Semiconductor photocatalysis has attracted much attention because of its potential applications in solving the problems of environmental pollution and energy shortage, such as photocatalytic degradation of organic pollutants [1-4], H₂ production by photocatalytic water splitting [5-7], photocatalytic reduction of CO₂ [8-11], and dye-sensitized solar cells [12]. TiO₂ is a typical semiconductor photocatalyst, and is the most studied photocatalytic material because of its excellent semiconducting properties, low cost, non-toxicity, and long-term stability [13-18]. In nature, TiO₂ mainly exists in three crystal forms: anatase, rutile, and brookite, of which rutile is the most stable form [19]. It is generally accepted that the photoreactivity of anatase TiO₂ is much higher than that of rutile, although some studies have reported a higher photoreactivity for rutile TiO₂ than anatase. Investigation of brookite TiO₂ is limited because pure brookite TiO₂ is difficult to synthesize [20].

Several research groups have recently investigated hollow structured TiO₂, such as TiO₂ hollow microspheres (TiO₂- HMSs) [12, 21-30]. This is because TiO₂-HMSs have high reactivity, low density, and good permeability, and are easy to recycle [31-34]. However, highly photoactive anatase TiO₂ transforms to the less reactive rutile phase at temperatures higher than 600 ℃ under normal conditions, which limits the commercial applications of TiO₂-HMSs in high temperature situations [35]. Doping TiO₂ with metal ions to improve the thermal stability of anatase TiO₂ has been investigated. For example, Shutilov et al. [36] prepared Al-doped anatase TiO₂ nanoparticles by incipient-wetness impregnation of xerogel (anatase) with tert-butyl aluminum solution, and they retained the anatase phase even when calcined at 950 ℃. However, a trace amount of Al₂O₃ also formed during calcination. Kumari et al. [37] showed that the thermal stability of anatase TiO₂ nanoparticles could be improved by using 1-thioglycerol as a capping agent during preparation. However, the thermal stability of the TiO₂ nanoparticles only improved from 500 to 650 ℃.

Non-metal ion dopants, such as S [38], N [39], and F [35, 40, 41], have also been used to increase the anatase–rutile phase transformation temperature, among which fluoride is effective for obtaining high thermal stability anatase TiO₂. According to the study of Yu et al. [42], the presence of fluoride ions can efficiently prevent the anatase–rutile phase transformation. Pillai and co-workers [43] prepared high thermal stability anatase TiO₂ (up to 900 ℃) with high photocatalytic activity using trifluoroacetic acid. Retention of the anatase phase at high temperatures was attributed to the presence of small amounts of fluorine in the lattice. Our group prepared high thermal stability anatase TiO₂ nanosheets with exposed high energy (001) facets, which were prepared by the hydrothermal reaction of a mixed solution of tetrabutyl titanate and hydrofluoric acid [35]. The high thermal stability of the anatase TiO₂ nanosheets was attributed to the strongly adsorbed fluoride ions on the surface. However, fabrication of highly photoreactive TiO₂-HMSs with high thermal stability has not been reported.

In this study, TiO₂-HMSs were prepared by hydrothermal treatment of a mixed Ti(SO₄)₂–NH₄F–H₂O₂ solution according to our previously reported method [44]. Considering that surface-fluorinated TiO₂-HMSs should possess high thermal stability, the effect of the calcination temperature on the structure and photocatalytic activity of the as-prepared TiO₂-HMSs was investigated, where the photocatalytic activity was evaluated by photocatalytic oxidation of acetone under ultraviolet (UV) irradiation.

The surface-fluorinated TiO₂-HMS precursor was prepared in the presence of fluoride ions according to our previously reported method [44]. First, 7.5 mmol of Ti(SO₄)₂ and 7.5 mmol of NH₄F were dissolved in 65 mL of water under magnetic stirring. Second, 10 mL of H₂O₂ (30 wt%) was added dropwise into the mixed solution. Third, the dark-brown solution that formed was transferred to a 100-mL Teflon-lined autoclave and maintained at 180 ℃ for 3 h. After cooling to room temperature, the precipitate was collected, washed with distilled water, and then dried in a vacuum oven.

