Photocatalytic Oxygen Evolution from Water Splitting

Abstract Photocatalytic water splitting has attracted a lot of attention in recent years, and O2 evolution is the decisive step owing to the complex four‐electrons reaction process. Though many studies have been conducted, it is necessary to systematically summarize and introduce the research on photocatalytic O2 evolution, and thus a systematic review is needed. First, the corresponding principles about O2 evolution and some urgently encountered issues based on the fundamentals of photocatalytic water splitting are introduced. Then, several types of classical water oxidation photocatalysts, including TiO2, BiVO4, WO3, α‐Fe2O3, and some newly developed ones, such as Sillén–Aurivillius perovskites, porphyrins, metal–organic frameworks, etc., are highlighted in detail, in terms of their crystal structures, synthetic approaches, and morphologies. Third, diverse strategies for O2 evolution activity improvement via enhancing photoabsorption and charge separation are presented, including the cocatalysts loading, heterojunction construction, doping and vacancy formation, and other strategies. Finally, the key challenges and future prospects with regard to photocatalytic O2 evolution are proposed. The purpose of this review is to provide a timely summary and guideline for the future research works for O2 evolution.


Introduction
Over the last few decades, the problems of energy shortage and environmental pollution have become the major issues that need to be solved for the sustainable development of modern society and economics. [1] Since TiO 2 photoelectrode was found to be able to decompose H 2 O into H 2 and O 2 under UV irradiation in 1972, semiconductor-based photocatalytic technologies requirement in the actual reaction situation due to the characteristic of multielectron/proton processes of water oxidation, the bandgaps of the used photocatalysts should be generally no smaller than 1.6 eV to conquer the defect of sluggish kinetics, which can be ascribed to the multistep reactions of the existent active intermediates during the O 2 evolution process. The reaction process is distinct under the acidic or alkaline condition, and the equations are as follows [26] 2H 2 O → O 2 + 4e − + 4H + (acidic condition) According to the three basic stages of photocatalytic reaction, the efficiency of photocatalytic O 2 evolution is also mainly determined by light absorption, photogenerated charge separation, and surface catalytic reaction, which can be expressed as follows [26] = absorption × separation × reaction In order to evaluate the photocatalytic activity and physicochemical stability of O 2 producing photocatalyst, the most common method is to test the turnover frequency (TOF), turnover number (TON), and apparent quantum efficiency (AQE) of the reaction system, and the corresponding formulas are in the following [27,28] In particular, the AQE is the ratio of the moles of generated O 2 per unit reaction time to the incident photons absorption number at a certain monochromatic wavelength, whereas the conversion efficiencies for O 2 evolution photocatalysts are distinct based on their different kinds of bandgaps (direct and indirect). Compared with indirect photocatalysts, the direct semiconductor materials have little change in the momentums during the migration process of photogenerated charge carriers after the absorption of incident photons, leading to the faster water oxidation rates than those of indirect ones. [26][27][28] Previously, the published reviews that are related to the photocatalytic O 2 evolution from water splitting mainly focused on some specific class of materials, such as BiVO 4 , metal-organic frameworks (MOFs), cobalt complexes, etc., [29][30][31] or limited to the theoretical calculations, defects design from crystal structures, etc. [32,33] So far, there is a lack of a systematic overview of the literatures on the photocatalytic O 2 evolution. Herein, we provide a review for the advances on the researches of photocatalytic O 2 evolution by starting with the fundamentals of photocatalysis and photocatalytic O 2 evolution, and then introduce various types of water oxidation photocatalysts based on different crystal structures, compositions, and morphologies. Then, we focus on the diverse strategies that improve the performance of photocatalytic O 2 evolution. Finally, the challenges and future prospects in this field are proposed. The summary contents are shown in Figure 1.

Photocatalysts for Oxygen Evolution
As O 2 evolution reaction is the rate-determining step during the process of photocatalytic OWS (schematic diagram shown in Figure 2), there are some critical requirements for these high efficiency water oxidation photocatalysts, including: i) the valence band (VB) potentials of the photocatalysts should be higher than those of the O 2 /H 2 O (1.23 eV); ii) to make more use of solar energy, bandgaps of the photocatalysts should be in the range of 1.23-3 eV; iii) the separation of the photogenerated electron-hole pairs within the photocatalysts should be fast under illumination, and then the corresponding redox reactions can occur in time on the surface of photocatalysts. [34,35] In addition, the water oxidation active sites should be sufficient on the surface of the photocatalysts in order to restrain the rapid recombination of the photogenerated electrons and holes on the surface. Nowadays, the heterogeneous photocatalysts are still the most commonly researched photocatalysts.

TiO 2
Titanium dioxide (TiO 2 ) can be mainly divided into four types: TiO 2 (B), brookite, anatase, and rutile, all of whose crystal structures are composed of TiO 6 octahedra (Figure 3a-d). As the bandgap of TiO 2 is 3.2 eV, the photogenerated charge carriers can only be excited by UV light with a wavelength of less than ≈390 nm. [36] The most common form of TiO 2 (B) is the layered titanate, and the brookite TiO 2 belonging to rhombic system has a low physicochemical stability. As anatase TiO 2 has more oxygen defects inside the crystal and greater band bending to capture the electrons and promote charge separation than the rutile, [37] the photocatalytic activity of the anatase TiO 2 is usually better than that of the rutile phase. [38] Commercial TiO 2 (P25) is composed of both anatase and rutile, which has a higher photocatalytic activity than the monocomponent one, due to the formation of energy barrier at the interface that improves the separation efficiency of the photogenerated charge carriers. [39,40] The commonly used routes for synthesis of anatase and rutile TiO 2 are hydrothermal and calcination methods. In general, the corresponding precursors were first synthesized with some titanium-containing substances and organic solvents, and then the hydrothermal/solvothermal or calcination processes at different temperatures were utilized to form the anatase, rutile or mixed-phase TiO 2 , [41][42][43] which were confirmed by transmission electron microscopy (TEM) (Figure 3e,f). For instance, with increasing the temperature, the crystalline phase gradually transformed from the TiO 2 (B) to anatase, and finally some of them were transformed into rutile, showing the 1D nanowires morphology with photocatalytic activity higher than that of the P25 nanoparticles due to the large specific surface area (Figure 3g). [44] In addition to the conventional preparation methods, researchers have also tried to combine the traditional routes with some additional conditions to synthesize TiO 2 with novel microstructures.  For example, the porous rutile TiO 2 nanorods were obtained with the synergistic silica as template under the hydrothermal environment, indicating that the crystal growth can be guided under external interference (Figure 3h-j). This 3D porous structure has a larger specific surface area and more exposed active crystal planes than the counterparts with traditional morphology, thus making TiO 2 more catalytically active. [45] 1D black TiO 2 nanotubes can be obtained through the annealing followed by aluminum reduction treatment, which can produce abundant oxygen vacancies (OVs) on the surface (Figure 3k,l). [46] Besides, different dimensions of TiO 2 can be obtained by the addition of diverse acids under the hydrothermal process, such as the 1D nanorods (by adding hydrochloric acid), 2D nanosheets (by adding hydrofluoric acid) (Figure 3m), etc. [47] Similarly, the 2D anatase TiO 2 flakes with microporous arrays were prepared after preliminary heat treatment and calcination with the rose petals as a natural mesoporous template, rendering TiO 2 more efficient light absorption and reactive sites, which is known as the petal effect (Figure 3n-r). [48] Compared with anatase TiO 2 , the rutile TiO 2 has a larger potential in the water oxidation due to the physicochemical characteristics. [49] In order to find out the relationship between O 2 generation capability and physicochemical properties of rutile TiO 2 , a series of TiO 2 were synthesized by calcination at temperature gradients. It was found that the crystallinity of the samples got better as the calcination temperatures increased (Figure 4g Figure 4i). The authors speculated that the water oxidation performance of the samples is related to not only the sizes and specific surface areas, but also the OVs. To verify this conjecture, the R-1073 sample was calcined in the hydrogen (H 2 ) atmosphere. There were no significant changes in the crystallinity after treatment, whereas the absorption intensity in the visible region of the H 2 -treated R-1073 was slightly higher than before, which may be due to the increase of the OVs concentration in the sample (Figure 4h). Regardless of NaIO 3 or FeCl 3 served as the sacrificial agent, the O 2 evolution performance of R-1073 after calcination under H 2 was greatly improved in comparison with the noncalcined one. [50]

