Hydrogen (H₂), a source of clean and high-heat-density secondary energy, is one of the promising energy carriers that can satisfy the ever-increasing energy and environmental demands [1-3]. Photocatalytic water splitting for H₂ production has received great attention owing to its simplicity and recyclability, and involves the conversion of solar energy to chemical energy with the help of a photocatalyst, sunlight, and water at room temperature [4-6]. Since the pioneering work of Fujishima and Honda on TiO₂ electrodes for photoelectrochemical water splitting, various semiconductors, including TiO₂ [7], CdS [8, 9], WO₃ [10], C₃N₄ [11, 12], and ZnO [13, 14], have been explored to develop highly active, robust, and low-cost photocatalysts and photoelectrodes for industrial application. By virtue of its high stability, environmental friendliness, and low cost [15], TiO₂ has been extensively studied and plays a vital role in H₂ production, pollutant degradation [16, 17], and CO₂ reduction [18, 19]. However, pure TiO₂ photocatalyst exhibits unsatisfactory photocatalytic activity owing to insufficient light harvesting (no more than 5%) and a high photogenerated electron–hole recombination rate [2, 20], which severely restrict its industrial application. To date, a myriad of strategies have been proposed for enhancing the photocatalytic activity, such as loading noble metals [21], doping metallic or non-metallic elements [22], and constructing heterojunctions [23-25].

Heterojunction construction by integrating two semiconductors is a universal strategy owing to its effectiveness in spatially separating the photoexcited electron–hole pairs via appropriate band alignments between the semiconductors [1, 26]. Recently, a new step-scheme heterojunction (S-scheme heterojunction) concept was proposed to explain the increased charge transfer rate observed at the interface of two photocatalysts with staggered band structures [27]. Typically, a S-heterojunction photocatalyst is formed between a reduction photocatalyst and an oxidation photocatalyst. In general, reduction photocatalysts display comparatively higher conduction bands (CBs) and Fermi levels. On the contrary, oxidation photocatalysts exhibit comparatively lower CBs and Fermi levels. Intriguingly, WO₃, with a relatively narrow bandgap, between 2.4 and 2.8 eV [28], is a representative oxidation-type photocatalyst that is amenable to the fabrication of S-heterojunction composites. Owing to its large work function (6.23 eV) and low Fermi level [27], the electrons in the CB of WO₃ can scarcely exhibit a reduction capacity, based on thermodynamics [29], and are therefore relatively useless for photocatalytic H₂-evolution reactions. By taking advantage of the S-scheme charge transfer mechanism, the unnecessary electrons in the CB of WO₃ can be recombined with the comparatively useless holes in the valence band (VB) of TiO₂. Consequently, the separation of useful photoinduced charge carriers at the heterogeneous interface is promoted, and the high redox capacities of the electrons in the CB of TiO₂ and the holes in the VB of WO₃ are retained. Additionally, WO₃ demonstrates strong light absorption and great resistance to photocorrosion [30]. As reported, WO₃ can extend its light absorption range to the visible or even near infrared (NIR) region when coupled with TiO₂ to construct a TiO₂/WO₃ S-scheme heterojunction [31, 32].

Reduced graphene oxide (rGO) with a two-dimensional (2D) honeycomb lamella of C atoms is a prospective candidate for the electron container and acceptor in heterogeneous photocatalytic systems that can facilitate the shuttling of photoexcited electrons for water-splitting reaction [33]. In addition to its excellent thermal conductivity and large carrier mobility, rGO exhibits a high theoretical specific surface area and thereby offers abundant adsorption and catalytic sites for photocatalytic reactions [34]. The presence of rGO can also extend the range of light absorption to the visible region and even to the infrared region, which may induce a positive photothermal effect and increase the photocatalytic H₂ yield. Recently, ternary composite photocatalysts have emerged as alluring candidates for photocatalytic H₂ evolution because of their cooperative effect on the successive charge transfer occurring between three different components, such as ZnO-MoS₂-rGO [35], TiO₂-MnO_x-Pt [36], and g-C₃N₄/RP/MoS₂ (RP denotes red phosphorus) [37]. Therefore, it is desirable to fabricate a rGO-based composite photocatalyst to enhance the photocatalytic activity and stability [38].

