Unravelling the Efficient Photocatalytic Activity of Boron-induced Ti3+ Species in the Surface Layer of TiO2

Ti3+ species are highly unstable in air owing to their facile oxidation into Ti4+ species, and thus they cannot concentrate in the surface layer of TiO2 but are mainly present in its bulk. We report generation of abundant and stable Ti3+ species in the surface layer of TiO2 by boron doping for efficient utilization of solar irradiation. The resultant photocatalysts (denoted as B-TiO2−x) exhibit extremely high and stable solar-driven photocatalytic activity toward hydrogen production. The origin of the solar-light activity enhancement in the B-TiO2−x photocatalysts has been thoroughly investigated by various experimental techniques and density functional theory (DFT) calculations. The unique structure invoked by presence of sufficient interstitial boron atoms can lead to substantial variations in density of states of B-TiO2−x, which not only significantly narrow the band gap of TiO2 to improve its visible-light absorption, but also promote the photogenerated electron mobility to enhance its solar-light photocatalytic activity.

Scientific RepoRts | 6:34765 | DOI: 10.1038/srep34765 of TiO 2−x 21 . Thus, the sufficiently high Ti 3+ in TiO 2−x should be essential to enhance its photocatalytic activity under solar-light irradiation. Up to now, many strategies, such as heating under vacuum or in reducing gas (i.e. H 2 ), combustion, laser irradiation, and high-energy particle (such as electrons or Ar + ions) bombardment 13,22 , have been applied to synthesize TiO 2−x . These methods, started from pure TiO 2 , possess more or less limitations such as multiple steps, harsh synthesis conditions and expensive facilities. Most importantly, the prepared Ti 3+ species are highly unstable in air owing to the facile oxidation of Ti 3+ to Ti 4+ 23 , and thus they cannot concentrate in the surface layer but are mainly present in the bulk of TiO 2−x . Therefore, it is a big challenge to produce stable and sufficiently abundant Ti 3+ species, especially the surface-layer Ti 3+ , for the application of TiO 2 with a simple and inexpensive approach.
Herein, we report a novel and simple approach to generate stable and sufficiently abundant Ti 3+ in TiO 2 by boron doping via a sol-gel synthesis method. The resultant photocatalyst (denoted as B-TiO 2−x ) exhibits efficient visible and infrared optical absorption and extremely high and stable photocatalytic activity for hydrogen production. To gain insight into the active center structures and plausible mechanisms associated with the photocatalysts, the results from density functional theory (DFT) calculations were correlated to those obtained from experimental studies, such as UV-vis absorption spectroscopy, X-ray photoelectron spectroscopy (XPS), electron spin resonance (ESR) and nuclear magnetic resonance (NMR) spectroscopy etc. In particular, XPS and ESR were employed to detect the variation of the chemical state of B and Ti, while 11 B solid-state magic-angle-spinning (MAS) NMR was used to identify the local structures of the contingent dopants. As a result, the exceptionally high photocatalytic activity observed for the B-TiO 2−x is ascribed to the synergistic effect of the interstitial B and Ti 3+ species.

Results and Discussion
Solar absorption and photocatalytic activity. The optical response of as-prepared pure TiO 2 and B-TiO 2−x with different B doping was evaluated by the diffusive reflectance spectroscopy, as shown in Fig. 1a. The band gap of the B-TiO 2−x samples can be derived from the plot of transformed Kubelka-Monk function against photo energy ( Figure S1 in Supplementary Information, SI). It can be found that the band gap of pure TiO 2 is deduced approximately at 3.25 eV, and the band gap of TiO 2 narrows gradually from 3.20 to 3.0 eV with the amount of doped B increasing from 2 to 10% ( Figure S1 in SI). More interestingly, for the B-TiO 2−x samples, the intensity of visible-light absorption between 400 and 780 nm (the photo energy between 3.1 and 1.59 eV) increases gradually with the increase of the amount of doped B (Fig. 1a). Especially, for the 10% B-TiO 2−x samples, considerable visible-light absorption is still observable at ca. 1350 nm (corresponding to an optical energy at 0.92 eV) in the UV-Vis spectrum ( Figure S2 in SI). All these experimental results suggest that rich mid-gap states should exist in the band gap of B-TiO 2−x samples, and the B doping is responsible for the high visible-light absorption of B-TiO 2−x photocatalysts.
