Noble Metal-Free TiO2-Coated Carbon Nitride Layers for Enhanced Visible Light-Driven Photocatalysis.

Composites of g-C3N4/TiO2 were one-step prepared using electron impact with dielectric barrier discharge (DBD) plasma as the electron source. Due to the low operation temperature, TiO2 by the plasma method shows higher specific surface area and smaller particle size than that prepared via conventional calcination. Most interestingly, electron impact produces more oxygen vacancy on TiO2, which facilitates the recombination and formation of heterostructure of g-C3N4/TiO2. The composites have higher light absorption capacity and lower charge recombination efficiency. g-C3N4/TiO2 by plasma can produce hydrogen at a rate of 219.9 μmol·g−1·h−1 and completely degrade Rhodamine B (20mg·L−1) in two hours. Its hydrogen production rates were 3 and 1.5 times higher than that by calcination and pure g-C3N4, respectively. Electron impact, ozone and oxygen radical also play key roles in plasma preparation. Plasma has unique advantages in metal oxides defect engineering and the preparation of heterostructured composites with prospective applications as photocatalysts for pollutant degradation and water splitting.


Introduction
Photocatalytic technology, a promising strategy for addressing energy shortages and environmental pollution, is important for the production of hydrogen via water splitting and the degradation of organic pollutants [1][2][3][4]. TiO 2 , discovered by Fujishima in 1972 [5], is the most widely studied and applied semiconductor photocatalyst and is non-toxic, stable, and cheap [6][7][8]. Many preparation methods of TiO 2 and TiO 2 composite have been used, such as the sol-gel method [9], solvothermal method [10], and chemical vapor deposition method [11] and so on. However, a green, simple, cheap and energy-efficient way for catalyst preparation is still necessary.
Non-thermal plasma, which has relatively low bulk temperature and extremely high electron temperature, has excellent advantages in preparing catalysts [12][13][14]. The catalyst can be prepared quickly without serious agglomeration due to the low temperature and high energy of plasma. On the other hand, the nucleation and crystallization process of the catalyst is very unique in plasma [13]. Furthermore, the catalysts prepared using the plasma method have small particles, strong interaction and specific structures [15,16]. However, TiO 2 alone can only absorb ultraviolet light (only 4% of solar energy), even if the light absorption properties of TiO 2 are improved. In addition, higher photo-generated charge recombination efficiency also affects its photocatalytic activity [17].
Strategies have been proposed to increase the photocatalytic activity of TiO 2 under visible light. For example, compounding TiO 2 with a narrow bandgap semiconductor catalyst can enhance its absorption of visible light and construct a special heterostructure [18,19]. Doping elements can reduce
Synthesis of TiO 2 and g-C 3 N 4 /TiO 2 g-C 3 N 4 was dispersed by ultrasound in 20 mL of ethanol. We added 3ml TBT slowly into the as-prepared g-C 3 N 4 suspension under adequate stirring. Then, the as-prepared hybrid suspensions were stand still for 24 h. The sample was filtered after 24 h. and drying at 80 • C. The samples were divided into two portions.
One portion was treated by DBD plasma to obtain g-C 3 N 4 /TiO 2 composites according to the procedure reported in previous work [32,33]. Figure S1a shows the DBD device. The plasma was generated by the high voltage generator that can provide a sinusoidal waveform at 22 kHz with a voltage range of 0 to 30 kV. There are two quartz plates and a quartz ring between the two electrodes, and the sample was placed in the quartz ring. The gas atmosphere of the DBD reactor was air. The average power and average voltage of DBD during catalyst preparation were 200 W and 100 V, respectively. One-time plasma operation proceeded for 3 min to restrict the heat effect, followed by manually stirring to expose the untreated samples outside. The operation was repeated 20 times, until total plasma treatment time was 1 h. As shown in Figure S1b, the infrared (IR) image taken by the IR camera (Ircon, 100PHT, Everett, WA, USA) shown that the temperature of DBD plasma was below 106 • C. Finally, the obtained samples were denoted as TCNX-D (X = 10, 30, 50, 70, 90). X is equal to the weight ratio of g-C 3 N 4 in the composites. D represents the samples prepared by DBD plasma. Figure S2 shows the schematic illustration of preparation of g-C 3 N 4 /TiO 2 composite.
For comparison, another portion of the sample was calcined 450 • C at a rate of 5 • C/min for 2 h and designated as TCN50-C (50 wt% g-C 3 N 4 /TiO 2 ).
The preparation method of pure TiO 2 is the same as the above method. The samples obtained were denoted as TiO 2 -D and TiO 2 -C according to the different preparation methods.

