Effects of Particle Size on the Structure and Photocatalytic Performance by Alkali-Treated TiO2

Particle size of nanomaterials has significant impact on their photocatalyst properties. In this paper, TiO2 nanoparticles with different crystalline sizes were prepared by adjusting the alkali-hydrothermal time (0–48 h). An annealing in N2 atmosphere after hydrothermal treatment caused TiO2 reduction and created defects, resulting in the visible light photocatalytic activity. The evolution of physicochemical properties along with the increase of hydrothermal time at a low alkali concentration has been revealed. Compared with other TiO2 samples, TiO2-24 showed higher photocatalytic activity toward degrading Rhodamine B and Sulfadiazine under visible light. The radical trapping and ESR experiments revealed that O2•- is the main reactive specie in TiO2-24. Large specific surface areas and rapid transfer of photogenerated electrons are responsible for enhancing photocatalytic activity. The above findings clearly demonstrate that particle size and surface oxygen defects can be regulated by alkali-hydrothermal method. This research will deepen the understanding of particle size on the nanomaterials performance and provide new ideas for designing efficient photocatalysts.


Introduction
Titanium dioxide (TiO 2 ) has been the most widely used multifunctional material due to its low cost, non-toxicity, wide source, acid-base resistance and high catalytic activity [1]. The outstanding features of TiO 2 greatly extends its application from traditional areas to recent promising fields such as photocatalysis, water splitting, lithium batteries and recently emerging perovskite solar cells [2,3]. However, large band gap of TiO 2 and high electron hole recombination limit its photocatalytic performance [4]. To modify these intrinsic drawbacks, tremendous efforts have been devoted to broadening the active spectrum and promoting the charge transport [5,6].
Particle size is a crucial factor affecting the performance of nano-photocatalytic materials. The size and shape of the catalyst influence its surface structure and then resulting in various catalytic performance [7,8]. Compared with bulk TiO 2 , granular TiO 2 nanoparticles have a large surface area and a broadened band gap. They contain more active sites and display improved photocatalytic activity, photoelectrochemical properties and gas sensitivity [9,10]. Lin et al. successfully synthesized TiO 2 nanoparticles of different particle sizes (12-29 nm) and found the band gap of the TiO 2 nanoparticles was a function of the primary particle size. The results also showed that the primary particle size is closely related to the electro-optical properties of photocatalysts. As the primary particle size increases, the photocatalytic rate constant decreases exponentially [11]. Cheng et al. found that the AgI/BiOI photocatalysts display size-dependent photocatalytic activity, which increases with the smaller size of the AgI NPs. This is believed to be related to the larger number of surface active sites and faster spatial charge transfer [12]. Kocı et al. have proved that as the particle size changed, the specific surface area, charge-carrier dynamics and light absorption efficiency of TiO 2 nanoparticles also changed, and then the yield changed accordingly [13]. Qi et al. summarized that the antimicrobial activity of ZnO NPs exhibits size dependency, that is, small ZnO NPs yield high antibacterial activities [14]. These researches have confirmed that the particle size has an important effect on the catalyst properties.
There are many methods that have been developed to prepare TiO 2 nanostructures with various morphology types and sizes, including the hydrothermal method, the sol-gel method, the vapor deposition method and the electrospinning method [9,15]. Nowadays the hydrothermal synthesis of TiO 2 nanostructures is a well-established technique and typically yields a large amount of TiO 2 nanostructures from solution [16,17]. Most of the current researches focus on the alkali-hydrothermal approaches with high alkali concentration (8-12 M), which usually produces different morphology types of 1D TiO 2 nanostructures (nanotube, nanorod, nanowire, nanofiber or nanobelt) [18,19]. Although the properties and applications of TiO 2 and nanostructures has been studied, the effect of alkali on TiO 2 morphology types, sizes and photocatalytic performance is still unavailable. The effect of low alkali concentration on TiO 2 is less studied.
In this study, we aimed to provide a comprehensive understanding of the effects of low alkali-hydrothermal treatment on the TiO 2 structure and photocatalytic performance. TiO 2 nanoparticles with different particle sizes were prepared by adjusting hydrothermal time by the same preparation method. The effects of particle size on photocatalytic activity and photocarriers migration rate were systematically compared. This research emphasizes the important role of particle size in regulating the photocatalyst structure and performance.

