Rapid economic development has resulted in severe environmental pollution, where industrial wastewater pollution represents a serious form. Currently, traditional methods such as physical adsorption [1], chemical redox [2], and biological treatment [3] have been adopted to mitigate pollution. However, these methods are inefficient and often lead to secondary pollution. Semiconductor photocatalysts are green materials that use sunlight to decompose organic contaminants. In 1972, a study on the water photolysis by a TiO₂ electrode was published in Nature [4]. Since then, rapid progress has been made in terms of semiconductor materials [5]. Semiconductor photocatalysis can directly utilize solar energy and does not cause secondary pollution, thereby enabling its wide applicability in environmental control.

TiO₂ is an ideal photocatalyst because of its low cost, high stability, and good photocatalytic effects [6]. However, this material has some disadvantages, including the limited absorption of ultraviolet light and serious photogenerated charge recombination. To solve these problems, significant efforts have been made. For example, to effectively increase the light absorption ability of TiO₂, its band gap has been adjusted to capture visible light [7]. In addition, heterojunction formation by combining TiO₂ with other semiconductors can effectively suppress photogenerated electron and hole recombination, improving the carrier utilization [8].

Recently, people have paid significant attention to the internal electron-hole transfer path of semiconductor-based composite photocatalysts. By studying the physicochemical and photocatalytic properties of different composite photocatalysts, a reasonable carrier transfer path can be proposed. This type of study has greatly enriched the theoretical understanding of photocatalysis. Heterostructure (including Type-I and Type-II heterojunctions [9]), Z-scheme heterojunctions (including direct Z-scheme [10], and indirect Z-scheme heterojunctions [11]) have been proposed as mechanistic explanations of performance enhancement. Currently, two kinds of heterostructures are popular and are briefly introduced below.

First, type-II heterojunction structure involves no other medium between the two semiconductors. As shown in Fig. 1, the two semiconductors directly contact each other. When the carriers are transferred according to a fixed path, a type-II heterojunction structure is formed [12]. Using TiO₂-CdS as an example, under simulated sunlight, TiO₂ and CdS are simultaneously excited and electrons jump from the valence band to conduction band, leaving holes in the valence band. Because the two semiconductors are in good contact and the valence band position of TiO₂ is more positive than that of CdS, the holes will spontaneously transfer from the valence band of TiO₂ to that of CdS. Similarly, the TiO₂ conduction band position is more negative than that of CdS. Therefore, the electrons transfer from the CdS conduction band to that of TiO₂. Thus, the holes become concentrated in the CdS valence band and the electrons concentrated in the TiO₂ conduction band for redox reactions [13]. This behavior can be effectively utilized for photogenerated electron and hole separation, greatly reducing the recombination rate of the carriers. However, the electrons transfer to the conduction band with low energy, leading to decreased reduction ability. Additionally, the holes moving to a lower valence band exhibit weakened oxidation power [14].

Second, indirect Z-scheme heterojunction structure involves two semiconductors connected by a conductive medium. As shown in Fig. 2, after the three structures are combined together, the carriers transfer through a fixed path, called indirect Z-scheme heterojunction structure [15]. Using TiO₂-Au-g-C₃N₄ as an example, gold is deposited on the surface of TiO₂, which is subsequently composited with g-C₃N₄. Thus, gold is present between TiO₂ and g-C₃N₄, and acts as a good electron conductor. After photoexcitation of TiO₂ and g-C₃N₄, the valence band electrons are excited to the conduction band and the electrons in the TiO₂ conduction band are transferred to the valence band of g-C₃N₄ through the gold. Thus, the holes remaining in the TiO₂ valence band can oxidize organic compounds, while the electrons in the g-C₃N₄ conduction band can undergo reduction. As such, g-C₃N₄ is prevented from being oxidized by the photogenerated holes [16].

However, contradictions remain between the heterojunction and Z-scheme structures, and the distinction is ambiguous. To date, comparison of the two structures has not yet been fully performed. If the two carrier transfer structures can be distinguished, it will be possible to design highly efficient photocatalysts for specific degradation conditions. Therefore, the preparation of highly efficient photocatalysts and internal principles must be explored by combining theory and experiment. In this manner, highly efficient catalysts and synergistic principles can be developed simultaneously.

