Enhanced photocatalytic activity of Se-doped TiO2 under visible light irradiation

Anatase TiO2 is a typical photocatalyst, and its excellent performance is limited in ultraviolet light range due to its wide band gap of 3.2 eV. A series of Se-doped TiO2 nanoparticles in anatase structure with various Se concentrations up to 17.1 at.% were prepared using sol-gel method. The doped Se ions are confirmed to be mainly in the valence state of + 4, which provides extra electronic states in the band gap of TiO2. The band gap is effectively narrowed with the smallest gap energy of 2.17 eV, and the photocatalytic activity is effectively improved due to the extended absorption range. The photocatalytic activity was evaluated by the degradation of Rhodamine B (RhB) in aqueous solution under visible light irradiation. The results show that Se doping significantly improves the photocatalytic activity of TiO2 and 13.63 at.% Se-doped TiO2 has the best performance.

without further purification. The distilled water was produced using a Direct-Q Millipore filtration system with resistivity of 18.2 MΩ·cm (Millipore Limited, Watford, UK).
A series of Se-doped TiO 2 nanoparticles with various Se concentration were prepared by sol-gel method. The Se-doped TiO 2 powders were synthesized as following steps: solution A was the mixture of C 16 H 36 O 4 Ti and C 2 H 5 OH which was stirred to transparent solution in dropping process of nitric acid, solution B was ethanol solution of different concentration of SeO 2 . The amount of SeO 2 used was based on the designed atomic concentration of Se in TiO 2 (Ti 1−x Se x O 2 ), which were 0 at.% (pure TiO 2 ), 5 at.%, 10 at.%, 15 at.%, 20 at.% and 25 at.%, respectively. Solution B was dropped into solution A by violent stirring. The uniform sol solution was kept at room temperature for 16 h till it became a gel solution. The gel solution was dried at 120 °C until the xerogel was generated, then the xerogel was calcined at 300 °C for 3 h to get the final powders. The obtained products are denoted as TiO 2 , TSe5, TSe10, TSe15, TSe20, TSe25, respectively, and the number denotes the designed atomic percentage of Se in Ti 1−x Se x O 2 .
Structural characterization. The crystal structure of Se-doped TiO 2 nanoparticles was analyzed by X-ray diffraction (XRD, Rigaku SmartLab3) using Cu Kα radiation (λ = 0.15418 nm). The UV-Vis diffuse reflectance spectra (DRS) were recorded using a UV-visible spectrophotometer (JASCO, UV-670) with a wavelength range of 200-800 nm. The photoluminescence (PL) and PL excitation (PLE) spectra were recorded on a PL spectrofluorometer (HJY-FL3-211-TCSPC) at an excitation wavelength of 365 nm. The morphologies were studied by a scanning electron microscope (SEM, FEI Inspect F50), equipped with an energy dispersive X-ray spectroscope (EDX). EDX was performed for the chemical analysis of the doped samples, and the measured concentrations are listed in Table 1. As can be seen, the measured concentrations of Se are smaller than the designed values, which might be due to the partial sublimation of SeO 2 during the calcining processes 28 . Raman studies were carried out on Raman spectrometer (Horiba Jobin Yvon Lab RAM HR 800) under the backscattering geometric configuration at room temperature. The valence states of each element were studied by X ray photoelectron spectroscopy (XPS, Thermo Fisher Scientific) with Al Kα X ray source (hν = 1486.6 eV). Binding energies were referenced to the C1s peak at 284.5 eV.