Chemicals
Anatase titanium oxide (CAS: 13463-67-7) was purchased from Sigma Aldrich (St Louis, MO, USA) and Sulfadiazine (SD) was purchased from Aladdin (Shanghai, China). All other chemicals were analytical reagent grade and used as received.

Preparation of TiO 2 Samples
The preparation of TiO 2 -X photocatalysts was performed by a modified alkali-hydrothermal method [20]. Alkali ions penetrate into the TiO 2 lattice, or combine with oxygen to form alkali metal oxides on the TiO 2 surface, which limit the growth of TiO 2 nanocrystals. Therefore, the alkali-hydrothermal method is used to adjust the size of titanium dioxide particles. Generally, 0.6 g Anatase TiO 2 , 60 mL D.I. water and 12 g NaOH were added into a beaker with stirring for 30 min. The mixture was transferred to a 100 mL Teflon-lined stainless steel autoclave and heated to 180 • C with different time. The obtained sediment was washed several times with D.I. water and HCl solution (0.6 M) to adjust the pH to remain neutral. The samples were freeze-dried and then calcined at 400 • C for 2 h under N 2 atmosphere. For the sake of distinction, we define the sample obtained by hydrothermal treatment for X hours as TiO 2 -X.

Characterization
The crystalline structure of the prepared TiO 2 nanomaterial was characterized by X-ray diffraction (XRD, D8 ADVANCE, Bruker AXS, Karlsruhe, Germany) with the scan range from 2 • to 80 • . The morphology of photocatalysts was observed by transmission electron microscope (TEM, JEM-2100F, JEOL, Tokyo, Japan). Elemental analysis in the material was characterized by X-ray photoelectron spectroscopy (XPS, ESCALAB Xi + , Thermo Fisher Scientific, Waltham, MA, USA) and Raman spectra (Raman, Senterra R200-L, Bruker Optics, Karlsruhe, Germany) with a 532 nm laser beam as the excitation source. N 2 adsorption-desorption isotherms were obtained using the Micromeritics (Gemini V2380, Micromeritics, Norcross, GA, USA). The optical absorption properties of photocatalysts were tested by the ultraviolet-visible diffuse reflectance spectroscopy (DRS, Hitachi U-300, Hitachi, Tokyo, Japan) with a scanning range of 200-800 nm. The electron spin resonance (ESR, A300, Bruker, Karlsruhe, Germany) spin-trap technique was employed to monitor the reactive oxygen radical species generated during the irradiation with spin-trap reagent 5,5-dimethyl-1-pirroline-N-oxide (DMPO). Photoluminescence (PL, Hitachi F-4500, Hitachi, Tokyo, Japan) spectra were measured with a continuous wave laser (325 nm) as the exciting source. The main oxidative species in the photocatalytic process could be detected through the free radical trapping experiment by using t-BuOH (hydroxyl radical scavenger), EDTA-2Na (hole scavenger) and circulating N 2 (superoxide radical scavenger).