BiVO 4
In recent years, the n-type semiconductor bismuth vanadate (BiVO 4 ) has attracted wide attention for photocatalytic water oxidation due to the moderate bandgaps (2.3-2.5 eV), suitable VB position, high photochemical stability, and low cost. [51] The typical crystalline phase of BiVO 4 is monoclinic scheelite, of which the crystal structure is composed of the distorted VO 4 tetrahedron and BiO 8 dodecahedron, as shown in Figure 5ac. [52] Density functional theory (DFT) calculation results demonstrated that BiVO 4 is a direct bandgap semiconductor with a bandgap of ≈2.2 eV (Figure 5d,f). It has a higher hole transferring rate than other water oxidation semiconductors, such as In 2 O 3 , [52] as verified by the density of states (DOS) (Figure 5e). [53] Figure 4. a-f) SEM images of samples calcined at varieties of temperatures. g) XRD patterns of samples calcined at varieties of temperatures. h) DRS of H 2 -treated/untreated R-1073. i) Time courses of oxidation evolution for untreated R-1073. Reproduced with permission. [50] Copyright 2014, American Chemical Society.
Fascinatingly, the large polaron can be formed under the polaronic state from holes within BiVO 4 , which contributes to the stability of the monoclinic scheelite BiVO 4 (Figure 5g,h). [54] BiVO 4 generally has a decahedral shape with smooth surfaces and sharp edges when prepared by hydrothermal reaction with NH 4 VO 3 and Bi(NO 3 ) 3 as the raw materials, and the diameters of these as-prepared BiVO 4 nanoparticles varied considerably depending on the different synthetic conditions (Figure 6a,b). [55,56] The lattice spacing of 0.308 nm was attributed to the {121} facets of BiVO 4 (Figure 6c,d). [56] Introduction of a certain amount of chelating agent to the hydrothermal environment usually produced BiVO 4 with the pure monoclinic phase and a starlike shape (Figure 6e,f). [57] Furthermore, the crystalline phases and photoabsorption of BiVO 4 also varied as the molar ratios of Bi(NO 3 ) 3 and NH 4 VO 3 changed (Figure 6g,h), leading to modulated photocatalytic properties of BiVO 4 . [58] Monoclinic BiVO 4 shows a large potential for water oxidation. [59,60] The active sites of photocatalytic reduction and oxidation reactions of the BiVO 4 decahedron are closely related to the exposing crystal facets, as revealed by photodeposition of metal/metal oxides. [53] As shown in Figure 6i-l, the BiVO 4 crystals presented a typical regular decahedron structure, and the front and side facets were {010} and {110} facets, respectively. Under illumination, Au and Pt particles were selectively deposited on the {010} facets, while MnO x was more prone to be deposited on the {110} facets, indicating that the reduction and oxidation active sites were located on the {010} and {110} facets, respectively. The following deposition of Au/MnO x and Pt/MnO x was also consistent with the above conclusions (Figure 6m,n). [61] As the exposing facets determine the photocatalytic activity, BiVO 4 with multiple exposing facets was studied. The BiVO 4 crystals with 30 facets were obtained by a decahedral-etching engineering under the participation of gold nanoparticles with the controllable concentrations (Figure 6o,p). In addition to the {010} and {110} facets, some uncommon ones, such as {121}, {321}, and {132} facets had also been exposed, offering more possibilities for the photocatalytic activity enhancement of BiVO 4 . The BiVO 4 with 30 facets showed a higher photocatalytic The band structure and f) corresponding brillouin zone of BiVO 4 . Reproduced with permission. [52] Copyright 2011, Royal Society of Chemistry. e) The total and projected density of states of BiVO 4 . Reproduced with permission. [53] Copyright 2015, Royal Society of Chemistry. g) The bond lengths of Bi-O bonds of BiO 8 dodecahedra from h) the density of states of nonpolaronic hole states for BiVO 4 . Reproduced with permission. [54] Copyright 2013, American Physical Society. O 2 evolution rate, 3-5 times that of decahedron BiVO 4 . The AQE reached 18.3% at 430 nm ( Figure 6q). [62] Besides exposing facets, the effect of morphology on the photocatalytic activity of BiVO 4 is another hot topic. By adjusting the pH value of reaction solution, BiVO 4 with the distinct morphologies were obtained. Under the acidic condition, BiVO 4 showed a regular decahedron structure, whereas leaf-like products were obtained in the alkaline media. Owing to the different crystal sizes and bandgaps, these BiVO 4 crystals synthesized at different pH showed different photocatalytic O 2 evolution performance, and the BiVO 4 decahedron exhibited the highest catalytic activity. [63]

WO 3
Tungsten trioxide (WO 3 ) has been widely used for photo(electro)catalytic water oxidation on account of its suitable bandgap (2.5-2.8 eV), positive VB, and high physicochemical stability. [64,65] The crystal structure of monoclinic WO 3 is composed of the distorted WO 6 octahedrons (Figure 7a), in which the marked A site is a good position for doping based on the perovskite-like structure of WO 3 . [66] The top of VB maximum is mainly dominated by the O 2p states (Figure 7e), which can be divided into three parts according to the energy size: low energy region (−7.5 Figure 6. a) SEM image and corresponding diagram (inset) of BiVO 4 . Reproduced with permission. [55] Copyright 2018, Royal Society of Chemistry. b) TEM image and size distributions (inset) of BiVO 4 nanoparticles. c) HRTEM image and d) lattice spacing of BiVO 4 nanoparticles. Reproduced with permission. [56] Copyright 2015, Wiley. e) XRD pattern and f) TEM image of starlike BiVO 4 . Reproduced with permission. [57] Copyright 2009, American Chemical Society. g) XRD patterns and h) DRS of BiVO 4 with Bi/V molar ratios of 1/1, 2/1, 4/1, and 6/1. Reproduced with permission. [58] Copyright 2015, Royal Society of Chemistry. i) SEM images of single BiVO 4 crystals, and BiVO 4 deposited by j) Au, k) Pt, l) MnO x , m) Au/MnO x , and n) Pt/MnO x ; the scale bar is 500 nm. Reproduced with permission. [61] Copyright 2013, Springer Nature. o) SEM image and p) corresponding diagram of 30-faceted BiVO 4 . q) Time courses of oxidation evolution for decahedron, truncated decahedron, and 30-faceted BiVO 4 . Reproduced with permission. [62] Copyright 2017, Wiley. a) The monoclinic structure of WO 3 ; blue and red spheres represent W and O atoms, respectively. Reproduced with permission. [66] Copyright 2008, American Physical Society. b-d) The electron charge density of WO 3 from the upper VB. e) The density of states of WO 3 , the zero of energy, and the CB minimum are indicated by the vertical solid line and vertical dashed line, respectively. Reproduced with permission. [67] Copyright 2012, American Chemical Society.
to −6 eV), medium energy region (−6 to −2 eV), and high energy region (−2 to 0 eV) (Figure 7b-d), while the bottom of conduction band (CB) minimum is mainly occupied by the W 5d states (Figure 7e). [67] It can also be inferred from the DOS of WO 3 that doping or defects with metal or nonmetal atoms is one of the effective ways to improve its photocatalytic performance.
Morphology regulation is the mostly involved among the researches of WO 3 photocatalysts. 1D WO 3 nanorod arrays on fluorine-doped tin oxide (FTO) were obtained by the hydrothermal-calcination method at different temperatures. When the calcination temperature increased to 450°C, the phase of as-synthesized 1D WO 3 nanorod arrays gradually transformed from orthorhombic to monoclinic phase (Figure 8a), as demonstrated by the emergence of (002), (020), and (200) peaks. After calcination at 500°C, the monoclinic WO 3 nanorod arrays with the exposed {202} facets were uniformly covered on the FTO substrate ( Figure 8b). [68] Similar to this convenient method, the 1D WO 3 nanowire arrays can be also successfully prepared on the FTO substrate under a relatively high temperature (≈550°C; Figure 8c,d). [69] Compared to 1D WO 3 , 3D WO 3 can promote the separation of photogenerated carriers more effectively. Using the wire mesh as a template, the 3D WO 3 nanosheets with a thickness of 200 nm were uniformly grown on this scaffold (Figure 8e-g). The exposed {002} facets indicated the successful synthesis of monoclinic phase (Figure 8h). [70] In addition to nanosheets, the monoclinic WO 3 can be deposited on FTO substrate in the form of nanofilms, enriching the 3D WO 3 materials (Figure 8i,j). [71] Due to the structure characteristic of WO 3 composed of WO 6 perovskite units, OVs are easier to be introduced into its lattice to obtain the defective 2D nanosheets through calcination in air, which improved the photo(electro)catalytic performance (Figure 8k-p). [72] The exposure ratios of the crystal facets have different effects on the photocatalytic O 2 evolution activity of WO 3 . The mono-clinic WO 3 cubes and WO 3 nanosheets with different exposing facets were obtained by hydrothermal/calcination treatment of the precursors (Figure 8q). The exposed facets of WO 3 cubes were mainly {002}, {020}, and {200} facets with exposing ratio of 1:1:1, while that of WO 3 nanosheets were mainly {002} facets (Figure 8r-y). The normalized O 2 evolution rate of the WO 3 cubes reached 5.9 µmol h −1 , which was more than eight times that of the WO 3 nanosheets (0.7 µmol h −1 ). In contrast, the WO 3 nanosheets showed high activity for photoreduction of CO 2 to CH 4 (0.34 µmol h −1 g −1 ), indicating that the exposing facets have large impacts on the photocatalytic activity of monoclinic WO 3 . [73] Besides the facets, the particle size is another factor to influence the photocatalytic activity of WO 3 . For instance, the O 2 evolution performance of as-synthesized WO 3 nanodots, nanoplates, and microcrystals was distinct from each other in the case of similar photoabsorption, among which the WO 3 nanodots had the highest O 2 production, implying the direct relationship between the particle sizes and photocatalytic activity of WO 3 . [74]

