A general nonaqueous sol-gel route to g-C3N4-coupling photocatalysts: the case of Z-scheme g-C3N4/TiO2 with enhanced photodegradation toward RhB under visible-light

The g-C3N4-coupling TiO2 photocatalysts with controllable particle size as well as the interface contact were prepared by a general nonaqueous sol-gel method. The structural and morphological features of g-C3N4/TiO2 were investigated through the X-ray diffraction, Fourier transformed infrared spectra, scanning electron microscopy and transmission electron microscopy, respectively. It is found the TiO2 nanoparticles with a size of 7.3 ± 1.6 nm are uniformly anchored on the surface of the g-C3N4 nanosheets in isolation. The photocatalytic properties of as-prepared g-C3N4/TiO2 were tested by degradation of Rhodamine B (RhB) under visible light, and an enhanced activity is observed. The mechanism of the enhanced activity was further investigated through N2 adsorption-desorption isotherms, UV-vis spectra, photoluminescence spectra, photoelectrochemical measurements, radical trapping experiments and X-ray photoelectron spectroscopy. Furthermore, the photocatalytic performances of obtained g-C3N4/TiO2 under sunlight were also evaluated in aspects of degradation efficiency and stability. The results indicate that the obtained g-C3N4/TiO2 is one promising photocatalyst for practical applications. The study of as-prepared g-C3N4/TiO2 also implies that the present method could be a general route of g-C3N4-coupling photocatalysts.


Results and Discussion
Structure and morphology of as-prepared g-C 3 N 4 /TiO 2 . The crystal structures of as-prepared samples were characterized by XRD, and the experimental data are shown in Fig. 1. For as-prepared pure g-C 3 N 4 , the curve exhibits a typical XRD pattern of g-C 3 N 4 , which consists of two characteristic peaks around 13.04° of (100) crystal plane, and 27.36° of (002) crystal plane [39][40][41] , respectively. All the peaks of pattern b match well with the (101), (103), (004), (112), (200), (105) and (211) reflections of anatase-type TiO 2 (PDF No. . The other patterns display the XRD curves of g-C 3 N 4 /TiO 2 . Comparing them with that of TiO 2 and g-C 3 N 4 , one can find that the major peaks are identical to that of TiO 2 , including the shape and position. It seems the relative amount of g-C 3 N 4 is very low so that only a small peak at 27.36°, the (002) reflection of g-C 3 N 4 , is observed, as shown in Fig. 1(b). The results indicate that the added g-C 3 N 4 does not influence the formation of TiO 2 . On the other hand, it is observed that the peaks belonging to TiO 2 possess an observed diffraction widening, so as to the overlapping of some adjacent peaks, such as the overlapping of (103), (004) and (112) peaks. These phenomena imply the TiO 2 ought to have a small particles size which is an important parameter determining the photocatalytic activity. Rietveld refinement was used to have further revealing of the structural characters of TiO 2 . The refined pattern together with the experimental data, Bragg position as well as difference curve are shown in Fig. S1. One can find that the black calculated pattern matches well with the experimental data and the difference curve only shows fluctuations, indicating that the calculated value is convinced. Through the refinement, the grain size of the TiO 2 is calculated to be 7.65 nm. Such the ultra-small particle size is identical to the reports of such the "benzyl alcohol route". Apart from the grain size, the other structural parameters including the space group, cell parameters, atom coordinate as well as microstrain are listed in Table S1.
Due to the overlapping and widening of the peaks, the XRD peaks of g-C 3 N 4 are undistinguished in the pattern of as-prepared g-C 3 N 4 /TiO 2 . In this case, the FTIR was carried out to further confirm the structure of the g-C 3 N 4 /TiO 2 , and the results are shown in Fig. 2. Compared with that of TiO 2 , a series of peaks in the range of 1700-1200 cm −1 is easily observed in the pattern of g-C 3 N 4 , and belongs to the typical stretching modes of CN heterocycles. These peaks are also can be found in the curves of g-C 3 N 4 /TiO 2 samples. Their intensity increases with the amount of g-C 3 N 4 . The results demonstrate that the structure of g-C 3 N 4 is remained after the growth of TiO 2 .  To observe the micro-morphology of as-prepared samples, FESEM was carried out. Figure 3(a) displays the morphology of g-C 3 N 4 . It can be seen that the g-C 3 N 4 is made up of stacking nanosheets whose thickness is easy to be measured and falls in tens nanometers. The surface morphology of TiO 2 and 5% g-C 3 N 4 /TiO 2 is shown in Fig. 3(b) and (c) which have the same magnification as Fig. 3(a). In Fig. 3(b), several submicron-size aggregations of TiO 2 nanoparticles are observed. Under the same magnification, the morphology of g-C 3 N 4 /TiO 2 is different from that of g-C 3 N 4 and TiO 2 . The surface of the composite is rough as that of TiO 2 , but the size of the composite is much larger than that of TiO 2 aggregations. For a better observation of the finer morphology, the surface of g-C 3 N 4 /TiO 2 is magnified and shown in Fig. 3(d). The finer structure consists of ultra-small particles. Combined the experimental section as well as the XRD results, it is believed that the ultra-small nanoparticles might belong to TiO 2 introduced by the benzyl alcohol reaction. Besides, the layered structure belonging to g-C 3 N 4 is also can be seen in the region marked by square in Fig. 3(c). The results imply that the morphology may result from that the TiO 2 nanoparticles are tightly anchored on the surface of g-C 3 N 4 . The SEM imagines of 2.5%, 7.5% and 10% g-C 3 N 4 /TiO 2 are shown in Fig. S2. The similar morphology can be observed in all these samples.
