Reactivity of Trapped and Accumulated Electrons in Titanium Dioxide Photocatalysis

Abstract

Electrons, photogenerated in conduction bands (CB) and trapped in electron trap defects (Ti_ds) in titanium dioxide (TiO₂), play crucial roles in characteristic reductive reactions. This review summarizes the recent progress in the research on electron transfer in photo-excited TiO₂. Particularly, the reactivity of electrons accumulated in CB and trapped at Ti_ds on TiO₂ is highlighted in the reduction of molecular oxygen and molecular nitrogen, and the hydrogenation and dehalogenation of organic substrates. Finally, the prospects for developing highly active TiO₂ photocatalysts are discussed.

1. Introduction

Since Fujishima and Honda discovered photoelectrochemical water splitting on titanium dioxide (TiO₂) photoelectrodes in the early 1970s [1], TiO₂ photocatalysis has been applied in various fields, such as the storage of solar energy [2,3,4,5], environmental purification [6], organic synthesis [7,8,9,10,11], anti-bacterial applications [12], and anti-fogging treatments [12,13]. These characteristic photo-functionalities are induced by incident light, in which the behavior of photogenerated electrons and holes, as well as the roles of defects formed on surface and in lattice, are of particular importance. The defect sites are the recombination centers for the photogenerated electrons and holes, because photocatalytic activities decrease with increasing the amount of defects created [2,6,8]. However, Amano et al. reported that the introduction of defect states in TiO₂ with H₂ reduction treatment greatly enhanced the photocatalytic activity for the water oxidation reaction in aqueous solution [14,15]. Moreover, Kong and coworkers claimed that tuning the relative concentration ratio of bulk defects/surface defects in TiO₂ nanocrystal improves the separation efficiencies of photogenerated electrons and holes, thereby enhancing the photocatalytic activity [16]. Thus, further understanding of the defects in TiO₂ necessitates the development of highly active photocatalysts.

The properties of defects—such as energy levels, structures, and interactions with adsorbates—have been reviewed by Diebold [17], Henderson [18], and Nowotny [19,20] in detail, but many unanswered questions remain. Recent studies in this field have made the considerable progress during the last decade. This review summarizes the recent progress in the research on the defects in TiO₂. Herein, we focus on the properties of electron trap defects formed within the bandgap of TiO₂ associated with Ti defects, specifically the intra-bandgap Ti states (Ti_ds). Firstly, the fate of photogenerated electron and holes in TiO₂ are described with respect to Ti_ds and hole trap sites in Section 2. Next, the origin of Ti_ds and their energy distribution in TiO₂ are considered in Section 3. In Section 4, the reactivity of electrons trapped at Ti_ds and accumulated in the conduction band (CB) on the representative reductive reactions are highlighted. Finally, the prospect for developing a highly active TiO₂ catalyst is discussed.

2. Fate of Photogenerated Electrons and Holes in TiO₂

Although several models exist for the charge transport, trapping, and the reaction of photogenerated electrons and holes on photoexcited TiO₂, we adopted a schematic model for the anatase TiO₂ based on the recent selected reviews and reports as illustrated in Figure 1 [11,21,22,23,24,25].

This model consists of several steps:

Step 1.

Electron–hole pair generation

TiO₂ + hν → TiO₂ (e⁻ + h⁺)

[<100 fs]

Step 2.

Trapping CB electrons (e_cb⁻) at defect Ti⁴⁺ sites

Ti_ds⁴⁺ + e_cb⁻ → Ti_ds³⁺

[100 fs–500 ps]

Step 3.

Trapping valence band holes (h_vb⁺) at terminal Ti–OH or surface Ti–O–Ti sites

Ti–O_sH or Ti–O_s–Ti + h_vb⁺ → Ti–O_sH^・+ or Ti–O_s^・+−Ti

[<100 fs–200 fs]

Step 4.

Reduction of adsorbed electron acceptor (A_ad) with e_cb⁻ at reduction sites

e_cb⁻ + A_ad → A_ad^・−

[>10 ns]

Step 5.

Reduction of A_ad with electrons trapped at defect sites (Ti_ds³⁺)

Ti_ds³⁺ + A_ad → Ti_ds⁴⁺ + A_ad^・−

[slow process]

Step 6.

Oxidation of adsorbed electron donor (D_ad) by trapped holes at oxidation sites

Ti–O_sH^・+ or Ti–O_s^・+–Ti + D_ad → Ti–O_sH or Ti–O_s–Ti + D_ad^・+

[100 ps–10 ns]

Step 7.