Characterization
A Rigaku D/Max-2500 V/PC diffractometer with Cu Kα1 radiation (Cu Kα1 α = 0.154 nm, 40 kV, 40 mA, 8 • ·min −1 , Rigaku, Tokyo, Japan) was used to analyze the crystal phase of samples. Fourier transform infrared (FTIR) spectra was analyzed by a Bruker Alpha FTIR-attenuated total reflection (ATR) instrument (Bruker, Karlsruhe, Germany). A Biaode SSA-7000 analyzer (Biaode Electronic Technology Ltd., Beijing, China) was applied to determine specific surface area and pore size. Thermogravimetry analysis was conducted on Perkin-Elmer TGA/DTA thermo-gravimetric analyzer (Waltham, MA, USA) in O 2 atmosphere. The temperature range was 20-800 • C. An ULVAC-PHI-5000versaprobe instrument (Tokyo, Japan) was used to obtain X-ray photoelectron spectra. The adventitious carbon C1s used for element correction is located at 284.8 eV. Electron paramagnetic resonance measurement was obtained on a Bruker A300 spectrometer (Karlsruhe, Germany) at room temperature. The ultraviolet-visible (UV-vis) diffuse-reflectance spectroscopy was recorded by a UV-2600 spectrophotometer (Shimadzu, Kyoto, Japan). Photoluminescence (PL) spectroscopy was analyzed on Hitachi F-4600 spectrometer (Tokyo, Japan) with excitation at 350 nm. Scanning electron microscopy (SEM) and energy dispersive X-ray spectroscopy were conducted on Hitachi S4800 instrument (Tokyo, Japan). High-resolution transmission electron microscopy (HRTEM) was obtained using a JEM-2100F instrument (JEOL, Tokyo, Japan).

RhB degradation
We mixed 50 mg samples and 100 mL RhB (20 mg·L −1 ) solution uniformly in continuous stirring. First the mixture was stirred without light for 0.5 h to obtain the balance of adsorption and desorption. Visible light was provided by Xenon lamp (HSX-F300, 300 W) with a 420 nm cut filter. The illumination power was 100 mW/cm 2 . The light source was placed about 10 cm above the RhB solution. In the experiment, 4 mL solution was taken every 20 min for RhB degradation rate test. The degradation rate of the obtained RhB solution was determined using a UV-2600 instrument.

Hydrogen Generation
The H 2 evolution experiment was conducted on the PerfectLight 3AG instrument (Beijing, China). The amount of H 2 was determined by online gas chromatography. 100 mg sample was added to 70 mL of solution containing 10% triethanolamine and 133 µL of H 2 PtCl 6 (0.5 wt%). The mixed solution was placed in 100 mL closed glass container. The 300 W xenon lamp was equipped with total reflection and 420 nm filter to provide full-spectrum light and visible light, respectively. The light source was placed 10 cm above the suspension. After full-spectrum irradiation for 3 h in vacuum, the Pt was loaded on the samples. Then the hydrogen production reaction started, and the content of H 2 in the system was measured by online gas chromatography per hour.

Photoelectrochemical Investigation
The photocurrent density was conducted at the LK2010 electrochemical system. It is a traditional three-electrode system, tin oxide mixed with fluorine (FTO) conductive glass loaded with sample as working electrode. The working electrode was made as follows: 50 mg of the sample and 10 µL of nafion (5%) were added to 500 µL absolute ethanol, 200 µL of the suspension was loaded on FTO glass after mixing evenly. Then the FTO working electrode was obtained by drying at room temperature for 12 h. The reference electrode and counter electrode were the saturated Hg/HgO electrode and Pt wire, respectively. The 0.1M Na 2 SO 4 solution was used as electrolyte. Visible light was provided by 300 W Xe lamp equipped with 420 nm filter. At the beginning of the test, we fully introduced N 2 into the electrolyte to remove dissolved oxygen. The experimental potential is 0.4 eV, which is the optimal value by testing (see Figure S3).

