Photocatalytic technology is a promising method for solving various energy and environmental problems [1-3]. Among the many developed photocatalysts, TiO₂ and g-C₃N₄ have shown excellent prospects in the fields of solar energy utilization and environmental protection because of their unique physical and chemical properties [4-10]. Compared with single photocatalytic materials, composite photocatalysts can improve quantum efficiency, increase specific surface area, promote sufficient contact between the catalyst surface and reactant molecules, and improve the separation efficiency of photogenerated electron-hole pairs. For example, the band structure of TiO₂ and g-C₃N₄ matches to a significant degree. In addition, the catalytic activity of the heterojunction composites is greatly enhanced when irradiation with ultraviolet or visible light [11-15].

Directly combining bulk g-C₃N₄ with other materials often results in small specific surface area and insufficient interface contact, limiting carrier migration between the two materials. Therefore, the abovementioned advantages of composite materials are not always fully realized. Researchers have improved the activity of g-C₃N₄ using various methods including fabrication of g-C₃N₄ nanotubes [16, 17], nanorods [18, 19], nanofibers [20], and nanosheets [21-23], coupled with direct mixing with TiO₂ to prepare composites. Yu et al. [24] used a chemical formation method, as opposed to the more common physical methods, to prepare a Z-scheme g-C₃N₄/TiO₂ system. The Z-scheme g-C₃N₄/TiO₂ heterojunction was fabricated by calcining a mixture of TiO₂ and urea. Subsequently, g-C₃N₄ was grown in situ on the surface of the TiO₂ particles.

Construction of high catalytic activity carbon-doped TiO₂ (C-TiO₂) is a popular research topic in the field of environmental photocatalysis [25-30]. However, significant drawbacks exist for traditional C-TiO₂ preparation methods, such as the need for additional carbon sources and carbon component aggregation. Thus, it is necessary to continuously optimize existing methods and design novel strategies for preparing multifunctional carbon-doped TiO₂ composite materials. To the best of our knowledge, few studies have considered the effects of the carbon doping state on TiO₂ photocatalytic activity or the synergistic effects of defect-assisted surface modification by in situ deposition of g-C₃N₄ on the photocatalytic performance of g-C₃N₄@C-TiO₂ heterojunctions.

To increase the number of active sites and defects in the TiO₂ material and achieve rapid and efficient transfer of photogenerated charges, a one-step hydrothermal method was used for in situ preparation of nano-carbon-doped TiO₂ to increase oxygen vacancy defects and surface active sites. A simple chemical vapor deposition method was subsequently used to crack urea through furnace heating. The effects of the carbon doping state and synergistic effects of surface modification by in situ deposition of g-C₃N₄ on the performance of the g-C₃N₄@C-TiO₂ composite photocatalysts were investigated.

C-TiO₂ (anatase) was synthesized using a hydrothermal combined calcination method. First, 3 mL of TiCl₄ was dissolved in 70 mL of glycol with concentrations of 5, 10, and 20 g/L. To the mixed solution, 2 mL of ammonia was added under magnetic stirring for 30 min to ensure complete hydrolysis. Subsequently, 5 mL of 0.01 g/mL glucose solution was added dropwise and left to stir for 1 h. The obtained mixture was subjected to hydrothermal treatment at 150 ℃ for 10 h. After centrifugation, washing, and drying, the precursor was prepared. The C-TiO₂ nanoparticle powder was obtained after the precursor was calcined at 450 ℃ for 3 h, where 5C-TiO₂, 10C-TiO₂, and 20C-TiO₂ represent carbon-doped anatase obtained using glucose concentrations of 5, 10, and 20 g/L, respectively. For comparison, pure TiO₂ and g-C₃N₄ samples were prepared where pure TiO₂ was prepared by the solvothermal method. A suitable amount of TiCl₄ was mixed with 70 mL of ethylene glycol and 2 mL ammonia was added dropwise to the mixed solution under magnetic stirring. Afterwards, the mixture was stirred for 1 h and placed in a hydrothermal kettle for reaction at 150 ℃ for 10 h. After the reaction was complete, the precipitate was filtered and washed three times with water and ethanol. The precipitate was then dried at 80 ℃ for 12 h. Finally, the dried powder was heated to 450 ℃ for 5 h at a heating rate of 5 ℃/min, resulting in an anatase-phase TiO₂ solid. Nanoporous g-C₃N₄ was prepared via pyrolysis of an aqueous urea solution, as described in detail in previous studies [9, 13]. Urea and distilled water were mixed in a mass ratio of 3:4 to form the aqueous urea solution, which was then placed in a corundum crucible with a cover and calcined at 400 ℃ for 1 h and subsequently at 500 ℃ for 2 h at a heating rate of 15 ℃/min.

