As a promising strategy to address environmental and energy issues, photocatalysis is attracting increasing attention [1−3]. Though extensive research on photocatalysis has been conducted, the crucial factors determining the efficiency of photocatalysis are still not well understood. Usually, photocatalytic reactions involve three major steps: (1) generation of electron-hole pairs on a semiconductor by absorption of photons with energy equal to or greater than its bandgap; (2) separation of electron-hole pairs and migration of these charge carriers to the surface of the semiconductor; (3) surface reduction and oxidation reactions involving electrons and holes, respectively, at the semiconductor surface. These three steps correspond to the three crucial factors of light absorption, charge separation and reaction kinetics. Among these, charge separation is the most complicated and important step.

Among semiconductor photocatalysts, TiO₂ is the earliest and most extensively investigated photocatalyst because it is cheap, stable, nontoxic, and environmentally friendly. Since the work by Fujishima and Honda published in 1972 ^[4], research on TiO₂ has dramatically increased and been extended to applications in environmental purification, self-cleaning, water splitting, CO₂ reduction, organic synthesis, and solar cells. An interesting phenomenon of TiO₂ is that the highest activity is usually not obtained for its two major crystalline forms, anatase and rutile, but from samples with a mixture of anatase and rutile phases. P25, a commercial TiO₂ material with mixed anatase and rutile phases, is the best example of this phenomenon. This produces questions about the intrinsic qualities that affect photocatalytic activity, and the step that the anatase: rutile mixed phase promotes, which need to be answered. Recently, the anatase:rutile mixed phase structures of TiO₂ have been systematically investigated [5−9]. The formation of anatase: rutile surface phase junctions has been confirmed, and is regarded as the main factor inducing the high photocatalytic activity of TiO₂ with mixed phase structure because of its excellent charge separation ability.

In this feature article, some sophisticated techniques and how they are used to characterize the surface phase structure of TiO₂ are first introduced. Then, research on the discovery of anatase: rutile TiO₂ phase junctions and how this concept has directed subsequent investigations is reviewed. Finally, the investigation of phase junctions by time-resolved mid-infrared (IR) spectroscopy to examine the mechanism behind its charge separation ability is discussed. This feature article also covers some other important factors related to the performance of TiO₂ photocatalysts.

As mentioned above, the phase structure of TiO₂ strongly affects its photocatalytic activity. Numerous techniques have been used to characterize the structure of TiO₂, such as X-ray diffraction (XRD), transmission electron microscopy (TEM), and Raman and Fourier transform infrared (FT-IR) spectroscopies. Herein, three typical spectroscopic techniques used to investigate the phase structure of TiO₂ are introduced. The unique features of these techniques can further our understanding of the structure and even the carrier dynamics of TiO₂.

In heterogeneous catalysis, reactions take place on the surface of the catalysts. Therefore, the catalytic performance largely depends on the surface properties of the catalyst. Ultraviolet (UV) Raman spectroscopy has recently been proved a powerful tool in materials science ^[10], mainly because it can efficiently avoid the interference from fluorescence in visible Raman spectra and simultaneously enhance the Raman signal because of its short wavelength and the resonance Raman effect ^[11]. Here, UV Raman spectroscopy is used to monitor the phase transformation process of TiO₂ ^[6]. Because TiO₂ strongly absorbs UV light, UV Raman spectroscopy is more sensitive to the surface structure of TiO₂ than XRD and visible Raman spectroscopy. The results obtained from these techniques for the phase transformation process of TiO₂ disagree, which indicates that the surface region of anatase phase detected by UV Raman spectra is more stable at higher temperatures than that in bulk TiO₂ detected by XRD and visible Raman spectra. TEM images show that the phase transformation process of TiO₂ is accompanied by growth of anatase particles. If the surface of anatase particles is covered by dispersed lanthanum oxide (La₂O₃), the phase transformation can be efficiently restrained ^[6]. Based on these results, a clear phase transformation mechanism for TiO₂ is shown in Fig. 1. With increasing temperature, the phase transformation process of TiO₂ initiates from the surface of integrated anatase particles. The bulk of the new larger particle then gradually transforms to rutile phase. Finally, the transformation is completed within the entire particle.

Fig. 1. The proposed mechanism for the phase transformation of TiO2 at elevated temperatures Reprinted with the permission from Ref. [11]. Copyright 2009 Editorial Office of Chinese Journal of Catalysis, Dalian Institute of Chemical Physics, Chinese Academy of Sciences.