Herein, we reported a facile one-step hydrothermal method to synthesize WO₃/TiO₂ S-heterojunction photocatalyst by graphene modification (WTG). TiO₂ and WO₃ formed a S-scheme heterostructure with a strong interaction at the interface, which facilitated charge transfer and separation. The S-scheme mechanism was evidenced by the greatly improved H₂ evolution performance. Additionally, rGO acted as a supporting matrix and an electron transfer channel, supplied surface active sites, enhanced the light absorption, and exhibited a unique photothermal effect in WTG composite. The cooperative effect of rGO and WO₃/TiO₂ heterojunction promoted the photocatalytic reaction. Consequently, WTG composites exhibited higher photocatalytic H₂ generation activities than pure TiO₂.

The WO₃/TiO₂/rGO composite was fabricated by a one-step hydrothermal approach. All the reagents employed were analytical grade and used without further purification. GO was synthesized from commercial graphite powder by using modified Hummer's method [39]. Typically, for the preparation of the WTG sample, a GO suspension (3 mL, 1 mg mL^–1) was diluted in a mixed solution of distilled water (35 mL) and ethanol (17.5 mL) and sonicated for 15 min. Then, D-(+)-glucose (20 mg) was dissolved in the GO suspension and stirred for 10 min. Subsequently, 1.3 mL of Ti(OC₄H₉)₄ (TBOT) as the Ti source, which was dissolved in ethanol (9 mL), was added dropwise to the mixture under magnetic stirring. After 30 min, 0.13 mmol of Na₂WO₄·2H₂O was added to the solution, and its pH was adjusted to 2 with hydrochloric acid (1 M). An hour later, the milky mixture was transferred to a Teflon-lined stainless-steel autoclave and kept at 180 ℃ for 12 h. The resultant dark grey precipitate was washed with distilled water and ethanol thoroughly and freeze-dried overnight to yield WO₃/TiO₂/rGO composite. In the composite, the theoretical mass ratio of rGO to TiO₂ was 1%, and the theoretical mass ratio of WO₃ to TiO₂ was 10%. The sample was designated as WTG.

Similarly, WO₃/TiO₂ was synthesized by following the same procedure as detailed above, except for the addition of GO. Likewise, TiO₂/rGO was prepared by removing the step involving Na₂WO₄·2H₂O and pH adjustment. Pristine WO₃ was produced from the Na₂WO₄·2H₂O solution of pH = 2. To prepare pure TiO₂, the TBOT ethanol solution was added to a mixture of distilled water and ethanol. All the samples were subjected to the aforementioned hydrothermal process. The as-prepared WO₃/TiO₂, TiO₂/rGO, and pristine WO₃ and TiO₂ were designated as WT, TG, W, and T, respectively. Inductively coupled plasma (ICP)-optical emission spectrometry was conducted to investigate the actual contents of Ti and W elements. The results are listed in Table 1.

Powder X-ray diffraction (XRD) was performed by using a D/MAX-RB diffractometer (Rigaku, Japan) with Cu K_α radiation. To characterize the morphological structures and microscopic details of the samples, field-emission scanning electron microscopy (FESEM) images were obtained by using a JSM-7500F scanning electron microscope (SEM; JEOL, Japan). Transmission electron microscopy (TEM) was carried out by using a JEM-2100F instrument (JEOL, Japan) with an accelerating voltage of 200 kV. The pore size distribution and specific surface area were obtained from nitrogen (N₂) adsorption-desorption isotherms on a Micromeritics ASAP 2020 instrument. The specific surface area results were determined by using the adsorption data in the relative pressure (P/P₀) range 0.05–0.3. Raman spectra were recorded with the help of a micro-Raman spectrometer (Renishaw inVia, U.K.). X-ray photoelectron spectroscopy (XPS) measurements were performed by using an ESCALAB 250 Xi electron spectrometer (Thermo Scientific Corporation, USA) to estimate the surface elemental composition and chemical states. To investigate the optical properties of the samples, the UV-vis diffuse reflectance spectra (DRS) were obtained by using a Shimadzu UV 2600 UV-vis spectrometer with BaSO₄ as the reference. The elemental contents were analyzed on an ICP-optical emission spectrometer (Prodigy 7, Leeman Labs Inc.).