Photocatalytic experiments. The visible-light-induced photocatalytic activity was evaluated in aqueous solution, by using Rhodamine B (RhB, Tianjin China Chemical Reagent Ltd.) as a model contaminant in this study. The light irradiation system contained a 500 W Xe lamp (Beijing Trusttech Co Ltd, CHF-XM), equipped with visible light pass filter (400-800 nm). In photocatalytic experiment: 0.4 g photocatalyst was mixed with a 150 mL RhB solution (7.5 mg·L −1 ), an adsorptive experiment primarily was proceeded in the dark for 30 min to achieve absorption-desorption equilibrium of RhB onto the catalyst surface before light irradiation. After interval of scheduled time, 2-3 mL RhB solution was collected for concentration analysis, which was monitored by using a UV-vis spectrophotometer (Hitachi U-3900). Figure 1 shows the XRD patterns of TiO 2 and Se-doped TiO 2 powders. As can be seen, the main diffraction peaks of TiO 2 powders can be indexed to anatase phase (PDF No. 21-1272), except for a small peak labeled by "o", indicating a small concentration of orthorhombic TiO 2 phase (PDF No. 65-2448). Interestingly, by doping with Se, the diffraction peak of orthorhombic phase disappears, and only anatase phase can be observed, suggesting that Se doping might promote the formation of anatase phase. During the polycondensation process, less TiO 6 octahedra units were formed due to the doping of Se. Thus, there was increased dispersion among the TiO 6 octahedra units and as a result the building units shared their corners and anatase phase formation took place 27 . Furthermore, no Se-related oxides or other impurities can be observed. This suggests the uniform distribution of Se ions without aggregation in doped-TiO 2 . The lattice constants a and c are calculated from the XRD patterns, and summarized in Table 1. As can be seen, with 5% Se doping, the lattice constants a and c decrease, and both are smaller than those of TiO 2 , which is due to the smaller radius of Se 4+ (0.64 Å) than Ti 4+ (0.745 Å) (The valence state of Se is determined to be + 4 by XPS, which will be discussed later) 29 . With further increasing Se doping concentration, the lattice constant a is still smaller than that of TiO 2 , but lattice constant c becomes larger than that of TiO 2 . The effect of Se doping on the crystallite size can be evaluated according to the Scherrer's equation, D = Kλ/βcosθ, where K is the constant depending how the half height width of selected diffraction peak is determined (here we use 0.89), λ is the wavelength of X ray (0.15418 nm), β is the half height width, and θ is the Bragg angle. Here we select (101) peak to estimate the crystallite size, which is the strongest diffraction peak for all the samples. As can be seen, the peak becomes narrower first with small doping concentration of TSe5, and then becomes broader with further increasing the doping concentration. The calculated crystallite size is listed in Table 1. Similar phenomenon was observed by Khan et al., though the doping concentration was much smaller in their work 27 . Due to the small concentration of Se in TSe5, the less formation of TiO 6 octahedra might facilitate the polycondensation process of TiO 6 octahedra in corner sharing arrangement, leading to the larger crystallite size of anatase phase 27 . However, with further increasing the doping concentration, the doped Se ions might not occupy the lattice site, and thus the growth of the crystallite size was prevented 30 . The similar rule but much higher Se concentration in our work compared with Khan et al. 's work might be due to the much lower calcining temperature in our work, leading to the much smaller crystallite size and much higher tolerable doping concentration 27 . The morphologies of the samples are further studied by SEM, and the images are shown in Fig. S1 (Supplementary materials). The particle sizes of all samples are all smaller than 10 nm, which is consistent with XRD results. Raman spectroscopy is a powerful technique for the investigation of crystalline and defect structure, and to elucidate the various phases of TiO 2 27 . The Raman spectra of Se-doped TiO 2 with various Se concentrations are shown in Fig. 2. The peak positions are observed at 147 cm −1 (1-E g ), 398 cm −1 (A 1g ), 514 cm −1 (B 1g ), 639 cm −1 (3-E g ) for TiO 2 , corresponding to the vibration modes of anatase phase 27 . All the Se-doped TiO 2 exhibit approximately the same bands, which is consistent with the pure anatase phase observed by XRD. TSe5 shows an extra peak at 195 cm −1 , which corresponds to the 2-E g band. This confirms the better crystalline structure of TSe5. The Se doping effect can be further evaluated from the shape and position of Raman peak, as shown the enlarged view of 1-E g band in inset of Fig. 2. The peak shows the general tendency of becoming broader and shifting of the peak position to smaller wavenumbers, indicating the increasing distortion and smaller crystallite size with increasing Se doping concentration 31 . Interestingly, TSe5 shows a distinct behavior, with much narrower peak and significant shift to smaller wavenumbers. This can be understood by the better crystalline structure with much larger crystallite size for TSe5.  XPS was used to study the chemical states and electronic structure of Se-doped TiO 2 . Figure 3 shows the XPS spectra of Ti, O and Se for TSe20. The Ti2p3/2 and Ti2p1/2 peaks can be deconvoluted into two sets of peaks, the stronger peaks at 458.7 eV and 464.3 eV correspond to Ti 4+ , confirming the main valence state of + 4 for Ti in Se-doped TiO 2 6 . A set of weaker peaks at 456.4 eV and 461.8 eV can be attributed to the Ti 3+6 . This might be due to the low calcining temperature, oxygen is not active enough to fully oxidize Ti, resulting in the formation of O vacancies and Ti 3+ sites in TiO 2 lattice 32 . To confirm the formation of Ti 3+ , we further annealed the TiO 2 powders at 500 °C for 1 hour, and the XPS spectrum is shown in Fig. 3(b). As can be seen, only Ti2p3/2 peak at 458.5 eV and Ti2p1/2 peak at 464.2 eV can be observed, indicating the valence state of + 4. In Fig. 3(c), O1s peak can be deconvoluted into two peaks, one at 529.8 eV, and the other at 528.2 eV . The peak at 529.8 eV is attributed to the Ti-O bond 32 . The observation of peak at 528.2 eV is quite abnormal, since the bonding energy of Se-O in SeO 2 should be higher. It is known that the O2p peak in La 2 O 3 is 528.8 eV, which is due to the higher ionic nature of the La-O bonding 33 . Thus O ions will have lower binding energy when they attract more electrons. In the Se-doped TiO 2 with much higher Se concentration, Se may not occupy the lattice sites due to the large lattice distortion. O may attract electrons not only from the neighboring Ti, but also from Se ions due to the formation of Ti-O-Se structure, leading to lower bonding energy, in comparison with O ions locating at regular lattice site 34 . Figure 3(d) shows the Se3d XPS peak, which can be deconvoluted into two peaks: one at 58.5 eV, and the other at 55.0 eV. The peak at 58.5 eV can be attributed to Se 4+ , and the slightly smaller value might be due to the doping in TiO 2 lattice environment 27 . The peak at 55 eV indicates the existence of small amount of Se 0 26 , which might be due to the reduction from C and H in the sources and further confirms the less chemical activity of O at such low calcining temperature.

