TiO₂, a white solid substance, has been widely used as a commercial pigment additive since the early 20th century ^[1]. Especially, since Fujishima et al. ^[2] reported for the first time the photoelectrochemical water splitting on a n-type TiO₂ electrode under ultraviolet (UV) light illumination in 1972, TiO₂ nanomaterials have been extensively applied in various areas including photovoltaics (dye-sensitized solar cells) ^{[3, 4, 5, 6, 7, 8]}, sensors ^[9], photo(electro)catalytic hydrogen production from water splitting ^{[1, 10, 11, 12, 13, 14, 15, 16, 17]}, degradation of pollutants ^{[18, 19, 20, 21, 22, 23, 24, 25]}, selective organic transformations ^{[26, 27]}, and CO₂ reduction for fuel generation ^{[11, 28, 29, 30, 31, 32, 33, 34]}. The global environment and energy issues, as two main challenges in the 21st century, are becoming increasingly severe with each passing day ^[35]. Importantly, TiO₂ is versatile, economical, stable, abundant, non-toxic, and environmentally friendly. And it has been regarded as one of the most ideal photocatalysts for energy and environment applications ^{[13, 16, 23, 36, 37]}. Thus, among these research fields, heterogeneous photocatalysis using solar energy and TiO₂ has drawn much attention in the past decade because it is a promising means to address both global energy crises and environmental pollution ^{[13, 38, 39]}.

To date, there has been a growing number of publications on the use of heterogeneous photocatalytic redox reactions over TiO₂ in energy and environmental applications ^[40], as contributed by scholars from various academic and industrial research communities. Clearly, the photocatalytic performance of TiO₂ is largely dependent on its optical and electronic structures, morphological properties, crystallinity, (nano) size, and surface chemistry ^[41]. However, its low quantum efficiency and wide band gap greatly limit its applications in the visible light region. Therefore, extensive research efforts have been devoted to improve its photocatalytic efficiency and absorption ability of visible light via diverse modification technologies ^{[12, 17, 25, 27]}.

The recent breakthroughs made in developing TiO₂-based photocatalysts have been presented in many excellent reviews ^{[11, 41, 42, 43, 44, 45]}. However, most of these reviews focus on the achievements made in the synthesis and modification of nanoscale titania and its diverse applications in specific areas of photocatalysis ^{[1, 11, 43, 45, 46]}. In contrast, some issues on the deeper understanding of the relationship between surface chemistry and photocatalysis enhancement of TiO₂ photocatalysts at the molecular level are often neglected despite the increasing interest in this exciting field of research and numerous review articles on TiO₂ photocatalysis. As much progress has been made in the development of TiO₂-based photocatalysts through suitable surface modification strategies, it is believed that a comprehensive review on the deeper understanding of the molecular mechanisms for various surface modification strategies of TiO₂ is timely to better design photocatalytic systems and promote state-of-the-art developments in this area. Such developments are also expected to offer additional unexplored opportunities regarding the design and improvement of other non-TiO₂- based photocatalysts. In this review, we highlight the fundamentals of photocatalysis, and surface chemistry and corresponding surface modification strategies of TiO₂ nanomaterials for various applications. Particularly, we focus on the relationship between surface chemistry and emerging surface modification strategies for the activity improvement of TiO₂-based photocatalysts. It is expected that this review will offer some new opportunities for designing and developing both TiO₂-based and non-TiO₂-based photocatalysts.

In general, there are many typical photocatalytic processes such as water splitting, degradation of pollutants, selective organic transformations, and CO₂ reduction for fuel generation. The unique photocatalytic property of TiO₂ originates from the thermodynamic properties of its band structure including the band gap and positions of the valence band (VB) and conduction band (CB). The band structural properties determine the light absorption capability and redox ability of TiO₂ nano-photocatalysts ^[11]. The thermodynamic fundamentals for photocatalytic water splitting, degradation of pollutants, and CO₂ reduction over TiO₂ nano-photocatalysts are presented in Fig. 1. As observed in Fig. 1, these processes share some typical features. First, the TiO₂ nano-photocatalyst absorbs light of wavelengths shorter than ~387.5 nm (λ ≤ hc/E_g ≈ 1240/3.2 nm = 387.5 nm; h = Planck’s constant, c = speed of light, E_g = band gap) upon exposure to incident light with sufficient energy. Then, electrons in the CB are photo-generated and holes in the VB are created, which promptly reach the surface of the TiO₂ nano-photocatalyst to initiate the corresponding reduction and oxidation reactions, respectively ^[47]. Therefore, the oxidation and reduction capabilities of trapped holes and electrons on the surface active sites of TiO₂ nano-photocatalysts are directly dependent on the positions of the VB and CB, respectively. The holes on the surface of TiO₂ can directly oxidize adsorbed water/OH⁻ to O₂ (Fig. 1(a) and (c)) or to reactive oxidants such as hydroxyl radicals toward the degradation of contaminants (Fig. 1(b)). Furthermore, it should be noted that the photo-generated electrons in anatase TiO₂ can initiate most reduction processes (including O₂ reduction to superoxides, H₂ generation, and CO₂ reduction to methane, methanol, or formaldehyde) except for CO₂ reduction to HCO₂H and CO. Accordingly, it is not surprising that anatase TiO₂ has been widely used in various photocatalytic fields.

Fig. 1. Similarities in diverse photocatalysis-related processes. (a) Water splitting; (b) Degradation of pollutants; (c) CO2 reduction over anatase TiO2. (HER, hydrogen evolution reaction; OER, oxygen evolution reaction; ORR, oxygen reduction reaction; WOR, water oxidation reaction; CRR, CO2 reduction reaction).

The free energy, ΔG, and standard redox potential, ΔE⁰, of the multi-electron water splitting (Eq. 1) and CO₂ reduction (Eqs. 2-6) processes are given below ^[30]. From the viewpoint of the Gibbs free energy change, the ΔG values of these reactions are highly positive, indicating that they cannot proceed spontaneously at ambient temperature because they are all thermodynamically unfavorable uphill processes with a multistep procedure. Especially, the more negative ΔG values of CO₂ reduction reactions indicate that such reactions are more difficult to proceed when compared with water splitting reactions. The ΔG values additionally indicate that the CO₂ reduction reactions can store more energy than the water splitting reactions ^{[48, 49]}. As observed from Fig. 1, for the water splitting and CO₂ reduction reactions, raising the position of the CB can enhance the driving forces for these reduction reactions to proceed.

H₂O(l) → H₂(g) + 1/2O₂(g) ΔG = −237 kJ/mol ΔE⁰ = 1.23 V (1)

CO₂(g) → CO(g) + 1/2O₂(g) ΔG = −257 kJ/mol ΔE⁰ = 1.33 V (2)

CO₂(g) + H₂O(l) → HCOOH(l) + 1/2O₂(g) ΔG = −286 kJ/mol ΔE⁰ = 1.48 V (3) (3)

CO₂(g) + H₂O(l) → HCHO (l) + O₂(g) ΔG = −522 kJ/mol ΔE⁰ = 1.35 V (4)

CO₂(g) + H₂O(l) → CH₃OH(l) + 3/2O₂(g) ΔG = −703 kJ/mol ΔE⁰ = 1.21 V (5)

CO₂(g) + H₂O(l) → CH₄(g) + 2O₂(g) ΔG = −818 kJ/mol ΔE⁰ = 1.06 V (6)

In contrast, photocatalytic environmental treatment processes, such as photo-oxidation of organic pollutants by O₂ molecules, are generally thermodynamically favorable as they are regarded as typical downhill-type processes ^[16]. Such processes have been extensively investigated using TiO₂ photocatalysts. The formation of highly oxidative species (e.g., hydroxyl radicals and O₂⁻) on the surface of TiO₂ semiconductors during catalytic processes is needed to overcome the activation barrier. However, these photo-generated OH radicals and superoxide anions are crucial for the oxidation of organic compounds in water or air. The CB energy level (E_CB = −0.51 V at pH 7) is slightly more negative than the one-electron reduction potential of oxygen (E⁰(O₂/O₂^−•) = −0.33 V), thus achieving facile reduction of O₂, which has been considered as the rate-determining step in photocatalysis for environmental purification (Fig. 1(b)) ^[25]. In contrast, the VB energy level of TiO₂ (E_VB = 2.69 V at pH 7) is more positive than the oxidation potential of surface-adsorbed H₂O or OH⁻ to produce ^•OH radicals (E⁰(H₂O/^•OH) = 2.29 V), thereby confirming that the photo-generated holes in VB of TiO₂ can directly oxidize organic substances or water to their radicals ^[24]. It has also been demonstrated that free ^•OH is more reactive than the surface-bound analogue owing to the reorganization of energy ^[25]. All these favorable thermodynamic factors make TiO₂ the most suitable semiconductor photocatalyst for various photocatalytic applications.

