Photocatalysis is an important type of heterogeneous catalysis that has been widely studied for environment remediation and energy production ^{[1, 2, 3, 4]}. Among various photocatalysts, TiO₂, the most widely studied semiconductor, has been used as an active photocatalyst in numerous applications such as water treatment ^[5], water splitting ^[6], CO₂ conversion ^[7], and air cleaning ^[8], because of its practicality and strong oxidation capacity. However, the band gap of TiO₂ is wide (about 3.2 eV), which limits its absorption range to less than 390 nm ^[9]. Therefore, it is crucial to enhance the light absorption of TiO₂ in the visible region. Numerous approaches to realize visible-light-driven TiO₂ have been developed. For example, the replacement of oxygen in TiO₂ with nonmetals such as N, C, and S to enhance visible-light absorption ^{[10, 11, 12, 13]}, introducing metals like La, Co, Fe, and Al into the TiO₂ lattice to narrow its bandgap ^[14], loading with plasmonic nanoparticles composed of metals such as Au, Ag, and Cu ^{[6, 8, 15]}, and coupling with narrow-band-gap semiconductors to form heterojunctions ^{[16, 17]}. However, lattice and doping impurity defects in TiO₂ tend to increase the recombination probability of photogenerated electron-hole pairs and decrease thermal stability, which may lower the overall quantum efficiency of TiO₂ ^{[18, 19]}. Therefore, the development of visible-light-responsive TiO₂ that displays efficient photogenerated electron-hole separation remains of great interest ^{[20, 21, 22]}.

Recently, Ti³⁺ self-doped TiO₂ showing enhanced visible-light absorption and photocatalytic performance, attributed to the introduction of a miniband below the conduction band minimum of TiO_2­, has been reported ^{[21, 23, 24]}. A series of synthetic approaches have been developed to prepare Ti³⁺ doped TiO₂ ^[25]. However, according to their electrode potentials, it is difficult to reduce Ti⁴⁺ to Ti³⁺, and the reduced Ti³⁺ tends to revert to its ground state. To date, hydrogen thermal treatment ^[26], plasma treatment ^[27], ion-beam-enhanced deposition ^[28], vacuum activation ^[29], and reduction using a strong reducing agent ^[19] have been used to synthesize Ti³⁺ self-doped TiO₂. Unfortunately, the multiple reaction steps, harsh reaction conditions, and complicated equipment required in these methods limit their practical application. Moreover, the post-reduction of Ti⁴⁺ to Ti³⁺ usually occurs on the surface of TiO₂ nanocrystals. Thus, the Ti³⁺ is easily oxidized via reaction with O₂ in air, resulting in poor stability.

Synthesis of nanostructured TiO₂ with well-defined pore size and high surface area has been considered another effective strategy to improve its photocatalytic performance ^[30]. It is expected that marked improvement of photocatalytic performance can be achieved using Ti³⁺ self-doped porous TiO₂ catalysts. However, high-temperature treatment of Ti³⁺-doped TiO₂ catalysts may cause particle aggregation and decease surface area, making it difficult to synthesize Ti³⁺ self-doped porous TiO₂ nanoparticles. Therefore, the development of a simple, mild strategy to synthesize Ti³⁺-doped mesoporous TiO₂ nanocrystals with high stability is still a challenge.

Evaporation-induced self-assembly (EISA) is a versatile method that has been widely used to fabricate mesoporous metal oxides ^[31]. In general, EISA simply involves the preparation of a solution containing the inorganic precursor, volatile solvent, and surfactant template, followed by slow evaporation of the volatile organic solvent. In the present work, a novel economic strategy for in situ synthesis of Ti³⁺ self-doped mesoporous TiO₂ nanocrystals involving EISA using water as the solvent without additional reducing agent is presented. Different from the typical preparation of Ti³⁺-doped TiO₂ by reduction, here Ti³⁺ is introduced into mesoporous TiO₂ nanocrystals through incomplete hydrolysis of TiCl₃, forming abundant oxygen vacancies in the bulk of TiO₂. These oxygen vacancies extend the light absorption of the photoactive TiO₂ to the visible region and effectively improve its photocatalytic performance in nitric oxide (NO) oxidation and methylene blue (MB) degradation. Unlike samples with surface Ti³⁺ doping, the Ti³⁺ self-doped mesoporous TiO₂ is stable up to 500 °C.