The C-TiO₂ powder was dispersed on a protruding platform in the corundum crucible and urea was uniformly spread on the bottom of the crucible. The covered corundum crucible was placed in a rapid heating sintering furnace, calcined at 400 ℃ for 1 h, and then heated to 500 ℃ for 2 h at a heating rate of 15 ℃/min. After cooling, the g-C₃N₄@C-TiO₂ composite photocatalyst was obtained. For comparison, a g-C₃N₄@TiO₂ (without carbon) sample was prepared using a similar process. A schematic diagram of the preparation process is shown in Fig. 1.

The phase composition of the obtained samples was characterized by X-ray diffractometry (XRD, D8 Advance, Bruker, Germany) using a Cu K_α radiation source. The microstructures and element mapping of the prepared samples were observed by transmission electron microscopy (TEM, Talos F200X, FEI, USA) equipped with an energy dispersive spectrometer (EDS). X-ray photoelectron spectroscopy (XPS) analysis was performed (Axis Ultra DLD, Shimadzu, Japan) with monochromatized Al K_α radiation. The UV-visible diffuse reflection spectra (DRS) were recorded using a UV-visible spectrophotometer (UV-3150, Shimadzu, Japan) equipped with an integrating sphere. The photoelectrochemical properties of the samples were characterized using an electrochemical analyzer (CHI660E, Shanghai Chenhua, China) with a three-electrode system. ITO glass substrate with a photocatalyst film (20 mg), calomel electrode, and Pt wire were served as the working, reference, and counter electrodes, respectively. A Na₂SO₄ solution (0.5 mol/L) was used as the electrolyte solution. Photoluminescence (PL) spectra were obtained using a fluorescence spectrophotometer (F-7000, Hitachi, Japan) at room temperature. Electron paramagnetic resonance (EPR) spectroscopy was performed using an EPR spectrometer (JES FA200, JEOL Ltd., Japan).

The photocatalytic activities of the samples were tested by decomposing Rhodamine B (RhB) and phenol pollutants in an aqueous solution under visible-light. A 350-W xenon arc lamp with a UV-cutoff filter (λ > 420 nm) was placed into a double-wall cylindrical quartz reactor with a water circulator. The composite photocatalyst powder (100 mg) was suspended in 100 mL of RhB or phenol solution (10 mg/L). The suspensions were magnetically stirred in the dark for 1–2 h to establish adsorption-desorption equilibrium before irradiation. For the photocatalytic activity test, approximately 5 mL of the mixture was collected every 0.5 h and centrifuged for 10 min at 9000 rpm to remove the photocatalyst. The supernatant was then analyzed using a UV-visible spectrophotometer (UV-3150, Shimadzu, Japan) and the characteristic absorption peaks were evaluated. The removal rates of RhB and phenol were calculated from the absorbance according to the Beer-Lambert law. The detection process of the active species was similar to that of the photocatalytic performance testing. Ammonium oxalate (AO, h⁺ capture agent), isopropyl alcohol (IPA, ·OH capture agent), and benzoquinone (BQ, ·O₂^-1 capture agent) were added to the photocatalytic reaction system to determine the active species in the photocatalytic process.

First-principles calculations were performed to investigate the electronic structure of the carbon-doped TiO₂ using the Vienna Ab initio Simulation Package (VASP) based on density functional theory (DFT) [31, 32]. The exchange correlation functional was introduced by the generalized gradient approximation (GGA) of Perdew-Berke-Ernzerhof (PBE) [33, 34]. A 2×2×1 supercell with 48 atoms was used for the doped TiO₂. In the supercell, 1 oxygen atom was substituted by a carbon atom and simultaneously formed a single oxygen vacancy. To examine the concentration of carbon atoms, 2 oxygen atoms were substituted by a single carbon atom and 1 oxygen vacancy in the supercell. All plane waves with a cutoff energy of 400 eV were used in the basis function. The convergence criteria for structural and energy optimization at each atom were set as 0.01 eV/Å and 10^-6 eV, respectively.

