Liquid-Plasma Hydrogenated Synthesis of Gray Titania with Engineered Surface Defects and Superior Photocatalytic Activity

Defect engineering in photocatalysts recently exhibits promising performances in solar-energy-driven reactions. However, defect engineering techniques developed so far rely on complicated synthesis processes and harsh experimental conditions, which seriously hinder its practical applications. In this work, we demonstrated a facile mass-production approach to synthesize gray titania with engineered surface defects. This technique just requires a simple liquid-plasma treatment under low temperature and atmospheric pressure. The in situ generation of hydrogen atoms caused by liquid plasma is responsible for hydrogenation of TiO2. Electron paramagnetic resonance (EPR) measurements confirm the existence of surface oxygen vacancies and Ti3+ species in gray TiO2−x. Both kinds of defects concentrations are well controllable and increase with the output plasma power. UV–Vis diffused reflectance spectra show that the bandgap of gray TiO2−x is 2.9 eV. Due to its extended visible-light absorption and engineered surface defects, gray TiO2−x exhibits superior visible-light photoactivity. Rhodamine B was used to evaluate the visible-light photodegradation performance, which shows that the removal rate constant of gray TiO2−x reaches 0.126 min−1 and is 6.5 times of P25 TiO2. The surface defects produced by liquid-plasma hydrogenation are proved stable in air and water and could be a candidate hydrogenation strategy for other photocatalysts.


Introduction
Titanium dioxide (TiO 2 ) due to its nontoxic, high stability, and high photocatalytic performance is considered as a promising photocatalyst, which has been widely used in environmental remedies and water splitting [1][2][3]. However, the photoactivity of TiO 2 is largely limited by its wide bandgap (3.2 eV) and only responds to UV light, nearly 4% part of solar energy, which severely restricts its practical applications. In order to enhance solar light absorption, much effort has been devoted to bandgap engineering, by means of noble metal (Au, Pt, Ag, etc.) and nonmental (N, S, I, etc.) doping, artificial crystal structure modification, and carbonaceous nanomaterial compositing (carbon nanotubes, graphene, and fullerenes) [4][5][6][7][8][9][10]. Nevertheless, the visible-light energy conversion remains insufficient due to little solar light absorption and many carrier-recombination centers [11,12]. In recent years, black TiO 2 has gained enormous attention owing to its extraordinary visible and infrared

Preparation of Gray TiO 2−x
Commercial white anatase TiO 2 nanopowders were bought from Shanghai Xieqing industrial Co., Ltd.(Shanghai, China), with 30 nm average size. P25 TiO 2 of 20-30 nm was bought from Degussa Corp, which consisted of 78% anatase and 28% rutile. Briefly, we mixed 300 mL of 100 mg/mL well-dispersed anatase TiO 2 solution and 7 mL of 0.1 mol/L nitric acid to get the electrolyte, which was afterwards transferred to electrolytic cell. One anodic platinum sheet (20 × 20 × 1 mm 3 , 99.99%) and two cathodic titanium rods (4-mm diameter, 99.9%) were sealed into a corundum tube in 300-mL nitric acid electrolyte as shown in Figure 1A. Two cathodes were used to generate glow discharges and avoid unbalanced flow and temperature gradients in the electrolyte. The HNO 3 electrolyte here acted as a conductive solution. Pulsed voltages were applied between anode and cathodes to produce intense plasma nearby the cathode surfaces. Liquid plasma was produced on the cathode surface when we applied an appropriate pulse voltage power (600 V). The produced plasma whose intensity is dependent on glow discharge power, induces the high temperature at plasma-electrolyte interface and hydrogen generation. To avoid the electrolyte evaporation, a water chiller was used to maintain the electrolyte temperature at 80 • C. The color of TiO 2 electrolyte gradually turned from white to dark gray within 1 h. The reaction mechanism for the formation of gray TiO 2−x can be ascribed to the hydrogenation effect of liquid plasma. Briefly, TiO 2 nanopowders well dispersed in electrolyte took disordered movements due to intense liquid-plasma shockwaves. Some TiO 2 nanoparticles moved to the plasma region and underwent hydrogenation under high temperature (2000-3000 K) owing to the hydrogen environment derived from water dissociation. Then, large numbers of nanoparticles repeated plasma hydrogenation treatments and hydrogenated gray TiO 2−x from white TiO 2 was finally obtained. Hereafter, we dubbed the sample according to the output plasma power, for example GT-360 refers to gray TiO 2−x obtained by applying an output plasma power of 360 W.