Fig. 2. Proposed mechanisms for the phase transformation of TiO2 with various particle size (a) smaller than 10 nm; (b) in the range of 10-60 nm; (c) larger than 60 nm. Reprinted with the permission from Ref. [11]. Copyright 2009 Editorial Office of Chinese Journal of Catalysis, Dalian Institute of Chemical Physics, Chinese Academy of Sciences.

Actually, the mechanism of phase transformation of TiO₂ is more complicated than one simple process. The following studies show that phase transformation of TiO₂ is a size-dependent process, where size determines not only the onset transition temperature but also the nucleation performance ^[7]. UV Raman spectroscopy is used to determine the variation of the surface phase structure of TiO₂, while XRD is used to examine the variation of the bulk. As depicted in Fig. 2, the phase transformation temperature increases with the initial particle size. For small particles (< 60 nm), rutile nucleates at the interfaces between contacting anatase grains. Conversely, for large particles (> 60 nm), the free surface, interface and bulk of anatase TiO₂ can all function as rutile nucleation sites. UV Raman spectroscopy allows the difference between the surface and bulk transformation processes of TiO₂ to be distinguished.

Although UV Raman spectroscopy can in principle provide more information about surfaces than the bulk for materials that absorb UV light, direct evidence of surface behavior is still needed. Therefore, IR spectra of probe molecules adsorbed on TiO₂ were used to support the results obtained from UV Raman spectroscopy ^[12]. Such FT-IR spectroscopy is more complicated than UV Raman spectroscopy but undoubtedly a surface technique. Because of the different surface properties of anatase and rutile TiO₂, carbon monoxide (CO) and carbon dioxide (CO₂) can exhibit different adsorption behavior on these two phases. For example, CO is weakly adsorbed on Ti⁴⁺ ions of the anatase phase, but hardly adsorbed on the rutile phase at room temperature. Moreover, CO₂ mainly adsorbs as bidentate carbonate on the anatase phase, but adsorbs as a bicarbonate species on the rutile phase. Therefore, IR spectra of CO and CO₂ adsorbed on TiO₂ can reveal its surface structure. The IR spectra agreed well with the findings of UV Raman spectra, but differed from those of XRD ^[12], further confirming that UV Raman spectroscopy is a surface-sensitive technique for TiO₂.

Photoluminescence (PL) spectroscopy is used to characterize the excited electronic states of materials. For semiconductor materials, PL spectroscopy can supply meaningful information about crystal structures, defect states and charge carrier dynamics. As an indirect wide-band-gap semiconductor, TiO₂ always displays trap-related luminescence bands, while the band-edge luminescence is difficult to observe. The position and intensity of PL from TiO₂ strongly depends on its crystal structure, particle size, dopants, impurities, annealing temperature and atmospheric environment during crystallization, and the ambient temperature and atmosphere during the PL measurements. The main luminescence features of TiO₂ are broad structureless emission bands. Anatase TiO₂ has a visible emission, which is attributed to self-trapped excitons ^[13], oxygen vacancies ^[14], defect sites ^[15], impurities or reduced metal ions ^[16], etc. Meanwhile, rutile TiO₂ exhibits a near-infrared (NIR) luminescence that is assigned to Cr³⁺ impurities ^[17], interstitial Ti³⁺ ions ^[16], intermediate species generated during the photooxidation reaction of water ^[18], or the intrinsic defects of rutile TiO₂ ^[19].

The PL characteristics of TiO₂ during phase transformation, in combination with UV Raman spectroscopy, reveal the progression of TiO₂ phase structure ^[20]. Anatase TiO₂ exhibits a visible emission band centered at about 505 nm, while rutile TiO₂ has a NIR emission band centered at about 835 nm. Comparison of the PL and UV Raman spectra of TiO₂ confirmed that the luminescence band position of TiO₂ is related to its crystalline structure. Namely, the visible emission band is related to anatase structure, while the NIR emission band is associated with rutile structure (Fig. 3). The visible emission of anatase and the NIR emission of the rutile phase both exhibit extremely long lifetimes of up to milliseconds under weak excitation conditions ^[21]. The power-law decay of the luminescence indicates that PL from both anatase and rutile TiO₂ has a close relation with trap states. The visible emission band of the anatase phase is assigned to donor-acceptor recombination, with mainly oxygen vacancies and hydroxyl groups as donor and acceptor sites, respectively. The NIR luminescence from rutile TiO₂ originates from the recombination of trapped electrons with free holes (Fig. 4).

Fig. 3. (a) Photoluminescence spectra of TiO2 calcined at different temperatures. (b) Schematic illustration of the photoluminescence evolution with the phase transformation of TiO2. Reprinted with the permission from Ref. [20]. Copyright 2007 American Chemical Society.