To verify the point that the TiO 2 is anchored on the surface of g-C 3 N 4 , element mapping of the g-C 3 N 4 /TiO 2 is carried out and its results are shown in Fig. 4. Figure 4(a) displays the SEM image of the g-C 3 N 4 /TiO 2 composite whose edge shows the layered structure. As shown in Fig. 4(b,c,d) and (e), N, C, Ti, O uniformly distribute on the region of the composite, even in the edge region, which fully proves the prediction about the distribution of g-C 3 N 4 and TiO 2 in the composite. In Fig. 4(f), it is also observed that the signal of Ti and O is much stronger than that of C and N. This phenomenon can be explained by the lower content of g-C 3 N 4 in the composite.
Due to the limited magnification of SEM, the morphology of the anchored TiO 2 particles is not clear. Hence, TEM were carried out to gain a better understanding of the morphological and structural features. Figure 5(a) and (b) display the TEM imagines of g-C 3 N 4 and 7.5% g-C 3 N 4 /TiO 2 , respectively. Comparing these two figures, the similar results with that of FESEM can be observed. The g-C 3 N 4 is made up of the stacking thin nanosheets whose surface is rather smooth. But it is clearly observed the surface of g-C 3 N 4 is full of tens thousands of ultrasmall nanoparticles. With the selected-area electron-diffraction (SAED), a series of clear diffraction rings corresponding to the reflections (101), (004), (200) and (105) of anatase-type TiO 2 is observed in the inset, which further reveals that the anchored nanoparticles are polycrystal anatase TiO 2 . Figure 5(c) exhibits the magnified TEM imagine of g-C 3 N 4 /TiO 2 . The well-define TiO 2 nanoparticles are randomly anchored on the surface of g-C 3 N 4 , and no hard aggregation is observed between the nanoparticles. The tight combination between g-C 3 N 4 and TiO 2 is favorable to charge transfer between these two semiconductors which possess the proper band structure. While the uniform and isolated distribution of the TiO 2 nanoparticles is beneficial for exposing more active site for the heterogeneous catalysis. The inset displays the HRTEM imagine of one isolated TiO 2 nanoparticle. The clear fringes make it easy to estimate the interplanar spacing which is 0.353 nm and corresponding to the (101) plane of anatase. On the other hand, since the anchored nanoparticles show a well-defined morphology, the grain size of these particles are obtained by measuring the size of five hundreds particles, and the results are shown in Fig. 5(d). The result, 7.3 ± 1.6 nm, matches well with refinement of XRD. Both of them indicate that the TiO 2 on the surface of g-C 3 N 4 /TiO 2 prepared through the "benzyl alcohol route" possesses an ultrasmall particle size.
With the XRD, FTIR, FESEM, element mapping and TEM toward the as-prepared samples, it is sure that g-C 3 N 4 is compatible with the "benzyl alcohol route" in the term of crystal structure and morphology. The addition of g-C 3 N 4 does not disturb the chemical reactions and crystallization in the nonaqueous sol-gel route based on benzyl alcohol. Moreover, the formed TiO 2 nanoparticles with ultrasmall grain size are uniformly anchored on the surface of g-C 3 N 4 without hard aggregations in a wide range. Such the scenario is exactly what we want and seldom reported before. And such the good combination of TiO 2 and g-C 3 N 4 can be attributed three main reasons: (1) the "benzyl alcohol route" gives the TiO 2 a very small particle size; (2) the electronegativity of the g-C 3 N 4 nanosheets endows a strong electrostatic adsorption of Ti 4+ , which leads to a further tight combination 42 ; (3) more importantly, g-C 3 N 4 nanosheets possesses very good compatible with the "benzyl alcohol route" like the GO does.