Besides the favorable thermodynamic aspects, the charge carrier dynamics properties in photocatalysis employing TiO₂ nanoparticles are important factors in determining the photocatalytic activity. It is commonly agreed that that the four-step tandem process (Fig. 2) must take place for achieving the final interfacial redox reactions of electrons and holes over TiO₂: light harvesting (step (1)), charge generation (step (2)), charge separation and transport (steps (3), (4), and (7)), and surface catalytic reactions (steps (5) and (6)). Therefore, only the photo-generated charge carriers that reach the surface of TiO₂ can be used for photocatalytic redox reactions. The overall efficiency of the various photocatalytic processes can be described as follows:

Fig. 2. Processes involved in photocatalytic redox reactions over TiO2 photocatalyst. (1) Light harvesting; (2) Charge generation; (3) Charge transport in bulk; (4) Charge recombination in bulk; (5) Surface reduction reactions; (6) Surface oxidation reactions; (7) Charge recombination on the surface (RC, reduction co-catalyst; OC, oxidation co-catalyst; BR, bulk recombination; SR, surface recombination).

η_c = η_abs × η_cs × η_cmt × η_cu (7)

where η_c is the solar energy conversion efficiency, η_abs is the light absorption efficiency, η_cs is the charge excitation/separation efficiency, η_cmt is the charge migration and transport efficiency, and η_cu is the charge usage efficiency for photocatalysis reactions. Among the different parameters efficiencies, high charge separation efficiencies (η_cs and η_cmt) and surface catalytic reaction efficiency (relating to η_cu) are the key factors determining the overall efficiency of the photocatalytic processes.

The fast charge recombinations of steps (4) and (7), as deactivation processes, take place in the time scale of femtoseconds to microseconds, and are unfavorable to improving the photocatalytic activity. Thus, investigations of charge carrier dynamics inside and on the surface of TiO₂ photocatalysts are essential for thoroughly understanding photocatalytic reaction mechanisms. Generally, the charge carrier dynamics in semiconductor nanoparticles can be directly measured by means of ultrafast time-resolved absorption spectroscopy ^[50]. The transient absorption spectra can efficiently identify the reactive species, trapped holes, trapped electrons, and quasi-free electrons, whereas the decay profiles can be used to determine the rate of recombination of electrons and holes ^[51]. In fact, time-resolved spectroscopic studies have revealed that the relatively low photocatalytic efficiency over TiO₂ is mainly due to the rapid recombination rate of photo-generated electron (e⁻)-hole (h⁺) pairs (∼90%) after excitation ^[21]. The time-resolved microwave conductivity technique can be employed to study charge carrier dynamics in TiO₂ nanoparticles; this technique is very sensitive to the particle size and morphology of the photocatalyst ^{[52, 53]}. For example, the mean lifetime of a single electron-hole pair in a TiO₂ particle (12 nm in diameter) was determined as 30 ns ^[54]. Furthermore, hole trapping processes are much slower (τ ~250 ns) than electron trapping processes (τ ~30 ps) ^[54]. As reported by Bahnemann et al. ^[55], flash radiolysis can provide strong evidence of the existence of trapped charge carriers. Studies on the complex relaxation dynamics of electrons revealed that photo-generated electrons (after reaching the surface within 170 fs) can migrate between the surface and shallow bulk trap sites that are in equilibrium (Fig. 3) ^[56]. These shallow-trapped electrons are capable of relaxing into deeper sites with a time constant of 500 ps through a hopping process involving energetically distributed trapping sites. More interestingly, TiO₂ colloids doped with Fe³⁺, V⁴⁺, Mo⁵⁺, or Ru³⁺ were found to exhibit much slower charge carrier recombination rates, of which the decay times could be extended to 50 ms ^[57]. The reaction dynamics of photo-generated charge carriers in TiO₂ are summarized in Fig. 4. As observed from Fig. 4, the trapping, recombination, and interfacial transfer of charge carriers are very fast processes. Therefore, employing strategies that can inhibit bulk and surface charge recombination and improve charge transport to the surface sites of TiO₂ can boost the charge carrier dynamics and the photocatalytic activities of TiO₂ photocatalysts. These strategies include decreasing the defect sites in the bulk and on the surface, increasing the accessible surface areas, coupling with nanocarbon materials, and constructing heterojunctions and nano-structured photocatalysts, all of which will be discussed in detail in Section 3.

Fig. 3. Schematic illustration of the spatial and energetic distribution of electron traps in a TiO2 film. Reprinted with permission from Ref. [56].

As the optical penetration depth, δ_p, in TiO₂ (δ_p ≈ 250 nm at λ = 308 nm) is very short, the photo-generated electron−hole pairs are generally created in the outer surface region of TiO₂ ^[58]. Thus, the photocatalytic reactions over TiO₂ are strongly dependent on its surface and interfacial properties. However, it is also seen from Fig. 4 that the interfacial charge transfer rates (steps (5) and (6), Fig. 2) are much slower than those of charge trapping and recombination in the bulk and on the surface of TiO₂ nano-photocatalyst. Hence, interfacial charge recombination will proceed rapidly if the interfacial charge transfer rates are not greatly enhanced. Therefore, improving the surface reaction kinetics is essential for suppressing the fast surface recombination of charge carriers and boosting the overall photocatalytic efficiency. Accordingly, to achieve this aim, many different strategies, such as loading co-catalysts, increasing the accessible surface active sites, and exposing highly reactive facets, have been widely studied during the past decades, and will be thoroughly discussed.

Fig. 4. Photo-induced reactions in TiO2 photocatalysis and the corresponding time scales. Reprinted with permission from Ref. [42].

The surface properties of TiO₂ nano-photocatalysts have significant effects on their photochemical applications. The complex surface physicochemical properties of TiO₂ nano- photocatalysts are shown in Fig. 5. Typically, the surface charge of TiO₂ plays an important role in determining the photocatalytic efficiency, especially for the kinetics of photo-oxidation of organic dyes. The surface charge properties of TiO₂ in water at different pH levels are typically influenced by its zero-point charge (zpc) as the acid-base equilibrium reaction can be observed for the surface titanol groups of TiO₂ as follows ^[25]:

Fig. 5. Surface physicochemical reactions on TiO2 photocatalyst.

>Ti-OH₂⁺ ↔ >Ti-OH + H⁺ (pK_a1) (8)

>Ti-OH ↔ >Ti-O⁻ + H⁺ (pK_a2) (9)

The pH of the zero point of charge is calculated as:

pH_zpc = 1/2(pK_a1 + pK_a2) (10)

The point of zero charge (pH_zpc), corresponding to the pH at the point of zero net proton charge (isoelectric point or zero zeta potential), is typically obtained by measuring the electrokinetic potentials (zeta potential, ZP) of suspended particles at different pH levels.For example, the pH_zpc of Degussa P25 TiO₂ is ~6.2 (pK_a1 = 4.5 and pK_a2 = 8) ^{[59, 60]}. Hence, at pH levels < 6.2, the particle surface is positively charged, whereas at pH levels > 6.2, it is negatively charged. In general, more efficient generation of hydroxyl radicals by TiO₂ is achieved in weak alkaline medium owing to the presence of an optimal concentration of OH⁻, which is beneficial for the photocatalytic degradation of organic compounds in aqueous solution. Furthermore, it was found that the relative dielectric constant, ε, of the solvent in which TiO₂ was suspended has a strong effect on the ZP of TiO₂ owing to the electrostatic interactions between the TiO₂ powder and solvent ^[61]. The ZP of TiO₂ powder suspended in different alcohol solvents can be calculated by an empirical formula (Eq.(11)). Studies showed that TiO₂ powder suspended in methanol exhibited the highest ZP when compared with that suspended in other alcohols (e.g., ethanol, 1-propanol, isopropanol, and 1-butanol) owing to the high ε of methanol ^[61]. These results indicate that TiO₂ is the most stable in methanol. Consequently, the effect of process parameters, such as pH and solvent, on the photocatalytic performance should be carefully investigated for practical applications.