The Ti³⁺ self-doped mesoporous TiO₂ nanocrystals were fabricated by EISA. In a typical process, a precursor solution was first prepared by dissolving polyalkylene oxide block copolymer (EO₁₀₆PO₇₀EO₁₀₆, Pluronic F127, 0.54 g) as a surfactant in water (10 mL) containing 10 wt% HCl and 15 wt% TiCl₃ by vigorous stirring for 1 h in an ice bath to form a transparent sol. The obtained sol was gelled in an open Petri dish at 40 °C in an oven for 4 d. The samples were then calcined in a muffle oven in air at 500 °C for 4 h to remove the surfactant. The obtained sample is denoted as TiO₂-0.54-500, where 0.54 refers to the amount of surfactant and 500 refers to the calcination temperature (°C). Other samples prepared with different surfactant amounts and calcination temperatures have been labeled in a corresponding manner.

The crystal phases of the final products were determined by X-ray diffraction (XRD, Bruker D8 Cu K_α). N₂ adsorption- desorption isotherms were recorded at -196 °C using a Micromeritics ASAP 2010 instrument. The Brunauer- Emmett- Teller (BET) specific surface area (S_BET), pore volume (V_P), and pore diameter (D_P) of sample were calculated by applying the BET and Barrett-Joyner-Halenda (BJH) models to desorption branches. Particle size and morphologies were observed by transmission electron microscopy (TEM; JEOL-2010F, 200 kV). Ultraviolet-visible (UV-Vis) diffuse reflectance spectra (DRS) were obtained on a Varian Cary 100 Scan UV-vis system using a barium sulfate standard as the reference. Electron paramagnetic resonance (EPR) spectra were collected on a Bruker EMX-8/2.7 EPR spectrometer at room temperature. Surface electronic state was analyzed by X-ray photoelectron spectroscopy (XPS, PerkinElmer PHI 5000).

The photocatalytic oxidation of NO in gas phase was carried out at ambient temperature in a continuous flow reactor with volume of 18 L (420 mm × 260 mm × 166 mm) to evaluate the photocatalytic activities of different samples. During the photocatalytic oxidation process, a light source was positioned vertically above the reactor. The amount of photocatalyst used for each experiment was 0.1 g. The initial concentration of NO was about 500 ppb in an air stream (humidity level of 83%) supplied by a zero air generator. After reaching adsorption-desorption equilibrium on the photocatalyst, the light source was turned on. The concentration of NO was continuously measured by using a chemiluminescence NO analyzer (Thermo Environmental Instruments Inc., 42c). The NO removal ratio (%) was calculated using the following equation: NO removal ratio (%) = ([NO]_in − [NO]_out)/[NO]_in × 100%, where [NO]_in and [NO]_out are the NO concentrations determined before and after photocatalytic reaction, respectively.

The photocatalytic degradation of MB was performed in aqueous solution with an initial MB concentration of 10 mg/L under the irradiation of a 300-W Xe arc lamp equipped with a UV cutoff filter (> 420 nm) positioned vertically above the suspension. A suspension of photocatalyst (50 mg) in MB solution (50 mL, 10 mg/L), was stirred for 30 min in the dark to establish adsorption-desorption equilibrium between dye and photocatalyst. At regular intervals, the catalyst was separated from the solution, and the MB concentration of the solution was analyzed by a UV spectrophotometer (UV 7504/PC) at its characteristic absorption wavelength (λ = 663 nm).