Herein, the all-solid-state Z-scheme and type-II heterojunction structures were studied [17] and TiO₂-C-C₃N₄ and TiO₂-C₃N₄ catalysts were prepared. Recently, C₃N₄ has been extensively studied due to its special structure and good performance. It exhibits a wide absorption spectrum, acting as a photocatalyst under visible light irradiation. Based on previous studies, C₃N₄ was prepared using melamine herein. The all-solid-state Z-scheme heterojunction structure refers to the introduction of a conductive material between two band-gap-matched semiconductors. The entire route adopts a "Z" shape that mimics the reaction pathway for photosynthesis. The TiO₂-C-C₃N₄ ternary catalyst was compared with the type-II heterojunction TiO₂-C₃N₄ formed by TiO₂ and C₃N₄. According to the experimental photocatalysis results, the differences in the photocatalytic mechanism for the two different composite materials were macroscopically explored. It was verified that the two different heterojunction structures resulted in different photocatalytic performances. The macroscopic experimental results were used to explore the degree of adaptation of different photocatalytic mechanisms in the same degradation system to design more efficient photocatalysts for different degradation systems.

First, 3 mL of tetrabutyl titanate was dispersed in 50 mL of glacial acetic acid through dropwise addition. This mixture was stirred at room temperature for 15 min and subsequently transferred to an oven for hydrothermal treatment. The temperature was set to 160 ℃ and the time was 12 h. After the hydrothermal reaction, the sample was cooled and centrifuged. The precipitate was washed three times with water and alcohol, then dried and calcined in a muffle furnace at 550 ℃ for 4 h. Finally, TiO₂ nanoparticles with a diameter of approximately 30 nm were prepared.

C₃N₄ was prepared by calcination using melamine as a precursor. For the synthesis of TiO₂-C₃N₄ composites, 0.75, 1.5, 3, and 6 g of melamine and 0.3 g of TiO₂ (with mass ratios of melamine to TiO₂ of 2.5:1, 5:1, 10:1, and 20:1, respectively) were simultaneously dispersed in 100 mL of water. The mixture was stirred for 2 h and subsequently centrifuged. Finally, the mixture was calcined at 550 ℃ for 4 h in a tube furnace under Ar gas to prepare the TiO₂-C₃N₄ composites. The samples were synthesized with mass ratios of melamine to TiO₂ of 2.5:1, 5:1, 10:1, and 20:1 denoted as TiO₂-2.5C₃N₄, TiO₂-5C₃N₄, TiO₂-10C₃N₄, and TiO₂-20C₃N₄, respectively.

First, 5 g/L glucose was added to 100 mL of deionized water, stirred for 15 min, and ultrasonically dispersed for another 15 min. Subsequently, 0.1 g of TiO₂ nanoparticles was dispersed in 100 mL of the glucose solution under stirring for 2 h, and subsequently centrifuged and dried, followed by calcination at 550 ℃ for 4 h in a tube furnace under an Ar gas flow. Thus, carbon-coated TiO₂ (TiO₂-C) particles were prepared.

In addition, 0.75, 1.5, and 3 g of melamine and 0.3 g of TiO₂-C with mass ratios of melamine to TiO₂-C of 2.5:1, 5:1, and 10:1, respectively, were simultaneously dispersed in 100 mL water. The mixture was stirred for 2 h and subsequently centrifuged, followed by calcination at 550 ℃ for 4 h in a tube furnace under Ar gas flow, to prepare the TiO₂-C-C₃N₄ composites. The samples prepared with mass ratios of melamine to TiO₂-C of 2.5:1, 5:1, and 10:1 are referred to as TiO₂-C-2.5C₃N₄, TiO₂-C-5C₃N₄, and TiO₂-C-10C₃N₄, respectively.

The composite morphology was measured by scanning electron microscopy (SEM; JEOL JSM-6700F) and transmission electron microscopy (TEM; JEOL JEM-2100F). The crystal structure was characterized using an X-ray diffraction (XRD) instrument (PANalytical diffractometer D/max 40 kV) with Cu Kα radiation (λ = 0.154598 nm). The N₂ adsorption-desorption isotherms were obtained at –196 ℃ using an Autosorb-iQ2 instrument. The UV-vis diffuse reflectance spectra (DRS) of the samples were recorded using a PerkinElmer Lambda 35 spectrophotometer from 200–800 nm. Photoluminescence (PL) spectra were measured using a PerkinElmer LS-55 fluorescence spectrophotometer at an excitation wavelength of 315 nm.

A 350 W Xe lamp (Shanghai Lansheng Lighting Appliance Co. Ltd.) was used to simulate AM 1.5 solar illumination. The methylene blue (MB) absorbance after photocatalytic experiments was measured using an ultraviolet-visible spectrophotometer (Shimadzu, UV-1700 or UV-2550). The electrochemical impedance, transient photocurrent response, and Mott-Schottky curves of the samples were acquired using an electrochemical workstation (Shanghai Chenhua Instrument Co., Ltd. CHI660B) with the newly prepared samples deposited onto conductive films as working electrodes, Pt as a counter electrode, and Ag/AgCl as a reference electrode. In addition, a 0.1 mol/L sodium sulfate solution was used as an electrolyte and low-pressure mercury lamp was used as the light source.