Results and Discussion
To investigate the optical absorption properties of Se-doped TiO 2 nanoparticles, the UV-vis DRS were recorded and the results are shown in Fig. 4. The absorption edges of TiO 2 shift to longer wavelength in the visible region, which is consistent with its brown-colored appearance, as shown in the inset of Fig. 4(a). This can be understood by the existence of Ti +3 due to the large concentration of O vacancies and disorder in TiO 2 with very low calcining temperature 35,36 . With Se doping, the TiO 2 powders first exhibits darker color, with the darkest color for TSe5. Then the color becomes pale with further increasing the Se doping concentration. TSe5 shows the highest absorption coefficient in the visible range, which continues to decrease with further increasing Se doping concentration. TSe25 shows almost the same absorption coefficient as TiO 2 . The optical band gap of Se-doped TiO 2 can be evaluated by its absorption spectrum. The optical absorption near the band edge of a semiconductor often obeys the Kubelka-Munk equation: (αhν) n = A(hν − E g ), where A is a constant, hv is corresponding to photon energy, E g is the band gap of the semiconductor, α is the absorption coefficient, and n is 0.5 for indirect band gap materials, such as TiO 2 37,38 . The approximate values of the band gap can be obtained from the intercept of the tangent to X-axis, as shown in Fig. 4(b). The calculated band gap values are listed in Table 1. As can be seen, TSe10 shows the smallest E g of 2.19 eV. There are two competing effects which may influence the final E g of Se-doped TiO 2 . The Se doping may introduce additional electronic energy levels inside the band gap, which will effectively narrow the band gap 24 . The suppressed PLE peaks from TiO 2 host lattice, shown in Fig. S2 (Supplementary materials), give an indirect evidence for the introduction of extra energy levels from doped Se ions. However, with further increasing Se concentration higher than that of TSe5, the crystallite size becomes smaller. The band gap increases with a decrease in crystallite size due to the quantum size effect 27 .
PL spectra have been widely used to investigate the change of surface states of TiO 2 , the efficiency of charge carrier trapping, immigration and transfer to understand the fate of electron-hole pairs in semiconductor particles 39 . Figure 5 shows the PL spectra of Se-doped TiO 2 . A strong peak at 394 nm can be found for Se-doped TiO 2 , but lacking in TiO 2 , indicating the significantly suppressed electron-hole recombination in TiO 2 , which might be due to the high defects concentration 40 . Slight Se doping improves the crystalline structure, which may suppress the defects concentration and increase the electron-hole recombination rate. Clear PL peaks can be observed in Se-doped TiO 2 . However, with increasing Se doping concentration, more defects are formed, which provides more electron trapping centers. Interestingly, no PL peaks can be observed for TSe20, suggesting the strongly suppressed electron-hole recombination rate 40 . Further increasing Se doping concentration to TSe25, strong PL peak can be observed again. This indicates that there must be an optimum level up to which the addition of doping ions can help to lower electron-hole recombination rate. Otherwise, the excess doping ions act as recombination centers, which can dramatically decrease the photocatalytic activity of the doped catalyst 27 .