The photocatalytic activities of the samples were evaluated based on the photocatalytic yields of H₂ from water splitting in a 100 mL Pyrex flask. First, 50 mg of a sample was suspended by ultrasonication in 80 mL of 20 vol% aqueous methanol solution. Then, the sealed system was bubbled with N₂ for 0.5 h to ensure anaerobic conditions during photocatalytic water splitting. A 350 W Xe arc lamp was used as the light source. After an hour of irradiation, 0.4 mL of gas was extracted from the flask by using a syringe to analyze the H₂ concentration by gas chromatography (GC-14C, Shimadzu, Japan) with the assistance of a thermal conductivity detector. Photocatalytic H₂ production cycled tests were conducted by continuously sampling gas from the flask at an interval of an hour and rebubbling the solution with N₂ for 0.5 h every 3 h. The apparent quantum efficiency (AQE) was measured under the irradiation of four low-power light-emitting diode (LED) lamps (λ = 365 nm). The AQE can be computed by the following equation.

(1)

Photoelectrochemical measurements were conducted on a typical three-electrode system by using a CHI-660C electrochemical workstation (Chenhua Instrument, Shanghai, China). A Pt wire and Ag/AgCl electrode were employed as the counter and reference electrodes, respectively. The working electrode was the photocatalyst film, which was deposited on the conductive surface of a F-doped tin oxide (FTO) glass with an active area of ca. 1 cm². The preparation process was as follows: 20 mg of the photocatalyst was added to 1 mL of ethanol and ground to a sticky slurry for approximately 0.5 h. Afterwards, the homogeneous paste was evenly applied on the conductive surface of the FTO glass by doctor blading technique and then dried in air. A LED (λ = 365 nm) with an intensity of ca. 44.0 mW cm^–2 acted as the light source. A volume of 20 mL of 0.5 M aqueous Na₂SO₄ solution was used as the electrolyte. The initial bias potential in the transient photocurrent responses and electrochemical impedance spectroscopy (EIS) tests was the open-circuit voltage.

A facile one-step hydrothermal method was utilized to fabricate the samples, and the as-obtained samples, including pristine WO₃ and TiO₂, WO₃/TiO₂, TiO₂/rGO, and WO₃/TiO₂/rGO, were designated as W, T, WT, TG, and WTG, respectively. The representative SEM images of WTG are shown in Fig. 1. The pristine TiO₂ and WO₃ tended to self‑aggregate with the irregular bulk. However, for the WTG composites, the TiO₂ and WO₃ particles were homogeneously mixed, thus forming a WO₃/TiO₂ heterostructure via strong interactions between TiO₂ and WO₃. The particles were closely and evenly grown in-situ on the anchoring sites of GO, which implied that the graphene sheets prevented the TiO₂ and WO₃ nanoparticles from assembling as an underlying substrate in WTG. This intimate interaction is beneficial for electron transfer from the TiO₂ or WO₃ particles to the graphene sheet during photoexcitation. After the hydrothermal treatment, the rGO retained the morphology of micron-sized 2D sheets with spontaneous wrinkles. The TiO₂ and WO₃ nanoparticles adhering to the oxidized graphene prevented the naked sheets from restacking, thus preserving the active surfaces. In addition, in the EDS elemental mapping images of WTG (Fig. 1(c)), the distributions of C, O, Ti, and W elements all matched well with the outline of the FESEM image, which suggested that the TiO₂ and WO₃ particles were evenly decentralized on the surface of rGO.