Characterization
The phase and crystallinity for all samples were tested by the X-ray powder diffraction (XRD) using a Rigaku Smartlab (Rigaku, Tokyo, Japan) machine equipped with Cu Ka irradiation (λ = 1.54056 Å). The morphology was characterized by TEM using a JEM-2500SE (JEOL, Tokyo, Japan) instrument operating at an acceleration voltage of 200 kV. UV-Vis diffused reflectance spectra were measured by the Shimadzu UV-2700 spectrophotometer at a wavelength range of 200-800 nm at room temperature. The Raman spectra of samples were recorded by Thermo scientific DXR Raman microscope with 532-nm laser excitation at room temperature. The X-ray photoelectron spectra (XPS) were recorded with thermos Escalab 250Xi. The existence of defects doped in the TiO 2−x nanoparticles was confirmed by the X-band EPR spectra recorded at room temperature. The plasma discharges were controlled by applying square-wave voltages between titanium cathodes and Pt anode, with variable pulsed voltage of 0~800 V, pulsed current of 0~5 A, and repetition rate of 0~5 kHz. The voltage was increased at a ratio of 20 V/s.

Visible-Light Photocatalytic Degradation
The evaluation of visible-light photoactivity was individually tested with three typical waste-water pollutants, including rhodamine B (RhB), methyl orange (MO), and phenol. We used a 300-W Xenon lamp with a 420-nm cutoff filter as the visible-light source. Degussa P25 TiO 2 was used as a standard photocatalyst for photoactivity comparison with as-prepared samples. The concentrations of RhB, MO, and phenol were 20, 20, and 10 mg/L, respectively. Firstly, 50 mg as-prepared TiO 2−x nanopowders and 50 mL of pollutant solution were put in a 500-mL beaker. Before illumination, the mixture was placed in a dark environment for 30 min with magnetic stirring for adsorption-desorption equilibrium. During photodegradation, we took out 1-mL of solution in the glass every 10 min and centrifuged it at 10,000 rpm for 11 min. Finally, the concentration of pollutant solution was analyzed at a specific wavelength by the UV-2700 spectrophotometer, where RhB, MO, and phenol are located at 554, 464, and 270 nm, respectively. Photodegradation experiments for whole tested samples were carried out under the same conditions.

Results and Discussion
The crystal structures of gray TiO 2−x obtained at different plasma discharge powers were detected by XRD analysis as shown in Figure 3. The XRD patterns of the as-grown gray TiO 2−x exhibit characteristic diffraction peaks matching the (101), (103), (004), (112), (200), (105), (211), (204), (116), (220), (215) facets of anatase crystal without any additional diffraction peaks, showing identical diffraction peaks with pristine TiO 2 . In addition, the full width at half maximum and height (FWMH), as well as the integrated peak area of the diffraction peak at 101 facets are nearly unchanged after plasma treatments as seen in Table S1 (see Support Information). This result indicates liquid plasma treatments neither destroy the long-range order lattice structure nor enlarge nanoparticle size of gray TiO 2−x . Nanomaterials 2020, 10, x FOR PEER REVIEW 5 of 15 phenol are located at 554, 464, and 270 nm, respectively. Photodegradation experiments for whole tested samples were carried out under the same conditions.

