Introduction

Solar-driven photocatalytic H2O2 production through oxygen reduction reaction (ORR) is a promising method for addressing energy and environmental crises due to its low energy consumption, safety, and environmental friendliness1,2,3,4,5. However, the production of H2O2 via photocatalysts is hindered by the low efficiency of electron transfer and interface reaction6,7,8, resulting in suboptimal H2O2 yields9,10. To address these challenges, cocatalyst deposition on the surface of photocatalysts can not only effectively promote electron transfer but also provide specialized active sites to facilitate interfacial ORR11,12,13. It is well known that photocatalytic H2O2 production via ORR on the active sites of cocatalysts involves multiple fundamental steps, such as O2 adsorption, intermediate *OOH formation, and H2O2 desorption8,14,15. Of these, O2 adsorption at active sites is one of the most important processes, as it facilitates the formation of intermediate *OOH and its further conversion into H2O216,17,18,19. Sabatier principle20. Suggests that the interaction between active sites and adsorbates must have an optimal binding energy. Further research indicates that the electron configuration of active sites fundamentally determines their interaction with adsorbates, influencing their adsorption/desorption performances21. However, in photocatalytic H2O2 production, current cocatalysts usually suffer from a mismatch in the electronic configuration between the active site and the adsorbed O2, leading to either excessively strong or weak O2 adsorption, which in turn limits H2O2-production rates22,23,24. Therefore, it is quite meaningful and challenging to modulate the electronic configuration of cocatalysts and optimize their oxygen adsorption strength to achieve efficient H2O2 production.

Currently, noble metal cocatalysts (Pd, Pt, and Au) have made significant advancements in improving photocatalytic ORR for H2O2 production25,26,27,28. Notably, Au cocatalysts usually exhibit a higher photocatalytic H2O2-production activity, attributed to the effective interfacial charge transfer between photocatalysts and Au nanoparticles29. This transfer enables the rapid movement of photogenerated electrons from the photocatalysts to the Au surface, facilitating the reduction of adsorbed O2 to H2O2 through either a two-step single-electron or a one-step two-electron ORR process30,31,32. However, metallic Au usually exhibits weak oxygen adsorption characteristics due to its intrinsic electronic structure (Fig. 1a-(1))33,34, which limits the formation of the *OOH intermediate and subsequent H2O2 production. Consequently, precise modulation of Au’s electronic structure is extremely crucial to optimize O2 adsorption and enhance photocatalytic H2O2 production. For instance, Tsukamoto et al. demonstrated increased electronic density in Au by forming an Au-Ag alloy, reducing H2O2 decomposition on Au sites35. Moreover, Wang et al. prepared an efficient core-shell Cu@Au-modified BiVO4 nanostructure. This structure reduces negative charge accumulation at the Au active sites by forming an ohmic contact with Cu/BiVO4, thereby enhancing the adsorption of O2 and its intermediate *OOH, leading to efficient photocatalytic H2O2-production performance36. Although the introduction of alloy and bimetallic core-shell structures have efficiently enhanced the photocatalytic activity of Au sites, the relationship about the H2O2-production activity, the Au-Oads bonds, and the electronic structure of Au remains unclear. Fortunately, the molecular orbital theory clearly states that the antibonding-orbital occupancy degree between a metal and its adsorbate usually determines its adsorption energy37, which provides a theoretical basis for the modulation of bond strength between Au and O2. Inspired by this, selectively decreasing the antibonding-orbital occupancy of Au-Oads is expected to further enhance the O2 adsorption on Au, potentially achieving efficient photocatalytic H2O2 production. However, there has been limited research focusing on this approach to date.

Fig. 1: Strategy to design efficient electron-deficient Auδ+ cocatalyst for improving H2O2-production kinetics.
figure 1

a Schematic illustration of electron-deficient Auδ+ formation to reinforce Au-Oads bond: (1) weak Au-Oads bond on Au surface; (2) the formation of electron-deficient Auδ+ sites by MoSx incorporation; (3) strong Au-Oads bond on MoSx-Au surface. b Schematic diagram about reducing the antibonding-orbital occupancy of Au-Oads by the free-electron transfer from Au to MoSx cocatalyst.