Recombination of e_cb⁻ with trapped holes

e_cb⁻ + Ti–O_sH^・+ or Ti–O_s^・+–Ti → Ti–O_sH or Ti–O_s–Ti

[1–10 ps]

Step 8.

Recombination of Ti_ds³⁺ with trapped holes

Ti_ds³⁺ + Ti–O_sH^・+ or Ti–O_s^・+–Ti → Ti_dt⁴⁺ + Ti–O_sH or Ti–O_s–Ti

[>20 ns]

where time scales for each step are described in brackets [21,22,23,24]. The time scales depend on the crystalline phases, crystallinity, specific surface area, and the presence of bulk and surface defect states in TiO₂. The following assumptions were applied to this model: (a) CB electrons (e_cb⁻) contain both electrons in CB and electrons trapped at shallow sites, located just below the CB edge of TiO₂ within 0–0.05 eV. These electrons were assumed to be in thermal equilibrium in the bulk CB and at the shallow trap sites; (b) Valence band holes (h_vb⁺) are rapidly transported to the surface hole trap sites (Ti–O_sH or Ti–O_s–Ti) (Step 3); (c) trapped holes (Ti–O_sH^・+ or Ti–O_s^・+–Ti) are the main oxidants for the adsorbed electron donor (D_ad) (Step 6); and (d) charge carrier recombination occurs between e_cb⁻ and holes trapped at the surface trap sites (Step 7), as well as between electrons trapped at Ti defect states (Ti_ds³⁺) and holes trapped at the surface trap sites (Step 8), whereas the interband electron–hole carrier recombination (e⁻ + h⁺→hv or heat) is negligible.

These assumptions can be justified as follows. Tamaki and coworkers described the charge carrier dynamics under weak excitation conditions for nano-crystalline anatase TiO₂ samples in femtosecond to microsecond time scales [22,23], which should be compatible with the actual photocatalytic reactions under the usual UV irradiation conditions. They observed the e_cb⁻ and h_vb⁺ pair generation within 100 fs, and the e_cb⁻ migration between CB and shallow trap sites in equilibrium. These electrons then relaxed to deep trap sites (Ti_ds) with an approximate 500 ps time constant. Meanwhile, h_vb⁺ was rapidly trapped to the surface terminal Ti–O_sH sites within 100 fs to create Ti–O_sH^・+ [22,23]. If the photoinduced event occurred in alcohols, the lifetime of the Ti–O_sH^・+ generated on the TiO₂ surface would be in the nanosecond or sub-nanosecond time scale (approximately 0.1–3 ns in alcohols) due to the fast reaction of Ti–O_sH^・+ with the abundant alcohol adsorbed on the TiO₂ surface [24]. Therefore, the free h_vb⁺ rarely presents in the bulk or on the surface of TiO₂, so that e_cb⁻ may recombine only with the trapped holes.

3. Origin and Energy Distribution of Electron Trap Defects (Ti_ds)

The bulk and surface Ti_ds are formed in reduced or doped TiO₂ in both rutile and anatase phases [17,18,19,20]. As depicted in Diebold’s review [17], the bulk Ti_ds are easily created in the rutile single crystal by thermal annealing in a vacuum, resulting in the formation of blue color centers, indicating high conductivity. Therefore, TiO₂ is classified as an n-type semiconductor. The H₂ reduction of TiO₂ creates both oxygen vacancies and Ti³⁺ ions, which is an electron trapped in a Ti⁴⁺ lattice site, as described in Reaction (1) using Kröger–Vink notation [14,15,19,20]

$${\mathrm{O}}_{\mathrm{O}}^{\mathrm{x}}+2{\mathrm{Ti}}_{\mathrm{Ti}}^{\mathrm{x}}+{\mathrm{H}}_2\to {\mathrm{V}}_{\mathrm{O}}^{••}+2{\mathrm{Ti}\prime }_{\mathrm{Ti}}+{\mathrm{H}}_2\mathrm{O}$$