Photocatalytic properties of as-prepared g-C 3 N 4 /TiO 2 under visible light. It has been proved that g-C 3 N 4 is compatible with the "benzyl alcohol route" to prepare g-C 3 N 4 /TiO 2 nanocomposites from the aspects of structure and morphology. In this section, the photocatalytic properties of as-prepared g-C 3 N 4 /TiO 2 under visible light are evaluated. And RhB, a typical organic dye which is a common water pollutant and could cause long-term environmental toxicity and short-term public health damage [43][44][45] , is chosen as the model pollutant. Prior the irradiation under the light source, half hour dark treatment was carried out to realize adsorption equilibrium and eliminate the influence of adsorption on degradation. Figure 6 shows the evolution of degradation rate along with the time. All the curves have a similar linear downtrend which seems to be zero-order kinetic process. Hence, the data points after the dark treatment were fitted according to the zero-order kinetic process: where k is the degradation rate constant, and b represents the residual composition at 0 h. The k, b, and R 2 values of the fitted curves for all samples are listed in Table 1. Firstly, all the R 2 values higher than 0.9842 indicating the fitted curves match well with the obtained data, and the degradation surely follows the zero-order kinetic process. The insets of (b) and (c) correspond to the SAED pattern of 7.5% g-C 3 N 4 /TiO 2 and HRTEM of one isolated TiO 2 nanoparticle anchored on the surface of g-C 3 N 4 , respectively.
Secondly, the as-prepared g-C 3 N 4 /TiO 2 nanocomposites show higher degradation rate constant than both TiO 2 and g-C 3 N 4 , and even the commercial P25 photocatalyst, which demonstrates a high photocatalytic activity of as-prepared g-C 3 N 4 /TiO 2 nanocomposites. The compounding of g-C 3 N 4 and TiO 2 through combining g-C 3 N 4 and "benzyl alcohol route" is meaningful. Thirdly, the composite containing 7.5% g-C 3 N 4 displays the highest k value, which means 7.5% is the best ratio of g-C 3 N 4 in the presented preparation method. On the other hand, the b in the fitted equation can reflect the adsorption ability of the photocatalyst, because it represents the residual composition at 0 h and after dark treatment. From the listed b values in Table 1, the as-prepared g-C 3 N 4 /TiO 2 nanocomposites have the lower value than as-prepared g-C 3 N 4 and TiO 2 . Such the stronger adsorption ability can be understood from the FESEM and TEM imagines. The morphology that the ultrasmall TiO 2 nanoparticles uniformly anchored on the surface of g-C 3 N 4 prevents the aggregations of TiO 2 and increase the surface roughness of g-C 3 N 4 . But for the g-C 3 N 4 /TiO 2 with different content of g-C 3 N 4 , the values of b are random which indicates that the enhanced activity is not only resulted from the better adsorption. For example, the sample with 7.5% g-C 3 N 4 possesses the best degradation efficiency, but its b value is the highest among the other samples of g-C 3 N 4 / TiO 2 . Hence, it is believed that there are some other reasons leading to the enhanced activity apart from the better adsorption. In the introduction, it is predicted that the coupling of g-C 3 N 4 and wide-band semiconductor could result in a better utilization of visible light, and separation as well as transfer of photogenerated carries. All of them could enhance the visible-light driven photocatalysis activity.
Mechanisms of the enhanced photocatalytic activities of g-C 3 N 4 /TiO 2 . In this section, the mechanisms of the enhanced photocatalytic activities of g-C 3 N 4 /TiO 2 were studied from the adsorption, utilization of visible light, and charge transfer, respectively. In usual, a better adsorption is linked to a higher specific surface area 46 , so the N 2 adsorption-desorption experiments toward the as-prepared TiO 2 , g-C 3 N 4 and g-C 3 N 4 /TiO 2 were carried out. Since the content of g-C 3 N 4 is relative low, and there are four samples with different content, the 10% g-C 3 N 4 /TiO 2 with highest amount of g-C 3 N 4 was chosen as the example to make the results clearer and easier to be analyzed. The isotherms and corresponding pore size distribution curves are shown in Fig. 7. In the similar isotherms of TiO 2 and g-C 3 N 4 /TiO 2 , the evident hysteresis loops from 0.5 to 0.9 can be observed, indicating the presence of mesopores within the materials. This can be further verified from the pore size distribution curves. The TiO 2 and g-C 3 N 4 /TiO 2 show the pores centered at 7.211 and 6.15 nm, respectively. According to the shape of the hysteresis loops and the TEM results, the observed mesoporous is resulted from the accumulation of TiO 2 nanoparticles. The isotherm of g-C 3 N 4 also shows non-closed and porous character as shown in the Fig. 7 and its inset. But its specific surface area is much lower than that of TiO 2 and g-C 3 N 4 /TiO 2 . To be more specific, the value from BET measurement of TiO 2 , g-C 3 N 4 /TiO 2  and g-C 3 N 4 are 141, 166 and 21 m 2 g −1 , respectively. These results well prove the prediction made from the b values listed in Table 1. Figure 8 displays the UV-vis spectra of as-prepared samples. The curve of TiO 2 displays an absorption edge in ultraviolet region, which corresponds to the band gap of TiO 2 . But the as-prepared TiO 2 still shows some absorption in visible light. Such the absorption backs the visible-light driven photocatalytic activity of TiO 2 . Due to the narrower band gap of g-C 3 N 4 (2.67 eV), the edge absorption of g-C 3 N 4 arise at the region with longer wavelength than TiO 2 . Comparing the curves of g-C 3 N 4 /TiO 2 with that of TiO 2 , one can find a better absorption of visible light from 400 to 550 nm which matches well with the spectra of g-C 3 N 4 , and the absorption is increased with the content of g-C 3 N 4 . Hence, it is clear that g-C 3 N 4 enhances the visible-light utilization of TiO 2 , owing to the narrower band gap of 2.67 eV.