MB was used as a simulated pollutant to examine the photocatalytic efficiency of the prepared samples. First, a high-concentration (5 × 10^–5 mol/L) MB solution was prepared. Then, 25 mg of the prepared photocatalyst was placed in a beaker with 100 mL of a 5 × 10^–5 mol/L MB solution, which was ultrasonically dispersed for 15 min and stirred for another 15 min in the dark to obtain a dark adsorption equilibrium. The reactor was then irradiated with a Xe lamp at an energy intensity of 140 W/m² (Licor Biosciences, LI-250 A). During the photocatalytic reaction, O₂ was flowed into the reactor and 2 mL of the suspended solution was removed every 10 min for centrifugation. The supernatant was subjected to UV-visible spectroscopy (Shimadzu UV-1700) measurement at 664 nm. Thus, the concentration of MB was determined; this process was also used for the scavenger capture experiments.

As the prepared samples were all powder catalysts, it was necessary to evenly deposit the sample onto a conductive glass (ITO) surface. The treated ITO glass was fixed at the center of a homogenizer and the suspension containing the prepared photocatalysts and nafion solution was added dropwise to the center of the ITO glass. After standing for 30 s, the homogenizer switch was turned on and held for 20 s. The whole process was repeated twice, and the samples deposited onto the ITO glass were dried in the oven for subsequent use.

To investigate the structures of the prepared samples, the XRD patterns of the TiO₂-C, TiO₂-C₃N₄, and TiO₂-C-C₃N₄ composites were compared. Fig. 3A shows peaks at 27.5° and 13.5° for pure C₃N₄ [22], which are characteristic of this material. In addition, the anatase crystal structure was attributed to TiO₂ in the TiO₂-C₃N₄ composite, but no characteristic peaks of C₃N₄ were observed. The XRD pattern for the TiO₂-C-C₃N₄ compound was similar to that of TiO₂-C₃N₄ as the ratio of C₃N₄ in that composite was small and the graphite carbon crystallinity was poor. Therefore no characteristic peaks for C and C₃N₄ were detected [23].

The elemental composition and state of the prepared TiO₂-C₃N₄ powder were further analyzed by X-ray photoelectron spectra (XPS). The XPS peak position was corrected using the C 1s peak with a binding energy of 284.8 eV. Fig. 3B shows the full spectrum of TiO₂-C₃N₄, showing that the sample was composed of C, N, O, and Ti. Fig. 3C shows that the sample exhibited two C 1s peaks at 284.6 and 288.0 eV, with the former assigned to the synergistic effect of the C–C bonds, including accidental hydrocarbons from the XPS instrument itself and sp²-hybridized carbon atoms present in C₃N₄. The other peak at 288.0 eV corresponded to the synergistic effect of N–C=N bonds [24]. In Fig. 3D, the two peaks for O 1s at 529.8 and 532.1 eV were fitted, corresponding to Ti–O and O–H bonds, respectively. The peak at 532.1 eV was related to the presence of hydroxyl or water molecules on the surface of TiO₂-C₃N₄ [25]. From Fig. 3E, the Ti 2p spectrum shows two peaks at 458.6 and 464.3 eV, corresponding to Ti 2p_3/2 and Ti 2p_1/2, respectively. Fig. 3F shows three peaks at 398.5, 399.5, and 401.1 eV for N 1s, with those at 398.5 and 401.1 eV arising from the N bonded in the triazine ring (C–N=C) and the amino group (C–N–H) [26], while that at 399.5 eV arising from the tertiary nitrogen in N–(C)₃ groups [27].