Full size image

In this work, we propose an approach to strengthen the Au-Oads bonds by modifying the electronic structure of Au active site. This is achieved through the introduction of molybdenum sulfide (MoSx) as an electron mediator to decrease the antibonding-orbital occupancy of Au-Oads. In this case, the MoSx mediator serves to adjust the electronic structure of Au cocatalyst, resulting in the creation of electron-deficient Auδ+ active sites and subsequently accelerating H2O2 production (Fig. 1a-(2) and -(3)). To this end, TiO2/MoSx-Au photocatalyst was synthesized by a two-step method. This process involves the initial MoSx deposition onto the TiO2 surface and subsequent S-induced selective photodeposition of Au cocatalyst onto the MoSx surface. The resulting TiO2/MoSx-Au photocatalyst achieves a boosted H2O2-production rate of 30.44 mmol g−1 h−1, which is 25.4 and 1.3 times higher than TiO2 and TiO2/Au, respectively. Density functional theory (DFT) calculations and ex-situ X-ray photoelectron spectroscopy (XPS) analysis have confirmed the effective reduction of d-orbital electron on Au cocatalyst upon the introduction of MoSx, leading to a decrease in antibonding-orbital occupancy in Au-Oads (Fig. 1b). Consequently, the Au-Oads bonds are significantly reinforced, which in turn enhances the rate of photocatalytic H2O2 production. This work focuses on modifying the electron structure of Au cocatalyst to reduce the antibonding-orbital occupancy of Au-Oads, offering a promising approach to enhance O2 adsorption for efficient photocatalytic H2O2 production.

Results and discussion

Synthesis and characterization of TiO2/MoSx-Au

To realize the successful deposition of MoSx-Au cocatalyst on the surface of TiO2 photocatalyst, a facile two-step route was carried out at room temperature (Fig. 2a), including the initial deposition of MoSx on the TiO2 surface and the subsequently selective photodeposition of Au onto the MoSx surface (Supplementary Fig. 1). First, (NH4)2MoS4 solution was mixed with a lactic acid solution to form brown MoSx colloidal nanoparticles (Supplementary Fig. 1a-(1)). Subsequently, TiO2 nanoparticles were uniformly dispersed into this colloidal solution with constant stirring. The positive charge on the TiO2 nanoparticles in the lactic acid solution allowed for the efficient adsorption of MoSx colloidal nanoparticles onto the TiO2 surface via electrostatic self-assembly (Supplementary Fig. 1a-(2), c). The deposition of MoSx on the TiO2 surface can be verified by Fourier transform infrared (FTIR) spectroscopy and Raman spectra analyses (Supplementary Fig. 2). A new FTIR peak for S-S vibration and a new Raman peak for Mo-S can be observed, providing strong evidence for the MoSx formation38,39. With the further addition of HAuCl4 solution into the TiO2/MoSx suspension, AuCl4- ions can be selectively adsorbed onto the MoSx surface via the strong interaction between S and Au atoms (Fig. 2a)40. Upon light irradiation, the AuCl4- ions were in situ reduced to form Au nanoparticles on the MoSx surface (Supplementary Figs. 1a-(3), d), evident from a color change from light brown to purple (Supplementary Fig. 1b), revealing the successful synthesis of TiO2/MoSx-Au photocatalyst. The above result can further be supported by X-ray diffraction (XRD) and Raman spectra. Compared with the TiO2/MoSx sample, a new XRD peak of Au at 38.1° and an Au-S Raman peak41,42 confirm the selective deposition of Au nanoparticles on the MoSx surface (Supplementary Figs. 3 and 4).

Fig. 2: Synthetic strategy and morphology characterization.
figure 2

a Schematic illustration for the synthesis of TiO2/MoSx-Au by the initial lactic acid-induced MoSx deposition on the TiO2 surface and subsequent S-induced selective photodeposition of Au cocatalyst onto the MoSx surface. b, c TEM, d HRTEM, e, f HAADF-STEM, and g–l elemental mapping pictures of TiO2/MoSx-Au photocatalyst.

Full size image

Transmission electron microscopy (TEM) analysis was employed to further verify the selective deposition of Au on the MoSx surface within the TiO2/MoSx-Au photocatalyst. As depicted in Fig. 2b–d, numerous dark spots on the TiO2 surface are observed, which can be attributed to the MoSx-Au cocatalyst. The high-angle annular dark-field (HAADF) images (Fig. 2e, f) further indicate that the MoSx-Au nanoparticles were effectively deposited on the TiO2 surface. The corresponding energy-dispersive X-ray spectroscopy (EDS) mapping images (Fig. 2g–l) show that four Au nanoparticles exhibit uniform and same distribution with the Mo and S elements on the TiO2 particles, unequivocally indicating that all the Au nanoparticles are selectively deposited on the MoSx surface via the self-assembly of S-Au. Ultraviolet-visible absorption spectra (UV-Vis) demonstrate a typical surface plasmon resonance (SPR) absorption of Au (ca. 540 nm) (Supplementary Fig. 5)40,43,44, suggesting that Au was effectively deposited on the surfaces of both TiO2 and TiO2/MoSx. Noticeably, the SPR signal in the TiO2/MoSx-Au appears relatively weaker than that of the TiO2/Au, which can be attributed to the strong interaction between Au and S atoms. According to the inductively coupled plasma optical emission spectrometry (ICP-OES) results (Supplementary Table S1), the contents of Mo, S, and Au elements in the TiO2/MoSx-Au photocatalyst are 0.62, 0.8, and 3.34 wt%, respectively, indicating the presence of MoSx and Au in the TiO2/MoSx-Au system. These results (HRTEM, UV-Vis, and ICP-OES) collectively support the selective deposition of Au on the MoSx surface.

