TiO₂ has attracted great attention since the demonstration of its excellent photoelectrochemical activity for water splitting in 1972 [1]. Under light illumination, TiO₂ materials can be used for both renewable energy generation and pollutant degradation [2, 3]. In addition, TiO₂ itself is a clean, stable, nontoxic, and low-cost material; hence, it is often used as a prototype material for photocatalytic applications [4-6]. These advantages highlight the importance of further studies aimed at improving the photocatalytic activity or visible light utilization of TiO₂. Many factors, including crystalline structure [7], specific surface area [8], exposed facets [9-14], morphology [15-17], and defects [18, 19], can affect the photocatalytic activity of TiO₂ materials. Surface decoration with noble metal nanoparticles [20-22] and doping [23-27] are the two main approaches used to enhance the visible light utilization of TiO₂ photocatalysis [28, 29].

The photocatalytic activity of TiO₂ materials is associated with their crystallinities. It was shown that increasing the crystallinity of anatase TiO₂ (a-TiO₂) can result in an enhancement of its photocatalytic activity [30, 31]. Synthesizing highly crystalline anatase TiO₂ or, equivalently, increasing the anatase-rutile transition temperature, appears as a suitable way to improve the photocatalytic properties of this material. It was shown that highly crystalline TiO₂ materials exhibit high photocatalytic activities for hydrogen generation and reactive brilliant red X3B degradation [20, 32]. By using TiOF₂ as precursor, the presence of F^– ions can inhibit the anatase-to-rutile phase transition during the calcination process, and can thus be used to prepare highly crystalline anatase TiO₂. Although TiOF₂ is often used as a precursor to prepare nanostructured TiO₂ materials [32-36], its role as a precursor of highly crystalline anatase TiO₂ and its effect on the photocatalytic properties have not been thoroughly investigated [31]. Controlling the {001}/{101} ratio in anatase crystals is also important for enhancing the photocatalytic activity, owing to the roles of {001} and {101} exposed facets as hole and electron collectors, respectively, which can effectively promote the spatial separation of electrons and holes. In addition, the special surface heterojunction formed by {001} and {101} facets can further prevent the recombination between holes and electrons in the {001} and {101} surfaces, respectively [9]. Therefore, another attractive task involves the preparation of highly crystalline anatase TiO₂ crystals with controlled {001}/{101} ratios. Because F^– ions are well-known surfactants limiting crystal growth along the {001} direction, it is possible to obtain TiO₂ materials with the above features by direct annealing of TiOF₂ precursors.

Although intensive research has focused on TiO₂ for many years, its visible light utilization is still an important issue that needs to be resolved. Although new types of semiconductors such as g-C₃N₄ and Cu₂O-based materials [37-39] have attracted considerable attention in recent years, enabling the visible light utilization of TiO₂ materials in photocatalytic applications still represents a significant challenge. Although a highly crystalline structure may be beneficial to improve the UV light-induced photocatalytic applications of TiO₂, the photocatalytic properties of these materials are still unresponsive to visible light; as a consequence, the preparation of highly crystalline TiO₂ with visible light activity represents an important goal. Surface decoration with noble metals [40] is an efficient way to enable the use of visible light by TiO₂ photocatalysts, due to the local surface plasmon resonance (LSPR) absorption [41]. In particular, gold is a viable choice for the decoration because of its remarkable light absorption at wavelengths around 550 nm. Even small amounts of Au decoration can provide intense absorption of visible light.

Herein, we first synthesized TiOF₂ materials by a traditional hydrothermal route; the samples were then calcined at different temperatures to obtain nanocrystals and sub-microcrystals of TiO₂ with different crystallinities and {001}/{101} facet ratios. The relation of the photocatalytic activity with the crystallinity and the {001}/{101} facet ratio was investigated. Au nanoparticles were also decorated on highly crystalline TiO₂ surfaces to enhance their photocatalytic activities in the visible region.

TiOF₂ precursors were prepared using a simple one-step hydrothermal method, using 15 mL titanium tetrabutoxide (TBOT) and 6 mL HF solution (40%) as raw materials, operating with great caution due to the strongly corrosive nature of the HF solution. The hydrothermal reaction was performed at 180 ℃ for 24 h. After being cooled to room temperature, the as-prepared TiOF₂ materials were washed several times with deionized water as well as alcohol, and then dried at 80 ℃ for further use. Finally, a-TiO₂ samples were synthesized by calcination of the as-prepared TiOF₂ at different temperatures for 2 h. The samples were labeled T0, T500, T600, T700, and T800, where the number denotes the calcination temperature. For example, T600 represents TiO₂ calcined from TiOF₂ at 600 ℃.