ZP = 13.43 − 213.63/ε (11)

Suitable control of surface hydroxyl groups can lead to the formation of other interesting photo-induced phenomena such as superhydrophilicity. These phenomena are intrinsically different from the well-known photocatalytic processes. In 1997, the UV light-induced superhydrophilicity behavior of TiO₂ film surfaces in both water and oil was first discovered by Fujishima and coworkers ^[62]. This unique photo-induced wetting behavior of TiO₂ under UV illumination can be elucidated by the following mechanism. The Ti³⁺ state and oxygen vacancies can be produced by the photo-generated electrons and holes, respectively (the right section of Fig. 5). Then, these oxygen vacancies can be occupied by water molecules, thus leading to the formation of adsorbed OH groups on the surface defect sites. Consequently, the increased density in surface hydroxyl groups gives rise to the surface hydrophilic properties ^{[23, 63, 64, 65]}. Furthermore, as reported, ultrasonic treatment of amphiphilic TiO₂ surfaces in pure water can reduce the degree of surface hydrophilicity owing to partial re-oxidation of the photo-reduced surface upon generation of OH radicals, further demonstrating the hydrophilic−hydrophobic conversion mechanism ^[66]. It was also found that the contact angle of water on porous TiO₂ coating films decreased with increasing adsorbed hydroxyl content, capillary force, and roughness of the resultant coating films ^[67]. The addition of 10-20 mol% SiO₂ in the TiO₂ films generated the lowest water contact angle ^[68]. Further investigations confirmed that suitable mesopores, and surface microstructures and roughness could enhance the photo-induced superhydrophilic properties of porous TiO₂ thin films ^{[69, 70]}. Moreover, TiO₂ nanofilms coated on suitable substrates exhibit superior photo-induced hydrophilicity, probably due to their higher surface roughness and hydroxyl content ^{[71, 72, 73]}. It is expected that TiO₂-based thin films, exhibiting superhydrophilicity under visible light irradiation, will be further developed in the future ^[74].

The surface Ti³⁺ states and oxygen vacancies on TiO₂ photocatalysts can be created by many other strategies such as by thermal treatment under vacuum conditions or a reducing or oxygen-depleted atmosphere, bombardment using high energy particles, or doping with metal or non-metal ions ^[75]. The as-obtained defective or self-doped TiO₂ generally exhibits excellent activities toward various photocatalytic applications ^{[76, 77, 78, 79, 80, 81, 82]}, which will be not discussed in this review. For more information, recently published reviews are available ^{[41, 75, 83]}.

The density of surface hydroxyl groups can be altered by employing high-temperature calcination treatments ^{[84, 85, 86, 87]} and by exchange of surface -OH groups ^{[88, 89]}. For example, using a simple sulfuric acid treatment can significantly enhance the photocatalytic activity of TiO₂ films toward the oxidation of acetone in air owing to the reduction of sodium ions and increase in the surface-adsorbed hydroxyl content ^[90]. Further research studies showed that SO₄²⁻ species could replace the surface -OH groups at high temperatures and introduce strong acidic sites onto the surface of mesoporous TiO₂ molecular sieve films (Fig. 6(a)) ^[89]. The highly polarized state of surface strong acid sites would allow more effective hole trapping, thus suppressing the photo-carrier recombination, which can be further evidenced by time-resolved photoluminescence studies (Fig. 6(b)). The resulting mesoporous sulfated TiO₂ films, as solid superacid photocatalysts, exhibited a 2-fold enhancement in the photocatalytic oxidation of CH₃Br in air when compared with mesoporous TiO₂ films ^[89]. Moreover, the crystallization of TiO₂ or the number of OH groups per TiO₂ particle can be significantly increased by increasing the calcination temperature, thus leading to enhanced photocatalytic performances ^{[87, 91, 92]}. However, if the calcination temperature is excessively high, the photocatalytic activity typically decreases because of the formation of rutile and sintering and growth of TiO₂ crystallites that result in a reduction in the surface area ^{[84, 85, 86, 93]}. Therefore, the calcination temperature should be optimized to obtain optimal photocatalytic performances. Furthermore, using suitable inhibitors for suppressing grain growth and rutile phase formation of TiO₂ (such as carbon layer, ZrO₂, phosphoric acid, and Na₂SO₄) is favorable for obtaining anatase TiO₂ with high crystallinity and surface area, thus achieving higher photoactivities ^{[94, 95, 96, 97]}. The use of surface fluorination modification for enhancing the photocatalytic activity will be thoroughly discussed in Section 3.5.

Fig. 6. (a) Schematic representation of mesoporous nanocrystalline SO42−/TiO2 superacid molecular sieve film formation. (b) Time-resolved photoluminescence profiles of mesoporous TiO2 and mesoporous sulfated TiO2 films. Reprinted with permission from Ref. [89].

Based on the above analysis, it is clear that pristine TiO₂ typically has unfavorable charge carrier dynamics and a wide band gap. Therefore, surface modification is often necessary to enhance charge carrier separation and visible light absorption to improve the photocatalytic activity. Based on the photocatalytic mechanism, some typical surface modification strategies can be proposed to design high activity TiO₂ nano- photocatalysts, as shown in Fig. 7. First, the wide band gap of pristine TiO₂ (3.0-3.2 eV) greatly limits its visible light absorption capabilities. Thus, narrowing the band gap of TiO₂ by doping or loading suitable visible light photosensitizers onto TiO₂ has been widely employed to enable usage of a wider range of the solar spectrum. Second, constructing surface heterojunctions has been used to improve the charge carrier dynamics and prolong the charge lifetime, thus resulting in enhanced photocatalytic performances. Third, loading nano-sized co-catalysts can also greatly accelerate the surface reaction kinetics, thus suppressing rapid surface charge recombination processes. Fourth, increasing accessible surface areas can efficiently promote light harvesting, and the adsorption and diffusion of reactants, thereby achieving improved photocatalytic activities. Fifth, introducing surface F effects on TiO₂ can increase the concentration of free OH radicals with larger oxidation potential and alter the photodegradation selectivity, thereby facilitating enhancements in the photodegradation performance. Finally, exposing highly reactive facets is beneficial for increasing the surface active sites and redox ability for target reactions and for improving photocatalytic reactivity. All these strategies are efficient for improving several unfavorable factors during photocatalysis, and have been widely studied by different groups. Herein, we select some typical examples to discuss these strategies. Special attention is devoted to highlight potential research hots and directions for each strategy, thus providing new insights into the surface modification of TiO₂ nano-photocatalysts for further development in these fields.

Fig. 7. Surface modification strategies for TiO2 nano-photocatalysts.

Conceptually, the low visible light absorption of TiO₂ nanomaterials is a key factor restricting their photocatalytic efficiency. Owing to their wide band gap, TiO₂ nanomaterials can only use 3%-4% of UV light, and cannot fully exploit the visible light spectrum (~48% of the total sunlight energy). Thus, to address this issue, research has focused on increasing the visible light activity of TiO₂ nanomaterials by shifting the onset of their light response from the UV to visible region. During the past decades, second-generation (metal-doped) and third-generation (non-metal-doped) TiO₂ nanomaterials have been successively developed and widely applied in various photocatalytic fields ^[1]. Five strategies have been used to enhance the visible light activity of TiO₂ nanomaterials: impurity doping, structural defects insertion, surface sensitization, surface plasmon resonance effect generation, and solid solution ^[30]. Herein, two promising strategies i.e., impurity doping and surface sensitization, are highlighted.