Photocatalytic Experiments
The photocatalysis was carried out using a 300 W Xe arc lamp (PLS-SXE 300, Perfectlight Co. Ltd., Beijing, China) equipped with a UV-cutoff filter (λ > 400 nm). RhB (Rhodamine B) and SD (Sulfadiazine) were selected as organic contaminants. 0.15 g photocatalyst was placed in 150 mL 20 ppm RhB or 150 mL 10 ppm SD aqueous solution and stirred in the dark to reach an adsorption-desorption equilibrium before illumination. During the period of visible light illumination, the temperature was kept at 20 • C by recirculating cooling water. A 5 mL aliquot was taken at 15 min intervals and filtered through a 0.45 micron filter. The residual concentration of RhB was estimated by visible spectrophotometer (722N, Shanghai Precision & Scientific Instrument Co. Ltd., Beijing, China). The concentration of SD was measured by high performance liquid chromatography (LC-20AT, Shimadzu, Kyoto, Japan) with a C18 (4.6 × 250) reversed phase column. The wavelength of photodiode array detector was 269 nm for SD. The mobile phase was selected as acetonitrile and water (V:V = 1:3) at a flow rate of 0.6 mL/min. Figure 1 shows the TEM images of pristine TiO 2 and prepared TiO 2 with different hydrothermal times (TiO 2 -X). The particle diameters of pristine TiO 2 are distributed in the range from 50 to 250 nm. After hydrothermal treatment for 3 h (Figure 1b), the TiO 2 particles show clear grain boundaries and agglomerations. The particle size decreases obviously to an average of 100 nm and a small amount of particles about 20 nm can be seen. TiO 2 -12 ( Figure 1c) exhibits an irregular morphology. Most of the crystal decreased to 10-40 nm and stacked together, which might indicate the raw materials begin to dissolve, decompose and nucleate. TiO 2 -24 is composed of 10-20 nm nanoparticles. As for TiO 2 -48, TiO 2 nanocrystals experience a growth process along a certain crystallographic orientation to form a rod-like structure.

Characterization of Prepared TiO 2 -X
The porous structure and specific surface areas of TiO 2 -X were characterized by nitrogen adsorption-desorption isotherms as well as the pore size distributions in Figure 2. All samples exhibit type IV nitrogen isotherm with a hysteresis loop, indicating the mesopores features [21]. The hysteresis loops of pristine TiO 2 , TiO 2 -3, TiO 2 -24 pertain to H3-type, that is caused by the unsaturated adsorption of nitrogen in relatively high P/P 0 [22,23]. For TiO 2 -12, a hysteresis loop is observed with distinct saturated adsorption platform in a wide relative pressure region. That belongs to the H2-type and reflects a relatively complex pore structure, which is related to its irregular morphology and stacking between layers in figure TEM [24,25]. TiO 2 -48 shows H2-type hysteresis loops distributed in the whole relative pressure region, mainly due to the irregular morphology and uneven pore size distribution caused by the gradual transformation of nanoparticles into nanorods [26]. The specific surface area, pore size and pore volume data are listed in Table 1 It is indicated that the specific surface area and pore volume increases with decreasing particle size. In Table 1, TiO 2 -24 possessed the largest surface areas (78.888 m 2 /g) and pore volume (0.312 cm 3 /g) nearly six times and 10 times larger compared to the pristine TiO 2 (13.060 m 2 /g and 0.035 cm 3 /g), respectively. For long treatment, particles get more and more aggregated, leading to reduction in surface area and pore volume (growth also inside the pores). in Table 1 It is indicated that the specific surface area and pore volume increases with decreasing 140 particle size. In Table 1, TiO2-24 possessed the largest surface areas (78.888 m 2 /g) and pore volume 0.035 cm 3 /g), respectively. For long treatment, particles get more and more aggregated, leading to 143 reduction in surface area and pore volume (growth also inside the pores). 146 Table 1. The specific surface area (SBET), pore size and pore volume, band gap (Eg) and crystalline 147 size (dXRD) of TiO2-X.    Table 1 displays the crystallite sizes (dXRD) estimated from the half-band width of the corresponding 160 X-ray spectral peak by the Scherrer Formula [30]. As the hydrothermal time increases, the grain size 161 of TiO2 gradually decreases. This may be due to the alkali ions penetrated into TiO2 lattice, or they 162 probably are bonded with oxygen onto TiO2 surface to form alkali metal oxides, both of which 163 restricted the growth of TiO2 nanocrystals [31]. TiO2-48 possesses the smallest crystallite size