The XRD patterns obtained from the pure TiO₂, 5C-TiO₂, 10C-TiO₂, 20C-TiO₂, and g-C₃N₄ @10C-TiO₂ samples are shown in Fig. 2. The crystal structure of TiO₂ (anatase) was not changed by the combination of carbon and g-C₃N₄ and no obvious carbon peaks were observed in the samples, indicating that carbon was doped into the anatase crystals. Fig. 2(b) shows a partially enlarged view of the 23°–27° diffraction angle, where all C-TiO₂ samples were shifted to a lower angle. This indicated that some carbon was doped in the interstitial space in anatase, leading to increased interplanar spacing. However, as the carbon content increased, the diffraction peaks of 10C-TiO₂ and 20C-TiO₂ were not further shifted to lower angle, indicating that carbon may be doped in the form of lattice position substitution. In addition, the angle of the g-C₃N₄@10C-TiO₂ composite was shifted to a higher angle. The stress induced by the interaction of g-C₃N₄ and 10C-TiO₂ may have led to the deformation and shrinkage of the TiO₂ lattice during thermal deposition.

To further confirm the chemical state of carbon on the surface of TiO₂, XPS characterization of the C-TiO₂ samples was performed. Fig. 3 shows the XPS spectra of the 5C-TiO₂, 10C-TiO₂, and 20C-TiO₂ samples. It could be seen from the carbon peaks in Fig. 3(a)–(c) that the peak intensity of the adsorbed C-C and C-OH bonds formed after pyrolysis of the organic matter on the surface of TiO₂ increased with increasing carbon content. This indicated that the amount of carbon adsorbed on the surface increased with increasing concentration of the organic precursor. The Ti-C bond formed by substitution of the lattice oxygen position with carbon followed the same trend. However, the peak intensity of the O-C bond formed by carbon gap doping initially increased with increasing carbon content and then plateaued. These results showed that when the carbon content was increased to a certain level, the carbon-doped TiO₂ gradually changed from gap to substitution doping modes. The Ti 2p peak in Fig. 3(d) further supports the above conclusion. That is to say, with increasing carbon doping content, the peak strength of the Ti-C peak gradually increased and the corresponding peak intensity of the Ti-O peak gradually decreased, indicating that carbon replaced the position of oxygen in the lattice.

The UV-vis DRS of the pure TiO₂, 5C-TiO₂, 10C-TiO₂, 20C-TiO₂, g-C₃N₄@5C-TiO₂, g-C₃N₄@10C-TiO₂ and g-C₃N₄@20C-TiO₂ samples are shown in Fig. 4. As the carbon content increased, the absorption edge of C-TiO₂ was significantly red-shifted and absorbance was enhanced. When combined with g-C₃N₄, the red shift of the absorption edge was more dramatic, especially for the g-C₃N₄@10C-TiO₂ sample. This indicated that the g-C₃N₄@10C-TiO₂ sample exhibited improved visible light response. The band gap energy (E_g) of the C-TiO₂ samples were calculated using the Tauc/Davis-Mott expression, as shown in Fig. 4(b). The band gap of C-TiO₂ gradually decreased with increasing carbon doping content, which also explained the red shift of the absorption edge.

The microstructures of the g-C₃N₄@10C-TiO₂ material were characterized by TEM and element mapping was achieved using EDS, as shown in Fig. 5. The black C-TiO₂ nanoparticles with a diameter of 20–30 nm were surrounded by a thin layer of g-C₃N₄. As seen in the HRTEM image (Fig. 5(b)), the thin g-C₃N₄ layer was tightly coated around the C-TiO₂ nanoparticles, indicating that the C-TiO₂ nanoparticles were successfully surface-modified, which improved the transmission of photogenerated carriers. In addition, the lattice fringes of TiO₂ exhibited significant defects (indicated by the red circle in Fig. 5(b)), mainly due to interstitial carbon doping. In the HRTEM image in Fig. 5(c) most of the TiO₂ lattice fringes were arranged regularly. However, a small number of defects were found in the middle of the periodically arranged lattice fringes (as shown in the dotted box), indicating the presence of partial oxygen vacancies in C-TiO₂. From the electron diffraction pattern in Fig. 5(d), it was determined that g-C₃N₄@10C-TiO₂ exhibited a polycrystalline structure. The element mappings of the rectangular frame shown in Fig. 5(a) are displayed Fig. 5(e)–(h), corresponding to C, N, O, and Ti, respectively. These maps further demonstrate that carbon and g-C₃N₄ were successfully incorporated into the g-C₃N₄@10C-TiO₂ composites.