Fig. 4 shows SEM images of the prepared TiO₂-C₃N₄ and TiO₂-C-C₃N₄ samples with the mass ratios of 1:2.5, 1:5, and 1:10 as well as HRTEM images for TiO₂-5C₃N₄ and TiO₂-C-10C₃N₄. From Fig. 4A, C₃N₄ adopted a two-dimensional structure. When melamine and TiO₂ were mixed and calcined, the images for the samples calcined at different ratios showed significantly different morphologies. Fig. 4B shows the image for TiO₂-2.5C₃N₄, which did not display a two-dimensional sheet structure for C₃N₄ with partially agglomerated TiO₂ particles into spherical shapes. The amount of melamine precursor was insufficient so the formed C₃N₄ was not observable. The addition of melamine disturbed the homogeneity of TiO₂ and a random spherical shape was formed in a localized range. When the mass ratio of TiO₂ to melamine was increased to 1:5, the TiO₂ nanoparticles became uniformly dispersed on the C₃N₄ sheet to form a two-dimensional layered structure (Fig. 4C). This was defined as a moderate ratio. For the TiO₂-10C₃N₄ sample (Fig. 4D) the excess melamine resulted in poor uniformity of the calcined sample. C₃N₄ did not show a sheet structure. The TiO₂ content was insufficient to wholly cover all C₃N₄. Therefore, TiO₂ was randomly deposited on the C₃N₄ surface. The morphology change of TiO₂-C-C₃N₄ was similar to that of the TiO₂-C₃N₄ composite. From Fig. 4E, the TiO₂-C dispersion showed no agglomeration and was relatively uniform. At the mass ratio of TiO₂-C to melamine of 1:2.5, the structure of the formed TiO₂-C-2.5C₃N₄ was irregular (Fig. 4F) and the two-dimensional structure of C₃N₄ was not observed. At a 1:5 mass ratio, the composite showed a flat two-dimensional structure composed of C₃N₄ with TiO₂-C uniformly covering the surface (Fig. 4G). At a 1:10 mass ratio, a large amount of C₃N₄ appeared in the sample, removing the two-dimensional morphology (Fig. 4H). Excess C₃N₄ agglomerated to form a random structure. The TiO₂-C particles were not completely covered on the C₃N₄ surface. To further investigate the sample microstructure, HRTEM images of TiO₂-5C₃N₄ and TiO₂-C-5C₃N₄ were also collected. From Fig. 4L, it is clear that TiO₂ was loaded on the surface of C₃N₄ in the TiO₂-5C₃N₄ sample. For TiO₂-C-5C₃N₄, the TiO₂ nanoparticle surfaces were coated with a carbon layer due to the introduction of the carbon medium (Fig. 4M). In addition, TiO₂ exhibited an apparent lattice spacing of approximately 0.34 nm, similar to the lattice constant of the (101) crystal plane for anatase TiO₂. Because of the interface between C₃N₄ and TiO₂, the two semiconductors were not simply physically mixed, confirming the formation of a close interface connection [28].

N₂ adsorption-desorption isotherms were measured to determine the surface areas of the prepared photocatalysts. Fig. 5 shows the N₂ adsorption-desorption isotherms and corresponding pore-size distribution curves (inset) for pure C₃N₄, TiO₂-5C₃N₄, and TiO₂-C-5C₃N₄. According to Brunauer-Deming-Deming-Teller (BDDT) classification, the pure C₃N₄, TiO₂-5C₃N₄, and TiO₂-C-5C₃N₄ exhibited type IV isotherms, indicating the presence of mesopores (2–50 nm) [29]–[31]. The hysteresis line was a H3 type with a high relative pressure range of 0.8–1.0, demonstrating the presence of slit-like pores. The isotherms of TiO₂-5C₃N₄ and TiO₂-C-5C₃N₄ showed higher adsorption at higher relative pressures (P/P₀ close to 1.0) so that the two composites contained large mesopores and macrospores [29]–[31]. The pore-size distribution curves of TiO₂-5C₃N₄ and TiO₂-C-5C₃N₄ (inset of Fig. 5) further confirmed the presence of mesopores and macropores. The parameters of the composites are summarized in Table 1. The Brunauer-Emmett-Teller (BET) surface area of TiO₂-5C₃N₄ and TiO₂-C-5C₃N₄ greatly increased compared to that of pure C₃N₄ due to the many smaller nanoparticles of TiO₂ loaded on the C₃N₄ surface. In addition, compared to the other samples, TiO₂-C-5C₃N₄ exhibited the lowest average pore-size. Overall, TiO₂-5C₃N₄ and TiO₂-C-5C₃N₄ were demonstrated to be mesoporous materials with high surface areas, which may facilitate reactant adsorption for improved photocatalytic performance.

The absorbance of the prepared samples was tested by UV-visible diffuse reflection spectroscopy. From these spectra, the forbidden band width was calculated using the formula proposed by Tauc/Davis and Mott et al. [32] as (αhv)^1/n = A(hv–E_g), in which α is the absorbance index, h is Planck's constant, v is the frequency, A is a constant, E_g is the semiconductor forbidden band width, and n is the semiconductor type. For direct bandgap semiconductors, n = , and for indirect bandgap semiconductors, n = 2. Both TiO₂ and C₃N₄ are indirect bandgap semiconductors. After (αhν)^1/n was plotted as a function of hv, the straight line portion was extended to the X-axis and the intersection was taken as the band gap width.