For a better illustration of the microstructure, the WO₃/TiO₂/rGO composite was characterized by TEM. Fig. 2(a) clearly shows that highly dispersed nanoparticles of TiO₂ and WO₃ were deposited on the surface of the transparent and wrinkled rGO nanosheets. The particle sizes were ca. 10 nm on average. Moreover, the WO₃ particles were tightly connected with the TiO₂ particles even after ultrasonication, which suggested the formation of a heterostructure interface instead of a simple physical mixture. In the HRTEM image (Fig. 2(b)), the lattice fringe spacings of 0.35 and 0.19 nm corresponded to the (101) and (200) planes of the TiO₂ nanoparticles [40], whereas the lattice spacing of 0.27 nm corresponded to the (022) planes of monoclinic WO₃, which suggested that the WO₃ existed in the monoclinic form [10].

XRD (Fig. 3) was employed to investigate the phase structures of the samples. For the T sample, the characteristic diffraction peaks at 25.1°, 38.6°, 46.4°, and 55.1° well matched with those of anatase (JCPDS # 21-1272) [41]. The diffraction peaks of W indicated that WO₃ crystallized in the monoclinic phase (JCPDS # 71-2141) [30]. For TG, WT, and WTG, the observed characteristic peaks could be attributed to anatase. No evidence of a tungsten oxide phase was found in the XRD patterns of WTG and WT. Besides, no peak shift was observed in the composites, compared with the case of pristine TiO₂, which suggested that the crystal structure of anatase was maintained. Similar broad diffraction peaks of the anatase phase were observed, which indicated small crystallite sizes. Monoclinic WO₃ was not observed in the WTG and WT samples, because the amount of WO₃ was too low to be revealed by XRD. The amounts of W element in WTG and WT were 6.52 and 6.58 wt%, respectively, which were obtained by ICP analysis. According to a previous report, WO₃ is present in the form of highly dispersed nanocrystallites or clusters when its content is very low. At least 8 wt% of WO₃ is necessary to cover the TiO₂ surface with a monolayer, the thickness of which can be determined by XRD [42-44]. Moreover, after the hydrothermal treatment, no characteristic diffraction peak of GO or graphene was found in the patterns of TG and WTG, which was ascribed to the low concentration and inherently low diffraction intensity of rGO relative to those of metallic oxides [45].

Raman spectra were recorded to verify the presence of C in the composites and reveal the significant structural change from GO to rGO after the hydrothermal reaction. In Fig. 4(a), the Raman spectrum of GO showed two typical peaks at approximately 1351 (D band) and 1586 cm^–1 (G band), which were attributed to the breathing mode of the k-point phonons of A_1g symmetry and first-order scattering of the E_2g vibration mode of sp²-bonded C atoms [46], respectively. The D band moved to 1344 cm^–1 for WTG, while the G band moved to 1600 cm^–1 for both TG and WTG (Fig. 4(b)), which indicated the reduction of GO and interaction between rGO and particles [47]. In comparison with that of pure GO, the D/G intensity ratios (I_D/I_G) of WTG and TG apparently increased. The increase in I_D/I_G indicated that the size of the sp² domains within the plane decreased in terms of the average value, which further testified the transformation from GO to rGO [48, 49]. For the T sample shown in Fig. 4(a), the several typical bands observed at 399, 518, and 640 cm^–1 were ascribed to the B_1g(1), A_1g + B_1g(2), and E_g(2) modes of anatase, respectively. For W, the major Raman bands at 807 and 717 cm^–1 were ascribed to W–O stretching modes, and the other two bands at 273 and 327 cm^–1 corresponded to the W–O bending modes of bridging O, which are characteristic of monoclinic WO₃ [50]. Bands of anatase were observed in the spectra of the corresponding composites, whereas no traces of monoclinic WO₃ were found in the WT and WTG samples. The results reflected the fact that the signal of WO₃ is too weak relative to that of TiO₂. Strikingly, in the TiO₂ composites, no shift in the main Raman scattering bands was found, compared with the case of pure TiO₂, which was sensitive to the insertion of ions into the TiO₂ unit cell, suggesting that ion diffusion had not occurred in the composites [51].