Results and Discussion
The crystal structures of gray TiO2−x obtained at different plasma discharge powers were detected by XRD analysis as shown in Figure 3. The XRD patterns of the as-grown gray TiO2−x exhibit characteristic diffraction peaks matching the (101), (103), (004), (112), (200), (105), (211), (204), (116), (220), (215) facets of anatase crystal without any additional diffraction peaks, showing identical diffraction peaks with pristine TiO2. In addition, the full width at half maximum and height (FWMH), as well as the integrated peak area of the diffraction peak at 101 facets are nearly unchanged after plasma treatments as seen in Table S1 (see Support Information). This result indicates liquid plasma treatments neither destroy the long-range order lattice structure nor enlarge nanoparticle size of gray TiO2−x.  Figure 4A shows that pristine TiO2 nanoparticles possessed serious agglomeration. Conversely, as shown in Figure 4B, little agglomeration was observed in gray TiO2−x, which could be ascribed to increasement of surface energy in nanoparticles owing to plasma treatments. Both pristine TiO2 and gray TiO2−x nanocrystals are on average ~30 nm in diameter, which is well consistent with XRD measurements and suggests that hot temperature in liquid plasma does not prompt TiO2 crystal growth. Figures 4C,D present the high-resolution morphological structures of pristine TiO2 and TiO2−x nanoparticles. From Figure 4C, marginal defects of pristine TiO2 are observed and could be produced during the crystal growth procedure. As indicated in Figure 4D, it seems that a thin disorder shell, nearly 1-2 nm in thickness, appears on the gray TiO2−x surface, while the bulk maintains good crystallinity. The morphology and lattice information of pristine TiO 2 and gray TiO 2−x nanoparticles are presented in Figure 4A-D. Figure 4A shows that pristine TiO 2 nanoparticles possessed serious agglomeration. Conversely, as shown in Figure 4B, little agglomeration was observed in gray TiO 2−x , which could be ascribed to increasement of surface energy in nanoparticles owing to plasma treatments. Both pristine TiO 2 and gray TiO 2−x nanocrystals are on average~30 nm in diameter, which is well consistent with XRD measurements and suggests that hot temperature in liquid plasma does not prompt TiO 2 crystal growth. Figure 4C,D present the high-resolution morphological structures of pristine TiO 2 and TiO 2−x nanoparticles. From Figure 4C, marginal defects of pristine TiO 2 are observed and could be produced during the crystal growth procedure. As indicated in Figure 4D, it seems that a thin disorder shell, nearly 1-2 nm in thickness, appears on the gray TiO 2−x surface, while the bulk maintains good crystallinity.
The UV-Vis diffuse reflectance spectra (DRS) are used to evaluate the light absorption properties of gray TiO 2−x samples. As shown in Figure 5A, pristine TiO 2 shows merely ultraviolet light absorption, while all as-prepared gray TiO 2−x samples exhibit a broadband absorption from ultraviolet to visible even infrared regions. Moreover, from the inset of Figure 5A, the as-prepared TiO 2−x gradually exhibited a deepened color with the change of output power from 360 to 480 W. Figure 5B plots (αhν) 1/2 versus hν by using the Kubelka-Munk function [33]. From which we calculate the bandgap of pristine TiO 2 , GT-360, GT-420, and GT-480 as 3.14, 3.07, 3.00, and 2.90 eV, respectively. Apparent red-shifts were observed for the absorption edge of gray TiO 2−x compared with pristine TiO 2 . The higher discharge power is, the narrower bandgap of gray TiO 2−x is obtained. The decreased bandgap in our case implies some self-doped crystal defects, for instance oxygen vacancy and/or Ti 3+ centers, has been produced after plasma hydrogenation. Since self-doped defects inevitably introduce doped states between the bandgap such as oxygen-vacancy and Ti 3+ induced intermedium states, which accounts for the visible-light absorption of gray TiO 2−x . The color depth of defective TiO 2−x is reported and can be related to the amount of surface defects, which plays a role of "color centers", and the more surface defects TiO 2−x possesses, the more visible light TiO 2−x captures [34,35]. To verify the existence of surface defects, gray TiO 2−x was heated at 300 • C under atmosphere for 3 h. Unexpectedly, the color was changed from gray to white, which can be attributed to the fact that surface defects diffuse from surface to bulk or can be oxidized by air. Thus, we conclude that some surface defects most likely exist in gray TiO 2−x and its quantity could be controlled directly by varying plasma discharge power.  The UV-Vis diffuse reflectance spectra (DRS) are used to evaluate the light absorption properties of gray TiO2−x samples. As shown in Figure 5A, pristine TiO2 shows merely ultraviolet light absorption, while all as-prepared gray TiO2−x samples exhibit a broadband absorption from ultraviolet to visible even infrared regions. Moreover, from the inset of Figure 5A, the as-prepared TiO2−x gradually exhibited a deepened color with the change of output power from 360 to 480 W. Figure 5B plots (αhν) 1/2 versus hν by using the Kubelka-Munk function [33]. From which we calculate the bandgap of pristine TiO2, GT-360, GT-420, and GT-480 as 3.14, 3.07, 3.00, and 2.90 eV, respectively. Apparent red-shifts were observed for the absorption edge of gray TiO2−x compared with pristine TiO2. The higher discharge power is, the narrower