The obtained white powder (TiO₂-HMS precursor) was then calcined at different temperatures (300–1100 ℃) for 2 h. For simplicity, the samples are denoted Tx, where x is the calcination temperature (see Table 1).

X-ray diffraction (XRD) was performed with a D8 Advance X-ray diffractometer (Bruker, Germany). The accelerated voltage and applied current were 15 kV and 20 mA. The morphology and structure of the samples were characterized by field emission scanning electron microscopy (FE-SEM, s-4800, Hitachi, Japan) with an acceleration voltage of 10 kV and transmission electron microscopy (TEM, Tecnai G20, USA) with an acceleration voltage of 200 kV. The nitrogen adsorption– desorption isotherms were obtained with an ASAP 2020 accelerated surface area and porosimetry system (Micromeritics Instruments, USA). All of the samples were degassed at 150 ℃ prior to the measurements. The Brunauer–Emmett–Teller (BET) specific surface area (A_BET) was determined by the multipoint BET method and the pore size distribution was obtained by the Barrett–Joyner–Halenda method. The nitrogen adsorption volume at P/P₀ = 0.994 was used to determine the pore volume and average pore size. The UV–visible (UV–vis) diffuse reflectance spectra were obtained with a UV–vis spectrophotometer (LambdaBio 35) using BaSO₄ as the reference. The surface chemical state of the samples was analyzed by X-ray photoelectron spectroscopy (XPS) using a Multilab 2000 XPS system with a monochromatic Al source and a charge neutralizer. All of the binding energies were referenced to the C 1s peak of surface adventitious carbon at 284.4 eV.

The photocurrent measurements were performed with an Electrochemical Station 5 (CHI660D, China). A 3-W light- emitting diode (LED) lamp mainly emitting at 365 ± 5 nm was used as the light source for excitation of the FTO/TiO₂ electrode. The measurements were performed with a standard three- electrode assembly. The FTO/TiO₂ electrode, a Pt plate, and a saturated calomel electrode (SCE) electrode were used as the working, counter, and reference electrodes, respectively. The FTO/TiO₂ electrode was prepared using the as-prepared TiO₂ sample as the TiO₂ precursor by the doctor-blade method. Na₂SO₄ solution (0.4 mol L^-1) was used as the electrolyte and saturated with air.

The photocatalytic activities of the TiO₂-HMSs were evaluated by photocatalytic mineralization of gaseous acetone under UV light irradiation in a 15-L reactor at ambient temperature [35]. The prepared TiO₂-HMS samples (0.3 g) were first dispersed in 30 mL of double distilled water. The dispersions were then evenly transferred into three culture dishes with diameters of 7.0 cm. The TiO₂ samples were dried in an oven at 80 ℃ for about 2 h to evaporate water and then cooled to room temperature before use. After the dishes coated with TiO₂ powder were placed in the reactor, 10 μL of acetone was injected into the reactor with a microsyringe. The acetone vapor was allowed to reach adsorption–desorption equilibrium with the catalyst in the reactor prior to UV light irradiation. The concentrations of acetone and CO₂ in the reactor were determined online by an INNOVA 1412 Photoacoustic IR Multigas Monitor (Air Tech Instruments). The initial concentration of acetone after reaching the adsorption equilibrium was about 300 ppm, which remained constant for about 5 min before UV lamp irradiation (15 W at 365 nm). Each set of experiments was performed for 120 min. The photocatalytic activity of the powders was evaluated by comparing the apparent pseudo-first-order reaction rate constant (K_app) based on ln(C₀/C) = k_appt, where k_app is the apparent reaction rate constant, and C₀ and C are the initial and reaction concentrations of acetone, respectively.