(1)

where ${\mathrm{O}}_{\mathrm{O}}^{\mathrm{x}}$ is an O²⁻ ion in the oxygen lattice site, ${\mathrm{V}}_{\mathrm{O}}^{••}$ is an oxygen vacancy with a double positive charge, and Ti’_Ti is a Ti³⁺ ion in the titanium lattice site. The two Ti’_Ti that are created per ${\mathrm{V}}_{\mathrm{O}}^{••}$ have two excess electrons, which are responsible for the n-type conductivity, the blue-black colorization, and the enhancement of photocatalytic activity on TiO₂. The H₂ reduction on TiO₂ can also induce a disordered structure in the surface layer of TiO₂ nanocrystals, indicated by black TiO₂ [26,27,28]. Black TiO₂ exhibits high photocatalytic performance in decomposing organic pollutants and in generating hydrogen gas in an aqueous methanol solution under solar light irradiation. The other titanium oxides that have Ti_ds are the F-doped or Nb-doped TiO₂, in which oxygen atoms are substituted with fluorine atoms or Ti atoms are replaced with Nb atoms, respectively [29]. Another type of Ti defect in TiO₂ is titanium interstitials (${\mathrm{Ti}}_{\mathrm{i}}^{•••}$) possessing excess Ti atoms or ions in the lattice or in the near-surface region on TiO₂ surface [17,18,19,20,30].

Facile laser ablation and processing techniques have been developed to introduce the defects into TiO₂ nanocrystals and colloids in liquid [31,32,33,34]. In a typical procedure, TiO₂ suspensions are irradiated by a high-intensity pulsed laser with frequent repetition rates to produce the characteristic blue-black TiO₂. The obtained TiO₂ nanoparticles enhanced the photocatalytic activities in decomposing an organic dye [31] and in a water splitting reaction [32].

The electronic energy of Ti_ds is located just below the CB edge in the band gap of TiO₂ in a broad range of 0–1.8 eV [35]. Di Valentin and co-workers theoretically calculated the energy levels of point defects in bulk anatase TiO₂, which are located at 0.3, 0.4, 0.7, and 0.8 eV below the CB edge, for six-fold-coordinated Ti introduced by F- or Nb-doping, Ti–OH species associated with hydrogen doping, five-coordinated Ti’_Ti associated with the oxygen vacancy site, and titanium interstitials ${\mathrm{Ti}}_{\mathrm{i}}^{•••}$, respectively [29]. Deskins et al. calculated the relative energies of Ti_ds formed in the {110}-terminated rutile TiO₂ surface by means of the density functional theory (DFT), known as the DFT + U method [36,37]. They modeled the formation of Ti_ds at various Ti sites, such as the five-coordinated Ti and oxygen vacancies [37]. The calculation for the five-coordinated Ti in the presence of surface hydroxyls indicated that deep Ti_ds sites may exist in the second Ti layer from the surface or under the five-coordinated Ti rows [36].

The presence of these Ti_ds species, such as ${\mathrm{Ti}\prime }_{\mathrm{Ti}}$ and ${\mathrm{Ti}}_{\mathrm{i}}^{•••}$, can be experimentally confirmed by means of electron spin resonance (ESR) [38,39,40,41,42], infrared radiation (IR) [40,42,43,44,45,46,47,48], ultraviolet-visible absorption (UV–vis) [14,42,49,50,51], photoluminescence (PL) [52], photoacoustic [53,54,55,56], and photoelectron spectroscopies [17,18,30]. Scanning tunneling microscopy (STM) and atomic force microscopy (AFM) are powerful tools for the direct observation of surface Ti_ds [17,18,30,57,58,59]. The oxygen vacancy (Ti’_Ti accompanied by ${\mathrm{V}}_{\mathrm{O}}^{••}$) [57,58,59] and interstitial Ti (${\mathrm{Ti}}_{\mathrm{i}}^{•••}$) [30] sites were directly observed on the TiO₂ surface.

