the existence of H218O.

For oxygen source investigation with isotopic labelled H218O: 20 mg Cu0.029-def-WO3 photocatalyst was dispersed in 3 mL H2O, then the reactor was vacuumed for 30 min and refilled with 1 bar 18O2 and 19 bar CH4. The reaction was conducted for 6 h to gain more concentrated products for detection facilitation. The as-produced HCH18O was measured with GC-MS (QP2020, Shimadzu Co., Ltd) equipped with the Cap WAX column to identify the existence of H218O.

Photoelectrochemical measurements

Mott-Schottky plots, electrochemical impedance spectroscopy (EIS) and photocurrent density were measured on the three-electrode system with an electronic workstation (CHI660E). Photocatalyst coated by indium tin oxide (ITO) glass (10 × 10 mm), Ag/AgCl electrode and platinum sheet electrode were respectively employed for working, reference and counter electrodes, with 0.1 mol·L−1 Na2SO4 solution as the electrolyte. The working electrodes were prepared by scraping the paste-like mixture containing 100 mg of different photocatalysts, 450 μL ethanol and 50 μL Nafion solution (Shanghai Adamas Reagent Co., Ltd). Before measurement, the working electrodes were dried at 60 °C in vacuum.

Monitor of the reactive oxygen species

5, 5-dimethyl-1-pyrroline N-oxide (DMPO) was used as the trapping agent for the monitoring of the reactive oxygen species, including hydroperoxyl (·OOH) and hydroxyl (·OH) radicals. For monitoring the generation of ·OOH radical, 10 mg Cu0.029-def-WO3 photocatalyst was dispersed in 5 mL methanol in the dark and purged with ultrapure O2 (99.99 vol.%) for 20 min. For monitoring the generation of ·OH radical, 5 mg Cu0.029-def-WO3 photocatalyst was dispersed in 5 mL distilled H2O in the dark and purged with ultrapure argon (99.99 vol.%) for 20 min. In situ EPR spectra of the above suspension in dark and under light were then obtained on the ELEXSYS II EPR instrument.

Analysis the productivity of ·OH radical

Coumarin was used as the probe to evaluate the productivity of ·OH radical due to the reaction between coumarin and ·OH to produce 7-hydroxycoumain (7HC). Typically, 20 mg Cu0.029-def-WO3 photocatalyst was dispersed in 100 mL aqueous coumarin solution (0.5 mM). After stirring in dark for 30 min, 5 mL reactant was sampled every 5 min under 420 nm light irradiation. After filtration, the PL spectra of the formed 7HC were measured by F4500 spectrofluorometer.

Theoretical calculations

All the calculations were performed based on density functional theory (DFT), implemented in the Vienna ab initio Package (VASP)69. Electron-ion interactions were described with the projector-augmented-wave (PAW) method with an energy cut-off set to 500 eV70. The PBE functional was used to optimise the structural models and to analyse the electronic structures71. The structures were fully relaxed until the changes in energy and in the force upon ionic displacement were no greater than 10−5 eV and 0.02 eV/Å, respectively. The delocalised electron was treated by the (DFT + U) Dudarev approach with the effective U value set to 5 and 8 for tungsten and copper, respectively72. The U values are in line with previous calculations on WO373. Due to the complex oxidation states of Cu in Cu-def-WO3, we have adopted the U value from one report74, as this U value has been well assessed for Cu2O, CuO and Cu3O4. The partial charges were determined using the Bader charge analysis75. The chemical potential of Cu single atom is derived from the bulk copper. The entropy of gas phase H2O, O2, HCHO are obtained from the NIST database under standard conditions76.

The adsorption energy (Ead) is calculated by

$${E}_{{{{{{{\mathrm{ad}}}}}}}}={E}_{{{{{{{\mathrm{total}}}}}}}}-{E}_{{{{{{{\mathrm{adsorbent}}}}}}}}-{E}_{{{{{{{\mathrm{sub}}}}}}}}$$
(2)

where the Etotal, Eadsorbent and Esub is the total energy of the system, the energy of the adsorbent and the substrate, respectively.

The reaction free energy is calculated according the following equation:

$$\triangle G={E}_{{{{{{{\mathrm{ad}}}}}}}}+\triangle {{{{{{\mathrm{ZPE}}}}}}}-T\triangle S$$
(3)

where the Ead is the adsorption energy, ΔZPE is the changes of the zero point energy, ΔS is the changes of the entropy of the reaction. The internal thermal energy U0→T is calculated based on

$${U}_{0\to T}=R{T}^{2}{\left(\frac{\delta {{{{{\rm{ln}}}}}}q}{\delta T}\right)}_{V}$$
(4)

The reaction temperature is set to 300 K.