The structure is one of the most important factors that affects the photocatalytic activity of semiconductor photocatalysts. XRD was performed to investigate the changes of the phase structure and crystallite size of the TiO₂-HMSs with the calcination temperature. Fig. 1 shows the XRD patterns of the TiO₂-HMS precursor and TiO₂-HMS samples calcined at various temperatures. In the XRD pattern of the TiO₂-HMS precursor, there is a sharp peak at 2θ = 25.3° corresponding to the (101) plane diffraction of anatase TiO₂, indicating that it has good crystallinity [45, 46]. After calcination at temperatures from 300 to 1000 ℃, the peak intensities of anatase increase, which indicates enhancement of crystallization. From Table 1, the average crystalline size steady increases from 79.0 to 97.7 nm and the relative crystallinity of the nanocrystals increases from 1.00 to 1.53 with increasing calcination temperature from 300 to 1000 ℃. However, with a further increase of the calcination temperature to 1100 ℃, a small peak at 2θ = 27.4° appears for the T1100 sample, which correspond to the (110) plane diffraction of rutile TiO₂ [41, 47]. This indicates that anatase–rutile phase transformation begins at about 1100 ℃. At this temperature, 18.4% of the anatase TiO₂ transformed to the rutile phase.

Anatase TiO₂ usually begins to transform to the low photoreactivity rutile phase at about 600 ℃ [48, 49]. However, in the present study, the prepared TiO₂-HMS precursor shows very high thermal stability with an anatase–rutile phase transformation temperature as high as 1100 ℃, indicating that the TiO₂-HMSs can be used in high-temperature environments.

Evolution of the morphology of the TiO₂-HMSs with the calcination temperature can be clearly seen in the corresponding scanning (Fig. 2) and transmission electron microscopy images (Fig. 3). The hollow interior of the TiO₂-HMS precursor is shown in Fig. 2(a), which shows that the hollow microspheres are composed of hollow truncated bipyramidal nanoparticles, indicating exposure of high energy (001) facets [50]. Lu and co-workers [51] reported synthesis of anatase TiO₂ microcrystals with 47% (001) facets on the surface by reversing the relative stabilities of the (101) and (001) facets by doping with fluoride ions [51]. Therefore, it is not unusual to obtain TiO₂-HMS assemblies from nanoparticles with exposed (001) facets. Formation of the hollow interior of the TiO₂ nanoparticles may be because of etching by fluoride ions [52].

With increasing calcination temperature from 300 to 1100 ℃, the morphology of the hollow microspheres remains almost unchanged. However, the shape of the building blocks evolves from hollow truncated bipyramidal nanoparticles to solid nanospheres. It has been reported that almost all of the surface-adsorbed fluoride ions can be removed by calcination at 500 ℃ [51, 53]. Therefore, the exposed high energy (001) facets of the TiO₂ nanoparticles disappear after calcination at high temperature, and the particles evolve into solid nanospheres to reduce the surface energy. Even after calcination at 1100 ℃, the hollow microstructure of the photocatalyst is still clearly observed (Figs. 2(g) and (d)), further indicating the thermal stability of the TiO₂-HMSs.

The BET surface area and pore structure are important parameters that affect the photoreactivity of photocatalysts. Fig. 4 shows the nitrogen adsorption–desorption isotherms and corresponding pore size distribution curves of the TiO₂-HMS samples. The isotherm of the TiO₂-HMS precursor is type Ⅳ (Brunauer, Deming, Deming, and Teller classification) with a hysteresis loop at high relative pressures between 0.6 and 1.0, suggesting the presence of mesopores and macropores [54]. The corresponding pore size distribution (Fig. 4(b)) has a wide range from 4 to ca. 100 nm. The BET surface area and pore volume of the anatase TiO₂-HMS precursor before calcination are 20.8 m² g^-1 and 0.11 cm³ g^-1, respectively. The nanopores (or generation of hysteresis loops) are from aggregation of nanoparticles and the hollow interiors of the truncated bipyramidal nanocrystals (see the arrows in Fig. 2(b)) [55]. Such organized porous structures are extremely useful in photocatalysis because they provide efficient transport pathways for the reactant molecules and products.

After calcination, the adsorption isotherm of the photocatalyst shifts downward and the area of the hysteresis loops decreases, indicating that the photocatalyst has a smaller BET surface area and pore volume. From Table 1, with increasing calcination temperature from 300 to 1100 ℃, the BET surface area decreases from 15.1 to 3.2 m² g^-1 and the pore volume decreases from 0.09 to 0.01 cm³ g^-1.