-Fe 2 O 3
According to the different elemental ratios and crystal structures, iron oxides can be divided into ferrous oxide (FeO), iron trioxide (Fe 2 O 3 ), and ferroferric oxide (Fe 3 O 4 ), among which the hematite Fe 2 O 3 ( -Fe 2 O 3 ) has a good performance in the photocatalytic field due to its narrow bandgap (1.9-2.2 eV), superior stability, and abundant natural resources compared with other forms of iron oxides. [75,76] From the perspective of crystal structure (Figure 9a,b), there are six equivalent crystalline directions in (001) plane in hematite, which are perpendicular to c-axis, based on the Cornell and Schwertmann theory. As the closely packed plane, the crystal growth rates of -Fe 2 O 3 along [100] direction and other five  [68] Copyright 2016, Elsevier. c) SEM image and d) TEM image of WO 3 nanowires. Reproduced with permission. [69] Copyright 2014, Royal Society of Chemistry. e-g) SEM images of wire mesh and corresponding WO 3 nanosheet arrays grown on it. h) TEM and HRTEM images (inset) of WO 3 nanosheet arrays. Reproduced with permission. [70] Copyright 2014, Wiley. i,j) SEM images of WO 3 nanofilms. Reproduced with permission. [71] Copyright 2014, Royal Society of Chemistry. k) TEM image, l) STEM image, and m) HAADF-STEM image of WO 3 nanosheets with nondefect treatment. n) STEM image and o) HAADF-STEM image of WO 3 nanosheets with defect treatment. p) Lattice diagram with OVs of as-synthesized WO 3 nanosheets with defect treatment. Reproduced with permission. [72] Copyright 2016, American Chemical Society. q) Schematic diagram of the synthesis of H 0.23 WO 3 crystal and monoclinic WO 3 nanosheets from tungsten boride precursor. r) XRD patterns of tungsten boride precursor and WO 3 crystals series. s,t) SEM images of H 0.23 WO 3 crystals, u) quasi-cubic-like WO 3 crystals, and v-x) monoclinic WO 3 nanosheets. y) Crystal models of (002), (020), and (200) facets of monoclinic WO 3 . Reproduced with permission. [73] Copyright 2012, Royal Society of Chemistry.  [110] direction. Reproduced with permission. [77,78] Copyright 2010, Wiley and 2006, American Chemical Society. c) The band structure and d) partial density of states of -Fe 2 O 3 . Reproduced with permission. [79] Copyright 2010, American Chemical Society. equivalent crystalline directions in (001) plane would be more slowly than that of [001] direction, making the crystal of -Fe 2 O 3 multifaceted. [77,78] As seen from the band structure and DOS of -Fe 2 O 3 (Figure 9c,d), the CB minimum is mainly occupied by the Fe 3d orbitals, whereas the O 2p orbitals constitute the VB maximum. According to the electronic structure of -Fe 2 O 3 , doping with IIIA elements (e.g., Al, Ga, In) in the lattice of -Fe 2 O 3 can promote the photocatalytic activity of -Fe 2 O 3 and have little effect on its bandgap and band edge energy at the same time, which can be regarded as a good strategy. [79] As the O 2 evolution performance of -Fe 2 O 3 can be greatly influenced by the morphology and size, a large number of investigations on the microstructure regulation of -Fe 2 O 3 were conducted by using different synthetic techniques under various conditions. The 0D -Fe 2 O 3 quantum dots were obtained by the facile microwave-assisted reverse micelle route (Figure 10a,b). It was shown that the narrow size distribution of as-synthesized quantum dots was 2-5 nm, which endowed -Fe 2 O 3 with higher photocatalytic activity (Figure 10c). [80] While 1D -Fe 2 O 3 nanorods (Figure 10d-f), nanotubes ( Figure 10g) and nanofibers (Figure 10h-j) can be obtained by high-temperature calcination, deposition, combustion, and electrospinning. [81][82][83][84][85] Compared to 0D or 1D -Fe 2 O 3 , the most widely used strategy for synthe-sizing 2D and 3D -Fe 2 O 3 is the hydrothermal or solvothermal route. 2D hexagonal -Fe 2 O 3 nanosheets with a thickness of 15 nm were prepared with a facile solvothermal approach, with the lattice spacing of 0.25 nm for the exposed (110) plane (Figure 11a-c). When they are combined with g-C 3 N 4 nanosheets to form a Z-scheme system, the OWS reaction with H 2 and O 2 evolution rate of 38.2 and 19.1 µmol h −1 g −1 , respectively, were achieved. [86] The truncated nano-octahedra structure of nanoparticles were obtained by hydrothermal method (Figure 11d,e), and the average size of these particles was about 800 nm. The uneven edges of these particles increased the exposure of crystal facets and specific surface areas, which were beneficial to the photocatalytic reactions. [87] The 3D -Fe 2 O 3 are generally constructed by 0D or 2D -Fe 2 O 3 . For instance, the solid/hollow -Fe 2 O 3 nanospheres were always assembled by the nanoparticles or nanosheets through hydrothermal process, leading to the unique hierarchical structure that allowed higher specific surface area and more catalytic active sites compared to nanoparticles or nanosheets (Figure 11f,g). [88,89] With potassium ferricyanide as the precursor, -Fe 2 O 3 dendrites were synthesized by hydrothermal treatment at 180°C for 12 h with trunk length of 6-7 µm and branches of 2-2.5 µm (Figure 11h). Under hydrothermal reaction at 180°C for 12 h, porous -Fe 2 O 3 nanocubes with average  [80] Copyright 2016, Royal Society of Chemistry. d) TEM image of -Fe 2 O 3 nanorods. Reproduced with permission. [81] Copyright 2010, American Chemical Society. e) SEM image and f) AFM image of -Fe 2 O 3 nanorods arranged on silicon substrates. Reproduced with permission. [82] Copyright 2009, Elsevier. g) SEM image and partial enlarged section (inset) of -Fe 2 O 3 nanotubes. Reproduced with permission. [83] Copyright 2010, Elsevier. h) Schematic diagram of electrostatic spinning. i) SEM image of -Fe 2 O 3 nanofibers. Reproduced with permission. [84] Copyright 2012, Royal Society of Chemistry. j) SEM image of -Fe 2 O 3 nanofibers. Reproduced with permission. [85] Copyright 2011, Springer Nature. edge length of 100 nm and various sizes of pores on their surface were fabricated. The hollow structure made -Fe 2 O 3 much easier to access reactants for accelerating photocatalytic reactions (Figure 11i,j). [90] The transition of dimensions of -Fe 2 O 3 can occur during the synthetic process of -Fe 2 O 3 . As illustrated in Figure 12a, the directional assembly from 1D iron oxide hydroxide chloride nanorods to 3D porous -Fe 2 O 3 nanocages was realized by introducing metal ions (Ni 2+ ) and surfactant (PVP) as the precursors. The -Fe 2 O 3 hollow nanocages were composed of nanorods with rough surfaces. The hematite phase was confirmed by the exposed crystal facets (Figure 12b-d). To study the transformation mechanism during the preparation process, the intermediate products synthesized at different times were observed. As shown in Figure 12e-g, the products gradually changed from nanorods to microspheres and then to hollow nanocages with increasing the reaction time, which may provide a new insight into the morphology regulation of -Fe 2 O 3 . [91] The water oxidation performance of -Fe 2 O 3 was also explored in terms of the morphology. The O 2 evolution rate of the -Fe 2 O 3 hollow nanospheres assembled from ultrathin nanosheets reached 70 µmol h −1 g −1 , which was much higher than that of -Fe 2 O 3 nanorods (32 µmol h −1 g −1 ) and commercial Fe 2 O 3 nanoparticles (14 µmol h −1 g −1 ) (Figure 12h,i), emphasizing the importance of the morphology regulation for the photocatalytic O 2 evolution performance of -Fe 2 O 3 . [92] similarly, the as-synthesized single crystalline nanospheres also demonstrated higher O 2 evolution rate than bulk crystals and irregular particles under either solar Figure 11. a) SEM image, and b,c) TEM and HRTEM images of -Fe 2 O 3 nanosheets. Reproduced with permission. [86] Copyright 2017, Wiley. d) XRD pattern and e) SEM image (structural representation inset) of -Fe 2 O 3 octahedral nanoparticles. Reproduced with permission. [87] Copyright 2012, Royal Society of Chemistry. f) SEM image of -Fe 2 O 3 spheres. Reproduced with permission. [88] Copyright 2012, Royal Society of Chemistry. g) SEM image of -Fe 2 O 3 hollow spheres. Reproduced with permission. [89] Copyright 2013, Royal Society of Chemistry. h) SEM image of -Fe 2 O 3 . i) SEM image and j) dark-field TEM image of porous -Fe 2 O 3 nanocubes. Reproduced with permission. [90] Copyright 2013, Springer Nature.
or visible light irradiation. Moreover, the O 2 evolution rates of these catalysts were found to be inversely proportional to the logarithm of the sizes (Figure 12j). [93]

Sillén-Aurivillius Perovskites
Sillén et al. found a series of compounds that were made up of the [M 2 O 2 ] 2+ metal oxide layers (M = Ca, Sr, Ba, Cd, Li, Na, Sb, Bi) separated by the halide layers and proposed the formula of X n (n = 1, 2, 3) to represent the number of halide layers that separated the metal oxide layers, such as BiPbO 2 X (X = Cl, Br, I) and BiOX (X = Cl, Br, I) are the X 1 and X 2 type of structure, respectively. Later, Aurivillius discovered in 1949 that some layered bismuth oxide compounds consist of the alternately stacked [Bi 2 O 2 ] 2+ units and perovskite layers, like Bi 2 MoO 6 , Bi 4 Ti 3 O 12 , Bi 3 TiNbO 9 , etc., the general formula can be described as (Bi 2 O 2 ) 2+ (A n−1 B n O 3n+1 ) 2− , where A and B represent the cations. In recent years, it has been discovered that some compounds possess both of the above two structural characteristics, and they are called Sillén-Aurivillius perovskites; the crystal structures can be expressed as [(Bi 2 O 2 ) 2 X] 2+ (A n−1 B n O 3n+1 ) 2− , where A and B represent the cations and X represents the halide anions. This type of material has recently been found to demonstrate excellent performance of photocatalytic O 2 evolution from water splitting. [94][95][96][97][98] In 2016, Ryu Abe's group first reported the photocatalytic activity for O 2 production over the Sillén-Aurivillius structured  Reproduced with permission. [92] Copyright 2013, Royal Society of Chemistry. j) Relationship between particle size and oxidation evolution rate for bulk crystals, irregular particles, and single crystalline nanospheres of -Fe 2 O 3 . Reproduced with permission. [93] Copyright 2011, Royal Society of Chemistry. mainly occupied by the highly dispersive O 2p orbitals and is not prone to self-oxidation due to the stable oxygen anions, so it has great potential for the catalytic water oxidation (Figure 13a,b). [99] In recent years, Abe's group has gained abundant experiences on the synthesis of Sillén-Aurivillius structured photocatalysts. Bi 4 NbO 8 Cl was traditionally synthesized by solid-state reaction (SSR), which led to the morphology of irregular chunks of the as-prepared samples (Figure 13c). Based on SSR experience, a certain proportion of alkali metal chlorides (CsCl and NaCl) were be added into the halogen precursors to make up for the halogen volatilization and to provide a solid solution environment (called as flux method). As a result, the Bi 4 NbO 8 Cl nanosheets with welldefined (00l) planes were obtained when calcination temperature reached above the melting points of CsCl and NaCl like 650°C (Figure 13d,e,h), 700°C (Figure 13f), or 800°C ( Figure 13g). [100] When FeCl 3 serves as the sacrificial agent, the photocatalytic O 2 evolution performance of Bi 4 NbO 8 Cl is the best (Figure 13l). [99] Importantly, the O 2 evolution rates of Bi 4 NbO 8 Cl synthesized by the flux method were much higher than those by SSR regardless of with or without the addition of cocatalysts (e.g., RuO 2 , Pt) (Figure 13m). [100] Inspiringly, the OWS reaction was realized when Bi 4 NbO 8 Cl was coupled with the H 2 evolution photocatalyst such as Rh-doped SrTiO 3 to construct the Z-scheme systems, and the system with Bi 4 NbO 8 Cl synthesized by flux method exhibited a higher photocatalytic activity than that prepared by SSR.
The different morphologies of Bi 4 NbO 8 Cl were obtained by the two-step polymerized complex method (2PC), solid-state reaction (2SSR), and conventional single-step solid-state reaction (1SSR) at the calcination temperature of 973 K (Figure 13i-k), and Bi 4 NbO 8 Cl synthesized by 2PC route had much smaller size and larger specific surface area than the samples synthesized by 2SSR and 1SSR, implying that the morphology of Bi 4 NbO 8 Cl was associated with the synthetic routes. The O 2 evolution production of Bi 4 NbO 8 Cl prepared by 2PC was also higher than those of samples synthesized by 2SSR and 1SSR routes ( Figure 13n). Naturally, the OWS activity of the 2PC synthesized Bi 4 NbO 8 Cl was also higher than that of 1SSR prepared sample when combined with Rh-doped SrTiO 3 with the existence of Fe 3+ /Fe 2+ redox mediator, highlighting the status of O 2 evolution reaction that as the decisive step for the overall water splitting. [101] Besides Bi 4 NbO 8 Cl, other Sillén-Aurivillius perovskites with similar crystal structures were also synthesized, including Bi 4 TaO 8 Cl and Bi 4 TaO 8 Br. The DOS plots in Figure 13s,t stated that the O 2p orbitals were powerfully hybridized with Bi 6s orbitals, which is a common characteristic of the electronic structures of most bismuth-based materials. Similar to Bi 4 NbO 8 Cl, the morphologies of Bi 4 TaO 8 X (X = Cl, Br) were transformed from the original bulk particles to the regular nanosheets with the thickness of ≈100 nm when the flux method was adopted with NaCl and KCl as the flux agents (Figure 13o-r). The O 2 evolution rates of Bi 4 TaO 8 X treated with fluxes were higher than that without fluxes. The O 2 production activity of all the Bi 4 TaO 8 X was promoted after loading the cocatalyst, and flux-treated Bi 4 TaO 8 X showed larger improvement (Figure 13u). [102] The interfacial modulation is considered to be an effective way to enhance the photocatalytic performance of photocatalysts. The O 2 evolution performances of Bi 4 TaO 8 X were substantially improved after being decorated with MoO 3 particles by the impregnation route ( Figure 13v). After the introduction of MoO 3 , the photogener-ated electrons would migrate from the VB of Bi 4 TaO 8 X to the CB of MoO 3 due to their distinct Fermi levels, thus forming a builtin electric field to effectively separate the charge carriers. Taking into account that the bandgap positions of both of Bi 4 TaO 8 Cl and Bi 4 TaO 8 Br meet the thermodynamic requirements of photocatalytic water splitting reaction (Figure 13w), when Bi 4 TaO 8 Br was coupled with Rh-doped SrTiO 3 to form the Z-scheme system, the yield ratio of H 2 to O 2 was close to 2:1 and it also showed a high stability, which provides an new reference for OWS (Figure 13x,y). [103] To make up for the volatilization effect of halogen species, excess halogen precursors were added in the SSR process of Bi 4 MO 8 X (M = Nb, Ta; X = Cl, Br) to obtain the samples denoted as ex-Bi 4 MO 8 X (M = Nb, Ta; X = Cl, Br). [100] As can be seen from Figure 14a-d, both Bi 4 MO 8 X and ex-Bi 4 MO 8 X were pure phase, and the addition of excess halogen precursors did not affect the crystal structure of Bi 4 MO 8 X. When calcination temperature increased to 1073 K, some small-sized particles appeared in the as-synthesized Bi 4 MO 8 X, which could be ascribed to the halogen species volatilization. Note that this phenomenon was effectively alleviated for ex-Bi 4 MO 8 X, indicating that excess halogen species compensated for the volatilization of halogen species (Figure 14e-l). With the increase of calcination temperature, the O 2 evolution rate of Bi 4 MO 8 X first increased and then decreased, reaching the maximum in 1073 K (Figure 14m), indicating that excessive anionic defects emerged at high temperature was not conducive to the O 2 evolution activity of Bi 4 MO 8 X. Obviously, the O 2 evolution rate of ex-Bi 4 MO 8 X was higher than Bi 4 MO 8 X at each calcination temperature, which was attributed to the less anionic defects in ex-Bi 4 MO 8 X than in Bi 4 MO 8 X. [104] Lately, a new series of double-layered Sillén-Aurivillius perovskites (denoted as A 4 A′M 2 O 11 Cl, A, A′ = Bi, Pb, Ba, and Sr; M = Ta, Nb, and Ti) were synthesized via the polymerized complex (PC) route by Abe's group (Figure 14n,o). The VB positions of the as-synthesized novel double-layered perovskites were generally more negative than those of some metallic oxides due to the interaction between O 2p orbitals and Bi/nearby anion/cation 6s and 6p orbitals (Figure 14p,q). Compared to the single-layered perovskites Bi 4