Fig. 6A shows the light absorption spectra with an inset of the band gap conversion plot. The absorption band edges of TiO₂ and C₃N₄ were observed at 350–400 and 42–250 nm, respectively[33]. After TiO₂ was hybridized with C₃N₄, the absorption band edge of the composite was largely unchanged, but the absorbance intensity increased from 400 to 500 nm due to the presence of C₃N₄. After the carbon layer was introduced into the composite, the absorption edge did not shift but the absorption intensity at 450–800 nm was significantly enhanced. The color of TiO₂ changed from white to black after carbon was coated on the surface, increasing the light absorption performance. In addition, the light absorption capacity of TiO₂-C-C₃N₄ was stronger than those of the other samples. In the inset, it is clear that the forbidden band widths of the TiO₂-C, TiO₂-C₃N₄, and TiO₂-C-C₃N₄ composites ranged between 3.22 and 3.24 eV. Thus, the differences between samples were minor, indicating that carbon and C₃N₄ had no effect on the band gap of TiO₂. C₃N₄ exhibits a forbidden band width of approximately 2.72 eV.

For an n-type semiconductor, the flat band potential can be approximated as the conduction band potential [34], [35]. The flat band potential can be measured according to the Mott-Schottky equation: , where in ε is the relative dielectric constant, ε₀ is the vacuum dielectric constant, N is the donor (for n-type semiconductors) or acceptor (for p-type semiconductors) density, and Δϕ_SC is the absolute value of the difference between the electrode and flat potentials. The slopes of the curves shown in Fig. 6B were all positive, indicating that both TiO₂ and C₃N₄ were n-type semiconductors [36]. After fitting the curve to a straight line, the extended straight line intersected the X-axis to provide a flat band potential. When the Ag/AgCl electrode was converted to a standard hydrogen electrode, the flat band potential can be calculated using the formula:

The forbidden band widths of C₃N₄ and TiO₂ were obtained from the UV-visible diffuse reflection data. The approximate position of the conduction band was calculated from the Mott-Schottky curve. Therefore, the valence band position of the two semiconductors was calculated. Fig. 7 shows a coordinate system with the band positions, which are also listed in Table 2. From Fig. 7, it is clear that the band gaps of TiO₂ and C₃N₄ overlaps and are quite similar. TiO₂ exhibited a lower valence band position, whereas C₃N₄ had a higher conduction band position, generating a type-II heterojunction structure.

Fig. 8A shows the photocatalytic performance of the prepared TiO₂-C₃N₄ sample. This test was performed to determine the optimal ratio of TiO₂ to C₃N₄. The sample containing the TiO₂ and melamine mass ratio of 1:5 (TiO₂-5C₃N₄) showed the best performance where 90% of the MB was removed within 40 min. This indicated that TiO₂ modified with a suitable amount of C₃N₄ can improve photocatalytic performance. The performance of pure TiO₂ was lower than that of TiO₂-5C₃N₄, but higher than those of the other samples tested herein. Therefore, when the amount of C₃N₄ in the composite was excessively small or large, the modification was deleterious. In addition, the performance of most composite samples were superior to pure C₃N₄, indicating that pure C₃N₄ exhibited poor photocatalytic performance under simulated solar light.

Fig. 8B compares the photocatalytic performance of the best TiO₂-C₃N₄ sample with that of TiO₂-C-C₃N₄ containing different proportions of constituent materials. The performance of TiO₂-C was the best, but when TiO₂-C was combined with C₃N₄ to form an all-solid-state Z-scheme heterojunction structure, the performance significantly decreased. The decreased photocatalytic performance can also explain the heterojunction structure differences between the TiO₂-C₃N₄ and TiO₂-C-C₃N₄ samples. For the TiO₂-C-C₃N₄ ternary composite, TiO₂-C-5C₃N₄ showed the best performance, which was attributed to its relatively good two-dimensional morphology and tight TiO₂ particle distribution. The TiO₂-C-2.5C₃N₄ sample with agglomerated morphology showed poor particle dispersion and inferior performance. The C₃N₄ content in TiO₂-C-10C₃N₄ was excessively high so that TiO₂-C particles could not be completely covered on the C₃N₄ surface. Additionally, the C₃N₄ performance was poor and the composite light absorbance was negatively affected, resulted in decreased performance compared to that of TiO₂-C-5C₃N₄.