Physico-Chemical Properties
The crystal phases of g-C 3 N 4 , TiO 2 , and g-C 3 N 4 /TiO 2 composites were analyzed by using wide-angle X-ray diffraction (XRD). As shown in Figure 1a, g-C 3 N 4 exhibits (100) and (002) planes at 13.2 • and 27.6 • , which correspond to tri-s-triazine structure and stacking of the conjugated aromatic system, respectively [34]. The as-prepared TiO 2 present the diffraction peaks at 25.  [35]. It indicated TiO 2 was successfully prepared by DBD plasma. As the proportion of g-C 3 N 4 increased in g-C 3 N 4 /TiO 2 composites, the intensities of g-C 3 N 4 peaks increased gradually. However, the no peak position shift of TiO 2 indicates that g-C 3 N 4 has no influence on the crystal structure of TiO 2 . Moreover, compared with samples (TiO 2 and g-C 3 N 4 /TiO 2 composites) prepared by high-temperature calcination, the samples by DBD plasma have lower peak intensity and wider peak width. On the basis of the full width at half maximum (FWHM) of (101) crystal plane and Debye-Scherrer equation, the average crystalline sizes of TiO 2 in the g-C 3 N 4 /TiO 2 composites were calculated and are listed in Table 1 [36]. The crystalline sizes of TiO 2 in the composites decreased gradually as g-C 3 N 4 increased. It is worth noting that TiO 2 has smaller crystallite size than that by the calcination method. The relative low temperature of plasma is the main reason restricting the agglomeration of particles. Nanomaterials 2020, 10, x FOR PEER REVIEW 4 of 15

Physico-Chemical Properties
The crystal phases of g-C3N4, TiO2, and g-C3N4/TiO2 composites were analyzed by using wideangle X-ray diffraction (XRD). As shown in Figure 1a, g-C3N4 exhibits (100) and (002) planes at 13.2° and 27.6°, which correspond to tri-s-triazine structure and stacking of the conjugated aromatic system, respectively [34]. The as-prepared TiO2 present the diffraction peaks at 25.1°, 37.7°, 47.8°, 54.0°, 55.1° and 62.5°, which corresponded to the (101), (004), (200), (105), (211) and (204) crystal planes of anatase TiO2 (JCPDS 71-1166), respectively [35]. It indicated TiO2 was successfully prepared by DBD plasma. As the proportion of g-C3N4 increased in g-C3N4/TiO2 composites, the intensities of g-C3N4 peaks increased gradually. However, the no peak position shift of TiO2 indicates that g-C3N4 has no influence on the crystal structure of TiO2. Moreover, compared with samples (TiO2 and g-C3N4/TiO2 composites) prepared by high-temperature calcination, the samples by DBD plasma have lower peak intensity and wider peak width. On the basis of the full width at half maximum (FWHM) of (101) crystal plane and Debye-Scherrer equation, the average crystalline sizes of TiO2 in the g-C3N4/TiO2 composites were calculated and are listed in Table 1 [36]. The crystalline sizes of TiO2 in the composites decreased gradually as g-C3N4 increased. It is worth noting that TiO2 has smaller crystallite size than that by the calcination method. The relative low temperature of plasma is the main reason restricting the agglomeration of particles.  Fourier transform infrared spectra (FTIR) of g-C3N4, TiO2, and g-C3N4/TiO2 composites are shown in Figure 1b. It reveals the composition and chemical bonding of samples. For pure g-C3N4, the peak of 807 cm −1 is due to the tri-s-triazine unit structure, and four intense bands in the region 1240-1640 cm −1 are attributed to the stretching of the C-N heterocycle in g-C3N4 [37]. For TiO2-C and TiO2-D, they showed similar peaks, and the shoulder bands between 3000-3500 cm −1 can be attributed to -OH stretching vibration [38]. The shoulder bands near 3200 cm −1 in TCN50 composites are  Fourier transform infrared spectra (FTIR) of g-C 3 N 4 , TiO 2 , and g-C 3 N 4 /TiO 2 composites are shown in Figure 1b. It reveals the composition and chemical bonding of samples. For pure g-C 3 N 4 , the peak of 807 cm −1 is due to the tri-s-triazine unit structure, and four intense bands in the region 1240-1640 cm −1 are attributed to the stretching of the C-N heterocycle in g-C 3 N 4 [37]. For TiO 2 -C and TiO 2 -D, they showed similar peaks, and the shoulder bands between 3000-3500 cm −1 can be attributed to -OH stretching vibration [38]. The shoulder bands near 3200 cm −1 in TCN50 composites are contributed by the N-H stretching vibration modes [37,39]. TCN50 composite exhibited both the g-C 3 N 4 and TiO 2 characteristic peaks. This shows that g-C 3 N 4 /TiO 2 composites were directly prepared by DBD plasma and is agreement with XRD analysis results. During plasma process, high energy electron bombard samples. It can be deduced that the surface hydroxyl and amino groups enhance interaction between TiO 2 and g-C 3 N 4 [40]. The strong interfacial connection can be used as the channel for charge conduction to improve charge separation efficiency.
Nitrogen adsorption-desorption isotherms and pore-size distribution curves of g-C 3 N 4 , TiO 2 and TCN50 samples are shown in Figure S4. It is clearly observed that samples prepared by plasma have larger N 2 adsorption capacity. The specific data are listed in Table 1. The specific surface area of TiO 2 -D (64.56 m 2 /g) is approximately 2.3 times that of TiO 2 -C (28.18 m 2 /g), which can be ascribed to the low agglomeration of the sample's particles prepared by low-temperature plasma. The specific surface area of g-C 3 N 4 merely is 7.13 m 2 /g. As shown in Table S1, the specific surface area and pore volume of g-C 3 N 4 /TiO 2 composites decreased and the average pore radius increased gradually as the proportion of g-C 3 N 4 increased. Compared with TCN50-C, TCN50-D has higher specific surface area, pore volume, and smaller average pore radius, which indicates that the surface of the plasma-treated sample can expose more active sites. In addition to the low temperature, the repulsion between the electrons attached to the particles prevents the agglomerating of particles [12], resulting in the excellent dispersion of TiO 2 on g-C 3 N 4 to enhance photocatalytic activity. Thermogravimetry (TG) was used to analyze thermal stability of the samples. As shown in Figure S5, the weight of TiO 2 prepared by DBD plasma only reduced by 2%, which indicated that DBD plasma can complete decompose TBT to TiO 2 at moderate temperature. The g-C 3 N 4 gradually lost weight from 550 • C to 730 • C. The composites TCNX gradually lost weight from 520 • C to 650 • C, mainly due to the burning of g-C 3 N 4 . The thermal stability of g-C 3 N 4 was reduced after coating with TiO 2 . The reason is catalytic action of TiO 2 and the cross-linking ring of g-C 3 N 4 after compounding [41]. Ignoring the slight weight loss due to water, the actual proportions of g-C 3 N 4 in the composites TCNX-D (X = 10, 30, 50, 50, 70, 90) were 13.0, 30.3, 52.9, 71.3, and 90.0 wt%, respectively. The proportion of g-C 3 N 4 in TCN50-C was 53.6%, which was approximately equal to the ideal ratio of TCN50-D.