The deposition-precipitation (DP) method [42] is a common approach used for decorating metal Au nanoparticles on TiO₂ surfaces. In particular, in our work, 1.5 mL of 0.1 mol L^–1 HAuCl₄ solution was added to 1 g TiO₂ (T700 and T800) in 20 mL containers. Then, 10 mL water was added and the reactants were fully mixed by magnetic stirring. After that, the solution was kept in the dark at room temperature for 1 h to deposit Au nanoparticles. The precipitates obtained by filtering the suspensions were washed several times with deionized water and alcohol, followed by drying at 80 ℃ in air. The resulting samples were labeled Au-T700 and Au-T800.

X-ray diffraction (XRD) patterns were obtained through a D8 Advance (Bruker, Germany) diffractometer using Cu K_α irradiation with a scan rate of 0.02° s^-1. The phase compositions and relative crystallinities were determined from the XRD patterns. A Raman microscope spectrometer (InVia, Renishaw, England) was used to record the Raman scattering spectra. Brunauer-Emmett-Teller specific surface areas (S_BET) were measured using a multichannel surface area and porosity analyzer (TriStar Ⅱ 3020, Micromeritics, USA). Surface morphologies were inspected by field-emission scanning electron microscopy (FESEM, Ultra Plus, Zeiss, Germany) with an acceleration voltage of 5 kV and a working distance of 4 mm. UV-vis diffuse reflectance spectra were recorded by an UV-vis spectrophotometer (UV-2600, Shimadzu, Japan). X-ray photoelectron spectroscopy (XPS, Escalab 250Xi, Thermo Scientific, USA) with an analyzer pass energy of 30 eV was used to determine the binding energies of the material surface.

A MB aqueous solution was chosen as target pollutant to evaluate the aqueous photocatalytic activity. A 0.1 g amount of TiO₂ powder was dispersed in a ø 60 mm glass container, followed by the addition of 15 mL of 1×10^–5 mol L^–1 MB solution. The solutions containing TiO₂ powders were first placed in the dark for 1 h to reach adsorption-desorption equilibrium, after which the solutions were exposed to the UV light illumination provided by four UV fluorescent tube lamps with emission centered at 375 nm. The temperature and light intensity were kept at ~25 ℃ and ~4.5 mW cm^–2 in all experiments. To follow the changes in MB concentration, the UV-vis absorption spectra were monitored for 10 min using a UV-vis spectrometer (UV-2600, Shimadzu, Japan). The apparent rate constants (k_app) were evaluated by linearly fitting the time-dependent ln(C_t/C₀) parameter (where C₀ and C_t are the initial and reacted MB concentrations, respectively) according to quasi-first order kinetics.

The gas-phase photocatalytic properties of the present materials were evaluated by studying the oxidation of formaldehyde and acetone, which are the main indoor pollutants harmful to human health [43]. A 0.1 g amount of TiO₂ powder was ultrasonically dispersed in a ø 60 mm glass container containing 10 mL deionized water and then dried at 80 ℃ in a dry box, to form a thin coating on the bottom of the glass container after removing water. Before the photocatalysis experiments, the sample was pretreated with UV light illumination for 24 h to remove surface carbonate contaminants. A photoacoustic multi-gas monitor (Model 1312, Innova Air Tech Instruments, USA) and a gas chromatograph (GC 2014, Shimadzu, Japan) were used to record the concentration of CO₂ generated from formaldehyde and acetone by the original and Au-decorated samples, respectively. The experimental device is shown in Scheme 1. Before starting the photocatalytic reactions, dry air was allowed to flow through the reactor for at least 10 min to drain out CO₂, until the CO₂ concentration was lower than 20 ppm. A 2 μL aliquot of liquid formaldehyde was then injected into the reactor for photocatalytic oxidation. After dark adsorption for 30 min, an Hg lamp (SP9, Ushio, USA) was switched on to start the photocatalytic reaction under UV light illumination at 15 mW cm^–2 and 365 nm. The reaction temperature was maintained at 40℃ with a heater. The photocatalytic rate was estimated by linearly fitting the CO₂ evolution profiles. The photocatalytic oxidation of acetone by the Au/TiO₂ samples under visible light illumination was investigated following the same procedure, with the difference that a Xe lamp (PLS-SXE300, PerfectLight, China) was adopted as the light source using a 420 nm short-cut optical filter, which only allows the transmittance of long-wavelength light. The light intensity was adjusted to 270 mW cm^–2.