The –ln (C/C₀) vs time curve was fitted by first-order kinetics using the data provided in Figs. 8A and 8B and a linear graph was obtained (Fig. 8C). The value of –k was calculated from the straight line slope. Based on the –k value of TiO₂, the photocatalytic performance of the samples was normalized and plotted as a histogram, as shown in Fig. 8D. The –k value of TiO₂-5C₃N₄ was 1.45 times that of TiO₂, indicating that C₃N₄ addition significantly improved the photocatalytic performance. The –k values of the TiO₂-2.5C₃N₄ and TiO₂-10C₃N₄ samples were low, indicating that the photocatalytic performance of the sample with the 1:5 ratio of TiO₂ to melamine was optimal. Overall, for the ternary Z-scheme system, the –k value of TiO₂-C-5C₃N₄ sample was optimal with a TiO₂-C to melamine ratio of 1:5. Therefore, it was confirmed that type-II heterojunction was achieved in the TiO₂-5C₃N₄ sample and that assigned to the Z-scheme was TiO₂-C-5C₃N₄ since the photocatalytic performance of the two samples were the best. In the cycling experiment of TiO₂-5C₃N₄ shown in Fig. 8E, the decomposition ratio of MB after the 5th cycle was ˃90% compared to original value, indicating that TiO₂-5C₃N₄ performance attenuation was small and the photocatalyst was stable.

The photocurrent and electrochemical impedance spectra were obtained using a three-electrode system and the results are shown in Fig. 9. Fig. 9A shows that all the samples with TiO₂ exhibited a positive photocurrent, which could be described as an n-type semiconductor, consistent with the Mott-Schottky curve. The its Mott-Schottky curve for C₃N₄ exhibited n-type character, but the photocurrent curve had a negative current likely due to the excessively small photocurrent of C₃N₄ (on the order of 10^–7) to be recognized as negative considering the error range of the instrument. In addition, TiO₂-C₃N₄ exhibited the highest photocurrent, even larger than the sum of TiO₂ and C₃N₄. Thus, the prepared composite exhibited high charge separation efficiency. Moreover, the two semiconductors were hybridized and not simply mixed, similar to the result inferred from the TEM images, which can be attributed to the formation of type-II heterojunction structure. This structure is beneficial for directional migration of the photogenerated charges, greatly reducing the recombination rate of carriers and promoting their transfer. The TiO₂-C photocurrent was also improved due to the spontaneous transfer of photogenerated electrons over TiO₂ to the carbon coated surface. The carbon layer acted as a container for electron storage, inhibiting carrier recombination [37]. However, it should be noted that although the carbon and C₃N₄ modifications increased the photocurrent of TiO₂, the photocurrent of TiO₂-C-C₃N₄ significantly decreased, indicating inferior charge separation efficiency in this system. This is likely due to the special carrier transfer path for the Z-scheme heterojunction structure of TiO₂-C-C₃N₄. The experimentally determined photoelectrochemical performance showed that it was difficult to achieve complete 1:1 recombination of carriers in the ideal model of the Z-scheme heterojunction structure, which likely resulted in the poor photocatalytic effect of the TiO₂-C-C₃N₄ composite. Therefore, it was speculated that the Z-scheme heterojunction structure of TiO₂-C-C₃N₄ may not be ideal. However, it is possible to achieve ideal Z-scheme heterojunction structure when the carrier concentrations of the two semiconductors (TiO₂ and C₃N₄) are identical. Moreover, the photoelectrochemical performance of pure TiO₂ was good due to the large number of electrons transferred to the surface.

Fig. 9B shows the electrochemical impedance spectroscopy (EIS) of the prepared samples. The curvature radii directly reflected the magnitude of interface charge transfer resistance. Smaller curvature radii indicated smaller interface resistances [38], [39]. The impedance of TiO₂-C was the smallest of the prepared sample, indicating that the carbon layer coating changed the space charge distribution of the surface, accelerating the charge separation and changing the charge distribution of the depletion layer. Thus, the interface charge transfer resistance decreased. After hybridization with C₃N₄, the EIS curvature radius of TiO₂-C₃N₄ became slightly smaller than that of pure TiO₂ and C₃N₄ due to the good conductivity of C₃N₄. This effectively improved the electron transfer efficiency and reduced interface resistance, increasing the quantum efficiency and photocatalytic performance of TiO₂. However, for the Z-scheme heterojunction structure of TiO₂-C-C₃N₄, the radius was higher than that of all the other samples, indicating that the charge transfer resistance was very large. This negatively affected the carrier separation, reducing photocatalytic performance. The results of the above photocurrent and impedance tests are consistent with those of the photocatalytic experiments.