Characterizations of Oxygen Vacancies
Chemical states of elements in g-C 3 N 4 , TiO 2 , and g-C 3 N 4 /TiO 2 composites were investigated by X-ray photoelectron spectroscopy (XPS). Figure S6a shows the high-resolution C 1s spectrum of the prepared samples. All samples have two C 1s peaks at 284.8 and 288.3 eV, which can be ascribed to the inherent adventitious carbon and N-C-N coordination, respectively [42]. Compared with the N-C-N peak of pure g-C 3 N 4 , TCN50 exhibit weaker peak intensity and positive shifts of 0.4 eV in binding energies, which indicates there is a chemical interaction between TiO 2 and g-C 3 N 4 , hence, g-C 3 N 4 has close surface contact with TiO 2 [43]. As shown in Figure S5b, for the N 1 s high-resolution spectrum of pure g-C 3 N 4 and g-C 3 N 4 /TiO 2 composites, three peaks were observed at about 398.8, 399.6, and 401.2 eV. The first peak is attributed to sp2 hybridized nitrogen (C=N-C), the second peak is due to the tertiary N in N-(C) 3 groups, and the last peak corresponds to the existence of amino groups (C-N-H) [44,45]. Figure 2a,b shows O 1s and Ti 2p spectrum region of pure TiO 2 . Figure 2a shows the O 1s spectrum region of TiO 2 -D have three peaks at about 529.7, 532.0 and 533.4 eV, which corresponds to the lattice oxygen, the oxygen vacancy and the adsorbed oxygen, respectively [46,47]. By calculating the ratio of the area occupied by oxygen vacancies, it was 23.4% and 18.6% in TiO 2 -D and TiO 2 -C, respectively, indicating that plasma method can produce more oxygen vacancies than that of the calcination method in TiO 2 . Figure 2b shows the Ti 2p spectrum of pure TiO 2 , There are four peaks at about 458.2, 463.7, 458.9, and 464.6 eV, corresponding to Ti 3+ 2p 3/2 , Ti 3+ 2p 1/2 , Ti 4+ 2p 3/2 and Ti 4+ 2p 1/2 , respectively [46,47]. It was found that the area of Ti 3+ in TiO 2 -D and TiO 2 -C were 15.6% and 11.1%, respectively. Compared with TiO 2 by calcination, the valence of more Ti in TiO 2 by plasma is reduced Nanomaterials 2020, 10, 805 6 of 14 from Ti 4+ to Ti 3+ . This is consistent with the observation for the oxygen vacancy. During the plasma preparation of TiO 2 , the oxygen atoms escaped to form oxygen vacancies and trivalent Ti [46].  show similar peaks at O 1s and Ti 2p. However, the intensity of the peaks representing oxygen vacancies and Ti 3+ decreased. The ratio of oxygen vacancies and Ti 3+ were 23.4% and 11.2% in TCN50-D, 18.6% and 7.2% in TCN50-C, respectively. It is indicated that g-C3N4 occupies the oxygen vacancies of TiO2 after TiO2 coated on g-C3N4. This will undoubtedly strengthen the interaction of TiO2 and g-C3N4. It is in agreement with previous results.
To further verify the existence of oxygen vacancies, electron paramagnetic resonance (EPR) analysis of the TCN50 composites was performed. As shown in Figure 3a, both TCN50-C and TCN50-D have the EPR signal at g = 2.003, which means the appearance of oxygen vacancies [48]. The peak intensity of TCN50-D is stronger than TCN50-C, indicating the plasma method produces more oxygen vacancies. As can be seen from Figure S6, the EPR spectra of TiO2-C and D exhibit the same characteristics as that of TCN50. This is consistent with XPS results. In addition, Figure 3b shows different colors between TiO2-D and TiO2-C. The color of TiO2-C is white, while TiO2-D is gray. This indicates oxygen vacancies can narrow band gap to promote light harvesting, leading significant color change [49].  decreased. The ratio of oxygen vacancies and Ti 3+ were 23.4% and 11.2% in TCN50-D, 18.6% and 7.2% in TCN50-C, respectively. It is indicated that g-C 3 N 4 occupies the oxygen vacancies of TiO 2 after TiO 2 coated on g-C 3 N 4 . This will undoubtedly strengthen the interaction of TiO 2 and g-C 3 N 4 . It is in agreement with previous results.
To further verify the existence of oxygen vacancies, electron paramagnetic resonance (EPR) analysis of the TCN50 composites was performed. As shown in Figure 3a, both TCN50-C and TCN50-D have the EPR signal at g = 2.003, which means the appearance of oxygen vacancies [48]. The peak intensity of TCN50-D is stronger than TCN50-C, indicating the plasma method produces more oxygen vacancies. As can be seen from Figure S7, the EPR spectra of TiO 2 -C and D exhibit the same characteristics as that of TCN50. This is consistent with XPS results. In addition, Figure 3b shows different colors between TiO 2 -D and TiO 2 -C. The color of TiO 2 -C is white, while TiO 2 -D is gray. This indicates oxygen vacancies can narrow band gap to promote light harvesting, leading significant color change [49]. Nanomaterials 2020, 10, x FOR PEER REVIEW 7 of 15