The photocatalytic activities of the pure TiO₂, g-C₃N₄, g-C₃N₄@TiO₂, g-C₃N₄@5C-TiO₂, g-C₃N₄@10C-TiO₂, and g-C₃N₄@20C-TiO₂ samples were examined for their RhB and phenol photodegradation ability under visible-light irradiation (Fig. 6). The g-C₃N₄@C-TiO₂ sample exhibited higher photocatalytic degradation efficiency than those of the pure TiO₂, g-C₃N₄, and g-C₃N₄@TiO₂ samples (without carbon). With increasing carbon content, the photocatalytic degradation efficiency of g-C₃N₄@C-TiO₂ initially increased and subsequently decreased. Calculation of the Langmuir-Hinshelwood mechanism (shown in Fig. 6(b) and (d)) showed that g-C₃N₄@10C-TiO₂ exhibited the largest apparent constant (k) of 0.036 and 0.039 min^-1 for RhB and phenol photodegradation, respectively, which was 150/139, 6.4/6.8, 2.3/3.0, and 1.7/2.1 times higher than those of pure TiO₂, 10C-TiO₂, g-C₃N₄, and g-C₃N₄@TiO₂, respectively. For comparison, the degradation rates and k values for RhB and phenol degradation by the as-prepared samples and other reported photocatalysts are listed in Table 1 [35-38].

The UV-vis absorption spectra (Fig. 6(e) and (f)) of the RhB and phenol solutions in the presence of g-C₃N₄@10C-TiO₂ under visible-light irradiation show that the intensity of the RhB and phenol absorption peaks gradually decreased with increasing degradation time and were gradually blue-shifted, indicating that the photocatalyst induced mineralization and decomposition of the organic pollutants. The degradation cycle performance (Fig. 6(g)) of RhB and phenol (inserted image) by g-C₃N₄@10C-TiO₂ showed that the degradation activity towards RhB and phenol did not decrease significantly after four cycles, indicating good cyclic stability.

To test the photocharge transfer efficiency of the catalyst, the sample was measured using impedance spectroscopy and the results are shown in Fig. 7(a). Similar to the photocatalytic degradation activity, the C-TiO₂ samples exhibited small impedance values, with g-C₃N₄@10C-TiO₂ exhibiting the lowest impedance, indicated that it had a higher photogenerated electron transfer efficiency. The PL spectra of the as-prepared pure g-C₃N₄, g-C₃N₄@TiO₂, g-C₃N₄@5C-TiO₂, g-C₃N₄@10C-TiO₂, and g-C₃N₄@20C-TiO₂ are shown in Fig. 7(b). Compared to pure g-C₃N₄, g-C₃N₄@TiO₂, g-C₃N₄@5C-TiO₂, and g-C₃N₄@20C-TiO₂, g-C₃N₄@10C-TiO₂ exhibited the lowest peak intensity, indicating that it had the lowest charge recombination probability. From the EPR spectra (shown in Fig. 7(c)), it is clear that the peak intensity gradually increased with increasing carbon content, which can be attributed to the formation of oxygen vacancies by carbon reduction and doping. The active species capture experiment confirmed that ·OH and ·O₂^- radicals were the dominant in the RhB and phenol degradation by g-C₃N₄@10C-TiO₂, while the h⁺ radical was less significant (Fig. 7(d)).

To further investigate the effects of carbon doping on the band structure of TiO₂, the energy band structures of TiO₂ and C-TiO₂ were calculated from first principles, the results of which are shown in Fig. 8. The band gap between the conduction and valence bands of the pure TiO₂ was empty. Some impurity energy levels were introduced between the conduction and valence bands when a certain amount of carbon was doped. However, when carbon doping was excessive, the Fermi level of the 20C-TiO₂ sample passed through the conduction band and the impurity level was localized, which hindered electron transfer [39]. This phenomenon may be caused by excessive carbon doping leading to the formation of bulk oxygen vacancies, which can easily bind photogenerated electrons and became recombination centers for photogenerated carriers.