XPS was employed to investigate the surface chemical states of the WTG composite photocatalysts. Fig. 5(a) shows that the XPS C 1s pattern of the WO₃/TiO₂/rGO composite can be deconvoluted into four peaks at 284.8, 285.3, 286.6, and 289.1 eV, which can be assigned to the surface adventitious C (C–C), epoxide and hydroxyl C (C–O), and carbonyl (C=O) and carboxyl (O–C=O) groups, respectively [52]. The results testified to the existence of rGO. The O 1s XPS pattern (Fig. 5(b)) displayed three peaks at 530.3, 531.9, and 533.2 eV, corresponding to lattice O (Ti–O, W–O), hydroxyl O (–OH), and the O of the C–O or C=O bonds [53], respectively. Fig. 5(c) shows the high-resolution Ti 2p spectra of T, WTG, and WTG obtained under light irradiation that reveals two symmetrical peaks corresponding to Ti 2p_3/2 and Ti 2p_1/2 of TiO₂, respectively [54]. Notably, the binding energy of Ti 2p for WTG showed a positive shift, compared with that for pristine TiO₂, suggesting that electrons transferred from TiO₂ to rGO or WO₃ at the interfaces.

In the W 4f spectrum of W, shown in Fig. 5(d), the main peaks at approximately 37.7 and 35.6 eV are observed, which result from W 4f_5/2 and W 4f_7/2 of W⁶⁺, respectively [55]. Two shoulder peaks assigned to W⁵⁺ and centered at approximately 37.5 and 35.4 eV are also noticed [56]. The appearance of W⁵⁺ implied the existence of O vacancies, which was in accordance with the UV-vis DRS. From the peak areas, we could roughly estimate that the ratio of W⁵⁺ to W⁶⁺ was 0.279:1. For the WTG sample, a well-resolved doublet at 36.6 (W 4f_7/2) and 38.7 eV (W 4f_5/2) indicated the existence of the W⁶⁺ species of stoichiometric WO₃ [57], which overlapped with Ti 3p at 37.6 eV [58]. Therefore, both TiO₂ and WO₃ were successfully anchored on the rGO nanosheets. As reported [56], the W 4f peaks of the composite shifted to higher binding energies, which confirmed the massive decrease in the number of unsaturated W ions.

N₂ adsorption-desorption measurements were conducted to characterize the specific surface areas and pore size distributions of the samples. In Fig. 6, the representative N₂ adsorption-desorption isotherms of T and WT (Fig. 6(a)), and TG and WTG (Fig. 6(b)), correspond to type IV, which indicate the existence of large mesopores and macropores. According to the International Union of Pure and Applied Chemistry classification, the isotherms of the T and WT samples exhibited an obvious H2‑type hysteresis with ink-bottle mesopores that are derived from the aggregation of the TiO₂ and WO₃ nanoparticles. For the TG and WTG samples shown in Fig. 6(b), the isotherms exhibited a combination of H2 and H3 hysteresis loops, which were associated with the pores caused by the aggregated nanoparticles and the slit-shaped mesopores due to the stacking of the rGO sheets [59], respectively. Accordingly, the Barret-Joyner-Halenda pore size distributions (shown in the insets of Fig. 6(a) and (b)) displayed a relatively wide range, 2–100 nm, with a distinct peak at ca. 10 nm, which coincided with the results of TEM analysis. In addition, the Brunauer-Emmett-Teller (BET) specific surface areas of WTG and TG (Table 2) were 165 and 160 m² g^–1, respectively, which were larger than those of the other two samples, indicating that the presence of graphene tended to boost the specific surface areas and provided abundant active sites for adsorption and surface reactions.