The Ti_ds described above are not the only type of point defects. There are many types of lattice defects including step edges, line defects, grain boundaries, and impurities [17]. The Ti_ds energies can be strongly affected by site heterogeneity due to the local structures. Furthermore, under actual photocatalytic conditions, the TiO₂ surface is always covered by adsorbates, especially solvent molecules; thus, the energy of Ti_ds should depend on the adsorbed species [19,20]. Therefore, the information from the theoretical calculations and the photoelectron spectroscopy, applied to the clean catalyst surfaces under ultra-high vacuum conditions, may be limited for actual photocatalytic systems. To address this issue, Ohtani et al. developed a powerful tool for measuring the energy-resolved distribution of electron trap states for many types of TiO₂ powders, composed of anatase, rutile, and brookite in methanol-containing gas phase, by means of reversed double-beam photoacoustic spectroscopy (RDB-PAS) [56]. They showed that the electron energy of the trap states is distributed around the CB edge of TiO₂ and the distribution range of anatase TiO₂ is relatively broader than that of rutile TiO₂ (Figure 2). The energy distribution patterns for both anatase and rutile TiO₂ powders are similar to those obtained by the photochemical method, which uses the surface reaction of the trapped electrons with methyl viologen to release its cation radical in de-aerated aqueous solution containing methanol as a sacrificial reagent [60]. They also revealed that the total density of the traps is well-correlated to the specific surface area of TiO₂ powders, suggesting that the electron trap sites are predominantly located on the surface of TiO_2, and they do not depend on the type of crystallites in anatase, rutile, or their mixtures.

The energy distribution of Ti_ds can also be obtained by an electrochemical method [61,62]. In this method, the potential variation in accumulated charge at the TiO₂ electrode can be measured in aqueous solution. By calculating the derivative of the accumulated charge (Q) versus the applied potential (U), the energy density of Ti_ds is directly proportional to dQ/dU and the plot of dQ/dU vs. U reflects the energy distribution of Ti_ds. The maximum density of Ti_ds is located around 0.25–0.4 eV below the CB edge for nanostructured anatase TiO₂ and P25 TiO₂ samples with a ratio of anatase to rutile of approximately 80:20. Thus, the distribution of Ti_ds within 0–0.4 eV is predominant, so that electrons trapped in these trap states may participate in the reductive reaction on TiO₂.

4. Reactivity of Trapped and Accumulated Electrons

This section highlights the reactivity of electrons trapped at Ti_ds and accumulated in CB in the reductions of molecular oxygen O₂ and molecular nitrogen N₂, and the hydrogenation and dehalogenation of selected organic substrates occurring on TiO₂. The reductive reactions associated with Ti_ds have been extensively investigated under various experimental conditions and through theoretical calculation methods. Although many studies have been performed for clean surfaces on TiO₂ under high-vacuum conditions [17,18,30,57,58,59], here we focus on the reactions occurring on powder or colloidal TiO₂ under conventional gas or liquid phase conditions.

Here we define the terms ‘accumulated electrons’ (e_cb⁻) and ‘trapped electrons’ (Ti_ds³⁺) to distinguish them. Accumulated electrons contain both electrons in CB and electrons trapped at shallow Ti states, located just below the CB edge of TiO_2, within 0–0.05 eV. These electrons can be in thermal equilibrium between the bulk CB and at the shallow trap sites, and easily migrate through these states. The accumulation of electrons in these states should occur after saturation of the intra-bandgap Ti states (Ti_ds) during UV irradiation. These electrons are highly reactive at the TiO₂ interface (Step 4 in Figure 1).The trapped electrons Ti_ds³⁺ mean the electrons trapped at Ti_ds are located in relatively deep energy from the CB edge. Therefore, the trapped electrons Ti_ds³⁺ cannot be excited thermally to the CB or the shallow states, exhibiting either low or no reactivity at the TiO₂ interface (Step 5 in Figure 1).

The trapped electrons Ti_ds³⁺ and the accumulated electrons e_cb⁻ are quite stable in the presence of a good sacrificial hole scavenger, such as alcohols or amines, and in the absence of electron acceptors such as O₂, the lifetime may exceed several hours [50,51,63]. This extremely long lifetime of Ti_ds³⁺ and e_cb⁻ is attributable to the excellent hole scavenging ability of the sacrificial reagents on the TiO₂ surface [24], which prevents the recombination of Ti_ds³⁺ and e_cb⁻ with the surface trapped holes. In other words, the hole scavengers inhibit Steps 7 and 8 in Figure 1. The excess charges caused by electrons Ti_ds³⁺ and e_cb⁻ on the irradiated TiO₂ are balanced by the insertion (intercalation) of protons into the TiO₂ lattice [47,62,64,65]. The electrons Ti_ds³⁺ and e_cb⁻ show the unique blue-black coloration from the visible to the IR region [14,40,42,43,44,45,46,47,48,49,50,51], which enables the tracing of the lifetimes of the species generated on the irradiated TiO₂. Interestingly, the electrons Ti_ds³⁺ and e_cb⁻ are distinguishable by measuring IR spectra. The free electrons e_cb⁻ exhibit the typical exponential frequency-dependent spectrum that is attributed to the intra-CB transition [43,44,46,48], whereas the trapped electrons Ti_ds³⁺ are characterized by a broad absorption in the mid-IR region that is ascribed to a direct optical transition [46,48].