It is generally accepted that the photoluminescence (PL) spectra can be used to check the separation and transfer of photogenerated carries 47 . Due to the separation and transfer, the recombination of the photogenerated carries are decreased, which further leads to reduced PL of the coupling. However, for the coupling of g-C 3 N 4 and TiO 2 , such the method loses its function. Because the g-C 3 N 4 shows very strong PL which even can be observed by naked eyes, but PL intensity of TiO 2 is much lower. Hence, the composite shows stronger PL than TiO 2 , but lower PL than g-C 3 N 4 , as shown in Fig. S3. Such the result matches with other reported literatures 23,48,49 . In this situation, the photoelectrochemical measurement was carried out to verify the separation of the photogenerated carries. Figure 9(a) displays the transient photocurrent responses of TiO 2 , g-C 3 N 4 and their composite through typical switch on-off cycles [50][51][52] . It is observed all the materials show fast response and recovery. And the composite shows the highest current, followed by TiO 2 . The lowest current of g-C 3 N 4 may be attributed to the high recombination of the photogenerated carries which can be understood from Fig. S3. The highest photocurrent directly proves the separation of photogenerated carries in g-C 3 N 4 /TiO 2 . By the way, the electrochemical impedance spectroscopy (EIS) was carried out and its results are shown in Fig. 9(b). From the figure, the arc character of all samples can be observed. It is reported that the charge lifetime is related to the recombination resistance, and large semicircle radius of EIS curves implies large recombination resistance [53][54][55] . From the obtained EIS curves,  one can find that the g-C 3 N 4 /TiO 2 has the largest radius, while the g-C 3 N 4 possesses the smallest. The results match well with that of photocurrents. From the photoelectrochmical measurement, it is known that the combination of g-C 3 N 4 and TiO 2 increase the recombination resistance and lifetime of photogenerated carries, which is resulted from the charge separation.
On the other hand, it is known to us that the reactive species trapping experiment is effective in judgment of active species. Further step, the mode of carries transfer as well as separation could be indirectly studied. Hence, it was carried out to study the possible separation and transfer of photogenerated carries. In this work, ethylenediamine teracetic acid disodium salt (EDTA-2Na, 0.01 M), p-benzoquinone (BZQ, 0.001 M) and isopropanol (IPA, 0.02 M) were used as scavengers for photogenerated holes, superoxide anion radicals and hydroxyl radicals, respectively 56,57 . The removal rates of the dye at 7 h with different catalysts were used for comparison. Because the behaviors of the composite highly depend on its components, and the function of the conduction band and valence band directly influence the judgment of the charge transfer model, the behavior of two components, TiO 2 and g-C 3 N 4 , are firstly studied. Figure 10(a) displays the results of TiO 2 . Without any trapping reagent, the as-prepared TiO 2 shows the ability of breaking down the RhB dye under visible light. With addition of the sacrifice reagents, the removal rates of the dye decrease, which means photogenerated holes, superoxide anion radicals and hydroxyl radicals are produced and active during the degradation process. Especially the active photogenerated holes exclude the dye-sensitization mechanism of the degradation, because the hole is not involved in the dye-sensitization process 58 . However, it is generally accepted that the band gap of anatase-type TiO 2 is 3.2 eV which is too wide to produce photogenerated carries and consequent corresponding redox reactions. But the result of the trapping reagents indirectly proves the existence of aforementioned process. For these, we attribute these to the possibly existing oxygen vacancies in the TiO 2 nanoparticles. Nakamura et al. created oxygen vacancies in TiO 2 by plasma treatment to increase its photocatalytic activities under visible light, and brought out that the oxygen vacancies facilitate visible light absorption by generating discrete state below the conduction band of TiO 2 59 . Ihara et al. also reported that oxygen vacancies can be easily created in the grain boundaries 60 . In this work, the TiO 2 nanoparticles anchored on the surface of g-C 3 N 4 have a very small grain size, and they are easy to form varies defects on their surface, naturally include the oxygen vacancy. It is believed that the oxygen-related defects including the oxygen vacancy can be indirectly observed in the O1s XPS spectra in the form of absorbed oxygen ion. To verify the existence of the absorbed oxygen ion, XPS toward TiO 2 and 7.5% g-C 3 N 4 /TiO 2 were carried out. The magnified O1s XPS spectra are shown in Fig. S4(a) and (b), respectively. The peak with lower binding energy around 530 eV is associated to the oxygen atom in the crystal lattice, while the peak with higher binding energy is assigned to the absorbed oxygen ion 61 . It proves that there surely exist the oxygen related defects including the oxygen vacancy. And it also forebodes that the TiO 2 in the g-C 3 N 4 /TiO 2 could also show the similar photocatalytic behaviors to the pure TiO 2 . Based on the oxygen vacancy introducing defect level and the results of sacrifice reagent, the degradation mechanism of as-prepared TiO 2 under visible light is obtained as shown in Fig. 10(b).