The XRD patterns of the TiO₂ samples obtained from the calcination of TiOF₂ at different temperatures are shown in Fig. 1a, which also indicates the crystal planes corresponding to each peak [44]. The characteristic peaks of T0 match well with those of TiOF₂ (JCPDS No. 01-0490), which indicates the TiOF₂ phase of the hydrothermal product. On the other hand, all XRD peaks of the annealed samples correspond to diffractions of the anatase TiO₂ planes (JCPDS No. 99-0008), revealing that TiOF₂ transformed into the anatase TiO₂ phase after calcination. The weak peak observed for the sample calcined at 800℃ corresponds to the (110) diffraction of rutile TiO₂ (JCPDS No. 88-1175), which indicates that that rutile phase was also present in this sample. To confirm the presence of the rutile phase, Fig. 1b shows enlarged views of the T700 and T800 patterns in the 24°–28° region. Compared to that from the pure anatase phase of T700, a weak XRD peak at 27.5° arising from rutile TiO₂ can be clearly seen in T800. However, the amount of rutile phase is much lower than that of anatase phase. T500 was selected as reference sample with a relative crystallinity (RC) of 1, and the RC of the other samples was estimated based on the intensity of their (101) peak relative to that of T500. Table 1 shows that the RC increases with an increase in annealing temperature. The RC of T800 shows a slight decrease due to the phase transformation from anatase to rutile. The samples maintain an almost pure anatase structure at the calcination temperature of 800 ℃, much higher than the 600 ℃ value generally associated with the transformation from anatase to rutile. F^– ions in the samples are considered to play an important role in inhibiting the anatase-to-rutile transformation [45], increasing the phase transition temperature by about 200 ℃. The average sizes of anatase TiO₂ crystallites were estimated from the (101) peak using the Scherrer equation, which yielded values of 55.4, 66.4, 65.4, and 71.3 nm for the T500, T600, T700, and T800 samples, respectively. Because the presence of F^– ions favors the formation of (001) exposed facets, the ratios of {001} to {101} exposed facets were calculated and are shown in Table 1. In order to calculate the {001}/{101} ratio, a schematic diagram of the anatase TiO₂ structure [46] is shown in Fig. 3f.

Fig. 2a shows the Raman scattering spectra of the synthesized samples. The Raman peaks at 396.2, 515.0, and 638.2 cm^{– 1} reveal the formation of anatase TiO₂ after calcination, which is in full agreement with the analysis of the XRD patterns. These peaks correspond to the Raman-active B_1g, A_1g, and E_g lattice vibration modes, respectively. No Raman signals corresponding to the rutile phase were detected for the T800 sample; this is due to the very low amount of this phase, even though it was observed in the XRD patterns. The intensities of these peaks show a trend similar to that of the XRD patterns, with an initial increase with an increase in calcination temperature, followed by a slight decrease due to the phase transition in the T800 sample. In terms of the specific vibration modes involved, it has been reported [47] that the E_g, B_1g, and A_1g peaks mainly originate from the symmetric stretching vibration (SSV), symmetric bending vibration (SBV), and antisymmetric bending vibration (ABV) of O-Ti-O linkages, respectively. The {101} facets contain saturated 6c-Ti and 3c-O as well as unsaturated 5c-Ti and 2c-O sites, while the {001} facets only contain unsaturated 5c-Ti and 2c-O atoms. When the percentage of {001} facets increased, the intensity associated with the SSV mode decreased, while that of the SBV and ABV modes increased. The XRD analysis shows that the {001}/{101} ratios tend to increase with an increase in calcination temperature, which may also affect the Raman spectra. Fig. 2b shows the normalized Raman scattering spectra of the samples, with the insets displaying partially enlarged view of the B_1g (396.2 cm^–1) and A_1g (515.0 cm^–1) peaks. The figure does not show a regular variation in the intensity of the E_g mode relative to that of the A_1g and B_1g modes, which may be due to the low percentage of exposed facets. Increasing crystallinity and {001}/{101} ratios result in an increase in the intensity of the B_1g mode (396.2 cm^{– 1}) relative to that of the A_1g mode (515.0 cm^–1). The T0 sample does not show any Raman signals, due to its TiOF₂ form.