The typical time-dependent UV-Vis spectra of RhB solution by TiO 2 and TSe20 in photochemical reaction are shown in Fig. 6. The intensity of the characteristic absorption peak of RhB solution decreases with time. It can be seen that TiO 2 can effectively decompose RhB under the visible light irradiation. TSe20 shows much faster decomposition of RhB under visible irradiation, confirming the promotion of Sedoping on the photocatalytic activity of TiO 2 . The absorption peak position shifts to shorter wavelength, revealing that RhB is de-ethylated in a stepwise manner (i.e., ethyl groups are removed one by one as confirmed by the gradual peak wavelength shifts toward the blue region) 41 . However, after 45 minute, for RhB under irradiation in the presence of TSe20, the shift of wavelength reaches maximum. This indicates the decomposition mechanism of RhB changes to destroy its conjugated structure 42 . Photocatalytic reactions for the degradation of RhB aqueous solutions using Se-doped TiO 2 are shown in Fig. 7. For comparison, the photocatalytic activity of Degussa P25 was also measured. For the first 30 min in dark, it can be seen that Se-doped TiO 2 shows higher adsorption ability than TiO 2 . Since the crystallite size of all the samples is comparable, the surface area should not be the main reason. It is supposed to be due to more active sites on the surface of Se-doped TiO 2 43 . TiO 2 shows superior photocatalytic performance than Degussa P25. With Se doping, the photocatalytic capability is significantly improved. After irradiation for 30 min, nearly 91.3% of RhB was degraded by the sample TSe20, while only 60.3% Rhb was degraded by TiO 2 . Experimental dependencies of the molar concentration of RhB in the presence of undoped/doped TiO 2 and Degussa P25 powders in visible irradiation exhibit pseudo first order kinetics, as shown in Fig. 7(b) by plotting ln(C/Co) versus irradiation time,  t. The apparent reaction rate constant (k app ) is calculated from the slope of the curve, as illustrated in Fig. 7(c). With small concentration of Se doping, TSe5 shows much smaller k app value. With further increasing Se doping concentration, k app continuously increases. TSe20 exhibits the highest k app . With further increasing Se doping concentration, k app of TSe25 drastically decreases. Comparing the rate constant values under the visible light irradiations, the powder TSe20 attains the highest k app value of 0.088 min −1 , which is much higher than that of undoped TiO 2 (0.030 min −1 ) and Degussa P25 (0.0004 min −1 ). TiO 2 exhibits better photocatalytic activity than Degussa P25, mainly because of the synergetic effect of narrow band gap in the visible light range and mixed phases (orthorhombic and anatase) which hinder the recombination of generated electron-hole pairs 44 .
As shown in Fig. 8  Se 3p states and not populated by electrons. They are not donor states but allowed energy states hybridized with the Ti 3d states in the conduction band. The increase in the concentration of dopant Se 4+ introduces more electronic states into the band-gap, enhancing the density of electronic states in the gap. These intermediate energy levels offer additional steps for the absorption of low energy photons through the excitation of valence band electrons to these intermediate energy levels, from where they can be excited again to the conduction band. Furthermore, the photocatalytic capability also depends on the separation efficiency of generated electron-hole pairs. To improve the photocatalytic activity, the photoinduced electrons and holes inside the particles should survive during the transportation to the surface of particles, and the recombination should be avoided. As can be seen, no obvious peaks can be observed in the PL spectrum of TSe20, indicating the strongly suppressed recombination rate of electron-hole pairs. The defects, such as O vacancies, Ti 3+ , etc. may play significant role, since they can trap the electrons and suppress the recombination of electron-hole pairs 27 . With slight concentration of Se doping, the crystal structure is improved, thus defects concentration is decreased, leading to the worse photocatalytic capability for TSe5. With further increasing Se doping concentration, more defects will be introduced in Se-doped TiO 2 , the photo-generated electrons can be effectively trapped, and the recombination of electron-hole pairs is effectively suppressed, leading to the improved photocatalytic performance. As can be seen from PL measurements, TSe20 exhibits the negligible PL intensity, indicating the lowest recombination rate of electron-hole pairs. Together with the band gap in visible range, TSe20 exhibits the best photocatalytic performance. With further increasing the Se doping concentration beyond this optimum value, the average distance between the trapping sites is so small that large amount of them will be confined within the crystal lattice, increasing the recombination of electron-hole pairs, as can be seen the significant PL intensity for TSe25 27 . The photocatalytic performance of TSe25 decreases drastically. Thus, it can be concluded that the improved photocatalytic activity in Se-doped TiO 2 with optimum concentration (TSe20) is a synergetic contribution from the narrow band gap by Se 4+ doping and suppressed electron-hole pairs recombination due to the optimum defects concentration.

Conclusions
A series of Se-doped TiO 2 nanoparticles of anatase structure with various Se concentrations up to 17.1 at.% were prepared using sol-gel method. Slight concentration of Se doping has been confirmed to improve the crystalline structure of TiO 2 , while higher concentration of Se doping deteriorates the crystalline structure. The doped Se ions are confirmed to be mainly in the valence state of + 4, which provides extra electronic states in the band gap of TiO 2 . The band gap is effectively narrowed with the smallest gap energy of 2.17 eV by Se doping of concentration of 6.76 at.%. With further increasing Se doping concentration, Se doping significantly improves the photocatalytic activity of TiO 2 and 13.63 at.% Se-doped TiO 2 has the highest photocatalytic activity from the photo degradation of RhB in aqueous solution under visible light irradiation, which is attributed to the synergetic contribution of narrowed band gap by Se doping and suppressed electron-hole recombination due to the optimum defect density.