4.1. Reduction of Molecular Oxygen and Hydrogen Peroxide

The interfacial electron transfer between molecular oxygen (O₂) and electrons Ti_ds³⁺ or e_cb⁻ on nanocrystalline TiO₂ films was examined using a transient UV–vis absorption spectroscopy in gas phase [66]. In the presence of ethanol, as a sacrificial hole scavenger under ethanol-saturated conditions (5.8%), the half-life (t_50%) of the electron-species Ti_ds³⁺ and e_cb⁻ was approximately 0.5 s in the absence of O₂. The t_50% value drastically decreased with increasing O₂ concentration, to approximately 12 μs in an oxygen concentration of 21% (air saturated conditions). Thus, the efficient electron transfer of molecular oxygen in a gaseous phase occurred on TiO₂ films by using ethanol as the hole scavenger. The dynamics of the electron transfer between O₂ and the nanosized TiO₂ particles in a liquid phase were also investigated by employing a simple and facile stopped flow technique [50,51]. With methanol as the hole scavenger, the electrons on the TiO₂ particles, in an argon-purged and de-aerated aqueous solution, were accumulated by pre-UV irradiation, causing blue colorization characterized by a broad absorption band in the visible light region of 400–800 nm. In the stopped flow experiment, the TiO₂ solution containing the electron-species Ti_ds³⁺ and e_cb⁻ was mixed with the aqueous solution containing O₂ at pH2.3, and the change in absorbance of the electrons was recorded at 600 nm. The absorbance signal decreased slowly with a rate constant of 8.9 × 10⁻⁷ mol L⁻¹ s⁻¹ in the absence of O₂, whereas the signal rapidly disappeared within a few seconds under O₂-saturated conditions. Thus, efficient electron transfer from the accumulated TiO₂ to O₂ proceeded even in the aqueous solution. The reduction of hydrogen peroxide (H₂O₂) was also confirmed by the same stopped flow experiment [50,51]. The decay rate of the electron-species Ti_ds³⁺ and e_cb⁻ in the H₂O₂ reduction was slower than in the O₂ reduction under similar conditions.

The reduction of O₂ on TiO₂ results in the formation of reactive oxygen species (ROS), such as superoxide anion radicals (O₂^˙−), hydroperoxy radicals (˙OOH), hydrogen peroxide (H₂O₂), and hydroxyl radicals (˙OH), under both aqueous and aerated conditions [6,21]. These ROS play a crucial role in the photocatalysis on TiO₂ for water purification, air cleaning, self-cleaning, self-sterilization, etc. The sequential ROS generation on photo-irradiated TiO₂ under acidic conditions can be depicted as follows [6,21,50,51]:

Photoreduction:	O2 + (Tids3+ and ecb−) → O2˙−
Protonation:	O2˙− + H+ →˙OOH
Disproportionation:	2˙OOH → H2O2 + O2
Photoreduction:	H2O2 + (Tids3+ and ecb−) →˙OH + OH−

Though the CB edge of anatase TiO₂ is located at −0.27 V vs. standard hydrogen electrode (SHE) at pH2.3 [67], anatase TiO₂ produced the superoxide anion radial by photoexcitation; O₂/O₂^˙− was approximately 0.33 V vs. SHE [68]. This could be due to the strong adsorption of O₂ at the Ti_ds sites, such as ${\mathrm{V}}_{\mathrm{O}}^{••}$, which may lead to a positive shift of the redox potential (O₂/O₂^˙−). In this electron transfer step, an electron seems to be transferred from the Ti 3d orbital to the π* orbital of the adsorbed O₂.