Photocatalytic performance

The photocatalytic activities for H2O2 production were conducted in an O2-saturated ethanol solution under Xe lamp irradiation. As shown in Fig. 3a, TiO2 exhibits a low H2O2-production rate of 1.20 mmol g−1 h−1. However, the introduction of Au nanoparticles onto the TiO2 surface leads to an improved H2O2-production rate (24.22 mmol g−1 h−1) for the resulting TiO2/Au. With the further incorporation of MoSx cocatalyst into TiO2/Au, the TiO2/MoSx-Au shows a significant enhancement in the photocatalytic H2O2-production activity. Further investigation indicated that the H2O2-production activity of TiO2/MoSx-Au photocatalysts is dependent on the Au amount (Supplementary Fig. 6). When the Au is precisely maintained at 3%, the resulting TiO2/MoSx-Au sample exhibits the highest H2O2-production rate with a value of 30.44 mmol g−1 h−1 (Fig. 3a and Supplementary Fig. 6), which is 25.4 and 1.3 times higher than that of TiO2 and TiO2/Au, respectively. However, increasing the Au content beyond this point to 5% leads to a reduction in H2O2 yield due to the light-shielding effect. According to the wavelength-dependent H2O2-evolution activity (Fig. 3b), the AQY of TiO2/MoSx-Au achieves an impressive value of 7.2% at 365 nm. The present H2O2-evolution performance is much higher than most reported results in TiO2-based photocatalysts or other inorganic photocatalysts (Supplementary Table S2). To investigate the application potential of present TiO2/MoSx-Au, its photocatalytic H2O2-evolution performance was further tested under different conditions. As exhibited in Fig. 3c and Supplementary Fig. 7, it is clear that there is almost no H2O2 generation over TiO2 in the air environment. In contrast, the H2O2 yield in the TiO2/MoSx-Au still maintains a high concentration (14.45 mmol g−1 h−1), indicating the great application potential of the TiO2/MoSx-Au photocatalyst. Besides, no significant decrease in the H2O2 concentration is observed after four cycles of photocatalytic reaction (Fig. 3d), revealing the robust reusability of the TiO2/MoSx-Au. From the above results, it can be concluded that the introduction of MoSx mediator into TiO2/Au can effectively improve the photocatalytic H2O2-production activity.

Fig. 3: Photocatalytic H2O2-production activities and stability.
figure 3

a Photocatalytic H2O2-production performance for different samples in an ethanol-water solution (10% vol.): (1) TiO2, (2) TiO2/MoSx, (3) TiO2/Au, and (4) TiO2/MoSx-Au. The error bars (mean ± standard deviation) were calculated based on three independent photocatalytic experiments. b The AQY of H2O2 production as a function of wavelength on TiO2/MoSx-Au (red dot), and the UV-vis absorbance spectrum (blue curve). c The photocatalytic H2O2-production performance over (1) TiO2 and (4) TiO2/MoSx-Au in the O2-saturated and air condition. d Recycling H2O2-production performance of TiO2/MoSx-Au.