Unlike wide-band TiO 2 , g-C 3 N 4 , a visible-light semiconductor with a narrow band gap of 2.67 eV, is easy to be excited by visible light. Its result of trapping experiments is shown in Fig. 10(c). Two unusual phenomena are observed, including that the addition of hole trapping reagent (EDTA-2Na) intensively increases the removal rate, and the hydroxyl radical arises and show its function in the degradation process. The active species in the degradation process can be predicted through comparing the top of valence band (TVB) potential and bottom of conduction band (BCB) potential with the standard redox potential (SRP) of corresponding reactions. Since the E BCB and E TVB of g-C 3 N 4 have been widely studied, − 1.10 eV and 1.57 eV are adopted to be the E BCB and E TVB , respectively 16 . Also, it is known to us that the SRP of O 2 /·O 2 − and OH − /·OH are − 0.046 and + 2.7 V vs. NHE, respectively 12 . Due to the E BCB of g-C 3 N 4 is more negative than the SRP of O 2 /·O 2 − , the photogenerated electrons can react with O 2 to form ·O 2 − . But the E TVB of g-C 3 N 4 is lower than the SRP of OH − /·OH, so ·OH can not be formed through the reaction of photogenerated holes and OH − . Although the photogenerated holes can not produce ·OH, it is believed the generated holes still can oxidize the pollutant. So, it is generally accepted that the effective species during the photocatalysis reaction of g-C 3 N 4 are holes, and ·O 2 − 62,63 . According to the result shown in Fig. 10(c), the prediction can not explain the behavior of the g-C 3 N 4 for the degradation of RhB. Therefore, the effect of the conduction band and valence band of g-C 3 N 4 need further analysis. As we know, the potential of TVB and BCB decides the redox ability of the photogenerated carries. More negative BCB means stronger reduction photogenerated electrons, and more positive TVB means stronger oxidation photogenerated holes. The g-C 3 N 4 possesses very negative BCB, but its TVB is weak. Hence, the oxidation ability of the photogenerated hole is weak which makes the hole show few effects on the RhB molecules just like the hole can not react with OH − to form ·OH. Moreover, the excess photogenerated holes would recombine with the photogenerated electrons which can form the effective ·O 2 − species. Especially, the g-C 3 N 4 always shows a strong blue photoluminescence, indicating a strong recombination of the photogenerated carries. With the addition of EDTA-2Na, the trapping reagent of photogenerated holes, the inactive holes are consumed, and the reduced amount of generated holes decreases the recombination possibility of photogenerated carries. That is to say, more photogenerated electrons are kept on the conduction band, so the removal rate is increased. On the other hand, due to the strong reduction of the photogenerated electrons, the formed ·O 2 − also possesses very strong reduction which could leads to the further reactions and formation of ·OH radicals through an intermediate of ·OOH 64,65 . That is why the addition of IPA, ·OH trapping reagent, causes the lower removal rate. Hence, the degradation mechanism of g-C 3 N 4 toward RhB solution is summarized in Fig. 10(d). From the trapping experiments toward TiO 2 and g-C 3 N 4 , it is found that the most difference between them during the degradation process under visible light in the aspect of effective species is that the photogenerated holes in the valence band of TiO 2 is active, but inactive for g-C 3 N 4 . Such the conclusion is different from the common reported literatures in which the difference is thought to be whether there are OH − species according the compassion of calculated potent ion of BCB and TVB with the SRP of corresponding reactions 21,49,63 . At the same time, the conclusion is very important to the analysis of charge transfer model and mechanism of enhanced photocatalytic activity of g-C 3 N 4 /TiO 2 nanocomposite. Hence, the analysis of trapping experiments also should base on the experiments of its components rather than simple theoretical comparisons.
The active specie trapping experiments toward 7.5% g-C 3 N 4 /TiO 2 are carried out, and the results are shown in Fig. 11. General speaking, all three tested species including photogenerated hole, superoxide anion radical and hydroxyl radical are active during the photocatalysis process. Comparing with Fig. 10(a) and (c), one can find that the active species of the composite are different from that of g-C 3 N 4 , but similar to that of TiO 2 . Further step, the removal rate are normalized according to the removal rate with no additive. For pure TiO 2 , the normalized removal rate with BZQ, EDTA-2Na and IPA are 22.91%, 13.60%, and 26.59%, respectively. For 7.5% g-C 3 N 4 /TiO 2 , the normalized removal rate with BZQ, EDTA-2Na and IPA are 18.90%, 4.48%, and 38.45%, respectively. It is clear that the experimental data between TiO 2 and 7.5% g-C 3 N 4 /TiO 2 are distinguished, and it also could not be the simple superposition of g-C 3 N 4 and TiO 2 . Especially for the removal rate with EDTA-2Na, the photogenerated hole is effective for TiO 2 , and inactive for g-C 3 N 4 . While for their composite it shows stronger effect than that for TiO 2 . Hence, the degradation model of g-C 3 N 4 /TiO 2 is different from their components, which is resulted from the separation and transfer of the photogenerated carries.