The B-TiO 2−x photocatalysts exhibit extremely high solar-driven photocatalytic activity toward H 2 production from water in the presence of methanol solution (5.0 volume %). Figure 1b shows the time course of H 2 evolution for the photocatalysts with different B doping. The photocatalytic activity generally increases with increasing the content of dopant. Particularly, when the amount of dopant B is about 10%, the B-TiO 2−x catalyst shows the best photocatalytic activity (also see Figure S3 in SI). We found that 1 hour of solar irradiation generates 0.059 mmol of H 2 using 0.005 g of the 10% B-TiO 2−x sample and 0.6% wt. Pt as the co-catalyst (11.8 mmol/hour/g of photocatalyst). The photocatalytic activity of 10% B-TiO 2−x is ca. 3.5 times as much as that of a commercial P25. (Fig. 1b).
Besides its excellent activity, as shown in Fig. 1c, the 10% B-TiO 2−x photocatalyst also exhibits excellent stability in the photocatalytic production of hydrogen under solar-light irradiation. There is no noticeable decrease in H 2 production rate for the 10% B-TiO 2−x sample in a 21 cycling test within a 42 h photocatalytic reaction period. The energy conversion efficiency, defined as the ratio of the energy of solar-produced H 2 and the energy of incident sunlight, reaches ca. 21% for the 10% B-TiO 2−x sample.
To establish the structure-activity relationship, the detailed structural characteristis of B-TiO 2−x photocatalysts was investigated with various analytical techniques, including X-ray photoelectron spectroscopy (XPS), electron spin resonance (ESR), transmission electron microscopy (TEM) and solid-state nuclear magnetic resonance (NMR).
Chemical states and electronic structure analysis. As shown in Fig. 2a, in the B 1s XPS spectra of the B-TiO 2−x samples, the only signal at 191.8 eV is ascribed to the doped B species weaving into the interstitial sites of TiO 2 lattice (B-O-Ti, called as interstitial B) 7 . Figure 2b shows the Ti 2p XPS spectra of as-prepared pure TiO 2 and B-TiO 2−x photocatalysts. Two major peaks at 459.0 (Ti 2p 3/2 ) and 464.7 eV (Ti 2p 1/2 ) are observable for pure TiO 2 , being indicative of the sole presence of Ti 4+ 24,25 . Upon introduction of 2% dopant B into TiO 2 , a new Ti 2p 3/2 XPS shoulder peak appears at 457.5 eV, which can be ascribed to the formation of Ti 3+ species 26,27 . As a surface analysis technique, the detection depth of XPS with the kinetic energy of 40 eV is normally less than 1.0 nm on metallic oxides 28 . Thus, the observed Ti 3+ species should mainly exist in the surface layer of B-TiO 2−x . By spectral deconvolution, it was found that the content of Ti 3+ amounts to ca. 10% of the total surface-layer Ti in the 2% B-TiO 2−x photocatalyst. Most interestingly, the Ti 2p 3/2 XPS signal of Ti 3+ at 457.5 eV gradually grows up at the expense of the signal of Ti 4+ at 459.0 eV with the amount of doped B being increased from 2% to 10% in the B-TiO 2−x photocatalysts, which unambiguously demonstrates that the additional B doping favors the formation of surface-layer Ti 3+ species. For the 10% B-TiO 2−x photocatalyst, the Ti 2p 1/2 XPS peak of Ti 3+ species is clearly observable at 457.3 eV and the content of Ti 3+ goes up to 87.5% of the total suface-layer Ti. Since Ti 3+ species is paramagnetic, its existence in B-TiO 2−x can be also confirmed by ESR experiments. As shown in Fig. 2c, an ESR signal due to Ti 3+ with a g-value of 1.987 29 is observable, and it gradually grows up with the amount of doped B being increased from 2% to 10%, indicating a gradual growth of Ti 3+ . For the 10% B-TiO 2−x sample, the concentration of Ti 3+ species is experimentally estimated to be ca. 30 μ mol/g. The high stability of the surface-layer Scientific RepoRts | 6:34765 | DOI: 10.1038/srep34765 Ti 3+ species is evidenced by the fact that no obvious change was observable for the ESR and XPS signals after the B-TiO 2−x photocatalysts were synthesized and exposed in atmosphere for 18 months.