Optical Properties
Ultraviolet-visible diffuse reflection spectrum (UV-vis DRS) is shown in Figure 4a and S10 reveal the light absorption capacity and bandgaps of the as-prepared g-C3N4, TiO2, and TCNX. The critical values of the light response of TiO2-D and TiO2-C were found at 398 nm and 388 nm, indicating that bandgaps were 3.12 and 3.19 eV, respectively [50]. The absorption starting point of g-C3N4 was located at 458 nm. The absorption range of visible light of TCN50 composites was significantly expanded due to introducing g-C3N4. As shown in Figure 4b, on the basis of the Kubelka-Munk formula, the bandgaps of g-C3N4, TCN50-C, and TCN50-D were calculated to be 2.71, 2.70, and 2.63 eV, respectively. Based on the above results, the samples prepared by plasma had a narrower bandgap. This can be attributed to oxygen vacancies [51]. Oxygen vacancies can introduce a defect status below the conduction band, and thus narrows the bandgap to improve the light absorption range [46,52]. Hence, TCN50-D has a relatively strong visible light response capacity.  Figure 4a and Figure S8 reveal the light absorption capacity and bandgaps of the as-prepared g-C 3 N 4 , TiO 2 , and TCNX. The critical values of the light response of TiO 2 -D and TiO 2 -C were found at 398 nm and 388 nm, indicating that bandgaps were 3.12 and 3.19 eV, respectively [50]. The absorption starting point of g-C 3 N 4 was located at 458 nm. The absorption range of visible light of TCN50 composites was significantly expanded due to introducing g-C 3 N 4 . As shown in Figure 4b, on the basis of the Kubelka-Munk formula, the bandgaps of g-C 3 N 4 , TCN50-C, and TCN50-D were calculated to be 2.71, 2.70, and 2.63 eV, respectively. Based on the above results, the samples prepared by plasma had a narrower bandgap. This can be attributed to oxygen vacancies [51]. Oxygen vacancies can introduce a defect status below the conduction band, and thus narrows the bandgap to improve the light absorption range [46,52]. Hence, TCN50-D has a relatively strong visible light response capacity.