Figs. 3a–e show the FESEM images of T0, T500, T600, T700, and T800, respectively. Fig. 3a shows that TiOF₂ exhibits an irregular morphology and multiple crystal sizes. The calcined samples show the presence of truncated octahedral bipyramids with high-energy {001} facets, indicating the transformation from TiOF₂ to anatase. The formation of the regular truncated octahedral bipyramids is ascribed to the presence of F^– ions during the calcination process; these ions have been reported to be effective capping agents to form the exposed {001} facets [48]. As the calcination temperature increases from 500 to 800 ℃, the ratio of truncated octahedral bipyramids to irregular particles increases. When the calcination temperature increases to 800 ℃, Fig. 3e shows that the truncated octahedral bipyramids become smooth and unclear compared to those of the other calcined samples. The grain size distributions of the samples are not uniform and their sizes increase with an increase in temperatures, in good agreement with the XRD analysis. The S_BET values and other properties of the as-prepared samples are shown in Table 1. A rapid decrease in S_BET is observed when TiOF₂ transforms into anatase. In the case of the TiO₂ materials, the calcination temperature cannot have a strong impact on the S_BET values, although the grain size shows a clear increase, which may be due to the aggregation of TiO₂ particles. The T700 sample shows the most regular morphology in the form of truncated octahedral bipyramids, despite the different sizes of its particles.

The UV-vis diffuse reflectance spectra of the samples are shown in Fig. 4. The Tauc plots of (αhν)^1/2 vs. hν shown in the inset of the figure were used for estimating the bandgap energies. The transformation from TiOF₂ to anatase TiO₂ is accompanied by a clear red shift of the optical absorption edges. However, the absorption edges hardly change with the calcination temperature when the samples are in the anatase phase. The estimated bandgap energies of TiOF₂ (3.10 eV) and anatase TiO₂ (3.04 eV) are indicated in the inset of Fig. 4. The bandgap energy of the anatase TiO₂ phase is lower than its theoretical value of 3.2 eV, possibly due to the effect of F^– ions, which has been theoretically [49] and experimentally [50] proved. The T800 sample shows an increased intensity between 380 and 420 nm, attributed to the existence of a small amount of rutile, which is a direct bandgap material. The UV-vis diffuse reflectance results are in good agreement with the XRD and Raman analyses.

Fig. 5a shows the time dependence of ln(C₀/C_t) during the photocatalytic degradation of MB under UV light illumination of the T0, T500, T600, T700, and T800 samples. TiOF₂ shows the lowest photocatalytic activity, while the calcined samples present much higher activity. The k_app values of the different samples, normalized to the S_BET values, were estimated from Fig. 5a and are shown in Fig. 5b. The figure shows that the photocatalytic activity for the photodegradation of MB aqueous solutions is not significantly affected by the crystallinity. The T700 material shows the highest photocatalytic activity among the investigated samples.

Fig. 6a shows the CO₂ evolution profiles during the photocatalytic oxidation of formaldehyde under UV light illumination, which exhibit an almost linear increase in the CO₂ amounts. Fig. 6b shows the increasing trend of the rate constants of CO₂ evolution, normalized to the S_BET values. The data corresponding to commercial P25 is also shown for comparison. As shown in the figure, the photocatalytic activity of the calcined samples first increases with the calcination temperature, and then decreases for the T800 sample. The T700 presents the best photocatalytic activity among all calcined samples, whereas TiOF₂ shows the lowest activity due to the absence of the anatase phase. After taking the specific surface area into account, the activity of all as-prepared TiO₂ samples is higher than that of commercial P25, measured under the same experimental conditions. The k_app value of T700 is about six times higher than that of commercial P25, which denotes a substantially improved performance in the degradation of formaldehyde.

Fig. 7 shows the UV-vis diffuse reflection spectra of the T700 and T800 samples and of their Au-decorated counterparts. The inset shows a partial magnification of the curves in Fig. 7. The absorption peaks at ~595 nm for T700 and Au-T800 are ascribed to the local surface plasmon resonance of Au nanoparticles, which shows that metal Au nanoparticles were successfully loaded on the TiO₂ samples, resulting in their absorption in the visible region.