Scavenger experiments were performed for the TiO₂-C₃N₄ catalyst as its activity was the highest of all prepared samples. The photocatalytic performance was tested in the presence of different scavengers, and the main active species in the photocatalytic reaction were determined by analyzing the decay of the photocatalytic performance. MB was used as the simulated pollutant, TiO₂-C₃N₄ as the photocatalyst, EDTA-2Na as a hole trapping agent, TBA as a hydroxyl radical trapping agent, and carbon tetrachloride as an electron trapping agent [40]. The obtained experimental results are shown in Fig. 10. After addition of the hole trapping agent, the photocatalytic performance was greatly reduced, demonstrating that the main active species for the MB degradation were holes. After addition of ·OH, the MB degradation performance was negatively affected because holes easily produce ·OH in aqueous systems and ·OH is an active species for MB degradation. In contrast, the photocatalytic performance was slightly improved after addition of the electron trapping agent because electron trapping suppressed the recombination of electrons and holes, resulting in more free holes for MB degradation, improving photocatalytic performance. Therefore, both holes and ·OH were determined to be active species in the reaction. By comparing the band gap diagrams of TiO₂ and C₃N₄, their valence band positions were determined at 2.122 and 2.782 eV, respectively. For a neutral solution system, a valence band hole of 1.99 eV is sufficient to generate hydroxyl radicals [41]. Therefore, both TiO₂ and C₃N₄ were capable of generating hydroxyl radicals in the photocatalytic system, which was also demonstrated in PL spectra. It is wellknown that hydroxyl radicals, as the main active species of photocatalysis, are responsible for organic pollutant degradation.

To further explore the photocatalytic mechanism of the samples, a hydroxyl radical capture experiment was performed. Terephthalic acid (TA) reacts with hydroxyl radicals to produce 2-hydroxyterephthalic acid which is a fluorescence emitter and is often used as a probe to analyze hydroxyl radicals by fluorescence spectroscopy [42]. From Fig. 11, it is clear that the TiO₂-5C₃N₄, TiO₂-C-5C₃N₄, and C₃N₄ samples showed distinct fluorescent peaks at 435 nm. In addition, the fluorescence intensity was almost linear as a function of illumination time (Fig. 11D). The increased fluorescence intensity with increasing illumination time indicates that more and more hydroxyl radicals were produced in the photocatalytic system. This demonstrates that the TiO₂-5C₃N₄, TiO₂-C-5C₃N₄, and C₃N₄ samples continuously produced hydroxyl radicals under simulated sunlight, consistent with the scavenger experiment results. However, the fluorescence peak intensities of TiO₂-C-5C₃N₄ and C₃N₄ over time were significantly lower than that of TiO₂-5C₃N₄, indicating that TiO₂-5C₃N₄ produced more hydroxyl radicals than TiO₂-C-5C₃N₄ and C₃N₄. In addition, the carrier transfer path of TiO₂-C-5C₃N₄ likely differed from that of TiO₂-5C₃N₄. Because hydroxyl radicals were the main active species (as demonstrated by the previous scavenger experiment), the photocatalytic properties of C₃N₄ and TiO₂-C-5C₃N₄ were judged to be lower than that of TiO₂-5C₃N₄, which was also demonstrated by the photocatalytic performance. In addition, C₃N₄ prepared with melamine can generate hydroxyl radicals because of the high positive potential of its valence band (2.122 eV) compared to that of ·OH/H₂O (1.99 eV). Therefore, it can be inferred that due to the Z-scheme heterojunction structure, the holes in the C₃N₄ valence band were transferred to the TiO₂ conduction band through the C medium. Thus, the holes in the C₃N₄ valence band were wasted and not able to generate hydroxyl radicals for organic pollutant degradation. This is the major reason for the decreased photocatalytic performance of the Z-scheme heterojunction structure.

In the photocatalytic experiment, TiO₂ was modified using two different carbonaceous materials. Although the performance was improved in both cases, the extent of improvement differed due to the mechanism of electron transfer and storage for photocatalysis in TiO₂-C. In addition, the holes in the valence band of TiO₂ remained. As shown in Fig. 12A, the photogenerated electrons and holes were separated by the mechanism of type-II heterojunction structure in the TiO₂-C₃N₄ material [43]. In principle, the photogenerated electrons in the conduction band of C₃N₄ should move toward the conduction band of TiO₂, and the valence band holes of TiO₂ toward the valence band of C₃N₄. This resulted in high charge separation efficiency and enhanced photocatalytic performance of TiO₂-C₃N₄, consistent with the performance results of other type-II heterojunction structures [44]. Although the carrier separation effect in TiO₂-C₃N₄ was excellent and the corresponding photocurrent was large, the catalytic performance was significantly dependent on the oxidation ability of the holes. The valence band position of C₃N₄ was higher than that of TiO₂, reducing the oxidation ability of the hole [45]. Therefore, the performance of TiO₂-C₃N₄ was lower than that of TiO₂-C. Moreover, TiO₂-C₃N₄ was formed by TiO₂ nanoparticle-coated C₃N₄ two-dimensional nanosheets with poor dispersion compared to that of the pure TiO₂ nanoparticles. Thus, the exposed area for photocatalysis was limited, resulting in lower photocatalytic performance. TiO₂-C maintained a high degree of dispersibility and the surface carbon layer promoted adsorption, which is beneficial for degradation of the organic pollutants [46]. Furthermore, the conductivity of C₃N₄ was inferior to that of graphitic carbon, which negatively impacted electron transfer [47].