Full size image

Photocatalytic mechanism of TiO2/MoSx-Au

Considering the boosted H2O2-production activity, the effect of the MoSx mediator on the Au electronic structure is primarily investigated by the first-principles calculations and XPS technology. For comparison, three slab models of Au, MoSx, and MoSx-Au are reasonably selected and optimized (Supplementary Fig. 8). Based on the optimized models, the work functions (Φ) of MoSx (001) and Au (111) are calculated to be 5.86 and 5.20 eV, respectively (Fig. 4a, b). In this case, when Au is loaded onto the MoSx surface, free electrons would migrate inevitably from Au nanoparticles to MoSx (Supplementary Fig. 9)45,46, thus inducing the formation of electron-deficient Auδ+ sites. This electron transfer is further substantiated by examining the local charge density difference and the corresponding planar-averaged electron density difference (Fig. 4c, d)47. Obviously, the MoSx-Au cocatalyst shows a distinct electron-enriched region on the MoSx side, while positive charges predominantly accumulate on Au atoms, leading to the production of an electron-deficient Auδ+ layer (Fig. 4d). To quantify the above charge transfer between Au and MoSx, Bader charge calculation was performed and shown in Fig. 4e. Clearly, they indicate a more negative charge density for S and Mo atoms in MoSx-Au compared to pure MoSx. Conversely, the charge density of Au atoms is slightly increased (+0.07) to produce Auδ+ active sites in the MoSx-Au cocatalyst. The formation of above electron-deficient Auδ+ can further be verified by XPS analysis (Fig. 4f). Compared with the TiO2/Au, a clear shift of binding energy (from 83.2 to 83.6 eV, ∆ = 0.4 eV) to a higher value is observed for Au 4f in the TiO2/MoSx-Au. In addition, the XPS peaks of S 2p (162.0 eV) and Mo 3d (231.6 eV) shift to lower values (∆= 0.6 eV for S 2p, ∆= 0.3 eV for Mo 3d) than those of TiO2/MoSx (Supplementary Fig. 10), indicating the efficient electron transfer from Au to MoSx. The above electron transfer can also cause a slight change of binding energy for the Ti element (Supplementary Fig. 11). Unquestionably, both DFT calculation and experimental results strongly support that the introduction of MoSx mediator can effectively modulate the Au electronic structure to induce the formation of electron-deficient Auδ+ sites in the MoSx-Au cocatalyst (Fig. 4g).

Fig. 4: MoSx-induced electron-deficient Auδ+ formation and mechanism.
figure 4

a, b Calculated average potential profiles of MoSx and Au. c The local charge density difference of MoSx-Au, where the light yellow and cyan areas represent electron accumulation and depletion, respectively. d Planar-averaged electron density difference ∆ρ(z) in MoSx-Au. e The charge density distributions of MoSx, MoSx-Au, and Au. f High-resolution XPS spectra of Au 4f in the TiO2/Au and TiO2/MoSx-Au. g Schematic illustration of the formation of electron-deficient Auδ+ sites in the MoSx-Au cocatalyst.

Full size image

The generation of electron-deficient Auδ+ impacts the O2 adsorption capability of TiO2/MoSx-Au photocatalyst, as evidenced by various analyses: O2 adsorption energy, crystal orbital Hamilton population (COHP), bonding distance analysis, and partial density of states (PDOS) calculations. Based on the optimized models in Fig. 5a and Supplementary Fig. 12, the O2 adsorption energies (Ea) of Au sites before and after MoSx introduction were first calculated (Fig. 5b). Clearly, the electron-deficient Auδ+ sites in the MoSx-Au cocatalyst exhibit more negative adsorption energy (Ea = −0.14 eV) than pure Au (Ea = 0.48 eV), indicating stronger O2 adsorption capability48. Further evidence of this enhanced adsorption is provided by COHP calculations49, revealing a smaller integrated COHP value (−1.40) for Au-Oads bonds in MoSx-Au than in pure Au (−1.38), and shorter bond length (2.01 Å) compared to pure Au (2.10 Å), signifying stronger Au-Oads bonding in electron-deficient Auδ+ sites (Fig. 5c)50. In this case, the d-band center mechanism helps explain this enhanced Au-Oads bonding in the MoSx-Au51. As depicted in Fig. 5d, g, the d-band center of Au 5d in pure Au cocatalyst is −2.23 eV, significantly far away from Ef. Consequently, the lower d-band center causes exorbitant antibonding-orbital occupancy, leading to a weak Au-Oads bond. However, when Au is loaded on the MoSx surface, the resulting d-band center of Au 5d in the MoSx-Au is closer to the Ef (−1.91 eV) than that of pure Au suggesting that modulating the electron structure of the Au cocatalyst to form electron-deficient Auδ+ sites can significantly elevate the d-band center (Fig. 5g). In this case, when O2 is adsorbed on the electron-deficient Auδ+ sites, the antibonding-orbital occupancy of Au-Oads is decreased, resulting in a reinforced O2 adsorption. Therefore, the presence of MoSx cocatalyst in the MoSx-Au can effectively raise the d-band center of Auδ+ sites for improved O2 adsorption, which is one of the essential steps for the following H2O2 production.

Fig. 5: Modifying the electron structure of Au cocatalyst to decrease the antibonding-orbital occupancy of Au-Oads for enhancing Au-Oads bonds.
figure 5

a Optimized configurations for O2 adsorption on Au and MoSx-Au, and b the corresponding adsorption energies and schematic illustration about the improved O2-adsorption ability. c COHP analyses of O2 adsorption on Au and MoSx-Au. d PDOS diagrams of Au 5d orbitals in Au and MoSx-Au. e COHP analyses of *OOH adsorption on Au and MoSx-Au (* represents active site). f ∆GHOO* over TiO2, Au and MoSx-Au. g Schematic diagram about the enhanced O2 adsorption on Au sites by decreasing antibonding-orbital occupancy via raising its d-band center.