bandgap of gray TiO2−x is obtained. The decreased bandgap in our case implies some self-doped crystal defects, for instance oxygen vacancy and/or Ti 3+ centers, has been produced after plasma hydrogenation. Since self-doped defects inevitably introduce doped states between the bandgap such as oxygen-vacancy and Ti 3+ induced intermedium states, which accounts for the visible-light absorption of gray TiO2−x. The color depth of defective TiO2−x is reported and can be related to the amount of surface defects, which plays a role of "color centers", and the more surface defects TiO2−x possesses, the more visible light TiO2−x captures [34,35]. To verify the existence of surface defects, gray TiO2−x was heated at 300 °C under atmosphere for 3 h. Unexpectedly, the color was changed from gray to white, which can be attributed to the fact that surface defects diffuse from surface to bulk or can be oxidized by air. Thus, we conclude that some surface defects most likely exist in gray TiO2−x and its quantity could be controlled directly by varying plasma discharge power. The Raman spectra of gray TiO2−x and pristine TiO2 are shown in Figure 6. Both samples exhibit similar Raman peaks at 142, 194, 395, 514, and 637 cm −1 , respectively, which is ascribed to anatase TiO2, well consistent with the XRD results. It is observed that the Eg mode of anatase associated with the symmetric stretching vibration of O-Ti-O is shifted to higher wavenumbers compared with pristine TiO2, which reveals that the Ti-O bands environment is changed. In principle, this The Raman spectra of gray TiO 2−x and pristine TiO 2 are shown in Figure 6. Both samples exhibit similar Raman peaks at 142, 194, 395, 514, and 637 cm −1 , respectively, which is ascribed to anatase TiO 2 , well consistent with the XRD results. It is observed that the E g mode of anatase associated with the symmetric stretching vibration of O-Ti-O is shifted to higher wavenumbers compared with pristine TiO 2 , which reveals that the Ti-O bands environment is changed. In principle, this phenomenon can result from lattice defects and/or finite size effect (<10 nm). Our TEM measurements exclude the size effect as pristine TiO 2 and gray TiO 2−x have almost the same nanoparticle size. Consequently, we infer that the difference between gray TiO 2−x and pristine TiO 2 is mainly due to the lattice defects effect. The Raman spectra of gray TiO2−x and pristine TiO2 are shown in Figure 6. Both samples exhibit similar Raman peaks at 142, 194, 395, 514, and 637 cm −1 , respectively, which is ascribed to anatase TiO2, well consistent with the XRD results. It is observed that the Eg mode of anatase associated with the symmetric stretching vibration of O-Ti-O is shifted to higher wavenumbers compared with pristine TiO2, which reveals that the Ti-O bands environment is changed. In principle, this phenomenon can result from lattice defects and/or finite size effect (<10 nm). Our TEM measurements exclude the size effect as pristine TiO2 and gray TiO2−x have almost the same nanoparticle size. Consequently, we infer that the difference between gray TiO2−x and pristine TiO2 is mainly due to the lattice defects effect.  Figure 7A shows the full-scale XPS spectra of pristine TiO2 and gray TiO2−x. No obvious differences are observed between pristine TiO2 and gray TiO2−x, indicating no other dopant elements in gray TiO2−x. Figure 7B shows a high-resolution Ti 2p XPS spectra for both samples. There are two main peaks for pristine TiO2 centered at a binding energy of 458.6 and 464.4 eV, which are ascribed to Ti 4+ 2p3/2 and Ti 4+ 2p1/2, respectively [36]. In comparison with pristine TiO2, two similar peaks of gray TiO2−x are detected but with a slight shift toward lower binding energy at 458.3 and 464 eV, assigned to Ti 3+ 2p3/2 and Ti 3+ 2p1/2, respectively [37]. Figure 7C shows the XPS O 1s spectra for pristine TiO2 and gray TiO2−x. From the O 1s spectrum of pristine TiO2, two deconvoluted peaks at 529.6 and 531.4 eV are related to lattice oxygen (Ti-O) and surface absorption hydroxyl (Ti-OH), respectively [38]. Comparably, the fitting peaks of O 1s spectrum of gray TiO2−x show an obvious shift toward  Figure 7A shows the full-scale XPS spectra of pristine TiO 2 and gray TiO 2−x . No obvious differences are observed between pristine TiO 2 and gray TiO 2−x , indicating no other dopant elements in gray TiO 2−x . Figure 7B shows a high-resolution Ti 2p XPS spectra for both samples. There are two main peaks for pristine TiO 2 centered at a binding energy of 458.6 and 464.4 eV, which are ascribed to Ti 4+ 2p 3/2 and Ti 4+ 2p 1/2 , respectively [36]. In comparison with pristine TiO 2 , two similar peaks of gray TiO 2−x are detected but with a slight shift toward lower binding energy at 458.3 and 464 eV, assigned to Ti 3+ 2p 3/2 and Ti 3+ 2p 1/2 , respectively [37]. Figure 7C shows the XPS O 1s spectra for pristine TiO 2 and gray TiO 2−x . From the O 1s spectrum of pristine TiO 2 , two deconvoluted peaks at 529.6 and 531.4 eV are related to lattice oxygen (Ti-O) and surface absorption hydroxyl (Ti-OH), respectively [38]. Comparably, the fitting peaks of O 1s spectrum of gray TiO 2−x show an obvious shift toward higher binding energy at 529.7 and 531.5 eV, which is mainly due to the existence of oxygen vacancies. Moreover, the area of fitting peak of Ti-OH in gray TiO 2−x is little larger than that of pristine TiO 2 , indicating that many Ti dangling bonds induced by oxygen vacancies are produced.