Other Photocatalysts for Oxygen Evolution
H 2 WO 4 belongs to the orthorhombic phase and its crystal structure consists of WO 5 (H 2 O) octahedral units, which has a wide absorption region for visible light. It was synthesized by the dehydration of H 4 WO 5 , generating products composed of aggregated plate-like particles with the sizes of 50-500 nm (Figure 15a). The O 2 evolution performance of H 2 WO 4 was the best when using Fe(NO 3 ) 3 as the sacrificial agent, whereas Fe 2 (SO 4 ) 3 was not  [100] Copyright 2019, American Chemical Society. SEM images of Bi 4 NbO 8 Cl synthesized through i) two-step polymerized complex method, j) solid-state reaction, and k) conventional single-step solid-state reaction. Reproduced with permission. [101] Copyright 2018, Royal Society of Chemistry. l) Time courses of oxidation evolution for Bi 4 NbO 8 Cl with different sacrifices. Reproduced with permission. [99] Copyright 2016, American Chemical Society. m) The O 2 evolution rates of pristine Bi 4 NbO 8 Cl, RuO 2 -loaded Bi 4 NbO 8 Cl, and Pt-loaded Bi 4 NbO 8 Cl synthesized through flux method and solid-state reaction. Reproduced with permission. [100] Copyright 2019, American Chemical Society. n) Time courses of oxidation evolution for Bi 4 NbO 8 Cl synthesized through two-step polymerized complex method, solidstate reaction, and conventional single-step solid-state reaction. Reproduced with permission. [101]  Reproduced with permission. [102] Copyright 2017, Wiley. v) The O 2 evolution production of Bi 4 TaO 8 X (X = Cl, Br) and MoO 3 -Bi 4 TaO 8 X with Fe(NO 3 ) 3 and Ag(NO 3 ) 3 as the sacrificial agent, respectively. Reproduced with permission. [103] Copyright 2018, Wiley. w) Band structures of Bi 4 TaO 8 X (X = Cl, Br). x) Time courses of overall water splitting evolution for Bi 4 TaO 8 Br combined with Ru(0.1 wt%)/Rh-doped SrTiO 3 with Fe 3+ /Fe 2+ as redox cycle mediator. y) Schematic diagram of overall water splitting evolution for Bi 4 TaO 8 Br-based Z-scheme system in Fe 3+ /Fe 2+ redox cycle mediator. Reproduced with permission. [102] Copyright 2017, Wiley.   [107] Copyright 2015, Royal Society of Chemistry. j-q) Digital pictures and r,s) DRS of M 3 ReO 8 . t) Total and partial density of states of Y 3 ReO 8 . u) Amount of oxidation evolution for M 3 ReO 8 in 10 h. Reproduced with permission. [108] Copyright 2017, Royal Society of Chemistry. v) Time courses of overall water splitting evolution for H 2 WO 4 combined with Ru(0.7 wt%)/Rh-doped SrTiO 3 in i) Fe(ClO 4 ) 3 aqueous solution and ii) distilled water. Reproduced with permission. [106] Copyright 2017, Royal Society of Chemistry. Time courses of overall water splitting evolution for w) PtO(0.5 wt%)/KCa 2 Nb 3 O 10 or x) PtO(0.5 wt%)/ex-Ca 2 Nb 3 O 10 /K + combined with Pt(0.5 wt%)/Rh-doped SrTiO 3 in KI aqueous solution. y) Time courses of overall water splitting evolution for ex-Ca 2 Nb 3 O 10 /K + /H + combined with Ru(0.7 wt%)/Rh-doped SrTiO 3 in Fe(ClO 4 ) 2 aqueous solution. Reproduced with permission. [107] Copyright 2015, Royal Society of Chemistry.
conducive to the O 2 evolution production, which was attributed to the fact that it was more easy for Fe 3+ to form the cation complex with SO 4 2− (Figure 15b). The O 2 evolution rate of H 2 WO 4 was directly proportional to its specific surface area, demonstrating that the larger specific surface area favors the reduction of Fe 3+ (Figure 15c). Similar to the above perovskites, H 2 WO 4 can also achieve OWS reaction when combined with Rh-doped SrTiO 3 , and the evolution activity of H 2 and O 2 was closely related to the solution environment, e.g., the gas yield in the Fe(ClO 4 ) 3 solution was more higher than that in distilled water (Figure 15v). [106] As one kind of cation-exchangeable layered metal oxides, KCa 2 Nb 3 O 10 can be easily intercalated by small molecules, which is beneficial for water oxidation. After exfoliating the H + /KCa 2 Nb 3 O 10 (K + of KCa 2 Nb 3 O 10 was replaced by H + ) into nanosheets (denoted as ex-Ca 2 Nb 3 O 10 /K + ) by continuous stirring, the specific surface area was greatly improved compared to that of KCa 2 Nb 3 O 10 (Figure 15d-g). With NaIO 3 as sacrificial agent, the O 2 evolution performance of the exfoliated samples was higher than that of KCa 2 Nb 3 O 10 with or without the cocatalyst RuO 2 (Figure 15h), owing to the more exposed active sites of ex-Ca 2 Nb 3 O 10 /K + . Some other cocatalysts, such as IrO 2 , can also promote the O 2 evolution of ex-Ca 2 Nb 3 O 10 /K + , and PtO was the most effective one (Figure 15i). The constructed Z-scheme system of ex-Ca 2 Nb 3 O 10 /K + -SrTiO 3 exhibited a more efficient OWS reaction with higher H 2 and O 2 yields than that of KCa 2 Nb 3 O 10 -SrTiO 3 in the KI aqueous solution, owing to the lower O 2 evolution activity and selectivity for KCa 2 Nb 3 O 10 (Figure 15w,x). Moreover, when RuO 2 and Fe(ClO 4 ) 2 were selected as the cocatalyst and reaction solution for this Z-scheme system, respectively, both H 2 and O 2 productions were greatly improved (Figure 15x,y). [107] The rare-earth rhenates M 3 ReO 8 (M = Y, La, Nd, Sm, Eu, Gd, Dy, and Yb) were obtained by the SSR route (Figure 15j-q). Compared to the corresponding oxides Re 2 O 7 , M 3 ReO 8 had more intense photoabsorption in the visible light region (Figure 15r,s). Taking Y 3 ReO 8 as the representative, the DOS revealed that the CB minimum of Y 3 ReO 8 was mainly occupied by Re 5d and O 2p orbitals, while the O 2p orbitals nearly took up the VB maximum of Y 3 ReO 8 (Figure 15t). Based on the previous findings, the photogenerated carries were easier to be recombined in the band, which was occupied by the R 4f orbitals of RVO 4 compounds (R = Ce, Pr, Nd, Sm, Eu, Tb, Dy, Ho, Er, Tm, and Yb), thus Gd 3 ReO 8 and Y 3 ReO 8 exhibited the higher O 2 production than other M 3 ReO 8 samples (Figure 15u). [108] In the homogenous system, the onset potentials of some ruthenium-based compounds (Figure 16a) were lower than those of Ru 3+/2+ of some normally used photosensitizers, such as polypyridyl ruthenium compounds [Ru(bpy) 3 ] 2+ (R 1 = COOEt, R 2 = H) (Figure 16b), so the water oxidation reactions can be driven by [Ru(bpy) 3 ] 2+ . In a three-components photocatalytic reaction system composed of the photocatalyst, photosensitizer, and sacrificial agent (Na 2 S 2 O 8 ) (Figure 16c (Figure 16d), [109] and the reaction equations are as follows 4Ru Since some ruthenium-based compounds have the good water oxidation performance under the assistance of the photosensitizer, some researchers then focused on the specific cobalt-based porphyrins in the same three-components reaction system (Figure 16e). The as-synthesized cobalt(II) tetrakis(psulfonatophenyl)porphyrin (COTPPS) showed the highest O 2 evolution production with the largest TOF of 0.17 s −1 , when the pH of the reaction solution was 11.0 and [Ru II (bpy) 3 ](NO 3 ) 2 and Na 2 S 2 O 8 served as the photosensitizer and sacrificial agent, respectively (Figure 16f-h). Additionally, the O 2 evolution rate exhibited a linear relationship with the concentration of COTPPS (Figure 16i). The corresponding reaction equations are as follows [110] Ru II ( bpy As a kind of organic-inorganic hybrid material, MOFs are generally constructed by the self-assembly of organic ligands and metal ions/clusters through coordination bonds, featuring the advantages of low density, high porosity, and large specific surface area. [111,112] In recent years, people have stepped up the researches of water oxidation of MOFs. Cu 3 PO 4 (C 2 N 3 H 2 ) 2 OH has a monoclinic crystal structure, in which the copper-polyhedral layers and copper-triazole layers were alternately stacked along [001] direction to form a 3D structure (Figure 17a,b). The more positive VB potential than O 2 /H 2 O endowed Cu 3 PO 4 (C 2 N 3 H 2 ) 2 OH with the potential of superior water oxidation ( Figure 17c). As seen from Figure 17d, the O 2 evolution performance of Cu 3 PO 4 (C 2 N 3 H 2 ) 2 OH reached ≈4.2 mL after 1 h photoirradiation with the addition of sulfite as sacrificial agent. [113] As introduced above, the bismuth-based photocatalysts have become a series of potential materials for water oxidation due to the abundant resources and environment-friendly feature. Bi 3+ ions easily coordinate with the organic ligands, so the Bi-based MOFs are ideal candidates worth to be explored. For instance, a orthorhombic bismuth-based metal-organic framework (denoted as Bi-mna) was obtained by solvothermal method with Bi(NO 3 ) 3 ·5H 2 O and 2-mercaptonicotinic acid as the precursors (Figure 17e). The O 2 evolution rate reached 216 µL h −1 with AgNO 3 as the sacrificial agent (Figure 17f). The calculated result of the electron localization function in Figure 17g indicated that the Bi atoms allowed the increase of delocalization of electrons in the covalent bond, which facilitates the migration of charge carriers. The Fukui functions showed that the photogenerated electrons were transferred from S atoms to pyridine rings with Bi atoms as the bridge, and this conclusion was also confirmed by the DOS of Bi-mna (Figure 17h). Furthermore, the measured time-resolved fluorescence decay spectra revealed that the average lifetime of Bi-mna (1.1 ns) was much longer than H 2 mna (100 ps), indicating that Bi-mna was able to more efficiently separate photogenerated charge carriers (Figure 17i). [114] The O 2 evolution production performance of non-strategic photocatalysts under distinct reaction conditions is shown in Table 1.  [110] Copyright 2013, Royal Society of Chemistry.