Full size image

It is worth emphasizing that a suitable Au-Oads bond can effectively contribute to the formation of Au-OOHads intermediate, thus greatly improving the selectivity and activity of photocatalytic H2O2 production (Supplementary Fig. 13). Hence, to explore the effect of improved Au-Oads bonds on the formation of Au-OOHads intermediate in MoSx-Au cocatalyst, COHP calculation was carried out. As illustrated in Fig. 5e, both COHP value (−1.66) and bond length (2.08 Å, inset) of Au-OOHads in MoSx-Au are smaller than those in pure Au (−0.99 and 2.18 Å), forcefully manifesting that the MoSx-Au possess stronger Au-OOHads bonds, which is beneficial to improving the selectivity of photocatalytic H2O2 production. The above selectivity and activity of photocatalytic H2O2 production on MoSx-Au cocatalyst can further be confirmed by the free energy of *OOH intermediate (∆GHOO*) based on the optimized models in Supplementary Fig. 1422. As shown in Fig. 5f, the ∆GHOO* values on TiO2 and Au sites are estimated to be 5.01 and 4.5 eV, respectively, which are significantly higher than the ideal ∆GHOO* value (4.2 eV)52. These results suggest that TiO2 and Au have relatively weak adsorption and easier detachment for *OOH intermediates, leading to sluggish interfacial H2O2-production kinetics (Supplementary Fig. 13a). In contrast, the electron-deficient Auδ+ sites in the MoSx-Au cocatalyst showcase the best ∆GHOO* values (4.2 eV), aligning with the ideal energy for *OOH adsorption and facilitating rapid H2O2 generation (Supplementary Fig. 13b).

In addition to enhancing O2 adsorption for efficient H2O2 production, the MoSx-Au cocatalyst also promotes the rapid transfer of photogenerated electrons in the TiO2/MoSx-Au photocatalyst, which is supported by the subsequent Kelvin probe force microscopy (KPFM) and in situ XPS53,54. A scanning probe microscopy (SPM) system with KPFM was employed to analyze the distribution and transfer pathways of photogenerated electron-hole pairs in photocatalysts. The resulting AFM topography image, KPFM image, and the corresponding contact potential difference (CPD) profiles of TiO2 and TiO2/MoSx-Au are shown in Fig. 6a, b, and Supplementary Figs. 15 and 16. Obviously, the TiO2/MoSx-Au particles are observed, and a line scan across the above sample pre- and post-light irradiation is used to evaluate the CPD change. Upon light irradiation, the CPD value of TiO2 shows a slight increase from −14.7 to 8.6 mV owing to the spontaneous transfer of photogenerated holes onto the TiO2 surface (Supplementary Fig. 16)55. After the loading of MoSx-Au, the TiO2/MoSx-Au exhibits an obvious CPD value increase of about 185.4 mV (from −51.3 to 134.1 mV) during light irradiation, accompanied by a color change from blue to red owing to enhanced hole accumulation on the TiO2 surface (Fig. 6a, b), strongly indicating that the photogenerated electrons are efficiently transferred from TiO2 to MoSx-Au cocatalyst56. To further validate the above transfer of photogenerated electrons and their enrichment on the Au active sites of TiO2/MoSx-Au, in situ XPS was performed (Fig. 6c). The peaks of Au 4f7/2 and Au 4f5/2 in the TiO2/MoSx-Au are remarkably shift toward lower binding energies (from 83.6 eV to 83.5 eV, ∆ = 0.1 eV) upon light irradiation, suggesting that the photogenerated electrons are directionally transferred from TiO2 to MoSx-Au and mainly enriched on the electron-deficient Auδ+ sites, thereby promoting the photocatalytic H2O2-production kinetics57.

Fig. 6: Photogenerated electron transfer mechanism and dynamics.
figure 6

a, b KPFM images and the corresponding surface potential profiles of the TiO2/MoSx-Au in the dark and light illumination, a 365 nm-LED light as the light source. c ISI-XPS spectra of Au 4f for TiO2/MoSx-Au before and after light illumination. Pseudocolor plots of d TiO2 and e TiO2/MoSx-Au (GSB represents ground-state bleaching). f, g Femtosecond transient absorption spectra of TiO2 and TiO2/MoSx-Au within 20 ps. All data were obtained under excitation of 330 nm and optical power of 600 μW cm−2. Schematic illustration of the decay pathways of photogenerated electrons in h TiO2 and i TiO2/MoSx-Au.