The formation mechanism of the g-C₃N₄@C-TiO₂ composite and its mechanism of influence on electron transfer are shown in Fig. 9(a). According to the experimental results, in addition to the solid solution of carbon atoms in the TiO₂ lattice by substitution and interstitial doping, some carbon atoms were adsorbed on the TiO₂ surface via C-C and C-OH bonding. Therefore, g-C₃N₄ was tightly bound to the surface of C-TiO₂ by π-conjugate and hydrogen bonding. Photogenerated electrons in the g-C₃N₄@C-TiO₂ composite were susceptible to electron transport through internal doping and surface hybridized carbon atoms. The proposed schematic of the photocatalytic mechanism for the g-C₃N₄@C-TiO₂ heterojunction is shown in Fig. 9(b). Based on the results described above, it can be concluded that the most likely mechanism for the g-C₃N₄@C-TiO₂ heterojunction was a direct Z-scheme mechanism. There are two main reasons for this. First, the active species for RhB and phenol degradation by g-C₃N₄@10C-TiO₂ are ·OH and ·O₂^-. Only h⁺ on the valence band of C-TiO₂ can produce ·OH, while the oxidation ability of h⁺ on the valence band of g-C₃N₄ is insufficient to form ·OH. Therefore, the holes cannot migrate from the valence band of C-TiO₂ to the valence band of g-C₃N₄ and remain in their original position with a higher oxidation potential to produce ·OH. Second, the Fermi level of the surface of C-TiO₂ is lower than that of g-C₃N₄. Therefore, when C-TiO₂ and g-C₃N₄ are combined, an inherent electric field from g-C₃N₄ to C-TiO₂ forms between the g-C₃N₄@C-TiO₂ heterojunctions [39, 40]. The holes in the valence band of C-TiO₂ cannot move to the valence band of g-C₃N₄ because of rejection by the built-in electric field. In contrast, the electrons on the C-TiO₂ conduction band can migrate to the valence band of g-C₃N₄ owing to the attraction of the inherent electric field, forming a direct Z-type heterojunction. The presence of a certain amount of surface oxygen vacancies in C-TiO₂ can generate a new defect level, which can effectively separate and transfer electron-hole pairs to the catalyst surface. The highly conductive carbon connecting TiO₂ and g-C₃N₄ can transfer electrons rapidly to the valence band of g-C₃N₄, which forms a direct Z-type photocatalytic heterojunction. This significantly improves the separation and transfer efficiency of photogenerated carriers and photocatalytic activity of the g-C₃N₄@C-TiO₂ heterostructured photocatalysts.

In summary, g-C₃N₄@carbon-doped anatase TiO₂ heterostructure photocatalysts were prepared by a facile hydrothermal combined vapor deposition method. With increasing carbon content, the diffraction peaks of C-TiO₂ initially shifted to lower angles but remained static after a critical doping content. Thus, at a low carbon content, carbon was doped into TiO₂ via interstitial doping. As the carbon content increased, carbon was gradually doped via lattice substitution. This was further confirmed by the XPS results, which showed that with the increasing carbon content, the peak intensity arising from the O-C bond produced by gap doping initially increased and then plateaued, while the peak intensity arising from the Ti-C bond increased gradually. The decreased Ti-O peak intensity further proved that more carbon replaced the oxygen in the lattice. The band gap of C-TiO₂ gradually decreased with increasing carbon doping content, which explained the red shift of the absorption edge. With increasing carbon content, the photocatalytic degradation efficiency of g-C₃N₄@C-TiO₂ initially increased and subsequently decreased. The g-C₃N₄@10C-TiO₂ sample exhibited the largest apparent constant of 0.036 (0.039) min^-1 for RhB (phenol) photodegradation. g-C₃N₄ was tightly bound to the surface of C-TiO₂ by π-conjugate and hydrogen bonding. The novel g-C₃N₄@C-TiO₂ heterojunction exhibited a direct Z-scheme photocatalytic mechanism. Appropriate carbon doping content and surface modification of g-C₃N₄ introduced new non-local impurity levels and surface states in the catalyst, which more efficiently separated and transferred photogenerated electrons, enhancing the photocatalytic activity of the g-C₃N₄@C-TiO₂ heterostructure photocatalysts.