Ultraviolet-visible diffuse reflection spectrum (UV-vis DRS) is shown in
Photoluminescence (PL) spectra was shown in Figure 4c. The PL peak intensity is proportional to the recombination rates of the photogenerated electron-hole pairs [53]. g-C 3 N 4 has strong peak intensity, indicating that its charge separation efficiency is low. TiO 2 -D has weaker PL peak strength than that of TiO 2 -C, indicating that TiO 2 -D is more conducive to charge transport. After TiO 2 coated on g-C 3 N 4 , PL peak intensity was greatly reduced, because electrons can be transmitted through the interface of g-C 3 N 4 and TiO 2 [54]. TCN50-D shows lower peak intensity than TCN50-C, resulting in that TCN50-D has lower charge recombination rate and stronger photocatalytic activity. Meanwhile, the presence of oxygen vacancies can promote the separation of electrons and holes [55]. The oxygen vacancies not only restrain the recombination of charges, but also narrow the band gap to increase light absorption [52,56]. As shown in Figure 4d, TCN50-D exhibits the highest photocurrent density than TCN50-C, indicating that TCN50-D has higher photoelectric conversion efficiency [57].

Optical Properties
Ultraviolet-visible diffuse reflection spectrum (UV-vis DRS) is shown in Figure 4a and S10 reveal the light absorption capacity and bandgaps of the as-prepared g-C3N4, TiO2, and TCNX. The critical values of the light response of TiO2-D and TiO2-C were found at 398 nm and 388 nm, indicating that bandgaps were 3.12 and 3.19 eV, respectively [50]. The absorption starting point of g-C3N4 was located at 458 nm. The absorption range of visible light of TCN50 composites was significantly expanded due to introducing g-C3N4. As shown in Figure 4b, on the basis of the Kubelka-Munk formula, the bandgaps of g-C3N4, TCN50-C, and TCN50-D were calculated to be 2.71, 2.70, and 2.63 eV, respectively. Based on the above results, the samples prepared by plasma had a narrower bandgap. This can be attributed to oxygen vacancies [51]. Oxygen vacancies can introduce a defect status below the conduction band, and thus narrows the bandgap to improve the light absorption range [46,52]. Hence, TCN50-D has a relatively strong visible light response capacity.  Photoluminescence (PL) spectra was shown in Figure 4c. The PL peak intensity is proportional to the recombination rates of the photogenerated electron-hole pairs [53]. g-C3N4 has strong peak intensity, indicating that its charge separation efficiency is low. TiO2-D has weaker PL peak strength than that of TiO2-C, indicating that TiO2-D is more conducive to charge transport. After TiO2 coated on g-C3N4, PL peak intensity was greatly reduced, because electrons can be transmitted through the interface of g-C3N4 and TiO2 [54]. TCN50-D shows lower peak intensity than TCN50-C, resulting in that TCN50-D has lower charge recombination rate and stronger photocatalytic activity. Meanwhile, the presence of oxygen vacancies can promote the separation of electrons and holes [55]. The oxygen vacancies not only restrain the recombination of charges, but also narrow the band gap to increase light absorption [52,56]. As shown in Figure 4d, TCN50-D exhibits the highest photocurrent density than TCN50-C, indicating that TCN50-D has higher photoelectric conversion efficiency [57]. Figure 5 shows the morphology and microstructure of TiO2, g-C3N4 and g-C3N4/TiO2 composites. As shown in Figure 5a,b, TiO2-D has less particle aggregation than TiO2-C, which agreed well with previous result. As shown in Figure 5c,d, g-C3N4 has an anomalous layered structure and smooth flat surface. After TiO2 coating process by plasma, TiO2 was uniformly dispersed on the surface of g-C3N4, and the surface of g-C3N4 changes from smooth to rough. Figures 5e and S7 show the energy dispersive X-ray spectroscopy (EDS) mapping of TCN50-D, the four elements (C, N, O, Ti) are welldispersed in the TCN50-D composite. It can form a heterostructure for electron transport and reduce charge recombination efficiency.