Full size image

For a comprehensive understanding of electron-transfer dynamics in the TiO2/MoSx-Au, femtosecond transient absorption spectroscopy (fs-TAS) was carefully performed58. As shown in Fig. 6d, e and Supplementary Fig. 17a, a typical photobleaching peak (~380 nm) is displayed in the pseudocolor plots of TiO2, TiO2/Au, and TiO2/MoSx-Au. These signals are assigned to the ground-state bleaching (GSB), which reflects the excited state relaxation59. Further monitoring of the GSB signal (380 nm) within 20 ps reveals stronger intensity in the TiO2/Au (Supplementary Fig. 17b) and TiO2/MoSx-Au (Fig. 6g) compared to TiO2 (Fig. 6f), indicating enhanced electron enrichment in both TiO2/Au and TiO2/MoSx-Au39. Further analysis of the interfacial electron transfer involved fitting decay kinetics within 25 ps at 380 nm using biexponential equations, with fitting results of normalized curves shown in Supplementary Fig. 18 and Table S3. The short-lived τ1 corresponds to the electron trapping by electron trapping state (e-TS), while the long-lived τ2 is related to the interfacial electron transfer from TiO2 to cocatalyst. Meanwhile, A1 and A2 represent the decay proportion of photogenerated electrons during the electron trapping and transfer, respectively. Obviously, the TiO2 primarily undergoes a short-lived process within 25 ps under irradiation, and the corresponding τ1 is 1.63 ps, which is primarily attributed to electron trapping in the e-TS (Fig. 6h). Interestingly, the τ1 value in the TiO2/Au and TiO2/MoSx-Au significantly decreases to 0.36 and 0.92 ps, respectively, suggesting rapid transfer of a portion of photogenerated electrons from TiO2 to Au (τ2 = 5.88 ps) and MoSx-Au (τ2 = 7.10 ps) cocatalysts (Fig. 6i). Noticeably, the TiO2/MoSx-Au exhibits a larger A2 value (A2 = 39.6%) compared to the TiO2/Au (A2 = 32.3%), indicating more effective transfer of photogenerated electrons from TiO2 to Au facilitated by the MoSx mediator60. The improved electron transfer on TiO2/MoSx-Au is well consistent with the results of photoelectrochemical and transient-state photoluminescence (TRPL) (Supplementary Fig. 19). These above results provide concertedly evidence that the MoSx-Au cocatalyst serves as an efficient platform for rapid transfer of photogenerated electrons to engage in the subsequent H2O2-production reaction at the electron-deficient Auδ+ sites (Fig. 6i), thus achieving high photocatalytic H2O2 yields.

Overall, a strategy of electronic structure modification for the Au cocatalyst has been proposed to effectively reinforce the Au-Oads bonding at electron-deficient Auδ+ sites within the MoSx-Au cocatalyst, which can achieve an enhanced O2 adsorption for fast H2O2-production kinetics. As a result, an exceptional H2O2-production rate of 30.44 mmol g−1 h−1 has been achieved in the resulting TiO2/MoSx-Au, which is 25.4 and 1.3 times higher than that of TiO2 and TiO2/Au, respectively. Theoretical simulations and experimental results consistently support the notion that the introduction of MoSx mediator induces the formation of electron-deficient Auδ+ active sites in the MoSx-Au cocatalyst by free-electron transfer from the Au cocatalyst to MoSx, which decreases the antibonding-orbital occupancy of the Au-Oads, thereby enhancing the O2-adsorption ability to realize efficient H2O2-production performance. In addition, the MoSx-Au cocatalyst can also provide an efficient platform for the rapid transfer and enrichment of photogenerated electrons from TiO2, leading to a distinct improvement of photocatalytic H2O2-production activity for the TiO2/MoSx-Au. This work emphasizes a feasible strategy for optimizing the O2-adsorption strength to efficiently accelerate H2O2-production kinetics, offering a very promising approach for the rational design of electronic structure for efficient artificial photosynthesis.

Methods

Preparation of TiO2/MoSx photocatalyst

TiO2 photocatalyst (P25) was calcined at 550 °C for 2 h in a muffle furnace before being used. The TiO2/MoSx sample was synthesized by one-step lactic acid-induced method, as schematically demonstrated in Supplementary Fig. 1. First, 624 μL (NH4)2MoS4 (0.02 mol/L) solution was dropped into 160 mL lactic acid solution (10 vol%) under stirring. In this case, the H+ was released from lactic acid and would induce the transformation of MoS42- into MoSx colloidal nanoparticles. Subsequently, the as-prepared TiO2 nanoparticles (0.1 g) were dispersed into the above solution. After stirring for 2 h, a brown product was collected by centrifugation and washing with deionized water and ethanol. Finally, the obtained product was dried at 80 °C for 12 h. The resulting brown powder was denoted as TiO2/MoSx. In addition, the pure MoSx product was also obtained by a similar synthesis route of the above TiO2/MoSx in the absence of TiO2.