It should be noted that the performance of the TiO₂-C-C₃N₄ ternary catalyst was the lowest of all tested materials in the photocatalytic experiment, inconsistent with those of other Z-scheme heterojunction structures[48]. According to the photocatalytic activity and PL results, the carrier in TiO₂-C-C₃N₄ transferred in a different path compared to that in TiO₂-C₃N₄. In summary, the prepared TiO₂-C-C₃N₄ was determined to be an all-solid-state Z-scheme heterojunction structure catalyst [49], with a carrier transfer mode that differed from those of TiO₂-C₃N₄ and TiO₂-C. In TiO₂-C-C₃N₄, the intermediate carbon layer acted as a bridge for electron transfer to connect the two bandgap-matched semiconductors. As shown in Fig. 12B, under simulated sunlight illumination, both TiO₂ and C₃N₄ were simultaneously excited and electrons and holes were generated on the respective conduction and valence bands. The photogenerated electrons in TiO₂ migrated along the conductive carbon layer to the valence band of C₃N₄ for recombination with the holes, leaving the holes in the TiO₂ valence band for subsequent oxidation reaction. This effectively reduced the recombination of electrons and holes.

Nevertheless, the two photocatalysts of TiO₂-C₃N₄ and TiO₂-C exhibited improved performance compared to that of TiO₂-C-C₃N₄. Theoretically, ternary composites with Z-scheme heterojunction structure often contain photogenerated electrons and holes with reductive and oxidative ability, respectively. Because of the high photocatalytic and photocurrent properties of TiO₂-C, the performance of the TiO₂-C-C₃N₄ ternary photocatalyst for MB degradation was unrelated to the effect of the carbon layer. This was clear from the mechanism comparison of the type-II heterojunction structure of TiO₂-C₃N₄ and Z-scheme heterojunction structure of TiO₂-C-C₃N₄. The Z-scheme heterojunction structure showed recombination of the electrons in the conduction band of one semiconductor with the holes in the valence band of the other semiconductor in the conductive medium. Therefore, some carriers with weak redox ability were sacrificed to leave the holes and electrons with strong redox ability [50]. Although the C₃N₄ valence band was lower than that of TiO₂, its band position of 2.12 eV was able to produce the hydroxyl radicals. Therefore, its holes exhibited good photocatalytic ability for MB degradation, as shown in the photocatalytic performance and PL experiments. In the Z-scheme heterojunction structure, the holes in the C₃N₄ valence band were sacrificed due to recombination with electrons from the TiO₂ conduction band, which negatively impacted photocatalysis. In contrast, the composite photocatalyst with a 1:1 component ratio is often required for the formation of ideal Z-Scheme heterojunction structure. However, the C₃N₄ and TiO₂ contents in TiO₂-C-C₃N₄ did not match, and the ratio of the photogenerated electrons and holes in the Z-scheme heterojunction structure was not in accordance. Thus, the TiO₂ electrons were not fully consumed and the transfer path of the Z-scheme heterojunction structure was interrupted rendering it impossible for TiO₂-C-C₃N₄ to operate according to the ideal model. Therefore, carrier recombination inside TiO₂ was responsible for the poor photocatalytic performance of TiO₂-C-C₃N₄.

TiO₂ was modified by graphitic carbon and C₃N₄ to synthesize type-II heterojunction TiO₂-C₃N₄ and all-solid-state Z-scheme heterojunction TiO₂-C-C₃N₄ semiconductor photocatalysts. In addition, the prepared photocatalysts were optimized and the TiO₂-5C₃N₄ and TiO₂-C-5C₃N₄ showed the best performance for their respective categories. The photocatalytic performance of TiO₂-C₃N₄ with type-II heterojunction structure was higher than that of TiO₂-C-C₃N₄ with all-solid-sate Z-scheme heterojunction structure. This may be because in the Z-Scheme heterojunction structure, the holes in the C₃N₄ valence band were transferred to the TiO₂ conduction band via the C medium so that the holes in the C₃N₄ valence band were wasted by following the mechanism of the all-solid Z-scheme heterojunction structure. In addition, TiO₂-C-C₃N₄ did not exhibit the ideal all-solid Z-scheme heterojunction structure. Thus, photogenerated electrons in TiO₂ cannot be fully consumed and their transfer path was interrupted. This study proposes a method based on the experimental results and internal mechanisms to determine the photocatalytic mechanism. By exploring the suitability of type-II heterojunction and all-solid-state Z-scheme heterojunction structures in the same degradation system, it was possible to design efficient photocatalysts for organic pollutant degradation according to different requirements.