Strategies for Enhancing Photocatalytic Oxygen Evolution
The challenges of photocatalytic O 2 evolution with the participation of sacrificial agents are mainly from the inferior kinetics of the four-electron migration of water oxidation and intrinsic drawbacks of photocatalysts, such as the low utilization of visible light, easy recombination of photogenerated carriers, and selfoxidation poisoning from photogenerated holes. [126,127] To overcome the disadvantages originating from the above processes, the researchers have put forward various strategies, including cocatalysts loading, heterojunction fabrication, doping and defects Figure 17. a) XRD pattern of Cu 3 PO 4 (C 2 N 3 H 2 ) 2 OH. b) Crystal structure of Cu 3 PO 4 (C 2 N 3 H 2 ) 2 OH along [001] direction. c) Schematic diagram of O 2 evolution of Cu 3 PO 4 (C 2 N 3 H 2 ) 2 OH. d) Time courses of oxidation evolution for Cu 3 PO 4 (C 2 N 3 H 2 ) 2 OH with/without sacrifices. Reproduced with permission. [113] Copyright 2014, Elsevier. e) XRD pattern of Bi-mna. f) Time courses of oxidation evolution for Bi-mna. g) Electron localization function plots and h) total and partial density of states of Bi-mna. i) Time-resolved fluorescence decay spectra of Bi-mna and H 2 mna. Reproduced with permission. [114] Copyright 2014, Wiley.  formation, and others strategies, such as formation of special microstructure, surface modification, and solid solution construction.

Cocatalysts Loading
In the photocatalytic water splitting process, oxidation reaction requires more rigorous thermodynamic and kinetic conditions. As a small amount of cocatalysts can provide the active sites and trap the photogenerated charge carriers, they can reduce the activation energy and enhance the photocatalytic performance. [128] Therefore, the strategy of loading suitable cocatalyst on the surface of semiconductor is necessary. For the O 2 evolution reaction, the cocatalysts can be divided into the noble metals (e.g., Ag, Ru, Rh, Ir), noble metal oxides (e.g., RuO 2 , RhO 2 , IrO 2 ), and transition metal oxides (TMOs) (e.g., MnO 2 , Co 3 O 4 ). [129,130] Pt is the mostly used noble metal to decrease the overpotential and to promote the charge separation by collecting electrons. For example, when the loading amount of Pt was 0.5 wt% at 823 K by the impregnation method, the uniformly dispersed Pt species were observed clearly on the surface of WO 3 (Figure 18a), and the as-synthesized PtO x /WO 3 exhibited the highest O 2 evolution activity (Figure 18b,c). The secondary loading of metal oxides (CoO x , MnO x , RuO 2 , IrO 2 ) further boosted the O 2 evolution performance of PtO x /WO 3 , and the most effective one was RuO 2 with a O 2 evolution rate of 41 µmol h −1 (Figure 18d). [131] Au nanoparticles with diameters of ≈2 and ≈11 nm were in situ loaded on the surface of -Fe 2 O 3 particles by photodeposition (Figure 18e-g). The -Fe 2 O 3 loaded with larger size Au species showed higher O 2 evolution production, due to the stronger surface plasmon resonance (SPR) effect (Figure 18h). [132] As a typical nitrogen oxide, TaON has attracted widespread attention in water oxidation by virtue of the suitable band position and intense visible light absorption. [133,134] To explore the influence of cocatalyst on the O 2 evolution performance of TaON, Ruloaded TaON was synthesized by the impregnation method. It was found that the O 2 evolution production of TaON reached the maximum when the calcination temperature and the loading amount of Ru was 623 K and 0.5 wt%, respectively (Figure 18i,j).  [132] Copyright 2012, Royal Society of Chemistry. i) Time courses of oxidation evolution for TaON loaded with 0.5 wt% Ru at varieties of calcination temperatures. j) Time courses of oxidation evolution for TaON with various loading amounts of Ru. SEM images of TaON calcined at k) 623 K and l) 723 K, respectively. Reproduced with permission. [135] Copyright 2011, American Chemical Society. m) The O 2 evolution rates of TaON loaded with Rh, Co, and Ir as a function of calcination temperatures. TEM images of TaON n) loaded with 1 wt% Rh and o) pure sample. p) Time courses of oxidation evolution for TaON loaded/coloaded with Rh, Ru, and Co at varieties of calcination temperatures. Reproduced with permission. [136] Copyright 2019, Royal Society of Chemistry. SEM images revealed that the Ru precursor was mostly transformed into RuO 2 with grain size of ≈30 nm when the calcination temperature was 623 K (Figure 18k), while the agglomeration of RuO 2 appeared upon higher calcination temperature (Figure 18l), which led to the decreased O 2 evolution activity of TaON. When Pt/ZrO 2 /TaON and RuO 2 /TaON were separately used as the reduction and oxidation catalysts to form a Z-scheme system, the yield of H 2 and O 2 was close to 2:1, achieving the OWS in NaI aqueous solution. [135] On the basis of aforementioned research work, the O 2 evolution activity of TaON loaded with other metal cocatalysts (Rh, Co, Ir) by the impregnation route was also explored. As shown in Figure 18m-o, the most effective cocatalyst for the O 2 evolution of TaON was Rh species, and the rate reached ≈43 µmol h −1 when the calcination temperature was 500°C. In addition, the O 2 evolution rate of TaON loaded with two cocatalysts was higher than those of solely loaded samples, demonstrating the superiority of dual-loading of cocatalysts (Figure 18p). Besides, the Z-scheme system consisting of Rh/Ru/TaON and Pt/ZrO 2 /TaON showed higher photocatalytic efficiency than that of solely loaded TaON. [136] Considering the rare resources and high cost of noble metals, some common transition metals have been employed as cocatalysts to replace the noble metals in recent years. As shown in Figure 19a-c, Co 3 O 4 nanoparticles with diameters of 3, 10, and 40 nm were fabricated, in which the 3 nm Co 3 O 4 presented a good colloidal stability in solution (Figure 19d). Though high O 2 evolution activity was obtained, it decreased for all the Co 3 O 4 after a period of time due to the flocculation effect. Fortunately, this phenomena can be alleviated by using porous SBA-15 as the a carrier (Figure 19e). [137] The loading of TMOs is a good strategy to improve the photocatalytic activity of TiO 2 . [138,139] When TMOs were loaded on the surface of the TiO 2 nanosheets through solvothermal method (Figure 19f-h). It can be seen from the scanning transmission electron microscopy (STEM) images that there were irregular clusters (MnO x ) on the surface of the TiO 2 nanosheets compared to pristine TiO 2 nanosheets (Figure 19i,j). The very similar phenomenon was also observed for other cocatalysts. Interestingly, the O 2 evolution rate of TiO 2 nanosheets loaded with CoO x (47 µmol h −1 ) was higher than that of Ru/Irloaded counterparts, reflecting the superiority of TMO cocatalysts (Figure 19k,l). [140] Since the photogenerated electrons and holes have the feature of selective migration based on different crystal facets of some specific semiconductors (e.g., BiVO 4 ), [141] it is significant to explore the distributions of different types of cocatalysts on the crystal facets. Under photoirradiation, the Pt and CoO x were respectively accumulated on {010} and {110} facets of BiVO 4 (Figure 19m-o), which was distinct from the random deposition of cocatalysts resulted by the impregnation method. The selective photodeposition feature of BiVO 4 was also confirmed by other single deposition (MnOx, Au) (Figure 19p,q) and dual deposition (Pt and MnO x ) (Figure 19r). Notably, the O 2 evolution performance of BiVO 4 loaded with MnO x or Co 3 O 4 was improved by deposition of Pt, Au, and Ag (Figure 19s), attributing to the synergetic effect of the reductive and oxidative cocatalysts. [142] As one of the most abundant metals on earth, Fe species as cocatalysts have a great potential in the substitution of the traditional noble metal oxides. Due to the poor conductivity of Fe(oxy)hydroxides, 2D ultrathin FeOOH nanosheets with a thick-ness of ≈1.8 nm were prepared by a bottom-up strategy, which had abundant OVs compared to the bulk FeOOH and FeOOH nanoparticles (Figure 20a-c). The FeOOH nanosheets can be easily loaded on BiVO 4 by electrostatic interaction, leading to the O 2 evolution rate of 67.2 µmol h −1 with 0.5% loading amount of FeOOH nanosheets (Figure 20d). [143] In addition to the TMOs as cocatalysts, transition metal complexes also play an important role in promoting the water oxidation activity of photocatalysts. BiVO 4 loaded with three kinds of cobalt-based complexes (Co 4 O 4 (O 2 CMe) 4 L 4 , L = py, Mepy, and CNpy, denoted as 1-3, respectively; Figure 20e) were prepared by heating and refluxing routes, among which the BiVO 4 loaded with No. 3 cobalt-based molecular catalyst exhibited an excellent photocatalytic activity with a high TOF of 2 s −1 (Figure 20f). Kelvin probe force microscopy (KPFM) and surface photovoltage (SPV) measurements revealed that the ΔCPD signal (reflecting the migration extent of photogenerated holes to the surface of photocatalysts) and SPV amplitude were enhanced for BiVO 4 /Co 4 O 4 (O 2 CMe) 4 L 4 (Figure 20g,h), indicating that the separation efficiency of photogenerated carriers of BiVO 4 was improved by loading of cobalt-based complexes. [144] Due to the stable structure, the MOF materials can also be used as the supports of photocatalysts to promote the O 2 evolution. To improve the stability and photocatalytic activity of the polyoxometalates [Co(H 2 O) 2 (PW 9 O 34 ) 2 ] 10− (denoted as CoPOM), the MOF MIL-101 was applied as the support to encapsulate CoPOM inside the cavity. TEM and energy dispersive X-ray (EDX) mapping images (Figure 20i,j) showed that the CoPOM nanoparticles were evenly distributed in the channel of MIL-101 by ion exchange approach. After being immobilized inside the cavity of MIL-101, the O 2 evolution production of the composite was higher than that of pristine CoPOM, and the slight decrease in O 2 evolution activity after one circle reaction was ascribed to the leakage of CoPOM within MIL-101 (Figure 20k). [145] Compared with traditional metal-based cocatalysts, the nonmetallic oxide cocatalysts can be modified by the heteroatoms, such as nitrogen doping. Through calcination under ammonia atmosphere, WO 3 loaded with B 2 O 3−x N x nanoclusters was synthesized. TEM images (Figure 20m,n) showed obvious nanocluster layers of both WO 3 /B 2 O 3 and WO 3 /B 2 O 3−x N x . The O 2 production amount of WO 3 /B 2 O 3−x N x was higher than WO 3 /B 2 O 3 (Figure 20o), demonstrating positive effect of B 2 O 3−x N x in promoting the O 2 evolution of WO 3 than B 2 O 3 . It provided an insight for the exploration of nonmetallic cocatalysts. [146] Another widely used 2D cocatalyst is graphene, which can act as the carrier of photocatalysts to promote the separation of photogenerated carriers owing to the excellent conductivity. In Figure 20p-r, the WO 3 /graphene nanowires with different mass ratios of graphene that were prepared by a facile hydrothermal route displayed higher O 2 evolution performance than the WO 3 nanowires, because graphene promoted the separation efficiency of charge carriers by accepting the photogenerated electrons from WO 3 . [147] Due to the existence of the low resistance conduction path, the WO 3 /graphene with the particle diameter of ≈12 nm exhibited higher O 2 evolution production than bare WO 3 and physically mixed WO 3 and graphene (Figure 20s-u). [148] Similarly, the photocatalytic activity of -Fe 2 O 3 was also elevated by coupling with reduced graphene oxide (rGO) as the electron transfer platform (Figure 20v-x). [149]  Reproduced with permission. [142] Copyright 2014, Royal Society of Chemistry.