The existence of surface defects was confirmed by EPR spectra at room temperature as shown in Figure 8. Unexpectedly, the pristine TiO 2 shows an apparent paramagnetic signal centered at g = 2.001 ascribed to surface oxygen defects, which is consistent with HRTEM measurement. In comparison with TiO 2 , gray TiO 2−x possesses two extinguished peaks at g = 2.001 and g = 1.974, corresponding to surface oxygen vacancies and surface Ti 3+ species, respectively, which is in good accordance with XPS results [39,40]. As is well known, the formation of Ti 3+ centers is always accompanied by the presence of oxygen vacancies due to the charge unbalance. It should be noticed that both of the EPR signals gradually increase with plasma discharge power. It is therefore proved that the high concentration of surface defects can be obtained when applying sufficient output energy.
higher binding energy at 529.7 and 531.5 eV, which is mainly due to the existence of oxygen vacancies. Moreover, the area of fitting peak of Ti-OH in gray TiO2−x is little larger than that of pristine TiO2, indicating that many Ti dangling bonds induced by oxygen vacancies are produced. The existence of surface defects was confirmed by EPR spectra at room temperature as shown in Figure 8. Unexpectedly, the pristine TiO2 shows an apparent paramagnetic signal centered at g = 2.001 ascribed to surface oxygen defects, which is consistent with HRTEM measurement. In comparison with TiO2, gray TiO2−x possesses two extinguished peaks at g = 2.001 and g = 1.974, corresponding to surface oxygen vacancies and surface Ti 3+ species, respectively, which is in good accordance with XPS results [39,40]. As is well known, the formation of Ti 3+ centers is always accompanied by the presence of oxygen vacancies due to the charge unbalance. It should be noticed that both of the EPR signals gradually increase with plasma discharge power. It is therefore proved that the high concentration of surface defects can be obtained when applying sufficient output energy.  The existence of surface defects was confirmed by EPR spectra at room temperature as shown in Figure 8. Unexpectedly, the pristine TiO2 shows an apparent paramagnetic signal centered at g = 2.001 ascribed to surface oxygen defects, which is consistent with HRTEM measurement. In comparison with TiO2, gray TiO2−x possesses two extinguished peaks at g = 2.001 and g = 1.974, corresponding to surface oxygen vacancies and surface Ti 3+ species, respectively, which is in good accordance with XPS results [39,40]. As is well known, the formation of Ti 3+ centers is always accompanied by the presence of oxygen vacancies due to the charge unbalance. It should be noticed that both of the EPR signals gradually increase with plasma discharge power. It is therefore proved that the high concentration of surface defects can be obtained when applying sufficient output energy. Based on the obtained results, the generation of surface defects in gray TiO 2−x could be understood that hydrogen atoms generated by liquid-plasma electrolysis reduced TiO 2 to produce Ti 3+ and oxygen vacancies. Furthermore, hydrogen atoms entered in the TiO 2−x lattice and some atoms bonded with Ti atoms while others bonded with O atoms, and thus created the surface lattice disorders [41]. Massive hydrogen atoms produced by enhanced output power create an intensified plasma hydrogenation effect, which induces the increased contents of surface Ti 3+ and oxygen vacancies [42]. In addition, there is no color fade for gray TiO 2−x even if exposed in air and water for six months, indicating the existences of surface defects are quite stable. The robust surface defects could be attributed to the wrapped outer disorder surface with 1-2 nm thickness, hindering further strong oxidation by liquid plasma. On the other hand, considering the presence of surface Ti 3+ and oxygen vacancies, the rather stable point defect structure of Ti 3+ -Vo-Ti 3+ could be formed in order to maintain the electrostatic balance [43].