UV-vis DRS were utilized to determine the optical absorption properties of the samples (Fig. 7). Obviously, pure TiO₂ exhibited an intrinsic absorbance edge at 388 nm, which was ascribed to its bandgap of 3.20 eV. For the W sample, WO₃ revealed a large light absorption edge at 466 nm, corresponding to its inherently narrow bandgap of 2.66 eV. Interestingly, the absorption curve of WO₃ revealed a slight increase from 500 to 900 nm, which was a combined result of the excitation via the d → d internal transitions of the W ions [60], polaron hopping [61], and the absorption induced by localized surface plasmon resonance [62]. In contrast with the commercial yellow-colored WO₃, the blue-green color of the tungsten oxide obtained in this work (inset of Fig. 7) proved that it existed partially in its sub-stoichiometric phases, which was associated with the defect levels introduced by O vacancies below the CB. The presence of O vacancies was beneficial for the photocatalytic reactions [63]. The WO₃/TiO₂ composite displayed increased absorption in the wavelength range corresponding to the UV, visible, and NIR regions, which was attributed to the intrinsic optical absorption of TiO₂ and WO₃ with O vacancies. With the introduction of graphene, the TG and WTG samples extended their broad background absorption range to the visible-light region, because graphene not only decreased the optical reflection but also was associated with the electronic transitions n → π* between the n-orbits of the O species and graphene and the π → π* of graphene [64]. Moreover, graphene can convert light to heat under optical irradiation, which accelerates the reaction. The photothermal conversion abilities of the samples were measured by using a thermos-imager. Fig. 8 exhibits the variations in the surface temperatures of the T, W, GO, and WTG samples. In fact, the surface temperatures of T and W changed slightly under LED illumination (365 nm, 44.0 mW cm^–2) for 20 s. Under the same conditions, the average temperatures of the GO and WTG surfaces increased by 9.1 and 4.5 ℃, respectively. These results revealed the photothermal conversion characteristic of GO.

Electron spin resonance (ESR) spectroscopy was employed to further monitor the presence and concentrations of O vacancies, as it is a pivotal tool that is used to examine unpaired spins in magnetism. Fig. 9 shows that the spectrum of WO₃ contains an asymmetric resonance signal (signal A), identified as that of W⁵⁺ [65], and another signal (signal B) at the g-value of 1.9036, which was typical of O (paramagnetic O^–) defects [66]. The pristine TiO₂ exhibited a weak, slightly anisotropic paramagnetic response that was centered at approximately g = 2.0027; it resulted from the localized Ti³⁺ 3d¹ states, which was consistent with the results of a previous study [67]. Notably, the spectrum of WT showed superparamagnetic behavior at g = 2.0034. The slight increase in g indicated interaction at the heterogeneous surface between TiO₂ and WO₃, which influenced the magnetic surroundings of the unpaired spins. Furthermore, the signal intensity was significantly enhanced, which reflected the increase in the concentration of O vacancies; this can be observed in the orange line of WT in the figure. The presence of O vacancies, which serve as shallow donors, is favorable for enhancing the electron density in n-type semiconductors and thus increasing the electrical conductivity. Moreover, they can improve the adsorption of surface species, making these composites attractive as photocatalytic materials [68].

The performances of the photocatalysts in H₂ generation via water splitting were evaluated under irradiation with a Xe arc lamp. In the preliminary control experiments, we did not detect appreciable H₂ evolution in the absence of irradiation or photocatalysts, which suggested that H₂ was generated by photocatalysts through a series of reactions under irradiation. Fig. 10(a) shows that the W, G, and WG samples exhibit no H₂ generation activity, while pure TiO₂ displays a relatively low photocatalytic activity, attributable to the large recombination rate of the electrons in the CB and the holes in the VB. After combination with 1 wt% graphene, the H₂ evolution activity of the TG sample increased, and was approximately 2.6-fold that of pure TiO₂, which was ascribed to the formation of a Schottky junction between TiO₂ and rGO. The electrons in the TiO₂ CB tended to migrate to rGO, with a lower Fermi level and greater carrier mobility, which inhibited the recombination of the charge carriers. Furthermore, the presence of rGO as an ideal support for photocatalytic reactions offered numerous adsorption and catalytic sites, which extended the light absorption range to the visible region and even to the IR region and induced a positive photothermal effect.