4.2. Reduction of Molecular Nitrogen, Nitrate, and Nitrite Ions

Molecular nitrogen (N₂) is chemically stable, so photocatalytic reduction of N₂ to ammonia (NH₃) under ambient temperature and pressure is challenging. Hirakawa and coworkers recently reported that the photocatalytic conversion of N₂ to ammonia with water occurred on the bare TiO₂ powders under ambient conditions [69]. They stated that the active sites for N₂ reduction are the Ti_ds with oxygen vacancies mainly formed on the rutile {110} surface. They investigated the photocatalytic reductions of nitrate (NO₃⁻) and nitrite (NO₂⁻) ions to ammonia and N₂ on bare TiO₂ under ambient conditions [70]. They proposed that the Ti_ds sites selectively promoted the eight-electron reduction of NO₃⁻ to NH₃ (NO₃⁻ + 9H⁺ + 8e⁻ → NH₃ + 3H₂O), while the Lewis acid site promoted nonselective reduction, resulting in N₂ and NH₃ formation. Thus, TiO₂ with many Ti_ds and a small number of Lewis acid sites produced ammonia with very high selectivity. The use of artificial fertilizers in agriculture has caused a great deal of concern for water pollution caused by the production of NO₃⁻ and NO₂⁻ ions from fertilizers [71]. Therefore, a chemical process for the reduction of NO₃⁻ and NO₂⁻ ions on TiO₂ may be useful for an environmental recycling process.

4.3. Hydrogenation of Carbonyl Compounds

Kohtani et al. examined whether electrons Ti_ds³⁺ and e_cb⁻ transfer to acetophenone (AP) derivatives adsorbed on TiO₂ [63,72]. The photoreductive hydrogenation of several aromatic carbonyl compounds was confirmed to occur on UV-irradiated P25 TiO₂ as illustrated in Scheme 1 [72,73]. They evaluated the number of transferred electrons in the injection experiment using a pre-irradiated TiO₂ suspension. When the P25 TiO₂ powder was dispersed in de-aerated ethanol as a hole scavenger and irradiated with UV light for 2 h, the white color of TiO₂ powder changed to blue-gray. After the blue-gray color change was confirmed, a large amount of AP derivatives was injected into this TiO₂ suspension in the dark. In this experiment, ethanol acted not only as a solvent but also as a hole scavenger.

Figure 3a shows that the blue-gray color of the pre-irradiated P25 TiO₂ suspension remained for a few days in the absence of AP derivatives [63], meaning that the electrons accumulated on the P25 TiO₂ surface are quite stable in the de-aerated ethanol. Figure 3b,c show the color change induced by the addition of aromatic carbonyl compounds. The blue-gray color of TiO₂ rapidly changed after the injection of 2,2,2-trifluoacetophenone (TFAP) where the aromatic ring (Ar) was C₆H₅ and the R was CF₃ (Scheme 1). The change from blue-gray to white was completed within 3 h in the TFAP solution as shown in Figure 3c. This result indicates that all the trapped and accumulated electrons on TiO₂ were consumed in the reduction of TFAP within 3 h. On the other hand, with AP, a part of the blue-gray species on TiO₂ was remarkably stable even after 50 h, as shown in Figure 3b, which may be due to the remaining electrons trapped at the deep states of TiO₂.

Figure 4 depicts the time evolution of the secondary alcohols 1-phenylethanol (AP-OH) or 1-phenyl-2,2,2-trifluoroethanol (TFAP-OH), as products after the injection of substrates AP or TFAP, into the sufficient pre-irradiated TiO₂ suspension, respectively [63]. The amount of each product quickly grew within 0.5 h, which agrees with the observation of color change in the TiO₂ suspension (Figure 3b,c). The amount of TFAP-OH product obtained from the reactive substrate TFAP rapidly increased, and reached 5.4 μmoL within 1 h. Assuming all Ti_ds³⁺ and e_cb⁻ electrons were consumed in the reductive hydrogenation of TFAP, the total amount of Ti_ds³⁺ and e_cb⁻ on the TiO₂ powder was estimated to be about 100 μmol∙g⁻¹. The time evolution of AP-OH from the less reactive substrate AP consisted of a fast component within 0.5 h, and a slow component after 0.5 h, which increased gradually to reach 4.1 μmol (Figure 4). The slow component represents the slow electron transfer event from middle Ti_ds, between shallow and deep states, to AP adsorbed on TiO₂. The total amount of AP-OH production was about 25% smaller than that of TFAP-OH production. In the reduction of the less reactive substrate AP, the deep Ti_ds³⁺ species remained on the TiO₂ surface for a long time (>20 h) after the injection of AP. The amount of deep Ti_ds³⁺ that remained on TiO₂ was roughly estimated to be 25% given the difference between the amounts of TFAP-OH (5.4 μmoL) and AP-OH (4.1 μmoL). Thus, the residual 25% electrons could not react with AP and remained at the deep trap sites.