Impurity doping can introduce localized electronic states into the band gap of wide band gap semiconductors, thus achieving an obvious red shift in the optical response and a significant enhancement in the visible light activity of the semiconductor material through a two-step photo-excitation by low energy visible light photons ^{[12, 30]}. For example, mesoporous Cr-doped TiO₂ exhibits a marked red shift with respect to pure TiO₂ (Fig. 8(a)). This shift can be attributed to the charge-transfer band Cr³⁺ → Ti⁴⁺ or 4A_2g → 4T_1g of Cr³⁺ in an octahedral environment ^[98]. The resulting homogeneous mesoporous Cr-TiO₂ shows a significant enhancement in the photocatalytic decomposition rate of methylene blue (MB) in the visible light region (Fig. 8(b)) owing to enhanced visible light absorption, crystallinity, and three-dimensional- interconnected mesochannels ^[98]. Although some doped TiO₂ solid solutions feature wide band gaps, improved photocatalytic activities can be achieved owing to the increased concentration of hydroxyl radicals ^[99], oxidation-reduction potential ^[100], or improved lattice parameters ^[101]. Generally, suitable dopants can efficiently narrow the band gap of TiO₂-based photocatalysts by forming new localized electronic states or mid-gap states, thus leading to enhanced photoactivities ^{[12, 102]}. To date, among various doped TiO₂ nanomaterials, non-metal nitrogen element has proved to be the best dopant for TiO₂ with respect to photocatalytic activity for environmental applications ^[45]. It was theoretically and experimentally demonstrated that the substitutional doping of N for O in anatase TiO₂ could efficiently narrow its band gap through the formation of mixed N 2p and O 2p states, thus achieving improved visible light photocatalytic activities ^[103]. Although there are debates on the origins of the red shift of the absorption onset of N-doped TiO₂ to the visible region ^[12], thousands of research papers on N-doped TiO₂ nanomaterials as visible light sensitive photocatalysts have been reported in the past 15 years ^{[45, 96, 104, 105, 106, 107, 108, 109]}. For example, N-doped TiO₂ nanomaterials with mesoporous structures and exposed (001) facets have been reported to exhibit excellent visible light photocatalytic activities owing to improved visible light absorption and increased surface active sites content ^{[104, 105, 110]}. More interestingly, it was also reported that ZrO₂-modified TiO_2−xN_x displayed higher porosity, higher specific surface area, and an improved thermal stability over the corresponding unmodified TiO_2−xN_x samples ^[96]. Recently, the doping of other non-metal species, such as F, S ^{[111, 112]}, P ^{[113, 114]}, B, and C ^{[115, 116, 117, 118, 119]}, and co-doping of these non-metal species ^{[120, 121, 122, 123, 124, 125, 126, 127]} and metal species ^{[128, 129, 130]} have also been widely investigated to further increase the activity and stability of TiO₂ nanomaterials. However, to date, the photocatalytic reaction rates of N-doped TiO₂ nanomaterials remain low owing to their poor visible light absorption. In the future, based on the deeper understanding of mechanisms in photo-excited states, it is expected that the visible light photocatalytic rates and stabilities of N-doped TiO₂ nanomaterials can be further improved through different engineering modification strategies such as co-doping, surface co-catalyst and passive layer loading, surface heterojunction construction, and usage of their favorable synergistic effects, thus extending the practical applications of non-metal-doped TiO₂ materials ^{[12, 45]}.

Fig. 8. (a) UV-visible absorption spectra of pure TiO2 (1) and Cr-TiO2 (2). (b) Comparison of the photoactivities of pure TiO2 (1) and Cr-TiO2 (2) toward the decomposition of MB under visible light irradiation. Reprinted with permission from Ref. [98].

Also, sensitizing TiO₂ using semiconductors with narrower band gaps has proven to be a promising and practical strategy to enhance the visible light activity of TiO₂. Composites of TiO₂ and visible light semiconductors can not only enhance the visible light response of TiO₂, but also improve the stability of sensitizers with narrower band gaps owing to the efficient charge separation in the semiconductor-sensitizer interface electrical fields ^[30]. For example, CdS-sensitized TiO₂ crystallites generally exhibit significantly enhanced visible light activities toward the photodegradation of organic pollutants ^{[131, 132, 133]}, photocatalytic CO₂ reduction ^[32], and hydrogen evolution in water splitting ^[134] owing to the improved charge separation and visible light absorption capabilities. In another example, the quantum size effects of MoS₂ and WS₂ nanoclusters were examined to alter the energy levels of the CB and VB edges to favor interparticle charge carrier separation and enhance the visible light photocatalytic activities of TiO₂ toward the degradation of MB and 4-chlorophenol in aqueous suspension (Fig. 9) ^[135]. Other sensitizers, such as CdSe nanoparticles ^[136], metal phthalocyanine ^[137], graphene quantum dots ^[138], and plasmonic metals ^{[139, 140, 141, 142, 143, 144, 145]}, have also been successfully used to enhance the visible light photocatalytic activity of TiO₂. In the future, further modifications should be investigated to enhance the stability and activity of the above systems. More importantly, green sensitizers, such as graphene, carbon, and C₃N₄ quantum dots, deserve more attention in developing sensitized TiO₂-based photocatalysts ^{[146, 147, 148, 149, 150]}.

Fig. 9. (a) Photodegradation of 4-chlorophenol over pure TiO2 and MoS2 and WS2 nanocluster-sensitized TiO2 under visible light irradiation (λ > 400 nm). (b) Energy levels of the CB and VB edges versus normal hydrogen electrode (NHE) for pure TiO2, WS2, and MoS2 with various sizes at pH 7 (Q-size, quantum size). Reprinted with permission from Ref. [135].

Constructing surface heterojunctions has been demonstrated to be another promising strategy to promote the photocatalytic activities of TiO₂-based nanomaterials. It is generally believed that well-defined junctions effectively facilitate charge transfer and suppress recombination of photo-generated electrons and holes, leading to extremely high activity and stability ^{[12, 151, 152, 153, 154, 155, 156]}. Herein, special emphasis will be placed on the heterojunctions between TiO₂ and other semiconductors or nano carbon materials.

To date, TiO₂ semiconductor heterojunctions including n-n and p-n junctions have been widely constructed to improve charge separation and prolong the lifetime of photo-excited charge carriers. For instance, TiO₂-based n-n heterojunctions, including CdS/TiO₂ ^{[32, 157, 158]}, CeO₂/TiO₂ ^[159], WO₃/TiO₂ ^{[160, 161, 162]}, SrTiO₃/TiO₂ ^[163], ZnO/TiO₂ ^[164], BiVO₄/TiO₂ ^{[165, 166]}, TiO₂/In₂O₃ ^[167], Ag₂CO₃/TiO₂ ^[168], and C₃N₄/TiO₂ ^{[169, 170, 171, 172, 173, 174]}, have been available for various photocatalytic applications. Such semiconductor heterojunctions can suppress the recombination rate of e⁻-h⁺ pairs owing to generation of interfacial electrical fields. In contrast, TiO₂-based p-n heterojunctions, such as BiOI/TiO₂ ^[175], NiO/TiO₂ ^[176], Cu₂O/TiO₂ ^{[177, 178, 179]}, CuCrO₂/TiO₂ ^[180], and MoS₂/TiO₂ ^[181], have been widely examined for extending the visible light absorption of TiO₂ semiconductors with wider band gap and enabling effective charge separation driven by the photo-induced generated p-n junction interface. For example, Dai et al. ^[175] constructed a p-n junction of BiOI/TiO₂ nanotube arrays (NTs) by coating the walls of TiO₂ NTs with BiOI using a novel impregnation-hydroxylation method. The results demonstrated a 3-fold enhancement in the photoelectrocatalytic degradation of methyl orange (MO) over the p-n junction of BiOI/TiO₂ under visible light irradiation. The schematic diagrams for the formation of the p-n junction between p-BiOI and n-TiO₂ are shown in Fig. 10. As displayed in Fig. 10, upon light irradiation, an equilibrium state is formed between BiOI and TiO₂ by raising the energy band of BiOI and reducing the Fermi level of TiO₂. The resulting inner electric field can efficiently drive the photo-generated electrons and holes to TiO₂ and BiOI, respectively, thus reducing the recombination of e⁻-h⁺ pairs and accelerating the photocatalytic reactions with the reactants adsorbed on the surfaces of the heterojunction ^[175]. In the future, well-defined TiO₂ semiconductor heterojunctions, such as junctions with exposed facets ^{[166, 172, 177]} and core-shell structures ^{[182, 183, 184]}, are believed to be very promising and generate much interest in constructing highly efficient TiO₂-based heterojunction systems.