Preparation of TiO2/MoSx-Au photocatalyst

The MoSx-Au modified TiO2 photocatalyst (TiO2/MoSx-Au) was synthesized by a two-step route, including the initial deposition of MoSx on the TiO2 surface and the subsequently selective photodeposition of Au onto the MoSx surface. First, 0.1 g of the MoSx/TiO2 was mixed with 80 mL ethanol aqueous solution (20 vol%) in a 100 mL three-necked flask. Then, a known amount of chloroauric acid (0.1 mol/L, HAuCl4·4H2O) was added. After the above system was evacuated with N2 for 20 min and then irradiated with a 300 W Xenon lamp for 1 h, the resultant suspension was collected by centrifugated, rinsed, and dried at 80 °C for 12 h to obtain the final TiO2/MoSx-Au. To investigate the effect of Au amount on the structure and photocatalytic performance, the amount of Au in the TiO2/MoSx was controlled to be 1, 1.5, 2, 3, and 5 wt%, respectively, and the resultant sample was referred to be TiO2/MoSx-Au-X% (X is the amount of Au).

Preparation of TiO2/Au photocatalyst

Au nanoparticle-loaded TiO2 (TiO2/Au) was prepared by a photodeposition method. First, 0.1 g of the TiO2 was dispersed into a 100 mL of ethanol aqueous solution (20 vol%). Subsequently, a known amount of chloroauric acid (0.1 mol/L, HAuCl4·4H2O) was added dropwise to the above mixture solution. Before irradiation, the above system was bubbled with N2 for 20 min and then irradiated for 1 h. Finally, a light-purple product was collected by centrifugated, rinsed, and dried at 60 °C for 12 h. The resulting powder was denoted as TiO2/Au.

Characterization

The microstructure of the samples was characterized by transmission electron microscopy (TEM; Thermal Fisher Talos F200X). X-ray photoelectron spectroscopy (XPS) was performed on a Thermo Scientific ESCALA 210 XPS spectrometer system (USA) with 300 W Al Kα radiation to survey the elemental composition and valence states of these photocatalysts. The X-ray diffraction (XRD) patterns of the samples were obtained on an X-ray diffractometer (Shimadzu XRD-6100) with Cu Kα radiation. The elemental content was performed by inductively coupled plasma optical emission spectrometry (ICP-OES). The ultraviolet-visible spectra (UV-vis DRS) were obtained on a UV-vis spectrophotometer (UV-2600i, Shimadzu, Japan). Time-resolved photoluminescence (TRPL) spectra were acquired on a fluorescence lifetime spectrophotometer (FLS 1000, Edinburgh, UK). The photo-irradiated Kelvin probe force microscopy (KPFM) (SPM-9700, Shimadzu, Japan) was carried out to test the contact potential difference of the samples.

Photocatalytic H2O2 production test

Photocatalytic H2O2-production activity was examined in an O2-saturated aqueous solution with ethanol as a hole scavenger, and a 300 W Xenon arc lamp was selected as the light source. First, 10 mg of as-prepared photocatalyst and 100 mL of ethanol solution (10 vol%) were mixed in a 100 mL three-necked flask reactor. Before irradiation, the system was purged with oxygen for 30 min to obtain an O2-saturated solution. During the photocatalytic H2O2-production test, 1 mL of solution was sampled from the reactor. Finally, the H2O2 concentration was examined via an iodometry method by using a UV-visible spectrophotometer (UV-1240, Japan). The concentration of H2O2 was calculated by the equation (\(y{{\mbox{=}}}0.00771x{{\mbox{+}}}0.0218\)), and the reaction mechanism was shown in Eq. (1). The absorbance of I3- at 350 nm can be recorded by UV-vis spectroscopy.

$${{{\mbox{H}}}}2{{{\mbox{O}}}}_{2}+3{{{{{{\rm{I}}}}}}}^{-}+2{{{\mbox{H}}}}^{+}\to {{{{{{\rm{I}}}}}}}_{3}^{-}+2{{{\mbox{H}}}}_{2}{{\mbox{O}}}$$
(1)

Photoelectrochemical measurements

The photoelectrochemical (PEC) properties were assessed using an electrochemical workstation (CHI 760E, China) within a three-electrode system. The working electrode was prepared by applying the photocatalyst onto a 1.0 cm2 FTO glass substrate. The Ag/AgCl (in a saturated KCl solution) and Pt foil were designated as the reference and counter electrodes, respectively. The PEC assessments were carried out in a 0.5 M Na2SO4 solution using a 300 W Xenon arc lamp for illumination.