172
and 634 cm −1 (Eg) in Figure S1 [34]. Notably, the intensity of the peaks become weaker after hydrothermal treatment, possibly due to the poorer crystallinity [35]. It is reported that the higher 174 the crystallinity, the fewer defects in the crystal. Therefore, defects (oxygen vacancies) may exist in  The X-ray diffraction patterns recorded from TiO 2 with different hydrothermal treatment time are shown in Figure 2f. All the relatively sharp peaks could be indexed as anatase TiO 2 which are basically in agreement with the anatase phase (JCPDS No. 84-1285). Firstly, the intensity of the diffraction peaks decreased with the increasing hydrothermal time, implying the decreased crystallinity of samples. Secondly, the broadening of the diffraction peaks also be observed, which may be due to the combined effect of decreasing grain size and increasing microstrains [27][28][29]. Table 1 displays the crystallite sizes (d XRD ) estimated from the half-band width of the corresponding X-ray spectral peak by the Scherrer Formula [30]. As the hydrothermal time increases, the grain size of TiO 2 gradually decreases. This may be due to the alkali ions penetrated into TiO 2 lattice, or they probably are bonded with oxygen onto TiO 2 surface to form alkali metal oxides, both of which restricted the growth of TiO 2 nanocrystals [31]. TiO 2 -48 possesses the smallest crystallite size calculated by Scherrer Formula, however, the specific surface area of TiO 2 -48 is not the largest. The contradiction arises because the morphology of TiO 2 -48 has changed and the calculation results of XRD only represent the size of the (101) crystal plane instead of the true size of TiO 2 -48 [32,33].
The anatase phase of the TiO 2 samples before and after hydrothermal process is further confirmed by the typical Raman bands at 138 cm −1 (E g ), 191 cm −1 (E g ), 391 cm −1 (B 1g ), 510 cm −1 (A 1g ) and 634 c −1 (E g ) in Figure S1 [34]. Notably, the intensity of the peaks become weaker after hydrothermal treatment, possibly due to the poorer crystallinity [35]. It is reported that the higher the crystallinity, the fewer defects in the crystal. Therefore, defects (oxygen vacancies) may exist in TiO 2 -X samples after hydrothermal treatment [36]. Moreover, compared with the pristine the TiO 2 , low frequency peak at 138 cm −1 (E g ) of the TiO 2 -X samples after hydrothermal treatment exhibits slight blue shift and significant broadening. And the E g peak of TiO 2 -24 is particularly prominent. It shifts from 137.82 cm −1 to 140.07 cm −1 and the full width at half maximum (FWHM) increases from 8.29 to 12.65. Generally, the quantum size limitation effect (that is to say decreased crystalline size) causes red shift and broadened peak while the presence of defects cause blue shift [37,38]. This result shows the effect of defects are dominant compared to crystalline size. To sum up, the variations of Raman shift and FWHM are the combined effect of crystalline size and defects.
The surface component and chemical valence state of TiO 2 -X are further investigated by XPS spectra. The high-resolution Ti 2p XPS spectrum of pristine TiO 2 and TiO 2 -24 are compared. In Figure S2a, the fitting result of pristine TiO 2 shows two peaks at 457.2 eV and 463.0 eV, this corresponds to the Ti 3+ 2p 3/2 and Ti 3+ 2p 1/2 , indicating that Ti 3+ already exists in pristine TiO 2 . Similarly, Ti 3+ 2p 3/2 and Ti 3+ 2p 1/2 peaks can also been detected in Figure 3c, certifying the existence of Ti 3+ in addition to the Ti 4+ oxidation state for TiO 2 -24 samples [39]. TiO2-X shifted to a lower energy because of the band bending effect caused by extra electrons from 197 oxygen vacancies on TiO2 crystal lattice or to satisfy the requirement of charge equilibrium [34,43].   According to the peak areas, the Ti 3+ /Ti 4+ ratio of TiO 2 -24 was calculated to be 85.9%, while the Ti 3+ /Ti 4+ ratio of pristine TiO 2 was 67.2% (Table S2). Usually, Ti 4+ could be reduced to Ti 3+ via trapping electron from oxygen vacancy, so the increased Ti 3+ /Ti 4+ ratio was able to verify indirectly the existence of surface oxygen vacancies [40,41]. The high-resolution O 1s XPS spectrum of pristine TiO 2 and TiO 2 -24 both be fitted into three peaks of around 529 eV, 531 eV and 533 eV, which are ascribed to lattice oxygen (Ti-O), surface hydroxyl groups (-OH) and the oxygen vacancies (O v ) in the vicinity of Ti 3+ [42]. Compared with the pristine TiO 2 , the high-resolution O 1s XPS spectrum of TiO 2 -X shifted to a lower energy because of the band bending effect caused by extra electrons from oxygen vacancies on TiO 2 crystal lattice or to satisfy the requirement of charge equilibrium [34,43]. Ti 3+ and oxygen vacancies can serve as photoinduced charge traps as well as adsorption sites to prevent the electron-hole recombination, which are conducive to the improving photocatalytic activity [44,45].
Ultraviolet-visible diffuse reflectance spectroscopy (DRS) is a common method to characterize the optical absorption properties of materials. As displayed in Figure 4a, the pristine TiO 2 shows a strong absorption in ultraviolet region and the absorption edge is located at 400 nm, which is consistent with the light absorption characteristics of anatase TiO 2 [46]. For TiO 2 -X, it exhibits increased absorption in the region of 200-800 nm after hydrothermal treatment, which can be attributed to surface oxygen vacancies and Ti 3+ [44,47]. According to the Kubelka-Munk function, the band gap energies can be calculated from the plots of (αhv) 1/2 versus photon energy [43]. The band gap energies of pristine TiO 2 , TiO 2 -3, TiO 2 -12, TiO 2 -24 and TiO 2 -48 are 3.27 eV, 3.26 eV, 3.16 eV, 3.42 eV and 3.09 eV, respectively ( Table 1). The wider the band gap, the stronger the redox capability of the photocatalysts. The enhanced visible light absorption is attributed to the presence of surface oxygen vacancies and Ti 3+ , which could form a shallow donor lever just below the conduction band, thus narrowing the band gap and response to visible light [32].