Heterojunction Construction
Two crucial factors that hinder the photocatalytic performance of semiconductor photocatalysts are the insufficient photoabsorption and the rapid recombination of photogenerated electron and holes during their migration from the bulk to the surface of photocatalysts. By combining two or more photocatalysts with the suitable band structures to form the II-type or Z scheme heterojunction, the weak photoabsorption of wide-bandgap semiconductors can be compensated by narrow-bandgap semiconductor. Moreover, the photogenerated electrons and holes can be effectively separated between different components on the basis of their band structures, collectively resulting in enhanced photocatalytic O 2 evolution activity.

II-type Heterojunction
II-type heterojunction is formed by two or more semiconductors with staggered band structures. The energy potential difference can propel the photogenerated electrons and holes to migrate to the semiconductors with more positive CB and more negative VB, respectively. II-type heterojunction is a p-n junction if it consists of a p-type semiconductor and an n-type semiconductor. The as-formed built-in electric field in the p-n junction will further boost the directional transfer of photogenerated carriers of the photocatalysts. [150][151][152] The schematic diagrams of II-type heterojunction and p-n heterojunction are shown in Figure 21.
Through a facile in situ crystallization, the CoAl-layered double hydroxide (CoAl-LDH) was grown on the surface of TiO 2 hollow nanospheres (TiO 2 @CoAl-LDH) to construct TiO 2 @CoAl-LDH hollow nanospheres with average size of ≈250 nm, which achieved the utilization of the full sunlight spectrum (Figure 22ac). Under the irradiation of simulated sunlight, the O 2 evolution rate of TiO 2 @CoAl-LDH (2.34 mmol h −1 g −1 ) was much higher than that of the pristine TiO 2 hollow nanospheres (0.27 mmol h −1 g −1 ) (Figure 22d). The DOS calculations revealed that the coupling of TiO 2 and CoAl-LDH made the CB potential of TiO 2 decreased and bandgap of CoAl-LDH narrowed at the same time, thus the photogenerated electrons can migrate from the CB of CoAl-LDH to that CB of TiO 2 , while holes from VB of TiO 2 were injected into that of CoAl-LDH, leading to the effective separation of charge carriers (Figure 22e). [153] The FeTiO 3 -TiO 2 porous hollow architectures with the interconnected nanosheets anchored on the surface were synthesized by a two-step solvothermal method followed by the calcination treatment (Figure 22g,h).
X-ray absorption near-edge spectra (XANES) of Ti K-edge revealed that the FeTiO 3 -TiO 2 and pristine component had the similar local symmetry (Figure 22i). When the atomic ratio of Ti to Fe was 0.75, the O 2 evolution rate of FeTiO 3 -TiO 2 hollow sphere reached the maximum, ≈2 times that of pristine FeTiO 3 (Figure 22j). The improved O 2 evolution performance was attributed to the formation of the II-type heterojunction between FeTiO 3 and TiO 2 , which promoted the separation of photogenerated charge carriers (Figure 22l). [154] Interestingly, inspired by the architecture of butterfly wings in nature, 3D WO 3 /BiVO 4 heterojunction was fabricated by a facile one-step sol-gel route with Paris papilio as the biological template ( Figure 22m). As shown in Figure 22n,o, both the as-prepared WO 3 and WO 3 /BiVO 4 photocatalysts retained the quasi-honeycomb morphology after the calcination treatment. The O 2 yield of WO 3 /BiVO 4 was 950 µmol after 5 h irradiation, which was 7.6-fold higher than that of pristine BiVO 4 (Figure 22p). The synergetic effects of the porous quasihoneycomb structure and the formed II-type heterojunction between WO 3 and BiVO 4 that boosted the visible light absorption and photogenerated charge separation, thereby contributing to the increased photocatalytic O 2 evolution production (Figure 22q). [155] Through a facile impregnation route followed by calcination treatment, the B-doped g-C 3 N 4 was deposited on the dual-phases BiVO 4 (monoclinic and tetragonal phases, denoted as BVOMT) to form the p-n heterojunction (Figure 23a). TEM and highresolution TEM (HRTEM) images demonstrated that the surface of g-C 3 N 4 nanosheets was decorated with BiVO 4 cashew nut (Figure 23b,c). The BVOMT showed higher O 2 evolution rate (452.8 µmol h −1 g −1 ) than the monoclinic-phase BiVO 4 (321 µmol h −1 g −1 ) and tetragonal-phase BiVO 4 (256 µmol h −1 g −1 ) (Figure 23d,e). Furthermore, when BVOMT was combined with Bdoped g-C 3 N 4 , the as-synthesized BVCN-50 with the strongest visible light absorption exhibited the highest O 2 evolution rate (1027.2 µmol h −1 g −1 ), which was attributed to the formation of p-n heterojunction (Figure 23f). [156]