Fig. 4. Schematic illustration of relaxation processes of the photogenerated carriers in anatase (a) and rutile TiO2 (b). (A) Electrons quenched to the bottom of the conduction band, (B) carriers captured by traps, (C) thermal excitation from traps, (D) electrons captured directly by deep traps, (E) carriers relax to luminescent sites, (F) nonradiative recombination process. Reprinted with the permission from Ref. [21]. Copyright 2010 Royal Society of Chemistry. Acknowledgements to be used by RSC authors.

Based on the relationship between PL bands and the crystalline structure of TiO₂, PL measurements were used to study the role of trap states of TiO₂ in photocatalysis. The visible emission band of anatase TiO₂ is easily quenched by the deposition of Pt on its surface, while the NIR luminescence band of rutile TiO₂ is not ^[20]. The lifetimes of anatase TiO₂ shortened considerably after it was treated under vacuum at 150 ℃ for 1 h, because of the desorption of H₂O molecules and change of surface hydroxyl groups ^[21]. In contrast, the luminescence decay profiles of NIR luminescence from rutile TiO₂ were not affected by the same thermal treatment. These results suggest that the trapped carriers in anatase TiO₂ may be involved in the photocatalytic reactions and the slow decay processes may be beneficial for photocatalysis performance, while the trapped carriers in rutile TiO₂ may be less likely to participate in photocatalytic reaction processes.

The above characterization results provide us with an opportunity to better understand the phase transformation mechanism and structure of TiO₂, which may in turn allow us to optimize the performance of TiO₂ in photocatalysis by revealing what we can do to improve its activity. TiO₂ samples with different surface and bulk crystalline phases were prepared by thermal treatment of Ti(OH)₄ in air from 500 to 800 ℃ ^[5]. The rutile content of the bulk region and anatase content of the surface region were estimated from visible and UV Raman spectra, respectively. Maximum activity was observed for TiO₂ samples calcined at 700-750 ℃, where the bulk of TiO₂ was mostly rutile while the surface was a mixture of anatase and rutile phases. When the calcination temperature was further increased to 800 ℃, the anatase phase at the surface region was completely transformed to rutile and correspondingly the photocatalytic activity decreased dramatically. These results show that the presence of anatase phase on the surface of rutile particles can maintain high photocatalytic activity. This led to the concept of surface phase junctions formed between anatase and rutile, which is proposed to facilitate charge separation at the surface of TiO₂. To further clarify this concept, anatase nanoparticles were deposited on rutile particles by a wet-impregnation method followed by thermal treatment. As shown in Fig. 5(a), the photocatalytic activity of the samples increased with the amount of anatase nanoparticles up to a certain threshold. However, when the amount of anatase was increased above the threshold, the activity decreased. This is because too many anatase nanoparticles will decrease the area of phase junctions exposed on the surface of TiO₂. The high-resolution transmission electron microscopy (HRTEM) image in Fig. 5(b) reveals that the loaded anatase nanoparticles closely contacted with rutile particles to form phase junctions.

Fig. 5. (a) Photocatalytic activities for H2 evolution of rutile TiO2 loaded with increasing amount of anatase nanoparticles. (b) Surface phase-junction formed between anatase and rutile. Reprinted with permission from Ref. [5]. Copyright 2008 Wiley-VCH.

Degussa P25 is composed of approximately 80% anatase and 20% rutile phases and is a benchmark type of TiO₂ because of its high photocatalytic activity. The synergistic effect between anatase and rutile phases is considered the origin of its excellent performance. The activity of P25 can be further increased via elaborately controlled thermal treatment ^[8], which transforms P25 to a catalyst with optimized anatase-rutile structure. As shown in Fig. 6, in the photocatalytic reforming of methanol, propanetriol and glucose, the overall activity of P25 can be enhanced up to 3-5 times by thermal treatment. These results also show that the degree of crystallization is not the major cause of the high photocatalytic activity of thermally treated P25, because further increasing the temperature during thermal treatment did not increase its photocatalytic activity. Therefore, the optimized phase structure obtained by carefully controlled thermal treatment of P25 is the main contributor to the enhancement of its photocatalytic activity.

Fig. 6. (a) Average hydrogen production in photocatalytic reforming methanol, propanetriol and glucose and (b) surface-specific photocatalytic activity for hydrogen production in photocatalytic reforming methanol on Pt/P25-w%R photocatalysts (where w indicates the rutile content estimated by XRD. For P25-100%R-a, b, c, d photocatalysts, where a, b, c and d indicate different treatment conditions), the surface areas of P25-w%R photocatalysts are also displayed (dash dot). Reprinted with the permission from Ref. [8]. Copyright 2011 Elsevier.