For the WT sample, the H₂ yield rate was 105.2 μmol g^–1 h^–1, which was approximately 1.5 times that of pure TiO₂. Hence, the improved photocatalytic activity can not only be attributed to the improved light harvesting, because WT displayed limited UV-visible light absorption. This finding might be attributed to the formation of the WO₃/TiO₂ S-scheme heterostructure, which expedited both the separation of useful electrons and holes and the recombination of relatively useless electrons and holes. Over the WTG composite, the H₂ evolution rate was greatly enhanced to 245.8 μmol g^–1 h^–1, which was approximately 2.3 times that over WT. Further, the AQE of the WTG composite was measured as 1.4% at the wavelength of 365 nm. The superior activity of WTG implied that rGO and the S-scheme heterojunction between WO₃ and TiO₂ had a positive synergistic effect that maximized the separation and transfer efficiency of the useful photogenerated electrons and holes. Fig. 10(c) shows the photocatalytic stability of the WTG sample, which was investigated over four consecutive cycles under the same conditions. Obviously, no remarkable deactivation was detected after four runs, indicating that the WO₃/TiO₂/rGO composite displayed good stability.

Photoelectrochemical measurements were employed to investigate the migration of the photoinduced charge carriers. Fig. 11(a) and (b) shows that all five samples exhibit relatively settled photocurrent curves with several on-off runs under intermittent ultraviolet (365 nm) irradiation. The WTG sample showed remarkably increased photocurrent density, compared with those of the other four monocomponent and dual-component photocatalysts, which was consistent with the photocatalytic performances. This increase in the photocurrent indicated an enhancement in the photogenerated charge carrier separation efficiency due to the TiO₂/WO₃ S-scheme heterostructure in the presence of the highly conductive rGO for electron transfer.

In addition, the photocurrent responses of the five samples revealed different characteristics. When light was turned off, the photocurrents of the T, W, and TG samples sharply dropped to zero. Afterwards, the photocurrent intensity of the W sample gradually increased even after the removal of irradiation, whereas those of the T and TG samples remained zero. By considering the change in the color of WO₃ to blue from yellow upon irradiation, we can deduce that electrons were released during reoxidation of W⁶⁺/W⁵⁺, which was reported in previous work [69, 70]. By contrast, the photocurrents of the WT and WTG samples slowly decreased to zero upon turning off the light, due to the reversible oxidation of W⁵⁺ to W⁶⁺.

The charge transfer kinetics was further investigated by using EIS. Fig. 11(c) and (d) shows that the WTG sample displayed the smallest semicircle relative to those of the other samples, which suggested that the WO₃/TiO₂/rGO composite exhibited the lowest interfacial electron-transfer resistance, which was beneficial for improving the separation and migration rates of the photoinduced charge carriers.

The Mott-Schottky plots were analyzed to ascertain the band structures of the T and W samples. In Fig. 12, the Mott-Schottky plots of TiO₂ and WO₃ exhibited positive slopes at 1000 Hz, which was typically characteristic of an n-type semiconductor. From the x-intercepts, the extrapolated flat-band potentials of TiO₂ and WO₃ were obtained as –0.84 and –0.50 V versus the saturated Ag/AgCl reference electrode (pH = 7), respectively.

With reference to the Mott-Schottky plots, the flat-band potentials of TiO₂ and WO₃ can be calculated according to the following conversion formula:

(2)

where E^θ (Ag/AgCl) = 0.197 V, and the flat-band potentials of TiO₂ and WO₃ were –0.23 and 0.11 V (vs. RHE, pH = 0), respectively. It was speculated that the bandgaps of TiO₂ and WO₃ were approximately 3.2 and 2.6 eV, based on the UV-vis DRS. Both the CB and VB energy levels of TiO₂ were higher than those of WO₃. The redox potential of graphene/graphene^•- was –0.08 V [71].