Fig. 10. Schematic diagrams of the energy bands of p-BiOI and n-TiO2 before light irradiation (a) and formation of a p–n junction under visible light irradiation (b). Reprinted with permission from Ref. [175].

TiO₂ exists as four main crystalline polymorphs in nature: tetragonal rutile, tetragonal anatase, orthorhombic brookite, and monoclinic TiO₂(B) ^{[11, 41]}. Distortion of the uniformly shaped TiO₆ octahedron units in the lattice structure of the four types of TiO₂ can cause different mass densities and band gaps in the different phase forms of TiO₂ ^[11]. Therefore, phase junctions would form between two phases upon contact, with suitable band alignments. Such phase junctions can greatly promote the photo-carrier separation and enhance the photocatalytic quantum yields ^{[185, 186, 187, 188]}. Yu et al. ^[189] demonstrated that the high brookite content in mesoporous TiO₂ with a bicrystalline (anatase and brookite) framework played an important role in enhancing its photocatalytic activity toward the oxidation of n-pentane in air. In another study, Li and coworkers ^[190] demonstrated that the formation of exposed surface anatase-rutile phase junction is critical to achieve enhanced photocatalytic activity of TiO₂ nanoparticles owing to the efficient synergic effects of the two phases. Notably, the phase junctions between most frequently studied TiO₂ phases (rutile or anatase) and less common TiO₂ phases (brookite and B phase) and charge-separation mechanisms deserve more attention in future studies ^[191].

Apart from TiO₂ semiconductor heterojunction and phase junctions, the junctions between TiO₂ and nano carbon materials have attracted much attention owing to their superior photocatalytic performance. For instance, the junctions between TiO₂ and conductive carbon nanotubes ^{[192, 193, 194, 195, 196, 197, 198, 199, 200]}, graphene ^{[138, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211]}, carbon black ^{[212, 213, 214, 215]}, graphitic carbon ^{[95, 216, 217]}, or C60 ^{[218, 219]} have been extensively fabricated to efficiently enhance the photocatalytic activity for various applications. The improved photoactivities can be primarily attributed to the enhanced charge separation by the Schottky junction between TiO₂ and the highly conductive nano carbon materials and the improved adsorption performance through the introduction of the carbon materials. For example, Xiang et al. ^[220] fabricated graphene-modified TiO₂ nanosheets with exposed (001) facets using a two-step hydrothermal process. The H₂ production rate of the composite with an optimal graphene content of 1.0 wt% was 41 times higher than that achieved by pure TiO₂ nanosheets. The schematic illustration for the charge transfer/separation and the proposed mechanism are shown in Fig. 11. As displayed in Fig. 11, it is believed that the graphene sheets, as an electron acceptor, can efficiently extract the photo-generated electrons from the TiO₂ nanosheets and thus promote charge carrier separation owing to the outstanding electrical conductivity of the graphene sheets and suitable potential of graphene/graphene^−• (−0.08 V vs. standard hydrogen electrode (SHE), pH = 0). In the future, more attention should be paid to the in situ growth of TiO₂ on three-dimensional graphene ^{[221, 222]} with high quality junctions and the hybridization of TiO₂ and composite carbon materials doped with bi- or multi-heteroatoms owing to the excellent synergistic effects exhibited by combining different carbon materials and heteroatom doping ^{[218, 223, 224, 225]}.

Fig. 11. (a) Schematic illustration of the charge transfer and separation in a graphene-modified TiO2 nanosheets system under UV light irradiation. (b) Proposed mechanism of photocatalytic H2 production under UV light irradiation. Reprinted with permission from Ref. [220].

As shown in Fig. 4, slow surface reaction kinetics is an important factor negatively affecting the photocatalytic activity of TiO₂, especially for uphill H₂ evolution and CO₂ reduction reactions. To address this issue, an efficient strategy is to load suitable co-catalysts to accelerate surface reaction kinetics ^[12]. To date, various co-catalysts, including noble metal, earth-abundant metal, and metal-free and hybrid co-catalysts, have been developed to enhance photocatalytic H₂-/O₂ evolution and CO₂/O₂ reduction reactions, as summarized in Table 1. For further information, the reader is invited to refer to some recently published reviews ^{[12, 226, 227, 228]}. As shown in Table 1, noble metal, earth-abundant metal (such as Ni(OH)₂ ^[229], Cu(OH)₂ ^[230], MoS₂ ^[206], NiS_x ^[231], and NiO_x^{[232, 233, 234]}), and metal-free and hybrid co-catalysts (Table 1) have been proven to greatly enhance the H₂ evolution activities of TiO₂ nano-photocatalysts. For example, Hu et al. ^[235] reported that Pt₃Co alloy co-catalyst loaded onto TiO₂ could achieve a 1.5-fold enhancement in the photocatalytic activity toward H₂ evolution when compared with Pt/TiO₂ photocatalyst. The high activity was attributed to the accumulation of photo-excited electrons onto Pt₃Co and improved H₂ evolution kinetics ^[235]. In another example, low-cost Cu(OH)₂ clusters, as an excellent co-catalyst, have been demonstrated to significantly enhance the H₂ evolution activity of TiO₂ (Fig. 12(a)) ^[230]. It is believed that the suitable potential of Cu(OH)₂/Cu lies between the CB of anatase TiO₂ (−0.26 V) and reduction potential of H⁺, thus favoring the efficient transfer of electrons and enhancements in the photocatalytic H₂ production activity (Fig. 12(b) and (c)). Typical O₂ evolution co-catalysts, which can be used in water oxidation or overall water splitting (combined with the H₂ evolution co-catalysts), are also listed in Table 1.

Table 1 Summary of various types of co-catalysts employed for photocatalysis. Photocatalytic reaction Noble metal co-catalysts Earth-abundant metal co-catalysts Metal-free and hybrid co-catalysts H2 evolution Pt, Au, Ag [242], Pt3Co [235], AuPd [243,244] Ni(OH)2 [229], WS2 [245–247], CuO [248,249], Cu(OH)2 [230], MoS2 [250,251], NiSx [231,252–254], NiOx [232–234,255], Co and Ni [256], Mo2C and Mo2N [257], Ni2P [258] g-C3N4/N-doped graphene [259–262], (N,P)-co-doped graphene [263], graphene [220,264], MoS2/graphene [206,265], MoS2 (or WS2) clusters/ N-doped graphene [266–268] O2 evolution RuO2 [269], IrO2 [269], Ru, Ir, and Pt [270] NiCo2O4 [271,272], CoMoO4 [273], Mn2O3 [274], CoOx [275], Co(OH)2 [276], cobalt phosphate (Co-Pi) [277], Ni1–xFexOOH [278] g-C3N4/carbon nanotubes [279], Co3O4-carbon [280], graphene/ZnCo (or NiCo) layered double hydroxide [281,282], NiCo2O4/N-doped graphene [283], Ni@N-doped graphene [284], NiCo2O4/carbon nanotubes [285] CO2 reduction Pt [236,237], Au [286,287], Pd [288], Ag [289–294], RuO2 [295–299], bimetallic Ag-Pt [300], Au-Cu [301], Cu-Pt [302,303] Cu [304], CuO [305,306], Cu2O [30,38], NiOx [307,308], CoOx [309] Graphene [239,310,311], carbon nanotubes [312,313], carbon dots [314,315], NiOx/graphene [316], carbon nanotubes@Ni [317], Pt@Cu2O [318] O2 reduction Pt [319,320], Ag [321,322], Au25 nanoclusters [323], bimetallic Au/Pt [324] NiCo2O4 [325–327], MnO2 [328], MxFe3−xO4 (M = Fe, Cu, Co, Mn) [329], CaMnO3 [330,331] g-C3N4@carbon [332,333], N/S or (B,N)-co-doped graphene [334,335], N-doped graphene/porous carbon [336], Co3O4/graphene [337], polyelectrolyte/carbon nanotubes [338], Mn3O4/N-doped graphene [339], graphene [340–344], N-doped graphene (carbon spheres, nanotubes, or nanocages) [345–349]

Table 1 Summary of various types of co-catalysts employed for photocatalysis.