Z-Scheme Heterojunction
Although the II-type/p-n heterojunction can effectively improve the separation efficiency of the photogenerated charge carriers, the photogenerated carriers are all transferred to lower energy levels, which reduces the redox driving force of photocatalysts. [157,158] As a result, the concept of Z-scheme has been proposed by recombination of the photogenerated holes of  3 . Reproduced with permission. [146] Copyright 2012, Royal Society of Chemistry. p,q) TEM images of WO 3 nanowires and WO 3 /graphene nanowires. r) Time courses of oxidation evolution for WO 3 nanowires and WO 3 /graphene nanowires. Reproduced with permission. [147] Copyright 2014, Hindawi Publishing Corporation. SEM images of s) graphene and t) WO 3 /graphene. u) Time courses of oxidation evolution for graphene, WO 3 , WO 3 /graphene, and physical mixed WO 3 /graphene. Reproduced with permission. [148] Copyright 2012, Royal Society of Chemistry. v) TEM image of -Fe 2 O 3 /rGO. w) Time courses of oxidation evolution for pristine -Fe 2 O 3 and -Fe 2 O 3 /rGO. x) Schematic diagram of reaction mechanism of -Fe 2 O 3 /rGO. Reproduced with permission. [149] Copyright 2013, American Chemical Society. PS I and the electrons of PS II (Figure 24). The advantage of the Zscheme heterojunction is that it can spontaneously promote the separation of photogenerated carriers and maintain the strong redox abilities of photocatalysts, which has attracted enormous attentions in photocatalytic water splitting in recent years. [159] The photodeposition experiment revealed that the {010} and {110} facets of BiVO 4 were the reduction and oxidation facets, respectively (Figure 25b), while the endpoint and surface of ZnO nanorods were the accumulation regions of electrons and holes, respectively. To probe the charge separation and migration behavior between ZnO and BiVO 4 , the 1D/3D ZnO/BiVO 4 was assembled by hydrothermal route followed by calcination treatment (Figure 25a). SEM images and electron paramagnetic resonance (EPR) spectra confirmed that the OV-rich ZnO nanorods were vertically anchored on the surface of BiVO 4 (Figure 25c-e). Under visible light illumination, the O 2 evolution rate of ZnO/BiVO 4 (68 µmol h −1 ) was much higher than that of the pristine BiVO 4 (Figure 25f). It was demonstrated that the photogenerated electrons and holes were separately gathered on {010} and {110} facets of BiVO 4 driven by the internal electric field. Partial holes on {110} facets were recombined with the electrons that enriched on the endpoints of ZnO nanorods to form a Z-scheme heterojunction, which led to efficient charge separation. [160] Besides, the mixed-phase BiVO 4 (monoclinic and tetragonal phases) were applied as the bridge to construct BiVO 4 /g-C 3 N 4 Z-scheme junction for accelerating the charge separation and water oxidation (Figure 25g). [161] To realize the tight coupling between -Fe 2 O 3 and g-C 3 N 4 , an SSR route was carried out to synthesize the 2D/2D -Fe 2 O 3 /C-g-C 3 N 4 (C-g-C 3 N 4 indicates g-C 3 N 4 with amorphous carbon). The weaker peak in XRD patterns of -Fe 2 O 3 /C-C 3 N 4 attributed to the exfoliation and partial carbonization of g-C 3 N 4 after calcination (Figure 25h), as confirmed by the obvious amorphous carbon around the edges of g-C 3 N 4 (Figure 25i). Owing to the formed Z-scheme junction between -Fe 2 O 3 and C-g-C 3 N 4 , the O 2 evolution production rate of -Fe 2 O 3 /C-g-C 3 N 4 (22.3 µmol h −1 ) was 30-fold higher than that of the pristine g-C 3 N 4 (0.7 µmol h −1 ) (Figure 25j,k). [162] As a layered perovskite, Bi 2 MoO 6 has a great potential in the field of photocatalysis due to the layered crystal structure and suitable bandgap. The heterojunction of Bi 2 MoO 6 hybridized with the good electron transporters, such as g-C 3 N 4 , has shown excellent photocatalytic performances in terms of the degradation of organic pollutants, water splitting for H 2 evolution, etc. [163][164][165][166] To further improve the activity or stability, it is significant to introduce appropriate electron transfer media, like Au, Rh, and Ru. Recently, the Bi 2 MoO 6 /Ru/g-C 3 N 4 catalyst was fabricated by the solvothermal method combined with the reduction of precursor (Figure 26a,d). DRS indicated that the incorporated Ru endowed the material with the absorption in the entire visible light region (Figure 26e). The calculation results of charge density difference and band structure manifested the formation of chemical bonds between Ru and O atoms of Bi 2 MoO 6 (Figure 26g-j). Under visible light irradiation, Bi 2 MoO 6 /Ru/g-C 3 N 4 exhibited the highest O 2 evolution rate of 328.34 µmol h −1 g −1 , which was ≈3 and 25 times that of Bi 2 MoO 6 and g-C 3 N 4 , respectively (Figure 26f). It was attributed to that the metallic Ru as the electron transfer media promoted the recombination of photogenerated electrons from Bi 2 MoO 6 and holes from g-C 3 N 4 , allowing Bi 2 MoO 6 to retain the strong oxidizing capability (Figure 26k). [167]

Doping and Vacancy Formation
Doping or vacancy creation can break the periodicity of crystal atomic arrangement and induce the lattice distortion, which lead to the formation of impurity states or defect states in the forbidden band, thus extending the light response range. [168] Besides, they can also promote the separation of photogenerated charge carriers, enhancing the photocatalytic O 2 evolution performance.

Doping
Since 1982, it has been found that the incorporation of a certain amount of transition metals, such as Cr and Ru, into TiO 2 allows it to absorb more visible light. After that, metallic doping was employed to improve the photocatalytic activity of semiconductors. [169] At the beginning of this century, nonmetallic elements doping (e.g., B, N) has gradually become the mainstream, as they bring more flexible tunability in the improvement of photocatalytic activity. [170,171] Nowadays, a generally accepted viewpoint is that the effect of metallic/nonmetallic doping for photocatalysts can be affected by the types of element, doping methods, concentration of dopant, and doping position.
Based on the mature research works of elemental doping in TiO 2 , the concept of gradient doping with nonmetallic heteroatoms in TiO 2 to improve the electronic structure has been Figure 22. a) Schematic diagram of the synthesis and O 2 evolution of TiO 2 @CoAl-LDH hollow nanospheres. b) SEM image and c) TEM image of TiO 2 @CoAl-LDH hollow nanospheres. d) Time courses of oxidation evolution along with three cycling tests for TiO 2 , CoAl-LDH, and TiO 2 @CoAl-LDH hollow nanospheres. e) Schematic diagram of reaction mechanism of TiO 2 @CoAl-LDH hollow nanospheres. Reproduced with permission. [153] Copyright 2015, Wiley. f) TEM image of FeTiO 3 . g,h) TEM and HRTEM image of FeTiO 3 -TiO 2 hollow spheres. i) The X-ray absorption near-edge spectra of Ti Kedge of TiO 2 , FeTiO 3 , and FeTiO 3 -TiO 2 . j) The O 2 evolution rates of pristine TiO 2 and FeTiO 3 -TiO 2 hollow spheres with mass ratios of 1:0.25, 1:0.5, 1:0.75, and 1:1. k) UV-vis spectrum and wavelength-dependent O 2 evolution rates of FeTiO 3 -TiO 2 hollow spheres. l) Schematic diagram of reaction mechanism of FeTiO 3 -TiO 2 hollow spheres. Reproduced with permission. [154] Copyright 2015, Royal Society of Chemistry. m) Schematic diagram of synthesis of WO 3 /BiVO 4 . SEM images of n) pristine WO 3 and o) WO 3 /BiVO 4 . p) Time courses of oxidation evolution for WO 3 , BiVO 4 , and WO 3 /BiVO 4 . q) Schematic diagram of reaction mechanism of WO 3 /BiVO 4 . Reproduced with permission. [155] Copyright 2017, Royal Society of Chemistry.  proposed. As shown in Figure 27a,b, the anatase TiO 2 microspheres doped with boron were synthesized by hydrothermal method and heat treatment with TiB 2 as the precursor of titanium and boron. The X-ray photoelectron spectroscopy (XPS) spectra with argon ion sputtering revealed that the binding energy of B 1s in the TiO 2 microspheres changed from 187.9 to 192.2 eV, indicating the transition of substitutional boron (B − ) to interstitial boron (B + ) during the heat treatment process (Figure 27c). It was found that the O 2 evolution of TiO 2 microspheres after thermal treatment was 4.5 times that of previous one (Figure 27d), which was attributed to the diffusion of boron from the core to the edges of microspheres, making the VB of TiO 2 shifted to a more positive energy level with stronger oxidation ability (Figure 27e). [172] Since the effect of doping can be affected by the types of element, it is interesting to investigate the influence of the valence state of doped elements in photocatalysts. As shown in Figure 27f, the O 2 evolution rates of TiO 2 doped with W 6+ , Ta 5+ , or Nb 5+ were higher than that of the pristine TiO 2 , while doping of Zr 4+ , Sn 4+ , or Ge 4+ had little effect on the photocatalytic activity of TiO 2 . In contrast, the O 2 evolution performance of TiO 2 was reduced by doping In 3+ , Ga 3+ , or Al 3+ . According to the classical Kröger-Vink theory, when the metallic cations with a higher valence than Ti 4+ were doped into the TiO 2 lattice (denoted as donor doping), the electron will increase in concentration and be captured into the Ti 4+ lattice, tuning the latter into Ti 3+ species. Then, the Fermi energy level was shifted upward and a built-in electric field was formed, promoting the separation of photogenerated charge carriers (Figure 27g). [173] The butterfly wings in nature are composed of uniformly arranged structures, in which the morphology of porous honeycomb endows them with the strong absorbance for the external sunlight (Figure 27h,i). Inspired by this architecture, the C-doped BiVO 4 was synthesized by the sol-gel route followed by subsequent thermal treatment with butterfly wings as the sacrificial template. As shown in Figure 27j,k, the original porous honeycomb structure was basically retained for the as-synthesized material and the lattice spacing of 0.308 nm was assigned to the {121} facets of BiVO 4 . DRS revealed that the incorporation of C  [160] Copyright 2018, Elsevier. g) Schematic diagram of water oxidation of mixed-phase BiVO 4 /g-C 3 N 4 , BVO-T, and BVO-M are represented the monoclinic scheelite and tetragonal zircon phase, respectively. Reproduced with permission. [161] Copyright 2019, Royal Society of Chemistry. h) XRD patterns of -Fe 2 O 3 /C 3 N 4 -r and -Fe 2 O 3 /C-g-C 3 N 4 . i) TEM image of -Fe 2 O 3 /C-C 3 N 4 . j) The O 2 evolution production of varieties of samples. k) Schematic diagram of reaction mechanism of -Fe 2 O 3 /C-C 3 N 4 . Reproduced with permission. [162] Copyright 2018, American Chemical Society.   (Figure 27l). When the temperature of thermal treatment and doping amount of C was 400°C and 1.5 wt%, respectively, the sample showed the highest O 2 evolution activity (800 µmol L −1 ; Figure 27m), which was ascribed to the synergistic effect of the unique structure and C doping, resulting in enhanced absorption of visible light and separation of the carriers. [174] In order to investigate the main factor that hinder the photocatalytic water oxidation, WO 3 treated with cesium (denoted as Cs-WO 3 ) was synthesized by the impregnation approach. After the introduction of cesium, an amorphous layer appeared on the surface of Cs-WO 3 with slightly changed size (Figure 27n,o). Compared to pristine WO 3 , the O 2 evolution rate of Cs-WO 3 increased by three times (51.3 µmol h −1 ) (Figure 27p). Contrary to previous view that the lower energy barrier was beneficial to the photocatalytic reaction, Cs-WO 3 with a higher energy barrier had a longer photogenerated carrier lifetime than that of WO 3 as revealed by the open circuit voltage decay curves (Figure 27q,r). Based on these advances, Li's group proposed that the main bottleneck for O 2 evolution is to enrich the long-lived photogenerated holes that were involved in the oxidation reaction on the surface of photocatalysts (Figure 27s), [175] as widely supported by other works. [176,177] The band position of TiO 2 can be adjusted through the nitridation, and its water oxidation performance can be improved by the donor doping, [173,178] thus the synergistic effect of metal and nonmetal codoping on the water oxidation activity of rutile TiO 2 was investigated. Kazuhiko's group reported that the rutile TiO 2 nanorods codoped with Ta/N species (denoted as TiO 2 :Ta/N) were synthesized by the microwave-assisted hydrothermal route followed by the subsequent calcination treatment in the ammonia atmosphere (Figure 28a-c). The doping sites of Ta 5+ within the crystal of TiO 2 were mainly along the [101] and [110] directions (Figure 28d,e). DRS and schematic diagram of band positions in Figure 28f,h revealed that the increased absorbance ability of visible light for TiO 2 :Ta/N was ascribed to the doping of N species, while the doping of single Ta species enlarged the bandgap of TiO 2 . The transient absorption spectra indicated that the as-synthesized TiO 2 :Ta/N showed a much lower concentration of deeply trapped charge carriers compared with N-doped TiO 2 , implying a higher photocatalytic activity of TiO 2 :Ta/N (Figure 28g). Based on above results, it can be proclaimed that the role of Ta doping is to improve the separation efficiency of photogenerated carriers of TiO 2 , while the incorporation of N species endows TiO 2 with stronger visible light absorbance. Additionally, the O 2 evolution rate rose with increasing the doping amount of Ta species and the temperature of microwave-assisted treatment, indicating that the photocatalytic activity of rutile TiO 2 was affected by both the doping level and the crystallinity of the catalysts (Figure 28i,j). Importantly, the photocatalytic activity of OWS was closely related to the types of redox cycle mediator and the kinds of O 2 evolution photocatalyst. It was revealed that the TiO 2 -SrTiO 3 Z-scheme system exhibited a higher catalytic activity with Fe 3+ /Fe 2+ as the redox cycle mediator than that with IO 3 − /I − because the Fe 2+ electron donor has a stronger capability in hindering the backward reaction of SrTiO 3 (Figure 28k,l). Furthermore, the TiO 2 :Ta/N showed the highest OWS performance (Figure 28m), and a solar-to-hydrogen conversion efficiency (STH) of 0.021% was obtained with Fe 3+ /Fe 2+ as the redox cycle mediator, and a higher STH of 0.039% was achieved with TiO 2 :Ta/N (IrO 2 as the cocatalyst) as the O 2 evolution component in the Rh-doped SrTiO 3 -based Z-scheme system. [179,180]