Thermodynamically, the CB electrons of WO₃ and the internal electrons of rGO cannot participate in the photocatalytic H₂-production reactions. Interestingly, the WO₃/TiO₂/rGO composite exhibited 3.5-fold H₂ evolution activity compared to that of pure TiO₂. If the charge carriers of the WTG sample are transferred in accordance with the traditional type-II model, the electrons in the CB of TiO₂ would migrate to the CB of WO₃ with decreased reduction ability, while the holes in the VB of WO₃ would migrate to the VB of TiO₂ with decreased oxidization ability, under illumination. However, transfer of the electrons accumulated in the CB of WO₃ to rGO would be thermodynamically unfavorable and retard the continual charge transfer from the CB of TiO₂. In this case, the WTG composite would show almost no photocatalytic H₂-production activity, which is distinctly opposite to the result obtained experimentally.

The TEM and XPS analyses showed that the S-scheme heterojunction was formed at the interface between TiO₂ and WO₃ in the WTG and WT samples. The work function of WO₃ is greater than that of TiO₂, and the difference drives the charge transfer from TiO₂ to WO₃ upon contact until the Fermi levels equilibrate [72]. Consequently, a built-in electric field can be formed at the heterogeneous interface that is directed from TiO₂ to WO₃ that favors the transfer and separation of the photogenerated charge carriers. In light of the S-scheme mechanism (Fig. 13), under optical irradiation, the relatively useless photoinduced electrons in the CB of WO₃ will migrate to the VB of TiO₂ via the intimate interface and recombine with the relatively useless holes under the driving force of the built-in electronic field. This process expedites the separation of the relatively useful electrons in the CB of TiO₂ and the holes in the VB of WO₃ and maintains their high reduction and oxidation abilities, respectively. Furthermore, the electrons collected in the CB of TiO₂ are prone to migration to the surface of rGO via the Schottky junction owing to their 2D π-conjugation structure and relatively lower potential than the CB of TiO₂, which facilitates free flow along the conductive network. The process further facilitates the migration of electrons, thus avoiding the accumulated electrons in the CB of TiO₂ from retarding the continual photocatalytic excitation of electron from the VB of TiO₂ in the S-scheme heterojunction system. As for the in situ XPS pattern of Ti 2p of the WTG sample, the binding energy of Ti 2p under light irradiation exhibited a slight shift relative to that measured in darkness, which was a result of the combination of the S-scheme heterojunction and the Schottky junction, where electrons migrated from WO₃ across TiO₂ to rGO. From a macroscopic viewpoint, the transfer of photogenerated electrons occurs like a "step, " with TiO₂ acting as a "bridge".

In summary, in the ternary WO₃/TiO₂/rGO composite system, the cross-coupling effect of rGO and the S-scheme heterojunction formed between TiO₂ and WO₃ suppresses the recombination of comparatively useful carriers and provides a variety of potential pathways for charge carrier transfer. In addition, the use of rGO as a supporting matrix can provide more surface-active sites and enhance the light absorption. The collective and positive synergy of rGO and the S-scheme heterojunction formed between TiO₂ and WO₃ results in a large driving force for the photocatalytic reaction, which increases the water-splitting H₂-evolution activity.

We successfully fabricated a graphene-modified WO₃/TiO₂ S-scheme heterojunction photocatalyst. The WTG composite was prepared by a facile one-step hydrothermal approach, and exhibited superior photocatalytic activity for H₂ production through water splitting, which was 3.5-fold that of bare TiO₂ under the same conditions. It was observed that the positive cooperative effect between the S-scheme heterojunction formed between WO₃ and TiO₂ and the Schottky heterojunction formed between TiO₂ and graphene sheets could effectively suppress the recombination of useful carriers, enhance the light harvest, and increase the number of active sites for the reduction reaction. This work provides new insights for the design and fabrication of novel S-scheme heterojunction photocatalysts and highlights the potential of utilizing the unique properties of rGO to design inexpensive, active photocatalysts with durability and stability for H₂ production.