On the other hands, according to the degradation mechanism of TiO 2 ( Fig. 10(b)) and g-C 3 N 4 ( Fig. 10(d)), the two possible separation and transfer model of carries for g-C 3 N 4 /TiO 2 nanocomposite are shown in Fig. 12(a) and (b), which correspond to typical heterojunction and Z-scheme charge transfer models, respectively. Comparing these two models, one can find that the active species in the Z-scheme model are photogenerated hole, superoxide anion radical and hydroxyl radical, but the active species of heterojunction-type model only involves superoxide anion radical. For the heterojunction-type model, although it is not sure whether the superoxide anion radicals generated in the conduction band of TiO 2 can produce hydroxyl radical, it is certain that the holes formed in the valence band of g-C 3 N 4 is inactive. Comparing this certain difference between the two models with the experimental data, it can be concluded that the separation and transfer model of as-prepared g-C 3 N 4 / TiO 2 is Z-scheme. Through the Z-scheme, the different behaviors of g-C 3 N 4 /TiO 2 and TiO 2 can be explained. The superoxide anion radical is generated on the conduction band of g-C 3 N 4 whose E BCB is more negative than that of TiO 2 . With stronger reduction, the effect of superoxide anion radical is enhanced compared with TiO 2 . Although the valence band is the place forming the active holes for both TiO 2 and g-C 3 N 4 /TiO 2 , there is a lower possibility of carries recombination of g-C 3 N 4 /TiO 2 than TiO 2 due to the Z-scheme charge transfer in g-C 3 N 4 / TiO 2 . For hydroxyl radical, it is formed through the photogenerated holes on the valence band of TiO 2 . Since the photogenerated holes directly consumed by the dye molecules in g-C 3 N 4 /TiO 2 , the residual holes for generation of hydroxyl radical are decreased. Hence, the effect of hydroxyl radical is decreased for g-C 3 N 4 /TiO 2. Through the trapping experiments, it proves the Z-scheme charge transfer between g-C 3 N 4 and TiO 2 in g-C 3 N 4 /TiO 2 . And the enhanced activity with the Z-scheme charge transfer can be attributed to the consequent stronger reduction ability of the material, and effective charge separation. Hence, the enhanced photocatalytic activity of g-C 3 N 4 /TiO 2 under visible light is resulted from the increased adsorption, better utilization of visible light, Z-scheme charge transfer as well as consequent stronger reduction ability toward the dye molecules.
Photocatalytic performances of as-prepared g-C 3 N 4 /TiO 2 under sunlight. In order to test the potential of g-C 3 N 4 /TiO 2 in practical application, the photocatalytic performances of as-prepared g-C 3 N 4 /TiO 2 nanocomposites under sunlight are tested. Because the 7.5% is the optimal content of g-C 3 N 4 , 7.5% g-C 3 N 4 /TiO 2 was chosen as the object. Figure 13 displays the evolution of degradation rate along with the time. It is observed that the data points in the blank group fluctuates around 1, indicating that the degradation caused by the sunlight can be neglected during the test. With the addition of 7.5% g-C 3 N 4 /TiO 2 , the concentration of the RhB is decreased rapidly. It only takes 40 min to remove 96.37% RhB. 99.5% RhB was broken down at 1 h. Such the removal efficiency is about ten times faster than that under visible light. On the other hand, de-ethylation of RhB is always observed in its photodegradation, and the de-ethylation process could lead to the shift of absorption peak from 550 to 500 nm, and consequent decrease of absorbance at its initial position of absorption peak 66 . But the de-ethylation only takes a small part of the whole of degradation process, and double time is usually needed. In this term, the UV-vis spectra of the centrifuged solutions were tested and shown in Fig. 13(b). It is observed that the strong absorption peak has no shift during the degradation reaction. That is to say the rapid degradation of RhB pollutant is not resulted from the de-ethylation. Hence, the as-prepared 7.5% g-C 3 N 4 /TiO 2 certainly possesses quite high photocatalytic activity. And the inset shows the photo of the centrifuged solutions. The color of the solution fades along with the time, and turns to be colorless and transparent solution. While the higher efficiency can be attributed to the different light resource: the first, the sunlight is much stronger than the 24 W lamp used as the visible-light resource; there are some ultraviolet in the sunlight, which is beneficial for exciting more carries for TiO 2 component.