Fig. 12. (a) Comparison of the photocatalytic activities of different samples CX (X, nominal molar ratios of Cu(OH)2 to TiO2+Cu(OH)2, mol%) toward the photocatalytic H2 production from ethylene glycol aqueous solution under UV-light-emitting diode irradiation. (b) Proposed mechanism of photocatalytic H2 production under UV light irradiation. (c) Schematic illustration of charge transfer and separation. Reprinted with permission from Ref. [230].

Interestingly, for the photocatalytic reduction of CO₂, co-catalyst loading can not only accelerate the kinetics of CO₂ reduction, but also achieve improved selectivity. A selective production of hydrocarbon (such as CH₄, other alkanes, and olefin) with a rate of 160 µL/(g·h) was reported over nitrogen-doped TiO₂ nanotube arrays loaded with both Cu and Pt co-catalysts under outdoor global AM1.5 sunlight (Fig. 13(a)) ^[236]. More recently, an optimum CH₄ yield of 1361 μmol/(g·h) was obtained over one-dimensional TiO₂ single crystal films coated with ultrafine Pt nanoparticles (0.5−2 nm in diameter, Fig. 13(b)) ^[237]. Also, Cu₂O, as a co-catalyst, can enhance the photocatalytic selectivity toward the generation of methanol ^{[30, 238]}. Other co-catalysts, such as nano carbon materials, and NiO_x and their hybrids (Table 1), have proven to be good co-catalysts for the selective production of CH₄ and CH₃OH, respectively ^{[30, 239, 240, 241]}.

Fig. 13. (a) Product generation rates obtained using a nitrogen-doped nanotube array film (annealed at 600 °C) surface-loaded with both Pt and Cu co-catalysts [236]. (b) CO and CH4 yields obtained using P25, pristine TiO2 columnar films (TiO2 film), and Pt-TiO2 films with different Pt deposition times [237].

Additionally, co-catalysts have been widely used to accelerate O₂ reduction reactions, which have been regarded as the rate-determining step in solar photocatalytic mineralization of organic pollutants ^{[25, 350]}. For example, Yu et al. ^[351] fabricated Ag-TiO₂ multi-phase nanocomposite thin films using liquid-phase deposition method. The results demonstrated that the photocatalytic activity of the Ag-TiO₂ multiphase nanocomposite thin films toward the degradation of MO was about 6.3 times higher than that of pure TiO₂ thin films. The enhanced activity was attributed to the formation of multi-junctions and improved O₂ reduction kinetics over Ag co-catalyst (Fig. 14). Similarly, Ag@TiO₂ core-shell nanocomposite nanowires exhibited significantly enhanced photocatalytic decolorization of rhodamine B (RhB) in aqueous solution at ambient temperature under UV light irradiation ^[352]. Furthermore, Au ^{[323, 353]} and Pt ^[354] nanoparticles/clusters have been demonstrated to be excellent co-catalysts toward the photocatalytic oxidation decomposition of organic pollutants over TiO₂ nano-photocatalysts. The improved charge separation and O₂ reduction rate were responsible for the significantly enhanced activities. In the future, it is expected that further earth-abundant, metal-free, and hybrid co-catalysts (Table 1) will be developed and applied in the photocatalytic degradation of organic compounds.

Fig. 14. Interface charge carrier transfer dynamics of anatase–rutile, Ag–anatase, and Ag–rutile within Ag-TiO2 composite thin films. Reprinted with permission from Ref. [351].

Fabricating hierarchical nanopores in TiO₂ nano- photocatalysts has also proved to be a feasible strategy to efficiently increase the accessible surface areas, prevent the aggregation of nano-photocatalysts, and enhance the photocatalytic activity and selectivity of nano-photocatalysts. Importantly, these hierarchically porous structures in TiO₂ photocatalysts feature readily accessible pore walls, high specific surface areas, and interconnected pore systems, which can not only enhance light harvesting and reactant adsorption efficiencies but also optimize the transport of guest species to framework binding sites (Fig. 15). Thus, the synergetic effects of different dimensional levels and multi-stage pore structures can result in superior photocatalytic performances in hierarchical photocatalysts. Unlike the application of TiO₂ nano- photocatalysts with hierarchical nanopores in photocatalytic H₂ evolution ^{[355, 356]} and CO₂ reduction ^{[357, 358]}, their usage in photocatalytic pollutant degradation have been more extensively studied ^{[205, 359, 360, 361, 362, 363, 364, 365, 366, 367, 368, 369, 370, 371, 372, 373, 374]}. Thus, we focus on the application of hierarchical TiO₂ photocatalysts in the photocatalytic degradation of organic compounds.

Fig. 15. Three-dimensional mesoporous architecture in a TiO2 nanocrystalline thin film with continuous and periodically porous networks that offer transport routes for gaseous reactants to move into and out of the framework. Reprinted with permission from Ref. [368].

Generally, such hierarchical TiO₂ photocatalysts are fabricated using (surfactant-)templated methods ^{[365, 368]} or oriented assembly strategies of nano-sized building blocks such as nanorods, nanotubes, and nanoplates ^{[375, 376, 377, 378, 379, 380, 381]}. For example, Wang et al. ^[365] fabricated macro/mesoporous TiO₂ photocatalysts via a surfactant-templated method. The resulting photocatalysts featured an intact macro/mesoporous structure (Fig. 16), which exhibited a 1.6-fold enhancement in photocatalytic activity toward ethylene photodegradation in gas-phase medium when compared with commercial P25 TiO₂. The enhanced activity can be attributed to the increased light absorption efficiency and improved diffusion of gaseous molecules owing to the existence of macrochannels. In contrast, a 30% drop in photocatalytic activity was observed in the absence of light-penetrating and molecular-diffusing macrochannels, further indicating the key roles of macrochannels in improving the photocatalytic activity. Enhanced photocatalytic activities toward the decomposition of formaldehyde and acetone in air were also demonstrated using hierarchical TiO₂ photocatalysts ^{[205, 359, 362, 382]}. For example, hierarchically flower-like TiO₂ superstructures (HFTS) were fabricated through self-assembly of anatase TiO₂ nanosheets with exposed (001) facets (up to 87%, Fig. 17(a)-(d)) ^[360]. The resulting HFTS exhibited a higher photocatalytic activity than tabular-shaped TiO₂ particles (TSTP) and P25 owing to the synergistic effect of the intermeshed nanosheets, hierarchical porous structures, and exposed (001) facets (Fig. 17(e) and (f)). Similarly, hierarchically porous TiO₂ photocatalysts exhibited enhanced photocatalytic activities toward the degradation of organic dyes in aqueous solution under UV illumination ^{[360, 361, 367]}. Additionally, it was found that photocatalytic selectivity in decomposing MB and MO could be achieved by simple surface modification of hollow TiO₂ spherical photocatalysts ^[383]. Recently, Li et al. ^[384] reported that porous TiO₂ materials with a hierarchically interconnected structure could be fabricated by templating Pickering high-internal phase emulsions. It was demonstrated that the parent emulsion droplets and the interfacially adsorbed microgel stabilizers could promote the formation of macropores/interconnecting large pores with sizes of ∼50 and ∼10 μm and smaller pores (∼100 nm). The photocatalytic degradation of RhB over porous materials with such bimodal porous structures was considerably more efficient than that over crystalline commercial P25 sample owing to the existence of the hierarchical pore structure. However, there are very few reports on the application of hierarchically structured semiconductors in CO₂ photoreduction and photocatalytic organic synthesis ^{[358, 385, 386]}. Therefore, it is highly expected that low-cost, hierarchically structured semiconductors with high activities will be subsequently exploited and applied in photocatalytic CO₂ reduction and organic synthesis in the future. Furthermore, different modification strategies for such hierarchical TiO₂ photocatalysts, such as doping, formation of hetero-junctions, and loading of suitable co-catalyst, would be worth investigating in future studies.