Fig. 5 compares the light-harvesting abilities of the TiO₂-HMSs calcined at different temperatures. The T300 and T900 samples have similar optical properties with an absorption onset of 388 nm (bandgap of 3.2 eV), while the absorption onset of the T1100 sample begins at 431 nm (bandgap of 2.88 eV). The smaller bandgap of T1100 further confirms phase transformation from anatase TiO₂ to the rutile phase [56], which is consistent with the XRD characterization results. Although the TiO₂-HMS precursor, T300, and T900 samples are the anatase phase, the TiO₂-HMS precursor shows absorption in the visible region (400–450 nm). The enhanced absorption is probably because of the complex reaction between H₂O₂ and titanium ions [44].

The XPS survey spectra of the TiO₂-HMSs before and after calcination are shown in Fig. 6(a). Both the TiO₂-HMS precursor and T300 sample contain Ti (458 eV), O (530 eV), C (284 eV), and F (684 eV) elements. The carbon peak is attributed to adventitious carbon from the XPS instrument [57]. However, the peak of the F element is not present in the XPS spectra of the T900 and T1100 samples. This is because of desorption of the fluoride ions from the surface of the TiO₂-HMSs with calcination. It has been reported that almost all of the adsorbed fluoride ions can be removed when the calcination temperature is higher than 500 ℃ [51].

The high-resolution Ti 2p XPS spectra of the samples are shown in Fig. 6(b). There are two peaks at binding energies of 458.6 and 464.4 eV, which correspond to the signals of Ti⁴⁺ 2p_3/2 and Ti⁴⁺ 2p_1/2 [58, 59]. The high-resolution XPS spectra in the F 1s region (Fig. 6(c)) only contain a single peak centered at 683.8 eV, which is from surface-adsorbed F ions. No signal corresponding to F doped in the lattice of TiO₂ is observed at a binding energy of about 688.6 eV [42, 60]. Therefore, the possibility of F doping in the TiO₂-HMSs can be excluded. Fig. 6(d) shows the high-resolution XPS spectra of TiO₂-HMSs in the O 1s region. The O 1s region is composed of two contributions. The main contribution (529.8 eV) is from lattice oxygen and the minor contribution (532.0 eV) is from surface-adsorbed oxygen (or –OH groups) [61, 62]:

(1)

(2)

(3)

where □ is an oxygen vacancy. The XPS signals of adsorbed oxygen are very weak for the TiO₂-HMS precursor and T300 sample compared with those for the T900 and T1100 samples. This is because of substitution of surface –OH groups by fluoride ions to (Eq. (1)) [63]. The surface-adsorbed fluoride ions can be removed by calcination, forming surface oxygen vacancies (Eq. (2)) [35]. However, oxygen vacancies can be recovered to produce surface adsorbed –OH groups by adsorption of water gas from air (Eq. (3)) [64]. Therefore, for the T900 and T1100 samples (Fig. 6(d)), it is not unusual that there are strong XPS signals for the –OH groups.

Generally, the photocurrent indirectly reflects the semiconductor's ability to generate and transfer photogenerated charge carriers under irradiation [28, 65]. Fig. 7 shows the photocurrent responses of the TiO₂-HMS samples coated on FTO electrodes in photoelectrochemical cells for several on–off cycles. Reproducible prompt photocurrents are generated when the FTO/TiO₂ electrodes are irradiated with an UV LED lamp. This indicates that most of the photogenerated electrons are transferred to the cathode (across the sample) to produce a photocurrent under irradiation. Note that there are initial anodic photocurrent spikes with irradiation of the samples, except for sample T900. The initial anodic photocurrent spikes originate from separation of the photogenerated electron–hole pairs within the photoelectrode. The holes are then transferred to the sample surface and trapped or captured by a reduced species in the electrolyte, while the electrons are transported to the cathode. After the spike current is attained, the photocurrent continuously decreases before reaching a constant current. This indicates that the holes accumulated at the sample surface competitively recombine with electrons rather than being trapped or are captured by a reduced species in the electrolyte [66]. No obvious spike photocurrent is observed for the T900 sample. This indicates that the number of accumulated holes on the surface of the TiO₂-HMSs of the T900 sample sharply decreases, indicating a low recombination rate of photogenerated electron–hole pairs. Notably, the TiO₂-HMS precursor shows the weakest photocurrent (less than 0.2 μA cm^-2), while the T900 sample shows the largest photocurrent (1.2–1.6 μA cm^-2). This indicates that calcination favors efficient separation of the photogenerated carrier, enhancing the photocatalytic activity.