As we know, the photocatalytic stability of the catalysts is one of the most important parameter evaluating the possibility of practical application. Hence the degradation evolution in five cycles of degradation tests under sunlight was recorded as shown in Fig. 14. Unexpectedly, the photocatalyst not only keep a good stability, but also has a better efficiency with the increase of the cycle time. The degradation rate reaches higher than 99% for all the cycles at 60 min. At 25 min, the degradation rate reaches 83.77%, 89.55%, 94.48%, 97.40% and 99.02% for first, second, third, fourth and fifth cycle, respectively. Because the property is always associated with the microstructure of the catalysts, the TEM toward the catalyst after the 5 cycles was carried out, and shown in Fig. S5. In the Fig. S5, the composite of the sheet-like g-C 3 N 4 and ultrasmall TiO 2 nanoparticles are still observed. But the amount of the TiO 2 nanoparticles loaded on the surface of g-C 3 N 4 is much smaller than the primary sample shown in Fig. 5(b). The degradation efficiency is increased with the decreased amount of TiO 2 loaded on the surface of g-C 3 N 4 according to the results of TEM and stability tests. And Fig. 6 and Table 1 have proved that 7.5% is the best ration amount 0, 2.5%, 5.0%, 7.5% and 10.0%. Hence, the contradiction arises. Such the unusual phenomenon might be explained through the presented experiment details. In the preparation experiment, TiO 2 is the matrix, and its amount is constant. The g-C 3 N 4 is used as enhancer, and its amount is calculated according to the mass ratio of g-C 3 N 4 /TiO 2 . Such the design of the experiment ignores that although the g-C 3 N 4 is compatible with benzyl alcohol, its dispersed amount is limited. The amount of g-C 3 N 4 added in the preparation of 7.5% g-C 3 N 4 / TiO 2 could be the maximum. With further addition of g-C 3 N 4 , it stacks with each other, and the exposed surface is limited for TiO 2 to attach. Hence, 7.5% is the best in the projecting experiments. But the optimal ratio should be much higher than 7.5% so that the catalyst's efficiency increases along with smaller loaded TiO 2 nanoparticles. From this point of view, it suggests that the g-C 3 N 4 should be used as the matrix or the baseline to calculate the amount of the other components. And the limited dispersing amount of g-C 3 N 4 can be improved through decrease the thick of the g-C 3 N 4 nanosheets, which also could make the "benzeyl alcohol route" more economic and the products more active. Anyhow, it has proven g-C 3 N 4 is compatibility with the benzyl-alcohol-based nonaqueous sol-gel route, and the combination could be a general route to g-C 3 N 4 -coupling photocatalysts with controlled morphology and enhanced photocatalytic activity.

Conclusions
Through the combination of g-C 3 N 4 and classical nonaqueous sol-gel route based on benzyl alcohol, g-C 3 N 4 /TiO 2 coupling photocatalysts were obtained. Owing to the contribution of benzyl-alcohol-based nonaqueous sol-gel route, the TiO 2 on the surface of g-C 3 N 4 shows ultrasmall grain size. Due to the excellent compatibility of g-C 3 N 4 and the benzyl alcohol route, the TiO 2 nanoparticles are uniformly distributed on the surface of g-C 3 N 4 . For the present experimental parameters, the optimal mass ratio of g-C 3 N 4 /TiO 2 is 7.5%. And the enhanced photocatalytic activity under visible light is attributed to the Z-scheme heterojunction, and consequent better absorption  of visible light, enhanced reduction ability as well as decreased recombination. The as-prepared g-C 3 N 4 /TiO 2 also has high efficiency and good cycle performance under sunlight, indicating its practical application. Through the detailed study of as-prepared g-C 3 N 4 /TiO 2 and considering that "benzyl alcohol route" is one general route to numerous size-controlled photocatalysts, it is believed that such the nonaqueous sol-gel route could be a general method to g-C 3 N 4 -coupling photocatalysts with desired morphology and contact interface.

Methods
Materials and synthesis. All the chemical reagents used in the experiments were obtained from commercial sources as guaranteed-grade reagents and used without further purification.
Well-defined g-C 3 N 4 nanostructure was synthesized through thermal condensation of cyanuric acidmelamine complex 67 . Briefly, 6.45 g grinded cyanuric acid was added to 100 mL ethanol with stirring. 6.3 g melamine was then grinded and added to the above suspension. The mixture was continuing stirred for 8 h, and then dried at 60 °C. The obtained white powder was subsequently calcined at 550 °C under N 2 atmosphere for 4 h with a heating rate of 2.3 °C min −1 to gain the final g-C 3 N 4 powder.