The phase structure and crystallite size of the TiO₂-0.54-500 sample were investigated by XRD, as shown in Fig. 1(a). All of the diffraction peaks could be assigned to TiO₂, and suggested that the obtained sample was anatase phase mixed with small amount of rutile phase. The broad (101) with moderate intensity indicates that the sample is highly crystalline with a small grain size, which was calculated to be dozens of nanometers using the Scherrer formula. Fig. 1(b) reveals that the sample displayed a clear optical response in the visible region ranging from 400 to 800 nm. This is different from typical TiO₂, which only absorbs in the UV region. Moreover, the absorption edge of the sample exhibited a considerable red shift to 422 nm compared with that of common anatase TiO₂ with an absorption edge of around 387 nm ^[32]. This shift is attributed to the Ti³⁺ doped in the sample originating from the incompletely reacted TiCl₃ caused by the low hydrolysis temperature used in EISA. The presence of Ti³⁺ in the sample extends the absorption of the sample to the visible region and narrows the band gap of TiO₂ by forming a miniband below the conduction band minimum of TiO₂. The band gap energy calculated from a plot of (ahv)^1/2 versus photon energy (inset of Fig. 1(b)) was 2.6 eV ^[5], which is much narrower than the theoretical band gap of anatase TiO₂ (3.2 eV), providing further evidence for the successful doping of the sample with Ti³⁺.

Fig. 1. Wide-angle XRD pattern (a), UV-vis absorption spectrum (b), TEM image (c), and nitrogen adsorption-desorption isotherms (d) of the TiO2-0.54-500 sample. The inset in (b) is a plot of (ahv)1/2 against photo energy, where a represents the absorption coefficient, and that in (d) is the pore size distribution calculated using the BJH method.

TEM was used to further investigate the particle size and pore morphology of the sample (Fig. 1(c)). Unique worm-like particles with wider interparticle spaces were clearly observed. The particle size estimated from the TEM image was around 15 nm. Fig. 1(d) shows the N₂ adsorption-desorption isotherm and pore size distribution plots (inset) of the sample. The sample exhibited characteristic type-IV isotherm and H2-type hysteresis loops in the relative pressure (p/p₀) range of 0.6-0.8 typical of cylindrical mesopores. S_BET and V_p of the sample were calculated to be 71.8 m²/g and 0.129 cm³/g, respectively. The average pore diameter of the sample was about 7.13 nm estimated from the desorption branch of the isotherm, with a narrow pore size distribution ranging from 2 to 15 nm. This observation is consistent with the TEM analysis. This unique Ti³⁺ self-doped mesoporous TiO₂ sample is expected to meet the requirements for high photocatalytic performance because of its narrow band gap, suitable structure to allow efficient charge separation, and large specific surface area.

To confirm the presence of Ti³⁺ embedded in the mesoporous TiO₂ sample, EPR measurements were conducted. Fig. 2(a) displays the EPR spectrum of the TiO₂-0.54-500 sample. It has been reported that paramagnetic Ti³⁺ has a g-value of around 1.95 ^[19]. An intense peak centered at g = 1.979 was observed, demonstrating the presence of Ti³⁺ in the sample. It is well known that surface Ti³⁺ is not stable when exposed to air or water because it is easily oxidized oxygen (O₂), which is reduced to O₂^- and shows a corresponding peak around g = 2.02. The absence of this signal in Fig. 2(a) indicates that Ti³⁺ was mainly present in the bulk of the sample, which explains the excellent stability of our sample even under thermal treatment at 500 °C for 4 h. The surface Ti state was also investigated by XPS. As shown in Fig. 2(b), two Ti 2p binding energies located at 464.7 and 458.8 eV were attributed to Ti⁴⁺. No Ti³⁺ signal was detected. It is well known that the surface Ti³⁺ is highly reactive with water and oxygen when the sample is exposed at high temperature in air, resulting in the hard detection of surface Ti³⁺. Therefore, EPR and XPS results confirm that the Ti³⁺ was only existed in the bulk of mesoporous TiO₂.

Fig. 2. EPR spectrum (a) and Ti 2p XPS spectrum (b) of the TiO2-0.54-500 sample.