The photocatalytic performance for whole samples was tested with degradation of RhB under visible-light illumination (λ > 420 nm), as shown in Figure 9A. It is clearly seen that gray TiO 2−x shows a much higher photoactivity than that of pristine TiO 2 and P25. Figure 9B shows the calculated ln(C 0 /C t ) with the irradiation time (t), i.e., ln(C 0 /C t ) = kt, where k is the apparent rate constant, which Nanomaterials 2020, 10, 342 9 of 13 follows first-order kinetic behavior and can directly give the efficiency of photodegradation. The best performance of gray TiO 2−x is 0.126 min −1 , which is far higher than pristine TiO 2 and is 6.5 times of P25. In order to evaluate the stability of gray TiO 2−x , the photocatalytic reaction was repeated four times under the same experimental conditions. As shown in Figure 9C, after four repeated cycles, the photocatalytic performance has no obvious differences, which indicates that the photoactivity of gray TiO 2−x is rather stable. Additionally, the visible-light degradation of methyl orange and phenol are shown in Figure 9D. After 2 h of photodegradation, the degradation rate of MO using gray TiO 2−x is 96.25%, while pristine TiO 2 shows only 7.52%. The degradation rate of phenol for pristine TiO 2 is 4.26%, while gray TiO 2−x shows a much higher rate than pristine TiO 2 and reaches 89.2%. Furthermore, some hydrogenated titania and doped TiO 2 with high performance in RhB photodegradation are provided for comparison with gray TiO 2−x as reflected in Table 1.
On basis of the above results, we can conclude that the superior photocatalytic activity of gray-TiO 2−x is mainly due to the existence of surface defects. It is generally accepted that photocatalytic activity is largely dependent on the quantity of absorbed photons. The more photons TiO 2 can absorb, the more photoinduced electrons and holes it can produce. Owing to the existence of oxygen vacancies and Ti 3+ centers, dopants states such as Ti 3+ states and oxygen states appear between the bandgap, which can narrow the bandgap and allow more long-wavelength light absorption. Even though, lots of photoinduced e-h pairs yet are easily recombined in the bulk of TiO 2 crystal, leading to a serious waste of photoinduced charges. Surface defects serve as holes absorbers and thus cause the directional diffusion of photogenerated e-h pairs to the TiO 2 surface. Therefore, benefits of surface defects in gray TiO 2−x not only narrow its bandgap, but also reduce the recombination rate of e-h pairs. On the other hand, with the increase of output energy, the concentrations of surface defects are correspondingly improved, and lead to a narrower bandgap, as well as faster e-h pairs diffusion. It thus explains discharge power-dependent photoactivity of gray TiO 2−x . Nanomaterials 2020, 10, x FOR PEER REVIEW 9 of 15 Based on the obtained results, the generation of surface defects in gray TiO2−x could be understood that hydrogen atoms generated by liquid-plasma electrolysis reduced TiO2 to produce Ti 3+ and oxygen vacancies. Furthermore, hydrogen atoms entered in the TiO2−x lattice and some atoms bonded with Ti atoms while others bonded with O atoms, and thus created the surface lattice disorders [41]. Massive hydrogen atoms produced by enhanced output power create an intensified plasma hydrogenation effect, which induces the increased contents of surface Ti 3+ and oxygen vacancies [42]. In addition, there is no color fade for gray TiO2−x even if exposed in air and water for six months, indicating the existences of surface defects are quite stable. The robust surface defects could be attributed to the wrapped outer disorder surface with 1-2 nm thickness, hindering further strong oxidation by liquid plasma. On the other hand, considering the presence of surface Ti 3+ and oxygen vacancies, the rather stable point defect structure of Ti 3+ -Vo-Ti 3+ could be formed in order to maintain the electrostatic balance [43].