The g-C 3 N 4 /TiO 2 nanocomposites were in-situ prepared on the pre-synthesized nanostructured g-C 3 N 4 through the "benzyl alcohol route". Firstly, 10 mL xylol was poured into a closed conical flask, and then 0.69 mL TiCl 4 was dissolved with magnetic stirring. Five minutes later, 66 mL benzyl alcohol was gently poured into the system. After stirring for another five minutes, the calculated amount of pre-synthesized g-C 3 N 4 was further grinded and added to the titanium precursor solution. The amount of g-C 3 N 4 was calculated according to mass ratios of g-C 3 N 4 /TiO 2 = 0%, 2.5%, 5.0%, 7.5% as well as 10%, respectively. The corresponding samples were named as TiO 2 and x% g-C 3 N 4 /TiO 2 (x = 2.5, 5.0, 7.5, 10.0), respectively. To gain a better dispersing of g-C 3 N 4 and promote the adsorption of Ti 4+ on the g-C 3 N 4 surface with electronegativity, the mixtures were treated with ultrasound for 2 h. Later, with further stirring for 30 s, 70 mL of the homogeneous precursors were transferred into the Teflon-lined stainless steel autoclaves with a capacity of 83 mL and reacted at 180 °C for only 4 h, respectively. The autoclaves were cooled down to room temperature, and the resulting precipitates were washed with ethanol for four times and dried at 60 °C overnight.
Characterization of as-prepared samples. X-ray diffraction (XRD, Rigaku D/MAX-3B powder diffractometer) with a copper target and K α radiation (λ = 1.54056 Å) was used for the phase identification, where the diffracted X-ray intensities were recorded as a function of 2θ. The samples were scanned from 10° to 80° (2θ ) in steps of 0.02°. Fourier transformed infrared (FTIR) spectra were recorded on AVATAR 360 FT-IR spectrophotometer. The microstructures of the samples were tested on Hitachi S-4800 field emission electron microscopy (FESEM) with the energy dispersive X-ray (EDX) device. The transmission electron micrographs (TEM) were obtained with a Zeiss EM 912 Ω instrument at an acceleration voltage of 120 kV, while high-resolution transmission electron microscopy (HRTEM) characterizations were carried out using a Philips CM200-FEG microscope (200 kV, C s = 1.35 mm). The samples used for TEM were prepared by dispersing the products in ethanol with ultrasound treatment, and the dispersion was then dropped on carbon-copper grids. The nitrogen adsorption-desorption technique was measured at 77.3 K with Autosorb iQ Station 1. Prior to the measurement, the sample was degassed at 300 °C for 28.4 h under a vacuum situation. UV-vis measurements were made with a Hitachi U4100 spectrophotometer with a wavelength range between 300 to 700 nm. Photoluminescence (PL) spectra were made with a FLS 980 spectrophotometer of Edinburgh Instruments. X-ray photoelectron spectroscopy (XPS) was carried out at room temperature in ESCALAB 250 system. During XPS analysis, an Al Kα X-ray beam was adopted as the excitation source and the vacuum pressure of the instrument chamber was 1 × 10 −7 Pa as read on the panel. Measured spectra were decomposed into Gaussian components by a least-square fitting method. Bonding energy was calibrated with reference to C1s peak (285.0 eV).
Evaluation of the photocatalytic activity of the samples. Since the strong absorption peaks in the visible light, which is favorable the direct evaluation, Rhodamine B (RhB, C 28 H 31 ClN 2 O 3 ) is chosen as the model pollutant and indicator of the photocatalytic activity in this work. The photocatalytic properties under visible light of the prepared sample were evaluated with a 24 W visible-light lamp as light source, and the light below 400 nm was removed using a glass filter. In a typical degradation test, 40 mg of the as-synthesized sample was added into a quartz beaker with addition of 50 mL RhB solution (10 mg/L). The distance between the lamp and the solution surface was 6 cm. The dispersion (6 mL) was extracted and centrifuged to separate the catalyst and dye solution at 8500 rev. per min for 10 min at different intervals. The changed RhB concentration of the centrifuged solutions was recorded using a UV-1800 spectrophotometer according to Beer-Lambert's Law. The UV-vis spectra of the centrifuged solution were also measured using the UV-1800 spectrophotometer from 300 to 600 nm. The photocatalytic performances under sunlight of the g-C 3 N 4 /TiO 2 were tested through the similar procedures. For the photocatalytic performance of the g-C 3 N 4 /TiO 2 under sunlight, except that sunlight was directly used as the light source, all other processes are similar with that of visible-light tests. And all the sunlight tests were carried out in sunny days in March 2016 between 11 am and 14 pm in Kunming, China.
Photoelectrochemical measurements of the samples. The photoelectrochemical measurements were performed on CHI660E (Chenhua Instrument, Shanghai, China) in a neutral aqueous system (0.1 M Na 2 SO 4 ) using a three-electrode system. A platinum (Pt) plate and a saturated calomel electrode (SCE, 0.2415 V vs SHE) were used as the counter and the reference electrode, respectively. Light source is provided by 8 W UV lamp (Spectroline EA-180/FE). The working electrode was prepared as follows: 20 mg of the sample was dispersed in 20 mL ethanol with ultrasound for 30 min; 5 μ L of the mixture was dropt to an ITO glass (1.25 × 2.5 cm 2 ); then the electrode was obtained after drying at 60 °C for 1 h and sintering at 300 °C for 1 h.