The calcination temperature used to prepare TiO₂ samples is of tremendous importance to achieve a mesoporous network by removing the surfactant micelles. To create porosity and initiate the crystallization of the as-prepared samples, the optimum calcination temperature and duration were investigated. Samples were prepared by changing the calcination temperature from 400 to 550 °C while maintaining the amount of surfactant F127 at 0.54 g. The changes in the degree of crystallization and phase composition of the resulting samples were revealed by XRD (Fig. 3). The textural characteristics of the calcined samples are summarized in Table 1. The crystallinity of the samples increased with calcination temperature, and no peak corresponding to the rutile phase was found for the sample calcined at 400 °C in air. A new peak located at 27.5° appeared when the calcination temperature was increased from 450 to 500 °C. This new peak can be assigned to the diffraction of the (110) plane of rutile, indicating the formation of the rutile phase. The intensity of this diffraction dramatically increased for the sample calcined at 550 °C. These observations correspond well with the reported phase transition temperature of 500 °C for the transition from anatase to thermodynamically stable rutile ^[33]. Meanwhile, the surface area of the sample increased with calcination temperature from 400 to 500 °C because of the efficient removal of surfactant micelles. However, a further increase of calcination temperature to 500 °C caused a decrease in surface area because of the collapse of the mesoporous structure and aggregation of TiO₂ into large nanoparticles.

Fig. 3. Wide-angle XRD patterns of TiO2 samples synthesized using different calcination temperatures and amounts of surfactant.

Table 1 Physicochemical properties of different Ti3+ self-doped TiO2 samples. Sample SBET (m2/g) Pore volume (cm3/g) Pore size (nm) TiO2-0.54-400 61.0 0.128 8.41 TiO2-0.54-450 72.2 0.120 6.64 TiO2-0.54-500 71.8 0.129 7.13 TiO2-0.54-550 40.4 0.108 10.7 TiO2-0.70-500 62.0 0.154 9.93 TiO2-0.38-500 61.0 0.128 8.41 TiO2-0.22-500 52.6 0.122 9.25

Table 1 Physicochemical properties of different Ti3+ self-doped TiO2 samples.

To maintain the open mesopores and initiate TiO₂ crystallization, the dosage of F127 used to prepare the samples was also investigated. The mass of F127 used to synthesize Ti³⁺ self-doped TiO₂ was changed from 0.22 to 0.70 g while the calcination temperature was maintained at 500 °C. Increasing the surfactant amount slightly affected the phase composition and crystallinity of the samples, as shown in Fig. 3, but strongly affected the formation of mesopores. Using a small amount of F127 also led to a low S_BET. As shown in Table 1, S_BET initially increased from 52.6 to 71.8 m²/g as the F127 amount was increased from 0.22 to 0.54 g. The optimum amount of surfactant F127 to maximize S_BET was found to be 0.54 g. Further increasing the surfactant amount above 0.54 g caused a decrease of S_BET.

To evaluate the photocatalytic performance of the mesoporous TiO₂ samples, their photocatalytic oxidation for NO gas under simulated sunlight irradiation in a single-pass flow was investigated. Preliminary tests demonstrated that no substantial reaction of NO was observed in the absence of either light irradiation or photocatalyst, which suggests that NO was quite stable in air and the oxidation of NO was mainly driven by photocatalysis. Figure 4(a) and (b) reveal that the NO removal rate over TiO₂-0.54-500 reached 83% and 40% under irradiation with simulated sunlight for 15 min and visible light for 30 min, respectively. The TiO₂-0.54-500 photocatalyst showed little deactivation because only a 1% decrease in activity was observed after 1 h of photocatalytic reaction, demonstrating the excellent photocatalytic performance of the TiO₂-0.54-500 sample under solar-light irradiation. As well known that, hydroxyl radicals and superoxide radicals, which are respectively produced by oxidation of water with photo-generated holes and reduction of oxygen with photo-generated electrons, play a vital role in photocatalytic oxidation of NO ^{[5, 8]}. In this case, the formation processes of hydroxyl radicals and superoxide radicals were facilitated by introducing Ti³⁺ into the framework of mesoporous TiO₂ to enhance the visible light absorption. Moreover, the mesoporous structure has great advantage of a rapid photo-charge and mass transfer. As shown in Fig. 5, the recovered TiO₂-0.54-500 sample displayed excellent durability and could be continuously used for 3.5 h without marked loss in activity, further suggesting the high stability of the Ti³⁺ self-doped sample.