In addition, the XPS technique was used to determine the surface chemical composition of the Au-TiO₂ samples and confirm the presence of metallic Au. Figs. 8a–d show the of O 1s, Ti 2p, C 1s, and Au 4f core-level XPS spectra of Au-T700 and Au-T800. The C 1s peaks shown in Fig. 8c arise from contamination from surface carbonate, which was used as reference for the XPS calibration. According to Chen's work [51], the O 1s, Ti 2p_3/2, and Ti 2p_1/2 peaks located around 530, 459, and 465 eV, respectively, are attributed to the lattice O^2– and Ti⁴⁺ ions. The additional O 1s peaks at 531.8 eV (Au-T800) and 531.0 eV (Au-T700) are ascribed to the hydroxyl (OH^–) species on the TiO₂ surface [52]. The Au 4f peaks located at binding energies around 86 eV (Au 4f_5/2) and 83 eV (Au 4f_7/2) reveal the presence of metallic Au on the TiO₂ surface [53]; the Au/Ti atomic ratios were calculated to be 1.0% for Au-T700 and 0.3% for Au-T800. The different location of the Ti 2p, O 1s, and Au 4f peaks in the Au-T700 and Au-T800 samples is due to their different annealing temperature.

Figs. 9a and b show the CO₂ evolution profiles of TiO₂ and Au/TiO₂ samples under visible light illumination and a comparison between the CO₂ evolution rate constants of the two samples, respectively. The non-decorated samples show very low photocatalytic activities, which substantially increase after Au decoration. In particular, the photocatalytic activities of T700 and T800 increase by 2.6 and 4.8 times, respectively, after Au modification. This Au-induced visible photocatalytic activity is likely due to the local surface plasmon resonance absorption of Au nanoparticles.

The observed increase in photocatalytic activity with an increase in calcination temperature may be due to two reasons. First, the higher crystallinity may inhibit the electron-hole recombination; second, the increase in the {001}/{101} ratios is another important factor, as shown in Scheme 2a. The FESEM images highlight the increased proportion of truncated octahedral bipyramidal structures with an increase in calcination temperature. In addition, the XRD analysis also shows an increase in the {001}/{101} ratio with an increase in calcination temperature. Because the surface energies of the {001} and {101} facets (0.90 and 0.44 J m^–2, respectively) are different [54], electrons tend to assemble on the low-energy {101} facets while holes transfer to the high-energy {001} facets, which favors the separation between holes and electrons. In photocatalysis, the consumption of electrons by O₂ leads to chemisorbed O₂^– species [55], which are considered important reactive oxygen species for organic oxidation. However, if the electron and hole accumulation sites are not well separated from each other, the generated O₂^– species can be consumed by holes on the TiO₂ surface, resulting in the desorption of gaseous O₂. Therefore, the coexistence of {001} and {101} facets can reduce the probability for holes to consume O₂^– species through the separation of electrons and holes into different facets, finally resulting in an enhanced photocatalytic activity. It is worth noting that Yu and coworkers [9] suggested the existence of a heterojunction between the {001} and {101} facets, belonging to the traditional heterojunction type, whose occurrence in a crystal is caused by the presence of two energy states. The presence of this heterojunction enables further separation of electrons and holes accumulated in different facets.

The above results show that the Au decoration accelerates the degradation of acetone in TiO₂ samples, compared to that in the undecorated materials. The visible light-induced photocatalytic mechanism is illustrated in Scheme 2b. The local surface plasmon resonance absorption of the decorated Au nanoparticles can result in the generation of hot electrons. The latter can move to the conduction band (CB) of TiO₂ and reduce O₂ to O₂^–, which is one of the factors promoting the photocatalytic oxidation of acetone. In addition, the hot holes left behind in the excited Au nanoparticles could accept electrons from the organic species and also lead to acetone oxidation. Owing to these reasons, Au decoration can enhance the photocatalytic activity under visible light illumination.

Highly crystalline anatase TiO₂ crystals were prepared by direct calcination of TiOF₂ precursors obtained through a simple hydrothermal method. The present systematic investigations show that a higher crystallinity can improve the UV light-induced photocatalytic activity toward both liquid MB solutions and gaseous formaldehyde. The F^– ions result in the formation of truncated octahedral bipyramidal morphologies after calcination; the increase in {001}/{101} ratio with the calcination temperature is considered beneficial for the separation of light-induced carriers and also contributes to enhancing the photocatalytic activity. The highly crystalline anatase TiO₂ samples decorated with Au, labeled T700 and T800, show 2.6 and 4.8 times higher photocatalytic activities, respectively, under visible light illumination compared to their undecorated counterparts. Our results indicate that combining high crystallinity and surface decoration may be a promising way to improve the visible photocatalytic properties of TiO₂ materials.