Evaluation of the Photocatalytic Efficiency
Photocatalytic performance of TiO 2 -X samples was evaluated by degradation of RhB and SD under visible light. Pristine TiO 2 , P25 and the pollutants without photocatalyst were used as control samples. P25 is a commercially produced titanium dioxide material with high photocatalytic activity and is often used as a benchmark photocatalyst [48]. The control experiment in Figure S3 showed that all SD samples reach the adsorption/desorption equilibrium in the dark already after 1 h. In Figure 5a,b, All TiO 2 -X samples exhibited better photocatalytic performance under visible light than pristine TiO 2 for RhB and SD degradation. The TiO 2 -24 showed the best performance, approximately 72.7% of RhB was degraded within 120 min and 59.5% of SD was degraded within 6 h. For pristine TiO 2 , TiO 2 -3, TiO 2 -12 and TiO 2 -48, the degradation efficiency of RhB exhibited 6.2%, 16.5%, 20.4%, 23.9% within 120 min and the degradation efficiency of SD was 21.7%, 12.2%, 24.3%, 28.3% within 6 h, respectively. The photodegradation curves of RhB and SD were fitted by pseudo-first-order reaction kinetics. The rate constant (k) of degrading RhB ( Figure S4) was calculated to be 0.0088 min −1 for the TiO 2 -24 and 0.00258 min −1 for P25, and the k of degrading SD was calculated to be 0.13931 h −1 for the TiO 2 -24 and 0.03303 h −1 for P25. Whether degrading RhB or SD, the degradation rate constant of TiO 2 -24 is much higher than P25, and the degradation rate constant of TiO 2 -24 is about 4 times that of P25. The degradation performance of SD is consistent with the degradation trend of RhB. This indicates that TiO 2 samples treated by alkaline hydrothermal method can be used as a useful photocatalyst for both dyes and likely also for other organic pollutants. This also demonstrates that the morphology of the nanomaterial has a certain influence on its photocatalytic effect.  [44,47]. According to the Kubelka-Munk function, the band gap energies can be calculated from the plots of (αhv) 1/2 versus photon energy [43]. The band gap energies of pristine TiO2, TiO2-3, TiO2-12, TiO2-24 and TiO2-48 are 3.27 eV, 3.26 eV, 3.16 eV, 3.42 eV and 3.09 eV, respectively ( Table 1). The wider the band gap, the stronger the redox capability of the photocatalysts. The enhanced visible light absorption is attributed to the presence of surface oxygen vacancies and Ti 3+ , which could form a shallow donor lever just below the conduction band, thus narrowing the band gap and response to visible light [32].