Assuming that all Ti_ds³⁺ and e_cb⁻ electrons react with TFAP, the percentages of reacted electrons were estimated for other AP derivatives, as summarized in

Table 1. Reduction potentials, amount of reacted electrons, percentages of reacted electrons, and reaction rates at maximum concentration of substrates [72]. Adapted with permission from reference [72]. Copyright (2014) The Royal Society of Chemistry.

Substrate	Ered/V a	Amount of Reacted Electrons b/μmoL	Percentage c/%	Reaction Rate d/mMh−1
1 (TFAP)	–1.35	10.2	100	–
2	–1.59	8.22	81	3.4 ± 0.2
3	–1.62	6.32	62	2.2 ± 0.2
4	–1.80	6.09	60	2.0 ± 0.1
5 (AP)	–1.89	7.38	72	1.9 ± 0.1
6	–0.92	5.70	56	1.2 ± 0.1
7	–1.94	4.76	47	0.75 ± 0.05

^a Reduction potentials vs. SHE (standard hydrogen electrode). ^b Molar number of reacted electrons estimated by the injection experiment using a pre-irradiated TiO₂ suspension. ^c Percentage of the reacted electrons per the total amount of Ti_st³⁺ and e_cb⁻ (10.2 μmol) generated on 0.10 g TiO₂. ^d Reaction rates at the maximum concentration of substrates.

[72]. The number of reacted electrons showed a tendency to decrease with decreasing the reduction potential (E_red) of the substrates according to the dependence on the actual reaction rates as listed in Table 1. This implies that the rates of photocatalytic hydrogenation of AP derivatives are governed by the electron transfer efficiency from the Ti_ds sites to the adsorbed AP sites. Notably, the relative position between the Ti_ds energy distributed within the bandgap and the acceptor level of the AP derivatives (the solid Gaussian curve) should be appropriate (Figure 5) [72]. The energy distribution of Ti_ds in TiO₂ powders and colloids can be obtained by photochemical [60], electrochemical [61,62], and RDB-PAS [56] methods. The acceptor levels of AP derivatives can be estimated by the Marcus theory [72,74,75,76].

Kohtani et al. also reported the photohydrogenation of AP derivatives on P25 TiO_2, modified with metal-free organic dyes such as rhodamine B, fluorescein, and coumarin derivatives [77,78]. The use of these organic dyes successfully extended the UV response of TiO₂ to the visible light region, though these reaction rates were much slower than the hydrogenation rate using UV excitation of non-modified TiO₂. In this dye-sensitized system, the electron injection from dye into TiO₂ can take place in two different ways: (1) injection via a lowest unoccupied molecular orbital (LUMO) level of the excited dye to the CB of TiO₂, and (2) direct injection of TiO₂ to CB on the excitation of the charge transfer complex (TiO₂^δ− dye^δ+) [77]. The injected electrons should then be distributed to the Ti_ds sites on the P25 TiO₂ surface. The accumulated electrons were observed with the blue-gray color for all dye-TiO₂ powders during visible light irradiation.

4.4. Defluorination of Fluorinated AP Derivatives

Compounds containing fluorine atoms are often used as pharmaceutical and agrochemical reagents. Since the C–F bond is one of the strongest bonds, C–F bond activation and cleavage is a field of current interest in organic chemistry [79], although less is known about the catalytic activation and cleavage methods. Photocatalytic reaction is one of the promising methods to promote the activation and cleavage of the C–F bond of fluorinated compounds under mild conditions. Therefore, an attempt was made to use trapped and accumulated electrons on TiO₂ for the sequential multi-step electron transfer in the reduction of fluorinated AP derivatives (Figure 6) [80]. The reaction of fluorinated AP derivatives on TiO₂ showed the two photocatalytic reductive transformations, i.e., the defluorination and reduction of the carbonyl group. The reduction of 2-fluoromethylacetophenone (MFAP) only provided the ketone AP because of the C–F bond cleavage, whereas the reaction of TFAP only provided the alcohol TFAP-OH as a result of the reduction of the carbonyl group. Interestingly, the reduction of 2,2-difluoromethyacetophenone (DFAP), possessing characteristics between those of MFAP and TFAP, gave the defluorinated ketones, MFAP and AP, as well as hydrogenated alcohol 1-phenyl-2,2-difluoroethanol (DFAP-OH), as shown in Figure 6. The defluorination reactions became unfavorable with increasing the number of fluorine atoms on the substrates. This mainly arises because of the increase in the bond dissociation energy of the C–F bond and the positive shift of the reduction potential of fluorinated AP derivatives with the increasing number of fluorine atoms [80].