Morphologies
To further investigate the microstructure of g-C3N4/TiO2 composites, HRTEM was conducted. As can be seen from Figure 5f,g, the particle size of TiO2 in TCN50-C composite is around 18 nm, that in the TCN50-D composite is around 12 nm. This proves that plasma treated sample has smaller particle size. The lattice spacing of TiO2 is 0.35 nm representing the (101) plane of anatase titanium oxide. Figure 5h also shows the clear lattice fringe of g-C3N4 is 0.32 nm, corresponding to the (002) lattice plane of g-C3N4 [54]. Compared with TCN50-C composite, TCN50-D have more and tighter heterojunctions, which also means extensive interfacial contacts to enhance photocatalytic activity.  Figure 5 shows the morphology and microstructure of TiO 2 , g-C 3 N 4 and g-C 3 N 4 /TiO 2 composites. As shown in Figure 5a,b, TiO 2 -D has less particle aggregation than TiO 2 -C, which agreed well with previous result. As shown in Figure 5c,d, g-C 3 N 4 has an anomalous layered structure and smooth flat surface. After TiO 2 coating process by plasma, TiO 2 was uniformly dispersed on the surface of g-C 3 N 4 , and the surface of g-C 3 N 4 changes from smooth to rough. Figure 5e and Figure S9 show the energy dispersive X-ray spectroscopy (EDS) mapping of TCN50-D, the four elements (C, N, O, Ti) are well-dispersed in the TCN50-D composite. It can form a heterostructure for electron transport and reduce charge recombination efficiency.

Morphologies
To further investigate the microstructure of g-C 3 N 4 /TiO 2 composites, HRTEM was conducted. As can be seen from Figure 5f,g, the particle size of TiO 2 in TCN50-C composite is around 18 nm, that in the TCN50-D composite is around 12 nm. This proves that plasma treated sample has smaller particle size. The lattice spacing of TiO 2 is 0.35 nm representing the (101) plane of anatase titanium oxide. Figure 5h also shows the clear lattice fringe of g-C 3 N 4 is 0.32 nm, corresponding to the (002) lattice plane of g-C 3 N 4 [54]. Compared with TCN50-C composite, TCN50-D have more and tighter heterojunctions, which also means extensive interfacial contacts to enhance photocatalytic activity.

Photocatalytic Activity
RhB degradation and H 2 production under visible light were conducted to evaluate the photocatalytic performance of the samples. As shown in Figure 6a, the concentration of RhB hardly changed within 0.5 h of dark adsorption. For TiO 2 -C and TiO 2 -D, RhB was degraded by 14.3% and 33.7% under two hours of the irradiation of visible light, respectively, which can be ascribed to the sensitization of dye [58]. Plasma-treated TiO 2 show better photoactivity. Pure g-C 3 N 4 degraded 63.7% of RhB under two hours of visible light irradiation. With the increase of proportion of g-C 3 N 4 in g-C 3 N 4 /TiO 2 , the degradation efficiency increased first and then decreased; 50 wt% g-C 3 N 4 in g-C 3 N 4 /TiO 2 composite exhibited the highest RhB degradation efficiency. When the amount of g-C 3 N 4 is low, TiO 2 covers its surface and prevents absorption of visible light. However, when the amount of TiO 2 is low, it is insufficient to promote charge separation. Therefore, the appropriate ratio of g-C 3 N 4 /TiO 2 has higher photocatalytic activity. Compared with 79.7% RhB degradation efficiency of TCN50-C, TCN50-D can completely degrade RhB in two hours, due to its stronger visible light absorption and charge separation efficiency.  Furthermore, Figure 6b shows the reusability experiment of the TCN50-D sample. The photocatalytic activity for degradation of RhB decreased from 95.5% to 90.2% within 6 h under visible light after the third cycle. This indicated that the g-C 3 N 4 /TiO 2 composite by plasma has high stability.
As shown in Figure 6c, the hydrogen production of pure TiO 2 was not detected in the system. Pure g-C 3 N 4 only produced 72.1 µmol·g −1 H 2 per hour. The hydrogen production of TCN50-D and TCN50-C composites was 219.9 and 174.3 µmol·g −1 ·h −1 , respectively. TCN50-D exhibited higher photocatalytic activity and approximately 3 times that of pure g-C 3 N 4 and 1.26 times that of TCN50-C. is low, TiO2 covers its surface and prevents absorption of visible light. However, when the amount of TiO2 is low, it is insufficient to promote charge separation. Therefore, the appropriate ratio of g-C3N4/TiO2 has higher photocatalytic activity. Compared with 79.7% RhB degradation efficiency of TCN50-C, TCN50-D can completely degrade RhB in two hours, due to its stronger visible light absorption and charge separation efficiency. Furthermore, Figure 6b shows the reusability experiment of the TCN50-D sample. The photocatalytic activity for degradation of RhB decreased from 95.5% to 90.2% within 6 h under visible light after the third cycle. This indicated that the g-C3N4/TiO2 composite by plasma has high stability.
As shown in Figure 6c, the hydrogen production of pure TiO2 was not detected in the system. Pure g-C3N4 only produced 72.1 μmol·g −1 H2 per hour. The hydrogen production of TCN50-D and TCN50-C composites was 219.9 and 174.3 μmol·g −1 ·h −1 , respectively. TCN50-D exhibited higher