Boron-induced structural disorder. As revealed by X-ray diffraction and high-resolution TEM, the as-synthesized pure TiO 2 is highly crystallized and exists in form of anatase phase, and the size of individual TiO 2 nanocrystals is about 21 nm in diameter (see Fig. 3a,b). After the interstitial B doping, the anatase phase remains almost unchanged, while the surface layer of B-TiO 2−x becomes disordered ( Fig. 3b and Figure S4 in SI), in which the thickness of disordered layer surrounding the crystalline core is ca. 1.5 nm for the 10% B-TiO 2−x sample. These structural properties of B-TiO 2−x photocatalysts were further examined by Raman scattering spectra (Fig. 3c). For the pure TiO 2 , six Raman-active modes of anatase phase occur with frequencies at ca. 142.5 (E g ), 196 (E g ), 396 (B 1g ), 515 (B 1g + A 1g ), and 640 (E g ) cm −1 , respectively 30 . For the B-TiO 2−x photocatalysts, the most intense peak at 142.5 cm −1 gradually broadens and blueshifts with the amount of doped B being increased from 2% to 10% (see the inset of Fig. 3c). All these features indicate that notable structural changes occur after the B doping, resulting in the structural disorder in the surface and subsurface layer of B-TiO 2−x , and the more the B doping the more the structural disorder. Similar conclusion can also be derived from the appearance of new signals at 242.4, 308.2, 358.9, 451.0, and 611.6 cm −1 . The structural disorders can lead to photon confinement effects, and activate Raman-forbidden modes by breaking down the Raman selection rule, generating the new Raman signals 11,31 . Combined with our XPS results, we conclude that sufficient interstitial B doping can generate not only sufficient surface-layer Ti 3+ species, but also structural disorder in the surface layer of B-TiO 2−x photocatalysts.
Boron environment analysis. The detailed local structure of dopant B in the B-TiO 2−x photocatalysts was thoroughly characterized by one-dimensional (1D) and two-dimensional (2D) solid-state 11 B MAS NMR techniques. When an observed nucleus is in close proximity to a paramagnetic center, its spin-lattice relaxation time (T 1 ) would be dramatically reduced. Therefore, we can utilize an extremely short recycle delay (3.0 ms) in the 2D 11 B triple-quantum z-filtering (3QZ) MAS NMR experiment to emphasize B sites that are in close proximity or bound to the paramagnetic center (surface-layer Ti 3+ ). As shown in Fig. 4a, four 11 B signals (B 1~4 sites) are resolved in the 11 B 3QZ MAS NMR spectrum of 10% B-TiO 2−x . According to the derived isotropic chemical shifts (Table 1), B 1~3 sites are assigned to tetrahedral-coordinated boron while the B 4 site is associated with tricoordinated boron 7 . When a long recycle delay of 2 s was used, the B 1 and B 2 sites could be much well resolved in the 11 B 3QZ MAS spectrum (Fig. 4a, insert, upper). In order to detect other possible boron sites with large quadrupolar interactions, we incorporated fast amplitude-modulated (FAM) radiofrequency (RF) pulse trains into the 3QZ MAS pulse sequence, namely the so-called 3QZ-FAM MAS NMR technique 32 , to explore the chemical environments of these boron sites. In addition to the B 4 signal, a new 11 B signal (B 5 site) with a typical second-order quadrupolar line shape appears in the 11 B 3QZ-FAM MAS NMR spectrum (Fig. 4a, insert, down) acquired with a long recycle delay of 2 s. Therefore, there are five different B sites present in 10% B-TiO 2−x , which can be utilized for the deconvolution of the 1D 11 B MAS spectrum (Fig. 4b) to determine the relative content of various B sites (Table 1).