Fig. 4. Influence of calcination temperature (a, b) and surfactant amount (c, d) on the photocatalytic oxidation of NO under irradiation with either simulated sunlight (2×150-W halogen lamps) or visible light (2×150-W halogen lamps equipped with a UV cutoff filter (>420 nm)).

Fig. 5. Stability of recovered TiO2-0.54-500 in photocatalytic oxidation of NO. The photocatalyst was washed thoroughly with water after 1 h of photocatalytic reaction and dried under vacuum at 80 °C before reuse.

As illustrated in Fig. 4(a) and (c), the calcination temperature and surfactant amount only slightly affected the photocatalytic performance of the samples toward NO oxidation under simulated solar irradiation because of the presence of 2%-5% UV light. Therefore, all samples were effectively activated under simulated solar light. However, the TiO₂-0.54-400, TiO₂-0.54-450, and TiO₂-0.54-550 samples displayed poor NO oxidation reactivity under visible-light irradiation. This could be attributed to the following reasons: (1) a low calcination temperature resulted in low crystallinity and inefficient removal of the surfactant; (2) a high calcination temperature caused a decrease in specific surface area and increase of TiO₂ particle size (Fig. 3 and Table 1); (3) a high calcination temperature promoted oxidation of Ti³⁺, leading to a decrease in the concentration of Ti³⁺ doped in mesoporous TiO₂; (4) inefficient absorption of visible light because of the loss of Ti³⁺ at high temperature. Therefore, the photocatalytic activity of TiO₂-0.54-550 was much lower than that of TiO₂-0.54-500. In contrast, the dosage of F127 used to synthesize mesoporous TiO₂ has only a slight influence on photocatalytic activity under visible-light irradiation. The optimum surfactant amount for the photocatalytic oxidation of NO was found to be 0.54 g, which might be because the dosage of surfactant only affected the mesoporous structure and surface area of the samples. Based on the above photocatalytic performance, we can conclude that the presence of Ti³⁺ played important roles in extending the light absorption of the samples to the visible region and narrowing the band gap of TiO₂ by forming a miniband below the conduction band minimum of TiO₂.

To further investigate the photocatalytic performance of the samples, the degradation of MB in aqueous solution under visible-light irradiation was measured as another photoreaction probe. As shown in the Fig. 6, no significant degradation of MB was observed in the absence of catalyst under visible light irradiation. After 30 min dark reaction, about 6% of MB was adsorbed on catalysts, which was due to the large surface area of catalysts. Figure 6 reveals that the concentration of MB decreased continuously with increasing irradiation time in the presence of TiO₂-0.54-500. Almost 100% of MB was photocatalytically degraded after 2 h, so TiO₂-0.54-500 displayed excellent visible-light-driven photocatalytic activity. The MB solution changed from dark blue to colorless during this time (lower left inset of Fig. 6), visually demonstrating the high photocatalytic activity of TiO₂-0.54-500. These results show that the photocatalytic degradation efficiency of TiO₂ under visible-light irradiation can be markedly enhanced by introducing Ti³⁺ into the framework of mesoporous TiO₂. Therefore, the synergistic effect between Ti³⁺ and mesoporous structure of TiO₂ on narrowing the band gap of TiO₂, enhancing visible light harvesting, and promoting photo-excited charge separation can actually be applied to improve photocatalytic activity ^[34].

Fig. 6. Degradation of MB in the presence of TiO2-0.54-500. The lower left and upper right insets are photographs and UV-vis spectra of the MB solutions after different reaction time, respectively.

We developed a new approach for in situ synthesis of Ti³⁺ self-doped mesoporous TiO₂ via a simple one-step EISA method without additional reductant and subsequent calcination. The mesoporous structure and self-doped Ti³⁺ of the sample are responsible for its enhanced light absorption and photocatalytic activity toward the oxidation of gas-phase NO and photocatalytic degradation of MB in the liquid phase under light irradiation. The present work demonstrates a simple, mild, and economical method to narrow the band gap of TiO₂ through one-step Ti³⁺ self-doping, allowing the development of a highly active and stable photocatalyst that may solve various global environmental problems.