4.5. Hydrogenation of Nitroaromatic Compounds

Several organic nitroaromatic compounds can be easily hydrogenated to create the corresponding amino compounds on the UV-irradiated TiO₂ in the presence of 2-propanol as a sacrificial hole scavenger (Scheme 2).

Shiraishi et al. reported that some kinds of rutile TiO₂ particles promote the highly efficient photocatalytic hydrogenation of nitroaromatic compounds [81,82]. They claimed that the oxygen vacancy sites on the rutile {110} behave as the adsorption sites for the nitroaromatic compounds and the electron trap sites, resulting in the formation of aniline derivatives with significantly high yields of greater than 25% at 370 nm [81]. They also found that the activity of rutile particles depends on the number of defects on the particles [82]. The inner (bulk) defects behave as the deactivation sites for the recombination of electrons and holes, whereas the surface defects behave as the active reaction sites as well as the deactivation sites. As a result, the reaction rate is proportional to the ratio of the amount of surface defects to that of total defects (N_surface/N_total) in the rutile TiO₂ particles, which aligns with the report of Kong and coworkers [16].

Molinari and coworkers examined the selective hydrogenation of NO₂–C₆H₄–CHO, bearing the two reducible functional groups, –NO₂ and –CHO [62]. They found that the nitro group was easily reduced by the trapped electrons at Ti_ds within the bandgap, whereas the aldehyde group was reduced by electrons accumulated on CB. Therefore, the chemoselective reduction of functional groups can be controlled by the energy distribution patterns of Ti_ds, which may depend on the type of TiO₂ powders. The selective reduction of functional groups is difficult to achieve through conventional thermal catalysis. Therefore, this topic is an interesting issue for chemoselective photocatalysis.

5. Summary and Potential for the Development of Efficient Photocatalysis

According to the reports of Kong et al. [16] and Shiraishi et al. [82], the photocatalytic efficiencies increased with increasing the ratio of the amount of surface to bulk defects (N_surface/N_bulk) or the amount of surface to total defects (N_surface/N_total). For example, molecular oxygen (O₂) in a gaseous phase would be easily adsorbed on the surface Ti_ds and reduced by electrons Ti_ds³⁺ and e_cb⁻, resulting in the efficient formation of ROS, which oxidize benzene efficiently [16]. Further, the surface Ti_ds behaves as the active reaction site in the efficient reduction of nitrobenzene [81]. Thus, the surface defects favorably act as adsorption sites as well as reaction sites.

In addition, the relative position between the energy distributions of Ti_ds and the acceptor level (reduction potential) of substrates should be appropriate as indicated in Figure 5 [72]. The electrons accumulated in CB and trapped at shallow Ti_ds can easily participate in the reaction, whereas those trapped at deep Ti_ds cannot [63]. These unreacted electrons remain at the deep Ti_ds sites and exhibit an extremely long lifetime in alcohols. Thus, the shallow traps enhance photocatalytic activity, while the deep traps cause a reduction. Furthermore, Amano et al. proposed that the creation of shallow Ti_ds greatly enhances electrical conductivity, thereby facilitating the charge transport and separation caused by the formation of band bending in the space charge layer at the TiO₂-liquid interface [14,15].

In conclusion, the development of highly active photocatalysts necessitates precise control of the structural properties; the density of surface shallow traps should be maximized and the density of deep traps as well as inner (bulk) traps minimized, as proposed by Ohtani [83]. One of the promising strategies for meeting these requirements is the use of highly uniform TiO₂ nanocrystals with specific exposure of the reactive facets [42,84,85,86,87]. In particular, the anatase {101} and {001} facets have been reported to be favorable for the reductive and oxidative reactions in TiO₂ photocatalysis, respectively [87,88,89,90,91]. If the reductive and oxidative facets could be selectively covered with a large amount of the active shallow Ti_ds and the terminal Ti–O_sH hole trapping sites, respectively, the photocatalytic activity on the TiO₂ nanocrystals would be greatly enhanced by the effective electron-hole charge separation, followed by the subsequent charge transfer reactions at the specific reductive and oxidative sites. Therefore, special attention should be directed toward the development of TiO₂ nanocrystals with precisely controlled facets.