As the importance of phase junctions has been realized, new methods to prepare TiO₂ that allow phase control have attracted increasing attention ^[22]. The phase transformation of TiO₂ from anatase to rutile can be restrained by surface modification with Na₂SO₄ ^[9]. When the content of SO₄²⁻ in TiO₂ was increased from 0 to 3 wt%, the percentage of anatase in the surface region increased from 2% to 75%. It should be mentioned that this method can be used to prepare TiO₂ with different phase structures at the same temperature, which facilitates comparison of these catalysts. Compared with P25, the as-prepared TiO₂-SO₄²⁻ samples showed activities for H₂ production via photocatalytic reforming of methanol that were up to six times higher. TEM images revealed that small anatase particles were dispersed on the edge of large rutile particles to form the junction structure. Because the junction formed during the phase transformation process, intimate contact formed between anatase and rutile phases. The example in Fig. 7 shows that the junction is formed at the atomic level, which consequently facilitates charge transfer. Besides Na₂SO₄, other additives such as NaNO₃, NaHCO₃, Na₃PO₄, Na₂SiO₃, and Na₂MoO₄ can also be used to control the phase of TiO₂ ^[23]. However, the photocatalytic activity of TiO₂ is not only influenced by its phase structure, but also by its surface chemical properties. Therefore, poorer performance might be obtained using certain control agents even when a perfect phase-junction structure is formed.

Since the anatase:rutile phase-junction concept was proposed, it has become a guideline for the preparation of TiO₂ nanoparticles ^{[24, 25]}, nanorod arrays ^[26], and nanobelts ^[27]. Similar junctions, such as anatase: TiO₂(B) phase junctions [28−33], and even other semiconductors such as the α:β phase junction of Ga₂O₃, have since been successfully used in photocatalytic reactions ^[34].

Besides phase junctions, some other aspects that affect the performance of TiO₂ photocatalysts, such as cocatalysts, by-products, and reaction environment, should also be considered. Recently, a synergetic effect of dual cocatalysts Pd-IrO_x loaded on TiO₂ in photocatalytic H₂ production reactions was found ^[35]. Pd and IrO_x particles located close together showed better performance than separated ones, which is quite different from the common situation. Undesired by-product CO obtained from the reactions of photocatalytic H₂ production via reforming of methanol or other biomass derivatives is always a problem, because there is a very strict CO limit that must be met for fuel cell applications ^[36]. CO-suppressing methods such as addition of a small amount of inorganic anions ^[37], using a smaller size of cocatalyst ^[38], and controlling the surface acidity of TiO₂ ^{[8, 9, 23]} have been developed.

Fig. 7. Atomic level contact of anatase: rutile phase-junction. Reprinted with the permission from Ref. [9]. Copyright 2012 Royal Society of Chemistry. Acknowledgements to be used by RSC authors.

Meanwhile, the overall water splitting reaction is seldom stoichiometric on TiO₂. However, it has been found that the addition of Cl⁻ can facilitate the water splitting reaction ^[39]. Cl⁻ can not only trigger O₂ evolution, but also promote the H₂ evolution simultaneously. Besides overall water splitting, TiO₂-based catalysts can also be used to convert biomass to value-added chemicals. For example, transformation of glycerol to hydroxyacetaldehyde ^[40] and glucose to sugar aldose ^[41] have been achieved recently, extending the scope of TiO₂ photocatalysts.

Anatase:rutile mixed-phase TiO₂ is more active than anatase or rutile TiO₂ in photocatalytic reactions. It has been proposed that anatase: rutile phase junctions are responsible for the improved photoactivity. Such phase junctions are supposed to increase charge separation and then prolong charge lifetime to facilitate photocatalytic reactions. Many researchers are devoted to confirming the role of anatase:rutile phase junctions in photocatalysis both experimentally and theoretically.

Based on electron spin resonance experiments, Hurum et al. ^[42] claimed that the electron transfer from rutile to anatase occurred at the transition points between anatase and rutile in Degussa P25, while Komaguchi et al. ^[43] proposed that photoinduced electron transfer proceeded from anatase to rutile in partially reduced P25. Kawahara et al. ^[44] suggested that the high photocatalytic activity of P25 was mainly caused by the increase in charge separation efficiency resulting from interfacial electron transfer from anatase to rutile phases based on TEM analysis of patterned TiO₂(anatase)/TiO₂(rutile) bilayer-type photocatalysts.