Vacancy Formation
In addition to doping, another strategy for improving the O 2 evolution performance based on defect modulation is the vacancy creation. Vacancy is one of the intrinsic defects, which can introduce the defect energy level into the forbidden band of semiconductors, thereby enhancing the light absorbance ability and charge separation. [181][182][183][184][185] OVs are one of the most common vacancies, and the construction of surface OVs is more beneficial to the separation of photogenerated carriers than the OVs in the bulk. [186] In recent years, 2D monocrystalline nanosheets are the ideal candidates to create OVs. For instance, the OVs were in situ generated on the surface of BiOCl single-crystalline nanosheets with the exposed {010}/{001} facets by the hydrothermal method followed by UV-light irradiation. The water molecules adsorbed over OVs on the {010} facets in the dissociated manner were more easily to be oxidized than that on {001} facets in the molecular manner, thus leading to an enhanced O 2 production activity. [187] Yan et al. fabricated WO 3 single-crystalline nanosheets with OVs introduced by the exfoliation and the following calcination treatment (Figure 29a). Compared to pristine WO 3 nanosheets, the morphology was hardly changed after the calcination procedure, while HRTEM images illustrated that there were amorphous layers with a thickness of ≈1 nm on the edge of WO 3 nanosheets via calcination in both vacuum and H 2 atmospheres (Figure 29bd). The slight shift of (020) diffraction peak to a higher 2 angle and the localized surface plasmon resonance (LSPR) peaks in the infrared region of the DRS spectra also proved the successful creation of OVs on the outside surface of WO 3 nanosheets (Figure 29e,f). The free carrier density of WO 3 nanosheets calcined in vacuum and H 2 atmospheres was calculated to be 2.5 × 10 21 and 2.0 × 10 21 by the following Drude formula, respectively where p , N e , e, 0 , and m e denote the bulk plasma frequency, charge carrier density, elementary charge, permittivity of free space, and effective mass of an electron, respectively. Under simulated solar light, WO 3 nanosheets calcinated in H 2 atmosphere demonstrated the highest O 2 evolution rate of 1593 µmol h −1 g −1 , ≈2.6-fold higher than that of the pristine WO 3 nanosheets (606 µmol h −1 g −1 ), which was attributed to the LSPR effect induced by the OVs that promoted the utilization of the solar energy ( Figure 29g). [188]

Other Strategies
Apart from the above strategies for improving the O 2 evolution performances of photocatalysts, others methods including the formation of special microstructure, surface modification, and solid solution construction, can also effectively promote the O 2 Figure 28. TEM images of a) pristine rutile TiO 2 , and b) TiO 2 doped with Ta and c) codoped with Ta/N. Reproduced with permission. [179] Copyright 2017, Royal Society of Chemistry. HAADF-STEM images, enlarged views, and crystal structures of TiO 2 codoped with Ta/N in d) [121] and e) [111] directions. Reproduced with permission. [180] Copyright 2019, Royal Society of Chemistry. f) DRS of pristine TiO 2 and varieties of samples. g) Transient absorption spectra of TiO 2 doped with N and codoped with Ta/N. h) Schematic diagram of band positions of TiO 2 doped with N, Ta, and codoped with Ta/N. Reproduced with permission. [179] Copyright 2017, Royal Society of Chemistry. The O 2 evolution rates of TiO 2 codoped with Ta/N as a function of i) doping amounts of Ta species and j) temperatures. Reproduced with permission. [180] Copyright 2019, Royal Society of Chemistry. Time courses of overall water splitting evolution for RuO 2 /Ta-N codoped TiO 2 combined with RuO 2 /Rh-doped SrTiO 3 in k) NaIO 3 aqueous solution and l) FeCl 3 aqueous solution. m) Time courses of overall water splitting evolution for varieties of O 2 evolution photocatalysts combined with RuO 2 /Rh-doped SrTiO 3 in FeCl 3 aqueous solution. Reproduced with permission. [179] Copyright 2017, Royal Society of Chemistry. Reproduced with permission. [188] Copyright 2015, Wiley.

Conclusions, Challenges, and Perspectives
Under the social background of advancing the sustainable development of energy resources, the photocatalytic water splitting has been gradually become the focus due to the features of abundant resources, environmental-friendly, etc. As the rate-determining step of the water splitting reaction, water oxidation is the key bottleneck that restricts the efficiency during this process. This review summarizes the latest research progresses of photocatalytic water oxidation. The content includes the introduction of several classical water oxidation photocatalysts (e.g., TiO 2 , BiVO 4 , WO 3 ), featuring the crystalline structures, synthesis approaches, and morphologies. On this basis of the critical issues that hinder the photocatalytic activity of photocatalysts, such as the low utilization of visible light and fast recombination of photogenerated charge carriers, the corresponding effective solutions, including the cocatalyst loading, heterojunction construction, doping and vacancy formation, and other strategies, are summarized.
In the last ten years, although a series of oxygen evolution photocatalysts have been developed, there is still a long way to go before the practical industrial applications. The photocatalysts for oxygen evolution still suffer from low efficiency or poor physicochemical stability, and especially most of them require the presence of sacrificial agents and cocatalysts, which also undoubtedly increase the economic costs to the industrial applications. In the future, more efforts are in need: i) The development of efficient water oxidation photocatalysts is still the present research focus, and the Sillén-Aurivillius perovskites will show significant potential for photocatalytic O 2 evolution. First, the VB of these perovskites such as Bi 4 NbO 8 Cl is mainly occupied by the O 2p orbitals. It makes them not be easily corroded by the photogenerated holes, thus demonstrating high photochemical stability. Second, the synthesis methods of this kind of materials are diverse, including 1SSR, 2SSR, 2PC, and flux, which allows the preparation process to be flexible and easy to optimize the photocatalytic performance. Additionally, the Sillén-Aurivillius perovskites have diverse compositions, which enables the adjustable light absorption or bandgap and photocatalytic O 2 evolution activity. Therefore, the rational design strategies based on crystal structure and band structure are expected to yield high-performance perovskites for O 2 evolution in the future.
ii) At present, the strategies for O 2 evolution performance enhancement are mainly achieved by improving the light absorption ability and charge separation efficiency, whereas the researches on the surface catalytic reaction are rarely involved. Actually, the reactive sites of photocatalysts are closely related to the adsorption of reactants and the reaction activation energy. For example, the construction of the surface defects can obviously enrich the active sites of photocatalysts. Therefore, the exploration of the catalytic active sites for O 2 evolution is expected to be one of the research priorities in the future. iii) Most of the reported works mainly focused on promoting the separation efficiency of photogenerated charge carriers to achieve the purpose of improving the O 2 evolution performance. However, less attention has been paid to the physicochemical stability of photocatalysts during the photocatalytic reaction process, whereas the actual situation is that the physicochemical stability of most of as-synthesized materials is often affected by many factors, such as the synthetic routes and conditions. In addition, some O 2 evolution photocatalysts are prone to be self-poisoned by the photogenerated holes, which results in the inactivation after photoreactions. Thus, developing effective tactics for improving the chemical and physicochemical stability of O 2 evolution photocatalysts are necessary.
Besides, the spatial separation for the occurrence of reduction and oxidation reactions should be considered, which can effectively inhibit the inverse reaction that usually occurs on the surface of photocatalyst. For instance, the hydrogen farm strategy proposed by Li's group is a promising direction, [222] and in which we believe more breakthroughs will be achieved.