Average decay time (τ average) calculation

The decay curves of as-prepared samples from the TRPL can be effectively fitted using the following biexponential Eq. (2), and the fluorescent lifetime (\({\tau }_{{{\mbox{a}}}}\)) is calculated by Eq. (3).

$${A}_{(t)}={A}_{(0)}+{A}_{1}\exp (-t/{\tau }_{1})+{A}_{2}\exp (-t/{\tau }_{2})$$
(2)
$${\tau }_{{{{{{\rm{a}}}}}}}=({A}_{1}{{\tau }_{1}}^{2}+{A}_{2}{{\tau }_{2}}^{2})/({A}_{1}{\tau }_{1}+{A}_{2}{\tau }_{2})$$
(3)

where \({A}_{1}\) and \({A}_{2}\) represent the weight factors, while \({\tau }_{1}\) and \({\tau }_{2}\) are the short and long fluorescent lifetimes, respectively.

Apparent quantum yield (AQY) calculation

The AQY measurement was conducted in an O2-saturated aqueous solution with ethanol as a hole scavenger by utilizing a 300 W Xenon arc lamp as the light source. In detail, the as-prepared photocatalyst (10 mg) and ethanol solution (100 mL, 10 vol%) were mixed in a 100 mL three-necked flask reactor, which was oxygenated for 30 min to obtain the O2-saturated solution.

The apparent quantum yields for H2O2 were calculated from the following Eq. (4):

$$\eta=\frac{{N}_{e}}{{N}_{p}}\times 100\%=\frac{2\times M\times {N}_{{{{{{\rm{A}}}}}}}\times h\times c}{S\times P\times t\times \lambda }\times 100\%$$
(4)

where M represents the amount of produced H2O2 molecules (mol), NA is the Avogadro constant (6.022 × 1023/mol), h is the Planck constant (6.626 × 10 −34 J s), c is the speed of light (3 × 108 m/s), S is the irradiation area (15.83 cm2), P is the average intensity of irradiation (365 nm, 13.35 mW/cm2), t is the irradiation time (s), and λ is the wavelength of the incident monochromatic light (nm).

Ultrafast transient absorption (TA) tests

Femtosecond transient absorption spectra of the as-prepared photocatalysts were obtained on a pump-probe system (Helios, Ultrafast System) with a maximum time delay of ~8 ns using a motorized optical delay line under ambient conditions. The 330 nm-pump pulses (600 μW average at tested samples) were generated by the 1 kHz regenerative amplifier (Coherent Libra, 800 nm, 35 fs, 5 mJ) in an optical parametric amplifier (OPerA Solo), seeded with a mode-locked Ti: sapphire oscillator (Coherent Vitara, 800 nm, 80 MHz) and pumped with an LBO laser (Coherent Evolution-50C, 1 kHz system). To generate the white light from 320 to 750 nm, the 800 nm-femtosecond pluses were pumped by a constantly rotating sapphire crystal.

Computational details

The density functional theory (DFT) calculations were performed using the Vienna Ab initio Simulation Package (VASP) with the generalized gradient approximation (GGA) employing the revised Perdew–Burke–Ernzerhof (PBE) functional for the exchange-correlation interaction. The convergence threshold for total energy converged within 10−4 eV/atom, and the average force was 0.01 eV/Å. Grid integration utilized a cutoff energy of 450 eV, and projector-augmented wave (PAW) potentials characterized the ion cores. To rationalize the calculation, unsaturated sulfur (S) atoms were obtained on the edge of MoSx model by creating a vacuum layer in the y-axis direction, which displayed a similar local coordination structure as amorphous MoSx cocatalyst. Moreover, (3 × 3 × 1) Monkhorst-Pack grids and (2 × 2) surface cells were used for oxygen adsorption. The adsorption energy Eads was defined as \({E}_{{{{{{\rm{ads}}}}}}}={E}_{{{{{{\rm{total}}}}}}}-{E}_{{{{{{\rm{surface}}}}}}}-{E}_{{{{{{\rm{O}}}}}}2}\), where Etotal, Esurface, and EO2 represent the energy of adsorption configurations, the energy of metallic surfaces, and the energy of molecular O2, respectively. In addition, the ΔG of HOO∗ intermediate on the surface was calculated by the equation \(G=E+{ZPE}-{TS}\), where E is the total energy, ZPE is the zero-point energy, T is the temperature (298.15 K), and S is the entropy. Several configurations of the adsorbed models were considered in the simulation, and the most favorable ones are presented based on the adsorption energy.