Evaluation of the Photocatalytic Efficiency
Photocatalytic performance of TiO2-X samples was evaluated by degradation of RhB and SD under visible light. Pristine TiO2, P25 and the pollutants without photocatalyst were used as control samples. P25 is a commercially produced titanium dioxide material with high photocatalytic activity and is often used as a benchmark photocatalyst [48]. The control experiment in Figure S3 showed that all SD samples reach the adsorption/desorption equilibrium in the dark already after 1 hour. In Figure 5a and b, All TiO2-X samples exhibited better photocatalytic performance under visible light than pristine TiO2 for RhB and SD degradation. The TiO2-24 showed the best performance, approximately 72.7% of RhB was degraded within 120 min and 59.5% of SD was degraded within 6 hours. For pristine TiO2, TiO2-3, TiO2-12 and TiO2-48, the degradation efficiency of RhB exhibited 6.2%, 16.5%, 20.4%, 23.9% within 120 min and the degradation efficiency of SD was 21.7%, 12.2%, 24.3%, 28.3% within 6 hours, respectively. The photodegradation curves of RhB and SD were fitted by pseudo-first-order reaction kinetics. The rate constant (k) of degrading RhB ( Figure S4) was calculated to be 0.0088 min −1 for the TiO2-24 and 0.00258 min −1 for P25, and the k of degrading SD was calculated to be 0.13931 h −1 for the TiO2-24 and 0.03303 h −1 for P25. Whether degrading RhB or SD, the degradation rate constant of TiO2-24 is much higher than P25, and the degradation rate constant of TiO2-24 is about 4 times that of P25. The degradation performance of SD is consistent with the degradation trend of RhB. This indicates that TiO2 samples treated by alkaline hydrothermal me

240
In addition to the high photocatalytic activity of catalysts, stable recyclability is also essential 241 for their practical application. In Figure 5c, the photocatalytic stability of TiO2-24 was investigated by  In addition to the high photocatalytic activity of catalysts, stable recyclability is also essential for their practical application. In Figure 5c, the photocatalytic stability of TiO 2 -24 was investigated by cyclic photocatalytic degradation of RhB solution. No obvious reduction was observed after three cycles. It is indicated that TiO 2 -24 exhibited relatively high photodegradation performance after a long time irradiation.
3.3. The mechanism of Photocatalytic Performance Improvement of the TiO 2 -X.

257
To further confirm the radicals generated in the reaction process of TiO2-24, ESR analysis was 258 also performed using DMPO as the radical spin-trapping probe [50,51]. After illuminating with 259 visible light for 10 min, TiO2-24 exhibits the strongest intensity of DMPO-O2 •− signals, indicating that 260 the amounts of O2 •− produced in TiO2-24 is much larger than that for pristine TiO2 and TiO2-48,

261
which is consistent with the photodegradation performance.

262
ESR can also be used to confirm the presence of oxygen vacancies. As shown in Figure 6d, there  To further confirm the radicals generated in the reaction process of TiO 2 -24, ESR analysis was also performed using DMPO as the radical spin-trapping probe [50,51]. After illuminating with visible light for 10 min, TiO 2 -24 exhibits the strongest intensity of DMPO-O 2 •− signals, indicating that the amounts of O 2 •− produced in TiO 2 -24 is much larger than that for pristine TiO 2 and TiO 2 -48, which is consistent with the photodegradation performance.
ESR can also be used to confirm the presence of oxygen vacancies. As shown in Figure 6d, there is a sharp signal at g = 2.004, which is identified as the electrons trapped on surface oxygen vacancies. Pristine TiO 2 , TiO 2 -24 and TiO 2 -48 exhibited the enhanced signal intensity of oxygen vacancies after illuminating 10 min under visible light, indicating that the oxygen vacancies were also involved in the degradation process under visible irradiation [52,53]. The single intensity increased from TiO 2 to TiO 2 -48, indicating the surface oxygen vacancies concentrations of TiO 2 increased with extending hydrothermal time [54]. However, when the surface oxygen vacancies concentration is too high, bulk oxygen vacancies can easily be generated [9]. The crystalline performance decreases, the bulk defects increases, the results of XRD also verified the existence of bulk defects in TiO 2 -48. Surface oxygen defects can serve as photoinduced charge traps as well as adsorption sites where the charge transfer to adsorbed species can prevent the electron-hole recombination, which is important for improving of photocatalytic activity; whereas bulk oxygen vacancies only act as charge traps where electron-hole recombines, resulting in the decrease of photoactivity [43,55].