Time-resolved spectroscopic techniques have also been used to study mixed-phase TiO₂. Carneiro et al. ^[45] claimed that photogenerated holes were captured by rutile in mixed-phase TiO₂ according to the results of time-resolved microwave conductance spectroscopy of mixed-phase TiO₂. The kinetics of photoinduced electrons in anatase, rutile, and mixed-phase TiO₂ have also been studied using time-resolved mid-IR spectroscopy ^[46], which has been proved to be a powerful tool to monitor the kinetics of photogenerated electrons in semiconductor photocatalysts ^[47]. It was found that the dynamics of the transient mid-IR absorption of anatase and rutile were different. Anatase showed a transient mid-IR absorption signal, whereas rutile did not display any detectable mid-IR absorption on the microsecond time scale. Based on the relationship between initial mid-IR absorption and phase composition, a charge transfer process across the anatase:rutile phase junction was confirmed, and electron transfer from anatase to rutile was proposed to occur in mixed-phase TiO₂ (Fig. 8).

Fig. 8. (a) Transient mid-IR absorption decays of (1) anatase, (2) anatase-rutile mixed phase TiO2-600, (3)anatase-rutile mixed phase TiO2-700, and (4) rutile in a vacuum excited by 355 nm laser pulse of 6−8 ns duration. (b) Schematic band structure of anatase and rutile under UV light irradiation. Reprinted with the permission from Ref. [50]. Copyright 2014 American Chemical Society.

To understand the charge transfer at the interfaces of anatase and rutile phases, the band alignment between anatase and rutile has been calculated theoretically. Kang et al. ^[48] claimed that the conduction band minimum of anatase was about 0.2 eV higher than that of rutile with similar valence band maxima for anatase and rutile, indicating that electron transfer from anatase to rutile is more likely than vice versa. Conversely, Scanlon et al. ^[49] reported that a type II staggered band alignment of 0.4 eV existed between anatase and rutile with rutile possessing the higher conduction band minimum, which is beneficial for electron transfer from rutile to anatase.

To date, the direction of charge transfer at anatase:rutile phase junctions and its role in photocatalysis are still debated, even though much work has been done using various techniques as discussed above. These points are still debated because charge-transfer behavior strongly depends on both the phase composition and preparation method of mixed-phase TiO₂. Therefore, further work is still required to understand anatase:rutile junctions in greater detail and optimize the photocatalytic performance of TiO₂.

In this feature article, characterization techniques used to investigate the phase structure of TiO₂ were first introduced, including UV Raman, FT-IR and PL spectroscopies. UV Raman spectroscopy and FT-IR spectroscopy of adsorbed probe molecules provided unique evidence for the surface phase structure and phase transformation of TiO₂. PL spectra not only revealed the phase structure of TiO₂ and its emission-related trap states, but also the role of trap states in photocatalysis. The roles of surface phase junctions of TiO₂ in photocatalytic H₂ evolution reactions were then reviewed in detail. The presence of anatase:rutile surface phase junctions can increase the photocatalytic activity of TiO₂-based materials, which is ascribed to efficient charge separation between anatase and rutile phases. Some other factors that can affect the performance of photocatalysts in different ways, such as cocatalysts, by-products, and reaction environment, were then briefly introduced. Finally, the mechanism of charge transfer at anatase:rutile phase junctions was discussed. Electron transfer from anatase to rutile TiO₂ was confirmed by time-resolved mid-IR spectroscopy.

Conversion of solar energy into fuels via photocatalysis is considered a promising way to solve energy and environmental issues. As a benchmark photocatalyst, TiO₂ is the most widely investigated semiconductor photocatalyst. However, the majority of research in this field still focuses on environmental photocatalysis, mainly involving the degradation of pollutants such as organic wastes and dyes. Increasing attention is now being paid to fuel generation reactions such as overall water splitting and CO₂ reduction. However, the photocatalytic H₂ production in the presence of sacrificial reagent achieved to date cannot be considered true water splitting. Despite this, these reactions can provide meaningful insights into the mechanism of photocatalytic reactions to aid design of improved photocatalysts.

It should be mentioned that TiO₂ may not be a promising catalyst for photocatalytic water splitting reaction because of its intrinsic properties, such as only UV light absorption. It is also not expected that a great breakthrough on TiO₂ can be achieved in the near future. However, strategies such as extending the light absorption range of TiO₂ and combination with other materials to construct a photocatalyst system may overcome some disadvantages of TiO₂ itself. Furthermore, TiO₂ is still an ideal model for understanding the mechanisms of photocatalytic processes, which will help in the design and construction of more efficient photocatalyst systems. Therefore, research on TiO₂ will continue.