We also measured the spin-lattice relaxation time (T 1 ) of the various B sites and the result was shown in Fig. 4c. Except for the B 5 site, the B 1~4 sites have extremely short T 1 , indicating that they are in spatial proximity to the surface-layer Ti 3+ center. Especially, the T 1 values of B 3 (2.4 ms) and B 4 (5.8 ms) sites are even much shorter than those of B 1 (42 ms) and B 2 (34 ms) sites, implying that the B 3 and B 4 sites may be directly bound to the surface-layer Ti 3+ center. Since the B 5 site exhibits a typical T 1 value (ca. 1000 ms), it should be far away from the surface-layer Ti 3+ center. According to their isotropic chemical shifts and our previous work 7 , the signals corresponding to B 1 and B 3 sites can be assigned to pseudotetrahedral coordinated interstitial boron with the latter being bound to the Ti 3+ center, while the signal corresponding to B 2 is due to the surface tetrahedral coordinated boron 7 . Likewise, the signals corresponding to B 4 and B 5 sites are assigned to tricoordinated interstitial boron with the former being bound to the surface-layer Ti 3+ center, and the latter being present in the bulk of B-TiO 2−x and far away from the Ti 3+ center. On the basis of our solid-state 11 (Table 1), which not only favors the formation of the Ti 3+ defects, but also provokes the surface-layer disorders.
Density of states (DOS) analysis and theoretical calculations. The DOSs of the valence band of as-synthesized pure TiO 2 and B-TiO 2−x photocatalysts were measured by valence band XPS technique (Fig. 5a).  A similar mid-gap state with an edge of maximum energy at ca. − 0.92 eV was also found in hydrogenated TiO 2 nanocrystals, which was due to surface-layer disorders 11 . The intensity of the mid-gap states of B-TiO 2−x gradually increases with the amount of doped B being increased. Additionally, the interstitial B doping can lead to a blue-shift for the maximum edge of the original valence band of TiO 2 (from 1.27 for pure TiO 2 to 0.97 eV for 10% B-TiO 2−x ), in good agreement with our theoretical results (see below), and the blue-shift increases with the increase of interstitial B doping (Fig. 5a, insert). In order to determine the structure of active centers and their influence on the electronic structure of the B-TiO 2−x , we performed theoretical calculations using 2 × 2 × 1 supercell models ( Fig. 5b and Figure S5 in SI). It was found that removing a oxygen atom from pure TiO 2 to form TiO 2−x needs to overcome an energy barrier of 132.7 kcal/mol, while this barrier is considerably decreased to 88.9 kcal/mol when a oxygen atom is removed from B-TiO 2 to form B-TiO 2−x , which demonstrates that the interstitial B doping favors the generation of Ti 3+ species in B-TiO 2−x . To further understand the origin of the change in the electronic and optical properties of B-TiO 2−x , we also calculated the energy band structures. The schematic illustrations of TiO 2 , TiO 2−x and B-TiO 2−x supercell models used in the calculations are shown in Fig. 5b and Figure S5 in SI, and the structural parameters of the relaxed structures are listed in Table S1 in SI. It is noteworthy that the supercell models of B-TiO 2−x were constructed on the basis of our NMR results. Compared with the standard anatase TiO 2 supercell, the TiO 2−x supercell contains variable O vacancies (associated with Ti 3+ sites). When one B atom weaves into the interstitial site around an O vacancy, denoted as 1B-TiO 2−x , the TiO 2−x structure is slightly distorted. With further increasing the amount of B and Ti 3+ in the B-TiO 2−x supercell, its lattice becomes more distorted (as for 2B-TiO 2−x , two O vacancies and two B atoms are present in this case), the symmetry of TiO 2 lattice is damaged, and eventually the lattice becomes disordered (as for 4B-TiO 2−x , four O vacancies and four B atoms are present in this case), in consistent with our TEM and Raman experimental observations. These structural changes lead to substantial variations in DOS of B-TiO 2−x . Figure 5c plots the calculated DOSs of B-TiO 2−x along with those of the pure TiO 2 and TiO 2−x . Compared with pure TiO 2 , the DOS of TiO 2−x shows new electronic states (Ti 3+ states) below the bottom of the conduction band at 0.8 eV (Fig. 5c), in consistent with the previous report 21 . Although the Ti 3+ state grows up gradually with the increase of the oxygen vacancy amount, the band gap of TiO 2−x remains almost unchanged (see Figure S6 in SI). For the DOS of 1B-TiO 2−x , a blue-shift (ca. 0.1 eV) is present simultaneously for both valence band and conduction band. In this case, its band gap is similar to that of TiO 2−x . When more B atoms and O vacancies are gradually doped into the lattice, beside the more blue-shift of valence band and conduction band, two groups of mid-gap states (ranging from ca. of 2B-TiO 2−x and 4B-TiO 2−x , accompanied by a significant reduction of the band gap. The high-energy mid-gap states (ca. − 1.4~0.5 eV) can be mainly ascribed to the formation of Ti 3+ (O vacancy) states, while the low-energy mid-gap states (ca. − 3.8~− 2.0 eV) are mainly due to the lattice disorder. The project DOSs of B-TiO 2−x are shown in Fig. 5d and Figure S6 in SI. It can be seen that the two groups of mid-gap states are hybridized with Ti, O, and B states. Therefore, the B atoms coupling to the neighboring lattice O and Ti atoms in the interstitial or defect sites of B-TiO 2−x do make a substantial contribution to the two mid-gap states, suggesting that the B doping can favor the formation and stabilization of the Ti 3+ species as well as the lattice disorder.
Understanding the high catalytic activity of B-TiO 2−x . Based on our experimental and theoretical results, a schematic illustration of the DOS of B-TiO 2−x is proposed in Fig. 5e. The band gap of pure TiO 2 is ca. 3.25 eV. According to our theoretical calculations, the existence of abundant Ti 3+ species induces an electronic state just below the conduction band minimum as shown in Fig. 5c and Figure S6 in SI, in consistent with the previous reports 21 . However, when the amount of Ti 3+ is high enough (i.e. in the TiO 2−x (4Vo) sample), more Ti 3+ states can be introduced in the band gap of DOS as shown in Figure S5, which may prevent the improvement of photocatalytic H 2 production due to the electric potential of some Ti 3+ states being less than that of H + reduction (ϕ(H + /H 2 )). When the additional interstitial B is doped, a blue-shift is simultaneously present for valence band and conduction band in the DOSs of B-TiO 2−x compared with the corresponding DOSs of TiO 2−x . Furthermore, when an appropriate amount of interstitial B (in the 2B-TiO 2−x ) is doped arround the Ti 3+ site, the blue-shift is also present for the Ti 3+ states ( Figure S6 in SI). The blue-shift of both conduction band and Ti 3+ states can result in a higher electric potential than that of TiO 2−x with respect to ϕ(H + /H 2 ), and thus easier H 2 production. It has been validated experimentally and theoretically that the sufficient interstitial B doping not only favors the formation of abundant Ti 3+ defects, but also provokes the surface-layer disorder. According to the valence band XPS spectra and theoretical calculations (Fig. 5), the surface-layer disorder causes the valence band maximum energy blue-shift by ca. 2.5 eV toward the vacuum level. Thus, a much narrowed band gap of ca. 0.47 eV is present in the DOS of B-TiO 2−x , in consistent with our optical measurement result (< 0.92 eV). As such, it can be concluded that the visible-light absorption of B-TiO 2−x would be much better than those of both TiO 2−x and pure TiO 2 . More interestingly, the structural changes (including the formation of B-O-Ti 3+ bond, the resultant structural distortion, and the surface disorder) can result in the formation of band tail state (Fig. 5). As a result, a better overlapping between the mid-gap states and the original energy band states of TiO 2 is present in the B-TiO 2−x photocatalysts as shown in Fig. 5e. Therefore, the band gap engineering can promote the mobility of photogenerated carriers (electron and hole), while the corresponding two mid-gap states, provoked by surface disorder and B-O-Ti 3+ structure, may trap shallowly the photogenerated carriers to prevent them from rapid recombination, which eventually improves the solar-light photocatalytic activity of B-TiO 2−x .