Photoinduced Electron Transfer Properties in the TiO 2 -X.
Photoluminescence (PL) spectra is a useful technique to investigate the oxygen vacancies through the measurement of the charge carrier trapping, immigration and recombination [56]. In Figure 7, the highest intensity emission peak around 433 nm corresponds to the emission of self-trapped excitons located on TiO 6 , which is consistent with the position of absorption edge band in DRS [57,58]. The peaks in the range of 450-500 nm are originated from surface oxygen vacancies and Ti 3+ [59,60]. The peaks at around 530 nm are derived from the oxygen vacancies buried in the bulk of materials [61]. The surface defects can form the exciton energy level near the bottom of the conduction band, which allows them to act as trapping sites and hinders the recombination of photogenerated chargers [36,62]. In Figure 7, the signals of TiO 2 -X are similar, but the peak intensity changed obviously. According to XRD, XPS and ESR results, the pristine TiO 2 possesses good crystallinity as well as little defects. After alkali-hydrothermal treatment, the decreased crystallinity of TiO 2 -X is accompanied by the increased defects, which is beneficial to hinder the recombination of photogenerated chargers, thus the emission peak intensity of TiO 2 -X is weakened [35,63]. However, TiO 2 -48 possesses an excess of surface defects and forms bulk defects, thus the peak intensity was higher than TiO 2 -24.   On the basis of the above results, the surface defects and crystalline sizes played a dominating role on photocatalytic performance. Scheme 1 is the photocatalytic process of TiO 2 -24 with Ti 3+ and oxygen vacancies (O v  in DRS [57,58]. The peaks in the range of 450-500 nm are originated from surface oxygen vacancies 283 and Ti 3+ [59,60]. The peaks at around 530 nm are derived from the oxygen vacancies buried in the 284 bulk of materials [61]. The surface defects can form the exciton energy level near the bottom of the 285 conduction band, which allows them to act as trapping sites and hinders the recombination of 286 photogenerated chargers [36,62]. In Figure 7, the signals of TiO2-X are similar, but the peak intensity 287 changed obviously. According to XRD, XPS and ESR results, the pristine TiO2 possesses good 288 crystallinity as well as little defects. After alkali-hydrothermal treatment, the decreased crystallinity 289 of TiO2-X is accompanied by the increased defects, which is beneficial to hinder the recombination of 290 photogenerated chargers, thus the emission peak intensity of TiO2-X is weakened [35,63]. However,

291
TiO2-48 possesses an excess of surface defects and forms bulk defects, thus the peak intensity was 292 higher than TiO2-24.

Conclusions
The TiO 2 samples were modified at different hydrothermal times (0-48 h) with a low alkali-hydrothermal method. The physicochemical properties and photocatalytic performance of the prepared TiO 2 -X nanoparticles with different crystalline sizes were explored. With the increased time of hydrothermal treatment, the crystalline size of TiO 2 -X decreased first and then increased, and the crystallinity decreased. An annealing in N 2 atmosphere after hydrothermal treatment caused TiO 2 reduction and created defect. The appropriate concentration of surface oxygen vacancies is beneficial to separate photogenerated charges. Due to the combined effects of crystalline size and surfaces defects, the efficiency of photocatalytic degradation of SD and RhB by TiO 2 -24 nanoparticle under visible light was obviously improved. This study highlights the effects of particle size and surface oxygen defects on the structure and performance of photocatalysts. It opens a new window for the purification of water environmental pollution.

Conflicts of Interest:
The authors declare no conflict of interest.