The photocatalytic performance for whole samples was tested with degradation of RhB under visible-light illumination (λ > 420 nm), as shown in Figure 9A. It is clearly seen that gray TiO2−x shows a much higher photoactivity than that of pristine TiO2 and P25. Figure 9B shows the calculated ln(C0/Ct) with the irradiation time (t), i.e., ln(C0/Ct) = kt, where k is the apparent rate constant, which follows first-order kinetic behavior and can directly give the efficiency of photodegradation. The best performance of gray TiO2−x is 0.126 min −1 , which is far higher than pristine TiO2 and is 6.5 times of P25. In order to evaluate the stability of gray TiO2−x, the photocatalytic reaction was repeated four times under the same experimental conditions. As shown in Figure 9C, after four repeated cycles, the photocatalytic performance has no obvious differences, which indicates that the photoactivity of gray TiO2−x is rather stable. Additionally, the visible-light degradation of methyl orange and phenol are shown in Figure 9D. After 2 h of photodegradation, the degradation rate of MO using gray TiO2−x is 96.25%, while pristine TiO2 shows only 7.52%. The degradation rate of phenol for pristine TiO2 is 4.26%, while gray TiO2−x shows a much higher rate than pristine TiO2 and reaches 89.2%. Furthermore, some hydrogenated titania and doped TiO2 with high performance in RhB photodegradation are provided for comparison with gray TiO2−x as reflected in Table 1.  Considering the high temperature involved in liquid plasma treatments (2000-3000 K), the possible modification of black TiO 2 deriving from the titanium cathodes should be discussed. Originally, we carried out the liquid plasma experiment to produce black TiO 2 , which was reported by others [16][17][18]. Typically, Ti cathodes (99.9%) and Pt anode (99.99%) were used and all other experimental conditions and setup are same with this work, except no TiO 2 powder was mixed in the electrolyte. The black TiO 2 nanoparticles were derived from Ti cathodes after plasma anodization and heat treatment, which shows abundant surface oxygen vacancies. However, the quantity of black TiO 2 is quite few (about 10 mg), and it cannot meet our demands in the experiment and some practical applications. Furthermore, visible-light photodegradation of black TiO 2 shows a heavy decrease in photocatalytic performance, which was ascribed to unstable existences of surface oxygen vacancies due to photocorrosion. To improve the quantity and photocatalytic performance, lots of anatase TiO 2 powders were added and mixed in the electrolyte, which subsequently produced gray TiO 2−x with mass production and stable surface defects. Overall, the contamination of black TiO 2 generated from Ti cathodes existed. Nevertheless, the stabilities of surface oxygen defects, as well as photodegradation performance exhibited in gray TiO 2−x , which confirms that the mechanism of plasma power-related hydrogenation with TiO 2 plays the major role in whole synthesis procedure. Therefore, contamination from Ti cathodes does not affect the performance of gray TiO 2−x , and the modification can be ignored. On the other hand, platinum sheet as inert metals is resisted by heavy corrosion due to intense plasma anodization, and there is no mass decay of Pt anode after plasma treatments. Transmission electron microscopy-energy dispersive X-ray spectroscopy (TEM-EDS) was carried out to exclude the influence of Pt element as seen in Figure S1.

Conclusions
In summary, a simple green method assisted by liquid plasma has been demonstrated for the synthesis of gray TiO 2−x at a low temperature and atmospheric pressure. The synthesis mechanism can be explained that the in situ production of plasma-induced hydrogen atoms plays significant roles in the hydrogenation reactions of gray TiO 2−x . The liquid-plasma technique can effectively manipulate the optical properties and band structures of gray TiO 2−x , and herein enhances accompanying visible-light absorption. One interesting finding is that the engineered surface defects can be realized through directly controlling the plasma discharge power. EPR measurements confirm the existences of surface oxygen vacancies and Ti 3+ species in gray TiO 2−x . Both kinds of defects concentrations are well controllable and increase with the output plasma power. The synergistic effect with narrowed bandgap and directional e-h pairs diffusion results in superior visible-light photoactivity. Rhodamine B was used to evaluate the visible-light photodegradation performance. The results show that the removal rate constant of gray TiO 2−x reaches 0.126 min −1 , and is 6.5 times of P25. Moreover, photodegradations of methyl orange and phenol confirm that gray TiO 2−x exhibits much higher photoactivity than pristine TiO 2 . We believe that this liquid plasma hydrogenation strategy can be extended to other photocatalysts with mass stable surface defects. Overall, our gray TiO 2−x with improved solar energy absorption possesses potential for more efficient polluted water cleaning.

Conflicts of Interest:
The authors declare no competing financial interest.