Mechanism
In order to investigate the phtotcatalytic degradation mechanism of the TCN50-D composite, different scavengers were added before RhB degraded by TCN50-D. As shown in the Figure S10, the degradation activity of TCN50-D changed slightly after the addition of the hydroxyl radical scavenger isopropyl alcohol (IPA) and the superoxide radical scavenger p-benzoquinone (BQ). However, after the addition of the hole scavenger Ethylene Diamine Tetraacetic Acid-2Na (EDTA-2Na), a significant decrease in photocatalytic activity for RhB degradation was observed, indicating that the hole is the main active species for degrading RhB. By measuring the intersection of the slope of the XPS valence band curve and the X axis, the valence band (VB) of the samples can be determined. As shown in Figure S11, the VB of TCN50-C and TCN50-D is 2.32 eV and 2.38 eV, respectively, indicating that TCN50-D possesses stronger oxidizing ability [40]. The VB of TiO 2 -D and TiO 2 -C is 2.91 eV and 2.84 eV, respectively. The conduction band (CB) of TiO 2 -D can be calculated that is −0.20 eV, according to its band gap of 3.11 eV. The VB and CB of g-C 3 N 4 are 2.24 eV and −0.47 eV, respectively [40]. Therefore, the energy band structure model of TCN50-D can be constructed and shown in Figure 6d. Under visible light irradiation, g-C 3 N 4 captures photons to generate electron-hole pairs, then electrons can be transported from the CB of g-C 3 N 4 to the CB of TiO 2 -D to reduce H + to produce H 2 . Meanwhile, the holes in the VB of g-C 3 N 4 can oxidize RhB to exert photodegradation [59].
The mechanism of TiO 2 and g-C 3 N 4 /TiO 2 composite prepared by dielectric barrier discharge plasma is proposed as follows. The high-energy electron bombardment plays a key role in the process.
During the plasma process, there are a large number of electrons that can decompose titanium hydroxide through breaking bonds by electron bombardment (Equation (1)). This also can produce oxygen vacancies.
As shown in Equations (5) and (6), the oxygen atom and ozone can also lead to Ti(OH) 4 decomposition.

Conclusions
TiO 2 and g-C 3 N 4 /TiO 2 composites were prepared by DBD plasma, which is a green, easy, and efficient method for catalyst preparation. Here, TiO 2 by plasma has enriched oxygen vacancies and larger specific surface area due to the electron impact. The electron impact, ozone and oxygen radical play important role in plasma preparation, which facilitate the interaction of TiO 2 and g-C 3 N 4 and forms heterojunctions. g-C 3 N 4 /TiO 2 composites by plasma have stronger light absorption capacity and higher charge separation efficiency. TCN50-D exhibited the highest photocatalysis activity on RhB degradation and hydrogen production, which was ascribed to enriched oxygen vacancies and special heterostructures. Plasma, green and convenient technology, provides a promising strategy for oxide defect engineering and composite preparation.