Fig. 16. SEM images of TiO2 monolithic particles calcined at 350 °C (a-c) and 500 °C (d). Reprinted with permission from Ref. [365].

Fig. 17. SEM (a, b), TEM (c) and high-resolutionTEM images (d) of HFTS. Inset in (a) is a magnified image of an individual spherical superstructure. (e) Comparison of the photocatalytic activities of HFTS, TSTP, and P25 toward the photocatalytic decomposition of acetone in air. (f) Dependence of the concentration of produced CO2 on irradiation time using HFTS, TSTP, and P25 as photocatalysts. Reprinted with permission from Ref. [360]

Recently, it was shown that using surface F effects could also significantly improve the physicochemical and photocatalytic properties of TiO₂ photocatalysts toward the degradation of organic compounds ^{[25, 88]}. Generally, surface F doping of TiO₂ cannot lead to an obvious red shift in the fundamental absorption edge of TiO₂ ^{[387, 388]}. However, surface fluorination can significantly alter the hydrophobic-hydrophilic properties of TiO₂, increase the photogeneration of ^•OH and surface acidity of TiO₂ particles (Fig. 18), and reduce the recombination rate of e⁻-h⁺ pairs, thus achieving a significant enhancement (by a factor of 1.4-6.3) in the photocatalytic oxidation rates of various (in)organic compounds (Table 2) ^[25]. It is believed that surface fluorination can not only trap the CB electrons owing to the high electronegativity of fluorine and thus accelerate the O₂ reduction, but also favor the transformation of VB holes into free OH radical species, thereby reducing the recombination of photo-generated electrons and holes and enhancing the photocatalytic activity. Importantly, the enhanced formation rate of free (unbound) hydroxyl radicals was evidenced on the surface of F-TiO₂ ^[25]. When compared with the surface-adsorbed OH radicals on TiO₂, free OH radicals are highly mobile and more oxidative owing to their higher redox potential and the existence of remote photocatalytic oxidation ^{[25, 88]}. For example, Yu et al. ^[389] fabricated F⁻-doped TiO₂ photocatalyst with anatase and brookite phases by hydrolysis of titanium tetraisopropoxide in a mixed NH₄F/H₂O solution. The prepared F⁻-doped TiO₂ powders exhibited enhanced photocatalytic activity toward the oxidation of acetone in air. It was believed that F⁻ doping improved the crystallinity of anatase, suppressed the formation of brookite phase, prevented phase transition of anatase to rutile, and enhanced the visible light absorption, thus greatly promoting the photocatalytic activity. Further research demonstrated that all fluorinated, F-doped, and trifluoroacetic acid-modified TiO₂ photocatalysts exhibited remarkably enhanced activities and stabilities toward the degradation of organic pollutants in aqueous solution or oxidation decomposition of acetone in air (Table 2) ^{[390, 391, 392, 393, 394, 395]}. The enhanced activity was ascribed to the reduced recombination rate of photo-generated electrons and holes and enhanced formation of free OH radicals ^{[392, 394, 395]}. The combination of fluorination and other surface modification strategies, such as impregnation of co-catalysts ^[396], co-doping ^{[397, 398]}, construction of hierarchical mesoporous structures and heterojunctions, would be worth investigating in future studies.

Fig. 18. Hole transfer and transformation into hydroxyl radicals on bare TiO2 and fluorinated TiO2 (F-TiO2) surfaces under UV irradiation. Reprinted with permission from Ref. [25].

Table 2 Summary of surface-fluorinated TiO2 for photocatalysis. Photocatalyst Preparation method Application (Enhancement factor) Enhancement mechanism Ref. (Year) TiO2 powders Hydrolysis of titanium tetraisopropoxide in a solution of NH4F Oxidation of acetone in air (1.4×) Improved crystallinity of anatase, suppressed formation of brookite, and enhanced visible light absorption [389] (2002) Mesoporous TiO2 thin films Treatment with trifluoroacetic acid (TFA) Oxidation of acetone in air (2×) Reduced recombination rate of e−-h+ pairs by the chemisorbed TFA complex [390] (2003) Mesoporous TiO2 microspheres Sonochemical/hydrothermal treatment of titanium isopropoxide in the presence of NaF Degradation of MB (7.6×) Increased content of hydroxyl radicals and adsorption of MB, reduced recombination rate of e−-h+ pairs [391] (2009) Mesoporous TiO2 powders Hydrothermal method in a NH4HF2/H2O/C2H5OH mixed solution Oxidation of acetone in air (3×) Increased content of hydroxyl radicals and crystallinity of anatase, reduced recombination rate of e−-h+ pairs [394] (2009) TiO2 hollow microspheres Hydrothermal treatment of Ti(SO4)2 in the presence of TFA Oxidation of acetone in air (2×) Enhanced transfer of light and acetone, reduced recombination rate of e-/h+ pairs [395] (2010) TiO2 hollow microspheres Hydrothermal strategy using NH4HF2 as a controlling agent Decolorization of RhB (2×) Improved crystallinity, reduced recombination rate of e−-h+ pairs [393] (2010) Square-shaped TiO2 nanocrystals Sonochemical treatment of titanium isopropoxide in the presence of NH4F Degradation of phenol (6.3×) Reduced recombination rate of e−-h+ pairs, increased content of surface hydroxyl groups and crystallinity of anatase [392] (2012) Hemisphere-like TiO2 Hydrothermal treatment of TiF4 aqueous solution in the presence of starch Degradation of acid orange II under visible light Formation of oxygen vacancies and good crystallinity [387] (2010) Hierarchically nanoporous TiO2 spheres Hydrothermal treatment of TiF4 powder in HCl solution Degradation of 4-chlorophenol under visible light Creation of oxygen vacancies, efficient transport of light and reactant molecules [388] (2006) TiO2 powders Spray pyrolysis from an aqueous solution of H2TiF6 Decomposition of acetaldehyde and trichloroethylene (16.4×) Creation of oxygen vacancies [402] (2005) Porous (F,Ce)-co-doped TiO2 Ultrasound irradiation of titanium isopropoxide in the presence of NaF Degradation of acid orange II (3×) Increased content of surface hydroxyl groups, reduced recombination rate of e−-h+ pairs [403] (2012) Hollow TiO2 microspheres Hydrothermal treatment of Ti(SO4)2 in the presence of NH4F and ethanol Selective decomposition of MO and MB Tunable adsorption selectivity, controlled hierarchical assembly, exposed (001) facets [383] (2010) Hierarchically porous TiO2 films Hydrothermal treatment of Ti foil in a dilute aqueous HF solution Selective degradation of MO and MV Controlled surface charge and adsorption selectivity [361] (2011) (F,N)-co-doped TiO2 nanocrystals Solvothermal treatment of Ti(OBu)4 in the presence of NH4F and ethanol Degradation of MB under visible light Formation of Ti3+ centers and oxygen vacancies [404] (2013) F-doped TiO2 nanoparticles Hydrolysis of tetrabutyl titanate in the presence of NH4F and ethanol Degradation of MB under visible light Reduced recombination rate of e−-h+ pairs, increased content of hydroxyl groups and visible light absorption [405] (2014)

Table 2 Summary of surface-fluorinated TiO2 for photocatalysis.