The photocatalytic activity of the TiO₂-HMSs was evaluated by photocatalytic oxidation of acetone. Fig. 8(a) shows the degradation profile of acetone and simultaneous formation of CO₂ in the reactor using the T900 sample as a photocatalyst. The photocatalytic oxidation of acetone is based on

(4)

From Fig. 8(a), about 40 ppm acetone decomposes and a stoichiometric amount of CO₂ (about 120 ppm) is produced after irradiation for 120 min. The molar ratio of acetone to CO₂ is nearly 1:3, indicating complete mineralization of acetone over the irradiated TiO₂-HMSs rather than a simple adsorption process.

Fig. 8(b) shows the relative photocatalytic activities of the TiO₂-HMSs calcined at different temperatures. After calcination at 300 ℃, the photocatalytic activity of the TiO₂-HMSs slightly decreases from 20.8 (TiO₂-HMS precursor) to 15.1 min^-1 (T300) (a 40.1% decrease). This can be interpreted by desorption of surface fluoride ions from the surfaces of the TiO₂-HMSs (Eq. (2)). It has been reported that surface fluorination enhances the photocatalytic activity of TiO₂ by formation of mobile free ^•OH radicals (^•OH_free), which are more reactive than surface-adsorbed ^•OH radicals (^•OH_ads) [67, 68]:

(5)

(6)

According to the XPS characterization results, the adsorbed fluoride ion contents of the TiO₂-HMS precursor and T300 sample are 2.89 and 2.18 at%, respectively. Therefore, it is understandable that the photoreactivity of the T300 sample is less than that of the TiO₂-HMS precursor.

With increasing calcination temperature from 300 to 900 ℃, the photoreactivity of the TiO₂-HMSs steadily increases. This can be ascribed to enhanced crystallization (Fig. 1 and Table 1). However, with a further increase of the calcination from 900 to 1100 ℃, the reactivity of the photocatalyst decreases. This can be attributed to the sharp decrease of the BET surface area (Table 1) and the appearance of the rutile phase.

Anatase TiO₂ will transform to the low photoreactivity rutile phase when calcined at a temperature higher than 600 ℃. However, in the present study, the thermal stability of the anatase TiO₂-HMSs is as high as 1100 ℃ (Scheme 1). The high thermal stability of the TiO₂-HMS precursor is attributed to surface fluorination during preparation:

(7)

(8)

It has been reported that fluoride ions show a strong affinity for TiO₂ by forming a layer of fluorine species (≡Ti–F, Eq. (1)) [63, 69]. The strong adsorption of fluoride ions on the surface of TiO₂ makes them difficult to remove. After calcination at 300 ℃, the adsorbed fluorine content only decreases from 2.89 (TiO₂-HMS precursor) to 2.18 at% (T300, Fig. 6(c)). According to the literature, adsorbed fluorine can only be completely removed at calcination temperatures higher than 500 ℃ [51]. Removal of the surface-adsorbed fluoride ions results in formation of surface oxygen vacancies (Eq. (2)), which inhibits growth of TiO₂ nanocrystals by preventing formation of ≡Ti–O–Ti≡ chains between neighboring TiO₂ nanoparticles (comparing Eqs. (7) and (8)) [35]. Therefore, a higher temperature is needed for diffusion of lattice oxygen to the oxygen vacancy. Fusion of neighboring TiO₂ nanoparticles then begins, and phase transformation occurs. Although adsorption of fluoride ions efficiently prevents the anatase–rutile phase transformation, crystallization of TiO₂ nanocrystals is greatly improved during calcination at high temperature. Therefore, high photoreactivity TiO₂-HMSs with high thermal stability are obtained.

Highly photoactive TiO₂-HMSs have been fabricated by assembly of hollow nanoparticles. The TiO₂-HMSs show very high thermal stability with an anatase–rutile phase transformation of as high as 1100 ℃. Adsorbed fluoride ions are considered to be the most important factor for the high thermal stability of the TiO₂-HMSs. The present study provides a new way to design (photo)catalysts with high thermal stability.