Conclusion
In summary, we have developed a novel and simple approach to generate and stabilize the sufficient Ti 3+ species in the surface layer of TiO 2 by introducing interstitial B. Our experimental and computational results unambiguously demonstrate that the introduction of interstitial B not only leads to the formation of abundant and stable Ti 3+ species but also generates structural disorder in the surface layer of B-TiO 2−x photocatalysts. These structural changes lead to substantial variations in their density of states, that is, two new groups of mid-gap states are formed, accompanied by a significant reduction of the band gap. These band gap engineering not only can promote the photogenerated electron mobility but also provide shallow trapping sites (B-O-Ti 3+ ) for photogenerated carriers to prevent them from rapid recombination, which eventually improves the solar-light photocatalytic activity of B-TiO 2−x toward hydrogen production water splitting. The experimental and theoretical calculation results presented herein should not only facilitate a better understanding of the photocatalytic mechanism at the atomic level but also be helpful for rational design of highly efficient titania-based photocatalysts.

Synthesis.
A series of B-TiO 2−x photocatalysts were prepared with a precursor solution consisting of titanium tetrachloride, boric acid, ethanol, and deionized water, with molar ratios of TiCl 4 /H 3 BO 3 /NaCl/H 2 O/ethanol at 1 : × : 2 : 60 : 20 (x = 0, 0.02, 0.05, and 0.10). The solution was stirred for ca. 2.5 days in ice bath, and then its pH value was adjusted to 7.0 by ammonia. The mixture was aged for 24 h at room temperature, washed with water to remove NH 4 Cl and NaCl during the filtration, and then dried at 80 °C. After being calcined at 450 °C in N 2 /air for 5 h, the resultant grey and blue powder was well-crystallized with features assignable to anatase crystalline phase as determined by X-ray diffraction. The actual B doping amount (molar ratio, B/Ti) is deterimned by solid-state NMR analysis, and it is ca. 1.6%, 3.9%, and 7.2% in 2% B-TiO 2−x , 5% B-TiO 2−x , and 10% B-TiO 2−x , respectively. Photocatalytic activity. After loading a sample (0.005 g) with 0.6% Pt (0.00003 g), the photocatalyst (Pt/B-TiO 2−x ) was placed into an aqueous methanol solution (5.0 ml, 5.0 volume %. Concerning Pt deposition, 0.2 g of the as-prepared TiO 2 photocatalysts were dispersed in a 10 mL aqueous solution (0.618 mM H 2 PtCl 6 • 6H 2 O) with stirring, which was irradiated using a 300 W Hg lamp for 2 h. Methanol acts as the sacrificial reagent in a Quartz glass container. The simulated solar light by Xe lamp (300 W) was used as the excitation source, which produced a power of ca. 130 mW/cm 2 . The hydrogen gas in reactor was replaced with nitrogen for 30 min before each cycling test. The energy conversion efficiency was calculated based on the equation 2H/P. Here H was the number of moles of hydrogen gas produced during a specific time, while P was the number of photons incident on the photocatalytic system within the same time period. Characterization Methods. The crystalline structures of various photocatalysts were determined by X-ray diffraction (XRD) on a Bruker D/max2550 instrument using Cu KR radiation (40 kV, 40 kmA). All UV-Vis diffuse reflectance spectroscopic (DRS) studies were carried out on an Agilent Technologies Cary 4000/5000