Furthermore, it should be pointed out that the surface fluorination of TiO₂ in acidic environment could result in the generation of a positively charged surface, thus promoting preferential adsorption of negatively charged contaminants (such as MO) and favoring their degradation ^[361]. As shown in Fig. 19, fluorinated TiO₂ films exhibited higher selectivity toward the photodecomposition of MO. In contrast, selective degradation of MB could be achieved over TiO₂ films subjected to NaOH washing treatment ^[361]. Thus, the surface-bonded fluoride and hydroxyl groups of hollow TiO₂ microspheres could achieve selective adsorption and photocatalytic decomposition of negatively charged MO and positively charged MB, respectively ^{[361, 383]}. More interestingly, the periodic density functional theory calculations revealed that F doping is not only thermodynamically stable for rutile, anatase, and brookite polymorphs of TiO₂, but also can preferentially stabilize the (001) over the (101) facets of TiO₂, thus achieving stable and enhanced photocatalytic activity over TiO₂ anatase nanoparticles with a reactive (001) surface and adsorption of F atoms ^{[399, 400]}. Time-resolved PL spectroscopy studies also demonstrated that the amount of surface traps originating from the long-lasting PL signal could be increased by fluorine doping, followed by calcination at high temperatures, which are beneficial for the improvement in photoactivity by ensuring long-living photo-produced charge couples ^[401]. In the future, more attention should be paid to more comprehensively understand the mechanisms of surface fluorides in enhancing photocatalysis.

Fig. 19. Comparison of the photocatalytic activity and selectivity of TiO2 films calcined at 180 °C for 1 h (T1) toward the degradation of MO and methyl violet (MV) before (T1-F) and after NaOH washing (T1-OH). The photocatalytic activity and selectivity are defined by the apparent rate constant (k) and the ratio (r) of k, respectively. Reprinted with permission from Ref. [361].

Since the breakthrough in synthesizing TiO₂ crystals constituting of 47% (001) and 53% (101) facets was first reported by Yang et al. ^[406] in 2008, exposing highly reactive facets has become a promising strategy to enhance the activity of TiO₂ for different photocatalytic applications ^{[43, 407, 408, 409, 410]}. Usually, different facets of TiO₂ crystals possess distinctive physiochemical properties, such as adsorption, and catalytic reactivity and selectivity, owing to the different optical/electronic structures ^[411]. Also, the crystal faces can facilitate the separation of electrons and holes ^[29]. Thus, exposing highly reactive facets of TiO₂ has gained much interest in the different photocatalytic applications ^{[29, 110, 121, 383, 412, 413, 414]}. For example, Zhang et al. ^[410] synthesized a micro-sheet anatase TiO₂ single crystal photocatalyst with a remarkable 80% level of reactive (001) facets using a microwave-assisted hydrothermal route involving titanium tetrafluoride and a tetrafluoroborate-based ionic liquid. The results showed that the anatase TiO₂ single crystals with exposed (001) facets exhibited considerably higher activities than those of the (001) unexposed TiO₂ single crystals (Fig. 20). The higher activities were due to the strong ability of the exposed facets to dissociatively adsorb water to form hydrogen peroxide and peroxide radicals. Furthermore, it was demonstrated that N-doped and (N,S)-co-doped TiO₂ nanosheets with exposed (001) facets exhibited significantly enhanced visible light activities toward photocatalytic H₂ production and degradation of 4-chlorophenol, respectively ^{[110, 121]}. The increased visible light absorption and highly reactive exposed (001) facets of TiO₂ nanosheets were thought to be the main reasons for the improved activities. In another example, Yu and coworkers ^[29] demonstrated that an optimal ratio of exposed (101) and (001) facets (45:55) of a TiO₂ nanosheet could achieve optimal activity toward the photoreduction of CO₂ to CH₄. It is believed that the synergistic effect of an overflow effect and surface heterojunction between the co-exposed (101) and (001) facets of anatase played an important role in enhancing the photocatalytic activity (Fig. 21). Importantly, the concept of surface heterojunction between two facets of TiO₂ nanocrystals may offer a new strategy for designing sufficiently efficient TiO₂-based photocatalysts ^[29].

Fig. 20. Photocatalytic efficiencies of photocatalyst samples with (☆) and without (∆) exposed reactive (001) facets toward the degradation of 4-chlorophenol under UV irradiation. For comparison, the photocatalytic result in the absence of photocatalyst is also shown. Reprinted with permission form Ref. [410].

Fig. 21. Overflow effect and hetero-facet junction between (101) and (001) facets in anatase. Reprinted with permission from Ref. [29].

To date, TiO₂ nano-photocatalysts have been widely applied in different photocatalytic fields owing to their superior properties. Based on the analysis of photocatalysis fundamentals and the surface chemistry of TiO₂,in the current review, we summarized and discussed some of the most recent state-of-the-art advancements in different surface modification strategies such as surface doping and sensitization, construction of surface heterojunctions, loading of nano-sized co-catalysts, increase in accessible surface areas, usage of surface F effects, and exposure of highly reactive facets. However, many challenges still remain in fabricating efficient and stable TiO₂-based visible light photocatalysts through suitable surface modification strategies and comprehensively understanding the associated photocatalytic enhancement mechanisms.

Special emphasis should be placed on the development of visible light TiO₂-based photocatalysts using simple and facile strategies such as co-doping, and loading of sensitizers and plasmonic metals. Especially, determining the mechanisms of visible light photocatalytic enhancements of N-doped TiO₂ nanomaterials deserves more attention. Importantly, their photocatalytic activity and stability should be improved further to meet requirements of practical applications. New, cheap, robust, efficient, and stable co-catalysts should be designed and developed in the near future. Furthermore, many non-TiO₂-based visible light photocatalysts are unstable under light illumination. Thus, fabricating suitable protection layers using amorphous or crystalline TiO₂ ^{[415, 416, 417, 418, 419, 420, 421, 422, 423, 424, 425]} and other Ti-containing semiconductors with wide band gaps (E_g > 3.0 eV) and high stability (such as SrTiO₃) ^{[426, 427, 428]} should be a promising strategy to enhance the activity and stability of other unstable/visible light photocatalysts and should be thoroughly investigated. Additionally, hybrids of TiO₂-based nanomaterials and earth-abundant semiconductors with narrow band gaps, such as Bi-based oxides ^{[429, 430, 431, 432]}, sulfides ^{[433, 434, 435]}, (oxy)nitrides ^[436], metal-free g-C₃N₄ ^{[6, 437]}, elemental α-sulfur ^[438], and red phosphorus ^{[439, 440, 441, 442]}, are a promising type of visible light photocatalyst. TiO₂/biomolecular composites are also very promising visible light photocatalysts for energy and environment applications ^[443]. In the future, it is expected that more efficient, cheap, and stable TiO₂-based visible light photocatalysts would be constructed through combined strategies.

The surface chemistry of TiO₂ photocatalysts and the complex photocatalysis mechanism remain interesting topics worth investigating, which will favor the design and development of highly reactive TiO₂-based photocatalysts. Although the quantum size effect can partially decease the utilization efficiency of light, increasing the CB/VB potentials of TiO₂ nanomaterials is an excellent potential strategy to improve their photocatalytic activities as a result of enhanced redox abilities ^[28]. In contrast, TiO₂ nanocrystals doped with rare earth metals or upconversion luminescence agents can exhibit much higher photocatalytic activity owing to improved utilization of visible or near infrared light ^{[444, 445]}. Furthermore, the important roles of surface F effects, hetero(phase)junctions, reactive facets, and co-catalysts in enhancing the photocatalytic activity of TiO₂ photocatalysts should be further examined and elucidated. In this regard, first-principle periodic calculations based on density functional theory have shown great potential for elucidating the molecular-level mechanisms of TiO₂ photocatalysis and thus deserve more attention ^{[29, 123, 124, 408, 446]}. Accordingly, the synergistic effect of these different surface modification strategies is also very interesting in developing highly reactive TiO₂-based photocatalysts ^{[341, 447]}.

Additionally, TiO₂ and their modified nanomaterials have been widely used in different fields related to photocatalysis including solar cells ^{[5, 6, 448, 449]}, light-induced superhydrophilicity ^[63], and photoelectrochemistry ^{[450, 451, 452]}. However, it should be noted that there has been no significant breakthrough in TiO₂-based photocatalysts since the work reported by Fujishima et al. ^[11] in 1972, and a breakthrough is unlikely to occur even in the near future. Importantly, TiO₂-based photocatalysts provide an excellent platform for thoroughly studying the mechanisms of photocatalytic reactions, which should be very helpful for the development of other highly efficient photocatalyst systems. We believe that the current review will contribute significantly toward further progress in the field of TiO₂-based photocatalysts for emerging applications in the photocatalytic fields in the future.